OUP user menu

Identifying the confounding factors in resolving phylogenetic relationships in Vespertilionidae

Justin B. Lack, Ronald A. Van Den Bussche
DOI: http://dx.doi.org/10.1644/09-MAMM-A-354.1 1435-1448 First published online: 16 December 2010

Abstract

Resolving phylogenetic relationships within Vespertilionidae has been difficult, with large data sets (>100 taxa, >7 kilobases) resolving portions of the phylogeny but leaving intertribal relationships within the Vespertilioninae unresolved. As a result the evolutionary history of the most speciose chiropteran family is largely unknown. The presence of short internodes followed by long terminal branches relative to other chiropteran phylogenies suggests that evolutionary rates of DNA substitution and lineage diversification could be inhibiting phylogenetic resolution. To test this hypothesis we obtained sequences of the mitochondrial DNA (mtDNA) 12s rRNA, tRNAVAL, and 16s rRNA, and the nuclear exon RAG2, resulting in more than 3 kilobases of digenomic DNA sequence data for representatives of all subfamilies and tribes within Vespertilionidae and Phyllostomidae, a family of bats that radiated at approximately the same time as Vespertilionidae. Analyses revealed that substitution rates for Vespertilionidae were significantly higher than those for Phyllostomidae, with the majority of fast-evolving lineages found within Vespertilioninae. Cladogenesis analyses characterized the vespertilionid radiation as compressed toward the root, with a rapid initial diversification, but the phyllostomid diversification was much more gradual. We suggest that ecological differences between tropical and temperate environments could have influenced diversification rates for Vespertilionidae and Phyllostomidae.

Key words
  • adaptive radiation
  • evolutionary rates
  • Phyllostomidae
  • Vespertilionidae

The chiropteran family Vespertilionidae is the 2nd most speciose mammalian family, with approximately 407 species and an essentially worldwide distribution (Simmons 2005). These bats are almost exclusively insectivorous and exhibit little morphological variation even among the most distantly related members (Corbet and Hill 1991; Nowak 1999). The lack of morphological variation has created difficulties resolving relationships within Vespertilionidae based on morphology (Jones et al. 2002), with significant conflict among data sets consisting of different combinations of morphological characters. This has led to a dependence upon molecular data sets (i.e., DNA sequences) to determine evolutionary relationships, often revealing significant convergence (Ruedi and Mayer 2001) and conflict with morphological phylogenies (Giannini and Simmons 2005).

The most comprehensive vespertilionid phylogenetic study to date was that of Hoofer and Van Den Bussche (2003), in which approximately 2,700 base pairs (bp) of mitochondrial DNA (mtDNA) sequence data were generated for 110 vespertilionid species representing 41 of 48 genera (since corrected for taxonomic changes—Simmons 2005). Although this study was extensive in terms of taxonomic sampling and DNA sequence data generated, only about 70% of the nodes within Vespertilionidae were supported statistically by a posterior probability of 0.95 or greater in the Bayesian analysis. The majority of unsupported relationships occurred within or at the base of the subfamily Vespertilioninae. Of these unsupported nodes, the majority were among tribes. One particularly problematic group within Vespertilioninae has been the tribe Plecotini. Both an initial mtDNA systematic approach (Hoofer and Van Den Bussche 2003) and an expanded digenomic data set (Roehrs et al. 2010) failed to obtain statistical support for this traditionally recognized tribe. Outside of Vespertilioninae, almost all nodes within Vespertilionidae were resolved. Not only was the lack of resolution confined to a particular group, but it also was confined to a particular snapshot in time. Roehrs et al. (2010) generated an additional 4,000 base pairs of nuclear intron and exon sequence data for representatives of all subfamilies and tribes within Vespertilionidae. Although this expanded data set resolved many of the remaining unresolved nodes above and below the tribal level, at the tribal level essentially no additional resolution was acquired. With this pattern having been established, the issue becomes identifying the underlying reason(s) for the lack of resolution.

The 1st, and probably the most easily arrived at, explanation is that the markers used in the previous studies (Hoofer and Van Den Bussche 2003; Roehrs et al. 2010) were inappropriate at the unresolved taxonomic level. However, the same mtDNA segment used by Hoofer and Van Den Bussche (2003) was used in combination with one of the nuclear exons (RAG2) used by Roehrs et al. (2010) to examine relationships among the New World leaf-nosed bats (Phyllostomidae— Baker et al. 2003). For phyllostomids, resolution was obtained at 48 of 55 nodes with no clear pattern of insufficient resolution at any taxonomic level (Baker et al. 2003). In addition, by conducting a more thorough Bayesian analysis, partitioning by molecular marker and then by codon for RAG2, resolution was achieved at 4 additional nodes (described in “Materials and Methods”; Fig. 1). Another possibility could be that taxonomic sampling was insufficient. However, both Baker et al. (2003) and Hoofer and Van Den Bussche (2003) similarly sampled nearly every genus occurring in each respective family and included multiple representatives of all previously hypothesized tribes.

Fig. 1

Bayesian phylogram resulting from combined analysis of mitochondrial DNA (16s rRNA, tRNAVAL, and 12s rRNA) and nuclear (RAG2) sequences under the GTR + I + Γ model of nucleotide substitution. Nodes indicated by # were not statistically supported (posterior probability [PP] < 0.95); all other nodes were statistically supported (PP ≥ 0.95).

One situation that can lead to a lack of resolution is among-lineage heterogeneity in evolutionary rate (Felsenstein 1978; Hillis et al. 1994; Li 1997). Elevated lineage-or clade-specific molecular rates can produce high levels of systematic error that lead to incorrect but statistically supported phylogenetic relationships (Rodriguez-Ezpeleta et al. 2007). Difficulty resolving basal relationships for Metazoa (Baurain et al. 2007; Brinkmann and Phillipe 2008) and superfamilial relationships for Rodentia (Montgelard et al. 2008) have been attributed to systematic error. However, recent maximum-likelihood and Bayesian inference methods (and more specifically, their flexible evolutionary models) and a more dense taxonomic sampling to break up long branches can improve significantly the ability of phylogenetic analyses to overcome this issue. Conversely, a reduced evolutionary rate is a problem less easily overcome. In this situation internodes on a given phylogeny will be relatively short and the number of apomorphic characters defining a clade will be few. In these data sets the number of phylogenetically informative characters per lineage can be drastically reduced or unevenly distributed across the topology. To overcome this problem acquisition of more sequence data is most likely the solution, although in large-scale phylogenies it has been remedied by removal of the slowly evolving lineage and replacing it with a more rapidly evolving lineage to represent the group (Baurain et al. 2007). However, when resolving relationships at the species or genus level, this is often not an option.

Another potential issue in obtaining phylogenetic resolution is the occurrence of rapid species radiations. Rapid species radiations have occurred in many taxa, including African cichlids, Hawaiian silverswords, and Caribbean Anolis (Givnish and Sytsma 1997; Schluter 2000), and have been implicated as a possible explanation for difficulty in resolving some phylogenetic relationships (Kraus and Miyamoto 1991; Rokas et al. 2005), but their role often has been difficult to quantify (Bokma 2002). A rapid accumulation of lineages would lead to extremely short internodes on a phylogeny. If internodes are short and effective population sizes between subsequent divergences are sufficiently large, incomplete lineage sorting can occur, resulting in ancestral polymorphisms being retained in gene trees (Avise et al. 1983; Maddison 1997; Neigel and Avise 1986; Rosenberg 2002). In this situation gene tree topologies conflicting with the species tree topology are more likely than gene tree topologies that match the species tree topology (Degnan and Rosenberg 2006). Most importantly, adding additional data is not a likely remedy (Edwards 2008; Kubatko and Degnan 2007). To illustrate the point rapid cladogenesis has been identified as a possible source of problems in resolving relationships at the base of the metazoan radiation. Rokas et al. (2005) generated sequence data for all metazoan phyla from 50 nuclear markers with more than 6,500 variable amino acid positions and were still unable to gain additional phylogenetic resolution.

Our primary objective was to determine the source of phylogenetic uncertainty in evolutionary relationships within Vespertilionidae. Phylograms produced by Hoofer and Van Den Bussche (2003) and Roehrs et al. (2010) show signs of both rate disparity and rapid cladogenesis. Branch lengths for clades at similar taxonomic levels were highly variable, and internodes at the base of Vespertilioninae and leading to the various tribes appeared to be very short. Therefore, we propose to use phylogenetic methods to determine if among-lineage rate variation or rapid cladogenesis has played a significant role in the evolution of vespertilionid bats. In addition, if either of these processes contributed significantly, we attempt to determine the historical processes that could have brought on the condition(s). Phylogenetics can be a powerful tool in examining evolutionary histories, potentially filling in gaps left by an incomplete fossil record and the frequently misleading phenotypic traits of extant taxa. In addition, we use molecular dating techniques to place a timescale on the evolution of vespertilionid and phyllostomid bats.

Materials and Methods

We obtained from GenBank previously published sequence data from multiple chiropteran systematic studies to incorporate both nuclear and mitochondrial sequence data and to sample Vespertilionidae and Phyllostomidae as thoroughly as possible. We used a continuous ∼2,700-bp mtDNA segment consisting of the 16s rRNA, tRNAVAL, and 12s rRNA, and exon 2 of the nuclear recombination activation gene (RAG2), to circumvent any genomic bias, because mitochondrial and nuclear genomes can have different evolutionary histories (Avise 1994). For taxonomic sampling of Vespertilionidae we included representatives of all hypothesized subfamilies and tribes, resulting in 107 taxa representing 35 of 48 genera currently recognized for the family (Simmons 2005). However, the estimate of 48 genera is inflated, because Miniopterus has been elevated to its own family (Miniopteridae—Hoofer and Van Den Bussche 2003), Histiotus has been religated to Eptesicus (Hoofer and Van Den Bussche 2003), Cistugo likely constitutes a distinct family (Lack et al. 2010), and several of the genera we did not include have had an inconsistent taxonomic past, often being lumped into other genera as frequently as they are considered distinct—that is, Scoteanax included in Nycticeius (Simmons 2005); Scotorepens included in Nycticeius (Simmons 2005); Glischropus included in Pipistrellus (Menu 1987); and Scotozous included in Pipistrellus (Bates and Harrison 1997; Koopman 1993, 1994). Even with missing genera, the portions of the phylogeny we are most concerned with (intertribal relationships) have no missing lineages in our data set, and any missing taxa will underestimate diversity at lower taxonomic scales.

As a reference group outside of Vespertilionidae, we sampled the 3rd most speciose family of chiropterans, Phyllostomidae (Simmons 2005), which has been referred to as the most morphologically diverse extant mammalian family (Monteiro and Nogueira 2010; Simmons 1998). The radiation of this family is of similar age as the vespertilionids (Miller-Butterworth et al. 2007; Teeling et al. 2005), and the same molecular markers described above have been used to resolve almost completely relationships at all levels within the family (Baker et al. 2000, 2003). We sampled 46 taxa representing all hypothesized subfamilies, including all but 9 of the currently recognized genera within Phyllostomidae (Simmons 2005). For Vespertilionidae mtDNA sequences were obtained from GenBank and had been published by Van Den Bussche and Hoofer (2000), Hoofer and Van Den Bussche (2003), and Roehrs et al. (2010). All vespertilionid RAG2 sequences were downloaded from GenBank and previously published by Hoofer et al. (2003), Stadelmann et al. (2007), and Roehrs et al. (2010). All phyllostomid mtDNA and RAG2 sequences were obtained from GenBank and had been published by Baker et al. (2000) or Baker et al. (2003). In addition to the 2 families mentioned above, we included representatives of all families included in the superfamily Vespertilionoidea (Vespertilionidae, Molossidae, Miniopteridae, and Natalidae) and all but 1 family (Mystacinidae) in the superfamily Noctilionoidea (Phyllostomidae, Noctilionidae, Furipteridae, Thyropteridae, Myzopodidae, and Mormoopidae) following the classification of Teeling et al. (2005). A representative of the family Emballonuridae (Saccopteryx) was included to serve as outgroup for all analyses. A complete list of all taxa included in this study is given in Appendix I.

Sequences were aligned independently using Clustal X software (Thompson et al. 1997), and the resulting multiple alignments were imported into MacClade version 4.08 (Maddison and Maddison 2000) for visual inspection and editing. The mtDNA fragment contained regions of questionable positional homology, and all of these positions were excluded from analyses. Edited alignments have been deposited in TreeBASE (submission 10593). A Bayesian phylogenetic analysis was conducted on both the total data set and a pruned data set of only Vespertilionidae, Phyllostomidae, and the outgroup Saccopteryx using MrBayes version 3.1.2 (Huelsenbeck and Ronquist 2001). Akaike information criterion was used to identify the most appropriate model of nucleotide substitution for the individual nuclear and mtDNA markers, and for the combined data set using the program MrModeltest version 2.2 (Nylander 2004). The GTR + I + Γ model was indicated to be the most appropriate for all data sets. The data set was partitioned by marker and then subsequently by codon position for RAG2, and search settings of 5,000,000 generations with a sampling frequency of 100 generations were used. The analyses employed the GTR + I +Γ model of nucleotide substitution, and values for model parameters were not defined a priori but treated as unknown variables with uniform priors. Resulting burn-in values (the point at which the model parameters and tree scores reached stationarity) were determined empirically by evaluating likelihood scores. For the Bayesian analysis all runs were checked for sufficient mixing, stable convergence on a unimodal posterior, and effective sample sizes (Drummond et al. 2002) > 100 for all parameters using TRACER version 1.4 (Drummond and Rambaut 2003).

To estimate divergence times among taxa we used the relaxed lognormal molecular clock as implemented in BEAST version 1.4.8 (Drummon and Rambaut 2007). Fossil calibrations were used to place a prior on 3 nodes, and to ensure conservative estimates with realistic confidence intervals a uniform prior distribution was used for all fossil calibrations. The calibrations were: 1) A prior age of 30 million years ago (mya) for Phyllostomidae. The oldest probable stem phyllostomids are found in the Whitneyan (32–30 mya—Czaplewski et al. 2008). Therefore, the maximum root age of Phyllostomidae was set at the Eocene-Oligocene boundary (34 mya). 2) A minimum of 37 mya for the split between Vespertilionidae and Molossidae, because verified vespertilionid and molossid fossils have been found from the middle Eocene (McKenna and Bell 1997). A maximum of 48 mya was used on this calibration to encapsulate the entire middle Eocene in the prior. 3) A minimum age of 30 mya for the Phyllostomidae-Mormoopidae divergence with a uniform distribution (Teeling et al. 2005). The oldest fossils uniting this group are found in the Whitneyan (32–30 mya—Morgan 2002). Therefore, the maximum age of the Mormoopidae-Phyllostomidae divergence was set at 37 mya to encapsulate the late Eocene. The data set was partitioned as in the Bayesian phylogenetic analysis described above. Using the above calibrations as point estimates, we generated a chronogram using r8s version 1.60 (Sanderson 2003) to provide a good starting tree for the dating analysis. This was accomplished by allowing r8s to rescale the branch lengths on the Bayesian phylogram shown in Fig. 1 to time rather than substitutions per site. To test the effects of the each calibration prior on posterior divergence estimates, initial analyses of 20,000,000 generations were run with each individual calibration and with all possible combinations. Although confidence intervals were significantly wider with fewer calibrations, mean estimates varied by <2 million years, indicating calibrations were robust. An initial analysis of 50,000,000 generations with 10% burn-in was run to tune operators. The final analysis consisted of 2 separate runs of 50,000,000 generations, each with 10% burn-in. Results of these final 2 runs were combined to obtain final estimates of divergence, and all runs were checked for sufficient mixing, stable convergence on a unimodal posterior, and effective sample sizes (Drummond et al. 2002) > 100 for all parameters using TRACER version 1.4 (Drummond and Rambaut 2003).

For analyses of rate variation we 1st determined if rates across the entire topology were evolving in a clocklike fashion. To do so we conducted a likelihood ratio test (Felsenstein 1981) of the null molecular clock model versus the alternative model, in which each branch was allowed its own unique rate. To determine if the vespertilionids and phyllostomids were evolving at significantly different rates we used the Baseml program (part of the PAML version 4.2 package—Yang 2007). For this analysis the pruned topology consisting of only Vespertilionidae, Phyllostomidae, and the outgroup Saccopteryx was used. We then compared 2 models: a 3-partition model where Vespertilionidae, Phyllostomidae, and the lineage leading from the outgroup to the ingroup each were constrained to having separate rates; and a 2-partition model where all ingroup lineages were constrained to have the same rate and the lineage leading from the outgroup to the ingroup was allowed to have its own rate. We then compared these models using a likelihood ratio test (Felsenstein 1981). In addition, we conducted a pairwise relative rate test using HyPhy version 1.0 (Kosakovsky Pond et al. 2005) to identify significant rate differences between individual taxa (the total data set was used). This analysis makes all possible pairwise comparisons under 2 models, a null model where taxon pairs are constrained to equal rates, and an alternative model with that constraint relaxed. For all Baseml and HyPhy analyses the GTR + I + Γ model of nucleotide substitution was used.

For tests of cladogenesis asymmetry we used multiple approaches. For all of these analyses each genus was reduced to a single representative to reduce sampling bias. However, in instances where genera were not monophyletic multiple lineages were included (i.e., EptesicusLack et al. 2010; Roehrs et al. 2010). Because of the high diversity of Myotis (>100 species—Simmons 2005), representatives of each subclade outlined by Hoofer and Van Den Bussche (2003), Lack et al. (2010), Stadelmann et al. (2004), and Stadelmann et al. (2007) were retained. First, we employed the whole-tree method implemented in SymmeTREE (Chan and Moore 2005) to test for asymmetry on 2 pruned topologies: Vespertilionidae only and Phyllostomidae only. SymmeTREE generates test statistics that differentially combine node probabilities generated under the equal-rates Markov model of stochastic diversification (Moore et al. 2004). It does this by examining large samples of simulated topologies under the equal-rates Markov model with the same number of taxa as the observed phylogeny. Each of the 6 test statistics is sensitive to topological asymmetry at different relative positions on the tree (B1 < MΣ < MΣ* < MΠ < MΠ* < IC, in order of increasing large-scale sensitivity), giving an estimate of the portion of the tree where asymmetrical diversification rates become significant. Also, to identify the specific nodes at which significant shifts in diversification rate have occurred we calculated the shift statistics Δ1 and Δ2 (Moore et al. 2004) in the SymeTREE package. This software is especially useful for phylogenies that have incomplete resolution because it incorporates phylogenetic uncertainty by randomly resolving clades with polytomies and reporting the distribution of test statistics and probability estimates. We used 1,000,000 random resolutions to generate frequentiles for all test statistics while generating 1,000,000 simulated topologies. As a 2nd indication of asymmetry in diversification rates we used the program Genie version 3.0 (Pybus and Rambaut 2002) to generate lineage-through-time plots for Vespertilionidae and Phyllostomidae. Also, to compare tree skewness between Vespertilionidae and Phyllostomidae we generated gamma statistics (γ—Pybus and Harvey 2000). This statistic depends on the distribution of internal nodes on the observed phylogeny as compared to those predicted under a model of constant diversification (γ = 0). A significant negative value indicates that internal nodes occur closer to the root than is expected under the null model.

Results

For the total data set combined alignment of the mtDNA and RAG2 sequences resulted in 4,296 aligned positions. Following exclusion of positions of questionable homology (984 positions), the final alignment consisted of 3,338 aligned positions. Of the total aligned positions, 1,763 were variable and 1,430 were potentially phylogenetically informative. When partitioned by family, Phyllostomidae had 721 informative characters and Vespertilionidae had 1,177 informative characters. When corrected for the number of taxa, Phyllostomidae had 15.7 informative characters per lineage and Vespertilionidae had 11.0 informative characters per lineage. The likelihood ratio test significantly rejected the molecular clock hypothesis (2ΔL = 1,010.88, d.f. = 169, P < 0.001). The Bayesian analysis reached stationarity at approximately 275,400 generations. Therefore, only trees following a conservative burn-in of 300,000 generations were retained. Relationships were considered statistically supported by a Bayesian posterior probability ≥ 0.95. For Phyllostomidae 91.0% (40/44) of the nodes were supported, but only 77.0% (82/106) were supported for Vespertilionidae (Fig. 1).

The PAML (Yang 2007) likelihood ratio test for a significant rate difference between Vespertilionidae and Phyllostomidae was significant (2ΔL = 23.206, d.f. = 1, P < 0.001), indicating that the 2 families are evolving at significantly different rates. Baseml calculates rates for each partition relative to the rate of branches not included in the partitions. Therefore, the rate of the unpartitioned branches (in this case, the lineage leading from Saccopteryx to the ingroup) is set to 1.0, and all other rates are multiples of that rate. Clade-specific relative rates for this data set were 1.48 for Vespertilionidae and 0.988 for Phyllostomidae, indicating the substitution rate for Vespertilionidae has been approximately 50% faster than for Phyllostomidae. Pairwise relative rate tests indicated that the majority of taxa within Vespertilioninae are evolving at a significantly higher rate than Phyllostomidae (Fig. 2).

Fig. 2

Results of pairwise relative rate comparisons among all ingroup taxa. Each square represents a comparison between a species in the left tree (the Bayesian phylogram of Fig. 1 with Saccopteryx outgroup removed) and a species in the mirrored tree on the right. Dark squares (blue squares in online figure) indicate that the top taxon has a significantly lower rate than the left taxon, and light squares (red squares in online figure) indicate that the top taxon has a significantly higher rate than the left taxon (P < 0.05). White squares indicate that differences between taxa were not statistically significant.

For Vespertilionidae the whole-tree statistics indicate significant diversification rate variation, with 5 of 6 measures either indisputably significant, or with significant P-values included in the minimum and maximum tail probabilities (Table 1). The relatively large-scale sensitive statistics were highly significant, but significance was decreased in the smaller-scale sensitive MΣ* and MΣ, and B1 was not significant. For Phyllostomidae the pattern of significance was reversed. The large-scale sensitive statistics were not significant, but significance increased in the smaller-scale sensitive measures (Table 1). The shift statistics identified multiple nodes within both Vespertilionidae and Phyllostomidae where significant shifts in diversification rate have occurred (Table 2). For Vespertilionidae a diversification shift was detected at 2 nodes: at the base of Vespertilioninae (labeled 1 in Fig. 3), and at the base of Plecotini and the remainder of Vespertilioninae, excluding Lasiurus and Scotophilus (labeled 2 in Fig. 3). For Phyllostomidae a diversification shift also was detected at 2 nodes: the divergence between Lonchorhina and all other phyllostomids (labeled 3 in Fig. 4), and the divergence between Rhinophylla and the Stenodermatinae (labeled 4 in Fig. 4). For the lineage-through-time plots (Fig. 5), the slope of the initial diversification is much steeper at the base of Vespertilionidae than for Phyllostomidae. For Vespertilionidae the gamma statistic was significantly negative (γ = −4.676, P < 0.001), indicating that internal nodes were skewed significantly towards the root of the tree, and the gamma statistic was near 0 and nonsignificant for Phyllostomidae (γ = 0.226, P = 0.648).

Fig. 3

Vespertilionid chronogram resulting from the relaxed, uncorrected lognormal clock molecular dating analysis of the combined data set conducted in BEAST version 1.4.8. Shaded bars represent the 95% highest posterior density interval for divergence estimates. Divergence estimates correspond to the mean node ages in units of millions of years before present. Arrows and numbered nodes correspond to significant shift values shown in Table 2.

Fig. 4

Phyllostomid chronogram resulting from the relaxed, uncorrected lognormal clock molecular dating analysis of the combined data set conducted in BEAST version 1.4.8. Shaded bars represent the 95% highest posterior density interval for divergence estimates. Divergence estimates correspond to the mean node ages in units of millions of years before present. Arrows and numbered nodes correspond to significant shift values shown in Table 2.

Fig. 5

Lineage-through-time plots showing diversification in Vespertilionidae (circles) and Phyllostomidae (triangles). The vertical dashed line indicates the point at which sampling of generic lineages becomes incomplete and after which diversity is likely underestimated.

View this table:
Table 1

The P-values for the 6 test statistics used to examine among-clade diversification rates. Significance is determined by comparison of observed versus simulated values based on 1,000,000 simulated topologies under the equal-rates Markov model. The range of values corresponds to the upper and lower values resulting from 1,000,000 random resolutions of polytomies.

IcMП*MПMΣ*MΣB1
Vespertilionidae0.00001–0.00340.00001–0.016650.00001–0.016650.0247–0.33650.0114–0.08860.1207–0.4454
Phyllostomidae0.0833–0.15610.1137–0.57560.0334–0.27620.0361–0.24290.0018–0.10370.0032–0.0195
View this table:
Table 2

Clade names and node numbers from Figs. 3 and 4 for which shift statistics indicate a significant (P < 0.05) increase in diversification rate. Upper and lower bounds for the shift statistics and P-values are the result of 1,000,000 random resolutions of polytomies.

CladeNode no.Δ1PΔ2P
Vespertilioninae14.319–4.6620.009–0.0264.025–4.2850.013–0.041
Plecotini22.212–3.0120.043–0.1881.989–2.6640.046–0.201
Lonchorhina32.915–2.9430.037–0.0412.639–2.8010.044–0.048
Rhinophylla/Stenodermatinae42.5310.0362.3030.048

Discussion

The confounding factors in resolving relationships within Vespertilionidae appear to be 2-fold. Both the rate-partitioning analysis and the pairwise relative rate test were suggestive of elevated evolutionary rates within Vespertilionidae. In this situation multiple substitutions can obscure phylogenetic signal, and when evolutionary models mishandle the data, relationships are either unresolved or resolved incorrectly. This problem has been identified for a number of taxonomic groups and is likely to add complexity to resolving deep nodes on the tree of life, even with enormous amounts of data (Jeffroy et al. 2006; Phillipe and Telford 2006). Although comparisons with Phyllostomidae indicate rate disparity, it is evident by looking within Vespertilionidae that elevated evolutionary rates are obscuring relationships. The pairwise relative rate test indicates that the subfamily Myotinae has evolved at a rate significantly slower than that of the rest of Vespertilionidae (Fig. 2). Consequently, relationships within Myotinae are resolved nearly completely with relatively smaller data sets (Hoofer and Van Den Bussche 2003; Ruedi and Mayer 2001; Stadelmann et al. 2004, 2007), and the same molecular markers that failed to resolve relationships within Vespertilioninae (Roehrs et al. 2010) resolved nearly all myotine relationships (Lack et al. 2010).

Since the initial suggestion of the constancy (with respect to time) of evolution by Zuckerkandl and Pauling (1962), a multitude of examples that violate the molecular clock hypothesis have been described (Bromham and Penny 2003). Although these exceptions have complicated the production of an accurate evolutionary timescale for the tree of life, relaxed clock approaches have been developed to provide more realistic divergence estimates. However, elevated substitution rates as a problem in resolving systematic relationships only recently have been addressed (Jeffroy et al. 2006). Solutions to data sets with high systematic error are beginning to surface (Baurain et al. 2007; Montgelard et al. 2008; Rodriguez-Ezpeleta et al. 2007), and it is likely that these approaches (in addition to the generation of more data) will be necessary to correctly resolving relationships within Vespertilioninae.

In addition to evolutionary rate disparity, differences in the rate with which Vespertilionidae and Phyllostomidae initially diversified also appear to have influenced resolution. For Phyllostomidae cladogenesis analyses indicate that diversification was more gradual, with a relatively linear lineage-through-time plot and a nonsignificant gamma, and whole-tree asymmetry statistics increasing in significance with progression towards the tips. Diversification of Vespertilionidae is best described as a left-skewed radiation in which a common ancestor rapidly diversified into a large number of distinct lineages followed by a period of reduced diversification. The whole-tree statistics, lineage-through-time plot, and the gamma statistic characterize Vespertilionidae as a radiation compressed at the root, and the nodes identified by the shift statistic analysis all were located in the older portion of the phylogeny, including the node basal to Vespertilioninae (Fig. 3). This is further supported by the presence of extremely short internodes among the lineages representing individual tribes (Fig. 3). Although Phyllostomidae possessed the same number of nodes significant for the shift statistic analysis as Vespertilionidae, the lineages separating those nodes in Phyllostomidae are spaced more evenly, indicating a longer period of stasis before subsequent speciation.

Cladogenesis asymmetry has been observed for a wide range of taxa (Magallon and Sanderson 2001; Mayhew 2002; Purvis 1995), and statistical methods for its detection are accumulating and their power increasing (Bokma 2002, 2003; Chan and Moore 2002; Pybus and Harvey 2000; Rabosky 2006; Ree 2005). Detection of these events is useful for 2 reasons: a rapid accumulation of lineages could create a problem in resolving evolutionary relationships; and events of rapid cladogenesis denote snapshots in time where some conditional threshold, whether intrinsic (e.g., genetic or physiological), extrinsic (e.g., environmental), or a combination of the 2 was met that allowed a mass accumulation of diversity over a short time period. For these reasons it is important not only to identify these situations but also to develop ways to circumvent them and obtain a resolved and correct phylogeny.

Previous treatments of diversification and adaptive radiation (Lack 1947; Mayr 1942; Simpson 1953) suggested that competition hinders the process, and a release from competitive stifling can lead to a sudden increase in the rate of diversification. Schluter (2000) outlined this process as the filling of an empty or underutilized adaptive zone either through the acquisition of a novel phenotype (the key innovation hypothesis—Simpson 1953), leading to a competitive advantage, or extinctions rendering a previously occupied adaptive zone empty. Following this competitive release, competition among individuals within the radiating lineage promotes divergence (Schluter 1994). A key innovation allowing for rapid diversification of Vespertilionidae is unlikely (or at least not apparent) because of the lack of a single morphological character that distinguishes the group from other bat families and unites the constituent taxa. Another possible explanation is climatic fluctuations. During the Eocene-Oligocene transition, the most significant climatic shift of the Cenozoic era occurred (∼33.5 mya—Zanazzi et al. 2007). The tropical climate of the Eocene was replaced by a more modern climate characterized by ice caps at the poles, dramatic drops in temperature, and strong seasonality (Wolfe 1978). In addition, the widespread Eocene forests were replaced by vast grasslands in temperate regions. However, debate has raged over the effect this climatic event had on prehistoric North American mammalian fauna, with fossil evidence lacking for a significant mammalian extinction or radiation event (Prothero and Heaton 1996; Zanazzi et al. 2007). The lack of chiropteran fossils has prohibited examination of the effect of the Eocene-Oligocene transition on bats, and the conclusions of Prothero and Heaton (1996) and Zanazzi et al. (2007) did not take chiropterans into consideration. Nonetheless, the rapid diversification of Vespertilionidae, which our molecular dating analyses date very close to the Eocene-Oligocene transition (∼32 mya), suggests that this event might have had a significant effect on temperate bats.

Regardless of the actual event that stimulated diversification, an ecological component to adaptive radiations exists, and differences in the environment in which these 2 families originated must also be considered. Phyllostomid bats are endemic to the New World tropics (Czaplewski et al. 2008), but the place of origin for the globally distributed Vespertilionidae is less clear. Teeling et al. (2005) attempted a biogeographic reconstruction to identify the continent of origin for all chiropteran families based on current distributions and the fossil record but were able to narrow down only the vespertilionid ancestral range to the Laurasian supercontinent. In addition, the largely unresolved vespertilionid phylogeny precludes any further inference of ancestral origins. However, if one considers only the fossils, the majority of stem vespertilionids (those dating prior to the Miocene) have been found in temperate Europe, Asia, or North America (McKenna and Bell 1997). In addition, peak generic diversity of extant vespertilionid taxa is also found in temperate Europe, Asia, and North America (Hutson et al. 2001). If vespertilionids originated in the temperate Northern Hemisphere, this might help to explain their diversification.

Patterns of climatic fluctuations suggest that the tropics have been relatively constant through significant climatic fluctuation that had dramatic effects on temperate organisms (Jansson and Dynesius 2002). For example, the majority of plant taxa occurring in tropical rain forests persisted with little change in their range throughout the last glacial maximum (Flenley 1998), whereas the temperate flora underwent significant shifts with each glacial advance (Huntley and Birks 1983). However, considerable difficulty has occurred in reconciling the differential effects of climatic fluctuations and many hypotheses that attempt to explain the increased diversity at the tropics. The latitudinal gradient in diversity is an old dilemma for which hypotheses have been numerous but explanatory power limited (Hillebrand 2004; Pianka 1966; Rohde 1992). A common theme of many hypotheses is the suggestion that speciation rates are higher in the tropics (Cardillo 1999; Rohde 1992); therefore, a tropical clade will have higher diversity than a temperate clade of the same age. Problems with this hypothesis have arisen on 2 levels. Weir and Schluter (2007) tested this speciation rate hypothesis on a large sample of birds and mammals and found trends opposite of those predicted, with speciation rates significantly decreasing as the equator is approached. In terms of time, recent analyses have provided substantial evidence against clade age as an explanation for clade diversity, especially at higher taxonomic levels (Rabosky 2009a, 2009b; Ricklefs 2007). Our molecular dating analyses suggest that the radiations of Vespertilionidae and Phyllostomidae occurred at approximately the same time (∼32 mya), and the total number of generic lineages present in each clade is comparable, but the rate with which they were accumulated differs. Weir and Schluter (2007) and several others before them (Fisher 1960; Wallace 1878; Weir 2006) suggested that the latitudinal diversity gradient can be explained by reduced speciation rates at the equator due to low species turnover. This could be due to the number of species in the tropics being maintained at or near carrying capacity, offering relatively little opportunity for rapid speciation events (Weir 2006). Conversely, a higher species turnover in temperate environments could translate into large fluctuations in total species diversity and therefore a higher frequency of rapid speciation events. The data presented here seem to support this hypothesis, although we do admit it is dependent on a temperate origin for Vespertilionidae, an assumption that is somewhat speculative.

Further evidence of a different evolutionary history for Vespertilionidae relative to Phyllostomidae lies in morphology. Although a rapid diversification will result in new diversity in terms of lineages, phenotypic diversity can be reduced among newly formed lineages, because reproductive isolation forms extremely quickly. The strength of natural selection following divergence dictates phenotypic variation; if niche utilization does not shift, one would not expect phenotypic optima to shift. This pattern is evident in the vespertilionid radiation, where low amounts of behavioral and morphological variation exist among lineages (Corbet and Hill 1991; Koopman 1994; Nowak 1999; Simmons 1998). Conversely, Phyllostomidae has been described as the most morphologically and behaviorally (i.e., 6 distinct feeding strategies) diverse family of extant mammals (Baker et al. 2000, 2003; Simmons 1998). This pattern of extreme phenotypic diversity could be explained by the intense resource competition in the tropics, leading to rapid morphological specialization; for Vespertilionidae rapid diversification in a temperate environment followed by a quick spread of species to nearly every part of the globe reduced the need to specialize.

In conclusion, the confounding factors in resolving relationships among vespertilionine bats appear to be derived from both an elevated rate of molecular evolution and a rapid initial diversification. To overcome these issues several approaches must be used. First, the use of more slowly evolving molecular markers (i.e., nuclear exons, expressed sequence tags, etc.) combined with more effective models of the substitution process might be necessary to overcome the systematic error introduced by multiple substitutions. Overcoming incomplete lineage sorting due to a rapid radiation is significantly more difficult because traditional concatenation and consensus approaches accompanied by an increased molecular data set will not improve phylogenetic inference (Kubatko and Degnan 2007). Coalescent approaches that can infer species trees in the presence of incomplete lineage sorting (Edwards et al. 2007; Heled and Drummond 2010; Liu et al. 2008) accompanied by an expansion of the current data set will be necessary.

Although it is clear that Vespertilionidae and Phyllostomidae have had dramatically different evolutionary histories, as indicated by phylogeny, behavior, morphology, and ecology, it is unclear exactly what led to the differences in the basal radiation for each family. We suggest that the ecological history of tropical versus temperate environments played a significant role, but this is somewhat speculative. To validate our hypotheses it is crucial that a resolved vespertilionid phylogeny be obtained. Gaps in the chiropteran fossil record also are problematic in distinguishing evolutionary history from speculation. However, a resolved phylogeny for Vespertilionidae would allow for a more thorough treatment of the fossil record and a better understanding of the evolutionary history of the most speciose chiropteran family.

Acknowledgments

We extend our sincere gratitude to Z. Roehrs, C. Stanley, and J. Wilkinson for assistance in the laboratory and to Z. Roehrs and A. Doust for their comments regarding previous versions of this manuscript. Financial support for this project was provided by National Science Foundation grants DEB-9873657 and DEB-0610844 to RAVDB. Any opinion, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

Appendix I

Specimens examined.—All molecular sequences used in this study were downloaded from GenBank, and accession numbers are given below.

View this table:

Specimens examined.—All molecular sequences used in this study were downloaded from GenBank, and accession numbers are given below.

TaxonmtDNARAG2
Emballonuridae
Saccopteryx bilineataAF263213AY141015
Furipteridae
Furipterus horrensAF345921AF316451
Miniopteridae
Miniopterus australisAY395864GU328066
Miniopterus fraterculusAY495486GU328067
Miniopterus inflatusAY495487GU328068
Miniopterus schreibersiiAY395865GU328070
Miniopterus tristisAY495489GU328071
Molossidae
Chaerephon pumilusAY495454GU328051
Otomops martiensseniAY495459GU328097
Sauromys petrophilusAY495460GU328116
Mormoopidae
Mormoops megalophyllaAF263220AY141020
Pteronotus parnelliiAF263221AF316482
Myzopodidae
Myzopoda auritaAF345926AY141022
Natalidae
Natalus micropusAF345925AY141023
Natalus stramineusAF345924AY141024
Noctilionidae
Noctilio albiventrisAF263223AF330811
Noctilio leporinusAF263224AF316477
Phyllostomidae
Ametrida centurioAY395802AF316430
Anoura caudiferAY395835AF316431
Ardops nichollsiAY395803AF316434
Artibeus jamaicensisAF263226GU328048
Artibeus obscurusAY395805DQ985504
Brachyphylla cavernarumAY395806AF316436
Carollia perspicillataAY395836AF316437
Centurio senexAF263227AF316438
Choeroniscus minorAY395809AF316440
Choeronycteris mexicanaAY395808AF316441
Chrotopterus auritusAF411538AF316442
Desmodus rotundusAF263228AF316444
Dermanura cinereaAY395810AF316443
Diaemus youngiAF411534AF316445
Diphylla ecaudataAF411533AF316447
Ectophylla albaAY395811AF316448
Enchisthenes hartiiAY395838AF316449
Erophylla sezekorniAY395839AF316450
Glossophaga soricinaAY395840AF316452
Glyphonycteris daviesiAY395812AF316464
Hylonycteris underwoodiAY395813AF316453
Lampronycteris brachyotisAF411536AF316463
Leptonycteris curasoaeAY395814AF316454
Lonchorhina auritaAY395843AF316457
Macrophyllum macrophyllumAF411540AF316458
Macrotus waterhousiiAF263229AF316460
Micronycteris hirsutaAY395819AF316466
Micronycteris megalotisAY395821AF316467
Micronycteris minutaAY395823AF316468
Micronycteris schmidtorumAF411535AF316470
Mimon crenulatumAF411543AF316472
Monophyllus redmaniAY395824AF316473
Musonycteris harrisoniAY395844AF316475
Phylloderma stenopsAF411542AF316480
Phyllostomus elongatusAF411541AF316479
Platyrrhinus brachycephalusAY395825AF316481
Pygoderma bilabiatumAY395826AF316483
Rhinophylla pumilioAY395827AF316484
Sphaeronycteris toxophyllumAY395828AF316486
Stenoderma rufumAY395829AF316487
Tonatia saurophilaAF411530AF316489
Trachops cirrhosusAF411539AF316490
Uroderma bilobatumAY395831AF316491
Vampyriscus bidensAY395833AF316492
Vampyrodes caraccioliAY395846AF316494
Vampyrum spectrumAF411537AF316495
Thyropteridae
Thyroptera tricolorAF263223GU328118
Vespertilionidae
Antrozous pallidusAF326088GU328047
Arielulus aureocollarisHM561621HM561643
Barbastella barbastellaAF326089GU328049
Baeodon alleniAF326108HM561632
Bauerus dubiaquercusAY395863GU328050
Chalinolobus gouldiiAY495461HM561665
Chalinolobus morioAY495462GU328054
Cistugo seabraeGU328039GU328052
Corynorhinus mexicanusAF326090GU328053
Corynorhinus rafinesquiiAF326091GU328055
Corynorhinus townsendiiAF263238AY141029
Eptesicus brasiliensisAY495464HM561644
Eptesicus diminutusAY495465GU328056
Eptesicus dimissusGU328040GU328057
Eptesicus furinalisAF263234AY141030
Eptesicus fuscusAF326092GU328058
Eptesicus hottentotusAY495466GU328059
Eptesicus macrotusAY495472HM561646
Eptesicus magellanicusHM561625HM561649
Eptesicus serotinusAY495467HM561650
Euderma maculatumAF326093GU328060
Glauconycteris argentataAY495468HM561652
Glauconycteris beatrixAY495469HM561653
Glauconycteris egeriaAY495470HM561654
Glauconycteris variegataAY495471HM561655
Harpiocephalus harpiaAF263235AY141031
Hypsugo cadornaeGU328041GU328061
Hypsugo eisentrautiAY495473HM561666
Hypsugo nanusAY495474GU328062
Hypsugo saviiAY495475HM561667
Idionycteris phyllotisAF326094GU328063
Kerivoula hardwickiiAF345928AY141034
Kerivoula lenis (analyzed previously
as K. papulosa)AF345927AY141035
Kerivoula pellucidaAY495476GU328064
Laephotis namibensisAY495477HM561668
Lasionycteris noctivagansAF326095GU328065
Lasiurus atratusAY495478HM561635
Lasiurus blossevilliiAY495479HM561636
Lasiurus borealisAY495480HM561637
Lasiurus cinereusAY495482HM561638
Lasiurus egaAY495483HM561639
Lasiurus intermediusHM561627HM561640
Lasiurus seminolusAY495484HM561641
Lasiurus xanthinusAY495485HM561642
Murina cyclotisGU952767GU328072
Murina huttoniAY495490GU328073
Murina tubinarisGU952768GU328074
Myotis albescensAY495492GU328076
Myotis bocagiiAF326096GU328077
Myotis cf. browni (analyzed previously
as M. muricola)AY495504GU328086
Myotis californicusAY495495GU328078
Myotis capacciniiAY495494GU328079
Myotis ciliolabrumAY495497GU328080
Myotis dominicensisAY495500GU328081
Myotis fortidensAY495502GU328082
Myotis keaysiAY495503GU328083
Myotis latirostrisGU952769GU328084
Myotis levisAF326097GU328085
Myotis moluccarum (analyzed previously as M. adversus)AY495491GU328075
Myotis myotisAF326098GU328087
Myotis nigricansAF326099GU328088
Myotis ripariusAF263236GU328089
Myotis septentrionalisAY495507GU328090
Myotis thysanodesAF326100GU328091
Myotis veliferAF263237AY141033
Myotis volansAY495510GU328092
Myotis welwitschiiAY495511GU328093
Myotis yumanensisAY495512GU328094
Neoromicia brunneusAY495514HM561669
Neoromicia rendalliAY495515HM561670
Neoromicia somalicusAY495516HM561671
Nyctalus leisleriAY495517HM561657
Nyctalus noctulaAY495518HM561658
Nycticeinops schlieffeniAF326101GU328095
Nycticeius humeralisAF326102GU328096
Otonycteris hemprichiiAF326103GU328098
Parastrellus hesperusAY495522GU328099
Perimyotis subflavusAY495523GU328103
Pipistrellus coromandraAY495524GU328102
Pipistrellus hesperidusHM561628HM561659
Pipistrellus javanicusAY495525GU328105
Pipistrellus nathusiiAF326104HM561660
Pipistrellus paterculusHM561629HM561661
Pipistrellus pipistrellusAY495529HM561662
Pipistrellus pygmaeus (analyzed previously as P. pipistrellus)AF326105GU328107
Pipistrellus tenuisAY495529HM561663
Plecotus auritusAF326106GU328100
Plecotus austriacusAF326107GU328101
Plecotus gaisleriGU328043GU328104
Rhogeessa aeneusAY495530HM561633
Rhogeessa miraAY495531HM561634
Rhogeessa parvulaAF326109GU328108
Rhogeessa tumidaAF326110GU328109
Scotoecus hirundoAY495536HM561664
Scotomanes ornatusAY495537HM561656
Scotophilus borbonicusAY495532GU328110
Scotophilus dinganiiAY495533GU328111
Scotophilus heathiiAY495534GU328112
Scotophilus kuhliiAF326111GU328113
Scotophilus leucogasterAY395867GU328114
Scotophilus nuxAY495535GU328115
Scotophilus viridisAF326112GU328117
Tylonycteris pachypusAY495538HM561672
Tylonycteris robustulaHM561631HM561673
Vespadelus regulusAY495539GU328119
Vespadelus vulturnusAY495499GU328119
Vespertilio murinusAY395866HM561676

Footnotes

  • Associate Editor was David L. Reed.

Literature Cited

View Abstract