3.4 | Genetic load
At the interspecific level (S. catenatus and S. tergeminussamples combined), we identified 28,493 sites containing derived nonsynonymous substitutions from 3,800 BUSCOs. When we analyzed the degree of evolutionary conservation of these sites using PROVEAN, we characterized 91% (26,035) of these sites as benign mutations, 8% (2,189) as deleterious mutations, and only 1% (269) as loss of function mutations. Collectively these represent Species level mutations in our analyses.
Based on these variants, on average, the genomes of inbred S. catenatus individuals had lower absolute numbers and lower relative proportion of all classes of mutations compared to the genomes of outbred S. tergeminus individuals (P < 0.001 in all cases; Figures 6a-b). There were ~2-fold differences in absolute numbers of benign (S. tergeminus : mean = 7054, SE = 94; S. catenatus : mean = 4999, SE = 47) and loss of function (S. tergeminus : mean = 65, SE = 2; S. catenatus : mean = 24, SE < 1) mutations compared to a smaller, but still significant, difference in deleterious mutations (S. tergeminus : mean = 290, SE = 6; S. catenatus : mean = 253, SE = 3). Differences in the relative proportion of deleterious variants are highest for loss of funtion alleles (S. tergeminus : mean = 0.147, SE = 0.003; S. catenatus : mean = 0.047, SE = 0.001), intermediate for benign mutations (S. tergeminus : mean = 0.154, SE = 0.001;S. catenatus : mean = 0.102, SE = 0.001), and lowest for deleterious mutations (S. tergeminus : mean = 0.071, SE = 0.001;S. catenatus : mean = 0.061, SE = 0.001) (Figure 6b).
Under the assumption that most deleterious mutations that arise in a population are recessive (Agrawal & Whitlock, 2011) the genotypic configuration of mutations can have a substantial impact on their realized negative impacts on individual fitness (Mathur & DeWoody, 2021). In this respect, S. catenatus has a significantly greater proportion of homozygotes relative to S. tergeminus , demonstrating greater levels of realized genetic load on a per-individual basis for two of three classes of mutations (Figure 6c). The proportion of deleterious homozygous genotypes was > 2-fold higher in S. catenatus compared to S. tergeminus(S. tergeminus : mean = 0.017, SE = 0.001; S. catenatus : mean = 0.037, SE = 0.001), while benign homozygous genotypes were also found at a significantly greater frequency in S. catenatus(S. tergeminus : mean = 0.060, SE = 0.001; S. catenatus : mean = 0.074, SE = 0.001) (Figure 6c). The exception is loss of function homozygous genotypes, which are significantly less often found inS. catenatus (S. tergeminus : mean = 0.043, SE = 0.003; S. catenatus : mean = 0.025, SE = 0.001; Figure 6c). Our interpretation is that this may represent extreme purging of highly deleterious mutations in small S. catenatus populations, as has been found in extremely small and repeatedly bottlenecked populations of Alpine ibex (Grossen et al., 2020).
Finally, although S. tergeminus had overall more deleterious mutations than S. catenatus , the fitness related impacts of mutations may be reduced because a greater proportion of all classes of deleterious mutations are found in heterozygous genotypes in S. tergeminus individuals (Figure 6d). This difference is highly significant for all three mutation classes (P < 0.001). This result supports the interpretation that the greater number of derived deleterious mutations in S. tergeminus may have reduced phenotypic expression because they are more often found in heterozygotes and less often found in homozygotes.
Mean PROVEAN scores of alleles found in both homozygote and heterozygote genotypes are also significantly lower (more deleterious) in S. catenatus compared to S. tergeminus (Figure 6e). These results support the interpretation that high levels of drift present in smallS. catenatus populations have allowed the persistence of mutations with greater potential negative impacts regardless of genotype class.
In contrast, analysis that use Population level mutations that capture population level processes show the opposite pattern in the form of higher levels of deleterious mutations with greater impact in S. catenatus (Fig. 7). For example, when data is pooled across all individuals within each species, the proportion of all variants made up by all classes of Population specific deleterious mutations is significantly higher in S. catenatus compared to S. tergeminus (Fig. 7a) as is the frequency of homozygous genotypes (Fig. 7b). Patterns of variation in heterozygotes are less clear: there is the expected increases in heterozygote frequency of beneficial mutations but no significant different in heterozygote frequency between species for other two more negative classes of mutations possibly due to the high variation present among individuals (Fig. 7 c). Nonetheless, the overall pattern is that, as a class, individual deleterious mutations that escape the “drift sieve” at the species level are more abundant within individual populations and potentially have more impact on individual fitness.
Comparisons based on Population level mutations across individualS. catenatus populations also show substantial and often significant differences in genome-wide proportion of deleterious mutations, and homozygote and heterozygote frequency across all mutation classes (Figure 8). If we use the proportion of each class of deleterious mutations as a measure of load (Figure 8a-c) then there is > 2-fold variation across populations for each class of deleterious mutation (benign: 0.384 [KPWA] – 0.585 [MOSQ]; deleterious: 0.238 [SSSP] – 0.493 [MOSQ]; loss of function: (0.228 [KPWA] – 0.410 [MOSQ]). These analyses identify specific populations that have high levels of load and are potentially subject to significant negative genetic impacts (MOSQ, CEBO, and SPVA) but also populations with the lowest levels of load (KPWA).
Finally, to assess possible drivers of the high level of variation in load across populations, we estimated correlations between estimates of contemporary effective size (Sovic et al., 2019) and mean proportion of deleterious mutations and levels of inbreeding (estimated as mean FROH) and mean frequency of homozygote deleterious genotypes. There were negative and marginally significant correlations between effective size and each class of mutations, supporting the predicted impact of drift acting through differences in population size on overall levels of genetic load (benign: Spearman rank correlation (R ) = -0.714;P = 0.06; deleterious: R = -0.667; P = 0.08; loss of function: R = -0.381; P = 0.3). There were also the expected positive correlations between F ROH and frequency of homozygotes, but none approached significance (results not shown).