Percina

James D. Williams, David A. Neely, Stephen J. Walsh & Noel M. Burkhead, 2007, Three new percid fishes (Percidae: Percina) from the Mobile Basin drainage of Alabama, Georgia, and Tennessee., Zootaxa 1549, pp. 1-28 : 21-25

[[ Genus Percina ]]

Morphometric comparisons

Proportional measurements for the three species of Percina are provided in Table 11. Results of the sheared principal component analysis (PCA) are presented in Table 12 and Figures 4-5. In the PCA of all species, the first three components accounted for 91% of the morphometric variance. Variable loadings on the first principal component were positive and indicative of PC I as a general size factor (Bookstein et al. 1985), and this component accounted for 83.5% of the total variance in the complete covariance matrix; values for PC II and PC III were 4.6% and 2.8%, respectively. The scatter of component scores on sheared PC II and sheared PC III revealed little differentiation among all three species (Fig. 4), with greatest overlap on both component axes between P. sipsi and P. smithvanizi . Heaviest loadings on sheared PC II were snout length and head length (both positive) and anal fin base length and body depth (both negative). Heaviest loadings on sheared PC III were body depth, pectoral fin length, pelvic fin origin to spinous dorsal fin origin (all positive), and anal and soft dorsal fin base lengths (both negative) (Table 12).

To further compare shape differences between Percina kusha and P. smithvanizi , a second principal component analysis was done with the removal of measurements for P. sipsi from the data matrix. In this analysis, the first three components accounted for 88% of the morphometric variance, with loadings on PC I through PC III of 76.8%, 5.3%, and 2.6%, respectively. A scatter plot of component scores revealed moderate separation between the two species along PC II (Fig. 5). Variables that loaded most heavily on sheared PC II were anal fin base length and body depth (both positive) and snout length and head length (both negative); proportional differences between P. kusha and P. smithvanizi in these measurements were accompanied by substantial overlap in range values (Table 11).

Based on results of the morphometric analysis, the combination of mensural data were of limited use in distinguishing specimens among the three Percina species. In spite of broad overlap for all morphometric characters among species, a slight difference in body depth between P. smithvanizi and the other two species was evident in a bivariate plot of log body depth versus log SL (Fig. 6). Analysis of covariance (ANCOVA; SPSS ver. 12.0) confirmed homogeneity of variance in the regression lines (Table 13). Using the Bonferroni procedure to control for type I error across three comparisons (p <0.0167), post hoc contrasts indicated that P. smithvanizi differed significantly in adjusted mean log body depth from specimens of both P. kusha (F = 10.99; p = 0.002) and P. sipsi (F = 7.75; p = 0.007), but that the latter two did not differ from each other.

Molecular analyses

Our mtDNA data set consisted of 83 OTUs, including seven outgroup taxa. Thirty-three most-parsimonious topologies (L = 3972, CI = 0.22, RC = 0.13) were recovered in the unweighted MP analysis. The Bayesian topology (Fig. 7) was nearly identical to the consensus tree of the MP topologies, differing primarily in the degree of resolution of basal branches. Several features of these trees are worthy of note. While support for many terminal nodes was generally high, the basal relationships of the genus Percina are poorly resolved, consistent with the analyses of Near (2002) and Sloss et al. (2004). The subgenera Cottogaster , Imostoma , Percina and Swainia were all resolved as monophyletic groups, with strong support. Sufficient structure was resolved, however, to clearly support a sister-taxon relationship between P. kusha and P. smithvanizi , as well as a strongly supported clade consisting of P. aurolineata + ( P. sciera + P. sipsi ). The three taxa described herein were never recovered as a monophyletic group in any analyses. There was no consistent geographic structure observed in populations of either P. sipsi or P. smithvanizi , however, populations of P. kusha from the Conasauga and Etowah River were always recovered as reciprocally monophyletic groups.

Populations of Percina sipsi , P. smithvanizi , and P. kusha displayed relatively low levels of variation (Table 14). The five sampled specimens of P. kusha possessed three haplotypes; there were seven substitutions (four transitions and three transversions) and no indels. All sampled individuals from the Conasauga River shared a single haplotype, which differed from the two haplotypes sampled from the Etowah River by 0.53- 0.64% pairwise sequence divergence. The two Etowah haplotypes were 0.09% divergent from each other. The six sampled specimens of P. smithvanizi (representing the entire geographic range of the taxon) possessed four haplotypes; there were ten substitutions (five transitions and five transversions) and no indels. The haplotypes had pairwise sequence divergence values between <0.01% and 0.26%.

The four sampled specimens of Percina sipsi each had unique haplotypes; there were five substitutions (two transitions and three transversions) and no indels. Haplotypes had pairwise sequence divergence values between 0.18% and 0.35%. One specimen from the Sipsey Fork identified as P. sciera on the basis of lateral blotching, fully scaled nape and a well-developed suborbital bar, had a haplotype that was very similar to that of P. sipsi (and which only differed from P. sipsi by between 0.26% and 0.44%). This specimen was resolved with P. sipsi in all analyses.

Although P. sipsi and P. sciera are morphologically similar (see Comparison with congeners, above), recovery of a sister-taxon relationship between them was unexpected.

The specimens used in the molecular analysis were adult, morphologically "good" P. sipsi and P. sciera . Our data do not allow us to determine whether this represents a natural relationship or an artifact of the mitochondrial locus examined. Considerable evidence exists for occasional mitochondrial introgression in darters (Page et al. 2003, Mendelson & Simons 2006, N.J. Lang & J.M. Ray unpublished data), and we suspect that our recovery of P. sipsi stems from introgression with P. sciera , but require additional nuclear data to test this hypothesis. Minimally, the haplotypic variation observed within P. sipsi suggests that if the mitochondrial genome is introgressed with P. sciera , it is not the result of a recent event.

Discussion

Relationships among species of Percina have been the subject of considerable attention during the past 50 years beginning with a proposed alignment of the recognized species into eight subgenera by Bailey & Gosline(1955). Using a variety of characters, Page (1974) employed a numerical taxonomic approach to reexamine the subgeneric alignment of the 32 species of Percina known at that time and recognized nine subgenera. Only one of the three species described herein, P. kusha , was included in the analysis by Page (1974) and it was assigned to the subgenus Alvordius . Based on color and morphological characters all three species were subsequently considered to belong to the subgenus Alvordius . More recently Near (2002) examined the phylogenetic relationships among 40 species of the genus Percina using mitochondrial cytochrome b gene sequences. This analysis did not result in the recovery of a monophyletic Alvordius , although many taxa formerly allocated to this subgenus were resolved in a large polytomy. Based on the results of his analysis, Near (2002) recognized eight clades with a group of eight species, including P. smithvanizi , that were not classified (placed) in one of the monophyletic clades. Near (2002) recovered P. smithvanizi as sister to P. palmaris although support for this clade was low. Five of the clades recognized by Near (2002) were in agreement with subgenera delineated by Page (1974).

The additional taxon sampling in the current study supports a close relationship between P. kusha and P. smithvanizi , but do not provide a robust hypothesis of the relationships of this clade.

Additional analyses based on different markers will be required to produce a robust phylogeny of the group and in particular the three taxa described herein, as well as to resolve the discord between morphologically diagnosed subgenera (Page 1974) and clades inferred from mitochondrial data (e.g., Near 2002).

Percina kusha superficially resembles P. macrocephala in having a dark, boldly contrasting, undulating lateral stripe, but the latter species differs in having the snout longer than the orbit length; cheeks, opercles and breast typically naked; it has smaller scales (70-90 lateral line scales, 8 or 9 scales above and 11 or 12 scales below the lateral line, 20-28 transverse scales), more fin rays (dorsal spines 11-14, dorsal rays 12-13), and more vertebrae (44-45, data from Bailey & Gosline 1955). It also differs in details of pigmentation, particularly in having a well-developed subocular bar, dark saddles along the midline of the dorsum, dark vermiculations above the lateral stripe and a dusky bar connected to and extending downward from the rounded basicaudal spot. Based on color and morphology, P. kusha and P. macrocephala do not appear to be closely related.

Percina sipsi and P. smithvanizi superficially resemble P. aurolineata and P. sciera in having a series of dark blotches connected to form a dark lateral stripe along the side of the body. However, both P. aurolineata and P. sciera usually have one or more pored lateral line scales on the caudal fin base. The majority of individuals of the three species described herein have a complete lateral line but typically lack pored scales on the base of the caudal fin beyond the hypural flexure. Percina aurolineata and P. sciera also have moderately joined branchiostegal membranes and serrate preopercular margins. In addition, P. kusha , P. sipsi , and P. smithvanizi are unusual among species of the genus in possessing a modal count of 10 dorsal soft rays and 8 anal rays.

Based on morphological characters, Percina smithvanizi was originally aligned with the subgenus Alvordius and was thought to represent an upland species isolated from related taxa in drainages below the Fall Line in the Mobile Basin. Based on this assumption, P. smithvanizi was often compared to P. maculata , the only other species in the subgenus Alvordius in the Mobile Basin. In the eastern Mobile Basin, P. maculata is confined to the Alabama, Cahaba, and Tallapoosa River drainages below the Fall Line. Based on color and morphology P. smithvanizi and P. maculata are probably not closely related. Percina maculata has a black blotch on the base of the first 3 to 4 membranes of the spinous dorsal fin, a well-defined black basicaudal spot, and 1 or more modified ctenoid scales on the center of the breast near the anterior ends of the pelvic girdle -a character it shares in common with P. gymnocephala , P. notogramma , P. pantherina and P. peltata .

While the broader relationships of the new species are equivocal, our mitochondrial data does not support recognizing them as a monophyletic group, as prior investigators have hypothesized. Additional sequence data from a different gene or genes are required to further resolve relationships within Percina and, in particular, of the three new species described herein.