Multi-model seascape genomics identifies distinct environmental drivers of selection among sympatric marine species

Abstract

Background As global change and anthropogenic pressures continue to increase, conservation and management increasingly needs to consider species’ potential to adapt to novel environmental conditions. Therefore, it is imperative to characterise the main selective forces acting on ecosystems, and how these may influence the evolutionary potential of populations and species. Using a multi-model seascape genomics approach, we compare putative environmental drivers of selection in three sympatric southern African marine invertebrates with contrasting ecology and life histories: Cape urchin (Parechinus angulosus), Common shore crab (Cyclograpsus punctatus), and Granular limpet (Scutellastra granularis). Results Using pooled (Pool-seq), restriction-site associated DNA sequencing (RAD-seq), and seven outlier detection methods, we characterise genomic variation between populations along a strong biogeographical gradient. Of the three species, only S. granularis showed significant isolation-by-distance, and isolation-by-environment driven by sea surface temperatures (SST). In contrast, sea surface salinity (SSS) and range in air temperature correlated more strongly with genomic variation in C. punctatus and P. angulosus. Differences were also found in genomic structuring between the three species, with outlier loci contributing to two clusters in the East and West Coasts for S. granularis and P. angulosus, but not for C. punctatus. Conclusion The findings illustrate distinct evolutionary potential across species, suggesting that species-specific habitat requirements and responses to environmental stresses may be better predictors of evolutionary patterns than the strong environmental gradients within the region. We also found large discrepancies between outlier detection methodologies, and thus offer a novel multi-model approach to identifying the principal environmental selection forces acting on species. Overall, this work highlights how adding a comparative approach to seascape genomics (both with multiple models and species) can elucidate the intricate evolutionary responses of ecosystems to global change.

Distinct interspecific and intraspecific vulnerability of coastal species to global change

“…Allele frequency datasets were generated using a pooled restriction site‐associated sequencing (RAD‐seq) approach, specifically using ezRAD; ezRAD uniquely allows for a combination of high coverage at specific loci and low coverage across the entire genome (Toonen et al, 2013). To ensure accurate inferences, we included a large number of individuals per pool, and imposed stringent coverage and quality filtering criteria, as well as custom scripts to account for linkage disequilibrium (see Nielsen, Henriques, et al, 2020 for details). Inferences of population differentiation were shown to be robust to changes in bioinformatic filtering parameters such as minimum coverage and read count (Nielsen, Henriques, et al, 2020).…”

“…To ensure accurate inferences, we included a large number of individuals per pool, and imposed stringent coverage and quality filtering criteria, as well as custom scripts to account for linkage disequilibrium (see Nielsen, Henriques, et al, 2020 for details). Inferences of population differentiation were shown to be robust to changes in bioinformatic filtering parameters such as minimum coverage and read count (Nielsen, Henriques, et al, 2020).…”

“…All model types were chosen as they either use relatedness to account for neutral population structure or are based on null models of inferred demographic histories, both of which have been shown to lead to more robust outlier identification (Hoban et al, 2016). Moreover, our study system and species are well suited to avoid high false discovery rates in outlier dectection analyses, as there is evidence of high gene flow and demographic homogeneity (Nielsen, Henriques, et al, 2020), as well as the axes of historical expansion and contemporary environmental variation being aligned in space (Frichot et al, 2015; Mmonwa et al, 2015; Muller et al, 2012).…”

“…Eight environmental variables were selected for the GF models based on their importance for rocky shore ecology (Branch & Branch, 2018), marine species distributions (Bosch et al, 2018), and significance in shaping genomic variation within the study species (Nielsen, Henriques, et al, 2020). These variables were the means and ranges of sea surface temperature (SST), sea surface salinity (SSS), air temperature (T), and precipitation (P).…”

“…As only SNPs with a R 2 > 0 are included in the turnover functions and mapping analyses, the allele frequencies of the subset of SNPs retained in the GF models were visualized in PCAs created using vegan and ggplot2 (Wickham, 2016) R packages. This was done to assess how well the SNPs in the GF models reflect patterns seen in the entire SNP datasets, which were previously used to assess neutral and outlier population structure (Nielsen, Henriques, et al, 2020). To assess whether the different sets of environmental predictor variables lead to differences in the SDM and GF vulnerability outputs, additional GF models were run with the subset of uncorrelated environmental variables similarly used in the SDM models (referred to in the following section), as predictor variables.…”