Phylogeny, taxonomy, genetics and global heritage ranks of an imperilled, freshwater snail genus Lithasia (Pleuroceridae)

Specimen collection and sequencing

Specimens representing all recognized (Burch & Tottenham 1980) Lithasia species and subspecies were collected live from throughout their ranges where available (Fig. 1, Table 2) and stored frozen or in 95% ethanol. Specimen vouchers were deposited at the Illinois Natural History Survey as well as in the University of Alabama Gastropod Collection. L. hubrichti is considered extinct (P. Hartfield pers. comm.), and individuals bearing the L. curta and L. salebrosa subglobosa phenotypes were not found. Specimens of Io fluvialis and Leptoxis species were also collected and formed the ingroup with Lithasia. Three species of Pleurocera and Elimia hydei were chosen as outgroup taxa, as the two genera are sister to Io + Lithasia + Leptoxis (Holznagel & Lydeard 2000).

Genomic DNA was isolated from head tissues using standard phenol/choloroform methods. Mitochondrial DNA sequences were obtained for an amplified segment of the mitochondrial cytochrome oxidase c subunit I gene using primers GASCOIH (5′-TTTAT-TACTACTATTTTTAATATACG) and GASCOIL (5′-GGAATATTATTATCTCA-TTCTAAAGC) designed from consensus sequences of Littorina saxatilis (Wilding et al. 1999) and Katharina tunicata (Boore & Brown 1994). The primers amplify an approximately 1000 base pairs (bp) fragment of the gene and were used for amplification and sequencing. Double-stranded amplifications via PCR were generated using 50–500 ng of template genomic DNA in 25 µl volumes (10 mm Tris, 50 mm KCl, 2.5 mm MgCl2, 1 µm each primer, 0.1 mm each dNTP, 1.5 units Taq DNA polymerase; Fisher Scientific). The amplification regime consisted of: initial denaturation (92 °C for 120 s); 30 cycles of denaturation (92 °C for 40 s); annealing (40 °C for 40 s) and extension (72 °C for 90 s); and a final extension (72 °C for 180 s). After the first five cycles, the annealing temperature was raised to 50 °C. Double-stranded products were concentrated using Millipore Ultrafree MC filters and provided the template for cycle sequencing using the ABI BigDye kit following manufacturer's instructions. Reactions were purified using Qiagen DyeEx spin columns and sequenced on an ABI 3100 genetic analyser.

Sequence analyses

Sequences were entered in the software program xesee (version 3.0; Cabot & Beckenbach 1989) and aligned by eye with the aid of L. saxatilis sequence. The aligned matrix is available from the authors. To account for the possibility of multiple COI haplotypes within a single population, additional individuals from each population, when available, were sequenced and included in the analysis. Sample sizes are shown in Table 2.

Because any analytical method, including parsimony, tends to group sequences of similar nucleotide composition together regardless of evolutionary history (Lockhart et al. 1994), a χ² test of base frequency homogeneity was run in paup* (version 4.0b10; Swofford 2002) to test for composition bias. Sequences were examined for evidence of saturation by plotting the number of transitions and transversions at each codon position against uncorrected p-distance using mega 2.0 (Kumar et al. 2001). Skewness of tree-length distributions as a measure of phylogenetic signal (Hillis & Huelsenbeck 1992) was estimated by generating 10 000 random trees in paup*. To account for genetic variation within and among species and genera, uncorrected p-distances between all samples were calculated.

Phylogenetic hypotheses were generated under parsimony and maximum likelihood. For the parsimony analysis, 50 replicates of heuristic search with random addition were run in paup* with the following settings: random sequence addition, tree-bisection–reconnection (TBR) branch swapping and branches with minimum length equal to zero collapsed. To test the internal stability of the data, a jackknife analysis (Farris et al. 1996; 1000 replicates, 10 random additions per replicate) was performed using xac (Farris unpublished) and Bremer support values (1000 trees; Bremer 1994) were generated in nona 2.0 (Goloboff 1998). For the ML analysis, the matrix was analysed using modeltest 3.06 (Posada & Crandall 1998) to determine an appropriate model of evolution. The data were then analysed in paup* using the same settings as the parsimony analysis, with empirical nucleotide frequencies used and remaining values estimated under the TrN + G model (Tamura & Nei 1993) with six categories of substitution and a gamma distribution.

Molecules and morphology

Because most classifications of Lithasia are based on morphological features (Tryon 1873; Goodrich 1940; Burch & Tottenham 1980), 25 shell and radula characters were coded (Minton 2002) and added to the molecular data. Minton (2002) showed that these characters roughly define pleurocerid genera and some previously considered species groups (Fig. 2). The resulting matrix was analysed under maximum parsimony in paup*, again using Pleurocera and Elimia taxa as outgroups and jackknife support determined as before using pac (Farris unpublished) with the same options as in the parsimony sequence analysis.

Biogeography

In addition to noting gross patterns of geographical division in the molecular analyses, we constructed area cladograms based on distribution and phylogeny of Lithasia only and compared them to speciation hypotheses to look for any replicated geographical or faunal distribution patterns. Using the method of Wiley (1988), river drainages were viewed as ‘taxa’, and the taxa occurring in those drainages as ‘characters’ (see Crandall & Templeton 1999). River drainage designations followed Mayden (1988), with two exceptions. Garrison Creek was considered a unique drainage; while it is part of the upper Duck River drainage, it is considered to have a Cumberlandian faunal distribution (Ortmann 1924; van der Schalie 1973; Ahlstedt 1980). The lower Ohio River was also considered a distinct drainage, as the area where included taxa occur does not appear in any of the predetermined drainages. Using the strict consensus phylogeny generated by total evidence analysis, terminal taxa were coded for what drainage they occurred in, and interior nodes were coded for those taxa that were in the clade and their distributions (Wiley 1988). The resulting matrix (Table 3) was analysed using paup*.

Molecular analyses

Aligned sequences with no necessary gaps, representing 47 total and 23 Lithasia haplotypes resulted in a data matrix of 890 characters, of which 306 were variable and 257 were parsimony informative (aligned matrix available from authors). A χ² test of base frequency homogeneity showed no composition bias (χ² = 27.64, d.f. = 138, P = 1.0000). While most sequence diversity occurred at the third codon position, transition/transversion plots showed no evidence of saturation at any codon position (data not shown). Random tree length distribution indicated significant phylogenetic signal in the data (g1 = −0.4722).

Recent studies (e.g. Sunnucks & Hales 1996; Zhang & Hewitt 1996; Lovette et al. 1999; Williams & Knowlton 2001) have reported the presence of nuclear copies of mitochondrial genes (‘pseudogenes’/‘numts’) that may be amplified and sequenced, thereby confounding phylogenetic analyses. We believe that the sequences reported here are mitochondrial in origin, based on a single amplicon after PCR, ‘clean’ sequencing reads of both strands, a lack of internal stop codons or frameshifts and base frequency and substitution patterns comparable to other pleurocerid studies. The existence of or changes in any of the above may indicate the presence of nuclear sequence (Collura & Stewart 1995; Sunnucks & Hales 1996; Sorenson & Quinn 1998).

Maximum parsimony analysis yielded 15 trees (766 steps, CI = 0.52) rendering Lithasia nonmonophyletic (Fig. 3). Specimens of L. obovata formed a clade between Elimia and Pleurocera species, while two specimens of L. geniculata pinguis from the Collins River fell between Elimia hydei and the Leptoxis clade. The remaining Lithasia species formed a clade sister to Io. Within the ‘Lithasia’ clade, two major clades were recovered: one containing specimens of L. lima, L. salebrosa and its subspecies and L. verrucosa; and one comprised of L. armigera, L. duttoniana, L. geniculata and its subspecies, and L. jayana. Within the L. verrucosa clade, three subclades were recovered: one containing specimens from the White River drainage of Arkansas; one containing lower Ohio River specimens; and one of Tennessee River drainage specimens. The position of Tennessee River verrucosa could not be determined in relation to L. lima and L. salebrosa and its subspecies. Within the second clade, two subclades were recovered. All L. armigera specimens grouped together sister to a clade of L. geniculata fuliginosa from the Cumberland and upper Duck rivers. The second subclade showed L. g. fuliginosa specimens from the Buffalo River sister to a clade containing taxa from the lower Duck River. This lower Duck River clade contained specimens of L. geniculata and its subspecies along with L. duttoniana and L. jayana. L. g. pinguis specimens in this clade possessed unique haplotypes, as did L. jayana, while the remaining L. geniculata and L. duttoniana shared a single haplotype. The relationships within the clade could not be resolved. Many currently recognized taxa were not recovered as monophyletic. Only L. armigera was recovered in all trees. L. geniculata and L. verrucosa were rendered polyphyletic, and the remaining taxa were unresolved.

Maximum likelihood analysis yielded a single tree (Fig. 4). Lithasia was rendered nonmonophyletic, with E. hydei, L. geniculata pinguis C1 and C2 and L. obovata 1 and 2 forming a clade. The remaining Lithasia species formed a single clade with a topology consistent with that achieved under parsimony.

Genetic variation within and among groups

Of all the populations sampled in the ‘Lithasia’ clade, only L. g. pinguis from the Duck River showed the presence of multiple haplotypes, differing by an uncorrected p-distance of 0.45%. Genetic distances for all groups are shown in Table 4. The maximum interpopulation difference within a putative species was 4.27% between L. g. fuliginosa from the Duck and Red Rivers, while many pairs (e.g. L. verrucosa B and W, O and I, etc.) showed no difference. The maximum distance seen within the ‘Lithasia’ clade was 9.10% between L. jayana and L. s. salebrosa. Within-genus distances were less than between genus distances in all cases except for the ‘Elimia’ clade and between Athearnia and Leptoxis.

Molecules and morphology

Analysis of the combined data sets yielded 12 trees of 872 steps (CI = 0.47; Fig. 5). Lithasia was again rendered polyphyletic by the positions of L. g. pinguis from the Collins River and L. obovata. The overall topology was consistent with that generated using only molecular data, although a number of clades (e.g. L. verrucosa specimens and L. armigera) were better resolved. Jackknife support was similar to that generated solely using molecular data.

Biogeography analysis

Clear biogeographical divisions were seen in the analyses. Using molecular data, all remaining Lithasia taxa formed clades based on river drainages with the exception of L. armigera. The subclades containing L. verrucosa were divided into Ohio, White and Tennessee river drainages, although the positions of each within the larger clade could not be resolved. Lithasia geniculata taxa were divided into Buffalo, Cumberland/upper Duck and lower Duck river groups. All lower Duck River taxa formed a single clade. Jackknife analysis strongly supported the division of Lithasia taxa into their respective clades.

Parsimony analysis of area relationships produced two trees (57 steps, CI = 0.91; Fig. 6). Two major clades were resolved: Cumberland/Duck river drainages; and a Tennessee, lower Ohio and Ozark (Black and White rivers) drainages. This pattern is inconsistent with previously generated analyses (e.g. Mayden 1988; Crandall & Templeton 1999), as the ‘lower Ohio’ (Cumberland, Duck, lower and upper Tennessee) and Interior Highland rivers (Wabash, lower Ohio, Black and White) did not group together. The pattern is also inconsistent with other speciation hypotheses in North American rivers (Mayden 1988).

Species composition of Lithasia

Although historically considered a genus based on morphological characters, Lithasia was rendered polyphyletic in all analyses, with L. obovata and L. g. pinguis from the Collins River falling outside the clade of all remaining species. Burch (1982) noted that based on shell characters, L. obovata and L. g. pinguis appear to belong to Elimia/Pleurocera and Leptoxis, respectively, but are placed in Lithasia because of their radulae. This hypothesis was supported by Minton (2002), who showed that L. obovata fell between Elimia and Pleurocera, and L. g. pinguis grouped with Leptoxis using shell and radula characters. Based on mitochondrial DNA sequences alone and combined with morphological data, L. obovata again fell between Pleurocera and Elimia taxa, suggesting that it is not a Lithasia species and needs to be studied further for proper taxonomic placement. Taxa believed to be L. g. pinguis were present in two parts of the trees. Specimens from the Duck River grouped with other Duck River taxa in the main ‘Lithasia’ clade, while two specimens from the Collins River each possessed unique haplotypes and grouped with Elimia species. L. g. pinguis is known from the Caney Fork and its branches and the Duck River, Coffee County, TN (Burch & Tottenham 1980). These two populations represent at least two separate taxa, and only those from the Duck River should be considered Lithasia. Again, further work needs to be undertaken to determine the identity and placement of the Collins River taxa. The remaining taxa form a well-supported clade that should be recognized as the genus Lithasia.

Genetic variability within Lithasia

Genetic divergences seen in Lithasia, as defined above, were comparable to literature values for mitochondrial 16S rDNA sequence variation in pleurocerids (Lydeard et al. 1997, 1998; Holznagel & Lydeard 2000). The maximum difference within a putative species was 4.27% for L. g. fuliginosa, exceeding the largest published value of 3.7% at 16S for populations of Elimia carinocostata (Lydeard et al. 1998). The maximum intrageneric difference within Lithasia was 9.10%, well within the 8.69% and 10.1% reported by Holznagel & Lydeard (2000) at 16S for Pleurocera and Elimia, respectively, and below the 12–15% they reported for Leptoxis. Overlapping intra- and intergeneric divergences were seen in ‘Elimia’ and between Athearnia and Leptoxis. Athearnia is frequently considered congeneric with Leptoxis (Burch & Tottenham 1980; Turgeon et al. 1998; Holznagel & Lydeard 2000; but see Morrison 1971; Dillon & Ahlstedt 1997), and similar patterns of overlap were reported for Elimia (Lydeard et al. 1998; Holzangel & Lydeard 2000).

Species delineation

Several versions of the PSC recognize species as minimal monophyletic groups (de Queiroz & Donoghue 1988; Mishler & Donoghue 1982; Mishler & Brandon 1987). This has proved useful in a variety of studies on freshwater bivalves (Mulvey et al. 1997; Roe & Lydeard 1998; King et al. 1999; Lydeard et al. 2000) and gastropods (Lydeard et al. 1998). The PSC further provides a framework for recognizing unique monophyletic lineages based on sequence divergence when morphological distinctness historically used to define species is problematic (e.g. Wilke & Falinowski 2000). This is useful in pleurocerids, where cryptic species that are not closely related have been observed (Chambers 1978). Using this concept, seven phylogenetic species (Fig. 3) would be recognized based on the molecular phylogeny of Lithasia (3, 4): (1) L. ‘fuliginosa’, a composite of what were formerly L. duttoniana, L. g. fuliginosa, L. g. geniculata, L. jayana and L. g. pinguis, all of which are from the Duck River; (2) a new, previously undescribed species comprised of L. g. fuliginosa from the Buffalo River; (3) L. armigera, the only currently recognized morphospecies recovered as monophyletic; (4) L. ‘geniculata’, a group comprised of L. g. fuliginosa from the upper Duck River tributaries and Cumberland River drainage; (5) L. lima, a composite comprised of L. lima, L. verrucosa from the Elk and Tennessee rivers, L. s. salebrosa and L. s. florentiana; (6) a new, previously undescribed species comprised of L. verrucosa from the White River drainage, Arkansas; and (7) L. ‘verrucosa’, comprised of L. verrucosa from the Ohio River drainage. Based on the combined morphological and molecular data and biogeography analyses, it is possible that L. jayana and L. salebrosa (salebrosa and florentiana) may warrant recognition as morphologically diagnosable, phylogenetic species. However, we await the acquisition of additional genetic data to examine this further.

The molecular and combined molecular + morphology data analyses uncovered hidden diversity within what were once considered widely distributed, highly variable morphospecies. Many of the newly recognized phylogenetic species have distributions that correspond quite well with geography (Fig. 6). The distinction between Cumberlandian and upper Duck river fauna from lower Duck River mollusc fauna has been noted by other authors. In their works on freshwater mussels, Ortmann (1924) and van der Schalie (1973) both noted the upper Duck River contains Cumberlandian species while the lower Duck River contains species of the Ohioan/Interior faunal group and those of undetermined origin (Ahlstedt 1980). The sister relationships of the Duck and Buffalo river taxa is not surprising, as the Buffalo River is the main tributary to the Duck River and they share faunal assemblages (Ortmann 1924; van der Schalie 1939, 1973). Lithasia ‘lima’, the undescribed sp. from the White River, and L. ‘verrucosa’ showed similar geographical division. The undescribed sp. is from the Black, Spring and White river drainages of Arkansas, L. ‘lima’ is from widely distributed in the Tennessee River and L. ‘verrucosa’ is from the Wabash and lower Ohio rivers (Goodrich 1940; Burch & Tottenham 1980). This general pattern of separation is consistent with freshwater fish and crayfish distributions and river drainages (Crandall & Templeton 1999; Mayden 1985, 1988).

Conservation implications from a PSC–based classification

Employing the PSC considerably modified the traditional classification of Lithasia, resulting in the synonymization of some subspecies and species and recognition of newly delineated species. The Natural Heritage Network has developed a method for estimating the degree of relative imperilment (Master et al. 2000). Global conservation status ranks are based on a one-to-five scale, ranging from critically imperilled (G1) to widespread, abundant and demonstrably secure (G5). With the new phylogenetically based classification, however, the conservation status has been altered appreciably. For example some taxa, such as L. lima, which is now part of the phylogenetic ‘lima’ has gone from G2 to G3G4, while some subspecies now constitute more narrowly restricted phylogenetic species, making them rarer (e.g. two new undescribed species have gone from G3G4 to G2; Table 1). The findings of this study suggest that other widely distributed pleurocerid taxa in particular (e.g. Elimia livescens, Leptoxis praerosa, Pleurocera canaliculatum and P. vestitum), and poorly delineated taxonomic groups in general, may actually be comprised of several phylogenetic species which warrant protection at a different level than currently recognized.