| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Department of Biological Sciences, Southeastern Louisiana University, Hammond, Louisiana 70402 USA; 3Plant Resources Center and Section of Integrative Biology, The University of Texas, Austin, Texas 78712 USA; 4Department of Biology, Duke University, Durham, North Carolina 27708 USA
Received for publication October 31, 2003. Accepted for publication April 8, 2004.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: Bayesian phylogenetic analysis • Convolvulaceae • Ipomoea • ITS • Markov chain Monte Carlo • molecular phylogenetics • morning glory
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) or up to 1000 species in recent phylogenetic conceptions of the group (Ipomoea and its segregates; e.g., Argyreia, Turbina, Astripomoea, Stictocardia, Lepistemon, and Rivea pro synonymo; Wilkin, 1999
; Manos et al., 2001
; Miller et al., 2002
). Most Ipomoea occur in tropical and subtropical climates throughout the world, but representative elements of the genus are in all known biomes (McDonald, 1991
; Wilkin, 1999
).
Infrageneric classifications of Ipomoea were provided by Choisy (1845)
, Hallier (1893a
, b
), and House (1908a)
. Relationships among Old World Ipomoea species were further refined by van Ooststroom (1953)
, who recognized seven infrageneric taxa in his studies on Asian species. Borrowing liberally from van Ooststroom's concepts, Verdcourt (1957
, 1963
) recognized eight infrageneric taxa in his treatment of African species. American Ipomoea have received more attention than those of the Old World (e.g., House, 1908a
; Matuda, 1963
; Standley and Williams, 1970
; Austin, 1975a
, b
, 1979
, 1997
; McPherson, 1979
; McDonald, 1982
, 1991
; Austin and Huáman, 1996
). Austin and colleagues have provided the most recent and comprehensive treatment of American Ipomoea, recognizing three subgenera within the genus: Eriospermum (Hallier) Verdcourt ex Austin, Ipomoea, and Quamoclit (Moench) Clarke (Austin and Huáman, 1996
; Austin, 1997
; Austin and Bianchini, 1998
). The treatments of Austin are similar in most respects to those of McDonald (1991)
on Mexican Ipomoea species. It is important to point out, however, that both McDonald and Austin have encouraged caution with their systems and have consistently recognized the enormous challenges presented by the study of this group of plants (e.g., McDonald, 1991
; Austin and Huáman, 1996
). This provides one of the motivations for the research program presented here of turning to additional sources of phylogenetic data to attempt to better understand the relationships among morning glories.
The present study focuses on species of the subgenus Quamoclit, as well as the species of section Pharbitis (formerly aligned in subgenus Ipomoea, Table 1). The nested relationship of Pharbitis species within subgenus Quamoclit has been established on the basis of DNA sequence data (Miller et al., 1999
) and is supported by morphological evidence (Wilkin, 1999
). There also are indications of this relationship in a chloroplast DNA RFLP study of McDonald and Mabry (1992)
, though this result does not hold for all species of section Pharbitis included in their study. The Quamoclit group (subgenus Quamoclit, plus section Pharbitis) forms a well-supported clade (Miller et al., 1999
) within the clade /Astripomoeinae (Stefanovic et al., 2003
) based on both separate and combined analyses of internal transcribed spacers (ITS) and waxy sequence data.
|
; Austin and Huáman, 1996
). A number of infrageneric taxa within this species group have been the subject of various revisional studies, including sections Calonyction (Gunn, 1972
), Exogonium (Austin, 1977
; McDonald, 1987
), Leptocallis (McDonald, 1995
), Mina (O'Donell, 1959
), and series Tyrianthinae of section Pharbitis (McDonald, 2001
). The chloroplast DNA restriction fragment length polymorphism (RFLP) study of McDonald and Mabry (1992)
included 31 New World species of Ipomoea, 22 of which were members of the Quamoclit group. Their study identified the morphologically distinct sections Calonyction and Mina as well-supported groups. They also found that three species of Ipomoea section Tricolores form a well-supported monophyletic group, although the placement of I. parasitica (section Tricolores) was problematic. Evidence was provided that series Pharbitis and series Heterophyllae of section Pharbitis are sister taxa, though not closely associated with species of series Tyrianthinae of this same section. Species of section Exogonium and series Tyrianthinae are highly variable (McDonald, 1987
and McDonald, 2001
, respectively). The three Exogonium species sampled by McDonald and Mabry (1992)
do not form a distinct clade, but this might owe in part to a small sample of species in this group. This is also the case for four species of Ipomoea series Tyrianthinae included in their study.
The ultimate goal of the research presented here is to develop a well-resolved phylogeny for the morning glories in the Ipomoeeae (Miller et al., 1999
). The objectives of this specific study are to further examine relationships among the species of Ipomoea subgenus Quamoclit and section Pharbitis. To meet these objectives, we sampled species from sections within subgenus Quamoclit and from series within section Pharbitis and obtained intraspecific samples when possible. We examined the correspondence between a molecular data set constructed with ribosomal nuclear DNA sequences from the 5.8S gene and associated internal transcribed spacers (ITS region) and traditional classifications of this group (Table 1). In particular, we wanted to determine whether or not the sections and series recognized of Quamoclit are supported by these molecular data. The results of the molecular analyses based on the ITS region are compared to earlier molecular studies based on chloroplast DNA restriction site variation (McDonald and Mabry, 1992
), as well as on other recent molecular phylogenetic results (Miller et al., 1999
; Manos et al., 2001
). In carrying out the analyses of the molecular data, a Bayesian phylogenetic analysis is emphasized (e.g., Huelsenbeck et al., 2002
; Miller et al., 2002
). An additional objective of this study was to further develop this approach for the analysis of phylogenetic data. In particular, methods to examine convergence of the Markov chain Monte Carlo (MCMC) sampling are emphasized.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Austin, 1997
; Austin and Bianchini, 1998
). In addition, multiple accessions for several individual species were included, resulting in a total of 68 samples for this investigation (Appendix 1; see Supplemental Data accompanying the online version of this article). Many of the plants used in this study were grown from the seed collection of J. A. M. (Appendix 1). The remaining plants were obtained by R. E. M. from various seed sources (Appendix 1), as well as one sample for this study provided by P. Wilkin (Royal Botanic Gardens, Kew). Vouchers specimens were deposited in one of the following: SLU, TEX (Appendix 1). Two species, Ipomoea cairica and I. sepiaria, were selected as outgroup taxa based on previous, higher-level analyses (Miller et al., 1999
; Manos et al., 2001
).
Molecular methods
Total genomic DNA was obtained using the DNeasy Plant mini kit (Qiagen, Valencia, California, USA) from live plants grown from seed in the Duke University Greenhouse. Extracted DNA is under the care of R. E. M. Molecular methods for PCR (polymerase chain reaction) and sequencing of the internal transcribed spacers of nuclear ribosomal DNA, or ITS region (ITS 1–5.8S-ITS 2), generally followed Miller et al. (1999)
. For the amplification of ITS, we substituted standard PCR components with an Advantage-GC cDNA PCR kit (Clontech, Palo Alto, California, USA), which permitted direct sequencing of PCR products.
Sequences
The sequence data from the ITS region (ITS 1–5.8S-ITS 2) included 58 new sequences and 10 previously published sequences (Miller et al., 1999
; Manos et al., 2001
) (Appendix 1; see Supplemental Data accompanying the online version of this article). The 68-taxon ITS data set was examined for identical sequences. The sequences for seven species included identical sequences for some of the multiple accessions, in which case a single sequence was used to represent these identical sequences. This reduced data set included a total of 54 taxa. Sequences were aligned manually. Fourteen sites were excluded due to ambiguous alignment. Sequences are available from GenBank (accessions AY538275-AY538332), and the aligned Nexus data file is available from R. E. M. (www.selu.edu/Academics/Faculty/rickmiller) and from TreeBase (www.treebase.org).
Phylogenetic analyses
We adopted a general time-reversible model of DNA substitution with among-site rate variation drawn from a gamma distribution (GTR + 
) for the analysis. This model was selected from a comparison of 56 models using the Akaike information criterion (Akaike, 1974
) as implemented in Modeltest version 3.0 (Posada and Crandall, 1998
). A Bayesian phylogenetic analysis was used to examine tree topology, support for clades, and to address specific questions about relationships (MrBayes version 2.0 software; Huelsenbeck and Ronquist, 2001
). The posterior probabilities of the phylogenetic model were estimated as part of the Bayesian analyses using Markov chain Monte Carlo (MCMC) sampling with the Metropolis– Hastings–Green algorithm running four chains, three heated and one cold chain. The analysis used uniform prior distributions for the alpha-shape parameter of the gamma distribution (0–10), proportion of invariable sites (0– 1), rate matrix parameters (1–100), and branch lengths (1–10). A flat prior was used for the topology and a Dirichlet distribution was used for the base frequencies. Unique random starting trees were used for each of 15 separate analyses (see Results). Every hundredth tree was sampled from the MCMC analysis to increase independence of samples.
The MCMC analysis starts at a random tree. The trees from the sample of interest are those within the stationary distribution. There is an initial burn-in period before the MCMC sampling is within the stationary distribution. To determine the burn-in period, both likelihood values and tree lengths were graphed against generation determining the number of generations at which these values reached a plateau.
One of the most important and difficult components of a Bayesian phylogenetic analysis is determining when MCMC analyses have run for enough generations for posterior probabilities to be sufficiently close to their true values, in other words, estimating convergence. Two approaches were adopted in this study to obtain estimates of convergence. In both cases, five separate analyses were carried out for a particular number of generations. Then the number of generations was increased for an additional set of five analyses, increasing the number of generations for sets of analyses until convergence was obtained. One approach used was a heuristic method. In this case, the index of convergence was when consistent results were obtained among five estimates of the parameters of the phylogenetic model. Specifically, we focused on tree topology and Bayesian posterior probabilities of individual clades. The indication of convergence used for posterior clade probabilities was when these values for all nodes of the five trees fell within a range of 3%. In addition, convergence of the parameters of the phylogenetic model were also evaluated more formally by adapting a method of Gelman et al. (1995)
; within-run variation was compared to between-run variation (details presented in Appendix 2; see Supplemental Data accompanying the online version of this article). An index of convergence is when the ratio of these two estimates of variation equal one. We examined log-likelihoods and tree lengths for this index of convergence. The analysis of the data consisted of constructing a 50% majority rule consensus tree from the concatenated set of trees from the final set of five analyses, once convergence had been met by the most conservative criterion.
Application of Bayesian analyses to phylogenetic systematics is still in the exploratory phase, and certain aspects of these analyses are still being evaluated (Huelsenbeck et al., 2002
). Therefore, a separate parsimony analysis was compared to the Bayesian phylogenetic analyses. The parsimony analysis was carried out using weighted parsimony with a six-parameter weighting scheme based on the model of DNA substitution obtained from the Bayesian analyses (gt = 1.0, ct = 4.99, cg = 0.54, at = 1.26, ag = 2.82, ac = 1.71). Heuristic searches were used with 1000 random-addition replicates using MULPARS, TBR, and AMB options as implemented in PAUP* version 4.0b10 (Swofford, 2000
). Branch support was estimated using bootstrap sampling with 1000 pseudoreplicates and 10 random-addition replicates with a full heuristic search.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Bayesian phylogenetic analyses
Burn-in period
A graph of log likelihoods and tree lengths for each generation was used to identify the burn-in period. In a previous study, tree length was found to be one of the most variable parameters in Bayesian phylogenetic analyses (Miller et al., 2002
). This was consistent with the analyses presented here. Log likelihoods reached a plateau at 10 000 generations, whereas tree lengths reached a plateau at 150 000 generations. Therefore, a burn-in period of 150 000 generations was used in the analyses.
Convergence
Sets of five analyses were used to evaluate convergence, running each set for 3, 5, and 7 x 106 (million) generations. The R statistic of Gelman et al. (1995)
was close to one after three million generations for both likelihood values and tree length (Table 2). This index of convergence remained essentially unchanged over the range of runs examined here.
|
Posterior clade probabilities were variable among the set of 5 three-million-generation analyses. Many of the nodes differed by as much as 6%, and three differed by over 30% (Fig. 1). Among the 5 five-million-generation analyses, the range of posterior probabilities were within 3% or less for all nodes. An additional set of 5 seven-million-generation analyses was obtained to ensure the narrow range of posterior clade probabilities detected among the set of five-million-generation runs was a consistent result. The range of posterior probabilities continued to be within 3% or less for all nodes for the latter set of analyses. Taken together, these results indicate that convergence had been met for these data by five million generations and certainly by seven million generations.
|
Bayesian posterior clade probabilities
The 50% majority rule consensus tree of the 342 500 trees provided estimates of posterior clade probabilities. Well-supported nodes (e.g., posterior probabilities >95%) were found throughout the topology for the Ipomoea taxa (Fig. 2). Over 50% of the clades received 100% support. Furthermore, over 60% of the clades were found in 95% or greater of the sampled trees. There also were numerous nodes without strong support, demonstrating the preliminary nature of these molecular data for certain regions of the topology. Twenty-two percent of the clades received less that 60% support, with two clades receiving as little as 23% support, the latter including terminal taxa on short branches (Fig. 3).
|
|
|
Bayesian posterior probabilities vs. bootstrap support
In general, posterior probabilities were greater than bootstrap values (Figs. 2, 4). For one node of 22 for which bootstrap values differed from posterior clade probabilities, the bootstrap value was greater than the posterior probability. In contrast, for 21 of the same comparisons posterior clade probabilities were greater than bootstrap values. Both bootstrap support and posterior probabilities were 100% for 10 nodes. Bootstrap values ranged from 63 to 99%, while the corresponding posterior probabilities were 100% for 15 nodes. However, the relationship between these two measures of support is not merely a scaling difference, as indicated in Fig. 5.
|
) unites sections Mina and Leptocallis, which was found in all 342 500 trees sampled in the Bayesian analysis (Fig. 2). In contrast, this node collapses to a polytomy in the strict consensus of 192 most-parsimonious trees (Fig. 4). Clade 2 (clade 2A–2 of Miller et al., 1999
) includes the species from section Pharbitis with species from sections Calonyction and Tricolores. This clade also was identified in the maximum parsimony analysis (Fig. 4). Species of sections Calonyction, Mina, and Leptocallis are all identified as monophyletic groups with at least 95% support in the Bayesian analyses (Fig. 2). The results from the maximum parsimony analysis are consistent with these findings (Fig. 4). Three species of section Tricolores form a clade with 100% support in both analyses, although species I. parasitica of section Tricolores does not join this clade in the analysis (Figs. 2, 4).
Species of series Pharbitis and series Heterophyllae, both of section Pharbitis, form a well-supported monophyletic group in both the Bayesian analyses (100% support) and maximum parsimony analyses (83% support) (Figs. 2, 4). Species of series Heterophyllae are identified as a clade derived from within subgenus Pharbitis, with the addition of I. purpurea (series Pharbitis) as a member of the Heterophyllae clade. In contrast, the three species of section Pharbitis series Tyrianthinae do not form a distinct clade, nor are they sister to the other Pharbitis species.
The species of section Exogonium do not form a clade, but two species pairs are identified in both analyses; I. dumetorum united with I. purga, and I. expansa united with I. seducta (Figs. 2, 4).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Austin and Huáman, 1996
; Austin, 1997
; Austin and Bianchini, 1998
). These molecular results corroborate the results from previous molecular studies (Miller et al., 1999
; Manos et al., 2001
) and the systematic study of some of these same species based on chloroplast DNA RFLP variation (McDonald and Mabry, 1992
). Congruence between these hypotheses and revisional treatments indicates that the relationships developed in former studies may represent a classification that reflects evolutionary history. This is particularly true for species groups within Ipomoea that are easily recognized based on morphology, such as Ipomoea sections Calonyction, Leptocallis, Mina, and series Pharbitis and series Heterophyllae (section Pharbitis).
This study included taxa that were formerly aligned in two separate subgenera: Ipomoea and Quamoclit (Table 1). Support for these subgenera as distinct clades is not provided by the data. More specifically, species of section Pharbitis (subgenus Ipomoea) were nested within species of subgenus Quamoclit. This result was shown previously by Miller et al. (1999)
with a broader sample of Ipomoea species for both ITS and waxy sequence data, as well as from a combined analysis. Wilkin (1999)
also observed this same result based on a morphological cladistic study of 142 Ipomoea species. The results of McDonald and Mabry (1992)
do not support these two subgenera as distinct clades, but the specific nesting of Pharbitis species within subgenus Quamoclit was not indicated in their study. Within the Quamoclit group, two major clades were identified (Figs. 2, 3). The first clade (clade 1) includes sections Mina and Leptocallis. The second clade (clade 2) includes sections Tricolores, Calonyction, and Pharbitis. The Quamcolit group was identified as clade 2A in Miller et al. (1999)
; while clades 1 and 2 here were identified as clade 2A– 1 and clade 2A–2, respectively, in the previous paper. Unfortunately, no obvious morphological features have been recognized that define the members of these two different groups.
Clade 1
Within clade 1, strong support was found for sections Mina and Leptocallis as monophyletic groups. Specifically, the Mina clade received 100% and 82% support and the Leptocallis clade received 95% and 75% support from Bayesian and maximum parsimony analyses, respectively. Species of section Mina have long been recognized as a distinct group of morning glories and at times have received generic status (e.g., McPherson, 1979
). These species can be clearly defined on the basis of various synapomorphic features: corollas ornithophilous; yellow-, orange-, or red-pigmented; tubes narrow (3–5 mm in diameter) the limb often flaring abruptly, style and stamens exserted, sepals bearing a single, fleshy subterminal appendage; and capsules four-locular (O'Donell, 1959
; McDonald, 1987
, 1993
). Species of section Leptocallis also usually possess a distinctive morphology: pedately dissected laminas (McDonald, 1995
). Another important result from the analysis presented here was obtaining good resolution among the species within section Mina. Only one node of 10 was not well supported (i.e., one clade with 36% support). It is interesting to note that there are approximately 18 red-flowered species within the Quamoclit group (House, 1908a
; MacBride, 1959
; Kearney and Peebles, 1960
; Matuda, 1963
; Radford et al., 1968
; Standley and Williams, 1970
; Adams, 1972
; Gunn, 1972
; Austin, 1975b
, 1982
; Correll and Johnston, 1979
; McPherson, 1980
; Austin and Cavalcante, 1982
; McDonald, 1982
, 1987
, 1993
, 1994
, 1995
; Eckenwalder, 1989
; Wagner et al., 1990
; Wilkin, 1995
). Of these, 12 are members of section Mina. Zufall and Rausher (2004)
recently demonstrated that red flowers among morning glory species is due to unique genetic changes in the anthocyanin biosynthetic pathway. Included in their study was the characterization of the genetic basis of red flowers in Ipomoea quamoclit, a member of section Mina. The results demonstrate that I. quamoclit produces pelargonidin-based anthocyanin pigments and that two genetic changes can account for red flowers in this species. One change is down regulation of transcription leading to decreased expression of one of the protein-coding genes, flavonid 3'-hydroxylase (f3'h). In addition, substrate specificity of dihydroflavonol reductase (DFR) for the precursor dihydrokaempferol leads to the production of pelargonidin-based anthocyanin pigments and also could account for red flowers in I. quamoclit. It is not clear, however, what the order of these two mutations was in the evolution of red flowers in this species. This line of investigation, coupled with the well-resolved relationships among the Mina species, could be used to determine whether the production of red flowers in section Mina is due to these specific mutations, as well as to determine the possible order of the origin of these mutations.
Clade 2
Within clade 2 100% support was observed for section Calonyction based on both methods of analysis (Bayesian and parsimony). McDonald and Mabry (1992)
also obtained support (80%) for the monophyly of these three species (I. muricata (L.) Jacq. = I. turbinata Lag.). The species of section Calonyction as presently defined are morphologically unique (Austin and Huáman, 1996
; Austin, 1997
). Ipomoea alba is typical of species in this taxon (Verdcourt, 1957
; Austin, 1997
), having a perennial habit, twining vines, stems armed with numerous herbaceous warts, sepals unequal bearing a long thick awn, fruit with greatly enlarged pedicels, corollas white, salverform, and fragrant, stigmas exserted, capsules two-celled, four-seeded, and seeds large (10–12 mm long) (Gunn, 1972
; Austin, 1997
). Within section Calonyction, I. alba and I. santillanii were found to be sister species, with I. turbinata sister to these taxa. Both I. alba and I. santillanii are white-flowered taxa, while I. turbinata is a smaller-flowered, autogamous species with pigmented flowers. Whether the white flowers of I. alba and I. santillanii represent independent loss-of-function mutations resulting in white flowers remains to be determined.
According to Austin (1997)
, Ipomoea section Pharbitis includes three series: Heterophyllae, Pharbitis, and Tyrianthinae. Section Pharbitis is distinguished on the basis of foliose, hispid sepals (McDonald and Mabry, 1992
). While there is no support for the monophyly of all three series within section Pharbitis, there is strong support (100% from Bayesian analyses and 83% support from maximum parsimony analyses) for a clade composed of species that present a three-part gynoecium—those of series Pharbitis and series Heterophyllae (section Pharbitis) (McPherson, 1979
; Manos et al., 2001
). Separation of species in series Pharbitis from those of series Heterophyllae is well supported in the Bayesian analyses with the Heterophyllae clade receiving 96% support. However, Ipomoea purpurea, traditionally placed within series Pharbitis, may belong to series Heterophyllae. The similarities between I. purpurea and species of series Heterophyllae have been noted previously (Austin et al., 2001
), though considering this species a member of series Heterophyllae has not been suggested previously. Although this conclusion conflicts with cladistic analyses of section Pharbitis based on morphological criteria (Austin et al., 2001
), we note that independent data using waxy sequences identifies a clade of series Heterophyllae species with 100% bootstrap support from a maximum parsimony analysis that includes I. purpurea within this clade (Miller et al., 1999
).
Ipomoea hederacea and I. nil are two annual species that may be regarded as variants of a singular species (Austin, 1975b
). Morphological distinctions between these two species are based on subtle differences in sepal shape. From the analysis presented here, I. hederacea and I. nil appear to be closely related to I. indica. The resolution among these species is not clear based on these data. These species display a variety of mating systems: I. hederacea is largely a selfing species with almost no anther–stigma separation; I. nil is self-compatible with notable anther–stigma separation (and therefore probably has a mixed mating system); while I. indica is self-incompatible (Martin, 1970
; R. Miller, Southeastern Louisiana University, unpublished data). Another set of closely related species with contrasting mating systems include I. purpurea and I. pubescens. Ipomoea purpurea is a self-compatible species with a mixed mating system (Rausher and Fry, 1993
), while I. pubescens appears to set seed in the bud (R. Miller, Southeastern Louisiana University, unpublished data). It also is noteworthy that the species of Ipomoea series Tyrianthinae (section Pharbitis) are not part of the clade that unites series Pharbitis and series Heterophyllae. In fact, species from series Tyrianthinae do not form a monophyletic group in our analysis. This is consistent with earlier findings by McDonald and Mabry (1992)
and Austin and Huáman (1996)
, who note that series Tyrianthinae is a heterogeneous group of species (McDonald and Mabry, 1992
). Nevertheless, McDonald (2001) recently regarded the Tyrianthinae complex as a monophyletic group based on morphological synapomorphies. These contrasting views of the relationships of these species within subgenus Quamoclit demands resolution with greater sampling of species of Tyrianthinae, as well as additional sources of phylogenetic data.
Three of the four species of section Tricolores included in this study form a well-supported clade. Ipomoea tricolor is a widely distributed species with relatively large flowers, while I. cardiophylla and I. marginisepala are small-flowered species (discussed in detail in McDonald, 1982
). Ipomoea tricolor is a predominately outcrossing species, while I. cardiophylla and I. marginisepala are selfing species. Furthermore, I. tricolor is a widely distributed Mexican species, while I. cardiophylla is distributed in the northern Chihuahuan desert region and I. marginisepala is found in the Argentine desert. These two disjunct desert species may represent two independent origins of selfing species from a widespread species with a mixed mating system (McDonald, 1982
), as suggested by the molecular phylogenetic results presented here. The fourth species of section Tricolores included in this study, Ipomoea parasitica, is not part of this clade. The Bayesian analysis suggested I. orizabensis (section Pharbitis series Tyrianthinae) is sister to I. parasitica (section Tricolores), which is then sister to other species of section Pharbitis, while the parsimony analysis showed I. parasitica as part of a polytomy. McDonald and Mabry (1992)
obtained two different results for the placement of I. parasitica with chloroplast DNA RFLP data depending on the particular analysis used. In contrast, morphological evidence would suggest I. parasitica is closely related to species of either section Tricolores or section Calonyction. Ipomoea parasitica shares the tricolored corollas (blue throat, white limb, and yellow throat) of the other Tricolores species (McDonald, 1982
) and the highly distinctive muricate stems of Ipomoea section Calonyction.
Taxa that fall outside of clade 1 or clade 2 include elements of Ipomoea section Exogonium. The species belonging to Exogonium have traditionally included morning glories with red flowers, salverform corollas, and exserted stamens and stigmas (House, 1908b
). There has been a dramatic redefinition of which species should be included in the section (Austin, 1977
; McDonald, 1987
). The instability of this taxon points to the difficulty in morphologically defining the group. For example, the species Austin and Huáman (1996)
indicate belong to subgenus Quamoclit section Exogonium were placed in three different species assemblages by McPherson (1979)
(his Purga group, Thurberi group, and Tyrianthina group). Wilkin (1999)
also found that section Exogonium was not monophyletic based on morphological data. According to our results, Ipomoea section Exogonium does not form a monophyletic group, but rather a basal grade relative to clades 1 and 2. Within Exogonium, two informal groups have been identified, the purgoid and suffultoid complexes (McDonald, 1987
). There is no correspondence between these groups and the results presented here, supporting McDonald's (1987)
statement that these morphological distinctions may best represent ecological adaptations rather than phylogenetic relationship. Ipomoea seducta and I. expansa were recognized as sister taxa in this study, a result consistent with McDonald and Mabry (1992)
. Section Exogonium, as it is currently defined, includes 20 species with striking floral diversity (McDonald, 1987
). This includes species with various floral shapes as well as the complete spectrum of corolla colors found within morning glories. Well-targeted sampling within this group could easily lay the foundation for studies of the evolution of floral form (e.g., Armbruster et al., 1994
, 2002
; Goldblatt et al., 1995
; Barrett et al., 1996
).
Bayesian phylogenetic analyses
The application of Bayesian analyses to phylogenetic studies is still being developed (Huelsenbeck et al., 2001
, 2002
; Lewis, 2001
). Important progress has been made in the application of this method to phylogenetic and evolutionary studies (e.g., Huelsenbeck et al., 2000
; Buckley et al., 2002
; Huelsenbeck and Imennov, 2002
; Leache and Reeder, 2002
; Miller et al., 2002
), but outstanding questions remain. One parameter to estimate in carrying out a Bayesian phylogenetic analysis is the burn-in period. This is the initial phase of MCMC sampling between the random starting point and when the sampling is within the stationary distribution of tree space. A widely used approach to determine the burn-in period is to plot likelihood values vs. generations of the search and determine when the likelihood value reaches a plateau and does not increase (e.g., Huelsenbeck and Bollback, 2001
; Huelsenbeck and Imennov, 2002
; Leache and Reeder, 2002
). Results from this study suggest that monitoring tree length provides a more conservative estimate of the burn-in period than monitoring likelihood values. Specifically, in an earlier Bayesian phylogenetic analysis, tree length was one of the most variable parameters in the phylogenetic model (Miller et al., 2002
). In the analysis presented here, this was confirmed where an order of magnitude difference in the burn-in period is suggested by these different parameters (10 000 generations for likelihood values vs. 150 000 generations for tree length).
Convergence of Markov chain Monte Carlo
A difficult question to address in a Bayesian phylogenetic analysis is determining how many generations to run the analysis (Lewis, 2001
; Huelsenbeck et al., 2002
; Miller et al., 2002
). This is a different question from establishing the burn-in period and relates to obtaining estimates of all the parameters in the phylogenetic model that are sufficiently close to the true estimates. Comparisons among Bayesian posterior probabilities for individual clades from different analyses have been used to establish whether or not an adequate sample has been obtained or whether results are consistent among separate analyses. To evaluate this question, a simple approach is to plot the posterior clade probabilities of one analysis against another and look for consistency between the results (e.g., Huelsenbeck et al., 2001
; Huelsenbeck and Imennov, 2002
; Leache and Reeder, 2002
). An alternative to this approach requires plotting the correlation between posterior probabilities for pairs of analyses with increasing generations of MCMC sampling and looking for a plateau in the correlation estimates (Miller et al., 2002
). However, it is important to remember that the objective of evaluating convergence is to obtain consistent results among sets of analyses. The index of convergence proposed here, a narrow range of values among sets of five separate analyses, is perhaps the most conservative approach proposed to date. This method provides assurance that the final analysis has been run for an adequate length of time for the results to provide a reliable representation of the parameter estimates. Furthermore, applying the criteria of accepting a narrow range of values among sets of analyses would result in a plot of the posterior probabilities of clade support of one analysis against another that would be essentially a straight line (e.g., r = 0.9996 using the results presented here for two 7-million generation runs). For the data presented here, a MCMC sampling of five million generations was required to obtain consistent posterior clade probabilities (a range of 3% for all nodes). We included an additional set of longer seven-million generation analyses. This additional set of runs was informative, but if further studies demonstrate that a narrow range of posterior clade probabilities provides a robust indicator of convergence, then the additional seven-million runs would not be necessary. It also is noteworthy that we would have been content with three million generations if we had used Gelman's R statistic as applied here as an indicator of convergence, with the potential of reporting variable and possibly misleading estimates of posterior clade probabilities.
Comparison of phylogenetic methods
Phylogenetic analysis of the ITS region involving 54 samples of 36 Ipomoea species resulted in a phylogenetic hypothesis that included both well-resolved clades and clades with weak support (Figs. 2, 4). For some taxa, these results provide an excellent statement of relationships, although particular regions of the tree are likely to require additional phylogenetic data to develop a complete well-resolved hypothesis. This combination of well-resolved and poorly-resolved relationships is most apparent in the results produced by the maximum parsimony analysis, as indicated by numerous clades with support well over 70%, as well as recovering four polytomies (Fig. 4). Two of the four polytomies include closely related species that have only a few base-pair differences among the ITS sequences (e.g., Ipomoea hederacea, I. nil, I. indica and I. tricolor, I. cardiophylla, I. marginisepala) (Fig. 3). Therefore, part of the varying resolution within the cladograms stems from the different levels of sampling within Ipomoea subgenus Quamoclit.
Comparing the different levels of support for particular nodes of the Bayesian analyses and the parsimony analyses highlights differences between measures of support provided by posterior clade probabilities (Fig. 2) and bootstrap values (Fig. 4). Discussion of these two measures of node support is a burning issue in the systematics literature (e.g., Huelsenbeck et al., 2002
; Suzuki et al., 2002
; Wilcox et al., 2002
; Alfaro et al., 2003
; Cummings et al., 2003
; Douady et al., 2003
; Holder and Lewis, 2003
). A general trend is that posterior clade probabilities are usually higher than bootstrap values (e.g., Leache and Reeder, 2002
; Miller et al., 2002
; Soltis et al., 2002
). Simulation studies generally support the accuracy of posterior probabilities (Wilcox et al., 2002
; Alfaro et al., 2003
), although tendencies toward over-credibility of posterior clade probabilities have been identified (Suzuki et al., 2002
; Alfaro et al., 2003
; Cummings et al., 2003
; Douady et al., 2003
). It is important to recognize that bootstrap values and Bayesian posterior probabilities of node support measure two different processes (Alfaro et al., 2003
). Bayesian posterior probabilities determine the strength of the data in supporting particular nodes, whereas bootstrap values indicate areas where additional data is needed to resolve relationships. Therefore, one should not expect these measures of support to be equal in value.
|
FOOTNOTES |
|---|
5 E-mail: rickmiller{at}selu.edu
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Akaike H. 1974 A new look at the statistical model identification. IEEE Transactions on Automatic Control AC-19 716-723
Alfaro M. E. S. Zoller F. Lutzoni 2003 Bayes or bootstrap? A simulation study comparing the performance of Bayesian Markov chain Monte Carlo sampling and bootstrapping in assessing phylogenetic confidence. Molecular Biology and Evolution 20: 255-266
Armbruster W. S. M. E. Edwards E. M. Debevec 1994 Floral character displacement generates assemblage structure of western Australian triggerplants (Stylidium). Ecology 75: 315-329[CrossRef][Web of Science]
Armbruster W. S. C. P. H. Mulder B. G. Baldwin S. Kalisz B. Wessa H. Nute 2002 Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae s. l). American Journal of Botany 89: 37-49
Austin D. F. 1975a Typification of the New World subdivisions of Ipomoea L. (Convolvulaceae). Taxon 24: 107-110[CrossRef]
Austin D. F. 1975b Convolvulaceae. Flora of Panama. Annals of the Missouri Botanical Garden 62: 157-224[CrossRef][Web of Science]
Austin D. F. 1977 Realignment of the species placed in Exogonium (Convolvulaceae). Annals of the Missouri Botanical Garden 64: 330-339[CrossRef][Web of Science]
Austin D. F. 1979 An infrageneric classification for Ipomoea (Convolvulaceae). Taxon 28: 359-361[CrossRef][Web of Science]
Austin D. F. 1982 Convolvulaceae. In G. Harling and B. Sparre [eds.], Flora of Ecuador. Swedish Research Councils, Stockholm, Sweden
Austin D. F. 1997 Dissolution of Ipomoea series Anisomerae (Convolvulaceae). Taxon 28: 359-361[CrossRef]
Austin D. F. R. S. Bianchini 1998 Additions and corrections in American Ipomoea (Convolvulaceae). Taxon 47: 833-838[CrossRef][Web of Science]
Austin D. F. P. B. Cavalcante 1982 Convolvulaceaes da Amazonia. Museu Paraense Emilio Goeldi, Belem, Para, Brazil
Austin D. F. Z. Huáman 1996 A synopsis of Ipomoea (Convolvulaceae) in the Americas. Taxon 29: 501-502[CrossRef]
Austin D. F. K. Kitajima Y. Yoneda L. Qian 2001 A putative tropical American plant, Ipomoea nil (Convolvulaceae), in pre-Columbian Japanese art. Economic Botany 55: 515-527[Web of Science]
Barrett S. C. L. D. Harder A. C. Worley 1996 The comparative biology of pollination and mating in flowering plants. Philosophical Transactions of the Royal Society of London, Series B 351: 1271-1280[CrossRef]
Buckley T. R. P. Arensburger C. Simon G. K. Chambers 2002 Combined data, Bayesian phylogenetics, and the origin of the New Zealand cicada genera. Systematic Biology 51: 4-18[CrossRef][Web of Science][Medline]
Choisy J. D. 1845 Convolvulaceae. In A. P. DeCandolle [ed.], Prodromus systematis naturalis regni vegetabilis, vol. 9, 323–462. Paris, France
Correll D. S. M. C. Johnston 1979 Manual of the vascular plants of Texas. University of Texas at Dallas, Richardson, Texas, USA
Cummings M. P. S. A. Handley D. S. Myers D. L. Reed A. Rokas Katarina Winka 2003 Comparing bootstrap and posterior probability values in the four-taxon case. Systematic Biology 52: 477-487[CrossRef][Web of Science][Medline]
Douady C. J. F. Delsuc Y. Boucher W. F. Doolitle E. J. P. Douzery 2003 Comparison of Bayesian and maximum likelihood bootstrap measures of phylogenetic reliability. Molecular Biology and Evolution 20: 248-254
Eckenwalder J. E. 1989 A new species of Ipomoea section Quamoclit (Convolvulaceae) from the Caribbean and a new combination for a Mexican species. Brittonia 41: 75-79
Gelman A. J. Carlin H. Stern D. Rubin 1995 Bayesian data analysis. Chapman and Hall, London, UK
Goldblatt P. J. C. Manning P. Bernhardt 1995 Pollination biology of Lapeirousia subgenus Lapeirousia (Iridaceae) in southern Africa; floral divergence and adaptation for long-tongued fly pollination. Annals of the Missouri Botanical Garden 82: 517-534[CrossRef][Web of Science]
Gunn C. R. 1972 Moonflowers, Ipomoea section Calonyction, in temperate North America. Brittonia 24: 150-168
Hallier H. 1893a Versuch einer natuerlichen Gliederung der Convolvulaceen auf morphologischer und anatomischer Grundlage. Botanische Jahrbucher fur Systematik 16: 453-591
Hallier H. 1893b Convolvulaceae Africanae. Botanische Jahrbucher fur Systematik 18: 81-160
Holder M. and P. O. Lewis 2003 Phylogeny estimation: traditional and Bayesian approaches. Nature Reviews Genetics 4: 275-284[CrossRef][Web of Science][Medline]
House H. D. 1908a The North American species of the genus Ipomoea. Annals of the New York Academy of Sciences 18: 181-263[CrossRef]
House H. D. 1908b Studies in the North American Convolvulaceae IV. The genus Exogonium. Bulletin of the Torrey Botanical Club 35: 49-96[CrossRef]
Huelsenbeck J. P. J. P. Bollback 2001 Empirical and hierarchical Bayesian estimation of ancestral states. Systematic Biology 50: 351-366[Web of Science][Medline]
Huelsenbeck J. P. N. S. Imennov 2002 Geographic origin of human mitochondrial DNA: accommodating phylogenetic uncertainty and model comparison. Systematic Biology 51: 155-165[CrossRef][Medline]
Huelsenbeck J. P. B. Larget R. E. Miller F. Ronquist 2002 Potential applications and pitfalls of Bayesian inference of phylogeny. Systematic Biology 51: 673-688[CrossRef][Web of Science][Medline]
Huelsenbeck J. P. B. Rannala B. Larget 2000 A Bayesian framework for the analysis of cospeciation. Evolution 54: 353-364
Huelsenbeck J. P. F. Ronquist 2001 MrBayes: Bayesian inference of phylogenetic trees. Bioinformatics 17: 754-755
Huelsenbeck J. P. F. Ronquist R. Nielsen J. P. Bollback 2001 Bayesian inference of phylogeny and its impact on evolutionary biology. Science 294: 2310-2314
Kearney T. H. R. H. Peebles 1960 Arizona flora. University of California Press, Berkeley, California, USA
Leache A. D. T. W. Reeder 2002 Molecular systematics of the eastern fence lizard (Sceloporus undulatus): a comparison of parsimony, likelihood, and Bayesian approaches. Systematic Biology 51: 44-68[CrossRef][Web of Science][Medline]
Lewis P. O. 2001 Phylogenetic systematics turns over a new leaf. Trends in Ecology and Evolution 16: 30-37
MacBride J. F. 1959 Convolvulaceae. Flora of Peru. Field Museum of Natural History—Botany 13: 455-536
Manos P. S. R. E. Miller P. Wilkin 2001 Phylogenetic analysis of Ipomoea, Argyreia, Stictocardia, and Turbina suggests a generalized model of morphological evolution in morning glories. Systematic Botany 26: 585-602[Web of Science]
Martin F. W. 1970 Self- and interspecific incompatibility in the Convolvulaceae. Botanical Gazette 131: 139-144[CrossRef][Web of Science]
Matuda E. 1963 El genero Ipomoea en Mexico. Anales del Instituto de Biologia. Universidad Nacional Autonoma de Mexico Serie Botanica 34: 85-145
McDonald J. A. 1982 Biosystematics of the Ipomoea tricolor complex. Ph.D. dissertation, University of Texas, Austin, Texas, USA
McDonald J. A. 1987 Revision of Ipomoea section Exogonium (Choisy) Griseb (Convolvulaceae). Brenesia 28: 41-87
McDonald J. A. 1991 Origin and diversity of Mexican Convolvulaceae. Anales del Instituto de Biologia Universidad Nacional Autonoma de Mexico Serie Botanica 62: 65-82
McDonald J. A. 1993 A new species, name change, and associate for Ipomoea section Mina (Cerv.) Griseb (Convolvulaceae). Harvard Papers 4: 49-52
McDonald J. A. 1994 Convolvulaceae II. Flora de Veracruz. Instituto de Ecologia, A. C. Xalapa, Veracruz, Mexico
McDonald J. A. 1995 Revision of Ipomoea section Leptocallis (Convolvulaceae). Harvard Papers in Botany 6: 97-122
McDonald J. A. 2001 Revision of Ipomoea series Tryianthinae (Convolvulaceae). Lundellia 4: 76-93
McDonald J. A. T. J. Mabry 1992 Phylogenetic systematics of New World Ipomoea (Convolvulaceae). Plant Systematics and Evolution 180: 243-259[CrossRef][Web of Science]
McPherson G. D. 1979 Studies in the genus Ipomoea. Ph.D. dissertation, University of Michigan, Ann Arbor, Michigan, USA
McPherson G. D. 1980 Eight new species of Ipomoea and Quamoclit from Mexico. Contributions of the University of Michigan Herbarium 14: 85-97
Miller R. E. T. R. Buckley P. S. Manos 2002 An examination of the monophyly of morning glory taxa using Bayesian phylogenetic inference. Systematic Biology 51: 740-753[CrossRef][Web of Science][Medline]
Miller R. E. M. D. Rausher P. S. Manos 1999 Phylogenetic systematics of Ipomoea (Convolvulaceae) based on ITS and waxy sequences. Systematic Botany 24: 209-227[CrossRef][Web of Science]
O'donell D. A. 1959 Convolvulaceas Argentinas. Lilloa 29: 87-343
Posada D. K. A. Crandall 1998 Modeltest: testing the model of DNA substitution. Bioinformatics 14: 817-818
Radford A. E. H. E. Ahles C. R. Bell 1968 Manual of the vascular flora of the Carolinas. University of North Carolina Press, Chapel Hill, North Carolina, USA
Rausher M. D. J. D. Fry 1993 Effects of a locus affecting floral pigmentation in Ipomoea purpurea on female fitness components. Genetics 134: 1237-1247[Abstract]
Soltis D. E. P. S. Soltis M. J. Zanis 2002 Phylogeny of seed plants based on evidence from eight genes. American Journal of Botany 89: 1670-1681
Standley P. C. L. O. Williams 1970 Tubiflorae, Flora of Guatemala. Fieldiana 24: 1-85
Stefanovic S. D. F. Austin R. G. Olmstead 2003 Classification of Convolvulaceae: a phylogenetic approach. Systematic Botany 28: 791-806[Web of Science]
Suzuki Y. G. V. Glazko M. Nei 2002 Overcredibility of molecular phylogenies obtained by Bayesian phylogenetics. Proceedings of the National Academy of Sciences, USA 99: 16138-16143
Swofford D. L. 2000 PAUP*. Phylogenetic analysis using parsimony (* and other methods), version 4.0. Sinauer, Sunderland, Massachusetts, USA
Van Ooststroom S. J. 1953 Convolvulaceae. In C. G. G. J. van Steenis [ed.], Flora Malesiana, 458–489. Woltors-Nordhoff, Groningen, Netherlands
Verdcourt B. 1957 Typification of the subdivisions of Ipomoea L. (Convolvulaceae) with regard to the East African species. Taxon 6: 150-152[CrossRef]
Verdcourt B. 1963 Convolvulaceae. In C. E. Hubbard and E. Milne-Redhead [eds.], Flora of tropical East Africa, 1–161. Whitefriars Press, London, UK
Wagner W. L. D. R. Herbst S. H. Sohmer 1990 Manual of the flowering plants of Hawai'i. Bishop Museum Special Publication 83. University of Hawaii Press, Honolulu, Hawaii, USA
Wilcox T. P. D. J. Zwickl T. A. Heath D. M. Hillis 2002 Phylogenetic relationships of the dwarf boas and a comparison of Bayesian and bootstrap measures of phylogenetic support. Molecular Phylogenetics and Evolution 25: 361-371[CrossRef][Web of Science][Medline]
Wilkin P. 1995 A new species of Ipomoea (Convolvulaceae) from Mexico State, Mexico, and its evolution. Kew Bulletin 50: 93-102[CrossRef]
Wilkin P. 1999 A morphological cladistic analysis of the Ipomoeeae (Convolvulaceae). Kew Bulletin 54: 853-876[CrossRef]
Zufall R. A. M. D. Rausher 2004 Genetic changes associated with floral adaptation restrict future evolutionary potential. Nature 428: 847-850[CrossRef][Medline]
This article has been cited by other articles:
|
|
 |
|
  R. A. Levin, N. R. Myers, and L. Bohs Phylogenetic relationships among the "spiny solanums" (Solanum subgenus Leptostemonum, Solanaceae) Am. J. Botany, January 1, 2006; 93(1): 157 - 169. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. A. MATTHIAS, M. M. DIAZ, K. J. CAMPOS, M. CALDERON, M. R. WILLIG, V. PACHECO, E. GOTUZZO, R. H. GILMAN, and J. M. VINETZ DIVERSITY OF BAT-ASSOCIATED LEPTOSPIRA IN THE PERUVIAN AMAZON INFERRED BY BAYESIAN PHYLOGENETIC ANALYSIS OF 16S RIBOSOMAL DNA SEQUENCES Am J Trop Med Hyg, November 1, 2005; 73(5): 964 - 974. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
