Genomic variation within and among populations is shaped by the interplay between natural selection and the effects of genetic drift and gene flow. Adaptive divergence can be found in small scale natural systems even when population sizes are small and the potential for gene flow is high, suggesting that local environments exert selection pressures strong enough to counteract the opposing effects of drift and gene flow. Here, we investigated genomic differentiation in nine moor frog (Rana arvalis) populations in a small-scale network of local wetlands using 16 707 ddRAD-seq SNPs, relating levels of differentiation with local environments as well as with properties of the surrounding landscape. We characterized population structure and differentiation, and partitioned the effects of geographic distance, local larval environment and landscape features on total genomic variation. We also conducted gene-environment association studies using univariate and multivariate approaches. We found small-scale population structure corresponding to 6-8 clusters. Local larval environment was the most influential component explaining 2.3% of the total genetic variation followed by landscape features (1.8%) and geographic distance (0.8%), indicative of isolation-by-environment, -by-landscape and -by-distance, respectively. We identified 1000 potential candidate SNPs putatively under divergent selection mediated by the local larval environment. The candidate SNPs were involved in, among other biological functions, immune system function and development. Our results suggest that small scale environmental differences can exert selection pressures strong enough to counteract homogenizing effects of gene flow and drift in this small-scale system leading to observable population differentiation.

Rana arvalis eggs were collected from populations inhabiting ponds in a small scale network of interconnected wetlands and raised in a common garden untill metamorphosis. We also measured a number of environmental and landscape variables within and around each pond. DNA was extracted from individual metamorphs and sequenced using a ddRAD-seq approach and sequencing data were processed in STACKS v2.1. Filtered variant files were analyzed using various population genetics/landscape ecology packages in R.