The relative influence of geography, currents and environment on gene flow within sessile marine species remains an open question. Detecting subtle genetic differentiation at small scales is challenging in benthic populations due to large effective population sizes, general lack of resolution in genetic markers, and because barriers to dispersal often remain elusive. Marine lakes can circumvent confounding factors by providing discrete and replicated ecosystems. Using high-resolution double digest restriction-site associated DNA sequencing (4,826 Single Nucleotide Polymorphisms, SNPs), we genotyped populations of the sponge Suberites diversicolor (n=125) to test the relative importance of spatial scales (1-1,400km), local environmental regimes, and permeability of seascape barriers in shaping population genomic structure. With the SNP dataset we show strong intra-lineage population structure, even at scales <10km (average F_ST = 0.63), that was not detected previously using single markers. Most variation was explained by differentiation between populations (AMOVA: 48.8%) with signatures of population size declines and bottlenecks per lake. Though the populations were strongly structured, we did not detect significant effects of geographic distance, local environments, or degree of connection to the sea on population structure, suggesting mechanisms such as founder events with subsequent priority effects may be at play. We show that the inclusion of morphologically cryptic lineages that can be detected with the COI marker can reduce the obtained SNP set by almost 90%. Future work on sponge genomics should confirm that only one lineage is included. Our results call for a reassessment of poorly dispersing benthic organisms that were previously assumed to be highly connected based on low-resolution markers.

Tissue samples from the sponge Suberites diversicolor were sampled from two lagoon locations and nine marine lakes in the Indo-Pacific (Australia and Indonesia). DNA was extracted using the Blood & Tissue kit from Qiagen. In total 125 individuals were successfully sequenced using Illumina HiSeq 2500, employing a RAD-seq strategy. Custom scripts (provided) and open source programs (see methodology of paper) were employed to create a de-novo reference and align reads.