Recent molecular evidence has shown that the largest genus of the family Hipposideridae, Hipposideros, is paraphyletic with respect to H. commersonii sensu lato and H. vittatus, both belonging to a species complex referred to as the commersonii group. The taxonomic issues at the generic level of certain species of Hipposideros remain unresolved in part related to insufficient material in previous molecular studies. Herein, we expand sampling of the commersonii group and include H. commersonii sensu stricto from its type locality, Madagascar. Our phylogenetic analysis revealed that the commersonii group forms a highly supported monophyletic clade with H. cyclops, which is sister taxa to Aselliscus and Coelops. A combination of phylogenetic and comparative morphological analyses, as well as divergence time estimates, were used to provide compelling evidence to support the placement of the clade containing the commersonii group and that with H. cyclops in two resurrected genera, Macronycteris and Doryrhina, respectively. Divergence time estimates indicated that Macronycteris and Doryrhina diverged 19 mya and separated from Coelops and Aselliscus in the Oligocene, about 31 mya. The commersonii group underwent a rapid radiation as recently as 3 mya likely in response to favourable climatic conditions during the Late Pliocene in Africa. Phylogenetic analysis of Cyt-b could not resolve relationships within this morphologically conserved complex. Further sampling is necessary to fully elucidate the evolutionary history of Doryrhina. Given that cryptic species are widespread among bats, including within the genus Hipposideros, this study highlights the shortcomings of current chiropteran taxonomy to describe hidden diversity.
Introduction
Morphologically similar or identical cryptic species pose a challenge to traditional systematics, which typically has relied on discrete morphological differences to delimit species. However, since the advent of PCR, DNA sequence data has meant that the detection of speciation in the absence of evident morphological change has become possible (Bickford et al., 2007; Tsang et al., 2016). Cryptic speciation is thought to be particularly prevalent in taxa which communicate by non-visual means or are subject to evolutionary constraints induced by extreme environmental conditions (Bickford et al., 2007). Particularly, due to their use of echolocation as a complex and non-visual means of communication (Schuchmann et al., 2012; Puechmaille et al., 2014a), cryptic diversity is expected to be widespread in echolocating bats (Jones, 1997). Indeed, this has been found to be true of species across the bat phylogenetic tree, which despite cryptic taxa being almost identical morphologically, have been shown to be distinct at the molecular level (Castella et al., 2000; Khan et al., 2010; Puechmaille et al., 2014b; Dool et al., 2016; Gager et al., 2016). Furthermore, these studies also highlight how current taxonomy, which is often central to management and conservation planning, may be insufficient to describe the diversity of extant species (Bickford et al., 2007).
For many years the relationships within the Hipposideridae have been considered a grand challenge in chiropteran systematics with Bogdanowicz and Owen (1998) comparing the phylogeny of the Hipposideridae with the mythological Minotaur's labyrinth of ancient Greece. Their seminal paper begins by highlighting the need for studies which sample the entire family and re-examine the validity of long recognised Hipposideros species complexes (Bogdanowicz and Owen, 1998); however, most likely due to the morphological stasis, many relationships remained unresolved based on morphological analysis alone (Bogdanowicz and Owen, 1998; Hand and Kirsch, 1998).
Recently, sampling all recognized hipposiderid genera, an analysis of a large ca. 10 kb molecular dataset, better suited to resolving relationships between morphologically similar taxa, resulted in the recognition of a distinct family, Rhinonycteridae (Foley et al., 2015). The re-examination by these authors of morphological characters in light of evidence from phylogenetic analysis and divergence time estimates revealed that several genera previously assigned to Hipposideridae, were in fact mor phologically, as well as molecularly, sufficiently divergent from Hipposideridae to warrant being placed into a separate family.
Hipposideros, the most speciose genus of the family Hipposideridae is estimated to contain approximately 70 species (Simmons, 2005), and has extant taxa distributed throughout the Old World spanning Asia, Australia, Middle East and Africa (Nowak and Paradiso, 1999). The current number of recognized species represents an underestimate of the actual diversity of the genus, as in recent years it has been shown to be rife with cryptic diversity and several new taxa described since Simmons' tabulation (Thabah et al., 2006; Monadjem et al., 2010, 2012, 2013; Murray et al.; Thong et al., 2012a, 2012b; Rakotoarivelo et al., 2015). This genus has been further subdivided into nine different species groups (Simmons, 2005) based on morphological similarities; these species groups have yet to be thoroughly evaluated using molecular data (Bogdanowicz and Owen, 1998; Murray et al., 2012).
Adding to the taxonomic challenges presented by cryptic diversity within hipposiderids, the monophyly of the genus Hipposideros has also long been questioned by studies based on morphology (Sige, 1968; Legendre, 1982; Bogdanowicz and Owen, 1998). In contrast, a recent molecular phylogeny of South East Asian Hipposideros taxa indicate that members of the genus, at least from this geographic region, show no evidence of paraphyly based on a dataset consisting of a single mitochondrial and single nuclear marker (Murray et al., 2012). The same pattern has been found for other mitochondrial DNA based studies focused solely on a few African taxa (Vallo et al., 2008; Monadjem et al., 2013). In contrast, Foley et al. (2015) sampling a range of Asian and African Hipposideros taxa have confirmed that the genus is paraphyletic, with H. commersonii sensu lato (s.l.) and H. vittatus forming a strongly supported sister-taxa relationship with Aselliscus and Coelops. At present the taxonomic implications of this result remain unaddressed due to incomplete sampling of H. commersonii sensu stricto (s.s.) from its type locality, Madagascar, which is necessary to provide correct assignment of the clade representing this species.
Hipposideros commersonii s.l. and H. vittatus are part of the notably large-bodied commersonii group, which also contains H. gigas and H. thomensis (Simmons, 2005), and the more recently described H. cryptovalorona (Goodman et al., 2016). Previously designated subspecies of H. commersonii s.l., H. vittatus, H. gigas and H. thomensis are now considered distinct species by some authors, within the commersonii species group, based on differences in echolocation calls and morphology (Simmons, 2005 and references therein). However, this classification is largely untested and the explicit character differences, particularly between vittatus and gigas, and their relationship to commersoni s.s., needs to be examined in detail (Monadjem et al., 2010). The morphological distinctiveness of various species currently retained in Hipposideros was noted by Gray (1866), who placed gigas in the genus Macronycteris and by Peters (1871), who placed cyclops in the subgenus Doryrhina. Later authors reassessing aspects of Gray's and Peters' generic allocations, adding taxa to the comparisons, and de scriptions of new species, proposed different arrangements below the level of genus (Tate, 1941).
Recently studies combining morphological and molecular data have discovered that H. commersonii sampled across Madagascar, the type locality of this species (Geoffroy Saint-Hilaire, 1813), contains unique evolutionary lineages indicative of cryptic diversity on the island. Analysis of two mitochondrial markers and two nuclear introns of H. commersonii s.s., revealed clades with varying support depending on marker; clade C occurring in the northern portion of the island; clade B found predominantly in the south-west; and, a notably divergent clade A found in the south-west that was more closely related to H. vittatus and H. gigas from continental Africa (Rakotoarivelo et al., 2015). Rakotoarivelo et al. (2015) classified clades B and C as the true representatives of H. commersonii. Subsequently, clade A was recognized as a distinct species, H. cryptovalorona, based on molecular data, but no clear morphological difference separated this taxon from H. commersonii (Goodman et al., 2016). These morphological results suggest that for the commersonii group, specifically extant taxa on Madagascar and Africa, field identifications should be treated cautiously until confirmed with appropriate molecular data. Despite extreme morphological conservatism, molecular clock analysis using Cyt-b indicated that this species group arose and began to diverge 5.8 mya during the Miocene (Rakotoarivelo et al., 2015).
This current study aims to assess the appropriate generic rank for the forms making up the commersonii group, which were shown by Foley et al. (2015) to be paraphyletic with respect to numerous other species of Hipposideros, by increasing taxon sampling, including material of H. commersonii s.s. from Madagascar, and any other Hipposideros species demonstrating paraphyly with respect to the genus, and generating novel molecular data for key taxa. A comparative morphological analysis of Hipposideros species, including the neotype of H. commersonii and holotype of H. cryptovalorona, were carried out to provide a morphological description for taxonomic rank for members of the commersonii group. Furthermore, Cyt-b was sequenced for different individuals from the commersonii group included in this study to verify field identifications and allocation to currently recognised species. A combination of cutting edge phylogenetic methods and molecular clock analysis provide further evidence towards elucidating the outstanding taxonomic issues surrounding this species group.
Materials and Methods
Sampling and Taxonomic Aspects
Samples for genetic analysis were either tissue samples from museum collections or wing biopsies obtained from animals captured and released in the context of recent fieldwork (see Table 1 for sample details). A number of other sequences were downloaded from GenBank (Table 1).
The taxonomy and diagnoses of African species of the commersonii group remain unresolved (Monadjem et al., 2010). Currently the morphological characters used to distinguish H. gigas from H. vittatus are ambiguous and, at least in part, based on habitat and distribution, which is further complicated by their partially overlapping distributions (Monadjem et al., 2010; Happold, 2013a, 2013b). As H. thomensis is an isolated insular population that is distinguished based on echolocation and external characters (Simmons, 2005; Rainho et al., 2010), the genetic diagnosis of this taxon is not a critical issue for this current project. In contrast, a number of synonyms exist for continental African members of the commersonii species group, including gambensis, marungensis, niangarae and viegasi, and to date all taxonomic work and considerations on these animals have been based on non-molecular characters (Tate, 1941). More detailed genetic analyses are needed for the resolution of several taxonomic questions of African members of the commersonii group and our utilization of the names gigas and vittatus for certain samples incorporated herein should not be considered as definitive identifications. In the context of this study, we did not have access to specimen material of H. thomensis.
Specimens
Material for the morphological and molecular genetic analyses presented herein are housed in the following museums: AMNH, American Museum of Natural History, New York; DM, Durban Natural Science Museum, Durban; FMNH, Field Museum of Natural History, Chicago; and UADBA, Département de Biologie Animale, Université d'Antananarivo, Antananarivo.
Comparison to Type Specimens
In the context of this study, we were able to make comparisons to the following type specimens: neotype of H. commersonii (FMNH 175972) and holotype of H. cryptovalorona (FMNH 175970).
DNA Extraction and Amplification
DNA was extracted using a Qiagen DNeasy kit following the manufacturer's instructions. On the basis of recent papers highlighting the utility of fast evolving nuclear introns to resolve phylogenetic relationships at these levels of divergence compared to exons (Foley et al., 2015; Dool et al., 2016), we amplified six nuclear intron gene fragments (ABHD11, BGN, PRKC1, STAT5A, ROGDI and THY) for 12 species of Hipposideros. Primer sequences, PCR cycles, thermocycler settings and PCR product purification procedures are as described in Foley et al. (2015). Cyt-b was amplified for eight individual Hip po sideros captured in continental Africa assigned to the com mersonii group and one outgroup taxon, H. cyclops, using the previously described primers mtDNA-R3-F and mtDNAF2-R and thermocycling conditions (Puechmaille et al., 2011). All samples were sequenced with mtDNA-R3-F and mtDNAF3-R (Puechmaille et al., 2011), with the exception of H. cyclops, which was sequenced using a combination of mtDNAR3-F and MVZ16 (Smith and Patton, 1993). All purified PCR products were Sanger sequenced in both directions by Ma crogen (Europe). The newly sequenced intron dataset of 12 individuals from Hipposideros species were supplemented with GenBank data for 13 additional hipposiderid taxa and two rhinonycterid outgroups to bring the total dataset to 27 taxa (Table 1). The Cyt-b data was combined with 31 individuals from the Cyt-b dataset of Rakotoarivelo et al. (2015) to confirm field identification and clade membership.
Phylogenetic Analysis
Raw sequence data were trimmed, assembled into contigs and quality checked using CodonCode Aligner 3.7.1 (Codon-Code Corporation, Dedham, MA). Single gene fragments were aligned using Muscle (Edgar, 2004), visually inspected and ambiguous positions manually removed. An appropriate model of sequence evolution was chosen using jModelTest v2.1.4 (Posada, 2008) using the AIC criteria for model selection. Where the model specified by jModelTest was not implemented in software for downstream analysis the next most complicated model was used. Individual intron gene trees were generated using BEAST v1.8.0 (Drummond and Rambaut, 2007). The Markov Chain Monte Carlo (MCMC) was run for 10 million generations, sampling every 1,000 generations. Convergence of all Bayesian analyses was confirmed with Tracer v1.5 (Rambaut and Drummond, 2007) and was deemed sufficient when all Effective Sample Size (ESS) values were > 200. A consensus tree was generated using Tree Annotator (Drummond and Rambaut, 2007), discarding 10% of initial trees as burn-in.
Table 1.
List of taxa used in this analysis — taxonomy following Simmons (2005). Also included are sample associated voucher numbers, sampling location and sample/data source. GenBank accession numbers for data generated and used in this study are also provided. *** denotes missing data; ‘a’ denotes data downloaded from GenBank. Acronyms for museum voucher specimens include: AMNH — American Museum of Natural History; DM — Durban Natural Science Museum; FMNH — Field Museum of Natural History; UADBA — Universite d'Antananarivo, Departement de Biologie Animale
Continued
All individual intron gene fragments were concatenated to form a super-matrix for subsequent analysis. Phylogenetic analyses were carried out using both Maximum Likelihood (ML) and Bayesian Analysis (BA) approaches, implemented using the RAxML gui (Stamatakis, 2006; Silvestro and Michalak, 2012) and MrBayes (Ronquist et al., 2012), respectively. The ML analysis was carried out using a fully partitioned model in which each gene fragment had its own model of sequence evolution,with all other settings as default and confidence was estimated using 500 bootstrap replicates. The MrBayes analysis was carried out using an MCMC run for 10 million generations, sampled every 1,000 generations, under a fully partitioned model. All other parameters were set as default. Cyt-b dataset was analysed separately following Dool et al. (2016). Our Cyt-b analyses included H. cyclops (not shown in Fig. 2) and two H. cryptovalorona individuals as outgroups and included 37 individuals from the commersoni species group. This dataset was analysed using ML and BA optimality criteria as described above, except the MrBayes MCMC was run for 20 million generations sampling every 1,000 generations to achieve convergence.
Molecular Clock Analysis
Divergence time estimates were obtained for our tree using the MCMCTREE program in PAML (Yang, 2007). MCMCTREE requires a fully bifurcating tree. As our MrBayes tree contained a polytomy among several commersonii species group taxa, this polytomy was randomly resolved using the APE package (Paradis et al., 2004) in R v3.0.3 (R Core Team, 2016). As it is known that Hipposideros is paraphyletic (Foley et al., 2015), it is difficult to confidently assign fossil material attributed to Hipposideros to a given branch in our tree for use as fossil calibrations to constrain our molecular clock analysis. To circumvent this problem, we used the 95% confidence interval obtained for the divergence time estimates for the node denoting the split between the Rhinonycteridae and Hipposideridae obtained in the analysis of Foley et al. (2015) as a minimum and a maximum soft bound calibration (36–41 mya) to constrain the root of our tree following (Dool et al., 2016). A further soft bound constraint of 16–34 mya was applied to the split between Rhinonicteris and Cloeotis following Foley et al. (2015). The MCMCTREE analysis was run under the HKY85 model of sequence evolution, as it was the closest in complexity to the optimum model chosen for our dataset. The analysis was carried out using both correlated and independent rate models.
Results
Alignment, Model Selection and Phylogenetic Analysis
Concatenation of the six nuclear introns resulted in a supermatrix of 2,926 aligned positions for 27 taxa. The following models were selected under the AIC criteria in jModelTest for each gene fragment; ABHD11 — TPM1uf + G; BGN — TIM1 + G; PRKC1 — TVM; ROGDI — TPM2uf + G; STAT5A — TVM + G; THY — TIM1 + G. The Cyt-b dataset comprised 594 base pairs (bp) and the GTR+G model was selected as above.
Intron tree
Analysis with Maximum Likelihood and Bayesian approaches yielded highly congruent topologies, which differed only with respect to several intracommersonii group relationships, all of which are poorly resolved in both analyses. The sister-taxa relationship between the rhinonycterid outgroups, Rhinonicteris and Cloeotis, was consistently recovered with full support (ML 100, BA 100). Higherlevel relationships among the Hipposideridae were well resolved (Fig. 1). Asellia was consistently recovered as the basal clade (ML 100, BA 100). The sister-taxa relationship between Aselliscus and Coelops was highly supported in both analyses (ML 100, BA 100). The genus Hipposideros is paraphyletic, and formed two distinct highly supported clades. One mono phyletic clade labelled as Hipposideros in Fig. 1 (ML 100, BA 100) formed a sister taxa relationship with the clade containing Aselliscus and Coelops. Within the Hipposideros clade, H. armiger, H. larvatus, H. diadema and H. lekaguli formed a strongly supported monophyletic basal clade (ML 100, BA 100). The relationship between H. galeritus and the remaining Hipposideros taxa was uncertain and received poor support (ML 58, BA 88), as was the position of H. jonesi (ML 43, BA 50). All remaining Hipposideros spp. formed a clade but support for intra-generic relationships was variable (ML > 50, BA > 80) (see Fig. 1). A distinct, highly supported clade containing H. cyclops and members of the commersonii group (ML 100, BA 100) was basal to the clade containing Aselliscus, Coelops, and the remaining Hipposideros taxa.
The commersonii group, composed of H. commersonii s.s. and individuals tentatively identified as H. vittatus and H. gigas, formed a highly supported monophyletic group in all analyses (ML 100, BA 100); however, intraspecific relationships were poorly resolved (see Fig. 1). This problem at least in part is associated with the lack of currently recognized diagnosable characters for gigas and vittatus and the historical placement of all of these forms within H. commersonii s.l., that is to say African individuals labelled as ‘H. commersonii’. Additional fieldwork across sub-Saharan Africa and off-shore islands and associated laboratory analyses are needed to resolve these aspects, as well as the possible validity of different forms placed in synonymy. To address at least a portion of the paraphyly of Hipposideros presented herein, we propose to elevate two clades to the rank of distinct genera: 1) members of the H. commersonii group and 2) H. cyclops (see Systematic Summary below).
Cyt-b tree
Topologies obtained from ML and BA analysis were highly congruent. The analysis recovered clades identified in previous studies (Rakotoarivelo et al., 2015; Goodman et al., 2016): clade A (H. cryptovalorona) (ML 89, BA 81), clade B (ML 85, BA 96), clade C (ML 92, BA 98), and the mixed H. vittatus (ML 91, BA 99) and H. gigas clades (ML 89, BA 97) (Fig. 2). However, the relationships between the commersonii group clades are not highly supported. Clade B and clade C are sister taxa (ML 73, BA 92). The H. gigas clade is basal to clades B and C, but this position is poorly resolved (ML 53, BA 56). The H. vittatus clade is sister to a clade containing H. gigas, clade B and clade C but this is not strongly supported in either analysis (ML 78, BA 78). A basal position for the H. cryptovalorona clade is highly supported. Apart from clade A, intra-clade relationships are poorly resolved and contain numerous polytomies (Fig. 2). The Cyt-b analysis confirms that our H. commersonii samples (FMNH 217931 and FMNH 209236) are representatives of the true H. commersonii belonging to clade B. Sam ples identified in the field as H. vittatus (FMNH 219682 and a re-sequenced FMNH 192857) were consistent with other H. vittatus sequences. H. gigas samples DM 14118, DM 12602 and AMNH 269871 formed a monophyletic clade. Samples of H. vitattus (FMNH 219682 and ML-129) were also found to group within the H. gigas clade, as did a H. vittatus/gigas sample (72).
Molecular Clock Analysis
Divergence times estimated using both the correlated and independent rate models were highly congruent. Results for the correlated rates model are used throughout. Divergence time estimates indicate that Rhinonycteridae and Hipposideridae diverged 39 mya during the Eocene (Fig. 3). The crown group hipposiderids began to radiate shortly after 33 mya. The split between the paraphyletic Hipposideros groups took place about 31 mya. An estimated 28 mya during the Oligocene Coelops and Aselliscus split from Hipposideros, with the genera Coelops and Aselliscus diverging 14 mya. Hipposideros cyclops and the commersonii group diverged during the Neogene 19 mya, with the crown commersonii group radiating both rapidly and recently about 3 mya.
Systematic Summary
Order Chiroptera Blumenbach, 1799 (pp. 58, 74)
Suborder Yinpterochiroptera Springer, Teeling, Madsen, Stanhope & de Jong, 2001 (p. 6243)
Superfamily Rhinolophoidea Gray, 1825: Teeling, Springer, Madsen, Bates, O'Brien & Murphy, 2005 (p. 581)
Family Hipposideridae Lydekker, 1891 (p. 657)
Genus Macronycteris Gray, 1866 (p. 82)
New combination — Macronycteris gigas (Wagner, 1845).
Also including the following species M. commersonii (E. Geoffroy, 1813), M. cryptovalorona (Goodman, Schoeman, Rakotoarivelo & Willows-Munro, 2016), M. gigas, M. thomensis (Bocage, 1891) and M. vittatus (Peters, 1852)
Synonyms
Rhinolophus Commersonii E. Geoffroy Saint-Hilaire, 1813.
Rhinolophus gigas Wagner, 1845.
Phyllorhina vittata Peters, 1852.
Phyllorhina Commersoni Peters, 1871.
Phyllorhina commersonii Dobson, 1878.
Phyllorhina commersoni var. thomensis Bocage, 1891.
Hipposideros commersoni Andersen, 1906.
Hipposideros gigas Wagner, 1845.
Hipposideros thomensis Bocage, 1891.
Hipposideros Commersoni Dorst, 1948.
Hipposideros vittatus Monadjem et al., 2010.
Hipposideros cryptovalorona Goodman et al., 2016.
Macronycteris Gray, 1866
Description of the Genus Macronycteris
Morphological characters
Gray (1866) in his description of Macronycteris, focused exclusively on the forehead and noseleaf structure of this genus, and the type species was designated as M. gigas. Here we provide further details on Gray's diagnosis and some other characters to differentiate Macronycteris from Hipposideros.
All five species of Macronycteris have a frontal sac (Hill, 1963; Happold, 2013c) on the forehead, in a central portion behind the posterior noseleaf. It varies in shape and size from a slightly oblong to longitudinal slit, but in all species of this genus (Fig. 4), as well as in ‘H. cyclops’, it is distinctly vertical in shape; the latter species differs notably from Macronycteris in aspects of the noseleaf (Happold, 2013c: Fig. 68a and see below). The shape and size of the frontal sac shows intraspecific variation, associated with sexual dimorphism (Andersen, 1906), being more developed in males of at least three out of the five species now placed in this genus. The structure opening reaches its maximum length of 9 mm in M. gigas, which is the largest species in the genus (see below). In M. commersonii M. gigas and M. vittatus, the frontal sac and the distal portion of the forehead is covered with fine hair (Fig. 4), although in large males of at least M. gigas it can be largely naked.
The noseleaves of Macronycteris have several particular structures, that when taken together separate them from all other genera of hipposiderids. The lateral fleshy leaflets number 4 and occasionally 3 (Fig. 4). Most species of Hipposideros either lack the leaflets or have a maximum of 2, the exceptions being H. abae, H. alongensis, H. cervinus (apparently in some cases only 2), H. dinops, H. larvatus, H. lekaguli and H. speoris with up to three leaflets and H. armiger, H. diadema, H. griffini, H. lankadiva, H. papua, H. pendleburyi and H. turpis with four leaf lets (Andersen, 1905, 1906; Payne et al., 1985; Strahan, 1995; Flannery, 1995a, 1995b; Bates and Harrison, 1997; Francis, 2008; Thong et al., 2012a, 2012b; Happold, 2013c); in all cases these species have other morphological aspects of the noseleaves that separate them from members of the genus Macronycteris. Further, the largely unornamented noseleaves of Macronycteris include the following characters: 1) anterior portion is broad and without any median modifications, internarial septa form small and narrow structures not obscuring the slightly deep-set nasal passages, and small but distinct lappets surround the nasal passages; 2) middle portion of noseleaf is simple, slightly expanded and lacking other structures; and 3) posterior portion of noseleaf without lateral process or other adornment, with a vertical medial septum, and two prominent vertical lateral septa, which divide the structure into four separate cells. (These three septa in some cases are not prominent and poorly defined.) The different aspects of the noseleaf and forehead pore are unique to Macronycteris and are not found in any other genus of the family Hipposideridae (Rosevear, 1965; Payne et al., 1985; Flannery, 1995a, 1995b; Bates and Harrison, 1997; Francis, 2008; Monadjem et al., 2010; Happold, 2013c).
The separated ears in Macronycteris are constricted towards the base, triangular in shape and with pointed distal tips and curved posterior margins that give a limp or droopy appearance. The ear length average in M. commersonii 29.6 mm (♂♂, n = 27) and 28.8 mm (♀♀, n = 76), in M. cryptovalorona 26 mm and 27 mm (n = 2), in M. gigas 31.4 mm (sexes combined, n = 34) and in M. vittatus 29.5 mm (sexes combined, n = 95) (Happold, 2013c; Goodman et al., 2016). Antitragus not notably developed. The genus is the largest amongst living hipposiderids, often showing sexual dimorphism in external measurements. The size of these animals can be best expressed by mean forearm length, which in M. commersonii is 90.7 mm (♂♂, n = 27) and 86.4 mm (♀♀, n = 76), in M. cryptovalorona 80 and 81 mm (n = 2), in M. gigas 107.9 mm (♂♂, n = 39) and 103.8 (♀♀, n = 39), in M. thomensis 79–82 mm (n = 3) and in M. vittatus 101.5 mm (♂♂, n = 78) and 93.9 (♀♀, n = 58) (Andersen, 1906; Happold, 2013c; Goodman et al., 2016). The only Hipposideros approaching the size of Macronycteris are H. dinops from Papua New Guinea with a forearm of 93.3 mm (♂♂, n = 3) and 91.6 mm (♀♀, n = 5) and H. inexpectatus from northern Sulawesi with a forearm of 100.8 mm (Laurie and Hill, 1954; Flannery, 1995b). All species of Macronycteris have relatively dense and short fur, with a mixture of rich fawn to rufus on the dorsum and slightly lighter ventrum, and distinctive white fur on the shoulders; certain species have a distinct rufus-orange colour phase. The unfurred wing and tail membranes are pale to dark brown. The tail is distinctly shorter than the length of the extended hind foot. The plagiopatagium attaches at approximately ankle level.
The skull of the different members of Macronycteris are notably large, with very pronounced lambdoid and sagittal crests, being more developed in adult males and in M. gigas, the largest member of the genus. The zygomatic arches are robust and form the widest portion of the skull. The rostrum is notably wide and in dorsal view the naso-frontal portion of the skull has a distinct pentagonal shape. The mandible is notably large, with a deep symphysis, large angular process and elevated coronoid process.
Dentition notably massive, clearly indicative of a powerful bite for subduing prey. Dental formula — premolars 1/2, canines 1/1, premolars 2/2 and molars 3/3. The upper incisors are widely spaced on the outer portion of the premaxillae and in fresh adult dentitions weakly lobed. The upper canine is massive and in direct contact with the second upper premolar, with the first upper premolar being small and external to the toothrow. The upper canine is grooved and with a distinct cusp on the upper posterior edge. In general, cusp structure of premolars and molars similar to large members of the genus Hipposideros following the description of Miller (1907).
Karyological characters
On the basis of karyological information from the literature, further evidence can be found for the resurrection of the genus Macronycteris for species formerly placed in the commersonii group. Hipposideros is characterised by its extreme karyological conservatism, with nearly all species examined so far having a 2n complement of 32 (Harada et al., 1982; Hood et al., 1988; Rautenbach et al., 1993; Bogdanowicz and Owen, 1998; Koubinova et al., 2010; Mao et al., 2010; Porter et al., 2010). However, Macronycteris taxa, previously attributed to Hipposideros, differ markedly from this conserved karyotype with M. commersonii s.l., M. commersonii s.s. and M. gigas having 2n = 52 (Rautenbach et al., 1993; Koubinova et al., 2010), as well as M. commersonii s.s. having chromosome characters unlike those in Hipposideros (Richards et al., 2016). The exception to the n = 32 complement for members of the genus Hipposideros is ‘H. cyclops’ having 2n = 36 (Koubinova et al., 2010); herein we present evidence that this species is best placed in the genus Doryrhina (see below).
Order Chiroptera Blumenbach, 1799 (pp. 58, 74)
Suborder Yinpterochiroptera Springer, Teeling, Madsen, Stanhope & de Jong, 2001 (p. 6243)
Superfamily Rhinolophoidea Teeling, Springer, Madsen, Bates, O'Brien & Murphy, 2005 (p. 581)
Family Hipposideridae Lydekker, 1891 (p. 657)
Genus Doryrhina Peters, 1871 (p. 314)
New combination — Doryrhina cyclops (Peters, 1871).
Synonyms
Phyllorrhina cyclops Temminck, 1853.
Phyllorrhina cyclops Temminck, 1853 = Doryrhina cyclops (Temminck, 1853), see Peters (1871).
Rhinolophus micaceus de Winton, 1897.
Hipposideros cyclops de Winton, 1899.
Hipposideros langi J. A. Allen, 1917.
Doryrhina Peters, 1871
Description of the Genus Doryrhina
Morphological characters
Peters (1871) in his naming of the subgenus Doryrhina presented different characters associated with the noseleaf structure, and the type species for the subgenus was Phyllorhina (Doryrhina) cyclops; a tropical African taxon (Fahr, 2013). Here we elevate the subgenus Doryrhina to the level of genus for cyclops and expand Peter's diagnosis.
The single species we place herein in the genus Doryrhina, D. cyclops, has a frontal sac on the forehead (Hill, 1963; Fahr, 2013), in a central position behind the posterior noseleaf, and opening as a relatively small vertical slit (Fig. 4). Within the slit are stiff white hairs that when everted form a distinct hair tuft (Rosevear, 1965). The noseleaf of Doryrhina shows several particular aspects which differ from other hipposiderids, including Macronycteris. Members of the genus Doryrhina have two lat eral fleshy leaflets (Fig. 4), a common configuration in Hipposideros, although several members of this genus, as well as Macronycteris, have four leaflets (Andersen, 1905, 1906; Payne et al., 1985; Strahan, 1995; Flannery, 1995a, 1995b; Bates and Harrison, 1997; Francis, 2008; Thong et al., 2012a, 2012b; Happold, 2013c). The 2nd lateral leaflet in D. cyclops extends posteriorly and forms a continuous extension of the posterior leaf (Fig. 4). Further, this species has a club-shaped process projecting from the posterior portion of the noseleaf, which distinguishes it from all African members of the genus Hip posideros, as well as Macronycteris. Other aspects of the noseleaf of D. cyclops include: 1) anterior portion is broad and with a club-like structure commencing at anterior margin of lateral leaflet and extending posteriorly to middle portion of noseleaf, internarial septa form relatively large and sculpted structures that partially obscure the deep-set nasal passages, and distinct lappets surrounded the nasal passages; 2) middle portion of noseleaf is absent; 3) posterior portion with a vertical medial septum (that extends as the club-shaped process mentioned above) and two prominent vertical lateral septa on either side of the vertical medial septum, which divide the structure into 6 separate cells, the lateral-most cell merging with the 2nd lateral leaflet, and posterior margin is a thin structure with little expansion. The different aspects of the noseleaf described herein are unique to Doryrhina and not found in another genus of the family Hipposideridae (Rosevear, 1965; Payne et al., 1985; Flannery, 1995a, 1995b; Bates and Harrison, 1997; Francis, 2008; Monadjem et al., 2010; Happold, 2013c). The exception is the African species camerunensis, apparently closely related to cyclops (Hill, 1963). Further, Tate (1941), Hill (1963) and Koopman (1994) based on anatomical characters considered the H. cyclops group to be distinct from other groups in this genus and composed of the African species H. cyclops and H. camerunensis, and the Australian-New Guinea species H. muscinus, H. wollastoni, H. corynophyllus, H. semoni and H. stenotis. With the exception of cyclops, these other species are not represented in our molecular dataset and are in need of study to determine if they are best placed in the genus Doryrhina or should be retained in Hipposideros.
The separated ears of D. cyclops are long, narrow and terminating with pointed tips. The average ear length in this species is 33.5 mm (sexes combined, n = 125) (Fahr, 2013). Antitragus not present. Other measurements of D. cyclops, which show sexual dimorphism, include: mean forearm length, 65.3 mm (♂♂, n = 53) and 68.3 mm (♀♀, n = 45); tail length, 26.5 mm (♂♂, n = 42) and 29.2 mm (♀♀, n = 77); and hindfoot length (with claws), 19.8 mm (♂♂, n = 41) and 20.5 mm (♀♀, n = 73) (Decher and Fahr, 2005). Doryrhina cyclops has dense, long and woolly pelage, on the dorsum generally blackish-brown and often with a frosted tint, while the ventrum is lighter and not frosted. No rufus or orange colour morphs are known. The wing and interfemoral membranes are blackish-brown and skin on forearm, wing digits and tibia are paler reddishbrown.
The skull of D. cyclops, is proportionately large, with a lengthened braincase, and low sagittal crest. The internarial septum is not enlarged. Rostrum is distinctly broad. Zygomatic arches are slender and form the widest portion of the skull. Premaxillae posteriorly wide and in broad contact with the palate. Anterior palatal foramina enclosed. Cochleae greatly expanded.
Dental formula — premolars 1/2, canines 1/1, premolars 2/2 and molars 3/3. The upper incisors and associated cusps are not well developed. The upper canines lack well-defined cusps, but have distinct cingula. The upper anterior premolar is reduced, visible in lateral view, and in contact with the canine and posterior premolar. Anterior lower premolar is distinctly small.
Karyological characters
On the basis of karyological information from the literature, further evidence can be found to support the resurrection of Doryrhina for a species previously placed in the genus Hipposideros, H. cyclops. Hipposideros shows extreme karyological conservatism, with nearly all species examined to date possessing a 2n complement of 32 (Harada et al., 1982; Hood et al., 1988; Rautenbach et al., 1993; Bogdanowicz and Owen, 1998; Koubínová et al., 2010). However, in the case of H. cyclops, 2n = 36 (Koubínová et al., 2010), which is also different from Macronycteris, 2n = 52 (Rautenbach et al., 1993; Koubínová et al., 2010).
Discussion
Phylogenetic Relationships
An outstanding taxonomic issue surrounding paraphyly of certain species of Hipposideros has been resolved with the resurrection of a previously synonymised genus, Macronycteris, for members of the commersonii species group and Doryrhina, for D. cyclops (see Fig. 1 and Systematic Summary). Further work is needed on African mainland and offshore island populations within the commersonii group, specifically to better diagnose M. gigas and M. vittatus; this might be best accomplished by obtaining sequence data from type specimens, or at least topotypic material, and based on new material with tissues across the African range of this genus overlying morphological characters on phylogenetic trees derived from the same specimens. In the same manner, it will also be necessary to also assess a number of taxa that are currently considered junior synonyms under M. gigas and M. vittatus. The generic placement for several African and Australasian species considered part of the H. cyclops group (sensu Koopman, 1994; Simmons, 2005), namely H. camerunensis, H. muscinus, H. wollastoni, H. corynophyllus, H. semoni and H. stenotis is unresolved based on the results presented herein. Further molecular work is needed to decide if these are best placed in the genus Doryrhina or maintained in the genus Hipposideros.
Higher level relationships within Hipposide ridae, derived from the intron analysis, were strongly supported and in agreement with the intron topology of Foley et al. (2015). Hipposideros formed a highly supported monophyletic group, sister to Aselliscus and Coelops. Phylogenetic analysis further confirmed that Hipposideros is paraphyletic. The component taxa of the resurrected genus Macronycteris formed a fully supported monophyletic group in all analyses. The association between ‘D. cyclops’ and the commersonii group, named here as Macronycteris, was previously recovered with strong support in a molecular analysis of 684 bp of Cyt-b, which focussed on Hipposideros species from West Africa (Monadjem et al., 2013). As this analysis did not contain any Asian Hipposideros taxa, the genus appeared to form a monophyletic group, also seen in a mitochondrial investigation of the caffer complex and other African taxa (Vallo et al., 2008). The relationships between the taxa common to this study and that of Monadjem et al. (2013) were recovered where D. cyclops, M. vittatus and M. gigas were basal to all other West African species including H. caffer, H. abae and H. jonesi, which in this study and in Foley et al. (2015) were recovered as part of the monophyletic Hipposideros clade also containing Asian members of this genus. Vallo et al. (2008) also recovered this relationship between D. cyclops and M. gigas, basal to all other Hipposideros, with strong support in their analysis of the complete Cyt-b mitochondrial fragment. Furthermore, they report that the genetic divergence between D. cyclops and M. gigas compared to all other Hip posideros taxa is actually greater than the intergeneric distance observed between Aselliscus and Hipposideros reported by Wang et al. (2003), which led them to suggest that D. cyclops and M. gigas could represent a distinct genus. Together these results provide further evidence from an independent data source, namely mitochondrial data, to support that D. cyclops falls outside of the genus Hipposideros and is closely related to the genus Macronycteris, which we resurrect herein.
Intra-generic relationships were relatively well resolved amongst Hipposideros, except the position of H. jonesi, which remains tenuous as it was recovered with poor support in both the ML and BA analyses (ML 43, BA 50). Herein we found strong support for the sister taxa relationship between H. diadema and H. lekaguli reported in the Esseltyn et al. (2012) analysis of the mitochondrial gene fragments ND2 and Cyt-b. A clade containing the Asian taxa H. armiger, H. larvatus, H. diadema and H. lekaguli formed a highly supported monophyletic clade basal to all other Hipposideros species, whose associations were also recovered in the combined RAG1 and ND2 analysis of Murray et al. (2012). While the phylogeny of Murray et al. (2012) contained a number of cryptic species and polytomies, they did find strong support for a clade that grouped together H. ater, H. khaokhouayensis and H. pomona, which was also recovered in this analysis. This clade of Asian taxa, including H. halophyllus, formed a well-supported sister taxa relationship with the African taxa H. caffer and H. abae, with H. galeritus basal to this African/Asian clade (see Fig. 1). These phylogenetic relationships suggest that extant Hipposideros likely diverged in Asia before dispersing back into Africa where the family Hipposideridae is thought to have originated (Foley et al., 2015), although this remains to be formally tested and confirmed with greater species sampling and detailed biogeographical analysis. However, providing a robust biogeographical analysis of this genus will be non-trivial owing to known differences in topologies obtained from exon and intron data relating to the position of Macronycteris (Foley et al., 2015), as well as difficulties in attempting to provide sufficient sampling of this geographically widespread speciose group, a problem which is known to greatly effect biogeographic estimates (Ruedi et al., 2012; Foley et al., 2015).
Intra-species relationships within the commersonii group are poorly resolved using the intron dataset. This is likely due to a recent and rapid diversification, as indicated by short branch lengths between component taxa in Fig. 1 as well as divergence time estimates from Fig. 3 (discussed below). Recently, in an examination of cryptic diversity in Malagasy M. commersonii, Rakotoarivelo et al. (2015) described three clades; clades B and C referable to M. commersonii s.s. and a highly divergent clade A, which was subsequently recognised as a new species, M. cryptovalorona (Goodman et al., 2016); the latter was more closely associated with continental African M. gigas and M. vittatus. Our Cyt-b analysis shows that M. commersonii s.s. samples sequenced as part of this study (FMNH 217931 and FMNH 209236) are part of the M. commersonii clade B.
Following phylogenetic analysis of Cyt-b, the monophyletic M. gigas clade contained two samples identified in the field as M. vittatus (FMNH 219682 and ML-129), as well as a sample field identified as H. commersonii and best referred to as H. vittatus/ gigas (72). The relationships within African members of the genus Macronycteris are poorly resolved and a larger and wider geographic sampling is needed. This approach should be combined with the utilization of faster evolving markers such as nuclear microsatellites, which will probably be necessary to resolve these relationships, as well as provide clearer insight into the delimitation of the M. vittatus and M. gigas clades, and associated morphological characters that diagnose these species. These problems highlight the need for routine and systematic molecular identification of samples thought to belong to cryptic species group, such as Macronycteris, or other cryptic complexes, which are widespread among bats.
Inclusion of mitochondrial and nuclear markers is recommended for analysis of cryptic species complexes following a recent study of two Pipistrellus kuhlii lineages across its European range, where mito chondrial data suggested the presence of two spe cies and nuclear (microsatellite) data provided evidence for one (Andriollo et al., 2015). These authors showed that in this instance the mitochondrial data was confounded by past population dynamics and so highlights the importance of a combined, diverse data approach, which is of particular importance when dealing with cryptic species complexes where alternative data sources such as morphology or echolocation data are uninformative (Ramasindrazana et al., 2011; Furman et al., 2010; Goodman et al., 2012). Studies should systematically include informative nuclear DNA if the aim is to resolve the taxonomy as inferences made only from mitochondrial DNA can be misleading (see Dool et al., 2016 and references therein). Future work to resolve relationships within Macronycteris would also benefit in particular from analyses of echolocation call frequencies in an effort to provide aids for field identification of these cryptic species. This type of acoustic information is particularly promising as echolocation variations were already described in H. commersonii (=M. gigas/vittatus) from mainland Africa, including differences from sympatric individuals also differing in size (Pye, 1972).
Divergence Time Estimates
In general, divergence time estimates for the hipposiderid tree were congruent with estimates obtained from previous studies (Foley et al., 2015; Rakotoarivelo et al., 2015). Our analysis shows that the split between Aselliscus stoliczkanus and Coelops frithii occurred 14 mya during the Miocene, which is in close agreement with estimates obtained elsewhere in the literature (Li et al., 2007; Lavery et al., 2014; Foley et al., 2015). Our analysis also supports an Oligocene split between Hipposideros from Aselliscus and Coelops around 28 mya. The crown group Hipposideros began to radiate 16 mya, during the Miocene (Li et al., 2007; Taylor et al., 2012; Lavery et al., 2014; Foley et al., 2015; Rakotoarivelo et al., 2015). Macronycteris and Doryrhina split from Aselliscus, Coelops and Hipposideros 31 mya during the Oligocene. This relatively deep divergence, which lends further support for the resurrection of Macronycteris, is independently supported by several molecular clock analyses of Cyt-b (Taylor et al., 2012; Rakotoarivelo et al., 2015). We show that D. cyclops diverged from Macro nycteris 19 mya during the Miocene (Taylor et al., 2012; Rakotoarivelo et al., 2015). Our analysis indicates that the crown form of Macronycteris began to diverge relatively recently at 3 mya. This estimate, in conjunction with the short branch lengths observed for this group (see Fig. 1), suggest that this group is characterised by a rapid and recent diversification event. Our estimate is also in line with the 5.81 mya obtained by Rakotoarivelo et al. (2015). While it remains to be confirmed with increased taxonomic sampling, the biogeographic analysis of Foley et al. (2015) indicated the crown Macronyc teris group diverged in Africa. The timing of the rapid radiation 3 mya in this group is consistent with the Late Pliocene Faunal Turnover observed in Africa at this time (Behrensmeyer et al., 1997; deMenocal, 2004), which has also been suggested to have been a potential driver in the diversification of African Rhinolophus (Dool et al., 2016). Climatic conditions at that time induced a change towards cooler, drier climates, an expansion of open habitats with persistent woodlands and forests that ultimately are thought to have led to an increase in mammal diversity and potentially changes in speciation rates. These conditions may have contributed to the rapid radiation observed in the commersonii species group at this time.
Navigating the Minotaur's Labyrinth
This study represents a further step towards navigating the Minotaur's labyrinth by resolving outstanding taxonomic issues surrounding the speciose hipposiderid genus Hipposideros. This study provides evidence from phylogenetic analysis, divergence time estimates, morphological and karyological data to support the resurrection of the genus Macronycteris for several species previously placed in the commersonii group and Doryrhina for H. cyclops. Given the cryptic nature of many taxa in the family Hipposideridae, molecular studies carried out in recent years have yielded taxonomic revisions at the level of family (Foley et al., 2015), genus (this study; Thong et al, 2012c), and in numerous cases species (Thabah et al., 2006; Monadjem et al., 2010; Murray et al., 2012; Thong et al., 2012a, 2012b; Monadjem et al., 2013; Rakotoarivelo et al., 2015; Tsang et al., 2016). While it is certainly no trivial undertaking, given the speciose nature of Hipposideros, the next step in navigating the Minotaur's labyrinth should focus on the evaluation and resolution of phylogenetic affinities between the Hipposideros species groups and species' delimitation. On a broad scale, our study highlights the shortcomings of present chiropteran taxonomy in describing hidden diversity within cryptic complexes of species and much work remains to be done to navigate the Minotaur's labyrinth of hipposiderid taxonomy, as well as more broadly across bats.
Acknowledgements
We are grateful to the Département de Biologie Animale, Université d'Antananarivo; Madagascar National Parks; and the Ministère de l'Environnement et des Forêts of Madagascar for their support and for providing research permits to conduct work on Madagascar; this was done in conjunction with Erwan Lagadec, C. Fabienne Rakotondramanana, Beza Ramasindrazana, Leigh Richards, Corrie Schoeman, Peter Taylor, and Mboho ahy Tsibara. Raphaël Colombo, Pipat Soisook and Vu Dinh Thong kindly provided samples. Velizar Simeonovski kind ly drew Figure 4. We also thank the European Research Council, ERC-2012-StG311000 for supporting ECT and NMF.