Yellow perch genetic structure and habitat use among connected habitats in eastern Lake Michigan

Abstract Maintenance of genetic and phenotypic diversity is widely recognized as an important conservation priority, yet managers often lack basic information about spatial patterns of population structure and its relationship with habitat heterogeneity and species movement within it. To address this knowledge gap, we focused on the economically and ecologically prominent yellow perch (Perca flavescens). In the Lake Michigan basin, yellow perch reside in nearshore Lake Michigan, including drowned river mouths (DRMs)—protected, lake‐like habitats that link tributaries to Lake Michigan. The goal of this study was to examine the extent that population structure is associated with Great Lakes connected habitats (i.e., DRMs) in a mobile fish species using yellow perch as a model. Specifically, we tested whether DRMs and eastern Lake Michigan constitute distinct genetic stocks of yellow perch, and if so, whether those stocks migrate between the two connected habitats throughout the year. To do so, we genotyped yellow perch at 14 microsatellite loci collected from 10 DRMs in both deep and littoral habitats during spring, summer, and autumn and two nearshore sites in Lake Michigan (spring and autumn) during 2015–2016 and supplemented our sampling with fish collected in 2013. We found that yellow perch from littoral‐DRM habitats were genetically distinct from fish captured in nearshore Lake Michigan. Our data also suggested that Lake Michigan yellow perch likely use deep‐DRM habitats during autumn. Further, we found genetic structuring among DRMs. These patterns support hypotheses of fishery managers that yellow perch seasonally migrate to and from Lake Michigan, yet, interestingly, these fish do not appear to interbreed with littoral fish despite occupying the same DRM. We recommend that fisheries managers account for this complex population structure and movement when setting fishing regulations and assessing the effects of harvest in Lake Michigan.


| INTRODUC TI ON
Understanding population genetic structure is crucial for conservation and management of fishes. Many fish species represent genetically and phenotypically diverse stocks (subpopulations with attributes relevant to their management) that were not recognized in many species before molecular population genetics was applied to fisheries management (Begg, Friedland, & Pearce, 1999;Stephenson, 1999). This diversity is important to fisheries management because it can stabilize populations of exploited species to natural and human-induced disturbances (Schindler et al., 2010;Schindler, Armstrong, & Reed, 2015). For example, historically minor producing stocks of the sockeye salmon (Oncorhynchus nerka) fishery in Bristol Bay, Alaska, became the dominant producers of the fishery after climate warming in the past 20 years, highlighting the need to conserve stock diversity in exploited fishes (Hilborn, Quinn, Schindler, & Rogers, 2003). However, most fish stocks are not as easy to identify as salmonids; therefore, population genetics is central for identifying cryptic stocks (distinct stocks of an exploited species residing in sympatry) in the sustainable management of exploited fishes.
Cryptic stock sorting may occur in species whose ranges span a variety of connected habitats that allow distinct stocks to reside in sympatry during different seasons or stages of the life cycle (Brenden et al., 2015;Hilborn et al., 2003;Wilson, Liskauskas, & Wozney, 2016). This is relevant to management, because unexpected harvest of a specific stock can occur when the stock is harvested in a location that is used seasonally and/or during a specific life stage migration. For instance, stocks of walleye (Sander vitreus) from Lake St. Clair migrate into Saginaw Bay, Lake Huron. Genetic analysis of harvested walleye indicated that 26% of the total catch taken from Saginaw Bay in 2008 and 2009 was Lake Huron and Lake St. Clair fish that were not accounted for in their respective management units (Brenden et al., 2015). Therefore, understanding cryptic stock sorting is especially important where different stocks can reside in sympatry during seasons where harvesting occurs.
Yellow perch (Perca flavescens) is an ecologically valuable and exploited fish species in North America that has long been a focus of fisheries managers (Scott & Crossman, 1973;Becker, 1983). Yellow perch normally spawn in the spring shortly after ice-out over a 1to 3-week timeframe, have relatively unspecialized requirements for spawning substrate in slow-moving or standing waters, and do not construct nests or guard their eggs and young (Becker, 1983).
The species once provided a prominent commercial fishery in the Laurentian Great Lakes (avg. 1.1 million kg/year in Lake Michigan during 1889-1970Becker, 1983). Yet, yellow perch suffered dramatic declines in recruitment in the late 1980s across the Great Lakes and commercial fishing has since slowed dramatically (Marsden & Robillard, 2004 Yellow perch has since remained at a much lower abundance in Lake Michigan than historical levels (Clapp & Dettmers, 2004). Yellow perch reside in both Lake Michigan and connecting drowned river mouths (DRMs; Janetski, Ruetz, Bhagat, & Clapp, 2013). DRMs are lake-like habitats that connect tributaries to Lake Michigan (Janetski & Ruetz, 2015), and receive inputs of water and nutrients from both the tributary and Lake Michigan (Larson et al., 2013;Liu et al., 2018;Wilcox et al., 2002). Recreational harvest of yellow perch in Lake Michigan and DRMs was managed as distinct units with lower limits in Lake Michigan (35 fish/day) than most DRMs (50 fish/day; MDNR, 2016), although a statewide harvest limit of 25 fish/day was enacted during the 2019 fishing season. We know that yellow perch exhibit genetic structure at large, geographic scales corresponding to recolonization of most of the Great Lakes from the Mississippian refugium and eastern lakes from the Atlantic refugium (Sepulveda-Villet & Stepien, 2012). At finer spatial scales, there also is structuring of yellow perch in Lake Erie and Lake Michigan (Glover, Dettmers, Wahl, & Clapp, 2008;Miller, 2003;Sepulveda-Villet, Stepien, & Vinebrooke, 2011). However, the question remains whether yellow perch in DRMs are genetically distinct stocks from Lake Michigan, and if so, whether these stocks ever reside in sympatry, such that harvest in one habitat might result in unaccounted take from a different stock.
Previous studies of yellow perch morphology, movement, and genetics provide evidence that DRMs may represent distinct stocks from Lake Michigan and that those stocks may mix at certain times of the year. Fisheries managers hypothesize that larger, lightercolored yellow perch (sometimes termed "white bellies") that are caught by recreational anglers in DRMs during autumn and winter are yellow perch that seasonally migrate from Lake Michigan into DRMs (Schneider, O'Neal, & Clark, 2007). Yellow perch may migrate into DRMs in late autumn and winter given the higher productivity of DRMs relative to nearshore Lake Michigan (Höök, Rutherford, Mason, & Carter, 2007;Janetski & Ruetz, 2015) for feeding and possibly spawning. These anecdotal accounts are corroborated by a recent otolith microchemistry study that found evidence for at least two types of yellow perch: resident DRM (which use DRM wetlands throughout life) and Lake Michigan fish that appear to return to DRM wetlands annually (Schoen, Student, Hoffman, Sierszen, & Uzarski, 2016). Morphological and genetic differences also were found between yellow perch captured in Lake Michigan and DRM wetlands (Parker, Stepien, Sepulveda-Villet, Ruehl, & Uzarski, 2009); however, the number of fish and DRMs sampled previously for genetic assessment was inadequate to answer whether true stock divisions exist.
Therefore, the existence of stock divisions in yellow perch between these connected habitats, and the extent that stocks may move between DRMs and Lake Michigan, remains unclear.
Our goal was to assess the stock structure of the exploited yellow perch in eastern Lake Michigan both spatially and temporally.
We aimed to address three questions: (a) Are Lake Michigan yellow perch genetically distinct stocks from DRM yellow perch? (b) If so, do yellow perch from Lake Michigan stocks use DRM habitats during any season where they may be harvested at the higher rate allowed in DRMs? and (c) Are yellow perch stocks distinct between DRM lakes? To evaluate these questions, we sampled fish for microsatellite analysis in deep and littoral habitats of DRMs in spring, summer, and autumn. We also sampled yellow perch in nearshore Lake Michigan in autumn and spring seasons.

| Field sites and sample collections
In an effort to sample possible resident and transient yellow perch in DRM lakes, we collected yellow perch in both deep and littoral habitats of 10 DRMs along the eastern shore of Lake Michigan ( Figure 1). We hypothesized that Lake Michigan fish use the more productive deep-DRM habitats (Höök et al., 2007;Janetski & Ruetz, 2015) during late autumn and winter (prior to the spawning season), when oxygen concentrations and temperatures of the deep-DRM are most similar to nearshore Lake Michigan (Altenritter, Wieten, Ruetz, & Smith, 2013;Biddanda et al., 2018;. Yellow perch from deep-DRM habitats were captured using 5.08-and 7.62-cm stretch-mesh gill nets placed on the bottom in the deepest part (range = 8.4-20.5 m) of each DRM where dissolved oxygen concentration was >2 mg/ml. Littoral-DRM habitats were sampled using boat electrofishing. We divided the shoreline of each DRM into 200-m numbered transects, and then, three transects were selected randomly and electrofished for 20 min. If the target number of yellow perch (40 individuals) was not achieved at the randomly selected transects, additional transects were chosen based on habitat (e.g., presence of submerged aquatic vegetation) to reach the target number of fish.
To account for temporal patterns of habitat use by yellow perch, we sampled fish from deep and littoral habitats of DRM lakes during spring, summer, and autumn 2015-2016, although not all DRMs or habitats were sampled in every season (see Table 1). We defined summer as when the DRM lake was thermally stratified, autumn F I G U R E 1 Map of eastern Lake Michigan showing drowned river mouth lakes sampled for yellow perch in deep and littoral habitats between summer 2015 and autumn 2016. Triangles indicate the two nearshore Lake Michigan sampling locations as after turnover (i.e., loss of thermal stratification) and before ice cover, and spring as after ice-out and before thermal stratification.
Yellow perch were captured in nearshore Lake Michigan by the Michigan Department of Natural Resources (MDNR) during late summer and spring seasons (2016) using gill nets and trawling as part of their biannual survey of yellow perch (Fitzgerald, Clapp, & Belonger, 2004;Fetzer et al., 2017). Two sites were sampled in nearshore Lake Michigan. One site was adjacent to the furthest north DRM, Charlevoix, and the other site was located between the two most southern DRMs, Macatawa and Muskegon ( Figure 1). We supplemented our sampling with yellow perch fin clips collected from northern and southern Lake Michigan by MDNR and from the deep-DRM habitat in Muskegon Lake during 2013 as part of another project (see Harris, Ruetz, Wieten, Altenritter, & Smith, 2017). These fin clips were genotyped with the fish collected for this study.

| MOLECUL AR ME THODS
We clipped tissue from the anal fin of each yellow perch. Clips were either stored in ethanol or dried in a scale envelope. We then extracted whole DNA from ~4 mm 2 of fin tissue using a modified method from Walsh, Metzger, and Higuchi (1991). Approximately Notes: In total, 1,022 yellow perch were captured and genotyped, but after quality control, only 975 individuals were included in analyses (see Table S2). "-" indicates that the habitat was not sampled, whereas "0" indicates that habitat was sampled but no fish were captured. for 30 s, an annealing step for 1 min (temperatures varied, see Table   S1), and 72°C for 30 s. A final extension for 10 min at 72°C finished the amplification. The exceptions to this method were a touchdown PCR on Svi-6 and an extra 5 cycles on Pfla-L6 (see Table S1). We visualized microsatellite markers on a 3130xl genetic analyzer using HiDi chemistry (Applied Biosystems).

| Data analysis
We scored markers blindly to their collection location in GeneMapper v5 (Applied Biosystems). In total, we collected DNA from 1,022 yellow perch. We performed quality control using the STRATAG package (Archer, Adams, & Schneiders, 2017) Table S2). In Lake Michigan, the sample size was 60 fish at the southern site and 47 fish at the northern site.
Two loci (Pfla-L3 and Pfla-L4) had intense stutter in their chromatograms, which likely caused unreliable genotype scoring. These loci were excluded from analyses because they were out of HWE in more than 60% of the site-habitat groupings. Yellow perch col- We assessed individual-level genetic clustering using the Bayesian clustering program STRUCTURE v2.3.2 (Pritchard, Stephens, & Donnelly, 2000). Yellow perch were clustered using the admixture model, λ = 1, and a burn-in period of 100,000 and a run time of 200,000 MCMC reps, 10 iterations at each value of K (1-17). We ran STRUCTURE both with and without priors. We used sampling location (site-habitat grouping) as a priori population indicators. We found the most likely values of K using the ΔK method from Evanno, Regnaut, and Goudet (2005) calculated in STRUCTURE HARVESTER v0.6.93 (Earl & vonHoldt, 2012), which are reported in Figure S1. We found consensus clusters across iterations of STRUCTURE by permuting and matching clusters using the large K greedy algorithm with a random input and 1,000 repeats in CLUMPP v1.1.2 (Jakobsson & Rosenberg, 2007), and we used distruct v1.1 (Rosenberg, 2004) to draw the final STRUCTURE plots.
We performed further clustering of individuals combining sites by habitat type (deep-DRM, littoral-DRM, nearshore Lake Michigan) and using discriminant analysis of principal components (DAPC) in the adegenet v2.0.1 (Jombart, 2008)  We calculated pairwise F ST (Weir & Cockerham, 1984) between all site-habitat groupings in STRATAG (Archer et al., 2017) to test whether sampling locations were genetically distinct. We applied a Holm-Bonferroni sequential correction (Holm, 1979) to pairwise F ST calculations to correct for multiple comparisons, which we recognize is conservative but did not affect our conclusions given the magnitude of the differences. Finally, we tested for isolation by distance (IBD) using the natural logarithm distances between DRMs and the linearized pairwise F ST (Slatkin, 1993)

| RE SULTS
The overall F ST was 0.024. Marker diversity was higher among DRM lakes (9.79-13.43 average alleles/locus) compared to Lake Michigan sites (9.64-9.93 avg. alleles/locus; Table S2). Private alleles ranged between zero and seven within a population, and heterozygosity within populations was fairly even (0.53-0.65; Table S2).
We found littoral-DRM and Lake Michigan yellow perch to be distinct from one another. In our STRUCTURE analysis at K = 2, Lake Michigan yellow perch and littoral-DRM yellow perch clearly exhibited different group memberships (Figure 2a), and this general pattern held when we did not use location as prior and as K increased ( Figure S2; Figure 2b). Similar to STRUCTURE, the results of DAPC showed distinct clusters between individuals sampled in littoral-DRMs and Lake Michigan habitats ( Figure 3) and no alleles weighed disproportionately on either axis of the ordination (Table   S3). Further, yellow perch from all littoral-DRM habitats were significantly different from both Lake Michigan sites in pairwise F ST comparisons ( Table 2).
Deep-DRM yellow perch were genetically distinct from littoral-DRM fish but were not distinct from Lake Michigan fish. At K = 2 and 5, most yellow perch from deep-DRM habitats and Lake Michigan sites belonged to the same genetic cluster (majority "orange" cluster; ral-DRM habitats and nearshore Lake Michigan was much higher (mean and median ≈ 0.034) than between deep-DRM habitats and nearshore Lake Michigan (mean and median ≈ 0.005), supporting the grouping of Lake Michigan and deep-DRM fish together but distinct from littoral-DRM yellow perch in the clustering analyses. However, yellow perch from the deep-water habitats in DRMs were not always significantly different from Lake Michigan sites based on F ST . The deep-water fish from Muskegon Lake were not significantly different from northern Lake Michigan, and yellow perch from the deepwater habitat of Pentwater Lake were not significantly different from either northern or southern Lake Michigan sites in pairwise F ST comparisons ( Table 2). The deep habitat of White Lake was significantly different from both Lake Michigan sites, and the deep habitat of Muskegon Lake was significantly different from southern Lake Michigan, although values of F ST were small (range = 0.007-0.008) for each comparison. Additionally, the divergence between the northern and southern Lake Michigan sites was small but significant (F ST = 0.008, Table 2).
Finally, littoral-DRM yellow perch were distinct between DRMs. Although pairwise F ST values were low (mean ≈ 0.017, median ≈ 0.016), all yellow perch from littoral-DRM habitats were significantly different from one another (Table 2). Further, there was a significant pattern of isolation by distance between DRMs (Mantel's R = 0.41, p = 0.001; Figure 4). When STRUCTURE was run for five genetic groups (K = 5), littoral-DRM lakes clustered distinctly from one another and lakes geographically close to each other shared clusters (Figures 1, 2b), supporting the results of pairwise F ST and the Mantel test. However, the clusters were not clearly defined, possibly due to the continuous isolation-by-distance pattern exhibited between DRMs.

| D ISCUSS I ON
Our results suggest that yellow perch in littoral-DRMs form distinct stocks from eastern Lake Michigan. Although divergences between littoral-DRM populations and Lake Michigan fish were small (F ST : Charlevoix = 0.008; others = 0.018-0.053), all of the differences were significant. Furthermore, both STRUCTURE and DAPC analyses consistently grouped littoral-DRM yellow perch distinctly from F I G U R E 2 Distruct plot of STRUCTURE cluster assignments for each yellow perch (n = 975). Sampling locations listed were used as a priori population indicators. Each vertical bar represents one fish, and color indicates the probability of an individual belonging to one of two (a) or five (b) genetic clusters. Dark vertical bars separate lake/habitats and lakes are ordered north to south from left to right Lake Michigan yellow perch (Figures 2, 3). Our finding that littoral-DRM and Lake Michigan yellow perch are distinct genetic stocks is supported by previous otolith and morphology studies that suggested these may be distinct stocks (Parker et al., 2009;Schoen et al., 2016).
Although littoral-DRM and Lake Michigan yellow perch are genetically distinct stocks, we found evidence that Lake Michigan yellow perch use deep-DRM habitats. Both STRUCTURE and DAPC grouped yellow perch collected in deep-DRM habitats during autumn with Lake Michigan fish (see Figures 2, 3). Our findings support the hypothesis that some Lake Michigan yellow perch migrate into DRMs in autumn and winter (prior to the spring spawning season), which was further supported by a recent study of otolith microchemistry that found evidence that yellow perch can primarily reside in Lake Michigan and make annual migrations into DRM wetlands (Schoen et al., 2016). The reasons for the annual migration of Lake Michigan yellow perch into DRMs are unclear. One explanation may be that Lake Michigan yellow perch utilize DRMs for foraging habitats because prey densities are higher than in Lake Michigan (Höök et al., 2007).
F I G U R E 3 First two axes of discriminant analysis of principal components (DAPC) of genotypes for all individuals (n = 975). Each dot represents an individual, and symbols represent sampling location. Individuals are also color-coded and grouped by habitat type. Lower right of plot shows eigenvalues of two axes relative to one another. Each single-axis distribution is plotted outside corresponding ordination axes Both managers and researchers have suggested that yellow perch from Lake Michigan may migrate into DRMs to spawn (Perrone, Schneeberger, & Jude, 1983;Schneider et al., 2007;Schoen et al., 2016). If so, then the genetic divergence of yellow perch captured in littoral-DRM habitats versus Lake Michigan suggests that Lake Michigan yellow perch either predominantly spawn in different locations or at different times than littoral-DRM fish.
Regardless of the reasons for migration of Lake Michigan fish into DRMs, yellow perch appear to begin entering deep-DRM habitats in autumn and return to Lake Michigan by the spring since we did not catch Lake Michigan fish in DRMs during our spring sampling (see Table 1). This timing of the migration may reflect when environmental conditions in deep-DRM habitats match the conditions of Lake Michigan (Altenritter et al., 2013;Biddanda et al., 2018). For instance, dissolved oxygen concentrations often become lower in the deepest parts of DRMs during summer when the water is thermally stratified ( Figure S3; see Altenritter et al., 2013;Biddanda et al., 2018). Future studies implementing telemetry (Hayden et al., 2014) could elucidate the timing and duration of DRM use by Lake Michigan yellow perch, which may facilitate accounting for the harvest of Lake Michigan fish in DRMs. Moreover, studies of otolith isotopes (Dufour, Patterson, Höök, & Rutherford, 2005;Dufour, Höök, Patterson, & Rutherford, 2008) or trace elements (Schoen et al., 2016) could provide insights on movement patterns between the two habitats.
Our results also suggest that there is a portfolio of yellow perch genetic diversity in the DRMs of eastern Lake Michigan. Previously, yellow perch stocks were found to be distinct between Green Bay and the southern basin of Lake Michigan (Miller, 2003;Glover et al., 2008). However, we found littoral yellow perch populations from all DRMs exhibited significant differences based on pairwise F ST (Table 2), suggesting that yellow perch population structure is more complex than originally reported in Lake Michigan. Furthermore, genetic divergence between DRMs exhibited a significant pattern of isolation by distance (Figure 4), which was similar to the pattern found for invasive round goby (Neogobius melanostomus) among Lake Michigan pierheads (LaRue, Ruetz, Stacey, & Thum, 2011).
Therefore, we hypothesize that DRMs are loosely distinct populations connected by gene flow. The importance of this underlying population structure to fisheries management is unclear; however, managers should think about yellow perch genetic diversity within and among DRMs when setting harvest regulations.
In conclusion, understanding population structure of species residing in connected habitats and in sympatry at certain times of the year and/or life cycle has proved critical for maintaining sustainable fisheries (Hilborn et al., 2003;Brenden et al., 2015;Wilson et al., 2016). Here, we showed that yellow perch from Lake Michigan and DRMs represent distinct stocks and that Lake Michigan yellow perch can reside in DRMs during autumn, where they may be subjected to greater harvest than in Lake Michigan. Maintaining yellow perch stock diversity may therefore depend on where they reside both spatially and temporally, and future studies should aim to support management strategies that will preserve yellow perch stocks.
Until then, we recommend that results from our study be considered when managing yellow perch and setting harvest limits in Lake Michigan, especially with respect to DRMs.

ACK N OWLED G EM ENTS
We thank the many graduate students and interns at the Robert