Bones and genes: resolution problems in three Vietnamese species of Crocidura (Mammalia, Soricomorpha, Soricidae) and the description of an additional new species

Abstract Recent investigations of Southeast Asian white toothed shrews belonging to the genus Crocidura have revealed discrepancies between the results of morphological and molecular studies. The following study concerns three species of Crocidura occurring in Vietnam, namely Crocidura attenuata, Crocidura tanakae and Crocidura wuchihensis, and an undescribed fourth species revealed by molecular analysis. For many years Crocidura attenuata has been known to occur in Vietnam but, until very recently, the morphologically similar and comparably sized Crocidura tanakae was believed to be restricted to Taiwan. Following several molecular studies over the last few years, this species is now believed to be considerably more widespread and recognised as occuring also in Vietnam. The results of one of these recent molecular studies also revealed the presence of an undescribed species of Crocidura, similar in size and morphology to Crocidura wuchihensis, which is herein described. Data are provided on geographical variation in Vietnam and the problems of defining morphologically similar yet molecularly disparate species are discussed.


Crocidura tanakae
from Taiwan was originally described as a full species but was subsequently considered to be either a synonym or subspecies of C. attenuata (Ellerman and Morrison-Scott 1951, Jameson and Jones 1977, Fang et al. 1997, Jiang and Hoffmann 2001, Han et al. 2002. Motokawa et al. (2001) demonstrated that the karyotype of Taiwanese specimens differed from that of C. attenuata from mainland southern China and suggested that it might represent a distinct species. Although Ohdachi et al. (2006) observed phylogenetic differentiation between the two samples they used from Taiwan and Vietnam, these authors continued to consider the Taiwanese samples as a subspecies of C. attenuata.  and  studying Southeast Asian shrews, recognised the relationship between samples of C. tanakae from Taiwan and a sample from northeastern Vietnam, which they identified in these studies as Crocidura cf. tanakae. The following year, Esselstyn and Oliveros (2010) demonstrated the presence of two similar sized species of Crocidura in Vietnam, namely C. attenuata based on samples from northern Vietnam and C. cf. tanakae based on samples from four separate localities in northern and central Vietnam. Bannikova et al. (2011) included their own recently collected samples from northern, central and southern Vietnam plus information derived from GenBank. They were similarly able to demonstrate the presence of two separate species, C. attenuata confined to a single locality in northeastern Vietnam and C. tanakae which was widespread in northern, central and southern localities. With a minimum distance of 9.91% between the haplotypes, their cytb tree showed good support for the distinction of C. tanakae from a multi-species group comprising C. attenuata, C. dsinezumi, C. indochinensis, C. lasiura, C. tadae kuroda, C. wuchihensis, C. sp. AB1 and C. zaitsevi.

Crocidura wuchihensis, C. indochinensis and C. sp. AB1
Although known from few specimens at any one location Crocidura indochinensis Robinson & Kloss, 1922 was considered to have a wide, disjunct distribution, occuring in a few widely separated locations in Vietnam and extralimitally in Myanmar and China (Osgood 1932, Anthony 1941, Heaney and Timm 1983, Jiang and Hoffmann 2001, Hutterer 2005, Jenkins et al. 2009). Additional specimens from southern Vietnam were recorded recently (Abramov et al. 2009;Jenkins et al. 2010) and tissue samples from these specimens were included in the molecular analysis by Bannikova et al. (2011).
Crocidura wuchihensis Shaw, Wang, Lu and Chang, 1966 was originally described on the basis of two specimens from Hainan Island, China. Specimens collected from two localities in Vietnam (Lunde et al. 2003;2004) were referred to this species and specimens from several other locations in Vietnam were also considered to represent C. wuchihensis (Jenkins et al. 2009). Samples of C. wuchihensis from Ha Giang Province, Mt. Tay Con Linh II, northern Vietnam were included in molecular analyses of Ohdachi et al. (2006) and Esselstyn and Oliveros (2010). A separate analysis of a sample from Vinh Phu Province, Tam Dao, northern Vietnam (Meegaskumbura et al. 2007) from a specimen misidentified as C. fuliginosa, also proved to be a representative of C. wuchihensis (see Bannikova et al. 2011). Bannikova et al. (2011) added GenBank data from samples from these two localities to their cytb analysis of samples from northern, central and southern Vietnam. Their analysis of cytb revealed the presence of three taxa forming a group with 100% support: C. wuchihensis in the two localities (Ha Giang Province and Vinh Phu Province) in northern Vietnam; C. indochinensis in a single locality (Bi Doup -Nui Ba Nature Reserve) in southern Vietnam; and an unnamed species, designated as C. sp. AB1, from Sa Pa in northern Vietnam. The average distance on the cytb tree between C. wuchihensis and the combined C. indochinensis / C. sp. AB1 branch was 7.75%. The p-distance from the cytb tree separating C. wuchihensis and C. indochinensis is 7.6%, and separating C. wuchihensis and C. sp. AB1 was 8.0%. There was 100% bootstrap support on the cytb and COI trees for the C. indochinensis and C. sp. AB1 group but these two taxa were respectively separated at p-distances of 4.1% (for cytb) and 4.0% (in the COI analysis).

Methods
This morphological study draws on specimens from a wide range of geographical locations in Vietnam (see Fig. 1) and includes those specimens from which tissue samples were analysed in the papers by Ohdachi et al. (2006) Measurements in millimetres were taken with digital callipers. Cranial and dental nomenclature follows that of Meester (1963), Mills (1966), Butler and Greenwood (1979) and Dannelid (1998). Definitions of skull measurements follow Jenkins et al. (2009). Diagnosis. A small shrew distinguished by the mitochondrial genes cytochrome b (cytb) and cytochrome oxidase c subunit I (COI) and by the shape of the talonid of the third lower molar (m3).

Crocidura sapaensis
Description. Size small (see Table 1) with a moderately long tail relative to head and body length (62-84%). Dorsal pelage dark greyish brown; tail dark grey dorsally, slightly paler below (see Fig. 2). Skull with a rounded, short rostrum: moderately broad interorbital region; rounded, relatively deep braincase with subangular superior articular facets and lambdoid crests just evident laterally near the junction with the mastoid (see Fig. 3). The first upper incisor is slender with a relatively small posterior cusp, less than half the height of the first upper unicuspid; posterolingual border of upper premolar (P4) deep and rounded, in close contact with the anterolingual margin of M1 in occlusal view; last upper molar (M3) relatively narrow. Lower incisor with two distinct cusps on the occlusal surface in unworn dentition; posterolingual cuspid present on lower premolar (p4); talonid basin of m3 broad and deep with an entoconid ridge and low entoconid (see Fig. 4).
Comparison with other species. Crocidura sapaensis averages larger than the very small species of Crocidura recorded from Vietnam. The condyloincisive length is greater than that of C. guy, C. annamitensis and C. kegoensis, within the upper part of the range of C. zaitsevi and the braincase is deeper than that of all four small species. Crocidura wuchihensis and C. sapaensis are in the same size range. Crocidura sapaensis is smaller than or at the lower end of the size range of C. indochinensis with a relatively shorter tail (see Table 1 and Fig. 5).
Crocidura sapaensis and C. wuchihensis are distinguished by differences in cytb sequences. Crocidura sapaensis differs from C. zaitsevi and C. indochinensis in the cytb and COI gene sequences.
Differences in the shape of the talonid of m3 in northern Vietnamese populations serve to distinguish C. sapaensis and C. wuchihensis (see Fig. 4). In specimens of C. sapaensis from northern Vietnam the talonid basin is broad and deep with an entoconid ridge and low entoconid, whereas in C. wuchihensis the talonid basin is narrow. In C. indochinensis the talonid basin is broad and deep with a hypoconid, entoconid and marked entoconid ridge (see Fig. 4).
Etymology. The new species is named after Sa Pa, the capital of Sa Pa District in Lao Cai Province of northern Vietnam, with the Latin suffix -ensis (belonging to).
Natural history. The series of type specimens was collected from a variety of habitats in the vicinity of Tram Ton Station of Hoang Lien National Park: mixed evergreen forest; forested banks of small streams; open grassy glades (Fig. 6); primary forest with large trees at an elevation 1930-2200m (Abramov et al. 2008b). During 2005-2010 a total of 190 shrews was captured in this area, including 4 species (Crocidura sapaensis, Blarinella griselda Thomas, 1912, Anourosorex squamipes Milne Edwards, 1872, and Episoriculus leucops (Horsfield 1855)). Crocidura sapaensis was the most numerous species (90% of the total captures), followed by A. squamipes and B. griselda (5.3% and 4.2% respectively), while only one E. leucops was captured . Crocidura sapaensis was more abundant in slightly disturbed mixed forest (2.2-3.0 specimens per 100 trap/nights), the occurrence in open glades, amongst shrubs on stream banks and in undisturbed primary forest was 0.3-2.6 specimens per 100 trap/nights. The proportion of males to females in C. sapaensis was greater in all seasons; on average the male to female ratio is 2.3. Pregnant females were recorded from May to mid-July. Mean litter size in C. sapaensis was 3.0 (2-4, n=15).
Distribution. Confirmed specimens of C. sapaensis are recorded from Lao Cai Province, Sa Pa District on the basis of cytb analysis and morphology of m3. On the basis of morphology, specimens from the northern part of Lao Cai Province, Ngai Tio (elevation 1450m) and from the vicinity of Cat Cat Village near Sa Pa Town (elevation 1400-1450m) in relatively close geographical proximity also probably belong to the same species.
Populations of C. wuchihensis identified on the basis of cytb and those probably representing this species on the basis of morphology (from Pa Kha and Thai Nien,
The population of C. wuchihensis recorded from Huong Son, Ha Tinh Province in the southern Annamites by Lunde et al. (2004) and Jenkins et al. (2009), does of course, occur west of the Song Hong and samples have not been included in any of the previous molecular studies. Specimens from Mt Tay Con Linh II are larger on average (CIL 15.7-17.1, mean 16.4) than those from Huong Son mean 16.0). The Canonical Variate Analysis reported in Jenkins et al. (2009: Fig. 10) shows that these two groups respectively from northern Vietnam and the southern Annamites are moderately well separated from each other. In view of the problems outlined in this paper, lacking further evidence from molecular studies, it is impossible to predict if the population from Huong Son is correctly assigned to C. wuchihensis, could belong to C. sapaensis, or might indeed represent a further undescribed species.

Characters separating C. attenuata and C. tanakae
The population from Mt Tay Con Linh II, Ha Giang Province in northern Vietnam recognised by molecular analysis of cytb as C. attenuata, falls within the size range of C. tanakae and both species are morphologically very similar in appearance. The two species may be separated by the following characters. The basioccipital region in C. attenuata is narrow and ridged particularly anterior to the position of the basioccipital suture, whereas in C. tanakae the basioccipital region is broad and flat to concave (see Fig. 7). The palatal suture in C. attenuata is a rounded to flat-topped 'n' shape, whereas in C. tanakae the suture is a shallow to more marked 'm' shape (see Fig. 8). The two species also differ in the shape of the talon of the upper premolar (P4). In occlusal view, the talon of C. attenuata is broader and more angular than that of C. tanakae, the lingual border is straight to concave, with the posterior border shallowly indented so that the whole tooth looks larger in occlusal view. In C. tanakae the lingual border of P4 is rounded and the posterior border of the tooth is deeply indented (see Fig. 8).
Geographical variation in C. tanakae Esselstyn and Oliveros (2010) did not provide detailed results about Vietnamese samples in the text of their analysis of Asian C. tanakae, nevertheless they demonstrated apparent geographical variation in Vietnam. Their illustration of a statistical parsimony network of mitochondrial haplotypes (Esselstyn and Oliveros 2010: Fig. 4) shows northeastern Vietnam samples from Tam Dao (21°27'N, 105°38'E) and Tuyen Quang (22°20'N, 105°25'E) grouped relatively closely, separated from each other by relatively few steps but separated by multiple steps from samples from the other two localities, Ha Tinh (18°21'N, 105°13'E) and Quang Nam (15°12'N, 108°02'E), which form a looser group. In their analysis of the COI gene, Bannikova et al. (2011) demonstrated the presence of two clearly defined haplogroups within Vietnamese C. tanakae: C. tanakae B restricted to the northern part of the country (Hoang Lien Mountains, Van Ban District) and C. tanakae A which was more widespread in Central and South Vietnam (Huong Hoa, Phong Nha-Ke Bang, Ngoc Linh, Hon Ba and Bi Doup). The uncorrected p-distance between these two groups using the COI gene was about 2.5%.
Populations of C. tanakae in Vietnam show a distinct clinal variation in skull size, populations at higher latitudes averaging smaller in size than those at lower latitudes (see Table 2). Although sample sizes are small, this observation is a possible example of the converse Bergmann's rule where body size decreases with latitude.
Geographical variation in C. attenuata Abramov et al. (2012) demonstrated apparent geographical variation of C. attenuata in southern China and northern Vietnam. Genetic differentiation of C. attenuata is notable and reveals a phylogeographic structure with four haplogroups. The specimens from Cat Ba Island (Hai Phong Province, northeastern Vietnam) formed a single cluster closely related to the group of specimens from northern Vietnam (Ha Giang Province) and southeastern China (Guangxi Province). The genetic distance (p-distance) between specimens from Cat Ba / Ha Giang as well as Cat Ba / Guangxi is about 2.1%. The specimen of C. attenuata from the more north-eastern region of China (Hunan Province) appears basal among all samples of C. attenuata from China and Vietnam. Thus, the genetic distance between two specimens from China (Hunan / Guangxi) is 4.3%, which is nearly the same as the distance between C. indochinensis and C. sapaensis (see Bannikova et al. 2011).

Distribution
While C. attenuata in Vietnam appears to occur only to the east of the Song Hong (Red River), in northeastern Vietnam (Bannikova et al. 2011, Abramov et al. 2012, C. tanakae does not appear to be so constrained and has been recorded on both sides of the river in northern Vietnam and also in central and southern Vietnam Oliveros 2010, Bannikova et al. 2011, this study).

Discussion
For sister species that are recognised on the basis of morphology, cytb distance values typically exceed 5% (Baker and Bradley 2006). The p-distance values of at least 9.91% between haplotypes belonging to the C. tanakae and C. attenuata groups, of 7.6% separating C. wuchihensis and C. indochinensis, and of 8.0% separating C. wuchihensis and C. sapaensis provide compelling evidence in support of their taxonomic distinction (Bannikova et al. 2011). The lower p-distance values of 4.1% separating C. in- dochinensis from C. sapaensis are somewhat less convincing, were it not for the ready morphological distinction of the two species. The conundrum is the morphological distinction yet relatively low molecular separation between C. indochinensis and C. sapaensis, in contrast to C. wuchihensis and C. sapaensis which are not easily defined by morphological features but are distinguished by high molecular values. Similar situations are found in the literature, for example amongst different species of Eumops Miller, 1906(bonneted bats) (McDonough et al. 2008. It is possible to speculate that this implies a relatively closer relationship between C. indochinensis and C. sapaensis but morphological convergence of C. wuchihensis and C. sapaensis to meet similar ecological requirements. Crocidura tanakae is currently recognised as being widely distributed in Southeast Asia including southern China, Vietnam, Taiwan and the Philippines Oliveros 2010, Bannikova et al. 2011). Crocidura attenuata has a more northerly distribution (Esselstyn and Oliveros 2010) and furthermore evidence suggests that, at least in Vietnam, the distribution may be more restricted than formerly understood (Bannikova et al. 2011, this study). It is conceivable that these two morphologically convergent species have secondarily come into contact with each other as there is evidence that the two species are sympatric in at least one Chinese locality (Esselstyn and Oliveros 2010, Judith Eger pers. comm.).