Population Genetic Structure Analysis
Individuals named by species and location, we estimated the phylogenetic relationships of otherAquilegia species with the A. viridiflora complex to determine whether the A. viridiflora complex in our study shared an MRCA using IQ-TREE multicore version 1.6.12 (Nguyen, Schmidt, Von Haeseler, & Minh, 2015) and MEGA X (Kumar, Stecher, Li, Knyaz, & Tamura, 2018) with 1000 bootstrap replicates. Both the ML tree and NJ tree indicated that 20 populations of the A. viridiflora complex shared an MRCA. All trees were illustrated in iTOL (http:// itol.embl.de). Therefore, 672,439 high-quality SNPs in 20 populations of the A. viridiflora complex were used for downstream analysis after removing other Aquilegia species. To explore the patterns of genetic structure of the A. viridiflora complex, we used a phylogenetic network by the Neighbor-Net algorithm in the software Splits Tree (Huson & Bryant, 2006) with 1000 bootstrap replicates. ADMIXTURE v.1.3.0 (Alexander, Novembre, & Lange, 2009) was applied to investigate the maximum likelihood of the ancestry of all individuals with K values ranging from 2 to 10 with 10 replicates for each K value and examined the optimum K value according to the lowest value of the error rate. Principal component analysis (PCA) was performed using EIGENSOFT v.6.1.4 (Price et al., 2006) to infer population genetic structure. By combining the results of the phylogenetic relationship and genetic structure analysis to establish lineages for downstream analysis, we divided 20 populations into four lineages. Since the results of the population genetic structure showed a mixed genetic background in some individuals, the Python package HyDe was used to identify hybridization events at the individual level (Blischak, Chifman, Wolfe, & Kubatko, 2018). Among them, P2 was an individual who did not show a mixed genetic background, and P1 and P3 were individuals of other populations, respectively.