Bioinformatic analysis of epigenetic and microRNA mediated regulation of drought responsive genes in rice

Abstract

Drought stress response is a complex trait regulated at multiple levels. Changes in the epigenetic and miRNA regulatory landscape can dramatically alter the outcome of a stress response. However, little is known about the scope and extent of these regulatory factors on drought related cellular processes and functions. To this end, we selected a list of 5468 drought responsive genes (DRGs) of rice identified in multiple microarray studies and mapped the DNA methylation regions found in a genome wide methylcytosine immunoprecipitation and sequencing (mCIP-Seq) study to their genic and promoter regions, identified the chromatin remodeling genes and the genes that are targets of miRNAs. We found statistically significant enrichment of DNA methylation reads and miRNA target sequences in DRGs compared to a random set of genes. About 75% of the DRGs annotated to be involved in chromatin remodeling were downregulated. We found one-third of the DRGs are targeted by two-thirds of all known/predicted miRNAs in rice which include many transcription factors targeted by more than five miRNAs. Clustering analysis of the DRGs with epigenetic and miRNA features revealed, upregulated cluster was enriched in drought tolerance mechanisms while the downregulated cluster was enriched in drought resistance mechanisms evident by their unique gene ontologies (GOs), protein-protein interactions (PPIs), specific transcription factors, protein domains and metabolic pathways. Further, we analyzed the proteome of two weeks old young rice plants treated with a global demethylating agent, 5-azacytidine (5-azaC), subjected to drought stress and identified 56 protein spots that are differentially expressed. Out of the 56 spots, 35 were differently expressed in the sample with both demethylation and drought stress treatments and 28 (50%) were part of DRGs considered in the bioinformatic analysis.

Figure 1. Classification of Drought Responsive Genes (DRGs) into nine clusters based on epigenetic/miRNA features and differential expression.

Cluster D: all DRGs, DU: upregulated DRGs, DD: downregulated DRGs, E: Genes with any or all epigenetic/miRNA features, NE: no epigenetic/miRNA features, EU: E with upregulated DRGs only, ED: E with downregulated DRGs only, NEU: NE with upregulated DRGs only and NED: NE with downregulated DRGs only.

Figure 2. Enrichment of epigenetic features in DRGs.

A) Epigenetic features of DRGs versus random set. The number of mCIP-reads mapped to the genic region of DRGs compared to the random set. In the same way mCIP-reads that mapped to promoter region were compared. Total numbers of miRNAs from PMRD database targeting the DRGs were compared to the random set. B) Distribution of multiple instances of epigenetic features on DRGs. ∧Each of the bars represents number of DRGs mapped with given number of mCIP-reads only. #Each of the bars represents number of DRGs targeted by given number of miRNA only. C) Comparison of sets of genes with unique epigenetic features in DRGs with the random set. Unique represents the set of genes with only one of the three epigenetic/miRNA features and all represents number of genes with a particular feature and with one or more other epigenetic/miRNA features. D) Distribution of the average of absolute fold change of gene expression from for the nine clusters. * indicates significant Z-score at p<0.05 and ** indicates significant Z-score at p<0.01.

Figure 3. Four way venn diagrams depicting overlap of different characteristics between the clusters NED, NEU, ED and EU.

A) GO terms analyzed by AgriGO, B) Protein domain families as per Pfam database, C) Metabolic pathways by RiceCyc and D) Types of transcription factors as reported by PlnTFDB.

Figure 4. Examples of enrichment/depletion of significant GO terms in DRG clusters.

A) and B) show increase and C) shows decrease in the significance of GO terms as we move to sub clusters indicated by vertical arrows. Also shown are the changes in significance of GO terms from a parent to child GO term indicated by horizontal arrows. White boxes denote GO terms that are not significant (p>0.05).

Figure 5. Number of protein protein interactions (PPIs) found in the four leaf clusters.

The numbers above the bars represent average number of PPIs per gene over total number of PPIs found in the cluster and average number of PPIs per gene among genes with ≥1 PPI over total number of PPIs found in the cluster.

Figure 6. Protein-protein interactions (PPIs) network of cluster-EU.

A) Network diagram showing DRGs as circles with size and color corresponding to number of PPIs. The large size of the circle and color intensity indicate higher number of PPIs,. Thickness of edges between two nodes is based on String-DB score. Thicker edges denote high String-DB scores, B) The number of genes with different number of PPIs (X-axis).