Rapid sex-specific evolution of age at maturity is shaped by genetic architecture in Atlantic salmon

Abstract

Understanding the mechanisms by which populations adapt to their environments is a fundamental aim in biology. However, it remains challenging to identify the genetic basis of traits, provide evidence of genetic changes and quantify phenotypic responses. Age at maturity in Atlantic salmon represents an ideal trait to study contemporary adaptive evolution as it has been associated with a single locus in the vgll3 region and has also strongly changed in recent decades. Here, we provide an empirical example of contemporary adaptive evolution of a large-effect locus driving contrasting sex-specific evolutionary responses at the phenotypic level. We identified an 18% decrease in the vgll3 allele associated with late maturity in a large and diverse salmon population over 36 years, induced by sex-specific selection during sea migration. Those genetic changes resulted in a significant evolutionary response only in males, due to sex-specific dominance patterns and vgll3 allelic effects. The vgll3 allelic and dominance effects differed greatly in a second population and were likely to generate different selection and evolutionary patterns. Our study highlights the importance of knowledge of genetic architecture to better understand fitness trait evolution and phenotypic diversity. It also emphasizes the potential role of adaptive evolution in the trend towards earlier maturation observed in numerous Atlantic salmon populations worldwide.

Figure 1. Change in mean age at maturity in the a, Tenojoki and b, Inarijoki populations.

Females are in red (N Tenojoki = 467, N Inarijoki = 261) and males in blue (N Tenojoki = 699, N Inarijoki = 570). Lines represent fitted values from the generalized additive model ± 1.96 SE, points are observed annual means.

Figure 2. Mean age at maturity as a function of vgll3 genotype in the a, Tenojoki and b, Inarijoki populations.

Females are in red (N Tenojoki = 522, N Inarijoki = 286) and males in blue (N Tenojoki = 804, N Inarijoki = 612). Means are calculated from multinomial models fitted values, averaged over years. Error bars represents 95% bootstrap confidence intervals based on 1000 replicates.

Figure 3. Temporal changes in vgll3 L allele frequency associated with late maturation in the a, Tenojoki and b, Inarijoki populations.

The lines represent fitted values from the quasibinomial model with ± 1.96 SE (N Tenojoki = 1166, N Inarijoki = 765). The vgll3 log-odd slope was estimated at -0.014 (CI95 = [-0.004, -0.024], P = 0.009) in Tenojoki and -0.009 (CI95 = [-0.023, 0.006], P = 0.26) in Inarijoki. Insets show the absolute estimated changes in allele frequencies of each SNP as a function of initial allele frequency in a, Tenojoki (144 loci) and b, Inarijoki (135 loci) over 36 and 37 years, respectively. The line represents the expected amount of drift at the 97.5 quantile. The vgll3 locus is indicated in red.

Figure 4. Model predicted mean vgll3 L allele frequency as a function of the sex and reproductive status in Tenojoki and Inarijoki.

Error bars indicate 95% confidence intervals. Adult allele frequencies are from years 2006 and 2007 for females (red circles) and males (blue circles), respectively, in the Tenojoki and Inarijoki populations (F tests, N Tenojoki = 1166 and N Inarijoki = 831). Juveniles allele frequencies (triangles, Likelihood-Ratio Tests) are from 2014-2015 in Tenojoki (143 individuals, 2-3 years old) and 2016 in Inarijoki (108 individuals, 1-3 years old).