Arabidopsis HEAT SHOCK TRANSCRIPTION FACTORA1b regulates multiple developmental genes under benign and stress conditions

Abstract

In Arabidopsis thaliana, HEAT SHOCK TRANSCRIPTION FACTORA1b (HSFA1b) controls resistance to environmental stress and is a determinant of reproductive fitness by influencing seed yield. To understand how HSFA1b achieves this, we surveyed its genome-wide targets (ChIP-seq) and its impact on the transcriptome (RNA-seq) under non-stress (NS), heat stress (HS) in the wild type, and in HSFA1b-overexpressing plants under NS. A total of 952 differentially expressed HSFA1b-targeted genes were identified, of which at least 85 are development associated and were bound predominantly under NS. A further 1780 genes were differentially expressed but not bound by HSFA1b, of which 281 were classified as having development-associated functions. These genes are indirectly regulated through a hierarchical network of 27 transcription factors (TFs). Furthermore, we identified 480 natural antisense non-coding RNA (cisNAT) genes bound by HSFA1b, defining a further mode of indirect regulation. Finally, HSFA1b-targeted genomic features not only harboured heat shock elements, but also MADS box, LEAFY, and G-Box promoter motifs. This revealed that HSFA1b is one of eight TFs that target a common group of stress defence and developmental genes. We propose that HSFA1b transduces environmental cues to many stress tolerance and developmental genes to allow plants to adjust their growth and development continually in a varying environment.

Fig. 1.

Alteration of HSFA1b binding in response to HS. (A) Heat map with k-means clustering showing the enrichment of ChIP-seq signals from NS and HS samples at a 4 kb window around HSFA1b-bound regions in the genome. (B) Venn diagram showing the number of HSFA1b target genomic features in Groups I–III. (C) GO Slim analysis heat map comparing genomic features of enriched Biological Process terms in Group I–III with Benjamini–Hochberg FDR <0.05. (D) Three histograms showing the frequency of HSFA1b binding relative to the distance from the TSS of target genomic features in Groups I–III. (E) Pie charts showing the distribution of HSFA1b binding on target genomic features in Groups I–III.

Fig. 2.

Genome browser view of normalized ChIP-seq tags in NP:HSFA1b NS and NP:HSFA1b HS along with controls showing examples of HSFA1b binding to genes in (A) Group I, (B) Group II, and (C) Group III. (D) Examples of HSFA1b targeting the promoter of HSFB2a (red rectangle) as well as a region known to code for an antisense RNA that targets HSFB2a (asHSFB2a; black rectangle). (E and F) HSFA1b targeting a region that codes for asCRY1 and asHSFC1, respectively.

Fig. 3.

Most HSFA1b target genes are responsive to HS. (A) Scatter plots of transcript abundance of HSFA1b target genes in response to HS. Red and blue dots represent up- and down-regulated genes, respectively (q≤0.05), and black dots show genes with expression q>0.05. The numbers shown in each panel are the number of DEGs in each group. (B) GO Slim analysis heat map comparing the enriched BP terms of up- (+) and down- (–) regulated HS-responsive genes in Groups I–III (Benjamini–Hochberg FDR <0.05).

Fig. 4.

35S:HSFA1b plants under NS partially mimic HS WT plants. (A) Heat map comparing normalized FPKM values for HSP genes in NS and HS WT and 35S:HSFA1b NS plants. (B) GO Slim analysis comparing the enriched Biological Process terms (Benjamini–Hochberg FDR <0.05) of up- (+) or down (–)-regulated DEGs in HS WT and NS 35S:HSFA1b plants compared with NS WT plants.

Fig. 5.

HSFA1b can indirectly control expression through its regulation of TF gene expression. (A) Heat map comparing normalized FPKM values for 28 TF genes bound by HSFA1b and differentially expressed in 35S:HSFA1b and WT HS plants. Asterisks (*) indicate development-associated TF genes. (B) Confirmation of the regulation of selected TF genes by clade A1 HSF genes. Quantitative real-time RT–PCR was conducted on RNA from qKO rosettes under NS and HS in comparison with its parental genotypes Col-0 (C) and Ws-0 (W). The suffixes ‘a’ and ‘b’ are where the qKO mutant shows a significant difference (P<0.05) under the same conditions (NS or HS) from Col-0 and Ws-0, respectively. (C) Venn diagram showing the overlap between HSFA1b target genes scored from the ChIP-seq data (Supplementary Data S1) and the target genes bound by HSFA1b (HSF3) from the Arabidopsis Cistrome Atlas (http://neomorph.salk.edu/dev/pages/shhuang/dap_web/pages/index.php). The boxed callout number is the P-value for the significance of the overlap between the two data sets (hypergeometric distribution test). (D) An overview of a Cytoscape-generated HSFA1b hierarchical TF gene network using the data outputs from the Cistrome Atlas with the ChIP-seq and RNA-seq data from this study. The yellow node is HSFA1b, red nodes are TF genes bound by HSFA1b, and blue nodes are differentially expressed TF genes that respond to HS and HSFA1b overexpression, are not bound by HSFA1b, but are scored as binding to the red node TFs. An interactive version of this network is available as an interactive file (Supplementary Cytoscape File S1).

Fig. 6.

HSFA1b regulates expression of cisNAT genes and their target sense genes. (A) Heat map showing differentially expressed cisNAT genes and their putative sense targets in WT plants in HS compared with NS and in 35S:HSFA1b NS plants compared with WT NS plants. (B) Linear correlation plots showing the relationship between cisNAT and sense target transcript abundance for WT plants subjected to HS compared with NS (top panel) and for 35S:HSFA1b NS plants compared with WT NS plants (bottom panel). (C) Transcript levels determined by qPCR of selected cisNAT genes and their TF gene sense target in HS and NS qKO plants compared with their parental genotypes as in the legend of Fig. 5B. The suffixes ‘a’ and ‘b’ are where the qKO mutant shows a significant difference (P<0.05) under the same conditions (NS or HS) from Col-0 (C) and Ws-0 (W) respectively.

Fig. 7.

Seven TFs co-ordinate with HSFA1b. (A) All significant motifs within HSFA1b peaks discovered by MEME (P<0.0001) in the NS and HS data sets. (B) Density heat maps showing enrichment of ChIP-seq signals of the seven TFs on 10 kb around the regions bound by HSFA1b. (C) Combined Jaccard index and pHYPER correlation matrix showing the significance of overlaps between the target genes of the seven TFs and HSFA1b under NS conditions only. Numbers of genes in each overlap are in parentheses. (D) GO Slim analysis comparing the enriched Biological Proess terms of the common targets between HSFA1b and the other seven TFs. (E) Genome browser view of normalized ChIP-seq tags from the ChIP-seq data of the eight TFs showing examples of target genes bound by HSFA1b and up to seven other TFs.