Littorina saxatilis is becoming a model system for understanding the genomic basis of ecological speciation. The parallel formation of crab-adapted ecotypes that exhibit partial reproductive isolation from wave-adapted ecotypes has enabled genomic investigation of conspicuous shell traits. Recent genomic studies suggest that chromosomal rearrangements may enable ecotype divergence by reducing gene flow. However, the genomic architecture of traits that are divergent between ecotypes remains poorly understood. Here, we use 11,504 single nucleotide polymorphism (SNP) markers called using the recently-released L. saxatilis genome to genotype 462 crab ecotype, wave ecotype, and phenotypically-intermediate L. saxatilis individuals with scored phenotypes. We used redundancy analysis to study the genetic architecture of loci associated with shell shape, shape corrected for size, shell size, and shell ornamentation, and to compare levels of co-association among different traits. We discovered 341 SNPs associated with shell traits. Loci associated with trait divergence between ecotypes were often located inside putative chromosomal rearrangements recently characterized in Swedish L. saxatilis. In contrast, we found that shell shape corrected for size varied primarily by site rather than by ecotype and showed little association with these putative rearrangements. Together, these results reveal that genomic regions of elevated divergence with putative rearrangements are associated with divergence along steep environmental axes in L. saxatilis ecotypes, consistent with models of adaptation with gene flow, but these regions are distinct from genomic architecture associated with site-specific variation. Our findings here further support predictions from models indicating the importance of genomic regions of reduced recombination allowing co-association of loci during ecological speciation with ongoing gene flow.

SNP genotype data in PLINK format output from STACKS 2 SNP genotyping using reference genome-aligned double digest RADseq.