Genomic clusters represent different admixture histories
In accordance with previous results, Mentha is resolved as monophyletic in relation to Salvia in the plastome phylogeny (Figure S3a; Li et al., 2016). However, the taxa within Menthaare polyphyletic although M. longifolia and M. suaveolensare mostly separated into different lineages (Figure S3b). Specimens identified as the hybrid M. × rotundifolia are found intermingled with both M. longifolia and M. suaveolens(Figure S3b). In contrast, with the exception of a single specimen, all specimens identified as M. spicata are monophyletic and found together with a subset of the M. longifolia specimens (Figure S3b).
Genomic clustering was performed by means of PCA, and similar to the morphological analyses there is a separation of M. longifolia(long) and M. suaveolens (sua) on the first PC (11%; Figure 2a). Intermediate to these taxa there are three relatively distinct groups of specimens that are consistent with the hybrid M. ×rotundifolia although a few samples of M. longifolia are also found in these clusters (rot1-3; Figure 2a). The genetic distinction between M. longifolia and M. spicata is more elusive with specimens only somewhat separating along the second PC (4%; Figure 2a). Some specimens mostly morphologically identified asM. spicata (spi1) do however, form a separate cluster on the positive extremes of the second PC (spi1; Figure 2a).
Overall, the admixture analysis reflects the genomic clustering with the best-fit number of population groups (K ) being two with other smaller optima at four and six (Figures 2 and S4). The specimens separating on the first genomic PC form five groups with distinct admixture profiles (Figure 2b). Based on the morphology of these specimens the three admixed groups genomically resemble F1 crosses between M. longifolia and M. suaveolens (M. ×rotundifolia ) as well as backcrosses to either parent (Figure 2b). Most specimens morphologically recognized as M. spicata are also admixed (Figures 2b). However, when the number of populations (K ) is increased it is clear that the admixture histories ofM. × rotundifolia and M. spicata are different (Figure 2b and S4b). When three respective four populations are considered, the two M. spicata clusters (spi1 and spi2; Figure 2a) forms separate populations but a few specimens morphologically identified as M. spicata still appear admixed (Figure 2b). The specimens with more central positions in the genomic PCA show admixture proportions consistent with mixing of all gene pools (Figures 2 and S4). However, our admixture analyses cannot decisively exclude that some patterns are caused by admixture with other species not included in our study.
Although, there is an overall agreement between the morphological and genomic analyses there are also a few cases of miss-matches. In particular a few specimens morphologically identified as M. ×rotundifolia are found intermingled with other genomic groups, as well as intermediate to the two groups of M. spicata (rot4; Figure 2a) and these specimens show admixture proportions consistent with their genomic cluster assignments (Figure 2b). In addition, a few specimens morphologically identified as M. spicata are genomically indistinguishable from M. longifolia and a few specimens morphologically identified as M. longifolia were found to be admixed (Figure 2b).