Restricted gene flow between resident Oncorhynchus mykiss and an admixed population of anadromous steelhead

Abstract The species Oncorhynchus mykiss is characterized by a complex life history that presents a significant challenge for population monitoring and conservation management. Many factors contribute to genetic variation in O. mykiss populations, including sympatry among migratory phenotypes, habitat heterogeneity, hatchery introgression, and immigration (stray) rates. The relative influences of these and other factors are contingent on characteristics of the local environment. The Rock Creek subbasin in the middle Columbia River has no history of hatchery supplementation and no dams or artificial barriers. Limited intervention and minimal management have led to a dearth of information regarding the genetic distinctiveness of the extant O. mykiss population in Rock Creek and its tributaries. We used 192 SNP markers and collections sampled over a 5‐year period to evaluate the temporal and spatial genetic structures of O. mykiss between upper and lower watersheds of the Rock Creek subbasin. We investigated potential limits to gene flow within the lower watershed where the stream is fragmented by seasonally dry stretches of streambed, and between upper and lower watershed regions. We found minor genetic differentiation within the lower watershed occupied by anadromous steelhead (FST = 0.004), and evidence that immigrant influences were prevalent and ubiquitous. Populations in the upper watershed above partial natural barriers were highly distinct (FST = 0.093) and minimally impacted by apparent introgression. Genetic structure between watersheds paralleled differences in local demographics (e.g., variation in size), migratory restrictions, and habitat discontinuity. The evidence of restricted gene flow between putative remnant resident populations in the upper watershed and the admixed anadromous population in the lower watershed has implications for local steelhead productivity and regional conservation.

watershed regions. We found minor genetic differentiation within the lower watershed occupied by anadromous steelhead (F ST = 0.004), and evidence that immigrant influences were prevalent and ubiquitous. Populations in the upper watershed above partial natural barriers were highly distinct (F ST = 0.093) and minimally impacted by apparent introgression. Genetic structure between watersheds paralleled differences in local demographics (e.g., variation in size), migratory restrictions, and habitat discontinuity. The evidence of restricted gene flow between putative remnant resident populations in the upper watershed and the admixed anadromous population in the lower watershed has implications for local steelhead productivity and regional conservation.

K E Y W O R D S
admixture, Columbia River, habitat heterogeneity, immigration, resident redband trout, steelhead trout, sympatry

| INTRODUCTION
The highly diverse life history of Oncorhynchus mykiss has been widely documented among populations throughout the species range along the North Pacific Rim (Beacham, Pollard, & Le, 1999;Busby et al., 1996;McPhee et al., 2007). Genetic differentiation varies broadly, with distinctions dictated by myriad factors including complex mating behaviors, habitat distribution, and the prevalence of plastic traits such as migration time (Crozier, Scheuerell, & Zabel, 2011;Hendry, Wenburg, Myers, & Hendry, 2002). Population distinctions are often more discernable at the regional level, in association with greater distances between spawning aggregates Garza et al., 2014;Nielsen, Byrne, Graziano, & Kozfkay, 2009), but diverse life histories can lead to fine-scale population distinctions (Benjamin, Connolly, Romine, & Perry, 2013;Thorpe, 2007), even within a discrete stream network (Kozfkay, Meyer, Schill, & Campbell, 2011;Nielsen, Pavey, Wiacek, & Williams, 2005). The migratory behavior of O. mykiss is one of the most influential life history attributes in terms of contributions to population genetic structure. Resident O. mykiss remain in fresh water from rearing to maturity, while conspecific anadromous steelhead trout migrate to the ocean where they typically grow for one to 3 years before returning to spawn in natal streams (Behnke, 2002).
The Rock Creek Subbasin of Washington State, USA, is part of the middle Columbia River distinct population segment (DPS) where steelhead trout are currently listed as "threatened" under the Endangered Species Act (Busby et al., 1996;NMFS 2009 appendix C). Rock Creek and its tributaries are devoid of dams and other anthropogenic structural barriers, and opportunities to monitor the local O. mykiss population (e.g., placement of adult weir or juvenile traps) have been limited by the generally inaccessible landscape dominated by rangeland. Based on its geographic isolation, the Rock Creek subbasin is believed to support a distinct native population of steelhead trout (ICTRT 2003), but habitat conditions have had adverse effects on the abundance and productivity of the local fish population/s (WDF and WDW, 1993;Harvey 2014). Introgressive influences from nonlocal hatchery (or wild) fish remain generally unknown, and assumptions about the distinctiveness of O. mykiss in the Rock Creek subbasin have not been sufficiently verified through genetic monitoring or abundance surveys.
In our study, we have conducted a genetic evaluation to characterize the Rock Creek steelhead trout population. We focused on testing three primary hypotheses: 1) the Rock Creek population is a single distinct population within the Columbia River Basin, 2) seasonally fragmented habitat does not influence genetic variation, and 3) potential natural barriers restrict gene flow between upper and lower sections of the watershed influencing sympatry between putative resident and anadromous life histories, respectively. Our objectives serve to inform regional conservation of the species. Understanding the degree of genetic distinctiveness in the Rock Creek population, where hatchery impacts are presumed small, will provide a more comprehensive understanding of immigrant or exogenous influences throughout the Columbia River Basin.

| Study area and steelhead habitat
Rock Creek is a tributary of the Columbia River located approximately 363 river kilometers (rkm) upstream of the Pacific Ocean. The small subbasin (~558 square kilometers) is dominated by rangeland, where developed land accounts for <1% of total area. The two main water courses within the subbasin are Rock Creek and Squaw Creek ( Figure 1). There is no regulated flow and only minor water diversions to accommodate livestock watering and irrigation (Aspect Consulting and Watershed Professionals Network 2004;Harvey, 2014). The steep bedrock terrain contributes to short but intense flow responses following precipitation. Throughout the summer dry season, low flows and high water temperatures create long stretches of dewatered streambed, leaving a patchwork of isolated perennial pools. Unlike many watersheds in the Middle Columbia River region, Rock Creek has incurred no direct stocking of hatchery fish.
The migratory limit of anadromous steelhead occurs approximately 0.5 rkm above the Quartz Creek confluence with Rock Creek (rkm 28.6), and approximately 1.3rkm upstream of the Squaw Creek/ Harrison Creek confluence (rkm 20.3; Figure 1). The historical range of anadromy in Rock Creek is largely unknown. During spawning surveys between 2009 and 2014, we observed the majority of steelhead redds in the mainstem of Rock Creek below rkm 22, and in Squaw Creek 9 rkm upstream of the Rock Creek confluence (Harvey, 2014). Habitat surveys were conducted in September 2011 and September to October 2012 throughout most of the anadromous fish-bearing reaches of Rock Creek to determine the spatial distribution and seasonal use of habitat by juvenile O. mykiss prior to smolt outmigration in spring (Allen, Munz, & Harvey, 2014;Harvey, 2014).
Sections of stream were partitioned into three stream categories: 1) nonpool dry, defined as dewatered areas, 2) nonpool wet-defined as wetted gravel but no pools, and 3) perennial pool habitat. The latter represents rearing and refuge habitat during the dry season.
Latitude and longitude coordinates were recorded at transitions (way-points) between habitat categories, and habitat was mapped to the stream network using ArcMap version 10.0 (copyright © 1999-2010 ESRI Inc.).  (Table 1). Fish <70 mm in spring and <90 mm in the fall were deemed age-0, larger fish up to 149 mm were grouped as age-1+ (see Gallagher, 2004 for comparison), and those at least 150 mm FL were presumed to be mature age2+ O. mykiss. Fish ≥70 mm were PIT-tagged following accepted procedures (Columbia Basin Fish and Wildlife Authority 1999), and juvenile O. mykiss and adult steelhead movement was monitored between 2009 and 2012 using PTIS transceivers installed at rkm 5 and rkm 13 of Rock Creek ( Figure 1). Large numbers of juveniles were handled and interrogated for PIT tags (n = 6,218). Sampling for genetic analyses was limited by encounter rate and permitting (as per NOAA fisheries). We targeted a random sample of five age-0 and five age-1+ juveniles from each pool, and all age-2+ fish, which were less abundant. Samples were grouped by spatial distribution into stream sections loosely bounded by extended stretches of "nonpool dry" habitat (i.e., dewatered river bed) or gaps in sampling coverage ( Figure 1). The resulting six lower watershed groups were identified by upstream distance from the Columbia River confluence: rkm1-9, rkm15-19, and rkm20-22 in Rock Creek and rkmS13-15, rkmS15-21, and rkmS21-22 in Squaw Creek ( Figure 2).
Sample sizes among sections ranged from n = 47 to n = 154 (Table 1).
Two locations were sampled in the upper watershed of Rock Creek, beyond the presumed limit of anadromy where resident O. mykiss may reside (Courter, Justice, & Cramer, 2009). One collection included a combination of fish sampled directly above and below Ekone falls located at ~rkm 33 of Rock Creek (n = 56; Figure 1). A second collection was sampled approximately 12 rkm upstream of a partial barrier in Quartz Creek (n = 21).

| Laboratory protocol and genetic analyses
Genomic DNA was extracted using a standard Qiagen ® DNeasy™ protocol. Samples were genotyped at 192 SNP loci using TaqMan chemistry (Life Technologies) and Fluidigm 96.96 dynamic array T A B L E 1 Summary statistics for the Rock Creek Oncorhynchus mykiss genetic sample. Samples are grouped by sample year and river section within watershed. Age proportions were inferred from fork lengths: age-1+ ≥70 mm in spring and ≥90 mm in the fall. Smaller fish were designated age-0. Fish ≥150 mm FL were deemed age-2+ or "mature." Mean genetic diversity is observed (H O ) and expected (H E ) heterozygosity. Differences in age structure within and between watersheds are shaded gray  Polymerase chain reaction (PCR) amplification followed the protocol of Hess, Campbell, Matala, and Narum (2012). Quality control measures allowed for no more than 10% missing data per individual.
The panel included three species diagnostic markers used for hybrid screening between O. clarkii (cutthroat trout) and O. mykiss, and a sex-determining SNP locus (OmyY1_2SEXY) provided identification of genetic sex.
Allele frequency and heterozygosity were estimated using GenAlEx version 6.5 (Peakall & Smouse, 2006). Deviations from Hardy-Weinberg equilibrium (HWE) expectations, and tests of linkage disequilibrium were evaluated using GENEPOP v. 3.3 (Raymond & Rousset, 1995), applying an adjusted significance level for multiple tests (Benjamini & Yekutieli, 2001;Narum, 2006). The program ARLEQUIN version 3.5 (Excoffier, Laval, & Schneider, 2005) was used to calculate pairwise population F ST among sample years and among river sections to evaluate temporal and spatial differentiation, respectively. A pairwise matrix of Nei's standard genetic distance (Nei, 1972) and an unrooted neighbor-joining (NJ) tree were generated using PHYLIP version 3.68 (Felsenstein, 2008). Relatedness was estimated using the program ML-Relate (Kalinowski, Wagner, & Taper, 2006) to test for potential kinship bias in the form of full-sibling groups within each stream section.

| Habitat and demographic descriptions
The distribution of dewatered stream sections and dispersed perennial pools was consistent over two survey seasons, suggesting interannual variation in stream hydration patterns may be minimal from early summer to early fall (Figure 2). The average proportion of age1+ O. mykiss among lower watershed groups ranged from 6% (rmk1-9) to 86% (rkm20-22) in the spring and from 18% (rkm15-19) to 66% (rkm20-22) in the fall (Table S2). The average age-1+ proportions in the upper watershed groups were 59% and 53% in the spring and fall, respectively. The proportion of fish among lower watershed groups that were considered reproductively mature ranged from a low of 2% to 6% in the spring, and from 6% to 13% in the fall. In the upper watershed, the proportion of mature fish in the sample ranged from 31% in the spring (Ekone Falls) to 48% in the fall (Quartz Creek).
Relative proportions of age-0 and age-1+ O. mykiss in the genetic subsample (Table 1) (Table 1). Full-sibling proportions ranged from 8% at Ekone Falls to 4%-5% among all remaining groups. Results were variable among sites but nominally different, with no apparent upstream kinship bias.

Ekone and Quartz groups assigned to the lower watershed of Rock
Creek with highest likelihood. (Table 3). Assigned origins among six lower watershed groups were dominated by self-assignments, ranging from 54% to 61%, however; corresponding likelihood scores were low ranging from LS = 69.6 to LS = 81.6. The largest proportion of out-ofbasin assignments among all Rock Creek lower watershed groups were to the MGILCS and UPSALM reporting groups. Only four individuals sampled in the lower watershed (0.7%) were assigned to either the Ekone Falls or Quartz Creek groups in the upper watershed, including three in the farthest upstream group (rkm20-22) and one in the farthest downstream (rkm1-9; Table 3).

| Gene flow and immigration in the lower watershed
The genetic structure among O. mykiss in the lower watershed of Rock Creek was spatially consistent, with limited differentiation among groups. In our stock identification analyses, we saw variable results with regard to out-of-basin influences in Rock Creek.
Some steelhead detected at PIT-tag transceivers in Rock Creek had been previously tagged as juveniles in Rock Creek, suggesting that steelhead continue to spawn naturally in Rock Creek. Further studies would be necessary to gauge the abundance or effective population size of a putative native spawning population, but it is likely to be small given the availability of spawning habitat. However, the moderate likelihood scores for self-assignments as well as significant numbers of out-of-basin assignments point to a clear lack of genetic distinction of the Rock Creek population within the Columbia River region. The insufficient power to confidently (i.e., LS > 0.9) assign origins of fish sampled from Rock Creek is consistent with results for adjacent Middle Columbia River populations. Minimal GSI resolution has been influenced by significant regional straying or admixture (see , likely contributing to the broad geographic expanse of the MGILCS reporting group (Hess, Campbell, et al., 2016;Hess, Ackerman, et al., 2016). In fact, most PIT-tag detections in Rock Creek were from individuals that had known out-of-basin juvenile release locations, and the probable reproductive success of immigrants is manifest as significant admixture in the lower watershed (Araujo, Candy, Beacham, White, & Wallace, 2014;Hess & Matala, 2014)  Over the past 25-50 years, hatchery stocking has been used extensively in the Columbia River Basin for fishery enhancement and harvest opportunities, or for supplementation to aid steelhead conservation and recovery goals (Paquet et al., 2011). However, research has shown that hatchery straying from both types of programs can influence local population structure in salmonids (e.g., Hess & Matala, 2014;Van Doornik, Berejikian, & Campbell, 2013;Van Doornik, Eddy, et al., 2013). Significant straying of nonlocal hatchery stocks has been observed in large subbasins adjacent to Rock Creek including the Deschutes River (Hand & Olson, 2003; with its own local hatchery program, and the John Day River (Ruzycki & Carmichael, 2010; Figure 1) with no direct hatchery stocking. Hatchery steelhead are believed to exhibit a higher stray rate compared to natural-origin steelhead (Keefer & Caudill, 2014;Quinn, 1993), but immigrants in Rock Creek identified via PIT-tag detection revealed that most (26 of 37) were natural-origin steelhead (Table S3). One caveat to consider is that not all hatcheries physically mark (e.g., adipose fin clip) fish, which is used to distinguish them from wild fish. The absence of population trend data or archival samples in our study precluded us from accurately assessing historical steelhead immigration rates in Rock Creek or the relative impacts associates with supplementation activity. Nevertheless, the accumulated effects of introgression over an unknown number of generations have likely diminished our ability to confidently differentiate local from exogenous sources of anadromous productivity in the lower watershed of Rock Creek.

| Migratory life history differences among watersheds
Overall Creek. Typically, the resident O. mykiss phenotype contributes to the overall productivity of sympatric anadromous steelhead. Resident trout may produce progeny that undergo smoltification and migrate to sea, while offspring of anadromous steelhead can adopt a nonmigrator, resident life history (Courter et al., 2013). However, gene flow between resident and anadromous population components is contingent to a large degree on their respective distributions (Narum et al., 2008;Sogard et al., 2012). Partial barriers, variable flows and other physical elements of stream habitat can restrict movement and influence opportunities for interactions. Based on characteristics of both Ekone Falls and the falls in Quartz Creek (e.g., height, flow, plunge pools), it is thought that neither is a complete barrier to upstream migrating steelhead, but we could not conclusively confirm this. Even with obstructed upstream movement, individuals in upstream reaches can become entrained (fall) or swim downstream (Hayes et al., 2012;Pearse et al., 2009;Van Doornik, Berejikian, et al., 2013;Van Doornik, Eddy, et al., 2013). In fact, approximately half of our Ekone Falls samples were taken below the falls. It would therefore be reasonable to assume that unidirectional gene flow in a downstream direction could occur. However, limited gene flow was evident between watersheds in our analyses.
Growth trajectory, generation time, and body size can affect the propensity to migrate (Benjamin et al., 2013;Letcher et al., 2007;Sogard et al., 2012), and upper watershed groups were dominated by larger, putatively mature fish. From an evolutionary perspective, it is likely that selection operates against migratory phenotypes in above barrier populations (Pearse et al., 2009)  O. mykiss juveniles (Ohms, Sloat, Reeves, Jordan, & Dunham, 2014). Rundio et al. (2012) found that sex ratios of outmigrants were skewed toward females in a sympatric resident and anadromous population.