American Journal of Botany 97(3): 493–510. 2010.
ARE SPURRED CYATHIA A KEY INNOVATION?
MOLECULAR SYSTEMATICS AND TRAIT EVOLUTION IN
THE SLIPPER SPURGES (PEDILANTHUS CLADE: EUPHORBIA ,
EUPHORBIACEAE)1
N. Ivalú Cacho2,6, Paul E. Berry3, Mark E. Olson4, Victor W. Steinmann5, and
David A. Baum2,6
2Department
of Botany, University of Wisconsin-Madison, 430 Lincoln Drive, Madison, Wisconsin 53706 USA; 3Department of
Ecology and Evolutionary Biology, University of Michigan, 830 N. University, Ann Arbor, Michigan 48109 USA; 4Instituto de
Biología, Universidad Nacional Autónoma de México, Departamento de Botánica, Tercer Circuito s/n, Ciudad Universitaria,
Copilco, Coyoacán A.P. 70-367, México, Distrito Federal, C.P. 04510 México; and 5Instituto de Ecología, A.C., Centro
Regional del Bajío, Av. Lázaro Cárdenas 253, A.P. 386 61600 Pátzcuaro, Michoacán, México
The study of traits that play a key role in promoting diversification is central to evolutionary biology. Floral nectar spurs are
among the few plant traits that correlate with an enhanced rate of diversification, supporting the claim that they are key innovations. Slight changes in spur morphology could confer some degree of premating isolation, explaining why clades with spurs tend
to include more species than their spurless close relatives. We explored whether the cyathial nectar spur of the Pedilanthus clade
(Euphorbia) may also function as a key innovation. We estimated the phylogeny of the Pedilanthus clade using one plastid (matK)
and three nuclear regions (ITS and two G3pdh loci) and used our results and a Yule model of diversification to test the hypothesis
that the cyathial spur correlates with an increased diversification rate. We found a lack of statistical support for the key innovation
hypothesis unless specific assumptions regarding the phylogeny apply. However, the young age (hence small size) of the group
may limit our ability to detect a significant increase in diversification rate. Additionally, our results confirm previous species
designations, indicate higher homoplasy in cyathial than in vegetative features, and suggest a possible Central American origin of
the group.
Key words: cyathium; diversification rate; Euphorbia; Euphorbiaceae; G3pdh; key innovation; matK; nectar spur;
Pedilanthus; phylogeny.
Understanding the factors that promote disparities in the rate
of diversification among lineages is central to evolutionary
biology. The concept of key innovation was used by Simpson
(1953) to refer to a trait or group of traits that allow a lineage to
occupy a new adaptive zone. Because the occupation of a novel
adaptive zone tends to promote diversification and the accumulation of more species, the term key innovation has come to
refer to traits that contribute to an increase in the intrinsic species diversification rate of a taxon (Hunter, 1998; Galis, 2001;
Ree, 2005; Kay et al., 2006). Under this revised definition, the
hypothesis that a trait is a key innovation is ideally supported by
three kinds of evidence (Hunter, 1998; Galis, 2001, and references therein): (1) The taxon having the trait has a higher rate
of diversification than closely related taxa lacking the trait, (2)
there is a reasonable ecological or functional model to justify a
1
causal link between the trait and increased diversity, and (3)
analogous traits are consistently associated with increased
diversification rates. The inherent association between floral
traits and reproduction in angiosperms has led evolutionary
biologists to focus on these traits as key players in the differential
diversification of clades of flowering plants (Kay et al., 2006).
In this paper, we explore whether the unusual spurred inflorescences that characterize the Pedilanthus clade of Euphorbia
L. have played a key role in increasing its diversification rate
compared to close relatives.
Among the few plant traits that have been studied carefully
and have been shown to meet all three criteria for key innovations are floral nectar spurs (Hodges, 1997; Ree, 2005; Kay
et al., 2006). Floral spurs are formally defined as hollow, slender,
sac-like appendages of a perianth organ, typically containing
nectar (Harris and Harris, 2001; Neilson et al., 1950). The floral
spurs of Aquilegia L. are a thoroughly studied example of this
morphological trait. It has been shown that the rate of diversification is higher in Aquilegia than in closely related, spurless
Ranunculaceae (Hodges and Arnold, 1995). Additionally, a
plausible causal model of how nectar spurs could promote
increased diversification has been proposed: nectar spurs allow
for the distance between the floral reward and the reproductive
organs to evolve rapidly without concomitant changes in the
floral reproductive organs themselves, thereby increasing the
rate at which lineages develop premating isolation (Hodges,
1997). It has been suggested that nectar spurs could promote
reproductive isolation, and therefore the rate of speciation,
Manuscript received 31 March 2009; revision accepted 18 December 2009.
The authors thank Trinidad and Martha G. Pérez, J. Ma. Cárdenas, J.
Luis T. Sánchez, and Raymundo and Vicky Ramírez for valuable help
during fieldwork; Cécile Ané, Johanne Brunet, Bret Larget, and Kenneth
Sytsma, for helpful discussion; and Kandis Elliot for assistance with
illustrations. Funding was provided by the National Science Foundation
through a Doctoral Dissertation Improvement Grant (DEB-0608428
to N.I.C. and D.A.B.), and a Planetary Biological Inventory Grant
(DEB-0616533 to P.E.B. and D.A.B.).
6 Authors
for correspondence (e-mail: ivalu.cacho@gmail.com,
dbaum@wisc.edu)
doi:10.3732/ajb.0900090
American Journal of Botany 97(3): 493–510, 2010; http://www.amjbot.org/ © 2010 Botanical Society of America
493
494
American Journal of Botany
through either specialization on different pollinators or differential placement of pollen on the same pollinators. In support
of this inference, there is a statistically significant correlation
between floral spurs and increased species number in multiple
spurred/spurless sister clades (Hodges, 1997; Kay et al., 2006).
Sixteen independent origins of nectar spurs have been documented so far, 12 of which are associated with clades that are
more species rich than their sister clades (Kay et al., 2006).
Until now, studies have focused primarily on florally derived
nectar spurs. However, floral organs are not unique in their ability to produce nectar, and in some taxa, extrafloral nectaries
have been shown to play a role in pollination (e.g., in Euphorbia
L., Acacia L., and Marcgraviaceae). Of the extrafloral nectaries
that function in pollination, to our knowledge, only one clade
produces an extrafloral nectar spur: the Pedilanthus clade of
Euphorbia. Like all Euphorbia, the Pedilanthus clade has reduced
flowers organized in a specialized pseudanthial inflorescence,
the cyathium. Members of the Pedilanthus clade are unusual in
having a strongly zygomorphic cyathium with a nectar-containing
spur that is derived from the fusion of petaloid appendages of
nectar glands associated with an inflorescence involucre (Fig. 1).
To reach the nectar reward, pollinators (primarily hummingbirds; N. I. Cacho, personal observation; Dressler, 1957) probe
the cyathial spurs and in so doing contact the staminate or pistillate flowers. Thus, these cyathial spurs in the Pedilanthus clade
have an analogous function to floral nectar spurs, but develop at
a distinct level of organization, the inflorescence rather than the
flower. Therefore, they present a unique opportunity to determine
whether the pattern of increased diversity in clades associated
with nectar spurs is limited to floral spurs or could extend to all
nectar spurs that function in plant–pollinator interactions.
The 15 species that comprise the Pedilanthus clade are characterized by zygomorphic cyathia that usually resemble slippers, as reflected in their many common names (slipper spurges,
zapatitos, and queen’s slipper) and also account for the scientific name of the genus to which these species were traditionally
assigned: Pedilanthus Necker ex Poit. (“foot flower”). However, molecular phylogenetic research has shown that the slipper spurges form a clade that is embedded within Euphorbia
sensu lato (Steinmann and Porter, 2002). Based on this result,
all species names have been transferred from Pedilanthus to
Euphorbia (Steinmann, 2003).
[Vol. 97
The Pedilanthus clade exhibits great morphological and ecological diversity for a group of its size (Fig. 2). Habit ranges
from succulent leafless shrubs about a meter in height [e.g.,
E. cymbifera (Schltdl.) V.W.Steinm.], to evergreen treelets a few
meters tall [e.g., E. finkii (Boiss.) V.W.Steinm., E. peritropoides
(Millsp.) V.W.Steinm.], to deciduous trees up to 8 m tall [e.g.,
E. coalcomanensis (Croizat) V.W.Steinm.]. Slipper spurges
occur in diverse habitats, including mesic tropical forests, dry
deciduous forests, and true deserts such as the Sonoran Desert
or the Tehuacán Desert in central Mexico. Leaf size, shape,
persistence, and indumentum are all variable. Some species
produce tuberous roots, adventitious root buds, or rhizomatous
stems, whereas other species lack any obvious adaptations for
vegetative reproduction or perennation (Dressler, 1957; N.
I. Cacho, personal observations). There is abundant variation
in cyathium size and color pattern, spur elongation and coloration, and cyathium bract morphology and phenology (Fig. 3).
This variation may correlate with different pollination systems
to some extent. Most species of the group are thought to be
hummingbird-pollinated (Dressler, 1957), but E. diazlunana
(J.Lomelí & Sahagún) V.W.Steinm. has been reported to be
pollinated by hymenopterans (Sahagún-Godínez and LomelíSención, 1997) and, based on its morphology, E. tehuacana
(Brandegee) V.W.Steinm. is also likely to be insect-pollinated,
although formal pollination studies in the group are lacking.
Twelve of the fifteen species in the Pedilanthus clade are
restricted to Mexico. The Mexican species vary in their geographical range from Euphorbia lomelii V.W.Steinm., estimated
to occupy some 300 000 km2 in the deserts around the Gulf of
California, to the microendemic E. conzattii V.W.Steinm. with
a range of 0.2 km2 on a single mountaintop (Olson et al., 2005).
Of the three species with distributions extending beyond Mexico’s borders, the southernmost populations of E. calcarata
(Schltdl.) V.W.Steinm. occur in northern Guatemala, whereas
E. personata (Croizat) V.W.Steinm. has disjunct populations as
far south as Costa Rica. Finally, E. tithymaloides L., by far the
most widespread species of the clade, has a range that includes
Mexico, Florida, northern South America, Central America,
and most islands in the Caribbean. In addition to an unusually
broad distribution, E. tithymaloides is also notable in the group
for the degree of infraspecific differentiation, with eight
subspecies recognized (Dressler, 1957).
Fig. 1. Longitudinal midsection of the spurred zygomorphic cyathium of Euphorbia tithymaloides subsp. padifolia. The terminal pistillate flower,
single staminate flowers, involucral tube, nectar glands, and spur concealing nectar glands are indicated.
March 2010]
Cacho et al.— Are spurred cyathia a key innovation?
495
Fig. 2. Diversity in habit in the Pedilanthus clade. (A) Euphorbia diazlunana. (B) E. tehuacana. (C) E. personata. (D) E. lomelii. (E) E. finkii.
(F) Glossy leaves of E. finkii. (G) E. calcarata. (H) Root of E. calcarata. (I) Yellow latex of E. diazlunana. (J) White latex of E. peritropoides. (K) E. cyri.
(L) E. tithymaloides. (M) E. peritropoides. (N) E. conzattii. (O) E. coalcomanensis during the dry season. (P) Canopy of E. coalcomanensis during rainy
season. (Q) Leaves of E. coalcomanensis.
496
American Journal of Botany
[Vol. 97
Fig. 3. Diversity in reproductive morphology in the Pedilanthus clade and some close relatives. (A) E. bracteata, in cultivation. (B) E. bracteata
in the wild; note bract and fruit color. (C) E. personata, inflorescence cluster. (D) E. personata, fruit. (E) E. diazlunana. (F) E. diazlunana, in cultivation.
(G) E. tithymaloides subsp. tithymaloides fruit. (H) E. tithymaloides subsp. padifolia. (I) E. lomelii, image by T. B. Kinsey (http://www.fireflyforest.net/
firefly/), reproduced with the author’s permission. (J) Persistent, dry inflorescence of E. cyri; note large and persistent bracts. (K) E. calcarata. (L) E. finkii
from herbarium material. (M) E. colligata. (N) E. cymbifera. (O) E. conzattii. (P) E. peritropoides; note peduncles. (Q) Persistent, dry inflorescence
of E. coalcomanensis. (R) E. coalcomanensis; note bright bracts. (S) Gland of E. umbelliformis, image by P. E. Berry. (T) Cyathia of E. umbelliformis, image
by P. E. Berry (U) E. gollmeriana. (V) E. pteroneura. (W) E. leucocephala.
March 2010]
Cacho et al.— Are spurred cyathia a key innovation?
Prior phylogenetic hypotheses for the Pedilanthus clade are
based exclusively on morphological characters (Dressler, 1957)
or considered only a few species (Steinmann and Porter, 2002);
and both are poorly resolved. In recent years, new species have
been described (Dressler and Sacamano, 1992; SahagúnGodínez and Lomelí-Sención, 1997), and extant populations of
several others have been located (Lomelí-Sención and SahagúnGodínez, 2002; Olson et al., 2005), giving us the opportunity
to conduct a comprehensive phylogenetic study in this group.
In this paper, we provide the first detailed molecular phylogenetic analysis of the Pedilanthus clade and use the results to
test the hypothesis that the cyathial spur, like the floral spur of
Aquilegia, is a key innovation. We also use our phylogenetic
results to study trait evolution in this small yet morphologically
and ecologically diverse group, to assess the exclusivity of
some traditionally named species, to test Dressler’s (1957)
hypothesized species groups, and to test the hypothesized
Mexican origin of the group (Dressler, 1957).
MATERIALS AND METHODS
Taxon sampling—Plant material was collected in the wild for 14 of the 15
described species of the Pedilanthus clade. We were unable to locate Euphorbia
dressleri V.W.Steinm. despite intensive fieldwork in and around the sole locality from which it has been reported. To our knowledge, there are no cultivated
individuals of this species. We therefore think that this species has likely
become extinct. For the remaining species of the Pedilanthus clade, representatives of multiple populations per species (up to six) were included when possible
(see Appendix 1 for taxa included in this study).
We sequenced selected outgroups to represent three of the four major clades
of Euphorbia, as defined by Steinmann and Porter (2002): clade B (E. esula L.,
E. cyparissias L.), clade C [E. umbelliformis (Urb. & Ekman) V.W.Steinm. &
P.E.Berry, Euphorbia pteroneura A.Berger, E. gollmeriana Klotzsch ex Boiss.,
E. milii Des Moul., Euphorbia umbellata (Pax) Bruyns], and clade D [E. leucocephala Lotsy, E. heterophylla L., E. oerstediana (Klotzsch & Garcke) Boiss.].
Representatives of eight outgroups were collected in the field, two (E. milii,
Manihot esculenta Crantz) were collected from cultivation, and sequences for
other Euphorbia outgroup taxa were downloaded from GenBank (see Appendix
1 for detailed information).
DNA extraction, PCR amplification, and sequencing—Genomic DNA was
extracted from plant tissue dried in silica gel (Chase and Hillis, 1991) using
either the CTAB method as outlined by Doyle and Doyle (1987) or using the
DNeasy Plant Mini Kit (Qiagen, Valencia, California, USA). The internal transcribed spacer region (ITS), consisting of ITS1 and ITS2 and the 5.8S ribosomal nuclear gene, was amplified with primers ITS1 and ITS4 (White el al.,
1990; Appendix S1, see Supplemental Data with the online version of this
article), and sequenced with the ITS1, ITS2, ITS3, and ITS4 primers. About
1500 bp of the matK sequence were amplified with primers trnK3914F
(Johnson and Soltis, 1994) and p6R (Nyffeler et al., 2005). Due to the presence
of extremely AT-rich regions, six additional newly designed sequencing primers
were used (Appendix S1, see online Supplemental Data). The glyceraldehyde
3-phosphate dehydrogenase-subunit C (G3pdhC) gene was initially amplified
with the primers GPDX7F and GPDX9F (Strand et al., 1997), gel purified with
QIAquick Gel Extraction Kit (Qiagen, Valencia, California, USA), ligated into
a pGEM T-Vector (Promega, Madison, Wisconsin, USA), cloned in E. coli DHB5α competent cells (Invitrogen, Carlsbad, California, USA), reamplified, and
sequenced. Two loci were separated by this cloning procedure, here referred to
as G3pdhC-A and G3pdhC-B. Locus-specific primers were designed and used
for amplification, cloning when necessary (eight clones per accession, same
procedure as already outlined), and sequencing (AF1/AR1 and BF1/BR1 in
online Appendix S1).
Most PCR reactions included 2.5 µL of M891A PCR-Buffer (Promega), 2.5
µL of 25mM MgCl2, 0.5 µL dNTP mix (2.5 mM each), 0.5 µL of each primer
(10 µM), and 0.625 units of Flexi-Taq (Promega) in a 25-µL reaction. The
amplification of matK required the use of buffer M890A (Promega) and BSA
(0.8%). For ITS and G3pdhC-A/B, the cycling conditions consisted of an initial
denaturation at 94°C for 10 min, followed by a three-cycle touchdown decreasing
497
2° per cycle (94°C for 30 s, 58°/56°/54°C for 60 s, 72°C for 90 s); 31 additional
cycles of 94°C for 30 s, 54°C for 60 s, 72°C for 90 s; and, a final extension of 7
min at 72°C. The cycling conditions for matK were initial denaturation at 95°C
for 5 min; 30 cycles of 94°C for 50 s, 52°C for 70 s, 72°C for 90 s; and a final
extension of 5 min at 72°C. PCR products were cleaned and diluted using
Ampure Magnetic Beads (Agencourt Biosciences, Beverly, Massachusetts,
USA) following the manufacturer’s protocol.
Sequencing reactions consisted of 0.5 µL of BigDye Terminator v. 3.1 mix
(Applied Biosystems), 2.0 µL of 5× dilution buffer (Applied Biosystems), 5
pmol of primer, dmso (10%), and ~0.2 µg of template DNA in a final reaction
volume of 10 µL. Cycle conditions consisted of an initial denaturation at 95°C
for 3 min; 50 cycles of 96°C for 10 s, 58°C for 4 min; and a final extension
of 7 min at 72°C. Excess dye terminators were removed using the CleanSeq
magnetic bead sequencing reaction clean up kit (Agencourt Biosciences).
Samples were electrophoresed on an Applied Biosystems 3730xl automated
DNA sequencing instrument, using 50-cm capillary arrays and POP-7 polymer,
at the University of Wisconsin-Madison Sequencing facility.
Sequences were assembled and edited in the program Sequencher v. 4.7
(Gene Codes Corp., Ann Arbor, Michigan, USA), and manually aligned in
the program MaClade v. 4.05 (Maddison and Maddison, 2002). In general,
alignments were unambiguous, at least within the ingroup. Where alternative
alignments that invoked a similar number of indel and substitution events could
be identified, we selected the one that minimized the number of parsimony
informative characters that were generated. For both loci of G3pdhC, a minimum
of eight clones per individual was examined. When more than one allele was
recovered from an individual, from five to eight additional clones were sequenced.
PCR error and PCR recombination were assessed by manual examination of the
sequences; potential PCR recombinants were excluded from the analyses.
Phylogenetic analyses—We examined our data sets for phylogenetic signal
using g1 statistics and permutation tail probability tests (PTP) as implemented
in the program PAUP* v. 4.0b10 (Swofford, 2002). Data sets were then analyzed
individually and in combination. Only accessions with data for three or more
partitions were included in combined analyses. Congruence among data partitions was explored in a parsimony framework using the incongruence length
difference (ILD) test (Farris et al., 1994) as implemented in PAUP*. Sources of
conflict were identified by deletion of potentially conflicting taxa (based on
examination of individual gene topologies). For ILD analyses, 10 000 replicates
of flat-weighted parsimony heuristic searches were conducted, with 10 random
additions, holding 10 trees per step, tree-bisection-reconnection (TBR) branch
swapping, and saving no more than one tree per replicate. Data sets that showed
no evidence of conflict were concatenated for combined analysis.
As many as six accessions per species were included in both individual
and combined data sets to assess species monophyly. We conducted a set
of combined analysis that retained all individuals, which we will refer to as
“combined-all”. There were only three cases in which distinct alleles were
recovered from a single, presumably heterozygous, individual (E. diazlunana,
E. calcarata_01, and E. colligata V.W.Steinm._02 for G3pdhC-A). We also
performed a final combined analysis with a single accession per species
(“combined-one”). For this combined analysis, we decided which alleles to use
by conducting preliminary parsimony searches with all possible combinations
of alleles and selected the set of alleles that yielded the shortest trees.
Maximum parsimony (MP)—For both separate and combined data sets,
flat-weighted MP heuristic searches were performed in PAUP* v. 4.0b10
(Swofford, 2002). Starting trees were obtained by 10 000 random addition replicates holding 10 trees per step and keeping best trees only. Searches used the
TBR branching swapping algorithm and saved only one tree per replicate. A
second search was run to completion starting from the set of most-parsimonious
trees and swapping (TBR). Clade support was assessed by 10 000 bootstrap
replicates as implemented in PAUP* with the following search settings: 10
random addition replicates, hold = 1, keep = best, TBR, not more than one
tree saved per replicate.
Model selection—An appropriate and not overly complex model of molecular
evolution was selected under a decision theory framework as implemented in
the program DT-ModSel (Minin et al., 2003). We modified the code to evaluate
alternative models on a most parsimonious tree rather than a neighbor-joining
tree. We only considered models that account for site-to-site rate heterogeneity
using a discrete approximation to a gamma distribution of rates (Gamma) rather
than those that also allow for a fixed proportion of invariant sites (P-invar), as
recommended by Stamatakis et al. (2008; RAxML v. 704 manual).
American Journal of Botany
498
[Vol. 97
Table 1.
Data sets analyzed, including matrices’ dimensions, models implemented, and trees obtained. Character status is in reference to the aligned
matrices (Cte = constant; PIC = parsimony informative characters; NPIC = nonparsimony informative characters). Number of most parsimonious
trees (MPTs), their length (L), consistency (CI), retention (RI), and rescaled consistency (RC) indices are provided for the maximum parsimony
(MP) analyses. For the maximum likelihood (ML) analyses, the model, program used, and optimal likelihood score are provided. For ML analyses
with RAxML, the numbers in parenthesis for the combined analyses refer to the number of partitions considered. The likelihood score provided for
Bayesian analysis corresponds to the likelihood of best state for “cold” chain of run 1; sf = average split frequencies.
Data set
Generals
MP
ML (Garli)
ML (RAxML)
MrBayes
G3pdhC-A
Outgroup: E. cyparissias
no. taxa = 41
no. chars = 966
cte = 416
NPIC = 176
PIC = 374
no. MPTs = 9079
L = 1035
CI = 0.775
RI = 0.816
RC = 0.632
HKY+G
–Ln L = 5831.5463
GTR+G
–Ln L = 5830.689202
HKY+G
–Ln L = 5861.54
sf = 0.006415
G3pdhC-B
Outgroup: E. cyparissias
no. taxa = 35
no. chars = 798
cte = 332
NPIC = 146
PIC = 320
no. MPTs = 156
L = 767
CI = 0.850
RI = 0.897
RC = 0.762
HKY+G
–Ln L = 4111.9547
GTR+G
–Ln L = 4111.678
HKY+G
–Ln L= 4140.33
sf = 0.006610
ITS
Outgroup: M. esculenta
no. taxa = 64
no. chars = 782
cte = 378
NPIC = 69
PIC = 335
no. MPTs = 5302
L = 1448
CI = 0.489
RI = 0.756
RC = 0.370
GTR+G
–Ln L = 7532.9602
GTR+G
–Ln L = 7535.598
GTR+G
–Ln L = 7597.88
sf = 0.005813
matK
Outgroup: M. esculenta
no. taxa = 40
no. chars = 1646
cte = 1284
NPIC = 205
PIC = 157
no. MPTs = 125
L = 462
CI = 0.877
RI = 0.895
RC = 0.784
GTR+G
–Ln L = 4915.4074
GTR+G
–Ln L = 4916.249
GTR+G
–Ln L = 4948.77
sf = 0.004931
Combined-all
Outgroup: E. cyparissias
no. taxa = 37
no. chars = 4161
cte = 2957
NPIC = 435
PIC = 769
no. MPTs = 84
L = 1799
CI = 0.833
RI = 0.867
RC = 0.723
GTR+G
–Ln L = 14886.503
GTR+G
–Ln L (1 partition) = 14889.380
–Ln L (4 partitions) = 14238.121
unlinked
–Ln L = 14737.85
sf = 0.003480
Combined-one
Outgroup: M. esculenta
no. taxa = 25
no. chars = 4161
cte = 2550
NPIC = 639
PIC = 972
no. MPTs = 1
L = 2934
CI = 0.757
RI = 0.772
RC = 0.584
GTR+G
–Ln L = 19184.935
GTR+G
–Ln L (1 partition) = 19186.333
–Ln L (4 partitions) = 18332.947
unlinked
–Ln L = 18941.15
sf = 0.002337
Maximum likelihood (ML)—Searches were performed in the program Garli
v. 0.95 (Zwickl, 2006) under the optimal model of evolution for each data set
and under the GTR+G model in the program RAxML (Stamatakis et al., 2008;
see Table 1). The combined data sets were analyzed as a single partition under
the GTR+G model of evolution both in Garli and in RAxML, and as four partitions in RaxML, taking advantage of the “per gene branch optimization” setting
that allows parameters to be optimized independently among genes even when
analyzed assuming the same model of evolution. Support values were obtained
by ML bootstrapping with automatic estimation of replicate number in Garli and
RAxML (RAxML calculations were carried out on the CIPRES cluster, at the San
Diego Supercomputer Center; Miller et al., 2009). For the combined-one analyses,
we tested resolved branches to see whether they were significantly better than a
polytomy using a likelihood ratio test (as described by Baum et al., 2004).
Bayesian Markov chain Monte Carlo analysis—Metropolis coupled Markov
chain Monte Carlo (MCMCMC) tree sampling (Larget and Simon, 1999; Mau et
al., 1999) was implemented in the program MrBayes v. 3.1.2 (Huelsenbeck and
Ronquist, 2001; Ronquist and Huelsenbeck, 2003) under the optimal model of
evolution. For the combined data sets, we analyzed each of the four partitions
under its best fitting model, linking only the topology and branch lengths across
partitions (unlinked parameters: tratio, revmat, statefreq, and shape). Two independent analyses of two runs each were performed, with the following parameters: nchains = 4, ngens = 1 000 000, sampfreq = 100, temp = 0.2 (G3pdhC-A/B),
0.04 (ITS), and 0.07 (matK). Heat was selected based on a preliminary evaluation
of mixing guided by split frequencies and acceptance rates. Based on generationby-likelihood plots, from 10–15% of the samples were discarded as burn-in.
Topology tests—Support for alternative topologies was explored in the MP
framework using Wilcoxon sign-ranked tests (Templeton, 1983) as described by
Larson (1994) and implemented in PAUP*. In the MCMC framework, we assessed
support for alternative topologies by determining their frequency in the posterior
distribution. We tested the following clade relationships suggested by Dressler
(1957): (E. conzattii + E. coalcomanensis + E. cymbifera); (E. lomelii + E. bracteata Jacq. + E. tehuacana + E. cyri V.W.Steinm.); [E. peritropoides (Millsp.)
V.W.Steinm. + E. finkii], and; (E. tithymaloides + E. personata). We were not able
to test (E. calcarata + E. dressleri) because E. dressleri was not available.
Trait evolution—Thirty-four discrete morphological characters were scored
based on field observations, herbarium specimens, and the literature. Character
states and scoring are presented in Appendices S2 and S3 (see online Supplemental Data).
The 34 characters were mapped onto the ML phylogeny derived from the
combined-one data set (one accession per species). We used both MP and ML
approaches to map characters, as implemented in the program Mesquite v. 2.01
(Maddison and Maddison, 2009). For ML, a one-parameter Markov model
(MK-1) of character evolution was implemented, then visualized according to
the proportional likelihood at each node.
Testing cyathial spurs as key innovations—New methods for studying
differential diversification of lineages have been developed in recent years
(e.g., Ree, 2005; Maddison et al., 2007). Most of these newer methods are premised on including a representative sample of taxa that have or lack the putative
key innovation. The difficulties we faced in obtaining a representative sample
March 2010]
Cacho et al.— Are spurred cyathia a key innovation?
of outgroups, combined with our nearly exhaustive sampling in the ingroup,
make these methods invalid for our data set. Instead, we decided to use an older
method (Sanderson and Donoghue, 1994) that makes use of a phylogeny and
relative age estimates for key nodes, but does not stipulate sampling beyond
that needed to establish the number of species per clade.
To estimate a shift in diversification rate associated with cyathial nectar
spurs, we used our combined-one ML phylogeny and a “pure birth” approach
to modeling diversification as implemented in the program LRDiverse v. 0.8
(M. Sanderson, as described in Sanderson and Donoghue [1994, 1996]). This
method uses a three-taxon analysis of two ingroup clades that have the putative
key innovation, the spurred zygomorphic cyathium in this case, and an outgroup
taxon that lacks the trait. Based on the relative ages of the root and ingroup nodes,
and the number of species in each of the three clades, the program evaluates the
likelihood of clades of the observed size given one of several nested models of
diversification and uses likelihood ratio tests to identify the optimal model.
The sister group to the Pedilanthus clade most likely includes between one
and seven species (Steinmann and Porter, 2002; Steinmann et al., 2007; V. W.
Steinmann, unpublished data). Guided by the most recent publications on phylogenetics of the New World members of clade C of Euphorbia, we conducted
the analysis under four alternative species counts for the sister group of the
Pedilanthus clade: one species (E. sinclairiana Benth. [= E. elata Brandegee]),
as suggested by the analysis of Steinmann and Porter (2002), two species, four
species, or seven species (either the Euphorbia section Cubanthus or the E.
pteroneura clades). We also considered the scenario of granting species status
to the subspecies of E. tithymaloides.
We estimated relative nodal ages for the nodes “Outgroup + Pedilanthus
clade” (stem node), and “Pedilanthus subclades” (crown node; see Fig. 4) in our
ML phylogeny using the program r8s (Sanderson, 2003) under three different
methods: molecular clock (Langley-Fitch; LF), nonparametric rate smoothing
(NPRS), and penalized likelihood (PL). For each method, we obtained 95%
confidence intervals for each of the two estimated relative nodal ages by recalculating ages on 100 trees of the appropriate topology randomly selected from
the Bayesian posterior distribution (Perl script for sampling trees at random
from the posterior available upon request). For the PL method, the appropriate
value of λ was estimated using a cross validation procedure (15 increments of
0.25, starting at zero) on 10 trees of appropriate topology randomly sampled
from the Bayesian posterior distribution following Scherson et al. (2008). We
then used both the upper and lower bounds of the nodal ages confidence intervals for analyses of diversification rate in LRDiverse.
LRDiverse assumes that branching events follow a Poisson distribution in
any given lineage, governed by a single rate parameter that is assigned to each
Fig. 4. Three-taxon model used to assess diversification rates in the
Pedilanthus clade. The three taxa in the model correspond to: Outgroup
(see Results section for the alternative scenarios considered), the “core”
Pedilanthus clade (ingroup 1), and the PT subclade of the Pedilanthus
clade (ingroup 2). The branches evaluated for a potential shift in diversification rate correspond to the ones subtending these three clades. Relative
ages were estimated using r8s for the stem and crown nodes (marked with
stars) of the branch subtending the ingroup.
499
branch of a three-taxon phylogeny (“pure birth” or Yule model). A likelihood
ratio test (β = −2LR) is used to assess goodness of fit of successively less constrained models relative to the completely unconstrained model under which
each branch has its own rate of diversification. The five models compared in
LRDiverse are, in order of increasing complexity (see LRDiverse manual;
Sanderson and Donoghue, 1994, 1996): a one-parameter model (no shift in diversification rate; model 0); three two-parameter models (models 1–3), only
one of which (model 1) is compatible with a key innovation hypothesis, and the
unconstrained, three-parameter model (model 4). We analyzed our data under
the assumption that the internal branch has the same diversification rate as the
one subtending the outgroup (mode 0 in LRDiverse). In all cases, we used 1000
replicates of Monte Carlo simulation to statistically evaluate whether a simpler
model is rejected in favor of a more complex model.
Finally, we added Pedilanthus to Hodges’ (1997) one-tailed sign test analysis
of nectar spurs as key innovations across angiosperms and conducted a sister
group comparison as outlined by Slowinski and Guyer (1993) under the different
outgroup and species scenarios already outlined.
RESULTS
Phylogenetics— Individual data matrices for G3pdhC-A,
G3pdhC-B, ITS, and matK consisted of 966, 903, 782, and 1461
characters, respectively, with differing numbers of taxa. The
combined-one molecular data set consisted of 4161 characters
and 25 taxa, each of which had data for at least three genes.
Details on data matrices, diagnostic statistics, model selection,
and trees obtained are presented in Table 1. Figures 5 and 6 show
the ML trees for each of the four genes analyzed individually.
Four main subclades of the Pedilanthus clade were found in multiple single-gene analyses. For ease of communication, these are
labeled M (most species live in mesic environments), X (most
species live in xeric environments), PT (the two included species
are E. personata and E. tithymaloides), and F (E. finkii).
Discordance among molecular markers—An ILD test applied
to the four partitions in the combined-all data set rejected the null
hypothesis of congruence among datasets (P < 0.001). When
the four partitions were compared in a pairwise fashion and a
Bonferroni correction was used for multiple tests (six comparisons, α = 0.05/6 = 0.0083), significant incongruence was only
found for the gene pairs: G3pdhC-A/ITS and G3pdhC-A/G3pdhCB. Incongruence between these data sets appears to be due
to three alleles from Euphorbia calcarata (E. calcarata_01,
E. calcarata_04, E. calcarata_06,) and one allele from E. colligata
(E. colligata_03) when only ingroup taxa were considered, and
E. leucocephala, E. oerstediana, E. milii, and Manihot when
ingroup and outgroup taxa were considered. There is, thus, no
evidence of incongruence between the four genes in regards to
interspecific relationships within the Pedilanthus clade.
Species monophyly—Substantial infraspecific differentiation
was observed in cases where enough sampling was available
(i.e., E. calcarata, E. bracteata, E. tithymaloides), but the individual genes often supported species monophyly. For instance,
G3pdhC-B has enough resolution to resolve E. calcarata, E.
peritropoides, E. colligata, and E. lomelii as monophyletic entities. However, it fails to resolve species as monophyletic in the
PT subclade (Fig. 6B). Conversely, ITS resolves species as
monophyletic in the PT clade, but it fails to support monophyly
of E. bracteata and E. colligata (Fig. 6C).
Combined analyses with multiple accessions per species
(selecting alleles from heterozygotes that minimize tree length)
resulted in good support for monophyly of all species represented by multiple accessions within the core Pedilanthus clade.
500
American Journal of Botany
[Vol. 97
Fig. 5. Maximum likelihood (ML) trees of individual gene analyses showing high support for monophyly of the Pedilanthus clade and its relationships
to other Euphorbia lineages. (A) G3pdh-A. (B) G3pdh-B. (C) ITS. (D) matK. ML bootstrap values ≥50% are shown above branches.
In contrast, reciprocal monophyly of E. tithymaloides and E. personata is not well supported. Because there is consistent support
for monophyly of the PT subclade, ambiguity over reciprocal
monophyly of these two species would not substantially influence
the results of a combined analysis that uses a single accession per
species. The interspecies relationships with the Pedilanthus clade
implied by the latter analysis (Fig. 7) are identical to those found
when multiple accessions per species are included.
Phylogenetics of the Pedilanthus clade—Monophyly of the
Pedilanthus clade was very strongly supported in all analyses
and tests. A Templeton test rejected the optimal tree lacking a
Pedilanthus clade for the combined-one data (P < 0.0001).
These results together with the cyathial synapomorphies that
unify the group leave no reason to suspect a lack of monophyly
of the Pedilanthus clade.
While our outgroup sampling is not extensive, our combined
and individual gene analyses (Figs. 5, 7) support previous
hypotheses (Steinmann and Porter, 2002, Wurdack et al., 2005,
Bruyns et al., 2006, Steinmann et al., 2007) that the closest relatives of the Pedilanthus clade are New World members of clade
C of Euphorbia, here represented by E. gollmeriana, E. pteroneura, and E. umbelliformis. Our sampling is most complete in
our ITS analyses. These suggest the single species E. sinclairiana as the sister group to the Pedilanthus clade in agreement with
Steinmann and Porter (2002), although clade support is lacking
for such relationship. The tree estimated from G3pdhC-B shows
anomalous outgroup relationships—a result that may reflect
long-branch attraction in the sparsely sampled outgroup.
Within the Pedilanthus clade, the PT subclade consists of
E. personata and the E. tithymaloides species complex. The PT
clade is well supported by the ITS (MLBS = 90; MPBS = 100;
PP = 1.0) and matK (MLBS = 93; MPBS = 93; PP = 1.0) markers, and is not meaningfully contradicted by either of the G3pdh
markers. It is also strongly supported by the combined-one
(MLBS = 100; MPBS = 99; PP = 1.0) data set, but more weakly
supported by the combined-all data set. The shortest trees lacking the PT clade are significantly rejected by the combined-one
March 2010]
Cacho et al.— Are spurred cyathia a key innovation?
501
Fig. 6. Maximum likelihood (ML) trees of individual gene analyses showing relationships within the Pedilanthus clade. (A) G3pdh-A. (B) G3pdh-B.
(C) ITS. (D) matK. ML bootstrap values ≥50% are shown above branches.
502
American Journal of Botany
[Vol. 97
Fig. 7. Maximum likelihood (ML) tree of the combined-one data set. The placement of the Pedilanthus clade (top) and the relationships within it
(bottom) are illustrated. Support values ≥50% are as follows: ML bootstrap, MP bootstrap, Bayesian posterior probability.
dataset (P = 0.0158) but not by the combined-all data set (P =
0.3458). The PT subclade was the only group proposed by Dressler
(1957) that gained support from our data. Indeed, the other three
clades proposed by Dressler (1957) were rejected based on a Templeton test: E. conzattii + E. coalcomanensis + E. cymbifera (P <
0.0001); E. lomelii + E. bracteata + E. tehuacana + E. cyri (P <
0.0027), and; E. peritropoides + E. finkii (P < 0.0001).
With the exception of E. finkii, an evergreen species of mesic
forests and sole member of subclade F, all five species that
occur in mesic forests occur in clade M. Most of these species
present a tree habit and large glossy leaves. This contrasts with
the xeric clade, X, which includes succulent shrubs of tropical
deciduous forest, scrub or desert. Monophyly of the X and M
subclades of the Pedilanthus clade is strongly supported by all
analyses of individual genes and combined molecular data sets
(Figs. 6, 7). Indeed, Templeton tests show that trees lacking
either the M or X clades are significantly rejected by the combined-one data (P = 0.0158 and 0.0005, respectively).
G3pdhC-A and G3pdhC-B provide moderate support for subclades F and X as sister clades and for these together as sister to
March 2010]
Cacho et al.— Are spurred cyathia a key innovation?
subclade M. Neither matK nor ITS conflict with this topology.
It is, therefore not surprising that the combined-one data set
supports both the (F, X) clade (MLBS = 88; MPBS = 91; PP =
0.99), and the (M (F, X)) clade (MLBS = 70; MPBS = 95; PP =
0.95). However, a Templeton test, conducted with either combined data set, does not allow one to reject alternative relationships among the four subclades.
Trait evolution— Scoring of morphological characters and
the resulting morphological matrix are presented in online Appendices S2 and S3. Figure 8 illustrates ancestral state reconstruction (using ML) of six selected traits for the Pedilanthus
clade. Reconstruction of habit evolution (Fig. 8A) suggests a
single origin of the tree habit with one reversal in subclade M.
There seems to be a strong correlation between the tree habit
503
and seasonally mesic environments (Fig. 8B). Two wood anatomy traits that might be expected to influence performance in
dry vs. wet habitats, vessel-element grouping (Fig. 8C) and intervessel pitting (Fig. 8D), also seem to show some correlation
with habitat. Yellow latex was found to be a consistent synapomorphy of a subset of subclade X (Fig. 8E). Spur projection
(Fig. 8F) was inferred to show homoplasy, as were spur coloration, cyathium color patterning, the shape of the spur apex, and
the spur to involucral tube ratio (data not shown).
Testing key innovation hypotheses— To test the hypothesis
that the spurred cyathium correlates with an increased diversification rate, we evaluated the rate of species diversification of
the two major subclades of the Pedilanthus clade (PT and core
Pedilanthus) relative to the inferred sister to the Pedilanthus
Fig. 8. Maximum likelihood (ML) mapping of six selected morphological traits (only the ingroup is shown). (A) Habit: shrub (white), tree/treelet
(black). (B) Habitat: mesophytic (white), TDF (hatched), xeric (gray), desertic (black). (C) Vessel element grouping: solitary (white), grouped (black).
(D) Intervessel pitting: scalariform-pseudoscalariform (white); alternate-opposite dominant (black). (E) Latex color: white (white), yellow (black). (F) Spur
projection: conspicuous (black), inconspicuous (white).
American Journal of Botany
504
group. We used the program LRDiverse, which models diversification as a pure birth process and has the advantage over simple comparisons of the numbers of species between sister clades
that it takes into account the relative ages of the crown and stem
nodes to evaluate alternative models for changes in diversification rate. Relative nodal ages and confidence intervals on those
ages were similar for all three alternative dating methods (LF,
NPRS, and PL).
As summarized in Table 2, the preferred model depended
heavily on whether we count the diversified subspecies of E.
tithymaloides as being species and on how many species are
inferred to be in the sister group to the Pedilanthus clade. The
model of diversification that is predicted under the key innovation hypothesis requires that the diversification rate for the two
ingroup clades differs (higher) relative to that of the outgroup.
The P-values corresponding to models of increased complexity
successively compared to the “unconstrained” model (each
branch its own rate) are shown in Table 2. It can be seen that in
all cases that the “outgroup vs. ingroups 1+2” model is no worse
than the unconstrained model (and therefore preferred to that
more complex model). However, for most clade size assignments, there is at least one other model that is also favored relative to the unconstrained model. In many cases, the one-rate
model is also not significantly different than the unconstrained
model and, as the simpler model, is preferred. When E. tithymaloides is treated as a single species and when the outgroup is
assumed to be just a single species, the data are unable to distinguish whether an increase in diversification rate occurred in the
[Vol. 97
branch subtending the Pedilanthus clade (compatible with the
key innovation hypothesis), or the branch subtending the core
Pedilanthus clade (suggesting an increased diversification rate
in the core Pedilanthus subclade relative to both the PT subclade and the outgroup). The only case in which the key innovation hypothesis is uniquely supported (indicated with an
asterisk) is when the E. tithymaloides subspecies are counted as
species and when the sister group comprises just one species.
The Pedilanthus clade is larger in size than its sister group regardless of which of the putative sister groups is considered (one,
two, four, or seven species). This fits with the pattern observed by
Hodges (1997) that clades bearing spurs contain more species
than their spurless sister clades. Adding the Pedilanthus clade to
Hodges’ (1997) analysis of spurred vs. spurless sister clades, increases the significance of his one-tailed sign test from P = 0.0352
to 0.0195. On the other hand, for the Pedilanthus clade, significance under Slowinski and Guyer’s (1993) metric for contrasting
species numbers is only achieved (P = 0.0476) when considering
the sister clade as a single species and when the subspecies of
Euphorbia tithymaloides are treated as individual species.
DISCUSSION
This study contributes to our understanding of traits hypothesized to promote diversification in angiosperms through the
analysis of the spurred cyathia, a unique feature of the Pedilanthus clade that is analogous to floral nectar spurs. The evolution
Table 2.
Shift in diversification rate in the Pedilanthus clade inferred using LRDiverse. We considered four alternative clade sizes for the outgroup: 1,
2, 4, or 7 species. We also considered two ways of counting species within the ingroup: using the traditional species limits (upper part of the table)
or treating the recognized subspecies as species (lower part of table). The P-values evaluate whether the model explains the data significantly worse
than the unconstrained three-rate model. P-values less than 0.95 (boldface) suggest that the corresponding simpler model is preferred. The range of
P-values corresponds to the upper and lower confidence intervals of relative nodal ages for each of three molecular dating methods: Langley-Fitch
molecular clock (LF), nonparametric rate smoothing (NPRS), penalized likelihood (PL). A one- or two-rate model is considered uniquely supported
(indicated with an asterisk) when it is the only model that is found to not be significantly worse than the unconstrained model.
Out
group
Ingroup
1 (core)
Ingroup
2 (PT)
E. tithymaloides treated as one species
1
12
2
Dating
method
LF
NPRS
PL
2
12
2
LF
NPRS
PL
4
12
2
LF
NPRS
PL
7
12
2
LF
NPRS
PL
E. tithymaloides subspecies each treated as species
1
12
9
LF
NPRS
PL
2
12
9
LF
NPRS
PL
4
12
9
LF
NPRS
PL
7
12
9
LF
NPRS
PL
Two-rate models (clade subtended by the branch where the shift in rate is inferred)
One-rate model
Pedilanthus clade
Core Pedilanthus subclade
PT Pedilanthus subclade
0.992–0.997
0.988–0.997
0.997–0.999
0.925–0.968
0.906–0.959
0.965–0.996
0.827–0.899
0.797–0.882
0.883–0.98
0.739–0.856
0.713–0.808
0.801–0.96
0.779–0.787
0.79–0.809
0.778–0. 8
0.773–0.807
0.775–0.81
0.781–0.814
0.797–0.798
0.767–0.79
0.767–0.772
0.773–0.785
0–0.735
0.764–0.799
0.913–0.946
0.881–0.956
0.941–0.952
0.396–0.402
0.396–0.448
0.436–0.521
0
0
0–0.337
0
0
0
0.993–1.0
0.994–1.0
0.999–1.0
0.955–0.982
0.942–0.978
0.994–1
0.903–0.973
0.877–0.954
0.955–0.999
0.826–0.987
0.761–0.902
0.902–0.993
0.978–0.994
0.966–0.993
0.992–1.0
0.903–0.949
0.849–0.941
0.943–0.999
0.768–0.884
0.709–0.863
0.864–0.988
0.628–0.816
0.546–0.796
0.773–0.964
0*
0*
0*
0
0
0
0
0
0
0
0
0
0.989–1.0
0.983–1.0
0.984–1.0
0.947–0.978
0.887–0.948
0.942–1
0.813–0.904
0.757–0.9
0.886–0.995
0.684–0.843
0.621–0.812
0.784–0.969
0.995–1.0
0.992–0.995
0.999–1.0
0.958–0.992
0.934–0.982
0.992–1
0.88–0.959
0.857–0.945
0.932–0.998
0.799–0.909
0.73–0.889
0.872–0.988
March 2010]
Cacho et al.— Are spurred cyathia a key innovation?
of other morphological traits in the group was also examined
and we used the inferred phylogenetic relationships to shed
light on general aspects of the evolution of this small yet diverse clade of Euphorbia.
Systematics of the Pedilanthus clade— Species monophyly
and infraspecific discordance —We recovered a general pattern
of species exclusivity in our data. Three instances of discordance were observed and will be discussed in turn.
Euphorbia calcarata—The monophyly of Euphorbia calcarata is highly supported in all analyses performed. However,
there is incongruence in the relationships of individual alleles,
especially when comparing G3pdhC-A and ITS (Fig. 6A, 6C).
One individual, E. calcarata_01, from Chiapas, was found to
be a heterozygote at G3pdhC-A. One of its alleles (E. c._01a) is
closely related to the alleles from accessions of Chiapas and
Guatemala (E. c._04, and E. c._06), confirming a southern
clade. However, the second allele (E. c._01b) forms a clade
with more northern accessions [E. c._02 (Colima), E. c._03
(Michoacán)]. The genetic distinction between northern and
southern accessions of E. calcarata is also supported by ITS,
while matK only shows weak support for the northern clade. A
second source of conflict between the G3pdhC-A and ITS data
sets is in the placement of E. calcarata_06 (Guatemala). The
G3pdhC-A gene tree shows E calcarata_06 in the southern
clade, which is concordant with biogeography, whereas ITS
places it in the northern clade. Given the limited seed dispersal
ability of these taxa and the great distances involved, it seems
most likely that alleles having discordant patterns reflect incomplete lineage sorting.
Euphorbia diazlunana and E. bracteata—G3pdhC-A showed
a lack of reciprocal monophyly for E. diazlunana (Jalisco) and
E. bracteata (Sinaloa). Considering the geographical proximity
of these species, one might be tempted to invoke recent introgression. However, given that these two taxa appear to have
diverged from common ancestry only recently and appear to
have premating barriers to gene flow (E. bracteata is presumed
to be pollinated by hummingbirds, and E. diazlunana by hymenopterans; Sahagún-Godínez and Lomelí-Sención, 1997; N.
I. Cacho, personal observations), we believe that incomplete
lineage sorting is a better explanation for the pattern observed.
Euphorbia tithymaloides and E. personata—The exclusivity
of E. tithymaloides is supported by ITS (MLBS = 72), and this
support increases in the combined-one data set (MLBS = 82).
Nonetheless, the matK tree suggests some introgression or incomplete lineage sorting between E. tithymaloides and E. personata. Furthermore, the combined data set with multiple accessions
per species fails to resolve E. personata as monophyletic.
Despite the lack of consistent reciprocal monophyly, there is
morphological (vegetative and reproductive) and ecological
evidence of differentiation. Euphorbia personata has a discontinuous distribution, with known populations restricted to the
northernmost portion of the Yucatan Peninsula, Honduras, and
the Santa Rosa National Park in Costa Rica. The populations in
Honduras were not studied, but at the other two localities, E.
tithymaloides also occurs in nearby sites. It is notable that while
there are soil differences, in both cases the E. personata plants
occur in relatively dry deciduous forests, whereas E. tithymaloides occurs in more inland forests that have a higher
proportion of evergreen species.
505
Morphologically, E. personata is distinguished by having two
exposed lateral glands such that the cyathium resembles a face
with two eyes (Fig. 3C). These exposed glands make nectar more
accessible to insects and might account for floral visits by both
hummingbirds and bees (I. N. Cacho, personal observation). This
contrasts with E. tithymaloides cyathia of nearby populations,
which seem to receive almost exclusively hummingbird visits.
Also, when entering the staminate phase, the style bends backward to a much greater extent in E. personata than in E. tithymaloides. The fruit in E. personata is densely tomentose (Fig.
3D), whereas in E. tithymaloides it is mostly glabrous. Vegetatively, E. personata is distinguishable from E. tithymaloides by
its glaucous, erect, and mostly leafless stems that form more upright shrubs, noticeably taller than plants of E. tithymaloides (Fig.
2). The leaves of E. tithymaloides are much larger, much glossier,
and notably less puberulent than those of E. personata. Given the
morphological and ecological distinctiveness of E. personata and
E. tithymaloides, we infer that the lack of resolution in the molecular data reflects a lack of variation or possibly incomplete lineage sorting in the markers used rather than nonmonophyly for
much of the genome. However, the possibility of recent gene
flow between these two entities cannot be ruled out and remains
an interesting topic for future research.
Subclade relationships—Our data show a sister relationship
between the PT clade, comprising E. personata and E. tithymaloides, and the rest of the Pedilanthus clade, which we refer
to as the core Pedilanthus clade. In the combined data set, this
result gains solid clade support (MLBS = 70; MPBS = 95; PP =
0.95; Fig. 7). Several vegetative and reproductive characters are
consistent with this result. Members of the core Pedilanthus
clade (M, X, and F) tend to share the trait of having larger cyathia, generally with a well-developed and conspicuous spur
(Fig. 8F), in contrast with the smaller cyathia and truncate spurs
of the PT subclade. Also, the wood anatomy of the core Pedilanthus clade has been suggested to show a number of derived
features (Cacho, 2003; Carlquist, 1975, 2001; Dressler, 1957),
for example, the presence of wide and clustered vessels (Fig.
8C), alternate or opposite intervessel pits (Fig. 8D), and uniseriate and homogeneous rays. Other traits that appear to be derived within or at the base of the core Pedilanthus clade include
yellow latex (Fig. 8E), large cyme bracts, unicolored cyathia,
and inaperturate pollen.
Within the core Pedilanthus clade, both loci of G3pdhC show
a sister relationship between the F and X clades, while ITS and
matK fail to resolve any relationship with even moderate support. The G3pdhC resolution is well supported in the combined
analysis (MLBS = 88; MPBS = 91; PP = 0.99). Euphorbia
finkii, the only member of the F clade, is an evergreen, woody
(81% xylem), mesic shrub with large, glabrous, glossy leaves
and a spur that is bent forward (Fig. 3L). While represented by
a single individual in three of the four markers analyzed, the
morphology and ecology of E. finkii suggest reproductive and
ecological isolation from closely related entities and give no
reason to question its monophyly.
The six species that constitute the X clade (MLBS = 100;
MPBS = 100; PP = 1.0) are all succulent shrubs that inhabit
deserts, thorn scrubs, and tropical deciduous forests. These taxa
are either practically leafless (E. lomelii, E. cymbifera), or
markedly deciduous (E. diazlunana, E. bracteata, E. cyri, E.
tehuacana), and leaves, when present, are mostly densely pubescent. Both of our combined analyses resolve relationships
within the xeric clade with reasonably good support, despite a
506
American Journal of Botany
lack of resolution in the individual markers. The desert-inhabiting species E. lomelii and E. cymbifera form a moderately supported clade (MLBS = 76; MPBS = 71; PP = 0.94) that is sister
to a core xeric clade (MLBS = 100; MPBS = 100; PP = 1.0). Both
E. lomelii and E. cymbifera exhibit some sort of underground
dispersion (root adventitious buds in E. lomelii and rhizomes in
E. cymbifera) and have heavily cutinized, glaucous stems.
Two clades of two species comprise the core xeric subclade.
Euphorbia bracteata is distributed along the interior slopes of
the Sierra Madre Oriental, with disjunct populations from Sinaloa to Guerrero. Its sister species (MLBS = 93; MPBS = 95;
PP = 1.0), E. diazlunana, has a restricted distribution in the Sierra de Manantlán area of Jalisco. In spite of their shared, succulent, shrubby habit, these two taxa are morphologically
distinct. Euphorbia bracteata is taller, with thicker stems and
has brightly colored, persistent bracts that enclose a bright green
cyathium with prominent spur and involucral tube. Euphorbia
diazlunana, in contrast, is a shorter shrub with thinner but more
numerous stems (Fig. 2A), has much smaller green bracts, and
both the spur and the involucral tube are short and pale (Fig. 3E,
3F). These traits suggest insect pollination in E. diazlunana
(Sahagún-Godínez and Lomelí-Sención, 1997).
The final pair of species in the X subclade is E. tehuacana and
E. cyri, which form a clade that receives moderate support in the
combined analysis (MLBS = 83; MPBS = 76; PP = 0.97). Both
species occur in flat scrubland around the city of Oaxaca, an area
whose development places both taxa under threat (Olson et al.,
2005). These two species are very similar in habit, although E.
cyri forms much larger clumps than E. tehuacana does. Leaf size
and indumentum are also very similar, if not indistinguishable
under cultivation. However, these two species differ in cyathial
morphology. Euphorbia cyri has reddish and persistent bracts
that enclose its cyathium, whose involucral tube and spur are
quite prominent, the spur lobes in this species are fused to a degree that suggests that access to the nectar chamber requires some
considerable force. In contrast, E. tehuacana has a shortened involucral tube and a spur whose lobes do not enclose the gland
chamber as tightly as those of E. cyri. Additionally, the style in E.
tehuacana is shorter and bent back toward the gland chamber
(rather than projecting forward), and the staminate flowers are
only shortly exserted beyond the involucral tube. Given that a
sister relationship between E. diazlunana and E. tehuacana is
convincingly rejected by a Templeton test (P = 0.0143) and that
such a relationship is not present in any of the trees retained in the
Bayesian posterior distributions, the similarity of the cyathia of
E. tehuacana and E. diazlunana could reflect independent transitions from bird to insect pollination.
All five members of the M clade (MLBS = 100; MPBS = 99;
PP = 1.0) are woody, with a high percentage of xylem in their
stems (average 67% vs. 44% in the X subclade). The species in
this clade vary from woody shrubs (E. colligata) or treelets (E.
conzattii, E. peritropoides), to true trees (E. calcarata, E. coalcomanensis). Our data support a sister relationship of E. conzattii with the rest of the M clade (MLBS = 90; MPBS = 94; PP =
1.0). This taxon is the most restricted in distribution, with a
single population of about 20 individuals at the very top of a
single mountain in southwestern Mexico (Olson et al., 2005).
Individuals of E. conzattii are evergreen treelets about 1 m tall
that have caducous cyathial bracts and bright red cyathia (Fig.
3O) that contrast very prominently with the dark green of the
surrounding vegetation. The type specimen for this taxon is
mixed with material of E. calcarata (Dressler, 1957; Olson
et al., 2005), but our data show no evidence of introgression or
[Vol. 97
incomplete lineage sorting between these two taxa. On the basis
of our field observations and the data here presented, we believe
that this taxon is a distinct morphological and genetic entity.
Our data are unable to resolve the relationships among E. peritropoides, E. calcarata, and the two-species clade formed by E.
colligata and E. coalcomanensis within the core mesic clade. All
markers but G3pdhC-A resolve E. peritropoides as a monophyletic entity with high support, and not surprisingly, the combinedall data do so as well. Euphorbia peritropoides is an understory
treelet of mesic, seasonal forests with glabrous and glossy leaves
and a light pink to bright red spur, with an extremely reduced,
green involucral tube and no bracts. Unlike all other taxa in the
Pedilanthus clade, cyathia in E. peritropoides are borne on pendent inflorescence shoots, each with several cyathia (Fig. 3P).
Euphorbia coalcomanensis, a tree of tropical deciduous forests,
is well supported as sister (MLBS = 100; MPBS = 99; PP = 1.0) to
E. colligata, the only woody shrub of the mesic clade. Euphorbia
coalcomanensis has densely puberulent leaves that are somewhat
succulent (Fig. 2P). In contrast, E. colligata leaves are completely
glabrous and coriaceous, much like the leaves of the oak forests in
which it grows. These two sister species differ in cyathial characteristics as well: cyathia of E. coalcomanensis are green with
brightly colored and persistent bracts (Fig. 3R), whereas those of
E. colligata are red with caducous bracts (Fig. 3M).
Relationships of the Pedilanthus clade—Traditionally, sampling issues (both taxonomic and of molecular characters) have
played a role in the persistence of unresolved phylogenetic
relationships among New World members of the clade C of
Euphorbia. In addition to the long branch subtending the
Pedilanthus clade, sampling issues might contribute to the uncertainty regarding the sister group of the Pedilanthus clade.
Our results do not show conclusive support for a Mexican origin for the Pedilanthus clade, as was suggested by Dressler
(1957). The PT subclade, which is sister to the rest of the Pedilanthus clade, includes one Mesoamerican species (E. personata) and one species that occurs throughout Central America,
coastal northern South America and the Caribbean (E. tithymaloides). Furthermore, many of the putative closest relatives
of the Pedilanthus clade [E. sinclairiana, E. comosa, E. pteroneura, E. hoffmanniana (Klotzsch & Garcke) Boiss., E. weberbaueri Mansf., E. cestrifolia Kunth, E. calyculata Kunth, E.
lagunillarum Croizat, E. tanquahuete Sessé & Moc.], occur either in Mexico, Central America, the Caribbean, or northern
South America (Steinmann et al., 2007). While a Mexican and
even a Caribbean origin of the group remains plausible, it seems
likely that the Pedilanthus group had a Central American ancestor that later diversified in central Mexico giving rise to the core
Pedilanthus clade. More precise inferences about the phylogenetic relationships of the Pedilanthus clade and its close relatives require a comprehensive study of clade C of Euphorbia,
with special emphasis on its New World members. Our results
suggest that G3pdhC might prove to be an excellent marker for
such an expanded study. Whether the Pedilanthus clade diversified from a succulent or woody ancestor is not certain at this
time because many of the New World clades that are closely
related to the Pedilanthus clade present varying degrees of succulence (e.g., E. pteroneura and E. gollmeriana), while others
are rather woody (E. sinclairiana).
Character evolution in the Pedilanthus clade— We explored
character evolution in the Pedilanthus clade to identify morphological synapomorphies for its different subclades. Our ML
March 2010]
Cacho et al.— Are spurred cyathia a key innovation?
reconstructions suggested that some vegetative characters
including habit, vessel grouping, or intervessel pit morphology
(Fig. 8A, 8C, 8D) might be quite good predictors of phylogeny
at a coarse scale. Other characters might be helpful “locally.”
For example, yellow latex is a synapomorphy unifying the core
xeric clade (Fig. 8E).
There is high homoplasy within reproductive morphological
traits and thus a clear absence of simple cyathial synapomorphies for any particular clade. Involucre color, spur size, spur
projection (Fig. 8F), spur shape and color, bract morphology
(persistence, size, coloration), and gland exposure all contribute
to the striking morphological variation of the zygomorphic cyathium in the Pedilanthus clade. Three distinct strategies seem
to serve the same ecological function of rendering the cyathium
visually conspicuous (equivalent to floral display): (1) numerous small cyathia with brightly colored involucral tubes and
truncate spurs; (2) few, large, usually green cyathia with welldeveloped spurs and involucral tubes and large, persistent, colorful bracts; and (3) few, large cyathia with brightly colored
spurs and involucral tubes but inconspicuous, often caducous
bracts. Strategy one is found in E. tithymaloides and E. personata of the PT subclade; in the X and M clades, there is a mixture of strategies two and three. For example, large, persistent,
colored bracts are present in E. coalcomanensis, E. bracteata,
E. cyri, and E. tehuacana, with a maximum of four mature cyathia at a given time (e.g., Fig. 3A), in contrast to 8–12 in the
PT clade (Fig. 3C). In cases where there is a single mature cyathium at a given time (e.g., E. conzattii, E. cymbifera, E. calcarata), cyathia have well-developed spurs and involucral
tubes, both solidly and brightly colored with caducous bracts
(e.g., Figs. 3I, 3K, 3O, 3N). Altogether, character-mapping
experiments reveal a lack of defining synapomorphies for the
major subclades of the core Pedilanthus clade and suggest that
reproductive morphology is more labile than vegetative morphology in the group, as might be expected if a reproductive
trait, the cyathial spur, has served as a key innovation.
Is the spurred zygomorphic cyathium a key innovation?— There is now a body of evidence that supports that nectar
spurs tend to increase diversification rate. This has been established both with one-tailed sign tests (Hodges, 1997) and with
analyses using independent contrasts on a comprehensive supertree of angiosperm families (Kay et al., 2006). This repeated
pattern, combined with a plausible causal hypothesis of how
nectar spurs might influence diversification, support the claim
that floral nectar spurs are key innovations (Hodges, 1997; Kay
et al., 2006). Here we explored whether this inference can be
extended to the extrafloral nectar spur found in the Pedilanthus
clade, which serves an analogous function to that of floral nectar spurs in other angiosperms.
In the Pedilanthus clade, the nectar glands are tightly enclosed
within a chamber, the spur, in such a way that full access to the
reward (nectar) requires forceful entry by a hummingbird or large
insect. Together with the strong zygomorphy of the Pedilanthus
cyathium, it has been argued that the presence of nectar glands
tightly enclosed in the cyathial spur has opened the adaptive zone
of hummingbird pollination in Euphorbia, a group that is otherwise mostly insect-pollinated (Dressler, 1957). Observations of
plants of the Pedilanthus clade in the wild suggest that hummingbirds play a role in pollen transfer by contacting male and female
flowers when probing spurred cyathia for nectar. A high diversity
in spur morphology coupled with low sequence variation in the
Pedilanthus clade compared to its outgroups, and a pattern of a
507
long branch followed by a sudden diversification in the group,
are suggestive of a rapid radiation in the Pedilanthus clade after
the evolution of the nectar spur.
In the light of our current knowledge of phylogenetic relationships among the New World members of the Clade C of
Euphorbia, the most likely sister group to the Pedilanthus clade
includes from one to seven taxa (Steinmann and Porter, 2002;
Steinmann et al., 2007; V. W. Steinmann, unpublished data).
Using the approach of Sanderson and Donoghue (1994), which
considers rates of diversification in a relative temporal framework, we found support for a model consistent with the spurred
cyathium being a key innovation only when we assumed that
the sister group to the Pedilanthus clade is a single species and
when the subspecies of E. tithymaloides are treated as species
(Table 2). Under other scenarios, the data are compatible with a
key innovation hypothesis but do not uniquely support it over
alternative models. These results are in complete agreement
with Slowinski and Guyer’s method: significance (P = 0.0476)
is only achieved when the sister clade to the Pedilanthus clade
is a single species and when the subspecies of E. tithymaloides
are treated as species. On the other hand, the Pedilanthus clade
is larger than its sister group regardless of which of the putative
sister groups is considered and whether the subspecies of E. tithymaloides are granted species status. Thus, adding the Pedilanthus clade to Hodges’ (1997) analysis of spurred vs. spurless
sister clades, increases the significance of his one-tailed sign
test from P = 0.0352 to 0.0195.
Our analysis of nectar spurs as key innovations faces two
major limitations. One is the relatively small size of the Pedilanthus clade, which is illustrated by a lack of significance when
only clade species numbers are taken into account. The minimal species proportion between sister clades to identify a significant directional shift in species diversity based on clade
species numbers alone has been shown to be 20 : 1 (Slowinski
and Guyer, 1993). The other limitation is an unequal density of
sampling of the Pedilanthus clade and its close relatives: while
our sampling of members of the Pedilanthus clade is virtually
complete, our sampling of outgroups is quite sparse. This unequal sampling density poses serious limitations for methods of
analysis that take into account the length of branches in all parts
of the tree (e.g., Ree, 2005; Maddison et al., 2007).
We are optimistic that it will be possible to implement approaches with higher statistical power than the one used here in
the near future, as new information on the phylogenetic relationships within Euphorbia comes to light, especially with respect to
the New World members of the clade C of Euphorbia. Also, improved knowledge of Euphorbia phylogenetics will make it possible to statistically test the role of other traits that have been
proposed to increase the rate of diversification in this huge clade
of flowering plants. For example, the cyathium itself has been
proposed as a key innovation explaining the tremendous diversification of Euphorbia (Steinmann and Porter, 2002; Prenner and
Rudall, 2007; Prenner et al., 2008). This hypothesis seems plausible considering that Euphorbia is the second largest plant genus
(ca. 2100 species) and is sister to a relatively species-poor group,
Neoguillauminiinae (six species), which lacks cyathia.
This study was not intended as a test of a causal hypothesis
but rather as an evaluation of whether a general correlation
found for floral nectar spurs and increased diversification rates
might hold for an analogous structure, the cyathial spur of Pedilanthus. In Aquilegia, pollinator shifts drive diversification by
imposing directional selection on nectar spurs (Whittall and
Hodges, 2007). In the case of the Pedilanthus clade, our analy-
American Journal of Botany
508
sis neither supports nor rejects the hypothesis that pollinator
shifts have played a role in promoting diversification in this
group. Slight changes in the morphology of the spur of Pedilanthus could either cause shifts in pollinator identity, whether to
different hummingbird species or, in two cases, to insect pollinators, or it could alter the location of pollen deposition on a
specific pollinating species and thus increase the rate at which
premating reproductive isolation can evolve. Future pollination
studies of the Pedilanthus clade will allow these causal hypotheses to be tested. Also, future systematic research that enables
the use of more powerful methods, clarifies the composition of
the sister group to the Pedilanthus clade, and reevaluates the
status of infraspecific taxa within E. tithymaloides could lead to
a more definitive statement as to whether the evolution of the
spurred cyathium correlates with increased species diversification. However, by analogy to floral spurs, and allowing for the
fact that the Pedilanthus clade could be too young a group to
show statistically significant evidence of increased diversification under the methods here applied, it remains plausible that the
cyathial nectar spur of the Pedilanthus clade of Euphorbia has
spurred the diversification of this distinctive group of spurges.
LITERATURE CITED
Baum, D. A., S. D. Smith, A. Yen, W. S. Alverson, R. Nyffeler, B. A.
Whitlock, and R. L. Oldham. 2004. Phylogenetic relationships of
Malvatheca (Bombacoideae and Malvoideae; Malvaceae sensu lato)
as inferred from plastid DNA sequences. American Journal of Botany
91: 1863–1871.
Bruyns, P. V., R. J. Mapaya, and T. Hedderson. 2006. A new subgeneric classification for Euphorbia (Euphorbiaceae) in southern Africa
based on ITS and psbA-trnH sequence data. Taxon 55: 397–420.
Cacho, N. I. 2003. Correlaciones en forma de vida y ecología con características de elementos de vaso en Pedilanthus (Euphorbiaceae),
con comentarios sobre su estado de conservación actual basados
en trabajo de campo reciente. Senior thesis (Biology), Universidad
Nacional Autónoma de México, México D.F., México.
Carlquist, S. 1975. Ecological strategies of xylem evolution. University
of California Press, Berkeley, California, USA.
Carlquist, S. 2001. Comparative wood anatomy: Systematic, ecological
and evolutionary aspects of dicotyledon wood, 2nd ed. Springer, New
York, New York, USA.
Chase, M. W., and H. H. Hillis. 1991. Silica gel: An ideal material for field
preservation of leaf samples for DNA studies. Taxon 40: 215–220.
Doyle, J. J., and J. L. Doyle. 1987. A rapid DNA isolation procedure from
small quantities of fresh leaf tissues. Phytochemical Bulletin 19: 11–15.
Dressler, R. L. 1957. The genus Pedilanthus (Euphorbiaceae). Contributions from the Gray Herbarium of Harvard University, vol 182,
1–188. Harvard University Press, Cambridge, Massachusetts, USA.
Dressler, R. L., and C. M. Sacamano. 1992. Pedilanthus connatus
(Euphorbiaceae), especie nueva y llamativa de Jalisco. Acta Botanica
Mexicana 18: 21–24.
Farris, J. S., M. Källersjö, A. G. Kluge, and C. Bult. 1994. Testing
significance of incongruence. Cladistics 10: 315–319.
Galis, F. 2001. Key innovations and radiations. In G. P. Wagner [ed.],
The character concept in evolutionary biology, 583–607. Academic
Press, San Diego, California, USA.
Harris, J. G., and M. W. Harris. 2001. Plant identification terminology:
An illustrated glossary, 2nd ed. Spring Lake Publishing, Spring Lake,
Utah, USA.
Hodges, S. 1997. Floral nectar spurs and diversification. International
Journal of Plant Sciences 158: S81–S88.
Hodges, S., and M. Arnold. 1995. Spurring plant diversification: Are
floral nectar spurs a key innovation? Proceedings of the Royal Society
of London, B, Biological Sciences 262: 343–348.
Huelsenbeck, J. P., and F. Ronquist. 2001. MRBAYES: Bayesian inference of phylogeny. Bioinformatics 17: 754–755.
[Vol. 97
Hunter, J. P. 1998. Key innovations and the ecology of macroevolution.
Trends in Ecology & Evolution 13: 31–36.
Johnson, L. A., and D. E. Soltis. 1994. matK DNA-sequences and phylogenetic reconstruction in Saxifragaceae s.st. Systematic Botany 19:
143–156.
Kay, K. M., C. Voelckel, J. Y. Yang, K. M. Hufford, D. D. Kaska, and
S. A. Hodges. 2006. Floral characters and species diversification. In
L. D. Harder and S. C. H. Barrett [eds.], Ecology and evolution of flowers, 311–325. Oxford University Press, New York, New York, USA.
Larget, B., and D. L. Simon. 1999. Markov chain Monte Carlo algorithms for the Bayesian analysis of phylogenetic trees. Molecular
Biology and Evolution 16: 750–759.
Larson, A. 1994. The comparison of morphological and molecular data in
phylogenetic systematics. In B. Schierwater, B. Streit, G. P. Wagner,
and R. DeSalle [eds.], Molecular ecology and evolution: Applications
and approaches, 371–390. Birkhauser Verlag, Basel, Switzerland.
Lomelí-Sención, J., and E. Sahagún-Godínez. 2002. Rediscovery of
Pedilanthus coalcomanensis (Euphorbiaceae), a threatened endemic
Mexican species. American Journal of Botany 89: 1485–1490.
Maddison, D. R., and W. P. Maddison. 2002. MacClade 4.05:
Analysis of phylogeny and character evolution. Sinauer, Sunderland,
Massachusetts, USA.
Maddison, W. P., P. E. Midford, and S. P. Otto. 2007. Estimating
a binary character’s effect on speciation and extinction. Systematic
Biology 56: 701–710.
Maddison, W. P., and D. R. Maddison. 2009 onward. Mesquite: A modular system for evolutionary analysis, version 2.6. Computer program
and documentation, website http://mesquiteproject.org.
Mau, B., M. A. Newton, and B. Larget. 1999. Bayesian phylogenetic
inference via Markov chain Monte Carlo methods. Biometrics 55:
1–12.
Miller, M. A., M. T. Holder, R. Vos, P. E. Midford, T. Liebowitz, L.
Chan, P. Hoover, and T. Warnow. 2009. CIPRES [Cyberinfrasructure
for Phylogenetic Research]. Website http://www.phylo.org/sub_sections/
portal; archived at http://www.webcitation.org/5imQlJeQa.
Minin, V., Z. Abdo, P. Joyce, and J. Sullivan. 2003. Performancebased selection of likelihood models for phylogeny estimation.
Systematic Biology 52: 674–683.
Neilson, W. A., T. A. Knott, and D. W. Carhart [eds]. 1950.
Webster’s new international dictionary of the English language, 2nd
ed., unabridged. G. & C. Merriam, Springfield, Massachusetts, USA.
Nyffeler, R., C. Bayer, W. S. Alverson, A. Yen, B. A. Whitlock,
M. W. Chase, and D. A. Baum. 2005. Phylogenetic analysis of
the Malvadendrina clade (Malvaceae s.l.) based on plastid DNA sequences. Organisms, Diversity & Evolution 5: 109–123.
Olson, M., J. A. Lomelí, and N. I. Cacho. 2005. Extinction threat in the
Pedilanthus clade (Euphorbia, Euphorbiaceae) with special reference
to the recently rediscovered E conzattii (P. pulchellus). American
Journal of Botany 92: 634–641.
Prenner, G., S. D. Hopper, and P. J. Rudall. 2008. Pseudanthium
development in Calycopeplus paucifolius, with particular reference
to the evolution of the cyathium in Euphorbieae (Euphorbiaceae–
Malpighiales). Australian Systematic Botany 21: 153–161.
Prenner, G., and P. J. Rudall. 2007. Comparative ontogeny of the
cyathium in Euphorbia (Euphorbiaceae) and its allies: Exploring the
organ-flower-inflorescence boundary. American Journal of Botany
94: 1612–1629.
Ree, R. 2005. Detecting the historical signature of key innovations using
stochastic models of character evolution and cladogenesis. Evolution
59: 257–265.
Ronquist, F., and J. P. Huelsenbeck. 2003. MRBAYES 3: Bayesian
phylogenetic inference under mixed models. Bioinformatics 19:
1572–1574.
Sahagún-Godínez, E., and J. A. Lomelí-Sención. 1997. Pedilanthus
diazlunanus (Euphorbiaceae): Pollination by hymenopterans in a birdpollinated genus. American Journal of Botany 84: 1584–1587.
Sanderson, M. J. 2003. r8s: Inferring absolute rates of molecular evolution and divergence times in the absence of a molecular clock.
Bioinformatics 19: 301–302.
March 2010]
Cacho et al.— Are spurred cyathia a key innovation?
Sanderson, M. J., and M. J. Donoghue. 1994. Shifts in diversification
rate with the origin of Angiosperms. Science 264: 1590–1593.
Sanderson, M. J., and M. J. Donoghue. 1996. Reconstructing shifts
in diversification rates on phylogenetic trees. Trends in Ecology &
Evolution 11: 15–20.
Scherson, R. A., R. Vidal, and M. J. Sanderson. 2008. Phylogeny,
biogeography, and rates of diversification of New World Astragalus
(Leguminosae) with an emphasis on South American radiations.
American Journal of Botany 95: 1030–1039.
Simpson, G. G. 1953. The major features of evolution. Columbia
University Press, New York, New York, USA.
Slowinski, J. B., and C. Guyer. 1993. Testing whether certain traits
have caused amplified diversification: An improved method based on
a model of random speciation and extinction. American Naturalist
142: 1019–1024.
Stamatakis, A., P. Hoover, and J. Rougemont. 2008. A fast bootstrapping algorithm for the RAxML Web servers. Systematic Biology
57: 758–771.
Steinmann, V. W. 2003. The submersion of Pedilanthus into Euphorbia
(Euphorbiaceae). Acta Botanica Mexicana 65: 45–50.
Steinmann, V. W., and J. M. Porter. 2002. Phylogenetic relationships
in Euphorbieae (Euphorbiaceae) based on its and ndhF sequence data.
Annals of the Missouri Botanical Garden 89: 453–490.
Steinmann, V. W., B. Van Ee, P. E. Berry, and J. Gutierrez. 2007.
The systematic position of Cubanthus and other shrubby endemic
509
species of Euphorbia (Euphorbiaceae) in Cuba. Anales del Jardín
Botánico de Madrid 64: 123–133.
Strand, A. E., J. Leebens-Mack, and B. B. Milligan. 1997. Nuclear
DNA-based markers for plant evolutionary biology. Molecular
Ecology 6: 113–118.
Swofford, D. L. 2002. PAUP*: Phylogenetic analysis using parsimony (*and other methods), version 4.0b10. Sinauer, Sunderland,
Massachusetts, USA.
Templeton, A. R. 1983. Phylogenetic inference from restriction endonuclease cleavage site maps with particular reference to the humans and
apes. Evolution 37: 221–244.
White, T. J., T. Bruns, S. Lee, and J. W. Taylor. 1990. Amplification
and direct sequencing of fungal ribosomal RNA genes for phylogenetics. In M. A. Innis, D. H. Gelfand, J. J. Sninsky, and T. J. White
[eds.], PCR protocols: A guide to methods and applications, 315—
322. Academic Press, New York, New York, USA.
Whittall, J. B., and S. A. Hodges. 2007. Pollinator shifts drive increasingly long nectar spurs in columbine flowers. Nature 447: 706–707.
Wurdack, K. J., P. Hoffman, and M. W. Chase. 2005. Molecular phylogenetic analysis of uniovulate Euphorbiaceae (Euphorbiaceae sensu
stricto) using plastid rbcL and trnL-F DNA sequences. American
Journal of Botany 92: 1397–1420.
Zwickl, D. J. 2006. Genetic algorithm approaches for the phylogenetic analysis of large biological sequence data sets under the maximum likelihood
criterion. Ph.D. dissertation, University of Texas, Austin, Texas, USA.
Appendix 1. Voucher and GenBank accession information for specimens included in this study. Identification number, locality information, and herbaria are provided
for specimens collected for this study; sequences downloaded from GenBank are identified as “downloaded”). Abbreviations: DAV= Herbarium at University
of California-Davis; HAJB = Jardín Botánico de la Habana, Cuba; MEO = Mark E. Olson; MEXU = Herbario Nacional de México; NIC = N. Ivalú Cacho;
UCB = University of California-Berkley Botanical Garden; UCD = University of Califorinia-Davis Botanical Garden; UC/JEPS: The University and Jepson
Herbaria at University of California-Berkeley; UWBot = University of Wisconsin-Madison Botany Greenhouses; UWDCS = University of Wisconsin-D. C.
Smith Greenhouse; WIS = Wisconsin State Herbarium.
Taxon, Voucher (or other identification number), Locality, Herbaria, GenBank accessions G3pdhC-A, G3pdhC-B, ITS, matK
Ingroup—E. bracteata_1, MEO & NIC 845, México: Sinaloa, WIS, MEXU,
[GU214886 (direct sequencing), GU214892 (clone 3)], GU214954,
GU214909, n/a; E. bracteata_2, MEO & NIC 1011, México: Guerrero,
WIS, MEXU, GU214887, GU214967, GU214910, GU214846; E.
bracteata_3, MEO & NIC 1010, México: Guerrero, WIS, MEXU,
GU214888, GU214976, GU214911, GU214854; E. bracteata_4,
downloaded, n/a, n/a, n/a, n/a, AF537489.1, n/a; E. calcarata_1, MEO
806, México: Chiapas, WIS, MEXU, [GU214895 (clone 4), GU214896
(clone 9)], GU214957, GU214912, GU214835; E. calcarata_2, MEO
& NIC 896, México: Colima, WIS, MEXU, GU214874, GU214962,
GU214913, GU214840; E. calcarata_3, MEO & NIC 900, México:
Michoacán, WIS, MEXU, GU214875, GU214963, GU214914,
GU214841; E. calcarata_4, MEO & NIC 939A, México: Chiapas, WIS,
MEXU, GU214897, GU214964, GU214915, GU214843; E. calcarata_5,
MEO & NIC 939, México: Chiapas, WIS, MEXU, n/a, GU214969,
GU214916, GU214848; E. calcarata_6, NIC 407, Guatemala: Nentón,
WIS, GU214883, GU214980, GU214917, GU214857; E. calcarata_7,
downloaded, n/a, n/a, AF537492.1, n/a; E. coalcomanensis, MEO &
NIC 886, México: Michoacán, WIS, MEXU, GU214873, GU214961,
GU214918, GU214839; E. colligata_1, MEO & NIC 866, México:
Jalisco, WIS, MEXU, [GU214870 (direct sequencing), GU214889 (clone
9)], GU214958, GU214919, GU214836; E. colligata_2, MEO & NIC
867, México: Jalisco, WIS, MEXU, [GU214871 (direct sequencing),
GU214890 (clone 2), GU214891 (clone 3)], GU214959, GU214920,
GU214837; E. colligata_3, MEO & NIC 867A, México: Jalisco, WIS,
MEXU, GU214872, GU214960, GU214921, GU214838; E. colligata_4,
downloaded, n/a, n/a, AF537493.1, n/a; E. conzattii, MEO & NIC 971,
México: Oaxaca, WIS, MEXU, GU214880, GU214972, GU214922,
GU214851; E. cymbifera, MEO & NIC 979, México: Puebla, WIS,
MEXU, GU214869, GU214956, GU214923, GU214834; E. cymbifera_1,
downloaded, n/a, n/a, AF537491.1, n/a; E. cyri, MEO & NIC 973,
México: Oaxaca, WIS, MEXU, GU214894, n/a, GU214926, GU214833;
E. diazlunana, MEO & NIC 888, México: Jalisco, WIS, MEXU,
[GU214893 (clone 4), GU214901 (clone 1)], GU214968, GU214927,
GU214847; E. finkii, MEO & NIC 917, México: Oaxaca, WIS,
MEXU, GU214898, GU214973, GU214929, GU214852; E. finkii_1,
downloaded, n/a, n/a, AF537520.1, n/a; E. lomelii_4, downloaded, n/a,
n/a, AF537490.1, n/a; E. lomelii_1, MEO & NIC 852, México: BCS, WIS,
MEXU, GU214868, GU214955, GU214933, GU214831; E. lomelii_2,
UCD 99263, México: BCS, DAV, GU214885, GU214981, GU214934,
GU214860; E. lomelii_3, UCB 62.0776, México: BCS, UC/JEPS, n/a,
n/a, GU214935, n/a; E. peritropoides_1, MEO & NIC 974, México:
Oaxaca, WIS, MEXU, GU214877, GU214966, GU214937, GU214845;
E. peritropoides_2, MEO & NIC 996, México: Guerrero, WIS, MEXU,
GU214878, GU214970, GU214938, GU214849; E. personata_1, MEO
& NIC 955, México: Yucatán, WIS, MEXU, GU214899, GU214974,
GU214939, GU214832; E. personata_2, NIC 343, Guatemala, WIS, n/a,
GU214979, GU214940, GU214856; E. t. subsp. angustifolia_1, NIC
059.2, USA: USVI: St. John, WIS, GU214881, GU214977, GU214945,
GU214855; E. t. subsp. angustifolia_2, NIC 073.2, Dominican Republic,
WIS, GU214882, GU214978, GU214946, n/a; E. t. subsp. padifolia,
Iltis 30229, Statia, WIS, GU214879, GU214971, GU214947, GU214850;
E. t. subsp. tithymaloides_1, MEO & NIC 926, México: Oaxaca, WIS,
MEXU, n/a, n/a, GU214948, GU214842; E. t. subsp. tithymaloides_2,
MEO & NIC 947, México: Oaxaca, WIS, MEXU, GU214876,
GU214965, GU214949, GU214844; E. t. subsp. tithymaloides_3, NIC
139, Guatemala: Cuija, WIS, GU214884, n/a, GU214950, GU214858;
E. t. subsp. tithymaloides_4, NIC 140, Guatemala: Cahuí, WIS, n/a,
n/a, GU214951, GU214859; E. t. subsp. tithymaloides_5, downloaded,
n/a, n/a, AF537494.1, n/a; E. tehuacana, MEO & NIC 981, México:
Puebla, WIS, MEXU, GU214900, GU214975, GU214944, GU214853;
E. tehuacana_1, downloaded, n/a, n/a, AF537488.1, n/a
510
American Journal of Botany
Outgroups—E. cestrifolia Kunth, downloaded, n/a, n/a, AF537521.1, n/a; E.
comosa Vell., downloaded, n/a, n/a, AF537503.1, n/a; E. cymosa Poir., NIC
083, Jamaica, WIS, GU214906, GU214988, GU214924, GU214865; E.
cyparissias L., NIC 429, USA: Wisconsin, WIS, GU214908, GU214984,
GU214925, GU214866; E. esula L., NIC 428, USA: Wisconsin, WIS,
GU214907, GU214985, GU214928, n/a; E. gollmeriana Klotzsch ex
Boiss., NIC 126, Venezuela: Falcón, WIS, GU214904, GU214982,
GU214930, n/a; E. heterophylla L., NIC 044, Puerto Rico: Manatí,
WIS, n/a, n/a, GU214931, GU214861; E. hoffmanniana (Klotzsch &
Garcke) Boiss., downloaded, n/a, n/a, AF537508.1, n/a; E. humifusa
Willd., downloaded, n/a, n/a, n/a, AB233780.1; E. leucocephala Lotsy,
NIC 414, Guatemala: Nentón, WIS, GU214902 (clone 4), GU214986,
GU214932, GU214862; E. lindenii (S.Carter) Bruyns, downloaded,
n/a, n/a, AF537473.1, n/a; E. milii Des Moul., NIC 626, UWBot,
WIS, GU214903 (clone 2), GU214987, GU214936, GU214864;
E. obesa Hook.f., downloaded, n/a, n/a, AF537566.1, n/a; E. petiolaris
Sims, NIC 054, USA: USVI: St. John, WIS, n/a, n/a, GU214941,
n/a; E. polyacantha Boiss., downloaded, n/a, n/a, n/a, AY491656.1;
E. pteroneura A.Berger_1, NIC 411, Guatemala: Nentón, WIS, n/a,
n/a, GU214942, GU214867; E. pteroneura _2, downloaded, n/a, n/a,
AF537506.1, n/a; E. pulcherrima Willd. ex Klotzsch_1, NIC 406,
Guatemala:Petén, WIS, n/a, n/a, GU214943, n/a; E. pulcherrima
_2, downloaded, n/a, n/a, AF537432.1, n/a; E. sinclairiana Benth.,
downloaded, n/a, n/a, AF537495.1, n/a; E. umbellata (Pax) Bruyns,
downloaded, n/a, n/a, AF537469.1, AB233784.1; E. umbelliformis
(Urb. & Ekman) V.W.Steinm. & P.E.Berry, HAJB 81901, Cuba, WIS,
GU214905, GU214983, GU214952, n/a; E. weberbaueri Mansf.,
downloaded, n/a, n/a, AF537519.1, n/a; Manihot esculenta, NIC 625,
UWDCS, WIS, n/a, n/a, GU214953, GU214863; Neoguillauminia
cleopatra (Baill.) Croizat, downloaded, n/a, n/a, AF537581.1, n/a.