OUP user menu

Molecular Phylogenetics and Taxonomic Review of Noctilionoid and Vespertilionoid Bats (Chiroptera: Yangochiroptera)

Steven R. Hoofer, Serena A. Reeder, Eric W. Hansen, Ronald A. Van Den Bussche
DOI: http://dx.doi.org/10.1644/BWG-034 809-821 First published online: 29 August 2003


Relationships of 5 families within traditional Yangochiroptera (Furipteridae, Mystacinidae, Myzopodidae, Natalidae, and Thyropteridae) have been debated considerably for more than a century, resulting in several alternative hypotheses for their evolution and zoogeography. A recent cladistic assessment of diverse morphologic traits and another of mitochondrial deoxyribonucleic acid (mtDNA) sequences have contradicted both traditional classification of yangochiropterans and each other. Our goals were to test alternative phylogenetic hypotheses by examining DNA sequences (1.4 kb) from a nuclear gene (recombination-activating gene-2; Rag2) and, if appropriate, to examine family-level relationships based on concatenation of Rag2 and mtDNA sequences (4.1 kb). Relationships suggested by parsimony and Bayesian analysis of Rag2 data were congruent with mtDNA, and combined analysis afforded high statistical support for all but 2 clades. These molecular data strongly support Noctilionoidea (Furipteridae, Mormoopidae, Mystacinidae, Noctilionidae, Phyllostomidae, and Thyropteridae) and Vespertilionoidea (Molossidae, Natalidae, and Vespertilionidae) and strongly contradict traditional association between Myzopodidae and other noctilionoid or vespertilionoid families. This study, current distributions, and limited fossil record strengthen the hypothesis of a New World origin for Noctilionoidea and suggest an Old World origin for Myzopodidae. Despite considerable statistical support afforded by these molecular data, further study using extensive taxonomic sampling of all bat families is needed to fully resolve relationships among Myzopodidae, Noctilionoidea, and Vespertilionoidea.

Key words
  • Bayesian inference
  • classification
  • mtDNA
  • Noctilionoidea
  • phylogeny
  • Rag2
  • systematics
  • Vespertilionoidea
  • Yangochiroptera

Relationships among family-level lineages of bats (Chiroptera) have been conjectural since bats were 1st described, resulting in considerable debate and several alternative hypotheses for their evolution and zoogeography. Much debate has focused on relationships of 5 families (Furipteridae, Mystacinidae, Myzopodidae, Natalidae, and Thyropteridae) within the traditional infraorder Yangochiroptera (Koopman 1984; Smith 1976). A prevalent hypothesis has been that all 5 families are allied closely with Molossidae and Vespertilionidae (superfamily Vespertilionoidea), but there has been little consensus about their exact affinities within traditional Yangochiroptera (Koopman 1984, 1993; Miller 1907; Smith 1976, 1980).

Several factors have led to difficulty in elucidating relationships of these 5 families: 1) bats belonging to each family are peculiar morphologically, ecologically, behaviorally, and to some extent geographically (Miller 1907; Nowak 1999); 2) morphologic adaptations to a great diversity of feeding strategies, associated echolocation and flight strategies, and other life history traits of yangochiropterans have produced a diverse array of phenotypes, the procession of which is at best ill-defined, often even within families; 3) rapid radiation of bats (Baker et al. 1991) has left a paucity of synapomorphies among all familial lineages; and 4) each family except Mystacinidae has received little systematic attention, perhaps by virtue of limited availability of museum specimens (and tissue samples) or access to areas or countries in which they occur.

Simmons and Geisler's (1998) parsimony analysis (superseding that of Simmons 1998) of primarily morphologic features essentially was the 1st explicit cladistic assessment of all yangochiropteran families (and all bat families). Their analysis produced little support for most relationships but strongly supported monophyly of traditional Yangochiroptera and Noctilionoidea. Position of Mystacinidae was ambiguous within Yangochiroptera. Most notably, there was strong support for a clade containing Furipteridae, Myzopodidae, Natalidae, and Thyropteridae, which Simmons (1998) recognized as a new superfamily (Nataloidea) within Yangochiroptera.

The latest cladistic analysis of all yangochiropteran families, by Van Den Bussche and Hoofer (2001), assessed about 2.6 kb of a mitochondrial deoxyribonucleic acid (mtDNA) sequence (12S ribosomal ribonucleic acid, [rRNA]; transfer RNA, [tRNA]Val; and 16S rRNA). Their results differed markedly from those of Simmons and Geisler (1998), especially with regard to the 4 families of Nataloidea. Analysis of mtDNA strongly supported Furipteridae, Mystacinidae, and Thyropteridae in a clade with traditional Noctilionoidea, although it weakly supported positions of Molossidae, Natalidae, and Vespertilionidae. Perhaps of greater significance, mtDNA analysis suggested a basal position for Myzopodidae such that Emballonuridae (a traditional yinochiropteran) is allied to other yangochiropterans more closely than is Myzopodidae. This has important implications zoogeographically because it excludes the problematic hypothesis that Myzopodidae, endemic to Madagascar, shared a most recent common ancestry with families endemic to the New World. However, due to limited statistical support for positions of Myzopodidae and some other families and due to marked topologic differences compared with Simmons and Geisler (1998), Van Den Bussche and Hoofer (2001) delayed taxonomic suggestions until data from other characters could be explored.

Our goals were to test alternative phylogenetic hypotheses for traditional Yangochiroptera by examining DNA sequences (about 1.4 kb) from a nuclear gene (recombination-activating gene-2; Rag2) and, if appropriate, to examine relationships based on concatenation of Rag2 and mtDNA sequences (about 4.0 kb). We chose Rag2 because it should be unlinked from the mitochondrial genome, thereby providing data independent of those of mtDNA; its immunologic function (Akamatsu and Oettinger 1998; Oettinger et al. 1990; Schatz et al. 1989; Swanson and Desiderio 1999) should provide an estimate of evolution largely uncorrelated with morphologic adaptations of yangochiropterans, and recent study of Rag2 sequences illustrated their utility for resolving higher-level relationships of phyllostomid bats (Baker et al. 2000) and other mammalian taxa (Murphy et al. 2001a, 2001b; Teeling et al. 2000).

Materials and Methods

Rag2 data collection and analysis.—We examined 30 specimens representing 13 chiropteran families (Appendix I). We extracted genomic DNA from skeletal muscle or liver tissue (Longmire et al. 1997) and amplified a portion of the Rag2 protein-coding sequence (about 1.4 kb) by polymerase chain reaction (PCR). Internal and flanking primers and PCR conditions and thermal profile were according to Baker et al. (2000). We purified double-stranded PCR amplicons with Wizard PCR Prep DNA Purification System (Promega Corporation, Madison, Wisconsin) and sequenced both strands with Big-Dye™ chain terminators followed by electrophoresis on a 377 automated DNA sequencer (Applied Biosystems, Inc., Foster City, California). We used a combination of flanking and internal primers (Baker et al. 2000) to sequence each strand entirely and AssemblyLIGN™ 1.0.9 software (Oxford Molecular Group PLC 1998) to assemble resulting, overlapping fragments. Rag2 alignment was accomplished in CLUSTAL W software (Thompson et al. 1994), with default parameters for costs of opening and extending gaps.

We coded nucleotides as unordered, discrete characters (G, A, T, and C), multiple states as polymorphisms, and gaps as missing. In PAUP* software (test version 4.0b10; Swofford 2002), we examined level of phylogenetic signal using the g1-statistic (Hillis and Huelsenbeck 1992) for 10,000 randomly drawn trees and assessed degree of substitutional saturation (i.e., multiple hits) by plotting observed numbers of transitions and transversions (each codon position separately and combined) against Kimura (1980) index of corrected sequence divergence.

We inferred phylogenetic relationships by Bayesian analysis (Li 1996; Mau 1996; Rannala and Yang 1996) implemented in MrBayes 2.01 software (Huelsenbeck and Ronquist 2001). This relatively new approach to phylogeny reconstruction is similar to maximum likelihood analysis in that both are optimality criteria, use character-based data (such as each site along a sequence alignment), and search for trees and branch lengths that are most consistent with the data and specified model of sequence evolution. However, their objective functions are defined somewhat differently such that Bayesian methods are much quicker and efficient and provide a powerful alternative for relatively large molecular data sets when maximum likelihood analysis (especially with subsequent bootstrapping) requires an inordinate amount of computing time (Murphy et al. 2001b). Further, we relied on Bayesian analysis, rather than parsimony analysis, because it provides an objective system for character weighting (Felsenstein 1981), a more efficient system with which to reconcile important biologic phenomena for molecular data (e.g., among-site rate variation, unequal base frequencies, nonindependence of substitutions), and provides access to maximum amount of information in a set of DNA sequences (Whelan et al. 2001). However, we did perform parsimony analyses in PAUP* (test version 4.0b10—Swofford 2002) to provide a check of Bayesian results because parsimony remains a governing method for phylogeny reconstruction from discrete character data (morphologic and molecular—Lewis 2001). Parsimony and bootstrapping methods were according to Van Den Bussche et al. (2002), which weighted all characters and substitution types equally and designated 4 outgroups (Emballonuridae: Diclidurus, Saccopteryx; Pteropodidae: Pteropus; Rhinopomatidae: Rhinopoma).

We ran all Bayesian analyses at least 1 × 106 generations with 1 cold and 3 incrementally heated Markov chains, random starting trees for each chain, and trees sampled (saved) every 10 generations. General time-reversible model with allowance for a gamma distribution of rate variation (Γ) and proportion of invariant sites (I) best fit our data (Modeltest—Posada and Crandall 1998). We did not define values for model parameters (from Modeltest) a priori but instead treated them as unknown variables (with uniform priors) to be estimated in each Bayesian analysis (Leaché and Reeder 2002). We also used the general time-reversible model with estimates of among-site rate variation partitioned by codon position as described for protein-coding genes (Huelsenbeck and Ronquist 2001). We ran sets of 3 independent analyses for each model and each specified outgroup (Pteropus and Rhinopoma) with burn-in values (initial set of unstable generations to be ignored) based on empirical evaluation of likelihoods converging on stable values (=stationarity).

Conditional combination of Rag2 and mtDNA.—We assessed combinability of Rag2 and mtDNA data sets according to Wiens (1998); see also Leaché and Reeder 2002). However, because we were able to generate (or obtain from GenBank) Rag2 sequences for most, but not all, taxa examined by Van Den Bussche and Hoofer (2001), we truncated both data sets to include only those taxa shared between studies. In doing so, 1 pteropodid (Nyctimene), 1 rhinopomatid (Rhinopoma), 1 emballonurid (Emballonura), 1 furipterid (Furipterus), 3 molossids (Eumops, Molos sus rufus, and Promops), and 1 phyllostomid (Artibeus) were excluded from mtDNA data set. One phyllostomid (Ariteus) likewise was excluded from Rag2 data set. We also examined Rag2 sequences for the emballonurid Diclidurus albus (available from GenBank), which was not examined by Van Den Bussche and Hoofer (2001). We did not exclude Diclidurus, however, because the same mtDNA fragment was available for D. scutatus (Van Den Bussche and Hoofer, in litt.).

With several taxonomic deletions and 1 addition (Diclidurus), it was necessary to realign mtDNA sequences. We performed 2 alignments in CLUSTAL X software (Thompson et al. 1997), differing in gap cost ratios: 1 using the default ratio (15.00:6.66) and the other with a smaller ratio (5:4—Hickson et al. 2000). Both were refined by eye according to secondary structural models (Anderson et al. 1982; de Rijk et al. 1994; Springer and Douzery 1996). We analyzed this slightly different set of mtDNA data with Bayesian methods (as described above) and general time-reversible + F + I model (Modeltest—Posada and Crandall 1998) and relied on parametric posterior probabilities from Bayesian analysis (P ≥ 0.95) to indicate strongly supported relationships (Wiens 1998).


Rag2 data.—The 21 Rag2 sequences we generated have been submitted to GenBank (AY141015AY141035) . Alignment of these plus 9 sequences obtained from GenBank resulted in 1,374 aligned sites. We deleted the last 12 sites, however, because these data were missing for Diclidurus (obtained from GenBank). Of 1,362 aligned sites, 1,043 (77%) were constant or parsimony-uninformative. Fifty-seven (18%) of the informative sites were at 1st-codon positions, 38 (12%) at 2nd positions, and 224 (70%) at 3rd positions. Only a single 3-bp gap was inserted in the alignment, corresponding to a deletion of codon 41 (or 42) in Thyroptera discifera. Whereas it is not obvious exactly which codon was deleted (41 or 42), deletion of codon 41 perhaps represents a slightly more parsimonious solution; codon 41 is variable among all taxa (glutamine, histidine, proline, and tyrosine), and codon 42 is constant (glutamine). Furthermore, for this set of sequences, the exact placement for the gap is trivial because separate phylogenetic analyses with the gap inserted at either codon position gave the same topology and support values. Plots (not shown) of observed numbers of transitions and transversions against corrected sequence divergence revealed no evidence for saturation even for 3rd-position transitions. Thus, we did not remove 3rd-position data before phylogenetic analysis. g1 statistic was skewed significantly left (g1 = −0.78; P < 0.01), indicating strong phylogenetic signal (Hillis and Huelsenbeck 1992).

Bayesian likelihoods reached stationarity by 50,000 generations (i.e., burn-in = 5,000), which thinned our data to 95,000 sample points. Topology and posterior probabilities for nodes and model parameters for all sets of runs (3 runs each) were in excellent agreement regardless of specified model of sequence evolution or choice of outgroup. Bayesian analyses strongly supported (P ≥ 0.95) monophyly of all families for which ≥2 representatives were examined (Fig. 1). Relationships among families that received strong support in Rag2 tree were limited to a clade containing Furipteridae, Mormoopidae, Noctilionidae, Phyllostomidae, and Thyropteridae and a clade containing Molossidae, Natalidae, and Vespertilionidae; within each clade, relationships among families were resolved but were supported weakly (i.e., P < 0.95). Bayesian analysis of Rag2 data offered no supported resolution to positions of Emballonuridae, Mystacinidae, and Myzopodidae. Parsimony analysis also produced limited resolution with concomitant statistical support. There were no supported conflicts between parsimony and Bayesian analyses, and all clades supported by posterior probabilities (P ≥ 0.95) also were supported in >50% bootstrap iterations.

Fig. 1

Best tree (mean In 1 = −7,217.262) from Bayesian analysis of Rag2 data (about 1.40 kb) based on general time-reversible + Γ + I model. Bayesian posterior probabilities (rounded to nearest 10th) are shown above each branch; asterisks denote strongly supported branches (≥0.95). K. = Kerivoula, M. = Myotis, N. = Noctilio, Nat. = Natalus, and T. = Thyroptera.

mtDNA data.—Complete sequence (12S rRNA–tRNAVal–16S rRNA) for D. scutatus has been submitted to GenBank (AY141036) . Alignment of this plus 27 sequences obtained from GenBank resulted in 2,753 aligned sites, of which 1,148 (41.7%) were parsimony-informative: 406 (35.4%) in 12S rRNA, 26 (2.2%) in tRNAVal, and 716 (62.4%) in 16S rRNA. We excluded 599 characters (115 were constant or parsimony-uninformative) from all analyses because positional homology was ambiguous (12S rRNA, 162; tRNAVal, 14; 16S rRNA, 424). Most ambiguous sites were within loop regions of ribosomal genes, but some were in stem regions and sites within tRNA gene. Alignment using 5:4 gap cost ratio was longer (2,833 aligned sites) than default alignment due to insertion of several more gaps primarily in large loop regions of ribosomal genes. Number (and percentage) of parsimony-informative characters per gene was similar to default alignment. There were slightly more ambiguous characters, however, corresponding to the increase in inserted gaps in variable loop regions.

Bayesian likelihoods reached stationarity by 70,000 generations (i.e., burn-in = 7,000), which thinned our data to 93,000 sample points. Topologies and posterior probabilities for nodes and model parameters for all sets of runs (3 runs each) were in excellent agreement regardless of choice of outgroup. Also, analysis of both alignments produced identical topologies and essentially identical posterior probabilities. There were no supported conflicts (P ≥ 0.95) between our analysis of mtDNA (not shown) and Rag2; therefore, we combined data sets.

Combined data.—Combined alignment for 29 taxa shared between data sets (4,114 characters) gave 977 parsimony-informative characters, of which 665 (68%) were mitochondrial and 312 (32%) nuclear. Bayesian likelihoods reached stationarity by 100,000 generations (i.e., burn-in = 10,000), which thinned our data to 90,000 sample points. Topology and posterior probabilities for nodes and model parameters for all sets of runs again were in excellent agreement regardless of choice of outgroup.

Combined analysis produced high statistical support for all but 2 clades (Fig. 2). Of 26 clades, 24 were supported by a posterior probability ≥0.95, and 22 were supported by >75% bootstrap iterations. The 2 clades supported weakly by all analyses were 1 containing Thyropteridae, Furipteridae, and Noctilionidae and 1 containing Noctilionoidea and Vespertilionoidea.

Fig. 2

Best tree (mean In 1 = −24,517.333) from Bayesian analysis of combined Rag2 and mtDNA data (about 4.0 kb) based on general time-reversible + Γ + I model. Bayesian posterior probabilities (rounded to nearest 10th) are shown above each branch; asterisks denote strongly supported branches (≥0.95). Abbreviations as for Fig. 1.


Rag2 and combined analyses.—Phylogenetic analysis of DNA sequences from nuclear Rag2 gene provides a novel assessment of yangochiropteran systematics and represents only the 3rd independent cladistic assessment of all families within traditional Yangochiroptera. Relationships suggested in Rag2 tree (Fig. 1) affirm some traditional views for an association of Mormoopidae, Noctilionidae, and Phyllostomidae and of Molossidae and Vespertilionidae (Smith 1976) but disagree strongly with proposed arrangements for Furipteridae, Myzopodidae, Natalidae, and Thyropteridae (Simmons 1998; Smith 1976; Van Valen 1979). At the same time, relationships suggested by Rag2 are congruent with those suggested by our analysis of mtDNA. Combined analysis afforded much greater support for almost all relationships (Fig. 2) than did analysis of individual data sets, suggesting that data concatenation helped to amplify phylogenetic signal that was masked by homoplasy in individual data sets (Teeling et al. 2000). In addition to sampling more characters from other nuclear genes, it is likely that sampling more taxa for Emballonuridae and Rhinopomatidae would help resolve the 2 clades not supported in our combined tree.

No relationships were contradictory between our separate and combined analyses, but exact position of Mystacinidae was conflicted between our analysis and that of Van Den Bussche and Hoofer (2001). Their results supported Mystacinidae internal within Noctilionoidea, whereas our results supported a basal position within Noctilionoidea. However, support for the internal position seems to have come only from characters that are questionable regarding positional homology and that should be excluded from phylogenetic analysis. We have become increasingly more conservative in our evaluation of positional homology within rDNA alignments, especially for alignments that include several highly divergent taxa. We excluded 599 ambiguous characters corresponding to both stem and loop regions of both ribosomal genes. Van Den Bussche and Hoofer (2001) excluded only 272 characters in the 12S rRNA, corresponding to hypervariable regions identified by Springer (1997). They did not report this, however, primarily because there was no difference between analyses with and without these characters. Thus, we excluded >350 more characters in our analysis of mtDNA, which apparently were the characters supporting an internal, rather than basal, position within Noctilionoidea. A basal position for Mystacinidae is congruent with Rag2 data. The fact that we excluded many characters also may help to explain why analysis using either alignment (15.00:6.66 versus 5:4 cost ratios) yielded identical results. That is, even though the 5:4 alignment was 80 sites longer, corresponding to insertion of several more gaps in variable loop regions, those entire regions were ambiguous and consequently excluded from both alignments.

Taxonomic and Zoogeographic implications.—Our study strongly supports inclusion of Furipteridae, Mystacinidae, and Thyropteridae within Noctilionoidea, which differs somewhat from past studies (Pierson 1986; Simmons 1998; Simmons and Geisler 1998; Smith 1976). A close association between Mormoopidae, Noctilionidae, and Phyllostomidae has long been recognized (Miller 1907; Smith 1972), and an association between Mystacinidae and traditional noctilionoids has been affirmed repeatedly by analysis of molecular (Kennedy et al. 1999; Kirsch et al. 1998; Pierson 1986; Van Den Bussche and Hoofer 2000, 2001) and morphologic (Simmons and Conway 2001) data. Despite congruency among studies, there has been no consensus on specific branching order between Mystacinidae and other families. Mystacinidae (Kirsch et al. 1998), Noctilionidae (Van Den Bussche and Hoofer 2000, 2001), and Phyllostomidae (Pierson 1986) all have been proposed as the basal lineage to this group of bats, although only 1 of those studies examined all families within traditional Yangochiroptera. Van Den Bussche and Hoofer (2001) proposal, however, apparently was incorrect based on a more conservative assessment of positional homology within their data set. Our study of Rag2 and a slightly different set of mtDNA suggest a basal position for Mystacinidae within Noctilionoidea. Specific branching order of other noctilionoid families (namely, Thyropteridae) remains unresolved, however, and future study of independent data is needed to test our hypothesis for Noctilionoidea and to help further resolve relationships among these families.

Furipteridae and Thyropteridae typically are considered related closely to each other and to molossids and vespertilionids based on heavily weighted consideration of characters of the wing and shoulder joint (Koopman 1984; Miller 1907; Smith 1976). Thus, our gene trees suggest that transformation of the chiropteran wing and shoulder has had a different history than previously thought. Furthermore, origins and current distributions of Furipteridae and Thyropteridae in the New World are more easily explained within the framework of our phylogeny than it is within those of previous phylogenies. It has long been suggested that traditional noctilionoids (Mormoopidae, Noctilionidae, and Phyllostomidae) originated in the New World, perhaps from some emballonurid-like stock (Smith 1972), because all 3 families always have been endemic there, appearing 1st as Late Paleogene and Miogene fossils (Czaplewski and Morgan 2002; Koopman 1970; Morgan and Czaplewski 2002). A shared recent ancestry of Furipteridae, Thyropteridae, and traditional noctilionoids presents no challenge to this hypothesis and in fact strengthens the hypothesis in terms of optimization of current distributions on our gene trees. Of 6 noctilionoid families, Mystacinidae is the only Old World member. Both dispersal (Pierson 1986; Simmons 1998) and vicariance (Kirsch et al. 1998) have been used to explain the distribution of Mystacinidae in New Zealand. We cannot exclude either of these hypotheses, but again, our results strongly corroborate association between these New and Old World bats.

Our study also strongly supports monophyly of Natalidae, Molossidae, and Vespertilionidae ( = Vespertilionoidea). Morphologic data traditionally have placed Natalidae nearest Furipteridae and Thyropteridae, but they also have always associated Natalidae with Molossidae and Vespertilionidae (Koopman 1984; Koopman and Jones 1970; Miller 1907; Smith 1976). Recent molecular data also affirm association between Natalidae, Molossidae, and Vespertilionidae (Teeling et al. 2002). Our study offers no new insight into the origin of these 3 families, primarily due to complex worldwide distributions of molossid and vespertilionid lineages. A possibility supported by our study is that Vespertilionoidea originated in the New World because Natalidae, endemic to the New World, represents the basal vespertilionoid lineage in our trees. However, our study does not adequately distinguish whether these bats originated in the New World and subsequently invaded the Old World 1 or several times (and perhaps back to the New World) or vice versa. Obviously, more thorough taxonomic sampling, especially of vespertilionid lineages, is needed to address such questions.

Support for traditional placement of Myzopodidae (specifically with Vespertilionidae) has been based almost exclusively on resemblance in derived conditions of humerus and shoulder joint (Koopman 1984; Miller 1907; Smith 1976), even though other parts of the myzopodid skeleton apparently do not exhibit such derived conditions. Miller (1907):194) wrote, “The skeleton is, as pointed out by Thomas, remarkable for its lack of special modifications.” Separate and combined analysis of Rag2 and mtDNA contradict traditional association between Myzopodidae and other noctilionoid or vespertilionoid families, suggesting a basal position such that Emballonuridae (a traditional yinochiropteran— Koopman 1994) is allied to other yangochiropterans more closely than is Myzopodidae. Despite congruence between Rag2 and mtDNA data and high statistical support from combined analysis, our inferences regarding position of Myzopodidae are qualified by taxonomic sampling. Specifically, our meager sampling of “non-yangochiropteran” families cannot distinguish with confidence whether Myzopodidae is the basal lineage of traditional Yangochiroptera or has closer affinities with other (unsampled) families. Better taxonomic sampling of families within traditional Yinochiroptera (emballonuroids and rhinolophoids) will be necessary for meaningful resolution to position of Myzopodidae (and Emballonuridae). Our intention in this study essentially was to duplicate Van Den Bussche and Hoofer (2001) experimental design, which sampled only a few outgroup taxa representing families outside traditional Yangochiroptera, thereby facilitating a comparative framework for the 2 studies (and Simmons and Geisler 1998) and conditional combination of molecular data. Despite these qualifications, our study provides strong support for monophyly of both Noctilionoidea and Vespertilionoidea (as defined herein) to the exclusion of Myzopodidae.

It always has been difficult to speculate on Zoogeographic history of Myzopodidae, given its limited distribution in Madagascar (and Africa, historically—Hill and Smith 1984) and the long-held inference of shared common ancestry with more derived and largely New World Vespertilionoidea (sensu Koopman 1993) or Nataloidea (sensu Simmons 1998). Even without full resolution to Myzopodidae relative to other (unsampled) families, our data clearly exclude the hypothesis that Myzopodidae was part of a New World radiation of bats (i.e., Nataloidea), suggesting, more intuitively, an Old World origin.

Summary.Smith (1976), (1980) summarized classifications from the 17th century through the 1970s and produced a phylogeny representing what at that time was the “generally accepted view” of extant bat families (Smith 1976:56). This view was modified only slightly in subsequent years, most notably by Koopman (1984), (1993), (1994), and has served as traditional classification for bats in recent decades. Although primitive versus derived similarity were considered by Koopman (1984), (1993), (1994) and Smith (1976), (1980), the traditional view of bat relationships was not conceived on the basis of explicit phylogenetic analysis of discrete characters. Recent study of a variety of characters, both morphologic and molecular, that has used explicit character analysis has been largely incongruent with traditional classification of bats (Simmons and Geisler 1998; Van Den Bussche and Hoofer 2000).

This study represents the 3rd cladistic assessment of all traditional noctilionoid and vespertilionoid families. Relationships suggested by Rag2 and mtDNA data are congruent, suggesting that there is sufficient signal for recovery of organismal phylogeny in both genomes. However, both molecular data sets contradict morphologic data (Simmons and Geisler 1998), essentially due to morphologic support for Nataloidea (Furipteridae, Myzopodidae, Natalidae, and Thyropteridae). All other morphologic relationships are either unresolved (based on statistical support) or compatible with our molecular data (Simmons and Geisler 1998). Further study is necessary to assess whether the contradiction may be caused by differences in taxonomic sampling between molecular and morphologic data sets or “incompleteness” of the morphologic data set (Simmons 1998; Simmons and Geisler 1998). As molecular data become available for more taxa, we expect better (supported) resolution for relationships among Myzopodidae, Noctilionoidea, and Vespertilionoidea as well as all other chiropteran families.


We thank R. J. Baker of the Natural Sciences Research Laboratory, the Museum of Texas Tech University, N. B. Simmons of the American Museum of Natural History, M. D. Engstrom of the Royal Ontario Museum, J. A. W. Kirsch of the University of Wisconsin Zoological Museum, J. L. Patton of the Museum of Vertebrate Zoology, Berkeley, and K. McBee of the Oklahoma State University Collection of Vertebrates for generously loaning tissues for this study. We thank R. DeBry, A. A. Echelle, G. Naylor, C. Simon, and 2 anonymous reviewers for comments and suggestions that improved the manuscript and R. Fonseca and S. Solari for kindly providing the Spanish resumen. Thanks also go to the Oklahoma Cooperative Fish and Wildlife Research Unit and personnel of the Oklahoma State University Recombinant DNA/ Protein Resource Facility for synthesis and purification of synthetic oligonucleotides. This study was supported by National Science Foundation grant DEB-9873657 and a REU supplement to R. A. Van Den Bussche and by Grantsin-Aid of Research from the American Society of Mammalogists, Sigma Xi, and the Theodore Roosevelt Memorial Fund, American Museum of Natural History, to S. R. Hoofer.

Appendix I

Specimens examined.—All tissues used in this study are represented by voucher specimens deposited in the American Museum of Natural History (AMNH), Oklahoma State University Collection of Vertebrates (OK), Royal Ontario Museum (ROM), Natural Sciences Research Laboratory of the Museum of Texas Tech University (TK), or University of Wisconsin Zoological Museum (UWZM). Tissue and DNA samples for Myzopoda aurita have been accessioned in the Oklahoma State University Collection of Vertebrates and Museum of Vertebrate Zoology, Berkeley, but the voucher specimen was accessioned in the United States National Museum (USNM 448885).

We generated Rag2 sequence data for 21 bats: Emballonuridae (Saccopteryx bilineata AMNH 267842), Furipteridae (Furipterus horrens ROM 100202), Molossidae (Molossus molos sus AMNH 269105, Nyctinomops macrons TK 78908, and Tadarida brasiliensis OK 430), Mormoopidae (Mormoops megalophylla TK 78661), Mystacinidae (Mystacina tuberculata UWZM—M27027), Myzopodidae (Myzopoda aurita OK 4246); Natalidae (Natalus micropus TK 19454 and N. stramineus TK 15660), Pteropodidae (Pteropus hypomelanus TK 20225), Rhinopomatidae (Rhinopoma hardwickei TK 40884), Thyropteridae (Thyroptera discifera TK 17210 and T. tricolor AMNH 268577), and Vespertilionidae (Corynorhinus townsendii TK 83182, Eptesicus furinalis AMNH 268583, Harpiocephalus harpia TK 21258, Myotis riparius AMNH 268591, Myotis velifer TK 79170, Kerivoula hardwickei ROM 110829, and K. papillosa ROM 110850).

Rag2 sequences for 9 other bats were obtained from GenBank (accession numbers indicated parenthetically): Emballonuridae (Diclidurus albus AF316446), Mormoopidae (Pteronotus davyi AF316482), Noctilionidae (Noctilio albiventris AF316476; Noctilio leporinus AF316477), and Phyllostomidae (Ariteus flavescens AF316435; Centurio senex AF316438; Desmodus rotundus AF316444; Macrotus waterhousii AF316461; Tonatia brasiliensis AF316489). Baker et al. (2000) generated all Rag2 sequences that we obtained from GenBank, and vouchers for each can be cross-referenced in their “specimens examined” (pp. 4–5). We also generated mtDNA sequence data (12S rRNA, tRNAVal, 16S rRNA) for the emballonurid D. scutatus (AMNH 267832).


  • Associate Editor was William L. Gannon.

Literature Cited

View Abstract