Tests for nuclear introgression
Nuclear introgression from albacore to ABFT was tested by applying the statistical model implemented in TreeMix (Pickrell and Pritchard 2012) and ABBA/BABA analyses (Kulathinal et al. 2009, Green et al. 2010, Durand et al. 2011) on ‘nuclear mapped + other’ dataset with only one SNP per tag. The latter test was also performed excluding or including only those SNPs located within the genomic region found under high linkage disequilibrium (scaffolds BKCK01000075 (partially) and BKCK01000111). TreeMix was used to estimate historical relationships among populations and species by estimating the maximum likelihood tree for a set of populations allowing historical gene flow events. TreeMix was run allowing from 0 to 10 migration events, obtaining an increasing number of possible gene flow events and associated likelihood values. We followed the author’s recommendations (Pickrell and Pritchard 2012) to select the most probable number of migration events as that showing best associated likelihood values after stopping adding additional migration events so that the results remained interpretable. The ABBA/BABA test, which measures the excess of derived alleles shared between a candidate donor species and one of two tested groups (in this case, one ABFT group) compared with the other group taken as a reference (a different ABFT group), was performed on the allele frequencies of the derived allele in albacore and ABFT locations, based on the ancestral state defined by the Southern bluefin tuna taken as an outgroup. Patterson’s D statistic was calculated for all possible combinations of target and reference groups of ABFT, always considering albacore as the candidate donor species. Additionally, inter-species absolute divergence (dxy) between Mediterranean ABFT larvae and albacore individuals was estimated at each polymorphic position from the ‘nuclear mapped + other’ catalog. PCAs were performed using the adegenet R package (Jombart and Ahmed 2011) based on all filtered SNPs and only those SNPs from the region under high linkage disequilibrium from the ‘nuclear mapped + others’ catalog.
Results
Genetic differentiation between Atlantic bluefin tuna spawning components
We studied the population genetic structure and connectivity of Atlantic bluefin tuna using a genome-wide single nucleotide polymorphisms (SNP) dataset. Our study includes reference samples (i.e., larvae and YoY ABFT captured at or close to where they were hatched and adults caught on the spawning grounds during the spawning season) from the Gulf of Mexico and Mediterranean Sea, and from a more recently discovered spawning ground in the Slope Sea used by ABFT of unknown origin (Figure 1A, Table S1). Significant genetic differentiation between samples from these three spawning grounds was revealed by both an unsupervised clustering analysis of genetic ancestry (ADMIXTURE) (Figures 1B and S2A) and a principal component analysis (PCA) (Figure 1C and S2B), with significant pairwise genetic differentiation (FST) between reference larvae from different spawning grounds ranging from 0.0007 to 0.003 (Table S4). This contemporary genetic structure was associated with a mixture of two genetic ancestries (Figure S3), hereafter called GOM-like, predominant in the Gulf of Mexico (average GOM-like proportion across Gulf of Mexico individuals was 0.81 SD ± 0.22), and MED-like, predominant in the Mediterranean Sea (average MED-like proportion across Mediterranean individuals was 0.82 SD ± 0.11). Compared to the Gulf of Mexico and the Mediterranean Sea, the Slope Sea showed a large variance in individual ancestries ranging from GOM-like to MED-like, with a high proportion of admixed ancestries (average GOM-like proportion across Slope Sea individuals was 0.68 SD ± 0.22) (Figure 1D). This supports the Slope Sea as a mixed spawning ground where genetic admixture occurs. Additionally, whereas all Mediterranean individuals had a homogeneous MED-like genetic origin, ancestry profiles of Gulf of Mexico individuals were more variable, including 15 adults (out of 156) (Table S1) with a clear MED-like genetic profile (average GOM-like proportion across Gulf of Mexico individuals excluding 15 MED-like was 0.86 SD ± 0.14) (Figure 1D, C). Otolith microchemistry composition available for 6 of these 15 MED-like adults is compatible with Mediterranean Sea origin (Figure S4). In agreement with these results, admixture tests (F3 statistics) showed that the Slope Sea component is the result of admixture between the two other components (Figure 1E, S5 and Table S5). No admixture was found in the Mediterranean Sea, neither in larvae from the Gulf of Mexico. In contrast, admixture was detected in adult samples from the Gulf of Mexico (Figure 1E). Demographic history inferences (მaმi) (Figures 1F, S6 and Table S6) supported that the Slope Sea and the Gulf of Mexico spawning components share a recent common ancestry, and that there is strong contemporary migration from the Mediterranean Sea and the Gulf of Mexico towards the Slope Sea. Migration rates in all other directions are much weaker, the strongest being the migration from the Slope Sea back to the Gulf of Mexico, which is three times lower than in the opposite direction.
Overall, these results support a historical scenario of two ancestrally differentiated populations that interbred in the Slope Sea and to a lesser extent in the Gulf of Mexico. This later finding is supported by the presence of spawning capable MED-like individuals in the Gulf of Mexico, including one female that had ovulated less than 48 hours prior to capture (Table S1). A high level of gene flow between GOM-like and MED-like ancestries, as evidenced by the detection of frequent admixture, could imply either recent demographic changes (e.g. in migration rates or effective population sizes), or the existence of mechanisms that maintain genetic differentiation through time (e. g. cryptic barriers to gene flow or introgression from another source lineage).