| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Herbarium and Division of Biology, Kansas State University, Manhattan, Kansas 66506-4901 USA; 3Section of Integrative Biology, Plant Resources Center, and Institute of Cellular and Molecular Biology, University of Texas, Austin, Texas 78712 USA
Received for publication January 3, 2002. Accepted for publication April 12, 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: hybridization • incongruence • introgression • Phlox • phylogeny • Polemoniaceae • restriction site data
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Levin and Smith, 1966
; Levin, 1967
; Ferguson, Levin, and Jansen, 1999
), genetic barriers to hybridization are generally weak (Levin, 1966
), and some taxa have been considered hybrid derivatives (at both the diploid and polyploid levels; Levin and Smith, 1966
; Levin and Schaal, 1970
; Levy and Levin, 1974
, 1975
). A recent study estimated relationships of eastern Phlox using sequence data from the internal transcribed spacer (ITS) region of the nuclear ribosomal DNA (Ferguson, Krämer, and Jansen, 1999
). That phylogeny was valuable inasmuch as it revealed some interesting natural groups within the genus and pointed to several inconsistencies in the current taxonomy. Overall, however, relationships were complicated, and placement of some samples in the tree was consistent with the hypothesis that hybridization has affected those relationships (i.e., through hybrid origin of taxa and/or introgression). Thus, the ITS study underscored the need for comparison with an independent data set—to add support to the previous findings, some of which were surprising, and to gain insight into the evolutionary processes that have been active in Phlox.
The value of comparing multiple phylogenies for a given group of organisms is well known. We may generally be confident that strongly supported, congruent patterns between trees reliably indicate organismal relationships (see Hillis, 1987
, 1995
; Miyamoto and Fitch, 1995
); in many cases it is appropriate to combine data sets for a stronger phylogenetic hypothesis (Olmstead and Sweere, 1994
; Whitten, Williams, and Chase, 2000
). On the other hand, well-corroborated conflict between independent data sets is intriguing because it can point to the evolutionary processes that have resulted in the differing patterns of relationship (see Bull et al., 1993
; Huelsenbeck, Bull, and Cunningham, 1996
; Wendel and Doyle, 1998
). For example, several workers have found that, for groups thought to have reticulating histories due to hybridization, different gene trees conflict (e.g., Kellogg, Appels, and Mason-Gamer, 1996
; Rieseberg, Whitton, and Linder, 1996
; Sang, Crawford, and Stuessy, 1997
; Seelanan, Schnabel, and Wendel, 1997
). Because of evidence for hybridization in Phlox, we set out to compare independent phylogenies (a conditional combination approach; see Bull et al., 1993
; Huelsenbeck, Bull, and Cunningham, 1996
). Incongruence can be assessed by a variety of statistical tests (see Mason-Gamer and Kellogg, 1996
; Cunningham, 1997a
; Johnson and Soltis, 1998
; Larson, 1998
). A current challenge, however, is inferring the effects of particular evolutionary processes given the many possible causes of phylogenetic incongruence (see Wendel and Doyle, 1998
; Sang and Zhong, 2000
).
Organellar DNA is an obvious source of data for development of a Phlox phylogeny independent of the nuclear-based ITS phylogeny. Restriction site analysis of the chloroplast genome enables resolution of relationships at low taxonomic levels and has other advantages as well (Olmstead and Palmer, 1994
; Jansen, Wee, and Millie, 1998
). Here we present a chloroplast DNA (cpDNA) restriction site phylogeny for eastern Phlox, allowing direct comparison with the ITS phylogeny (Ferguson, Krämer, and Jansen, 1999
). We discuss noteworthy congruence between the phylogenies, and then focus on statistically supported incongruence. This study adds to our understanding of relationships of Phlox and is an important step on the road to understanding the evolutionary processes that have led to those complicated patterns.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). Representatives of all 22 eastern species and most eastern subspecies were included, as well as multiple populations of some taxa, particularly members of the P. pilosa complex, which is thought to have a history of hybridization (e.g., Levin, 1966
; Levin and Smith, 1966
; Levin and Schaal, 1970
; Levy and Levin, 1974
). In addition, six samples of western taxa were included (four species). We also sampled the monotypic genus Microsteris, which is clearly closely related to Phlox based on molecular phylogenetic studies of the family Polemoniaceae (Steele and Vilgalys, 1994
; Johnson et al., 1996
; Porter, 1996
; Bell, Patterson, and Hamilton, 1999
; Prather, Ferguson, and Jansen, 2000
) as well as morphological data (Wherry, 1943
; Grant, 1959
, 1998
). The segregation of this genus has been controversial: some workers have recognized Microsteris gracilis within Phlox as P. gracilis (Hook.) Greene (e.g., Mason, 1941
; see also Patterson and Wilken, 1993
). We therefore included M. gracilis as part of the ingroup. Two species of Leptosiphon (see Porter and Johnson, 2000
) were chosen as outgroups (Appendix 1, http://ajbsupp.botany.org/v89); Leptosiphon, previously discussed as a lineage of Linanthus (Ferguson, Krämer, and Jansen, 1999
), is most closely related to the Phlox-Microsteris lineage (Bell, Patterson, and Hamilton, 1999
; Bell and Patterson, 2000
).
Since the publication of the ITS study, Grant (2001)
has addressed problems with sectional nomenclature of Phlox that arose through difficulty in interpretation of Asa Gray's treatments (Gray, 1870
, 1878
). The correct sectional names in Phlox are: sect. Phlox, sect. Divaricatae Peter (previously called sect. Annuae; see Prather, 1994
; Ferguson, Krämer, and Jansen, 1999
), and sect. Pulvinatae Peter (previously called sect. Occidentales; see Grant, 1959
; Ferguson, Krämer, and Jansen, 1999
). This new understanding of the nomenclature (Grant, 2001
) is followed here (Appendix 1, http://ajbsupp.botany.org/v89).
Leaf material from 3 to 15 individuals was collected and pooled as a bulk sample for each population. Bulk samples were necessary to obtain adequate amounts of highly purified DNA. The only exception was P. nivalis subsp. texensis, a federally protected plant; in that case, part of a single large individual was sampled. Total genomic DNA was extracted using the method of Doyle and Doyle (1987)
and further purified by cesium chloride/ethidium bromide gradients (Sambrook, Fritsch, and Maniatis, 1989
).
cpDNA restriction site variation
Eighteen restriction endonucleases were used: AseI, AvaI, AvaII, BamHI, BanII, BclI, BglII, BstNI, BstXI, ClaI, DraI, EcoO109I EcoRI, EcoRV, HincII, HindIII, NciI, and NsiI. Digestion of DNA, gel electrophoresis, bidirectional DNA transfer to nylon membrane (ZetaBind, AMF Cuno) by Southern blotting, radioactive labeling of probe DNA by nick-translation, filter hybridization, and autoradiography were all performed as described by Palmer (1986)
except that agarose gel concentrations ranged from 1.0 to 1.5% (see Olmstead and Palmer, 1992
). Tobacco cpDNA probes (SolClone Top40 clone bank; Olmstead and Palmer, 1992
) were used in the hybridizations at 60°C. Probes 1–29b and 35–40 were used for all of the samples (the excluded clones, 30–34, cover most of the conserved inverted repeat region). Up to three adjacent probes were combined and used in the hybridizations; combinations were based on probe sizes and relative number of enzyme cuts per probe in tobacco cpDNA. To facilitate detailed mapping where necessary, hybridizations were also performed using each individual tobacco probe for a set of three DNAs (P. pilosa subsp. fulgida MN, M. gracilis, and L. nuttallii; Appendix 1, http://ajbsupp.botany.org/v89).
In most cases, inference of restriction sites from the fragment patterns was straightforward, and construction of complete restriction site maps was deemed unnecessary (see Jansen, Wee, and Millie, 1998
). In cases where interpretation was more difficult (most often due to differences between Phlox and Leptosiphon), mapping of localized areas was carried out using the data from hybridizations done with individual probes. Length variants were scored, but care was taken to compare equivalent regions for different enzymes so that the variants were scored only once. Finally, in some cases the presence of a small fragment had to be inferred. Such fragments were presumably not seen due to our use of heterologous probes: the family Polemoniaceae is not closely related to the Solanaceae despite traditional classifications (see Porter and Johnson, 1998
; Prather, Ferguson, and Jansen, 2000
). For each restriction site, samples were scored for presence, absence, or polymorphism.
Phylogenetic analyses
Restriction site data were analyzed by unweighted Wagner parsimony using PAUP* (version 4.0b4; Swofford, 1999
); polymorphisms in the data matrix were treated as such, rather than as uncertainties. A heuristic search was conducted with simple addition, tree bisection-reconnection (TBR) branch swapping, and the MULTREES option on. MAXTREES was set to 20 000. The consistency index (CI; Kluge and Farris, 1969
) and the retention index (RI; Farris, 1989
) were calculated in PAUP*. Support for branches was evaluated by bootstrapping (Felsenstein, 1985
) and decay analyses (Bremer, 1988
; Donoghue et al., 1992
). One thousand bootstrap replicates were performed using the same settings described above, and no more than 100 trees were retained per replicate. Constraint statements for the decay analyses were generated using the software program TreeRot (version 2; Sorenson, 1999
), and the analyses were then conducted in PAUP* using the same settings described above for the original search.
The resulting cpDNA phylogeny could be directly compared to the complete ITS phylogeny (Ferguson, Krämer, and Jansen, 1999
). However, for increased efficiency and ease of comparison, the complete trees were also pruned, resulting in reduced data sets containing 55 of the 79 samples. The removed samples represented additional populations of species remaining in the reduced data sets. Moreover, prior to pruning, placement of samples in both complete consensus trees was considered so that those samples (or representatives of groups of samples) that were differently placed between the complete trees were retained.
The reduced, 55-sample cpDNA data set was analyzed as described above for the complete data set, except that 100 random additions were used and no limit was set on the maximum number of trees to be retained. In the reduced ITS data set, gaps were coded as a new character state and multistate characters were treated as polymorphisms (see Ferguson, Krämer, and Jansen, 1999
, for discussion). Insertion/deletion (indel) events for which at least one sample exhibited polymorphism were coded as interleaved characters, and the corresponding nucleotide characters were excluded, because it is not possible to code a sample as polymorphic for a base and a gap (see Ferguson, Krämer, and Jansen, 1999
). The reduced ITS data set was analyzed using Fitch parsimony under the same search conditions as the reduced cpDNA data set, except that the maxtree limit was set to 1000 for each random addition sequence replicate. Bootstrap analyses of 1000 replicates were performed using simple addition, TBR branch swapping and MULTREES on, and no more than 100 trees were retained at each replicate. Decay analyses were again conducted using TreeRot (Sorenson, 1999
). For the reduced cpDNA data set, the default TreeRot settings were used, which specify 20 random additions in PAUP*. Twenty random additions were also conducted for the reduced ITS data set, but a limit of 1000 trees was specified for each replicate (MAXTREES = 20 000).
Assessment of congruence between the cpDNA and ITS trees
Congruence can be assessed in terms of topological patterns, levels of character congruence, and degree of homogeneity of data sets (see Johnson and Soltis, 1998
, for review). In this study, comparison of the trees by direct observation initially allowed assessment of areas of topological congruence and incongruence. Strict topological agreement between the 55-sample strict consensus trees was ascertained by construction of largest common pruned trees (Finden and Gordon, 1985
; Kubicka, Kubicki, and McMorris, 1995
), accomplished using the "agreement subtrees" option of PAUP*. The consensus information index (CI1) of Rohlf (1982)
provides a quantitative measure of topological congruence: a CI1 value of 1.0 indicates complete congruence among all trees summarized in the consensus tree, with full resolution. CI1 values were calculated in PAUP* (as part of the consensus indices) for each of the 55-sample strict consensus trees and for the combined strict and semistrict consensus trees.
The incongruence index of Mickevich and Farris (1981)
, IMF, is one of several indices of character congruence (see Swofford [1991]
and Johnson and Soltis [1998]
for discussion and reviews). The IMF is the proportion of between data set homoplasy (the difference between the extra number of steps required for a most parsimonious tree from the combined data set and the extra number of steps required by each data set on its most parsimonious trees) relative to the extra homoplasy required by the combined data set; data sets with no conflict will yield an IMF value of 0.0. Analyses for this test on the 55-sample data sets were conducted using the same settings as the original analyses (see above), except that uninformative characters were excluded. The minimum numbers of synapomorphies (for each data set separately and for the combined data set) were obtained in PAUP* as part of the consistency index measures (under "tree scores, parsimony").
A global test for homogeneity was accomplished by the incongruence length difference (ILD) test of Farris et al. (1995)
, implemented in PAUP* as the partition homogeneity test. The data sets are combined into a single data matrix with two partitions: the test compares the sum of shortest tree lengths based on the original partitions (the two separate 55-sample data sets) to a distribution of sums of lengths of trees generated by random repartitioning of all of the data (i.e., the combined 55-sample data set; see discussion by Mason-Gamer and Kellogg [1996]
; see also Yoder, Irwin, and Payseur [2001]
for discussion on shortcomings of the ILD test). For this test, invariant characters were excluded (see Cunningham, 1997b
), and heuristic searches were conducted with simple addition, TBR branch swapping, and the MULTREES option on. Ninety-nine random repartitions of the data were performed with the MAXTREES limit set to 1000.
While a global analysis may point to a lack of homogeneity between the data sets generally, it is also useful to consider evidence for incongruence on a more local level by examining conflict relative to one or more clades of particular interest. The Wilcoxon signed-ranks (WSR) test as adapted by Templeton (1983)
was chosen to assess incongruence between the trees with respect to well-supported lineages occurring in one tree but not the other (see Mason-Gamer and Kellogg, 1996
; Larson, 1998
). A constraint analysis is first conducted: a tree defining the node (or nodes) of interest present in the rival topology is specified and that topological constraint is enforced during a reanalysis of the data set. The number of steps required by each character on the favored tree relative to the constraint tree is then evaluated. Generally, some characters require more steps given the topological constraint, while other characters require fewer steps. If this increase and decrease do not differ significantly based on the WSR test, then the constraint tree is not considered incongruent—the data show some support for the alternative topology. Here, WSR tests were conducted based on well-supported nodes (with bootstrap support 
80) occurring in one of the 55-sample strict consensus trees but not the other. All constraint analyses were conducted using the same settings as the original analyses (see above). For each data set, a single most parsimonious tree from the original analysis was compared with a single most parsimonious tree from each constraint analysis. The WSR tests were conducted in PAUP* using the nonparametric pairwise comparison test option (under "tree scores, parsimony"), and two-tailed probabilities were recorded. Because multiple comparisons were made for each phylogeny, a sequential Bonferroni adjustment was calculated (Rice, 1989
; Baum, Small, and Wendel, 1998
; Flores-Villela et al., 2000
); this may cause the WSR tests to be overly conservative (see discussion by Lee, 2000
). There is current discussion regarding the appropriate use of WSR tests (Goldman, Anderson, and Rodrigo, 2000
); yet some studies employing this and related strategies suggest that the WSR tests are useful and, indeed, conservative (e.g., Townsend and Larson, 2002
). The WSR tests are included here (using two-tailed probabilities and the Bonferroni adjustment) for consideration in conjunction with other measures of incongruence.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Appendices 2–4 at http://ajbsupp.botany.org/v89; also available from the first author). Twenty of the differences were length variants (13 being informative), and these were concentrated in variable parts of the genome (as ascertained relative to studies of the Ranunculaceae, which employed the same tobacco probes; Johansson and Jansen, 1993
). One hundred twenty-five data points were scored as uncertainties and constituted 0.26% of all 48 269 data points. Twenty clear polymorphisms were detected (0.04% of all data points), and 18 of those occurred in two samples: P. divaricata TN2 (eight polymorphisms) and P. pilosa subsp. latisepala TX2 (10; Appendix 1, http://ajbsupp.botany.org/v89). We do not know whether this reflects heteroplasmy because bulk population samples were used here, but individual plants usually possess only one chloroplast type (see Johnson and Palmer, 1989
; Mason, Holsinger, and Jansen, 1994
; Bain and Jansen, 1996
). Comparison of the restriction site data for these particular samples with those data from other samples, particularly of the same taxa and from the same geographic regions, did not allow inference of the individual haplotypes present, suggesting that haplotype variation in these populations is more extensive than that detected in this study (see Ferguson [1998]
for direct data comparisons).
Phylogenetic analyses
Analysis of the complete cpDNA restriction site data set (79 samples) resulted in 20 000 most parsimonious trees, the maximum number saved. Trees were 854 steps long including autapomorphies and had a CI of 0.62 excluding uninformative characters and an RI of 0.86. A comparison with relevant statistics for the complete ITS data set (Ferguson, Krämer, and Jansen, 1999
) and the 55-sample cpDNA and ITS data sets is provided in Table 1. The complete cpDNA tree (Figs. 1 and 2) was entirely congruent with the 55-sample tree (Fig. 3), and three additional nodes were resolved in the smaller tree (although, without strong support). The 55-sample ITS phylogeny (Fig. 4) was largely congruent with the complete tree (Ferguson, Krämer, and Jansen, 1999
); the branching pattern differed in one area that was previously found to be unstable, and support for the differing patterns was weak. One node was resolved with weak support in the 55-sample ITS tree that was not resolved in the complete tree (Fig. 4; Ferguson, Krämer, and Jansen, 1999
).
|
|
|
|
|
) revealed areas of agreement as well as conflict, and these patterns were observed for the 55-sample trees as well (Figs. 3 and 4). For example, the annual Phlox taxa form a monophyletic group in each tree (clade "A" in Figs. 3 and 4), yet in the cpDNA tree they are placed distantly from some samples with which they share a relatively close relationship in the ITS tree (e.g., P. floridana FL1, P. floridana FL2, P. pilosa subsp. pilosa AL, and P. pilosa subsp. sangamonensis, among others; Figs. 3 and 4). Samples of two different populations of P. pilosa subsp. pilosa collected from adjacent counties in Texas (P. pilosa subsp. pilosa TX1 and P. pilosa subsp. pilosa TX2; Appendix 1, http://ajbsupp.botany.org/v89) group in the same area of the ITS tree but are relatively distantly related based on the cpDNA tree (Figs. 3 and 4). The largest common pruned trees required removal of 38 samples from the 55-sample trees (Fig. 5), demonstrating that the cpDNA and ITS trees differ greatly overall in their topologies. Finally, the CI1 values for the 55-sample cpDNA and ITS strict consensus trees were 0.63 and 0.54, respectively, while those for the combined consensus trees were 0.08 (strict consensus) and 0.09 (semistrict). The values for the cpDNA and ITS strict consensus trees are below 1.0 due to lack of resolution at some nodes (a CI1 of 1.0 would indicate complete topological congruence among all of the trees in the set, with full resolution); the very low CI1 values for the combined consensus tree sets (0.08 and 0.09) are an indicator of the poor topological agreement between the cpDNA and ITS trees.
|
; Seelanan, Schnabel, and Wendel, 1997
; Johnson and Soltis, 1998
). Since the IMF measure does not address the nature of the incongruence (for example, is the conflict due to strongly supported differing placement of one or a few samples, or, alternatively, are there weakly supported differences with respect to many samples?), it must be interpreted in concert with other measures of incongruence (see Johnson and Soltis [1998]
for discussion).
|
80 occur in the 55-sample cpDNA strict consensus tree but do not occur in the corresponding ITS tree (a–j; Fig. 3), while seven such nodes occur in the ITS tree but not the cpDNA tree (k–q; Fig. 4). Significance could not be evaluated for two of these tests because fewer than six characters varied (nodes e and i, Fig. 3; Table 3; Siegel, 1956
; Sokal and Rohlf, 1995
; Johnson and Soltis, 1998
). For the remaining 15 tests, five of the nodes were statistically incongruent with the rival tree with P < 0.01 (nodes a, b, c, and d, Fig. 3; and node n, Fig. 4; Table 3), while one was statistically incongruent with P < 0.001 (node l, Fig. 4; two-tailed P values were used, followed by a sequential Bonferroni adjustment). These strong incongruencies highlight differences between the cpDNA and ITS phylogenies with respect to placement of members of the P. pilosa and P. glaberrima complexes in particular (Figs. 3 and 4; Appendix 1, http://ajbsupp.botany.org/v89).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
): the cpDNA phylogeny exhibits 45 resolved nodes in the strict consensus tree relative to 37 for the ITS phylogeny and has 24 branches with bootstrap values 80 relative to 17 for the ITS phylogeny (Table 1). These results may be attributable to the increased number of informative characters in the cpDNA data set (346 vs. 134 for the ITS data set; Table 1). The CI and RI values are lower for the cpDNA phylogeny (0.62 and 0.86, respectively, vs. 0.80 and 0.91 for the ITS phylogeny; Table 1), but more characters would also be expected to lead to increased levels of homoplasy.
A similar pattern of improved resolution and support in the cpDNA phylogeny relative to the ITS phylogeny is found in the trees resulting from analyses of the 55-sample data sets (Table 1; Figs. 3 and 4). It is noteworthy that the CI and RI values for these analyses are lower than those for the complete data sets in each case (Table 1); we would generally expect a data set with fewer samples to result in increased CI and RI values (Sanderson and Donoghue, 1989
, 1996
). These findings may be due to the pruning strategy. Because we were particularly interested in relationships at lower taxonomic levels (e.g., between taxa and populations of P. pilosa) as well as in samples that differed in topological position between the two complete phylogenies, the samples removed may have provided much structure to the complete trees. For example, all but a single sample each of P. drummondii and P. cuspidata were pruned in constructing the 55-sample data sets, and those additional samples had added phylogenetic signal in a well-supported and well-structured part of each complete tree.
Implications of congruence between the cpDNA and ITS phylogenies
The cpDNA phylogeny is in agreement with the ITS phylogeny with respect to several points of taxonomic importance: (1) Microsteris gracilis comprises a lineage separate from all samples of Phlox; (2) Wherry's (1955)
sections Phlox and Divaricatae do not represent natural, monophyletic assemblages; (3) the eastern mat-forming taxa form a single lineage; and (4) P. roemeriana is closely related to the other annuals of central Texas, P. cuspidata and P. drummondii. All of these comparisons can be visualized in the complete trees (Fig. 2; fig. 2 of Ferguson, Krämer, and Jansen, 1999
) as well as in the 55-sample trees (Figs. 3 and 4).
The placement of Microsteris as a lineage separate from all samples of Phlox in both the cpDNA and ITS phylogenies is consistent with its segregation as a genus distinct from Phlox. It is likely that this relationship will hold with increased sampling of western Phlox taxa and additional sampling of M. gracilis. Notable differences exist between the genera with respect to flower and seed morphology (see Wherry, 1943
, 1955
; Grant, 1959
); moreover, the annual habit is independently derived in Microsteris, which is clearly not closely related to P. drummondii as hypothesized by Mason (1941)
when he argued for lumping the two genera. Segregation of the monotypic Microsteris thus becomes a judgment call on the taxonomic significance of the characters in question. Recent treatments of the Polemoniaceae recognize Microsteris (Grant, 1998
; Porter and Johnson, 2000
), and we advocate this segregation.
Lack of monophyly of the sections containing eastern Phlox taxa (sects. Phlox and Divaricatae; Appendix 1, http://ajbsupp.botany.org/v89) is apparent in both the cpDNA and ITS phylogenies. The eastern mat-forming taxa provide an excellent example of this. Samples of the four species (clade "B" in Figs. 2–4) are grouped together in both phylogenies, although Wherry (1955)
grouped P. bifida and P. subulata into sect. Phlox because of their long styles and P. oklahomensis and P. nivalis into sect. Divaricatae because of their short styles (and associated floral characters; see Wherry, 1955
). The phylogenetic data thus highlight a traditional overemphasis on particular morphological characters (i.e., style length) in the classification of the genus Phlox. Information from phylogenetic studies will be valuable to the development of a more natural classification, a goal of ongoing work including western members of the genus.
The grouping of the eastern mat-forming species (clade "B" in Figs. 2–4) is less surprising in light of the habit they share: all have narrow, evergreen leaves and tend to spread, forming mats of large colonies. These features are common among many of the western Phlox species, and this makes the placement of the eastern mat-forming taxa particularly intriguing. Clade "B" is nested within a group of western species with moderate to strong support in the cpDNA phylogeny (Figs. 2 and 3), a relationship not suggested by the ITS tree, which has poor resolution and support at some of the interior nodes (Fig. 4; fig. 2 of Ferguson, Krämer, and Jansen, 1999
). The cpDNA phylogeny thus not only supports the monophyly of the eastern mat-forming species (as first detected in the ITS phylogeny), but also suggests they may be closely related to some of the western Phlox taxa.
Finally, the cpDNA tree strongly corroborates the conclusion, based initially on the ITS tree, that the central Texas annuals P. cuspidata, P. drummondii, and P. roemeriana are a monophyletic group. As discussed previously (Ferguson, Krämer, and Jansen, 1999
), P. roemeriana was placed by Wherry (1955)
with western taxa in the P. nana complex, its unusual morphological features obscuring its evolutionary relationships. In the cpDNA tree, the annuals form a strongly supported clade within which resolution is identical to that in the ITS phylogeny (clade "A" in Figs. 2–4; fig. 2 of Ferguson, Krämer, and Jansen, 1999
). We consider the phylogenetic data taken together to be convincing evidence of the close relationship between P. cuspidata, P. drummondii, and P. roemeriana.
Incongruence between the cpDNA and ITS phylogenies and the possible role of hybridization
Classical evolutionary studies in Phlox have suggested that past hybridization has played an important role in the evolution of some eastern species. In particular, several taxa in the P. pilosa complex were proposed to be hybrid derivatives based on multiple lines of evidence, as outlined in Table 4. In undertaking phylogenetic studies of the genus Phlox, the first author initially anticipated straightforward evidence for or against these hypotheses of hybridization (particularly with regard to the tetraploid taxa; Table 4). For example, some ITS studies have found support for hybrid origin in multiple ITS types of putative hybrid derivatives (e.g., Sang, Crawford, and Stuessy, 1995
; Campbell et al., 1997
). In Phlox, there are no such clear patterns, though multiple ITS sequences were detected (presumably due to incomplete concerted evolution; cloning yielded multiple sequences that still grouped together in the phylogeny; Ferguson, Krämer, and Jansen, 1999
). The lack of monophyly of the P. pilosa complex detected in the ITS study was intriguing given the clear circumscription, based on morphology, of many members of that group. Because comparison of multiple phylogenies can provide strong support for hypotheses of hybrid origin of taxa (e.g., Wendel, Schnabel, and Seelanan, 1995
; Sang, Crawford, and Stuessy, 1997
; Sang and Zhang, 1999
), the present study was undertaken. A hybrid derivative might group with one of its "parental" taxa in one tree and the other taxon in a tree based on a different data set, for example. Of course, a history of hybridization and introgression could also yield more complicated patterns, and complexity of relationships seems to be a key feature of Phlox.
|
for review of congruence tests, and references under MATERIALS AND METHODS for additional criticisms of the ILD and WSR tests). The WSR tests identify several particular clades (labeled a, b, c, d, l, n; Figs. 3 and 4) that are incongruent with the alternative topology. All of this evidence highlights strong incongruence involving many samples in our study. Members of the well-sampled P. pilosa complex (and possibly some other eastern taxa, including members of the P. glaberrima complex; Appendix 1, http://ajbsupp.botany.org/v89; Figs. 2–4) are players in the observed incongruence, and the patterns of incongruence are far from simple. Wendel and Doyle recently reviewed the possible causes of phylogenetic incongruence, including the organism-level processes of rapid diversification (short internal branches), hybridization, lineage sorting, and horizontal gene transfer (Wendel and Doyle, 1998
; see also Avise, 1989
; Rieseberg and Soltis, 1991
). The effects of hybridization (both in terms of the origin of new taxa and, perhaps more commonly, in terms of gene flow) are considered to be the major cause of statistically supported incongruence between cpDNA- and nuclear-based plant phylogenies (Rieseberg, Whitton, and Linder, 1996
); yet it is difficult to rule out other possibilities without more in-depth studies of processes active in particular groups (see Wendel and Doyle [1998]
for review).
In the case of Phlox, there is additional evidence that hybridization occurs and may affect patterns of relationship. Among the eastern taxa (the best-studied members of the genus), many species are cross-compatible (including species pairs within and between the P. pilosa and P. glaberrima complexes; Levin, 1966
), and some species are known to hybridize in nature (e.g., Wherry, 1955
; Levin, 1978
; Ferguson, Levin, and Jansen, 1999
). Our knowledge of the natural history of eastern Phlox suggests that natural hybridization is not common, though. In fact, it is very rare to come across an herbarium specimen that appears to be a hybrid plant, and active field work searching for hybrid zones yields few such areas (C. J. Ferguson, unpublished data). Most sympatric eastern Phlox species are somewhat isolated by ecology and potentially by differences in effective pollinators (S. Wiggam-Harper and C. J. Ferguson, unpublished data). One study suggested hybridization occurred in central Tennessee between P. amoena and P. bifida (based strictly on morphological data; Anderson and Gage, 1952
), and that is not supported by our data (see the placement of the eastern mat-forming taxa, which include P. bifida subsp. stellaria from Tennessee; Figs. 3 and 4). Nonetheless, the occasional hybridization that does occur currently (and presumably has occurred in the past) suggests that there has been ample opportunity for gene exchange among some of these eastern Phlox species. It is noteworthy that hybridization can potentially have a variety of effects on phylogenetic relationships: (1) it may result in the hybrid origin of a new taxon at the polyploid level (Soltis and Soltis, 1995
, 1999
; Ramsey and Schemske, 1998
; Wendel, 2000
), or, perhaps rarely, at the diploid level (Rieseberg, 1997
; Buerkle et al., 2000
), (2) it may lead to gene flow, or introgression (see Rieseberg and Soltis, 1991
; Rieseberg, Whitton, and Linder, 1996
), and this may be historical and/or ongoing, or (3) a group may contain one or more hybrid derivative taxa as well as taxa that have experienced gene flow from congeners. Clearly, a history of hybridization can yield complex patterns of relationship, and processes may be difficult to infer from the resulting patterns.
Particular long-standing hypotheses of hybrid origin of Phlox taxa by Levin and co-workers (Table 4) are not directly supported by the phylogenetic data. Although placement of the putative hybrid derivatives does differ between the different phylogenies (Figs. 3 and 4), relationships of eastern Phlox are more complicated than previously thought. We do not suggest that other lines of evidence employed in Phlox (karyology, flavonoids, seed protein electrophoresis, and experimental crosses; see Table 4 for references) are in disagreement with our findings; on the contrary, all available data support the hypothesis that hybridization has affected relationships within Phlox. Yet, previous workers proposed the simplest reasonable relationships to explain their data, and our present phylogenetic information, based on extensive sampling, reveals a more complicated pattern.
Incongruence due to gene flow is certainly possible for eastern Phlox. The introgression of DNA markers, particularly from the chloroplast genome, has been well-documented and is considered fairly common (see Rieseberg, Whitton, and Linder [1996]
for review). There has also been a sense that a phylogeny developed from nuclear-based markers, such as the ITS tree here, is more likely to be "correct" (i.e., reflect species relationships), while the cpDNA tree may be more likely to reflect introgression (e.g., Soltis and Kuzoff, 1995
; Hardig, Soltis, and Soltis, 2000
). In fact, many of the circumstances leading to gene flow would generally favor cpDNA introgression, although the possibility of introgression of nuclear genes is also present (reviewed by Rieseberg, Whitton, and Linder, 1996
). In the Phlox phylogenies, the grouping of some samples may be due to localized cpDNA introgression (e.g., P. carolina subsp. turritella and P. pilosa subsp. pilosa AL came from the same geographical area and are grouped in the cpDNA tree; Figs. 3 and 4; Appendix 1, http://ajbsupp.botany.org/v89). Yet, relationships are clearly problematic in both the cpDNA and ITS phylogenies, as demonstrated by placement of samples of the P. pilosa and P. glaberrima complexes (Figs. 2–4). Pairs of samples of the clearly circumscribed P. pilosa are placed differently between the two phylogenies; for example, P. pilosa subsp. pilosa TX1 and P. pilosa subsp. pilosa TX2 are found in the same clade in the ITS tree but very different parts of the cpDNA tree, while the reverse is true for P. pilosa subsp. sangamonensis and P. pilosa subsp. pilosa IN (Figs. 3 and 4). On the whole, there are no generalized patterns of geographical groupings for the samples involved in the incongruence in one phylogeny vs. the other. And, even occasional introgression over time could theoretically yield such complicated patterns.
The challenges our data present to understanding the evolutionary history of Phlox highlight the importance of careful sampling in phylogenetic studies, particularly for groups including widespread taxa thought to undergo some hybridization. The complexity of relationships described here may turn out to be more commonplace as workers initiate more phylogenetic studies with multiple data sets that include many samples at the population level, as recommended by Rieseberg (1997)
.
Ongoing questions in Phlox
The data presented for eastern Phlox demonstrate the value of a phylogenetic approach to the study of this important system. Cases of congruence between the cpDNA and ITS phylogenies support some strong taxonomic conclusions: the genus Microsteris is a lineage separate from Phlox, the currently recognized sections within Phlox are not monophyletic, and the eastern mat-forming species and the annuals each form monophyletic groups. Even more interesting are cases of incongruence between the two trees, supported by a variety of statistical tests: there is strong conflict with respect to placement of many samples from the P. pilosa complex, as well as some of the other eastern taxa. This incongruence is consistent with a hypothesized history of hybridization (gene flow and/or possibly hybrid origin of taxa) among eastern Phlox, yet at present the complicated patterns leave us with more questions than answers. These findings form the basis for ongoing studies at different taxonomic levels. Phylogenetic investigations sampling the entire genus, two-thirds of which are distributed in western North America, will lead to an improved understanding of broad patterns of diversification of Phlox. Work at the population level, considering patterns of genetic variation and pollinator movement, will enable further consideration of the role hybridization may have played in causing the observed patterns of incongruence. Overall, as shown here, phylogenetic studies involving multiple data sets and including strong intraspecific sampling can contribute greatly to our knowledge of patterns of diversification of species groups and the processes leading to the diversity we see.
|
FOOTNOTES |
|---|
4 Author for correspondence (ferg{at}ksu.edu
)
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Avise J. C. 1989 Gene trees and organismal histories: a phylogenetic approach to population biology. Evolution 43: 1192-1208[CrossRef][ISI]
Bain J. F. R. K. Jansen 1996 Numerous chloroplast DNA polymorphisms are shared among different populations and species in the aureoid Senecio (Packera) complex. Canadian Journal of Botany 74: 1719-1728
Baum D. A. R. L. Small J. F. Wendel 1998 Biogeography and floral evolution of Baobabs (Adansonia, Bombacaceae) as inferred from multiple data sets. Systematic Biology 47: 181-207[CrossRef][ISI][Medline]
Bell C. D. R. W. Patterson 2000 Molecular phylogeny and biogeography of Linanthus (Polemoniaceae). American Journal of Botany 87: 1857-1870
Bell C. D. R. W. Patterson L. A. Hamilton 1999 Sectional integrity in Linanthus (Polemoniaceae): a molecular phylogeny of section Dianthoides. Systematic Botany 24: 632-644[CrossRef][ISI]
Bremer K. 1988 The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 795-803[CrossRef][ISI]
Buerkle C. A. R. J. Morris M. A. Asmussen L. H. Rieseberg 2000 The likelihood of homoploid hybrid speciation. Heredity 84: 441-451
Bull J. J. J. P. Huelsenbeck C. W. Cunningham D. L. Swofford P. J. Waddell 1993 Partitioning and combining data in phylogenetic analysis. Systematic Biology 42: 384-397[CrossRef][ISI]
Campbell C. S. M. F. Wojciechowski B. G. Baldwin L. A. Alice M. J. Donoghue 1997 Persistent nuclear ribosomal DNA sequence polymorphism in the Amelanchier agamic complex (Rosaceae). Molecular Biology and Evolution 14: 81-90[Abstract]
Cunningham C. W. 1997a Can three incongruence tests predict when data should be combined?. Molecular Biology and Evolution 14: 733-740[Abstract]
Cunningham C. W. 1997b Is congruence between data partitions a reliable predictor of phylogenetic accuracy? Empirically testing an iterative procedure for choosing among phylogenetic methods. Systematic Biology 46: 464-478[CrossRef][ISI][Medline]
Donoghue M. J. R. G. Olmstead J. F. Smith J. D. Palmer 1992 Phylogenetic relationships of Dipsacales based on rbcL sequences. Annals of the Missouri Botanical Garden 79: 249-265
Doyle J. J. J. L. Doyle 1987 A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11-15
Farris J. S. 1989 The retention index and homoplasy excess. Systematic Zoology 38: 406-407[CrossRef]
Farris J. S. M. Källersjö A. G. Kluge C. Bult 1995 Testing significance of incongruence. Cladistics 10: 315-319[CrossRef][ISI]
Felsenstein J. 1985 Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783-791[CrossRef][ISI]
Ferguson C. J. 1998 Molecular systematics of eastern Phlox L. (Polemoniaceae). Ph.D. dissertation, University of Texas, Austin, Texas, USA
Ferguson C. J. F. Krämer R. K. Jansen 1999 Relationships of eastern North American Phlox (Polemoniaceae) based on ITS sequence data. Systematic Botany 24: 616-631[CrossRef][ISI]
Ferguson C. J. D. A. Levin R. K. Jansen 1999 Natural hybridization between an outcrossing and a selfing Phlox (Polemoniaceae): the maternal species of F1 hybrids. Plant Systematics and Evolution 218: 153-158[CrossRef][ISI]
Finden C. R. A. D. Gordon 1985 Obtaining common pruned trees. Journal of Classification 2: 255-276[CrossRef][ISI]
Flores-Villela O. K. M. Kjer M. Benabib J. W. Sites Jr 2000 Multiple data sets, congruence, and hypothesis testing for the phylogeny of basal groups of the lizard genus Sceloporus (Squamata, Phrynosomatidae). Systematic Biology 49: 713-739[CrossRef][ISI][Medline]
Goldman N. J. P. Anderson A. G. Rodrigo 2000 Likelihood-based tests of topologies in phylogenetics. Systematic Biology 49: 652-670[CrossRef][ISI][Medline]
Grant V. 1959 Natural history of the Phlox family: systematic botany. Martinus Nijhoff, The Hague, The Netherlands
Grant V. 1998 Primary classification and phylogeny of the Polemoniaceae, with comment on molecular cladistics. American Journal of Botany 85: 741-752[Abstract]
Grant V. 2001 Nomenclature of the main subdivisions of Phlox (Polemoniaceae). Lundellia 4: 25-29
Gray A. 1870 Revision of the North American Polemoniaceae. Proceedings of the American Academy of Arts and Sciences, Philadelphia 8: 247-282
Gray A. 1878 Synoptical flora of North America. Ivison, Blakeman, Taylor, New York, New York, USA
Hardig T. M. P. S. Soltis D. E. Soltis 2000 Diversification of the North American shrub genus Ceanothus (Rhamnaceae): conflicting phylogenies from nuclear ribosomal DNA and chloroplast DNA. American Journal of Botany 87: 108-123
Hillis D. M. 1987 Molecular versus morphological approaches to systematics. Annual Review of Ecology and Systematics 18: 23-42
Hillis D. M. 1995 Approaches for assessing phylogenetic accuracy. Systematic Biology 44: 3-16[CrossRef]
Huelsenbeck J. P. J. J. Bull C. W. Cunningham 1996 Combining data in phylogenetic analysis. Trends in Ecology and Evolution 11: 152-157[CrossRef]
Jansen R. K. J. L. Wee D. Millie 1998 Comparative utility of chloroplast DNA restriction site and DNA sequence data for phylogenetic studies in plants. In D. E. Soltis, P. S. Soltis, and J. J. Doyle [eds.], Molecular systematics of plants II: DNA sequencing, 87–100. Kluwer Academic, Boston, Massachusetts, USA
Johansson J. T. R. K. Jansen 1993 Chloroplast DNA variation and phylogeny of the Ranunculaceae. Plant Systematics and Evolution 187: 29-49[CrossRef][ISI]
Johnson L. A. J. L. Schultz D. E. Soltis P. S. Soltis 1996 Monophyly and generic relationships of Polemoniaceae based on matK sequences. American Journal of Botany 83: 1207-1224[CrossRef][ISI]
Johnson L. A. D. E. Soltis 1998 Assessing congruence: empirical examples from molecular data. In D. E. Soltis, P. S. Soltis, and J. J. Doyle [eds.], Molecular systematics of plants II: DNA sequencing, 297–348. Kluwer Academic, Boston, Massachusetts, USA
Johnson L. B. J. D. Palmer 1989 Heteroplasmy of chloroplast DNA in Medicago. Plant Molecular Biology 12: 3-11
Kellogg E. A. R. Appels R. J. Mason-Gamer 1996 When genes tell different stories: the diploid genera of Triticeae (Gramineae). Systematic Botany 21: 321-347[CrossRef][ISI]
Kim K.-J. R. K. Jansen 1994 Comparisons of phylogenetic hypotheses among different data sets in dwarf dandelions (Krigia, Asteraceae): additional information from internal transcribed spacer sequences of nuclear ribosomal DNA. Plant Systematics and Evolution 190: 157-185[CrossRef][ISI]
Kluge A. G. J. S. Farris 1969 Quantitative phyletics and the evolution of anurans. Systematic Zoology 18: 1-32
Kubicka E. G. Kubicki F. R. McMorris 1995 An algorithm to find agreement subtrees. Journal of Classification 12: 91-99
Larson A. 1998 The comparison of morphological and molecular data in phylogenetic systematics. In R. DeSalle and B. Schierwater [eds.], Molecular approaches to ecology and evolution, 275–296. Birkhäuser Verlag, Basel, Switzerland
); the positions of particular sets of samples of P. pilosa discussed in the text are also indicated by symbols (P. pilosa subsp. pilosa TX1 and P. pilosa subsp. pilosa TX2, 
; P. pilosa subsp. sangamonensis and P. pilosa subsp. pilosa IN, •). The three nodes resolved here that were not resolved in the complete cpDNA tree (
