Threatened and endangered populations of salmonid fishes in the Pacific northwest of North America are the subject of intensive conservation efforts and stocking programmes aimed at restoring these historic populations. Often, hatchery-raised fish have lower fitness than wild-born fish raising serious concerns about how hybridization may affect the long-term fitness of these threatened populations.
To assess to what extent stocking programmes impact the fitness of wild populations, wildlife biologists have invoked a variety of molecular-based techniques in combination with parentage analysis to measure how much gene flow from hatchery-born fish ends up in wild gene pools. However, populations that display a diversity of life history strategies pose some methodological issues as differences in migration timing or whether fishes migrate at all can obfuscate the true sources of gene flow in these populations.
A threatened population of steelhead trout (O. mykiss) residing in the Hood River, Oregon, offers a perfect platform to elucidate multiple sources of gene flow in a population with multiple life history strategies. This population consists of wild anadromous (i.e., migrating) steelhead supplemented by an extensive stocking programme of juvenile hatchery fish. The natural anadromous life cycle of wild steelhead has been impacted for nearly a century by the construction of the Powerdale dam (Fig. 1). For nearly two decades prior to its recent decommissioning and removal in the summer of 2010, all anadromous fish returning to their spawning grounds were passed over the dam, at which point they were genetically sampled. The net result of this sampling was the creation of a rich genotypic data set of hundreds of migrating steelhead each year.
Despite the effective sampling of all migrating steelhead, the assessment of hybrid fitness and introgression in this population is hampered because a fraction of wild and hatchery steelhead trout do not complete their anadromous life cycle. Instead, these fish spend their entire life upstream of the Powerdale dam as year-round residents rather than going out to the sea (Fig. 2). Because these resident fish are in low densities and complete sampling of individuals is not an option, these fish represent a previously unmeasured avenue of gene flow between wild and hatchery fish.
So how can we measure the genetic contribution of residualized hatchery fish to the next generation of steelhead with accuracy? To piece together the genetic contributions of residualized vs. anadromous steelhead, Christie et al. (2011) first use parentage analysis to match offspring with one or both anadromous parent(s). If an offspring can only be assigned to one anadromous parent, then Christie et al. (2011) employ a previously undescribed method of grandparentage analysis to test whether that offspring matches a known hatchery stock grandparental cross. Taken together, the combined approach of parentage and grandparentage analyses shows that while almost half of the genes in steelhead originate from resident parents, only a small fraction of the unsampled parents appear to be residualized hatchery fish. Moreover, their analyses reveal asymmetric gene flow from hatchery stock into wild populations, demonstrating that hatchery males produce many more offspring with wild females rather than hatchery females. Although their analyses cannot say with certainty how many resident matings originate from hatchery stock, this represents the most complete account of gene flow within this population to date.
Christie et al. (2011) new method of grandparentage analysis is a logical extension of well-established exclusion principles in parentage analysis. The power of exclusionary methods lies in the ability to capitalize on genotypic incompatibilities between parents and offspring (in this case grandparents and grandoffspring) to reject particular parent–offspring matches (Jones & Ardren 2003). In this case of steelhead trout, knowledge of hatchery breeding stock is sufficient to exclude putative grandparent, grandoffspring pairs.
The benefit of grandparentage analysis is that neither parental genotypes need to be known. However, the power to detect true grandparent–grandoffspring matches increases exponentially with the addition of one putative parent in addition to known grandparental crosses. Therefore, the application of grandparent analysis is particularly beneficial to multigenerational data sets where at least knowledge of grandparental breeding stock exists.
Grandparentage analysis is not without its problems, as any grandoffspring that fail to share at least one allele at any particular locus with a putative grandparent cross would be rejected under strict exclusion criteria. Thus, grandparentage analysis suffers from the same Achilles heel of all parentage analyses based on exclusionary methods—namely rejection of true parent–offspring pairs based on germ-line mutations, nonamplifying (null) alleles and genotyping errors (Jones & Ardren 2003). Compounding this issue is the fact that grandparentage analysis requires at least double the amount of microsatellite loci or alleles to have comparable levels of assignment success to traditional parentage analysis (Letcher & King 2001).
To circumvent these issues, exclusion assumptions can be relaxed which may also increase the chance of falsely assigning parent–offspring pairs. Christie et al. (2011) route this criticism by testing the exclusionary power of their methods under both stringent and relaxed criteria. They demonstrate that hatchery X hatchery crosses are under-represented under both criteria lending support to their hypothesis that these crosses are not as successful as crosses between hatchery and wild trout. Christie et al. (2011) also extend recently developed methods (Christie 2010) to determine the probability of false grandparent–grandoffspring trios and, using simulated data sets, demonstrate that their methods are remarkably precise.
With the proliferation of large, multigenerational data sets, it is simply a matter of time before grandparentage analysis will be routinely conducted to help piece together multigenerational pedigrees. The application of grandparentage analysis to breeding programmes and hybrid zone dynamics is obvious, and the implementation of grandparentage analysis to various other fields such as quantitative trait loci mapping, inbreeding and heritability estimation seems to be an appropriate next step. Given the high interest in standardized methods of parentage analysis (see Jones et al. 2010 for a recent review), the multitude of existing parentage computer programmes should incorporate multigenerational pedigree construction utilizing combined parentage and grandparentage analysis. Although we have just scratched the surface of what grandparentage analysis can accomplish, it will undoubtedly be a useful tool in the construction of wild pedigrees for generations to come.