Old World fruit bats (Chiroptera: Yinpterochiroptera: Pteropodidae) are a diverse radiation endemic to the tropics of Africa, Asia, Australia, and nearby island archipelagos. Recent molecular analyses have provided considerable resolution of phylogenetic relationships within this group, but many points of uncertainty have remained including the position of several enigmatic taxa (e.g., Notopteris, Eidolon), relationships among species in more diverse subfamilies and genera (e.g., Pteropodinae, Pteropus, Epomophorus), and topology of the backbone of the tree. Here we provide a new, synthetic analysis including representatives of all 45 currently recognized genera and enhanced sampling in several speciose genera. Our matrix included four nuclear genes regions (vWF, RAG1, RAG2, and BRCA1) and four mitochondrial gene loci (Cytb, tRNA valine, 12S rRNA, and 16S rRNA) for a total of >8000 bp including new sequence data for 13 species. Maximum likelihood and Bayesian analyses resulted in trees supporting recognition of six main suprageneric clades similar in content to those identified in our previous studies. We did not recover strong support for relationships among the main clades along the backbone of the tree, but identified many well-supported clades within all of the major groups. Based on these results, we propose a new classification for Pteropodidae comprising eight subfamilies and 14 tribes, and including several new and/or replacement higher-level taxonomic names for which we provide morphological diagnoses.
INTRODUCTION
Old World fruit bats, also known as megabats, comprise the Family Pteropodidae, the third largest bat family (Simmons and Cirranello, 2019). Pteropodidae includes 45 currently recognized genera and a minimum of 200 species, six of which are extinct beyond reasonable doubt (Giannini, 2019; Simmons and Cirranello, 2019). Megabats are found throughout Africa, the Asian tropics, Australia, and many islands of the Indian and Pacific oceans where they play important ecological roles in seed dispersal and pollination (Nowak, 1994). Although some pteropodids use echolocation (Yovel et al., 2011; Lee et al., 2016; Boonman et al., 2014), most rely on vision and olfaction to navigate and find plant food sources. Pteropodidae is also functionally diverse; for example, it exhibits the widest variation in body size seen in any family of bats, ranging from 14 g to more than 1000 g (Moyers-Arévalo et al., 2018). Relative gigantism observed in some pteropodid species has been associated with the evolutionary loss of echolocation abilities, which may have released selective constraints on body size (Moyers-Arévalo et al., 2018). The majority of pteropodids are frugivorous; however, some species feed exclusively or opportunistically on nectar, pollen, and/or floral parts, while others include leaves, bark, and possibly sap in their diets (Kunz and Pierson, 1994; Giannini, 2019). In the Old World tropics, some pteropodid species are thought to be reservoirs of zoonotic viruses, since evidence of infection from filoviruses (e.g., ebola), henipaviruses (e.g., hendra), and coronaviruses (e.g., SARS) has been detected in healthy animals (Leroy et al., 2005; Towner et al., 2007; Hayman et al., 2010; Halpin et al., 2011; Olival et al., 2013).
The monophyly of Pteropodidae is widely accepted and the group has been consistently recovered in numerous molecular phylogenetic studies independent of taxonomic sampling, analytic methods, and the genetic markers used (Teeling et al., 2002, 2005; Shi and Rabosky, 2015; Amador et al., 2018). Similarly, its sister position to the insectivorous echolocating superfamily Rhinolophoidea has been confirmed in numerous studies since Hutcheon et al. (1998) first suggested “microbat” paraphyly due to “megabats” recovered as sister to rhinolophoids (e.g., Teeling et al., 2002, 2005; Shi and Rabosky, 2015; Amador et al., 2018). By contrast, some relationships within Pteropodidae remain contentious, although interesting results have accumulated over the past 20 years as a series of molecular phylogenetic studies have approached the problem of internal relationships of Pteropodidae with ever-increasing sampling of its diversity (e.g., Hollar and Springer, 1997; Juste et al., 1999; Álvarez et al., 1999; Giannini and Simmons, 2003, Almeida et al., 2011, Shi and Rabosky, 2015; Amador et al., 2018). These studies have uncovered many disagreements between topologies derived from molecular analyses and morphology-based classifications originally proposed by Andersen (1912) and significantly improved by Bergmans (1997; see also Giannini and Simmons, 2005). Some recent studies focused on particular groups and proposed relevant changes in the classification (Giannini et al., 2006, 2008, 2009; Nesi et al., 2013; Hassanin, 2014; Almeida et al., 2016). A new subfamily Eidolinae was erected to contain the rather enigmatic genus Eidolon (Almeida et al., 2016), and Boneia joined Dobsonia, Aproteles, and Harpyionycteris in a substantially expanded Harpyionycterinae (Giannini et al., 2006, 2009). Other studies made progress sorting out relationships within the more speciose subfamilies; e.g., Cynopterinae (Almeida et al., 2009) and Pteropodinae (Almeida et al., 2014). Finally, Rousettinae was redefined to accommodate the African endemic tribes Scotonycterini, Epomophorini, Stenonycterini, Myonycterini, and Plerotini, together with the widespread Rousettini (Almeida et al., 2016, Nesi et al., 2013).
Despite these recent advancements in phylogenetics and classification of Pteropodidae, some important relationships within the clade remain unresolved. This is the case with the problematic genus Notopteris Gray (1859), an endemic of Fiji, New Caledonia, and Vanuatu, which includes just two species that share unusual morphological characters, such as a long free tail (Andersen, 1912; Hill, 1983; Giannini and Simmons, 2005). Notopteris was included in the subfamily Pteropodinae by Bergmans (1997) but was recovered as sister to Nyctimene by Almeida et al. (2011) in an arrangement that nonetheless lacked statistical support. Although this result could reflect the large amount of missing data for this taxon in the matrix (only mitochondrial loci rRNA 12S and 16S), another possibility is that Notopteris might represent a relatively independent lineage without close affinities to any other pteropodid genus or subfamily (see Almeida et al., 2011).
Almeida et al. (2011) included 56 terminals representing 42 pteropodid genera. Their study, based on 8181 bp (including both nuclear and mitochondrial loci), recovered several clades with high statistical support, among them many of Bergmans' (1997) suprageneric groups. However, the analyses of Almeida et al. (2011) revealed an apparent polytomy at the base of the pteropodid tree. A careful inspection of the data showed that such a result could not be explained by bias or phylogenetic conflict between the eight loci represented in the matrix, and additional analyses suggested that some branches might indeed have zero length for the dataset used. As a conclusion, Almeida et al. (2011) hypothesized that crown pteropodids experienced an explosive radiation, which would account for the blurred phylogenetic signal at the base of their phylogeny. Several characteristics of the family are compatible with a scenario of explosive radiation, including high diversity in species richness and relative morphological disparity, the origin of key innovations related to the loss of echolocation, and acquisition of new feeding habits (Almeida et al., 2011). Demographic factors may also have played a role as many pteropodid species have insular distribution, especially the most speciose genus, Pteropus. While Almeida et al. (2011) included just a handful of Pteropus species, an important fraction of them were represented in another study that focused on pteropodines (Almeida et al., 2014). In that contribution, the authors reallocated many Pteropus species to mostly new species groups that differed considerably from previous arrangements that were all essentially traceable back to Andersen (1912).
Here we readdress the phylogenetics of Pteropodidae using a comprehensive species-level sampling, including representatives of all 45 currently recognized genera and sequence data from a total of 139 ingroup terminals. This taxonomic sample improves sequence sampling for several problematic taxa including Notopteris in addition to enhanced coverage for species in many other groups not included in previous simultaneous analyses. Our goal is to propose a new formal, fully revised and updated classification of the Pteropodidae that reflects the structure of monophyletic groups recovered, and also to resolve most of long-standing taxonomic issues (e.g., preoccupied names improperly used for over a century) in the Old World fruit bats.
MATERIAL AND METHODS
Taxonomic Sampling
Our matrix included a total of 145 species, 139 of which were pteropodids. The ingroup species represent all 45 currently recognized pteropodid genera. The majority of the ∼47 pteropodid species missing from the matrix belong to the highly diverse genera Nyctimene (a genus including ≥16 species), Dobsonia (≥14 species), and Pteropus (≥65 species; Simmons and Cirranello, 2019), which were represented in our matrix by six, eight, and 51 species, respectively. Our approach in this study was to use total evidence, including all species with available sequences, no matter the number of loci or length of the sequences available. The six outgroup species belong to the superfamily Rhinolophoidea, a group that has been consistently shown to be sister to Pteropodidae (e.g., Teeling et al., 2002, Amador et al., 2018; Lei and Dong, 2016). A list of species included in this study is provided in the supplementary material (table S1, available online at doi.org/10.5531/sd.sp.39).
Molecular Matrix
Our matrix included a total of eight loci. Four of those were nuclear gene regions: exon 28 of the von Willebrand factor gene (vWF, 1230 bp), partial recombination activating gene 1 (RAG1, 1084 bp), partial recombination activating gene 2 (RAG2, 760 bp), and partial breast cancer 1 gene (BRCA1, about 1370 bp). The other four loci were genes of the mitochondrial genome: complete cytochrome b gene (Cytb, 1140 bp), partial rRNA 12S gene (1069 bp), the valine tRNA gene, and partial rRNA 16S gene (1330 bp), totaling about 2550 bp. We obtained many of these sequences in previous studies (Giannini et al., 2006, 2008, 2009; Almeida et al., 2009, 2011, 2014, 2016), while sequences for 13 species were newly acquired for the current study using protocols and primers as previously published. Other sequences, however, were obtained by other authors, and so we downloaded these from GenBank. Accession numbers of each sequence used in this study are listed in supplementary table S2.
Multiple sequence alignments were obtained for each locus individually with the software MAFFT version 7 (Katoh and Standley, 2013). Exception was made for the three RNA genes, which were treated as a single continuous sequence due to their molecular contiguity in the mitochondrial genome. This particular alignment was treated with Gblocks for the removal of “gappy” regions that could have been misaligned (Castresana, 2000). Individual gene trees were then obtained and checked for unexpected results that could point to sample misidentification. The Almeida et al. (2011) study used the same set mitochondrial and nuclear loci sampled in a similar diversity of genera, and, through statistical tests, demonstrated that there is no incongruence of phylogenetic signal between this set of genes in this clade. We therefore proceeded with the concatenation of all six alignments into a single matrix with 7866 characters for the current study.
Phylogenetic Analysis
The best partition scheme for the data in our matrix was evaluated with the program PartitionFinder v. 1.1, which was prompted with the beginning and ending of each protein-coding gene separated into coding positions plus the fragment containing the mtRNA genes. The best scheme thus identified included nine partitions of the data, and these were later employed in both maximum likelihood and Bayesian inference tree searches. The former searches were done with RAxML v. 8 (Stamatakis, 2014), applying the GTRCAT substitution model to all partitions, which were treated separately for model-parameter estimation. The best tree was obtained from 100 independent runs. Statistical support for clades was assessed with 1000 nonparametric bootstrap replicas, with clade recovery percentages later plotted on the most likely tree. Bayesian inference searches were run on MrBayes (Ronquist and Huelsenbeck, 2003) employing for each partition the optimum substitution model according to PartitionFinder. Two MCMC were run for 15 million generations, with trees and parameters sampled every 5000 generations. Runs were checked for convergence using Tracer v. 1.6 (Rambaut et al., 2014), using as criteria effective sample sizes for each parameter larger than 200. The first 25% of trees of each run were discarded as burn-in; the remaining trees were then summarized into a majority rule consensus tree that was visualized with FigTree (Rambaut, 2009).
Additionally, we obtained an ultrametric dated tree using the Bayesian approach implemented in the program BEAST 1.8.4 (Drummond et al., 2012). This approach allows for dating a phylogenetic tree, but at the same time performs Bayesian tree searches that will be constrained for some clades depending on the age calibration used. In view of a dearth of good fossil records for Pteropodidae, we employed dates estimated by Amador et al. (2018) in a study based on more than 800 bat species of all families and 44 fossil-based calibration points. Following their results, we chose the divergence-date estimates obtained for six highly supported clades (so to have minimal effect on the topology) as age constraints to calibrate our pteropodid phylogeny. Age constraints were set with a normal distribution (including the confidence interval of the previous estimates) with the following taxa and parameters: split between Rhinolophidae and Hipposideridae (mean = 50, standard deviation = 5), crown Pteropodidae (25, 2), Hipposideros (24.5, 2), Rousettus (9, 4), Cynopterini (8, 5), and Pteropus (7, 2). The sequence data employed in our study were separated into the optimum partitioning scheme to which were applied the best substitution models according to PartitionFinder. We employed an uncorrelated relaxed clock (Drummond et al., 2006), birth-death process with incomplete sampling as tree prior, and soft bound age constraints with normal distribution based on the dates' confidence intervals from Amador et al. (2018). The MCMC was done in two runs of 15 million generations, sampling every 5000, after which convergence was checked with Tracer. As expected, resulting divergence-date estimates for all clades within Pteropodidae were very similar to those obtained in Amador et al. (2018), generally within ±1 million years.
RESULTS AND DISCUSSION
Phylogeny and Evolution
ML and BI trees were very similar, differing topologically only in a few clades and in some clade support measures (fig. 1). All analyses recovered Pteropodidae as monophyletic and strongly supported (supplementary figs. S1 and S2). As we previously observed in a genus level study (Almeida et al., 2011), six main suprageneric clades were recovered mostly with high statistical support, whereas two genera (Eidolon and Notopteris) did not fall in any of these suprageneric clades. Also confirming our previous findings, relationships between the main clades received low support in general, with a few exceptions in the BI inference (fig. 1): macroglossines + Harpyionycterinae in one clade, and Nyctimeninae + Pteropodinae + Notopteris in another clade, which were recovered in more than 95% of the posterior samples of the BI trees. Cynopterinae appeared as sister to all other pteropodids, although statistically significant support was recovered only in the dated tree, making this position unreliable (fig. 1). Cynopterines contained its two tribes, Balionycterini and Cynopterini, as previously reported (see Almeida et al., 2009). The next split in the BI trees (from both the dated and the unconstrained analyses) separates Rousettinae, a fully supported grouping of seven successive tribes, from the remaining subfamilies. Rousettinae as currently understood (see Almeida et al., 2016) replaces a previous, more restricted biogeographic concept of an endemic African clade proposed by Hollar and Springer (1997); in its new definition it includes also some typically Asian forms (Eonycteris and basal Rousettus), an arrangement also recovered here. Tribes in Rousettinae were recovered as monophyletic with Scotonycterini successively sister to Eonycterini, Rousettini, Stenonycterini, Myonycterini, Plerotini, and Epomophorini (see also Nesi et al., 2013; Almeida et al., 2016). The monotypic subfamily Eidolinae appeared in different unsupported positions, depending on the analysis.
A suprageneric group appears in all trees, including two previously recognized subfamilies Macroglossinae (a preoccupied name corrected here to Macroglossusinae; see below, Taxonomic updates) and Harpyionycterinae; no morphological synapomorphies have been detected for this unranked clade, and statistical support for this grouping was present only in the Bayesian trees (BI, pp [posterior probability] = 0.96). The other suprageneric clade that was consistently recovered consists of a heterogeneous group inclusive of Notopteris, Nyctimeninae, and Pteropodinae (BI, pp = 0.98). Placement of Notopteris has been contentious, often classified either among typical nectar-feeding pteropodids (Andersen 1912), with flying foxes in the Pteropodinae (e.g., Bergmans 1997), or in a subfamily of its own, as adopted here (see below and also Giannini, 2019). The largest pteropodid group in terms of species diversity, Pteropodinae, contained three well-defined clades classified here as tribes (see below): monkey-faced bats and relatives, blossom bats in the genera Melonycteris and Nesonycteris, and flying foxes. The latter included the genera Styloctenium, Neopteryx, Acerodon, and the speciose Pteropus.
The genus Pteropus was found to be monophyletic in our analyses except for the unsupported position of the form currently classified as P. personatus, whose position remains uncertain, appearing either linked with Acerodon (in the BI tree) or outside the Acerodon + Pteropus clade, and in both cases with little or no support. Pteropus scapulatus was sister of all typical Pteropus species (to the exclusion of problematic personatus) and represents an early invasion of the Australian continent, which, interestingly, had no consequences for the diversification of Pteropus as it remains the single species originating in this branch. By contrast, high speciation rates appear to have been the emerging process governing all subsequent diversification in this genus based on our results. In our analysis, four major clades of Pteropus species branched after P. scapulatus and included members of the lombocencis and pelagicus species groups (first clade, predominantly micronesian); livingstonii and vampyrus species groups (second clade, SE Asian and W Indian Ocean); capistratus, vetulus, and samoensis species groups (third clade, meganesian and Central Polynesian); and poliocephalus, ocularis, and griseus species groups (widespread in Australasia). Group membership in our study matches that of Almeida et al. (2014; see fig. 1), with addition here of P. melanotus to the griseus species group (as in previous study by Phalen et al., 2017). This arrangement is also compatible to that of Tsang (2015). Only one species group is missing in our analysis (the melanopogon species group, linked to griseus species group by Tsang, 2015), and the extinct P. subniger from the Mascarenes (incertae sedis; see Almeida et al., 2014).
The age of the Pteropodidae (∼25 mybp; fig. 2) makes it one of the youngest bat families. This late Oligocene age was closely followed by independent origination of phytophagous bats (Stenodermatinae, Phylostomidae) in the New World (Rojas et al., 2016; Amador et al., 2018). Both events occurred during a phase of climate warming that saw global temperatures increase to end-Eocene levels, and likely affected positively plant diversity (Zachos et al., 2001). The specialization of phytophagous bats in both the Old (Pteropodidae) and New World (clades inside Phyllostomidae) suggests that they found a highly reliable source of energy and nutrients in plants, but in all cases, plant clades were older than bat clades, indicating that pteropodids and phyllostomids jumped into previously established plant-animal interaction processes, likely between arboreal mammals and these plants millions of years before (see Sánchez and Giannini, 2018; Fleming and Kress, 2011).
CLASSIFICATION AND TAXONOMIC UPDATES
The phylogenetic structure recovered in our study indicates that several taxonomic changes in the higher-level classification of pterodids are necessary, and our recommendations are shown in table 1 (complete down to species level in table S2, available online at: doi.org/10.5531/sd.sp.39). We recognize eight subfamilies and 14 tribes. This classification resembles that of Bergmans (1997), the last formal classification of the entire family Pteropodidae, but it also differs from it in some important aspects. We erect four new tribes, correct two preoccupied subfamily names, and propose the emended diagnosis of Epomophorus now inclusive of Micropteropus and of one species formerly included in Epomops. Remaining issues in pteropodid classification chiefly revolve around uncertainties related to topology of the backbone of the pteropodid tree, poorly sampled species, and lack of morphological correlates of molecular groupings. Pteropus in particular, comprising, in our sample, some eight major subclades that include a total of 13 species groups (Almeida et al., 2014), still awaits formal classification in subgenera (see Almeida et al., 2014).
FIGURE 1.
Maximum likelihood tree (see text for details) with statistical support values showed on nodes. Asterisks designate nodes with pp ≥ 0.95 (from Bayesian inference, shown as percentages) and bootstrap (maximum likelihood) ≥ 90. Nodes without values had pp ≤0.90 and bootstrap ≤70. Species names on tree reflect the nomenclatural changes proposed herein, with older names for the affected taxa as follows: 1 = Megaerops wetmorei, 2 = Epomops dobsonii, 3 = Micropteropus pusillus, 4 = Melonycteris woodfordi, 5 = Melonycteris fardoulisi.
FIGURE 2.
Dated tree with genera represented by more than one species in the analysis collapsed. Circles designate nodes that were used in the calibration of the molecular clock. Scale in million years ago.
TABLE 1.
New classification of Pteropodidae based on the phylogenetic results presented herein.
On the basis of our phylogeny, we formalize below taxonomic changes in the classification of Pteropodidae (summarized in table 1) that include new taxonomic units for which we provide descriptions, replacement of preoccupied names, and a formal reclassification of several species that are allocated to other genera; for some of these taxa, an emended diagnosis is provided as needed.
Notopterisinae, New Subfamily
Type genus: Notopteris Gray, 1859.
Contents: Contains one genus, Notopteris Gray, 1859, and two species: N. macdonaldi Gray, 1859, and N. neocaledonica Trouessart, 1908.
Synonyms: Notopterini Koopman and Jones, 1970 (unavailable and preoccupied; see below).
Description and Diagnosis: A moderately sized pteropodid (FA 60 –72 mm) with a long tail (subequal in length to the forearm) and wing membranes that meet at the middorsal line, giving a naked-backed appearance. Dental formula i2/2, c1/1, p2/2, m2/3 × 2 = 28, although upper i1 is often lost in adults; anterior lower premolar larger than any other lower premolar or molar; premaxillae well developed and coossified anteriorly, facial process of premaxilla approximately 2× wider dorsally than ventrally; infraorbital canal long; metacarpal formula III< IV≤V; terminal phalanx of digit III subequal to or longer than metacarpal of that digit; no claw on second digit of wing; tibia length approximately half that of forearm; tongue long, approximately 1.5× length of the mandible; no obvious sexual dimorphism.
Comments: Koopman and Jones (1970) coined the name “Notopterini” for a tribe of pteropodids including Notopteris and Melonycteris, but they did not make this name available under the rules of the International Commission on Zoological Nomenclature (ICZN) because this taxon was not identified as new and the authors failed to provide a description or diagnosis indicating the characters purported to differentiate this taxon. As reviewed by Baker et al. (2016), the current Code (ICZN, 1999) is unambiguous about what is required for a family-group name (including subfamilies and tribes) to be available. In addition to requirements for publication, the name must be a noun in the nominative plural formed from the stem of an available generic name (articles 11.7.1 and 13.2) or the whole genus name (art. 29.1), which must be cited in the description (art. 16.2); and the name must end with an appropriate family-group name suffix (arts. 11.7.1.3 and 29.2). For all names published after 1930, the name must be accompanied by a description or definition that states in words characters that are purported to differentiate the taxon (art. 13.1.1) or a bibliographic reference to such a statement (art. 13.1.2); it may include a diagnosis to differentiate it from related and similar groups (recommendation 13A). And, as with all taxonomic names, family-group names are required to conform to the principle of homonymy (arts. 52, 53.1, 55), which notes that when two or more taxa are distinguished from each other, they must not be denoted by the same name.
The rules for formation of family-group names are straightforward and involve application of appropriate suffixes (-idae for a family name, -inae for a subfamily name, and –ini to the name of a tribe) to the Greek or Latin stem of a genus name (ICZN, 1999: art. 29). In the case of Notopteris, the Greek stem is composed of two parts: noto- (Greek for “back”) and pter(Greek for “wing”) forming notopter- (back wing). The appropriately formed family group names for this taxon would thus be Notopteridae, Notopterinae, and Notopterini, the latter of which was used by Koopman and Jones (1970) for the pteropodid bat group including Notopteris and Melonycteris (the latter presumably including Nesonycteris). In grouping these taxa together those authors were apparently following Andersen (1912), who did not propose a formal name for this group but variously referred to it as the “Notopterine section” of the Macroglossinae (Andersen, 1912: lxii), or as the “Notopterides” (Andersen, 1912: lxv). Bergmans (1997) noted this history and attributed authorship of the Tribe Notopterini to Andersen (1912), but this was incorrect – authorship of the formal family-group name Notopterini as applied to bats dates to Koopman and Jones (1970), a fact noted by Koopman (1994).
Regardless of issues of availability and authorship, yet another problem exists with the use of Notopterini as a family-group name in Chiroptera: it is preoccupied by Notopteridae Bleeker, 1859, a family of osteoglossiform fishes commonly known as knifefishes. Notopteridae Bleeker, 1859, is based on Notopterus Lacépède, 1800, which has the same Greek stem as Notopteris Gray, 1859. To avoid homonymy, any family-group names based on the stem notopter- can be applied only to fishes including Notopterus based on the principle of priority (which applies to family-group names and covers homonyms as well as synonyms; ICZN, 1999: arts. 23.1, 23.4) because these names were used in fishes prior to their use in bats. Homonymy from similar but not identical generic names that share the same stem is recognized by the Code (arts. 29.3, 29.6, 55.3), and recommendation 29A explicitly offers a preferred means of addressing this problem: authors are advised to use the entire generic name as the stem in formulating a new family-group taxon. In the case of bats, notopteris- thus would be the appropriate stem for formation of new family-group names. We thus propose Notopterisinae as a new family-group name for the clade including the type genus Notopteris. The new subfamily Notopterisinae differs from prior concepts (e.g., Notopterini Koopman and Jones, 1970) additionally in being restricted to only the genus Notopteris.
Macroglossusinae, New Subfamily
Type genus: Macroglossus F. Cuvier, 1824.
Contents: Includes two genera, Macroglossus F. Cuvier, 1824, and Syconycteris Matschie, 1899, and five species classified into two genera: M. sobrinus Andersen, 1911, M. minimus (E. Geoffroy, 1810), Syconycteris australis (Peters, 1867), S. hobbit Ziegler, 1982, and S. carolinae Rozendaal, 1984.
Synonyms: Macroglossinae Harris, 1939 (preoccupied; see below).
Diagnosis: Small to small-medium (in adults, interspecific ranges of head-body length 49–97 mm, forearm 37–61 mm, body weight 13–47 g; Giannini et al., 2019), long-snouted nectarivorous pteropodids with tail, uropatagium, and calcar rudimentary (Macroglossus) or absent (Syconycteris), rostrum long, strongly deflected ventrally with respect to the basicranial axis; dentition generally weak; tongue greatly elongated with long, unfringed filiform papillae on tip.
Comments: As in the case of Notopteris developed above, the widely used subfamily name Macroglossinae is preoccupied as it designates sphingid lepidopterans of subfamily Macroglossinae Harris, 1839, tribe Macroglossini Harris, 1839, and subtribe Macroglossina Harris, 1839, referred to genus Macroglossum Scopoli, 1777. We applied the same concept as in Notopteris (see above), providing a replacement name by changing the name of this subfamily to Macroglossusinae, using the entire genus name Macroglossus as valid stem and adding the suffix -inae, again following recommendation 29A of the Code. The content is now restricted to genera Macroglossus and Syconycteris to the exclusion of other nectar-feeding pteropodids previously thought to be closely related: the “macroglossine section” of Andersen (1912) that formerly included Eonycteris (relocated to Rousettinae), Melonycteris and Nesonycteris (relocated to Pteropodinae), and Notopteris (placed on a family of its own; see above).
Melonycterini, New Tribe
Type genus: Melonycteris Dobson, 1877.
Contents: Includes two genera, Melonycteris Dobson, 1877, and Nesonycteris Thomas, 1887, and three species: M. melanops Dobson, 1877, Nesonycteris fardoulisi (Flannery, 1995), and N. woodfordi (Thomas, 1887).
Description and Diagnosis: Small to moderately sized pteropodids (FA 42 –63 mm) with vestigial anterior upper and lower premolars; dental formula i2/1–2, c1/1, p3/3, m2/3 × 2 = 32–34; middle lower premolar smaller than either molar; premaxillae well developed and in simple contact (or not in contact) anteriorly; facial process of premaxilla approximately 2–3× wider dorsally than ventrally; infraorbital canal long; metacarpal formula III< IV≤V; terminal phalanx of digit III subequal to or longer than metacarpal of that digit; claw present on digit II of wing in some species; tibia much less than half the length of the forearm; tail absent.
Comments: This tribe is equivalent to the genus Melonycteris as recognized by many authors (e.g., Flannery, 1995; Simmons and Cirranello, 2019). We treat Nesonycteris as a genus distinct from Melonycteris; see comments under Nesonycteris below.
Pteralopini, New Tribe
Type genus: Pteralopex Thomas, 1888.
Contents: Contains three genera, Pteralopex Thomas, 1888, Mirimiri Helgen, 2007, and Desmalopex Miller, 1907, and eight species: Pteralopex anceps K. Andersen, 1909, P. atrata Thomas, 1888, P. flanneryi Helgen, 2005, P. pulchra Flannery, 1991, and P. taki Parnaby, 2002; Mirimiri acrodonta (Hill and Beckon, 1978); and Desmalopex leucopterus (Temminck, 1853) and D. microleucopterus Esselstyn et al., 2008.
Description and Diagnosis: Large to very large flying-fox-like pteropodids (FA 116–171 mm); dental formula i2/2, c1/1, p3/3, m2/3 × 2 = 34; upper molar teeth subquadrate to quadrate in occlusal view; premaxillae well developed and either in simple contact or co ossified anteriorly; postorbital process very long, often contacts zygoma to completely enclose the orbit posteriorly; second phalanx of digit IV of wing longer than first phalanx; claw present on digit II of wing.
Comments: The recent segregation of the Fijian monkey-faced bat in a genus of its own (Mirimiri) is strongly supported here by the very long branch separating the two Pteralopex species included in this study (P. atrata and P. taki) and Mirimiri acrodonta; mutual synapomorphies are listed in Helgen (2005). In addition, the relationship between Desmalopex (formerly in Pteropus) and monkey-faced bats (Pteralopex + Mirimiri) is also strongly supported (see also Almeida et al., 2014, 2018). Therefore, inclusion of the three genera in this new tribe is confirmed.
Harpyionycterini, New Tribe
Type genus: Harpyionycteris Thomas, 1896.
Contents: Includes two genera, Harpyionycteris Thomas, 1896 and Boneia Jentink, 1879, and three species: Harpyionycteris whiteheadi Thomas, 1896, H. celebensis Miller and Hollister, 1921, and Boneia bidens Jentink, 1879.
Diagnosis: No morphological diagnostic characters have been identified for this tribe. However, Harpyionycteris + Boneia compose a well-supported group (fig. 1).
Comments: This new tribe is one of two tribes (see also below) recognized here within the subfamily Harpyionycterinae Miller, 1907, whose contents were established by Giannini et al. (2009) as including Harpyionycteris, Boneia, Dobsonia, and Aproteles. Previously, Boneia was treated as a subgenus of Rousettus (e.g., Bergmans, 1994, 1997; Bergmans and Rozendaal, 1988), or as a valid genus associated with Rousettus (e.g., Andersen, 1912; Koopman, 1993) on the basis of morphological similarity. However, Boneia has been consistently recovered in phylogenetic analyses as sister to Harpyionycteris (Giannini et al., 2009, Almeida et al., 2011, Amador et al., 2018), a result reproduced herein as well, which is the basis of recognizing this tribe within Harpyionycterinae. Harpyionycteris has long been considered an early offshoot of the pteropodid tree chiefly due to its unusually multicuspidate dentition resembling a (primitive mammalian) tribosphenic pattern (e.g., Slaughter, 1970). Significantly, Andersen (1912) nevertheless noted the close affinity with Dobsonia, which was first confirmed phylogenetically only recently (in Giannini et al., 2006). Therefore, the dentition and other characters such as the procumbent premaxilla, mandibular symphysis, and incisors of Harpyionycteris have been interpreted as uniquely derived traits of this genus (Giannini et al., 2006). Hassanin et al. (2016) used Boneini as a tribe-level name for a group whose membership corresponds to that of Harpyionycterinae Miller, 1907, as modified by Giannini et al. (2009), and Boneini is therefore considered a junior synonym.
Dobsoniini Andersen, 1912
Type genus: Dobsonia Palmer, 1898.
Contents: Includes two genera, Dobsonia Palmer, 1898, and Aproteles Menzies, 1977, and 14 species: Dobsonia anderseni Thomas, 1914, D. beauforti Bergmans, 1975, D. chapmani Rabor, 1952, D. crenulata K. Andersen, 1909, D. emersa Bergmans and Sarbini, 1985, D. exoleta K. Andersen, 1909, D. inermis K. Andersen, 1909, D. minor (Dobson, 1879), D. moluccensis (Quoy and Gaimard, 1830), D. pannietensis (De Vis, 1905), D. peronii (E. Geoffroy Saint-Hilaire, 1810), D. praedatrix K. Andersen, 1909, D. viridis (Heude, 1896); and Aproteles bulmerae Palmer, 1977.
Diagnosis: Midsized to large pteropodids uniquely characterized by the combination of wings that originate from the spine, thus giving the appearance of bare-backed bats, and the dental formula I 0–1/0, C 1/1, P 2/3, M 2/3. Other bare-backed bats (Rousettus spinalatus and Notopteris spp.) always exhibit at least one lower incisor, but in addition these species greatly differ in numerous other cranial and dental features as they belong in different subfamilies.
Comments: Bergmans (1997) attributed the authorship of this tribe to Andersen (1912). Besides the extensive description of Dobsonia in Andersen (1912), to our knowledge there is no proper diagnosis of this tribe as currently composed (Dobsonia plus the more recently described Aproteles), only the mention of its constituent genera by Bergmans (1997), and hence the diagnosis provided above. Dobsonia and Aproteles are quite alike and differ chiefly in dental formula, with Aproteles completely lacking incisors, both upper and lower, and Dobsonia species lacking the claw on index finger (present in Aproteles). Aproteles is monotypic and contains the critically endangered, New Guinea highlands endemic, A. bulmerae, while Dobsonia is more speciose and geographically widespread, albeit the distribution of the genus is centered also in New Guinea. First discovered as a subfossil dated 12000 years old, some evidence suggests that Aproteles has been in decline perhaps for millennia, probably due to hunting by humans both prehistorically (as first encountered as food remains in archaeological sites), and currently in its limited range (Flannery, 1995). Competition for caves with Dobsonia species may be an additional cause of decline (Bonaccorso, 1998); our tree shows that Dobsonia species are all considerably younger than its sister Aproteles, thus contributing some support to this hypothesis.
Epomophorus Bennett, 1835
Type species: Pteropus gambianus Ogilby, 1835.
Contents: Includes 11 species: Epomophorus angolensis Gray, 1970, E. anselli Bergmans and van Strien, 2004, E. crypturus Peters, 1852, E. dobsonii Bocage, 1889, E. gambianus (Ogilby, 1835), E. grandis (Sanborn, 1950), E. labiatus (Temminck, 1837), E. minor Dobson, 1879, E. wahlbergi (Sundevall, 1846), E. pusillus Peters, 1867, and E. intermedius (Hayman, 1963).
Synonyms: Micropteropus Matschie, 1899.
Emended diagnosis: Typical epomophorine pteropodids with epaulettes in adult males and tufts of white hairs medially and laterally on the ears in both sexes; five to six thickened palatal ridges present; and postdental palate deeply concave.
Comments: Epomophorus minimus Claessen and De Vree, 1991, is provisionally considered a junior synonym of E. minor Dobson, 1879, pending thorough revision of small-sized Epomophorus in the minor-minimus-labiatus complex, of which E. anselli may also be part. Our revised definition of Epomophorus contains species historically assigned to this genus, along with species subsequently transferred to Epomops (i.e., dobsonii), species included in Micropteropus as subgenus of Epomophorus (i.e., pusillus), and species originally described in Micropteropus (i.e., grandis, intermedius). Therefore, in this treatment Epomophorus contains, in addition to its typical species, all three species previously assigned to Micropteropus (thus considered a junior synonym at the genus level), as well as the form dobsonii (formerly in Epomops). This broadly defined, inclusive Epomophorus is justified based on findings that the aforementioned species nest within the smallest clade containing the typical Epomophorus species in all recent phylogenies (Almeida et al., 2016; Amador et al., 2018; Nesi et al., 2011, 2013), including this study. This arrangement is supported by our reinterpretation of palatal-ridge patterns (discussed further below), which provides a clear synapomorphy for this redefined Epomophorus.
Epomophorus, as defined and diagnosed here, exhibits 5–6 thickened palatal ridges in all 11 species, with the following variation among species. The other three genera in tribe Epomophorini (Hypsignathus, Epomops and Nanonycteris) bear just three such thickened ridges. Typical Epomophorus (E. gambianus, E. crypturus, E. wahlbergi, E. angolensis, E. labiatus, E. minor, E. minimus, and E. anselli) all have the typical pattern of six thickened palatal ridges, with some interspecific variation that includes ridges with or without a median notch, and ridges whose location varies with respect to the dentition (e.g., one versus two postdental ridges; Bergmans, 1997). Species previously referred to Micropteropus (pusillus, grandis, and intermedius) all have six thickened palatal ridges, of which the first is large, undivided, and hastate, while the next five ridges are divided by a deep, continuous median groove; furthermore, ridges two through six are in close contact in intermedius and pusillus (see Bergmans, 1988, 1997). Epomophorus dobsonii has five ridges, but the second one is incipiently split; we observe that this may be the result either of fusion (of formerly separate ridges two and three) or of a division of the second ridge (Bergmans, 1997). In addition, E. dobsonii lacks the typical condition of the pterygoid bone seen in Epomops (the genus to which it was referred to for most of its taxonomic history), namely, a pterygoid bone not rolled up to form a deep cavity as is typically present in Epomops buettikoferi and E. franqueti (this trait was figured in Bergmans, 1989: figs. 6, 12). Instead, the pterygoid in Epomophorous dobsonii resembles that of other epomophorines including all Epomophorus (Bergmans, 1989: fig. 9). The condition of the postdental palate in E. dobsonii and all species previously referred to Micropteropus also matches typical Epomophorus, which have deeply concave postdental palate (see Bergmans, 1988, 1989, 1997; Giannini and Simmons, 2005), as opposed to palate only weakly concave in Epomops (Bergmans, 1989).
Nesonycteris Thomas, 1887
Type species: Nesonycteris woodfordi Thomas, 1887.
Contents: Includes two species: Nesonycteris fardoulisi (Flannery, 1993) and N. woodfordi Thomas, 1887.
Synonyms: None.
Diagnosis: Nesonycteris species lack a claw on the second digit (present in Melonycteris), lack the first lower incisor (present in Melonycteris), and present dorsal and ventral body fur similarly light orange brown (Melonycteris has a dark brown venter and orange dorsum).
Comments: Nesonycteris is often included in Melonycteris (e.g., Flannery, 1993; Simmons and Cirranello, 2019), but the phylogenetic results reported here recovered the two species of Nesonycteris forming a strongly supported clade that is sister to Melonycteris and separated by a very long branch dated with a point estimate of their split at 10 mybp. Moreover, Melonycteris and Nesonycteris each exhibit apomorphies (see Diagnosis above) that, together with our phylogenetic results, warrant recognition of two genera.
Megaerops Peters, 1865
Type species: Pachysoma ecaudatus Temminck, 1837.
Contents: Includes four species: Megaerops ecaudatus (Temminck, 1837), M. kusnotoi Hill and Boeadi, 1978, M. niphanae Yenbutra and Felten, 1983, and M. albicollis (Francis, 1989). It does not include the form wetmorei, transferred to Ptenochirus as P. wetmorei.
Synonyms: None; included as a subgenus of Pachysoma (= Cynopterus).
Emended diagnosis: Cynopterine genus with tail generally absent (extremely short, up to 4 mm only in some western Malaysian specimens of M. albicollis), rostrum short and deep, extremely deep in type species, with upper incisors unequal (I2 about one half of I1), upper canine relatively short and with mesial surface convex (without mesial groove), single lower incisor represented by i2 (see Giannini and Simmons, 2007), lower canine very short.
Comments: Simmons (2005) recognized four species of Megaerops: M. ecaudatus, M. niphanae, M. kusnotoi, and M. wetmorei, the latter with two distinct subspecies, the Philippine M. w. wetmorei and the Sundaic M. w. albicollis Francis, 1989. Recently, Giannini (2019) recognized albicollis as a separate, valid species of Megaerops, and transferred typical wetmorei to Ptenochirus, as P. wetmorei. Here we provide additional justification for these changes. On the basis of three specimens (two from Borneo and one from peninsular Malaysia), Francis (1989) described a distinct cynopterine taxon, Megaerops wetmorei albicollis, noting, however, that “differences [with respect to M. w. wetmorei from Mindanao] seem sufficient to warrant taxonomic recognition, but because of the geographic separation and the small number of specimens available, I consider it most appropriate to recognize them at the subspecific level” (Francis, 1989: 2878). We believe that Francis's intuition was correct about the species-level status of this new taxon. New specimens have expanded the previously known distribution to Sumatra and confirmed the distinctiveness of albicollis (see Maharadatunkamsi and Maryanto, 2002). Furthermore, Francis (1989) correctly specified a number of external and craniodental characters that distinguish albicollis from wetmorei. It turns out that these differences largely serve to distinguish Megaerops from Ptenochirus at the genus level, which is in line with the consistent recovery of wetmorei grouping in Ptenochirus and separate from Megaerops in phylogenetic analyses (e.g., Almeida et al., 2009; this study). These observations led Giannini (2019) to recognize albicollis as a valid species within Megaerops (as M. albicollis, new combination) with Sundaic distribution as typical of Megaerops, and to include wetmorei in the Philippine Ptenochirus (as monotypic P. wetmorei) with a distribution in Mindanao only. Megaerops albicollis has not been included in any phylogeny to date, but it is similar to other Megaerops in craniodental characters and in the uniform light-brown, creamy pelage, particularly apparent in M. niphanae, from which it can be distinguished by the distinctive, ample white ruff (the species epithet means “white collar”). By contrast, the pelage pattern of Ptenochirus wetmorei is typical of Ptenochirus species, i.e., dark gray head, dark muzzle, rusty ruff, and grayish-brown to rusty-brown body pelage. The tail is absent in Megaerops except in some specimens of albicollis (up to 4 mm; see Francis 1989, Giannini 2019). Craniodental characters of M. albicollis are typical of Megaerops and distinct from Ptenochirus, as follows: rostrum short, deep, and wide (vs. longer, shallower, and narrower in Ptenochirus); palate wide and semicircular (vs. narrow with slightly divergent tooth rows in Ptenochirus); dentition weak, with upper canine and premolars almost vertical, short (vs. dentition generally stronger, with noticeably longer, procumbent upper canines, and premolars with longer, shorter main cusp in Ptenochirus). These differences are schematically illustrated by Francis (1989: fig. 2).
Ptenochirus Peters, 1861
Type species: Ptenochirus jagori (Peters, 1861).
Contents: Includes three species: Ptenochirus jagori (Peters, 1861), P. minor (Yoshiyuki, 1979), and P. wetmorei (Taylor, 1934). Ptenochirus as thus recognized does not include albicollis, currently treated as Megaerops albicollis but originally described as Megaerops wetmorei albicollis Francis, 1999 (see details above).
Synonyms: None; described by Peters as a subgenus of Pachysoma (= Cynopterus).
Diagnosis: As in Andersen (1912), a cynopterine genus with upper incisors unequal (inner incisors considerably bulkier and longer than minute outer incisors), upper canine with deep anteromedial (i.e., mesial) groove (inconspicuous in P. wetmorei), single lower incisor representing i2 (see also Giannini and Simmons, 2007), tail always present (range across species 3–18 mm), head pelage significantly darker than body pelage, rusty brown ruff.
Comments: Ptenochirus wetmorei was originally described in Megaerops chiefly because it lacks a tail, which is characteristic of Megaerops among cynopterine genera (see Taylor, 1934). Ptenochirus wetmorei indeed appears intermediate between typical Ptenochirus and Megaerops in some morphological aspects (very short tail, upper canine with barely noticeable or lacking groove), but a closer relationship of this form to Ptenochirus jagori (and P. minor) than to Megaerops species was found in the first comprehensive study of cynopterine relationships by Almeida et al. (2009), a finding replicated in this study.
CONCLUSIONS
A dated phylogeny of the speciose bat family Pteropodidae is presented here sampling ∼70% of currently recognized species and all 45 genera. We recovered support for eight major clades, all now identified at subfamily level, although relationships among these taxa along the backbone of the tree remained mostly unresolved. Several subfamilies contain well-supported internal structure that is recognized in tribes both previously proposed and new taxa described herein. We applied taxonomic changes to the classification of Pteropodidae to reflect our phylogenetic results, including proposal of new tribes, correction of preoccupied names at subfamily level, redefinition with emended diagnosis of a key African genus, Epomophorus, recognition of Nesonycteris as a genus distinct from Melonycteris, and reorganization of species and subspecies previously included in Megaerops wetmorei. In sum this represents a new phylogeny-based classification of the Pteropodidae, and invites a reinterpretation of evolutionary history of this diverse clade. A preliminary view of this history based on our results suggests a South-East Asian origin for the family associated with a complete shift to phytophagy; an early, rapid dispersal of main-subclade ancestors to reach most continents and near-continent islands presently occupied by the descendant species; and a recent phase of oceanic dispersal and spectacular diversification in flying foxes (Pteropus), which successfully reached remote-island environments of the western Indian and northwestern Pacific oceans and central Polynesia.
ACKNOWLEDGMENTS
We are indebted to Rob DeSalle (Sackler Institute for Comparative Genomics, American Museum of Natural History) for providing laboratory space and reagents for the molecular part of this work. We thank Lawrence Heaney (Field Museum of Natural History, Chicago), Denis O'Meally (Australian Museum), Jeremy Jacobs, Louise Emmons, and James Mead (Smithsonian Institution, National Museum of Natural History, Washington DC), and Wim Bergmans (ZMA Zoölogisch Museum, Amsterdam) for access to tissue samples used to generate new data for this contribution. Our thanks also to Deanna P. Byrnes for her generosity in providing two unpublished sequences for use in this study and for her thoughtful review of the manuscript. N.P.G. and F.C.A. were funded by the Agencia Nacional de Promoción Científica y Tecnológica (Argentina) grants PICT 2015-2389, PICT 2016-3682 (NPG), and PICT2015-2009 (FCA).
