Coselected genes determine adaptive variation in herbivore resistance throughout the native range of Arabidopsis thaliana

Abstract

The "mustard oil bomb" is a major defense mechanism in the Brassicaceae, which includes crops such as canola and the model plant Arabidopsis thaliana. These plants produce and store blends of amino acid-derived secondary metabolites called glucosinolates. Upon tissue rupture by natural enemies, the myrosinase enzyme hydrolyses glucosinolates, releasing defense molecules. Brassicaceae display extensive variation in the mixture of glucosinolates that they produce. To investigate the genetics underlying natural variation in glucosinolate profiles, we conducted a large genome-wide association study of 22 methionine-derived glucosinolates using A. thaliana accessions from across Europe. We found that 36% of among accession variation in overall glucosinolate profile was explained by genetic differentiation at only three known loci from the glucosinolate pathway. Glucosinolate-related SNPs were up to 490-fold enriched in the extreme tail of the genome-wide [Formula: see text] scan, indicating strong selection on loci controlling this pathway. Glucosinolate profiles displayed a striking longitudinal gradient with alkenyl and hydroxyalkenyl glucosinolates enriched in the West. We detected a significant contribution of glucosinolate loci toward general herbivore resistance and lifetime fitness in common garden experiments conducted in France, where accessions are enriched in hydroxyalkenyls. In addition to demonstrating the adaptive value of glucosinolate profile variation, we also detected long-distance linkage disequilibrium at two underlying loci, GS-OH and GS-ELONG. Locally cooccurring alleles at these loci display epistatic effects on herbivore resistance and fitness in ecologically realistic conditions. Together, our results suggest that natural selection has favored a locally adaptive configuration of physically unlinked loci in Western Europe.

Keywords: Arabidopsis thaliana; adaptation; genome-wide association mapping; glucosinolates; linkage disequilibrium.

Fig. 1.

Map of methionine-derived GSL variation in Europe. The x and y axes correspond to longitude and latitude, respectively. Dots indicate collection sites, and the color reflects the score of each accession along the first (top map) and second (bottom map) principal component describing GSL profile variation. The gradient from blue to yellow represents an increase in alkenyl and hydroxyalkenyl GSLs (Fig. S2).

Fig. 2.

Confirmed glucosinolate genes, GWA mapping, and $F_{ST}$ scan. (A) Positions of confirmed genes from the methionine-derived GSL biosynthesis pathway along the five chromosomes of A. thaliana. (B) Manhattan plot of mapping results using EMMAX pooled from all 22 molecules analyzed. SNPs with significant association scores are marked in gold. (C) Wright’s fixation index $\left(F_{ST}\right)$ scan along the Arabidopsis genome, based on 1,080 European accessions, divided into nine populations (15). The blue dotted line represents the 99.5% quantile of the $F_{ST}$ distribution. SNPs significantly associated with GSL variation are marked in gold.

Fig. 3.

LD among SNPs located within 20 kb of the three regions associated with GSL variation, MAM, AOP, and GS-OH. (A) Heat map of LD between pairs of SNPs located near the loci associated with GSL variation. GS-OH, AOP, and MAM are located on chromosomes 2, 4, and 5, respectively. Note the high values of LD between SNPs from chromosomes 2 and 5 at the upper left of the heat map. (B) LD between the associated SNPs in the GS-OH (chromosome 2) and MAM (chromosome 5) regions (in red) compared with a null distribution of $r^2$ for 1,000,000 random pairs of SNPs located on different chromosomes (in blue). The vertical dashed lines mark the 0.95, 0.99, 0.999, and 0.9999 quantiles of the null distribution.