Fig 1.
The Japan Sea stickleback is a separate species.
(A) Rooted nuclear consensus tree for Japan Sea, Pacific Ocean and Atlantic Ocean stickleback lineages from 10 kb non-overlapping sliding windows across the autosomes. Red trees indicate species clustering; blue trees indicate geographical clustering and green trees reflect ancestral polymorphism. NB: Only 1,000 subsampled species trees are shown here to aid illustration. (B) Mitogenome Bayesian consensus tree shows divergence between two mitochondrial clades–all Japanese sticklebacks (G. nipponicus and G. aculeatus) and G. aculeatus occurring in Europe and North America. (C) Present day distribution of G. aculeatus (blue) and G. nipponicus (red) around the Japanese archipelago. The two species overlap in Hokkaido, Northern Japan and samples for this study were collected in Bekanbeushi River in Akkeshi unless noted. (D) PSMC plot of 26 resequenced genomes shows a steady effective population size in the Pacific Ocean lineage (blue) but a bottleneck around 0.15–0.3 million years before present and a subsequent increase in the Japan Sea lineage (orange). The effective population size of the Atlantic Ocean lineage is shown in blue green.
Fig 2.
ABC analysis supports isolation with gene flow.
(A) A model of isolation with migration and a bottleneck in the Japan Sea lineage is best supported by ABC analysis using ~2,000 nuclear loci (see Table 1). Posterior probability densities for model parameters estimated using neural network analysis with a tolerance of 1% and 20 summary statistics. Parameters are: T = time of split, m12 = the proportion of the Japan Sea population that are migrants from the Pacific Ocean per generation, m21 = the proportion the Pacific Ocean population that are migrants from the Japan Sea per generation (note that m is the migration rate backward in time); TG = timing of bottleneck, NPO = Pacific Ocean effective population size, NJS = Japan Sea effective population size and NJSB = Japan Sea bottleneck effective population size. Posterior probability density curves for (B) Japan Sea and Pacific Ocean divergence time and timing of bottleneck in the Japan Sea lineage, (C) Japan Sea, Pacific Ocean and Japan Sea bottleneck effective population sizes, and (D) migration rates averaged across the genome, shown as m in Fig 2A. Figures on each panel are median parameter estimates.
Table 1.
Posterior probability values for models for final ABC model selection using neural network rejection.
All estimates produced using a tolerance of 1% and 20 summary statistics. Bold text indicates the model where posterior probability provides the highest support. Models are I = isolation, IM = isolation with migration, IAM = isolation and ancient migration, IRM = isolation and recent migration, IARM = isolation with ancient and recent migration.
Fig 3.
Genomic divergence is lower in sympatry than in allopatry between species.
Histograms of (A) relative (FST) and (B) absolute (dXY) differentiation measures for each of the species comparisons. (C) Mean genome-wide FST of the Japanese species pair compared with those of other stickleback systems taken from previously published studies [35–37,56].
Fig 4.
Genome-wide distribution of divergence and introgression.
Divergence was measured using FST and dXY, while introgression was measured using GMIN and fd. Data plotted here is from 50 kb non-overlapping genome windows. Blue and yellow lines indicates allopatric (Japan Sea vs Atlantic) and sympatric (Japan Sea vs Pacific Ocean) comparisons, respectively.
Table 2.
Genome-wide averages for measures of divergence and introgression.
FST, dXY, GMIN, and fd for all pairwise comparisons of Japan Sea (JS), Pacific Ocean (PO) and Atlantic Ocean sticklebacks (AT) are shown. Mean ± SD and lower and upper limits of the 95% confidence interval (in parenthesis) are shown. NA, not analysed.
Fig 5.
Fewer introgression valleys occur on the neo-X chromosome.
(A) A greater number of GMIN valleys occur in sympatry than in allopatry between species. (B) GMIN valleys and fd peaks also occur in regions of the genome with a higher recombination rate. Fewer valleys occur on the neo-X chromosome (chrIX; shown in pink) compared to autosomes (C), even when chromosome length is taken into consideration (D); N.B.–data for (C) and (D) were measured using females only. Green shows the ancestral sex chromosome (chrXIX).