Haplotype network variation across the genome
Despite clear genomic and morphological differences among A. marina subspecies, we found evidence of gene flow. Thus, the genomes of these groups are expected to be mosaic, with regions of high divergence interspersed among undifferentiated loci subject to exchange of genetic material among subspecies. To verify this, we inferred haplotype networks across the close to 100 loci we sequenced on the genome. Using an expectation-maximization method to infer the linkage of the SNPs detected from these loci, we split these regions into 454 linked segments (Supplementary Table 3). Segments with missing data and with length less than 100bp were discarded and 231 segments were retained for haplotype network reconstruction, with A. alba as the outgroup (Figure 5).
Among these segments, 134 (58.0%) were not genetically distinguishable among groups with only one or a few haplotypes identified and all haplotypes closely related to each other and shared among the three subspecies. The other 66 segments (28.6%) reliably distinguishedA. m. australasica from the other two subspecies. Among these 66 segments, the maBB population shared haplotypes with A. m. australasica rather than A. m. marina in seven loci. The third type of segments, which was 14 in total (6.1%), delimited A. m. marina from the other two subspecies. Five segments (2.2%) distinguishA. m. eucalyptifolia , but maBB shared haplotypes with A. m. eucalyptifolia in all cases. Most importantly, 11 segments (4.8%) allow clear delimitation of all three subspecies, with haplotypes split into three groups and each subspecies containing haplotypes from a single group. However, in eight of the 11 segments, maBB shared haplotypes with A. m. eucalyptifolia . Finally, one segment (0.4%) separated A. m. marina and A. m. australasica, butA. m. eucalyptifolia contained haplotypes from both subspecies.