SYSTEMATICS OF MUSTELID-LIKE CARNIVORES

JERRY W. DRAGOO AND RODNEY L. HONEYCUTT

Departmentof Wildlifeand Fisheries Sciences, TexasA&M University,College Station, TX 77843 Present address of JWD: Museumof SouthwesternBiology, Departmentof Biology, Universityof New Mexico, Albuquerque,NM 87131

The phylogenetic relationships of the to the and other caniform carni- vores were examined using mitochondrial-DNA (mtDNA) sequence data from portions of the 12S and 16S ribosomal RNA (rRNA) genes. Data were combined with partial sequences of the cytochrome b gene and morphological data obtained from the literature, and used in a total-evidence analysis. The Mustelidae represented a paraphyletic group, with the skunks (Conepatus, , and Spilogale) and the Oriental stink (Mydaus) forming a monophyletic clade separate from a clade containing the rest of the Mustelidae and the monophyletic . Within the Mustelidae, minus the skunks and , only one currently recognized subfamily, the Lutrinae, represented a monophyletic group. The families Phocidae, Otariidae, and Odobenidae formed a monophyletic group that was the sister group to the clade composed of the skunks, procyonids, and mustelids. The families Ursidae and occurred at the base of the clade. It is proposed that the skunks be elevated to the level of family and be referred to as the . The family Mephitidae includes the genera Mephitis (striped and hooded skunks), Conepatus (hog-nosed skunks), Spilogale (spotted skunks), and Mydaus (Oriental stink ).

Key words: Mephitidae, Mustelidae, , skunks, phylogeny,

In F. Gabriel Sagard-Theodat's "Histoire itinae (skunks), Melinae (badgers), Lu- du Canada," skunks were described as "les trinae (), and (the rest of enfants du diable," children of the devil the mustelids). Classically, the Mellivorinae (Coues, 1877; Seton, 1926), and their () has been recognized as a unique method of defense and self- fifth subfamily (Anderson, 1989; Simp- preservation has sustained that reputation. son, 1945; Wozencraft, 1993). Wozencraft The enlarged anal scent glands of skunks (1989, 1993) placed the have long been one of the characters that (Taxidea) as a separate subfamily, the Taxid- unites them with the family, Musteli- iinae, and Anderson (1989) recognized the dae. However, this character alone does not South American genera Eira, , and define the relationship of skunks to any of Lyncodon as components of a distinct sub- the other mustelids. family, the Galictinae. Anderson (1989) also Extant members of the Mustelidae are di- mentioned the possibility of distinguishing agnosed as a monophyletic group on the the South American subgenus Grammogale basis of the loss of the carnassial notch on ( Mustela) as a distinct subfamily. the upper fourth premolar, the loss of the Not only has the designation of subfami- upper second molar, as well as the enlarged lies in the Mustelidae been problematic, but scent glands (Bryant et al., 1993; Martin, the relationships among taxa within sub- 1989; Wozencraft, 1989). Four, to as many families and among many subfamilies have as seven, subfamilies are recognized cur- been difficult to decipher. The monophyly rently (Anderson, 1989; Eisenberg, 1989; of the North American skunks (Mephitinae) Wozencraft, 1989, 1993). The four sub- is well supported (Anderson, 1989; Dragoo families commonly accepted are the Meph- et al., 1993; Wozencraft, 1989, 1993), with

Journal of Mammalogy, 78(2):426-443, 1997 426 May 1997 DRAGOO AND HONEYCUTT-SYSTEMATICS OF SKUNKS 427 three extant genera recognized, including (1993) based on a cladistic analysis of the Mephitis (hooded and striped skunks), Spi- Mustelidae using cranial, post-cranial, and logale (spotted skunks), and Conepatus soft anatomy suggested that Mydaus is a (hog-nosed skunks). Two Old World taxa, member of the Mephitinae. Ictonyx (African zorilla) and Mydaus (Ori- Recent morphological data (several fea- ental stink badger), among others, have tures of the cranium, dentition, and soft been included at various times in the sub- anatomy) have indicated a sister-group re- family Mephitinae. lationship between the Lutrinae and the Me- The inability to determine monophyletic phitinae (Hunt, 1974; Wozencraft, 1989; groups within the Mustelidae has contrib- Wyss and Flynn, 1993). These studies, how- uted to confusion regarding sister-group ever, were interested in higher-level rela- relations within the family. The African tionships and, thus, did not examine rela- zorilla (Ictonyx) has a color pattern that tionships below the familial level, except converges on that of the North American within the Mustelidae, which was analyzed spotted skunks (Spilogale). Many of the at the subfamilial level. early naturalists confused the two genera One of the primary difficulties in deter- (Nowak, 1991). Coues (1877) recognized mining the relationships among subfamilies the African Zorillinae as the Old World of mustelids is the diagnosis of monophy- representative nearest to the Mephitinae. letic groups on the basis of synapomorphies O'Brien et al. (1989), in a protein electro- as opposed to symplesiomorphies. The mor- phoresis study of black-footed and phological data uniting the skunks with any other in the genus Mustela, used particular subfamily of the Mustelidae, or skunks, including Ictonyx (which was re- even to the family, have been based on ple- ferred to as the African striped ) as an siomorphic character states and convergent outgroup. They suggested an ancient split similarity. Recent examinations of non- between the skunks sensu lato and the ge- morphological characters have revealed a nus Mustela. somewhat different picture of relationships Traditionally, Mydaus has been placed of mustelids. For example, Wurster and Be- within the subfamily Melinae, yet some nirschke (1968:374) studied chromosomal consider the Melinae to be a polyphyletic data of various carnivores and indicated that group, primarily based on its inclusion the "skunks are remarkably different from (Bryant et al., 1993; Petter, 1971; Pocock, the rest of the family." These data suggest 1921; Radinsky, 1973; Simpson, 1945). that, karyologically, skunks are apomorphic Earlier researchers (Petter, 1971; Pocock, relative to other mustelids; however, these 1921; Simpson, 1945) suggested a sister- characters provide little information about group relationship between the skunks and the relationships of skunks to the mustelids. the badgers based on similar cranial char- Ledoux and Kenyon (1975) studied serum acters shared by the stink badger and the proteins and suggested that the Mustelinae, Mephitinae. Radinsky (1973) asked the Melinae, and Lutrinae shared a common an- question: Are stink badgers skunks? His an- cestry long after the lineage leading to the swer was inconclusive because the charac- modem Mephitinae diverged. ters he identified as being shared were Recent molecular studies of relationships symplesiomorphies. Although the shared of carnivores based on DNA hybridization characters were plesiomorphic, Radinsky (Arnason and Widegren, 1986; Wayne et al., (1973) argued that the fossil record pro- 1989) suggested that the family Musteli- vided support for the skunks and stink bad- dae is paraphyletic. Although nucleotide- gers being a monophyletic group, with fos- sequence data (Vrana et al., 1994) have pro- sil skunks dating from the Miocene and vided additional support for this observa- Pliocene in the Old World. Bryant et al. tion, these authors did suggest that data for 428 JOURNALOF MAMMALOGY Vol. 78, No. 2 more taxa of both skunks and mustelids vores by increasing the sampling of taxa, es- were needed. Finally, Ledje and Arnason pecially within the family Mustelidae, and (1996) examined sequence data for the en- in so far as possible, taxa were selected in tire cytochrome b gene and concluded that a manner that would allow for the consid- Mephitis and Spilogale were not part of the eration of both morphological and molecu- mustelid clade. lar data. Is the family Mustelidae monophyletic? Part of the answer to this question pertains MATERIALSAND METHODS to the of the placement subfamily Mephiti- Molecularcharacters were used to examinere- nae. The mustelids have been a difficult lationships among 26 species of carnivores to and the Mustelidae is in group classify, (Table 1). Representativesof the (Fe- need of systematic revision. The objective lidae andHerpestidae) were used as the outgroup of this research was to address, using a mo- for analyses of the suborderCaniformia, which lecular approach, the phylogenetic relation- includes the skunks and the mustelids. The mo- ships among taxa of mustelids, especially lecular data consisted of mitochondrial-DNA those pertinent to the problem of mono- (mtDNA) sequences obtained during this study phyly of mustelids. Sequence data from two and, wheneverpossible, comparedto data in the GenBank databaseand recent mitochondrial genes (12S and 16S riboso- publications(Ar- nason and Johnsson, 1992; Janczewski et al., mal RNA) were used to assess the phylo- 1992; Vranaet al., 1994). genetic relationships among lineages of car- Total genomic DNA was isolated from frozen nivores, with a focus on primary diagnos- tissue, blood, and skin samples using the the clades within the Mustelidae ing major method of Hillis et al. (1990). Specific regions and examining the relationship of skunks to of mtDNA were amplifiedusing the polymerase the mustelids and other carnivores. The re- chain reaction (PCR). The primers used to sults also were compared with findings in amplify portions of the 12S and 16S rRNA recent publications of molecular and mor- genes were (LGL Ecological Genetics, Inc., phological studies investigating relation- Bryan, TX); LGL284 (5' TGG GAT TAG ATA ships of carnivores. CCC CAC TAT 3') and LGL383 (5' ATT GGT The 12S and 16S ribosomal RNA (rRNA) GGC TGC TTT TAG GCC 3'). These primers to 1,432 and 2,554 of the genes were selected for several reasons. correspond positions mtDNA of the , vitulina There is a large existing database of these (Arnason and Johnsson, 1992). The amplifica- for carnivores and other , genes tion productfor the rRNA genes was ca. 1,110 thus the examination of more taxa. allowing base pairs, including 518 base pairs of the 12S Relative to other genes within the mamma- rRNA gene, 513 base pairs of the 16S rRNA lian mitochondrial genome, the rate of se- gene, and 79 base pairs of the transfer RNA quence divergence for these genes is suit- valine gene. Double-strandedDNA (dsDNA) able for the suggested time periods over productswere obtainedwith PCR amplification which these lineages of carnivores diverged using Taq DNA polymerase (Saiki et al., 1986, (Janczewski et al., 1992; Lento et al., 1995; 1988), following similar procedures used by et al. All of the taxa were Miyamoto et al., 1990; Zhang and Ryder, Dragoo (1993). on an 373A DNA 1993). recent studies of sequenced AppliedBiosystems Finally, systemat- were attained the ics of carnivores have shown these to Sequencer. Sequences using genes Terminator be robust in terms of their in- Taq DyeDeoxy@ Cycle Sequencing phylogenetic Kit. Two procedureswere used for nucleotide- formation content and to ability diagnose sequence alignment. The Clustal V program et monophyletic groups (Janczewski al., (Higgins and Sharp, 1989) was used to align se- 1992; Lento et al., 1995; Vrana et al., 1994; quences, followed by visual inspection and Zhang and Ryder, 1993). We expanded alignment,taking into accounthighly conserved upon existing molecular studies of carni- areas. Sequences reported in this paper have May 1997 DRAGOOAND HONEYCUTT-SYSTEMATICSOF SKUNKS 429

TABLE 1.-Carnivore taxa used for molecular analyses of phylogenetic relationships. Taxonomy follows Wilson and Reeder (1993).

Family Subfamily Scientific name Common name Sourcea Mustelidae Mephitinae Conepatus mesoleucus Hog-nosed skunk JWD423 Mephitis mephitis JWD389 Spilogale putorius JWD405 Lutrinae Enhydra lutris Sea LGL canadensis River otter Vrana et al. (1994) cinerea Small-clawed otter IZL2940 Melinae Mydaus marchei Stink badger CMNH meles Eurasian badger IZL1111 Taxidiinae Taxidea taxus American badger TK26747 Mustelinae Ictonyx striatus Zorilla SP7550 Gulo gulo AF0354 Martes americana Pine AF0055 Mustela vison AK11208 Mustela frenata Long-tailed weasel JWD421 Mustela putorius European IZL686 Phocidae Phocinae Phoca vitulina Harbor Seal Amrnasonand Johnson (1992) Odobenidae Odobenus rosmarus LGL Otariidae Otariinae californicus Sea LGL Ursidae Ursinae americanus Black AMNH Procyonidae Procyoninae lotor JWD329 Bassaricus astutus Ringtail JWD392 Procyonidae Potosinae Potos flavus NK 13928 Canidae latrans H170 cinereoargenteus Gray JWD516 Herpestidae Herpestinae pulverulenta Cape gray TM1495 pardalis TAMU Kingsville a Collectionsloaning frozen tissues includingthe Universityof Alaska,Fairbanks (AF); TexasA&M University,College Station (AK, H, and JWD); Texas A&M University,Kingsville (TAMU Kingsville);American Museum of NaturalHistory, New York (AMNH);Cincinnati Museum of NaturalHistory (CMNH); Institute of Zoology, London(IZL); LGL Ecological Genetics, Bryan, Texas (LGL);Museum of SouthwesternBiology, Universityof New Mexico, Albuquerque(NK); Section of Mammals,Carnegie Museumof NaturalHistory, Pittsburgh, Pennsylvania (SP); The Museum,Texas Tech University,Lubbock (TK); TransvaalMu- seum, Johannesburg,South Africa (TM).

been deposited in the GenBank database (acces- 0.6 as suggested by Springer et al. (1995) for the sions U78326-U78350). 12S and 16S rRNA genes. Maximum parsimony was used to derive a Sequences for the 12S and 16S datasets were phylogeny from the nucleotide sequence data. analyzed both separately and in combination. All analyses were conducted using PAUP, ver- Due to the large number of taxa involved, all sion 3.1.1 (Swofford, 1993). The characters were maximum-parsimony analyses were performed treated as unordered, discrete characters with using the heuristic-search option with 100 rep- four possible states (A, C, G, and T). Gaps were lications, a random addition of taxa, and tree coded as present or absent, and an interleave ma- bisection-reconnection (TBR) branch swapping. trix was appended to the end of the sequence Homoplasy was evaluated using the consis- data. Both equal and unequal weighting schemes tency index (CI-Kluge and Farris, 1969) and were used. In equally weighted parsimony, all the retention index (RI-Farris, 1989). Tree substitutions were used regardless of either the length was used to determine the most- type (transition or transversion) or position in the parsimonious solution, and support for indi- gene. Unequal weighting involved the use of vidual clades was evaluated using both the transversions only, and an evaluation of stems decay index, the number of extra steps needed and loops where stems were down weighted to to collapse a node (Bremer, 1988), and boot- 430 JOURNALOF MAMMALOGY Vol. 78, No. 2

strapresampling using 500 replications(Felsen- when a strong transition-transversion bias is stein, 1985). absent (Kumar et al., 1993; Nei, 1991). Nei- ther transition nor transversion saturation RESULTS was apparent in either gene relative to the Patterns of sequence variation.-Patternms genetic distances compared. However, there of nucleotide-base composition were exam- were more transition changes than transver- ined using a chi-square test. No bias is in- sion changes in the 12S gene, especially at dicated by equal frequency (50%) of pu- lower levels of divergence. In the 16S gene, rines and pyrimidines and equal frequency the transition and transversion changes were (25%) of each base (A, C, G, and T). Tests about equal, again with a bias toward tran- were performed on each species to deter- sitions between less-divergent taxa. Beyond mine if nucleotide composition varied be- divergences of 10-15%, the ratio of transi- tween sequences (Lockhart et al., 1994). tions to transversions was relatively low, av- The results were nearly identical among eraging 1.75:1 and 1.22:1 for the 12S and taxa and between the two gene fragments. 16S genes, respectively. Although the frequency of purines and py- There were 103 insertion-deletion events rimidines in the 12S rRNA gene did not in the sequences examined. About 65% of differ from that expected by chance alone the gaps that occurred in the 12S sequences (0.5 > a > 0.1), there was a general ten- were the result of single insertion-deletions dency for more purines than pyrimidines. events, whereas 70% of the gaps were The frequencies of purines and pyrimidines single insertion-deletion events in the 16S also were about equal in the 16S sequence. sequences. The largest insertion-deletion There were, however, significantly more ad- (seven bases) occurred between Conepatus enines than guanines in both genes (a < and the outgroup in the 12S gene; no gaps 0.005). The percentage composition of ad- were longer than four bases in the 16S gene. enine ranged from 35.6 to 39.6% for 12S Phylogenetic analyses.-Parsimony anal- and 35.6 to 41.8% for 16S, cytosine ranged yses of the 12S and 16S sequence data were from 20.6 to 24.9% for 12S and 15.8 to performed separately. Five equally parsimo- 21.7% for 16S, guanine ranged from 16.7 nious trees (not shown) resulted from the to 19.4% for 12S and 16.1 to 19.8% for 16S, analysis of the 12S gene; analysis of the 16S and thymine ranged from 20.7 to 24.6% for gene produced nine most-parsimonious so- 12S and 22.6 to 27.0% for 16S. lutions (trees not shown). In all 14 trees, the The type of nucleotide substitution also taxa currently comprising the family Mus- was similar in all pairwise comparisons be- telidae did not form a monophyletic group. tween sequences. Transition changes be- The Mephitinae (skunks) and Mydaus (tra- tween thymine and cytosine tended to be ditionally placed in Melinae) formed a more common than changes between ad- monophyletic clade separate from a larger enine and guanine; changes between ad- clade containing the remainder of the Mus- enine and thymine were, on average, the telidae (Lutrinae, Melinae, and Mustelinae) more common transversions. Few transver- and the family Procyonidae. The relation- sions involving guanine and either cytosine ships of taxa within the Mustelidae were un- or thymine were observed. resolved. The genus Ictonyx (subfamily Genetic distances were calculated by the Mustelinae) was placed either sister to the method of Tajima and Nei (1984) using the Lutrinae or outside a clade containing Lutri- program MEGA, version 1.01 (Kumar et nae and the genus Mustela (Mustelinae), al., 1993). This distance, which provides a thereby making the Mustelinae paraphy- better estimate of the number of nucleotide letic. The placement of Gulo and Martes substitutions when nucleotide frequencies outside a clade containing Ictonyx, Mustela, deviate substantially from 0.25%, is useful and the Lutrinae also resulted in a paraphy- May1997 DRAGOOAND HONEYCUTT-SYSTEMATICSOF SKUNKS 431

97/11 Conepatus 99/11 1 Mephitis Mephitinae 85/8 Spilogale Mydaus Melinae 100/14 57/2 AonyxAoy 57/2 Lutra <50/1 Lutrinae <50/1 Enhydra 71/6 <50/1 Ictonyx S 99/12 Mustelavison 95/10 M.frenata <50/• Mustelinae M. putorius 96/7 79/6 Gulo Martes <50/1 99/14 Meles Melinae STaxidea Taxidiinae 88/7 90/9 [89/7 FProcyon Procyonidae Potos 52/1 Odobenus 89/7 Zalophus Phoca <50/1 Ursus Ursidae Canis i Canidae 100/18 Urocyon 100/35_Leopardus 100/35 Galerella FelifoiaI

FIG. 1.-Phylogenetic relationshipsamong carnivoresbased on parsimony analysis of all nucle- otide substitutions(tree length = 2,036; CI = 0.365; RI = 0.444) among 12S and 16S rRNA gene sequences. Numbers representbootstrap values/decay indices. letic Mustelinae. Other discrepancies among nodes not well supported by a single gene the nine trees obtained from the 16S rRNA alone (Miyamoto et al., 1994; Olmstead and sequences involved the placement of the Sweere, 1994). Parsimony analysis of the Mephitinae and Mydaus clade relative to the combined 12S and 16S data resulted in a pinnipeds, and the placement of the Cani- single most-parsimonious tree (Fig. 1). The dae relative to the Ursidae. bootstrap value and decay index for the The two datasets were combined because node uniting Mydaus with the Mephitinae both genes are linked and, thus, share a were 99% and 11 additional steps, respec- common evolutionary history (Miyamoto et tively. As was found with the separate al., 1994), and demonstrate similar patterns analyses of the 12S and 16S data, the mus- of divergence and base composition. Com- telids did not form a monophyletic group. bining data has the potential of increasing The Mephitinae and Mydaus grouped out- resolution due to the greater number of po- side the clade (bootstrap = 71%, decay in- tentially informative characters that support dex = 6 steps) containing a monophyletic 432 JOURNAL OF MAMMALOGY Vol. 78, No. 2

Procyonidae and the rest of the mustelids, of Canidae, Procyonidae, and the sister taxa a clade supported by a bootstrap of 88% and Mustelidae and Phocidae. Within Musteli- decay index of seven steps. dae, the subfamilies Mustelinae and Meli- Within the mustelid clade (minus the Me- nae were sister groups, as were Mephitinae phitinae and Mydaus), there was little sup- and Lutrinae. port for any particular grouping among the To evaluate how well our molecular data three recognized subfamilies. The Musteli- conformed to Wozencraft's (1989) hypoth- nae was paraphyletic based on the place- esis derived from morphology, we produced ment of Ictonyx, Gulo, and Martes. There a constraint tree for the taxa used in our was, however, strong support (bootstrap = study, based on the topology presented by 96%, decay index = 7 steps) for the mono- Wozencraft (1989). The length of the tree phyly of the mustelid taxa minus the Me- was determined using our 12S and 16S se- phitinae and Mydaus. quence data. Fifty-six additional steps (rela- An analysis using transversions only pro- tive to the most-parsimonious tree shown in duced a tree with a topology similar to that Fig. 1) would be needed for the molecular seen in Fig. 1, with the exception of the data to support Wozencraft's (1989) hypoth- placement of the Ursidae relative to the esis. This suggests a lack of congruence be- Canidae and the pinnipeds. As was found tween Wozencraft's (1989) morphological in the equal-weighting analysis, the Mus- results and those obtained from the molecu- telidae was paraphyletic when only trans- lar data presented here. versions were used. When stems were down Wyss and Flynn (1993), in an investi- weighted relative to loops, the resulting to- gation of the relationships among families pology was identical to that seen in Fig. 1. of carnivores, re-examined Wozencraft's (1989) characters and made the following DISCUSSION changes in their interpretation: 1) the scor- In recent years, several cladistic analyses ing of some characters for extant taxa was of morphological characters have addressed changed based on the inclusion of fossil relationships among carnivores. Flynn et al. taxa; 2) characters not considered indepen- (1988) reviewed morphological characters dent were combined; 3) ambiguous charac- used in studies of phylogenetics of carni- ters were rejected; 4) series of complex vores and suggested that two major clades characters with co-varying components were present in the monophyletic Car- were combined as a single character; 5) the nivora. One clade, the Caniformia, con- polarity of some characters was corrected sisted of Cynoidea (Canidae) and . based on outgroup comparison; 6) poorly The Arctoidea was an unresolved tri- justified ordered transformations were dis- chotomy comprised of Ursidae, the mono- missed. Their final dataset consisted of 64 phyletic Pinnipedia, and Musteloidea (Pro- morphological characters and both extinct cyonidae and Mustelidae). Relationships of and extant taxa. A phylogenetic analysis of subfamilies within Mustelidae were not ex- the data produced the following results. amined. Canidae was the sister group to the Arc- Wozencraft (1989) examined 100 mor- toidea as defined by Flynn et al. (1988), phological characters in a cladistic analysis Mustelidae was at the base of the Arctoidea of extant carnivores and proposed a differ- and the sister group to the remaining taxa, ent hypothesis for the phylogeny of Cani- Pinnipedia was the sister group to Ursidae, formia. He suggested that there were two and Procyonidae was the sister group to the superfamilies, Ursoidea and Canoidea. The clade comprised of Pinnipedia and Ursidae. clade containing Otariidae and Odobenidae In addition, they examined relationships with Ursidae as the sister taxon represented among subfamilies of Mustelidae and found the Ursoidea. The Canoidea was comprised that Mephitinae and Lutrinae were sister May 1997 DRAGOO AND HONEYCUTT-SYSTEMATICS OF SKUNKS 433 taxa. When the 12S and 16S data were con- rRNA and cytochrome b genes to ascertain strained to Wyss and Flynn's (1993) hy- relationships of carnivores. They deter- pothesis, the resultant phylogeny was 29 mined that Ursidae and pinnipeds were steps longer than the tree shown in Fig. 1. sister taxa. Procyonidae was paraphyletic; This result also suggested incongruence be- the sister taxa Procyon and Bassariscus tween the morphological and molecular were placed at the base of the ursid- data. clade and Potos was the sister In addition to the morphological studies taxon to this larger clade. Mustelidae was noted above, several molecular studies have paraphyletic and basal to an arctoid clade examined phylogenetic relationships among containing the ursids, pinnipeds, and pro- carnivores (Ledje and Amason, 1996; Miy- cyonids. Mephitinae, represented by Me- amoto and Goodman, 1986; Vrana et al., phitis, was placed outside the rest of the 1994; Wayne et al., 1989). Miyamoto and arctoid carnivores, and Canidae was the Goodman (1986) studied amino acid se- sister group to this entire clade. This hy- quences and suggested that within the Cani- pothesis required 18 additional steps rela- formia, Canidae and Ursidae were sister tive to our tree (Fig. 1). taxa, pinnipeds were monophyletic, and Ledje and Arnason (1996) examined Procyonidae and Mustelidae were sister DNA sequences for the entire mitochondrial taxa. Miyamoto and Goodman (1986) did cytochrome b gene. Although they were not not examine all the carnivores used in the able to resolve the major clades of arctoid present study, and did not include any taxa carnivores, they found that the skunks were of the subfamily Mephitinae. Nevertheless, not a part of the mustelid clade and sug- the 12S and 16S data constrained on their gested that they should be recognized as a phylogenetic hypothesis was only eight separate family. Our results, based on 12S steps longer than the most parsimonious and 16S data, are consistent with theirs, phylogeny (Fig. 1). The absence of taxa of based on cytochrome b. skunks in their analysis may be the primary Based on the above mentioned studies, difference between the 12S and 16S data one can conclude that the pinnipeds are part and the amino-acid-sequence data. Whereas of the arctoid carnivore radiation and are the 12S and 16S data suggested that the monophyletic, a result supported by other family is paraphyletic, the results of Miy- molecular studies (Lento et al., 1995; amoto and Goodman (1986) indicated the Sarich, 1969). However, the relationships Mustelidae to be monophyletic. among the major clades in the Caniformia In a DNA hybridization study by Wayne are unresolved. A strict consensus of the to- et al. (1989), the relationships of many of pologies derived from five of the studies the major clades in the Caniformia were discussed above resulted in a polytomy unresolved, with the exception of the Cani- for the Caniformia. The Lutrinae, Melinae dae, which was the sister taxon to the (minus Mydaus), Procyonidae, Canidae, arctoid group. Within the arctoid carni- and Mephitinae (including Mydaus) each vores, the Procyonidae, Ursidae, the pinni- formed separate branches. The pinnipeds peds, and part of the Mustelidae were part represented three distinct branches, and the of a polytomy. The subfamily Mephitinae Mustelinae also was separated into three was basal to the arctoid bush. The results different branches. A 50% majority rule of the present study support Wayne et al. consensus for these topologies revealed a (1989) in placing the skunks outside the polytomy consisting of five lineages, in- mustelids and also resolves the arctoid cluding pinnipeds, Canidae, Procyonidae, polytomy. Ursidae, and Mustelidae. Within the mus- Vrana et al. (1994) examined DNA se- telid clade, Mephitinae (including Mydaus) quence from partial sequences of the 12S was basal, and Melinae was the sister group 434 JOURNAL OF MAMMALOGY Vol. 78, No. 2 to an unresolved clade containing Lutrinae Phocidae as was found by Wyss and Flynn and Mustelinae. (1993). Finally, Canidae and Ursidae repre- To examine all of the existing evidence sented more basal caniform families. used to describe the families of caniform Total-evidence and missing data.-The carnivores, the 12S and 16S rRNA gene se- phylogenetic results from the total-evidence quences from the present study were com- analysis (Fig. 2), were similar to the results bined with character-state data from mo- of the 12S and 16S rRNA gene analysis lecular studies (Vrana et al., 1994) and mor- (Fig. 1). This is surprising given the lack phological studies (Bryant et al., 1993; of congruence among independent analyses Wozencraft, 1989; Wyss and Flynn, 1993) of morphological and molecular data. in a maximum-parsimony analysis. The Recently, de Queiroz (1993) summarized short-tailed shrew (Blarina) and the extinct the two general approaches used to evalu- order Creodonta were used as outgroups. A ate phylogenetic hypotheses derived from total of 62 specimens representing 44 taxa multiple datasets. One approach, which de- of carnivores, (including multiple individu- rives a consensus tree from independently als of a species and multiple species in a derived phylogenies, has the advantage of genus) and 2 outgroup taxa were analyzed. providing equal weights for each dataset Sequences from the 12S and 16S rRNA and eliminates any potential swamping of genes and gap codes, sequences from the small datasets by larger amounts of data cytochrome b gene, and morphological from another source. Nevertheless, this ap- characters from Wyss and Flynn (1993) and proach may not always provide the most- Bryant et al. (1993) were combined to pro- parsimonious solution for character change duce a dataset of 1,647 characters. A con- (Miyamoto, 1985) and will result in de- sensus of the 48 most-parsimonious trees creased resolution. With regard to the evalu- was generated (Fig. 2). ation of relationships among caniform car- Several interesting results were obtained nivores, a consensus approach results in a with this combined analysis. The two ma- total lack of resolution due to low taxo- jor groups of carnivores, Caniformia and nomic congruence among the various exist- Feliformia (Wozencraft, 1989; Wyss and ing hypotheses derived using morphological Flynn, 1993) formed monophyletic groups. and molecular data. As with the molecular data, the family Mus- A second approach, combining all data telidae was paraphyletic, with the subfam- into a single phylogenetic analysis, has been ily Mephitinae (plus Mydaus) being the sis- advocated by several authors (Barrett et al., ter taxon to a clade containing Procyonidae 1991; Kluge, 1989; Miyamoto, 1985). Phy- and the remaining Mustelidae. In contrast to logenies derived from a broad range of the molecular data, except for Ictonyx, the characters may overcome many of the bi- Mustelinae formed a monophyletic group ases associated with a particular set of char- that was the sister group to the monophy- acters that individually yield erroneous phy- letic Lutrinae. This clade and Ictonyx were logenies (Bull et al., 1993). Arguments for the sister group to the American and Euro- combining data prior to analysis include: 1) pean badgers (Melinae). The , different character classes may provide bet- Ailurus, and Procyonidae were sister taxa; ter levels of resolution at different nodes of the , , grouped with a tree (Hillis, 1987); 2) individual datasets Ursidae. The pinnipeds formed a mono- may contain only a weak phylogenetic sig- phyletic clade that was the sister group to nal and adding multiple datasets with low the clade containing the mustelids, procy- signal should be additive and override the onids, and skunks. The basal taxon of the noise; 3) phylogenetic analyses should ex- pinniped clade, the extinct Desmatophoci- plain all of the data simultaneously (Kluge, dae, was placed as the sister group to the 1989). May 1997 DRAGOOAND HONEYCUTT-SYSTEMATICSOF SKUNKS 435

Conepatus Mephitis* Spilogale Mephitinae Mydaus Melinae Aonyx* Lutra* Lutrinae EnhydraGulo Martes Mustela frenata* Mustelinae M. vison* M. putorius Ictonyx Meles Melinae Taxidea* Taxidiinae Canifoia Bassariscus* Caniformia Procyon* Procyonidae Potos* Ailurus Odobenus* Zalophus* Phoca* Pinnipeds Halichoerus Desmatophocid Ursus*+ Ailuropoda Ursidae Amphicyonidae Canis*+ Urocyon Canidae Leopardus pardalis * + HyaenidaeFeliformia Galerella Feliformia Suricata Nimravidae Viverravidae Miacidae Creodonta Blarina FIG. 2.-Strict consensus tree (tree length 3,534; CI = 0.346; RI = 0.591) of the Carnivora and outgroups derived from 48 equally parsimonious trees based on total evidence including 12S and 16S rRNA sequences, cytochrome b sequences, and morphological characters. Taxa followed by (*) are represented by multiple individuals and taxa followed by (+) are represented by multiple species. Terminal taxa denoted by (*) or (+) have been combined into a single branch for ease of presenta- tion. 436 JOURNAL OF MAMMALOGY Vol. 78, No. 2

When different datasets are congruent, it approach (Fig. 2), except that several nodes can be assumed that the resultant phylog- received less support as indicated by the eny approximates the true species phylog- low bootstrap values and decay indices. The eny (Kluge, 1989). Alternatively, different resultant phylogeny clearly demonstrates traits or sets of characters may not share a more resolution than seen by the consensus common evolutionary history in terms of analyses performed earlier. As suggested by the species phylogeny or differ with respect Kluge (1989) and Eernisse and Kluge to the amount of homoplasy associated with (1993), combining all of the characters, re- each type of data. In such cases, combining gardless of the degree to which they differ highly heterogeneous data may result in an in terms of homoplasy, results in a consis- inaccurate representation of relationships tent phylogeny that, for the most part, is among species. Bull et al. (1993) have ar- identical to the molecular analysis. gued that prior to combining datasets one A potential problem with combining sets should test for homogeneity among the dif- of characters is missing data. This is a com- ferent sets of characters. If individual data- mon occurrence when analyzing both extant sets demonstrate considerable heterogeneity and extinct taxa in a combined analysis that and produce highly incongruent phylog- includes morphological and molecular data. enies, then the data should not be combined. In such an analysis, the extinct taxa will They suggest that these character sets may have missing data associated with all the have different evolutionary histories, be molecular characters. According to Swof- evolving at different evolutionary rates, and ford (1993), only those characters that have have different levels of associated ho- non-missing values will affect the location moplasy. For example, individual gene of any taxon on the tree. Nevertheless, there trees, derived from independently evolving is an indication that missing data can be loci, may not reflect the phylogeny of spe- problematical (Platnick et al., 1991), espe- cies. Such a result is likely when compar- cially when taxa have large amounts of ing multiple nuclear-gene loci (Moore, missing data. 1995). Incongruence among phylogenies Wyss and Flynn's (1993) analysis of ex- derived from different mitochondrial genes tant and extinct carnivores produced a phy- may be a result of differences in levels of logeny in which the extinct Desmatophoci- constraint, amounts of homoplasy, and rates dae was the sister group to Phocidae. In the of nucleotide substitution, because the mi- total-evidence analysis (Fig. 2), the Des- tochondrial genome is evolving as a single matophocidae was at the base of the pin- locus (Miyamoto et al., 1994). Morphologi- niped radiation. Only 38 morphological cal data also can show different degrees of characters were needed to place the des- homoplasy associated with particular suites matophocid on the tree, because of a num- of characters as a result of primarily paral- ber of synapomorphies with the Phocidae. lel and convergent evolution. However, 1,500 characters were used to To avoid some of the potential problems place Phocidae on the same tree (Fig. 2). with combining the morphological and mo- Due to the close relationship among the lecular data, an analysis was conducted us- other pinnipeds, the larger number of syn- ing mitochondrial gene sequences (12S, apomorphies in the molecular data swamped 16S, and cytochrome b) to examine rela- the morphological data so that the phocids tionships among 31 taxa of carnivores ob- became the sister group to the otariid and tained from the present study and Vrana et odobenid clade rather than to the desmato- al. (1994). A single most-parsimonious tree phocids. was obtained (Fig. 3). The result is congru- In contrast to the apparent problems as- ent with both the analysis of the 12S and sociated with taxa represented only by mor- 16S genes (Fig. 1) and the total-evidence phological characters in the total-evidence May 1997 DRAGOOAND HONEYCUTT-SYSTEMATICSOF SKUNKS 437

94/13 Conepatus 100/16 Mephitis Mephitinae 88/9 Spilogale Mydaus Melinae 60/3 Aonyx <50/1 Lutra Lutrinae <50/1 Enhydra <50/2 93/6 Mustelavison <50/2<50/1 M. frenata 90/19.. M. putorius Mustelinae 95/10 -Ictonyx80/5 Gulo <50/1 Martes Meles Melinae <50/1 - 95/6 Taxidea Taxidiinae <50/2 100/15 E:Bassariscus <50/2 Procyonidae PotosProcyons <50/2 Ailurus 83/6 Odobenus -- 80/2 L Zalophus 1 Phoca Pinnipeds 100/14 Halichoerus <50/1 Ursus <50/1 Ailuropoda Ursidae Canis 97/6 Urocyon Canidae 99/9 100/15 Leopardus 100/15Panthera Feliformia Galerella 97/8 Suricata

FIG. 3.-Phylogenetic relationships among 31 taxa of carnivores based on parsimony analysis (tree length 2,429; CI = 0.363; RI = 0.437) of the 12S, 16S, and cytochrome b gene sequences. Some taxa have partial data. Numbers represent bootstrap values/decay indices. analysis, there appears to be less sensitivity phitis and Spilogale. This clade was joined to missing data associated with the genes to the clade containing the paraphyletic examined in this study. To examine the af- Mustelinae and the monophyletic Lutrinae. fects of missing data, a parsimony analysis A second analysis (Fig. 4B) was performed (Fig. 4) was performed using the 12S, 16S, on the datasets where the cytochrome b data and cytochrome b datasets for the three (ca. 25% of the data) for Mephitis, Mustela North American genera of skunks, Mustela vison, Ictonyx, and Lutra were scored as frenata and M. vison, the three genera of ot- missing. These taxa were chosen because of ters, Ictonyx, and Canidae. The tree topol- the varying amounts of support for their ogy (Fig. 4A) is similar to that of the 12S placement on the tree shown in Fig. 1. The and 16S analysis (Fig. 1). Conepatus was single most-parsimonious tree had lower the sister group to the clade formed by Me- bootstrap values and support indices, com- 438 JOURNALOF MAMMALOGY Vol. 78, No. 2

100/36 Conepatus 100/29 Conepatus Mephitis Mephitis* 97/10 Spilogale 89/5 Spilogale 100/36 Aonyx 100/44 Aonyx <50/1o Lutra 69/3 Lutra* <501 76/3 Enhydra Enhydra Ictonyx Ictonyx* 100/29 99/15 Mustelavison 100/20 100/18 Mustela vison* M. frenata M. frenata Canis Canis A B

100/17 Conepatus 99/7 Conepatus E Mephitis* Mephitis* 65/3 Spilogale Spilogale 100/15 Aonyx 99/6 Aonyx 5/ 56/3 Lutra* 76/1 Lutra* Ictonyx* 7Enhydra Enhydra Ictonyx* 100/12 _ 52/3 Mustela vison* 97/3 - Mustela vison* M. frenata M. frenata Canis Canis C D FIG. 4.-The most-parsimonious trees obtained from analyses of datasets containing various amounts of missing data for: A) complete dataset consisting of 12S, 16S, and cytochrome b sequences; B) cytochrome b sequence scored as missing (*) for Mephitis, Lutra, Ictonyx, and Mustela vison; C) cytochrome b and 16S sequence scored as missing for the same taxa; D) all, but ca. 200 base pairs, of the 12S sequence scored as missing. Numbers represent bootstrap values/decay indices. pared to the first analysis (Fig. 4A), for the The above results suggest that the sever- Mephitis-Spilogale clade, yet higher sup- ity of the effects from missing data may re- port for the placement of Ictonyx and En- late more to the relationship of the taxa in- hydra. Next, the 16S data was scored as volved and the phylogenetic signal in the missing for the same taxa (Fig. 4C). A data than to the amount of missing data. For single most-parsimonious tree was found example, if several taxa from a strongly and the bootstrap values and support indi- supported monophyletic group, such as ces were lower still, compared to the first Mustela or the river otters, are included in two analyses (Figs. 4A and 4B), for the an analysis, missing data from a single Mephitis-Spilogale clade, as well as the taxon may not affect its placement. This Mustela clade; the placement of Ictonyx and suggests that clades supported by a large Enhydra also changed positions on the tree, number of characters from one dataset can resulting in a paraphyletic Lutrinae. When withstand relatively high degrees of miss- data were scored as missing for all but ca. ing data and still maintain the same phylo- 200 base pairs of the 12S gene for the four genetic position on the tree. A second ex- taxa, eight equally parsimonious trees were ample can be seen for members of the Fe- found and the resulting topology (Fig. 4D) liformia examined in this study. The mono- consisted of a polytomy for the skunks and phyly of this group was strongly supported a polytomy for M. vison, Ictonyx, Enhydra, by all analyses as suggested by the morpho- and the other two otters. logical characters reviewed by Wyss and May 1997 DRAGOO AND HONEYCUTT--SYSTEMATICS OF SKUNKS 439

Flynn (1993), despite the fact that some fe- terpretation of some morphological charac- liform taxa had large amounts of missing ter state changes may be compromised data. (Anderson, 1989; Bryant et al., 1993). Missing data may be more problematical Many of the morphological studies that when a taxon has no close relatives, such examined the higher-level relationships of as monotypic taxa. For example, when more carnivores assumed not only that the fam- and more data were recorded as missing for ily Mustelidae was monophyletic, but that Ictonyx, the placement of this taxon became the subfamilies were monophyletic (Wozen- obscure. The red panda, Ailurus, also was craft, 1989; Wyss and Flynn, 1993). There difficult to place, with trees only one or two are few synapomorphies to support the steps longer resulting in entirely different monophyly of the Mustelidae, including topologies to that shown in Fig. 3. In gen- enlarged scent glands, loss of the carnassial eral, the addition of Ailurus lowered boot- notch on the upper fourth premolar, and strap values and decay indices for well- the loss of the upper second molar (Bryant supported clades. Vrana et al. (1994) indi- et al., 1993; Martin, 1989; Wozencraft, cated that Ailurus may be an interesting 1989). These characters, however, may be taxon to examine in terms of evolution of homoplasious. For example, Wozencraft arctoid carnivores. The red panda has been (1989) suggested that the loss of the notch considered a procyonid (Nowak, 1991), a on the camassial occurred more than once bear (Wozencraft, 1989), at the base of the in independent lineages of carnivores. All radiation of and pinnipeds (Vrana et carnivores have scent glands; they are en- al., 1994), at the base of a clade comprised larged in mustelids, and greatly enlarged in of ursids and procyonids (Zhang and Ryder, skunks. The association of a nipple with the 1993), and as a separate family (Wozen- scent gland, rather than a duct as in mustel- craft, 1989). In an analysis of mitochondrial ids, suggests that the scent gland in skunks and nuclear-molecular data, Slattery and is apomorphic. Other characteristics shared O'Brien (1995) found Ailurus to be a mem- between skunks and mustelids may repre- ber of the procyonid clade. Ledje and Ar- sent symplesiomorphies (e.g., type of audi- nason (1996) also used mitochondrial mo- tory bulla-Hunt, 1974). lecular data, but suggested that Ailurus In a recent study of the Mustelidae, Bry- should be relegated to a separate family. Al- ant et al. (1993) used 46 morphological though the red panda was the sister taxon characters and 23 extant genera of the pre- to Potos in the Procyonidae clade, in analy- sumed monophyletic Mustelidae to test the sis of the total evidence (Fig. 2) and the subfamilial relationships based on Simp- analysis of the molecular data (Fig. 3) in the son's (1945) classification. They suggested present study the bootstrap values and sup- a sister-group relationship between the port indices were low. monophyletic subfamilies Lutrinae and Me- Monophyly of the Mustelidae.-Most phitinae. The subfamily Mustelinae was morphological studies (Bryant et al., 1993; found to be paraphyletic, which agreed with Wozencraft, 1989; Wyss and Flynn, 1993) the findings of earlier studies (Anderson, place skunks as sister to the mustelid sub- 1989; Pocock, 1921; Radinsky, 1973). The family Lutrinae, whereas molecular data Melinae was polyphyletic; three of the five (Amrnasonand Widegren, 1986; Ledje and genera (Meles, Arctonyx, and Mydaus) were Arnason, 1996; Vrana et al., 1994; Wayne members of the clade containing Lutrinae et al., 1989), including the present study, and Mephitinae. show the skunks to be outside the mustel- Although morphological studies (Bryant ids. That the family Mustelidae, as currently et al., 1993; Wyss and Flynn, 1993) have recognized, is not monophyletic, as shown suggested a monophyletic Mustelidae, this by the molecular data, suggests that the in- is not strongly supported by their data. Only 440 JOURNAL OF MAMMALOGY Vol. 78, No. 2 three additional steps are required to col- the sister group to the other mustelids. lapse the mustelid clade derived from mor- Although the remaining mustelid taxa form phology, whereas the 12S and 16S data in a monophyletic group, the subfamily Mus- the present study (Fig. 1) require an addi- telinae is paraphyletic in relation to the tional 11 steps to make the Mustelidae Lutrinae. monophyletic and 15 steps to place the Me- The results of the present study, based on phitinae sister to the Lutrinae. The weak molecular data and total-evidence data, sup- support for a monophyletic Mustelidae port the paraphyly of the Mustelidae. How based on morphology may be due to ple- can the patterns of morphological evolution siomorphic or homoplastic characters unit- in skunks and mustelids be explained with ing the skunks and the mustelids. Bryant et respect to our results? Part of the answer al. (1993) reported that two of the shared may be obtained by considering existing in- character states were plesiomorphic and formation from the fossil record. The earli- the polarity of two additional characters est Carnivora, which first appeared during was uncertain. Of the five characters the late Paleocene-early Eocene, were shared between Mephitinae and Lutrinae, small, arboreal, viverrid-like (or weasel- four were based on tooth morphology, and like) forms belonging to the extinct Viver- the fifth (auditory bullae) was plesiomor- ravidae and Miacidae (Martin, 1989). Ac- phic. cording to Anderson (1989) and Martin Other data also suggest that the skunks (1989), during the late Eocene and early may not be a member of the Mustelidae. these forms gave rise to the Cani- Based on chromosomal evidence (Wurster formia (Mustelidae, Canidae, Procyonidae, and Benirschke, 1968), the Mustelidae is at and Ursidae) and the Feliformia (Felidae, the base of the Caniformia radiation. Dip- Viverridae, and Hyaenidae). loid and fundamental numbers are variable The early mustelid-like forms that ap- within Mustelidae, compared to the rest of peared in the late Eocene cannot be traced the Carnivora, and they noted that the to the modem mustelids (Kurten and Ander- skunks had a unique karyotype relative to son, 1980), which first appeared in the Old the mustelids. Arnason and Widegren World during the mid-Miocene (Kurten and (1986) examined the relationship of the pin- Anderson, 1980; Martin, 1989). Miomephi- nipeds using DNA hybridization of highly tis, one of the first recognizable skunks repetitive DNA components and found that (Anderson, 1989; Kurten and Anderson, the pinnipeds were monophyletic and more 1980), retained many of the plesiomorphic closely related to the Mustelidae (not in- traits associated with the Eocene mustelid- cluding the Mephitinae) than to any other like carnivores. Radinsky (1973) suggested carnivore family. The pinnipeds and the that most of the earliest known modem procyonids were more genetically similar to mustelids (20-25 X 107 years ago) exhibit the Mustelidae than was the subfamily Me- several advanced cranial features compared phitinae. These observations are congruent with modem skunks and the stink badger. with the 12S and 16S rRNA gene data pre- In addition, Wayne et al. (1989) suggested sented here. Parsimony analyses of the 12S that the origin of the skunk lineage occurred and 16S data in this study suggest that the in the Oligocene (ca. 40 x 107 years ago), family Mustelidae is not monophyletic, which is prior to the appearance in the fos- with the skunks being more divergent from sil record of the musteloid stem group that the rest of the mustelids than the family gave rise to the Procyonidae. Because mus- Procyonidae. Our data also provide sup- telids, in general, and skunks in particular, port for the Melinae being diphyletic with retain many plesiomorphic traits (Anderson, Mydaus being the sister group to skunks, 1989), it is difficult to find synapomorphies and the American and European badgers as for a monophyletic Mustelidae as well as May 1997 DRAGOO AND HONEYCUTT-SYSTEMATICS OF SKUNKS 441 other groups of mustelid-like carnivores. As LITERATURE CITED a the Mustelidae the Mus- result, (especially ANDERSON, E. 1989. The phylogeny of mustelids and telinae) has been a catch-all category for the systematics of ferrets. Pp. 10-20, in Conserva- of the undifferentiated taxa as tion biology and the biology of the black-footed fer- many early, ret S. E. T. M. A. and S. well as of doubtful affin- (U. Seal, Thorne, Bogan, divergent genera H. Anderson, eds.). Yale University Press, New Ha- ity (Anderson, 1989). ven, Connecticut, 209 pp. Taxonomic considerations.-Although ARNASON,U., ANDE. JOHNSSON.1992. The complete mitochondrial DNA sequence of the harbor seal, the relationships among some lineages of Phoca vitulina. Journal of Molecular Evolution, carnivores are uncertain, the results of the 34:493-505. as well as other recent mo- ARNASON, U., AND B. WIDEGREN. 1986. Pinniped present study, molecular lecular studies and phylogeny enlightened by hybridizations (Amrnason Widegren, using highly repetitive DNA. Molecular Biology and 1986; Ledje and Amrnason,1996; Vrana et Evolution, 3:356-365. al., 1994; Wayne et al., 1989) suggest a need BARRETT, M., M. J. DONOGHUE, AND E. SOBER. 1991. Against consensus. Systematic Zoology, for some changes in the classification of 40:486-493. caniform carnivores. If the Mustelidae are BONAPARTE,C.-L. J. L. 1845. Catalogo methodico to be treated as a then dei mammiferi Europei. L. di Giacomo Pirola, Mi- monophyletic group, 36 the skunks need to be reclassified. The so- lano, Italy, pp. BREMER,K. 1988. The limits of amino acid sequence lution is to elevate the subfamily, Mephiti- data in Angiosperm phylogenetic reconstruction. nae (Bonaparte, 1845), to a distinct Evolution, 42:795-803. family, H. A. P. AND W. D. FITCH. the This includes the ex- BRYANT, N., RUSSELL, Mephitidae. family 1993. Phylogenetic relationships within the extant tant genera Mephitis (the striped and Mustelidae (Carnivora): appraisal of the cladistic hooded skunks), Conepatus (the hog-nosed status of the Simpsonian subfamilies. Zoological Journal of the Linnean Society, 108:301-334. skunks), Spilogale (the spotted skunks), and BULL, J. J., P. HUELSENBECK,C. W. CUNNINGHAM,D. Mydaus (the stink badger). L. SWOFFORD,AND P. J. WADDELL.1993. Partition- ing and combining data in phylogenetic analysis. Systematic Biology, 42:384-397. ACKNOWLEDGMENTS COUES,E. 1877. Fur-bearing : a monograph of North American Mustelidae. Miscellaneous Pub- We thank the following people who have con- lications of the Department of the Interior, United tributed DNA or tissue samples from various States Geological Survey, Washington, D.C., 8:1- taxa of carnivores; J. Cooke, J. W. Bickham, 348. DE A. 1993. For consensus P. B. Vrana, L. A. Ruedas, R. K. J. C. QUEIROZ, (sometimes). Wayne, Systematic Biology, 42:368-372. T. L. D. R. J. M. Patton, Yates, Schlitter, Baker, DRAGOO, J. W., R. D. BRADLEY, R. L. HONEYCUTT, Tewes, and K. Rautenbach. For help in the labo- AND J. W. TEMPLETON. 1993. Phylogenetic relation- ratory, technical assistance, data analyses, and fi- ships among the skunks: a molecular perspective. Journal of Mammalian Evolution, 1:225-267. nancial support, we thank R. D. Bradley, J. Derr, EERNISSE,D. J., ANDA. G. KLUGE. 1993. Taxonomic M. T. P. L. A. J. Smolen, Lee, Lee, Ruedas, congruence versus total evidence, and amniote phy- Templeton, K. Rittenhouse, R. Adkins, M. Ned- logeny inferred from fossils, molecules, and mor- bal, A. Walton, J. Rice, C. Walker, J. Patton, and phology. Molecular Biology and Evolution, 10:1170-1195. J. Wickliffe. Finally, we thank G. A. Dragoo, T. EISENBERG, J. F. 1989. An introduction to the Car- Taylor, S. Davis, D. Schmidly, J. Bickham, J. nivora. Pp. 1-9, in Carnivore behavior, ecology, and McEachran, and two anonymous reviewers for evolution (J. L. Gittleman, ed.). Comstock Publish- reviewing earlier drafts of this manuscript. This ing Associates, Ithaca, New York, 620 pp. research was supported by grants from the Na- FARRIS, J. S. 1989. The retention index and the rescaled indes. 5:417-419. tional Science Foundation to J. W. Bickham, R. consistency Cladistics, FELSENSTIEN,J. 1985. Confidence limits on L. and J. W. phylog- Honeycutt, Dragoo (DEB-9310553), enies: an approach using the bootstrap. Evolution, and to R. L. Honeycutt (DEB-9208022). Data 39:783-791. analyses were performed at the Center for Bio- FLYNN,J. J., N. A. NEFF, AND R. H. TEDFORD.1988. systematics and Biodiversity, a facility funded in Phylogeny of the Carnivora. Pp. 73-116, in The phy- logeny and classification of the tetrapods (M. J. Ben- part by the National Science Foundation (DIR- ton, ed.). Oxford University Press, New York, 2:1- 8907006). This paper represents contribution 58 329. of the center. HIGGINS, D. G., ANDP. M. SHARP. 1989. Fast and sen- 442 JOURNAL OF MAMMALOGY Vol. 78, No. 2

sitive multiple sequence alignments on a microcom- L. L. JANECEK, AND R. L. HONEYCUTT. 1994. puter. Cagios, 5:151-153. A congruence test of reliability using linked mito- HILLIS,D. M. 1987. Molecular versus morphological chondrial DNA sequences. Systematic Biology, approaches to systematics. Annual Review of Ecol- 43:236-249. ogy and Systematics, 18:23-42. MOORE, W. S. 1995. Inferring phylogenies from HILLIS, D. M., A. LARSON, S. K. DAVIS, AND E. A. ZIM- mtDNA variation: mitochondrial-gene trees versus MER. 1990. Nucleic acids III: sequencing. Pp. 318- nuclear-gene trees. Evolution, 49:718-726. 370, in Molecular systematics (D. M. Hillis and C. NEI, M. 1991. Relative efficiencies of different tree Moritz, eds.). Sinauer Associates, Inc., Sunderland, making methods for molecular data. Pp. 90-128, in Massachusetts, 588 pp. Phylogenetic analysis of DNA sequences (M. M. HUNT, R. M., JR. 1974. The auditory bulla in Car- Miyamoto and J. L. Cracraft, eds.). Oxford Univer- nivora: an anatomical basis for reappraisal of carni- sity Press, New York, 358 pp. vore evolution. Journal of Morphology, 143:21-76. NOWAK, R. M. 1991. Walker's mammals of the JANCZEWSKI, D. N., N. YUHKI, D. A. GILBERT, G. T. world. Fifth ed. The Johns Hopkins University Press, JEFFERSON, AND S. J. O'BRIEN. 1992. Molecular Baltimore, 2:643-1629. phylogenetic inference from saber-tooth fossils O'BRIEN, S. J., J. S. MARTENSON,M. A. EICHEL- of Rancho La Brea. Proceedings of the National BERGER, E. T. THORNE, AND F. WRIGHT. 1989. Academy of Sciences, 89:9769-9773. Genetic variation and molecular systematics of the KLUGE,A. G. 1989. A concern for evidence and a black-footed ferret. Pp. 21-33, in Conservation bi- phylogenetic hypothesis of relationships among Epi- ology and the biology of the black-footed ferret crates (Boidae, Serpentes). Systemic Zoology, 38:7- (U. S. Seal, E. T. Thomrne,M. A. Bogan, and S. H. 25. Anderson, eds.). Yale University Press, New Haven, KLUGE, A. G., AND J. S. FARRIS. 1969. Quantitative Connecticut, 209 pp. phylogenetics and the evolution of anurans. System- OLMSTEAD, R. G., AND J. A. SWEERE. 1994. Combin- atic Zoology, 18:1-32. ing data in phylogenetic systematics: an empirical KUMAR,S., K. TAMURA,AND M. NEI. 1993. MEGA: approach using three molecular data sets in the So- molecular evolutionary genetics analysis, version lanaceae. Systematic Biology, 43:467-481. 1.01. The Pennsylvania State University, University PETTER, G. 1971. Origine, phylogenie et system- Park, 130 pp. atique des blaireaux. Mammalia, 35:567-597. KURTI'N,B., AND E. ANDERSON. 1980. Pleisticene PLATNICK, N. I., C. E. GRISWOLD, AND J. A. CODDING- mammals of North America. Columbia University TON. 1991. On missing entries in cladistic analy- Press, Irvington, New York, 442 pp. sis. Cladistics, 7:337-343. LEDJE, C., AND U. ARNASON. 1996. Phylogenetic POCOCK,R. I. 1921. On the external characters and analyses of complete cytochrome b genes of the or- classification of the Mustelidae. Proceedings of the der Carnivora with particular emphasis on the Cani- Zoological Society of London, 1921:803-837. formia. Journal of Molecular Evolution, 42:135- RADINSKY,L. 1973. Are stink badgers skunks? Im- 144. plications of neuroanatomy for mustelid phylogeny. LEDOUX, R. G., AND A. J. KENYON. 1975. Protides Journal of Mammalogy, 54:585-593. of the Mustelidae-II. Immunologic relatedness. SAIKI, R. K., T. L. BUGAWAN, G. T. HORN,K. B. MUL- Comparative Biochemistry and Physiology, A. Com- LIS, AND H. A. ERLICH. 1986. Analysis of enzymati- parative Physiology, 51:213-217. cally amplified beta-globin and HLA-DQalpha DNA LENTO, G. M., R. E. HICKSON, G. K. CHAMBERS, AND with allele-specific oligonucleotide probes. Nature, D. PENNY. 1995. Use of spectral analysis to test hy- 324:163-166. potheses on the origin of pinnipeds. Molecular Bi- SAIKI, R. K., ET AL. 1988. Primer-directed enzymatic ology and Evolution, 12:28-52. amplification of DNA with a thermostable DNA LOCKHART, P. J., M. A. STEEL, M. D. HENDY, AND P. polymerase. Science, 239:487-491. DAVID. 1994. Recovering evolutionary trees under SARICH,V. M. 1969. Pinniped phylogeny. Systematic a more realistic model of sequence evolution. Mo- Zoology, 18:416-422. lecular Biology and Evolution, 11:605-612. SETON, E. T. 1926. Lives of game animals. Double- MARTIN, L. D. 1989. Fossil history of the terrestrial day, Doran & Company, Inc., Garden City, New Carnivora. Pp. 536-568, in Carnivore behavior, ecol- York, 746 pp. ogy, and evolution (J. L. Gittleman, ed.). Comstock SIMPSON, G. G. 1945. The principles of classification Publishing Associates, Ithaca, New York, 620 pp. and a classification of mammals. Bulletin of the MIYAMOTO, M. M. 1985. Consensus cladograms and American Museum of Natural History, 85:1-350. general classifications. Cladistics, 1:186-189. SLATTERY,J. P., AND S. J. O'BRIEN. 1995. Molecular MIYAMOTO, M. M., AND M. GOODMAN. 1986. Biomo- phylogeny of the red panda (Ailurus fulgens). Jour- lecular systematics of eutherian mammals: phyloge- nal of Heredity, 86:413-422. netic patterns and classification. Systematic Zool- SPRINGER, M. S., L. J. HOLLAR, AND A. BURK. 1995. ogy, 35:230-240. Compensatory substitutions and the evolution of the MIYAMOTO, M. M., F KRAUS, AND O. A. RYDER. mitochondrial 12S rRNA gene in mammals. Molecu- 1990. Phylogeny and evolution of antlered deer de- lar Biology and Evolution, 12:1138-1150. termined from mitochondrial DNA sequences. Pro- SWOFFORD,D. L. 1993. PAUP: phylogenetic analysis ceedings of the National Academy of Sciences, using parsimony, version 3.1.1. Computer program 87:6127-6131. distributed by the Illinois Natural History Survey, MIYAMOTO, M. M., M. W. ALLARD, R. M. ADKINS, Champaign, Illinois, 264 pp. May 1997 DRAGOO AND HONEYCUTT-SYSTEMATICS OF SKUNKS 443

TAJIMA, F., AND M. NEI. 1984. Estimation of evolu- . 1993. Carnivora. Pp. 279-348, in tionary distance between nucleotide sequences. Mo- species of the world: a taxonomic and geographic lecular Biology and Evolution, 1:269-285. reference (D. E. Wilson and D. M. Reeder, eds.). VRANA, P. B., M. C. MILINKOVITCH,J. R. POWELL, AND Smithsonian Institution Press, Washington, D. C., W. C. WHEELER. 1994. Higher level relationships 1206 pp. of the arctoid Carnivora based on sequence data and WURSTER, D. H., AND K. BENIRSCHKE.1968. Com- "total evidence." Molecular Phylogenetics and Evo- parative cytogenetic studies in the order Carnivora. lution, 3:47-58. Chromosoma (Berlin), 24:336-382. WAYNE, R. K., R. E. BENVENISTE, D. N. JANCZEWSKI, WYss, A. R., AND J. J. FLYNN. 1993. A phylogenetic AND S. J. O'BRIEN. 1989. Molecular and biochemi- analysis and definition of the Carnivora. Pp. 32-52, cal evolution of the Carnivora. Pp. 465-494, in Car- in Mammal phylogeny: placentals (F. S. Szalay, nivore behavior, ecology, and evolution (J. L. Gittle- M. J. Novacek, and M. C. McKenna, eds.). Springer- man, ed.). Comstock Publishing Associates, Ithaca, Verlag, New York, 321 pp. New York, 620 pp. ZHANG, Y., AND O. A. RYDER. 1993. Mitochondrial WILSON, D. E., AND D. M. REEDER (EDS). 1993. DNA sequence evolution in the Arctoidea. Proceed- Mammal species of the world: a taxonomic and geo- ings of the National Academy of Science, 90:9557- graphic reference. Second ed. Smithsonian Institu- 9561. tion Press, Washington, D.C., 1206 pp. WOZENCRAFT,W. C. 1989. The phylogeny of the Re- Submitted 9 December 1995. Accepted 16 July 1996. cent Carnivora. Pp. 495-535, in Carnivore behavior, ecology, and evolution (J. L. Gittleman, ed.). Com- Associate Editor was Janet K. Braun. stock Publishing Associates, Ithaca, New York, 620 pp.