2.3 Population genetic analysis
Individual fish with genotype call rates below 0.7 (n =10) were excluded, resulting in a total of 2,572 individuals used in further analyses. We quantified genetic diversity in several ways described here and these measures were used for the indicators (section 2.4). We assessed the most likely number of populations (K ) using structure (v.2.3.4; Pritchard et al., 2000; Falush et al., 2003). For this, we pooled the material from time points and localities within metapopulation, in order to investigate whether the same genetic populations appear in multiple lakes and/or are stable over time. In total, seven metapopulations and three separate lakes were analyzed (Table S1) like this with structure. We used the default model allowing population admixture and correlated allele frequencies, applying the alternative (population-specific) ancestry prior, with ALPHA=1/number of samples (i.e. the number of lake and time point combinations; Wang, 2019). No á priori information was used. The burn-in length was 250,000 and the number of Markov chains (MCMC) 500,000. Estimations of Q (assignment probability; the mean individual probability of belonging to a certain genetic cluster) and the most likely value of K (simulated K =1-15) was repeated over 20 runs, with the output analysed using kfinder (v.1.0; Wang, 2019) and structure harvester (v.0.6.94; Earl & vonHoldt, 2012). Mean individual Qover the 20 runs was derived from the clumpp software (v.1.1.2; Jakobsson & Rosenberg, 2007). The most likely number of K was based on the parsimony index (PI ) recommended by Wang (2019). Individuals were assigned to the cluster for which they had the highestQ .
We defined the clusters identified by structure as populations and all further analyses are based on these populations. We use “population” and “cluster” synonymously from here on, and if such populations occur within the same metapopulation we also use term “subpopulation” for such populations/clusters). We measured genetic diversity at two points in time for each population by estimating observed and expected heterozygosity (H O;H E), the average number of alleles per locus (N A) using genalex v.6.5 (Peakall & Smouse, 2006, 2012), allelic richness (A R) using fstat (v.2.9.4; Goudet, 2003), and the proportion of polymorphic loci (P L). Confidence intervals for diversity measures, as well as tests for normality of data were calculated in statistica (v.7.1; StatSoft, Inc., 2005). To test for changes in the genetic diversity measures over time we performed non-parametric Wilcoxon matched pairs test as well as Student’s t test for paired samples (heterozygosity, average number of alleles per locus, allelic richness), and χ 2 tests (proportion of polymorphic loci) using statistica and Microsoft excel. Statistical analyses investigating relationships between genetic diversity and the physical parameters of the localities were carried out using statistica.
We estimated effective population size (N e) for individual populations (cluster) with the temporal method using tempofs (sampling plan II; Jorde & Ryman, 2007), as well as with the linkage disequilibrium method (Hill, 1981; Waples, 2006; Waples & Do, 2010) as implemented in neestimator v.2.1 (Do et al., 2014). Confidence intervals for N e were obtained from the respective software. N e for metapopulations were estimated using the temporal method only.N e estimation in substructured (non-isolated) populations is complex and we follow suggestions from Ryman et al. (2014, 2019) when interpreting the estimations (below).
F ST (Weir & Cockerham, 1984) quantifying temporal genetic heterogeneity within populations and genetic heterogeneity among subpopulations within metapopulations were obtained using genepop (v.4.3; Raymond & Rousset, 1995; Rousset, 2008). chifish (v.5.0; Ryman, 2006) was used for significance testing of allele frequency change over time, while testing for spatial genetic differentiation was performed in genepop and statistica. The relationship among genetic clusters and metapopulations was illustrated with a phylogenetic tree from Poptree2 (Takezaki et al., 2010).