© 2008 Society of Systematic Biologists
Rooting and Dating Maples (Acer) with an Uncorrelated-Rates Molecular Clock: Implications for North American/Asian Disjunctions
1 Department of Biology, University of Munich D-80638 Munich, Germany; E-mail: renner{at}lrz.uni-muenchen.de
2 Institute of Geosciences, University of Tuebingen D-72076, Tuebingen, Germany
3 Department of Biogeography and Botanical Garden, University of Vienna A-1030 Vienna, Austria
4 Department of Systematic and Evolutionary Botany, University of Vienna A-1030 Vienna, Austria
5 Department of Biology, University of Missouri–St. Louis St. Louis, Missouri 63121, USA
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionAcknowledgmentReferences |
|---|
Simulations suggest that molecular clock analyses can correctly identify the root of a tree even when the clock assumption is severely violated. Clock-based rooting of phylogenies may be particularly useful when outgroup rooting is problematic. Here, we explore relaxed-clock rooting in the Acer/Dipteronia clade of Sapindaceae, which comprises genera of highly uneven species richness and problematic mutual monophyly. Using an approach that does not presuppose rate autocorrelation between ancestral and descendant branches and hence does not require a rooted a priori topology, we analyzed data from up to seven chloroplast loci for some 50 ingroup species. For comparison, we used midpoint and outgroup rooting and dating methods that rely on rooted input trees, namely penalized likelihood, a Bayesian autocorrelated-rates model, and a strict clock. The chloroplast sequences used here reject a single global substitution rate, and the assumption of autocorrelated rates was also rejected. The root was placed between Acer and Dipteronia by all three rooting methods, albeit with low statistical support. Analyses of Acer diversification with a lineage-through-time plot and different survival models, although sensitive to missing data, suggest a gradual decrease in the average diversification rate. The nine North American species of Acer diverged from their nearest relatives at widely different times: eastern American Acer diverged in the Oligocene and Late Miocene; western American species in the Late Eocene and Mid Miocene; and the Acer core clade, including A. saccharum, dates to the Miocene. Recent diversification in North America is strikingly rare compared to diversification in eastern Asia.
Keywords: Bayes factors; biogeography; Bayesian relaxed clock; Beringian disjunctions; rooting of phylogenies; substitution rates
Received November 1, 2007; Revised January 28, 2008; Accepted July 15, 2008
Rooted phylogenetic trees are essential for inferring the sequence of evolutionary events. Trees can be rooted with outgroups (Maddison et al., 1984; Wheeler, 1990), midpoint-rooting (the tree is rooted at the midpoint of the longest path connecting any pair of taxa; Swofford and Begle, 1993), suitable gene duplications (Iwabe et al., 1989; Mathews and Donoghue, 1999), insertion/deletion-rooting (Lake et al., 2007), paleontological or other circumstantial evidence (Huelsenbeck, 1994), nonreversible models of substitution and asymmetric step matrices (Swofford et al., 1996; Yap and Speed, 2005), or molecular clocks (Huelsenbeck et al., 2002). All ultrametric trees, constructed assuming a clock, have a root as the starting point and tips, or leaves, as the end points, usually in the present. Using real and simulated datasets, Huelsenbeck et al. (2002) compared outgroup rooting, molecular clock–based rooting, and nonreversible models of DNA substitution and found that the outgroup and molecular clock criteria correctly identified the root a large proportion of the time. As expected, the outgroup criterion performed best when the length of the branch leading to the outgroup was short (Wheeler, 1990). With increasing genetic distance between the ingroup and the outgroup, the performance of outgroup-rooting decreased, while molecular clock–based rooting correctly identified the root of the tree even when the clock criterion was severely violated. Under the most extreme conditions of rate heterogeneity, the clock criterion identified the correct root roughly four times more often than did random rooting. (Rooting via a nonreversible model of DNA substitution performed poorly under most circumstances.) These results suggest that the root of a phylogenetic tree can be identified with "clock rooting" even when the substitution process is not strictly clocklike.
Unclock-like data are the rule, and many methods of estimating divergence times therefore relax the clock assumption (Rambaut and Bromham, 1998; Thorne et al., 1998; Aris-Brosou and Yang, 2002; Sanderson, 2002; Thorne and Kishino, 2002; Britton et al., 2007). These relaxed clock approaches all require a "known" rooted input tree. Recently an approach has become available that coestimates phylogeny and divergence times (Drummond et al., 2006). The new method, Bayesian evolutionary analysis sampling trees, or BEAST, by relying neither on an input topology nor on a (potentially misleading and/or too distant) outgroup, can be used to identify the root of a phylogenetic tree (Drummond et al., 2006). Here, we explore the application of relaxed-clock-rooting to resolve a difficult problem, namely the root of the genus Acer. This problem likely has resulted from the effects of numerous extinctions and a possible shift in net diversification rate (speciation rate minus extinction rate) in the genus, which we therefore also address.
Acer comprises 124 to 156 species (van Gelderen et al., 1994; de Jong, 2002) and is the largest tree genus of the Northern Hemisphere besides Quercus. Knowing the root of Acer is relevant for reconstructing its evolution and biogeography, and in molecular-clock dating, the correct position of the root is essential (Sanderson and Doyle, 2001). An erroneous rooting will greatly influence (and often invalidate) all estimates of divergence times. Because Acer is such a large and widespread clade, inferring its diversification has implications for our understanding of the Tertiary vegetation history of the Northern Hemisphere (Wolfe and Tanai, 1987; Boulter et al., 1996; Manchester, 1999). Today, most species of Acer occur in Eurasia, particularly eastern Asia, whereas North America harbors only nine species: A. circinatum, A. glabrum, and A. macrophyllum from the western United States; A. pensylvanicum and A. spicatum from the northeastern United States; and the widespread A. negundo, A. rubrum, A. saccharinum, and A. saccharum. During the Tertiary, however, the diversity of Acer in North America appears to have been much greater, especially in the western United States (Wolfe and Tanai, 1987). In Europe, as well, many species of maples appear to have gone extinct during the Late Oligocene–Early Miocene (Walther, 1972; Wolfe and Tanai, 1987; Boulter et al., 1996).
The closest relative of Acer is Dipteronia, a genus of two species endemic to mainland China (Qiu et al., 2007). Whereas Acer has two-winged fruits (samaras) of a type that evolved several times in Sapindales (e.g., in Thouinia, Thouinopsis, and Serjania), Dipteronia has circular-winged fruits that are almost unique among angiosperms. The two species, D. sinensis and D. dyeriana, are so similar to each other that they have sometimes been ranked as subspecies of a single species (D. sinensis subsp. dyeriana). However, molecular studies of Acer, which have relied mainly on nuclear rDNA (the ITS region), found D. sinensis nested inside Acer (Cho et al., 1996; Suh et al., 2000; Pfosser et al., 2002, Tian et al., 2002; Grimm et al., 2006). This was especially surprising in view of Paleocene fruit fossils of both genera (Wolfe and Tanai, 1987; Crane et al., 1990; Manchester, 1999; McClain and Manchester, 2001; Kittle et al., 2005), suggesting that they existed a long time ago.
The closest relative of Acer and Dipteronia within Sapindaceae is a clade comprised of Aesculus, with 13 to 19 species in China, Europe, and North America; Billia, with two species in Central America and northern South America; and Handeliodendron, with a single species in China (Harrington et al., 2005). There is thus a striking asymmetry in extant species diversity between Acer and its relatives, perhaps pointing to differential rates of extinction and/or speciation in the geographic areas occupied by the maple lineage as well as the more inclusive Hippocastanoideae lineage.
To reconstruct the evolutionary unfolding of Acer, it is necessary to know both the root of the genus and the absolute divergence times of key lineages. Here we use combined data from up to seven chloroplast loci and relaxed-clock rooting, as well as outgroup and midpoint rooting, to address three issues: (i) the identity of the root of Acer/Dipteronia and the age of the Acer crown group; (ii) whether the diversification rate in the genus has shifted over the Tertiary; and (iii) the phylogenetic placement and temporal context of the extant North American species of Acer. We focus on the American species because our species-level sampling of North American maples is complete, and the timing of the American–East Asian disjunctions has been the focus of a previous investigation (Wolfe, 1981).
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentReferences |
|---|
Taxon Sampling
Appendix 1 (http://www.systematicbiology.org) lists the 74 species and subspecies of Acer and Dipteronia that were sequenced, with their sources and NCBI GenBank accession numbers. Date matrices have been deposited in TreeBASE as study accession number S2163. Species concepts follow the monograph of van Gelderen et al. (1994). The sampled species represent all sections accepted by de Jong (1994, 2002), except for Wardiana, which contains only A. wardii W. W. Smith. Previous studies found A. wardii embedded in the Palmata clade (nuclear rDNA ITS data of Grimm et al., 2006) or in a large polytomy (chloroplast data of Li et al., 2006).
We sequenced most species for the chloroplast (cp) rbcL gene, the rpl16 intron, the psbA-trnH spacer, and the trnL intron and adjacent trnL-F spacer. The slowly evolving rbcL gene was not sequenced for each species of Acer (see Appendix 1). Sequences of the cp trnD-trnT and psbM-trnD spacers from Li et al. (2006) were downloaded from GenBank as specified in Appendix 1. Where spacer regions could not be reliably aligned between ingroup and distant outgroups, they were replaced by question marks. To maximize signal, a few analyses included just 24 species (6674 characters) selected to represent the most divergent species.
For outgroup rooting, we included species of Aesculus and Handeliodendron from subfamily Hippocastanoideae of Sapindaceae plus representatives of Dodonaeoideae (Conchopetalum and Dodonaea) and Sapindoideae (Koelreuteria), all based on Harrington et al. (2005). Two of the outgroup taxa were composites because some loci could not be sequenced for each species (Appendix 1). This did not affect tree topology, as tested by running analyses with and without the Aesculus and Koelreuteria composites (data not shown). To assess a possible substitution rate change in the stem of the Acer/Dipteronia clade, we added rbcL sequences of Acer and Dipteronia to an rbcL data set of all Sapindales families (Muellner et al., 2007) for a total of 98 or 128 sequences. GenBank numbers for the non-Acer/Dipteronia rbcL sequences can be found in Muellner et al. (2007). Sapindales contain several taxa with fossil records, and therefore the large rbcL data set also permitted estimating divergence times in Acer and Dipteronia without relying on ingroup fossils.
DNA Sequencing, Editing, and Alignment
DNA extraction, cloning, and sequencing followed the methods described in Renner et al. (2007), which also contains information on the primers used. Forward and reverse reads were obtained for most samples. Sequences were edited with Sequencher (4.6; Gene Codes, Ann Arbor, Michigan) and aligned by eye, using MacClade 4.06 (Maddison and Maddison, 2003). Species newly sequenced for this study include A. fabri, A. kweilinense, and Handeliodendron bodinieri.
Phylogenetic Analyses
Data matrices for the various DNA loci were analyzed separately under parsimony optimization and in the absence of statistically supported conflict were combined. Parsimony searches relied on PAUP version 4.0b10 (Swofford, 2002), using a parsimony ratchet analysis command block (Müller, 2004), with 200 replicates, 10 random taxon-addition cycles, and a single ratchet run. Gaps were treated as missing data.
Maximum likelihood (ML) searches were carried out with RAxML version 2.2.0 (Stamatakis, 2006). Selection of the best-fit models relied on ModelTest 3.7 (Posada and Crandall, 1998) and is described in Rooting and Divergence Time Estimation. Tree searches in RAxML used the GTR+
model, the only model implemented in RAxML version 2.2.0. Model parameters were estimated over the duration of specified runs and searches started from several random parsimony trees.
Statistical support was measured by maximum likelihood bootstrapping performed in RAxML, with 1000 replicates, and by maximum parsimony bootstrapping in PAUP, with 100,000 replicates, a single tree as the start tree, tree bisection-reconnection branch swapping, and the MulTrees option deactivated. For visualization of competing phylogenetic signal, a bipartition network based on 100 maximum likelihood bootstrap replicates (6577 bp from six chloroplast loci with two partitions and a GTR+ model each) was constructed using SplitsTree 4 (Huson and Bryant, 2006).
Rooting and Divergence Time Estimation
We used three methods to find the root of Acer: midpoint rooting, clock rooting, and outgroup rooting with more or less numerous representatives of Sapindaceae. Midpoint rooting forces the use of MINF (minimization of the F-value) optimization (Swofford and Begle, 1993), which assigns states at internal nodes such that the F-value of Farris (1972) is minimized. The effect of minimizing the F-value is that length is transferred from interior branches towards terminal branches whenever possible, minimizing the risk that groups will be arbitrarily resolved internally (Swofford and Begle, 1993).
Clock rooting relied on BEAST version 1.4.6 (Drummond and Rambaut, 2007). BEAST analyses were conducted on two matrices, one with the combined plastid sequences (6487 base pairs [bp]) from 53 species of Acer and Dipteronia, the other with 98 Sapindaceae rbcL sequences (1387 bp), including 18 from Acer and Dipteronia. Concatenated datasets were analyzed under partition-specific models, one for the rbcL gene, the other for all non-coding plastid regions. The best-fit substitution models under the Akaike information criterion in ModelTest were TrN++I for the coding region (rbcL) in the 53-taxon matrix, TVM++I for the non-coding regions in the 53-taxon matrix, and GTR++I for the coding region (rbcL) in the 98-taxon matrix. To reduce possible overparameterization, and given that the gamma shape parameter () and the proportion of invariable sites (I) are strongly correlated (Sullivan and Swofford, 2001), we dropped the estimation of I, instead increasing the number of rate categories for to 8, and we also used the slightly simpler TrN+ model for both partitions in the 53-taxon matrix.
To quantify the support for alternative root positions we used Bayes factors, which measure the change in support for one model versus another given the data (Suchard et al., 2001). They are the Bayesian analogue of the likelihood ratio test but have the advantage that they can be used to compare non-nested hypotheses, such as different topologies (Suchard et al., 2001, 2005). To this end, we obtained prior probabilities for root placements at 37 different positions along the backbone of the Acer/Dipteronia topology obtained with the 53-taxon dataset (13 of these positions resulted from alternative resolutions of polytomies) by sampling from the prior only, using BEAST (computations conducted at the computer cluster of the University of Vienna Schrödinger III; http://www.univie.ac.at/ZID/schroedinger/). Despite the comparatively large number of sampling points (350,000), several root positions were estimated to have prior probabilities of zero. To be able to include these in the calculation of Bayes factors, their prior probabilities were set to 2.857 x 10–6 (= 1/350,001) as an upper bound for the prior or to 3.527 x 10–83 (the number of rooted topologies for 53 taxa) as a reasonable lower bound. We use 2lnBF12 as the test statistic and consider 2lnBF > 10 as strong support for model 1. This widely used cutoff value (Kass and Raftery, 1995) has been found to give a 5% type I error rate, at least in the context of data partitioning (Brown and Lemmon, 2007).
After tuning the operators using BEAST's auto-optimization option, analyses in BEAST used UPGMA trees as the starting trees and a speciation model following a Yule process as tree prior, with two runs of 1 x 107 generations each, sampling every 250th generation, with a burn-in of 1 x 106 generations each. Results from individual runs were combined as recommended for phylogenetic MCMC analyses, instead of running single longer chains (Beiko et al., 2006), and effective sample sizes (ESS) for all estimated parameters and node ages were well above 100. Clades of interest were defined via most recent common ancestors (MRCA), and they could thus have varying taxon composition in the posterior. This effect will be particularly pronounced where relationships are uncertain because of weak signal. In our dataset, this affected our ability to calibrate the tree. In particular, the ambiguity concerning the monophyly of Dipteronia (see Results) affects the application of fossil calibrations for Acer and Dipteronia; if Dipteronia is inferred as paraphyletic or polyphyletic, a calibration point derived from Dipteronia fossils (not assignable to either of the two extant taxa, see below) cannot be interpreted as a minimum age of Dipteronia divergence.
The earliest fossils of Dipteronia (McClain and Manchester, 2001) are from the Paleocene Fort Union Formation in central Wyoming and are 60 to 63 Myr. Dipteronia fruits are also well represented at Middle Eocene localities in Washington, Oregon, and British Columbia, and at Late Eocene localities in Colorado, western Montana, and Oregon (McClain and Manchester, 2001). The only locality of Oligocene age is from the Bridge Creek flora of Oregon, 32 Myr. There are no Asian Dipteronia fossils. The oldest described Acer fossils are Paleocene fruits from North America and Eurasia (Wolfe and Tanai, 1987; Crane et al., 1990; Manchester, 1999; Kittle et al., 2005), but still older fruits from the Maastrichtian (66.5 Ma) Hell Creek formation in South Dakota also conform morphologically to Acer (K. R. Johnson, personal communication, August 2008). None of these ancient isolated fruits, however, can securely be assigned to an extant clade.
Based on this fossil record, we applied the following alternative constraints to the 53-taxon matrix (using BEAST). We first used an exponentially distributed prior with an offset of 60 and a mean of 1, which gave a prior mean age of 72.184 Ma and a median of 62.054 Ma for the age of the MRCA of Acer and Dipteronia. Alternatively, we employed a model that additionally constrained the Acer stem age. Here we used a normal prior distribution with a mean of 62 Ma and a standard deviation of 4 Ma, which resulted in a complex prior resembling an exponential distribution with an overall mean of 61.6 and a median of 61 Ma. We chose these different models for the prior distribution of fossil ages because the Dipteronia fossil from central Wyoming is more precisely dated than are the early Acer fossils. An exponential distribution of rates across branches is a one-parameter distribution (the parameter determines both the mean and the variance); it requires only the specification of the offset and the mean. By contrast, the normal prior distribution requires a mean and a standard deviation, which are not easy to choose. The normal distribution has the advantage, however, of not placing as much prior probability on a relatively narrow time frame as does an exponential distribution.
For BEAST analyses of the 98-taxon rbcL data set, minimal age constraints were again modeled as exponentially distributed priors, with the offsets given by fossils (below) and means of 1. The prior for the age of the root followed a uniform distribution between 0 and 137 Ma, based on the onset of angiosperm radiation (Hughes, 1994; Brenner, 1996). The minimal age of the stem lineage of Aesculus + Handeliodendron was constrained to 60 Ma (with the prior distribution following a simple exponential distribution with mean/median of 63.9/63.0 Ma), based on the oldest Aesculus fossils (Manchester, 2001). Of the more distant Acer outgroups, Biebersteiniaceae were constrained to minimally 54.8 Myr old, based on fossil pollen from the Neomugen Formation of Inner Mongolia (Late Paleocene, 57.0 to 54.8 Ma; Muellner et al., 2007); the prior distribution again followed an exponential distribution with 55.6/55.3 Ma mean/median. The Ailanthus stem lineage (Simaroubaceae) was constrained to minimally 52 Myr old, based on the fossil fruit taxon A. confucii, thought to be related to extant A. altissima (Corbett and Manchester, 2004), the species included in our DNA analyses. Here, the prior distributions were complex, comprising two overlapping exponential distributions with modes of ca. 52 and 55 Ma and an overall mean/median of 54.0/54.3 Ma. The stem of Cedreleae was constrained to minimally 49 Myr old, based on fruit and seed fossils of Toona from the London Clay (DeVore et al., 2005) that share morphological features of modern Toona and Cedrela (Muellner et al., 2007). The prior distribution for Cedreleae showed three overlapping exponential distributions with modes of ca. 49, 52, and 55 Ma and an overall mean/median of 50.9/50.2 Ma. To assess rate heterogeneity in the 98-taxon Sapindales rbcL dataset, we carried out local-clock analyses in PAML 3.14 (Yang, 1997) to compare the fit of models that assumed either a single rate for all Sapindales or two rates, one for Acer + Dipteronia, the other for the remaining Sapindales.
For comparison with the estimates obtained with the uncorrelated approach (in BEAST), we used three other clock approaches: the strict clock assumption of a single substitution rate (mostly carried out in PAUP); semiparametric rate smoothing (penalized likelihood) as implemented in r8s version 1.71 (Sanderson, 2002, 2004); and a Bayesian approach that assumes autocorrelated rates (and which therefore requires a rooted input phylogeny) as implemented in multidivtime (Thorne and Kishino, 2002). These analyses were necessarily carried out on data matrices that included outgroups. For strict clock analyses, we imported the highest likelihood tree found with RAxML into PAUP, outgroup-rooted it, and calculated branch lengths under a GTR++I+ clock model, with model parameters estimated in PAUP. The resulting branch length table was saved, the distance between a calibration node and the tip was then divided by the age of the calibration node to obtain a substitution rate, and this rate used to calculate the ages of other nodes of interest. For calibration of the strict clock, we sequentially explored the fossils described above.
For penalized likelihood analyses, we again imported the preferred ML topology into PAUP, outgroup-rooted it, and described its branch lengths under the GTR++I model. The resulting rooted trees with branch lengths became the input for r8s, outgroups were pruned from the tree, and cross-validation then used to determine the best level of clock enforcement (i.e., the smoothing parameter).
Multidivtime analyses were carried out as described in Renner et al. (2007), using an input topology obtained with RAxML and rooted on Aesculus. The minimal age of the Acer/Dipteronia stem lineage (i.e., the divergence of these two from Aesculus/Handeliodendron) was constrained to 63 Ma and the maximal age of the Hippocastanoideae stem to 72 Ma. This maximal age was chosen based on the age estimates for Sapindales clades of Muellner et al. (2007) and the oldest fossils of Acer, Dipteronia, and Aesculus, none of which predate 65 Ma; multidivtime does not perform well without at least one minimal and one maximal constraint (R. Thorne, multidivtime read-me file, 2003). Absolute ages and delimitation of fossil eras is from the geologic timescale of Gradstein et al. (2004).
Modeling the Acer Diversification Rate
We used a lineage-through-time (LTT) plot derived from an ultrametric tree (chronogram) for Acer/Dipteronia to graphically assess species diversification over time (Harvey et al., 1994). Taxon sampling in the chronogram included 50 of the estimated 124 or 156 species of Acer plus Dipteronia (van Gelderen et al., 1994; de Jong, 2002), and the LTT plot was therefore brought up to either 124 or to 156 living species. We also fit different survival models implemented in the APE package (Paradis, 1997; Paradis et al., 2004) to the chronogram to detect significant departures from a constant-rate null model of diversification (CR-BD). Two alternatives were compared to the CR-BD model, namely a model of a gradually changing diversification rate and one that assumes an abrupt change in rate before and after some breakpoint in the past. Because the constant-rate model is nested in both variable-rate models, it can be compared with each in a likelihood-ratio test (Paradis, 1997). Survival models can incorporate missing data (species known to belong to Acer but not sequenced) provided a minimum age is known for these species. In this study, we explored adding varying proportions of the 70 missing species at 5, 10, 15, 20, 25, 30, and 35 Myr. We did not use older origination times because we sampled all of the major clades within Acer. A worldwide analysis of maple fossils found striking decreases in occurrences of Acer macrofossils between 35 and 25 Myr in Europe, West Asia, and East Asia (Boulter et al. 1996: fig. 4), followed by a renewed increase after that period.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentReferences |
|---|
The Root of Acer
An outgroup-rooted ML tree obtained with the combined chloroplast data for 24 taxa (6674 characters, almost no missing data) is shown in Figure 1; Dipteronia and Acer appear mutually monophyletic, although the monophyly of Dipteronia is only weakly supported. With the dataset enlarged from 24 to 74 taxa by the inclusion of species that had not necessarily all been sequenced for rbcL, there was continued weak support for the mutual monophyly of Acer and Dipteronia. Midpoint rooting of the ML trees obtained for the 24 and 74 taxon datasets placed the root between Dipteronia and Acer. The backbones of the Acer trees were essentially unresolved, while there was considerable support for tip clades (labeled A to H in Figs. 1 to 3).
|
|
|
A relaxed clock model with uncorrelated rate change (in BEAST) applied to the 53-taxon Acer/Dipteronia dataset usually placed the root between Acer (posterior probability [PP] 0.95) and Dipteronia (PP 0.51), rendering each monophyletic (Fig. 3). The minority of topologies showed the root between the two Dipteronia species (with D. dyeriana usually sister to all other species), rendering Dipteronia paraphyletic. A strict clock model applied to the same data also placed the root between D. dyeriana and all other species.
To assess the decisiveness of the data, we compared 37 alternative rootings using Bayes factors. The results (Table 1) show that most root positions with prior probabilities of zero either were not significantly worse than the preferred root position (when using the upper bound prior probability values; see Materials and Methods) or received strong support in their favor (when using the lower bound prior probability values). Because Bayes factors measure the change in support for one hypothesis versus the other coming from the data, they can be very high when prior probabilities are essentially zero. Thus, a root position at branch 5, between a clade including all Acer species except A. glabrum and A. pseudoplatanus and the remaining taxa, changes from a prior probability of essentially zero to a posterior propability of 0.011, resulting in a large BF in its favor (Table 1). In the following, we only consider root positions that received posterior probabilities at or above an arbitrary threshold of 0.01 (branches 2 to 5 and 21 in Table 1 and Fig. 3). Among those, root positions at branches 3 (to D. sinensis) and 21 (to A. tataricum subsp. ginnala) are rejected in favor of the preferred root position by BFs > 10, whereas roots at branch 2 (to D. dyeriana) or 4 (to the clade of A. glabrum and A. pseudoplatanus) are not (Table 1, Fig. 3). When the root position at a given branch is compared with that at any other branch (Table 1, column 5), BFs again "overrate" positions 1, 2, 4, and 5, which receive significant support in their favor (BF > 10) because their prior probabilities are essentially zero.
|
Substitution rate heterogeneity along the tree is substantial as judged by the high coefficient of variation (mean 0.234; 95% highest posterior density [HPD] 0.111 to 0.362). Also, a model of autocorrelated rate change, as assumed by penalized likelihood and multidivtime relaxed clock models, is rejected (covariance mean 0.007; 95% HPD –0.186 to 0.202). However, testing for autocorrelation in BEAST may be problematic because the current implementation does not allow the two branches chosen at random to have their rates switched to have identical rates, which biases against finding positive autocorrelation (Rannala and Yang, 2007). However, because the 95% HPD interval found here safely includes zero, the inferred lack of autocorrelation appears robust.
Divergence Times and Shifts in Diversification Rates
Table 2 lists the estimates for when the North American species of Acer split from their closest Asian relatives. In spite of poor model fit, divergence time estimates obtained with all three relaxed clock approaches agree with each other (Table 2). Figure 3 shows a BEAST chronogram with the time estimates and their errors. Increasingly wide error bars towards the older parts of the BEAST chronogram indicate the lack of phylogenetic signal in deeper parts of the tree. This is supported by the bipartition network (Fig. 4), which depicts the distribution of conflicting signal in the data. Competing splits are confined almost entirely to the central part of the network (the grey-shaded area in Fig. 4), where they cause very low bootstrap support values (generally below 25%); in contrast, the mid-level and near-tip clades A through H are relatively well circumscribed and contain few contradictory splits.
|
|
Acer circinatum, from the northwestern United States, apparently diverged from its Asian relatives (in the Palmata clade) in the Early to Mid Miocene, A. macrophyllum diverged from its Asian sisters A. caesium and A. pilosum already in the Late Eocene, and the third western North American species, A. glabrum, separated from A. pseudoplatanus in the Late Eocene or Oligocene (Table 2). The northeastern American A. spicatum appears to have split from the Japanese A. carpinifolium in the Oligocene, whereas A. pensylvanicum split from A. crataegifolium and A. tegmentosum in the Late Miocene. The estimates for the crown of the Acer core clade, which includes the North American A. saccharum, range from Eocene and Oligocene to the Miocene (Table 2), whereas A. rubrum and A. saccharinum (both widespread throughout the eastern United States) appear to have split from their Asian relatives in Late Oligocene to Early Miocene and from each other in the Pliocene. The ninth North American species, A. negundo, is an isolated, apparently early-divergent species (Mid to Late Eocene), as are the Japanese A. distylum (sister to the remainder of clade G in Figs. 2 and 3), A. nipponicum, the Eurasian A. tataricum, and the two species of Dipteronia from China. The most tropical species of Acer, A. laurinum, which is widespread in Hainan, Vietnam, Sumatra, Borneo, Celebes, Java, and Indonesia's Lesser Sunda Islands, belongs to a clade that is at least 30 Myr old (Table 2).
The LTT plot (Fig. 5) derived from an optimal chronogram suggests that diversification (speciation minus extinction) in Acer began leveling off between 30 and 20 Myr. The results of censored survival analyses under three models of diversification varied with the way the missing 70 species were added to the LTT plot (Materials and Methods). Models of constant diversification were favored when about half the missing species were allocated to the period between 35 and 20 Myr, whereas models of a gradual decrease in diversification were favored when the bulk of missing species was allocated to younger periods. In other words, when the censored species were moved towards the present, the decrease in diversification through time is less steep (as expected). These results suggest that the inferences are robust to this amount of missing data.
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentReferences |
|---|
Tree Rooting with a Relaxed Clock
With some 6500 base pairs of chloroplast sequence data, outgroup rooting, and a species sample that spanned the root of each genus, Dipteronia and Acer are mutually monophyletic, although this has weak statistical support. Midpoint rooting also places the root between Dipteronia and Acer. Accepting the sister group relationship between Dipteronia and Acer as true, a relaxed clock model with uncorrelated rate change between ancestors and decedents correctly specified this root without an outgroup, albeit also without significant support (Table 1, Fig. 3). To our knowledge, this might be the first exploration of the use of clock analyses in BEAST for rooting purposes. Explicit tests of alternative root positions via Bayes factors revealed that a rooting between Dipteronia and Acer is better supported than most alternative rootings, although it is not significantly better than root positions 2, 4, and 5 (Table 1, Fig. 3). Bayes factors crucially depend on good estimates of both prior and posterior probabilities. More intensive sampling of tree space to obtain stronger prior probabilities might have improved the reliability of the BFs calculated here, but the possible resulting changes are unlikely to affect our main conclusions because we only considered root placements with posterior probabilities
0.01. Regardless of the specifics of the Acer/Dipteronia system, the possibility of using uncorrelated rates relaxed clock models to root datasets for which outgroup rooting is problematic (or not possible) presents an exciting application of clock analyses. This was presaged by the simulation study of Huelsenbeck et al. (2002), who compared outgroup rooting, molecular clock–based rooting, and nonreversible models of DNA substitution and found that the molecular clock criterion correctly identified the root a large proportion of the time. Clock-based rooting, however, was not possible in practice before the development of an approach that can handle large (real-life) datasets. With the availability of uncorrelated rates models of molecular evolution (Drummond et al., 2006), one can now use clock-based rooting to explore whether outgroup rooting might be biased by long-branch attraction or simply as an independent test of correct rooting (as done here).
In the 53-taxon-6847-bp Acer/Dipteronia data, the correct root (assuming that it lies between Dipteronia and Acer) was found only under a relaxed clock, while under a strict clock (in BEAST), the root was placed between D. dyeriana and all other species. The separation of D. dyeriana from the remaining species might be explained by its higher substitution rate compared to D. sinensis and the majority of Acer species (Fig. 3). Although a strict clock model performed well in most of the experimental datasets analyzed by Huelsenbeck et al. (2002), misleading results could be obtained if a crucial taxon close to the root had a substitution rate substantially different from taxa nearby.
Shifts in Diversification Rates and Timing of the Beringian Disjunctions
Inspection of the lineage-through-time plot for Acer and Dipteronia suggests a gradual decrease in the diversification rate sometime between 30 and 20 Myr (Fig. 5), but the ratio of some 50 sequenced species to 70 (or more) not-sequenced species cautions against attaching too much weight to the LTT plot. A worldwide analysis of maple fossils revealed a striking decrease in occurrences of Acer megafossils between 35 and 25 Myr in Europe, West Asia, and East Asia, but not in North America (Boulter et al., 1996). These apparent waves of extinctions were followed by new radiations in Europe, East Asia, and Western Siberia (Boulter et al. 1996; Kva
ek, 1996). Doubtless, both Acer and Dipteronia have suffered substantial extinction since the Early Oligocene (33 to 32 Ma), when Dipteronia was still present in Oregon (McClain and Manchester, 2001). The number of combinations of extinction and speciation rates that can result in the same net diversification is infinite, and LTT plots are therefore poor tools for distinguishing these two processes. Nevertheless, changes in diversification can be inferred as long as early lineages are reasonably well sampled (Harvey et al., 1994; Pybus and Harvey, 2000; Nee, 2006; Ricklefs, 2007), which we believe is the case for Acer. Overall, we sampled 74 of the 124 to 156 species (Fig. 2), of which the chronogram includes 53 selected to span the root.
Divergence time estimation under a model of uncorrelated rate change, by not relying on a resolved input topology, may be relatively robust against unresolved relationships in the deeper part of a phylogeny (as is the case here), and even clock approaches that assume autocorrelation are surprisingly little affected by polytomies or zero-length branches. Thus, Won and Renner (2006) analyzed a gymnosperm dataset, using either different resolutions of a polytomy or maintaining the polytomy in the input tree, and found that time estimates were essentially unaffected. Apparently, polytomies impact posterior branch lengths only where branch length uncertainty occurs throughout the tree, which clearly is not the case in the Acer data. In the present study, Bayesian relaxed clocks yielded similar time estimates, whether they relied on an input topology and autocorrelated rates or not (Table 2). In general, however, autocorrelated models outperform uncorrelated models when age estimation is the goal (Lepage et al., 2007).
The following inferences about the Beringian disjunctions in Acer concern only nodes towards the tips of the tree that have statistical support; for absolute times we rely on the estimates obtained with the two Bayesian relaxed clock approaches, calibrated with a Dipteronia fossil (Table 2). The pre-Oligocene burst of Acer radiation (Figs. 3 and 5) inferred in this study fits with the abundant Eocene and Oligocene record of the genus (Wolfe and Tanai, 1987; Boulter et al., 1996; Manchester, 1999; McClain and Manchester, 2001; Kittle et al., 2005). Much of this diversity, especially in Europe and North America, appears to have gone extinct during the Late Oligocene–Early Miocene (Walther, 1972; Wolfe and Tanai, 1987; Boulter et al., 1996). Nevertheless, living A. macrophyllum may descend from a lineage with fruits and foliage resembling the extinct A. osmontii, which occurred in Oregon during the Early Oligocene at ca. 33 Ma. Similarly, the living European species A. heldreichii (incl. A. trautvetteri) and A. pseudoplatanus have leaf traits that closely resemble those of A. haselbachense, which lived in Central Europe some 35 Myr (Walther, 1972; Grimm et al., 2007). A difficulty with assigning fossils to living clades of Acer, however, is that similar leaf and fruit characters have evolved repeatedly. For example, leaves of North American A. spicatum closely resemble those of A. caudatum subsp. ukurunduense, and leaves of North American A. negundo resemble those of Japanese A. cissifolium and Chinese A. henryi (van Gelderen et al., 1994: 64), yet these species are not closely related (Fig. 2; further examples are illustrated in Grimm et al., 2007: fig. 7). This uncertainty is exemplified by the recent reassignment of most "Acer arcticum" leaves from the North American Paleocene and Eocene (and also Eocene floras of northeastern China) to Deviacer wolfei, whereas "Acer arcticum" fruits have been reassigned to Polygalaceae (Manchester, 1994; Pigg et al., 2004).
In spite of large errors around the molecular divergence times it is clear that the extant North American species of Acer diverged from their respective Asian sister clades at widely different times (Table 2). Wolfe (1981), based on an extensive knowledge of Acer macrofossils, arrived at the same conclusion. He thought that Acer originated in (western) North America (his fig. 10.1) and that some lineages spread to eastern Asia in the Early Eocene, others from Asia to North America in the Early Miocene (the A. pensylvanicum–containing section Macrantha) or Late Pliocene (A. circinatum, section Palmata). Wolfe also thought that the Negundo and Rubra lines (which in his view did not include A. laurinum) originated in North America during the Middle Eocene and that the A. macrophyllum line originated in western North America in the Late Eocene. The rough coincidence with several of our divergence time estimates (Table 2) is striking since we did not use any internal constraints on node times. One of Wolfe's main conclusions was that "the distribution patterns of fossil Acer do not support the hypothesis of an Arcto-Tertiary Geoflora," because many of the vicariance events occur later than the abrupt climate cooling at the end of the Eocene. Nor do they support the concept of fragmentation of one biota. "Rather, the history of Acer is replete with repeated dispersals and vicariant events through a period of about 60 million years" (Wolfe, 1981:422).
Based on the average of the estimated divergence times, North American taxa split from their Asian sister taxa at a rate of about once per 5 Ma beginning about 40 Myr. What is remarkable about diversification in Acer is that except for the recent divergence of A. rubrum and A. saccharinum, the North American lineages have not formed new species, at least among those that are still extant. This is consistent with the discrepancy in diversity favoring eastern Asia over eastern North America in the number of species in disjunct genera (Qian and Ricklefs, 2000, 2004) and analyses that found diversification of sister lineages to have been more rapid in Asia than in North America (Xiang et al., 2004).
Our PAML and BEAST analyses of the 98-taxon Sapindales rbcL dataset rejected a molecular clock but yielded no evidence of a strong rate change in the common ancestor of Acer and Dipteronia. A PAML-estimated non-clock model had a likelihood score of –12,476.6, a single rate clock model a score of –12,858.4, and a two-rate local clock model with one rate for Acer/Dipteronia, the other for the remaining Sapindales had a score of –12,858.0, not significantly different from the single-rate model. When the 98-taxon set was analyzed in BEAST with minimal age constraints only on the Aesculus and Dipteronia crown groups, the Acer crown was dated to 30 (18–44) Ma and the divergence of Acer and Dipteronia to 34 (20–48) Ma. Even younger ages were obtained when neither Aesculus nor Dipteronia were constrained (in runs relying on other Sapindales fossils); the crown-group age of Acer/Dipteronia was then less than 20 Myr old (data not shown). These age estimates are in the range of those obtained by Wikström et al. (2001), who in an angiosperm-wide chronogram dated the split between Acer and Aesculus to 20 to 26 Ma (using a single calibration point far from Sapindaceae, namely the divergence between Fagales and Cucurbitales). Such young ages for the Acer/Dipteronia split are strongly contradicted by the Paleocene fossil records of both genera. One explanation for the discrepancy might lie in the saturation of conserved genes, such as the Rubisco enzyme (rbcL), which is under strong selection (Kapralov and Filatov, 2007), suggesting that molecular dating might fare better with a mix of coding and non-coding gene regions. Indeed, some of the worst artifacts in angiosperm-wide time estimation occurred when non-synonymous substitutions were used, and the bias was often in opposite directions (Sanderson and Doyle, 2001: fig. 12). Alternatively, strongly delayed sorting of ancestral chloroplast lineages in Acer might explain some of the relatively young molecular estimates. Testing this possibility would require a comparison with time estimates from nuclear data.
|
Acknowledgment |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentReferences |
|---|
We thank M. Sanderson and V. Savolainen for constructive suggestions; M. Suchard for clarifications concerning the Bayes factor calculations; S. Ho for initial help with BEAST; and Natalie Cusimano for an introduction to the APE package in R. The study benefited from discussions during a workshop on Northern Hemisphere biogeography supported by the National Evolutionary Synthesis Center (NESCent), NSF grant EF-0423641, and from financial support by the Austrian Science Fund (FWF; P14825 to T.F.S.).
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentReferences |
|---|
-
Aris-Brosou S., Yang Z. Effects of models of rate evolution on estimation of divergence dates with special reference to the metazoan 18S rRNA phylogeny. Syst. Biol. (2002) 703–714.
Beiko R. G., Keith J. M., Harlow T. J., Ragan M. A. Searching for convergence in phylogenetic Markov chain Monte Carlo. Syst. Biol. (2006) 55:553–565.
Boulter M. C., Benfield J. N., Fisher H. C., Gee D. A., Lhotak M. The evolution and global migration of the Aceraceae. Phil. Trans. R. Soc. Lond B (1996) 351:589–603.
Brenner G. J. Evidence for the earliest stage of angiosperm pollen evolution: A paleoequatorial section from Israel. In: Flowering plant origin, evolution and phylogen—Taylor D. W., Hickey L. J., eds. (1996) New York: Chapman and Hall. 91–115.
Britton T., Anderson C. L., Jacquet D., Lundqvist S., Bremer K. Estimating divergence times in large phylogenetic trees. Syst. Biol. (2007) 56:741–752.
Brown J. M., Lemmon A. R. The importance of data partitioning and the utility of Bayes factors in Bayesian phylogenetics. Syst. Biol. (2007) 56:643–655.
Cho H.-J., Kim S., Suh Y., Park C.-W. ITS sequences of some Acerspecies and phylogenetic implication. Korean J. Plant Taxon. (1996) 26:271–291.
Corbett S. L., Manchester S. R. Phytogeography and fossil history of Ailanthus(Simaroubaceae). Int. J. Plant Sci. (2004) 165:671–690.[CrossRef][Web of Science]
Crane P. R., Manchester S. R., Dilcher D. L. A preliminary study of fossil leaves and well-preserved reproductive structures from the Sentinel Butte Formation (Paleocene) near Almont, North Dakota. Fieldiana Geology, new series, (1990) 20:1–63.
DeVore M. L., P. Kenrick, K. B. Pigg, R. A. Ketcham. CT-scanning the London clay: An excellent noninvasive technique for studying pyritized fossil fruits (2005) Abstract 122, Botany 2005.
Drummond A. J., Ho S. Y. W., Phillips M. J., Rambaut A. Relaxed phylogenetics and dating with confidence. PLoS Biol. (2006) 4.
Drummond A. J., A. Rambaut. BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol. Biol. (2007) 7:214. (doi:10.1186/1471-2148-7-214).[CrossRef][Medline]
Farris J. S. Estimating phylogenetic trees from distance matrices. Am. Nat. (1972) 106:645–668.[CrossRef][Web of Science]
Gelderen D. M., van P.C. de Jong, H.J. Oterdoom. Maples of the world (1994) Portland, Oregon: Timber Press.
Gradstein F.M., J.G. Ogg, A.G. Smith. A geologic time scale (2004) Cambridge, UK: Cambridge University Press.
Grimm G. W., Denk T., Hemleben V. Evolutionary history and systematics of Acersection Acer—A case study of low-level phylogenetics. Plant Syst. Evol. (2007) 267:215–253.[CrossRef]
Grimm G. W., Renner S. S., Stamatakis A., Hemleben V. A nuclear ribosomal DNA phylogeny of Acerinferred with maximum likelihood, splits graphs, and motif analysis of 606 sequences. Evol. Bioinformatics (2006) 2:7–22.
Harrington M. G., Edwards K. J., Johnson S. A., Chase M. W., Gadek P. A. Phylogenetic inference in Sapindaceae sensu lato using plastid matK and rbcL DNA sequences. Syst. Bot. (2005) 30:366–382.[CrossRef]
Harvey P. H., May R. M., Nee S. Phylogenies without fossils. Evolution (1994) 48:523–529.[CrossRef][Web of Science]
Huelsenbeck J. P. Measuring the stratigraphic record to estimates of phylogeny. Paleobiology (1994) 20:470–483.[Abstract]
Huelsenbeck J. P., Bollback J. P., Levine A. M. Inferring the root of a phylogenetic tree. Syst. Biol. (2002) 51:32–43.
Hughes N.F. The enigma of angiosperm origins (1994) Cambridge, UK: Cambridge Palaeobiology Series 1, Cambridge University Press.
Huson D., Bryant D. Application of phylogenetic networks in evolutionary studies. Mol. Biol. Evol. (2006) 23:254–267.
Iwabe N., Kuma K. I., Hasegawa M., Osawa S., Miyata T. Evolutionary relationship of archaebacteria, eubacteria, and eukaryotes inferred from phylogenetic trees of duplicated genes. Proc. Natl. Acad. Sci. USA (1989) 86:9355–9359.
de. Jong P. C. Taxonomy and reproductive biology of maples. In: Maples of the worl—van Gelderen D. M., de Jong P. C., Oterdoom H. J., eds. (1994) Portland, Oregon: Timber Press. 69–104.
Jong P. C. de. Worldwide maple diversity. In: Proceedings of the International Maple Symposiu—Wiegrefe S. J., Angus H., Otis D., Gregorey P., eds. (2002) Gloucestershire, England: Westonbirt Arboretum and the Royal Agricultural College. 2–11.
Kapralov M. V., Filatov D. A. Widespread positive selection in the photosynthetic Rubisco enzyme. BMC Evol. Biol. (2007) 7:73.[CrossRef][Medline]
Kass R. E., Raftery A. E. Bayes factors. J. Am. Stat. Assoc. (1995) 90:773–795.[CrossRef][Web of Science]
Kittle A., DeVore M. L., Wall B., Pigg K. Acerfruits from the Paleocene of North Dakota. Georgia J. Sci. (2005) 63:32.
Kvaek Z. Are the Turgayan floras homogeneous. In: Memorial Conference dedicated to Vsevolod Andreevich Vakhrameev, Abstracts and Proceeding—Akhmetiev M. A., Doludenko M. P., eds. (1996) Moscow: Russ. Acad. Sci. Geol. Inst. GEOS Press. 29–33.
Lake J. A., Herbold C. W., Rivera M. C., Servin J. A., Skophammer R. G. Rooting the tree of life using nonubiquitous genes. Mol. Biol. Evol. (2007) 24:130–136.
Lepage T., Bryant D., Philippe H., Lartillot N. A general comparison of relaxed molecular clock models. Mol. Biol. Evol. (2007) 24:2669–2680.
Li J., Yue J., Shoup S. Phylogenetics of Acer(Aceroideae, Sapindaceae) based on nucleotide sequences of two chloroplast non-coding regions. Harvard Pap. Bot. (2006) 11:101–115.[CrossRef]
Maddison D.R., W.P. Maddison. MacClade 4.0. Sinauer Associates, Sunderland. (2003) Massachusetts.
Maddison W. P., Donoghue M. J., Maddison D. R. Outgroup analysis and parsimony. Syst. Zool. (1984) 33:83–103.
Manchester S. R. Fruits and seeds of the Middle Eocene Nut Beds Flora, Clarno Formation, Oregon. Palaeontogr. Am. (1994) 58:1–205.
Manchester S. R. Biogeographical relationships of North American Tertiary floras. Ann. Missouri Bot. Gard. (1999) 86:472–522.[CrossRef][Web of Science]
Manchester S. R. Leaves and fruits of Aesculus(Sapindales) from the Paleocene of North America. Int. J. Plant Sci. (2001) 162:985–998.[CrossRef][Web of Science]
Mathews S., Donoghue M. J. The root of angiosperm phylogeny inferred from duplicate phytochrome genes. Science (1999) 286:947–950.
McClain A. M., Manchester S. R. Dipteronia(Sapindaceae) from the Tertiary of North America and implications for the phytogeographic history of the Aceroideae. Am. J. Bot. (2001) 88:1316–1325.
Muellner A. N., Vassiliades D. D., Renner S. S. Placing Biebersteiniaceae, a herbaceous clade of Sapindales, in a temporal and geographic context. Plant Syst. Evol. (2007) 266:233–252.[CrossRef]
Müller K. PRAP—Computation of Bremer support for large data sets. Mol. Phylogenet. Evol. (2004) 31:780–782.[CrossRef][Web of Science][Medline]
Nee S. Birth-death models in macroevolution. Ann. Rev. Ecol. Syst. (2006) 37:1–17.[CrossRef][Web of Science]
Paradis E. Assessing temporal variations in diversification rates from phylogenies: Estimation and hypothesis testing. Proc. R. Soc. Lond. B (1997) 264:1141–1147.[CrossRef]
Paradis E., Claude J., Strimmer K. APE: Analyses of phylogenetics and evolution in R language. Bioinformatics (2004) 20:289–290.
Pfosser M. F., Guzy-Wróbelska J., Sun B.-Y., Stuessy T. F., Sugawara T., Fujii N. The origin of species of Acer(Sapindaceae) endemic to Ullung Island, Korea. Syst. Bot. (2002) 27:351–367.
Pigg K.B., M.F. Wojciechowski, M.L. DeVore. Samaras from the Late Paleocene Almont and Beicegel Creek floras of North Dakota, U.S.A. with potential affinities to Securidaca(Polygalaceae) (2007) Abstracts. Botany 2004.
Posada D., Crandall K. A. ModelTest: Testing the model of DNA substitution. Bioinformatics (1998) 14:817–818.
Pybus O. G., Harvey P. H. Testing macro-evolutionary models using incomplete molecular phylogenies. Proc. R. Soc. Lond. B (2000) 267:2267–2272.[Medline]
Qian H., Ricklefs R. E. Large-scale processes and the Asian bias in species diversity of temperate plants. Nature (2000) 407:180–182.[CrossRef][Medline]
Qian H., Ricklefs R. E. Geographical distribution and ecological conservatism of disjunct genera of vascular plants in eastern Asia and eastern North America. J. Ecol. (2004) 92:253–265.[CrossRef]
Qiu Y.-X., Luo Y.-P., Comes H. P., Ouyang Z.-Q., Fu C.-X. Population genetic diversity and structure of Dipteronia dyerana(Sapindaceae), a rare endemic from Yunnan Province, China, with implications for conservation. Taxon (2007) 56:427–437.[Web of Science]
Rambaut A., Bromham L. Estimating divergence dates from molecular sequences. Mol. Biol. Evol. (1998) 15:442–448.[Abstract]
Rannala B., Yang Z. Inferring speciation times under an episodic molecular clock. Syst. Biol. (2007) 56:453–466.
Renner S.S., L. Beenken, G.W. Grimm, A. Kocyan, R.E. Ricklefs. The evolution of dioecy, heterodichogamy, and labile sex expression. Acer. Evolution (2007) 61:2701–2719.[CrossRef]
Ricklefs R. E. Estimating diversification rates from phylogenetic information. Trends Ecol. Evol. (2007) 22:601–610.[CrossRef][Medline]
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, version 1.70. User's manual. (2004) http://ginger.ucdavis.edu/r8s/.
Sanderson M. J., Doyle J. A. Sources of error and confidence intervals in estimating the age of the angiosperms from rbcLand 18S rDNA data. Am. J. Bot. (2001) 88:1499–1516.
Stamatakis A. RAxML-VI-HPC: Maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics (2006) 22:2688–2690.
Suchard M. A., Weiss R. E., Sinsheimer J. S. Bayesian selection of continuous-time Markov chain evolutionary models. Mol. Biol. Evol. (2001) 18:1001–1013.
Suchard M. A., Weiss R. E., Sinsheimer J. S. Models for estimating Bayes factors with applications to phylogeny and test of monophyly. Biometrics (2005) 61:665–673.[CrossRef][Web of Science][Medline]
Suh Y., Heo K., Park C.-W. Phylogenetic relationships of maples (AcerL.; Aceraceae) implied by nuclear ribosomal ITS sequences. J. Plant Res. (2000) 113:193–202.[CrossRef][Web of Science]
Sullivan J., Swofford D. L. Should we use model-based methods for phylogenetic inference when we know that assumptions about among-site rate variation and nucleotide substitution pattern are violated? Syst. Biol. (2001) 50:723–729.
Swofford D.L. PAUP*. Phylogenetic analysis using parsimony (*and other methods) (2002) Sunderland, Massachusetts: Version 4. Sinauer Associates.
Swofford D.L., D.P. Begle. PAUP Version 3.1 User's manual (1993) Washington, DC: Smithsonian Institution. (freely available online).
Swofford D. L., Olsen G., Waddell P., Hillis D. M. Phylogenetic inference. In: Molecular systematics, 2nd editio—Hillis D., Moritz C., Mable B., eds. (1996) Sunderland, Massachusetts: Sinauer Associates. 407–511.
Thorne J. L., Kishino H., Painter I. S. Estimating the rate of evolution of the rate of molecular evolution. Mol. Biol. Evol. (1998) 15:1647–1657.[Abstract]
Thorne J. L., Kishino H. Divergence time and evolutionary rate estimation with multilocus data. Syst. Biol. (2002) 51:689–702.
Tian X., Guo Z.-H., Li D.-Z. Phylogeny of Aceraceae based on ITS and trnL-F data sets. Acta Bot. Sin. (2002) 44:714–724.
Walther H. Studien über tertiäre AcerMitteleuropas. Abh. Staatl. Mus. Mineral. Geol. Dresden (1972) 19:1–309.
Wheeler W. C. Nucleic acid sequence phylogeny and random outgroups. Cladistics (1990) 6:363–368.[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]
Wolfe J.A. Vicariance biogeography of angiosperms in relation to paleobotanical data. Pages 413–427, in Vicariance biogeography—G. Nelson, D.E. Rosen, eds. (1981) New York: Columbia University Press.
Wolfe J. A., Tanai T. Systematics, phylogeny, and distribution of Acerin the Cenozoic of western North America. J. Fac. Sci. Hokkaido Univ. IV: Geol. Mineral. (1987) 22:1–246.
Won H., Renner S. S. Dating dispersal and radiation in the gymnosperm Gnetum (Gnetales) – clock calibration when outgroup relationships are uncertain. Syst. Biol. (2006) 55:610–622.
Xiang Q. Y. J., Zhang W. H., Ricklefs R. E., Qian H., Chen Z. D., Wen J., Li J. H. Regional differences in rates of plant speciation and molecular evolution: A comparison between eastern Asia and eastern North America. Evolution (2004) 58:2175–2184.[CrossRef][Web of Science][Medline]
Yang Z. PAML: A program package for phylogenetic analysis by maximum likelihood. Comp. Appl. BioSci. (1997) 13:555–556.[Medline]
Yap V. B., Speed T. Rooting a phylogenetic tree with nonreversible substitution models. BMC Evol. Biol. (2005) 5:2.[CrossRef][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||