THE approximately 33 species in the genus Fundulus are widely distributed in freshwater, brackish, and coastal marine environments of North America, coastal areas of the Yucatan Peninsula, and Bermuda. The studfish clade, subgenus Xenisma, includes five extant species, Fundulus bifax, Fundulus catenatus, Fundulus julisia, Fundulus rathbuni, and Fundulus stellifer, and one recently extinct species, Fundulus albolineatus. These fishes are typically found in clear freshwater habitats of the central and eastern United States. Unlike the other subgenera of Fundulus, all species in Xenisma naturally occur only in low salinity environments. Studfishes are among the largest fundulids and nuptial males of most Xenisma species are very brightly colored. The species are allopatric except that F. julisia and F. catenatus are often syntopic (Etnier and Starnes, 1993), and F. bifax and F. stellifer both occur in the Lower Coosa drainage (Cashner et al., 1988). Most populations are confined to highland stream systems.

Phylogenetic relationships within Xenisma have never been comprehensively addressed using morphological data. Williams and Etnier (1982) described F. julisia, diagnosed the subgenus, and suggested that F. julisia is more closely related to F. catenatus and F. stellifer than to F. rathbuni. Wiley (1986) in his study of relationships of Fundulus addressed relationships within Xenisma but included only characters that were previously mentioned in the literature. Wiley (1986:125) stated that his discussion of character state support for relationships within subgenus Xenisma did not constitute a phylogenetic analysis and that “much work has to be done.” Wiley's (1986) morphological data supported the monophyly of Xenisma and F. catenatus plus F. stellifer (populations now recognized as F. bifax were considered part of F. stellifer). Morphological variation within the subgenus Xenisma has been described (Brown, 1957; Thomerson, 1969; Williams and Etnier, 1982), but the morphology of Xenisma has not yet been subjected to a phylogenetic analysis.

A series of studies using allozyme data (Rogers and Cashner, 1987; Cashner et al., 1988; Grady et al., 1990) first described F. bifax from populations previously referred to F. stellifer and suggested that F. rathbuni was more closely related to the clade containing F. bifax, F. catenatus, and F. stellifer, than to F. julisia. Allozyme (Rogers and Cashner 1987, Cashner et al. 1988) and RFLP data (Strange and Burr 1997) support F. bifax and F. catenatus as sisters, and this clade as sister to F. stellifer. There are various DNA sequence studies that included some species in subgenus Xenisma (Bernardi and Powers, 1995; Bernardi, 1997). However, none of these DNA sequence studies include all extant Xenisma species.

Although Wiley's (1986) study has been the only hypothesis of relationships among all Xenisma based on morphological data, this study was not a phylogenetic analysis. We were interested in how a more comprehensive examination of morphological data would contribute to knowledge of relationships within Xenisma. In this study, we used osteological and external morphological characteristics, to assess phylogenetic relationships within the subgenus Xenisma. In addition, the primarily allopatric and highland distribution of Xenisma species allowed clear evaluation of the biogeographic history of the group.

Materials and Methods

Specimens examined were whole alcohol preserved or cleared and stained for bone and cartilage following Dingerkus and Uhler (1977). See the Materials Examined section for sample sizes. Counts and measures follow methods outlined in Hubbs and Lagler (1947). Dissection of cleared-and-stained specimens followed Weitzman (1974), except that the branchial basket was removed prior to removal of the suspensorium. Dissection for examination of ventral coelomic viscera involved a right parasaggital cut through the body wall into the coelom from the anus around the right side of the pelvic girdle and to the pectoral girdle.

All specimens were examined using a Leica stereomicroscope for potentially phylogenetically informative variation among included members of the subgenera Xenisma and Fontinus. Specimens of the extinct species F. albolineatus were not available for direct osteological or visceral examination, and some data were unavailable from external examination of specimens or from Williams and Etnier (1982). Digital x-rays showing ventral and left lateral views of the lectotype (USNM 125055) and a paralectotype (USNM 225996) were examined. Because some data were available, we included this species in the phylogenetic analysis. Character states that were not identifiable were coded as unknown (?).

We obtained color pattern data for all species from photographs and descriptions of color pattern in Thomerson (1969), Williams and Etnier (1982), Cashner et al. (1988), Robison and Buchanan (1988), Page and Burr (1991), Etnier and Starnes (1993), Jenkins and Burkhead (1993), Mettee et al. (1996), Wildekamp (1996), Pflieger (1997), and Ross (2001) and from examination of ethanol preserved specimens and field observations of F. catenatus. Institutional abbreviations are as listed in Leviton et al. (1985).

Fundulus chrysotus (subgenus Zygonectes), Fundulus heteroclitus (subgenus Fundulus), Fundulus diaphanus (subgenus Fontinus), Fundulus seminolis (subgenus Fontinus), and Fundulus zebrinus (subgenus Plancterus) were included as outgroups based on the family level phylogenies of Parenti (1981) and Wiley (1986). We rooted with F. zebrinus because it has been recognized as basal within the Fundulidae (Parenti, 1981), the genus Fundulus (Bernardi and Powers, 1995), or within a clade of unresolved relationships composed of F. zebrinus, all other Fundulus, and Lucania (Wiley, 1986).

Transformation series (TS) are grouped into traditional anatomical units and assigned numbers. The character state for F. zebrinus was designated 0. Each account includes a description of each character state and a short discussion of the historic usage of each transformation series if applicable. All transformation series were unordered.

Maximum parsimony analyses used the exhaustive search option in PAUP 3.1.1 (D. L. Swofford, Illinois Natural History Survey, 1993, unpubl.). Decay indices (Bremer, 1988) were determined using TreeRot (M. D. Sorenson, University of Michigan, 1996, unpubl.). A Templeton's test (Templeton, 1983) was not run to test statistically the significance of the difference between the phylogeny and the phylogeny obtained from allozyme data (Rogers and Cashner, 1987; Cashner et al., 1988; Grady et al., 1990). The number of character states (4) that varied between the two topologies was low and the power of a Templeton's test to statistically reject a hypothesis was correspondingly low.

Results

No gross characteristics of the coelomic viscera were discretely variable and useful for this analysis. Thomerson (1969) described the molariform pharyngeal teeth and enlarged tooth-bearing bones of F. stellifer and noted the absence of these pharyngeal characteristics in F. catenatus. This putatively unique trophic morphology suggests that studfishes' diets vary and might be correlated with differing alimentary tract morphology. However, there were no gross morphological differences in digestive tract morphology among Xenisma examined. All species had a short digestive tract and a simple intestine with a single U-shaped bend (Fig. 1A).

Twenty-one transformation series were coded for phylogenetic analysis. See Appendix 1 for descriptions of transformation series. See Table 1 for distribution of character states among taxa. Six equally most parsimonious cladograms were found (length = 40, CI = 0.66, RI = 0.74, RC = 0.48; Fig. 2). Incongruities among the shortest trees involved relationships among Fundulus bifax, F. catenatus, and F. stellifer and among the outgroups. However, all shortest trees include a Xenisma + Fontinus clade. Monophyly of F. bifax, F. catenatus, and F. stellifer, a sister group relationship between F. albolineatus and F. julisia, a clade composed of all Xenisma except F. rathbuni, and monophyly of the subgenus Xenisma were supported by all most parsimonious trees.

Discussion

Our morphological data support many of the previously proposed relationships among Xenisma species. Monophyly of Xenisma (Williams and Etnier, 1982; Wiley, 1986) is supported by the unique and unreversed character state of the absence of barring in nuptial males (TS 13–1). Three other homoplastic character states diagnose this node (TS 11–0, 14–1, 18–1). However, this node has a decay index of one because one of the seven trees at 41 steps includes the two Fontinus species in this analysis as sister to a clade composed of all Xenisma species except F. rathbuni, which is sister to the Fontinus + other Xenisma clade. Previously proposed relationships among species of Xenisma corroborated by this analysis are a sister relationship between F. albolineatus and F. julisia (Williams and Etnier, 1982), and monophyly of the studfishes sensu strictu (F. bifax, F. catenatus, and F. stellifer; Wiley, 1986, Rogers and Cashner, 1987). Both of these relationships have a decay index of greater than one.

The monophyly of F. bifax, F. catenatus, and F. stellifer was well supported by the morphological data with a decay index of three. Although well diagnosed as a group, morphological variation among these three species was limited to three transformation series, which did not suggest a single most parsimonious resolution. The low level of morphological variation is not surprising since F. bifax was not recognized as distinct from F. stellifer until relatively recently (Cashner et al., 1988) when allozyme data highlighted the distinctness of these populations. Three of the six trees F. bifax and F. catenatus as sister taxa, which is supported by the arrangement of lateral spots into horizontal lines (TS 15–1). This also occurs homoplastically in F. albolineatus. The other three shortest trees support a sister relationship between F. bifax and F. stellifer, which is supported by a loss of a yellow band in the caudal fin of nuptial males (TS 18–0). The former relationship is supported by both allozyme and restriction analysis (Rogers and Cashner, 1987; Cashner et al., 1992; Strange and Burr, 1997). Therefore, the hypothesis of relationships among these species supported by consideration of most data when considering all published data form a sister-group relationship between F. bifax and F. catenatus.

The morphological data suggest a novel relationship of F. rathbuni as sister to all other Xenisma. This relationship is supported by the shared possession by all members of Xenisma except F. rathbuni of a long ascending process of the prootic that does not fuse to the alisphenoid (TS 4–2), teeth on the second pharyngobranchial toothplate present in a narrow band of only one or two rows (TS 7–1), the absence of teeth on the third ceratobranchial (TS 8–1), and distinct iridescent blue coloration in nuptial males (TS 16–1). This relationship differs from Rogers and Cashner's (1987) hypothesis based on allozyme data, which recognized F. julisia as sister to all other Xenisma and F. rathbuni as sister to F. bifax, F. catenatus, and F. stellifer. No morphological character states support this relationship. Although the clade including all Xenisma species except F. rathbuni is supported by a decay index of one, none of the seven trees one step longer (41 steps) than the most parsimonious tree include F. rathbuni as sister to the F. bifax, F. catenatus, and F. stellifer clade. The difference between our best supported hypothesis and trees derived from the allozyme data (Rogers and Cashner, 1987) also could be the result of different rooting of subgenus Xenisma because of outgroup choice. In both studies species in subgenus Fontinus, hypothesized to be sister to Xenisma by Wiley (1986), and subgenus Fundulus were included as outgroups. However, we also included F. chrysotus and F. zebrinus, in subgenera Zygonectes and Plancterus, respectively. Although the outgroups used in both studies were similar, outgroup choice could account for the differences between the morphological and molecular (allozyme and RFLP) trees.

Relationships within the subgenus Xenisma suggest that both dispersal and vicariance have shaped the geographic distributions of Xenisma species. Fundulus rathbuni, the sister to the rest of the subgenus, occurs in four Atlantic slope drainages, the Roanoke, Neuse, Peedee, and Santee River drainages, within the Piedmont physiographic province of Virginia and North Carolina. Meristic variation among populations from these drainages (Brown, 1955) suggests limited gene flow. The remaining species of Xenisma occur in the Mississippi and Mobile basin drainages. A pattern in which a species distributed in Atlantic slope drainages is sister to a more widespread taxon distributed west of the Appalachians is repeated in several distantly related fish groups. For example, Percina rex is restricted to the Roanoake and Chowan River drainages of Virginia (Jenkins and Burkhead, 1993) and is sister to the rest of the subgenus (Near, 2002), which is widely distributed in the Mississippi and Gulf coast drainages. Also, Hypentelium roanokense occurs in the Roanoke River drainage of Virginia and is sister to the rest of the genus (Berenzen et al. 2003); sympatry with H. nigricans in the Roanoke is likely the result of recent dispersal (Berendzen et al. 2003). Phylogenetic and geographic concordance among the unrelated groups suggest a historic event dividing the ranges of the common ancestors of subgenera Xenisma and Percina and the genus Hypentelium between the Atlantic and Gulf drainages.

Fundulus albolineatus and F. julisia had or have very restricted ranges in the Tennessee and Cumberland River drainages. The extinct F. albolineatus was known from a single spring in what is now downtown Huntsville, Alabama (Mettee et al., 1996). The ancestor of F. albolineatus and F. julisia separated from the ancestor of F. bifax, F. catenatus, and F. stellifer after the isolation of F. rathbuni on the Atlantic slope. The syntopic occurrence of the narrowly distributed F. julisia and the widely distributed F. catenatus suggests dispersal. This was likely dispersal by F. catenatus rather than F. julisia because molecular data suggest that much of the disjunct distribution of F. catenatus can be explained by more recent dispersal than by vicariance (Strange and Burr, 1997).

Fundulus bifax, F. catenatus, and F. stellifer exhibit a much wider distribution than their putative sister clade. Fundulus stellifer is known from the Coosa River system and the Alabama River system of the Mobile basin (Mettee et al., 1996). Fundulus bifax also is distributed in the Mobile basin in the Tallapoosa river system with a single population in a tributary to the lower Coosa (Cashner et al., 1988; Mettee et al., 1996). Fundulus catenatus has a widespread distribution with disjunct populations in the Ozark, Ouachita, and Eastern Highlands and small populations in the White River of Indiana and the Homochitto drainage of Mississippi. We accept the relationships among these species supported by the allozyme (Rogers and Cashner, 1987; Cashner et al., 1988; Grady et al., 1990) and restriction fragment (Strange and Burr, 1997) analysis data since the morphological data are equivocal. Fundulus bifax and F. catenatus are recognized as sister taxa, yet the current distribution of F. stellifer lies between them. The existence of a connection between the Tennessee River system, which is part of the Mississippi basin and the Mobile basin has been proposed by Mayden (1988) and Etnier and Starnes (1993). Some examples of clades where there is support for a sister group relationship between taxa in the Mobile and Mississippi basins include Hypentelium etowanum and Hypentelium nigricans (Berendzen et al., 2003); Percina antesella, Percina tanasi, and Percina uranidea (Near, 2002); Percina evides, Percina aurantiaca, Percina palmaris, and Percina (Alvordius) sp.; and Phenacobius catostomus, Phenacobius uranops, and Phenacobius crassilabrum (Mayden, 1989; Dimmick and Burr, 1999). However, in these cases, there is no evidence of paraphyly of taxa in the Mobile basin with respect to taxa in the Mississippi basin as is the case with F. stellifer and F. bifax (Rogers and Cashner, 1987; Cashner et al., 1988). A better understanding of this unique distributional pattern requires further phylogenetic and geologic study.

Material Examined

The number of alcohol preserved specimens examined is indicated after the catalog number followed by cleared and stained specimens examined in parentheses. Ingroup taxa are listed first followed by outgroup taxa. Fundulus albolineatus: UMMZ 157692, cotype, Alabama, Madison Co., Spring Cr. Fundulus bifax: JFBM 35197, 5(3), Alabama: Tallapoosa Co., Josie Leg Cr; UMMZ 213930, holotype, Alabama, Tallapoosa Co., Tallapoosa R; UMMZ 213931, 4 paratypes, Alabama, Tallapoosa Co., Tallapoosa R. Fundulus catenatus: JFBM 37334, 3 (2), Mississippi: Lincoln Co., Homochitto R.; JFBM 37590, 5(1); Tennessee; Blount Co.; Little R.; JFBM 37773, 4(1); Indiana, Johnson Co., Leatherwood Cr.; JFBM 37821, 4(1), Tennessee, Lewis Co., Buffalo R.; JFBM 37920, 4(2), Arkansas, Pike Co., Caddo R.; KU 11550, 4(1), Tennessee, Jackson Co., Roaring R.; KU 17616, (7), Missouri, Jefferson Co., Big R. Fundulus julisia: KU 20999, 2(1), Tennessee, Cannon Co., tributary to McMahan Cr.; UMMZ 120861, 3 paratypes, Tennessee, Coffeee Co., Spring Branch.; UMMZ 120914, 8 paratypes, Tennessee, Coffee Co., Little Duck R.; UMMZ 121013 and 21014, (1)7 paratypes, Tennessee, Coffee Co., Hunt Cr.; UMMZ 207690, 2, Tennessee, Coffee Co., W Fork Hickory Cr. Fundulus rathbuni: JFBM 38544, 5, North Carolina, Caswell Co., County Line Cr.; JFBM 38634, 4(1), North Carolina, Randolph Co., Uwharrie R.; JFBM 38675, 3(2), North Carolina, Randolph Co., Deer R.; UMMZ 147615, (5), North Carolina, Sugartree Cr. Fundulus stellifer: JFBM 19834, 2(1), Tennessee, Bradley Co., Conasauga R.; JFBM 35224, 4(2), Alabama, Coosa Co., Hachemendega Cr.; KU 18168, (3), Alabama, Calhoun Co., Little Hillabee Cr.; KU 20263, 9, Alabama, Calhoun Co., Little Hillabee Cr. Fundulus chrysotus: KU 18165, 5(6), Florida, Lake Co., Hogeye Sink. Fundulus diaphanus: JFBM 12707, 20, Polk Co., Minnewaska L.; KU 18192, 5(3), Pennsylvania, Franklin Co., Conococheaque Cr. Fundulus heteroclitus: KU 15351, 5(5), Massachusetts, Norfolk Co., Quincy. Fundulus seminolis: KU 18195, 3(11), Florida, Putnam Co., St. John's R. Fundulus zebrinus: KU 5362, 16, Kansas, Barton Co., Arkansas R.; KU 14726, (5), Kansas, Edwards Co., Arkansas R.