| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© 2008 Society of Systematic Biologists
Resolving an Ancient, Rapid Radiation in Saxifragales
Edited by Vincent Savolainen
1 South China Botanical Garden, the Chinese Academy of Sciences Guangzhou 510650, China
2 Department of Botany, University of Florida Gainesville, FL 32611, USA; E-mail: dsoltis{at}botany.ufl.edu
3 Florida Museum of Natural History, University of Florida Gainesville, Florida 32611, USA
4 Department of Ecology and Evolutionary Biology, University of Michigan Ann Arbor, Michigan 48109, USA
5 Department of Horticulture and Landscape Architecture, Washington State University Pullman, Washington 99164, USA
6 Department of Biological Sciences, University of New Orleans New Orleans, Louisiana 70148, USA
7 Current Address: Biology Department, Oberlin College Oberlin, Ohio 44074-1097, USA
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
Despite the prior use of
9000 bp, deep-level relationships within the angiosperm clade, Saxifragales remain enigmatic, due to an ancient, rapid radiation (89.5 to 110 Ma based on the fossil record). To resolve these deep relationships, we constructed several new data sets: (1) 16 genes representing the three genomic compartments within plant cells (2 nuclear, 10 plastid, 4 mitochondrial; aligned, analyzed length = 21,460 bp) for 28 taxa; (2) the entire plastid inverted repeat (IR; 26,625 bp) for 17 taxa; (3) "total evidence" (50,845 bp) for both 17 and 28 taxa (the latter missing the IR). Bayesian and ML methods yielded identical topologies across partitions with most clades receiving high posterior probability (pp = 1.0) and bootstrap (95% to 100%) values, suggesting that with sufficient data, rapid radiations can be resolved. In contrast, parsimony analyses of different partitions yielded conflicting topologies, particularly with respect to the placement of Paeoniaceae, a clade characterized by a long branch. In agreement with published simulations, the addition of characters increased bootstrap support for the putatively erroneous placement of Paeoniaceae. Although having far fewer parsimony-informative sites, slowly evolving plastid genes provided higher resolution and support for deep-level relationships than rapidly evolving plastid genes, yielding a topology close to the Bayesian and ML total evidence tree. The plastid IR region may be an ideal source of slowly evolving genes for resolution of deep-level angiosperm divergences that date to 90 My or more. Rapidly evolving genes provided support for tip relationships not recovered with slowly evolving genes, indicating some complementarity. Age estimates using penalized likelihood with and without age constraints for the 28-taxon, total evidence data set are comparable to fossil dates, whereas estimates based on the 17-taxon data are much older than implied by the fossil record. Hence, sufficient taxon density, and not simply numerous base pairs, is important in reliably estimating ages. Age estimates indicate that the early diversification of Saxifragales occurred rapidly, over a time span as short as 6 million years. Between 25,000 and 50,000 bp were needed to resolve this radiation with high support values. Extrapolating from Saxifragales, a similar number of base pairs may be needed to resolve the many other deep-level radiations of comparable age in angiosperms.
Keywords: Large data sets; long branch attraction; mtDNA; nuclear ribosomal DNA; plastid inverted repeat; rapid radiation; Saxifragales; slowly evolving genes
Received June 9, 2007; Revised August 14, 2007; Accepted September 26, 2007
Deep-level organismal relationships, such as those resulting from ancient, rapid radiations, are notoriously difficult to resolve because of the confounding problems of few synapomorphies, homoplasy, and extinction. Major radiations in plants include red algae (Yoon et al., 2006), mosses (Shaw et al., 2003; Shaw and Renzaglia, 2004), basal land plants (Qiu et al., 2006b), seed plants (Burleigh and Mathews, 2004), and the early angiosperms (Crane, 1995; P. Soltis and Soltis, 2004; D. Soltis et al., 2005). Questions of general interest relating to the resolution of ancient radiations include (1) Can ancient radiations be resolved with confidence? Some apparent radiations may be resolvable given sufficient taxon and character sampling, whereas others may accurately depict evolutionary history. Furthermore, resolving deep-level radiations is complex and involves more than simply adding numerous base pairs (Rokas et al., 2005). Lack of resolution may be the result of a number of factors, including taxon density and model of sequence evolution (Baurain et al., 2007). (2) If adding numerous base pairs with adequate taxon density ultimately provides resolution, which genes are the best choices for phylogenetic sequencing (Goldman, 1998; Townsend, 2007)? Despite the fact that slowly evolving genes (i.e., those with uniformly slow rates of substitution across the gene) may have relatively few phylogenetically informative characters, some have argued that sequencing many such genes is the best approach for inferring deep-level phylogeny because the frequency of sites with multiple substitutions that may obscure relationships is lower, resulting in a higher proportion of phylogenetically reliable characters (Felsenstein, 1983; Moritz et al., 1987; Yang, 1998; Graham and Olmstead, 2000; Wortley et al., 2005). However, others (e.g., Hillis, 1998; Hilu et al., 2004; Borsch et al., 2003) have advocated the use of rapidly evolving regions, arguing that many characters of such regions are phylogenetically informative, given sufficient taxon sampling, so that homoplasy will be dispersed appropriately relative to signal (see Källersjö et al., 1998, 1999). (3) How does taxon density affect divergence time estimation and the use of this information for inferring the duration of a radiation? Although increasing the number of base pairs and, as a result, resolution and support, likely increases the accuracy of molecular age estimates, the combined effect of increased sampling of characters and taxa on divergence time estimation has rarely been addressed (see Linder et al., 2005).
The flowering plant clade Saxifragales represents part of the early diversification of the large eudicot clade, the latter comprising approximately 75% of all angiosperms (Drinnan et al., 1994). Although small compared with other eudicot orders, Saxifragales (only about 2470 species) are a morphologically highly diverse group, including annual and perennial herbs, succulents, aquatics, shrubs, vines, and large trees (Cronquist, 1981; Takhtajan, 1997; D. Soltis et al., 2005). The minimum age of Saxifragales based on the fossil record is 89.5 Ma (Magallón et al., 1999; > 90 Ma: Hermsen et al., 2006), but initial molecular estimates based on phylogenetic analyses of all major groups of angiosperms suggest older ages of 111 to 121 Ma (Wikström et al., 2001) or 102 Ma (crown group) and 108 Ma (stem group; Anderson et al., 2005). Saxifragales are just one of several examples of angiosperm clades whose deep-level relationships have been difficult to resolve due apparently to a rapid, ancient radiation (Fishbein et al., 2001). Other examples include Malpighiales (Davis et al., 2001), Caryophyllales (D. Soltis et al., 2005; Cuénoud et al., 2002), Ericales (e.g., Anderberg et al., 2002), and Lamiales (Olmstead et al., 2001; Bremer et al., 2002).
The composition of Saxifragales is one of the major surprises of molecular phylogenetic studies of angiosperms (reviewed in D. Soltis et al., 2005). Saxifragales (sensu APG II, 2003) include Altingiaceae, Cercidiphyllaceae, Crassulaceae, Daphniphyllaceae, Grossulariaceae, Haloragaceae sensu lato (expanded to include Tetracarpaeaceae, Penthoraceae, and Aphanopetalum), Hamamelidaceae, Iteaceae, Paeoniaceae, Pterostemonaceae (the latter sometimes included in Iteaceae), Saxifragaceae, and Peridiscaceae (Davis and Chase, 2004; D. Soltis et al., 2007; see summary in Fig. 1). The molecular circumscription of Saxifragales departs markedly from previous morphology-based classifications, which placed these same families in three or four different subclasses, Hamamelidae, Rosidae, Dilleniidae, or Magnoliidae (e.g., Cronquist, 1981; Takhtajan, 1997; Morgan and Soltis, 1993).
|
A recent analysis of 18S and 26S rDNA and mitochondrial matR sequence data placed the parasite Cynomorium coccineum (Cynomoriaceae) in Saxifragales (Nickrent et al., 2005). However, bootstrap support for this relationship was less than 50%, and expanded analyses (Jian et al., unpublished data) of 561 angiosperms using five genes place Cynomoriaceae in Santalales, in agreement with traditional classifications (e.g., Cronquist, 1981). We therefore will not consider Cynomoriaceae further.
Whereas the monophyly of Saxifragales is well supported (e.g., D. Soltis et al., 1997a, b, 1998, 2000; D. Soltis and Soltis, 1997; Hoot et al., 1999), deep-level relationships within the clade remain poorly resolved and supported. Based on sequences from five genes (9237 bp), bootstrap (BS) and Bayesian posterior probability (pp) values are very high for two crown clades within Saxifragales (BS = 99–100; pp = 1.0): (1) Crassulaceae + Haloragaceae sensu lato (s.l.); and (2) the "Saxifragaceae alliance" of Iteaceae + Pterostemonaceae as sister to Grossulariaceae + Saxifragaceae (Fishbein et al., 2001; D. Soltis et al., 2007). Values are lower (pp = 0.92 and BS = 68%) for a "core Saxifragales" clade of Crassulaceae + Haloragaceae s.l. as sister to the Saxifragaceae alliance. However, the relationships of the remaining families to other Saxifragales are unclear. In Bayesian and maximum likelihood analyses, Peridiscaceae appear as sister to the rest of Saxifragales followed by a clade or grade of Altingiaceae, Cercidiphyllaceae, Daphniphyllaceae, and Hamamelidaceae as sister(s) to the core members noted above (D. Soltis et al., 2007). However, relationships among these five woody families remain uncertain, a result attributed to an ancient, rapid radiation. The placement of Paeoniaceae (comprising the single genus Paeonia) within Saxifragales has varied in previous parsimony-based analyses; the family is characterized by a long branch (Fishbein et al., 2001; D. Soltis et al., 2007).
Analysis of > 9000 bp has not resolved deep-level relationships in Saxifragales (Fishbein et al., 2001; D. Soltis et al., 2007). The primary purpose of this study was therefore to assess whether much larger data sets (up to approximately 50,000 bp) could provide strong support for deep-level relationships within this problematic group, as a test case for resolving ancient, rapid radiations in angiosperms. In addition, we explored whether rapidly and slowly evolving gene regions provide equivalent estimates of phylogeny in Saxifragales by comparing the phylogenetic signal of 10 relatively rapidly evolving protein-coding genes in the two single-copy (SC) regions of the plastid genome to that of the eight protein-coding genes within the plastid inverted repeat (IR), a slowly evolving region of approximately 25,000 bp. Rates of substitution at synonymous sites are about 4 to 5 times lower in the IR than in SC regions of the plastid genome (based on our data for Saxifragales, as well as other comparisons; e.g., Wolfe et al., 1987). The slow rate of evolution of this region appears to be due, in part, to the presence of the highly conserved 16S and 23S rRNA genes, as well as frequent intramolecular recombination that homogenizes the two IR copies (Palmer, 1985). Sequencing of the entire IR region has recently been facilitated by the rapid PCR-based "ASAP" method (Amplification, Sequencing and Annotation of Plastomes; Dhingra and Folta, 2005). Finally, given this large amount of data, we also examined the impact of using different numbers of taxa (17 cersus 28) to estimate divergence times within Saxifragales.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
Sampling
Each family of Saxifragales (sensu APG II, 2003; D. Soltis et al., 2005) was represented in our analyses. Nearly all of the species are in just three families—Saxifragaceae (1400), Crassulaceae (500), and Hamamelidaceae (100). For these larger families (plus Haloragaceae), multiple genera spanning the phylogenetic diversity of the group were included. Most families of Saxifagales are small, consisting of only one genus (e.g., Cercidiphyllaceae, Daphniphyllaceae, Grossulariaceae, Pterostemonaceae, and Paeoniaceae) or two or three genera (e.g., Altingiaceae, Iteaceae, Peridiscaceae). The final data set included 25 taxa of Saxifragales. Three outgroups were sampled: one from the rosid clade, Vitis (Vitaceae); and two basal eudicots, Platanus (Platanaceae) and Trochodendron (Trochodendraceae). Species, voucher/collection information, and GenBank accession numbers are given in Supplementary Table 1 (www.systematicbiology.org).
|
DNA Amplification and Sequencing
We isolated DNA following standard CTAB protocols (Doyle and Doyle, 1987; D. Soltis et al., 1991). We targeted 11 specific genes for sequencing—7 plastid genes from the large and small SC regions as well as 4 mitochondrial genes; all targeted genes and primers used for PCR and sequencing are provided in Supplementary Table 2 (www.SystematicBiology.org). These 11 targeted genes were added to an existing data set (Fishbein et al., 2001; D. Soltis et al., 2007) of 3 SC plastid genes (rbcL, atpB, matK) and 2 nuclear genes (18S rDNA and 26S rDNA) for a total of 16 targeted genes (2 nuclear; 4 mitochondrial; 10 plastid; Table 1). We also used the ASAP method (Dhingra and Folta, 2005) to obtain the sequence of the entire IR for 15 of the 25 taxa of Saxifragales in our analyses (due to expense, we did not sequence the IR for all 25 taxa). The complete plastid sequences of Platanus (Moore et al., 2006) and Vitis (Jansen et al., 2006) were used as outgroups for the IR data set, yielding a total data set of 17 taxa. The 15 ingroup taxa were chosen to be phylogenetically representative of all the major diversity in Saxifragales. We used the primers of Dhingra and Folta (2005) for IR amplification and sequencing. The PCR reactions and thermocycler profiles for the targeted plastid and nuclear genes followed the general methods of Qiu et al. (2005) and Mavrodiev et al. (2005), those for the mtDNA genes followed Qiu et al. (2006a), and those for the inverted repeat followed Dhingra and Folta (2005). PCR products were purified using the Wizard SV PCR Clean-up System for the targeted plastid genes from the SC regions, Qiagen QIAquick PCR purification kit for the mtDNA genes, and ExoSAP for the IR. Most sequences were generated on an ABI 3730 XL DNA sequencer (Applied Biosystems, Inc., Fullerton, CA) following the manufacturer's protocols; mtDNA products were sequenced on an ABI 3100 Genetic Analyzer. The IR sequences were annotated using DOGMA (Wyman et al., 2004).
|
Alignment and Phylogenetic Analysis
DNA sequences were aligned using Clustal X (Thompson et al., 1997; Jeanmougin et al., 1998). We used the default gap penalties to do the initial alignment. For several areas that were difficult to align, we adjusted the gap-opening penalty from 15 (default) to 20. When our visual inspection of the CLUSTAL alignments suggested that any portion of these alignments was clearly in error, we adjusted the alignment manually, as needed. To assist in the alignment of coding regions, sequences were also aligned by amino acid. Although alignment was straightforward across almost all sequence data, several short, highly variable regions that were difficult to align were excluded from analyses, as were several short regions at the beginnings and ends of genes for which sequences were incomplete. There were also three short inversions that were corrected in the IR portion of the alignment: positions 2008 to 2015 of the alignment for Myriophyllum and Penthorum, positions 4876 to 4880 for Kalanchoe, and positions 24,947 to 24,966 of Paeonia. The total aligned lengths and the aligned, analyzed lengths of all genes are given in Table 1.
To test the phylogenetic effects of analyzing different organellar genomes as well as different regions of the plastid genome, we analyzed the following data partitions for both the 17-and 28-taxon data sets: (1) 4 mitochondrial genes; (2) 2 nuclear ribosomal RNA genes (18S rDNA, 26S rDNA); (3) 10 plastid genes from the SC region; (4) entire plastid IR; and (5) all plastid regions (3 plus 4). Several total evidence data sets were also analyzed: (1) all sequence data for 17 taxa (equivalent to combining partitions 1 to 4 above); (2) all sequence data for 28 taxa, excluding the IR (which was not sequenced for 11 of the 28 taxa)—this is referred to as the 16-gene data set (equivalent to combining partitions 1 to 3 above); and (3) all sequence data for 28 taxa, including the IR data when present (equivalent to combining partitions 1 to 4 above). Within most of the above partitions, protein-coding plastid and mitochondrial genes were additionally analyzed by codon position (first and second positions combined, and third positions separately) using parsimony.
The plastid protein-coding genes in the data set were also partitioned into three rate classes as follows. First, we calculated the average of the uncorrected pairwise (p) distances across all taxa for each gene in PAUP, as well as the proportion of parsimony-informative (PI) characters as a function of analyzed, aligned sequence length for each gene. We then ranked the genes according to average p distance. This ranking was strongly correlated with the ranking based on PI characters/sequence length. A clear break was observed between rpoC2 and rps4 in both p distance and PI characters/sequence length, and a reasonably clear break was observed between psbN and ycf1 in PI characters/sequence length (Table 2). These breaks were used to partition the plastid data into "fast-evolving" genes (average pairwise distance greater than 0.080, proportion of parsimony-informative characters greater than 0.130), "intermediate-evolving" genes (average pairwise distance between 0.026 and 0.080, proportion of parsimony-informative characters between 0.050 and 0.130), and "slow-evolving" genes (average pairwise distance less than 0.026, proportion of parsimony-informative characters less than 0.050) for the 17-taxon analyses (Table 2). The "slow" category coincided with the IR protein-coding genes. We recognize that categorization of rates of overall gene evolution is complex and that "fast-evolving" genes may have slowly evolving sites, and vice versa (e.g., Olmstead et al., 1998; P. Soltis et al., 1999), but we used these categories in an effort to address the general question of how genes with different substitution rates perform at resolving deep nodes in a problematic group.
We used maximum parsimony (MP), maximum likelihood (ML), and Bayesian analysis to infer phylogeny. MP analyses were conducted using PAUP* 4.0 (Swofford, 2000). Shortest trees were obtained using heuristic searches and 1000 replicates of random taxon addition with tree-bisection-reconnection (TBR) branch swapping, saving all shortest trees per replicate. Bootstrap support for relationships (Felsenstein, 1985) was estimated from 1000 bootstrap replicates using 10 random taxon additions per replicate, with TBR branch swapping (saving all trees).
For ML analyses, we employed the program GARLI (Genetic Algorithm for Rapid Likelihood Inference; version 0.942; Zwickl, 2006). Unlike MrBayes, which can apply different models to different partitions within a data set, GARLI can only use a single model for the entire data set. With this restriction, MrModelTest indicated that GTR+I+
was the preferred model for all data sets for the GARLI analyses. Analyses were run with default options, except that the "significanttopochange" parameter was reduced to 0.01 to make searches more stringent. ML bootstrap analyses were conducted with the default parameters with 100 replicates.
Bayesian analyses were conducted with the MPI-enabled version of MrBayes 3.1.2 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003; Altekar et al., 2004), splitting runs across processors. Each analysis consisted of six independent runs, with four chains (one cold and three hot) each, of two million generations. Chains were sampled every 100 generations. For each data set, MrModelTest 2.2 (Nylander, 2004) was used to determine the appropriate evolutionary model. Data sets consisting of only protein-coding genes were also partitioned by codon position. Codon position-based character partitions were established in MrBayes based on the model selected by the Akaike information criterion (AIC), unlinking the substitution rates, character state frequencies, gamma shape parameter, and proportion of invariable sites among partitions. All other parameters were left at their default values. Based on visual inspection of log-likelihood over time plots, burn-in was set at 20,000 generations for all runs. In addition, the web tool AWTY (Wilgenbusch et al., 2004) was used to plot posterior probabilities of splits over time, and the MrBayes output was examined to ensure that the standard deviation of splits among multiple runs was under 0.01 in all cases. Posterior probabilities were calculated using the resulting trees from all six runs.
Divergence Time Estimates
Trees used in divergence time analyses were those that resulted from GARLI searches. Branch lengths and model parameters of GARLI trees, however, were reestimated using PAUP* under a GTR+I+ model of sequence evolution (Zwickl, 2006). In all cases, a likelihood-ratio test (LRT) was performed to test for rate constancy among lineages (Felsenstein, 1981). The molecular clock hypothesis was strongly rejected (P < 0.001) for all of the data sets. Because of the absence of rate constancy in our data, we used the penalized likelihood (PL: Sanderson, 2002, 2003) method to estimate divergence times in Saxifragales. PL is a semiparametric smoothing method that assumes an autocorrelation in substitution rates and attempts to minimize rate changes between ancestral/descendant branches on a tree (i.e., at the nodes). Branches are allowed to change rates of molecular evolution but are penalized when rates change from ancestral to descendant branches. A smoothing parameter (
) was determined by cross-validation.
Several fossil dates were used as calibration points or minimum age constraints: Hamamelidaceae (84 to 86 Ma; Zhou et al., 2001), Cercidiphyllum (Cercidiphyllaceae; 65 to 71 Ma; Magallón et al., 1999), Altingia (Altingiaceae; 88.5 to 90.4 Ma; Zhou et al., 2001), Liquidambar (Altingiaceae; 11.6 to 15.9 Ma; Pigg et al., 2004), and Divisestylus (89.5 to 93.5 Ma; Hermsen et al., 2003). Recent molecular analyses suggest that Altingia and Liquidambar may not be distinct lineages (i.e., Liquidambar is derived from within Altingia s.l.; however, morphological analysis strongly supports Altingia and Liquidambar as mutually exclusive sister clades (Wen, 1999; Zhang et al., 2003; Ickert-Bond et al., 2005, 2007). We applied the Liquidambar constraint to the node representing the common ancestor of Liquidambar and Altingia but realized it may be nested higher within a clade of Altingiaceae s.l.
Three pairs of data sets and trees were used to estimate divergence times: (1) 16 targeted genes from three genomes for 17 and 28 taxa; (2) all plastid genes for 17 and 28 taxa; and 3) "total evidence" for 17 and 28 taxa. Tree topologies and branch lengths were those that resulted from the ML estimates of the corresponding data set. For each of the six data sets, two different sets of analyses were performed to estimate divergence times. First, each of the five fossils was used separately as a fixed calibration point, and ages were estimated using PL. Second, each of the fossils was separately treated as a fixed calibration point, whereas the remaining fossils were enforced as minimum age constraints. We adopted a nonparametric bootstrap approach (Efron, 1981; Felsenstein, 1985; Baldwin and Sanderson, 1998) to generate standard errors for divergence time estimates. We used PAUP* to generate all bootstrap samples of trees with branch lengths.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
Characteristics of Gene Sequences
The total aligned length for each gene is given in Table 1. The total evidence data sets (17 and 28 taxa) consisted of 50,845 bp. Of these, 27,647 bp represented the entire IR region of the plastid genome. In addition, 23,198 bp represented 16 targeted genes (12,059 bp for 10 SC plastid genes; 5946 bp for 4 mtDNA genes; 5193 bp for 2 nuclear genes). Complete (or nearly so) sequences were obtained for all 10 SC plastid regions for all 28 taxa (Table 1), with the exceptions of rps4 and rpoC2 for Tetracarpaea and matK for Peridiscus. For the mtDNA partition, data are missing for all four genes for Choristylis, Pterostemon, and Saxifraga integrifolia; matR, nad5, and rps3 for Crassula; and rps3 for Aphanopetalum. These missing data resulted from having only small amounts of partially degraded DNA for taxa that were difficult to resample.
Phylogenetic Relationships: Data Partitions and Total Evidence
We consider first the three subcellular data partitions (nuclear, mtDNA, plastid) and combinations thereof (e.g., all plastid genes; total evidence); we then provide results for "fast," "intermediate," and "slow" genes. Summaries of relationships and internal support for clades are given in Table 3 and Table 4. Only the trees from the largest combined data sets are shown. The data matrix file and all tree files are provided as supplementary data on the Systematic Biology Web site (www.systematicbiology.org) and in TreeBASE (accession S1902). The analyzed length (equal to the aligned length minus excluded sites), number of parsimony-informative characters, consistency index (CI), and retention index (RI) are given for all data partitions in Table 5.
|
|
|
Bayesian and ML analyses recovered the same topology in all partitions except the nuclear data (Table 3; Fig. 1). In Bayesian and ML analyses of total evidence, plastid, and mitochondrial partitions, Peridiscaceae are sister to all remaining Saxifragales, whereas Paeoniaceae are sister to the woody clade (Cercidiphyllaceae, Daphniphyllaceae, Altingiaceae, and Hamamelidaceae). This Paeoniaceae + woody clade is, in turn, sister to core Saxifragales. Core Saxifragales comprise two subclades: Crassulaceae + Haloragaceae s.l. (i.e., Aphanopetalum, Tetracarpaeaceae, Penthoraceae, and Haloragaceae) and the Saxifragaceae alliance. Within the Saxifragaceae alliance, Saxifragaceae + Grossulariaceae are sister to Iteaceae + Pterostemonaceae. In contrast, combined nuclear data (18S and 26S rDNA) provide generally poor resolution and support within Saxifragales. For example, with the nuclear data set, parsimony recovered only one major subclade with BS > 50%: Haloragaceae s.l. + Crassulaceae (BS = 80%). This low support and resolution agrees with earlier analyses of these genes in Saxifragales (Fishbein et al., 2001).
The consensus topology described above is obtained using Bayesian and ML methods for Saxifragales with both the 17-and 28-taxon data sets. The posterior probabilities for most of these relationships are very high (pp > 0.95) and increase with the addition/combination of partitions (Table 3). For example, in the complete plastid tree (28 taxa), all clades have pp = 1.0, except the clade of Altingiaceae + Hamamelidaceae (pp = 0.94). Similarly, in the total evidence tree (28 taxa), all clades have posterior probabilities of 1.0, except one relationship within Saxifragaceae (Fig. 1).
In contrast to the identical (or nearly so) topologies obtained in Bayesian and ML analyses, the topologies vary across gene partitions (Table 3), as well as across codon positions (Table 6a, Table 6b), in MP analyses. In particular, the position of Paeoniaceae varies as sister to Peridiscaceae, the Saxifragaceae alliance, Crassulaceae, or Crassulaceae + Haloragaceae s.l. (Table 3). However, in most MP trees (particularly of the largest data sets; i.e., total evidence, all plastid data), Paeoniaceae are sister to Peridiscaceae, and this sister group is sister to the remaining Saxifragales. For example, MP analysis of the total evidence data set for 28 taxa resulted in a single shortest tree (Fig. 2; Table 3) with Peridiscaceae + Paeoniaceae (BS = 63%) sister to all other Saxifragales (BS = 54%). The woody clade (BS = 87%) is sister to core Saxifragales (BS = 78%) consisting of Crassulaceae and Haloragaceae s.l. (BS = 100%) plus the Saxifragaceae alliance (BS = 100%). Analysis of the 17-taxon total evidence data set yielded the same topology and similar bootstrap support (Table 3).
|
|
|
Paeoniaceae are characterized by a long branch, and the phylogenetic position of this family varies. Removal of long-branch taxa and subsequent reanalysis of the data set can provide important phylogenetic insights (Bergsten, 2005). Removal of Paeonia from all but the nuclear data set yielded a topology identical (or nearly so) to the Bayesian and ML trees obtained for most partitions (Fig. 1). The nuclear topology was also similar to these topologies, but with Paeonia removed, Peridiscaceae are sister to the core Saxifragales clade rather than to all remaining Saxifragales. Furthermore, most nodes have high BS values with Paeoniaceae removed (supplementary trees). For example, when Paeoniaceae are removed from the 28-taxon total evidence data set, Peridiscaceae are sister to the rest of Saxifragales (BS = 86%), and the woody clade has BS = 91%.
Significantly, MP analyses of the slow-evolving IR (see section below) and mtDNA genes also yielded the topology consistently found with Bayesian and ML methods. Four shortest mtDNA trees were obtained, and two of those matched the Bayesian and ML trees (Fig. 1), with Paeoniaceae sister to the woody clade; in the other two trees Paeoniaceae are sister to core Saxifragales. MP analysis of all coding plastid data recovered a topology similar to the Bayesian and ML tree that was repeatedly obtained; however, Paeoniaceae are sister to Haloragaceae s.l. + Crassulaceae (and not sister to the woody clade).
MP Analyses of Fast, Intermediate, and Slow Plastid Genes
MP analysis of the 17-taxon "slow" gene data set revealed a topology essentially identical to that obtained in ML and Bayesian analyses of multiple partitions (Fig. 3a; Table 4). With slow genes, MP placed Peridiscaceae sister to the remainder of the clade (BS = 88%) and recovered two major clades with bootstrap support > 50%. One of these (BS = 79%) consists of Paeoniaceae as sister to the woody families (BS = 51%). The second major clade is core Saxifragales (BS = 90%), which is composed of two subclades as in the total evidence tree: Saxifragaceae alliance (BS = 98%) and Crassulaceae + Haloragaceae s.l. (BS = 99%). The slow-gene tree differed from the Bayesian and ML total-evidence tree only in the lack of resolution and support within the woody Saxifragales clade (Fig. 3a).
|
MP analysis of the combined first and second codon positions for the coding regions of the slow data set resulted in one shortest tree essentially identical to those described above with all IR sequence but with more resolution within the woody clade. However, whereas Paeoniaceae are again sister to the woody clade, BS support for the monophyly of the woody clade is < 50% (Table 6a, Table 6b). MP analysis of the third codon positions for the slow data set (Table 6a, Table 6b) resulted in eight shortest trees; the strict consensus is again nearly identical to those described for other IR data sets, but support for relationships is often lower. Peridiscaceae are sister to all other Saxifragales (but without BS > 50%). Paeoniaceae are part of a clade (BS = 50%) with the woody families, forming a trichotomy with Hamamelidaceae and a clade of Cercidiphyllaceae, Daphniphyllaceae, and Altingiaceae. That is, the woody clade is not recovered. A core Saxifragales clade (BS = 53%) comprises the Saxifragaceae alliance (BS = 86%) and Crassulaceae + Haloragaceae s.l. (BS = 81%). The lower resolution and support observed in the third codon position analysis likely resulted from the lower number of parsimony-informative characters at this codon position (95) versus the first and second positions combined (142), especially because all other tree statistics (CI, RI) are similar between these partitions (Table 5).
Parsimony analysis of "fast" genes alone (all codon positions; Fig. 3c; Table 4) recovered one shortest tree with two major clades: (1) Peridiscaceae + the Saxifragaceae alliance; (2) Paeoniaceae + (Haloragaceae s.l. + Crassulaceae) as sister to the clade of woody families. The positions of neither Peridiscaceae nor Paeoniaceae receive BS > 50%. Only three major subclades (all tip clades) have BS > 50%: (1) Saxifragaceae alliance (BS = 100%); (2) Crassulaceae + Haloragaceae s.l. (BS = 95%); and (3) the woody clade (BS = 57%). Thus, MP analysis of these fast genes does not resolve deep-level relationships to the extent observed with other partitions. Fast genes performed better at resolving relationships at the tips of clades; they recovered a monophyletic Hamamelidaceae (BS = 51%), Saxifragaceae (BS = 100%), Saxifragaceae + Grossulariaceae (BS = 100%), and Crassulaceae (BS = 100%).
MP analysis of the "intermediate" data set resulted in six shortest trees (Fig. 3b). The core Saxifragales clade (BS = 61%) is recovered with two subclades: the Saxifragaceae alliance (BS = 81%) and Crassulaceae + Haloragaceae s.l. (BS = 82%). Relationships among the remaining members of Saxifragales were poorly resolved and supported. Peridiscaceae and Paeoniaceae formed a clade in five of the shortest trees (without BS > 50%). A woody clade of Cercidiphyllaceae, Daphniphyllaceae, Hamamelidaceae, and Altingiaceae was not recovered.
When intermediate and fast genes (i.e., all SC plastid genes) were combined, the tree (not shown) recovers the 17-taxon total evidence tree obtained with MP but with lower support. Paeoniaceae are sister to Peridiscaceae (without BS > 50%), and this clade is sister to the rest of Saxifragales (again without BS support > 50%). Two large clades are recovered: woody Saxifragales (BS = 52%) and core Saxifragales (BS = 66%), the latter with two subclades: (1) Crassulaceae + Haloragaceae s.l. (96%) and (2) the Saxifragaceae alliance (100%).
The higher resolution and support (especially at deep levels) with slow compared to fast genes is not the result of more parsimony-informative sites—in fact, the slow data set had far fewer parsimony-informative sites (237 of 11,538 bp) than did either the fast (940 of 5847 bp) or intermediate (464 of 5079 bp) data sets (Table 2).
Divergence Time Estimates
For all trees used for divergence time estimation, likelihood scores calculated with PAUP* were nearly identical to those produced by GARLI. Differences in branch lengths between the two optimization methods had no effect on divergence times for these data sets.
Taxon density greatly affected the age estimates (Supplementary Table 3, Table 4; www.SystematicBiology.org), both for clades within Saxifragales as well as for Saxifragales itself. Age estimates obtained using the 17-taxon data sets are generally much older than those based on the 28-taxon data set (Fig. 4). For example, the age estimate for the Saxifragaceae alliance based on the total-evidence 28-taxon data set and the Cercidiphyllaceae fossil without constraints is 87.6 (±6.3) Ma, but that from the 17-taxon data set is 152.5 (±10.2) Ma. The 28-taxon age estimates are more in line with fossil dates and expectations; across the three data sets, the 17-taxon estimates for the age of Saxifragales range from 189.7 to 257.7 Ma, a time frame that is older than most current age estimates for the origin of the angiosperms (e.g., Sanderson et al., 2004; Bell et al., 2005). On average, age estimates based on the 28-taxon data set (for 16 genes) were younger than those based on the 28-taxon data set for plastid genes only. This pattern was also observed in the data sets consisting of 17 taxa (Supplementary Table 3, Table 4).
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
Resolving Ancient Radiations
Saxifragales are considered to represent an ancient, rapid radiation that occurred during the early diversification of the eudicot angiosperms (Fishbein et al., 2001; D. Soltis et al., 2005). Fossil evidence suggests that the radiation may have occurred 89.5 to 110 Ma (Magallón et al., 1999; Hermsen et al., 2006). Based on the total-evidence 28-taxon data set, our age estimates indicate that Saxifragales originated sometime before 112 (±9.7) to 120 (±10.2) Ma and diversified rapidly. Importantly, a similar window of some time before 100 to 120 Ma appears to be common for the diversification of other major clades of angiosperms, including some rosid lineages such as Malpighiales (Davis et al., 2005); several major asterid lineages, including Ericales, Cornales (Bremer et al., 2004; Sytsma et al., 2006), and Dipsacales (Bell and Donoghue, 2005); and most major clades of monocots (Bremer, 2000).
Our molecular estimates also suggest that the early radiation of Saxifragales may have occurred over a time span as short as 6 million years. The woody clade appears to represent a second early radiation within Saxifragales. Age estimates for this clade range from 90 (±0) to 106 (± 7.7) Ma (with constraints), and it, too, may have diversified over a narrow time span (perhaps only 3 to 6 million years). These calculations are based on differences in age inferred between the crown group nodes between the two groups. The variation observed in age estimates based on different fossil treatments is much greater than that based on standard errors calculated from bootstrap resampling of branch lengths (Supplementary Table 3, Table 4). These estimates for Saxifragales and the woody clade represent a timeframe comparable to the rapid radiation of the Hawaiian silversword alliance (Asteraceae-Madiinae), which putatively arose from a North American ancestor 5 Ma (Baldwin et al., 1991; Baldwin and Sanderson, 1998; Barrier et al., 1999).
In prior analyses, deep-level relationships within Saxifragales were not resolved, and internal support for deep nodes was low, despite the use of > 9000 bp of sequence data (Fishbein et al., 2001; D. Soltis et al., 2007). In this study, combining data sets yielded increased internal support for clades. Considering the Bayesian and ML topologies, total evidence (50,845 bp) provided the best-resolved and supported topology for Saxifragales, with higher posterior probabilities than obtained with smaller data sets, such as 10 plastid genes (12,059 bp), 16 genes (23,198 bp), or the inverted repeat (27,647 bp). The Saxifragales clade is just one of several examples of deep-level phylogenetic problems in the angiosperms (many recognized at the ordinal level in APG II) that appear to be of comparable ages (D. Soltis et al., 2005; Wikström et al., 2001). Extrapolating from the current investigation, reliable resolution of other deep-level radiations in the angiosperms (that date to 90 to 130 Ma) may require well over 10,000 bp and perhaps as many as 25,000 to 50,000 bp of sequence data.
Problems with Parsimony and Paeoniaceae
With the largest data sets (16-gene, IR, all plastid sequences, total evidence), Bayesian analyses resulted in identical topologies, and most nodes possessed pp = 1.0. Only a few internal nodes (within Saxifragaceae; within Hamamelidaceae) had a pp < 1.0. The Bayesian results suggest in particular that the ancient, rapid radiation has been resolved.
In contrast, deep relationships within Saxifragales were not well resolved or supported based on parsimony analyses. Across the different partitions and sampling schemes, parsimony varied in the placement of Paeoniaceae (Table 3, Table 4). The family appears in several positions, including as sister to Peridiscaceae, the woody clade, Crassulaceae, Haloragaceae s.l. + Crassulaceae, and the Saxifragaceae alliance. However, bootstrap support is generally < 50%, and the highest values are < 70%. Even across codon positions, the placement of Paeoniaceae varied (Table 5). The most common placement of Paeoniaceae with parsimony is sister to Peridiscaceae, with this clade sister to the remainder of Saxifragales.
Perhaps as a result of these problems with Paeoniaceae, deep-level relationships are not strongly supported with parsimony. The largest data sets (all plastid data and total evidence) place Peridiscaceae + Paeoniaceae as sister to the remainder of Saxifragales. With both of these data sets, parsimony support for the core Saxifragales clade is also low (< 80%). The woody clade receives BS support of only 83% and 87%, respectively, with these data sets. Only the Saxifragaceae alliance and Crassulaceae + Haloragaceae s.l. are consistently strongly supported (BS = 100%) with the largest data sets.
Parsimony versus ML and Bayesian methods provided different placements of Paeoniaceae in earlier analyses with five genes (Fishbein et al., 2001; D. Soltis et al., 2007), a result attributed to the long branch of Paeoniaceae. In MP analyses of a five-gene (> 9000 bp) data set, Paeoniaceae appeared as sister to Peridiscaceae (with this sister group embedded within Hamamelidaceae); in Bayesian and ML analyses Peridiscaceae appeared as sister to the remainder of Saxifragales, and Paeoniaceae were part of the core Saxifragales (D. Soltis et al., 2007). These results are similar to those obtained here with many more genes, but key basal nodes received low pp values (less than 0.85) and low bootstrap values in the five-gene studies.
Previous efforts to break up the long branch to Paeoniaceae by adding additional species of Paeonia (the only genus in the family) failed (Fishbein et al., 2001). Parsimony is known to be inconsistent under certain conditions, wherein the addition of characters leads to increased confidence in an incorrect topology (Felsenstein, 1978; Huelsenbeck and Hillis, 1993; Sullivan and Swofford, 1997; Bergsten, 2005). The addition of 40,000 new base pairs to the Saxifragales data set increased the severity of the long-branch problem for Paeoniaceae. With the three largest data sets for both 17 and 28 taxa, bootstrap support was highest for the placement of Paeoniaceae with Peridiscaceae. MP analysis of the total evidence data set, the plastid data set, and the 16-gene data set all revealed that Paeoniaceae are sister to Peridiscaceae with BS values > 65%.
Impact of Substitution Rate on Phylogeny Estimation
Wortley et al. (2005) suggested that increasing sequence length using genes with a slow or moderate substitution rate is a better way to improve accuracy than employing sequences with a faster substitution rate because the latter may increase the likelihood that nucleotide substitutional saturation will result in erroneous phylogenetic reconstruction. For Lamiales, a large clade of perhaps 30,000 species, Wortley et al. (2005) suggested that at least 2,000 parsimony-informative characters, corresponding to a sequence data set of 10,000 bp or more, should result in a fully resolved and well-supported phylogeny.
In parsimony analyses (Fig. 2, Fig. 3; Table 4), slowly evolving genes generally perform better in elucidating deep-level relationships within Saxifragales than do genes with intermediate or fast rates of evolution. That is, slowly evolving genes appear to provide higher phylogenetic accuracy than do fast or intermediate genes (reviewed in Hillis, 1995). With slowly evolving genes (the IR data set), a topology comparable to the total evidence Bayesian and ML trees was obtained, despite the fact that the slow-evolving genes had by far the fewest parsimony-informative sites (Table 2). In contrast, the intermediate and fast-evolving gene data sets provide fully resolved trees when analyzed with parsimony, but recover deep-level relationships that appear to be incorrect. These data support the contention (Felsenstein, 1983; Graham and Olmstead, 2000; Wortley et al., 2005) that slow-evolving genes provide fewer sites, but those sites are phylogenetically informative and less homoplasious than variable sites in fast genes. However, these results are somewhat at odds with studies that have shown the value of rapidly evolving genes and third codon positions in resolving green plant (Källersjö et al., 1998, 1999) and angiosperm (Hilu et al., 2003; Borsch et al., 2003) relationships. Importantly, the fast and slow data sets for Saxifragales are also somewhat complementary in that fast genes provide resolution and support for some clades not recovered by slow genes, and vice versa. For example, the intermediate and fast genes provided resolution within the woody Saxifragales that was not realized with slow-evolving genes.
Our analyses also suggest that the plastid IR region may be an ideal source of genes for resolution of deep-level (90 to 130 Ma) angiosperm relationships because of its size ( 25,000 bp) and relatively low substitution rate. Furthermore, the IR region can be readily amplified and sequenced across the angiosperms using the rapid PCR-based ASAP method (Dhingra and Folta, 2005). With recent reductions in sequencing costs, we estimate that the entire IR can be sequenced for $200 to $250 per species (this includes DNA isolation, primer costs, PCR, and sequencing), making it practical to employ this region routinely.
Systematic Implications
Five families of Saxifragales are composed primarily of trees; the placements of all have been considered problematic. One of these, Peridiscaceae, is sister to the remainder of the Saxifragales clade. The other woody families (Hamamelidaceae, Altingiaceae, Cercidiphyllaceae, and Daphniphyllaceae) are here shown to form what we term the woody clade. Within this clade, Altingiaceae are sister to the remaining members, with Cercidiphyllaceae and Daphniphyllaceae, two problematic families in Saxifragales in previous studies (Fishbein et al., 2001; D. Soltis et al., 2007), forming a subclade.
Peridiscaceae and the woody clade share a number of anatomical features, including imperforate tracheary elements with bordered pits, apotracheal, diffuse wood parenchyma, and vessels with scalariform perforation plates (reviewed in D. Soltis et al., 2007). However, these features may be plesiomorphic based on the placement of these families in the shortest trees. Hermsen et al. (2006) failed to find morphological synapomorphies for this clade. Hence, clear non-DNA synapormorphies of the woody clade remain uncertain.
In Bayesian and ML analyses, Paeoniaceae are sister to the woody clade with all but the nuclear data set. PP and BS values for this placement are high (pp = 1.00; BS = 100%) for the largest data sets (e.g., total evidence, IR, all plastid genes). Given the variability in placement of Paeoniaceae in parsimony analyses, and the long branch to this taxon, is its placement with the woody clade supported by nonmolecular evidence? The relationships of Paeoniaceae have long been problematic based on morphology (reviewed in D. Soltis et al., 2007). Despite diverse placements, a relationship to these woody families had not been suggested previously. Paeoniaceae, Altingiaceae, Cercidiphyllaceae, Hamamelidaceae, and Daphniphyllaceae share vessels with scalariform perforation plates, an anatomical feature that occurs only rarely in core Saxifragales (Aphanopetalum of Haloragaceae). But this feature is also present in Peridiscaceae (see above) and may therefore be symplesiomorphic or perhaps synapomorphic for Saxifragales. Hence, there is no strong corroborating morphological evidence to support the Bayesian placement of Paeoniaceae with the woody clade (see also Hermsen et al., 2006) despite high posterior probability values (pp = 1.0 with the largest data sets).
A clade of core Saxifragales composed of primarily herbaceous taxa was suggested in earlier molecular analyses (Fishbein et al., 2001; D. Soltis et al., 2007) and in an analysis of molecular plus morphological data (Hermsen et al., 2006); it emerges with high pp and BS values with most data sets in Bayesian and ML analyses. With parsimony, core Saxifragales have BS support of 88% in the entire IR analysis and of 78% in the total evidence analysis. Again, however, morphological synapomorphies for this clade remain unknown (D. Soltis et al., 2005; Hermsen et al., 2006). Within core Saxifragales, relationships routinely receive high pp and BS values and agree with previous results (D. Soltis et al., 2000, 2007; Fishbein et al., 2001; Davis and Chase, 2004).
Divergence Time Estimates and Taxon Density
Age estimates for Saxifragales, as well as clades within Saxifragales, are greatly affected by taxon sampling. In general, the age estimates obtained using the 17-taxon data sets are substantially older than those for 28 taxa. The 28-taxon age estimates are more in line with fossil dates and expectations and from previous age estimates of angiosperms (Fig. 4; Supplemental Table 3 and Table 4). For example, the 17-taxon estimates (with constraints) for the age of the ingroup (142 ± 12.8 to 257 ± 17 Ma) is older than most current age estimates for the origin of the angiosperms (e.g., Sun et al., 2002; Sanderson et al., 2004; Bell et al., 2005). In contrast, the age of the Saxifragaceae alliance is generally between 89 (±6.2) and 96 (±7.3) Ma with the 28-taxon data set (with constraints), which is close to the age (90 Ma) of the recently described fossil Divisestylus (Iteaceae; Hermsen et al., 2003), a member of the Saxifragaceae alliance. This estimate is also in agreement with minimum age mapping estimates (90 Ma, Hermsen et al., 2006) and some molecular estimates based on rbcL alone (91 Ma; Anderson et al., 2005).
Our results indicate that sufficient taxon density, and not simply numerous base pairs, is crucial for reliably estimating ages. The caveats of divergence time estimation have been discussed elsewhere (Hillis et al., 1996; Sanderson, 1998; Sanderson and Doyle, 2001; P. Soltis et al., 2002; Sanderson et al., 2004; Bell and Donoghue, 2005; Bell et al., 2005; Yang and Rannala, 2006). However, until recently, much less attention has been paid to the effect of taxon sampling (see Linder et al., 2005). If the molecular clock hypothesis held across taxa, taxon sampling would not be a critical issue, as long as enough taxa were sampled to place fossils as calibration points or minimum age constraints on a tree. Just as sufficient taxon density is important for estimating phylogeny, it is also important for estimating the ages of clades in the absence of rate homogeneity. Many of the widely used methods that relax the assumption of a strict molecular clock (e.g., nonparametric rate smoothing, NPRS; penalized likelihood, PL; and Bayesian relaxed clock, BRC) try to accommodate rate variation among lineages into the estimation procedure by assuming that rate changes across a tree are autocorrelated from parental to descendant branches. This assumption of rate autocorrelation, however, may be violated in some cases and therefore could lead to an error in age estimates (Ho et al., 2005; Drummond et al., 2006). As with the estimation of phylogenies, sparse taxon sampling could lead to long branches in an inferred topology. Transition in rates from parental branches to these long branches would potentially be inferred as large shifts in rates because these branches are not subdivided by additional unsampled taxa. In our case, we observed similar mean rates across branches regardless of the number of taxa sampled; however, the variance was substantially larger in the 17-taxon data sets. These taxon sampling effects have been investigated using empirical data (Linder et al., 2005); all methods are sensitive to undersampling, but dates estimated using PL were less affected by taxon sampling than those inferred using other approaches. Furthermore, the effect is most severe in analyses that use more extreme rate smoothing (e.g., NPRS). "Oversmoothing" results in overestimation of ages of deep nodes and underestimation of ages of shallow nodes. Additionally, the undersampling effect is positively related to the distance of an estimated clade from the calibration point used in the tree. Our results thus support those of Linder et al. (2005) and further underscore the importance of taxon sampling in molecular dating analyses. In Saxifragales, the combined effect of undersampling and distance from the calibration point and age constraints resulted in as much as a threefold difference in ages estimated using 17-versus 28-taxon data sets, with the same partition and calibrations (Supplementary Table 3, Table 4).
|
|
Acknowledgments |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
This research was supported in part by an Assembling the Tree of Life (AToL) grant EF-0431266 (NSF) and by the Deep Time RCN, DEB-0090283 (NSF). The authors thank Jeff Thorne and Tom Britton for much discussion concerning divergence time estimation and Bin Wang and Libo Li for help with data collection. The University of Florida High-Performance Computing Center and Phyloinformatics Cluster for High-Performance Computing in the Life Sciences provided computational resources and support that have contributed to the research results reported in this paper.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
-
Altekar G., Dwarkadas S., Huelsenbeck J. P., Ronquist F. Parallel Metropolis-coupled Markov chain Monte Carlo for Bayesian phylogenetic inference. Bioinformatics (2004) 20:407–415.
Anderberg A. A., Rydin C., Källersjö M. Phylogenetic relationships in the order Ericales s.l.: Analyses of molecular data from five genes from the plastid and mitochondrial genomes. Am. J. Bot. (2002) 89:677–687.
Anderson C. A., Bremer K., Friis E. M. Dating phylogenetically basal eudicots using rbcL sequences and multiple fossil reference points. Am. J. Bot. (2005) 92:1737–1748.
APG II. An update of the Angiosperm Phylogeny Group classification for the orders and families of flowering plants. Bot. J. Linn. Soc. (2003) 141:399–436.[CrossRef][Web of Science]
Baldwin B. G., Kyhos D. W., Dvorak J., Carr G. D. Chloroplast DNA evidence for a North American origin of the Hawaiian silversword alliance (Asteraceae). Proc. Natl. Acad. Sci. USA (1991) 88:1840–1843.
Baldwin B. G., Sanderson M. J. Age and rate of diversification of the Hawaiian silversword alliance (Compositae). Proc. Natl. Acad. Sci. USA (1998) 95:9402–9406.
Barrier M., Baldwin B. G., Robichaux R. H., Purugganan M. D. Interspecific hybrid ancestry of a plant adaptive radiation: Allopolyploidy of the Hawaiian silversword alliance (Asteraceae) inferred from floral homeotic gene duplications. Mol. Biol. Evol. (1999) 16:1105–1113.[Abstract]
Baurain D., Brinkmann H., Philippe H. Lack of resolution in the animal phylogeny: Closely spaced cladogeneses or undetected systematic errors. Mol. Biol. Evol. (2007) 24:6–9.
Bell C. D., Donoghue M. J. Dating the Dipsacales: Comparing models, genes, and evolutionary implications. Am. J. Bot. (2005) 92:284–296.
Bell C. D., Soltis D. E., Soltis P. The age of the angiosperms: A molecular time scale without a clock. Evolution (2005) 59:1245–1258.[CrossRef][Web of Science][Medline]
Bergsten J. A review of long-branch attraction. Cladistics (2005) 21:163–193.[CrossRef][Web of Science]
Borsch T., Hilu K. W., Quandt D., Wilde V., Neinhuis C., Barthlott W. Non-coding plastid trnT-trnF sequences reveal a well resolved phylogeny of basal angiosperms. J. Evol. Biol. (2003) 16:558–576.[CrossRef][Web of Science][Medline]
Bremer B., Bremer K., Heirdari N., Erixon P., Olmstead R. G., Källersjö M., Anderberg A. A., Barkhordarian E. Phylogenetics of asterids based on 3 coding and 3 non-coding chloroplast DNA markers and the utility of non-coding DNA at higher taxonomic levels. Mol. Phylogenet. Evol. (2002) 24:274–301.[CrossRef][Web of Science][Medline]
Bremer K. Early Cretaceous lineages of monocot flowering plants. Proc. Natl. Acad. Sci. USA. (2000) 97:4707–4711.
Bremer K., Friis E. M., Bremer B. Molecular phylogenetic dating of asterid flowering plants shows early Cretaceous diversification. Syst. Biol. (2004) 53:496–505.
Burleigh J. G., Mathews S. Phylogenetic signal in nucleotide data from seed plants: Implications for resolving the seed plant tree of life. Am. J. Bot. (2004) 91:1599–1613.
Crane P. R. The origin and early diversification of angiosperms. Nature (1995) 374:27–33.[CrossRef]
Cronquist A. An integrated system of classification of flowering plants (1981) New York: Columbia University Press.
Cuénoud P., Savolainen V., Chatrou L. W., Powell M., Grayer R. J., Chase M. W. Molecular phylogenetics of Caryophyllales based on nuclear 18S rDNA and plastid rbcL, atpB, matK DNA sequences. Am. J. Bot. (2002) 89:132–144.
Davis C. C., Anderson W. R., Donoghue M. J. Phylogeny of Malpighiaceae: Evidence form chloroplast ndhF trnL-F nucleotide sequences. Am. J. Bot. (2001) 88:1830–1846.
Davis C. C., Chase M. W. Elatinaceae are sister to Malpighiaceae; Peridiscaceae belong to Saxifragales. Am. J. Bot. (2004) 91:262–273.
Davis C. C., Webb C. O., Wurdack K. J., Jaramillo C. A., Donoghue M. J. Explosive radiation of Malpighiales supports a mid-Cretaceous origin of modern tropical rain forests. Am. Nat. (2005) 165:E36–E65.[CrossRef][Web of Science][Medline]
Dhingra A., Folta K. M. ASAP: Amplification, sequencing and annotation of plastomes. BMC Genomics (2005) 6:176.[CrossRef][Medline]
Doyle J. J., Doyle J. L. A rapid DNA isolation procedure from small quantities of fresh leaf tissue. Phytochem. Bull. (1987) 19:11–15.
Drinnan A. N., Crane P. R., Hoot S. B. Patterns of floral evolution in the early diversification of non-magnoliid dicotyledons (eudicots). Plant Syst. Evol. Suppl. (1994) 8:93–122.
Drummond A. J., Ho S. Y. W., Phillips M. J., Rambaut A. Relaxed phylogenetics and dating with confidence. PLoS Biol. (2006) 4:e88.[CrossRef][Medline]
Efron B. Nonparametric estimates of standard error: The jackknife, the bootstrap and other methods. Biometrika (1981) 68:589–599.
Felsenstein J. Evolutionary trees from DNA sequences: A maximum likelihood approach. J. Mol. Evol. (1981) 17:368–376.[CrossRef][Web of Science][Medline]
Felsenstein J. Parsimony in systematics: Biological and statistical issues. Annu. Rev. Ecol. Syst. (1983) 14:313–333.[CrossRef][Web of Science]
Felsenstein J. Confidence limits on phylogeny: An approach using the bootstrap. Evolution (1985) 39:783–791.[CrossRef][Web of Science]
Fishbein M., Hibsch-Jetter C., Soltis D. E., Hufford L. Phylogeny of Saxifragales (angiosperms, eudicots): Analysis of a rapid, ancient radiation. Syst. Biol. (2001) 50:817–847.
Goldman N. Evaluationg the phylogenetic utility of genes: A search for genes informative about deep divergences among vertebrates. Syst. Biol. (1994) 43:174–193.
Graham S. W., Olmstead R. G. Utility of 17 chloroplast genes for inferring the phylogeny of the basal angiosperms. Am. J. Bot. (2000) 87:1712–1730.
Hermsen E. J., Gandolfo M. A., Nixon K. C., Crepet W. L. Divisestylus gen. nov. (aff. Iteaceae), a fossil saxifrage from the late Cretaceous of New Jersey, USA. Am. J. Bot. (2003) 90:1373–1388.
Hermsen E. J., Gandolfo M. A., Nixon K. C., Crepet W. L. The impact of extinct taxa on understanding the early evolution of angiosperm clades: An example incorporating fossil reproductive structures of Saxifragales. Plant Syst. Evol. (2006) 260:141–169.
Hillis D. M. Approaches for assessing phylogenetic accuracy. Syst. Biol. (1995) 44:3–16.
Hillis D. M. Taxonomic sampling, phylogenetic accuracy, and investigator bias. Syst. Biol. (1998) 47:3–8.
Hillis D. M., Mable B. K., Moritz C. Applications of molecular systematics: The state of the field and a look to the future. In: Molecular systematics—Hillis D. M., Moritz C., Mable B. K., eds. (1996) Sunderland, Massachusetts: Sinauer Associates. 515–543.
Hilu K. W., Borsch T., Müller K., Soltis D. E., Soltis P. S., Savolainen V., Chase M. W., Powell M. P., Alice L. A., Evans R., Sauquet H., Neinhuis C., Slotta T. A. B., Rohwer J. G., Campbell C. S., Chatrou L. W. Angiosperm phylogeny based on matK sequence information. Am. J. Bot. (2004) 90:1758–1766.[CrossRef][Web of Science]
Ho S. Y. W., Phillips M. J., Drummond A. J., Cooper A. Accuracy of rate estimation using relaxed-clock models with a critical focus on the early metazoan radiation. Mol. Biol. Evol. (2005) 22:1355–1363.
Hoot S. B., Magallón S., Crane P. R. Phylogeny of basal eudicots based on three molecular data sets: atpB rbcL, and 18S nuclear ribosomal DNA sequences. Ann. Missouri Bot. Gard. (1999) 86:1–32.[CrossRef][Web of Science]
Huelsenbeck J. P., Hillis D. M. Success of phylogenetic methods in the four-taxon case. Syst. Biol. (1993) 42:247–264.
Huelsenbeck J. P., Ronquist F. MrBayes: Bayesian inference of phylogeny. Bioinformatics (2001) 17:754–755.
Ickert-Bond S. M., Pigg K. B., Wen J. Comparative infructescence morphology in Liquidambar (Altingiaceae) and its evolutionary significance. Am. J. Bot. (2005) 92:1234–1255.
Ickert-Bond S. M., Pigg K. B., Wen J. Comparative infructescence morphology in Altingia (Altingiaceae) and discordance between morphological and molecular phylogenies. Am. J. Bot. (2007) 94:1094–1115.
Jansen R. K., Kaittanis C., Saski C., Lee S-B., Tomkins J., Alverson A. J., Daniell H. Phylogenetic analyses of Vitis (Vitaceae) based on complete chloroplast genome sequences: Effects of taxon sampling and phylogenetic methods on resolving relationships among rosids. BMC Evol. Biol. (2006) 6:32.[CrossRef][Medline]
Jeanmougin F., Thompson J. D., Gibson T. J., Gouy M., Higgins D. G. Multiple sequence alignment with Clustal X. Trends Biochem. Sci. (1998) 23:403–405.[CrossRef][Web of Science][Medline]
Källersjö M., Albert V. A., Farris J. S. Homoplasy increases phylogenetic structure. Cladistics (1999) 15:91–93.[CrossRef][Web of Science]
Källersjö M., Farris J. S., Chase M., Bremer B., Fay M. F., Humphries C. J., Peterson G., Seberg O., Bremer K. Simultaneous parsimony jackknife analysis of 2538 rbcL DNA sequences reveals support for major clades of green plants, land plants, seed plants, and flowering plants. Plant Syst. Evol. (1998) 213:259–287.[CrossRef]
Linder H. P., Hardy C. F., Rutschmann F. Taxon sampling effects in molecular clock dating: An example from the African Restionaceae. Mol. Phylogenet. Evol. (2005) 35:569–582.[CrossRef][Web of Science][Medline]
Magallón S., Crane P. R., Herendeen P. S. Phylogenetic pattern, diversity and diversification of eudicots. Ann. Missouri Bot. Gard. (1999) 86:297–372.[CrossRef][Web of Science]
Mavrodiev E. V., Tancig M., Sherwood A. M., Gitzendanner M. A., Rocca J., Soltis P. S., Soltis D. E. Phylogeny of Tragopogon L. (Asteraceae) based on internal and external transcribed spacer sequence data. Int. J. Plant Sci. (2005) 166:117–133.[CrossRef][Web of Science]
Moore M. J., Dhingra A., Soltis P. S., Shaw R., Farmerie W. G., Folta K. M., Soltis D. E. Rapid and accurate pyrosequencing of angiosperm plastid genomes. BMC Plant Biol. (2006) 6:17.[CrossRef][Medline]
Morgan D. R., Soltis D. E. Phylogenetic relationships among Saxifragaceae sensu lato based on rbcL sequence data. Ann. Missouri Bot. Gard. (1993) 80:631–660.[CrossRef][Web of Science]
Moritz C., Dowling T. E., Brown W. M. Evolution of animal mitochondrial DNA: Implications for population biology and systematics. Ann. Rev. Ecol. Syst. (1987) 18:269–292.[CrossRef][Web of Science]
Nickrent D. L., Joshua P. D., Frank E. A. Discovery of the photosynthetic relatives of the "Maltese mushroom" Cynomorium. BMC Evol. Biol. (2005) 5:38.[CrossRef][Medline]
Nylander J. A. A. MrModelTest v2. Program distributed by the editor (2004) Uppsala, Sweden: Evolutionary Biology Centre, Uppsala University.
Olmstead R. G., dePamphilis C. W., Wolfe A. D., Young N. D., Elisens W. J., Reeves A. Disintegration of the Scrophulariaceae. Am. J. Bot. (2001) 88:348–361.
Olmstead R. G., Reeves P. A., Yen A. C. Patterns of sequence evolution and implications for parsimony analysis of chloroplast DNA. In: Molecular systematics of plants II: DNA sequencing—Soltis D. E., Soltis P. S., Doyle J. J., eds. (1998) Boston: Kluwer Academic Publishers. 164–187.
Olmstead R. G., Sweere J. A. Combining data in phylogenetic systematics: An empirical approach using three molecular data sets in the Solanaceae. Syst. Biol. (1994) 43:467–481.
Palmer J. D. Evolution of chloroplast and mitochondrial DNA in plants and algae. In: Molecular evolutionary genetics—MacIntyre R. J., ed. (1985) New York: Plenum Press. 131–240.
Pigg K. B., Ickert-Bond S. M., Wen J. Anatomically preserved Liquidambar (Altingiaceae) from the middle Miocene of Yakima Canyon, Washington State, USA, and its biogeographic implications. Am. J. Bot. (2004) 91:499–509.
Qiu Y.-L., Chase M. W., Hoot S. B., Conti E., Crane P. R., Sytsma K. J., Parks C. R. Phylogenetics of the Hamamelidae and their allies: Parsimony analyses of nucleotide sequences of the plastid gene rbcL. Int. J. Plant Sci. (1998) 159:891–905.[Web of Science]
Qiu Y.-L., Dombrovska O., Lee J., Li L., Whitlock B. A., Bernasconi-Quadroni F., Rest J. S., Davis C. C., Borsch T., Hilu K. W., Renner S. S., Soltis D. E., Soltis P. S., Zanis M. J., Cannone J. J., Gutell R. R., Powell M., Savolainen V., Chatrou L. W., Chase M. W. Phylogenetic analysis of basal angiosperms based on nine plastid, mitochondrial, and nuclear genes. Int. J. Plant Sci. (2005) 166:815–842.[CrossRef][Web of Science]
Qiu Y.-L., Li L., Hendry T. A., Li R., Taylor D. W., Issa M. J., Ronen A. J., Vekaria M. L., White A. M. Reconstructing the basal angiosperm phylogeny: Evaluating information content of the mitochondrial genes. Taxon (2006a) 55:837–856.[Web of Science]
Qiu Y.-L., Li L., Wang B., Chen Z., Knoop V., Groth-Malonek M., Dombrovska O., Lee J., Kent L., Rest J., Estabrook G. F., Hendry T. A., Taylor D. W., Testa C. M., Ambros M., Crandall-Stotler B., Duff R. J., Stech M., Frey W., Quandt D., Davis C. C. The deepest divergences in land plants inferred from phylogenomic evidence. Proc. Nat. Acad. Sci. USA (2006b) 103:15511–15516.
Rokas A., Krüger D., Carroll S. B. Animal evolution and the molecular signature of radiations compressed in time. Science (2005) 310:1933–1938.
Ronquist F., Huelsenbeck J. P. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics (2003) 19:1572–1574.
Sanderson M. J. Estimating rate and time in molecular phylogenies: Beyond the molecular clock? In: Molecular systematics of plants II: DNA sequencing—Soltis D. E., Soltis P. S., Doyle J. J., eds. (1998) Boston: Kluwer Academic Publishers. 242–264.
Sanderson M. J. Estimating absolute rates of molecular evolution and divergence times: A penalized likelihood approach. Mol. Biol. Evol. (2002) 19:101–109.
Sanderson M. J. r8s: Inferring absolute rates of molecular evolution and divergence times in the absence of a molecular clock. Bioinformatics (2003) 19:301–302.
Sanderson M. J., Doyle J. A. Sources of error and confidence intervals in estimating the age of angiosperms from rbcL and 18S rDNA data Am. J. Bot. (2001) 88:1499–1516.[CrossRef]
Sanderson M. J., Thorne J. L., Wikström N., Bremer K. Molecular evidence on plant divergence times. Am. J. Bot. (2004) 90:1656–1665.
Shaw J., Cox C. J., Goffinet B., Buck W. R., Boles S. B. Phylogenetic evidence of a rapid radiation of pleurocarpous mosses (Bryophyta). Evolution (2003) 57:2226–2241.[Web of Science][Medline]
Shaw J., Renzaglia K. Phylogeny and diversification of bryophytes. Am. J. Bot. (2004) 91:1557–1581.
Soltis D. E., Clayton J. W., Davis C. C., Gitzendanner M. A., Cheek M., Savolainen V., Amorim A. M., Soltis P. S., Chase M. W. Monophyly and relationships of the enigmatic amphitropical family Peridiscaceae. Taxon, (2007) 56:65–73.[Web of Science]
Soltis D. E., Hibsch-Jetter C., Soltis P. S., Chase M., Farris J. S. Molecular phylogenetic relationships among angiosperms: An overview based on rbcL and 18S rDNA sequences. In: Evolution and Diversification of Land Plants—Watsuki K., Raven P. H., eds. (1997a) Tokyo: Springer. 157–178.
Soltis D. E., Soltis P. S. Phylogenetic relationships among Saxifragaceae sensu lato: A comparison of topologies based on 18S and rbcL sequences. Am. J. Bot. (1997) 84:504–522.[Abstract]
Soltis D. E., Soltis P. S., Chase M. W., Mort M. E., Albach D. C., Zanis M., Savolainen V., Hahn W. H., Hoot S. B., Fay M. F., Axtell M., Swensen S. M., Prince L. M., Kress W. J., Nixon K. C., Farris J. S. Angiosperm phylogeny inferred from a combined data set of 18S rDNA, rbcL atpB sequences. Bot. J. Linn. Soc. (2000) 133:381–461.[CrossRef][Web of Science]
Soltis D. E., Soltis P. S., Collier T. G., Edgerton M. L. Chloroplast DNA variation within and among genera of the Heuchera group: Evidence for chloroplast capture and paraphyly. Am. J. Bot. (1991) 78:1091–1112.[CrossRef][Web of Science]
Soltis D. E., Soltis P. S., Endress P. K., Chase M. W. Phylogeny and evolution of angiosperms (2005) Sunderland, Massachusetts: Sinauer Associates.
Soltis D. E., Soltis P. S., Mort M. E., Chase M. W., Savolainen V., Hoot S. B., Morton C. M. Inferring complex phylogenies using parsimony: An empirical approach using three large DNA data sets for angiosperms. Syst. Biol. (1998) 47:32–42.
Soltis D. E., Soltis P. S., Nickrent D. L., Johnson L. A., Hahn W. J., Hoot S. B., Sweere J. A., Kuzoff R. K., Kron K. A., Chase M., Swensen S. M., Zimmer E., Chaw S. M., Gillespie L. J., Kress W. J., Sytsma K. Angiosperm phylogeny inferred from 18S ribosomal DNA sequences. Ann. Missouri Bot. Gard. (1997b) 84:1–49.[CrossRef][Web of Science]
Soltis P. S., Soltis D. E., Chase M. W. Angiosperm phylogeny inferred from multiple genes as a research tool for comparative biology. Nature (1999) 402:402–404.[CrossRef]
Soltis P. S., Soltis D. E., Savolainen V., Crane P. R., Barraclough T. Rate heterogeneity among lineages of land plants: Integration of molecular and fossil data and evidence for molecular living fossils. Proc. Nat. Acad. Sci. USA. (2002) 99:4430–4435.
Sullivan J., Swofford D. L. Are Guinea pigs rodents? The importance of adequate models in molecular phylogenetics. J. Mamm. Evol. (1997) 4:77–86.[CrossRef]
Sun G., Ji Q., Dilcher D. L., Zheng S., Nixon K. C., Wang X. Archaefructaceae, a new basal angiosperm family. Science (2002) 296:899–904.
Swofford D. L. PAUP*. Phylogenetic analysis using parsimony (*and other methods), version 4 (2000) Sunderland, Massachusetts: Sinauer Associates.
Sytsma K. J., Walker J. B., Schönenberger J., Anderberg A. A. The evolution of Ericales: Recent insights using both morphology and molecules (2006) Botany (abstract).
Takhtajan A. Diversity and classification of flowering plants (1997) New York: Columbia University Press.
Thompson J. D., Gibson T. J., Plewniak F., Jeanmougin F., Higgins D. G. The CLUSTAL_X windows interface: Flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. (1997) 25:4876–4882.
Townsend J. P. Profiling phylogenetic informativeness. Syst. Biol. (2007) 56:222–231.
Wen J. Evolution of eastern Asian and eastern North American disjunct distributions in flowering plants. Annu. Rev. Ecol. Syst. (1999) 30:421–455.[CrossRef][Web of Science]
Wikström N., Savolainen V., Chase M. W. Evolution of the angiosperms: Calibrating the family tree. Proc. R. Soc. Lond. B (2001) 268:2211–2220.[Medline]
Wilgenbusch J. C., Warren D. L., Swofford D. L. AWTY: A system for graphical exploration of MCMC convergence in Bayesian phylogenetic inference (2004) http://ceb.csit.fsu.edu/awty.
Wolfe K. H., Li W-H., Sharp P. M. Rates of nucleotide substitution vary greatly among mitochondrial, chloroplast, and nuclear DNAs. Proc. Natl. Acad. Sci. USA (1987) 84:9054–9058.
Wortley A. H., Rudall P. J., Harris D. J., Scotland R. W. How much data are needed to resolve a difficult phylogeny? Case study in Lamiales. Syst. Biol. (2005) 54:697–709.
Wyman S. K., Jansen R. K., Boore J. L. Automatic annotation of organellar genomes with DOGMA. Bioinformatics (2004) 20:3252–3255.
Yang Z. On the best evolutionary rate for phylogenetic analysis. Syst. Biol. (1998) 47:125–133.
Yang Z., Rannala B. Bayesian estimation of species divergence times under a molecular clock using fossil calibrations with soft bounds. Mol. Biol. Evol. (2006) 23:212–226.
Yoon H. S., Müller K. M., Sheath R. G., Ott F. D., Bhattacharya D. The major lineages of red algae (Rhodophyta). J. Phycol. (2006) 42:482–492.[CrossRef][Web of Science]
Zhang Z.-Y., Zhang H.-T., Endress P. K. Hamamelidaceae. In: Flora of China—Wu Z.-Y., Raven P. H., Hong D.-Y., eds. (2003) Beijing, China: SciencePress. 18–42.
Zhou Z.-K., Crepet W. L., Nixon K. C. The earliest fossil evidence of the Hamamelidaceae: Late Cretaceous (Turonian) inflorescences and fruits of Altingioideae. Am. J. Bot. (2001) 88:753–766.
Zwickl D. J. Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the Maximum Likelihood criterion (2006) Austin: University of Texas. PhD dissertation.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  K. J. Wurdack and C. C. Davis Malpighiales phylogenetics: Gaining ground on one of the most recalcitrant clades in the angiosperm tree of life Am. J. Botany, August 1, 2009; 96(8): 1551 - 1570. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. E. Pfeil and M. D. Crisp The age and biogeography of Citrus and the orange subfamily (Rutaceae: Aurantioideae) in Australasia and New Caledonia Am. J. Botany, December 1, 2008; 95(12): 1621 - 1631. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||