1 | INTRODUCTION
Understanding the resilience of biological marine resources to changing oceanographic conditions is central to ecosystem-based fisheries management and the implementation of adaptive sustainable harvest strategies (Link 2002; Levin and Möllmann 2015). The ability of populations to respond to disturbances in their habitat is in part determined by the genetic diversity present in the population (Parker et al. 2000; Hoffman and Sgro 2011). This genetic diversity, exhibited by a portfolio of available gene variants, allows for a quick response if selectively advantageous variants are already present in the population (Sunday et al. 2011; Pacifici et al. 2017). Understanding the response of marine populations to environmental perturbations will allow us to readily assess the resilience or vulnerability of these populations and species.
Measuring differential survival between subpopulations reveals how environmental conditions can influence the overall productivity of exploited populations. For example, Schindler et al. (2010) demonstrated that environmental conditions favored the production of discrete salmon populations residing in Bristol Bay, Alaska. Maintenance of a portfolio of locally adapted genotypes ensured adult returns to the region, but returns to different streams were maximized under different environmental conditions. This maintenance of a portfolio of adaptive alleles may be a key aspect of resilience of populations to natural environmental variability.
The idea of a temporal portfolio effect, in which adaptive variation is maintained by overlapping generations in a temporally variable environment, has been studied in general (Ellner and Hairston 1994). For example, many freshwater zooplankton taxa have relatively short-lived adults that may be subject to strong selection, but eggs can remain viable for decades in sediment, resulting in persistent egg banks that are relatively buffered from environmental variation (Brendonck and DeMeester 2003). We hypothesize that a similar effect may occur in marine fish species with highly dispersive larvae and long-lived, relatively sedentary adults. The genetic composition of each recruitment cohort may reflect relatively strong selection during the larval stage, while the adult population would maintain genetic variation reflecting multiple cohorts. Here we test this hypothesis using genomic methods for detecting population structure and adaptive loci.
Genomic data allow us to scan for individual and population-level differences across the whole genome, and genomics is becoming integral in answering a wide array of previously unresolved questions in conservation biology with numerous applications in fisheries (Wenne et al., 2007; Barrio et al. 2016; Jasonowicz et al., 2016; Valenzuela-Quiñonez, 2016; Kumar and Kocour, 2017). It is now possible to estimate, with a high level of precision and certainty, the demographic structure of fish populations at small spatio-temporal scales, and to identify local adaptation from genomic data (Wang and Höök, 2009; Barrio et al., 2016; Catchen et al. 2017; McKinney et al. 2017). RADseq approaches have been extensively used to describe various biological and ecological phenomena, such as phylogeography, population differentiation and structure, population and individual admixture (composition of lineages), genetic diversity, and outlier loci detection, among others (Alexander et al. 2009; Andrews et al. 2016; Narum et al. 2013).
Our model species, the Pacific ocean perch (POP; Sebastes alutus ), is a long-lived species with its oldest individuals being over 100 years old (Conrath and Knoth 2013). POP are the most abundant and economically important rockfish species in the Gulf of Alaska (Conrath and Knoth 2013) with landings in excess of 55,000 tons in 2017 (NOAA 2019). The fishery is managed using an age-structured model where the vital population rates are derived from the abundances of different ages in the catch (Megrey 1988; Hulson et al. 2017), but the relationship between the abundance of spawning fish and their offspring cohort is highly variable and unpredictable. This extreme annual fluctuation in success and failures of various year classes has been noted as a characteristic of this and many other commercially exploited species (Westrheim 1958; Carlson and Haight 1976).
POP in the Gulf of Alaska live on the upper slope of the continental shelf. They spawn from September through November, with parturition occurring in April through May the following year, when larvae rise from demersal spawning habitats on the continental shelf break (150 – 400m depth) to surface waters. They then become part of the ichthyoplankton and within a few weeks metamorphose to a young-of-the-year form (YOY). They are carried in the surface waters by currents and settle out of the water column in nearshore rocky habitat by the end of their first year (Carlson and Haight 1976; Major and Shippen 1970). During their shoreward movement, larvae grow rapidly and allocate significant amounts of energy to creating lipid tissue. This lipid tissue is apparently lost during settlement (Moss et al. 2016), suggesting energy acquisition and growth are important determinants of settlement success (Hoey and McCormick 2004). They remain in the nearshore habitat for the next few years until they join the discrete adult schools residing on the continental shelf and slope (Love et al. 2002). They reach sexual maturity at eight to ten years of age and repeatedly spawn until their hundredth year or longer (Hulson et al. 2017). These adult schools are genetically differentiated and the degree of their differentiation (FST) is correlated to the geographic distance between them (Palof et al. 2011).
Larval dispersal pathways in POP may be highly variable from year-to-year since they mostly depend on ocean currents in a given year (Mundy et al. 2010). An important prediction of ocean current and dispersal models (Stockhausen 2009; Stockhausen and Hermann 2007) is that in each year, the larvae at a given pelagic location are comprised of mixtures of individuals from different spawning locations indicating a high degree of mixing among them. However, population genetic studies of young-of-the-year and adults indicate there is limited mixing among subpopulations (Palof et al., 2011; Kamin et al., 2014). The Kamin et al. (2014) follow up study examined the YOY POP catches corresponding to locations near the adults caught by Palof et al. (2011). Their work showed that the collections of YOY POP were most related to the linearly closest adult populations. Either widely dispersed juveniles are able to return to their natal areas, or survival is maximized among locally retained larva, possibly due to local adaptation.
Here, we test whether POP larvae exhibit signatures of selection that could allow for the maintenance of a portfolio of adaptive variation in the multi-cohort adult population. We examined YOY POP collected from the eastern Gulf of Alaska during two years (2014 and 2015), when the oceanographic conditions were drastically different, with 2014 being an average temperature year, and 2015 being anomalously warm which is expected to have a negative impact on the fish (Cavole et al. 2016; Gentemann et al. 2017; Jones et al. 2018). We evaluated the potential for differences in selection strength for YOY POP across years by testing whether the fish differed in physiological conditions in 2014 and 2015, measured as a body condition index based on weight-length relationships, and total lipid content. We then used genotype environment association (GEA) tests with RADseq genomic data to test for differences in selection acting on the genome to favor different phenotypes between the two years. Finally, we identified candidate biological pathways on which selection was acting in the two different YOY cohorts. We predicted that the strength of selection would be higher in 2015 than 2014 due to the unusually high 2015 sea temperatures, resulting in poor body condition and a greater number of SNPs associated with environmental variables and physiological condition in the 2015 dataset. These findings may explain the difference in recruitment for the 2014 and 2015 cohorts as estimated in the 2017 stock assessment (Hulson et al., 2017).