Going against the flow: barriers to gene flow impact patterns of connectivity in cryptic coral reef gobies throughout the western Atlantic

Aim: Complex oceanographic features have historically caused difficulty in understanding gene flow in marine taxa. Here, we evaluate the impact of potential phylogeographic barriers to gene flow and assess demography and evolutionary history of a coral reef goby species complex. Specifically, we test how the Amazon River outflow and ocean currents impact gene flow. Location: Western Atlantic. Taxon: The bridled goby (Coryphopterus glaucofraenum) and sand-canyon goby (C. venezuelae) species complex. Methods: We used mitochondrial DNA and 2401 genomic SNPs to investigate evolutionary history and test hypotheses of how major barriers impact species-level differentiation. We used clustering algorithms and pairwise FST to assess population differentiation caused by minor barriers within and among regions. Finally, we tested alternate hypotheses of demographic history via coalescent simulations to determine the most plausible spread across the Western Atlantic. Results: We found two unique clades of C. glaucofraenum along the Brazilian coast and Atol das Rocas (AR) that are more closely related to C. venzuelae. Further genetic structure within the Caribbean and separately along the Brazilian coast led to at least two distinct populations in each location. Coalescent simulations indicated that an ancestral population of C. venezuelae split from C. glaucofraenum in the Caribbean, dispersed to Brazil, then spread to AR. Main Conclusions: Species-level genetic differentiation has resulted from the Amazon River outflow and isolation of AR. Population differentiation within the Caribbean matched previous studies indicating an east-west pattern of divergence. Brazilian population differentiation was impacted by the cold-water upwelling at Cabo Frio. Overall, this research highlights how barriers to gene flow impact speciation and genetic structure within western Atlantic gobies and provides insight into the role oceanographic features have in the speciation process of fishes.The data file is provided in STRUCTURE format where missing values are denoted as "-9" and SNPs are defined as G = 1, C = 2, A = 3, and T = 4. Individuals have two fragments, comprising one diploid individual. There is a metadata file which provides location information for each individual. Funding provided by: University of Central FloridaCrossref Funder Registry ID: http://dx.doi.org/10.13039/100007900Award Number: Funding provided by: Brazilian Marine Biodiversity Network – SISBIOTA-Mar*Crossref Funder Registry ID: Award Number: Funding provided by: Conselho Nacional de Desenvolvimento Científico e TecnológicoCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100003593Award Number: 563276/2010-0Funding provided by: Fundação de Amparo à Pesquisa e Inovação do Estado de Santa CatarinaCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100005667Award Number: 6308/2011-8Funding provided by: Brazilian Marine Biodiversity Network – SISBIOTA-MarCrossref Funder Registry ID:A reduced-sample SNP dataset was generated using 103 total individuals from 16 populations across the range of both C. glaucofraenum and C. venezuelae including three individuals of C. tortugae as an outgroup. Genomic DNA was converted into nextRAD genotyping-by-sequencing libraries (SNPsaurus, LLC) as in Russello, Waterhouse, Etter, & Johnson (2015). The nextRAD libraries were sequenced on an Illumina HiSeq 4000 with one lane of single-end 150 bp reads (University of Oregon). After de novo assembly and initial filtering, there were 9,003 SNPs. Following additional locus filtration for 10x coverage, 20% missing data and HWE, samples were thinned to include only one SNP per fragment resulting in a final dataset of 2,401 SNPs. Detailed information on the standard SNPsaurus SNP generation and filtering can be found in Supplemental Document 1. To identify loci under selection we used the program Bayescan v.2.1 (Foll & Gaggiotti, 2008) using default parameters for a neutral model: 100,000 iterations and 10 prior odds. The maximum allowable false discovery rate (FDR) was set to 0.05. We ran Bayescan within each putative species (i.e. Brazilian C. glaucofraenum (with southern Brazil vs. central and northern Brazil); C. venezualae; and Caribbean C. glaucofraenum; see Results). All outlier loci were removed from downstream analyses