Genetic changes caused by restocking and hydroelectric dams in demographically bottlenecked brown trout in a transnational subarctic riverine system

Abstract Habitat discontinuity, anthropogenic disturbance, and overharvesting have led to population fragmentation and decline worldwide. Preservation of remaining natural genetic diversity is crucial to avoid continued genetic erosion. Brown trout (Salmo trutta L.) is an ideal model species for studying anthropogenic influences on genetic integrity, as it has experienced significant genetic alterations throughout its natural distribution range due to habitat fragmentation, overexploitation, translocations, and stocking. The Pasvik River is a subarctic riverine system shared between Norway, Russia, and Finland, subdivided by seven hydroelectric power dams that destroyed about 70% of natural spawning and nursing areas. Stocking is applied in certain river parts to support the natural brown trout population. Adjacent river segments with different management strategies (stocked vs. not stocked) facilitated the simultaneous assessment of genetic impacts of dams and stocking based on analyses of 16 short tandem repeat loci. Dams were expected to increase genetic differentiation between and reduce genetic diversity within river sections. Contrastingly, stocking was predicted to promote genetic homogenization and diversity, but also potentially lead to loss of private alleles and to genetic erosion. Our results showed comparatively low heterozygosity and clear genetic differentiation between adjacent sections in nonstocked river parts, indicating that dams prevent migration and contribute to genetic isolation and loss of genetic diversity. Furthermore, genetic differentiation was low and heterozygosity relatively high across stocked sections. However, in stocked river sections, we found signatures of recent bottlenecks and reductions in private alleles, indicating that only a subset of individuals contributes to reproduction, potentially leading to divergence away from the natural genetic state. Taken together, these results indicate that stocking counteracts the negative fragmentation effects of dams, but also that stocking practices should be planned carefully in order to ensure long‐term preservation of natural genetic diversity and integrity in brown trout and other species in regulated river systems.


| INTRODUC TI ON
Long-term persistence of natural populations depends on a complex interplay of ecoevolutionary forces affecting genetic diversity and local adaptation to environmental conditions (Bijlsma & Loeschcke, 2012;Mimura et al., 2017). Habitat destruction and overexploitation of species may lead to isolated and small populations with declining effective population size and genetic diversity, and increasing genetic drift (Bijlsma & Loeschcke, 2012;Mimura et al., 2017;Piccolo, Unfer, & Lobón-Cerviá, 2017;Vøllestad, 2017). Introduction of foreign genetic material by immigration or the release of translocated and/or hatchery-reared individuals into the wild may replace gene pools (Laikre, Schwartz, Waples, & Ryman, 2010;Quiñones, Johnson, & Moyle, 2014). Conservation and management actions commonly aim to increase connectivity and demographic robustness by, for example, establishing dispersal corridors or releasing individuals into the wild to support local populations (Quiñones et al., 2014). However, potential genetic effects of such mitigating actions are often not considered or monitored, which increases the risk of losing natural genetic diversity in disturbed and altered populations (Araguas et al., 2009(Araguas et al., , 2017. Therefore, assessments of genetic diversity, connectivity, and structure are essential for ensuring long-term benefits of management actions (Quiñones et al., 2014).
The Eurasian brown trout (Salmo trutta L. 1758, Figure 1) is a socioeconomically important freshwater fish that is widespread in the Northern Hemisphere (Jonsson & Jonsson, 2006;Laikre, 1999;Vøllestad, 2017). It is ecologically and morphologically variable including, resident and migratory life-history forms like anadromous (i.e., natal rivers-sea-natal rivers' migrations), as well as potamodromous (i.e., natal rivers-lakes-natal rivers' migrations) that can coexist in the same habitat. Although not endangered as such, natural brown trout populations with high genetic integrity are becoming increasingly rare across the distribution range (Araguas et al., 2009(Araguas et al., , 2017Baric et al., 2010) due to anthropogenic habitat destruction and the long-term practice of translocations and stocking. Hydropower developments are widespread in many countries and affect a wide range of formerly continuous brown trout populations (Heggenes & to ensure long-term preservation of natural genetic diversity and integrity in brown trout and other species in regulated river systems.

K E Y W O R D S
fish stocking, genetic diversity, genetic erosion, genetic integrity, habitat fragmentation, Salmo trutta Røed, 2006;Vøllestad & Hesthagen, 2001). Therefore, supportive breeding has been widely applied to support local populations that are under pressure of fragmentation effects of dams and/or overfishing (Vøllestad & Hesthagen, 2001).
Targeted studies of how anthropogenic disturbances, such as hydropower developments, and mitigating actions, such as stocking, work in concert to affect the genetic integrity of natural populations are warranted to support long-term conservation goals in brown trout and other species facing similar challenges (Araguas et al., 2009(Araguas et al., , 2017Baric et al., 2010;Berrebi et al., 2019;Piccolo et al., 2017). In this study, we used brown trout as a model species and compared genetic diversity and differentiation patterns along a transnational subarctic riverine system that has been partitioned by hydroelectric power dams, and that additionally is regularly stocked with fish in parts of the river system but not in others. The Pasvik River, shared between Norway, Russia, and Finland, and for a large part constituting the border between Norway and Russia, is one of the largest and most species-rich subarctic river systems in northwestern Eurasia, and is described in sport fishing literature as one of Norway's best trout fishing destinations. However, the construction of seven hydroelectric dams from 1932 to 1978 hindered trout migration (Arnesen, 1987) and led to the destruction of many natural spawning and nursery areas (Amundsen et al., 1999;Jensen, Bøhn, Amundsen, & Aspholm, 2004). To strengthen the breeding population, approx. 5,000 offspring of local specimens are released annually between dams in the Norwegian-Russian parts of the river (Amundsen et al., 1999;Jensen et al., 2004; Table 1). Each year new parents are caught, but from the same location (i.e., zone H; Figure 2). Surveys show that 70%-90% of the trout caught in the Norwegian-Russian part of the river are stocked fish (Haugland, 2014). By contrast, no stocking of brown trout takes place in the upper Russian sections of the river, where the natural trout population is presumably more intact. Due to the dams, there is assumedly little or no gene flow between the stocked (Norwegian-Russian) and nonstocked

| Study system and sampling
The study area comprises the Pasvik River, which originates in Lake Inari in Finland and runs 147 km in a northeastward direction  Table 1). None of the dams have fish passes facilitating migration between river sections.
Additionally, two more sections were introduced and sampled (F and I, Figure 2) that are separated by rapids from adjacent sections, but not by dams, and that mark a significant change in habitat to allow for comparative fine-scale genetic analyses. Regular stocking in the Norwegian-Russian river part has occurred in fairly equal numbers in sections E/F and H/I with about 50% of the stocking pool being released in these river parts. Section G has been stocked in the past with thousands of hatchery-bred individuals, but it seems that no F I G U R E 1 A brown trout specimen from the Pasvik River (Photo: Valery Buzun)

| Development of multiplex PCRs
In total, 40 previously described microsatellite or short tandem repeat (STR) loci were tested in the development of multiplex PCR for genotyping of brown trout. When needed, original primers were redesigned by adding PIG-tails (Brownstein, Carpten, & Smith, 1996) and changing both amplicon lengths and annealing temperatures to provide adequate peak separation and minimize noise on the genetic analyzer (Applied Biosystems 3730xl) used in this study. The STR markers were tested first on a small sample set (N = 7), and later the novel multiplexes were validated on a larger sample set (N = 75) consisting of fish collected in the field (sections F and G; Figure 2). First, STR markers were run singly (0.5 µM primer concentration, PCR program as described under DNA extraction and multiplex PCR-STR analysis-58°C annealing) to exclude those that were difficult to score and those that were in physical linkage (Gharbi et al., 2006).

| DNA extraction and multiplex PCR-STR analysis
Genomic DNA was extracted from the samples using a DNeasy Blood & Tissue kit (Qiagen) and genotyped at 16 STR loci (Table 2) The PCR cycling profile consisted of an initial denaturation step at 95°C for 10 min, followed by 28 cycles including a denaturation step at 94°C for 30 s, annealing at 55°C/58°C (depending on multiplex; Genetic Analyzer (Applied Biosystems), sized, and scored using GeneMapper 5.0 (Applied Biosystems), and manually verified.

| Analysis of genetic variation
Tests for linkage disequilibrium and deviations from Hardy-Weinberg equilibrium (HWE) were carried out with the software GENEPOP 4.7 (Rousset, 2008 AB001062.1 Presa and Guyomard (1996) R: GTTTCTTCACAAGTCATCTGGGCATCT a a chain method with 10,000 dememorization steps, 5,000 batches, and 10,000 iterations each was used to estimate exact p-values for deficiency of heterozygotes and likelihood ratio statistics, respectively.
Observed and expected heterozygosity and inbreeding coefficient for the seven river sections were calculated with GenAlEx 6.51b2 (Peakall & Smouse, 2012). Allelic richness and private allelic richness were estimated with ADZE 1.0 (Szpiech, Jacobsson, & Rosenberg, 2008) based on a standardized sample size of 16.

| Population-genetic differentiation
GenAlEx 6.51b2 was used to estimate pairwise population-genetic differentiation based on G ST (Nei & Chesser, 1983) and Jost's D (Jost, 2008) and to test their significance based on 9,999 random permu-

| Genetic structure analysis
The Bayesian clustering method implemented in STRUCTURE 2.3.4 (Pritchard, Stephens, & Donnelly, 2000) was applied to detect the presence of distinct genetic clusters and to identify individuals of potentially admixed ancestry. The admixture model with correlated allele frequencies (Falush, Stephens, & Pritchard, 2003) was run twice, once using the LocPrior option and once without. Forty replicates were conducted for each K from 1 to 10, with 1,000,000 MCMC steps and a burn-in period of 100,000. The LocPrior option was chosen to assess whether additional population-genetic structure could be detected, as it has been shown that including information on the sampling location of individuals improves clustering without leading to the detection of nonexisting populationgenetic structure (Hubisz, Falush, Stephens, & Pritchard, 2009).
Additionally, we used STRUCTURE to test for hierarchical population structure, in order to assess whether higher-level genetic structure masks fine-scale genetic clustering. The number of genetic clusters present in the data set was estimated by four recently proposed estimators (Puechmaille, 2016): the median of means (MedMeaK), maximum of means (MaxMeaK), median of medians (MedMedK), and maximum of medians (MaxMedK), using the program STRUCTURESELECTOR (Li & Liu, 2018) to account for uneven sample sizes in the data set. The program CLUMPAK (Kopelman, Mayzel, Jakobsson, Rosenberg, & Mayrose, 2015) was used to visually summarize results from the separate STRUCTURE runs. Finally, significance of differences in observed heterozygosity between the genetic clusters was tested with FSTAT 2.9.3.2, based on 5,000 permutations (Goudet, 2002).

| Bottleneck analysis
To test for recent reductions in effective population sizes (i.e., genetic bottlenecks), the program BOTTLENECK 1.2.02 (Piry, Luikart, & Cornuet, 1999) was used. The algorithm in BOTTLENECK assumes that allelic diversity is lost at a faster rate than heterozygosity and, therefore, tests for an excess of heterozygosity compared to expectations at mutation-drift equilibrium (Cornuet & Luikart, 1996). Following recommendations by Peery et al. (2012), two mutation models were assessed, the infinite-allele model (IAM) and the two-phase model (TPM). The TPM model allows different proportions of microsatellites to follow either the IAM or the stepwise mutation model (SMM) and so the model was run three times for each population, assuming that the percentage of stepwise mutations was 20%, 50%, and 70%, respectively. The 1-way Wilcoxon sign-rank test (Luikart, 1997) was applied to assess significance.

| Genetic variation
The result of the microsatellite optimization was five novel multiplexes consisting of 16 STR markers (

| Population-genetic differentiation
Estimates of population-genetic differentiation (G ST and Jost's D,

| Genetic structure results
The different estimators in STRUCTURESELECTOR provided support for 3-4 genetic clusters in the whole data set (Figure 4ac) regardless of whether the LocPrior option was used or not.
Nevertheless, the posterior LocPrior parameter (mean r = 1.64 for K = 3) indicated that location information was fairly informative for assisting in genetic clustering. At K = 3, the three groups generally corresponded to the nonstocked river sections A-C in the Russian part and to the stocked sections G-J in the Norwegian-Russian part, while sections in-between (i.e., E and F) partially showed assignment to a third genetic cluster (Figure 5a,b). A few individuals in section G showed assignment to this third cluster as well.
Although less supported, it is noteworthy that at K = 2, genetic structuring essentially separated the nonstocked Russian sections from stocked sections in the Norwegian-Russian part of the river.
Individuals that were assigned to the third cluster at K = 3 were mostly unassigned at K = 2, and were concentrated in the area connecting the stocked and nonstocked river sections. Additional STRUCTURE runs in which 25 individuals were randomly selected from sections B and G to reduce unevenness in the data set in terms of sample sizes verified the presence of the main genetic clustering described above as well as the main admixture patterns present in the center of the river (results not shown

| D ISCUSS I ON
The current study provides evidence that patterns of fine-scale genetic diversity and differentiation are governed by both hydroelectric dams and restocking in a transnational subarctic riverine system. In nonstocked parts of the Pasvik River, dams contributed to significant genetic differentiation between adjacent river sections, whereas this effect was absent in the stocked parts.
Additionally, heterozygosity was comparatively low in nonstocked compared to stocked river sections. However, in the stocked river  Theoretical and simulation studies (Morrissey & de Kerckhove, 2009;Paz-Vinas, Loot, Stevens, & Blanchet, 2015) predict that population sizes and genetic diversity might be naturally higher in downstream than in upstream river sections. Two main processes have been proposed to explain this prediction: (a) downstream-biased gene flow caused by asymmetric dispersal costs in unidirectional water flow (Morrissey & de Kerckhove, 2009) and (b) variation in habitat availability, with larger habitat areas usually being available downstream due to increased river width (Carrara, Rinaldo, Giometto, & Altermatt, 2014).

| Effects of stocking on genetic diversity and demography
Generally, it is more likely that potential higher genetic diversity results from more variation in habitat availability in the current study because downstream-biased gene flow is low or absent because of the dams. The prediction of higher genetic diversity in downstream river sections may be partially supported by this study based on relatively high heterozygosity values found downstream. However, this F I G U R E 5 STRUCTURE bar plots. (a) STRUCTURE bar plots for K = 2 to K = 3 for all river sections (not using LocPrior). (b) STRUCTURE bar plots for K = 2 to K = 3 for all river sections (using LocPrior). (c) STRUCTURE bar plots for K = 2 for the upper Russian part (using LocPrior) (a) All sections, no LocPrior (c) Russian sections, with LocPrior

| Effects of stocking on population-genetic differentiation
The most likely partitioning scheme resulting from the Bayesian shown to contribute to substantial genetic differentiation and drift effects in salmonids, leading to rapid divergence of wild and captivity-bred fish (Hansen et al., 2009;Laikre et al., 2010;Quiñones et al., 2014). Similarly, stocking probably provides an explanation for the genetic differentiation between stocked and nonstocked river sections in our study. However, the origin of the third genetic cluster is less clear. Looking at the K = 3 STRUCTURE bar plot (Figure 5a,b), one possibility is that admixture of individuals from stocked and nonstocked parts or the river led to an additional genetically differentiated cluster over time in the center of the watercourse. We cannot rule out that occasional downstream migration of individuals through dams contributes to this pattern. An alternative explanation is that flooding led to a carry-over of fish specimens in this river section. Both migration and flooding also allow for the introduction of genetically different individuals from side rivers and hence the introduction of new genetic material. In connection to this, sections E and F, including their side rivers, are assumed to have larger intact spawning sites and nursery areas for brown trout, probably allowing for the retention of natural genetic diversity.
Despite the existence of four hydroelectric dams, genetic differentiation was generally low among sections of the Norwegian-Russian part of the Pasvik River. Within this stocked part, differentiation was statistically significant only for two pairwise comparisons involving section G. This particular section is short and therefore possibly has a lower population size than other river sections. It also shows the lowest private allelic richness, suggesting that the genetic differentiation in this case is caused by drift.
Finally, although stocking in section G was stopped more than a decade ago, it is possible that past extensive stocking introduced alleles that led to the significant genetic differentiation. The general genetic uniformity of the brown trout population of the Norwegian-Russian river part indicates that stocking has homogenized the gene pools of the separate river sections. Contrastingly, in the upstream Russian part of the river, sections B and C showed significant genetic differentiation and were identified as distinct genetic clusters in the STRUCTURE analysis. Hence, in these nonstocked parts of the river, dams contributed to genetic differentiation. This finding is in line with previous studies in brown trout and other salmonids showing that artificial barriers can quickly lead to isolation and genetic differentiation (Heggenes & Røed, 2006;Meldgaard et al., 2003;Stelkens et al., 2012).
Paradoxically, our results suggest that the three hydroelectric dams of the headwater sections of Pasvik River may protect wild trout populations in Russia from genetic swamping by hatchery-reared fish of Norwegian-Russian origin by preventing their upstream migration. Indeed, several studies have pointed out that removal of barriers or the installment of fish passes could have unintended detrimental consequences for unique genetic diversity in natural populations of brown trout and other aquatic species (Baric et al., 2010;Van Houdt et al., 2005;Rahel, 2013). Due to widespread anthropogenic habitat destruction across the distribution range of the Eurasian brown trout, as well as the long-term practice of translocations and stocking, natural and genetically unaltered trout populations are becoming increasingly rare (Araguas et al., 2009(Araguas et al., , 2017Baric et al., 2010). Consequently, the identification and preservation of populations with high genetic integrity is a primary longterm conservation goal in brown trout (Araguas et al., 2009(Araguas et al., , 2017Baric et al., 2010). The Pasvik River is one of the largest and most species-rich subarctic river systems in northwestern Eurasia, with partially remaining wild brown trout populations; conservation of these natural populations is important to preserve unique genetic diversity in the region. Although brown trout as a species is not endangered, harvest, translocations, and stocking have heavily altered natural populations, and it is therefore crucial to preserve remaining wild populations in order to retain the full spectrum of intraspecific genetic and ecological diversity (Berrebi et al., 2019;Piccolo et al., 2017).
To conclude, our results indicate that stocking alleviated the negative genetic effects of habitat fragmentation caused by dams by reducing genetic differentiation, but also that stocking is the likely cause of genetic diversity loss, as demonstrated by the significant bottleneck results. This suggests that the stocking program in the Pasvik riverine system would benefit from a larger parental breeding pool, in order to prevent further genetic diversity loss and to ensure long-term population viability. Likewise, increasing natural recruitment by restoration of spawning areas may be a viable option to support a larger breeding pool. In a broader context, the present study points to the fact that contrasting management strategies across international borders were insufficient to address certain sustainable management goals that aim for the long-term protection of natural genetic diversity, connectivity, and evolutionary potential. Consequently, our results suggest that transnational harmonization of mitigation strategies based on more research may be warranted to develop efficient large-scale conservation management strategies that link local genetic diversity to large-scale genetic integrity for native brown trout in the Pasvik riverine system and across Eurasia.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
SH conceptualized and designed the study and wrote first draft together with CFCK. CFCK additionally analyzed the data. KF carried out laboratory analyses. NP contributed samples from Russia. All authors commented, interpreted, and discussed the results and contributed critical feedback for the final version of the study, including analytical steps and background information. PEA provided background information on ecosystem specifics. All authors approved the final version of the article.

DATA ACCE SS I B I LIT Y
The microsatellite data set has been deposited in the Dryad Digital Repository (DOI: https://doi.org/10.5061/dryad.45482t5).