In this study we used the original genetic data from 119 specimens of stripe-backed shrews (Supporting Information Table S1). Samples from China (tissue biopsies, muscles preserved in 96° ethanol and vouchers) were obtained from the collection of the Zoological Museum of Lomonosov Moscow State University.

The total genomic DNA was extracted using a standard protocol of proteinase K digestion, phenol-chloroform deproteinization and isopropanol precipitation (Sambrook et al. 1989). We amplified and sequenced mitochondrial cyt b gene in 116 specimens and segments of apolipoprotein B (ApoB), exon 11 of the breast cancer type 1 susceptibility protein (BRCA1), and melanocortin 1 receptor (MCIR) in 32 specimens. Primers and polymerase chain reaction protocols for ApoB and BRCA1 are described in Bannikova and Lebedev (2010); primers for MCIR were used from Rosenblum et al. (2004). General methods of cyt b extraction, amplification and sequencing follow Bannikova et al. (2010). PCR products were visualized on 1.5% agarose gel and then purified using the Diatom DNA CleanUp kit (Isogen). Approximately 10–30 ng of the purified PCR product was used for sequencing with each primer by the autosequencing system ABI 3100 Avant using ABI PRISM BigDyeTM Terminator v. 3.1 (Applied Biosystems, Foster City, CA, USA). The sequences obtained in this study can be accessed via GenBank (www.ncbi.nlm.nih.gov/Genbank) (Accession numbers PV201060–PV240092, Supporting Information Table S1).

Besides, sequences of eight nuclear genes obtained in the study by Chen et al. (2015, 2022) and sequences of cyt b obtained by Sheftel et al. (2018), Bannikova et al. (2018) and Chen et al. (2015, 2022) were downloaded from GenBank and used in the phylogenetic analysis (see Figs S1–S3 and table S3 in Chen et al. (2022).

All sequences were aligned by eye using Bioedit version 7.0.9.0 (Hall 1999). Heterozygous positions in nuclear gene sequences were coded using the IUB Ambiguity Codes. For the nuclear based tree reconstructions, we used the concatenation of eight nuclear genes sequenced by Chen et al. (2022). To improve resolution, the sample of Chen et al. (2022) was reduced to the specimens for which all eight nuclear loci were present. These alignments were expanded by including our original sequences (three nuclear loci for 32 shrews), which was sufficient for specimen assignment. The concatenation consisted of ApoB (509 bp), BRCA1 (794 bp), MC1R (681 bp), GHR (528 bp), RAG2 (594 bp), BDNF (482 bp), DBY7 (405 bp), UTY11 (562 bp).

The cyt b alignment consisted of 1140 bp and included 119 our own sequences and 475 downloaded from GB. Estimation of the cyt b genetic p-distances was conducted in MEGA11 (Tamura et al. 2021).

For the cyt b dataset and the concatenated alignment of nuclear genes, the phylogenetic trees were reconstructed under maximum likelihood (ML) and Bayesian inference (BI) criteria. In all analyses, the cyt b dataset was partitioned into three codon positions while the nuclear dataset was partitioned by gene.

The Bayesian Maximum Credibility tree was obtained using BEAST 1.10.4 (Suchard et al. 2018). The maximum likelihood (ML) analyses were performed in IQTREE v.1.6 (Nguyen et al. 2015). The ML trees were rooted via midpoint and clade support was assessed using ultrafast bootstrap (Hoang et al. 2018) with 10,000 replicates.

A karyotype of a stripe-backed shrew belonging to Sorex aff. cylindricauda sensu Bannikova et al. (2018) (female specimen S207220/G18-104 in Table S1) from vicinities of Langmusi (Gansu, Luqu County, 34.0789°N 102.634°E, 3,500 m a.s.l.) was examined. For cytogenetic analysis, a chromosome suspension from a short-term culture of bone marrow and spleen was used that is stored in the cryocollection of cell cultures at A. N. Severtsov Institute of Ecology and Evolution, the Russian Academy of Sciences (RAS; Moscow).

Air-dried chromosome spreads of a specimen were conventionally stained with 2% Giemsa for 1–2 min and sequentially subjected to differential staining. The standard trypsin/Giemsa staining procedure (GTG-banding) was carried out for the identification of each chromosome arm by G-bands (Seabright 1971). CBG-banding was performed by the standard technique (Sumner 1972) to determine C-heterochromatin blocks. NORs were detected by silver nitrate staining (Howell and Black 1980).

An ES-Experts BMR-1400HM-U CCD camera mounted on an Olympus BХ43F fluorescence light microscope was employed to capture images using Argus KARIO software package (ArgusSoft, St. Petersburg, Russia). All images were processed in Adobe Photoshop 2021.

For the morphological study, 70 voucher specimens in the collection of the Zoological Museum of Lomonosov Moscow State University (ZMMU) were compared with 44 historical specimens in the collection of the Natural History Museum, London (NHMUK) (Supporting Information Table S2). For the historical specimens, place names and their coordinates were determined from field notes in combination with information obtained from the United States Board on Geographic Names (USBGN), the GEOnet Names Server (GNS) (https://libraries.indiana.edu/databases/geonet) and Google Earth (https://earth.google.com).

Cranial and dental nomenclature, and measurement definitions follow that of Hoffmann (1987), Dannelid (1998), Maldonado et al. (2004) and Poroshin et al. (2010). Abbreviations used in the text for the dental nomenclature are incisor (I/i), unicuspid (Un), lower antemolar (a), premolar (P/p), molar (M/m) with premaxillary and maxillary teeth denoted by uppercase and mandibular teeth by lowercase letters.

External measurements of head and body length (HB), tail length (TL), hindfoot length (HF) and ear length (EAR) are recorded by collectors on specimen labels. The following cranial measurements (millimetres) were taken with digital calipers and using a microscope measuring stage.

Condyloincisive length (CIL): distance from anterior of first upper incisors (I1) to posterior of the occipital condyles.

Upper toothrow length (UTL): anterior of I1 to posterior of third upper molar (M3).

Length of incisor and unicuspids (IuniL): anterior of I1 to posterior of fifth unicuspid (Un5).

Length of unicuspid row (Until): anterior of first unicuspid (Un1) to posterior of Un5.

Length of molariform row (MolL): anterior of premolar (P4) to posterior of M3.

Palatal length (PIL): length in midline from I1 to posterior of palate.

Post palatal length (PstPaL): basioccipital length in midline from posterior of palate.

Incisor breadth (IncB): breadth across the labial margins of I1 to I1.

Rostral breadth (Un4B): breadth across the labial margins of Un4–Un4.

Maxillary breadth (MAXB): breadth across the labial margins of M2–M2.

Interorbital breadth (IoB): measured across the narrowest portion of the interorbital region.

Braincase breadth (BB): greatest breadth measured across the mastoids.

Braincase length (BL): length from the inferior articular facet to the posterior margin of the occipital condyle.

Braincase height (BH): depth of the braincase in midline.

Mandible length including i1 (MandLi1): length from the tip of the lower incisor (i1) to the posterodorsal facet of the condylar process.

Mandible length (MandL): length from the base of i1 to the posterodorsal facet of the condylar process.

Mandible toothrow length including i1 (MTLi1): length from the tip of i1 to the posterior face of the third lower molar (m3).

Mandible toothrow length (MTL): measured from the anterior face of the first lower antemolar (a1) to the posterior face of m3.

Mandible height (MH): the height of the coronoid process of the ascending ramus.

In the description of the results of the phylogenetic analysis, we mostly followed the clade designations by Chen et al. (2022) to facilitate the comparison among the patterns obtained. However, we accepted the following important modifications. The assignment of specimen to clades was based on their position in the nuclear but not the mitochondrial tree.

Nuclear trees

The ML and Bayesian trees as inferred from the concatenated sequences of eight nuclear genes (Figs 1, S1, S2) support the following monophyletic groups (clades): B (S. wardi), G+H (S. cylindricauda), N (S. excelsus), K (Sorex Tax.K), A, D, J, L, and I+M. The monophyly of group C (S. bedfordiae) is recovered in the Bayesian tree but violated in the ML tree due to the placement of a single specimen (C_CBBX71) which does not form a monophyletic group with other samples of group C. The clades G and H are well supported in the Bayesian tree, however in the ML tree only H is monophyletic. Group I (gomphus) is paraphyletic relative to the clade M. The samples that were assigned to the lineages E (nepalensis) and F do not form monophyletic groups being subdivided correspondingly into two and three well supported lineages.

Figure 1. 

ML phylogeny of the stripe-backed shrews complex as inferred from eight concatenated nuclear autosome and Y-chromosome genes. Numbers above the branches indicate Bayesian posterior probabilities (≥0.90) and ultrafast bootstrap support (≥80%). A tree with the uncollapsed clades is shown in Figure S1.

In the Bayesian tree the relationships among most of the clades including basal nodes are not resolved. However, both Bayesian and ML trees support the clade comprising monophyletic K and the E/F association. Besides, the ML tree supports the monophyly of a large clade including all groups with the exception of B, D, G+H and I+M, with the latter two clades being the most divergent.

The cyt b tree and comparison between nuclear and mitochondrial results

The main clades found in the nuclear tree demonstrate obvious correspondence to well supported divergent lineages in the cyt b tree (Figs 2, S3, S4); mitochondrial lineages are designated using a combination of letters A–N with the “m” superscript). The mitochondrial results strongly support the monophyly of lineages B^m, C^m, D^m, E^m, F^m, G^m, H^m, J^m, K^m, L^m, M^m, N^m. Lineage I^m is embedded within the lineage A^m thus making it paraphyletic. Ninety percent of inter-lineage distances fall within the range between 5.0 and 9.6% (mean 7.0%).

Figure 2. 

Bayesian Maximum Credibility tree (BEAST 1.10.4) based on the cyt b gene. Numbers above branches indicate Bayesian posterior probabilities (≥0.90) and ultrafast bootstrap support (≥80%). A tree with uncollapsed clades is shown in Figure S3. A dendrogram on the left depict phylogenetic relationships among genetic lineages as follows from the nuclear data analyses; dashed lines indicate lineages with discordant positions in the nuclear and mitochondrial trees.

As a modification to the delimitation scheme suggested by Chen et al. (2022) we recognize an additional mitochondrial lineage which is represented in our sample by three specimens from SE Qinghai (Q18-75, Q18-88, Q18-89). Previously, this lineage was regarded as a part of the clade A^m, from which it is, however, well divergent (p-distance = 5.1%). From the nuclear perspective the three above specimens belong to the clade K. The same is true for the only specimen (AEBT11) in Chen et al.’s sample that has a similar mitotype and for which nuclear data are available (see Fig. S1B in Chen et al. 2022). Thus, this mitochondrial lineage appears to have no matching unique nuclear clade. Therefore, we designate the new lineage as K^m+.

Compared to the nuclear tree, the mitochondrial topology is marginally better resolved. Several associations of the above-listed lineages are well supported. Among these are the sister relationships of lineages B^m and J^m, G^m and H^m, D^m and K^m, E^m and F^m (ML tree only for the latter pair). The E^m and F^m lineages are clearly distinct, however deep structure is shown within both. The clade E^m is split into several divergent sub-lineages that show correspondence to those in the nuclear tree. At the same time, the subdivision into three groups within the clade F demonstrates no evident correlation with the mitochondrial pattern. The distribution of group F is interesting in that it occupies a south-western Yunnan region divided by almost parallel ridges and rivers cutting through them. Unfortunately, the coordinates of the locality for each of the specimens studied by Chen et al. (2022) are not known, although they could clarify the subdivision of F into subgroups. The current data does not allow the resolution of the basal radiation of the S. cylindricauda species complex, however lineages M^m and L^m appear as the most divergent.

The conflict between nuclear and mitochondrial trees is twofold. First, several specimens that are robustly supported as members of a particular clade in the nuclear tree (Figs S1, S2) are falling into non-matching mitochondrial lineages (Figs S3, S4). For example, the specimen AXcsd25181 from the clade C in the nuclear tree appeared within the large subclade of clade A^m in the cyt b tree. The specimens AEDege11 and CCMGLL61 from the clade A in the nuclear tree clustered within the lineages N^m and C^m, respectively. In the cyt b tree, the clade D^m includes three specimens (G17-74, G17-75, G18-146) that belong to the clade K in the nuclear tree. Second, the positions of several clades in the nuclear and mitochondrial trees are strongly discordant (Fig. 2), (for example, the position of lineage I^m versus group I). An important point is the phylogenetic position of the lineage D^m. In the mitochondrial tree D^m stands as a close relative of K^m (p-distance of just 2.4%) in contrast in the nuclear tree the clade D is placed as one of the basal clades while the clade K forms a supported association with groups E and F.

The autosomal set of Sorex Tax.K (specimen G18-104/S207220) was composed of 20 biarmed, four single-­armed autosomes and an unpaired smallest biarmed chromosome, which apparently is an extra or B chromosome (Fig. 3A); 2n = 26 + 1B karyotype with a fundamental autosomal number (NFa) of 44. Among autosomes, there are one large metacentric pair (chromosome 1), six medium-to-small meta/submetacentric pairs (chromosomes 2–7), three large-to-medium subtelocentric pairs (chromosomes 8–10) and two medium-to-smallest acrocentric pairs (chromosomes 11–12). After the G-banding pattern was assessed (Fig. 3B), the mid-sized acrocentrics were designated as X chromosomes.

Figure 3. 

The female karyotype of the stripe-backed shrew from Langmusi (Gansu Province, Luqu County, China; G18-104/S207220): A conventional Giemsa staining, B G-banding, C C-banding and D silver-nitrate staining. 2n = 26 + 1B, NFa = 44. XX – sex chromosomes. В – B chromosome.

C-positive heterochromatic blocks were found to be located in centromeric regions of three autosome pairs (chromosomes 1–3), while five pairs 4–6 and 8–9 were completely C-negative. Visible C-blocks were also revealed at a terminal position of p- or q-arms in three pairs (chromosomes 7, 10–11); the smallest acrocentric pair (chromosome 12) as well as B chromosome were completely C-positive. The X chromosomes were C-negative (Fig. 3C). Silver nitrate staining revealed terminal localisation of NORs on short p-arms of two acrocentric pairs, chromosomes 11 and 12 (Fig. 3D).

Morphological data (Table 1 and Fig. 4) indicate that Sorex Tax.K is clearly larger than S. wardi and S. cansulus Thomas, 1912, which occur in geographical proximity to this taxon. Sorex Tax.K differs in skull proportions and is greater in skull length than S. cylindricauda and S. sinalis, however it should be noted that the sample sizes are small. The dental proportions of Un4, Un5 and P4 of S. sinalis differ from Sorex Tax.K and S. cylindricauda in both of which Un4 is broader than Un5 (Table 1). The morphological differences are subtle and may be partly obscured by geographical variation but suggest that Sorex Tax.K differs from S. cylindricauda. The skulls of Sorex Tax.K, S. cylindricauda, and S. sinalis are presented in Figure 5.

Figure 4. 

Comparison of anterior skull size in five Chinese species of Sorex. Maxillary breadth (MAXB) is plotted against upper tooth row length (UTL).

Figure 5. 

A Comparison of skulls from left to right of Sorex cylindricauda NHMUK 1911.9.8.14 from Sichuan, Yingjing, the holotype of S. sinalis NHMUK 1912.8.5.3 from Shaanxi, Feng Siang, and Sorex Tax.K ZMMU S-207218 (holotype of S. nivicola sp. nov.) from Gansu, Luqu County, Langmusi. Top row dorsal view, middle row ventral view, bottom row left lateral view of skull and mandible. B Upper dentition from left to right of S. cylindricauda NHMUK 1911.9.8.14, the holotype of S. sinalis NHMUK 1912.8.5.3 and Sorex Tax.K ZMMU S-207218 (holotype of S. nivicola sp. nov.). Occlusal view above, left lateral view below.

Our results are consistent with the existence of at least 14 clades in the stripe-backed shrew group which indicates that the group is a species complex comprising a number of cryptic taxa as suggested by Chen et al. (2022). However, the relationships among putative species are found to be significantly more complex than previously apparent.

First, the data demonstrate multiple instances of mitonuclear discordance, which is likely to be a result of mtDNA introgression. In such a case, specimen assignments to clades should be based on multiple independently evolving nuclear loci rather than on mtDNA. However, in Chen et al. (2022) the assignments were performed following the results of the nuclear + mitochondrial concatenated analysis. Careful comparison of mitochondrial and nuclear trees indicates that when mitochondrial and nuclear data are combined, mitochondrial genes impose their phylogenic signal, which can be attributed to much higher information content of mitochondrial data. As a result, in the concatenated analysis of Chen et al. (2022), specimens with discordant positions in the nuclear and mitochondrial trees were classified in accordance with their mitotype but not the nuclear gene affinities. This may bias our understanding of evolutionary relationships among lineages as well as their geographical distribution.

According to the molecular clock analyses based on nuclear data (Bannikova et al. 2018; Chen et al. 2022) the onset of group radiation is attributed to Early Pleistocene. The age of splits among most lineages is close to (or younger than) 1 Mya (species tree-based estimates in Bannikova et al. 2018).

Most of the lineages are distributed allopatrically or parapatrically, the most evident exception is S. cylindricauda (lineages G and H) which is sympatric with either A, B, C, D, E or N across the entirety of its range. The other case of relatively wide sympatry is observed between lineages B and K (SW Gansu, NW Sichuan, E Qinghai) (our data, Chen et al. 2022). In addition, according to the latter study lineages B and D are sympatric in some parts of N Sichuan. In neither of the above cases is there evidence of gene flow between sympatric taxa (as follows from the lack of mitonuclear discordance).

In contrast, our results show that reproductive barriers among many of the allo- or parapatric putative species are not effective enough to prevent hybridization and gene introgression. On one hand, cases of discordant mitochondrial versus nuclear assignments are likely explained by recent hybridizations between neighboring lineages C and D, C and A, and N and A (see Figs S1–S4). On the other hand, contradictory positions of certain clades may signify ancient reticulation events. For example, as discussed by Chen et al. (2022), the origin of the group I^m deep within the clade A^m in the mitochondrial tree is likely due to a single ancient event of mitochondrial introgression. Similarly, the position of the D^m clade as a very close sister to K^m can be explained by ancient introgression of its mtDNA from K. At the same time, the fact that three specimens of K have D^m mitotypes can indicate ongoing gene flow in the opposite direction (although no zone of sympatry has been revealed so far). This result by no means implies that D and K should be treated as conspecifics. The two species are pronouncedly different in their morphology: Sorex Tax.K is one of the largest shrews within the species complex with no clear dorsal stripe while D specimens are relatively small with a conspicuous stripe (Fig. 9C, see also fig. 2 in Chen et al. 2022). If our interpretation of the molecular trees is correct the clade D is a separate rather divergent species, which has lost its authentic mitochondrial lineage completely. A similar case is exemplified by relationships between brown (Ursus arctos Linnaeus, 1758) and polar bears (U. maritimus Phipps, 1774) (Hailer et al. 2012). The potential scheme of introgression events in the complex of stripe-backed shrews is presented in Figure 6.

Figure 6. 

Scheme of potential introgression events in the complex of stripe-backed shrews. Positions of the ellipses roughly correspond to geographic distribution of the genetic lineages; arrows represent introgressive events between them as follows from cases of cytonuclear discordance; dashed line indicates potential gene exchange between lineages I and M.

Ancient mitochondrial capture may also explain the existence of the K^m+ lineage, which is found in western populations of Sorex Tax.K; most likely this represents introgression from an unknown taxon. At the same time, one can find no evidence of hybridization between Sorex Tax.K and the geographically proximal S. excelsus (clade N). Other examples of apparently effective isolation between neighboring groups are illustrated by clades F and N, F and I (in both cases there can be a narrow zone of sympatry).

Thus, one can conclude that reticulation events are widespread within the species complex of the stripe-backed shrews, and gene exchange is common between them. However, it does not mean that all forms of the stripe-backed shrews should be lumped into a few species. A better interpretation is to consider the clades as nascent species, for some of which we find effective barriers (e.g., S. wardi = clade B), but for some they remain permeable. Plausibly, the situation can be interpreted within the framework of speciation with gene flow (Rice and Hostert 1993; Bolnick and Fitzpatrick 2007; Butlin et al. 2008; Arnold et al. 2015). The complexity of the group evolution is augmented by the specific nature of the area (Hengduan with its multiple geographic barriers) and possible gradient speciation driven by altitudinal adaptations.

Many questions on lineage interrelationships remain yet unanswered due to insufficient geographic sampling (primarily for groups L, M, J, I). In particular, close relationships between geographically distant I and M and the role of hybridization between I and M require additional study. Another issue is the relationships between and within F and E, which are likely sister groups for Sorex Tax.K. The nuclear tree neither supports nor strongly contradicts the monophyly of either lineage, which may be accounted for by insufficient numbers of studied loci. Hence, one may currently accept the mtDNA-based phylogeny as the correct one. At the same time, both clades appear deeply structured, so the status of the divergent subclades within E and F deserve separate consideration.

The karyotype of the female Sorex Tax.K studied here for the first time is characterized by a diploid number of 26 with an additional B chromosome (26+1B) and NFa = 44. Previous few cytogenetic studies have uncovered karyotypic variation within Sorex “bedfordiae”; 2n = 24–26 and 28 have been reported (Moribe et al. 2009; Motokawa et al. 2009; Pavlova et al. 2021) but the differences in 2n were mainly owing to В chromosome polymorphism (2n = 26+0–2Bs). Among eight cytogenetically examined individuals from Mt Emei Shan, Sichuan Province (Motokawa et al. 2009), karyotype variation (2n = 24+0–1B or 2n = 24–25 as presented by the authors) was associated with both the number of B chromosomes and the centric fusion of two acrocentrics, resulting in the formation of a large subtelocentric pair and, consequently, a decrease in 2n. Apparently, the increase of 2n in the karyotype of a male S. “bedfordiae” (2n = 26+0–2Bs) from Mt Laojun, Lijiang District, Yunnan Province, was also attributed to two additional small metacentric B chromosomes, but only the routine karyotype was represented in this study (Moribe et al. 2009). All other stripe-backed shrews karyotyped, which were described as S. “bedfordiae”, had a stable 2n = 26, NF = 46 (Moribe et al. 2009; Sheftel et al. 2018; Pavlova et al. 2021).

Detailed comparison of the new Sorex Tax.K karyotype with all previously published ones of S. “bedfordiae” showed that despite the same 2n and NF values, there are differences in the morphology of some bi-armed autosomes. Except В chromosome polymorphism, the structure of routine karyotype of Sorex Tax.K is closer to four S. “bedfordiae” (2n = 26+0–2Bs) karyotyped from Mt Laojun (Moribe et al. 2009); these specimens likely belonged to clade F by Chen et al. (2022). The lack of G-banded karyotypes from this locality did not allow us to compare the homology of chromosomes between them and Sorex Tax.K.

Motokawa et al. (2009) showed that karyotypes of S. cylindricauda 2n = 30 and S. bedfordiae, 2n = 24+1B, from Mt. Emei Shan differed by seven Robertsonian translocations and a couple of unidentified minor rearrangements. Furthermore, in our previous study we compared two S. “bedfordiae” karyotypes, a 2n = 26 karyotype from N Sichuan (clade B, here S. wardi, Songpan County, G18-221 in Table S1) and 2n = 24+1B from Emei Shan (clade С by Chen et al. 2022, presumably S. bedfordiae sensu stricto). It was demonstrated that although all chromosome arms were matched in G-bands pattern, the karyotypes differed by several whole-arm rearrangements (Pavlova et al. 2021). Comparing homologous chromosome arms between Sorex Tax.K and two available abovementioned G-banded karyotypes (2n = 24+1B and 2n = 26), we found that only three (Fig. 7A) and seven (Fig. 7B) autosome pairs, respectively, as well as XX chromosomes were completely matched in G-bands. Thus, despite having the same diploid chromosome number and NF, the two previously published and one new karyotype differ significantly in the combinations of chromosomal arms in biarmed autosomes.

Figure 7. 

G-band matching between Sorex Tax.K (S207220/G18-104 specimen from Langmusi, 2n = 26+1B) and A S. bedfordiae with 2n = 24 from Mt Emei Shan (Motokawa et al. 2009) and B S. wardi with 2n = 26 from Songpan County (Pavlova et al. 2021) karyotypes. Homologous chromosomes of Sorex Tax.K are numbered in accordance with Figure 3 and additionally marked double asterisks, while these of 2n = 24 and 2n = 26 specimens (taken with permission from Pavlova et al. 2021) are numbered and marked by an asterisk, respectively (in accordance with fig. 7 in Pavlova et al. 2021). Homologous regions are indicated by vertical and horizontal lines. XX: sex chromosomes, B: B chromosome.

In contrast to S. wardi (G18-221) (Pavlova et al. 2021), in the new Sorex Tax.K karyotype (G18-104) we visualized silver-stained nuclear organizers (AgNORs) on two acrocentric pairs (chromosomes 11 and 12). Motokawa et al. (2009) presented C-banded metaphase of an individual with 2n = 24+1B S. “bedfordiae” from Emei Shan to demonstrate C-positive В and Y chromosomes, but unfortunately, they did not report any C-heterochromatic blocks on autosomes. In the case of the female Sorex Tax.K, the B chromosome was also completely C-positive, but a lack of description in Motokawa’s work did not allow us to compare the distribution of constitutive heterochromatin on autosomes.

Taking into account all available chromosome data, it can be concluded that stripe-backed shrews demonstrate a high level of karyotypic variation owing to both supernumerary (B) chromosomes and whole-arm structural rearrangements (possible Robertsonian type, Rb). Robertsonian chromosomal polymorphism is well known, for instance, for the common shrew Sorex araneus, Eurasian species belonging to the “araneus” species group. This species is subdivided into more than 75 intraspecific chromosomal races (Searle et al. 2019); their karyotypes are characterized by a combination of a constant number of chromosomal arms in Rb metacentrics, resulting from Rb or whole-arm translocations WART (Bulatova et al. 2019). Apparently, similar mechanisms may operate in karyotype evolution of stripe-backed shrews; further cytogenetic investigation is needed.

Thus, the taxonomy of the stripe-backed shrew complex appears challenging due to remarkable variation in morphological traits, complex (likely reticulate) phylogenetic pattern and relatively recent times of diversification. But there is little doubt that S. cylindricauda, S. bedfordiae, S. wardi, S. excelsus, S. nepalensis, S. gomphus and Sorex Tax.K deserve species status as was previously suggested by Chen et al. (2022).

It should be noted that the available data on morphological variation within the species complex (e.g., Chen et al. 2014) are insufficient for a morphology-based revision and identification of diagnostic traits. We believe that in order to determine the species limits for such taxa as S. bedfordiae, S. gomphus and S. nepalensis and to provide adequate taxonomic descriptions for yet nameless clades (such as A, D, L, F) a comprehensive analysis integrating morphological, cytogenetic and multilocus molecular data is necessary.

In the current study, we revise the status of Sorex Tax.K, which was treated as S. sinalis by Chen et al. (2022). On the one hand, the molecular data supported the species status of Sorex Tax.K, which is pronouncedly divergent from other recognized species including similar sized and geographically proximal S. cylindricauda and S. excelsus. On the other hand, our morphological analysis of skulls and skins shows that the specimens of clade K attributed to S. sinalis by Chen et al. (2022) are not such according to morphological characteristics. The genotype of the holotype of S. sinalis is absent, however the skulls of shrews from clade K do appear to be distinct from S. sinalis and S. cylindricauda. Interestingly, one of the specimens included in Thomas’ type description of S. sinalis, the single specimen from Gansu, is very similar in size and morphology to our specimens from Gansu and Sichuan (Sorex Tax.K). The skin of this specimen has an indistinct dorsal stripe, similar to the skins of Sorex ­Tax.K and unlike the type series of S. sinalis from Shaanxi, so we consider that it may also represent Sorex Tax.K. Therefore, based on the combination of available morphological and genetic data we believe that Sorex ­Tax.K should be described as a new species.

Family Soricidae G. Fischer, 1814

Subfamily Soricinae G. Fischer, 1814

Genus Sorex Linnaeus, 1758

Sorex nivicola Bannikova, Jenkins, Lebedev, Pavlova, Sheftel, sp. nov.

https://zoobank.org/347B9872-AF9E-4CB8-90E5-B4021365D3D4

Holotype.

ZMMU S-207218 (field number G18-87), GenBank [PV240087, PV201067, PV201094, PV239965], collectors: Lebedev V., Zhu Y., juvenile male, skin and skull, weight 7.5 g, head and body length 71 mm, tail length 53 mm, hind foot length 13 mm.

Type locality.

CHINA, Gansu Province, Luqu County, Langmusi; 34.0789°N 102.634°E.

Paratypes.

S-207219 (G18-96) GenBank [PV240088, PV201068, PV201095, PV239966] • juvenile female, skin and skull; S-207220 (G18-104) GenBank [PV240089, PV201069, PV201096, PV239967] • adult female, skin and skull; S-207221 (G18-116) GenBank [PV240085, PV201070, PV201097, PV239968] • juvenile male, skin and skull; S-207222 (G18-117) GenBank [PV240090, PV201071, PV201098, PV239969] • juvenile female, skin; S-207223 (G18-119) GenBank [PV240091, PV201072, PV201099, PV239970] • juvenile female, skin and skull; S-207224 (G18-120) GenBank [PV240092, PV201073, PV201100, PV239971] • juvenile male, skin and skull. All from the same locality as the holotype.

Other material.

S-195208 (Chi11-96) GenBank [MH332013, PV240034, PV201086, PV239957] adult female, skin and skull, from S Gansu, Taizishan Nature Reserve, 35.2667°N, 103.4333°E; NHMUK 1912.8.5.11 female, skin and skull, Gansu, 17 miles [27.4 km] S.E. Tao-chou [Lintan], 800–900 ft [244–275 m], c. 34.6667°N 103.3833°E; S 207225 (G18-146) GenBank [PV240067, PV201075, PV201102, PV239972] juvenile male, skin; S-207226 (G18-174) GenBank [PV240086, PV201076, PV201103, PV239973] juvenile male, skin; S-199343 (G17-75) GenBank [PV239985, PV201083, PV201110, PV239974] juvenile male, skin and skull from Sichuan, 16 km E from Ruoergai (= Zoigê) 33.5847°N 103.1453°E.

Diagnosis.

Defined by a combination of characters. Large sized shrew, dorsal stripe present. Skull with narrow interorbital region relative to maxillary breadth. Upper unicuspids with Un2 larger than Un3, Un5 broad, similar in size to Un4.

Etymology.

The name is derived from the Latin nivis – snow, with the suffix -cola – dweller. The name is treated as a noun in apposition even though the generic name is masculine.

Description.

Skin (Fig. 8A): Dorsal pelage brown with a conspicuous to inconspicuous darker dorsal stripe (4–6 mm wide); dorsal pelage colour grading in the lateral region into the lighter brown of the ventral pelage. Dorsal pelage dark grey at the base, the upper portion of the hairs pale brown, with some darker brown at the tips; ventral pelage dark grey at the base, pale drab brown at the tips. Tail brown dorsally, clearly demarked from the pale brown ventral colour. Tail with pencil of hairs at the tip (4–5 mm). Dorsal surface of feet light in colour. Two pairs of inguinal mammae are present on the adult female S-207220. — Skull (Fig. 5A): Skull large, upper toothrow relatively long, narrow interorbital region relative to maxillary breadth. — Dentition (Fig. 5B): Un5 clearly visible in toothrow in labial view although slightly obscured by anterobuccal cingulum of P4; slightly less than 0.5 height of parastyle of P4. In occlusal view, Un5 similar in size to Un4 with a broad posterolingual cingulum; breadth equivalent to or slightly less than distance between parastyle and base of protocone of P4.

Figure 8. 

Dry skins of Sorex nivicola sp. nov. in dorsal and ventral views: A holotype S-207218 and paratypes: B S-207222, C S-207221, D S-207220.

Comparison with other species.

Sorex nivicola sp. nov. is larger in body size than S. cansulus and S. wardi. The head and body length of S. nivicola sp. nov. is slightly greater than in S. sinalis but comparable to S. cylindricauda. — Skin: The presence of a dorsal stripe in S. cylindricauda, S. wardi and S. nivicola sp. nov. (Figs 8, 9), serves to distinguish these three species from S. cansulus, S. excelsus and S. sinalis. — Skull: The skull of S. nivicola sp. nov. is similar in size to that of S. sinalis and S. cylindricauda (Fig. 5A) but notably larger than that of S. cansulus and S. wardi (Table 1). Sorex nivicola sp. nov. is distinguished from S. sinalis and S. cylindricauda by the longer toothrow, greater maxillary breadth and greater mandible length. The rostrum of S. nivicola sp. nov. is generally longer and broader than that of S. sinalis and on average longer than in S. cylindricauda. The ratio of interorbital breadth to maxillary breadth is on average lower and mandible height greater than in S. cylindricauda. — Dentition: The relatively large size of Un5 in S. nivicola sp. nov., S. cylindricauda and S. bedfordiae distinguishes these species from S. sinalis and S. cansulus in both of which Un5 is much smaller than Un4. The size of Un5 relative to Un4 differs in all three of the larger species (Fig. 5B). In labial view, Un5 of S. nivicola sp. nov. and S. cylindricauda is clearly visible in the toothrow, whereas in S. sinalis Un5 is visible but partially obscured by the parastyle of P4. In occlusal view, Un5 is noticeably smaller and narrower than Un4 in S. sinalis and narrower than the distance between the parastyle and protocone of P4; in S. nivicola sp. nov. Un5 is similar in size to Un4 and slightly less or equivalent in breadth to the distance between parastyle and protocone of P4; in S. cylindricauda Un5 is as large or larger than Un4 and nearly as broad or broader than the distance between the parastyle and the protocone of P4. Sorex nivicola sp. nov. differs from S. cylindricauda and S. wardi in the relative size of the second and third unicuspids; Un2 is larger than Un3 in S. nivicola sp. nov., whereas Un3 is larger than Un2 in S. cylindricauda and S. wardi.

Figure 9. 

Dry skins in dorsal and ventral views of Sorex wardi (A) S-199291 and (B) S-207194; Sorex aff. bedfordiae, clade D S-199290 (C), S. cylindricauda S-207217 (D).

Distribution.

Sorex nivicola sp. nov. is known from SW Gansu, NW Sichuan, E Qinghai. This species mostly lives in high mountain shrubland and grasslands at altitudes of 3400–4500 m (although it can also be found at elevations of just 2500 m). It primarily inhabits stony valleys of small streams covered with bushes.

Sorex sinalis – remains unexplored genetically. This species is known only from the type series collected in the high-altitude belt >3000 m asl in the Qinling mountains (72.4 km S.E. of Fengxiang, Shaanxi) (Thomas 1912). As of now, the only genetically analysed Sorex from the nearby mountains belong to S. wardi. Having examined the type material, Dolgov (1985) concluded that S. sinalis is conspecific with S. isodon, a species which belongs to different species group including also S. cansulus, S. unguiculatus and S. caecutiens (Bannikova et al. 2018). The same view was previously maintained by Corbet (1978). However, Hoffmann (1987) disagreed with Dolgov and rejected conspecificity of isodon and sinalis. We believe that the status and phylogenetic position can be established either by the analysis of the type specimens or de novo sampling in the high-altitude zone of Qinling.