Computational protocol: Low levels of hybridization between sympatric Arctic char (Salvelinus alpinus) and Dolly Varden char (Salvelinus malma) highlights their genetic distinctiveness and ecological segregation

Similar protocols

Protocol publication

[…] We used MICRO-CHECKER (van Oosterhout et al. ) to check the microsatellite data for errors in scoring such as stuttering and null alleles which can compromise subsequent analyses. Next, we used FSTAT ver 2.9.3 (Goudet ) and ARLEQUIN ver 3.5 (Excoffier and Lischer ) to generate the basic descriptive statistics of sample size (N), number of alleles (NA), allelic richness (AR), and observed (HO) and expected (HE) heterozygosity. Tests for departures from Hardy–Weinberg equilibrium (HWE) for each locus–locality combination were performed with GENEPOP ver 4.2 (Raymond and Rousset ) using an exact test in which probability values were determined using a Markov chain method (P) controlling for multiple tests as per Narum ().To obtain a general assessment of the number and distinction of genetic groups in Lake Aleknagik and Lake Nerka without a priori assignment to taxon, we conducted a factorial correspondence analysis (FCA) in GENETIX (Belkhir et al. ) on allele frequencies across all loci of Lake Aleknagik and Lake Nerka char alone, and with reference populations of allopatric Arctic char from the central Canadian Arctic (N = 43) and a reference sample Dolly Varden (N = 29) from the Egegik fishing district located south of Dillingham and Bristol Bay, Alaska, at the outflow of Becharof Lake (see Taylor et al. and Hart et al. for details). Learning samples of Arctic char and Dolly Varden were assigned to species using a combination of morphology, biogeography, and diagnostic microsatellite markers (see below). The FCA is a multiallelic factor analysis that summarizes variation in categorical variables (allele counts across multiple loci) in allelic space across a small number of dimensions. To test for population subdivision for all samples pooled within Lake Aleknagik and Lake Nerka, we ran simulations of K = 1 to K = 10, repeated five times in the Bayesian program STRUCTURE ver. 2.3.4 (Pritchard et al. ) for each sampling year (2012, 2013) separately to detect any interannual differences that might exist and were the data using DISTRUCT (Rosenberg ). STRUCTURE is a model-based clustering algorithm that assigns individuals to a K number of genetic groups that minimizes departures from HWE and LD within populations based on their multilocus genotypes. To infer support for the most probable number of subpopulations for each sampling year, ΔK (Evanno et al. ) was calculated across multiple runs of STRUCTURE using STRUCTURE HARVESTER (Earl and von Holdt ).To assign individuals as Arctic char or Dolly Varden after confirming that a K = 2 was an appropriate model (see above and Results), we used an ad hoc approximation of species identity where fish whose Q values were ≥ 0.95 (i.e., at least 95% of the genome characteristic of Dolly Varden, QDV) were classified as Dolly Varden and fish whose QDV values were ≤ 0.05 were classified as Arctic char. Preliminary assignment of Q value groups to species was based on established morphological differences between Arctic char and Dolly Varden (e.g., see McPhail ). Any fish that had QDV values ≥ 0.05, but ≤ 0.95 were noted as possible hybrids. Next, we used a simulation approach to generate a range of QDV values that would be indicative of hybrids (see Vähä and Primmer ). First, we used the program HYBRIDLAB (Nielsen et al. ) to create 200 simulated F1, F2 and backcrossed hybrids generated from the random selection of alleles from each locus in the reference population of Arctic char from the central Canadian Arctic (same as above) and in Dolly Varden sampled from Vancouver Island and the Queen Charlotte Islands in British Columbia (BC, N = 40). We used BC Dolly Varden from northern BC because we were unsure of the exact river of origin of the Egegik Dolly Varden and whether or not they were truly allopatric relative to Arctic char. The Dolly Varden from BC are a different subspecies (S. malma lordi) than the Egegik Dolly Varden (S. m. malma, see review by Kowalchuk et al. ; Taylor and May-McNally ), but the results of simulations using either subspecies did not change and we gained the advantage of knowing that the BC Dolly Varden are allopatric with respect to Arctic char. Next, these reference populations were designated as “learning samples” by implementing the USEPOPINFO model combined with simulated hybrids and char from Lake Aleknagik and Lake Nerka through five replicated analyses in STRUCTURE (Pritchard et al. ). We took the average values between the upper and lower limits of possible QDV values for the simulated hybrids across the five replicated analyses to create a “zone of hybridity”. Admixed individuals from the Wood River basin whose QDV values fell inside this range were classified as putative hybrids and were used to generate the level of hybridization for each lake. Our sample sizes for allopatric Arctic char and Dolly Varden used in the simulation above, especially for Arctic char, were relatively modest. Consequently, we generated a second simulated zone of hybridity using 50 Arctic char from Lake Aleknagik and Lake Nerka with QDV values closest to 0 (QDV < 0.05, pure Arctic char) and 50 Dolly Varden with QDV values closest to 1.0 (QDV ≥ 0.88, pure Dolly Varden) using the same methods as above. Both zones of hybridity were then compared to the range generated by Taylor et al. () for Arctic char and Dolly Varden in southwestern Alaska. The STRUCTURE analyses were conducted using a burn-in period of 50,000 Markov Chain Monte-Carlo (MCMC) iterations proceeded by an additional 450,000 steps, replicated five times to verify consistency across runs.We also used NEWHYBRIDS vers.1.1 Beta 3 (Anderson and Thompson ) as an alternative procedure to identify parental species and hybrids. NEWHYBRIDS is also a model-based approach, but uses the multilocus genotypes to assign individual fish to one of six genotypic classes: parental Arctic char, parental Dolly Varden, F1 hybrids, F2 hybrids, backcrosses to Arctic char or backcrosses to Dolly Varden. An estimate of the posterior probability of belonging to each of the six genotype classes was obtained for each individual fish and assigned to that class for which this probability was highest. The same allopatric populations used in the STRUCTURE analysis were also run in NEWHYBRIDS among the samples as nonadmixed reference populations for a minimum of 200,000 MCMC steps under the uniform priors option. […]

Pipeline specifications

Software tools Arlequin, Genepop, DISTRUCT, Structure Harvester, hybridlab, NewHybrids
Application Population genetic analysis
Organisms Salvelinus alpinus, Danio rerio, Hemisus marmoratus