Analysis of metabolic dynamics during drought stress in Arabidopsis plants

Abstract

Drought is a major cause of agricultural losses worldwide. Climate change will intensify drought episodes threatening agricultural sustainability. Gaining insights into drought response mechanisms is vital for crop adaptation to climate emergency. To date, only few studies report comprehensive analyses of plant metabolic adaptation to drought. Here, we present a multifactorial metabolomic study of early-mid drought stages in the model plant Arabidopsis thaliana. We sampled root and shoot tissues of plants subjected to water withholding over a six-day time course, including brassinosteroids receptor mutants previously reported to show drought tolerance phenotypes. Furthermore, we sequenced the root transcriptome at basal and after 5 days drought, allowing direct correlation between metabolic and transcriptomic changes and the multi-omics integration. Significant abiotic stress signatures were already activated at basal conditions in a vascular-specific receptor overexpression (BRL3ox). These were also rapidly mobilized under drought, revealing a systemic adaptation strategy driven from inner tissues of the plant. Overall, this dataset provides a significant asset to study drought metabolic adaptation and allows its analysis from multiple perspectives.

Fig. 1

Experimental design and sampling procedure. Scintillation vials represent metabolite sampling points whereas microtubes represent RNA sampling points (Days 1 and 5).

Fig. 2

Technical validation. (a) Boxplot with the distribution of metabolomic measures (log-scale) per sample. Note the few samples showing an evident outlier distribution. These are likely artifacts and were deleted for subsequent analyses. Red dashed lines denote the Q1 and Q3 of the whole dataset. (b) Graphical representation of Principal Component Analysis (PCA). Axis represent the two first components explaining most of the dataset variability. Data matrix was centered and scaled to unit variance with R prcomp function. (c) Same than in (b) but limited to root samples only. (d) Profile of the osmoprotectant sugar raffinose along the drought time course in WT shoots. Points denote individual measurements and lines denote the medians. Note the large accumulation in later stages of drought, whereas the watered control remains unchanged. (e) Same than in (d) but in WT roots.

Fig. 3

Validation of RNAseq. (a) PCA plot of samples based on gene counts (NOISeq R package). PC1 roughly coincide with drought, which clearly separates samples. PC2 roughly corresponds with the genotype, which also separates well the samples. Data matrix was centered and scaled to unit-variance. (b) GO enrichment analysis of differentially upregulated genes on the pairwise comparison WT drought vs. WT control. The great enrichment values obtained for response to water (GO:0009415), response to water deprivation (GO:0009414) and other stress-related categories validate the effect of drought on transcriptomics, thus supporting the quality of the dataset. The p.adjust parameter is the FDR-adjusted p-value of the enrichment test. Count is the number of deregulated genes annotated in a particular GO category and GeneRatio is the Count number divided by number of deregulated genes that are not annotated in such category. GO enrichment analysis performed with ClusterProfiler package, based on org.At.tair.db annotation.

Fig. 4

Usage examples: (a) Time 0 comparison between WT and BRL3ox. Boxplots represents the (log) fold-change relative to the WT median. Points are the particular relativized values of each replicate. Important osmoprotectant metabolites appeared accumulated already in basal conditions in BRL3ox roots, supporting the priming hypothesis of these plants. (b) Metabolites whose ratio shoot/root is significantly altered after six days of drought in WT plants. Boxplot represent the (log) ratio between shoot and roots. Points represent the particular values of each replicate. (c) Median profile of a cluster of metabolites that follow differential dynamics between BRL3ox and WT along drought. Note how both genotypes exponentially accumulate these osmoprotectant metabolites along drought, however in BRL3ox this accumulation is way steeper. (d) Median profiles of the same metabolites than in (c) but in the watered series. Any metabolite was identified as significantly affected by time in the watered series. Note the y-axis scale, despite the apparent fluctuations in BRL3ox, these changes are very small compared to drought.

Fig. 5

The dynamic evolution of the calculated Composite Index (CI) for each cluster in Col-0 and BRL3ox genotypes. Each line represents a metabolite cluster. Dashed lines are CI of clusters that are not yielding DNBs. Solid lines are the cluster in which a DNB is identified at the critical point (encircled). Metabolite names of the DNBs are denoted next to the critical point for each genotype. (a) Analysis of CI evolution in shoots. (b) Analysis of CI evolution in roots.