Genome-wide co-expression analysis predicts protein kinases as important regulators of phosphate deficiency-induced root hair remodeling in Arabidopsis

Background: Phosphorus (P) is one of the essential but often limiting elements for plants. Based on transcriptional profiling we reported previously that more than 3,000 genes are differentially expressed between phosphate (Pi)-deficient and Pi-sufficient Arabidopsis roots (MCP 11(11):1156-1166, 2012). The current study extends these findings by focusing on the analysis of genes that encode protein kinases (PK) and phosphatases (PP) by mining PK and PP genes that were differentially expressed in response to Pi deficiency. Results: Subsets of 1,118 and 205 annotated PK and PP genes were mined on the basis of the TAIR10 release of the Arabidopsis genome. Analysis of RNA-seq data showed that 92 PK and 19 PP genes were not detected in roots (zero reads in three biological repeats); 96 PK and 10 PP showed low abundance (<= 10 reads). Gene ontology analysis revealed that the 188 PK genes with no or low expression level in Arabidopsis roots are mainly involved in pollen recognition, pollen tube growth or other processes not relevant for root hair formation. More than 50% of the cysteine-rich RLK (receptor-like protein kinase) subfamily genes belong to this group. Among the 29 PP genes with no or low expression level, purple acid phosphatases, haloacid dehalogenase-like hydrolases, and PP2C genes with functions in the dephosphorylation of RNA polymerase II C-terminal domain and mRNA capping were enriched. Subsets of 173 PK and 35 PP genes were differentially expressed under Pi-deficient conditions. Putative functional modules (clusters) of these PK and PP genes were constructed based on co-expression analysis using the MACCU toolbox. A co-expression network comprising 65 known or annotated PK and PP genes (60 PK and 5 PP genes, respectively) was subdivided into several highly co-expressed gene sub-clusters. The largest sub-cluster was composed of 22 genes, most of which have been assigned to the RLK superfamily and were associated with cell wall metabolism, pollen tube and/or root hair development and growth. Conclusions: We here provide comprehensive 'digital' transcriptional information on PK and PP genes in Arabidopsis roots. The co-expression network derived from our data mining approach sets the stage for follow-up experimentation that helps to complete our understanding of the post-translational regulation of Pi deficiency-induced changes in root hair morphogenesis.