| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Institute of Systematic Botany, University of Zurich, Zollikerstrasse 107, 8008 Zurich, Switzerland; 3Department of Biology, Colorado College, Colorado Springs, Colorado 80903 USA
Received for publication August 7, 2003. Accepted for publication January 22, 2004.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: adaptation • buzz pollination • Cortusa • Dodecatheon • exaptation • heterostyly • Primula subgenus Auriculastrum • solanoid flowers
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
During the buzz pollination of Dodecatheon, pollen is removed from the anthers by vibrations that originate with a settled bee's indirect flight muscles and are transferred to the androecium via its legs and mandibles (Harder and Barclay, 1994
). Pendant flowers with reflexed petals and large, conspicuous, connivent, poricidal anthers, as seen in Dodecatheon, were dubbed "solanoid" by Vogel (1978)
. Solanoid flowers are independently derived in numerous buzz-pollinated genera (Vogel, 1978
; Buchmann, 1983
; Faegri, 1986
; Endress, 1994
), including Ardisia, Lysimachia, and Cyclamen, three close relatives of Dodecatheon in the Myrsinaceae (sensu Källersjö et al., 2000
).
While the solanoid flowers of Dodecatheon make them instantly recognizable at anthesis, sterile, pre-anthetic, and fruiting individuals of Dodecatheon are strikingly similar to members of Primula subgenus Auriculastrum (ca. 30 spp. from North America, Pacific coastal Asia, and Europe). Both genera have members that grow in moist alpine meadows and streamsides, have chromosome numbers of 2n = 44, and produce valvate capsules on long scapes arising from a rosette of fleshy, lance-shaped leaves with involute vernation. In an often quoted observation, Thompson (1953
, p. 75) noted that Dodecatheon jeffreyi and Primula parryi are "virtually indistinguishable when the corollas and inserted anthers are removed." Recent authors (Thompson, 1953
; Wendelbo, 1961b
; Richards, 1993
, 2002
; Holmgren, 1994
) have taken this similarity as evidence for a close phylogenetic relationship between Dodecatheon and Primula subgenus Auriculastrum (particularly Primula section Parryi).
This view that similarities between Dodecatheon and Primula subgenus Auriculastrum have shared evolutionary origins is bolstered by the results of phylogenetic analyses of cpDNA (Källersjö et al., 2000
; Mast et al., 2001
; Trift et al., 2002
) and the internal transcribed spacers of the nuclear ribosomal DNA (Martins et al., 2003
). In each of these studies, the single sampled exemplar of Dodecatheon is resolved as most closely related to subgenus Auriculastrum. In the only molecular study to date to sample each of the four sections of subgenus Auriculastrum, Mast et al. (2001)
found weak support (66% bootstrap frequency) for an unexpected sister relationship between Dodecatheon and the monotypic Primula section Suffrutescens (Richards, 2002
). Primula suffrutescens, a sub-shrubby species with wedge-shaped leaves from rock crevices and granite scree in the Sierra Nevada of California, lies close to the Pacific Northwest center of diversity for Dodecatheon, yet it shares fewer morphological and ecological similarities with Dodecatheon than does Primula section Parryi.
Unsampled in these previous molecular studies is a section of Primula that could modify our views on the proper circumscription of Primula subgenus Auriculastrum and thus the evolutionary context for the origin of Dodecatheon. This section, the Asian Primula section Amethystina (eight spp.), has fleshy, toothed leaves and globose capsules that are reminiscent of subgenus Auriculastrum (Smith and Fletcher, 1942
). Furthermore, uncertainty remains regarding its leaf vernation, because published observations (Smith and Fletcher, 1942
, 1950
; Wendelbo, 1961a
; sketches by J. Haldova in Halda, 1992
) are conflicting and inconclusive (Mast et al., 2001
). Leaf vernation has seldom changed state in Primula (Mast et al., 2001
), and it could support the membership of section Amethystina in the revolute subgenus Aleuritia, in which it has been placed by most recent authors (Wendelbo, 1961a
; Fenderson, 1986
; Halda, 1992
; Richards, 1993
), or in the involute subgenus Auriculastrum, in which it was placed by Richards (2002)
. Inclusion of section Amethystina in subgenus Auriculastrum would bridge an existing biogeographic gap between that subgenus and the center of greatest diversity for Primula in the Himalayan Mountains (Mast et al., 2001
).
The phylogenetic position of Dodecatheon within Primula supports a scenario in which the monomorphic, solanoid flowers of Dodecatheon arose from a heterostylous (specifically, distylous) ancestor. Ninety-one percent of the 430 spp. of Primula are distylous, as are all taxa but one (Primula cuneifolia subsp. saxifragifolia) in Primula subgenus Auriculastrum (Richards, 2002
). In distylous species, populations maintain a genetic polymorphism that produces two flower morphs, each differing reciprocally in the position of their anthers and stigmas ("reciprocal herkogamy;" Webb and Lloyd, 1986
). In the "pin" morph, the stigma is positioned high and the anthers low in the flower ("approach herkogamy"), whereas the reverse ("reverse herkogamy;" Webb and Lloyd, 1986
) is true in the "thrum" (Darwin, 1862
). Ancillary features, including pollen size and stigmatic papillae length, differ between the two morphs in Primula and other groups (reviewed in Dulberger, 1992
). These morphological differences are often, though not always, accompanied by a sporophytically controlled incompatibility system that largely limits successful crosses to those between different morphs ("heteromorphic incompatibility;" reviewed by Barrett and Cruzan, 1994
). Reciprocal herkogamy reduces pollen wastage (and thus increases male fitness), whereas heteromorphic incompatibility protects against self-fertilization and inbreeding depression (and thus increases female fitness; Barrett, 2002
). Distyly, and the functionally similar tristyly, are currently known in 28 flowering plant families (Barrett et al., 2000
; Barrett, 2002
).
Within Primula, distyly is thought to be controlled by at least three tightly linked loci (Ernst, 1925
, 1936
, 1957
; Lewis, 1949
; Dowrick, 1956
), that are frequently referred to as the "heterostyly supergene" (e.g., Ganders, 1979
; Barrett, 2002
). The style length, stigmatic papillae length, and female mating type are thought to be controlled by locus G, the pollen size and male mating type by locus P (or two separate loci Pp and Pm, respectively; Kurian and Richards, 1997
), and the anther height by locus A. Pins are homozygous recessive for these loci (gpa/gpa) and thrums are heterozygous (GPA/gpa; Bateson and Gregory, 1905
; Lewis and Jones, 1992
). The rarity of homozygote thrums might be due to the presence of recessive sublethal alleles linked to GPA (Kurian and Richards, 1997
; Richards, 1998
). The loci are likely ordered GPA, at least in Primula subgenus Auriculastrum (Kurian and Richards, 1997
), based on the frequency of observed recombinations (Dowrick, 1956
; Lewis and Jones, 1992
; but see Charlesworth and Charlesworth, 1979
).
The objectives of this study are to reconstruct the chloroplast phylogeny of Dodecatheon and Primula subgenus Auriculastrum and to use this to address four questions: (1) Are traditionally recognized taxa (Dodecatheon, Primula subgenus Auriculastrum, and sections therein) monophyletic? (2) If Dodecatheon is nested in Primula, does our increased character sampling support a sister relationship between it and the morphologically and ecologically similar Primula section Parryi? (3) Is the enigmatic, and previously unsampled, Asian Primula section Amethystina nested in the extra-Himalayan Primula subgenus Auriculastrum? (4) What new insights does the phylogeny bring to our understanding of the buildup of floral traits associated with the buzz pollination of Dodecatheon, including its monomorphy and the morphology of its corolla and androecium?
|
MATERIAL AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) of the remaining section Auricula, two species from the enigmatic section Amethystina, and the other major clades of /Primula. Mast et al. (2001)
defined /Primula as all descendents of the most recent common ancestor (MRCA) of genus Primula using the "clademark" convention of Baum et al. (1998a). In addition, we sequenced the matK gene in five additional genera of the Primulaceae that are outside of /Primula; we did not expand the taxonomic sampling for the six noncoding datasets to minimize the number of alignment-ambiguous positions in them. Prior cpDNA studies (Källersjö et al., 2000
; Mast et al., 2001
; Trift et al., 2002
) strongly support the position of two of these five additional genera (Androsace and Douglasia) as sister to the remaining sampled taxa in the Primulaceae. Consequently, we used these two genera for outgroup comparisons in the analysis of our matK data set. We present the results that include only the 70 taxa in /Primula with a basal trichotomy, as supported by the results of Mast et al. (2001)
.
DNA extraction, amplification, and sequencing
We extracted total genomic DNA from 20–30 mg of fresh leaf material (dried mass after lyophilization), silica-dried material, or herbarium material. Fresh material was lyophilized for 24–48 h in a Lyovac GT 2 (Leybold Vacuum, Cologne, Germany). Dry leaf tissue was disrupted with glass beads using a Retsch MM 2000 shaker (Retsch, Haan, Germany) at an amplitude of 80 for 2 min. We used the DNeasy Plant Mini kit (Qiagen, Valencia, California, USA) for fresh and silica-dried material and an SDS extraction protocol (Eichenberger et al., 2000
) for herbarium material.
Each DNA region was amplified using the polymerase chain reaction (PCR; Mullis and Faloona, 1987
): the matK gene with primers 1F and 1R of Sang et al. (1997)
, the rpl16 intron with primers F71 of Jordan et al. (1996)
and R1516 of Baum et al. (1998a), the rps16 intron with primers F and R2 of Oxelman et al. (1997)
, the trnL intron and trnL/F spacer with primers c and f of Taberlet et al. (1991)
, the trnS/G spacer with primers trnS(GCU) and trnG(UCC) of Hamilton (1999)
, and the trnT/L spacer with primers a and b of Taberlet et al. (1991)
. The most effective thermal cycling program proved to be 34 cycles of 0.5 min at 95°C, 1 min at 52°C, and 1.7 min at 72°C, with a terminal extension of 10 min at 72°C. A TGradient thermocycler (Whatman Biometra, Göttingen, Germany) performed all PCR reactions. To detect successfully amplified DNA and the possible contamination of negative controls, we examined PCR products on agarose gels. We purified successful PCR reactions with the QIAquick PCR purification kit (Qiagen).
A Gene Amp PCR system 9700 (Perkin Elmer, Boston, Massachusetts, USA) performed cycle-sequencing reactions that we prepared with the ABI PRISM dye terminator cycle sequencing ready reaction kit (Perkin Elmer). The primers used in the sequencing reactions were the same as in the amplifications with the addition of the internal matK primer 3F (Sang et al., 1997
). We cleaned the sequenced products with microspin G-50 columns (Amersham Biosciences, Piscataway, New Jersey, USA) or 96-well multiscreen filtration plates (Millipore, Billerica, Massachusetts, USA) to remove excess dye terminators before we ran them out on an ABI Prism 377 or 3100 DNA sequencer (Perkin Elmer). To detect mistakes and correct uncertainties in the computer-generated sequence, we compared aligned trace-files in Sequencher 3.0 (Gene Codes Corporation, Ann Arbor, Michigan, USA).
Defining substitution and indel characters
We reviewed each variable position in the alignment twice in Sequencher 3.0 to confirm that base calls were consistent at informative positions. In the six noncoding regions, nucleotide (nt) substitutions and insertion and/or deletion (indel) events occasionally combined to produce a region that could be aligned reasonably in more than one way. This occurred most frequently when poly-N stretches of variable length were embedded with variant nts. We segregated the nts in those regions and in regions where two or more nonidentical insertions arose independently into invariant alignment positions with gaps so that informative characters were not produced. In the alignment of the matK gene, we inserted gaps in multiples of three in positions that maintained the reading frame. The aligned matrices are available at TreeBase (www.treebase.org).
We determined the borders of each region from comparisons with the complete cpDNA sequence of Nicotiana tabacum (GenBank accession NC_001879). Each data set represents the complete sequence for each region and does not include portions of adjacent regions, with one exception. The rpl16 intron data set is missing 50 nts on the 5' end of the intron due to the position of the priming site.
We coded indels as additional characters for maximum parsimony (MP) analyses if they were informative, bordered by stretches of unambiguously aligned nts, and could be described as binary (presence/absence) characters. We coded insertions as the same state if they were the same size and identical (
2 nts long) or had no more than one nt substitution (>2 nts long) and did not appear to be duplications of different parts of the adjacent sequence. We did not code single-nt indels if they were adjacent to strings (>1 nt long) of the same nt (e.g., AAA vs. AAAA). Other researchers have also excluded this type of indel because it may arise from experimental error (Downie et al., 1998
; McDade and Moody, 1999
), or they have noted its evolutionary lability (Small et al., 1998
). When taxa had long deletions that stretched over positions containing coded indels, we coded those taxa as uncertain for the intervening indels.
Phylogenetic reconstruction
We used the MP optimality criterion to reconstruct the chloroplast phylogeny of the 70 intensively sampled taxa with each of the six noncoding regions separately and with these six together in combination with the matK gene. We also used MP to reconstruct the chloroplast phylogeny for these 70 taxa plus the five additional genera sampled in the matK data set. PAUP* 4.0b10 (Swofford, 2002
) performed all MP calculations. For the MP analyses and the calculation of the consistency index (CI; Kluge and Farris, 1969
) and the retention index (RI; Farris, 1989
), we included only informative nt and indel characters and weighted all of the characters and character state transitions equally. The heuristic MP searches employed 10 random addition sequences and tree bisection-reconnection (TBR) branch swapping.
To ascertain the degree of support for branches in the shortest MP trees of the 70-taxa, seven-region analysis and the 75-taxa, matK analysis, we used PAUP*4.0b10 to calculate nonparametric bootstrap frequencies (Felsenstein, 1985
) in 1000 replicates. The MP analyses of each replicate employed 10 random addition sequences and TBR branch swapping with the maximum number of saved trees per replicate set at 5000.
To further ascertain the degree of support for branches in the shortest MP trees of the 75-taxa, matK analysis, we calculated Bayesian posterior probabilities with Mr. Bayes 3.0b4 (Huelsenbeck and Ronquist, 2001
). We used hierarchical likelihood ratio tests (hLRTs) and the Akaike Information criterion (AIC) in MrModeltest 1.1b (J. A. A. Nylander, Department of Systematic Zoology, Uppsala University, Sweden, unpublished program) to choose the adequately parameter-rich model of sequence evolution for the matK gene from among 24 models. MrModeltest is a simplified version of Posada and Crandall's (1998)
Modeltest 3.06 that only considers those models that can be implemented in Mr. Bayes. We used the chosen model for the maximum likelihood (ML) calculations involved in hypothesis testing as well.
For the Bayesian analysis, Mr. Bayes 3.0b4 ran one cold and three heated chains for 1 x 106 generations in a Markov Chain Monte Carlo (MCMC). Mr. Bayes sampled the cold chain of the MCMC every 1000th generation, and we plotted the likelihood values of these sampled trees in Microsoft Excel to determine the point at which the chain reached stationarity. In PAUP*4b10, we constructed a majority rule consensus of the trees sampled following this initial "burn-in" period to determine posterior probabilities for the internal branches. Mr. Bayes performed two separate MCMC runs using these settings, and we compared the posterior probabilities calculated with each of them.
Hypothesis testing
We tested the significance of the evidence against prior hypotheses that were not supported by the cpDNA data in an MP framework using the winning sites test (WS; Prager and Wilson, 1988
) and in a ML framework using the Shimodaira-Hasegawa test (SH; Shimodaira and Hasegawa, 1999
). We used the WS test for both the 70-taxa, seven-region data set and the 75-taxa, matK data set, but we used the SH test only with the 75-taxa, matK data set. For the tests, we compared a randomly chosen tree found in an unconstrained MP search with a randomly chosen tree found in an MP search constrained to recover the feature of interest. PAUP*4b10 performed the MP searches as described with the exception that the constrained searches involved 100 (rather than 10) random addition sequences. We included only parsimony informative nt and indel characters in the constrained and unconstrained MP searches and in the WS test, but we included all aligned nt positions (and only nts) for the SH test. For the one-tailed SH test, PAUP*4b10 used a distribution derived from 1000 resampling estimated log-likelihood bootstrap replicates.
Pollen size measurements
We collected pre-anthetic flowers of Dodecatheon alpinum subsp. majus (Feller & Mast 592; all vouchers in Z), D. frigidum (Feller & Mast 531), D. pulchellum subsp. pulchellum var. alaskanum (Feller & Mast 510), Primula suffrutescens (thrum and pin; Feller & Mast 577), and P. cuneifolia subsp. saxifragifolia (Feller & Mast 500) in FAA. After one month, we transferred the material to 70% ethanol. For each taxon or flower morphology, we measured the size of 100 pollen grains from one individual along the longest axis using an optical micrometer at 400x magnification. With the exception of the ellipsoidal pollen of Dodecatheon alpinum subsp. majus, the taxa have nearly spherical pollen grains (Wendelbo, 1961b
).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
parsimony, using the topologies of Figs. 1 and 3, reconstructs 25 of these (54%) as insertions, 17 (37%) as deletions, three as either insertions or deletions (depending upon the resolution of the basal trichotomy in /Primula), and one as an insertion that is later deleted. Of the 47 indels, eight (17%) are homoplasious on Figs. 1 or 3.
|
|
|
Phylogenetic reconstructions
The strict consensus of the 28 shortest MP trees found for the 75-taxa sampling of matK data is shown in Fig. 1 with the tree statistics. /Primula is resolved as monophyletic in the strict consensus (98% bootstrap frequency and 100% posterior probability), as is the clade composed of Primula subgenus Auriculastrum and Dodecatheon (100% bootstrap frequency and 100% posterior probability). Of the 47 branches resolved in this strict consensus, 35 (74%) have strong bootstrap support (frequency greater than 70%). The frequency range is 54–100%, and the mean is 85%.
MrModeltest 1.1b chose the SYM substitution model (Zharkikh, 1994
) as adequately parameter-rich using both the hLRT and AIC approaches. The two separate MCMC runs by Mr. Bayes 3.0b4 with the 75-taxa sampling of matK and the SYM substitution model each appeared to reach stationarity after 25 000 generations (further significant changes in the likelihood score beyond 1 x 106 generations cannot be ruled out), and thus we excluded the first 26 sampled trees. PAUP*4.0b10 calculated the majority rule consensus of the remaining 975 trees for each run, and we took the frequency of the branches in these consensuses as their posterior probabilities. The posterior probabilities for 43 of 47 branches were identical in each of the two MCMC runs and only differed by 1% for the four remaining branches. The posterior probabilities calculated in the first MCMC run are shown beneath the branches in Fig. 1. The range of posterior probabilities for the branches resolved in the strict consensus of the shortest MP trees is 85– 100%, and the mean is 99%.
The strict consensus of the shortest MP trees found when analyzing each of the six noncoding regions separately are shown in Fig. 2 with the tree statistics. The ranges of CI values (0.716–0.794) and RI values (0.889–0.938) are similar across the individual analyses. Only seven of the branches (marked with an asterisk in Fig. 2; three in the trnL intron consensus and four in the trnL/F spacer consensus) are incongruent with the topology found in the strict consensus for the combined seven-region data set (Fig. 3). Because chloroplast inheritance is predominantly uniparental in angiosperms (Corriveau and Coleman, 1988
; Harris and Ingram, 1991
; Morgenson, 1996
) and recombination is rare when biparental inheritance does occur (Birky, 1995
), the entire cpDNA is expected to have a shared history. Thus, we ascribe the minor incongruence observed to homoplasy amongst the small number of parsimony-informative characters in the trnL intron and trnL/F spacer (50 and 85 characters, respectively, vs. 112–327 characters in any of the remaining regions; Table 1).
|
Hypothesis testing
We used the SH and WS tests to determine whether or not the cpDNA data significantly rejected four topologies: (1) Dodecatheon sister to /Primula, (2) Primula section Amethystina sister to the clade composed of Primula subgenus Auriculastrum and Dodecatheon, (3) Primula section Parryi sister to Dodecatheon, and (4) a monophyletic Dodecatheon section Purpureo-tubulosa. We considered the first topology listed above solely with the expanded 75-taxa, matK data set. All of the alternative topologies but one are rejected in all of the tests at P 0.05. The one that is not rejected in any of the tests at this threshold is Primula section Parryi sister to Dodecatheon. However, it comes close to the threshold in two of the three tests (P = 0.054 and 0.0654; Table 3). The high P value (0.5000) for the test of this topology using the WS test on the expanded matK data set is due to the small number of characters (just two) relevant to the test.
|
; Fig. 4). Differences between his and our measurements might be explained by slight pollen swelling in glycerol jelly, which he used (Wendelbo, 1961b
; Moore et al., 1991
), and/or slight pollen shrinkage in 70% ethanol, which we used. These might be the reasons for his larger size measurement of D. frigidum (15 µm; Fig. 4), though his measurement of D. alpinum is low within the range we observed for that species.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Wendelbo, 1961b
) and Primula (Richards, 1993
, 2002
), as well as all previous molecular phylogenetic studies that sampled both genera using cpDNA (Källersjö et al., 2000
; Mast et al., 2001
; Trift et al., 2002
) or nuclear DNA (Martins et al., 2003
) characters, have supported a close relationship of Dodecatheon to Primula subgenus Auriculastrum. The current cpDNA data set (ca. 6.8 kb) is the largest sampling of molecular characters and the most complete sampling of taxa in Dodecatheon and Primula subgenus Auriculastrum to date. It significantly rejects the divergence of Dodecatheon prior to the MRCA of /Primula (Table 3), and it strongly supports Dodecatheon as a descendent of the MRCA of Primula subgenus Auriculastrum (high posterior probabilities of Fig. 1 and high bootstrap frequencies of Fig. 3). This has significant implications for the evolution of the solanoid, buzz-pollinated flowers of Dodecatheon. We will discuss this at length after a consideration of the additional phylogenetic relationships reconstructed here (Figs. 1 and 3). Transfer of the species of Dodecatheon to their own section in Primula subgenus Auriculastrum merits consideration, but it will be discussed elsewhere.
Phylogenetic relationships in Dodecatheon
The monophyly of Dodecatheon is strongly supported in the cpDNA phylogeny (100% posterior probability, Fig. 1; 97% bootstrap frequency, Fig. 3), and previous suggestions to the contrary are unknown in the literature. In contrast, the basal split in Dodecatheon that leaves Dodecatheon section Purpureo-tubulosa paraphyletic is novel and has not been previously suggested. However, with the one exception discussed later (the putative allopolyploid D. poeticum), it cleanly splits the genus into a clade with rugose anther connectives and a clade with smooth connectives (Figs. 4 and 5). Smooth connectives in Primula (Fig. 5) support the rugose state as the derived condition and thus the monophyly of the former clade. We are as yet unaware of any morphological synapomorphies supporting the monophyly of the clade with smooth connectives, but the support for it in the cpDNA results is high (100% posterior probability, Fig. 1; 89% bootstrap frequency, Fig. 3).
|
, who circumscribed the three sections of Dodecatheon, noted that anther connective rugosity is useful taxonomically in Dodecatheon, but he maintained the section Purpureo-tubulosa as an assemblage with both smooth and rugose anthers because of the operculate capsules and similar seedling development of its members. The other sections of Dodecatheon (Thompson, 1953
) and the species of Primula subgenus Auriculastrum (Richards, 2002
) have valvate capsules, and thus the operculate character state appears to provide a synapomorphy for Dodecatheon section Purpureo-tubulosa, in conflict with anther connective rugosity and the cpDNA phylogeny. We have not examined seedling development in closely related species of Primula, and thus we cannot comment on the polarity of this character and its implications for the monophyly of section Purpureo-tubulosa.
In his extensive biosystematic study of Dodecatheon, Thompson (1953)
split section Purpureo-tubulosa into two groups, one that might prove to be monophyletic with further character sampling (the Dodecatheon hendersonii complex) and one that is not monophyletic in the cpDNA phylogeny (the Dodecatheon frigidum complex; Fig. 3). He did this primarily based on the presence or absence of a well-developed filament tube and the species' geographic distributions. Species of the Dodecatheon hendersonii complex (four spp.; D. subalpinum is unsampled here) possess a well-developed filament tube (as do other taxa in both of the major clades of Dodecatheon), and their Pacific coastal distributions overlap in central California (Thompson, 1953
; Cholewa and Henderson, 1993
). In the combined cpDNA phylogeny (Fig. 3), D. hendersonii subsp. hendersonii is in a trichotomy with a clade composed of the remaining, sampled members of the D. hendersonii complex and with the monophyletic section Capitulum. The trichotomy is not resolved in any of the 76 shortest MP trees for the combined data.
In contrast, the two species of the paraphyletic Dodecatheon frigidum complex are resolved as the earliest diverging species in the two major clades of Dodecatheon. They both lack a filament tube, but this is the ancestral character state for Dodecatheon (it is absent in the closely related species of Primula) and thus is not a synapomorphy. These two species are also disjunct in their distributions, with D. conjugens occurring in the mountains of the northwestern US and adjacent Canada (Thompson, 1953
; Cholewa and Henderson, 1993
) and D. frigidum occurring in Alaska, adjacent Canada, and across the Bering Strait in Siberia (Thompson, 1953
; Hultén, 1968
). Thompson's (1953
, p. 101) observation that the two species are both "continental. . . in the sense that they avoid the maritime areas of the Pacific Coast adjacent to their ranges" in support of grouping them together ignores the maritime range of D. frigidum along the Bering Sea and Arctic Ocean. Thus, like the absence of a filament tube, we do not consider their geographic distributions to conflict with anther rugosity and the cpDNA phylogeny.
The monophyly of section Capitulum (three spp.) is strongly supported (100% posterior probability, Fig. 1; 98% bootstrap frequency, Fig. 3) in the cpDNA phylogeny. Its monophyly is further corroborated by its unique possession (within Dodecatheon) of the following derived character states: (1) enlarged stigmatic papillae, (2) angular seeds with thin membranes along the edges, and (3) first true leaves arising from the base of the cotyledons (Thompson, 1953
).
However, within section Capitulum, a reconsideration of species circumscriptions appears warranted (Figs. 1 and 3). Differentiation of the species in the section (D. jeffreyi, D. redolens, and D. alpinum) hinges upon flower merosity (four- and five-merous, solely four-merous, or solely five-merous, respectively), capsule dehiscence (usually with an irregular operculum and occasionally valvate in D. jeffreyi or solely valvate in the other two), pubescence (glandular-pubescent to glabrous, solely glandular-pubescent, or solely glabrous, respectively), and features of the corolla tube (completely yellow tube covering the anther bases in D. redolens or tube maroon at base and not covering anther bases in the other two; Thompson, 1953
). Dodecatheon jeffreyi is the most widespread of the three species, intergrades with the other two where they are sympatric (Thompson, 1953
; Cholewa and Henderson, 1993
; but see Hall, 1912
and Clausen et al., 1940
), and encompasses the morphological variation in the other two (with the exception of the corolla tube character mentioned).
Given this, it seems reasonable to expect that the other two taxa are derived from D. jeffreyi once (D. redolens, if the point of corolla tube reflexion and tube coloration are viewed as synapomorphies) or perhaps many times (D. alpinum, if it is a high elevation ecotype of D. jeffreyi). Both D. redolens and D. alpinum have been previously recognized as varieties of D. jeffreyi (by Hall, 1901
and Gray, 1886
, respectively), though not recently. The accessions of D. jeffreyi and D. alpinum that are placed together in the phylogeny were collected from locations geographically closer to each other than the second accession assignable to the same species. Dodecatheon jeffreyi subsp. jeffreyi (Feller & Mast 580) and D. alpinum subsp. majus (Feller & Mast 592) were both collected in northern California (Siskiyou and Plumas counties, respectively); D. jeffreyi subsp. jeffreyi (Feller & Mast 630) and D. alpinum subsp. majus (Feller & Mast 631) were both collected in central California (Yosemite National Park).
The monophyly of section Dodecatheon (five spp.) is strongly supported (100% posterior probability, Fig. 1; 100% bootstrap frequency, Fig. 3) in the cpDNA phylogeny (Fig. 3), and the undescribed species from the Pacific Northwest can be considered as a member. Its monophyly is further corroborated by its unique seedling development (within Dodecatheon), in which the origin of the first true leaves is above ground from the hypocotyl (Thompson, 1953
).
However, as in section Capitulum, species circumscriptions within section Dodecatheon (Figs. 1 and 3) merit a reexamination. Dodecatheon dentatum appears to be a paraphyletic species with respect to other members of section Dodecatheon and the undescribed species. Features cited by Thompson (1953
, p. 121) to uniquely distinguish D. dentatum from other species in the genus include its consistently white corolla limbs and its subulate anthers. Dodecatheon dentatum subsp. utahense has pink or violet limbs (Holmgren, 1994
; he recognized it as a variety carved out of D. dentatum var. ellisiae), and Thompson (1953
, p. 151) reports both subulate and lanceolate anthers in D. dentatum subsp. ellisiae (the latter type is the anther type seen in the remaining members of section Dodecatheon). Thus, the combination of white corolla limbs and subulate anthers would only consistently distinguish D. dentatum subsp. dentatum and some specimens of D. dentatum subsp. ellisiae from other taxa of Dodecatheon. Dodecatheon dentatum subspp. dentatum, utahense, and ellisiae are geographically disjunct (by more than 500 km between any two of them) in the Pacific Northwest, Utah, and Arizona and New Mexico, respectively. The genetic distance between the subspecies of D. dentatum (up to 0.7% using the Hasegawa et al. [1985]
model of sequence evolution for the combined seven-region data set) is in striking contrast to that seen between the four remaining species in section Dodecatheon (up to 0.1%). It is somewhat surprising that the undescribed species from the North Coast Range (and possibly also the Siskiyou Mountains; Chambers, 1999
) is resolved as most closely related to subsp. utahense, rather than the geographically closer subsp. dentatum from the Pacific Northwest. However, like subsp. utahense, and unlike subsp. dentatum, the undescribed species has a non-white (red-purple; Chambers, 1999
) corolla. If population-level sampling corroborates the genetic distinctiveness of the subspecies of D. dentatum, and the paraphyly of D. dentatum relative to the rest of section Dodecatheon, then the subspecies are best elevated to specific rank.
Among the remaining species of section Dodecatheon is D. poeticum, a species with both smooth and rugose anther connectives in a clade with smooth connectives (Fig. 4). A possible interpretation for the presence of rugose connectives in D. poeticum would involve two independent origins of the rugose phenotype in D. poeticum and in the MRCA of the large clade with rugose connectives. However, D. poeticum (2n = 88; Thompson, 1953
) is likely an allopolyploid with D. hendersonii subsp. hendersonii (2n = 44; from the clade with rugose connectives) and D. pulchellum subsp. cusickii (2n = 44, 88; Suttill and Allen, 1992
; from the clade with smooth connectives) as parents (Thompson, 1953
).
Dodecatheon hendersonii subsp. hendersonii is common west of the Cascade Mountains and extends up the Columbia River Gorge. Dodecatheon pulchellum subsp. cusickii is common to the east of the Cascade Mountains and extends down the Columbia River Gorge. Dodecatheon poeticum occurs where these two taxa meet, and the phenotype of D. poeticum appears to be a chimera of the two parents (Thompson, 1953
). Given the position of D. poeticum in the combined cpDNA phylogeny, D. pulchellum subsp. cusickii was likely the maternal parent (for at least the allopolyploid event sampled here; there might have been more than one to produce the heterogeneous D. poeticum). Thus, the genetics underlying the rugose phenotype is likely of common origin in all species with rugose connectives in Dodecatheon, with a transfer of it to D. poeticum during the polyploidization event(s). Suttill and Allen's (1992)
report of an accession of D. poeticum with 2n = 44 is contrary to this and merits further examination.
A phylogeny reconstructed using a bi-parentally or paternally inherited molecular marker will provide a further test of this hypothesis for D. poeticum and could provide insights into the degree to which extant taxa of Dodecatheon are reproductively isolated and the extent to which hybridization and introgression and allopolyploidization have played roles in their evolutionary history. This additional data will be critical to making informed taxonomic realignments in the group.
Phylogenetic relationships in Primula subgenus Auriculastrum
The sister relationship between Dodecatheon and Primula section Suffrutescens, first observed by Mast et al. (2001)
and strongly supported here with increased taxon and character sampling (100% posterior probability, Fig. 1; 100% bootstrap frequency, Fig. 3), is unexpected given the greater morphological and ecological similarities between members of Dodecatheon and Primula section Parryi. The special attention drawn to the similarities between Dodecatheon jeffreyi and Primula parryi by Thompson (1953
, p. 75) is warranted. Each grows in moist mountain meadows and streamsides and has long (up to ca. 50 cm) scapes that arise from a rosette of fleshy, oblanceolate leaves and support an umbel of ca. 5–20 flowers with pedicels of similar length (ca. 3–5 cm) at flowering (Thompson, 1953
; S. Kelso, unpublished manuscript). These features differ from the smaller P. suffrutescens, a mat-forming species of rocky alpine slopes, having shorter, fleshy, cuneate leaves and fewer flowers on shorter pedicels and scapes (Kelso, 1991
). Chromosome numbers (2n = 44), also cited as similar between D. jeffreyi and P. parryi, are, however, shared between Dodecatheon and both Primula sections Parryi and Suffrutescens (Fig. 4). One might conclude that the morphological features shared by D. jeffreyi and P. parryi are the ancestral states for their MRCA and that P. suffrutescens simply represents divergence from it in its rocky, high alpine habitat. However, morphologies similar to P. suffrutescens occur in sections Cuneifolia (both species) and Auricula (e.g., P. hirsuta and P. minima) and thereby complicate the reconstruction of ancestral states in the group.
The clade formed by Dodecatheon and Primula sections Suffrutescens and Parryi (21 spp. in total) is poorly supported in the cpDNA phylogeny (37% bootstrap frequency, Fig. 3). However, members of the clade have similar distributions and chromosome numbers. This is the largest clade (21 spp.) in the Primulaceae (sensu Källersjö et al., 2000
) for which all members can be found in North America (principally western North America; Fig. 4). Furthermore, Primula sections Suffrutescens and Parryi and the earliest diverged species of Dodecatheon have chromosome numbers of 2n = 44 (Fig. 4), as opposed to 2n = 66 in the European section Auricula (Richards, 2002
) or 2n = 22 in the Pacific coastal Cuneifolia (Kelso, 1991
).
Ours is the first published phylogenetic study to sample more than two taxa of the western North American Primula section Parryi, and we sampled all eight taxa recognized as members of the section by Holmgren and Kelso (2001)
. The section offers a model for understanding the effects of Pleistocene climatic and vegetation changes on the flora of the Intermountain Region of western North America (Holmgren and Kelso, 2001
), but much of the elucidation of this will come with finer-scale population sampling of the taxa. At present, we will let it suffice to note the strong support for the section's monophyly (85% bootstrap frequency, Fig. 3), the basal split between P. parryi (the most widespread taxon of the section and that species most similar to Dodecatheon) and the remaining taxa, and the possible, though poorly supported (50% bootstrap frequency, Fig. 3) paraphyly of P. cusickiana with respect to P. capillaris.
The monophyly of the two remaining sections, Cuneifolia (as recognized to exclude P. suffrutescens; Richards, 2002
) and Auricula, is strongly supported (100% and 85% posterior probability, respectively, Fig. 1; both at 100% bootstrap frequency, Fig. 3), and all relationships among their sampled members in the seven-region strict consensus (Fig. 3) are consistent with previous studies by Fujii et al. (1995
, 1999
; Cuneifolia) and Zhang (2002
; Auricula). These authors provide abundant observations regarding the implications of the inferred relationships.
Conspicuously absent in the clade of Dodecatheon and Primula subgenus Auriculastrum are the sampled members of Primula section Amethystina (Fig. 3). Mast et al. (2001)
drew special attention to this small section, which can be found in moist, peaty meadows in Asia, because (1) it remained unsampled in that study, (2) its thick, toothed leaves and globose capsules are reminiscent of subgenus Auriculastrum (Smith and Fletcher, 1942
), (3) uncertainty remains regarding the leaf vernation type in the section (Smith and Fletcher, 1942
, 1950
; Wendelbo, 1961a
; sketches by J. Haldova in Halda, 1992
), and (4) its distribution in the Himalayan Mountains and neighboring regions would bridge the gap between the currently recognized distribution of Primula subgenus Auriculastrum and the region of greatest diversity in Primula. The species of section Amethystina uniquely (in Primula) share pitted glands in their leaves (Smith and Fletcher, 1942
), a synapomorphy that supports the section's monophyly.
Our analysis places the two sampled members of this section with P. eximia among the outgroup taxa with strong support (100% posterior probability, Fig. 1; 100% bootstrap frequency, Fig. 3), and a sister relationship between this section and the clade composed of Dodecatheon and Primula subgenus Auriculastrum is significantly rejected by the data (Table 3). Thus, the descendents of the MRCA of subgenus Auriculastrum remain extra-Himalayan in their distribution, and the similarities of subgenus Auriculastrum and section Amethystina appear to be independently derived. The persistent confusion surrounding the leaf vernation type of this section (a character otherwise thought to have undergone few evolutionary shifts in the Primulaceae; Mast et al., 2001
) merits a comparative developmental study.
Dodecatheon represents a breakdown in heterostyly
The likely shift from distylous flowers that offered nectar as a reward in the MRCA of Primula suffrutesens and Dodecatheon to monomorphic, solanoid flowers that offered pollen as a reward in the MRCA of Dodecatheon raises at least two interesting questions: (1) Which alleles were fixed in the heterostyly linkage group of Dodecatheon, if it persists in that genus? and (2) What further morphological changes were necessary to get from one floral type to the other?
The evolutionary steps traditionally posed for the shift from self/intra-morph incompatible heterostyly to self/intra-morph compatible monomorphy in Primula involve a recombination in the linkage group between the locus responsible for style length and female mating type (G) and the loci responsible for pollen size and male mating type (P; we will discuss this as a single locus for simplicity) and anther height (A) (Ernst, 1955
; Dowrick, 1956
; Charlesworth and Charlesworth, 1979
; Wedderburn and Richards, 1992
; Al Wadi and Richards, 1993
; Richards, 1993
, 1997
). This is followed by the fixation in the population of the recombinant with the long style length of the pin and the recessive female mating type (g) and the high anther height, dominant male mating type, and larger pollen size of the thrum (PA). This allows the pollen of any flower to pollinate any ovule in the population, and it positions the anther and stigma at the same height in the flower (the long "homostyle;" Darwin, 1877
; Path A in Fig. 6). These events are thought to have occurred in most of the 17 sections of Primula in which long homostyles are found (Wedderburn and Richards, 1992
; Richards, 1993
, 1997
). However, there is evidence derived from their incompatibility reactions that some of these might represent monomorphic lineages diverged from the MRCA of Primula prior to the origin of heterostyly in the group (Ernst, 1955
; Wedderburn and Richards, 1992
; Al Wadi and Richards, 1993
; Richards, 1993
, 1997
, 2002
; but see Baker, 1966
). Our observation that the pollen of P. cuneifolia subsp. saxifragifolia is the same size as, or slightly larger than, the thrum pollen of P. suffrutescens is consistent with a recombinatory origin of long homostyly in P. cuneifolia subsp. saxifragifolia. Charlesworth and Charlesworth (1979)
have shown mathematically why other single recombinants (Gpa, GPa, and gpA) in the heterostyly linkage group would be at a selective disadvantage relative to the nonrecombined dominant and recessive alleles or the "long homostyle" alleles. Only once has the "short homostyle" phenotype (presumably the Gpa alleles) been fixed in a taxon of Primula (P. septemloba var. minor; Richards, 2002
).
|
). This is the range of sizes that we observed for the pin pollen of P. suffrutescens (Fig. 4) and smaller than that species' thrum pollen. Dodecatheon alpinum and D. jeffreyi of section Capitulum have ellipsoidal pollen that is 20–24 µm along the long axis, suggesting an increase in pollen size from the smaller size that we infer to be primitive for the genus. This is unexpected, given that small pollen size is commonly described as an adaptation to buzz pollination (e.g., Buchman, 1983
), but it suggests that pollen sizes within the range observed in Dodecatheon (10–24 µm) might be equally suited to the syndrome. Apparently concurrent with this increase in pollen size was an increase in length of the stigmatic papillae, for section Capitulum is unique in Dodecatheon in its possession of an enlarged stigma (Thompson, 1953
).
The origin of monomorphy in Dodecatheon from a heterostylous ancestor requires a shift from self/intra-morph incompatibility to intra-morph compatibility either by overcoming the incompatibility by recombination (path A of Fig. 6) or by at least a partial breakdown in it (path B in Fig. 6). Partial breakdowns of self/intra-morph incompatibility are common in Primula, where 77% of 35 examined species set seed after the self-pollination of pins (Wedderburn and Richards, 1990
). In the extreme case, 61.8% of pin ovules pollinated with self-pollen in P. vulgaris (of Primula subgenus Primula) set viable seed, compared to 70.1% of pin ovules pollinated by thrum pollen (Wedderburn and Richards, 1990
). This incompleteness in the self/intra-morph incompatibility system has permitted a number of true-breeding pin lines (homozygous recessive individuals) in cultivation (Richards, 1993
).
The fixation of the recessive alleles could have arisen by stochastic events (founder events or genetic drift in small populations) or by selection. The MRCA of Dodecatheon likely arose in western North America, where Primula sections Suffrutescens and Parryi and the earliest diverged species of Dodecatheon are found. The extensive climatic, orogenic, and biological changes in this region during the Tertiary and Quaternary are well documented (Graham, 1999
), and we can reasonably assume then-extant populations to have gone through numerous founding events, population size fluctuations, and periods of isolation during that time. Founder events and genetic drift have been hypothesized to play a role in the generation of populations lacking one or two (in the case of tristyly) floral morphs in other heterostylous species (e.g., Barrett et al., 1989
; Husband and Barrett, 1992
; Eckert and Barrett, 1992
) between the constrained and unconstrained trees. P values 
