Genomic region associated with run timing has similar haplotypes and phenotypic effects across three lineages of Chinook salmon

Abstract Conserving life‐history variation is a stated goal of many management programs, but the most effective means by which to accomplish this are often far from clear. Early‐ and late‐migrating forms of Chinook salmon (Oncorhynchus tshawytscha) face unequal pressure from natural and anthropogenic forces that may alter the impacts of genetic variation underlying heritable migration timing. Genomic regions of chromosome 28 are known to be strongly associated with migration variation in adult Chinook salmon, but it remains unclear whether there is consistent association among diverse lineages and populations in large basins such as the Columbia River. With high‐throughput genotyping (GT‐seq) and phenotyping methods, we examined the association of genetic variation in 28 markers (spanning GREB1L to ROCK1 of chromosome 28) with individual adult migration timing characteristics gleaned from passive integrated transponder recordings of over 5000 Chinook salmon from the three major phylogeographic lineages that inhabit the Columbia River Basin. Despite the strong genetic differences among them in putatively neutral genomic regions, each of the three lineages exhibited very similar genetic variants in the chromosome 28 region that were significantly associated with adult migration timing phenotypes. This is particularly notable for the interior stream‐type lineage, which exhibits an earlier and more constrained freshwater entry than the other lineages. In both interior stream‐type and interior ocean‐type lineages of Chinook salmon, heterozygotes of the most strongly associated linkage groups had largely intermediate migration timing relative to homozygotes, and results indicate codominance or possibly marginal partial dominance of the allele associated with early migration. Our results lend support to utilization of chromosome 28 variation in tracking and predicting run timing in these lineages of Chinook salmon in the Columbia River.


| INTRODUC TI ON
It is generally understood that demographic or life-history variations within a population, such as age at maturity or reproductive investments, act to buffer the population against short-term environmental fluctuations (Hoelzel et al., 2019) and thereby provide stability in the ecosystem roles and services in which each species participates, commonly known as "portfolio effects" (Moore et al., 2014;Schindler et al., 2010Schindler et al., , 2015. Many life-history traits are known to exhibit significant heritability, suggesting not only that they are under selection and provide adaptive benefits within those populations, but also that they may be subject to influence by anthropogenic effects, such as habitat modification or fishery-induced evolution (Cartwright et al., 2014;Heino et al., 2015;Waples & Audzijonyte, 2016). Moreover, the recognition that this genetic variation may be eroded by anthropogenic effects has prompted petitions for specific conservation management of vulnerable species that is directly tied to life-history traits and associated genes (National Marine Fisheries Service, 2018). However, these conservation approaches need a clear understanding of the nature of the heritability and the strength of association to measurable markers (Heino et al., 2013;Waples & Lindley, 2018;Wray et al., 2013). This has added fuel to the search for genetic loci under selection or with strong association to known phenotypic variation in wild populations, a search which has been greatly empowered by the rapidly falling cost of sequencing and the many bioinformatic pipelines available to identify trait-associated or outlier loci (Hoban et al., 2016).

Chinook salmon (Oncorhynchus tshawytscha) is a highly exploited
Pacific salmon species that exhibits a number of heritable life-history traits that have been used to identify and manage stocks for many years, in particular variation in adult migration patterns (Myers et al., 1998). These patterns have led to the identification of different "runs" of adult Chinook salmon commonly referred to by their peak season of migration (winter, spring, summer, fall), with early runs of fish entering freshwater several months prior to spawning in contrast to late runs that typically return in the fall and spawn shortly after (Ford et al., 2020;Quinn et al., 2016). While some rivers may host multiple runs, early-returning fish tend to spawn higher in watersheds than late-returning fish, and since many of these higher reaches are only accessible with high flow, this has led to the hypothesis that heritable earlier arrival was advantageous for populations accessing habitat that became inaccessible later in the year (Quinn et al., 2016). Different runs in the same river in coastal regions tend to be closely related relative to runs of the same phenotype in other areas, indicating either low-to-moderate ongoing gene flow in coastal populations (Hecht et al., 2015;Matala et al., 2011;Narum et al., 2008;O'Malley et al., 2013;Waples et al., 2004). In contrast, different runs of Chinook salmon in the interior Columbia River remain reproductively isolated and represent distinct lineages from those found in coastal regions including tributaries near the Columbia River estuary (Hecht et al., 2015;Narum et al., 2010;Waples et al., 2004). Previous management of runs of Chinook salmon has largely been based on the idea that most complex life-history traits are polygenic and genetic variation for run timing should be broadly distributed across fish from both early and late runs (Thompson et al., 2019;Waples & Lindley, 2018;Waples et al., 2004).
Chinook salmon can be divided up into a number of hierarchical phylogeographic lineages (Beacham et al., 2006;Moran et al., 2013), and the populations in the Columbia River Basin exhibit three genetic lineages (Hecht et al., 2015;Narum et al., 2010;Waples et al., 2004). The "interior stream-type" lineage (hereafter "iST") is strongly divergent from the other two Columbia River Basin lineages (Hecht et al., 2015) known as "interior ocean-type" (hereafter "iOT"; found exclusively east of the Cascade divide) and "Lower Columbia" (hereafter "LC"; found primarily in reaches nearest the estuary and coast).
The LC and iOT lineages are more similar to one another than the iST lineage despite that the two interior lineages occur in greater sympatry.
Populations of Chinook salmon in the interior lineages are characterized by a narrower life-history variation than the LC lineage (Healey, 1991;Myers et al., 1998). While iOT Chinook salmon exhibit both summer and fall runs, they exhibit outmigration generally as subyearlings. In contrast, iST Chinook salmon outmigrate as yearlings and show a more constrained return time, in spring to early summer.
While the "stream-type" and "ocean-type" terminology apply poorly in coastal regions (Moran et al., 2013), they effectively describe interior lineages of the Columbia River and we include the "interior" designation for consistency with other literature on these Columbia Basin lineages and their life-history variation Moran et al., 2013;Narum et al., 2018). These life-history variations have allowed managers to identify different stocks within the Columbia Basin prior to genetic identification, since only a subset of stocks were present during predefined periods (Myers et al., 1998).
However, understanding genetic associations of run timing variation in Chinook salmon returning to the Columbia Basin provides another tool to apply for fisheries conservation and management (Waples & Lindley, 2018).
Several recent studies of genomic variation in Chinook salmon have identified a region of large effect for run timing on chromosome 28 that contains two genes, the human homologs for which are "GREB1 Like retinoic acid receptor coactivator" (GREB1L) and "rho associated coiled-coil containing protein kinase 1" (ROCK1; Narum et al., 2018;Prince et al., 2017). Both genes are understood to be involved in pathways affecting gonadal development, actin-myosin contraction, and renal development and are expressed in renal and reproductive tissues (Brophy et al., 2017;De Tomasi et al., 2017;Mizuno et al., 2013;Nakagawa et al., 1996;Oviedo et al., 2011;Sanna-Cherchi et al., 2017), and so could credibly be involved in reproductive maturation and environmental acclimation in anadromous salmonids. This gene region explained >70% of variance in run timing in some instances Narum et al., 2018;Prince et al., 2017;Thompson et al., 2019) and may have partially dominant phenotypic expression, in which case it is unlikely that a late run could act as a refuge or reservoir for genetic variation to bolster a beleaguered early run Thompson et al., 2019). This revelation has led some conservationists to petition for runs of Chinook salmon to be considered separate stocks in order to ensure that the full extent of life-history variation is maintained, but as yet we do not have a complete understanding of the effects of these genes on run timing variation in different stocks, or the predictive value of the different molecular markers currently available (Ford et al., 2020;Waples & Lindley, 2018).
The large variation in life-history suites of the three Columbia Basin Chinook salmon lineages make it challenging to predict the utility of chromosome 28 markers in predicting run timing. While a study using whole-genome variation identified the same region of chromosome 28 in association with run timing in each of the lineages , it is as yet unclear whether markers in this region have the same predictive value. For example, in a study of hatchery-spawned Chinook salmon and using categorial run timing phenotypes for arrival timing, Koch and Narum (2020) discovered that the strength of association varied strongly between Columbia River Basin lineages, with up to 9% of phenotypic variance explained in ST Chinook salmon but >70% in OT.
Here, we expand upon previous studies by coupling individual genetic and run timing phenotypes for Chinook salmon that were sampled and PIT tagged early in their return migration to freshwater and then tracked through the Columbia River Basin until they arrived to natal tributaries near their spawning grounds. This study addressed three specific questions regarding the association of GREB1L and ROCK1 with two adult migration phenotypes (freshwater entry and arrival timing) in Chinook salmon: (1) Is the pattern and strength of association the same among the three lineages? (2) How much variation for the two migratory phenotypes is explained by candidate genes (GREB1L/ROCK1)? and (3) Do the candidate genes demonstrate dominant or additive inheritance patterns for migratory phenotypes? Our characterization of these patterns improves the utility of these candidate genomic regions for management applications in Chinook salmon populations that exhibit variation in adult migration timing.

| Samples
All samples were collected with nonlethal methods at the Bonneville Dam Adult Fish Facility (BONAFF) between 2017 and 2019. This facility is located adjacent to a north shore fish ladder of Bonneville Dam and enables a portion of returning fish to be nonlethally sampled during their adult migration. Samples were collected during the entire run from April through October except for when the water temperature exceeding 22.2°C. Total sample sizes were estimated to be <1% of the run , and higher sampling rates were not possible given limitations of the facility. Metadata was recorded for each corresponding fish sampled including passage date at the facility and fork length, a PIT tag was inserted (unless previously tagged), scales were taken for aging, and a fin clip was taken for genetic analysis. Fish were then released into a recovery pool to volitionally continue their migration. Scales were read to determine ocean duration following standard practice for aging salmonids.
The majority of markers in the SNP panel were putatively neutral, but also included a marker predictive of sex (Brunelli et al., 2008), as well as F ST outliers from landscape analyses (Hecht et al., 2015).
Further, the panel included 28 markers in the region of chromosome 28 associated with migration timing, with five in the GREB1L gene, eleven from the intergenic region, and twelve from the ROCK1 gene . All genotype and phenotype data, and R code to process them, have been deposited in the DRYAD repository (https://datad ryad.org/stash/ share/ 4t7rm cqThS poOCn -YSRHz 7jYwv AFxpU hDnE3 ubmSt7Y).
Data were filtered to retain samples with <10% missing data and formatted using custom code in R (R Corp.) and PGDsPiDer (Lischer & Excoffier, 2012). Putative neutral loci (265) were used to identify individuals to each of the three lineages (LC, IOT, and IST) based on a principal components analysis performed using aDeGenet v2 (Jombart, 2008) in R; individuals that exhibited ambiguous lineage affinity were omitted ( Figure 1). For each lineage separately, this neutral dataset was filtered for linkage disequilibrium and effects of selection (outlier loci) using the exact test implemented in GenePoP v1.1.7 (Raymond & Rousset, 1995), with 50 batches of 50 iterations after 50 of dememorization, and outflank v0.2 (Whitlock & Lotterhos, 2015), with default options except for a RightTrimFraction of 0.1, respectively. Loci with q-values <0.1 were omitted, producing a "neutral and unlinked" marker set for each lineage. While different programs for outlier identification vary in their efficacy across demographic scenarios and genetic data types (Hoban et al., 2016), by using liberal thresholds with this relatively heuristic method, our "neutral" dataset could be considered conservative with respect to the potential effects of selection and was congruent with previous studies of neutral variation . We examined linkage disequilibrium in the candidate region of chromosomes 28 for each lineage using the program HaPloview v4.2 (Barrett et al., 2005). For simplicity hereafter, we refer to the SNP markers in the candidate region of chromosomes 28 by their order of sequence along the chromosome: 1 through 28, respectively. These numbers differ slightly than those used in Koch and Narum (2020) since five markers were dropped due to inconsistent or poor amplification (see Table S1).

| Migration timing phenotypes
For all Chinook salmon sampled at Bonneville between 2017 and 2019 (N = 8327), we obtained complete PIT histories from PTAGIS (https://www.ptagis.org). These were filtered using custom code in R, as follows. Chinook salmon migrating to the interior Columbia R. encounter Bonneville Dam at river kilometer 234, and the date of this arrival ("Passage Day") serves as a common reference point for fish passage (Figure 1), a date that may be proportional to freshwater entry. Subsequently, the most upstream array at which each fish was recorded was expected to be the closest to the tributary in which they spawned, which approximates arrival date ("Arrival Day"). Mortalities that occurred due to harvest, predation, or other unknown causes exhibited incomplete migration histories, providing abbreviated estimates of arrival date. These were recognized when an individual was last recorded at a site where spawning was not expected, and these samples (~3k) were omitted. Because fish in different lineages and runs tend to use different reaches for spawning, the arrays from which records were omitted were different. Recordings at arrays lower in the basin were omitted for iST Chinook salmon but retained for iOT, with the exception of some middle Columbia river mouths, which were omitted for summerrun iOT Chinook salmon ( Figure 2 and Figure S1). Although some prespawn mortalities were likely included in the final dataset, being relatively conservative in which arrays were allowed ensured that predicted Arrival days could be off by only a small amount. Passage Day was used to discriminate summer and fall run within the iOT lineage, based on the observed minimum in passage (July 31, ordinal F I G U R E 1 Loadings on the first two principal components from an analysis of putatively neutral loci in Columbia River Basin Chinook salmon. Solid lines demarcate where fish of indeterminate genetic affiliation were omitted. The ordinal date at which each fish were tagged crossing Bonneville Dam is imposed as color on these loadings (red, earlier; blue, later) F I G U R E 2 Locations of passive integrated transponder (PIT) arrays in the Columbia River Basin where Chinook salmon tagged at Bonneville Dam were later recorded for the last time. Active arrays for each lineage separately are included as Figure S1. Lower Columbia lineage fish were only recorded at Bonneville Dam, identified as the farthest west gray array day 212), which permitted identifying incomplete migration histories for summer-run iOT fish. Some LC lineage fish spawn in east of the Cascades in the reach above Bonneville Dam and the lower reaches of nearby tributaries (i.e., Spring Creek). The last PIT array these fish pass before spawning will often be at Bonneville Dam, making Passage and Arrival Days the same for most LC lineage fish.
For this reason, we did not utilize Arrival Day for LC Lineage fish. PIT arrays were also grouped by subbasin ( Figure S2) to calculate two relative statistic pairs: each fish's Passage and Arrival day was transformed relative to the (1) median or (2) Figure S3) and categorized each fish in these subbasins as arriving before or after this day.

| Association testing
We tested the association of all 342 autosomal SNPs with run timing in each lineage using GAPIT v. 20190926 (Lipka et al., 2012). We used the 200 (ST), 223 (OT), and 265 (LC) "neutral and unlinked" loci to infer a kinship matrix and three principal components reflecting underlying population structure in each lineage. Fork length (mm), ocean duration (years), sex (as 1 or −1), and hatchery origin (0 or 1) were included as covariates. Missing values in length and ocean duration were imputed as the median value for that lineage, while missing sex was coded as zero (missing covariates <10%). The putative genetic stock and hatchery origin of each fish were estimated elsewhere using parentage-based tagging and genetic stock assignment . We ran the "mixed linear model" (MLM) of GAPIT. To filter SNPs that are in perfect linkage disequilibrium in each lineage, we estimated LD among all SNPs using PLINK v1.9 (Purcell et al., 2007) (Balding, 2006).
For ordinal Passage and Arrival Days, we also ran the "Bayesian information and Linkage disequilibrium Iteratively Nested Keyway" (BLINK) model in GAPIT, which groups loci according to linkage disequilibrium (LD) thresholds and tests a representative SNP from each LD group, orders these by association, and then tests the significance of subsequent LD groups while including higher ranked LD groups as covariates. We modified the BLINK code to group only loci in near-perfect LD (≥0.99), implying that differences in significance between the MLM and BLINK models reflect the redundancy in explanatory power produced by the remaining linkage among loci.
The BLINK model does not report the variance explained by each LD group (R 2 ), so for the first 12 LD groups reported by BLINK, we used sequential MLM models in which cluster-representative SNPs were included as covariates to estimate R 2 . Missing covariate genotypes in these sequential MLM models were input as heterozygotes.
Since SNPs in LD were not independent, further examination was made through haplotype association in the candidate regions.
However, as results from haplotype-based analyses were strongly consistent with the SNP-based methods, we describe these analyses and results in the supplementary information archived with DRYAD.  Table S2). We omitted 20 individuals that exhibited ambiguous lineage affinity; these fish were distributed across all three management periods, but were mostly comprised of LC lineage fish (N = 16), many of which were clipped hatchery-origin fish from the Willamette River spring-run stock (Table S2). We have described straying above Bonneville Dam from this LC stock previously (Hess et al., 2014). As expected given the stock composition, the iST Chinook salmon all arrived early in the year, but overlapping with "early" or summer-run iOT fish (Figure 3).  Figure S9 and Files S1-S3). In iOT and iST fish, markers in the right-central linkage block (ROCK1-side) showed the strongest association with both ordinal Passage Day and Arrival Day, followed by markers in the left-central block (GREB1L-side; Figure 5 and Figure   S10). For LC lineage fish, there was no distinction among SNPs in the single central LD block (markers 6-19), which showed the strongest association ( Figure 5). However, the maximum variance in Passage Day explained by these SNPs was different in each lineage, with iOT Chinook salmon showing the strongest association (max R 2 = 0.476), followed by iST Chinook salmon (R 2 = 0.353) and LC (R 2 = 0.279), though we note the small sample size of spring-run LC Chinook salmon compromises the precision of the variance estimate for that lineage (Files S1-S3). The maximum variance explained in Arrival

| RE SULTS
Day was also higher in iOT (R 2 = 0.396) than iST fish (R 2 = 0.098), and, notably, each was less than the respective values for Passage Day, though this discrepancy was larger for iST fish. The pattern and amount of variance explained in the subbasin relative variables showed similar patterns and in each case were less than the raw ordinal variables (Files S1,S2). Results from the BLINK analyses of ordinal Passage and Arrival Days reflected these results, with the group containing the right-central LD block (ROCK1-side) identified as the most significant, while all other LD groups explained a much smaller amount of variance in run timing after accounting for the earlier LD groups ( Figure S11). In most cases, the second most significant LD group was also in the chromosome 28 candidate region. Association testing with haplotypes estimated from genotypes of the chromosome 28 candidate markers also substantiated these observations (results archived on DRYAD).
The evidence for dominance effects in chromosome 28 variation differed by lineage, although for the SNPs in the most strongly associated linkage groups identified above, heterozygotes were generally intermediate in Passage Day to homozygotes, both overall ( Figure 6) and within individual subbasins with sufficient F I G U R E 3 Probability Density Function for the date at which each Columbia River Basin Chinook salmon was tagged crossing Bonneville Dam between April and October, for each lineage (blue, interior stream-type; orange, interior ocean-type; gray, Lower Columbia). Probability Density Function is akin to the proportion of fish in a lineage that passed on each day. For each color, the area under the curve sums to 1, but curves are not proportional to each other in counts; the sample size representative of each curve is indicated. Vertical dotted and dashed lines represent divisions between the spring, summer, and fall management periods, ordinal, and calendar dates for which are noted, along with the first and last day a Chinook salmon in this dataset migrated sampling of each genotype ( Figure S12). For iST Chinook salmon, coding heterozygotes so that they were either of the two homozygotes did not change the patterns of run timing association within the chromosome 28 region, and neither of these dominance codings explained as much variation as codominance, although the strength of association was stronger when coding heterozygotes as early-migrating homozygotes than later migrating (File S1

| DISCUSS ION
Recent studies have shown that the GREB1L/ROCK1 region on chromosome 28 has a strong effect on run timing in Chinook salmon and steelhead trout (Oncorhynchus mykiss), but many questions remain about the consistency and strength of this genetic variation among the various populations of each species and thus the utility of identified molecular markers from this region (Waples & Lindley, 2018).
Our results confirm that genetic variations in this region are associated with effects in the same direction in each of the three lineages of Chinook salmon present in the Columbia River Basin, albeit in a lineage-specific strength and pattern, and thus that genetic markers in this region will be useful in predicting run timing distributions of stocks in the Columbia Basin. Moreover, heterozygotes at the most strongly associated markers were largely intermediate in their migration timing relative to homozygotes, suggesting that strong selection for/against a particular run will largely preclude the retention of variation for alternative migration phenotypes.
We observed that markers in the chromosome 28 region explained a large portion of phenotypic variation in the date around which Chinook salmon began their migration (Passage Day), and to a lesser extent, the approximate date on which they arrived to their spawning grounds (Arrival Day). Consistent with the linkage among these markers and the strong haplotype blocks in the central portion of this genomic region, the same haplotypes (combination of SNP alleles) were strongly associated with run timing variation in each of the three lineages. This provides strong evidence that the same genetic variation has similar effects in each of the three lineages, albeit in a lineage-specific manner. For example, although Lower Columbia (LC) and interior ocean-type (iOT) Chinook salmon appear to exhibit similarly large differences in run timing phenotypes, both presenting a larger range between early and late migrators than interior streamtype (iST), the strength of association between chromosome 28 alleles was considerably higher for iOT Chinook salmon than for LC (max R 2 = 0.476 vs. 0.279, respectively). While this difference could conceivably result from strong asymmetry in representation of the two runs in our LC samples (N = 6 vs. N = 216), a previous study ) observed a similar effect size (max R 2 = 0.280) but with symmetrically distributed collections of Chinook salmon from the Cowlitz tributary of the lower Columbia (N = 94 each for early (spring) vs. late (fall) collections). In contrast, this same previous study also observed considerably higher association values for iOT Chinook salmon collected at Prosser and Wells Dams (max R 2 = 0.71), suggesting that the estimates made herein may yet be conservative. Two differences between these two studies should be noted. First, while the current study considered iOT Chinook salmon from a number of different interior spawning groups (tributaries and reaches), those authors used early-and late-migrating collections from within a single representative population of each lineage , thus avoiding some potential confounding variation that we may not have been able to completely accommodate through covariates here. Second, while we considered individual run timing data (i.e., continuous response variable) gleaned from PIT tag recordings and passage date, this previous study used a categorical (binary) response variable based on the designation of each population as early or late arrival for spawning. However, we did not see strong differences in results after transforming our migration data into a categorical response, except for changing rank among markers in the left-central (GREB1L-side) and right-central (ROCK1-side) LD blocks, suggesting this coding is not responsible for the difference in the magnitude of association between the two studies.
The relative strength of the association of run timing of Columbia Basin Chinook salmon with markers in the intergenic region between the GREB1L and ROCK1 genes is notable for its consistency both here and in Koch and Narum (2020), as well as similar studies of spring vs. fall Chinook salmon from coastal populations (Thompson et al., 2019;Thompson et al., 2020). While the strongest associations appeared to be with markers linked to the most upstream portion of the ROCK1 gene both here and in Koch and Narum (2020), the markers in the linkage group including the most upstream portion of GREB1L showed values of nearly equal magnitude, while markers adjacent but further downstream in the ROCK1 gene showed considerably weaker associations. Interestingly, Columbia Basin steelhead, which show a similar association of chromosome 28 variation as Chinook salmon, in some cases explaining as much as 50% of variation in run timing, exhibited the strongest associations with the upstream portion of GREB1L and adjacent intergenic region . This slight difference in genomic localization of the strongest signal of association across these two species is unexpected because parsimony would suggest similar effects of chromosome 28 genetic variation on these two species (Ford et al., 2020).
However, it is worth noting that, despite the fact that these fishes have so far been the only salmonid species in which chromosome 28 variation has been implicated in run timing, there are reasons to expect that patterns of association would differ between them, including that they are not each other's closest relatives. Moreover, as these two genes are situated with opposing transcriptional frames, they may share a regulatory region, and so their effects on run timing may not be independent . While both genes are expressed in reproductive and renal tissues and appear to be integral to development in humans (Brophy et al., 2017;De Tomasi et al., 2017;Mizuno et al., 2013;Nakagawa et al., 1996;Oviedo et al., 2011;Sanna-Cherchi et al., 2017), their functions in fishes are still uncertain. Thus, while markers in this chromosome 28 region appear to have strong predictive value for run timing in Chinook salmon, studies illuminating the functional structure of the intergenic sequence in this region and the effects of observed genetic variation would be very useful.
Although interior stream-type (iST) Chinook salmon showed a similar pattern of association as the other lineages with strongest associations of markers 15-17, in the intergenic region, and similar association signals for similar haplotypes, which is notable considering how divergent this lineage is from the other two, there were two additional noteworthy observations specific to iST Chinook salmon.
The first is that the magnitude of association between individual run timing and chromosome 28 variation was similar between the three lineages despite earlier median run date and more constrained overall variation in run timing in iST Chinook salmon compared to iOT or LC, and all iST Chinook salmon are reproductively immature at the time of freshwater entry (Myers et al., 1998 This observation suggests that more diverse or extreme barriers or environmental variation along the migration paths of iST Chinook salmon act to diminish the correlation between freshwater entry and arrival to spawning grounds relative to iOT fish (Hecht et al., 2015).
While spring and summer-run Chinook salmon are at an advantage over later returning fish by utilizing habitats that might be inaccessible later in the year, there appear to be tradeoffs associated with an early run timing strategy (Quinn et al., 2016). In particular, later-migrating fish avoid the highs in river temperature during the summer that make migration physiologically more taxing (Quinn et al., 2016), and given the more extreme temperatures and variation in flows associated with climate change, and it is not surprising that early-migrating (spring and summer run) Chinook salmon have been beleaguered disproportionately to later-migrating (fall or winter run) fish, even so far as being extirpated completely in some cases despite the persistence of strong fall runs (Ford et al., 2020;Thompson et al., 2019). An important uncertainty is if the conditions for early-migrating fish were to improve where such a run once existed, could early-migrating runs re-emerge from nearby later-migrating fish of the same lineage (Ford et al., 2020;Thompson et al., 2019;Waples & Lindley, 2018)?
Since we now recognize the importance of chromosome 28 variation in shaping migration phenotypes, the answer to this often depends on the migration phenotype of heterozygotes carrying the early migration (premature) allele, and the degree to which these persist in the fall run (Ford et al., 2020;Thompson et al., 2019;Waples & Lindley, 2018). Examining phenotype plots of the most significantly associated marker with run timing, heterozygotes in iOT and iST fish appear to migrate at an intermediate time to either homozygote, suggesting the premature allele is not recessive and will not persist cryptically within later-migrating runs. We note that slightly more heterozygote iOT fish migrated with the peak of homozygous early/premature than with the peak of homozygous late/mature fish, but this may be a product of low heterozygote fitness during the period of highest temperatures rather than direct genetic effects . While our examination of patterns of inheritance of the chromosome 28 alleles did not reveal clear evidence for dominance, there were indications of partial dominance of the early migrating (premature) rather than late-migrating (mature) alleles, further eroding the hope that genetic variation for premature migration could be retained within later-migrating runs.
Finally, these genetic markers have the potential to be useful for tracking and predicting migration phenotypes for closely man- Periods based on the chromosome 28 genotypes of broodstock fish that were crossed in a hatchery setting. This would provide further information to assist with fisheries management in the region that is heavily based on run timing.

| CON CLUS IONS
Despite strong genomic divergence among lineages of Chinook salmon, genetic variation in the region of chromosome 28 containing the GREB1L and ROCK1 genes was found to have moderate to strong association and similar effects on adult migration timing in each of the three lineages that inhabit the Columbia River Basin. These results point to the potential for these genetic markers to be useful for tracking and predicting migration phenotypes for managed stocks of Chinook salmon, though we recommend that their adoption be made with robust adaptive implementation. Our results align with previous inferences that chromosome 28 variation is codominant, or perhaps marginal partial dominance of the premature allele, suggesting that fall runs of Chinook salmon will be poor long-term reservoirs and ineffective sources of re-emergence for premature migration alleles and extirpated spring and summer runs.

ACK N OWLED G EM ENTS
The authors thank colleagues at the Columbia River Inter-Tribal Administration (Grant no. 2008-907-00).

CO N FLI C T O F I NTE R E S T
The authors declare they have no conflict of interest regarding the data or inferences discussed in this manuscript.