3.5 | Trait-associated loci
Difference in phenotypic values between the inland and coastal sites and
dependence of phenotypic values on the Qd ancestry varied among
the eight traits (Figures S8, S9). Phenotypic values of each trait were
scaled in each site separately. Scaled phenotypic values of the 224
sampled trees in both sites were pooled and applied to mixed linear
models with effects of genetic structure and kinship relationship. Based
on PCA of SNP genotypes at ddRAD loci, genetic variation betweenQc and Qd mainly determined genetic structure in the first
PC (Figure S10), and some northern-edge Qd populations had unique
genetic properties in the second, third, and fourth PCs (Figure
S10b–d). Q-Q plots of association mapping of SNP genotypes and
admixture mapping of local ancestry showed consistent relationships
between expected and observed log-likelihood values at loci with a lower
range of –log10 p values, which represented the
majority of unassociated loci, indicating successful fitting (Figures
S11, S12).
We did not detect any loci significantly associated with the eight
traits (FDR > 0.05) but found some loci potentially
associated with one taxon-specific trait and three habitat-specific
traits (0.051 ≤ FDR ≤ 0.148; Figure 7, Table S2). Association mapping
detected three loci in chromosomes 2, 7, and 9 associated with lateral
vein interval (Figure 7c), and admixture mapping also supported one
locus in chromosome 2 of the three loci (Figure 7d). For loci associated
with stellate hair density, association mapping detected one locus in
chromosome 1 (Figure 7i), and admixture mapping detected loci in two
genomic regions in chromosomes 10 and 11 (Figure 7j). Admixture mapping
detected loci associated with LMA in chromosome 11 (Figure 7l), and
association mapping detected one locus associated with shoot diameter in
chromosome 1 (Figure 7m). Among them, a locus in chromosome 9
potentially associated with lateral vein interval showed a high value of
the Patterson’s D statistics (D = 0.51; Figure S7), and a
region of loci potentially associated with stellate hair density in
chromosome 11 showed high local ancestry from Qd (1.33 Qdalleles; Figure 6b).
From 31 genes around the loci potentially associated with the four
traits, we obtained 39 GO terms, in which most frequently observed terms
were DNA binding and protein binding (Table S2). Among the 31 genes, we
obtained 10 protein names (Table S2), some of which were related to
drought stress resistance (Lorenzo et al. 2002; Kushiro et al. 2004),
disease resistance (Van Damme et al. 2009; Mokryakova et al. 2014), and
senescence (Engqvist et al. 2011).