Convergent evolution of SWS2 opsin facilitates adaptive radiation of threespine stickleback into different light environments

Abstract

Repeated adaptation to a new environment often leads to convergent phenotypic changes whose underlying genetic mechanisms are rarely known. Here, we study adaptation of color vision in threespine stickleback during the repeated postglacial colonization of clearwater and blackwater lakes in the Haida Gwaii archipelago. We use whole genomes from 16 clearwater and 12 blackwater populations, and a selection experiment, in which stickleback were transplanted from a blackwater lake into an uninhabited clearwater pond and resampled after 19 y to test for selection on cone opsin genes. Patterns of haplotype homozygosity, genetic diversity, site frequency spectra, and allele-frequency change support a selective sweep centered on the adjacent blue- and red-light sensitive opsins SWS2 and LWS. The haplotype under selection carries seven amino acid changes in SWS2, including two changes known to cause a red-shift in light absorption, and is favored in blackwater lakes but disfavored in the clearwater habitat of the transplant population. Remarkably, the same red-shifting amino acid changes occurred after the duplication of SWS2 198 million years ago, in the ancestor of most spiny-rayed fish. Two distantly related fish species, bluefin killifish and black bream, express these old paralogs divergently in black- and clearwater habitats, while sticklebacks lost one paralog. Our study thus shows that convergent adaptation to the same environment can involve the same genetic changes on very different evolutionary time scales by reevolving lost mutations and reusing them repeatedly from standing genetic variation.

Fig 1. Scan for selective sweeps around cone opsins in Haida Gwaii threespine stickleback.

One of the strongest outlier regions for H12 and iHS in the genome of Haida Gwaii stickleback from 28 populations is centered on SWS2 and LWS, indicating a past selective sweep in this region. No such signature is found around the other two cone opsins. Grey dots show H12 (upper panel) and absolute iHS (lower panel) values for each SNP with a minor allele frequency (MAF) greater 5%, with top 0.1% outlier SNPs highlighted as red dots. The proportion of iHS-SNPs with a value greater than 2 per 10 kb window is shown as a blue line. The 99.9%-quantile boundaries are shown as red- and blue-dashed lines for SNPs and windows, respectively. Genomic coordinates in Mb based on an improved version of the stickleback reference genome [51] are given on the x-axis. The position of the four cone opsin genes is highlighted with black boxes and vertical dashed lines; grey boxes indicate other genes. Depicted values for H12, iHS, window-iHS, and 99.9%-quantile boundaries can be found in S1 Data.

Fig 2. Haplotype structure and extended haplotype homozygosity around SWS2 and LWS.

Left panels show phased haplotypes, right panels the decay of haplotype homozygosity around selected SNPs, each for the adaptive radiation (upper panels) and the selection experiment dataset (middle and lower panels). The selective sweep signature seen in Fig 1 is caused by a long run of reference alleles (blue, upper left panel). This led to an extended haplotype homozygosity (EHH) signal at SNPs in the top H12-window (dashed lines in top right panel): haplotype homozygosity decays slowly around the reference allele (blue) for these SNPs, but rapidly around their alternate alleles (red). Decay is similar for SWS2 key amino acid substitutions A109G and S97G (full lines). In the selection experiment (middle/lower panels), the same EHH decay is found for the reference haplotype, but the alternate haplotype shows reduced decay, in particular in the transplant population Roadside Pond. Rows in the left panel show haplotypes with imputed SNPs and monomorphic sites with missing data (white). Columns represent SNPs with color code relative to the threespine stickleback reference genome, a freshwater stickleback female [6]. Gene positions are indicated with boxes above the figure. The haplotype matrix and EHH values can be found in S1 Data.

Fig 3. Signature of selective sweep in blackwater-adapted stickleback and “reversed sweep” after transplant to clearwater habitat.

Significantly reduced nucleotide diversity (middle left) and Tajima’s D (T_D, bottom left) in the blackwater source population Mayer Lake compared to genome-wide expectations (right) indicate a selective sweep in this population, consistent with the signature seen across the adaptive radiation (cf. Figs 1 and 2). After transplant to a clearwater habitat, however, the alternate haplotype has increased in frequency (Fig 2), leading to one of the strongest differentiation signals in the genome (F_ST, top left panel) and a significantly positive Tajima’s D. Such a pattern is consistent with a transient phase of a reversed selective sweep. Dashed horizontal lines indicate the 0.1%- and 99.9%-quantiles for each statistic based on their genome-wide distribution in regions with similar recombination rates (right panels); boxes above the figure indicate the position of genes, with black boxes and vertical dashed lines highlighting the position of the two cone opsins SWS2 and LWS. Sliding-window F_ST, nucleotide diversity and Tajima’s D values and genome-wide distributions of each statistic can be found in S1 Data.

Fig 4. Allele-frequency change around SWS2 and LWS in the selection experiment.

Distribution of single SNP allele-frequency change in the selection experiment (Roadside Pond minus Mayer Lake) around the selective sweep region on chromosome XVII (a,b). Note the unexpectedly strong increase of low frequency alleles linked to blue-shifting SWS2 coding variation after the transplant to a clear water pond. Color codes in (a) and (b) show “starting” allele frequencies, i.e., minor allele frequencies (MAFs) in Mayer Lake, color codes in (c) indicates the genome-wide distribution of allele-frequency change for Mayer Lake MAF bins of 0.05 width. Symbol shapes and colors in (b) and (c) indicate associations with different genes and predicted effects. Black and grey boxes below (a) and above (b) are exons of genes surrounding the sweep region. Depicted allele frequencies and allele-frequency change quantiles can be found in S1 Data.

Fig 5. Cone opsin amino acid polymorphisms across the Haida Gwaii stickleback radiation and selection experiment.

Note the high frequency and nearly perfect linkage of SWS2 amino acid polymorphisms. Columns represent single individuals per population, rows represent amino acid polymorphisms. Color codes show the genotype probabilities and amino acid alleles relative to the threespine stickleback reference genome (S97C: reference = S, alternate = C; rr: homozygous for reference amino acid, aa: homozygous for alternate amino acid, ra: heterozygous genotype). Amino acid positions are relative to the bovine rhodopsin protein. Light transmission ratios at 400 nm (T400) for the different water bodies are given in percent in the lower panels, with the grey area representing blackwater lakes [16]. Genotype probabilities can be found in S1 Data, T400 values in Table 1.

Fig 6. Accelerated protein evolution, evolutionary origin of threespine stickleback SWS2, and association with blackwater habitation.

(a) SWS2 shows an unexpected high frequency of amino acid polymorphisms given the total number of mutations in SWS2. The distribution of mean pairwise relative protein sequence divergence (mean ratio of non-synonymous to synonymous substitutions [dN/dS]) is based on 17,846 functional protein coding genes in the stickleback genome. Each value is based on pairwise comparisons between all 56 haplotypes in the “adaptive radiation dataset” containing one individual per population from the Haida Gwaii adaptive radiation, two mainland freshwater, and a marine site (see materials and methods). (b) Rooted phylogenetic tree of SWS2 opsins in stickleback and related taxa, which have retained two ancient SWS2 paralogs. Both red-shifted (“sweep”) and blue-shifted threespine stickleback haplotypes are derived from the ancestral SWS2A paralog, and the haplotype under selection has rapidly accumulated amino acid changes (dN/dS = 1.64). Branches are color coded with maximum-likelihood dN/dS estimates, and node labels show Bayesian branch credibility. (c) Genetic variation at SWS2, LWS, and SWS1, each summarized in single multidimensional scaling (MDS) coordinates, is associated with light transmission at 400nm (T400), but only variation at the sweep haplotype carrying coding and noncoding SWS2 and noncoding LWS variation is associated with the colonization of blackwater habitats. The grey area indicates blackwater lakes, following [16]. (d) Gene conversion did not contribute to molecular convergence between stickleback haplotypes and SWS2 paralogs in other fish: synonymous divergence (dS) between the two threespine stickleback haplotypes is larger to the SWS2B paralog than to the SWS2A paralog, except for a region around bp 700–800, a region without intraspecific amino acid variation. Mean pairwise dN/dS values for all genes shown in (a), MDS values for the three opsins SWS1, SWS2, and LWS in (c), and dS values in (d) can be found in S1 Data.

Fig 7. Median-joining network of SWS2 coding sequence haplotypes.

Virtually all populations across the Haida Gwaii adaptive radiation recruited the same red- or blue-shifted SWS2 allele. Population codes are shown in Table 1, numbers in brackets indicate the number of haplotypes among all 58 sequenced individuals, and the color code indicates haplotypes predicted to be red- or blue-shifted or intermediate, based on the two SWS2 opsin key sites at position 97 and 109.

Fig 8. Convergent evolution of the blue-sensitive SWS2 opsin at the molecular, functional, and ecological level.

The duplication of SWS2 in the ancestor of most spiny-rayed fish 198 million years ago was followed by a red-shift in SWS2A and a blue-shift in SWS2B [22, 47], paralogs that are divergently expressed among bluefin killifish (L. goodei) living in blackwater and clearwater habitats [24]. Two key amino acid polymorphisms of the ancient paralogs causing shifts in their absorption spectra have reevolved within threespine stickleback and are now divergently selected between blackwater and clearwater habitats in Haida Gwaii—convergent evolution at the molecular, functional, and ecological level. Clearwater spectra (left photo) are blue-shifted with increasing depth, typical of marine habitats and oligotrophic clearwater lakes on Haida Gwaii [18]. Tannin-stained blackwater (right photo) absorbs almost all up- and sidewelling light, making it a nearly nocturnal habitat, except for red-shifted downwelling light visible in a small cone above the observer. Unedited photos taken with a GoPro Hero 4 (GoPro Inc.) pointed towards the zenith at approximately 20 m depth in clearwater (Palau) and at 4 m depth in blackwater (Drizzle Lake).