INTRODUCTION

Small, isolated populations lose rare alleles through genetic drift, and such reductions in genetic diversity can make species more vulnerable to extinction because greater diversity increases adaptability and long-term population persistence (Reed and Frankham 2003; Hoffman et al. 2017). As a result, imperiled species face increased probabilities of extinction because small populations leave them vulnerable to the interacting effects of genetic drift, demographic change, and environmental stochasticity. Such species are at risk of entering a so-called extinction vortex (Gilpin and Soulé 1986). More than 70% of North American freshwater mussel species (families Unionidae and Margaritiferidae) are imperiled, and many survive only in small, isolated populations (Williams et al. 1993).

Lampsilis powellii (Lea, 1852), the Arkansas Fatmucket, is a federally threatened unionid mussel with a narrow historical distribution restricted to Interior Highlands portions of the Ouachita, Saline, and Caddo rivers in south-central Arkansas, USA (U.S. Fish and Wildlife Service 1990). Populations of upland fishes and mussels in many Interior Highlands streams are proposed to have been isolated from each other historically by long stretches of lowland habitat in the Mississippi Embayment, created when the lower reaches of these streams were buried under sediment deposited by high sea levels during interglacial periods (Mayden 1988; Haag 2010). The range of L. powellii has been fragmented further by dams, which destroyed suitable habitat for this species. Dams also restrict fish movement, which, in turn, restricts mussel dispersal and gene flow because mussel larvae are obligate parasites on fishes. Lampsilis powellii currently survives in only two isolated populations separated by impoundments, in the upper Ouachita and upper Saline river systems.

We examined population genetic structure and genetic variation of L. powellii from remaining populations using a 607-base-pair (bp) region of the mitochondrial COI gene and 14 nuclear microsatellite loci. Given historical and recent barriers to dispersal, we expected to observe reduced genetic diversity and limited connectivity among populations compared to more widespread mussel species. We identified patterns of genetic variation and population structure in L. powellii, which will be useful for assessing its conservation status and implementing population restoration efforts.

METHODS

We obtained swab samples from the mantle or foot of 42 individuals: 15 individuals from one site on the South Fork of the Ouachita River, and 27 individuals from four sites on the Saline River (Fig. 1). Genomic DNA was extracted using ArchivePure DNA Cell/Tissue kits (5 Prime, Gaithersburg, MD, USA). The mitochondrial cytochrome c oxidase subunit I (COI) gene was amplified with primers LCO22me2 and HCO700dy2 using conditions developed by Walker et al. (2006) following the specifications included with GoTaq Master Mix (Promega Corporation, Madison, WI, USA) in 20 µL reactions. Clean-up of polymerase chain reactions (PCRs) was performed with a QIAquick gel extraction (Qiagen, Inc., Germantown, MD, USA) with 35 µL of the PCR product. We used the above primers for cycle sequencing and carried out reactions using BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Inc., Waltham, MA, USA) with the default protocol. Cycle sequencing products were purified with the EDTA/sodium-acetate/ethanol protocol included with the BigDye kit and analyzed on an ABI Genetic Analyzer (Applied Biosystems).

We assembled, edited, and aligned sequences with Sequencher (Gene Codes Corporation, Ann Arbor, MI, USA) and verified an open reading frame and absence of primer sequences. DNASP v5.10 (Librado and Rozas 2009) was used to estimate population genetic indices from mtDNA sequences, including number of haplotypes (H), mean number of bp differences (K), and mean nucleotide diversity (π) over the pooled data set and within each river. A 95% confidence parsimony network was created in PopArt v1.7 (Leigh and Bryant 2015), and multiple connections between haplotypes were simplified using methods described by Fetzner and Crandall (2003).

Fourteen of 15 microsatellite primers designed for Lampsilis abrupta (Eackles and King 2002) were successfully optimized in L. powellii. Forward primers for each PCR were labeled with a 5′ fluorescent tag (6-FAM, NED, PET, or VIC) for visualization. We amplified microsatellite loci in 10 µL reactions using GoTaq Master Mix (Promega), 0.5 µM of fluorescently labeled forward and reverse primer, and 10 ng of DNA template carried out under the following conditions: initial denaturing at 94°C for 2 min; 35 cycles of 94°C for 40 s, variable annealing temperature for 40 s (Appendix A1), 72°C for 1 min; and a final extension at 72°C for 5 min. We performed fragment analysis on an ABI 3730 Genetic Analyzer with LIZ600 size standard (Applied Biosystems). PEAKSCANNER v1.0 (Applied Biosystems) was used to score alleles, and TANDEM v1.07 (Matschiner and Salzburger 2009) was used to assign integer numbers to DNA fragment sizes.

We estimated null allele frequencies using the Brookfield I (1996) method as calculated in MICROCHECKER (van Oosterhout et al. 2004). We used GenAlEx v6.3 (Peakall and Smouse 2006) to estimate mean number of observed alleles per locus (N_A) and number of private alleles (N_P) and to calculate observed and expected heterozygosities (H_O and H_E) for each population. Allelic richness (A_R) was computed via rarefaction using FSTAT v2.9.3 (Goudet 1995) to estimate the total number of alleles in the population given the sample size. GENEPOP v4.0.10 (Rousset 2008) was used to conduct exact tests of pairwise linkage disequilibrium and to test whether genotype frequencies met Hardy–Weinberg expectation.

Population genetic structure was visualized by conducting a principal coordinate analysis (PCoA) in GenAlEx using a covariance matrix created from all polymorphic loci. A permutational multivariate analysis of variance (PERMANOVA; n = 9999 permutations) was used to determine the significance of the PCoA clusters utilizing the adonis function within the vegan package (Oksanen et al. 2017) in R (R Core Team 2016). FreeNA was used to assess the degree of genetic differentiation among rivers by calculating pairwise F_ST values employing the ENA (excluding null alleles) correction, which has been shown to effectively correct for the positive bias of F_ST that may result from the presence of null alleles (Chapius and Estoup 2007).

RESULTS

We successfully sequenced mitochondrial DNA for 29 individuals and recovered nine COI haplotypes with a length of 607 bp. Sequences are available on GenBank (accession nos. MT762680–MT762708).

Three clusters of haplotypes were present, separated by 34–46 mutational steps (>5% of total sequence length; Fig. 2). The first cluster consisted of a common haplotype found in 13 individuals from the Saline River and seven individuals from the Ouachita River, two individuals of a second haplotype limited to the Saline River that differed from the common haplotype by two bp, and a singleton haplotype from the Ouachita River that differed from the common haplotype by four bp. Our most common haplotype was identical to a GenBank L. siliquoidea haplotype from the upper Mississippi River (accession no. MK672781.1; see Inoue et al. 2019).

A second cluster contained three singleton haplotypes from two sites in the Saline River that differed from each other by four or five bp. These haplotypes differed from those in the first cluster by 34–39 bp and corresponded to Lampsilis hydiana sequences on GenBank (accession no. MK672683.1).

Figure 1.

Sampling localities for Lampsilis powellii in Arkansas, USA.

Table 1.

Summary statistics for COI sequences and 14 microsatellite loci for Lampsilis powellii in the Ouachita and Saline rivers, Arkansas, USA. Individuals identified as Lampsilis hydiana and Actinonaias ligamentina based on COI sequences are not included (see text). N = number of individuals sampled. Genetic diversity metrics are number of haplotypes (H), mean number of bp differences between all pairs of individuals (K), nucleotide diversity (π), mean number of observed alleles (N_A), rarefied allelic richness (A_R), mean number of private alleles (N_P), mean observed heterozygosity (H_O), and mean expected heterozygosity (H_E).

A third cluster contained three singleton haplotypes from the Saline River that differed from each other by six to nine bp. These haplotypes differed from those in the first cluster by 43–46 bp and were most similar to GenBank sequences for Actinonaias ligamentina (accession no. MK672757.1). We omitted from further analysis the six individuals in clusters two and three representing L. hydiana and A. ligamentina.

There was little genetic diversity among the 23 remaining COI sequences representing L. powellii, and patterns were similar for both rivers (Table 1). There were two haplotypes in each river, the mean number of base-pair differences between all individuals was three, and nucleotide diversity was low.

Microsatellite analyses also showed low genetic variation. We scored a total of 172 alleles over the 14 loci examined, ranging from four to 19 per locus (Tables 1 and 2). Microsatellite genotypes in GENEPOP format are deposited in Dryad ( https://doi.org/10.5061/dryad.3r2280gd6). Mean rarefied allelic richness and heterozygosity across the 14 loci were similar between rivers (Table 1). The mean number of private alleles was twice as high in the Saline than in the Ouachita. We found considerable evidence of null alleles (present at 82.1% of all river-by-locus pairs) and no evidence of linkage disequilibrium. Eighteen of 28 locus-by-river combinations deviated from Hardy–Weinberg expectations after Bonferroni correction, with all showing a heterozygote deficiency consistent with the presence of null alleles.

Table 2.

Genetic diversity at 14 microsatellite loci in Lampsilis powellii from the Ouachita and Saline rivers, Arkansas, USA. N = number of individuals genotyped. Genetic diversity metrics are allelic richness (N_A), observed and expected heterozygosity (H_O and H_E), and null allele frequency. * indicates deviation of heterozygosity from Hardy–Weinberg expectation after sequential Bonferroni correction. Bold values indicate null allele frequencies that are significantly different from zero at P ≥ 0.05.

Figure 2.

Haplotype network for samples identified as Lampsilis powellii derived from 607 bp region of the mitochondrial cytochrome c oxidase subunit I (COI) gene. Each circle represents a unique haplotype. The number within each circle represents the number of individuals having that haplotype; circles without numbers represent a single individual. Black dots and tick marks represent missing haplotypes.

Populations of L. powellii showed little evidence of significant divergence. The first PCoA axis separated the two rivers but explained only 11.6% of total microsatellite variation (Fig. 3a). The second and third axes accounted for 20.5% and 27.3% of the variation, respectively; however, we could not infer any geographical patterns from these axes (Fig. 3b). The ENA-corrected estimate of pairwise F_ST among rivers was 0.076, while the PERMANOVA failed to reject the null hypothesis of no difference between rivers (F = 1.72, P = 0.11). The uncorrected estimate of F_ST (0.085) was similar to the ENA-corrected estimate; its slightly higher value was likely due to the presence of null alleles.

Figure 3.

Results of principal coordinates analysis (PCoA) for 14 microsatellite loci of Lampsilis powellii from Arkansas, USA. Red squares represent individuals from the Saline River, and blue circles represent Ouachita River individuals.

DISCUSSION

Lampsilis powellii seems to have maintained only a limited amount of intraspecific genetic diversity at both mtDNA and nDNA loci compared with other mussel taxa (Elderkin et al. 2007, 2008; Jones 2009). This limited amount of genetic diversity may be a result of the small sample sizes in our study, which is a common issue with rare species. However, the number of mtDNA haplotypes we found was lower than the number found in other endangered mussel species, even though samples sizes were comparable (Zanatta and Murphy 2008; Jones 2009; Menon et al. 2019). In the Ouachita River, Cumberlandia monodonta showed similarly low levels of mtDNA diversity with the presence of a single COI haplotype (Inoue et al. 2014b). Additionally, genetic diversity in L. powellii was lower than in other unionid studies that employed the Eackles and King (2002) microsatellite primers (Kelly and Rhymer 2005; Menon et al. 2019). For example, genetic diversity was lower in L. powellii than in Great Lakes populations of L. cardium (N_A range: 3.3–13.7; H_O range: 0.42–0.85; Hewitt et al. 2019). Additionally, mean allelic richness was lower in L. powellii than for L. siliquoidea in the St. Clair River (A_R = 8.7–11.4; Rowe and Zanatta 2015). The latter two comparisons are noteworthy because those populations are geologically young and colonized the Great Lakes region only after Pleistocene glaciation. Mussel populations in the Interior Highlands are likely much older and would be expected to contain higher genetic diversity. The low genetic diversity of L. powellii may reflect its small population size due to its long history of isolation compounded by anthropogenic habitat destruction and fragmentation.

The Interior Highlands of North America is a biodiversity hotspot with high endemicity for a variety of aquatic organisms including mussels, fishes, crayfishes, and salamanders (Mayden 1985; Haag 2010). The region's distinctive aquatic fauna is believed to have been formed by vicariant pre-Pleistocene events, followed by repeated periods of isolation due to Pleistocene glaciations (Mayden 1988), resulting in genetically divergent populations. Not only do isolated populations with restricted gene flow lose genetic diversity, they also can exhibit an increase in population structuring (Paetkau et al. 1995). However, the limited genetic divergence between rivers that we found suggests a single panmictic population in relatively recent times. Despite the presence of the potential dispersal barrier of lowland habitats in the lower reaches of the Saline and Ouachita rivers, a mobile fish host may have allowed some dispersal across this barrier (e.g., Galbraith et al. 2015). Host fishes for L. powellii are currently unknown, but most Lampsilis are specialists on black basses (Micropterus), which can undertake substantial movements (Haag 2012; Schall et al. 2019). Identification of fish hosts of L. powellii is needed for inferring its potential dispersal ability and gene flow.

Interestingly, we were unable to differentiate L. powellii from L. siliquoidea COI sequences from the Mississippi River (Inoue et al. 2019). Whole mitochondrial genome sequencing identified a divergent protein-coding region, the f-orf gene, which distinguished between L. siliquoidea and L. powellii (Robicheau et al. 2018). However, when we examined the COI region of the whole mtDNA genome submitted to GenBank by Robicheau et al. (2018; accession no. NC_037720) and a partial COI gene sequence (Breton et al. 2011; accession no. HM849705) labeled L. powellii, they matched sequences identified as L. hydiana from another study (Inoue et al. 2019). At this time, we are unable to make conclusions about the taxonomic distinctiveness of L. powellii. Because the available genetic data are limited to mtDNA sequences and a few microsatellite loci, a high-resolution genomic approach may be necessary to elucidate the taxonomic position of L. powellii among other Lampsilis species from this region.

Lampsilis siliquoidea and L. powellii are difficult to distinguish morphologically (Harris et al. 2004; Harris et al. 2010; Krebs et al. 2013). Geometric morphometric analyses can be used to differentiate morphologically similar species (Inoue et al. 2014a; Riccardi et al. 2019), and this technique also may be useful for differentiating L. powellii from other co-occurring lampsilines from the Interior Highlands. The presence in our dataset of samples referable to A. ligamentina and L. hydiana from individuals originally identified in the field as L. powellii underscores the difficulty of identifying lampsiline species in this region of high evolutionary diversification.

Our results have important conservation implications. The lack of a clear genetic signal for differentiating L. powellii, L. siliquoidea, and L. hydiana shows that further research is necessary to evaluate the phylogenetic status of these species. Our results suggest that between-river genetic divergence in L. powellii is very low, although both populations contain private microsatellite alleles. However, our small samples sizes make this conclusion preliminary; population genomic methods are likely necessary to determine whether significant between-river divergence exists for L. powellii. Resolution of these issues is critical for designing conservation strategies for the small populations of L. powellii that remain in the Saline and upper Ouachita basins.