Assisted gene flow from outcrossing shows the potential for genetic rescue in an endangered salmon population

Abstract

Genetic rescue has emerged as an important tool to prevent extinction and improve fitness of declining populations. In principle, genetic rescue increases genetic variation in a population through translocation of unrelated individuals from an outside source population. Genetic rescue remains uncommon in conservation management due to concerns about the risk of outbreeding depression. Lack of data evaluating success of genetic rescue interventions has hampered willingness to use this technique to improve the genetic viability of inbred, bottlenecked populations. Here, we evaluate the success of a genetic rescue intervention within the endangered Central California Coast Coho Salmon Evolutionarily Significant Unit using ∼17 years of genetic and demographic data, including pre- and postoutcrossing with fish from a nearby watershed. We assessed fitness of outcrossed F1 and F2 progeny in a captive setting, and then used mark–recapture to estimate survival of juveniles released into streams. We found that outcrossing decreased relatedness among adults, and their hybrid progeny had higher fitness in both captive and stream settings relative to nonhybrids. Importantly, we did not observe evidence of outbreeding depression in either the F1 or F2 generations. This study highlights that genetic rescue can be a useful tool in the conservation of imperiled salmonids.

INTRODUCTION

The ongoing biodiversity crisis has motivated investigations into strategies to prevent extinction and improve fitness of declining populations. One way to improve fitness is to increase genetic diversity by introducing additional variation through assisted gene flow to “genetically rescue” a population (Hedrick et al., 2011; Tallmon et al., 2004). Assisted gene flow is often accomplished through translocations of individuals from a genetically and demographically compatible population, and previous studies have found that the introduction of just a few individuals can help reduce adverse effects of small population size (Nathan et al., 2017; Tallmon et al., 2004). Indeed, genetic rescue has successfully contributed to population recovery of critically endangered taxa including the Florida panther (Puma concolor coryi) and Mexican wolf (Canis lupus baileyi) (Fredrickson et al., 2007; Hedrick & Fredrickson, 2010; Pimm et al., 2006).

Despite the relevance of genetic rescue for improving population fitness, it is underused as a conservation tool (Frankham et al., 2017; Whiteley et al., 2015). While increasing gene flow can reduce inbreeding and increase population growth, introducing novel genetic variation to a population also has the potential to lower fitness by disrupting coadapted gene complexes or introducing deleterious alleles, resulting in outbreeding depression (Edmands, 2007). However, recent work suggests that outbreeding depression may be of less concern than previously thought, and can be minimized by using source populations that occur in similar habitats and have low genetic divergence from the focal population (Frankham et al., 2011). For example, recent meta-analyses revealed that introduction of novel genetic variation into inbred populations usually resulted in higher fitness outcomes (Frankham, 2015; Ralls et al., 2020). Thus, the potential fitness benefits of genetic rescue generally outweigh outbreeding depression risks, particularly when inbreeding depression in recipient populations is severe (Ralls et al., 2018).

Such a genetic rescue intervention was undertaken with a severely inbred population of coho salmon (Oncorhynchus kisutch) from California's Russian River in the endangered Central California Coast (CCC) Coho Salmon Evolutionarily Significant Unit (ESU). A captive breeding program was established in the Russian River in 2001, to supplement the remaining coho salmon population and restore viability. While the captive breeding program uses careful protocols to avoid inbreeding (Conrad et al., 2013), this alone was not sufficient to overcome the negative effects associated with a small founding population and associated low genetic diversity. In 2008, outcrossing began with fish from nearby Olema Creek, also in the CCC ESU. The intent was to infuse additional genetic variation and decrease inbreeding, while preserving any regional genetic adaptation in the Russian River population.

Here, we assess the genetic and demographic effects of this genetic rescue intervention across the coho salmon life cycle. First, we tracked changes in relatedness among the broodstock between pre- and postoutcrossing periods. We then assessed fitness of F1 and F2 progeny in both captive and in-stream settings and used mark–recapture to estimate survival of juveniles released into the Russian River. We report that outcrossing had an overall positive impact on the Russian River coho salmon population by reducing genetic relatedness among spawning adults and improving fitness-related traits of hybrid fish relative to nonhybrids in both captive and in-stream settings.

METHODS Study system

A recovery program for coho salmon from the Russian River was initiated in 2001 following a severe population decline. Remaining juveniles in the watershed were captured and raised in captivity until reproductive maturity, with their offspring eventually released into Russian River tributaries. In California, coho salmon tend to have a 3-year life cycle. This leads to three distinct “brood cycles,” with somewhat distinct demography and ancestry. There is some gene flow among brood cycles due to precocious reproduction (i.e., age-2 maturation). At initiation of the captive-breeding program, the remaining natural-origin Russian River coho salmon available for use as broodstock were limited by the severe and recent population decline. Because of the limited number of founders and consequent low effective population size, best practices were implemented to decrease the risk of inbreeding in the captive broodstock program. These practices included bringing in both additional juveniles from Russian River tributaries streams, and adults returning from the ocean to spawn in the Russian River, when available. Pairing fish as mating partners across brood cycles (i.e., using precocious age-2 individuals) was also used to avoid severe inbreeding. However, this could only reduce relatedness to a point so the 2008–2009 spawning season marked the first year of outcrossing the Russian River captive broodstock with fish from Olema Creek (Figure 1) for approximately 25% of the total crosses (see Supporting Information S1; Tables S1 and S2). Olema Creek was considered a suitable source population based on persistently higher abundance, as well as genetic similarity and geographic proximity (Garza & Gilbert-Horvath, 2003). Outcrossing was performed in the captive facility rather than direct release/translocation of Olema Creek individuals into the Russian River. Using a captive facility ensured successful reproduction of Olema Creek and Russian River crosses as well as allowed control of the amount of outcrossing that occurred. For example, with direct release there would be no way to guarantee whether or not Olema Creek fish would reproduce with Russian River fish and/or control to what degree they reproduce relative to Russian River fish. Early concerns about potential outbreeding depression were overshadowed by clear signs of inbreeding and associated inbreeding depression. Conservation managers agreed that a controlled crossing and release with intensive monitoring, starting with a modest fraction of captive broodstock production, was the most prudent strategy to evaluate the risks of outcrossing in the face of critically low abundance.

FIGURE 1. Map showing (a) the study area within California, USA; (b) the Russian River Watershed and location of Olema Creek within the Lagunitas Watershed (gray shaded areas); (c) four Russian River tributaries (Mill, Willow, Dutch Bill, and Green Valley creeks) where juveniles were tracked using paired PIT antenna arrays (black circles); (d) adult coho salmon returns (note that observed returns are overlaid on top of estimated return numbers, Table S7) to the Russian River (adult return panel adapted from Obedzinski et al., 2021; see Supporting Information S2)

Genetic analysis

Tissue samples were collected from all broodstock individuals belonging to spawning cohorts from 2003 to 2018, and genotyped at 18 microsatellites (2003–2012) or 91 single-nucleotide polymorphisms (2013–2018) (see Supporting Information S1). To evaluate how outcrossing affected the Russian River adult broodstock over time, we estimated pairwise relatedness (r_XY) for each spawning year using the estimator of Queller and Goodnight (1989) as implemented in the “Related” package (Pew et al., 2015) in R (R Core Team, 2022). Mean individual relatedness measures the proportion of shared alleles between a pair of individuals relative to that of “unrelated” individuals from that population (Blouin, 2003), and was calculated for each spawning year as the mean of all pairwise r_XY estimates. We tested for a change in the slope of r_XY after outcrossing began in 2008 via a breakpoint-regression model (segmented package in R; Muggeo, 2008), with spawn year regressed against each mean r_XY estimate.

F1 hybrid performance in captivity

Following fertilization, eggs were incubated in trays with four sections (hereafter “sublots”). Fertilized eggs could be from one of three different crosstypes, depending on each parent's origin (Russian River [RR] or Olema Creek [OC]; sample sizes in Table S2): nonhybrid (RRRR, both parents RR; or OCOC, both parents OC) and reciprocal hybrids (OCRR and RROC with the origin of the female parent noted first). For each sublot, we modeled the proportion of individuals surviving to the hatch, eye-up, and swim-up stages of development (definitions in Supporting Information S1) and the proportion of individuals with deformities, using logistic regression models in R (glm function in R stats package; R Core Team, 2022). We first compared hybrids (OCRR and RROC combined) to RRRR nonhybrids, as well as differences among all three crosstypes (OCRR vs. RROC vs. RRRR), to test whether male/female parent origin influenced fitness. For this analysis, we built a total of five models each with either one or two predictors: year, crosstype, hybrid, year + crosstype, and year + hybrid status, which we regressed against the proportion of individuals that survived to each life stage or had deformities. Next, we constructed another set of models where we investigated how hybrid captive fitness compared to that of the source population (OCOC). For this analysis, we again regressed crosstype against the proportion of individuals surviving and the proportion of deformities, but with OCOC as the reference, instead of RRRR. Models were compared using Akaike's information criterion adjusted for small sample size (AIC_C; Burnham & Anderson, 2002). Models were considered competing for support if they were within 0–2 AIC_C units and/or if they carried >10% of total model weight.

F1 hybrid in-stream survival

After RRRR, OCRR, and RROC crosstypes reached ≥56 mm and 2 g in size, they were implanted with unique passive integrated transponder (PIT) tags and released into one of four Russian River tributaries (Dutch Bill, Green Valley, Mill, and Willow creeks; Figure 1c) at two different times of year: late spring (June) and fall (October–December). OCOC juveniles were not released into the Russian River. Channel-spanning paired antenna arrays were installed in Mill Creek in 2008, and in the three other streams in 2013, to detect fish postrelease until the outmigration stage. Between March and June, a fish trap was operated downstream of each antenna array and all captured individuals were scanned for PIT tags, which provided another opportunity to detect tagged individuals during the smolt outmigration season. The data from the PIT antennas and outmigrant trap were used to construct detection histories and estimate survival to outmigration (see Supporting Information S2).

Fitness outcomes from genetic interventions have been demonstrated to vary across time and space (Armbruster & Reed, 2005). Therefore, we investigated how outcrossing impacted hybrid survival using three different mark–recapture datasets of PIT-tagged fish. We created two temporal datasets for Mill Creek, one for spring-released fish and one for fall-released fish. Spring-released fish spend ∼4−6 more months in the Russian River than fall-released fish, so they were analyzed separately. We estimated survival of each crosstype in years with data available for all three crosstypes: fall-released fish (2009–2013) and spring-released fish (2009–2012) (Table S8[a and b]). We used a third dataset to investigate spatial variability in survival among crosstypes for fall-released fish in all four streams in 2013 (Table S8[c]).

We constructed multistate mark–recapture models (Schwarz et al., 1993) using Program MARK (White & Burnham, 1999) to estimate survival from postrelease to the outmigration stage using the capture histories of PIT-tagged fish (application of model described in Horton et al., 2011; see Supporting Information S1). Multiple models were constructed to evaluate if survival probability varied among crosstypes. For the multiyear Mill Creek datasets, models allowed survival to vary by year, crosstype, hybrid, both year and crosstype, or both year and hybrid status. Crosstype allowed survival to vary across three possible crosses (RRRR vs. RROC vs. OCRR), whereas hybrid allowed survival to vary between RRRR and the combined RROC and OCRR crosses. A null model in which survival was held constant across both year and crosstype was also included. In all of these models, detection probability (p) was allowed to vary by year. We constructed similar models for the spatial comparison, in which survival and detection varied by stream instead of year. We constructed a total of six models for each candidate model set and compared them using AIC_C.

F2 hybrid performance

Since fitness reductions from outbreeding depression can manifest in later generations, we also evaluated hybrid performance for available F2 crosses (see Supporting Information S3). We compared three crosstypes—RRRR×RRRR, RRRR×RROC, and RROC×RRRR—in the captive environment in spawn years 2011–2014, and 2018 (Table S11[a]). We also compared four crosstypes—RRRR×RRRR, RRRR×RROC, RROC×RRRR, and RROC×RROC—for in-stream survival of fall-released fish in 2013 (Table S11[b]). We constructed models for captive and in-stream metrics similar to the methods described above for the F1s. If F2 hybrids had significantly lower performance than RRRR fish, we interpreted this as evidence of outbreeding depression. If F2 hybrids had the same (or higher) performance, this was evidence that outbreeding depression had not occurred.

RESULTS Adult relatedness

Genetic analysis of the adult broodstock revealed that mean relatedness (r_XY) continued to decrease over time from the first broodstock spawning year in 2003 through 2018 (min = 0.0001, max = 0.23; Figure 2). As noted above, in the early years of the program, a variety of strategies was used to decrease relatedness, including incorporation of additional wild Russian River fish into the broodstock, when available, and pairing mates across cohorts (i.e., precocious males with age-3 females). However, this could only decrease relatedness to a point, given the small number of families contributing to the Russian River population. The breakpoint analysis showed that the slope of mean relatedness versus year changed from −0.0213 in 2003–2010 to −0.0052 in 2011–2018, coincident with the start of outcrossing. After a sharp decrease in relatedness in years 2011–2013, concurrent with the initial outbreeding, the slope asymptoted, as relatedness reached its practical minimum (since asynchronicity in reproductive maturation limits the ability to cross optimally unrelated fish for every spawning event).

FIGURE 2. Mean pairwise relatedness (rXY) of the adult broodstock for each spawn year. Brood cycles are arbitrarily numbered starting at 1 with the cohorts that were spawned in 2003. The first instance of outcrossing began in 2008–2010 for cohorts 1, 2, and 3, respectively. The black line represents a breakpoint regression result indicating a change in slope in 2011.

F1 hybrid performance in captivity

Model selection results showed strong support for differences among crosstypes for each of the four fitness metrics (survival at eye-up, hatch, swim-up stages, and proportion with deformities, Table 1[a–d]). In the crosstype + year models, OCRR juveniles had higher survival at all three life stages (effect size range = 0.37–0.75) and lower deformity rates (effect size = −0.48) than RRRR juveniles (Table S4). In contrast, there were no significant effects for RROC versus RRRR juveniles, but there may be biological significance of RROC hybrid status, given that the majority of the effect sizes indicated either higher survival or lower deformities. Model selection results also pointed to differences in fitness among progeny groupings when compared to OCOC crosses (Table S5). OCRR was not significantly different from OCOC for all three life stages or deformities, and RROC was found to have significantly lower survival compared to OCOC for all three life stages (effect size range = −0.388 to −0.688), but no significant differences in deformities (Table S6).

TABLE 1 Here, we include the most supported model(s) for each model comparison set performed to test if captive (model sets a–d) and in-stream (model sets e–g) survival varied by crosstype (RRRR vs. OCRR vs. RROC), hybrid, and year variables for the F1 generation. Model rankings were based on Akaike's information criterion adjusted for small sample size (AIC_C), differences in AIC_C (ΔAIC_C), model weight (w_i), model log-likelihood (log(L)), and the number of parameters (K).

Model	AICC	ΔAICC	wi	Log(L)	K
F1 hatchery metric AIC results
(a) Eye-up
Crosstype + Year	8219.24	0	1	–4096.59	13
(b) Hatch
Crosstype + Year	4069.43	0	1	–2021.69	13
(c) Swim-up
Crosstype + Year	7991.14	0	1	–3982.55	13
(d) Deformities
Year	2015.03	0	0.47	–996.49	11
Crosstype + Year	2015.33	0.30	0.41	–994.64	13
Hybrid + Year	2017.83	2.81	0.12	–996.89	12
F1 survival AIC results
(e) Mill Creek: Spring-release
S(hybrid, year)	14023.31	0	0.61	–6979.48	32
S(crosstype, year)	14024.22	0.91	0.38	–6972.85	39
(f) Mill Creek: Fall-release
S(hybrid, year)	28964.29	0	0.99	–14437.01	45
(g) Spatial
S(hybrid, stream)	17424.57	0	0.99	–8677.20	35

Note: Models were considered competing for support if they were within 0–2 AIC_C units and/or if they carried >10% of total model weight, in which case multiple, top supported models are presented. The most supported models are presented for each of the hatchery metrics separately: (a) Eye-up, (b) Hatch, (c) Swim-up, and (d) Deformities, and each of the in-stream survival models (S): (e) Mill Creek: Spring-released, (f) Mill Creek: Fall-released, and (g) Spatial. Note that (a) to (g) are all separate model comparison sets showing results for models with the most support only, and thus AIC_C values cannot be compared across candidate sets. Fully expanded AIC_C tables for each model set with all models constructed are in Table S3 (hatchery metrics) and Table S9 (in-stream survival).

F1 hybrid in-stream survival

In-stream survival model selection results showed that differences in survival among progeny groups extended postrelease (Table 1[e and f]). For both Mill Creek datasets, we found that survival probability was higher for hybrids than nonhybrids (Figure 3). There was a larger difference in survival probability between hybrids and nonhybrids (Table S10) in the spring-released fish dataset (mean difference = 0.04 across all years) compared with fall-released fish (mean difference = 0.03). Fall-released fish also tended to have a higher overall probability of survival. While allowing survival to vary by hybrid status was most supported, OCRR also had higher survival relative to RROC and RRRR crosstypes. This difference was more pronounced in the spring-released fish, with OCRR 0.08 higher on average than RRRR, and RROC 0.03 higher than RRRR (Figure 3). Survival differences for fall-released fish were again slightly smaller than in spring-released fish, with OCRR 0.04 higher than RRRR, and RROC 0.02 higher than RRRR.

FIGURE 3. F1 in-stream survival estimates and 95% confidence intervals for spring- (a, b) and fall-released (c, d) Mill Creek fish. The top row represents results from the S(Hybrid + Year) and the bottom row represents results from the S(Crosstype + Year) model (see Table 1). Survival estimates are presented for each year of tracking. For a given year, spring-released fish spend 8–12 months in freshwater and fall-released fish spend 3–8 months in freshwater before emigrating to the ocean.

For the spatial dataset, survival varied by hybrid status and stream (Table 1[g]). Hybrids on average had 0.07 higher survival probability relative to nonhybrids, but this varied by stream (range = 0.01−0.11). While the model that allowed survival to vary by all three crosstypes and stream was less supported, survival followed a similar pattern. Survival probabilities for the OCRR and RROC crosstypes were 0.08 and 0.05 higher, respectively, than for RRRR, but the magnitude of difference varied by stream (Figure 4).

FIGURE 4. F1 in-stream survival estimates and 95% confidence intervals for (a) S(Hybrid + Stream) and (b) S(Crosstype + Stream) models for all four streams (fall-released fish in 2013) in the spatial dataset in Table 1

F2 hybrid performance

We did not observe evidence of outbreeding depression, as F2 hybrids did not have significantly lower fitness outcomes than RRRR×RRRR individuals in either the captive or stream setting. In the captive environment, model selection showed some support for differences among F2 progeny groupings (model weight ranged from 0.12−0.38; Table S12), where RROC×RRRR hybrids had higher survival metrics (eye-up, hatch, swim-up) and lower deformity rates relative to RRRR×RRRR than when comparing RRRR×RROC to RRRR×RRRR, but the majority of effect sizes were nonsignificant (Table S13). Postrelease, there was some support for in-stream survival differences among hybrid groupings (model weight = 0.37; Table S14), where F2 hybrids with Olema ancestry had higher in-stream survival (mean effect size = 0.02) relative to RRRR×RRRR fish (Table S15; Figure 5), but the magnitude of this difference varied across streams (range = −0.0007 to 0.05).

FIGURE 5. F2 in-stream survival estimates and 95% confidence intervals for (a) S(Hybrid + Stream) and (b) S(Crosstype + Stream) models for all four streams (fall-released fish in 2013) in the spatial dataset in Table S12

DISCUSSION

Here, we demonstrate that introducing genetic variation from another population provides significant fitness benefits to a declining population of coho salmon. Outcrossing decreased relatedness and alleviated inbreeding risk among the adult broodstock, and improved values of fitness-related traits in outcrossed progeny in both captivity and natural habitat. We found no evidence of outbreeding depression in the F1 and F2 generations due to this genetic intervention, which may be due to the proximity and habitat similarity of the donor and recipient streams (Garza & Gilbert-Horvath, 2003; Gilbert-Horvath et al., 2016). Our work adds to the evidence that genetic rescue provides fitness benefits for imperiled populations and highlights its importance as a strategy for conservation and recovery.

Whether assisted gene flow results in genetic rescue depends on interactions among factors such as genetics, evolutionary history, demography, and environmental context (Fitzpatrick et al., 2019; Tallmon et al., 2004). Notably, we observed that differences in all fitness-related traits, in both captive and in-stream settings, were greatest for hybrids when the female parent was external (i.e., from Olema Creek). This suggests maternal effects, which can arise when maternal investment in offspring (e.g., nutrients, hormones, parental care) is greater than paternal investment (Mousseau & Fox, 1998), although contributions from genetic factors, such as mitochondrial–nuclear interactions or sex-linked traits, may also play a role (Bertho et al., 2021; Moran et al., 2022). Thus, consideration of maternal effects may be important when devising outcrossing protocols for conservation breeding programs. Postrelease in the stream setting, the magnitude of the differential effects by crosstype was partly dependent on environmental context, with variation in survival observed across both time and space. Environments can vary in their selection regimes, and the amount of fitness variance (i.e., opportunity for selection) can determine the strength and direction of selection (Brodie et al., 1995; Siepielski et al., 2009). The opportunity for selection is maximized when the survival rate is intermediate, and this is also when we observed the largest effect of crosstype. Specifically, we observed little difference among crosstypes when there was little variance in relative fitness, that is, during periods of higher survival (e.g., Mill Creek fall-released fish) or very low survival (e.g., in 2013). In the Russian River, variability in spatial patterns of wetted habitat (Moidu et al., 2021) and flow conditions influence survival of coho salmon (Obedzinski et al., 2018; Vander Vorste et al., 2020). Exploring the impacts of environmental variation on survival and genetic rescue outcomes is an important area for future study.

The fitness differences among crosstypes measured here demonstrate the potential for genetic rescue to aid small and declining populations of coho salmon and other species. Until recently, there has been little guidance on how to evaluate genetic rescue, but there is now some consensus that an increase in population growth rate is a necessary criterion for success (Bell et al., 2019; Robinson et al., 2021). However, demographic increases may not be evident or possible if there is limited habitat for population expansion or environmental conditions are deteriorating (Hedrick et al., 2011). A marked increase in returns of coho salmon to the Russian River was observed beginning in 2010/2011, which coincided with the first outcrossed progeny returning as adults (Figure 1d). Adult returns to the basin have been monitored since the inception of the program, but it is challenging to quantify the impacts of outcrossing on population growth given that so few individuals survive to adulthood and because so few fish in the system are handled as adults, so we cannot directly attribute this change to outcrossing. Moreover, monitoring protocols have changed over time, juvenile releases have increased, and environmental conditions may have been more favorable in certain years than in others (although drought conditions have dominated in the latter part of the time series).

Our study differs from earlier work on genetic rescue in which we used captive breeding to implement a carefully controlled set of reciprocal crosses, coupled with subsequent evaluation of hybrid survival in intensively monitored natural habitat. The majority of published work on genetic rescue involved direct translocations into a recipient population (e.g., Fitzpatrick et al., 2016; Fredrickson et al., 2007; Pimm et al., 2006; Westemeier et al., 1998). An advantage of translocation is that it generally involves lower costs and effort relative to ex situ methods. However, in our study, captive breeding allowed generation of individuals from different crosstypes in sufficient numbers to adequately assess fitness effects. The use of controlled crosses in a captive setting also ensures that new genetic variation is incorporated into the recipient population. However, effects of gene flow can be very different under captive and natural conditions. In natural settings, environmental stress can increase the effects of both inbreeding and heterosis following gene flow (Crnokrak & Roff, 1999; Fitzpatrick et al., 2016).

Additionally, genetic rescue studies have varied in the magnitude of gene flow or the number and frequency of translocated individuals. While outcrossing has occurred in the Russian River captive program every year since 2008, other studies have undertaken a single translocation event. In fact, Frankham (2016) demonstrated that genetic rescue benefits can persist to at least the F3 generation. Management guidelines often recommend “adequate” numbers of individuals be released to counteract inbreeding and additional loss of genetic diversity, but these numbers are rarely quantified (but see Nathan et al., 2017; Tracy et al., 2011). Important work remains to be done on the optimal frequency of gene flow and number of individuals released or incorporated into the recipient population. The numbers will depend on population size, prior inbreeding, life-history traits, and the carrying capacity of the release sites, among other things.

Given continuing habitat degradation and fragmentation, it is important for conservation plans to better address potential inbreeding depression and loss of genetic diversity (Ralls et al., 2018). Assisted gene flow may provide the variation needed to persist, and increase adaptive potential. Indeed, recent work has shown that selection can resist swamping even under high gene flow (Fitzpatrick et al., 2019) and that differently adapted populations may also benefit from genetic rescue when crossed (Fitzpatrick et al., 2020). The ESU concept was developed for the purpose of preservation of unique adaptive genetic diversity (Moritz, 1994). Even so, the potential for outbreeding depression exists in salmon, given the prevalence of local adaptation (Waples, 1991), even among differentially adapted populations within an ESU. However, small populations that have undergone strong genetic drift are generally more at risk from demographic stochasticity than from loss of fine-scale local adaptations (Lande & Barrowclough, 1987; Wright, 1931). So it is important for conservation policy to include a risk assessment to weigh the costs and benefits of not introducing novel genetic variation, alongside the risk of outbreeding depression for doing so (Liddell et al., 2021; Love Stowell et al., 2017; Weeks et al., 2016) and decision-support tools for accomplishing such a risk assessment are available (e.g., Frankham et al., 2017; Ralls et al., 2018). This should be coupled with careful consideration of the source populations for outcrossing, especially when potential source populations are from another ESU or otherwise distinct environment, or when declining population size in the recipient population has not yet induced strong genetic drift. Regardless, our study provides strong evidence for the direct benefit of genetic rescue to fitness at multiple life stages in an imperiled salmon population, and contributes to the growing literature on the benefits of genetic rescue for the conservation of endangered species.

AUTHOR CONTRIBUTIONS

Kasey C. Pregler led composition of the manuscript and conducted the statistical analyses. Stephanie M. Carlson and John Carlos Garza helped frame the paper and provided feedback on multiple manuscript iterations. John Carlos Garza and Elizabeth A. Gilbert-Horvath provided genotype data and guided the crossing design. Benjamin White provided early life-history data. Mariska Obedzinski led postrelease data collection and conducted survival analyses. All co-authors contributed to development of the study design, writing, and providing feedback on the manuscript and analyses.

ACKNOWLEDGMENTS

K.C.P. was supported by a UC Berkeley Chancellor's Fellowship. We are grateful to lab staff, hatchery, and field crews for collecting data, and the many stakeholders that support the Russian River Coho Salmon Captive Broodstock Program. We thank G. Horton for assistance with survival models, and B. Coey and the anonymous reviewers for feedback on an earlier version of this manuscript. Publication made possible in part by sypport from the Berkeley Research Impact initiative (BRII) sponsored by the UC Berkeley Library.

FUNDING INFORMATION

Analysis and writing of this paper was supported by the University of California, Berkeley Chancellor's Postdoctoral Fellowship (no grant number). The US Army Corps of Engineers provided funding support for hatchery operations, field monitoring, and genetic analysis. California Department of Fish and Wildlife also contributed funding for field monitoring activities.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT

Data used in this study are available upon request.