Effect of lyophilisation, refrigerated storage and frozen storage on the coagulant activity and microbiological quality ofCynara cardunculus L. extracts

luis tejada

BACKGROUND:Cheese-makers have traditionally kept vegetable coagulants refrigerated until use, even though little was known of their microbiological quality or coagulant activity during storage. This study aimed to assess the efficacy of lyophilisation, refrigerated storage and frozen storage of fresh vegetable extract as a means of standardising coagulant activity in terms of coagulation times, pH and microbiological quality.RESULTS:Neither the pH nor the coagulation time of lyophilised extracts was significantly modified during 1 year; however, changes were observed following frozen storage, and more notable following refrigerated storage. Lyophilisation of aqueous extracts prompted the destruction of most micro-organisms; low counts initially noted for total mesophiles, lactic acid bacteria and yeasts disappeared during the first few days of storage, due to low water activity. There was a generalised decrease in micro-organism counts during frozen storage. Refrigeration was found to be unsuitable for storing of cardoon extract; an increase of roughly 2 log unit counts was recorded in total mesophile, lactic acid bacteria, yeast and mould counts after 1 year of refrigerated storage.CONCLUSION:Refrigerated storage cannot be considered a suitable method for prolonged conservation of aqueous cardoon extract. Both lyophilisation and frozen storage of aqueous extracts proved ideal for prolonged storage of vegetable coagulant. Lyophilisation additionally had certain advantages over frozen storage. Copyright © 2008 Society of Chemical Industry

Journal of the Science of Food & Agriculture Volume 88 Issue 8 , Pages 1301 - 1485 (June 2008) Research Articles Effect of lyophilisation, refrigerated storage and frozen storage on the coagulant activity and microbiological quality of Cynara cardunculus L. extracts (p 1301-1306) Luis Tejada, Montserrat Vioque, Rafael Gómez, José Fernández-Salguero Published Online: Apr 14 2008 4:50AM DOI: 10.1002/jsfa.3193 Polycyclic aromatic hydrocarbons in smoked cheese (p 1307-1317) Marie Suchanová, Jana Haj lová, Monika Tomaniová, Vladimír Kocourek, Lubo Published Online: Mar 28 2008 5:08AM DOI: 10.1002/jsfa.3198 Babi ka Effect of inulin and Lactobacillus paracasei on sensory and instrumental texture properties of functional chocolate mousse (p 1318-1324) Haíssa R Cardarelli, Lina C Aragon-Alegro, João H A Alegro, Inar A de Castro, Susana M I Saad Published Online: Mar 31 2008 9:07AM DOI: 10.1002/jsfa.3208 Fruit quality and volatile fraction of Pink Lady apple trees in response to rootstock vigor and partial rootzone drying (p 1325-1334) Riccardo Lo Bianco, Vittorio Farina, Giuseppe Avellone, Felice Filizzola, Pasquale Agozzino Published Online: Mar 11 2008 10:04AM DOI: 10.1002/jsfa.3210 A chemometric study of pesto sauce appearance and of its relation to pigment concentration (p 1335-1343) Francesca Masino, Giorgia Foca, Alessandro Ulrici, Laura Arru, Andrea Antonelli Published Online: Apr 2 2008 9:27AM DOI: 10.1002/jsfa.3221 Effect of electrical stimulation, delayed chilling and post-mortem aging on the quality of M. longissimus dorsi and M. biceps femoris of grass-fed steers (p 1344-1353) Regina H Razminowicz, Michael Kreuzer, Martin RL Scheeder Published Online: Mar 28 2008 5:04AM DOI: 10.1002/jsfa.3222 Predictability of price of tea from sensory assessments and biochemical information using data-mining techniques (p 1354-1362) Sanjoy K Paul Published Online: Apr 9 2008 6:32AM DOI: 10.1002/jsfa.3223 Comparison of volatile emissions from undamaged and mechanically damaged almonds (p 1363-1368) John J Beck, Bradley S Higbee, Glory B Merrill, James N Roitman Published Online: Apr 14 2008 4:45AM DOI: 10.1002/jsfa.3224 Cadmium accumulation in Agaricusblazei Murrill (p 1369-1375) Jian Cheng Huang, Kai Ben Li, Ying Rui Yu, Hanwen Wu, De Li Liu Published Online: Apr 7 2008 6:31AM DOI: 10.1002/jsfa.3225 Comparison of postmortem changes in goose cardiac and breast muscles at 5 °C (p 1376-1379) Sy-Shyan Ho, Chia-Ying Lin, Rong-Ghi R. Chou Published Online: Apr 7 2008 4:07AM DOI: 10.1002/jsfa.3227 Effects of pressure toasting on in situ degradability and intestinal protein and protein-free organic matter digestibility of rapeseed (p 1380-1384) Arash Azarfar, Claudio S Ferreira, Jacob O Goelema, Antonius FB Van der Poel Published Online: Apr 18 2008 9:03AM DOI: 10.1002/jsfa.3228 Rye bread reduces plasma cholesterol levels in hypercholesterolaemic pigs when compared to wheat at similar dietary fibre level (p 1385-1393) Helle Nygaard Lærke, Camilla Pedersen, Marianne Asp Mortensen, Peter Kappel Theil, Torben Larsen, Knud Erik Bach Knudsen Published Online: Apr 9 2008 6:33AM DOI: 10.1002/jsfa.3229 Color stability of frozen whole tilapia exposed to pre-mortem treatment with carbon monoxide (p 1394-1399) David Mantilla, Hordur G Kristinsson, Murat O Balaban, W Steven Otwell, Frank A Chapman, Sivakumar Raghavan Published Online: Apr 17 2008 7:10AM DOI: 10.1002/jsfa.3230 Antioxidant activity of the ethanolic extract from the bark of Chamaecyparis obtusa var. formosana (p 14001405) Palanisamy Marimuthu, Chi-Lin Wu, Hui-Ting Chang, Shang-Tzen Chang Published Online: Apr 17 2008 7:15AM DOI: 10.1002/jsfa.3231 Cloning of an alfalfa polyphenol oxidase gene and evaluation of its potential in preventing postharvest protein degradation (p 1406-1414) Michael L Sullivan, Ronald D Hatfield, Deborah A Samac Published Online: Apr 16 2008 3:45AM DOI: 10.1002/jsfa.3232 Application of surface response methodology to optimize hydrolysis of wheat gluten and characterization of selected hydrolysate fractions (p 1415-1422) Silvina R Drago, Rolando J González, María C Añón Published Online: Apr 18 2008 9:05AM DOI: 10.1002/jsfa.3233 High-performance liquid chromatography procedure for the determination of flavor enhancers in consumer chocolate products and artificial flavors (p 1423-1430) Charles H Risner, Melissa J Kiser Published Online: Apr 17 2008 7:14AM DOI: 10.1002/jsfa.3234 Fatty acid and fat-soluble antioxidant concentrations in milk from high- and low-input conventional and organic systems: seasonal variation (p 1431-1441) Gillian Butler, Jacob H Nielsen, Tina Slots, Chris Seal, Mick D Eyre, Roy Sanderson, Carlo Leifert Published Online: Apr 18 2008 9:04AM DOI: 10.1002/jsfa.3235 Protective effect of polyphenol-rich extract prepared from Malaysian cocoa (Theobroma cacao) on glucose levels and lipid profiles in streptozotocin-induced diabetic rats (p 1442-1447) Azli Mohd Mokhtar Ruzaidi, Maleyki Mhd Jalil Abbe, Ismail Amin, Abdul Ghani Nawalyah, Hamid Muhajir Published Online: Apr 17 2008 7:13AM DOI: 10.1002/jsfa.3236 Nutritional and sensory qualities of raw meat and cooked brine-injected turkey breast as affected by dietary enrichment with docosahexaenoic acid (DHA) and vitamin E (p 1448-1454) Carmen Sárraga, M Dolors Guàrdia, Isabel Díaz, Luis Guerrero, Jacint Arnau Published Online: Apr 17 2008 9:10AM DOI: 10.1002/jsfa.3238 Microbiological hazards involved in fresh-cut lettuce processing (p 1455-1463) Adriano G da Cruz, Sergio A Cenci, Maria Cristina A Maia Published Online: Apr 17 2008 6:36AM DOI: 10.1002/jsfa.3240 Effects of nitrogen application on malt modification and dimethyl sulfide precursor production in two Japanese barley cultivars (p 1464-1471) Masahito Nanamori, Ryoichi Kanatani, Makoto Kihara, Kazumitsu Kawahara, Katsuhiro Hayashi, Toshihiro Watanabe, Takuro Shinano, Mitsuru Osaki Published Online: Apr 17 2008 7:11AM DOI: 10.1002/jsfa.3241 Basis for the new challenges of growing broccoli for health in hydroponics (p 1472-1481) Diego A Moreno, Carmen López-Berenguer, M. Carmen Martínez-Ballesta, Micaela Carvajal, Cristina García-Viguera Published Online: Apr 23 2008 11:04AM DOI: 10.1002/jsfa.3244 Short Communications Anti-sickling potential of Aloe vera extract (p 1482-1485) Agunna Everest Ejele, Pascal Chukwuemeka Njoku Published Online: Apr 7 2008 4:05AM DOI: 10.1002/jsfa.3036 Journal of the Science of Food and Agriculture J Sci Food Agric 88:1301–1306 (2008) Effect of lyophilisation, refrigerated storage and frozen storage on the coagulant activity and microbiological quality of Cynara cardunculus L. extracts Luis Tejada,1∗ Montserrat Vioque,2 Rafael Gómez2 and José Fernández-Salguero2 1 Universidad Católica San Antonio, Departamento de Tecnologı́a de la Alimentación y Nutrición Campus de Los Jerónimos s/n 30107, Guadalupe, Murcia, Spain 2 Universidad de Córdoba, Departamento de Bromatologı́a y Tecnologı́a de los Alimentos, Campus de Rabanales, Edificio Darwin, 14014, Córdoba, Spain Abstract BACKGROUND: Cheese-makers have traditionally kept vegetable coagulants refrigerated until use, even though little was known of their microbiological quality or coagulant activity during storage. This study aimed to assess the efficacy of lyophilisation, refrigerated storage and frozen storage of fresh vegetable extract as a means of standardising coagulant activity in terms of coagulation times, pH and microbiological quality. RESULTS: Neither the pH nor the coagulation time of lyophilised extracts was significantly modified during 1 year; however, changes were observed following frozen storage, and more notable following refrigerated storage. Lyophilisation of aqueous extracts prompted the destruction of most micro-organisms; low counts initially noted for total mesophiles, lactic acid bacteria and yeasts disappeared during the first few days of storage, due to low water activity. There was a generalised decrease in micro-organism counts during frozen storage. Refrigeration was found to be unsuitable for storing of cardoon extract; an increase of roughly 2 log unit counts was recorded in total mesophile, lactic acid bacteria, yeast and mould counts after 1 year of refrigerated storage. CONCLUSION: Refrigerated storage cannot be considered a suitable method for prolonged conservation of aqueous cardoon extract. Both lyophilisation and frozen storage of aqueous extracts proved ideal for prolonged storage of vegetable coagulant. Lyophilisation additionally had certain advantages over frozen storage.  2008 Society of Chemical Industry Keywords: Cynara cardunculus; powdered vegetable coagulant (PVC); refrigerated vegetable coagulant (RVC); frozen vegetable coagulant (FVC) INTRODUCTION Written references dating back to Columella (De Re Rustica, c. 50 BC) testify to the use of cardoon extracts of the genus Cynara L. (Cynara cardunculus, Cynara humilis, Cynara scolymus) as a milk coagulant in cheesemaking in Mediterranean countries. The increasing consumption of cheese and the decreasing number of calves slaughtered in the 1970s led to an increase in the price of calf rennet, to a shortage in chymosin-rich rennet, and to a search for alternative milk coagulants. More recently, this shortage was exacerbated by the outbreak of BSE in dairy cattle, which was diagnosed in 1986 in the United Kingdom and later spread to the rest of Europe and elsewhere. Moreover, the use of animal rennet may be limited for religious reasons, diet (e.g. vegetarianism) or opposition to genetically engineered foods.1 Enzymes present in the flowers of Cynara cardunculus (cyprosins) are used in the production of some traditional Spanish and Portuguese cheeses, replacing calf rennet. These cheeses are currently enjoying considerable commercial success, and are increasingly prized by the consumer for their soft, creamy texture and their characteristic exquisite flavour, sometimes slightly bitter but piquant.2 Three proteinases of C. cardunculus L. have been isolated, purified, and partially characterised in terms of activity.3 They are thus acidic proteinases belonging to the aspartic proteinase group called ‘cynarases’ or ‘cyprosins’.4 Also, two additional aspartic proteinases were isolated from fresh stigmas of a standard variety of C. cardunculus grown from selected seeds, namely, cardosins A and B.5,6 ∗ Correspondence to: Luis Tejada, Universidad Católica San Antonio, Departamento de Tecnologı́a de la Alimentación y Nutrición Campus de Los Jerónimos s/n 30107, Guadalupe, Murcia, Spain E-mail: ltejada@pdi.ucam.edu (Received 5 March 2007; revised version received 6 September 2007; accepted 10 September 2007) Published online 14 April 2008; DOI: 10.1002/jsfa.3193  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 L Tejada et al. Crude vegetable coagulant obtained from directly harvested cardoons does not meet the minimum health and hygiene standards required of any food ingredient, because it displays high total viable germ counts, as well as elevated concentrations of enterobacteria.7,8 In addition to this lack of food safety, use of this natural vegetable coagulant cannot be standardised, since flowers gathered by different pickers may well include other cardoon species such as Centaurea calcitrapa, Silybum marianum and Cynara humilis, whose coagulant activity is either non-existent or very different from that displayed by Cynara cardunculus.9 Cardoon flowers are picked and dried in the open air; they are not specially treated, and may thus contain impurities and fragments of brownish bracts. Other authors10 found that drying the flowers affected the coagulant activity of the extract obtained. There are no standardised conditions for cutting and drying. The coagulant activity of cardoon extracts varies considerably as a function of variety, stage of maturity, part of the flower used, drying time and final moisture content.11,12 Experiments have shown that protease mRNA increases as the flower matures, and is not present in leaves, suggesting that gene expression is tissue specific.4 In an attempt to surmount these obstacles, a powdered concentrate has been obtained from coagulant enzymes from cardoon flowers.13 Lyophilisation of the aqueous extract provides a form of vegetable coagulant which is more manageable, stable and hygienic, and also more readily standardised. Use of a hygienic, standardised coagulant facilitates the production of more uniform batches of cheese, thus offering improved health guarantees and greater commercial and financial viability, without modifying biochemical parameters or losing the distinctive sensory features characteristic of this type of cheese.14,15 Ewe’s milk cheese has also been produced using recombinant cyprosin obtained by genetic engineering; no significant differences have been observed between this cheese and those made with fresh vegetable coagulant.16 Cheese-makers have traditionally kept vegetable coagulants under refrigeration until use, even though little was known of their microbiological quality or coagulant activity during storage. It has lately been suggested that frozen storage may be suitable for crude vegetable coagulants; this has the advantage of allowing storage of large volumes of extract under appropriate hygienic conditions, and ensuring both the conservation of organoleptic features and stable coagulant activity, thus enabling the production of more uniform batches of cheese. There is thus a need for research into the effect of refrigerated and frozen storage on the coagulant activity and microbiology quality of aqueous extracts obtained from the C. cardunculus cardoon flower. This study aimed to assess the efficacy of lyophilisation, refrigerated storage and frozen storage of fresh vegetable extract as a means of standardising 1302 coagulant activity in terms of coagulation times, pH and microbiological quality. MATERIALS AND METHODS Preparation of aqueous and lyophilised cardoon extracts Five aqueous extracts were prepared from each of five batches of Cynara cardunculus picked in various parts of southern Spain, following a procedure previously reported.17 We used 70 g of dry flowers, ground in a mortar and soaked in water (1 L) at room temperature for 24 h and filtered through a cheese cloth. Each of the five extracts was divided into three parts. One third of the extracts was lyophilised (prior freezing at between −25 ◦ C and −35 ◦ C; lyophilisation at 4–13 Pa for 24–36 h), using a Thermo Savant Freeze Dryer Modulyo D, as described previously.14 The lyophilised extract was then stored in sealed containers at ambient temperature and away from the light. Another second part of aqueous extracts was placed in Eppendorf tubes and kept in refrigerated storage at 4 ◦ C. The remaining third was placed in Eppendorf tubes and kept stored in a chest freezer at −30 ◦ C. Analytical determinations The pH of the extract was measured by direct reading, placing the combined pH electrode in the crude aqueous extract. Lyophilised extracts were first diluted in distilled water to 2.18% (percentage of powdered vegetable extract obtained after lyophilisation). Clotting activity was measured using the Berridge method;18 the substrate for clotting tests was skimmed milk powder reconstituted just before use (12%, w/v) in CaCl2 10 mmol L−1 solution, pH being adjusted to 6.40 with 10% lactic acid or 0.1 mol L−1 NaOH. Activity was expressed in coagulant units (CU). The CU unit is defined as the amount of extract required to clot 10 mL of milk in 100 s.19 Counts were performed in the lyophilised coagulant and in the refrigerated and frozen aqueous extracts for aerobic bacteria, enterobacteria, lactic acid bacteria (LAB), yeast and mould, Salmonella–Shigella, Listeria, Clostridium perfringens and enterococci, using procedures described previously.7 Coliforms were also analysed on violet red bile agar (VRBA) incubated at 37 ◦ C for 24 h.20 For microbiological testing of crude aqueous extract, 10 mL of the macerate obtained was homogenised in 90 mL sterile 0.1% peptone water. Frozen extracts were thawed in refrigerated conditions for 8 h. For lyophilised coagulant extracts, 0.2 g of extract (equivalent to 10 mL of aqueous extract) was dissolved in 100 mL peptone water. Testing was performed in duplicate at 0, 30, 60, 90, 140, 200, 270 and 360 days’ storage. Statistical analysis Statistical analysis of results was performed using the SPSS ver. 13 software package. A two-factor ANOVA J Sci Food Agric 88:1301–1306 (2008) DOI: 10.1002/jsfa Effects of storage on quality of C. cardunculus L. extracts was performed for storage method and storage time; Fisher’s least significant difference (LSD) test was used to compare differences between means; principal component analysis (PCA) were performed for storage time and type of coagulant. RESULTS AND DISCUSSION Effect of lyophilisation, refrigerated storage and frozen storage on pH, and clotting activities Mean values for pH and clotting activity of crude aqueous extract from C. cardunculus flowers kept in refrigerated and frozen storage, and for lyophilised extract, are shown in Figs 1 and 2, respectively. The changes of pH in refrigerated extracts were highly irregular: values declined sharply at 60 days, gradually rising thereafter to the end of the study period (Fig. 1). A strong positive correlation was noted between pH and most micro-organism counts, suggesting that increased counts might be attributable to an accumulation of alkaline substances arising from microbial metabolism. Coagulant activity rose progressively from the start of storage to 140 days. Other authors21 report similar behaviour for the clotting activity of C. cardunculus extracts kept in refrigerated storage for 30 days. However, a marked decline was recorded from 140 days to the end of Figure 1. Mean values for pH of crude aqueous extract from C. cardunculus flowers kept in refrigerated ( ) and frozen storage ( ), and for lyophilised extract (△). ° Figure 2. Mean values for clotting activity of crude aqueous extract from C. cardunculus flowers kept in refrigerated () and frozen storage ( ), and for lyophilised extract (△). ° J Sci Food Agric 88:1301–1306 (2008) DOI: 10.1002/jsfa the study (Fig. 2). This was probably due to elevated microbial activity, mainly involving lactic flora (peak counts were observed at 200 days, Table 1), to the loss of enzyme activity of cynarases, or to the increase of the pH. Milk pH is reported to have a great influence on the gelation properties of plant coagulants,22,23 and the clotting activity of chymosin appears to be more influenced than that of cyprosins by milk pH.24 Frozen storage of aqueous extracts prompted a slight increase in coagulant activity from 30 to 140 days. However, a significant decline was noted from 200 days onwards with respect to other storage periods, probably due to changes in cyprosin activity (Fig. 2). Other authors9 noted an increase of 40 s in clotting time for aqueous C. cardunculus extract kept in frozen storage for 415 days. There was also a significant increase in pH values (P < 0.05) after 140 days’ frozen storage in the present study. As shown in Figs 1 and 2, lyophilisation prompted no significant change (P < 0.05) in pH or clotting activity after 1 year’s storage at ambient temperature. This means that the lyophilisation does not affect the cyprosin activity, without modifying its coagulant activity, becoming an adequate method for the preservation of vegetable extracts used in the cheese production. Effect of lyophilisation, refrigerated storage and frozen storage on microbiological quality Table 1 shows average values and standard deviations of the counts (log cfu g−1 ) for the main microbial groups in crude aqueous extracts of C. cardunculus kept in refrigerated storage and frozen storage, and for lyophilised extracts at 0, 30, 60, 90, 140, 200, 270 and 360 days. The results of multiple comparison of individualised means by the LSD test are also shown. Neither Salmonella–Shigella nor Listeria were detected in any of the extracts studied (25 mL of sample). There was also no evidence of enterococci (10 mL of sample) or Clostridium in any of the samples. Similar findings were reported by Fernandez-Salguero et al.7 Time and storage method prompted a highly significant (P < 0.001) change in mean total viable, enterobacteria, coliform, lactic acid bacteria (LAB), yeast and mould counts. Lyophilisation of aqueous extracts destroyed micro-organisms due to the conditions reached during processing and to low water activity in lyophilised extracts (0.200–0.275; unpublished data). Lyophilised extracts therefore comply with the microbiological limits set down in the general Standard regarding identity and purity for coagulants and other milk-clotting enzymes intended for the domestic market.25 (P < 0.01). During refrigerated storage, mean total viable, enterobacteria, coliform, LAB, yeast and mould counts increased significantly (P < 0.001). There was thus a deterioration in the hygiene quality of extracts; total mesophile, LAB, yeast and mould counts increased by roughly 2 log units after 1 year of refrigerated storage. Enterobacteria and coliform 1303 L Tejada et al. 1304 Table 1. Effect of refrigerated storage (RVC), frozen storage (FVC) and lyophilisation (PVC) on different microbial groups (log10 cfu g−1 ; mean values and standard deviation) of C. cardunculus aqueous extracts Days of storage Variable Total viable Enterobacteria Coliforms Lactic acid bacteria Yeasts Moulds J Sci Food Agric 88:1301–1306 (2008) DOI: 10.1002/jsfa a–j Batch RVC FVC PVC RVC FVC PVC RVC FVC PVC RVC FVC PVC RVC FVC PVC RVC FVC PVC 0 ± 0.00g 5.80 5.80 ± 0.76g 0a 5.21 ± 0.52de 5.21 ± 0.52de 0a 5.13 ± 0.63def 5.13 ± 0.63def 0a 4.43 ± 0.51h 4.43 ± 0.51h 0a 5.57 ± 0.45g 5.57 ± 0.45g 0a 1.29 ± 0.14cd 1.29 ± 0.14cd 0a 30 ± 0.79h 6.59 4.19 ± 0.93f 0a 6.75 ± 0.43g 2.88 ± 0.93c 0a 6.52 ± 0.23g 2.26 ± 0.88b 0a 4.59 ± 0.49h 3.94 ± 0.57g 0a 6.66 ± 0.41h 4.74 ± 0.40f 0a 1.62 ± 0.38de 1.23 ± 0.23cd 0a 60 ± 0.90h 6.86 3.61 ± 0.77e 0a 6.48 ± 0.85fg 2.22 ± 0.86c 0a 5.90 ± 0.82efg 1.48 ± 0.76b 0a 4.85 ± 0.65h 3.20 ± 0.71f 0a 6.88 ± 0.34hi 3.97 ± 0.39e 0a 2.11 ± 0.31e 0.94 ± 0.14c 0a 90 140 7.41 ± 0.83i 7.60 ± 0.8i 3.22 ± 0.83de 2.79 ± 1.04d 0a 5.40 ± 1.35de 1.25 ± 0.73b 0a 4.86 ± 1.47d 0.63 ± 0.57a 0a 5.64 ± 0.75i 2.10 ± 0.92e 0a 7.03 ± 0.23i 2.86 ± 0.77d 0a 2.71 ± 0.51f 0.54 ± 0.48b 0a 0a 4.65 ± 1.77d 0.05 ± 0.16a 0a 3.34 ± 2.49c 0.21 ± 0.45a 0a 5.89 ± 0.81i 1.24 ± 1.11d 0a 7.21 ± 0.31j 1.42 ± 0.69c 0a 3.15 ± 0.46h 0.22 ± 0.37ab 0a Means of the same microbial groups for each type storage in the same row without a common superscript are different (P < 0.05). 200 270 ± 0.46i ± 0.39i 7.81 2.02 ± 1.10c 0a 5.23 ± 2.00de 0a 0a 5.01 ± 1.98de 0a 0a 7.46 ± 0.29j 0.71 ± 0.92c 0a 7.21 ± 0.11j 0.34 ± 0.45b 0a 3.03 ± 0.61h 0a 0a 7.63 0.38 ± 0.39b 0a 5.68 ± 1.93ef 0a 0a 5.78 ± 1.70efg 0a 0a 7.41 ± 0.47j 0.13 ± 0.28ab 0a 7.21 ± 0.41j 0a 0a 2.94 ± 0.88gh 0a 0a 360 7.52 ± 0.46i 0a 0a 5.67 ± 2.10ef 0a 0a 5.87 ± 1.63fg 0a 0a 7.30 ± 0.39j 0.00 ± 0.00a 0a 7.15 ± 0.42j 0a 0a 2.93 ± 1.02g 0a 0a Effects of storage on quality of C. cardunculus L. extracts counts displayed more irregular behaviour, with moderate increases at the end of the study period (log 5.7 cfu g−1 and log 5.9 cfu g−1 , respectively). The accumulation of substances deriving from microbial metabolism, the changes of the pH values and the marked proteolytic activity of proteases may have triggered the destruction of enterobacteria and coliforms between 60 and 140 days’ refrigerated storage. Total mesophile and enterobacteria counts at the end of storage were very similar to those reported by other authors;7 by contrast, LAB, yeast and mould counts were appreciably higher. Thus, unless a microbial inhibitor is added, aqueous extracts kept in refrigerated storage do not meet the health and hygiene requirements laid down in current Spanish legislation.25 In extracts kept in frozen storage (Table 1), a generalised decline was observed in micro-organism counts over the storage period; this was particularly marked for enterobacteria and coliforms, which were not detected from 200 days onwards. Total viable and LAB showed the greatest resistance to freezing: counts were still obtained at 270 days (log 0.4 cfu g−1 and log 0.1 cfu g−1 , respectively). In contrast to traditional assumptions, a number of authors have reported that freezing kills certain micro-organisms, particularly pathogens; freezing is thus recommended as a suitable way of improving food safety.26 – 28 Although freezing may reduce the numbers of some pathogens, it may not eliminate the all pathogens. In addition, all pathogens are not equally susceptible to freezing. Microbial destruction appears to be more intense below pH 6 (the pH of extracts ranged from 4.77 to 5.01; Fig. 1) and in Gram-negative bacteria. The death of micro-organisms due to freezing may be ascribed to a number of factors, including extracellular and intracellular ice formation, the concentration of extracellular and intracellular solutes, and low temperatures.28 This may account for the greater destruction of enterobacteria and coliforms than of LAB. Prolonged freezing of crude aqueous extract of Cynara cardunculus improves their microbiological quality; even after 200 days’ storage, the extract still meets legal requirements. There was no change in the general appearance of lyophilised or frozen extracts over the storage period. By contrast, after 60 days’ storage refrigerated extracts began to show unmistakable signs of deterioration, which became more acute over time including lumps, deposits, opacity and anomalous odours. Multivariate analysis PCA was performed to determine principal components. The selection criterion factor was based on an eigenvalue greater than 1, without any type of rotation of the factor matrix. Figure 3 shows the variables studied in the plain defined by the first two components, and the correlation between them. Two components were extracted that accounted for 96.2% of total variance. The first component, which accounted for J Sci Food Agric 88:1301–1306 (2008) DOI: 10.1002/jsfa 76.45% of total variance, was positively correlated with all variables. Microbial counts were the variables with the greatest differentiating effect. The second component accounted for 18.77% of total variance, was positively correlated with pH and clotting activity and correlated negatively with microbial counts. Figure 4 shows factor scores obtained by regression in two-dimensional space. The first component perfectly differentiated the three storage forms, and differentiated length of the storage period for frozen extracts over which microbial counts declined. The second component length of the storage period for refrigerated extracts, over which microbial counts rose. Storage time was never differentiated for lyophilised extract, indicating that the extract was not affected by storage time. Figure 3. Representation of variables studied in the plane defined by the first two principal components. Figure 4. Principal component analysis. Representation of extracts in terms of two components detected. 1305 L Tejada et al. CONCLUSIONS Refrigerated storage cannot be considered a suitable method for prolonged conservation of aqueous cardoon extract, since it prompts an increase in clotting time and microbial contamination, as well as causing some deterioration of organoleptic properties. Both lyophilisation and frozen storage of aqueous extracts proved ideal for prolonged storage of vegetable coagulant. Lyophilisation additionally had certain advantages over frozen storage. Firstly, after 4–5 months’ frozen storage, clotting activity declined, whereas the clotting activity of lyophilised extract remained constant over at least 1 year. Secondly, viable micro-organisms virtually disappeared following lyophilisation, thus meeting current regulations, whilst micro-organism counts were still detected in frozen extracts after 7–9 months’ storage. Thirdly, lyophilised extracts require less space for storage and transport, occupying roughly 2% of the space occupied by frozen aqueous extracts. Since lyophilisation does not affect proteinase conformation or modify clotting power, it is a suitable method for storing the vegetable extracts used in cheese-making. ACKNOWLEDGEMENTS The authors are grateful to the Spanish Ministry of Science and Technology for funding through Project AGL2002-02752. 9 10 11 12 13 14 15 16 17 18 19 REFERENCES 1 Roseiro LB, Barbosa M, Ames JM and Wilbey RA, Cheesemaking with vegetable coagulants – the use of Cynara L. for the production of ovine milk cheeses. Int J Dairy Technol 56:76–85 (2003). 2 Tejada L, Gómez R and Fernández-Salguero J, Sensory characteristics of ewe milk cheese made with three types of coagulant: calf rennet, powdered vegetable coagulant and crude aqueous extract from Cynara cardunculus. J Food Quality 30:91–103 (2007). 3 Heimgartner U, Pietrzak M, Geertsen R, Brodelius P, Da Silva Figueiredo AC and Pais MS, Purification and partial characterization of milk clotting proteases from flowers of Cynara cardunculus. Phytochemistry 29:1405–1410 (1990). 4 Cordeiro MC, Pais MS and Brodelius PE, Tissue-specific expression of multiple forms of cyprosin (aspartic proteinase) in flowers of Cynara cardunculus L. Physiol Plantarum 92:645–653 (1994). 5 Verissimo P, Esteves C, Faro C and Pires EV, The vegetable rennet on Cynara cardunculus L. contains two proteinases with chymosin and pepsin-like specificities. Biotechnol Lett 17:621–626 (1995). 6 Ramalho-Santos M, Verı́ssimo P, Faro C and Pires EV, Action on bovine αs -casein of cardosins A and B, aspartic proteinases from the flowers of the cardoon Cynara cardunculus L. Biochim Biophys Acta 297:83–89 (1996). 7 Fernández-Salguero J, Sánchez E, Gómez R, Mata C, Vioque M and Tejada L, A preliminary study of microbiological quality of cardoons of genus Cynara, L. used in manufacture of traditional cheese. Milchwissenschaft 54:688–690 (1999). 8 Gómez R, Sánchez E, Vioque M, Ferreira J, Tejada L and Fernández-Salguero J, Microbiological characteristics of ewes’ milk cheese manufactured using aqueous extracts of 1306 20 21 22 23 24 25 26 27 28 flowers from various species of cardoon Cynara L. Milchwissenschaft 56:16–19 (2001). Sanjúan E and Fernández-Salguero J, Influencia de algunos factores sobre el tiempo de coagulación por cuajo vegetal (Cynara sp.). Aliment Equ Tecnol 13:69–73 (1994). Martins APL, de Vasconcelos MMP and de Sousa RB, Thistle (Cynara cardunculus L) flower as a coagulant agent for cheesemaking – Short characterization. Lait 76:473–477 (1996). Macedo A, Malcata FX and Oliveira JC, The technology, chemistry and microbiology of Serra cheese. A Review. J Dairy Sci 76:1725–1739 (1993). Vioque M and Gómez R, Coagulantes vegetales empleados en la fabricación de quesos tradicionales. Caracterización de algunas variedades. Aliment Equ Tecnol 24:48–55 (2005). Fernández-Salguero J, Gómez R, Tejada L and Vioque M, A powdered vegetable coagulant, procedure for its preparation and their applications to cheese-making. Spanish Patent 2166719 (2003). Fernández-Salguero J, Tejada L and Gómez R, Use of powdered vegetable coagulant in the manufacture of ewe’s milk cheeses. J Sci Food Agr 82:464–468 (2002). Tejada L and Fernández-Salguero J, Chemical and microbiological characteristics of ewe milk cheese (Los Pedroches) made with a powdered vegetable coagulant or calf rennet. Ital J Food Sci 15:125–131 (2003). Fernández-Salguero J, Prados F, Calixto F, Vioque M, Sampaio P and Tejada L, Use of recombinant cyprosin in the manufacture of ewe’s milk cheese. J Agr Food Chem 51:7426–7430 (2003). Fernández-Salguero J and Sanjuán E, Influence of vegetable and animal rennet on proteolysis during ripening in ewe’s milk cheese. Food Chem 64:177–183 (1999). International Dairy Federation. Determination milk clotting time of milk to which milk clotting enzyme has been added. Document 110 A: Appendix A. IDF, Brussels (1987). International Dairy Federation. Bovine rennets: Determination of total milk-clotting activity. IDF (FIL-IDF Standard no. 157A). IDF, Brussels (1997). American Public Health Association, Standard Methods for the Examination of Dairy Products, 15th edition. APHA, Washington, DC (1984). Tavaria FK, Sousa MJ and Malcata FX. Storage and lyophilization effects of extracts of Cynara cardunculus on the degradation of ovine and caprine caseins. Food Chem 72:79–88 (2001). Esteves CL, Lucey JA, Wang T and Pires EM, Effect of pH on the gelation properties of skim milk gels made from plant coagulants and chymosin. J Dairy Sci 86:2558–2567 (2003). Zhao J, Chen SJ and Agboola S. Characterisation of milkclotting properties of Australian cardoon (Cynara cardunculus L.) extracts and comparison with commercial coagulants on different substrates. Milchwissenschaft 59:59–63 (2004). Silva SV, Allmere T, Malcata FX and Andren A, Comparative studies on the gelling properties of cardosins extracted from Cynara cardunculus and chymosin on cow’s skim milk. Int Dairy J 13:559–564 (2003). Ministerio de Sanidad y Consumo. Orden del 14 de enero de 1988. Norma general de identidad y pureza para el cuajo y otros enzimas coagulantes de leche destinados al mercado interior. BOE n◦ 17, de 20.01.1988 (1988). Tejada L, Sánchez E, Gómez R, Vioque M and FernándezSalguero J, Effect of freezing and frozen storage on chemical and microbiological characteristics in ewes’ milk cheese. J Food Sci 67:126–129 (2002). Vasudevan P, Marek P, Daigle S, Hoagland T and Venkitanarayanan KS, Effect of chilling and freezing on survival of Vibrio parahaemolyticus on fish fillets. J Food Safety 22:209–217 (2002). Archer DL, Freezing: an underutilized food safety technology? Int J Food Microbiol 90:127–138 (2004). J Sci Food Agric 88:1301–1306 (2008) DOI: 10.1002/jsfa Journal of the Science of Food and Agriculture J Sci Food Agric 88:1307–1317 (2008) Polycyclic aromatic hydrocarbons in smoked cheese Marie Suchanová,1 Jana Hajšlová,1∗ Monika Tomaniová,1 Vladimı́r Kocourek1 and Luboš Babička2 1 Institute of Chemical Technology, Prague, Faculty of Food and Biochemical Technology, Department of Food Chemistry and Analysis, Technická 3, 166 28 Prague 6, Czech Republic 2 Czech University of Life Sciences, Prague, Faculty of Agrobiology, Food and Natural Resources, Department of Quality of Agricultural Products, Kamýcká 129, 165 21 Prague 6 - Suchdol Abstract BACKGROUND: Polycyclic aromatic hydrocarbons (PAHs) represent a group of organic compounds containing two or more aromatic rings. Their control in the human food chain is required due to the mutagenic and carcinogenic potential, exhibited in vertebrates. In the present study, the occurrence of PAHs in 36 cheeses smoked by various processes was investigated. RESULTS: PAH concentrations (sum of 15 US EPA PAHs) found in samples smoked under controlled industrial conditions were at level 0.11 µg kg−1 , whereas in ‘home-made’ cheeses, the PAH content was up to 10 times higher. A similar trend was observed for B[a]P, a marker compound representing carcinogenic PAHs. While its levels in commercial products prepared by controlled smoking technologies were close to the limit of quantification (0.03 µg kg−1 ); in household samples, the B[a]P content ranged from 0.6 to 0.9 µg kg−1 . Significantly higher amounts of PAHs (up to three to six times) were found in surface layers as compared to internal parts of cheese. CONCLUSION: Although smoked cheese is a popular food, only several papers have focused on PAH levels in these products. This paper evaluates the contribution of different smoking technologies to PAH contamination of several cheeses and thus can help in a risk assessment associated with their consumption. Moreover, the study shows the concentration ratios of selected PAHs, from which the type of smoking technology can be indicated. The results obtained in this study also supported the suggestion of the EU Scientific Committee on Food to use benzo[a]pyrene as an indicator of the occurrence of higher-molecular mass PAHs.  2008 Society of Chemical Industry Keywords: Polycyclic aromatic hydrocarbons (PAHs); smoked cheese; HPLC/FLD INTRODUCTION Polycyclic aromatic hydrocarbons (PAHs) constitute a large group of organic compounds containing two or more fused aromatic rings. PAHs occur widely in the human environment, mainly as a result of incomplete combustion of organic matter; for example, it may take place during fires, in various industrial processes or in car engines. In toxicological studies, several PAHs have been demonstrated to be carcinogenic, and therefore they represent an issue of a great concern to health. Although air or drinking water may be responsible for some human exposure, the highest PAHs intake is typically associated with the occurrence of these hazardous chemicals in diet. Contamination of food crops by PAHs may be caused by environmental emissions; nevertheless, fairly high levels in some food commodities may be due to the processing practices such as drying and/or smoking. Also grilling, roasting and frying are high temperature processes potentially generating ‘food-borne’ PAHs. For instance, in barbecued meat the total PAHs were found to be present at levels up to 164 µg kg−1 with benzo[a]pyrene (B[a]P) being present at levels as high as 30 µg kg−1 , whereas in uncooked foods the average background values are usually in the range of 10−1 to 1 µg kg−1 .1 The health risk associated with dietary PAHs has recently been evaluated by Scientific Committee on Food (SCF).2 In European Commission (EC) recommendations3 the member states have been advised to monitor 15 SCF priority PAHs together with one additional PAH benzo[c]fluorene, recommended by the JECFA (Joint FAO/WHO Expert Committee on Food Additives).4 The maximum levels recently set by EC on B[a]P for smoked meat products is 5 µg kg−1 ; however, no limits are given for smoked cheese products at present.5 ∗ Correspondence to: Jana Hajšlová, Institute of Chemical Technology, Prague, Faculty of Food and Biochemical Technology, Department of Food Chemistry and Analysis, Technická 3, 166 28 Prague 6, Czech Republic E-mail: jana.hajslova@vscht.cz Contract/grant sponsor: Ministry of Education, Youth and Sports of the Czech Republic; contract/grant number: MSM 6046137305 (Received 13 June 2007; revised version received 26 November 2007; accepted 26 November 2007) Published online 28 March 2008; DOI: 10.1002/jsfa.3198  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 M Suchanová et al. Although smoked cheese is a popular delicacy in many countries, only a few papers published in recent two decades have focussed on PAH levels in these products.6 – 12 As a widely used contamination marker, carcinogenic B[a]P was determined in all the studies. In addition to B[a]P, another 15 US EPA PAHs were determined by Bosset et al.11 in a set of 67 smoked cheeses. Pagliuca et al.9 reported on the occurrence of six heavy US EPA PAHs (B[a]P, B[a]A, Chr, B[b]F, B[k]F). The broadest spectrum of PAHs in smoked cheeses manufactured from various types of milk (cows’, sheep’s, goats’) was monitored by Guillén and Sopelana.6,7 Besides the ‘classic’ analysis of PAHs in this smoked commodity, there is also the possibility of their ‘head-space’ determination being examined. Until now, none of studies were concerned with determination of the whole range of ‘SCF PAHs’ in smoked cheese. As regards analytical strategy, several approaches for PAH isolation from smoked cheeses have been described in literature: (1) alkaline treatment followed by liquid–liquid extraction,6,10,11 (2) extraction of freeze dried cheese by organic solvent supported by sonication12 or (3) pressurised liquid extraction (PLE).8 Clean-up of crude extracts can be performed using either SPE cartriges6,9 – 11 or gel permeation chromatography (GPC).8 For the final identification/quantification step, high-performance liquid chromatography with fluorescence detection (HPLC–FLD)8 – 12 or gas chromatography with mass spectrometric detection (GC–MS)6,7 are mostly used. An overview of methods employed for analysis of PAHs in food and environmental matrices and their relevance to control new EU regulations has been critically discussed in the recent paper by Wenzl et al.13 In the current study, a wide range of smoked cheeses available at a Czech market has been examined for 15 PAHs. The aim was to assess the increase (if any) of these hazardous chemicals due to the various (industrial and household) smoking processes. In order to document the distribution of PAHs within the cheese, surface parts were analysed separately. EXPERIMENTAL Cheese samples Altogether, 36 smoked cheese samples were analysed in this study. Twenty-four samples of smoked cheeses were obtained from three Czech cheese producers (in the following text identified as cheese companies A, B and C) and three samples from small, private manufacturers (home-made products, D) employing traditional household practices. Nine samples were obtained from a common market (E). Smoking conditions and other cheese characteristics are summarised in Table 1. An unsmoked cheese product (Edam) obtained from the common market was used as the reference sample. When supplied, samples were stored at −18 ◦ C. 1308 Chemicals and materials The standard mixture NIST 1647d of 16 priority PAHs: naphthalene (Naph), acenaphthene (Ace), fluorene (Fln), phenanthrene (Phe), anthracene (Ant), fluoranthene (Flt), pyrene (Pyr), benz[a]anthracene (B[a]A), chrysene (Chr), benzo[b]fluoranthene (B[b]F), benzo[k]fluoranthene (B[k]F), benzo[a] pyrene (B[a]P), dibenz[a,h]anthracene (DB[ah]A), benzo[g,h,i]perylene (B[ghi]P) and indeno[1,2,3cd]pyrene (I[1,2,3-cd]P) dissolved in acetonitrile was supplied by National Institute of Standards and Technology (NIST, Gaithesburg, MD, USA). Working standard solutions (concentrations in the range 0.01–60 ng mL−1 ) were prepared in acetonitrile and stored at 4 ◦ C. Before use, all glassware was washed with detergent, rinsed with distilled water and acetone and then dried at 220 ◦ C. Chloroform and acetone (analytical reagent grade, Lach-Ner, Neratovice, Czech Republic) were redistilled in glass before use. Acetonitrile (gradient grade, for chromatography), n-hexane (for organic trace analysis), dichloromethane (for gas chromatography; all Merck, Darmstadt, Germany), were used as supplied. Deionised water was obtained from Milli-Q water purification system (Millipore, Bedford, MA, USA). Anhydrous sodium sulfate (Penta Praha, Prague, Czech Republic) was dried at 500 ◦ C for 5 h and then stored in a tightly capped glass bottle. Equipment A laboratory blender (Waring blender, 38BL-40; Waring Commercial, New Hartford, CT, USA) and stainless-steel grater were used for homogenisation of cheese samples. A Soxhlet extractor with cellulose extraction thimbles (Filtrak, Niederschlag, Germany) was used for sample extraction. An automated gel permeation chromatography (GPC) system consisting of 305 MASTER pump, fraction collector, automatic regulator of loop 231 XLI, microcomputer (software 731 PC via RS232C), dilutor 401C (GILSON, Paris, France) and stainless steel column 500 × 8 mm i.d. packed with gel BioBeads S-X3, 200–400 mesh (Bio-Rad Laboratories, Philadelphia, USA) was used for clean-up of extracts. A vacuum evaporator (Büchi Rotavapor R-114 a Waterbath B-480, Postfach, Switzerland) was used for concentration of extracts. A high-performance liquid chromatographic system (HPLC), Hewlett-Packard 1100 Series, composed of a quarternary pump system with a degasser, an autosampler, a column thermostat, a fluorescence detector (FLD) (Hewlett Packard, Palo, Alto, CA, USA) and a Supelcosil LC-PAH (250 mm × 4.6 mm i.d.; 5 µm) column with the guard column Supelcosil LC-18 (20 mm × 4.0 mm i.d., 5 µm; Supelco, Bellefonte, USA) was used for analysis of sample extracts. Analytical procedure PAH analysis consisted of following steps. J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa Table 1. Overview and characteristics of the smoked cheese samples examined in this study Category of cheese producer/source A (Industrial) Cheese commercial name Sample code Cheese type Weight of single cheese package (g) Dry matter/fat content (%, w/w) 2000 1500 200 1500 1500 90 2000 50/42 56/45 54/40 42/45 42/45 45/45 57/45 56/45 Friction smoke technique Friction smoke technique Friction smoke technique Friction smoke technique Dry smoke flavouring powder Liquid smoke flavouring Smoke generation from burning wood – industrial Smoke generation from burning wood – household Smoke generation from burning wood – household – 280 43/38 N/A Liquid smoke flavouring 100 100 200 115 330 125 90 43/44 54/37 48/27 48/25 50/40 50/40 45/24 B (Industrial) C (Industrial) Mozzarella Edam Jadel Smoke processed cheese Smoke processed cheese Tizian Edam Pasta filata Semi-hard White, brined, steamed Processed, bar-shaped Processed, bar-shaped Processed Semi-hard D (Household products) D25–D26 Gazdovsky ostepek Semi-hard 600 50/40 D27 Klobucik Semi-hard 500 50/40 E28 Edam Semi-hard 260 E29 Akawi E30 E31 E32 E33 E34 E35 E36 Smoked processed cheese Jadel Koliba-mini Parenica Gazdovsky ostepek Koliba Tizian White cheese with smoked flavour Processed Steamed Semi-hard, steamed Semi-hard, steamed Semi-hard Semi-hard, steamed Processed – – – – – – – – – – – – – – 1309 PAHs is smoked cheese N/A, not applied; –, information not available. Type of smoking process 4 h at 32 ◦ C, beech wood 4 h at 32 ◦ C, beech wood 4 h at 32 ◦ C, beech wood 4 h at 32 ◦ C, beech wood N/A N/A 4–5 h at 36–38 ◦ C, smoke from beech wood 4 h, cherry tree and beech wood 4 h, cherry tree and beech wood – A1–A3 A4–A15 A16–A17 A18 A19 B20 C21–C24 E (Samples collected from retailed market) Smoking conditions M Suchanová et al. Soxhlet extraction Ten grams of homogenised cheese sample obtained by homogenisation either of the whole sample or the cheese rind (represented typically by 1–2 mm thick surface layer) was thoroughly mixed with 25 g of anhydrous sodium sulfate in a grinding mortar, then placed into the extraction cellulose thimble, covered with glass wool and inserted into the Soxhlet extractor. Prior to use, the thimbles were pre-extracted for 2 h with an extraction solvent to minimise PAHs procedure blank. Extraction was carried out with 170 mL of hexane–dichloromethane mixture (1:1, v/v) for 7 h (10 cycles h−1 ). The Soxhlet apparatus was covered with an aluminium foil to avoid access of daylight (to prevent the risk of photodegradation). The obtained extract was then carefully evaporated by rotary vacuum evaporator at 40 ◦ C, just to dryness, and the residue was quantitatively transferred into a 10-mL volumetric flask by chloroform. Liquid–liquid extraction of liquid smoke flavouring agent For extraction of liquid smoke flavouring agent, modified procedure published by Pagliuca et al.9 was used. Briefly, 10 g of sample were extracted with 100 mL of n-hexane in a separatory funnel; after 4 min of vigorous shaking the mixture was allowed to separate into two phases. The lower phase was re-extracted with 50 mL of n-hexane. This process was repeated twice. The three combined hexane extracts were concentrated by a rotary vacuum evaporator at 40 ◦ C, just to dryness, and the residue quantitatively transferred into a 10-mL volumetric flask by chloroform. Clean-up of crude extracts The clean-up step (separation of lipids) was carried out by GPC employing gel Bio-Beads S-X3. The flow rate of the mobile phase (chloroform) was set at 0.6 mL min−1 ; and the volume of sample injected onto the GPC column was 1 mL. After discarding the first 15.5 mL of eluate, the next 15.5 mL were collected. The purified extracts were subsequently subjected to concentration by rotary vacuum evaporator at 40 ◦ C just to dryness. The residue obtained after evaporation of solvent was dissolved in 0.5 mL of acetonitrile before the HPLC–FLD determinative step; this solution was then transferred into a 2 mL amber vial. Identification and quantification The HPLC–FLD was carried out under the following conditions: The high performance liquid chromatography with fluorescence detection (HPLC-FLD) was carried out under the following conditions: gradient elution (0 min–55% acetonitrile + 45% water, 20 min–100% acetonitrile, 32 min–100% acetonitrile), mobile phase flow rate 1 ml min-1, injection volume 20 µl, column temperature 35 ◦ C, FLD settings are shown in Table 2. The external standard calibration method based on peak heights was used for quantification of PAHs. 1310 Table 2. FLD settings for PAH detection Target PAHs Time window (min) Excitation wavelength (nm) Emission wavelength (nm) 7.2–10.7 10.7–11.1 11.1–12.2 12.2–13.2 13.2–14.3 14.3–16.0 16.0–19.3 19.3–23.6 23.6–25.8 25.8–30.5 216 240 248 248 232 236 270 250 295 248 336 320 368 404 448 384 388 430 405 484 Naph Ace, Fln Phe Ant Flt Pyr B[a]A, Chr B[b]F, B[k]F, B[a]P DB[ah]A, B[ghi]P I[1,2,3-cd]P Performance characteristics and quality assurance Since suitable matrices with certified concentrations of PAHs (CRM) are not available commercially; spiked samples at 0.5, 5 and 10 µg kg−1 were analysed within the validation study (50 µL, 50 µL and 100 µL of standard solution containing 100, 1000 and 1000 ng mL−1 of PAHs, respectively, were carefully incorporated into 10 g of cheese before extraction). Recovery was obtained as a slope of dependence of the measured values and the theoretical values (multiplied by 100). Repeatability of method was calculated as a relative standard deviation (RSD, %) from six parallel measurements of cheese with native PAHs content (0.1–45 µg kg−1 ). The overview of selected performance characteristics is shown in Table 3. RESULTS AND DISCUSSION In the first phase of our experiments, we had to decide on a choice of optimal analytical strategy enabling Table 3. Method performance characteristics obtained in validation study PAH Naph Ace Fln Phe Ant Flt Pyr B[a]A Chr B[b]F B[k]F B[a]P DB[ah]A B[ghi]P I[1,2,3-cd]P Recovery (%) Repeatability∗ (%) Limit of detection (µg kg−1 ) 52 68 57 94 86 87 73 71 72 70 75 71 78 70 94 34 16 20 9 11 11 11 13 14 14 14 14 15 12 16 0.05 0.05 0.05 0.25 0.09 0.23 0.12 0.05 0.04 0.08 0.01 0.01 0.02 0.04 0.07 ∗ Repeatability was calculated as a relative standard deviation (RSD, %), n = 6. Limit of quantification (LOQ) was not lower than three times of LOD level. J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa B[a]A PAHs is smoked cheese LU −240 −260 2000 18 20 B[ghi]P B[b]F Fln −255 DB[ah]A Naph −250 2500 22 26 24 1500 I[1,2,3-cd]P −245 3000 B[a]P B[k]F Chr 3500 Ant Phe LU 28 min Zoom I[1,2,3-cd]P 1000 B[ghi]P DB[ah]A B[a]P B[k]F B[b]F B[a]A Chr Pyr 0 Flt Ace 500 −500 0 5 10 15 20 25 min Figure 1. Example of an HPLC–FLD sample chromatogram obtained by analysis of sample D26 (PAH concentrations: 0.03–60 µg kg−1 ). to obtain accurate data even at low concentration levels of PAHs potentially occurring in smoked cheese. Considering our experience regarding limits of detection attainable either by GC–MS employing unit resolution mass analyser (quadrupole operated in selected ion monitoring mode) or HPLC–FLD, the latter technique was the preferred option, because of a better potential to detect even very low levels of carcinogenic PAHs. An example of cheese sample chromatogram obtained by this procedure is demonstrated in Fig. 1. Regarding the key analyte, B[a]P, all the quality assurance criteria required by EU Directive 2005/10/EC14 were met. Also, for most of the other PAHs (three-, four- and five-ring) the recoveries and repeatability of measurements were in a satisfactory range (70–94% and 9–16%, respectively). On the other hand, lower recoveries (52–68%) and poor RSDs (16–34%) were obtained for the most volatile PAHs such as Naph, Ace and Fln (see Table 3). It should be noted that these species are not a health concern in terms of carcinogenicity and therefore no modification of analytical procedure was carried out since it would have increased both time and labour demands. According to the data, results for Naph, Ace and Acy are semi-quantitative. PAH levels (in µg kg−1 ; values not corrected for recovery) determined in the set of examined cheeses are shown in Table 4. The sum of 12 PAHs (sum of all target PAHs except volatiles, Naph, Ace and Fln) and the sum of eight carcinogenic PAHs (B[a]A, Chr, B[b]F, B[k]F, B[a]P, DB[ah]A, B[ghi]P and I[1,2,3-cd]P) are presented in Fig. 2. The ‘background level’ obtained for the unsmoked reference sample, relevant to the latter PAH group, is shown in this figure as a dashed line. In commercial smoked cheese samples (categories A, B, C and E in Table 1), the concentrations of 12 PAHs and eight carcinogenic Sum of 12 PAHs (µg kg-1) 6 100 Sum of 12 PAHs Sum of carcinogenic PAHs 5 80 4 60 3 Background level (sum of 8 carcinogenic PAHs) 40 2 20 1 0 E28 E29 E30 E31 E32 E33 E34 E35 E36 D25 D26 D27 C21 C22 C23 C24 B20 A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 A13 A14 A15 A16 A17 A18 A19 0 Sum of carcinogenic PAHs (µg kg-1) 120 Cheese code Figure 2. PAH content in smoked cheeses (homogenate taken for analysis was prepared from the whole cheese sample, including cheese rind), for cheese codes see Table 1. Sum of 12 PAHs = sum of Phe, Ant, Flt, Pyr, B[a]A, Chr, B[b]F, B[k]F, B[a]P, DB[ah]A, B[ghi]P and I[1,2,3-cd]P. Sum of eight carcinogenic PAHs = sum of B[a]A, Chr, B[b]F, B[k]F, B[a]P, DB[ah]A, B[ghi]P and I[1,2,3-cd]P. J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa 1311 M Suchanová et al. Table 4. PAH content in smoked cheese and reference unsmoked sample (mean value, n = 3) (µg kg−1 ) Cheese code A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 A13 A14 A15 A16 A17 A18 A19 B20 C21 C22 C23 C24 D25 D26 D27 E28 E29 E30 E31 E32 E33 E34 E35 E36 Reference Naph 5.1 2.7 13 40 55 14 58 9.8 28 34 35 56 9.2 9.0 53 27 7.3 7.6 15 10 22 29 14 19 36 60 24 14 0.5 8.5 0.2 2.1 0.2 3.8 3.9 1.2 12.9 Ace 1.3 1.1 1.4 5.1 5.2 2.6 4.1 2.4 4.0 2.2 3.9 4.6 1.9 1.8 5.0 2.4 2.0 0.4 0.5 0.5 3.0 1.9 7.2 4.1 4.1 7.1 5.4 1.3 0.4 0.9 2.1 0.2 0.4 2.7 1.2 0.4 0.6 Fln 5.0 5.0 5.7 19 17 11 12 11 17 7 15 14 8.7 8.6 15 8.6 11 0.9 0.9 0.8 6.7 3.9 17 8.6 18 30 27 4.9 1.0 1.8 16 1.1 1.1 12 11 1.2 1.1 Phe 12 8.5 7.7 20 15 17 14 15 19 11 19 15 12 12 18 12 15 4.8 1.6 2.6 19 11 34 26 40 63 60 9.0 4.4 5.0 39 7.0 5.3 18 17 3.9 4.1 Ant Flt Pyr B[a]A Chr 1.4 1.5 0.8 1.2 1.3 0.8 1.3 1.1 0.6 3.8 2.2 1.6 4.0 1.7 1.0 2.6 2.2 1.3 2.7 1.7 0.9 2.6 2.2 1.4 3.8 2.2 1.3 1.6 1.2 0.6 3.6 2.5 1.4 2.8 2.1 1.2 2.0 1.7 0.9 1.9 1.7 0.9 3.4 2.1 1.2 1.8 1.6 0.8 3.1 1.4 0.9 0.1a 0.3a 0.4 0.1a ND 0.4 0.1a 0.3a 0.4 3.2 1.9 1.7 2.2 1.2 1.1 10.4 3.6 3.4 5.6 2.5 2.3 14 7.1 7.1 23 12 11 23 12 12 1.2 0.7 0.5 0.3 0.3a 0.3 0.3 0.3a 0.2a 8.1 5.7 3.7 0.1a 1.1 0.7 0.1a 0.9 0.5 4.7 4.3 2.8 5.6 3.6 3.1 0.3 0.1a 0.2a 0.1a 0.1a 0.4 0.08a 0.06a 0.08a 0.08a 0.08a 0.08a 0.2 0.08a 0.08a 0.2 0.08a 0.2 0.08a 0.08a 0.08a 0.08a 0.08a 0.08a ND ND 0.08a 0.08a 0.08a 0.2 0.2 1.3 1.9 2.2 0.08a ND ND 0.3 0.08a ND 0.6 0.9 ND ND 0.1 0.06a 0.2 0.1 0.3 0.2 0.2 0.2 0.1 0.3 0.3 0.1 0.1 0.2 0.2 0.2 ND ND ND 0.06a 0.06a 0.1 0.1 0.8 1.2 1.3 0.06a 0.06a 0.06a 0.3 0.06a 0.06a 0.5 0.7 0.1 ND B[b]F ND ND ND ND n.d n.d ND ND 0.04a ND ND ND ND ND ND ND ND 0.04a 0.04a 0.1a ND ND ND 0.04a 0.3 0.4 0.5 0.04a 0.04a ND 0.04a 0.04a ND 0.1a 0.2 0.04a 0.04a B[k]F B[a]P DB[ah]A B[ghi]P I[1,2,3-cd]P ND ND ND 0.01a 0.01a 0.02a ND 0.01a 0.02a ND 0.02a ND ND ND ND ND ND 0.01a 0.01a 0.06 ND ND 0.02 0.04 0.2 0.3 0.3 0.01a 0.01a 0.01a 0.06 0.3 0.01a 0.07 0.1 0.02a 0.01a 0.02a ND ND ND ND 0.01a ND ND ND ND ND ND ND ND ND ND ND ND ND ND 0.01a ND ND ND 0.01a 0.01a 0.03a 0.01a ND ND ND 0.01a ND ND 0.01a 0.03a 0.01a ND ND ND ND 0.02a ND 0.02a ND 0.02a 0.02a ND 0.06a 0.06a 0.06a 0.06a 0.02a 0.02a ND 0.06a 0.06a 0.1 ND ND ND ND 0.6 0.8 0.7 0.1 0.06a 0.06a 0.1 0.06a 0.06a 0.1 0.3 0.1 ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND 0.1a ND ND ND ND 0.3 0.5 0.5 ND ND ND 0.1a ND ND ND 0.3 ND ND 0.01a 0.01a 0.03 0.03 0.05 0.03 0.05 0.05 0.01a 0.04 0.04 0.01a 0.01a 0.03 0.02a 0.02a 0.01a 0.01a 0.07 0.01a 0.01a 0.03 0.02a 0.6 0.6 0.9 0.04 0.03 0.02a 0.1 0.03 0.02a 0.2 0.5 0.05 0.01a a Estimate; analyte concentration below limit of quantification. ND, not detected (for LODs see Table 3). PAHs ranged from 2.3 to 57 µg kg−1 and from 0.1 to 2.7 µg kg−1 , respectively. Significantly higher PAHs levels were found in ‘home-made’ samples (category D), where the content of total PAHs was in the range from 73 to 114 µg kg−1 and the sum of carcinogenic PAHs ranged from 3.9 to 6.2 µg kg−1 . Higher contamination of these ‘traditional’ cheeses was obviously due to deposition of PAHs-containing solid particles on their surface. Purification of the smoke generated within industrial process under carefully controlled conditions allows only flavouring compound to reach the cheese, while most of hazardous products of wood pyrolysis are removed. In addition to examination of the whole cheese, the rinds of selected samples were analysed separately (Table 5). The PAH levels (µg kg−1 ; values not corrected for recovery) detected in these surface layers (thickness, 1–2 mm) were three to six times higher 1312 (Fig. 3) regardless the type of smoking technique used. Fairly lower overall contamination, similar to ‘background levels’ detected in unsmoked cheese was found in edible parts of samples that were aromatised either by dry smoke flavouring powder (A18) or under the controlled pyrolysis conditions (A19). Both these bar-shaped cheeses were coated by inedible waxy coating, which accumulated most of PAHs (i.e. served as contamination barrier). Compared to edible rind (cheese layer under the waxy coating), the PAHs content in the coating was 14 times higher; the sum of carcinogenic PAHs even 18 times higher. In summary, according to our estimates, in most cases, removing the 1–2 mm surface layer from the smoked cheese eliminates approximately 50–100% of PAHs detectable in particular product. These observations are in agreement with results of study published by Guillén and Sopelana.6 J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa Table 5. PAH content in smoked cheese rind samples (µg kg−1 ) Cheese code Naph Ace Fln Phe Ant Flt Pyr B[a]A Chr B[b]F B[k]F B[a]P DB[ah]A B[ghi]P I[1,2,3-cd]P Rind portion taken for analysis (% of cheese total weight) A4–A15 A19 B20 C21–C22 C23–24 D25 D26 D27 E28 E29 E30 E31 E32 E33 E34 E35 E36 184 9.2 16 346 326 98 167 249 60 6.8 10 137 1.2 3.3 39 84 ND 24 0.8 1.0 46 61 23 28 43 12 1.3 0.4 23 0.8 1.2 15 9.2 0.2 101 1.7 2.7 267 164 108 137 191 62 3.4 2.2 120 3.3 3.9 83 66 1.6 93 6.5 8.1 148 204 244 371 339 65 20 16 270 17 20 99 85 16 28 0.3 0.4 42 58 88 132 119 17 1.0 0.7 61 0.5 0.6 29 25 0.6 8.8 1.2 0.7 11 24 47 77 51 6.5 3.4 3.1 37 2.8 3.0 19 18 3.0 7.2 1.1 0.9 8.0 23 44 75 48 5.0 3.2 2.4 25 2.3 2.4 14 14 2.4 0.9 0.2 0.08a 0.8 1.3 8.6 14 9.4 0.5 0.2 0.3 3.0 0.2 0.2 3.6 4.4 0.2 0.4 0.06a 0.1 0.5 0.8 5.0 7.9 5.5 0.3 0.2 0.4 1.2 0.2 0.2 2.3 2.8 0.2 0.2 0.1a 0.04a 0.1a 0.3 1.7 2.9 2.1 0.2 0.1a 0.2 0.6 0.1a 0.1a 0.7 1.0 0.2 0.1 0.01a 0.01a 0.08 0.1 1.04 1.6 1.3 0.1 0.08 0.1 0.3 0.06 0.07 0.4 0.7 0.09 0.3 0.07 0.01a 0.2 0.5 3.4 5.4 4.4 0.3 0.4 0.1 0.7 0.08 0.08 1.2 2.1 0.1 0.06 0.06 0.01a ND ND 0.09 0.1 0.1 0.03a ND ND ND ND ND ND ND 0.03a 0.1 0.1 ND 0.3 0.4 3.0 4.2 3.7 0.3 0.2 0.2 0.5 0.2 0.2 0.5 0.9 0.1 0.2 0.1a ND ND 0.4 1.9 2.8 2.5 ND ND ND ND ND ND ND 0.9 ND 7–13 13 7 8–9 8–9 10 10 10 6 13 20 16 13 10 10 20 11 a Estimate; analyte concentration below limit of quantification. ND, not detected. A4–A15; C21–C22; C23–24: pooled samples. PAHs is smoked cheese 1313 45 40 700 Sum of 12 PAHs Sum of carcinogenic PAHs 35 600 30 500 25 400 20 300 15 Rind E36 Rind E35 Rind E34 Rind E33 Rind E32 Rind E31 Rind E30 Rind E29 Rind E28 Rind D27 Rind D26 Rind D25 0 Rind C23-24 0 Rind C21-C22 5 Rind B20 10 100 Rind A19 (edible part) 200 Rind A4-A15 Sum of 12 PAHs (µg kg-1) 800 Sum of carcinogenic PAHs (µg kg-1) M Suchanová et al. Cheese code Figure 3. PAH content in smoked cheese rinds (pooled samples A4–A15, C21–C22 and C23–24), for cheese codes see Table 1. Sum of 12 PAHs = sum of Phe, Ant, Flt, Pyr, B[a]A, Chr, B[b]F, B[k]F, B[a]P, DB[ah]A, B[ghi]P and I[1,2,3-cd]P. Sum of eight carcinogenic PAHs = sum of B[a]A, Chr, B[b]F, B[k]F, B[a]P, DB[ah]A, B[ghi]P and I[1,2,3-cd]P. Table 6. Comparison of B[a]P levels found in different studies B[a]P content (µg kg−1 ) 0.01–0.88 ND–0.55 <0.1–3.8 <0.2–1.7 0.03–0.39 0.1–0.75 <0.1–4.2 ND–0.91 Subject of study/ type of sample Commercially and home-made smoked cheeses Rind of commercial smoked cheeses Home-made smoked cheeses Effect of different smoking conditions on B[a]P Effect of different smoking conditions incl. smoke-flavoured and liquid flavoured cheeses on PAHs Investigation of different smoking conditions on B[a]P Investigation of different smoking conditions on PAHs Commercially smoked cheeses Reference This study 6 8 8 9 10 11 12 Considering for comparison the maximum level 5 µg kg−1 given in EC Regulation No 1881/20065 for B[a]P in smoked meat, we classify the contamination as low. As shown in Table 6, concentrations of B[a]P documented in smoked cheese from the Czech market were comparable to levels reported for similar products by Guillén and Sopelana6 , Garcia Falcon et al.12 Anastazio et al.10 and Pagliuca et al.9 from Spanish and Italian markets and even lower than reported by Bosset et al.11 for Swiss cheeses and home-made products from Slovakia examined by Michalski and Germuska.8 Relative abundances of individual PAH groups, when applying classification based on the number of fused aromatic rings (see Table 7), varied largely among examined cheeses, reflecting differences in 1314 smoking practices, as shown in Fig. 4. Regardless the cheese type, 70–90% of total 12 PAHs were those with three aromatic rings: Phe and Ant (i.e. more volatile PAHs). PAHs containing four rings typically contributed to 10–20% of the total PAH content. In unsmoked cheeses and those smoked under the industrial conditions (friction technique and/or controlled wood burning) carcinogenic fiveand six-ring species accounted only for 0.1–0.2% of PAH content. In home-smoked products and cheeses aromatised by liquid and/or dry smoke flavourings, the contribution of this hazardous fraction to the overall contamination was 2–12%. The analysis of liquid smoke-flavouring agent, which was used for sample B20 aromatisation, showed the presence of a high total PAH concentration (182 µg kg−1 ), even those carcinogenic (10 µg kg−1 ). PAH levels found in the liquid smoke flavouring agent are very similar to those published for liquid flavouring agents by Guillén et al.15 but rather higher than those found by Pagliuca et al.9 It should be noted that in spite of this relatively high PAH content, only the background level was found in the final cheese product (see Fig. 2). Several authors in earlier studies concerned with a similar topic reported on a correlation between concentrations of Pyr and B[a]P in some matrices, such as smoke flavourings,15,16 smoked cheese6,7 or charbroiled hamburgers.17 Similarly, a relationship between Phe/Pyr6,7 and Phe/B[a]P7 was investigated. Table 7. PAH groups according to number of aromatic rings Number of rings 2,3 4 5,6 PAH Naph, Ace, Fln, Phe, Ant Flt, Pyr, B[a]A, Chr B[b]F, B[k]F, B[a]P, DB[ah]A, B[ghi]P, I[1,2,3-cd]P J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa PAHs is smoked cheese Dry smoke agent Liquid smoke agent Wood burning - household Wood burning - industrial Friction smoke technique Non-smoked 0 20 40 60 80 100 Relative PAHs abundance (%) SUM of 3-ring PAHs SUM of 4-ring PAHs SUM of 5- and 6-ring PAHs Figure 4. Relative abundances of individual PAH groups in tested cheeses. Data are aggregated based on the smoking technology. Table 8. Ratios of selected PAH concentrations calculated for smoked cheese samples PAH ratio Smoking technology Average Median Min. Max. Phe/Pyr Unsmoked cheese Friction smoke (A5–A9; A11–A12) Wood burning – industrial (C21–C24) Wood burning – household (D25–D27) Liquid smoke (B20) Dry smoke (A19) 11 14 11 6 7 4 10 14 11 6 7 4 8 11 10 5 – – 12 15 11 6 – – Pyr/B[a]P Unsmoked cheese Friction smoke (A5–A9; A11–A12) Wood burning – industrial (C21–C24) Wood burning – household (D25–D27) Liquid smoke (B20) Dry smoke (A19) Unsmoked cheese Friction smoke (A5–A9; A11–A12) Wood burning – industrial (C21–C24) Wood burning – household (D25–D27) Liquid smoke (B20) Dry smoke (A19) 40 30 127 14 6 40 405 408 1359 79 37 160 40 30 114 13 – – 410 382 1220 67 – – 33 26 110 12 – – 390 298 1117 67 – – 45 35 170 18 – – 431 510 1880 104 – – Phe/B[a]P Samples with B[a]P content close to the limit of detection were discarded from the calculation. According to Franklach and Warnatz,18 such relations typically should exist, since heavy PAHs are derived through pyrosynthesis from the lighter PAHs by addition of small units (i.e. acetylene or aryl radicals) what means that an increase of a precursor group during the smoking is accompanied by originating of final reaction products at higher degree. This observation could be useful for predicting the high molecular PAH levels based on the concentrations of lighter PAHs, whose determination is easier, due to their higher concentrations usually presented in analysed matrices. Under these conditions, uncertainty of measurement is lower and accuracy of data is better. Table 8 shows concentration ratios of Phe and Pyr, Pyr and B[a]P, and Phe and B[a]P found for each group of cheeses (grouping based on the smoking procedure). In the case of the Phe/Pyr ratio, relatively consistent results were obtained, both within the each individual group and between the groups (with median J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa values in the range of 4–14), which is in a good agreement with values published by Guillén and Sopelana7 (2.4–12). On the other hand, as regards Pyr/B[a]P and Phe/B[a]P concentration ratios, a distinct diversity between individual cheese groups representing different smoking procedures was found. However, ratios obtained within each group were relatively consistent, at least in terms of orders of magnitude. These results indicate that predicting the levels of high molecular PAHs (typically carcinogenic) from the concentrations of lighter PAHs is not a straightforward approach; nevertheless, it seems that availability of these ratios might be useful to identify the type of smoking technology. Similar strategy employing various PAHs ratios for the identification of the emission sources responsible for environmental pollution was reported in some studies.19,20 Recently, the EU Scientific Committee on Food concluded, that B[a]P can be employed as an indicator 1315 M Suchanová et al. Table 9. Correlations between B[a]P concentrations and other PAH groups PAH groups Sum of 5- and 6-ring PAHs Sum of 4-ring PAHs Sum of 3-ring PAHs Sum of 8 carcinogenic PAHs Sum of 15 PAHs Correlation coefficient 0.993 0.947 0.848 0.995 0.728 of occurrence and concentrations of higher-molecular mass PAHs (from benzofluoranthenes upwards) in food, whereas it cannot be used as an indicator of lower-molecular mass PAHs.3 The possibility of using this approach was demonstrated in a study by Kazerouni et al.21 where the correlation coefficient between concentrations of B[a]P and the sum of the carcinogenic PAHs was 0.98, while it decreased to 0.87 when B[a]P and the total of 15 PAHs were correlated. Almost the same trend was obtained in our study. A high correlation coefficient (0.995) was found for B[a]P concentrations and the sum of eight carcinogenic PAHs (Table 9), while the correlation between B[a]P content and the total content of 15 PAHs was weaker (0.728). CONCLUSIONS In this study, we examined 36 cheese samples smoked by various smoking technologies for the presence of major PAHs. The conclusions based on generated data can be summarised as follows. Firstly, both the use of smoke flavourings and smoking procedure achieved under industrial conditions (friction smoke and wood burning) led to only slightly elevated PAH levels as compared to unsmoked cheeses. Distinctly the highest PAH levels were determined in home-made smoked samples. Secondly, the analysis of cheese rinds has shown that the surface layers are three to six times more contaminated by PAHs compared to the whole sample. Their removal reduced the total PAH content by approximately 50–100%. Thirdly, B[a]P concentrations in smoked cheese strongly correlated with the sum of carcinogenic PAHs, what confirmed the applicability of suggestion of EU SCF to use this analyte as an indicator of the carcinogenic higher-molecular PAH fraction. Lastly, PAH profiles expressed as Pyr/B[a]P and Phe/B[a]P concentration ratios differed according to the smoking technology with the lowest values for home-smoked and liquid smoke flavouring treated cheeses. On the other hand Phe/Pyr ratios were similar for all smoking technologies, and hence not diagnostic. ACKNOWLEDGEMENTS This work was a part of the research project MSM 6046137305 granted by the Ministry of Education, Youth and Sports of the Czech Republic. 1316 REFERENCES 1 Phillips DH, Polycyclic aromatic hydrocarbons in the diet. Mutat Res 443:139–147 (1999). 2 Scientific Committee on Food. Opinion of the Scientific Committee on Food on the risks to human health of polycyclic aromatic hydrocarbons in food. SCF/CS/CNTM/PAH/29 Final. Health and consumer protection directorate - general, Brussels (2002). 3 European Commission. Commission recommendation of 4 February 2005 on further investigation into the levels of polycyclic aromatic hydrocarbons in certain foods (2005/108/EC). Off J Eur Union 48:L34/43–L34/45 (2005). 4 Joint FAO/WHO Expert Committee on Food Additives, Evaluation of Certain Food Contaminants. Report of the 64th meeting, Rome, 8 to 17 February 2005, No. 930. WHO, Geneva (2006). 5 European Commission. Commission regulation (EC) No 1881/2006 of 19 December 2006 setting maximum levels for certain contaminants in foodstuffs. Off J Eur Union 49:L364/5–L364/2 (2006). 6 Guillén MD and Sopelana P, Occurrence of polycyclic aromatic hydrocarbons in smoked cheese. J Dairy Sci 87:556–564 (2004). 7 Guillén MD and Sopelana P, Headspace solid-phase microextraction as a tool to estimate the contamination of smoked cheeses by polycyclic aromatic hydrocarbons. J Dairy Sci 88:13–20 (2005). 8 Michalski R and Germuska R, The content of benzo(a)pyrene in Slovekian smoked cheese. Polish J Food Nutr Sci 12/53:33–37 (2003). 9 Pagliuca G, Gazzotti T, Zironi E, Serrazanetti GP, Mollica D and Rosmini R, Determination of high molecular mass polycyclic aromatic hydrocarbons in typical Italian smoked cheese by HPLC-FLD. J Agr Food Chem 51:5111–5115 (2003). 10 Anastasio A, Mercogliano R, Vollano L, Pepe T and Cortesi ML, Levels of benzo[a]pyrene (BaP) in ‘‘Mozzarella di Bufala Campana’’ Cheese smoked according to different procedures. J Agr Food Chem 52:4452–4455 (2004). 11 Bosset JO, Butikofer U, Dafflon O, Koch H, ScheuererSimonet L and Sieber R, Occurrence of polycyclic aromatic hydrocarbons in cheese with and without a smoke flavour. Sci Aliment 18:347–359 (1998). 12 Garcı́a Falcón SM, González Amigo S, Lage Yusty MA, Lopéz de Alda Villaizán MJ and Simal Lozano J, Enrichment of benzo[a]pyrene in smoked food products and determination by high-performance liquid chromatography–fluorescence detection. J Chromatogr A 753:207–215 (1996). 13 Wenzl T, Simon R, Kleiner J and Anklam E, Analytical methods for polycyclic aromatic hydrocarbons (PAHs) in food and the environment needed for new food legislation in the European Union. Trends Anal Chem 25:716–725 (2006). 14 European Commission. Commission Directive 2005/10/EC of 4 February 2005 laying down the sampling methods of analysis for the official control of the levels of benzo[a]pyrene in foodstuffs. Off J Eur Union 48:L34/15–L34/20 (2005). 15 Guillén MD, Sopelana P and Partearroyo MA, Determination of polycyclic aromatic hydrocarbons in commercial liquid smoke flavorings of different compositions by gas chromatography-mass spectrometry. J Agr Food Chem 48:126–131 (2000). 16 Guillén MD, Sopelana P and Partearroyo MA, Polycyclic aromatic hydrocarbons in liquid smoke flavorings obtained from different types of wood. Effect of storage in polyethylene flasks on their concentrations. J Agric Food Chem 48:5083–5087 (2000). 17 Greenberg A, Hsu CH, Rothman N and Strickland PT, PAH profiles of charbroiled hamburgers: Pyrene/B[a]P ratios and presence of reactive PAHs. Polycyclic Aromatic Compounds 3:101–110 (1993). J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa PAHs is smoked cheese 18 Frenklach M and Warnatz J, Detailed monitoring of PAH profiles in sooting low-pressure acetylene flame. Comb Sci Technol 51:265–283 (1987). 19 Vasilakosa Ch, Levia N, Maggosa T, Hatzianestisb J, Michopoulosa J and Helmisc C, Gas–particle concentration and characterization of sources of PAHs in the atmosphere of a suburban area in Athens, Greece. J Hazard Mater 140:45–51 (2007). J Sci Food Agric 88:1307–1317 (2008) DOI: 10.1002/jsfa 20 Juany B, Zhen H, Juany G, Din H, Li X, Suo H, et al, Characterization and distribution of polycyclic aromatic hydrocarbon in sediments of Haihe River, Tianjin, China. J Environ Sci 19:306–311 (2007). 21 Kazerouni N, Sinha R, Hsu CH, Greenberg A and Rothman N, Analysis of 200 food items for benzo[a]pyrene and estimation of its intake an epidemiologic study. Food Chem Toxicol 39:423–436 (2001). 1317 Journal of the Science of Food and Agriculture J Sci Food Agric 88:1318–1324 (2008) Effect of inulin and Lactobacillus paracasei on sensory and instrumental texture properties of functional chocolate mousse Haı́ssa R Cardarelli,1 Lina C Aragon-Alegro,2 João H A Alegro,1 Inar A de Castro2 and Susana M I Saad1∗ 1 Department 2 Department of Biochemical Technology, University of São Paulo, Av. Prof. Lineu Prestes, 580, 05508-000, São Paulo, SP, Brazil of Food and Experimental Nutrition, University of São Paulo, Av. Prof. Lineu Prestes, 580, 05508-000, São Paulo, SP, Brazil Abstract BACKGROUND: This study evaluated the effect of a potentially probiotic bacteria (Lactobacillus paracasei subsp. paracasei LBC 82), added solely or together with the prebiotic ingredient inulin on instrumental texture attributes and sensory properties of a functional chocolate mousse during storage at 4 ± 1 ◦ C for up to 28 days. RESULTS: The addition of Lactobacillus paracasei resulted in a firmer and more adhesive chocolate mousse. This effect was intensified with the presence of inulin in the synbiotic formulation (5.24 N and −0.956 N, respectively, for firmness and adhesiveness after 28 days of storage) (P < 0.05). L. paracasei population did not vary (P > 0.05) during storage (always between 7.27 and 7.35 log cfu g−1 ), both for the probiotic and the synbiotic mousses. Synbiotic mousse differed from control and probiotic mousses during storage with respect to the color attribute. Moreover, both probiotic and synbiotic mousses presented taste, aroma and texture perceptions which were different from one another and from the control mousse after 14 and 21 days of storage. CONCLUSION: The use of inulin, together with the potentially probiotic strain of Lactobacillus paracasei subsp. paracasei, is advantageous, conferring potentially symbiotic potential to the chocolate mousse, as well as favorable texture and sensory properties.  2008 Society of Chemical Industry Keywords: inulin; Lactobacillus paracasei; sensory evaluation; texture profile analysis; chocolate mousse INTRODUCTION The use of foods that promote a state of well-being, better health and reduction of the risk of diseases has become popular as consumers become more health conscious.1 Some good examples are foods containing physiologically active components such as probiotics and prebiotics. Lactobacillus casei/paracasei strains, for instance, have been widely studied owing to their health benefits, and applied as food probiotics.2 – 4 Prebiotics are ‘dietary carbohydrates having a selective metabolism within the gut flora thereby shifting the community towards a more advantageous structure’.5 Inulin type fructans have been widely investigated as prebiotics and have the most extensive and widespread evidence to support their prebiotic efficacy.6,7 Apart from its nutritional benefits, inulin is used as an ingredient in the formulation of new foods for fat or sugar replacement, as a low-caloric bulking agent and as a texturizing agent.8 The term ‘synbiotic’ is used when a product contains both probiotic and prebiotic ingredients.9 Thus, a product containing inulin and the probiotic L. paracasei, for instance, would fulfill this definition. A broad range of ready-to-eat dairy desserts is available to the consumer, offering a wide variety of textures, flavors and appearances.10 The nutritional and sensory characteristics stimulate their consumption by several groups of consumers, including children and elderly people. Variations in the characteristics of these desserts and the interactions with their ingredients produce noticeable differences in the physical and sensory properties of the formulated products. These differences could influence their acceptability by consumers.11 ∗ Correspondence to: Susana M I Saad, Department of Biochemical Technology, University of São Paulo, Av. Prof. Lineu Prestes, 580, 05508-000, São Paulo, SP, Brazil E-mail: susaad@usp.br Contract/grant sponsor: Coordenação de Aperfeiçoamento de Pessoal de Ensino Superior (CAPES) Contract/grant sponsor: Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP; contract/grant number: 01/10055-0; 00/14681-0; 02/14185-8; 03/13748-1 (Received 26 June 2007; revised version received 12 December 2007; accepted 16 December 2007) Published online 31 March 2008; DOI: 10.1002/jsfa.3208  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Effect of inulin and Lactobacillus paracasei on chocolate mousse properties Sensory attributes can have different levels of importance depending upon the type of food. Texture makes a significant contribution to the overall food quality, contributing more or less equally with both flavor and appearance.12 The texture of semisolid food products is also of importance for their acceptability.13 Texture profile analysis (TPA) is an imitative test that mimics mastication with a texturometer through a compressive force deformation and has proved a valuable aid to assessing food texture. However, care should be exercised in accepting the results for purposes other than comparative evaluation14 and a combination of techniques could provide scientists with an effective tool for predicting the commercial success of a product.15 Probiotics and inulin may influence sensory features of foods. There are a number of studies dealing with the isolated effects of inulin and L. paracasei additions on the sensory properties of non-fermented dairy desserts, including ice cream, frozen vegetarian dessert, starch-based dairy dessert, and coconut flan.16 – 20 However, reports on synbiotic dairy-based products other than yoghurt, fermented milks, and infant formulas are still scarce.21 – 23 Chocolate mousse is an aerated dairy dessert with a stabilized foamy structure that is becoming increasingly popular in the refrigerated dessert market.24 The industrial production of aerated dairy desserts is delicate, requiring knowledge of the formation and stabilization of foam, the use of functional ingredients (emulsifiers, stabilizers), and the interaction and interference of process parameters in the properties of the resulting product.25 It has been seen as an excellent vehicle for the incorporation of probiotics,23 and may even be improved by the presence of inulin. Moreover, to our knowledge, the simultaneous effects of both inulin and probiotics on the instrumental texture and individual sensory attributes (appearance, flavor, and texture) of this nonfermented synbiotic dairy dessert are non-existent. Therefore, this study has aimed to evaluate a functional chocolate mousse to which probiotic (L. paracasei) and prebiotic (inulin) ingredients were added, focusing on the effects of both ingredients on instrumental texture attributes and sensory properties during refrigerated storage (4 ± 1 ◦ C) for up to 28 days. MATERIAL AND METHODS Ingredients for chocolate mousse The following commercial ingredients were employed for the production of chocolate mousse: whole milk cream (25% fat, Nestlé, Araçatuba, Brazil), cocoa powder (Cacau em Pó do Padre, Nestlé, Araras, Brazil), chocolate powder (Nescau, Nestlé, Araras, Brazil), unflavored gelatine powder (Oetker, São Paulo, Brazil), emulsifying agent (Glintex, Lida Mercantil, Diadema, Brazil), sucrose (União, Coopersucar União, Limeira, Brazil), skimmed milk J Sci Food Agric 88:1318–1324 (2008) DOI: 10.1002/jsfa powder (Ninho, Nestlé, Ituiutaba, Brazil), and UHT skimmed milk (Parmalat, Muriaé, Brazil). Production of chocolate mousse Three pilot-scale chocolate mousse formulations, denoted C (control), P (probiotic), and S (synbiotic), were produced in triplicate. Formulations P and S were supplemented with the potentially probiotic microorganism L. paracasei subsp. paracasei LBC 82 (Danisco, Dangé, France) and formulation S was also supplemented with the prebiotic ingredient inulin (Beneo GR, Orafti, Oreye, Belgium). Formulation C was a control, prepared without the addition of L. paracasei or inulin. The ingredients employed for the production of the three formulations are shown in Table 1. Each lot of chocolate mousse was produced in amounts to obtain 2–3 kg of the final product. For this purpose, the ingredients were weighed, mixed (except for the emulsifying agent), heated to 80–85 ◦ C in a water bath, and cooled to 40 ◦ C in an ice bath, being continuously stirred. After reaching 40 ◦ C, L. paracasei subsp. paracasei was added to formulations P and S, in order to obtain concentrations of approximately 7 log cfu g−1 in the final product. Inulin was also added to formulation S. The mixtures were homogenized and, after the addition of the emulsifying agent, they were blended with a domestic mixer (model Pérola Plus, Britânia, São José dos Pinhais, Brazil) at 14 ◦ C in an ice bath, for the incorporation of air, until the volume of the mixture doubled. The resulting products were transferred to individual plastic cups (approximately 35 g per cup), sealed with a metallic cover, and stored at 4 ± 1 ◦ C for up to 28 days. Physical–chemical and compositional analysis After one day of storage at 4 ± 1 ◦ C, the chocolate mousse formulations were freeze-dried, grated, and analyzed for moisture, ash, fat, and protein. Three Table 1. Ingredients and respective quantities (g kg−1 ) employed for the production of the three formulations of chocolate mousse studied Trials Ingredients Milk cream (25% fat) Cocoa powder Chocolate powder Unflavoured gelatine powder Emulsifying agent Sucrose Skimmed milk powder UHT skimmed milk Inulin L. paracasei LBC82 Total C P S 279.00 25.00 10.00 12.50 12.50 110.00 39.00 512.00 – – 1000.00 279.00 25.00 10.00 12.50 12.50 110.00 39.00 511.90 – 0.10 1000.00 267.90 24.00 9.60 12.00 12.00 95.10 37.40 491.80 50.10 0.10 1000.00 C, control: no addition of the probiotic microorganism or of the prebiotic ingredient; P, Probiotic: addition of L. paracasei subsp. paracasei LBC 82; S, synbiotic: addition of L. paracasei subsp. paracasei LBC 82 plus the prebiotic ingredient inulin. 1319 HR Cardarelli et al. independent determinations were made for each formulation, in triplicate. For moisture content determination, 5 g of the samples was submitted to drying at 70 ◦ C under vacuum (Marconi MA030112, Piracicaba, Brazil) for 24 h. Ash was determined gravimetrically by heating the 2 g sample at 550 ◦ C, until completely ashed. Protein was estimated by measuring the nitrogen content using the Kjeldahl method and multiplying it by a conversion factor (6.38), after drying 5 g of mousse samples. Fat was determined through lipid extraction with ethyl ether, using the Soxhlet device. Carbohydrate content was calculated by difference to achieve total content of 1000 kg. Analytical procedures followed the appropriate standard methods.26 Texture profile analysis The TPA method was adapted from an application study described by Stable Micro Systems for mousse (http://www.stablemicrosystems.com/). TPA was determined in quintuplicate samples of the final product (day 1), and after 7, 14, 21, and 28 days of storage at 4 ± 1 ◦ C, using a TA-XT2 texture analyzer (Stable Micro Systems, Haslemere, UK). Individual refrigerated mousse cups containing 35 g of product were submitted to a single compression, using a 5 kg load cell and a 25 mm diameter aluminium cylinder probe set to penetrate to a depth of 10 mm, at 1 mm s−1 speed, and then returned to the starting point, at the same speed. The parameters determined included firmness and adhesiveness, using the Texture Expert for Windows software version 1.2 (Stable Micro Systems). Firmness (N) was recorded as the height of the force peak during the compression and adhesiveness (N s) as the negative force area of the compression. Probiotic viability Viability of L. paracasei subsp. paracasei in chocolate mousses P and S was monitored during the storage period on duplicate samples of the final product (day 1) and after 7, 14, 21, and 28 days of storage. On each sampling day, portions of 25 g were collected aseptically, blended with 0.225 L of 1.00 g L−1 peptone water in a Bag Mixer 400 (Interscience, St Nom, France), and submitted to serial decimal dilutions with the same diluent. L. paracasei subsp. paracasei was counted by pourplating 1 mL of each dilution in DeMan-RogosaSharpe agar (MRS agar, Oxoid Ltd, Basingstoke, UK), acidified to pH 5.4 with acetic acid, after 3 days of anaerobic incubation (Anaerogen Anaerobic System, Oxoid) at 37 ◦ C.27 The colonies were counted and the results were expressed in logarithm of colony-forming units per gram of product (log cfu g−1 ). Experimental design and statistical analysis The experimental treatment and levels constituted a randomized complete block design which was replicated three times, with repeated measures at five time points. The treatments had a factorial structure. 1320 Analysis of variance (ANOVA) for repeated measures, followed by Tukey post hoc test, was used to determine significant differences (P < 0.05) for the viability of the probiotic microorganism added to formulations P and S, for the results on the TPA and for physical–chemical compositional analysis. Data were analyzed using the Statistica 7.0 software (Statsoft, Tulsa, OK, USA). Sensory analysis Formulations C, P, and S were submitted to sensory evaluation, using a randomized complete block design and employing the difference-from-control test,28 after 7, 14, and 21 days of refrigerated storage of the products. Sensory evaluation was carried out at the Faculty of Pharmaceutical Sciences (São Paulo, Brazil) by 18 trained panelists from the Faculty, including professors, students and staff. Sessions were carried out in individual booths between 3:00 and 4:00 p.m., under fluorescent light. Samples were presented in white plastic cups already removed from the refrigerator, and the panel was asked to evaluate the three-digit coded samples from three different mousse formulations (C, P, and S, all of them from the same replicate), using a structured scale from 0 (no difference from the control) to 7 (considerable difference from the control), based on color, aroma, texture, taste, and firmness of each sample compared to the control sample. Sensory data were analyzed by ANOVA two-way (panelists and treatment) for each storage time (7, 14, and 21 days), followed by the Tukey HSD post hoc test for comparison of the treatments. Approval for the study was obtained from the Ethics Committee of the Pharmaceutical Science Faculty, and written consent was given by all volunteers. RESULTS AND DISCUSSION Physical–chemical and compositional analysis The chemical composition of the three chocolate mousse formulations is presented in Table 2. The Table 2. Chemical compositiona of the three chocolate mousse formulations studied Chocolate mousse formulations C Ash Fat Protein Total carbohydratesb Moisture P S 12.20 [0.00]B 12.10 [0.00]B 11.70 [0.10]A 55.50 [0.10]B 55.70 [0.10]B 52.00 [0.30]A 54.80 [0.20]B 55.00 [0.20]B 51.80 [0.20]A 229.30 [2.90]A 229.70 [1.30]A 269.10 [1.00]B 648.10 [2.70]B 647.50 [1.40]B 615.50 [1.00]A Mean g kg−1 [SD] (n = 3). Values obtained by difference [1000 − (ash + fat + protein + moisture)]. C, control; P, probiotic; S, synbiotic. Different capital letters in the same line denote significant differences (P < 0.05) between different formulations. a b J Sci Food Agric 88:1318–1324 (2008) DOI: 10.1002/jsfa Effect of inulin and Lactobacillus paracasei on chocolate mousse properties Texture profile analysis Table 3 shows the results of the relevant instrumental texture attributes – firmness and adhesiveness – during the refrigerated storage of mousses. Firmness is defined as the ‘force required to compress a substance between tongue and palate’ and physically defined as the ‘maximum force required to compress the sample’.14,29 Firmness was statistically different among the formulations studied, for all storage periods, and the synbiotic mousse presented the highest values (5.24 N after 28 days of storage) (P < 0.05) (Table 3). The synbiotic mousse also presented higher total solids content when compared to the other formulations, probably because of the addition of inulin. This additive was described as capable of increasing the viscosity of food products at low concentrations.30 Similar results were described for low-fat dry-fermented sausage prepared with inulin. The meat product became softer but with springiness and adhesiveness comparable to conventional high-fat sausages.31 In the present study, when the same formulation during the storage period evaluated was considered, all formulations (C, P, and S) varied significantly (P < 0.05), presenting increased firmness throughout storage (Table 3). This finding might be attributable to the low storage temperature and interaction between the ingredients present in the formulations. Table 3. Instrumental texture attributes obtained for the chocolate mousse formulations studied during refrigerated storage (4 ± 1 ◦ C) Formulations Control Probiotic Synbiotic Days of storage Firmness (N) Adhesiveness (N s) 1 7 14 21 28 2.03 [0,05]aA 2.31 [0.02]bA 2.88 [0.08]cA 3.28 [0.07]dA 3.85 [0.06]eA −0.721 [0.030]aA −0.763 [0.026]bA −0.793 [0.022]cA −0.822 [0.020]dA −0.863 [0.015]eA 1 7 14 21 28 2.03 [0.04]aA 2.38 [0.05]bB 3.09 [0.05]cB 3.45 [0.05]dB 4.32 [0.09]eB −0.727 [0.022]aA −0.781 [0.023]bA −0.805 [0.023]bA −0.846 [0.022]cA −0.894 [0.021]dB 1 7 14 21 28 2.29 [0.04]aB 3.12 [0.06]bC 3.92 [0.06]cC 4.73 [0.08]dC 5.24 [0.05]eC −0.792 [0.030]aB −0.868 [0.028]bB −0.897 [0.028]bcB −0.928 [0.029]cdB −0.956 [0.033]dC Adhesiveness is defined as the ‘force required to remove the material that adheres to the mouth – generally the palate – during the normal eating process’ and physically defined as ‘the work required to pull the sample away from a surface’.14,29 During the storage period of mousses, the absolute adhesiveness values increased significantly (P < 0.05). The highest adhesiveness results were also found for the synbiotic mousse (−0.956 N after 28 days of storage) when all time points during the storage period are considered together, followed by the probiotic and control mousses. Inulin certainly contributed to the increased adhesiveness of the synbiotic formulation. This is one of the technological properties of inulin, as previously reported by Franck.32 Results found in the present study indicated a clear influence of both probiotic and prebiotic ingredients on firmness and adhesiveness presented by the products. Consequently, a firmer and more adhesive chocolate mousse was obtained, with an intensified effect promoted by the inulin present in the synbiotic formulation. Probiotic viability L. paracasei population in the probiotic and the synbiotic mousses did not vary (P > 0.05) during refrigerated storage (Fig. 1). Populations were always between 7.27 and 7.35 log cfu g−1 . Similar results were previously reported by Aragon-Alegro et al.23 Heenan et al.18 incorporated L. paracasei ssp. paracasei into a vegetarian frozen soy dessert and observed that the population levels remained above 107 cfu g−1 during the 6-month storage period. These authors considered the dessert as a suitable food for the delivery of bacterial probiotic strains, with excellent viability and acceptable sensorial features. Helland et al.,33 studying the growth and metabolism of four probiotic strains (Lactobacillus acidophilus La5 and 1748, Bifidobacterium animalis Bb12, and Lactobacillus rhamnosus GG) in puddings, found viability varying from 8 to 9.1 log cfu g−1 during 21 days of refrigerated storage. However, survival of probiotic strains in 7.80 P S 7.60 Log cfu g -1 significant variations in chemical composition (P < 0.05) observed for the synbiotic mousse were due to the supplementation with inulin, which led to increased total carbohydrate contents, when compared to the control and probiotic mousses. 7.40 7.20 7.00 6.80 1 7 14 21 28 Storage (days) C, control; P, probiotic; S, synbiotic. Mean [SD] (n = 5). Within a column, for each day of storage, different capital letters denote significant differences (P < 0.05) between different trials. For each trial, within a column, different lower-case letters denote significant differences (P < 0.05) between different days of storage. J Sci Food Agric 88:1318–1324 (2008) DOI: 10.1002/jsfa Figure 1. Viability of Lactobacillus paracasei (mean log cfu g−1 ± SD) during the refrigerated storage of the chocolate mousse formulations studied (n = 3); P, addition of Lactobacillus paracasei subsp. paracasei LBC 82; S, addition of L. paracasei subsp. paracasei LBC 82 plus the prebiotic ingredient inulin. 1321 HR Cardarelli et al. chocolate mousses seems to be strain-dependent, as Borges et al.34 observed that Lactobacillus acidophilus, microencapsulated in a calcium alginate matrix, decreased 2 logs in chocolate mousse after 20 days of storage. Sensory evaluation The sensory attributes evaluated through the difference-from-control test are presented in Table 4. Synbiotic mousse differed from the control and probiotic mousses with respect to the color attribute during all the storage periods studied. Apparently, the presence of inulin greatly interfered with panelists’ perception of color. Color is an important appearance characteristic, and its perception differs from person to person, depending on lighting and numerous other factors.35 Moreover, according to Rosenthal,14 color and texture also influence perception of flavor. The sensory panel mentioned that the flavor of the probiotic and of the synbiotic mousses were significantly different, when compared to the control product after 14 days of storage (P < 0.05). When compared to the control product, the synbiotic mousse presented a significantly different taste during the entire shelf-life, whereas the taste of the probiotic mousse was different only after 14 days of storage (P < 0.05). The panelists reported that the texture of the probiotic and mainly of the synbiotic mousse were different from the control mousse during the whole experimental period (Table 4). The sensory firmness of the synbiotic mousse differed significantly from the control mousse during the whole storage period (P < 0.05), similarly to what was observed for the instrumental firmness. Tárrega and Costell11 studied variations in the consistency of commercial samples of Spanish ‘natillas’ (a type of semi-solid dairy dessert) and the relationships between instrumental and sensory measurements. The authors concluded that differences in thickness between samples affected their acceptability by consumers, and the samples showing intermediate values of instrumental consistency were preferred. Composition and structure of food systems interfere with the perception of the oral texture of food, and the use of gelling agents or thickeners modifies the texture of a specific food product. However, the textural agents might also modify flavor perception and vice versa.36 In the present study both probiotic and synbiotic mousses presented taste, aroma and texture perceptions which were different from one another and from the control mousse after 14 and 21 days of storage (Table 4). The presence of inulin in the synbiotic mousse was probably responsible for the markedly modified sensory perception. Buriti et al.37 also reported that fresh cream cheese containing L. paracasei and inulin was described by the sensory panel as having an ‘improved sensory texture’. Tárrega and Costell19 reported that inulin increased thickness and creaminess (texture attributes), and sweetness and vanilla flavor (taste and aroma attributes) in fat-free starch-based dairy desserts, especially in samples with low starch concentrations. In another study, the same authors have pointed out that interactions between ingredients (milk fat and carrageenan gum) not only influenced rheological characteristics of semi-solid dairy desserts but also greatly affected their texture and flavor.38 Some prior studies have dealt with the effects of inulin and L. paracasei additions on sensory properties of dairy desserts and reported divergent results.16 – 18 In the present study, the addition of L. paracasei did not modify the taste and aroma at 7 days of storage. Perception of differences in taste and aroma between the probiotic and the synbiotic formulations and the control occurred only after 14 days of storage (Table 4). On the other hand, Favaro-Trindade et al.,39 studying the sensory acceptability of fermented acerola ice cream, reported that products prepared with the traditional yogurt culture were better accepted than those produced with probiotic cultures, although no strange taste or aroma in the probiotic samples was reported by the panelists. In this study, the trained panel found differences in the sensory attributes of the tested mousses (Table 4). Table 4. Sensory scores for the attributes color, aroma, texture, taste and firmness obtained for the three chocolate mousse formulations studied Sensory attributes Color Aroma Taste Texture Firmness Days of storage 7 14 C 0.00a 1.00a P 1.00a 1.00a S 3.50b 3.00b P <0.01 <0.01 21 7 0.00a 1.00a 2.00b 0.02 1.00a 1.00a 1.50a 0.12 14 21 7 14 21 0.00a 0.00a 1.00a 0.50a 0.00a 2.00b 2.00b 2.00a 2.50b 3.00b 2.00b 2.00b 3.00b 3.50b 3.00b 0.03 <0.01 <0.01 <0.01 <0.01 7 14 1.00a 0.00a 1.50b 1.00b 2.50c 3.00c 0.01 <0.01 21 7 14 21 0.00a 1.00b 2.00c 0.04 0.00a 0.00a 1.50b 0.04 0.00a 0.00a 2.00b 0.01 0.00a 1.00a 2.00b 0.03 C, control; P, probiotic; S, synbiotic. Values expressed by median. Score zero in the scale means ‘no difference from the control’ and score seven means ‘extremely different from the control’. Within a column, for each day of storage, different lower-case letters denote significant differences (P < 0.05) between trials. Values of probability (P) were obtained by ANOVA for each storage day. 1322 J Sci Food Agric 88:1318–1324 (2008) DOI: 10.1002/jsfa Effect of inulin and Lactobacillus paracasei on chocolate mousse properties These differences would possibly not compromise the acceptability of synbiotic and probiotic mousses considering that, in a preliminary study carried out with consumers,23 the differences in preference between samples of control, probiotic and synbiotic mousses after 7 days of storage were not significant. CONCLUSIONS The use of the potentially probiotic strain of Lactobacillus paracasei subsp. paracasei and inulin for the production of chocolate mousse was shown to be advantageous, conferring symbiotic potential to this non-fermented dairy dessert, as well as favorable texture and sensory properties. These are very important attributes for the functional appeal and also quality of the product, leading to good perspectives for its future commercial production. Nevertheless in vivo studies are necessary to confirm its functionality. ACKNOWLEDGEMENTS The authors would like to thank Coordenação de Aperfeiçoamento de Pessoal de Ensino Superior (CAPES) and Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP, grants 01/100550, 00/14681-0, 02/14185-8, and 03/13748-1), for financial support and fellowships. The authors also wish to thank Orafti – Active Food Ingredients, Clariant and Danisco for providing the inulin and the culture employed and Prof Dr Bernadette D.G.M. Franco for her useful suggestions. REFERENCES 1 Mussatto SI and Mancilha IM, Non-digestible oligosaccharides: a review. Carbohydr Polym 68:587–597 (2007). 2 Itsaranuwat P, Al-haddad KSH and Robinson RK, The potential therapeutic benefits of consuming ‘health-promoting’ fermented dairy products: a brief update. Int J Dairy Technol 56:203–210 (2003). 3 Médici M, Vinderola CG and Perdigón G, Gut mucosal immunomodulation by probiotic fresh cheese. Int Dairy J 14:611–618 (2004). 4 Vásquez A, Molin G, Pettersson B, Antonsson M and Ahrne S, DNA-based classification and sequence heterogeneities in the 16S rRNA genes of Lactobacillus casei/paracasei and related species. Syst Appl Microbiol 28:430–441 (2005). 5 Gibson GR, Fibre and effects on probiotics (the prebiotic concept). Clin Nutr Suppl 1:25–31 (2004). 6 Roberfroid MB, Introducing inulin-type fructans. Br J Nutr 93:S13–S25 (2005). 7 Gibson GR, Probert HM, Van Loo J, Rastall RA and Roberfroid MB, Dietary modulation of the human colonic microbiota: updating the concept of prebiotics. Nutr Res Rev 17:259–275 (2005). 8 Tungland BC and Meyer D, Non digestible oligo-andpolysaccharides (dietary fiber): their physiology and role in human health and food. Compr Rev Food Sci Food Safety 1:73–92 (2002). 9 Crittenden R, Karppinen S, Ojanen S, Tenkanen M, Fagerström R, Mättö J, et al., In vitro fermentation of cereal dietary fibre carbohydrates by probiotic and intestinal bacteria. J Sci Food Agric 82:781–789 (2002). J Sci Food Agric 88:1318–1324 (2008) DOI: 10.1002/jsfa 10 Verbeken D, Bael K, Thas O and Dewetiinck K, Interactions between κ-carrageenan, milk proteins and modified starch in sterilized dairy desserts. Int Dairy J 16:482–488 (2006). 11 Tárrega A and Costell E, Colour and consistency of semi-solid dairy desserts: instrumental and sensory measurements. J Food Eng 78:655–661 (2007). 12 Bourne M, Food Texture and Viscosity: Concept and Measurement. Academic Press, San Diego, CA (2002). 13 Engelen L, de Wijk RA, Prinz JF, Janssen AM, van der Bilt A, Weenen H, et al., A comparison of the effects of added saliva, α-amylase and water on texture perception in semi-solids. Physiol Behav 78:805–811 (2003). 14 Rosenthal AJ, Food texture: Measurement and Perception. Aspen, Gaithersburg, MD (1999). 15 Kealy T, Application of liquid and solid rheological technologies to the textural characterization of semi-solid foods. Food Res Int 39:265–276 (2006). 16 Schaller-Povolny LA and Smith DE, Sensory attributes and storage life of reduced fat ice cream as related to inulin content. J Food Sci 64:555–559 (1999). 17 El-Nagar G, Clowes G, Tudorica M, Kuri V and Brennan CS, Rheological quality and stability of yog-ice cream with added inulin. Int J Dairy Technol 55:89–93 (2002). 18 Heenan CN, Adams MC, Hosken RW and Fleet GH, Survival and sensory acceptability of probiotic microorganisms in a nonfermented frozen vegetarian dessert. LWT Food Sci Technol 37:461–466 (2004). 19 Tárrega A and Costell E, Effect of inulin addition on rheological and sensory properties of fat-free starch-based dairy desserts. Int Dairy J 16:1104–1112 (2006). 20 Corrêa SBM, Castro IA and Saad SMI, Probiotic potential and sensory properties of coconut flan supplemented with Lactobacillus paracasei and Bifidobacterium lactis, during shelf life of the product. Int J Food Sci Technol (in press). DOI: 10.1111/j.1365-2621.2007.01585.x (2007). 21 Buriti FCA, Cardarelli HR and Saad SMI, Biopreservation by Lactobacillus paracasei in co-culture with Streptococcus thermophilus in potentially probiotic and synbiotic fresh cream cheese. J Food Prot 70:228–235 (2007). 22 Buriti FCA, Cardarelli HR, Filisetti TMCC and Saad SMI, Synbiotic potential of fresh cream cheese supplemented with inulin and Lactobacillus paracasei in co-culture with Streptococcus thermophilus. Food Chem 104:1605–1610 (2007). 23 Aragon-Alegro LC, Alegro JHA, Cardarelli HR, Chiu MC and Saad SMI, Probiotic and synbiotic chocolate mousse. LWT Food Sci Technol 40:669–675 (2007). 24 Müller-Fisher N and Windhab EJ, Influence of process parameters on microstructure of food foam whipped in a rotor-stator device within a wide static pressure range. Colloids Surf A 263:353–362 (2005). 25 Pires MH, Sobremesas lácteas aeradas: sistemas de estabilização e tecnologia de aeração. Food Ingredients 6:74–77 (2004). 26 Instituto Adolfo Lutz, Normas analı́ticas do Instituto Adolfo Lutz: métodos quı́micos e fı́sicos para análise de alimentos. São Paulo (1985). 27 Oliveira MN, Sodini I, Remeuf F and Corrieu G, Effect of milk supplementation and culture composition on acidification, textural properties and microbiological stability of fermented milks containing probiotic bacteria. Int. Dairy J 11:935–942. 28 Meilgaard M, Civille GV and Carr BT, Sensory Evaluation Techniques. CRC Press, Boca Raton, FL (1999). 29 Szczesniak AS, Texture is a sensory property. Food Qual Prefer 13:215–225 (2002). 30 Murphy O, Non-polyol low-digestible carbohydrates: food applications and functional benefits. Br J Nutr 85:47–53 (2001). 31 Mendoza E, Garcı́a ML, Casas C and Selgas MD, Inulin as fat substitute in low fat, dry fermented sausages. Meat Sci 57:387–393 (2001). 32 Franck A, Technological functionality of inulin and oligofrutose. Br J Nutr 87:287–291 (2002). 33 Helland MH, Wicklund T and Narvhus JA, Growth and metabolism of selected strains of probiotic bacteria in milk1323 HR Cardarelli et al. and water-based cereal puddings. Int Dairy J 14:957–965 (2004). 34 Borges JQ, Ferreira SRSS and Costa GW, Cinética de sobrevivência de Lactobacillus acidophilus microencapsulados em matriz de alginato de cálcio e veiculados em musse de chocolate, in Proceedings of the 19th Congresso brasileiro de ciência e tecnologia de alimentos, ed. by Sociedade Brasileira de Ciência e Tecnologia de Alimentos, Campinas (2004). 35 Pinho O, Mendes E, Alves MM and Ferreira IMPLVO, Chemical, physical, and sensorial characteristics of ‘Terrincho’ ewe cheese: changes during ripening and intravarietal comparison. J Dairy Sci 87:249–257 (2004). 36 Lethuaut L, Brossard C, Rousseau F, Bousseau B and Genot C, Sweetness–texture interactions in model dairy desserts: effect 1324 of sucrose concentration and the carrageenan type. Int Dairy J 13:631–641 (2003). 37 Buriti FCA, Cardarelli HR and Saad SMI, Influence of Lactobacillus paracasei and inulin on instrumental texture and sensory evaluation of fresh cream cheese. Bras J Pharm Sci (in press). 38 Tárrega A and Costell E, Effect of composition on the rheological behavior and sensory properties of semisolid dairy desserts. Food Hydrocoll 20:914–922 (2006). 39 Favaro-Trindade CS, Bernardi S, Bodini RB, Balieiro JCC and Almeida E, Sensory acceptability and stability of probiotic microorganisms and vitamin C in fermented acerola (Malpiguia emarginata, DC.) ice cream. J Food Sci 71:S492–S495 (2006). J Sci Food Agric 88:1318–1324 (2008) DOI: 10.1002/jsfa Journal of the Science of Food and Agriculture J Sci Food Agric 88:1325–1334 (2008) Fruit quality and volatile fraction of ‘Pink Lady’ apple trees in response to rootstock vigor and partial rootzone drying Riccardo Lo Bianco,1∗ Vittorio Farina,1 Giuseppe Avellone,2 Felice Filizzola2 and Pasquale Agozzino2 1 Dipartimento 2 Dipartimento S.En.Fi.Mi.Zo., Sezione di Frutticoltura Mediterranea, Tropicale e Subtropicale, Viale delle Scienze 11, 90128 Palermo, Italy di Chimica e Tecnologie Farmaceutiche, Via Archirafi 32, 90123 Palermo, Italy Abstract BACKGROUND: Partial rootzone drying (PRD) is a novel deficit irrigation technique consisting in the alternated wetting of only one side of the rootzone, which induces partial stomatal closure and increased water use efficiency. The effect of PRD and rootstock vigor on ‘Pink Lady’ apple fruit quality and aroma profile was studied using solid-phase micro-extraction in headspace and gas chromatography/mass spectrometry. RESULTS: PRD irrigation generally did not affect quality attributes, whereas it influenced the aroma of the apple fruit. In particular, PRD improved the aroma of the fruit flesh, while it decreased the volatile fraction in the peel, where most of the compounds are concentrated. Taking into account the relative contribution of the flesh and peel (w/w) to the apple fruit, the volatile content of the entire fruit was increased by PRD irrigation in less vigorous trees on M.9 rootstock, but reduced in more vigorous trees on MM.106 rootstock. CONCLUSIONS: Differences between the two rootstocks were probably due to different ability to extract soil water by the two types of trees. A combination of the less vigorous rootstock and PRD irrigation may induce an improvement in the aroma composition of the apple fruit.  2008 Society of Chemical Industry Keywords: apple flavor; aroma; deficit irrigation; fruit peel; gas chromatography; mass spectrometry; odor units; rootstock vigor; volatile compounds INTRODUCTION Maximizing fruit production and quality with minimum irrigation inputs is essential and regulated deficit irrigation (DI) strategies have been shown to induce beneficial consequences on fruit quality while limiting shoot growth.1,2 In species like apple (Malus domestica Borkh.), however, fruits and shoots grow concurrently3 and water deficit usually reduces fruit size and yields irrespective of timing.4 – 8 Further efforts toward improving irrigation efficiency of grapes (Vitis vinifera L.) in Australia has led to the development of a novel technique – partial rootzone drying (PRD) – in which only one half of the rootzone is irrigated, while the other half is not.9 The physiological basis for PRD is that roots in drying soil produce abscisic acid (ABA), which is translocated to the shoots, indicating a developing soil-water deficit.10 In leaves, ABA induces partial stomatal closure, which reduces transpiration and may increase water use efficiency. Since the other half of the rootzone is kept well watered, the effect on plant water potential is minimal.11 Other metabolic and physiological processes associated with water stress are not affected during PRD.10,12 Rootstock vigor may also alter biomass allocation between roots and shoots,13 especially in response to water deficit,14 and consequently affect resource acquisition. In particular, apple trees grown on vigorous rootstocks like MM.111 show higher root–shoot ratios than trees grown on M.913 and for this reason may perform better under limiting water resources. In addition, water limitation and biomass allocation can have beneficial consequences on apple fruit quality, such as increased flesh firmness, soluble solids, and red peel color.15 – 17 Aroma volatiles are responsible for odor and contribute to the general flavor of the fruit and ultimately to the final perception of the apple fruit quality by the consumer. More than 300 volatile compounds have been identified in apple fruit, with esters accounting for 80–98% of the total fraction,18 – 21 and fatty acids representing the major precursors for volatile formation during fruit ∗ Correspondence to: Riccardo Lo Bianco, Dipartimento S.En.Fi.Mi.Zo., Sezione di Frutticoltura Mediterranea, Tropicale e Subtropicale, Viale delle Scienze 11, 90128 Palermo, Italy E-mail: rlb@unipa.it Contract/grant sponsor: Intramural Scientific Research Fundings of the University of Palermo (Received 19 July 2007; revised version received 23 November 2007; accepted 17 December 2007) Published online 11 March 2008; DOI: 10.1002/jsfa.3210  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 RL Bianco et al. maturation.22 Specifically, palmitic acid, stearic acid, oleic acid, linoleic acid, and triacontane are the main lipids detected in the apple peel at harvest,23 and they may serve as precursors of important regulatory (jasmonates, phosphoinositides) and volatile aroma substances.24 Information on the effects of DI on apple volatiles is rather scarce. Behboudian et al.25 found increased volatile concentrations in fruit from late-season DI but this was not confirmed in another season.17 In a subsequent study, DI increased volatile concentration only after apple ripening or cold storage and not at harvest.26 In grapes, both DI and fixed PRD, but not alternated PRD, tend to increase fruit aroma concentrations.27 This inconsistency could be due to different degrees of water stress developed, as demonstrated in grapes where aroma is enhanced by mild stress but inhibited by severe stress.28 Alternatively, different degrees of fruit maturity, which is known to affect volatile production significantly,29 may have caused those differences. Recent studies conducted on ‘Braeburn’, ‘Fuji’, and ‘Gala’ apples7,30,31 indicate that PRD should allow for apple fruit quality and yields similar to those of conventionally irrigated trees along with a significant reduction in irrigation water. Our main objective was to examine whether PRD, in combination with rootstock vigor, would have beneficial or detrimental effects on the fruit aroma profile as compared to that of fruit from conventionally irrigated trees. MATERIAL AND METHODS The study was conducted near Caltavuturo (37◦ 49′ N and 850 m a.s.l.), in central Sicily. Trees were 49 uniform 5-year-old ‘Pink Lady’ apple trees, 22 grafted on M.9 and 27 on MM.106 rootstock, and trained to a central leader. Trees were planted in single rows (north–south oriented), spaced at 3.5 m between rows and 1 m within the row, and arranged in a randomized block design with three blocks each of three to five trees per rootstock–irrigation combination. The soil type was a sandy clay loam (53.3% sand, 17.6% silt, and 29.1% clay) with pH 7.3 and 1.8% active carbonates. Soil moisture content at field capacity was about 0.26 m3 m−3 and soil water tension around −17 kPa. Trees were drip irrigated using one emitter every 1 m and received conventional cultural care. For the conventional irrigation treatment (CI), all drip emitters on the line located between consecutive trees along the row were left open so that trees were receiving water on both north and south sides of the rootzone. Irrigation of CI treatment supplied 220 mm of water (768 L per tree) distributed in 24 events from 15 July to 29 August. For the PRD treatment, the drip emitter on one side of each tree was closed and the emitter on the other side was left open so that trees were receiving 50% of the CI irrigation water only on one side of the rootzone. Wet and dry sides were alternated every 2–3 weeks 1326 when soil water tension in the dry side reached values of approximately −100 kPa (Fig. 1). The period of irrigation, the interval between irrigation events, and the duration of each event (maximum 4 h) were adjusted to maintain soil moisture above 80% of field capacity (−50 kPa) in the rootzone of CI trees but avoid spreading of wet areas into the dry sides of PRD trees. Climatic parameters and soil water tension were recorded continuously with a µMetos weather station (Pessl, Weiz, Austria) equipped with six Watermark sensors (one for CI and one for each side of PRD, repeated in two blocks) positioned 50 cm away from the drip emitter toward the aisle and at a fixed depth of 45 cm. Fruits were harvested on 7 November and a subsample of 10 fruits per tree was taken to the laboratory to determine average fresh weight, size (height and width with a digital caliper), flesh firmness on two peeled surfaces of each apple using a handheld pressure tester (TR di Turoni & Co., Forlı̀, Italy) mounting an 11 mm plunger tip), total soluble solids with an Atago Palette PR-32 digital refractometer (Atago Co. Ltd, Tokyo, Japan), pH and acidity with a Crison S compact titrator (Crison Instruments SA, Alella, Barcelona, Spain) and expressed as grams of malic acid per liter of juice, percentage and intensity of peel red color (by digital image analysis after photographing each fruit on two opposite sides), and starch pattern index (by I2 –KI staining and digital image analysis of each stained section). The volatile fraction was analyzed separately on fruit peel and flesh by solid-phase microextraction technique in headspace followed by gas chromatography/mass spectrometry (HS-SPMEGC/MS).32 Owing to the time needed for volatile analysis and the fact that extractions were made on fresh samples, each extraction was carried out on three pooled fruits (one per block), and repeated five times per irrigation treatment and rootstock. Fresh peel and flesh were homogenized separately and 2 g of each tissue was transferred into 7 mL vials with pierceable silicone rubber septa coated with polytetrafluoroethylene (PTFE) film. Ten microliters of 1-heptanol water solution (0.822 µg mL−1 ) was used as an internal standard. Two different polydimethylsyloxane (PDMS) fibers (PMDS-100 and 50/30 µm DVB/carboxen/PDMS) were tested in order to choose the one with the best extraction efficiency for our samples. Selection was based on the Fij criterion function introduced by Zuba et al.33 and modified by Hamm et al.34 Fij values were, respectively, 0.47 for peel and 0.33 for flesh with the PDMS-100 fiber and 1.53 for peel and 1.66 for flesh with the DVB/carboxen/PDMS fiber. Chromatograph separation was performed with two different columns: a 30 m long and 0.25 mm inner diameter fused-silica J & W DB5 capillary column (coated with a 0.25 µm thick film of 5% diphenyl–95% PDMS); and a 30 m long and 0.25 mm inner diameter fused-silica Supelcowax-10 (Supelco, Bellafonte, PA, J Sci Food Agric 88:1325–1334 (2008) DOI: 10.1002/jsfa Controlled irrigation and apple fruit flavor USA) column (coated with a 0.25 µm thick film of polyethylene glycol). The best results were obtained with a combination of DVB/carboxen/PDMS fiber and the Supelcowax-10 column. The vials were stirred in a water bath for 15 min at a controlled temperature (27 ± 0.5 ◦ C) in order to reach equilibrium. On the basis of preliminary tests, a 15 min exposition time was suitable for fiber saturation and for reproducibility of the extraction procedure. Desorption time in the chromatograph injector at 250 ◦ C was fixed at 5 min in splitless mode. Separation was performed with a gas chromatograph/mass spectrometer Varian Saturn3 Ion Trap System (Varian Inc., Palo Alto, CA, USA) and the carrier gas (helium) pressure was fixed at 82.7 kPa on the column head. The transfer line and ion trap temperature were 180 ◦ C. The gas chromatograph was programmed for a starting temperature of 40 ◦ C (5 min hold), a ramp of 5 ◦ C min−1 and a final temperature of 220 ◦ C (5 min hold). In the ion trap mass spectrometer, electron ionization mode was set at 70 eV, mass range at 40–400 thompson and frequency at 3 scans s−1 . Collected data were processed with the instrument data system and chromatographic and spectrometric results showed excellent reproducibility (SD ≤ 4%). Each determination was repeated three times. All standard reagents used, namely 1-butanol, 1-hexanol, 1-heptanol, hexyl acetate, ethyl hexanoate, and 2hexen-1-ol, were purchased from Fluka (Buchs, Switzerland). Linear retention indices (LRI) were calculated using Kovats’ equation35 and a sequence of linear hydrocarbons from C10 to C26 . Apple volatile compounds were identified first by a critical and reasoned comparison with mass spectral data within the NIST 2002 library. Subsequently, compounds 1, 3, 4, 5, 7, 10, 11, 12, 13, 22, 27, 28, 29, and 36 (listed in Tables 2 and 3) were verified on the LRI list. In addition, compounds 8, 14, 16, 23, and 26 were compared to their related standards. Semiquantitative determination was carried out by the method of internal standard. The calibration curve was constructed with readings on six 1-heptanol water solutions ranging from 1.64 to 0.10 µg mL−1 (R2 = 0.995); for all compounds identified in apple headspace the relative response factor was assumed to be 1. Odor thresholds of 21 compounds (listed in Tables 5 and 6) found in the literature were used to calculate the corresponding odor units and their log10 was used to express the relative contribution of each volatile to the formation of the final aroma. Data of volatile compounds were reported individually and grouped by chemical class. Data of quality attributes were analyzed by two-way analysis of variance with rootstock and irrigation as the two main factors and block as a non-interacting factor using SYSTAT (Systat Software Inc., Richmond, CA, USA) procedures. Data of volatile compounds were analyzed J Sci Food Agric 88:1325–1334 (2008) DOI: 10.1002/jsfa by two-way analysis of variance with rootstock and irrigation as the two sole factors. Means were separated by Tukey’s multiple range test at P ≤ 0.05. Multivariate linear discriminant analysis (LDA) was performed using all peel and flesh volatile compounds to attempt classification of treatment groups and individuate the set of compounds that would allow for discrimination of treatments. RESULTS Mean daily vapor pressure deficit before the irrigation period ranged between 0.9 and 1.82 kPa, during the irrigation period between 0.75 and 3.35 kPa, and after the interruption of irrigation between 0.26 and 1.67 kPa. Frequent rainfall events during September and throughout fruit harvest and leaf fall (over 150 mm in 25 events) allowed for maintenance of soil moisture levels above 80% of field capacity, which corresponds to a soil water tension of −50 kPa, without any need for irrigation (Fig. 1). The PRD irrigation regime imposed in this experiment did not induce significant differences in plant water status or fruit yields (data not shown). Fruit of trees grown on the M.9 rootstock exhibited greater weight and firmness, but lower acidity than those of trees grown on MM.106, whereas other quality attributes were similar in both rootstocks (Table 1). On the other hand, irrigation did not affect fruit quality attributes, with the exception of an increase in firmness under PRD (Table 1). Thirty-six volatile compounds were found in the fruit tissues of ‘Pink Lady’ apple. LRI and concentrations are reported in Tables 2 and 3 for both rootstocks and irrigation regimes. Profiles included 29 esters, four alcohols, two terpenes, and one aldehyde. Fruit flesh of all treatments showed a similar volatile profile, with a few differences (Table 2). In particular, hexyl acetate was the most abundant compound, followed by 2-methylbutyl acetate, whereas compounds 9, 17, 18, 20, 25, 30, 31, 32, 34, and 35 were generally absent or barely detectable. Hexanal concentration was higher in M.9 than in MM.106, whereas compounds 7, 8, 10, 11, 13, 16, 19, 23, 28, and 29 tended to be more concentrated in PRD than in CI, regardless of the rootstock. Compounds 3 and 26 responded differently to water regimes depending on the rootstock. Specifically, in response to PRD irrigation they tended to increase in M.9 and to decrease in MM.106 (Table 2). Fruit peel of all treatments also showed similar volatile profile and significantly higher concentrations than in the fruit flesh (Table 3). Hexyl acetate was again the most abundant compound, followed by 2methylbutyl acetate, but in this case all compounds were detectable. Compounds 2, 13, 20, and 29 were more concentrated in fruit peel of trees on M.9, whereas compounds 14 and 25 were more concentrated in fruit peel of trees on MM.106. On the other hand, compounds 14, 16, 20, 25, and 34 1327 RL Bianco et al. Figure 1. Soil-water tension at 45 cm of depth for conventional irrigation (CI) and partial rootzone drying (PRD) treatments. PRD north and PRD south indicate position of sensors relative to tree trunk. Data for each line are the average of two measurements. Table 1. Fruit quality attributes of 6-year-old ‘Pink Lady’ apple trees grown on M.9 and MM.106 rootstocks and under conventional irrigation (CI) and partial rootzone drying (PRD) Rootstock Irrigation Weight (g) Diameter (mm) Color index Cover color (%) M.9 CI PRD CI PRD 197 ± 7.80 202 ± 18.8 181 ± 4.30 176 ± 10.2 71.3 ± 3.30 73.9 ± 2.69 71.9 ± 0.71 69.4 ± 1.68 n.s. n.s. 0.94 ± 0.003 0.94 ± 0.006 0.93 ± 0.004 0.94 ± 0.004 n.s. n.s. 91.1 ± 1.54 90.0 ± 2.44 90.5 ± 1.53 94.9 ± 1.91 n.s. n.s. MM.106 ∗ Rootstock effect Irrigation effect n.s. Rootstock Irrigation M.9 CI PRD CI PRD MM.106 Rootstock effect Irrigation effect Firmness (kg cm−2 ) Starch index Soluble solids (◦ Brix) Acidity (g L−1 ) 9.23 ± 0.10 9.80 ± 0.17 8.89 ± 0.12 8.93 ± 0.19 0.20 ± 0.006 0.22 ± 0.013 0.20 ± 0.007 0.19 ± 0.009 n.s. n.s. 14.6 ± 0.34 14.4 ± 0.34 14.6 ± 0.15 15.0 ± 0.16 n.s. n.s. 1.96 ± 0.09 1.91 ± 0.12 2.23 ± 0.08 2.27 ± 0.06 ∗∗ ∗ ∗∗ n.s. Values are means ± standard error. Rootstock × irrigation interaction non-significant. tended to be more concentrated in the peel of CI fruit than in the peel of PRD fruit, regardless of the rootstock, while compound 29 behaved in opposite ways depending on the rootstock (Table 3). Volatile compounds were also grouped into four chemical classes, namely esters, terpenes, alcohols, and aldehydes. In the flesh, esters were the most abundant, followed by alcohols, aldehydes, and terpenes, while in the peel esters were followed by terpenes, alcohols, and aldehydes, regardless of irrigation or rootstock (Table 4). In the flesh, there was a significant irrigation effect on terpenes and alcohols indicating a significant increase in response to PRD, whereas aldehydes were more concentrated in M.9 compared to MM.106, regardless of irrigation. As a result, PRD fruit flesh tended to show a higher total volatile concentration or content per fruit than CI fruit. 1328 In the peel, CI fruit tended to show higher (although non-significant) ester, alcohol, and total volatile concentrations than PRD fruit (Table 4). On the other hand, terpenes increased in PRD fruit of trees on M.9 only, while aldehydes increased in CI fruit of trees on M.9 and in PRD fruit of trees on MM.106 (Table 4). If flesh and peel compounds are pooled together, the total fruit volatile fraction (µg g−1 ) is 143.5 (CI) and 113.4 (PRD) for M.9, and 146.9 (CI) and 112.1 (PRD) for MM.106. This shows a general decrease of volatile concentration in response to PRD irrigation, specifically by 21% in M.9 and by 24% in MM.106. On the other hand, if concentrations of CI and PRD volatile compounds are averaged, trees on M.9 and MM.106 show similar fruit volatile concentrations (128.5 and 129.5 µg g−1 , respectively). However, since the peel represents only approximately 10% of the J Sci Food Agric 88:1325–1334 (2008) DOI: 10.1002/jsfa Controlled irrigation and apple fruit flavor Table 2. Volatile compounds (µg g−1 ) in the fruit flesh of 6-year-old ‘Pink Lady’ apple trees grown on M.9 and MM.106 rootstocks and under conventional irrigation (CI) and partial rootzone drying (PRD) M.9 MM.106 n. LRI Rootstock Irrigation CI PRD CI PRD 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 1041 1044 1054 1063 1080 1110 1131 1141 1151 1189 1208 1218 1232 1235 1268 1278 1283 1315 1323 1333 1341 1345 1359 1379 1405 1414 1418 1421 1433 1483 1485 1517 1616 1620 1731 1753 Ethyl butanoate Propyl propanoate Ethyl, 2-methylbutanoate Butyl acetate Hexanal 2-Methylbutyl acetate Butyl propanoate 1-Butanol Isobutyl butanoate Pentyl acetate 1-Butanol, 3-methyl Butyl butanoate Butyl, 2-methyl butanoate Ethyl hexanoate Isopentyl butanoate Hexyl acetate 2-Methylbutyhyl 2-methylbutanoate (3E),-3-Hexenyl acetate, (E) (3Z),-3-Hexenyl acetate, (Z) Isopentyl 2-methylbutanoate (2E),-2-Hexenyl acetate, (E) Hexyl propanoate 1-Hexanol Heptyl acetate (2E),-2-Hexenyl propionate 2-Hexen-1-ol, (Z) Butyl hexanoate Hexyl butanoate Hexyl 2-methylbutanoate (2E),-2-Hexenyl butyrate (2E),-2-Hexenyl pentanoate Pentyl hexanoate Hexyl hexanoate Butyl octanoate α-Farnesene (Z, E) α-Farnesene (E, E) 0.23 0.02 0.04b 1.82 0.15a 2.91 0.04b 0.07b 0.01 0.19b 0.12b 0.11 0.14b 0.12 0.01 4.66b 0.02 n.d. 0.06b 0.01 0.77 0.11 0.49b 2.80 n.d. 0.05ab 0.02 0.16ab 0.19b 0.01 n.d. n.d. 0.01 n.d. n.d. 0.03 0.27 0.05 0.12ab 1.97 0.19a 3.10 0.11a 0.11a 0.02 0.27a 0.19a 0.12 0.20ab 0.18 0.02 7.71ab 0.01 n.d. 0.09ab 0.01 1.38 0.03 1.07a 3.02 n.d. 0.11a 0.04 0.21ab 0.34ab 0.01 n.d. n.d. 0.03 n.d. n.d. 0.11 0.26 0.03 0.18a 1.74 0.06b 2.64 0.05b 0.05b 0.01 0.16b 0.08b 0.07 0.09b 0.14 0.01 4.65b n.d. 0.01 0.06b 0.02 0.51 n.d. 0.52b 2.45 n.d. 0.05ab 0.02 0.10b 0.22b 0.01 n.d. n.d. 0.02 n.d. n.d. 0.09 0.32 0.04 0.05b 2.31 0.07b 4.51 0.11a 0.09a 0.02 0.31a 0.18a 0.13 0.29a 0.14 0.01 8.59a 0.04 n.d. 0.16a 0.02 0.84 0.09 0.96a 2.97 n.d. 0.04b 0.04 0.35a 0.46a 0.02 0.01 n.d. 0.03 n.d. n.d. 0.09 LRI, linear retention index. Values are means of five replicates. Different letters indicate statistical differences among treatments and within each compound (Tukey’s multiple range test). entire fruit, data of volatile content expressed in mg per fruit show values of the same order for flesh and peel (Table 4). In this case, if flesh and peel contents (mg per fruit) are pooled together, the total volatile content is 5.26 (CI) and 6.09 (PRD) for M.9, and 5.24 (CI) and 4.67 (PRD) for MM.106. When volatile concentrations were converted into odor units by using odor thresholds found in the literature, aroma of both peel and flesh was characterized predominantly by compounds 3, 6, 16, and 29 (Tables 5 and 6). On the other hand, compounds 2, 8, 11, 26, 27, and 28 of the flesh and compounds 8, 11, and 26 of the peel reported log10 of odor units smaller than zero, meaning that they did not contribute to the formation of the fruit flavor. Total odor units were generally higher in fruit tissues of trees on M.9 (60.4) than in those of trees on MM.106 (51.3). In the flesh, total odor units were J Sci Food Agric 88:1325–1334 (2008) DOI: 10.1002/jsfa higher in PRD fruit than in CI fruit (Table 5), while an opposite trend was observed in the peel with CI fruit showing higher odor units (Table 6). LDA of all volatile compounds (concentration) in fruit flesh and peel was able to separate completely the four treatment groups (Fig. 2). Specifically, the canonical score plot shows a greater distance between PRD and CI in M.9 than in MM.106. In this case, the canonical discriminant functions after a forward step analysis included 10 flesh compounds (1,3,14,19,20,21,22,24,25, and 28) and three peel compounds (4, 22, and 23). LDA of flesh or peel volatile compounds alone was not able to separate completely all the four groups of fruit (data not shown). On the other hand, if we consider the 21 available volatile compounds of fruit flesh and peel expressed as odor units, LDA was also able to separate completely 1329 RL Bianco et al. Table 3. Volatile compounds (µg g−1 ) in the fruit peel of 6-year-old ‘Pink Lady’ apple trees grown on M.9 and MM.106 rootstocks and under conventional irrigation (CI) and partial rootzone drying (PRD) M9 MM106 n. LRI Rootstock Irrigation CI PRD CI PRD 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 1041 1044 1054 1063 1090 1110 1131 1141 1151 1189 1208 1218 1232 1235 1268 1278 1283 1315 1323 1333 1341 1345 1359 1379 1405 1414 1418 1421 1433 1483 1485 1517 1616 1620 1731 1753 Ethyl butanoate Propyl propanoate Ethyl, 2-methyl butanoate Butyl acetate Hexanal 2-Methylbutyl acetate Butyl propanoate 1-Butanol Isobutyl butanoate Pentyl acetate 3-Methyl 1-butanol Butyl butanoate Butyl, 2-methyl butanoate Ethyl hexanoate Isopentyl butanoate Hexyl acetate 2-Methylbutyl 2-methylbutanoate 3-Hexenyl acetate, (E) 3-Hexenyl acetate, (Z) Isopentyl 2-methyl butanoate 2-Hexenyl acetate, (E) Hexyl propanoate 1-Hexanol Heptyl acetate 2-Hexenyl propionate, (E) 2-Hexen-1-ol, (Z) Butyl hexanoate Hexyl butanoate Hexyl 2-methylbutanoate (2E),-2-Hexenyl butyrate (2E),-2-Hexenyl pentanoate Pentyl hexanoate Hexyl hexanoate Butyl octanoate α-Farnesene (Z, E) α-Farnesene (E, E) 0.98 0.26a 0.62 10.14 1.91 20.17 2.00 0.22 0.12 2.00 0.69 2.20 2.24a 0.75b 0.18 43.42a 1.29 0.14 1.74 0.43a 14.64 1.02 2.19 3.84 0.18b 0.18 2.91 2.11 13.30b 0.27 0.08 0.24 2.91 0.63a 0.72 12.11 0.79 0.13b 0.69 2.68 0.94 10.41 0.55 0.11 0.10 0.72 0.18 1.21 2.16a 0.66c 0.13 18.98c 0.25 0.05 0.81 0.16b 4.81 1.06 1.49 2.36 0.11c 0.22 2.90 2.29 15.71a 0.15 0.07 0.42 3.29 0.51a 0.83 14.83 0.87 0.11b 0.80 5.53 1.70 21.79 0.85 0.12 0.11 1.20 0.29 1.62 2.06a 0.88a 0.22 33.17b 0.27 0.09 1.16 0.14b 13.56 0.86 1.78 3.36 0.22a 0.33 2.59 2.53 9.73c 0.27 0.09 0.31 3.62 0.66a 0.71 13.09 0.59 0.08b 0.45 3.02 1.14 17.55 0.52 0.08 0.11 0.78 0.20 1.18 1.12b 0.73b 0.21 24.81b 0.57 0.08 0.69 0.09c 3.63 1.11 1.85 2.43 0.15b 0.19 2.29 2.52 7.56d 0.10 0.04 0.27 3.25 0.38b 0.48 8.69 LRI, linear retention index. Values are means of five replicates. Different letters indicate statistical differences among treatments and within each compound (Tukey’s multiple range test). Table 4. Concentration and content per fruit of volatile compounds grouped by chemical class in the fruit flesh and peel of ‘Pink Lady’ apple trees grown on M.9 and MM.106 rootstocks and under conventional irrigation (CI) and partial rootzone drying (PRD) Esters (µg g−1 ) Terpenes (µg g−1 ) Alcohols (µg g−1 ) Aldehydes (µg g−1 ) Total (µg g−1 ) Total (mg per fruit) Flesh M.9 MM.106 CI PRD CI PRD Rootstock effect Irrigation effect Peel M.9 MM.106 Rootstock effect Irrigation effect CI PRD CI PRD 14.6 ± 1.7 17.6 ± 3.1 16.1 ± 3.1 17.7 ± 4.0 n.s. n.s. 0.03 ± 0.01 0.11 ± 0.01 0.06 ± 0.02 0.10 ± 0.03 n.s. 0.72 ± 0.10 1.55 ± 0.20 0.69 ± 0.23 1.12 ± 0.24 n.s. 0.15 ± 0.03 0.18 ± 0.02 0.06 ± 0.02 0.07 ± 0.01 n.s. 15.5 ± 1.8 23.4 ± 4.9 17.9 ± 4.0 19.1 ± 4.3 n.s. n.s. 2.74 ± 0.32 4.26 ± 0.88 2.91 ± 0.65 3.04 ± 0.68 n.s. n.s. ∗∗ ∗∗ 110 ± 26 71 ± 12 112 ± 19 78 ± 19 n.s. n.s. 12.8 ± 1.1b 19.7 ± 1.6a 12.6 ± 1.5b 9.2 ± 1.4b – – 2.80 ± 0.51 2.12 ± 0.31 2.53 ± 0.42 1.93 ± 0.21 n.s. n.s. 2.19 ± 0.10a 0.72 ± 0.15c 0.39 ± 0.06c 1.27 ± 0.07b – – 128 ± 26 90 ± 16 129 ± 20 93 ± 22 n.s. n.s. 2.52 ± 0.50 1.83 ± 0.32 2.33 ± 0.37 1.63 ± 0.39 n.s. n.s. ∗∗ Values are means of five replicates ± standard error. In cases of non-significant interaction, only effects of main factors are reported (two-way ANOVA; ∗∗ P ≤ 0.01; ∗ P ≤ 0.05; n.s., non-significant). In cases of significant interaction, different letters indicate significant differences (Tukey’s multiple range test) among groups and within chemical class and fruit tissue. 1330 J Sci Food Agric 88:1325–1334 (2008) DOI: 10.1002/jsfa Controlled irrigation and apple fruit flavor Table 5. Aroma value in log10 of odor units in the fruit flesh of ‘Pink Lady’ apple trees grown on M.9 and MM.106 rootstocks and under conventional irrigation (CI) and partial rootzone drying (PRD) M.9 n. Fruit flesh 1 2 3 4 5 6 7 8 10 11 12 13 14 16 18 22 23 26 27 28 29 Ethyl butanoate Propyl propanoate Ethyl, 2-methylbutanoate Butyl acetate Hexanal 2-Methylbutyl acetate Butyl propanoate 1-Butanol Pentyl acetate 1-Butanol, 3-methyl Butyl butanoate Butyl, 2-methylbutanoate Ethyl hexanoate Hexyl acetate (3E),-3-Hexenyl acetate, (E) Hexyl propanoate 1-Hexanol 2-Hexen-1-ol, (Z) Butyl hexanoate Hexyl butanoate Hexyl 2-methylbutanoate Odor threshold (µg kg−1 ) 142 5743 0.144 6642 542 1145 2542 50042 544 475046 10043 1743 145 242 244 843 50042 670047 70043 25045 643 Total MM.106 CI PRD CI PRD 2.36 <0 2.63 1.44 1.48 2.42 0.23 <0 1.58 <0 0.04 0.90 2.09 3.37 <0 1.12 <0 <0 <0 <0 1.49 2.43 <0 3.07 1.48 1.57 2.45 0.65 <0 1.73 <0 0.09 1.07 2.25 3.59 0.05 0.52 0.33 <0 <0 <0 1.75 2.41 <0 3.25 1.42 1.10 2.38 0.27 <0 1.51 <0 <0 0.74 2.13 3.37 0.54 <0 0.02 <0 <0 <0 1.56 2.51 <0 2.66 1.54 1.14 2.61 0.63 <0 1.80 <0 0.10 1.23 2.15 3.63 0.24 1.05 0.28 <0 <0 0.14 1.88 21.18 23.03 14.38 17.86 Table 6. Aroma value expressed in log10 of odor units in the fruit peel of ‘Pink Lady’ apple trees grown on M.9 and MM.106 rootstocks and under conventional irrigation (CI) and partial rootzone drying (PRD) M.9 n. Fruit peel 1 2 3 4 5 6 7 8 10 11 12 13 14 16 18 22 23 26 27 28 29 Ethyl butanoate Propyl propanoate Ethyl, 2-methylbutanoate Butyl acetate Hexanal 2-Methylbutyl acetate Butyl propanoate 1-Butanol Pentyl acetate 1-Butanol, 3-methyl Butyl butanoate Butyl, 2-methylbutanoate Ethyl hexanoate Hexyl acetate (3E),-3-Hexenyl acetate, (E) Hexyl propanoate 1-Hexanol 2-Hexen-1-ol, (Z) Butyl hexanoate Hexyl butanoate Hexyl 2-methylbutanoate Odor threshold (µg kg−1 ) 142 5743 0.144 6642 542 1145 2542 50042 544 475046 10043 1743 145 242 244 843 50042 670047 70043 25045 643 Total the four groups of fruit (Fig. 3), but the canonical score plot shows a greater distance between PRD and CI in MM.106 than in M.9. In this case, the canonical J Sci Food Agric 88:1325–1334 (2008) DOI: 10.1002/jsfa MM.106 CI PRD CI PRD 2.99 0.65 3.80 2.28 2.58 3.26 1.90 <0 2.60 <0 1.34 2.12 2.88 4.34 1.84 2.10 0.64 <0 0.62 0.93 3.35 2.90 0.35 3.84 1.61 2.27 2.98 1.35 <0 2.16 <0 1.08 2.10 2.82 3.98 1.37 2.12 0.48 <0 0.62 0.96 3.42 2.94 0.30 3.90 1.92 2.53 3.30 1.53 <0 2.38 <0 1.21 2.08 2.94 4.22 1.65 2.03 0.55 <0 0.57 1.01 3.21 2.77 0.16 3.65 1.66 2.36 3.20 1.31 <0 2.19 <0 1.07 1.82 2.86 4.09 1.60 2.14 0.57 <0 0.51 1.00 3.10 40.21 36.40 37.43 32.82 discriminant functions after a forward-step analysis included seven flesh compounds (2, 3, 7, 13, 14, 22, and 23) and five peel compounds (1, 6, 12, 22, and 27). 1331 RL Bianco et al. Figure 2. Canonical score plot from linear discriminant analysis for concentration (µg g−1 ) data of volatile compounds detected in fruit flesh and peel of ‘Pink Lady’ apple trees grown on M.9 and MM.106 rootstocks and under conventional irrigation (CI) and partial rootzone drying (PRD). Figure 3. Canonical score plot from linear discriminant analysis for odor values (log10 of odor units) of volatile compounds detected in fruit flesh and peel of ‘Pink Lady’ apple trees grown on M.9 and MM.106 rootstocks and under conventional irrigation (CI) and partial rootzone drying (PRD). DISCUSSION AND CONCLUSIONS The PRD irrigation regime generally did not affect fruit quality attributes of ‘Pink Lady’ apples, regardless of the rootstock. Only flesh firmness was significantly increased by PRD, in agreement with data reported for ‘Fuji’ apples.8 Contrasting responses have been reported by various authors concerning other quality attributes. For example, Caspari et al.7 in ‘Braeburn’, Einhorn and Caspari31 in ‘Gala’ and Van Hooijdonk et al.36 in ‘Pacific Rose’ did not find any reduction in fruit quality due to PRD; whereas Lombardini et al.37 and Caspari et al.30 found a significant reduction in ‘Fuji’ final fruit size. These discrepancies may 1332 be attributed to different irrigation (system, delivery rate and timing, amounts, etc.), soil, and climate conditions in the various field trials which may have resulted in different degrees of water stress experienced by the plants. The volatile fraction was more concentrated in fruit peel than in the flesh, and differences in the aroma profile were mainly due to the irrigation regime rather than to the rootstock. Our observations prove that the apple peel is very important for fruit volatile formation. In fact, volatile compounds in apples are primarily synthesized in the peel after lipid and amino acid catabolism.38,39 Lipid and amino acid concentration in the fruit peel may therefore represent a serious limiting factor for aroma volatile production.40 Moreover, in the fruit flesh, PRD irrigation induced higher total volatile concentration or content than CI irrigation. These results in part agree with those reported by Behboudian et al.,25 who observed that late-season DI increases volatile concentrations, although their DI treatment is substantially different from the PRD technique imposed in this study. Also, volatile emissions from entire apple trees was associated with water deficit as shown by a significant negative correlation between rainfall and emissions of α-pinene, the monoterpene camphene, and the two green leaf volatiles, (E)-2-hexenal and (Z)-3-hexen-1ol.41 However, the PRD treatment in our study did not induce significant changes in plant water status, fruit water content, starch index, or flesh firmness compared to the CI irrigation, and the observed increases in volatile concentrations should be due to mechanisms other than water stress, dilutions due to fruit water content, or degree of fruit maturation. For example, in PRD trees, hormonal signals generated in response to soil drying in the non-irrigated portion of the rootzone may have some direct role in the metabolic events that lead to fruit volatile formation, despite the unaltered plant water status. When flesh and peel compounds were pooled together, the total fruit volatile fraction showed a general decrease of volatile concentration in response to PRD irrigation. These apparent volatile reductions in response to PRD irrigation were mainly due to changes in the concentration of peel volatile compounds. However, the peel represents approximately 10% of the entire fruit and the impact of peel volatiles may therefore be greatly reduced if this is taken into account. In addition, data of total volatile content indicate that trees on rootstocks of different vigor may respond differently to irrigation for what concerns fruit volatiles. In particular, PRD seems to increase total fruit volatiles of trees on M.9 but to decrease total fruit volatiles of trees on MM.106. This is in part due to differences in fruit size between rootstocks as in larger fruits (M.9) the flesh:peel ratio increases and the net irrigation effect of the flesh (PRD increasing volatiles) overcomes the effect of the peel (PRD decreasing volatiles). Yet, data of volatile J Sci Food Agric 88:1325–1334 (2008) DOI: 10.1002/jsfa Controlled irrigation and apple fruit flavor concentration (mg g−1 tissue, Table 4) show that the PRD effect in the fruit flesh of trees on M.9 tends to be more marked than in the fruit flesh of trees on MM.106. Hence, the resulting opposite trends for the two rootstocks, in addition to the fruit size component, could also be due to a greater ability of more vigorous trees (MM.106) to extract soil water, by exploring larger soil volumes, compared to less vigorous trees (M.9), which in turn reduces the intensity of rootgenerated hormonal signal. When volatile concentrations were converted into odor units by using odor thresholds found in the literature, very high odor threshold values for a few compounds prevented their contribution to fruit flavor formation. This is true only in theory because single odor units do not take into account possible interactions and synergisms among volatile compounds and with the fruit matrix, which could change the human perception of the fruit flavor. Nevertheless, data expressed in odor units, although determined on only 21 volatile compounds, were consistent with those from the total concentration of the 36 volatiles detected. LDA was able to separate completely the four treatment groups when fruit volatile compounds were expressed both in concentration or odor units. However, for volatile concentration, LDA revealed a stronger effect of PRD on the fruit aroma composition of trees on M.9 than on that of trees on MM.106, whereas, in terms of odor units, LDA indicated a stronger effect of PRD on the fruit flavor expression of trees on MM.106 than on that of trees on M.9. In both cases, the flesh seems to participate more than the peel in the differentiation of the four treatment groups. Our observations show that reducing water inputs according to the PRD irrigation strategy has an impact on the aroma of the apple fruit. In particular, it improves the aroma of the fruit flesh, while it decreases the volatile fraction in the peel where most of the compounds are concentrated. As a result the volatile concentration of the fruit as a whole is partly reduced by PRD. If the relative contribution (in terms of weight) of the flesh and peel to the apple fruit is then considered, the volatile content of the entire fruit is generally improved by PRD irrigation in less vigorous trees on M.9 rootstock, but reduced in more vigorous trees on MM.106 rootstock probably due to different degrees of water stress developed by the two types of trees. Therefore, a combination of the less vigorous rootstock and PRD irrigation may induce an improvement in the aroma composition of the apple fruit. ACKNOWLEDGEMENTS This research was financially supported by the Intramural Scientific Research Fundings of the University of Palermo (ex quota 60%) for year 20042005. J Sci Food Agric 88:1325–1334 (2008) DOI: 10.1002/jsfa REFERENCES 1 Yang J, Zhang J, Huang Z, Zhu Q and Wang L, Remobilization of carbon reserves is improved by controlled soil-drying during grain filling of wheat. Crop Sci 40:1645–1655 (2000). 2 Higgs KH and Jones HG, Water relations and cropping of apple cultivars on a dwarfing rootstock in response to imposed drought. J Hortic Sci 66:367–379 (1991). 3 Forshey CG, Weires RW, Stanley BH and Seem RC, Dry weight partitioning of ‘McIntosh’ apple trees. J Am Soc Hortic Sci 108:149–154 (1983). 4 Lötter J De V, Beukes DJ and Weber HW, Growth and quality of apples as affected by different irrigation treatments. J Hortic Sci 60:181–192 (1985). 5 Ebel RC, Proebsting EL and Patterson ME, Regulated deficit irrigation may alter apple maturity, quality, and storage life. HortScience 28:141–143 (1993). 6 Mpelasoka BS, Behboudian MH and Green SR, Water use, yield and fruit quality of lysimeter-grown apple trees, responses to deficit irrigation and to crop load. Irrig Sci 20:107–113 (2001). 7 Caspari HW, Neal S and Alspach P, Partial rootzone drying. A new deficit irrigation strategy for apple? Acta Hortic 646:93–100 (2004). 8 Leib BG, Caspari HW, Redulla CA, Andrews PK and Jabro JJ, Partial rootzone drying and deficit irrigation of ‘Fuji’ apples in a semi-arid climate. Irrig Sci 24:85–99 (2006). 9 Dry PR and Loveys BR, Factors influencing grapevine vigour and the potential for control with partial rootzone drying. Aust J Grape Wine Res 4:140–148 (1998). 10 Dry PR, Loveys BR, Botting DG and Düring H, Effects of partial rootzone drying on grapevine vigour, yield, composition of fruit and use of water, in Proceedings of the Ninth Australian Wine Industry Technical Conference; 128–131 (1995). 11 Gowing DJG, Davies WJ and Jones HG, A positive root-sourced signal as an indicator of soil drying in apple, Malus × domestica Borkh. J Exp Bot 41:1535–1540 (1990). 12 Dry PR, Loveys BR and Düring H, Partial drying of the rootzone of grape. I. Transient changes in shoot growth and gas exchange. Vitis 39:3–7 (2000). 13 Stutte GW, Baugher TA, Walter SP and Leach DW, Rootstock and training system affect dry-matter and carbohydrate distribution in ‘Golden Delicious’ apple trees. J Am Soc Hortic Sci 119:492–497 (1994). 14 Atkinson CJ, Policarpo M, Webster AD and Kuden AM, Drought tolerance of apple rootstocks: Production and partitioning of dry matter. Plant Soil 206:223–235 (1999). 15 Mills TM, Behboudian MH and Clothier BE, Plant water status and fruit quality in ‘Braeburn’ apples. HortScience 29:1274–1278 (1994). 16 Kilili AW, Behboudian M and Mills T, Water relations, photosynthesis, growth, and yield of ‘Braeburn’ apples under reduced irrigation applied at different stages of the growing season. Gartenbauwissenschaft 61:267–273 (1996). 17 Mpelasoka BS, Behboudian MH, Dixon J, Neal SM and Caspari HW, Improvement of fruit quality and storage potential of ‘Braeburn’ apple through deficit irrigation. J Hortic Sci Biotechnol 75:615–621 (2000). 18 Lopez ML, Lavilla MT, Riba M and Vendrell M, Comparison of volatile compounds in two seasons in apples: ‘Golden Delicious’ and ‘Granny Smith’. J Food Qual 21:155–166 (1998). 19 Lopez ML, Lavilla MT, Recasens I, Riba M and Vendrell M, Influence of different oxygen and carbon dioxide concentrations during storage on production of volatile compounds by ‘Starking Delicious’ apples. J Agric Food Chem 46:634–643 (1998). 20 Lavilla T, Puy J, Lopez ML, Recasens I and Vendrell M, Relationships between volatile production, fruit quality, and sensory evaluation in ‘Granny Smith’ apples stored in different controlled atmosphere treatments by multivariate analysis. J Agric Food Chem 47:3791–3803 (1999). 1333 RL Bianco et al. 21 Echeverria G, Fuentes MT, Graell J and Lopez ML, Relationships between volatile production, fruit quality and sensory evaluation of ‘Fuji’ apples stored in different atmospheres by means of multivariate analysis. J Sci Food Agric 84:5–20 (2003). 22 Fellman JK, Miller TW, Mattinson DS and Mattheis JP, Factors that influence biosynthesis of volatile flavor compounds in apple fruits. HortScience 35:1026–1033 (2000). 23 Noro S, Kudo N and Kitsuwa T, Changes of lipids of ‘Jonagold’ apple peel in the harvest time. J Jpn Soc Hortic Sci 54:116–120 (1985). 24 Song J and Bangerth F, Fatty acids as precursors for aroma volatile biosynthesis in pre-climacteric and climacteric apple fruit. Postharvest Biol Technol 30:113–121 (2003). 25 Behboudian MH, Dixon J and Pothamshelty K, Plant and fruit responses of lysimeter-grown ‘Braeburn’ apple to deficit irrigation. J Hortic Sci Biotechnol 73:781–785 (1998). 26 Mpelasoka BS and Behboudian MH, Production of aroma volatiles in response to deficit irrigation and to crop load in relation to fruit maturity for ‘Braeburn’ apple. Postharvest Biol Technol 24:1–11 (2002). 27 El Ansary DO and Okamoto G, Vine water relations and quality of ‘Muscat of Alexandria’ table grapes subjected to partial root-zone drying and regulated deficit irrigation. J Jpn Soc Hortic Sci 76:13–19 (2007). 28 Peyrot des Gachons C, Van Leeuwen C, Tominaga T, Soyer JP, Gaudillere JP and Dubourdieu D, Influence of water and nitrogen deficit on fruit ripening and aroma potential of Vitis vinifera L cv Sauvignon blanc in field conditions. J Sci Food Agric 85:73–85 (2005). 29 Song J and Bangerth F, The effect of harvest date on aroma compound production from ‘Golden Delicious’ apple fruit and relationship to respiration and ethylene production. Postharvest Biol Technol 8:259–269 (1996). 30 Caspari HW, Einhorn TC, Leib BG, Redulla CA, Andrews PK, Lombardini L, et al, Progress in the development of partial rootzone drying of apple trees. Acta Hortic 664:125–132. 31 Einhorn T and Caspari HW, Partial rootzone drying and deficit irrigation of ‘Gala’ apples in a semi-arid climate. Acta Hortic 664:197–204 (2004). 32 Vas G and Vékey K, Solid-phase microextraction: a powerful sample preparation tool prior to mass spectrometric analysis. J Mass Spectrom 39:233–254 (2004). 33 Zuba D, Parczewski A and Reichenbächer M, Optimization of solid phase microextraction conditions for gas chromatographic determination of ethanol and other volatile compounds in blood. J Chromatogr B 773:75–82 (2002). 34 Hamm S, Lesellier E, Bleton J and Teapla A, Optimization of headspace solid phase microextraction for gas chromatogra- 1334 35 36 37 38 39 40 41 42 43 44 45 46 47 phy/mass spectrometry analysis of widely different volatility and polarity terpenoids in olibanum. J Chromatogr A 1018:73–83 (2003). Kovats ES, Gas chromatographic characterization of organic compounds. L Retention Index of aliphatic halides, alcohols, aldehydes, and ketones. Helv Chim Acta 6:1915–1932 (1958). van Hooijdonk BM, Dorji K and Behboudian MH, Responses of ‘Pacific Rose’ apple to partial rootzone drying and to deficit irrigation. Eur J Hortic Sci 69:104–110 (2004). Lombardini L, Caspari HW, Elfving DC, Auvil TD and McFerson JR, Gas exchange and water relations in ‘Fuji’ apple trees grown under deficit irrigation. Acta Hortic 636:43–50 (2004). Fan X, Mattheis JP, Fellman JK and Patterson ME, Effect of methyl jasmonate on ethylene and volatile production by Summered apples depends on fruit developmental stage. J Agric Food Chem 45:208–211 (1997). Rudell DR, Mattinson DS, Mattheis JP, Wyllie SG and Fellman JK, Investigations of aroma volatile biosynthesis under anoxic conditions and in different tissues of ‘Redchief Delicious’ apple fruit (Malus domestica Borkh.). J Agric Food Chem 50:2627–2632 (2002). Bangerth F, Streif J, Song J and Brackmann A, Investigation into flavor physiology in apple fruit. Acta Hortic 464:189–194 (1998). Vallat A, Gu H and Dorn S, How rainfall, relative humidity and temperature influence volatile emissions from apple trees in situ. Phytochemistry 66:1540–1550 (2005). Flath RA, Black DR, Guadagni DG, McFadden WH and Schultz TH, Identification and organoleptic evaluation of compounds in ‘Delicious’ apple essence. J Agric Food Chem 15:29–35 (1967). Takeoka GR, Flath RA, Mon TR, Teranishi R and Guentert M, Volatile constituents of apricot (Prunus armeniaca). J Agric Food Chem 38:471–477 (1990). Herrmann K, Die Aromastoffe des Obstes. Teil I: Allgemeine Angaben zu den Aromastoffen, ihren Schwellenwerten und ihrer Zusammensetzung. Erwerbsobstbau 33:4–7 (1991). Takeoka GR, Buttery RG and Flath RA, Volatile constituents of Asian pear (Pyrus serotina). J Agric Food Chem 40:1925–1929 (1992). Karahadian C and Johnson KA, Analysis of headspace volatiles and sensory characteristics of fresh corn tortillas made from fresh masa dough and spray-dried masa flour. J Agric Food Chem 41:791–799 (1993). van Gemert LJ and Nettenbreijer AH, Compilation of odour threshold values in air and water. National Institute for Water Supply, Voorburg, The Netherlands (1977). J Sci Food Agric 88:1325–1334 (2008) DOI: 10.1002/jsfa Journal of the Science of Food and Agriculture J Sci Food Agric 88:1335–1343 (2008) A chemometric study of pesto sauce appearance and of its relation to pigment concentration Francesca Masino, Giorgia Foca, Alessandro Ulrici,∗ Laura Arru and Andrea Antonelli Dipartimento di Scienze Agrarie e degli Alimenti, Università degli Studi di Modena e Reggio Emilia, Padiglione Besta - Via Amendola 2, 42100 Reggio Emilia, Italy Abstract BACKGROUND: Pesto sauce is a typical example of a food matrix in which aspect is of key importance to the final judgment of the consumer, and whose color strongly depends on the production process and on the ingredients. In view of this, the aim of the present work is to evaluate the possibility of quantifying the variability of visual aspect of different brands of pesto sauce, and its relation to the concentration of the main pigments. RESULTS: Sensory evaluation of the appearance of 12 commercial pesto samples was carried out by a panel of 16 assessors who evaluated quantitatively six visual attributes, suitably defined for the description of pesto aspect. A quantitative estimate of the performance of the panel was carried out by means of both univariate and multivariate–multiway chemometric tools (parallel factor analysis, PARAFAC). In addition, the relationship between the mean sensory scores values and the concentrations of chlorophylls, pheophytins and carotenoids was investigated by principal components analysis (PCA). Both PCA and PARAFAC showed good clustering of the samples and a satisfactory degree of homogeneity of the assessors. CONCLUSION: Data analysis showed that assessors fundamentally agree about the main visual characteristics of pesto sauces, which are partly correlated with the concentration values of the main pigments.  2008 Society of Chemical Industry Keywords: pesto sauce; sensory evaluation; color; chemometrics; PCA; PARAFAC INTRODUCTION Notwithstanding their increasing importance in the food market, individual perceptions of sensory attributes are frequently subjective and difficult to estimate, because of the influence of various factors, such as age, sex, health conditions and nutritional habits, in addition to social and cultural traditions. In order to reduce the effects of such factors, several experimental methods have been optimized to obtain significant information.1 For these reasons, the application of sensory panels and the development of appropriate methods to handle the sensory data are rapidly increasing with the expansion of the processed food industry,2 and recent surveys suggest that it will continue in this direction in the coming years.3 – 5 Among all food sensory attributes, those which often play the main role in the consumer decisions during purchase are undoubtedly the visual characteristics, such as color and texture, since packaging does not permit the consumer to use senses other than sight.6 – 10 Also a typical Ligurian pasta sauce known as pesto alla genovese, or simply pesto, is usually presented in transparent pots so that the perceived quality can be determined by its visual appearance. In the last few years the diffusion of pesto has been increasing rapidly all over the world and many food industries are working to improve its preparation process.11,12 In Italy, the production of pasta sauces is about 47 000 tons per year with a value of about ¤300 million, of which 10 000 tons per year are of pesto, with a value of about ¤110 million. Moreover, concerning its diffusion throughout the world, pesto sauce is the second most popular pasta sauce after tomato sauce. Pesto sauce is essentially made up of mashed basil leaves, pine nuts, cheese, olive oil, garlic and salt, its overall aspect resembling that of minced leaves in oil (the reader is referred to the literature13 for some sample images). Chlorophylls are the pigments mainly responsible for the green hue of pesto, even though oil, cheese, pine nuts, garlic and other minor compounds play also a role in the overall sauce appearance. In particular, cheese and pine nuts contribute to the white and yellow hue components, the dimensions of their particles also being influential. In addition, pesto sauce commonly also presents brown particles in the green bulk. These dark hues are mainly due to the ∗ Correspondence to: Alessandro Ulrici, Dipartimento di Scienze Agrarie e degli Alimenti, Università degli Studi di Modena e Reggio Emilia, Padiglione Besta Via Amendola 2, 42100 Reggio Emilia, Italy E-mail: alessandro.ulrici@unimore.it (Received 28 August 2007; revised version received 28 November 2007; accepted 7 January 2008) Published online 2 April 2008; DOI: 10.1002/jsfa.3221  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 F Masino et al. degradation of chlorophylls a and b into the respective brownish pheophytins, by substitution of the Mg atom in the porphyrinic ring with two hydrogen atoms.14 – 17 In the industrial production different processes are often used to enhance the shelf life of pesto sauce, such as pasteurization or addition of small amounts of organic acids,12 but these treatments tend to compromise the bright green color of the untreated product due to chlorophyll degradation.18 These characteristics of pesto color, i.e., its inhomogeneous aspect and its strong color variability, led us to deepen our knowledge of its aspectrelated properties by means of different approaches. In a previous study, an automated color-based classification algorithm was developed, which allowed classification of different brands of pesto on the basis of digital RGB images.13 The results obtained showed that heterogeneous food matrices can also be submitted to color analysis when the appropriate data acquisition and elaboration techniques are used, leading to very satisfactory classification models. Moreover, in a subsequent study, the concentration values of the main pigments of pesto sauce were determined by analysis using high-performance liquid chromatography (HPLC).19 In this work, with the aim of further objectively defining the appearance of this food product, the application of sensory analysis to a set of pesto samples is presented. In particular, a group of trained assessors (judges) was asked to define color properties of a set of pesto samples using descriptive sensory techniques.2,20 The measured scores were then analyzed by ANOVA,21 by other univariate approaches, and by multivariate tools. In particular, the three-way nature of the dataset (samples × attributes × judges) was analyzed by parallel factor analysis (PARAFAC),22 – 24 a three-way multivariate explorative tool. This method provided a unique opportunity to obtain interesting indications both on the behavior of the assessors as well as on the properties of the different samples of pesto. Moreover, to obtain an objective confirmation of the dependence of visual aspect on pigment concentration, the relationships among the mean sensory scores values and the pigment concentrations were also investigated by principal components analysis (PCA).21,25 EXPERIMENTAL Samples Twelve different kinds of pesto sauce (samples), indicated by letters A–L, were considered in this study. Samples from A to I, corresponding to nine different brands (one lot for each brand), were purchased from local markets, while the remaining samples J, K, and L were directly supplied by a producer, and correspond to three different lots. Details on the different recipes were unknown, but samples A–I were subjected to pasteurization and addition of small 1336 amounts of organic acids by the producers, while samples J–L, by contrast, had added water-binding substances to lower water activity. For each sample, pesto sauce from two different jars was analyzed. The content of the two jars for each sample was mixed together and an appropriate amount (stored in a dark place at 4 ◦ C and analyzed in the following days) was kept for the determination of pigment concentration. The remaining part was used to fill three anonymous glass pots (subsamples) identified by a numerical code, each one containing about 40 g of pesto. Similarly to the part of sample to be used for the determination of pigment concentrations, the 36 identical glass pots containing the subsamples of pesto were stored in a dark place at 4 ◦ C, to be used for the panel test sessions in the following 3 days. Sensory analysis Education of the panel Sixteen judges (nine men and seven women, with ages ranging from 24 to 55 years) indicated by abbreviations from JUD1 to JUD16, took part as volunteers to the panel test. The panel was selected on the basis of general guidance26 and their interest and delight in pesto. All assessors had already participated in various panel sessions on different food products in the past. For this reason, they were trained in the senses of taste, smell and visual appearance and also in the general rules of sensory analysis, as reported in the standard methods.27 Moreover, a specific training session was dedicated to the visual appearance of pesto samples, where the different attributes used to define aspect-related properties of pesto sauce were discussed, using as examples some of the anonymous pots containing the pesto samples. Evaluation card The sensory evaluation consisted of a descriptive analysis carried out by evaluating different attributes on an interval scale,1,20 using an evaluation card analogous to that reported in Fig. 1. Six sensory attributes pertinent to color and texture of the analyzed samples were chosen, according to the discussion with the assessors in the training section.3 The following terms were selected to describe pesto sauce: green hue (GH), yellow hue (YH), brown hue (BH), white amount (WA), color homogeneity (CH), and particle size (PS). Moreover, the personal preference of the judge (PR), which is a hedonic attribute, was also included. Concerning this latter attribute we are absolutely aware that hedonic scales are usually reserved for consumer populations greater than 30 in number and ideally many times larger than that. However, we decided to include it anyway, in order to have just a preliminary estimate of the relation between delight for pesto and its aspect. Moreover, it is worth noting that multivariate analysis (PCA and J Sci Food Agric 88:1335–1343 (2008) DOI: 10.1002/jsfa Chemometric study of pesto sauce appearance PARAFAC) was performed both with and without PR attribute, and the results were essentially the same (results without PR are not reported for reasons of conciseness). Assessors were asked to record the intensity of each attribute for each subsample, i.e., for each anonymous glass pot, by placing a mark on an unstructured line scale 10 cm long. Evaluation sessions Evaluation sessions were carried out in individual sensory booths to avoid exchange of opinions about samples. The sensory evaluation of the visual characteristics of pesto was conducted in three separate sessions, planned on three consecutive days. In each session, 16 glass pots containing the pesto sauce (12 different samples + 4 replicate samples) were presented to the judges, following a Latin square design25 in order to avoid any time-related effect in the evaluation. This particular sample subdivision and session organization was such that, at the end of the evaluation sessions, each of the 12 pesto samples was submitted four times to the judgment of each assessor. The procedure adopted also allowed checking for possible changes in the samples’ aspect over time. To this end, the presence of a time (i.e., session)related effect was tested by ANOVA and results (not reported for reasons of conciseness) indicated that the aspect of the analyzed samples remained essentially unchanged during the 3 days of panel test sessions. The visual evaluation of the samples was conducted by the assessors without opening the glass pots, thus avoiding use of senses other than sight during evaluation. Determination of pigments In the same samples subjected to sensory evaluation, the amount of the main pigments was also determined. In particular, the following species were quantified: chlorophylls a and b (Chl a and Chl b); lutein (Lut); β-carotene (Car); and pheophytins a and b (Pht a and Pht b). All pigments were determined by reverse-phase HPLC, except for Car, which was directly quantified on the purified extract by visible spectrophotometry. For a detailed description of the procedures adopted for extraction and quantification of pigments, also including data on pigment concentrations, the reader is referred to the original article.19 Data analysis First, the performance of the panel was investigated by one-way and two-way ANOVA. In the mixed models the assessors were considered as a random factor, while the samples were considered as a fixed factor.1,28,29 Moreover, to obtain an estimate of the repeatability in the evaluation of each sample by each assessor with respect to each attribute, a parameter named Score% was defined. In particular, the Score% value has been computed using the following procedure: first, the score values obtained from the evaluation session were autoscaled for each judge and attribute, so that their variance for each judge and attribute (VJUD,ATTR ) equals 1. The variance of the four replicated score values of each sample i obtained from each judge for each attribute was then calculated (VJUD,ATTR,SMPLi ) and the corresponding variance ratio was defined as follows: FJUD,ATTR,SMPLi = = VJUD,ATTR VJUD,ATTR,SMPLi 1 VJUD,ATTR,SMPLi (1) Then, FJUD,ATTR,SMPLi was compared with the corresponding critical value, FCRIT , and the corresponding score value, SJUD,ATTR,SMPLi , was defined as follows: if FJUD,ATTR,SMPLi > FCRIT , then SJUD,ATTR,SMPLi = 1; Figure 1. English version of the evaluation card used for evaluation sessions (the original version was in Italian). In the original version, line scales were 10 cm long. J Sci Food Agric 88:1335–1343 (2008) DOI: 10.1002/jsfa if FJUD,ATTR,SMPLi ≤ FCRIT , then SJUD,ATTR,SMPLi = 0. 1337 F Masino et al. Finally, the Score% value was calculated for each judge over each attribute as follows: N Score%JUD,ATTR = 100 · xijk = SJUD,ATTR,SMPLi i=1 N residual sum of squares, eijk , as defined by the following model equation: (2) where N is the number of samples (12 in this case). Thus, the Score% parameter can be related to the ability of discriminating as many different samples as possible, since it equals 100 when the variance of the scores for each single sample is significantly lower than the variance of all the scores for all the samples, i.e., when the range of the line scale used for evaluating each single sample is considerably shorter than the whole range used from the assessor for the considered attribute. Moreover, in order to properly account for the threeway nature of the data matrix obtained by the sensory evaluation sessions (12 samples × 7 attributes × 16 judges), PARAFAC was employed.22 – 24 PARAFAC is substantially an extension of PCA21,25 to three-way data arrays and is particularly useful for exploration of datasets presenting a three-way nature, such as those derived from sensorial experiments. In fact, this kind of data can be arranged in a cube-shaped array reporting the evaluated samples in the first mode, the sensory variables in the second mode and the judges in the third one. Similarly to PCA, which decomposes the information contained in the (two-way) data matrix into one part reflecting object variation (scores) and another part related to variables (loadings), when dealing with a three-way array the PARAFAC model provides for each one of the three ways (modes) a set of parameters (labeled as loadings of mode 1, 2, and 3), which directly reflect the variability in the respective modes. Therefore in PARAFAC, as in PCA, the variation along each way of the dataset is described by a restricted number of underlying latent variables or factors. The three loading matrices that contribute to the model are defined in a manner to minimize the F aif bjf ckf + eijk (3) f =1 where xijk represents the elements of the three-way array X, F is the number of significant latent variables (factors), and aif , bjf , and ckf are the elements of the loading matrices A, B, and C, respectively. In this way, the PARAFAC model provides information about the variability in the modes of interest and reveals the latent relationships among samples, attributes and assessors. In the present study, no pretreatment was necessary to analyze the threeway array, thus PARAFAC was applied to the raw data matrix. Based on the clear indications given from the core consistency plots and on other diagnostics like residual variation and computation time,23 two factors have been selected. Moreover, the mean scores calculated over all the assessors for each sample and each attribute were analyzed together with the pigment concentration values by PCA. The analysis of variance was conducted by means of the Statistics Toolbox ver. 5 (The Mathworks Inc., Natick, MA, USA), while PCA and PARAFAC were performed using the PLS Toolbox ver. 4 (Eigenvector Research Inc., Wenatchee, WA, USA), both toolboxes running the Matlab 7.0 environment (The Mathworks Inc.). RESULTS AND DISCUSSION As a first screening, one-way ANOVA was carried out to evaluate the discriminant ability of the judges. To this end, for each assessor and each attribute the difference among the values of the given attribute for the different samples (between-samples variance) was compared with the variability of repeated estimations of the same sample (within-sample variance). For clarity of representation, in Fig. 2 we report the Figure 2. Values of log(F/Fcrit ) derived from the one-way ANOVA calculated for each assessor and each attribute over the different samples. 1338 J Sci Food Agric 88:1335–1343 (2008) DOI: 10.1002/jsfa Chemometric study of pesto sauce appearance logarithm of the ratio between the calculated F values and the critical F value, log(F/Fcrit ), with Fcrit corresponding to P = 90%. Positive log(F/Fcrit ) values indicate significant differences among the samples for the given attribute as evaluated by the given judge. The general performance of the judges was fairly similar, even if they showed different behaviors towards the evaluated attributes. In fact, the lines connecting the symbols of the different attributes are frequently crossed, and there are no attributes showing generally higher or lower values for all the assessors. The only partial exception observed was concerning the behavior of color homogeneity (CH) which, even if showing positive values of log(F/Fcrit ) for all the judges, frequently had rather low values. Judges 15 and 16 showed greater discriminant ability on the whole, while judges 3 and 4 showed lower values. In any case, each assessor was able to discriminate significantly among the samples for all the attributes, except for two subjects, who did not discriminate among the pesto samples for the PR attribute (JUD3) and the GH attribute (JUD4). However, since the exclusion of JUD3 and JUD4 did not lead to significant improvements in the global performance of the panel, we decided to present the results of the whole sensory dataset. To estimate the differences among samples considering all the judges contemporarily, in order to evaluate how much the single assessors differed in their judgments, and how much these differences varied with varying samples, the sensory data for each attribute were then subjected to a two-way ANOVA with samples, judges, and their interactions as effects (Table 1). The F values highlighted statistically significant differences among the pesto samples for all the evaluated attributes. In particular, the F values for PS, WA and YH attributes were greater than for GH, suggesting that, conversely to what was expected, the green hue is not the only parameter having a remarkable effect on the perceived quality by consumers. The F values for the judges and for the (samples × judges) interactions were also statistically significant for all the attributes, which is likely due to individual differences in the use of the scale,29 including: (i) the use of different ranges (i.e., more or less variability in the score values), (ii) shift of the mean values (i.e., more or less generous scoring on average) and (iii) nonlinearity (i.e., different increase in the score values with increasing values of the given attribute). An example of this fact (or, at least of the differences (i) and (ii)) is furnished by the box and whiskers plot reported in Fig. 3, representing the distribution along the sensory scale of the score values for the GH attribute expressed by each judge. The different judges tended to use the given scale in a quite different manner. For example, JUD15 used the sensory scale in a perfectly symmetric way, spreading his marks over the whole range. Also JUD9 used approximately all the scale range but in an asymmetric way, usually preferring marks at lower values of the scale; on the contrary, other judges used a smaller J Sci Food Agric 88:1335–1343 (2008) DOI: 10.1002/jsfa portion of the scale, such as JUD10, who concentrated 50% of his marks in the region between 6.5 and 7.5 cm. Similar box and whiskers plots were also been obtained for the other sensory attributes. When considering the differences among samples for a given attribute by means of one- or two-way ANOVA, the corresponding F value indicates how different the samples are, but not how many samples are evaluated as different. In other words, high F values can be obtained by ANOVA even if only one sample differs significantly from the others. Thus F values do not furnish any indication about the number of samples identified as different. Therefore, in order to express the performance of the judges in terms of their ability to separate for each attribute as many different samples as possible, the Score% parameter was calculated, as defined in the ‘Data analysis’ section above. Figure 4 represents the Score% values for each judge with respect to the different attributes. This figure shows how this measure of the judges’ repeatability varies considerably from one attribute to another. In particular, JUD15 seems to be the most consistent, while JUD4 seems the most unreliable, as is clearly pointed out by the solid line, which is a mean of the Score% values for each assessor over Table 1. Mixed analysis of variance (two-way ANOVA) results for the seven sensory attributes (P < 0.001, except where otherwise specified) F values Attributes GH YH BH WA CH PS PR a Sample 12.36 52.60 10.51 37.18 19.75 35.72 7.31 Judge 5.20 13.68 8.25 7.61 6.66 2.71a 7.42 Interaction 6.00 2.21 4.12 3.27 2.14 2.70 4.85 P < 0.01. Figure 3. Box and whiskers plot of the score value distribution for the GH attribute. 1339 F Masino et al. Figure 4. Score% values for the evaluation of each attribute by the 16 judges. The solid line (TOT) represents the mean of the Score% values for each judge. all the attributes. It must be underlined that these observations partially coincide with the considerations made for Fig. 2, indicating that these two evaluation parameters furnish complementary information, but are not redundant. In fact, even if the assessors with the best and the worst overall performances coincide, there are also cases where log(F/Fcrit ) and Score% show different trends. For example, JUD3 gives rather reproducible evaluations, since his Score% values are high on average, but at the same time he does not possess as many high F values for many attributes. Summarizing, based on the results of one-way ANOVA, two-way ANOVA and Score% it can be stated that, though further training of judges could have produced a better overall level of agreement, the results are nevertheless sufficiently coherent. The results obtained by univariate analysis are surely valuable as a first screening of the sensory data. However, more powerful multivariate analysis methods are needed both to gain a better knowledge about the dataset, and to consider the complex mechanisms driving the behavior of the assessors with respect to their interpretation of the meaning of the attributes and to their ability to characterize the different samples. Since the acquired dataset presents a three-way nature, we decided to explore the 3D array (48 samples × 7 attributes × 16 judges) by means of the three-way multivariate PARAFAC method. On the basis of different diagnostics (including core consistency plots, residual variation and computation time) two factors have been selected, which explained 78.72% and 8.94% of the dataset variance, respectively. The scatter plot of the loadings of the two selected factors for mode 1, i.e., the samples mode (Fig. 5), reveals good overall repeatability in the evaluation of the different samples, since the measures over the four replicated subsamples are generally fairly Figure 5. PARAFAC loadings plot for the first two factors for mode 1 (samples mode). 1340 J Sci Food Agric 88:1335–1343 (2008) DOI: 10.1002/jsfa Chemometric study of pesto sauce appearance well grouped together. Moreover, the samples can be roughly divided into five different clusters: the largest one composed of samples D, E, G, H and I, the second one composed of samples J, K and L, then the cluster of samples B and F and finally two isolated clusters for samples A and C, respectively. In particular, the separation of the cluster of samples J, K and L from the others is very likely due to the fact that they are the only samples that had not been subjected to pasteurization and addition of organic acids. The scatter plot of the loadings for mode 2, i.e., the attributes mode (Fig. 6), shows that factor 1 is mainly influenced by GH and CH, and in the second place by BH, YH and PR. This suggests that the panel in general preferred samples with a homogeneous aspect and more brilliant colors (in particular, green). Factor 2 discriminates among two groups of attributes. At positive values are located the WA, YH and PS attributes, evidencing a certain correlation between the attributes related to ‘fair hues’ and particle size. In fact, white and yellow are more evident in those samples having a coarser size, where cheese and pine nut particles are generally more visible. At negative values of factor 2, where all the other attributes are located, it can be seen that GH and CH are particularly close to each other. This fact could indicate that the judges estimated as showing a brighter green hue those samples presenting a more homogeneous aspect. To explain this behavior it must be considered that, though generally pesto is composed of relatively large particles lying in a clear oil phase, in some cases the presence of emulsions is also possible. The turbid aspect of the emulsified oil and water phases could enhance the color homogeneity, contemporarily affecting the lightness of pesto color, due to a lightscattering effect.30,31 Finally, the scatter plot of the loadings for mode 3, i.e., the judges mode (Fig. 7), shows in general a good agreement among assessors, which all lie on the first quadrant. Only JUD3 is partly separated from Figure 6. PARAFAC loadings plot for the first two factors for mode 2 (attributes mode). Figure 7. PARAFAC loadings plot for the first two factors for mode 3 (judges mode). J Sci Food Agric 88:1335–1343 (2008) DOI: 10.1002/jsfa 1341 F Masino et al. Figure 8. Biplot of the first two principal components with the respective explained variance reported in brackets. Circles identify samples and plus signs identify variables. the other judges, closer to the origin of the axes, thus indicating a minor contribution of this judge to the overall assessment of the analyzed samples. As stated previously, the elimination of this assessor from the panel has not led to significant variations in the overall results. Therefore, these results essentially corroborate the previous observations obtained by univariate statistical tools, confirming an overall sufficient agreement among judges. PCA analysis (Fig. 8) was then applied to the (12 samples × 13 variables) autoscaled data matrix containing the average values for the seven sensory attributes plus the average values for the six chemical parameters (pigment concentrations) as variables. Only the first two principal components were found to be significant, explaining more than 70% of the total variance of the data matrix (49.91% for PC1 and 20.35% for PC2). The sensory attribute GH and the chlorophyll content are located very close together in the same quadrant of the figure, while CH remains in the upper right quadrant together with the PR attribute. Confirming the observations with PARAFAC, it seems that the main contribution to the sensory characterization of the analyzed samples is due to green hue and color homogeneity, which exert their positive influence on preference (GH, CH and PR having positive values on PC1), in opposition to fair hues and to particle size (WA, YH and PS having negative values on PC1). The opposite position of WA and PS with respect to CH is the natural consequence of their opposite effects. In fact, the larger is the particle size, the more particles of pine nuts and cheese are visible, and the less homogeneous in general is the product color. Analogous considerations can be made for the opposite position of GH with respect to YH and BH. In the case of Car and Lut, their closeness is very likely the consequence of the common biosynthetic pathway, as for the Chl a–Chl b and Pht a–Pht b couples. As could be expected, GH appears 1342 strictly correlated to the amount of Chl a and b. Conversely, the correlations between the other pigments (pheophytins and carotenoids) and their hues (BH and YH) are less clear. In order to explain this behavior, first it must be noted that chlorophyll and carotenoid concentrations increase contemporarily with plant growth.19 Subsequently, the almost quantitative conversion of chlorophylls into pheophytins in the pasteurized samples can explain the proximity of carotenoids and pheophytins. For the same reason, the almost opposite position of chlorophylls with respect to pheophytins seems not accidental. Moreover, a high amount of cheese and pine nuts (not strictly correlated with carotenoid content) may alter color perception as a consequence of their white and yellow light hue, thus making YH not strictly related to Car and Lut. Also the interpretation of the spatial distribution of the PCA scores seems coherent with the results obtained from PARAFAC. In fact, Fig. 8 shows that the three non-pasteurized pesto samples J, K, and L, are definitely separated from the others, being the only ones showing positive values on PC1, which corresponds to brighter green hues, higher chlorophyll content and higher preference by the assessors. On the contrary, all the other samples are positioned on the opposite hand of the plot, suggesting evident discrimination between non-pasteurized and pasteurized products. Similarly to what was observed in the corresponding PARAFAC plot of Fig. 5, sample C is positioned apart from the others, mainly for its high WA and PS values. Samples D, E, G, H, and I are grouped together in the PCA scores plot as they were in the corresponding plot of PARAFAC, but in the PCA scores plot also sample A and, to a minor extent, sample B become part of the cluster. Conversely, sample F shows a greater separation from this cluster, which can be mainly explained by its higher pheophytin content. The results obtained from PCA analysis on the whole – sensory and compositive – variables set seem therefore to be in agreement with those gained by PARAFAC analysis carried out on sensory attributes only. CONCLUSIONS The results presented suggest that a properly trained panel of judges can furnish useful information about aspect-related attributes of pesto sauce and about their dependence on pigment composition which, in turn, depends on the amount of ingredients and on the production process. Probably the most remarkable evidence furnished by the present study regards the objective evaluation of the agreement between the assessments made by the judges and pigment composition. Notwithstanding the significant interaction found between pesto samples and judges, the overall evaluations are clearly coherent and reproducible, and partially consistent with the results obtained by chromatographic analysis of J Sci Food Agric 88:1335–1343 (2008) DOI: 10.1002/jsfa Chemometric study of pesto sauce appearance pigments. In particular, chlorophyll content has a neat influence on green hue which, together with color homogeneity, seems to have a positive influence on the general appreciation of the product. Conversely, pheophytins and carotenoids have a negative influence on pesto appearance, even if these pigments have only a partial effect on the brown and yellow hues of the product. The strong relations between pigment concentrations and visual aspect, together with the good repeatability of the measured sensory attributes, are leading us to continue research work in this direction, with the aim of evaluating the possibility of quantifying visual attributes and pigment concentrations by appropriate multivariate digital image analysis methods. ACKNOWLEDGEMENTS The authors wish to gratefully thank the judges who kindly took part in the panel test. REFERENCES 1 Lea P, Naes T and Rodbotten M, Analysis of Variance for Sensory Data. Wiley, New York (1997). 2 Lawless HT and Heymann H, Sensory Evaluation of Food: Principles and Practices. Chapman & Hall, New York (1998). 3 Murray JM, Delahunty CM and Baxter IA, Descriptive sensory analysis: past, present and future. Food Res Int 34:461–471 (2001). 4 Muñoz AM, Sensory evaluation in quality control: an overview, new developments and future opportunities. Food Qual Prefer 13:329–339 (2002). 5 Sidel JL and Stone H, The role of sensory evaluation in the food industry. Food Qual Prefer 4:65–73 (1993). 6 Maga JA, Influence of color on taste thresholds. Chem Sens Flav 1:115–119 (1974). 7 DuBose CN, Cardello AV and Maller O, Effects of colorants and flavorants on identification, perceived flavor intensity and hedonic quality of fruit-flavored beverages and cake. J Food Sci 45:1393–1399 (1980). 8 Johnson J and Clydesdale FM, Perceived sweetness and redness in colored sucrose solution. J Food Sci 47:747–752 (1982). 9 Christensen CM, Effect of color on aroma, flavor and texture judgements of foods. J Food Sci 48:787–790 (1983). 10 Christensen CM, Effect of color on judgements of food aroma and flavor intensity in young and elderly adults. Perception 14:755–762 (1985). 11 Previdi MP, Vicini E, Squarcina N and Lusardi C, Aspetto igienico-sanitario e stabilizzazione microbiologica del Pesto Ligure. Ind Conserv 73:272–277 (1997). J Sci Food Agric 88:1335–1343 (2008) DOI: 10.1002/jsfa 12 Vicini E and Previdi MP, Aspetti microbiologici del Pesto Ligure. Ind Conserv 67:426–429 (1992). 13 Antonelli A, Cocchi M, Fava P, Foca G, Franchini GC, Manzini D, et al, Automated evaluation of food colour by means of multivariate image analysis coupled to a waveletbased classification algorithm. Anal Chim Acta 515:3–13 (2004). 14 Di Cesare LF, Forni E, Viscardi D and Nani RC, Changes in the chemical composition of basil caused by different drying procedures. J Agric Food Chem 51:3575–3581 (2003). 15 Lau MH, Tang J and Swanson BG, Kinetics of textural and colour changes in green asparagus during thermal treatments. J Food Eng 45:231–236 (2000). 16 Teng SS and Chen BH, Formation of pyrochlorophylls and their derivatives in spinach leaves during heating. Food Chem 65:367–373 (1999). 17 Forni E, Grezzi M and Polesello A, Determinazione mediante HPLC delle clorofille e delle feofitine nei vegetali freschi e congelati. In Proceedings of 2nd Forum Analitico HP, 22–24 June (1988). Rome, pp. 75–89. 18 Fabiano B, Perego P, Pastorino R and Del Borghi M, The extension of the shelf-life of ‘pesto’ sauce by a combination of modified atmosphere packaging and refrigeration. Int J Food Sci Technol 35:293–303 (2000). 19 Masino F, Ulrici A and Antonelli A, Extraction and quantification of main pigments in pesto sauces. Eur Food Res Technol 226:569–575 (2007). 20 Meilgaard M, Civille GV and Carr BT, Descriptive Analysis in Sensory Evaluation Techniques (3rd edn). CRC Press, Boca Raton, FL (1999). 21 O’Mahony M, Sensory Evaluation of Food: Statistical Methods and Procedures. Marcel Dekker, New York (1986). 22 Bro R, PARAFAC: tutorial and application. Chemom Intell Lab Syst 38:149–171 (1997). 23 Smilde A, Bro R and Geladi P, Multi-way Analysis: Applications in the Chemical Sciences. Wiley, Chichester (2004). 24 Henrion R, N-way principal component analysis theory, algorithm and applications. Chemom Intell Lab Syst 25:1–23 (1994). 25 Massart DL, Vandenginste BGM, Buydens IMC, De Jong S, Lewi PJ and Smeyers-Verbeke J, Handbook of Chemometrics and Qualimetrics: Part A. Elsevier, Amsterdam (1997). 26 ISO 8586-1, (1993). Sensory analysis – general guidance for the selection, training and monitoring of assessors. Part 1: Selected assessors (1993). 27 ISO 3972, Sensory analysis – Methodology – Method investigation sensitivity of taste (1991). 28 Lundahl DS and McDaniel MR, The panelist effect fixed or random. J Sens Stud 3:113–121 (1988). 29 Naes T and Langsrud O, Fixed or random assessors in sensory profiling? Food Qual Pref 9:145–152 (1998). 30 McClements DJ, Colloidal basis of emulsion color, Curr Opin Colloid Interface Sci 7:451–455 (2002). 31 McClements DJ, Theoretical prediction of emulsion color, Adv Colloid Interface 97:63–89 (2002). 1343 J Sci Food Agric 88:1344–1353 (2008) Journal of the Science of Food and Agriculture Effect of electrical stimulation, delayed chilling and post-mortem aging on the quality of M. longissimus dorsi and M. biceps femoris of grass-fed steers† Regina H Razminowicz, Michael Kreuzer and Martin RL Scheeder∗ ETH Zurich, Institute of Animal Science, Universitätstrasse 2, CH-8092 Zurich, Switzerland Abstract BACKGROUND: Roughage-based low-input beef production systems are gaining increasing interest owing to the perceived ecological advantages and potential health benefits associated with the favourable fatty acid composition of such beef. The low plane of nutrition may on the other hand yield less tender beef by affecting growth, carcass weight and fatness and therefore, indirectly, early post-mortem (p.m.) proteolytic enzyme activity and sarcomere shortening. This study aimed to examine delayed chilling and electrical stimulation as promising techniques to control early p.m. muscle metabolism in a way that improves the tenderness of beef from purely grass-fed steers in comparison with that from steers receiving a finishing diet with concentrates. RESULTS: Electrical stimulation decreased the pH at 1.5 and 3 h p.m. in the M. longissimus dorsi (LD) and M. biceps femoris (BF) of the treated carcass sides as well as the maximum shear force in the LD, while delayed chilling had no effect on pH or texture. The interactions of carcass fatness with electrical stimulation (P = 0.025) and delayed chilling (P = 0.089) indicated more pronounced effects of the p.m. treatments on beef texture in leaner carcasses. CONCLUSION: Electrical stimulation, but not delayed chilling, could markedly improve pasture beef texture and reduce the aging period needed for proper tenderisation.  2008 Society of Chemical Industry Keywords: beef; electrical stimulation; chilling; tenderness; aging; pasture INTRODUCTION It has been argued1 that eating satisfaction depends on a combination of pleasant flavour, juiciness and tenderness and that all three components should be considered for meat quality assessment. Nevertheless, toughness seems to represent the most important cause of consumer dissatisfaction in beef, and variability in meat quality, particularly in texture, is regarded as a major problem worldwide for both the meat industry and consumers.2 Toughness depends mainly on the amount and maturity of connective tissue in the meat (background toughness), the myofibrillar toughness, depending on the contraction state and fragmentation of contractile proteins in the muscles,3 and the interaction between these compounds. The properties of these proteinaceous structural compounds of beef texture and the enzymes controlling their synthesis and degradation may depend on the breed and genotype of the animal. For instance, a mutation in the myostatin gene, causing the double-muscled syndrome, improves beef tenderness.4 There are also DNA tests for genetic markers of tenderness available.5 However, slaughter and chilling technology and aging are further major factors influencing beef texture. Myofibrillar toughness of meat is mainly influenced by two types of process: development of rigor mortis and enzymatic tenderisation during aging.6 Rapid chilling of beef carcasses after slaughter is common practice today in order to limit weight loss and maintain a satisfactory microbial status, but it also facilitates the development of myofibrillar toughness. Since the degree of shortening of pre-rigor muscle, which is temperature-dependent,7 is a key factor in meat tenderness,8 it is clear that the time–temperature relationship during rigor development has a most profound effect on myofibrillar toughness.3 This may explain why studies comparing the palatability of beef from cattle fattened on forage- versus concentrate-based diets have produced inconsistent results.9,10 The outcome of such studies may well be ∗ Correspondence to: Martin RL Scheeder, Swiss College of Agriculture, Länggasse 85, CH-3052 Zollikofen, Switzerland E-mail: martin.scheeder@suisag.ch † The experiment was approved by the Cantonal Veterinary Office, Zug, Switzerland under approval number ZG 33/04 Contract/grant sponsor: Hermann Herzer Foundation (Received 3 July 2007; revised version received 26 December 2007; accepted 7 January 2008) Published online 28 March 2008; DOI: 10.1002/jsfa.3222  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Beef quality in grass-fed steers: effect of chilling regime and electrical stimulation biased by interactions between the chilling regimens applied and carcass weight and fatness. Because cerealbased (high-energy) diets often yield heavier and fatter carcasses than forage-based diets, tenderness might be increased as a result of decelerated post-mortem (p.m.) temperature decline in the muscles.11 Simultaneously, enhanced early p.m. activity of proteolytic enzymes in the muscles may additionally increase the ultimate tenderness of beef from fatter carcasses.11,12 Reduced tenderness in relation to extensive feeding could therefore be at least partly the result of interactions with slaughter/chilling technology. Supplementation of grass with some concentrate may produce more tender meat,13,14 although this was not always found,15 and no difference between pasture beef and conventional beef obtained at the point of sale was found in a Swiss retail study.16 Nevertheless, forage-only fattening, applied to make use of the ecological and economic merits of extensive (low-input) beef production on grasslands as well as for the potential health benefits associated with changes in the fatty acid composition of such beef,16 bears the risk of increased toughness.17 Appropriate p.m. treatment of carcasses may help to prevent this. Technologies to decrease myofibrillar toughness by preventing cold shortening or reduced proteolytic activity in the muscles include electrical stimulation of carcasses18 and delayed chilling, either accelerating p.m. muscle metabolism and inducing a rapid pH decline19 or simply reducing the cooling rate.20 The hypothesis to be tested in the present investigation, therefore, was that different p.m. technologies, namely electrical stimulation and delayed chilling, improve the texture of beef from grass-fed animals, particularly in lean carcasses. For this purpose, steers fattened in a controlled feeding experiment were employed.13 By supplementation with concentrate during finishing of some of the animals, variation among carcasses in fat cover was obtained. The effects of the technologies applied were tested within animals, with one carcass side having been exposed to either electrical stimulation or delayed chilling and the other side acting as the conventionally treated control. MATERIALS AND METHODS Animals Thirty-four Limousin-sired crossbred steers out of Brown Swiss (n = 17) and Holstein-Friesian (n = 17) dams were fed on a forage-only diet (grass in summer, grass silage and hay in winter) until reaching 470 kg live weight. During subsequent finishing up to 560 kg (±6 kg standard deviation) live weight and 560 days (±16 days) of age on average the diet was based on a 95:5 (w/w) grass/hay mixture offered ad libitum. Half of the steers additionally received 3 kg of cerealbased concentrate. The diets were supplemented by a 1:1 (w/w) mixture of sodium chloride and commercial vitamin/mineral mix at 100 g day−1 per head. Eight Brown Swiss and nine Holstein-Friesian J Sci Food Agric 88:1344–1353 (2008) DOI: 10.1002/jsfa crossbred animals were allocated to the grass/hay-only treatment, while nine Brown Swiss and eight HolsteinFriesian crossbreds were allocated to the concentratesupplemented feeding. The animals were housed in six pens of six animals each and had responder-controlled access to individual feeding places. Slaughter process and chilling procedure Animals were fasted for 12 h before being weighed in the morning of the slaughter day. Within the next few hours the animals were transported to a commercial slaughter plant. Slaughter took place within 2 h of departure from the farm. The steers were stunned with a captive bolt, exsanguinated and dressed following commercial procedures. Hot carcass weights were recorded 30 min p.m., and the right carcass sides of 16 animals (eight from each dam breed group) were electrically stimulated (ES) with 230 V and 60 Hz for 30 s (Elektrostimulationsgerät Type IMA, Schermer & Co., Apparatebau, Ettlingen, Germany). The current was applied through clips attached to the muscles of the neck region and the Achilles tendon. After stimulation, both carcass sides were chilled for 90 min in a blast chiller at a temperature of −5 ◦ C, 90% relative humidity and an air speed of about 4 m s−1 . These carcasses are further referred to as data set 1. For the other 18 animals (nine per dam breed group; data set 2) the left carcass sides were chilled in the same way, while the right sides of the carcasses were subjected to delayed chilling (DC) by holding them for 90 min at a temperature of 15 ◦ C. From then on, all carcasses were stored in a chilling room at a temperature of 2 ◦ C on average (air speed 0.2 m s−1 , humidity 90%). Early post-mortem temperature and pH measurement Temperature and pH were measured in the M. longissimus dorsi (LD; between the 10th and 11th ribs) and M. biceps femoris (BF) at 1.5 h p.m. (pH1.5h ) and 3 h p.m. (pH3h ) using an IP67 electrode (model SenTix 21) attached to a WTW-340 pH meter (Wissenschaftliche Technische Werkstätte, Weilheim, Germany). The temperature of both muscles was measured with a PT-100 probe mounted on a TTX 290 SKW (Ebro, Ingolstadt, Germany) at depths of 3, 5.5 and 8 cm. Sample preparation Two days after slaughter the left side of each carcass was dissected according to a standardised industry procedure and the amount of saleable meat, bone and dissected fat (further referred to as ‘cutting fat’) was recorded. The amount of cutting fat was on average 8.4% of the cold carcass weight and ranged from 5.2 to 13.9%. Following dissection, meat samples from the LD (∼2 kg; between the 9th and 12th ribs) and BF (∼2 kg; between the line from the ischium to the acetabulum and a parallel cut further distal at about 30% of the length of the femur) were taken from both 1345 RH Razminowicz, M Kreuzer, MRL Scheeder carcass sides and transported in a refrigerated box to the laboratory. Directly after transport the ultimate pH was measured and 2.5 cm thick slices were cut for colour measurements and later texture analysis. The remaining part was vacuum packaged and kept to age at 2 ◦ C for 15 days, while another 2.5 cm slice was taken and stored frozen. From the LD an additional 2.5 cm slice was aged for a total of 29 days and then stored at −20 ◦ C until analysis. Meat colour and texture measurements Meat colour was always measured at the same three positions on the freshly cut surface after blooming for 1 h at 4 ◦ C, using a Chroma Meter (model 300CR, Minolta, Dietikon, Switzerland) with an observer angle of 0◦ and applying the L∗ , a∗ , b∗ colour system with D65 as light source. However, because no treatment effect on beef colour was detected, data of the colour measurements are not reported in the tables nor referred to in the discussion. After the colour measurements the slices were weighed, vacuum packaged and kept frozen at −20 ◦ C until used for texture measurements. For those, the samples were thawed for 24 h at 4 ◦ C, then grilled to a core temperature of 72 ◦ C in an electrical double-contact grill (model TURMIX 246, Beer Grill, Zurich, Switzerland) heated to 240 ◦ C. The core temperature was controlled by two thermocouples (Thermo ZA9020-FS, NiCr-Ni Type K) attached to a data logger (ALMEMO model 3290-8, Ahlborn, Holzkirchen, Germany) and placed in the centre of the slices. Cooking loss was determined after cooling the grilled slices on a grid for 30 min at ambient temperature. From each of the LD and BF samples a total of six cores (cylinders of 1.27 cm diameter) and six strips (1 × 1 cm2 cross-section) were prepared by drilling and cutting respectively parallel to the muscle fibre direction. The LD samples were taken from six defined positions in the arrangement shown in Fig. 1. The BF samples were also taken from six defined position of two rows from the superficial to the deep side of the muscle and three columns from the lateral to the medial side of the muscle. The cores were sheared by a modified Warner-Bratzler shear blade (slot width 3.3 mm, 3 mm blade, measuring in compression) and the strips by a Volodkevich device, both mounted on a TA-XT2 texture analyser (Stable Micro System, Godalming, UK). The texture measurements were performed on slices at all aging stages (2, 15 and 29 days for the LD and 2 and 15 days for the BF). Statistical analyses Data were statistically analysed with the SAS program (Version 8.0, SAS Institute Inc., Cary, NC, USA), applying the General Linear Model (GLM) procedure for analyses of variance and using different models in a block design with the animals as block. Data sets 1 and 2, i.e. ES and DC, were analysed separately. In both cases the model included aging time, p.m. treatment (ES versus non-stimulated (CO1) 1346 2 1 1 2 3 4 4 5 5 6 3 6 Figure 1. Locations within M. longissimus dorsi for collection of cores (circles 1–6) for determination of maximum shear force (Warner-Bratzler device) and strips (squares 1–6) for measurement of compression energy (Volodkevich device). or DC versus conventional (CO2)), dam breed (Brown Swiss or Holstein-Friesian), feeding treatment and the interactions aging time × p.m. treatment, p.m. treatment × genotype and p.m. treatment × cutting fat as effects. The model to calculate the effects of core/strip location within beef slices included position, aging time, treatment and the interactions between these effects. Data were further subjected to Pearson correlation analyses. Additional GLM procedures considering the proportion of cutting fat or the slaughter weight as co-variables were applied to evaluate possible linear effects of these variables under different chilling regimens on meat texture properties. RESULTS Meat temperature and pH The pH and temperature profiles measured in the LD and BF illustrate that ES and DC resulted in different rates of pH and temperature decline when compared with the conventionally chilled controls (Fig. 2). The temperature of the LD at 1.5 and 3 h p.m. was slightly higher with ES compared with CO1. This effect was more pronounced deep inside the muscle. The major effect of ES, however, was a more rapid pH decline in both muscles. The mean pH3h was slightly below 6 in the LD and slightly above 6 in the BF in ES, while it was still at above 6.5 in CO1. As intended, the DC treatment resulted in a higher temperature compared with CO2 in both muscles and at all locations and times of measurement. In contrast, there was no or only a very small effect on pH, resulting in a pH3h of around 6.4 in both treatments. A relationship between either cutting fat proportion or carcass weight and early p.m. muscle temperature or pH in the LD was only observed in the DC treatment, with correlations between carcass weight and temperature taken at the J Sci Food Agric 88:1344–1353 (2008) DOI: 10.1002/jsfa m. longissimus dorsi CO1/ES 40 35 o 30 25 CO1 ES * n.s. * o 20 15 8 *** Temperature °C Temperature °C Beef quality in grass-fed steers: effect of chilling regime and electrical stimulation *** pH pH 4 30 *** *** *** 25 20 15 8 n.s. o 3 cm 5.5 cm 8 cm 1.5 h 3 cm 5.5 cm 8 cm 3h 4 3 cm 5.5 cm 8 cm 3h m. biceps femoris CO1 / ES * CO1 ES n.s. n.s. n.s. n.s. n.s. *** Temperature °C 3 cm 5.5 cm 8 cm 1.5 h Temperature °C 35 CO2 DC *** 0 0 m. biceps femoris *** CO2 / ES 40 30 CO2 DC *** *** 35 *** *** 25 *** 20 15 8 *** 6 n.s. n.s. 3 cm 5.5 cm 8 cm 1.5 h 3 cm 5.5 cm 8 cm 3h 6 pH pH *** 2 2 4 4 2 2 0 40 6 6 40 35 30 25 20 15 8 m. longissimus dorsi CO2 / ES *** 3 cm 5.5 cm 8 cm 1.5 h 3 cm 5.5 cm 3h 8 cm 0 Figure 2. Temperature (lines) and pH (bars) at 1.5 and 3 h p.m. in M. longissimus dorsi and M. biceps femoris from steer carcasses treated by two different regimes: CO1, control group 1; ES, electrically stimulated group; CO2, control group 2; DC, delayed chilling group. Error bars indicate standard deviations. Comparisons between p.m. treatments within the same experimental data set were done with the t test: ∗∗∗ P < 0.001; ∗ P < 0.05; ◦ P < 0.1; n.s., not significant. different depths (3, 5.5 and 8 cm) and times (1.5 and 3 h p.m.) ranging from 0.44 (P = 0.06) to 0.6 (P = 0.009). The p.m. treatments had no effect on the ultimate pH of meat aged for 2 and 15 days. However, in both muscles of CO1/ES and in the BF of CO2/DC the pH increased with aging time (Table 1). No differences in early p.m. temperature decline in the LD and BF of steers of different dam breeds were observed (data not shown). In experimental data set 1 (ES and CO1), pH1.5h and pH3h were lower in the LD of the Limousin × Holstein steers (6.42 and 6.15 respectively) compared with the LD of the Brown Swiss progeny (6.62 and 6.35 respectively). A significant interaction between breed type and chilling treatment occurred only in the LD in experimental data set 1, with the LD of steers from Brown Swiss dams showing a higher pH than the LD of steers from Holstein dams only in the non-stimulated LD (data not shown in table), while no dam breed effects on the ultimate pH in the aged LD were found (Table 2). No such effects could be observed in experimental data set 2. J Sci Food Agric 88:1344–1353 (2008) DOI: 10.1002/jsfa Cooking loss and meat texture Cooking loss was higher in the BF than in the LD, which can be explained by the longer cooking time required to reach the target core temperature in the BF (∼10 min compared with ∼7 min for the LD). No significant effect of chilling regimen and dam breed on cooking loss was found and no relevant interactions of the p.m. treatments with other factors were observed for meat colour traits and cooking loss. The ES-treated LD showed lower maximum shear force values compared with CO1 at 2 and 15 days p.m. (Fig. 3). The treatment differences were no longer significant after 29 days of aging, as the LD shear force did not decrease further in ES but did do so in CO1. In both ES and CO1, Volodkevich compression energy (VOL) in the cooked LD decreased with aging, until mean VOL values of 1.04 and 1.22 J (P = 0.10) respectively were reached after 29 days of aging. The DC treatment remained ineffective in decreasing the maximum shear force in the LD at any stage of aging. The mean level of maximum shear force remained above 40 N even after 29 days of aging. There were 1347 RH Razminowicz, M Kreuzer, MRL Scheeder Table 1. Effect of electrical stimulation, delayed chilling and post-mortem aging on pH, cooking loss and texture properties (least square means and standard error of least square means (SEM)) Post-mortem treatment (PT)a Aging (A; days p.m.) M. longissimus dorsi Ultimate pH Cooking loss (%) Compression energyb (J) M. biceps femoris Ultimate pH Cooking loss (%) Maximum shear forcec (N) Compression energyb (J) Post-mortem treatmentd M. longissimus dorsi Ultimate pH Cooking loss (%) Compression energyb (J) M. biceps femoris Ultimate pH Cooking loss (%) Maximum shear forcec (N) Compression energyb (J) Control (CO1) Electrically stimulated (ES) P level 2 15 2 15 SEM PT A PT × A CFe PT × CF 5.46 25.0 1.98 5.50 24.0 1.46 5.47 24.7 1.56 5.52 24.3 1.15 0.010 0.53 0.069 0.748 0.837 0.017 <0.001 0.006 <0.001 0.693 0.831 0.184 0.453 0.003 0.001 0.925 0.805 0.154 5.40 31.1 35.4 1.38 5.45 29.6 43.0 1.29 5.40 32.1 38.5 1.36 5.47 30.6 42.9 1.30 0.012 1.01 1.90 0.048 0.324 0.678 0.647 0.749 <0.001 0.148 0.002 0.118 0.477 0.976 0.394 0.702 0.541 0.342 0.139 0.044 0.228 0.535 0.531 0.744 Control (CO2) Delayed chilling (DC) 5.46 24.7 2.13 5.45 24.8 1.69 5.48 24.4 1.93 5.46 25.6 1.51 0.019 0.79 0.088 0.374 0.954 0.264 0.442 0.609 <0.001 0.953 0.726 0.182 0.080 0.069 0.005 0.462 0.933 0.403 5.38 31.4 38.8 1.54 5.44 30.9 40.2 1.39 5.38 32.8 40.9 1.52 5.45 31.1 43.2 1.48 0.025 0.91 1.92 0.052 0.796 0.133 0.175 0.046 0.007 0.185 0.310 0.063 0.752 0.508 0.807 0.286 0.064 0.007 0.038 0.001 0.764 0.087 0.097 0.030 a n = 16 per treatment. Volodkevich device. c Warner-Bratzler device. d n = 18 per treatment. e Cutting fat proportion of carcass weight. b Table 2. Interaction of dam breed and post-mortem treatment on quality of M. longissimus dorsi (average of 2 and 15 days of aging) from steers treated by two different regimes (least square means and standard error of least square means (SEM)) Dam breed (D) Post-mortem treatment (PT)a Ultimate pH Cooking loss (%) Maximum shear forceb (N) Compression energyc (J) Post-mortem treatmentd Ultimate pH Cooking loss (%) Maximum shear forceb (N) Compression energyc (J) Brown Swiss Holstein-Friesian P level Control (CO1) Electrically stimulated (ES) Control (CO1) Electrically stimulated (ES) SEM D D × PT 5.48 25.4 48.1 1.53 5.49 24.9 38.7 1.25 5.49 24.4 52.2 1.56 5.50 25.1 38.7 1.26 0.011 0.48 2.56 0.063 0.332 0.442 0.483 0.778 0.947 0.149 0.395 0.838 Control (CO2) Delayed chilling (DC) Control (CO2) Delayed chilling (DC) 5.44 25.6 65.9 1.87 5.47 25.3 64.4 1.67 5.47 23.9 57.0 1.57 5.46 25.0 59.4 1.58 0.020 0.66 3.37 0.074 0.556 0.128 0.041 0.008 0.298 0.303 0.553 0.142 a n = 8 per sub-treatment; the P level for PT is given in Table 1. Warner-Bratzler device. c Volodkevich device. d n = 9 per sub-treatment. b also no DC effects on VOL values of the LD, finally reaching 1.42 and 1.35 J in DC and CO2 respectively. In the BF, no effect of the chilling treatments on maximum shear force was observed and, in contrast to the LD, the shear force of the BF did not decrease from 2 to 15 days of aging and was even higher after aging in CO1. The VOL values decreased only marginally with aging. 1348 The texture traits were affected by the fatness of the animals (proportion of cutting fat) for both experimental data sets, particularly in the LD (e.g. P < 0.01 and P = 0.09 for maximum shear force in CO1/ES and CO2/DC respectively). Additionally, interactions of cutting fat with p.m. treatments occurred for maximum shear force in CO1/ES (P < 0.05) as well as in CO2/DC (P < 0.1) (Fig. 4). Accordingly, increasing J Sci Food Agric 88:1344–1353 (2008) DOI: 10.1002/jsfa Beef quality in grass-fed steers: effect of chilling regime and electrical stimulation m. longissimus dorsi Maximum shear force (N) n.s. 120 *** n.s. 80 80 n.s. ** n.s. 40 40 0 2d 15 d 29 d 2d 15 d Maximum shear force (N) CO1 ES CO2 DC 120 0 29 d CO1 / ES 50 40 30 20 10 0 3 6 9 CO2 / DC 50 y = -2.9878 x + 37.975 R2 = 0.24 PInteraction = 0.025 12 15 Cutting fat (%) Delta maximum shear force Delta maximum shear force Figure 3. Effect of aging time on maximum shear force (Warner-Bratzler device) of M. longissimus dorsi: CO1, control group 1; ES, electrically stimulated group; CO2, control group 2; DC, delayed chilling group. Error bars indicate standard deviations. Comparisons between treatments within the same experimental data set were done with the t test: ∗∗∗ P < 0.001; ∗∗ P < 0.01; n.s., not significant. P levels for aging, p.m. treatment × aging and cutting fat were <0.001, 0.137, <0.001 and <0.001, 0.710, 0.09 for the CO1/ES and CO2/DC p.m. treatments respectively. y = -5.0529 x + 40.818 R2 = 0.27 PInteraction = 0.089 40 30 20 10 0 -10 3 6 9 12 -20 -30 -40 Cutting fat (%) Figure 4. Interaction between amount of cutting fat and p.m. treatments on M. longissimus dorsi maximum shear force (Warner-Bratzler device): CO1, control group 1; ES, electrically stimulated group; CO2, control group 2; DC, delayed chilling group. Delta maximum shear force is the difference between the shear force measured in the steak from the CO side and that measured in the steak from the p.m. treated (ES or DC) side of the same animal. cutting fat proportions decreased the effects of the p.m. treatments ES and DC on the LD shear force relative to the control (Fig. 4). Therefore at low carcass fatness there was a certain positive effect of DC which was reversed at high fatness, together leading to a zero net effect of DC. In contrast to the LD, in the BF of CO1/ES treatments, cutting fat affected only compression energy, and no significant interactions with p.m. treatment were observed (Table 1). However, both texture traits of the BF were affected by cutting fat proportion in CO2/DC, and significant interactions with p.m. treatments were observed (Table 1). Maximum shear force and Volodkevich compression energy were lower for Holstein compared with Brown Swiss offspring only in experimental data set 2 (CO2/DC) (Table 2). No interactions occurred between dam breed and p.m. treatments. Within the LD slice, maximum shear force was lower for the core locations (Fig. 1) 1, 2 and 4 than for location 5 and particularly locations 3 and 6, J Sci Food Agric 88:1344–1353 (2008) DOI: 10.1002/jsfa producing a tenderness gradient from dorso-medial to ventro-lateral (Fig. 5). These differences were observed in the LD of all experimental treatments and at all stages of aging (data not shown). Additionally, significant interactions between ES treatment and position were observed, with the ES-stimulated LD showing less variation among positions and therefore not only a more favourable but also a more consistent texture over the cross-sections of the LD slices. DISCUSSION Effects of production factors on beef quality Extensive grass-based feeding is an inexpensive method of fattening. However, the value of beef from this production method is often discounted compared with beef from concentrate-dominated diets because of perceived or assumed differences in meat quality, an assumption which, however, is not necessarily supported by investigations of beef obtained at the 1349 RH Razminowicz, M Kreuzer, MRL Scheeder m. longissimus dorsi 120 120 CO1 ES CO2 DC o n.s. n.s. n.s. 80 *** * ** o 80 n.s. n.s. * o 40 0 40 1 2 3 4 5 6 1 2 3 4 5 6 0 Core location Figure 5. Effect of core location within M. longissimus dorsi (see Fig. 1) on maximum shear force (Warner-Bratzler device) after 15 days of aging: CO1, control group 1; ES, electrically stimulated group; CO2, control group 2; DC, delayed chilling group. Error bars indicate standard deviations. Comparisons between treatments within the same core location were done with the t test: ∗∗∗ P < 0.001; ∗∗ P < 0.01; ∗ P < 0.05; ◦ P < 0.1; n.s., not significant. retail level.16 As grass-based fattening can be accompanied by less intensive growth of the animals, they are often older at slaughter, which might be associated with less tender17 and darker meat21 and a lower proportion of carcass fat than in other, more intensive, feeding systems.10 Early animal production studies11 indicated that fatter animals usually produced meat that was more tender than that from leaner animals. This was partly confirmed by the results of the present study, where a higher cutting fat proportion was associated with a reduction in shear force in the 15 day aged LD of CO1 (r = −0.52, P = 0.04) and CO2 (r = −0.36, P = 0.14). There was no correlation between intramuscular fat content and shear force in the 15 day aged LD of the carcasses of data set 1 (either in ES or in CO1 carcasses; P > 0.25), while in data set 2 the correlation coefficients were −0.49 (P = 0.04) for CO2 and −0.47 (P = 0.048) for DC. However, intramuscular fat content and cutting fat proportion were significantly correlated in data set 2 (r = 0.59, P = 0.01), indicating collinearity. This correlation was not that pronounced in data set 1 (r = 0.34, P = 0.2). According to Koohmaraie and Geesink,2 the effect of intramuscular fat on tenderness is often overemphasised and was estimated to contribute about 5% of the variability in tenderness. In the present study the correlation of intramuscular fat content and shear force is inconsistent in data sets 1 and 2, and the proportion of cutting fat correlated with shear force only in the conventionally chilled carcass sides but not in the ES or DC ones. A favourable effect of higher carcass fatness may therefore be explained by the insulating effect of the carcass fat cover and bigger-sized carcasses, both decelerating the chilling rate. Production factors influencing tenderness of grass-fed cattle by increasing 1350 fatness include gender (steers, as used here, generally deposit more body fat than bulls), supplementary feeding in the finishing period (varied in the present study and reported to affect carcass composition and meat quality in cattle and lambs13,10,22 ) and breed. Meat quality characteristics of different cattle breeds have been investigated in various studies.20,23 Such comparisons, however, mostly refer to different sire breeds used for crossbreeding with dairy cows. The potential role of the dam breed in beef production is less well documented. In the two data sets of the present study, some dam breed effects were found. However, these were not always consistent across the two data sets. In the second data set the beef from steers born to Holstein dams was significantly more tender than the meat from steers of Brown Swiss dams (confirmed by both measures applied). Dam breed differences could result from genetic differences in enzymatic activity, fibre types and structural characteristics.24 Differences in the rate of early p.m. pH decline could affect the activity of enzymes2 but can hardly explain the lower Warner-Bratzler values found in the LD of the Holstein offspring in experimental data set 2, because differences in p.m. pH were found only in experimental data set 1, which in turn exhibited no difference in meat texture characteristics between dam breed groups. Therefore no conclusive evidence for dam breed effects can be drawn from this study. Anyway, according to Koch et al.,25 only about 30% of the variation in beef tenderness between breeds can be explained by additive gene effects, whereas 70% is explained by environmental and non-additive gene effects. J Sci Food Agric 88:1344–1353 (2008) DOI: 10.1002/jsfa Beef quality in grass-fed steers: effect of chilling regime and electrical stimulation Effects of post-mortem treatments on beef quality The first 24 h after the animal is slaughtered, complex biochemical and structural processes take place while the muscle converts to meat. During this period, rapid chilling of the carcass to a temperature of 7 ◦ C or less within 24 h is desired in order to improve shelf-life and to reduce losses, labour and costs.26 However, under rapid chilling conditions the sarcomeres might undergo cold-induced shortening and the meat may consequently suffer from toughening,27 a major factor impairing beef quality. Normally, the pH in beef declines from 7 upon slaughter to approximately 5.3–5.8 within 18–40 h.28 The interrelationship of time, temperature and pH differs between and within muscles, causing variation in the susceptibility of individual muscles to cold shortening.29 Although shortening of sarcomeres cannot be completely avoided, there are several known efficient means to reduce the extent and toughening effects of this process after slaughter.27 Electrical stimulation during the slaughter process may reduce the susceptibility of muscles to cold shortening by inducing muscle contraction, accelerating anaerobic glycolysis and, consequently, increasing the rate of pH decline, thus reducing the time until rigor mortis develops.27 The faster decline of pH recorded in the present study with ES aligns with the results of Eilers et al.30 However, ES did not accelerate early p.m. temperature decline in muscle. The temperature of the LD at 1.5 h p.m. was lower for control carcasses compared with ES carcasses in the present study and others.30 Jones and Tatum31 reported a positive linear relationship between LD pH3h and shear force among steaks obtained from commercially processed carcasses. In particular, a low shear force was found when pH3h was below 6.2. In our study, this recommended pH3h level was reached only in ES carcasses, and these carcasses actually expressed a lower maximum shear force of the LD compared with the control carcasses. However, the lower pH3h level in the BF of ES carcasses did not lead to a reduced shear force, which was on a rather low level already at 2 days p.m. A second promising technique could be to delay chilling of the meat and thus accelerate rigor development and subsequent tenderisation. Keeping carcasses out of the chilling room for a certain period of time has been proposed as a means of DC28 and was found to have a positive influence on meat quality.27 In the present study, DC significantly reduced the rate of early p.m. LD and BF temperature declines, but resulted in only a slightly lower pH than in control beef. Additionally, DC failed to generally improve texture attributes. This would suggest that the overall effect of DC on p.m. muscle metabolism was too weak to exert changes in texture. Nevertheless, a certain interaction between CO2/DC treatment and carcass fatness was observed, which is discussed below. A well-known technique to increase tenderness is aging during refrigerated storage. This is mainly J Sci Food Agric 88:1344–1353 (2008) DOI: 10.1002/jsfa due to the effect of weakening myofibrillar and cytoskeletal proteins by endogenous enzymes.2 Bruce et al.32 demonstrated that aging of LD muscles can reduce the maximum shear force to less than 40 N, a level that Shorthose et al.33 found to correspond with a consumer rating of ‘tender’ beef. In the present study, there were clear aging effects in the LD when comparing the texture of samples aged for 2 or 15 days. The decrease in maximum shear force when aged for 29 days was less pronounced, and LD samples from data set 2 did not decline below the threshold of 40 N for tender meat. Different from the LD, aging the BF for 15 days instead of 2 days did not result in a lower shear force, which was not expected and cannot be explained. It may be speculated that the myofibrillar texture compound is of low importance in this muscle, which is consistent with the lower initial shear force than in the LD. It was also reported by Hostetler et al.34 that, in contrast to other major muscles, the shear force of the BF did not correspond to the sarcomere length. Interactions among factors affecting texture traits A major objective of the present study was to investigate the effects of interactions between production factors (e.g. resulting in different degrees of carcass fatness) and p.m. treatments as well as between early p.m. treatments and aging on beef texture. Increasing fatness caused both ES and DC treatments to become less effective. However, while ES improved the tenderness of the LD on average, DC did not affect the texture of the LD in general, although it tended to improve the texture in lean and impair it in fatter carcasses. Fatness may reflect the effects of various production system factors, particularly finishing regime and breed. In another approach, French et al.14 concluded that supplementing grass with concentrate would suffice to produce tender and acceptable meat already at 2 days p.m., and that further aging eliminated the treatment effects on eating quality of beef observed at 2 days p.m. Realini et al.10 found, despite differences in carcass weight, fatness and temperature during chilling, a similar initial shear force in steaks from pasture- and concentrate-fed steers; more extensive aging of the steaks from pasture-fed animals resulted at 7 and 14 days p.m. in an even lower shear force than in the steaks from concentrate-fed animals. The interaction between p.m. treatment and aging was such that the CO1 LD tenderised slower during aging than the ES LD. ES accelerated tenderisation to a degree that, at day 2, the shear force of the LD was nearly as low as in the LD of CO1 after 15 days of aging. ES beef fell below the threshold of 40 N from 15 days of aging onwards, and prolongation of the aging period was ineffective. This finding supports the hypothesis of King et al.35 that the accelerated p.m. metabolism induced by ES would increase the myofibrillar fragmentation in the muscles, resulting 1351 RH Razminowicz, M Kreuzer, MRL Scheeder in an improved tenderness already early in the aging period. Another relationship of texture traits and p.m. processes becomes evident when looking at the positional variation in texture within the same slice of beef.36,37 Homogeneity of tenderness is important for consumer satisfaction and purchase loyalty, which might be particularly relevant for branded beef. In the present study a systematic variation in texture within LD steaks was observed. The highest shear force values occurred at the ventro-lateral locations, where the muscle touches the ribs, and the lowest values in the dorso-medial region, which is in line with the study of Kerth et al.38 Scheeder39 found the highest shear force at the ventral location and the lowest at the dorso-medial and dorso-lateral locations. At the same time the different sarcomere lengths corresponded well with the shear force data. Thus variation in sarcomere length can be one underlying cause of the variation in texture over the cross-section of LD steaks and may partly explain the elevated maximum shear force at certain locations.39 In muscle, attached to the skeleton, shortening of the muscle fibres occurring in one zone of the muscle may be compensated by stretching of the fibres elsewhere.40 It may therefore be speculated that a faster rate of temperature decline at the ventro-lateral site can promote myofibrillar shortening because of the close contact with the bones and the low tissue mass surrounding the muscle at this site. The temperature differences in the LD at 3 and 8 cm depth were of the order of 10 ◦ C at 1.5 h p.m. and on average still above 5 ◦ C at 3 h p.m. Because the temperature at which rigor develops affects the activity of µ-calpain and calpastatin,12,41 the observed variation in texture over the cross-section of the LD may also be partly due to differences in early p.m. tenderisation rates. In contrast to the LD, the texture of the BF did not respond to any of the production factors and p.m. treatments. Overall, it can be concluded that the tenderness of the M. longissimus dorsi from grass-fed steers can be improved by electrical stimulation and prolonged aging. Electrical stimulation accelerated and enhanced tenderisation. It was increasingly more effective as carcass fatness decreased, thus compensating for potentially detrimental effects of grass-only feeding, and it reduced within-slice variation in tenderness. Electrical stimulation therefore proved to be an efficient measure to control the tenderness of beef from cattle fed grass only. ACKNOWLEDGEMENTS We are grateful to the staff of the cutting plant Traitafina AG and Rolf Bickel for their assistance with sampling and to the Hermann Herzer Foundation for financial support. REFERENCES 1 Carpenter CE, Cornforth DP and Whittier D, Consumer preferences for beef colour and packaging did not affect eating satisfaction. Meat Sci 57:359–363 (2001). 1352 2 Koohmaraie M and Geesink GH, Contribution of postmortem muscle biochemistry to the delivery of consistent meat quality with particular focus on the calpain system. Meat Sci 74:34–43 (2006). 3 Tornberg E, Biophysical aspects of meat tenderness. Meat Sci 43:175–191 (1996). 4 Caballero B, Sierra V, Olivan M, Vega-Naredo I, TomasZapico C, Alvarez-Garcıa O, et al., Activity of cathepsins during beef aging related to mutations in the myostatin gene. J Sci Food Agric 87:192–199 (2007). 5 Van Eenennaam AL, Li J, Thallman RM, Quaas RL, Dikeman ME, Gill CA, et al., Validation of commercial DNA tests for quantitative beef quality traits. J Anim Sci 85:891–900 (2007). 6 Olsson U, Herzman C and Tornberg E, The influence of low temperature, type of muscle and electrical stimulation on the course of rigor, ageing and tenderness of beef muscles. Meat Sci 37:115–131 (1994). 7 Locker RH and Hagyard CJ, A cold shortening effect in beef muscles. J Sci Food Agric 14:787–793 (1963). 8 Locker RH, Degree of muscular contraction as a factor in tenderness of beef. J Food Sci 25:304–312 (1960). 9 Muir PD, Deaker JM and Brown MD, Effects of forage- and grain-based feeding systems on beef quality: a review. NZ J Agric Res 41:623–635 (1998). 10 Realini CE, Duckett SK, Brito GW, Dalla Rizza M and De Mattos D, Effect of pasture vs. concentrate feeding with or without antioxidants on carcass characteristics, fatty acid composition, and quality of Uruguayan beef. Meat Sci 66:567–577 (2004). 11 Smith GC, Duston TR, Hostetler RL and Carpenter ZL, Fatness, rate of chilling and tenderness of lamb. J Food Sci 41:748–756 (1976). 12 Hwang IH and Thompson JM, The interaction between pH and temperature decline early postmortem on the calpain system and objective tenderness in electrically stimulated beef longissimus dorsi muscle. Meat Sci 58:167–174 (2001). 13 Razminowicz RH, Kreuzer M, Leuenberger H and Scheeder MRL, Efficiency of extruded linseed for the finishing of grassfed steers to counteract a decline of omega-3 fatty acids in the beef. Livest Sci 114:150–163 (2008). 14 French P, O’Riordan EG, Monahan FJ, Caffrey PJ, Vidal M, Mooney MT, et al., Meat quality of steers finished on autumn grass, grass silage or concentrate based diets. Meat Sci 56:173–180 (2000). 15 French P, O’Riordan EG, Monahan FJ, Caffrey PJ, Mooney MT, Troy DJ, et al., The eating quality of meat of steers fed grass and/or concentrates. Meat Sci 57:379–386 (2001). 16 Razminowicz RH, Kreuzer M and Scheeder MRL, Quality of retail beef from two grass-based production systems in comparison with conventional beef. Meat Sci 73:351–361 (2006). 17 Mitchell GE, Reed AW and Rogers SE, Influence of feeding regimen on the sensory qualities and fatty acid contents of beef steaks. J Food Sci 56:1102–1106 (1991). 18 Davey CL and Chrystall BB, Conditions for an efficient postmortem electrical stimulation. Ann Technol Agr 29:547–561 (1980). 19 Specht H and Kunis J, Kälteverkürzung und Elektrostimulation. Auswirkungen auf die Beschaffenheit von Schaf- und Rindfleisch. Fleischwirtschaft 69:1275–1280 (1989). 20 Monson F, Sanudo C and Sierra I, Influence of cattle breed and ageing time on textural meat quality. Meat Sci 68:595–602 (2004). 21 Priolo A, Micol D and Agabriel J, Effects of grass feeding systems on ruminant meat colour and flavour. A review. Anim Res 50:185–200 (2001). 22 Priolo A, Micol D, Agabriel J, Prache S and Dransfield E, Effect of grass or concentrate feeding systems on lamb carcass and meat quality. Meat Sci 62:179–185 (2002). 23 Chambaz A, Kreuzer M, Scheeder MRL and Dufey PA, Characteristics of steers of six beef breeds fattened from eight J Sci Food Agric 88:1344–1353 (2008) DOI: 10.1002/jsfa Beef quality in grass-fed steers: effect of chilling regime and electrical stimulation 24 25 26 27 28 29 30 31 32 months of age and slaughtered at a target level of intramuscular fat. II. Meat quality. Arch Anim Breed 44:473–488 (2001). Wulf DM, Tatum JD, Green RD, Morgan JB, Golden BL and Smith GC, Genetic influences on beef longissimus palatability in Charolais- and Limousin-sired steers and heifers. J Anim Sci 74:2394–2405 (1996). Koch RM, Cundiff LV and Gregory KE, Heritabilities and genetic, environmental, and phenotypic correlations of carcass traits in a population of diverse biological types and their implication in selection programs. J Anim Sci 55:1319–1325 (1982). Bowater FJ, Rapid carcass chilling plants compared to conventional systems. [Online]. International Institute of Refrigeration (2001). Available: http//www.fjb.co.uk [3 July 2007]. Savell JW, Mueller SL and Baird BE, The chilling of carcasses. Meat Sci 70:449–459 (2005). Smulders FJM, Toldrá F, Flores J and Prieto M (eds), New Technologies for Meat and Meat Products. ECCEAMST/Audet Tijdschriften, Utrecht, pp. 182, 186–188 (1992). Hannula T and Puolanne E, The effect of cooling rate on beef tenderness: the significance of pH at 7 ◦ C. Meat Sci 67:403–408 (2004). Eilers JD, Tatum JD, Morgan JB and Smith GC, Modification of early-postmortem muscle pH and use of postmortem aging to improve beef tenderness. J Anim Sci 74:790–798 (1996). Jones BK and Tatum JD, Predictors of beef tenderness among carcasses produced under commercial condition. J Anim Sci 72:1492–1501 (1994). Bruce HL, Stark JL and Beilken SL, The effects of finishing diet and postmortem ageing on the eating quality of the M. longissimus thoracis of electrically stimulated Brahman steer carcasses. Meat Sci 67:261–268 (2004). J Sci Food Agric 88:1344–1353 (2008) DOI: 10.1002/jsfa 33 Shorthose WR, Harris PV, Hopkins AF and Kingston OL, Relationships between some objective properties of beef and consumer perceptions of meat quality. Proc Int Congr Meat Sci Technol 34:667–669 (1988). 34 Hostetler RL, Link BA, Landmann WA and Fitzhugh Jr HA, Effect of carcass suspension on sarcomere length and shear force of some major bovine muscles. J Food Sci 37:132–135 (1972). 35 King DA, Voges KL, Hale DS, Waldron DF, Taylor CA and Savell JW, High voltage electrical stimulation enhances muscle tenderness, increases aging response, and improves muscle color from cabrito carcasses. Meat Sci 68:529–535 (2004). 36 Crouse JD, Theer LK and Seideman SC, The measurement of shear force by core location in longissimus dorsi beef steaks from four tenderness groups. J Food Qual 11:341–347 (1989). 37 Berry BW, Tenderness of beef loin steaks as influenced by marbling level, removal of subcutaneous fat, and cooking method. J Anim Sci 71:2412–2419 (1993). 38 Kerth CR, Montgomery JL, Lansdell JL, Ramsey CB and Miller MF, Shear gradient in longissimus steaks. J Anim Sci 80:2390–2395 (2002). 39 Scheeder MRL, Age-related changes in meat quality of growing cattle, in Proceedings of the Symposium on Growth in Ruminants: Basic Aspects, Theory and Practice for the Future, ed. by Blum JW, Elsasser T and Guilloteau P. University of Berne, Berne, pp. 265–275 (1998). 40 Marsh BB and Leet NG, Studies in meat tenderness. The effect of cold shortening on tenderness. J Food Sci 31:450–459 (1966). 41 Dransfield E, Optimisation of tenderization, ageing and tenderness. Meat Sci 36:105–121 (1994). 1353 J Sci Food Agric 88:1354–1362 (2008) Journal of the Science of Food and Agriculture Predictability of price of tea from sensory assessments and biochemical information using data-mining techniques Sanjoy K Paul∗ DTU, OCDEM, University of Oxford, Oxford OX3 7LJ, UK Abstract BACKGROUND: The valuation of tea depends on the sensory assessments made by the Brokers and Buyers (Tea Tasters) to a large extent, though the market conditions and the requirements of a particular Buyer play an important role in determining the basic prices of teas. Again, there are several biochemical quality parameters in tea on which the quality of a particular tea depends. It is not straightforward to establish the reflection of biochemical quality characteristics in tea on the Taster’s sensory assessments and price because of the complex dynamics within chemical properties and the inherent subjectivity of quality evaluation through the Taster’s scores. It is, however, important to judge the market valuation of teas from quality assessments and biochemical properties. This paper describes the advantages of using statistical data-mining techniques to explore the association of biochemical quality parameters in teas with the Taster’s sensory assessments, and the application of a nonparametric statistical technique, multivariate adaptive regression splines (MARSplines), to establish the predictability of the realised prices of teas from sensory assessments. RESULTS: The price of tea is significantly associated with various quality attributes and some of the biochemical parameters. The MARSplines technique successfully demonstrated the predictability of price through Tasters’ sensory assessments and also raised the issue of inherent subjectivity of the Tasters’ assessments. CONCLUSION: It is important to explore appropriate statistical techniques to assess the subjectivity in Tasters’ assessments, and a better-designed study needs to be conducted to understand the complex biochemical reflections on the price of tea.  2008 Society of Chemical Industry Keywords: biochemical quality parameters; sensory assessments; tea price; data mining; regression splines INTRODUCTION It is well acknowledged that the quality of tea is crucially dependent on some inherent chemical characteristics. In practice, many biochemical properties of tea can be measured fairly satisfactorily.1,2 However, the quality of black tea cannot be judged solely on the basis of chemical information. Here the professional Tea Tasters play a vital role in grading various types of tea in terms of different quality attributes, e.g. colour, brightness, strength, taste (involving non-volatile compounds) and aroma (involving volatile compounds). The most desirable biochemical quality parameters in black tea are theaflavins (TFs) and thearubigins (TRs). They significantly influence the Taster’s perception and also play important roles in the ultimate valuation of tea at the auction centres.2 – 20 The Tea Broker Houses have their own Tasters who evaluate the samples in terms of overall quality. The final price is significantly influenced by their quality evaluation. Thus we can clearly hypothesise a direct relationship between the Taster’s assessment and the price of tea. It is generally conceded that the evaluation of quality by Tea Tasters is not completely unbiased. Market conditions in general and the requirements of the Broker or Buyer whose needs the Tasters serve could significantly affect the evaluation of tea quality. The producers and Brokers will naturally try to obtain the best possible evaluations on their teas. However, the method of assessment should be objective as far as possible. As the quality attributes assessed by the Taster are results of the influence of inherent biochemical quality parameters, we can also postulate a relationship between the chemical information and the Taster’s judgement. There have been a number of published efforts to correlate in a quantitative manner the chemistry of tea with the Taster’s descriptions and cash valuations.3 – 24 Attempts were made by researchers to explain the quality and various liquor characteristics of manufactured teas in terms of the chemical composition and biochemical behaviour of both the unprocessed tea shoots and the manufactured teas. Some earlier ∗ Correspondence to: Sanjoy K Paul, DTU, OCDEM, University of Oxford, Oxford OX3 7LJ, UK E-mail: sambhupaul@hotmail.com (Received 17 April 2007; revised version received 22 December 2007; accepted 3 January 2008) Published online 9 April 2008; DOI: 10.1002/jsfa.3223  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Predictability of price of tea using date-mining techniques studies in this field are due to Roberts and Smith,9 Wood and Roberts,22 Deb and Ullah6 and Biswas and Biswas,3 among many others. Roberts and Smith9 found that TFs and TRs were largely responsible for colour and strength and that TFs were factors in quality and briskness. They also found that highest cash values were given to teas with high TF levels, as long as the TR content was also at a satisfactorily high level. Wood and Roberts22 observed that the Taster’s scores for colour and strength were related to the TF and TR contents of the manufactured teas. They also observed that scores for briskness and quality depend to some extent on TFs, with caffeine contributing to briskness. According to their observations, cash valuation would be more closely related to TFs than to TRs. Wickremasinghe and Swain23 discussed the relationship between the quantities of phenolic compounds and commercial valuation and the contributions of volatile compounds to the flavour of Ceylon tea. They observed that the quality of black tea might be predicted from an estimation of the polyphenol content before processing the tea shoots, because the amount of polyphenols in black tea depends on the amount originally present in the unprocessed tea shoots. All these studies are based on total correlation between the individual biochemical constituents and the Taster’s scores on the individual liquor characteristics or on the cash valuation of the manufactured teas. A well-known study on the statistical association of liquor characteristics with the cash valuation of NE Indian black tea is due to Biswas and Biswas.3 They used a multiple regression technique to determine whether the term ‘quality’ of NE Indian plain black tea has its own single characteristic as recognisable by Tasters or whether it is the integration of some of the other important liquor characteristics. The influence of different quality characteristics on the cash valuation of tea was explored. According to their observations, the quality of NE Indian plain black tea depended mainly on briskness, with quality being increased by an increase in briskness. Cash valuations of Crush, Tear and Curl (CTC) as well as orthodox teas, in general, depended mainly on quality and/or briskness. They related the biochemical quality parameters with individual Tasters’ choices and studied the significance of different biochemical parameters. One relatively recent study aimed at assessing the correlation of chemical quality parameters with quality attributes and the market price of black teas is due to Wright et al.24 Another study on the estimation of black tea quality is due to Liang et al.21 The latter study, based on Chinese tea, finds significant correlations among phenolic compounds, tea pigments, nitrogen-containing compounds and sensory evaluation. We note here that these studies do not address the inherent subjectivity of Tasters’ choices. There have been very few studies on the statistical association of sensory scores and biochemical information, specific to tea quality assessment, after J Sci Food Agric 88:1354–1362 (2008) DOI: 10.1002/jsfa eliminating the bias associated with sensory scores. Aspects of inherent subjectivity in Tasters’ sensory assessments and methodologies to eliminate the bias have been addressed by Pal and co-workers.25 – 27 These statistical approaches reduce the problem of uncertainty in sensory evaluation, making the derived quality scores more objective. All previous studies relating biochemical aspects in tea with sensory assessments and valuation of teas have basically drawn inferences based on simple correlation and linear regression analyses, without recognising the inherent complex dynamics and likely interactions among biochemical quality parameters and the nonlinear systems of relationships between biochemistry and Tasters’ perceptions. The main objectives of the present paper are twofold. First we explore the patterns of association between biochemical properties in teas, their quality attributes as observed by Tea Tasters and the realised price using advanced data-mining techniques. The second objective is to examine the relationship between price and quality of tea, with the exploration of possible interactions of various quality aspects, using the multivariate adaptive regression splines (MARSplines) technique. To the best of our knowledge, such a rigorous statistical analysis has not been attempted before to address the complex issues of predicting valuation of teas from biochemical properties and sensory attributes. THE DATA The data for this study are based on the published work of Wright et al.,24 which investigates the predictability of quality and price of black teas produced in Central and Southern Africa from biochemical parameters, especially the TF contents. Forty African tea clones (samples) from Malawi were processed and biochemical measurements were obtained on TF-f, TF-A, TF-B, TF-dg and sum of individual TFs (SIT), flavognost (FLAV), caffeine (CAF), total polyphenols (TP), crude fibre (CF), epicatechin (EC), epigallocatechin (EGC), epicatechin-3-gallate (EGCg), epicatechin-3-gallate 1 (ECg1), gallated catechins (GALEC), non-gallated catechins (NGALEC), gallocatechins (GALOC), non-gallocatechins (NGALOC) and sum of individual flavognost (SIF). Two professional Tasters evaluated the tea samples on various quality attributes, but on two different scales. Taster A evaluated colour of liquor (COL-A), strength of liquor (SOL-A), colour of infusion(COI-A), colour with milk (CWM-A), briskness (BRSK-A) and brightness (BRIG-A). Taster B evaluated the samples on a 20-point scale in terms of colour of liquor (COL-B), strength (STR-B), brightness (BRIG-B), briskness (BRSK-B), quality (QAL-B) and valuation (VAL-B). The prices of the tea samples are presented in USc/ kg−1 . These 40 tea clones are differentiated by ‘good (high) quality’ and ‘poor 1355 SK Paul (low) quality’ in terms of biochemical properties and Tasters’ perception. The basic statistics on sensory attributes by Taster A and realised prices of tea samples are presented in Table 1, differentiated by the quality status of the tea clones. The basic statistics on all biochemical parameters are given in Table 2. As evident from Table 1, the within-sample variability in all quality attributes is very low as assessed by Taster A. However, the dispersions in the measures of most of the chemical Table 1. Basic statistics on price and sensory attributes by quality status Quality attribute/price Colour of liquor Colour of infusion Colour with milk Brightness Briskness Strength of liquor Price (USc / kg−1 ) Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Minimum Maximum Mean SD 3.90 4.30 1.70 2.60 2.50 3.20 1.00 1.20 1.00 2.00 3.50 4.10 88.60 109.40 4.90 5.90 3.50 5.70 4.30 5.60 2.10 3.20 2.00 3.50 4.60 6.20 142.00 153.00 4.39 4.75 2.67 4.18 3.43 4.47 1.49 2.34 1.52 2.56 3.88 4.70 112.35 136.44 0.26 0.35 0.59 0.69 0.49 0.60 0.35 0.50 0.29 0.39 0.28 0.49 12.42 13.74 Table 2. Basic statistics on biochemical parameters by quality status Biochemical parameter TF-f TF-A TF-B TF-dg TP CF CAF EC EGC ECg1 EGCg FLAV GALEC GALOC NGALEC NGALOC SIF SIT 1356 Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Low quality High quality Minimum Maximum Mean SD 2.96 4.51 2.22 3.11 0.70 1.32 0.85 0.86 104.03 131.27 2.38 3.91 130.52 145.25 16.32 24.82 28.22 61.97 43.32 54.73 165.90 174.94 4.82 6.56 238.44 233.03 204.78 305.39 49.44 93.68 64.09 90.76 306.18 413.78 7.29 12.33 11.64 16.72 4.96 6.28 2.49 2.89 4.34 3.68 290.09 276.08 12.32 13.39 243.10 267.10 55.75 114.67 139.88 250.50 82.09 131.23 294.18 329.29 12.32 18.44 368.52 428.67 412.00 461.37 198.76 351.30 134.82 255.97 532.85 717.34 19.33 26.62 6.99 11.21 3.58 4.72 1.53 2.08 1.96 2.06 178.76 199.85 6.61 8.73 187.42 190.97 34.68 61.05 97.39 141.88 63.13 80.43 237.69 235.95 8.55 12.88 300.82 316.38 335.08 377.83 138.69 211.66 104.42 150.22 439.50 528.04 14.07 20.07 2.39 3.58 0.80 0.90 0.48 0.49 0.82 0.75 41.93 40.00 2.44 2.57 30.43 26.69 11.81 23.68 36.53 48.92 10.36 21.91 30.73 32.54 2.35 3.77 34.97 47.95 50.18 45.85 47.74 70.79 17.62 39.82 60.91 76.45 3.46 4.57 J Sci Food Agric 88:1354–1362 (2008) DOI: 10.1002/jsfa Predictability of price of tea using date-mining techniques parameters are very high (Table 2). We have not compared the significance of differences in the average levels of the biochemical measurements and quality attributes, as this is not the aim of our study and we are only interested in assessing the patterns of association and predictability irrespective of the quality status of the tea clones. STATISTICAL METHODS AND ANALYSIS The data-mining approach (feature selection) The patterns of association of various biochemical parameters with different sensory characteristics cannot be satisfactorily explored using estimates of correlation coefficients and linear regression techniques. The main reasons for this are the nonlinear relationships between individual chemical parameters and quality scores and the unknown levels of interactions among various chemical parameters which influence a particular quality attribute in tea. For example, there are various levels of TF measures, and appropriate methodology needs to be employed to determine the specific TF measures affecting the particular quality attribute without constraining the statistical exploration by prior assumptions (e.g. normality of the data, non-existence of outliers, etc.) about the relationships. Data mining is an analytical approach designed to explore data in search of consistent patterns and/or systematic relationships among variables and then to validate the findings by applying the detected patterns to subsets of data. After recognising the pattern(s) of relationship(s) among variables or systems, the goal of data mining is prediction, and predictive data mining is the most common type of data mining. The analytical methods used in data mining are often well-known mathematical and statistical algorithms and techniques. The process of data mining generally consists of three stages: (1) the initial exploration, (2) model building or pattern identification and (3) deployment (i.e. application of the model to new data for prediction and validation). Detailed explanations of data-mining techniques with applications in the field of biochemistry can be found in the books by Edelstein28 and Hastie et al.29 Association of biochemical parameters with sensory scores and price We first adopt the feature selection and variablescreening approach to explore the statistical associations of TF-f, TF-A, TF-B and TF-dg with all the individual quality attributes assessed by both Tasters and price. We do not consider SIT here, as this is the sum of the measured TFs and therefore not independent of the other TF measures. This initial step identifies the order of the strength of association of chemical parameters with quality attributes using estimated probability values (P values) based on the F test. This feature selection procedure is essentially based on the robust analysis of variance (ANOVA) J Sci Food Agric 88:1354–1362 (2008) DOI: 10.1002/jsfa technique which uses MM estimates and is suitable under the existence of outliers and non-normality. This approach is often more powerful compared with general linear regression-type analysis and leads to a better model choice in terms of inclusion or exclusion of possible interaction effects. The individual significance levels (P values) for all TF measures in association with quality attributes scored by both Tasters are presented in Table 3. The significance levels of associations of other chemical quality parameters with sensory scores and price are presented in Table 4. As evident from the results, the digallated TFs are not related to any of the quality attributes assessed by the two Tasters. The monogallated TFs are very highly significantly associated with strength, colour of liquor and infusion, briskness and brightness. It is interesting to observe that, although the colour of liquor assessed by Taster A (COL-A) is only marginally significantly associated with TF-B and insignificantly associated with TF-A, the scenario is completely different for the patterns of association for the colour of liquor as assessed by Taster B. This raises the issue of inconsistency in sensory assessments by Tasters, though is not conclusive based on this small set of tea samples. The biochemical parameters TF-f, TF-A and TF-B, among others, were significantly associated with price. The patterns of relationships of the quality attributes assessed by Taster A with the realised price are presented using scatter plots with Kernel smoothers in Fig. 1. As evident from the scatter plots, the relationships between sensory scores and price cannot be claimed to be linear. We have also assessed the significance of association of all the quality attributes with the realised price (Table 3) using the same feature selection procedure, and the associations were very highly statistically significant, except for the colour of liquor assessed by Taster A. This definitely establishes our primary hypothesis of the reflection of quality in price through sensory assessment. This approach of assessing the significance of association of biochemical information with sensory assessments and valuation using the nonparametric data-mining technique is more reliable and more Table 3. Significance (P values) of theaflavins and price in association with quality attributes Attribute TF-f TF-A TF-B TF-dg Price COL-A SOL-A CWM-A BRSK-A BRIG-A COI-A COL-B BRIG-B STR-B BRSK-B QAL-B VAL-B 0.01 0.07 <0.001 0.001 0.001 0.001 0.07 0.04 0.08 0.08 0.04 0.5 0.23 0.01 <0.01 0.001 0.001 0.001 0.001 0.001 0.01 0.04 0.02 0.02 0.04 0.00 <0.001 0.001 0.001 0.001 0.13 0.01 0.05 0.02 0.05 0.02 0.22 0.14 0.83 0.21 0.58 0.90 0.50 0.35 0.31 0.55 0.55 0.58 0.25 0.02 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 1357 SK Paul Table 4. Significance (P values) of biochemical parameters in association with quality attributes (empty cells indicate P values greater than 0.05) Attribute COL-A SOL-A CWM-A BRSK-A BRIG-A COI-A COL-B BRIG-B STR-B BRSK-B QAL-B VAL-B Price TP CF CAF 0.05 EC EGC 0.01 <0.001 0.01 0.002 0.003 <0.001 0.01 <0.001 <0.001 0.006 0.01 0.02 0.03 0.03 0.02 0.02 0.01 0.03 0.008 0.03 0.006 0.004 0.05 ECg1 EGCg GALEC GALOC NGALEC 0.001 0.003 0.02 0.01 <0.001 0.04 0.03 0.04 0.04 0.04 0.04 0.01 NGALOC SIF SIT FLAV 0.002 0.05 0.005 0.004 0.007 0.02 0.03 <0.001 0.004 0.03 0.001 <0.001 <0.001 0.008 0.003 <0.001 <0.001 0.04 0.005 0.002 0.03 <0.001 0.04 <0.001 0.003 <0.001 <0.001 0.04 <0.001 0.002 0.001 <0.001 0.002 <0.001 <0.001 0.005 0.007 0.02 0.02 0.008 0.01 0.007 TF-f, TF-A and TF-B were significantly associated with price (P < 0.001). 140 Price Price 140 120 100 100 80 80 1.0 1.5 2.0 2.5 Briskness 3.0 3.5 1.0 120 100 2.0 2.5 Brightness 3.0 120 100 80 4.0 4.5 5.0 5.5 6.0 1 Colour of Liquor 140 2 3 4 5 Colour of Infusion 6 140 Price Price 1.5 140 Price Price 140 80 3.5 120 120 100 120 100 80 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 Colour With Milk 80 3.5 4.0 4.5 5.0 5.5 6.0 Strength of Liquor Figure 1. Scatter plots with Kernel smoothers of price against quality attributes scored by Taster A. authentic, as it suggests an exact significance level (P value) and we do not have to base our inferences on estimated correlation coefficients, which only suggest the degree of linear relationships and are highly sensitive to outliers or extreme values. To demonstrate the effects of outliers and the possible existence of nonlinear relationships in ‘correlation coefficient’-based analysis, we have estimated both 1358 robust and classical correlation coefficients between different types of TFs and the overall quality scores given by the two tasters (TA and TB). The mean (SD) values of TA and TB are 20.32 (3.59) and 57.73 (11.33) respectively. Details on the robust estimation of correlation coefficients can be found in the book by Wilcox30 and the Research and Statistical Support section of the University of North Texas J Sci Food Agric 88:1354–1362 (2008) DOI: 10.1002/jsfa Predictability of price of tea using date-mining techniques (www.unt.edu) and may be easily implemented in freely available R software. Both robust and classical estimates of correlation coefficients along with the Mahalanobis distance plot are presented in Fig. 2. Clearly there are differences in the classical and robust estimates, the reason being the existence of outliers explored in the distance plot. For example, the robust and classical correlations between TF-A and TA are 0.86 and 0.54 respectively, a difference of 0.32. From the distances based on classical correlation estimates, we do not see any outliers in the data set. In contrast, the distances based on robust correlation estimates reveal quite a few outliers in the data set that are otherwise hard to find. Figure 2 compares the robust and classical correlation matrix estimates by interpreting the correlations in the upper triangle of each matrix as ellipses. The ellipses are drawn such that the ijth ellipse is the contour of a bivariate normal distribution with correlation rij . The lower triangle contains the numerical correlations. The overlaid ellipses are particularly useful for spotting where the robust and classical correlation estimates differ. Figure 2. Comparison of robust and classical estimates of correlation coefficients and detection of outliers through robust Mahalanobis distance approach. J Sci Food Agric 88:1354–1362 (2008) DOI: 10.1002/jsfa Predicting price from quality attributes (the regression spline approach) Here we discuss the aspects of developing a functional relationship between quality attributes and the realised price of tea and the possibility of predicting price from the sensory assessments. This is more formally known as the ‘hedonic price function’ in the economics literature. Hedonic analysis is a well-established method for the study of market-based product differentiations. The hedonic approach implies that the final price of a product will be functionally related to the product’s characteristics. The general standard used to select characteristics for inclusion in the hedonic price equation is due to Griliches.31 This suggests the inclusion of only those variables that are ‘direct characteristics’, i.e. quality characteristics ‘inherent’ in the commodity. Most of the studies in this field are based on parametric specification of the hedonic price function, which potentially leads to consequences of misspecification. More flexible regression approaches attempted involve nonlinear transformation models such as Box–Cox transformation.32 The nonparametric additive regression model in a more general set-up for the hedonic price function is due to Martins-Filho and Bin,33 whose study addresses the very important issues related to specification of regression class, choice of the smoother underlying the estimation method and choice of the smoothing or bandwidth parameter. This type of methodology is especially warranted in situations like ours, where the relationships are nonlinear and the levels of interaction effects of the sensory quality attributes on the valuation of tea are not known. We adopt the well-established multivariate adaptive regression splines (MARSplines) approach to associate the different quality attributes in tea assessed by the same Taster with price. MARSplines is a non-parametric procedure that makes no assumptions about the underlying functional relationship between the quality attributes and price. This regression approach constructs the relation from a set of coefficients and basis functions that are entirely datadriven. This makes MARSplines particularly suitable for developing the hedonic price function in tea, where the relationships between individual quality attributes and price are clearly nonlinear and not very straightforward to model (Fig. 1). The MARSplines technique is particularly popular in the field of data mining, as it does not assume or impose any particular type or class of relationship (e.g. linear, logistic, etc.) between the predictor variables (quality attributes) and the dependent variable (price) of interest. Instead, models that yield accurate predictions can be derived even in situations where the relationship between the predictors and the dependent variables is non-monotonic and difficult to approximate with parametric models. The special advantage with this nonparametric regression method is that it automatically searches for the statistically significant possible interaction effects among the quality attributes, apart from the individual 1359 SK Paul quality attributes, to include in the model to predict price. Technical details and application aspects of this regression technique can be found in the paper by Friedman34 and the book by Hastie et al.29 A detailed simplistic description of MARSplines with the interpretation of results from the regression fit may be found in the freely available online textbook provided by StatSoft (www.statsoft.com). There are always issues related to over-fitting a statistical predictive model. In general, nonparametric models are adaptive and can exhibit a high degree of flexibility that may ultimately result in over-fitting if adequate measures are not taken to counteract it. It has been observed that such models can achieve zero prediction error on training data under certain formulations; however, they have the tendency to perform poorly when presented with new observations or instances. This means that on many occasions these predictive regression models do not generalise well to the prediction of ‘new’ samples. MARSplines, like most methods of this kind, tends to over-fit the data as well. To address this problem, we can use a pruning technique to limit the complexity of the model by reducing the number of its basis functions. The features of the selection of pruning of basis functions make this technique a very powerful tool for predictor selection. The MARSplines algorithm can pick up only those basis functions and relevant predictor variables that make a significant contribution to the prediction. Although the two Tasters scored the samples on different scales and different quality attributes, we have fitted regression models for both Tasters to compare the predictive ability. This will also address the inherent subjectivity in Tasters’ assessments to some extent. Taster A assessed the samples in terms of colour of liquor, strength of liquor, colour with milk, briskness, brightness and colour of infusion, while Taster B considered the attributes colour of liquor, brightness, strength, briskness, quality and value. Details of the model-fitting criteria along with the regression results are presented in Table 5. We set 25 basis functions with two levels of interactions between the quality attributes. The choice of basis functions is basically judgemental and some guidelines on choosing the ideal basis functions are due to Hastie et al.29 We have considered only the secondorder interactions between sensory attributes for two reasons: simplicity and the relatively small size of tea samples available for analysis. For Taster A the final fitted model retained five regressors, with strength of liquor, briskness and brightness appearing independently along with the interaction effects of briskness*brightness and strength*briskness. The number of times specific regressors were referred to the basis function is indicated in Table 5, with the highest number of references occurring for briskness. The influence of these three quality attributes on the valuation of the particular African tea types is very much in line with the belief of the Tea Tasters. The additional new information we have now is the specific significant interaction effects among these quality attributes. In the results we have presented the generalised crossvalidation (GCV) error and the root mean square error of cross-validation (RMSECV). The RMSECV measures the goodness of fit, which takes into account not only the residual error but also the model complexity. This error estimate also helps guide the Table 5. Results from MARSpline fit Information Criterion Initial conditions Number of basis functions Order of interaction Pruning Predictors Output Number of terms retained Number of basis functions Terms retained Interaction terms GCV error Number of times particular predictor referenced to basis function Mean (SD) of observed price (USc / kg−1 ) Mean (SD) of predicted price (USc / kg−1 ) Mean (SD) of residual Adjusted R2 RMSECV Spline model for Taster A Taster A Taster B 25 2 Yes COL, SOL, CWM, BRSK, BRIG, COI 25 2 Yes COL, BRIG, BRSK, QAL, VAL, STR 5 12 SOL, BRSK, BRIG BRSK∗ BRIG, SOL∗ BRSK 80.54 SOL(3), BRSK(5), BRIG(4) 3 3 BRSK, BRIG BRSK∗ BRIG 180.38 BRIG(2), BRSK(1) 125.03 (17.79) 125.01 (17.06) −0.06 (5.03) 0.89 1.1874 125.03 (17.79) 124.28 (13.29) 0.09 (11.82) 0.52 2.9948 Price = 160.11 − 51.1∗ max(0, 2.1 − BRSK) − 34.63∗ max(0, SOL − 3.50) − 472.83∗ max(0, 2.10 − BRSK)∗ max(0, BRIG − 1.80) + 39.38∗ max(0, SOL − 3.50)∗ max(0, BRSK − 2.30) + 38.23∗ max(0, SOL − 3.50)∗ max(0, 2.30 − BRSK) + 58.98∗ max(0, BRIG − 2.20) − 17.18∗ max(0, 2.20 − BRIG) − 74.28∗ max(0, BRSK − 2.10)∗ max(0, BRIG − 2.20) 1360 J Sci Food Agric 88:1354–1362 (2008) DOI: 10.1002/jsfa Predictability of price of tea using date-mining techniques choice of basis functions. In the MARSpline method, after implementing the forward stepwise selection of basis functions, a backward procedure is applied in which the model is pruned by removing those basis functions that are associated with the smallest increase in the (least squares) goodness of fit. The GCV error is a type of least squares error function (inverse of goodness of fit). The performance of the predictive model is quite impressive for Taster A, with very small average residual and relatively small standard error. The fitted model explains 89% of the variation in price levels based on the Taster’s assessment (quality attributes), with a relatively small RMSECV estimate of 1.19. However, the nature of the fit for Taster B was quite different, with a poor quality of fit. The scatter plot of the predicted price of tea samples versus the observed price along with the density plot of the residual from the fitted model for Taster A (Fig. 3) clearly illustrates the successful prediction. CONCLUSION We have considered in this paper the potential application of statistical data-mining techniques to understand the complex association of biochemical aspects in tea with human perception of its quality and the realised price. This statistical approach is clearly advantageous in exploring the association of various chemical quality parameters and their possible interactions with sensory assessments, which is otherwise not possible through simple correlation and linear regression-based analyses. Also, the interesting issue, from a business point of view, of the predictability of price through inherent quality in tea has been addressed objectively using a nonparametric adaptive regression technique. As far as we are aware, this is the first methodological exercise to analyse the hedonic price function for tea, which would allow the Brokers to judge the worth of sensory analysis in price realisation. Alhough this hedonic approach has the advantage of being based on the ‘realised’ price, it still depends on subjective sensory assessments. However, in the beverage industry, quality assessments are based on sensory analysis only. It is important to address the subjectivity of sensory assessment methodologically so that a biascorrected derived sensory score can be associated with biochemical aspects and price. A well-designed study on a relatively large tea sample with exhaustive assessment of biochemical parameters along with sensory assessments by several Tasters would help Figure 3. Observed versus predicted price and residual density from MARSpline fits. J Sci Food Agric 88:1354–1362 (2008) DOI: 10.1002/jsfa 1361 SK Paul address predictability issues supported by appropriate statistical methodological development. 17 ACKNOWLEDGEMENT I would like to thank Dr Zeno Apostolides of the University of Pretoria (South Africa) for providing data for this study. 18 19 20 REFERENCES 1 Gilchrist RCJH and Wight W, The concept of kind of tea. Nature 191:14–60 (1961). 2 Baruah DN, Tea quality. Two and a Bud 39:2–6 (1992). 3 Biswas AK and Biswas AK, Biological and chemical factors affecting the valuations of north-east Indian plains teas: II. Statistical evaluation of the biochemical constituents and their effects on briskness, quality and cash valuations of black teas. J Sci Food Agric 22:196–204 (1971). 4 Biswas AK and Biswas AK, Statistical evaluation of biochemical constituents and their effects on colour, brightness and strength. J Sci Food Agric 24:1457–1477 (1972). 5 Davies AG, Theaflavins – objective quality indicators. Tea Coffee Trade J 155–157 (1983). 6 Deb SB and Ullah MR, The role of theaflavins and thearubigins in the evaluation of black tea. Two and a Bud 15:101–102 (1968). 7 Ellis RT and Cloughley JB, The importance of theaflavins in black tea liquors. Int Tea J 2:7–8 (1981). 8 Roberts EAH, The phenolic substances of manufactured tea. II. Their origin as enzymic oxidation products in fermentation. J. Sci. Food Agric 9:212–216 (1950). 9 Roberts EAH and Smith RF, Spectrophotometric measurements of theaflavins and thearubigins in black tea liquors in assessment of quality of teas. Analyst 86:94–98 (1961). 10 Roberts GR and Fernando RSS, Some observations on the correlation of polyphenol content to the quality of tea clones. Tea Q 50:30–31 (1981). 11 Hilton PJ and Ellis RT, Estimation of market value of Central African tea by theaflavins analysis. J. Sci. Food Agric 23:227–232 (1972). 12 Hilton PJ and Palmer-Jones RW, Relation between flavanol composition of tea shoots and theaflavin content of manufactured tea. J. Sci. Food Agric 24:813–818 (1973). 13 Hilton PJ and Palmer-Jones RW, Chemical assessment of quality in tea and its relation to market over extended period. J Sci Food Agric 26:681–684 (1975). 14 Thanaraj SNS and Seshadri R, Influence of polyphenol oxidase activity and polyphenol content of tea shoot on quality of black tea. J Sci Food Agric 51:57–70 (1990). 15 Owuor PO and Obanda M, Clonal variation in the individual theaflavins and their impact on astringency and sensory evaluation. Food Chem 54:273–277 (1995). 16 Owuor PO, Obanda M, Nyirenda HE, Mphangwe NIK, Wright WP and Apostolides Z, The relationship between some 1362 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 chemical parameters and sensory evaluations for plain black tea (Camellia sinensis) produced in Kenya and comparison with similar teas from Malawi and South Africa. Food Chem 97(4):644–653 (2006). Owuor PO, Theaflavins content as a basis for pricing Kenyan black tea. Tea 4:11–16 (1983). Owuor PO, Othieno CO and Reeves SG, The economic impact of theaflavins in Kenyan teas. Tea 7:88–93 (1986). Owuor PO, Reeves SG and Wanyoko JK, Correlation of theaflavins content and valuations of Kenyan black teas. J Sci Food Agric 37:507–513 (1986). Owuor PO, Reeves SG and Wanyoko JK, Estimation of the market value of central African tea by theaflavin analysis. J Sci Food Agric 23:227–232 (1972). Liang Y, Lu J, Zhang L, Wu S and Wu Y, Estimation of black tea quality by analysis of chemical composition and color difference of tea infusions. Food Chem 80:283–290 (2003). Wood DJ and Roberts WA, The chemical basis of quality in tea. III. Correlation of analysis results with tea tasters’ reports and valuations. J Sci Food Agric 15:19–25 (1964). Wickremasinghe RL and Swain T, Studies of the quality and flavour of Ceylon tea. J Sci Food Agric 16:57–64 (1965). Wright LP, Mphangwe NIK, Nyirenda HE and Apostilides Z, Analysis of theaflavin composition in black tea (Camellia sinensis) for predicting the quality of tea produced in Central and Southern Africa. J Sci Food Agric 82:517–525 (2002). Pal M and Paul SK, Small sample estimation problem with Box–Cox transformation: application to tea quality data. Statist Appl 12:378–393 (2000). Pal M, Paul SK and Das AK, A statistical approach to study the subjectivity of tea tasters’ scores. Two and a Bud 44:18–25 (1997). Pal M, Paul SK and Das AK, Assessment of tea quality associating biochemical parameters and the tasters’ choice. Two and a Bud 45:16–30 (1999). Edelstein A, Introduction to Data Mining and Knowledge Discovery. Two Crows Corporation; 2 edition (1998). Hastie T, Tibshirani R and Friedman J, The Elements of Statistical Learning, Data Mining, Inference, and Prediction. Springer, New York, NY (2001). Wilcox RR, Fundamentals of Modern Statistical Methods. Springer, New York, NY (2001). Griliches Z, Price Indexes and Quality Change: Studies in New Methods of Measurement. Harvard University Press, Cambridge, MA (1971). Wooldridge J, Some alternatives to Box–Cox regression model. Int Econ Rev 4:935–955 (1991). Martins-Filho C and Bin O, Estimation of hedonic price function via additive nonparametric regression. Empir Econ 30:93–114 (2005). Friedman JH, Multivariate adaptive regression splines. Ann Statist 19:1–67 (1991). McDowell I, Feakes J and Gay C, The phenolic pigment composition of black tea liquors – Part I: Predicting quality. J Sci Food Agric 69:467–474 (1995). J Sci Food Agric 88:1354–1362 (2008) DOI: 10.1002/jsfa Journal of the Science of Food and Agriculture J Sci Food Agric 88:1363–1368 (2008) Comparison of volatile emissions from undamaged and mechanically damaged almonds† John J Beck,1∗ Bradley S Higbee,2 Glory B Merrill1 and James N Roitman1‡ 1 United States Department of Agriculture, Agricultural Research Service, Western Regional Research Center, Plant Mycotoxin Research, 800 Buchanan Street, Albany, CA 94710, USA 2 Paramount Farming Company, 33141 East Lerdo Highway, Bakersfield, CA 93308, USA Abstract BACKGROUND: The navel orangeworm (NOW) Amyelois transitella (Walker) is a major insect pest of almonds causing considerable monetary setbacks for both growers and processors, and thus control of NOW is one of the top priorities for the almond industry. Field observations purport that NOW is attracted to previously injured almonds. Accordingly, in this study the volatile output of damaged almonds was investigated in an effort to identify potential attractants for further studies into the control and/or monitoring of NOW. Mature almonds from the Monterey variety were evaluated for their volatile composition after mechanical damage and compared with the volatile composition of undamaged almonds. RESULTS: Volatile organic compounds (VOCs) were collected on Tenax, desorbed and identified via gas chromatography/mass spectrometry analysis. VOCs unique to the damaged tree nuts included trace amounts of 3-pentanol and isomers of the spiroketal chalcogran. VOCs that increased in relative amounts after damage include the spiroketal conophthorin and numerous four-carbon ester and ketone as well as alcohol derivatives, in addition to two eight-carbon chain compounds. CONCLUSION: Several VOCs, both unique and in increased amounts, were identified from damaged almonds. Their presence in damaged almonds warrants further investigation into their role in NOW response to damaged almonds, which may lead to insights into the control and/or monitoring of NOW. Published in 2008 by John Wiley & Sons, Ltd. Keywords: almond; damaged; spiroketal; Tenax; volatile INTRODUCTION The navel orangeworm (NOW) Amyelois transitella (Walker) is a major insect pest of almonds grown in California and causes considerable monetary setbacks for both growers and processors. Control of NOW has been stated as one of the top priorities for the almond industry, with another priority being the development of new pest management tools.1 There is twofold interest in controlling NOW, namely its direct damage to tree nuts and the associated contamination of toxin-producing fungi (mycotoxins) resulting from NOW feeding damage, which provides avenues for infection by mycotoxigenic fungi. The point of damage into the tree nut from the pest insect exposes the protective layers (hull, shell, seed coat) surrounding the kernel. This point of entry allows for ambient spores of aspergilli to enter and thus contaminate the nut.2 Contamination of tree nuts by mycotoxins is a chief concern for both human food and animal feed safety, with both areas experiencing major export issues as a result of the contamination.3 The aflatoxin-producing (aflatoxigenic) fungi most relevant to agriculture include Aspergillus flavus and Aspergillus parasiticus. Aflatoxin is presently a significant food safety problem owing to its carcinogenic and teratogenic attributes. The current total aflatoxin action threshold for international export of tree nuts is set at 4 ppb compared with the domestic level of 20 ppb set by the Food and Drug Administration (FDA).2,3 California is the top producer of almonds, supplying 75% of the world’s needs.1 Approximately 5% of California’s cropland is dedicated to almond production.4 The ∗ Correspondence to: John J Beck, USDA-ARS, WRRC, Plant Mycotoxin Research, 800 Buchanan Street, Albany, CA 94710, USA E-mail: john.beck@ars.usda.gov † Segments of this report were presented at the 48th Annual Meeting of the American Society of Pharmacognosy, Portland, ME 04101, USA, 14–18 July 2007 and at the 17th Annual Multi-Crop Aflatoxin Elimination Workshop, Sacramento, CA 95814, USA, 25–28 October 2004 ‡ Retired Contract/grant sponsor: USDA-ARS; contract/grant number: CRIS 5325-42000-036-00 Contract/grant sponsor: Paramount Farming Company; contract/grant number: CRADA #58-3K95-7-1198 (Received 18 September 2007; revised version received 10 December 2007; accepted 28 December 2007) Published online 14 April 2008; DOI: 10.1002/jsfa.3224 This article is a US Government work and is in the public domain in the USA. J Sci Food Agric 0022–5142/2008/$30.00 JJ Beck et al. California almond industry generates approximately $2 billion annually, with the total California tree nut industry reporting over $3.5 billion. About 50–70% of California tree nuts are exported overseas annually, with 80% of almond production alone being exported.1 The strict export action levels for aflatoxin have resulted in mycotoxin management issues for producers as well as state and federal governments. Actual costs of crop loss due to aflatoxin contamination in California were estimated to have been $23–47 million over the period 1995–2001.3 Moreover, the economic and health impacts of mycotoxins have been stated to be severe for developing nations.5 In a recent investigation, researchers reported the observation that female NOW moths were attracted to injured almonds.6 Current attractants used in the field and/or lab for NOW include the female sex pheromone of NOW, (Z,Z)-11,13-hexadecadienal,7 a pheromone blend of (Z,Z,Z,Z,Z)-3,6,9,12,15-tricosapentaene, (Z,Z,Z,Z,Z)-3,6,9,12,15-pentacosapentaene, ethyl palmitate and ethyl-(Z,Z)-11,13-hexadecadien-1-yl acetate8 and the almond oil fatty acids myristic, palmitic, stearic, oleic and linoleic.9 Investigations on VOCs from almonds report the detection of 2hexyl-3-methylmaleic anhydride10 and various alkane, alkene, alkanol, aromatic and furan VOCs.11 However, a search of the literature does not provide examples of VOC emission as a result of injury to the almond. As part of our ongoing efforts to address the concerns regarding NOW, our labs investigated the VOC output of mechanically damaged (DMG) almonds from the Monterey variety and compared the VOC fingerprint with that of undamaged (CTRL) almonds to ascertain what VOCs, if any, were unique to DMG almonds. The major VOCs from the CTRL and DMG experiments were compared and contrasted. MATERIALS AND METHODS Plant material Fruits of Prunis dulcis (P. Mill.) D.A. Webb, variety Monterey, common name sweet almond, were collected in two batches during mid to late June 2006 from the groves of Paramount Farming Company, Bakersfield, CA, USA. Each batch was replicated in triplicate over different days. Batch 1 consisted of almonds that had been injured while intact on the tree, allowed to remain on the tree for approximately 14 days, then removed and placed in glass jars with a Teflon paper seal between the cap and jar. The injury/damage consisted of hull penetration with an 8 penny nail (3 mm diameter). Batch 2 consisted of control almonds that were not injured, removed from the tree and placed in glass jars with a Teflon paper seal between the cap and jar. Batches 1 and 2 were collected during concurrent time frames. Batches were sent via overnight delivery to the USDA-ARS facility in Albany, CA, USA for volatile evaluation. 1364 Collection of VOCs6 Almonds (ca 500 per experiment) were transferred to a 12 L round-bottomed flask fitted with an inlet for purified airflow at 1 L min−1 and a Tenax (25 g) collection system. VOCs were collected for 18 h and desorbed with freshly distilled diethyl ether (100 mL), then the ether was concentrated to a volume of ca 1 mL with a warm water bath and a Vigreux distillation column. Gas chromatography/mass spectrometry (GC/MS) analysis Separation of the collected VOC mixture was achieved with a DB-Wax column (60 m × 0.32 mm i.d. × 0.25 µm; J&W Scientific, Folsom, CA, USA) installed on an HP 6890 gas chromatograph (GC) coupled to an HP 5973 mass selective detector (MSD) (Hewlett Packard, Palo Alto, CA, USA). Extracts were analysed with the following method: 1 µL injections; injector temperature, 150 ◦ C; splitless mode; inlet temperature, 150 ◦ C; inlet pressure, 7.7 psi; total flow, 11.9 mL min−1 ; He carrier gas at 7.7 psi; flow, 1.5 mL min−1 ; velocity, 31 cm min−1 ; constant flow; oven settings: initial temperature, 30 ◦ C; hold time, 4 min; ramp, 2 ◦ C min−1 ; final temperature, 200 ◦ C; hold time, 30 min. The MSD parameters were as follows: source temperature, 230 ◦ C; MS quadrupole temperature, 150 ◦ C; electron impact (EI) mode, 70 eV; solvent delay, 1 min; scan group 1, 40–300 amu; scan group 2 at 20 min, 40–450 amu. National Institute of Standards and Technology (NIST), Wiley and internally generated databases were used for fragmentation pattern identification. Retention indices (RIs) were calculated using a homologous series of nalkanes on a DB-Wax column. Compounds that did not match the RIs of known VOCs from our database and/or did not provide sufficient mass fragmentation pattern matches were assigned as unknown in Table 1. Statistical analysis GC/MS analysis was performed on each of the three separate samples for both the CTRL and DMG batches of almonds. The relative areas for each of the compounds from the GC/MS runs were normalised to the internal standard cyclodecanone (15 µg) and the means, standard deviations and confidence limits (95%) in Table 1 and Fig. 3 were calculated with Microsoft Excel software (Redmond, WA, USA). RESULTS AND DISCUSSION Analysis of the major VOCs emitted by both the CTRL and DMG almonds provides a wide range of compounds, which corroborated other reports and added to the volatile fingerprint of almonds in the literature. Table 1 provides a list of the major VOCs detected from both experiments. Examination of Table 1 showed a number of monoterpenes common to citrus and other plants,12 namely αpinene, camphene, β-pinene, β-myrcene, limonene J Sci Food Agric 88:1363–1368 (2008) DOI: 10.1002/jsfa Volatile emissions from undamaged and mechanically damaged almonds Figure 1. Total ion chromatogram (relative abundance versus time) illustrating a typical elution pattern of DMG almond VOCs. Unique, increased and/or notable compounds are labelled with numbers corresponding to compounds listed in Table 1. and cymene. The compounds α-pinene and camphene were noted to be in relatively large amounts in the CTRL almonds and underwent a small decrease in volatile output for the DMG almonds (Fig. 1 illustrates a typical GC elution pattern for DMG VOCs). Camphene and α-pinene are both common, non-specific plant VOCs that have a wide range of semiochemical activity,13 – 15 but neither has been reported for activity against NOW. The remaining monoterpenes are ubiquitous as plant VOCs, and several have been noted as semiochemicals.14 The spiroketal conophthorin (7-methyl-1,6-dioxaspiro[4.5]decane), in unknown configuration, was also observed to undergo a small increase in relative amounts in several of the DMG almond volatile analyses. Conophthorin is present in several insects and plants and in varying concentrations of isomers (Fig. 2).16 The sesquiterpenes bourbonene (as a mixture with benzaldehyde), β-copaene and aromadendrene also increased in relative amounts in the DMG almonds. These particular sesquiterpenes have been noted to occur together in potato leaf VOCs.17 Bourbonene and β-copaene are pheromones for the European birch aphid,18 and aromadendrene has been reported to be an attractant for the Brazilian eucalyptus brown looper.15 However, none of the noted sesquiterpenes has been implicated as possessing activity against NOW. The only compounds to demonstrate corroboration of previous reports of almond VOCs were 2pentylfuran, nonanal, 1-octen-3-ol, benzaldehyde O O O Z-(5S,7R) O O Z-(5R,7S) O E-(5R,7R) O O E-(5S,7S) Figure 2. Stereoisomers of conophthorin (7-methyl-1,6-dioxaspiro [4.5]decane). J Sci Food Agric 88:1363–1368 (2008) DOI: 10.1002/jsfa and 2-phenylethyl alcohol.11,19 Notable differences between the work performed by Buttery et al.,11 which evaluated VOCs from almond hulls, and the VOCs collected in the present study were the detection here of numerous four-carbon ester and ketone as well as alcohol derivatives. Specific examples were the compounds that also showed a general increase in amounts between the CTRL and DMG almond VOCs, namely 2-butanol, ethyl 2methylbutyrate, ethyl isovalerate, ethyl 2-butenoate, ethyl 3-methylbut-2-enoate, ethyl tiglate and 3hydroxy-2-butanone. Several of these VOCs have been attributed to fruity, wine aroma and smoky odours20,21 and are known semiochemicals,22 – 24 yet are not associated with NOW semiochemicals. The compounds that demonstrated statistically valid increases were ethyl 2-methylbutyrate, 2-methyland 3-methyl-1-butanol, ethyl tiglate and β-copaene (Fig. 3), in addition to one unknown compound. Several compounds in Table 1 were noted to be indicative of fungal growth. Of particular interest were 2-methyl- and 3-methyl-1-butanol and 2-pentyfuran owing to their relatively large amounts. The butanol Figure 3. VOCs showing a statistically significant increase (95% confidence limit) in DMG almonds. 1365 JJ Beck et al. Table 1. Major volatile components of Monterey (MO) damaged (DMG) and control (CTRL) almondsa Relative amountb No. Compound RIc RI (PMR)d RI (lit.) MO CTRL MO DMG 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 α-Pinene 2-Butanol Ethyl butyrate Unknown Ethyl 2-methylbutyrate Camphene Ethyl isovalerate β-Pinene Diethyl carbonate 3-Pentanol Ethyl 2-butenoate β-Myrcene Limonene 3-Methyl- and 2-methyl-1-butanol Ethyl 3-methylbut-2-enoate 2-Pentylfuran Ethyl tiglate, ethyl hexanoate 1-Dodecene Styrene 3-Octanone Cymene isomer (para- 1264) 3-Hydroxy-2-butanone Conophthorin Unknown E-4,8-Dimethyl-1,3,7-nonatriene Chalcogran isomer #1 Chalcogran isomer #2 Nonanal Tetradec-1-ene 1-Octen-3-ol Bourbonene/benzaldehyde mix trans-α-Bergamotene β-Copaene Aromadendrene 1-Hexadecene Ethyl benzoate 1-Methyl-2-pyrrolidinone isomer Cyclodecanonee Unknown 2-Phenylethyl alcohol 1-Dodecanol 2-Phenoxyethanol Docosane Unknown Unknown Unknown Vanillin 1016 1032 1037 1044 1053 1058 1067 1092 1100 1111 1158 1160 1188 1207 1219 1227 1231 1243 1247 1249 1261 1274 1280 1286 1301 1343 1348 1387 1444 1451 1505 1577 1582 1606 1645 1654 1662 1726 1870 1899 1968 2126 2187 2264 2279 2471 2541 1020 1027 1029 1014 1019, 1025 2.21 (1.42) 0.24 (0.12) 0.51 (0.37) ND 0.32 (0.09) 2.27 (1.24) 2.57 (0.25) 0.36 (0.14) 0.53 (0.67) ND 0.14 (0.25) 0.77 (0.04) 1.06 (0.36) 1.66 (0.19) 0.16 (0.28) 1.81 (0.04) 0.38 (0.07) 0.69 (0.75) 0.30 (0.02) Tr 0.16 (0.04) 1.83 (0.76) 0.34 (0.26) ND 0.11 (0.10) ND ND 0.30 (0.36) 1.02 (1.77) 0.28 (0.06) 0.21 (0.07) 0.19 (0.19) 0.64 (0.20) 0.11 (0.18) 0.63 (1.09) 0.51 (0.60) 0.60 (0.09) 15.00 (0.00) 0.58 (0.28) 0.16 (0.02) 0.37 (0.13) 0.21 (0.06) 0.52 (0.12) 0.17 (0.03) 0.56 (0.28) 0.18 (0.04) 1.88 (0.57) 1.78 (0.74) 0.32 (0.20) 0.59 (0.35) 0.19 (0.26) 0.62 (0.04) 1.98 (0.92) 3.34 (2.10) 0.32 (0.12) ND 0.08 (0.08) 0.20 (0.27) 0.56 (0.13) 0.80 (0.18) 3.63 (2.95) 0.23 (0.30) 0.93 (0.38) 0.74 (0.14) ND 0.52 (0.56) 0.12 (0.11) 0.11 (0.10) 2.55 (1.85) 0.79 (0.88) Tr 0.15 (0.04) Tr Tr ND ND 0.42 (0.12) 0.47 (0.66) 0.13 (0.02) 1.04 (0.27) 0.24 (0.12) ND 0.18 (0.20) 0.55 (0.22) 15.00 (0.00) 0.53 (0.21) 0.42 (0.62) 0.18 (0.17) 0.14 (0.12) 0.30 (0.29) Tr 0.53 (0.33) 0.21 (0.21) 0.72 (0.68) 1046 1063 1062 1106 1101 1103 1158 1157 1197 1205 1226 1232 1242 1252 1251 1250 1278 1048 1088, 1094 1108 1154, 1159 1182, 1195 1190 1220, 1224 1302 1389 1446 1448 1516 1582 1589 1390, 1400 1428, 1446 1605 1647 1661 1744 1910 2142 2200 2585 1848 a Almonds collected on three different days. Volatile amounts reported as mean, normalised to 15 µg of internal standard, with standard deviation in parentheses; ND, not detected; Tr, trace amount (<0.10 µg). c Retention index relative to n-alkanes on DB-Wax column. d RI of volatile compounds based on in-house database. e Internal standard. b VOCs increased in amounts from CTRL to DMG, while 2-pentylfuran decreased between these two experiments. The VOCs noted to occur during fungal growth, particularly Aspergillus species, are 2-methyland 3-methyl-1-butanol, 2-pentylfuran, 1-octen-3-ol 1366 and 3-octanone.25,26 In addition to its previously reported occurrence in almonds,11,19 it should be noted that 1-octen-3-ol is also a plant volatile of numerous plants, including genera of the Orchidaceae, as well as a semiochemical for several different J Sci Food Agric 88:1363–1368 (2008) DOI: 10.1002/jsfa Volatile emissions from undamaged and mechanically damaged almonds insects. 2-Pentylfuran is also an Orchidaceae plant volatile, but to a much lesser extent (The Pherobase, www.pherobase.com, accessed 22 August 2007). The eight-carbon VOCs, in addition to the sesquiterpenes myrcene, limonene and copaene, have been reported to be produced by Penicillium species.27 Moreover, sesquiterpene VOCs unique to A. flavus28 were not found in this study, thus indicating the possible presence of Penicillium more so than Aspergillus, yet this information did not provide enough evidence to exclusively implicate one particular microbe. Both Aspergillus and Penicillium are known to be present on almonds.29 The compounds unique, albeit only present in trace amounts, to the DMG almond VOCs were 3-pentanol, two chalcogran isomers (Fig. 4) and one unknown compound (No. 24, Table 1). 3-Pentanol is relatively new as a semiochemical, with only one study that demonstrated its ability to provoke a response in the male sugarcane weevil.30 Interestingly, the same study reported ethyl butyrate, among other esters, as eliciting an antennal response in the female sugarcane weevil. The chalcogran isomers, however, have a long history of semiochemical activity, primarily with the European spruce bark beetle Pityogenes chalcographus.31,32 The (2S,5R) and (2S,5S) configurations of chalcogran are found in P. chalcographus, and as two isomers with unknown configurations in the bark beetle Pityogenes quadridens.33 It is interesting to note that Byers et al.34 used combinations of chalcogran, camphene and α- and β-pinene, all VOCs detected in DMG almond VOCs, along with the compound methylE,Z-2,4-decadienoate to enable host recognition of the bark beetle. Other correlations between DMG almond VOCs and semiochemicals from bark beetles are similar VOCs, among others, emitted from Ips typographus males under stress, namely α- and βpinene, camphene, myrcene, limonene and cymene, and similar VOCs from Pityogenes species, namely limonene, chalcogran, 1-octen-3-ol and 2-phenylethyl alcohol.35 The occurrence of the chalcogran isomers in this and the one associated previous study6 does not conclusively determine whether the spiroketals are emitted as a result of damage to the almonds or formed by fungal growth. The detection of several VOCs indicative of fungal growth brings into question whether or not the method of removing the DMG almonds after several days on the tree and subsequent transportation to the laboratory allows ambient fungi to initiate growth on the almonds. Investigations into this matter are ongoing. O O O O Z-(2R,5R) Z-(2S,5S) O O E-(2R,5S) E-(2S,5R) O Figure 4. Stereoisomers of chalcogran (2-ethyl-1,6-dioxaspiro [4.4]nonane). J Sci Food Agric 88:1363–1368 (2008) DOI: 10.1002/jsfa O CONCLUSION The VOC emissions of control and damaged almonds were investigated. VOCs unique to damaged almonds include 3-pentanol and two isomers of the spiroketal chalcogran (unknown configuration) in trace amounts. Other VOCs that increased in relative quantity include the spiroketal conophthorin (unknown configuration), numerous four-carbon ester and ketone as well as alcohol derivatives, in addition to two eightcarbon chain compounds. VOCs suggestive of fungal growth were noted and brought to question whether the chalcogran isomers are damage-induced or a result of fungal growth. Also notable was the apparent correlation between several bark beetle semiochemicals and VOCs from the CTRL and DMG almonds. The detection of the VOCs noted above provides evidence that further investigation into their role in NOW response to damaged almonds is required. ACKNOWLEDGEMENTS This research was conducted under USDA-ARS CRIS project 5325-42000-036-00 and a cooperative research and development agreement with Paramount Farming Company (CRADA #58-3K95-7-1198). REFERENCES 1 Almond Board of California, The Foundation for a Pest Management Strategic Plan in Almond Production: Summary of a Workshop Held December 12, 2002. California Pest Management Center, Modesto, CA (2003). 2 Campbell BC, Molyneux RJ and Schatzki TF, Current research on reducing pre- and post-harvest aflatoxin contamination of U.S. almond, pistachio, and walnut. J Toxicol Toxin Rev 22:225–266 (2003). 3 Robens J and Cardwell W, The costs of mycotoxin management to the USA: management of aflatoxins in the United States. J Toxicol Toxin Rev 22:143–156 (2003). 4 Epstein L, Bassein S, Zalom FG and Wilhoit LR, Changes in pest management practice in almond orchards during the rainy season in California, USA. Agric Ecosyst Environ 83:111–120 (2001). 5 Wu F, Mycotoxin risk assessment for the purpose of setting international regulatory standards. Environ Sci Technol 38:4049–4055 (2004). 6 Roitman JN, Merrill GB and Higbee BS, What attracts navel orangeworm to oviposit preferentially on wounded almonds rather than unblemished fruit? The search for a volatile attractant. Proc 4th Annu Fungal Genomics, 5th Annu Multicrop Fumonisin, 17th Annu Multi-crop Aflatoxin Elimination Workshops, pp. 27 (2004). 7 Coffelt JA, Vick KW, Sonnet PE and Doolittle RE, Isolation, identification, and synthesis of a female sex pheromone of the navel orangeworm, Amyelois transitella (Lepidoptera: Pyralidae). J Chem Ecol 5:955–966 (1979). 8 Leal WS, Parra-Pedrazzoli AL, Kaissling K-E, Morgan TI, Zalom FG, Pesak DJ, et al, Unusual pheromone chemistry in the navel orangeworm: novel sex attractants and a behavioral antagonist. Naturwissenschaften 92:139–146 (2005). 9 Phelan PL, Roelofs CJ, Youngman RR and Baker TC, Characterization of chemicals mediating ovipositional hostplant finding by Amyelois transitella females. J Chem Ecol 17:599–613 (1991). 1367 JJ Beck et al. 10 Buttery RG, Seifert RM, Haddon WF and Lundin RE, 2Hexyl-3-methylmaleic anhydride: an unusual volatile component of raisins and almond hulls. J Agric Food Chem 28:1336–1338 (1980). 11 Buttery RG, Soderstrom EL, Seifert RM, Ling LC and Haddon WF, Components of almond hulls: possible navel orangeworm attractants and growth inhibitors. J Agric Food Chem 28:353–356 (1980). 12 Choi H-S, Lipolytic effects of citrus peel oils and their components. J Agric Food Chem 54:3254–3258 (2006). 13 Byers JA, Birgersson G, Lofqvist J, Appelgren M and Bergstrom G, Isolation of pheromone synergists of bark beetle, Pityogenes chalcographus, from complex insect–plant odors by fractionation and subtractive-combination bioassay. J Chem Ecol 16:861–876 (1990). 14 Whitehead AT, Electroantennogram responses by mountain pine beetles, Dendroctonus ponderosae Hopkins, exposed to selected semiochemicals. J Chem Ecol 12:1603–1621 (1986). 15 Batista-Pereira LG, Fernandes JB, Correa AG, Da Silva MFGF and Vierira PC, Electrophysiological responses of eucalyptus brown looper Thyrinteina arnobia to essential oils of seven Eucalyptus species. J Braz Chem Soc 17:555–561 (2006). 16 Zhang Q-H, Rolasch T, Schlyter F and Francke W, Enantiospecific antennal response of bark beetles to spiroacetal (E)-conophthorin. J Chem Ecol 28:1839–1852 (2002). 17 Khalilov LM, Khalilova AZ, Odinokov VN, Baltaev UA, Paramonov EA and Dzhemilev UM, Identification and biological activity of the volatile organic substances emitted by plants and insects. II. Sesquiterpene compostion of the native scent of leaves of the potato Solanum tuberosum. Chem Nat Comp 35:422–426 (1999). 18 Francis F, Vandermoten S, Verheggen F, Lognay G and Haubruge E, Is (E)-β-farnesene the only volatile terpenoid in aphids? J Appl Entomol 129:6–11 (2005). 19 Caja MM, Ruiz del Castillo ML, Martinez Alvarez R, Herraiz M and Blanch GP, Analysis of volatile compounds in edible oils using simultaneous distillation–solvent extraction and direct coupling of liquid chromatography with gas chromatography. Eur Food Res Technol 211:45–51 (2000). 20 Marti MP, Mestres M, Sala C, Busto O and Guasch J, Solidphase microextraction and gas chromatography olfactometry analysis of successively diluted samples. A new approach of the aroma extract dilution analysis applied to the characterization of wine aroma. J Agric Food Chem 51:7861–7865 (2003). 21 Joulain D, Casazza A, Laurent R, Portier D, Guillamon N, Pandya R, et al, Volatile flavor constituents of fruits from southern Africa: mobola plum (Parinari curatellifolia). J Agric Food Chem 52:2322–2325 (2004). 22 Bartelt RJ, Schaner AM and Jackson LL, Aggregation pheromones in Drosophila borealis and Drosophila littoralis. J Chem Ecol 14:1319–1327 (1988). 1368 23 Pinero JC and Prokopy RJ, Field evaluation of plant odor and pheromonal combinations for attracting plum curculios. J Chem Ecol 29:2735–2748 (2003). 24 Malo EA, Cruz-Lopez L, Toledo J, Del Mazo A, Virgen A and Rojas JC, Behavioral and electrophysiological responses of the Mexican fruit fly (Diptera: Tephritidae) to guava VOCs. Fla Entomol 88:364–371 (2005). 25 Magan N and Evans P, VOCs as an indicator of fungal activity and differentiation between species, and the potential use of electronic nose technology for early detection of grain spoilage. J Stored Prod Res 36:319–340 (2000). 26 Gao P, Korley F, Martin J and Chen BT, Determination of unique microbial volatile organic compounds produced by five Aspergillus species commonly found in problem buildings. Am Ind Hyg Assoc J 63:135–140 (2002). 27 Nilsson T, Larsen TO, Montanarella L and Madsen JO, Application of head-space solid-phase microextraction for the analysis of volatile metabolites emitted by Penicillium species. J Microbiol Meth 25:245–255 (1996). 28 Zeringue HJ, Bhatnagar D and Cleveland TE, C15H24 volatile compounds unique to aflatoxigenic strains of Aspergillus flavus. Appl Environ Microbiol 59:2264–2270 (1993). 29 Bayman P, Baker JL and Mahoney ME, Aspergillus on tree nuts: incidence and associations. Mycopathologia 155:161–169 (2002). 30 Perez AL, Campos Y, Chinchilla CM, Oehlschlager AC, Gries G, Gries R, et al, Aggregation pheromones and host kairomones of West Indian sugarcane weevil, Metamasius hemipterus sericeus. J Chem Ecol 23:869–888 (1997). 31 Francke W, Heemann V, Gerken V, Renwick JAA and Vite JP, 2-Ethyl-1,6-dioxaspiro[4.4]nonane, principal aggregation pheromone of Pityogenes chalcographus (L.). Naturwissenschaften 64:590–591 (1977). 32 Byers JA, Hogberg H-E, Unelius CR, Birgersson G and Lofqvist J, Structure–activity studies on aggregation pheromone components of Pityogenes chalcographus (Coleoptera: Scolytidae), all stereoisomers of chalcogran and methyl 2,4-decadienoate. J Chem Ecol 15:685–695 (1989). 33 Francke W and Kitching W, Spiroketals in insects. Curr Org Chem 5:233–251 (2001). 34 Byers JA, Birgersson G, Lofqvist J and Bergstrom G, Synergistic pheromones and monoterpenes enable aggregation and host recognition by a bark beetle. Naturwissenschaften 75:153–155 (1988). 35 Francke W, Bartels J, Meyer H, Schroder F, Kohnle U, Baader E, et al, Semiochemicals from bark beetles: new results, remarks, and reflections. J Chem Ecol 21:1043–1063 (1995). J Sci Food Agric 88:1363–1368 (2008) DOI: 10.1002/jsfa Journal of the Science of Food and Agriculture J Sci Food Agric 88:1369–1375 (2008) Cadmium accumulation in Agaricus blazei Murrill Jian Cheng Huang,1∗ Kai Ben Li,1 Ying Rui Yu,1 Hanwen Wu2 and De Li Liu2 1 Institute of Plant Protection, Fujian Academy of Agricultural Sciences, Fuzhou, Fujian 350013, China Graham Centre for Agricultural Innovation (a collaborative alliance between NSW Department of Primary Industries and Charles Sturt University), Wagga Wagga Agricultural Institute, Wagga Wagga, NSW 2650, Australia 2 E.H. Abstract BACKGROUND: Agaricus blazei Murrill has become a popular food source of high nutritional and medicinal value in many Asian countries. However, there are growing health concerns on cadmium (Cd) accumulation by A. blazei. Experiments were conducted to investigate Cd accumulation patterns in A. blazei and to identify key factors contributing to Cd accumulation. RESULTS: Cd concentrations in the substrates and subsequent fruit bodies declined rapidly after the first, second and third harvest wave. A quick rinse of the fruit bodies in water reduced the Cd concentration by 27–54% depending on the strain. The Cd concentration in the fruit bodies decreased as the fruit body yield m−2 or fruit body number m−2 increased, while it increased as the substrate Cd content or fruit body size increased. Cd accumulation was positively associated with phosphorus (P) uptake. CONCLUSION: The results suggests that, in the A. blazei-growing region studied, Cd and P concentrations in the substrates and casing soil should be reduced in order to effectively minimise Cd accumulation in the fruit bodies. It is also necessary to improve fruit body yield by increasing the number of medium-sized fruit bodies. Overgrowth of individual fruit bodies stimulates Cd accumulation in A. blazei.  2008 Society of Chemical Industry Keywords: Agaricus blazei Murrill; cadmium; substrate; fruit body; accumulation; yield components; modelling INTRODUCTION Agaricus blazei Murrill is a mushroom species of the phylum Basidiomycota originating from Brazil.1 It is a medicinally important fungus that is widely consumed and prescribed in countries such as China, Japan, South Korea and Turkey.2 – 4 Agaricus blazei is rich in protein, amino acids, microelements and polysaccharides. Since its introduction into Japan in 1965 by Takatoshi Furumoto, it has gradually gained popularity and is now widely known as ‘himematsutake’.1,3 In 1992 the mushroom was introduced from Japan into China and successfully cultivated and is now in large-scale commercial production.5,6 The high nutritional and medicinal value of A. blazei has been well documented.3 It has been widely used as a folk medicine for various diseases, including diabetes, hyperlipidaemia, arteriosclerosis, osteoporosis, gastric ulcer and cancer.1,7,8 Extensive research has been conducted to identify bioactive substances from A. blazei. A number of such substances have been identified, e.g. polysaccharides, cytotoxic steroids, lectin and antimutagens.9 – 11 Among them, various polysaccharides and protein–polysaccharide complexes isolated from the fruit bodies of A. blazei possess antitumour activity.3,12 Ohno et al.12 reported that a highly branched 1,3-βglucan, a main component of polysaccharides, showed strong antitumour activity. It has been demonstrated that the aqueous extract of A. blazei provides significant protection against mutagenicity induced both in vivo by cyclophosphamide and in vitro by methyl methanesulfonate.7 The most anticipated pharmacological effect of A. blazei is modulation of the immune system against cancer. Recent research has shown that extracts of A. blazei selectively up-regulate the genes related to immune function, particularly proinflammatoric genes such as the interleukins IL1B and IL8.13 However, accumulation of cadmium (Cd) in A. blazei has received increasing attention in the past few decades owing to the potential health hazard of this heavy metal. The genus Agaricus is able to accumulate high concentrations of Cd, up to 100–300 mg kg−1 dry matter (DM).14,15 Accumulation of Cd in A. blazei has been found to range from 3 to 35 mg kg−1 DM.16 Cadmium is accumulated mainly in the kidneys, spleen and liver, and its concentration in ∗ Correspondence to: Jian Cheng Huang, Institute of Plant Protection, Fujian Academy of Agricultural Sciences, Fuzhou, Fujian 350013, China E-mail: huangjc12@sina.com Contract/grant sponsor: Department of Science and Technology of Fujian, China; contract/grant number: 2003-N-011 (Received 16 September 2007; revised version received 12 December 2007; accepted 11 January 2007) Published online 7 April 2008; DOI: 10.1002/jsfa.3225  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 JC Huang et al. blood serum increases markedly following mushroom consumption.17 Cadmium is a toxic element exerting profound adverse effects on many life processes, e.g. kidney function and the male reproductive system.18,19 Uptake of a high concentration of Cd via the food chain is detrimental to the human body, causing anaemia, high blood pressure, osteoporosis, kidney dysfunction, lung impairment, reproductive function damage and carcinogenesis.20,21 The FAO/WHO has listed Cd as the third most important food pollutant after aflatoxins and arsenic (As). The FAO/WHO provisional tolerable weekly intake for Cd has a limit of 0.007 mg kg−1 body weight, and the maximum permissible dose for an adult is 0.5 mg Cd week−1 .22 If an average person with a body weight of 60–70 kg consumes a 300 g portion of fresh mushrooms per meal, which generally contains 100 g DM kg−1 ,3 the tolerable weekly intake is thus easily reached by a single portion of 300 g of fresh mushrooms containing 14 mg Cd kg−1 DM. Therefore Cd accumulation in A. blazei could potentially affect food safety and eventually have a direct or indirect impact on human health. However, there is little information available on the distribution and accumulation of Cd in A. blazei. The present study aimed to investigate the accumulation patterns of Cd in A. blazei and to determine the impact of fruit body yield, yield components and phosphorus (P) uptake on Cd concentration in order to maximise mushroom production with minimum accumulated Cd in the fruit body. MATERIALS AND METHODS Agaricus blazei strains and site selection Two strains of A. blazei, N-2 (narrow-stalk type) and A-6 (broad-stalk type), were obtained from the Institute of Plant Protection, Fujian Academy of Agricultural Sciences, Fujian, China. Experiments were conducted in the township of Xin Du in southeastern Fujian. The research site was located in a rural area about 30 km away from industrial sites. Local farmers have a long history of growing common mushroom (Agaricus bisporus) and A. blazei. Agaricus blazei is normally grown twice a year in spring and autumn. The present experiment was carried out in the spring season. Drinking water was used to minimise the risk of water contamination from other sources. Substrate preparation and cultivation The substrates consisted of 1000 kg of rice straw, 200 kg of dry cow manure, 150 kg of commercial fertilisers (N:P:K = 2.5:1:2.5), 200 kg of ammonium carbonate and 100 kg of gypsum. On day 1 a layer of rice straw 20 cm thick was spread onto levelled ground to form a rectangular pile of 2 m × 4 m. A thin layer of cow manure and fertilisers was then evenly spread on top of the straw and covered with another layer of rice straw. This process was alternated six times until the pile reached a height of 1.5 m. The pile was turned on 1370 day 5 to mix the ingredients thoroughly. This turning process was periodically repeated three more times on days 9, 12 and 14. Gypsum was added during the first turning operation. Drinking water was used to maintain a moisture level of 650 g kg−1 during layering and turning and throughout the entire composting period. The substrate was then subjected to steam pasteurisation and evenly spread onto culture beds in a temperature-controlled culture room. After the temperature of the pasteurised substrate had dropped to about 25 ◦ C, wheat grain spawn was introduced at a rate of 1000 mL m−2 and thoroughly mixed into the substrate. A layer of casing soil (peat moss) was spread over the surface of the substrate beds to a depth of 3 cm when the substrate surface was completely colonised by the spawn. The final substrate bed had a height of 15 cm. The bed was irrigated periodically. Common district practices were used to manage the continuous growth of A. blazei.6 Substrate and fruit body sampling Substrates were consecutively sampled at different times, i.e. immediately after the steam pasteurisation stage, after complete substrate colonisation by the spawn (colonisation stage) and at the first, second and third harvest stages. The substrate samples were oven dried at 65 ◦ C for 48 h. Fruit bodies of the first and second harvests were sampled from nine fixed quadrats of 1 m2 each, which were randomly selected to cover the range of various numbers of fruit bodies and hence yields m−2 . The samples of the third harvest were taken from another nine random quadrats (1 m2 ). The fruit bodies from each quadrat were assessed for fresh weight (g m−2 ) and number of fruit bodies m−2 . The fresh fruit bodies were then bulked, oven dried and weighed prior to analyses. Effects of different Agaricus blazei strains, rinsing treatments and fruit body parts on Cd accumulation For each of the two A. blazei strains A-6 and N-2 a 1 kg sample of fresh fruit bodies was randomly obtained from the first harvest and subjected to a quick rinse with tap water for 2 min to remove the scales. Rinsed and non-rinsed samples (1 kg fresh weight each) were oven dried prior to Cd quantification. Another 1 kg sample of non-rinsed fresh fruit bodies was taken and separated into caps and stalks. The caps and stalks were oven-dried prior to Cd analyses. Quantification of Cd in cultivation substrates and fruit bodies Dried samples of fruit bodies and substrates were subjected to atomic absorption spectrometry measurement by Fujian Centre for Instrumental Analysis. A national standard protocol for element analysis was used for the determination of Cd (GB/T 5009.152003).23 Briefly, each dried sample was homogenised and stored in pre-cleaned polyethylene bottles until analysis. For analysis, a 5 g sample was placed in J Sci Food Agric 88:1369–1375 (2008) DOI: 10.1002/jsfa Cadmium accumulation in Agaricus blazei Murrill a porcelain crucible and ashed at 500 ◦ C for 8 h. The ash was then dissolved in 1 mL of a 4:1 (v/v) mixture of nitric acid (HNO3 ) and perchloric acid (HClO4 ), evaporated to dryness and heated again at 500 ◦ C for 4 h. The material obtained from this mineralisation process was further treated with 10 mL of 0.5 mol L−1 hydrochloric acid (HCl) and diluted with deionised water to a final volume of 50 mL. Three blank samples were treated in the same manner. An atomic absorption spectrometer (model AA6800, Shimadzu, Shanghai, China) operating at a wavelength of 228.8 nm was used to determine the Cd concentrations in the mushroom samples. Quantification of P in cultivation substrates and fruit bodies Phosphate was determined in accordance with a national standard protocol (GB/T5009.87–2003)23 based on the molybdenum blue method. Briefly, each dried sample was acid digested and subjected to reaction with molybdic acid to form a heteropoly molybdenum blue complex. Phosphorus (as P2 O5 ) was then determined spectrophotometrically by measurement of the absorbance at a wavelength of 660 nm with a UV–visible spectrophotometer (model 7220, Beijing Rayleigh Instrument Corp., Beijing, China). Data analysis From the sampling procedure a wide range of data for fruit body yield m−2 , fruit body number m−2 , dry weight per fruit body and Cd concentration were obtained. The data were subjected to linear and nonlinear regression analyses. The assumption was made that Cd accumulation in the fruit body (Cd, g m−2 ) is primarily related to fruit body yield (Y , g m−2 ) and follows the principle of enzyme kinetics.24 Thus Cd accumulation is a function of fruit body yield as described by Cd = Cdmx Y /(Y0.5 + Y ) (1) where Cdmx is the asymptote for Cd accumulation as the fruit body yield approaches infinity and Y0.5 is the fruit body yield for Cd accumulation at a half of Cdmx . The Cd concentration (CCd , g g−1 ) can be defined as CCd = Cd/Y (2) Combining Eqns (1) and (2) gives CCd = Cdmx /(Y0.5 + Y ) (3) Equation (3) was used to express the Cd concentration in the fruit body as a function of fruit body yield by a nonlinear least squares regression analysis. In addition, the relationships between fruit body yield, average fruit body dry weight per fruit and fruit body number m−2 were analysed using a similar equation. The nonlinear least squares regression analysis was conducted by a nonlinear module in SHAZAM.25 J Sci Food Agric 88:1369–1375 (2008) DOI: 10.1002/jsfa RESULTS AND DISCUSSION Dynamics of Cd accumulation in substrates and fruit bodies during fructification and harvesting period There were slight changes in Cd concentration in the substrates between the steam pasteurisation and substrate colonisation stages, with values of 0.29 and 0.32 × 10−6 g g−1 DM respectively. However, the Cd concentration in the substrates declined rapidly after the first, second and third harvests of fruit bodies, resulting in 29, 44 and 58% reductions respectively compared with the Cd concentration at the stage of complete colonisation by the spawn. Heavy metals are transported to the fruit body via the mycelium during the start of fructification.17 After the first harvest a significant amount of Cd in the substrates was exported owing to the removal of fruit bodies, thereby resulting in a decreasing Cd concentration being detected in the substrates. Similarly, the Cd concentration in the fruit bodies decreased after the consecutive harvests of first, second and third flushes of mushrooms, with values of 5.51, 5.22 and 4.73 × 10−6 g g−1 DM respectively. Kalac and Svoboda17 also reported that metal concentrations were highest in the initial harvest wave of A. bisporus. This declining trend of Cd concentration in the fruit bodies after consecutive harvests is a result of the decreasing Cd concentration in the substrates due to the harvesting of fruit bodies. Cd concentrations as affected by Agaricus blazei strains, body parts and rinsing Rinsing is an effective tactic to reduce the Cd concentration in the fruit bodies after harvest. After a quick rinse in water for 2 min to remove the scales, the Cd concentration in the fruit bodies was reduced from 10.6 to 7.70 × 10−6 g g−1 DM for the A. blazei strain N-2 and from 10.9 to 5.00 × 10−6 g g−1 DM for the strain A-6. In a similar study, Zrodlowski26 found that washing and peeling of A. bisporus decreased the contents of cadmium, lead, copper and zinc by 30–40%. Cadmium concentrations differ between body parts. The Cd concentrations in the cap and stalk of the strain N-2 were 14.3 and 6.20 × 10−6 g g−1 DM respectively. For the A. blazei strain A-6 the Cd concentration was 15.0 × 10−6 g g−1 DM in the cap and 5.00 × 10−6 g g−1 DM in the stalk. The uneven distribution of heavy metals within fruit bodies has also been reported by Kalac and Svoboda.17 They found that metal concentrations were highest in the sporophore (but not in the spores), lower in the rest of the cap and lowest in the stalk. Similarly, Melgar et al.27 reported that the Cd concentration in Agaricus macrosporus was about two times higher in the cap than in the other parts of the fruit body. Seeger15 found that the Cd concentration in the gills was five times higher than in the other parts. 1371 JC Huang et al. Relationship between Agaricus blazei yield and its components The fruit body dry weight per fruit of A. blazei decreased as the fruit body number m−2 increased, following a linear relationship between the reciprocal of fruit body dry weight and fruit body number m−2 (Fig. 1). Similar trends were found for the first and second harvests as well as for the two harvests combined. The total dry weight of fruit bodies m−2 can be fitted by a hyperbolic relationship to fruit body number m−2 by adopting a characteristic yield–density relationship described by Willey and Heath28 (Fig. 2). Although a lower number of fruit bodies m−2 resulted in a higher weight per single fruit body, a high fruit body yield m−2 can only be achieved by increasing the number of fruit bodies m−2 rather than the size (weight) of each individual fruit body. Effect of fruit body yield m−2 on Cd accumulation The Cd concentration in the fruit body decreased as the fruit body yield m−2 increased (Fig. 3), following a similar relationship between fruit body size (weight) and fruit body number m−2 to that shown in Fig. 1. The fitted parameters for Eqn (3) suggest that the relationship is valid when the fruit body yield is greater than 60 g m−2 . Within the experimental range of fruit body yields a relatively low Cd content (<6.0 mg kg−1 ) in the fruit body can be maintained when a fruit body yield higher than 110 g m−2 is achieved. Figure 2. Relationship between fruit body yield m−2 (Y) and fruit body number m−2 (ρ). The full line is fitted to the combined data of the two harvests by Y = Ymx ρ/(ρ0.5 + ρ), where Ymx = 187.75 and ρ0.5 = 28.08 (R2 = 0.69). Cd accumulation as affected by yield components Linear regression analyses showed that the Cd concentration in the fruit body can be well related to the yield components of A. blazei, i.e. fruit body Figure 3. Relationship between Cd concentration in fruit body (CCd ) and fruit body yield m−2 (Y). The full line is fitted to the combined data of the two harvests by CCd = Cmx /(Y0.5 + Y), where Cmx = 289.01 and Y0.5 = −59.20 (R2 = 0.87). Figure 1. Relationship between fruit body dry weight per fruit (w) and fruit body number m−2 (ρ). The full line is fitted to the combined data of the two harvests by w = wmx /(ρ0.5 + ρ), where wmx = 193.55 and ρ0.5 = 30.53 (R2 = 0.87). 1372 number m−2 , fruit body size (weight) and their interaction (Fig. 4). Fruit body number m−2 made the greatest positive contribution to Cd accumulation, while the interaction between fruit body size and fruit body number m−2 showed a smaller but negative effect. The results indicate that the decline in Cd concentration in the fruit body might be due to a dilution effect as a result of the increased number of fruit bodies per unit area. J Sci Food Agric 88:1369–1375 (2008) DOI: 10.1002/jsfa Cadmium accumulation in Agaricus blazei Murrill Figure 4. Relationship between predicted Cd concentration (CCd,prd ) and observed Cd concentration (CCd,obs ) in fruit body as a function of yield components of Agaricus blazei. CCd,prd is fitted by CCd,prd = E0 + E1 w + E2 ρ + E3 wρ, where E0 = 2.40, E1 = 3.87, E2 = 0.091 and E3 = −0.094. Cd accumulation as affected by substrate Cd concentration and fruit body size The Cd content in the substrates is the source for Cd accumulation in the fruit body. The Cd concentration in the fruit body can be well predicted as a function of the Cd concentration in the cultivation substrates (CCd,E ) together with the fruit body size (w). The Cd concentration in the fruit body can approach an asymptote of 13.4 mg kg−1 dry weight as the Cd concentration in the cultivation substrates approaches infinity (Fig. 5). The effect of fruit body size was expressed as a negative value of the coefficient β (−0.37), showing an increase in the Cd concentration in the fruit body as the size of the fruit body increased. The model for the Cd concentration in the fruit body as a function of the Cd concentration in the cultivation substrates and the size of the fruit body accounted for 87% of the variance in the observed Cd concentration in the fruit body. It is therefore imperative to control the overgrowth of individual fruit bodies to ensure the production of a large number of medium-sized and uniform fruit bodies. On the other hand, it is also critical to limit the initial Cd concentration in the substrates. Substrate composition is considered to be an important factor in determining heavy metal concentrations in fruit bodies.17 The Cd concentration in the fruit body increases as the Cd concentration in the substrates increases.29 Careful selection of suitable materials for substrate preparation is therefore an important step in reducing Cd concentrations in A. blazei. In the present study the initial Cd concentrations were estimated as 0.06, 0.35, 0.52 and 0.1 mg kg−1 in the rice straw, cow manure, calcium superphosphate (Ca(H2 PO4 )2 · H2 O) and casing soil used respectively. Potassium dihydrogen phosphate (KH2 PO4 ) and gypsum (CaSO4 · 2H2 O) were then J Sci Food Agric 88:1369–1375 (2008) DOI: 10.1002/jsfa Figure 5. Relationship between predicted Cd concentration (CCd,prd ) and observed Cd concentration (CCd,obs ) in fruit body as a function of substrate Cd content and fruit body size. CCd,prd is fitted by CCd,prd = CCd,mx CCd,E /(1 + αCCd,E + βw), where CCd,E is the substrate Cd content, w is the dry weight per fruit body, CCd,mx is the maximum Cd concentration in the fruit body, α is the coefficient for the effect of CCd,E and β is the coefficient for the effect of w. The parameter values are CCd,mx = 13.40, α = 2.16 and β = −0.37. used to replace calcium superphosphate, which contained relatively high concentrations of Cd and As. The usage of cow manure was also significantly reduced, from the commercial rate of 22–30% to a rate of 12%, but this had no significant impact on fruit body yield. The cultivation substrates prepared and used in the present research had a Cd concentration of around 0.17 mg kg−1 . The bioaccumulation factor (ratio of Cd in the fruit body to Cd in the substrates) was estimated as 27.8–32.4 in the present study. Kalac and Svoboda17 reported that the bioaccumulation factor of Cd ranged from 50 to 300 depending on the mushroom species. Cd accumulation and P uptake The Cd concentration in the fruit body was a modified exponential function of the P concentration in the fruit body (Fig. 6). The results suggest that Cd accumulation is highly associated with P uptake. However, there was no significant association between Cd accumulation and calcium content (data not shown). The positive relationship between Cd accumulation and P absorption suggests that it might be possible to limit Cd accumulation in the fruit body by reducing P input in the cultivation substrates. Meisch and Schmitt30 isolated a Cd–mycophosphatin organic complex from Agaricus macrosporus. The P-containing organic compound had a molecular weight of 12 000 Da. The presence of this Cd–P organic complex might contribute to the positive association of P and Cd. Higher P levels might also stimulate Cd accumulation and increase Cd tolerance in A. blazei. 1373 JC Huang et al. 2 3 4 5 6 7 8 Figure 6. Relationship between Cd concentration (CCd ) and P concentration (CP ) in fruit body: CCd = CCd,0 + CCd,k /[1 − γ exp(µCP )]. The coefficient values are CCd,0 = 3.55, CCd,k = 0.16, γ = 0.35 and µ = 0.10 (R2 = 0.93). In a separate study we found that there was a slight increase (2.2%) in the content of total amino acids as the Cd concentration in the fruit body increased (unpublished data). Among the 17 amino acids tested, arginine, proline, phenylalanine and alanine were the top four amino acids that were significantly increased. It is not clear if amino acids are involved in the chelation process of Cd. CONCLUSIONS Cadmium accumulation in A. blazei poses a potential health threat to consumers. The popular demand for A. blazei products has prompted the need to significantly reduce Cd accumulation in the harvested fruit bodies. The present research has identified a number of control measures that can be taken by local farmers in this key A. blazei-growing region of southeastern Fujian to effectively reduce the Cd concentration in fruit bodies: (1) limiting the initial Cd and P concentrations in the substrates, (2) rinsing the fruit bodies in water and (3) improving cultivation techniques to increase fruit body yields m−2 through the production of sufficient numbers of medium-sized fruit bodies. Overgrown fruit bodies should be avoided owing to the risk of high Cd accumulation. 9 10 11 12 13 14 15 16 17 18 19 20 ACKNOWLEDGEMENT The authors acknowledge funding support for the project 2003-N-011 from the Department of Science and Technology of Fujian, China. REFERENCES 1 Barbisan LF, Spinardi-Barbisan AL, Moreira EL, Salvadori DM, Ribeiro LR, da Eira F, et al., Agaricus blazei 1374 21 22 23 (Himematsutake) does not alter the development of rat diethylnitrosamine-initiated hepatic preneoplastic foci. Cancer Sci 94:188–192 (2003). Duan XP, Ma JF and Zhao XH, Advances in bioactive substances from Agaricus blazei Murrill. Anim Sci Vet Med 21:15–17 (2006). Firenzuoli F, Gori L and Lombardo G, The medicinal mushroom Agaricus blazei Murrill: review of literature and pharmaco-toxicological problems. Evid based Compl Altern Med 5:3–15 (2008). DOI: 10.1093/ecam/nem007. Tuzen M, Sesli E and Soylak M, Trace element levels of mushroom species from East Black Sea region of Turkey. Food Control 18:806–810 (2007). Chen TQ, Li KB and Huang DB, Introduction and successful cultivation of Agaricus blazei Murrill in Fujian. Edible Fungi China 13:41 (1994). He XJ, Yang PY and Chen TQ, Cultivation of medical mushroom Agaricus blazei Murrill in Fujian. Fujian J Agric Sci 14:56–58 (1999). Delmanto RD, de Lima PLA, Sugui MM, da Eira AF, Salvadori DMF, Speit G, et al, Antimutagenic effects of Agaricus blazei Murrill mushroom on the genotoxicity induced by cyclophosphamide. Mutat Res 496:15–21 (2001). Matuo R, Oliveira RJ, Silva AF, Mantovani MS and Ribeiro LR, Anticlastogenic activity of aqueous extract of Agaricus blazei in drug-metabolizing cells (HTCs) during cell cycle. Toxicol Mech Meth 17:147–152 (2007). Kawagishi H, Nomura A, Yumen T, Mizuno T, Hagiwara T and Nakamura T, Isolation and properties of a lectin from the fruiting bodies of Agaricus blazei. Carbohydr Res 183:150–154 (1988). Kawagishi H, Katsumi R, Sazawa T, Mizuno T, Hagiwara T and Nakamura T, Cytotoxic steroids from the mushroom Agaricus blazei. Phytochemistry 27:2777–2779 (1988). Osaki Y, Kato T, Yamamoto K, Okubo J and Miyazaki T, Antimutagenic and bactericidal substances in the fruit body of a Basidiomycete Agaricus blazei. Yakugaku Zasshi 114:342–350 (1994). Ohno N, Furukawa M, Miura NN, Adachi Y, Motoi M and Yadomae T, Antitumor β-glucan from the cultured fruit body of Agaricus blazei. Biol Pharmaceut Bull 24:820–828 (2001). Ellertsen LK, Hetland G, Johnson E and Grinde B, Effect of a medicinal extract from Agaricus blazei Murrill on gene expression in a human monocyte cell line as examined by microarrays and immuno assays. Int Immunopharmacol 6:133–143 (2006). Schmitt JA and Meisch HU, Cadmium in mushrooms, distribution growth effects and binding. Trace Elem Med 2:163–166 (1985). Seeger R, Toxische Schwermetalle in Pilzen. Dtsch Apotheker Zeitung 122:1835–1844 (1982). Xu CS, Effects of pernicious trace elements on human health. Guangdong Trace Elem Sci 6:1–3 (1999). Kalac P and Svoboda L, A review of trace element concentrations in edible mushrooms. Food Chem 69:273–281 (2000). Shukla GS and Singhal RL, The present status of biological effects of toxic metals in the environment: lead, cadmium, and manganese. Can J Physiol Pharmacol 62:1015–1031 (1984). Nordberg GF, Cadmium and health in the 21st century – historical remarks and trends for the future. Biometals 17:485–489 (2004). Waalkes MP, Cadmium carcinogenesis. Mutat Res 533:107–120 (2003). Wang H, Zhu G, Shi Y, Weng S, Jin T, Kong Q, et al, Influence of environmental cadmium exposure on forearm bone density. J Bone Miner Res 18:553–560 (2003). FAO/WHO, Joint FAO/WHO Expert Committee on Food Additives 61st Meeting, JECFA/61/SC. [Online]. (2003). Available: http://www.who.int/ipcs/food/jecfa/summaries/en/ summary 61.pdf. Accessed 28 March 2008. CNSSE (China National Standards Series Editor), China National Standards Compilation (2003 revision). Standards Press of China, Beijing (2003). J Sci Food Agric 88:1369–1375 (2008) DOI: 10.1002/jsfa Cadmium accumulation in Agaricus blazei Murrill 24 Thornley JHM and Johnson IR, Plant and Crop Modelling: a Mathematical Approach to Plant and Crop Physiology. Oxford University Press, New York, NY (1990). 25 Whistler D, White KJ, Wong SD and Bates D, SHAZAM User’s Reference Manual. Version 9.0. Northwest Econometrics, Vancouver (2001). 26 Zrodlowski Z, The influence of washing and peeling of mushrooms Agaricus bisporus on the level of heavy metals contamination. Pol J Food Nutr Sci 45:26–33 (1995). 27 Melgar MJ, Alonso M, Perez-Lopez M and Garcıa MA, Influence of some factors on toxicity and accumulation of cadmium J Sci Food Agric 88:1369–1375 (2008) DOI: 10.1002/jsfa from edible wild macrofungi in NW Spain. J Environ Sci Health B 33:439–455 (1998). 28 Willey RW and Heath SB, The quantitative relationships between plant population and crop yield. Adv Agron 21:282–321 (1969). 29 Si QQ, Lin L and Chen ZC, Studies on the accumulation of heavy metals and their effects on the growth and metabolism in edible fungi. Acta Mycol Sin 10:301–310 (1991). 30 Meisch HU and Schmitt JA, Characterization studies on cadmium-mycophosphatin from the mushroom Agaricus macrosporus. Environ Health Perspect 65:29–32 (1986). 1375 J Sci Food Agric 88:1376–1379 (2008) Journal of the Science of Food and Agriculture Comparison of postmortem changes in goose cardiac and breast muscles at 5 ◦C Sy-Shyan Ho, Chia-Ying Lin and Rong-Ghi R. Chou∗ Department of Animal Science, National Chiayi University, Chiayi City, Taiwan Abstract BACKGROUD: The tenderness of goose heart is an important consideration in its utilization as a popular meat product. It is generally thought that postmortem degradation of myofibrillar proteins may improved meat tenderness. Little information, however, is available regarding the postmortem changes in goose cardiac muscle. Therefore, the postmortem proteolysis between goose cardiac and breast muscles at 5 ◦ C is compared. RESULTS: The pH is higher (P < 0.05) in cardiac samples than in breast samples. Sodium dodecyl sulfate–polyacrylamide gel electrophoresis and western blots results indicate that postmortem degradation of titin and desmin and the appearance of the 28 and 30 kDa components are faster in breast muscle than in cardiac muscle. CONCLUSION: Our results may suggest that goose postmortem proteolysis occurs more rapidly in breast muscle than in cardiac muscle at 5 ◦ C.  2008 Society of Chemical Industry Keywords: white Roman goose; cardiac muscle; breast muscle; postmortem changes INTRODUCTION Poultry heart is considered as an edible by-product in poultry production, but its texture is generally perceived as tough by consumers. The tenderness of heart, therefore, becomes an important consideration in its utilization as a popular meat product. It is generally believed that postmortem proteolysis may improve the meat tenderness.1 The postmortem changes in skeletal muscle from various animal origins have been extensively studied.2 – 4 Degradation of myofibrillar proteins and disintegration of Z-lines are two main changes occurring in postmortem skeletal muscle (for reviews, see Robson et al.5 ). It has been suggested that µ-calpain is a key contributor in postmortem proteolysis in bovine muscle at 5 ◦ C (for reviews, see Koohmaraie and Geesink6 ). On the other hand, little information is available regarding the postmortem changes in goose cardiac muscle. The purpose of this study, therefore, is to compare the postmortem proteolysis between goose cardiac and breast muscles at 5 ◦ C. The changes in pH and degradation of myofibrillar proteins are examined. MATERIALS AND METHODS Sample preparation White Roman geese (Anser anser, ∼100 days old with an average live weight of 4.5 kg) hearts and carcasses (3–4 h postmortem) were obtained from a local slaughterhouse using standard commercial practices. Cardiac muscle from the left ventricle and breast muscle (pectoralis major) excised from the carcass of each bird were vacuum-packed and stored at 5 ◦ C. Samples were taken at 0, 1, 3, 7, and 14 days of storage. This experiment was replicated three times and 30 geese were used in each replication. Geese were randomly assigned to each of the five time periods, resulting in six geese per time. After sampling, the cardiac or breast muscle specimens from all six birds per time period were combined into an averaged sample and then ground through a 3 mm plate for pH measurement, sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and western blot analysis. The pH was determined by the method of Farouk and Swan7 and statistically analyzed by the split-plot design.8 The main plots were cardiac and breast muscles and the subplots were the meat specimens sampled at each time period. SDS-PAGE analysis Cardiac and breast myofibrils from the pooled samples were isolated via the method of Huff-Lonergan et al.9 The protein concentration of myofibrils was determined using a modified biuret method.10 The myofibril samples for SDS-PAGE were prepared by the method of Wang et al.11 SDS-PAGE was routinely analyzed in a 12% Tris–glycine slab gel ∗ Correspondence to: Rong-Ghi R. Chou, Department of Animal Science, National Chiayi University, 300 University Road, Chiayi City, 60083 Taiwan E-mail: chourg@mail.ncyu.edu.tw (Received 27 August 2007; revised version received 9 January 2008; accepted 10 January 2008) Published online 7 April 2008; DOI: 10.1002/jsfa.3227  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Postmortem changes in goose cardiac and breast muscles Western blot analysis Proteins were transferred from a 12% slab gel to a nitrocellulose membrane by the method of Towbin et al.13 After the transfer, the membrane was incubated in a 5% bovine serum albumin–phosphate buffer solution (BSA-PBS) for 30 min at 37 ◦ C and then washed three times in a 0.1% BSA-PBS solution for 5 min each time at room temperature. A monoclonal antibody (mAb) of desmin (Clone DE-U-10) (D1033, Sigma, St Louis, MO, USA) was used as a primary antibody. The membrane was incubated with the primary antibody for 2 h at room temperature, washed three times in 0.1% BSA-PBS for 5 min each time, incubated with immunogold-labeled secondary antibody for 2 h at room temperature, and washed twice in 0.1% BSA-PBS solution for 5 min each time and twice in deionized water for 1 min each. The gold label was enhanced by silver staining.14 the findings of Sekikawa et al.,15 who have shown that the final pH is higher in postmortem bovine cardiac than skeletal muscle. That cardiac muscle has a higher ultimate pH than skeletal muscle is likely due to cardiac muscle having a greater aerobic capacity than skeletal muscle, so that much of the postmortem metabolism in cardiac muscle occurs via the citric acid cycle and does not result in lactic acid production.16 Titin is first found in rabbit skeletal muscle17 and then in cardiac muscle.18 It has been reported that skeletal titin is degraded during postmortem storage.2,4,9 Our SDS-PAGE analysis of both CM and BM samples with an 8% gel shows that two closely spaced titin bands (T1 and T2) near the top of the gel are present in 0-day samples (lane 1 in Fig. 2). These migrations of titin bands in SDS-PAGE are typical, as described previously for rabbit skeletal muscle by Wang et al.17 Our results show that the T1 band remains visible in CM samples during the entire 14 days of storage at 5 ◦ C (Fig. 2) and that little degradation of titin is found in CM samples. Locker and Wild19 have also found no degradation of titin in ox and chicken hearts stored at 15 ◦ C for 24 h. In BM samples, however, the T1 band is visible on days 0 and 1, becomes faintly seen by day 3, and nearly disappears by day 7 (Fig. 2(B)). Our results show that titin degrades more rapidly in BM than in CM samples. One of the most noticeable changes during postmortem aging is the appearance of a ∼30 kDa 6.5 6.3 pH (acrylamide:methylenebisacrylamide was 37.5:1, w/w) for proteins migrating below myosin heavy chains according to the method of Laemmli.12 Degradation of titin was examined in an 8% Tris–glycine slab gel (acrylamide:methylenebisacrylamide 200:1, w/w).11 The same amount of protein (150 µg) from each sample was loaded on each well of the 12% and 8% gels. Molecular weight markers ranging from 14 400 to 97 000 (Amersham Biosciences Ltd, Piscataway, NJ, USA) were used as protein standards. All gels were run at 15 mA at 25 ◦ C. SE 400 slab gel electrophoresis units (Hoofer Scientific Instruments, San Francisco, CA, USA) were used. Gels were stained in a solution of 0.05% (w/v) Compassion Blue R-250, 45% (v/v) methanol and 9.2% (v/v) acetic acid for 4 h and distained in 10% (v/v) methanol and 7.5% (v/v) acetic acid. 6.0 5.8 RESULTS AND DISCUSSION Figure 1 shows the changes in pH of cardiac (CM) and breast (BM) muscle samples during 14 days of storage at 5 ◦ C. The overall average pH is significantly higher (P < 0.05) in CM (6.12 ± 0.06) than in BM (5.92 ± 0.04) samples. This result was consistent with 5.5 0 2 4 6 8 10 Day of Storage 12 14 Figure 1. Changes in the pH of goose cardiac and breast muscles at 5 ◦ C. Standard error of means (SEM) = 0.06. , cardiac samples; , breast samples. ž Figure 2. Changes in titin of goose cardiac (A) and breast (B) muscles at 5 ◦ C. These gels are representative of three replications of combined samples. T1, titin 1; T2, titin 2; N, nebulin; M, myosin heavy chains. Lane 1 = 0 days; lane 2 = 1 day; lane 3 = 3 days; lane 4 = 7 days; lane 5 = 14 days. J Sci Food Agric 88:1376–1379 (2008) DOI: 10.1002/jsfa 1377 S-S Ho, C-Y Lin, R-GR Chou component, a degradation product of troponin-T.20 The 12% gels of CM samples apparently show that little changes in electrophoresis patterns can be found in the molecular weight region between 30 and 40 kDa in CM samples during the entire 14 days of storage at 5 ◦ C (Fig. 3(A)). The 30 and 28 kDa bands in BM samples, on the other hand, are faintly seen by day 0 and day 3, respectively, and become more apparent by days 7 and 14 (Fig. 3(B)). These results indicate that the appearance of the 30 kDa component occurs more rapidly in BM samples than in cardiac samples. However, the molecular weight of cardiac troponin-T is higher than that of skeletal troponin-T in mammals.21 This implies that the degradation products of goose cardiac troponin-T may be larger than 30 kDa. More studies are therefore needed to identify the degradation products of cardiac troponin-T. Desmin is known as a major component of desmincontaining intermediate filaments in muscle cells.22 Several studies have reported that desmin degrades in postmortem skeletal muscles.23 Our western blots labeled with an mAb to desmin demonstrate that intact desmin remains unchanged in CA samples (Fig. 4(A)). In BM samples (Fig. 4(B)), however, intact desmin decreases on day 3 and completely disappears by day 7. These results, again, indicate that degradation of desmin is more rapid in BM samples than in CM samples. Our results show that postmortem degradation of titin and desmin and appearance of ∼30 kDa components occur more rapidly in BM samples Figure 3. Changes in myofibrillar proteins of goose cardiac (A) and breast (B) muscles at 5 ◦ C. These gels are representative of three replications of combined samples. M, myosin heavy chains; αA, α-actinin; A, actin; 30K, 30 kDa component; 28K, 28 kDa component; f, dye front. Lane 1 = 0 days; lane 2 = 1 day; lane 3 = 3 days; lane 4 = 7 days; lane 5 = 14 days; lane S = standard proteins; 97K = phosphorylase b; 66K = bovine serum albumin; 30K = carbonic anhydase; 20K = trypsin inhibitor; 14K = α-lactalbumin (14.4 kDa). Figure 4. Western blotting showing the changes in desmin of goose cardiac (A) and breast (B) muscles at 5 ◦ C. These gels are representative of three replications of combined samples. D, Desmin. Lane 1 = 0 days; lane 2 = 1 day; lane 3 = 3 days; lane 4 = 7 days; lane 5 = 14 days. 1378 J Sci Food Agric 88:1376–1379 (2008) DOI: 10.1002/jsfa Postmortem changes in goose cardiac and breast muscles than in CM samples. It has been reported that µ-calpain is responsible for postmortem proteolysis in bovine muscle at 5 ◦ C,6,24 and titin, desmin and troponin-T can be degraded by µ-calpain.25 Huff-Lonergan and Lonergan26 have also proposed that calpain activation may explain the variation of desmin postmortem degradation. This may imply that the calpain activation is more rapid in BM than in CM samples. On the other hand, our results show that CM sample (>6.0) has a higher ultimate pH than SM sample (<6.0). It may be assumed that the calpains in CM sample are more active than those in SM sample and favor more extensive postmortem proteolysis. However, it has been shown that little µ-calpain, which is essential for postmortem proteolysis, is observed in chicken (avian) heart.27 Therefore, the difference between skeletal muscle and cardiac muscle is probably more important than the observed difference in final pH between the two muscles examined. Thus, our results show that goose postmortem proteolysis occurs more rapidly in BM than in CM at 5 ◦ C. In summary, the postmortem changes in goose CM and BM at 5 ◦ C are compared in the present studies. The results show that the pH is higher (P < 0.05) in CM than in BM. The SDS-PAGE and western blot results indicate that postmortem degradation of titin and desmin and appearance of the 28 and 30 kDa components are more rapid in goose BM than in CM. Therefore, our results may suggest that goose postmortem proteolysis occurs more rapidly in BM than in CM at 5 ◦ C. REFERENCES 1 Hopkins DL and Thompson JM, Factors contributing to proteolysis and disruption of myofibrillar proteins and the impact on tenderization in beef and sheep meat. Aust J Agric Res 53:149–166 (2002). 2 Ho C-Y, Stromer MH and Robson RM, Effect of electrical stimulation on postmortem titin, nebulin, desmin and troponin-T degradation and ultrastructural changes in bovine longissimus muscle. J Anim Sci 74:1563–1575 (1996). 3 Wheeler TL, Shackelford SD and Koohmaraie M, Variation in proteolysis, sarcomere length, collagen content, and tenderness among major pork muscles. J Anim Sci 78:958–965 (2000). 4 Tsai S-F, Lin C-Y, Lu J-J and Chou R-GR, Postmortem proteolysis of breast and leg muscles from Taiwan colored chickens and Silkie bantam. Asia Aust J Anim Sci 19:739–743 (2006). 5 Robson RM, Huff-Lonergan E, Parrish FC Jr, Ho C-Y, Stromer MH, Huiatt TW, et al, Postmortem changes in the myofibrillar and other cytoskeletal proteins in muscle. Proc Recip Meat Confer 50:43–52 (1997). 6 Koohmaraie K and Geesink GH, Contribution of postmortem muscle biochemistry to the delivery of consistent meat quality with particular focus on the calpain system. Meat Sci 74:34–43 (2006). 7 Farouk MM and Swan JE, Acceptability and functional properties of restructured roast from frozen pre-rigor injected beef. Meat Sci 46:57–66 (1997). J Sci Food Agric 88:1376–1379 (2008) DOI: 10.1002/jsfa 8 SAS Institute, User Guide: Statistics, version 6. SAS Institute, Cary, NC (1986). 9 Huff-Lonergan E, Parrish Jr FC and Robson RM, Effects of postmortem aging time, animal age, and sex on degradation of titin and nebulin in bovine longissimus muscle. J Anim Sci 73:1064–1073 (1995). 10 Robson RM, Goll DE and Temple MJ, Determination of protein in ‘Tris’ buffer by the biuret reaction. Anal Biochem 24:339–341 (1968). 11 Wang S-M, Greaser ML, Schultz E, Bulinski JC, Lin JJ-C and Lessard JL, Studies on cardiac myofibrillogenesis with antibodies to titin, actin, tropomyosin, and myosin. J Cell Biol 107:1075–1083 (1988). 12 Laemmli UK, Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227:680–685 (1970). 13 Towbin H, Staehelin T and Gordon J, Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheet: procedure and some applications. Proc Natl Acad Sci USA 76:4350–4354 (1979). 14 Moeremans M, Daneels G, De Raeymaeker M, De Wever B and De Mey J, The use of colloidal gold particles for testing the specificity of antibodies and/or the presence of antigen, in Immuno-Gold Labeling in Cell Biology, ed. by Verkleij AJ and Leuniissen JLM. CRC Press, Boca Raton, FL, pp. 17–27 (1989). 15 Sekikawa M, Shimada K, Fukushima M, Ishikawa T, Wakamatsu J and Mikami M, Presence of ubiquitin in bovine post-mortem cardiac muscle. Food Chem 69:315–318 (2000). 16 Katz AM, Contractile proteins of heart. Physiol Rev 50:63–158 (1970). 17 Wang K, McClure J and Tu A, Titin: major myofibrillar components of striated muscle. Proc Natl Acad Sci USA 76:3698–3702 (1979). 18 Hu DH, Kimura S and Maruyama K, Sodium dodecyl sulfate gel electrophoresis studies of connectin-like high molecular weight proteins of various types of vertebrate and invertebrate muscles. J Biochem 99:1485–1492 (1986). 19 Locker RH and Wild DJ, A comparative study of huge molecular weight proteins in various muscles across the animal kingdom. J Biochem 99:1473–1484 (1986). 20 Ho C-Y, Stromer MH and Robson RM, Identification of the 30 kDa polypeptide in post mortem skeletal muscle as a degradation product of troponin-T. Biochimie 76:369–375 (1994). 21 Perry SV, Troponun-T: genetics, properties and function. J Muscle Res Cell Motil 19:575–602 (1998). 22 Lazarides E, Intermediate filaments: a chemically heterogeneous, developmentally regulated class of proteins. Annu Rev Biochem 51:219–250 (1982). 23 Hwan S-F and Bandman E, Studies of desmin and α-actinin degradation in bovine semitendinosus muscle. J Food Sci 54:1426–1430. 24 Koohmaraie M, Seideman SC, Schollmeyer JE, Dutson TR and Crouse JD, Effect of post-mortem storage on Ca++ dependent proteases, their inhibitor and myofibril fragmentation. Meat Sci 19:187–196 (1987). 25 Huff-Lonergan E, Mitsuhashi T, Beekman DD, Parrish FC Jr, Olson DG and Robson RM, Proteolysis of specific muscle structural proteins by µ-calpain at low pH and temperature is similar to degradation in postmortem bovine muscle. J Anim Sci 74:933–1008 (1996). 26 Huff-Lonergan E and Lonergan SM, Mechanism of watercapacity of meat: the role of postmortem biochemical and structural changes. Meat Sci 71:194–204 (2005). 27 Lee HL, Sante-Lhoutellier V, Vigouroux S, Briand Y and Briand M, Calpain specificity and expression in chicken tissue. Comp Biochem Physiol B 146:88–93 (2007). 1379 Journal of the Science of Food and Agriculture J Sci Food Agric 88:1380–1384 (2008) Effects of pressure toasting on in situ degradability and intestinal protein and protein-free organic matter digestibility of rapeseed Arash Azarfar,1∗ Claudio S Ferreira,2 Jacob O Goelema3 and Antonius FB Van der Poel4 1 Faculty of Agriculture, University of Lorestan, PO Box 465, Khorramabad, Iran do Frade 54, Chaneca, Casalinhos de Alfaite, 2560 Torres Vedras, Portugal 3 De Heus Voeders BV, Rubensstraat 175, 6717 VE, Ede, The Netherlands 4 Animal Nutrition Group, Department of Animal Sciences, Wageningen University, Marijkeweg 40, 6709 PG Wageningen, The Netherlands 2 Avenu Abstract BACKGROUND: Rapeseed is a protein supplement that contains up to 40% crude protein (CP) on a dry matter (DM) basis, but a large part of its protein can be easily degraded in the rumen. Therefore, before inclusion in ruminant’s diet, the extent of its protein degradation in the rumen must be reduced without altering its intestinal digestibility. A study was conducted to investigate the effects of pressure toasting (T, 130 ◦ C) at two residence times (1.5 and 10 min) alone or in combination with soaking in water (ST, 4 h) on ruminal degradability and intestinal digestibility of CP and protein-free organic matter (PFOM) in whole full-fat rapeseed. RESULTS: Regardless of the processing time (1.5 or 10 min), T significantly (P < 0.05) increased the fraction of undegraded intake protein (UIP) compared to the untreated rapeseed samples. Soaking prior to further toasting did not improve the rumen degradation characteristics of rapeseed CP. Compared to the untreated rapeseed samples, both T and ST significantly (P < 0.0001) improved the true protein digested in the small intestine (DVE) and degraded protein balance (OEB), effects that were more evident in samples heated for 10 min. Soaking prior to pressure toasting, however, did not further improve the DVE or OEB in the rapeseed samples in comparison with T treatment. CONCLUSIONS: It was concluded that ruminal protein degradability of rapeseed decreased after pressure toasting, without seriously affecting its intestinal digestibility.  2008 Society of Chemical Industry Keywords: Brassica napus; intestinal digestibility; pressure toasting; rumen degradability; soaking INTRODUCTION In European countries, a large part of the protein sources used in animal nutrition is imported from abroad. Home-grown protein sources (e.g., seeds as peas, faba beans, rapeseed, lupins) are available, but because of the high rumen degradability of their proteins and starches these feedstuffs are not regularly included in compound feeds for dairy cattle. Various chemical and physical methods for the processing of feedstuffs have been proposed to improve the degradability characteristics of protein sources for ruminants.1 One of the possibilities to reduce degradation of crude protein (CP) in the rumen is thermal processing.2 Heat treatments have been reported to reduce ruminal degradability of CP with a concomitant increase in the lower gastrointestinal tract digestibility.3 The effect of heat treatment is, amongst others, a function of the temperature reached and the duration of exposure.1 According to Aguilera et al.,4 heat treatment may result in improved weight gain, feed efficiency and nitrogen retention in ruminants. Heat treatment may also destroy a number of antinutritional factors in some legume seeds, an effect that may have a beneficial effect on digestion in the small intestine.4 Rapeseed is a protein supplement that contains up to 40% CP on a dry matter (DM) basis. However, when rapeseed meal is fed to ruminants, a large part of the protein is degraded in the rumen.5 Moreover, if rapeseed protein is to be used more efficiently in ruminant nutrition, for instance in the diets of rapidly growing calves and high-yielding dairy cows, the extent of protein degradation in the rumen must be reduced without altering its intestinal digestibility. ∗ Correspondence to: Arash Azarfar, Faculty of Agriculture, University of Lorestan, PO Box 465, Khorramabad, Iran E-mail: Arash.Azarfar@gmail.com (Received 27 June 2007; revised version received 9 January 2008; accepted 9 January 2008) Published online 18 April 2008; DOI: 10.1002/jsfa.3228  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Processing of rapeseed Pressure toasting as an available technology has been proposed to improve the nutritive value of legume seeds. In former experiments it was shown that pressure toasting reduced rumen protein degradability of peas, lupins and faba beans.6 McKinnon et al.2 investigated the effect of dry heat treatment of canola meal using a vacuum tumble dryer at 125 or 145 ◦ C for 10, 20 or 30 min on ruminal, intestinal and total tract disappearance of the dry matter (DM) and crude protein (CP) fractions of canola meal. They concluded that heating canola meal to a temperature of 145 ◦ C reduced ruminal and total tract availability of the DM and CP fractions. Heating to 125 ◦ C for 10, 20 or 30 min reduced rumen disappearance of CP over the total tract of the ruminant. Limited work has been done on the effects of soaking before the toasting of rapeseed, on the degradability of protein in the rumen. Therefore, in this study, samples of untreated rapeseed will be compared with soaked/toasted or toasted samples to determine the effects of the treatments on rumen degradation, as well as on intestinal digestibility of CP and protein-free organic matter (PFOM). MATERIALS AND METHODS Experimental setup The effects of using pressure toasting at different residence time (1.5 versus 10 min) alone or in combination with prior soaking on the rumen protein and PFOM degradability and intestinal digestibility in the samples of whole, full-fat rapeseed (Brassica napus) were studied. The treatments were untreated whole full-fat rapeseed (C), toasted rapeseed at 130 ◦ C for 1.5 min (T 130/1.5), toasted rapeseed at 130 ◦ C for 10 min (T 130/10), soaked (4 h) and subsequently toasted rapeseed at 130 ◦ C for 1.5 min (ST 130/1.5) and soaked (4 h) and subsequently toasted rapeseed at 130 ◦ C for 10 min (ST 130/10). Treatments (from the same batch) were repeated on two consecutive days. Rapeseed processing and sample preparation Whole, full-fat rapeseed (Brassica napus) of French origin was supplied by a commercial supplier (Agrifirm, Meppel, The Netherlands). Processing was carried out at the Wageningen Feed Processing Centre using a small-scale pressurized toaster, as described by Van der Poel.7 Throughout the experiment, the toaster was used for a batch-type direct steam heating of 2–2.5 kg of the samples. The speed of the conveyor was adjusted to achieve the required residence times (1.5 or 10 min). For the soaking/toasting treatment, samples were soaked in tap water for 4 h prior to sieving in a tray to remove the water prior to toasting. The samples were then dried in a forced oven for 16 h at 35 ◦ C prior to grinding (Retsch ZM100, GmbH & Co., Hanover, Germany; 3 mm screen) and storage at 4 ◦ C. J Sci Food Agric 88:1380–1384 (2008) DOI: 10.1002/jsfa In situ incubations Rumen incubations were carried out according to Dutch standard method8 with four rumen-cannulated lactating Holstein–Friesian cows. The cows received about 17 kg dry matter (DM) daily of a ration consisting of grass silage (65% of DM intake) and a commercial concentrate (90 g intestinal absorbable protein and 6.5 MJ NEL kg−1 ). To incubate the samples, 5 g DM of each ground sample (Retsch ZM100 centrifugal mill; 3 mm) were weighed into pre-weighed nylon bags (10 × 17 cm; pore size 40 µm; PA 40/30, Nybolt, Zurich, Switzerland) and incubated in the rumen for 0, 4, 8, 12, 48 and 120 h. The all-out method was applied. For incubation times 4 and 8 h two bags, and for 12 and 48 h three bags, of each sample were incubated in the rumen of each cow. All treatments were randomly assigned to cows. After incubation, bags were immediately immersed in ice water and rinsed with tap water to stop the fermentation and the residues were pooled per time per treatment. The bags were washed using a domestic washing machine for 52 min with 70 L of cold tap water, without centrifuging. The bags were dried in a forcedair oven at 60 ◦ C for 24 h, air equilibrated and weighed. Residues from the bags were then pooled within time and treatment and ground prior to further chemical analysis. Intestinal incubations Intestinal digestibility of crude protein and PFOM was measured as described by Goelema et al.9 Two non-lactating Holstein cross Friesian cows, fitted with a cannula in the proximal duodenum, were used to measure intestinal protein digestibility. The cows received a daily ration consisting of 22.5 kg maize silage and 7.5 kg grass silage. The feed was offered at 06.00 h (40%) and 18.00 h (60%). Nylon with a pore size of 40 µm (PA 40/30, Nybolt) was used to prepare bags with an inner size of 3 × 7 cm. The bags were filled with approximately 0.5 g DM of the 12 h rumen incubation residue. Prior to incubation, the rumen incubation residue was prepared and handled as described above, but with freeze-drying instead of oven-drying. Prior to incubation in the proximal duodenum the bags were incubated in a solution containing 1 g (2000 FIP U g−1 ) pepsin in 1 L of 0.1 mol L−1 HCl at 37 ◦ C for 1 h. Three bags were inserted into the proximal duodenum through the cannula of each cow after every 5 min for a period of 7 h. After insertion of 12 bags, a 20 min break was taken, after which the procedure continued. Bags were retrieved from the faeces every 2 h and stored at −20 ◦ C until all the bags had been recovered. After thawing, the bags were rinsed, washed and freeze dried as described above. Residues were pooled within treatment and ground through a 1 mm sieve. 1381 A Azarfar et al. Chemical analysis Rapeseed samples were analysed for DM, Ash, CP, neutral detergent fibre (NDF) and acid detergent fibre (ADF) as described by Goelema et al.9 Calculations of ruminal degradability and intestinal digestibility Both CP and PFOM were classified into three fractions: a washable fraction (W), the fraction that disappears from the nylon bags after washing; an undegradable fraction (U), measured as the asymptote of the degradation curve at infinite time; and a potentially degradable fraction (D) measured as 1000W-U. The fractional rate of degradation of D fraction of CP and PFOM (kd , h−1 ) was measured using a first-order kinetic degradation model.10 Undegraded intake CP (UIP, as g kg−1 of CP) and undegraded intake of PFOM (UIPFOM, as g kg−1 of PFOM) were calculated assuming a ruminal outflow rate (kp , h−1 ) of 0.06 h−1 as described by Goelema et al.9 The residues of CP and PFOM after intestinal incubations were used to calculate intestinal digestibility of UIP (DUP, as g kg−1 of UIP) and intestinal UIPFOM (DUPFOM, as g kg−1 of UIPFOM). True protein digested in the small intestine (DVE, g kg−1 DM) and degraded protein balance (OEB, g kg−1 DM) were calculated as described by Tamminga et al.11 Statistical analysis The fractional rate of degradation and the U fraction were estimated with the NLIN procedure in SAS 9.1 (SAS/STAT package, SAS Institute Inc., Cary, NC, USA). Analysis of variance was conducted using the GLM procedure of SAS 9.1, with treatment (four replicates in the in situ incubations and two replicates in the intestinal incubations) and treatment day effect as independent variables in the model. Differences between the treatment means were analysed by a multiple comparison test (Tukey/Kramer), and the least square means were listed using the LSMEANS statement of SAS 9.1. Differences among treatments were separated by contrast statements, using the GLM procedure of SAS 9.1. RESULTS Chemical analysis The DM, ash, CP, NDF and ADF content of the untreated whole, full-fat rapeseed sample were 956.7 (g kg−1 ), 43.1, 205.4, 155.5 and 141.0 g kg−1 DM, respectively. Compared to the reference tabulated values (DM, ash, CP, NDF and ADF were 923.0 (g kg−1 ), 39.0, 198.0, 198.0 and 181.0 g kg−1 DM, respectively)12 the rapeseed samples used in the current study contained a higher organic matter and CP whereas the NDF and ADF contents were lower. Rumen degradation and intestinal digestion of CP and PFOM In Tables 1 and 2, rumen CP and PFOM degradation and intestinal digestion characteristics of untreated and treated rapeseed samples are presented. Significant differences were observed between C and other treatments with regard to U of CP (135.0 versus 179.0 g kg−1 CP, P < 0.01) and UIP (387.0 versus 425.0 g kg−1 , P < 0.05). Residence time (1.5 and 10 min) had no significant effect on rumen degradation and intestinal digestion of CP (Table 1). Regardless of the residence time, differences were observed between T and ST with respect to U of CP (T 2.5% lower than ST, P < 0.05), kd of CP (T 2.6% h−1 lower than ST, P < 0.01) and UIP (T 3.5% higher than ST). Compared to the C, ST for 1.5 min significantly decreased the D fraction of CP. Toasting (T and ST) regardless of residence time significantly increased the DVE in the samples of rapeseed. The effect of toasting, however, was more profound in those samples which were heated for 10 min compared to those which were heated for 1.5 min (116.8 and 121.1 g kg−1 DM in 1.5 and 10 min, respectively; P < 0.01). As opposed to DVE, toasting drastically decreased (77.5%) OEB. However, like DVE, the effect of toasting on decreasing OEB was more profound in those samples which were heated for 10 min. Statistical analysis revealed that, compared to the C, both T and ST increased the W and U of the PFOM (9.4% and 11.7%, respectively) and increased Table 1. Ruminal degradation characteristics and intestinal digestion of crude protein (CP) in untreated and treated samples of rapeseed Treatment W U D kd UIP DUP DVE OEB Significance (P) C T 130/1.5 ST 130/1.5 T 130/10 ST 130/10 SEM Controls vs. others 1.5 vs. 10 T vs. ST 220.0a 135.0b 645.0a 0.094a 387.0c 773.0a 60.7c 102.8a 270.0a 167.0a 563.0ab 0.064c 440.0ab 785.0a 116.4b 23.5bc 278.0a 194.0a 528.0b 0.090ab 405.0c 781.0a 117.1b 26.4b 234.0a 166.0a 600.0ab 0.069bc 445.0a 787.0a 122.5a 19.7c 224.0a 189.0a 587.0ab 0.099a 411.0bc 777.0a 119.6ab 23.2bc 18.2 7.6 25.1 0.059 7.9 4.4 0.9 1.2 NS <0.01 NS NS <0.05 NS <0.0001 <0.0001 NS NS NS NS NS NS <0.01 <0.05 NS <0.05 NS <0.01 <0.05 NS NS NS W, washable fraction (g kg−1 CP); U, undegradable fraction (g kg−1 CP); D, potentially degradable fraction (g kg−1 CP); kd , fractional rate of degradation of D (h−1 ); UIP, undegraded intake protein (g kg−1 CP); DUP, intestinal digestibility (g kg−1 UIP); DVE, true protein digested in the small intestine (g kg−1 DM); OEB, degraded protein balance (g kg−1 DM); SEM, standard error of least square mean; NS, not significant. Means with different letters within a row differ significantly (P < 0.05). 1382 J Sci Food Agric 88:1380–1384 (2008) DOI: 10.1002/jsfa Processing of rapeseed Table 2. Ruminal degradation characteristics and intestinal digestion of protein-free organic matter (PFOM) in untreated and treated samples of rapeseed Treatment W U D kd UIPFOM DUPFOM Significance (P) C T 130/1.5 ST 130/1.5 T 130/10 ST 130/10 SEM Controls vs. others 1.5 vs. 10 T vs. ST 203.0b 187.0b 610.0a 0.056b 502.0a 518.0c 324.0a 323.0a 353.0b 0.063b 499.0a 558.0bc 319.0a 317.0a 364.0b 0.079ab 473.0ab 631.0a 277.0ab 286.0a 437.0ab 0.067ab 489.0a 617.0ab 266.0ab 289.0a 445.0ab 0.107a 450.0b 644.0a 25.2 19.3 44.0 0.011 9.6 18.1 <0.05 <0.01 <0.05 NS NS <0.01 NS NS NS NS NS NS NS NS NS NS <0.05 <0.05 W, washable fraction (g kg−1 PFOM); U, undegradable fraction (g kg−1 PFOM); D, potentially degradable fraction (g kg−1 PFOM); kd , fractional rate of degradation of D (h−1 ); UIPFOM, undegraded intake PFOM (g kg−1 of PFOM); DUPFOM, intestinal digestibility (g kg−1 UIPFOM); SEM, standard error of least square mean; NS, not significant. Means with different letters within a row differ significantly (P < 0.05). DUPFOM (9.5%), whereas D was decreased by 21% (Table 2). Residence time had no significant effect on ruminal degradation characteristics and intestinal digestion of PFOM in the rapeseed samples. Compared to the T samples, a decreased UIPFOM was observed for ST samples which was caused by a numerically decreased kd . However, intestinal digestibility of UIPFOM was higher in the ST samples than in the T samples. DISCUSSION Several studies have been carried out to investigate the influence of heat treatment on rumen degradability and intestinal digestibility of diet ingredients.13 Goelema et al.6 studied the influence of pressure toasting and flaking on nutritive value of lupins (Lupinus albus) and rapeseed (Brassica napus) for dairy cows. They concluded that pressure toasting is a suitable method to improve nutritive value of lupins, but for rapeseed steam toasting, either at 100 ◦ C for 20 min or at 130 ◦ C for 1.5 min, did not improve the protein value of rapeseed since the fractional rate of degradation of the nitrogen was increased. In contrast with their results, our findings show, that regardless of residence time, toasting significantly decreased the kd of whole full-fat rapeseed (Table 1). In agreement with our findings, Sadeghi and Shawrang14 reported a reduced kd when canola meal samples were subjected to microwave irradiation for 2, 4 and 6 min. Kaldmäe et al.15 also observed that heat treatment of rape seed cake (100 ◦ C for 15–20 min) decreased effective degradability of protein. However, when the rapeseed samples were soaked in water prior to toasting no significance differences were observed between C and heat-treated (ST) samples with regard to kd . In contrast to findings of Goelema et al.9 with pressuretoasted peas, lupins and faba beans, toasting and soaking before toasting increased the washable fraction of CP (numerically) and PFOM (significantly) in our study. These changes may be attributed to the added water, which may have created a more favourable environment for the endogenous enzymes present in the plant material. According to Davies et al.,16 J Sci Food Agric 88:1380–1384 (2008) DOI: 10.1002/jsfa endogenous enzyme activity may be enhanced in a liquid environment, thereby increasing nutrient availability to the animal. Compared to the C, UIP was increased by both the T and ST (for 1.5 or 10 min) in the current study, which is in agreement with the results of Goelema et al.9 The elevated UIP due to the T process was mainly due to a lowered D (with a lower kd ) and an increased U fraction. Moshtaghi Nia and Ingalls17 investigated the effect of heating on canola meal protein degradation in the rumen and digestion in the lower gastrointestinal tract of steers. They reported a lower ruminal degradability and a higher intestinal disappearance after the treatments. According to these authors, heat treatment of protein sources decreases the ruminal degradability of CP by creating cross-linkages between peptide chains and carbohydrates, thereby increasing their resistance to proteolysis in the rumen. However, in the current study the effect of pressure toasting in increasing UIP was more profound when toasting was applied without soaking. Rape seed proteins consist of two subunits: napin (2S albumin) and cruciferin (12S globulin).14 Napin is an albumin that is characterized with a high water solubility and ruminal degradability. The cruciferin subunits have a lower water solubility compared to the napin subunits, and degrade slower in the rumen than dose napin. It is likely that during soaking the napin subunits were washed out of the samples, which may explain the significantly lower UIP in the ST sample compared to that of the T samples. Indeed, the lower UIP in the ST samples was concomitant with a lower W of CP (Table 1). Despite an elevated UIP, intestinal digestibility of UIP was not decreased by T and ST, resulting in a higher DVE in these samples compared to the C samples. Concomitantly, both T and ST with a more profound effect for 10 min processing significantly improved the OEB, which implies a shift from rumen degradation to intestinal digestion of CP. This was in the line with findings of Goelema et al.,9 who found a drastically lower OEB in the samples of pressure toasted (136 ◦ C; 3, 7 and 15 min) faba beans, peas and lupins compared to the untreated samples. 1383 A Azarfar et al. Although compared with untreated samples, pressure toasting (T) irrespective of residence time increased the W and U fractions of PFOM, but UIPFOM was not affected by these treatments. This effect can be explained by a significantly lower D in the T samples compared to the untreated samples. Maillard polymerization thought to be responsible for decreased D of PFOM after toasting.18 Except in pressure-toasted samples at 130 ◦ C for 1.5 min, pressure toasting alone or in combination with soaking significantly improved the intestinal digestibility of PFOM compared to the C samples. It is documented that when sufficient enzymatic activity for hydrolysis of PFOM is provided in the duodenum, an increase of undegraded PFOM after toasting improves the nutritive value of the feedstuffs.19 However, in the current study the improved intestinal digestibility of PFOM due to the pressure toasting was not concomitant with an increased UIPFOM. CONCLUSIONS Soaking prior to further toasting did not enhance the intestinal digestibility of UIP in rapeseed samples; therefore, it is not recommended. The lack of significant differences between processing times (1.5 versus 10 min) with regard to ruminal degradation characteristics of CP and PFOM suggest that toasting can be regarded as a high-temperature short-time technological process in this respect. Although the processing time (1.5 versus 10 min) did not have any effects on ruminal degradation characteristics of either CP or PFOM, OEB and DVE were improved by increasing the processing time during toasting from 1.5 to 10 min. Pressure toasting irrespective of residence time improved OEB and DVE; therefore, it is recommended to improve the protein quality of rapeseed for ruminants. REFERENCES 1 Waltz DM and Stern MD, Evaluation of various methods for protecting soya bean protein from degradability by rumen bacteria. Anim Feed Sci Technol 25:111–122 (1989). 2 McKinnon JJ, Olubobokun JA, Mustafa A, Cohen RDH and Christensen DA, Influence of dry heat treatment of canola meal on site and extent of nutrient disappearance in ruminants. Anim Feed Sci Technol 56:243–252 (1995). 3 Arieli A, Ben-Moshe A, Zamwel S and Tagari H, In situ evaluation of the ruminal and intestinal digestibility of heat treated whole cotton seed. J Dairy Sci 72:1228–1233 (1989). 1384 4 Aguilera JF, Bustoa M and Molina E, The degradability of legume seed meals in the rumen: effect of heat treatment. Anim Feed Sci Technol 36:101–112 (1992). 5 De Boer G, Murphy JJ and Kennely JJ, Mobile nylon bag for estimating intestinal digestibility of rumen undegradable protein. J Dairy Sci 70:977–982 (1987). 6 Goelema JO, Hof G, Tamminga S and Van der Poel AFB, Influence of pressure toasting of lupins and rapeseed and subsequent expander treatment and/or pelleting on rumen degradation and intestinal digestibility of concentrate for dairy cows. Internal report, Wageningen Institute of Animal Sciences, The Netherlands (1995). 7 Van der Poel AFB, Effects of processing on bean (Phaseolus vulgaris L.) protein quality. PhD thesis, Wageningen University (1990). 8 CVB, Voorloping protocol voor In situ pensincubatie voor het meten van eiwitbestendigheide, Centraal Veevoeder Bureau, Lelystad, NL (2003). 9 Goelema JO, Spreeuwenberg MAM, Hof G, van der Poel AFB and Tamminga S, Effect of pressure toasting on the rumen degradability and intestinal digestibility of whole and broken peas, lupins and faba beans and a mixture of these feedstuffs. Anim Feed Sci Technol 76:35–50 (1998). 10 Robinson PH, Fadel JG and Tamminga S, Evaluation of mathematical models to describe neutral detergent residue in terms of its susceptibility to degradation in the rumen. Anim Feed Sci Technol 15:249–271 (1986). 11 Tamminga S, van Straalen WM, Subnel APJ, Meijer RMG, Steg A, Wever CJG, et al, The Dutch protein evaluation system: DVE/OEB system. Liv Prod Sci 40:139–155 (1994). 12 CVB, Veevoedertable, gegevens over chemische samenstelling, verteerbaarheid en voederwaarde van voedermiddelen, Centraal Veevoeder Bureau, Lelystad, NL (2005). 13 Van der Poel AFB, Presløkken E and Goelema JO, Feed processing: effects on nutrient degradation and digestibility, in Quantitative Aspects of Ruminant Digestion and Metabolism, ed. by Dijkstra J, Forbes JM and France J. CABI Publishing, Wallingford, UK, p. 734 (2005). 14 Sadeghi AA and Shawrang P, Effects of microwave irradiation on ruminal degradability and in vitro digestibility of canola meal. Anim Feed Sci Technol 127:45–54 (2006). 15 Kaldmäe H, Kass M, Kärt O and Olt A, Effect of temperature on the degradation of rapeseed cake protein. Vet Zootech 36:30–34 (2006). 16 Davies ZS, Mason D, Brooks AE, Griffith GW, Merry RJ and Theodorou MK, An automated system for measuring gas production from forages inoculated with rumen fluid and its use in determining the effect of enzymes on grass silage. Anim Feed Sci Technol 83:205–221 (2000). 17 Moshtaghi Nia SA and Ingalls JR, Evaluation of moist heat treatment of canola meal on digestion in the rumen, small intestine, large intestine and total digestive tract of steers. Can J Anim Sci 75:279–283 (1992). 18 Hurrel RF, Carpenter KJ, Sinclair WJ and Otterburn MS, Mechanisms of heat damage in proteins. 7. The significance of lysine-containing isopeptides and lanthionine in heated proteins. Br J Nutr 35:383–395 (1976). 19 Van Soest PJ, Nutritional Ecology of the Ruminant. Cornell University Press, Ithaca, NY (1994). J Sci Food Agric 88:1380–1384 (2008) DOI: 10.1002/jsfa Journal of the Science of Food and Agriculture J Sci Food Agric 88:1385–1393 (2008) Rye bread reduces plasma cholesterol levels in hypercholesterolaemic pigs when compared to wheat at similar dietary fibre level‡ Helle Nygaard Lærke,∗ Camilla Pedersen,† Marianne Asp Mortensen, Peter Kappel Theil, Torben Larsen and Knud Erik Bach Knudsen University of Aarhus, Faculty of Agricultural Sciences, Department of Animal Health, Welfare and Nutrition, PO Box 50, 8830 Tjele, Denmark Abstract BACKGROUND: Rye is a whole-grain cereal with the potential of contributing to a healthy diet, but research on rye in relation to chronic diseases is scarce compared to wheat and oats. In this study, a total of 17 hypercholesterolaemic pigs were fed high-fat high-cholesterol rye (n = 9) or wheat-based buns (n = 8) with similar dietary fibre (DF) content for 9–10 weeks to study the effect on cardiovascular risk factors. RESULTS: Ingestion of rye bread resulted in a 40% lower plasma total and LDL cholesterol compared to the wheat group, whereas HDL cholesterol, insulin and glucose concentrations were not affected by dietary treatments. Intestinal viscosity was 7.2 times higher, and organic matter and fat digestibility significantly reduced in the pigs fed rye buns. The hepatic expression of the cholesterol 7α-hydroxylase gene (CYP7A1) was lower in rye-fed pigs, whereas four other key genes involved in cholesterol metabolism were not affected. Plasma cholesterol correlated inversely with intestinal viscosity and organic matter digestibility. CONCLUSION: The ability of DF from rye to interfere with digestion and absorption is more important for whole-body cholesterol homeostasis than regulation in the liver at gene level.  2008 Society of Chemical Industry Keywords: dietary fibre; viscosity; cereals; metabolic syndrome; gene expression INTRODUCTION Whole-grain products are recommended as part of a healthy diet preventing several lifestyle diseases, including cancer, type 2 diabetes and cardiovascular disease (CVD). Regarding the prevention of CVD, most attention has been paid to cereals rich in βglucans (oats and barley) with profound evidence of hypocholesterolaemic effects.1,2 Rye, which is the only grain traditionally used as whole-grain and contributes extensively to dietary fibre (DF) intake in Scandinavia,3 contains relatively little βglucan, but has a high content of both soluble and insoluble DF in the form of arabinoxylans (AX).4 To date, research on the effect of rye on chronic diseases is scarce compared to wheat and oats, but animal and human studies indicate lipid-lowering properties similar to oats, although conflicting results occur. Hence, further evidence of the effects of rye consumption on cardiovascular risk factors and knowledge of the mechanisms are needed. In this respect, the pig is considered an excellent model as its lipoprotein and apolipoprotein profiles are similar to humans; it can develop hypercholesterolaemia and atherogenic lipoproteins within few a weeks, and develop complex atherosclerotic lesions similar to those found in humans under experimental conditions.5,6 In the present study we examined the effects of rye DF on central risk markers of CVD (total plasma cholesterol, LDL, HDL, triglycerides, insulin, and glucose levels) in pigs fed an atherogenic diet for several weeks. We hypothesize that the risk markers of CVD are influenced either through events incurred by the action of DF in the gut compartments (viscosity in the small intestine or short-chain fatty acids produced in the large intestine) or by regulatory hepatic enzymes involved in the regulation of cholesterol synthesis. Therefore, ∗ Correspondence to: Helle Nygaard Lærke, University of Aarhus, Faculty of Agricultural Sciences, Department of Animal Health, Welfare and Nutrition, PO Box 50, 8830 Tjele, Denmark E-mail: Hellen.laerke@agrsci.dk † Present address: School of Biomedical and Molecular Sciences, University of Surrey, Guildford, Surrey GU2 7XH, UK. ‡ Preliminary results were presented at the LMC International Food Congress 2006 ‘Nutrigenomics and Health – From Vision to Food’ held at the Royal Veterinary and Agricultural University, Denmark, 15–16 March 2006. An abstract from the congress is published in Scand J Food Nutr 50(suppl 1): 29. (Received 19 June 2007; revised version received 7 January 2008; accepted 13 January 2008) Published online 9 April 2008; DOI: 10.1002/jsfa.3229  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 HN Lærke et al. we studied five key genes involved in cholesterol homeostasis: sterol regulatory element binding protein (SREBP2), HMG-CoA reductase (HMGR), acylCoA:cholesterol O-acyltransferase (ACAT ), hepatic LDL receptors (LDLr) and cholesterol 7α-hydroxylase (CYP7A1). EXPERIMENTAL The study was divided into two parts: (I) immediate changes in clinical chemical parameters when changing from a low-DF low-fat diet to high-fat high-DF wheat or rye buns; (II) medium-term effects of ingesting high-fat high-DF wheat or rye-based buns on selected key parameters of glucose, lipid and cholesterol metabolism. The study consisted of two blocks: based on the experiences from block 1 (five pigs), the study protocol was subjected to minor improvements to ensure the health of the pigs in block 2 (12 pigs). In part I of the study only pigs from block 2 were included, whereas both blocks were included in part II. The experiment was conducted according to protocols approved by the Danish Animal Experiments Inspectorate and complied with Danish Ministry of Justice Law no. 382 (10 June 1987) and Acts 739 (6 December 1988) and 333 (19 May 1990). Diets An atherogenic wheat-flour based ‘selection diet’ supplemented with egg powder (Danæg Products A/S, Roskilde, Denmark), lard, rape seed oil, whey protein concentrate (Lacprodan 87, Arla Foods Ingredients amba, Viby J, Denmark), pure cholesterol, and vitamin–mineral mixture (Table 1) was used to select for pigs responding to dietary cholesterol. A wash-out diet used in block 1 consisted of 867 g kg−1 wheat flour, 73 g kg−1 whey protein concentrate, 30 g kg−1 rape seed oil and 30 g kg−1 vitamin–mineral mixture. However, as constipation was observed in the wash-out period in the pigs from block 1, 60 g kg−1 cellulose in the form of Vitacel WF 600 (LCH A/S, Frederiksberg, Denmark) was added to the wash-out diet in block 2. The diets during the experimental period consisted of high-DF buns enriched with fat and cholesterol (Table 1). The rye buns contained finely ground whole-grain rye (Valsemøllen, Esbjerg, Denmark) and ground rye bran (Nordmills, Uppsala, Sweden), whereas the wheat buns contained white wheat flour with added Vitalcel WF 600 in order to obtain the same DF level. Whey protein concentrate was used to adjust the protein content, and lard, egg powder and cholesterol to increase the fat and cholesterol content. The wheat buns were produced at Lantmännen Unibake, Karup, Denmark, and the rye buns at Holstebro Technical College, Denmark. The buns were stored at −20 ◦ C until consumption. Animals and feeding A total of 30 Duroc × Danish Landrace × Yorkshire 4-month-old gilts (weighing approximately 70 kg) 1386 Table 1. Ingredients of experimental diets (g kg−1 ) Atherogenic Wheat flour Rye whole meal Rye bran Cellulose Whey protein concentrate Yeast Sugar Egg powder Rape seed oil Lard Cholesterol Vitamin–mineral mixb Experimental bun dietsa Selection diet Wheat Rye 735 0 0 0 10 0 0 150 20 50 5 30 528 0 0 157 25 20 15 150 20 50 5 30 0 310 400 00 0 20 15 150 20 50 5 30 a Before water addition and baking. Providing in mg kg−1 diet: 6642 Ca(H2 PO4 )2 , 4122 NaCl, 16 580 CaCO3 , 286 FeSO4 .7H2 O, 114 ZnO, 41 Mn3 O4 , 92 CuSO4 .5H2 O, 0.3 KI, 0.8 Na2 SeO3 .5H2 O, 2.1 retinoacetate, 0.03 cholecalciferol, 69 α-tocopherol, 2.52 menadione, 4.58 riboflavin, 12.59 D-pantothenic acid, 0.025 cyanocobalamine (B12 ), 2.52 thiamin (B1 ), 25.2 niacin, 3.78 pyridoxine (B6 ), 0.063 biotin. b obtained from the swine herd at the Faculty of Agricultural Sciences, Foulum, Denmark, were fed the atherogenic selection diet twice daily (1 kg per meal) for 2 weeks. Based on the cholesterol response, hyperresponders (cut-off value of 3.5 mmol L−1 plasma) were chosen as study subjects for the remaining part of the study. For the following 2 weeks the pigs were fed the wheat-flour based wash-out diet at a level of 2 kg.d−1 (blocks 1 and 2) increasing to 2.5 kg d−1 in the third week (block 2 only). After the wash-out period the pigs were fed the experimental diets (eight pigs on wheat buns, nine pigs on rye buns) for 6 (block 1) or 7 weeks (block 2) at an initial level of 2 kg d−1 ) increasing to 3 kg d−1 for the last 2.5 weeks of the study, where the buns were ground and chromic oxide (2 g kg−1 on as-is basis) was added. In the second week after chromium addition the pigs were transferred to metabolic cages for 7 days, where faeces and urine were collected quantitatively on a daily basis and pooled. The pigs were subsequently returned to their pens and slaughtered 2–4 days after the end of the balance period. When not placed in metabolic cages, the pigs were housed individually in 4 m2 smooth-walled pens with a concrete floor. The pigs were weighed once weekly. Blood sampling After 2 weeks on the atherogenic wheat-flour based diet a fasting blood sample was taken by venipuncture. During the wash-out period, 8 out of 10 selected pigs in block 1 were fitted with a jugular vein (JV) catheter with the aim of repeated blood samplings. However, the catheters were removed after a few days due to technical problems (infections and/or blockage of catheters), and the pigs treated with 0.05 mL kg−1 Streptocillin Vet. (Boehringer Ingelheim, J Sci Food Agric 88:1385–1393 (2008) DOI: 10.1002/jsfa Rye bread reduces plasma cholesterol levels in hypercholesterolaemic pigs Copenhagen, Denmark) for 3 days. Consequently, five pigs (three rye, two wheat) in block 1 completed the study, and only data from the balance period and slaughter are included. In block 2 fasting blood samples were taken by venipuncture on day −2, 8, and 12, from 12 pigs (six rye, six wheat) to study the development in clinical chemical blood parameters after transfer to the experimental diets. Postprandial blood samples from vena jugularis, lateral arteria auricularia, vena hepatica, and vena portae were taken 3 h after the morning meal from anaesthetized pigs using 10 mL Zolitil mixture containing 50 mg mL−1 of tiletamine/zolazepam (Vibrac SA, Carros, France), 2.5 mg mL−1 Torbugesic Vet (Scan Vet Animal Health A/S, Fredensborg, Denmark), 12.5 mg mL−1 Ketaminol Vet (Intervet Danmark, Skovlunde, Denmark), and 12.5 mg mL−1 Rompun (Bayer Health Car AG, Animal Health Division, Leverkusen, Germany). Anaesthesia was maintained using Dipivan (propofol, Astra-Zenica A/S, Albertslund, Denmark) and Haldid (Fentanyl, Janssen-Cilag A/S, Birkerød, Denmark) perfused through an ear vein at rates of 140 and 50 mL min−1 , respectively. Sampling from visceral organs The pigs were euthanized with an overdose of pentobarbital sodium (Pharmacy of Copenhagen University, Faculty of Life Science, Frederiksberg, Denmark) followed by exsanguination. A biopsy was taken from the left lobe of the liver, quick-frozen in liquid nitrogen, and stored at −80 ◦ C until analysis. The bowel content in the distal third of the small intestine, caecum, and the distal third of the colon were homogenized, an aliquot was stored frozen at −20 ◦ C, and the remainder freeze-dried for further analyses. Analyses of diets and digesta All chemical analyses were performed in duplicate. Freeze-dried samples of diets, digesta and faeces were ground to a particle size of less than 0.5 mm prior to analysis. The dry matter content of feed, freezedried faeces, and intestinal content was determined by drying to constant weight at 105 ◦ C for 20 h. Organic matter (OM) content was determined as the difference between the content of DM and ash,7 chromic oxide concentrations in digesta, faeces, and diets were analysed as described by Schürch et al.,8 and HClfat after the Stoldt procedure.9 Starch was analysed enzymatically as described by Knudsen10 without prior extraction of low-molecular-weight sugars, and with the use of enzymes supplied by Megazyme International Ireland Ltd, Wicklow, Ireland. Neutral non-starch polysaccharides (NSP) and constituent sugars and Klason lignin was determined as described by Knudsen.10 Viscosity of the digesta liquid phase was measured at 40 ◦ C in a Wells-Brookfield Dial cone/plate viscometer after thawing and centrifugation of digesta J Sci Food Agric 88:1385–1393 (2008) DOI: 10.1002/jsfa at 10 000 × g for 20 min at 4 ◦ C. Values of apparent viscosity are reported at shear rate 30 s−1 . Clinical chemical analyses on blood plasma Plasma glucose, triglycerides, total cholesterol, LDL and HDL cholesterol, and bile acids were analysed using an auto analyser, ADVIA 1650 Chemistry System (Bayer Corporation, Tarrytown, NY, USA) using human standards and calibration materials. Glucose, triglycerides, total cholesterol and LDL and HDL cholesterol were analysed according to standard procedures (Bayer HealthCare LLC). Total bile acids were analysed by oxidizing bile acids using 3α-hydroxysteroid dehydrogenase with subsequent reduction of thio-NAD to thio-NADH according to the Randox assay procedure (Randox Laboratories Ltd, Crumlin, UK). Immunoreactive insulin was analysed as described by Tindal et al.,11 and shortchain fatty acid (SCFA) concentrations in plasma were determined by gas chromatography according to Brighenti et al.12 Real-time reverse transcriptase–polymerase chain reaction (RT-PCR) Approximately 30 mg of liver samples was weighed and homogenized in 1000 µL TriReagent solution. The RNA was purified using the TriReagent kit (Sigma, St Louis, MO, USA) according to manufacturer’s protocol, except that liver samples were added to 200 µL of BCP solution (1-bromo-3-chloropropane). Synthesis of cDNA and quantification of gene transcription using real-time RT-PCR were carried out as previously described.13 Primers were designed across exon boundaries (exon structures reported for human were used) by using Primer Express 2.0 software (Applied Biosystems, Foster City, CA, USA). The design of primers and probe for ACAT did not allow discrimination between ACAT1 and ACAT-2. Details of primer (and probe) design and runs of real-time RT-PCR are given in Table 2. Transcription of target genes were normalized according to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) transcription. Calculations and statistical analysis Apparent digestibility of organic matter and fat were calculated relative to the concentration of chromic oxide. The flux of propionate in the portal vein (PV) was calculated as the product of an estimated portal flow of 25 mL kg−1 min−1 based on interpolation of values previously reported for this parameter,14,15 weight of the pigs, and the concentration of propionate in the PV. The estimated absorption of organic acids was calculated from the porto-arterial difference using the same portal flow values as mentioned above. For real-time RT-PCR all statistics were performed at the Ct stage defined as the difference between Cycles at threshold (Ct) of the target gene and Ct of GAPDH, as described in detail by Theil et al.13 1387 HN Lærke et al. Table 2. Accession numbers, amplicon location, amplicon length, slope of standard curve and PCR efficiency of target genes Gene Accession no. Amplicon Amplicon Slope of location length standard (exon–exon) (bp) curve CYP7A1 NM 001005352 3–4 146 −3.49 SREBP2 DQ020476 7–8 89 −3.32 LDLr AF067952 6–7 86 −3.43 HMGR S79678 16–17 83 −3.31 ACAT AK230873 12–13 119 −3.34 GAPDH AF017079 2–3 76 −3.47 Primer design (forward, reverse) 5′ -gaatgacaccctctccacct 5′ -gaatgacaccctctccacct 5′ -gatgcaaaggtcaaagacga 5′ -aaggtgaggacacacagcag 5′ -aaggagtgtgggaccaacga 5′ -aggcactcatagccgatcttg 5′ -tcgtgactgccatctacattgc 5′ - cgcttccattaaagtaataca gttgga 5′ -tcaatgcctttgctgagatgtt 5′ -cagccagtcatggaccacaac 5′ -gtcggagtgaacggatttgg 5′ -caatgtccactttgccagagttaa Probe/detection PCR efficiency 5′ -aagctgctttcattgcttca (Taqman probe)a SYBR 0.94 SYBR 0.96 SYBR 1.00 5′ -tggtggaa (LNA probe # 31)b 5′ -cgcctggtcaccagggctgct (Taqman probe)a 0.99 1.00 0.94 a Taqman probe labelled with FAM (carboxyfluorescein). Locked nucleic acid probe labelled with FAM (Universal ProbeLibrary (human); https://www.roche-applied-science.com/sis/rtpcr/upl/index.jsp; Roche Diagnostics A/S, Hvidovre, Denmark). b The relative mRNA quantity was calculated by using the formula: Relative mRNA quantity = 2−Ct . For target genes with PCR efficiencies lower than 1.0, the base number of two was adjusted accordingly (e.g., a base number of 1.96 for LDL receptor was used to account for a PCR efficiency of 0.96). All statistical analyses were carried out using SAS for Windows, version 8.2 (SAS Institute Inc., Cary, NC, USA). Level of significance was P < 0.05 in all analyses. Variables meeting the criteria of normal distribution are reported as least square means (lsmeans) ± their standard error of mean (SEM). When variables were not normally distributed, statistical evaluation was performed on logarithmized values, and results are reported as geometric means with 95% confidence intervals. Relative gene expression is also reported as 95% confidence intervals. The effect of diet in the initial phase of dietary intervention, of site of blood collection on plasma concentrations at slaughter, and on digestibility along the gut was examined by PROC MIXED, taking repeated measurements made on the same individuals (time, site or segment) into account.16 However, as there was no difference between sites of total HDL and LDL cholesterol, means of sites were subsequently subjected to one-way ANOVA using a general linear model. The effect of diet on faecal digestibility, apparent ileal viscosity, SCFA absorption, propionate flux and concentrations and proportions of SCFA at distinct collection sites, as well as the mRNA quantities, were subjected to one-way ANOVA analysis using PROC GLM. Only block 2 (n = 12) was included in the analysis of the development of blood cholesterol in phase I. For none of the studied responses were there any measurable long-term effects of differences in experimental handling between block 1 and 2, and 1388 both blocks (n = 17) were included in the results obtained from the balance period and at slaughter. Pearson correlation coefficients between parameters were obtained using the PROC CORR procedure. RESULTS Diets and animals The rye and wheat high-DF buns were formulated to be of similar macronutrient composition and energy content (Table 3). Protein provided 14%, fat 28% and carbohydrates 58% of the gross energy. The DF content was approximately 200 g kg−1 DM in both diets, but the chemical composition was different. The rye buns contained more Klason lignin, arabinoxylans (AX) and less cellulose than the wheat buns. The rye buns also contained three times more soluble NSP, of which AX is the main component. The pigs tolerated the experimental diets well. During the wash-out period, the pigs weighed 89.7 ± 1.0 kg. At the end of the balance period, the mean weight was 136.3 ± 0.9 kg with no effect of dietary treatment. Gastrointestinal effects The apparent viscosity of the supernatant of digesta in the distal small intestine was 7.2 times higher when pigs consumed the rye diet compared to the wheat diet (Table 4). Concurrently, the small intestinal OM and HCl-fat digestibility was significantly reduced in the rye group compared to the wheat-fed pigs, an effect that remained throughout the colon. Quantitatively, the reduction corresponded to 32 g (81%) more HClfat being excreted per day although the fat intake was 7% higher in the rye group at the end of the experiment. Total plasma cholesterol, LDL and HDL Initially, when fed the atherogenic selection diet the 17 hyper-responders completing the study had J Sci Food Agric 88:1385–1393 (2008) DOI: 10.1002/jsfa Rye bread reduces plasma cholesterol levels in hypercholesterolaemic pigs DF, dietary fibre as the sum of NSP and Klason lignin; NSP, nonstarch polysaccharides; NCP, non-cellulosic polysaccharides; AX, arabinoxylans as the sum of anhydro-arabinose and anhydro-xylose; A:X ratio, arabinose:xylose ratio. Values in parentheses are soluble NSP. Subsequently, after ingestion of both types of high-fat and cholesterol-rich buns there was a rapid rise in plasma total cholesterol (Table 5). However, the increase in cholesterol level was less rapid in pigs fed the rye buns. The development of LDL cholesterol levels followed the trend of total cholesterol, but the effect of diet did not reach statistical significance (Pdiet = 0.06). HDL cholesterol levels increased significantly in the initial phase of intervention (Ptime < 0.0001), but did not differ between dietary treatments. Hence, the LDL:HDL ratio increased also with both dietary treatments from baseline to day 12, but the difference between dietary treatments was not significant (Pdiet = 0.06). Whereas the cholesterol level increased throughout the experiment in pigs fed the wheat buns, the rye buns led to an apparent small decrease toward the end of the experiment (postprandial values, Table 6), and pigs fed the rye diet had significantly lower total and LDL cholesterol concentrations and LDL:HDL ratio than the wheat-fed pigs. a plasma cholesterol level of 7.2 ± 0.51 mmol L−1 (range: 4.5–12.0 mmol L−1 ). Then, after 2–3 weeks on the wash-out diet, plasma total cholesterol fell to 2.5 ± 0.08 mmol L−1 . Plasma insulin, glucose and triglycerides There was no effect of diet on fasting insulin or plasma glucose level between wash-out and day 12. The triglyceride concentration remained stable in pigs fed rye whereas it decreased in pigs fed the wheat buns (Table 5). However, no dietary differences were seen Table 3. Chemical composition (g kg−1 DM) and gross energy content (MJ kg−1 DM) of the experimental high-DF diets Wheat diet Dry matter (as-is basis) Energy (MJ) Starch and sugars (g) Protein (g) Fat (g) Ash (g) DF (g) Klason lignin (g) NSP (g) Cellulose (g) NCP-glucose (g) AX (g) A:X ratio Rye diet 661 21.10 437 182 153 39 194 22 173 (18) 110 16 (6) 36 (6) 0.26 680 21.16 384 172 159 50 203 40 163 (53) 22 31(13) 93 (35) 0.64 Table 4. Extract viscosity (mPa s) in the contents of the distal small intestine (SI) and digestibility of organic matter (OM) and HCl-fat in gut segments and faecesa Diet Viscosity Organic matter Fat P-value Segment Wheat Rye Diet (D) Segment (S) D×S SI3 Distal SI Caecum Distal colon Faeces Distal SI Caecum Distal colon Faeces 3.2 (1.97–5.11) 0.73 ± 0.010 0.82 ± 0.008 0.90 ± 0.003 0.91 ± 0.004 0.76 ± 0.020 0.87 ± 0.010 0.88 ± 0.006 0.87 ± 0.010 22.9 (14.21–36.82)  0.63 ± 0.010 0.79 ± 0.009  0.86 ± 0.003 0.86 ± 0.004  0.68 ± 0.020 0.78 ± 0.010  0.81 ± 0.006 0.78 ± 0.010 <0.0001 – – <0.0001 <0.0001 0.002 <0.0001 – – <0.0001 <0.0001 0.45 0.0003 – – a Calculated separately on the basis of faeces samples in the balance period. Values are lsmeans ± SEM or in parentheses 95% confidence intervals, n = 8 for wheat, n = 9 for rye, except for viscosity, where n = 8 for rye. Table 5. Development in fasting plasma concentrations of glucose (mmol L−1 ), insulin (µg L−1 ), triglycerides (mmol L−1 ), total, LDL and HDL cholesterol (mmol L−1 ) and LDL:HDL ratio in the immediate period after transfer to the high-fat, high-cholesterol DF-rich wheat or rye based buns Wash-out Wheat Day 8 Rye Wheat Day 12 Rye Wheat P-values Rye Glucose 4.6 ± 0.1 4.5 ± 0.1 4.5 ± 0.1 4.5 ± 0.1 4.5 ± 0.1 4.6 ± 0.1 Insulin 0.12 ± 0.06 0.19 ± 0.06 0.14 ± 0.06 0.15 ± 0.06 0.13 ± 0.03 0.09 ± 0.03 Triglycerides 0.20 ± 0.02 0.22 ± 0.02 0.13 ± 0.02 0.22 ± 0.02 0.13 ± 0.02 0.23 ± 0.02 Total cholesterol 2.5 ± 0.1 2.5 ± 0.1 8.8 ± 0.5 6.7 ± 0.5 10.2 ± 0.6 7.9 ± 0.6 HDL cholesterol 0.81 ± 0.04 0.87 ± 0.04 1.60 ± 0.09 1.58 ± 0.09 1.64 ± 0.10 1.67 ± 0.10 LDL cholesterol 1.3 ± 0.1 1.3 ± 0.1 5.4 ± 0.4 4.3 ± 0.4 6.2 ± 0.4 4.9 ± 0.4 LDL:HDL ratio 1.6 ± 0.1 1.5 ± 0.1 3.4 ± 0.2 2.7 ± 0.2 3.9 ± 0.4 2.9 ± 0.4 Diet Time Diet × Time 0.94 0.81 0.009 0.021 0.80 0.06 0.06 0.60 0.65 0.16 <0.0001 <0.0001 <0.0001 <0.0001 0.69 0.58 0.16 0.06 0.80 0.12 0.16 Values are lsmeans ± SEM, n = 8 for wheat, n = 9 for rye. J Sci Food Agric 88:1385–1393 (2008) DOI: 10.1002/jsfa 1389 HN Lærke et al. Table 6. Postprandial plasma concentrations of glucose (mmol L−1 ), insulin (µg L−1 ), triglycerides (mmol L−1 ), total, LDL and HDL cholesterol (mmol L−1 ) and LDL:HDL ratio, in the portal vein (PV), hepatic vein (HV), ear artery (Art) and jugular vein (JV) at the time of slaughter Diet Glucose PV HV Art JV PVa HV Art JV PV HV Art JV Insulin Triglycerides Total cholesterolb LDL cholesterolb HDL cholesterolb LDL:HDL ratiob P-value Wheat Rye Diet (D) Site (S) D×S 7.5 ± 0.9 18.2 ± 3.6 5.2 ± 0.3 5.2 ± 0.3 0.18 ± 0.05 0.10 ± 0.03 0.05 ± 0.008 0.04 ± 0.006 0.85 ± 0.10 0.97 ± 0.12 0.59 ± 0.07 1.03 ± 0.18 11.9 ± 0.9 6.9 ± 0.6 1.86 ± 0.11 3.8 ± 0.3 7.1 ± 0.9 19.1 ± 3.8 5.1 ± 0.3 4.8 ± 0.3 0.13 ± 0.05 0.02 ± 0.03 0.05 ± 0.008 0.03 ± 0.006 0.99 ± 0.10 1.12 ± 0.12 0.75 ± 0.07 0.82 ± 0.17 7.2 ± 0.8 4.2 ± 0.6 1.55 ± 0.11 2.7 ± 0.3 0.99 0.003 0.59 0.30 0.0006 0.09 0.55 0.02 0.35 0.001 0.005 0.06 0.03 – – – – – – Values are lsmeans ± SEM. PV, Art and JV: n = 8 for wheat, n = 9 for rye; HV: n = 8 for wheat, n = 7 for rye. a n = 7 for wheat, n = 7 for rye. b Values are calculated from means of the different sites, as no effect of site was found. on postprandial values at the end of the experiment (Table 6). Postprandial glucose and triglyceride concentrations were highest in the hepatic vein (HV), whereas insulin concentration was highest in PV (Table 6). The HV glucose concentrations were much higher than PV values, probably reflecting stress-induced conversion of hepatic glycogen to glucose during blood sampling. Bile acids The concentration of bile acid was 3.6–9.2 times higher in PV compared to HV (Table 7), and more than nine times higher than in JV and arterial plasma (P < 0.0001). Overall, there was no significant effect of diet, or interaction between diet and site of plasma collection. Eliminating the non-significant interaction led to an overall significant effect of diet (P = 0.04). However, separate analyses by one-way ANOVA of dietary effects at the different sites showed no difference between diets in portal, jugular, and arterial plasma, and only a tendency to reduced bile acid concentrations in hepatic plasma in rye-fed compared to wheat-fed pigs (P = 0.06). Short-chain fatty acids SCFA concentrations in plasma were very variable, and values were not significantly different between ryefed and wheat-fed pigs in PV (1544 ± 138 µmol L−1 , P = 0.41) and JV (339 ± 24 µmol L−1 , P = 0.11). However, in artery plasma (Art), the concentration was lower (P = 0.003) in rye-fed pigs (350 ± 21 µmol L−1 ) compared to pigs fed the wheat buns (461 ± 22 µmol L−1 ), and there was a similar tendency in HV (583 ± 159.6 versus 1049 ± 159.6 µmol L−1 , P = 0.053). The proportion of propionate was higher in portal plasma of rye-fed pigs (23.1 ± 1.50) compared to pigs fed the wheat buns (18.1 ± 1.59, P = 0.036). Although numerically quite small, there was a significant change in composition from acetate (94.2 ± 0.26 versus 95.6 ± 0.27, P = 0.002) to propionate (4.0 ± 0.29 versus 2.9 ± 0.31, P = 0.019) in arterial blood when feeding rye buns at the expense of wheat buns. Dietary treatment had no influence on the apparent daily absorption of total organic acids amounting on average to 5641 ± 658 mmol 24 h−1 , or the portal propionate flux amounting to 1547 ± 201 mmol 24 h−1 . Table 7. Postprandial plasma concentrations (µmol L−1 ) of total bile acids in the portal vein (PV), hepatic vein (HV), ear artery (Art), and jugular vein (JV) at the time of slaughter Wheat PV HV JV Art 113.0 30.9 10.8 11.6 Rye (89.7–142.3) (24.5–38.9) (8.3–14.1) (9.0–14.9) 126.4 13.7 8.4 8.6 P-values (101.7–157.1) (10.9–17.4) (6.5–10.8) (6.8–10.9) Pdiet : 0.27 Psite : <0.0001 Pdiet×site : 0.13 Values in parentheses are 95% confidence intervals obtained from PROC MIXED analysis. PV, Art and VJ: n = 8 for wheat, n = 9 for rye; HV: n = 8 for wheat, n = 7 for rye. 1390 J Sci Food Agric 88:1385–1393 (2008) DOI: 10.1002/jsfa Rye bread reduces plasma cholesterol levels in hypercholesterolaemic pigs 2.0 Relative expression 1.5 **** 1.0 0.5 0.0 Wheat Rye Wheat Rye Wheat Rye Wheat Rye Wheat Rye SREBP2 HMGR LDLr ACAT CYP7A1 Figure 1. Expression of liver mRNA of genes involved in cholesterol metabolism in rye bun-fed pigs compared to wheat-fed pigs. Gene expression in liver tissue With expressions of genes encoding for SREBP2, HMGR and LDLr of 1.21–1.25 and 0.99 for ACAT, there was no significant difference (P > 0.2) compared to wheat (Fig. 1). However, for CYP7A1, encoding cholesterol 7α-hydroxylase, the rate-limiting enzyme in bile acid biosynthesis in the liver, there was a significantly lower expression (P < 0.0001) in the pigs fed rye compared to the pigs fed the wheat diet. Correlations between responses The log-transformed apparent ileal viscosity was negatively correlated with total plasma cholesterol (r = −0.61, P = 0.012), and LDL cholesterol concentrations (r = −0.54, P = 0.032), and was also negatively associated with OM digestibility in the distal small intestine (r = −0.65, P = 0.006). There were no significant correlations between acetate, propionate or butyrate concentrations or their proportions and total plasma cholesterol in the portal vein, hepatic vein and jugular vein. However, in the ear artery acetate concentration (r = 0.51, P = 0.035) and acetate proportion (r = 0.52, P = 0.032) were positively correlated with total plasma cholesterol concentration. DISCUSSION Both types of high-fat, high-cholesterol buns increased plasma total and LDL cholesterol compared to the wash-out period, but the rye buns blunted the increase and maintained a lower total cholesterol level relative to the wheat buns. The results support other studies showing a hypocholesterolaemic effect of rye products in hamsters,17,18 rats,19 chickens20 and moderately hypercholesterolaemic men,21 but are in contrast to other studies with hamsters,22 human ileostomates23 and hypercholesterolaemic women.21 Several factors may count for the discrepancies between different studies; Firstly, the impact of DF is influenced by the initial cholesterol level, as has J Sci Food Agric 88:1385–1393 (2008) DOI: 10.1002/jsfa previously been documented in numerous human and animal intervention studies.1,17,22 Secondly, differences both between species and individuals may be explained by differences in their ability to maintain cholesterol homeostasis.24 – 27 Impaired glucose tolerance, insulin resistance and elevated triglyceride level are components of the metabolic syndrome, which increases the risk of developing type 2 diabetes and cardiovascular disease.28 However, neither fasting nor 3 h postprandial glucose and triglyceride levels were affected by diet. This supports previous observations on the effect of rye in hypercholesterolaemic subjects21,29 and in normocholesterolaemic pigs.14 However, we cannot exclude that the two experimental diets would result in different insulin sensitivity, as arabinoxylan-rich wheat endosperm DF and rye bread have previously been shown to reduce the insulin response to a meal in both type 2 diabetics and healthy subjects.29 – 33 The ability of rye DF to create a viscous environment leading to impaired absorption of fat and cholesterol and increased excretion of bile acids is probably a major factor accounting for the hypocholesterolaemic effect of the rye buns. Knudsen et al.14 has previously reported an approximately 50% higher viscosity when pigs were fed a rye diet when compared to a high-DF wheat diet. In rats and hamsters intestinal content viscosity has been reported to be inversely correlated to total liver cholesterol and plasma cholesterol,34 and the importance of viscosity is underlined by results showing that bile acid excretion in ileostomy subjects is reduced, and hypocholesterolaemic properties are abolished when soluble DF is hydrolysed.20,35 Interference of enterohepatic circulation and increased faecal excretion of bile acids and/or sterols were previously suggested as a mechanism for hypocholesterolaemic effects of DF.36 – 38 However, the relation is not straightforward, and it is still intensively discussed to what extent plasma cholesterol is regulated by intracellular cholesterol and bile acid levels, bile acid pool size, bile acid composition and 1391 HN Lærke et al. excretion.24 – 26,38 – 40 In the current study there was no difference in the concentration of bile acids in portal blood. Instead, there was a trend for a lower concentration of bile acid in plasma from the hepatic vein of the rye-fed pigs, which might indicate a lower bile acid synthesis compared to the wheat-fed pigs, although the majority of bile acids will flow into the bile duct. These results were supported by the lower hepatic expression of CYP7A1. None of the other key genes involved in hepatic cholesterol metabolism were significantly altered, although metabolic profiling using 1 H NMR suggests differences in hepatic fatty acid and cholesterol metabolism.41 As bile acid sequestrants, DF may lead to increased faecal excretion, thereby diminishing the bile acid pool size, and counteracting a suppressive effect of bile acids on CYP7A1.36,37 The lower expression of CYP7A1 in the current study is opposite to other studies on the effect of DF.37,38 However, since dietary cholesterol also induces a higher expression and activity of CYP7A1,24,25 the lower expression of CYP7A1 in the current study may be an effect of a lower fat and cholesterol absorption in the rye-fed pigs, as indicated by the lower small intestinal digestibility of HCl-fat. In this respect it must be emphasized that ileal and faecal bile acid excretion was not measured in the present study, but bile acids may have contributed to the increased excretion of HCl-fat in the rye-fed pigs. Another proposed mechanism for hypocholesterolaemic action of DF is via fermentation in the large bowel. Propionate has been suggested to attenuate hepatic cholesterol synthesis by decreasing the activity of the rate-limiting enzyme HMGR.42 – 44 Neither propionate concentration nor the proportion of SCFA in the form of propionate was correlated with plasma cholesterol levels, and the concentrations of propionate in the portal vein found in this study (∼300 µmol L−1 ) are hardly sufficient to exert a significant effect on cholesterol synthesis. Furthermore, neither total SCFA absorption nor propionate flux was influenced by diet. Consistently, Knudsen et al.14 found no differences in total SCFA absorption, whereas butyrate absorption was enhanced in pigs consuming high-DF wheat or rye diets. Collectively, these results suggest that SCFA is not a mediator of the hypocholesterolaemic effect of rye, as also suggested by others.45 – 47 Beside the difference in DF composition, the buns also differed in content of plant lignans, plant sterols and several other phenolic compounds,48 which may have the potential to affect blood lipids. Collectively, these confounding compounds may have added to the hypocholesterolaemic effect observed with the rye buns. CONCLUSIONS The present study demonstrated that a high intake of rye in the form of buns had hypocholesterolaemic 1392 properties when compared to a wheat-based diet at similar DF level. The hypocholesterolaemic effect of rye is presumably a result of the combination of soluble DF intake as well other bioactive components. However, reduced absorption rather than hepatic regulation is most likely the main factor contributing to the hypocholesterolaemic effect seen in the present study. However, the mechanisms involved require further research, and need to be confirmed at lower dose levels and in human subjects. ACKNOWLEDGEMENTS This project was financially supported by a grant from the Nordic Research Council (NKJ-121: Rye Bran for Health). Skilful assistance by the technical staff at the Institute of Animal Health, Welfare and Nutrition and the Department of Farm and Animal Research Facilities is greatly appreciated. REFERENCES 1 Ripsin CM, Keenan JM, Jacobs DR, Elmer PJ, Welch RR, Vanhorn L et al., Oat products and lipid lowering: a metaanalysis. JAMA 267:3317–3325 (1992). 2 Aman P, Cholesterol-lowering effects of barley dietary fibre in humans: scientific support for a generic health claim. Scand J Food Nutr 50:173–176 (2006). 3 Buttriss JL, Rye: the overlooked cereal. BNF Nutr Bull 31:3–5 (2006). 4 Hansen HB, Rasmussen CV, Knudsen KEB and Hansen A, Effects of genotype and harvest year on content and composition of dietary fibre in rye (Secale cereale L) grain. J Sci Food Agric 83:76–85 (2003). 5 Knudsen KEB and Canibe N, Changes in pig plasma lipids to dietary cholesterol, source and level of dietary fibre and caecal infusion of propionate, in COST 92: Metabolic and Physiological Aspects of Dietary Fibre in Foods. Mechanisms of Action of Dietary Fibre on Lipid and Cholesterol Metabolism, Carry le Rouet-Marseille, France, 23–24 October 1993, ed. by Lairon D. Commission of the European Communities, Luxembourg; pp. 123–130 (1993). 6 Dixon JL, Stoops JD, Parker JL, Laughlin MH, Weisman GA and Sturek M, Dyslipidemia and vascular dysfunction in diabetic pigs fed an atherogenic diet. Arterioscler Thromb Vasc Biol 19:2981–2992 (1999). 7 AOAC, Official Methods of Analysis. Arlington, VA (2006). 8 Schürch AF, Lloyd LE and Crampton EW, The use of chromic oxide as an index for determining the digestibility of a diet. J Nutr 50:629–636 (1950). 9 Stoldt W, Vorschlag zur Vereinheitlichung der Fettbestimmung in Lebensmitteln. Fette und Seifen/(Fette, Seifen, Anstrichmittel) 54:206–207 (1952). 10 Knudsen KEB, Carbohydrate and lignin contents of plant materials used in animal feeding. Anim Feed Sci Technol 67:319–338 (1997). 11 Tindal JS, Knaggs GS, Hart IC and Blake LA, Release of growth-hormone in lactating and non-lactating goats in relation to behavior, stages of sleep, electroencephalograms, environmental stimuli and levels of prolactin, insulin, glucose and free fatty-acids in circulation. J Endocrinol 76:333–346 (1978). 12 Brighenti F, Testolin G, Canzi E, Ferrari A, Wolever MS, Ciappellano S et al., Influence of long-term feeding of different purified dietary fibers on the volatile fatty acid (VFA) profile, pH and fiber-degrading activity of the cecal contents in rats. Nutr Res 9:761–772 (1989). J Sci Food Agric 88:1385–1393 (2008) DOI: 10.1002/jsfa Rye bread reduces plasma cholesterol levels in hypercholesterolaemic pigs 13 Theil PK, Sørensen IL, Therkildsen M and Oksbjerg N, Changes in proteolytic enzyme mRNAs relevant for meat quality during myogenesis of primary porcine satellite cells. Meat Sci 73:335–343 (2006). 14 Knudsen KEB, Serena A, Kjaer AKB, Jorgensen H and Engberg R, Rye bread enhances the production and plasma concentration of butyrate but not the plasma concentrations of glucose and insulin in pigs. J Nutr 135:1696–1704 (2005). 15 Serena A, Jorgensen H and Bach Knudsen KE, The absorption of lactic acid is more synchronized with the absorption of glucose than with the absorption of short-chain fatty acids: a study with sows fed diets varying in dietary fibre. Livest Sci 109:118–121 (2007). 16 Littell RC, Milliken GA, Stroup WW and Wolfinger RD, SAS System for Mixed Models. SAS Institute, Cary, NC (1996). 17 Rieckhoff D, Trautwein EA, Malkki Y and Erbersdobler HF, Effects of different cereal fibers on cholesterol and bile acid metabolism in the Syrian golden hamster. Cereal Chem 76:788–795 (1999). 18 Zhang JX, Lundin E, Reuterving CO, Hallmans G, Stenling R, Westerlund E et al., Effects of rye bran, oat bran and soyabean fibre on bile composition, gallstone formation, gallbladder morphology and serum cholesterol in Syrian golden hamsters (Mesocricetus auratus). Br J Nutr 71:861–870 (1994). 19 Lund EK, Salf KL and Johnson IT, Baked rye products modify cholesterol-metabolism and crypt cell-proliferation rates in rats. J Nutr 123:1834–1843 (1993). 20 Frigard T, Pettersson D and Aman P, Fiber-degrading enzyme increases body-weight and total serum-cholesterol in broilerchickens fed a rye-based diet. J Nutr 124:2422–2430 (1994). 21 Leinonen KS, Poutanen KS and Mykkanen HM, Rye bread decreases serum total and LDL cholesterol in men with moderately elevated serum cholesterol. J Nutr 130:164–170 (2000). 22 Zhang JX, Lundin E, Hallmans G, Bergman F, Westerlund E and Petterson P, Dietary-effects of barley fiber, wheat bran and rye bran on bile composition and gallstone formation in hamsters. Apmis 100:553–557 (1992). 23 Zhang JX, Lundin E, Hallmans G, Adlercreutz H, Andersson H, Bosaeus I et al., Effect of rye bran on excretion of bile-acids, cholesterol, nitrogen, and fat in human-subjects with ileostomies. Am J Clin Nutr 59:389–394 (1994). 24 Pandak WM, Schwarz C, Hylemon PB, Mallonee D, Valerie K, Heuman DM et al., Effects of CYP7A1 overexpression on cholesterol and bile acid homeostasis. Am J Physiol-Gastr L 281:G878–G889 (2001). 25 Tiemann M, Han ZH, Soccio R, Bollineni J, Shefer S, Sehayek E et al., Cholesterol feeding of mice expressing cholesterol 7 alpha-hydroxylase increases bile acid pool size despite decreased enzyme activity. PNAS 101:1846–1851 (2004). 26 Ratliff EP, Gutierrez A and Davis RA, Transgenic expression of CYP7A1 in LDL receptor-deficient mice blocks diet-induced hypercholesterolemia. J Lipid Res 47:1513–1520 (2006). 27 Patterson BW, Wong WW, Sheng HP, Mersmann HJ, Insull W, Klein PD et al., Neonatal genetically lean and obese pigs respond differently to dietary cholesterol. J Nutr 122:1830–1839 (1992). 28 Dyson MC, Alloosh M, Vuchetich JP, Mokelke EA and Sturek M, Components of metabolic syndrome and coronary artery disease in female ossabaw swine fed excess atherogenic diet. Comp Med 56:35–45 (2006). 29 Juntunen KS, Laaksonen DE, Poutanen KS, Niskanen LK and Mykkanen HM, High-fiber rye bread and insulin secretion and sensitivity in healthy postmenopausal women. Am J Clin Nutr 77:385–391 (2003). 30 Leinonen K, Liukkonen K, Poutanen K, Uusitupa M and Mykkanen H, Rye bread decreases postprandial insulin response but does not alter glucose response in healthy Finnish subjects. Eur J Clin Nutr 53:262–267 (1999). J Sci Food Agric 88:1385–1393 (2008) DOI: 10.1002/jsfa 31 Lu ZX, Walker KZ, Muir JG and O’dea K, Arabinoxylan fibre improves metabolic control in people with Type II diabetes. Eur J Clin Nutr 58:621–628 (2004). 32 Lu ZX, Walker KZ, Muir JG, Mascara T and O’dea K, Arabinoxylan fiber, a byproduct of wheat flour processing, reduces the postprandial glucose response in normoglycemic subjects. Am J Clin Nutr 71:1123–1128 (2000). 33 Juntunen KS, Niskanen LK, Liukkonen KH, Poutanen KS, Holst JJ and Mykkanen HM, Postprandial glucose, insulin, and incretin responses to grain products in healthy subjects. Am J Clin Nutr 75:254–262 (2002). 34 Gallaher DD, Hassel CA and Lee KJ, Relationships between viscosity of hydroxypropyl methylcellulose and plasma cholesterol in hamsters. J Nutr 123:1732–1738 (1993). 35 Lia A, Hallmans G, Sandberg AS, Sundberg B, Aman P and Andersson H, Oat beta-glucan increases bile acid excretion and a fiber-rich barley fraction increases cholesterol excretion in ileostomy subjects. Am J Clin Nutr 62:1245–1251 (1995). 36 Marlett JA, Hosig KB, Vollendorf NW, Shinnick FL, Haack VS and Story JA, Mechanism of serum-cholesterol reduction by oat bran. Hepatology 20:1450–1457 (1994). 37 Goel V, Cheema SK, Agellon LB, Ooraikul B and Basu TK, Dietary rhubarb (Rheum rhaponticum) stalk fibre stimulates cholesterol 7 alpha-hydroxylase gene expression and bile acid excretion in cholesterol-fed C57BL/6J mice. Br J Nutr 81:65–71 (1999). 38 Buhman KK, Furumoto EJ, Donkin SS and Story JA, Dietary psyllium increases expression of ileal apical sodium-dependent bile acid transporter mRNA coordinately with doseresponsive changes in bile acid metabolism in rats. J Nutr 130:2137–2142 (2000). 39 Moundras C, Behr SR, Remesy C and Demigne C, Fecal losses of sterols and bile acids induced by feeding rats guar gum are due to greater pool size and liver bile acid secretion. J Nutr 127:1068–1076 (1997). 40 Levrat-Verny MA, Behr S, Mustad V, Remesy C and Demigne C, Low levels of viscous hydrocolloids lower plasma cholesterol in rats primarily by impairing cholesterol absorption. J Nutr 130:243–248 (2000). 41 Bertram HC, Duarte IF, Gil AM, Knudsen KEB and Laerke HN, Metabolic profiling of liver from hypercholesterolemic pigs fed rye or wheat fiber and from normal pigs: highresolution magic angle spinning H-1 NMR spectroscopic study. Anal Chem 79:168–175 (2007). 42 Chen WJ, Anderson JW and Jennings D, Propionate may mediate the hypocholesterolemic effects of certain soluble plant fibers in cholesterol-fed rats. Proc Soc Exp Biol Med 175:215–218 (1984). 43 Wright RS, Anderson JW and Bridges SR, Propionate inhibits hepatocyte lipid-synthesis. Proc Soc Exp Biol Med 195:26–29 (1990). 44 Cheng HH and Lai MH, Fermentation of resistant rice starch produces propionate reducing serum and hepatic cholesterol in rats. J Nutr 130:1991–1995 (2000). 45 Nishina PM and Freedland RA, Effects of propionate on lipid biosynthesis in isolated rat hepatocytes. J Nutr 120:668–673 (1990). 46 Kishida T, Nogami H, Ogawa H and Ebihara K, The hypocholesterolemic effect of high amylose cornstarch in rats is mediated by an enlarged bile acid pool and increased fecal bile acid excretion, not by cecal fermented products. J Nutr 132:2519–2524 (2002). 47 Illman RJ, Topping DL, McIntosh GH, Trimble RP, Storer GB, Taylor MN et al., Hypocholesterolemic effects of dietary propionate: studies in whole animals and perfused rat liver. Ann Nutr Metab 32:97–107 (1988). 48 Liukkonen KH, Katina K, Wilhelmsson A, Myllymaki O, Lampi AM, Kariluoto S et al., Process-induced changes on bioactive compounds in whole grain rye. Proc Nutr Soc 62:117–122 (2003). 1393 Journal of the Science of Food and Agriculture J Sci Food Agric 88:1394–1399 (2008) Color stability of frozen whole tilapia exposed to pre-mortem treatment with carbon monoxide David Mantilla,1 Hordur G Kristinsson,1∗ Murat O Balaban,1 W Steven Otwell,1 Frank A Chapman2 and Sivakumar Raghavan1 1 Aquatic Food Products Program, Department of Food Science and Human Nutrition, University of Florida, Gainesville, FL 32611, USA of Fisheries and Aquatic Sciences, University of Florida, Gainesville, FL 32611, USA 2 Department Abstract BACKGROUND: Color of muscle foods plays a major role in consumer perception of meat quality. Carbon monoxide (CO) has been successfully used for improving color of packaged meat and fish products. In this study, we wanted to investigate pre-mortem treatment of live tilapia using 100% CO for its ability to improve the color of frozen whole tilapia. We compared untreated and CO-treated whole, gutted tilapia, frozen for 2 and 4 months at −20 ◦ C. Frozen tilapia samples were thawed overnight at 4 ◦ C, filleted and analyzed for their color, heme peak wavelength and CO concentration. RESULTS: Euthanasia using CO significantly increased redness (a∗ value) and lightness (L∗ value) of tilapia white and red muscle. Frozen storage significantly (P < 0.05) decreased redness of both CO-treated and untreated tilapia. However, even after 4 months of frozen storage, a∗ -value of CO-treated tilapia was similar to fresh untreated tilapia fillets. Heme peak wavelengths of CO-euthanized tilapia were higher than in untreated tilapia and there was no significant (P > 0.05) decrease in heme peak wavelengths of CO-treated tilapia white and red muscle during frozen storage. The CO content of frozen euthanized tilapia fillets was significantly (P > 0.05) higher than in untreated fillets. In general, red muscle tissue of euthanized tilapia had a higher concentration of CO than white muscle. CONCLUSION: Color stability of tilapia fillets was significantly improved by pre-mortem CO treatment. The color of CO-treated fillets was retained during frozen storage compared to untreated fillets. Hence, pre-mortem CO treatment could be used as a new method for improving color of tilapia.  2008 Society of Chemical Industry Keywords: carbon monoxide; tilapia; euthanasia; frozen storage; color; carbon monoxide concentration INTRODUCTION The color of fresh meat and meat products is one of the parameters by which consumers judge the quality of muscle foods.1 Typically, fresh meat is cherry-red in color due to the presence of oxy-myoglobin. However during storage, the oxy- form of myoglobin can oxidize into the undesirable brown-colored met-myoglobin.2,3 Antioxidants are usually added to muscle foods to prevent oxidation and to maintain the red color of meat. During the last decade, numerous researchers have started using carbon monoxide (CO) as a foodprocessing tool to preserve the red color of muscle foods.4 – 6 When muscle foods are treated with CO, it binds with the ferrous iron of myoglobin and forms a stable complex. The stable CO–myoglobin complex provides the desirable red color of meat. Usually, meat products are treated with CO as a part of modified atmospheric packaging.7 – 9 The use of CO in muscle foods has created controversy as the cherry red color can last well beyond the microbial shelf life of the meat and thus may mask spoilage10 and more importantly could potentially hide underlying safety problems. The US Food and Drug Administration (USFDA) has reviewed the data on CO and filtered smoke (which contains CO) and has not objected to its use as a preservative on tuna as long as the product is frozen and properly labeled, with the goal of preserving taste, aroma, texture, and color.11 Hence, for this reason it is important to investigate the effect of CO on the quality and safety of other seafood products such as tilapia. In this study, we wanted to examine the pre-mortem treatment of live tilapia (Oreochromis aureus) with CO as a method to improve the color stability of frozen whole tilapia. ∗ Correspondence to: Hordur G Kristinsson, Aquatic Food Products Program, Department of Food Science and Human Nutrition, University of Florida, Gainesville, FL 32611, USA E-mail: HGKristinsson@mail.ifas.ufl.edu (Received 10 December 2007; revised version received 21 January 2008; accepted 22 January 2008) Published online 17 April 2008; DOI: 10.1002/jsfa.3230  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Color stability of CO-treated frozen tilapia Several methods of fish slaughter, such as electrical stunning,12 immersion in liquid ice,13 and stunning using carbon dioxide (CO2 ),14 are currently being used in the aquatic food industry. Although several researchers have shown the ability of CO to improve the color and quality of fish,8,15,16 little or no research has studied 100% CO dissolved in water as a way of euthanizing tilapia and its effect on the quality of frozen tilapia fillets. CO has been known to induce loss of consciousness without pain and with minimal discomfort in small animals.17 Hypoxemia induced by CO could lead to death at concentrations as low as 4–6%. Hence, our objective in this study was to investigate the color retention of fillets obtained from CO-euthanized and then frozen tilapia and compare them with frozen non-treated tilapia fillets. Parameters such as change in the color of tilapia fillets due to CO treatment, change in the absorbance of heme protein, and CO retention in muscle tissues were investigated. MATERIALS AND METHODS Materials Live tilapia (Oreochromis niloticus) were obtained from a farm in Pierson, FL, USA. They were transferred live to the laboratory and kept in holding tanks prior to euthanasia using CO. Chemicals were of American Chemical Society (ACS) grade and were purchased from Fisher Scientific (Santa Clara, CA, USA). Treatment of tilapia using a saturated solution of CO A holding tank was constructed using transparent Plexiglas (36 × 16 × 12 in.), in which tilapia were euthanized with CO-saturated water. CO (100%) was flushed into the circulatory system of the holding tank to saturate the water with gas.18 All experiments were performed at ambient room temperature (21 ◦ C). Tilapia were maintained in the euthanizing tank until they were confirmed dead by visual inspection. On average, 31 min were needed for the completion of the euthanasia process. During every trial, 13 fish were euthanized. To flush the remaining CO out of the tank, air was flushed into the tank and CO was converted to CO2 by passing it through a Hopcalite catalyst tube. After CO treatment, tilapia were immediately gutted, vacuum packaged in high-density polyethylene (HDPE) bags, and frozen at −20 ◦ C. A set of normally slaughtered (i.e., concussive blow to the head of the fish, without CO treatment), tilapia were also gutted, vacuum packed and stored at −20 ◦ C. The latter were used as control. For storage studies, frozen tilapia were removed at the end of 2 and 4 months, thawed for 24 h at 4 ◦ C, filleted, and then analyzed for various parameters. All analyses were replicated at least twice. Color analysis A digital color machine vision system (CMVS) was used following the procedures outlined by Balaban et al.19 for detailed color analysis of RGB and L∗ J Sci Food Agric 88:1394–1399 (2008) DOI: 10.1002/jsfa (lightness), a∗ (redness), and b∗ (yellowness) values, along with hue values and identifying important color blocks for each treatment. The average L∗ , a∗ , and b∗ values were reported. The color analysis was done separately for white and dark lateral muscle of tilapia. The dark muscle was collected from the left lateral side of tilapia fillets. The reverse side, which had practically no red muscle, was used for analysis of the white muscle. Quantification of CO in tilapia fillets The concentrations of CO in the white and dark muscle of tilapia fillets were determined separately using the method of Miyazaki et al.20 Briefly, 6 g of minced muscle was introduced into a 60 mL headspace bottle. Three drops of 1-octanol (antifoaming agent) and 12 mL of 10% sulfuric acid were then added to the sample. Sulfuric acid was added to denature heme proteins and release CO. The mixture was shaken for 10 s and then incubated for 5 min at 40 ◦ C. After incubation, the tubes were shaken at room temperature for 15 min and 100 µL of the headspace gas was injected into a gas chromatography (GC) system (Agilent Technologies Inc. Santa Clara, CA, USA) equipped with a stainless steel Poropak Q column (3.17 mm i.d. × 1.82 m; 80–100 mesh), a methanizer (to convert CO to methane) and a flame ionization detector. Helium was used as the carrier gas at a flow rate of 29.7 mL min−1 . The injection port, column, methanizer and detector temperatures were maintained at 100 ◦ C, 35 ◦ C, 320 ◦ C and 250 ◦ C, respectively. Hydrogen was used as the reducing gas at a flow rate of 40 L min−1 . The retention time and area of the methane peak were compared to those obtained with a calibration CO gas. CO levels were then calculated based on a standard curve constructed by injecting different known levels of 100% CO. For CO quantification, three fillets from CO treatment and the control were retrieved from the cold room (4 ◦ C) and were analyzed. Heme protein extraction and spectrophotometric analysis The amount of CO taken up by tilapia muscle and the stability of CO bound to heme proteins during storage were determined using spectrophotometric analysis. Heme was extracted by the method of Huo and Kristinsson.21 A 10 g sample of tilapia muscle was mixed with 100 mL of 20 mmol L−1 Na2 HPO4 buffer (pH 8), followed by homogenization with an Ultra-Turrax T19 homogenizer (IKA Works, Inc., Wilmington, NC) at its lowest speed. The homogenate was filtered through Whatman no. 1 filter paper followed by centrifugation at 3000 × g. The entire extraction procedure was done at 4 ◦ C to avoid protein denaturation and loss of CO from the heme protein. The absorbance spectra of the supernatant was read between 350 and 700 nm. The absorption maxima for the heme protein was determined at 419 nm for COHb/Mb (carboxy-hemoglobin/myoglobin), 414 nm for oxy-Hb/Mb, and 408 nm and below for met-Hb/Mb. 1395 D Mantilla et al. 1396 (a) 35 25 b b 30 a*-value RESULTS AND DISCUSSION Effect of pre-mortem CO treatment on color of tilapia fillets The change in the muscle color of CO-treated whole tilapia, frozen at −20 ◦ C for 4 months, was analyzed using a CMVS. In this, method, the machine analyzes the color of the entire tilapia fillet. One advantage of using CMVS for color measurement is the ability to measure uneven surface color distribution such as the surface of tilapia fillets. The degree of redness (a∗ value) is often used as an indicator of quality and freshness of fish rich in dark muscle.22 In this study, we found that euthanizing tilapia with CO significantly (P < 005) increased the redness (a∗ value) of both red (Fig. 1(a)) and white muscle (Fig. 1(b)) compared to untreated tilapia fillets. The average a∗ value of fresh untreated tilapia red muscle was around 21 and white muscle around 17. Immediately after CO euthanization, the a∗ value of red muscle increased to 27 and white muscle to 23. The control a∗ values decreased progressively during frozen storage. During the first 2 months of frozen storage, a∗ value of the control decreased, but insignificantly (P > 0.05) compared to fresh untreated tilapia. However, during 4 months of frozen storage, there was a significant (P < 0.05) decrease in the redness of untreated tilapia fillets (Fig. 1(a, b)). The loss of redness in tilapia fillets during storage is due to the oxidation of the bright red oxy-myoglobin to brownish met-myoglobin.23 The a∗ values of CO-euthanized tilapia also decreased during frozen storage. There was no significant (P > 0.05) change during the first 2 months of storage, but both red and white muscle had significantly (P < 0.05) lower a∗ values after 4 months of storage. However, a∗ values of the fillets from the CO-euthanized fish were significantly (P < 0.05) higher throughout the study compared to the control. Even after 4 months of frozen storage, the redness of CO-euthanized tilapia red and white muscle was similar to the a∗ value of untreated fresh tilapia fillets. These results suggest significantly higher stability of a∗ values for frozen CO-euthanized tilapia compared to the untreated tilapia. Interestingly, an increase in a∗ values of red muscle was observed after 2 months of frozen storage. Danyali24 also noted an increase in a∗ values of CO-treated yellowfin tuna steaks subjected to 30 days of storage at −25 ◦ C. The increase in red color of the tilapia muscle is due to an increase in the concentration of bound CO in the muscle. Upon treatment, significant amounts of CO could remain unbound in the muscle, and during freezing and also thawing the unbound CO could bind to free reduced heme proteins and lead to an increase in redness. The yellowness (b∗ value) of CO-euthanized tilapia fillets was not significantly (P > 0.05) affected by frozen storage (Fig. 2). However, the b∗ value of both CO-treated and untreated samples increased during the initial 2 months and decreased during the subsequent 2 months of frozen storage. For the untreated (control) tilapia fillets, the b∗ value of red and white muscle increased significantly (P < 0.05) during the first 2 months of frozen storage and then decreased during the subsequent 2 months of storage. The increase in yellowness (b∗ value) could be due to the oxidation of lipids and proteins and the subsequent formation of yellow pigments.25 Also, the reduction in a∗ values for the tilapia fillets (Fig. 1) could correspond to oxidation and met-myoglobin formation, which could produce a brown-yellowish appearance in the red muscle, resulting in an increased b∗ value. The lightness (L∗ ) values of the CO-euthanized tilapia red and white muscle decreased significantly (P > 0.05) during the 4 months of frozen storage (Fig. 3). However, at the end of 4 months, the L∗ value of the CO-treated fillets was not significantly (P > 0.05) different from the untreated fresh tilapia muscle. The results may suggest that euthanasia with 100% CO yielded more ‘natural’ fresh-looking tilapia fillet than untreated fillets, at the end of frozen storage. The L∗ value of untreated tilapia red muscle showed a significant decrease (P < 0.05) during 4 months’ storage (Fig. 3(a)). The decrease in lightness values may indicate oxidation in tilapia red muscle. In tilapia white muscle (Fig. 3(b)), no significant decrease (P > 0.05) in L∗ value was observed. The difference in L∗ values between the red and white muscle of tilapia a Control 100% CO Euthanized d a 20 15 c 10 5 0 Fresh 2 months Storage time 30 25 20 4 months Control 100% CO Euthanized (b) 35 a*-value Statistical analysis All analyses were conducted in triplicate. Analysis of variance (ANOVA) and t-test were used to determine significant differences between treatments and among treatments. Statistical Analysis Software (SAS; Cary, NC, USA) and Microsoft Excel were used for analyzing the data. b b a d a 15 c 10 5 0 Fresh 2 months Storage time 4 months Figure 1. Effects of euthanasia with 100% CO and no treatment on a∗ values of (a) red muscle and (b) white muscle of fresh tilapia and tilapia stored frozen for up to 4 months. Columns having different letters are significantly different (P < 0.05). J Sci Food Agric 88:1394–1399 (2008) DOI: 10.1002/jsfa Color stability of CO-treated frozen tilapia b,c b*-value 16 12 a,b b,c a a a 8 (a) 80 a Control 100% CO Euthanized b 50 b 40 30 20 10 0 Fresh 2 months Storage time (b) 20 b,c a,b b,c 0 4 months Control 100% CO Euthanized a a a 12 Fresh (b) 80 70 60 50 40 30 20 10 0 a,b 2 months Storage time a a a 4 months Control 100% CO Euthanized b b L-value b*-value a a,b 60 4 16 a 70 L-value Control 100% CO Euthanized (a) 20 8 4 0 2 months Storage time Fresh 4 months Figure 2. Effect of euthanasia with 100% CO and no treatment on b∗ values of (a) red muscle and (b) white muscle of fresh tilapia and tilapia stored frozen for up to 4 months. Columns having different letters are significantly different (P < 0.05). could be due to a lower amount of heme pigments present in the white muscle than in red muscle and hence a lesser degree of oxidation in the white muscle of tilapia fillets. Spectroscopic analysis of heme pigments Hemoglobin/myoglobin in the muscle tissue of tilapia exists primarily in three different forms: oxy-, deoxyand met- forms. However, treatment with CO would change the UV-visible spectra of heme proteins. Kristinsson et al.26 had shown earlier that at pH 6.5 oxy-, met- and carboxy- forms of myoglobin would Fresh 2 months Storage time 4 months Figure 3. Effect of euthanasia with 100% CO and no treatment on L∗ values of (a) red muscle and (b) white muscle of fresh tilapia and tilapia stored frozen for up to 4 months. Columns having different letters are significantly different (P < 0.05). absorb around 414, 408 and 418 nm, respectively. Heme proteins from the red and white muscle of untreated and CO-treated tilapia were extracted and analyzed spectroscopically between 350 and 700 nm. For tilapia red muscle, a heme peak wavelength of 415 nm was observed for the untreated fresh sample, signifying the presence of oxy-hemoglobin/myoglobin (Fig. 4(a)). This peak was expected since the fillets were exposed to air during the filleting and skinning process (which was part of the sample preparation), which would have allowed oxygen from the air 100% CO Euthanized Control Wavelength (nm) (a) 420 418 416 414 412 410 408 406 404 Fresh (b) 420 418 416 414 412 410 408 406 404 2 months Storage time 4 months Wavelength (nm) 100% CO Euthanized Control Fresh 2 months Storage time 4 months Figure 4. Maximum heme peak values for (a) red muscle and (b) white muscle of euthanized (100% CO) and untreated fresh and frozen whole tilapia. J Sci Food Agric 88:1394–1399 (2008) DOI: 10.1002/jsfa 1397 D Mantilla et al. 1398 (a) CO (µg kg-1 of muscle) Quantification of CO in tilapia fillets Quantifying the concentration of CO in the muscle tissue of tilapia fillets is important for differentiating CO-treated products from untreated ones. Currently, a variety of methods are available for CO detection in food materials. These include spectrophotometric27,28 and gas chromatographic methods.29,30 In our present study, we used GC equipped with a flame ionization detector for quantifying CO in tilapia fillets. We chose the GC method for two reasons: (i) ability to detect CO at µg kg−1 level and (ii) availability of limited amount of muscle tissue for CO analysis. The amount of CO present in tilapia muscle varied with the type of muscle (white or red muscle). Untreated tilapia red (Fig. 5(a)) and white muscles (Fig. 5(b)) had low levels of CO, around 1100 µg and 900 µg respectively per kilogram of muscle. Euthanizing tilapia with CO increased the concentration of CO in the red and white muscle tissues to 3800 and 1300 µg/kg, respectively. In the untreated tilapia red and white muscle, there was no significant decrease (P > 0.05) in the concentration of CO during 4 months of frozen storage. In general, both the untreated and CO-treated white muscle of tilapia had lower amounts of CO than their corresponding red muscle tissue. This was expected since white muscle tissue contains lesser amounts of heme proteins and hence less ability to bind CO than the red muscle tissue. In the CO-euthanized tilapia samples, the amount of CO increased during the first 2 months of frozen storage and then decreased for the subsequent 2 months of storage. The increase and decrease in CO concentration were more significant in the red muscle (Fig. 5(a)) than in the white muscle (Fig. 5(b)) of COtreated tilapia fillets. It is possible that CO taken up via gills and bloodstream had not all been delivered into the muscle when it was sampled. Red muscle tissue, having a higher content of heme, would be able to trap greater amounts of released CO, leading to a higher concentration of CO and greater a∗ value or redness (Fig. 1(a)) compared to the white muscle. However, the concentration of CO decreased during 4 months of frozen storage in both the red and white muscle of CO-treated tilapia, indicating oxidation and loss of CO from the CO–Fe2+ heme complex. In conclusion, the color stability of tilapia fillets was significantly (P < 0.05) improved by pre-mortem treatment with 100% CO. Also, the red color of COtreated tilapia white and red muscle was retained during frozen storage. Hence, euthanasia of tilapia using 100% CO-saturated water could be used as a new method for improving the color of tilapia fillets. 12000 c 10000 (b) Control 100% CO Euthanized d 8000 b 6000 4000 2000 a a a 0 2 months Storage time Fresh CO (µg kg-1 of muscle) to bind to the heme proteins. The coupling of oxygen with heme would give a bright red color, explaining the high a∗ values obtained for the control in the fresh state (Fig. 1(a)). The heme peak wavelengths decreased during frozen storage, which indicates oxidation of heme proteins and loss of oxygen. These results correlate well with the loss of a∗ value of tilapia red muscle. At the end of 4 months’ frozen storage, the heme peak wavelength was around 409 nm, which indicates that the heme proteins were significantly oxidized, leading to methemoglobin/myoglobin formation. The red muscle from CO-euthanized tilapia fillets had higher heme peak wavelengths (∼418 nm) than the control, indicating the presence of CO-hemoglobin (Fig. 1(a)). The heme proteins in the euthanized fish were also significantly (P < 0.05) more stable as they maintained high heme peak wavelengths during 4 months of frozen storage. This stability accounts for the high a∗ values of CO-treated tilapia red muscle (Fig. 1(a)). Spectroscopic analysis showed that the heme peak absorbance values of tilapia white muscle (Fig. 4(b)) were not as high as those of red muscle (Fig. 4(a)). However, the absorbance values of CO-euthanized tilapia white muscle were higher compared to the control fillets. The high peak wavelengths of COtreated tilapia white muscle indicates CO binding with heme proteins. Binding of CO to the heme proteins would also result in a greater degree of redness (higher a∗ value), as observed with tilapia white muscle (Fig. 1(b)). The lower wavelength value of the euthanized white muscle compared to the red muscle suggests that the values represent an average of all three different heme protein forms, i.e., met-, oxy-, and carboxy-, while the control might be a mixture of met- and oxy- forms.21 The increase in heme peak wavelength, increased heme protein stability, and red color of the euthanized white muscle during 4 months of frozen storage may hence be due to the partial binding of CO to the heme proteins in the white muscle of tilapia. 2500 d 2000 4 months Control 100% CO Euthanized c,d 1500 a,b c,d a,c b 1000 500 0 Fresh 2 months Storage time 4 months Figure 5. Concentration of CO (µg/kg) in (a) red muscle and (b) white muscle of fresh and frozen untreated and euthanized (100% CO) tilapia. Columns having different letters are significantly different (P < 0.05). J Sci Food Agric 88:1394–1399 (2008) DOI: 10.1002/jsfa Color stability of CO-treated frozen tilapia ACKNOWLEDGEMENTS The authors would like to express their thanks to Mr. Gene Evans at Evans Farms in Pierson, FL for supplying tilapia for this study. REFERENCES 1 Kanner J, Oxidative processes in meat and meat-products: quality implications. Meat Sci 36:169–189 (1994). 2 Millar S, Wilson R, Moss BW and Ledward DA, Oxymyoglobin formation in meat and poultry. Meat Sci 36:397–406 (1994). 3 Faustman C and Cassens RG, The biochemical basis for discoloration in fresh meat: a review. J Muscle Foods 1:217–243 (1990). 4 Brewer MS, Wu SY, Field RA and Ray B, Carbon-monoxide effects on color and microbial counts of vacuum-packaged fresh beef steaks in refrigerated storage. J Food Qual 17:231–244 (1994). 5 Jayasingh P, Cornforth DP, Carpenter CE and Whittier D, Evaluation of carbon monoxide treatment in modified atmosphere packaging or vacuum packaging to increase color stability of fresh beef. Meat Sci 59:317–324 (2001). 6 Martinez L, Djenane D, Cilla I, Beltran JA and Roncales P, Effect of different concentrations of carbon dioxide and low concentration of carbon monoxide on the shelf-life of fresh pork sausages packaged in modified atmosphere. Meat Sci 71:563–570 (2005). 7 Wicklund RA, Paulson DD, Tucker EM, Stetzer AJ, DeSantos F, Rojas M, et al, Effect of carbon monoxide and high oxygen modified atmosphere packaging and phosphate enhanced, case-ready pork chops. Meat Sci 74:704–709 (2006). 8 Brown WD, Albright M, Watts DA, Heyer B, Spruce B and Price RJ, Modified atmosphere storage of rockfish (Sebastes miniatus) and silver salmon (Oncorhynchus kisutch). J Food Sci 45:93–96 (1980). 9 Hunt MC, Mancini RA, Hachmeister KA, Kropf DH, Merriman M, DelDuca G, et al, Carbon monoxide in modified atmosphere packaging affects color, shelf life, and microorganisms of beef steaks and ground beef. J Food Sci 69:C45–C52 (2004). 10 Kropf DH, Effects of retail display conditions on meat color, in Proceedings of the Annual Reciprocal Meat Conference, American Meat Science Association, IN, pp. 15–32 (1980). 11 Hahn MJ, The tasteless smoke process: preserving seafood with tasteless smoke, in Proceedings of the 25th Annual Meeting of the Seafood Science and Technology Society of the Americas, Longboat Key, FL (2000). 12 Robb DHF, O’Callaghan M, Lines JA and Kestin SC, Electrical stunning of rainbow trout (Oncorhynchus mykiss): factors that affect stun duration. Aquaculture 205:359–371 (2002). 13 Huidobro A, Mendes R and Nunes ML, Slaughtering of gilthead seabream (Sparus aurata) in liquid ice: influence on fish quality. Eur Food Res Technol 213:267–272 (2001). 14 Kestin S, Wotton S and Adams S, The effect of CO2 concussion or electrical stunning of rainbow trout (Oncorhynchus mykiss) on fish welfare, in Quality in Aquaculture, ed. by Svennevig N and Krogdahl A. EAS Special Publication 23, Ghent, Belgium, pp. 380–381 (1995). J Sci Food Agric 88:1394–1399 (2008) DOI: 10.1002/jsfa 15 Hsieh PP, Chow CJ, Chu YJ and Chen WL, Change in color and quality of tuna during treatment with carbon monoxide gas. J Food Drug Anal 6:605–613 (1998). 16 Chow CJ, Hsieh PP, Tsai ML and Chu YJ, Quality changes during iced and frozen storage of tuna flesh treated with carbon monoxide gas. J Food Drug Anal 6:615–623 (1998). 17 American Veterinary Medical Association, AVMA Guidelines on Euthanasia. AVMA, Schaumburg, IL, pp. 1–39 (2007). 18 Mantilla TD, Euthanasia of tilapia using carbon monoxide for color fixation and color stabilization. MS thesis, Department of Food Science and Human Nutrition, University of Florida, Gainesville, FL, p. 74 (2005). 19 Balaban MO, Kristinsson HG and Otwell WS, Evaluation of color parameters in a machine vision analysis of carbon monoxide treated fish. Part I: Fresh tuna. J Aquat Food Prod Technol 14:5–24 (2005). 20 Miyazaki H, Abe M, Asanoma M, Nagai Y, Nakajima M and Miyabe M, Simple determination of carbon monoxide in fish meat by GC. J Food Hyg Soc Japan 38:233–239 (1997). 21 Huo L and Kristinsson H, Rapid detection of carbon monoxide treated seafood products based on spectral properties of heme proteins, in Institute of Food Technologists Annual Meeting, New Orleans, LA, pp. 89A–35 (2005). 22 Hiroyasu O, Packaging for freshness and the prevention of discoloration of fish fillets. Packag Technol Sci 2:201–213 (1989). 23 Lee S, Joo ST, Alderton AL, Hill DW and Faustman C, Oxymyoglobin and lipid oxidation in yellowfin tuna (Thunnus albacares) loins. J Food Sci 68:1664–1668 (2003). 24 Danyali N, The effect of carbon monoxide and filtered smoke on quality and safety of yellowfin tuna, in Food Science and Human Nutrition, University of Florida, Gainesville, FL, pp. 24–43 (2004). 25 Thanonkaew A, Benjakul S, Visessanguan W and Decker EA, Yellow discoloration of the liposome system of cuttlefish (Sepia pharaonis) as influenced by lipid oxidation. Food Chem 102:219–224 (2007). 26 Kristinsson HG, Mony SS and Petty HT, Properties of tilapia carboxy- and oxyhemoglobin at postmortem pH. J Agric Food Chem 53:3643–3649 (2005). 27 Bylkas SA and Andersson LA, Microburger biochemistry: extraction and spectral characterization of myoglobin from hamburger. J Chem Educ 74:426–430 (1997). 28 Smulevich G, Droghetti E, Focardi C, Coletta M, Ciaccio C and Nocentini M, A rapid spectroscopic method to detect the fraudulent treatment of tuna fish with carbon monoxide. Food Chem 101:1071–1077 (2007). 29 Oritani S, Zhu BL, Ishida K, Shimotouge K, Quan L, Fujita MQ, et al, Automated determination of carboxyhemoglobin contents in autopsy materials using head-space gas chromatography/mass spectrometry. Forensic Sci Int 113:375–379 (2000). 30 Ishiwata H, Takeda Y, Kawasaki Y, Yoshida R, Sugita T, Sakamoto S, et al, Concentration of carbon monoxide in commercial fish flesh and in fish flesh exposed to carbon monoxide gas for color fixing. J Food Hyg Soc Japan 37:83–90 (1996). 1399 J Sci Food Agric 88:1400–1405 (2008) Journal of the Science of Food and Agriculture Antioxidant activity of the ethanolic extract from the bark of Chamaecyparis obtusa var. formosana Palanisamy Marimuthu, Chi-Lin Wu, Hui-Ting Chang and Shang-Tzen Chang∗ School of Forestry and Resource Conservation, National Taiwan University, Taipei, Taiwan Abstract BACKGROUND: Chamaecyparis obtusa var. formosana (Taiwan hinoki) is an endemic conifer in Taiwan and the purpose of this study is to evaluate the antioxidant activity of various fractions obtained from the bark of this plant material. The ethanolic extract of the bark was sequentially separated into three fractions, including n-hexane, ethyl acetate and ethanol soluble fractions, by liquid–liquid partition. Then the antioxidant activities of crude extract and three fractions along with 13 subfractions obtained from the ethyl acetate (EA) soluble fraction were tested for several antioxidant assays. RESULTS: The total phenolic content of the samples varied from 27.71 to 102.86 mg GAE g−1 dry weight for fractions, and from 49.94 to 206.46 mg GAE g−1 for subfractions (where GAE is milligrams of gallic acid per gram of extract). The Trolox equivalent antioxidant capacity (TEAC) ranged from 0.15 to 0.26 mmol L−1 Trolox equivalents. The EA soluble fraction was found to be the best antioxidant-rich fraction in terms of DPPH and reducing power assays. With further data analysis it was found that there was a positive correlation between the total phenolic content of extracts and TEAC is R2 = 0.61. CONCLUSION: Results from various antioxidant assays showed that the EA fraction possessed strong antioxidant activity. This would provide additional information about the antioxidant activity of bark extract of this plant species.  2008 Society of Chemical Industry Keywords: antioxidant activity; bark extract; β-carotene bleaching assay; Chamaecyparis obtusa var. formosana; total phenolic content; Trolox equivalent antioxidant capacity INTRODUCTION Plants have been used in many domains including medicine, nutrition, flavourings, beverages, dyeing, repellents, fragrances, cosmetics and other industrial purposes. Since the prehistoric era, plants have been the basis for nearly all medicinal therapy until synthetic drugs were developed in the 19th century.1,2 The preservative effect of many plant extracts suggests the presence of antioxidative and antimicrobial constituents in their tissues.3,4 Recently, interest has increased considerably in finding naturally occurring antioxidants for use in foods or medicinal materials to replace synthetic antioxidants, which are being restricted due to their carcinogenicity.5 Many medicinal plants contain large amounts of antioxidants such as polyphenols, which can play an important role in adsorbing and neutralising free radicals, quenching singlet and triplet oxygen, or decomposing peroxides. Many of these phytochemicals possess significant antioxidant capacities that are associated with lower occurrence and lower mortality rates of several human diseases.6 It has been reported that there is an inverse relationship between the antioxidative status occurrences of human diseases.7 In addition, antioxidant compounds which are responsible for such antioxidant activity could be isolated and then used as antioxidant for the prevention and treatment of free radical-related disorders.8 Therefore, research to identify antioxidative compounds is an important issue. Although it remains unclear which of the compounds from medical plants are the active ones, polyphenols have recently received increasing attention because of some interesting new findings regarding their biological activities. From pharmacological and therapeutic points of view, the antioxidant properties of polyphenols, such as free-radical scavenging and inhibition of lipid peroxidation, are the most crucial. There are seven species of the genus Chamaecyparis (Cupressaceae), but only two endemic species, C. formosensis and C. obtusa var. formosana, are found in the central mountains of Taiwan. Five new cadinane-type sesquiterpenes were isolated from the heartwood extract of C. obtusa var. formosana9 and also a recent report explains the dominancy of this plant partly due ∗ Correspondence to: Shang-Tzen Chang, School of Forestry and Resource Conservation, National Taiwan University, Taipei 106, Taiwan E-mail: peter@ntu.edu.tw (Received 29 November 2007; revised version received 11 January 2008; accepted 17 January 2008) Published online 17 April 2008; DOI: 10.1002/jsfa.3231  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Antioxidant activity of C. obtusa to the allelopathic potential, when it was planted along with other plant species.10 Both of these species are important building materials, and the latter species is believed to be more resistant to wood-decaying fungi. The choice of our investigated plant is based on two criteria: firstly, in this domain there is no antioxidative study that deals with this plant; and secondly, this plant is believed to be more resistant to wood-decaying fungi.11 Although there are several earlier studies on the chemical constituents12,13 and antitermite14 studies on C. obtusa var. formosana, there are no studies reported to antioxidant activity of bark of this plant species. This evaluation is related to the total phenolic content and antioxidant activity to determine new potential sources of natural antioxidants from the bark of C. obtusa var. formosana. MATERIALS AND METHODS Plant material The bark was collected from the 600-year-old Chamaecyparis obtusa var. formosana tree located in the central Taiwan. A voucher specimen was deposited in the laboratory of wood chemistry, School of Forestry and Resource Conservation, National Taiwan University. Chemicals 2,2′ -Diphenyl-1-picrylhydrazyl radical (DPPH), potassium dihydrogen phosphate, trichloracetic acid, 3-(-2-pyridyl)-5,6-bis(4-phenyl-sulfonic acid), βcarotene, lipoxidase (type I) from Glycine max Merrill (soybean), Folin–Ciocalteu reagent, quercetin, β-carotene and (+)-catechin were purchased from Sigma Chemical Co., St Louis, MO. Linoleic acid was from Acros, Morris Plains, NJ, USA. All other solvents and reagents were purchased from Sigma. Extraction and isolation Fifty kilograms of bark from the plant material was dried at room temperature and milled. The milled material was percolated in 95% ethanol (2 × 60 L) at room temperature for 6 days. Then, the extract was filtered and solvent was evaporated under reduced pressure gave an extract (472.4 g), which was subjected to liquid–liquid partition successively with n-hexane then ethyl acetate (EA) to give the n-hexane (180.5 g) and EA soluble fractions (166 g), respectively. The remaining fraction was considered as the ethanol soluble fraction (22.6 g). Based on the results obtained from antioxidant assays (DPPH and reducing power), the EA soluble fraction was considered to be the antioxidant-rich fraction. One hundred grams of the EA soluble fraction was applied to a silica gel open column and eluted with a stepped gradient consisting of n-hexane, EA, acetone, ethanol and water. The samples collected were screened by thin-layer chromatography (TLC) profile and fractions having similar TLC patterns were combined: 13 fractions (SF1–SF13) were obtained. J Sci Food Agric 88:1400–1405 (2008) DOI: 10.1002/jsfa Based on the DPPH screening assay, fraction 11 was found to be the antioxidant-rich fraction. It was applied (20 g) to a RP-18 open column and eluted with a gradient of MeOH–H2 O and further fractionated into 13 subfractions. Then the fractions were studied for antioxidant activity by using the DPPH assay and determining the reducing power, and also for their total phenolic content. Antioxidant assays DPPH assay The DPPH assay was carried out as reported previously.15 Fifty microliters of sample solution (100, 50, 25, 12.5 µg mL−1 as per final concentration) were added to 450 µL of Tris-HCl buffer and 1 mL of 0.1 mmol L−1 methanol solution of DPPH. After a 30 min incubation at room temperature, the absorbance was read against a blank at 515 nm in a Jasco V-550 UV–visible spectrophotometer (Tokyo, Japan). The assay was carried out in triplicate and results were averaged. (+)-Catechin was used as a positive reference. Reducing power The reducing power was determined as described previously.16 Various amounts (final concentration of 50, 25, 12.5, 6.25 µg mL−1 ) of fractions or subfractions (dissolved in methanol) were mixed with 0.5 mL of 0.2 mol L−1 phosphate buffer (pH = 6.6) and 0.5 mL of 1% potassium ferricyanide, and the mixture was incubated at 50 ◦ C for 20 min. After adding 0.5 mL of 10% trichloroacetic acid, the mixture was centrifuged at 976 × g for 10 min in a Hettich Micro 22R model centrifuge (Tuttlingen, Germany). The supernatant (0.5 mL) was mixed with 0.55 mL of distilled water and 0.1 mL of 0.1% ferric chloride and the absorbance read at 700 nm in a Jasco V-500 UV–visible spectrophotometer. Quercetin was used as a positive control. β-Carotene bleaching assay The β-carotene antioxidant assay was carried out as given by Chaillou and Nazareno17 and Kulisica et al.18 with slight modifications. All the reagents and solutions were prepared according to the procedure reported in the literature. Initially, 2.5 mL of βcarotene solution was thoroughly mixed with 200 µL of linoleic acid. Then, 200 µL of lipoxidase were added followed by 100 µL of sample (final concentrations of 100, 50, 25, 12.5, 6.25 µg mL−1 ). The absorbance of the control sample was measured immediately (t = 0) and t = 10 min. Reading of samples containing antioxidants were measured at 10 min at 460 nm in a Jasco V-550 UV–visible spectrophotometer. All determinations were performed in triplicate. The percentage of β-carotene inhibition was calculated as % inhibition = (1 − ((AS(0) − AS(10) )/ (AC(0) − AC(10) ))) × 100 1401 P. Marimuthu et al. where AS(0) is the absorbance of the sample at t = 0 min; AS(10) is the absorbance of the sample at t = 10 min; AC(10) is the absorbance of the control at t = 10 min; and AC(0) is the absorbance of the control at t = 0 min. Quantification of total antioxidant activity The total antioxidant activity values were estimated by the Trolox equivalent antioxidant capacity (TEAC) assay.19 In this test, we measured the relative capacity of antioxidants to scavenge the 2,2-azinobis(3ethylbenzothiazoline-6-sulfonic acid) diammonium salt (ABTS) radical compared to the antioxidant potency of Trolox is used as a standard. The ABTS radical generated by mixing a solution of ABTS (7 mmol L−1 ) with K2 S2 O8 (2.45 mmol L−1 ). Before use, the ABTS solution was diluted with water to obtain an absorbance of 0.700 ± 0.020 at 734 nm. Upon adding 1485 µL of the diluted ABTS solution to 15 µL of antioxidant sample or Trolox standard, the absorbance at 734 nm was recorded by a Jasco V-550 UV–visible spectrophotometer 6 min after initial mixing. Appropriate solvent blanks were run in each assay, and all measurements are done at least three times. Decreases in absorbance were noted and then calculated and plotted with respect to absorbance and concentration of the standard and samples. The final TEAC value of the antioxidant compound was calculated by comparing ABTS decolorisation with Trolox, which gives a useful indication of the antioxidant potential of the specimen. Determination of total phenolics according to the Folin–Ciocalteu method The amount of total phenolics was measured by the Folin–Ciocalteu method15 using gallic acid as standard, for which a calibration curve was obtained with solutions of 0.08, 0.04, 0.02, 0.01, and 0.005 mg mL−1 of this compound (y = 37.907x − 0.093, R2 = 0.9991). A 0.4 mL aliquot of diluted extract (all fractions were diluted with methanol to adjust the absorbance within the calibration limits), 0.4 mL of 1 mol L−1 Folin–Ciocalteu reagent, and 0.8 mL of Na2 CO3 (20%, w/v) were mixed. After 8 min, the mixture was centrifuged at 15 616 × g for 10 min. Then the absorbance of the supernatant solution was measured at 730 nm by using a Jasco V-550 UV–visible spectrophotometer and against a blank prepared similarly but containing distilled water instead of extract. The concentration of phenolics thus obtained was multiplied by the dilution factor and the results were expressed as the equivalent to milligrams of gallic acid per gram of extract (mg GAE g−1 ). Statistical analysis For all the extracts three samples were prepared for assays of every antioxidant attribute. The data were presented as mean ± standard deviation of three determinations. The significance of difference was 1402 analysed using SAS Scheffe’s statistics software and a value of P < 0.05 was considered significant. RESULTS AND DISCUSSION Antioxidant activities of bark extract and its fractions from C. obtusa var. formosana DPPH and reducing power assay The scavenging effect of crude (71.72–12.06%), n-hexane (38.68–6.81%), EA (83.06–32.12%) and ethanol (80.56–24.31%) soluble fractions on DPPH radical increased linearly with increasing concentration (Fig. 1) at 100, 50, 25 and 12.5 µg mL−1 . The IC50 values of EA, ethanol, crude and (+)catechin were found to be 21.88, 31.07 56.65 and 2.18 µg mL−1 , respectively. Wang et al.20 reported that the IC50 of an ethanolic extract of Calocedrus formosana bark, which belongs to the same family, was 23 µg mL−1 , which is higher in comparison with our plant. The reducing power of various soluble fractions increased with increasing concentration (Fig. 2). Based on optical density values of the fractions, the antioxidant activity can be ranked in the following descending order: EA fraction > ethanol fraction > crude extract > n-hexane fraction. Total phenolic content and TEAC assay The amount of total phenolics varied in different fractions and ranged from 27.71 to 102.86 mg GAE g−1 of dry material (Table 1) while for the subfractions it was ranged from 49.94–206.46 mg GAE g−1 (Table 2). EA soluble fraction showed higher phenolic content (102.86 mg GAE g−1 ) followed by ethanol, crude and n-hexane extract (27.71 mg GAE g−1 ). TEAC values are expressed in mmol L−1 Trolox equivalent. The ethanol fraction has a higher TEAC value of 0.26 mmol L−1 Trolox equivalent followed by the EA, crude and n-hexane fractions (Table 1). Wang et al.20 also demonstrated that the total phenolic content of the ethanol extract from Calocedrus formosana heartwood (159.5 ± 1.9 mg GAE Figure 1. Antioxidant activity of extract and various soluble fractions of C. obtusa var. formosana bark in terms of the DPPH radical scavenging assay. J Sci Food Agric 88:1400–1405 (2008) DOI: 10.1002/jsfa Antioxidant activity of C. obtusa Table 2. Total phenolic content, DPPH (IC50 value) and reducing power of subfractions fractionated on a reverse-phase open column Figure 2. Antioxidant activity of extract and various soluble fractions of C. obtusa var. formosana bark in terms of the reducing power assay. g−1 ) was higher than that of the bark extract (115.3 mg GAE g−1 ), which is relatively lower (90.72 ± 0.18) in Chamaecyparis obtusa var. formosana. The report21 on antioxidant activity of n-hexane, EA, n-butanol and water extracts from the bark of Chamecyparis lawosoniana revealed that the EA fraction exhibited a higher total phenolic content (337 mg GAE g−1 ) and a lower IC50 value (6.53 µg mL−1 ) in the DPPH assay. The EA fraction from our bark extract belonging to the same family showed lower total phenolic content (102.86 mg GAE g−1 ) and higher IC50 value (21.88 µg mL−1 ) in the DPPH assay. At the same time, our crude bark extract showed relatively higher total phenolic content in comparison with Juniperus oxycedrus22 from the same family. β-Carotene bleaching assay This method is based on the loss of the yellow colour of β-carotene due to its reaction with radicals that are formed by oxidation of linoleic acid, induced by lipoxidase in the emulsion. It was reported that linoleic acid is the preferred substrate for lipoxidase as it particularly attacks the fatty acid containing a 1-cis, 4-cis-pentadiene system.23 The rate of βcarotene bleaching can be reduced in the presence of antioxidants. This fact is used in the determination Specimen Total phenolic content (mg GAE g−1 ) DPPH IC50 (µg mL−1 ) Reducing power∗ SF1 SF2 SF3 SF4 SF5 SF6 SF7 SF8 SF9 SF10 SF11 SF12 SF13 Catechin Querectin 49.94 ± 1.15i 84.86 ± 0.75h 134.77 ± 0.50e 177.52 ± 0.50b 206.46 ± 1.34a 141.02 ± 0.57d 138.52 ± 0.78e,d 172.58 ± 1.78c 139.75 ± 0.32d 135.04 ± 0.98e 113.02 ± 0.76f 91.60 ± 0.84g 109.21 ± 1.37f – – 82.87 ± 1.73a 49.93 ± 0.22b 37.52 ± 0.07c 26.43 ± 0.72d 14.82 ± 0.47g 16.82 ± 0.17g,f 17.00 ± 0.24g,f 15.05 ± 0.21g 17.30 ± 0.29g,f 23.89 ± 1.41e 26.79 ± 0.52d 22.63 ± 0.18f 23.14 ± 0.25e 2.18 ± 0.03h – 0.75 ± 0.01h 1.29 ± 0.10f 1.27 ± 0.03f 1.77 ± 0.04e,d 2.22 ± 0.01b 2.10 ± 0.05c,b 1.94 ± 0.01c,d 2.06 ± 0.03c,b 1.77 ± 0.07e,d 1.39 ± 0.15g,f 1.66 ± 0.05e,f 2.19 ± 0.05c,b 1.76 ± 0.09e,d – 2.49 ± 0.01a Numbers followed by different letters (a–d) are statistically different at the probability level of P < 0.05 according to Scheffe’s analysis. SF, subfraction. ∗ Absorbance value at 700 nm, subfractions at final concentration of 50 µg mL−1 . Each value is mean ± SD of three measurements. of antioxidant activity of fractions obtained from C. obtusa var. formosana. Figure 3 shows the antioxidant activity of various fractions, among which the EA fraction showed strong antioxidant activity with 75.82% inhibition of β-carotene at a concentration of 100 µg mL−1 . IC50 values of the ethanol fraction, EA fraction, n-hexane fraction, crude extract and quercetin were found to be 43.16, 43.90, 92.45, 65.76 and 4.27 µg mL−1 , respectively. In terms of percent inhibition of β-carotene bleaching at 100 µg mL−1 , antioxidant activity of various fractions can be ranked as EA fraction = ethanol fraction > crude extract > n-hexane fraction. In the modified β-carotene assay, the percent inhibition of various fractions at higher concentration (100 µg mL−1 ) were significantly different, but at lower Table 1. Total phenolic content and Trolox equivalent antioxidant capacity values for extract and various fractions of C. obtusa var. formosana bark Specimen Crude extract n-Hexane fraction EA fraction Ethanol fraction Quercetin Total phenolic content (mg GAE g−1 ) TEAC (mmol L−1 trolox equivalent) 50.86 ± 1.61c 27.71 ± 0.18d 102.86 ± 0.78a 90.72 ± 0.18b – 0.19 ± 0.01b 0.15 ± 0.00b 0.21 ± 0.05b 0.26 ± 0.01b 3.55 ± 0.11a Numbers followed by different letters (a–d) are statistically different at the probability level of P < 0.05 according to Scheffe’s analysis. Each value is mean ± SD of three measurements. J Sci Food Agric 88:1400–1405 (2008) DOI: 10.1002/jsfa Figure 3. Antioxidant activity of extract and various soluble fractions of C. obtusa var. formosana bark in terms of the modified β-carotene bleaching assay. 1403 P. Marimuthu et al. concentration, almost all the specimens (except the nhexane fraction) showed similar percent inhibition. At the lower concentration (6.25 µg mL−1 ), the nhexane fraction gave a negative antioxidant value (−0.81%); this represents a pro-oxidant effect of the n-hexane fraction in this system. This result is in accord with the pro-oxidant effect of cinnamic acid shown by Chaillou and Nazareno.17 The presence of different antioxidant components in the plant tissues makes it relatively difficult to quantify each antioxidant component separately. Therefore, in many studies, several intermediate extractions are used to ensure a maximum extraction of the available antioxidants.24 The antioxidant activity of phenolics is mainly due to their redox properties which make them act as reducing agents, hydrogen donors, and singlet oxygen quenchers. They may also have a metallic chelating potential.25 Antioxidant activities of the subfractions from the ethyl acetate soluble fraction DPPH and reducing power assay Subfractions obtained from the RP-18 open column have IC50 values in the range 14.82–82.87 µg mL−1 (Table 2). Subfractions SF5, SF6, SF7, SF8 and SF9 exhibited strong radical scavenging effects and these subfractions have IC50 values of 14.82– 17.30 µg mL−1 . As for the reducing power, subfractions SF5, SF6, SF8 and SF12 showed higher reducing power (2.10–2.22) in comparison with other samples (Table 2). Total phenolic content RP open column considerably increased total phenolic content in the subfractions SF5, SF4 and SF8. Subfraction SF5 exhibited higher total phenolic content followed by subfractions SF4 and SF8 in comparison with other fractions (Table 2). The correlation coefficient, R2 , between TEAC and total phenolic content of various solvent fractions is 0.61 (Fig. 4). The antioxidant activity of fractions may not only be due to the presence of phenolic compounds but also related to the presence of some individual active components in the extracts. The unclear relationship between the antioxidant activity and total phenolic content may be explained by the fact that the total phenolic content does not incorporate all the antioxidants. In addition, the synergism between the antioxidants in the mixture makes the antioxidant activity not only dependent on the concentration but also on the structure and interaction between the antioxidants. This is why the EA and ethanol fractions, which have similar total phenolic contents, varied in antioxidant performance in the TEAC assay. Phenolic groups play an important role in antioxidant activity.26,27 It has been reported that most natural antioxidative compounds often work synergistically with each other to produce a broad spectrum of antioxidative activities that create an effective defence system against free-radical attack.28 The composition 1404 Figure 4. Linear correlation of Trolox equivalent antioxidant capacity (TEAC) with respect to the total phenolic content of various fractions obtained from C. obtusa var. formosana. of the extract is very complex; it consists of various classes of organic compound which may exert opposite effects on the process of lipid oxidation. Based on the results obtained, it is highly possible that some constituents of different polarity may contribute to the antioxidative activity of the extract. CONCLUSIONS It has also been noted in this study that the EA fraction of C. obtusa var. formosana bark extract showed strong radical scavenging and can be considered a good source of natural antioxidants for medicinal and commercial use. However, due to the diversity and complexity of the natural mixtures of phenolic compounds in this plant extract, it is not easy to characterise every compound and assess the antioxidant activity of each one. Each plant generally contains different phenolic compounds with different amounts of antioxidant activity. As a result of this study, we believe that in vivo studies are needed to further confirm the advantageous quality of these natural products. In order to confirm the antioxidative effect of these promising plant extracts, a further survey, which uses other types of antioxidant assay, is now under way. This survey also includes the characterisation of active phenolic antioxidants. ACKNOWLEDGEMENT We thank the National Science Council, Taiwan for generous financial support (NSC95-2313-B-002012). REFERENCES 1 Dahanukar SA, Kulkarni RA and Rege NN, Pharmacology of medicinal plants and natural products. Indian J Pharmacol 32:81–118 (2000). 2 Exarchou V, Nenadis N, Tsimidou M, Gerothanassis IP, Troganis A and Boskou D, Antioxidant activities and phenolic composition of extracts from Greek oregano, Greek sage and summer savory. J Agric Food Chem 50:5294–5299 (2002). J Sci Food Agric 88:1400–1405 (2008) DOI: 10.1002/jsfa Antioxidant activity of C. obtusa 3 Singh G, Marimuthu P, de Heluani CS and Catalan CAN, Antimicrobial and antioxidant potentials of essential oil and acetone extract of Myristica fragrans Houtt. (aril part). J Food Sci 70:141–148 (2005). 4 Cheng SS, Liu JY, Hsui YR and Chang ST, Chemical polymorphism and antifungal activity of essential oils from leaves of different provenances of indigenous cinnamon (Cinnamomum osmopholeum). Bioresource Technol 97:306–312 (2006). 5 Velioglu YS, Mazza G, Gao L and Oomah BD, Antioxidant activity and total phenolics in selected fruits, vegetables, and grain products. J Agric Food Chem 46:4113–4117 (1998). 6 Anderson KJ, Teuber SS, Gobeille A, Cremin P, Waterhouse AL and Steinberg FM, Walnut polyphenolics inhibit in vitro human plasma and LDL oxidation. Biochemical and molecular action of nutrients. J Nutri 131:2837–2842 (2001). 7 Rice-Evans CA, Sampson J, Bramley PM and Holloway DE, Why do we expect carotenoids to be antioxidants in vivo? Free Radical Res 26:381–398 (1997). 8 Middleton EJ, Kandaswami C and Theoharides TC, The effects of plant flavonoids on mammalian cells: implications for inflammation, heart disease, and cancer. Pharmacol Rev 52:673–751 (2000). 9 Kuo YH, Chen CH, Chien SH and Lin YL, Five new cadinanetype sesquiterpenes from the heartwood of Chamaecyparis obtusa var. formosana. J Nat Prod 65:25–28 (2002). 10 Tseng MH, Lai WR, Hsieh CL and Kuo YH, Allelopathy on bark of downed logs of Chamaecyparis obtusa Sieb. and Zucc. var. formosana (Hayata) Rehder. J Chem Ecol 33:1283–1296 (2007). 11 Kuo YH, Chen CH and Huang SL, New diterpenes from the heartwood of Chamaecyparis obtusa var. formosana. J Nat Prod 61:829–831 (1998). 12 Su YC, Ho CL and Wang EIC, Analysis of leaf essential oils from the indigenous five conifers of Taiwan. Flavour Frag J 21:447–452 (2006). 13 Wang SY, Wang YS, Tseng YH, Lin CT and Liu CP, Analysis of fragrance compositions of precious coniferous woods grown in Taiwan. Holzforschung 60:528–532 (2006). 14 Cheng SS, Chang HT, Wu C-L and Chang ST, Anti-termitic activities of essential oils from coniferous trees against Coptotermes formosanus. Bioresource Technol 98:456–459 (2007). 15 Tung YT, Wu JH, Kuo YH and Chang ST, Antioxidant activities of natural phenolic compounds from Acacia confusa bark. Bioresource Technol 98:1120–1123 (2007). J Sci Food Agric 88:1400–1405 (2008) DOI: 10.1002/jsfa 16 Oyaizu M, Studies on products of browning reactions: antioxidative activities of products of browning reactions prepared from glucosamine. Japanese J Nutr 44:307–315 (1986). 17 Chaillou LL and Nazareno MA, New method to determine antioxidant activity of polyphenols. J Agric Food Chem 54:8397–8402 (2006). 18 Kulisica T, Radonic A, Katalinic V and Milosa M, Use of different methods for testing antioxidative activity of oregano essential oil. Food Chem 85:633–640 (2004). 19 Miller NJ, Rice-Evans CA and Papaganga G, Antioxidant properties of phenolic compounds. Trends Plant Sci 4:152–159 (1997). 20 Wang SY, Wu JH, Cheng SS, Lo CP, Chang HN, Shyur LF et al., Antioxidant activity of extracts from Calocedrus formosana leaf, bark and heartwood. J Wood Sci 50:422–426 (2004). 21 Gao H, Shupe TF, Hse CY and Eberhardt TL, Antioxidant activity of extracts from the bark of Chamaecyparis lawsoniana (A. Murray) Parl. Holzforschung 60:459–462 (2006). 22 Djeridane A, Yousfi M, Nadjemi B, Boutassouna , Stocker P and Vidal N, Antioxidant activity of some Algerian medicinal plants extracts containing phenolic compounds. Food Chem 97:654–660 (2006). 23 Belitz HD and Grosch W. Lipids, in Food Chemistry, 2nd edition, ed. by Belitz HD and Grosch W. Springer-Verlag, Berlin, Heidelberg, pp. 199–200 (1999). 24 Kakoen MP, Hopia AI, Vuorela HJ, Rauha JP, Pihlaja K and Kujala TS, Antioxidant activity of plant extracts containing phenolic compounds. J Agric Food Chem 47:3954–3962 (1999). 25 Rice-Evans CA, Miller NJ, Bolwell PG, Bramley PM and Pridham JB, The relative antioxidant activities of plant derived polyphenolic flavonoids. Free Radical Res 22:375–383 (1995). 26 Huang SW and Frankel EN, Antioxidant activity of tea catechins in different lipid systems. J Agric Food Chem 40:3033–3038 (1997). 27 Baratta MT, Dorman HJD, Deans SG, Figueiredo AC, Baroso JG and Ruberto G, Antimicrobial and antioxidant properties of some commercial essential oils. Flavour Frag J 13:235–244 (1998). 28 Lu F and Foo LY, Phenolic antioxidant component of evening primrose, in Nutrition, Lipids, Health and Diseases, ed. by Ong ASH, Niki E and Packer L. Champaign: American Oil Chemists Society Press, pp. 86–95 (1995). 1405 Journal of the Science of Food and Agriculture J Sci Food Agric 88:1406–1414 (2008) Cloning of an alfalfa polyphenol oxidase gene and evaluation of its potential in preventing postharvest protein degradation Michael L Sullivan,1∗ Ronald D Hatfield1 and Deborah A Samac2 1 US Dairy Forage Research Center, 1925 Linden Drive West, Madison, WI 53706, USA Plant Science Research Unit and Department of Plant Pathology, University of Minnesota, 1991 Upper Buford Circle, St Paul, MN 55108, USA 2 USDA-ARS Abstract BACKGROUND: Ensiling forages often leads to degradation of protein to non-protein nitrogen (NPN), which is poorly utilized by ruminants. Postharvest protein degradation is especially high in alfalfa (Medicago sativa L.). In contrast, red clover (Trifolium pratense L.) has up to 90% less protein loss during ensiling due to polyphenol oxidase (PPO) forming o-quinones from endogenous o-diphenols and subsequent binding of o-quinones to cytoplasmic proteins. Here we determined whether an endogenous PPO might be exploited for postharvest protein protection in alfalfa. RESULTS: We isolated an alfalfa PPO gene (MsPPO1) that shares limited sequence identity (70–72%) with red clover PPO genes. MsPPO1 is expressed primarily in flowers and developing seed pods, but not in leaves or stems. Expression of MsPPO1 from a strong constitutive promoter in transgenic alfalfa results in accumulation of PPO transcripts in leaves, but little enzyme activity is detected using a variety of o-diphenol substrates unless assayed in the presence of sodium dodecyl sulfate (SDS). Under this SDS-activated condition, preference of MsPPO1 for tested substrates is catechol ≥ (−)-epicatechin > caffeic acid. PPO activity in unactivated MsPPO1-alfalfa extracts is sufficient to inhibit proteolysis in the presence of catechol, but not caffeic acid or (−)-epicatechin. Inhibition is less than in extracts of alfalfa expressing the red clover PPO1 gene. CONCLUSION: Endogenous alfalfa PPO, even if expressed in appropriate target tissues, would be less effective at preventing proteolytic losses in ensiled forages than red clover PPO. Published in 2008 by John Wiley & Sons, Ltd. Keywords: polyphenol oxidase; alfalfa; forage legumes; proteolysis; protein preservation INTRODUCTION Polyphenol oxidases (PPOs; EC 1.14.18.1, 1.10.3.1) are capable of catalyzing the oxidation of o-diphenols to their corresponding o-quinones. Although they are nearly ubiquitous among plants,1 the exact roles they play in plant growth, development, and physiology are not entirely clear. In at least some cases, PPOs appear to be involved in pathogen defense responses.2,3 Some PPOs have been implicated in biosynthetic pathways including the biosynthesis of yellow aurone pigments in snapdragon flowers4 and a specific lignan in creosote bush.5 PPOs are probably best known for their negative impact on the quality of fresh fruits and vegetables as a result of PPO-mediated oxidation of endogenous o-diphenols to o−quinones.6 The resulting o-quinones covalently couple to a number of cellular nucleophiles and consequent secondary quinone reactions lead to the formation of brown and black colored polymers (the browning reaction). Despite the usual negative association of PPO with agricultural crops, we have recently demonstrated that oxidation of o-diphenols by PPO can be exploited as a natural system of protein protection in forage crops preserved by ensiling.7 Postharvest protein degradation in ensiled forage crops results in the conversion of true protein to amino acids and peptides (non-protein nitrogen, NPN). This conversion of true protein to NPN is problematic because dairy cows and other ruminant animals poorly utilize excess non-protein nitrogen, resulting in economic losses to farmers, who must supplement rations with other sources of true protein. Such losses are especially high in alfalfa (Medicago sativa L.), approaching $100 million annually in the United States alone.7,8 Further, loss of true protein in preserved forages has negative environmental impacts, as excess NPN is excreted by ruminants as urea, increasing N burdens to the environment. In contrast to alfalfa, the forage legume red clover (Trifolium pratense L.) experiences up to 90% less proteolysis ∗ Correspondence to: Michael L Sullivan, US Dairy Forage Research Center, 1925 Linden Drive West, Madison, WI 53706, USA E-mail: michael.sullivan@ars.usda.gov (Received 18 October 2007; revised version received 16 January 2008; accepted 17 January 2008) Published online 16 April 2008; DOI: 10.1002/jsfa.3232 This article is a US Government work and is in the public domain in the USA. J Sci Food Agric 0022–5142/2008/$30.00 Cloning and characterization of an alfalfa polyphenol oxidase gene when ensiled.9 We have recently demonstrated that red clover’s lower level of postharvest proteolysis is due to the oxidation of endogenous o-diphenols by red clover PPO.7 Using genetically modified alfalfa expressing the red clover PPO1 gene (TpPPO1), we also demonstrated that this natural system of protein protection could be transferred to alfalfa, relatively little PPO activity is required, and a number of odiphenol PPO substrates are effective in the process. Although the mechanism of protein protection by PPO-generated o-quinones is not clear, it seems likely the quinones react with nucleophilic sites on cellular proteins resulting in direct inactivation of proteases, modification of proteins such that they become poor substrates for endogenous proteases, or both. PPOgenerated o-quinones also appear to prevent lipid breakdown in ensiled red clover10 and have a positive impact on the lipid profile of products derived from animals fed diets high in red clover.11,12 Red clover leaves accumulate high levels of both PPO activity (as high as 70 nkat mg−1 protein)7,13 and caffeic acid derived o-diphenols phasalic acid and clovamide14,15 (Winters A, personal communication) that seem to be good substrates for the predominant foliar PPO enzymes.16,17 In contrast to red clover, alfalfa, including a collection of almost 200 perennial Medicago accessions, has little if any PPO activity in its leaves and lacks significant levels of o-diphenol PPO substrates.13,17 – 19 Here we have identified and characterized an alfalfa PPO gene (MsPPO1) to determine whether alfalfa’s endogenous PPO enzyme might be exploited for preserving forage protein. EXPERIMENTAL Plant materials A highly regenerable clone of Regen-SY20 was used for alfalfa (Medicago sativa L.) transformation. Additionally, alfalfa expressing red clover PPO1 and control alfalfa transformed with pILTAB 357 empty vector were the same Regen-SY background.13 Transformed alfalfa was maintained in a growth chamber at 26 ◦ C with 16 h d−1 of approximately 3000 lx illumination. Alfalfa clone P derived from variety ‘Blazer XL’21 used in some experiments was maintained in a greenhouse year-round at 20–30 ◦ C with light intensities between 25 000 and 60 000 lx. Supplemental lighting (13 h d−1 ) was used during all but summer months. All plants were fertilized weekly with Peter’s soluble 20-20-20 (Scott’s, Marysville, OH). DNA and RNA methodologies Preparation and characterization of nucleic acids Genomic DNA from young leaves of alfalfa plants, plasmid DNA, and lambda DNA were prepared using the DNeasy Plant Maxi Kit, QIAprep Spin Miniprep Kit, and Lambda Midi Kit, respectively (Qiagen, Valencia, CA, USA). For DNA blotting, J Sci Food Agric 88:1406–1414 (2008) DOI: 10.1002/jsfa DNA digested with restriction endonucleases was fractionated on agarose gels and transferred to nylon membranes essentially as described by Sambrook et al.22 DNA sequence was determined by cycle sequencing using Big Dye v3.1 (Applied Biosystems, Foster City, CA, USA) and run on ABI automated sequencers by the University of Wisconsin Biotechnology Center. Sequence analyses were carried out using the Wisconsin Package, Version 10 (Accelrys, San Diego, CA, USA), ChloroP23 and SignalP24 algorithms available online at http://www.cbs.dtu.dk/services, and BLAST using the National Center for Biotechnology Information (www.ncbi.nlm.nih.gov) and the Institute for Genome Research (www.tigr.org) web sites. Total RNA was prepared from plant tissues using an RNeasy kit (Qiagen) or by the method of Chang et al.25 Formaldehyde agarose gels were run and RNA blots prepared as previously described.26 DNA and RNA hybridizations were carried out as previously described,13 except that in some experiments 32 Plabeled riboprobes, prepared using the Strip-EZ RNA Kit (Ambion, Austin, TX, USA), were used instead of 32 P-labeled DNA probes. Generation of PPO gene fragments by polymerase chain reaction (PCR) A previously described primer pair corresponding to the conserved region flanking the CuA copper binding site of plant PPOs13 was used to amplify a PPO gene fragment from an alfalfa DNA preparation. The primers were 5′ GGGGAATTCCAACAAGCTARKRTHCATTG TGCTT-3′ (sense) and 5′ -GGGAAGCTTATCC CAATTCCARWAHGG-3′ (antisense). A series of three G deoxyribonucleotides and either an EcoRI or HindIII restriction endonuclease site (underlined) were included in each primer to facilitate subsequent cloning. These primers (50 pmol) and alfalfa clone P genomic DNA (50 ng) were used in a 50 µL PCR reaction using Taq polymerase (Promega Corp., Madison, WI, USA) in the supplied reaction buffer supplemented with 2 mmol L−1 MgCl2 for 30 cycles of 30 s at 94 ◦ C, 30 s at 50 ◦ C, and 1 min at 72 ◦ C. The resulting DNA fragment was digested with EcoRI and HindIII and cloned into pBluescript SKII(−) (Stratagene, La Jolla, CA, USA) digested with the same enzymes using standard methodologies.22 PCR of full-length MsPPO1 from alfalfa cDNA was accomplished using the primers 5′ -ATGGCATCTA TCTCACCCCTTG-3′ (sense) and 5′ -TCAATCTT CAAGCTCTATC-3′ (antisense), KlenTaq LA (Sigma, St Louis, MO, USA) and cDNA template prepared from total RNA. cDNA was prepared using Superscript III reverse transcriptase according to the manufacturer’s protocol (Invitrogen, Carlsbad, CA, USA) from DNase I-treated total RNA isolated from MsPPO1-expressing transgenic alfalfa (described in detail below). 1407 ML Sullivan, RD Hatfield, DA Samac Library screening Approximately 5 × 105 phage from an alfalfa genomic library (derived from cultivar Saranac) in the Lambda DASH II vector27 were plated and lifted to nylon or nitrocellulose filters using standard protocols,22 then screened by hybridization with a cloned 196 bp PPO fragment derived by PCR (described above). This screening procedure was repeated once for a total of 106 phage screened. Overexpression of MsPPO1 in transgenic alfalfa An MsPPO1 overexpression construct was prepared in a manner analogous to that previously described for red clover PPO genes13 using the primers 5′ GGGGAATTCAAACAATGGCATCTATCTCACCCCTTG-3′ (sense) and 5′ -GGGGAATTCAGATCTTCAATCTTCAAGCTCTATC-3′ (antisense) to generate an MsPPO1 coding region fragment by PCR from the cloned gene. These primers incorporated the proposed dicot consensus sequence AAACA28 immediately upstream of the initiating Met codon and EcoRI restriction sites (underlined) to facilitate cloning behind the cassava vein mosaic virus (CsVMV) promoter of the pILTAB 357 plant transformation vector.29 The cloned insert was sequenced to ensure that no mutations were introduced that would alter the sequence of the translated protein. The MsPPO1 construct was transferred to Agrobacterium tumefaciens strain LBA4404 and transformed into a Regen-SY alfalfa clone as previously described.13,30,31 Putative transformants were initially screened for the presence of the npt II gene by PCR as previously described.32 Analysis of PPO-mediated browning, PPO activity, protein accumulation, and proteolytic inhibition For PPO activity and immunoblots, leaf tissues were extracted with 100 mmol L−1 CH3 COONH4 , 20 mmol L−1 Tris, pH 7.5 (3 mL g−1 fresh weight) and protein content determined as previously described.13 For immunoblotting, a protease inhibitor cocktail (P9599, Sigma) was included in the extraction buffer according to the supplier’s instructions. Flower and seed pod protein samples were prepared by phenol extraction as previously described.13 Browning assays were carried out on plant leaf extracts by addition of 100 mmol L−1 o-diphenol solution in ethanol to 3 mmol L−1 final concentration or ethanol to 3% (v/v) as a negative control followed by incubation at room temperature. PPO activity assays were carried out essentially as described by Esterbauer et al.33 as previously detailed,13 except assays were carried out in 0.1× McIlvaine’s citrate–phosphate buffer, pH 7.0 (15 mmol L−1 Na2 HPO4 , 2.3 mmol L−1 citric acid). For comparison of different PPO substrates, data for a given extract were normalized to the activity for catechol. Immunoblotting was carried out using antiserum raised against red clover PPO1 as previously described.13 Proteolysis in leaf extracts 1408 (protein content adjusted to 2 mg mL−1 protein with extraction buffer) was determined by measuring the release of tricholoroacetic acid (TCA, 50 g L−1 ) soluble amino acids and peptides over time using previously described procedures.7 For quantitative PPO activity and proteolytic inhibition experiments, the results using two extracts prepared from different leaf tissue samples harvested on different days were averaged and error reported as standard error of the mean (SEM). For comparison of two sample means, statistical significance was determined using the t-test. For comparison of more than two sample means, data were subjected to single-factor ANOVA using the statistical package of Excel (Microsoft Corp., Redmond, WA, USA) and Tukey’s HSD post hoc test. RESULTS AND DISCUSSION Cloning and sequencing of a PPO gene from alfalfa Degenerate primers flanking the conserved CuA copper binding site of several previously cloned plant PPO genes, and previously used to amplify approximately 200 bp PPO gene fragments from red clover cDNA,13 amplified a DNA fragment of the expected size from alfalfa clone P DNA. The 196 bp fragment was cloned and sequenced (Genbank accession EU168793). The resulting nucleotide and predicted protein sequences (excluding sequence derived from the PCR primers), were used in BLASTn and tBLASTn searches of the NCBI database. Sequences with the highest probability matches corresponded to PPO genes from a wide variety of plant species, including red clover, Ipomoea batatas L., Malus domestica Borkh., and Triticum aestivum L., with most showing 60–70% amino acid identity. These results suggested the gene fragment derived from the PCR reaction corresponds to a polyphenol oxidase gene. Interestingly, no high identity matches (i.e., >80%) with Medicago truncatula Geartn. ESTs or genomic sequences were identified. The 196 bp PCR fragment was used as a hybridization probe to screen an alfalfa genomic library.27 A single hybridizing clone with an approximately 12 kbp insert was obtained following a screen of approximately 5 × 105 lambda clones. By Southern blotting of restriction enzyme-digested clone DNA followed by hybridization with the PCR-derived probe, an approximately 6 kbp EcoRI fragment containing the hybridizing DNA sequence was isolated and subcloned into a plasmid cloning vector. Sequence analysis indicated the 6 kbp EcoRI fragment contains an entire PPO coding region as well as 1.2 kbp of upstream sequence (Genbank Accession AY283062). This alfalfa PPO coding region contains no introns, typical of PPO genes isolated from dicotyledonous plants.34 The gene (designated MsPPO1) is predicted to encode a protein of 607 amino acid residues (68.5 kDa) (Fig. 1). Based on ChloroP23 and Signal P24 algorithms, the J Sci Food Agric 88:1406–1414 (2008) DOI: 10.1002/jsfa Cloning and characterization of an alfalfa polyphenol oxidase gene Figure 1. Predicted amino acid sequence encoded by an alfalfa PPO gene, MsPPO1. Predicted cleavage sites for chloroplast transit and thylakoid lumen signal peptides are indicated by open and filled arrowheads, respectively. The conserved CuA and CuB copper binding sites (starting at amino acids 205 and 363, respectively) are underlined. protein is predicted to be localized to the thylakoid lumen following cleavage of a 69 amino acid transit peptide and 26 amino acid signal peptide. Localization of the protein to the thylakoid lumen would be similar to that observed for nearly all characterized PPO proteins. If the alfalfa PPO protein is targeted to the chloroplast thylakoid lumen as predicted, the mature protein would be 512 amino acid residues (58.0 kDa). The sequence of MsPPO1 isolated from the library had only limited nucleotide sequence identity (70%, excluding primer sequence) with the PCR-derived gene fragment used for the screen. Screening of an additional 5 × 105 genomic clones from the alfalfa genomic library using the 196 bp PCR fragment failed to identify a clone containing its corresponding gene, although three additional clones identical or nearly identical to MsPPO1 were identified. This result suggests that the original PCR product amplified from clone P alfalfa DNA was amplified from contaminating DNA in the original DNA preparation or PCR reactions, represents a PCR artifact, or is derived from a divergent alfalfa gene not present in the genomic library that was screened. That we were unable to identify any potential homologs in searches of M. truncatula sequence databases nor detect hybridzing fragments in Southern blotting experiments with clone P DNA (data not shown) suggests the 196 bp PCR fragment does not represent an actual alfalfa PPO sequence. Using the 1824 bp coding region sequence from the full-length MsPPO1 clone in BLASTn searches of the NCBI Genbank database we identified an M. truncatula BAC clone (Genbank accession AC157507, clone mt2-78b21) containing two tandem PPO genes with a high degree of sequence similarity to MsPPO1. The first gene (from position 53 210 to 51 387) and the second gene (from position 59 385 to 57 560) are 95% and 88% identical to MsPPO1, respectively. The two M. truncatula genes are 89% identical to each other. We currently have no evidence of tandemly arranged PPO genes in M. sativa, although Southern blotting experiments using DNA derived from clone P and Regen SY27 alfalfa genotypes indicate the presence of multiple PPO genes (data not shown). In addition, a J Sci Food Agric 88:1406–1414 (2008) DOI: 10.1002/jsfa BLASTn search of the TIGR M. truncatula gene index with the full-length MsPPO1 coding region identified a tentative consensus sequence (TC101697) sharing 91% sequence identity with MsPPO1. Interestingly, TC101697 is not identical to either of the PPO genes contained on mt2-78b21, although sequence comparisons suggest TC101697 is comprised of cDNAs originating from both genes. MsPPO1 shares a high degree of sequence identity (82% over the coding region) with Vicia faba PPO (Genbank accession Z11702). MsPPO1 shares lower sequence identity (70%) with PPO cDNAs and genomic clones from red clover (Genbank accessions AY017302, AY017303, AY01704, EF183483, EF183484), suggesting the red clover PPOs may have diverged from the other legume enzymes. MsPPO1 is expressed in seedpods and flowers We analyzed MsPPO expression in alfalfa by RNA blotting and hybridization with a probe derived from MsPPO1. Hybridization of a blot containing RNA from young leaves, mature leaves, stems, flower buds, flowers, seed pods 2 weeks post pollination, and 3-day-old seedlings was carried out using an in vitro-labeled RNA corresponding to a 1.5 kbp XbaI–HindIII fragment of the MsPPO1 coding region. Hybridization and washes were at moderate stringency (approximately 25 ◦ C below probe-target Tm calculated according to Sambrook et al.22 ) to detect MsPPO1 as well as any related PPO sequences. RNA from leaves of transgenic alfalfa expressing MsPPO1 under the control of a constitutive promoter (detailed below) served as a positive control. Under the hybridization and wash conditions used, PPO transcripts of the expected size (approximately 2 kb) were detected in developing seed pods, and to a lesser extent flowers and flower buds (Fig. 2). No transcripts of the approximate expected size of an MsPPO1 transcript were apparent in lanes corresponding to leaves, stems, and seedlings. Although faint higher and lower molecular weight bands are present in several of the lanes, these bands are likely an artifact because their locations on the blot correspond to ribosomal RNA bands that had been visualized by ethidium 1409 ML Sullivan, RD Hatfield, DA Samac Figure 2. Expression of MsPPO1 as detected by RNA blot hybridization. Total RNA (5 µg) from Regen SY alfalfa young leaves (Y. Leaf), mature leaves (M. Leaf), stems, flower buds (F. Bud), flowers, seed pods (Pod), or 3-day seedlings was used to make an RNA blot. Seed pods (2 weeks post-pollination) and seeds were produced by self-pollination. RNA (1 µg) from leaves of genetically modified MsPPO1-alfalfa (GM Leaf) served as a positive control. Migration position of a 1.9 kb ribosomal RNA (detected on the ethidium stained gel, not shown) is indicated by an arrow. To facilitate comparison of expression levels, a 50-fold shorter exposure of the GM Leaf sample is shown. bromide staining of the original gel. These results are consistent with a more limited expression analysis that we previously carried out on RNA derived from tissues of clone P alfalfa,16 in which PPO mRNA was detected in seed pods and to a lesser extent flowers, but not in leaves, stems, or apical shoots. The RNA blotting results are also consistent with M. truncatula EST data. Of 14 M. truncatula EST sequences with nucleotide identity ≥90% to MsPPO1, seven are from a pod wall library (representing 0.2% of the ESTs from this library). Of the remaining seven ESTs, four are from a drought-stressed plantlet library (representing 0.05% of the ESTs from this library), and one each from developing leaf, insect herbivory, and virus-infected leaf libraries (representing approximately 0.01% of the ESTs from each of these libraries). Together these results suggest that under normal growing conditions PPO genes are expressed at exceedingly low levels, if at all, in alfalfa leaves. MsPPO1 has low activity in leaves of genetically modified alfalfa Because alfalfa leaves and stems fail to express MsPPO1 mRNA at levels detectible by RNA blotting and alfalfa leaves have little if any detectible endogenous PPO activity,13,17 – 19 we decided to characterize the enzymatic activity of MsPPO1 by ectopically expressing it in alfalfa leaves. Creation of genetically modified alfalfa expressing MsPPO1 in its leaves would also allow us to test whether endogenous alfalfa PPO might be useful in preventing postharvest proteolytic losses if it were expressed in leaves. The entire coding region of MsPPO1 including plastidtargeting signals was inserted behind the CsVMV promoter29,35 in a plant transformation vector, the resulting construct was transformed into alfalfa, and 16 independent alfalfa transformants containing the npt II selectable marker gene were identified. Protein extracts 1410 were made from leaves of the alfalfa plants transformed with the MsPPO1 gene and used in quantitative assays of PPO activity. Caffeic acid was used as a substrate in this screening experiment since it had proven to be among the best substrates for red clover PPOs.16,18 A leaf extract of alfalfa transformed with empty vector (control alfalfa) served as a negative control. For the 16 MsPPO1-transformed alfalfa analyzed in this experiment, measured PPO activities were indistinguishable from that of control alfalfa (0.04 nkat mg−1 in this experiment). Because enzyme activities in leaf extracts of MsPPO1-transformed alfalfa were not detectible above background in the quantitative assay, we screened the plants for enzyme activity using a more sensitive qualitative assay. MsPPO1transformed alfalfa or control alfalfa leaf extracts were prepared and incubated at room temperature in the presence of catechol, caffeic acid, or (−)-epicatechin, and color changes in the extracts characteristic of PPO activity were observed (Fig. 3). These o-diphenols were chosen because they are structurally diverse, representing the simplest o-diphenol and two distinct classes of naturally occurring o-diphenols (caffeic acid derivatives and flavanol o-diphenols). Following o-diphenol addition and incubation at 25 ◦ C, for four of the MsPPO1-transformed alfalfa plants, color changes were apparent after 1–6 h for all three substrates. Maximum color development was observed at 6–24 h. Extracts of the remaining MsPPO1transformed and control alfalfa extracts failed to exhibit discernible color changes, even after 2-day incubations. For the four MsPPO1-transformed alfalfa plants that did have PPO activity based on this assay (hereafter referred to as MsPPO1-alfalfa), the length of time for color changes to become apparent were substantially longer (several hours) than that previously observed in leaf extracts of alfalfa expressing TpPPO genes (2–5 min).13 Interestingly, we observed slight darkening of MsPPO1-alfalfa leaf extracts following a 3-day incubation in the absence of added o-diphenol compared to extracts of control alfalfa (no PPO transgene) or alfalfa expressing TpPPO1 (data not shown), although the biological significance of this observation is unclear. A plant consistently showing the most rapid extract color changes in the presence of added o-diphenol substrate, and hence the highest levels of PPO activity, was used for the analyses detailed below. MsPPO1 protein accumulates as a truncated form in genetically modified alfalfa leaves The relatively low PPO activity detected in leaves of MsPPO1-alfalfa could be due to low-level expression of the transgene, poor accumulation of the active gene product, or poor activity against the tested substrates or under the conditions of the assay. To avoid the first possibility, MsPPO1 was expressed from the CsVMV promoter, which has been shown to direct high-level expression of transgenes in alfalfa.29,35 An appropriately sized approximately 2 kb transcript corresponding to the MsPPO1 transgene was easily detected J Sci Food Agric 88:1406–1414 (2008) DOI: 10.1002/jsfa Cloning and characterization of an alfalfa polyphenol oxidase gene Figure 3. Qualitative assay for PPO activity. Extracts of leaves of MsPPO1- or control alfalfa (Cont.) were supplemented (+) with the indicated o-diphenols at 3 mmol L−1 . Extracts are shown following a 24 h incubation at 25 ◦ C. A non-supplemented MsPPO1-alfalfa extract (−) incubated for 24 h is shown for comparison. Figure 4. Immunological detection of MsPPO1 protein. An immunoblot of 10 µg protein from untransformed Regen SY alfalfa leaves, flowers, or seed pods; or leaves of MsPPO1-alfalfa (GM Leaf) was developed with anti-TpPPO1 antiserum. Red clover PPO1 expressed in Escherichia coli (TpPPO1, approximately 5 ng) served as a positive control.13 An asterisk indicates the full-length mature (lacking chloroplast targeting signals) TpPPO1 used to produce the antiserum. Lower molecular weight bands in the TpPPO1 lane correspond to truncation products that accumulate when the protein is overexpressed in E. coli (Sullivan M, unpublished result). in leaves of MsPPO1-alfalfa (Fig. 2), indicating the MsPPO1 transgene is well expressed at the mRNA level, accumulating to approximately 250-fold higher levels than the endogenous transcript in seed pods. To make sure that the MsPPO1 protein was accumulating in transgenic plants, leaf extracts were fractionated on SDS-PAGE and analyzed by immunoblotting using anti-TpPPO1 antiserum,13 which cross-reacts with PPOs from several dicot (Sullivan M, unpublished data) and monocot (Anderson JV and Marita J, personal communication) species. As shown in Fig. 4, a J Sci Food Agric 88:1406–1414 (2008) DOI: 10.1002/jsfa faint band running at approximately 54 kDa is associated with expression of the MsPPO1 transgene and is close to the expected size (58 kDa) of the mature, plastid targeted MsPPO1 gene product. However, a much more intense band associated with expression of the MsPPO1 transgene corresponds to an approximately 42 kDa protein. It is unlikely that this 42 kDa protein represents post-isolation cleavage of the expected mature protein product since a protease inhibitor cocktail was included in the buffer used to prepare this protein extract. A faint band <18 kDa in size is also present in the MsPPO1-alfalfa leaf extract and could be the other cleavage product derived from the expected full-length MsPPO1 protein. To rule out the possibility that the MsPPO1 expression construct contained a premature stop codon, we resequenced the MsPPO1 coding region in the binary vector used for the alfalfa transformation. We also sequenced a PCR fragment of the entire MsPPO1 coding region derived from MsPPO1-alfalfa leaf cDNA and found no coding region mutations had been introduced during the transformation process. These results suggest the 42 kDa form of MsPPO1 in the transgenic plants is a truncated form produced in vivo from the full-length protein. This 42 kDa form could represent the mature form of the protein since similar truncation products have been purified as active PPO from other plant species.36 We have recently demonstrated that for red clover PPOs expressed in alfalfa post-isolation proteolytic cleavage in alfalfa leaf extracts to a 45 kDa form may be partly responsible for enzyme activation.17 Alternatively, the 42 kDa form of the alfalfa protein could be the result of aberrant processing due to high-level ectopic expression in leaves. If the protein is being aberrantly processed, this might explain the relatively low activity levels observed despite high levels of MsPPO1 mRNA and protein detected in the genetically modified alfalfa. In extracts of leaves from control plants not transformed with MsPPO1, a faint band of approximately 36 kDa molecular weight cross-reacting with the TpPPO1 antibody was detected. It is not clear if this represents a breakdown product of an endogenous PPO protein or an unrelated alfalfa protein that fortuitously contains an epitope recognized by the red clover PPO antiserum. Given the lack of MsPPO1 mRNA detected in leaves, the latter explanation seems more likely. We also examined extracts of alfalfa flowers and seed pods, since these were the only tissues in which we identified PPO1 mRNA. Faint bands co-migrating with recombinant TpPPO1 (apparent molecular mass of 65 kDa) and cross-reacting with TpPPO1 antiserum were detected. It is not clear if these bands are the proteins translated from the mRNA observed in Fig. 2, since the immunoblot band intensities for flowers and seed pods are similar but the corresponding mRNA levels are different. We cannot rule out the possibility that the bands seen on the immunoblot for flower and pod proteins are not related to PPO, and that actual 1411 ML Sullivan, RD Hatfield, DA Samac alfalfa PPO protein in these tissues is below the limit of detection in this experiment. MsPPO1 activity is enhanced by sodium dodecyl sulfate Many PPO enzymes are activated by small amounts of detergent or other denaturants. These agents presumably cause conformational changes in the enzyme that allow substrates better access to the active site.37 We assayed leaf extracts from control and MsPPO1-alfalfa for PPO activity with caffeic acid, (−)-epicatechin, and catechol, in standard assay buffer or assay buffer supplemented with 8.7 mmol L−1 sodium dodecyl sulfate (SDS) to test for activation (Table 1). As in the screening experiment described above, in standard assay buffer (lacking SDS), PPO activity measurements for extracts of MsPPO1-alfalfa could not be distinguished from those of control alfalfa for any of the tested substrates (P > 0.05). This result underscores the limitations of this quantitative assay with extremely low activity levels, since the MsPPO1-alfalfa extracts could mediate the substrate-dependent color changes described above, whereas control alfalfa extracts could not. Addition of SDS to the assay buffer resulted in a slight increase in enzyme activity measurements made for extracts of control alfalfa with catechol (P < 0.025), but values for caffeic acid and (−)-epicatechin were unchanged by SDS addition (P > 0.3). For control alfalfa extracts it is unclear what any of these background values represent, since extract browning is not seen in the presence of PPO substrate, even with SDS addition (data not shown), and although these background assay values are extract-dependent, addition of more or less extract does not result in the expected corresponding changes in measured activity (data not shown). In contrast to control alfalfa extract, MsPPO1-alfafla extracts showed substantial increases in PPO activity when assayed in the presence of SDS. A high degree of variability was observed in the SDS-activated measurements for MsPPO1-alfalfa leaf extracts prepared from different tissue samples, presumably due to uncontrolled environmental variables that affected enzyme accumulation. This high degree of variability resulted in a relatively high Pvalue for the difference between non- and SDS-treated MsPPO1 alfalfa extracts (0.07 < P < 0.09) despite the obvious increase in activity. We have previously observed similar high levels of variation when working with transgenic alfalfa expressing red clover PPO genes from the CsVMV promoter, although normalizing activity data to a given substrate greatly reduces variability between extracts of different tissue samples expressing the same PPO.7,13,17 Using this approach allowed us to evaluate the relative substrate specificities for MsPPO1 (Table 2). SDS-treated MsPPO1 activity for catechol and (−)-epicatechin was approximately the same, although catechol may be favored (P = 0.08), and about 10-fold higher than for caffeic acid (P < 0.01). It is unclear what the natural substrates for MsPPO1 might be, but with its preference for a flavanoid o-diphenol and seedpod expression pattern, a role in the formation of condensed tannins during seed development is an intriguing possibility (e.g., see Dixon et al.38 ). Interestingly, the substrate preference of activated MsPPO1 is different from those of characterized red clover PPO gene products, which seem to prefer caffeic acid over (−)-epicatechin and catechol.16,17 With the SDS-induced increase in the enzyme activity measured for MsPPO1-alfalfa leaf extracts, PPO activity levels approach those of freshly prepared (and not SDS-activated) leaf extracts of alfalfa expressing TpPPO1 from the CsVMV promoter (up to approximately 5 nkat mg−1 ), although additional activation of the clover enzyme in alfalfa extracts results in a 5- to 10-fold increase in PPO activity.13,17 This finding suggests the truncated form of the protein accumulating in MsPPO1-alfalfa leaves represents the active form of the enzyme. MsPPO1 inhibits proteolysis in alfalfa extracts in the presence of catechol Previous studies utilizing red clover PPO1 (TpPPO1) ectopically expressed in alfalfa indicated that even relatively low levels of PPO activity can inhibit postharvest proteolysis in alfalfa leaf extracts.7 To determine whether MsPPO1 is capable of inhibiting Table 2. Relative preference of SDS-activated MsPPO1 for various substrates Percent maximum activitya Substrate Catechol (−)-Epicatechin Caffeic acid 100a 74 ± 9a 8±2 a Average of two experiments ± SEM. For each experiment, activities for each substrate were corrected for background activity measured in control alfalfa extract and normalized to the catechol value. Difference marked with ‘a’ is significant at P = 0.08. Other differences are significant at P < 0.01. Table 1. PPO activitiesa (nkat mg−1 ) in MsPPO1- and control alfalfa leaf extracts without and with SDS Without SDS Substrate Catechol (−)-Epicatechin Caffeic Acid a With SDS MsPPO1 Control MsPPO1 Control 0.09 ± 0.01b 0.06 ± 0.03b 0.05 ± 0.04b 0.13 ± 0.01a 0.11 ± 0.01 0.07 ± 0.05 4.22 ± 1.69b 3.13 ± 1.49b 0.43 ± 0.17b 0.22 ± 0.02a 0.15 ± 0.03 0.07 ± 0.05 Average of two experiments ± SEM. Differences within a row marked with the same letter are significant at P < 0.025 (a) and P < 0.10 (b). 1412 J Sci Food Agric 88:1406–1414 (2008) DOI: 10.1002/jsfa Cloning and characterization of an alfalfa polyphenol oxidase gene Table 3. Proteolysis in control, MsPPO1-, and TpPPO1-alfalfa extracts in the presence of various o-diphenols Amino acid release (4 h), µmol mg−1a Extract Control MsPPO1 TpPPO1 Amino acid release (20 h), µmol mg−1a Caffeic (−)-Epi Catechol Caffeic (−)-Epi Catechol 0.74 ± 0.07a 0.69 ± 0.04a 0.09 ± 0.02 0.68 ± 0.02a 0.57 ± 0.11a 0.09 ± 0.02 0.75 ± 0.01 0.45 ± 0.01 0.13 ± 0.01 1.50 ± 0.07a 1.41 ± 0.08a 0.21 ± 0.03 1.47 ± 0.05a 1.35 ± 0.22a 0.17 ± 0.03 1.43 ± 0.00 1.00 ± 0.05 0.32 ± 0.02 a Average of two experiments ± SEM. Data marked with the same letter within a column are not significantly different (P > 0.56). All other differences within a column are significant (P < 0.025). proteolysis in alfalfa extracts in the presence of PPO substrate, leaf extracts of control and MsPPO1-alfalfa were incubated at 37 ◦ C in the presence of 3 mmol L−1 caffeic acid, (−)-epicatechin, or catechol. An extract of alfalfa expressing TpPPO17 served as a positive control. These assays were carried out without SDS activation, since during ensiling such extraordinary steps would be impractical. Proteolysis was measured as release of TCA-soluble amino acids after 4 and 20 h (Table 3). Trends at the 4 h and 20 h time points were similar. As previously observed,7 in the TpPPO1-alfalfa extract proteolysis was reduced by >75% in the presence of any of the tested o-diphenols relative to a control extract. While no significant reduction in proteolysis was seen for MsPPO1alfalfa extracts in the presence of either caffeic acid or (−)-epicatechin (P > 0.5), a reduction in amino acid release was observed when catechol was the o-diphenol used. Although the reduction in proteolysis seen for MsPPO1 was not as great as that seen for TpPPO1, it was substantial given the relative lack of quantifiable PPO activity of these extracts. This finding underscores our previous observation that even small amounts of PPO activity are capable of inhibiting postharvest proteolysis.7 Although (−)-epicatechin is almost as good a substrate as catechol for SDS-treated MsPPO1, it is unclear why it failed to significantly reduce proteolysis in MsPPO1-alfalfa extracts. Many factors may influence the efficacy with which a given PPO-generated o-quinone results in proteolytic inhibition (e.g., rate of quinone formation, relative reactivity towards various nucleophiles, half-life, crosslinking potential). Until the roles these parameters play in proteolytic inhibition are better understood, it may be difficult to fully predict optimal PPO/quinone combinations. The high levels of enzyme activity available from TpPPO1 without additional activation may be sufficient to allow proteolytic inhibition with a wide variety of o-diphenol substrates.7 Unfortunately, the relative lack of enzymatic activity and/or limited substrate specificity of MsPPO1 make it, overall, less effective at preventing postharvest proteolysis than its red clover counterpart. CONCLUSIONS The properties of the MsPPO1 gene (e.g., natural expression pattern limited to flowers and seed pods, poor activity of the encoded enzyme in the absence J Sci Food Agric 88:1406–1414 (2008) DOI: 10.1002/jsfa of activation against naturally occurring substrates most likely to be available for an ensiling process) make it a poor target for development of a PPO/ o-diphenol-based system of protein protection in alfalfa. These results are consistent with a recent screen of approximately 200 accessions of perennial Medicago species for useful levels of foliar PPO activity against caffeic acid, (−)-epicatechin, and catechol.19 In that screen, no accessions were identified with potential for development of a PPO-based protein preservation system. Together, these results suggest development of a PPO-based protein protection system in alfalfa will likely require the use of transgenic approaches (e.g., use of TpPPO1 as a transgene). Additionally, the results reported here underscore the uniqueness of the red clover PPOs among legume species with respect to sequence, expression pattern, and enzymatic activities. ACKNOWLEDGEMENTS We thank Merici Evans Awe, Sara Zerbel, and Mindy Dornbusch for excellent technical assistance, and Dr George Schmitz for helpful comments on the manuscript. Mention of trade names or commercial products in this article is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the US Department of Agriculture. REFERENCES 1 Sherman TD, Gardeur TL and Lax AR, Implications of the phylogenetic distribution of polyphenol oxidase in plants, in Enzymatic Browning and its Prevention, ed. by Lee CY and Whitaker JR. American Chemical Society, Washington, DC, pp. 103–119 (1995). 2 Li L and Steffens JC, Overexpression of polyphenol oxidase in transgenic tomato plants results in enhanced bacterial disease resistance. Planta 215:239–247 (2002). 3 Thipyapong P and Steffens JC, Tomato polyphenol oxidase: differential response of the polyphenol oxidase F promoter to injuries and wound signals. Plant Physiol 115:409–418 (1997). 4 Nakayama T, Soto T, Fukui Y, Yonekura-Sakakibara K, Hayashi H, Tanaka Y, et al, Specificity analysis and mechanism of aurone synthesis catalyzed by aureusidin synthase, a polyphenol oxidase homolog responsible for flower coloration. FEBS Lett 499:107–111 (2001). 5 Cho MH, Moinuddin SGA, Helms GL, Hishiyama S, Eichinger D, Davin LB, et al, (+)-Larreatricin hydroxylase, an enantiospecific polyphenol oxidase from the creosote bush (Larrea tridentata). Proc Natl Acad Sci USA 100:10 641–10 646 (2003). 1413 ML Sullivan, RD Hatfield, DA Samac 6 Vamos-Vigyazo L, Polyphenol oxidase and peroxidase in fruits and vegetables. CRC Crit Rev Food Sci Nutr 15:49–127 (1981). 7 Sullivan ML and Hatfield RD, Polyphenol oxidase and o-diphenols inhibit postharvest proteolysis in red clover and alfalfa. Crop Sci 46:662–670 (2006). 8 Rotz CA, Pitt RE, Muck RE, Allen MS and Buckmaster DR, Direct-cut harvest and storage of alfalfa on the dairy farm. Trans ASAE 36:621–628 (1993). 9 Papadopoulos YA and McKersie BD, A comparison of protein degradation during wilting and ensiling of six forage species. Can J Plant Sci 63:903–912 (1983). 10 Lee MRF, Winters AL, Scollan ND, Dewhurst RJ, Theodorou MK and Minchin FR, Plant-mediated lipolysis and proteolysis in red clover with different polyphenol oxidase activities. J Sci Food Agric 84:1639–1645 (2004). 11 Dewhurst RJ, Shingfield KJ, Lee MRF and Scollan ND, Increasing the concentrations of beneficial polyunsaturated fatty acids in milk produced by dairy cows in high-forage systems. Anim Feed Sci Technol 131:168–206 (2006). 12 Lee MRF, Minchin FR, Hatfield RD and Sullivan ML, Red clover polyphenol oxidase reduces ruminal lipolysis in in vitro batch culture, in Proceedings of the British Society of Animal Science, Southport, UK, p. 25 (2007). 13 Sullivan ML, Hatfield RD, Thoma SL and Samac DA, Cloning and characterization of red clover polyphenol oxidase cDNAs and expression of active protein in Escherichia coli and transgenic alfalfa. Plant Physiol 136:3234–3244 (2004). 14 Yoshihara T, Yousikawa H, Sakamura S and Sakuma T, Clovamides; L-DOPA conjugated with trans- and cis-caffeic acids in red clover (Trifolium pratense). Agric Biol Chem 38:1107–1109 (1974). 15 Hatfield RD and Muck R, Characterizing proteolytic inhibition in red clover silage, in XIIth International Silage Conference, Swedish University of Agriculture and Sciences, Upsala, Sweden, pp. 147–148 (1999). 16 Sullivan M, Thoma S, Samac D and Hatfield R, Cloning of red clover and alfalfa polyphenol oxidase genes and expression of active enzymes in transgenic alfalfa, in Molecular Breeding of Forage and Turf, ed. by Hopkins A, Wang ZY, Mian R, Sledge M and Barker R. Kluwer, Dordrecht, pp. 189–195 (2004). 17 Schmitz GE, Sullivan ML and Hatfield RD, Three polyphenol oxidases from red clover (Trifolium pratense) differ in enzymatic activities and activation properties. J Agric Food Chem 56:272–280 (2008). 18 Jones BA, Hatfield RD and Muck RE, Screening legume forages for soluble phenols, polyphenol oxidase and extract browning. J Sci Food Agric 67:109–112 (1995). 19 Sullivan ML and Bringe A, A plate-based browning assay to screen alfalfa germplasm for polyphenol oxidase activity and o-diphenols, in 2004 Research Report, US Dairy Forage Research Center, Madison, WI, pp 13–15 (2005). 20 Bingham ET, Registration of alfalfa hybrid Regen-SY germplasm for tissue culture and transformation research. Crop Sci 31:1098 (1991). 21 Dilkova M and Bingham ET, Microsporogenesis of alfalfa cultivars and selected genotypes. http://www.medicagoreports.org/pdfs/microsporogenesis.pdf [17 October 2007]. 1414 22 Sambrook J, Fritsch EF and Maniatis T, Molecular Cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY (1989). 23 Emanuelsson O, Nielsen H and von Heijne G, ChloroP, a neural network-based method for predicting chloroplast transit peptides and their cleavage sites. Prot Sci 8:978–984 (1999). 24 Nielsen H, Engelbrecht J, Brunak S and von Heijne G, Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites. Prot Eng 10:1–6 (1997). 25 Chang S, Puryear J and Cairney J, A simple and efficient method for isolating RNA from pine trees. Plant Mol Biol Rep 11:113–116 (1993). 26 Newman TC, Ohme-Takagi M, Taylor CB and Green PJ, DST sequences, highly conserved among plant SAUR genes, target reporter transcripts for rapid decay in tobacco. Plant Cell 5:701–714 (1993). 27 Gregerson RG, Miller SS, Petrowski M, Gantt JS and Vance CP, Genomic structure, expression and evolution of the alfalfa aspartate aminotransferase genes. Plant Mol Biol 25:387–399 (1994). 28 Joshi CP, Zhou H, Huang XQ and Chiang VL, Context sequences of translation initiation codon in plants. Plant Mol Biol 35:993–1001 (1997). 29 Verdaguer B, de Kochko A, Beachy RN and Fauquet C, Isolation and expression in transgenic tobacco and rice plants, of the cassava vein mosaic virus (CVMV) promoter. Plant Mol Biol 31:1129–1139 (1996). 30 Austin S, Bingham ET, Matthews D, Shahan M, Will J and Burgess RR, Production and field performance of transgenic alfalfa expressing alpha-amylase and manganese-dependent lignin peroxidase. Euphytica 85:381–393 (1995). 31 Samac DA and Austin-Phillips S, Alfalfa (Medicago sativa L.), in Agrobacterium Protocols (2nd edn), ed. by Wang K. Humana Press, Totowa, NJ, pp. 301–311 (2006). 32 Saruul P, Srienc F, Somers DA and Samac DA, Production of a biodegradable plastic polymer, poly-beta-hydroxybutyrate, in transgenic alfalfa. Crop Sci 42:919–927 (2002). 33 Esterbauer H, Schwarzl E and Hayn M, A rapid assay for catechol oxidase and laccase using 2-nitro-5-thio benzoic acid. Anal Biochem 77:486–494 (1977). 34 Gooding PS, Bird C and Robinson SP, Molecular cloning and characterisation of banana fruit polyphenol oxidase. Planta 213:748–757 (2001). 35 Samac DA, Tesfaye M, Dornbusch M, Saruul P and Temple SJ, A comparison of constitutive promoters for expression of transgenes in alfalfa (Medicago sativa). Transgenic Res 13:349–361 (2004). 36 Eicken C, Zippel F, Buldt-Karentzopoulos K and Krebs B, Biochemical and spectroscopic characterization of catechol oxidase from sweet potatoes (Ipomoea batatas) containing a type-3 dicopper center. FEBS Lett 436:293–299 (1998). 37 Jimenez M and Garcia-Carmona F, The effect of sodium dodecyl sulfate on polyphenol oxidase. Phytochemistry 42:1503–1509 (1996). 38 Dixon RA, Xie DY and Sharma SB, Proanthocyanidins: a final frontier in flavonoid research? New Phytol 165:9–28 (2005). J Sci Food Agric 88:1406–1414 (2008) DOI: 10.1002/jsfa J Sci Food Agric 88:1415–1422 (2008) Journal of the Science of Food and Agriculture Application of surface response methodology to optimize hydrolysis of wheat gluten and characterization of selected hydrolysate fractions Silvina R Drago,1,3∗ Rolando J González1 and Marı́a C Añón2,3 1 Instituto de Tecnologı́a de Alimentos, FIQ-UNL, 1◦ de Mayo 3250 (3000) Santa Fe, Argentina calle 47 y 116, La Plata, Argentina 3 CONICET 2 CIDCA Abstract BACKGROUND: The aim of this work was to optimize the hydrolysis experimental conditions to obtain wheat gluten hydrolysates and to characterize fractions from hydrolysates with different hydrolysis degrees in order to develop protein functional ingredients. Surface response methodology was used to analyze the effect of reaction factors on the degree of hydrolysis to assess the conditions for maximum fungal protease enzyme activity. Hydrolysates having three different trichloroacetic acid indices (TCAI) were prepared. Soluble fractions at pH 4, 6.5 and 9 from these hydrolysates were characterized by electrophoresis, reverse-phase high-performance liquid chromatography, free amino group content and peptide chain length. RESULTS: Temperature and pH ranges for highest enzyme activity at 2.5 h were 54–58 ◦ C and 4.2–4.4, respectively. Hydrolysate fraction composition differs according to the hydrolysis degree and extracting pH, the difference being more pronounced at low TCAI. Hydrolysate having 32% TCAI is composed of peptides whose size is lower than 18.5 kDa, with an average peptide chain length of 14 amino acid residues. CONCLUSION: The combination of hydrolysis degree and pH of extraction allows fractions of different peptide composition to be obtained, which could be taken into account when trying to find a defined composition related to determined functional characteristics.  2008 Society of Chemical Industry Keywords: hydrolysis; wheat gluten; enzymes; experimental design; surface response methodology INTRODUCTION Gluten is one of the products obtained from the wheat wet milling process, which consists of several steps: hydration of flour, gluten formation, washing and drying the final vital gluten under mild conditions, which ensures that the unique viscoelastic properties are retained. It is particularly used to improve commercial wheat bread flours of average quality and, also, to reinforce viscoelastic properties in special formulations1,2 that require functional properties related to viscoelasticity or gluten vitality.3 Due to the relative low price of gluten in comparison with other protein ingredients, there is interest in expanding the scope of gluten application by means of structural transformations which could provide other functional properties. Hydrolysis is one of the alternatives that allow protein modification, which can be carried out either by chemical (acid or alkaline), physical, or enzymatic methods, the latter having remarkable advantages over the traditional chemical one.4,5 Although different food-grade proteases can be used, the only information usually available is that offered by the manufacturer, which is limited to conditions of use based on tests carried out on a standard substrate, the structural characteristics of which differ from those corresponding to the protein system under study. By controlling the reaction conditions during enzymatic hydrolysis, hydrolysates of various characteristics can be obtained. The degree of protein hydrolysis depends on the hydrolysate’s expected use,6,7 and low degrees of hydrolysis are required in order to maintain functional properties.8,9 The aim of the present work was to optimize the hydrolysis experimental conditions to obtain wheat gluten hydrolysates using surface response methodology and to characterize fractions from hydrolysates with different hydrolysis degrees in order to develop functional protein ingredients. ∗ Correspondence to: Silvina R Drago, Instituto de Tecnologı́a de Alimentos, FIQ-UNL, 1◦ de Mayo 3250 (3000) Santa Fe, Argentina E-mail: sdrago@fiq.unl.edu.ar (Received 11 October 2007; revised version received 6 February 2008; accepted 7 February 2008) Published online 18 April 2008; DOI: 10.1002/jsfa.3233  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 SR Drago, RJ González, MC Añón EXPERIMENTAL The enzyme used in the experiments, provided by Genencor SA (Arroyito, Córdoba, Argentina), is a fungal protease derived from Aspergillus oryzae (31 000 HU g−1 ). The activity is expressed in hemoglobin units, where 1 HU is that activity which will liberate 0.0447 mg of non-protein nitrogen in 30 min under the conditions of the assay (denatured hemoglobin, pH 4.7; T : 40 ◦ C). The enzyme is a mixture of endo/exopeptidases whose characteristics are as follows: effective pH 3.5–9, optimum pH 4.3–5; temperature range 30–50 ◦ C. Commercial vital gluten, provided by Molinos Semino SA (Carcarañá, Santa Fe, Argentina) was used as a substrate. Gluten composition was as follows: moisture 5.95% (AACC 44-15A method);10 protein (N × 5.7) 77.20% dry basis (d.b.) (Kjeldahl–AACC 46-11 method);10 starch 13.15% d.b. (Ewers polarimetric method); ether extract 0.71% d.b. (AACC 30-25 method);10 and ash 0.834% d.b. (ICC No. 104–IRAM No. 15 851, standard technique). additional two central points). Time was kept constant (2.5 h). The data obtained were modeled by a second-degree polynomial function. In every case, hydrolysis was carried out in a laboratory batch system with temperature and agitation control. Protein concentration was 5% and enzyme/substrate (E/S) ratio 3%, pH being adjusted with HCl or NaOH, as necessary. A parallel reaction blank was carried out. Thermal treatment of vital gluten In order to disperse vital gluten in water and secure a uniform suspension, a moderate thermal treatment was carried out. Vital gluten was placed in screw-cap cylindrical tubes (16 mm diameter) which were placed in a boiling water bath for 15, 25 and 35 min. The sample was then cooled by placing the tubes in a 21 ◦ C water bath. To estimate the thermal treatment intensity, the impact on the viscoelastic properties of each sample was analyzed by means of an alveogram, carried out with a Chopin alveograph (Paris, France), which used a gluten–starch mixture as a model. The thermal treatment time was selected based on the capacity of the treated gluten to form a uniform dispersion, considering the alveogram as an indicator of thermal treatment intensity. Enzyme/substrate ratio The conditions selected from the experimental design were pH 4.25 and temperature 55 ◦ C and substrate concentration [S] was 5%. The E/S ratios under study were 1.87%, 3%, 3.75%, 5% and 10% (w/w). TCAI was determined by using the Lowry et al. technique12 to measure protein concentration. Effect of operating variables on degree of hydrolysis as studied by experimental designs Surface response methodology was used to analyze the effect of reaction factors (pH, temperature and time) on the degree of hydrolysis with two experimental designs.11 Experimental design 1, ‘central composite blocked cube star’, consisted of a total of 20 experiments which included 15 treatments with an additional five central points. Reaction factor levels were selected by taking into account the pH and temperature data supplied by the manufacturer, and the time, which ranged between 2 and 5 h, considering a time variation between each experimental level sufficient to show the difference between each treatment. In order to be more precise in the assessment of maximum enzyme activity for this particular substrate, a second experimental design was used, ‘central composite design 22 + star’, using pH (between 3 and 5) and temperature (between 35 and 55 ◦ C) as independent variables, with a total of 11 experiments (nine treatments with an 1416 Hydrolysis reaction progress The progress of hydrolysis was followed by means of the trichloroacetic acid index (TCAI), using TCA 20% and diluting the sample in a 1:1 ratio. N was measured by the Semimicro–Kjeldahl method. The TCAI, which was used as an indirect measurement of degree of hydrolysis (DH), was calculated as follows: TCAI = [N soluble in TCA (hydrolysate) − N soluble in TCA (blank) × 100]/total aminic N Preparation of hydrolysates Hydrolysates were prepared in a 5 L batch reactor with agitation, using a thermostated bath kept at a constant temperature of 55 ◦ C. HCl (3 mol L−1 ) was added in order to maintain a constant pH of 4.25. A protein concentration of 8% (w/w) and E/S ratio of 5% were used. Hydrolysates were obtained at different reaction times: 31 min, 2 h and 6 h. Enzyme inactivation was carried out at 70 ◦ C for 15 min. The hydrolysates were frozen at −20 ◦ C and lyophilized. Preparation of fractions at different pH In order to obtain hydrolysate fractions at different pH (4, 6.5 and 9), a 2% (w/w, d.b.) solution of the different hydrolysates was prepared.13 The pH was achieved by adding 0.8 mol L−1 HCl or 0.8 mol L−1 NaOH. The samples were stirred for 1 h at room temperature, and then centrifuged for 15 min at 8000 × g at room temperature. The supernatant (the extract at each pH) was frozen and protein content determined using the semimicro–Kjeldahl method. Characterization of soluble fractions at different pH of thermally treated gluten (TTG) and hydrolysates Sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) (EF) Electrophoresis was performed according to Laemmli,14 in a 4–15% gradient buffer system using a J Sci Food Agric 88:1415–1422 (2008) DOI: 10.1002/jsfa Wheat gluten hydrolysates and characterization of selected fractions Mini-Protean II electrophoresis cell (Bio-Rad, Hercules, CA, USA) with a Model 200/2.0 Bio-Rad source. The gel plates were fixed and stained with a solution containing 0.125% Coomassie Blue R250, 50% methanol and 10% acetic acid in water. The pH extracts were mixed with sample buffer containing stacking buffer, glycerol and Bromophenol Blue, using a 1.33 dilution, with or without 5% β-mercaptoethanol. The mixtures were heated in a boiling bath for 90 s and loaded. Reverse-phase high-performance liquid chromatography (RP-HPLC) of extracts The extracts were diluted to a protein concentration (N × 5.7) of 2.5 mg mL−1 . A Sephasil peptide C8 column of 12 µm ST 4.6/250 (Pharmacia Biotech, Piscataway, NJ., USA) was used, together with an auto injector Waters 717 Plus Auto sampler (Millipore, Billerica, MA, USA), a Waters 600 E pump (multisolvent delivery system, Millipore), and a diode array detector (Waters 996, Millipore). Peptides were separated and eluted at 1.1 mL min−1 , at 60 ◦ C, using the following buffers: buffer A: acetonitrile–water 2:98, with 650 µL L−1 trifluoroacetic acid (TFA); buffer B: acetonitrile–water 65:35 with 650 µL L−1 TFA, and detected at 210 nm. Since elution profiles of RP-HPLC can be grouped into categories according to increasing hydrophobicity of the eluted peptides,15 chromatogram analysis was carried out by integrating peak areas in three sections of each chromatogram: (a) components of low molecular weight and low hydrophobicity: 0–20 min of elution range; (b) components of medium molecular weight and medium hydrophobicity: 20–40 min of elution range; (c) components of high molecular weight and high hydrophobicity: 40–60 min of elution range. Results were expressed as a percentage of each section with respect to the total area. Determination of free amine group content The o-phthaldialdehyde (OPA) technique16 was used for this purpose. Free amine group content was used to calculate the number of peptide bonds cleaved during hydrolysis. Estimation of peptide chain length (PCL) PCL can be estimated by means of the following expression:17 PCL (soluble fraction) = 100 × S/(h/htot )% where htot is the total number of peptide bonds in the protein substrate (8.3 mEq g−1 protein), h is the number of peptide bonds cleaved during hydrolysis, and S is the fraction of soluble proteins. J Sci Food Agric 88:1415–1422 (2008) DOI: 10.1002/jsfa Statistical analysis Software Statgraphics Plus 3.0 was used for statistical analysis. The effect of variables on TCAI was analyzed by response surface methodology and the least significant difference (LSD) test was used to determine statistical differences among samples (p < 0.05). RESULTS AND DISCUSSION Thermal treatment of vital gluten Table 1 shows the alveographic test results. A significant increase in gluten strength (P) is observed, even with slight thermal treatment (15 and 25 min). These results are in agreement with those reported by Jeanjean et al.18 and Autran and Berrier.19 Thermal treatment also produced a decrease in total absorbed energy (W ) and in the swelling index (G). Values corresponding to 15 and 25 min samples were similar; however, a 35 min treatment proved excessive, which explains the P drop and also indicates a decrease in dough-forming ability. The increase in P/G ratio at the beginning of the thermal denaturation process is related to the decrease in cohesivity. These results indicate that the decrease in gluten-forming ability caused by heat treatment is better described by G and W . Gluten treated for 25 min produced an alveographic diagram similar to that frequently observed in wheat samples taken from silos with overheating.20 The loss of gluten bread-making quality could be associated with the formation of aggregates between the different protein fractions,21 which could be the result of polymerization by sulfhydryl–disulfur interchange.22 The selected thermal treatment was 25 min, since better gluten dispersion was obtained when compared with that corresponding to 15 min, which seems to be insufficient for an easy dispersion, and also with 25 min treatment gluten is not over-treated as in the case of 35 min treatment. Effect of operating variables on degree of hydrolysis as studied by experimental designs Table 2 shows the results of TCAI corresponding to the central composite blocked cube star experimental design. It is observed that the highest TCAI value was obtained for one extreme of the design (pH 3.8 and 40 ◦ C) and the lowest ones were around pH 6.5 Table 1. Alveographic test Sample Vital gluten 15 min gluten 25 min gluten 35 min gluten W P G P/G 172c 135b 135b 35a 82b 111c 119d 71a 15.5c 11b 10b 8.5a 5.29a 10.1c 11.9d 8.35b W, total absorbed energy; P, strength; G, swelling index; P/G, alveographic equilibrium index. For each column, different letters represent significant differences between the samples (p < 0.05). 1417 SR Drago, RJ González, MC Añón Experiment no. pH Temp. (◦ C) Time (h) TCAI 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 6.5 8.0 8.0 5.0 5.0 6.5 5.0 8.0 8.0 5.0 6.5 6.5 6.5 9.2 6.5 6.5 3.8 6.5 6.5 6.5 40 32 32 48 48 40 32 48 48 32 40 40 25.5 40 54.5 40 40 40 40 40 3.5 2.0 5.0 2.0 5.0 3.5 5.0 2.0 5.0 2.0 3.5 3.5 3.5 3.5 3.5 3.5 3.5 6.21 0.79 3.5 3.29 0.44 1.05 11.20 17.00 2.60 9.65 0.70 1.35 4.62 2.42 2.22 1.07 0.90 5.30 2.85 17.08 3.56 0.73 2.96 Standardized Pareto Chart for TCAI A:pH AA B:Temp. AB C:time AC BB BC CC 0 4 8 12 16 20 24 Standardized effect Figure 1. Pareto chart corresponding to central composite blocked cube star experimental design. [S] = 5%, E/S = 3%; TCAI, trichloroacetic acid index. TCAI Table 2. Results of TCAI corresponding to the central composite blocked cube star experimental design. [S] = 5%, E/S = 3% 30 25 20 15 10 5 0 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 pH 50 55 40 45 35 Temp °C 25 30 Figure 2. Surface response for TCAI as a function of pH and temperature, for 3.5 h, corresponding to central composite blocked cube star experimental design. [S] = 5%, E/S = 3%; TCAI, trichloroacetic acid index. or higher. The Pareto chart (Fig. 1) shows that pH is the variable of highest impact, both in its linear and quadratic term. The effects of the three variables were significant, particularly in the case of the linear terms, the pH quadratic term and the interactions pH × temperature and pH × time. The time quadratic term and interaction time × temperature, on the contrary, were not significant (P < 0.05). Figure 2 shows the response surface for TCAI as a function of pH and temperature, for 3.5 h. A minimum TCAI was observed around 6.5–7.5 and low temperature. Bombara,23 working with wheat flour, reported similar TCAI values at 54 ◦ C, pH 6.5 and time less than 6 h. However, Adler-Nissen,17 working with soy protein and an enzyme derived from A. oryzae, found that the DH obtained at pH 7 was similar to that at pH 5. In the present work, the TCAI obtained for pH 7 is surprisingly lower than that at pH 5, this difference being probably due to the fact that experiments were carried out with different substrates and also a different method of measuring DH. Gluten isoelectric point is around pH 7, which favors molecular association by hydrophobic interactions, and makes them less susceptible to enzymatic attack by aggregate formation. According to the results obtained, the highest TCAI values will be obtained in the pH range between 3.7 and 5. Table 3 shows the results of TCAI corresponding to the central composite design 22 + star. Analysis of variance showed that the two variables were significant in every term and that the correlation coefficient was r = 0.9907. Figure 3 shows the surface response for TCAI as a function of pH and temperature, and shows a maximum around pH 4.2–4.4 and at 54 and 58 ◦ C. 1418 Table 3. Results of TCAI corresponding to the central composite experimental design 22 + star. [S] = 5%, E/S = 3% Experiment no. 1 2 3 4 5 6 7 8 9 10 11 pH Temp. (◦ C) TCAI 4.25 3.0 3.0 5.5 5.5 4.25 4.25 4.25 2.48 6.02 4.25 45 35 55 35 55 45 30.9 59.1 45 45 45 17.600 12.484 12.630 3.886 13.568 17.466 8.500 20.036 9.060 3.860 17.000 The polynomial corresponding to this experimental design was TCAI% = −46.5172 + 19.1734 × pH + 0.842979 × Temperature − 3.44202 × pH2 + 0.19072 × pH × Temperature − 00 147 418 × Temperature2 According to the information given by the manufacturer, the operating conditions (pH and temperature) of this enzyme include conditions at which very low activity was observed in the present work. This underlines the importance of checking the operating conditions for each particular enzyme–substrate system; surface response methodology could be used to find the best conditions. J Sci Food Agric 88:1415–1422 (2008) DOI: 10.1002/jsfa Wheat gluten hydrolysates and characterization of selected fractions Table 4. Solubility of gluten hydrolysates and thermally treated gluten (TTG) 20 TCAI 16 12 Samples pH TTG 4.0 6.5 9.0 58.04 ± 0.41d 5.88 ± 0.13h 25.58 ± 0.88g TCAI 14% 4.0 6.5 9.0 80.47 ± 0.28a,b 35.87 ± 0.06f 57.77 ± 0.37d TCAI 22% 4.0 6.5 9.0 81.01 ± 0.50a 53.74 ± 0.32e 73.88 ± 0.63c TCAI 32.6% 4.0 6.5 9.0 77.58 ± 0.16a,b 67.54 ± 0.69b,c 79.94 ± 0.06a 8 4 0 3 3.5 4 4.5 5 5.5 35 40 45 50 55 60 Temp. °C pH Figure 3. Surface response for central composite experimental design 22 + star. [S] = 5%, E/S = 3%; TCAI, trichloroacetic acid index. Selection of E/S ratio Figure 4 shows the hydrolysis curves corresponding to different E/S ratios at pH 4.25 and 55 ◦ C, expressed as percentage (%w/w). It is observed that TCAI increases with an increase in E/S. However, when selecting the enzyme concentration to reach a particular degree of hydrolysis, the relationship effectiveness–economy has to be taken into account. A higher enzyme concentration could be used in order to increase hydrolysis rate, but in this case a cost analysis of raw material versus reaction rate should be carried out. In our case, a 5% E/S ratio was chosen because it was considered to be a reasonable ratio, although with products having a high added value (for example, peptides with specific functions) a higher E/S ratio could be selected. Analysis of hydrolysates and solubility Three hydrolysate samples were prepared under the conditions previously selected (pH, temperature and E/S) for times 0.5, 2 and 6 h and their TCAI were 14%, 22% and 32.6%, respectively. Table 4 shows the protein solubilities corresponding to thermally treated gluten and the three hydrolysates. At pH 6.5 (isoelectric point region), hydrolysate solubility was always lower than that corresponding 30 25 TCAI 20 15 10 5 0 0 20 40 60 Time (min) 1.87% 3% 3.75% 5% 10% Figure 4. Effect of E/S ratio (%w/w) on TCAI; [S] = 5%, pH = 4.25, T = 55 ◦ C; TCAI, trichloroacetic acid index. J Sci Food Agric 88:1415–1422 (2008) DOI: 10.1002/jsfa x ± SD Different letters represent significant differences between the samples (P < 0.05); x ± SD, mean of triplicate ± standard deviation. to other pH values. The enzyme could hydrolyze the protein, producing large-size polypeptides, whose solubility is more pH dependent, together with small peptides. The pH dependence could be associated not only with the size of the peptides produced by hydrolysis but also their net charge. Linares et al,15 working with wheat gluten hydrolysate (DH 1.4%, measured as the increase in free amino groups), observed that a pH increase from 4 to 6.5 produced a decrease in the solubility of the more hydrophobic peptides. Molecular size decrease allows hydrolysates to be obtained which are more soluble than the native proteins and within a wider pH range. The solubility profile of the partially hydrolyzed protein improves along the whole pH scale, being generally higher with higher TCAI, since the protein does not form large aggregates even at the isoelectric pH. The solubility increase obtained from a limited proteolysis is attributed to the formation of either smaller and more hydrophilic polypeptide unities or soluble hydrophobic peptides,24 as well as the disruption of protein aggregates.25 Characterization of soluble fractions at different pH of thermally treated gluten (TTG) and hydrolysates SDS-PAGE When observing the electrophoresis pattern corresponding to non-hydrolyzed gluten (TTG) (Fig. 5(a), lanes 4, 5 and 6), without mercaptoethanol (ME), a diffuse stain can be observed for the extracts obtained at the three pH values. These could be attributed to the presence of protein aggregates distributed along the range of the molecular weights (MW) under study. When using ME (Fig. 5(b), lanes 4, 5 and 6) these aggregates break up and form well-defined bands, which show that disulfur bridges may be involved in their formation. Moreover, pH has a marked influence on the fractions of solubilized proteins. The extracts at pH 4 and 9 treated with ME (Fig. 5(b), lanes 4 and 5) are formed by components of MW higher than 94 1419 SR Drago, RJ González, MC Añón and lower than 14.4 kDa, and the extract at pH 6.5 (Fig. 5(b), lane 6) is mainly formed by components whose MW is around 43 and 14.4 kDa. The differences between the fractions extracted at pH 4 and 9 are particularly evident at MW ranging between 67 and 30 kDa. According to Cornec et al,26 gliadins are present between fractions of 30–40 kDa, except for ω-gliadin, which is present in the fractions rich in glutenins. Glutenins of high MW are present in fractions of more than 50 kDa and those of low MW are present within the range 40–43 kDa. Fractions lower than 20 kDa are non-storage proteins, but membrane and lipid binding proteins. The extracts at pH 9 and 4 from hydrolysate with TCAI = 14.1% (Fig. 5(a), lanes 1 and 3, respectively) have components of MW higher than 43 kDa which, when treated with ME (Fig. 5(b), lanes 1 and 3), show that they corresponded to subunits of MW below 43 kDa, linked by disulfur bridges. That is, with a low hydrolysis, the extracts at these pH showed components of high MW, which corresponded to glutenin fractions associated with S–S, though already hydrolyzed. The extracts at pH 4 and 9 have different components between 40 and 30 kDa. In the case of the pH 6.5 fraction (Fig. 5(b), lane 2) the extracted (a) S 1 2 3 4 5 6 kDa 94 67 43 30 20.1 14.4 (b) S 1 2 3 4 5 6 polypeptides are mainly formed by monomers of MW below 20.1 kDa, though showing an MW dispersion, like those components of extracts at pH 9. The extracts from hydrolysates with higher TCAI (22% and 32.6%) had components whose MW are higher than 43 kDa (electrophoresis without ME is not shown), which correspond to S–S bridge associations of peptides whose MW is lower than 43 kDa for TCAI 22% and lower than 20.1 kDa for TCAI 32.4%. This is shown in Fig. 6 for samples treated with ME. In every case there seem to be components of 14.4 kDa, which do not hydrolyze, since there is always a band corresponding to that MW. The extracts at pH 4 and 9 are similar but differ from that at pH 6.5, the later lacks the components corresponding to the 20–40 kDa zone (ITCA 32.6%) or have low quantities of them (TCAI 22%), this probably being related to the presence of a limiting peptide size for the enzyme action in the range of hydrolysis degrees evaluated in this work. Average peptide chain length Figure 7 shows the average peptide chain length of the soluble fraction. The extracts at pH 6.5 for the same TCAI have the peptides with the smallest sizes. For the hydrolysate with 32.6% TCAI, protein extractability was not much affected by pH, and their extracts have peptides of about 14 amino acid residues. The way the enzyme acts can be seen with the PCL measured at the pH corresponding to the isoelectric point (pH 6.5) or pH 9, where the solubility of the thermally treated gluten is low. PCL shows a rapid decrease with progress of the enzymatic reaction during the first 30 min. This increase is then slower, which shows that soluble peptides are degraded during hydrolysis until they reach a certain length. After that, the enzyme would not act on them but on the substrates of higher MW, which increases solubility. Integrated areas of RP-HPLC Figure 8 shows the integrated areas of RP-HPLC chromatograms. The extracts at pH 6.5 have a higher kDa S 94 67 kDa 43 94 30 20.1 14.4 Figure 5. (a) EF-SDS-PAGE without mercaptoethanol of pH extracts from thermally treated gluten (TTG) and hydrolysate (TCAI: 14%). S, standard; 1, 14%–pH 9; 2, 14%–pH 6.5; 3, 14%–pH 4; 4, TTG–pH 9; 5, TTG–pH 6.5; 6, TTG–pH 4. (b) EF-PAGE-SDS with mercaptoethanol of pH extracts from TTG and hydrolysate (TCAI: 14%). S, standard; 1, 14%–pH 9; 2, 14%–pH 6.5; 3, 14%–pH4; 4, TTG–pH 9; 5, TTG–pH 4; 6, TTG–pH 6.5. 1420 1 2 3 4 5 6 67 43 30 20.1 14.4 Figure 6. EF-PAGE-SDS with mercaptoethanol of pH extracts from hydrolysate (TCAI: 22% and 32.6%). S, standard; 1, 32.6%–pH 9; 2, 32.6%–pH 6.5; 3, 32.6%–pH 4; 4, 22%–pH 9; 5, 22%–pH 6.5; 6, 22%–pH 4. J Sci Food Agric 88:1415–1422 (2008) DOI: 10.1002/jsfa Wheat gluten hydrolysates and characterization of selected fractions 40 a 35 30 b c PCL 25 20 f e d 15 g g 10 h 5 0 0 10 20 30 40 TCAI (%) pH 4 pH 6.5 pH 9 These results suggest that enzyme action on gluten proteins reduced the size of gliadin and glutenin molecules. The resulting hydrolysate varied in composition with the degree of hydrolysis reached, though in general peptide size distribution would not be unimodal. At TCAI 32%, most of the hydrolysate is composed of peptides whose size is less than 18.5 kDa, with an average PCL of 14 amino acid residues. PCL is useful to evaluate the hydrolysis process at pH values at which the proteins have a low solubility. However, it is important to point out that PCL does not represent the average of a unimodal population, particularly at low degrees of hydrolysis. Figure 7. Average peptide chain length versus TCAI. Different letters represent significant differences between the samples (P < 0.05). CONCLUSIONS Surface response methodology was used to find the experimental conditions required for the highest enzymatic activity for hydrolysis of wheat gluten by fungal protease. Fractions at different pH values had different compositions, mainly at low TCAI (14% and 22%), which could be taken into account when trying to find a defined composition related to determined functional characteristics. 70 % total Area 60 50 40 30 20 10 0-20 min 20-40 min 32 _ 32 4 _6 .5 32 _9 22 _ 22 4 _6 .5 22 _9 14 _ 14 4 _6 .5 14 _9 G TT G _4 TT _9 0 40-60 min Figure 8. Area percentage of each eluate section (0–20, 20–40 and 40–60 min) with respect to the total area, corresponding to different extracts at pH 4, 6.5 and 9 of hydrolysates (TCAI: 14%, 22% and 32.6%) and thermally treated gluten (TTG). content of components with low hydrophobicity and MW (0–20 min zone) compared to the other extracts, which is in agreement with the lower PCL values obtained. As TCAI increases, the quantity of components with average hydrophobicity and MW (20–40 min zone) increases, except at pH 6.5, where it decreases. It is also observed that as TCAI increases there is a clear decrease in hydrophobic components (40–60 min zone), though there are practically no changes for pH 6.5 extracts. This could be attributed to the break pattern of this protease. Since hydrolysis products have components of high and low MW, at pH 6.5 the solubilized proteins would be poorer in hydrophobic components of high MW, and richer in components of low hydrophobicity, which is reversed at TCAI 32.6%, when solubility increases. It is then confirmed that during the first reaction stages (TCAI 14% and 22%) this enzyme hydrolyzes the protein, generating components of high MW, the solubility of which is more pH dependent, together with soluble components of low MW. At TCAI 32.3%, the enzyme generates peptides which are soluble (solubility higher than 76%) and less pH dependent. J Sci Food Agric 88:1415–1422 (2008) DOI: 10.1002/jsfa ACKNOWLEDGEMENTS This work was partly supported by Proyecto CAI+D 2005-005-25, Universidad Nacional del Litoral, Santa Fe, Argentina. REFERENCES 1 Kalin F, Wheat gluten applications in Food Products. J Am Oil Chem Soc 56:477–479 (1979). 2 Pecquet C and Lauriere M, New allergens in hydrolysates of wheat proteins. Rev Fr Allergol Inmmunol Clin 43:21–23 (2003). 3 Popineau Y, Huchet B, Larré C and Bérot S, Foaming and emulsifying properties of fractions of gluten peptides obtained by limited enzymic hydrolysis and ultrafiltration. J. Cereal Sci 35:327–335 (2002). 4 Guadix A, Guadix E, Páez-Dueñas MP, González-Tello P and Camacho F, Procesos tecnológicos y métodos de control en la hidrólisis de proteı́nas. Ars Pharm 41:79–89 (2000). 5 Löffler A, Proteolytic enzymes: sources and applications. Food Technol 40:63–70 (1986). 6 Spellman D, McEvoy E, Cuinn GO and FitzGerald RJ, Proteinase and exopeptidase hydrolysis of whey protein: comparison of the TNBS, OPA and pH-stat methods for quantification of degree of hydrolysis. Int Dairy J 13:447–453 (2003). 7 Mullally MM, O’Callaghan DM, Fitzgerald RJ, Donnelly WJ and Dalton JP, Zymogen activation in pancreatic endoproteolytic preparations and influence on some whey protein hydrolysate characteristics. J Food Sci 60:227–233 (1995). 8 Vioque J, Clemente A, Pedroche J, Yust MM and Millán F, Obtención y aplicaciones de hidrolizados proteicos. Grasas y Aceites 52:132–136 (2001). 9 Lee JY, Duck H and Lee CH, Characterization of hydrolysates produced by mild-acid treatment and enzymatic hydrolysis of defatted soybean flour. Food Res Int 34:217–222 (2001). 10 AACC, Approved Methods of the American Association of Cereal Chemists (8th edn). AACC, St Paul, MN (1983). 1421 SR Drago, RJ González, MC Añón 11 Cochran WG and Cox GM, Diseños experimentales. Trillas, Mexico (1978). 12 Lowry OH, Rosenbrough NJ, Farr AL and Randall RJ, Protein measurement with the Folin-phenol reagent. J Biol Chem 193:265 (1951). 13 Drago SR and González RJ, Foaming properties of enzymatically hydrolysed wheat gluten. Innov Food Sci Emerg Technol 1:269–273 (2001). 14 Laemmli UK, Cleavage of structural proteins during the assembly of the head of bacterophage T4. Nature 227:680 (1970). 15 Linares E, Larré C and Popineau Y, Freeze or spray-dried gluten hydrolysates. 1. Biochemical and emulsifying properties as a function of drying process. J Food Eng 48:127–135 (2001). 16 Nielsen PM, Petersen D and Dambmann C, Improved method for determining food protein degree of hydrolysis. J Food Sci 66:642–646 (2001). 17 Adler-Nissen J, Enzymic Hydrolysis of Food Proteins. Elsevier Applied Science, London (1986). 18 Jeanjean MF, Damidaux R and Feillet P, Effect of heat treatment on protein solubility and viscoelastic properties of wheat gluten. Cereal Chem 57:325–331 (1980). 19 Autran JC and Berrier R, Durum wheat functional protein subunits revealed through heat treatments: biochemical and genetical implications, in Gluten Proteins: Proceedings of the 2nd International Workshops on Gluten Protein, Wageningen 1422 20 21 22 23 24 25 26 Netherlands, ed. by Graveland A and Moonen JEH, pp. 175–183 (1984). Tosi E, Ré E, Catalano O and Cazzoli A, Secado de trigo por lecho fluidizado. Alim Latinoam 16(133):61–80 (1982). Lavelli V, Guerrieri N and Cerletti P, Controlled reduction study of modifications induced by gradual heating in gluten proteins. J Agric Food Chem 44:2549–2555 (1996). Schofield JD, Bottomley RC, LeGrys GA, Timms MF and Booth MR, Effects of heat on wheat gluten, in Gluten Proteins: Proceedings of the 2nd International Workshops on Gluten Protein, Wageningen, Netherlands, ed. by Groveland A and Moonen JHE, pp. 81–90 (1984). Bombara N, Modificación de las propiedades de la harina de trigo mediante hidrólisis enzimática. Doctoral thesis, Universidad de Mar del Plata (1995). Linares E, Larré C, Lemestre M and Popineau Y, Emulsifying and foaming properties of gluten hydrolysates with an increasing degree of hydrolysis: role of soluble and insoluble fractions. Cereal Chem 77:414–420 (2000). Masson P, Tomé D and Popineau Y, Peptic hydrolysis of gluten, glutenin and gliadin from wheat grain: kinetics and characterization of peptides. J Sci Food Agric 37:1223–1235 (1986). Cornec M, Popineau Y and Lefebvre J, Characterization of gluten subfractions by SE-HPLC and dynamic rheological analysis in shear. J Cereal Sci 19:131–139 (1994). J Sci Food Agric 88:1415–1422 (2008) DOI: 10.1002/jsfa J Sci Food Agric 88:1423–1430 (2008) Journal of the Science of Food and Agriculture High-performance liquid chromatography procedure for the determination of flavor enhancers in consumer chocolate products and artificial flavors Charles H Risner∗ and Melissa J Kiser RJ Reynolds Tobacco Company, Bowman Gray Technical Center, Winston-Salem, NC 27102-1487, USA Abstract BACKGROUND: A number of individual high-performance liquid chromatography (HPLC) procedures exist for the analysis of maltol, theobromine, ethyl maltol, catechin, vanillic acid, caffeine, vanillin, epicatechin, and ethyl vanillin. A single procedure utilizing simple sample preparation and less sophisticated equipment would be advantageous for the analysis of different sample types containing these compounds. RESULTS: An HPLC procedure has been developed using water as the extract for consumer products and artificial flavors. A methanol–water gradient was used to elute the compounds of interest using a reverse-phase column. Absorbance detection using a wavelength of 273 nm was used to monitor the eluent. Recoveries for these compounds ranged from 88% to 104%. CONCLUSIONS: Results obtained for theobromine, catechin, caffeine, and epicatechin in Standard Reference Material 2380 Baking Chocolate compare well with those found in its certificate of analysis verifying that the procedure is valid. Vanillic acid and ethyl vanillic acid were found as oxidation products of vanillin and ethyl vanillin in both standards and some consumer products.  2008 Society of Chemical Industry Keywords: artificial flavors; consumer products; flavor enhancers; HPLC INTRODUCTION Several individual reverse-phase high-performance liquid chromatography (RP-HPLC) procedures are reported in the literature for the determination of flavor enhancers: theobromine, theophylline, (+)-catechin, caffeine and (−)-epicatechin in chocolate;1 adenine, theobromine, theophylline, and caffeine in cola beverages, tea, coffee, and cocoa;2 theobromine, theophylline, and caffeine in cocoa, coffee, tea, chocolate, coconut water, and baking chocolate;3 – 5 theobromine and caffeine in cocoa and chocolate products;6,7 (+)catechin and (−)-epicatechin in cocoa beans, powders, chocolates, and baking chocolate;8,9 theobromine, caffeine, and vanillin in a cocoa drink;10 and maltol and ethyl maltol in beverages.11 Preparation of these samples includes extraction in methanol,1,8 – 10 defatting with petroleum ether or hexane prior to further sample treatment,2 – 7,9 and, in some cases, solid-phase extraction.2,7 Although amperometric,2 electrochemical,8 and mass selective detection are used,9 most procedures employ more common ultraviolet (UV) detection for the determination of the analytes.1,3 – 7,10,11 No single procedure exists for the simultaneous determination of maltol, theobromine, ethyl maltol, (+)-catechin, vanillic acid, caffeine, vanillin, (−)epicatechin, and ethyl vanillin in consumer products and artificial flavors used to enhance the flavor of these products using RP-HPLC. Since all of these compounds have some solubility in water, it was decided to use water as the extract and avoid the use of solvents such as petroleum ether or hexane for fat removal and methanol to dissolve the compounds of interest. Sample preparation consists of simple dilution of the sample in water or warming the water extract to melt the sample in the case of solid samples, after which the extract is filtered prior to analyses. No solidphase extraction is used for sample preparation as is done in other procedures where loss of the analyte can occur.2,7 UV detection, which is more commonplace in many laboratories, is employed for the analyses of nine flavor enhancers. The method described here has the capability of quantitatively determining maltol, theobromine, ethyl maltol, (+)-catechin, vanillic acid, caffeine, vanillin, (−)-epicatechin and ethyl vanillin from the water extract of consumer products and artificial flavors. ∗ Correspondence to: Charles H Risner, RJ Reynolds Tobacco Company, Bowman Gray Technical Center, Winston-Salem, NC 27102-1487, USA E-mail: risnerc@rjrt.com (Received 17 August 2007; revised version received 29 January 2008; accepted 30 January 2008) Published online 17 April 2008; DOI: 10.1002/jsfa.3234  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 CH Risner and MJ Kiser EXPERIMENTAL Samples All consumer products were purchased from local supermarkets in the United States in January 2006. Dove milk chocolate, Dove dark chocolate, Hershey’s chocolate syrup, Hershey’s chocolate milk, Tootsie Roll and Swiss Miss Cocoa were selected as research material. Proprietary formulations of commercially available, artificially compounded cocoa/chocolate flavors in propylene glycol were obtained from three different suppliers. Standard Reference Material 2384 baking chocolate was purchased from the National Institute of Standards and Technology (NIST) Gaithersburg, MD, USA. Chemicals and reagents Adenine, maltol, theobromine, theophylline, ethyl maltol, (+)-catechin hydrate (catechin), vanillic acid, caffeine, vanillin, (−)-epicatechin (epicatechin), and ethyl vanillin were purchased from Sigma/Aldrich (St Louis, MO, USA) and used as received. In-house water was passed through a Nanopure system which consisted of an organic removal cartridge, two anion/cation mixed-bed cartridges, a carbon and mixed-bed cartridge and a 0.20 µm filter (Barnstead/Thermolyne, Dubuque, IA, USA). Methanol (MeOH) (Burdick & Jackson, Muskegon, MI, USA) was used as a portion of the mobile phase. Glacial acetic acid was purchased from Fisher Scientific Company LLC, Suwanee, GA, USA. Stock standard preparation The stock standard concentrations were chosen so as not to exceed their solubility in water. Stock standards of adenine were prepared at 150 µg mL−1 ; maltol 150 µg mL−1 ; theobromine 300 µg mL−1 ; theophylline 150 µg mL−1 ; ethyl maltol 1300 µg mL−1 ; catechin 130 µg mL−1 ; vanillic acid 100 µg mL−1 ; caffeine 200 µg mL−1 ; vanillin 500 µg mL−1 ; epicatechin 130 µg mL−1 ; and ethyl vanillin 300 µg mL−1 in water. The stock solutions and working standards were stored at 5 ◦ C in the dark when not in use. The standards were stable for at least 2 weeks when stored under these conditions. Extraction and sample preparation Owing to the variation of analyte concentration in the samples and their different molar absorptivities, sample weights ranged from 0.001 to 1 g in the 5 mL water extract (no constant ratio of sample weight to water extract volume was used). The variation in sample weight was done to keep the concentration of the analytes in the range of the working standards. Vortexing for 15 min (VWR VX-2500 Multi-Tube Vortexer; Henery Troemner LLC, Thorofare, NJ, USA) was used to extract the samples. NIST baking chocolate, Dove milk chocolate and Dove dark chocolate sample extracts were heated to 60 ± 2 ◦ C for 10 min in a water bath to melt the sample. After extraction, the extract was then filtered using a 0.45 µm 1424 pore size polyvinylidene fluoride Whatman Autovial (Whatman, Clifton, NJ, USA) into a glass autosampler vial which was sealed with a screw cap containing a septum. Spectrophotometer, HPLC system and conditions UV absorbance spectra were determined using an Agilent 8453 ultraviolet-visible absorbance spectrophotometer (Agilent Technologies Deutschland gmbH, Waldbronn, Germany) of the standard compounds in water to find the wavelength of maximum absorption (λmax ) to be used for their detection. The HPLC system consisted of a 680 gradient controller, two 515 pumps, a 717 plus autosampler and a 2487 absorbance detector (Waters, Milford, MA, USA). Analysis was conducted using a 2.0 mm × 150 mm, 5 µm particle size Gemini C18 analytical column preceded by a guard column containing a Gemini C18, 2 mm cartridge (Phenomenex, Torrance, CA, USA). The injection volume was 5 µL and the autosampler temperature set at 10 ◦ C. The absorbance detector wavelength was set at 273 nm using different sensitivity settings (absorbance units full scale, AUFS) during and within a run to avoid saturation of the detector or non-detection of the analyte. The mobilephase flow rate was 500 µL min−1 using a gradient composed of 0.3% (volume fraction) acetic acid (A) and MeOH (B). Initial conditions consisted of 85% A and 15% B (volume fractions) for 10 min, after which the composition was changed to 75% A and 25% B and held for 8 min. After 18 min the mobile phases were changed to 70% A and 30% B and held for 7 min. This was followed by a column wash of 100% B for 5 min and a 5 min conditioning of the column using 100% A prior to the next injection. Data were acquired and results calculated in mg kg−1 using EZChrom Elite version 3.0.0 (Scientific Software, Pleasanton, CA, USA). Method validation procedures Quantification was determined by injecting 5 µL of each sample extract into the high-performance liquid chromatograph and the height of the peak observed at the retention time corresponding to each authentic standard was recorded. This value was used to calculate the amount (mg kg−1 ) of flavor enhancer in the samples using the linear equation obtained from the standard curve, the extract volume and sample weight. Precision was obtained by analyzing each sample at least three times. Recoveries and y-intercept values from standard addition experiments were determined in duplicate by spiking pure standards at one half, one, and two times the amounts found in the samples to the extract solutions prior to sample analysis. All statistical and mathematical calculations were done using Microsoft Excel 2003 (Microsoft Corporation, Redmond, WA, USA). J Sci Food Agric 88:1423–1430 (2008) DOI: 10.1002/jsfa HPLC method for flavor enhancers the C18 columns using water as the mobile phase was phase collapse, resulting in poor or no resolution after only about 10 injections.12 The gradient described in the Experimental section was chosen because 11 compounds found in consumer products can be separated in a single run. Figure 1 is a chromatogram of a composite standard containing the 11 analytes evaluated in this work. Adenine (Aden) and theophylline (Theop) were not detected in the samples analyzed as determined by comparison of retention times with authentic standards and spiking of the sample extract with solutions of authentic standards. Portions of the gradient may not be needed. For example, after screening a sample extract, if only maltol, theobromine, and ethyl maltol are observed the first two segments of the gradient would only need to be applied (85:15, 0.3% acetic acid:MeOH and 75:25, 0.3% acetic acid:MeOH). The 100% MeOH column wash would be optional, depending on the cleanliness of the sample extract, but may be included to prevent phase collapse. The last segment of 100% 0.3% acetic acid may also be an option, but proved to help better resolve early-eluting compounds such as adenine. RESULTS AND DISCUSSION Ultraviolet absorbance spectra UV-visible spectra were obtained on standards dissolved in water at a concentration of 10 ± 1 µg mL−1 . Table 1 gives the λmax and the absorbance: concentration ratios (AUFS/µg mL−1 ) of these compounds. Even though some of these compounds do not have their maximum absorbance at 273 nm, this wavelength was used for detection since a sufficient response was obtained. Other workers who used UV absorbance detection for the determination of these compounds used wavelengths from 273 to 280 nm.1,3 – 11 Columns and conditions evaluated Initially, water was used as the mobile phase on reverse-phase octadecylsilane (C18) columns using column temperatures as high as 55 ◦ C. Caffeine was strongly retained (>60 min) on the columns using water as the mobile phase, but the main drawback of Table 1. Wavelengths of maximum absorbance and absorbance:concentration ratios of compounds evaluated in watera AUFS/µg mL−1 261 272 273 271 275 279 257 273 279 279 279 0.0888 0.0651 0.0503 0.0574 0.0696 0.0127 0.0582 0.0521 0.0646 0.0124 0.0576 Instrument precision, linearity of response, limit of detection, and limit of quantitation Table 2 lists the instrument precision, linearity of response, limit of detection (LOD), and limit of quantitation (LOQ) of the standards used in this work. The percent relative standard deviations (%RSD) were less than 3%, except for catechin and epicatechin. This may be due to the low sensitivity setting of 0.008 absorbance units full scale (AUFS) required to detect catechin and epicatechin at these concentrations due to their lower absorbance:concentration ratio at 273 nm (see Table 1). λmax , wavelength of maximum absorption; AUFS, absorbance units full scale. a Concentration of 10 ± 1 µg mL−1 . Malt 1.0 0.5 Van Vana 0.4 Caff 0.6 Theop 0.3 0.2 Cat Absorbance Units 0.7 EtMalt 0.8 Theob Aden 0.9 EtVan Adenine Maltol Theobromine Theophylline Ethyl maltol Catechin Vanillic acid Caffeine Vanillin Epicatechin Ethyl vanillin λmax (nm) 0.1 Ecat Compound 0.0 -0.1 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 Min Figure 1. Chromatogram of adenine (Aden), maltol (Malt), theobromine (Theob), theophylline (Theop), ethyl maltol (EtMalt), catechin (Cat), vanillic acid (Vana), caffeine (Caff), vanillin (Van), epicatechin (Ecat), and ethyl vanillin (EtVan) standards in water. Conditions: Gemini C18 (2.0 mm × 150 mm, 5 µm); mobile phase gradient 0.3% acetic acid and MeOH at 500 µL min−1 ; injection volume 5 µL; autosampler temperature 10 ◦ C; absorbance detection at 273 nm and 0.350 absorbance units full scale (AUFS); concentration of all components 10 ± 1 µg mL−1 . J Sci Food Agric 88:1423–1430 (2008) DOI: 10.1002/jsfa 1425 CH Risner and MJ Kiser Table 2. Instrument precision (n = 12), linearity of response, limit of detection and limit of quantitationa Compound Adenine Maltol Theobromine Theophylline Ethyl maltol Catechin Vanillic acid Caffeine Vanillin Epicatechin Ethyl vanillin % Relative standard deviation Concentration range (µg mL−1 )b r2 LOD (µg mL−1 )c LOQ (µg mL−1 )d 1.66 0.76 2.74 2.47 2.58 4.01 1.20 2.61 1.59 3.89 2.55 1.00–16.56 1.50–18.60 1.00–15.84 1.00–16.32 1.50–19.96 0.50–10.10 1.00–14.90 1.00–15.50 1.00–14.80 0.50–11.70 1.00–15.36 0.9970 0.9994 0.9986 0.9996 0.9998 0.9976 0.9999 0.9995 0.9993 0.9976 0.9958 0.05 0.03 0.08 0.07 0.11 0.06 0.04 0.06 0.05 0.06 0.08 0.17 0.10 0.26 0.24 0.35 0.19 0.13 0.19 0.16 0.19 0.26 % Relative standard deviation = height standard deviation/height average × 100; r 2 , linear correlation coefficient. Absorbance units full scale (AUFS) = 0.100 for all compounds except catechin and epicatechin, where AUFS = 0.008. b Five levels. c LOD, limit of detection, three times the standard deviation of the lowest standard analyzed as a sample. d LOQ, limit of quantitation, ten times the standard deviation of the lowest standard analyzed as a sample. a Table 3. Percent recovery, amount found by standard addition and amount found by external standard of compounds from Flavor 1, Hershey’s chocolate syrup and Hershey’s chocolate milk Compound Maltola Theobromineb Ethyl maltola Catechinb Vanillic acidc Caffeineb Vanillinb Epicatechinb Ethyl vanillina % Recovery ± SD (n = 2) 101.7 ± 1.7 102.1 ± 0.8 94.7 ± 6.3 97.6 ± 0.9 88.3 ± 0.8 102.2 ± 2.1 103.8 ± 0.6 102.1 ± 2.6 101.6 ± 1.5 mg kg−1 ± SD standard addition (n = 2) mg kg−1 ± SD external standard (n = 3) 1 576 ± 1 1 788 ± 22 29 661 ± 93 54 ± 1 19 ± 0.1 130 ± 1 148 ± 1 144 ± 3 137 246 ± 683 1 515 ± 20 1 843 ± 22 30 871 ± 328 67 ± 2 19 ± 0.6 132 ± 3 147 ± 2 150 ± 2 130 875 ± 1 411 SD, standard deviation. a Flavor 1, absorbance units full scale (AUFS) = 0.200, 0.030 ± 0.001 g in 5 mL water for maltol and ethyl maltol, 0.016 ± 0.001 g in 5 mL water for ethyl vanillin. b Hersey’s chocolate syrup, AUFS = 0.350 initially, changing to 0.100 after 10 min, 0.15 ± 0.01 g in 5 mL water for theobromine and caffeine, AUFS = 0.200, 0.16 g ± 0.01 in 5 mL water for vanillin, AUFS = 0.008, 0.12 ± 0.01 g in 5 ml for catechin and epicatechin. c Hersey’s chocolate milk, AUFS = 0.040, 1.02 ± 0.01 g in 5 mL water for vanillic acid. The linear correlation coefficients (r 2 ) for the standards over the specified concentration ranges are greater then 0.99. The LOQ values were well below the concentration of the standards used to analyze the samples. Recovery and standard addition Table 3 shows the percent recoveries to be acceptable, except for vanillic acid found at a low level in Hershey’s chocolate milk, which was 88%. The amounts found by standard addition are in the same range as those found by external standard, showing no interferences with the compounds analyzed in their sample matrix. Comparison of results with National Institute of Standards and Technology (NIST) baking chocolate Table 4 compares the results for theobromine, catechin, caffeine, epicatechin, and theophylline in a baking chocolate standard reference material (NIST Standard Reference Material 2384) obtained by the 1426 procedure described in the Experimental section with those stated by NIST.5 The results are comparable to the NIST certified results except that no theophylline was detected in the baking chocolate using the procedure described here. Theobromine, caffeine, and theophylline results obtained by NIST were determined on a sample defatted with hexane four times, dried in a stream of nitrogen, and reconstituted in water. The extract was analyzed using absorbance detection at 274 nm and quantified using an internal standard (β-hydroxyethyltheophylline) and an isocratic mobile phase of acetonitrile, water, and acetic acid.5 This is similar to the procedure described here except that hexane was not used, results were calculated based on external standard, and an MeOH, water, acetic acid mobile phase gradient was employed. Theophylline is below detection limits (<11 mg kg−1 ), but an amount of 151 mg kg−1 should have given a response around 11 min (see Fig. 1), since catechin at 222 mg kg−1 , which has a fivefold lower absorbance:concentration ratio than theophylline, was detected (Table 1). Other workers have also reported J Sci Food Agric 88:1423–1430 (2008) DOI: 10.1002/jsfa HPLC method for flavor enhancers Table 4. Comparison of results with the National Institute of Standards and Technology (NIST) Standard Reference Material 2384, baking chocolate (n = 6, mg kg−1 ), percent recovery (n = 2, percent ± SD) and standard addition from NIST baking cocoa (n = 2, mg kg−1 ± SD) Theobromine Catechin Caffeine Epicatechin Theophylline Results by this procedurea Average SD % RSD 12 532 1 166 9.30 222 16 7.22 1117 40 3.56 1463 47 3.18 <11b Certified results of NIST Average SD % RSD 11 600 1 100 9.48 245 51 20.82 1060 50 4.72 1220 240 19.67 151 3 1.96 102.1 ± 2.5 1126 ± 2c 101.0 ± 0.4 1517 ± 15d Recovery and standard addition from NIST baking cocoa % Recovery 99.6 ± 1.2 104.3 ± 0.9 Standard Addition 11 659 ± 122c 211 ± 1d SD, standard deviation; % RSD, percent relative standard deviation = standard deviation/average × 100. 0.020 ± 0.001 g in 5 mL 60 ◦ C water. b Based on S/N ≥ 2 at 0.002 AUFS. c Absorbance units full scale (AUFS) = 0.6 initially, changing to 0.04 AUFS after 10 min. d AUFS = 0.002 initially, changing to 0.008 AUFS after 18 min. a not detecting theophylline in Theobroma seeds and dark and milk chocolates, where theophylline was detected in some samples but not in others.1,13 The NIST results for catechin and epicatechin (quantified using an internal standard of trytophan methyl ester hydrochloride) were obtained using mass selective detection after removing the fat from the sample with three extractions of hexane, drying under nitrogen, extracting two times in MeOH, combining the MeOH extracts, and final dilution in water.9 Table 4 shows the percent recovery of theobromine, catechin, caffeine, and epicatechin to be acceptable (∼100%). The amounts found by standard addition compare with those obtained by external standard, indicating no interferences with the analytes in the baking chocolate sample matrix. The overall results obtained here, except for possibly catechin, are higher than those using the two NIST procedures. This may be due to the additional sample preparation steps used in the NIST procedures, which can result in the loss of the analyte or change in the internal standard concentration. Application to consumer products and artificial flavors Table 5 gives the results for seven compounds found in consumer chocolate products, sample weights extracted in 5 mL water, and the absorbance detector sensitivity (AUFS) setting. Maltol and ethyl maltol were not detected in these samples. Theobromine, catechin, caffeine, vanillin, and epicatechin were higher in Dove dark chocolate than in Dove milk chocolate. Hersey’s chocolate syrup was purchased on two occasions and analyzed three separate times, yielding similar results. In both Hersey’s chocolate syrup samples there was a large peak at approximately 20 min (not shown), which was not observed in other consumer product samples. Table 5. Results for consumer products (n = 3, mg kg−1 ± SD) Compound Dove dark chocolate Dove milk chocolate Hershey’s chocolate syrup Hershey’s chocolate milk Tootsie Roll Swiss Miss cocoa Theobromine Sample wt/AUFS Catechin Sample wt/AUFS Vanillic acid Sample wt/AUFS Caffeine Sample wt/AUFS Vanillin Sample wt/AUFS Epicatechin Sample wt/AUFS Ethyl vanillin Sample wt/AUFS 6184 ± 31 0.04/0.350 287 ± 21 0.04/0.020 <2a 0.12/0.040 1072 ± 16 0.04/0.020 401 ± 13 0.04/0.020 630 ± 160 0.04/0.020 <2a 0.10/0.160 2168 ± 21 0.08/0.350 123 ± 4 0.08/0.020 <2a 0.12/0.040 355 ± 3 0.08/0.020 134 ± 3 0.08/0.020 284 ± 10 0.08/0.020 <2a 0.10/0.160 1843 ± 22 0.15/0.350 67 ± 2 0.12/0.008 <2a 0.12/0.040 132 ± 3 0.15/0.100 147 ± 2 0.16/0.200 150 ± 2 0.12/0.008 <2a 0.12/0.160 139 ± 2 1.0/0.350 <2a 0.60/0.008 19 ± 1 1.0/0.040 11 ± 1 1.0/0.100 <1a 0.60/0.160 <1a 0.60/0.008 <1a 0.60/0.160 488 ± 8 0.20/0.200 14 ± 1 0.60/0.040 <2a 0.60/0.040 59 ± 3 0.20/0.200 51 ± 1 0.40/0.160 29 ± 1 0.60/0.040 21 ± 1 0.40/0.160 1626 ± 34 0.04/0.200 28 ± 2 0.40/0.040 <2a 0.60/0.040 140 ± 4 0.04/0.200 <4a 0.50/0.160 26 ± 1 0.40/0.040 81 ± 2 0.10/0.160 Sample wt, weight of sample (g, two significant figures); AUFS, absorbance units full scale; SD, standard deviation. a Based on S/N ≥ 2. J Sci Food Agric 88:1423–1430 (2008) DOI: 10.1002/jsfa 1427 CH Risner and MJ Kiser This peak did not match the retention time of epicatechin as determined by spiking experiments and was estimated to have a concentration of 2325 ± 19 mg kg−1 (n = 6, 0.1690 ± 0.004 g in 5 mL water using 0.200 AUFS, based on the response of ethyl vanillin) – much higher than the amounts of epicatechin found in the other consumer products (Table 5). Hersey’s chocolate milk was also purchased twice and analyzed three times. Vanillic acid was only found in Hersey’s chocolate milk. This may be due to the oxidation of vanillin since in the second sample of Hersey’s chocolate milk vanillin was detected (32 mg kg−1 ) and the sample contained less vanillic acid (14 mg kg−1 ). The oxidation of vanillin to vanillic acid will be discussed later, as well as the oxidation of ethyl vanillin to ethyl vanillic acid in Hersey’s chocolate syrup. Tootsie Roll contained theobromine, catechin, caffeine, vanillin, epicatechin and ethyl vanillin. These same compounds, with the exception of vanillin, were found in Swiss Miss cocoa mix. The sample weights varied from 0.04 ± 0.01 g to 1.00 ± 0.01 g, the higher weights being used to detect compounds of low concentration such as vanillic acid and caffeine in Hershey’s chocolate milk. The detector sensitivity settings were also varied from 0.0008 to 0.350 AUFS, the lower, more sensitive settings required to detect compounds of low absorbance:concentration ratios like catechin and epicatechin (Table 1). Higher sample weights and or low sensitivity setting were used to determine the minimum detectable quantity based on an approximate height in the extract chromatogram of two times the background signal of the detector. Figure 2 is a chromatogram of the water extract of Dove dark chocolate. A lower sensitivity setting (0.02 AUFS) was used after 10 min to better detect the compounds of lower concentration than theobromine. The two peaks on either side of epicatechin may be either (−)-epigallocatechin gallate or (−)-epigallocatechin from the cocoa added to the chocolate formulation, but their presence was not confirmed by the use of authentic standards.1 Table 6 gives the results for the major components found in four different artificial flavors, sample weights extracted in 5 mL water, and the absorbance detector sensitivity (AUFS) setting. As seen in Table 6, maltol, ethyl maltol, vanillin and ethyl vanillin are present in substantial amounts in some of these samples, vanillin, for example, being over 300-fold the amount found in consumer products (Table 5). Catechin and epicatichin may be present in these samples, but are difficult to detect since an increase in sample weight or decrease in extract volume would be required to determine catechin and epicatichin at levels of mg kg−1 . Since maltol, ethyl maltol, vanillin, and ethyl vanillin are present at g kg−1 levels in these samples, an increase in sample weight or decrease in extract volume may not be practical. Attempts at increasing sample extract concentration in this work using these artificial flavors often resulted in column overload, and hence a decrease in retention time of the analytes in the sample matrix and compound carried over from previous injections (instrument contamination). Because of these drawbacks caused by concentrated extracts, catechin and epicatechin were not determined in Flavor 2, Flavor 3, and Flavor 4. Favor 1 contains large amounts of vanillin and ethyl vanillin. Favor 2 is similar to Flavor 1, but contained no detectable theobromine and less caffeine. Flavor 3 contains lesser amounts of maltol, theobromine, and ethyl maltol than Flavor 1, but vanillin and ethyl vanillin are in the same range as Flavors 1 and 2. Flavor 4 appears unusual in that the amount of caffeine is greater than theobromine. It may be that this flavor was not entirely formulated from cocoa, which usually has a ratio of theobromine:caffeine of at least 2.5 to 1.6 As done for consumer products (Table 5), various sample weights (0.001 to 0.100 ± 0.001 g) and detector sensitivity settings (0.100 to 1.000 AUFS) 1.6 Theob 1.2 1.0 Van 0.8 0.6 Ecat 0.4 Cat Absorbance Units 1.4 0.2 Caff 0.0 -0.2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 Min Figure 2. Chromatogram of 60 ◦ C water extract of Dove dark chocolate. Conditions: Gemini C18 (2.0 mm × 150 mm, 5 µm); mobile phase gradient of 0.3% acetic acid and MeOH at 500 µL min−1 ; injection volume 5 µL; autosampler temperature 10 ◦ C; absorbance detection at 273 nm and 0.35 AUFS initially then 0.02 AUFS after 10 min; 0.04 g in 5 mL−1 ; theobromine (Theob), catechin (Cat), caffeine (Caff), vanillin (Van), epicatechin (Ecat). 1428 J Sci Food Agric 88:1423–1430 (2008) DOI: 10.1002/jsfa HPLC method for flavor enhancers Table 6. Results for artificial chocolate flavors (n = 3, mg kg−1 ± SD) Compound Maltol Sample wt/AUFS Theobromine Sample wt/AUFS Ethyl maltol Sample wt/AUFS Catechin Sample wt/AUFS Caffeine Sample wt/AUFS Vanillin Sample wt/AUFS Epicatechin Sample wt/AUFS Ethyl vanillin Sample wt/AUFS Flavor 1 Flavor 2 Flavor 3 Flavor 4 1515 ± 20 0.030/0.200 961 ± 11 0.020/0.350 30871 ± 328 0.030/0.200 619 ± 20 0.020/0.100 294 ± 6 0.020/0.100 131173 ± 1468 0.016/0.200 269 ± 3 0.020/0.100 130875 ± 1411 0.016/0.200 1363 ± 94 0.012/1.000 <9a 0.020/0.200 32264 ± 159 0.012/1.000 NA 83 ± 7 0.100/1.000 591 ± 7 0.010/0.200 3285 ± 9 0.100/1.000 NA <9a 0.100/0.100 753 ± 4 0.010/0.200 <16a 0.100/0.100 NA 81 ± 2 0.024/0.200 132540 ± 1786 0.001/0.160 NA 217 ± 2 0.010/0.200 14257 ± 143 0.006/0.160 NA 2267 ± 22 0.010/0.200 <9a 0.100/0.100 NA 129838 ± 918 0.001/0.160 14563 ± 205 0.006/0.160 115544 ± 701 0.002/0.160 SD, standard deviation; Sample wt, weight of sample (g, three significant figures); AUFS, absorbance units full scale; NA, not analyzed. a Based on S/N ≥ 2. 0.14 0.12 Theob 0.08 0.06 Caff 0.04 Malt EtMalt -0.02 -0.04 0 2 4 6 8 EtVan 0.00 Ecat Cat 0.02 Van Absorbance Units 0.10 10 12 14 16 18 20 22 24 26 28 30 32 34 Min Figure 3. Chromatogram of the water extract of Flavor 1. Conditions: Gemini C18 (2.0 mm × 150 mm, 5 µm); mobile phase gradient of 0.3% acetic acid and MeOH at 500 µL min−1 ; injection volume of 5 µL; autosampler temperature 10 ◦ C; absorbance detection at 273 nm and 0.350 AUFS initially then 0.100 AUFS after 10 min; 0.02 g in 5 mL−1 ; maltol (Malt), theobromine (Theob), ethyl maltol (EtMalt), catechin (Cat), caffeine (Caff), vanillin (Van), epicatechin (Ecat), and ethyl vanillin (EtVan). were used for the different analytes, since the concentrations varied from not detected (<9 mg kg−1 ) for maltol and vanillin in Flavor 4 to 132 540 mg kg−1 vanillin in Flavor 2. Figure 3 is a chromatogram of the water extract of Flavor 1 intentionally enlarged to illustrate the chromatographic problems associated with these artificial flavors when large amounts of other compounds are present in the extract. One of the two peaks on either side of ethyl maltol may be (−)-epigallocatechin from the cocoa added to the formulation.1 These peaks may also be 4-hydroxybenzoic acid or 4hydroxybenzaldehyde from a vanilla extract used in Flavor 1.10,14 The peak around 20 min may be ethyl vanillic acid since Flavor 1 contained a large amount of ethyl vanillin. The presence of these compounds was not confirmed by authentic standards. J Sci Food Agric 88:1423–1430 (2008) DOI: 10.1002/jsfa Stability of standards It was observed when catechin and epicatechin standards were injected before and after the sample extracts (bracket calibration) that the set of standards analyzed after the sample extracts gave about 10% less response than the standards analyzed prior to the sample extracts. This occurred after a 16 h runtime with the autosampler temperature compartment set at 10 ◦ C. Standards of catechin and epicatechin also became discolored under ambient conditions in light after 3 days, with a response loss of around 20%. Some workers have used ascorbic acid to reduce the degradation of catechins.15 The vanillin standard began to yield two peaks after about a month under ambient conditions in the presence of light, the unknown peak having a retention time of approximately 2 min less than 1429 CH Risner and MJ Kiser vanillin (not shown). Since vanillin is an aldehyde, the expected oxidized form of vanillin would be vanillic acid. An authentic standard of vanillic acid was obtained and the additional peak in the vanillin standard was confirmed by retention time and spiking experiments to be vanillic acid. An analogous situation also occurred with ethyl vanillin, with an unknown peak having a retention time of about 2 min less than ethyl vanillin (not shown). Although an authentic standard of ethyl vanillic acid was not used to confirm the presence of this compound (not commercially available), it does show that ethyl vanillin, like vanillin, oxidizes similarly to vanillin, forming a more polar compound which is less retained on the C18 column. These two degradation products of vanillin and ethyl vanillin were seen in Hershey’s chocolate milk and Hershey’s chocolate syrup. Even though these consumer products were refrigerated (5 ◦ C) as recommended and were analyzed within 1 day after purchase, vanillic acid was found in Hershey’s chocolate milk and ethyl vanillic acid appeared in Hershey’s chocolate syrup. CONCLUSIONS An accurate and precise method has been developed which is capable of simultaneously quantifying theobromine, catechin, vanillic acid, caffeine, vanillin, epicatechin, and ethyl vanillin from the water extract of consumer products and maltol, theobromine, ethyl maltol, caffeine, vanillin, and ethyl vanillin from artificial flavors. Minimal sample preparation is required and equipment found in most laboratories can be used to perform the analyses. Results from this procedure compare well with other RP-HPLC techniques and is the first reported single procedure to determine theobromine, catechin, caffeine, and epicatechin in Standard Reference Material 2380 baking chocolate without extensive sample preparation and mass selective detection. Vanillic acid and ethyl vanillic acid, found in some consumer products, may be indicators of shelf-life stability since it would be not likely that these compounds were added intentionally to consumer products. 2 3 4 5 6 7 8 9 10 11 12 13 14 15 compounds in chocolate using RP-HPLC. Eur Food Res Technol 215:340–346 (2002). Meyer A, Ngiruwonsanga T and Henze G, Determination of adenine, caffeine, theophylline and theobromine by HPLC with amperometric detection. Fresenius J Anal Chem 356:284–287 (1996). Hurst WJ, Snyder KP and Martin RA, Use of microbore highperformance liquid chromatography for the determination of caffeine, theobromine and theophylline in cocoa. J Chromatogr 318:408–411 (1985). Bispo MS, Veloso MCC, Pinheiro HLC, DeOliveira RFS, Reis JON and DeAndrade JB, Simultaneous determination of caffeine, theobromine and theophylline by high performance liquid chromatography. J Chromatogr Sci 40:45–48 (2002). May EW and Watters RL, National Institute of Standards and Technology, Certificate of Analysis Standard Reference Material 2384 Baking Chocolate, Department of Commerce, USA (2004). Kreiser WR and Martin RA, High pressure liquid chromatographic determination of theobromine and caffeine in cocoa and chocolate products. J Assoc Off Anal Chem 61:1424–1427 (1978). Naik JP, Improved high-performance liquid chromatography method to determine theobromine and caffeine in cocoa and cocoa products. J Agric Food Chem 49:3579–3583 (2001). Subagio A, Sari P and Morita N, Simultaneous determination of (+)-catechin and (−)-epicatechin in cacao and its products by high performance liquid chromatography with electrochemical detection. Phytochem Anal 12:271–276 (2001). Nelson BC and Sharpless KE, Quantification of the predominate monomeric catechins on baking chocolate standard reference material by LC/APCI-MS. J Agric Food Chem 51:531–537 (2003). Herrmann A and Stoeckli M, Rapid control of vanillacontaining products using high performance liquid chromatography. J Chromatogr 246:313–316 (1982). Peng SH, Ma WH and Di JW, Determination of maltol and ethyl maltol by spectrophotometry, HPLC and voltammetry. Guangpu Shiyanshi 22:680–682 (2005). Nagae N, Enami T and Doshi S, The retention behavior of reversed-phase HPLC columns with 100% aqueous mobile phase. LCGC North America 20:964–972 (2002). Marx F and Maia JGS, Purine alkaloids in seeds of Theobroma species from the Amazon. Z Lebensm Unters 193:460–461 (1991). Scharrer A and Mosandi T, Reinvestigation of vanillin contents and component ratios of vanilla extracts using highperformance liquid chromatography and gas chromatography. Dtsch Lebensm Rundsch 97:449–456 (2001). Yang SY, Lee MJ and Chen L, Human salivary tea catechin levels and catechin esterase activities: implication in human cancer prevention studies. Cancer Epidemiol Biomarkers Prev 8:83–89 (1999). REFERENCES 1 Tokusoglu O and Uenal MK, Optimized method for simultaneous determination of catechin, gallic acid and methylxanthine 1430 J Sci Food Agric 88:1423–1430 (2008) DOI: 10.1002/jsfa J Sci Food Agric 88:1431–1441 (2008) Journal of the Science of Food and Agriculture Fatty acid and fat-soluble antioxidant concentrations in milk from high- and low-input conventional and organic systems: seasonal variation Gillian Butler,1∗ Jacob H Nielsen,2 Tina Slots,2 Chris Seal,1 Mick D Eyre,1 Roy Sanderson3 and Carlo Leifert1 1 School of Agriculture, Food and Rural Development, Newcastle University, Nafferton Farm, Stocksfield, Northumberland, NE43 7XD, UK of Food Science, Danish Institute for Agricultural Science (DIAS), PO Box 50 DK-8830 Tjele, Denmark 3 Institute for Research on Environment and Sustainability, Newcastle University, Devonshire Building, Devonshire Place, Newcastle Upon Tyne NE1 7RU, UK 2 Department Abstract BACKGROUND: Previous studies showed differences in fatty acid (FA) and antioxidant profiles between organic and conventional milk. However, they did not (a) investigate seasonal differences, (b) include non-organic, lowinput systems or (c) compare individual carotenoids, stereoisomers of α-tocopherol or isomers of conjugated linoleic acid. This survey-based study compares milk from three production systems: (i) high-input, conventional (10 farms); (ii) low-input, organic (10 farms); and (iii) low-input non-organic (5 farms). Samples were taken during the outdoor grazing (78 samples) and indoor periods (31 samples). RESULTS: During the outdoor grazing period, on average, milk from the low-input systems had lower saturated FAs, but higher mono- and polyunsaturated FA concentrations compared with milk from the high-input system. Milk from both the low-input organic and non-organic systems had significantly higher concentrations of nutritionally desirable FAs and antioxidants – conjugated linoleic (60% and 99%, respectively) and α-linolenic (39% and 31%, respectively) acids, α-tocopherol (33% and 50%, respectively) and carotenoids (33% and 80%, respectively) – compared with milk from the high-input system. Milk composition differed significantly between the two low-input systems during the second half of the grazing period only; with milk from non-organic cows being higher in antioxidants, and conjugated linoleic acid, and that from organic cows in α-linolenic acid. In contrast, few significant differences in composition were detected between high-input and low-input organic systems when cows were housed. CONCLUSIONS: Milk composition is affected by production systems by mechanisms likely to be linked to the stage and length of the grazing period, and diet composition, which will influence subsequent processing, and sensory and potential nutritional qualities of the milk.  2008 Society of Chemical Industry Keywords: milk; low-input farming; organic farming; fatty acid profiles INTRODUCTION The fatty acid (FA) and fat-soluble antioxidant composition in milk fat is known to affect processing and sensory quality of dairy products,1,2 and may also affect their nutritional value.3 – 5 The degree of saturation in milk fat has a bearing on the hardness, texture and taste of manufactured dairy products, particularly butter and cheese.6 The presence of longer-chain saturated fatty acids (SFA) increases the hardness of butter, while milk with a high proportion of unsaturated FA content (typical range 275–400 g kg−1 fat) tends to give softer products (e.g., more spreadable butter). Unsaturated (especially polyunsaturated) FAs are also more prone to oxidation, which results in the development of off-flavour and reduced shelf-life in milk and dairy products.6 However, the sensory quality and shelf-life of milk and dairy products is determined by the balance of unsaturated FAs and fat-soluble antioxidants, which protect against oxidation and off-flavour development.6 – 8 High dietary intakes of SFA (which account for 60–70% of milk fat) is a risk factor for development of obesity, cardiovascular disease (CVD), impaired insulin sensitivity and the ‘metabolic syndrome’.4 In ∗ Correspondence to: Gillian Butler, School of Agriculture, Food and Rural Development, Newcastle University, Nafferton Farm, Stocksfield, Northumberland, NE43 7XD, UK E-mail: Gillian.Butler@ncl.ac.uk (Received 26 July 2007; revised version received 15 January 2008; accepted 15 January 2008) Published online 18 April 2008; DOI: 10.1002/jsfa.3235  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 G Butler et al. contrast, dietary intake of certain unsaturated fatty acids, in particular conjugated linoleic acid (CLA) and omega-3 fatty acids (n-3 FA), and fat-soluble antioxidants (e.g., α-tocopherol, carotenoids) has been linked to potential health benefits.3,9,10 CLA and n3 FA have been shown to counteract the negative physiological effects of SFA, and CLA has also been linked to anticancer properties, reduced risk of type 2 diabetes, CVD and enhanced immune function.11 – 13 However, while CLA isomer C18:2 c9 t11 (CLA9) was only linked to beneficial health impacts, another CLA isomer, C18:2 t10 c12 (CLA10), was also associated with some negative health impacts in cell culture and animal models.13 In studies comparing the impact of different (e.g., organic and conventional) production systems on milk fat composition, it is therefore important to compare concentrations of both CLA isomers. Most previous comparative studies14 – 16 only reported concentrations of individual isomers or total CLA and also did not report concentrations of vaccinic acid (VA), the precursor for CLA. Milk contains significant concentrations of VA and, since a proportion can be readily converted to CLA9 in the human body, the total potential CLA9 supply can only be estimated if both VA and CLA9 levels are known.17 Previous studies showed that the feeding regime has a major effect on the FA profiles of milk, but that other factors (including breed/genotype, stage and number of lactations) may also influence milk composition.17 – 19 Dietary unsaturated fatty acids are likely to undergo hydrogenation by rumen microorganisms and long-chain fatty acids may be subjected to desaturase activity in the mammary gland.17 – 20 The FA profile of milk, therefore, is primarily determined by: (i) the balance of fatty acids in the diet; (ii) the extent of rumen hydrogenation; and (iii) mammary desaturase activity. CLA levels are linked to dietary supply of α-linolenic acid (αLA) and linoleic acid.17 However, while 70–90% of CLA9 (which constitutes >70% of total CLA in milk) is generated from desaturation of VA in the mammary gland, all other CLA isomers (including CLA10) are generated as intermediates of rumen biohydrogenation and are therefore found at much lower concentrations than CLA9 in milk.17 Fat-soluble antioxidants/vitamins present in milk are derived from dietary sources, either from (i) natural constituents in feedstuff (especially the forage component of the diet)21 or (ii) synthetic compounds added as supplements to the diet of lactating cows.22 Carotenoids derived from fresh forage are dominated by β-carotene, but also include lutein, zeaxantin, cryptoxanthin, lycopene and α-carotene.23 The main vitamin E activity in fresh forage is associated with the RRR isomer of α-tocopherol (the only isomer synthesized by plants), with some activity being associated with 1432 β-, γ - and δ -tocopherol and α-, β,- γ - and δ tocotrianol.24 Most high-input conventional dairy production systems supplement diets with proprietary mineral and/or vitamin products containing A vitamins, vitamin D3 and E vitamins (in particular αtocopherol); such supplements are prohibited in organic production.25 The naturally occurring RRR isomer of αtocopherol has a higher vitamin E activity (1.49 IU mg−1 ) than synthetic vitamin E (1.0 IU mg−1 ), which contains equal proportions of the eight different stereoisomers of α-tocopherol.24 Synthetic α-tocopherol products are referred to as ‘all rac’ αtocopherol and consist mainly of 2R stereoisomers. Synthetic α-tocopherol is absorbed with the same efficiency as the RRR stereoisomer of α-tocopherol, but levels of uptake into key tissues (e.g. the brain) are lower.24 Also, a recent study with dairy cows found higher α-tocopherol concentrations in blood and milk following supplementation of RRR compared with ‘allrac’ α-tocopherol and reported preferential transfer of RRR isomers into milk by cows receiving the synthetic isomer mix.22 Milk and dairy products from certified organic dairy production systems have been reported to contain higher concentrations of polyunsaturated fatty acids (PUFA), αLA (the main n-3 FA in milk), and/or CLA, and fat-soluble antioxidants than those from high-input conventional production.14 – 16 These studies did not include non-organic, low-input systems in comparisons. However, an increasing number of dairy farms in Europe, New Zealand/Australia and North America are adapting ‘lower-input’ production methods similar to those used in organic farming, but do not comply with all input restrictions prescribed by organic farming standards.26 Most importantly, these systems use mineral NPK fertilizers, but often at reduced levels compared with conventional highinput systems. It is unclear whether such non-organic, low-input systems can provide similar benefits in milk composition to certified, organic dairy production systems. Milk composition is known to change when switching from outdoor grazing to indoor forage-based diets in winter;6,12,20,27 however, little is known about whether this dietary change affects the differential in milk quality between organic and conventional systems reported previously.14 – 16 There is also limited information on differences in the composition of fat-soluble antioxidants in milk from high- and low-input dairy systems and the few studies available show contradictory results.14,28,29 Such information would, however, be essential to assess (i) the overall nutritional value of milk from lowinput systems and (ii) whether the higher unsaturated fat content of organic milk (and associated risk of oxidation and off-flavour development) is compensated for by higher concentrations of antioxidants. J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa Fatty acid and fat-soluble antioxidant concentrations in milk The objectives this study were therefore to: (i) compare the fatty acid and fat-soluble antioxidant composition of milk from three UK production systems – certified-organic ‘low-input’ (O-LI), nonorganic certified ‘low-input’ (NO-LI) and standard ‘high-input’ (HI) conventional production systems, during the outdoor grazing period; (ii) quantify differences in fatty acid and fat-soluble antioxidant content of milk between O-LI and HI systems, during the winter indoor (conserved forage-based) feeding period; and (iii) identify whether there are differences in milk composition between certifiedorganic ‘low-input’ (O-LI) and non-certified ‘lowinput’ (NO-LI) systems that use spring block calving systems and graze cows outdoors throughout lactation. EXPERIMENTAL Farm details and milk survey design One hundred and nine milk samples were collected from 25 commercial farms categorized into three different production systems. Management and production parameters were recorded for each farm and sampling date using a standard questionnaire (see Table 1 for the most important parameters recorded). The number of cows in early lactation (first 100 days) was also recorded. Live weights (LW) of cows were estimated based on mean weights of breeds (Holstein–Friesian = 650 kg; Jersey = 450 kg; Ayrshire = 550 kg; Brown Swiss and Scandinavian Red = 575 kg)30 and the proportion of each breed in the genetic make-up of the herd. Total dry matter intakes (DMI) were estimated from average milk yields (bulk tank contents divided by the number of milking cows recorded by farmers) Table 1. Differences in management and production system parameters between high-input conventional (HI), organically certified (O-LI) and non-organic (NO-LI) low-input farms (mean values over all samples, with standard deviation in parentheses) Production system Parameters recorded Herd characteristics Herd size (milking cows)∗ Breed Indexa % primiparous cows∗ Live weight of cows (kg)b Dry matter intake (kg d−1 )c Diet composition 1. Outdoor period Fresh forage (proportion DMI) Conserved forage (proportion DMI) • Grass silagee∗ • Maize silagee∗ • Other silaged,e∗ • Straw/haye∗ Concentrate (proportion of DMI) • Cereals∗ • By-products∗ g • Other concentrates∗h,i Mineral supplements∗ (g cow−1 day−1 ) Vitamin E supplement∗ (iu cow−1 day−1 ) 2. Indoor period Fresh forage (proportion of DMI)f Conserved forage (proportion of DMI) • Grass silagee∗ • Maize silagee∗ • Other silaged,e∗ • Straw/haye∗ Concentrate (proportion of DMI) • Cereals∗ • By-products∗g • Other concentrates∗h,i Mineral supplements∗ (g cow−1 day−1 ) Vitamin E supplement∗ (iu cow−1 day−1 ) HI O-LI NO-LI 252 (125) 0.0 (0) 25 (7) 650 (0) 19.5 (0.5) 160 (93) 0.2 (0.3) 27 (12) 610 (34) 17.6 (1.0) 322 (141) 0.3 (0.1) 30 (8) 588 (21) 16.9 (0.7) 0.37 (0.24) 0.29 (0.15) 0.84 (0.23) 0.08 (0.16) 0.95 (0.07) 0 (0) 0.73 (0.28) 0.10 (0.20) 0.13 (0.18) 0.04 (0.09) 0.72 (0.40) 0 (0) 0 (0) 0.28 (0.40) 0.34 (0.13) 0.08 (0.09) 0.05 (0.07) 0.31 (0.24) 0.30 (0.23) 0.40 (0.31) 142 (75) 450–750 0.23 (0.40) 0.20 (0.40) 0.57 (0.49) 8 (17) 0 0.05 (0.14) 0.52 (0.50) 0.43 (0.53) 3 (13) 0 0 (0) 0.56 (0.08) 0.24 (0.38) 0.54 (0.30) 0.69 (0.29) 0.05 (0.12) 0.24 (0.28) 0.02 (0.04) 0.80 (0.19) 0 (0) 0.20 (0.19) 0 (0) NA NA NA NA NA NA NA NA NA NA NA 0 0.44 (0.08) 0.23 (0.10) 0.31 (0.17) 0.24 (0.16) 0.45 (0.24) 150 (53) 250–674 0.42 (0.16) 0.07 (0.11) 0.51 (0.23) 22 (31) 0 ∗ Based on farm records and collected by questionnaire; a estimated proportion of non-Holstein–Friesian genetics in the herd; b estimated based on breed index; c estimated based on live weight and milk yield; d whole-crop wheat, barley and/or oats, dry matter; e proportion of total conserved forage intake; f when weather permitted, most organic herds were grazed in the day; g brewing and distillers’ waste and/or sugarbeet pulp; h bought in or farm produced compound/mixed concentrate feeds; i no oilseed or fat supplementation was recorded by farmers; NA, not applicable (NO-LI cows were grazed throughout the lactation). J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa 1433 G Butler et al. and assumed live weight (DMI = 0.025 LW + 0.125 milk yield). Grazing was calculated at the herd level by difference: DMI (fresh grass) = total DMI − DMI (conserved forage + concentrate; recorded by producers). Since cow live weight varied between farming systems, recorded levels of dietary components were used to calculate proportions of total intake, to allow a more relevant comparison between systems. Tables 1 and 2 list diet composition for each production system during grazing and the housed periods of this study. Conventional ‘high-input’ (HI) farms Ten farms were selected representing common conventional production and feeding systems in the UK. HI farms used predominantly pure ryegrass swards during the grazing period, winter diets based on grass silage and higher concentrate:conserved forage ratio diets during the indoor feeding period than LI farms (see Table 1 for the diets used during the outdoor grazing and indoor feeding periods). The HI group did not include farms with extremely highinput/output systems (e.g., farms which use more than 50% of the diet coming from concentrates, regularly milk three times per day and/or those that house animals throughout their lactation). All farms were all-year round-calving and had similar proportions of cows in early lactation at all sampling dates. Organically certified ‘low-input’ (O-LI) farms Ten farms were selected representing two principal organic dairy systems found in the UK: (a) an all-yearround calving system (five farms) in which lactating cows are grazed when conditions allow (spring to autumn), but fed on conserved forage-based diets during the winter indoor period (see Table 1); and (b) a spring block calving system in which cows Table 2. Diet composition in organic (O-LI) and non-organic (NO-LI) low-input dairy production systems (spring calving herds only), at different sampling dates during the outdoor period (mean values, with standard deviation in parenthesis) Production system Sampling date Dietary components (proportion of DMI) August O-LI NO-LI Fresh forage Conserved forage Concentrate 0.96 (0.04) 0 (0) 0.04 (0.04) 0.92 (0.08) 0 (0) 0.08 (0.08) October Fresh forage Conserved forage Concentrate 0.88 (0.11) 0.04 (0.06) 0.08 (0.08) 0.95 (0.08) 0 (0) 0.05 (0.08) March Fresh forage Conserved forage Concentrate 0.86 (0.20) 0.11 (0.15) 0.03 (0.06) 0.95 (0.07) 0 (0) 0.05 (0.07) May Fresh forage Conserved forage Concentrate 0.96 (0.06) 0 (0) 0.04 (0.06) 1.00 (0) 0 (0) 0 (0) DMI, dry matter intake. 1434 are grazed throughout lactation (March to October) and were only indoors when not lactating between November to February. All-year-round calving farms had similar proportions of cows in early lactation at all sampling dates. Diets used in both organic systems were similar during the outdoor grazing period (Table 1); all O-LI farms used mixed grass–clover swards and did not apply mineral N or water-soluble P fertilizers. Where appropriate, on the basis of soil analyses, finely ground rock phosphate fertilizers were applied. Non-organically certified ‘low-input’ (NO-LI) farms Five farms representing the main non-organic, ‘lowinput’ system found in the UK were selected. All farms used a New Zealand-type production system26 with spring block calving, in which cows were grazed throughout the lactation and no, or low levels of concentrate and/or other feed supplements included in the diet (see Table 1). As with the organic spring block calving herds, cows were only housed when not lactating between November and February. NO-LI farms selected used mixed grass–clover swards, but applied up to 120 kg N ha−1 per year of mineral N and water-soluble P fertilizer at levels determined from soil analyses. Samples were taken in August and October in 2004 and in January, March and May in 2005 from all farms. In January 2005 samples could only be collected from O-LI and HI farms that used an all-year-round calving system. Samples of milk were taken from the stirred bulk tank after two milkings (representing a 24 h production period), at each participating farm and frozen immediately after sampling and kept at −20 ◦ C until dispatched for analysis. Extraction of fat from milk The extraction of fat from the milk was carried out as described by Havemose et al.,23 with minor modifications. Milk fat was extracted from milk (2 mL) by adding methanol (2 mL) and chloroform (4 mL). The mixture was shaken vigorously for 1 min, then centrifuged for 10 min at 3000 × g at 4 ◦ C. The lower phase containing the lipid fraction was isolated and evaporated to dryness under nitrogen. Methylation of fatty acids from milk The methylation of fatty acids extracted from the milk was carried out as described by Havemose et al.,23 with minor modifications. Fat (approx. 10 mg) was dissolved in sodium methylate solution (2 g L−1 methanol) in sealed glass tubes filled with argon, incubated at 60 ◦ C for 30 min, and then cooled on ice. Saturated sodium chloride solution (4 mL) and pentane (1 mL) were added. The samples were mixed on a vortex mixer for 1 min and centrifuged at 1700 × g for 10 min. The upper pentane phase was collected and used for gas chromatographic analysis. J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa Fatty acid and fat-soluble antioxidant concentrations in milk Analysis of fatty acid composition by gas chromatography Separation and quantification of the fatty acids isolated from milk was carried out as described by Havemose et al.,23 with modifications. Samples (1 µL) of the pentane phase containing the fatty acid methyl esters were analysed by gas chromatography (HP6890 GC system, Hewlett Packard Co., Palo Alto, CA, USA) with a flame ionization detector and a Supelco SI 2560 column (100 m × 0.25 mm × 0.20 µm, Supelco, Bellafonte, PA, USA). The inlet temperature was 275 ◦ C with a split ratio of 40:1, and the carrier gas was helium with a constant flow of 1.5 mL min−1 . The starting temperature of 140 ◦ C was held for 5 min and increased by 4 ◦ C min−1 to an end temperature of 240 ◦ C. The detector temperature was 300 ◦ C. The concentrations of saturated (SFA), monounsaturated (MUFA) and polyunsaturated (PUFA) fatty acids and the ratio of n-3 and n-6 isomers of linolenic acid (C18:3) were then calculated as a proportion of total fatty acids recovered, based on the use of external standards. To calculate the n-3:n-6 FA ratio, the concentration of the main n-3 FA (α-LA) was divided by the sum of the concentrations of the following n-6 FA isomers: 18:2 t9 t12, 18:2 t10 t12, 18:2 c9 c12, 18:3 c6 c9 c12 and 20:4 c5 c8 c11 c14. Analysis of fat-soluble antioxidant composition Fat-soluble antioxidants (α-tocopherol, β-carotene, lutein and zeaxantin) were analysed using the high-performance liquid chromatographic method described by Havemose et al.23 Isomers of αtocopherol were analysed using the methods described by Meglia et al.22 Statistical analysis Linear mixed-effects models31 were used to investigate differences in milk quality parameters under the different systems (HI, O-LI and NO-LI). These models use two types of explanatory variables: fixed effects, which affect the mean of the response variable; and random effects, which affect the variance of the response. In these analyses, farm identifier was used as a random effect. Three sets of analyses were undertaken: (i) comparison of milk samples from all three systems (HI, OLI and NO-LI) taken during the outdoor grazing period (samples from the spring block and allyear calving organic farms were pooled, because no major differences could be detected in preliminary analyses; results not shown); (ii) comparison of samples taken from HI and all-year calving O-LI farms during the indoor period when cows were on conserved forage-based diets; and (iii) comparison of samples taken from spring block calving OLI and NO-LI herds at four different sampling dates using a two-factorial model (system and date), adapted to account for repeated measures from the four dates, to identify (a) whether at any time during the grazing period milk quality differed J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa between the two LI systems and (b) interactions between the two factors for any of the milk quality parameters assessed. All proportion data were arcsine transformed prior to statistical analysis, but means presented were calculated from non-transformed data. Pairwise comparisons of means were carried out, where appropriate, using Tukey’s honest significant difference tests. All statistical analyses were carried out using the R statistical environment.32 RESULTS Comparison of milk fat composition during the outdoor period (fresh forage-based diets) On average the total fat content was higher in milk from LI systems compared with the HI system, and was significantly higher for the NO-LI system compared with the HI system (Table 3). When the composition of milk fat was compared, on average, the percentage of SFAs in milk fat was lower, while percentages of both MUFA (of which >80% was oleic acid C18:1 cis9) and PUFA were higher in milk from LI systems, compared with the HI system, and was significantly higher for the NO-LI system compared with the HI system (Table 3). Percentages of the nutritionally desirable FAs (αLA and CLA9) were significantly higher, while levels of total n-6 PUFAs were significantly lower in milk from both LI systems, when compared with milk from HI farms (Table 3). As a result, the n3:n6 ratio was also higher in milk from LI systems (Table 3). CLA10 was found in low concentrations in milk from all production systems and was not affected by production system (Table 3). Differences between O-LI and NO-LI were generally smaller than those between HI and LI systems, but the percentage of CLA was significantly higher in milk from NOLI systems and the percentage of total n-6 FA was significantly higher in milk from O-LI systems (Table 3). The concentrations of most antioxidants (the RRR stereoisomer of α-tocopherol, β-carotene, lutein and zeaxanthin) were highest in milk from NO-LI, at intermediate concentrations in milk from O-LI and lowest in milk from HI systems (Table 3) during the outdoor period. Concentrations of the 2R stereoisomer of α-tocopherol were not significantly different between systems, but were slightly lower in milk from NO-LI systems. Comparison of milk fat composition during the indoor period (conserved forage-based diets) Since the spring, block-calving NO-LI and O-LI systems did not produce milk during the indoor period only milk from all-year calving O-LI and HI systems was compared. 1435 G Butler et al. Table 3. Fatty acid composition and fat-soluble antioxidant concentrations in milk from conventional high-input and organic and non-organic low input dairy production systems, during the outdoor, fresh forage-based feeding period (mean values, with standard error of means in parentheses) Production system Low-input Characteristic assessed High-input O NO ANOVA (P-value) Number of samples Milk yield/cow (kg) Protein content (g kg−1 ) Fat content (g kg−1 ) 24 34 20 26.2 (0.7)a 33.1 (2.3)c 39.6 (3.1)b 18.4 (0.8)b 34.1 (3.5)b 42.0 (6.9)ab 17.4 (0.9)b 35.9 (3.9)c 45.5 (9.0)a 691 (59)b 275 (54)b 59 (20)b 672 (55)ab 289 (51)ab 82 (17)a 660 (64)a 305 (57)a 78 (22)ab Omega 3 and 6 FAs (g kg−1 milk fat) α-LA C18:3 c9 c12 c15 6.2 (0.5)b γ LA C18:3 c6 c9 c12 0.26 (0.01) Total n-6 20.1 (1.3)a n-3:n-6 ratio 0.37 (0.13)b 10.2 (0.3)a 0.26 (0.06) 15.2 (1.0)b 0.79 (0.09)a 9.0 (0.3)a 0.14 (0.01) 10.6 (0.4)c 0.88 (0.01)a <0.0001 0.242 <0.0001 <0.0001 VA and CLA isomers (g kg−1 milk fat) VA C18:1 t11 22.5 (1.8)b CLA C18:2 c9 t11 8.8 (0.7)c CLA C18:2 t10 c12 0.31 (0.03) 35.5 (1.6)a 14.1 (0.6)b 0.33 (0.03) 41.9 (1.9)a 17. 5 (1.4)a 0.38 (0.07) <0.0001 <0.0001 0.589 <0.0001 0.0006 0.0004 Fatty acid groups (g kg−1 milk fat) Total SFA Total MUFA∗ Total PUFA 0.042 0.017 0.0017 Fat-soluble antioxidants (mg kg−1 milk fat) α-Tocopherol 2R α-toc RRR α-toc Total α-tocopherol 2.6 (0.1) 18.8 (0.8)c 21.4 (0.8)b 2.5 (0.3) 26.0 (0.9)b 28.5 (0.9)a 1.8 (0.2) 30.2 (1.0)a 32.0 (1.1)a 0.123 <0.0001 <0.0001 Carotenoids β-Carotene Lutein Zeaxantin Total carotenoids 5.35 (0.33)c 0.46 (0.03)c 0.11 (0.01)c 5.91 (0.35)c 6.95 (0.29)b 0.77 (0.04)b 0.16 (0.01)b 7.88 (0.32)b 9.29 (0.48)a 1.14 (0.05)a 0.20 (0.01)a 10.64 (0.52)a <0.0001 <0.0001 <0.0001 <0.0001 O, organic; NO, non-organically certified; SFA, saturated fatty acids; MUFA, monounsaturated fatty acids (∗ >80% oleic acid); PUFA, polyunsaturated fatty acids; α-LA, α-linolenic acid; 2R α-toc, 2R stereoisomers of α-tocopherol; RRR α-toc, 3R stereoisomers of α-tocopherol; means within a row with different letters are significantly different (P < 0.05). In contrast to results from the outdoor rearing period, there were few differences in milk composition during the housed period. The percentages of total SFA in milk fat were significantly higher (4%) and MUFA significantly lower (10%) in milk from the O-LI system compared with milk from HI systems (Table 4). There was also a significantly lower (24%) content of n-6 fatty acids and trends towards a higher content (38%) of α-linolenic acid (P = 0.052) and a higher (30%) lutein content (P = 0.081) in O-LI milk compared with HI milk (Table 4). Comparison of milk fat composition during the grazing period between O-LI and NO-LI spring block calving dairy systems Apart from CLA9 isomer (which was present in significantly higher percentages in milk from NOLI farms on the August and May sampling dates), significant differences in FA composition between OLI and NO-LI block calving systems were found only late in the outdoor grazing period (August and October sampling date, Fig. 1). The percentages of total SFA 1436 and αLA were higher in milk from O-LI systems, while percentages of MUFA, PUFA, VA and CLA9 were higher in milk from NO-LI systems. No significant differences in the percentages of CLA10 and n-6 FAs were detected (data not shown). There were also significant interactions between LI production system and date for PUFA (P = 0.020; Fig. 1(c)), VA (P = 0.029; Fig. 1(e)) and CLA (P = 0.030; Fig. 1(f)). The concentration of most antioxidants changed significantly over time, and at specific dates significant differences in the concentrations of individual antioxidants between the two LI systems could be detected. Concentrations of 2R toc were significantly higher in milk from O-LI systems in May, while concentrations of 3R toc were significantly higher in NO-LI systems in October. Levels of total and all three individual carotenoids were significantly higher in milk from NOLI-systems in August and May (and for lutein also in October) (Fig. 2). A significant interaction between LI production system and date was only identified for the 2R stereoisomer of α-tocopherol (P = 0.003; Fig. 2(a)). J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa Fatty acid and fat-soluble antioxidant concentrations in milk 350 * 660 600 * * 300 Oct Mar Aug May 12 Oct Mar Aug CLA (g kg-1 total FA) VA (g kg-1 total FA) (e) * 40 Aug Oct Mar Aug May Mar May (f) * 20 * 10 20 8 Oct 30 (d) 10 * 90 May 60 * (c) 60 250 Aug ALA (g kg-1 total FA) 120 (b) PUFA (g kg-1 total FA) (a) * MUFA (g kg-1 total FA) SFA (g kg-1 total FA) 720 Oct Mar May Aug Oct Mar May Figure 1. Effect of organic (black bars) and non-organic (white bars) low-input production systems on the fatty acid composition of milk fat. (a) SFA, saturated fatty acids; (b) MUFA, monounsaturated fatty acids; (c) PUFA, polyunsaturated fatty acids; (d) ALA, α-linolenic acid; (e) VA, vaccinic acid; (f) CLA, conjugated linoleic acid isomer C18:2 c9 t11; ∗ means for organic and non-organic low input systems are significantly different according to Tukey’s honest significant difference test. Error bars indicate standard error of mean values. Two-way ANOVA (with production system and date as factors) identified significant differences (a) between production systems for VA (P = 0.041) and CLA (P = 0.012) and (b) between dates for PUFA (P = 0.028), VA (P = 0.005) and CLA (P < 0.0001). Significant interactions between system and date were identified for PUFA (P = 0.020), VA (P = 0.029) and CLA (P < 0.030). 3 2 * 30 25 20 1 Aug Oct 12 Mar Aug May Oct 1.5 (d) * * 8 Lutein (mg kg-1 fat) b carotene (mg kg-1 fat) (b) Mar * * 1.0 0.5 4 Aug Oct Mar May (c) * * 10 5 Aug Oct 0.3 (e) * 15 May Zeaxanthin (mg kg-1 fat) * Total carotenoids (mg kg-1 fat) 35 (a) 3R a toc (mg kg-1 fat) 2R a toc (mg kg-1 fat) 4 Mar May (f) * * 0.2 0.1 Aug Oct Mar May Aug Oct Mar May Figure 2. Effect of organic (black bars) and non-organic (white bars) low-input production systems on the levels of fat-soluble antioxidants in milk fat. (a) 2R α-toc, 2R stereoisomers of α-tocopherol; (b) 3R α-toc, 3R stereoisomers of α-tocopherol; (c) total carotenoids; (d) β-carotene; (e); lutein, (f); zeaxantin; ∗ means for organic and non-organic low-input systems are significantly different according to Tukey’s honest significant difference test. Error bars indicate standard error of mean values. Two-way ANOVA (with production system and date as factors) identified significant differences (a) between production systems for β-carotene (P = 0.003), lutein (P = 0.004), zeaxantin (P = 0.027) and total carotenoids (0.002), and (b) between dates for 2R α-toc (P = 0.0005), 3R α-toc (P = 0.0005), β-carotene (P = 0.005), lutein (P = 0.0008), zeaxantin (P = 0.002) and total carotenoids (0.003). A significant interaction between system and date was only identified for 2R α-toc (P = 0.003). DISCUSSION AND CONCLUSIONS Effect of feeding regimes on milk fat composition: outdoor grazing period The finding of lower percentages of SFA and contrasting higher percentages of MUFA in milk from the NO-LI system and higher PUFA (specifically α-LA and CLA9) and antioxidant content (α-tocopherol and carotenoids) of milk from both LI systems, compared J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa with that from HI farms during the outdoor grazing period, is not surprising in view of the contrasting diets. The two LI systems used a high level of fresh forage (>80% of DMI), with only half that level (<40%) used in HI systems. Increasing the level of fresh forage by similar margins was previously shown to elevate nutritionally desirable PUFA, CLA, α-LA and antioxidant percentages in milk17 – 21,27 to those 1437 G Butler et al. found in milk from LI and HI systems here. For example, CLA concentrations were previously shown to increase with the proportion of fresh grass intake, while high proportions of maize silage and/or cerealbased concentrates reduced CLA content.17,18,33 Also cutting and transport of grass to housed animals (a practice used to increase milk yield in zero-grazing systems) was also shown to decrease the CLA and VA content of milk by 50% and that of αLA content by 30%, compared to milk from cows grazing pasture.34 This response may have been due to rapid lipolysis of PUFA after harvest and/or a modification of rumen biohydrogentation.27 The finding that concentrations of CLA9 were significantly higher in milk from LI than HI systems, while concentrations of CLA10 were similar in both systems, was likely to be caused by contrasting effects of LI and HI diets on the biosynthesis of CLA9 which is mainly (70–90%) generated from VA in the mammary gland, and that of CLA10 which is a minor intermediate of rumen biohydrogenation.19 Table 4. Fatty acid composition and fat-soluble antioxidant concentrations in milk from conventional high-input and organic and non-organic low-input dairy production systems, during the indoor conserved forage-based feeding period (mean values, with standard error of means in parentheses) Characteristic assessed High-input Low-input organic ANOVA (P-value) Number of samples Milk yield/cow (kg) Protein content (g kg−1 ) Fat content (g kg−1 ) 21 10 26.5 (1.0) 33.0 (0.3) 40.8 (0.5) 19.1 (1.3) 33.1 (0.6) 42.1 (0.7) 0.0014 0.803 0.235 712 (6) 254 (5) 53 (2) 740 (11) 228 (10) 51 (4) 0.041 0.028 0.730 milk fat) 5.3 (0.5) 0.2 (0.02) 21.7 (1.3) 0.30 (0.04) 7.3 (0.9) 0.2 (0.03) 16.4 (0.7) 0.42 (0.06) 0.052 0.127 0.018 0.114 VA and CLA isomers (g kg−1 milk fat) VA C18:1 t11 16.4 (1.0) CLA C18:2 c9 t11 6.2 (0.04) CLA C18:2 t10 c12 0.31 (0.01) 17.5 (2.3) 7.8 (0.21) 0.34 (0.02) 0.636 0.111 0.139 Fatty acid groups (g kg−1 milk fat) Total SFA Total MUFA∗ Total PUFA −1 Omega 3 and 6 FA (g kg α-LA C18:3 c9 c12 c15 γ LA C18:3 c6 c9 c12 Total n-6 n-3:n-6 ratio Fat-soluble antioxidants (mg kg−1 milk fat) α-Tocopherol 2R α-toc RRR α-toc Total α-tocopherol 3.5 (0.4) 20.4 (0.9) 23.9 (1.0) 2.8 (0.4) 20.3 (1.5) 23.1 (1.6) 0.360 0.776 0.513 Carotenoids B-carotene Lutein Zeaxantin Total carotenoids 5.49 (0.41) 0.37 (0.03) 0.12 (0.01) 5.98 (0.44) 6.29 (0.64) 0.48 (0.06) 0.14 (0.01) 6.90 (0.68) 0.359 0.081 0.265 0.314 SFA, saturated fatty acids; MUFA, monounsaturated fatty acids (∗ >80% oleic acid); PUFA, polyunsaturated fatty acids; α-LA, αlinolenic acid; 2R α-toc, 2R stereoisomers of α-tocopherol; RRR α-toc, 3R stereoisomers of α-tocopherol. 1438 Previous studies have shown that VA in the rumen increases with increasing fresh forage and decreasing concentrate levels in dairy diets, while CLA10 generation in the rumen is relatively unaffected by changes in the diet except at very high levels of concentrate feeding.17,20 The greater dietary contribution from fresh forage is also the most likely explanation of elevated levels of RRR tocopherol and carotenoids in milk from the LI herds during the grazing period, compared to the HI milk. Transfer of β-carotene and α-tocopherol into milk was reported to be directly proportional to dietary supply, being highest in spring grazing.21 Effect of feeding regimes on milk fat composition: indoor period Few significant differences and trends in milk fat composition were found between HI and O-LI production systems during the indoor period when cows were fed conserved forage-based diets. This may have been due to feeding regimes used by OLI and HI herds being more similar during the indoor compared with the outdoor feeding period. The higher SFAs and lower MUFA content of organic milk during this feeding period are difficult to explain, since previous studies have shown that fresh forage intake (24% in organic as opposed to none in conventional winter diets) increases dietary PUFA supply.20,27 However, some previous studies have reported lower biohydrogenation rates for highconcentrate indoor diets,17,20 suggesting that the higher proportion of concentrate in the HI diets results in lower biohydrogenation and thereby lower SFA and higher MUFA, and that this effect overrides the effect of higher fresh forage intake in the O-LI animals. In order to allow milk from organic or LI production systems to be marketed as having ‘added nutritional value’ throughout the year, efforts need to be made to achieve higher concentrations of at least some to the nutritionally desirable compounds during the indoor feeding period, if year-round grazing is not an option. This could be achieved by supplementation of conserved forage-based winter diet with oil seeds (e.g., rapeseed, linseed, sunflower seed), a practice shown to significantly improve α-LA, VA, CLA9 and/or fatsoluble antioxidant concentrations in milk.12,17,33,35 – 37 Changes to the forage conservation methods may also increase the content of desirable FAs. For example, using hay rather then silage was also shown to increase the α-LA content in milk by up to 50%.33,38 It is interesting to note that in the UK it is very difficult to find farms feeding hay rather than silage, except among very traditional organic producers that work to biodynamic farming principles (which strongly recommend the use of hay for milking cows). Effect of vitamin feed supplements on antioxidant concentrations in milk Results of the study reported here suggest that the addition of synthetic vitamin/antioxidant supplements J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa Fatty acid and fat-soluble antioxidant concentrations in milk to feed in HI systems has a relatively minor effect on antioxidant concentrations in milk. For example, milk from HI herds, which received high levels of vitamin E supplements (in our study between 450 and 750 IUs vitamin E per day) contained significantly lower concentrations of total α-tocopherol during grazing than milk from farms working to organic farming standards, which do not permit feed supplementation with synthetic vitamins. It is particularly interesting that the concentration of the 2R stereoisomer of αtocopherol was not significantly higher in milk from the HI systems. The 2R stereoisomers account for most of the α-tocopherol in synthetic vitamin E supplements, but are virtually absent from natural sources of αtocopherol such as forage. This indicates either poor uptake of the 2R stereoisomers in the gastrointestinal system and/or preferential/selective uptake/transfer of 3R stereoisomers from the blood into milk in the udder, as reported previously.22 Potential effects of seasonal forage composition and availability on milk fat Differences in milk quality (both fatty acid profiles and antioxidant levels) were also detected between spring block calving O-LI and NO-LI systems which appeared to have very similar dietary regimes. These were more likely due to variation in the composition and/or total forage availability between the two systems over the season, since both systems grazed cows throughout the lactation and used very low levels of supplementary feeds such as conserved forage or concentrate. The finding that, in August, milk from O-LI systems had higher percentages of α-LA than milk from NO-LI systems is not surprising, and is likely to be due to a combination of two factors. Firstly, the use of mineral (especially N) fertilizers in the NO-LI system, a practice which has been shown to suppress the relative amounts of white clover in grass clover swards,39,40 and secondly, the impact of higher clover content causing elevation in concentrations of n-3 FAs in milk compared with ryegrass.27 However, it should be noted that most of the studies reviewed by Dewhurst et al.27 that compared the effect of clover and rye grass used ensiled forage, where reduced lipolysis in clover would have a greater influence over PUFA supply compared with fresh forage. The significantly higher CLA and antioxidants in milk from NO-LI systems are more difficult to explain, but may be related to differences in the nutritional composition of the herbage resulting from the grazing systems used (e.g., the length of time allowed for pasture regrowth between grazing periods), which has also been shown to affect the fatty acid composition of milk.27 Milk yields, protein and urea content in this study (data not shown) did not differ at times when differences in milk fat composition were detected between the two LI systems. This suggests that differences in milk fat composition were unlikely to be linked to contrasting energy or protein supply levels. However, since sward composition and total forage availability J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa were not monitored in the study reported here this will have to be tested in future studies. Potential effects of dairy genotypes on milk fat composition The higher proportion of fresh forage in the dairy diet is likely to have been the main reason for the differences in milk composition. However, since contrasting dairy genotypes (breed index) were used in different production systems this may also have contributed to the differences in milk composition recorded between systems. There is relatively little quantification of the effect of breed on fatty acid composition, although breed effects on CLA and antioxidant content have been reported to vary by up to 15–20% between breeds.21,35 This differential is considerably lower than the 60–99% for CLA9 and 30–140% for antioxidants measured between HI and LI systems recorded in this study. The finding of substantial differences in milk fat composition between HI and LI systems during the outdoor grazing period, but similar milk composition during the indoor feeding period, also suggests that the differences in feeding regimes (rather than dairy genotypes) were the main factors responsible for the milk composition differences between systems. However, the exact influence of breed relative to dietary supply and possible interaction needs to be determined in future studies. Potential nutritional impacts of differences in milk fat composition Differences in nutritionally desirable FA and antioxidants between HI and LI systems during the grazing period were generally quite large (65% and 45% for α-LA, 60% and 99% for CLA9, 33% and 50% for α-tocopherol, 30% and 74% for β-carotene, 67% and 148% for lutein and 46% and 82% for zeaxanthin, for O-LI and NO-LI systems, respectively). This confirms previously published comparisons of conventional and organic, low-input production systems carried out in Germany, Italy and the UK.14 – 16 Consumption of milk and milk products from LI systems produced during this period may therefore contribute significantly to increasing the intake of these compounds in line with nutritional recommendations. Importantly, the higher percentages of nutritionally desirable PUFA (CLA9 and α-LA) found in milk from LI systems did not coincide with a significant increase in nutritionally less desirable PUFA (e.g., CLA10, total n-6 FA). Also, the higher n-3 FA and lower n-6 FA percentages found in milk from LI systems resulted in a higher n-3:n-6 FA ratio, which is also considered nutritionally desirable.4,10,12,27,41 Even if trends of elevated α-LA and lutein in organic milk produced during housing were confirmed, it is clear that consumption of organic milk produced during the indoor winter period will not increase the intake of nutritionally desirable compounds to the same extent as low-input milks produced during the outdoor grazing period. 1439 G Butler et al. While CLA9 and n-3 FA have been linked to a range of beneficial impacts on health,10 – 13 it should be pointed out that it is currently uncertain whether the main n-3 FA found in milk, α-linolenic acid (αLA; C18:3 c9 c12 c15), has similar effects on human health as the long-chain n-3 FAs found mainly in fish oil (C20 or longer), which have been shown to protect against coronary heart disease, associated with improved neurological function and linked to reduced risk of type 2 diabetes, hypertension and certain cancers.10,12,41,42 These long-chain n-3 fatty acids are known to be present at low levels in milk fat27 and were not determined in this study. However, there is now both direct and indirect evidence that significant levels of longer-chain n-3 FAs, especially eicosapentaenoic acid (EPA; C20:5 n-3) and to a lesser extent docosahexaenoic acid (DHA; C22:6 n-3), are generated from αLA in humans.42 The impact of fat-soluble antioxidants/vitamins on human health has been reviewed extensively.24,43 – 45 Beneficial effects of increased dietary α-tocopherol (a compound belonging to the vitamin E group) intake on human health have mainly been linked to its ability to reduce oxidative stress, which was shown to be a risk factor for a number of chronic health conditions including cardiovascular disease, cancer, impaired immunity and premature ageing.45 Carotenoids can act as precursors for vitamin A, although a range of health benefits were linked to their antioxidant properties, and thought to be independent from their contribution to vitamin A generation.46 With respect to the current availability of milk from LI systems for consumers, it should be emphasized that milk from organic producers is identifiable and widely available, while milk from the non-organically certified LI farms is currently mixed with milk from HI conventional systems in the supply chain and is not available to consumers. Given the apparently high nutritional quality of milk produced in NO-LI-systems it is important that this practice is reviewed in order to take advantage of the price premiums that can currently be achieved by ‘nutritionally enhanced’ food products.47 When data for all sampling dates were pooled, the concentration of α-LA was elevated by 60% and that of CLA9 by 64% in the organic compared to HI milk (α-LA; mean = 9.4, SE = 0.3 versus mean = 5.7, SE = 0.3 g kg−1 fat, P < 0.001 and CLA9; mean = 12.2, SE = 0.7 versus mean = 7.5, SE 0.4 g kg−1 fat, P < 0.001 for O-LI and HI milk, respectively). These data may help explain why consumption of organic dairy produce has been shown to have a significant impact on the CLA content of breast milk in lactating woman48 and on the eczema risk during the first 2 years of life.49 It is now important to (a) identify exactly those production system components in organic, LI and conventional farming systems that are responsible for differences in milk composition and (b) to allow agronomic strategies in dairy production 1440 to be optimized further with respect to compounds that can be linked to positive health impacts. ACKNOWLEDGEMENTS The authors gratefully acknowledge financial support from the European Community under the 6th framework programme Integrated Project QualityLowInputFood, FP6-FOOD-CT-2003-506358 and the UK Red Meat Industry Forum (RMIF). The help and advice of the Grasshoppers dairy producers group, Acorn Dairies and all producers taking part in the study are also gratefully acknowledged. REFERENCES 1 Jones EL, Shingfield KJ, Kohen C, Jones AK, Lupoli B, Grandison AS, et al, Chemical, physical, and sensory properties of dairy products enriched with conjugated linoleic acid. J Dairy Sci 88:2923–2937 (2005). 2 Kristensen D, Hedegaard RV, Nielsen JH and Skibsted LH, Oxidative stability of buttermilk as influenced by the fatty acid composition of cows’ milk manipulated by diet. J Dairy Res 71:46–50 (2004). 3 Thorsdottir I, Hill J and Ramel A, Short communication: seasonal variation in cis-9, trans-11 conjugated linoleic acid content in milk fat from Nordic countries. J Dairy Sci 87:2800–2802 (2004). 4 Givens DI, The role of animal nutrition in improving the nutritive value of animal-derived foods in relation to chronic disease. Proc Nutr Soc 64:395–402 (2005). 5 Havemose MS, Weisbjerg MR, Bredie WLP and Neilsen JH, Oxidative stability of milk influenced by fatty acids, antioxidants, and copper derived from feed. J Dairy Sci 89:1970–1978 (2006). 6 Chen S, Bobe G, Zimmerman S, Hammond EG, Luhman CM, Boylston TD, et al, Physical and sensory properties of dairy products from cows with various milk fatty acid compositions. J Agric Food Chem 52:3422–3428 (2004). 7 Gonzalez S, Duncan SE, O’Keefe SF, Sumner SS and Herbein JH, Oxidation and textural characteristics of butter and ice cream with modified fatty acid profiles. J Dairy Sci 86:70–77 (2003). 8 Lynch JM, Lock AL, Dwyer DA, Noorbakhsh R, Barbano DM and Bauman DE, Flavour and stability of pasteurized milk with elevated levels of conjugated linoleic acid and vaccenic acid. J Dairy Sci 88:489–498 (2005). 9 Parodi PW, Conjugated linoleic acid in food, in Advances in Conjugated Linoleic Acid Research (2nd edn), ed. by Christie WW, Sébédio JL and Adlof RO. AOCS Press, Champaign, IL, pp. 101–122 (2003). 10 Connor WE, Importance of n-3 fatty acids in health and diseases. Am J Clin Nutr 7:171S–175S (2000). 11 Pariza MW, The biological activities of conjugated linoleic acid, in Advances in Conjugated Linoleic Acid Research (2nd edn), ed. by Christie WW, Sébédio JL and Adlof RO. AOCS Press, Champaign, IL, pp. 12–20 (2003). 12 Lock AL and Bauman DE, Modifying milk fat composition of dairy cows to enhance fatty acids beneficial to human health. Lipids 39:1197–1206 (2004). 13 Wahle KWJ, Heys SD and Rotondo D, Conjugated linoleic acids: are they beneficial or detrimental to health? Rec Prog Lipid Res 43:553–587 (2004). 14 Bergamo P, Fedele E, Iannibeli L and Marzillo G, Fat soluble vitamin contents and fatty acid composition in organic and conventional Italian dairy products. Food Chem 82:625–631 (2003). 15 Ellis KA, Innocent D, Grove-White D, Cripps P and McLean WG, Comparing the fatty acid composition of organic and conventional milk. J Dairy Sci 89:1938–1950 (2006). J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa Fatty acid and fat-soluble antioxidant concentrations in milk 16 Jahreis G, Fritsche J and Steinhart H, Conjugated linoleic acid in milk fat: high variation depending on production system. Nutr Res 17:1479–1484 (1997). 17 Chilliard Y, Glasser F, Ferlay A, Rouel J and Doreau M, Diet, rumen biohydrogenation and nutritional quality of cow and goat milk fat. Eur J Lipid Sci Technol 109:828–855 (2007). 18 Jensen RG, The composition of bovine lipids: January 1995 to December 2000. J Dairy Sci 85:295–350 (2002). 19 Walker GP, Dunshea FR and Doyle PT, Effects of nutrition and management on the production and composition of milk fat and protein: a review. Aust J Agric Res 55:1009–1028 (2004). 20 Dewhurst RJ, Targets for milk fat research: nutrient, nuisance or nutraceutial? J Agric Sci 143:359–367 (2005). 21 Jensen SK, Johannsen AKB and Hermansen JE, Quantitative secretion and maximal secretion capacity of retinol, βcarotene and α-tocopherol into cows’ milk. J Dairy Res 66:511–522 (1999). 22 Meglia GE, Jensen SK, Lauridsen C and Waller KP, αTocopherol concentration and stereoisomer composition in plasma and milk from dairy cows fed natural and synthetic vitamin E around calving J Dairy Res 7:227–234 (2006). 23 Havemose MS, Weisbjerg MR, Bredie WLP and Neilsen JH, Influence of feeding different types of roughage on the oxidative stability of milk. Int Dairy J 14:563–570 (2004). 24 Schneider C, Review: chemistry and biology of vitamin E. Mol Nutr Food Res 49:7–30 (2005). 25 Soil Association Certification, Organic Standards, No. 10.13.34. Soil Association Certification, Bristol, UK (2005). 26 McCall DG and Clark DA, Optimized dairy grazing systems in the northeast United States and New Zealand. II. System analysis. J Dairy Sci 82:1808–1816 (1998). 27 Dewhurst RJ, Shingfield KJ, Lee MRF and Scollan ND, Increasing the concentrations of beneficial polyunsaturated fatty acids in milk produced by dairy cows in high-forage systems. Anim Feed Sci Technol 131:168–206 (2006). 28 Toledo P and Andren A, Content of beta carotene in organic milk. J Food Agric Environ 1:122–125 (2003). 29 Neilsen JH, Lund-Neilsen T and Skibsted L, Higher antioxidant content in organic milk than in conventional milk due to feeding strategy, in Newsletter from Danish Research Centre for Organic Farming. 2004:3. [Online]. Available: http://www.darcof.dk/enews/sep04/milk.html [July 2007]. 30 Oklahoma State University, Breeds of Livestock (2007). [Online]. Available: www.ansi.okstate.edu/breeds/cattle/ [27 March 2008]. 31 Pinheiro JC and Bates DM, Mixed-Effects Models in S and S-Plus. Springer, New York (2000). 32 R Development Core Team, A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria (2005). 33 Shingfield KJ, Reynolds CK, Lupoli B, Tolvonen V, Yurawecz MP, Delmonte P, et al, Effect of forage type and proportion of concentrate in the diet on milk fatty acid composition in cows given sunflower oil and fish oil. Anim Sci 80:225–238 (2005). J Sci Food Agric 88:1431–1441 (2008) DOI: 10.1002/jsfa 34 Offer NW, Effects of cutting and ensiling grass on levels of CLA in bovine milk, in Proceedings of the 13th International Silage Conference, ed. by Gechie LM and Thomas C. Scottish Agricultural College, Auchincruive, Ayr, UK, pp. 16–17 (2002). 35 Dhiman TR, Nam S and Ure AL, Factors affecting conjugated linoleic acid content in milk and meat. Crit Rev Food Sci Nutr 45:463–482 (2005). 36 Collomb M, Schmid A, Sieber R, Wechsler D and Ryhänen EL, Conjugated linoleic acids in milk fat: variation and physiological effects. Int Dairy J 16:1347–1361 (2006). 37 Collomb M, Sollberger H, Bütikofer U, Sieber R, Stoll W and Schaeren W, Impact of a basal diet of hay and fodder beet supplemented with rapeseed, linseed and sunflower seed on the fatty acid composition of milk fat. Int Dairy J 12:549–559 (2004). 38 Shingfield KJ, Salo-Väänänen P, Pahkala E, Toivonen V, Jaakkola S, Piironen V, et al, Effect of forage conservation method, concentrate levels and propylene glycol on the fatty acid composition and vitamin content of cows’ milk. J Dairy Sci 72:349–361 (2005). 39 Frame J and Newbould P, Agronomy of white clover. Adv Agron 40:1–88 (1986). 40 Marriott CA, Seasonal variation in white clover content and nitrogen fixing (acetylene reducing) activity in a cut upland sward. Grass Forage Sci 43:253–262 (1988). 41 Sanderson P, Finnegan YE, Williams CM, Calder PC, Brudge GC, Wootton SA, et al, UK Food Standards Agency αlinolenic acid workshop report. Br J Nutr 88:573–579 (2002). 42 Burdge GC and Calder PC, α-Linolenic acid metabolism in adult humans: the effects of gender and age on conversion to longer-chain polyunsaturated fatty acids. Eur J Lipid Sci Technol 107:426–439 (2005). 43 Bendich A, Physiological role of antioxidants in the immune system. J Dairy Sci 76:2789–2794 (1993). 44 Kayden HJ and Traber MG, Absorption, lipoprotein transport and regulation of plasma concentrations of vitamin E in humans. J Lipid Res 34:343–358 (1993). 45 Willcox JK, Ash SL and Catignani GL, Antioxidants and prevention of chronic disease. Crit Rev Food Sci Nutr 44:275–295 (2004). 46 Spears JW, Micronutrients and immune function in cattle. Proc Nutr Soc 59:587–594 (2000). 47 Lampkin N, Measures M and Padel S, 2002/03 Farm Management Handbook, ed. by Lampkin N, Measures M and Padel S. Organic Farming Research Centre, Ceredigion, UK (2002). 48 Rist L, Muller A, Barthel C, Snijders B, Jansen M, SimõesWüst AP, et al, Influence of organic diet on the amount of conjugated linoleic acids in breast milk of lactating women in the Netherlands. Br J Nutr 97:735–743. 49 Kummeling I, Thijs C, Huber M, van de Vijver LPL, Snijders BEP, Penders J, et al, Consumption of organic foods and risk of atopic disease during the first 2 years of life in the Netherlands. Br J Nutr 99:598–605. 1441 J Sci Food Agric 88:1442–1447 (2008) Journal of the Science of Food and Agriculture Protective effect of polyphenol-rich extract prepared from Malaysian cocoa (Theobroma cacao) on glucose levels and lipid profiles in streptozotocin-induced diabetic rats Azli Mohd Mokhtar Ruzaidi,1,2 Maleyki Mhd Jalil Abbe,1 Ismail Amin,1∗ Abdul Ghani Nawalyah1 and Hamid Muhajir3 1 Department of Nutrition and Dietetics, Faculty of Medicine and Health Sciences, Universiti Putra Malaysia, 43400 Serdang, Selangor, Malaysia 2 Biotechnology Research Institute, Universiti Malaysia Sabah, 88999 Kota Kinabalu, Sabah, Malaysia 3 Department of Microbiology, Faculty of Biotechnology and Biomolecular Sciences, Universiti Putra Malaysia, 43400 Serdang, Selangor, Malaysia Abstract BACKGROUND: Cocoa beans are used for preparing cocoa liquor and cocoa powder, which are the main ingredients of cocoa-based products. Previous studies have reported the health benefits of cocoa polyphenols in reducing the risk of cardiovascular diseases. However, there is no report on the efficacy of cocoa polyphenols on diabetes mellitus. Therefore this study was designed to evaluate the protective effect of cocoa polyphenol-rich extract (CE) on glucose levels and lipid profiles in streptozotocin (STZ)-induced diabetic rats. Male SpragueDawley rats were divided into diabetic control, diabetic CE and diabetic glibenclamide groups. RESULTS: Three different dosages of CE (10, 20 and 30 mg per 100 g body weight) were administered orally once a day for 1 week before STZ injection and for 3 weeks thereafter. The results showed that CE could normalise the body weight loss caused by STZ. In the 20 mg CE-pretreated group there was a 143% increase in plasma glucose levels, compared with a 226% increase in diabetic control rats. CE could also normalise total cholesterol, triglycerides and high-density lipoprotein cholesterol at the end of the experiment compared with the baseline. CONCLUSION: The present study suggests that pretreatment with CE from roasted cocoa beans could prevent the development of diabetes induced by STZ injection in rats.  2008 Society of Chemical Industry Keywords: Theobroma cacao; cocoa extract; glucose levels; hypoglycaemic INTRODUCTION There are several ways of preventing diabetes and/or controlling its progression. Public and professional awareness of the risk factors and symptoms of diabetes is an important step towards its prevention and control. There is increasing demand by patients for natural products with antihyperglycaemic activity owing to the side effects associated with insulin and oral hypoglycaemic drugs.1,2 Therefore it has become necessary to look for an economical as well as a therapeutically effective use of natural products in prevention and treatment, especially in developing and underdeveloped countries. The search for safer and more effective compounds to protect β-cells from inflammatory destruction is still in progress. Several compounds such as metallothionein, nicotinamide and (−)-epicatechin have been reported to inhibit the diabetogenic action of streptozotocin (STZ) or alloxan in animal studies.3,4 Palm Vitee (palm oil vitamin E) has a protective action against the toxic inflammation caused by STZ.5 Cocoa beans have been reported to be a rich source of polyphenols, especially (−)-epicatechin. Some of the earliest studies established that the major flavanoids in cocoa beans were catechin, epicatechin, the dimers epicatechin-(4β → 8)-catechin (procyanidin B-l) and epicatechin-(4β → 8)-epicatechin (procyanidin B-2) and the trimer [epicatechin-(4β → 8)]2 epicatechin (procyanidin C-l).6,7 In a previous study we showed that a diet containing cocoa polyphenolrich extract reduced the glucose levels and lipid profiles in STZ-induced diabetic rats.8 In similar work, ∗ Correspondence to: Ismail Amin, Department of Nutrition and Dietetics, Faculty of Medicine and Health Sciences, Universiti Putra Malaysia, 43400 Serdang, Selangor, Malaysia E-mail: amin@medic.upm.edu.my (Received 3 July 2007; revised version received 21 January 2008; accepted 21 January 2008) Published online 17 April 2008; DOI: 10.1002/jsfa.3236  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Protective effect of polyphenol-rich cocoa extract in diabetic rats Osakabe et al.9 demonstrated that proanthocyanidins derived from cocoa inhibited diabetes-induced cataract formation. However, limited research has been conducted on the protective effect of cocoa polyphenol-rich extract (CE) against STZ diabetogenic action. Therefore the present study was focused on evaluating the protective action of CE against the destruction of insulin-producing β-cells of the pancreas in STZ-induced diabetic rats. MATERIALS AND METHODS Preparation of ethanolic extract Raw (fermented and dried) Malaysian cocoa beans were purchased from KL-Kepong Cocoa Products Sdn. Bhd. (Port Klang, Selangor, Malaysia). The beans were roasted in an air oven for 20 min at 140 ◦ C.10 After cooling to room temperature, the roasted beans were deshelled using a cocoa breaker (Limprimita, John Gordon & Co., Laneashire UK). The cotyledons were ground and defatted with petroleum ether (b.p. 40–60 ◦ C) in a Soxhlet apparatus. The defatted cotyledons were air dried to remove the solvent residue. The extract was prepared by treating the defatted powder with 80% (v/v) ethanol for 2 h. The ethanol residue was removed from the extract using a rotary evaporator (Rotavor R-200, Büchi, Flawil, Switzerland) for 20 min at 70 ◦ C and the resulting extract was lyophilised. This ethanolic extract was considered to be cocoa polyphenol-rich8 and was used for total phenolic determination and the animal study. Determination of total phenolics Total phenolic content was estimated according to the Folin–Ciocalteu assay.11 Briefly, CE was dissolved in 80% (v/v) ethanol and centrifuged (Rotofix 32, Hettich Zentrifugen, Tuttlingen, Germany) at 1000 × g for 15 min. Following centrifugation, 100 µL of the supernatant was mixed with 0.75 mL of Folin–Ciocalteu reagent (previously diluted 1:10 with distilled water) and allowed to stand at room temperature for 5 min. Sodium carbonate solution (0.75 mL) was then added to the mixture. After standing for a further 90 min at room temperature, the absorbance at 725 nm was recorded using a UV–visible spectrophotometer (Anthelie Advanced 5, Secomam, Ales, France). A standard calibration curve was constructed using 0.02–0.12 mg mL−1 (−)epicatechin (Sigma, St Louis, MO, USA). Results were expressed as mg epicatechin equivalents g−1 extract. Animal study Preparation of animals This study has been approved by the Animal Care and Use Committee of the Faculty of Medicine and Health Sciences, Universiti Putra Malaysia. Fifty male Sprague-Dawley rats (200–350 g initial weight) were purchased from Syarikat Usaha Cahaya Sdn. Bhd. J Sci Food Agric 88:1442–1447 (2008) DOI: 10.1002/jsfa (Batu Caves, Selangor, Malaysia). The rats were housed in individual plastic cages with stainless steel covers and kept at room temperature (24–28 ◦ C) under a 12/12 h dark/light cycle. Animals were allowed free access to their respective diets and water. All rats were allowed 7 days to adapt to the environment before being given the treatment. The experiment was conducted for 28 days. Body weights, food intakes and blood glucose levels were recorded weekly. The rats were divided into five groups, each consisting of ten rats (n = 10): • group 1 – diabetic rats administered normal (DC); • group 2 – diabetic rats administered 10 mg CE (DCE1); • group 3 – diabetic rats administered 20 mg CE (DCE2); • group 4 – diabetic rats administered 30 mg CE (DCE3); • group 5 – diabetic rats administered 100 mg glibenclamide (DG). saline mL−1 mL−1 mL−1 mL−1 CE (10, 20 and 30 mg mL−1 ) and glibenclamide (100 mg mL−1 ) were suspended in 0.9% (w/v) normal saline and given daily (1 mL per 100 g body weight) to the experimental rats by gastric intubation using a force-feeding needle. The animals were given CE once daily for 7 days before STZ injection and for 21 days thereafter. At day 7 the rats were given CE 1 h before STZ injection. Induction of diabetes STZ (Sigma) was used for inducing diabetes in the rats at day 7. After overnight fasting, the rats were injected intravenously with 45 mg kg−1 body weight of STZ dissolved in 0.05 mol L−1 citrate buffer (pH 4.5). Rats injected with the same volume of 0.05 mol L−1 citrate buffer served as the control group. Determination of glucose levels and lipid profiles At days 0, 9 and 28 of the experiment, 5 mL of blood was collected from the abdominal aorta of each animal, placed in a Vacutainer tube and centrifuged (Universal 32 , Hettich Zentrifugen) at 1000 × g for 10 min at room temperature. The supernatant was collected and kept at −20 ◦ C for further analysis. Plasma glucose levels and lipid profiles were measured using a chemistry analyser (Automatic Analyser 902, Hitachi, Tokyo, Japan). Statistical analysis Data were expressed as mean ± standard error of mean (SEM). One-way analysis of variance (ANOVA) was applied to determine differences between groups. Duncan’s multiple range test was used to find significant differences among means. Results were considered significantly different at the level of P < 0.05. 1443 Table 1. Effect of cocoa extract (CE) on body weight of rats Body weight (g)a Group Diabetic control (DC) Diabetic + 10 mg CE (DCE1) Diabetic + 20 mg CE (DCE2) Diabetic + 30 mg CE (DCE3) Diabetic + glibenclamide (DG) Initial Final 328.6 ± 13.7b 325.8 ± 13.4b 323.1 ± 13.3b 320.8 ± 11.8b 313.4 ± 19.3b 222.0 ± 38.0a 251.3 ± 26.3a 280.3 ± 29.2ab 251.0 ± 8.2a 224.0 ± 26.9a a Body weights were measured weekly. Values are expressed as mean ± SEM. Different letters indicate significant differences (P < 0.05). Body weight gain (g) 20 0 1 –20 –40 –60 –80 –100 –120 2 3 4 DC DCE1 DCE2 DCE3 DG 25 d d 20 bc b a a a a a DCE1 DCE2 Groups DCE3 5 0 week 0 week 1 DG week 4 Figure 2. Plasma glucose levels of diabetic rats pretreated with cocoa extract (CE): DC, diabetic control; DCE1, diabetic + 10 mg CE; DCE2, diabetic + 20 mg CE; DCE3, diabetic + 30 mg CE; DG, diabetic + glibenclamide. Values with different letters are significantly different (P < 0.05) between groups and weeks. levels compared with the DC group at the end of the experiment (day 28). The plasma total cholesterol levels in all groups were significantly higher (P < 0.05) at day 9 after STZ injection compared with day 0 (Fig. 3). At the end of the experiment (day 28) the total cholesterol levels in all treated animals were normalised. Figure 4 shows the effect of CE on plasma highdensity lipoprotein (HDL) cholesterol levels in STZinduced diabetic rats. After STZ injection, all rats exhibited a significant decrease (P < 0.05) in HDL cholesterol levels at day 9 compared with day 0. The reduction in HDL cholesterol levels was in the range 54–61%. Interestingly, HDL cholesterol levels were normalised in CE- and glybenclamide-pretreated rats at the end of the study. However, no significant change in plasma low-density lipoprotein (LDL) cholesterol levels was observed during the 4 week experimental period (Fig. 5). There were significant increases (P < 0.05) in plasma triglyceride levels in DC, DCE1 and DG rats after STZ injection (Fig. 6). However, in the DCE2 3.5 b b 3 b b b b 2.5 2 1.5 a a a a a a a a a 1 0.5 0 DCE1 DCE2 DCE3 DG Groups Weeks week 0 1444 cd bc bc 10 DC Figure 1. Pattern of body weight in control rats and those pretreated with cocoa extract (CE) and glibenclamide: DC, diabetic control; DCE1, diabetic + 10 mg CE; DCE2, diabetic + 20 mg CE; DCE3, diabetic + 30 mg CE; DG, diabetic + glibenclamide. The coefficient of variation for all data was less than 24%. Body weights were measured weekly. d bc cd 15 DC Plasma Total Cholesterol Level (mmol L–1) RESULTS The initial body weights of rats were in the range 200–350 g. The body weights of each group of rats were not significantly different before STZ injection (Table 1). At 21 days after STZ injection the body weights of DC, DCE1, DCE3 and DG rats were significantly decreased (P < 0.05) compared with their initial weights. However, there was no significant decrease in body weight in the DCE2 group. All rats exhibited a decrease in body weight gain after STZ injection at week 1 (Fig. 1). The body weights of DC and DG rats were drastically decreased at week 2. However, in the CE-pretreated groups (DCE1, DCE2 and DCE3) the body weight loss was much lower compared with the DC group at weeks 2, 3 and 4, though there was no significant difference. Figure 2 shows the protective effect of CE on plasma glucose levels in STZ-induced diabetic rats. At 2 days after STZ injection, i.e. at day 9, glucose levels increased significantly (P < 0.05) in all groups compared with the initial glucose levels. In the CEpretreated groups (DCE1, DCE2 and DCE3) and the glibenclamide-pretreated group (DG) the increase was significantly lower (P < 0.05) compared with the DC group at day 9. The increments in glucose levels in the DC, DCE1, DCE2, DCE3 and DG groups were 226, 163, 143, 156 and 148% respectively. There was no significant increase in glucose levels in treated rats at the end of the study (day 28) compared with day 9, except for the DCE3 group. Only the DCE2 group showed a significant decrease (P < 0.05) in glucose Plasma Glucose Level (mmol L–1) AMM Ruzaidi et al. week 1 week 4 Figure 3. Plasma total cholesterol levels of diabetic rats pretreated with cocoa extract (CE): DC, diabetic control; DCE1, diabetic + 10 mg CE; DCE2, diabetic + 20 mg CE; DCE3, diabetic + 30 mg CE; DG, diabetic + glibenclamide. Values with different letters are significantly different (P < 0.05) between groups and weeks. J Sci Food Agric 88:1442–1447 (2008) DOI: 10.1002/jsfa 1.2 c 1 c bc 0.6 a a a 0.4 c c bc bc ab 0.8 c c a a DCE2 Groups DCE3 0.2 0 DC DCE1 week 0 week 1 DG Plasma Triglyceride Level (mmol L–1) Plasma HDL-cholesterol Level (mmol L–1) Protective effect of polyphenol-rich cocoa extract in diabetic rats 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 e de de de abc abc DC a a DCE1 week 0 Plasma LDL-cholesterol Level (mmol L–1) 0.5 b b ab b ab ab ab ab 0.4 ab ab ab ab ab ab a 0.3 0.2 0.1 0 DC DCE1 week 0 DCE2 Groups week 1 DCE3 DG week 4 Figure 5. Plasma LDL cholesterol levels of diabetic rats pretreated with cocoa extract (CE): DC, diabetic control; DCE1, diabetic + 10 mg CE; DCE2, diabetic + 20 mg CE; DCE3, diabetic + 30 mg CE; DG, diabetic + glibenclamide. Values with different letters are significantly different (P < 0.05) between groups and weeks. and DCE3 groups the elevation was significantly lower (P < 0.05) compared with the DC group at day 9. At the end of the experiment a normalisation of plasma triglyceride levels was observed in CE- and glibenclamide-pretreated rats. DISCUSSION STZ is a specific β-cell toxin and can be used to chemically induce hyperglycaemia in rats and mice. It is taken up by pancreatic β-cells via a glucose transporter (GLUT2) and causes alkylation of deoxyribonucleic acid (DNA).12,13 DNA damage induces activation of poly adenosine diphosphate (ADP)-ribosylation, a process that is more important for the diabetogenicity of STZ than DNA damage itself.14 Poly ADP-ribosylation leads to the depletion of cellular nicotinamide adenine dinucleotide (NAD+ ) and adenosine triphosphate (ATP).15 Enhanced ATP dephosphorylation after STZ treatment supplies a substrate for xanthine oxidase (XOD), resulting in the formation of superoxide radicals. Consequently, J Sci Food Agric 88:1442–1447 (2008) DOI: 10.1002/jsfa abc a a DCE2 Groups DCE3 ab abc DG week 4 Figure 4. Plasma HDL cholesterol levels of diabetic rats pretreated with cocoa extract (CE): DC, diabetic control; DCE1, diabetic + 10 mg CE; DCE2, diabetic + 20 mg CE; DCE3, diabetic + 30 mg CE; DG, diabetic + glibenclamide. Values with different letters are significantly different (P < 0.05) between groups and weeks. 0.6 cd bcd week 1 week 4 Figure 6. Plasma triglyceride levels of diabetic rats pretreated with cocoa extract (CE): DC, diabetic control; DCE1, diabetic + 10 mg CE; DCE2, diabetic + 20 mg CE; DCE3, diabetic + 30 mg CE; DG, diabetic + glibenclamide. Values with different letters are significantly different (P < 0.05) between groups and weeks. hydrogen peroxide and hydroxyl radicals are also generated. Furthermore, STZ liberates toxic amounts of nitric oxide (NO), which inhibits aconitase activity and participates in DNA damage.15 As a result of the action of STZ, β-cells undergo destruction by necrosis. In this study the total polyphenol content of CE was in the range 190–286 mg epicatechin equivalents g−1 extract. A previous study showed that cocoa beans are rich in polyphenols such as (−)-epicatechin, (+)catechin, quercetin and procyanidin.16 To evaluate the protective effect of CE against STZ-induced diabetes in rats, CE (10, 20 and 30 mg mL−1 ) was force-fed daily to the rats for 1 week before STZ injection. On the last day of pretreatment (day 7), CE was given to the rats 1 h before STZ injection. This procedure was based on the findings of Baba et al.,17 which indicated that (−)-epicatechin metabolite occurred at its maximum level in plasma between 30 and 60 min after rats were given a cocoa beverage. Our results showed that CE administration significantly lowered (P < 0.05) the hyperglycaemic action of STZ in the DCE2 group. Treatment with CE also seemed to prevent body weight loss and improve body weight to some extent. Thus pretreatment with CE could be effective in preventing the development of hyperglycaemia following STZ injection. Kamtchouing et al.18 also reported that Anacardiaceae (Anacardium occidentale) extract showed a protective effect against the diabetogenic action of STZ. Moreover, Gupta et al.19 showed that neem seed extract had a protective effect on the heart and erythrocytes of diabetic rats. It is suggested that CE may have reacted with or scavenged STZ. Superoxide dismutase (SOD) is an enzyme known to be part of the antioxidant defence system of cells and a scavenger of free radicals. Vucic et al.20 reported that the activity of SOD is low in diabetes mellitus. CE may have acted by increasing the resistance of β-cells through activating SOD and scavenging free radicals caused by STZ. This scenario is supported by Sabu et al.,21 who found that green 1445 AMM Ruzaidi et al. tea polyphenols improved SOD levels in diabetic rats. The actual mechanisms of this pharmacological effect have yet to be determined. The selection of the dose of glibenclamide employed in this study was based on previous research by Nagappa et al.22 Glibenclamide is one of the most widely used orally active drugs (sulfonylureas) for the treatment of type 2 diabetes mellitus. The acute hypoglycaemic action of glibenclamide involves stimulation of insulin and inhibition of glucagon secretion.23 However, glibenclamide is only effective when there are still surviving β-cells in the pancreas. In the present study, glibenclamide tended to lower plasma glucose levels, which may be due to activation of pancreatic β-cells to secrete insulin after a single administration of glibenclamide. It could be suggested that the diabetic rats in this study still have some surviving β-cells in the pancreas, though not sufficient to significantly decrease the plasma glucose levels. Generally, diabetic models are also used to investigate successful treatments for hypercholesterolaemia.24 Abnormalities in lipids and lipoproteins play key roles in the development and progression of atherosclerotic vascular diseases in type 1 diabetes mellitus.25 The most common lipid abnormalities in diabetes mellitus are changes in plasma cholesterol and triglyceride levels, which certainly contribute to the development of cardiovascular diseases.26 Hypercholesterolaemia and hypertriglyceridemia have been reported to occur in STZ-induced diabetic rats in several studies.24,27,28 As to the protective role of CE against STZ action, rats pretreated with CE and glibenclamide did not seem to be protected from elevation of plasma cholesterol levels by STZ. However, after 3 weeks of further treatment with CE, total cholesterol levels decreased significantly (P < 0.05) and were normalised in diabetic rats. In contrast, elevation of triglyceride levels seemed to be significantly suppressed (P < 0.05) by pretreatment with 20 and 30 mg mL−1 CE in diabetic rats. Thus the present study indicates that CE exhibits protective effects on triglyceride levels in STZ-induced diabetic rats. This study suggests that polyphenols, the main components of CE, may be involved in the improvement of lipid profiles in diabetic rats. The main cause of elevated cholesterol and triglyceride levels in STZinduced diabetic rats is insulin deficiency. It is well known that, under normal circumstances, insulin activates the enzyme lipoprotein lipase (LpL), which then hydrolyses very-low-density lipoprotein (VLDL) cholesterol.29 However, in insulin-deficient diabetic rats, LpL is not activated, resulting in hypercholesterolaemia and hypertriglyceridemia. Our present study already suggests that the glucose-lowering effect of CE is due to the stimulation of insulin secretion in β-cells. This result is in agreement with a previous study which demonstrated that cocoa supplementation could increase postprandial insulin secretion and, to greater extent, improve insulin resistance.30,31 1446 Thus it is possible that the hypocholesterolaemic and hypotriglyceridemic effect of CE is also due to an increase in insulin secretion. Although LDL cholesterol levels did not seem to be affected by CE treatment, total cholesterol levels were normalised in diabetic rats after oral administration of CE. Therefore it can be suggested that CE might possess hypocholesterolaemic and hypotriglyceridemic activity in STZ-induced diabetic rats. The most common abnormalities in humans with poorly controlled type 1 or 2 diabetes are hypertriglyceridemia and low HDL cholesterol levels.32 HDL is the smallest, densest lipoprotein with the lowest amount of triglyceride. Lower total cholesterol and higher HDL cholesterol levels represent a very desirable biochemical state for prevention of atherosclerosis and ischaemic conditions.33 In its protective role against STZ action, CE pretreatment did not seem to prevent plasma HDL cholesterol levels from being reduced by STZ, but the levels were significantly enhanced (P < 0.05) after 3 weeks of further treatment with CE. CONCLUSIONS The underlying mechanisms responsible for the lack of a protective effect of CE on lipid profiles are not entirely understood and still to be determined. This study indicated that crude cocoa bean extract containing polyphenols and other components might not have a protective effect against hypercholesterolaemia, but it does exert a hypocholesterolaemic effect in STZinduced diabetic rats. ACKNOWLEDGEMENTS The authors would like to acknowledge the financial assistance provided by the Ministry of Science, Technology and Innovation of Malaysia (project IRPA 01-02-04-0013-EA001) and the laboratory facilities of Universiti Putra Malaysia. REFERENCES 1 Holman RR and Turner RC, Oral agents and insulin in the treatment of NIDDM, in Textbook of Diabetes, ed. by Pickup J and Williams G. Blackwell, Oxford, pp. 467–469 (1991). 2 Kameswara RB, Kesavulu MM, Giri R and Apparao C, Antidiabetic and hypolipidemic effects of Momardica cymbalaria Hook. fruit powder in alloxan diabetic rats. J Ethnopharmacol 67:103–109 (1999). 3 Bone AJ, Hii CST, Brown D, Smith W and Howell SL, Assessment of the antidiabetic activity of epicatechin in streptozotocin-diabetic and spontaneously diabetic BB/E rats. Biosci Rep 5:215–221 (1985). 4 Yang J and Cherian MG, Protective effects of metallothionein on streptozotocin-induced diabetes in rats. Life Sci 55:43–51 (1994). 5 Wan Nazaimoon WM and Khalid BAK, Palm vitamin E reduced serum levels of glycated hemoglobin, advanced glycosylation end-products and malondialdehyde of STZinduced diabetic rats. Diabetes Res Clin Prac 50:S357 (2000). 6 Osakabe N, Yamagishi M, Sanbongi C, Natsume M, Takizawa T and Osawa T, The antioxidative substances in cacao liquor. Int J Vitam Nutr Res 44:313–321 (1998). J Sci Food Agric 88:1442–1447 (2008) DOI: 10.1002/jsfa Protective effect of polyphenol-rich cocoa extract in diabetic rats 7 Porter LJ, Ma Z and Chan G, Cacao procyanidins: major flavonoids and identification of some minor metabolites. Phytochemistry 130:1657–1663 (1999). 8 Ruzaidi A, Amin I, Nawalyah AG, Hamid M and Faizul HA, The effect of Malaysian cocoa extract on glucose levels and lipid profiles in diabetic rats. J Ethnopharmacol 98:55–60 (2005). 9 Osakabe N, Yamagishi M, Natsume M, Yasuda A and Osawa T, Ingestion of proanthocyanidins derived from cacao inhibits diabetes-induced cataract formation in rats. Exp Biol Med 229:33–39 (2004). 10 Jinap S, Wan Rosli WI, Russly AR and Nordin LM, Effect of roasting time and temperature on volatile component profiles during nib roasting of cocoa bean. J Sci Food Agric 77:441–448 (1998). 11 Velioglu YS, Mazza G, Gao L and Oomah BD, Antioxidant activity and total phenolics in selected fruits, vegetables, and grain products. J Sci Food Agric 46:4113–4117 (1998). 12 Elsner M, Guldbakke B, Tiedge M, Munday R and Lenzen S, Relative importance of transport and alkylation for pancreatic beta-cell toxicity of streptozotocin. Diabetologia 43:1528–1533 (2000). 13 Thulesen J, Orskov C, Holst JJ and Poulsen SS, Short-term insulin treatment prevents the diabetogenic action of streptozotocin in rats. Endocrinology 138:62–68 (1997). 14 Sandler S and Swenne I, Streptozotocin, but not alloxan, induces DNA repair synthesis in mouse pancreatic islets in vitro. Diabetologia 25:444–447 (1983). 15 Pieper AA, Verma A, Zhang J and Snyder SH, Poly (ADPribose) polymerase, nitric oxide and cell death. Trends Pharmacol Sci 20:171–181 (1999). 16 Sanbongi C, Osakabe N, Natsume M, Takizawa T, Gomi S and Osawa T, Antioxidative polyphenols isolated from Theobroma cacao. J Agric Food Chem 46:452–457 (1998). 17 Baba S, Osakabe N, Natsume M, Yasuda A, Takizawa T, Nakamura T, et al, Cocoa powder enhances the level of antioxidative activity in rat plasma. Br J Nutr 84:673–680 (2000). 18 Kamtchouing P, Sokeng SD, Moundipa PF, Watcho P, Jatsa HB and Lontsi D, Protective role of Anacardium occidentale extract against streptozotocin-induced diabetes in rats. J Ethnopharmacol 62:95–99 (1998). 19 Gupta S, Kataria M, Gupta PK, Murganandan S and Yashroy RC, Protective role of extracts of neem seeds in diabetes caused by streptozotocin in rats. J Ethnopharmacol 90:185–189 (2004). 20 Vucic M, Gavell M, Bozikov V, Ashcroft JH and Rocic B, Superoxide dismutase activity in lymphocytes and polymorphonuclear cells of diabetic patients. Eur J Clin Chem Biochem 35:517–521 (1997). J Sci Food Agric 88:1442–1447 (2008) DOI: 10.1002/jsfa 21 Sabu MC, Smitha K and Kuttan R, Anti-diabetic activity of green tea polyphenols and their role in reducing oxidative stress in experimental diabetes. J Ethnopharmacol 83:109–116 (2002). 22 Nagappa AN, Thakurdesai PA, Venkat Rao N and Singh J, Antidiabetic activity of Terminalia catappa Linn fruits. J Ethnopharmacol 88:45–50 (2003). 23 Landstedt-Hallin L, Adamson U and Lins PE, Oral glibenclamide suppresses glucagon secretion during insulin-induced hypoglycaemia in patients with type 2 diabetes. J Clin Endocrinol Metab 84:3140–3145 (1999). 24 Ohara I, Tabuchi R, Onai K and Econ MH, Effects of modified rice bran on serum lipids and taste preference in streptozotocin-induced diabetic rats. Nutr Res 20:59–68 (2000). 25 Ginsberg HN, Lipoprotein physiology in nondiabetic and diabetic states. Relationship to atherogenesis. Diabetes Care 14:839–855 (1991). 26 Sachdewa A and Khemani LD, Effect of Hibiscus rosa sinensis Linn. ethanol flower extract on blood glucose and lipid profile in streptozotocin induced diabetes in rats. J Ethnopharmacol 89:61–66 (2003). 27 Cho SY, Park JY, Park EM, Choi MS, Lee MK, Jeon SM, et al, Alteration of hepatic antioxidant enzyme activities and lipid profile in streptozotocin-induced diabetic rats by supplementation of dandelion water extract. Clinical Chimica Acta 317:109–117 (2002). 28 Sharma SR, Dwivedi SK and Swarup D, Hypoglycaemic, antihyperglycaemic and hypolipidemic activities of Caesalpinia bonducella seeds in rats. J Ethnopharmacol 58:39–44 (1997). 29 Rensen PC and Van Berkel TJ, Apolipoprotein E effectively inhibits lipoprotein lipase-mediated lipolysis of chylomicronlike triglyceride-rich lipid emulsions in vitro and in vivo. J Biol Chem 271:14791–14799 (1996). 30 Brand-Miller J, Holt SHA, de Jong V and Petocz P, Cocoa powder increases postprandial insulinemia in lean young adults. J Nutr 133:3149–3152 (2003). 31 Grassi D, Necozione S, Lippi C, Croce G, Valeri L, Pasqualetti P, et al, Cocoa reduces blood pressure and insulin resistance and improves endothelium-dependent vasodilation in hypertensives. Hypertension 46:398–405 (2005). 32 Ginsberg HN, Diabetic dyslipidemia: basic mechanisms underlying the common hypertriglyceridemia and low HDL cholesterol levels. Diabetes 45:S27–S30 (1996). 33 Luc G and Fruchart JC, Oxidation of lipoproteins and atherosclerosis. Am J Clin Nutr 53:206S–209S (1991). 1447 J Sci Food Agric 88:1448–1454 (2008) Journal of the Science of Food and Agriculture Nutritional and sensory qualities of raw meat and cooked brine-injected turkey breast as affected by dietary enrichment with docosahexaenoic acid (DHA) and vitamin E Carmen Sárraga,∗ M Dolors Guàrdia, Isabel Dı́az, Luis Guerrero and Jacint Arnau IRTA, Food Technology, Finca Camps i Armet, E-17121 Monells, Girona, Spain Abstract BACKGROUND: This aim of this study was to evaluate the effects of feeding turkeys with docosahexaenoic acid (DHA) and vitamin E on the fatty acid profile, proteolytic enzyme activities and oxidative status of raw breast meat and cooked brine-injected breast meat. Four treatments were investigated: T1, basal diet (control); T2, basal diet plus 15 g kg−1 DHA; T3, basal diet plus 100 mg kg−1 vitamin E; T4, basal diet plus 5.4 g kg−1 DHA plus 100 mg kg−1 vitamin E. A sensory analysis of cooked brine-injected breasts was conducted in order to assess the sensory characteristics of these products and relate them to the expected nutritional benefits. RESULTS: Among the four treatments tested, no differences were observed in enzyme activities. No activities of cathepsin B, cathepsins B + L and catalase were detected in cooked brine-injected breast meat. Glutathione peroxidase activity was reduced and superoxide dismutase activity was similar to that measured in raw meat. The diets supplemented with DHA increased eicosapentaenoic acid and DHA levels in comparison with the control in both raw and cooked products. The increase in n-3 polyunsaturated fatty acids (PUFAs) led to a reduction in n-6/n-3 PUFA ratio to values between 1.01 ± 0.11 and 2.94 ± 0.47. In cooked brine-injected breast, treatment with 15 g kg−1 DHA (T2) induced a considerable fishy flavour, while treatment with 5.4 g kg−1 DHA plus 100 mg kg−1 vitamin E (T4) induced a slight fishy flavour. However, fishy odour in T4 did not differ significantly from that of the control. CONCLUSION: By feeding turkeys with 5.4 g kg−1 pure DHA plus 100 mg kg−1 vitamin E, the nutritional quality is improved through the introduction of a natural antioxidant and the reduction in n-6/n-3 PUFA ratio. With this treatment the sensory characteristics were similar to those of control samples, except for the fishy flavour, which could probably be masked by modifying the technological process.  2008 Society of Chemical Industry Keywords: DHA; vitamin E; nutritional quality; enzyme activities; sensory quality; turkey meat INTRODUCTION Enrichment of diets with long-chain n-3 polyunsaturated fatty acids (PUFAs), specifically eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), gives potential benefits to consumers by reducing the risk of a number of cardiovascular diseases1,2 and certain neuropsychiatric conditions.3 The n-6/n-3 PUFA ratio should be reduced to below 4 for a healthy human diet according to recent studies.4 – 6 Increasing the intake of fish and derived products can achieve this recommended n-6/n-3 PUFA ratio. However, fish is not always available to or appreciated by many consumers. In modern dietary trends, consumption of turkey meat is indicated as an adequate source of essential fatty acids with a low content of total lipids. However, the n-3 content can be considered too low to achieve a healthy n-6/n-3 PUFA ratio when birds are fed on standard diets. Feeding monogastric animals with n-3 PUFA-supplemented diets has been recommended as a feasible way to achieve this nutritional objective.7,8 Vegetable sources of n-3 PUFAs seem to be less efficient than marine sources in terms of the modification of fat composition,9,10 although too high dietary levels can produce some negative sensory characteristics.11 Poultry meat is prone to oxidation owing to its inherent relatively high content of PUFAs, so an enhancement in n-3 content could lead to a reduction in shelf life and sensory quality.12 – 14 Studies on the supplementation of diets with different natural antioxidants such as β-carotene, ascorbic acid, rosemary extracts and oregano15 – 18 have been carried out with varying degrees of success, and dietary vitamin E has demonstrated a high efficiency in preventing oxidative deterioration and reducing the development of off-flavours in poultry meat.19,20 ∗ Correspondence to: Carmen Sárraga, IRTA, Food Technology, Finca Camps i Armet, E-17121 Monells, Girona, Spain E-mail: carmen.sarraga@irta.es (Received 20 April 2007; revised version received 18 January 2008; accepted 15 February 2008) Published online 17 April 2008; DOI: 10.1002/jsfa.3238  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Effect of DHA and vitamin E on turkey nutritional and sensory qualities The oxidative status of meat can be estimated by the activity of the enzymes glutathione peroxidase (GSHPx), catalase (CAT) and superoxide dismutase (SOD),21,22 among others. Moreover, cathepsins B and L are considered to be the enzymes that are most involved in texture development during the processing of some meat products.23 As cathepsins B and L are cystein proteinases, their activity can be affected by the oxidative status of meat.24 Despite the numerous studies made on dietary supplementation of poultry meat with DHA and vitamin E, there has been hardly any research on cooked brine-injected turkey products. Therefore the objective of the present study was to evaluate the proteolytic activity, antioxidant status and fatty acid profile of raw breast meat and cooked brine-injected breast meat from turkeys fed on different DHA- and α-tocopheryl-enriched diets. In addition, a sensory quantitative descriptive analysis of the cooked brineinjected product was carried out to evaluate the commercial potential of the different diets. MATERIALS AND METHODS Birds and dietary treatments One-day-old female turkeys of the BUT 9 strain were used. They were housed in flat-deck batteries (6 cm2 ) in a flat-deck battery room. The birds were raised according to routine practice in terms of light and temperature and allowed ad libitum access to water and feed. They were fed on three different basal diets which contained lard as saturated fat and 20 mg kg−1 vitamin E (Table 1) throughout the raising period, which lasted 12 weeks. The birds were distributed randomly in four experimental treatments. The test period covered the fattening time during the last 4 weeks before slaughter. • Treatment 1 (T1): control birds fed on the basal diets. • Treatment 2 (T2): birds fed on the basal diets with 50% of the lard replaced by 15 g kg−1 pure DHA from tuna oil (Algatrium , Brudy SL, Barcelona, Spain). • Treatment 3 (T3): birds fed on the basal diets plus 100 mg kg−1 α-tocopheryl acetate. • Treatment 4 (T4): birds fed on the basal diets with 50% of the lard replaced by 0.54 g kg−1 pure DHA from Algatrium plus 100 mg kg−1 α-tocopheryl acetate. Seven birds from each treatment were slaughtered at a commercial processing plant. The breast (Pectoralis major) of each bird was removed and divided into two sections. One half was vacuum packed in an aluminium foil bag and stored at −20 ◦ C pending raw meat analysis. The other half was used to prepare a cooked product. J Sci Food Agric 88:1448–1454 (2008) DOI: 10.1002/jsfa Table 1. Composition of basal diets Ingredient (g kg−1 ) Wheat Lard Soybean meal 48% Extruded soybean DL-Methionine L-Lysine Calcium carbonate Dicalcium phosphate Salt Choline chloride 50% Monensine (mg kg−1 ) Minerals and vitaminsa MEa (MJ kg−1 ) Gross protein Gross fibre 0–4 weeks 5–8 weeks 9–12 weeks 418.34 34.09 473.63 20 2.27 0.37 19.50 23.64 3.74 0.42 0.09 4 11.7 280 31 484.63 39.78 416.18 20 1.69 0.95 11.09 18.46 3.18 0.04 0.09 4 12.1 260 31 564.77 60 328.16 – 1.16 1.9 22.63 15.01 2.38 – – 4 12.55 220 28 a A 1 kg portion of feed contained the following: vitamin A, 12 000 IU; vitamin D3 , 2400 IU; vitamin E, 20 IU; vitamin K3 , 3 mg; thiamine, 2.2 mg; riboflavin, 8 mg; pyridoxine, 5 mg; vitamin B12 , 11 µg; folic acid, 1.5 mg; biotin, 150 µg; calcium pantothenate, 25 mg; nicotinic acid, 65 mg; Mn, 60 mg; Zn, 40 mg; I, 0.33 mg; Fe, 80 mg; Cu, 8 mg; Se, 0.15 mg; ethoxyquin, 150 mg. Metabolisable energy. Preparation of cooked brine-injected breast product Breast muscles were injected with 180 g kg−1 brine consisting of sodium chloride and pentasodium tripolyphosphate, then tumbled at 7 rpm under vacuum for 45 min at 2–5 ◦ C, followed by a resting period of 18 h at 2–5 ◦ C. Afterwards, they were vacuum packed in plastic casings (CN330, Sealed Air, Passirana di Rho, Italy) and closed at both ends. The cooking process was carried out in a steam oven at 70 ◦ C until the internal temperature of the product reached 68 ◦ C. Extraction and activity of lysosomal proteolytic enzymes Lysosomal enzymes were extracted according to the method of Etherington et al.25 A portion of ground muscle was homogenised (1:4 w/v) in 50 mmol L−1 sodium acetate buffer (pH 5) containing 1mmol L−1 ethylene diamine tetraacetic acid (EDTA) and 2 mg mL−1 Triton X-100. The extract was stirred for 1 h at 4 ◦ C and then centrifuged (10 000 × g). The supernatant was filtered to remove debris and used as the source of enzymes. Cystein proteinases B and L were determined fluorimetrically using the method of Etherington and Wardale.26 Cathepsins B and L were assayed with the common substrate N-CBZ-Lphenylalanyl-L-arginine 7-amido-4-methylcoumarin (Z-Phe-Arg-NHMec; Sigma, Madrid, Spain). Cathepsin B was measured with N-CBZ-L-arginyl-L-arginine 7-amido-4-methylcoumarin (Z-Arg-Arg-NHMec; Sigma). One unit of activity was defined as the amount of enzyme hydrolysing 1 nmol substrate min−1 g−1 tissue at 37 ◦ C. 1449 C Sárraga et al. Determination of antioxidant enzyme activities The enzymatic extract was obtained according to the procedure of DeVore and Greene.27 A portion of tissue was homogenised (1:5 w/v) in 50 mmol L−1 Tris-HCl buffer (pH 7). The homogenate was centrifuged at 17 000 × g for 30 min at 4 ◦ C. The supernatant was recovered and filtered through deactivated glass wool. Subsequent centrifugation of the filtrate at 100 000 × g for 1 h at 4 ◦ C gave the final supernatant extract used for GSHPx, SOD and CAT activity assays. GSHPx activity was assessed according to DeVore and Greene.27 The assay medium included 0.5 units of glutathione reductase, 1 mmol L−1 reduced glutathione, 0.15 mmol L−1 nicotinamide adenine dinucleotide phosphate (NADPH) and 0.12 mmol L−1 hydrogen peroxide (H2 O2 ). The decrease in absorbance at 340 nm was recorded for 5 min and the activity was expressed as nmol NADPH oxidised min−1 g−1 meat. SOD activity was measured according to the method of Marklund and Marklund,28 based on the ability of SOD to inhibit the autoxidation of pyrogallol. The rate of autoxidation was determined by the decrease in absorbance measured at 420 nm. Enzyme activity was calculated according to a SOD standard curve (0–200 ng). One unit of activity was defined as the amount of enzyme inhibiting the autoxidation of pyrogallol by 50%. The activity was expressed as units SOD g−1 meat. CAT activity was determined by measuring the decrease in H2 O2 at 240 nm over 5 min.29 One unit of activity was defined as the amount of enzyme decomposing 1 µmol H2 O2 min−1 . The activity was expressed as µmol H2 O2 decomposed min−1 g−1 meat. Determination of thiobarbituric acid-reactive substances (TBARS) The assay method was based on the procedure of Botsoglou et al.,30 modified according to the current assay conditions. A 1 g sample was homogenised in 20 mL of ultrapure water. After the addition of 5 mL of 0.25 g mL−1 trichloroacetic acid (TCA), the homogenate was left to stand at 4 ◦ C for 15 min and then centrifuged at 12 000 × g for 15 min at 4 ◦ C. The supernatant was filtered and 3.5 mL of the filtrate was added to 1.5 mL of 6 mg mL−1 thiobarbituric acid (TBA). The mixture was incubated at 70 ◦ C for 30 min, chilled and the absorbance at 532 nm was recorded. The results, expressed as µg malondialdehyde (MDA) g−1 tissue, were calculated according to a standard curve (0–2.5 µg MDA) of hydrolysed 1,1,3,3-tetraethoxypropane (TEP). Determination of α-tocopherol levels A portion of muscle was sonicated in n-hexane/2propanol (3:2 v/v) to extract α-tocopherol. The sample was centrifuged and the supernatant was evaporated to dryness in a stream of nitrogen. The residue was redissolved in 1 mL of n-hexane/ethyl acetate (80:20 v/v). A 20 µL aliquot of the filtered extract was injected 1450 into the high-performance liquid chromatography (HPLC) system. Samples and standards (5 and 10 mg L−1 αtocopheryl acetate in mobile phase) were analysed by normal phase HPLC using an LKB Bromma system (Stockholm, Sweden) with an aminopropylsilica NH2 -NP column (5 µm, 250 mm × 4.6 mm i.d.; (Supelco-Sigma, Bellefonte PA, USA). The mobile phase consisted of n-hexane/ethyl acetate (80:20 v/v) at a flow rate of 1.2 mL min−1 . Detection was by fluorescence measurement at an excitation wavelength of 290 nm and an emission wavelength of 330 nm.31 Fatty acid profile Fatty acids were analysed according to Mach et al.32 Briefly, 1 mg of the internal standard tripentadecanoic acid (Sigma-Aldrich, Madrid, Spain) was added to 2 g of sample and homogenised in 100 mL of chloroform/methanol (2:1 v/v) for 24 h. The mixture was subsequently filtered through a separating funnel, mixed with 0.1 g mL−1 NaCl and extracted twice, after which the solvent was evaporated. Fatty acid methyl esters (FAMEs) were obtained by the ISO method33 and analysed using an HP 5890 Series II gas chromatograph (GC; Hewlett Packard SA, Barcelona, Spain). Samples were introduced by split injection into a BPX70 fused silica capillary column (30 m × 0.25 mm i.d., 0.25 µm film thickness; SGE, Milton Keynes, UK). Helium was the carrier gas at 30 cm s−1 . The GC temperature was held at 150 ◦ C for 1 min, then increased at 4 ◦ C min−1 to 200 ◦ C and held for 10 min. Individual fatty acids were identified by comparison of their retention times with those of standards (lipid standard: fatty acid methyl ester mixture 189-19, Sigma-Aldrich). Sensory analysis A quantitative descriptive analysis was conducted by a six-member trained panel.34,35 The panellists had a minimum experience of 10 years in descriptive analysis of a wide range of foods. The generation/selection of descriptors was carried out by open discussion in two previous sessions. Each panellist assessed the different descriptors using a rating scale where 0 means absence and 10 very high intensity of the descriptor. Sensory evaluations were conduced in a standardised sensory testing room36 equipped with ten individual sensory booths lit by red lights in order to mask the colour effect when odour and flavour attributes were evaluated. In each of the seven sessions performed, each assessor evaluated the same four products (one from each treatment), which were coded using random three-digit numbers. Slices of 2 mm from each sample were provided to the assessors in individual plastic containers covered with plastic film. The presentation order of samples and the first-order and carry-over effect were blocked.37 The descriptors assessed were: darkness (evaluated in individual booths using a standardised 8 W day light), turkey odour, fishy odour, metallic flavour, turkey J Sci Food Agric 88:1448–1454 (2008) DOI: 10.1002/jsfa J Sci Food Agric 88:1448–1454 (2008) DOI: 10.1002/jsfa Values are mean ± SD (n = 7 for each dietary treatment); ND, not detected. Means in the same row with different letters are significantly different (P < 0.05): lowercase letters indicate differences between treatments; uppercase letters indicate differences between raw and cooked breasts. Experimental treatments: T1, basal diet; T2, basal diet + 15 g kg−1 pure DHA; T3, basal diet + 100 mg kg−1 vitamin E; T4, basal diet + 5.4 g kg−1 pure DHA + 100 mg kg−1 vitamin E. Units: a nmol substrate min−1 g−1 meat; b µmol H2 O2 min−1 g−1 meat; c nmol NADPH min−1 g−1 meat; d units SOD g−1 meat; e µg MDA g−1 meat; f µg g−1 meat. ND ND ND 41 ± 8.7B 136 ± 12 2.42 ± 0.4bB 1.39 ± 0.2bB ND ND ND 40 ± 13B 134 ± 8.5 1.45 ± 0.2cB 1.89 ± 0.3aB ND ND ND 46.3 ± 13B 141 ± 14 4.68 ± 1.0aB 0.78 ± 0.1cB ND ND ND 39.5 ± 23B 141 ± 13 2.09 ± 0.6bcB 1.25 ± 0.3b 0.77 ± 0.2 2.79 ± 0.78 1.45 ± 1.6 205 ± 70A 154 ± 8.4 0.31 ± 0.1bA 1.85 ± 0.3bA 0.74 ± 0.23 2.45 ± 0.55 1.13 ± 0.7 227 ± 85A 153 ± 7.5 0.16 ± 0.06bA 2.31 ± 0.3aA 0.89 ± 0.16 2.65 ± 0.43 0.98 ± 0.5 199 ± 67A 161 ± 12 0.58 ± 0.1aA 1.11 ± 0.3cA 1.02 ± 0.15 2.11 ± 0.52 0.69 ± 0.4 200 ± 61A 155 ± 11 0.22 ± 0.1bA 1.29 ± 0.2c Cathepsin Cathepsins B + La CATb GSHPxc SODd TBARSe Vitamin Ef T3 T2 T4 T1 Cooked breast product T3 T2 T1 Ba Parameter RESULTS AND DISCUSSION The activities of the most representative cystein proteinases and antioxidant enzymes are shown in Table 2. No differences among treatments were found in any of the enzymes studied, neither in raw breast nor in cooked breast, although there were evident differences between raw and cooked breast samples. Cathepsin B and B + L and CAT activities were not detected and GSHPx activity was severely reduced in cooked breast samples. In agreement with previous results for cooked breast turkey22 and porcine cooked ham,39 SOD activity was not affected by the thermal process, it being the most stable among the enzymes studied. Results of vitamin E determination suggested an interactive effect between vitamin E and DHA, since vitamin E accumulation was significantly higher in T3 than in T4 in both raw and cooked meat, taking into account that the supplementation dose of 100 mg kg−1 was the same in both treatments. This finding could be explained by the rapid consumption of vitamin E in meat from the DHA-supplemented diet due to the high oxidative instability, which prevents antioxidant accumulation. Results from T1 and T2 showed significant differences in cooked breast (P < 0.05) but not in raw meat. Except for the control samples, cooked brine-injected breasts showed a lower level of vitamin E than raw meat because of the dilution effect of the brine and the effect of the cooking process on vitamin E degradation. TBARS values were closely related to the supplementation of the diets. As expected, significant differences were observed between raw and cooked breast. Cooked product presented the highest TBARS value in samples from birds fed with 15 g kg−1 DHA. The lowest TBARS level was found with the vitamin E-supplemented diet (T3). An intermediate level was found in samples of T4 where vitamin E and DHA had Raw breast meat Statistical analysis All data were analysed using the MEANS and GLM procedures of the SAS statistical package.38 One-way analysis of variance (ANOVA) was performed for each product (raw and cooked breasts) in order to assess differences between treatments. Then, for each treatment, an additional one-way ANOVA was carried out in order to assess differences between products. For sensory data the ANOVA was performed with the mean values from the six assessors for each session. In this case the two-way ANOVA model included treatment, session and their interaction as fixed effects. No significant interaction was observed between treatment and session. Mean comparisons were performed using the Tukey test. Table 2. Proteolytic (cathepsin B and L) and antioxidant (CAT, GSHPx and SOD) enzyme activities, TBARS and vitamin E levels in raw meat and cooked brine-injected breast from turkeys fed on different supplemented diets flavour, fishy flavour, salty flavour and rancid flavour. The products were evaluated 10 min after slicing, and the average assessors’ score for each sample was recorded. T4 Effect of DHA and vitamin E on turkey nutritional and sensory qualities 1451 Cooked breast product T1 T2 T3 T4 9.18 ± 2.52 8.40 ± 1.65 9.04 ± 3.30 8.67 ± 2.47 Values are mean ± SD (n = 7 for each dietary treatment). Experimental treatments: T1, basal diet; T2, basal diet + 15 g kg−1 pure DHA; T3, basal diet + 100 mg kg−1 vitamin E; T4, basal diet + 5.4 g kg−1 pure DHA + 100 mg kg−1 vitamin E. 1452 18.36 ± 0.75b 0.91 ± 0.08bA 5.31 ± 1.35abB 1.12 ± 0.34b 5.68 ± 1.97b 2.94 ± 0.47b 19.80 ± 1.13ab 0.90 ± 0.09b 6.25 ± 1.36a 0.30 ± 0.06c 1.12 ± 0.37c 11.77 ± 0.74a 15.85 ± 0.95c 0.92 ± 0.15bA 4.25 ± 0.81ab 2.75 ± 0.19a 13.90 ± 1.76a 1.15 ± 0.13c 20.80 ± 1.43a 1.23 ± 0.17aA 4.05 ± 1.51bB 0.28 ± 0.25c 1.30 ± 1.45c 12.12 ± 0.87a 18.33 ± 1.06b 0.75 ± 0.15B 6.91 ± 1.28aA 1.31 ± 0.14b 6.92 ± 1.61b 2.89 ± 0.54c 20.37 ± 0.69a 0.91 ± 0.10 6.86 ± 1.48a 0.37 ± 0.11c 1.58 ± 0.61c 10.75 ± 0.58b 14.67 ± 0.97c 0.73 ± 0.13B 4.77 ± 1.12b 3.05 ± 0.43a 15.63 ± 2.05a 1.01 ± 0.11d T3 T2 T2 T3 T4 T1 Cooked breast product Raw breast meat Values are mean ± SD (n = 7 for each dietary treatment). Means in the same row with different letters are significantly different (P < 0.05): lowercase letters indicate differences between treatments; uppercase letters indicate differences between raw and cooked breasts. Experimental treatments: T1, basal diet; T2, basal diet + 15 g kg−1 pure DHA; T3, basal diet + 100 mg kg−1 vitamin E; T4, basal diet + 5.4 g kg−1 pure DHA + 100 mg kg−1 vitamin E. 8.30 ± 2.86 7.89 ± 1.74 8.09 ± 1.61 8.82 ± 3.67 20.28 ± 1.62a 0.93 ± 0.18B 6.23 ± 1.51abA 0.45 ± 0.39c 1.27 ± 0.83c 11.74 ± 0.54a Raw breast meat T1 T2 T3 T4 Linoleic (18:2n-6) α-Linolenic (18:3n-3) Arachidonic (20:4n-6) EPA (20:5n-3) DHA (22:6n-3) n-6/n-3 ratio Fatty acid concentration (mg g−1 ) T1 Treatment Fatty acid (%) Table 3. Total fatty acid concentrations in raw meat and cooked brine-injected breast from turkeys fed on different supplemented diets Table 4. Percentages of n-6 and n-3 fatty acids in raw meat and cooked brine-injected breast from turkeys fed on different supplemented diets been added. The supplementation with 100 mg kg−1 vitamin E contributed to the decrease in TBARS levels when T2 was compared with T4 (P < 0.05) in both raw and cooked samples. The total fatty acid concentrations (mg g−1 ) found in samples were not affected by different treatments (Table 3). The profile of n-6 and n-3 fatty acids in raw and cooked turkey is shown in Table 4. Raw meat in the two DHA-supplemented treatments (T2 and T4) showed a significantly lower level of linoleic acid in comparison with T1 and T3. The lowest percentage of 18:2n-6 was detected in T2, while T4 showed an intermediate value. Similar tendencies were observed in cooked products. The diets supplemented with 15 g kg−1 (T2) and 5.4 g kg−1 (T4) pure DHA from a marine source increased the EPA and DHA levels in relation to the control level for both raw and cooked turkey. With the exception of α-linolenic acid, the n-3 PUFA percentages were maintained after the cooking process, because no differences were found between raw breast and cooked brine-injected product. The reduction in n-6 and large increase in n-3 PUFAs resulted in a significant reduction in n-6/n-3 PUFA ratio in T2 and T4 to values between 1.01 and 2.94 respectively (Table 4). No significant difference was observed between raw breast and cooked brineinjected breast. It is worth noting that the reduction in n-6/n-3 PUFA ratio was in accordance with the dietary DHA amount. Results from the sensory analysis of cooked breast products are shown in Table 5. Samples from birds fed on the vitamin E-supplemented diet (T3) showed a significant increase in darkness, as was reported by Ruiz et al.40 in raw broiler meat. Rancidity was very low, possibly because samples were consumed within 1 week after elaboration and 10 min after slicing. The batches enriched in DHA were slightly less rancid than the control group, maybe because of a masking effect due to their stronger fishy flavour. The main differences among dietary treatments were in turkey T4 C Sárraga et al. J Sci Food Agric 88:1448–1454 (2008) DOI: 10.1002/jsfa Effect of DHA and vitamin E on turkey nutritional and sensory qualities Table 5. Sensory analysis of cooked brine-injected breast product from turkeys fed on DHA- and vitamin E-supplemented diets Attributea T1 T2 T3 T4 Darkness 3.3 ± 1.0b 3.5 ± 1.3b 3.9 ± 1.1a Odour Turkey 2.7 ± 1.3ab 0.9 ± 0.9c 3.1 ± 1.7a Fishy 0.6 ± 1.3b 4.9 ± 2.3a 0.4 ± 1.1b 2.3 ± 1.5b 0.9 ± 1.3b Flavour Metallic Turkey Fishy Salty Rancid 1.6 ± 1.7 2.9 ± 2.1b 1.5 ± 1.5b 3.2 ± 1.7 0.1 ± 0.2b 1.6 ± 1.3 3.4 ± 2ab 0.3 ± 0.7c 3.2 ± 1.3 0.4 ± 0.6a 1.2 ± 1.1 0.5 ± 0.8c 7.1 ± 1.7a 3.8 ± 1.9 0.1 ± 0.1b 1.1 ± 0.8 3.9 ± 2.5a 0.2 ± 0.6c 3.1 ± 1.1 0.3 ± 0.6ab 3.2 ± 1.1b Values are mean ± SD (n = 7 for each dietary treatment). Means in the same row with different letters are significantly different (P < 0.05). Experimental treatments: T1, basal diet; T2, basal diet + 15 g kg−1 pure DHA; T3, basal diet + 100 mg kg−1 vitamin E; T4, basal diet + 5.4 g kg−1 pure DHA + 100 mg kg−1 vitamin E. a Assessed according to a scale from 0 (absence) to 10 (intense). and fishy flavours and odours. Turkey flavour and odour had the lowest values in T2 samples compared with the other treatments owing to the presence of fishy notes. Samples supplemented with vitamin E (T3) showed the highest score, which was only significantly different from T2 and T4. Fishy flavour followed the opposite pattern of turkey flavour. Samples from birds fed with 15 g kg−1 DHA presented the highest level, which was perceived as an intermediate value in T4 and nearly absent in T1 and T3. Fishy flavour was also significantly higher in T4 than in T1 and T3. Maruri and Larick41 suggested that the negative effects of PUFA-enriched diets on the sensory profile could be reduced by supplementation with vitamin E. Nevertheless, Trout42 reported that vitamin E is not so effective at minimising fishy flavours when fish oil is added to the animal diet and recommended a fish oil withdrawal period before slaughter. According to Rymer and Givens,8 feeding with marine algae can result in better sensory efficacy than feeding with fish oil to improve consumer acceptability. Thus the quality and composition of the DHA source seem to be basic parameters that influence off-flavour production. In addition to these suggestions, the offflavour perception of cooked brine-injected breast could be eliminated by modifying the technological process via the addition of smoke or other antioxidant compounds. CONCLUSIONS Feeding turkeys with 5.4 g kg−1 pure DHA plus 100 mg kg−1 vitamin E improves the nutritional quality of raw meat and cooked brine-injected breast owing to the increase in vitamin E and the reduction in n6/n-3 PUFA ratio. With this treatment the sensory characteristics of cooked brine-injected turkey breasts were similar to those of control samples, except for a slight fishy flavour. J Sci Food Agric 88:1448–1454 (2008) DOI: 10.1002/jsfa ACKNOWLEDGEMENTS Financial support for this work was provided by FEDER and INIA (Government of Spain) project CAL02-00. Algatrium was donated by Brudy SL (Barcelona, Spain). The authors wish to thank Narc ı́s Sa ı́s, Eugeni Anselmet and Quim Arbonés for their technical assistance. REFERENCES 1 Moreno JJ and Mitjavila MT, The degree of unsaturation of dietary fatty acids and the development of atherosclerosis. J Nutr Biochem 14:182–195 (2003). 2 Sanders TAB, Gleason K, Griffin B and Miller GJ, Influence of an algal triacylglycerol containing docosahexaenoic acid (22:6n-3) and docosapentaenoic acid (22:5n-6) on cardiovascular risk factors in healthy men and women. Br J Nutr 95:525–531 (2006). 3 Young G and Conquer J, Omega-3 fatty acids and neuropsychiatric disorders. Reprod Nutr Dev 45:1–28 (2005). 4 Mantzioris E, Cleland LG, Gibson RA, Neumann MA, Demasi M and James MJ, Biochemical effects of a diet containing foods enriched with n-3 fatty acids. Am J Clin Nutr 72:42–48 (2000). 5 Williams C, Dietary fatty acids and human health. Ann Zootechnol 49:165–180 (2000). 6 Harbige LS, Fatty acids, the immune response, and autoimmunity: a question of n-6 essentiality and the balance between n-6 and n-3. Lipids 38:323–341 (2003). 7 Raes K, DeSmet S and Demeyer D, Effect of dietary fatty acids on incorporation of long chain polyunsaturated fatty acids and conjugated linoleic acid in lamb, beef and pork meat: a review. Anim Feed Sci Technol 113:199–221 (2004). 8 Rymer C and Givens DI, n-3 Fatty acid enrichment of edible tissue of poultry: a review. Lipids 40:121–130 (2005). 9 Hulan HW, Ackman RG, Ratnayake WM and Proudfoot FG, Omega-3 fatty acid levels and general performance of commercial broilers fed practical levels of redfish meal. Poultry Sci 68:153–162 (1989). 10 Schreiner M, Hulan HW, Razzazi-Fazeli E, Böhm J and Moreira RG, Effect of different sources of dietary omega3 fatty acids on general performance and fatty acid profiles of thigh, breast, liver and portal blood of broilers. J Sci Food Agric 85:219–226 (2005). 11 López-Ferrer S, Baucells MD, Barroeta AC and Grashorn MA, n-3 Enrichment of chicken meat using fish oil: alternative substitution with rapeseed and linseed oils. Poultry Sci 78:356–365 (1999). 12 Mooney JW, Hirschler EM, Kennedy AK, Sams AR and van Elswyk ME, Lipid and flavour quality of stored breast meat from broilers fed marine algae. J Sci Food Agric 78:134–140 (1998). 13 Bou R, Guardiola F, Tres A, Barroeta AC and Codony R, Effect of dietary fish oil, α-tocopheryl acetate, and zinc supplementation on the composition and consumer acceptability of chicken meat. Poultry Sci 83:282–292 (2004). 14 Lee S, Faustman C, Djordjevic D, Faraji H and Decker EA, Effect of antioxidants on stabilization of meat products fortified with n-3 fatty acids. Meat Sci 72:18–24 (2006). 15 Ruiz JA, Pérez-Vendrell AM and Esteve-Garcı́a E, Effects of β-carotene and vitamin E on the oxidative stability in leg meat of broilers fed different supplemental fat. J Agric Food Chem 47:448–454 (1999). 16 Yu L, Scanlin L, Wilson J and Schmidt G, Rosemary extracts as inhibitors of lipid oxidation and colour change in cooked turkey products during refrigerated storage. Food Chem Toxicol 67:582–585 (2002). 17 Botsoglou NA, Govaris A, Botsoglou EN, Grigoropoulou SH and Papegeorgiou G, Antioxidant activity of dietary oregano essential oil and α-tocopherol acetate supplementation in 1453 C Sárraga et al. 18 19 20 21 22 23 24 25 26 27 28 long-term frozen stored turkey meat. J Agric Food Chem 51:2930–2936 (2003). Young JF, Stagsted J, Jensen SK, Karlsson AH and Henckel P, Ascorbic acid, α-tocopherol and oregano supplements reduce stress-induced deterioration of chicken meat quality. Poultry Sci 82:1343–1351 (2003). Carreras I, Guerrero L, Guàrdia MD, Esteve-Garcı́a E, Garcı́a Regueiro JA and Sárraga C, Vitamin E levels, thiobarbituric acid test and sensory evaluation of breast muscles from broilers fed α-tocopheryl acetate and β-carotene supplemented diets. J Sci Food Agric 84:313–317 (2004). Sárraga C, Carreras I, Garcı́a Regueiro JA, Guàrdia MD and Guerrero L, Effects of α-tocopheryl acetate and β-carotene dietary supplementation on the antioxidant enzymes, TBARS and sensory attributes of turkey meat. Br Poultry Sci 46:700–707 (2006). Maraschiello C, Sárraga C and Garcı́a Regueiro JA, Glutathione peroxidase activity, TBARS and α-tocopherol in meat from chickens fed different diets. J Agric Food Chem 47:867–872 (1999). Carreras I, Castellari C, Valero A, Garcı́a Regueiro JA and Sárraga C, Influence of enrofloxacin administration on the proteolytic and antioxidant enzyme activities of raw and cooked turkey products. J Sci Food Agric 85:2407–2412 (2005). Arnau J, Guerrero L and Sárraga C, The effect of green ham pH and NaCl concentration on cathepsin activities and the sensory characteristics of dry cured hams. J Sci Food Agric 77:387–392 (1998). Sárraga C and Garcı́a Regueiro JA, Membrane lipid oxidation and proteolytic activity in thigh muscles from broilers fed different diets. Meat Sci 52:213–219 (1999). Etherington DJ, Taylor MAJ, Wakefield DK, Cousins A and Dransfield E, Proteinase (cathepsin B, D, L and calpains) levels and conditioning rates in normal, electrically stimulated and high ultimate pH chicken muscle. Meat Sci 28:99–109 (1990). Etherington DJ and Wardale RJ, The mononuclear cell population in rat muscle: its contribution to the lysosomal enzyme activities of whole muscle extracts. J Cell Sci 58:139–148 (1982). DeVore VR and Greene BE, Glutathione peroxidase in postrigor bovine Semitendinosus muscle. J Food Sci 47:1406–1409 (1982). Marklund S and Marklund G, Involvement of the superoxide anion radical in the autoxidation of pyrogallol and a conventional assay for superoxide dismutase. Eur J Biochem 47:469–474 (1974). 1454 29 Mei L, Crum AD and Decker EA, Development of lipid and protein oxidation and incubation of antioxidant enzymes in cooked pork and beef. J Food Lipids 1:273–283 (1994). 30 Botsoglou NA, Fletouris DJ, Papageorgiou GE, Vassilopoulos VN, Mantis AJ and Trakatelis AG, Rapid, sensitive and specific thiobarbituric acid method for measuring lipid peroxidation in animal tissue, food and feedstuff samples. J Agric Food Chem 42:1931–1937 (1994). 31 Casademont G and Garcia Regueiro JA, Residuos en productos animales. Problemática de su control analı́tico. Eurocarne 75:45–58 (1999). 32 Mach N, Devant M, Bach A, Dı́az I, Font-Furnols M, Oliver MA, et al, Increasing the amount of omega-3 fatty acid of meat from young Holstein bulls through nutrition. J Anim Sci 84:3039–3048 (2006). 33 ISO, Animal and Vegetable Fats and Oils – Preparation of Methyl Esters of Fatty Acids. ISO 5509(E), pp. 1–6 (1978). 34 Stone H, Sidel J, Oliver S, Woolsey A and Singleton RC, Sensory evaluation by quantitative descriptive analysis. Food Technol 28:24–34 (1974). 35 Cross HR, Moen R and Stanfield MS, Training and testing judges for sensory analyses of meat quality. Food Technol 32:48–54 (1978). 36 ISO, Sensory Analysis – General Guidance for the Design of Test Rooms. ISO 8589, pp. 1–9 (1989). 37 Macfie HJ, Bratchell N, Greenhoff H and Vallis LV, Designs to balance the effect of order of presentation and first-order carry over effects in hall test. J Sens Stud 4:129–149 (1989). 38 SAS, SAS/STAT User’s Guide: Release 8.03. SAS Institute, Cary, NC (2000). 39 Sárraga C, Guàrdia MD, Dı́az I, Guerrero L, Garcı́a Regueiro JA and Arnau J, Nutritional and sensory quality of porcine raw meat, cooked ham and dry-cured shoulder as affected by dietary enrichment with docosahexaenoic acid (DHA) and α-tocopheryl acetate. Meat Sci 76:377–384 (2007). 40 Ruiz JA, Guerrero L, Arnau J, Guàrdia MD and EsteveGarcı́a E, Descriptive sensory analysis of meat from broilers fed diets containing vitamin E or β-carotene as antioxidants and different supplemental fats. Poultry Sci 80:976–982 (2001). 41 Maruri JL and Larick DK, Volatile concentration and flavour of beef as influenced by diet. J Food Sci 57:1275–1281 (1992). 42 Trout GR, Biochemistry of lipid and myoglobin oxidation in post-mortem muscle and processed meat products – effects on rancidity. Proc. 49th ICoMST, pp. 50–55 (2003). J Sci Food Agric 88:1448–1454 (2008) DOI: 10.1002/jsfa Journal of the Science of Food and Agriculture J Sci Food Agric 88:1455–1463 (2008) Microbiological hazards involved in fresh-cut lettuce processing Adriano G da Cruz,1∗ Sergio A Cenci2 and Maria Cristina A Maia3 1 Programa de Pós-Graduação em Ciência de Alimentos, Instituto de Quı́mica, Universidade Federal do Rio de Janeiro, CEP 21949-900, Rio de Janeiro, Brazil 2 Embrapa Agroindústria de Alimentos, Av. das Américas, CEP 23020-470, 29501 Guaratiba, Rio de Janeiro, Brazil 3 Departamento de Engenharia Bioquı́mica, Escola de Quı́mica, Universidade Federal do Rio de Janeiro, CEP: 21949-900, Rio de Janeiro, Brazil Abstract BACKGROUND: The increasing consumption of produce all over the world has resulted in increasing concern by the regulatory agencies with respect to the level of safety performed by the processors. The objective of the present study was to investigate the hazards involved in the various steps of fresh-cut lettuce processing (reception/selection of raw material, washing, rinsing, sanitisation and final product) by means of microbiological analyses of microbial groups used as indicators of hygienic conditions and of pathogens. RESULTS: High microbial loads of mesophilic and psychrotrophic bacteria and Pseudomonas spp. were found in the ram reception (∼6 log colony-forming units (CFU) g−1 ), which were reduced by a single logarithmic cycle for the last two microbial groups after the sanitisation step (P < 0.05), the latter being ineffective against the first microbial group (P > 0.05). Lower counts of yeasts and moulds, total coliforms (35 ◦ C) and faecal coliforms (44 ◦ C) were observed in the initial step (3.49–4.53 log CFU g−1 , 0.65–1.55 log most probable number (MPN) g−1 and 0.50–0.90 log MPN g−1 respectively), these values increasing significantly after the sanitisation step for yeasts and moulds (∼5 log CFU g−1 ) but remaining unaltered for coliforms (P > 0.05). Salmonella spp. were not found in any of the experiments carried out, while the presence of Escherichia coli was observed in the final product. CONCLUSIONS: Practices compromising the hygienic quality of the final product during commercial storage were observed and corrective measures suggested. To the best of the authors’ knowledge, these are the first data on microbiological safety in Brazilian fresh-cut processing plants.  2008 Society of Chemical Industry Keywords: fresh-cut lettuce; processing; safety INTRODUCTION Minimally processed vegetables (MPVs) have become popular amongst Brazilians. Although no official specific numbers exist for the production and consumption of these foods, it is known that the ready-to-eat food industry in Brazil increased by 148% between 1998 and 2000.1,2 This is due to a change in the behaviour of Brazilian society, in which 49% of women now work outside the home, be it of necessity or choice, apart from the facilities this type of product provides, such as economy of time and reduction of residues. Lettuce (Lactuca sativa) is amongst the most highly consumed vegetables in Brazil, with a production of 9091 t in the state of Rio de Janeiro in 1997, being an obligatory ingredient in salads, sandwiches and other items found in self-service restaurants.3 It is, nevertheless, known as a vehicle for pathogens and toxins, being one of the main leafy vegetables involved in cases of salmonellosis in the USA. A recent study showed the survival of Salmonella baildon in this product even after storage for 12 h at 4 ◦ C and exposure to 200 mg kg−1 chlorine.4 In São Paulo, Brazil, 61% of the samples of this vegetable analysed registered populations of faecal coliforms above 102 colony-forming units (CFU) g−1 , while in Campinas, Brazil, this index only reached 3.2% of the total, with a mean population of 1.2 most probable number (MPN) g−1 , the positivity index oscillating according to the season of the year, being 62.5% in summer and 12.5% in winter.5,6 In Nebraska, USA, 72 clients of a restaurant were infected with Escherichia coli O157:H7 and, after epidemiological studies, minimally processed lettuce was implicated as the probable source.7 A routine operation in a restaurant in California, USA resulted in an interstate food poisoning outbreak in which 61 people took sick and at least 21 were interned in hospital, three developing serious complications, the food ∗ Correspondence to: Adriano G da Cruz, Programa de Pós-Graduação em Ciência de Alimentos, Instituto de Quı́mica, Universidade Federal do Rio de Janeiro, CEP 21949-900, Rio de Janeiro, Brazil E-mail: food@globo.com (Received 21 August 2007; revised version received 1 November 2007; accepted 1 February 2008) Published online 17 April 2008; DOI: 10.1002/jsfa.3240  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 AG da Cruz, SA Cenci, MCA Maia vehicle being contaminated lettuce.8 Karenlampi and Hanninen9 reported better survival of Campylobacter jejuni strains in lettuce compared with other MPVs for times sufficient to cause food poisoning. Croci et al.10 reported better survival of the hepatitis A virus in this vegetable as compared with other products, and washing did not reduce the count. Doller et al.11 reported a food poisoning outbreak involving 34 people in a restaurant in Germany, caused by Cyclospora cayetanensis, a parasitic protozoa, after consumption of a salad containing imported minimally processed lettuce, resulting in clinical symptoms such as diarrhoea and vomiting and the need to take leave from work for an average period of 80 days. A survey carried out in the USA from 1973 to 1997 showed 25 food poisoning outbreaks caused by lettuce as the vehicle, resulting in 2078 sick people, 181 interned in hospital and six deaths, corresponding to 29% of the food poisoning outbreaks in the period caused by a single source.12 Although in Canada there are no specific statistical data available for this vegetable, it is known that there is a significant annual variation with respect to the occurrence of food poisoning outbreaks with MPVs as the vehicle, leading to an underestimation of the data.13 Another survey between 1982 and 2002, with the sole objective of discriminating the sources of food poisoning outbreaks related to E. coli O157:H7 in the USA, identified that 21% of these were related to MPVs, 74% occurring between the months of July and October and 34% being associated with lettuce.14 In Brazil, although the consumption of this type of food is on the rise, no data are available on food poisoning outbreaks related to MPVs, nor is there any information on the hygienic quality of the processing carried out by the agro-industries. The main sources of pathogenic bacteria in salads elaborated with fresh raw vegetables are the food handlers and the processing environment.15 The fact that these products are consumed by people with an elevated degree of sensibility, such as families with a high educational level, including children and elderly individuals, whose motivation for purchasing such products is convenience and speed of preparation,16 has increased the concern of the food industry and regulatory agencies with respect to the safety of such products. The objective of the present study was to identify the hazards involved in the various steps of minimal processing of lettuce by means of microbiological analyses of microbial groups used as indicators of hygienic conditions and of pathogens, with the aim of determining the sources of contamination and possible hazards offered by processing. MATERIALS AND METHODS Lettuce samples The study was carried out in an MPV production unit located in the state of São Paulo in the southeast 1456 of Brazil. This unit cultivates and processes various products such as lettuce, carrot, kale, tomato and mixed salads, their consumer market being the two largest centres in Brazil, i.e. São Paulo and Rio de Janeiro. Iceberg lettuce was chosen for this study owing to its importance in the diet of the Brazilian population, plus the fact that it is processed both as a single product and as an ingredient of salads, being of economic importance in agro-industrial sales. Lettuce samples were taken aseptically in duplicate during each of the different steps of the minimal processing carried out in the production unit, as described below. The temperature of the processing plant was approximately 10 ◦ C during the execution of the study. • Raw material reception/selection: carried out manually by handlers who discarded external and damaged leaves. • Washing: carried out with a detergent specific for vegetables, prepared according to the manufacturer’s instructions. • Rinsing: carried out with running water at a mean temperature of 10 ◦ C. • Sanitisation: carried out using ozone at a nonstandardised concentration and a mean temperature of 10 ◦ C. • Final product: plastic packaged in a passive way and stored in a cold chamber at 2 ◦ C until distributed. The samples were transported in insulating polystyrene containers under refrigeration to the Laboratory of Food Microbiology of Embrapa (Rio de Janeiro, Brazil) and analysed immediately on arrival. Each microbiological analysis was carried out in duplicate using ten samples per experiment and two samples per processing step. Since the experiment was repeated three times, a total of 30 samples were analysed. Microbiological analyses The microbiological tests consisted of analyses of microbial groups used as indicators of hygienic conditions (mesophilic bacteria, yeasts and moulds, total coliforms and faecal coliforms), pathogens (E. coli and Salmonella spp.) and inherent micro-organisms of interest for the minimal processing of vegetables (Pseudomonas spp. and psychrotrophic bacteria). Lettuce samples (25 g) were weighed into sterile stomacher bags, diluted with 225 mL of buffered peptone water (Merck, Rio de Janeiro, Brazil) and stomached. Aerobic and psychrotrophic plate counts were performed using Plate Count Agar (PCA; Merck) and subsequent incubation at 35 ◦ C for 48 h and at 5 ◦ C for 7–10 days respectively. Total and faecal coliforms were determined by the MPN method using triplicate Brilliant Green agar (BG; Merck) and E. coli broth (EC; Merck) and subsequent incubation at 35 ◦ C for 48 h and at 44 ◦ C for 24 h respectively. Escherichia coli was confirmed by subculturing the gas-positive tubes from EC broth into Eosin Methylene Blue agar (EMB; Merck) at 37 ◦ C for 24 h and performing appropriate biochemical tests. J Sci Food Agric 88:1455–1463 (2008) DOI: 10.1002/jsfa Microbiological hazards in fresh-cut lettuce processing Pseudomonas spp. counts were performed using KingB agar (KB; Merck) and subsequent incubation at 35 ◦ C for 48 h. Yeasts and moulds were determined using Potato Dextrose Agar (PDA; Merck) with 40 g L−1 tartaric acid solution and incubation at 22–25 ◦ C for 5 days. Additionally, for Salmonella spp. determination, lettuce samples (25 g) were weighed into sterile stomacher bags, diluted with 225 mL of buffered peptone water (Merck), stomached and incubated overnight at 35 ◦ C. A 1 mL aliquot of the pre-enriched sample was subsequently inoculated into 10 mL of mannitol Selenite Cysteine broth (SC; Merck) and Tetrationate broth (TT; Merck) and incubated at 42 and 35 ◦ C respectively. Then 1 mL of an appropriate dilution was spread-plated onto Salmonella–Shigela agar (SS; Merck), Hektoen enteric agar (HEK; Merck) and Brilliant Green agar (BG; Merck). These plates were incubated at 37 ◦ C for 24 h. Typical colonies were picked from each plate and inoculated into Triple Sugar Iron agar (TSI; Merck) and Lysine Iron Agar (LIA; Merck) slants. Each culture showing presumptive-positive TSI and LIA results was submitted to appropriate biochemical tests of confirmation. Statistical analysis Microbiological data were transformed into log MPN g−1 or log CFU g−1 according to the microbial group. Data were subjected to one-way analysis of variance (ANOVA), followed by the Tukey test in the case of significant differences. Probability P < 0.05 was considered significant. RESULTS AND DISCUSSION Tables 1–6 show the results of the microbiological analyses of the various microbial groups carried out during the minimal processing of lettuce. High microbial loads of mesophilic and psychrotrophic bacteria and Pseudomonas spp. were found in the ram reception (6 log CFU g−1 ), which were reduced by a single logarithmic cycle for the last two microbial groups after the sanitisation step (P < 0.05), the latter being ineffective against the first microbial group (P > 0.05). Lower counts of yeasts and moulds, total coliforms and faecal coliforms were observed in the initial step (3.49–4.43 log CFU g−1 , Table 1. Total mesophilic aerobic bacteria counts (log CFU g−1 ) Raw reception Washing Rinsing Sanitisation Packaged product I II III 6.88a 6.89a 6.83a 6.78a 6.49a 6.57a 6.47a 5.59a 6.43a 6.52a 6.38a 6.39a 7.04b 6.11a 6.22a Different letters in the same column indicate statistically significant differences according to the Tukey test (P < 0.05). J Sci Food Agric 88:1455–1463 (2008) DOI: 10.1002/jsfa Trial Processing step Raw reception Washing Rinsing Sanitisation Packaged product I II III 6.13a 6.14a 6.10a 6.08a 5.86a 6.54b 6.49a 6.35a 6.45a 5.59b 6.58acd 6.47ac 6.52acd 6.62ad 5.73b Different letters in the same column indicate statistically significant differences according to the Tukey test (P < 0.05). Table 3. Total yeast and mould counts (log CFU g−1 ) Trial Processing step Raw reception Washing Rinsing Sanitisation Packaged product I II III 3.49a 3.55a 3.50a 3.57a 4.41a 4.53a 4.28c 4.50b 4.58a 4.77a 4.20a 4.35b 4.51ac 4.77a 5.06a Different letters in the same column indicate statistically significant differences according to the Tukey test (P < 0.05). Table 4. Total Pseudomonas spp. counts (log CFU g−1 ) Trial Processing step Raw reception Washing Rinsing Sanitisation Packaged product I II III 6.13a 6.15a 6.65b 6.20a 6.38a 6.07a 6.78b 6.26c 6.09a 5.91d 6.11acd 6.88b 6.27ac 6.12acd 5.94ad Different letters in the same column indicate statistically significant differences according to the Tukey test (P < 0.05). Table 5. Total coliform counts (log MPN g−1 ) Trial Processing step Raw reception Washing Rinsing Sanitisation Packaged product I II III 0.65a 1.55a 1.44a 0.74a 0.55a 1.55d 1.61ad 1.73a 1.41b 1.28c 1.49c 1.65bd 1.74d 1.42a 1.25a Different letters in the same column indicate statistically significant differences according to the Tukey test (P < 0.05). Trial Processing step Table 2. Total psychrotrophic bacteria counts (log CFU g−1 ) 0.65–1.55 log MPN g−1 and 0.50–0.90 log MPN g−1 respectively), these values increasing significantly after the sanitisation step for yeasts and moulds (4.41–5.06 log CFU g−1 ) but remaining unaltered for coliforms (P > 0.05). Salmonella spp. were not found in any of the trials carried out (30 samples analysed), while the presence of E. coli was observed in the packaged product (two samples). 1457 AG da Cruz, SA Cenci, MCA Maia Table 6. Fecal Coliform count (log MPN g−1 ) Trial Processing step Raw reception Washing Rinsing Sanitisation Packaged product I II III 0.63a 0.58a 0.53a 0.58a 0.38a 0.90a 0.83a 0.59a 0.92a 0.73a 0.50a 0.55a 0.47a 0.80a 0.60a Different letters in the same column indicate statistically significant differences according to the Tukey test (P < 0.05). The micro-organisms found in MPVs should not prejudice the consumer, since they are also present in the raw material in the field. Thus high counts of Pseudomonas spp., Erwinia herbicola and lactic acid bacteria such as Leuconostoc mesenteroides and Lactobacillus spp. are commonly detected, the first normally being found in greater numbers and corresponding to 50–90% of the microbial load.17 Coliforms usually represent a smaller proportion of the initial bacterial contamination, as do yeasts and moulds.18 The results obtained in the present study indicated that Pseudomonas spp. represented 90% of the bacterial count on average. In general, the results suggested the need for refinement of the food safety prerequisite programmes such as Good Manufacturing Practices (GMP) and Sanitation Standard Operational Procedures (SSOP) carried out by the production unit, since the data indicated that the sanitisation step was not producing a significant reduction in numbers of the various microbial groups (P > 0.05) with respect to the final product, the counts remaining high, thus compromising product quality and shelf-life stability and leading to an acceleration of product deterioration. It must be emphasised that in the minimal processing of vegetables the function of the sanitisation step is not to eliminate the microbial population normally present, since that would be impossible. The idea is to maintain this parameter at low levels so that it can exert a positive role in the competition with pathogenic micro-organisms, impeding or inhibiting their growth throughout product shelf-life and thus retarding deterioration. In this context it is essential that the initial quality of the raw material be high and that the product be handled and packaged adequately, without temperature abuse, since the initial microbial load and product condition will have a considerable effect on product shelf-life.17 The high counts found for the various microbial groups in the raw material on reception, i.e. mesophilic bacteria (6.38–6.88 log CFU g−1 ), psychrotrophic bacteria (6.13–6.58 log CFU g−1 ), yeasts and moulds (3.49–4.53 log CFU g−1 ), Pseudomonas spp. (6.07–6.13 log CFU g−1 ), total coliforms (0.65–1.55 log MPN g−1 ) and faecal coliforms (0.50–0.90 log MPN g−1 ), suggested the need for integrated action at the field level by implementing quality norms such 1458 as Good Agricultural Practices (GAP), since a high microbial load in the raw material makes the action of the sanitising agent used in the process nonviable, whatever it may be, even if there is strict control of the conditions used during refrigerated storage of the product. In addition, the counts raise an alert for the presence of pathogenic microorganisms with psychrotrophic characteristics, e.g. Yersinia spp., Listeria spp. and Aeromonas ssp. De Curtis et al.19 reported the presence of Listeria spp. in 25% of MPVs sampled in Caracas, Venezuela, 30% of which were Listeria monocytogenes. Kaneko et al.20 investigated the microbiological quality of various ready-to-eat products, including vegetables before and after minimal processing, in supermarkets in Tokyo and found that mesophilic counts in the postprocessed vegetables ranged from 103 to 107 CFU g−1 , with 50% of the samples positive for coliforms, and E. coli and Bacillus cereus present in 9.2% of the samples. Before processing, L. monocytogenes was found in only two samples. The authors also concluded that, of the samples analysed, the vegetables showed a greater degree of contamination before processing, with a mesophilic count in the range 102 –107 CFU g−1 , which remained unaltered after processing and storage at 10 ◦ C for 3 days. Thunberg et al.21 analysed various MPVs sold in retail outlets in the USA and found mesophiles in the range from 2.0 × 106 to 5.0 × 108 CFU g−1 , with lettuce showing the highest counts and 89–96% of the bacteria present being Gram-negative. Yeasts and moulds were found in lower numbers, Salmonella spp. and Campylobacter spp. were shown to be absent and Listeria ssp. (including L. monocytogenes) were isolated in the experiments. A previous study carried out by our group in the same production unit identified various faults in the execution of GAP, which were translated as a potential to obtain raw materials with high microbial loads, representing a public health hazard.22 Different results were found for producers in Iowa, USA, where the adoption of measures such as the use of water with a high degree of potability (tested at intervals by official laboratories) for irrigation and sanitisation of the equipment and utensils, adequate composting practices and training of the handlers for the operations showed an effect leading to greater adherence to the GAP norms, although some adjustments were still necessary.23 The adoption of quality systems in the minimal processing of vegetables obligatorily passes through a quality guarantee at the field level, signifying the unrestricted adoption of GAP in order to obtain raw materials with minimal physical, chemical and microbiological hazards for the processing units. In addition, it can serve as a marketing instrument for the product, adding value to the product and giving greater impact when faced with competitors, and an important factor in conquering new markets. Kokkinakis and Fragkiadakis24 investigated the effect J Sci Food Agric 88:1455–1463 (2008) DOI: 10.1002/jsfa Microbiological hazards in fresh-cut lettuce processing of adopting quality systems (GAP-certified raw material) in catering establishments in Greece. Establishments that acquired GAP-certified products presented significantly better microbiological quality than the others, as expressed by low counts of total coliforms and mesophilic bacteria, low E. coli populations and the absence of L. monocytogenes. These values were reduced even more by processing the products, reaching 0% E. coli and maintaining the absence of L. monocytogenes in the final product. On the other hand, establishments not maintaining this practice showed increases in the microbial load of coliforms and mesophilic bacteria and 38% presence of L. monocytogenes in the final product. Little et al.25 reported 82% acceptable microbiological quality in imported minimally processed lettuce commercialised in England, verifying the absence of diverse pathogens such as E. coli O157:H7, Shigella spp., Salmonella spp., Campylobacter spp. and Vibrio spp. in 100% of the samples, which indicated adoption of the quality norms in the field and in the industries of the products analysed. A subsequent, wider study in the same country in 2002, with 3852 samples of salads elaborated with MPVs, indicated 99.3% of samples with satisfactory or acceptable microbiological quality, only 0.7% being condemned owing to the presence of E. coli and Listeria spp. at levels above 102 CFU g−1 and a positive result for Salmonella spp.26 Similar results were observed in Mexico, where the absence of pathogens such as Shigella spp., Salmonella spp. and E. coli O157:H7 was observed in the analysis of various fresh cuts of domestic origin, with only 1% of samples positive for L. monocytogenes.27 These results show that it is possible to achieve the food safety and quality assurance requirements in MPVs. Nevertheless, the results obtained in the present study are in agreement with various earlier studies that report mesophilic bacterial loads of 3–6 CFU g−1 .18 Tauxe et al.28 indicated that the interaction of four factors was responsible for the survival, growth and inactivation of micro-organisms in fresh cuts: the characteristics of the micro-organisms present; the physiological state of the vegetables and their resistance to the microbial metabolic process; the characteristics of the processing environment; and the effects of the processing steps on the microbial population. Thus an overall analysis from the raw material in the field to the final, ready-to-eat product becomes necessary for a successful process assurance programme. Another relevant factor observed was the nonexistence of parameters associated with the action of the sanitising agent (contact time, temperature and concentration) during the process. As observed visually, immersion of the vegetables in the ozone sanitisation tank failed to obey any standardisation with respect to contact time or temperature, being random and depending on the will of the handler. This situation contributed to the results obtained, since various recent studies have shown the efficiency J Sci Food Agric 88:1455–1463 (2008) DOI: 10.1002/jsfa of ozone with/without other organic agents in decreasing and/or inactivating the microbial count and/or pathogens in MPV plants, without prejudice to the sensory quality of the vegetable, with results many times superior to those of chlorine.29 – 39 Thus a change in the process layout was suggested, substituting the first step of washing with a detergent specific for vegetables, with washing with chlorine associated with ozone in the sanitisation step, to take advantage of the synergistic effect of the sanitising agents, since many studies have shown the efficiency of the use of chlorine in MPV washing, especially for lettuce, with an improvement in the safety level of the process and without prejudice to the sensory quality.40 – 43 A lower temperature in the washing tank, close to 5 ◦ C, was also suggested, since higher temperatures create conditions for the internalisation of pathogens present in the vegetable by chance, which would probably not be removed in the subsequent sanitisation step with ozone.44 Internalisation of pathogens in MPVs is of great concern, control being more effective by controlling the microbial load of the raw material in the field.44 Additional experiments have become necessary to determine the optimal values for the parameters of the sanitising agents, for a better analysis of the process with respect to the reduction in microbial load. As a starting point, the adoption of values in the range from 100 to 200 ppm chlorine was suggested, with continuous monitoring of the pH throughout the processing line, since the germicidal form of chlorine is potentised in hypochlorous acid (HClO), which, since it is a weak acid, is susceptible to dissociation in aqueous solution. Allende et al.45 showed a significant reduction in psychrotrophic micro-organisms and coliforms (3 and 2 log CFU g−1 respectively) after washing with 160–180 µg g−1 chlorinated water at 10 ◦ C for 1 min in a minimally processed red lettuce line, demonstrating the importance of this step. Soriano et al.46 analysed 144 lettuce samples from university restaurants in Spain, obtaining values between 103 and 107 CFU g−1 for mesophilic organisms and from <3 to >2400 MPN g−1 for faecal coliforms, with Aeromonas hydrophila, Pseudomonas aeruginosa, E. coli and Staphylococcus aureus also being detected in 10.4, 2.8, 25.7 and 22.9% of the samples respectively. Washing with sodium hypochlorite and potassium permanganate reduced the mesophilic count by one logarithmic cycle and the total coliform count by two logarithmic cycles. Simons and Sanguansri47 warned that the washing system should take the characteristics of the product into consideration, so lettuce, being a fragile, delicate product, should be washed gently to avoid physical damage. The rinsing process presented significantly higher counts for faecal coliforms and yeasts and moulds (P < 0.05) and was inefficient for the other microbial groups analysed (P > 0.05), suggesting that the quality of the water used by the industry was inadequate. In 1459 AG da Cruz, SA Cenci, MCA Maia fact, this water was merely filtered through a sand filter, which was not periodically changed, suffering no chemical disinfection before contact with the vegetables. A previous study detected the presence of total and faecal coliforms (0.48 and 1.20 log MPN per 100 mL respectively) in this water supply, not complying with the conditions required for food processing.48 Howard and Gonzalez49 warned that water quality is of extreme importance, since water serves as a vehicle for many pathogenic microorganisms, and that all operations in which it takes part in washing or rinsing processes deserve special attention by the processors. McKnight50 emphasised that the potability of the water should be assured in all steps of the food production chain, from start to finish, and that this parameter should be documented and regularly monitored. Leoni et al.51 suggested that the microbiological quality of the washing water could be used as an indicator of the microbiological quality of the vegetables. On analysing samples from a processing unit, they found faecal coliforms, E. coli and Enterococcus spp. in 42.9, 40.0 and 77.1% of the samples respectively, demonstrating the existence of faecal material in the vegetables. The lower counts found for yeasts and moulds (3–4 log CFU g−1 ) throughout the processing steps can be explained by the characteristics of the vegetables, i.e. low acidity and high water activity, conditions in which yeasts and moulds compete badly with bacteria and can be eliminated or reduced by good manufacturing practices in the processing line, especially by optimising the operations involving the cutting and removal of external leaves with a minimum of damage to the vegetable, plus continuous sanitisation of the handlers and utensils and of the air present in the processing line. The need to implement these actions must also be emphasised in order to minimise the potential danger of these organisms in producing metabolites such as mycotoxins, which could be present in the final product, representing a hazard to public health, in addition to reducing the time in which the product is fit for consumption. Tournas52 carried out a mycological analysis of various fresh-cut products commercialised in Washington, USA and found a high prevalence of this microbial group, especially in lettuce, with counts between 3.11 and 5.97 log CFU g−1 . The isolation of Alternaria spp., Cladosporium spp. and Penicillium spp. was observed in 10.3, 23.3 and 5.1% of the samples respectively, these indices being 12.5, 40.0 and 20.0% for lettuce, which, allied to their capacity to grow at refrigeration temperatures, suggests a situation of alert with respect to the production of mycotoxins. The centrifuge was considered to be a potential source of contamination, since a significant increase in the microbial load (P < 0.05) was shown during this step, maintaining the previously observed levels or a non-significant decrease (P > 0.05) for yeasts and moulds and the other microbial groups in the final product. This reinforces the need for periodic 1460 sanitisation of this equipment, as expressed in a written procedure understandable by all the handlers, since it reflects directly on the hygienic quality of the final product. The operational conditions, e.g. rotational speed, should also be standardised. This step contributes to the removal of excess water from the surface of the vegetable and, if done badly, has a direct implication on the maintenance of optimal conditions for microbial growth in the final product and also for the activity of enzymes naturally present in the vegetable. Badly sanitised equipment represents a potential source of contamination in the food industry and should be taken into account for the safety of the process. Garg et al.53 carried out microbiological analyses on various MPVs sampled from processing lines before and after their passage through various equipments/utensils (cutters, peelers, slicers and centrifuges) and showed an increase of two logarithmic cycles in the microbial count, the lettuce cutter (from 4.25 to 6.15 log CFU g−1 ) and onion slicer (from 3.6 to 5.08 log CFU g−1 ) presenting the greatest increases, reflecting the source of contamination they represented. Kaneko et al.20 analysed various equipments/utensils in vegetable processing units before and after sanitisation, obtaining increases of two logarithmic cycles in the microbial load after passage of the vegetables through these equipments. Allende et al.45 reported increases of 1.24 log CFU g−1 for psychrotrophs and 1.0 log CFU g−1 for coliforms in a recent experiment on a minimally processed red lettuce line and confirmed the increase in microbial load in the final product after passage of the product through the centrifuge. Previous experiments by this research group revealed the precarious hygienic quality of the equipments and utensils of the production line in the agro-industry studied – specifically the centrifuge and transport belt – with an elevated index of total and faecal coliforms, confirming previously discussed hypotheses.48 Jay54 pointed out that the mesophilic count is commonly used to determine the hygienic quality of foods, an elevated count being a sign of unwholesomeness of the food, caused by at least one of the following factors: contaminated raw material; unsatisfactory processing with abuse of the time/temperature binomial; and precarious hygienic conditions of the process, equipment and handlers. In addition, since the majority of pathogenic bacteria are mesophilic, high counts signify that there were adequate conditions for them to multiply. However, in the minimal processing of vegetables, such data are of limited value and do not necessarily reflect the true number of live microbial cells and their action on the shelf-life of the product. For this type of food, psychrotrophic bacteria are the group which predominate and are responsible for product deterioration and are therefore a good indicator of the hygienic quality of the process, since their analysis simulates the conditions found in the processing plants, i.e. low-temperature J Sci Food Agric 88:1455–1463 (2008) DOI: 10.1002/jsfa Microbiological hazards in fresh-cut lettuce processing environments.17 Szabo et al.55 analysed 120 samples of minimally processed lettuce samples in Australian supermarkets and found counts of 103 –109 CFU g−1 for mesophilic organisms, with 76% of the samples in the range 105 –107 CFU g−1 , and the presence of L. monocytogenes and of A. hydrophila and Yersinia enterocolitica in 36 and 71 samples respectively. Froder et al.6 found 51 and 68% of minimally processed salads marketed in São Paulo, Brazil with populations of mesophilic and pyschrotrophic micro-organisms above 106 CFU g−1 respectively. Jay54 defined total coliforms as a group of bacteria of the Enterobacteriaceae capable of fermenting lactose with the production of gas when incubated at 35–37 ◦ C for 48 h, normally found in faeces and soil and on the surface of vegetables, where they survive longer than pathogenic bacteria such as E. coli and Salmonella spp. High counts of total coliforms in MPVs can indicate that the raw material was cultivated in soil rich in organic matter and/or with an abnormal bacterial incidence on the vegetable, not indicating, however, a problem related to product quality and being of limited use in a programme for the microbiological quality of minimally processed foods, serving only as an indicator of the efficiency of the sanitisation process.17 The faecal coliform group of bacteria corresponds to that part of the total coliforms with the capacity to continue fermenting lactose with the production of gas when incubated at a temperature of 44–45.5 ◦ C. They normally grow in human and animal intestines and are associated with faecal material,54 the main representative being E. coli, which is thus used as an indicator of faecal contamination. In food processing, especially in the minimal processing of vegetables, the presence of E. coli indicates contamination by a source that could contain pathogenic enterobacteria such as Salmonella spp. and Shigella spp., resulting in the product being rejected for human consumption owing to being in an unsatisfactory sanitary condition according to the norms of Brazilian legislation.56 Thus it is important to test for faecal coliforms positive for E. coli.17 In the particular case of the production unit under study, low counts of total and faecal coliforms were observed during the processing steps, specifically 0.65–1.74 and 0.47–0.92 log MPN g−1 respectively. These results were not in accordance with other research, as normally a high incidence of this microbial group is observed in foods of vegetable origin, suggesting an occasional situation. Acevedo et al.57 observed higher counts on analysing salads elaborated with fresh vegetables, finding 5.16 log MPN g−1 of total coliforms and 4.66 log MPN g−1 of faecal coliforms and the absence of E. coli, which could be related to the local cultivation conditions of the vegetables. Nevertheless, the presence of E. coli in the product serves as a signal for concern with processing safety, its presence in the final product step being attributed J Sci Food Agric 88:1455–1463 (2008) DOI: 10.1002/jsfa to handlers in the final product packaging sector, demonstrating the need for training of the handlers in hygiene and food microbiology as part of the refinement of GMP, one of the non-conforming items detected in the audit.48 Many buyers impose limits on the mesophilic and total coliform counts when buying raw materials of vegetable origin destined for minimal processing, rejecting samples above these limits. It must be understood that the existence of a microbiological monitoring programme with a systematic interpretation of the results, aiming to improve the sanitisation programmes of the processing line, is of great utility and that high counts make sense when considered together with a poor history of processing results, encouraging improvements aimed at the new limits.17 Salmonella spp. were not detected in any of the trials. This could be related to the methodology used, which, although official, includes a large number of steps, and also to the low level of competitiveness in the presence of other microbial groups, it being important to develop methodologies with greater sensitivity, since the shelf-life of MPVs is intrinsically short. In addition, one should not ignore the possibility of continuous monitoring and a greater number of analyses, especially with respect to the handlers, who could become non-symptomatic carriers, excreting in their faeces for a long period of time. Various studies have reported the presence of Salmonella spp. at rates varying from 7.5 to 68%, with lettuce showing the highest incidence.58 Salleh et al.59 showed the presence of Salmonella spp. in 35% of raw vegetables commercialised in Selangor, Malaysia, of which S. weltevreden, S. agona, S. senftenberg and S. Albany were the most commonly isolated serotypes. Up to the present date, Brazilian legislation has not set specific microbiological standards for MPVs, and only recently was the intention announced to create legislation for GMP in agro-industries of these products at the level of the state of São Paulo. It contemplates guidance with respect to the layout of the establishment, with prior obligation to register in the sanitation organs, and to raw material quality, periodic checking of water potability, effluent treatment, sanitary installations, handling area, obligation to have a laboratory for analyses, packaging and labelling, and a Sanitary Certificate being awarded annually to those establishments complying with the specified demands,60 showing the concern of the authorities. However, if these products were inserted in the category ‘fresh vegetables, in natura, prepared, sanitised, refrigerated or frozen for direct consumption’ with the recommended maximum limit of 2 log CFU g−1 for faecal coliforms and the absence of Salmonella spp., 100% of the samples obtained in the final product step would comply with these limits,56 representing a dangerous situation considering the results obtained, and aggravated by the positive result for E. coli. The establishment of standards specific for these products in Brazilian 1461 AG da Cruz, SA Cenci, MCA Maia legislation is essential, respecting the peculiarities involved in their processing and attributing maximum standards for microbial groups such as mesophilic and psychrotrophic bacteria and for pathogens such as L. monocytogenes and E. coli amongst others. Studies similar to the present one should be carried out simultaneously in other processing plants in Brazil to provide a greater number of results for use in the elaboration of this legislation. CONCLUSIONS The results of this research show the level of safety offered by an MPV production unit, especially with respect to the minimal processing of lettuce, being a pioneer study with respect to Brazilian agro-industries. Practices compromising the hygienic quality of the product and its period of commercial storage were detected, all of which are covered by food safety programmes such as GMP and SSOP. Corrective actions must be taken in order to produce products with no hazard to public health. REFERENCES 1 Branco A, Dole investe US$ 100 mil. em legumes. Gazeta Mercantil, 2–3, São Paulo (1997). 2 Nascimento EFN, Molica EM and de Moraes JS, Hortaliças minimamente processadas – mercado e produção. Brası́lia (2000). 3 Embrapa, Controle de qualidade em alface (Lactuca sativa) cultivada em sistema hidropônico. Brası́lia (1999). 4 Weissinger WR, Chantarappanont W and Beuchat LR, Survival and growth of Salmonella baildon in shredded lettuce and diced tomatoes and effectiveness of chlorinated water as sanitizer. Int J Food Microbiol 62:123–131 (2000). 5 Porto E and Elroa MNU, Occurence of Listeria monocytogenes in vegetables. Dairy, Food, Environ Sanit 21:282–285 (2001). 6 Froder H, Martins CG, de Souza KLO, Landgraf M, Franco BDG and Destro MT, Minimally processed vegetables salads: microbial quality evaluation. J Food Protect 5:1277–1280 (2007). 7 Wachtel MR and Charkowski AO, Cross-contamination of lettuce with Escherichia coli O157:H7. J Food Protect 65:465–470 (2002). 8 De Wall CS, Safe food from a consumer perspective. Food Control 14:75–79 (2003). 9 Karenlampi R and Hanninen ML, Survival of Campylobacter jejuni on various fresh produce. Int J Food Microbiol 97:187–195 (2004). 10 Croci L, de Medici D, Scalfaro C, Flore A and Toti L, The survival of hepatitis A virus in fresh produce. Int J Food Microbiol 73:29–34 (2002). 11 Doller PC, Dietrich K, Filipp N, Brockmann S, Dreweck C, Vonthein R, et al, Cyclosporiasis outbreak in Germany associated with the consumption of salad. Emerg Infect Dis 8:992–994 (2005). 12 Sivapalasingam S, Friedman CR, Cohen L and Cauxe RV, Fresh-produce: a growing cause of outbreaks of foodborne illness in the United-States, 1973 through 1997. J Food Protect 67:2342–2353 (2004). 13 Sewell AM and Farber JM, Foodborne outbreaks in Canada linked to produce. J Food Protect 64:1863–1877 (2001). 14 Rangel JM, Sparling PH, Crowe C, Griffin PM and Swerdlow DL, Epidemiology of Escherichia coli O157:H7 outbreaks, United States, 1982–2002. Emerg Infect Dis 11:603–08 (2005). 1462 15 Fain RF, A review of the microbiological safety of fresh-salads, Dairy, Food, Environ Sanit 16:146–149 (1996). 16 Ragaert P, Verbeke W, Devlieghere F and Debever J, Consumer perception and choice of minimally processed vegetables and packaged fruits. Food Qual Pref 14:259–270 (2004). 17 International Fresh-cut Produce Association, Food Safety Guidelines for the Fresh-cut Industry (3rd edn). International Fresh-cut Produce Association, Washington, DC (1998). 18 Nguyen-the C and Carlin F, The microbiology of minimally processed fresh fruits and vegetables. Crit Rev Food Sci Nutr 34:371–401 (1994). 19 De Curtis ML, Franceschi O and De Castro N, Listeria monocytogenes in vegetables minimally processed ready-to-use. Arch Latino Am Nutr 52:282–287 (2002). 20 Kaneko K-I, Hayashidani H, Ohtomo Y, Kosuge J, Kato M, Takahashi K, et al, Bacterial contamination of ready-to-eat foods in fresh products in retail shops and food factories. J Food Protect 62:644–649 (1999). 21 Thunberg RL, Tran TT, Bennett RW, Matthews RN and Belay N, Microbial evaluation of selected fresh produce obtained in retails markets. J Food Protect 65:677–682 (2002). 22 Da Cruz AG, Cenci SA and Maia MCA, Good agricultural practices in a Brazilian produce plant. Food Control 17:781–788 (2006). 23 Ellis JD, Strohbehn CH and Henroid Jr DH, Assessing on-farm food handling practices of Iowa grown produce and eggs in regard to food safety. Food Protect Trends 25:758–761 (2005). 24 Kokkinakis EN and Fragkiadakis GA, HACCP effect on microbiological quality of minimally processed vegetables: a survey in six mass-catering establishments. Int J Food Sci Technol 42:18–23 (2006). 25 Little C, Roberts D, Youngs E and de Louvois J, Microbiological quality of recall imported unprepared whole lettuces: a PHLS Food Working Group study. J Food Protect 62:325–328 (1999). 26 Sagoo SK, Little CL, Gillespie LA and Mitchell RT, Microbiological study of ready-to-eat salad vegetables from retail establishments uncovers a national outbreak of salmonellosis. J Food Protect 66:403–409 (2003). 27 Johnston LM, Jaykus L-A, Moll D, Anciso J, Mora B and Moe CL, A field study of the microbiological quality of domestic and Mexican origin. Int J Food Microbiol 112:83–95 (2006). 28 Tauxe R, Kruse H, Hedberg C, Potter M, Madden J and Wachsmuth K, Microbial hazards and emerging issues associated with produce: a preliminary report to the National Advisory Committee on Microbiological Criteria for Foods. J Food Protect 60:1400–1408 (2006). 29 Kim JG, Yousef AE and Dave S, Application of ozone for enhancing the microbiological safety and quality of foods: a review. J Food Protect 62:1071–1087 (1999). 30 Kim J-G, Yousef AE and Chism GW, Use of ozone to inactivate microorganisms on lettuce. J Food Safety 19:7–34 (1999). 31 Siveira NFA, Berbari SG, Prestes EB, Bliska FMM and Silva N, Efeito do uso do ozônio na água de lavagem para desinfecção de alface minimamente processada. Anais do XIX Congresso de Ciência e Tecnologia de Alimentos, Recife (2004). 32 Singh N, Singh RK, Bhunia AK and Stroshine RL, Efficacy of chlorine dioxide, ozone and thyme essential oil or a sequential washing in killing Escherichia coli O157:H7 on lettuce and baby carrots. LWT – Food Sci Technol 35:720–729 (2002). 33 Garcia A, Mount JR and Davidson PM, Ozone and chlorine treatment of minimally processed lettuce. J Food Sci 68:2747–2751 (2003). 34 Baur S, Klaiber R, Hammes WP and Carle R, Sensory and microbiological quality of shredded packaged iceberg lettuce as affected by pre-washing procedures with chlorinated and ozonated water. Innov Food Sci Emerg Technol 5:45–55 (2004). 35 Beltran D, Selma MV, Marin A and Gil MJ, Ozonated water extends the shelf-life of fresh-cut lettuce. J Agric Food Chem 53:5654–5663 (2005). J Sci Food Agric 88:1455–1463 (2008) DOI: 10.1002/jsfa Microbiological hazards in fresh-cut lettuce processing 36 Uyttendaele M, Neyts K, Vanderswalmen H, Notebaert E and Debevere J, Control of Aeromonas on minimally processed vegetables by decontamination with lactic acid, chlorinated water or thyme essecial oil solution. Int J Food Microbiol 90:263–271 (2006). 37 Yuk HG, Yoo MY, Moon KD, Marshall DL and Oh DH, Effect of combined ozone and organic acid treatment for control of Escherichia coli O157:H7 and Listeria monocytogenes on lettuce. J Food Sci 71:M83–M87 (2006). 38 Koseki S and Isobe S, Effect of ozonated water treatment on microbial control and on browning of iceberg lettuce (Lactuca sativa L.). J Food Protect 69:154–160 (2006). 39 Selma MV, Beltrán D, Allende A, Vera-Chacón E and Gil MI, Elimination by ozone of Shigella sonnei in shredded lettuce and in water. Food Microbiol 24:492–499 (2007). 40 Berbari SA, Paschoalino JE and Silveira NFA, Effect of chlorine in the washing water for desinfection of minimally processed lettuce. Ciênc Tecnol Alim 21:197–201 (2001). 41 Li Y, Brackett RE, Shewfelt RL and Beuchat LR, Changes in appearance and natural microflora on iceberg lettuce treated in warm, chlorinated water and then stored at refrigeration temperatures. Food Microbiol 18:299–308 (2001). 42 Delaquis PJ, Fukumoto LR, Toivonen PMA and Cliff MA, Implications of wash water chlorination and temperature for the microbiological and sensory properties of fresh-cut iceberg lettuce. Postharv Biol Technol 31:81–91 (2004). 43 Baur S, Klaiber R, Hammes WP and Carle R, Effect of temperature and chlorination of pre-washing on shelf-life and physiological properties of ready-to-use lettuce. Innov Food Sci Emerg Technol 6:171–182 (2005). 44 Aruscavage D, Lee K, Miller S and LeJeune JT, Interactions affecting the proliferation and control of human pathogens on edible plants. J Food Sci 71:R89–R94 (2006). 45 Allende A, Aguayo E and Artés F, Microbial and sensory quality of commercial fresh red lettuce through the production chain and shelf life. Int J Food Microbiol 91:109–117 (2004). 46 Soriano JM, Rico H, Molto JC and Manes J, Assessment of the microbiological quality and wash treatments of lettuce served in University restaurants. Int J Food Microbiol 58:123–128 (2000). 47 Simons LK and Sanguansri P, Advances in the washing of minimally processed vegetables. Food Aust 49:75–80 (1997). J Sci Food Agric 88:1455–1463 (2008) DOI: 10.1002/jsfa 48 Da Cruz AG, Cenci SA and Maia MCA, Pré-requesitos para implementação do sistema APPCC em uma unidade produtora de vegetais minimamente processados. Ciênc Tecnol Alim 26:104–109 (2006). 49 Howard LR and Gonzalez AR, Food safety and produce operation: what is the future? HortScience 36:33–39 (2001). 50 McKnight S, Issues in water and safety. Food Protect Trends 22:512–513 (2002). 51 Leoni E, Legnani PP and Lucano A, Analysis of a critical control point in the production and distribution line of fresh vegetable products. Ind Alim 42:361–366 (2003). 52 Tournas VH, Moulds in fresh and minimally processed vegetables and sprouts. Int J Food Microbiol 99:71–77 (2005). 53 Garg N, Churey JJ and Splittstoesser DF, Effect of processing conditions on the microflora of fresh-cut vegetables. J Food Protect 53:701–703 (1990). 54 Jay JM, Modern Food Microbiology. Springer, New York, NY (2005). 55 Szabo EA, Scurrah KJ and Burrows JM, Survey for psychrotrophic bacterial pathogens in minimally processed lettuce. Lett Appl Microbiol 30:456–460 (2000). 56 Agência Nacional de Vigilância Sanitária, Regulamento técnico sobre padrões microbiológicos para alimentos. Resolução Diretiva Colegiada No. 12. [Online]. Ministério da Saúde, Brazil (2001). Available: http://www.anvisa.gov.br [13 February 2007]. 57 Acevedo L, Mendoza C and Oyon R, Total and fecal coliform, some enterobacteria, Staphylococcus sp and moulds in salads for hot-dogs sold in Maracay, Venezuela. Arch Latino Am Nutr 51:366–370 (2001). 58 Francis GA, Thomas C and O’Beirne D, The microbiological safety of minimally processed vegetables. Int J Food Sci Technol 34:1–22 (1999). 59 Salleh NA, Rusul G, Hassan Z, Reezal A, Isa SH, Nishibuchi M, et al, Incidence of Salmonella in raw vegetables in Selangor, Malaysia. Food Control 14:475–479 (2003). 60 De Brito PF, Guimarães AB, Salate CA and Alonso CAM, Legislação para agroindustrias de vegetais minimamente processados. Anais do I Simposio ı́bero-Americano de Vegetais Frescos e Cortados, Piracicaba, pp. 41–49 (2006). 1463 Journal of the Science of Food and Agriculture J Sci Food Agric 88:1464–1471 (2008) Effects of nitrogen application on malt modification and dimethyl sulfide precursor production in two Japanese barley cultivars Masahito Nanamori,1,2∗ Ryoichi Kanatani,1 Makoto Kihara,1 Kazumitsu Kawahara,1 Katsuhiro Hayashi,1 Toshihiro Watanabe,2 Takuro Shinano3 and Mitsuru Osaki2 1 Department of Bioresources Research and Development, Sapporo Breweries Ltd, 37-1 Nittakizaki, Ota, Gunma 370-0393, Japan School of Agriculture, Hokkaido University, Kitaku, Sapporo 060-0859, Japan 3 Creative Research Initiative ‘Sousei’, Hokkaido University, Kitaku, Sapporo 001-0021, Japan 2 Graduate Abstract BACKGROUND: Nitrogenous components have a great influence on both malt and beer qualities. Barley storage proteins are degraded during the germination process, in which amino acids and small peptides are released. Some of these compounds relate to dimethyl sulfide precursor production in the malting process. In this study, barley and malt qualities were investigated using two Japanese barley cultivars, Sukai Golden and Mikamo Golden, with several different nitrogen (N) treatments. RESULTS: Nitrogen top-dressing treatments efficiently increased N and sulfur (S) concentrations in grains. A difference in malt modification was induced by these treatments without any change in protease activity in malts. S-Methyl methionine (SMM) concentration in malt of Sukai Golden with low-N treatment was 1.8–2.1 times higher than that with higher-N treatments. Methionine concentration in malts was not significantly affected by N treatments of both cultivars, while grain S level was not consistent under any treatments. CONCLUSION: Results show that low-N treatment increases SMM concentration in malts despite major Scontaining amino acids of malts being not highly affected by the difference in nutrient status of grains. Further investigations are necessary into aspects of both metabolic profiles in barley germination and SMM degradation in the kilning process.  2008 Society of Chemical Industry Keywords: nitrogen fertilisation; barley quality; malt modification; dimethyl sulfide precursor; S-methyl methionine INTRODUCTION High-molecular-weight compounds such as starches and proteins in barley grain are degraded as germination proceeds and are used for subsequent seedling development. These compounds are also fermented with yeasts in brewing. In the brewing process it is important to understand the malting traits in order to control the malt qualities as well as the final beer quality. Barley with high nitrogen (N) content causes low recovery of extracts from malts and may influence the formation of haze in the finished beer. On the other hand, barley with low N content causes malnutrition of yeasts and affects the foam stability of the finished beer. Thus precise regulation of the N content in the seed is very important for the beer-brewing process. Storage proteins in barley are degraded into peptides and amino acids by four endoproteases.1 One of the most important proteases is cysteine protease, because it is considered to have the strongest influence on malt qualities such as soluble nitrogen (SN) in wort.2 The degradation of storage proteins by proteases is of importance in malting, because SN in wort is essential for fermentation by yeasts. The Kolbach index (KI), which indicates the degree of solubilisation of storage proteins in malts, has a great influence on beer quality.3 The patterns of uptake of individual amino acids formed by proteolysis in malting and mashing are categorised into four groups.4 In a study on the effects of individual amino acids, Edney and Langrell5 reported that the amino acid content in the wort had no influence on the final attenuation limit, ∗ Correspondence to: Masahito Nanamori, Department of Bioresources Research and Development, Sapporo Breweries Ltd, 37-1 Nittakizaki, Ota, Gunma 370-0393, Japan E-mail: masahito.nanamori@sapporobeer.co.jp (Received 29 November 2007; revised version received 21 January 2008; accepted 25 January 2008) Published online 17 April 2008; DOI: 10.1002/jsfa.3241  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Effects of nitrogen application on Japanese barley cultivars but that a difference in barley cultivars had a significant influence on the final attenuation limit. However, it has been suggested that amino acids such as methionine and lysine affect the fermentation performance.6,7 Dimethyl sulfide (DMS), which gives a boiled vegetable flavour to beer, is formed from its precursors S-methyl methionine (SMM) and dimethyl sulfoxide (DMSO). SMM is involved in the methionine cycle in plant metabolism, and it was shown that SMM contributed to phloem sulfur (S) transport in wheat during maturation, with 80% of the S in the grain being transported as SMM.8 There have been several reports on the physiological role of SMM,9,10 but it has not yet been fully elucidated. SMM is synthesised in germinating barley seed from L-methionine and S-adenosyl-L-methionine. The reaction is catalysed by S-adenosyl-L-methionine:L-methionine Smethyltransferase (MMT), which is localised in the shoot, scutellum and aleurone cells but not in the root or endosperm.11 After the germination period, some of the synthesised SMM decomposes into DMS or DMSO during the kilning process and wort boiling, and then DMSO is reduced to DMS by yeasts during fermentation.12 The protein content in barley and its malting traits are of great importance in the influence of N compounds on beer quality. Barley with 9–12% protein content is believed to be favourable for making malts in Japan, so adequate field management is necessary. Barley protein content is affected by several factors such as temperature, rainfall, amount of fertiliser and timing of its application. In particular, the management of N fertilisation is important in controlling barley protein content, and N fertilisation affects β-amylase activity and malt qualities.13 Barley protein content also affects endosperm texture and constituents including β-glucan and arabinoxylan in the grain.14 It is desirable to understand the barley qualities and its malting traits before using such material in malt-houses and breweries. Beer quality depends on the barley, and barley quality is dependent on the growing conditions in the field. The aim of this study was to investigate the influence of N fertilisation on malt qualities, particularly focusing on malt modification and dimethyl sulfide precursor (DMSP) production, using two Japanese barley cultivars. MATERIALS AND METHODS Barley cultivation A field trial was carried out in the 2005–2006 growing season. Two Japanese barley cultivars, Sukai Golden and Mikamo Golden, were grown in a test plot of Sapporo Breweries Ltd in Gunma, Japan (36◦ 16′ N, 139◦ 18′ E). Sukai Golden is a well-modified cultivar with rapid germination. Mikamo Golden has a well-balanced modification property and the largest cultivated acreage of Japanese malting barley cultivars. An N basal dressing of 30 or 60 kg N ha−1 was applied as urea. Plants were grown with or without an J Sci Food Agric 88:1464–1471 (2008) DOI: 10.1002/jsfa N top-dressing of 30 kg N ha−1 at the booting stage in each basal treatment respectively: ‘30/0’ means 30 kg N ha−1 of basal dressing without top-dressing; ‘30/30’ means 30 kg N ha−1 of basal dressing with 30 kg N ha−1 of top-dressing; ‘60/0’ means 60 kg N ha−1 of basal dressing without top-dressing; ‘60/30’ means 60 kg N ha−1 of basal dressing with 30 kg N ha−1 of top-dressing. Phosphorus and potassium were applied as a basal dressing of 60 kg ha−1 (equivalent amount of P2 O5 or K2 O respectively) for all treatments. Each treatment was carried out with three replicates. Measurement of N and S concentrations Barley and malt samples were milled and digested with sulfuric acid and copper sulfate for N measurement. Total N was determined by the Kjeldahl method.15 Malt protein was calculated by multiplying the total N of malts by 6.25. For S measurement, barley samples were analysed using a Vario Max CNS analyser (Elementar Analysensysteme GmbH, Hanau, Germany). Micro-malting and malt analysis Barley samples were malted on a micro-scale (using a 250 g barley sample) according to Kihara et al.2 a) Steeping – casting moisture was targeted to reach 43.0% for all samples. Since hydration was expected to be variable among samples, the steeping time was estimated for each sample preliminarily. Samples were intermittently put into the Automatic Micro-malting machine (Phoenix Biosystems, Adelaide, Australia) to allow them to be steeped for the required time to reach the desired moisture level. Steeping consisted of a repeated cycle of 5 h immersion and 7 h air rest (non-immersion period with over 95% humidity), both at 15 ◦ C. No aeration was applied in the steeping period. b) Germination – 15 ◦ C for 115 h with continuous water spraying. c) Kilning – 100% fresh air/0% recirculation for 13.5 h at 55 ◦ C; 100%/0% for 8 h at 65 ◦ C; 50%/50% for 3.5 h at 75 ◦ C; 50%/50% for 4 h at 83.5 ◦ C. Rootlets were removed prior to analysis. Malt moisture, wort colour (WC), boiled wort colour (BWC), extract (fine grind, 0.2 mm), SN, apparent attenuation limit (AAL), diastatic power (DP), Hartong index (VZ 45 ◦ C), viscosity, wort β-glucan (WBG) and friability were determined according to standard methods of the European Brewing Convention (EBC).15 SMM, DMSO and DMS were determined according to standard methods of the American Society of Brewing Chemists (ASBC).16 Barley β-glucan analysis For measurement of the β-glucan concentration in barley grain, 750 µL of water was added to 25 mg of ground barley and the sample was incubated in boiled 1465 M Nanamori et al. water for 60 min. After cooling, 750 µL of concentrated HCl was added and the sample was incubated in boiled water again for 10 min. After centrifugation at 18 500 × g for 15 min at 25 ◦ C the supernatant was collected and used for the measurement. The determination of β-glucan in the extract was achieved by detecting the fluorescence in the reaction between β-glucan and calcoflour using a flow injection analysis (FIA) system.17 Amino acid analysis Amino acids were extracted according to Kihara et al.17 with slight modification. Malt samples were milled and 25 mg of milled sample was mixed with 1 mL of cold water. The sample was then incubated at 5 ◦ C for 15 h. The extract was ultrafiltered and used for high-performance liquid chromatography (HPLC) analysis. The fluorescence reaction was achieved using Waters AccQ Fluor Reagent (6-aminoquinolyl-Nhydroxysuccinimidyl carabamate) and the concentration was determined by detecting the fluorescence at 395 nm. Enzyme extraction Malt protease was extracted using a buffer containing 10 mmol L−1 Tris-HCl (pH 6.8) and 10 mmol L−1 dithiothreitol (DTT). A 20 mg ground malt sample was incubated in the extraction buffer at 4 ◦ C overnight. After centrifugation at 18 500 × g for 10 min at 4 ◦ C the supernatant was collected as the crude extract and used for the assay according to Kihara et al.2 Malt α-amylase and barley β-amylase were extracted in the same way as malt protease except for the composition of the extraction buffer. The extraction buffer for malt α-amylase contained 50 mmol L−1 malate (pH 5.4), 50 mmol L−1 NaCl, 2 mmol L−1 CaCl2 and 0.77 mmol L−1 sodium azide. The extraction buffer for barley β-amylase contained 20 mmol L−1 sodium acetate (pH 5.5) and 10 mmol L−1 DTT. Enzyme activity Malt protease activity was measured using resorufinlabelled casein as the substrate (Roche Diagnostics GmbH, Mannheim, Germany).18 Casein solution (4 g L−1 , pH 5.0) containing 20 mmol L−1 succinic acid and 20 mmol L−1 CaCl2 was mixed with the crude extract. The mixture was incubated at 37 ◦ C for 4 h, then the reaction was stopped by adding 50 g L−1 trichloroacetic acid (TCA). After centrifugation at 18 500 × g for 10 min at 5 ◦ C the supernatant was collected and diluted by a factor of 2.5 with 0.5 mol L−1 Tris-HCl buffer (pH 8.8). Subsequently the absorbance at 574 nm was measured. For the measurement of malt α-amylase activity the reaction was started by mixing a solution containing p-nitrophenyl maltoheptaoside and α-glucosidase with the crude extract. Trisodium phosphate was added 1466 to stop the reaction after incubation at 37 ◦ C for 10 min, followed by measurement of the absorbance at 405 nm. Before the assay of barley β-amylase activity the crude extract was diluted by a factor of 500 with 50 mmol L−1 3-[N-Morpholino] propanesulfonic acid (MOPS) buffer. The reaction was started by mixing a solution containing p-nitrophenyl-α-maltoheptaoside and α-glucosidase with the diluted extract. Tris was added to stop the reaction after incubation at 37 ◦ C for 10 min, then the absorbance at 405 nm was measured. Statistical analysis The experimental data were first subjected to two-way analysis of variance (ANOVA). After confirmation that there was no significant interaction, the data were then subjected to one-way ANOVA for each cultivar. When a significant effect of N treatment was suggested, the difference was determined by the least significant difference (LSD) procedure (P < 0.05). RESULTS Grain yield and nutrient status in grains Grain yield in each treatment was increased by basal dressing and/or top-dressing of N fertiliser (Table 1). Basal dressing increased the yield of Sukai Golden efficiently, while the total amount of N fertiliser had a significant positive influence on the yield of Mikamo Golden. Nitrogen concentration in grains was effectively increased by top-dressing, indicating the importance of the timing of N application (Table 2). Basal dressing did not affect grain N concentration when top-dressing was applied in both cultivars. Sulfur concentration was also affected by top-dressing in Sukai Golden, and the change in S concentration corresponded with that in N concentration. However, the effect of N fertilisation was not statistically significant in Mikamo Golden. Malt quality Protein content in malts corresponded with N concentration in grains (Table 3). Nitrogen topdressing increased malt protein equally well in both cultivars. On the other hand, malt extract generally correlated negatively with malt protein. In this study, malt extract decreased as malt protein increased. Sukai Golden had slightly higher malt extract than Mikamo Golden. DP, which mainly reflects α-amylase and β-amylase activities, correlated positively with malt protein content. No significant difference in AAL within treatments was observed. SN, which includes amino acids and small peptides, was higher when malt protein was higher in all treatments. KI was significantly lower in the 30/0 treatment of Mikamo Golden, while there was no change in Sukai Golden. SN and KI were higher in Sukai Golden than in Mikamo Golden, indicating faster degradation of nitrogenous components of the former cultivar. Friability was decreased by top-dressing, the J Sci Food Agric 88:1464–1471 (2008) DOI: 10.1002/jsfa Effects of nitrogen application on Japanese barley cultivars Table 1. Grain yield (kg ha−1 ) in each N treatment N treatment Cultivar Sukai Golden Mikamo Golden 30/0 30/30 60/0 60/30 3.67 ± 0.31c 3.63 ± 0.39b 4.28 ± 0.35b 4.33 ± 0.30ab 4.80 ± 0.19b 4.38 ± 0.19ab 5.40 ± 0.11a 5.07 ± 0.15a Treatments: 30/0, 30 kg N ha−1 of basal dressing without top-dressing; 30/30, 30 kg N ha−1 of basal dressing with 30 kg N ha−1 of top-dressing; 60/0, 60 kg N ha−1 of basal dressing without top-dressing; 60/30, 60 kg N ha−1 of basal dressing with 30 kg N ha−1 of top-dressing. Values are mean ± SE of three replicates. The effect of N treatment was determined separately between cultivars by LSD (P < 0.05). Table 2. Nitrogen and sulfur concentrations (mg g−1 DW) in grain of each N treatment N treatment Cultivar/nutrient Sukai Golden Nitrogen Sulfur N/S ratio Mikamo Golden Nitrogen Sulfur N/S ratio 30/0 30/30 60/0 60/30 14.8 ± 0.3b 1.38 ± 0.02c 10.7 17.0 ± 0.1a 1.51 ± 0.03a 11.3 15.6 ± 0.1b 1.41 ± 0.02bc 11.0 16.8 ± 0.2a 1.46 ± 0.02ab 11.5 15.1 ± 0.4b 14.1 ± 0.02 10.6 17.4 ± 0.3a 1.50 ± 0.02 11.6 16.2 ± 0.5ab 1.48 ± 0.04 10.9 17.2 ± 0.3a 1.52 ± 0.04 11.3 Values are mean ± SE of three replicates. The effect of N treatment was determined separately between cultivars by LSD (P < 0.05). Table 3. Malt quality in each N treatment N treatment Cultivar/parameter Sukai Golden Moisture (%) WC (EBC) BWC (EBC) Extract (%) AAL (%) Hartong 45 ◦ C (%) DP (◦ WK) Protein (%) SN (%) KI Friability (%) WBG (mg L−1 ) Viscosity (mPa s) pH Mikamo Golden Moisture (%) WC (EBC) BWC (EBC) Extract (%) AAL (%) Hartong 45 ◦ C (%) DP (◦ WK) Protein (%) SN (%) KI Friability (%) WBG (mg L−1 ) Viscosity (mPa s) PH 30/0 30/30 60/0 60/30 4.5 ± 0.1a 3.8 ± 0.2ab 6.6 ± 0.5ab 84.6 ± 0.1a 86.5 ± 0.6 40.7 ± 2.1 242 ± 8a 8.8 ± 0.2c 0.757 ± 0.022c 53.6 ± 0.9 97.3 ± 0.1a 54 ± 1b 1.49 ± 0.01ab 6.04 ± 0.02ab 4.0 ± 0.2b 4.3 ± 0.3a 7.9 ± 0.6a 84.1 ± 0.2b 84.6 ± 0.4 42.9 ± 0.8 270 ± 6ab 10.4 ± 0.1a 0.873 ± 0.014a 52.7 ± 0.1 93.0 ± 1.8b 61 ± 1a 1.50 ± 0.00a 6.00 ± 0.01a 4.3 ± 0.0ab 3.6 ± 0.1b 6.2 ± 0.3b 84.5 ± 0.1ab 86.2 ± 0.9 38.8 ± 1.1 272 ± 7ab 9.4 ± 0.1b 0.786 ± 0.009bc 52.3 ± 0.8 98.9 ± 0.5a 52 ± 1c 1.49 ± 0.00b 6.05 ± 0.00b 4.5 ± 0.1a 3.6 ± 0.2b 6.1 ± 0.3b 84.3 ± 0.2ab 85.5 ± 0.5 38.5 ± 1.2 295 ± 24b 10.1 ± 0.2a 0.827 ± 0.013ab 51.2 ± 0.6 95.5 ± 1.4ab 50 ± 1d 1.48 ± 0.00b 6.05 ± 0.01b 4.5 ± 0.1 4.5 ± 0.1a 6.0 ± 0.2 83.9 ± 0.1a 84.1 ± 0.4 36.4 ± 0.3 252 ± 4c 9.0 ± 0.3b 0.679 ± 0.015b 47.2 ± 0.4a 90.2 ± 2.9a 90 ± 4b 1.51 ± 0.01 6.14 ± 0.01a 4.6 ± 0.1 4.4 ± 0.2ab 6.4 ± 0.2 82.7 ± 0.4b 83.7 ± 0.3 38.5 ± 0.4 296 ± 12ab 10.5 ± 0.3a 0.769 ± 0.019a 45.7 ± 0.5b 76.5 ± 2.9b 125 ± 6a 1.53 ± 0.01 6.10 ± 0.01b 4.4 ± 0.1 4.1 ± 0.1bc 6.0 ± 0.3 83.3 ± 0.2ab 83.4 ± 0.2 38.2 ± 1.5 268 ± 2bc 10.0 ± 0.3ab 0.729 ± 0.028ab 45.8 ± 0.4b 88.2 ± 1.5a 115 ± 5a 1.52 ± 0.01 6.11 ± 0.01b 4.6 ± 0.1 4.0 ± 0.1c 5.9 ± 0.2 82.8 ± 0.2b 84.0 ± 0.0 36.9 ± 0.1 303 ± 21a 10.6 ± 0.3a 0.771 ± 0.018a 45.6 ± 0.1b 78.1 ± 2.6b 116 ± 2a 1.51 ± 0.00 6.11 ± 0.01b Malt was analysed by EBC-recommended methods. Values are mean ± SE of three replicates. The effect of N treatment was determined separately between cultivars by LSD (P < 0.05). WC, wort colour; BWC, boiled wort colour; AAL, apparent attenuation limit; DP, diastatic power; SN, soluble nitrogen; KI, Kolbach index; WBG, wort β-glucan. J Sci Food Agric 88:1464–1471 (2008) DOI: 10.1002/jsfa 1467 M Nanamori et al. effect being more prominent in Mikamo Golden than in Sukai Golden. WBG was highest in the 30/30 treatment of both cultivars. DMSP concentration in malts SMM concentration was highest in the treatment with low N in both cultivars, with the difference being more obvious in Sukai Golden (Fig. 1(A)). SMM concentration in the 30/0 treatment of Sukai Golden was 1.8–2.1 times higher than that in other treatments. SMM is heat-labile and partly decomposes during kilning into DMS. Interestingly, DMS concentration in malt had an inverse relation with SMM concentration, so that DMS concentration in the 30/0 treatment was lower than that in any other 7.0 a SMM (mg kg-1) 6.0 a 5.0 ab b 4.0 c b 3.0 b bc 2.0 1.0 0.0 (B) Sukai Golden Mikamo Golden 8.0 a 7.0 ab ab DMSO (mg kg-1) 6.0 b 5.0 4.0 3.0 2.0 1.0 0.0 (C) Sukai Golden Mikamo Golden Amino acid concentration SMM, which may affect the final concentration of DMS, is synthesised from methionine. Concentrations of 19 amino acids, including methionine, in malts were measured, and Table 4 shows the ratio of each against protein concentration in malt. Amino acid concentration tended to be higher in malts with high protein content. However, there was no significant difference in methionine concentration among treatments. Enzyme activity Several types of protease are involved in the degradation of storage proteins in barley germination, with cysteine protease having the strongest influence on malt qualities.2 However, total protease activity has to be considered in the degradation of nitrogenous compounds. In the present study, protease activity in malts was not affected by N treatments, and there were no differences in activity between Sukai Golden and Mikamo Golden (Fig. 3). The activities of malt α-amylase and barley βamylase were not significantly influenced by N treatments (Fig. 4(A) and 4(B) respectively). However, there was an indication that these enzyme activities increased as protein content increased. 7.0 60 a 6.0 a a a 5.0 DMS (mg kg-1) β-Glucan concentration in grains To clarify the relationship between barley β-glucan concentration and malt modification, the β-glucan concentration in grains was determined (Fig. 2). Barley β-glucan concentration was stable despite the change in nitrogenous status in both cultivars. Sukai Golden had lower β-glucan concentration, around 72–76% of that in Mikamo Golden. ab 4.0 3.0 b ab b 2.0 1.0 0.0 Sukai Golden Beta-glucan (mg g-1 DW) (A) treatment (Fig. 1(C)). DMSO concentration in the 30/0 treatment of Sukai Golden was slightly higher than that in other treatments, but the difference was not significant (Fig. 1(B)). In Mikamo Golden, DMSO concentration in the 30/30 treatment was significantly higher than that in the 60/30 treatment. 50 40 30 20 10 Mikamo Golden 0 Figure 1. DMSP ((A) SMM, (B) DMSO, (C) DMS) concentrations in malt of each N treatment: 30/0, white bars; 30/30, diagonal bars; 60/0, dotted bars; 60/30, black bars. Values are mean ± SE of two replicates. The effect of N treatment was determined separately between cultivars by LSD (P < 0.05). 1468 Sukai Golden Mikamo Golden Figure 2. β-Glucan concentration in grain of each N treatment: 30/0, white bars; 30/30, diagonal bars; 60/0, dotted bars; 60/30, black bars. Values are mean ± SE of three replicates. J Sci Food Agric 88:1464–1471 (2008) DOI: 10.1002/jsfa Effects of nitrogen application on Japanese barley cultivars Table 4. Amino acid (AA) ratio against protein concentration (%) in malt of each N treatment N treatment Cultivar/AA Sukai Golden Gly Ala Val Leu Ile Ser Thr Lys Arg Asp Asn Glu Gln Cys Met Phe Tyr Pro His Mikamo Golden Gly Ala Val Leu Ile Ser Thr Lys Arg Asp Asn Glu Gln Cys Met Phe Tyr Pro His 30/0 30/30 60/0 60/30 0.22 ± 0.01 0.64 ± 0.06 0.86 ± 0.01a 0.98 ± 0.01 0.56 ± 0.01 0.45 ± 0.02 0.52 ± 0.03 0.58 ± 0.01a 0.95 ± 0.03 0.77 ± 0.05 1.47 ± 0.04 0.83 ± 0.03 1.98 ± 0.12 0.07 ± 0.01 0.23 ± 0.01 1.23 ± 0.01 0.79 ± 0.01 4.19 ± 0.27 0.39 ± 0.01 0.20 ± 0.01 0.54 ± 0.03 0.81 ± 0.02b 0.99 ± 0.02 0.55 ± 0.00 0.46 ± 0.02 0.46 ± 0.01 0.49 ± 0.02c 0.87 ± 0.01 0.69 ± 0.03 1.37 ± 0.11 0.72 ± 0.01 2.08 ± 0.07 0.05 ± 0.00 0.22 ± 0.01 1.18 ± 0.01 0.76 ± 0.01 4.18 ± 0.11 0.35 ± 0.01 0.21 ± 0.01 0.54 ± 0.01 0.86 ± 0.02a 1.01 ± 0.01 0.56 ± 0.00 0.48 ± 0.01 0.48 ± 0.01 0.56 ± 0.01ab 1.02 ± 0.04 0.70 ± 0.01 1.45 ± 0.12 0.77 ± 0.03 2.17 ± 0.02 0.06 ± 0.01 0.22 ± 0.00 1.23 ± 0.00 0.79 ± 0.01 4.72 ± 0.39 0.40 ± 0.02 0.22 ± 0.00 0.51 ± 0.03 0.83 ± 0.02ab 0.98 ± 0.02 0.55 ± 0.01 0.47 ± 0.01 0.45 ± 0.01 0.52 ± 0.01b 0.95 ± 0.04 0.65 ± 0.01 1.45 ± 0.06 0.77 ± 0.01 2.24 ± 0.06 0.06 ± 0.00 0.21 ± 0.00 1.21 ± 0.03 0.78 ± 0.01 4.46 ± 0.32 0.38 ± 0.01 0.21 ± 0.02 0.57 ± 0.01a 0.69 ± 0.09 0.75 ± 0.07 0.42 ± 0.05 0.37 ± 0.04 0.42 ± 0.01 0.47 ± 0.07 0.74 ± 0.06 0.67 ± 0.04 1.14 ± 0.10ab 0.65 ± 0.04 1.42 ± 0.14 0.03 ± 0.03 0.21 ± 0.01 0.84 ± 0.10 0.62 ± 0.07 3.81 ± 0.11ab 0.33 ± 0.03 0.21 ± 0.00 0.53 ± 0.02ab 0.73 ± 0.02 0.75 ± 0.03 0.44 ± 0.02 0.37 ± 0.01 0.42 ± 0.02 0.46 ± 0.02c 0.85 ± 0.08 0.64 ± 0.02 1.49 ± 0.06a 0.62 ± 0.02 1.51 ± 0.06 0.01 ± 0.01 0.19 ± 0.01 0.86 ± 0.02 0.62 ± 0.02 3.72 ± 0.07ab 0.35 ± 0.01 0.17 ± 0.01 0.45 ± 0.00b 0.64 ± 0.02 0.71 ± 0.01 0.40 ± 0.00 0.33 ± 0.00 0.37 ± 0.00 0.41 ± 0.02 0.68 ± 0.04 0.59 ± 0.04 1.02 ± 0.13b 0.56 ± 0.03 1.36 ± 0.02 0.05 ± 0.00 0.19 ± 0.01 0.82 ± 0.01 0.60 ± 0.01 3.09 ± 0.23b 0.31 ± 0.02 0.20 ± 0.01 0.51 ± 0.05ab 0.68 ± 0.02 0.72 ± 0.03 0.42 ± 0.02 0.36 ± 0.01 0.40 ± 0.03 0.43 ± 0.01 0.83 ± 0.05 0.60 ± 0.03 1.38 ± 0.17ab 0.61 ± 0.04 1.55 ± 0.07 0.03 ± 0.01 0.18 ± 0.01 0.82 ± 0.03 0.61 ± 0.02 3.93 ± 0.42a 0.33 ± 0.01 Values are mean ± SE of three replicates. The effect of N treatment was determined separately between cultivars by LSD (P < 0.05). DISCUSSION Nitrogen basal dressing increased grain yields, more prominently in Sukai Golden than in Mikamo Golden (Table 1). However, N and S concentrations in grains were not affected significantly by basal dressing in both cultivars (Table 2). Nitrogen concentration in grains was increased efficiently by top-dressing as reported by Chen et al.13 It was shown that the timing of N application is important for the nutrient status in grain (Table 2). Nitrogen applied as basal dressing was utilised mainly during the vegetative stage for plant growth, while that applied as top-dressing was translocated into grain efficiently. Grain S concentration correlated with grain N concentration as reported by Zhao et al.19 Therefore translocation of N may have a positive relation with S translocation in barley. J Sci Food Agric 88:1464–1471 (2008) DOI: 10.1002/jsfa The effects of cultivar and N fertilisation on malt qualities have been studied. Malt extract declined almost linearly with increasing N application.20 Agu21 discussed the relationships between barley N level and wort properties and showed that barley with higher N content had a slower rate of endosperm modification and lower extract yield, while barley with lower N content had a faster rate of endosperm modification and higher extract yield. In the present study, malt extract declined as protein content in barley increased, and the decline in Mikamo Golden was more dramatic than that in Sukai Golden, indicating that Sukai Golden is more manageable in terms of malt quality (Table 3). Both barley β-amylase activity and malt α-amylase activity tended to correlate with protein content in 1469 M Nanamori et al. Protease activity (pmol min-1 mg-1 DW) 15 12 9 6 3 0 Sukai Golden Mikamo Golden (A) 250 Alpha-amylase activity (nmol min-1 mg-1 DW) Figure 3. Protease activity in malt of each N treatment: 30/0, white bars; 30/30, diagonal bars; 60/0, dotted bars; 60/30, black bars. Protease activity was defined as the activity degrading 1 pmol resorufin-labelled casein min−1 . Values are mean ± SE of three replicates. 200 150 100 50 (B) 1.5 Beta-amylase activity (nmol min-1 mg-1 DW) 0 1.2 Sukai Golden Mikamo Golden Sukai Golden Mikamo Golden 0.9 0.6 0.3 0.0 Figure 4. Activities of starch-hydrolysing enzymes ((A) malt α-amylase, (B) barley β-amylase) of each N treatment: 30/0, white bars; 30/30, diagonal bars; 60/0, dotted bars; 60/30, black bars. Both enzyme activities were defined as the activity degrading 1 nmol substrate min−1 . Values are mean ± SE of three replicates. both cultivars, although there were no statistically significant differences (Fig. 4). DP indicates the level of starch-degrading ability, which refers to the combined activities of β-amylase, α-amylase, limit dextrinase and α-glucosidase. It has been shown that β-amylase is the only DP enzyme to correlate with DP.22 Our results support the possibility of the contribution of both β-amylase and α-amylase to DP. 1470 However, β-amylase might contribute more than αamylase to the changes in malt DP, because α-amylase activity showed larger fluctuations, suggesting the ratelimiting role of β-amylase. Leach et al.14 showed that barley with high protein content tended to have higher SN level and lower KI. Our results show that high protein content induces high SN in wort (Table 3). However, KI in Sukai Golden was not affected significantly by N treatment. This suggests that the barley protein content is critical for these cultivars in the aspect of N nutrient for yeasts. Amino acid content also tended to be higher in high-protein barley, but this did not affect the fermentability (Tables 3 and 4). Thus these barley cultivars resulted in less sensitivity of fermentability to N fertilisation even though the extract declined with increasing protein content. β-Glucan and protein content in barley is affected by environmental factors, which contributed the largest component of the variation in those compounds.23 Wang et al.24 showed that there are cultivar and environmental variations of β-glucan content in both malt and grain, but they are much higher in malt than in grain. Barley β-glucan concentration was lower in Sukai Golden than in Mikamo Golden (Fig. 2). Nitrogen treatment did not affect the levels of β-glucan in either cultivar. These results indicate that the difference in modification between N treatments is not due to the change in barley β-glucan concentration. On the other hand, MolinaCano et al.25 reported that a mutant that had low β-glucan content improved its malting quality and extract yield by enhancing germination speed and hence starch degradation. Rapid germination and good modification in Sukai Golden may be due to lower βglucan concentration, which enables easier enzymatic attacks on the endosperm. WBG was significantly affected by N treatments (Table 3). Although β-glucanase contributes to βglucan degradation, the factors influencing the differences in WBG were not revealed in this study, because the β-glucanase activity fluctuated greatly (data not shown). Friability decreased with increasing protein content in both cultivars, the effect being larger in Mikamo Golden than in Sukai Golden. Although it is obvious that higher protein content induces lower friability in malts, the marked decrease in friability in Mikamo Golden might be indicative of uneven modification, although the reason is unclear. SMM and DMSO are precursors of DMS. The former is thermally decomposed into DMS during the kilning process and wort boiling. Although the majority of DMS generated in the malting process is released in wort boiling, DMS generated after wort boiling might remain in the finished beer. DMSO is also derived from SMM and can be reduced to DMS during fermentation by yeasts. Therefore SMM concentration in malt is believed to be a key factor for DMS production. Our results show that low-N treatment increases SMM concentration in J Sci Food Agric 88:1464–1471 (2008) DOI: 10.1002/jsfa Effects of nitrogen application on Japanese barley cultivars malt (Fig. 1). The response to N treatment differed in the two cultivars. Interestingly, there was an inverse relationship between SMM and DMS in malt. Dethier26 showed that SMM and methionine increased in parallel in germinating barley, although there was a lag period between the two compounds. Protein is degraded in starchy endosperm, releasing methionine, which diffuses into the germ where SMM can be synthesised. Dethier26 suggested that SMM synthesis is strongly dependent on the methionine content in the corresponding fraction. However, neither methionine concentration in malt nor total DMSP (SMM, DMSO and DMS) differed among N treatments in our study (Table 4, Fig. 1). As N and S concentrations in grain increased with N top-dressing, the N/S ratio in grain also increased (Table 2). This indicates that the relative amount of S in grain was higher when N top-dressing was not applied. Zhao et al.19 showed that S application to S-deficient malting barley crop increased DMSP in the malt and hence could have a significant impact on the flavour of the beer. Our results suggest that barley can maintain a homeostasis of SMM synthesis from methionine in germination despite the difference in N status in grain. It was interesting that SMM concentration in the malt from barley with low N and S content was high. Yang and Schwarz27 suggested that steeping and germination temperature influences the synthesis and levels of SMM in green malt but that kilning temperature determines the final levels of SMM and DMSO in malt. The degradation of SMM in low-N malt may not be enhanced during kilning. Further investigation is necessary into detailed metabolic profiles in barley germination and/or SMM degradation in kilning to elucidate factors influencing the DMSP concentration in malt. ACKNOWLEDGEMENTS We are grateful for the cooperation of Mr Kiyoshi Takoi and Ms Kimiko Otake in the analysis of barley and malt qualities. We also thank Dr Isao Kishinami for advice in writing this paper. REFERENCES 1 Jones BL, Marinac LA and Fontanini D, Quantitative study of the formation of endoproteolytic activities during malting and their stabilities to kilning. J Agric Food Chem 48:3898–3905 (2000). 2 Kihara M, Saito W, Okada Y, Kaneko T, Asakura T and Ito K, Relationship between protease activity during malting and malt quality. J Inst Brew 108:371–376 (2002). 3 Enari TM, Proteinases and peptidases of malt and their influence on wort composition and beer quality. Cerevisia 11:19–28 (1986). 4 Clapperton JF, Simple peptides of wort and beer. J Inst Brew 88:244–252 (1971). 5 Edney MJ and Langrell DE, Effect of fermentable sugars and amino acids on fermentability of malts made from four barley varieties. Tech Q MBAA 42:101–106 (2005). 6 Lekkas C, Stewart GG, Hill A, Taidi B and Hodgson J, The importance of free amino nitrogen in wort and beer. Tech Q MBAA 42:113–116 (2005). J Sci Food Agric 88:1464–1471 (2008) DOI: 10.1002/jsfa 7 Lekkas C, Stewart CG, Hill AE, Taidi B and Hodgson J, Elucidation of the role of nitrogenous wort components in yeast fermentation. J Inst Brew 113:3–8 (2007). 8 Bourgis F, Roje S, Nucio ML, Fisher DB, Tarczynski MC, Li C, et al, S-Methylmethionine plays a major role in phloem sulfur transport and is synthesized by a novel type of methyltransferase. Plant Cell 11:1485–1498 (1999). 9 Gyetvai E, Racz I, Lasztity D, Szalai G, Janda T, Marton L, et al, Effect of S-methylmethionine as a protective compound on the metabolism of agricultural plants at low temperature. Acta Biol Szeged 46:95–96. 10 Manegus F, Lilliu I, Brambilla I, Bonfà M and Scaglioni L, Unusual accumulation of S-methylmethionine in aerobicetiolated and in anoxic rice seedlings: an 1 H-NMR study. J Plant Physiol 161:725–732 (2004). 11 Pimenta MJ, Kaneta T, Larondelle Y, Dohmae N and Kamiya Y, S-Adenosyl-L-methionine:L-methionine S-methyltransferase from germinating barley. Purification and localization. Plant Physiol 118:431–438 (1998). 12 Annes BJ and Bamforth CW, Dimethyl sulfide – a review. J Inst Brew 88:244–252 (1982). 13 Chen JX, Dai F, Wei K and Zhang GP, Relationship between malt qualities and β-amylase activity and protein content as affected by timing of nitrogen fertilizer application. J Zhejiang Univ Sci B 7:79–84 (2006). 14 Leach R, Li Y, Edney M, Izydorczyk M, Egi A and Sawatzky K, Effects of barley protein content on barley endosperm texture, processing condition requirements, and malt and beer quality. Tech Q MBAA 39:191–202 (2002). 15 EBC, Analytica-EBC. European Brewery Convention, Brussels (1998). 16 ASBC, Methods of Analysis of the American Society of Brewing Chemists (8th rev. edn). American Society of Brewing Chemists, St Paul, MN (1999). 17 Kihara M, Okada Y, Iimure T and Ito K, Accumulation and degradation of two functional constituents, GABA and betaglucan, and their varietal differences in germinated barley grains. Breed Sci 57:85–89 (2007). 18 Ahokas H and Naskali L, Variation of α-amylase, β-amylase, β-glucanase, pullulanase, proteinase and chitanase activity in germinated samples of the wild progenitor of barley. J Inst Brew 96:27–31 (1990). 19 Zhao FJ, Fortune S, Barbosa VL, McGrath SP, Stobart R, Self M, et al, Sulphur requirement of malting barley: effects on yield and quality and diagnosis of sulphur deficiency. HGCA Project Rep 369 (2005). 20 Eagles HA, Bedggood AG, Panozzo JF and Martin P, Cultivar and environmental effects on malting quality in barley. Aust J Agric Res 46:831–844 (1995). 21 Agu RC, Some relationships between malted barleys of different nitrogen levels and the wort properties. J Inst Brew 109:106–109 (2003). 22 Evans DE, Lance RCM, Eglinton JK, Logue SJ and Barr AR, The influence of beta-amylase isoform pattern on betaamylase activity in barley and malt. Proc 45th Aust Cereal Chemistry Conf, pp. 357–368 (1995). 23 Zhang GP, Chen JX, Wang JM and Ding SR, Cultivar and environmental effects on β-glucan and protein content in malting barley. J Cereal Sci 34:295–301 (2001). 24 Wang J, Zhan G, Chen J and Wu F, The changes of β-glucan content and β-glucanase activity in barley before and after malting and their relationships to malt qualities. Food Chem 86:223–228 (2004). 25 Molina-Cano JL, Roca de Togores F, Royo C and Perez A, Fast-germinating low β-glucan mutants induced in barley with improved malting quality and yield. Theor Appl Genet 78:748–754 (1989). 26 Dethier M, In vivo and in vitro investigations of the synthesis of S-methylmethionine during barley germination. J Am Soc Brew Chem 49:31–37 (1991). 27 Yang B and Schwarz P, Factors involved in the formation of two precursors of dimethylsulfide during malting. J Am Soc Brew Chem 56:85–92 (1998). 1471 Journal of the Science of Food and Agriculture J Sci Food Agric 88:1472–1481 (2008) Basis for the new challenges of growing broccoli for health in hydroponics Diego A Moreno,1∗ Carmen López-Berenguer,1,2 M. Carmen Martı́nez-Ballesta,2 Micaela Carvajal2 and Cristina Garcı́a-Viguera1 1 Department of Food Science and Technology, CEBAS-CSIC, Campus Universitario de Espinardo, PO Box 164, Espinardo 30100-Murcia, Spain 2 Department of Plant Nutrition, CEBAS-CSIC, Campus Universitario de Espinardo, PO Box 164, Espinardo 30100-Murcia, Spain Abstract BACKGROUND: Variations in the contents of phytochemicals with biological activity in broccoli could originate as a result of genetic and environmental factors. An understanding of the effects of growth conditions on the bioactive compounds in broccoli is essential for improving its quality and nutritive value. Using salinity (40 mmol L−1 NaCl), and foliar sprayed compounds (methionine, tryptophan and chitosan) as different stress conditions, broccoli developed in soilless culture in the greenhouse was analysed for biologically active phytochemicals (glucosinolates, caffeoyl-quinic, ferulic and sinapic derivatives and vitamin C). RESULTS: The application of elicitors during head formation could be beneficial for the enrichment in phytochemicals in broccoli. Management practices for increasing a given phytochemical (e.g., glucoraphanin or glucobrassicin) may be related to a decreased level of natural antioxidants (hydroxycinnamic acids). Growing broccoli hydroponically in the greenhouse in winter (Mediterranean climate) needs the supporting treatment of abiotic stress during development (i.e., NaCl, elicitors). CONCLUSION: The use of hydroponic growth conditions for broccoli and the application of stress factors (elicitors) at head induction and during development may serve the purpose of enhancing its nutritional quality to deliver a health-promoting food.  2008 Society of Chemical Industry Keywords: Brassica oleracea var. italica; hydroxycinnamic acids; flavonoids; glucosinolates; vitamin C; soilless culture INTRODUCTION Today, consumers are more proactive and conscious in managing their health and the prevention of dietrelated diseases. Many of these diseases are at epidemic levels, making the prevention of these lifestyle diseases attractive markets to exploit for both the food and pharmaceutical industries. The consumption of diets containing five to ten servings of fruits and vegetables daily is the foundation for cancer prevention, and combinations of tomato and broccoli in the Dunning R3327-H prostate adenocarcinoma model was more effective at slowing tumour growth than either tomato or broccoli alone. This supports public health recommendations to increase the intake of a variety of health-giving fruits and vegetables.1 Broccoli, the worldwide-known immature flower vegetable of Brassicaceae (Brassica oleracea L. (Italica group)), is also well recognized as a healthpromoting vegetable owing to its high content of beneficial biologically active compounds. Numerous epidemiological studies indicate that brassicas in general, and broccoli in particular, have potential for chemoprevention of degenerative diseases and certain types of cancer since they are rich sources of glucosinolates and dietary natural antioxidants: vitamins, flavonoids and hydroxycinnamic acids.2,3 Variations in the contents of phytochemicals with biological activity in broccoli could originate from genetic and environmental factors: variability of accessions, cultivars, organ, inflorescence development, temperature and radiation, growth system, fertilization practices, post-harvest storage and processing.4 – 11 In this respect, water and dissolved salts are essential to plant growth, but water reuse and high evaporation rates in arid or semi-arid regions such as southeastern Spain (Murcia) concentrate the salts and salinization occurs. Broccoli is considered moderately sensitive/tolerant to salinity.12 Looking into the glucosinolate composition of broccoli leaves and inflorescences, soilless-grown broccoli treated with ∗ Correspondence to: Diego A Moreno, Department of Food Science and Technology, CEBAS-CSIC, Campus Universitario de Espinardo, PO Box 164, Espinardo 30100-Murcia, Spain E-mail: dmoreno@cebas.csic.es (Received 27 June 2007; revised version received 30 January 2008; accepted 4 February 2008) Published online 23 April 2008; DOI: 10.1002/jsfa.3244  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Phytonutrients in hydroponically grown broccoli 40 mmol NaCl in outdoor cultivation showed higher glucosinolate content than untreated plants.13 The lack of reproducibility of phytochemical analysis or activities is a major obstacle when plants are regrown in the field, resampled and re-extracted over the years. The biochemical profiles of plants harvested at different times and locations vary greatly and whole plant elicitation may also increases the amounts of natural products widely used in different areas of research. However, hydroponic cultivation is fast, simple and applicable to a great majority of plant species. Chemical composition and bioactivity could be readily reproduced if plants are regrown and re-elicited under standard greenhouse conditions.13 – 15 It is well known that different stresses, location climates, microenvironments and physical and chemical stimuli (often called elicitors) qualitatively and quantitatively alter the content of bioactive secondary metabolites, and whole-plant elicitation increases the amounts of bioactive compounds in foods of plant origin.14 – 16 Clearly, genotypic variations observed in different plant species imposed both substantial variation and a genetic limit on the production of bioactive compounds. However, elicitation may be able to increase the production of some bioactive compounds up to the genetic limit.15 To date, there are almost no reports on the effect of methionine or tryptophan elicitation or fertilization on glucosinolate-producing vegetable crops. Scheuner et al.17 showed that methionine foliar fertilization increased the glucosinolate content in broccoli, but not in radish hypocotyls of greenhouse-grown plants. Chitosan has been reported as stimulating the growth and yield of soybean sprouts without adverse effects on vitamin C or their postharvest characteristics.18 Nonetheless, chitosan did not improve the production of glucotrapeolin hydrolysis products and the recorded levels were very close to control values in Farsetia aegyptia cultures.19 Studies on broccoli have been carried out either in open-air cultivation or in the growth chamber, but to date little is known about growing broccoli hydroponically in the greenhouse for the production of bioactive compounds of interest for human health. The objective of this investigation was to determine the effects of different stress factors under soilless hydroponic growth conditions on the contents of glucosinolates, phenolic compounds (flavonoids and hydroxycinnamic acids) and vitamin C in broccoli. An understanding of the effects of growth conditions on bioactive compounds in broccoli is essential for improving its quality and nutritive value. MATERIALS AND METHODS Conditions of plant growth The object of our investigations was ‘Marathon’ broccoli (Brassica oleracea (Plenck) Italica Group, cv. ‘Marathon’). The plants were cultivated in the autumn–winter season (October 2006 to February 2007) in the greenhouse of the CEBAS-CSIC located J Sci Food Agric 88:1472–1481 (2008) DOI: 10.1002/jsfa in Espinardo (Murcia, Spain) under a semi-arid Mediterranean climate. Broccoli and design of experiments Broccoli seeds obtained from Ramiro Arnedo SA (Murcia, Spain), were pre-hydrated with aerated, deionized water for 2 h and germinated in vermiculite, at 28 ◦ C in an incubator, for 2 days. They were then transferred to a controlled-environment growth chamber with a 16 h light–8 h dark cycle, and air temperatures of 25 and 20 ◦ C, respectively. The relative humidity (RH) was 60% (day) and 80% (night) and photosynthetically active radiation (PAR) was 400 µmol m−2 s−1 , provided by a combination of fluorescent tubes (Philips TLD 36W/83, NY, USA and Sylvania F36W/GRO Munich, Germany); metal halide lamps (Osram HGI.T400W, Munich, Germany). After 5 days, the seedlings were placed in 15 L containers with continuously aerated Hoagland nutrient solution (published elsewhere), replaced completely every week.12,13 Broccoli (49-day-old) plants were transplanted to the greenhouse (day ‘0’ after transplanting; 0 DAT). The experiments were conducted under a non-controlled environment in an aluminiumframed greenhouse with polycarbonate panels in a single gable structure. The humidity achieved in the greenhouse averaged 65%/85% (day/night) and the air temperature 16/9 ◦ C (Table 1). The greenhouse was vented when the temperature exceeded the norm. Daily mean temperature and relative humidity were calculated from measurements taken every 10 min using dataloggers (AFORA SA, Barloworld Scientific, Murcia, Spain). A total of 25 ‘Marathon’ broccoli plants were placed in a randomized design, using five plants per treatment, with each plant being grown in a perlite-filled 15 L container, spaced from each other, resulting in a density of 4 plants m−2 . All plants were grown under the same conditions and irrigated with complete Hoagland solution twice a week under natural light conditions, until 14 DAT (Table 2). At that moment, the application of 40 mmol NaCl in the nutrient solution started as treatment T1 (five plants), chosen because the previous studies, provided by López-Berenguer et al.,12,13 had shown that 40 mmol NaCl increased glucosinolate levels in leaves (metabolic active leaves) and inflorescences (commercial size heads) of broccoli. The untreated control (five plants) and remaining plants did not show any symptom of deficiency or toxicity. Plants being treated with NaCl were also randomly subdivided into groups of five and subjected to the different stress conditions (also called elicitors) once approximately 30% of the plants reached head induction (0.3–0.5 cm head arc diameter; 52–56 DAT). The plants were sprayed with 40 mL of elicitor dissolved in 0.04% ethanol. A full description of the treatments is given in Table 2. At harvest (75 DAT) all the plants were collected. Plants were separated into three parts: leaves 1473 DA Moreno et al. Table 1. Experimental conditions in the greenhouse (day/night) for growing ‘Marathon’ broccoli hydroponically Plant age (DAT)a (◦ C)b Air temperature Relative humidity (%)b a b 0 1–14 15–27 29–42 43–56 57–70 70–75 16.3/11.7 45.2/82.6 15.4/9.4 64.5/81.3 13.4/7.5 66.6/82.2 16.7/8.7 66.2/93.2 15.7/9.0 68.1/87.1 15.8/9.1 83.1/90.1 21.6/12.4 62.2/82.1 Days after transplanting (DAT) to greenhouse; 7-week-old plants = 0 DAT. Average values of the time period. Table 2. Description of treatments Key Control T1 T1 +E1 T1 +E2 T1 +E3 Treatment DATa NaCl treatedb (DAT) Elicitor treatedc (DAT) Complete Hoagland’s nutrient solution 40 mmol NaCl added to nutrient solution 200 mmol DL-methionine (Alfa Aesar GmbH & Co. KG, Kralsruhe, Germany) 200 mmol DL-tryptophan (Alfa Aesar GmbH & Co. KG, Kralsruhe, Germany) 1 g L−1 chitosan (from crab shells; Sigma-Aldrich Quı́mica SA, Tres Cantos, Madrid) 0–75 0–14 0–14 15–75 15–75 52–56 0–14 15–75 52–56 0–14 15–75 52–56 a Days after transplanting (DAT) to greenhouse; 7-week-old plants = 0 DAT. 40 mmol NaCl added to Hoagland’s complete nutrient solution. c Broccoli plants were sprayed at 11:00 a.m. and at 15–20 cm distance, to minimize differences due to daily fluctuations (solution in 0.04% ethanol). b (leaf blades and petioles), stalks/stems and heads (inflorescences). For analytical purposes, the sampled organs of each treatment were cut into pieces and mixed thoroughly, to be again separated into five well-mixed replicates per treatment. Fresh weight was determined. The plant material was then flash frozen using liquid nitrogen and kept at −80 ◦ C and dried in a freeze-drier Alpha (Type 1–4, Christ, Osterode am Harz, Germany). Dry weight was then determined and plant material was ground to a fine powder and stored at −20 ◦ C for further analysis. Extraction and determination of phenolic compounds Freeze-dried powder samples (1 g) were homogenized with 25 mL of 70% methanol three times. The homogenates were filtered through a cheesecloth and kept in ice. The homogenates were centrifuged (4000 × g, 5 min, 4 ◦ C) and the supernatants were evaporated under vacuum at 30 ◦ C to approximately 1 mL and diluted to 2 mL with water. The supernatants were filtered through a 0.45 µm Millex-HV filter (Millipore, Bedford, MA, USA). The extracted samples (20 µL) were analysed on a Waters highperformance liquid chromatography (HPLC) system (Waters Cromatografı́a SA, Barcelona, Spain) consisting of a W600E multi-solvent delivery system, inline degasser, W717plus autosampler and a W2996 photodiode array detector, using a Luna C18 column (25 × 0.46 cm, 5 µm particle size; Phenomenex, Macclesfield, UK) with a security guard C18-ODS (4 × 3.0 mm) cartridge system (Phenomenex). The mobile phase was a mixture of 1 mL L−1 TFA (A) and acetonitrile/TFA (99.9:0.1, v:v) (B). Phenolic compounds were eluted off the column in 35 min. The 1474 flow rate was 1 mL min−1 in a linear gradient starting with 0% B 0–5 min, reaching 17% B in 15 min, 17% B at 17 min, 25% B at 22 min, 35% B at 30 min and 50% B at 35 min. Chromatograms were recorded at 320 and 360 nm.20,21 Caffeoyl-quinic derivatives were quantified as chlorogenic acid (5-caffeoyl-quinic acid, Sigma, St Louis, MO, USA), flavonoids as quercetin 3-rutinoside (Sigma) and sinapic acid and ferulic derivatives as sinapinic acid (Sigma). Extraction and determination of vitamin C Ascorbic (AA) and dehydroascorbic (DHAA) acid contents were determined as described by Vallejo et al.10,21 Briefly, 200 mg of freeze-dried sample was homogenized in a vortex stirrer for 20 s with 10 mL of extractant solution consisting of MeOH–H2 O (5:95) plus citric acid 2.1%, ethylenediaminetetraacetic acid (EDTA) 0.05% and NaF 0.01%; the homogenate was filtered through a cheesecloth and the pH adjusted to 2.2–2.4 by addition of 6 mol HCl. The extract was centrifuged (3.600 × g for 15 min at 4 ◦ C) and the supernatant was recovered, filtered through a C18 Sep-Pack cartridge (Waters, Milford, MA, USA), previously activated with 10 mL of methanol followed by the same volume of water and then the same volume of air, and filtered through a 0.45 µm polyethersulfone filter (MillexHV, Millipore). HPLC analysis of vitamin C (AA + DHAA) was achieved after derivatization of DHAA into the fluorophore 3-(1,2-dihydroxyethyl)furo(3,4b) quinoxaline-1-one (DFQ) with fresh daily prepared 1,2-ortophenylenediamine (OPDA). OPDA solution was added to the water-soluble fraction eluted from a C18 solid-phase extraction cartridge Sep-Pak (1:3, v:v). Samples were incubated for 37 min at room J Sci Food Agric 88:1472–1481 (2008) DOI: 10.1002/jsfa Phytonutrients in hydroponically grown broccoli temperature in the dark, and 20 µL analysed with a Merck-Hitachi (Tokyo, Japan) HPLC, equipped with an L-4000 UV detector and an L-6000 pump. Separations of DFQ and AA were achieved on a Kromasil 100 C18 column (25 × 0.4 cm; 5 µm particle size; Tecnokroma, Barcelona, Spain). The mobile phase was methanol–water (5:95, v:v) containing 5 mmol cetrimide and 50 mmol potassium dihydrogen phosphate at pH 4.5. The flow rate was 0.9 mL min−1 ; the detector wavelength was initially set at 348 nm, and after elution of DFQ was manually shifted to 261 nm for AA detection. Extraction and determination of glucosinolates We followed the procedure as fully described in Mart ı́nez-Sánchez et al.22 for extraction of intact glucosinolates.3 Briefly, freeze-dried samples (50 mg) were extracted with 1.5 mL of 70% MeOH, and placed in a sonicator bath for 10 min to improve the methanol extraction. The mixture was heated at 70 ◦ C for 30 min with a heating bath and shaking every 5 min with a vortex stirrer, and centrifuged (30 min, 17 500 × g, 4 ◦ C). Supernatants were collected and methanol completely removed using a rotary evaporator; the obtained dry material was redissolved in 1 mL ultrapure water and filtrated through 0.45 µm Millex-HV filter. Each sample (20 µL) was analysed in a Waters HPLC system (Waters Cromatografı́a SA, Barcelona, Spain) under the same conditions mentioned above for polyphenolic compounds. Chromatograms were recorded at 227 nm. Samples were identified using the previously described intact glucosinolate LC-MS method and quantified by HPLC–diode array detection (HPLCDAD) using sinigrin (sinigrin monohydrate from Sinapis nigra, Phytoplan Diehm & Neuberger GmbH, Heidelberg, Germany) as standard. Statistical analysis All data were subjected to analysis of variance (ANOVA) using Statgraphics version 7.0 software (Manugistics, Inc.). The data shown are mean values and the significance of the differences was compared using a multiple comparison test at LSD P < 0.05 probability level (Duncan’s multiple range test). RESULTS AND DISCUSSION Broccoli parameters at harvest The harvested inflorescences (commercial size heads) were significantly heavier in the T1 + E2 treatment, surpassing the inflorescences of the control plants by 25% in fresh mass (Table 3). The T1 and T1 + E1 were intermediate treatments between T1 + E2 and the control, only 12% and 8% higher, respectively. The T1 + T3 inflorescences did not differ from the untreated control. The fresh weight of the young fully expanded leaves (metabolically active leaves) showed a similar result, with T1 + E2 and T1 + E1 being J Sci Food Agric 88:1472–1481 (2008) DOI: 10.1002/jsfa Table 3. Biomass parameters (g per plant) of greenhouse-grown ‘Marathon’ broccoli at harvest Treatment Broccoli head Leaves Control T1 T1 +E1 T1 +E2 T1 +E3 165.79ba 188.70ab 183.33ab 211.77a 168.89b 324.37c 352.14bc 427.36ab 483.48a 341.11bc ANOVA P-value LSD (P < 0.05) P < 0.05 26.05 P < 0.05 101.57 a Means (n = 5) within a column followed by the same lower-case letter are not significantly different at P < 0.05 according to Duncan’s multiple range test. 49% and 32% higher than the control, respectively (Table 3). Taking into account that the collected inflorescences were all of a very similar size (120–135 mm diameter in average), changes in fresh mass could be related more to density of the inflorescences. The application of abiotic stress through 40 mmol L−1 NaCl in the nutrient solution (T1) and the additional application of methionine (T1 + E1) or tryptophan (T1 + T2) would then prove positive in some way for the inflorescence biomass of broccoli grown hydroponically; this fact would be of interest in the event that a higher content of phytochemicals could be found in such inflorescences. On the contrary, the application of Chitosan (T1 + E3), showed values similar to the control. Hydroxycinnamic acid derivatives and flavonoids Hydroxycinnamoyl derivatives were identified by their chromatographic behaviour and UV spectra, using HPLC-MS and chromatographic comparisons with authentic markers.20,21 The pattern found in broccoli was similar to that previously described by other authors.8 A number of individual flavonoids (10–15, depending on the treatment) were detected but not fully identified, mainly quercetin and kaempferol glycosides, in agreement with previous reports on broccoli.20,21 The total flavonoid contents in broccoli inflorescences were significantly affected by the treatments (Table 4), T1 + E3 being the highest (by 52%) against the control. The caffeoyl-quinic derivatives were also higher in T1 + E3 than in the control, with intermediate values for the rest of the treatments. The same was found for the sinapic and ferulic derivatives, with the majority of the compounds also being higher in T1 + T3 and T1 + E1 than in the control. The 1,2diferuloylgentiobiose (3) was similar in the control and T1 + E1, and higher than the remaining treatments. On average, total phenolic contents reflected the results of the individual compounds (Fig. 1), and in T1 + E3 (chitosan-sprayed salt-stressed broccoli) the total phenolic content was improved by 44%, followed by T1 + E1 (methionine-sprayed NaCl-treated plants), a 39% higher than the inflorescences in the 1475 1476 control. The results of polyphenolics in broccoli were probably not related to a dilution effect, even though T1 + E3 inflorescences were smaller (Table 3). The correlation coefficients between individual and total phenolics and the weight of inflorescences were not significant (P > 0.1; data not shown). Thus, the application of elicitors (methionine and chitosan) during head formation could be beneficial for the enrichment in phytochemicals in broccoli grown hydroponically, and the effect was additional to the salt-induced stress in these plants, since T1 + E1, T1 + E2, and T1 + T3 were all significantly higher (total phenols; Fig. 1) than T1 and the untreated control. When looking at the effects on leaves – the metabolic factory of the plant, but in terms of phytochemical farming, treated as a byproduct of the broccoli agrifood activity – the hydroxycinnamic acids in young fully expanded leaves (Table 5) were also significantly affected by treatments. The total content of flavonoids was much higher than in the inflorescences (Table 4), but in this case the control, salt-stressed T1 and elicited T1 + E1 leaves were similar in content while the leaves of T1 + E2- and T1 + E3-sprayed plants were surpassed by the control. The trend was very similar for the caffeoyl-quinic derivatives and sinapic and ferulic derivatives (Table 5). Figure 2 shows the total content of phenolics in broccoli leaves, where T1 + E2 and T1 + T3 were surpassed by the control by 41% and 25%, respectively. The flavonoids in greenhouse-grown broccoli are stated to be present at lower levels than in field cultures.23 Broccoli produced in late summer and early autumn field crops in different parts of Europe presented different flavonol contents, ranging from 1.5–6.5 mg 100 g−1 fresh weight8 to 5.7–9.6 mg 100 g−1 fresh weight (‘Marathon’ broccoli).4 Growing broccoli hydroponically in the greenhouse during the autumn/winter in southeastern Spain, we found that total flavonoid contents were higher than previously 60 Inflorescence Total phenols (mg 100 g-1 of fresh weight) a C1 (neochlorogenic acid) and C2 (chlorogenic acid); sinapic and ferulic derivatives: 1 (1,2-disnapoylgentiobiose); 2 (1-sinapoyl-2-feruloylgentiobiose); 3 (1,2-diferuloylgentiobiose); 4 (1,2,2′ -trisinapoylgentiobiose); 5 (1,2′ -disinapoyl-2-feruloylgentiobiose); 6 (1-Sinapoyl-2,2′ -diferuloylgentiobiose); 7 (1,2,2′ -trisinapoylgentiobiose); 8 (1,2,2′ -triferuloylgentiobiose); compound identification according to HPLC-DAD-MS analysis.21 b Means (n = 5) within a column followed by the same lower-case letter are not significantly different at P < 0.05 according to Duncan’s multiple range test. P < 0.001 0.09 P < 0.01 0.07 P < 0.001 0.16 P < 0.001 0.20 P < 0.001 0.25 P < 0.001 0.09 P < 0.001 0.21 P < 0.001 0.22 ANOVA P-value LSD (P < 0.05) P < 0.001 3.96 P < 0.001 0.29 P < 0.001 0.39 0.05c 0.28ab 0.26ab 0.19b 0.32a 0.29ab 0.23bc 0.33a 0.18c 0.33a 1.41b 1.23c 1.68a 1.07c 1.77a 1.81b 1.84b 2.28a 1.53c 2.30a 0.24c 1.39a 0.79b 0.28c 0.18c 0.64a 0.54b 0.58ab 0.40c 0.54b 2.22c 2.45b 2.62ab 1.86d 2.82a 1.91b 2.61a 2.59a 2.04b 2.62a Control T1 T1 +E1 T1 +E2 T1 +E3 33.86b 36.26ab 32.44b 37.89a 1.07c 1.23c 1.74b 1.59b 2.11a 2.80a 2.94a 3.03a 2.18b 3.06a 7 6 5 4 1 C1 24.97cb Treatment Total flavonoids C2 2 3 Sinapic and ferulic derivativesa Caffeoyl-quinic derivativesa Table 4. Total flavonoids, caffeoyl-quinic derivatives, and individual sinapic and ferulic derivative levels (mg 100 g−1 fresh weight) in the inflorescences of greenhouse-grown ‘Marathon’ broccoli 8 DA Moreno et al. 40 Control T1 T1 + E1 T1 + E2 T1 + E3 LSD(P<0.05)= 4.59 ab a b c da 20 0 Treatments Figure 1. Total phenolics (mg 100 g−1 of fresh weight) in the inflorescences of greenhouse-grown ‘Marathon’ broccoli. a Means (n = 5; P < 0.001) with the same lower-case letter are not significantly different at P < 0.05 according to Duncan’s multiple range test. J Sci Food Agric 88:1472–1481 (2008) DOI: 10.1002/jsfa J Sci Food Agric 88:1472–1481 (2008) DOI: 10.1002/jsfa 100 Foliar Total phenols (mg·100g-1 f.w.) a C1 (neochlorogenic acid) and C2 (chlorogenic acid); sinapic and ferulic derivatives: 1 (1,2-disnapoylgentiobiose); 2 (1-sinapoyl-2-feruloylgentiobiose); 3 (1,2-diferuloylgentiobiose); 4 (1,2,2′ -trisinapoylgentiobiose); 5 (1,2′ -disinapoyl-2-feruloylgentiobiose); 6 (1-sinapoyl-2,2′ -diferuloylgentiobiose); 7 (1,2,2′ -trisinapoylgentiobiose); 8 (1,2,2′ -triferuloylgentiobiose); compound identification according to HPLC-DAD-MS analysis.21 b Means (n = 5) within a column followed by the same lower-case letter are not significantly different at P < 0.05 according to Duncan’s multiple range test. 0.05b 0.07ab 0.09a 0.06b 0.05b P < 0.05 0.03 0.27a 0.34a 0.38a 0.27a 0.23a P > 0.05 0.19 1.32bc 1.64a 1.54ab 1.27c 1.15c P < 0.01 0.24 1.35b 2.03a 2.00a 1.58b 1.51b P < 0.001 0.28 0.35bc 0.43ab 0.44a 0.25d 0.32cd P < 0.001 0.08 0.05ab 0.04b 0.06a 0.05ab 0.04b P < 0.001 0.01 5.06ab 5.26a 5.65a 3.66c 4.35bc P < 0.01 0.90 5.16ab 6.05a 5.22ab 4.25c 4.48bc P < 0.01 0.91 Control T1 T1 +E1 T1 +E2 T1 +E3 ANOVA P-value LSD (P < 0.05) 51.15ab 55.14a 49.65a 34.32b 40.35b P < 0.001 8.30 5.03a 4.39a 4.65a 3.62b 3.53b P < 0.001 0.67 4.25a 4.38a 3.99a 2.96b 3.25b P < 0.01 0.73 7 6 5 4 1 C1 Treatment Total flavonoids C2 2 3 Sinapic and ferulic derivativesa Caffeoyl-quinic derivativesa Table 5. Total flavonoids, caffeoyl-quinic derivatives, and individual sinapic and ferulic derivative levels (mg 100 g−1 fresh weight) in young fully expanded leaves of greenhouse-grown ‘Marathon’ broccoli 8 Phytonutrients in hydroponically grown broccoli 80 LSD(P<0.05)= 11.556 a aa a Control T1 T1 + E1 T1 + E2 T1 + E3 b 60 b 40 20 0 Treatments Figure 2. Total phenolics (mg 100 g−1 of fresh weight) in young fully expanded leaves of greenhouse-grown ‘Marathon’ broccoli. a Means (n = 5; P < 0.001) with the same lower-case letter are not significantly different at P < 0.05 according to Duncan’s multiple range test. reported. The agronomic conditions (growing in the greenhouse during winter) should be taken into consideration if we are intending to obtain broccoli for raw material/ingredients for the development of functional foods or phytochemically rich vegetables, growing the plants under the controlled greenhouse environment and with the putative use of stress factors (elicitors), increasing the content of phytochemicals in broccoli. The results of the phenolic contents in the leaves could be not be explained by a possible ‘dilution’ of the phytochemicals with development, because of the absence of correlation between these parameters (P > 0.05; data not shown). Instead, we could find at least part of the explanation in the physiological function of the leaves in a stressed plant, and the source–sink relationships for the biosynthesis and translocation of different phytochemicals to the inflorescences, to maximize the resource-use efficiency of the plant.24 Vitamin C The AA and the dehydroascorbic acid DHAA in the broccoli inflorescences were affected by the treatments, but not the total content of vitamin C (Fig. 3). Our ‘Marathon’ inflorescences were less rich in vitamin C than the field-produced inflorescences (74–107 mg 100 g−1 fresh weight),20,21 although the content could be considered as normal (25–80 mg 100 g−1 fresh weight).25 The remarkable data of the T1 + E3 (chitosan) inducing the significant highest and lowest AA and DHAA levels, respectively, confirmed the effects of this elicitor with no negative effect on ascorbate,18 but the general trend is the absence of effects on the total content of vitamin C, which is also a positive outcome. The effects on young fully expanded leaves were quite different (Fig. 4), because the higher content of vitamin C in leaves than in the inflorescences was owed to the high to high levels of DHAA, 1477 and in this case the T1 + E3 treatment induced the lowest content of vitamin C in leaves. The highest vitamin C content was found in the control plants, not different from T1 + E1 or T1 + E2. A possible explanation of these results may be related to the source/sink trends between leaves and inflorescences in broccoli (weak positive correlation coefficients of r 2 < 0.4; data not shown). From these results, it looks as though the changes in vitamin C and its components (AA and DHAA) could be more related to the environmental growth conditions (Table 1) than to the imposed stress factor treatments (Table 2, Fig. 4). Ascorbic acid declined under full sunlight conditions in mustard greens, and supplementing natural light with blue or sodium vapour light increased ascorbic acid concentrations in broccoli. Rainy, cloudy, cool conditions also decrease ascorbic acid.26 In such conditions, similar to our experimental setup, the relationships between size and phytonutrient concentrations may or may not always be linear, and may not always be negative for ascorbic acid. Glucosinolates The broccoli inflorescences (Table 6) contained the aliphatic glucosinolates glucoiberin (GI), glucoraphanin (GR) and glucoerucin (GE), as well as the indole glucosinolates 4-hydroxy-glucobrassicin (HGB), glucobrassicin (GB), 4-methoxyglucobrassicin (MGB) and neoglucobrassicin (NGB). The main glucosinolates were glucoraphanin and glucobrassicin, but other glucosinolates found in very small amounts (glucoalyssin, gluconapin, etc.) were taking into account only for the total content of glucosinolates (Table 6). Both aliphatic and indole glucosinolates were affected by the treatments, and T1 + E1 induced significantly higher amounts of the aliphatic glucosinolates GI (by 47%) and GR (by 21%) than the control. T1 + E2 also surpassed the control by 24% for GE. The indole glucosinolates were similar between the control, T1 (40 mmol NaCl) and T1 + E1, with significantly higher amounts of HGB, GB, MGB and NGB than T1 + E2, and with T1 + E3, the treatment with the lowest content in total glucosinolates. Inflorescence Vitamin C (mg 100 g-1 of fresh weight) DA Moreno et al. 80 Control T1 T1 + E1 T1 + E2 T1 + E3 60 LSD(P<0.05)= 13.47 40 LSD(P<0.05)= 5.11 a ab bcz bc c 20 LSD(P<0.05)= 7.12 a a a a b 0 AA DHAA Total Vitamin C Figure 3. Ascorbic acid (AA), dehydroascorbic acid (DHAA), and total vitamin C (mg100 g−1 of fresh weight), in the inflorescences of greenhouse-grown ‘Marathon’ broccoli. a Means (n = 5; AA P < 0.01; DHA P < 0.05) with the same lower-case letter are not significantly different at P < 0.05 according to Duncan’s multiple range test. GR concentration in the broccoli inflorescences was significantly increased compared to the control when methionine (T1 + E1) was applied at the time of head formation. In previous reports of greenhouse-grown broccoli under controlled conditions and fertilized with methionine, the same kind of response was found.17 In any case, differences in total contents were not relevant if compared to the untreated control. The metabolically active leaves showed that GR was highest in the T1 + E1 treatment (Table 7), surpassing the control by 78%, whereas GI and GE were significantly the lowest in T1 + E3. The indole glucosinolates showed a similar variation between treatments, with HGB, GB, MGB and NGB being higher in control, T1, T1 + E1 and T1 + E2, and lowest in T1 + E3, as repeated in the total content of glucosinolates, and similar to what was found for the inflorescences (Table 5). Table 6. Aliphatic, indole and total glucosinolates (mg 100 g−1 fresh weight) in the inflorescences (broccoli heads) of greenhouse-grown ‘Marathon’ broccoli Aliphatic glucosinolatesa Treatment Control T1 T1 +E1 T1 +E2 T1 +E3 ANOVA P-value LSD (P < 0.05) GI GR GE 14.62bb 14.61b 21.55a 15.63b 13.88b 38.12bc 42.73ab 46.12a 33.01c 33.62c 15.72b 15.50b 16.37b 19.56a 10.56c P < 0.01 4.26 P < 0.01 6.33 P < 0.001 2.95 Indole glucosinolatesa HGB 22.10abc 25.55ab 25.37a 20.44bc 17.86c P < 0.05 5.12 GB MGB NGB Total glucosinolates 68.36a 59.07b 52.49c 43.40d 49.68c 17.10a 16.68a 14.17b 12.61b 9.928c 58.47a 38.82b 36.52b 25.36c 23.03c 260.94a 235.39a 234.58a 201.82b 178.45b P < 0.001 6.08 P < 0.001 2.35 P < 0.001 6.23 P < 0.001 29.46 a GI: 3-methylslfinylpropyl-glucosinolate; GR: 4-methylsulfinylbutyl-glucosinolate; GE: 4-methylthiobutyl-glucosinolate; HGB: 4-hydroxyindol-3ylmethyl-gls; GB: 3-indolylmethyl-gls; MGB: 4-methoxy-3-indolylmethyl-gls; NGB: N-methoxy-3-indolylmethyl-gls. b Means (n = 5) within a column followed by the same lower-case letter are not significantly different at P < 0.05 according to Duncan’s multiple range test. 1478 J Sci Food Agric 88:1472–1481 (2008) DOI: 10.1002/jsfa Phytonutrients in hydroponically grown broccoli Table 7. Aliphatic, indole and total glucosinolates (mg 100 g−1 fresh weight) in the young fully expanded leaves of greenhouse-grown ‘Marathon’ broccoli Aliphatic glucosinolatesa Treatment GI GR GE HGB GB MGB NGB Total glucosinolates 24.56ab 17.81b 15.80b 26.54a 10.09c 17.93b 16.64b 31.98a 15.42b 10.32c 18.01a 18.73a 16.17a 15.39ab 12.54b 18.21a 10.27bc 12.27b 12.44b 7.67c 26.90bc 35.26bc 48.78a 36.09b 26.24c 27.91a 26.79a 25.61a 22.86ab 18.72b 15.97cd 21.91ab 18.97bc 23.24a 14.68d 208.95a 221.89a 207.83a 206.66a 140.77b P < 0.001 5.78 P < 0.001 5.10 P < 0.05 3.51 P < 0.001 2.818 P < 0.001 9.62 P < 0.05 5.41 P < 0.01 4.17 P < 0.001 21.69 Control T1 T1 +E1 T1 +E2 T1 +E3 ANOVA P-value LSD (P < 0.05) Indole glucosinolatesa a GI: 3-methylslfinylpropyl-glucosinolate; GR: 4-methylsulfinylbutyl-glucosinolate; GE: 4-methylthiobutyl-glucosinolate; HGB: 4-hydroxyindol-3ylmethyl-gls; GB: 3-indolylmethyl-gls; MGB: 4-methoxy-3-indolylmethyl-gls; NGB: N-methoxy-3-indolylmethyl-gls. b Means (n = 5) within a column followed by the same lower-case letter are not significantly different at P < 0.05 according to Duncan’s multiple range test. Foliar Vitamin C (mg ·100g-1 f.w.) 80 Control T1 T1 + E1 T1 + E2 T1 + E3 60 LSD(P<0.05)= 6.57 a a LSD(P<0.05)= 6.04 40 ab b a b a LSD(P<0.05)= 2.20 a c bz 20 c d c c c 0 AA DHAA Total Vitamin C Figure 4. Ascorbic acid (AA), dehydroascorbic acid (DHAA), and total vitamin C (mg 100 g−1 of fresh weight), in young fully expanded leaves of greenhouse-grown ‘Marathon’ broccoli. a Means (n = 5; P < 0.001) with the same lower-case letter are not significantly different at P < 0.05 according to Duncan’s multiple range test. Numerous studies on single vegetables and phytochemicals have demonstrated that pre-harvest variables are factors that have the potential to influence the phytochemical content in the final produce.5,11,27,28 In using immature broccoli flower crops for the production of glucosinolate-enriched raw plant material, for functional foods or supplements, cultivation of its cultivars (e.g. ‘Marathon’, ‘Shogun’) in seasons with relatively low daily mean temperatures (about 14 ◦ C; springtime in Germany) has been recommended, combined with rising daily mean irradiation up to 450 µmol m−2 s−1 of the photosynthetic photon flux density.28 Broccoli could be produced in Murcia (Spain) in the fall/winter and early spring seasons,8,11,20 at temperatures within that range (Table 1). The effects of treatments on individual and total glucosinolates of the inflorescences and leaves in this experiment (greenhouse soilless culture) could be due to the changing source–sink relationship of photoassimilates, and glucosinolate exchange between the J Sci Food Agric 88:1472–1481 (2008) DOI: 10.1002/jsfa individual plant organs is also possible, since the reason for the different responses amongst the glucosinolate groups could be because the various enzymes involved in the synthesis of each glucosinolate are affected differently depending upon treatment and environmental conditions,7 helping to clarify why the control, T1, T1 + E1 and T1 + E2 presented similar glucosinolate contents (individual and total glucosinolates in the inflorescences and leaves). In most markets broccoli is sold on a per head basis, not by weight. Total and individual glucosinolates per head may be an essential criterion in considering enhancement of phytochemicals (i.e., glucosinolates) in certain broccoli genotypes.27 In our work, there is no significant correlation between individual or total glucosinolate content and head weight, and there has been no indication of a dilution effect, only a weak correlation between GE and head weight (r 2 = 0.439, P < 0.05) or between leaf weight and total glucosinolates (r 2 = 0.415, P < 0.05). Management practices such as nutrient supply have been investigated for their specific influences on the contents of glucosinolates and aroma volatiles by using broccoli and radish as examples.8,11,29 An increased level of mineral nutrients (i.e., nitrogen, mineral and organic fertilization) in a field experiment with broccoli decreased the content of aliphatic glucosinolates. Using Hoagland’s solution for broccoli may have supplied enough nutrients to maintain a high level of glucosinolates in control and treated plants (T1, T1 + E1, T1 + E2), but not in T1 + E3, where the influence of chitosan could account for the effect on glucosinolates and phenolics more importantly than the plant nutrient status. Significant effects of the treatments on phytochemical content in broccoli indicate that the management practices for increasing one given phytochemical (i.e., glucoraphanin and glucobrassicin for chemoprevention) may be related to a decreased level of natural antioxidants (i.e., hydroxycinnamic acids). The response in the inflorescences for total glucosinolates (lowest in T1 + E3) and total phenolics (highest in T1 + E3) was somehow corroborated or supported by 1479 DA Moreno et al. the correlation between total glucosinolates and the total flavonoids (r 2 = 0.518; P < 0.01) and phenolics (r 2 = 0.418; P < 0.05) as a negative relationship. For the leaves, phenolics and glucosinolates were not significantly related (r 2 < 0.1). Obtaining broccoli that delivers high amounts of bioactive phytochemicals and chemoprotective potency as a feasible goal therefore needs the consideration of all the factors (i.e., environmental conditions, abiotic stress treatments, elicitors) responsible for the wide variation in phytochemical contents at harvest.27,28 2 3 4 5 CONCLUSIONS To satisfy the increasing health consciousness of consumers worldwide, the demand for broccoli enriched with phytochemicals – available as fresh produce or raw material for functional foods and supplements – would need the integration of total quality management strategy with respect to the genetic and environmental effects on the formation of bioactive compounds, selecting the correct time of planting and harvesting as well as the use of abiotic stress (NaCl), during the vegetative period and additional factors during head induction and development (i.e., sprayed elicitors), for the induction of phytochemical biosynthesis, to manipulate the response of plants to different environmental factors, and to enhance the amount of dietary antioxidants and phytonutrients (i.e., human wellness compounds) which along with consumption of five or more servings per day of fruits and vegetables will make for a healthier population. From our experience at this point, growing broccoli hydroponically in the greenhouse as a winter crop in Spain (Mediterranean climate) needs the supporting treatment of abiotic stress during development (i.e., NaCl). Additionally, the use of stress factors at head induction and development (i.e., elicitors) may serve the purpose of enhancing the nutritional quality to deliver a health-promoting food. ACKNOWLEDGEMENTS The authors wish to thank the CICYT National Programme for financial support of this work (AGL20066499/AGR). Part of this work was also funded by CSIC (Proyecto Intramural 200470E038). Carmen López-Berenguer thanks the Regional Government of Murcia for funding through ‘Science and Technology’ for PhD grants of the ‘Fundación Séneca’. Diego A Moreno thanks the European Social Fund (ESF) and the Spanish Ministerio de Educación y Ciencia and CSIC for funding through the ‘Ramon y Cajal’ S&T Programme. We thank Ascensión Mart ı́nez-Sánchez and Santiago Pérez-Balibrea for their valuable help and technical assistance. 6 7 8 9 10 11 12 13 14 15 16 17 18 REFERENCES 1 Canene-Adams K, Lindshield BL, Wang S, Jeffery EH, Clinton SK and Erdman JW Jr, Combinations of tomato and 1480 19 broccoli enhance antitumor activity in Dunning R3327-H prostate adenocarcinomas. Cancer Res 67:836–843 (2007). Higdon JV, Delage B, Williams DW and Dashwood RH, Cruciferous vegetables and human cancer risk: epidemiological evidence and mechanistics basis. Pharmacol Res 55:224–236 (2007). Moreno DA, Carvajal M, López-Berenguer C and Garcı́aViguera C, Chemical and biological characterisation of nutraceutical compounds of broccoli. J Pharmaceut Biomed Anal 41:1508–1522 (2006). Gliszczyńska-Świglo A, Kalużewicz A, Lemańska K, Knaflewski M and Tyrakowska B, The effect of solar radiation on the flavonol content in broccoli inflorescence. Food Chem 100:241–245 (2007). Jeffery EH, Brown AF, Kurilich AC, Keck AS, Matusheski N, Klein BP et al, Variation in content of bioactive compounds in broccoli. J Food Comp Anal 16:323–330 (2003). Sanwal SK, Laxminarayana K, Yadav DS, Rai N and Yadav RK, Growth, yield, and dietary antioxidants of broccoli as affected by fertilizer type. J Veg Sci 12:13–26 (2006). Schonhof I, Kläring H-P, Krumbein A, Claußen W and Schreiner M, Effect of temperature increase under low radiation conditions on phytochemicals and ascorbic acid in greenhouse grown broccoli. Agric Ecosys Environ 119:103–111 (2007). Vallejo F, Tomás-Barberán FA and Garcı́a-Viguera C, Potential bioactive compounds in health promotion from broccoli cultivars grown in Spain. J Sci Food Agric 82:1293–1297 (2002). Vallejo F, Tomás-Barberán FA and Garcı́a-Viguera C, Glucosinolates and vitamin C content in edible parts of broccoli florets alter domestic cooking. Eur Food Res Technol 215:310–316 (2002). Vallejo F, Tomás-Barberán FA and Garcı́a-Viguera C, Healthpromoting compounds in broccoli as influenced by refrigerated transport and retail sale period. J Agric Food Chem 51:3029–3034 (2003). Vallejo F, Tomás-Barberán FA, Benavente-Garcı́a AG and Garcı́a-Viguera C, Total and individual glucosinolate contents in inflorescences of eight broccoli cultivars grown under various climatic and fertilisation conditions. J Sci Food Agric 83:307–313 (2003). López-Berenguer C, Garcı́a-Viguera C and Carvajal M, Are root hydraulic conductivity responses to salinity controlled by aquaporins in broccoli plants? Plant Soil 279:13–23 (2006). López-Berenguer C, Moreno DA, Carvajal M and Garcı́aViguera C, Effect of salt stress on glucosinolate content in broccoli plants, in First International Conference on Glucosinolates: Glucosinolate Biology, Chemistry and Biochemistry and its Application to Human Health and Agriculture, Max Planck Institute for Chemical Ecology and the Phytochemical Society of Europe, Jena, Germany (2006). Demmig-Adams B and Adams WW III, Antioxidants in photosynthesis and human nutrition. Science 298:2149–2153 (2002). Poulev A, O’Neal JM, Logendra S, Pouleva RB, Timeva V, Garvey AS et al, Elicitation, a new window into plant chemodiversity and phytochemical drug discovery. J Med Chem 46:2542–2547 (2003). Van Dam NM, Witjes L and Svatos A, Interactions between aboveground and belowground induction of glucosinolates in two wild Brassica species. New Phytol 161:801–810 (2004). Scheuner ET, Schmidt S, Krumbein A, Schonhof I and Schreiner M, Effect of methionine foliar fertilization on glucosinolate concentration in broccoli and radish. J Plant Nutr Soil Sci 168:275–277 (2005). Lee Y-S, Kim Y-H and Kim S-B, Changes in the respiration, growth, and vitamin C content of soybean sprouts in the response to chitosan of different molecular weights. HortSci 40:1333–1335 (2005). Vallejo F, Garcı́a-Viguera C and Tomás-Barberán FA, Changes in broccoli (Brassica oleracea L. var italica) health-promoting J Sci Food Agric 88:1472–1481 (2008) DOI: 10.1002/jsfa Phytonutrients in hydroponically grown broccoli 20 21 22 23 24 compounds with inflorescence development. J Agric Food Chem 51:3776–3782 (2003). Al-Gendy AA and Lockwood GB, Production of glucosinolate hydrolysis products in Farsetia aegyptia suspension cultures following elicitation. Fitoterapia 76:288–295 (2005). Vallejo F, Tomás-Barberán FA and Garcı́a-Viguera C, Effect of climatic and sulphur fertilisation conditions, on phenolic compounds and vitamin C, in the inflorescences of eight broccoli cultivars. Eur Food Res Technol 216:395–401 (2003). Martı́nez-Sánchez A, Allende A, Bennet RN, Ferreres F and Gil MI, Microbial, nutritional and sensory quality of rocket leaves as affected by different sanitizers. Postharvest Biol Technol 42:86–97 (2006). Wildanger W and Hermann K, Flavonole und Flavone der Gemüsearten. I. Flavonole der Kohlarten. Eur Food Res Technol 152:134–137 (1973). Hikosaka K, Leaf canopy as a dynamic system: ecophysiology and optimality in leaf turnover. Ann Bot 95:521–533 (2005). J Sci Food Agric 88:1472–1481 (2008) DOI: 10.1002/jsfa 25 Singh J, Rai M, Upadhyay AK, Bahadur A, Chauraisa SNS and Singh KP, Antioxidant phytochemicals in broccoli (Brassica oleracea L. var. italica Plenck) cultivars. J Food Sci Technol 43:391–393 (2006). 26 Lester GE, Environmental regulation of human health nutrients (ascorbic acid, β-carotene, and folic acid) in fruits and vegetables. HortSci 41:59–63 (2006). 27 Farnham MW, Wilson PE, Stephenson KK and Fahey JW, Genetic and environmental effects on glucosinolate content and chemoprotective potency of broccoli. Plant Breeding 123:60–65 (2004). 28 Schreiner M, Vegetable crop management strategies to increase the quantity of phytochemicals. Eur J Nutr 44:. 85–94 (2005). 29 Krumbein A, Schonhof I, Rühlmann J and Widell S, Influence of sulphur and nitrogen supply on flavour and health-affecting compounds in Brassicaceae, in Plant Nutrition: Developments in Plant and Soil Sciences, Vol. 92, ed. by Horst WJ et al. Kluwer, Dordrecht, pp. 294–295 (2002). 1481 J Sci Food Agric 88:1482–1485 (2008) Journal of the Science of Food and Agriculture Short Communication Anti-sickling potential of Aloe vera extract Agunna Everest Ejele∗ and Pascal Chukwuemeka Njoku Department of Chemistry, Federal University of Technology, PMB 1526, Owerri, Nigeria Abstract: The effect of Aloe vera extract on the gelling time of human HbSS erythrocytes was investigated. The results showed that A. vera extract increased the gelling time of HbSS blood and inhibited sickling in vitro. In addition, a linear relationship was found between extract concentration and gelling time, suggesting that A. vera extract may have great potential in the management of sickle cell disease.  2008 Society of Chemical Industry Keywords: Aloe vera; sickle cell disease; red blood cells; gelling time INTRODUCTION Several attempts have been made to reduce or totally inhibit the occurrence of painful crisis in sickle cell disease using various drugs.1,2 However, it has been difficult to obtain an effective and safe antisickling drug for use in sickle cell therapy, because many drugs employed in the treatment of sickle cell disease are effective only at high concentrations.2 This has led to various attempts to introduce indigenous herbs in the therapy of sickle cell disorders. In this regard, Sofowora and Isaacs3 found that an extract from the root of Fagara zanthoxyloides had antisickling properties and reduced the frequency of painful crisis. Since then, several papers on the use of medicinal plants as remedies for painful crisis in sickle cell therapy have been published.4 – 6 Ekeke and Shode4 reported that a boiled extract from the edible bean Cajanus cajan (‘Fiofio’ in Igbo) of the family Papilionaceae brought enormous relief to a sickle cell patient. The extract not only inhibited sickling in vitro in sodium metabisulfite (Na2 S2 O5 ) solution but also reverted pre-sickled erythrocytes to their normal morphology. Another medicinal plant that has proved effective in the management of sickle cell disease is an extract of the leaves of Terminalea catappa L. (Indian almond) of the family Combretaceae.5 The extract was found to inhibit osmotically induced haemolysis of human erythrocytes in a dose-dependent manner. Mgbemene and Ohiri5 reported that a 1.0 mg mL−1 solution of the extract was effective in preventing and reversing the sickling of human HbS erythrocytes induced by 2% (w/v) Sodium metabisulphite (Na2 S2 O5 ) solution. Recently, Njoku and Ejele6 reported the prophylactic effect of a liquid extract of Telferia occidentalis (‘Ugu’ in Igbo) in the prevention of painful sickle cell crisis and showed a 95% reduction in the frequency of occurrence of sickle cell disease and a 100% reduction in painful crisis among five homozygous volunteers aged 5–10 who were studied for a period of 8 months. In this paper we report on the anti-sickling potential of Aloe vera extract, popularly referred to as ‘a pharmacy in a plant’. MATERIALS AND METHODS Aloe vera extract Whole fresh leaves of A. vera obtained from the open market in Owerri (Imo State, Nigeria) were used in this study. About 25 g of sun-dried leaves were extracted with 250 mL of ethanol for 12 h in a Soxhlet extractor equipped with a reflux condenser. The ethanolic extract was concentrated to 100 mL by distillation, cooled and filtered and the filtrate was used without further purification. Preliminary phytochemical tests carried out on the filtrate showed that alkaloids, tannins, saponins and glycosides were present. However, we did not isolate or separate any of the active principles. Cajanus cajan extract Fresh seeds of the local edible bean C. cajan obtained from the open market in Owerri were used in this study. About 35 g of sun-dried edible beans were extracted with 250 mL of ethanol for 12 h in a Soxhlet extractor equipped with a reflux condenser. The ethanolic extract was concentrated to 100 mL by distillation, cooled and filtered and the filtrate was used without further purification. Sickling test This test is based on the morphological change in HbS blood containing red blood cells when deoxygenated. When the cells of sickle cell blood are exposed to reduced oxygen tension through the use of a reducing ∗ Correspondence to: Agunna Everest Ejele, Department of Chemistry, Federal University of Technology, PMB 1526, Owerri, Nigeria E-mail: monyeejele@yahoo.com (Received 25 April 2006; revised version received 29 May 2007; accepted 6 June 2007) Published online 7 April 2008; DOI: 10.1002/jsfa.3036  2008 Society of Chemical Industry. J Sci Food Agric 0022–5142/2008/$30.00 Anti-sickling potential of Aloe vera extract agent such as Na2 S2 O5 solution,4 the erythrocytes assume a characteristic sickle shape. Venous blood was used in this study. Blood samples were obtained from healthy donors as well as sickle cell patients at the Federal Medical Centre, Owerri. Whole blood containing ethylene diamine tetraacetic acid (EDTA) as anticoagulant was centrifuged for about 5 min, the upper supernatant layer was discarded and the packed red blood cells were used for the gelling experiments. A 1 mL aliquot of HbAA red blood cells (RBCs) was mixed with two drops of freshly prepared 2% (w/v) NaS2 O5 solution with the aid of an applicator stick on a microscope slide labelled A, covered with a coverslip and sealed with molten paraffin wax to exclude air and prevent drying. The slide was then incubated at 37 ◦ C for 10 min and observed for sickling and RBC count. This was used as control 1. A 1 mL aliquot of HbSS RBCs was mixed with two drops of freshly prepared 2% (w/v) NaS2 O5 solution with the aid of an applicator stick on a microscope slide labelled B, covered with a coverslip and sealed with molten paraffin wax to exclude air and prevent drying. The slide was then incubated at 37 ◦ C for 10 min and observed for sickling and RBC count. This was used as control 2. The same procedure was followed for microscope slides labelled C1 − C5 , but here a 1 mL aliquot of HbSS RBCs was mixed with two drops of freshly prepared 2% (w/v) NaS2 O5 solution and different amounts of A. vera extract (see Table 1). Thereafter the slides were viewed under a microscope and RBCs were counted with a haemocytometer to determine the number of unsickled RBCs mL−1 blood. The number of unsickled RBCs was used as an index of sickling. The above experiments were repeated using C. cajan extract (see Table 2). Gelation of sickle cell blood In these experiments the effect of A. vera extract on sickle cell blood was determined and the time of gelation of erythrocytes in vitro was obtained. Ten test tubes labelled A–J were used and arranged in order. In test tube A a 0.5 mL aliquot of RBCs was mixed with two drops of 2% (w/v) Na2 S2 O5 solution. The mixture was stirred gently and the time at which gelling occurred was noted. This was taken as the blank. The same procedure was used for test tubes B–J, to which different concentrations of A. vera extract were added. The time taken for the blood samples in test tubes B–J to gel was determined (see Table 3). RESULTS AND DISCUSSION Table 1 shows the results of the sickling test carried out on HbSS erythrocytes to which different volumes (number of drops) of the ethanolic extract of A. vera Table 1. Anti-sickling characteristics of Aloe vera extract Red blood cell count (106 µL−1 ) Sickling test Sample Volume of extract (drops) HbAA HbSS HbAA HbSS Slide A Slide B Slide C1 Slide C2 Slide C3 Slide C4 Slide C5 – – 2 4 6 8 10 5.0 – 4.8 4.9 5.0 5.2 5.4 – 2.0 1.65 1.86 2.35 2.45 2.50 Negative – Negative Negative Negative Negative Negative – Positive Positive Positive Positive Positive Positive a Commenta # +++ ++ ++ + + + Key: +++ , marked degree of sickling; ++ , moderate degree of sickling; + , low degree of sickling; # , no sickling observed. Table 2. Anti-sickling characteristics of Cajanus cajan extract Red blood cell count (106 µL−1 ) Sickling test Sample Volume of extract (drops) HbAA HbSS HbAA HbSS Slide A Slide B Slide C1 Slide C2 Slide C3 Slide C4 Slide C5 – – 2 4 6 8 10 5.0 – 4.2 4.4 4.7 4.9 5.0 – 2.0 1.35 1.46 1.54 1.85 2.00 Negative – Negative Negative Negative Negative Negative – Positive Positive Positive Positive Positive Positive a Commenta # +++ ++ ++ + + + Key: +++ , marked degree of sickling; ++ , moderate degree of sickling; + , low degree of sickling; # , no sickling observed. J Sci Food Agric 88:1482–1485 (2008) DOI: 10.1002/jsfa 1483 AE Ejele, PC Njoku Table 3. Changes in gelling time due to addition of Aloe vera extract Sample A. 0.5 mL of blood + 2 drops of 2%NaS2 O5 B. 0.5 mL of blood + 2 drops of 2%NaS2 O5 C. 0.5 mL of blood + 2 drops of 2%NaS2 O5 D. 0.5 mL of blood + 2 drops of 2%NaS2 O5 E. 0.5 mL of blood + 2 drops of 2%NaS2 O5 F. 0.5 mL of blood + 2 drops of 2%NaS2 O5 G. 0.5 mL of blood + 2 drops of 2%NaS2 O5 H. 0.5 mL of blood + 2 drops of 2%NaS2 O5 I. 0.5 mL of blood + 2 drops of 2%NaS2 O5 J. 0.5 mL of blood + 2 drops of 2%NaS2 O5 Aloe vera extract added (v/v) Gelling time (min) – 7 1 drop of 1% solution 15 1 drop of 2% solution 20 1 drop of 5% solution 25 1 drop of 10% solution 27 1 drop of 20% solution 31 1 drop of 50% solution 35 1 drop of 60% solution 37 1 drop of 80% solution 40 1 drop of 100% solution 50 had been added. It can be seen that at all levels the extract inhibited the sickling of erythrocytes induced by two drops of 2% (w/v) Na2 S2 O5 , although to varying degrees. For example, slides C1 and C2 , which contained two and four drops of A. vera extract respectively, showed a marked degree of sickling, whereas slides C3 − C5 , which contained six–ten drops of the extract, showed a much lower degree of sickling. It was also observed that the number of RBCs counted via the haemocytometer increased markedly with increasing volume of A. vera extract from C1 to C5 . Similar results were obtained for the ethanolic extract of C. cajan (Table 2). Although the mechanism involved is not yet known, this observed increase in the number of unsickled RBCs may be explained in terms of a reversion of sickled erythrocytes as the concentration of the extract increased.4 A similar observation was made earlier by Ekeke and Shode,7 who studied the effect of phenylalanine on pre-sickled HbSS erythrocytes suspended in 0.15 mol L−1 phosphate buffer at pH 6.5. Table 3 shows the results of the gelation rate test on HbSS blood in the presence of different concentrations of A. vera extract and gives the time taken for the HbSS blood sample to clot under these conditions. It can be seen that, as the concentration of the extract was increased from 1 to 100% (v/v), the gelling time increased from 15 to 50 min. In other words, there is a linear relationship between the gelling time and the concentration (% v/v) of A. vera extract. The ability of an agent or compound to increase the gelling time of human HbSS blood can be taken as a measure of its anti-sickling potential and determines its capability of retarding the aggregation of erythrocyte cells in blood vessels and increasing the number of unsickled RBCs. Such reduction in aggregation rate is related to gelation 1484 inhibition. Na2 S2 O5 is a strong gelation-promoting agent, so any compound that inhibits the gelation of HbSS erythrocytes in the presence of Na2 S2 O5 will be a powerful gelation inhibitor.4,7 The gelling of human HbSS blood cells results from weak non-covalent tetramer–tetramer interactions, which hold the deoxygenated HbSS molecules together.1,2 These interactions may be classified into electrostatic forces, hydrogen bonds, van der Waals forces and hydrophobic effects and may be disrupted by the addition of a variety of reagents, including salts, urea and sucrose.8 Such reagents help to solubilise and stabilise proteins by making protein–solvent interactions more favourable than protein–protein contacts.2 The degree of sickling inhibition observed in the present study appeared to be a linear function of the concentration of A. vera extract, because an increase in the extract concentration (% v/v) increased the gelling time of the HbSS blood sample. Therefore this study indicates that A. vera extract may contain some active chemical agents that are possible gelation inhibitors. Various researchers have studied the anti-sickling potential of several indigenous medicinal plants.3 – 6 Ekeke and Shode4 demonstrated that a boiled or methanolic extract of the local edible bean C. cajan brought enormous relief to a known sickle cell patient. They showed that the extract inhibited sickling in vitro in the presence of Na2 S2 O5 solution, a strong reducing agent, which on its own is capable of inducing sickling in human HbSS blood.4 In a more recent study, Ekeke et al.7 carried out an amino acid analysis on the water-soluble fraction of a methanolic extract of C. cajan and showed that this fraction contained several free amino acids, which they thought were responsible for most of the anti-sickling properties of the extract. In addition, the authors7 demonstrated that the methanolic (water-soluble) extract contained as much as 26.3% of phenylalanine and concluded that the presence of phenylalanine alone could account for about 70% of the anti-sickling potency of C. cajan extract. Earlier studies have shown that several compounds, including amino acids such as Phe,9 LAsn, L-Gln and Ser,10 sugars such as glucose as well as salts such as potassium cyanate and sodium chloride,8 also possess anti-sickling properties to various degrees. We have not evaluated the amino acid composition of Table 4. Comparison of anti-sickling potential of Aloe vera and Cajanus cajan extracts Volume of Sample Slide C1 Slide C2 Slide C3 Slide C4 Slide C5 Red blood cell count (106 µL−1 ) Sickling time (min) extract (drops) A. vera C. cajan A. vera C. cajan 2 4 6 8 10 1.66 1.80 2.30 2.40 2.50 1.35 1.46 1.54 1.85 2.00 10.15 10.20 10.50 10.56 11.00 10.00 10.15 10.30 10.50 10.75 J Sci Food Agric 88:1482–1485 (2008) DOI: 10.1002/jsfa Anti-sickling potential of Aloe vera extract the ethanolic extract of A. vera in the present study, but we have compared the anti-sickling properties of the ethanolic extracts of C. cajan and A. vera (Table 4). There appears to be no significant difference in the anti-sickling properties of the two extracts. CONCLUSION It is common knowledge that sickle cell disease is always associated with pain, and painful crisis is one of the characteristic features of this disease. Although there have been tremendous advances in the basic molecular understanding of sickle haemoglobin and the processes of gelation and sickling, it is disappointing to note that so little basic scientific research has been carried out on the treatment and management of sickle cell disease, as reflected in the area of improved medical care for patients. The results obtained from this study have shown that A. vera extract can increase the gelling time of HbSS blood and inhibit sickling in vitro and may indeed have great potential in the management of sickle cell disease. 2 Pasvol G, Cellular Mechanism for protective effect of HbSS against P. falicparum malaria. Nature 274:701–708 (1978). 3 Sofowora EA and Isaacs WA, Reversal of sickling and cremation in erythrocytes by root extract of Fagara Zathoxyloides. Lloydia 34:383–389 (1971). 4 Ekeke GI and Shode FO, Phenylalanine is the predominant anti-sickling agent in Cajanus Caja seed extract. Planta Med 56:41–43 (1990). 5 Mgbemene CN and Ohiri FC, Anti-sickling potential of Terminalia Catappa leaf extract. Pharmaceut Biol 37:152–156 (1999). 6 Njoku PC and Ejele AE, Prophylactic effect of liquid extract of Telfereia Occi-dentalis in the prevention of painful sickle cell crisis. J Sci Ind Stud 1:8–12 (2003). 7 Ekeke GI and Shode FO, Edible legumes as nutritionally beneficial antisickling agents. Journal of Biochemistry and Molecular Biology 56:111–113 (1999). 8 Freedman ML, Weissmann G, Gorman BD and CunninghamRundles W, Effects of amino acids on gelation kinetics and solubility of sickle hemoglobin. Biochemica Biophysica Research community 22:637–643 (1973). 9 Noguchi CT and Schechter AN, The treatment of sickle cell disease. A historical and chronological literature of the therapies applied since 1910. Tropical and Geographical Medicine 74:1–26 (1977). 10 Rumen NM, Antisickling effect of dietary thiocyanate on prophylactic control of sickle anemia. Journal of Nat. Medical association 45:1053–1056 (1975). REFERENCES 1 Serjeant GR, Sickle Cell Disease. Oxford University Press, Oxford (1985). J Sci Food Agric 88:1482–1485 (2008) DOI: 10.1002/jsfa 1485

RELATED PAPERS

RELATED TOPICS

Log In

Effect of lyophilisation, refrigerated storage and frozen storage on the coagulant activity and microbiological quality ofCynara cardunculus L. extracts

Effect of lyophilisation, refrigerated storage and frozen storage on the coagulant activity and microbiological quality ofCynara cardunculus L. extracts

Related Papers

RELATED PAPERS

RELATED TOPICS