Marine and freshwater sticklebacks represent genetically, morphologically, and behaviorally distinct ecotypes that have evolved repeatedly throughout the Northern hemisphere [7
]. To identify potential adaptive loci that underlie repeated ecological differentiation, we searched for SNPs that consistently distinguish marine and freshwater ecotypes, using the method of Coop et al.
] to correct for overall structure between populations (, Table S3
). Five of the top six scoring SNPs fell near or within genomic regions previously associated with marine and freshwater ecotypes [4
], including two SNPs (chrIV:12,811,933 and chrIV:12,815,024, candidate class) near the ectodysplasin (EDA
) region underlying variation in armor plate phenotypes, and three SNPs (chrI:21,663,978, chrI:21,689,292, and 21,487034 candidate class) near the Na/K ATPase
) gene for salinity tolerance [18
] that shows strong differentiation along a marine–freshwater salinity gradient in Scotland [16
]. Although these genes have already been linked to morphological or physiological differences between particular marine and freshwater populations, allele frequencies have not previously been characterized in multiple individuals from a global set of populations. The repeated differentiation of these regions in many marine and freshwater sticklebacks provides strong additional support for the adaptive significance of these loci even after correcting for potential non-independence of populations–a confounding factor not explicitly adjusted for in previous studies [4
Bayesian scan for genomic regions consistently differentiated between marine and freshwater environments
Our analysis also identified a new marker associated with marine–freshwater differentiation on chromosome 2. This outlier SNP (chrII:418,094, candidate class) is located near multiple, potentially duplicated, genes belonging to the mucin gene family, Table S3
]). Mucus secretions in fish play important roles in osmoregulation, locomotion, and protection against pathogens [21
]. Genetic differences in epithelial barrier functions seem likely in marine and freshwater fish, and we propose that variation in the mucin cluster contributes to repeated transformation of marine sticklebacks to freshwater forms living in low ionic-strength environments.
To determine if we detect similar or different adaptive loci in Pacific and Atlantic sticklebacks, we also performed analyses separately in fish from each ocean basin. SNPs in the EDA
region showed high Bayes factor scores (BF >3.0) in sticklebacks from both basins (). In Atlantic sticklebacks the mucin gene region SNP scored even higher than EDA
loci, but was only weakly differentiated in Pacific sticklebacks. In contrast, four SNPs located on chromosomes IV, IX, XI and XVII had BF >2.5 in Pacific, but not Atlantic populations. These SNPs are located near genes that influence iron metabolism in humans and fish (chrIV:12,022,250, general class, , ABCB7
]), alter blood pressure and protect the heart, vasculature, and lungs from oxidative stress (chrIX:8,719,760, candidate class, , SOD3 gene [24
]), control ion gradients involved in sperm motility, resorption of bone, and digestion of microbes by phagocytes (chrXI:5,708,414, candidate class, , ATP6V0A1
]), and influence immune functions, pathogen clearing, and expression levels of a cell surface mucin gene required for epithelial barrier function and protection against infections (chrXVII:9,697,366, general class, , PRKCD
]). Given the low density of the genotyping array, the actual targets of selection might be other linked genes (see Table S3
). However we find it interesting that SNPs linked to mucin functions were recovered in both basins, highlighting the likely importance of epithelial barrier changes during recurrent evolution of marine and freshwater sticklebacks.
Recurrent genetic differentiation suggests that marine–freshwater stickleback adaptation proceeds, in part, by large shifts in allele frequency at some loci shared across multiple populations (). Many additional adaptive variants may exist that are specific to individual populations, and these would not be detected by the current method. Given the average spacing between markers in our study (~one marker per 400kb), we may have missed other strongly selected loci in the genome, even if they are repeatedly used in different populations. Previous studies show that linkage disequilibrium extends approximately 20–40 kb around major genes controlling armor and pelvic traits in natural populations [11
]. Selective sweeps that are tightly linked to the genotyped SNPs, and sweep events that are young, strong, or fall within regions of suppressed recombination are therefore the most likely to be detected in genome scans for adaptive loci using this SNP genotyping platform. Our results do not preclude the possibility that marine–freshwater adaptation might also involve other loci with smaller shifts in allele frequency (we note that a number of other markers show high or moderately high Bayes factors; , Table S3
). Nevertheless, these findings highlight how repeated evolution can be used to filter genetic drift from the signatures of adaptive loci, and show the potential of this approach for future studies using even higher densities of markers or whole genome sequencing.