Cryptic species diversity in the genus Tylonycteris

Previous studies have detected high levels of genetic and karyological variation among specimens identified as either T. pachypus or T. robustula collected from different geographic locations in mainland Southeast Asia (Francis et al. 2010; Huang et al. 2014). Our COI analyses further revealed the existence of three divergent geographic haplogroups for both T. pachypus and T. robustula, for each corresponding to Sumatra, northern Indochina and southern Indochina (but also northwestern India and Peninsular Malaysia for T. robustula). Our Cytb dataset confirmed the existence of these geographic haplogroups in mainland Southeast Asia. In addition, a specimen of T. pachypus collected from Borneo and originally described as T. robustula was found to be highly divergent from the two other Indochinese haplogroups (Fig. 2). In the absence of Cytb data for Sumatran specimens, it is impossible to know whether they share the same mitochondrial lineage as those from Borneo.

Taken together, our mtDNA analyses show that all haplotypes sequenced for insular Tylonycteris are very divergent from those identified in mainland Southeast Asia. However, genetic inferences based on the maternally inherited mitochondrial genes are prone to be discordant with the true evolutionary history of the taxa, owing to various evolutionary processes, such as mtDNA introgression, incomplete lineage sorting or female philopatry (Avise 2000; Ballard & Whitlock 2004; Hassanin & Ropiquet 2007; Nesi et al. 2011; Rivers et al. 2005). Here, the geographic pattern of mtDNA diversity observed for the two species of Tylonycteris could be the consequence of female philopatry, i.e., the behavior of remaining in, or returning to the natal territory. Indeed, bat species with philopatric females generally display high geographic structure when relationships are examined with maternally inherited markers, such as the mitochondrial DNA. This pattern can disappear with biparentally inherited markers when adult males are able to disperse over long distances, allowing gene flow between otherwise isolated populations (Castella et al. 2001; Hassanin et al. 2015; Hulva et al. 2010; Pereira et al. 2009; Rivers et al. 2005). Behavioral and population genetic studies in southern China have shown that T. pachypus bats are philopatric to their natal area and that philopatry is especially pronounced in females (Hua et al. 2011, 2013). Although no data are available for T. robustula, female philopatry can also be predicted for this species, because it shares similar morphological, behavioural and ecological traits with T. pachypus (Medway 1972; Medway & Marshall 1970, 1972; Zhang et al. 2007). The social organization of T. pachypus and T. robustula, combined with their fragmented habitats, is therefore expected to result in limited gene flow between populations, especially among matrilines from distant geographic localities. For both species, this prediction is corroborated by the analyses of mtDNA markers, with the identification of three divergent geographically non-overlapping haplogroups. For T. robustula, this phylogeographic pattern is also supported by the nuclear sequence data, as the two Indochinese clades were recovered monophyletic with all the three introns containing enough nucleotide variation at the intra-specific level, i.e., CHPF2, HDAC1 and TUFM (Appendix 8). By contrast, our nuclear analyses (Fig. 3; Appendices 4, 8) do not support the reciprocal monophyly of the two Indochinese clades of T. pachypus, suggesting that gene flow was maintained by male dispersal or, alternatively, that their separation was too recent to be detected with our nuclear genes.

Nucleotide distances estimated from mtDNA genes between northern and southern Indochinese populations of T. robustula were more than twice those of T. pachypus (6.5% vs 2.8% in COI; 9.5% vs 2.8% in Cytb), indeed supporting a more recent divergence for the latter taxon, if we assume equal evolutionary rates. Similarly, the nuclear distances between the two Indochinese clades of T. robustula were between 0.41 and 0.56%, which is more than twice those calculated between Indochinese individuals of T. pachypus (0–0.2%; Appendix 5) and in the range of interspecific distances found in other groups of Laurasiatheria, such as fruit bats of the tribes Myonycterini (Nesi et al. 2013) and Scotonycterini (Hassanin et al. 2015), or cattle and buffalo of the tribe Bovini (Hassanin et al. 2013). Although none of the nuclear markers could be sequenced for Sumatran and Bornean Tylonycteris, their high mtDNA divergence from Indochinese populations (> 5.7 % in both COI and Cytb sequences; Appendix 5) suggests they might represent distinct lineages based on nuclear markers as well. In agreement with this view, our multivariate morphological analyses revealed that Indochinese bats of the T. pachypus complex constitute a distinct group separated from those collected on Sumatra. For the T. robustula complex, morphological overlap between haplogroups is more extensive, but pairwise comparisons of their PC mean scores support the distinctness of adjacent geographical taxa, such as Tr2 and Tr3 on the Southeast Asian mainland.

The close morphological similarity among taxa of Tylonycteris suggests that they have evolved under the influences of similar and specialized habitats, i.e., woody bamboo vegetation. Molecular evidence indicates, however, that T. pachypus should be split into at least two distinct species, T. pachypus on the Sunda islands (Sumatra and/or Borneo) and T. fulvida in mainland Southeast Asia, and that T. robustula should be divided into at least three species, with T. robustula on Sumatra, T. malayana in southern and western mainland Southeast Asia, and T. tonkinensis sp. nov. in northern Indochina.

The evolution of Tylonyteris spp. in Southeast Asia during the Pleistocene

Given that both species complexes, here named T. pachypus s. lat. and T. robustula s. lat., are usually found in sympatry across most of their geographic ranges in Southeast Asia, they are expected to share a common phylogeographical history. Our estimates of divergence times based on mtDNA sequences suggest that the genus Tylonycteris diversified during the Pliocene epoch (Cytb: 5.92 ± 0.65 Mya; COI: 4.56 ± 0.72 Mya) (Table 1). During the Miocene and until the early Pliocene, Southeast Asia was generally covered by large tracks of rain forests as a consequence of warm and humid climatic conditions (Meijaard & Groves 2006; Morley 2000). Thus, ancestors of both Tylonycteris species complexes were presumably widely distributed across Southeast Asia during the Pliocene.

Our molecular dating estimates indicate that the basal geographic splits within the two species complexes, i.e., between mainland Southeast Asia and Sumatra, took place approximately at the same time during the Early Pleistocene (between 2.70 and 1.96 Mya for T. pachypus, between 3.07 and 2.22 Mya for T. robustula; Table 1). The Pleistocene epoch is characterized by the onset of repeated cycles of cold glacial and warm interglacial periods as the results of the glaciations/deglaciations of the Northern Hemisphere, which implied contraction and expansion of rain forests in Asia (An et al. 2001; Meijaard & Groves 2006; Morley 2000). As bats of the genus Tylonycteris are highly dependent on woody bamboo vegetation for roosting, foraging and mating (Kunz 1982; Medway 1972; Medway & Marshall 1970, 1972), their Pleistocene biogeographic history was firmly constrained by the distribution of such bamboo habitats. The current distribution of woody bamboo species in Asia (Bystriakova et al. 2003b; Fig. 6) indicates that eight disjunctive biogeographic regions have higher species richness (> 5 species) for some bamboo genera: southern India (Ochlandra Thwaites), northern Myanmar (Cephalostachyum Munro), southern China (Dendrocalamus Nees and Bambusa Schreb.), Hainan Island (Bambusa), northwestern Thailand (Dendrocalamus and Gigantochloa Kurz ex Munro), Peninsular Malaysia (Gigantochloa), Sumatra and Borneo (Gigantochloa and Schizostachyum Nees). All these regions may therefore have acted as distinct bamboo forest refugia during the glacial periods of the Pleistocene (Fig. 6). Evidence for a number of these postulated glacial refugia has been reported in previous studies for many organisms, including bats (Flanders et al. 2011; Khan et al. 2010; Lin et al. 2014; Mao et al. 2013). Accordingly, we propose that the contraction of woody bamboo forests into different glacial refugia had fragmented the distribution of the Pliocene ancestors of both T. pachypus s. lat. and T. robustula s. lat. In addition, we can assume that Pleistocene glacial periods resulted in higher interspecific competition between co-distributed species of Tylonycteris, because the supply of most suitable resources was more limited in glacial bamboo forest refugia (Medway & Marshall 1970). As noted in previous studies, T. pachypus s. lat. has a more manoeuvrable flight in cluttered habitats and forages on smaller insects than T. robustula s. lat. (Zhang et al. 2005, 2007). Moreover, Medway & Marshall (1970) found that the smaller T. pachypus s. lat. can roost in the internodes with small entrance holes, which the larger T. robustula s. lat. is unable to enter. These differences suggest, therefore, that the smaller T. pachypus s. lat. have greater advantages than the larger T. robustula s. lat. in interspecific competition when natural resources are limited. Hence, isolated populations of T. robustula s. lat. may have been more exposed to bottlenecks and therefore more vulnerable to local extinction than those of co-distributed T. pachypus s. lat.

During interglacial periods of the Early Pleistocene, warmer and humid conditions resulted in the expansion of woody bamboo forests, which in turn may have favored the restoration of connectivity between isolated populations of both complexes. However, the isolated populations may have been connected or not, depending on their dispersal capacity and the distances between refugia. In T. robustula s. lat., these processes may have taken longer, because of its lower population abundance (Lande & Barrowclough 1987; Shaffer 1981), and may have been prevented in cases of extinction of transitional populations (Huang et al. 2014 and references therein). This scenario is supported by the fact that T. pachypus s. lat. is usually found to be more abundant than T. robustula s. lat. in bamboo forests (Zhang et al. 2004; Medway & Marshall 1972) and by the wider geographic range of T. pachypus s. lat. (Bates et al. 2008a, 2008b; Fig. 1). Knowing this, the body size differences between the two species complexes may be the key factor explaining why the basal divergence of northern Indochinese populations occurred earlier in T. robustula s. lat. (i.e., T. tonkinensis sp. nov) than in T. pachypus s. lat., i.e., 2.97–1.70 vs 1.35–0.79 Mya (Table 1). During Pleistocene interglacials, exchanges of Tylonycteris spp. between the continent and the islands of Sundaland were probably prevented because of the long distances between the glacial forest refugia, as well as the higher sea levels (Fig. 1). Our molecular dating estimates corroborate this scenario, as continental populations of Tylonycteris spp. from Indochina and Peninsular Malaysia diverged from insular populations (Sumatra and Borneo) in the Early Pleistocene (Table 1).

Implications for conservation

Previous studies considered T. pachypus and T. robustula to be common species and thus classified them as Least Concern in the IUCN Red List (Bates et al. 2008a, 2008b). Since our study reveals that both species in fact represent several species with more restricted distributions, the IUCN status of the different taxa should be reassessed urgently, including that of the new species, T. tonkinensis sp. nov. In addition, our study suggests that several biogeographic regions have acted as Pleistocene glacial refugia. This information is very important for developing more effective conservation strategies, particularly given the high current rates of deforestation affecting most natural habitats in Southeast Asia (Kingston 2010; Sodhi et al. 2010; Tordoff et al. 2012).