Identification of priming-responsive and memory-associated NAC transcription factors
To investigate the expression pattern of NAC TFs in response to HS priming and during the memory phase, we treated Arabidopsis seedlings with a mild HS (priming) before exposing them to a severe HS (triggering stimulus). We collected samples at multiple time points after priming (up to 48 h into the thermomemory phase) and tested the expression of 104 NAC genes; seedlings kept in unprimed conditions were used as controls (Fig. 1).
Considering a 1.5-fold change as cut-off, 75 of the NACs were differentially expressed upon priming compared to control, while expression of 29 NAC genes was undetectable in the whole-seedling samples. The cut-off threshold was chosen considering that even moderate changes in TF expression levels can elicit strong downstream responses32. The expression of some NACs (including e.g., ANAC013 and ANAC029/AtNAP) in response to HS priming was similar to that reported in a previous study by Kilian, et al.30. NACs were grouped in clusters based on their expression pattern using k-means clustering. Changes in the expression of many NACs was already detectable immediately after the priming treatment (Fig. 1 and Supplementary Table S1). Three clusters comprising 33 genes in total (clusters 1, 2 and 3) appeared upregulated during early timepoints, and largely unchanged during later timepoints (Fig. 1), while 26 genes (in cluster 4) were initially strongly downregulated, and then strongly upregulated during later time points (Fig. 1). For the genes in cluster 5 we observed a strong downregulation at time point 0, which is immediately after completion of the 90 min HS, followed by a strong upregulation wthin the time frame of up to 2 h after the HS, and then a moderate to unchanged expression (Fig. 1). Taken together, the NAC TFs analyzed exhibited different expression profiles during HS priming and memory, suggesting different roles in the response to HS.
The effect of heat priming and triggering on the expression of NAC transcription factors
Pre-exposure of plants to a moderate stress (priming) alters their response to upcoming severe temperature stress. We tested this, at the molecular level, by investigating the transcriptional response of NAC TFs. We tested whether or not heat-priming affects the expression of NACs after the triggering HS. To this end, we determined the expression of NAC TFs after the triggering stimulus (T plants) and compared it with the expression in primed and triggered plants (P+T), using a 1.5-fold change cut-off. To identify specific expression patterns between P+T and T plants, a k-means clustering approach was applied, using 10 clusters (Fig. 2 and Supplementary Table S2). We noticed that expression of most NAC TFs was induced in primed + triggered plants, compared with triggered-only plants. Only few genes (cluster 9) showed an opposite expression pattern, being mostly repressed in primed + triggered plants and induced in triggered-only plants (Fig. 2). ATAF1 displayed a particularly interesting pattern: its expression remained almost unchanged in plants after priming + triggering (P+T), and expression was considerably induced in triggered-only conditions (without prior priming) (Fig. 2, cluster 9).
To test the hypothesis that ATAF1 is functionally involved in thermotolerance, additional physiological experiments were performed on ataf1 mutants and plants overexpressing ATAF1. Two other genes from clusters 4 (ANAC047/SHYG, AT3G04070) and 8 (ANAC013, AT1G32870) were included for comparison.
Functional analysis of selected NACs for thermomemory
To investigate the impact of the selected NAC genes on thermomemory, we assessed their functional involvement in the process by analyzing the phenotype of overexpressor and knockout (or knockdown) mutants in response to a triggering HS given 3 days after heat priming (3 days of thermomemory). As shown in Supplementary Figure S1, ANAC013 and SHYG overexpressors, shyg knockout and anac013 knockdown lines were not significantly different from wild type when compared in the thermomemory assay. However, ATAF1 appeared to be involved in thermomemory. A phenotypic analysis of ATAF1 transgenic plants for thermomemory showed a strong phenotype, where the ataf1-2 and ataf1-4 mutants had a significantly higher survival rate and fresh mass compared with WT plants, while plants overexpressing ATAF1 (called ATAF1-OE in the following) exhibited a severely reduced thermomemory (Fig. 3a, b and c). When the ataf1-4 mutant was transformed with an ATAF1 allele (pATAF1::ATAF1-GFP) which expresses an ATAF1-GFP fusion from the native ATAF1 promoter, thermomemory was restored to wild-type response (Supplementary Fig. S2). Taken together, these results suggest a negative regulatory role of ATAF1 in thermomemory.
Potential target genes of ATAF1
As our data suggested that ATAF1 is a negative regulator of thermomemory we sought to understand the role of ATAF1 in response to a heat priming stimulus by identifying its regulated target genes. To this end, WT, ATAF1-OE and ataf1-4 mutant seedlings were treated with the priming stimulus (37°C for 90 min), and samples were then collected at three time points (0 h, 1 h and 4 h; Fig. 4a) after the heat priming for gene expression profiling by RNA-seq. To eliminate the potential influence from other environmental factors, control samples (unprimed control) were collected at the same time points as the heat-treated samples.
To understand the relationship between samples in the RNA-seq, a hierarchical clustering analysis was performed and a heat map of the Pearson’s correlation coefficients of the expression profiles between all possible pairs of the samples was established (Fig. 4b). We observed a clustering between biological replicates, suggesting a high reproducibility of the experiment. The heat map also showed a strong separation between samples based on HS conditions, and by genotype (Fig.4b). The overall similarities and differences between samples were confirmed by multidimensional scaling (MDS) (Fig. 4c), supporting a stronger separation by condition (heat and control) than by genotype. We also noticed a similarity of samples under control temperature and those after 4 h of heat stress, suggesting a rapid recovery of most changes in gene expression. Of note, the expression of ATAF1 in WT seedlings during the memory phase, as revealed by RNA-seq, followed the pattern determined by qRT-PCR (Supplementary Fig. S3).
For all three genotypes, we found the highest number of differentially expressed genes (DEGs) at time point 0, right at the end of the 90-min heat priming treatment (Fig. 5a). To identify priming HS-associated genes, we used the same criteria that were previously used by Sedaghatmehr, et al.4. We investigated genes whose expression was induced after priming and remained high at all examined time points in the memory phase, as well as genes whose expression was down-regulated and remained low during the memory phase (Fig. 5a). Among the DEGs are thermomemory-associated HSPs4 whose expression in response to heat priming was similarly induced in ATAF1-OE, ataf1-4 mutant and WT, such as HSP22, HSP21, HSP17.4 and HSP18.2 (Supplementary Table S3). This finding corroborates the conclusion that these HSPs contribute to a general heat-response, which is common to all three genotypes.
To identify differential responses of ATAF1-OE and ataf1-4 mutant plants to heat priming, we analysed the RNA-seq data for genes upregulated in ATAF1-OE, compared to WT, but downregulated in ataf1-4 mutant plants, and vice versa (Fig. 5b). Next, to identify potential direct target genes of ATAF1, we analyzed the promoters of differentially expressed genes for either the presence of the ATAF1 binding site (reported by Garapati, et al.31) or for binding by ATAF1 as determined by DNA affinity purification sequencing (DAP-seq) experiments33. The analysis was performed for each time point after the heat priming (0 h, 1 h and 4 h). In total, we identified 64 genes as likely direct ATAF1 targets (Supplementary Table S4). We then refined our analysis by searching for those potential target genes that are commonly regulated by ATAF1 in all three, or two, time points after heat priming. Five genes, i.e., AT2G31260 (ATG9), AT2G41640 (GT61), AT3G44990 (XTH31), AT4G27720 and AT3G23540 were commonly regulated at two of the three timepoints, suggesting they might be priming-associated direct targets of ATAF1 (Fig. 5c); no gene was regulated at all three time points. ATG9 is an autophagy gene34 and experimental evidence indicates that autophagy plays a role in the heat stress response35,36. Its involvement in thermomemory is yet to be confirmed. The two genes GT61 and XTH31 are related to cell wall biosynthesis and expansion. Glycosyltransferase 61 (GT61) belongs to the glycosyltransferase (GT) family. GT proteins have diverse functions in plants, but most of them are likely involved in the biosynthesis of polysaccharides and glycoproteins in the cell wall. In grasses, including rice (Oryza sativa) and wheat (Triticumaestivum), GT61 family enzymes are involved in the synthesis of xylans, one of the main components of the cell wall37-39. GT61 has not yet been shown to be involved in heat stress priming or tolerance. XTH31 belongs to the xyloglucan endotransglucosylase/hydrolase (XTH) family. Generally, members of the XTH family are involved in cell wall remodeling, expansion and morphogenesis suggesting a potential role in stress responses40. In Arabidopsis, XTH31 is involved in regulating cell wall xyloglucan content41. The xth31 loss-of-function mutant has a reduced sensitivity to ABA, and seeds germinate faster than those of the WT41,42. Transgenic soybean (Glycine max) plants overexpressing XTH31 from Arabidopsis display enhanced tolerance to flooding along with more adventitious roots and longer primary roots43. The two genes AT4G27720 and AT3G23540 are not well characterized. AT4G27720 is annotated to encode a major facilitator superfamily protein with a molybdate ion transporter function, while AT3G23540 is annotated to encode a protein of the alpha/beta-hydrolase superfamily. The α/β-hydrolase enzymes are involved in various processes, including biosynthesis, metabolism, signal transduction, gene regulation44, and in the plants’ response and tolerance to salinity stress45.
Co-regulatory network of ATAF1 reveals potential co-regulation with ANAC055
Genes with similar expression patterns often share similar functions and are potentially regulated by the same transcription factors. In order to identify genes co-regulated with ATAF1 in response to heat priming, a co-expression analysis was performed using the transcriptomic data of ATAF1-OE, ataf1-4 mutant and WT plants, under control condition and upon exposure to heat. A weighted correlation network analysis was used to find modules of highly correlated genes. This method is commonly used to investigate relationships between genes and to identify gene candidates for further analysis46. The co-expression analysis revealed 40 modules in our data set (Fig. 6a). These modules contained groups of genes expressed in a similar way. It is likely that genes regulated directly by ATAF1, in response to heat, display similar expression patterns; hence, we took publicly available data on genes that have already been identified to be potentially targeted by ATAF1 (in general, not just by heat stress) and queried if these genes were present in our co-expression modules. The publicly available data of the potential direct targets were generated using a DAP-seq assay33. Putative target genes were designated as having a DAP-seq peak within the first 1,000 bp upstream of the transcription start site. Thus, we compared the list of genes that are proposed direct targets of ATAF1 (from O’Malley, et al.33) with the genes that are clustered in the co-expression analysis; we noticed that the potential targets of ATAF1 were overrepresented in co-expression clusters 10 and 13 (Fig. 6b). These clusters needed further investigation because they likely contained direct target genes of ATAF1, in response to heat stress. The clusters were further interrogated with the proposed TF targets identified using the DAP-seq assay developed by O’Malley, et al.33. We found that three of the clusters (10, 13, 21) showed significant enrichment for genes targeted by both ATAF1 and ANAC055, suggesting that the genes in these clusters might be co-regulated by the two NACs (Fig. 6b).
ATAF1 and ANAC055 appear to have a common role in thermomemory
Our co-expression analysis showed that ATAF1 and ANAC055 had a large number of shared co-expressed targets in response to HS (Supplementary Table S5). We therefore tested if ANAC055 is also functionally involved in regulating thermomemory. To this end, we assessed the thermomemory phenotype of transgenic lines with altered expression of ANAC055 (Fig. 7a,b). Two anac055 mutants (anac055-1 and anac055-2) were tested, and both showed a substantial increase in fresh mass ratio and survival compared with WT, while plants overexpressing ANAC055 showed a decrease in both parameters (Fig. 7b). These results indicate a negative role for ANAC055 in the regulation of thermomemory, similar to that of ATAF1. To test whether or not ATAF1 and ANAC055 worked in a fully or partially redundant manner, we generated an ataf1/anac055 double mutant and exposed seedlings to HS and compared their phenotypes with the phenotypes of the corresponding single-gene mutants. The fresh mass ratio of the ataf1/anac055 double-mutant was higher than that of the WT, but not significantly different from that of the single-gene knockout mutants ataf1-4 and anac055 (Fig. 7c). This result indicates that ATAF1 and ANAC055 require each others function to control thermomemory.