Discordance between mitochondrial and nuclear markers
The internal transcribed spacer (ITS1) of the ribosomal DNA has widely been used as a molecular marker for phylogenetic analysis at species level and for species identification of closely related molluscan taxa, including mussel species (Santaclara et al., 2006; Wood, Apte, MacAvoy, & Gardner, 2007). Nevertheless, unlike the deep divergence we uncovered in mitochondrial sequences, nuclear DNA sequences (ITS1) were not highly divergent (a maximum pairwise sequence divergence of 0.6%) and in the phylogenetic tree, all ITS1 sequences clustered together, irrespective of their geographic origin (Fig. 3A). The ITS1 network showed two genotype groups (a single step away from each other), each forming a star-like genealogy made of an admixture of northern and southern mitochondrial cox1 members (Fig. 3B). This type of mito-nuclear discordance is often found in biogeography studies and can be explained by many possible hypotheses (reviewed in Toews and Brelsford 2012). Further investigation will be needed to define the precise cause of mito-nuclear discordance in this species, but we discussed two hypotheses in the case of M. virgata : incomplete lineage sorting and selection on mitochondrial genes.
The first hypothesis invokes the stochasticity of coalescent processes at the two independent loci, resulting in divergent patterns. Incomplete lineage sorting is often mentioned in studies that uncover mito-nuclear discordance because nuclear markers have a larger effective population size (i.e. slower coalescence) than mitochondrial DNA (Moore, 1995; Pesole, Gissi, De Chirico, & Saccone, 1999). However, in the present study the difference between the cox1 and ITS1 results are so extreme (very deep divergence versus negligible divergence) that the incomplete lineage sorting hypothesis is inadequate given the inferred demographic history. Incomplete lineage sorting can only occur if two subpopulations maintain large population sizes before they merge into a common ancestral population; in contrast, our mtDNA cox1 and ITS1 genealogies both support very rapid and recent population growth following a population bottleneck. This is indicated by the genealogies with very short external branches only (a “star genealogy;” see Figs. 2A, 2B and Fig. 3), the significant negative values in Tajima’s Dand Fu’s FS test (Table 3), and the unimodal peak skewed over the Y-axis in the pairwise mismatch distribution analysis for the two mitochondrial lineages (Figs. 6A and 6B). Furthermore, the coalescence time between northern and southern cox1 lineages is estimated to be more than five times older than those of within-lineage coalescences (Fig. 5). Given a small ancestral population(s) and/or a very old splits of an ancestral population, incomplete lineage sorting at ITS1 or a long waiting time prior to the coalescence of the mtDNA lineages are unlikely to be the reason for mito-nuclear discordance.
Another hypothesis worth considering is that natural selection may have produced long-term, balanced mtDNA polymorphism. Although it is a long-held assumption that most of the sequence variation in mtDNA is selectively neutral (Ballard & Whitlock 2004; Meiklejohn, Montooth, & Rand, 2007), it is also well documented that balancing selection can occur under spatial heterogeneity (Hedrick, 2006; Scott et al., 2011). We note that the spatial distribution of the northern and southerncox1 lineages strongly correlates with surface seawater temperature (Fig. 1). Therefore, we suggest that these two divergentcox1 lineages are a proxy for two mitochondrial genotypes adapted to different surface water temperature zones, possibly due to divergence in the energy-metabolic functions in mitochondria. To test for divergent natural selection on the mt genome, we compared the whole mitochondrial genome sequences of two individuals representing the northern and southern mitochondrial lineages. Calculating the ratio of nonsynonymous (dN ) over synonymous (dS ) substitutions for the 12 mitochondrial protein-coding genes of the complete mitochondrial genomes revealed that none of the 12 protein-coding genes (includingcox1 , the genetic marker used in this study) showed a significant excess of dN/dS > 1 (Supplementary Table 2). These results do not provide supporting evidence for the selection on the mitochondrial genome; however, we cannot rule out the possibility that the two divergent lineages might still be maintained by balancing selection. The functional difference between two lineages might result from a small number of amino acid substitutions (not readily detectable in this case) or gene regulatory sequence differences. Additional mitochondrial genome sequencing that can cover the genetic diversity for the two mitochondrial lineages is further required to address whether deep divergence in mitochondrial genes is related to adaptation to different surface water temperature zones in the NWP.