Phylogenetic relationships among 31 operational taxonomic units of shrews (Soricidae, Mammalia), mainly from eastern Eurasia, were inferred from partial nucleotide sequences of the mitochondrial cytochrome b gene by maximum likelihood (ML) and neighbor joining (NJ) methods. Eleven monophyletic groups were recognized among the soricine shrews examined in the ML tree. However, branching orders of the groups were obscure judging from the local bootstrap values, and two out of the 11 groups were not monophyletic in the NJ tree. The phylogenetic relationships among Sorex caecutiens, S. shinto, and S. sadonis in the Japanese and Sakhalin islands, whose taxonomy was controversial, were clarified. S. shinto in the Honshu and Shikoku Islands is genetically differentiated enough to be considered a separate species from S. caecutiens, while S. sadonis could be treated as a subspecies of S. shinto. Some other taxonomic problems are also discussed.
INTRODUCTION
The modern Soricidae (Insectivora, Mammalia) is divided into two subfamilies, the Soricinae, red-toothed shrews, and the Crocidurinae, white-toothed shrews (Repenning, 1967). Some authors recognize a third subfamily, the Scutisoricinae (armored shrew), as being independent from the Crocidurinae (Simpson, 1945). Soricine shrews are distributed throughout the Holarctic region, consisting of approximately 110 species, while the crocidurines are diversified mainly in Africa and southern Asia (MacDonald, 1984; Abe, 1985; Corbet and Hill, 1986; Churchfield, 1990; Hutterer, 1993). In the northeastern Asiatic region, including eastern Siberia, up to 10 soricine species co-exist, with the highest species diversity in Eurasia (Dolgov, 1985). Therefore, northeastern Asia is thought to be a key region in which to investigate the evolution of soricine shrews.
It is quite difficult to reconstruct the speciation process of shrews, because their fossil record is poor owing to minute and fragile skeletons (but see, Repenning, 1967; Savage and Russell, 1983; Reumer, 1987) and because their morphology is similar to each other (Churchfield, 1990). The recent development of molecular phylogenetics has made it relatively easy to estimate phylogenetic relationships among modern stocks, even if they have very similar morphologies. Phylogenetic relationships of some Nearctic and Palaearctic soricine species have been investigated by biochemical techniques (George, 1988; George and Sarich, 1994) and mitochondrial DNA (Stewart and Baker, 1994; Taberlet et al., 1994). However, phylogenetic relationships based on molecular data among most Eurasian species remain unresolved, although Dannelid (1991) proposed a hypothetical phylogeny of Eurasian Sorex, mainly based on karyotypes. Furthermore, the taxonomy of some Japanese shrews (Sorex caecutiens, S. shinto, and S. sadonis) is controversial, because of the difficulty of interpreting morphological data (Imaizumi, 1960; Abe, 1967; Yoshiyuki and Imaizumi, 1986).
The major objective of this paper is to estimate the phylogenetical relationships and to clarify the taxonomic status among species or populations of Eurasian soricine shrews, especially, those from northeastern Asia, where there is high species diversity. We determined partial nucleotide sequences of the mitochondrial cytochrome b gene, which is thought to be an appropriate marker to estimate phylogenetic relations at species or subspecies level (Zhang and Ryder, 1993; Cao et al., 1994; Masuda and Yoshida, 1994a; Sturmbauer et al., 1994; but see William et al., 1995). Then, phylogeny of the shrews was estimated by using the maximum likelihood and the neighbor-joining methods.
MATERIALS AND METHODS
OTUs examined
Thirty one operational taxonomic units (OTUs) of shrews (Soricidae), a mole (Mogera wogura), and the rat (Rattus norvegicus) were used for phylogenetic analysis (Table 1). These OTUs include all of the soricine species in northeastern Asia. We basically followed Hutterer's (1993) nomenclature; however, Sorex caecutiens in Hokkaido (‘Ezo-togarinezumi’ shrew) and Mogera wogura (‘Kobe-mogura’ mole) were used (Abe, 1967, 1995), instead of S. shinto and M. robusta by Hutterer (1993), respectively. Tribe names followed those used by Repenning (1967).
Table 1
Abbreviations and sources of the OTUs examined
Sequencing of DNA
The first 402-base region of the mitochondrial cytochrome b gene was sequenced. Total DNAs were extracted by the phenol/proteinase K/SDS method (Sambrook et al., 1989) or by the Chelex® 100 method (procedure of ‘Chelex DNA extraction from hair’ in Walsh et al., 1991). The DNA extracts (ca. 500 μl) from dried tissues of old samples (Table 1) using the first method were concentrated to ca. 45 μl with TE buffer (10 mM Tris: 1 mM EDTA, pH 7.0–7.5) using Centricon™−30 microconcentrator columns (Amicon). The 402-base cytochrome b region was sequenced using the PCR (polymerase chain reaction) product/direct sequencing technique by Kocher et al. (1989) with a modification of Masuda and Yoshida (1994a, b). A slight modification of the present study was that 0.5–2 μl bovine serum albumin (20 μg/μl) was added to a 50-μl PCR reaction mixture to eliminate reaction inhibitors (Cooper, 1994), when DNA extracts were from dried tissues by the phenol/proteinase K/SDS method. For the PCR amplification, the forward primers on the light strand were designed as 5′-GACCAA-TGATATGAAAAACCATCG-3′ (L14721), 5′-GATATGAAAAACCATC-GTTG-3′ (L14724), and 5′-AAAAACCATCGTTGTTATTCAACT-3′ (L14734), and the reverse primer on the heavy strand as 5′-CTCAGAATGATATTTGTCCTCA-3′ (H15149) and 5′-GCCCCTCAG-AATGATATTTGTCCT-3′ (H15151). Codes of primers in parentheses identify the light (L) or heavy (H) strand and the 3′ end-position of the primer in the human mitochondrial DNA sequence (Anderson et al., 1981). Internal primers were 5′-AGCAATACACTACACATCAGA-3′ or 5′-CACATCTGCCGAGACGTAAA-3′ for forward sequencing, and their complements for reverse sequencing. These internal primers were also used as forward or reverse primers for PCR when a full 402-base region could not be amplified. Choice of these primers was determined by ‘trial and error’ for each sample. For doubtful results, both forward and reverse sequencings were done. Sequence data are available from DNA Data Bank of Japan ( http://www.ddbj.nig.ac.jp): accession numbers D85342-D85372 and D87025. Sequence of the rat was cited from Goetz and Feldman (1982).
Phylogenetic tree
A phylogenetic tree was inferred by the maximum likelihood (ML) method (Felsenstein, 1981), using the program NucML of MOLPHY Ver. 2.3 package (Adachi and Hasegawa, 1996). The third codon position of the cytochrome b gene (133 bases, excluding the first triplet) were included for analysis because few amino acid substitutions were found among the OTUs of the genus Sorex. A distance matrix was generated using a nucleotide substitution model by Hasegawa et al. (1985) and using the most likelihood value (25.5) of the rate of transition to transversion (α/β). The neighbor-joining tree (Saitou and Nei, 1987) was obtained using the distance matrix. Next, the tree which had the greatest log likelihood was searched for by using the local rearrangement search option of the NucML, starting from the neighbor-joining tree obtained (and also from some given topologies). To show the reliability of internal branches, local bootstrap values (Adachi and Hasegawa, 1996) of 1000 replications were calculated. The procedure using only the third codon position is not appropriate for the analysis among distantly related OTUs because nucleotide substitutions would be saturated. To complement the weakness of the ML method using only the third codon position, the neighbor-joining (NJ) method was also performed using the full 402-base data sets. The NJ tree was obtained by the computer program CLUSTAL W Ver. 1.5 (Thompson et al., 1994) with the multiple substitution correction (Kimura, 1980). Bootstrap values of 1000 replications were calculated to assess the confidence of internal branches. Relative efficiencies of some algorithms for phylogeny reconstruction using molecular data were examined by Hasegawa and Fujiwara (1993), where the ML method tended to find a correct phylogeny.
RESULTS
A 402-base region of the cytochrome b was successfully sequenced for all OTUs. No insertions or deletions were observed.
In the ML tree (Fig. 1), monophyly among the OTUs from the Soricinae and the genus Sorex was strongly supported by the local bootstrap value (≥89%). We tentatively propose eleven groups within the soricine shrews examined, according to the bootstrap values of the ML tree (≥ 65%); (1) Soriculus group (Soriculus nigrescens and So. caudatus, Asiatic mountain shrews), (2) Neomys/Chimarrogale group (Neomys fodiens and Chimarrogale platycephala, water shrews), (3) cinereus/fumeus group (Sorex cinereus and S. fumeus, the subgenus Otisorex in North America), (4) S. raddei, (5) unguiculatus/isodon/mirabilis group (S. unguiculatus, S. isodon, and S. mirabilis), (6) araneus/tundrensis/daphaenodon group (S. araneus, S. tundrensis, and S. daphaenodon), (7) S. gracillimus, (8) S. minutus, (9) S. roboratus, (10) minutissimus/hosonoi group (S. minutissimus and S. hosonoi), and (11) caecutiens/shinto group (S. caecutiens, S. shinto, and S. sadonis). Branching orders of these groups were obscure judging from the local bootstrap values.
Fig. 1.
The maximum likelihood (ML) tree by the nucleotide substitution model of Hasegawa et al. (1985), using third codon position (133 bases) of the mitochondrial cytochrome b gene sequences in insectivores. Branch length is proportional to the estimated number of nucleotide substitutions per site. Numbers near internal branches indicate the local bootstrap values (%) of 1000 replications.
Ignoring bootstrap values, nine out of the 11 groups proposed by the ML tree were also recognized in the NJ tree (Fig. 2), although branching orders of the groups were rather different from those of the ML tree. The unguiculatus/isodon/mirabilis group was non-monophyletic in the NJ analysis (Fig. 2). The Soriculus group was also non-monophyletic, but the four OTUs of the Neomyini were monophyletic in this tree (Fig. 2).
Fig. 2.
The neighbor-joining (NJ) tree of the mitochondrial cytochrome b gene sequences (402 bases) in insectivores. Branch length is proportional to the estimated number of nucleotide substitutions per site by Kimura's two-parameter method. Numbers near internal branches indicate the bootstrap values (%) of 1000 replications.
DISCUSSION
The hypothesis that all the OTUs from the Soricinae are monophyletic was strongly supported by the nucleotide sequences (Figs. 1 and 2). We did not examine any samples from the Blarini (short-tailed shrew in North America), the third tribe of the Soricinae, but a close relationship between the tribes Soricini and Blarini is accepted by morphological (Repenning, 1967) and biochemical (George, 1988; George and Sarich, 1994) data. Therefore, the Soricinae is presumably monophyletic.
The ML tree (Fig. 1) did not support a monophyletic relation between the two groups of the Neomyini, but the NJ tree (Fig. 2) demonstrated a monophyletic relation among the four species of the Neomyini. In the ML analysis, data included only the third codon position, and substitutions among the OTUs from the Neomyini, Crocidurinae, mole, and rat were almost saturated (Fig. 1). Therefore, non-monophyly of the Neomyini in the ML tree is probably an artifact of the analytical method used. In such a case, the results of the NJ analysis using the full 402-base data are supplementary to that of the ML analysis using only the third codon position. The NJ tree showed that the monophyly of the Neomyini although the Soriculus group was not monophyletic (Fig. 2). Non-monophyly of Soriculus in the NJ tree might be caused by insufficient data size, inferring from the low bootstrap values within the Neomyini. In addition, Reumer (1984) suggested that from the morphological and biogeographical view points the extant Neomyini should be divided into two different tribes, Notiosoricini and Soriculini. We did not analyze any of the samples from the Notiosoricini. Thus, the phylogenetic status of the Neomyini remains uncertain.
Hutterer (1993) listed some subgenera within Eurasian Sorex. However, the subgenera Stroganovia and Ognevia obviously should be included in the subgenus Sorex according to our result (Figs. 1 and 2), as Dannelid (1991) pointed out.
S. unguiculatus, S. isodon, and S. mirabilis are monophyletic in the ML tree (Fig. 1), but S. mirabilis was excluded from a cluster of the other two in the NJ tree (Fig. 2). Therefore, the phylogenetic status of S. mirabilis is not rigorously defined. Longer sequences and information of other gene sites are necessary to confirm its phylogenetic position. In contrast, a monophyletic relation of the araneus/tundrensis/daphaenodon group was strongly supported by both of the ML and NJ trees (Figs. 1 and 2). Their monophyletic relationships is also suggested by chromosome type (XY1Y2 sex-determination system; Fedyk and Ivanskaya, 1972; Reumer and Meylan, 1986; Dannelid, 1991).
Sorex gracillimus (Asiatic pygmy shrew) has long been regarded as S. minutus (European pygmy shrew), but is now accepted as a separate species (Hutterer, 1993). George (1988) also proposed from allozyme data that these two species should be separated, but showed that they still had a close relation. The present analysis based on mitochondrial DNA sequences, however, indicates that the two species are rather distantly related (Figs. 1 and 2), as proposed by Dannelid (1991).
There is taxonomic disagreement for the caecutiens/shinto group in Japan and Sakhalin. Abe (1967) did not recognize Sorex shinto. He regarded specimens from the Honshu, Shikoku, Hokkaido, and Sakhalin Islands all as being S. caecutiens. On the other hand, Imaizumi (1960) treated S. shinto as a separate species from S. caecutiens and included “S. caecutiens” of Hokkaido and Sakhalin in S. shinto (names between double quotation marks are the usage in the present study). However, both of the ML and NJ trees produced two clusters in the caecutiens/shinto group (Figs. 1 and 2). One cluster consists of S. caecutiens in Finland, Siberia, Sakhalin, and Hokkaido, and the other contains S. shinto in Honshu and Shikoku and S. sadonis. The genetic distance between the two clusters is great enough to be regarded as a between species difference (Figs. 1 and 2). George (1988) also treated S. shinto as a separate species based on allozyme analysis. Yoshiyuki and Imaizumi (1986) described S. sadonis as a new species from the Sado Island, Japan. According to our usage of species names (Table 1), there would be no inconsistency between the nomenclature for S. sadonis and its phylogeny (Fig. 1). However, the genetic distance between S. sadonis and S. shinto appears to be small enough for these OTUs to be regarded as subspecies or populations, when compared with the distances among the other species examined (Figs. 1 and 2). Consequently, the usage of S. shinto or S. shinto sadonis might be appropriate for S. sadonis.
Branching orders among the groups suggested by the ML tree and among OTUs within the group were obscure within the genus Sorex, judging from the bootstrap values (Fig. 1). Further, many bootstrap values within Sorex were very low in the NJ tree (Fig. 2). Such uncertainty of the phylogenetical relationships of Sorex was probably caused by the insufficient data size. Data from longer sequences and other genes are needed to confirm the phylogenetical relationships of Sorex.
Acknowledgments
We are grateful to V. Nesterenko, E. Pankakoski, H. Henttonen, and H. Shinyashiki for supplying some shrew samples. M. J. Toda and T. P. Craig critically commented on earlier versions of the manuscript. Without their cooperation, this work could not have been done. This study was supported in part by Grants-in-Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture, Japan and by the cooperative research projects of the Institute of Low Temperature Science, Hokkaido University.
