Across the great divide: revision of the genus Eupetaurus (Sciuridae: Pteromyini), the woolly flying squirrels of the Himalayan region, with the description of two new species

The woolly flying squirrel, Eupetaurus cinereus , is among the rarest and least studied mammals in the world. For much of the 20 th century it was thought to be extinct, until it was rediscovered in 1994 in northern Pakistan. This outlines the first taxonomic and biogeographical review of the genus Eupetaurus , which until now has contained only a single species. Careful review of museum specimens and published records of Eupetaurus demonstrates that the genus occurs in three widely disjunct areas situated on the western (northern Pakistan and north-western India), north-central (south-central Tibet, northern Sikkim and western Bhutan) and south-eastern margins (north-western Yunnan, China) of the Himalayas. Taxonomic differentiation between these apparently allopatric populations of Eupetaurus was assessed with an integrative approach involving both morphological examinations and molecular phylogenetic analyses. Phylogenetic reconstruction was implemented using sequences of three mitochondrial [cytochrome b ( Cytb ), mitochondrially encoded 12S and 16S ribosomal RNA (12S, 16S)] and one nuclear [interphotoreceptor retinoid-binding protein ( IRBP )] gene fragment. Morphological assessments involved qualitative examinations of features preserved on museum skins and skulls, supplemented with principal components analysis of craniometric data. Based on genetic and morphological comparisons, we suggest that the three widely disjunct populations of Eupetaurus are each sufficiently differentiated genetically and morphologically to be recognized as distinct species, two of which are described here as new.


INTRODUCTION
Tibet and Yunnan are deeply divergent genetically from those in the western Himalayas, probably reflecting a need to recognize more than one species in this genus. There are also clear morphological distinctions between allopatric populations of Eupetaurus (Jackson & Schouten, 2012). Therefore, the aim of this study was to undertake an integrative taxonomic review of the genus Eupetaurus across its distribution and make taxonomic recommendations based on interpretations of morphological and molecular data.

SpecimenS
A total of 24 specimens of Eupetaurus are known in natural history museums, including six specimens in the Natural History Museum, London, UK (BMNH); one in the Naturalis Museum, Leiden, The Netherlands (NML); four at the Bombay Natural History Society, Mumbai, India (BNHS); two in the Zoological Collection of the Zoological Survey of India, Kolkata, India (ZSI); two in the Florida Museum of Natural History, Gainesville, FL, USA (FLMNH); and nine in the KIZ (Supporting Information, Supplementary Data SD1). We have examined all of these specimens except the series at BNHS, but the only skull from the BNHS (specimen #7108, now missing according to Zahler & Woods, 1997) was measured and photographed by Joseph Curtis Moore (McKenna 1962). This allowed us to include it in our comparisons. Most available specimens were skins, and only nine skulls were available for study (two juveniles and seven adults; Supporting Information, Supplementary Data SD1). We also made comparisons with specimens of various other flying squirrel genera in these and other collections, including the American Museum of Natural History, New York, NY, USA (AMNH), the Field Museum of Natural History, Chicago, IL, USA (FMNH) and the National Museum of Natural History, Smithsonian Institution, Washington, DC, USA (USNM).

morphometric analySiS
Dental observations were made under a light microscope (Keyence digital microscope system VHX-2000C with a ×50 Kenyence VH-Z00R lens). Morphological nomenclature of the cheek teeth follows a modified version of that by McKenna (1962) and Li et al. (2021) (Fig. 1). Dental abbreviations in the text are D for deciduous, P for premolars, M for molars, superscript (P x , M x ) for upper premolars and upper molars, and subscript (P x , M x ) for lower premolars and lower molars.
A principal components analysis based on the covariance matrix was used to evaluate variation in all 34 craniodental variables taken from seven available adult skulls, which were divided into three allopatric populations (west Himalayan population, IMK 9492, FLMNH 28583 and BNHS 7108;central Himalayan population, NML 19524; and east Himalayan population, KIZ 034190, KIZ 035087 and KIZ 035088). However, owing to the small sample size, this was only an exploratory analysis. Missing data were filled with the mean of the population. The cheek teeth of FLMNH 28583 were worn excessively; therefore, we replaced its tooth-related measurements with the mean of the population. Raw measurements were log 10 -transformed to standardize variances and improve normality. All analyses were performed using PAST v.4.02 (Hammer et al., 2001).

Laboratory protocols
Total genomic DNA was extracted from the muscle tissues of four newly collected Yunnan specimens (KIZ 034189-KIZ 034192) using the DNeasy Tissue kit (Qiagen). Three mitochondrial [cytochrome b (Cytb), 12S ribosomal RNA (12S) and 16S ribosomal RNA (16S)] and one nuclear [interphotoreceptor retinoid-binding protein (IRBP)] gene fragment that are commonly used in mammalian molecular studies (e.g. Mercer & Roth, 2003;Yu et al., 2004) were selected for sequencing and analysis. All gene fragments were amplified with rtaq DNA Polymerase (Takara) [primers followed He et al. (2010) and Jansa et al. (2009) and are presented in Supporting Information, Supplementary Data SD2]. The polymerase chain reaction (PCR) products were purified with the QIAquick PCR Purification kit (Qiagen) and sequenced with the BigDye Terminator Cycle kit v.3.1 using an Applied Biosystems automated sequencer (ABI 3730xl). Sequencing reads were assembled and edited using the DNASTAR laSergene v.7.1.

Phylogenetic analyses
We downloaded all previously published sequences for Eupetaurus (Yu, 2002;Mercer & Roth, 2003;Yu et al., 2004), in addition to sequence data for various other flying squirrel taxa, from GenBank (Supporting Information, Supplementary Data SD3). We did not include the previously widely used Aeretes sequences [GenBank nos AY227593, AY227535 and AY227481] in our analyses, because these come from a misidentified specimen (Roth & Mercer, 2015: 223;Quan Li, pers. obs.). Squirrels from other sciurid subfamilies (Ratufinae and Sciurinae) and the mountain beaver [Aplodontia rufa (Rafinesque, 1817), family Aplodontiidae], which is the sister group to all squirrels (Huchon et al., 2002), were selected as the outgroup taxa for phylogenetic analysis.
All gene sequences were aligned using MUSCLE (Edgar, 2004) and assessed manually in MEGA 6 (Tamura et al., 2013). Pairwise Cytb divergences were calculated using the Kimura-2 parameter (K2P) model in MEGA 6 (Tamura et al., 2013). Our phylogenetic reconstructions were based on the mitochondrial DNA (mtDNA) plus nuclear DNA Figure 2. Craniodental measurements taken for this study. See main text for definitions of abbreviations. Owing to space constraints, for WPP 3 , WPP 4 , WPM 1 , WPM 2 and WPM 3 with similar measurement methods we show only the first and last one.
In ML analysis, the GTRGAMMA model was selected for each partition, and the rapid bootstrapping algorithm was set for running 500 replicates. In BI analysis, we used the relaxed uncorrelated lognormal clock models that allowed the rate in each branch to evolve independently. The BEAST analyses were repeated four times, each analysis for 50 million generations, with a sampling interval of 5000. TRACER v.1.6 was used to confirm the effective sample sizes (ESSs) > 200, and the first 25% of the generations were treated as burn-in. Both ML and BI analyses were performed using resources available in the CIPRES Science Gateway (Miller et al., 2010).

Molecular dating and species delimitations
Divergence times were estimated using BEAST v.1.7.5 (Drummond et al., 2012). The mtDNA + nDNA (12S + 16S + Cytb + IRBP) dataset was used for molecular dating analyses to avoid overestimation of divergence times. The oldest fossil squirrel, Douglassciurus jeffersoni (E. Douglass. 1902), dated to the late Eocene (~37.2-33.9 Mya), was used for calibration (Emry & Thorington, 1984;Goodwin, 2008). We used an exponential distribution for the prior of this calibration and set the lower boundary to 33.9 Mya. All other settings for the analysis were the same as those described for the previous phylogenetic reconstructions.
The single-threshold generalized mixed Yule coalescent (GMYC) model (Pons et al., 2006) was applied to delimit species boundaries and determine the number of putative species. Maximum-likelihood statistics and divergence times were used to estimate the split point from the species to the population level (using the R package 'splits'). We conducted this analysis based on the Cytb time-calibrated tree because the availability of Cytb sequences for all major geographical regions of occurrence for Eupetaurus (from Pakistan, Tibet and Yunnan) best facilitated our examinations of species delimitation in the genus. The Bayesian analysis was based on the Poisson Tree Process model (bPTP, http:// species.h-its.org/ptp/; Zhang et al., 2013) with the ML tree derived from the Cytb gene. In addition, the Bayesian species delimitation was evaluated further using BP&P v.3.1 (Yang & Rannala, 2010, analysed with both A10 and A11 models.

morphological compariSon
A principal components analysis was used to evaluate morphological variation in 34 craniodental variables among the three allopatric populations of Eupetaurus. The first component (PC1) separated individuals mainly on the basis of WIF, WM 1 , WM 2 , WLI, WM 3 and LAP (63.18% of variation), the second component (PC2) on the basis of WPP 3 , IOB, LCP, WN, WP 4 and WIF (16.76% of variation) and the third (PC3) on the basis of WPM 1 , WPP 3 and WPP 4 (8.51% of variation; Supporting Information, Supplementary Data SD4). The bivariate scatter plot of specimen scores on PC1 and PC2 clearly distinguished the three allopatric populations from each other (Fig. 3).
External body measurements and masses for each species from the available literature and specimens is found in Table 1. Measurements of variables for all nine available skulls is found in Table 2. Some easily observed distinguishing features of the three species of Eupetaurus are summarized in Table 3. Intergeneric relationships of Eupetaurus were not recovered consistently in different trees (Fig. 4). In the nDNA tree (Supporting Information, Supplementary Data SD5) and the ML tree of the mtDNA + nDNA dataset (Fig. 4A), Aeromys + Biswamoyopterus was recovered as the sister clade of Eupetaurus, whereas in the BI tree of mtDNA + nDNA dataset (Fig. 4B

Molecular divergence within Eupetaurus
Specimens from the three major geographical areas of occurrence for Eupetaurus (Pakistan, Tibet and Yunnan) all showed deep respective genetic divergences, grouping into two major clades, east and west. Divergences for Cytb measured 14-20% (K2P distance) between specimens from Pakistan and specimens from farther east (Tibet and Yunnan). Divergences for Cytb measured 4-5% (K2P distance) between the sample from Tibet and those from Yunnan. These are genetic divergences consistent with or greater than specieslevel distinctions in many other groups of mammals (Baker & Bradley, 2006), including other squirrel genera (e.g. Harrison et al., 2003;Baker & Bradley, 2006;Helgen et al., 2009). Genetic divergence within Eupetaurus samples from Yunnan was considerably smaller, consistent with intraspecific differences (K2P distance = 1.2%). Likewise, minimal genetic difference was observed within samples from Pakistan.

Molecular species delimitation
Cytb tree-based analyses using the single-threshold GMYC model recognized three putative species diverging > 0.51 Mya, indicating that all three regional groupings of Eupetaurus (from Pakistan, Tibet and Yunnan) might be recognized taxonomically at the species level (Fig. 5). This was confirmed in our bPTP analysis with strong Bayesian support (PP = 0.814, 0.947 and 0.942, respectively). This level of divergencetime depth is consistent with that between species of squirrels in various other genera (e.g. Hawkins et al., 2016a, b). Furthermore, the delimitation using BP&P has supported the new taxa (PP = 0.99 in A10 model and PP = 0.95 in A11 model; Supporting Information, Supplementary Data SD6).  Roberts (1997) and Zahler and Woods (1997 Eupetaurus are amongst the largest flying squirrels, with fur that is thick, long and soft. Dorsal coloration ranges from pale grey to brownish grey, and ventral coloration is typically whitish grey. The cylindrical, fox-like tail is long and bushy. The palmar and plantar surfaces are thickly furred between naked pads (Fig. 6). The skull is generally similar to other flying squirrels (reviewed in detail by McKenna, 1962), with the following diagnostic features (Fig. 7): (1) the snout is long and constricted at the base; (2) the incisors are ungrooved and relatively small, with pale yellow enamel faces; (3) the glenoid fossa opens posteriorly, with the opening placed high on the skull; (4) the coronoid process of the mandible is weak, not rising The abbreviations of skull measurements are defined in the main text and illustrated in Figure    above the condyle; (5) the cheek teeth are large, slow to erupt and strongly hypsodont in younger animals (the occlusal surfaces becoming nearly flat after wear); they are enormously expanded vertically (Supporting Information, Supplementary Data SD7), with cusps and their interconnecting ridges combined into lophs and cross-lophs (this remarkable hypsodonty is unlike any other sciurid); (6) the alveolar cavities are strongly swollen to house the unerupted portion of the hypsodont cheek teeth; and (7) 'the pterygoid fossae possess three diagonally elongate pits for the origin of powerful pterygoid musculature' (McKenna, 1962). These diagnostic features appear to relate to Eupetaurus feeding primarily on abrasive food, with the incisors reduced in functionality and chewing achieved mainly by anteroposterior trituration of the cheek teeth (McKenna, 1962). Limited available natural history information (mainly for E. cinereus) suggests that these squirrels might specialize on eating pine needles (Zahler & Khan, 2003). The upper cheek-tooth row is strongly convergent anteriorly in adults (although it remains more parallel in juveniles). The P 3 is peg-like and strong. The relative size of the upper cheek teeth can be given as P 4 > M 1 ≈ M 2 > M 3 . The P 4 (and DP 4 ) and the upper molars have similar structures: the endoloph is strongly folded to form the 'posterolingual diagonal flexus' (see below); the anteroloph, hypertrophic preparaconule crista, paraloph, metaloph and posteroloph are finger-like and connected, in turn, to the buccal side of the endoloph. The anterior valley is wide and divided by the preparaconule crista. When worn, the preparaconule crista gradually connects with the parastyle to isolate the lingual part of anterior valley into an anterofossette. The central valley is always open. The posterior valley is always closed into a posterofossette, except in the M 3 in early wear.
The relative size of the lower cheek teeth can be given as P 4 > M 1 ≈ M 2 ≥ M 3 . The structure of the lower molars is relatively simple, with a 'ring' at the front and back, connected by the ectolophid in the middle. The anterior ring is formed by the anterolophid between the protoconid and metaconid, and the protolophid between the protoconid and metastylid. The centre of the anterior ring is one (E. cinereus and E.tibetensis) or two (E. nivamons) anterior fossettids. The lingual edge of the anterior fossettid is not closed in early wear. The posterior ring is formed by the entolophid and posterolophid that run between hypoconid and entoconid. The centre of the posterior ring is one posterior fossettid (talonid basin remnant). The posterior fossettid of M 3 is not closed in early wear. The mesoconid develops into a buccal cusp as strong as the protoconid and hypoconid on the middle of the ectolophid, which divides the hypoflexid into two folds. The central flexid runs from the posterolingual side of the tooth to the base of the anterior fold of the hypoflexid. Generally, DP 4 is similar to the lower molars, but has two opened anterior fossettids and two extra small cusps:an isolated paraconid in front of the buccal anterior fossettid and an isolated ectostylid on the buccal side of the mesoconid (Fig. 8). Also, P 4 possess two open anterior fossettids and an isolated paraconid, but the paraconid will connect with the metaconid and protoconid as the teeth are worn. The mesoconid is very small on P 4 in Eupetaurus (Fig. 9F). The entire crown of P 4 is more diagonally oriented than DP 4 and lower molars.

Diagnosis: Eupetaurus cinereus differs from other
Eupetaurus species in its overall more grizzled-grey pelage, with more pronounced grey frosting to the dorsal hairs (more saturated brown tones in the pelage of other species). Interspecific distinctions in the skull mainly concern the breadth of the rostrum and the shape of the frontal bones: the rostrum of E. cinereus is similar to that of E. tibetensis and narrower than that of E. nivamons; the interorbital region of E. cinereus is broader than in the other two species, and its postorbital processes is larger; the temporal ridges of E. cinereus and E. tibetensis are convergent posteriorly rather than parallel in E. nivamons (Table 3). The cusps and ridges of cheek teeth of E. cinereus are markedly more robust than in the other two species. The developed protocone and hypocone make the upper cheek teeth heart shaped, similar to E. tibetensis, and differ from the round ones in E. nivamons. Its lower molars are subsquare, also more robust than the subrectangular ones in the other two species. Eupetaurus cinereus preserves only one short anterior fosettid (one long anterior fosettid in E. tibetensis; two short anterior fosettids in E. nivamons) (Figs 7, 8).
Descriptive notes: Size large: head-body length 420-510 mm, tail length 430-545 mm, hind foot length 83-93 mm and mass 2500 g (Zahler & Woods, 1997). The dorsal pelage is typically grizzled greyish, although a melanistic example was recorded by Chakraborty & Agrawal (1977) from Chitral, Pakistan. The pointed ears are hairy, with the outer fur black or brown and the fur of the inner ear greyish. The dorsal surface of the patagium is darker brown. The dorsal surfaces of the forefeet are black and those of the hindfeet are brown. The ventral  pelage is pale brownish grey, with the hairs slate coloured basally and dirty white terminally. The tail is long, cylindrical and exceedingly bushy, with similar coloration to the body, often with the portion of the black tail tip short. The skull and dentition are as described above for the genus Eupetaurus and diagnosis of the species.
Distribution: Until recently, E. cinereus was known only by specimens and sightings from a few localities between 2400 and 3600 m a.s.l. in the western Himalayan region, concentrated around Gilgit in northern Pakistan, where its range appears to fall within a small area where the Himalayan, Karakoram and Hindu Kush Mountain Ranges meet (Zahler, 1996;Zahler & Woods, 1997;Molur et al., 2005;Dinets, 2011;Qamar et al., 2012;Din et al., 2015).
More recently, Eupetaurus has been recorded by camera traps from the Upper Bhagirathi Basin, Uttarakhand, north-western India, at elevations of 2700 and 4800 m a.s.l. (Pal et al., 2018(Pal et al., , 2020. On the basis of the colour of the animals and their geographical location, we regard these records as representing E. cinereus. This indicates a wider geographical and elevational range of occurrence for this species along the western margins of the Himalayas. The increasing use of camera-trap survey techniques might demonstrate that the species occurs in areas intermediate between the widely separated known localities in the Gilgit region of Pakistan and the Bhagirathi Basin of India, perhaps even extending into western Nepal. Natural history: This species resides on cliffs and rock faces at high elevations, and individuals are not normally tree dwellers except when feeding (Roberts, 1997;Zahler & Woods, 1997). Within Pakistan, the habitat of E. cinereus is characterized by patchy forests, with herbs and shrubs of the genus Artemisia  (Zahler & Woods, 1997;Zahler & Karim, 1999). In India, alpine vegetation associated with the 2700 m record includes Bistorta affinis (D.Don) Greene, Gentiana L. spp., Potentilla L. spp., Primula L. spp., Rhodiola L. spp. and Rhododendron anthopogon D.Don. The nearest forest was located 4 km downhill of this sighting and was dominated by Abies pindrow (Royle ex D.Don) Royle, Betula utilis D.Don and Rhododendron campanulatum D.Don. The 4800 m record was lacking any woody or arborescent vegetation, which contrasts with other records that typically associate the species with scattered conifers (Pal et al., 2018).
In the original taxonomic description of the genus, Thomas (1888) speculated that it might feed on fungi, lichens, mosses and other rock-loving plants in its high-elevation environment. Roberts (1997) proposed that the buds and cones of the Morinda spruce (Picea smithiana) may be an important food source for E. cinereus, because buds occur in spring and cones in summer, which are shed in winter. Indeed, Zahler & Woods (1997) found that one captive specimen would eat the buds of Morinda spruce but refused all fruits, nuts and other food offered. More recently, faecal samples from four individuals were found to contain 92-100% pine needles (Zahler & Khan, 2003). Other observations have found this species to consume the buds of blue pine, juniper, spruce, oaks, Chilgoza pine and deodar [Cedrus deodara (Roxb. ex D.Don) G.Don]. In August and September, they are also thought to visit walnut trees (Juglans regia L.), although it is not clear what parts of the plants they eat (Mirza & Rasool, 2005 (Zahler, 1996(Zahler, , 1998Zahler & Dietemann, 1999;Qamar et al., 2012;Pal et al., 2020).
This species appears to be nocturnal, solitary and active all year round (Zahler, 1996). A nest attributed to this species was reported from a shelf in a cave at 3230 m a.s.l. in Jutal Nala, northern Pakistan, which was made of bark and branches of juniper and grasses and looked like a flat bowl measuring 500 mm × 400 mm (Oshida et al., 2005). A specimen has also been recorded in Sai Nullah (Gilgit, Pakistan) coming out of an old oak tree at ~2590 m a.s.l. Thus, there appears to be some flexibility in nesting locations if tree hollows are available (Mirza & Rasool, 2005). An immature specimen at BMNH collected on 17 April suggests that breeding occurs early in spring and that two litters in a year could be produced (Roberts, 1997). This is supported by observations of two or three young being produced per litter, with young observed in spring and in late summer (Zahler & Woods, 1997).
Conservation: Eupetaurus cinereus is threatened by habitat loss owing to large-scale clear cutting of forests, particularly the destruction of highelevation pine woodlands (Zahler & Woods, 1997;Zahler & Khan, 2003). It is also threatened to a lesser extent by expansion of agriculture, small-scale logging, infrastructure development and human settlements (Molur et al., 2005). The total population size was estimated to be between 1000 and 3000 individuals within the known range in Pakistan in the mid-1990s (Zahler & Woods, 1997). It is currently classified as Endangered by the International Union for Conservation of Nature (IUCN) (Zahler, 2010), because it was considered likely to have suffered a recent reduction in population numbers from habitat loss (Rao & Marwat, 2003;Hasan, 2008).
The recent discovery of the species in the state of Uttarakhand in north-western India (Pal et al., 2018(Pal et al., , 2020 demonstrates that it might have a considerably wider distribution than previously realized and indicates that protection and conservation in India is also fundamental for the long-term survival of E. cinereus. The species is included in Schedule II (Part II) of the Indian Wildlife (Protection) Act 1972 (Molur et al., 2005;Pal et al., 2018).
Hunting is not thought to be a major threat to this species. The species might receive some traditional respect and de facto protection because it is considered be a source of a material called salajit (also spelt salaajeet or shilajit), which is thought to be a mixture of flying squirrel urine and faecal pellets and rock leachate (Zahler, 1996;Zahler & Karim, 1998;Meena et al., 2010). Salajit is important in traditional medicine (Hooper, 1903) and is collected from caves and sold in medicinal potions used to treat various diseases (Zahler, 1996;Zahler & Karim, 1998;Meena et al., 2010). Historical references regarding the collection and use of salajit (e.g. Hooper, 1903)  Suggested common name: Tibetan woolly flying squirrel.
Z o o b a n k r e g i s t r a t i o n : u r n : l s i d : z o o b a n k . org:act:19DA79D4-7E20-4EB6-A1C7-2C22C6F21D58 Holotype: NML 19524, adult, skin and skull, collected in 'Tibet'. Sequences of Cytb for this specimen are deposited in GenBank (no. AY331673). This specimen was first discussed by Anderson (1879), who left it taxonomically unidentified. When describing E. cinereus as a new genus and species, Thomas (1888) speculated that Anderson's Tibetan specimen was also a Eupetaurus, a point later confirmed by Jentink (1890), who figured the skull of this specimen. Little is known about the collection details for this specimen other than that it is 'said to be from Tibet' (Jentink, 1890: 143). Corbet & Hill (1992: 314) noted that the association of the skull and skin for this specimen suggest that 'it is not a trade skin' (i.e. it is probably a wild-collected animal). To us, this specimen, when considered alongside the paratype skin obtained in a market in the Tibetan town of Gyantse (also spelt Gyangze, 江孜) in 1909, establishes the historical occurrence of this species in the south-central portion of the Tibetan Plateau (see below).
Type locality: Labels associated with the holotype indicate only that it was collected in 'Tibet'. The paratype specimen at BMNH was obtained from the bazaar at Gyantse (as 'Gyantsi Bazaar'). Here, we accordingly fix the type locality as 'vicinity of Gyantse Town', Tibet (= Xizang). Gyantse is situated at an elevation of 4000 m a.s.l. in the Nyang Chu Valley (年 楚河谷), Shigatse (日喀则), southern Tibet.
Diagnosis: Eupetaurus tibetensis differs from E. cinereus in having more saturated brown pelage, tinged with reddish tones (typically more clear grey in E. cinereus) and from E. nivamons in having a black tail tip that is much shorter (approximately half the length of the tail in E. nivamons). The rostrum is similar to E. cinereus and narrower than E. nivamons. The temporal ridges are convergent posteriorly, as in E. cinereus, rather than parallel, as in E. nivamons. The cusps and ridges of cheek teeth are more robust than in E. nivamons but less robust than in E. cinereus. Upper cheek teeth are heart shaped, as in E. cinereus, and differ from the rounded shape in E. nivamons. Lower molars are subrectangular, as in E. nivamons, and differ from the subsquare ones in E. cinereus, with only one long anterior fosettid per lower molar (two short anterior fosettids in E. nivamons; one short anterior fosettid in E. cinereus). Overall, the dentition of E. tibetensis resembles that of E. cinereus more than E. nivamons, but the external appearance of E. tibetensis is similar to E. nivamons.
Descriptive notes: In the available specimens of E. tibetensis, overall body size is similar to E. cinereus, with a head-body length of 468 mm in the holotype (with the imperfect, broken tail measuring 350 mm). The dorsal pelage is brownish grey, with less extensive frosting of the fur tips than seen in E. cinereus, and with reddish tones sometimes evident along the midline of the dorsum, on the rump and on the proximal portion of the tail. The underparts are greyish, washed with yellow. The dorsal surfaces of the forefeet and hindfeet are brown. The tail is bushy and cylindrical, with coloration similar to the body, except for a short black tip. The skull and dentition are as described above for the genus Eupetaurus and diagnosis of the species.
Etymology: The species name is derived from its occurrence in Tibet, where the holotype and paratype originated.
Distribution: Records of occurrence that we attribute to E. tibetensis originate from elevations of 2700-4000 m a.s.l. in Tibet (= Xizang) in China, Sikkim in India and (more tentatively) from Bhutan. This species is known for certain from two specimens from Tibet and two specimens from Sikkim, discussed above. Eupetaurus has also been reported from Bhutan (Wangchuk et al., 2004), and given the proximity of Bhutan to vouchered records from south-central Tibet and north Sikkim, we suggest that these Bhutanese observations are likely to represent E. tibetensis. We are not aware of vouchered specimens or specific published records from Bhutan, only general statements about its occurrence: Wangchuk et al. (2004) noted that it was found in coniferous, rhododendron and subalpine forests > 2700 m a.s.l.; Srinivasulu & Srinivasulu (2012) specified that it was known from western Bhutan; and Rinchen (2009) suggested that it might occur in Jigme Dorji National Park in north-western Bhutan. No records of Eupetaurus have yet been noted from Nepal (Mitchell, 1979;Thapa et al., 2016).
In summary, we suspect that the distribution of E. tibetensis comprises the far southern portion of the central Tibetan Plateau and the northern slopes of the Himalayas in Sikkim and Bhutan (Fig. 10).
Natural history: Currently, there is little known about the ecology of E. tibetensis. Commencing research on this species to understand its distribution, ecology and conservation status is an urgent need.

Conservation:
The conservation status of E. tibetensis remains unknown. We recommend an IUCN Red List categorization of Data Deficient for the species. Surveys are needed to determine the extent of its distribution, including whether it occurs in protected areas, such as Jigme Dorji National Park in Bhutan, and if it is facing particular threats, such as deforestation or hunting.  Referred specimens: KIZ 034189 (field number GS16078), immature female, skull, skin, and body preserved in ethanol, collected from Nanmowanshan Pass (贡山-巴坡人马驿道丫口), Mount Gaoligong, Gongshan, Yunnan. Two additional skins without skulls (KIZ 003299 and KIZ 003310), purchased at Lijiang or Gongshan, Yunnan in 1973 by Yingxiang Wang (Yang & Wang, 1989) (the original label locality was written as Lijiang, which was later altered to Gongshan).
Diagnosis: Eupetaurus nivamons differs from E. cinereus in its more saturated brown dorsal pelage (clearer grey dorsal pelage in E. cinereus) and closely resembles the external appearance of E. tibetensis, but its black tail tip is longer. The rostrum of E. nivamons is wider than in E. cinereus and E. tibetensis; its temporal ridges are parallel rather than posteriorly convergent as in E. cinereus and E. tibetensis (Table 3). Cheekteeth cusps and ridges are weakest among the taxa; the protocone and hypocone are not as developed as in the other two species, making the upper cheek teeth round in shape rather than heart shaped. Its lower molars are subrectangular, as in E. tibetensis, and less robust than the subsquare molars of E. cinereus. Eupetaurus nivamons has two short anterior fosettids in the lower molars (one long anterior fosettid in E. tibetensis; one short anterior fosettid in E. cinereus) (Figs 7, 8).
Descriptive notes: Eupetaurus nivamons is slightly smaller than E. cinereus (holotype head-body length 419 mm, tail length 440 mm, hind foot length 90 mm, ear length 44.5 mm and mass 1420 g). The dorsal pelage is grizzled greyish brown. The same colour extends to the forehead, turning pale grey along the cheeks to the throat. Ventral pelage is light ashy, with a median longitudinal line of coarse hairs. The margin of the patagium is black. The dorsal surfaces of the manus and pes are black, mixed with scattered straw-coloured hairs. The ears are pointed and hairy, externally black and internally white washed with khaki. The proximal half of the tail is similar in colour to the back but browner, and the distal half is black.
A pale marking round the lower half of the otherwise brown body has been observed in one camera-trapped animal at Mount Gaoligong, Yunnan Province, China (Wang Jianhua, pers. obs.). The skull and dentition are as described above for the genus Eupetaurus and diagnosis of the species.
Etymology: The specific name is composed of the Latin nivalis, snowy, and mons, mountain. It is a noun in apposition.
Distribution: According to specimens, camera-trap images/videos and sightings by local people, E. nivamons is currently known from the alpine zone (3400-4450 m) on Mount Gaoligong [the watershed of the Ayeyarwady (= Irrawaddy) River and the Nu (= Salween) River] and Biluo Snow Mountain (the watershed of the Nu River and the Mekong River) (Fig. 10). Given that the known distribution runs along the Chinese side of the border with Myanmar, the species is likely to occur in suitable habitats immediately across the border in north-eastern Myanmar (similar to the case of the recently described gibbon Hoolock tianxing Fan et al., 2017). It should also be sought in alpine areas of far south-eastern Tibet situated east of the Brahmaputra River (Fig. 10).
Natural history: The specimen KIZ 034189 was collected from a cliff at an elevation of 3700 m a.s.l. near the spine of Mount Gaoligong. The surrounding natural vegetation is a mosaic of swamp meadows and dense shrubs ~0.5 m high. The main plants are currants (Ribes L.), barberries (Berberis L.), willows (Salix L.), roses (Rosa L.), rhododendron (Rhododendron L.) and juniper (Juniperus squamata Buch.-Ham. ex D.Don). The mean annual temperature is 3.9 °C, ranging from a low of −12.5 °C in January to a high of 19.6 °C in August. Mean annual humidity is 94.4%, ranging from a low of 70% in November to a high of 100% from April to September, except August [data from a nearby combined temperature and humidity meter (ibutton-MAXIM DS1923); Supporting Information, Supplementary Data SD8]. All cameratrap photographs and videos of E. nivamons were taken from similar cliff habitats at 3700-4450 m a.s.l. on Mount Gaoligong (Fig. 11). Large, round faecal pellets and partly eaten plant material were scattered under juniper scrub (Juniperus squamata) near the high-elevation cliff of Mount Gaoligong. Given that no other large rodents were documented in this habitat, we interpret this as evidence that juniper is an important food plant for E. nivamons in the area.
Camera-trap observations (N = 31) show that E. nivamons is active throughout the year, with a peak in October; and it is active from 19.30 to 06.00 h, with a peak from 19.30 to 00.00 h. Even in winter, fresh faecal pellets can be found on top of the snow under juniper shrubs, indicating that E. nivamons does not hibernate or migrate from the area during the coldest months of the year (Ge Gao, pers. obs.).

Conservation:
We have documented this species at two principal localities, Mount Gaoligong and Biluo Snow Mountain. Mount Gaoligong is within a nationally protected nature reserve, whereas Biluo Snow Mountain has not been gazetted as a protected area. Owing to its high-elevation cliff habitat, far away from human settlements and covered by heavy snow for more than half of the year, E. nivamons is probably not heavily affected by human activities. The only known direct threat we documented is poaching for meat by herb collectors who visit the area from June to October. Given the apparent rarity of the animal and relatively small habitat area, we recommend an IUCN Red List classification of Near Threatened for this species. The loss of alpine habitats owing to ongoing climate change is a potential threat to the species in the near future.

DISCUSSION phylogenetic poSition oF EupEtaurus
The high-crowned cheek teeth of Eupetaurus have attracted biological interest since the genus was first described (Thomas, 1888). In rodents, the evolution of hypsodont teeth involves modifications that result in cheek-tooth cusps that cease to interlock, which can improve the mechanical strength and effectiveness of cheek teeth, but may also obscure homologous landmarks on premolars and molars, making morphological relationships difficult to decipher (McKenna, 1962). Hypsodonty has evolved repeatedly in various lineages across the Rodentia (Williams & Kay, 2001), but Eupetaurus is the only hypsodont member of the Pteromyini. The cheek teeth of Eupetaurus are thus superficially different from all other flying squirrels and also different from other hypsodont rodents. In particular, the lower cheek teeth have two obvious buccal flexa and only one prominent lingual flexus, the reverse situation to most hypsodont rodents (a single buccal flexus and two or more lingual ones) (McKenna, 1962).
The genus was removed from Sciuridae and placed in its own family, Eupetauridae, by Schaub (1953: 395) (also Grassé & Dekeyser, 1955;Schaub, 1958) partly because of its highly derived dentition, but also owing to fundamental misconceptions in interpreting published images of the teeth of Eupetaurus. However, in a major overview of the relationships and dental anatomy of Eupetaurus, McKenna (1962) demonstrated that Eupetaurus is clearly referable to the Sciuridae (along with other flying squirrels, recognized at that time within the subfamily Petauristinae). He suggested that Aeretes G. M. Allen, 1940, Eupetaurus and Petaurista Link, 1795 constitute a natural group (the 'Petaurista group') and considered the dentition of Eupetaurus to be most like Petaurista, probably originating from a cheek-tooth pattern similar to Petaurista xanthotis Milne-Edwards, 1872. In a later review, Mein (1970) augmented McKenna's (1962) scheme by also transferring the genera Belomys Thomas, 1908, Pteromys G. Cuvier, 1800, Pteromyscus Thomas, 1908and Trogopterus Heude, 1898 to the 'Petaurista group' based on perceived dental synapomorphies. Thorington et al. (2002) later moved an additional genus, Aeromys Robinson & Kloss, 1915, to the 'Petaurista group', drawing on a phylogenetic analysis of flying squirrels using 80 morphological characters.
More recent work has focused on reconstructing relationships among the various flying squirrel genera using genetic data. Yu (2002) Thomas, 1908, Hylopetes Thomas, 1908 or Trogopterus. Subsequently, researchers have used three gene fragments (12S, 16S and IRBP) to reconstruct the phylogenetic tree of most flying squirrel genera (Mercer & Roth, 2003;Lu et al., 2013;Casanovas-Vilar et al., 2018). These studies found that Eupetaurus falls into a major clade with generic membership similar to the Petaurista group as proposed by Thorington et al. (2002), and more specifically, with Eupetaurus recovered as sister to the genus Aeromys. More recently, Li et al. (2020) included the most recently described flying squirrel genus, Biswamoyopterus Saha, 1981(Sanamxay et al., 2013Li et al., 2019), not previously included in morphological and genetic comparisons, in a phylogenetic analysis drawing on the same markers (12S, 16S and IRBP).
That study recovered a sister relationship between Biswamoyopterus and Aeromys, which together were most closely related to Eupetaurus.
In the present study, our ML tree also recovered a sister relationship between Eupetaurus and Aeromys + Biswamoyopterus, but our BI tree showed a different topology, with Eupetaurus in a more isolated phylogenetic position, recovered as sister to a larger clade consisting of Aeromys, Belomys, Biswamoyopterus, Pteromys, Pteromyscus and Trogopterus (Fig.  4). Some intergeneric relationships are now clear. Five monophyletic clades can be identified clearly in the Pteromyina by the consensus of our ML and BI trees ( Fig. 4): (1) a Eupetaurus clade including Eupetaurus only; (2) a Pteromys clade including Pteromys only; (3) a Petaurista clade including Petaurista and Aeretes (genetic samples of Aeretes are unavailable, but the close relationship between Aeretes and Petaurista is a consensus of morphological studies; McKenna, 1962;Mein, 1970;Thorington et al., 2002); (4) an Aeromys clade including Aeromys and Biswamoyopterus; and (5) a Trogopterus clade including Trogopterus, Belomys, and Pteromyscus.
More comprehensive molecular phylogenetic comparisons are needed to resolve the deeper intergeneric splits between these lineages fully, including the exact phylogenetic placement of Eupetaurus. In the meantime, our morphological comparisons provide additional information that suggests a clearer placement of Eupetaurus. We note that: (1) the metaconule is similarly reduced in Eupetaurus and the Aeromys clade, but prominent in Pteromys, the Petaurista clade and the Trogopterus clade; (2) the interpremaxillary foramen (Hill, 1935) and greater palatine foramina are more expansive in Eupetaurus and the Aeromys clade than in other flying squirrels; and (3) the enamel faces of the incisors are pale yellow in Eupetaurus and Biswamoyopterus, but usually orange in other flying squirrels. Presuming that these might represent synapomorphic characters, we hypothesize that our ML tree is a better indication of the true evolutionary relationships in this group of genera: that is, Eupetaurus is the sister genus of the Aeromys clade (Aeromys + Biswamoyopterus).
new interpretationS oF Similarity oF cheek teeth McKenna (1962) proposed that the occlusal morphology of the cheek teeth of Eupetaurus was closest to, and derived from, a pattern most similar to Petaurista. Evidence for this was that a posterolingual diagonal flexus exists only in Eupetaurus and Petaurista/Aeretes, and this was interpreted as a homologous apomorphy (Fig. 9). We recognize that there are three different types of flexus/notch on the lingual side of the upper cheek teeth in Pteromyini: one in front of the protocone on the anteroloph (anterior flexus/notch); one between the protocone and hypocone on the endoloph (central flexus/notch); and one between the protocone and the metaconule-derived pseudohypocone on the posteroloph (posterior flexus/notch) (Fig. 9). Flying squirrel taxa can have no lingual flexus/notch or up to three types of lingual flexus/notch. Traces of the protocone and hypocone can be found on two sides of the 'posterolingual diagonal flexus' of Eupetaurus, and a possibly homologous central notch can be found in the Aeromys and Trogopterus clades, which are genetically close to Eupetaurus, demonstrating that the 'posterolingual diagonal flexus' of Eupetaurus is a central flexus. Petaurista and Aeretes do not develop the hypocone on the endoloph; their 'posterolingual diagonal flexus' is instead a posterior flexus, which lies between the protocone and the metaconule-derived pseudohypocone. McKenna (1962) also suggested that the lower cheek teeth of Eupetaurus evolved from a pattern most like Petaurista xanthotis, suggesting a possible phylogenetic relationship with Petaurista. We found that the paraconid and protoconid of Petaurista/Ae retes/Pteromys are not merged on the lower molars, revealing four distinct buccal cusps, whereas the paraconid and protoconid of the Eupetaurus/Aeromys clade/Trogopterus clade are merged on the lower molars, revealing three distinct buccal cusps (Fig. 9).

Biogeography and adaptation oF EupEtaurus
Eupetaurus species are endemic to alpine habitats across the Himalayan region, suggesting that their evolution is closely tied to the geological history and environment of the Himalayas. The geological history of the Himalayas and the surrounding area is complex, but some consensus has emerged from recent syntheses (Wang et al., 2014a;Deng & Ding, 2015;Favre et al., 2015;Renner, 2016). One is that a central 'proto-Qinghai-Tibetan Plateau' was uplifted as early as 40 Mya, with subsequent outward extensions by the early Miocene (Wang et al., 2014a). Then the mountains surrounding the Plateau to the south, west and north began to uplift by the late Miocene (Fang et al., 2005;Brookfield, 2008;Wang et al., 2012;Deng & Ding, 2015). The mountains on the south-eastern edge of the plateau uplifted more recently, mainly between the late Miocene and late Pliocene (Sun et al., 2011;Wang et al., 2014b).
The Himalayas are an evolutionary hotspot for flying squirrels, where many different genera and species co-exist regionally (Corbet & Hill, 1992;Jackson & Schouten, 2012;Li, 2019). Many taxa demonstrate patterns of elevational stratification, probably reflecting patterns of habitat specialization and local competition between related lineages (Smith & Xie, 2008). Eupetaurus occupies the highest elevational range of flying squirrels in the region, and its deep phylogenetic isolation and morphological specializations indicate that it has probably occupied this niche for many millions of years. Our study estimates that the evolutionary divergence between Eupetaurus and other flying squirrels occurred during the early to mid-Miocene (95% CI = 13.72-20.46 Mya), similar to the timing when at least the south-west Himalayas probably reached their current elevation (Garzione et al., 2000;Murphy et al., 2009;Saylor et al., 2009;Wang et al., 2012).
The Himalayas are not only the highest mountain range on Earth, but also possess the steepest river gorges, such as the Indus, Ganges, Brahmaputra, Salween, Mekong and other major river systems (Brookfield, 1998). These deep gorges cut the alpine habitats suitable for Eupetaurus into multiple 'sky islands', between which lowland tropical and subtropical habitats hinder the movement of alpine species, which can consequently result in the isolation of populations and allopatric evolution between 'islands' (He & Jiang, 2014;Ai et al., 2018;Wan et al., 2018;He et al., 2019). However, 10 Mya the drainages were completely unlike those of today, with all the major rivers of South Asia running along parallel structural belts (Brookfield, 1998). After the late Miocene, the uplift of the Himalayas began to be slower than the headward erosion, which extended back onto the plateau to capture major originally latitudinal drainages to form the great sigmoidal bends of the current river system (Brookfield, 1998). River capture first occurred in the Ganges system and later in the Indus and Brahmaputra river systems (Brookfield, 1998). The order of river capture is consistent with the pattern of Eupetaurus speciation documented in our study: E. cinereus on the west side of the Ganges first split with an eastern lineage (E. nivamons + E. tibetensis) in the mid-to late Miocene (95% CI = 6.89-13.83 Mya), after which E. nivamons on the east of the Brahmaputra split from E. tibetensis on the west side in the late Miocene or at the Pliocene-Pleistocene boundary (95% CI = 2.28-6.98 Mya). We imagine that a common ancestor of living Eupetaurus species might once have had a more continuous distribution across Himalayan alpine habitats, with the evolution of river systems since the mid-Miocene promoting allopatric speciation across the region.

reSearch outlook
For most of the 20 th century, Eupetaurus was the most enigmatic and least studied of all Asian squirrel genera. Only now is the biological diversity represented in the genus becoming understood, with our study demonstrating the presence of three distinctive species of Eupetaurus occurring across the Himalayan region. New fieldwork, especially using camera-trapping techniques, continues to expand the known range of the genus, with previously overlooked populations of E. cinereus and E. nivamons recently discovered in north-western India (Pal et al., 2018; and Yunnan (Gao et al., 2017), respectively. There are still large gaps between the known ranges of the three species, and targeted fieldwork is likely to document further populations and perhaps even additional species in the genus. For example, it is not clear whether Eupetaurus occurs in the eastern margin of the Tibetan Plateau in the Qinghai, Gansu and Sichuan Provinces of China. It is important to document the current range of occurrence for E. tibetensis, which has not been documented by a museum specimen or other firmly documented record for a half century, but the species is likely to occur in Sikkim, southern Tibet and western Bhutan. Considering that the evolution and speciation of Eupetaurus are closely related to the Himalayan orogeny, the discovery of additional geographical populations of Eupetaurus might further enhance our understanding of geological history of the Himalayas and forming of the major rivers in South Asia, in addition to the ecology and evolution of the largest flying squirrels of the world.