3.2 Phylogenetic trees and network
The phylogenetic trees reconstructed with the Angiosperms353 loci (hereafter, Angiosperms353 trees) show strong support, with bootstrap values above 90% for most inter-species nodes (Figs. 3A and Suppl. Fig. 2). The IQ-partition and RAxML trees are largely congruent, withA. cantabrica resolved as a sister to the /halleri clade (Fig. 3A). In the ASTRAL tree, a phylogenetic conflict was observed regarding the placement of A. rioxana A.Segura (Suppl. Fig. 2); here,A. rioxana was resolved as a sister to the /halleri and A. cantabrica clades rather than being embedded within the /halleri clade, as seen in other phylogenetic trees (Fig. 3). In the plastid tree,A. cantabrica is resolved as a sister to A. adfinis , with these two forming a clade that is sister to A. alpina (L.) Lam.. The /halleri clade is resolved as a sister to the clade formed byA. alpina , A. adfinis , and A. cantabrica (Fig. 3B).
Our ploidy estimation analysis predicts that A. rioxana , A. cantabrica and A. adfinis subsp. brigantiaca (Jord. & Fourr.) Kress are tetraploid, whereas all other species tested are estimated as diploid (Suppl. Table 5). The split network analysis reveals indicated gene flow and reticulate evolution in the formation of these species, with A. cantabrica emerging as a relatively independent evolutionary branch related to A. adfinis and the /halleri clade, yet showing more ancestral gene flow with the /halleri species (Fig. 3C).