Metabolomics analysis and metabolite-agronomic trait associations using kernels of wheat (Triticum aestivum) recombinant inbred lines

Abstract

Plants produce numerous metabolites that are important for their development and growth. However, the genetic architecture of the wheat metabolome has not been well studied. Here, utilizing a high-density genetic map, we conducted a comprehensive metabolome study via widely targeted LC-MS/MS to analyze the wheat kernel metabolism. We further combined agronomic traits and dissected the genetic relationship between metabolites and agronomic traits. In total, 1260 metabolic features were detected. Using linkage analysis, 1005 metabolic quantitative trait loci (mQTLs) were found distributed unevenly across the genome. Twenty-four candidate genes were found to modulate the levels of different metabolites, of which two were functionally annotated by in vitro analysis to be involved in the synthesis and modification of flavonoids. Combining the correlation analysis of metabolite-agronomic traits with the co-localization of methylation quantitative trait locus (mQTL) and phenotypic QTL (pQTL), genetic relationships between the metabolites and agronomic traits were uncovered. For example, a candidate was identified using correlation and co-localization analysis that may manage auxin accumulation, thereby affecting number of grains per spike (NGPS). Furthermore, metabolomics data were used to predict the performance of wheat agronomic traits, with metabolites being found that provide strong predictive power for NGPS and plant height. This study used metabolomics and association analysis to better understand the genetic basis of the wheat metabolism which will ultimately assist in wheat breeding.

Keywords: Triticum aestivum L.; agronomic trait; mature seed; metabolic prediction; metabolic quantitative trait loci.

Figure 1

Metabolic profiling in wheat RIL population. (a) Number of detected metabolites and their classification. (b) Distribution of the values of coefficient of variation (CV) and broad‐sense heritability (H ²) of metabolic traits in the RIL population. H ² was estimated using one‐way ANOVA, taking into account the variations between the three biological replicates as phenotypic variance derived from environmental factors. (c) Pairwise Pearson’s correlations are shown in a heat map, whereas metabolites are sorted according to correlation‐based hierarchical cluster analysis. The level of correlation is indicated by red (positive correlation) and blue (negative correlation).

Figure 2

Chromosomal distribution of metabolic QTLs (mQTLs) identified. (a) Chromosomal distribution of all mQTLs (1005) and mQTL hotspots. The horizontal dashed line indicates the threshold for mQTL hotspots, represented by the maximum number of mQTLs expected to fall into any interval by chance alone with a genome‐wide P = 0.01. The interval size is 10 cM. (b) Distribution of mQTLs of 467 known metabolites on chromosomes. Each row represents the QTL mapping of single metabolic traits. Metabolites from different chemical groups are marked by distinct colours. The x‐axis indicates the genetic positions across the wheat genome. The heat map under the x‐axis illustrates the density of QTL across the genome. The window size is 10 cM.

Figure 3

Functional annotation of candidate gene TraesCS2B01G012000 (a) LOD curves of QTL mapping of the mr1092 (Apigenin 7‐O‐rutinoside) accumulation on chromosome 2B. (b) Gene model of TraesCS2B01G012000. The black box represents the coding sequence. (c) Candidate gene encoded proteins were transiently expressed in N. benthamiana followed by a StrepII purification. Samples (5 µl) were taken at different stages of the purification. Lanes S1 to S5 are total soluble proteins; proteins not bound; last wash fraction; elution fraction; and proteins left on the matrix after elution, respectively. The arrowhead indicates the purified protein. CBB, Coomassie Brilliant Blue stain; WB, western blot. (d) Enzymatic reaction by the purified proteins. The structures of the substrates and products (left) and the chromatograms of the standards and the biochemical reaction.

Figure 4

Association network visualization of co‐detected metabolite‐agronomic traits and dissection of a candidate gene associated with number of grains per spike (NGPS). (a) Association analysis of 467 annotated metabolites with 17 agronomic traits. Co‐detected metabolites and agronomic traits are represented as nodes, and their correlation coefficient values as edges. The absolute values of the Pearson’s correlation coefficient values above the threshold (P < 0.01) are shown. Different colours represent different classes of metabolites. Circles and green hexagons are represented as metabolites and agronomic traits, respectively, where the size of the shape represents the number of associations. The level of correlation is indicated as red (positive correlation) or blue (negative correlation). The intensity of the colour indicates the correlation, where a darker colour denotes a stronger correlation. The yellow circles indicate metabolites that are significantly associated with the co‐localization of close agronomic traits. PR, panicle rate; YPP, yield per plant; NSPP, number of spikes per plant; AB, aboveground biomass; SDW, straw dry weight; LWR, length width ratio of seed; GW, grain width; NSPS, number of spikelets per spike; FLW, flag leaf width; FLA, flag leaf area; FLL, flag leaf length; KGW, kilo‐grain weight; HI, harvest index; NGPS, number of grain per spike; GWPS, grain weight per spike; SL, spike length; PH, plant height. (b) Correlation analysis between two metabolites (wm0034, 4‐indolecarbaldehyde; mr1346, tryptophan) and NGPS. (c) LOD curves of QTL mapping for number of grains per spike, wm0034 (4‐indolecarbaldehyde), and mr1346 (tryptophan) levels on chromosome 4B. Green, number of grains per spike; Blue, 4‐indolecarbaldehyde; Red, tryptophan.

Figure 5

Metabolic data used to predict plant height (PH) and number of grains per spike (NGPS) based on two models. The BLUP and LASSO models were used to predict the plant height and number of grains per spike, respectively. Right: BULP prediction result. Left: LASSO prediction result. The x‐axis indicates the predictive value of agronomic traits and the y‐axis indicates phenotypic observations. The image was made using R (http://www.r‐project.org/).