Protein sequences of DREB1 genes from Arabidopsis and rice were aligned using ClustalX. A neighbor-joining tree with a bootstrap setting of 1,000 was generated using MEGA7.0.14 (Kumar et al., 2016). Expression data of OsDREB1 genes in response to abiotic stresses were obtained from GENEVESTIGATOR.¹

All rice plants in this study are Oryza sativa cv. Nipponbare (NIP). Mutants were generated in NIP using CRISPR/Cas9 system (Xing et al., 2014). Specific gRNA targets were designed using CRISPR-PLANT website.² The gRNA targets were cloned into the pHUE411 vector (Xing et al., 2014), and the resulting constructs were transformed into NIP by Agrobacterium-mediated transformation (Nishimura et al., 2006). Genomic fragments around the target sites in the mutants were amplified by PCR, and the edited mutations were identified by sequencing the PCR products. The homozygous mutants were selected by PCR based genotyping. All the primers used in this study are listed in Supplementary Table 1.

All plants were grown in growth chamber with a photoperiod of 16h-light/8h-dark and at 28°C unless specified (such as for stress treatment). Seeds were sterilized with 5% bleach for 10 min followed by three washes with water. The sterilized seeds were soaked in water at 28°C for 3 days for germination. For hydroponic growth, seeds at a similar germination stage were transferred to a 96-well plate with the bottom cut out. Seedlings were grown hydroponically till the third leaf just emerged, and then cultured in Yoshida nutrient solution (Yoshida et al., 1971). For soil growth, the germinated seeds were transferred from water to soil (nutrient soil:vermiculite = 3:1) and grown in growth chamber.

All stress treatments were carried out in growth chambers. For chilling treatment, the hydroponically grown seedlings at three-leaf stage (when the third leaf was fully expanded) were subject to 6°C treatment for 2.5–4 days and then returned to 28°C for 7 days for recovery growth. For suboptimal low temperature treatment, germinated seeds were transferred to soil and grown under 16°C. For cold acclimation treatment, seedlings at three-leaf stage (grown hydroponically at 28°C) were subject to 1 day of 12°C growth (cold acclimation, CA) or kept at 28°C for 1 day (no acclimation, NA). They were then subject to a 6°C treatment for 3.5–4 days before a 7-day recovery growth at 28°C. For heat treatment, the soil grown seedlings at three-leaf stage were treated at 48°C for 2 days and then returned to 28°C for 7-day recovery growth. For salt treatment, the soil grown seedlings of 3-day-old were treated with 200 mM NaCl for 21–23 days. For drought treatment, the soil-grown seedlings at three-leaf stage were withheld of watering for 10 days and then re-watered for 7 days for recovery growth. For survival analysis, 24 hydroponically grown seedlings or 21 soil growth seedlings were used in a biological repeat. At least three biological replicates were performed for each analysis.

Accumulations of H₂O₂ and cell death were detected by 3,3′-diaminobenzidine (DAB) staining and trypan blue staining, respectively, as previously described (Bach-Pages and Preston, 2018). The second leaves of the soil grown seedlings at three-leaf stage were used.

The three-leaf stage seedlings that were soil grown at 28°C were treated at 6°C for 0, 4, and 24 h before all above ground tissue of three biological replicates were collected. Total RNA was extracted from tissues using an RNA extraction kit (RP1002, Bioteke Corporation). RNA sequencing (RNA-seq) was carried out at Berry Genomics.³ Approximately 6.0 GB of raw reads were generated from each sample. The clean reads were mapped to rice genome,⁴ and expression level for each gene was determined by its relative abundance or value of fragments per kilobase of exon model per million mapped (FPKM). Gene expression values in FPKM was used for principal component analysis.⁵ Differentially expressed genes (DEGs) were identified using edgeR package (3.32.1) by R Language (R Core Team, 2021). DEGs between two samples were defined by fold change ≥ 2 and false discovery rate (FDR) ≤ 0.01. Heat maps were plotted via TBtools or using pheatmap (1.0.12) and ggplot2 (3.3.3) packages by R language (Chen et al., 2020; R Core Team, 2021). Venn diagrams were plotted using Venn Diagram package (1.6.20) by R language (R Core Team, 2021). Gene ontology (GO) enrichment analysis was performed by PANTHER.⁶ GO annotation was downloaded from RiceXPro.⁷ The promoter analysis was performed by a manual search for the DRE/CRT motif (G/ACCGAC) in the 1.5 kb fragment 5′ to the translation start codon of the candidate gene. The RNA-seq data were submitted to National Center for Biotechnology Information (accession number: PRJNA797855).⁸

The DREB1 family has six members in Arabidopsis and 10 members (named OsDREB1A to OsDREB1J) in rice.⁹ Analysis of genes from 38 organisms revealed that this gene family diverged before the separation of Arabidopsis and rice, and the DREB1 genes have independently amplified in two species (Supplementary Figure 1). This suggests that the three CBF genes in Arabidopsis do not have a one-to-one orthologous DREB1 genes in rice. Protein sequence homology analysis of the six Arabidopsis and 10 rice DREB1 genes, however, revealed that OsDREB1C, OsDREB1E, OsDREB1F, and OsDREB1G are likely closer to the six Arabidopsis DREB1 genes in sequence identity (Figure 1A).

We further examined the expression patterns of these 10 OsDREB1 genes (except for OsDREB1H which is absent in the database) using the curated high-quality transcriptome data at GENEVESTIGATOR (see Footnote¹). All these nine OsDREB1 genes (1A, 1B, 1C, 1D, 1E, 1F, 1G, 1I, and 1J) have altered expression in response to one or multiple abiotic stresses. Chilling treatment (4°C for 4 h) increased the expression of 1A, 1B, 1C, 1E, 1F, 1G and 1I. Dehydration treatment (on 3 mm Whatman paper for 3 h) increased the expression of 1A, 1B, 1C, 1E, 1F, and 1G. Drought treatment (no watering till leaves were completely rolled) increased the expression of 1C. Heat treatment (42°C for 3 h) increased expression of 1I and repressed the expression 1A and 1B. Salt treatment (300 mM NaCl for 5 h) had no significant effect on the expression of any of these nine OsDREB1 genes. Simulated drought treatment (25% polyethylene glycol6000 for 1 h) increased the expression of 1C and 1G and repressed the expression 1A and 1B. Wounding treatment (needle puncturing 15 times in 6 h) increased expression of 1A and 1B (Figure 1B). Therefore, most of the OsDREB1 genes had increased expression in response to chilling and dehydration stresses.

We selected three OsDREB1 genes (OsDREB1C, OsDREB1E, and OsDREB1G) for further functional analysis because they are more related to each other and are more closely related to the Arabidopsis DREB1 genes than six other rice genes (Figure 1A). LOF mutants of these three genes were generated in the NIP background by using CRISPR/Cas9 gene editing system (Xing et al., 2014). Two gRNA targets were used to generate single mutants of dreb1c (via targets C1, C2), dreb1e (via targets E1, E3), and dreb1g (via targets G2, G3). A total of four dreb1c mutant alleles (dreb1c-1, dreb1c-2, dreb1c-3, and dreb1c-4), two dreb1e alleles (dreb1e-1 and dreb1e-2), and one dreb1g mutant (dreb1g-1) were obtained (Table 1). A gRNA target (C2) specific for OsDREB1C and gRNA targets (E2, G1) for both OsDREB1E and OsDREB1G was used to generate the dreb1ceg triple mutant (Table 1).

All these mutants were LOF as predicted by the mutation types. The dreb1c-1 mutant has an insertion of “C” at position 12 bp (relative to the translation initiation site) and a deletion at 13–279 bp, which are predicted to cause a reading frame shift. The dreb1c-2 mutant has a deletion of 35–37 bp and an insertion of “A” at 275 bp, leading to a reading frame shift. The dreb1c-3 mutant has an insertion of “T” and “A” at 38 and 275 bp, respectively, leading to coding termination. The dreb1c-4 mutant has a deletion at 32–67 bp and an insertion of “A” at 275 bp leading to a reading frame shift. The dreb1e-1 mutant has an insertion of “TGG” and “T” at 21 and 337 bp, respectively, leading to a reading frame shift. The dreb1e-2 mutant has an insertion of “G” at 21 bp and a deletion of “G” at 336 bp, leading to protein sequence changes starting from amino acid (aa) positions 8 and 112, respectively. The dreb1g-1 mutant has an insertion of “A” at 164 bp, leading to coding termination. The dreb1ceg mutant has a deletion of “G” at 274 bp in OsDREB1C, and an insertion of “A” and “C” at 50 bp in OsDREB1E and OsDREB1G, respectively, leading to a reading frame shift of OsDREB1E and coding terminations of OsDREB1C and OsDREB1G (Table 1). Homozygous single or triple mutants of these three genes were used for further analyses.

The role of OsDREB1C/E/G genes in chilling tolerance was assessed by chilling treatment of the single and triple mutants of dreb1c, dreb1e, and dreb1g. Three-leaf-stage seedlings were subjected to 2.5–4 days of chilling treatment at 6°C and allowed to recover for 7 days. All dreb1 single mutants had significantly lower survival rates than NIP, although the survival rate of the WT differed among different sets of experiments likely due to a slight variation of temperature and duration of chilling treatment (Figures 2A–F; Supplementary Figure 2). The dreb1c-1 and dreb1c-2 mutants had survival rates of 19% and 27%, respectively, lower than the rate of 42% for NIP (Figure 2A). The dreb1c-3 and dreb1c-4 mutants had survival rates of 55% and 49%, respectively, lower than the rate of 92% for NIP (Supplementary Figures 2A–C). The dreb1e-1 mutant had a survival rate of 36%, lower than the rate of 55% for NIP (Figure 2B). The dreb1e-2 mutant had survival rates of 53%, lower than the rate of 72% for NIP (Supplementary Figures 2D,E). The dreb1g-1 mutant had a survival rate of 29%, lower than the rate of 56% for NIP (Figure 2C). The dreb1ceg triple mutant also had lower survival rate than the wild type NIP (27% versus 57%; Figure 2D). Chilling tolerance levels among dreb1 single mutants and the triple mutants were also compared. The survival rates of dreb1e-1 and dreb1g-1 were 50% and 59%, respectively, higher than the rate of 23% for dreb1c-3 (Figure 2E). The survival rate of dreb1c-3 was 63%, higher than the rate of 30% for dreb1ceg (Figure 2F). These results indicate that all three rice genes, OsDREB1C/E/G, positively regulate chilling tolerance in rice at seedling stage.

We further analyzed H₂O₂ accumulation by DAB staining as ROS accumulation is highly related to chilling tolerance. All dreb1 mutants had more H₂O₂ accumulation in leaves than NIP after 6°C treatment for 24 h. The dreb1c-3 mutant had more H₂O₂ accumulation than dreb1e-1 and dreb1g-1, and dreb1ceg had the highest H₂O₂ accumulation than all single mutants (Figure 2G). Cell death caused by chilling was subsequently analyzed by trypan blue staining. All dreb1 mutants had more cell death than NIP after chilling treatment. The dreb1c-3 mutant had more cell death than dreb1e-1 and dreb1g-1, and dreb1ceg had the highest cell death than all single mutants (Figure 2H). Therefore, H₂O₂ accumulation and cell death are each associated with chilling susceptibility in the rice dreb1 mutants.

We analyzed the growth phenotypes of the dreb1 mutants at non-chilling temperatures. Under normal growth temperature (28°C), no growth difference was observed between the dreb1 mutants and NIP at seedling, tillering, or mature stages (Supplementary Figure 3). Under a suboptimal low temperature (16°C), none of the single dreb1 mutants showed detectable growth difference compared with NIP at the seedling stage (Figures 3A–C). In contrast, the dreb1ceg triple mutant was significantly shorter than the WT NIP at 16°C, with a height of 7.9 cm compared to 11.5 cm of the WT (Figure 3D). These results indicate that the OsDREB1C/E/G genes have an overlapping function in promoting rice growth at suboptimal low temperature.

We further determined whether OsDREB1C/E/G genes regulate cold acclimation. The WT NIP plants exhibited a higher survival rate at 6°C with a prior 12°C treatment of 1 day compared to no prior treatment (Figure 4), indicating that rice has cold acclimation for chilling tolerance. All the dreb1 mutants also exhibited cold acclimation. In one set of experiment, NIP, dreb1c-1, and dreb1c-2 had survival rates of 63%, 42%, and 53%, respectively, with prior cold treatment (cold acclimation, CA), higher than the rates of 37%, 0%, and 0% observed with no prior cold treatment (no acclimation, NA; Figure 4A). NIP and dreb1e-1 had survival rates of 57% and 34%, respectively, with CA, higher than the rates of 28% and 9% with NA (Figure 4B). NIP and dreb1g-1 had survival rates of 75% and 52%, respectively, with CA, higher than the rates of 54% and 25% with NA (Figure 4C). NIP and dreb1ceg had survival rates of 86% and 72%, respectively, with CA, higher than the rates of 53% and 29% with NA (Figure 4D). With CA, all mutants including dreb1c-1, dreb1e-1, dreb1g-1, and dreb1ceg plants showed more chilling susceptibility compared with WT NIP (Figures 4A–D). The only exception was dreb1c-2 which had a lower survival rate compared with NIP, but the difference was not significant (Figure 4A). This indicates that OsDREB1C/E/G genes do not affect cold acclimation in rice. We further analyzed the genotype by environment (G x E) interaction between each genotype pair (WT versus the dreb1 mutants) and cold acclimation (NA versus CA) on survival rates. No significant interaction was found for any of the OsDREB1 genotype and CA, indicating that the dreb1 mutants have a cold acclimation-independent chilling susceptibility (Figure 4). These results indicate that OsDREB1C/E/G are not important for cold acclimation in rice, or they have overlapping functions with other OsDREB1 genes in cold acclimation.

To further understand how OsDREB1 regulate chilling tolerance in rice, we carried out RNA-seq analysis on the WT, dreb1c-2, dreb1g-1, and dreb1ceg plants at 3-leaf-stage before and after a chilling treatment of 6°C. Principal component analysis (PCA) revealed that the WT and the dreb1 mutants were similar to each other at any time points in the first two principal components, and chilling treatment was the most important factor in differentiating the samples (Supplementary Figure 4).

This PCA showed that chilling treatment induced a similar global transcriptional change in WT and dreb1 mutants. A total of 747 chilling up-regulated and 406 chilling down-regulated DEGs were shared among the WT, dreb1c-2, dreb1g-1, and dreb1ceg at 4 h of 6°C treatment, accounting for 69%, 56%, 56%, and 70% of up-regulated DEGs as well as 52%, 45%, 42%, and 52% of down-regulated DEGs in the WT, dreb1c, dreb1g, and dreb1ceg, respectively (Supplementary Figure 5). Moreover, 2,774 up-regulated and 2,148 down-regulated DEGs were shared among the WT and three dreb1 mutants at 24 h, accounting for 77%, 84%, 71%, and 71% of up-regulated DEGs as well as 64%, 65%, 59%, and 57% of down-regulated DEGs in the WT, dreb1c, dreb1g, and dreb1ceg, respectively (Supplementary Figure 5). These genes represent the majority of the DEGs in WT or dreb1 mutants, indicating that the dreb1 mutations do not drastically alter chilling-induced or chilling-repressed DEGs. At 4 h, 895 up-regulated and 565 down-regulated DEGs were identified in both dreb1g and dreb1ceg, and 874 up-regulated and 506 down-regulated DEGs were identified in both dreb1c and dreb1ceg (Supplementary Figure 5). At 24 h, 3,419 up-regulated and 2,960 down-regulated DEGs were identified in both dreb1g and dreb1ceg, and 2,962 up-regulated and 2,570 down-regulated DEGs were identified in both dreb1c and dreb1ceg (Supplementary Figure 5). The dreb1g mutant shared more DEGs with dreb1ceg than dreb1c, suggesting that OsDREB1G has a larger functional contribution than OsDREB1C to gene expression changes under chilling.

Considering that OsDREB1 genes are likely transcriptional activators, we examined genes that had a reduced expression in the dreb1ceg mutant under chilling because they are potentially regulatory target genes of OsDREB1C/E/G. These include chilling up-regulated DEGs in WT but not in the triple mutant as well as chilling down-regulated DEGs in dreb1ceg but not in WT at 4 h and/or 24 h. These DEGs were analyzed for enrichment of Gene Ontology (GO) terms defined by FDR < 0.05. No enriched GO terms were identified for the 428 up-regulated DEGs that appeared in WT but not in dreb1ceg (Supplementary Figure 6A). The 1,092 down-regulated DEGs that appeared in dreb1ceg but not in WT were enriched for the GO terms “plant-type hypersensitive response”, “protein maturation”, and “peptidyl-amino acid modification”, with “plant-type hypersensitive response” having the highest fold enrichment of 24 (Supplementary Figures 6B,C). Under the GO term “plant-type hypersensitive response” were four genes that had a lower expression at 24 h in dreb1ceg compared to the WT (Supplementary Figure 6D). These four genes code for a transporter, a ubiquitin-protein ligase, and two MACPF domain-containing protein. One MACPF domain containing protein CAD1 is a negative regulator of the cell death in Arabidopsis, and its mutant exhibited a hypersensitive response-like cell death (Morita-Yamamuro et al., 2005). The ubiquitin-protein ligase OsNLA1 modulates phosphate accumulation in rice, and the loss of its function leads to lethality (Zhong et al., 2017). This suggests that the LOF of OsDREB1C/E/G leads to reduced expression of potential negative regulators of cell death, which might contribute to the increased cell death in the dreb1ceg mutant under chilling.

Chilling downregulated 3,568, 3,534, 4,006, and 4,006 genes in WT, dreb1c, dreb1g, and dreb1ceg, respectively (Supplementary Figure 7A). GO analysis on downregulated DEGs in WT and the three dreb1 mutants was performed to identify differential GO terms between WT and the mutants. One GO term “hydrogen peroxide catabolic process” was present in all dreb1 mutants but not in WT (Supplementary Figure 7B). Under this enriched GO term were 18 chilling downregulated DEGs (1 peroxiredoxin-2E-1, 17 peroxidase) which had a lower expression in both WT and triple mutant at 24 h compared to 0 h. Under this GO term were also 10 chilling downregulated genes (one peroxiredoxin-2E-1, nine peroxidase) that had a lower expression in dreb1ceg compared to the WT at 24 h (Supplementary Figure 7C). Therefore, more genes that are involved in ROS scavenging have a lower expression in the dreb1 mutant compared to the wild type at 24 h. Plant peroxidase scavenges ROS and protects host cells from damages made by excessive ROS (Liu et al., 2021a). These results suggest that OsDREB1C/E/G may promote ROS scavenging under chilling.

As transcriptional factors, OsDREB1 proteins are expected to bind to specific cis-element motifs in the promoter region of their target genes and affect their gene expression. Because AtDREB1s mainly function as transcriptional activators (Song et al., 2021), we searched among genes with reduced expression in the dreb1ceg mutant compared to the WT for potential direct regulatory targets of OsDREB1C/E/G by the criterion of having DRE/CRT element(s) in their promoter region. We first examined the genes under GO terms uniquely enriched in the mutant but not the WT for chilling down regulated DEGs. Among the four genes under the GO term “plant-type hypersensitive response”, one gene (LOC_Os07g07194, MACPF domain-containing protein) contained a DRE/CRT (G/ACCGAC) element in its promoter region. Among the 18 genes under the term “hydrogen peroxide catabolic process”, three genes (LOC_Os03g22020, LOC_Os06g42000, and LOC_Os11g43980) contained one DRE/CRT element in their promoter regions. Therefore, these four genes are potential direct targets of OsDREB1C/E/G.

In addition, we examined genes with reduced expression in dreb1ceg compared to the WT at each of the three time points (0, 4, and 24 h) of chilling treatment. With 21, 38, and 145 downregulated DEGs (mutant versus WT) at 0, 4, and 24 h, respectively, the dreb1ceg mutant had a total of 168 downregulated DEGs compared to the WT with three time points combined (Figure 5A). These genes were defined as potential OsDREB1C/E/G target genes.

The potential function of these 168 potential OsDREB1C/E/G target genes was subject to GO analysis. Among the 95 genes that had a GO annotation in the biological process, no enrichment of GO terms was identified for these genes. However, many of these genes have shared GO terms. Seventy genes had the GO term “metabolic process”, including carbohydrate metabolism, protein metabolism, lipid metabolism, photosynthesis, organic acid metabolism, protein modification process and nucleobase, nucleoside, nucleotide, and nucleic acid metabolic process (Supplementary Figure 8). A total of 33, 16, 37, 7, and 14 genes had the GO terms of “response to stimulus”, “transport”, “cellular process”, “signal transduction”, and “developmental process”, respectively (Supplementary Figure 8). Sixteen and 26 genes had the GO terms of “response to abiotic stimulus”, and “response to stress”, respectively. Two genes were under the term “cell death,” which is a related term of “plant-type hypersensitive response”.

We further examined the expression pattern of these 168 potential OsDREB1C/E/G target genes to determine whether their regulation by OsDREB1C/E/G genes is time-point specific. Heat map analysis revealed that these genes, although identified from a specific time point, mostly had a reduced expression at all three time points (Figure 5B). For the 12 DEGs of dreb1ceg that are unique at 0 h, almost all also had a lower expression in dreb1c (11 genes) and dreb1g (12 genes) than in WT. Ten of these genes had a lower expression in the dreb1c, dreb1g, and dreb1ceg mutants compared to WT at 4 h, and most of them (nine for dreb1c, seven for dreb1g, and five for dreb1ceg) had a lower expression at 24 h. A similar low specificity for time points was also observed for DEGs at 4 and 24 h where they had lower expression at the other two time points in the mutants compared to the WT. In sum, the majority of these genes had decreased expression in the dreb1ceg mutants at all three time points although the fold change is less than 2 and were therefore were not identified as DEGs at all time points.

Next, we search for the presence of sequence motif G/ACCGAC, the DRE/CRT element, in the promoter region of the potential OsDREB1C/E/G target genes. Because some genes do not have defined transcription initiation sites, we used the 1.5 kb fragment 5′ to the translation start codon of these genes as a proxy for promoter fragment. Among the 168 genes, 53 genes contained at least one DRE/CRT element in their promoter region, with 15 genes containing more than one element (Supplementary Table 2). Eight genes have only one element within the 100 bp region 5′ to the translation start codon (Supplementary Table 2), and whether those are in the 5’ UTR region needs further investigation. Function of ten out of these 53 genes have been previously studied. Four of them, OsHAN1, OsDREB1F, OsNAP, and OsGNA1, have been shown to be involved in chilling tolerance or growth. The OsHAN1 gene codes for an oxidase, and it LOF mutant had enhanced chilling tolerance (Mao et al., 2019). OsDREB1F is member of DREB1 family, and its overexpression has been shown to enhance chilling tolerance in rice (Wang et al., 2008). It is interesting that OsDREB1F is a potential target of the OsDREB1C/E/G proteins in the same family. OsNAP encodes a NAC transcription factor, and its overexpression also enhances chilling tolerance rice (Chen et al., 2014). OsGNA1, encoding a glucosamine-6-phosphate acetyltransferase, and its LOF mutant has a temperature-sensitive defect in root elongation (Jiang et al., 2005). In addition, OsGRX2 encodes glutaredoxin that is involved in detoxification processes (Jeong et al., 2018) and could contribute to stress tolerance. Five other genes are involved in growth and developmental processes. OsDof18 is involved in growth and ammonium transport regulation (Wu et al., 2017), OsDjA7/8 is essential for chloroplast development (Zhu et al., 2015), OsNIA2, encoding a nitrate reductase, increases nitrogen uptake capacity (Sun et al., 2015), OsMYB103L is involved in GA-mediated regulation of secondary wall biosynthesis (Ye et al., 2015), and OsSPW1 controls floral organ identity (Nagasawa et al., 2003). Thus, these potential target genes may contribute to chilling tolerance and chilling growth mediated by OsDREB1C/E/G.

To determine whether these OsDREB1C/E/G genes also play a role in tolerance to other abiotic stresses in rice, the dreb1 mutants were subjected to heat, salt, and drought treatments. The survival rates for the WT NIP and the mutant dreb1 plants grown in the same pot were compared in the same pot to control variability of stress treatment.

For heat stress, 10-day-old seedlings grown on soil were treated at 48°C for 2 days, followed by a recovery growth. The NIP had a survival rate of 36%, while the dreb1ceg mutant plants all died (Figures 6A,B). These data indicate that OsDREB1C/E/G genes are positive regulators of tolerance to heat in rice.

For salt stress, 3-day-old seedlings were treated with 200 mM NaCl for 21 or 23 days. The survival rates for the four dreb1c mutants were 76%, 71%, 57%, and 61%, respectively, lower than the rate of 92% for NIP (Supplementary Figure 9A). No significant difference in salt tolerance was observed between dreb1e and NIP (Supplementary Figure 9B). The survival rate for dreb1g-1 was 74%, lower than the rate of 96% for NIP (Supplementary Figure 9C). The survival rate for dreb1ceg was 58%, lower than the rate of 93% for NIP (Figure 6C). We further compared salt tolerance of the single and triple mutants. The survival rates for dreb1c-1, dreb1g-1, and dreb1ceg were 42%, 41%, and 35%, respectively, all lower than the rate of 64% for NIP (Supplementary Figure 9D). However, no significant difference was observed between single and triple mutants (Supplementary Figure 9D). These results indicate that the OsDREB1C and OsDREB1G genes, but not the OsDREB1E gene, are positive regulators of salt tolerance in rice.

For drought stress, seedlings at three-leaf stage were withheld of water for 10 days and then re-watered for 7 days. No significant difference in survival rate was observed among dreb1c-2, dreb1e-1, and NIP (Supplementary Figures 10A,B). The survival rate for the dreb1g-1 mutant was 17%, lower than 53% for NIP (Supplementary Figure 10C). The survival rate for the dreb1ceg mutant was 29%, lower than 52% for NIP (Figure 6D). Compared drought tolerance between the dreb1g and dreb1ceg mutant, no significant difference was observed (Supplementary Figure 10D). These data indicate that the OsDREB1G gene but not OsDREB1C or OsDREB1E positively regulates drought tolerance in rice.

Accumulation of ROS and cell death was examined in the dreb1ceg mutant after each of the stress treatments: 24 h of 48°C, 13 days of 200 mM NaCl, and 6 days of water withheld. DAB staining revealed a higher H₂O₂ accumulation in leaves of the dreb1ceg mutant compared to that in NIP after all three stress treatments: heat, salt, and drought (Figure 6E). Similarly, leaves of the dreb1ceg mutant had more cell death than NIP after these stress treatments as indicated by trypan blue staining (Figure 6F). Therefore, OsDREB1C/E/G genes function to repress ROS overaccumulation and cell death under heat, salt, and drought stresses.