Natural genetic variation in Arabidopsis thaliana defense metabolism genes modulates field fitness

Shown are the identities, functions and mutation types of nine genes representing seven loci important for aliphatic GSL synthesis and activation. These genes were chosen for mutant laboratory population of Arabidopsis thaliana accession Col-0 due to the fact that they represent the majority of aliphatic GSL variation observed in Arabidopsis. Each of these genes is naturally polymorphic among Arabidopsis accessions.

Shown are the genotypes in the mutant laboratory GSL population used in this study and the allelic state of each gene within each of them. Each gene in our population is naturally polymorphic among Arabidopsis accessions. See Table 1 for gene functions. For each gene listed, a ‘+’ indicates a functional allele and a ‘−’ indicates a non-functional allele. The loss-of-function and gain-of-function mutant lines shown in Table 1 were manually crossed to generate this population of genotypes, each of which vary from Col-0 at only these eight genes, including single, double, and triple mutants.

A linear mixed model was fitted to phenotypes measured on plants grown in the field. Variation was partitioned among the fixed effects, Genotype, Environment, and Treatment as well as a random factor, Plot, inside which Treatment and Environment were nested. Phenotypic data used in the model was collected on 17 genotypes from three environments and two treatment groups. df = degrees of freedom, SS = Type II Sums of Squares variation, MS = Mean Squared variation, F = F statistic (for fixed factors), χ2 = chi squared statistic (for random factors). p value = probability value from either an F test or a chi squared test, depeding on the source of variation. Non-significant p values (>0.05) are represented by a dash.

Glucosinolate analysis was conducted on the 17 genotypes within a Long-day growth chamber (16 hr light) set to match the median light regime for the three environmments. Principal component analysis was conducted on the mean glucosinolate accumulation for the aliphatic glucosinoles within the chamber environment. This creates a set of mathematical descriptors of the chemotype variation across the 17 genotypes. The first three eigenvectors were used to generate scores from the lsmeans of the glucosinolates across the 17 glucosinolate genotypes independently for the chamber and three different field environments values. These scores were then correlated to test if the GSL profiles were similar or not across the environments. The R of the correlation to the Chamber scores for the 17 genotypes for each of the three PCA vectors are shown in conjunction with the p value as determined by Pearson correlation. To the right of each PCA vector label is shown the fraction of total variance approximated by the given vector. In total, the four vectors describe >99% of the GSL profile variance.

Shown are Pearson's correlations for leaf damage between the different environments. R = correlation coefficients; p value = probability statistic. Non-significant p values are represented by a dash.

A linear mixed model was fitted to phenotypes measured on plants grown in the field. Variation was partitioned among the fixed effects, Genotype, Environment, and Treatment as well as a random factor, Plot, inside which Treatment and Environment were nested. Phenotypic data used in the model was collected on 17 genotypes from three environments and two treatment groups. df = degrees of freedom, SS = Type II Sums of Squares variation, MS = Mean Squared variation, F = F statistic. Non-significant p values (>0.05) are represented by a dash.

Shown are genetic correlations between absolute fitness and traits pleiotripically controlled by GSL genes. Pearson's correlation coefficients are on the top half of the tables and Spearman rank correlations are on the bottom. *p value < 0.05, **p value < 0.01, ***p value < 0.001.