Even with the present analysis, the family Rhinopomatidae remains an enigmatic group whose history, taxonomic content, patterns of variation and phylogenetic relationships are far of being properly comprehended. Nevertheless, the data summarized here substantially improve the scarce information on these subjects. We will discuss them in regard to (a) composition of the group, (b) possible phylogeographic patterns and (c) evolutionary history.
Composition of Rhinopomatidae
The analysis of the family Rhinopomatidae by Van Cakenberghe and De Vree [9] demonstrated that the genus consists of four species that differ in the shape of the palatal incision, the rostral ridges, the narial swellings, in the relative length of the tail, and in overall body size. The most distinctive in all these characters was R. microphyllum, whereas the differences between the remaining forms were less pronounced, exhibiting a broad measure of overlap in most metrical characters. Our analysis (n = 252) provided the same picture (Figure 2). All of these data suggest that the major phenotypic divergence within the genus is that between R. microphyllum (including R. m. kinneari, R. m. sumatrae and R. m. asirensis) and the remaining forms, which thought to be closely related to a medium-sized species, R. hardwickii (Figure 2c), see [18].
Our mtDNA study confirmed the existence of the same three deep lineages recognized as morphospecies by Koopman [4], Hill [8] and Van Cakenberghe and De Vree [9] (another recognized species, R. macinnesi, was not included in our comparison). In addition, we found (i) deep divergence within the R. hardwickii lineage, (ii) incongruency between genetic and phenotypic phylogeographic patterns in clade II, (iii) deep divergence within the R. muscatellum lineage, (iv) a very shallow distance between the samples of R. microphyllum, suggesting an unexpected genetic homogeneity of that species. Last but not least, we demonstrated that (v) R. muscatellum (including the Yemeni population) is not a sister group of hardwickii, but of microphyllum. All these results contradict the standard view of the taxonomic structure of the family (Table 1), as well as of its distributional history [e.g. [9, 15]] and call for a brief comment.
Phylogeographic patterns
(i) The genetic divergence found within R. hardwickii s.l.(= the R. hardwickii clade) splits the corresponding morphospecies into an Iranian clade I (R. hardwickii s.s.) and Afro-Arabian clade II (R. cystops). While there is a clear genetic continuity between the Levantine and Yemeni populations (e.g. haplotypes LE1, YE2, which are separated by approximately 3,000 km, differ by only 1 mutation step), the much smaller geographic distance between the Levantine and Iranian samples (approx. 1,200 km) is combined with deep genetic dissimilarities (the minimum genetic distance between haplotypes LE1 and IR3 is 34 mutation steps). We expect the divergence between these two groups represents real phylogeographic structure, a break crossing the Middle East from the north-west to the south-east. The boundary might be situated along the southern part of the Zagros Mountains, which represents a significant distribution barrier to many clades [17, 19]. Unfortunately, knowledge of the distribution of bats in upper Mesopotamia is too scarce [19] to allow further discussion. Thus, we are unable to answer whether there is continuous distribution of haplotype frequencies with a clinal transition between geographic extremes, whether there are two allopatric ranges separated by a distinct geographic gap, or whether the ranges meet at a distinct zone of parapatry or sympatry. Because of the extent of the genetic dissimilarity, we are rather skeptical about the first alternative. Rather, we expect that clades I and II are entities separated at species level. We propose a separate species status for the two clades as per the genetic species concept [20–22] which sets a cutoff based on empirical data (cytochrome b in the order Chiroptera) of about 5% of corrected sequence divergence [e.g. [23, 24]]. These two groups have almost double that divergence with 9% corrected divergence.
(ii) Within the clade II, a divergence of about 3% separates African and Asian haplotypes of R. hardwickii. Within the African group, our genetic data contradict the groupings proposed by previous studies [see [8, 9, 17] and [25]], which stress a separate status for the populations of the central Sahara (including that of Upper Egypt). The genetic relatedness of these small bats to the largest form in Libya suggests an unexpected degree of phenotypic plasticity in these bats, apparently driven by temporary local conditions rather than by the genotypic backgrounds of the respective populations. Here, the selection pressures of the extreme conditions of desert habitats may have played a key role. A similar pattern of morphological bimodality has been observed in other desert or semi-desert species of bats [16, 25, 26], such as Taphozous nudiventris Cretzschmar, 1830, Rhinolophus clivosus Cretzschmar, 1828, Asellia tridens (Geoffroy, 1813), or Pipistrellus kuhlii (Kuhl, 1817), and such an explanation could be also invoked with respect to the smaller Arabian form, Rhinopoma microphyllum asiriensis Nader and Kock, 1983.
(iii) Our study has revealed that the morphospecies R. muscatellum is composed of two distinct clades: clade III in Yemen and clade IV in Iran and, supposedly, in Oman. This split is supported by morphometric differences (Fig. 2). Recent allopatry is more obvious in this case because clades III and IV are geographically isolated by the Arabian Desert. However, geographic positioning of major genetic breaks in R. hardwickii and R. muscatellum lineages coincides with this division (Fig. 4). With respect to the genetic species concept, it is reasonable to consider species status also for clades III and IV [for taxonomic rearrangements in the R. muscatellum lineage, see Additional file 1].
(iv) Considering the relatively deep genetic divergences within the morphospecies R. hardwickii and R. muscatellum (in the sense of Van Cakenberghe and De Vree [9]), the surprisingly low degree of geographic divergence of mtDNA in R. microphyllum calls for a comment. At least two qualities of this species are worth discussing in this connection: (i) its larger body size, and (ii) the well-pronounced seasonality of its life cycle and reproduction, including regular seasonal movements [e.g. [27]]. Both of these factors may contribute to increases in vagility and the rate of gene flow.
(v) The sister status of the R. microphyllum and R. muscatellum phylogroups contradicts traditional arrangements of the family where R. hardwickii and R. muscatellum are considered as the most closely related taxa based on similarities in narial morphology and body size (Figure 2c) [18, 28]. The morphological polymorphism in genetically uniform populations of R. cystops (clade II) and R. microphyllum (clade I) does, however, indicate that the body size can undergo rapid rearrangement regardless whether in reaction to environmental conditions or as a character displacement due to interspecific interactions. Worth mentioning in this context is the large body size of the fossil Rhinopoma aff. R. hardwickii, which clearly exceeds the limits of the Recent R. hardwickii to which the fossil form is linked by its morphological characters. All these cases suggest that body size, traditionally applied as a significant character in taxonomy of the genus (because of considerable uniformity in other morphological characters) is controlled by ecological factors rather than by a strict taxon-specific developmental constraint.
Evolutionary history
The evolutionary history of Rhinopomatidae is a subject of particular interest, one which makes the group one of the most enigmatic clades of chiropterans. In the traditional view, Rhinopomatidae were regarded as the most primitive group of extant bats, the closest to the common ancestor of microbats and megabats [12, 14]. Indeed, compared to other families of Yangochiroptera and Yinpterochiroptera, the family Rhinopomatidae exhibits a set of unique plesiomorphies: (i) the trochiter of the humerus (tuberculum minor) is small and does not permit the scapulo-humeral lock found in other bats; (ii) the wing tip index has the lowest value of all Chiroptera; (iii) medial phalanx of the second wing finger is complete and well-ossified; (iv) the last cervical and first throracic vertebrae are free (not fused as in other bats); (v) individual sacral vertebrae have distinct boundaries; (vi) the uropatagium is incomplete; (vii) the calcar is absent; (viii) tail is long and mouse-like, not entirely integrated to the uropatagium; (ix) the premaxillae are not attached to each other or to maxillae; and (x) the premaxillae are developed at the palatal plane only. A few of these characteristics (i, iv, vii) are shared with Craseonycteridae, while the others are unique among both Yangochiroptera and Rhinolophoidea, partly resembling the condition in Pteropodidae (i, iii, iv, v, vi, ix, partly vii, viii, x).
In contrast to the major clades of Yinpterochiroptera [cf. [3, 15]], Rhinolophidae (1 genus, ca. 77 species), Hipposideridae (9 genera, ca. 81 species), and Pteropodidae (42 genera, ca 184 species), the family Rhinopomatidae is much less diversified [15]. In that respect it is similar to Megadermatidae (4 genera, 5 species) and Craseonycteridae (1 genus, 1 species), which are the sister clades of Rhinopomatidae according to the recent molecular data [3].
The present paper dates the beginning of radiation of extant clades of Rhinopomatidae (i.e. the separation of the R. hardwickii clade), to about 29 Ma in the Oligocene. Nevertheless, the datum is apparently not relevant for the beginning of the family which arose with the earliest divergence of Rhinolophoidea, which molecular clock studies place at 50–55 My [3, 7]in the Early Eocene. In contrast to other groups of Rhinolophoidea, whose early divergence is well represented in the fossil record, no such information is available for Rhinopomatidae and Craseonycteridae. In contrast to Creaseonycteriae, Rhinopomatidae occupies quite a large range comparable to that of other rhinolophoid families. At least for that reason, the absence of fossil record is unusual and calls for comment, at least as a background story to the discussion on meaning of the first Neogene record of the family reported in this paper.
Despite the fact that the fossil record of bats is sometimes regarded as being quite a poor [3], it is actually rich enough to enable discussions on major differences in phylogeny and early paleobiogeography of particular chiropteran clades at least in Europe and northern Africa. The remains of bats, including rich and taxonomically diversified assemblages, have been found in more than 130 European and North African sites of the Late Eocene, Oligocene and Early Miocene age [29–31] and current views on the structure of chiropteran fauna and the history of particular clades during that period [32] can be considered relevant and reliable. Among Rhinolophoidea, Hipposideridae and Rhinolophidae are particularly rich in their fossil record and, with a number of divergent clades, they have predominated the fossil assemblages in Europe, Africa, and even Australia since the Late Eocene [32–34]. In contrast, no relevant fossil record is available for Craseonycteridae or (until this paper) for Rhinopomatidae. The situation with Megadermatidae is more intricate. The first evidence of appearance of true Megaderma in Europe comes from the Upper Oligocene (MP25 Carrascosa del Campo, Spain [35]; MP29 Saint-Victor-la-Coste, France [36]; MP 29 Herrlingen 9, Germany [37]) and a number of further records are of Miocene and Pliocene age [37]. In contrast to hipposiderids or nycterids, the family is absent from African Oligocene sites (including Fayum or Taquah in Oman [30, 38] but appears in the Lower Miocene of Thailand and even in Australia [39]. The Late Eocene to Early Oligocene genus Necromantis, often regarded as a megadermatid [31], differs from true megadermatids in several characters (including basisphenoidal pits, a key character of emballonurids, which is invariably absent in rhinolophoids) and most probably does not belong to that stock. The absence of Rhinopomatidae and the late first appearance datum of Megadermatids in the fossil record contrasts with the fact that other groups such as Emballonuridae, Hipposideridae, Rhinolophidae, Molossidae, and Vespertilionidae s.l. are constant components of the western Palaearctic and African fossil record since the late Eocene [31, 29, 40]. All had already produced a number of subclades during the Oligocene and Early Miocene [32, 37, 41]. The absence of any rhinopomatids in the fossil record is surprising because these bats differ from all others in a number of conspicuous dental and skeletal specificities by which they are easily distinguishable, even based on a single fragmentary tooth. Moreover, rhinopomatids are cave-dwellers, which predisposes them to be particularly common in the fossil record. Under such conditions their absence in fossil record can be interpreted as a real fact which most probably reflects actual absence of the group in the western Palaearctics prior to the Miocene.
The fact that the phenotype of Rhinopomatidae (similarly as in Craseonycteridae) is composed almost exclusively of the ancestral characters not affected by adaptive rearrangements common in other chiropteran families, in contrast to other chiropteran families [42], suggests that (i) the clade was established at a very early stage of chiropteran radiation (prior to the first appearances of modern families, in the Middle Eocene or earlier), and (ii) that rhinopomatids were relatively little affected by the same adaptive processes that affected other all bat families, which all evolved under constant competitive pressure from other microbat clades. The latter could happen only under conditions of long-term isolation of rhinopomatids from other bats. An analogous case is the extinct clades of Palaeochiropterygidae and Hassianycteridae, which extensively diversified in Central Europe during geographic isolation of that region in the Early and Middle Eocene [43, 44]. These endemic groups were radically replaced by modern bat families soon after their invasion during the Late Eocene "grand coupure" [33]. The respective modern clades, Hipposideridae, Rhinolophidae, modern Emballonuridae, Vespertilionidae, Molossidae [comp. [29, 31]], arrived either from Africa and or Asia, and their early radiations most probably took place there (comp. also [3] for molecular support to that hypothesis). The complete absence of rhinopomatids in the fossil record and the lack of coevolutionary influence on their phenotype suggest that this group was absent in Europe and probably also in Africa and Asia. Of course, Tanzanycterididae with Tanzanycteris mannardi from the Early Lutetian (46 Ma) of Tanzania [45] may ultimately be shown to be closely related to Rhinopomatidae. Unfortunately the characters available in the specimen of Tanzanycteris provide only tentative support for such a possibility (e.g. enlarged cochlea, a lack of scapulo-humeral lock which is, common to all other Eocene bats).
The first appearance datum of Megadermatidae s. str. is nearly synchronous with dramatic rearrangements of the European mammalian fauna, with the appearance of the Asiatic elements (e.g. Cricetidae) and a considerable contribution of non-mammalian taxa from the Indian and Indomalaysian provinces such as Gavialosuchus, Tomistoma, and Varanidae. The spread of these taxa into Europe has been dated to ca. 18 Ma [46]. Recent paleogeographic analyses [47, 48] supplement the picture with further data that show continuity between the Mediterranean-Iranian and the eastern Indian-east African marine provinces until the final disappearance of the Western Tethyan seaway in the Early Miocene.
In case of Rhinopomatidae, no such evidence is available. The Late Miocene fossil record reported in this paper is apparently not related to the early history of the clade. Nevertheless, the results of molecular studies provide valuable information. The first dated split within the family (R. hardwickii s.l. vs. R. microphyllum-muscatellum: 28.1 Ma) shows no phylogeographic signal – both clades are broadly sympatric. Nevertheless, the next cladogenetic events (R. microphyllum vs. R. muscatellum: 20.9 Ma) have clear phylogeographic correlates. In the later events, the phylogeographic signals are even more pronounced: clades I vs. II (14.1 Ma): Iran vs. Levant to Africa, and clades IV and III (10 Ma): Iran vs. Yemen. According to traditional biogeography [49, 50], the region with the largest concentration of taxonomic diversity is the most likely candidate for being the source area of the group in question. In the case of rhinopomatids, the present results would suggest Iran to be such a candidate. At the very least, these results suggest that Iran was an area of paleoendemism that played host for the ancestral clades more than 11 Ma ago. Nevertheless, the terrestrial conditions appeared in the respective region first at the time of Oligocene/Miocene transition [51] and thus the source area of the clades that colonized at in that time was apparently situated in other regions.
Based on the above discussion, we proposed the following biogeographic hypothesis (Figure 5):
(i) Rhinopomatidae originated during the Eocene from the early diversification of rhinolophoid bats that remained isolated from competitive pressure of other chiropteran clades somewhere in the archipelago south of the Western Tethyan seaway or in India.
(ii) The group evolved in isolation until the Oligocene when the marine barrier between the Mediterranean and Indian Tethys provinces disappeared. Endemic adaptation to major chiropteran foraging strategies in ancient rhinolophoids produced clades whose ecomorphological design was much different corresponding foraging specialists on neighbouring continents. Some of those ancient clades survive now: rhinopomatids as aerial foragers, craseonycterids as foragers of small prey in cluttered habitats, and megadermatids as ground gleaners.
(iii) The land between Iran and western India, uplifted during the Oligocene, was subsequently invaded by rhinopomatids and became the key location of their early Neogene radiation. The westward invasion was from the south, then to Arabia (which came in contact with the Iran belt some 20 Ma ago) with the R. muscatellum lineage, and perhaps later to the northern part of the Iran and to the Mediterranean with the R. hardwickii lineage. The paleobiogeographic analyses of rodents [52] suggest that the respective westward migrations may begun even much earlier – in the late Eocene and early Oligocene via archipelagos south of the Western Tethyan seaway.
(iv) Extension of the paleogeographic and paleoenvironmental rearrangements of the Middle East during the Vallesian and Turolian stage, ca. 11 Ma [compare with [53–55]], fixed the already established divergences among the clades within both R. hardwickii and R. muscatellum lineages.
(v) The extension of the range of the large-sized Indian R. microphyllum may have appeared quite late after these events, possibly even during the Quaternary under the influence of more pronounced seasonality in the climate [53].