Sampling

A total of 28 species-group taxa (species and subspecies) were examined, of which 23 represented the carnivoran infraorder Arctoidea and five represented the carnivoran infraorder Aeluroidea (Table 1). For each of these species-group taxa, partial nucleotide sequences of two nuclear genes, the recombination-activating gene 1 (RAG1) and the gene encoding interphotoreceptor retinoid-binding protein (IRBP), were obtained by sequencing or extraction from the DDBJ/EMBL/GenBank databases. The RAG1 sequences are a fragment of the exon, 1095 base pairs (bp) in length, corresponding to human sites 1852–2946 (Schatz et al., 1989). The IRBP sequences are a fragment of exon 1, 1188 bp in length, corresponding to human sites 337–1317 and 1324–1530 (Fong et al., 1990). Because all examined Mustela lacked three IRBP base pairs (corresponding to human sites 1311–1313; Sato et al., 2003), this fragment of the gene was excluded from data analysis.

Table 1

Taxon, organism, and gene sampling, with DDBJ/EMBL/GenBank accession numbers

Isolation, amplification, and sequencing of DNA

Total genomic DNA was extracted from tissues preserved in ethanol by the conventional phenol-chloroform method. The amplification was performed via nested polymerase chain reactions (PCRs), using an automated thermal cycler (model PJ 2000, TAKARA). Each first PCR mix contained 10 mM Tris (pH 8.3), 50 mM KCl, 0.01% gelatin, 0.1% Triton X-100, 2.5 mM MgCl₂, 0.2 mM dNTP mix, 0.05 μM of each primer (1 pmol of each primer per reaction), 0.5 units of Amplitaq DNA polymerase (Applied Biosystems), and 0.1–0.5 μg of template total genomic DNA in a total volume of 20 μl. Thermal cycling parameters of the first PCR were as follows: RAG1-a cycle of denaturation at 95°C for 3 min and 30 cycles of denaturation at 95°C for 30 sec, annealing at 57°C for 30 sec, and extension at 72°C for 1 min; IRBP-30 cycles of denaturation at 94°C for 1 min, and annealing and extension at 70°C for 3 min each (Stanhope et al., 1992; Sato et al., 2003). A 1-μl aliquot of each reaction mixture after the first PCR was used as a template for the second PCR in a 20 μl reaction mixture with the same reagents except for the concentration of MgCl₂ (which was 1.875 mM) and the primer pairs. The second PCR was performed under the following conditions: RAG1-a cycle of denaturation at 95°C for 3 min and 30 cycles of denaturation at 95°C for 30 sec, annealing at 57°C for 30 sec, and extension at 72°C for 1 min; IRBP-35 cycles of denaturation at 96°C for 30 sec, annealing at 50°C for 30 sec, and extension at 60°C for 30 sec (Sato et al., 2003).

In the first PCR, a 1.1-kb fragment of RAG1 was amplified using primers RAG1F1842 and RAG1R2951, and a 1.3-kb fragment of IRBP was amplified using primers +IRBP217 and −IRBP1531 (Table 2). In the second PCR, two segments of RAG1 and three segments of IRBP were amplified against the respective products of the first PCR. For RAG1, the following two primer sets were used: (1) RAG1F1851 and RAG1R2486, and (2) RAG1F2357 and RAG1R2951. For IRBP, the three primer sets were used: (1) R +IRBP335 and U −IRBP734, (2) R +IRBP724 and U −IRBP1145, and (3) R +IRBP1085 and U −IRBP1532. The sequencing of the products of the second PCR was carried out according to the manufacturer's instructions, using automated sequencing (Big Dye Terminator cycle sequencing kit) on an ABI 310.

Table 2

Primers used to amplify the RAG1 and IRBP genes

Data analysis

A χ²-test of homogeneity was applied to test the assumption of base-compositional homogeneity within either single-gene data set. To test significance of incongruence between the two data sets, the partition homogeneity test was performed. The Kimura two-parameter method was used to compare per site substitution rates between the two genes.

Phylogenetic analysis was done on a combined data set (2280 bp) of the two genes by using the maximum-parsimony optimality criteria and equal weighting of nucleotide substitutions. The analysis was conducted using 100 heuristic tree-bisection reconnection searches in which the input order of taxa was randomized. As recommended by Barriel and Tassy (1998), and references therein, more than one outgroup taxon was used. Trees were rooted such that the collective aeluroid outgroup was forced to be monophyletic with respect to the monophyletic arctoid ingroup, in accordance with the current views on carnivoran phylogeny (e.g., Dragoo and Honeycutt, 1997; Flynn and Nedbal, 1998; Bininda-Emonds et al., 1999). To assess statistical support for hypothesized clades, bootstrap analysis was done with 1000 bootstrap replicates sampling 100 replicates of the random stepwise-addition option. In addition, the decay index, representing the number of extra steps required for a clade not to be unequivocally supported, was calculated.

All analyses were performed using PAUP* version 4.0b10 (Swofford, 1998). In addition, TreeRot version 2b (Sorenson, 1999) was used to calculate the decay index.

Nucleotide variation

Heterozygosity was found in five mustelid nucleotide sequences of RAG1 and four mustelid, two felid, and one viverrid sequences of IRBP (Table 3). Sequence composition statistics for both gene fragments are listed in Table 4. For either gene, the null hypothesis of homogeneity in base composition across the arctoid taxa was not rejected by the χ²-test (P > 0.05).

Table 3

Heterozygosity found in the RAG1 and IRBP fragments among the 28 carnivorans sampled

Table 4

Sequence composition statistics at different codon positions for the RAG1 and IRBP fragments from the 23 arctoids sampled

Phylogenetic inference

The partition homogeneity test did not reject the null hypothesis of homogeneity in phylogenetic signal between the RAG1 and IRBP data sets. In addition, a comparison of per site substitution rates between these genes (Fig. 1) indicates similar rates of evolution at least as far back as the divergence between the lineages leading to Arctoidea and Aeluroidea. Therefore, a combined analysis of the two data partitions is justified (Huelsenbeck et al., 1996).

Fig. 1

Comparison of per site substitution rates between the carnivoran RAG1 and IRBP, as estimated by using the Kimura two-parameter method. The single diamond is for a comparison between the subspecies of Mustela putorius. Open squares are for pairwise comparisons among species-group taxa within a genus (Martes, Mustela, Procyon, Panthera). Triangles are for pairwise comparisons among species-group taxa of different genera within a subfamily (Mustelinae, Ursinae). Open circles are for pairwise comparisons among species-group taxa of different subfamilies within a family (Mustelidae, Felidae). Filled circles are for pairwise comparisons among species-group taxa of different families within an infraorder (Arctoidea, Aeluroidea). Filled squares are for pairwise comparisons between species-group taxa of Arctoidea and Aeluroidea.

Maximum-parsimony analysis of the combined data set yielded eight shortest trees. The strict consensus of these trees is shown in Fig. 2. There is robust evidence of an out-group position of Ursidae with respect to a clade containing Mustelidae and Procyonidae. The family Mustelidae, as traditionally circumscribed, is found to be paraphyletic. The paraphyly is caused by the strongly supported outgroup position of Mephitis mephitis (representing Mephitinae) in relation to a clade containing the rest of the mustelids studied and Procyonidae. The subfamilies Melinae and Mustelinae, as traditionally circumscribed, are also found to be nonmonophyletic. While Melogale moschata (generally classified as a meline) is relatively strongly supported as basal to the strongly supported clade containing Mustela (the name-bearing type of Mustelinae) and Enhydra (representing Lutrinae), Meles meles (the type species of the type genus of Melinae) is placed in an unresolved trichotomy with the strongly supported clade containing Martes and Gulo (both usually classified as mustelines) and the Melogale-Mustela-Enhydra clade.

Fig. 2

Strict consensus of the eight shortest trees (length, 639 steps; consistency index, 0.70; retention index, 0.89) resulted from maximum-parsimony phylogenetic analysis of concatenated partial nucleotide sequences of RAG1 and IRBP, rooted using aeluroids as outgroups. Numbers above branches are percentage bootstrap values in support of adjacent nodes, and numbers below branches are the decay indices. A traditional circumscription for arctoid families and subfamilies (Wozencraft, 1993) is indicated.

The genus Martes is weakly supported as monophyletic, and there is strong evidence for the monophyly of the subgenus Martes (M. martes, M. americana, M. zibellina, M. foina). Within this subgenus, Martes foina and Martes zibellina are moderately supported as the closest relatives.

The monophyly of the genus Mustela is strongly supported. Strong support, too, is found for Mustela vison and Mustela erminea as basal and successively more closely related to a clade encompassing the remainder of the studied species of the genus. Within this clade there exists a relatively well-supported dichotomy between the small-sized (M. nivalis, M. altaica) and large-sized (M. lutreola, M. sibirica, M. eversmanii, M. putorius) species. Mustela lutreola is strongly supported as basal to the remaining large-sized species, which, in turn, are placed in an unresolved trichotomy. The conspecifity between Mustela putorius putorius and Mustela putorius furo is relatively weakly supported.

Phylogenetic and taxonomic position of the ferret-badgers

That the ferret-badgers Melogale moschata, Melogale personata (the name-bearing type of the genus), Melogale orientalis, and Melogale everetti are each other's closest relatives among extant mustelids is convincingly evidenced by morphological data (e.g., Everts, 1968; Long, 1981; Long and Killingley, 1983; Bininda-Emonds et al., 1999) and has probably never been questioned. The congeneric status of the four species is generally accepted (e.g., Corbet and Hill, 1992; Wozencraft, 1993; Nowak, 1999). Therefore, the use of Melogale moschata as a proxy for Melogale in exploring the phylogenetic position of this genus is justified. Hence, on the basis of the result of our analysis using Melogale moschata (Fig. 2), we put forward the hypothesis that Melogale is an outgroup to a clade containing Mustela and Enhydra, and is more closely related to these genera than it is to Meles, Martes, or Gulo.

Several competing, although less well statistically supported, hypotheses of phylogenetic relationships of Melogale have recently been presented. A phylogenetic analysis using 46 morphological characters, carried out by Bryant et al. (1993), weakly supported a basal placement of Melogale in relation to the remainder of the extant Mustelidae. On the basis of bacular morphology (17 characters), Baryshnikov and Abramov (1997, 1998) proposed placement of Melogale in a sister-group position to Gulo, also indicating only a distant relationship of Melogale with Meles. Using morphological data derived from literature sources, Bininda-Emonds et al. (1999) placed Melogale in a polytomy with a Meles-Arctonyx-Mydaus clade, Taxidea, a lutrine-mephitine clade, and a clade including the remaining extant mustelids. Finally, on the basis of karyological evidence, Nie et al. (2002) suggested an outgroup position of Melogale with respect to a clade containing Meles and Mustela.

Although all these phylogenetic hypotheses, including ours, differ from each other in the placement of Melogale within Mustelidae, they consistently indicate that this genus is not part of the monophyletic Melinae. This, together with morphological evidence presented in support of the nonmeline status for Melogale by earlier authors (Schlosser, 1888; Pocock, 1922; Petter, 1971; Rabeder, 1976; Schmidt-Kittler, 1981, 1984; Baryshnikov and Averianov, 1990), justify the exclusion of the ferret-badgers from the monophyletic Melinae. Therefore, we recommend that Melogale be classified in a distinct mustelid subfamily, the monotypic Helictidinae, as originally proposed by Gray (1865; his tribe Helictidina, elevated to subfamily rank by Gill [1872]).

Phylogenetic and taxonomic position of the skunks

While the monophyly of the skunks (including Mephitis, Spilogale, Conepatus, and Mydaus) is well supported by evidence from both morphology (e.g., Schmidt-Kittler, 1981; Bryant et al., 1993; Wolsan, 1999) and genetics (e.g., Dragoo and Honeycutt, 1997), the two databases indicate considerably different phylogenetic positions for this clade. Morphological evidence from both living and fossil taxa (e.g., Wozencraft, 1989; Bryant et al., 1993; Wyss and Flynn, 1993; Baskin, 1998; Wolsan, 1999), as well as some genetic evidence (protein electrophoresis-O'Brien et al., 1989), support the traditional placement of the skunks within the monophyletic Mustelidae. In contrast, genetic evidence from DNA hybridization (Árnason and Widegren, 1986; Wayne et al., 1989; Árnason and Ledje, 1993) and nucleotide sequencing (mitochondrial genes-Vrana et al., 1994; Ledje and Árnason, 1996a, b; Dragoo and Honeycutt, 1997; nuclear genes-this paper; combined mitochondrial and nuclear genes-Flynn et al., 2000), as well as combined evidence from mitochondrial genes and selected morphological characters (Vrana et al., 1994; Dragoo and Honeycutt, 1997), support the skunks as an outgroup to a clade containing the rest of Mustelidae, and also Procyonidae.

Taking account of the persisting conflict between phylogenetic interpretations based on primarily morphological and primarily genetic grounds, we consider the recent elevation of the skunks to the level of family (Ledje and Árnason, 1996a, b; Dragoo and Honeycutt, 1997) as premature, and conservatively recommend that this clade be left at the subfamily level (Mephitinae) within the family Mustelidae, pending further evidence.