Harpy fruit bats, two closely related species in the genus Harpyionycteris (Chiroptera: Pteropodidae), exhibit a suite of unique craniodental traits. For this reason, the affinities of these bats have remained unclear, and most systematists have placed them in a group of their own (Harpyionycterinae Miller, 1907). The multicuspidate pattern of the cheek teeth in Harpyionycteris has generated speculation that it may represent an ancestral tribosphenic pattern lost in other pteropodids. In this contribution we propose a phylogenetic placement of Harpyionycteris based on parsimony analysis of complete sequences from two coding genes, the nuclear vWF (exon 28) and the mitochondrial cytochrome b (cyt-b). Both datasets, independently and in combination, strongly support a close relationship between Harpyionycteris and Dobsonia, as originally proposed by Andersen (1912, Catalogue of Chiroptera, British Museum Trustees). In turn, this group nests deeply inside Pteropodidae but it is not closely related to any particular suprageneric clade. Based on other data, we postulate that Aproteles also belongs in this group and therefore propose the expansion of Harpyionycterinae to include Harpyionycteris, Aproteles, and Dobsonia. Regarding the dentition, our results strongly reject the tribosphenic hypothesis advanced by some authors. The multicuspidate cheek tooth pattern seen in Harpyionycteris appears uniquely derived and related to specialized feeding habits and it thus has no direct bearing on the evolution of the typical pteropodid dentition from the tribosphenic pattern of microchiropterans and other mammals.
Introduction
The systematic position of the Harpy fruit bats (Harpyionycteris Thomas, 1896) has been controversial since the original description of the genus. Thomas (1896: 243) described this genus for a single species, H. whiteheadi, from the Philippines. Although he associated it with rousettines, to him Harpyionycteris was “one of the most isolated of all genera” of Megachiroptera. This assessment was based on craniodental characters exhibited by Harpyionycteris, including uniquely multicuspidate cheek teeth (figs. 1, 2). The unusual nature of the skull and dentition subsequently prompted Miller (1907) to place this genus in a separate subfamily, Harpyionycterinae. In his influential monograph on the Megachiroptera, Andersen (1912: 800) followed Miller in recognizing Harpyionycterinae, emphasizing its highly distinct characters:
Figure 1
Harpyionycteris whiteheadi FMNH 142766, view of the upper (A) and lower (B) tooth rows. Abbreviations: c, lower canine; C, upper canine; I2, second upper incisor; m1, first lower molar; M1, first upper molar; m2, second lower molar; M2, second upper molar; m3, third lower molar; p1, first lower premolar; P1, first upper premolar; p3, third lower premolar; P3, third upper premolar; p4, fourth lower premolar; P4, fourth upper premolar. Homology of teeth according to Andersen (1912). Scale = 5 mm.
Figure 2
Harpyionycteris celebensis AMNH 196437, rostral view of the apex of the skull and mental view of the mandible. Abbreviations: c, lower canine; C, upper canine; i2, lower second incisor; I2, upper second incisor. Homology of teeth according to Andersen (1912). Scale = 1 mm.
Each of the following single characters appears to be absolutely diagnostic: (1) premaxillae, upper incisors, and upper and lower canines strongly proclivous (canines crossing each other at nearly right angles when the jaw is closed); (2) lower canines tricuspidate; (3) molariform teeth multicuspidate; (4) tibia less than one-third of forearm (only slightly longer than foot with claws).
However, Andersen (1912) also explicitly linked Harpyionycteris with Dobsonia and discussed at length both the characters supporting this perceived association and the features by which Harpyionycteris appeared more derived with respect to Dobsonia. Several years later, Miller and Hollister (1921) described a second species from Sulawesi, Harpyionycteris celebensis. Tate (1951: 4) examined new specimens from both the Philippines and Sulawesi and recognized the two taxa as distinct but he left open the question of whether they deserve the status of species or subspecies, noting that they differ by “extremely slight characters”. Subsequent authors have either treated celebensis as a subspecies of whiteheadi (e.g., Laurie and Hill, 1954; Koopman, 1994) or more commonly as a separate species (e.g., Bergmans and Rozendaal, 1988; Corbet and Hill, 1992; Hill,1983; Koopman, 1993; Peterson and Fenton,1970; Simmons, 2005), although often with some reservations.
Tate (1951) disregarded the association of Harpyionycteris and Dobsonia proposed by Andersen (1912), and he considered the dental characters of Harpyionycteris to be derived relative to those of other megachiropterans. By contrast, Slaughter (1970) suggested that Harpyionycteris represents a prototype from which all megachiropteran dentitions can be derived by simplification of the cusp pattern. To Slaughter (1970), the multicuspidate condition seen in Harpyionycteris could be interpreted as a retention of a tribosphenic pattern seen in microchiropterans and other mammals. This point of view was rejected by Hill and Beckon (1978), who supported the hypothesis advanced by Andersen (1912) regarding the origin of the multicuspidate cheek teeth of Harpyionycteris. The typical cusp pattern of a postcanine tooth in Megachiroptera consists of a median groove that separates two ridges, medial and lateral, each of which tapers to a point rostrally to form medial and lateral main cusps. In several megachiropterans (e.g., in Pteralopex and Pteropus species of the pselaphon species group; Andersen, 1912) the lateral and medial ridges are each subdivided into more than one cusp, although the main rostral cusp remains clearly distinct. To Andersen (1912) and Hill and Beckon (1978), the multicuspidate condition of Harpyionycteris is the result of extreme subdivision of the lateral and medial ridges, and reduction of the main cusp so that all cusps are essentially subequal. That is, the multicuspidate condition of the cheek teeth, as well as other dental traits in Harpyionycteris, are considered derived.
Authors proposing formal classifications of megachiropterans have generally accepted Miller's (1907) segregation of Harpyionycteris in a suprageneric taxon of its own. Koopman and Jones (1970) placed Harpyionycterini within Pteropodinae, as did McKenna and Bell (1997) and Koopman (1994), while Corbet and Hill (1992) and Bergmans (1997) recognized Harpyionycterinae Miller, 1907 as distinct at the subfamily level. Moreover, von Schultz (1970) elevated the taxon to family level on the basis of characters of the digestive tract.
Harpyionycteris has been included in two previous phylogenetic studies. Springer et al. (1995) and Romagnoli and Springer (2000) used groupings derived from molecular data to constrain searches based on a morphological matrix of 36 characters scored for 36 genera (Harpyionycteris was scored for the morphological data only). Both studies recovered Harpyionycteris as sister to Dobsonia within a larger clade of pteropodine bats, thus supporting Andersen's (1912) treatment. However, this result was not unexpected because the morphological matrix used by Springer et al. (1995) was based entirely on Andersen's (1912) key. The small number of characters, low character-to-taxon ratio, and the uncertainty that inevitably results from use of supraspecific terminals limited the resolving power and character support in Springer et al.'s (1995) study. The problem of placing Harpyionycteris in the megachiropteran tree clearly requires further study. Understanding the phylogenetic position of Harpyionycteris will also clarify attendant problems of character evolution in this form, particularly the derived versus primitive condition of its dentition, which is the most distinct found in megabats.
In the present study, we report the results of a parsimony analysis of one nuclear and one mitochondrial gene sampled in Philippine Harpyionycteris and representatives of all other major clades of megabats. Our results support Andersen's (1912) vision of relationships and indicate the need for an important taxonomic rearrangement of several megachiropteran genera to accommodate the recovered relationships. The polarity of the complex dental features seen in Harpyionycteris is also discussed.
Methods
Taxa
We included members of each major clade (subfamily or tribe) of Megachiroptera in our analyses in order to test the proposed affinities of Harpyionycteris. Based on the latest estimate of megabat higher level relationships (Giannini and Simmons, 2005), we included two Nyctimeninae (Nyctimene albiventer and N. vizcaccia), two Cynopterinae (Cynopterus sphinx and Ptenochirus jagori), three Dobsoniini (Dobsonia minor, D. moluccensis, and D. inermis), two Pteropodini (Pteropus hypomelanus and P. tonganus), two Rousettini (Rousettus aegyptiacus and R. amplexicaudatus), two Myonycterini (Myonycteris torquata and Megaloglossus woermanni), two Epomophorini (Epomophorus wahlbergi and Epomops franqueti), and two macroglossine megabats, Melonycteris fardoulisi and Macroglossus minimus. Two individuals of Harpyionycteris whiteheadi from Mindanao, Philippines, were included in our analyses (see appendix 1). To root our megabat subtree, and therefore to allow for testing character polarity within megabats, we included several microbats as outgroups. Specifically, we included two yangochiropterans (Artibeus jamaicensis and Plecotus auritus) and two yinochiropterans (Rhinopoma hardwickei and Rhinolophus creaghi), which were selected on the basis of sequence availability (see appendix 1). An additional analysis was done with Rhinolophus pumilus in place of R. creaghi (see below).
The ingroup taxonomic sampling, though not intended to cover the full diversity of megabats, includes most key suprageneric megabat clades, and certainly includes all relevant groups with which Harpyionycteris has variously been associated throughout its systematic history. Three hypotheses can be tested using our taxonomic sample: segregation of Harpyionycteris in a group of its own (sustained by most authors, including Thomas, 1896; Miller, 1907; Tate, 1951; Koopman and Jones, 1970; Koopman, 1994; Bergmans, 1997); association of Harpyionycteris with rousettines (Thomas, 1896); and association of Harypionycteris with dobsoniines (Andersen, 1912; Springer et al., 1995). In particular, the first hypothesis would be rejected if Harpyionycteris is found to nest within any of the suprageneric groups listed above.
Sequences
We used newly generated sequences as well as published sequences. Total DNA was obtained from preserved tissue samples (see voucher list in appendix 1) with the DNeasy tissue kit (QIAGEN). Sequences of two coding genes, one nuclear (1231 bp of exon 28 of the von Willerbrand factor gene) and one mitochondrial (1140 bp of the cyt-b), were generated using standard methods. PCR amplification was carried out using previously published primers (vWF: Porter et al., 1996; cytochrome b: Bastian et al., 2002). In order to obtain both forward and reverse sequences for each gene region, internal primers were used for sequencing in addition to the PCR primers (sequences available upon request). All sequences were obtained with an automated ABI 3730XL sequencer. Sequence editing and prealignment were done with the Sequencher 4.2 software (Gene Codes).
Accession numbers for the new sequences are DQ445684-714. Voucher information is provided in appendix 1. In addition, we used published sequences of Artibeus jamaicensis (AF447542, NC_002009), Cynopterus sphinx (AY629004), Dobsonia moluccensis (U31209, AF144064), Epomophorus wahlbergi (AF044642), Epomops franqueti (AF044639), Myonycteris torquata (AF044650), Nyctimene albiventer (AF447549, DQ14264), Plecotus auritus (AB079840, AB085734), Pteropus hypomelanus (AF203777, AB062472), Pteropus tonganus (AF044656), Rhinopoma hardwickei (AF447551, AY056462), Rhinolophus creaghi (AF447546, DQ178986), Rhinolophus pumilus (NC_005434), and Rousettus amplexicaudatus (AY057836, AB046329). Although for some species more than one sample was available, only one was included in the final analysis. The aligned dataset is provided at ftp://ftp.amnh.org/pub/group/mammalogy/downloads/.
Phylogenetic Analyses
Aligned sequences of the vWF and cyt-b genes were submitted to parsimony analysis both individually and in combination. In the cyt-b analysis, the outgroup Rhinolophus creaghi was identified as the cause of a collapse in the backbone of the tree due to its incomplete sequence; as a consequence, it was replaced by the complete sequence of its close relative R. pumilus and the problem was eliminated. However, sequences of the vWF are not available for R. pumilus, so in the combined analysis we used the complete vWF and the partial cyt-b sequences of Rhinolophus creaghi, which did not affect resolution. The tree search strategy consisted of 500 replicates of random addition sequences of taxa each followed by tree bisection reconnection branch swapping (TBR). An additional round of TBR was done on all optimal trees obtained. Clade support was estimated using both Bremer values (Bremer, 1994) and character resampling (Goloboff et al., 2003). For the first method, we followed Giannini and Bertelli (2004). Specifically, we implemented an incremental strategy for obtaining suboptimal trees in successive stages, saving up to 2000 suboptimals in each stage. We first searched for suboptimal trees 1 step longer than the optimal tree length, next saving suboptimals up to 2, 3, 4, 5, 6, 7, 8, 9, and 10 steps longer than the optimal trees. Second, the resampling technique used was a version of the jackknife developed by Goloboff et al. (2003). Group frequency (based on unbiased symmetric resampling) was calculated on the basis of 5000 replications. All analyses were executed in TNT (Goloboff et al., 2004).
Results
The parsimony analysis using the vWF gene resulted in two optimal trees of 688 steps (CI = 56, RI = 0.61). Bremer values were calculated using a sample of 8932 trees up to 10 steps longer than the optimal trees sampled incrementally (see Methods: Phylofgenetic Analyses). In the strict consensus topology (fig. 3), outgroup relationships were as expected (yinochiropteran terminals grouped together, as did yangochiropterans), and Megachiroptera was recovered as a monophyletic group with high support (Bremer support, BS > 10; jackknife frequency, JF = 100). Within Megachiroptera, successive sister groups were Nyctimeninae, a macroglossine clade formed by Melonycteris and Macroglossus, a Pteropus clade sister to Harpyionycteris + Dobsonia, Cynopterinae, and “clade C” of rousettines, myonycterines, and epomophorines found in previous analyses by Giannini and Simmons (2003, 2005). The latter was composed of Eonycteris, Rousettus, Myonycterini (Myonycteris + Megaloglossus), and Epomophorini (Epomops + Epomophorus). Support values along the backbone of the megabat subtree (linking subfamilies or tribes) were minimal (BS = 1, JF < 50) except for clade C (BS = 4, JF = 84), whereas most of the suprageneric groups were reasonably supported. A sister-group relationship between Harpyionycteris and Dobsonia was particularly well supported (BS = 7, JF = 97), as was the monophyly of Dobsonia (BS > 10, JF = 100).
Figure 3
Results of parsimony analysis using the vWF gene. ). Strict consensus of two trees of 688 steps. Numbers above branches are Bremer support values. Numbers below branches are jackknife frequencies (cutoff value = 50%
Parsimony analysis of the cyt-b gene resulted in a single optimal tree (fig. 4) of 1916 steps (CI = 0.42, RI = 0.35). Bremer values were calculated from an incremental sample of 10,670 suboptimal trees. This analysis recovered Megachiroptera as monophyletic but the support level was low (BS = 2, JF < 50). Successive sister groups were Macroglossus, Nyctimene + Melonycteris, Pteropodini + Cynopterinae, Harpyionycteris + Dobsonia, and the clade C. Support values were minimal along the backbone (BS = 1, JF < 50) and slightly higher in the suprageneric groups. Harpyionycteris + Dobsonia was recovered but support was minimal (BS = 1, JF < 50).
Figure 4
Results of parsimony analyses using the cyt-b gene. Single optimal tree of 1916 steps. Numbers above branches are Bremer support values. Numbers below branches are jackknife frequencies (cutoff value = 50%).
The combined analysis of vWF + cyt-b resulted in two trees (strict consensus in fig. 5) at 2507 steps (CI = 0.42, RI = 0.42). Using the same search strategy, just 519 suboptimal trees were found that were up to 10 steps longer than the optimal trees. Support in resolved backbone clades was low to moderately high (1 ≤ BS ≤ 6, 50 ≤ JF ≤ 81), as were all suprageneric groups recovered (3 ≤ BS ≤ 10 for most groups, but for some BS > 10, and most JF > 50). Successive sister groups in the optimal tree were Nyctimeninae, a macroglossine clade formed by Melonycteris and Macroglossus, and a trichotomy formed by Pteropodini, Cynopterinae, and a clade containing the remaining groups. The last clade included Harpyionycteris + Dobsonia, and the clade C. The clade formed by Harpyionycteris + Dobsonia was highly supported (BS > 10, JF = 89).
Discussion
Results based on both separate and combined analyses of our two genes strongly suggest a close phylogenetic relationship between Harpyionycteris and Dobsonia, as initially suggested by Andersen (1912) and later recovered by Springer et al. (1995) based on morphological evidence. The vWF analysis also suggests, albeit weakly, that Dobsonia + Harpyionycteris may be linked to pteropodines as found by Springer et al. (1995). However, the cyt-b signal recovered a clade formed by pteropodines + cynopterines. In the combined analysis, the conflicting signals caused a polytomy involving pteropodines and cynopterines. Harpyionycteris + Dobsonia formed a separate group in the next resolved node up in the tree, sister to clade C (fig. 5). Giannini and Simmons (2005) found that dobsoniine bats (Aproteles and Dobsonia) form a clade not closely associated with any particular suprageneric clade (Giannini and Simmons, 2005: fig. 7).
Andersen (1912) established a solid argument for the relationship between Harpyionycteris and the other megachiropterans known to him, specifically highlighting a phylogenetic link to Dobsonia. He did so on the basis of putatively derived traits shared between Dobsonia and Harpyionycteris, without losing sight of the overwhelmingly distinct craniodental features that make Harpyionycteris so easily diagnosable. For instance, he (1913: 804) noted that “If the extremity of the rostrum and mandible of the Harpyionycteris skull were removed from view, it would be difficult to point out any cranial character of generic importance to separate it from Dobsonia …”.
Andersen (1912) identified other taxonomically important similarities between the two genera, including: form of the braincase and postdental palate, position of all cranial foramina, deflection of basicranial axis, similar dental formulae (first upper premolar lost in Dobsonia), loss of the first upper and lower incisor, and left and right lower canines displaced rostromedially to nearly contacting each other. As expected, a phylogenetic analysis based on some of Andersen's characters recovered this association (Springer et al., 1995). Based on one nuclear and one mitochondrial gene, we independently recovered a well-supported sister group relationship between Harpyionycteris and Dobsonia.
We further postulate that Aproteles is also a member of the Harpyionycteris + Dobsonia group. In the original description of Aproteles, Menzies (1977) recognized its cranial and external resemblance to Dobsonia. Moreover, a close association of Aproteles and Dobsonia has been supported in all previous phylogenetic analyses in which Aproteles was included (Colgan and da Costa, 2002; Giannini and Simmons, 2003, 2005; Jones et al., 2002; Kirsch et al., 1995). In the present study we did not have access to tissue samples of Aproteles, and the genes we sequenced were not previously reported for this form, so we cannot resolve relationships among the three genera on the basis of our current molecular data. However, Aproteles undoubtely shares more derived morphological traits with Dobsonia (most notably, the bare back) than with Harpyionycteris. We predict that future studies including all three genera will demonstrate a sister-group relationship between Dobsonia and Aproteles, with Harpyionycteris as sister to that clade.
The recognition that Harpyionycteris, Dobsonia, and Aproteles belong in a monophyletic group to the exclusion of other genera requires changes in the higher level classification of megabats. We recommend that Harpyionycterinae Miller, 1907 be expanded to include Harpyionycteris Thomas, 1896, Dobsonia Palmer, 1898, and Aproteles Menzies, 1977. This arrangement differs from Bergmans (1997) in that the tribe Dobsoniini Andersen, 1912, formed by Dobsonia and Aproteles, is transferred from Rousettinae Andersen, 1912 to Harpyionycterinae, the oldest name available for this new group. Maintaining Dobsoniini as a tribe within Harpyionycterinae requires recognizing the tribe Harpyionycterini, which seems superfluous for such a small group.
Members of the newly defined Harpyionycterinae are distributed in the center of the Australasian region. Andersen (1912: 803) made a strong biogeographical statement based on his association of Harpyionycteris with Dobsonia: “So evident is the intimate phylogenetic connection between these two genera that Harpyionycteris may be said, almost with certainty, to be the peculiarly modified Philippine representative of the Austro-Malayan Dobsonia.”
The subsequent discovery of Harpyionycteris in Sulawesi (Miller and Hollister, 1921) substantiated the biogeographical link between the two genera. While Harpyionycteris is a Philippine/Wallacean genus with its two species separated by the Wallace's line, Dobsonia is essentially Wallacean/Papuan. However, 3 of the 14 currently recognized species of Dobsonia (Simmons, 2005) narrowly escape the boundaries of these two regions combined (Dobsonia chapmani in the Philippines, D. peronii in Bali, and D. magna in Cape York Peninsula, Australia). As far as is known, the recently discovered Aproteles is Papuan. Overlap exists in the distribution of the three harpyionycterine genera, but it is limited to few instances involving few species at a time. Aproteles and two species of Dobsonia (D. magna and D. minor) are sympatric in Papua New Guinea, but probably D. minor and A. bulmerae are not syntopic as they differ in their elevational range; that is, they are lowland and highland species, respectively, whereas D. magna is widespread (Flannery, 1995). Harpy fruit bats formerly assigned to Harpyionycteris whiteheadi negrosensis Peterson and Fenton, 1970 (not currently recognized as a separate subspecies; Simmons, 2005) and D. chapmani both occur in Negros Island (the Philippines). Harpyionycteris celebensis coexists with three species of Dobsonia in Sulawesi (D. crenulata, D. exoleta, and D. minor).
The relationships recovered in our analyses imply that the unusual multicuspidate molars of Harpyionycteris cannot possibly represent the retention of an ancestral tribosphenic pattern. Therefore, the most plausible explanation for the cusp pattern of Harpyionycteris is the one proposed by Andersen (1912): the multiple cusps of Harpyionycteris are derived from repeated subdivisions of the simple cusp pattern typical of pteropodids (fig. 1). The strong dentition of Harpyionycteris might represent a dietary specialization not shared with other megachiropterans. Before anything was known about the diet of Harpyionycteris, Peterson and Fenton (1970: 8) speculated that “it is a fruit eater, perhaps adapted for a particular type of tough-textured fruit for which the multicuspid teeth would be advantageous in extracting the juice”. This hypothesis has been confirmed to a large extent, at least in the Philippines. Harpyionycteris principally occurs in mid-elevation forests (Heaney et al., 1989, 1998) were it feeds on the hard-seeded, tough-skinned fruits with fibrous mesocarp of a locally abundant vine (Freycinetia, Pandanaceae; Heaney et al., 1998, 1999; Utzurrum, 1995). Although Harpyionycteris also feeds on Ficus as do all other Philippine megachiropterans, no other fruit bat appears to consume Freycinetia fruits (Utzurrum, 1995).
Acknowledgments
We thank Larry Heaney (Field Museum of Natural History, Chicago) and Jim Patton and Carla Cicero (Museum of Vertebrate Zoology, Berkeley) for access to tissue samples and specimens that made possible this contribution. We also thank Natalee Stephens for help in the laboratory and Paul Sweet, Esteban Sarmiento, Eric Stiner, and Darrin Lunde for tissue samples. Funding for this report was provided by the National Science Foundation (research grant DEB-9873663 to N.B.S.), Coleman and Vernay postdoctoral fellowships at the AMNH to N.P.G., and Henry MacCracken doctoral fellowship at New York University to F.C.A.
References
Appendices
Appendix 1
Abbreviations of Institutions: AMNH, American Museum of Natural History, New York; FMNH, Field Museum of Natural History, Chicago; AMCC, Ambrose Monell Cryo Collection (AMNH); MVZ, Museum of Vertebrate Zoology, University of California, Berkeley. Other abbreviations refer to collector's catalog.
