Introduction

Conservation management using hatchery supplementation continues to be effective at maintaining abundances of threatened or endangered fish species, such as in populations of Pacific salmon (Paquet et al. 2011; Fast et al. 2015). However, the underlying life history diversity in salmonids can complicate attainment of management goals. Beyond maintaining abundance, successful supplementation typically involves minimizing life history divergence between hatchery-reared fish and natural populations (Mobrand et al. 2005; Fraser 2008; Waters et al. 2015). A considerable degree of life history diversity occurs in Pacific salmon. For instance, variation occurs in adult migration timing (Quinn et al. 2016), distribution in marine environments (Healey 1983), juvenile emergence and smoltification/migration timing (Billman et al. 2014), and age of sexual maturity (Healey 1991). Thus, one goal of hatchery supplementation for conservation purposes is to produce fish with maturation trajectories reflective of the natural population.

The age an individual reaches sexual maturity (age-at-maturity) is believed to be a threshold trait influenced by genetics (Hankin et al. 1993; Heath et al. 1994) and responses to factors such as rearing environment, growth opportunities, and physiological assessments during specific “decision windows” that are recognized by photoperiodic changes (Thorpe 1989). For semelparous Chinook salmon (Oncorhynchus tshawytscha), age-at-maturity can vary greatly amongst populations, sexes, and individuals. Population level variation results from the differences and shifts of reaction norms that are the result of genetic and environmental interactions (Hutchings 2011). Within the interior Columbia River, there are two distinct lineages of Chinook salmon referred to as interior ocean-type and interior stream-type that have diverse patterns in freshwater residency of juvenile fish (Myers et al. 1998). Juveniles from the interior ocean-type lineage will migrate to the ocean as subyearlings within a few months after emergence, whereas interior stream-type juveniles will spend a full year in freshwater before migrating to the ocean as yearlings (Myers et al. 1998). Male Chinook salmon have a higher degree of variability in age-at-maturity than females. Typical anadromous adults mature at age 4 or 5, however, a notable proportion of males, referred to as jacks, mature at age 3 (Myers et al. 1998). Precocious maturation in males can occur in freshwater at age 1 (microjacks) and at age 2 (minijacks) within populations (Larsen et al. 2004). Smaller precociously maturing males attain reproductive success by utilizing a sneaking tactic to gain access to females amongst older and larger males (Ford et al. 2012; Schroder et al. 2012). In contrast, female fish rarely mature precociously, but instead maximize reproductive success by maturing at older ages and larger sizes leading to increased fecundity (Healey and Heard 1984; Larsen et al. 2013; Malick et al. 2023). The precocial male phenotypes represent a “bet hedging” tactic maintained within the species to increase population resilience against potential disastrous environmental conditions impacting other age classes of the same cohort or other life histories (Bourret et al. 2016). For individuals within a population, there is a tradeoff between maturing at a younger age and experiencing reduced reproductive success versus the increased risk of mortality associated with maturing at older ages.

In natural populations, male precocial maturation is maintained at a low level by frequency dependent selection, whereby the reproductive success of the precocious male phenotype is inversely related to the frequency of the phenotype (Pearsons et al. 2009; Berejikian et al. 2010). Observations of natural origin fish show that both age-1 and age-2 precocial males represent < 5% of populations (Gebhards 1960; Mullan et al. 1992; Larsen et al. 2004; Pearsons et al. 2009). However, hatcheries often observe elevated incidences of precocial maturation, particularly as minijacks. In Columbia River Basin Spring Chinook salmon hatcheries, minijacks can represent as much as 71% of the males produced, depending on hatchery and brood year (Harstad et al. 2014). The high proportion of minijacks is likely due to high growth rate and whole-body lipid content during the juvenile stage in the hatchery environment (Silverstein et al. 1998; Shearer and Swanson 2000; Larsen et al. 2006; Shearer et al. 2006). Elevated minijack production impedes attainment of hatchery goals by resulting in an unnatural distribution of ages-at-maturity and a reduction in the proportion of anadromous males present in fisheries. Some evidence shows precocial males experience high mortality before spawning, stray into other systems, or fail to return to spawning grounds (Pearsons et al. 2009, 2023). Additionally, there’s been an observed decrease in overall size and age-at-maturity in northeast Pacific Ocean salmon, possibly related to size selective fisheries and changing environmental conditions, resulting in a decrease of phenotypic diversity and population resiliency (Greene et al. 2010; Ohlberger et al. 2018). Identifying the physiological and genetic factors that regulate male Chinook salmon precocious maturation in males is a pathway towards effective management of the maturation phenotypes of fish produced by conservation hatcheries. In addition, efforts to better understand these mechanisms will inform on the more broadly observed shifts in population age structures.

Several studies have investigated the genetic basis for age-at-maturity in salmonids completing an anadromous lifecycle (i.e. multiple years in the ocean) including Chinook salmon. Evidence shows high heritability for this trait in multiple salmonid species indicating selection could occur for a given phenotype (Hankin et al. 1993; Heath et al. 2002; Carlson and Seamons 2008; Gamble and Calsbeek 2023). Most of the focus around explaining the genetic control of age-at-maturity has been in domesticated strains and wild populations of Atlantic salmon (Salmo salar) in Europe (Ayllon et al. 2015; Barson et al. 2015; Sinclair-Waters et al. 2020) and North America (Gutierrez et al. 2015; Kusche et al. 2017). In Atlantic salmon using a high-density single nucleotide polymorphism (SNP) array (Barson et al. 2015) and whole genome re-sequencing (Ayllon et al. 2015), a single large effect locus displaying sex dependent dominance was found. This locus, the vestigial-like family member 3 gene (VGLL3) on chromosome 25, explained 33–36% (Allon et al. 2015) and 39% (Barson et al. 2015) of the phenotypic variation in age-at-maturity. Additionally, Sinclair-Waters et al. (2020) identified a signal at the VGLL3 locus and an additional strong signal at the Six homeobox 6 (Six6) locus that provided evidence of a polygenic architecture for age-at-maturity in Atlantic salmon. Studies involving Pacific salmonids have found similar genetic association with age-at-maturity. Consistent with the Six6 locus in Atlantic salmon, Willis et al. (2020, 2024) found that variation in markers on chromosome 25 in and around the Six6 region was associated with ocean-age in steelhead trout (Oncorhynchus mykiss) and subsequently age-at-maturity. Furthermore, Waters et al. (2021) detected intra- and interspecific patterns in association of the Six6 genome region after targeted sequencing and whole-genome re-sequencing in steelhead trout and sockeye salmon (Oncorhynchus nerka). The variable pattern of association in these species indicates both a shared and possibly divergent evolutionary pathway towards regulation of maturation ages in Pacific salmonids. Using RADseq data, McKinney et al. (2020, 2021) performed a genome-wide association study (GWAS) and discovered genetic variation for age-at-maturity in the form of male specific haplotypes in Chinook salmon located near sex determining regions of the genome. The haplotypes associated with age-at-maturity (including minijacks) varied in frequency across geographical landscapes and maturation ages suggesting lineage specificity and varying environmental effects between test groups. While all these studies provide candidate regions within salmonids, none were completed using whole-genome data and specifically relevant for age-at-maturity of precocial male Chinook salmon. Therefore, the genetic basis for maturation as a minijack is not well understood. Since some of the evidence points towards a polygenic architecture in this complex trait, signal detections can benefit from approaches that include whole genome scans with high density markers across chromosomes.

To investigate the genetic basis for minijack maturation in a hatchery Spring Chinook salmon population in the Columbia River Basin (interior stream-type lineage), a GWAS was conducted over two replicate years with low coverage whole genome sequence data. This study was designed to address two primary questions: 1) Is there evidence for a region or multiple regions in the Chinook salmon genome that are associated with precocial maturation at age 2 male fish? 2) Are there any genes located close or within these regions that have a functional role in the sexual developmental process of precocial age 2 males? To answer these questions, individual immature males and minijacks were identified through measurement of plasma levels of 11-ketotestosterone (11-KT), an index of reproductive development in salmonids that has been shown to be 99% accurate in predicting minijack phenotypes within this population (Medeiros et al. 2018).

Methods

Study fish

The study fish were male Spring Chinook salmon from two-year classes (brood years BY2018 and BY2019) that were progeny of first-generation hatchery-origin anadromous parents (i.e. SH line; Fast et al. 2015; Waters et al. 2015) spawned at the Cle Elum Supplementation Research Facility (CESRF, Cle Elum, Washington). The CESRF supplements the Upper Yakima River Spring Chinook salmon population. In 2019 and 2020, shortly after swim-up (6 months post spawned) the fish were transferred to the University of Idaho Aquaculture Research Institute (ARI, Moscow, ID). Fish care in this study was in accordance with and approved by the University of Idaho Animal Care and Use Committee. In March 2020 (BY2018) and April 2021 (BY2019) fin clips, blood samples, body weights (g), and fork lengths (mm) were collected from juvenile fish at ~ 1.5 years old. Male fish were identified through dissection and visually inspecting the gonads and additionally through genotyping the sex identifying marker, Ots_SEXY3-1 (Hess et al. 2023). All the visually inspected males also genotyped as males. The individuals included in this study were a subset of fish from a previously published study (Hoffman et al. 2023) and were randomly selected from the control tanks (BY2018: N = 183, BY2019: N = 201).The maturation status of each fish (immature male or minijack) was determined based on concentrations of plasma 11-ketotestosterone (11-KT) approximately 6 months prior to spawning as previously described (Hoffman et al. 2023). Plasma levels of 11-KT in the spring at approximately 1.5 years of age can be used to distinguish, with 99% predictability, between these phenotypes (Medeiros et al. 2018). Two-way analysis of variance (ANOVA) followed by Tukey’s multiple comparison test was used to test for differences in body weights and fork lengths of minijacks and immature males within and between brood years (α = 0.05; PRISM software version 10.2.0, GraphPad Inc., La Jolla, CA). Body weight and fork length measurements were log10 transformed prior to analysis to meet the assumptions of normality but, non-transformed values are shown.

Library preparation and sequencing

Genomic DNA was extracted from the fin clips of 384 individuals using the chelex method for both SNP genotyping and lcWGS libraries (Sweet et al. 1996). SNP genotyping was done for a panel of 347 markers (Hess et al. 2023) using the GT-Seq method (Campbell et al. 2015). Of those, 265 SNPs were putatively neutral markers which were included in principal component analysis (PCA) and kinship analysis to assess population structure and relatedness of individuals used in the GWAS. Any extent of kinship was measured by calculating pairwise kinship in GAPIT version 3 (Wang and Zang 2021) and FIS (inbreeding coefficient) values using Gpoppin package in R version 2023.06.1. To examine potential population structure which could bias genomic association analyses, PCA was run with the Gpoppin package in R version 2023.06.1. Preparation of lcWGS was done on eight individually barcoded libraries, which averaged 48 individuals per library (total = 384 fish). Library preparation and lcWGS was done with the Pool-Seq method described by Micheletti and Narum (2018). To increase the number of available reads the libraries were sequenced twice on the NextSeq™2000 platform using P3 reagents with a read length of 2 × 150 paired end reads.

Test groups, alignment, and association tests

The phenotypes for association tests were either immature males (IM) or minijacks (MJ) based on plasma 11-KT concentrations. For associations tests, two sets of test groups were created. First, sequenced reads were distributed into four groups based on maturation status and brood year (Fig. 1). The number of individuals included in the four phenotypic test groups were: BY2018: 101 immature (IM, Group 1) and 82 minijack (MJ, Group 2), BY2019: 100 immature (IM, Group 3) and 101 minijack (MJ, Group 4). Secondly, a subset of individuals from the first four test groups were partitioned into four groups that only included individuals from the extreme ends of each 11-KT bimodal distribution. (i.e. had 11-KT values lower than the mean 11-KT in the low modes or higher than the mean 11-KT value in the high modes, Fig. S1 in Supplementary Information 1). Selecting individuals from the extreme ends of the distribution can increase power to detect strong adaptive variation when intermediate phenotypes are excluded (Kardos et al. 2016). The number of individuals in the subset test groups were: BY2018: 51 IM (Group 5, low 11-KT) and 32 MJ (Group 6, high 11-KT), BY2019: 51 IM (Group 7, low 11-KT) and 51 MJ (Group 8, high 11-KT). Alignment to the reference genome and association tests for the extreme groups were done separately from the first set of test groups. The reads were aligned to the Otsh_v2.0 (NCBI: GCF_018296145.1) Chinook salmon reference genome using the PPalign module in the Poolparty pipeline (Micheletti and Narum 2018; Willis et al. 2023). This module trimmed paired end reads based on a minimum base quality score of 20 and reads with a minimum length of 50 base pairs (bps) were selected for alignment. Additionally, SNPs were called with the PPalign module. The parameters used for global SNP calling included, removing SNPs with a minimum depth of coverage less than 10, a minor allele frequency less than 0.01, and a SNP quality score lower than 20. After applying those filters, alignment statistics were generated using the PPstats module.

Fig. 1
figure 1

Frequency distributions of plasma 11-ketotestosterone (11-KT) in male (1 + years old) Spring Chinook salmon from two brood years (BY2018 and BY2019). Genome-wide association tests were done to compare allele frequency differences between immature (group 1 & 3) and minijack (group 2 & 4) phenotypes

Full size image

To identify candidate regions in the genome associated with maturation as a minijack, genome-wide allele frequencies were compared between the four test groups that included all the individuals from each bimodal 11-KT distribution (i.e. Group 1–4), with Fisher’s exact test (FET) for BY2018 and BY2019 separately. Additionally, the same phenotypic groups in BY2018 and BY2019 were combined, and the FET test was used to test for differentiated regions between phenotypes (i.e. Groups 1 & 3 vs 2 & 4). The same comparisons were then made for the subset test groups that included only the individuals from the extreme ends of each bimodal distribution. Group filters were incorporated prior to these tests, so that only SNPs with a minor allele frequency greater than 0.05 and a minimum depth of coverage greater than 30X (10X for subset test groups on the extreme ends of 11-KT distributions) were included. The SNP p-values from FET tests were then used for local score analyses using a tuning parameter (ξ) ranging from 1.04 to 1.22 (Fariello et al. 2017). The ξ is a user specified parameter which is the threshold (log10) scale below which p-values are autocorrelated (linkage). Adjusting ξ results in different levels of background significance. Lower values increase detection power of significant signals and association intervals become wider but can increase background variation in significance. Whereas higher values decrease background significance and association intervals shorten. Local score analyses were performed with the PPanalyze module in the Poolparty pipeline. The local score approach includes false discovery rate correction (Bonferroni), and significance was based on α = 0.01.

Results

Size differences

Two-way ANOVA showed that BY2018 fish had lower mean body weight and smaller mean fork length compared to BY2019 regardless of maturation status (Fig. 2). Maturation status had the biggest effect on body weight and fork length, explaining 31.2% and 27.0% respectively of the total variation for those traits (both traits P < 0.0001). Brood year had a smaller but still significant effect on body weight (2.7% total variation) and fork length (2.1% total variation; P < 0.0001 and P = 0.0008 respectively). The interaction between maturation status and brood year had no significant effect on body weight and fork length (P = 0.8344 and P = 0.7183 respectively). The multiple comparisons test showed significant differences in mean body weight between BY2018 (23.9 g) and BY2019 (29.7 g; P = 0.0063) immature males and also differences in mean body weight of minijacks (BY2018 = 44.7 g and BY2019 = 56.4 g; P = 0.0040). Additionally, there was significant differences in mean fork length between BY2018 (128.4 mm) and BY2019 (134.8 mm; P = 0.0283) immature males and in mean fork length of minijacks (BY2018 = 151.5 mm and BY2019 = 160.4 mm; P = 0.0100).

Fig. 2
figure 2

Comparisons of body weight (left) and fork length (right; mean ± SE) of BY2018 and BY2019 male Spring Chinook salmon used for this study. Asterisks indicate significant differences of log10 measurements (P < 0.05)

Full size image

Alignment results

For the first set of test groups the application of the global read depth filters on lcWGS resulted in 1.9 Gbps that were covered sufficiently by all libraries and 84% coverage of the genome. After applying group level filters, this reduced coverage to 1.6 Gbps and 70% of the genome (Fig. S2). Alignment resulted in 21,046,904 SNPs and a range of 7,076,935 through 8,276,122 SNPs were retained after SNP quality filters for association tests. The subset test groups (only individuals from extreme ends of distributions), resulted in 1.7 Gbps covered by all libraires and 76% coverage of the genome (Fig. S3 in Supplementary Information 1). A total of 22,360,234 SNPs were called during alignment. After applying group level filters, a range of 8,864,326 through 10,709,429 SNPs were retained for association tests.

Population and family structure

PCA and kinship/relatedness analyses were run with a set of 265 putatively neutral markers. PCA identified a limited amount of population structure within each brood year (Figs. S4-5 in Supplementary Information 1). Pairwise kinship analysis amongst individuals indicated that samples represented several families within a brood year (Figs. S6-7). Negative FIS values were detected for BY2018 (Group 1 FIS = − 0.015 and Group 2 FIS = − 0.034) and BY2019 (Group 3 FIS = − 0.029 and Group 4 FIS = − 0.015), but values indicated limited levels of relatedness in each brood year (Table S1 in Supplementary Information 2).

Association tests

The association test for the BY2018 group showed two significantly differentiated regions containing 184 SNPs associated with precocial maturation as a minijack (Fig. 3a and Table 1a). The differentiated regions were located on chromosomes 4 and 14. The highest peak was on chromosome 14 (local score peak = 61.7), and the identified region contained two genes: immunoglobulin kappa variable 3–15-like (IGKV3-15) and immunoglobulin lambda-2 chain C region-like (IGLC2). For the BY2019 group, results showed 134 differentiated SNPs spread across regions on chromosomes 4, 5, 8, and 10 (Fig. 3b and Table 1b). The highest peak was on chromosome 10 (local score peak = 50.2) and the identified region contained a single gene: NACHT, LRR and PYD domains-containing protein 1 homolog (NLRP1). After combining the same phenotypic groups in BY2018 and BY2019, the association test revealed 1,549 SNPs within multiple regions on chromosomes 5, 9, 15, and 18 were associated with precocial maturation (Fig. 4 and Table 1c). The most differentiated region was on chromosome 15 (local score peak = 891.6), and the identified region contained five annotated genes (signal peptide, CUB and EGF-like domain-containing protein 1 (Scube1), sorting and assembly machinery component 50 homolog (SAMM50), mitochondrial 10-formyltetrahydrofolate dehydrogenase (ALDH1L2), protein chaperone 1 (NOP), and ethanolamine kinase 1 (ETNK1).

Fig. 3
figure 3

FET local score Manhattan plots showing the significantly differentiated regions within the Chinook salmon genome when comparing immature males to minijacks in the BY2018 group (panel a) and BY2019 group (panel b). Red threshold line indicates α = 0.01

Full size image
Table 1 List of genes, and their chromosomal location, from significantly (α = 0.01) differentiated regions across the male Chinook salmon genome, when comparing by precocial maturation status. Data is displayed for BY2018 (a), BY2019 (b), and combined years (c)
Full size table
Fig. 4
figure 4

FET local score Manhattan plot from the combined years association test showing the significantly differentiated regions across the Chinook salmon genome after combining the same phenotypes between years (red threshold line indicates α = 0.01)

Full size image

FET local score Manhattan plots and significant regions from the comparisons of the extreme ends of each bimodal distribution provided additional candidate regions of the genome (shown as supplementary results). The BY2018 association tests showed 279 SNPs within regions on chromosomes 5, 7, 14 and 18 (Fig. S8.a in Supplementary Information 1 and Table S2.a in Supplementary Information 2). The strongest signal was on chromosome 7 (local score peak = 59.5). There was a single gene (rho guanine nucleotide exchange factor 10-like protein (ARHGEF10) associated with the identified region. For BY2019, 201 SNPs showed significant differentiation on chromosomes 3, 17, 26, and 30 (Fig. S8.b in Supplementary Information 1 and Table S2.b in Supplementary Information 2). The largest peak was on chromosome 30 (local score peak = 46.6) and within the identified region there are two genes; disks large homolog 2 (DLG2) and CREB/ATF bZIP transcription factor-like (GREBZF). Lastly, the association test that included the combined year groups resulted in 712 SNPs spread across chromosomes 5, 7, and 26 (Fig. S9 in Supplementary Information 1 and Table S2.c in Supplementary Information 2). As in the BY2018 comparison, the strongest signal was on chromosome 7, with the identified region containing the single ARHGEF10 gene.

Discussion

Conservation through supplementation typically aims to produce fish that have similar life history trajectories as natural origin populations, including age-at-maturity. High proportions of minijacks produced by hatcheries relative to natural populations can impede conservation efforts and diminish the number of anadromous adults available in fisheries. Further, minijacks may experience high mortality, cause ecological and genetic risk to other populations through straying, or fail to contribute towards future generations by not returning to spawning grounds (Pearsons et al. 2009, 2023). Thus, information on the genetic control of this trait may facilitate conservation and broodstock monitoring for supplementation programs related to precocial maturation as minijacks in hatchery and natural origin populations. To better understand the genetic factors influencing this trait, a GWAS interrogating millions of SNPs across the Chinook salmon genome was conducted to compare allele frequencies between immature males and males maturing as minijacks. The immature and minijack phenotypes were differentiated based on their plasma concentrations of 11-KT, high levels signifying a minijack and low levels in immature male. This is a critical, yet reasonable, assumption based on the known reproductive physiology of salmonids. Substantial literature exists supporting the role of 11-KT as the final hormone in a steroid pathway that regulates testes development (Campbell et al. 2003; Taranger et al. 2010; Butts et al. 2012). The analyses detected differentiated signals on multiple chromosomes for a subset of fish that were offspring of first-generation hatchery-origin fish representing upper Yakima River Spring Chinook salmon. The results from analyzing both years separately, as well as combining the same phenotypic groups from each year, indicated that the genetic architecture controlling precocious maturation was polygenic and likely not under the control of a single region of major effect in this population of Chinook salmon. The polygenic architecture remained when assessing differences between fish from the extreme ends of the phenotypic distributions between and within two separate years. Additionally, the variation of significant genomic signals and the differences in body sizes between replicate years may indicate that maturation as a minijack may be influenced by differential parental genetic effects on offspring growth.

Multiple candidate regions overlapped with earlier studies that examined genomic association with age at maturity, and putative signals on chromosomes 17 and 18 appear to be the most consistent with previous studies (Micheletti and Narum 2018; McKinney et al. 2020, 2021). Candidate regions on chromosomes 17 and 18 were consistent with previous studies, with results found in the combined year analysis and in the extreme ends of the phenotypic distributions in both BY2018 and BY2019 analyses. However, there were also differentiated regions detected in this study that were novel. In the BY2018 group the most significant peak was on chromosome 14 which contained two immunoglobulin genes that were located within a 58 kb region. To our knowledge, there’s no current evidence showing a role these genes play in maturation age. However, immunoglobulins are documented to be involved in immune responses that may play an essential role in disease susceptibility during the energetically demanding process of maturation (Mashoof and Criscitiello 2016). An additional gene on chromosome 4 showed a significant association for the BY2018 comparison. This gene is involved in carbohydrate biosynthetic processes but also affects luteinizing hormone production (Chou et al. 2024). Luteinizing hormone is a gonadotropin and a key component of the hypothalamus-pituitary–gonadal endocrine axis. It functions to regulate gametogenesis and final maturation in teleost species (Taranger et al. 2010). Mutations in this region may mechanistically regulate activation of the hypothalamic-pituitary–gonadal endocrine axis leading to maturation as a minijack. In the BY2019 group the strongest signal contained a single gene that is located on chromosome 10. This gene family is involved in innate immunity and inflammation (Martinon et al. 2007). Therefore, like the immunoglobulins, it may be important for disease resistance or physiological maintenance during the maturation process. Additionally, this gene on chromosome 10 has also been associated with reproductive processes in mammals (Tong et al. 2000). There were two large peaks for the combined year analyses. The first signal was on chromosome 15 and within the significant region there were five annotated genes. The second large signal was on chromosome 18 and within two significant regions, containing a total of nine genes. While it is unclear what regulatory function genes on chromosomes 15 and 18 have in the maturation process, candidate regions on chromosome 18 have been previously identified as associated with age at maturity in another population of Chinook salmon (McKinney et al. 2020, 2021).

Previous studies in Chinook salmon show evidence for polygenic architecture associated with maturation as a minijack. McKinney et al. (2020 and 2021) performed genome-wide scans using RADseq data aligned to the Otsh_v1.0 reference genome (NCBI: GCA_002872995.1) and identified haplotypes on chromosomes 17 and 18, as well as other SNPs on multiple chromosomes that were associated with precocial males (minijacks). We detected candidate regions of the genome that were similar to McKinney et al. (2021) for chromosomes 5, 8, 10, 17, 18, 28, and 30. Blasting the Otsh_v1.0 genome for the regional nucleotide sequences containing the significant SNPs from the current study showed none of the 11 SNPs on the same chromosomes associated with precocial males reported by McKinney et al. (2021) were located within those regions. Furthermore, Micheletti and Narum (2018) did not include precocial males but did whole genome-resequencing (CHI06, GCA_002831465.1 genome assembly) and pairwise comparisons of 3-, 4-, and 5-year-old Chinook salmon males and identified regions on chromosome 4, 6, 15, 18, and 30 associated with age-at-maturity. Comparing chromosomes with significant signals from that study showed overlap with chromosomes 4, 15, 18, and 30.

A noteworthy result from the current study was the observation of differentiated regions on chromosomes 17 and 18 for precocial males that were also identified as candidate regions for alternative maturation ages in previous studies on Chinook salmon. McKinney et al. (2020, 2021) hypothesized that sex specific haplotypes associated with age-at-maturity linked to sex determining regions on chromosomes 17 and 18 may provide a mechanism for resolving sexual conflict between males and females. This architecture would enable alleles driving age-at-maturity in males to be maintained within a population without influencing female maturation ages. The main sex determining region, sdY, is associated with chromosome 17 in Chinook salmon, but there is evidence for this region being translocated onto chromosome 18 (Phillips et al. 2013; McKinney et al. 2021). However, sdY is not mapped on the current Otsh_v2.0 reference genome and determining if the significant SNPs found are linked to sdY was not possible with the data from this study. Additionally, the primary focus of this study was to detect signals associated with precocial maturation as a minijack and didn’t include female fish, so it is unclear if the detected signals on chromosomes 17 and 18 are sex specific. Further research would be needed to identify linkage between the differentiated regions on chromosomes 17 and 18 with the sex determining regions to determine whether this genetic architecture is a mechanism for resolution of the sexual conflict between sexes in this population of fish. Altogether, the significant SNPs on chromosomes 17 and 18 reported in this study provide further evidence for the role these regions may have in precocial maturation, which rarely occurs in female fish.

A surprising result from this study was the lack of consistency and strength of signals between test groups from each brood year. Although there were significant peaks consistently observed on chromosomes 17 and 18, the SNPs on those chromosomes did not overlap. However, two other chromosomal regions were significant in more than one set of analyses. One was on chromosome 5 between the combined year analyses of all individuals and the individuals from the extreme ends of the phenotypic distributions. The second was on chromosome 7 for the BY2018 extreme group and the combined year extreme group. A possible explanation for the variability between years may be the polygenetic architecture detected for this trait with regions of small effect and an interaction between somatic growth, energy stores, and maturation. A substantial amount of evidence exists that environmental conditions, such as food availability and growth rate, have a major impact on precocial maturation in Chinook salmon. The probability that a fish will precociously mature is dependent on growth opportunities at the juvenile stage and energy acquisition during maturation windows (Clarke and Blackburn 1994; Shearer and Swanson 2000; Larsen et al. 2006; Shearer et al. 2006). In this study, weights and lengths (a proxy for growth and energy stores) were less in the BY2018 brood year compared to the BY2019. Therefore, a higher proportion of fish in the BY2019 group would have been above energetic thresholds for maturation compared to the BY2018 brood year. However, all fish included in this study were reared in identical water temperatures and under identical feeding protocols for each brood year (Hoffman et al. 2023). Thus, in the presence of identical rearing conditions, the variation in growth between the two years is likely the result of different parental genetic effects on offspring growth. Indeed, Chinook salmon jack life history has shown a paternal effect on juvenile growth which may subsequently govern juvenile life history pathways due to the strong interaction between energy stores and the capacity to precociously mature (Berejikian et al. 2011). The ability to detect strong signals of association for this trait and additionally, similar signals between years, may have been diminished due to parental genetic effects on growth and energy stores in the hatchery environment impacting the maturation process. Additionally, there was a level of relatedness between individuals within brood years due to the half-sibling families used for this study. While high levels of kinship in GWAS could cause inaccurate or family specific associations for the specific trait, several family groups were represented to reduce the chances of spurious results. However, future work may benefit from an individual level GWAS with high coverage sequencing data where family effects and other factors (size) could be included as cofactors in a mixed linear model approach.

Several factors may be involved with the observation of overlapping candidate regions on the same chromosomes with previous studies in Chinook salmon, but an inconsistent detection of the same SNPs. First, different versions of the Chinook salmon reference genome, which vary in completeness and quality, were used for alignment and the generation of SNPs in previous studies. Second, breadth and depth of genome coverage differed depending on the approach taken, which resulted in differences in the number of genome-wide SNPs available for comparing phenotypes. Lastly, differences in evolutionary and environmental forces acting on the populations studied may result in a variation in the genetic architecture involved for this trait. For example, several studies have identified SNPs with major effect in the VGLL3 region associated with age-at-maturity in Atlantic salmon (Ayllon et al. 2015; Barson et al. 2015; Kusche et al. 2017). However, further research in Atlantic salmon showed the effect of VGLL3 varies for populations that are separated by long distances, reared in different hatchery conditions, and that are adapted to different environments (Boulding et al. 2019; Mohamed et al. 2019; Besnier et al. 2024; Kess et al. 2024). Likewise, the variation in frequencies of certain Ots17 and Ots18 haplotypes across geographically separated populations of Alaskan and Columbia River Chinook salmon may represent historical genetic divergence of the architecture driving this trait or possibly differences in the genetic effects interacting with environmental effects between the populations studied (McKinney et al. 2020, 2021). Therefore, the underlying evolutionary and environmental forces may differentially influence the genetic architecture controlling this trait in a lineage or even population specific manner.

Conclusion

In this study, genome-wide association methods identified a polygenic architecture associated with precocial maturation as a minijack in a population of Spring Chinook salmon. Since high proportions of minijacks can have negative ecological and genetic effects on natural populations, this information is important for conservation and broodstock monitoring related to this trait. Several of the chromosomes with differentiated regions were the same chromosomes found in other Chinook salmon populations, most notably chromosomes 17 and 18. Candidate regions from this study provided further evidence of the influence of genetic variation on age at maturity and precocial maturation as a minijack. However, many of the regions of association showed relatively weak signals which are common for polygenic traits, but the large effect that growth related factors have on this trait may have also limited detection. Following further marker design, the candidate SNPs in candidate regions may be useful for monitoring shifts in allele frequencies and genotypic association at these sites across brood years to assess changes in genetic composition relating to precocial maturation in this population of Chinook salmon. Further validation would be important to untangle whether variation at these sites is population, sex, or lineage specific. The effect of environmental factors that differ between populations may complicate determining whether the selective signals found in this population of fish are consistent with Chinook salmon across geographic distances that are influenced by differing selective and environmental forces. Furthermore, many of the genes contained with regions of association on multiple chromosomes have an uncertain role in the maturation process and further studies would be necessary to evaluate the biological pathways that influence maturation as a Chinook salmon minijack.