Guiding Species Recovery through Assessment of Spatial and Temporal Population Genetic Structure of Two Critically Endangered Freshwater Mussel Species (Bivalvia: Unionidae)


 The Cumberlandian Combshell (Epioblasma brevidens) and Oyster Mussel (E. capsaeformis) are critically endangered freshwater mussel species native to the Tennessee and Cumberland River drainages, major tributaries of the Ohio River in the eastern United States. The Clinch River in northeastern Tennessee (TN) and southwestern Virginia (VA) harbors the only remaining stronghold population for either species, containing tens of thousands of individuals per species; however, a few smaller populations are still extant in other rivers. We collected and analyzed genetic data to assist with population restoration and recovery planning for both species. We used an 888 base-pair sequence of the mitochondrial NADH dehydrogenase 1 (ND1) gene and ten nuclear DNA microsatellite loci to assess patterns of genetic differentiation and diversity in populations at small and large spatial scales, and at a 9-year (2004 to 2013) temporal scale, which showed how quickly these populations can diverge from each other in a short time period. Intraspecific mitochondrial DNA (mtDNA) and microsatellite DNA variation was higher in E. capsaeformis than in E. brevidens. These two species have maintained quite different levels of genetic diversity within their Clinch River stronghold and in their smaller peripheral populations in the Big South Fork Cumberland and Nolichucky rivers, TN. For instance, with only three mtDNA haplotypes detected overall across populations, E. brevidens’ capacity for maintaining genetic diversity appears to be less than that of E. capsaeformis, which had 18 haplotypes. At the relatively small spatial scales (15-30 kilometers) investigated in the Clinch River, demes of both species exhibited minimal genetic differentiation in either the 2004 or 2013 sampling periods, typically <0.02 based on FST and <0.1 based on Jost’s D. Our genetic data suggest that mussels at the numerous shoals in a 32-kilometer section of the Clinch River comprise a single, large population of each respective species with very high gene-flow among individual demes. However, we also observed a high level of genetic differentiation among demes at the 9-year temporal scale, with differentiation metrics for E. brevidens (D = 0.47 and FST = 0.12) and E. capsaeformis (D = 0.31 and FST = 0.05) proving higher than the within-year values. This result strongly suggests that genetic drift is playing an important role in allele frequency change over time in these populations. At the spatial and temporal scales investigated in this study, various demographic, life history, and environmental factors are influencing maintenance of genetic variation and need to be considered during conservation planning for each species.


Introduction
Habitat fragmentation and restricted gene ow often lead to a loss of genetic diversity and inbreeding within fragmented populations (Frankham et al. 2010). For imperiled species, this may impact individual tness and increase the probability of extinction of a species. Combined analysis of spatial and temporal variation during the investigation of population genetic structure enhances our ability to understand opposing forces of gene ow and genetic drift, while weighing the impacts of population fragmentation and facilitating gene ow through human-mediated conservation practices. In this study, we investigated the spatial and temporal genetic structure of the Cumberlandian Combshell Epioblasma brevidens (Lea, 1831) and Oyster Mussel E. capsaeformis (Lea, 1834), two critically endangered freshwater mussel species (U.S. Fish and Wildlife Service 2004) native to the Tennessee and Cumberland River drainages, major tributaries of the Ohio River in the eastern United States those at three sites or demes in the Clinch River, TN [Wallen Bend (WB), Frost Ford (FF), Swan Island (SI)] and for the global sampled population (i.e., data combined for all three demes within each species), in the Clinch River's largest tributary, the Powell River, and in the Big South Fork Cumberland River (Table 1). The ND1 DNA sequences from WB, FF, and SI in 2004 were obtained from Jones et al. (2015). Parameter estimates and associated variances were calculated using DNAsp 6.12.03X64 software (Rozas et al., 2018). Mean uncorrected genetic distance (D xy ) between populations was calculated in DnaSP, and genetic differentiation (F ST ) between populations was calculated in the program ARLEQUIN, version 3.0 (Exco er et al. 2005).

DNA microsatellites Analyses
Microsatellite loci and primers were developed and characterized using DNA of E. capsaeformis (Jones et al., 2004) and L. abrupta (Eackles & King, 2002) and were screened in all sampled individuals using a subset of the loci available in these two primer sets (Table 2). The DNA microsatellite genotypes from individuals collected at WB, FF, SI in 2004 were obtained from Jones et al. (2015). PCR ampli cation protocols followed Eackles & King (2002) and consisted of 100 ng of genomic DNA, 1x PCR buffer, 2 mM MgCl 2 , 0.25 mM dNTPs, 0.5 mM each primer, and 1.0 U AmpliTaq DNA polymerase (Perkin-Elmer Applied Biosystems, Inc., Foster City, CA) in a total volume of 20 mL. PCR thermal cycling conditions were: 94 °C for 2 min; followed by 35 cycles of 94 °C for 40 sec, 58 °C for 40 sec, and 72 °C for 1 min; followed by a nal extension at 72 °C for 1 min; and a hold at 4 °C (Eackles & King, 2002).
Ampli cation products containing microsatellite loci were examined for size polymorphism using an ABI 3100 automated sequencer and GENOTYPER (ABI) software to determine allele size. Results were stored as GENESCAN les; GENEMAPPER software (ABI) was used to visualize allele size and score the results. Microsatellite data sets were tested for genotyping errors caused by null alleles, stuttering, and short allele dominance using a Monte Carlo simulation of expected allele size differences using MICROCHECKER (Van Oosterhout et al. 2004). Populations were screened for linkage disequilibrium (LD) between all pairs of loci and for deviations from Hardy-Weinberg equilibrium (HWE) at each locus in ARLEQUIN. Signi cance of LD pairwise tests was determined using the likelihood-ratio test with 10,000 permutations (Slatkin and Exco er 1996), and HWE using the exact test with a Markov chain of 1,000,000 steps and 100,000 dememorisation steps (Guo and Thompson 1992). We used a sequential Bonferroni correction to account for Type I errors associated with all multiple pair-wise comparisons (Rice 1989).
Genetic variability across microsatellite loci for populations of E. brevidens and E. capsaeformis was quanti ed in terms of percentage of polymorphic loci, observed heterozygosity, expected heterozygosity, number of alleles per locus, total number of alleles, number of private alleles, and allele frequencies per locus. Evidence for a bottleneck in each population at each locus was tested using the Garza-Williams index (M-ratio) as implemented in ARLEQUIN, which is the ratio of the number of alleles to those possible within the observed range in allele size; values below 0.7 suggest the occurrence of a bottleneck (Garza and Williamson 2001). We used program BOTTLENECK, which tests for an excess of heterozygosity as a consequence of a genetic bottleneck to test for a recent bottleneck in each population (Cornuet and Luikart 1996;Piry et al. 1999). The Wilcoxon's sign-rank test was used to test for signi cant excess of heterozygosity in population samples across loci, which is interpreted as evidence of a recent bottleneck. This is a one-tailed test and is considered the most powerful of all similar tests available in program Bottleneck when the number of polymorphic loci tested is low (<20) (Cornuet and Luikart 1996). The Wilcoxon's sign-rank test was conducted using the two-phase mutation model (TPM), which is considered the best mutation model for DNA microsatellites. The TPM was implemented with the variance set to 12 and the proportion of stepwise mutations set to 90 (Piry et al. 1999;Garza and Williamson 2001). Evidence for inbreeding within each population was tested for using F IS , which measures the extent of departure from HWE due to inbreeding within a subpopulation(s) and can range from -1.0 (all heterozygotes) to +1.0 (all homozygotes) (Wright 1978).
Population differentiation was quanti ed using the F ST metric in ARLEQUIN, which ranges in value from 0 (no differentiation) to 1 (complete differentiation) (Wright 1978;Balloux & Lugon-Moulin 2002). However, because F ST will approach zero when gene diversity is high, D, -an estimator of actual differentiation developed by Jost (2008) -was estimated using GenAlEx v.6.5 (Peakall and Smouse 2006). This metric also ranges from 0 to 1, but provides a more accurate estimate of differentiation due to genetic drift and gene ow at highly polymorphic loci. The program POPTREE2 (Takezaki et al. 2010) was used to construct phylogenetic trees from our DNA microsatellite allele frequency data by using the neighbor-joining (NJ) method (Saitou and Nei 1987) and Nei's standard genetic distance (D ST ) (Nei 1972), utilizing a total of 10,000 bootstrap replications.
We used program STRUCTURE version 2.3.4 (Pritchard et al. 2000) to infer the most likely number of genetically distinct, multilocus genotypic clusters (~populations) and to assign individual genotypes to such populations. An admixture ancestry model with correlated allele frequencies was used to determine population structure of E. brevidens and E. capsaeformis individuals. No priors or population information were used in the analysis. The best-supported number of populations (K) was assessed by calculating the mean natural logarithm of the probability of K [mean LnP(K)] (Pritchard and Wen 2003) across iterations from K = 2-9 and with ten independent runs for each K. The largest mean LnP(K) is considered the optimal value of K. All runs consisted of a burn-in of 10,000 steps, followed by 100,000 iterations. Finally, runs generating the highest likelihood estimates were imported to the program STRUCTURE PLOT at http://btismysore.in/strplot/index.php (Ramasamy et al. 2014) to create a visualization of population structure.
Contemporary effective population sizes (N e ) were estimated for each population using the linkage disequilibrium (LD) method of Hill (1981). The method is known to be downwardly biased, but the program LDNe corrects the bias and was used to estimate N e (Waples 2006;Waples and Do 2007).

Results
Genetic variation of mitochondrial DNA sequences Intraspeci c mitochondrial DNA sequence variation at the ND1 gene generally was higher in E. capsaeformis than in E. brevidens (Table 3). The number of polymorphic nucleotides was higher in E. capsaeformis (n = 22) than in E. brevidens (n = 4), and a total of 18 haplotypes were observed among all E. capsaeformis populations, compared to just three haplotypes observed among all E. brevidens populations (Figure 3, Panels A and B; Table 4). Further, none of the three E. brevidens haplotypes were unique to the study samples; they were shared among the Big South Fork Cumberland River, Clinch River and Powell River populations, and shared between the 2004 and 2013 sampling periods. However, the Big South Fork Cumberland River sample only contained one haplotype (Ebrev2), and it was the same haplotype observed in the Clinch and Powell river samples (Table 4). In contrast, for E. capsaeformis, eleven of the thirteen haplotypes observed in the Clinch River population and ve of the seven haplotypes observed in the Nolichucky River population were unique. Two haplotypes (Ecap1 and Ecap5) were shared between the two populations, but at different proportions in the respective samples. Several haplotypes were unique to the three demes in the Clinch River (WB, FF, SI) and between the 2004 and 2013 sampling periods (Table 4).
To summarize the genetic diversity metrics for both species, global haplotype diversity for E. brevidens was h = 0.59, but ranged from a low of h = 0 in the Big South Fork Cumberland River sample to a high of h = 0.66 in the Clinch River sample at Frost Ford in 2004 (Table 3). Global nucleotide diversity was p = 0.0018, with p in population samples ranging from 0 to 0.0017. Mean k was 1.57 nucleotide differences among sequences, ranging from 0 to 1.82. Global haplotype diversity for E. capsaeformis was h = 0.68, but ranged from a low of h = 0.49 in the Clinch River sample at WB in 2013 to a high of h = 0.90 in the Nolichucky River sample in 2013 (Table 3). Global nucleotide diversity was p = 0.0024, with p in population samples ranging from 0.0017 to 0.0040. Mean k was 2.09 nucleotide differences among sequences, ranging from 1.54 to 3.59.
Mitochondrial DNA genetic distance and differentiation among populations  (Table 5). Similarly, pairwise F ST comparisons involving the Nolichucky River (range = 0.23 -0.39) with other populations were the highest and signi cantly different, whereas F ST among demes and years in the Clinch River were lower and mostly non-signi cant. However, pairwise comparisons involving WB (2013) and the other Clinch River demes were higher and signi cantly different.
Genetic variation of nuclear DNA microsatellites Genetic diversity at microsatellite loci generally was lower in E. brevidens than in E. capsaeformis (Table 6) After the threshold for the signi cance of p-values for the respective locus pairs was Bonferroni corrected (p < 0.001; a = 0.05), none of the E. brevidens populations in either the 2004 or 2013 samples contained loci with signi cant deviations from linkage equilibrium. Signi cant deviations (p < 0.001; a = 0.05) from Hardy-Weinberg equilibrium were observed at Ecap01, Ecap06, Ecap08, and Ecap09, but the loci in disequilibria were distributed randomly among 1-3 loci per population. Further, no evidence was found for genotyping errors or large-allele drop-out, although increased homozygosity at some loci suggested the possible segregation of null alleles, or inbreeding due to small population size and hermaphroditic reproduction (van der Schalie 1970). Each population contained a small number of private alleles not detected in other sampled populations, with 1-6 private alleles observed per population; however, the Big South Fork Cumberland River population contained a high number of 19 private alleles (Table 6; see Appendix).
After the threshold for signi cance of p-values for the respective locus pairs was Bonferroni corrected (p < 0.001; a = 0.05), none of the E. capsaeformis populations in either the 2004 or 2013 samples contained loci with signi cant deviations from linkage equilibrium. Signi cant deviations (p < 0.001; a = 0.05) from Hardy-Weinberg equilibrium were observed at Ecap01, Ecap02,Ecap04, Ecap06, Ecap08, and Ecap09, Lab206 and Lab213, but the loci in disequilibria were distributed randomly among a small number of loci per population. Further, no evidence was found for genotyping errors or large-allele drop-out, although increased homozygosity at some loci suggested the possible segregation of null alleles, or inbreeding due to small population size and hermaphroditic reproduction (van der Schalie 1970). Each population contained a small number of private alleles not detected at other sampled populations, with 2-6 private alleles observed per population; with the Nolichucky River population containing the highest number of private alleles, 7 (Table 6).

Genetic evidence for bottlenecks and inbreeding in populations
For E. brevidens, mean observed M-ratio values per population (Table 6) Table 6). Mean F IS values per population ranged from 0.05 at Swan Island in the Clinch River to 0.27 in the Big South Fork Cumberland River, indicating a low to moderate level of inbreeding in each population (Table 6).
For E. capsaeformis, mean observed M-ratio values per population (Table 5)

Discussion
In this study, we assessed patterns of genetic diversity and differentiation in populations of Epioblasma brevidens and E. capsaeformis at both small and large spatial scales; for the rst time with freshwater mussels, we assessed diversity and differentiation at a temporal scale, showing how quickly these populations can diverge from each other over a relatively short time period of just 9 years or approximately one generation for either species. These two species have maintained quite different levels of genetic diversity within their Clinch River stronghold, and also in much smaller peripheral populations in the Big South Fork Cumberland and Nolichucky rivers. For instance, with only three mtDNA haplotypes detected overall across populations, E. brevidens' capacity for maintaining genetic diversity appears to be less than that of E. capsaeformis. Hence, the spatial and temporal scales, demographic, life history, and environmental factors in uencing maintenance of genetic variation for these two species require more in-depth discussion.
At the relatively small spatial scales (15-30 kilometers) investigated within the Clinch River, the three demes exhibited minimal genetic differentiation for both species. While some statistically signi cant differences were observed between demes in 2004 and 2013, differentiation levels were very low, typically <0.02 based on F ST and <0.1 based on Jost's D for both species. The river is free-owing in Hancock County, TN, and therefore no barriers exist among demes that would prevent dispersal of host shes, downstream dispersal of adults or downstream drift of sperm during spawning. Thus, our genetic data suggests that at the dozen or so major shoal areas in this reach of the river (RKM 276.8-309.0), which collectively contain tens of thousands of individuals of E. brevidens and E. capsaeformis (Jones et al. 2018;Lane et al. 2021), would best be described as containing a single large functional population of each respective species, with very high gene-ow among the individual demes.
However, at a larger spatial scale, a much different pattern of genetic diversity and differentiation emerges. For E. brevidens, the Big South Fork Cumberland River population appears devoid of mtDNA diversity at the ND1 gene, and as a whole range-wide, the species appears to be low in variation at the mitochondrial genome, especially given that the Clinch River population is its last stronghold and mtDNA variation is low even there. This result is expected, given the anthropogenic fragmentation caused by construction of dams, impounded back-waters, and cold-water temperature zones below hypolimneticreleasing reservoirs. The ND1 haplotypes sampled in the Big South Fork Cumberland and Powell rivers were identical to those in the Clinch River, suggesting that mtDNA diversity is low and similar across the species' range in the Cumberland and Tennessee river valleys. However, at the mtDNA cytochrome-b gene, a 1-bp xed difference was observed between samples in the Big South Fork Cumberland and Clinch rivers (Johnson et al. 2006), indicating that some unique variation does exist even in the small remaining populations. Thus, assessing genetic diversity of the Buck Creek, KY and Bear Creek, AL and MS populations of E. brevidens would help test this interpretation. Genetic differentiation at nuclear DNA microsatellites among the Big South Fork Cumberland, Clinch and Powell river populations was very high based on D and moderate to high based on F ST , which is not surprising given that the remaining populations are small and bottlenecked and separated by hundreds of river miles and numerous barriers, including large hydro-power dams. While observed mtDNA variation was not unique among E. brevidens populations, the Big South Fork Cumberland River population contained numerous private microsatellite alleles, suggesting that this population has unique or novel genetic variation when compared to the Clinch River population for example, which is expected given the geographic distance (1,200 river kilometers) separating the two populations. For E. capsaeformis, the Nolichucky River population, while small, still retains relatively high mtDNA haplotype diversity. Our mtDNA sample size was small (N = 13) but a total of seven haplotypes, ve of which were unique, were observed in this small peripheral population. Given our small sample size, but high observed haplotype diversity, it is likely that additional haplotypes might be discovered in this population with additional sampling. However, with eighteen total haplotypes observed-thirteen unique to the Clinch River and ve unique to the Nolichucky River-global haplotype diversity is much higher in E. capsaeformis than in E. brevidens. Hence, two interpretations merit discussion here. First, the Nolichucky River may have supported a much larger population of E. capsaeformis in the not-too-distant past than previously thought. There are more than a dozen major island complexes located throughout a ~74 kilometer reach (RKM 19.6−93.8) of the river, each with signi cant shoal habitat for mussels (T. Lane, unpublished data). Nearly all of these sites are larger (>20,000 m 2 ) than the largest sites in the Clinch River, TN for example. Hence, the amount of physical habitat that is available in the river is more than enough to support a large population. Second, the capacity of E. capsaeformis to maintain genetic diversity seems to exceed that of E. brevidens. In the Clinch River, TN, where both species co-occur at all sites, E. capsaeformis genetic diversity is higher, where presumably both species have been exposed to the same environmental and historical conditions. Genetic differentiation at nuclear DNA microsatellites between the Clinch and Nolichucky river populations of E. capsaeformis was high based on D and moderate to low based on F ST , although these metrics were still much higher than the among-deme comparisons in the Clinch River, which again were very low. To reiterate, this increase in differentiation at a larger spatial scale is not surprising given that these two populations also are separated by hundreds of river miles and numerous large hydro-power dams.
Demographic and life-history factors, such as changes in population size, longevity, and contrasting sh host life-histories, offer the best explanations for why genetic diversity was greater in E. capsaeformis than in E. brevidens. The census population size of E. capsaeformis is much larger than that of E. brevidens in the TN section of the Clinch River (Figure 8). From 2004-2014, populations of both species were censused annually at Swan Island, Frost Ford, and Wallen Bend, which showed that across these three sites, combined mean total population size over the 10-year period was ~12,000 individuals of E. brevidens versus 285,000 individuals of E. capsaeformis (Jones and Neves 2011;Jones et al. 2014;Jones et al. 2018;Lane et al. 2021 in review). Further, uctuations in population size were much greater for E. capsaeformis relative to E. brevidens, with population size increasing ve-fold for E. capsaeformis from 2007-2009 for example, then decreasing signi cantly over the following years, whereas E. brevidens exhibited a much more stable population trend with only modest, non-signi cant uctuations in size over this same period (Figure 8). Hence, the E. capsaeformis population seems to have a greater inherent capacity to grow quickly and maintain a larger population size, allowing it to maintain more genetic variation, or stated differently, E. brevidens which has a much smaller population size making it more susceptible to losing genetic variation through random genetic drift. However, E. capsaeformis is a relatively short-lived species (maximum age of at least 12 years), and thus despite any advantages it may have in population growth rate and size under seemingly favorable conditions, E. brevidens is a longer-lived species (maximum age at least 28 years) and is better at persisting under unfavorable conditions (Jones and Neves 2011). For example, small recruiting populations of E. brevidens have persisted in the Big South Fork Cumberland and Powell rivers despite decades of impacts from coal mining, whereas E. capsaeformis was extirpated decades ago from both rivers (Johnson et al. 2012;Ahlstedt et al. 2016;Zipper et al. 2016).
Likely one of the biggest drivers of differences in population size between these two mussel species is the relative abundance and specialization in use of their primary sh hosts. In the Clinch River, E. brevidens uses two large, relatively mobile darters (Percidae), the blotchside logperch (Percina burtoni) and common logperch (P. caprodes), as its primary hosts (Yeager and Saylor 1995). Both sh species are typically 120-130 mm in total length, considered rare to uncommon in the river, and have a sturdy cartilaginous snout that it uses to ip small stones to hunt for aquatic insect prey items (Jenkins and Burkhead 1993). Importantly, the snout and skull are su ciently strong to withstand capture and holding during infestation of glochidia by females of E. brevidens (Barnhart et al. 2008). In contrast, E. capsaeformis uses smaller, presumably less mobile darters in the genus Etheostoma, such as Redline Darter (E. ru lineatum) and Bluebreast Darter (E. camurum) as primary hosts (Yeager and Saylor 1995). These sh species are typically 50-60 mm in total length, and are considered common to abundant in the river (Jenkins and Burkhead 1993). Females of E. capsaeformis utilize a cushioned mantle-pad to cradle these small darters to minimize harm to them when they are being infested with glochidia Barnhart et al. 2008;Jones et al. 2020).
The sh hosts of E. brevidens likely number in the dozens per site, whereas the sh hosts of E. capsaeformis likely number in the thousands per site, which allows this species to have much greater contact with its host sh and ultimately the capacity for reproduction to build up large local populations.
Perhaps the most interesting nding in our study was the high level of genetic differentiation observed among the demes in the Clinch River, TN between the 2004 and 2013 sample periods. Our within-year data clearly show that differentiation was very low and thus high gene-ow existed among demes, indicating that the mussels inhabiting the numerous habitat patches in this section of river are collectively acting together as a single large population. This low within-year differentiation also speak to the adequacy of our sample sizes for the demes. The 2004 and 2013 STRUCTURE analyses for each species showed that individuals from each deme shared essentially the same overall genetic ancestry and thus comprised single, large populations, respectively ( Figure 6). However, at a 9-yr. temporal scale, genetic differentiation increased greatly for both species, with differentiation values between 2004 and 2013 for E. brevidens (D = 0.47 and F ST = 0.12, Figure 4 Panel A) and E. capsaeformis (D = 0.31 and F ST = 0.05, Figure 4 Panel B) increasing greatly relative to the within-year values. This result strongly suggests that random genetic drift is playing a big role in driving allele frequency change over time in these two populations; further, the numerous individual demes within this reach of river are not drifting apart from each other, but rather they are staying interconnected through high gene-ow, but drifting genetically together in unison over time as a single, large population. In addition, within-year differentiation was slightly less among demes for E. brevidens compared to E. capsaeformis, suggesting higher geneow, perhaps due to its more mobile sh hosts; further, differentiation between years was much greater (nearly double) among demes for E. brevidens than for E. capsaeformis, suggesting that the smaller population size of this species is increasing the rate of random genetic drift. In contrast, corresponding increases in mtDNAbased genetic differentiation between 2004 and 2013 for E. brevidens (F ST = 0.012) and E. capsaeformis (F ST = 0.008) within the Clinch River were not observed, but rather differentiation was elevated geographically between the Clinch River populations and the small peripheral populations of each species (Table   5). The Clinch River population (all three sites combined) of E. brevidens changed very little with regard to haplotype frequency over the period from 2004-2013 (Figure 3, Panel A). However, the Clinch River population of E. capsaeformis showed noticeable changes in haplotype frequencies over the same time period, with some previously observed haplotypes absent in 2013 and several new haplotypes now frequently observed relative to the overall samples (Figure 3. Panel B).
Our estimates of N e in 2004 and 2013 were convergent with the census size (N c ) estimates made by Jones et al. (2014Jones et al. ( , 2018 in two important ways. First, values of N e and N c at all three demes were lower for E. brevidens than for E. capsaeformis, and second, estimates of population size were much less variable for E. brevidens relative to E. capsaeformis (Figures 7 and 8). Since our data showed minimal differentiation and high gene-ow among demes for both species, we combined the microsatellite DNA genotype data from all three demes to estimate N e ( Figure 5). By increasing the sample size, estimates of N e are expected to improve, i.e., to become more accurate and precise (Waples and Gagiotti 2006). Hence, the combined data indicate that N e was very stable over time and much lower for E. brevidens, whereas for E. capsaeformisN e was larger and more variable, and increased substantially from 2004 and 2013. Again, this pattern of lower N c but stable N e for E. brevidens likely is driven by the species' greater longevity and its less-common host shes, whereas for E. capsaeformis both N c and N e are larger but more variable, and likely tied to the species' short lifespan and its abundant host shes. Interestingly though, the N e /N c ratios were quite similar, essentially 10% in 2004 and 7% in 2013 for both species ( Figure 5). While our N e con dence intervals are large, we believe these estimated ratios should best be viewed as upper-end values, as the total N c used to calculate them were derived by adding the N c values only from Wallen Bend, Frost Ford, and Swan Island. For example, Jones and Neves (2011)  individuals for E. capsaeformis. Using these census values to estimate N e /N c ratios for 2004, for example, would be 1.3% for both species; when considered together with estimates derived from just the three sites, each species' respective N e /N c ratio ranges from 1-10% in the Clinch River, TN. An understanding of this ratio is important because it gives insight into the genetic and demographic viability of a population. Other invertebrate species (e.g., Paci c oysters) that broadcast-spawn have very low N e /N c ratios (<1%), meaning that they need large census sizes to support a breeding population capable of maintaining genetic variation over time (Hedgecock and Pudovkin 2011). Further, understanding the N e /N c ratio allows estimation of N c from N e when census data are unavailable. For instance, applying a N e /N c range of 1-10% to the Nolichucky River N e of 224 individuals of E. capsaeformis gives the N c estimate of 2,240-22,400 individuals, and to the Big South Fork Cumberland River N e of 95 individuals of E. brevidens gives the N c estimate of 950-9,500 individuals. Even though neither population has been quantitatively surveyed, these rough estimates of N c are likely to be useful to conservation planners, especially when they consider the long-term adaptive potential of focal populations.
Due to various anthropogenic changes to stream hydrology and pollutant stressors over the last 100 years, the range and abundance of E. brevidens and E. capsaeformis have contracted by >90%, causing both species to become critically endangered (USFWS 2004 (Hubbs 2020). Strategies to maintain and enhance the genetic representation of the native population of E. capsaeformis in the lower Nolichucky River should be given high priority. This small native population contains unique genetic variation for the species, and managers need to be careful that it is not lost due to further population decline or from genetic swamping from overstocking with Clinch River animals. Because the Clinch and Nolichucky River populations are strongly differentiated at nuclear and mitochondrial DNA markers, additional investigation of these populations using nuclear single nucleotide polymorphism (SNP) markers may help elucidate the phylogenetic depth of this differentiation, and further inform conservation planning for the species. Currently, the restoration site containing Clinch River individuals and the native Nolichucky River population are separated by more than 15 river kilometers, but in time mixing may occur. Hence, for now, they can be managed separately to maximize genetic representation. With so much genetic diversity still maintained in the Nolichucky River population, managers should consider using the native population as a source for broodstock alongside the Clinch River stock to further diversify augmentation of upstream and other recipient populations. Finally, consideration should be given to assessing the genetic diversity of restored populations of each species to see whether they are maintaining diversity levels similar to that in the lower Clinch River, TN populations, and to assessing growth, survival and phenotypic variation of the mantle-lure displays at the population level for both species.