Genetic Identification of Simulated Lake Trout from Lake Michigan: Efficacy of Genetic Algorithms to Determine Cross-Strain Hybrids based on Available Genetic Data

Abstract

In the 1950s, lake trout (Salvelinus namaycush) were functionally extirpated from Lake Michigan from a combination of overharvest and introduced species. Since the 1960s, restoration efforts to return lake trout to Lake Michigan have involved annual stocking of millions of yearling lake trout from four strains into the lake. However, restoration of self-sustaining lake trout populations has not been successful. If lake trout restoration is to be successful, fisheries managers will need to focus on stocking lake trout strains that are either adapted to their stocking locations or have adequate genetic diversity and shown reproductive success in past efforts. One of the best ways to assess strain fitness is to determine reproductive success and survival of different strains through genetic data analyses. While genetic data analyses are essential for determining which strains are most appropriate for stocking, questions remain if current genetic markers and statistical approaches are able to accurately determine the origin strain of an individual lake trout that is the result of hybridization between individuals of two different pure strains. Previous studies have found genetic markers to be useful for studying the origin of pure strain lake trout. But as lake trout aggregates of multiple broodstock origins congregate on spawning reefs, spawning between lake trout of different hatchery origins may create interstrain hybrid offspring. Many statistical computer algorithms have been written for identifiying individuals to populations, but as of yet, most of these models are untested for identifying individuals of hybrid origin. If fishery biologists rely on results from these models when making management decisions, models must be both accurate and unbiased, even for hybrid individuals. I tested six individual assignment or hybrid/admixture algorithms (i.e., BAPS v4.14, NewHybrids v1.1, STRUCTURE v2.2, GeneClass2, GMA, and WhichRun) to determine their accuracy and bias for assigning parentage to individuals from a simulated population of pure strain and interstrain hybrid offspring. These algorithms were moderately successful for assigning pure strain individuals, but less successful for assigning interstrain hybrid offspring. Success in assigning pure strain individuals ranged from 8.9% to 40.0% in the maximum likelihood (ML) algorithms and from 36.8% to 99.6% in the Bayesian clustering algorithms. For F1 hybrid individuals, assignment success ranged from 4.3% to 46.2% in the ML algorithms and from 7.0% to 95.7% in Bayesian clustering algorithms. Assignment success of advanced hybrid individuals (i.e., Bx and F2) only exceeded 65% in one algorithm (i.e., NewHybrids). While NewHybrids consistently had the highest assignment success, its use in studies of Lake Michigan‟s lake trout may be limited by the algorithm‟s limitation of a maximum of two reference populations during assignment. As a result of this poor performance, none of these methods were sufficient to be used for the assignment of lake trout of unknown origin with currently published genetic markers. Therefore, I was unable to determine the origins of lake trout eggs and fry collected from the mid-lake reef complex. Developing additional lake trout microsatellite genetic markers may provide tools necessary to distinguish among stocked strains and their potential hybrid offspring. Eleven lake trout specific microsatellite loci were developed and grouped into three multiplex reactions. These loci had an average heterozygosity of 0.450 while individual locus heterozygosity ranged from 0.023 to 0.810. The number of alleles per locus ranged from 2 – 22. These reactions were tested for performance with four other salmonid species with varying success. Between 2 and 7 loci were successfully amplified for each salmonid, and a subset of each were polymorphic.