1644 ARTICLE

Loss of SNP genetic diversity following population collapse in a recreational walleye (Sander vitreus) fishery Brandon E. Allen, Ella Bowles, Matthew R.J. Morris, and Sean M. Rogers

Abstract: Walleye (Sander vitreus) are in demand as a commercially and recreationally harvested freshwater fish in Canada. Managed populations may exhibit different phenotypic and genetic signatures from their natural counterparts. In , Canada, this fishery is recovering from population collapses attributed to intensive recreational angling. We hypothesized that historical population collapses would be associated with signatures of reduced genetic diversity. To address this question, we sampled six walleye lakes in , including historical tissue samples for one population, and used genotyping-by- sequencing to characterize 1081 single nucleotide polymorphisms (SNPs). Lakes were identified as unique genetic clusters except for two lakes that unexpectedly exhibited signs of genetic clustering. Using historical DNA samples, 428 homologous SNPs characterized in walleye between pre- and postpopulation collapse exhibited significant reductions in multiple estimates of genetic diversity. Collectively, our results illustrate that genotype-by-sequencing methods that integrate historical and contem- porary samples in association with managed populations provide insight into the consequences of harvest pressure causing collapse.

Résumé : Les dorés jaunes (Sander vitreus) sont prisés comme poissons d’eau douce exploités de manière commerciale et sportive au Canada. Les populations gérées peuvent présenter des signatures phénotypiques et génétiques différentes de celles des populations naturelles. En Alberta (Canada), cette ressource se remet d’un effondrement de la population attribué à la pêche sportive intensive. Nous avons postulé que les effondrements de populations passés seraient associés à des signatures d’une diversité génétique réduite. Pour examiner cette question, nous avons échantillonné six lacs à dorés jaunes dans le nord de l’Alberta, incluant des échantillons de tissus historiques pour une population, et utilisé le génotypage par séquençage pour caractériser 1081 polymorphismes mononucléotidiques (SNP). Il a été établi que les lacs présentent des groupements génétiques uniques, à l’exception de deux lacs qui, contrairement aux attentes, présentent des signes de regroupement génétique. Sur la base d’échantillons d’ADN historiques, 428 SNP homologues caractérisés chez des dorés jaunes d’avant et d’après l’effondrement de populations présentent des baisses significatives dans plusieurs estimations de la diversité génétique. Collectivement, nos résultats montrent que les méthodes de génotypage par séquençage qui incorporent des échantillons historiques et modernes en association avec des populations gérées fournissent de l’information sur les conséquences de la pression de la pêche menant à

For personal use only. l’effondrement. [Traduit par la Rédaction]

Introduction facilitates population response to changing environments (Aitken Wildlife management strategies attempt to balance the needs et al. 2008; Hamilton and Miller 2016), new diseases (Lamaze et al. of wildlife with the needs of people using the best available sci- 2012), or adaptive growth rates (Cena et al. 2006). Characterizing ence. These strategies historically have included harvest limits patterns of genetic variation within and among populations also (Pitcher 2001), antipoaching regulations (Walker et al. 2007), or provides a means to predict whether nonrandom mating struc- breeding programs (Kleiman 1989; Leberg and Firmin 2008)in tures the populations and contributes to isolating barriers (Milligan efforts to maintain stable population abundance. The relatively et al. 1994; McCracken et al. 2001; Pearse and Crandall 2004). recent integration of genetics with conservation and manage- The integration of genetic tools in heavily managed popula- ment has provided novel molecular tools that can be used to tions presents major challenges. Managed populations often mitigate the effects of inbreeding (Hedrick and Kalinowski 2000), deviate from the population genetic expectation for natural pop- identify the genetic effects of selective harvest on populations ulations (Waples 2010, 2015). Unreported historical stocking or (Allendorf et al. 2008), and detect hybridization (Neville and supplementation can contribute to uncertainty in the expected Dunham 2011; Allen et al. 2016). Where it was once difficult to population genetic profiles (Morán et al. 1991), in addition to the Can. J. Fish. Aquat. Sci. Downloaded from cdnsciencepub.com by UNIV CALGARY on 04/10/21 implement such tools for some species (DeSalle and Amato 2004; consequences of hybridization and competitive interactions be- Segelbacher et al. 2010), newer methods of sequencing are making tween wild and stocked individuals (Araki et al. 2007; Naish et al. it possible to rapidly integrate genetics into more progressive 2007; Yau and Taylor 2013; Allen et al. 2016). These challenges may management and conservation strategies for most species. be partially circumvented by the integration of fisheries genetics Identifying genetic variation is beneficial for the long-term con- in cases where there is access to historical samples, whereby tem- servation and management of a species because such variation poral comparisons can directly test the consequences of manage-

Received 24 April 2017. Accepted 17 December 2017. B.E. Allen,* E. Bowles, M.R.J. Morris,† and S.M. Rogers. Department of Biological Sciences, University of Calgary, 2500 University Drive NW, Calgary, AB T2N 1N4, Canada. Corresponding author: Brandon E. Allen (email: [email protected]). *Present address: Alberta Biodiversity Monitoring Institute, University of Alberta, CW 405 Biological Sciences Centre, Edmonton, AB T6G 2E9, Canada. †Present address: Ambrose University, 150 Ambrose Circle SW, Calgary, AB T3H 0L5, Canada. Copyright remains with the author(s) or their institution(s). Permission for reuse (free in most cases) can be obtained from RightsLink.

Can. J. Fish. Aquat. Sci. 75: 1644–1651 (2018) dx.doi.org/10.1139/cjfas-2017-0164 Published at www.nrcresearchpress.com/cjfas on 18 December 2017. Allen et al. 1645

Table 1. Population name, river basin, year samples were collected (Year), number of individuals sampled (N), mini- mum (MinAge) and maximum (MaxAge) age of samples, minimum (MinFL) and maximum (MaxFL) fork length (in mm), and GPS location (°N, °W) of six walleye populations in Alberta, Canada. MinFL MaxFL GPS location Population River basin Year N MinAge MaxAge (mm) (mm) (°N, °W) Gods Peace–Slave 2005 20 1 15 200 583 56.82, 114.28 Graham Peace–Slave 2004 20 20 21 475 648 56.56, 114.55 Rainbow Hay 2002 20 5 10 369 503 58.28, 119.28 Round Peace–Slave 2004 20 5 17 292 510 56.75, 114.56 Vandersteene Peace–Slave 2000 20 2 22 188 685 56.61, 114.46 Smoke Peace–Slave 2005 20 2 15 228 475 54.36, 116.94 Smoke Peace–Slave 1973 20 7 13 465 552 54.36, 116.94

ment decisions (Coltman et al. 2003; Uusi-Heikkilä et al. 2015; We had two objectives for this study. The first was to character- Haponski and Stepien 2016). ize genetic population structure and genetic diversity estimates The walleye fishery in Alberta, Canada, is an ideal system in using genome-wide SNPs. We hypothesized genetic diversity which to investigate the genetic consequences of management would vary among lakes and river basins, with lakes that have strategies. Walleye is a cool-water species that persists throughout experienced population collapse associated with reduced genetic North America, occupying around 32% of all available freshwater diversity. Alternatively, collapse may not have had an impact on habitats (Bozek et al. 2011), and were likely established in Alberta genetic diversity (Lippé et al. 2006). For example, copper redhorse during the Missourian refugium (Billington 1996). Fishing pres- (Moxostoma hubbsi) population abundance in Quebec, Canada, has sure on walleye is known to be greater in Alberta due to the decreased in response to breeding sites becoming increasingly popularity of the sport fishery and limited number of populations isolated with the construction of dams (Lippé et al. 2006). Yet, available for fishing (Sullivan 2003). This increased fishing pres- genetic diversity does persist despite low census population size sure has been hypothesized to be the main factor contributing to (Lippé et al. 2006; Labonne et al. 2016). In addition, we expect the observed collapse of several walleye populations (Post et al. populations from geographically proximate lakes to be more ge- 2002). When the Walleye Management and Recovery Plan was netically alike than lakes separated by greater distances. Our sec- implemented in 1996 (Berry 1995), 62 out of 177 water bodies ond objective was to investigate temporal changes in genetic containing walleye were designated as collapsed (<0.05 fish per diversity and length-at-age within a population that has continu- hour). The goal of the management plan was to recover and sus- ally experienced harvest over the last 40 years (1973 and 2005, tain Alberta walleye populations using an active management approximately four to eight generations) with documented popu- approach; this involved implementing a diversity of minimum lation collapses. We predicted that the genetic consequences of size and catch limits in addition to special harvest licenses. Special these processes would leave phenotypic and genetic signatures in harvest licenses are distributed via a lottery and allow limited contemporary individuals, generally characterized by lower ge- harvest of specified size classes at lakes with sustainable popu- netic diversity and decreased length-at-age.

For personal use only. lations (http://albertaregulations.ca/fishingregs/prov-regs.html). However, some walleye populations have begun to exhibit modi- Materials and methods fied growth curves (Spencer 2010), indicating a phenotypic or genetic shift. Whether such practices are associated with a size- Sample collection selective harvest pressure remains unknown. Small fin ray clips (ϳ50 mg tissue) were collected from six wall- Walleye were regularly stocked in Alberta, with records dating eye lakes within the and Slave– basins in back to 1926. There was a large increase in stocking effort in the late Alberta, Canada, between 2000 and 2005 by Alberta Environment 1980s to early 1990s (Johnston and Paul 2006). Eggs were collected and Parks (AEP) as part of their annual indexing program to mon- from six donor lakes, with , , and Prim- itor walleye populations (Fig. 1; Table 1). Smoke Lake had addi- rose Lake acting as the primary sources. Stocking efforts continued tional scale samples that were collected in 1973 before population into the 2000s but were stopped in 2014. Walleye populations collapse was documented. Twenty individuals from each of the six have begun to recover, but a large number of lakes are still at high populations and from the Smoke Lake 1973 collection were ran- risk of collapse (Walleye Population Status History 2015; available domly sampled (N = 140). Age (from pelvic fin estimates; Mackay from http://aep.alberta.ca/fish-wildlife/fisheries-management/fish- et al. 1990) and fork length (mm) were collected for each individ- sustainability-index/fsi-species-maps/documents/WalleyeLakes ual at the time of sampling by AEP (Table 1).

Can. J. Fish. Aquat. Sci. Downloaded from cdnsciencepub.com by UNIV CALGARY on 04/10/21 PopulationStatus-Sep23-2015A.pdf). Prolonged stocking events from a small number of donor lakes might have left unique DNA isolation and library preparation genetic signatures on this fishery. DNA was isolated using NucleoSpin Tissue kits from Macherey- The genetic consequences of walleye management in Alberta is Nagel, and sequencing was completed using a genotype-by- unknown (Johnston and Paul 2006). However, research in Ontario sequencing protocol with the restriction enzymes EcoRI and MseI and the Great Lakes provides insight into walleye population ge- as our rare- and frequent-cutter, respectively. Library preparation netics (e.g., McParland et al. 1999; Strange and Stepien 2007; consisted of the following steps: (i) digestion; (ii) ligation; (iii) size- Haponski and Stepien 2014, 2016; Haponski et al. 2014). In these selection; and (iv) amplification. studies, genetic markers (allozymes, mitochondrial DNA, and mi- Reactions and conditions for (i) digestion and (ii) ligation are as crosatellite markers) have characterized spatial and temporal follows. Both EcoRI and MseI adapters were annealed at 95 °C for genetic structure in managed fisheries. Haponski et al. (2014) de- 5 min and allowed to cool to room temperature. A total of 200 ng termined that genetic composition of walleye in Lake Erie re- of isolated DNA from each individual was digested in 9 ␮L vol- mained distinct despite large stocking efforts. This suggests wild umes containing 1.15 ␮L of 10× T4 buffer, 0.60 ␮L of 1 mol·L–1 NaCl, populations can be robust to the potential homogenizing effects 0.60 ␮L of bovine serum albumin (BSA; 1 mg·mL–1), 0.25 ␮Lof ␮ –1 ␮ of stocking. Understanding the genetic consequences of harvest ddH2O, 0.12 LofMseI enzyme (10 000 units·mL ), 0.28 L of HiFi and stocking could inform management strategies in this fishery. EcoRI enzyme (20 000 units·mL–1), and 6 ␮L of DNA at 37 °C for 8 h

Published by NRC Research Press 1646 Can. J. Fish. Aquat. Sci. Vol. 75, 2018

Fig. 1. Black shapes indicate lake locations (n = 6) analyzed in this study. Dark grey lines indicate provincial boundaries, light grey lines indicate major river systems, and black lines indicate drainage basins. The capital city of Alberta, Canada (Edmonton), is shaded in grey and labelled. For personal use only.

followed by 65 °C for 45 min. After digestion, 1.0 ␮L of a unique (Thermofisher) and Tapestation (Agilent) to confirm the target EcoRI barcode adaptor (1 pmol·␮L–1) and 1.4 ␮L of master mix was size distribution of all samples. Pooled samples were standardized added to each well for ligation. The master mix consisted of 1 ␮L to 2 nmol·L–1 and combined into a final DNA library, which was ␮ ␮ of the universal MseI adaptor, 0.072 LofddH2O, 0.1 Lof10× sequenced on a NextSeq 500 using a 300-cycle mid-output flow ␮ –1 ␮

Can. J. Fish. Aquat. Sci. Downloaded from cdnsciencepub.com by UNIV CALGARY on 04/10/21 T4 buffer, 0.05 L of BSA (1 mg·mL ), and 0.1675 L of T4 DNA ligase. cell. Ligation conditions were held at 16 °C for 6 h. Reaction products of individual samples were pooled together into groups of nine for Genetic bioinformatic analysis fragment size-selection (step iii) using SPRIselect beads from Beck- Sequence data were analyzed using the Stacks 1.32 software man Coulter. This selected distribution was created using left-side (Catchen et al. 2011, 2013). Raw reads were processed using and right-side ratios of 0.8x and 0.61x, respectively, to create a process_radtags, which (i) identifies the unique EcoRI barcodes normal distribution around 400 base pairs (bp) in length. and restriction site associated DNA (RAD) sites and (ii) performs a Final step (iv) PCR and sequencing were performed as follows. sliding window analysis to assess the quality score of reads. Se- Pooled samples were amplified in triplicate in 20 ␮L polymerase quences were trimmed to include only the first 150 bp as quality chain reactions (PCR) that consisted of 10 ␮L of 2× NEB Q5 High- scores decreased below the Q30 threshold. All individuals were ␮ ␮ Fidelity Master Mix, 5.34 LofddH2O, 4 L of size-selected pooled parsed per their unique sequence identifier. Resulting file sizes samples, and 0.66 ␮Lof10␮mol·L–1 primers that included the P5 ranged between 2 and 220 MB, with the bottom 5% of file sizes and P7 Illumina primer sequence. These primers amplified the removed to exclude individuals with minimal sequence data. Se- size-selected fragments, in addition to adding the unique P5 quences were aligned de novo using ustacks (Byrne et al. 2013; and P7 adapters necessary for fragments to bind to the sequenc- Everett and Seeb 2014), as the walleye genome has yet to be pub- ing flow cell. Amplified samples were quantified using a Qubit lished. For de novo assembly, a bounded model was used, using

Published by NRC Research Press Allen et al. 1647

alpha = 0.1, lower epsilon bound = 0, upper epsilon bound = 0.04, Table 2. Population, year samples were collected (Year), number of and a minimum depth of coverage of 4×. The upper epsilon individuals analyzed (N), number of private alleles (PA), observed

boundary was chosen based on the average sequencing error rate heterozygosity (Ho), expected heterozygosity (He), nucleotide diversity ␲ of the NextSeq 500, which was reported in Illumina’s Sequence ( ), percent polymorphic loci (% Poly), and inbreeding coefficient (FIS) Analysis Viewer version 1.9.1. A catalog of loci was built using in six walleye populations in Alberta, Canada. cstacks, which created a set of consensus loci, while stacks con- ␲ Population Year N PA Ho He % Poly FIS structed from ustacks were searched against the catalog of con- sensus loci using sstacks. After the creation of the catalog, Gods 2005 14 58 0.090 0.096 0.101 0.211 0.027 rxstacks was used to filter poor-quality stacks and filter putative Graham 2004 18 70 0.116 0.119 0.123 0.228 0.020 Rainbow 2002 20 184 0.092 0.094 0.097 0.186 0.011 sequencing errors. Poor-quality stacks were identified based on Round 2004 19 21 0.093 0.100 0.104 0.163 0.027 their assigned log-likelihood (LL) estimates, and stacks were re- Smoke 2005 18 13 0.085 0.104 0.108 0.137 0.050 moved from further analysis if they had an LL < –50. The cut-off Vandersteene 2000 15 45 0.108 0.108 0.113 0.208 0.012 value of –50 was chosen to maximize the number of SNPs in the final data set while filtering out low-quality stacks. Based on population-wide examination, nonsignificant nucleotides were recalled using the same outlined parameters in ustacks. The new were calculated for both the contemporary and historic popula- filtered data set was rerun through both cstacks and sstacks to tions using GenoDive (Meirmans and van Tienderen 2004), using create a final catalog of loci. Putative loci were analyzed using the the Weir and Cockerham method (Weir and Cockerham 1984). To program populations, which further filtered loci on the following correct for multiple pairwise comparisons, a Bonferroni correc- criteria: SNPs were required to be present in four of six popula- tion was applied (Dunn 1961). An analysis of molecular variance tions, in at least 65% of individuals, have a minimum minor allele (AMOVA) was performed using the R package Poppr (Kamvar et al. 2014) to confirm the DAPC results and assess hierarchical struc- frequency ≥0.01, and exhibit an FIS > –0.3. This threshold was chosen to filter out loci that had higher levels of heterozygosity turing of populations within river basins. than expected under a random mating model. This helped control for deviations from Hardy–Weinberg equilibrium. Results Samples from Smoke Lake, collected in both 1973 and 2005, Sequencing were analyzed together in a separate analysis using the same pa- One flow cell of sequencing produced approximately 174 mil- rameters as the above protocol. However, filtered loci must be lion sequenced nucleotide reads with more than 145 million reads present in both time periods. This was done to maximize the retained after filtering out reads with Q-scores ≤ Q30. Down- number of SNPs identified between the historical and contempo- stream filtering steps were categorized into two analyses: (i) con- rary samples, as the DNA quality was lower in the historical sam- temporary population structure analysis and (ii) comparing ples. historical and contemporary estimates of genetic diversity in Smoke Lake. A total of 104 individuals across six populations had Characterization of phenotypes and genetic diversity sufficient sequencing data to assess contemporary population analyses structure; and 13 individuals from the historical (1973) time period A Friedman’s rank test (Friedman 1937) was performed in R and 19 individuals from the contemporary time period (2005) of (R Core Team 2016) using the package PMCMR (Pohlert 2014)to Smoke Lake had sufficient data to estimate changes in genetic For personal use only. assess genetic differences among contemporary populations and diversity (Table 2). After applying our stringent filtering thresh- differences between temporal periods in Smoke Lake using four olds, a total of 1081 loci passed filter to assess genetic population genetic characteristics (e.g., number of private alleles, observed structure, and 428 loci passed filter to estimate changes in genetic heterozygosity, nucleotide diversity, and percent polymorphic). diversity between 1973 and 2005 in Smoke Lake. In addition, an ANOVA was performed to assess if length distribu- tions changed between temporal periods, with respect to age, in Characterization of phenotypes and genetic diversity Smoke Lake. An ANOVA was chosen because von Bertalanffy There was large variation in the number of private alleles (13– growth curves were unable to be fit due to low replication at age 184), observed heterozygosity (0.085–0.116), nucleotide diversity classes, resulting in linearization of the data. (0.097–0.123), and percent polymorphic loci (13.8–22.76%; Table 2) between contemporary lakes; however, there was no significant Population structure analyses difference among lakes (␹2 = 10.714, df = 5, p value = 0.057). We Population structure was assessed using Adegenet (Jombart observed a significant decrease in the number of private alleles 2008) and Admixture (Alexander et al. 2009). Adegenet uses dis- (296, 119), observed heterozygosity (0.075, 0.026), percent poly- criminant analysis of principal components (DAPC) to estimate morphic loci (47.58%, 20.32%), and nucleotide diversity (0.0786, the optimal number of genetic clusters (K). A DAPC analysis opti- 0.0302) between 1973 and 2005 (respectively) in Smoke Lake (␹2 =

Can. J. Fish. Aquat. Sci. Downloaded from cdnsciencepub.com by UNIV CALGARY on 04/10/21 mizes the variation between groups while minimizing the varia- 5, df = 1, p value = 0.025). The random sample of individuals col- tion within groups using discriminate functions. The number of lected from Smoke Lake in 1973 and 2005 had respective similar genetic clusters was not known; therefore, the optimal number of minimum (7, 2), maximum (13, 15), and mean (9.15, 8.55) ages. clusters was identified with K-means clustering using Bayesian There was a significant difference in fork lengths between 1973 information criterion (BIC) for K of 1–100. The optimal value of K and 2005 (mean = 496 ± 23.1 mm versus 394 ± 64.2 mm, respec- was identified as the value with the lowest BIC estimate. Admix- tively; ANOVA, df = 1, p value < 0.001), but there was no interaction ture uses a maximum likelihood approach using a block relax- between age and time period (ANOVA, df = 1, p value = 0.065). ation algorithm to calculate population structure. To determine the optimal value for K, Admixture uses a cross-validation (CV) Population structure approach similar to the program fastPHASE (Scheet and Stephens Adegenet estimated the number of genetic clusters as K =5 2006). K values of 1–6 were assessed, with the lowest cross- (BIC = 390.57) using 30 principal components and four discrimi- validation score chosen as the optimal K. Principal component nant functions. Discriminant functions 1 and 2 explained the analysis and K-means clustering group observations together most variation between clusters, 76.79% and 13.80%, respectively, if there are missing data at the same loci (Meirmans and and were the discriminant functions used for visualization. Pop- van Tienderen 2004). Therefore, missing data were filled in based ulations were clustered by individual lakes except for Graham

on the overall allele frequencies at each locus. Pairwise FST values Lake and Vandersteene Lake (Fig. 2) with pairwise FST values rang-

Published by NRC Research Press 1648 Can. J. Fish. Aquat. Sci. Vol. 75, 2018

Fig. 2. Scatter plot of principal components from a discriminant Table 3. Pairwise FST estimates for six lakes in Alberta, Canada, using analysis of principal components (DAPC) analysis where K = 5 using 1081 single nucleotide polymorphism (SNP) loci. Adegenet. Discriminant functions 1 and 2 are represented on the x Gods Graham Rainbow Round Smoke Vandersteene and y axes, respectively. Genetic clusters identify individual lakes except for Vandersteene Lake and Graham Lake (circles). Rainbow Gods Lake (triangles) is within the Hay River basin, and the remaining Graham 0.266 four clusters reside in the Peace– basin. Rainbow 0.592 0.566 Round 0.322 0.133 0.613 Smoke 0.396 0.352 0.609 0.375 Vandersteene 0.281 0.014 0.581 0.134 0.332 Note: All populations are significantly differentiated after implementing a Bonferroni correction (␣ = 0.003).

history affects the genetic structure of individual lakes, while colonization history acts on a larger geographic scale. These inter- acting factors would result in a complex genetic structure. To manage genetic diversity effectively in this fishery, lakes should have individual management objectives, aimed at preserving lo- cal genetic diversity, that are contextualized by the genetic struc- ture of these populations across the landscape. There was a significant decline in observed heterozygosity, nu- cleotide diversity, number of private alleles, and polymorphic loci between 1973 and 2005 in Smoke Lake (␹2 =5,df=1,p value = 0.025; Table 1). This aligns with previous research documenting the decline of genetic diversity in association with harvest (Hauser et al. 2002; van Wijk et al. 2013) and decreased catch rates in other species. In addition, there was a significant reduction in fork length between 1973 and 2005 in Smoke Lake (Table 1). The reduc- tion in genetic diversity and fish lengths corresponds with the known population collapse in Smoke Lake. Collapse may have been influenced by stochastic environmental conditions (e.g., pro- longed anoxic conditions, toxic pollution, and virulent disease), but harvest during this period likely played a primary role. Fisheries-induced evolution (Heino et al. 2015), as a result of a ing between 0.014 and 0.613 (Table 3). There was clear differenti- size-selective harvest pressure, could explain the corresponding ation between genetic clusters except for Round Lake and the reductions in both observed phenotypes and genetic diversity. cluster of Graham Lake and Vandersteene Lake. Based on K =5,the Experimental studies have shown strong size-selective harvest

For personal use only. AMOVA results indicated a significant proportion of the contem- pressure can reduce the length of fish (e.g., Conover and Munch porary variation was explained by river basins (36.29%, p value = 2002; Conover and Baumann 2009) and result in corresponding 0.001), populations within river basins (7.85%, p value = 0.001), and genetic changes (van Wijk et al. 2013; Heino et al. 2015). If suffi- individuals (55.86%, p value = 0.001). Admixture identified the cient genetic variability remains in the population, reductions in number of genetic clusters as K = 2 (CV = 0.339), where populations length can be reversed after approximately 12 generations of no clustered by river basins (Fig. 3). Analyzing only the 1973 and 2005 harvest (Conover et al. 2009). Closure of a fishery for decades is samples from Smoke Lake, both Adegenet and Admixture esti- impractical in many situations. However, these results reinforce mated the number of genetic clusters as K = 1, with a pairwise FST past studies that highlight the importance of incorporating genet- value of –0.00025 (p value = 0.480) between the two time periods. ics into management strategies that can inform managers of the evolutionary trajectories of their fishery (Spies and Punt 2015; Discussion Westgaard et al. 2017). The six contemporary walleye lakes were grouped in either two DNA quality was low in all samples, resulting in relatively or five genetic clusters, though estimates of genetic diversity were few SNPs identified compared with studies using high-quality similar among lakes (Table 2). Adegenet identified lakes as unique genomic DNA (e.g., Sachidanandam et al. 2001; Kjeldsen et al. clusters (Fig. 2) except for Graham Lake and Vandersteene Lake. 2016). Archived scale samples were dried and stored in envelopes,

Can. J. Fish. Aquat. Sci. Downloaded from cdnsciencepub.com by UNIV CALGARY on 04/10/21 These two lakes are significantly differentiated (FST = 0.014, leading to DNA fragmentation and (or) degradation. We at- p value = 0.002), but low FST estimates are supported by their geo- tempted to sequence an additional historical population for this graphic proximity (ϳ2 km) and connection via a small stream. study for which samples had been archived using this same

Alternatively, low FST estimates between these lakes could reflect method. Samples from that population did not amplify with suf- historical events; this could include both lakes being colonized by ficient success to be useable. Yet, our results are consistent with the same source population or perhaps a single historical origin an increasing number of studies that demonstrate genotype-by- for both lakes. The genetic cluster for Round Lake exhibited sub- sequencing methods to be more robust when comparing histori- stantial overlap with the Graham Lake and Vandersteene Lake cal with contemporary samples (Morin et al. 2004; Speller et al.

cluster. This overlap, in addition to low pairwise FST estimates 2012). The 428 shared loci identified between the 1973 and 2005 between these lakes (Table 3), suggests Round Lake may have a time periods are robust due to hygienic laboratory protocols and similar colonization history or historical source as Graham Lake stringent filtering steps that only included SNPs identified in both and Vandersteene Lake. Interestingly, Admixture identified two time periods. Phenotypic information was limited for historical genetic clusters with lakes categorized by river basin (Fig. 3). Clus- samples, preventing modeling of growth curves. To understand tering by river basin is support by the AMOVA, which estimates a long-term trajectories of fisheries, there needs to be effective ar- significant portion of the variation could be explained by river chiving protocols for both genetic and phenotypic data. Commu- basin (36% variation, p value = 0.001). Management and stocking nication with local fishery managers will be essential in

Published by NRC Research Press Allen et al. 1649

Fig. 3. Ancestry assignment using K = 2 for six walleye lakes in Alberta, Canada. Lakes are genetically clustered by river basin (black = Hay River basin, grey = Peace–Slave River basin). For personal use only.

developing long-term archiving methods that can facilitate future References research. Aitken, S.N., Yeaman, S., Holliday, J.A., Wang, T., and Curtis-McLane, S. 2008. In conclusion, the loss of genetic variability over time, and com- Adaptation, migration or extirpation: climate change outcomes for tree pop- plex genetic structure of walleye lakes in Alberta, indicate the ulations. Evol. Appl. 1(1): 95–111. doi:10.1111/j.1752-4571.2007.00013.x. PMID: need for incorporating SNPs into current genetic management 25567494. strategies. Alternative genetic tools can be informative, and Alexander, D.H., Novembre, J., and Lange, K. 2009. Fast model-based estimation of ancestry in unrelated individuals. Genome Res. doi:10.1101/gr.094052.109. widely accessible, for certain management questions and species. Allen, B.E., Anderson, M.L., Mee, J.A., Coombs, M., and Rogers, S.M. 2016. Role However, SNPs provide a viable approach for understanding a of genetic background in the introgressive hybridization of rainbow trout large range of genetic questions important to managers, even for (Oncorhynchus mykiss) with Westslope cutthroat trout (O. clarkii lewisi). Conserv. species that lack current genetic information. Our results suggest Genet. 17: 521. doi:10.1007/s10592-015-0801-6. Allendorf, F.W., England, P.R., Luikart, G., Ritchie, P.A., and Ryman, N. 2008.

Can. J. Fish. Aquat. Sci. Downloaded from cdnsciencepub.com by UNIV CALGARY on 04/10/21 that management needs to identify the genetic population struc- Genetic effects of harvest on wild animal populations. Trends Ecol. Evol. ture of walleye and develop a long-term storage of genetic and 23(6): 327–337. doi:10.1016/j.tree.2008.02.008. PMID:18439706. phenotypic data, as this could inform lake management and aid Araki, H., Cooper, B., and Blouin, M.S. 2007. Genetic effects of captive breeding sustainable fisheries. Such efforts will improve the long-term sta- cause a rapid, cumulative fitness decline in the wild. Science, 318(5847): bility of the walleye fishery and help minimize the evolutionary 100–103. doi:10.1126/science.1145621. PMID:17916734. consequences of size-selective harvest. Berry, D.K. 1995. Alberta’s walleye management and recovery plan. Report num- ber T/310 of Alberta Environment Protection, Natural Resources Service, Ed- Acknowledgements monton, Alberta, Canada. Billington, N. 1996. Geographical distribution of mitochondrial DNA (mtDNA) We thank AEP biologists S. Spencer, B. Lucko, M. Banko, and variation in walleye, sauger, and yellow perch. Ann. Zool. Fenn. 33: 699–706. M. Brown for assistance in collecting tissue samples and providing Bozek, M.A., Haxton, T.J., and Raabe, J.K. 2011. Walleye and sauger habitat. In access to their archival samples. We also thank members of the Biology, management and culture of walleye and sauger. pp. 133–197. Rogers and Vamosi labs for their helpful comments and P. Peller Byrne, S., Czaban, A., Studer, B., Panitz, F., Bendixen, C., and Asp, T. 2013. for the creation of Fig. 1. This research was supported by a Natural Genome Wide Allele Frequency Fingerprints (GWAFFs) of populations via genotyping by sequencing. PLoS ONE, 8(3): e57438. doi:10.1371/journal.pone. Sciences and Engineering Research Council of Canada Discovery 0057438. PMID:23469194. Grant to S.M.R. and by the Alberta Conservation Association Bio- Catchen, J.M., Amores, A., Hohenlohe, P., Cresko, W., Postlethwait, J.H., and diversity Program Grant to B.E.A. De Koning, D.-J. 2011. Stacks: Building and genotyping loci de novo from

Published by NRC Research Press 1650 Can. J. Fish. Aquat. Sci. Vol. 75, 2018

short-read sequences. G3 Genes, Genomes, Genet. 1(3): 171–182. doi:10.1534/ Pisces): the positive sides of a long generation time. Mol. Ecol. 15(7): 1769–1780. g3.111.000240. doi:10.1111/j.1365-294X.2006.02902.x. PMID:16689897. Catchen, J., Hohenlohe, P.A., Bassham, S., Amores, A., and Cresko, W.A. 2013. Mackay, W.C., Ash, G.R., and Norris, H.J. 1990. Fish ageing methods for Alberta. Stacks: an analysis tool set for population genomics. Mol. Ecol. 22(11): 3124– R.L & L. Environmental Services Ltd. In association with Alberta Fish and 3140. doi:10.1111/mec.12354. PMID:23701397. Wildlife Division and University of Alberta, Edmonton. Cena, C.J., Morgan, G.E., Malette, M.D., and Heath, D.D. 2006. Inbreeding, out- McCracken, K.G., Johnson, W.P., and Sheldon, F.H. 2001. Molecular population breeding and environmental effects on genetic diversity in 46 walleye (Sander genetics, phylogeography, and conservation biology of the mottled duck vitreus) populations. Mol. Ecol. 15(2): 303–320. doi:10.1111/j.1365-294X.2005. (Anas fulvigula). Conserv. Genet. 87(2): 87–102. doi:10.1023/A:1011858312115. 02637.x. PMID:16448402. McParland, T.L., Ferguson, M.M., and Liskauskas, A.P. 1999. Genetic population Coltman, D.W., O’Donoghue, P., Jorgenson, J.T., Hogg, J.T., Strobeck, C., and structure and mixed-stock analysis of Walleyes in the Lake Erie–Lake Huron Festa-Bianchet, M. 2003. Undesirable evolutionary consequences of trophy corridor using allozyme and mitochondrial DNA markers. Trans. Am. Fish. hunting. Nature, 426: 655–658. doi:10.1038/nature02177. PMID:14668862. Soc. 128: 1055–1067. doi:10.1577/1548-8659(1999)128<1055:GPSAMS>2.0.CO;2. Conover, D.O., and Baumann, H. 2009. The role of experiments in understanding Meirmans, P.G., and Van Tienderen, P.H. 2004. GENOTYPE and GENODIVE: Two fishery-induced evolution. Evol. Appl. 2(3): 276–290. doi:10.1111/j.1752-4571. programs for the analysis of genetic diversity of asexual organisms. Mol. Ecol. 2009.00079.x. PMID:25567880. Notes, 4(4): 792–794. doi:10.1111/j.1471-8286.2004.00770.x. Conover, D.O., and Munch, S. 2002. Sustaining fisheries yields over evolutionary Milligan, B.G., Leebens-Mack, J., and Strand, A.E. 1994. Conservation genetics: time scales. Science, 297: 94–96. doi:10.1126/science.1074085. PMID:12098697. beyond the maintenance of marker diversity. Mol. Ecol. 3: 423–435. doi:10. Conover, D.O., Munch, S.B., and Arnott, S.A. 2009. Reversal of evolutionary 1111/j.1365-294X.1994.tb00082.x. downsizing caused by selective harvest of large fish. Proc. R. Soc. B Biol. Sci. Morán, P., Pendás, A.M., Garcia-Vázquez, E., and Izquierdo, J. 1991. Failure of a 276(1664): 2015–2020. doi:10.1098/rspb.2009.0003. PMID:19324761. stocking policy, of hatchery reared brown trout, Salmo trutta L., in Asturias, DeSalle, R., and Amato, G. 2004. The expansion of conservation genetics. Nat. Spain, detected using LDH-5 as a genetic marker. J. Fish Biol. 39: 117–121. Rev. Genet. 5(9): 702–712. doi:10.1038/nrg1425. PMID:15372093. doi:10.1111/j.1095-8649.1991.tb05075.x. Dunn, O.J. 1961. Multiple comparisons among means. J. Am. Stat. Assoc. 56: Morin, P.A., Luikart, G., Wayne, R.K., and the SNP workshop group. 2004. SNPs in 52–64. doi:10.1080/01621459.1961.10482090. ecology, evolution and conservation. Trends Ecol. Evol. 19(4): 208–216. doi: Everett, M.V., and Seeb, J.E. 2014. Detection and mapping of QTL for temperature 10.1016/j.tree.2004.01.009. tolerance and body size in Chinook salmon (Oncorhynchus tshawytscha) using Naish, K.A., Taylor, J.E., Levin, P.S., Quinn, T.P., Winton, J.R., Huppert, D., and genotyping by sequencing. Evol. Appl. 7(4): 480–492. doi:10.1111/eva.12147. Hilborn, R. 2007. An evaluation of the effects of conservation and fishery PMID:24822082. enhancement hatcheries on wild populations of salmon. Adv. Mar. Biol. 53(7): 61–194. doi:10.1016/S0065-2881(07)53002-6. PMID:17936136. Friedman, M. 1937. The use of ranks to avoid the assumption of normality implicit in the analysis of variance. J. Am. Stat. Assoc. 32(200): 675–701. doi: Neville, H.M., and Dunham, J.B. 2011. Patterns of hybridization of nonnative 10.1080/01621459.1937.10503522. cutthroat trout and hatchery rainbow trout with native redband trout in the Boise River, Idaho. N. Am. J. Fish. Manage. 31(6): 1163–1176. doi:10.1080/02755947. Hamilton, J.A., and Miller, J.M. 2016. Adaptive introgression: a resource for man- 2011.647252. agement and genetic conservation in a changing climate. Conserv. Biol. 30(1): Pearse, D.E., and Crandall, K.A. 2004. Beyond FST: Analysis of population genetic 33–41. doi:10.1111/cobi.12574. PMID:26096581. data for conservation. Conserv. Genet. 5(5): 585–602. doi:10.1007/s10592-003- Haponski, A.E., and Stepien, C.A. 2014. A population genetic window into the 1863-4. past and future of the walleye (Sander vitreus): relation to historic walleye and Pitcher, T.J. 2001. Fisheries managed to rebuild ecosystems? Reconstructing the the extinct “blue pike” S. v. “glaucus”. BMC Ecol. Biol. 14: 133. doi:10.1186/1471- past to salvage the future. Ecol. Appl. 11(2): 601–617. doi:10.1890/1051- 2148-14-133. 0761(2001)011[0601:FMTRER]2.0.CO;2. Haponski, A.E., and Stepien, C.A. 2016. Two decades of genetic consistency in a Pohlert, T. 2014. The Pairwise Multiple Comparison of Mean Ranks Package reproductive population in the face of exploitation: patterns of adult and (PMCMR) [online]. R package. Available from http://CRAN.R-project.org/ larval walleye (Sander vitreus) from Lake Erie’s Maumee River. Conserv. Genet. package=PMCMR. 17: 1345–1362. doi:10.1007/s10592-016-0866-x. Post, J.R., Sullivan, M., Cox, S., Lester, N.P., Walters, C.J., Parkinson, E.A., Haponski, A.E., Dean, H., Blake, B.E., and Stepien, C.A. 2014. Genetic history of Paul, A.J., Jackson, L., and Shuter, B.J. 2002. Canada’s recreational fisheries: walleyes spawning in Lake Erie’s Cattaraugus Creek: a comparison of pre- the invisible collapse? Fisheries, 27(1): 6–17. doi:10.1577/1548-8446(2002)027 and poststocking. Trans. Am. Fish. Soc. 143: 1295–1307. doi:10.1080/00028487.

For personal use only. <0006:CRF>2.0.CO;2. 2014.935477. R Core Team. 2016. R: a language and environment for statistical computing Hauser, L., Adcock, G.J., Smith, P.J., Ramiréz, J.H.B., and Carvalho, G.R. 2002. Loss [online]. R Foundation for Statistical Computing, Vienna, Austria. Available of microsatellite diversity and low effective population size in an overex- from https://www.R-project.org/. ploited population of New Zealand snapper (Pagrus auratus). Proc. Natl. Acad. Sachidanandam, R., Weissman, D., Schmidt, S.C., Kakol, J.M., Stein, L.D., Sci. U.S.A. 99(18): 11742–11747. doi:10.1073/pnas.172242899. PMID:12185245. Marth, G., Sherry, S., Mullikin, J.C., Mortimore, B.J., Willey, D.L., Hunt, S.E., Hedrick, P.W., and Kalinowski, S.T. 2000. Inbreeding depression in conservation Cole, C.G., Coggill, P.C., Rice, C.M., Ning, Z., Rogers, J., Bentley, D.R., biology. Annu. Rev. Ecol. Syst. 31: 139–162. doi:10.1146/annurev.ecolsys.31.1. Kwok, P.Y., Mardis, E.R., Yeh, R.T., Schultz, B., Cook, L., Davenport, R., 139. Dante, M., Fulton, L., Hillier, L., Waterston, R.H., McPherson, J.D., Gilman, B., Heino, M., Pauli, B.D., and Dieckmann, U. 2015. Fisheries-induced evolution. Schaffner, S., Van Etten, W.J., Reich, D., Higgins, J., Daly, M.J., Blumenstiel, B., Annu. Rev. Ecol. Evol. Syst. 46: 461–481. doi:10.1146/annurev-ecolsys-112414- Baldwin, J., Stange-Thomann, N., Zody, M.C., Linton, L., Lander, E.S., and 054339. Altshuler, D. 2001. A map of human genome sequence variation containing Johnston, F.D., and Paul, A.J. 2006. Review and assessment of walleye genetics 1.42 million single nucleotide polymorphisms. Nature, 409(6822): 928–933. and stocking in Alberta. Alberta Conservation Association. doi:10.1038/35057149. PMID:11237013. Jombart, T. 2008. Adegenet: a R package for the multivariate analysis of genetic Scheet, P., and Stephens, M. 2006. A fast and flexible statistical model for large- markers. Bioinformatics, 24(11): 1403–1405. doi:10.1093/bioinformatics/btn129. scale population genotype data: applications to inferring missing genotypes PMID:18397895. and haplotypic phase. Am. J. Hum. Genet. 78: 629–644. doi:10.1086/502802. Kamvar, Z.A., Tabima, J.F., and Grüwald, N.J. 2014. Poppr: an R package for PMID:16532393. genetic analysis of populations with clonal, partially clonal, and/or sexual Segelbacher, G., Cushman, S.A., Epperson, B.K., Fortin, M.J., Francois, O.,

Can. J. Fish. Aquat. Sci. Downloaded from cdnsciencepub.com by UNIV CALGARY on 04/10/21 reproduction. PeerJ. 2: e281. doi:10.7717/peerj.281. PMID:24688859. Hardy, O.J., Holderegger, R., Taberlet, P., Waits, L.P., and Manel, S. 2010. Kjeldsen, S.R., Zenger, K.R., Leigh, K., Ellis, W., Tobey, J., Phalen, D., Melzer, A., Applications of landscape genetics in conservation biology: concepts and FitzGibbon, S., and Raadsma, H.W. 2016. Genome-wide SNP loci reveal novel challenges. Conserv. Genet. 11(2): 375–385. doi:10.1007/s10592-009-0044-5. insights into koala (Phascolarctos cinereus) population variability across its Speller, C.F., Hauser, L., Lepofsky, D., Moore, J., Rodrigues, A.T., Moss, M.L., range. Conserv. Genet. 17(2): 337–353. doi:10.1007/s10592-015-0784-3. McKechnie, I., and Yang, D.Y. 2012. High potential for using DNA from an- Kleiman, D.G. 1989. Reintroduction of captive mammals for conservation. Bio- cient Herring bones to inform modern fisheries management and conserva- science, 39(3): 152–161. doi:10.2307/1311025. tion. PLoS ONE, 8(7): e51122. doi:10.1371/journal.pone.0051122. Labonne, J., Kaeuffer, R., Guéraud, F., Zhou, M., Manicki, A., and Hendry, A.P. Spencer, S. 2010. The increasing prevalence of smaller fish in highly exploited 2016. From the bare minimum: genetics and selection in populations fisheries: concerns, diagnosis and management solutions. Ph.D. thesis, De- founded by only a few parents. Evol. Ecol. Res. 17(1): 21–34. partment of Renewable Resources, University of Alberta, Alberta, Canada. Lamaze, F.C., Sauvage, C., Marie, A., Garant, D., and Bernatchez, L. 2012. Dynam- Spies, I., and Punt, A.E. 2015. The utility of genetics in marine fisheries manage- ics of introgressive hybridization assessed by SNP population genomics of ment: a simulation study based on Pacific cod off Alaska. Can. J. Fish. Aquat. coding genes in stocked brook charr (Salvelinus fontinalis). Mol. Ecol. 21(12): Sci. 72(9): 1415–1432. doi:10.1139/cjfas-2014-0050. 2877–2895. doi:10.1111/j.1365-294X.2012.05579.x. PMID:22548328. Strange, R.M., and Stepien, C.A. 2007. Genetic divergence and connectivity Leberg, P.L., and Firmin, B.D. 2008. Role of inbreeding depression and purging in among river and reef spawning groups of Walleye (Sander vitreus vitreus)in captive breeding and restoration programmes. Mol. Ecol. 17(1): 334–343. doi: Lake Erie. Can. J. Fish. Aquat. Sci. 64(3): 437–448. doi:10.1139/f07-022. 10.1111/j.1365-294X.2007.03433.x. PMID:18173505. Sullivan, M.G. 2003. Active management of walleye fisheries in Alberta: dilem- Lippé, C., Dumont, P., and Bernatchez, L. 2006. High genetic diversity and no mas of managing recovering fisheries. N. Am. J. Fish. Manage. 23(4): 1343– inbreeding in the endangered copper redhorse, Moxostoma hubbsi (Catostomidae, 1358. doi:10.1577/M01-232AM.

Published by NRC Research Press Allen et al. 1651

Uusi-Heikkilä, S., Whiteley, A.R., Kuparinen, A., Matsumura, S., Venturelli, P.A., Mol. Ecol. Resour. 10(5): 785–796. doi:10.1111/j.1755-0998.2010.02876.x. PMID: Wolter, C., Slate, J., Primmer, C.R., Meinelt, T., Killen, S.S., Bierbach, D., 21565090. Polverino, G., Ludwig, A., and Arlinghaus, R. 2015. The evolutionary legacy of Waples, R.S. 2015. Testing for Hardy–Weinberg proportions: Have we lost the size-selective harvesting extends from genes to populations. Evol. Appl. 8: plot? J. Hered. 106(1): 1–19. doi:10.1093/jhered/esu062. PMID:25425676. 597–620. doi:10.1111/eva.12268. PMID:26136825. Weir, B.S., and Cockerham, C.C. 1984. Estimating F-statistics for the analysis of 38 van Wijk, S.J., Taylor, M.I., Creer, S., Dreyer, C., Rodrigues, F.M., Ramnarine, I.W., population structure. Evolution (N.Y.), (6): 1358–1370. doi:10.1111/j.1558- 5646.1984.tb05657.x. Van Oosterhout, C., and Carvalho, G.R. 2013. Experimental harvesting of fish Westgaard, J., Saha, A., Kent, M., Hansen, H., Knutsen, H., Hauser, L., Cadrin, S., populations drives genetically based shifts in body size and maturation. Albert, O.T., and Johansen, T. 2017. Genetic population structure in Green- Front. Ecol. Environ. 11(4): 181–187. doi:10.1890/120229. land halibut (Reinhardtius hippoglossoides) and its relevance to fishery manage- Walker, J.R., Foote, L., and Sullivan, M.G. 2007. Effectiveness of enforcement to ment. Can. J. Fish. Aquat. Sci. 74(4): 475–485. doi:10.1139/cjfas-2015-0430. deter illegal angling harvest of northern pike in Alberta. N. Am. J. Fish. Yau, M.M., and Taylor, E.B. 2013. Environmental and anthropogenic correlates of Manage. 27(4): 1369–1377. doi:10.1577/M06-011.1. hybridization between westslope cutthroat trout (Oncorhynchus clarkii lewisi) Waples, R.S. 2010. Spatial–temporal stratifications in natural populations and and introduced rainbow trout (O. mykiss). Conserv. Genet. 14(4): 885–900. how they affect understanding and estimation of effective population size. doi:10.1007/s10592-013-0485-8. For personal use only. Can. J. Fish. Aquat. Sci. Downloaded from cdnsciencepub.com by UNIV CALGARY on 04/10/21

Published by NRC Research Press