Herein we clarify the taxonomy of Fluminicola coloradensis Morrison (1940), which was described for populations in the Green River and Bonneville basins but has also been treated as restricted to the former watershed and conspecific with (currently unassigned) Snake River basin populations of Fluminicola hindsi sensu Taylor (1966). Bayesian analyses of DNA sequences from 2 mitochondrial genes congruently resolved F. coloradensis and Snake River basin populations of F. hindsi sensu Taylor as a strongly supported, shallowly structured clade. Haplotypes were extensively shared by Bonneville, Snake River and Green River populations; AMOVA did not detect significant variation among basins for either gene. Morphological variation was minor. Based on these results, we assign the Snake River basin populations to F. coloradensis. We attribute the limited differentiation of widely ranging F. coloradensis to its well-integrated habitats and to dispersal mediated by geologically recent drainage transfers. The broadly disjunct population in the Owyhee River drainage may be a product of translocation, as evidenced by detection of only the most common haplotypes in these snails. Our finding that F. coloradensis is more widely distributed than previously thought suggests that it may not require conservation attention rangewide, although some geographic subunits may be at risk.
The northwestern North American genus Fluminicola (containing 24 currently recognized species) is composed of small (1.2–12.0 mm shell height), gill-breathing gastropods (commonly known as pebblesnails) that usually live in lotic habitats. Although Fluminicola is a conspicuous component of benthic communities throughout much of the region and recently has become a focus of conservation activities (e.g., USDA Forest Service and USDI Bureau of Land Management 2001), it has received little taxonomic study across most of its broad range (Hershler and Frest 1996). One of the long-standing problems is the unsettled taxonomy of Fluminicola coloradensis (commonly known as the Green River pebblesnail) and the numerous currently unassigned Snake River basin populations that closely resemble this species morphologically.
Morrison (1940) described F. coloradensis based on the large pebblesnails in the upper Green River and Bonneville basins, which he differentiated from Fluminicola fuscus by their wider umbilicus and lighter shell color. Taylor (1966 : fig. 14) included F. coloradensis (as a synonym) in Lithoglyphus hindsii (Baird 1863) (= Fluminicola hindsi [Baird, 1863]), which he envisaged as ranging widely across the upper Green River, Bonneville, and Snake—Columbia River basins. His brief treatment consisted of a distribution map and associated figure caption which detailed the content of F. hindsi. In their subsequent revision of the genus, Hershler and Frest (1996: 12) synonymized the type material of F. hindsi (from the Kootenai River) with F. fuscus and resuscitated F. coloradensis, which they restricted to the upper Green River basin. They did not assign the Bonneville and most of the Snake River basin populations (those in the lower reaches of the Grande Ronde and Salmon rivers were allocated to F. fuscus), which they considered as belonging to one or more undescribed species. Hershler (1999) later reassigned the Bonneville Basin pebblesnails to F. coloradensis, however Frest and Johannes (2000: 13) recently suggested that this species “is likely a composite taxon” as currently constituted (also see Frest 1999).
Mitochondrial DNA sequences have proved useful for delimiting species in morphologically conservative Fluminicola (Hershler et al. 2007). Here we use partial sequences of cytochrome c oxidase subunit I (COI) and cytochrome b (cytb), together with morphological evidence, to resolve the taxonomy of F. coloradensis. In addition to filling an important gap in Fluminicola taxonomy (Hershler and Frest 1996), our findings provide additional insight into the complex biogeographic history of this genus. Our clarification of the geographic range of F. coloradensis may also assist state agencies in their efforts to conserve and manage this species.
METHODS
We analyzed mtDNA sequences from 18 populations of F. coloradensis (upper Green River basin, 6; Bonneville Basin, 12), 34 currently unassigned Snake River basin populations of F. hindsi sensu Taylor (1966), and 11 populations of F. fuscus (from the Snake— Columbia River basin), which has been delineated as the sister to F. coloradensis based on morphological (Hershler and Frest 1996) and molecular (Hershler et al. 2007) evidence. Fluminicola insolitus, which appears to be the closest relative to the F. fuscus + F. coloradensis clade (Hershler et al. 2007), was used to root the phylogenetic trees. Four COI and 3 cytb sequences used in our analyses were taken from GenBank, and the remaining 242 COI and 166 cytb sequences were newly obtained for this study. Sample codes, museum voucher numbers, locality details, and sample sizes are in Table 1, and the locations of sampling sites are shown in Fig. 1.
Genomic DNA was extracted from entire snails using a CTAB protocol (Bucklin 1992). Amplifications were conducted in a 25 µL total volume, containing 5 µL of 5X buffer, 0.5 µL of dNTPs (10mM), 2 µL of MgCl2 (25 mM), 1.25 µL of each primer (10 µM), 1 unit Taq polymerase, 1 µL of template DNA (ca. 100 ng double-stranded DNA), and 13.8 µL of sterile water. COIL1490 (Folmer et al. 1994) and COH654 (Hershler and Liu 2012) were used to amplify a 634 bp fragment (excluding primers) of the COI gene. Cytb284F and Cytb757R were used to amplify a 449 bp fragment (excluding primers) of the cytb gene. Cytb284F (5′ATT TAT TTW CAT ATY GGW CGA GG3′) and Cytb757R (5′TGG AAT RAA ATT TTC TGG GTC TG3′) were designed based on conserved regions of cytb in an alignment using previously published sequences from the following gastropods: Littorina saxatilis (AJ132137), Potamopyrgus antipodarum (GQ996433), and Oncomelania hupensis (NC013073). Thermal cycling was performed with an initial denaturation for 2 min at 94 °C, followed by 35 cycles of 1 min at 94 °C, 1 min at 50 °C, 2 min at 72 °C, with a final extension of 10 min at 72 °C. The amplified polymerase chain reaction product was cleaned using the Exonuclease I/Shrimp Alkaline Phosphatase method. Approximately 10–20 ng of cleaned polymerase chain reaction product was used as a template in a cycle sequencing reaction using the BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystem, Inc.). The following cycling conditions were used: 30 cycles of 94 °C for 20 s, 50 °C for 20 s, and 60 °C for 4 min. The cycle-sequenced product was cleaned using the ethanol precipitation method and then run on an ABI 310 genetic analyzer. Sequences were determined for both strands and then edited and aligned using Sequencher™ version 4.8. One to 10 specimens were sequenced for both genes for most of the sampled populations; we were unable to amplify and sequence cytb from 4 populations (Table 1).
Sequence divergences (uncorrected p distance) were calculated using MEGA5 (Tamura et al. 2011). Genetic structuring among drainage basins was assessed by analysis of molecular variation (AMOVA) as implemented in ARLIQUIN 3.5 (Excoffier et al. 2005). Haplotype networks were generated using TCS version 1.21 (Clement et al. 2000). Because of missing data in the cytb dataset, phylogenetic analyses of the 2 datasets were performed separately using Bayesian inference in MrBayes 3.2.1 (Ronquist et al. 2012). MrModeltest 2.3 (Nylander 2004) selected the HKY + I + G and HKY + G models for the COI and cytb datasets, respectively, which best fit these data under the Akaike information criterion. In the Bayesian analysis Metropolis-coupled Markov chain Monte Carlo simulations were performed with 4 chains for 5,000,000 generations, and Markov chains were sampled at intervals of 10 generations to obtain 500,000 sample points. We used the default settings for the priors on topologies and HKY + I + G model parameters for COI and HKY + G model parameters for cytb. The sampled trees with branch lengths were used to generate a 50% majority-rule consensus tree with the first 25% samples removed to ensure that the chain sampled a stationary portion.
Shells, radula, opercula, and soft anatomy of relevant pebblesnail collections in the Smithsonian Institution's National Museum of Natural History (USNM), the University of Minnesota Bell Museum of Natural History (UMBMNH), and the Orma J. Smith Museum of Natural History (ALBRCIDA) were studied using methods employed in recent taxonomic studies of Fluminicola (Hershler and Frest 1996, Hershler 1999, Hershler et al. 2007). Shells were cleaned with commercial bleach (to remove surface deposits) before being photographed. Fresh material was collected when needed for anatomical or molecular studies. All of the samples used for molecular studies were collected after 1990. Small samples (ca. 25–50 individuals) were collected by hand or by washing rocks and stones. Subsamples were anaesthetized overnight with menthol crystals and fixed in dilute formalin. Preservation of subsamples was in 70% ethanol for morphologic study and in 90% ethanol for molecular study.
RESULTS
Mitochondrial DNA Analyses
The alignment of COI sequences yielded 634 bp, of which 98 sites were variable (15.5%) and 46 were parsimony informative (7.3%). Average base frequencies for COI were 27.3% A, 31.7% T, 21.7% C, and 19.3% G. Base frequencies were homogeneous across all sites (χ2 = 13.75, df = 306, P = 1.00). For cytb, 449 bp were sequenced, of which 74 were variable (16.5%) and 30 were parsimony informative (6.7%). Average base frequencies for cytb were 26.5% A, 35.6% T, 20.6% C, and 17.3% G. Base frequencies were homogeneous across all sites (χ2 = 15.38, df = 222, P = 1.00). The new sequences that were generated for this study were deposited in GenBank under accession numbers JQ996156–JQ996230 (one sequence per haplotype).
In the Bayesian analysis, Metropolis-coupled Markov chain Monte Carlo simulations were performed with 4 chains for 5,000,000 generations. At the end of the analysis, the average standard deviation of split frequencies was 0.0064 for COI and 0.0059 for cytb and the Potential Scale Reduction Factor (PSRF) was 1, indicating that the runs had reached convergence. Haplotypes detected in F. coloradensis, unassigned pebblesnails from the Snake River basin, and the lower Salmon River population of F. fuscus (F59) formed a well-supported (100%) but weakly structured clade (Clade “A”) in both the COI (Fig. 2) and cytb (Fig. 3) Bayesian trees. Haplotypes from populations in the Bonneville, upper Green River, and Snake River basins were highly intermingled within clade A (Figs. 2, 3). The 41 COI and 21 cytb haplotypes detected are detailed in Appendices 1 and 2, respectively (GenBank accession numbers JQ996156–JQ996196 for COI haplotypes I–XLI, and JQ996205–JQ996225 for cytb haplotypes I–XXI). The pairwise sequence divergence among these ranged from 0.2%–1.6% for COI and 0.2%– 1.8% for cytb. Pairwise divergence among populations ranged from 0.0%–1.4% for COI and 0.0%–1.6% for cytb; divergence within populations ranged from 0.0%–1.3% for both genes.
TABLE 2.
Analysis of molecular variance (AMOVA). Groups are upper Green River basin (samples 18, 19, 26, 27, 28, 39), Bonneville Basin (16, 17, 20, 21, 22, 23, 24, 25, 33, 37, 38, 40, 45), Snake River basin (1, 2, 4, 5, 7, 8, 9, 10, 12, 13, 14, 15, 29, 30, 31, 32, 34, 35, 36, 41, 42, 43, 47, 48, 49, 50, 51, 52, 53, 56, 57, 58, 59, 60, 61, 63).
The most common COI haplotype (I) within clade A was detected in 5 upper Green River and 5 Bonneville basin populations of F. coloradensis; the lower Salmon River population of F. fuscus; and 18 unassigned populations spread among the upper reach, Owyhee River, and Salmon River segments of the Snake River watershed (Appendix 3). Most of the other haplotypes (30 of 40) were restricted to single populations and many were singletons. The most common cytb haplotype (III) was detected in 5 upper Green River and 5 Bonneville Basin populations of F. coloradensis; the lower Salmon River population of F. fuscus; and 16 unassigned populations spread among the upper Snake, Crooked Creek, and Salmon River segments of the Snake River watershed (Appendix 4). Most of the other haplotypes (14 of 20) were restricted to single populations, and many were singletons. Haplotypes were organized in a star-like pattern (with the common haplotype centrally positioned) in both of the TCS networks (e.g., Fig. 4, the COI topology), suggesting recent population expansion and diversification (Slatkin and Hudson 1991) within this clade.
When populations belonging to clade A were grouped by major drainage basin (upper Green River, Bonneville, Snake River) most of the COI and cytb variation was partitioned among populations within basins (64.3% and 85.4%, respectively) and within populations (33.9% and 12.2%); variation between basins was very small (1.8% and 2.4%) and nonsignificant (Table 2).
Morphology
Shell size, shape, and color varied considerably among populations of both F. coloradensis (Fig. 5A–G) and the unassigned Snake River basin pebblesnails (Fig. 5H–T). Variation was also marked within some of the populations (e.g., Fig. 5O–Q). Although most of the unassigned pebblesnails conformed well to F. coloradensis, some of the populations distributed along the Snake River between Murtaugh and King Hill were distinguished by their narrow, high-spired shells (Fig. 5N). The shells of the lower Salmon River populations currently identified as F. fuscus (e.g., Fig. 5U) also fall into the range of variation of F. coloradensis; note that none of the former have the well-developed subsutural angulation or keel on the later whorls that characterizes F. fuscus (Hershler and Frest 1996). The unassigned and lower Salmon River pebblesnails also closely resembled F. coloradensis anatomically, although some populations differed from previously studied material of this species in several features such as the shape of the distal section of the penis and the shape of bursa copulatrix (Fig. 6). However, similar variation was observed among populations of F. coloradensis that were re-studied during the course of this project. We did not observe significant differences in the number of cusps and shape of the radular teeth among F. coloradensis and the unassigned pebblesnails (Table 3; Fig. 7, compare with Hershler and Frest 1996 : fig. 8A–C).
DISCUSSION
Our findings do not provide any basis for restricting F. coloradensis to the upper Green River basin per recent proposals (Frest 1999, Frest and Johannes 2000). These results also indicate that F. coloradensis and the unassigned Snake River basin populations of F. hindsi sensu Taylor (1966) are closely similar both morphologically and genetically. Haplotypes observed in F. coloradensis and the unassigned populations are highly intermingled within a single, well-supported clade that lacks obvious structure. The variation among haplotypes in this clade was less than typically observed between congeners (>2% for both genes in most cases; Hershler et al. 2007). Based on these results we conclude that the unassigned Snake River basin populations are conspecific with F. coloradensis. The other taxonomic change that we are implementing based on our findings is to transfer the lower Salmon River populations currently classified as F. fuscus (per Hershler and Frest 1996:12) to F. coloradensis. The first author previously had difficulty identifying these pebblesnails because of their typically eroded shells (Fig. 5U). The material that we are newly referring to F. coloradensis is listed in Appendix 5.
TABLE 3.
Variation in number of cusps on the radular teeth among populations of Fluminicola coloradensis and unassigned Snake River basin pebblesnails; n = 5 for all samples (see Hershler et al. 2007 for methods).
The shallow genetic structuring of F. cohradensis across 3 major drainage basins (Colorado River, Snake River, Bonneville) detailed herein contrasts strikingly with the extensive diversification in the upper Sacramento River basin documented in the only previous molecular study of pebblesnails (Hershler et al. 2007) and suggests that the biogeographic history of Fluminicola in relation to topography and drainage has been complex. The locally extensive differentiation of pebblesnails in the upper Sacramento River watershed was attributed to the high fidelity of (most of) the members of that fauna to insular spring and spring-influenced habitats, and the complex late-Cenozoic tectonic-volcanic history of the region (Hershler et al. 2007). The limited differentiation of F. coloradensis across its broad geographic range may be attributed, in part, to the typical occurrence of this species in rivers and other large, well-integrated habitats. There is also abundant geologic evidence of late-Quaternary drainage exchanges between the Snake River and Bonneville watersheds—e.g., diversion of the Bear River from the former to the latter (Bouchard et al. 1998), spillover of Lake Bonneville into the Snake River basin (Malde 1968, O'Connor 1993)—that could have facilitated dispersal of F. coloradensis among these basins. Although there is no evidence that the Green River was integrated with either the Bonneville or Snake River basins during the late Cenozoic, dispersal of pebblesnails across the upper Green River drainage divide may have been facilitated instead by headwater stream captures. Hansen (1985, 1986) suggested that inter-basinal transfer of fishes was facilitated by (one or more of) a series of possible late-Quaternary headwater stream diversion points that he identified along this divide.
The disjunct distribution of F. coloradensis (as newly constituted herein) in the Snake River basin (Fig. 1) suggests additional complex aspects of biogeographic history. Most of these populations are distributed from the Snake River headwaters to just above the mouth of the Bruneau River where the ranges of F. coloradensis and F. fuscus abut. We are not aware of any sympatric occurrences of these sister species. The single, isolated occurrence of F. coloradensis in the middle Snake River basin (Crooked Creek), which is otherwise occupied only by F. fuscus (Fig. 1), may have resulted from translocation, as evidenced by the detection of only the most common COI (I) and cytb (III) haplotypes in this population. The occurrence of F. coloradensis in the Salmon River is intriguing given that the lower (and middle) Snake River basin is otherwise occupied only by F. fuscus. The close proximity of F. coloradensis populations on either side of the divide between the upper Salmon and upper Snake River watersheds suggests that the former may have been colonized as a result of a prior headwater transfer. Ruppel (1967) provides geologic evidence of late-Quaternary reversal of flow of the Lemhi River (now tributary to the Salmon River) consistent with this hypothesis.
Fluminicola coloradensis is currently ranked as imperiled or vulnerable (G2G3) by NatureServe (2011) and imperiled (G2) by the Idaho Department of Fish and Game (2005). Our finding that this species is much more widely distributed than previously thought suggests that it may not merit these rankings, at least on a rangewide basis. Conservation measures should perhaps be focused instead on geographic subunits of this species that may be at risk owing to local habitat degradation (i.e., the relatively small number of populations in the Bonneville and upper Green River basins).
ACKNOWLEDGMENTS
Susan E scher inked drawings and Yolanda Villacampa (Smithsonian Institution) prepared scanning electron micrographs and counted radula cusps. Peter Hovingh assisted in the field. We thank William H. Clark (ALBRCIDA) and Andrew Simons and Jonathan Slaght (UMBMNH) for loans of specimens under their care. We also thank Huyen Nguyen and Corbin Bradford for their assistance with the molecular labwork. The figures reproduced from Hershler and Frest (1996) are in the public domain. This project was supported by an award from the Idaho Department of Fish and Game (Agreement # T-3-7 0814).