Mitochondrial and nuclear gene-trees discordance
Our gene-trees inferred some conflicting or poorly supported phylogenetic relationships within Chinese Crocidura, and the molecular genetic interrelationships among several species are not consistent in mtDNA and nDNA trees (Fig. 2ab). This discordance could be attributed to explosive speciation [34], incomplete lineage sorting [35] and introgression [36]. Explosive speciation (i.e., rapid radiation) offers an explanation for the discrepancy owing to the very short branches that have poor support (Fig. 2). Our time-calibrated Bayesian analysis indicates rapid cladogenesis from the late Pliocene to the Early Pleistocene (Fig. 3), when most diversification occurred. Such rapid speciation can preclude high resolution of the phylogenetic trees because informative DNA substitutions will not have time to accumulate (e.g., [37,38]). A second explanation is that relatively recent radiation may not have provided enough time for a complete lineage sorting, further blurring the phylogenetic relationships of these species [39]. For example, there is a large genetic distance between C. sp. 2 and C. vorax and do not cluster together in mtDNA trees (Table 3), while relatively strong support for the clade C. sp. 2 + C. vorax in nDNA tree. However, distinguishing the two hypotheses will require additional nuclear loci [35,40]. Another explanation for the discrepancy is an old introgression event. For example, there is strong support for the clade C. lasuria + C. kurodai in mtDNA analysis and a small genetic distance between these two species (Table 3). However, these two species are very different and do not cluster together in nDNA trees.
Phylogeny and rapid radiation among China’s Crocidura species
Our fine-scale sampling of Crocidura from China reveals two well-supported major groups of Crocidura (Fig. 2). Some species in the group 1 seemingly represents north-western Crocidura species, while Some species in the group 2 represents south-eastern Crocidura species in China (Fig. 2). However, the phylogenetic relationships among many species in the group 2 remain unresolved (Fig. 2). To resolve phylogenetic relationships in recent, rapid radiation group, such as Crocidura, is a challenge task. Previous phylogenetic studies of Philippine Crocidura using multilocus data also yielded many poorly supported branches [41,14]. Fortunately, phylogenomic analyses often have the potential to resolve difficult phylogenetic relationships [42, 43]. However, using hundreds of ultraconserved elements (UCEs) and whole mitogenomes, Giarla and Esselstyn [44] could not fully resolve the phylogenetic relationships among the Philippine Crocidura based on coalescent-based approaches. Restriction site associated DNA sequencing (RADseq) is a commonly used approach for phylogeny estimation, and has been proved to be effective in solving some difficult phylogenetic relationships [43,45,46]. However, this RADseq could also lead to bias phylogenetic relationships when locus filtering techniques can not accurately identify homologous loci [44,47]. Thus, to completely resolve phylogenetic relationships using rapid high-throughput sequencing technology for this recent, rapid radiation Crocidura group may be still go a long way.
According to our divergence time estimation, China Crocidura species have rapidly diversified from the late Pliocene (3.66 Ma) to the early Pleistocene (2.29 Ma), followed by a series of diversifications through the Pleistocene (Fig. 3). What factors are responsible for the rapid diversification of Chinese Crocidura distributed? We propose that climate change and rapid uplifting of the Qinghai-Tibet Plateau may have driven the diversification of Chinese Crocidura. Prevailing trends towards cooling and desiccation at the Pliocene/Pleistocene boundary [48] have led to the diversification of some taxa (e.g., [24, 26, 38]). Shrews are extremely sensitive to changes in ambient temperature and humidity, which are key factors affecting the distribution and number of shrews [49]. An arid environment will most likely serve as a barrier to their dispersal, and subsequent allopatric diversification of Chinese Crocidura. At the same time, global cooling induced habitat turnover and consequent further accelerated the diversification, and eventually lead to the formation of Crocidura species in the region [48, 50]. In addition, some geological studies also supported that rapid uplifting of the Qinghai-Tibet Plateau at 3.6 Ma, 1.8 Ma and 1.1 Ma have resulted in the diversifications of Chinese Crocidura distributed on the Tibetan plateau [51, 52].
Taxonomic implications and diversity of Crocidura in China
Our data and analyses confirm the taxonomic status of the six well-recognized species: C. shantungensis, C. dracula, C. suaveolens, C. vorax, C. lasiura, and C. tanakae (Fig. 2; Additional file 2: Table S2). The mitochondrial and nuclear gene trees resolved each species as a well-demarcated clade, with substantial genetic differences between them (Table 3). Species delimitation analyses also identified each of them as a full species (Additional file 2: Table S2).
Jameson and Jones [53] originally described C. horsfieldii kurodai. Subsequently, Jiang and Hoffmann [2] revised the genus Crocidura in southern China and placed C. tadae and C. kurodai from Taiwan as junior synonyms of C. rapax. Hutterer [1] recognized three subspecies, C. rapax tadae, C. r. kurodai, and C. r. lutaoensis following Fang and Lee [19]. This view was followed by Hoffmann and Lunde [8], and Jenkins et al. [54]. Our analyses suggested the two subspecies, C. rapax rapax and C. rapax kurodai, should be elevated to full species status based on a high level of divergence (Table 3). GMYC and BPP analyses results (Fig. 3 and Additional file 2: Table S2).
Interestingly, when the C. attenuata samples from Vietnam were addedto the analysis, the phylogenetic tree showed that C. r. rapax was embedded within C. attenuata, making the latter a paraphyletic group (Fig. 3). Although the genetic distance value between these two species is only 0.049, this distance value is greater than the genetic distance value (= 0.039) between C. kurodai and C. lasiura (Table 3). GMYC and BPP analysis strongly supported that they were two different species. We also examined a number of specimens including near type site specimens (Baoxing of Sichuan), and found that they were completely different in body size, skull size and altitude distribution (unpublished data). In addition, a C. attenuata population from Vietnam is supported as a separate species in GMYC. Therefore, a further research is needed to clarify the taxonomic status of the C. attenuata complex.
Crocidura wuchihensis was originally described in Hainan by Shou et al. [55]. Subsequently, Lunde et al. [56] identified one specimen from Vietnam as C. wuchihensis. Some specimens of Crocidura from Guangxi and Vietnam also have been referred to as C. wuchihensis [41]. Jenkins et al. [54] stated that C. wuchihensis was widely distributed across Vietnam, including the provinces Lao Cai, Ha Giang and Lang Son in the north and Ha Thinh and Quang Nam in Central Vietnam. Bannikova et al. [3] considered the distribution of C. wuchihensis to be restricted to areas east and north of the Red River. However, our analyses contradicted these assertions. Our analyses suggested that the population of C. wuchihensis_HN from Hainan and those of C. wuchihensis_GX from Guangxi and Vietnam are valid species, which implies that the true C. wuchihensis may only occur on Hainan Island. Our results call for the reappraisal of the taxonomic status of shrews previously referred to as C. wuchihensis from Guangxi and Vietnam.
Crocidura indochinesis were previously considered a subspecies of C. horsfieldii [57]. Subsequent, Lunde et al. [58] considered C. indochinesis from Ke Go Nature Reserve, Vietnam as a full species. Our GMYC analysis strongly supported it as a species. It is still unknown whether C. indochinesis is distributed in China, because we do not know if the population in southern China is the conspecific with those in Vietnam.
Crocidura suaveolens is not known in China [2]. However, Dubey et al. [59] and Bannikova et al. [60] considered C. sibirica distributed in Xinjiang province (Northwest China) as Crocidura suaveolens, based on multilocus phylogenetic data. Subsequently, specimens of C. aff. suaveolens have been reported from southern Gansu(Northwest China) [61]. These specimens formed the sister-group of C. suaveolens in our phylogenic trees. Our GMYC analysis supported the population (C. aff. suaveolens) as a potential species (Fig. 3). In addition, Jiang and Hoffmann [2] listed C. gmelini from Central Asia, including Xinjiang in China. However, Ohdachi et al. [22] stated “It is possible that C. gmelini might be a synonym of C. sibirica (= C. suaveolens)”. All our Crocidura specimens from seven sample sites of Xinjiang in China appear to be C. suaveolens (Fig. 1). Therefore, it is necessary to sample C. gmelini to clarify the phylogenetic relationships and taxonomy between it and C. suaveolens. The occurrence of C. gmelini in China remains uncertain.
Crocidura dracula Thomas, 1912 [62] was first described from Mengzi County of Yunnan Province, China. Subsequently, the taxonomic decision was followed by Allen [4] and Ellerman and Morrison-Scott [63]. However, Jenkins [64] made C. dracula a subspecies of C. fuliginosa. Others have followed the arrangement [1, 2]. However, chromosomal and recently genetics studies suggested that C. dracula and C. fuliginosa are two different species [3,7,11,29, 65]. Burgin and He [7] considered these two forms as full species. Our analyses also supported three clades of C. fuliginosa groups that represented three different species: C. dracula from Northern Vietnam and southern China, C. fuliginosa from southern Vietnam, Cambodia and Malaysia, and undescribed species (C. sp. 3) from Motuo of Xizang (West China). It is worth noting that the conspecificity of C. fuliginosa from Southern Vietnam, Cambodia and Malaysia with the type locality in Burma has still not been tested. In addition, a population was also reported as C. dracula grisescens in Zhejiang [66]. Jiang and Hoffmann [2] supposed that the population represents probably new taxon.
Our analyses also resolved another undescribed species, Crocidura sp. 1, despite only one specimen being available from Zada Coutry in Tibet. This species is closely related to the Zarudny’s rock shrew (Crocidura zarudnyi) from Iran based on the cytb gene. It is the sister-group of C. shantungensis and C. suaveolens based on nuclear genes. The condyloincisive length of the undescribed white-toothed shrew (19.55 mm; our unpublished data) is close to C. attenuata in China. Despite considerable effort, only one specimen was collected. Thus, our conclusions are therefore tentative. Additional fieldwork is needed to acquire new specimens and allow a comprehensive taxonomic and population genetic analysis. The continuous discovery of new species such as Bufo zamdaensis [67] and Laudakia papenfussi [68] shows a strong need for further exploration in the region.
Another undescribed species C. sp. 2 from Hongjiang County of Hunan and Dongyang County of Zhejiang was also strongly supported. It was not identified as any known species based on our morphological data. Genetically, it has a high level of divergence from all other members of the group (10.2%–14.7%) (Table 3) and appears as a monophyletic group in the phylogenetic trees (Fig. 2; Fig. 3). Thus, the C. sp. 2 is likely another new species. Molecular analyses offer important insights, but extensive sampling, comprehensive morphological and morphometric comparisons are necessary to reach a final conclusion.