Comparative transcriptome meta-analysis of Arabidopsis thaliana under drought and cold stress

Abstract

Multiple environmental stresses adversely affect plant growth and development. Plants under multiple stress condition trigger cascade of signals and show response unique to specific stress as well as shared responses, common to individual stresses. Here, we aim to identify common and unique genetic components during stress response mechanisms liable for cross-talk between stresses. Although drought and cold stress have been widely studied, insignificant information is available about how their combination affects plants. To that end, we performed meta-analysis and co-expression network comparison of drought and cold stress response in Arabidopsis thaliana by analyzing 390 microarray samples belonging to 29 microarray studies. We observed 6120 and 7079 DEGs (differentially expressed genes) under drought and cold stress respectively, using Rank Product methodology. Statistically, 28% (2890) DEGs were found to be common in both the stresses (i.e.; drought and cold stress) with most of them having similar expression pattern. Further, gene ontology-based enrichment analysis have identified shared biological processes and molecular mechanisms such as-'photosynthesis', 'respiratory burst', 'response to hormone', 'signal transduction', 'metabolic process', 'response to water deprivation', which were affected under cold and drought stress. Forty three transcription factor families were found to be expressed under both the stress conditions. Primarily, WRKY, NAC, MYB, AP2/ERF and bZIP transcription factor family genes were highly enriched in all genes sets and were found to regulate 56% of common genes expressed in drought and cold stress. Gene co-expression network analysis by WGCNA (weighted gene co-expression network analysis) revealed 21 and 16 highly inter-correlated gene modules with specific expression profiles under drought and cold stress respectively. Detection and analysis of gene modules shared between two stresses revealed the presence of four consensus gene modules.

Fig 1. Workflow for data collection, curation and co-expression network analysis.

(A) In total 29 series (26 series from NCBI-GEO and 3 series from ArrayExpress) comprising of 241 and 149 Affymetrix A. thaliana arrays related to drought and cold stress were used in the analysis. (B) Workflow describes the steps for co-expression network generation and consensus module detection.

Fig 2. Number of unique and common differentially expressed genes (DEGs) found in A.thaliana under cold and drought stress.

Total number of genes is shown in bold, below which are the percentages of genes.

Fig 3. Abundance of transcription factors in transcription factor families expressed commonly under cold and drought stress.

Grey bar: Number of transcription factors in each transcription factor family expressed under cold stress. Black bar: Number of transcription factors in each transcription factor family expressed under drought stress.

Fig 4. Dendrogram and heatmap of DEGs found under cold and drought stress in A. thaliana.

The heatmap describes the Topological Overlap Matrix (TOM) for all DEGs used for co-expression network analysis. Color ranges from darker red to light red according to overlap i.e. higher to low and darker red color blocks along the diagonal are the modules. The gene dendrogram and modules (as different color bars) are also shown along the left side and the top.

Fig 5. The medianRank and Zsummary statistics of module preservation of cold modules in drought modules (y-axis) vs. module size (x-axis).

In plot colored circles represents gene modules of cold common gene co-expression network. The black borderline represents no preservation (Z-score = 0), blue borderline represents very weak preservation limits (Z-score = 2). The region between blue borderline and green borderline represents weak to moderate preservation zone (10 < Z-score > 2) and above green borderline represent strong preservation (Z-score> 10).

Fig 6. Consensus module detection and comparison.

(A) The plots depicting various network indices (y-axes) as functions of the soft- thresholding power (x-axes). Numbers in the plots represents soft-thresholding powers. The plots demonstrate that 10 is the smallest soft-thresholding power at which approximate scale-free topology is accomplished for both sets as the various connectivity measures decrease sharply with increasing soft-thresholding power. (B) and (C) Relation between consensus modules and modules found individually in cold and drought expression set. The table represents the analogy between consensus modules and modules of cold and drought stress related global co-expression network based on the expression values of the common genes. Each row of the table represent individual stress specific module and each column represent one consensus module. Numbers in the table represent genes common between individual stress module and consensus module. The table coloring pattern represents the negative log of Fisher’s exact test p-value for the overlap of the two compared modules. The darker red color represents more significant overlap.

Fig 7. Differential consensus module eigengene analysis between cold and drought consensus module eigengene networks.

Differential eigengene network analysis was used to address the strength of the correlation preservation for all eigengene pairs across the two networks: (A) and (B) Clustering dendrograms of consensus module eigengenes (ME) showing the presence of meta-modules as depicted by the presence of same major branching pattern in both cold and drought eigengene network dendrograms. (C) and (F) Heatmaps of eigengene adjacencies in each of the consensus eigengene networks for cold and drought dataset respectively. Each of the rows and columns represents an eigengene tagged by the consensus module color and within the heatmap, red color represents high adjacency and positive correlation, whereas blue represents low adjacency and negative correlation, as represented by the color legend. (D) Bar plot depicts the preservation measure for each consensus eigengene as the height of the bar (y-axis) where each colored bar corresponds to the eigengene of the associated consensus module. The high-density value D(Preserv^Drought,Cold) = 0.91 indicates the high overall preservation between the two networks. (E) Heatmap representation of Adjacencies of the pair-wise preservation network (Preserv^Drought,Cold); high values of (Preserv^Drought,Cold) implies that there is a strong correlation preservation between pairs of module eigengenes across the two networks. Each column and row represented by consensus module eigengene with the saturation of red color showing adjacency according to the color legend.