High diversity of small insectivorous mammals on Qinghai–Tibet Plateau and first description of karyotype for four endemics of China

Among seven species of the order Eulipotyphla (from southern Gansu and northern Sichuan Provinces, Central China) studied cytogenetically, karyotypes of one talpid species, Uropsilus aff. soricipes (2n = 36, NFa = 54), and three soricid species, Chodsigoa hypsibia (2n = 65, NFa = 66), Sorex cansulus (2n = 42, NFa = 64) and Sorex thibetanus (2n = 42, NFa = 60), are described cytogenetically for the first time. All four species are endemic to China with distribution ranges restricted to the Qinghai–Tibet Plateau and adjacent mountain ranges. The Ch. hypsibia karyotype consists of mostly acrocentric autosomes and one metacentric pair of autosomes; besides, a B chromosome was identified. No polymorphism was detected among karyotypes of other species, including shrews Sorex bedfordiae (2n = 26, NFa = 44), Anourosorex squamipes (2n = 48, NFa = 92) and Crocidura suaveolens (2n = 40, NFa = 44). The Chinese shrew mole U. aff. soricipes and three shrew species (S. bedfordiae, Ch. hypsibia and A. squamipes) represent autochthonous fauna of Central/Western China, whereas S. thibetanus, S. cansulus and C. suaveolens belong to phylogenetic groups occurring mostly to the north and west from China; therefore, they should be considered relatively recent colonisers. Thus, considering the relationships of the species within phylogenetic groups, our results on karyotypes are in good agreement with molecular genetic data.

Silver-positive dots were tentatively registered close to centromeres in one mid-size subtelocentric pair, chromosome 4 (Fig. 2c).
C-positive heterochromatic blocks were found to be located in centromeric regions of all autosomes except for two subtelocentric pairs, chromosomes 2 and 8, which have fully heterochromatic short p-arms. Visible C-blocks were revealed at a terminal position of q-arms in two autosome pairs, chromosomes 4 and 14. Less darkly stained C-bands were visible at interstitial positions of pair number 8. The X chromosome is C-negative, whereas the Y chromosome stains positively throughout the entire chromosome arm (Fig. 2d).
C-heterochromatic blocks were revealed in pericentromeric regions of 27 acrocentric pairs (chromosomes . Two large acrocentric pairs (chromosomes 1 and 2) and one smallest acrocentric pair (chromosome 30) as well as the small metacentric pair are C-negative. Slightly visible blocks of interstitial heterochromatin were noted on the B chromosome. Because one acrocentric chromosome has a large pericentromeric C-block, we identified it as an X chromosome (Fig. 5b). The Y chromosome was tentatively identified as an acrocentric www.nature.com/scientificreports/ chromosome and therefore is marked by an asterisk (as is the B chromosome). Silver nitrate staining uncovered centromeric localisation of NORs in three medium-to-small acrocentrics (Fig. 5c). Two specimens of the Chinese mole shrew A. squamipes Milne-Edwards, 1872 (G17-140 and G18-336), were collected at sites No. 1 and 3. The studied male and female karyotypes do not differ from those reported  www.nature.com/scientificreports/ previously 19 : 2n = 48, NFa = 92. The autosomal set is composed of 23 large-to-small meta-or submetacentric pairs gradually decreasing in size. In a male karyotype, the largest metacentric and small subtelocentric were identified as the X chromosome and Y chromosome, respectively (Fig. 6a).

Discussion
Among the seven studied species of the order Eulipotyphla from southern Gansu and northern Sichuan Provinces, China, four species were analysed cytogenetically for the first time (U. aff. soricipes, Ch. hypsibia, S. thibetanus and S. cansulus).
In Talpidae, the most primitive subfamily Uropsilinae includes a single genus (Uropsilus), which formerly consisted of Uropsilus andersoni, U. gracilis, U. soricipes, U. investigator and U. aequodonenia 20,21 ; later, U. nivatus and U. atronates were also recognised as valid species 9 . One more new species, U. dabieshanensis sp. nov., was recently described based on morphological and molecular genetic data 22 ; thus, now the genus is thought to comprise seven species. At least 15 monophyletic lineages of Uropsilus have been recognised using molecular markers, and it seems that the taxonomic diversity of this genus may still be underestimated 9,22 . Until now, karyotypes of only two species, U. nivatus and U. andersoni, have been reported. Note that during the description of the karyotype 23 , U. nivatus was regarded as a subspecies of U. gracilis 24 but was later elevated to full species rank 9 . The karyotype was described for an individual from an area inhabited only by U. nivatus; therefore, we use this name for the species, whose karyotype has been reported by Kawada et al. 23 . Despite the same 2n = 34, these karyotypes differ in chromosomal morphology as evidenced by a difference in NFa, which is 46 and 52 in U. nivatus and U. andersoni, respectively 13,23 . This variation is caused by a difference in the number of the subtelo-and acrocentric autosomes. Because only the conventional karyotype is reported for U. andersoni, it is difficult to compare the homologous arms and determine possible rearrangements between these two species. Moreover, the variation in the size of the short arms can be a consequence of some nuances of obtaining the chromosome preparations and then can lead to incorrect determination of NFa. Altogether, the authors 13 concluded that the karyotype of U. andersoni is similar to that of U. nivatus.
The U. aff. soricipes karyotype (2n = 36, NFa = 56), described here for the first time, differs from those of the two species mentioned above by smaller numbers of meta-and submetacentric pairs (two pairs instead of four in both U. nivatus and U. andersoni) and a larger number of subtelo-and acrocentric pairs (in total, 15 pairs instead of 12). The comparison of homologous chromosome arms between the U. nivatus and U. soricipes karyotypes indicates that all chromosome arms match between the two species, judging by the G-band pattern. Two small acrocentric pairs (chromosomes 15 and 16) in U. soricipes correspond to metacentric pair No. 2 in the U. nivatus karyotype. Accordingly, the lower 2n in U. nivatus, and apparently in U. andersoni, can be explained by a single Robertsonian fusion resulting in a new metacentric pair. Moreover, these species show a difference in the size of the acrocentric Y chromosome, which is larger in U. aff. soricipes than in the other two species. The silver nitrate staining performed on U. nivatus and U. soricipes chromosomes did not reveal any differences between   27 to contain a smaller number of chromosomal arms: 2n = 26, NF = 44. After re-checking of that material, it turned out that those two individuals are characterised by the same karyotype structure (2n = 26, NF = 46) as the three shrews described here.
Previously, karyotypes with 2n = 26, NF = 46 were detected in three S. bedfordiae individuals from Mt. Laojun (3900 m a.s.l.), Lijiang District in northern Yunnan province 26 . With the exception of one much larger subtelocentric pair (chromosome 10), those karyotypes were very similar to those described here. The lack of G-banded karyograms for the individuals from Yunnan makes it impossible to compare the homology of chromosomes between them in more detail. Additionally, these authors identified a 2n = 28 karyotype with an additional pair of   www.nature.com/scientificreports/ the 2n = 26, NF = 46 karyotypes described in our study are characterised by the presence of an additional small acrocentric pair (chromosome 12) and a different length of one subtelocentric pair (chromosome 11, Figs. 3, 7). A comparison of homologous chromosome arms between these G-banded karyotypes (2n = 24/25 and 2n = 26 from this study) indicated that they differ by several structural rearrangements (Fig. 7). Among the autosomes, a Robertsonian translocation (centric fusion) between telocentric pairs No. 11 and 12 in 2n = 26 karyotype resulted in the formation of a large subtelocentric pair No. 11* in 2n = 24/25 karyotype. Moreover, whole-arm translocations were detected in chromosome pairs Nos. 1, 2, 5, and 8 in 2n = 26 karyotype (Fig. 7). Nevertheless, all chromosome arms matched in the G-band pattern between the two S. bedfordiae karyotypes.
Molecular genetic studies point to the existence of eight highly divergent mitochondrial lineages among the shrews affiliated with the 'cylindricauda' group (five within S. bedfordiae, two in S. excelsus and one in S. cylindricauda) occurring on the eastern and southeastern edge of the QTP 7 . We did not aim to revise the intraspecific taxonomy of this group; however, it is worth noting that on the basis of the geographic origin and published genetic data 25 , all the karyotyped individuals presented here correspond to clade B, while those that have been reported by Moribe et al. 26 and Motokawa et al. 13 belong to other mitochondrial clades. Therefore, in this case, the molecular genetic differences match the karyotypic ones.
According to the recent phylogenetic analysis 25 , the Tibetan shrew belongs to the 'minutus s.l. ' group together with S. minutissimus, S. gracillimus and a species endemic to Honshu Island, S. hosonoi. Nonetheless, karyotypic characteristics of S. gracillimus and S. minutissimus are close to those of the 'caecutiens' group but not those of the 'minutus s.str. ' group 16,28 . Among these four species, S. gracillimus has a lower diploid number of chromosomes, 2n = 36, NFa = 60 29 , whereas both S. hosonoi 30 and the S. thibetanus described here are characterised by the same number, 2n = 42, but differ in NFa: at 60 and 66, respectively. The karyotypic status of S. minutissimus is still not clear; 2n = 38 and 42 have been reported for the species from two distinct areas, Finland 31 and Siberia 32 . Despite the absence of a 2n = 42 karyogram, in both cases, those authors reported 33 that the karyotypes consist of mostly bi-armed autosomes and single-armed sex chromosomes with the same NFa, 72. It is worth noting that the S. thibetanus karyotype contains 10 bi-armed autosomes, which is the smallest number of these among the other three species. Except for our results on S. thibetanus, there are no data on the localisation of NORs in these species.
On the basis of molecular data, the Tibetan shrew has been estimated to have diverged from other members of 'minutus s.l. ' (most likely from S. gracillimus) 2.0-2.5 million years ago (Mya) 25 . This age corresponds to the earliest possible time of QTP colonisation by the Tibetan shrew. The karyotypic data overall match the data from molecular genetic analysis indicating that S. thibetanus is affiliated with the eastern branch of the 'minutus s.l. ' group 25 .
The latest multilocus nuclear and mitochondrial phylogeny 25 of the nominal subgenus Sorex s.str. revealed that the 'caecutiens' species group is composed of five members: S. caecutiens, S. isodon, S. unguiculatus, S. shinto and S. cansulus. Among them, karyotypes of the first four species are known to be very similar (2n = 42, NFa = 64-66) 16,34,35 . The chromosomal set of the Gansu shrew S. cansulus has been unknown; here, for the first time, we demonstrated that this species has the same number of chromosomes, 2n = 42. A conventionally stained S. cansulus karyotype (NFa = 64) turned out to be quite similar in its morphological structure to species with NFa = 64 (S. isodon, S. unguiculatus and S. shinto). It has been documented that G-banding patterns are very conserved among S. caecutiens, S. shinto and S. unguiculatus karyotypes, which differ from each other by pericentric inversions 29 . The karyotype of S. caecutiens (NFa = 66) differs 35 from these species by an additional pericentric inversion in pair No. 8 and a centromeric shift in pair No. 12. There are no data on differentially stained karyotypes of the other species.
In S. cansulus, NORs were detected for the first time in satellite regions of two small telocentric pairs, No. 19 and 20. The same localisation of NORs has been found in S. unguiculatus 29 . As Zima et al. 16 have mentioned, 'all species of this group with 42 chromosomes possess pronounced satellites in two small autosomal pairs' . Evidently, these autosomes bear NORs in all species, but only S. cansulus and S. unguiculatus karyotypes have been stained.
According to molecular data, the Gansu shrew separated from its most probable sister group (S. caecutiens) 1.5-2.0 Mya 25 , which is somewhat later in comparison with S. thibetanus. Consequently, one may theorise that the Gansu shrew is a more recent coloniser than the Tibetan shrew. Our karyotypic data overall match the data from molecular genetic analysis suggesting that S. cansulus belongs to the eastern branch of the 'caecutiens s.l. ' group 25 .
Within the subfamily Soricinae, the tribe Anourosoricini is composed of four species 20 : Anourosorex assamensis Anderson, 1875, A. schmidi Petter, 1963, A. squamipes Milne-Edwards, 1872, and A. yamashinai Kuroda, 1935. Among them, karyotypic data have not been published for A. schmidi. A. yamashinai, endemic to Taiwan, has been studied using a set of differential stains and has 2n = 50, NFa = 96 36 , whereas only conventionally stained karyograms are available for A. squamipes (2n = 48, NFa = 92) collected in Fuleshan, Mianyang-shi, Sichuan Province, China 19 , and for A. assamensis from Tiddim Town of Chin State, western Myanmar (2n = 50, NFa = 96) 37 . Our results show that the structure of the A. squamipes routine karyotype does not differ from that reported previously: 2n = 48, NFa = 92. It has been demonstrated that the A. yamashinai karyotype 19 contains two pairs of the large subtelocentrics that are missing in A. squamipes. Thus, species of the genus Anourosorex share a similarity of the karyotype, even though they are geographically distant from each other, as is the case, for example, for A. yamashinai (Taiwan) and A. assamensis (south-western China, south-eastern Tibet, north-eastern India and west-central Myanmar).
The taxonomic status of shrews of the genus Chodsigoa has changed many times. Previously, eight species have been recognised in the genus 5,20 , then new species, Chodsigoa hoffmanni sp. nov., was described on the basis of sequencing data from one mitochondrial and two nuclear genes 8 . Now nine species are attributed to the genus.
De Winton's shrew Ch. hypsibia is endemic to China and occurs in central and southern provinces: southeastern Qinghai, southern Gansu, southern Shaanxi, eastern Tibet, Sichuan, northern Yunnan and southern Anhui 2 . So far, karyotypic data on the genus Chodsigoa have been limited to one chromosome set that is described for a species endemic to Taiwan, Chodsigoa sodalis (2n = 44, NF = 88) 38  www.nature.com/scientificreports/ this species consists of all bi-armed chromosomes; sex chromosomes have not been identified. The Ch. hypsibia karyotype stands out because all chromosomes are acrocentrics except one small metacentric pair of autosomes (chromosome 31) and an unpaired metacentric B chromosome (2n = 65, NFa = 66). On the basis of C-banding patterns, the X and Y sex chromosomes in the studied male were identified as a large acrocentric and medium acrocentric, respectively. The large differences between the Ch. hypsibia (almost completely acrocentric) and Ch. sodalis (almost completely metacentric) karyotypes may be explained by the more ancient origin of Ch. hypsibia. To some extent, this notion is confirmed by large molecular genetic differences between Ch. hypsibia and most of other species of this genus. In the genus Episoriculus, closely related to Chodsigoa, there are also large differences in the number of metacentric and acrocentric chromosomes between E. caudatus with 2n = 60, NF = 118 (M + SM = 19, ST = 9, A = 1) and E. soluensis with 2n = 74, NF = 126 (M + SM = 12, ST = 13, A = 11) 39 . Those authors consider E. soluensis a subspecies of E. sacratus soluensis.
The lesser white-toothed shrew C. suaveolens inhabits a huge area, from Europe to Asia. The species is characterised by a mostly stable karyotype with 2n = 40, NF = 50, but some authors have reported chromosome sets 2n = 41-42, NF = 52, 54, which result from the presence of supernumerary (B) chromosomes (see review 16 ). In a previous work, a 2n = 40, NF = 50 karyotype was identified in a female individual of C. suaveolens from the vicinity of the Goin Ba village in southern Gansu Province 27 . The specimen described here has 2n = 40 but a smaller number of chromosomal arms, NF = 48, owing to three bi-armed chromosome pairs instead of four. Unfortunately, highly condensed chromosomal arms in a low-quality chromosomal suspension did not allow us to clearly distinguish a possible fourth pair of bi-armed chromosomes. Thus, we cannot be completely sure that this specimen differs by a lower fundamental number of arms from those reported previously.
In a previous report, we showed that lesser white-toothed shrews from the QTP are close relatives of Cr. suaveolens s.str. (p-distance ~ 3%) 27 . Therefore, we can assume that this species is a relatively recent coloniser of the QTP and has not managed to achieve considerable genetic differentiation.

Conclusion: the QTP is a natural laboratory of speciation
The eastern edge of the QTP features high species diversity of small mammals, including insectivores (Eulipotyphla); this phenomenon can be attributed to high habitat richness of this relatively small area. This effect is caused by the alternation of mountain ranges with well-defined high-altitude vegetation belts as well as by low mountains with elements of subtropical vegetation. On the tops of some ridges directly connected to the QTP, there are plateaus with steppe and tundra vegetation, while the high-altitude parts of other remote ridges are isolated and form the so-called sky island 27,40,41 .
On the one hand, the high species diversity is a product of multiple faunal exchange events between the QTP and more northerly areas. In our case, Tibetan and Gansu shrews as well as the lesser white-toothed shrew belong to phyletic lineages that have their centres of origin in Northern Eurasia (Siberia and Amur Oblast). At present, geographic ranges of Tibetan and Gansu shrews are separated from those of their closest sister groups by a gap of 1000-1500 km. It must be pointed out that colonisation from the north can hardly be a one-time event; the exact dispersal routes remain to be elucidated.
On the other hand, autochthonous taxa showed high inter-and intraspecific diversity in our study (Chodsigoa, Uropsilus and the Sorex cylindricauda species group), consistently with long-term persistence of these species at the eastern edge of the QTP resulting in the presence of forms at different stages of speciation. In contrast to autochthonous species, no intraspecific variation was detected among the recent colonisers. This also applies to the Tibetan shrew, whose separation from the 'minutus s.l. ' group occurred more than 2 Mya. This observation suggests that autochthonous species have inhabited this area much longer.
The mountains at the eastern edge of the QTP can be called a natural laboratory of speciation, where both genetic and chromosomal alterations may promote the formation of new species. Further comprehensive comparative analyses of chromosomal and genetic variation within species groups of this region will enable an investigator to evaluate their contributions to barriers to gene flow. For instance, the genus Anourosorex features stable karyotypes and significantly lower genetic diversity than that in the other groups of species possessing similar geographic ranges, e.g. the 'cylindricauda' group. Nevertheless, this phenomenon may be explained alternatively as follows: representatives of the genus Anourosorex are more eurytopic, and the boundaries between habitats, including the 'sky of islands' , are less important for them than for other groups. In any case, we propose that the high species diversity of the QTP derives from the combined effects of genetic and ecological processes.

Materials and methods
Compliance with ethical standards. The study was carried out in compliance with the ARRIVE guidelines. All applicable international, national and/or institutional guidelines for the Care and Use of Animals were followed. The collection of animals and the trapping method complied with the guidelines of the A. Specimen collection and identification. Animals were caught alive at seven sites in Gansu and Sichuan Provinces of China during a joint Russian-Chinese scientific expedition conducted as part of scientific project 'RFBR-NSFC' in September-October 2017-2018 using home-made live-traps 42 . These traps work well for assessing the species diversity of small mammals 43 . Capture locations were determined via a GPS (Garmin) personal navigation system (Fig. 1). www.nature.com/scientificreports/ Regarding the Chinese shrew mole, on the basis of morphological characteristics and the geographic location, we tentatively assigned the found individual to U. soricipes 5 . Given that recent studies showed that the number of species in the genus Uropsilus may be much larger than is currently known 9,22 , we cannot be confident that this individual is a true U. soricipes. In this study, it is referred to as U. aff. soricipes. Further molecular genetic research is needed to clarify its true identity.

Cytological preparations.
Eighteen specimens belonging to seven species were karyotyped. Mitotic chromosome suspensions were made in the field from short-term culture of bone marrow and/or spleen after colchicine treatment in vivo as described previously 44 . For U. aff. soricipes, a chromosome suspension from primary fibroblast culture was prepared following the standard technique 45 . Air-dried chromosome spreads of all the specimens were conventionally stained with 2% Giemsa for 1-2 min and sequentially subjected to differential staining.
The standard trypsin/Giemsa staining procedure was carried out for the identification of each chromosome arm by G-bands 46 . C-banding was performed by the classic technique 47 . NORs were detected by silver nitrate staining 48 .
A Leica DFC-295 CCD camera mounted on a DM1000 (Leica) microscope or a Metasystems CCD (Zeiss) camera mounted on an AxioScope 2 (Zeiss) microscope was employed to capture images using MetaSystems Ikaros ver.5.3 and Leica Application ver.3.2 software packages, respectively.