The role of sorghum genotype in the interaction with the parasitic weed Striga hermonthica Promotor: Prof. dr. M. J. Kropff Hoogleraar in de Gewas- en Onkruidecologie Co-promotor: Dr. ir. L. Bastiaans Universitair docent, leerstoelgroep Gewas- en Onkruidecologie Promotiecommissie: Prof. dr. ir. P. Stam (Wageningen Universiteit) Prof. dr. Th. W. Kuyper (Wageningen Universiteit) Prof. dr. W. H. O. Ernst (Vrije Universiteit Amsterdam) Dr. J. A. C. Verkleij (Vrije Universiteit Amsterdam) Dr. B. I. G. Haussmann (ICRISAT Niger) The role of sorghum genotype in the interaction with the parasitic weed Striga hermonthica Jonne Rodenburg Proefschrift ter verkrijging van de graad van doctor op gezag van de rector magnificus van Wageningen Universiteit Prof. dr. M. J. Kropff in het openbaar te verdedigen op woensdag 16 november 2005 des namiddags te vier uur in de Aula Jonne Rodenburg (2005) The role of sorghum genotype in the interaction with the parasitic weed Striga hermonthica. Rodenburg, J. – [S.l.: s.n.]. Ill. PhD thesis Wageningen University. -With ref. – With summaries in English, French and Dutch ISBN: 90-8504-276-3 Abstract Rodenburg, J., 2005. The role of sorghum genotype in the interaction with the parasitic weed Striga hermonthica. PhD thesis, Wageningen University, Wageningen, The Netherlands, 138 pp. with English, French and Dutch summaries. This thesis presents a study on the interaction between the parasitic weed Striga (S. hermonthica [Del.] Benth.) and the cereal crop sorghum (S. bicolor [L.] Moench). Its main objective was to find suitable measures for the selection of breeding material (crop genotypes) with superior levels of resistance or superior levels of tolerance to Striga. To meet this objective the physiological background of tolerance, the relation between Striga infestation, infection and yield loss and the effect of host genotype on Striga parasitism and reproduction were studied. These host-parasite interactions were studied with 4-10 different sorghum genotypes differing in level and mechanism of defence against Striga. Field experiments carried out in Mali were used for yield assessments and development and validation of selection measures. Through pot and agar-gel experiments, aboveground resistance measures were validated with observations on belowground stages. Pot experimentation was also used to create infection response curves and to measure photosynthesis and chlorophyll fluorescence to develop tolerance measures. Striga parasitism and reproduction, and the detrimental effect of Striga on crop yield can significantly be reduced through crop genotype choice. Maximum aboveground Striga number is a reliable selection measure for resistance. Striga flowerstalk dry weight can be used to identify genotypes that reduce Striga reproduction. The maximum relative yield loss is a suitable selection measure for tolerance in susceptible genotypes, while for more resistant genotypes the relative yield loss per Striga infection seems more appropriate. For these tolerance measures, yield assessment of nearby uninfected controls is indispensable. Chlorophyll fluorescence, more precisely photochemical quenching and electron transport rate, may enable screening for tolerance without this requirement. Keywords: Striga hermonthica, Sorghum bicolor, selection measures, resistance, tolerance, genotypes. Preface In November 2000, I started as a PhD student at the Crop and Weed Ecology Group (CWE) of Professor Martin Kropff at Wageningen University (WU). For this opportunity as well as for all the support, trust and friendship I received over the past five years, I need to acknowledge a great number of people. I would like to thank my promotor Martin Kropff for his enthusiasm, faith and support as well as for his scientific input in the thesis. A special word of appreciation goes out to my co-promotor Lammert Bastiaans of the Crop and Weed Ecology Group (CWE) at WU. Dear Lammert, I admire your scientific rigor, perfectionism and dedication. I am thankful for the time you have put into this work and the very pleasant collaboration. My supervisor Aad van Ast is also kindly acknowledged. Dear Aad, you have been a great support to me and I would like to thank you for your time, useful advices and valuable contributions to this thesis. Dale Hess was my supervisor at the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) in the important first experimental season in Samanko, Mali. I would like to thank him for his willingness to collaborate with us, as well as for teaching me the basics of Striga research. When he left, Eva Weltzien became my supervisor at ICRISAT. I am thankful for her contribution to Chapter 2 as well as for her support, kindness and collaborative approach. In Mali, I could not have done the work without the unconditional help of my hardworking and dedicated assistant Ana Amossè Dolo. Thank you very much Ana. I thank my colleagues at CWE in Wageningen. In particular I would like to thank Tom van Mourik, my room mate, Striga-colleague and ‘paranimf’ for his time and help and for the many useful and less useful discussions on Striga and other urgent matters. Also, I thank my fellow-PhD-students that participated in our discussion group and provided me with valuable comments on my work: Nick den Hollander, Harm Smit, Jochem Evers, Marjolein Kruidhof, Bart Timmermans, Sanne Heijting, and Ingrid Haage. Peter van der Putten and Ad Schapendonk are kindly acknowledged for their help with the photosynthesis work. Furthermore, I would like to thank Paula Westerman, Tjeerd Jan Stomph, Maja Slingerland, Gon van Laar, Gijsbertje Berkhout, Hilde Holleman, and Leonie van Scherrenburg. I thank Ans Hofman and Henriette Drenth of CWE and Geurt Versteeg, Peter Saat, Taede Stoker and Henk Meurs of Wageningen Plants Sciences Experimental Center (WPSEC) for their assistance in the experiments in Wageningen. The kind permission of Aad Termorshuizen from the Biological Farming Systems Group (BFS) to use their lab-facilities, and the hospitality of Dine Volker and Oscar de Vos of BFS was much appreciated. Jacques Withagen kindly advised me on statistical matters. I thank ICRISAT for the opportunity to work at their station in Mali. There, I had the indispensable support of many people. In particular, I thank Ibrahima Sissoko, Badara Diallo, the late Idrissa Sanagaré, Moussa and Lassine Keita, Abdoulaye Bah, Abdoulaye Niambaly, Bouba Camara, Boubou Sankaré, Mafhouze Cissé, Safiatou Ba, Ramadjita Tabo, Fred Rattunde, Mamourou Sidibé, Benoit Clerget, Ousmane Youm, Inoussa Akintayo, and Boni N’Tare. Also the help of interns and students, both in Samanko and in Wageningen was of great value. Thanks to Lisette Arts, Ronny Riemens, Justin Lado Mairi, Almouner Yattara, and Mody Doucouré. I kindly acknowledge Bettina Haussmann of the University of Hohenheim (currently working for ICRISAT) for her availability and advice in the early stages, Professor Gebisa Ejeta of Purdue University for advice and sorghum seeds, and Mitch Tuinstra from Kansas State University for his comments on Chapter 3. I would like to express my gratitude to my current employer WARDA for facilitation and my colleagues at WARDA and IITA in Benin for useful discussions. In particular I would like to thank my supervisor at WARDA, Paul Kiepe, for his confidence and support and for proof reading some of the sections. Also many thanks to my colleague Philippe Morant and my friend and colleague Boubakary Cissé for assistance with French translations, and my colleagues David Millar and Bianca Beks for editorial comments. Finally, I thank my friends and family. Particular words of appreciation and thanks must go to Allard Kalverkamp for designing and testing the ethylene gas injector, and to Niels Dingemanse for useful discussions on science and for advice on statistics and quantitative genetics. A special thanks also to my other ‘paranimf’ Ellen Heeres for her critical attitude and unconditional friendship and help. Other friends that I would like to thank for their comprehension, patience and loyalty are: Enver Loke, Jessamijn Miedema, Piet Weber, Sander Zwart, Ruben van den Broek, Tahirou Santara, Coen Elemans, Patrick Zimmerman, Ard Schoemaker, Waldo Ogg, Peter Gildemacher, Frederik Oberthür, Ferdinand Ouedraogo, Kalifa Dembélé, and Bakary Sidibé. I am much indebted to my sisters Floor and Jessica, my parents Piet and Tineke, and to Hawa, for their love and support. Jonne Rodenburg Cotonou, October 2005 Funding Financial assistance for this study was made possible through the beneficence of the Netherlands Foundation for the Advancement of Tropical Research (WOTRO). Printing costs were co-financed by Wageningen University, the J.E. Jurriaanse Stichting and WOTRO. Contents CHAPTER 1 General introduction 1 CHAPTER 2 How can field selection for Striga resistance and tolerance in sorghum be improved? 7 CHAPTER 3 Characterization of host tolerance to Striga hermonthica 29 CHAPTER 4 Can host plant tolerance to Striga hermonthica be detected by photosynthesis measurements? 51 CHAPTER 5 Effects of host plant genotype and seed bank density on Striga reproduction 71 CHAPTER 6 General discussion 97 References 109 Summary 125 Samenvatting 129 Résumé 133 Publications 137 Curriculum vitae 138 CHAPTER 1 General introduction The parasitic weed Striga hermonthica One of the major biotic constraints to cereal production in sub-Saharan Africa is Striga hermonthica (Del.) Benth. (Sauerborn, 1991). This parasitic weed species from the Orobanchaceae (formerly: Scrophulariaceae) family parasitizes on cereals like rice (Oryza glaberrima [Steudel] and O. sativa [L.]), pearl millet (Pennisetum glaucum [L.] R. Br. or P. americanum [L.] K. Schum), maize (Zea mays [L.]) and sorghum (Sorghum bicolor [L.] Moench) (Parker, 1991; Johnson et al., 1997). In sub-Saharan Africa, where problems with Striga hermonthica (Striga) are most severe, 94 % of all the area under cereal production is cultivated with one of its host crops. Sorghum is the most widely cultivated cereal crop in this region with 25.5 million ha under cultivation, which is 30.6% of the total area under cereal crops in 2004 (FAOSTAT, 2004). Infection by Striga can cause yield losses of a few percentages up to complete crop failure, depending on crop species, crop variety and severity of Striga infestation (Doggett, 1965; Riches and Parker, 1995; Adetimirin et al., 2000a). Striga problems are often associated with low soil fertility and marginal environments with high cropping intensities and therefore mostly affect resource-poor subsistence farmers (Kroschel, 1999; Ransom, 2000). The life cycle of Striga Striga is an obligate hemi-parasitic plant implying that it needs a host plant to fulfil its life-cycle, but, having chlorophyllous leaves, is not entirely dependent on its host for its metabolite requirements (Kuijt, 1969). The life cycle of the parasite follows a series of developmental stages going from seed to seed producing plants. Like many other plant species, Striga seeds have a period of primary dormancy before the seeds are able to germinate. For Striga this period is 6 months (Vallance, 1950). A second prerequisite for germination is the preconditioning of the seed, which requires about two weeks of humid and warm (25-35˚ C) conditions (Vallance, 1950; Parker and Riches, 1993). Preconditioned Striga seeds will then need secondary metabolites (xenognosins), derived from the host root, for germination (Saunders, 1933; Vallance, 1 Chapter 1 1950; Yoder, 2001). These xenognosins also serve to direct the radicle of the Striga seedling towards the host root (Williams, 1961a;b). Within four days after germination the radicle needs to find a host root to start formation of a haustorium and penetrate the host root (Riopel and Timko, 1995). The haustorium is a specialized organ that connects the parasite to the xylem of the host root enabling the transport of water and solubles from the parasite to the host (Kuijt, 1977). The xylem to xylem connection is established soon (48-60 hours) after attachment and penetration of the haustorium into the host root tissue (Ramaiah et al., 1991; Riopel and Timko, 1995). Once the xylem connection is established, the Striga can start to develop and grow to the soil surface. First Striga emergence aboveground is reported to occur around 35 to 45 days after crop sowing (DAS) or three to six weeks after attachment (Doggett, 1988; Olivier et al., 1991; Webb and Smith, 1996; Adetimirin et al., 2000b; Haussmann et al., 2001a). Aerial parts of Striga turn green upon exposure to daylight. The aboveground vegetative stage is followed by flowering, which starts around 4 weeks after emergence, and seed production and dissemination, which start around 4 weeks after first flowering at 90 to 120 DAS (Doggett, 1988; Webb and Smith, 1996). Seed size of Striga hermonthica is between 0.2 and 0.3 mm (Parker and Riches, 1993). Estimates on seed production per plant vary from between 5,000 to 85,000 seeds per reproductive plant (Andrews, 1945; Stewart, 1990; Webb and Smith, 1996). Since large quantities of seed will survive the dry season following seed dispersion, a series of cropping seasons with the same host crop will lead to a quick build-up of the Striga seed bank (Weber et al., 1995) and consequently result in increasingly high infections and decreasing crop yields. Striga effects on its host Striga has lower leaf chlorophyll contents and lower photosynthetic rates than related non-parasitic plant species (De La Harpe and Visser, 1979; Shah et al., 1987; Tuquet et al., 1990). Therefore, despite the green leaves, Striga continues to benefit from its host after emergence (Seel et al., 1992; Pageau et al., 1998). Transpiration rates of aboveground Striga exceed that of its host, show little to no response to darkness and only reduce when the host is subjected to water stress (Press et al., 1987a, 1988; Ackroyd and Graves, 1997). This principle ensures a constant flux of water from the host to the parasite (Raven, 1983; Schulze et al., 1984; Press et al., 1987b; Shah et al., 1987; Pageau et al., 2003). Through this transfer Striga subtracts carbon assimilates (Rogers and Nelson, 1962; Okonkwo, 1966; Press et al., 1987b), water, nutrients (nitrate) and amino-acids (Pageau et al., 2003) from its host. 2 General introduction However, it was found that the loss in biomass production of a host plant caused by Striga infection largely outweighs the Striga biomass attached to it. It was therefore concluded that Striga does not only act as a sink for its host plant but has additional negative effects on the host plant (Press and Stewart, 1987). Upon Striga infection abscisic acid levels increase while levels of cytokinins and giberellic acid decrease (Drennan and El Hiweris, 1979). By changing this balance of plant growth regulators in the host, Striga negatively affects host photosynthesis (Press and Stewart, 1987; Gurney et al., 1995) and alters the biomass allocation of its host. More biomass is allocated to the roots at the expense of the stem (Graves et al., 1989). Furthermore Striga reduces the water use efficiency (Gebremedhin et al., 2000) and strongly affects the water economy of the host plant through its high transpiration rates (Shah et al., 1987; Press et al., 1987a, 1988; Ackroyd and Graves, 1997). These Striga-induced modifications of the host plant are thought to be the main causes for host yield loss. Control options Since Striga is primarily a problem in small-scale subsistence farming systems with few options for external inputs, control options must be low-cost and practical. A multitude of control options against Striga have been studied ranging from cultural measures like transplanting, delayed sowing or the use of trap crops (e.g. Doggett, 1988; Carsky et al., 1994b; Gbehounou and Adango, 2003; Gbehounou et al., 2004; Hess and Dodo, 2004), chemical control or soil fumigation (e.g. Bebawi and Eplee, 1986; Eplee and Norris, 1987; Carsky et al., 1994a), biological control (e.g. Kroschel and Muller Stover, 2004; Lendzemo et al., 2005) and host plant resistance (e.g. Williams, 1959; Kim et al., 1998). Despite the high potential of some of those solutions, no single option on its own has proven to be both sufficiently effective and durable as well as economically and practically applicable for low-input farming systems (Joel, 2000). Integration of various low-cost control options has proven to be a suitable approach (Berner et al., 1996; Schulz et al., 2003). An important element of this integrated approach is host plant defence. Two main groups of defence mechanisms against Striga can be distinguished: resistance and tolerance. Resistance against Striga reduces the infection level of a host plant, while tolerance enables the host plant to perform well, despite the parasitic infection. Host resistance is thought to be the most economical and potentially the most effective control option against root diseases and soil borne pathogens (Shew and Shew, 1994) and therefore a potentially acceptable Striga control option to resource-poor farmers (Hess and Ejeta, 1992; Debrah, 1994). Yet, complete resistance, or immunity, against 3 Chapter 1 Striga has not been found to date. Because few Striga infections can already seriously harm the host plant, resistance alone may not be enough to prevent crop losses. It is therefore recommended to direct breeding efforts towards finding varieties that combine resistance with high levels of tolerance (Haussmann et al., 2001a). Conceptions and knowledge gaps Over the past 75 years many breeders and researchers worked on resistance and tolerance against various species of Striga in a range of host plant species (e.g. Saunders, 1933; Williams, 1959; Doggett, 1965; Obilana, 1984; Ramaiah et al., 1990; Olivier et al., 1991; Hess et al., 1992; Efron, 1993; Cubero et al., 1994; Johnson et al., 1997; Kim et al., 1998; Haussmann et al., 2000a; Wilson et al., 2000). These efforts resulted in useful varieties with high resistance or tolerance and important insights in mechanisms behind these forms of defence. Examples of these achievements are the work on sorghum varieties Framida and SRN39 by various research groups (El Hiweris, 1987; Olivier et al., 1991; Hess and Ejeta, 1992; Arnaud et al., 1999; Mohamed et al., 2003). Still, many questions related to mechanisms behind host resistance and tolerance remain to be resolved. Subsequently, measures and methods for selection of good parent material for breeding need to be improved. Complete defence against Striga entails a combination of resistance and tolerance. While resistance lowers the number of Striga infections, tolerance reduces the negative effects of the infection. Breeding for resistant host plant varieties with superior levels of tolerance requires the separate selection of parental lines with either superior resistance or superior tolerance and hence appropriate selection measures for each trait. However, resistance and tolerance are often confounded, both in definitions and in selection measures. In Striga research resistance is often described as the mechanism that ensures lower infection and higher yields (Doggett, 1988; Hess and Haussmann, 1999). However, higher yields do not only depend on infection pressure (a result of resistance) but also on the consequence of infection on host performance (tolerance). Hence for identification of resistance one should focus on infection level alone. Resistance is often expressed in aboveground Striga numbers either at a fixed point in time or at its maximum (Olivier et al., 1991; Johnson et al., 1997; Adetimirin et al., 2000b; Wilson et al., 2000). It is, however, not clear whether this provides accurate information on what happens belowground and at what moment and what frequency the aboveground numbers should be counted. Tolerance is the ability of a variety to support equally severe levels of infection as other varieties of the same species, without the associated yield loss (Caldwell et al., 4 General introduction 1958; Doggett, 1988). The difficulty with the identification of tolerant lines is its entanglement with resistance. No one line will have exactly the same resistance level. Consequently, infection levels will also vary among tolerant lines. To assess the level of tolerance of a certain line, the Striga effects on yield need to be corrected for the infection load. But as long as the relation between Striga infection and yield loss is unknown, a fair correction cannot be carried out. Also, the physiological background for tolerance has been poorly understood. It was shown that tolerant varieties are often able to maintain high rates of photosynthesis under Striga infection (Gurney et al., 1995, 2002a). Principals behind this mechanism are not completely resolved yet and options for the application of physiological measurements to identify tolerant genotypes are not fully explored yet. It is expected that the development of fair and practical selection measures for tolerance, either based on crop yield or host plant physiological parameters, will greatly enhance breeding efforts against Striga. Host resistance is believed to reduce Striga seed production, through a reduction in Striga development rate or Striga numbers (Weber et al., 1995; Haussmann et al., 2000b). A reduction in aboveground Striga numbers, caused by resistance, does however not necessarily lead to a reduction in Striga seed production. The lower intraspecific competition with lower aboveground Striga numbers may enable higher seed production per individual Striga plant and hence compensate, at least partly, for the reduction in plant numbers. Whether or not selection for resistant host plant genotypes also implies selection for genotypes supporting lower Striga reproduction remains an important question to be solved. Objective and approach The objective of this study was to find suitable field selection measures that facilitate breeders in finding breeding material (genotypes) with superior levels of resistance and tolerance, serving both the short term goal of ensuring crop yield and the long term goal of lowering Striga seed bank density. The parasite - host plant interactions were studied with Striga (Striga hermonthica [Del.] Benth) and sorghum (Sorghum bicolor [L.] Moench). A selection of ten different sorghum genotypes with different levels and mechanisms of defence against Striga was used to study resistance and tolerance. This selection encompassed the sensitive and susceptible genotypes CK60-B and E36-1, the resistant genotypes N13 and Serena, the tolerant genotypes Seredo and Tiémarifing and the tolerant and resistant genotypes CMDT39, Framida, IS9830 and SRN39. The physiological work 5 Chapter 1 as well as the study on the relation between yield loss and infection was conducted with a selection of four of those ten genotypes: CK60-B, E36-1, Framida and Tiémarifing. Three field experiments (in 2001, 2002 and 2003) and two pot experiments (in 2001 and 2003) were conducted at the research station of the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) in Mali. One agar-gel test and two pot experiments (in 2003 and 2004) were conducted in the laboratory and the greenhouse of Wageningen University (WU) in The Netherlands. Field plots were artificially infested with known quantities of viable Striga seeds, over the whole surface to a depth of 5-10 cm, to simulate farmer field conditions. Adjacent Striga-free control plots were treated with ethylene gas (in cases of prior infestation) or kept free from Striga infestation (in the newly cleared Striga-free field of 2002). Field experiments were rain fed and lightly fertilized. Pot experiments consisted of pots with known Striga infestation levels and Striga-free controls. Sorghum plants in pots received regular water gifts to ensure non-water-limited conditions, and moderate fertilizer gifts comparable to the field experiments. Outline of the thesis In Chapter 2, selection measures for resistance are evaluated and selection measures and methods for tolerance are explored. This chapter particularly deals with the problems of quantifying host tolerance in the field. In the subsequent chapter (Chapter 3) the relations between Striga infestation, Striga infection and host plant yield loss are further studied and consequences and options for an adequate screening procedure for host plant tolerance are discussed. In Chapter 4, the host plant photosynthesis of tolerant versus sensitive genotypes is studied in order to enhance the understanding of some physiological principals that play a role in withstanding Striga effects. Additional objective of this study was to explore the options to use non-destructive and quick measurements as a selection tool for tolerance. The effects of host plant genotype and seed bank density on Striga reproduction are studied in Chapter 5. In the General discussion (Chapter 6) results of the present study are discussed and related to earlier work and outcomes of other studies. 6 CHAPTER 2 How can field selection for Striga resistance and tolerance in sorghum be improved?1 J. Rodenburga, L. Bastiaansa, E. Weltzienb, and D. E. Hessc a Group Crop and Weed Ecology, Wageningen University, Wageningen, the Netherlands b International Crops Research Institute for the Semi Arid Tropics, Bamako, Mali c Department of Agronomy, Purdue University, West Lafayette, Indiana, USA Abstract Breeding for high yielding Sorghum bicolor varieties with effective resistance and tolerance against the hemi-parasitic weed Striga hermonthica requires suitable selection measures for both characteristics. The objective of this research was to constitute a set of practical selection measures that contain independent, reliable and discriminative criteria for resistance and tolerance. Ten sorghum genotypes were grown in the field with and without Striga infestation in a split-plot design in 3 successive years (2001-2003) using different Striga infestation levels (low, high and intermediate). Resistance against Striga in the belowground stages was determined separately in an agar-gel assay and a pot trial. The addition of Striga-free control plots facilitated the calculation of the relative yield loss, which represents the result of resistance and tolerance combined. Correlation analysis indirectly demonstrated that both resistance and tolerance are important yield determining traits under Striga infestation. Tolerance was relatively more important under low Striga infestation levels, whereas resistance was relatively more important at high infestation levels. With respect to resistance, both the area under the Striga number progress curve (ASNPC) and maximum aboveground Striga number (NSmax) turned out to be discriminative and consistent selection measures. Both measures also corresponded well with the expression of resistance during belowground stages of the parasite. It proved more difficult to arrive at a satisfactory measure for tolerance. Inclusion of Striga-free plots is an essential step for the determination of tolerance, but in itself not sufficient. It provides a basis for the determination of the relative yield loss, which then needs to be corrected for differences in infection level resulting from genotypic differences in resistance. A linear correction for infection level disregards the density dependency of the relative yield loss function. It is expected that clarification of the relation between Striga infection level and yield loss, provides a solid basis for the development of unambiguous tolerance measures in the field. This will enable the breeder to select for resistance and tolerance separately, which is likely to result in the optimum combination of both defence mechanisms. 1 Published in: Field Crops Research 93 (2005) 34-50 7 Chapter 2 Introduction Striga hermonthica (Del.) Benth. (Scrophulariaceae, popular name: witchweed) is an out-crossing, obligate hemi-parasitic weed species that attacks roots of tropical Gramineae, including sorghum (Sorghum bicolor [L.] Moench), pearl millet (Pennisetum glaucum [L.] R. Br.), maize (Zea mays [L.]) and upland rice (Oryza sativa [L.]). Besides withdrawal of water, nutrients and assimilates, Striga damages its host by inducing enzyme and plant hormone changes, disrupting host water relations and carbon fixation (Press et al., 1996). According to Mboob (1989), 40% of the arable land in sub-Saharan Africa is infested with Striga. For six West African countries the total Striga-infested area was estimated at 5 million ha which is around 52% of the total grain production area (Sauerborn, 1991). Yield losses due to Striga infection of cereals in West Africa average 24% (10-31%), but in areas of heavy infestation losses reach 90-100% in some years (Sauerborn, 1991). Problems with Striga appear to be associated with degraded environments and are most severe in subsistence farming systems with little options for external inputs. Farmers are clearly in need of low-input solutions to Striga problems, for both the short and the long term. In the long term, the goal is to diminish Striga presence through depletion of Striga seed bank and limitation of Striga seed production (Obilana, 1988). In the short term, the goal is satisfactory grain yield under Striga infestation. Yield under Striga infestation is determined by the yield that would be achieved in the absence of Striga and the reduction caused by this biotic stress factor. This yield reduction is a function of the infection level and the response of the crop to this infection. Breeding for improved crop performance under Striga-infested conditions, which may benefit farmers without requiring high external inputs (Obilana, 1988), might consequently be focussed on resistance, to reduce the infection level, or on tolerance, to diminish the consequences of infection. According to the definitions of Parker and Riches (1993), resistance, the opposite of susceptibility, applies to genotypes that show fewer infections. A suitable selection measure for resistance should thus include the number of attached or emerged parasites. For practical reasons, selection for resistance is often based on number of aboveground Striga plants alone. A relevant question is whether this number is indeed a good selection criterion. Does it give a good reflection of the number of attached parasites? Furthermore, this number is the result of various belowground stages (e.g. germination, attachment, belowground development), and screening based on the overall result might unintentionally lead to the exclusion of 8 Field selection for Striga resistance and tolerance genotypes with a high level of partial resistance in one of these life-cycle stages. Such genotypes may in fact be good candidates for gene pyramiding. Resistance against Striga is sometimes used in a broader sense and described as a mechanism that ensures lower infection and higher (or satisfactory) host yields (Doggett, 1988; Hess and Haussmann, 1999). This definition not only includes the level of infection, but also the consequences of infection on host performance. Hence tolerance is included in this definition of resistance and no clear distinction is made between the two defence mechanisms (e.g. Kim et al., 2002). It is evident, that in the absence of immunity, the combination of resistance and tolerance is the most promising and durable breeding objective (Haussmann et al., 2001b). For obtaining the best combination of both traits, selection for both components separately seems the best approach. Tolerance, the opposite of sensitivity, is the ability to support equally severe levels of a pathogen, disease or parasitic weed as other varieties of the same species, without the associated impairment of growth or losses in grain yield or quality (Caldwell et al., 1958; Doggett, 1988; Ejeta et al., 1991). Tolerance on its own is difficult to quantify, as it is always confounded with a certain degree of resistance. Each genotype possesses its own level of resistance, making it difficult to directly assess the level of tolerance or compare the level of tolerance among genotypes. Furthermore, identification of tolerance requires Striga-free plots as a reference next to infested plots, as each genotype will have its own yield level, which will also be influenced by the specific environment where the screening takes place. The aforementioned constraints likely explain why research on defence against Striga in sorghum has been focussed more on resistance than on tolerance. A clear separation of tolerance and resistance as well as suitable characterisations for both traits seem beneficial to an efficient use of these defence mechanisms in crop improvement (Shew and Shew, 1994). Suitable measures should ideally meet various criteria like appropriateness (does the measure unambiguously represent the characteristic?), discriminativeness (is the measure making differences between genotypes sufficiently clear?), stability and objectivity (are selections based on the measure consistent over years and infestation levels?), repeatability (does the measure sufficiently express genetic variation?) and, last but not least, practicability (is the measure easy to determine?). The objective of this paper is to evaluate, improve and search for independent and practical field selection measures for resistance and tolerance against Striga hermonthica in sorghum, using Striga-free next to Strigainfested plots. 9 Chapter 2 Material and methods Genetic materials For all experiments, 10 sorghum genotypes were used: CK60-B, CMDT39, E36-1, Framida, IS9830, N13, Seredo, Serena, SRN39 and Tiémarifing. The objective was to use a range of genotypes that differed in degree and type of resistance and tolerance against Striga hermonthica (Table 1). Striga seed for field and pot infestation, was collected in Samanko (all experiments) and in Doumba, 80 km north-east of Samanko (agar-gel-assays only) and harvested from plants that parasitized sorghum. Field trials A series of field trials was conducted during three cropping seasons (2001-2003), at the ICRISAT-Mali field station in Samanko, 20 km south-west of Bamako, at the northern side of the river Niger (latitude 8°54”W and 12°54”N, altitude 329 m). Average mean temperature of the study site is 29.1°C during the cropping season (June-November). The climate type is Sudanese, characterised by one single rainy season between May and October. Mean annual rainfall at the field station is 950 mm, of which 96% falls between May and October. Experimental plots were laid on washed out, ferruginous tropical soils with wash-out spots and concretions and a sandy loam texture. Table 2 presents soil fertility parameters of the main plots of the three fields (2001, 2002 and 2003) after fertilization, as well as rainfall data of the three cropping seasons. In all years a split-plot design was used with either five (2001), eight (2002) or six (2003) replicates (Table 3). In 2001 and 2002 there were two main plot levels: Striga–free (control) and Striga-infested. In 2003 there were three main plot levels: Striga-free (control), low Striga infestation (L) and high Striga infestation (H). In each case, sorghum genotype was used as sub-plot factor. In each year a different field was used. The 2001 and 2003 experiments were sown in previously infested fields. Control plots were created through ethylene gas (C2H4, purity 99.98%) injections with a backpack ethylene applicator as described by Bebawi et al. (1985). The gas was injected twice, at a 4-day interval following a 0.5 - 0.5-m grid. Upon injection of the probe in the soil, gas was released for 3 s at a pressure of 3.5 bar. Ethylene injections resulted in nearly complete absence of Striga infection. The 2002 experiment was laid on a Striga-free field. Striga plots were created through artificial Striga infestation of the whole soil surface till a depth of 5 (2001) and 10 cm (2002 and 2003) with 45,000 (2001), 200,000 (2002), 30,000 and 150,000 viable Striga seeds m-2 (2003). 10 Field selection for Striga resistance and tolerance Table 1. Name, race, origin (NE = north-eastern, S = southern, E = eastern) and reported defence mechanism of the selected sorghum genotypes. Genotype Race Origin Defence mechanism CK60-B Kafir NE. Africa/ Sensitive/ USA Susceptible CMDT39 Guinea Mali Tolerant/ Resistant E36-1 Caudatum Ethiopia Susceptible Framida Caudatum S. Africa Tolerant/ Resistant IS9830 Caudatum Sudan Tolerant/ Resistant N13 Durra India Resistant Seredo Caudatum Uganda Tolerant Serena Caudatum E. Africa Resistant SRN39 Kafir Unknown Tolerant/ Resistant Tiémarifing Guinea Mali Tolerant Reference Olivier et al. (1991) ICRISAT/ IER (pers. commun.) ICRISAT (pers. commun.) El Hiweris (1987), Arnaud et al. (1996) El Hiweris (1987), Ramaiah (1988) Maiti et al. (1984) Haussmann et al. (2001a) El Hiweris (1987) El Hiweris (1987) ICRISAT (pers. commun.) Table 2. Soil fertility indicators: pH (H2O; 1:2.5), C-organic (% C.O.), P-available (Bray-1; mg P kg-1) and N-total (mg N kg-1) of the main plots of the study fields in 2001-2003 as determined shortly after fertilization, and cumulative rainfall (mm) at Samanko (Mali) for the three rainy seasons at three different moments (before sowing (at start), at 56 days after sowing (DAS) and at harvest). 2001 Control Striga pH 4.9 4.9 C-organic 0.3 0.3 P-available 10.3 9.2 N-Total 238.2 227.5 Cum. rainfall At start 233.1 At 56 DAS 758.5 At harvest 922.1 2002 Control 5.6 0.7 18.7 471.1 243.7 738.6 978.5 Striga 5.6 0.7 21.0 486.4 2003 Control 5.0 0.4 12.0 251.4 Striga (L) 4.9 0.4 12.2 248.4 Striga (H) 5.1 0.4 13.6 256.3 260.3 882.6 1147.3 In 2001, artificial Striga infestation was accomplished with seeds from 1998 (viability: 82.5%). In 2002 a mixture of Striga seeds was used from 1995, 1996, 1997 and 2001 (mean viability: 73%). In 2003 the mixture consisted of Striga seeds from 1995 to 1998 and 2001, but because of its low viability (10.5%) Striga seeds from 2002 (viability: 78.7%) were added to arrive at the desired infestation levels. 11 Chapter 2 Table 3. Information on field experiments in 2001- 2003 Year Parameter 2001 2002 2003 Replications 5 8 6 Fertilization 17-17-17 (N:P:K) kg ha -1 34-34-34 (N:P:K); 34-34-34 (N:P:K) -1 gypsum 100 kg ha kg ha-1 Sub-plot size 12.80 m2 24.32 m2 20.48 m2 Main-plot levels 2 (Striga, Striga-free) 2 (Striga, Striga-free) 3 (Striga low, Striga high, Striga-free) Spacing of plants 0.20-0.80 m 0.40-0.80 m 0.40-0.80 m Sowing date July 13 July 6 July 5 0 and 45,000 0 and 200,000 0, 30,000 and Striga infestation -2 levels (seeds m ) Striga infestation 150,000 0.05 m 0.10 m 0.10 m 1.60 m2/ 10 plants 3.20 m2/ 10 plants 2.56 m2/ 8 plants Two times None Two times depth Area/number of plants used to assess grain yield Ethylene injections Each sub-plot, representing one sorghum genotype, comprised four crop rows of 4.0 (2001), 7.6 (2002) and 6.4 m (2003) length with a row spacing of 0.8 m and a plant distance in the row of 0.2 (2001) and 0.4 m (2002 and 2003). After soil tillage (till 0.3 m depth), and levelling, the field was fertilised with 100 (2001) and 200 kg N-PK ha-1 (2002 and 2003) (17%N, 17%P, 17%K). In 2002 an additional 100 kg gypsum ha-1 was applied to raise soil pH. Sorghum was sown on 13 July 2001, 6 July 2002 and 5 July 2003 at six seeds per pocket and a depth of 2-4 cm. Plants were thinned to one plant per pocket at 21 days after sowing (DAS). Aboveground Striga numbers were counted every two weeks from Striga emergence till harvest of the crop. Simultaneously, in 2001 and 2002 Striga vigour scores, on a scale from 1 to 9, were given, depending on height and number of branches of individual plants (Haussmann et al., 2000b). Sorghum grain yield (Striga-infested and Striga-free) was determined, based on 10 (2001 and 2002) and 8 (2003) plants per sub-plot, representing an area of 1.6 (2001), 3.2 (2002) and 2.6 m2 (2003). Panicles were harvested at maturity and air dried before threshing and weighing. Maturity was determined for each genotype separately. 12 Field selection for Striga resistance and tolerance Resistance and tolerance of the various genotypes were estimated based on the field observations. Four Striga infection measures were used to indicate the level of resistance: (1) number of aboveground Striga plants at harvest (NSharvest), (2) maximum number of aboveground Striga plants (NSmax), (3) area under the aboveground Striga number progress curve (ASNPC) and (4) area under the Striga severity progress curve (ASVPC). Striga severity is the product of Striga number and Striga vigour score. The maximum number of aboveground Striga plants (NSmax) was introduced as, due to mortality, the maximum number was not always obtained at final harvest, but more often at earlier counts. The ASNPC, as outlined by Haussmann et al. (2000b) was calculated as: ASNPC = ∑ [S + S n −1 i i =0 (i + 1) 2](t (i + 1) − ti ) (1) where n is the number of Striga assessment dates, S i is the Striga number at the ith assessment date, ti the number of days after sowing at the ith assessment date. The ASNPC is a measure of the total Striga emergence throughout the season. ASVPC was calculated likewise, with Si representing the Striga severity score. Sorghum yield from Striga-free plots (YC; kg ha-1) was used as a control and represented the attainable yield. The attainable yield is the yield that could be obtained under the specific environmental conditions, in the absence of biotic stresses (Rabbinge, 1993). Combining this yield with the sorghum yield from adjacent Striga-infested plots (YS) was the basis for the derivation of tolerance measures. The first measure of tolerance was the relative yield loss due to Striga (RYL): RYL = (YC − YS ) YC (2) In an additional measure the RYL was divided by the maximum number of aboveground Striga plants, to obtain the RYL caused by a single Striga plant. This yields the second tolerance measure alinear. This measure implicitly assumes a linear relation between relative yield loss and Striga infection level. Pot trial A pot trial was conducted in 2001, at the same site as the field trials, in Samanko, Mali. The pot trial comprised a randomised block design in six replicates, with ten sorghum genotypes grown under Striga infestation. Plant distances were 0.35 m in the row and 0.7 m between rows. Pots of 10 L content were filled with 10 kg of a sand-soil-compost mixture (3:3:2). Striga infestation level was 4 viable Striga seeds cm-3 in the upper 5 cm (origin: Samanko, year: 1995, viability: 71.2%). After mixing through the soil, Striga seeds were preconditioned for 12 days in the pots. 13 Chapter 2 Sorghum was sown on 16 July (4-5 seeds per pot at 2-3 cm depth) and thinned to one plant per pot at 14 DAS. Number of below- and aboveground Striga plants (NSbg and NS ag respectively) were counted at 77 DAS. Laboratory trial Two agar-gel assays were conducted, in 2002 in a laboratory of Wageningen University, in Wageningen, The Netherlands, with ten sorghum genotypes and Striga seeds from two different locations in Mali (Samanko and Doumba) in eight replicates. The agar-gel assay developed by Hess et al. (1992) is a quick tool to screen sorghum genotypes for their ability to stimulate Striga seed germination. Agar-gel (0.7 % agar-agar) was added to a Petri dish containing sterilised and preconditioned (12 days at 28 oC in the dark) Striga seeds. The radicle of a 24 h old sorghum seedling was inserted in the solidified agar. After five days (at 28oC in the dark) the total number of Striga seeds as well as the number of germinated Striga seeds was counted and the fraction of germinated seeds (GS) calculated. Furthermore, the distance from the sorghum radicle to the furthermost germinated Striga seed (GD; mm) was determined. Statistical analyses An analysis of variance (ANOVA) was carried out to analyse the data, followed by a comparison of means with the least significant difference (L.S.D.) using the Genstat (release 6.1) statistical software package. To meet the assumptions of the analysis of variance some data were subjected to transformation prior to analysis, following procedures recommended by Sokal and Rohlf (1995, pp. 413-41). On field data involving Striga counts logarithmic transformations (log(X+c), where X is the original, individual observation and c=1.0) were applied. On belowground data involving counts with zeroes present, square root transformations ((X + c)1/2, where X is the original observation and c=0.5) were applied. Binomial distributed data, e.g. the fraction germinated Striga seeds, were subjected to a GLM regression analysis with binomial errors followed by a pairwise comparison of means by a t-test, in Genstat, following McCullagh and Nelder (1989, pp. 98-107) and Payne et al. (1993, pp. 413-26). Pearson’s correlations are presented throughout, based on treatment means, carried out with the SPSS (version 10.0) statistical software package. Correlations in this study were phenotypic correlations (r). Due to relative high environmental variation (see Results) genetic correlations could not be calculated. Repeatability (R) of resistance measures and yield were calculated following: R = (VG + VEg ) VP = 1 − (VEs VP ) 14 (3) Field selection for Striga resistance and tolerance where VP is the total phenotypic variance, which is composed of three components: (1) VG the genetic variance, (2) VEg the environmental variance due to permanent environmental effects on the phenotype and (3) VEs the environmental variance due to temporary or localized environmental effects on the phenotype (Falconer and Mackay, 1996, pp.136-37). Repeatabilities set an upper-limit to the heritability of a selection measure. Results Resistance Table 4 shows the mean, repeatability and ranking of all genotypes for each year and infestation level according to four different measures for resistance: NSharvest, NSmax, ASNPC and ASVPC. Only in 2003 the ASVPC was not determined. In 2002 and 2003H, the experiments with the highest infection levels, NSmax and ASNPC appeared more discriminative than NSharvest. Repeatabilities of NSmax and ASNPC were also higher than for NSharvest in most of the cases, except for 2003H. Comparison between measures shows that all measures, except NSharvest, appoint the same three most resistant genotypes within years. Also for the least resistant genotypes, ranking based on NS harvest deviated from that based on the other measures. There was a highly significant correlation between the different measures in all years except for NSharvest in 2002. In this year NSharvest did not show a significant correlation with one of the other resistance measures, while correlation between the other measures was still highly significant (Table 5). Ranking of most resistant and least resistant genotypes corresponded reasonably well between years, except for some cases. In 2001, representing the lowest infestation level, CMDT39 belonged to the group of three most resistant genotypes at the expense of IS9830. In 2002 (NSmax, ASNPC and ASVPC), CMDT39 was ranked within the group of three lowest resistant genotypes at the expense of Seredo. The three most resistant genotypes, based on NSmax and ASNPC, throughout the three years were N13, IS9830 and SRN39. CK60-B, E36-1 and Seredo showed to be poorly resistant, whereas CMDT39, Framida, Serena and Tiémarifing held an intermediate position. 15 Chapter 2 Table 4. Means, rankings (1-10) and repeatabilities (R) of different measures used to express resistance in the field in 2001, 2002 and 2003 (low infestation: L and high infestation: H). Mean Striga number at harvest (NSharvest), maximum aboveground Striga number (NSmax ), area under the Striga number progress curve (ASNPC) and area under the Striga severity progress curve (ASVPC). All measures are expressed per host plant. Year (level) 2001 2002 2003L 2003H 16 Genotype NSharvest CK60-B CMDT39 E36-1 Framida IS9830 N13 Seredo Serena SRN39 Tiémarifing c S.E.D R CK60-B CMDT39 E36-1 Framida IS9830 N13 Seredo Serena SRN39 Tiémarifing S.E.D. R CK60-B CMDT39 E36-1 Framida IS9830 N13 Seredo Serena SRN39 Tiémarifing S.E.D. R CK60-B CMDT39 E36-1 Framida IS9830 N13 Seredo Serena SRN39 Tiémarifing S.E.D. R 0.70 0.22 2.73 0.41 0.58 0.04 0.66 0.98 0.31 0.32 0.091 0.48 53.7 8.8 25.4 19.5 22.8 7.7 53.5 53.1 26.3 17.8 0.152 0.43 8.20 3.63 5.19 1.50 1.45 0.28 2.48 2.51 1.74 2.39 0.126 0.50 20.23 7.79 9.69 11.19 5.92 1.34 11.71 11.88 4.38 4.90 0.115 0.62 NSmax a b bc cd a bcd bcd d bc b bcd bcd 8 2 10 5 6 1 7 9 3 4 a cd b b b d a a ab bc 10 2 6 4 5 1 9 8 7 3 a bc ab d d e bcd bcd cd cd 10 8 9 3 2 1 6 7 4 5 a bcd bc b cd e ab ab d d 10 5 6 7 4 1 8 9 2 3 2.14 0.60 7.30 1.19 0.82 0.11 1.92 1.44 0.66 0.96 0.109 0.62 92.1 84.5 91.5 48.8 26.5 8.6 67.9 74.7 32.7 63.9 0.081 0.73 13.32 5.85 10.91 3.26 1.78 0.42 4.75 5.07 2.52 4.40 0.139 0.49 50.2 18.3 27.9 23.8 11.9 2.6 31.6 28.6 7.7 14.2 0.129 0.49 ASNPC b de a bcd cde e bc bcd de bcd 9 2 10 6 4 1 8 7 3 5 a a a b c d a ab c ab 10 8 9 4 2 1 6 7 3 5 a bc ab cde e f cd cd de de 10 8 9 4 2 1 6 7 3 5 a bcd ab bc de f ab ab e cde 10 5 7 6 3 1 9 8 2 4 73.3 16.1 187.4 34.3 16.0 3.9 60.2 33.3 23.8 29.5 0.255 0.48 3774.7 3356.4 3588.2 1895.7 925.8 308.0 2540.0 2876.4 1121.0 2448.1 0.074 0.84 473.2 165.3 307.3 97.6 47.9 5.6 138.0 162.7 47.9 146.2 0.256 0.55 1785.5 634.3 892.3 844.3 404.5 81.2 1139.2 951.8 290.1 508.3 0.148 0.67 ASVPC ab c a bc c d ab bc bc bc 9 3 10 7 2 1 8 6 4 5 a a ab ab bc bc c d d e 10 8 9 4 2 1 6 7 3 5 a ab ab bc c d bc ab c abc 10 8 9 4 2 1 5 7 3 6 a bcd bc bc de f ab abc e cde 10 5 7 6 3 1 9 8 2 4 226.0 31.5 473.2 62.4 32.7 6.7 145.6 68.5 53.5 61.8 0.302 0.46 31044.6 19723.2 17578.2 8413.0 4919.4 2141.9 10374.3 12501.6 5901.0 11375.3 0.117 0.66 ab d a bcd d e abc bcd cd bcd 9 2 10 6 3 1 8 7 4 5 a ab bc de e f cd bcd e cd 10 9 8 4 2 1 5 7 3 6 Field selection for Striga resistance and tolerance a Means in the same column followed by the same letter are not significantly different according to L.S.D. test (P<0.01). b Numbers 1-10 in the third column of each criterion, indicate ranking. c Data were analysed after log(X+1)-transformation. S.E.D.-values of transformed data are given. Means in table are back-transformed. Degrees of freedom: 36 (2001), 63 (2002) and 45 (2003). Table 5. Pearson’s correlation coefficients (one-tailed) between four different Striga resistance measures: Striga numbers at harvest (NSharvest ), maximum number of aboveground Striga plants (NSmax), area under the Striga number progress curve (ASNPC) and area under the Striga severity progress curve (ASVPC), for three different years, 2001, 2002 and 2003 (low infestation: L and high infestation: H) Year (level) Correlated traits 2001 2002 2003L a 2003H b NSharvest NSmax 0.975* 0.462 ns 0.977* 0.983* NSharvest ASNPC 0.947* 0.448 ns 0.984* 0.985* NSharvest ASVPC 0.923* 0.419 ns NSmax ASNPC 0.991* 0.998* 0.986* 0.997* NSmax ASVPC 0.974* 0.867* ASNPC ASVPC 0.993* 0.891* a Not significant; * Significant at the P< 0.01 level Belowground information A pot-trial was conducted to determine the extent to which the number of emerged Striga plants (aboveground: NSag) reflects the number of attached Striga plants (belowground: NSbg). The results presented in Table 6 show that the number of attached Striga plants correlated significantly with the number of emerged Striga plants (r=0.871, P<0.01). Repeatabilities of NSbg and NSag were however very low (0.25 and 0.31). By combining the results of the pot trial with an agar-gel assay it was assessed whether resistance against individual life-cycle stages of the parasite (germination, attachment and emergence) should be separately considered in the selection process. Table 6 shows the fraction of germinated seeds (GS) and the maximum germination distance from the sorghum root (GD) for the various genotypes. Germination of the two Striga batches with different origins did not differ significantly and consequently their results were combined. The two measures for germination stimulation (GS and GD) yielded similar results and correlated significantly with one another (r= 0.865, P< 0.01). None of the germination measures correlated significantly with numbers of attached or emerged Striga plants as observed in the pot experiment (r (GS-NSbg)=0.304; r (GS-NSag)=0.072). 17 Chapter 2 Table 6. Means, standard error’s (S.E.) or 95% confidence intervals (95% C.I.), repeatability (R) and rankings (1-10) of fraction of germinated Striga seeds (GS) and maximum germination distance (GD, mm) observed in the agar-gel tests and mean number of Striga attachments (NSbg) and emergence (NSag) at 77 DAS from the pot trial. Data are expressed per sorghum plant or sorghum seedling. Germination Genotype GSa S.E. GD (mm) b 95% C.I. CK60B 0.0258 0.0090 b 4 3.67 [1.80, 6.11] d 4 CMDT39 0.0974 0.0183 cd 9 13.06 [8.85, 18.04] ab 7 E36-1 0.1572 0.0196 d 10 17.72 [11.06, 25.90] ab 9 Framida 0.0003 0.0008 a 2 0.15 [0.0, 0.56] e 1 IS9830 0.0016 0.0019 a 3 0.41 [0.0, 1.01] e 3 N13 0.0788 0.0129 c 7 18.15 [11.51, 26.26] a 10 Seredo 0.0966 0.0146 cd 8 7.16 [3.55, 11.89] cd 5 Serena 0.0613 0.0112 bc 5 11.49 [6.11, 18.47] bc 6 SRN39 0.0003 0.0008 a 1 0.33 [0.0, 1.29] e 2 Tiémarifing 0.0738 0.0133 c 6 13.20 [8.29, 19.21] ab 8 R 0.57 Attachment and Emergence Genotype NSbgb 95% C.I. NSagb 95% C.I. CK60B 5.65 [3.97, 7.77] a 9 7.51 [2.63, 9.96] a 10 CMDT39 3.42 [2.29, 4.41] abc 5 2.74 [0.0, 6.50] abcd 7 E36-1 5.75 [1.85, 10.19] a 10 4.38 [0.18, 8.38] ab 8 Framida 4.70 [0.62, 9.95] ab 8 4.25 [0.0, 9.67] abc 9 IS9830 0.71 [0.00, 2.10] c 1 0.62 [0.0, 1.28] cd 2 N13 1.43 [-0.03, 4.30] bc 3 0.21 [0.0, 0.85] d 1 Seredo 2.19 [1.40, 3.65] abc 4 2.70 [0.48, 3.65] abcd 6 Serena 3.30 [0.93, 8.47] abc 7 1.78 [0.12, 2.98] bcd 5 SRN39 1.69 [-0.19, 3.08] abc 2 0.80 [0.0, 1.67] bcd 3 Tiémarifing 3.26 [1.19, 5.65] abc 6 1.32 [0.0, 2.28] bcd 4 R 0.25 0.31 GS has a binomial distribution and is analysed with a GLM regression analysis, degrees of freedom: 158. b )-1/2 Means of GD, NSbg and NSag are back-transformed from ANOVA with (X+0.5 transformed data. Means followed by the same letter are not different at the P=0.001 level of significance for GD and at the P=0.01 level of significance for GS, NSbg and NSag. Numbers 1-10 in the fourth column of each criterion, indicate ranking. Degrees of freedom are 159 (GD) and 45 (NSbg and NSag). a 18 Field selection for Striga resistance and tolerance Table 7. Means and rankings of 10 sorghum genotypes for grain yield (kg ha-1) under Striga infestation (YS) and control conditions (YC), relative yield loss due to Striga (RYL) and relative yield loss per Striga plant (alinear) per year (2001-2003) and level (L or H). Year (level) 2001 Genotype YS YC RYL alinear a b CK60-B 352 c 10 1093 abc 5 0.68 10 0.297 CMDT39 816 abc 6 1019 abc 6 0.20 5 0.321 E36-1 799 abc 7 798 bc 9 0.00 1 0.000 Framida 1164 ab 3 1481 a 2 0.21 8 0.162 IS9830 1405 a 1 1438 ab 4 0.02 2 0.024 N13 501 c 9 761 c 10 0.34 7 2.849 Seredo 1237 ab 2 1564 a 1 0.21 6 0.094 Serena 631 bc 8 1480 a 3 0.57 9 0.326 SRN39 888 abc 4 988 abc 7 0.10 4 0.144 Tiémarifing 886 abc 5 979 abc 8 0.09 3 0.083 S.E.D. 307.0 315.8 c 0.21 0.14 R 2002 CK60-B 188 e 10 1072 de 9 0.82 9 0.0088 CMDT39 333 de 9 1589 cd 7 0.79 8 0.0089 E36-1 346 de 8 2203 ab 4 0.84 10 0.0089 Framida 1543 b 2 2400 ab 3 0.36 4 0.0065 IS9830 2434 a 1 2178 ab 5 -0.12 1 -0.0041 N13 792 cd 5 900 e 10 0.12 2 0.0124 Seredo 1185 bc 3 2522 a 1 0.53 5 0.0064 Serena 698 cd 7 2477 a 2 0.72 7 0.0091 SRN39 990 c 4 1146 de 8 0.14 3 0.0040 Tiémarifing 711 cd 6 1893 bc 6 0.62 6 0.0094 S.E.D. 248.7 291.2 R 0.63 0.50 2003L CK60-B 546 e 10 1174 ef 9 0.53 10 0.0236 CMDT39 1481 bc 5 1955 bc 6 0.24 7 0.0332 E36-1 1063 cd 8 1970 bc 4 0.46 9 0.0231 Framida 1743 ab 3 1812 cd 7 0.04 1 0.0060 IS9830 1693 ab 4 2030 bc 3 0.17 2 0.0452 N13 702 de 9 931 f 10 0.25 6 0.2860 Seredo 1747 ab 2 2289 b 2 0.24 4 0.0239 Serena 1986 a 1 2658 a 1 0.25 5 0.0303 SRN39 1115 cd 7 1501 de 8 0.26 3 0.0568 Tiémarifing 1445 bc 6 1967 bc 5 0.27 8 0.0533 S.E.D. 217.1 182.9 R 0.59 0.71 2003H CK60-B 288 e 10 1174 ef 9 0.75 9 0.0113 CMDT39 1206 abc 3 1955 bc 6 0.38 4 0.0115 E36-1 411 de 9 1970 bc 4 0.79 10 0.0150 Framida 921 bcd 5 1812 cd 7 0.49 6 0.0121 IS9830 1576 a 1 2030 bc 3 0.22 1 0.0124 N13 708 de 8 931 f 10 0.24 2 0.0599 Seredo 863 bcd 6 2289 b 2 0.62 8 0.0144 Serena 1133 abc 4 2658 a 1 0.57 7 0.0152 SRN39 861 bcd 7 1501 de 8 0.43 5 0.0229 Tiémarifing 1327 ab 2 1967 bc 5 0.33 3 0.0109 S.E.D. 264.9 182.9 R 0.37 0.71 a Means in the same column followed by the same letter are not significant different according to the L.S.D. test (P<0.001). Exceptions are: YC 2001 (P=0.096) and YS 2001 (P=0.037). Degrees of freedom: 36 (2001), 63 (2002) and 45 (2003L and H). b c Numbers 1-10 in every second or third column, indicate ranking. R means Repeatability. 7 9 1 6 2 10 4 8 5 3 5 7 6 4 1 10 3 8 2 9 3 6 2 1 7 10 4 5 9 8 2 3 7 4 5 10 6 8 9 1 19 Chapter 2 These data showed low stimulation of germination (GS) and low numbers of attachments and emergence (NSbg and NS ag) at IS9830 and SRN39 and an absence of resistance in any of these stages for E36-1. At Framida and CK60-B, GS was low and medium-to-low but NSbg and NSag were relatively high, whereas at N13, GS was high but NSbg and NSag very low. Serena, Seredo, Tiémarifing and CMDT39 held an intermediate position in every stage. Tolerance Table 7 presents yield under Striga infestation (YS), yield under Striga-free conditions (YC), relative yield loss due to Striga (RYL) and relative yield loss per maximum aboveground Striga plant (alinear). The RYL was calculated directly from the yields presented in Table 7. The alinear was calculated by dividing RYL by the maximum number of aboveground Striga plants (NSmax, Table 4). In 2002 and 2003, YC was much higher (on average 1.6 times) than in 2001 for nearly all genotypes. Exceptions were CK60-B and N13 in 2002 and 2003 and Framida in 2003. For YS large differences in ranking between years were observed. CK60-B and E36-1 were consistently ranked within the group of lowest yielding genotypes. IS9830 and Framida belonged consistently to the highest yielding genotypes under Strigainfested conditions, except for Framida in 2003H. Tiémarifing was a rather constant intermediate genotype, concerning YS. Only in 2003H it was ranked somewhat higher. The repeatability of YS was low, especially in 2001 (0.21). This indicates a low upper-limit of heritability and a large contribution of environmental variation to the phenotypic variation of this trait. Rankings based on RYL were not very consistent. Throughout the years, seven genotypes were ranked among the three genotypes with the highest RYL. Only CK60-B (four times) and E36-1 (three times) appeared more than once in this group. Six genotypes were ranked among the three genotypes with the lowest RYL and only IS9830 appeared more than twice in this group. Relative yield loss is the result of resistance and tolerance combined. For a fair assessment of tolerance, the RYL needs to be corrected for infection level. The alinear expresses the average relative yield loss per emerged Striga plant. Correction of RYL for the infection level had important consequences for the ranking of the different genotypes. In 2003, CK60-B was the genotype that suffered most from Striga infection but if relative yield loss was related to the number of infections it was found that the yield loss per Striga plant was modest. For N13 exactly the opposite was found. Compared to the other genotypes RYL was either moderate (2003L) or even low (2003H). Relating this RYL to the number of Striga plants revealed that with this genotype the damage per Striga plant was by far the largest. The three most tolerant 20 Field selection for Striga resistance and tolerance genotypes based on alinear were difficult to identify due to inconsistency throughout the years and infestation levels. Table 7 shows that over the years and infestation levels, eight genotypes were ranked as most tolerant based on alinear, of which four of them only once (Seredo, SRN39, Framida and CMDT39). The other four genotypes all belonged two times to the group of three most tolerant genotypes (E36-1, Tiémarifing, IS9830, and CK60-B). Among the group of eight genotypes Tiémarifing (two times), SRN39, and CMDT39 were also ranked among the three least tolerant genotypes in other years or infestation levels. Phenotypic correlations In this study resistance, tolerance and yield under Striga-free conditions were used as a complementary set of traits that together determine yield under Striga. From a breeding perspective it is relevant to find out how well each of these traits correlates to the yield under Striga infestation, as an indication for their significance. Table 8 shows results of the phenotypic correlations between yield under Striga infestation (YS) and control yield (YC), relative yield loss (RYL), and maximum number of emerged Striga plants (NSmax). NS max represents resistance, whereas RYL represents the outcome of all defence mechanisms combined including resistance. Only in the two low infested fields (2001 and 2003L), YC was found to correlate significantly with YS (r= 0.584 and 0.886, P= 0.038 and < 0.01, respectively). The RYL was found to correlate significantly with YS in all situations. Significance of this correlation increased with infestation level (going from the lowest to the highest infested fields: P= 0.013, 0.016, 0.008 and 0.002). The NSmax correlated significantly with YS only in the highest infested field (2002; r= -0.633, P=0.025). A significant correlation between RYL and NSmax was found in all situations, except in 2001, the lowest infested field. Table 8. Pearson’s correlations coefficients between yield under Striga infestation (YS), yielding ability (YC), maximum Striga number (NSmax) and the relative yield loss (RYL) for 2001, 2002, 2003L (low Striga infestation level), and 2003H (high Striga infestation level). Year (level) Correlated traits 2001 2002 2003L 2003H YC 0.584* 0.390 0.886** 0.506 YS RYL -0.692* -0.809** -0.674* -0.730** YS NSmax -0.079 -0.633* -0.383 -0.521 0.835** 0.849** YS a RYL NSmax -0.218 0.944** Correlations are one-tailed. * Correlation is significant at the 0.05 level of significance. ** Correlation is significant at the 0.01 level of significance. a 21 Chapter 2 Discussion Factors determining yield under Striga infestation Abiotic growth factors, like temperature, radiation and availability of water and nutrients, combined with the physiological and morphological characteristics of a genotype determine the attainable yield of a crop (Rabbinge, 1993). The actual yield will in general be lower than the attainable yield, due to the presence of biotic stress factors, like Striga. Yield reduction due to Striga is determined by the infection level and the consequences of infection for crop production. Analogous to this, the defence mechanism of a crop can be separated into resistance, the ability to reduce the infection level, and tolerance, the ability to minimize the consequences of infection. Results of this study show that the correlation between RYL, representing the effect of resistance and tolerance combined, and the yield under Striga infestation becomes stronger with an increase in infestation level. Simultaneously, the correlation between attainable yield and yield under Striga infestation decreases at higher infestation levels. Moreover, the correlation study demonstrates that at high infestation levels resistance becomes an increasingly important component of the overall defence mechanism against Striga. Implicitly this suggests that tolerance is a relatively more important mechanism at low infestation levels. Combining host plant resistance with tolerance and high yielding ability has often been proposed as durable control measure against parasitic angiosperms (Kim, 1991; DeVries, 2000; Kling et al., 2000; Haussmann et al., 2001a,b; Pierce et al., 2003; Showemimo, 2003). Our findings support this approach. For obtaining the best combination of traits, the potentially best sources of resistance, tolerance and yielding ability need to be identified. In breeding programs against Striga, the number of emerged Striga plants, and the yield under Striga infestation are often important selection criteria. Selection based on those two traits alone unintentionally ignores tolerance. This can be illustrated by the results of CMDT39 and E36-1 in 2001. These genotypes had equal yields under Striga (816 and 799 kg.ha-1, respectively) but a significant difference in number of emerged Striga plants (0.6 and 7.3, respectively). In such a situation screening based on yield and Striga number alone would favour the genotype with the lowest Striga number (CMDT39) which implies a negative selection for tolerance. This could be avoided if a proper selection measure for tolerance would be available. For this reason this study explored the opportunities for defining a practical set of field selection measures that takes into account both resistance and tolerance. To achieve this, a group of genotypes was selected with a wide range of modes and levels of defence mechanisms against Striga. As a result the selected 22 Field selection for Striga resistance and tolerance group of genotypes consisted of different sorghum races (Guinea, Caudatum, Kafir and Durra) and origins with only two local sorghum genotypes (CMDT39 and Tiémarifing). The specific levels of control yield, tolerance and resistance of the various sorghum genotypes in this study may therefore be affected by genotype x environment interactions and Striga population (e.g. Botanga et al., 2002; Oswald and Ransom, 2004). For this reason it is often recommended to screen at multiple locations and with different Striga populations (Ramaiah, 1987a; Haussmann et al., 2000b; Omanya et al., 2004). However, the aim of this study was not to identify the best genotypes but to evaluate and improve the current screening procedures and measures. Complexity of tolerance Screening for tolerance requires a field design with Striga-free control plots next to Striga-infested plots. As sorghum yield is determined by many environmental factors, this set-up offers the best possibility for estimating the gap between attainable and actual yield. The ratio between this gap and the attainable yield expresses the relative yield loss (RYL). So far, only few studies have used a factorial design with Striga-infested and Striga-free control plots in the same field (Efron, 1993; Kim and Adetimirin, 1997a; Gurney et al., 1999; Adetimirin et al., 2000a,b; Kim et al., 2002). It requires infesting Striga free fields (Efron, 1993; this study), which is not always possible, or the creation of Striga-free control plots within Striga-infested fields. Technically this can be achieved by using ethylene gas (this study) or methyl bromide (Gurney et al., 1999) but this is very expensive. Furthermore, ethylene injections do not guarantee total absence of Striga (personal observation). In some situations it is already possible to separate tolerance from resistance based on RYL and infection level. In 2001 for instance, yield of E36-1 under Strigainfested conditions was identical to the yield under Striga-free conditions despite a relatively high infection level (NS max: 7.3 plants per host plant). This indicates the presence of a tolerance mechanism. For N13, with a mean NSmax of only 0.1, resistance seems the most important mechanism. However, not in all cases it is so easy to disentangle the contribution of tolerance and resistance to the overall defence mechanism. As mentioned earlier, tolerance is defined as the reaction of genotypes that germinate and support as many Striga plants as other genotypes without the same severity of yield reductions. In reality however, as shown in this study, clear differences in Striga infection level exist between genotypes. This implies that for obtaining an independent measure for tolerance, the yield reduction due to Striga should be corrected for Striga infection level. Consequently, RYL in 23 Chapter 2 itself is not an independent measure of tolerance, as it is always confounded with resistance. The high correspondence between the ranking based on NS max and the ranking based on RYL in 2002 for instance follows from the fact that resistance is included in RYL. As RYL depends on both resistance and tolerance, it is not surprising that rankings based on RYL are inconsistent over years. Infestation levels varied over years and, as earlier demonstrated, the importance of resistance and tolerance varies with infestation level. The importance of correction for Striga infection level is also demonstrated by data published by Efron (1993). Correction of the RYL of the low resistant maize hybrid 8338-1 for the simultaneously observed Striga counts, would appoint this genotype as the most tolerant instead of the most sensitive one. Contrary to earlier statements made by Kim (1991) and Efron (1993) Striga counts may be very important for the accurate assessment of tolerance. However, simply expressing the relative yield loss per aboveground Striga plant proved to be insufficient. Such a linear correction for infection pressure assumes an identical negative effect of every additional Striga plant on yield. Data presented in Table 7 illustrate this assumption to be incorrect. With an increase in aboveground Striga numbers, the a linear decreases drastically (e.g. 2001 vs. 2002). Additional evidence that the relation between RYL and Striga infection level is not linear is provided by data on CK60-B in Table 7. At a very low infection level (2001) already a RYL of 60% was attained, while at a 40 times higher infection level (2002) the RYL was only 82%. For a proper assessment of tolerance in the field, one needs to know how to correct for genotype-dependent differences in Striga infection level. This means that the relation between Striga infection and yield loss should be known. The correction factor for Striga infection should be obtainable from field observations, and preferably be based on an aboveground resistance measure such as NSmax. With nonparasitic weeds that mainly affect crop plants through resource competition, a progressively declining yield loss with increasing weed numbers is generally observed (e.g. Weaver et al., 1987; Spitters et al., 1989). This relation can be accurately described by a rectangular hyperbola, which is characterised by the initial slope, the yield loss caused by the first weed added to a weed free crop, and the maximum yield loss at high weed density (Cousens, 1985). Webb and Smith (1996) suggested that a similar relation would hold for parasitic weeds. For a single sorghum genotype, Gurney et al. (1999, 2000) observed a declining marginal yield loss with increasing Striga dry weight. Although Striga dry weight is not a straightforward resistance measure and not linearly related to Striga number, the observation confirms that the relation between yield loss and infection level is not proportional. 24 Field selection for Striga resistance and tolerance The initial slope (ahyperbolic) of the assumed hyperbolic relation between relative yield loss and number of Striga plants (NSmax or ASNPC), representing the yield reduction due to the very first Striga plant, could be a good measure to express tolerance. A preliminary calculation of the ahyperbolic was made, under the assumption that for each of the genotypes ultimately a maximum relative yield loss of 100% would be obtained. As expected, the rankings of alinear and ahyperbolic proved to be reasonably comparable at low infection levels (2001 and 2003L) but deviated significantly at higher infection levels (2002 and 2003H). However, the current data suggest that with genotypes such as IS9830 and Framida severe Striga infection will never result in complete failure of the host. This implies that tolerance might be characterised by two components: (1) the initial slope of the relation between relative yield loss and Striga infection level and (2) the attainable relative yield loss. It will then be valuable to assess tolerance at least at two infection levels: low (infection initiation), to get a good estimation of the initial slope, and high (infection saturation), to estimate the maximum relative yield loss. Furthermore, it is not evident that the relation between relative yield loss and Striga infection always obeys the same function. For instance, observations on E36-1 show that some genotypes may be very tolerant at low infection levels and very sensitive at high infection levels. This indicates the possible presence of an infection threshold beyond which the initial tolerance collapses. Further research is needed to resolve the relation between relative yield loss and Striga infection, and to investigate whether a similar relation holds for all Striga hosts (independent of genotype). This should lead to a practical field selection measure, which helps the cereal breeder to identify genotypes with superior tolerance. Field selection measure for resistance A reliable resistance measure is a prerequisite for the identification of both resistance and tolerance. Of the resistance measures, the Striga number at harvest (NSharvest) is an easy measure to obtain but not very discriminative. Moreover, selection based on NS harvest proved to be insufficiently consistent over years and infestation levels. This trait was characterised by low repeatabilities, especially in 2001 and 2002, implying large contributions of environmental and error variation to the phenotypic variation. Moreover, harvest time is genotype dependent and determines to a large extent the fraction of emerged Striga plants that still remain at the time of observation. The area under the Striga number progress curve, ASNPC, as introduced by Haussmann et al. (2000b) is an appropriate measure as it incorporates infection time. In order to avoid differences caused by the genotypedependent length of the growing season (harvest moment), the ASNPC was 25 Chapter 2 calculated between two fixed points in time (39 and 102 DAS) for all genotypes and all years. The ASNPC demonstrated to be one of the most discriminative, objective and complete measures. Repeatabilities of ASNPC were reasonably high, which confirms results of Omanya et al. (2004). Only in 2001, with a low infection level, repeatability was rather low. The ASVPC is considered less suitable as resistance measure because vigour scores are due to subjectivity and might also be affected by host tolerance. This might explain the somewhat lower repeatabilities observed for ASVPC compared to the repeatabilities of NSmax and ASNPC. Omanya et al. (2004) reported that expression of genetic variation (by sorghum genotypes) for vigour scores is rather inconsistent. Furthermore, assigning appropriate vigour scores to the counted Striga plants, requires additional time. Maximum aboveground number of Striga plants (NSmax), earlier used, with millet, by Wilson et al (2000, 2004), turned out to be a more objective measure than counts at harvest time. It proved to be very consistent over years and equally discriminative as the ASNPC. Correlation between NSmax and ASNPC was found to be highly significant irrespective of year and infestation level. A slight advantage of NSmax over ASNPC is that one could save time because regular counts can be started later, around the time when the maximum number of aboveground Striga plants is expected. Still more than one count is required for determining NS max, as it is not known on beforehand when exactly the maximum can be found and this moment will also differ between genotypes. Adetimirin et al. (2000b) who worked with maize, and Omanya et al. (2004), working with sorghum, proposed a single count at around 56 DAS and 77 DAS respectively. Additional analyses in the current study revealed that Striga numbers around 77 DAS correlated better with ASNPC and NSmax, and had a higher mean repeatability (averaged over years, R=0.64) than Striga numbers at 56 DAS (R=0.39). Selection based on a single count around 77 DAS is therefore expected to correspond well with selection based on ASNPC or NSmax. Usefulness of belowground observations Kim (1996) and Ejeta et al. (2000) stressed the importance of belowground Striga observations in the assessment of resistance. Because this kind of observations is difficult to make in the field, one has to find other media, such as Petri-dishes and pots to study belowground processes. Techniques, such as the agar-gel test or a pot trial, permit the researcher to get insight in resistance during the stages that are most harmful for the crop and to acquire this information within a relatively short period of time and at low costs (Omanya et al., 2004). Disadvantages of pot trials are its high labour requirements, artificial root conditions and, according to Haussmann et al. (2000b) and Omanya et al. (2000), inconsistent correlation with field 26 Field selection for Striga resistance and tolerance experiments. Results from the pot trial presented in this study showed nevertheless a ranking that corresponded reasonably well with the ranking based on maximum number of emerged Striga plants in the field. However, the 95% confidence intervals for NSbg and NSag, were very large and the repeatabilities of these measures were very low (0.25 for NSbg and 0.31 for NSag) which confirms earlier results from Omanya et al.(2004). The absence of correlation between the germination measures from the agar-gel test and the numbers of attached and emerged Striga plants in the pot trial suggests that genotypes with an effective belowground resistance mechanism in a very specific stage (germination) are not necessarily identified by aboveground counts. Therefore screening with the help of assays that only address a very specific life-cycle stage is indeed useful for detecting specific resistance mechanisms. This observation confirms earlier statements from Kim (1996) and Ejeta et al. (2000). Combination of aboveground measures and information on germination stimulation revealed a very effective resistance mechanism in N13. This genotype stimulates abundant Striga seed germination which nevertheless resulted in extreme low numbers of Striga infection. This suggests the presence of a resistance mechanism that operates after germination stimulation. For that reason, genotypes with high germination stimulation should not be discarded as they might have valuable other sources of resistance. Results from CK60-B show that low germination stimulation on its own is not a useful characteristic, as it can still result in abundant parasitism. These observations indicate that in a selection process genotypes should never be selected or rejected after evaluation of a single resistance mechanism alone. Following the ranking of resistance based on a single mechanism, SRN39, Framida and IS9830 (germination stage) and N13 (attachment stage) would be good sources for pyramiding resistance genes. This confirms results from Maiti et al. (1984), Ramaiah (1984, 1987a), Vasudeva Rao (1984), El Hiweris (1987), Olivier et al. (1991), Hess et al. (1992), Ejeta et al. (2000), Heller and Wegmann (2000), and Omanya et al. (2004). In conclusion, the maximum number of aboveground Striga plants showed to be a reliable measure for resistance as a reasonable correspondence between number of belowground attachments and maximum number of emerged Striga plants was observed. This measure also proved to be discriminative and consistent over years. Screening based on number of aboveground Striga plants in combination with yield under Striga infestation is likely to result in a negative selection for tolerance. The addition of Striga-free control plots allows the determination of the relative yield loss, which represents the effect of resistance and tolerance combined. Relative yield loss itself was found to be an inconsistent screening measure. The reason for 27 Chapter 2 this inconsistency might be that the relative contribution of resistance and tolerance to the overall defence against Striga depends on Striga infestation level. Tolerance was found to be relatively more important at low infestation levels, whereas resistance was found to be more important at high infestation levels. A fair comparison of tolerance among genotypes is difficult to make, as genotypic differences in resistance cause major differences in infection level. Corrections for these differences in infection level are difficult to make as long as the relation between relative yield loss and Striga infection level is not resolved. After clarification of this relation an independent tolerance measure can be derived. This will facilitate the breeder to identify genotypes with superior tolerance against Striga in the field. 28 CHAPTER 3 Characterization of host tolerance to Striga hermonthica1 J. Rodenburga , L. Bastiaansa and M. J. Kropffa a Group Crop and Weed Ecology, Wageningen University, Wageningen, The Netherlands Abstract One of the most promising control options against the parasitic weed Striga hermonthica is the use of crop varieties that combine resistance with high levels of tolerance. The aim of this study was to clarify the relation between Striga infestation level, Striga infection level and relative yield loss of sorghum (Sorghum bicolor) and to use this insight for exploring the options for a proper screening procedure for tolerance. Pot experiments in which four sorghum genotypes were exposed to a range of Striga infestation levels, ranging from 0.0625 to 16 seeds cm-3, were conducted in Mali in 2003 and in a greenhouse in The Netherlands in 2003 and 2004. Observations included regular Striga emergence counts and sorghum grain yield at maturity. There were significant genotype, infestation and genotype × infestation effects on sorghum yield. The relation between infestation level and infection level was density dependent. Furthermore, the relation between Striga infection level and relative yield loss was non-linear, though for the most resistant genotype Framida only the linear part of the relation was obtained, as even at high infestation levels only moderate infection levels were achieved. The results suggest that for resistant genotypes, tolerance can best be quantified as a reduced relative yield loss per aboveground Striga plant, whereas for less resistant genotypes the maximum relative yield loss can best be used. Whether both expressions of tolerance are interrelated could not be resolved. Complications of screening for tolerance under field conditions are discussed. 1 Euphytica (accepted) 29 Chapter 3 Introduction The obligate hemi-parasitic weed Striga hermonthica (Del.) Benth is a major constraint to cereal production in the semi-arid to sub-humid tropics of Africa. Yields of host plants infected by Striga can be severely reduced (Obilana, 1983; Rodenburg et al., 2005). Striga attacks most of the tropical Gramineae species, including several important agricultural species like sorghum (Sorghum bicolor [L.] Moench), pearl millet (Pennisetum glaucum [L.] R. Br.), maize (Zea mays [L.]) and upland rice (both Oryza glaberrima [Steudel] and O. sativa [L.] [Johnson et al., 1997]). One of the most promising control options against Striga is the use of crop varieties with improved levels of resistance and tolerance against this parasite. Resistant genotypes have fewer infections, while tolerant genotypes show less impairment of growth or losses in grain yield when exposed to similar levels of infection than other varieties of the same species (Parker and Riches, 1993). The converse of resistance is susceptibility, while the converse of tolerance is sensitivity. Every host genotype combines a specific level of resistance with a specific level of tolerance. Breeding for those characteristics requires suitable selection criteria. Many different selection measures have been developed for resistance. All of these measures are based on the number of aboveground Striga plants and vary from a single count at a specific moment in time (Adetimirin et al., 2000b; Omanya et al., 2004) or the maximum number of aboveground Striga plants (Wilson et al., 2000; 2004; Rodenburg et al., 2005) to the area under the Striga number progress curve (ASNPC) (Haussmann et al., 2000b; Omanya et al., 2004; Rodenburg et al., 2005). Complete resistance, also referred to as immunity against Striga, has not yet been found. Therefore, a host variety that combines superior levels of resistance and tolerance is an obvious breeding objective and has been proposed in many studies (e.g. Kim, 1991; DeVries, 2000; Kling et al., 2000; Haussmann et al., 2001a,b; Pierce et al., 2003; Showemimo, 2003; Rodenburg et al., 2005). Different measures of tolerance have been proposed, ranging from host plant damage scores to yield, yield loss, or relative yield loss under Striga infestation (Efron, 1993; Kim, 1994; Adetimirin et al., 2000b; Gurney et al., 2002a; Kim et al., 2002). None of these measures account for the difference in resistance among genotypes and hence they ignore the fact that the observed damage is due both to Striga infection level (resistance) and the extent to which the specific genotype endures these infections (tolerance). Consequently, differences among genotypes in level of yield reduction cannot simply be attributed to tolerance only. It seems that the only way to obtain an unbiased comparison of the level of tolerance among genotypes would be to create identical infection levels for all genotypes. Theoretically this might be achieved 30 Characterization of tolerance to Striga by exposing all genotypes to a range of infestation levels. However, realization of such a range under field conditions is difficult, if not impossible, and definitely costly. Another alternative might be to correct the observed damage of each genotype for its Striga infection level. Such a correction requires that the relation between Striga infection level and yield loss is known. Studies in which the biomass of the parasite was used as infection measure suggest that the relation between Striga infection level and yield loss is non-linear and characterized by a diminishing slope with increasing infection level (Gurney et al., 1999; 2000; Rodenburg et al., 2005). Whether this type of relation also holds for the relation between Striga number and yield loss is not yet clear. Nor is it known whether such a relation has a general validity or is genotype specific. The aim of this study was to resolve the relationship between Striga infestation level, Striga infection level and yield loss for a number of sorghum genotypes, and to explore options for the development of a screening procedure for tolerance to Striga infection. Material and methods Experimental sites and plant material Four sorghum (Sorghum bicolor L. Moench) genotypes were grown at a range of Striga (Striga hermonthica (Del.) Benth.) infestation levels, including Striga free controls, in pot experiments in Mali (2003) and The Netherlands (2003 and 2004). The sorghum genotypes used in this study (CK60-B, E36-1, Framida and Tiémarifing) were selected for their supposed differences in resistance and tolerance (Table 1). The Striga hermonthica seeds, used for infestation were collected at Samanko, Mali in 1998 (for experiments conducted in 2003 and 2004 in The Netherlands) and 2001 (for experiment conducted in 2003 in Mali) from plants parasitizing sorghum. The seed viability was 70% (2003, The Netherlands), 88% (2003, Mali) and 60% (2004, The Netherlands). In all experiments, only the upper 10 cm of the soil in each pot was infested with Striga seeds. Table 2 presents an overview of the materials and methods of the different experiments. Table 1. Overview of defence mechanisms (resistance and tolerance) against Striga hermonthica in the four selected sorghum genotypes: CK60-B, E36-1, Framida and Tiémarifing, based on literature sources and personal communication Defence mechanism/ Genotype CK60-B E36-1 Framida Resistance + Tolerance - Tiémarifing + Sources: El-Hiweris, 1987; Gurney et al., 1995; Ast et al., 2000; D. E. Hess/ ICRISAT, pers. commun. 31 Parameter/ Pot experiment 2003S 2003W 2004W Year 2003 2003 2004 Location Samanko (Mali) Wageningen Wageningen Environment Open-air Greenhouse Greenhouse Mean day temperature (°C) 29 28 28 Pot volume (L) 17 12 17 Soil mixture (sand: soil: compost) 3:1:2 3:1 3:1 50:42:75 50:42:75 Fertilizer application (kg ha-1 N:P:K) 42.5:42.5:42.5 Plant spacing (m) 0.4 - 0.8 0.3 - 0.8 0.3 - 0.5 Striga seed viability (%) 80 70 60 Striga infestation levels 0, 0.125, 0.25, 0.5, 1.0 and 2.0 0, 0.5, 1.0 and 3.0 0, 0.0625 (CK60-B and E36-1), 0.125, 0.25, 0.5, 1.0, -3 2.0, 4.0, 8.0 and 16.0 (Framida and Tiémarifing) (viable seeds cm ) Replicates 8 7 8 Sowing date 11 July 26 May 28 April Harvest time (DAS) 120 92-106 106 - 112 Chapter 3 32 Table 2. Summary of materials, methods and environmental conditions of the three experiments: 2003S, 2003W and 2004W Characterization of tolerance to Striga Open-air experiment One pot experiment (2003S) was carried out in the open-air from 11 July to mid November 2003 (2003S) at the ICRISAT field station in Samanko, 20 km Southwest of Bamako, the capital of Mali (latitude: 8°54”W and 12°54”N, altitude: 329 m). The climate type in this area is Sudanese, characterized by a single rainy season between May and October. The mean temperature during the cropping season (June-November) was 29.1°C and the mean annual rainfall at the field station was 950 mm. In addition to natural rainfall, pots were watered by hand to create conditions without water limitation. Pots used in this experiment had a volume of 17 litres and a diameter of 29 cm. Soil used for the experiment was a 3:1:2 quartz sand: arable soil: compost mixture. To improve the drainage capacity of the pots, a 3 cm layer of gravel was put on the bottom of each pot. After infesting the soil with Striga seeds, all pots were kept moist for 10 days to allow preconditioning of the Striga seeds. Pot spacing in the plot was 0.4 m (centre – centre) and plots were separated by an additional row of 0.4 m wide. An equivalent of 42.5 kg N, 42.5 kg P and 42.5 kg K per hectare was applied in a single fertilizer (N-P-K: 1-1-1) dressing just prior to sowing. The sorghum seeds were sown at a rate of five sorghum seeds per pot. Thinning to one plant per pot was done at 17 days after sowing (DAS). Harvests of all aboveground parts of sorghum and Striga plants were done at 120 DAS. This experiment consisted of a split-plot design in eight replicates with sorghum genotype at the plot level, and six Striga infestation levels at the sub-plot level. Striga seeds were mixed through the upper 10 cm of the soil at infestation levels of 0 (control), 0.125, 0.25, 0.5, 1.0 and 2.0 viable Striga seeds cm-3 of soil. Greenhouse pot experiments Two pot experiments were conducted in a tropical greenhouse of Wageningen University in The Netherlands from 26 May to mid-September 2003 (2003W) and from 28 April to mid-August 2004 (2004W). Day length was held constant at 12 h (between 08.00 and 20.00 h). Supplemental light was provided by 400 W sodium vapour lamps that automatically switched on during daytime when global solar radiation dropped below 400W m-2. Day temperatures did not fall below 28°C. Mean relative humidity was kept between 50 and 70% for the duration of the experiments. Pots received water every two days, to create conditions without water limitation. Soil used for the experiment was a 3:1 quartz sand: arable soil mixture. After infesting the pots with Striga seeds, all pots were kept moist for 10 days to allow preconditioning of the Striga seeds. The sorghum seeds were pre-germinated for 36 hours before they 33 Chapter 3 were sown at a rate of three seeds per pot. Thinning to one plant per pot was done at seven DAS. The pot experiment of 2003 (2003W) consisted of a split-plot design in seven replicates with sorghum genotype at the plot level and four Striga infestation levels at the sub-plot level. Striga seeds were mixed through the upper 10 cm of the soil at infestation levels of 0 (control), 0.5, 1.0 and 3.0 viable Striga seeds cm-3 of soil. Pots used for this experiment had a volume of 12 litres and a diameter of 28 cm. Pot spacing in each plot was 0.3 m and plots were separated by an additional row of 0.5 m wide. An equivalent of 50 kg N, 42 kg P and 75 kg K per hectare was applied in a single fertilizer (N-P-K: 12-10-18) dressing at 35 DAS. Plants were harvested at physiological maturity of the different sorghum genotypes at 92 (Framida), 99 (E36-1 and Tiémarifing) and 106 DAS (CK60-B). The pot experiment conducted in 2004 (2004W), consisted of a split-plot design in eight replicates with sorghum genotype at the plot level, and nine Striga infestation levels at the sub-plot level. A wider range of Striga infestation densities was chosen to facilitate the analysis at extreme low and high densities of Striga infection. Striga seeds were mixed through the upper 10 cm of the soil. Framida and Tiémarifing received Striga infestation densities of 0.0, 0.125, 0.25, 0.5, 1.0, 2.0, 4.0, 8.0 and 16.0 seeds cm-3 (0-10 cm). For the more susceptible genotypes, CK60-B and E36-1, the highest infestation level was replaced by an additional low infestation level of 0.0625 seeds cm-3. Pots used in this experiment had a volume of 17 litres and a diameter of 30 cm. Pot spacing in each plot was 0.3 m and plots were separated by an additional row of 0.2 m wide. An equivalent of 50 kg N, 42 kg P and 75 kg K per hectare of fertilizer (N-P-K: 12-10-18) was applied in a single dose before sowing. Harvests of all genotypes were conducted at 105 and 106 DAS, except for plants that were not yet mature. This last category of plants was harvested at 112 DAS. Observations Striga counts were performed every two to three days, up to 61 DAS (2003S), 56 DAS (2003W) and 49 DAS (2004W) and were conducted weekly after these dates. From these regular Striga counts the maximum aboveground Striga numbers (NSmax) were derived. At maturity of the cereal plants, sorghum panicles and aboveground Striga plants of every pot were harvested. Sorghum panicles were sun- (2003S) or oven(2003W and 2004W) dried. Panicles were threshed and kernel yield (DWkernel) was determined. The relative yield loss (RYL) was calculated as: RYL= [(Y c-Ys)/Y c] × 100 (%) 34 Characterization of tolerance to Striga where Y c is the average kernel yield of all control plants of a specific genotype and Ys is the observed yield (DWkernel) of an individual plant grown under Striga infestation. Average control yields were used to reduce variability of RYL. Statistical analyses Data on NSmax, DWkernel and RYL were subjected to analyses of variance (ANOVA), followed by a comparison of means with the least significant difference (L.S.D.), using the Genstat (release 7.1) statistical software package. NSmax and DWkernel were subjected to square root ([X+c]1/2 transformations, where X is the original, individual observation and c=0.5), prior to analysis, to meet the assumptions of the analysis of variance, following procedures recommended by Sokal and Rohlf (1995). The RYL percentages were arc-sinus (or angular) transformed prior to analysis of variances, following procedures recommended by Sokal and Rohlf (1995). Negative RYL values (six cases for 2003W with N =84, 11 cases for 2003S with N =160 and six cases for 2004W with N =256) were replaced by zeros before statistical analysis. Results Sorghum yields and infestation levels Table 3 shows the results of an analysis of variance on DWkernel for each experiment. In 2004, only eight infestation levels were used in the analysis of variance. To balance the experimental design, infestation levels of 0.0625 seeds cm-3 (CK60-B and E36-1) and of 16.0 seeds cm-3 (Framida and Tiémarifing) were left out of the analysis. Table 3. Analysis of variance of sorghum kernel dry weight per host plant of the three experiments: 2003S, 2003W and 2004W, with genotype and infestation level as factors a Experiment Source of Variation df Mean Square 2003W Genotype (G) 3 44.03 156.0* Infestation level (I) 3 29.22 136.7* GXI 9 6.19 29.0* Genotype (G) 3 337.98 38.9* Infestation level (I) 5 152.53 24.1* GXI 15 19.41 3.1* Genotype (G) 3 98.24 71.7* Infestation level (I) 7 39.05 51.2* GXI 21 3.95 5.18* 2003S 2004W F-value * Significant at the 0.01 probability level a Data are square-root-transformed ([X+0.5]1/2) to meet requirements for ANOVA 35 level (0, 0.125, 0,25, 0.5 1.0, 2.0, 3.0, 4.0 and 8.0 seeds cm-3) and experiment (2003S, 2003W and 2004W) Infestation level (seeds cm-3) 0 2003W 2003S 2004W 0.125 0.25 0.5 1.0 2.0 3.0 4.0 8.0 CK60-B 25.7a 0.3g 0.0g 0.0g E36-1 20.3b 6.6e 3.4f 1.9f Framida 24.5ab 21.1ab 20.7ab 15.6cd Tiémarifing 20.8ab 14.8cd 13.0d 18.1c CK60-B 47.9cd 11.8fgh 15.1efg 3.9gh 3.9gh 1.6gh E36-1 110.8a 5.2gh 0.6h 3.5gh 1.0h 4.7gh Framida 122.9a 45.5cd 102.5ab 80.1abc 36.8de Tiémarifing 99.5ab 79.8abc 59.3bcd 47.1cd 28.1def 13.3efg CK60-B 40.6a 2.6fg 1.2fgh 2.7fg 0.5gh 0.0h 0.0h 0.0h E36-1 23.3bc 2.8fg 2.5fg 2.5fg 3.3f 0.6fgh 1.2fgh 0.5gh Framida 30.1ab 22.9bc 23.9bc 20.8c 19.8cd 13.0de 9.8e 11.0e Tiémarifing 26.0bc 10.3e 12.1e 8.6e 10.0e 11.4e 11.2e 8.5e 57.7bcd Values in the same column or row, followed by a different letter are significantly different (P<0.01; S.E.D.’s: 1.30 [2003W], 0.26 [2003S] and 1/2 0.46 [2004W]). Means in the table are back-transformed from square-root transformation ([X+0.5] ), S.E.D. values are not. Chapter 3 36 Table 4. Total sorghum kernel dry weight (g plant-1) per genotype (CK60-B, E36-1, Framida and Tiémarifing), Striga infestation Characterization of tolerance to Striga Yields of E36-1, Framida and Tiémarifing in the control treatment in 2003S were much higher than yields in the control treatments obtained in the greenhouseexperiments in The Netherlands (Table 4). In Mali, yield of CK60-B in the control was significantly lower than that of the other genotypes, whereas in Wageningen yield of CK60-B in the control was as good or significantly better than that of other genotypes (E36-1 [2003W and 2004W] and Tiémarifing [2004W]). Yields of CK60-B and E36-1 plants infected with Striga were always significantly lower at comparable levels of infestation, than the yields of Framida and Tiémarifing, except for Tiémarifing at the highest infestation level in 2003S. Complete crop failure was only found with CK60-B in Wageningen at infestation levels of 1.0 (2003W) and 2.0 seeds cm-3 (2004W) and higher. In all experiments, yields of CK60-B and E36-1 at the lowest infestation level were already significantly lower than in the control. Tiémarifing also showed a significant yield decrease at the lowest infestation level in the two greenhouse experiments (2003W and 2004W). In 2003S, a significant yield reduction for Tiémarifing was observed beginning with the third infestation level (0.5 seeds cm-3). Significant yield reductions in Framida were only obtained at the higher infestation levels (3 seeds cm-3 in 2003W; ≥1 seeds cm-3 in 2003S and ≥0.5 seeds cm-3 in 2004W). The yield reduction of Framida obtained at an infestation level of 0.125 seeds cm-3 in 2003S was a clear exception. Relative yield loss as a function of infestation level Figure 1 shows fairly consistent genotype specific yield loss responses in relation to varying Striga infestation levels. Initial yield loss responses of CK60-B and E36-1 were much more severe than those of Framida and Tiémarifing. The maximum relative yield losses of CK60-B and E36-1 approached 100% and were generally much higher than those of Framida and Tiémarifing (always below 80%). In the 2004W experiment (Figure 1C), relative yield losses at the high infestation level of 16.0 seeds cm-3 were still only 75% (Framida) and 66% (Tiémarifing). While relative yield losses of Tiémarifing seemed to have reached a maximum at the applied Striga infestation levels, those of Framida seemed to continue to increase. Furthermore, at the given inoculum levels, the relation between Striga infestation level and relative yield loss seemed linear for Framida whereas a clear density dependency was observed for the other genotypes. 37 Chapter 3 A 100 RYL (%) 75 50 25 0 0 0.5 1 -3 Infestation level (seeds cm ) 3 B 100 RYL (%) 75 50 25 0 0 0.125 0.25 0.5 1 -3 Infestation level (seeds cm ) 2 C 100 RYL (%) 75 50 25 0 0 0.13 0.5 2 -3 Infestation level (seeds cm ) 8 Figure 1. Relative yield loss (RYL: %) of four sorghum genotypes (CK60-B [closed circles], E36-1 [open circles], Framida [closed triangles] and Tiémarifing [open triangles]) as a function of Striga infestation level (seeds cm-3) in 2003 in the greenhouse in Wageningen: 2003W (A), in the open-air in Samanko: 2003S (B) and in 2004 in the greenhouse in Wageningen: 2004W: (C) 38 Characterization of tolerance to Striga Relation between infestation and infection level Figure 2 shows the maximum aboveground Striga plant numbers (NSmax) per infestation level and sorghum genotype in each experiment. The maximum number of aboveground Striga plants per host plant was much higher for all genotypes in the experiment conducted in Mali. At an infestation level of 1 seeds cm-3, the average NSmax values were 68.6 for 2003S, 32.3 for 2003W and 21.2 for 2004W. The minimum and maximum Striga infestation levels within an experiment differed a factor 6 (2003W) and a factor 16 (2003S) in 2003. Despite this wide range, the differences in maximum number of aboveground Striga plants within a genotype were relatively small. In general, CK60-B and E36-1 always had large maximum aboveground Striga numbers, whereas Striga numbers on Framida were always relatively small. Tiémarifing had intermediate and erratic infection numbers, sometimes comparable to Framida and sometimes comparable to E36-1. In 2003W, maximum aboveground Striga numbers on CK60-B and E36-1 were always significantly higher than on Framida and Tiémarifing with the exception of infestation level 1.0 where NSmax on E36-1 was not significantly different from that on Tiémarifing. No significant genotype × infestation level effect on NSmax was observed in the 2003S experiment. Here Framida had a significantly lower NSmax than the other three genotypes. Based on these results, an even wider range of infestation levels was used in the 2004W experiment. For each genotype, the highest infestation level was 128 times higher than the lowest infestation level. This resulted in significant differences between the lowest and the highest maximum number of aboveground Striga plants for all genotypes. However, the wide infestation range still only resulted in a ratio of 2.4 (CK60-B), 4.4 (E36-1), 23.9 (Framida) and 4.4 (Tiémarifing) between the highest and the lowest infection level. Again, NSmax on CK60-B and E36-1 were always significantly higher than on Framida, while NSmax on Tiémarifing was intermediate. For CK60-B and Tiémarifing, it appeared that within this range of infestation levels, a maximum for NSmax was reached, whereas for E36-1 and Framida NSmax still gradually increased with an increase in infestation level. Relative yield loss per genotype and aboveground infection level The average infection levels of CK60-B and E36-1 were not significantly different from one another in any experiment (Table 5). Average infection levels of CK60-B and E36-1 were higher than that of Framida and Tiémarifing, except for 2003S. 39 Chapter 3 A 50 NSmax 40 30 20 10 0 0 NSmax B 0.5 1 -3 Infestation level (seeds cm ) 3 120 100 80 60 40 20 0 0 C 0.125 0.25 0.5 1 -3 Infestation level (seeds cm ) 2 50 NSmax 40 30 20 10 0 0 0.13 0.5 2 -3 Infestation level (seeds cm ) 8 Figure 2. Maximum aboveground Striga numbers (NSmax) at four sorghum genotypes (CK60B [closed circles], E36-1 [open circles], Framida [closed triangles] and Tiémarifing [open triangles]) as a function of Striga infestation level (seeds cm-3) in 2003 in the greenhouse in Wageningen: 2003W (A), in the open-air in Samanko: 2003S (B) and in 2004 in the greenhouse in Wageningen: 2004W (C). 40 Characterization of tolerance to Striga Table 5. Main genotype effects on maximum aboveground Striga numbers (NSmax) and relative yield loss (RYL: %) of the three experiments: 2003S, 2003W and 2004W NSmax 2003W CK60-B 45.1 aa 99.4 a E36-1 39.3 a 80.3 b Framida 9.0 c 19.1 c Tiémarifing 19.3 b 22.8 c b 2003S 2004W RYL S.E.D 0.28 4.16 CK60-B 81.4 a 85.0 b E36-1 77.1 a 97.8 a Framida 32.2 b 45.4 c Tiémarifing 69.7 a 56.5 c S.E.D 0.56 CK60-B 26.6 a 98.1 a E36-1 25.9 a 93.1 b Framida 7.1 c 38.8 d Tiémarifing 14.1 b 58.7 c S.E.D 0.24 4.99 3.82 a Means in the same column, followed by a different letter are significantly different at the 0.01 (RYL) 1/2 or 0.001 (NSmax) probability level. Data on NSmax were square root-transformed ([X+ 0.5] ) while data on RYL were arc-sinus transformed for ANOVA. b Test statistics (probabilities and S.E.D.’s) are based on transformed data, whereas values in table are back-transformed In two of the three experiments (2003W and 2004W), average relative yield loss of E36-1 was significantly lower than that of CK60-B, indicating that E36-1 could be less sensitive than CK60-B. In 2003, exactly the opposite was observed. Although the average infection level of Framida was always significantly lower than that of Tiémarifing, the average RYL of Tiémarifing was only significantly higher than that of Framida in the 2004W experiment. Relative yield reductions of both genotypes were significantly lower than those of CK60-B and E36-1 in all experiments. The relationship between infection level and relative yield loss is presented in Figure 3. It is obvious that both in 2003W and 2003S the range of infection levels for each of the genotypes was narrow (NSmax in 2003W: 43-50 [CK60-B], 34-50 [E36-1], 7-18 [Framida] and 11-33 [Tiémarifing]; in 2003S: 57-100 [CK60-B], 70-93 [E36-1], 31-41 [Framida] and 59-105 [Tiémarifing]). This makes it difficult to resolve the relation between infection level and relative yield loss of the genotypes. Furthermore, there was no specific range of infection levels in which all four genotypes were represented. 41 Chapter 3 100 A RYL (%) 80 60 40 20 0 0 10 20 30 40 50 NSmax 100 B RYL (%) 80 60 40 20 0 0 C 50 NSmax 100 100 RYL (%) 80 60 40 20 0 0 10 20 30 NSmax 40 50 Figure 3. Relative yield loss (RYL: %) of four sorghum genotypes (CK60-B [closed circles], E36-1 [open circles], Framida [closed triangles] and Tiémarifing [open triangles]) as a function of Striga infection level (NSmax: maximum Striga numbers) in 2003 in the greenhouse: 2003W (A) and in the open-air: 2003S (B) and in 2004 in the greenhouse: 2004W (C). Vertical lines in the figure indicate the highest infection level of the most resistant genotype (dotted) and the lowest infection level of the most susceptible genotype (solid). 42 Characterization of tolerance to Striga This is obvious from Figures 3A and 3B, where the dotted vertical line, representing the highest infection level of the most resistant genotype, is found at the left hand side of the solid vertical line, representing the lowest infection level of the most susceptible genotype. A broader range of infection levels was obtained for all four genotypes in 2004. CK60-B was characterized by infection levels ranging from 15-36 Striga plants. Even at the lowest infection level, RYL was already higher than 80%. At higher infection levels, 100% RYL was observed. E36-1 had an even broader range of infection levels, ranging from 11- 48 aboveground Striga plants. RYL was 55% at the lowest infection level and gradually increased until it reached nearly 100% at the highest infection level. The RYL of Framida was characterized by a nearly linear increase with Striga infection level. RYL at the lowest infection level was 22% and increased to 75% at the highest infection level (NSmax= 21.5). Infection levels for Tiémarifing varied from 9 to 27 Striga plants per pot, however, RYL did not show much variation and averaged 57%. Overlapping infection levels were observed in all four sorghum genotypes in 2004W (Figure 3C). This range varied from 15.4 (lowest infection of CK60-B obtained at an infestation level of 0.0625 seeds cm-3) to 21.5 (highest infection level of Framida obtained at an infestation level of 16 seeds cm-3) aboveground Striga numbers. In this range of infection levels, Tiémarifing and Framida showed to be significantly (P<0.001) more tolerant than CK60-B and E36-1, with relative yield reductions of 85.3% for CK60-B, 84.1% for E36-1, 67.5% for Framida and 60.2% for Tiémarifing. Discussion The results of the 2003 experiments showed that differences in Striga infestation level did not result in proportional differences in infection level. In 2003W, infestation levels that differed six-fold only resulted in infection levels that differed three-fold, whereas in 2003S infestation levels that differed sixteen-fold resulted in infection levels that differed less than two-fold. These results clearly indicate that the relation between Striga infestation and Striga infection is density dependent, confirming earlier observations by Smith and Webb (1996). The 2003 experiments confirmed the results of earlier studies (e.g. El Hiweris, 1987; Hess 1989; Arnaud et al., 1999; Ast et al., 2000) on the resistance of Framida and the susceptibility of CK60-B and E36-1. Since substantial numbers of aboveground Striga plants (> 25) were obtained at the lowest infestation levels for most genotypes, it was not possible to explore the relation between infection level and relative yield loss at low levels of infection. The 43 Chapter 3 relatively narrow range of infection levels also made it hard to conclude whether the maximum relative yield loss was attained at the highest infection level. This was particularly true for Tiémarifing and Framida, which did not yet reach 100% yield loss. Consequently, the exact course of the relation between Striga infection level and relative yield loss could not be completely resolved with the 2003 experiments. In two of the three experiments, the narrow range of infection levels for each genotype, combined with the distinct differences in resistance level among genotypes, resulted in the absence of a common infection range for all genotypes. Hence, a direct comparison of tolerance between the various genotypes was not possible. However, some indications for differences in tolerance between genotypes were obtained. In 2003S, Framida and Tiémarifing had comparable relative yield losses; however, the average Striga infection level of Tiémarifing was twice as high. This result suggests that Tiémarifing is the more tolerant genotype. In an attempt to overcome the aforementioned problems, the Striga infestation range in the 2004 experiment was expanded. Each genotype was exposed to infestation levels that differed 128-fold and the infestation range was made genotype specific. For the more susceptible genotypes (CK60-B and E36-1), infestation levels varied from 0.0625 to 8.0 seeds cm-3, whereas the more resistant genotypes (Framida and Tiémarifing) were exposed to infestation levels varying between 0.125 and 16.0 seeds cm-3. The range of infection levels was much smaller than the range of infestation levels. The size of these infection ranges, expressed as the ratio between maximum and minimum infection level, varied between genotypes (CK60-B: 2.4; Tiémarifing: 2.8; E36-1: 4.4; Framida: 23.9). Again this demonstrates the density dependence of the relation between infestation and infection. Main reason for the narrow range of infection levels for three of the four genotypes was the absence of low infection levels (< 10 aboveground Striga plants). This indicates that, in order to obtain such low infection levels for susceptible genotypes, extremely low infestation levels are required, which comprises the risk of not obtaining any infection at all. Despite the differences in Striga infection level among genotypes, a small overlapping range of infection levels was obtained. Within this range, each genotype was represented by data obtained from just two (CK60-B, E36-1 and Framida) or three (Tiémarifing) infestation levels. Under these conditions, Tiémarifing and Framida were significantly more tolerant than CK60-B and E36-1. Sensitivity of CK60-B was earlier reported by Gurney et al. (1995) while tolerance of Tiémarifing was observed by Ast et al (2000). However, to arrive at this conclusion, only 36% of the experimental units were used. This demonstrates that, regardless of practical difficulties, the strategy to create identical infection levels to facilitate a direct screening for tolerance is very inefficient. 44 Characterization of tolerance to Striga Table 6. Overview of defence mechanisms (resistance and tolerance) against Striga hermonthica in the four selected sorghum genotypes: CK60-B, E36-1, Framida and Tiémarifing, based on observations from the three experiments: 2003S, 2003W and 2004W Defence mechanism/ Genotype CK60-B E36-1 Framida Tiémarifing Resistance + +/Tolerance +/+ Based on the outcomes of this study two modifications concerning the information presented in Table 1 are made (Table 6). First, Tiémarifing appeared not as susceptible as CK60-B and E36-1, though clearly less resistant than Framida. Second, Framida proved more tolerant than CK60-B and E36-1 though still less tolerant than Tiémarifing. Main objective of the current study was not to compare genotypes at identical infection levels, but rather to resolve the relation between Striga infection level and yield loss of the host. It was anticipated that clarification of this relation would enable the development of a suitable screening procedure for tolerance. In Figure 4, a threequadrant representation of the relationship between Striga infestation level, infection level and relative yield loss is given for the results obtained in 2004W. This presentation form was adopted from the nutrient supply, nutrient uptake and crop yield response curves introduced by de Wit (1953). The figure is composed of three quadrants, where the upper-left quadrant (quadrant II) represents the relation between Striga infestation level and relative yield loss, the lower-right quadrant (quadrant IV) represents the relation between Striga infestation level and Striga infection level and the upper-right quadrant (quadrant I) represents the relation between Striga infection level and relative yield loss. Note that in this figure, in contrast to Figures 1 and 2, Striga infestation level is presented on a linear scale. Quadrant II shows two main response types to Striga infestation. CK60-B and E36-1 (Figure 4A and 4B, respectively) represent genotypes where complete or nearly complete yield losses were attained at low infestation levels. Framida and Tiémarifing (Figure 4C and 4D, respectively) represent genotypes where relative yield losses seem to stabilise around 60-70% at high infestation levels. The main difference between these two genotypes was that Tiémarifing obtained this level already at low infestation levels, whereas with Framida a more gradual increase in relative yield loss with infestation level was observed. For Orobanche (spp.) in carrot and pea, Bernhard et al. (1998) found a rectangular hyperbola describing the relation between seed infestation level and yield loss. At low infestation levels they observed a gradual increase in yield loss with increasing infestation level, comparable to what was observed with Framida, resulting in complete crop failure at high infestation levels, identical to the results obtained with CK60-B and E36-1. 45 Chapter 3 A B 100 RYL (%) RYL (%) 100 NSmax 16 Infestation 16 50 Infestation 50 Infestation 16 Infestation NSmax 16 D C 100 RYL (%) RYL (%) 100 NSmax NSmax 16 16 Infestation 50 Infestation 50 Infestation 16 Infestation 16 Figure 4. Three-quadrant representations of the relations between Striga infestation level, Striga infection level (NSmax) and relative yield loss (RYL: %) of four different sorghum genotypes: CK60-B (A), E36-1 (B), Framida (C) and Tiémarifing (D), as observed in 2004 in the greenhouse in Wageningen (2004W) The two quadrants on the right hand side provide further information on how the relation between Striga infestation level and relative yield loss was achieved. Quadrant IV contains the relation between Striga infestation and Striga infection and as such shows the level of resistance of a certain genotype. For both E36-1 and Framida this relation developed according to a rectangular hyperbola. Such a 46 Characterization of tolerance to Striga relationship between Striga infestation and Striga infection level was previously reported by Smith and Webb (1996) and confirms the earlier observation on density dependence. With E36-1, low infestation levels resulted in relatively high infection levels and the number of infections further increased in response to higher Striga seed densities. Framida was more resistant, with few infections at low infestation levels and the number of infections increased slowly as infestation level increased. Another type of response was observed with CK60-B and Tiémarifing. For those two genotypes, relatively high infection rates were observed at low seed densities: however, the infection rate did not continue to rise as infestation levels increased. CK60-B differed from Tiémarifing as it had a steeper initial increase in number of infections, it attained its maximum infection level at a lower infestation level, and its maximum number of infections was higher. The relationships observed for CK60-B and Tiémarifing could result from a reduced carrying capacity of the host plant at higher infestation levels, following reduced host vigour. It could also result from increased intra-specific competition following a higher number of belowground Striga attachments or from a combination of both. Consequently, screening for host plant resistance under very high infestation levels, using number of aboveground Striga plants as screening measure, might result in an overestimation of the level of resistance of susceptible genotypes. Kim et al. (1998) and Haussmann et al. (2000b) also suggested that this might be possible. Quadrant I represents the relationship between Striga infection and relative yield loss and conveys the level of tolerance of a certain genotype. Three genotypes (CK60-B, E36-1 and Tiémarifing) seemed to reach or approach their maximum relative yield losses in 2004. Complete crop failure was observed for CK60-B and E36-1, whereas the maximum relative yield loss for Tiémarifing was only around 57%. For these three genotypes, the relative yield loss at low infection levels was not observed and remained unresolved. For Framida, the relationship between infection and relative yield loss was observed over a wide range and in this trajectory a nearly linear increase in relative yield loss was observed with an average yield loss of 4% per Striga infection. Koskela et al. (2002) reported a similar relationship between parasite infection level and host damage for the holoparasite Cuscuta europaea parasitizing on Urtica dioica. Gurney et al. (1999) found a negative and exponential relationship between yield and parasite load for Striga hermonthica parasitizing on sorghum, where parasite load was expressed as Striga dry weight. These findings do not necessarily contradict findings of the current study. Rather, given the difficulties of obtaining a complete infection range for a single genotype, the current observations only cover parts of the relation between infection level and relative yield loss. Also with Framida one might 47 Chapter 3 expect that the relation between infection level and relative yield loss will eventually reach a saturation level. Whether this saturation level corresponds to complete crop failure or is found at a lower level of yield reduction remains unresolved. It is evident that the saturation level of the relation between infection level and relative yield loss is one of the ways through which tolerance can come to expression. Tiémarifing is an example of this. At the same time, the three genotypes for which a maximum relative yield loss was observed will possess an initial trajectory in which the relative yield loss increases with infection level. E36-1 already shows part of this trajectory. The steepness of this initial increase, expressed as relative yield loss per Striga plant, represents another expression of tolerance. The current results however do not allow verifying whether for the other three genotypes this increase deviates from the 4% yield loss per Striga plant obtained with Framida. For the same reason it remains unclear whether a lower maximum relative yield loss, as observed for Tiémarifing, goes along with a reduced initial slope, or whether those two exist independently. Conducting large scale screening for tolerance at multiple infestation levels is not realistic. Determination of the relative yield loss already requires the presence of Striga-free control plots adjacent to Striga infested plots (e.g. Gurney et al., 1999; Rodenburg et al., 2005). Control plots in the field can be created by use of methyl bromide (e.g. Gurney et al., 1999) or ethylene injections (e.g. Bebawi et al., 1985; Bebawi and Eplee, 1986) which are both rather expensive and laborious. An alternative is the infestation of Striga free fields, which is undesirable. Furthermore, measures should be taken to prevent contamination of control plots with Striga seeds from adjacent infested plots. Additionally, for a reliable selection, also sufficient replications (≥ 5) are needed as was already shown by Haussmann et al. (2000b). Due to variation in Striga virulence (e.g. Bebawi and Farah, 1981), and significant genotype × environment interactions (e.g. Haussmann et al., 2001a; Oswald and Ransom, 2004), stability of tolerance levels in a genotype should be tested at multiple locations. Compared to screening at one infestation level and a control plot, the installation of multiple infestation levels, to facilitate the estimation of tolerance, will only further increase these practical difficulties. Kim (1991) suggested that screening for tolerance could best take place at high infestation levels. The current results indicate that differences in maximum relative yield loss, and thus tolerance, between susceptible and moderately resistant genotypes can well be detected in this way, as was shown by the comparison between Tiémarifing, CK60-B and E36-1. For more resistant genotypes this approach proved less suitable, due to the fact that it was not possible to obtain infection levels that are high enough to cause the maximum relative yield loss. For breeding programs that try to develop genotypes that combine superior resistance with high levels of tolerance, as 48 Characterization of tolerance to Striga suggested by Ramaiah and Parker (1982), Haussmann et al. (2000b) and Pierce et al. (2003), screening based on the maximum relative yield loss seems less appropriate, as particularly the expression of tolerance at lower infection levels is of interest. For those resistant genotypes, expressing tolerance as the ratio between relative yield loss and infection level seems more appropriate. Main bottleneck here is that if the relation between relative yield loss and infection level is described by a rectangular hyperbola, this ratio will decrease with increasing infection level. Such a linkage with resistance hampers an unbiased estimation of tolerance. Screening at more than one infestation level might improve the estimation of the proposed ratio, but, as was mentioned earlier, is not a realistic option. In conclusion, two compatible tolerance measures are proposed based on yield response. For resistant genotypes a reduced relative yield loss per aboveground Striga plant indicates tolerance, whereas for less resistant genotypes the relative yield loss as such provides the best indication. Consequently, screening for tolerance based on the yield response of a genotype is difficult when the selection pool contains genotypes with largely different and unknown levels of resistance. As the need for unravelling resistance and tolerance is evident, the results of this study emphasize the need for a proper alternative method for screening for tolerance. 49 CHAPTER 4 Can host plant tolerance to Striga hermonthica be detected by photosynthesis measurements?1 J. Rodenburg1, 2, L. Bastiaans1, A. H. C. M. Schapendonk3, P. E. L. van der Putten1, A. van Ast1 and M. J. Kropff1 1 Crop and Weed Ecology Group, Department of Plant Sciences, Wageningen University, P.O. Box 430, 6700 AK Wageningen, The Netherlands 2 3 Africa Rice Center (WARDA), 01 BP 2031, Cotonou, Benin Plant Dynamics BV, Englaan 8, 6703 EW, Wageningen, The Netherlands Abstract The photosynthetic response of four sorghum genotypes (CK60-B, E36-1, Framida and Tiémarifing) differing in level of tolerance to Striga hermonthica was measured at different moments in time in pot experiments conducted in 2003 and 2004. Striga infection significantly reduced CO2 assimilation rate (A) of sorghum plants. This process was found indicative for tolerance, as sensitive genotypes were affected earlier, more severe and already at lower infestation levels than more tolerant genotypes. This observation was confirmed in 2004, when it was demonstrated that the CO2 assimilation rate of infected and uninfected sorghum plants, measured during the early stages (26 and 48 DAS) correlated very significantly with their final kernel yield. However, CO2 assimilation as screening measure was shown to have some serious constraints. The measure did not enable a clear distinction between superior and moderately tolerant genotypes, it still requires Striga-free controls due to genotype effects on assimilation rate and measurement systems based on gas exchange are costly. In 2004, photochemical quenching (Pq), non-photochemical quenching (NPq), electron transport rate through PSII (ETR) and the ratio of CO2 assimilation over electron flow (A ETR-1) were determined along with CO2 assimilation rates. All of these chlorophyll fluorescence parameters correlated highly significantly with CO2 assimilation rate. Based on discriminative ability, practicability and cost effectiveness, Pq and ETR were found to carry the highest potential to serve as a screening measure for tolerance to Striga. Screening is recommended to be conducted between first Striga emergence and sorghum flowering and at infestation levels of at least 300,000 viable Striga seeds m-2. In contrast to existing screening methods that need control plots that are expensive and difficult to obtain, both parameters facilitate screening at one infestation level and without the requirement of Striga-free control plots. 1 Submitted to New Phytologist 51 Chapter 4 Introduction Striga hermonthica (Del.) Benth. is an obligate hemi-parasitic C3 plant of the Orobanchaceae (formerly: Scrophulariaceae) family that parasitizes monocotyledonous hosts of the Gramineae and Poaceae families. Among the C4 hosts of Striga are some important cereal crops such as, pearl millet (Pennisetum glaucum [L.] R. Br. and P. americanum [L.] K. Schum), maize (Zea mays [L.]) and sorghum (Sorghum bicolor [L.] Moench) accounting for an estimated 25, 27 and 31% of the total area under cereal production in sub-Saharan Africa (FAOSTAT, 2004). More than half of the total cereal production area in this region is estimated to be infested with Striga (Sauerborn, 1991). Infection by Striga hermonthica seriously reduces host plant yield (Bebawi and Farah, 1981; Doggett, 1982; Vasudeva Rao et al., 1989). Average yield losses due to Striga in West Africa are estimated to range between 10 to 31% and can reach 100% in severely infested fields (Sauerborn, 1991). Striga is one of the most serious causes of yield reduction of sorghum and a major constraint to food production in semi-arid Africa (Doggett, 1982; Parker and Riches, 1993). Striga parasitizes on the host root, subtracting host carbon assimilates (Rogers et al., 1962; Okonkwo, 1966; Press et al., 1987b), water, nutrients (nitrate) and amino-acids (Pageau et al., 2003). However, the drain of assimilates and nutrients only accounts for 16 to 20% of the total growth reduction of the host (Press and Stewart, 1987; Graves et al., 1989, 1990). The remaining 80 to 84% is caused by other effects of Striga infection on host performance, often referred to as phytopathological or toxic reactions. Known biochemical reactions upon Striga infection are decreased levels of host growth regulators such as cytokinins and giberellic acid (Drennan et al., 1979), and increased levels of abscisic acid (ABA) in the host plant (Drennan et al., 1979; Taylor et al., 1996; Ackroyd et al., 1997; Frost et al., 1997). The affected hormone balance may be responsible for the modified host plant allometry as observed in many studies (e.g. Egley, 1971; Graves et al., 1989; Cechin et al., 1993; Clark et al., 1994; Gurney et al., 1995, 1999; Boukar et al., 1996; Frost et al., 1997; Watling and Press, 1997; Gebremedhin et al., 2000; Sinebo and Drennan, 2001). The increased levels of ABA cause reduced stomatal conductance, a phenomenon often observed, particularly during the early stages of infection (Press and Stewart, 1987; Press et al., 1987a; Gurney et al., 1995, 1999; Taylor et al., 1996; Ackroyd et al., 1997; Frost et al., 1997). Reduced stomatal conductance was found to be one of the reasons for reduced photosynthesis of infected hosts (Prabhakara Setty and Hosmani, 1981; Press and Stewart, 1987; Press et al., 1987; Graves et al., 1989; Smith et al., 1995; Gurney et al., 52 Detecting tolerance by photosynthesis measurements 1995, 1999). An increased level of photoinhibition was also found as a response to Striga infection (Gurney et al., 2002a). In some host plant species, genotypes with tolerance against Striga have been identified (e.g. Efron, 1993; Kim, 1994; Gurney et al., 2002a). Tolerance is the ability of the host plant to endure the presence of a pathogen, disease or parasite with minimized symptoms or damage (Parlevliet, 1979). Varieties with improved tolerance can play a key-role to increase cereal production in Striga infested areas (Gurney et al., 1999). Hence tolerance is an important breeding objective. One of the constraints to breeding for tolerance against Striga is the absence of a suitable selection procedure. The presence of tolerance in a host plant genotype results in a lower relative yield loss at comparable Striga infection levels than sensitive genotypes of the same host species. However, as much as genotypes differ in tolerance, they can differ in resistance. Consequently, it is difficult to compare genotypes at identical infection levels. A direct quantification of tolerance based on relative yield loss is thus hampered by the entanglement of this measure with resistance (Rodenburg et al., 2005). To overcome this problem it was recommended to conduct screening for tolerance at more than one Striga infestation level (Rodenburg et al., Accepted), whereas the inclusion of Strigafree control plots is required for the calculation of relative yield loss. These prerequisites make screening for tolerance expensive and laborious and hence create the need for an alternative procedure. A suitable selection measure should facilitate an easy, quick and reliable quantitative assessment that enables the comparison of tolerance among a group of genotypes without the need for various infestation levels and control plots. Some studies observed that tolerant host plant genotypes are able to maintain high levels of photosynthesis upon infection (Gurney et al., 1995, 2002a). The current study explored options for the use of photosynthesis and related chlorophyll fluorescence measurements in screening sorghum genotypes for tolerance to Striga. For that reason two pot experiments were conducted in which the leaf photosynthetic response of four sorghum genotypes with different levels of tolerance, to Striga infection, was measured at different moments in time. 53 Chapter 4 Table 1. Characterization of the four sorghum genotypes used in the 2003 and 2004 experiment CK60-B E36-1 Framida Tiémarifing origin USA/ NorthEthiopia Southern Mali type morphology race photoperiodicity cycle length (days) grain colour resistance tolerance east Africa short kafir insensitive 100-110 white very low very low Africa medium caudatum insensitive 120-130 cream very low low medium caudatum insensitive 120-130 red high medium long guinea sensitive 120-130 white medium high Materials and methods Experimental sites Two pot experiments, one in 2003 and the other in 2004, were carried out in the tropical greenhouse of Wageningen University, The Netherlands. Day length was held constant at 12 hours (08.00 am to 20.00 pm). Additional light was provided by highpressure sodium lamps (400W SON-T Agro-Philips lamps) when incoming radiation dropped below 910 µmol m-2 s-1 (PAR). Day temperatures did not fall below 28°C. Mean night temperature was 24°C. Mean relative humidity was kept between 50 and 70% for the whole duration of the experiments. Pots received water every two days, to create non-water-limited conditions. Plant material and genetics Host plant species used in this study was Sorghum bicolor (L.) Moench. The study comprised two sensitive and susceptible (CK60-B and E36-1), one resistant and moderately tolerant (Framida) and one tolerant and moderately resistant sorghum genotype (Tiémarifing). Furthermore, genotypes differed in origin, race, grain colour and morphology (Table 1). Striga hermonthica seeds were collected in 1998 in Samanko (Mali) from Striga plants parasitizing sorghum. Seed viability was 70% (2003) and 60% (2004). Experimental set-up Both pot experiments consisted of a split-plot design in seven (2003) and five (2004) replications, with four sorghum genotypes at the plot level, and four (2003) and two (2004) Striga infestation levels at the split-plot level. The Striga infestation levels, expressed in number of viable Striga seeds cm-3 of soil, were: 0 (control), 0.5, 1.0 and 54 Detecting tolerance by photosynthesis measurements 3.0 in 2003 and 0 (control) and 4.0 in 2004. A 3:1 quartz sand: arable soil mixture was used in both experiments. Striga seeds were mixed through the upper 10 to 12 cm soil layer. After infesting the soil with Striga seeds, all pots, including the uninfested ones, were kept moist for 10 days to allow preconditioning of the Striga seeds. The sorghum seeds were pre-germinated for 36 hours before they were sown at a rate of 3 seeds per pot, on 26 May (2003) and 28 April (2004). Thinning to one plant per pot was done at 7 days after sowing (DAS). Pot spacing in the plot was 0.3 m and between plots 0.8 m (2003) and 0.5 m (2004). Before sowing (2004) or 35 days after sowing (2003), fertilizer was applied in both experiments in a single dose, equivalent to 50 kg N, 42 kg P and 75 kg K per hectare (N-P-K; 12:10:18). Measurements and observations Leaf CO2 assimilation rates (A) of sorghum were measured at 19, 33, 47 and 61 DAS (2003) and at 26, 48, and 75 DAS (2004). In the 2003 experiment different plants were measured at each observation time, as after photosynthesis measurement these plants were used for destructive sampling (data not reported). In 2004 the same plants were used for repeated measurements. Measurements were always made halfway along the length of the youngest fully expanded leaf. This did not include the flag leaf. Stomatal conductance (gs) and intra-stomatal CO2 concentration (Ci) were calculated based on transpiration rates (Tr) and vapour pressure deficits of the leaves that were measured along with CO2 assimilation rates. Photosynthesis was measured with two different open systems. In 2003, photosynthesis was measured with the LCA2 from the Analytical Development Company (ADC) Hoddesdon, UK. An external heat filtered light source was used to maintain irradiance at a constant value of 1800 µmol m-2 s-1 (PAR) ensuring light saturation. Environmental conditions at the time of measurement were comparable to the growing conditions. Leaf chamber temperature ranged between 28 and 35 °C (mean: 31°C), the inlet CO2 concentration was around 360 ppm with generally less than 50 ppm depletion. Photosynthetic rate was recorded when the rate of CO2 exchange had been steady for 5 minutes. One single measurement (including adaptation time) took about 15 minutes. In 2004, photosynthesis and chlorophyll fluorescence were measured with the LICOR-6400-40. This system has an incorporated light source with a programmable light intensity and an integrated modulated chlorophyll fluorescence measurement system. After a dark adaptation period of 5 minutes, photosynthesis and fluorescence responses were measured at 0, 200, 400, 800 and 1600 (at 26 DAS) or 2000 (at 48 and 75 DAS) µmol m -2 s-1 (PAR). During the measurements, leaf temperature ranged 55 Chapter 4 between 28 and 33 °C (mean: 31°C), the inlet CO2 concentration was 400 ppm and depletion never exceeded 24 ppm. Chlorophyll fluorescence measurements were used to derive the electron transport rate through PSII (ETR), as well as the level of photochemical (Pq) and nonphotochemical quenching (NPq). For the derivation of ETR, first the electron transport efficiency of PSII (Φ2) was calculated as: Φ 2 = (1 − Ft ) / Fm' (1) where Ft is the steady-state fluorescence emission, and Fm' is the maximum fluorescence emission induced by a saturating light pulse in the light (Genty et al., 1989). ETR was then calculated as: ETR = Φ2 ρf abs I (2) where ρ is the factor to account for the partitioning of energy between the two photosystems (PSI and PSII), fabs is the absorbtivity of the leaf and I is the light intensity (PAR) (Genty et al., 1989). Parameter ρ was set to 0.5, which is a common value (Maxwell and Johnson, 2000; Rascher et al., 2000), and assumes that at any light level the excitation energy is equally distributed between PSI and PSII. The absorbtivity was set to 0.8, which indicates that of the incoming photosynthetically active radiation 80% is absorbed by the leaf (Goudriaan and Laar, 1994). Photochemical quenching of fluorescence (Pq) was computed as: Pq = ( Fm'− Ft ) ( Fm '− Fo' ) (3) where Fo' is the basic fluorescence in the light when all PSII centres are oxidized by a period of far-red light (Schreiber, 1986). Finally, non-photochemical quenching (NPq) was computed as: NPq = ( Fm − Fm' ) Fm' (4) where Fm is the maximum fluorescence emission induced by a saturating light pulse in the dark (Genty et al., 1989). In both experiments aboveground Striga counts were done at each photosynthesis measurement time. Sorghum kernel weight of each individual plant at harvest time was assessed after drying (48 h at 70° C) and threshing of the panicles. Statistical analysis All data were subjected to analysis of variance followed by a comparison of means with the least significant difference (L.S.D.) using the Genstat (release 6.1) statistical software package. Linear regression analysis and Pearson correlation tests were done with the SPSS (version 10.0) statistical software package. 56 Detecting tolerance by photosynthesis measurements Table 2. Average first Striga emergence time (DAS), average aboveground Striga numbers at time of photosynthesis measurements and average kernel dry weight (DW: g) for control and Striga infected plants and relative yield loss (RYL: %) of the four sorghum genotypes in the 2003 and the 2004 experiment. Striga numbers 2003 Emergence 33 DAS 47 DAS Kernel DW 61 DAS Control Striga- RYL (%) infected CK60-B 30.3a 2.5 8.9 34.5a 25.8a 0.0d 100 E36-1 28.3a 2.5 11.6 41.3a 20.3b 2.0d 90 Framida 33.0a 1.1 4.3 17.1b 24.5ab 15.8c 36 Tiémarifing 42.3b 2.5 7.9 10.7b 21.2b 18.2bc 14 Striga- RYL (%) S.E.D. 2.26 9.94 1.82 P 0.011 0.03 <0.001 2004 Emergence 26 DAS 75 DAS Control 48 DAS infected CK60-B 33.8a - 4.2a 22.8a 40.6a 0.0e 100 E36-1 40.0ab - 4.3a 23.3a 23.6c 1.6e 93 Framida 49.0bc - 1.3b 7.5b 30.5b 10.2d 67 - 0.2b 7.5b 26.4bc 14.3d 46 Tiémarifing 56.2c S.E.D. 5.19 0.84 4.71 2.31 P 0.002 <0.001 0.005 <0.001 a for 2003, only data of the highest Striga infestation level (3.0 seeds cm-3) are presented b values in the same column, within a year, followed by a different letter, differ significantly (P<0.05) Results Striga infection and sorghum yield loss In Table 2, results on Striga infection and sorghum yield loss are presented for the highest Striga infestation levels (2003: 3.0 seeds cm-3; 2004: 4.0 seeds cm-3). In 2003, first Striga emergence on CK60-B, E36-1 and Framida was significantly (P<0.05) earlier than on Tiémarifing. In the 2004 experiment, Striga on CK60-B emerged significantly (P<0.01) earlier than on Framida and Tiémarifing. Furthermore Striga emergence on Tiémarifing was significantly later than on E36-1. In 2003, at 61 DAS, aboveground Striga numbers on CK60-B and E36-1 were significantly (P<0.05) higher than on Framida and Tiémarifing. No differences were observed before that time. In 2004, differences between these genotypes were already found at 48 DAS (P<0.001) and were still present at 75 DAS (P<0.01). 57 Chapter 4 50 CK60-B 2003 2004 CK60-B 40 30 20 10 0 E36-1 E36-1 Framida Framida 50 40 30 20 10 0 50 40 30 20 10 0 Tiemarifing 50 Tiemarifing 40 30 20 10 0 15 25 35 45 55 65 75 25 35 45 55 65 75 Figure 1. CO2 assimilation rate (A; µmol CO2 m-2 s-1; y-axis) over time (DAS; x-axis) for four sorghum genotypes (CK60-B, E36-1, Framida and Tiémarifing) at different Striga infestation levels (seeds cm-3): 0.0 (open diamonds), 0.5 (squares), 1.0 (triangles) and 3.0 (closed diamonds) in 2003 (left) and 0.0 (open diamonds) and 4.0 (closed diamonds) in 2004 (right). Light intensity ([400-700 nm]; µmol photon m-2 s-1): 1800 (2003) and 1600-2000 (2004). Bars represent (genotype × Striga) L.S.D. (P<0.05) values from one-way ANOVA. 58 Detecting tolerance by photosynthesis measurements In 2003, Striga infection significantly (P< 0.001) reduced sorghum kernel dry weight of all genotypes except Tiémarifing (Table 2). Striga infection resulted in complete (CK60-B) to nearly complete (E36-1) yield loss for the sensitive genotypes, and yield losses of 36% for Framida and only 14% for Tiémarifing. In 2004, Striga infection significantly (P<0.001) reduced sorghum kernel dry weight of all genotypes. Again, yield loss of CK60-B was most severe (100%), followed by E36-1 (93%), Framida (67%) and Tiémarifing (46%). CO2 assimilation rate In both 2003 and 2004, CO2 assimilation rates (A in µmol m-2 s-1) of control plants gradually decreased over time (Figure 1). These observations had coefficients of variation (CV) between 8.5 and 17.2 %, but a much higher value at the last observation date in 2004 (75 DAS: CV = 32.4%). Significant differences between A of control plants of the four genotypes were observed in 2003 at 33 DAS (P<0.01) and 47 DAS (P<0.05) and in 2004 at both 26 (P<0.05) and 48 DAS (P<0.01). In both experiments, E36-1 was significantly lower than either CK60-B (2004 at 48 DAS), Tiémarifing (2003 at 47 DAS and 2004 at 26 DAS), or both of them (2003 at 33 DAS). The response of the various genotypes to Striga infection differed considerably. In 2003 (left side of Figure 1), CO2 assimilation rates of infected plants of CK60-B were significantly reduced compared to the control plants (P<0.01) at 19 and 47 DAS, irrespective of infestation level, at 33 DAS at infestation levels of 1.0 and 3.0 seeds cm-3 (P<0.05) and at 61 DAS at 0.5 (P<0.05) and 3.0 seeds cm-3 (P<0.01). The same was found for E36-1 at 47 DAS at all infestation levels (P<0.01) and at 61 DAS at 0.5 and 3.0 (P<0.05) and 1.0 seeds cm-3 (P<0.01). Significant reductions in CO2 assimilation rate of infected plants of Framida were only found at 47 DAS at the highest infestation level (P<0.01), while Tiémarifing was significantly (P<0.01) affected at 47 and 61 DAS at the two highest infestation levels. In 2004 (right side of Figure 1), Striga infection resulted in significant (P<0.01) reductions in CO2 assimilation rates of CK60-B and E36-1 at 26 and 48 DAS. CO2 assimilation rate of Framida was only significantly (P<0.01) reduced at 48 DAS. Tiémarifing was the only genotype without a significant reduction in CO2 assimilation rate due to Striga-infection at any observation date. As photosynthesis and kernel dry weight in 2004 were measured on the same plants, a direct correlation between CO2 assimilation rate and kernel dry weight could be made. A highly significant (P<0.001) correlation between kernel dry weight and CO2 assimilation measured at 26 and 48 DAS was found (r = 0.61, N = 40; for both dates). No such correlation was found with the CO2 assimilation rates measured at 75 DAS. 59 Chapter 4 26 DAS 48 DAS CK60-B 35 CK60-B 25 15 5 -5 35 E36-1 E36-1 25 15 5 -5 35 Framida Framida 25 15 5 -5 35 Tiemarifing Tiemarifing 25 15 5 -5 0 400 800 1200 1600 2000 0 400 800 1200 1600 2000 Figure 2. CO2 assimilation rate (A; µmol CO2 m-2 s-1; y-axis) measured at different light intensities (PAR [400-700 nm]; µmol photon m-2 s-1; x-axis) on four sorghum genotypes (CK60-B, E36-1, Framida and Tiémarifing) for control (open symbols) and Striga-infected plants (closed symbols) at 26 (left) and 48 DAS (right) in the 2004 experiment. Bars represent (+/-) standard errors of means. 60 Detecting tolerance by photosynthesis measurements In 2004, photosynthesis was measured at a range of light intensities and the corresponding photosynthesis-light response curves of control and infected plants are presented in Figure 2. At 26 DAS (left side of Figure 2), CO2 assimilation of CK60-B was negatively (P<0.01) affected by Striga at all light intensities. For E36-1 Striga effects were significant (P<0.05) at 200 and 800 µmol PAR m-2 s-1 and highly significant (P<0.01) at 400 and 1600 µmol PAR m-2 s-1, at this stage. CO2 assimilation of infected Framida and Tiémarifing plants were not significantly reduced at 26 DAS. At 48 DAS (right side of Figure 2), significant reductions (P<0.05) in CO2 assimilation rate of infected CK60-B plants were observed at 200 µmol PAR m-2 s-1 and highly significant (P<0.01) reductions at all other light intensities. For E36-1 and Framida highly significant reductions (P<0.01) were observed at the three highest light intensities. Striga had no significant effect on CO2 assimilation of Tiémarifing at any light intensity. Table 3. Transpiration rate (Tr; mmol H2O m-2 s-1), stomatal conductance (gs; mol m-2 s-1) and intra-stomatal CO2 concentrations (Ci; ppm) of Striga-infected (3.0 [2003] - 4.0 [2004] seeds cm-3) and uninfected (0.0) plants at 47 DAS (2003) and 48 DAS (2004). The r-values represent the correlation coefficients between the parameters and the CO2 assimilation rate. 2003 Tr gs Ci 2004 Tr gs Ci a Seeds cm-3 CK60-B E36-1 Framida Tiémarifing 0.0 3.0 S.E.D 0.0 3.0 S.E.D 0.0 3.0 4.6aa 3.1b 0.47 0.31a 0.15b 0.055 91.6 64.3 4.4a 2.9b 0.23 0.29a 0.13b 0.035 85.8 64.3 4.5a 3.0b 0.42 0.31a 0.16b 0.038 89.3 64.2 4.6 4.0 0.0 4.0 S.E.D 0.0 4.0 S.E.D 0.0 4.0 S.E.D 5.0a 3.4b 0.27 0.19a 0.12b 0.011 42.1 42.8 r(1-tailed) (N=80) 0.86**b 0.36 0.25 0.90** 115.6 99.5 -0.02 (N=40) 0.85** 3.5 2.3 4.4 2.9 4.7 5.1 0.13a 0.08b 0.017 24.5 29.1 0.16 0.10 0.18 0.20 0.92** 35.1a 22.1b 3.68 39.0 65.0 0.08 values in the same column within the same year followed by a different letter are significantly (P<0.05) different 61 Chapter 4 Transpiration and stomatal conductance Transpiration rate (Tr in mmol H2O m-2 s-1), stomatal conductance (gs in mol m-2 s-1) and intra-stomatal CO2 concentration (Ci in ppm) measured at 47 (2003) and 48 DAS (2004) are presented in Table 3. Values represent those associated with the measurement of CO2 assimilation rate (A) at the highest light intensity and conducted for control sorghum plants and sorghum plants exposed to the highest Strigainfestation level. Significant reductions in Tr and gs of Striga-infected plants were commonly found with the more sensitive cultivars. In both years Tr and gs correlated highly significantly (P<0.01) with CO2 assimilation rate (A). The significant negative Striga effects on CO2 assimilation rate measured at 47 and 48 DAS were always associated with significant reductions in transpiration rate and stomatal conductance. Only for E36-1 in 2004 the reduction in A was not accompanied with a significant reduction in Tr. A significant reduction in Ci was only observed with Framida in 2004. This reduction was not associated with a significant reduction in CO2 assimilation. In both 2003 and 2004 no correlation between intra-stomatal CO2 concentration and CO2 assimilation was observed. Chlorophyll fluorescence In Figure 3, various parameters that were calculated based on the chlorophyll fluorescence measurements conducted in 2004 at 26 (left side) and 48 DAS (right side) are presented. Apart from photochemical quenching (Pq), non-photochemical quenching (NPq) and electron transport rate (ETR in µmol m-2 s-1) the ratio of CO2 assimilation over electron flow through PSII (A ETR-1) is presented. Presented data are based on the measurements conducted at the highest light intensity (1600 µmol PAR m-2 s-1at 26 DAS; 2000 µmol PAR m-2 s-1 at 48 DAS). For control plants, the CV of all parameters at both observation dates was smaller than the CV of the CO2 assimilation rate (26 DAS: 8.5% and 48 DAS: 16.4%) measured at the same time, on the same plants. Particularly for NPq (CV = 2.0% and 3.1%), A ETR-1 (CV = 3.5% and 6.8%) and Pq (CV = 4.0% and 8.3%) the differences in CV, compared to those of the CO2 assimilation rate, were considerable, whereas for ETR a relatively small difference was observed (CV = 7.0% and 12.9%). Significant genotype differences between control plants were only observed at 26 DAS for Pq (P<0.05) and at 48 DAS for NPq (P<0.05) and A ETR-1 (P<0.01). Striga effects on each parameter were analysed per measurement time (26 DAS and 48 DAS). Values of Pq and ETR were significantly (at 48 DAS: P<0.05) to highly significantly (at 26 DAS: P<0.01) affected by Striga infection for CK60-B. 62 Detecting tolerance by photosynthesis measurements 26 DAS 48 DAS 0.8 Pq 0.6 0.4 0.2 0.0 0.8 NPq 0.6 0.4 0.2 0.0 250 ETR 200 150 100 50 A ETR -1 0 0.20 0.15 0.10 0.05 0.00 CK60-B E36-1 Framida Tiemarifing CK60-B E36-1 Framida Tiemarifing Figure 3. Photochemical quenching (Pq), non-photochemical quenching (NPq), electron transport rate (ETR: µmol m-2 s-1) and photosynthesis per electron transport (A ETR-1) measured on four sorghum genotypes (CK60-B, E36-1, Framida and Tiémarifing) for control (open bars) and Striga-infected plants (shaded bars) at 26 (left) and 48 DAS (right) in 2004. Bars represent (+/-) standard error of means. 63 Chapter 4 For E36-1 these parameters were also negatively affected at both dates (P<0.05 at 26 DAS; P<0.01 at 48 DAS), while for Framida highly significant (P<0.01) Striga effects on Pq and ETR were only found at 48 DAS. An increase in NPq was only observed for CK60-B and E36-1 at 26 DAS (P< 0.01) and for CK60-B at 48 DAS (P< 0.05), whereas the ratio A ETR-1 was only significantly (P<0.01) reduced for CK60-B at both observation dates. For Tiémarifing, none of these chlorophyll fluorescence parameters were negatively affected by Striga. All four chlorophyll fluorescence parameters, measured at both Striga infected and uninfected sorghum plants, correlated highly significantly (P< 0.01) with CO2 assimilation rate, at both measurements times (Table 4). Relative reductions in CO2 assimilation rate correlated highly significantly (P<0.01) with Pq, ETR and A ETR-1 values measured on Striga-infected plants at both observation dates (P<0.01). For NPq correlations were highly significant (P<0.01) at 26 DAS and significant (P<0.05) at 48 DAS. Figure 4 shows the observed and fitted relation between CO2 assimilation (A) and electron transport (ETR) measured at 200, 400, 800 and 1600 or 2000 µmol m-2 s-1 (PAR) for infected and uninfected plants of each genotype at both measurement times (26 DAS: left side of Figure 4, and 48 DAS: right side of Figure 4). For all genotypes a linear relation through the origin gave an adequate description of the relation between A and ETR, indicating that the number of molecules of CO2 reduced per electron flowing through PSII was independent of light intensity. Analysis of variance per genotype and measurement time revealed that CK60-B was the only genotype showing a significant (P<0.001) decrease in slope for the infected compared to the control plants at both measurement times. This observation is identical to the conclusion drawn from the observations at the highest light intensity only (Figure 3). Table 4. Correlations between CO2 assimilation rate (A), relative reduction in CO2 assimilation rate (∆ A) and electron transport (ETR) the ratio CO2 assimilation per electron transport (A ETR-1) and photochemical (Pq) and non-photochemical (NPq) quenching Na 26 DAS 48 DAS A ∆A a ETR A ETR-1 Pq NPq ETR A ETR-1 Pq NPq 40 20 0.94 0.70 0.91 -0.86 -0.91 -0.70 -0.85 0.89 **b ** ** ** ** ** ** ** 0.95 0.79 0.93 0.70 -0.86 -0.88 -0.84 0.51 ** ** ** ** ** ** ** * Correlations with photosynthetic rate were analyzed with all data (control and Striga-infected plants), correlations with relative photosynthetic rate where analyzed based on Striga-infected plants only * and ** indicate significant correlations at the 0.05 and 0.01 probability level, respectively. 64 Detecting tolerance by photosynthesis measurements 50 40 26 DAS CK60-B 48 DAS CK60-B A(Control)= 0.178 ETR A(Control)= 0.178 ETR 30 20 10 A(Striga )= 0.149 ETR 0 50 40 A(Striga )= 0.144 ETR E36-1 E36-1 A(Control)= 0.158 ETR A(Control)= 0.175 ETR 30 20 10 A(Striga )= 0.169 ETR 0 50 40 A(Striga )= 0.1468 ETR Framida Framida A(Control)= 0.175 ETR A(Control)= 0.162 ETR 30 20 10 A(Striga )= 0.168 ETR 0 50 40 A(Striga )= 0.156 ETR Tiemarifing Tiemarifing A(Control)= 0.182 ETR A(Control)= 0.165 ETR 30 20 10 A(Striga )= 0.1634 ETR A(Striga )= 0.180 ETR 0 0 100 200 300 0 100 200 300 Figure 4. CO2 assimilation rate (A; µmol CO2 m-2 s-1; y-axis) as a function of electron transport rate (ETR; x-axis) measured on four sorghum genotypes (CK60-B, E36-1, Framida and Tiémarifing) for control (open symbols) and Striga-infected plants (closed symbols) at 26 (left) and 48 DAS (right) in 2004. Lines represent best fits obtained through linear regression. 65 Chapter 4 Discussion Genotype specific Striga effects on CO2 assimilation rate The observations on CO2 assimilation rate largely correspond to the earlier classification of sorghum genotypes with respect to tolerance and resistance to Striga, which was based on relative yield loss due to Striga and aboveground Striga numbers (Rodenburg et al., Accepted). CK60-B was earlier identified as a very susceptible and very sensitive genotype. This genotype showed immediate and highly significant reductions in CO2 assimilation rate upon Striga infection during its vegetative stage. E36-1 was earlier identified as a very susceptible and sensitive genotype. This was reflected in the Striga effects on CO2 assimilation of this genotype. E36-1 showed highly significant negative Striga effects on CO2 assimilation but these effects were exposed in a later stage than those on CK60-B. Framida was earlier identified as a resistant and moderately tolerant genotype. The genotype showed highly significant negative photosynthesis responses upon Striga infection but not at all observation times and only at high infestation levels. Negative effects on CO2 assimilation appeared later than on CK60-B. Tiémarifing was earlier identified as tolerant and moderately resistant. This genotype showed highly significant Striga effects on CO2 assimilation in 2003 but only at the latest measurement dates and highest infestation level. In the 2004 experiments, despite the presence of Striga, effects on photosynthesis of Tiémarifing were even completely absent. The significant correlation between CO2 assimilation measured at 26 and 48 DAS and kernel dry weight is a further indication that photosynthesis is an important indicator of the performance of Striga-infected sorghum plants. The absence of a significant correlation between kernel dry weight and CO2 assimilation measured at 75 DAS shows that during later growth stages photosynthesis is less indicative. The ability to maintain high rates of CO2 assimilation as mechanism to endure parasite infection was earlier reported by Gurney et al (2002a). They also suggested that this characteristic could be used as a screening measure for tolerance. Photosynthesis as screening measure for tolerance against biotic stresses has earlier been proven useful with Septoria nodorum in wheat (Scharen and Krupinsky, 1969). Selection measures should enable a quick and reliable assessment and preferably be low-cost. Obviously, the selection measure should discriminate clearly between tolerant and non-tolerant genotypes. Another requirement would be that in the absence of Striga the measure shows little variation, both within and among genotypes. Absolute measurements under Striga infestation, without Striga free controls, would than be sufficient to make a selection among genotypes. 66 Detecting tolerance by photosynthesis measurements Despite the relatively high coefficients of variation (CV: 8.5% to 32.4%), rates of CO2 assimilation of control plants of different genotypes were significantly to highly significantly different in four of seven measurements. This indicates that measurements on control plants are required for a selection based on CO2 assimilation rate. Following this procedure, only in 2004 at 48 DAS, Tiémarifing could be identified as the most tolerant genotype. If selection would have been based on measurements of 2003 only, Framida would probably been falsely identified as the most tolerant genotype. Hence, CO2 assimilation rate might not always be a reliable selection measure. Other serious constraints to the use of CO2 assimilation rate as screening measure are the relatively long time needed per measurement and the high costs of photosynthesis measurement systems. In both experiments, CO2 assimilation rate correlated very strongly with transpiration rate and stomatal conductance. Significant reductions in CO2 assimilation were never associated with significant reductions in the intra-stomatal CO2 concentration. It must be noted that internal CO2 concentrations are not measured directly but result from calculations based on stomatal conductance. Stomatal conductance, in turn, is computed based on transpiration rate and the vapour pressure difference between inside and outside the leaf (Von Caemmerer and Farquhar, 1981). Due to that, the calculated internal CO2 concentrations are sensitive to small perturbations or measurement errors. Still the observations suggest that Striga effects on stomatal conductance are not the main reason for photosynthetic reduction because such a response would lead to a transitional decrease of the internal CO2 concentration due to the increased resistance for CO2 diffusion (Kropff, 1987). This confirms results of an earlier study by Press et al. (1987a), but seems to contradict findings from Frost et al. (1997). A direct effect of Striga on the photosynthetic apparatus of its host plant offers good perspectives for using chlorophyll fluorescence to replace the CO2 exchange measurements. Chlorophyll fluorescence parameters as screening measure The regulation of photosynthesis in response to stress involves the protection of the photosynthetic apparatus. Photochemical and non-photochemical quenching are two essential elements of this photoprotection (Ort and Baker, 2002). Photochemical quenching is proportional to the energy transfer to the functional photosynthesis reaction centres. Non-photochemical quenching (NPq) refers to the process of dissipation of the excess excitation energy in the PSII antennae as heat, whereby down-regulation of PSII electron transport efficiency is triggered. As the capacity for photochemistry of leaves reduces under stress conditions, both photochemical and non-photochemical quenching are potentially suitable measures for stress severity 67 Chapter 4 (Schreiber, 1986) or stress tolerance (Harbinson, 1995). This was demonstrated by studies on cold tolerance in maize (Schapendonk et al., 1989a; Fracheboud et al., 1999) and drought tolerance in wheat (Havaux and Lannoye, 1985), barley (Nogues et al., 1994; Olsovska et al., 2000) and potato (Schapendonk et al., 1989b, 1992). Some of the Striga effects show remarkable resemblance with drought stress effects (e.g.White and Wilson, 1965; Björkman and Powles, 1984) which opens the way for rapid selection of Striga tolerant genotypes, using fluorescence analyses. The observations in 2004, when chlorophyll fluorescence was measured in addition to CO2 assimilation, confirm this. All four chlorophyll fluorescence parameters (ETR, A ETR1 , Pq, NPq) of Striga infected plants showed a strong correlation with CO2 assimilation rate or the relative reduction in CO2 assimilation rate due to Striga infection. In addition, all parameters were characterized by a relatively small CV and in spite of this only in three occasions (Pq at 26 DAS and Nq and A ETR-1 at 48 DAS) a significant genotype difference between control plants was observed. This indicates that for screening purposes measurements on control plants do not always seem to be a prerequisite. As also the equipment required for measurement of chlorophyll fluorescence costs only about 10% of that of the gas exchange equipment, screening based on chlorophyll fluorescence is a cost effective alternative to screening using photosynthesis measurements based on gas exchange. Clear differences in suitability of the various fluorescence parameters for use as screening measure are present. The most straightforward approach to estimate photosynthetic capacity from fluorescence analysis is provided by ETR, which is based on the efficiency of electron transport through PSII (Φ 2) and the absorbed light intensity. Parameter Φ2 is based on Ft and Fm', which are both measured in the light. At 26 DAS, CK60-B and E36-1 could already be identified as sensitive genotypes and at 48 DAS Framida also showed a significant reduction in ETR and could be distinguished from Tiémarifing. Determination of ETR requires fairly stable or saturated light intensities throughout the selection procedure, as ETR is dependent on light intensity. The ratio A ETR-1 showed to be independent of light intensity. However, the photosynthesis per electron transport only helped to identify CK60-B as a sensitive genotype. Possibly a reduction in photosynthesis per electron transport only appears in very sensitive genotypes, or at relatively high levels of infection. For this reason it is not believed to be a helpful parameter for the identification of superior tolerant genotypes. Furthermore, the calculation of this ratio would require the determination of both chlorophyll fluorescence and CO2 assimilation rate. For the sensitive genotypes, significant increases in non-photochemical quenching (NPq) following Striga infection were observed at 26 (CK60-B and E36-1) and 48 DAS (only CK60-B). Based on NPq values, no distinction could be made 68 Detecting tolerance by photosynthesis measurements between Framida and Tiémarifing at any measurement time. Moreover, in absolute terms the increases observed with the sensitive genotypes were marginal. For instance the NPq values measured on infected plants of CK60-B and E36-1 at 26 DAS fall well in the range of NPq values obtained on non-infected plants at 48 DAS. For that reason, NPq is not considered a very suitable screening measure for tolerance, particularly not in the absence of control plants. Another disadvantage of NPq is that its calculation involves Fm, which means that chlorophyll fluorescence should also be measured in the dark, requiring a dark-adaptation period. Photochemical quenching (Pq) showed to alter significantly upon Striga infection in sensitive genotypes. Tiémarifing was the only genotype with virtual unchanged Pq values due to Striga infection throughout the experiment. Hence, through measurement of Pq it is possible to identify genotypes with superior levels of tolerance. Possible drawback is the observed presence of a significant genotype effect on Pq values of Striga-free control plants at 26 DAS, which would require the inclusion of control plants in the screening trial, to calculate relative changes. However, at 48 DAS the genotype effect on control values of Pq was not longer present. It is therefore expected that Pq of uninfected control plants only shows a genotype effect at certain (early) phenological stages. Major constraint for Pq measurements as selection tool seems to be the requirement of far-red light to determine Fo'. Provision of far-red light is often not available on standard equipment for measuring chlorophyll fluorescence. However, such a technical constraint could be overcome. Alternatively, a method is available which estimates Fo' through a simple equation involving the minimum fluorescence yield in the dark-adapted state (Fo), the maximum fluorescence yield in the dark-adapted state (Fm), and the maximum fluorescence yield in the light-adapted state (Fm') (Oxborough and Baker, 1997). Disadvantage of this alternative is again the requirement to conduct fluorescence measurements in both light and dark-adapted conditions. Development of a screening protocol Based on the current results it is concluded that chlorophyll fluorescence parameters, particularly ETR and Pq, carry good potential for the development of a discriminative and cost effective screening procedure for host plant tolerance to Striga hermonthica. This however requires the design of a measuring protocol that should be evaluated with a wider range of genotypes and preferably with combined measurements of gas exchange and fluorescence to test for what conditions the attempted screening protocol is actually valid. Earlier studies where chlorophyll fluorescence measurements were used as selection measure for tolerance (e.g. Havaux and Lannoye, 1985; Nogues et al., 1994; Fracheboud et al., 1999; Olsovska et al., 2000) dealt with abiotic stresses 69 Chapter 4 such as cold or drought. Striga is a biotic stress and this implies some additional difficulties for screening. As much as genotypes may differ in tolerance to an biotic stress factor, they may differ in resistance. Equivalence of parasite infection or removal of resistance effects as a confounding factor is one of the first requirements for measuring tolerance (Schafer, 1971), and typically this aspect was identified as the main constraint for developing a simple screening procedure based on actual yield data (Rodenburg et al., Accepted). Differences in Striga effects on photosynthesis and chlorophyll fluorescence between genotypes as found in this study may partly result from differences in resistance, since significant differences in aboveground Striga numbers were observed between Tiémarifing and Framida on the one hand and CK60B and E36-1 on the other. Completely cancelling out all differences in resistance seems impossible. The most practical solution would be to use very high infestation levels (three to four viable Striga seeds cm-3 equivalent to 300- 400,000 seeds m-2 of the upper 10 cm in the field). This should prevent that more resistant genotypes, such as Framida in this study, will be falsely identified as tolerant, due to the fact that they simply do not have enough infections to damage the host plant sufficiently. Use of high levels of Striga infestation in screening trials for tolerance against Striga was earlier recommended by Kim (1991). In agreement with other reports (e.g. Olivier et al., 1991a; Ast et al., 2000; Rodenburg et al., Submitted), resistance also resulted in significantly delayed parasitism and hence later effects. Because of differences in Striga emergence between genotypes the screening should not be carried out too early. For a reliable and fair screening based on photosynthesis or chlorophyll fluorescence measurements Striga should have had sufficient opportunity to establish on all genotypes. In addition, this study showed that screening also has an upper time limit. The 2004 experiment showed that after flowering (50% flowering stages were recorded between 52 and 61 DAS) of sorghum, photosynthesis did no longer correlate with yield and was no longer consistent with observed overall tolerance or sensitivity of the genotype to the parasite. The most significant differences in photosynthesis and chlorophyll fluorescence in this study were found in the early stages of the host-parasite relation. This is expected, based on the earlier observation by Parker and Riches (1993) that Striga exerts already severe effects on the host when host plants are young and Striga is still belowground. Early Striga effects on photosynthesis were observed in several other studies as well (e.g. Graves et al., 1989, 1990; Gurney et al., 1995). It is therefore recommended to do the screening shortly after first Striga emergence and well before sorghum flowering. A solution to correct for the delay in parasitism could be to make measurement time for each genotype dependent on first Striga emergence or to conduct measurements on two specific moments in time. 70 CHAPTER 5 Effects of host plant genotype and seed bank density on Striga reproduction1 J. Rodenburgab, L. Bastiaansa, M. J. Kropffaand A. van Asta a Group Crop and Weed Ecology, Wageningen University, the Netherlands b Africa Rice Center (WARDA), Cotonou, Benin Abstract Prevention of seed input in the seed bank of Striga hermonthica infested fields is an important objective of Striga management. In three consecutive years of field experimentation in Mali, Striga reproduction was studied for ten sorghum genotypes at infestation levels ranging from 30,000 to 200,000 seeds m-2. Resistance was identified as an important determinant of Striga reproduction, with the most resistant genotypes (N13, IS9830 and SRN39) reducing Striga reproduction with 70-93% compared to the most susceptible genotype (CK60-B). Seed bank density was another factor having a significant effect on Striga seed production. The relation between seed bank density and Striga reproduction was non-linear. Density-dependent reduction in seed production resulted mainly from intra-specific competition between aboveground Striga plants. For the most susceptible genotypes density dependence also occurred in the earlier belowground stages. Striga reproduction continued beyond crop harvest. At the high infestation level just 8% of the total reproduction was realized after harvest, whereas at the low infestation level 39% was attained after harvest. Even though host plant genotype plays a significant role in Striga reproduction, calculations indicated that only at very low infestation levels the use of the most resistant genotype was able to lower the Striga seed bank. 1 Submitted to Weed Research 71 Chapter 5 Introduction Cereal production in the semi-arid to sub-humid tropics is often limited by the obligate, out-crossing, hemi-parasitic weed Striga hermonthica (Del.) Benth, a member of the Orobanchaceae (formerly: Scrophulariaceae) family. Striga hermonthica parasitizes on roots of cereals like sorghum (Sorghum bicolor [L.] Moench), pearl millet (Pennisetum glaucum [L.] R. Br.), maize (Zea mays [L.]) and upland rice (both Oryza glaberrima [Steudel] and O. sativa [L.]; [Johnson et al., 1997]). Infection by Striga can cause severe yield losses of up to 80-85%, depending on the level of resistance and tolerance of the specific host genotype (Obilana, 1983; Rodenburg et al., 2005). The Striga problem has become increasingly important in the sub-Saharan regions of Africa (Lagoke et al., 1991). This increase is caused by the good reproduction opportunities for Striga plants that are created by an intensification of land use, where suitable host plants are grown continuously on the same fields (Weber et al., 1995). Striga produces numerous, very small seeds per plant. Striga seed size is 0.2 to 0.3 mm (Parker and Riches, 1993). Estimates of seed production per plant vary from 5,000 up to 85,000 seeds per reproductive plant (Andrews, 1945; Stewart, 1990; Webb and Smith, 1996). One host plant can support several seed producing Striga plants simultaneously and a substantial part of the newly produced Striga seeds survive the subsequent dry season until the next cropping season. Therefore, the Striga seed bank in the soil easily increases with every new cropping season with the same host species. Delft et al. (1997) concluded that only two to three seed producing Striga plants per m2 would be enough to balance the seed bank. Prevention of seed input in the seed bank and reduction of the soil seed bank of Striga infested fields are among the most important objectives for Striga management (Ramaiah, 1987b). Host plant defence mechanisms against Striga can contribute to this (Hess and Haussmann, 1999), particularly since host resistance against Striga is expected to reduce Striga seed production (Doggett, 1988; Ejeta et al., 2000). Complete resistance, or immunity, against Striga in cereals has not yet been reported. Even varieties possessing the most effective resistance mechanisms against Striga can still not completely prevent some individual Striga plants to emerge and complete their life-cycle (e.g. Rodenburg et al., 2005). The assumed reduction in seed production rate of resistant genotypes has been attributed to a slower development of Striga, a reduced number of emerged Striga plants, or a reduced number of flowering and capsule bearing Striga plants ( Weber et al., 1995; Carsky et al., 1996; Kim and Adetimirin, 1997b). Objective of this study was to test whether there is a significant host plant genotype and seed bank density effect on Striga reproduction. It was hypothesized that resistant genotypes, that reduce the number of aboveground Striga plants, are able to create a proportional reduction in Striga seed production. In that 72 Genotype and seed bank density effects on Striga reproduction case, screening for resistant genotypes will automatically yield genotypes that reduce Striga reproduction. An additional objective was to find a characteristic for Striga seed production that is simple and easy to measure. Furthermore, it was investigated whether Striga seed production continues after harvest of the host plant and how significant this additional post-harvest seed production is. Material and Methods Genetic material The host species in this study is Sorghum bicolor (sorghum), because it is the most important host species for reproduction of Striga hermonthica (Weber et al., 1995). Ten different sorghum genotypes were selected: CK60-B, CMDT39, E36-1, Framida, IS9830, N13, Seredo, Serena, SRN39 and Tiémarifing. These genotypes originate from different parts of the world and were selected for their differences in level and mechanism of defence against Striga. The genotypes ranged from resistant (N13) to susceptible (E36-1) and from tolerant (Tiémarifing) to sensitive (CK60-B) and comprised various combinations of these reaction types (Ast et al., 2000; Hess, 1989; Olivier et al., 1991; Rodenburg et al., 2005). Striga hermonthica seeds, used to create Striga infested plots, originated from Samanko, Mali and were derived from plants parasitizing on sorghum. In 2001, Striga seed from 1998 was used (viability 82.5%) and in 2002 a mixture of seed from 1995-1997 and 2001 (viability 73%) was used. In 2003, a mixture of Striga seed from 1995-1998 and 2001 (viability 10.5%) was used, supplemented with seeds from 2002 (78.7%) to arrive at the desired infestation level. Location The field trials were conducted in 2001, 2002 and 2003 at the ICRISAT field station in Samanko, 20 km Southwest of Bamako, the capital of Mali (latitude: 8°54”W and 12°54”N, altitude: 329 m). The climate type is Sudanese, characterized by a single rainy season between May and October. The average temperature during the cropping season (June-November) is 29.1°C. Mean annual rainfall at the field station is 950 mm, of which 96% falls between May and October. Each year a different field was used, adjacent to that of the previous year. Experimental fields had sandy-loam, ferruginous tropical soils with wash out spots and concretions. Table 1 presents soil fertility parameters of the main plots of the three fields (2001-2003) after fertilization. Figure 1 presents cumulative rainfall, in the 3 cropping seasons, over time. 73 1200 800 400 2001 2002 2003 0 Ap ri l M ay Ju ne Ju Au l y Se gu pt st em b O er ct N obe ov em r be r Cumulative rainfall (mm) Chapter 5 Figure 1. Cumulative rainfall (mm) during the cropping seasons of 2001-2003, in Samanko, Mali. Table 1. Soil fertility parameters after fertilization: pH (H2O; 1:2.5), C-organic (% C.O.), Pavailable (Bray-1; mg P kg-1) and N-total (mg N kg-1) of the main plots of the study fields from the three years (2001, 2002, 2003 low infestation [L] and 2003 high infestation [H]) 2001 2002 2003 (L) 2003 (H) pH 4.91 5.59 4.93 5.07 C-organic 0.27 0.70 0.35 0.37 P-available 9.17 21.01 12.15 13.56 N-Total 227.5 486.4 248.4 256.3 Experimental design In 2001 and 2002, a completely randomised block design with sorghum genotype (10) as treatment was used in either five (2001) or eight (2002) replicates. In 2003 a splitplot design in eight replications was used with two Striga infestation densities (high [H] and low [L]) at the main-plot level and ten sorghum genotypes at the sub-plot level. Each plot or sub-plot, representing one sorghum genotype, contained four crop rows, of 4 m (2001), 7.6 m (2002) and 6.4 m (2003) of which the middle two rows were used for observations. Row distance was 0.8 m and plant distance in the row was 0.2 (2001) and 0.4 m (2002 and 2003), corresponding to plant densities of 62,500 (2001) and 31,250 plants ha-1 (2002 and 2003). 74 Genotype and seed bank density effects on Striga reproduction Field preparation The soil of the experimental field was tilled, levelled and fertilized prior to Striga infestation and sorghum sowing. Fertilization was done at a rate of 100 (2001) and 200 kg ha-1 (2002 and 2003) of N-P-K (17-17-17) and in 2002 gypsum (100 kg ha-1) was added to increase soil pH. Artificial Striga infestation of the upper 5 (2001) and 10 cm (2002 and 2003) was created by dispersing Striga seeds mixed with sand. This was done 2 weeks before sowing of the host plant crop, to allow pre-conditioning of the Striga seeds. Striga infestation levels were 45,000 (2001), 200,000 (2002) and 30,000 (2003L) and 150,000 (2003H) viable Striga seeds m -2. In all years, blocks or mainplots were surrounded by small dikes to prevent Striga inflow from adjacent blocks through soil run-off after rain showers. Sorghum was hand sown at 13 (2001), 6 (2002) and 5 July (2003), with six seeds per pocket at 3 cm depth. Sorghum plants were thinned to one plant per pocket at 21 days after sowing (DAS). Throughout the season, experimental plots were kept free of weeds other than Striga hermonthica. Harvest time depended on sorghum genotype and year (110-119 DAS in 2001, 102-132 DAS in 2002 and 118-132 DAS in 2003). Observations and sampling In each plot, Striga sampling areas, containing either ten (2001 and 2002) or four (2003L and 2003H) host plants, were selected for observations on the parasitic weed. First Striga emergence date (Edate) and dates of first flowering (Fdate) were registered. Aboveground Striga plants (result of emergence and death) were counted weekly (2001) or bi-weekly (2002 and 2003). These counts were used to determine the maximum aboveground Striga number (NSmax). In 2001 and 2002, the number of generative Striga plants was counted at crop harvest, resulting in NSgen. In both years, all (living and dead; generative and vegetative) aboveground Striga plants of the sampling area were collected at harvest, dried and weighted for total aboveground Striga dry weight (DWtot). Dry weight of the flowerstalks was separately determined (DWstalks). Flowerstalks were defined as the generative part of the branches of a Striga plant, from the oldest flower or capsule to the top. In 2003, starting from the first Striga flowering date, dead generative Striga plants were collected weekly from the observation plots and together with the living generative plants that were sampled at harvest used for assessment of NSgen, DWtot and DWstalks. Additionally, Striga seed production in 2003 was estimated based on the number of capsules produced throughout the growing season. For all generative plants (dead or living) at harvest, the number of capsules was counted, resulting in the total number of capsules (NRcaps). This method ensured inclusion of all generative plants in 75 Chapter 5 the assessment of reproductive effort. In the adjacent observation area of equal size (four host plants) collection of dead Striga plants, and the observation of Striga characteristics, continued after harvest until all plants were dead. Comparison of the results of both sampling areas gives an indication of the importance of Striga seed production after harvest of the crop. Just after sorghum harvest in 2003, ten ripe seed capsules from five randomly selected Striga plants (two capsules per plant) were sampled in each individual sub-plot. Those seed capsules were dried and its content was weighted to get an estimate of the seed production per capsule. Additionally individual Striga seed weight was assessed. Statistical analyses Data on Edate, Fdate, NSmax, NS gen, DWtot, DWstalks and NRcaps were subjected to analyses of variance (ANOVA), followed by a comparison of means with the least significant difference (L.S.D.), using the Genstat (release 7.1) statistical software package. To meet the assumptions of the analysis of variance, prior to analysis all data, except Striga development dates, were subjected to square root ([X+c]1/2) transformation, where X was the original, individual observation and c was set to 0.5 (Sokal and Rohlf, 1995). In tables with comparisons between genotypes, rankings from 1 to 10 were used. Rank 1 was assigned to the best performing genotype, from the Striga control perspective, and rank 10 was assigned to the worst performing genotype (e.g. the genotype with the highest capsule number, Striga number or dry weight). Correlations in this study were one-tailed Pearson’s correlations, based on individual data points (original data or square-root transformed: [X+c]1/2, with c=0.5). Correlations and linear regressions were carried out with the SPSS (version 10.0) statistical software package. Correlations in this study were phenotypic correlations (r). Due to relatively high environmental variation it was not possible to calculate genetic correlations. Repeatabilities set an upper-limit to the heritability of a variable. Repeatability (R) of capsule numbers was calculated according to Falconer and Mackay (1996): R = (VG + V Eg ) / V P = 1 − (V Es / V p ) where VP is the total phenotypic variance, which is composed of three components: (1) VG the genetic variance, (2) VEg the environmental variance due to permanent environmental effects on the phenotype and (3) VEs the environmental variance due to temporary or localized environmental effects on the phenotype. 76 Table 2. First Striga emergence (Edate) and first flowering dates (Fdate), in days after sowing (DAS), and ranking (numbers 1-10) per sorghum genotype and infestation level, for three subsequent years (2001 - 2003) and infestation levels (2003L: low and 2003H: high). 2001 2002 Fdate Edate 2003L Fdate Edate 2003H Fdate Edate Fdate Edate 33.6 10 73.6 9 28.9 dea 9 66.9 d 10 37.3 b 10 77.5 e 10 28.8 10 75.1 10 CMDT39 38.4 5 79.2 4 32.5 abc 4 74.8 ab 3 43.9 ab 3 88.3 bcd 7 32.3 4 83.1 2 E36-1 37.6 6 73.4 10 28.6 e 10 70.1 bcd 8 37.9 b 8 85.4 de 9 30.6 8 79.3 7 Framida 37.2 7 81.0 2 32.0 abc 5 74.1 abc 5 47.5 a 2 93.9 abc 3 34.4 2 78.0 9 IS9830 44.2 2 78.2 6 33.6 ab 3 75.5 a 1 37.4 b 9 91.3 abcd 5 29.4 9 79.0 8 N13 40.8 3 78.9 5 31.6 bcd 6 69.4 cd 9 50.3 a 1 96.7 a 1 31.6 6 83.3 1 Seredo 37.0 8 80.6 3 33.6 ab 2 75.4 a 2 38.5 b 7 92.4 abcd 4 31.9 5 81.4 4 Serena 36.4 9 76.8 8 31.1 bcd 7 74.6 ab 4 43.8 ab 4 95.3 ab 2 37.3 1 81.3 5 SRN39 44.2 1 81.0 1 34.8 a 1 70.5 abcd 7 38.6 b 6 87.0 cd 8 33.8 3 80.9 6 Tiémarifing 39.2 4 77.2 7 30.5 cde 8 72.4 abc 6 41.9 ab 5 90.3 abcd 6 31.5 7 82.0 3 S.E.D. (df) nsb ns 1.42 (79) 2.56 (79) 4.38 (79) 4.05 (79) ns mean 38.9 78.0 31.7 41.7 ab 89.8 a 32.1 a 72.4 means in the same column followed by a different letter are significantly different at the 0.05 level ns means not significant c emergence and flowering are significantly affected by Striga infestation level (P<0.001) b ns b 80.3 b 77 Genotype and seed bank density effects on Striga reproduction CK60-B Chapter 5 Results Striga emergence and flowering Only in 2002 and in 2003L host plant genotype had a significant (P<0.05) effect on first Striga emergence and flowering dates (Table 2). In 2002, Striga plants on E36-1 and CK60-B emerged significantly earlier than on most other genotypes whereas on SRN39, Seredo and IS9830 emergence was significantly delayed compared to some other genotypes. In 2003L, emergence was significantly delayed on Framida and N13, compared to most other genotypes. In both years, genotype rankings based on flowering dates corresponded roughly with rankings based on emergence dates. However the data also show that early emergence did not automatically result in early flowering and vice versa (e.g. SRN39 and Serena in 2002 and CMDT39 in 2003L). Striga development response to host genotype is rather erratic. Throughout the years, early (rank 1 to 3) and late (8 to 10) emergence was observed at six different genotypes and early and late flowering was observed at seven and eight different genotypes, respectively. The only genotypes with a consistent fast development are CK60-B and E36-1. Striga on SRN39 emerged consistently late compared to the other genotypes. However, this was not followed by consistent later flowering. Both emergence and flowering dates of Striga were significantly (P<0.001) affected by Striga seed bank density in 2003 (Table 2). In the low infested plots, Striga development (both emergence and flowering) was on average 9.5 days later than in the high infested plots. Furthermore in 2001 at 45,000 seeds m-2, Striga emergence and flowering were on average 7 and 5.5 days (respectively) later than in 2002 at 200,000 seeds m-2. Striga numbers In all years, there was a significant genotype effect (P<0.01) on both maximum number of aboveground Striga plants per host plant (NS max) and number of generative Striga plants per host plant (NS gen) (Table 3). Within years and infestation levels, genotype rankings on NSmax corresponded to rankings on NSgen. Furthermore, rankings based on NSmax and NSgen were relatively stable over years and infestation levels, except for CMDT39. Throughout the years, the highest maximum number of Striga plants, as well as the highest number of generative plants was recorded on genotypes CK60-B, E36-1, Seredo and Serena (mostly ranked as 10, 9, 8 and 7). Genotype N13, always ranked as 1 (often significantly) and SRN39 and IS9830 had very low numbers as well while Framida and Tiémarifing held intermediate positions. In 2003, no major differences in genotype rankings based on NSmax were found between the low and high infestation level. 78 Genotype and seed bank density effects on Striga reproduction Table 3. Means and rankings (numbers 1-10) of maximum number of aboveground Striga plants (NSmax) and number of generative Striga plants (NSgen) per genotype, year (2001 2003) and infestation level (2003L:low and 2003H:high); all parameters are expressed per host plant (equivalent to 0.32 m-2). 2001 2002 NSmax NSgen NSmax NSgen CK60B 2.4 ba 9 0.4 bc 6 103.1 a 10 35.3 ab 9 CMDT39 0.6 cd 2 0.2 bc 2 92.6 a 8 10.5 de 2 E36-1 7.9 a 10 1.8 a 10 99.9 a 9 20.0 cd 7 Framida 1.3 bc 6 0.3 bc 5 54.0 c 4 16.4 cd 5 IS9830 0.9 bcd 4 0.4 bc 7 29.3 d 2 11.3 de 3 c 1 N13 0.1 d 1 0.0 c 1 9.3 e 1 5.7 Seredo 2.2 b 8 0.5 bc 8 81.8 ab 7 22.5 bc 8 Serena 1.7 bc 7 0.6 b 9 80.9 ab 6 42.3 a 10 SRN39 0.7 cd 3 0.3 bc 3 33.6 d 3 14.8 cd 4 Tiémarifing 1.2 bc 5 0.3 bc 4 67.9 bc 5 17.6 cd 6 0.61 (79) b S.E.D. (df) 0.26 (49) 0.15 (49) 2003L e 0.65 (79) 2003H NSmax NSgen NSmax NSgen CK60B 20.9 a 10 12.5 a 10 64.8 a 10 32.1 a 10 CMDT39 7.1 6 4.7 b 8 30.0 cd 5 14.9 bcd 5 E36-1 17.1 a 9 9.6 a 9 48.9 ab 9 19.7 b 9 Framida 5.7 bc 5 2.3 b 4 37.3 bcd 6 18.4 bc 7 IS9830 3.3 cd 2 2.0 bc 2 16.6 e 3 10.3 de 3 N13 0.8 d 1 0.2 c 1 4.2 f 1 2.6 f 1 Seredo 8.1 b 8 4.5 b 7 43.1 bc 8 19.5 b 8 Serena 7.9 bc 7 3.6 b 6 38.7 bcd 7 15.1 bcd 6 SRN39 3.4 cd 3 2.2 b 3 16.1 e 2 7.3 2 Tiémarifing 5.0 bc 4 3.2 b 5 25.6 de 4 11.9 cde 0.64 (79) b bc S.E.D. (df) 0.48 (79) 0.38 (79) e 4 0.43 (79) a Values in the same column followed by a different letter are significantly different from one another at the 0.01 level b 1/2 ANOVA was based on square root (X+0.5) transformations, S.E.D. -values of transformed data are given c Value NSgen N13 in 2001 = 0.04 79 Chapter 5 Striga biomass Sorghum genotype had a significant effect (P<0.05) on total aboveground Striga dry weight at harvest (DWtot) in 2001, 2002, 2003L and 2003H as well as on flowerstalk dry weight (DWstalks) in 2001, 2003L and 2003H (Table 4). In 2003, there were highly significant main effects of both genotype and infestation level (not shown) on DWtot and on DWstalks (P<0.001). There were significant (one-tailed) correlations between means of DWtot of different years (P<0.05; not shown), except between 2001 and 2002, the years with the lowest and highest infection level. The good correlations were also reflected in rather consistent genotype rankings for DWtot over years. Throughout the years and infestation levels, relative high values of DWtot were recorded on CK60-B (ranked 9-10), CMDT39 (except in 2001), E36-1 (except in 2002), Tiémarifing and Seredo whereas, relative low values of DWtot were recorded on N13 (ranked 1), SRN39 and IS9830. Genotype rankings based on DWtot and DWstalks were largely similar. There were significant (one-tailed) correlations between means of DWstalks of different years (P<0.05; not shown), except between 2001 and 2002. Consequently, genotype ranking based on DWstalks of 2002, 2003L and 2003H corresponded largely with one another. Throughout the years, consistently high DWstalks were recorded on CMDT39 and CK60-B. Striga on N13 produced the lowest flowerstalk dry weight throughout the years. On other resistant genotypes, SRN39 and IS9830, Striga also produced relatively low flowerstalk biomass, except in 2001 (IS9830 and SRN39) and 2002 (SRN39). Remarkable were the relatively low Striga dry weights in 2002. Both DWtot and DWstalks in 2002 were much lower than in plots of 2003H, whereas the maximum numbers of aboveground Striga plants (NSmax) in 2002 were higher than in 2003H. Again, this might be due to the difference in observation methods between both years. Striga numbers and reproduction Correlations between Striga number (NSmax and NSgen) and Striga reproduction parameters (DWstalks and NRcaps) were highly significant (P<0.01) in all years and at all infestation levels (Table 5). Figures 2A-D show observed variation in NSmax and DWstalks among genotypes for different years and infestation levels. Identical infection levels resulted often in strongly different DWstalks, whereas there were also cases where different infection levels resulted in comparable DWstalks. This shows that DWstalks is not only a result of NSmax and, consequently, that there are genotype effects on reproduction other than resistance alone. Dotted lines in these figures indicate the minimum and maximum flowerstalk production per emerged Striga plant. 80 Genotype and seed bank density effects on Striga reproduction Table 4. Means and ranking (numbers 1-10) of total aboveground Striga dry weight at harvest (DWtot : g) and dry weight of generative Striga parts at harvest (DWstalks: g) per genotype, year (2001 - 2003) and infestation level (2003L:low and 2003H:high); all parameters are expressed per host plant (equivalent to 0.32 m-2). 2001 2002 DWtot DWstalks DWtot DWstalks CK60-B 2.1 aa 9 0.7 a 10 40.3 a 10 10.7 10 CMDT39 0.2 b 2 0.1 c 2 38.7 ab 9 10.0 9 E36-1 2.9 a 10 0.5 ab 9 23.1 cd 5 5.3 5 Framida 0.4 b 5 0.1 c 3 16.0 cde 3 4.2 1 IS9830 0.4 b 4 0.2 bc 6 14.6 de 2 4.2 2 c 1 11.8 e 1 4.9 4 N13 0.2 b 1 0.0 c Seredo 0.7 b 7 0.1 c 4 23.5 bcd 6 4.9 3 Serena 0.7 b 6 0.2 bc 8 26.2 abcd 7 6.7 7 SRN39 0.4 b 3 0.2 bc 5 17.0 cde 4 5.7 6 Tiémarifing 0.7 b 8 0.2 bc 7 27.3 abc 8 7.3 8 b S.E.D. (df) 0.176 (49) 0.098 (49) 0.642 2003L (79) ns d 2003H DWtot DWstalks DWtot DWstalks CK60-B 57.7 a 10 20.8 a 10 98.9 a 10 30.9 a 10 CMDT39 24.2 bc 8 9.4 b 9 61.1 b 8 18.9 b 8 E36-1 28.7 b 9 8.7 b 8 58.9 b 7 16.3 bc 5 Framida 6.8 de 2 2.4 cd 2 51.9 bc 5 16.4 bc 6 IS9830 7.7 d 3 3.3 c 4 42.5 bc 3 14.6 bcd 4 N13 0.7 e 1 0.4 d 1 20.2 d 1 7.8 d 1 Seredo 15.1 bcd 7 5.3 bc 6 66.6 b 9 19.7 b 9 Serena 10.7 cd 5 3.7 c 5 49.6 bc 4 13.0 bcd 3 SRN39 7.9 d 4 3.1 cd 3 30.6 cd 2 9.0 cd 2 Tiémarifing 14.0 bcd 6 5.6 bc 7 55.2 bc 6 17.4 b 7 0.861 (79) 0.485 (79) 0.997 (79) 0.552 (79) b S.E.D. (df) a Values in the same column followed by a different letter are significantly different from one another at the 0.05 level b 1/2 ANOVA was based on square root transformations (X+0.5) , S.E.D. values of transformed data are given c Value of N13 2001 DWstalks = 0.004 d ns means not significant 81 Chapter 5 Table 5. Pearson correlations between maximum aboveground Striga numbers (NSmax) and number of generative Striga plants and harvest (NSgen), Striga flowerstalk dry weight at harvest (DWstalks), and capsule number per host plant (NRcaps) at harvest, per year (2001 2003) and infestation level (2003L:low and 2003H:high); all parameters are expressed per host plant. NSmax DWstalks 2001 2002 2003L 2003H 0.64 0.38 0.80 0.58 0.79 0.56 0.90 0.78 0.89 0.74 NRcaps NSgen DWstalks 0.69 0.49 NRcaps a All data were square root-transformed (X+0.5) 1/2 b all (one-tailed) correlations are significant (P<0.01) From these figures it appears that particularly on E36-1 and Seredo flowerstalk dry weight per emerged Striga plant was often low. For N13, the results depended a lot on infestation level. At low infestation levels (2001 and 2003L), flowerstalk production was negligible, whereas at high Striga infestation levels (2002 and 2003H) the relatively low total flowerstalk production resulted from few emerged Striga plants with a very high flowerstalk production per emerged Striga plant. Figures 3A-D show the relations between maximum aboveground Striga numbers (NSmax) and reproductive potential of Striga plants for ten genotypes at two infestation levels, as observed in 2003. A five times higher infestation level of 2003H compared to 2003L resulted in a proportional increase in NSmax for almost all genotypes, except the least resistant ones (E36-1 and CK60-B). This increase in number of emerged Striga plants was accompanied with a disproportional increase in DWstalks (Figure 3A), with the most resistant genotype (N13) as a clear exception. Figure 3B shows the relation between flowerstalk dry weight per aboveground Striga plant (DWstalks per NSmax) and maximum aboveground numbers (NSmax). This figure shows negative slopes for most genotypes, confirming that at higher infection levels the reproductive effort per Striga plant is diminishing. For some of the least resistant genotypes (E36-1, Seredo and Serena) the differences in Striga flowerstalk dry weight per plant between the low and the high infestation levels were relatively small, whereas for some of the more resistant genotypes these differences were relatively large (SRN39, Tiémarifing and CMDT39). For the most resistant genotype (N13) flowerstalk production per Striga plant increased with infestation level. Furthermore, from these observations it appears that at a low infestation level genotypes primarily differed in flowerstalk dry weight (reproductive potential) of individual Striga plants rather than in the number of Striga plants they support. 82 0.75 B DWstalks (g) DWstalks (g) A 0.5 0.25 0 12 9 6 3 0 0 2 4 6 8 0 20 80 100 21 DWstalks (g) D 14 7 0 32 24 16 8 0 0 7 14 NS max 21 0 15 30 NSmax 45 60 Figure 2. Striga flowerstalk dry weight (DW stalks g) as a function of maximum number of aboveground Striga plants (NS max ) for different years and infestation levels: 2001 (A), 2002 (B) 2003 low infestation (C) and 2003 high infestation (D); for 10 different sorghum genotypes: CK60-B (closed diamonds), CMDT39 (asterixes), E36-1 (closed squares), Framida (open circles), IS9830 (open squares), N-13 (open diamonds), Seredo (closed triangles), Serena (crosses), SRN39 (open triangles) and Tiémarifing (plus signs); dotted lines indicate the variation; all parameters are expressed per -2 83 host plant (equivalent to 0.32 m ). Genotype and seed bank density effects on Striga reproduction DWstalks (g) 60 NSmax NSmax C 40 16 8 -1 (g plant ) DWstalks NSmax -1 DWstalks (g) 2.0 B 24 Chapter 5 1.5 1.0 0.5 0.0 0 0 10 20 30 40 50 60 0 70 10 20 NSmax 30 40 50 60 70 50 60 70 NSmax 0.6 0.8 D NSgen NSmax -1 -1 C DWstalks DWtot 84 32 A 0.4 0.2 0.0 0.6 0.4 0.2 0.0 0 10 20 30 40 NSmax 50 60 70 0 10 20 30 40 NSmax -1 Figure 3. Flowerstalk dry weight (DW stalks g: A), flowerstalk dry weight per aboveground Striga plant (DW stalks NS max g -1 -1 plant : B), flowerstalk dry weight per total Striga dry weight (DW stalks DW tot : C) and fractions of generative plants of total -1 aboveground Striga number (NS gen NS max : D) as a function of number of aboveground Striga plants (NS max ) as observed in 2003 at two infestation levels and at 10 sorghum genotypes: CK60-B (closed diamonds), CMDT39 (asterixes), E36-1 (closed squares), Framida (open circles), IS9830 (open squares), N-13 (open diamonds), Seredo (closed triangles), Serena (crosses), SRN39 (open triangles) and Tiémarifing (plus signs); all parameters are expressed per host plant -2 (equivalent to 0.32 m ). Genotype and seed bank density effects on Striga reproduction The more resistant genotypes facilitated heavier flowerstalks per Striga plant than the more susceptible ones. At high infestation levels, differences in NS max between genotypes were more pronounced than differences in flowerstalk dry weight per Striga plant. Figure 3C shows the relation between aboveground Striga numbers and biomass partitioning to the generative parts (fraction DWstalks over DWtot). This relation shows a non-linear decrease of biomass allocated to the flowerstalks with increasing number of emerged Striga plants. The relation is rather consistent, independent of genotype, indicating that the variation in allocation pattern can merely be attributed to differences in number of emerged plants. The fraction of Striga shoot biomass allocated to the flowerstalks ranged from 0.55 to 0.25. Figure 3D shows the relation between the fraction generative Striga plants (NSgen over NSmax) and the aboveground Striga numbers (NS max), per genotype and infestation level. For most of the genotypes, the generative fraction decreased with increasing infection levels. Exceptions were N13 and IS9830, the most resistant genotypes and Framida, an intermediate genotype. Among the other seven genotypes, the more resistant ones (SRN39, CMDT39 and Tiémarifing) showed a steeper negative slope between the low and the high infestation level than the more susceptible ones (CK60-B, E36-1, Seredo and Serena). The relation between the fraction of generative plants and the maximum number of aboveground Striga plants is comparable with the relation between flowerstalk dry weight per Striga plant and aboveground Striga numbers as shown in Figure 3B. A wide variation in fraction generative plants was observed among genotypes, particularly at low infection levels. Striga dry weights and reproduction Figure 4 shows the relation between total aboveground Striga dry weight and number of capsules (4A) and between Striga flowerstalk dry weight and number of capsules (4B) as observed in 2003. Both total aboveground Striga dry weight and Striga flowerstalk dry weight correlate significantly with capsule number (P<0.01; r= 0.90 and 0.94 respectively). Both relations appear to be linear in the range of observed dry weights in this trial. According to the linear regression, one gram of flowerstalks corresponds to 18.6 capsules. 85 Chapter 5 Seed capsule number A 1000 800 600 400 200 0 0 30 60 90 120 150 Total aboveground Striga dry weight (g) Seed capsule number B 1000 800 600 400 200 0 0 10 20 30 40 50 Total Striga flowerstalk dry weight (g) Figure 4. Relation between total aboveground Striga dry weight (DWtot) and Striga seed capsule number (y=5.59 DWtot; R2=0.80; A) and flowerstalk dry weight (DWstalks) and Striga seed capsule number (y=18.57 DWstalks; R2=0.88; B). Both observations were done in 2003 at crop harvest; all parameters are expressed per host plant (equivalent to 0.32 m-2). Capsule production Mean time span between crop harvest and the end of the Striga life-cycle was 28 days but depended on host genotype (range: 24 [E36-1] to 34 days [IS9830]; data not shown). Table 6 presents the results of the ANOVA of genotype, infestation level and sampling time effects on capsule production per host plant (NRcaps) in 2003. There were significant main effects of genotype, infestation level and sampling time on capsule production (P<0.01) and a significant infestation level × sampling time interaction (P=0.016). 86 Genotype and seed bank density effects on Striga reproduction Table 6. ANOVA of infestation level, sorghum genotype and sampling time effects on Striga seed capsule number per host plant in 2003 Source of variation df MSS F F-prob. (P) Infestation level (I) 1 3274.44 20.66 0.003 Genotype (G) 9 610.80 15.09 <0.001 IxG 9 27.99 0.69 0.715 Time (T) 1 208.25 16.62 <0.001 IxT 1 74.07 5.91 0.016 GxT 9 11.73 0.94 0.496 IxGxT 9 17.77 1.42 0.186 Genotype 9 257.12 6.39 <0.001 Time 1 16.96 1.76 0.189 GxT 9 23.91 2.48 0.016 Genotype 9 381.67 9.38 <0.001 Time 1 265.35 17.22 <0.001 GxT 9 5.59 0.36 0.949 Overall Per infestation level High Low At the high infestation level significantly more Striga seed capsules were produced than at the low level and at the end of the life-cycle significantly higher numbers of seed capsules were produced than at crop harvest. This last phenomenon was only significant at the low infestation level, as became evident from the ANOVA’s that were separately conducted per infestation level. This last analysis also revealed a significant genotype × sampling time effect at the high infestation level. Table 7 presents genotype effects on seed capsule number per infestation level and per sampling time separately. At all sampling times and infestation levels NRcaps of Striga plants parasitizing N13 was significantly lower than on all other genotypes. Capsule production on CK60-B was significantly higher than on any other genotype, throughout sampling times and infestation levels. Apart from these differences there were significant differences between the high capsule production of Striga plants on CMDT39 and E36-1 and the low capsule production on Framida, IS9830 and SRN39 at the low infestation level. 87 Low High harvest a end mean harvest end mean CK60-B 383.3 10 523.0 10 450.6 aa 10 624.0 ab 10 637.1 a 10 630.5 a 10 CMDT39 151.3 8 235.4 9 191.0 b 9 332.2 bc 7 347.3 bc 8 339.5 bc 7 E36-1 156.5 9 201.7 8 178.3 bc 8 329.3 bc 6 288.5 bc 4 308.6 bc 6 Framida 36.5 2 99.1 4 64.0 d 3 272.1 bc 5 297.8 bc 5 284.8 bc 5 IS9830 39.6 3 105.4 5 68.6 d 4 256.5 bc 4 224.8 bcd 2 240.4 bc 4 N13 2.8 1 16.1 1 8.2 e 1 111.4 de 2 77.8 e 1 93.8 d 1 Seredo 92.6 6 97.5 3 95.1 bcd 6 364.7 b 9 372.8 b 9 368.9 b 9 Serena 60.5 5 115.9 6 86.0 cd 5 194.7 cd 3 245.7 bc 3 219.4 bc 3 SRN39 46.1 4 66.4 2 55.8 d 2 110.6 de 1 306.8 bc 6 196.3 cd 2 Tiémarifing 101.9 7 172.9 7 135.0 bcd 7 344.0 bc 8 345.8 bc 7 345.1 b 8 S.E.D. (df) nsd 2.26 (159) 2.5 (159) 2.24 (159) mean 86.2ac 141.0b 277.0 299.5 R 0.48 0.38 0.34 0.39 Values in the same column followed by a different letter are significantly different at the 0.01 probability level. Means in table are back-transformed 1/2 from (X+0.5) transformations, S.E.D. values of transformed data are given b Values of the high infestation level (harvest and end) followed by a different letter are significantly different at the 0.05 probability level c Values in the same row followed by a different letter are significantly different at the 0.001 level of probability d ns means not significant (genotype × time effect) Chapter 5 88 Table 7. Means, ranks (1-10) and repeatabilities (R) of number of Striga seed capsules produced in 2003 (NRcaps) per host genotype and infestation level (Low, High) and time (crop harvest, end life-cycle); capsule number is expressed per host plant (equivalent to 0.32 m-2). Genotype and seed bank density effects on Striga reproduction At the high infestation level, there were significant differences between the high Striga reproduction on Seredo and Tiémarifing and the low reproduction at SRN39. Genotype ranking at the low infestation level differed from the ranking at the high infestation level, except for four genotypes. At the low infestation level the highest capsule production was reached on CK60-B, CMDT39 and E36-1, whereas at the high infestation level this group consisted of CK60-B, Seredo and Tiémarifing. At the other extreme, the lowest Striga capsule production was found on N13, SRN39 and Framida in the low infested plots and on N13, SRN39 and Serena in the high infested plots. The differences in mean capsule numbers between the two infestation levels decreased between crop harvest and the end of the life-cycle. For the lower infestation level, NRcaps at harvest time was around 61% of the final capsule production (the capsule production till the end of the life-cycle), whereas at the higher infestation level, already around 92% of the final capsule number was produced at crop harvest. Repeatabilities (R) of NRcaps were lower than 0.5 in all cases, implying that more than half of the phenotypic variation observed should be attributed to environmental and error sources rather than genetic variation. There was no significant effect of sorghum genotype (P=0.592), or infestation level (P=0.324) on seed weight per capsule (data not shown). The grand mean of seed weight per capsule was 2.41 10 -3 g (S.E. = 0.053 10-3). As mean individual Striga seed weight was determined to be 4.5 10 -6 g, one capsule contained on average 536 seeds. Discussion Methodology for estimating Striga reproduction In the field, Striga seeds are formed and disseminated over a period of several weeks. Quantification of the total seed production of a single Striga plant or a population of plants at a single point in time is therefore less accurate. In 2001 and 2002, reproductive success was estimated based on DWstalks measured at sorghum harvest. As a result, plants that had decomposed (due to early death) before that time were not included and for that reason this estimate most likely represents an underestimate of the actual reproductive success. This also explains the relatively low flowerstalk dry weight per emerged Striga plant in these years, ranging from 0.04 to 0.53 g. In 2003, when the procedure was extended and dead generative Striga plants were collected weekly, this ratio ranged from 0.33 to 1.85 g per emerged plant. Analysis on these ratios between flowerstalk dry weights derived exclusively from plants at harvest (in 2003) resulted in fractions ranging from 0.07 to 0.57 g per emerged plant. These ratios are very much in line with those found in 2001 and 2002. The same explanation could 89 Chapter 5 be given for the observed lower ratios of NSgen to NSmax 2001 and 2002 (0.29 and 0.35, respectively) compared to 2003 (low [L]: 0.54; high [H]: 0.49). In 2003, not only the dry weight of the flowerstalks was determined, but also the number of seed capsules was counted. Capsule numbers were earlier used as estimate for seed production by Weber et al. (1995) and Webb and Smith (1996). In 2003, ten mature capsules from 5 different Striga plants were sampled in each subplot. No infestation density or host plant genotype effect on seed weight per capsule was found, indicating that capsule number is a reliable estimate for seed production. Counting of capsules is however a time consuming and therefore expensive method. This study showed that a very high correlation exist between Striga seed capsule number and total aboveground Striga dry weight or, even more accurate, flowerstalk dry weight (Figure 4). This confirms the broad applicability of relations between seed production and plant biomass as found by Samson and Werk (1986) and Thompson et al. (1991). The linearity of this relation and its independency from genotype makes total aboveground or flowerstalk dry weight a reliable measure for Striga seed production. A major constraint for the identification of genotypes that facilitate less Striga reproduction seems to be the high contribution of the environmental variance component to the phenotypic variation. The repeatability of capsule number was found to be low (R=0.34 to 0.48). Striga reproduction is an indirect genotypic trait, as it can only be measured at the Striga plants parasitizing the specific genotype. Therefore the phenotypic expression of the trait is subject to various factors. Deviations in rankings between years are most probably not due to differences in infestation level between the different years as in 2003 no significant genotype × infestation level interaction was found for either capsule production or flowerstalk dry weight. These deviations must therefore be caused by other environmental factors. Although the general genotype categorization in the various experiments was not affected, multiple location and multiple year testing for screening of genotypes that support low Striga reproduction is recommended. Host plant genotype effects on Striga seed production The results clearly show a strong significant effect of host plant genotype on Striga number, biomass and flowerstalk dry weight. Rankings for the various characteristics within an experiment were largely identical, with N13 as the variety with the lowest number and weight of Striga plants and CK60-B with the highest plant number and weight. Resistance thus proved to be important in reducing Striga reproduction. Several authors (Doggett, 1988; Hess and Haussmann, 1999; Ejeta et al., 2000) already suggested an effect of host plant resistance on Striga reproduction. Both capsule 90 Genotype and seed bank density effects on Striga reproduction number in 2003 and flowerstalks dry weight in 2001 and 2003 were significantly affected by genotype. Absence of this effect in 2002 (P=0.086) may be due to the overall high infection levels, levelling out differences between genotypes due to density effects. Total aboveground dry weight, also a good indicator for capsule production, was significantly affected by genotype in all years and at all infestation levels. Variability in the number of aboveground Striga plants (both maximum and generative numbers) is the main reason for differences in Striga reproduction between host plant genotypes. This is shown by the highly significant correlations between NSmax (and NS gen) and DWstalks in all years as well as between NSmax (and NSgen) and NRcaps in 2003. These observations confirm suggestions from Carsky et al. (1996), Kling et al. (2000) and Haussmann et al. (2001b). The majority of the genotypes supporting low Striga reproduction in this study (e.g. IS9830, SRN39 and N13) also showed low infection levels. These genotypes were classified as resistant in earlier studies (Maiti et al., 1984; Vasudeva Rao, 1984; Ramaiah, 1984, 1987b; Olivier et al., 1991; Ejeta et al., 2000; Heller and Wegmann, 2000; Omanya et al., 2004; Rodenburg et al., 2005). Additionally, genotypes with very high Striga numbers (CK60-B and E36-1) supported high Striga reproduction. High Striga numbers result in high aboveground Striga biomass and consequently high capsule production. Comparison between genotypes shows that similar infection levels do not necessarily result in equal reproduction, whereas equal reproduction may be obtained as a result of different infection levels (Figures 2A-D and Figure 3A). Hence resistance is not the only genotypic factor determining Striga reproduction. From the results of this study it appeared that the differences in Striga reproduction between genotypes were also related to single Striga plant dry weight (data not shown), the fraction of aboveground plants that reached the generative stage and flowerstalk dry weight per emerged Striga plant. These last two parameters showed largely identical trends, indicating that one of the main reasons for a low Striga flowerstalk production per emerged plant was the fact that fewer plants reached the reproductive phase. The aforementioned traits seem not to be linked to resistance, as they all showed variation among genotypes that could not be explained by differences in infection level. Genotypes with low flowerstalk dry weight per emerged Striga plant or a low fraction generative plants were found among the susceptible genotypes (E36-1, Serena and Seredo), but also among those classified as moderately resistant (Framida and CMDT39). Time, particularly first emergence time of Striga relative to the crop, has been suggested to have a major effect on Striga reproduction ( Weber et al., 1995; Kim and Adetimirin, 1997a), as it determines the total time for development and seed setting. 91 Chapter 5 Emergence date depended on seed bank density, as shown by the significant later emergence in 2003L compared to 2003H, as well as on genotype, as shown in the experiments of 2002 and 2003L. Genotype dependency of emergence date was previously reported by (Doggett, 1988; Olivier et al., 1991a; Webb and Smith, 1996; Adetimirin et al., 2000c; Ast et al., 2000; Haussmann et al., 2001b). In general, late emergence was observed on genotypes with a high level of resistance (e.g. SRN39, IS9830, N13 and Framida). The significant negative correlations between emergence date and aboveground Striga numbers (r = -0.25 with P<0.05 in 2001; -0.27, -0.51 and -0.24 in 2002, 2003L and 2003H respectively, with P<0.01) confirm the entanglement between Striga development and resistance that was earlier found by Ast et al (2000). Despite the significant negative correlations between emergence or flowering dates and flowerstalk dry weights in 2002 and 2003 (r = -0.42 [P<0.01], -0.47 [P<0.001] and –0.34 [P<0.01] for Edate-DWstalks and –0.33 [P<0.05], -0.67 [P<0.001] and –0.29 [P<0.01] for Fdate-DWstalks, for 2002, 2003L and 2003H respectively) the variation in emergence and flowering time among genotypes could not always explain observed variation in Striga reproduction. In 2003, among the genotypes supporting low Striga reproduction only Serena and N13 showed a significantly delayed Striga development compared to genotypes supporting high Striga reproduction such as CK60-B and E36-1. In 2002, Striga flowering of N13 was not significantly later than on CK60-B, while Striga flowerstalk dry weight of N13 was still much lower than that of CK60-B. In 2003H, Seredo supported rather high seed production while Striga flowered relatively late whereas Striga on SRN39 combined low reproduction with early development dates. As there are many inconsistencies in genotype rankings based on development dates throughout the years it may be concluded that differences in development rate of Striga on different host plants may be only one, among several factors, determining Striga reproduction. Density effects on seed production The experiment in 2003, conducted with two Striga infestation levels, revealed clear evidence of the existence of a density dependent relation between Striga number and Striga reproduction. A five times higher infestation level in the high density plots, resulted for most of the genotypes in a five times higher number of emerged Striga plants. At the same time, this increase in number of emerged Striga plants only resulted in a 2.5 to 3.5 times increase in flowerstalk dry weight and capsule number. It was observed that at the highest infestation level average Striga plant dry weight remained smaller, indicating the presence of intra-specific competition. On top of that, a smaller fraction of dry matter was allocated to the reproductive organs. This last aspect can partly be attributed to the smaller fraction of generative plants at the highest 92 Genotype and seed bank density effects on Striga reproduction infestation level. In this study, density dependence thus mainly resulted from intraspecific competition of emerged Striga plants. Density dependence was previously reported by Smith and Webb (1996), but they observed this phenomenon between seed bank density and Striga emergence. This type of density dependence was only observed with the most susceptible genotypes CK60-B and E36-1, for which only a three-fold increase in number of emerged Striga plants was observed between 2003L and 2003H. It can thus be concluded that density dependence is able to manifest itself in more than one stage. Secondly, the expression of density dependence showed to be genotype specific, with the more susceptible genotypes showing this phenomenon in an earlier development stage of the parasite. In many of the observed density dependency relations, N13 formed an exception on the rule. Instead of the generally observed decrease, this genotype showed an increase of Striga plant weights, generative fractions and flowerstalk dry weights per plant, with increasing aboveground Striga numbers. This may be caused by the extremely low numbers and the entanglement of these low numbers with time of emergence and generative development (e.g. mean first emergence at N13: 50.3 DAS compared to overall mean 41.7 DAS and mean first flowering at N13: 96.7 DAS compared to overall mean of 89.8 DAS). Consequently, the relation between individual Striga plant dry weight and infestation level is not a continuously descending curve; rather it is an optimum curve. At below optimum infestation levels few Striga plants emerge and these plants emerge relatively late and are not able to make up for the lost number of growing days. At infestation levels above the optimum, too many individuals emerge and in that situation intra-specific competition causes individual Striga plant dry weight to remain below the maximum. Striga seed production after crop harvest Striga seed production continued beyond crop harvest. The magnitude of the additional seed production after harvest depended on seed bank density. Continued Striga reproduction beyond harvest contributed significantly (39%) to the final reproduction under low infestation, whereas under high infestation the final capsule production was already almost reached at harvest (only 8% was produced after harvest). As a result, differences in capsule production between low and high-infested plots decreased between harvest and the end of the Striga life-cycle. The difference in Striga development between low and high infestation levels is likely to be one of the major reasons for the differences in seed production pattern. In the low infestation plots, first emergence of the parasite and first flowering were on average delayed with 9.5 days. The additional time after crop harvest offers the opportunity for late emerged Striga plants to complete their life cycle and this is particularly relevant for the low 93 Chapter 5 infestation plots. Hence manifestation of the density effects on Striga reproduction increases with time, reducing the initial differences between infestation levels. Host plant genotype effects on Striga seed bank In this study, estimated Striga seed number per generative plant averaged per genotype and infestation level ranged between 6,700 and 26,500. Maximum estimated seed number per generative plant found in a single observation plot was 72,000. These estimates are higher than those of Webb and Smith (1996) who estimated a range of 5,000 to 11,000, but within the range of Stewart (1990) who recorded up to 85,000 seeds per plant. Increases in soil seed banks of 270% (Murdoch and Kunjo, 2003) and 340% (Delft et al., 1997) through newly produced Striga seeds within one cropping season have been reported. In this study the average (averaged over genotypes) estimated seed production per m2 was 334 % (2003H) to 785 % (2003L) of the seed bank density at the onset of the experiments. Adetimirin et al. (2000c) and Haussmann et al. (2000b), suggested that reduction of the Striga seed bank of infested soils can be accomplished through the use of Striga resistant cultivars, as these lower the seed input through a reduced seed production rate. From this study, it appeared that resistant genotypes, those supporting fewer attachments or aboveground Striga infections, indeed affect Striga reproduction negatively. However, after multiplication of the capsule production per host plant with the estimated mean number of seeds per capsule (536 seeds capsule-1) and the sorghum plant density (3.125 plants m-2 in 2003), it appeared that for an initial seed bank of 30,000 seeds m-2 (as in 2003L) a production of 18 capsules per host plant would be sufficient to completely replace the original seed bank. For an initial seed-bank of 150,000 seeds m-2 (as in 2003H) a production of 90 capsules per host plant would be enough. In fact these values represent an upper limit to the required capsule production per host plant. Only in the situation of a complete depletion of the soil seed bank during the cropping season, these reproduction rates are required to just balance the seed bank. Rather than a complete depletion, Delft et al (1997) observed depletion rates of 50-70% depending on soil depth, whereas Murdoch and Kunjo (2003) found depletions rates of 46%. Based on the most conservative depletion rate (46%) , a host plant genotype in the 2003L plots of this study should not support Striga to produce more than eight (46% of 18) capsules and in the 2003H plot not more than 41 (46 % of 90) capsules to just balance the seed bank. This confirms earlier statements by Delft et al. (1997) that only a few Striga plants per m2 would be enough to balance or increase the seed bank. Out of the ten genotypes used in this study, only the capsule production rate of N13 in the low infestation plots at harvest did not surpass the estimated upper limit. From these results and calculations it is 94 Genotype and seed bank density effects on Striga reproduction concluded that only the use of extremely resistant varieties on fields that still have a very low seed bank density may prevent an increase of the seed bank density over time. At the same time, the results indicate that even with N13 not removing the Striga plants at crop harvest time would already be sufficient to initiate a further increase in Striga density. The calculations thus show that Striga problems easily increase and just the use of resistant genotypes is not enough to reduce the seed bank once the initial seed bank already surpassed a certain level. 95 CHAPTER 6 General discussion Orientation of the study Striga hermonthica is an important and persistent problem in cereal production of the semi-arid tropics. It is important as a problem because of its wide dispersion throughout this climatic zone, and because it seriously lowers crop yields thereby threatening the livelihood of many subsistence farmers. It is persistent because it seasonally produces thousands of tiny but viable and long-lived seeds per plant that cause a gradual increase of infection levels if suitable host plants remain to be grown in the same field. Hence, control options against Striga should reduce yield losses and at the same time minimize or prevent future infestations. This may be achieved through proper host plant defence mechanisms such as resistance and tolerance. Resistance reduces or prevents infection and reproduction (Shew and Shew, 1994), while tolerance lowers or prevents yield losses or damage due to infection (Caldwell et al., 1958). The use of crop varieties with improved resistance and tolerance against Striga is believed to be one of the most useful control options against this parasitic weed. For smallholder farmers the advantage of improved varieties compared to other options is that these varieties do not require extra work or additional inputs (such as pesticides) that can be costly or have undesired side-effects on environment and health (e.g. Hess and Haussmann, 1999). Ever since the early work on plant resistance by Williams (1959), breeders and Striga researchers have been working on breeding for resistance and tolerance against Striga (e.g. Ramaiah, 1987a; Hess and Ejeta, 1992; Olivier et al., 1992; Efron, 1993; Kim, 1994; Kim et al., 1998; Adetimirin et al., 2000c). Success of these breeding efforts depends largely on the availability of practical and cost efficient screening techniques that make use of selection measures. that are precise, reliable and discriminative (Hess and Haussmann, 1999). Host resistance and tolerance are intensively studied as control options against biotic stresses (e.g. Orton, 1909; Schafer, 1971; Clarke, 1986; Trudgill, 1991). Resistance and tolerance against Striga have been studied since the late 50’s up to date (e.g. Williams, 1959; Obilana, 1984; Haussmann et al., 2000a; Wilson et al., 2000; Gurney et al., 2002a). Work on host plant defence against Striga can roughly be classified as studies on mechanisms behind resistance or tolerance (e.g. El Hiweris, 1987; Vogler et al., 1996; Arnaud et al., 1999; Gurney et al., 2002a) and studies on selection measures and methods such as the design of screening trials and Striga 97 Chapter 6 infestation techniques (e.g. Cubero et al., 1994; Adetimirin et al., 2000b; Haussmann et al., 2000b; Omanya et al., 2004). However, studies in which the two aspects are studied in an integrated way are rare. The present study combined both aspects, as the objective was to enhance efficiency of selection for tolerance and resistance against Striga through better understanding and proper characterization of these mechanisms. Resistance Resistance against Striga has been widely studied and good measures for this defence mechanism are available. Resistance can be quantified based on the number of aboveground Striga plants at a specific moment in time (Adetimirin et al., 2000b; Omanya et al., 2004), the maximum number of Striga plants that emerged aboveground (Wilson et al., 2000), or the integration of regular Striga counts (Haussmann et al., 2000b). All these measures are based on aboveground Striga plants. For practical reasons aboveground measures are preferred over belowground measures. Likewise, selection measures that can be used in the field are preferred over those that require laboratories or pot experiments, as this environment resembles most the actual environment of a farmers’ field. Furthermore some studies found inconsistencies between results from the field and those from other test environments (Haussmann et al., 2000b; Omanya et al., 2000). Results of the current study support the suggestion that aboveground measures are adequate because they were found to reflect the number of belowground attachments reasonably well. Haussmann et al. (2000b) proposed the use of the Area under the Striga Number Progress Curve (ASNPC) as a suitable resistance measure. It provides a fair representation of the infection throughout the cropping season. In this study, the maximum aboveground Striga number (NS max) was found to be as good as the ASNPC. It was objective, as it takes into account the genotypic differences in host development, and proved discriminative, very consistent over years and easy in use. Its advantage over the ASNPC is that it requires less time and labour, because counting can start later (just before the maximum number is expected) and stop earlier (as soon as numbers for all genotypes start to decrease). Other measures that have been proposed for screening for resistance were Striga vigour (Haussmann et al., 2000a,b) and aboveground Striga dry weights (Kulkarni and Shinde, 1985). However, numbers are considered better for a variety of reasons. Compared to vigour scores, they are more objective and therefore less dependent on the person performing the screening. Numbers are also thought to be a more unambiguous resistance measure because they have less interaction with 98 General discussion tolerance. For a Striga plant to be vigorous or to gain dry weight, it needs a sufficiently tolerant host plant that is able to provide it with the required carbohydrates. Furthermore, contrary to dry weights, counts do not require drying and weighing facilities. This study showed that selection for resistance based on numbers implies a risk, if infestation levels are high. Relatively low Striga infection levels, compared to those on other genotypes, can be the result of a lower maximum attainable infection level. This may be due to a limited number of establishment sites on the host root (a resistance mechanism), or a limitation of the carrying capacity of the host (caused by sensitivity of the host plant) and subsequent intra-specific competition among Striga plants. Due to this limitation of the carrying capacity the infestation – infection curve shows an optimum, beyond which infection numbers start to decrease. The current study revealed two examples of genotypes showing such an optimum infection curve with increasing infestation (CK60-B and Tiémarifing). This typical interaction between susceptibility, sensitivity and intra-specific competition was earlier reported by Kim et al. (1998) and Haussmann et al. (2000b). They observed that susceptible or sensitive genotypes often support fewer emerged Striga due to reduced host vigour and underground competition. Hence the number of infections is not only a result from resistance but may be the outcome of a complex interaction between resistance, tolerance and intra-specific competition. To avoid this entanglement with tolerance it was recommended to conduct resistance screening at low infestation levels. This study showed that aboveground Striga numbers obtained at low infestation levels, representing the initial slope of the infestation - infection curve, is the most discriminative screening measure for resistance (Chapter 3). Lower initial infection response to infestation may be the result of resistance mechanisms discovered in earlier studies, such as a low germination stimulant production (Hess et al., 1992; Olivier and Leroux, 1992), germination inhibitors (Weerasuriya et al., 1993), mechanical resistance hampering penetration of host root cells (Maiti et al., 1984; Olivier et al., 1991b; Ejeta and Butler, 1993) or reduced metabolite flow from the host to the parasite in the early belowground stages (Arnaud et al., 1999). Complete resistance is defined as resistance that prevents growth and reproduction of the pathogen (Shew and Shew, 1994). Resistance as found against Striga is not complete, as it still allows some Striga plants to parasitize and hence reproduce. It has been hypothesized that despite the absence of complete resistance against Striga, resistance can lower the Striga seed production (Doggett, 1988; Ejeta et al., 2000). Resistance against Striga indeed proved an important factor to lower seed production. However, large variation in Striga reproductive efforts among equally 99 Chapter 6 resistant genotypes exist (Chapter 5). While Striga numbers were more relevant than dry weight for the assessment of resistance, Striga dry weight may be very useful to screen for genotypes supporting less Striga reproduction (Chapter 5). Seed production correlates very well with plant biomass (Samson and Werk, 1986; Thompson et al., 1991). This study showed that this is also true for Striga seed production and Striga dry weight. These measures can either be determined around crop harvest or at the end of the Striga life cycle. However if growth duration of genotypes under consideration varies a lot, an assessment at the end of the life cycle of Striga seems to result in a more fair comparison. In that way, at all genotypes Striga plants on all genotypes will have about the same development time, independent of the development time of the host genotype. Carsky et al. (1996) assumed that host plant varieties could also reduce Striga reproductive efforts through early host maturity. However, Striga continues to reproduce after removal of the aboveground parts of the host plant (Chapter 5) and there were no significant correlations between crop harvest time and seed production. The effect of early host maturity on Striga reproduction seems to be overruled by other effects such as host susceptibility. Main constraint for post-harvest Striga reproduction seems to be the water availability to the Striga, determined by the water retention of the soil and the rooting density and depth of the host. Magnitude of the post-harvest reproduction depends on the initial seed bank density. Due to delayed development in plots with low initial seed bank densities, Striga reproduction after harvest makes up a very significant part of the total reproduction. As a result of this compensational time the difference between seed production under low and high initial seed bank densities reduces towards the end of the Striga life cycle (Chapter 5). In the field, density dependent reduction in seed production mainly resulted from intra-specific competition between aboveground Striga plants. In pot experiments with much higher infestation levels than in the field, the relation between Striga infestation and Striga infection was also found to be density dependent, confirming the earlier conclusion of Smith and Webb (1996). However, in the field experiments this was only evident for the most susceptible genotypes, CK60-B and E36-1 (Chapter 5). Hence the density dependency between infestation and infection only becomes apparent when infection levels can reach sufficiently high levels. 100 General discussion Yield loss and tolerance Average estimated sorghum yield losses of 50% and more have been reported in Striga infested fields (Last, 1960; Bebawi and Farah, 1981). In this study (Chapter 2) relative yield loss estimates from the field ranged from 0 to 85 %. In the low infested fields mean relative yield loss over all genotypes was around 25% and in the high infested fields mean relative yield loss was just below 50%. Dogget (1965) estimated 2-3 kg yield loss ha-1 1000 Striga plants-1. In the field experiments of this study this estimate ranged from 0 to 10 kg yield loss ha-1 1000 Striga plants-1 in the low infested fields with estimated aboveground Striga densities ranging from 5,000 to 500,000 plants ha-1. In the high infested fields, with estimated aboveground Striga densities of 80,000 to 2,900.000 plants ha-1 the estimated yield loss ranged from 0 to 3 kg ha-1 1000 Striga plants-1. This lower average yield loss per 1000 Striga plants at higher Striga densities demonstrates that yield loss is not proportional to Striga density. Compared to resistance against biotic stresses or tolerance against abiotic stresses relatively few studies have focussed on tolerance against biotic stresses. This may have various reasons. Tolerance still allows the pathogen or the parasite to attack, develop and reproduce. Therefore Striga tolerance in itself is not considered a desirable defence mechanism, as it would cause a build-up of the seed bank and hence an aggravation of the problem in future cropping seasons. However, tolerance as an additional defence mechanism in a resistant host genotypes may be very useful (Ramaiah and Parker, 1982; Haussmann et al., 2000b, 2001a). Tolerance against Striga is especially required since to date no immunity against this parasite has been found. Hence parasites may still be able to attack crops and cause severe damage once resistance is combined with sensitivity. Another reason for the lower attention for tolerance against biotic stresses may be its entanglement with resistance. In plants, resistance and tolerance co-exist and may be represented at different degrees (Barker, 1993). In the separate quantification of both resistance and tolerance, this entanglement plays a complicating role. In general, selection for tolerance against abiotic stresses (such as salinity, heat or cold) is less complicated, as all genotypes in a selection trial can easily be exposed to the same stress level. In contrast, for biotic stresses, exposing host plants to equal disease or parasite pressure, a prerequisite for screening for tolerance according to Schafer (1971), proves difficult due to differences in levels of resistance against the disease or parasite (Chapter 3; Clarke, 1986). This study showed that it is extremely difficult to get overlapping ranges of infection levels if the genotypes differ widely in resistance level. An extremely wide range in infestation levels would be needed for such an overlap, which is too laborious and not practical for use in the field (Chapter 3). Another theoretical solution would be 101 Chapter 6 to rely on interpolation techniques to correct the tolerance measure (e.g. relative yield loss) for differences in Striga infection levels, found among genotypes. This requires the relation between infection and relative yield loss to be known and preferably also to be the same for all genotypes. In the pot experiments of this study, despite the wide range of infestation levels, for none of the genotypes the complete relation between infection and relative yield loss could be resolved. To explore relations between infestation, infection and yield loss in an integrated way, the three-quadrant figure, modified from the one introduced by de Wit (1953) for the analysis of fertilizer fate, was used. It shows the relation between Striga infestation and Striga infection (quadrant IV), Striga infestation and (relative) yield loss (quadrant II) and Striga infection and yield loss (quadrant I). Some of the abovementioned problems related to tolerance can be illustrated by this representation. B RYL I II II NSmax Infestation Infestation IV RYL I Infestation NSmax Infestation A IV Figure 1 Three-quadrant representation of three relations: (1) between Striga infestation and Striga infection (NSmax: maximum aboveground Striga numbers; quadrant IV), (2) between Striga infection (NSmax) and relative yield loss (RYL; quadrant I) and (3) between Striga infestation and relative yield loss (quadrant II). Figure 1A shows these relations for a fictive resistant (dotted line) and susceptible genotype (solid line) with equal tolerance levels. Figure 1B shows genotypes with equal resistance that differ in sensitivity to Striga infection; genotypes are either sensitive (solid line) or contain tolerance resulting in a lower maximum yield loss level (dotted line) or a lower yield loss per Striga plant (discontinuous line). 102 General discussion In Figure 1A, two fictive genotypes with equal tolerance levels (same relation NSmax and RYL) but different resistant levels are shown. Quadrant IV (Infestation level – NSmax) shows that a resistant genotype (dotted line) has very limited overlapping infection levels with a susceptible genotype (solid line), making a direct comparison for tolerance between those two genotypes extremely difficult. Due to the low infection levels, relative yield loss of the resistant genotype will not surpass a certain level, and is basically determined by the maximum number of attachments, whereas yield loss of the susceptible genotype may easily reach 100%. For the resistant genotype no information can be acquired on the maximum yield loss at high infection levels, whereas for the susceptible genotype, it proved very difficult to retain information on the initial slope of the relation between NSmax and RYL (Chapter 3). At the same time, it is not known whether the mechanism responsible for a lower initial slope is identical to the mechanism that is responsible for a reduced maximum yield loss. Tolerance could thus be based on a single mechanism, or on two separate mechanisms of which one causes a lower yield loss per Striga (initial slope), and the other a lower maximum attainable yield loss. Figure 1B shows both tolerance mechanisms. The figure represents genotypes with equal resistance but different tolerance levels. The solid line represents a sensitive genotype, characterized by a steep initial increase and a high (100%) maximum yield loss. For the other lines either the initial slope is less steep (discontinuous line) or the maximum yield loss is less high (dotted line). The intersections of the different curves show that it depends on the actual infection level which of these tolerance mechanisms is most efficient and consequently which of these genotypes would be selected in a screening trial. Obviously, a combination of a reduced initial slope and a reduced maximum yield loss level would yield the best tolerance. In case both expressions of tolerance are based on one and the same mechanism, such a combination would automatically be obtained through selection. It was concluded that for resistant genotypes, tolerance can be quantified as a reduced relative yield loss per aboveground Striga plant and for susceptible genotypes the maximum relative yield loss can be used (see 1A). One approach could be to make two separate groups of genotypes, one resistant and one susceptible group. This preselection could than be followed by two separate screenings for tolerance. For the group with resistant genotypes, the initial slope of the infection level-relative yield loss can be used as selection measure for tolerance, while for the susceptible genotypes one could use the maximum relative yield loss. Selected material from both groups could then be used in breeding programs, either in combination or as separate traits, depending on the objective and ease and costs of gene-localisation and transfer. Yield based measures have at least one main obstacle. Due to the large 103 Chapter 6 genotypic variation in uninfected crop yields, resulting from differences in yielding ability or differences in adaptability to the local screening environment, yield under Striga-infested conditions can not simply be used as screening measure. The presence of controls therefore remains an indispensable requirement when using a measure of tolerance based on grain yield. Consequently, alternative selection measures for tolerance, without this requirement, are desired. This study investigated the use of alternative measures that were based on possible mechanisms of tolerance. Striga infection lowers host plant photosynthesis, and this reduction was also shown to be reflected in chlorophyll fluorescence measurements. Tolerant genotypes appeared less affected than sensitive ones. Differences in Striga effects on chlorophyll fluorescence among genotypes were earlier found by Gurney et al. (2002a). Chlorophyll fluorescence measurements proved to be suitable as screening measure for abiotic stresses, such as drought or cold (e.g. Havaux and Lannoye, 1985; Schapendonk et al., 1989a; Nogues et al., 1994; Fracheboud et al., 1999; Olsovska et al., 2000). The current study strongly suggests that chlorophyll fluorescence measurements can also be used as an alternative screening measure for tolerance to Striga. Based on the results of this study the most suitable measurement for selection for Striga tolerance seemed to be photochemical quenching (Pq) and electron transport (ETR). Comparative advantage of these measurements, over yield-based selection measures, is the possibility to screen in the absence of control plots, as values of Pq an ETR of unstressed plants at comparable phenological stages showed to be rather stable among genotypes (Chapter 4). Whether the infestation level for tolerance screening should be high or low depends on the genotypes in the selection trials and the screening objective. For screening based on yield response as well as for screening based on a physiological response the differences between genotypes are most discriminative at high infestation levels (Chapter 3 and 4). This would confirm the earlier recommendations of Kim (1991). However, in view of the general objective of developing highly resistant varieties with high levels of tolerance, selection of tolerant parental lines may best be carried out at low levels of Striga infection. Since superior resistant genotypes guarantee low levels of infection, the desired tolerance is the one that is able to prevent yield loss at low infection levels. Hence, if the aim is to use tolerance as an additional defence to superior resistance, its selection measure could well be the initial relative yield loss per Striga infection. The maximum relative damage level (e.g. maximum relative yield loss) would be less useful as tolerance measure, simply because infection levels of superior resistant genotypes will never become high enough to attain the maximum relative yield loss. Furthermore, in the initial trajectories of the relation between infection and relative yield loss or chlorophyll fluorescence reduction, one 104 General discussion can assume a nearly linear slope as was observed with Framida. This assumption would facilitate the calculation of losses or reductions per Striga infection, which can be used as a tolerance measure. If variation in resistance among genotypes is not considerable, one infestation level would probably be enough. Implications for breeding Host plant resistance or tolerance is often believed to be one of the most promising solutions to Striga (e.g. Kim, 1996; Gurney et al., 1999). However, this study showed that, relative yield losses under field situations could still reach more than 22% in very tolerant and resistant genotypes (Chapter 2). Also, Striga may still be able to reproduce at very resistant genotypes and thereby completely replenish or even increase the seed bank (Chapter 5). Hence even with the use of improved varieties problems related to Striga persist and may even increase. It is therefore recommended that farmers combine the use of varieties with improved resistance and tolerance with agronomic Striga control options such as hand weeding before flowering, the use of trap crops in crop rotations or intercropping schemes (e.g. Hess and Dodo, 2004) and possibly biological control options such as the use of AM fungi as proposed by Lendzemo and Kuyper (2001) and Lendzemo et al. (2005). Such an integrated Striga management approach is probably the best solution to the problem. It means however that each component of this integrated control should be optimised. Hence, breeding efforts should be further enhanced to develop better varieties that prevent or highly reduce Striga infection and reproduction and yield well either in the absence and the presence of Striga. The idea of this study was that the separation of resistance and tolerance in definitions, selection measures and methods could result in the selection of separate breeding material with either superior levels of resistance or superior levels of tolerance. Through subsequent crossings of these selections new varieties could be developed that combine superior levels of each mechanism. The combination of disease resistance and tolerance was earlier suggested by Stakman and Christensen (1960), and later for Striga by Ramaiah and Parker (1982), Kim (1991), Haussmann et al. (2001a,b), Pierce et al. (2003) and many others. The conventional selection and breeding approach, as proposed in this study with improved methods and measures may be enhanced by combining it with other improved or more advanced methods or breeding strategies. Possibilities are to include wild relatives with natural resistance or tolerance in the breeding programme (Burdon, 1997; Wilson et al., 2000; Gurney et al., 2002b; Wilson et al., 2004), to pyramid genes 105 Chapter 6 for different resistance mechanisms in order to obtain more stable, polygenic resistance, or to transfer resistance or tolerance genes into well-adapted genotypes. The role of marker-assisted selection in these breeding strategies against Striga is very promising as it is a powerful tool for the incorporation of genes from selected genotypes or wild relatives into improved varieties (Tanksley et al., 1989). These additional techniques can, however, never completely replace the classical breeding (Hess and Haussmann, 1999). One of the remaining constraints to breeding for defence against Striga is the large error variation and consequently rather low repeatabilities of the selection measures (this study). It is therefore necessary to use enough replications within each experiment and to replicate experiments over years. Haussmann et al. (2000b) recommended the use of at least 4 replications in screening trials. Also the high genetic variation of Striga should be taken into consideration in breeding programmes (Verkleij and Pieterse, 1994), as well as the need to exclude effects caused by typical genotype × environment interactions. Another problem is that most of the resistant and tolerant genotypes are poorly adapted and have low yields at locations outside their region of origin (Hess and Haussmann, 1999). For these reasons screening should be done at multiple sites, in order to test the broad adaptability of the genotype and stable performance of the defence (Ramaiah, 1987a). Selection for genotypes that prevent or reduce Striga infection should be done in the field at low Striga infestation levels (around 25,000 viable Striga seeds m -2) and can be based on maximum aboveground Striga numbers. It requires the breeder to start frequent Striga counts around 70 days after sowing. Additionally, for an adequate estimation of the reproductive effort of Striga on a genotype, total aboveground Striga dry weights or Striga flowerstalk dry weights should be assessed. Preferably the breeder regularly collects dead Striga plants from a known surface, including a number of host plants, and continues this collection until all Striga plants are collected (the end of the Striga life cycle). If time is sparse, and genotypes do not differ much in season length the breeder could decide to do a single sampling of all aboveground Striga material (dead or alive) at crop harvest. Slightly more work but also more accurate would be to collect dead Striga plants in two or more rounds (e.g. around crop harvest and subsequently at the end when all Striga plants are dead). Selection for genotypes that endure Striga infection and maintain high crop yields should be done in the field. If this selection is based on yield response to Striga infestation, incorporation of Striga-free control plots is required. Alternatively a selection that does not require control plots can be carried out with a measurement system that is able to quantify photochemical quenching or electron transport rate through photosystem II. Considering that the objective is to find genotypes with 106 General discussion tolerance that will eventually be combined with superior resistance, ensuring low infection levels, the infestation levels for tolerance screening can be kept relatively low. If the objective is to find superior tolerance that needs to stand on its own, high infestation levels (at least 300,000 viable seeds m-2) are recommended, while the best bet might be to screen at two (low and high) infestation levels or with two groups of genotypes (susceptible and resistant). The latter solutions enable identification of possible different types of tolerance, one with expression at low infection and one with expression at high infection levels that could than both be incorporated in a superior resistant or adapted genotype. Finally, for varieties to be acceptable to farmers, they have to meet a whole range of other criteria than yielding ability, disease resistance and agro-climatic stability alone (Defoer et al., 1997). Criteria such as grain colour, taste and cooking qualities proved to be both region and gender specific. Understanding these criteria, early incorporation of these criteria in the breeding program, and tailoring varieties to local needs could enhance adoption of improved varieties by farmers (Kamara et al., 1996; Bengaly and Defoer, 1997; Defoer et al., 1997). 107 References Ackroyd, R. D. and J. D. Graves, 1997. The regulation of the water potential gradient in the host and parasite relationship between Sorghum bicolor and Striga hermonthica. Annals of Botany 80, 649-656. Adetimirin, V. O., M. E. Aken'Ova and S. K. Kim, 2000a. Effects of Striga hermonthica on yield components in maize. Journal of Agricultural Science 135, 185-191. Adetimirin, V. O., S. K. Kim and M. E. Aken'Ova, 2000b. An alternative method of screening maize for tolerance to Striga. African Crop Science Journal 8, 171-178. Adetimirin, V. O., S. K. Kim and M. E. Aken'Ova, 2000c. Expression of mature plant resistance to Striga hermonthica in maize. Euphytica 115, 149-158. Andrews, F. W., 1945. The parasitism of Striga hermonthica Benth. on Sorghum spp. under irrigation. I. Preliminary results and the effect of heavy and light irrigation on Striga attack. Annals of Applied Biology 32, 193-200. Arnaud, M. C., S. Renaudin, A. Villaine and P. Thalouarn, 1996. Why is Sorghum bicolor var Framida resistant to Striga hermonthica? A histological and molecular approach. In: M. T. Moreno, J. I. Cubero, D. Berner, D. Joel, L. J. Musselman and C. Parker (eds), Advances in parasitic plant research. Sixth International Parasitic Weed Symposium., Cordoba, Spain, pp. 574-580. Arnaud, M. C., C. Véronési and P. Thalouarn, 1999. Physiology and histology of resistance to Striga hermonthica in Sorghum bicolor var. Framida. Australian Journal of Plant Physiology 26, 63-70. Ast, A. v., L. Bastiaans and M. J. Kropff, 2000. A comparative study on Striga hermonthica interaction with a sensitive and a tolerant sorghum cultivar. Weed Research 40, 479493. Barker, K. R., 1993. Resistance/tolerance and related concepts/terminology in plant nematology. Plants Disease 77, 111-113. Bebawi, F. F. and A. F. Farah, 1981. Effects of parasitic and non-parasitic weeds on sorghum. Experimental Agriculture 17, 415-418. Bebawi, F. F., R. E. Eplee and R. S. Norris, 1985. The dispersion of backpack-applied ethylene in soil. Weed Science 33, 74-77. Bebawi, F. F. and R. E. Eplee, 1986. Efficacy of ethylene as a germination stimulant of Striga hermonthica seed. Weed Science 34, 694-698. Bengaly, M. and T. Defoer, 1997. Perception paysanne l'ampleur du Striga, repartition sur le terroir. Agriculture et Developpement 1997, 13, 52-57. 109 References Berner, D., R. Carsky, K. Dashiell, J. Kling and V. Manyong, 1996. A land management based approach to integrated Striga hermonthica control in sub-Saharan Africa. Outlook on Agriculture 25, 157-164. Bernhard, R. H., J. E. Jensen and C. Andreasen, 1998. Prediction of yield loss caused by Orobanche spp. in carrot and pea crops based on soil seedbank. Weed Research 38, 191-197. Björkman, O. and S. P. Powles, 1984. Inhibition of photosynthetic reactions under water stress: interaction with light level. Planta 153, 376-387. Botanga, C. J., J. G. Kling, D. K. Berner and M. P. Timko, 2002. Genetic variability of Striga asiatica (L.) Kuntz based on AFLP analysis and host-parasite interaction. Euphytica 128, 375-388. Boukar, I., E. Hess Dale and A. Payne William, 1996. Dynamics of moisture, nitrogen, and Striga infestation on pearl millet transpiration and growth. Agronomy Journal 88, 545549. Burdon, J. J., 1997. Crop resistance to parasitic plants. In: J. A. Lane, D. V. Child, G. C. Reiss, V. Entcheva, J. A. Bailey, I. R. Crute and E. B. Holub (eds). The gene for gene relationship in plant parasite interactions. Cab International, Wallingford, pp. 81-97. Caldwell, R. M., J. F. Schafer, L. E. Compton and F. L. Patterson, 1958. Tolerance to cereal leaf rust. Science 128, 714-715. Carsky, R. J., L. Singh and R. Ndikawa, 1994a. Effect of herbicide and handweeding on current and subsequent season Striga hermonthica density on sorghum. International Journal of Pest Management 40, 111-116. Carsky, R. J., L. Singh and R. Ndikawa, 1994b. Suppression of Striga hermonthica on sorghum using a cowpea intercrop. Experimental Agriculture 30, 349-358. Carsky, R. J., R. Ndikawa, R. Kenga, L. Singh, M. Fobasso and M. Kamuanga, 1996. Effect of sorghum variety on Striga hermonthica parasitism and reproduction. Plant Varieties and Seeds 9, 111-118. Cechin, I. and M. C. Press, 1993. Nitrogen relations of the sorghum-Striga hermonthica hostparasite association: growth and photosynthesis. Plant Cell and Environment 16, 237247. Clark, L. J., K. G. Shawe, G. Hoffmann and G. R. Stewart, 1994. The effect of Striga hermonthica (Del.) Benth. infection on gas-exchange characteristics and yield of a sorghum host, measured in the field in Mali. Journal of Experimental Botany 45, 281283. Clarke, D. D., 1986. Tolerance of parasites and disease in plants and its significance in hostparasite interactions. In: D. S. Ingram and P. H. Williams (eds). Advances in plant pathology. 5. Academic Press, New York, pp. 161-197. 110 References Cousens, R., 1985. An empirical model relating crop yield to weed and crop density and a statistical comparison with other models. Journal of Agricultural Science 105, 513521. Cubero, J. I., A. H. Pieterse, S. A. Khalil and J. Sauerborn, 1994. Screening techniques and sources of resistance to parasitic angiosperms. Euphytica 73, 51-58. De La Harpe, A. C. and J. H. Visser, 1979. The chlorophyll concentration and photosynthetic activity of some parasitic flowering plants. Zeitschrift fuer Pflanzenphysiologie 93, 83-87. Debrah, S. K., 1994. Socio-economic constraints to the adoption of weed control techniques: the case of Striga control in the West African Semi-Arid Tropics. International Journal of Pest Management 40, 153-158. Defoer, T., A. Kamara and H. De Groote, 1997. Gender and variety selection: farmers' assessment of local maize varieties in southern Mali. African Crop Science Journal 5, 65-76. Delft, G. J. v., J. D. Graves, A. H. Fitter and M. A. Pruiksma, 1997. Spatial distribution and population dynamics of Striga hermonthica seeds in naturally infested farm soils. Plant and Soil 195, 1-15. DeVries, J., 2000. The inheritance of Striga reactions in maize. In: B. I. G. Haussmann, D. E. Hess, M. L. Koyama, L. Grivet, H. F. W. Rattunde and H. H. Geiger (eds), Breeding for Striga resistance in cereals: proceedings of a workshop held at IITA, Ibadan, Nigeria. pp. 73-84. Doggett, H., 1965. Striga hermonthica on sorghum in East Africa. Journal of Agricultural Science 65, 183-194. Doggett, H., 1982. Factors reducing sorghum yields: Striga and birds. In: L. R. House, L. K. Mughogho and J. M. Peacock (eds). Sorghum in the eighties. Proceedings of the International Symposium on Sorghum, Vol. 1, Patancheru, India. pp. 313-320. Doggett, H., 1988. Sorghum. Longman Group, Essex. Drennan, D. S. H. and S. O. El Hiweris, 1979. Changes in growth regulator substances in Sorghum vulgare infected with Striga hermonthica. In: L. J. Musselman, A. D. Worsham and R. E. Eplee (eds). Proceedings of the Second International Symposium on Parasitic Weeds. North Carolina State University, Raleigh, pp. 144-155. Efron, Y., 1993. Screening maize for tolerance to Striga hermonthica. Plant Breeding 110, 192-200. Egley, G. H., 1971. Mineral nutrition and the parasite-host relationship of witchweed (Striga lutea). Weed Science 19, 528-533. Ejeta, G., L. G. Butler, D. E. Hess and R. K. Vogler, 1991. Genetic and breeding strategies for Striga resistance in sorghum. In: J. K. Ransom, L. J. Musselman, A. D. Worsham and 111 References C. Parker (eds). Proceedings of the 5th International Symposium of Parasitic Weeds, CIMMYT, Nairobi, Kenya, pp. 539-544. Ejeta, G. and L. G. Butler, 1993. Host plant resistance to Striga. In: D. R. Buxton, R. Shibles, R. A. Forsberg, B. L. Blad, K. H. Asay, G. M. Paulsen and R. F. Wilson (eds). International Crop Science Congress, Ames, Iowa, USA, pp. 561-569. Ejeta, G., A. Mohammed, P. Rich, A. Melake Berhan, T. L. Housley and D. E. Hess, 2000. Selection for specific mechanisms for resistance to Striga in sorghum. In: B. I. G. Haussmann, D. E. Hess, M. L. Koyama, L. Grivet, H. F. W. Rattunde and H. H. Geiger, Breeding for Striga resistance in cereals: proceedings of a workshop held at IITA, Ibadan, Nigeria, pp. 29-40. El Hiweris, S. O., 1987. Nature of resistance to Striga-hermonthica (Del.) Benth. parasitism in some Sorghum vulgare (Pers.) cultivars. Weed Research 27, 305-312. Eplee, R. E. and R. S. Norris, 1987. Chemical control of Striga. In: L. J. Musselman. (ed). Parasitic weeds in agriculture. Volume I. Striga. CRC Press Inc., Boca Raton, pp. 173182. Falconer, D. S. and T. F. C. Mackay, 1996. Introduction to Quantitative Genetics. Longman Group Limited, Essex. FAOSTAT, 2004. http://faostat.fao.org/faostat/ Fracheboud, Y., P. Haldimann, J. Leipner and P. Stamp, 1999. Chlorophyll fluorescence as a selection tool for cold tolerance of photosynthesis in maize (Zea mays [L.]). Journal of Experimental Botany 50, 1533-1540. Frost, D. L., A. L. Gurney, M. C. Press and J. D. Scholes, 1997. Striga hermonthica reduces photosynthesis in sorghum: the importance of stomatal limitations and a potential role for ABA. Plant Cell and Environment 20, 483-492. Gbehounou, G. and E. Adango, 2003. Trap crops of Striga hermonthica: in vitro identification and effectiveness in situ. Crop Protection 22, 395-404. Gbehounou, G., E. Adango, J. C. Hinvi and R. Nonfon, 2004. Sowing date or transplanting as components for integrated Striga hermonthica control in grain-cereal crops? Crop Protection 23, 379-386. Gebremedhin, W., J. Goudriaan and H. Naber, 2000. Morphological, phenological and wateruse dynamics of sorghum varieties (Sorghum bicolor) under Striga hermonthica infestation. Crop Protection 19, 61-68. Genty, B., J. M. Briantais and N. R. Baker, 1989. The relationship between the quantum yield of photosynthetic electron transport and quenching of chlorophyll fluorescence. Biochimica et Biophysica Acta 990, 87-92. Goudriaan, J. and H. H. v. Laar, 1994. Modelling potential crop growth processes. Kluwer Publishers, Dordrecht. 112 References Graves, J. D., M. C. Press and G. R. Stewart, 1989. A carbon balance model of the sorghum Striga hermonthica host-parasite association. Plant Cell and Environment 12, 101-108. Graves, J. D., A. Wylde, M. C. Press and G. R. Stewart, 1990. Growth and carbon allocation in Pennisetum typhoides infected with the parasitic angiosperm Striga hermonthica. Plant Cell and Environment 13, 367-373. Gurney, A. L., M. C. Press and J. K. Ransom, 1995. The parasitic angiosperm Striga hermonthica can reduce photosynthesis of its sorghum and maize hosts in the field. Journal of Experimental Botany 46, 1817-1823. Gurney, A. L., M. C. Press and J. D. Scholes, 1999. Infection time and density influence the response of sorghum to the parasitic angiosperm Striga hermonthica. New Phytologist 143, 573-580. Gurney, A. L., M. Adcock, J. D. Scholes and M. C. Press, 2000. Physiological processes during Striga infestation in maize and sorghum. In: B. I. G. Haussmann, D. E. Hess, M. L. Koyama, L. Grivet, H. F. W. Rattunde and H. H. Geiger (eds). Breeding for Striga resistance in cereals: proceedings of a workshop held at IITA, Ibadan, Nigeria, pp. 3-18. Gurney, A. L., A. Taylor, A. Mbwaga, J. D. Scholes and M. C. Press, 2002a. Do maize cultivars demonstrate tolerance to the parasitic weed Striga asiatica? Weed Research 42, 299-306. Gurney, A. L., M. C. Press and J. D. Scholes, 2002b. Can wild relatives of sorghum provide new sources of resistance or tolerance against Striga species? Weed research 42, 317324. Harbinson, J., 1995. Detection of stress in pot plants. Acta Horticultura 405, 320-334. Haussmann, B. I. G., D. E. Hess, B. V. S. Reddy, H. G. Welz and H. H. Geiger, 2000a. Analysis of resistance to Striga hermonthica in diallel crosses of sorghum. Euphytica 116, 33-40. Haussmann, B. I. G., D. E. Hess, H. G. Welz and H. H. Geiger, 2000b. Improved methodologies for breeding Striga-resistant sorghums. Field Crops Research 66, 195211. Haussmann, B. I. G., D. E. Hess, B. V. S. Reddy, S. Z. Mukuru, M. Kayentao, H. G. Welz and H. H. Geiger, 2001a. Pattern analysis of genotype × environment interaction for Striga resistance and grain yield in African sorghum trials. Euphytica 122, 297-308. Haussmann, B. I. G., D. E. Hess, B. V. S. Reddy, S. Z. Mukuru, M. Kayentao, H. G. Welz and H. H. Geiger, 2001b. Quantitative-genetic parameters of sorghum growth under Striga infestation in Mali and Kenya. Plant Breeding 120, 49-56. Havaux, M. and R. Lannoye, 1985. Drought resistance of hardy wheat cultivars measured by a rapid chlorophyll fluorescence test. Journal of Agricultural Sciences 104, 501-504. 113 References Heller, R. and K. Wegmann, 2000. Mechanisms of resistance to Striga hermonthica (Del.) Benth. in Sorghum bicolor (L.) Moench. In: B. I. G. Haussmann, D. E. Hess, M. L. Koyama, L. Grivet, H. F. W. Rattunde and H. H. Geiger (eds). Breeding for Striga resistance in cereals: Proceedings of a workshop held at IITA, Ibadan, Nigeria, pp. 1928. Hess, D. E., 1989. Resistance to Striga hermonthica (Del.) Benth. in sorghum. Ph.D. thesis, Purdue University. Hess, D. E. and G. Ejeta, 1992. Inheritance of resistance to Striga in sorghum genotype SRN39. Plant Breeding 109, 233-241. Hess, D. E., G. Ejeta and L. G. Butler, 1992. Selecting sorghum genotypes expressing a quantitative biosynthetic trait that confers resistance to Striga. Phytochemistry 31, 493-497. Hess, D. E. and B. I. G. Haussmann, 1999. Status quo of Striga control: Prevention, mechanical and biological control methods and host plant resistance. In: J. Kroschel, H. Mercer-Quarshie and J. Sauerborn (eds). Advances in Parasitic Weed Control at On-farm Level. Vol. 1. Joint Action to Control Striga in Africa. Margraf Verlag, Weikersheim, pp. 75-87. Hess, D. E. and H. Dodo, 2004. Potential for sesame to contribute to integrated control of Striga hermonthica in the west African Sahel. Crop Protection 23, 515-522. Joel, D. M., 2000. The long-term approach to parasitic weeds control: manipulation of specific developmental mechanisms of the parasite. Crop Protection 19, 753-758. Johnson, D. E., C. R. Riches, R. Diallo and M. J. Jones, 1997. Striga on rice in West Africa: crop host range and the potential of host resistance. Crop Protection 16, 153-157. Kamara, A., T. Defoer and H. d. Groote, 1996. Selection of new varieties through participatory research, the case of corn in south Mali. Tropicultura 14, 100-105. Kim, S. K., 1991. Breeding for tolerance and general resistance in Maize: A novel approach to combatting Striga in Africa. In: Improving Striga management in Africa: 2nd general workshop of the Pan-African Striga Network (PASCON), Nairobi, Kenya, pp. 168176. Kim, S. K., 1994. Genetics of maize tolerance of Striga hermonthica. Crop Science 34, 900907. Kim, S. K., 1996. Horizontal resistance: core to a research breakthrough to combat Striga in Africa. Integrated Pest Management Reviews 1, 229-249. Kim, S. K. and V. O. Adetimirin, 1997a. Striga hermonthica seed inoculum rate effects on maize hybrid tolerance and susceptibility expression. Crop Science 37, 1066-1071. Kim, S. K. and V. O. Adetimirin, 1997b. Responses of tolerant and susceptible maize varieties to timing and rate of nitrogen under Striga hermonthica infestation. Agronomy Journal 89, 38-44. 114 References Kim, S. K., J. M. Fajemisin, C. The, A. Adepoju, J. Kling, B. Badu Apraku, M. Versteeg, R. Carsky and S. T. O. Lagoke, 1998. Development of synthetic maize populations for resistance to Striga hermonthica. Plant Breeding 117, 203-209. Kim, S. K., V. O. Adetimirin, C. The and R. Dossou, 2002. Yield losses in maize due to Striga hermonthica in West and Central Africa. International Journal of Pest Management. 48, 211-217. Kling, J. G., J. M. Fajemisin, B. Badu Apraku, A. Diallo, A. Menkir and A. Melake Berhan, 2000. Striga resistance breeding. In: B. I. G. Haussmann, D. E. Hess, M. L. Koyama, L. Grivet, H. F. W. Rattunde and H. H. Geiger (eds). Breeding for Striga resistance in cereals: Proceedings of a workshop held at IITA, Ibadan, Nigeria, pp. 103-118. Koskela, T., S. Puustinen, V. Salonen and P. Mutikainen, 2002. Resistance and tolerance in a host plant-holoparasitic plant interaction: genetic variation and costs. Evolution 56, 899-908. Kropff, M. J., 1987. Physiological effects of sulphur dioxide. I. The effect of SO2 on photosynthesis and stomatal regulation of Vicia faba (L.). Plant, Cell and Environment 10, 753-760. Kroschel, J., 1999. Analysis of the Striga problem, the first step towards further joint actions. In: J. Kroschel, H. Mercer-Quarshie and J. Sauerborn. (eds). Advances in Parasitic Weed Control at On-Farm Level. Vol. 1 Joint Action to Control Striga in Africa. Margraf Verlag, Weikersheim, pp. 3-26. Kroschel, J. and D. Muller Stover, 2004. Biological control of root parasitic weed with plant pathogens. Weed Biology and Management, 423-438. Kuijt, J., 1969. The biology of Parasitic Flowering Plants. University of California Press, Berkeley. Kuijt, J., 1977. Haustoria of phanerogamic parasites. Annual Review of Phytopathology 17, 91-118. Kulkarni, N. and V. K. Shinde, 1985. Genetic analysis of Striga resistance in sorghum parameters of resistance. Indian Journal of Genetics and Plant Breeding 1985, 545551. Lagoke, S. T. O., V. Parkinson and R. M. Agunbiade, 1991. Parasitic weeds and control methods in Africa. In: S. K. Kim (ed). Combating Striga in Africa. Proceedings of an International Workshop organized by IITA, ICRISAT and IDRC, Ibadan, Nigeria, pp. 3-14. Last, F. T., 1960. Effect of cultural treatments on incidence of Striga hermonthica (Del.) benth. and yields of sorghum in the Sudan: field experiments 1957/58. Annals of Applied Biology 40, 207-229. 115 References Lendzemo, V. W. and Th. W. Kuyper, 2001. Effects of arbuscular mycorrhizal fungi on damage by Striga hermonthica on two contrasting cultivars of sorghum (Sorghum bicolor). Agriculture Ecosystems and Environment 87, 29-35. Lendzemo, V. W., Th. W. Kuyper, M. J. Kropff and A. v. Ast, 2005. Field inoculation with arbuscular mycorrhizal fungi reduces Striga hermonthica performance on cereal crops and has the potential to contribute to integrated Striga management. Field Crops Research 91, 51-61. Maiti, R. K., K. V. Ramaiah, S. S. Bisen and V. L. Chidley, 1984. A comparitive study of the haustorial development of Striga asiatica (L.) Kuntze on sorghum cultivars. Annals of Botany 54, 447-457. Maxwell, K. and G. N. Johnson, 2000. Chlorophyll fluorescence - a practical guide. Journal of Experimental Botany 51, 659-668. Mboob, S. S., 1989. A regional programme for Striga control in West and central Africa. In: T. O. Robson and H. R. Broad (eds). Proceedings of the FAO/OAU all African Government Consultation on Striga Control, 1988, Maroua, Cameroon. FAO plant protection paper 96, pp. 190-194. McCullagh, P. and J. A. Nelder, 1989. Generalized Linear Models. Chapman and Hall, London. Mohamed, A., A. Ellicott, T. L. Housley and G. Ejeta, 2003. Hypersensitive response to Striga infection in sorghum. Crop Science. 43: 4, 1320-1324. Murdoch, A. J. and E. M. Kunjo, 2003. Depletion of natural soil seedbanks of Striga hermonthica in West Africa under different integrated management regimes. Aspects of Applied Biology 69, 261-268. Nogues, S., L. Alegre, J. L. Araus, L. Perez-Aranda and R. Lannoye, 1994. Modulated chlorophyll fluorescence and photosynthetic gas exchange as rapid screening methods for drought tolerance in barley genotypes. Photosynthetica 30, 465-474. Obilana, A. B., 1983. Striga studies and control in Nigeria. In: 2nd International Workshop on Striga, 1981, Ouagadougou, Upper Volta. ICRISAT, Patancheru, India, pp. 87-98. Obilana, A. B., 1984. Inheritance of resistance to Striga (Striga hermonthica [Benth]) in sorghum. Protection Ecology 7, 305-311. Obilana, A. B., 1988. Strategies and recommendations for Striga control in sorghum and millets. In: The OAU/STRC-ICRISAT EARSAM 6th Regional Sorghum and Millet Workshop, Mogadishu, Somalia, pp. 461-473. Okonkwo, S. N. C., 1966. Studies on Striga senegalensis II. Translocation of C14 labelled photosynthate, urea-C14 and sulphur-35 between host and parasite. American Journal of Botany 53, 142-148. 116 References Olivier, A., K. V. Ramaiah and G. D. Leroux, 1991a. Selection of sorghum (Sorghum bicolor [L.] Moench) varieties resistant to the parasitic weed Striga hermonthica (Del.) Benth. Weed Research 31, 219-226. Olivier, A., N. Benhamou and G. D. Leroux, 1991b. Cell surface interactions between sorghum roots and the parasitic weed Striga hermonthica cytochemical aspects of cellulose distribution in resistant and susceptible host tissues. Canadian Journal of Botany 69, 1679-1690. Olivier, A. and G. D. Leroux, 1992. Root development and production of a witchweed (Striga spp.) germination stimulant in sorghum (Sorghum bicolor) cultivars. Weed Science 40, 542-545. Olivier, A., K. V. Ramaiah and G. D. Leroux, 1992. Assessment of the resistance of sorghum (Sorghum bicolor) lines to the parasitic weed Striga-hermonthica in Burkina Faso. Phytoprotection. 1992; 73, 13-23. Olsovska, K., M. Brestic and J. Hudec, 2000. Use of physiological characteristics of barley (Hordeum vulgare [L.]) for screening genotypes tolerant to drought. Acta Fytotechnica et Zootechnica 3, 103-106. Omanya, G. O., B. I. G. Haussmann, D. E. Hess, B. V. S. Reddy, S. Z. Mukuru, H. G. Welz and H. H. Geiger, 2000. Evaluation of laboratory, pot and field measures of Striga resistance in sorghum. In: B. I. G. Haussmann, D. E. Hess, M. L. Koyama, L. Grivet, H. F. W. Rattunde and H. H. Geiger (eds). Breeding for Striga resistance in cereals: Proceedings of a workshop held at IITA, Ibadan, Nigeria, pp. 59-72. Omanya, G. O., B. I. G. Haussmann, D. E. Hess, B. V. S. Reddy, M. Kayentao, H. G. Welz and H. H. Geiger, 2004. Utility of indirect and direct selection traits for improving Striga resistance in two sorghum recombinant inbred populations. Field Crops Research 89, 237-252. Ort, D. R. and N. R. Baker, 2002. A photoprotective role for O2 as an alternative electron sink in photosynthesis? Current Opinion in Plant Biology 5, 193-198. Orton, W. A., 1909. The development of farm crops resistant to disease. In: US Department of Agriculture Yearbook 1908, 453-464. Oswald, A. and J. K. Ransom, 2004. Response of maize varieties to Striga infestation. Crop Protection 23, 89-94. Oxborough, K. and N. R. Baker, 1997. Resolving chlorophyll fluorescence images of photosynthetic efficiency into photochemical and non-photochemical components calculation of qP and Fv/Fm without measuring Fo. Photosynthesis Research 54, 135142. Pageau, K., A. P. Simier, N. Naulet, R. Robins and A. Fer, 1998. Carbon dependency of the hemiparasite Striga hermonthica on Sorghum bicolor determined by carbon isotopic and gas exchange analyses. Australian Journal of Plant Physiology 25, 695-700. 117 References Pageau, K., P. Simier, B. Le Bizec, R. J. Robins and A. Fer, 2003. Characterization of nitrogen relationships between Sorghum bicolor and the root-hemiparasitic angiosperm Striga hermonthica (Del.) Benth. using K15NO3 as isotopic tracer. Journal of Experimental Botany 54, 789-799. Parker, C., 1991. Protection of crops against parasitic weeds. Crop Protection 10, 6-22. Parker, C. and C. R. Riches, 1993. Parasitic weeds of the world: Biology and control. Cab International, Wallingford. Parlevliet, J. E., 1979. Components of resistance that reduce the rate of epidemic development. Annual Review of Phytopathology 17, 203-222. Payne, R. W., A. E. Ainsley, K. E. Bicknell and M. F. Franklin, 1993. Genstat 5 release 3: reference manual. Oxford Science Publications, Oxford, Clarendon. Pierce, S., A. M. Mbwaga, M. C. Press and J. D. Scholes, 2003. Xenognosin production and tolerance to Striga asiatica infection of high-yielding maize cultivars. Weed Research 43, 139-145. Prabhakara Setty, T. K. and M. M. Hosmani, 1981. Effect of Striga infestation of sorghum. In: Proceedings of the Eight Asian-Pacific Weed Science Society, ICRISAT, Patancheru, India, pp. 287-289. Press, M. C. and G. R. Stewart, 1987. Growth and photosynthesis in Sorghum bicolor infected with Striga hermonthica. Annals of Botany 60, 657-662. Press, M. C., J. M. Tuohy and G. R. Stewart, 1987a. Gas exchange characteristics of the sorghum Striga host-parasite association. Plant Physiology 84, 814-819. Press, M. C., N. Shah, J. M. Tuohy and G. R. Stewart, 1987b. Carbon isotope ratios demonstrate carbon flux from C4 host to C3 parasite. Plant Physiology 85, 1143-1145. Press, M. C., J. D. Graves and G. R. Stewart, 1988. Transpiration and carbon acquisition in root hemi-parasites. Journal of Experimental Botany 39, 1009-1014. Press, M. C., A. L. Gurney, D. L. Frost and J. D. Scholes, 1996. How does the parasitic angiosperm Striga hermonthica influence host growth and carbon relations? In: M. T. Moreno, J. I. Cubero, D. Berner, D. Joel, L. J. Musselman and C. Parker (eds). Advances in parasitic plant research. Sixth International Parasitic Weed Symposium. Cordoba, Spain, pp. 303-310. Rabbinge, R., 1993. The ecological background of food production. In: Crop protection and sustainable agriculture. Ciba Foundation Symposium 177. John Wiley and Sons, Chichester, pp. 2-29. Ramaiah, K. V. and C. Parker, 1982. Striga and other weeds in sorghum. In: L. R. House, L. K. Mughogho and J. M. Peacock (eds). Sorghum in the eighties: Proceedings of the international Symposium on Sorghum, Patancheru, India, pp. 291-302. Ramaiah, K. V., 1984. Patterns of Striga resistance in sorghum and millets with special emphasis on Africa. In: E. S. Ayensu, H. Doggett, R. D. Keynes, J. Marton Lefevre, L. 118 References J. Musselman, C. Parker and A. Pickering (eds). Proceedings of the International Workshop on the Biology and Control of Striga, Dakar, Senegal. ICSU Press, Paris, pp. 71-92. Ramaiah, K. V., 1987a. Breeding cereal grains for resistance to Witchweed. In: L. J. Musselman (ed). Parasitic weeds in agriculture. CRC Press, Boca Raton, pp. 227-242. Ramaiah, K. V., 1987b. Control of Striga and Orobanche species - a review. In: H. C. Weber and W. Forstreuter (eds). Proceedings of the 4th ISPFP, Marburg, pp. 637-664. Ramaiah, K. V., 1988. Breeding for Striga resistance in sorghum and millet. In: S. K. Kim (ed). Combating Striga in Africa. Proceedings of an International Workshop organized by IITA, ICRISAT and IDRC, Ibadan, Nigeria, pp. 75-80. Ramaiah, K. V., V. L. Chidley and L. R. House, 1990. Inheritance of Striga seed-germination stimulant in sorghum. Euphytica 45, 33-38. Ramaiah, K. V., V. L. Chidley and L. R. House, 1991. A time-course study of early establishment stages of parasitic angiosperm Striga asiatica on susceptible sorghum roots. Annals of Applied Biology 118, 403-410. Ransom, J. K., 2000. Long-term approaches for the control of Striga in cereals: field management options. Crop Protection 19, 759-763. Rascher, U., M. Liebig and U. Lüttge, 2000. Evaluation of instant light-response curves of chlorophyll fluorescence parameters obtained with a portable chlorophyll fluorometer on site in the field. Plant Cell and Environment 23, 1397-1405. Raven, J. A., 1983. Phytophages of xylem and phloem: a comparison of animal and plant sapfeeders. Advances in Ecological Research 13, 135-234. Riches, C. R. and C. Parker, 1995. Parasitic plants as weeds. In: M. C. Press and J. D. Graves (eds). Parasitic plants. Chapman and Hall, London, pp. 226-255. Riopel, J. L. and M. P. Timko, 1995. Haustorial initiation and differentiation. In: M. C. Press and J. D. Graves (eds). Parasitic Plants. Chapman and Hall, London, 39-79. Rodenburg, J., L. Bastiaans, E. Weltzien and D. E. Hess, 2005. How can selection for Striga resistance and tolerance based on field observations be improved? Field Crops Research 93, 34-50. Rodenburg, J., L. Bastiaans and M. J. Kropff, Accepted. Characterization of host tolerance to Striga hermonthica. Euphytica. Rodenburg, J., L. Bastiaans, A. v. Ast and M. J. Kropff, Submitted. Effects of host plant genotype and seed bank density on Striga reproduction. Weed Research. Rogers, W. E. and R. R. Nelson, 1962. Penetration and nutrition of Striga asiatica. Phytopathology 52, 1064-1070. Samson, D. A. and K. S. Werk, 1986. Size-dependent effects in the analysis of reproductive effort in plants. American Naturalist 127, 667-680. 119 References Sauerborn, J., 1991. The economic importance of the phytoparasites Orobanche and Striga. In: J. K. Ransom, L. J. Musselman, A. D. Worsham and C. Parker (eds). Proceedings of the 5th International Symposium of Parasitic Weeds, CIMMYT, Nairobi, Kenya, pp. 137-143. Saunders, A. R., 1933. Studies in phanerogamic parasitism, with particular reference to Striga lutea (Lour). South Africa Department of Agriculture, Science Bulletin 128, 1-56. Schafer, J. F., 1971. Tolerance to plant disease. Annual Review of Phytopathology 9, 235252. Schapendonk, A. H. C. M., O. Dolstra and O. v. Kooten, 1989a. The use of chlorophyll fluorescence as a screening method for cold tolerance in maize. Photosynthesis Research 20, 235-247. Schapendonk, A. H. C. M., C. J. T. Spitters and P. J. Groot, 1989b. Effects of water stress on photosynthesis and chlorophyll fluorescence of five potato cultivars. Potato Research 32, 17-32. Schapendonk, A. H. C. M., P. E. L. v. d. Putten, O. Dolstra, S. R. Haalstra and W. J. M. Tonk, 1992. Chlorophyll fluorescence: a non-destructive method for detecting damage in the photosynthetic apparatus in plants. Acta Horticultura 304, 61-70. Scharen, A. L. and J. M. Krupinsky, 1969. Effect of Septoria nodorum infection on CO2 absorption and yield of wheat. Phytopathology 59, 1298-1301. Schreiber, U., 1986. Detection of rapid induction kinetics with a new type of high-frequency modulated chlorophyll fluorometer. Photosynthesis Research 9, 261-272. Schulz, S., M. A. Hussaini, J. G. Kling, D. K. Berner and F. O. Ikie, 2003. Evaluation of integrated Striga hermonthica control technologies under farmer management. Experimental Agriculture 39, 99-108. Schulze, E. D., N. C. Turner and G. Glatzel, 1984. Carbon, water and nutrient relations of the mistletoes and their hosts: an hypothesis. Plant Cell and Environment 7, 293-299. Seel, E. D., I. Cechin, C. A. Vincent and M. C. Press, 1992. Carbon partitioning and transport in parasitic angiosperms and their hosts. In: C. J. Pollock, J. F. Farrah and A. J. Gordon (eds). Carbon partitioning within and between organisms. Bios Scientific Publishers Ltd, Abingdon, Oxfordshire, pp. 199-223. Shah, N., N. Smirnoff and G. R. Stewart, 1987. Photosynthesis and stomatal characteristics of Striga hermonthica in relation to its parasitic habit. Physiologia Plantarum 69, 699703. Shew, H. D. and B. B. Shew, 1994. Host resistance. In: C. L. Campbell and D. M. Benson (eds). Epidemiology and Management of Root Diseases. Springer Verlag, Berlin, pp. 244-275. Showemimo, F. A., 2003. Selection Criteria for combining high yield and Striga resistance in sorghum. Tropicultura 21, 157-159. 120 References Sinebo, W. and D. S. H. Drennan, 2001. Vegetative growth of sorghum and Striga hermonthica in response to nitrogen and the degree of host root infection. European Journal of Plant Pathology 107, 849-860. Smith, L. H., A. J. Keys and M. C. W. Evans, 1995. Striga hermonthica decreases photosynthesis in Zea mays through effects on leaf cell structure. Journal of Experimental Botany 46, 759-765. Smith, M. C. and M. Webb, 1996. Estimation of the seedbank of Striga spp. (Scrophulariaceae) in Malian fields and the implications for a model of biocontrol of Striga hermonthica. Weed Research 36, 85-92. Sokal, R. R. and F. J. Rohlf, 1995. Biometry. W.H. Freeman and Company, New York. Spitters, C. J. T., M. J. Kropff and W. d. Groot, 1989. Competition between maize and Echinochloa crus-galli analysed by a hyperbolic regression model. Annals of Applied Biology 115, 541-551. Stakman, E. C. and J. J. Christensen, 1960. The problem of breeding resistant varieties. In: J. G. Horsfall and A. E. Dimond (eds). Plant Pathology 3. Academic Press, New York, pp. 567-624. Stewart, G., 1990. Witchweed: a parasitic weed of grain crops. Outlook on Agriculture 19, 115-117. Tanksley, S. D., N. D. Young, A. H. Paterson and M. W. Bonierbale, 1989. RFLP mapping in plant breeding: new tools for an old science. Biotechnology 7, 257-264. Taylor, A., J. Martin and W. E. Seel, 1996. Physiology of the parasitic association between maize and Witchweed (Striga hermonthica): is ABA involved. Journal of Experimental Botany 47, 1057-1065. Thompson, B. K., J. Weiner and S. I. Warwick, 1991. Size-dependent reproductive output in agricultural weeds. Canadian Journal of Botany 69, 442-446. Trudgill, D. L., 1991. Resistance to and tolerance of plant parasitic nematodes in plants. Annual Review of Phytopathology 29, 167-192. Tuquet, C., N. Farineau and G. Salle, 1990. Biochemical composition and photosynthetic activity of chloroplasts from Striga hermonthica and Striga aspera, root parasites of field-grown cereals. Physiologia Plantarum 78, 574-582. Vallance, K. B., 1950. Studies on the germination of the seeds of Striga hermonthica I. The influence of moisture treatment, stimulant dilution, and after-ripening on germination. Annals of Botany 14, 347-363. Vasudeva Rao, M. J., 1984. Patterns of resistance to Striga asiatica in sorghum and millets, with special reference to Asia. In: E. S. Ayensu, H. Doggett, R. D. Keynes, J. Marton Lefevre, L. J. Musselman, C. Parker and A. Pickering (eds). Proceedings of the International Workshop on the Biology and Control of Striga, Dakar, Senegal. ICSU Press, Paris, pp. 71-92. 121 References Vasudeva Rao, M. J., V. L. Chidley and L. R. House, 1989. Estimates of grain yield losses caused in sorghum (Sorghum bicolor [L.] Moench) by Striga asiatica (L.) Kuntze obtained using the regression approach. Agriculture Ecosystems and Environment 25, 139-150. Verkleij, J. A. C. and A. H. Pieterse, 1994. Genetic variability of Orobanche (broomrape) and Striga (witchweed) in relation to host specificity. In: A. H. Pieterse, J. A. C. Verkleij and S. J. t. Borg (eds). Biology and management of Orobanche, Proceedings of the Third International Workshop on Orobanche and Related Striga Research. Wageningen, pp. 67-79. Vogler, R. K., G. Ejeta and L. G. Butler, 1996. Inheritance of low production of Striga germination stimulant in sorghum. Crop Science 36, 1185-1191. Von Caemmerer, S. and G. D. Farquhar, 1981. Some relationships between the biochemistry of photosynthesis and the gas exchange of leaves. Planta 153, 376-387. Watling, J. R. and M. C. Press, 1997. How is the relationship between the C4 cereal Sorghum bicolor and the C3 root hemi-parasites Striga hermonthica and Striga asiatica affected by elevated CO2? Plant Cell and Environment 20, 1292-1300. Weaver, S. E., N. Smits and C. S. Tan, 1987. Estimating yield losses of tomatoes (Lycopersicon esculentum) caused by nightshade (Solanum spp.) interference. Weed Science 35, 163-168. Webb, M. and M. C. Smith, 1996. Biology of Striga hermonthica (Scrophulariaceae) in Sahelian Mali: effects on pearl millet yield and prospects of control. Weed Research 36, 203-211. Weber, G., K. Elemo, S. T. O. Lagoke, A. Awad and S. Oikeh, 1995. Population dynamics and determinants of Striga hermonthica on maize and sorghum in savanna farming systems. Crop Protection 14, 283-290. Weerasuriya, Y., S. B. A., D. Hess, G. Ejeta and L. G. Butler, 1993. Influence of conditions and genotype on the amount of Striga germination stimulants exuded by roots of several host crops. Journal of Agricultural and Food Chemistry 41, 1492-1496. White, P. C. and G. L. Wilson, 1965. Effects of water stress on the reproductive development of Sorghum vulgare (pers). University of Queensland Papers IV (14). Williams, C. N., 1959. Resistance of sorghum to Witchweed. Nature 184, 1511-1512. Williams, C. N., 1961a. Growth and morphogenesis of Striga seedlings. Nature 189, 378-382. Williams, C. N., 1961b. Tropism and morphogenesis of Striga seedlings in the host rhizosphere. Annals of Botany 25, 407-414. Wilson, J. P., D. E. Hess and W. W. Hanna, 2000. Resistance to Striga hermonthica in wild accessions of the primary gene pool of Pennisetum glaucum. Phytopathology 90, 1169-1172. 122 References Wilson, J. P., D. E. Hess, W. W. Hanna, K. A. Kumar and S. C. Gupta, 2004. Pennisetum glaucum subsp. monodii accessions with Striga resistance in West Africa. Crop Protection 23, 865-870. Wit de, C. T., 1953. A physical theory on placement of fertilizer. PhD thesis, Landbouw Hogeschool Wageningen. Yoder, J. I., 2001. Host-plant recognition by parasitic Scrophulariaceae. Current Opinion in Plant Biology 4, 359-365. 123 Summary Striga hermonthica (Del. Benth) is a parasitic weed on tropical cereals, such as sorghum (Sorghum bicolor [L.]), millet (Pennisetum glaucum [L.] R. Br. or P. americanum [L.] K. Schum), maize (Zea mays [L.]) and rice (Oryza glaberrima [Steudel] and O. sativa [L.]). It seriously reduces crop yields and is therefore one of the main biotic constraints to cereal crop production in the semi-arid tropics, endangering the livelihood of many subsistence farmers. Striga negatively affects its host by subtracting host assimilates, water and nutrients and by changing the plant hormone balance, which in turn alters host plant allometry (leaf/stem ratio) and lowers the photosynthetic rate of the host. Several control options have been proposed but none of these measures on its own is both sufficiently effective as well as practical and accessible for resource-poor farmers. Integration of various control options is thought to be the best approach to combat Striga. In integrated Striga management the role of crop varieties with improved resistance and tolerance is indispensable. For the development of such improved varieties the identification and selection of superior breeding material is of vital importance. This study focussed on the effects of Striga on host physiology and production and on the effects of host genotype on the Striga infection and reproduction rate with the aim of enhancing the understanding of their interaction and developing suitable tools for field selection of resistant and tolerant breeding material. Between 2001 and 2004, three field and two pot experiments were conducted in Samanko (Mali) and two pot experiments were conducted in a greenhouse in Wageningen (The Netherlands). In the three field experiments, 10 different sorghum genotypes (CK60-B, CMDT39, E36-1, Framida, IS9830, N13, Seredo, Serena, SRN39 and Tiémarifing) were grown in the presence and absence of Striga seeds, and in the last year Striga plots were infested at two different levels (‘low’ and ‘high’). In 2001 and 2003 previously infested fields were used and Striga-free plots were achieved through ethylene gas injections with a backpack ethylene gas injector, while existing Striga plots received additional Striga seeds. In 2002, a non-contaminated field was used and Striga plots were achieved through artificial infestation, while avoiding contamination of the Striga-free plots. Small dikes surrounding the different plots prevented the Striga-free plots from contamination (through water flow) by escaping Striga seeds from neighbouring Striga plots. In three of the four pot experiments a selection of four of the 10 genotypes (CK60-B, E36-1, Framida and Tiémarifing) were grown under different Striga infestation levels. Each year sorghum seeds were multiplied through self-pollination for use in the subsequent year. Striga seeds used in all experiments were derived from sorghum plants grown in Samanko. In all 125 Summary experiments, routine observations and measurements were done: Striga emergence counts, Striga dry weights and sorghum grain yield. More detailed observations, depending on the experiment, included photosynthesis and chlorophyll fluorescence of sorghum plants and flowering dates, flowerstalk dry weights, number of reproductive plants and seed production of Striga. In the first study of this thesis (Chapter 2), field selection measures for resistance and tolerance were evaluated and discussed, based on results of the three field experiments and one pot experiment in Samanko with all 10 genotypes. Both resistance and tolerance are important yield determining traits under Striga infestation. Under low infestation, tolerance was relatively more important for yield than resistance, whereas resistance was more important under high infestation. The area under the Striga number progress curve (ASNPC) and the maximum number of aboveground Striga plants (NSmax) were discriminative and consistent measures for resistance. It proved more difficult to find a satisfactory measure for tolerance. Genotype differences in resistance and the non-linearity of the relation between Striga infection and yield loss are the main reasons for this. In Chapter 3 an attempt was made to resolve the relationship between Striga infection and sorghum yield loss in order to find a suitable selection measure for tolerance against Striga. Data from three pot experiments, one in Samanko (2003) and two in the greenhouse in Wageningen (2003 and 2004) with CK60-B, E36-1, Framida and Tiémarifing, were used for this purpose. There were significant genotype, infestation and genotype × infestation effects on sorghum yield. The relation between infestation level and infection level was density dependent. As a result, the wide range of infestation levels resulted in a relatively narrow range of infection levels for each individual genotype. In the 2004 experiment the range of infestation levels was even further extended to obtain at least a narrow range of identical infection levels for all four genotypes. The relationship between Striga infection level and relative yield loss showed to be non-linear. The results suggested that for resistant genotypes tolerance could best be quantified by the reduction of relative yield loss per aboveground Striga plant, whereas for less resistant genotypes the maximum relative yield loss could best be used. Whether both expressions of tolerance are interrelated remained unresolved. Despite the identification of these tolerance measures a main bottle neck of selection for tolerance to Striga based on yield data remained the requirement of control plots. Chapter 4 explored options for the use of photosynthesis or related measurements in screening for tolerance to Striga hermonthica. This work was based on the results from two pot experiments conducted in a greenhouse in Wageningen with CK60-B, E36-1, Framida and Tiémarifing. The aim was to find a better measure for tolerance without the requirement of control plots and various infestation levels. 126 Summary CO2 assimilation rate of sorghum plants was significantly reduced by Striga infection. This was accompanied by a reduced stomatal conductance, which was shown not to be the main cause of the reduction in photosynthetic rate. Other processes related to photosynthesis that were affected by Striga infection were transpiration rate, photochemical quenching and non-photochemical quenching, electron transport and photosynthesis per electron transport. Sensitive genotypes were affected earlier, more severe and at lower infestation levels than tolerant genotypes. Tiémarifing, earlier identified as the most tolerant variety, was the only genotype showing no significant Striga effect on any parameter at any measurement time. Consequently, Striga tolerant genotypes may be identified through photosynthesis or related measurements. Particularly suitable for this purpose seemed measurements of photochemical quenching and electron transport rate through photosystem II. These parameters facilitate screening at one single infestation level and without the requirement of Striga-free control plots. It was recommended to screen between first Striga emergence and sorghum flowering and at infestation levels of at least 300,000 Striga seeds m-2. The objective of Chapter 5 was to study the genotype effects on Striga reproduction and to find a suitable selection measure for Striga seed production. Data for this study were derived from the three field experiments in Mali with 10 sorghum genotypes. There were significant genotype and infestation level effects on aboveground Striga numbers, aboveground Striga dry weights and Striga seed production. There were significant correlations between aboveground Striga numbers and seed production and highly significant correlations between aboveground Striga dry weights and Striga seed production. Aboveground Striga dry weight and Striga flowerstalk dry weight were found to be good indicators for Striga reproduction. An increase in infestation level in the field generally resulted in a proportional increase in infection level. Exceptions were the most susceptible genotypes (CK60-B and E36-1), for which high infestation levels resulted in a less than proportional increase in number of emerged Striga plants. Increasing infestation levels resulted in a disproportional increase in Striga dry weights (total and flowerstalks) and seed production for all genotypes. The relations between Striga infestation and Striga infection and between Striga infection and reproduction were both density and genotype dependent. Density dependence however was observed at much lower infestation levels for the second relationship (infection - reproduction) than for the first relationship (infestation - infection), where it only appeared at relatively high infestation levels. Striga reproduction continued after crop harvest. Differences in Striga seed production between Striga infestation levels decreased between harvest 127 Summary and the end of the Striga life-cycle. There were no significant genotype effects on seed dry weight per seed capsule. In this study, suitable selection measures were found for resistance against Striga parasitism and reproduction as well as for tolerance against Striga infection. Maximum aboveground Striga numbers is a reliable selection measure for resistance. Striga flowerstalk dry weight can be used to identify genotypes supporting low Striga reproduction. Screening for tolerance showed to be more complicated. For susceptible genotypes, the highest relative yield loss that was attained is a suitable selection measure for tolerance, whereas for more resistant genotypes, the relative yield loss per Striga infection seems more appropriate. However, Striga-free control plots are indispensable for selection of tolerant genotypes when selection is based on host plant yield. Chlorophyll fluorescence parameters, and especially photochemical quenching (Pq) and electron transport (ETR) were identified as promising alternative screening measures for tolerance, particularly since with the use of these measures tolerance can potentially be identified without the presence of Striga free controls. Host plant genotype choice affects Striga reproduction efforts. Differences in Striga seed production among sorghum genotypes can to a large extent be explained by differences in resistance level of the host plants. Resistance was responsible for a 70 to 93% reduction in Striga seed production in this study. To reduce the Striga seed bank, additional control options such as hand weeding before harvest are required. 128 Samenvatting Striga hermonthica (Del.) Benth is een onkruid dat parasiteert op tropische graangewassen zoals sorghum (Sorghum bicolor [L.]), gierst (Pennisetum glaucum [L.] R. Br. of P. americanum [L.] K. Schum), mais (Zea mays [L.]) en rijst (Oryza glaberrima [Steudel] en O. sativa [L.]). Het reduceert de opbrengst van zijn gastheer aanzienlijk en is daarom één van de meest belangrijke biotische problemen voor graanproductie in de semi-aride tropen en een groot gevaar voor de bestaanszekerheid van vele zelfvoorzienende boeren. Striga benadeelt zijn gastheer door onttrekking van assimilaten, water en nutriënten en doordat het de hormoonbalans van de gastheer verstoort, waardoor de allometrie van de plant (verhouding blad/stengel/wortel) verandert en de fotosynthesesnelheid afneemt. Verscheidene beheersmaatregelen zijn voorhanden, maar geen van deze maatregelen is zowel voldoende effectief als praktisch en toegankelijk voor arme boeren. Integratie van meerdere beheersmaatregelen lijkt daarom de beste benadering om Striga te bestrijden. In zo’n geïntegreerde aanpak is de rol van gewasvariëteiten, met verhoogde tolerantie en resistentie tegen de parasiet, essentieel. Het ontwikkelen van dergelijke variëteiten vereist het identificeren en selecteren van geschikt uitgangsmateriaal voor veredeling. Het doel van deze studie was het verkrijgen van een beter inzicht in de gastheer-Striga relatie om bruikbare methoden en selectiecriteria te ontwikkelen voor het verkrijgen van goed uitgangsmateriaal voor de verdere veredeling van rassen met een hoge mate van tolerantie en resistentie tegen Striga. Dit onderzoek concentreerde zich op het bestuderen van Striga effecten op de fysiologie en productie van zijn gastheer, alsmede op de effecten van het gastheer-genotype op Striga parasitisme en reproductie. Tussen 2001 en 2004 zijn drie veldexperimenten en twee potexperimenten uitgevoerd in Samanko (Mali) en twee potexperimenten in de tropische kas in Wageningen (Nederland). In de drie veldexperimenten werden 10 verschillende sorghum genotypen (CK60-B, CMDT39, E36-1, Framida, IS9830, N13, Seredo, Serena, SRN39 en Tiémarifing) geteeld op zowel Striga-vrije als Striga-besmette velden. In het laatste veldseizoen (2003) werden de Striga-velden op twee besmettingsniveaus aangelegd (‘laag’ en ‘hoog’). In 2001 en 2003 werden reeds besmette velden gebruikt, waarin Striga-vrije velden werden gecreëerd door middel van ethyleengas injecties, terwijl aan Striga velden extra Striga zaden werden toegevoegd. In 2002 werd een onbesmet veld gebruikt waarbinnen Striga velden werden gecreëerd door handmatige toediening van Striga zaad. Dijkjes om de verschillende velden beschermden Striga-vrije velden tegen besmetting (door 129 Samenvatting afstromend water) met Striga zaad uit naastgelegen velden. In drie van de vier potexperimenten werd een selectie van vier van de 10 genotypen (CK60-B, E36-1, Framida en Tiémarifing) gebruikt bij een reeks van Striga besmettingsniveaus. Elk jaar werd sorghumzaad vermenigvuldigd door middel van zelfbestuiving, voor gebruik in het daaropvolgende jaar. Het Striga zaad dat in de experimenten gebruikt werd was afkomstig van sorghumplanten die in Samanko (Mali) geteeld waren. In alle experimenten werden routinewaarnemingen en -metingen gedaan, waaronder Striga opkomst-tellingen en de bepaling van het Striga drooggewicht en de korrelopbrengst van sorghum. In sommige experimenten werden meer gedetailleerde waarnemingen verricht, zoals fotosynthese- en chlorofylfluorescentiemetingen aan sorghumplanten alsmede bloeitijdstip en zaadproductie karakteristieken van Striga planten. In hoofdstuk 2 werden selectiecriteria voor resistentie en tolerantie geëvalueerd aan de hand van gegevens van drie veldexperimenten en een potexperiment uitgevoerd in Samanko met alle 10 genotypen. Zowel resistentie als tolerantie bleken belangrijke opbrengstbepalende eigenschappen van sorghum in Striga-besmette velden. Onder lage besmettingsniveaus bleek tolerantie relatief belangrijker voor opbrengst dan resistentie, terwijl resistentie meer bepalend was onder hoge besmettingsniveaus. Het oppervlak onder de Striga-aantallen ontwikkelings-curve (ASNPC) en de maximale bovengronds zichtbare Striga-aantallen (NSmax) waren onderscheidende en consistente criteria voor resistentie. Het ontwikkelen van geschikte criteria voor tolerantie bleek moeilijker. De belangrijkste oorzaken hiervan waren de bestaande verschillen in resistentie tussen genotypen en het gegeven dat de relatie tussen Striga-infectie en het relatieve opbrengstverlies niet lineair is. In het derde hoofdstuk is gepoogd de relatie tussen Striga-infectie en het relatieve opbrengstverlies op te helderen om zodoende een gepast selectiecriterium voor tolerantie tegen Striga te vinden. Deze studie was gebaseerd op gegevens van een potexperiment in Samanko (2003) en twee potexperimenten in de tropische kas in Wageningen (2003 en 2004) met CK60-B, E36-1, Framida en Tiémarifing. Er werden significante genotype-, besmettings- en genotype × besmetting effecten op de sorghumopbrengst gevonden. De relatie tussen besmettings- en infectieniveau bleek dichtheidsafhankelijk. Als gevolg hiervan resulteerde de brede reeks aan besmettingsniveaus in de experimenten van 2003 in een relatief smalle reeks aan infectieniveaus per ras. Hierdoor bleef een overlap in infectieniveaus, die een directe vergelijking tussen de vier genotypen mogelijk zou maken, uit. Om deze reden werd er in het experiment van 2004 gebruik gemaakt van een nog bredere reeks aan besmettingsniveaus, waardoor in ieder geval een smalle reeks van overeenkomstige infectieniveaus voor alle vier de genotypen werd verkregen. De relatie tussen Strigainfectieniveau en het relatieve opbrengstverlies bleek niet lineair. De resultaten 130 Samenvatting suggereren dat voor resistente genotypen tolerantie het best kan worden gekwantificeerd als een gereduceerd relatief opbrengstverlies per bovengronds Striga plant, terwijl voor minder resistente genotypen het maximale relatieve opbrengstverlies kan worden gebruikt. Of beide maten van tolerantie aan elkaar gerelateerd zijn blijft onduidelijk. Ondanks de vaststelling van deze tolerantiematen blijft selectie op tolerantie tegen Striga gebaseerd op de sorghum korrelopbrengst het nadeel houden dat Striga-vrije controlevelden vereist zijn voor een goede onderlinge vergelijking. In hoofdstuk 4 werden mogelijkheden voor het gebruik van fotosynthese, of eraan gerelateerde metingen, in de selectie voor Striga tolerantie onderzocht. Hierbij werd gebruik gemaakt van de gegevens van de twee potexperimenten in de kas in Wageningen, met een selectie van vier sorghumrassen (CK60-B, E36-1, Framida en Tiémarifing). Het doel van deze proef was het vinden van een beter selectiecriterium voor tolerantie tegen Striga, zonder verdere vereisten, zoals Striga-vrije controlevelden en meerdere Striga besmettingsniveaus. De CO2-assimilatiesnelheid van sorghum werd door Striga-infectie significant gereduceerd. Dit ging gepaard met een daling van de stomataire geleidbaarheid. Andere aan fotosynthese gerelateerde processen die de reductie in fotosynthese door Striga infectie konden verklaren waren de transpiratiesnelheid, fotochemische en niet-fotochemische uitdoving van PSII, electronentransport door PSII en de verhouding tussen fotosynthese en electronentransport. Bij gevoelige genotypen traden deze negatieve effecten reeds in een eerder stadium, in sterkere mate, en bij een lager besmettingsniveau op dan bij tolerante genotypen. Tiémarifing, eerder geïdentificeerd als tolerant, was het enige genotype waarbij voor geen van deze parameters een significant Striga-effect werd waargenomen. Hieruit valt af te leiden dat tolerantie voor Striga via fotosynthesegerelateerde metingen kan worden aangetoond. Met name fotochemische uitdoving en electronen transportsnelheid door fotosysteem II lijken hiervoor geschikt te zijn. Deze metingen maken het mogelijk om een selectie uit te voeren bij slechts één besmettingsniveau, zonder de verplichte aanwezigheid van Striga-vrije controleplanten. Selectie tussen opkomst van Striga en de bloei van sorghum en daarnaast op besmettingsniveaus van ten minste 300.000 levensvatbare Striga zaden per vierkante meter lijkt het meeste perspectief te bieden. Doelstelling van het in hoofdstuk 5 beschreven onderzoek was het kwantificeren van het effect van genotype op de Striga reproductie alsmede een geschikte selectiemaat te vinden voor Striga zaadproductie. Hiervoor werden gegevens van drie veldexperimenten en van alle 10 genotypen gebruikt. Er waren significante effecten van genotype en besmettingsniveau op bovengrondse Striga aantallen, bovengrondse Striga drooggewichten en Striga zaadproductie. Bovendien waren er significante correlaties tussen bovengrondse Striga aantallen en zaadproductie en zeer 131 Samenvatting significante correlaties tussen bovengrondse Striga drooggewichten en Striga zaadproductie. Drooggewichten van totale bovengrondse Striga biomassa en van Striga bloeiwijzen werden als goede indicatoren voor zaadproductie aangemerkt. In het veld resulteerde een toename in besmettingsniveau van Striga veelal in een proportionele toename in infectieniveau. De twee meest vatbare genotypen (CK60B en E36-1) vormden hierop een uitzondering. Bij deze rassen bleef het infectieniveau bij hogere besmettingsniveaus achter bij de verwachting. Een toename in besmettingsniveau leidde bij alle rassen tot een minder dan evenredige toename in bovengronds Striga drooggewicht (zowel totaal als alleen de bloeiwijzen) en zaadproductie. De relatie tussen Striga infectie en zaadproductie was naast dichtheidsafhankelijk ook afhankelijk van het sorghum genotype. De Striga reproductie ging in alle gevallen door na de oogst van het gewas en was, vooral bij lage besmettingsniveaus, aanzienlijk. Verschillen in Striga zaadproductie tussen besmettingsniveaus werden zodoende steeds kleiner in de periode tussen oogst en het einde van de Striga levenscyclus. Er waren geen significante effecten van sorghum genotype op Striga zaadgewicht per zaadcapsule. In deze studie zijn geschikte selectiecriteria gevonden voor resistentie van de gastheer tegen Striga parasitisme, tegen Striga reproductie alsmede voor tolerantie van het gewas die optreedt na Striga infectie. Het maximum aantal bovengrondse Striga planten is een betrouwbaar selectiecriterium voor resistentie, terwijl het drooggewicht van Striga bloeiwijzen kan worden gebruikt om genotypen te identificeren die weinig Striga reproductie toestaan. Het vinden van een goed selectiecriterium voor tolerantie bleek het meest gecompliceerd. Voor vatbare genotypen is het hoogst verkregen relatieve opbrengstverlies een geschikt selectiecriterium voor tolerantie. Voor meer resistente genotypen lijkt het relatieve opbrengstverlies per opgekomen Striga plant beter geschikt als tolerantie criterium. Echter, als de selectie gebaseerd wordt op de opbrengst van de gastheerplant, zijn Striga-vrije controles onmisbaar voor selectie van tolerante genotypen. Een geschikt alternatief selectiecriterium voor tolerantie bleek te bestaan uit waarden afgeleid van chlorofylfluorescentiemetingen en dan in het bijzonder de fotochemische uitdoving en de electronentransportsnelheid. Een groot voordeel van het gebruik van chlorofylfluorescentiemetingen is dat tolerantie kan worden geïdentificeerd zonder de aanwezigheid van Striga-vrije controlevelden. De sorghum rassenkeuze beïnvloedt de reproductie van Striga. Verschillen in Striga zaadproductie tussen sorghumrassen kunnen voor een groot deel toe worden geschreven aan verschillen in resistentie. In dit onderzoek resulteerde het gebruik van resistente genotypen in een reductie in Striga zaadproductie van 70 tot 93% ten opzichte van vatbare genotypen. Echter, om de Striga zaadbank te reduceren zijn additionele maatregelen nodig zoals handmatig wieden voor de oogst. 132 Résumé Striga hermonthica (Del. Benth) est une plante adventice parasite des céréales tropicales, telles que le sorgho (Sorghum bicolor [L.]), le mil (Pennisetum glaucum [L.] R. Br. ou P. americanum [L] K. Schum), le maïs (Zea mays [L]) et le riz (Oryza glaberrima [Steudel] et O. sativa [L]). Elle réduit sérieusement les rendements des cultures et constitue par conséquent l’une des principales contraintes biotiques à la production des cultures dans les zones tropicales semi-arides, fragilisant les moyens de subsistance des petits paysans. Le Striga pénalise son hôte en lui prélevant des assimilats, de l’eau et des nutriments et en modifiant l’équilibre des phytohormones qui à son tour modifie l’allométrie (ratio feuille/tige) de la plante hôte et réduit les taux de photosynthèse. Plusieurs moyens de lutte ont été proposés, mais aucun n’est tout à la fois suffisamment efficace, pratique et accessible pour les paysans à faible revenue. L’association de plusieurs moyens de lutte semble être la meilleure approche pour combattre le Striga. Dans la gestion intégrée du Striga, le rôle des variétés améliorées, résistantes et tolérantes, est indispensable. Pour le développement de ces variétés améliorées, l’identification et le choix de matériel végétal performant est d’une importance vitale. La présente étude s’est focalisée sur les effets du Striga sur la physiologie et la production de l’hôte ainsi que les effets du génotype de l’hôte sur le taux d’infection et de reproduction du Striga. Tout ceci dans le but de mieux apprécier leur interaction et de développer des outils appropriés pour la sélection au champ de matériel végétal résistant et tolérant. Entre 2001 et 2004, trois expérimentations au champ et deux cultures en pots au Mali et deux cultures en pots en serre aux Pays-Bas ont été réalisées. Dans les trois expérimentations au champ, 10 différents génotypes de sorgho (CK60-B, CMDT39, E36-1, Framida, IS9830, N13, Seredo, Serena, SRN39 et Tiémarifing) ont été cultivés en présence et en absence de graines de Striga, et pour la dernière année les parcelles de Striga ont été infestées à deux niveaux (‘bas’ et ‘élevé’). En 2001 et 2003, des champs préalablement infestés ont été utilisés et des parcelles indemnes de Striga ont été obtenues à la suite d’injection de gaz éthylène à l’aide d’un injecteur portable de, tandis que les parcelles déjà infestées ont bénéficié d’apports complémentaires de graines. En 2002, dans un champ non contaminé une infestation artificielle contrôlée a permis d’installer des parcelles Striga tout en évitant de contaminer les parcelles indemnes. De diguettes entourant les différentes parcelles ont empêché les parcelles indemnes d’être contaminées par les graines de Striga provenant (via le ruissellement de l’eau de pluie) des parcelles infestées avoisinantes. Dans trois des quatre cultures en pots, une sélection de quatre des dix génotypes (CK60-B, E36-1, Framida et Tiémarifing) a été cultivée sous différents degrés d’infestation par le Striga. Chaque 133 Résumé année, les semences de sorgho ont été multipliées par autofécondation pour être utilisées l’année suivante. Les graines de Striga utilisées pour ces études ont été collectées sur des plants de sorgho cultivés à Samanko (Mali). Dans toutes les expérimentations, des observations et mesures de routine ont été faites: comptage de levée des plantes de Striga, poids de matière sèche du Striga et rendement en grains du sorgho. Selon l’expérimentation, des observations plus détaillées ont été réalisées: la photosynthèse et la fluorescence chlorophyllienne des plants de sorgho et les dates de floraisons, le poids de matière sèche des hampes florales de plantes de Striga ainsi que le nombre des plants en phase de reproduction et la production de graines de Striga par comptage de capsules de graines. Dans la première étude de cette thèse (chapitre 2), des critères de sélection au champ pour la résistance et la tolérance ont été évalués et comparés sur la base des résultats des trois expérimentations au champ et d’une culture en pot au Mali avec tous les dix génotypes. La résistance comme la tolérance du sorgho sont des déterminants importants du rendement dans le cas d’infestation de Striga. En condition de faible infestation, la tolérance est relativement plus importante que la résistance pour le rendement, tandis que la résistance est plus importante dans le cas de fortes infestations. L’ASNPC (la surface en dessous de la courbe de progression du nombre de plants de Striga) ainsi que le NSmax (le nombre maximum de plants de Striga émergé) étaient discriminants et fiables pour les critères de résistance. Il a été plus difficile de trouver un critère satisfaisant pour la tolérance. Les différences de résistance entre les génotypes et la relation non linéaire entre l’infection par le Striga et la perte de rendement en sont les principales explications. Le chapitre 3, avait pour objectif de démontrer la relation entre l’infection par le Striga et la perte de rendement pour obtenir un critère approprié de sélection pour la tolérance au Striga. Les données de trois expérimentations en pots, une au Mali (2003) et deux en serre aux Pays-Bas (2003 et 2004) avec CK60-B, E36-1, Framida et Tiémarifing, ont été utilisées à cet effet. Il y avait des effets significatifs sur le rendement du sorgho pour les génotypes, pour les infestations et pour l’interaction génotype × infestation. La relation entre le niveau d’infestation et le niveau d’infection était dépendant de la densité. Par conséquent, la large gamme des niveaux d’infestation n’a entraîné qu’une petite gamme de niveau d’infection pour les quatre génotypes dans les expérimentations de 2003, sans avoir des infections identiques pour tout les génotypes. L’expérimentation de 2004 a été conçue pour obtenir au moins une petite gamme d’infection identique pour tous les quatre génotypes. La relation niveau d’infection par le Striga et perte relative de rendement n’était pas linéaire. Les résultats montrent que pour les génotypes résistants, la tolérance pouvait être mieux évalué par la diminution de la perte relative de rendement reliée au nombre de plants de Striga 134 Résumé emergés, tandis que pour les génotypes moins résistants, la perte maximale de rendement était le meilleur indicateur. L’hypothèse selon laquelle les deux expressions de tolérance étaient liées n’est pas confirmée. Il a été conclu qu’un critère de tolérance sur la base du rendement nécessite toujours la présence de témoins. Le chapitre 4 explore les options d’utilisation de la photosynthèse ou des mesures s’y rapportant dans le criblage pour la tolérance au Striga hermonthica. Ce travail était basé sur les résultats des deux cultures en pots conduits aux Pays-Bas avec CK60-B, E36-1, Framida et Tiémarifing. L’objectif était de trouver un meilleur critère de criblage pour la tolérance avec moins d’exigences, telles la présence de parcelles témoins et de différents niveaux d’infestation. L’assimilation du CO2 du sorgho a été réduit significativement par l’infection du Striga. Cette diminution était accompagnée par une diminution de la conductance stomatique qui toutefois n’était pas la raison principale de la diminution de l’assimilation du CO2. Les autres processus importants affectés par Striga étaient le taux de transpiration, l’extinction photochimique et non photochimique, le transport d’électrons et la photosynthèse par transport d’électrons. Les génotypes sensibles ont été affectés plus tôt et à des niveaux d’infestation plus faibles que les génotypes tolérants. Tiémarifing, identifiée auparavant comme la variété la plus tolérante, était le seul génotype ne montrant aucun effet de Striga significatif pour tout les paramètres mesurés à différentes dates. En conséquence, les génotypes tolérants au Striga peuvent être détectés par des mesures de photosynthèse. Les mesures de l’extinction photochimique et du transport d’électron au photosystème II semblent particulièrement être adaptées à cet objectif. Elles facilitent le criblage à une période donnée, à un niveau d’infestation unique et sans parcelles témoins exemptes de Striga. Il a été recommandé d’effectuer le criblage entre la première émergence de Striga et la floraison du sorgho et à des niveaux d’infestation d’au moins 300 000 graines viables de Striga par m2. L’objectif du chapitre 5 était d’étudier les effets des génotypes sur la reproduction du Striga et de trouver un critère approprié de sélection pour la production de graines de Striga. Les données utilisées pour cette étude proviennent des trois expérimentations au champ au Mali avec tous les dix génotypes. Il y avait des effets significatifs des génotypes et de niveau d’infestation sur le nombre de Striga émergés du sol, sur les poids de matière sèche de Striga émergés et sur la production de graines de Striga. Il y avait également des corrélations significatives entre le nombre de Striga émergés et la production de graines et des corrélations très significatives entre le poids de matière sèche de Striga émergés et la production de graines de Striga. Les poids de matière sèche de Striga émergé du sol et de hampes florales de Striga sont apparus comme de bons indicateurs pour la reproduction du Striga. 135 Résumé Généralement une augmentation du niveau d’infestation au champ augmente proportionnellement le niveau d’infection (le nombre de plants de Striga émergés de la terre), exception faite pour les génotypes les plus susceptibles (CK60-B et E36-1), pour lesquels les hauts niveaux d’infestation ont conduit à une augmentation non proportionnelle du niveau d’infection. Une augmentation du niveau d’infestation a conduit à une augmentation non proportionnelle du poids de matière sèche de Striga (total et hampe florale) et de la production de graines pour tous les génotypes. La relation entre l’infection par le Striga et la production de graines semble être dépendant à la fois de la densité et du génotype. La reproduction du Striga a continué après la récolte de la culture. Les écarts de production de graines de Striga entre les deux niveaux d’infestation par le Striga ont baissé entre la récolte et la fin de cycle du Striga. Il n’y avait pas d’effet significatif des génotypes sur le poids sec des graines par capsule. Dans cette étude, des critères appropriés de criblage ont été trouvés pour la résistance de la plante hôte au parasitisme et à la reproduction du Striga ainsi que pour la tolérance à l’infection par le Striga. Le nombre maximum de Striga émergés du sol est un critère fiable de criblage pour la résistance. Le poids de matière sèche des hampes florales de Striga peut être utilisé pour identifier les génotypes qui supportent une faible reproduction du Striga. Comme le montre cette étude, le criblage de la tolérance est plus compliqué. Pour les génotypes susceptibles, la perte maximale relative de rendement semble être un critère de criblage approprié pour la tolérance. Pour les génotypes plus résistants, la perte relative de rendement par infection de Striga semble plus adéquate comme mesure de tolérance. Les parcelles témoins sans Striga sont indispensables pour la sélection des génotypes tolérants lorsque la sélection est basée sur les composantes du rendement de l’hôte. Un critère alternatif fiable pour la tolérance semble être les mesures de fluorescence chlorophyllienne comme l’extinction photochimique (Pq) ou le transport d’électron (ETR). Avec l’utilisation des mesures de la fluorescence chlorophyllienne, la tolérance peut être potentiellement identifiée en absence de témoins sans Striga. Le choix des génotypes affecte les efforts de reproduction du Striga. De façon générale les écarts de production de graines de Striga entre génotypes de sorgho peuvent être expliqués par les écarts de résistance de la plante hôte. La résistance était responsable de 70 à 93 % de la réduction de la production de graines de Striga dans le cadre de cette étude. Pour réduire la banque des graines de Striga dans le sol, des moyens supplémentaires de lutte telles que le sarclage manuel avant la récolte, sont nécessaires. 136 Publications Rodenburg, J., A. Stein, M. van Noordwijk, Q. M. Ketterings, 2003. Spatial variability of soil pH and phosphorus in relation to soil run-off following slash-and-burn land clearing in Sumatra, Indonesia. Soil and Tillage Research 71, 1-14. Rodenburg, J., A. van Ast, L. Bastiaans, D.E. Hess and M. J. Kropff, 2002. Defence mechanisms in sorghum against Striga hermonthica. In: Bastiaans, L., D. T. Baumann, S. Christensen, P.E. Hatcher, P. Kudsk, A. C. Grundy, E. J. P. Marshall, J. C. Streibig and F. Tei (eds). EWRS 12th Symposium, Wageningen, The Netherlands, pp. 388-389. Rodenburg, J., L. Bastiaans, E. Weltzien Rattunde and D. E. Hess, 2004. Yielding ability, resistance and tolerance as independent selection criteria for breeding against Striga. In: Rubin, B., P. J. Pieterse and C. Reinhardt. 4th International Weed Science Congress, Durban, South Africa, pp. 124. Rodenburg, J., L. Bastiaans, E. Weltzien, D. E. Hess, 2005. How can field selection for Striga resistance and tolerance in sorghum be improved? Field Crops Research 93, 34-50. Rodenburg, J., L. Bastiaans, M. J. Kropff, Accepted. Characterization of host tolerance to Striga hermonthica. Euphytica. 137 Curriculum vitae Jonne Rodenburg was born on July 9, 1975 in Hilversum, The Netherlands. He attended the Stichtse Vrije School in Zeist as secondary school and obtained his VWO diploma at the Teisterbant College in Culemborg in June 1994. In September 1994 he started his MSc studies Tropical land use at Wageningen Agricultural University (WAU). He specialized in agronomy and soil fertility and graduated at Wageningen University (WU) in 1999. For his studies he conducted on-farm research on the use and effects of compost and rock phosphate in sorghum production systems in Burkina Faso for a rural development project financed by the Dutch Directorate-General for International Cooperation (DGIS), in Koudougou. In Indonesia he studied the effects of slash-and-burn practices on soil erosion and soil fertility in rice-rubber agro-forests in Jambi, Sumatra, under supervision of Meine van Noordwijk (ICRAF, Bogor, Indonesia), Bert H. Janssen and Nico de Ridder (WAU, Wageningen, The Netherlands). After his graduation he worked on temporary basis at the research station of Boomteelt Praktijkonderzoek, in Horst, Limburg. In the fall of 2000 he started his PhD research at Wageningen University within the Group of Crop and Weed Ecology on defence mechanisms in sorghum against the parasitic weed Striga hermonthica. During three cropping seasons he collected data for his PhD thesis at the research station of the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), in Mali. He is currently working at Africa Rice Center (WARDA) in Cotonou, Benin, as agronomist in the Inland Valley Consortium. This APO position is financed by DGIS. 138