Genetic dissection of metabolite variation in Arabidopsis seeds: evidence for mQTL hotspots and a master regulatory locus of seed metabolism

Abstract

To gain insight into genetic factors controlling seed metabolic composition and its relationship to major seed properties, an Arabidopsis recombinant inbred line (RIL) population, derived from accessions Col-0 and C24, was studied using an MS-based metabolic profiling approach. Relative intensities of 311 polar primary metabolites were used to identify associated genomic loci and to elucidate their interactions by quantitative trait locus (QTL) mapping. A total of 786 metabolic QTLs (mQTLs) were unequally distributed across the genome, forming several hotspots. For the branched-chain amino acid leucine, mQTLs and candidate genes were elucidated in detail. Correlation studies displayed links between metabolite levels, seed protein content, and seed weight. Principal component analysis revealed a clustering of samples, with PC1 mapping to a region on the short arm of chromosome IV. The overlap of this region with mQTL hotspots indicates the presence of a potential master regulatory locus of seed metabolism. As a result of database queries, a series of candidate regulatory genes, including bZIP10, were identified within this region. Depending on the search conditions, metabolic pathway-derived candidate genes for 40-61% of tested mQTLs could be determined, providing an extensive basis for further identification and characterization of hitherto unknown genes causal for natural variation of Arabidopsis seed metabolism.

Keywords: Arabidopsis thaliana; gas chromatography–mass spectrometry; metabolic quantitative trait locus; primary metabolism; recombinant inbred line; seed biology..

Fig. 1.

Distribution of mQTLs for metabolites of known chemical structure. Chromosomal locations of significant mQTLs for the 58 metabolites of known chemical structure and the seed protein content are indicated by boxes representing the 1.5-LOD QTL support intervals. Vertical black lines within the boxes indicate the apices of the corresponding LOD curves. The mQTLs are color-coded according to their significance [threshold at alpha of 0.05 (yellow), 0.01 (orange), 0.001 (red)] derived from permutation results of the genome-wide maximum LOD scores. Vertical lines represent marker positions. For a subset, their approximate distance in centiMorgans is indicated. Asterisks at the bottom correspond to the position of identified mQTL hotspots.

Fig. 2.

mQTL analysis and candidate gene identification for leucine. (A) LOD profiles were plotted for all five Arabidopsis chromosomes. Gray lines represent LOD profiles calculated with the ‘cim’ function (composite interval mapping). Gray dots indicate selected cofactors. The horizontal dashed gray line corresponds to a CIM alpha threshold of 0.05, estimated by 10 000 permutations. The solid black lines indicate LOD profiles calculated with the ‘stepwiseqtl’ function using a multiple QTL model. The positions of the QTL apices in centiMorgans are given above the curves. (B) A simplified genetic map with known and putative genes involved in leucine biosynthesis and degradation. Purple horizontal lines indicate the locations of genes, directly or indirectly involved in leucine metabolism. Leucine mQTLs were identified on chromosomes II, III, IV, and V. Support intervals are shown as red vertical lines beside the chromosomes. Leucine-related genes, located within the confidence intervals of the mQTLs, are indicated. Identified candidate genes for chromosome II are AT2G23170 (GH3.3), AT2G26800 (HML1), and AT2G31810, for chromosome III AT3G48560 (AHAS) and AT3G49680 (BCAT3), for chromosome IV AT4G27260 (GH3.5), and for chromosome V AT5G65780 (BCAT5). (C) Boxplots of normalized and median divided leucine abundances in seeds of RILs. Samples were subdivided into four groups according to the allelic state at the epistatically interacting loci on chromosomes IV and V. Significant differences between the groups are indicated by upper case letters (ANOVA with post-hoc Tukey HSD, P_adj<0.001; number of individuals: n_C24/C24=113, n_Col-0/C24=20, n_C24/Col-0 =82, n_Col-0/Col-0=149). (D) Boxplots of normalized and median divided leucine abundances in seeds of parental and reciprocal F₁ hybrid plants derived from an independent experiment. Significant differences between the groups are indicated by upper case letters (ANOVA with post-hoc Tukey HSD, P_adj<0.05; number of individuals: n_C24=7, n_C24×Col-0=5, n_Col-0×C24=5, n_Col-0=5).

Fig. 3.

Principal component analysis of metabolite data. Score plot of the first two principal components PC1 and PC2 explaining 41% and 20% of variance of the data set, respectively. Samples were colored according to the genotype information on chromosome IV/marker: ‘MASC05042’ (12.90 cM). Black, red, and green circles correspond to Col-0, C24, and heterozygous alleles, respectively. Data were normalized, Pareto scaled, and mean centered prior to the calculation of the principal components.