Landscape characteristics can influence evolutionary processes. For instance, landscape fragmentation related to urbanization can restrict gene flow and decrease genetic diversity, potentially amplifying rates of local adaptation. For this study, our objectives were to compare genome-wide genetic diversity among urban and rural wood frog (Lithobates sylvaticus) populations and evaluate genetic selection associated with urban landscape features. We used a replicated sampling design with four urban-rural site pairs and 12 individuals per site to guard against false positive signals associated with other aspects of geographic variation. Replicates were selected with the goal of minimizing geographic distance within replicates and maximizing it between replicates. Genomic data were collected using double-digest RAD sequencing and were used to identify loci with consistently elevated genetic differentiation across replicates. A subsequent genotype-environment association analysis was conducted using logistic regression to identify loci with genotype frequencies that varied depending on their a priori urban/rural characterization. We then used a discriminant analysis of principal components (DAPC) to determine whether loci with evidence of selection could correctly assign individuals back to their respective rural or urban environment types. Finally, loci with signals of selection were aligned to the recently published American bullfrog (L. catesbeianus) genome to determine their functions.
Results/Conclusions
Our analysis evaluated 8344 loci following the removal of those that were non-variable, sequenced in fewer than 75% of samples per population, did not occur in every population, or had read depths less than 10x. We found no evidence of reduced allelic diversity or heterozygosity between urban and rural sites. However, we identified several loci with evidence of directional selection potentially associated with urbanization. This included 37 loci with elevated genetic differentiation values (FST) across all replicates with most showing a consistent shift to a particular genotype depending on the environment type. Logistic regressions independently indicated significant relationships between genotype frequencies and environment type for 12 of the 37 loci. Using the 37 outlier loci, DAPC could correctly discern whether individuals originated from a rural or urban site in over 90% of cases. Eighteen of the 37 outlier loci mapped to the bullfrog genome; however, none were in known genes, suggesting these loci are linked to coding regions rather than located directly within them. Overall our results demonstrate the capability of urban landscapes to drive contemporary evolution and represent a rare, although often advocated, attempt at conducting replicated landscape genomics research.