Measuring the effects of selection on the genome imposed by human-altered environment is currently a major goal in ecological genomics. Given the polygenic basis of most phenotypic traits, quantitative genetic theory predicts that selection is expected to cause subtle allelic changes among covarying loci rather than pronounced changes at few loci of large effects. The goal of this study was to test for the occurrence of polygenic selection in both North Atlantic eels (European Eel, Anguilla anguilla and American Eel, A. rostrata), using a method that searches for covariation among loci that would discriminate eels from 'control' vs. 'polluted' environments and be associated with specific contaminants acting as putative selective agents. RAD-seq libraries resulted in 23 659 and 14 755 filtered loci for the European and American Eels, respectively. A total of 142 and 141 covarying markers discriminating European and American Eels from 'control' vs. 'polluted' sampling localities were obtained using the Random Forest algorithm. Distance-based redundancy analyses (db-RDAs) were used to assess the relationships between these covarying markers and concentration of 34 contaminants measured for each individual eel. PCB153, 4'4'DDE and selenium were associated with covarying markers for both species, thus pointing to these contaminants as major selective agents in contaminated sites. Gene enrichment analyses suggested that sterol regulation plays an important role in the differential survival of eels in 'polluted' environment. This study illustrates the power of combining methods for detecting signals of polygenic selection and for associating variation of markers with putative selective agents in studies aiming at documenting the dynamics of selection at the genomic level and particularly so in human-altered environments.
Keywords: RAD sequencing; Random Forest algorithm; distance-based redundancy analysis; landscape genomics; polygenic selection.
© 2015 John Wiley & Sons Ltd.