3.8 | Population structure of B. schroederipopulations
We carried out whole-genome resequencing of 266 samples, including 240 from captive pandas of the Sichuan subspecies and 26 samples from individuals of the wild Qinling panda subspecies (Fig. 6a). The average sequencing coverage and sequencing depth reached 97.91% and 41-fold, respectively (Table S1 and S13). A total of 6.32 million SNPs were obtained after filtering (see methods). Principal components analysis (PCA) supported the clear separation between B. schroederi from the captive Sichuan and the wild Qinling panda subspecies (Fig. 6b), with PC1 separating the Qinling and Sichuan populations and PC2 separating the Sichuan population into two clusters (P <0.05). We constructed a phylogenetic tree using the maximum likelihood (ML) method, which showed two distinct clusters in the whole population, with all Qinling individuals forming a single cluster and all Sichuan individuals forming another single cluster (Fig. 6c). In addition, the results of structure analyses also indicated that there were almost no shared ancestral components between the Qinling and Sichuan populations, supporting the results from the phylogenetic tree and the PCA (Fig. 6d). Although the two populations showed extremely similar genetic diversity (Fig. S14 and Fig. S15), our results consistently supported two distinct groups corresponding to the Sichuan and Qinling populations.
3.9 | Recent positive selections in theB. schroederipopulations
We used integrated haplotype score (iHS) to detect genes under recent natural selection in the captive and wild populations. A total of 29,553 SNPs in captive and 18,953 SNPs in wild were identified within the top 1% iHS scores. By extending the 25 kb distance around the top 1% SNPs, filtering out SNPs in the non-gene regions, a total of 518 and 370 genes were located in the positively selected regions in captive and wild populations, respectively (Supplementary Data 3b and 3c). We further calculated the distribution of nucleotide diversity on the 21 superscaffolds (Fig S15), and found that the genetic diversity in some regions was significantly lower than in the flanking genome regions such as glutamate-gated chloride channel alpha (glc-1 , a receptor for anthelmintic ivermectin (Cook et al., 2006)), nose resistant to fluoxetine protein 6 (nrf-6 , a fluoxetine (Prozac) resistance gene (Choy, Kemner, & Thomas, 2006; Fares & Grant, 2002)), ABC transporter ced-7 (ced-7 , phagocyte corpse) (Wu & Horvitz, 1998) and β-1,4-N-acetylgalactosaminyltransferase bre-4 (bre-4 , resistance to pore-forming toxin (Griffitts et al., 2003)) genes (Fig. 7b). Interestingly, the four genes were only found under positive selection in the captive population, but not in the wild population. In addition, we also used the cross-population extended haplotype homozygosity (XP-EHH) method (Pardis C Sabeti et al., 2007) to screen for genes that might have been positively selected by different deworming selection pressures by comparing the captive and wild populations (Fig. S16, Fig S17, Supplementary Data 3d). Similarly, the genes encoding the multidrug resistance protein pgp-3 (Xu et al., 1998) which is related to ivermectin resistance, glc-1 , nrf-6 , cytochrome p450 (CYP ) family members and other drug resistance related genes, were also observed to be under strong positive selection in the captive population (Fig. 7c).