Population genetics of Bull Trout (Salvelinus confluentus) in the upper Athabasca River basin

Abstract Freshwater ecosystems are negatively impacted by a variety of anthropogenic stressors, with concomitant elevated rates of population decline for freshwater aquatic vertebrates. Because reductions in population size and extent can negatively impact genetic diversity and gene flow, which are vital for sustained local adaptation, it is important to measure these characteristics in threatened species that may yet be rescued from extinction. Across its native range, Bull Trout (Salvelinus confluentus) extent and abundance are in decline due to historic overharvest, invasive non‐native species, and habitat loss. In Alberta's Eastern Slope region, populations at the range margin have progressively been lost, motivating us to better understand the amount and distribution of genetic variation in headwater habitats and some downstream sites where they continue to persist. Across this region, we sampled 431 Bull Trout from 20 sites in the Athabasca and Saskatchewan River basins and assayed 10 microsatellite loci to characterize within‐ and among‐population genetic variation. The Saskatchewan and Athabasca River basins contained similar levels of heterozygosity but were differentiated from one another. Within the Athabasca River basin, five genetically differentiated clusters were found. Despite the evidence for genetic differentiation, we did not observe significant isolation‐by‐distance patterns among these sites. Our findings of ample genetic diversity and no evidence for hybridization with non‐native Brook Trout in headwater habitats provide motivation to ameliorate downstream habitats and remove anthropogenic barriers to connectivity towards the goal of long‐term persistence of this species.

is about twofold that of terrestrial and ocean-dwelling vertebrates (Tickner et al., 2020). Major contributors to these declines include habitat loss (i.e. destruction, degradation, fragmentation), climate change, overexploitation, invasive non-native species, and pollution (Dudgeon et al., 2006). These factors, acting either alone or in concert, can reduce population numbers and genetic diversity, the latter of which can then further reduce long-term population viability (Frankham, 2005;Gaggiotti, 2003).
When populations contain sufficient genetic diversity, natural selection can act on beneficial alleles to facilitate or maintain local adaption (Frankham, 2005;Tiffin & Ross-Ibarra, 2014). Especially if habitat fragmentation is occurring, functional connectivity and gene flow between populations typically decrease, constricting the species range and increasing isolation between populations (Pierce et al., 2013). These demographic changes have genetic consequences, including population differentiation, genetic bottleneck effects and inbreeding, especially at low population densities (Neraas & Spruell, 2001). Fragmenting populations exposes the resulting sub-populations to inbreeding effect and genetic drift, both of which further decrease genetic diversity and heterozygosity through allele fixation (Frankham, 2005). Following allele fixation, the potential for expression of deleterious recessive alleles increases, with subsequent decrease in fitness-related traits in that population (Ruiz-López et al., 2012). In the absence of genetic rescue via gene flow from migrants, these negative genetic consequences can lead to population extirpation (Hoglund, 2009). Quantifying and incorporating population genetic information can provide valuable guidance when attempting to identify and protect populations of a declining species (Epifanio et al., 2003;Muniz et al., 2019). Although there have been great advances in genomics of model species, our understanding of the extent and distribution of within-and among-population genetic diversity remains limited for many threatened species.
Bull Trout (Salvelinus confluentus, Suckley 1895) is an apex predator endemic to northwestern North America. Across this vast range, spanning different watersheds, ecoregions, and management areas, the species is declining in population size and spatial extent due to overharvest, hybridization with invasive non-native species, reduced habitat connectivity, and habitat degradation (Taylor et al., 1999).
Owing to Bull Trout's important ecological role, popularity as a recreational fishery, and significance to Indigenous peoples, we have additional incentive to understand and monitor this sensitive, coldwater dwelling species (Warnock et al., 2011). Population genetic approaches have contributed to a better understanding of how the species is responding to stressors, especially in the southern extent of their range Costello et al., 2003;Taylor et al., 1999). Bull Trout studied thus far typically exhibit hierarchical population structure (Evanno et al., 2005), whereby high levels of genetic differentiation between populations cause population substructuring within a river basin (Whiteley et al., 2004). Additionally, these patterns generally hold regardless of the presence or relative representation of different life-history forms (i.e. resident and migratory; Homel et al., 2008). Rare gene flow between spawning streams typically results in a positive, albeit weak correlation between distance and genetic similarity, although populations typically contain unique allelic variants (Warnock et al., 2010). Because populations of threatened species typically harbour reduced genetic heterozygosity compared to not-at-risk congeners, it is crucial to establish presence and distribution of genetic diversity while correction measures are still viable (Spielman et al., 2004).  Ripley et al., 2005). Currently, there are relatively little data on the genetic diversity or genetic relationships within and among these Bull Trout populations, especially in the headwaters region (Ripley et al., 2005;Taylor et al., 1999). With reductions to population numbers and to the overall range of Bull Trout, this gap in our knowledge of their genetics negatively affects our ability to identify vulnerable populations and adequately protect them.
Here, we assess the population genetic structure of Bull Trout, a globally recognized species as risk (COSEWIC, 2012), in the upper reaches of the Athabasca River basin. We used neutral markers to genetically identify putative populations and characterize genetic differentiation within and among populations and among river basins in Alberta. Our specific objectives were to (a) identify Bull Trout populations within the Athabasca River basin and (b) characterize within-and among-population genetic variation for three major river basins (Athabasca -DU4, North Saskatchewan -DU2, and Bow -DU2), which have evolved from the same genetic lineage (COSEWIC, 2012). Within the Athabasca River Basin, we expect that Bull Trout will follow the typical hierarchical population structuring seen in other river basins where intra-population genetic differentiation is low, inter-population genetic variation is high, and isolation by distance is weak. By measuring genetic diversity and differentiation, these data will help inform management decisions, such as instances where reintroductions are being considered for ameliorated sites or genetically depauperate populations suggest the existence of barriers to migration and gene flow . Overall, we provide insight into how cryptic diversity in Bull Trout may inform local management and conservation strategies.

| Study location and sample collection
We sampled 14 sites to characterize the genetic diversity of Bull
These loci were chosen based on the degree of polymorphism, ability to detect Bull Trout × Brook Trout hybrids and resolution to detect population structure of the samples (Warnock et al., 2010).
PCR products were run on a 1.5% w/v agarose gel using a stan-

| Hierarchical population structure
Among-river basin and within-and among-sub-basin relationships were evaluated using the Bayesian clustering method in STRUCTURE v.2.3.4 (Pritchard et al., 2000). All analyses in STRUCTURE utilized the correlated allele frequency model and flexibility in linkage disequilibrium parameters to allow the complexities of natural systems to be included into the STRUCTURE estimates (Evanno et al., 2005;Vähä et al., 2007). These parameters allow allele frequencies to be correlated between populations and are able to accurately assign individuals to closely related populations due to admixture or recent common ancestry (Falush et al., 2003). This scenario is biologically plausible in this study system due to the potential for mainstem river  sampling sites to contain migratory adults from different populations that could be classified as a single population.
Including all sites within the Athabasca and Saskatchewan watersheds, 10 independent STRUCTURE runs (Evanno et al., 2005) performed at K-values of 1-15 were performed to confirm genetic differentiation of Bull Trout between major river basins (Athabasca and Saskatchewan). Subsequent STRUCTURE analyses were performed on Bull Trout captured within the Athabasca River basin to determine sub-basin structure using 10 independent runs at Kvalues of 1-15. For all K-values in STRUCTURE, 10 independent replicates were run. Initial STRUCTURE runs were conducted at 100,000 burn-in lengths and 100,000 Markov chain Monte Carlo (MCMC) iterations (Warnock, 2008) to reveal coarse, large-scale structure among the water basins while subsequent sub-watershed STRUCTURE runs were conducted at 500,000 burn-in lengths and 500,000 MCMC iterations to give more resolution to the localized population structure within watersheds (Warnock, 2008).
To determine the optimal value of K, the simulation summary results of the 10 independent runs for each value of K were evaluated with Structure Harvester v0.6.94 (Earl & vonHoldt, 2012).
If multiple unique STRUCTURE plots were created at each K, CLUMPAK (Kopelman et al., 2015) was used to find the optimal alignment of runs at a given value of K using CLUMPP (Jakobsson & Rosenberg, 2007) and plotted in R version 3.4.4 (R Core Team) using the R package POPHELPER (Francis, 2017). Using the recommendations of Pritchard et al. (2000) and Evanno et al. (2005), the model in which the lowest value of K that encompassed the majority of the structure while also having the highest rate of change in the log probability of the data (ΔK) was deemed the most likely correct K value. The hierarchical partitioning of genetic variation among river basins and drainages using Analysis of Molecular Variance (AMOVA) in GenAlEx version 6.5 (Peakall & Smouse, 2006). The R package adegenetversion 2.1.1 was used to perform a principal components analysis (PCA; Jombart, 2008), which corroborates these findings.
Isolation by distance was tested in the Athabasca River basin using the Mantel Test to determine the relationship between pairwise genetic distance (F ST ) and geographic distance (km) with 1,000 randomizations in each scenario (Dennenmoser et al., 2014;Jensen et al., 2005). Because Euclidean distance may not adequately represent realized distance between sites, we also measured the shortest waterway distance. Geographic distances in both cases were obtained using GoogleEarth (v. 4.2.1, Google Inc.).

| Microsatellite loci
The degree of polymorphism displayed by the microsatellite loci of the 431 Bull Trout ranged from 7 (Sfo18) to 51 (Sco109).     Figures 2 and S3). Although clustered based on river basin, the majority of regional variation was explained when sampling sites were grouped based on their location within the Athabasca River basin and the Saskatchewan River basin (Table 2). Additionally, an

| Population genetics
Analysis of Molecular Variance (AMOVA) showed that across all measured regions, both among and within major river basins, the greatest amount of genetic variation was explained by population level groupings ( Table 2).
The principal component analysis of microsatellite data from all river basins is generally consistent with evolution from a common   Table 2). Regardless of regional grouping, the majority of the genetic variation was explained within sampling sites.
In the North Saskatchewan River basin ( Figure S4) and the Bow River basin ( Figure S5), only preliminary STRUCTURE analyses were performed due to small sample sizes within river basin. although Jacques was not differentiated on these two axes, its 95% confidence ellipse was by far the smallest of the five drainages. As with the full dataset, the remaining PCA axes explained low amounts of variance in the dataset.

| Overview of population structure of Bull Trout
Population genetics is an important conservation tool for understanding the genetic health of vulnerable populations (Tiffin & Ross-Ibarra, 2014). Evaluating the status of a species' genetic diversity through population genetic techniques enables us to detect population trends, such as inbreeding depression (Restoux et al., 2012), hybridization (Rhymer & Simberloff, 1996), and population isolation by a cryptic barrier (Bergek & Björklund, 2007). In this study, Bull Trout exhibited patterns of high inter-population genetic differentiation within river basins with the majority of variation explained at the population level, a pattern found in other Salmonids in highly fragmented systems using a similar suite of markers Dehaan & Ardren, 2005).
All metrics of genetic differentiation and diversity that we considered support the conclusion that all river basins have similar levels of diversity, albeit with unique alleles and allele frequencies F I G U R E 3 Principal components analysis for all individuals sampled in the Athabasca, Bow and North Saskatchewan River basins, with 95% confidence ellipses for each group. N = 431 F I G U R E 4 Principal components analysis for individuals sampled in the Athabasca River basin, with 95% confidence ellipses for the five population groups identified by STRUCTURE analyses (see Figure 2; N = 293) TA B L E 3 Pairwise F ST estimates (Weir & Cockerham, 1984)  in each that differentiate the basins. The Athabasca River basin was found to contain several genetic clusters that did not correspond to the differences based on drainage of origin. These clusters were identified based on allele frequencies (by STRUCTURE), differentiation (F ST , F IS , and private alleles) and genetic diversity (H E and allelic richness). The majority of genetic variation was explained within each population with less variation explained among populations, and even less variation explained by the separate river basins.

| River basin scale differentiation and diversity
Despite the finding that Bull Trout in the Athabasca and North Saskatchewan River basins have similar levels of heterozygosity (H E ), STRUCTURE results showed that the two river basins are differentiated from one another, exemplified by populations being clustered based on grouping fish to their origin in the Western Arctic or Saskatchewan-Nelson DU. In both river basins, high heterozygosity values and low F IS were observed, alluding to large genetic differences between populations, which is common for Bull Trout systems (Costello et al., 2003). On average, the Saskatchewan River basin had higher allelic richness than the Athabasca River basin. The higher al- Because fewer sites were sampled in the North Saskatchewan and Bow River basins compared to the Athabasca River basin, our values for genetic diversity and allelic richness may be underestimated in the former two basins due to limited localized sampling in those areas.
Anticipated differentiation for Bull Trout between the Athabasca and Saskatchewan watersheds was confirmed by STRUCTURE ( Figure 2). This supports the COSEWIC (2012) designation of the two river basins as separate conservation units, the Western Arctic and Saskatchewan-Nelson, despite deriving from a common genetic lineage. Genetic differentiation is just one method used to support this designation, but it provides context for considering how Bull Trout from these DUs may differentially respond to management strategies and climate change. Previous studies using microsatellite loci have illuminated regional genetic differentiation in Bull Trout, corroborating the designation of these groupings Taylor et al., 1999). This type of differentiation on a large scale is common among fish (McPhail & Lindsey, 1986) due to different refugia, extended isolation, and limited gene flow (Avise, 2004). Because the groups have been separated for an extended period of time, evolutionary processes (e.g. genetic drift purging or fixing mutations in the two river basin's populations) have influenced the genetic differentiation and divergence of Bull Trout in these two river basins.

| Sub-basin structure within the Athabasca River basin
Within the Athabasca River basin, additional substructuring was found. Genetic differences between populations were high but consistent with other Bull Trout studies (Warnock et al., 2010). For this dataset, differentiation among groups was best explained when using the five distinct clusters that were detected (

| Isolation by distance
Despite the evidence of substructuring in the Athabasca River basin, no evidence for isolation by distance was found. This result suggests that the observed genetic differences are at least partly influenced by physical barriers or other unmeasured natural occurrences that impede or facilitate movement between populations in ways that do not correlate linearly with geographical distance (Slatkin, 1993).
In lower, more homogenous stretches of the Athabasca River basin, another Salmonid, Arctic Grayling (Thymallus arcticus, Pallas 1776) exhibits moderate isolation-by-distance patterns, which is thought to be due to a combination of their large geographic ranges and population sizes (Reilly et al., 2014). Mountain Whitefish (Prosopium williamsoni, Girard 1856), also a Salmonid, tends to exhibit weak isolation-by-distance trends, although this has been attributed to large population sizes and high levels of gene flow, which prevent differentiation (Whiteley et al., 2006).

| Implications
Because the range of Bull Trout is declining, this elevates the importance of conserving genetic diversity in remaining extant populations as a means to withstand stochastic events (Rieman & McIntyre, 1995) and a source for adaptation in the future (Frankham, 2005). Shrinking populations on the periphery of the range are especially at risk of extirpation as a result of isolation and habitat fragmentation (Rieman & Myers, 1997). At present, many Bull Trout populations are listed as "data deficient" with little insight into regionally specific differences from the highly studied areas from which management plans were created (COSEWIC, 2012). Establishing baseline levels of genetic diversity allows for comparisons and detection of change in future stocks in addition to evaluation of adaptive management efforts (Epifanio et al., 2003). Uncovering areas that are genetically distinct from one another and determining the level of differentiation in local river basins can guide management strategies that work at local scales. In Alberta, the genetic integrity of fish species and populations is determined based on the degree of hybridization, genetic similarity to original stock, and genetic distinction (Rodtka, 2009).
In this study, we provide baseline genetic information from which to track genetic changes observed in Bull Trout populations in the Athabasca River basin. Our results suggest little hybridization with Brook Trout at present in this region, as no hybrids or backcrosses were observed or genetically detected. With regard to genetic distinctiveness, we found genetic substructuring within the Athabasca River basin. High pairwise F ST values between sites indicate differentiation among groups (Table 3). However, high heterozygosity (H E ) values and AMOVA results indicate that genetic variance largely resides within the population level. Thus, although there is differentiation between groups, there currently appears to be sufficient genetic diversity that can be drawn upon, acting as evolutionary potential to allow adaptation to changing habitat conditions and longterm persistence.

This research was supported by funding and in-kind support from
Parks Canada Agency and Alberta Environment and Parks and was supported by NSERC, QEII Scholarship, ACA, and ASPB Funds.
John Post, Shelley Alexander, and Ward Hughson provided project guidance and support. Peter Peller provided GIS support. Lindsay Dowbush and Traudi Golla assisted with field sampling.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest to declare.