Effectiveness of managed gene flow to reduce genetic and phenotypic change associated with captive breeding of Chinook salmon

Abstract

Captive breeding programs can rebuild depressed populations and aid in the recovery of threatened or endangered species. However, associated genetic and phenotypic changes may decrease the fitness of captive individuals when they are released into the wild and thus reduce restoration success. Genetic changes include loss of genetic diversity and divergence from the wild population, inbreeding, and adaptation to captivity, while changes in fitness traits, which affect both individual reproductive success and population productivity and resilience, may arise from genetic factors and differences between the captive and natural environments. Incorporating wild individuals as broodstock in captive breeding programs, which we refer to here as managed gene flow, is one strategy that may mitigate these potential risks. While this approach has been widely adopted in salmon hatchery management throughout the Pacific Northwest, it has not been empirically tested in a comparative framework over multiple generations. This dissertation characterizes genetic and phenotypic changes associated with captive breeding using novel genome-wide approaches and explicitly tests the effectiveness of managed gene flow to minimize these changes using two hatchery populations of Chinook salmon, Oncorhynchus tshawytscha. The hatchery populations were derived from the same wild population but are now managed as separate lines, one integrated with (i.e. managed gene flow) and one segregated from (i.e. no gene flow) the source stock, and thus provide an ideal system for comparing the two alternative management strategies. I used genomic and phenotypic data spanning five generations to examine the two hatchery populations across a range of measures. First, I used over 9000 loci to test whether managed gene flow between natural and captive environments, when compared to broodstock segregation, reduced genome-wide divergence from the wild founding population over four generations of captive rearing (Chapters 1-2). Genetic divergence from the source population was minimal in the integrated hatchery line, which implemented managed gene flow by using only naturally-born adults as captive broodstock, but significant in the segregated line, which bred only captive-origin individuals. Estimates of effective number of breeders revealed that the rapid divergence observed in the segregated line was largely attributable to genetic drift. However, we also identified temporally-consistent signatures of adaptive divergence within the segregated line, indicative of domestication selection. The results empirically demonstrated that using managed gene flow for propagating a captive-reared population reduces genetic divergence over the short term compared to one that relies solely on captive-origin parents. The findings also provided insight into the rate at which divergence may occur in integrated and segregated hatchery programs. Second, I computed genomic-based estimates of pairwise relatedness and individual inbreeding within the integrated and segregated hatchery lines across four generations and determined if managed gene flow successfully reduced the risks of inbreeding over time (Chapter 3). I also quantified the effect of inbreeding coefficient on eight fitness-related traits that had been measured in returning adults. The segregated line had slight but significantly lower levels of relatedness than the integrated line in the first generation but significantly higher levels in the third and fourth generations. Levels of inbreeding were similar between the two hatchery lines in the first, third, and fourth generations, despite 3- to 27-fold differences in estimates of effective numbers of breeders. However, inbreeding in the segregated line was significantly higher in the second generation. Inbreeding coefficient did not affect fecundity, reproductive effort, return timing, and fork length. In contrast, inbreeding significantly affected spawn timing, weight, condition factor, and daily growth coefficient, although the effects varied by sex, hatchery line, and generation. While the results indicated that managed gene flow may reduce the genetic risks of inbreeding, they also suggested that short-term risks may not be severe in small, segregated hatchery populations. The effects of inbreeding on fitness, however, require further examination, particularly at earlier life stages. Third, I identified loci associated with six fitness-related traits in adult Chinook salmon using an approach suitable for polygenic traits, Random Forest, and then explored the use of trait-associated loci within a management context; namely, whether they could serve as tools for monitoring the effects of alternative management approaches on genetic change underlying phenotypic traits (Chapter 4). I identified 226 unique loci associated with the six traits. Mapping of these trait-associated loci, gene annotations, and integration of results across multiple studies revealed candidate regions involved in fitness. Genotypes at trait-associated loci were then compared between the integrated and segregated hatchery lines. While no broad scale change was detected between the lines across four generations, there were numerous regions where trait-associated loci overlapped with signatures of adaptive divergence identified in Chapters 1 and 2. Many regions of overlap, primarily with loci linked to return and spawn timing, were either unique to, or more divergent in, the segregated line, suggesting that these traits may be responding to domestication selection. This chapter is one of the first studies to utilize genomic approaches to demonstrate the effectiveness of a conservation strategy, managed gene flow, on trait-associated – and potentially adaptive – loci. Last, I combined lessons learned from my analyses in Chapter 4 with information from colleagues and other studies to provide a simple, introductory guide to facilitate the use of Random Forest to identify genotype-phenotype associations in non-model organisms (Chapter 5). The guide first provides an overview of the Random Forest algorithm. Next, steps are described to prepare data for Random Forest, including initial data exploration and the identification of important covariates and possible confounding factors. Advice is then provided on the initiation and optimization of the Random Forest algorithm, along with a summary of methods for interpreting the results and identifying trait-associated, or predictor, loci. Annotated R tutorials are also included to assist users in implementing each step of the algorithm. This guide will hopefully facilitate the use of Random Forest in future ecological and evolutionary studies and contribute to the reporting of accurate and reproducible results. This dissertation research encompassed a broad perspective to characterize multiple genetic risks of captive breeding using novel genomic approaches, to link these risks to important fitness traits that were measured in adults, and to demonstrate that managed gene flow successfully mitigated potential adverse effects across four generations. By comparing two alternative management strategies, the study provides insight into the range of outcomes that may occur in captive breeding programs, which is highly relevant to risk assessment in realistic scenarios. Further, the results provide molecular tools to better monitor genetic change in hatchery and wild populations of salmon and further inform “best practices” in hatchery management to support declining wild populations. The findings also lay the foundation for future research efforts. For example, Chapters 1 and 2 revealed that using 100% natural-origin broodstock reduced genetic divergence across four generations. However, removing individuals from the wild for broodstock may also have genetic and demographic costs to the wild population, particularly since most populations supplemented by hatcheries are already in decline. A future study could use experimental populations to explicitly examine the effects of different levels of managed gene flow on divergence. Such information would enable conservation hatcheries to minimize genetic risks while also reducing potential costs of broodstock removal. Chapter 3 examined the genetic and phenotypic risks of inbreeding in returning adults. Yet, inbred fish may have significantly higher levels of mortality in the marine environment than non-inbred fish. Thus, the examination of inbreeding in adults may be inherently biased. A future study could sample both juveniles and returning adults from each hatchery line to determine if levels of inbreeding differ by life stage and if selection in the marine environment mitigates the risks of inbreeding in hatcheries. The identification of loci associated with six key traits in Chapter 4 is a first step towards characterizing the functional genetic basis of fitness in Chinook salmon. The regions where trait-associated and outlier loci overlapped will also provide useful starting points for future sequencing efforts that aim to identify the specific genes responding to domestication selection. In addition to specific research questions, this body work demonstrates the utility of genomic-based approaches in conservation monitoring and therefore provides an overall framework for other studies that aim to integrate genomics with the management of captive and wild populations.