Association mapping of local climate-sensitive quantitative trait loci in Arabidopsis thaliana

Abstract

Flowering time (FT) is the developmental transition coupling an internal genetic program with external local and seasonal climate cues. The genetic loci sensitive to predictable environmental signals underlie local adaptation. We dissected natural variation in FT across a new global diversity set of 473 unique accessions, with >12,000 plants across two seasonal plantings in each of two simulated local climates, Spain and Sweden. Genome-wide association mapping was carried out with 213,497 SNPs. A total of 12 FT candidate quantitative trait loci (QTL) were fine-mapped in two independent studies, including 4 located within ±10 kb of previously cloned FT alleles and 8 novel loci. All QTL show sensitivity to planting season and/or simulated location in a multi-QTL mixed model. Alleles at four QTL were significantly correlated with latitude of origin, implying past selection for faster flowering in southern locations. Finally, maximum seed yield was observed at an optimal FT unique to each season and location, with four FT QTL directly controlling yield. Our results suggest that these major, environmentally sensitive FT QTL play an important role in spatial and temporal adaptation.

Fig. 1.

Histogram of FT in four environments. Shown is the average of four replicates for each accession. DTF, days to flower.

Fig. 2.

GWA mapping results by EMMA for FT in four environments (MAF ≥ 0.1). (A) Spring in Spain; (B) summer in Spain; (C) spring in Sweden; and (D) summer in Sweden. Four a priori loci (shown in red by the gray dashed lines) for FT showed significant (P ≤ 1 × 10⁻⁵) in at least one environment. Eight novel loci for seasonal FT (SFT) are shown in blue.

Fig. 3.

(A) Allele effects of the 12 QTL on FT in each of the four environments (the difference in days when replacing homozygous Col allele with homozygous non-Col allele). The tagged SNPs for these QTL were listed in Table S3. The graphs in the black boxes represent the effects from the mutants at the corresponding loci. (B) Effect sizes of SNPs, SNP × season, and SNP × location on FT across four environments. (C) Cluster of the 12 QTL by their sensitivity to season vs. location (the difference in effect sizes of SNP × season and SNP × location). Error bars are 95% confidence intervals. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

Fig. 4.

(A) Histogram of the P value for correlation between SNP allele and latitude. Histogram in blue represents the genome-wide distribution (172243 SNPs with MAF > 10%), and red represents the 12 candidate QTL SNPs. (B) Latitudinal distribution of the alleles at the four QTL (Col allele in red and non-Col allele in orange).

Fig. 5.

Correlations between FT and seed yield. Black lines show segmented regression with optimal breaking points. The Pearson's correlation r is shown above each regression with the corresponding significance levels: *P < 0.05, **P < 0.01, ***P < 0.001.