Drought tolerance is promoted by reducing the endogenous GA level in plants (Colebrook et al., 2014). Therefore, we hypothesized that the negative regulators of GA signaling, the DELLA proteins, may also play important roles in the drought tolerance of B. napus. In B. napus, there are four BnaRGAs (B. napus REPRESSOR OF ga1-3) homologs of Arabidopsis RGA gene (Zhao et al., 2017). To identify RGA genes that are responsive to drought and ABA, we examined their expression patterns under drought or ABA conditions using quantitative RT-PCR. After the drought treatment, the expression of BnaA6.RGA was greatly induced at 1 h, being three times higher than in the untreated tissue, and then its level slightly decreased at 3 h (Figure 1A). Compared with BnaA6.RGA, the expression levels of the BnaA9.RGA and BnaC9.RGA genes were also induced by drought, but the BnaC7.RGA was not significant changed (Figure 1A). Similarly, the expression of BnaA6.RGA was also greatly induced by an exogenous application of ABA, reaching its highest level at 1 h (Figure 1B), which was almost same drought treatment. In addition, the expressions of other BnaRGAs were up-regulated by the ABA treatment, although to a lesser extent than that of BnaA6.RGA (Figure 1B). Thus, the expression of BnaA6.RGA is more sensitive than those of other DELLA genes in response to drought and ABA treatments, which suggests that BnaA6.RGA plays more important roles in the drought tolerance of B. napus.

Then, the genome editing of BnaRGAs was performed using CRISPR/Cas9 (Yang et al., 2017). Two types of mutants were generated among the transgenic plants according to the genotyping results. L4 and L6 are gain-of-function mutants of BnaA6.RGA, designated bnaa6.rga-D. In L4, there was a 6-nt deletion at the sgRNA2 target site that caused a two-amino acid deletion in the TVHYNP motif (Yang et al., 2017). In L6, there were both a 9-nt deletion at the sgRNA1 target site that caused a three-amino acid deletion in the DELLA motif and a 12-nt deletion at the sgRNA2 target site that caused a four-amino acid deletion in the TVHYNP motif (Yang et al., 2017). All the other lines possessed loss-of-function mutations. For example, there was a 118-nt deletion in L2 and 1-nt insertion in L8 at the sgRNA1 target site that caused frameshifts in BnaA6.RGA, designated bnaa6.rga. These mutations in bnaa6.rga also reduced the transcript level of BnaA6.RGA significantly (Supplementary Figure S1). These mutants provided precious materials for investigating BnaA6.RGA’s roles in drought tolerance in B. napus.

To determine whether BnaA6.RGA regulates drought resistance, 3-week-old wild type (WT; Westar), and bnaa6.rga and bnaa6.rga-D mutants, grown in pots were subjected to drought stress by withholding water for 20 days and then re-watering the plants for 3 days. After 20 days of water deprivation, the relative soil water content was almost the same in each pot (Supplementary Figure S2). We found almost half of the WT plants wilted, but the bnaa6.rga-D plants remained turgid (Figure 2A). The leaf relative water content (RWC) was consistent with the drought phenotype, being 77.3–82.7% in bnaa6.rga-D plants, which was greater than in WT (54.6%) (Figure 2B). After 3 days of re-watering, bnaa6.rga-D plants recovered well, and the survival rates of the two bnaa6.rga-D lines reached 91.7 and 100%. In contrast, the WT survival rate was only 55.6% (Figure 2C), which suggested that bnaa6.rga-D is more tolerant to drought stress. However, the survival rate and leaf RWC of bnaa6.rga were not significantly different from those of WT (Figures 2A–C).

Ion leakage is an important indicator of cell injury. Then, the ion leakage was measured in irrigated versus water-deprived WT, bnaa6.rga-D and bnaa6.rga leaves. Under irrigated condition, no significant differences were found between the different lines (Figure 2D). After 10 days without irrigation, the ratio of ion leakage of bnaa6.rga-D was about 33.9%, which was less than WT (46.2%) (Figure 2D). After 20 days, the differences of ratio of ion leakage between bnaa6.rga-D and WT were much greater (Figure 2D), suggesting that membrane damage of WT was more serious than bnaa6.rga-D. Meanwhile, at the same drought condition, the ion leakage of bnaa6.rga was no different from WT (Figure 2D). Then, malondialdehyde (MDA) was measured in WT, bnaa6.rga-D and bnaa6.rga leaves, before or after the drought treatment. Compared to the WT, the MDA was less accumulated in bnaa6.rga-D, following exposure to the same drought condition (Figure 2E). These results indicated that bnaa6.rga-D could enhance the drought tolerance in B. napus.

Previously, the gain-of-function mutants of BnaC7.RGA, ds-3, was obtained by screening rapeseed EMS library (Zhao et al., 2017). A substitution of proline to leucine was identified in ds-3 in the conserved VHYNP motif, which is essential for GA-dependent interaction between GA receptor GID1 and DELLA proteins (Zhao et al., 2017). Then, we also tested the ds-3 in response to drought. ds-3 and WT (Huashang5, HS5) were subjected to the same drought treatment. Like bnaa6.rga-D, ds-3 plants remained turgid (Supplementary Figure S3A) after 3-week without watering, and the leaf RWC was about 95% (Supplementary Figure S3B). After 3 days of re-watering, ds-3 plants recovered well, and the survival rates was about 100% (Supplementary Figure S3C). The ion leakage and MDA content of ds-3 were less than WT (Supplementary Figures S3D,E). In contrast, the HS5 survival rate was almost 0% (Supplementary Figure S3C), which suggested that ds-3 is more tolerant to drought stress. Although BnaC7.RGA was not significant induced by drought or ABA, the gain-of-function mutants of BnaC7.RGA also showed strong drought tolerance, indicating that RGA proteins play the same functions in rapeseed drought tolerance.

Because there are four BnaRGAs, we hypothesized that the other BnaRGA genes may act redundantly during drought tolerance. To test this hypothesis, single mutants of the other BnaRGAs (bnac7.rga, L7; bnaa9.rga, L5 and L16; and bnac9.rga, L3) and the quadruple mutant bnarga, L27 and L46 (Yang et al., 2017), were subjected to the same drought treatment. Like bnaa6.rga, the survival rates were similar among the single mutants and WT (Supplementary Figure S4). In contrast, a majority of the bnarga plants wilted after three weeks without watering (Figure 3A). The leaf RWCs of bnarga plants ranged from 37.9 to 40.6% (Figure 3B), which was lower range than that of the WT. After re-watering, the two lines of bnarga plants, L27 and L46, recovered poorly, with survival rates of only 34.4 and 31.0%, respectively, indicating that bnarga is more sensitive to drought stress than WT (Figure 3C). After 10 days without irrigation, the ratio of ion leakage of bnarga was about 55–60%, which was more than WT (46.2%) (Figure 3D). After 20 days, the differences of ratio of ion leakage between bnarga and WT were much greater (Figure 3D), suggesting that membrane damage of bnarga was more serious than WT. The MDA was more accumulated in bnarga, following exposure to the same drought condition (Figure 3E). Thus, BnaA6.RGA and BnaC7.RGA positively regulates drought tolerance in B. napus, while other BnaRGAs have redundant functions in this process.

Reducing water loss is a key determinant of drought tolerance (Xiong et al., 2002). The BnaRGA mutants exhibited different sensitivities to drought stress. To uncover the underlying causes, weights of the detached leaves from these mutants were measured every half hour for 3 h after detachment. The bnarga quadruple mutants displayed significantly greater water loss rate, while bnaa6.rga-D and ds-3 showed a significantly lower water loss rate in comparison to that of WT (Figure 4A, Supplementary Figure S5A). Thus, the expression of BnaA6.RGA and BnaC7.RGA appears to reduce drought-induced water loss.

The drought stress induces ABA biosynthesis (McAdam and Brodribb, 2016), which, in turn, promotes stomatal closure. We first analyzed ABA levels in well-watered versus water-deprived WT, bnaa6.rga-D, and bnarga leaves. After 10 days without irrigation, the soil RWC of all plants reached round 40%. Under irrigation, no significant differences in ABA content were found between the different lines (Figure 4B). While the water deficit treatment increased ABA levels in all lines, which was no significant differences in different lines (Figure 4B). These results suggest that BnaRGAs does not promote ABA accumulation in B. napus leaves.

Measuring ABA-induced stomatal closure is a well-established assay for studying plant responses to drought stress (Mustilli et al., 2002). To evaluate whether BnaRGAs affects stomatal response to ABA, we treated peeled abaxial epidermal strips taken from bnaa6.rga-D, ds-3, bnarga, and their relative wildtype leaves with ABA treatment and monitored stomatal closure. The stomata of WT (Westar), bnaa6.rga-D and bnarga were induced to be wide open before the ABA treatment (Figures 4C,D). After the ABA treatment, the stomatal apertures of all these plants decreased; however, bnaa6.rga-D was more sensitive than WT, while the bnarga plants were less sensitive (Figures 4C,D). Specifically, 1 h after the ABA treatment, the average stomatal aperture (width/length) of the WT was 0.19, while those of the two bnaa6.rga-D were 0.16 to 0.17, and that of bnarga was 0.26 (Figures 3C,D, P < 0.01). At 2 h after the ABA treatment, the differences between stomatal apertures were much greater (Figures 4C,D). The same pattern was found between ds-3 and HS5 (Supplementary Figures S5A–C). Thus, the stomatal closures of the BnaRGA mutants in response to the ABA treatment were consistent with the wilting phenotypes under drought-stress conditions.

DELLA proteins usually play roles by interacting with other transcription factors (Van De Velde et al., 2017). In Arabidopsis, RGA can interact with the bZIP transcription factor ABI5 to regulate seed germination (Lim et al., 2013). The bZIP transcription factor BnaA10.ABF2 positively regulates plant drought tolerance when transformed into Arabidopsis (Zhao et al., 2016). Thus, we hypothesized that BnaA6.RGA might directly interact with BnaA10.ABF2. We used a yeast two-hybrid system to test this possibility, and it indicated that these two proteins interact (Figure 5A). A phylogenetic analysis revealed that the BnaA10.ABF2 protein belongs to the ABF2 clade of the ABRE family along with three other paralogs (Supplementary Figure S6). Three other BnaRGA proteins could also physically interact with BnaA10.ABF2 (Figure 5A). Then, bimolecular fluorescence complementation (BiFC) was performed to examine the interactions between BnaA10.ABF2 and the four BnaRGA proteins. The YFP fluorescence occurred specifically in the nucleus only when BnaA10.ABF2-cYFP and each BnaRGA-nYFP were expressed simultaneously in tobacco leaves (Figure 5B). Without the GRAS domain, BnaA6.RGA and BnaC7.RGA could not interact with BnaA10.ABF2 (Supplementary Figure S7). These results suggest that BnaA10.ABF2 physically interacted with BnaRGAs to form a protein complex.

Because BnaRGA proteins physically interact with BnaA10.ABF2, we speculated that BnaRGAs may affect the expression of ABF2’s downstream genes. To investigate this possibility, the expression levels of its downstream genes were examined in bnaA6.rga-D, bnarga and WT after the drought or ABA treatment. After the drought treatment, the expression of BnaRAB18 was about 12.8–841.8 times induced in bnaa6.rga-D (Figure 6A), which was much greater than in WT. Under the same conditions, the expression of BnaRAB18 was induced to a lesser extent in bnarga (Figure 6A). Similarly, BnaRD29A and BnaRD29B were also significantly up-regulated in bnaa6.rga-D and up-regulated to a lesser extent in bnarga compared with WT (Figures 6B,C). After the ABA treatment, these three downstream genes were all greatly induced in bnaa6.rga-D, but were induced to a significantly lesser extent in bnarga (Figures 6D–F). The same results were observed in ds-3 after drought treatment (Supplementary Figures S8A–C). Thus, BnaA6.RGA and BnaC7.RGA may promote drought tolerance by enhancing the expression of BnaA10.ABF2’s downstream genes.

In Arabidopsis, ABF2 activates expression of RAB18 (Fujita et al., 2005). Therefore, we used the dual luciferase reporter assay to determine whether BnaA10.ABF2 could promote the expression of BnaRAB18 in B. napus. The reporter vector contained a firefly luciferase gene driven by the BnaC9.RAB18 promoter and a renilla luciferase (REN) gene driven by the CaMV 35S promoter. BnaA6.RGA, BnaC7.RGA and BnaA10.ABF2 were each driven by the CaMV 35S promoter in independent effector vectors (Figure 7A). The LUC activity was measured for different combinations. Compared with the expression of pBnaC9.RAB18-LUC only, the LUC enzyme’s activity level was 3.7 times greater when BnaA10.ABF2 was co-expressed (Figure 7B). Furthermore, when BnaA6.RGA or BnaC7.RGA was expressed together with BnaA10.ABF2, the activity of pBnaC9.RAB18-LUC was increased by 6.5 or 11.8 times, respectively (Figure 7B). Without the GRAS domain, BnaA6.RGA and BnaC7.RGA could not interact with BnaA10.ABF2, and the expression levels of the reporter genes were less enhanced (Supplementary Figure S9). Collectively, these results suggested that BnaA6.RGA and BnaC7.RGA may form a complex with BnaA10.ABF2 to promote the expression of BnaC9.RAB18.

In Arabidopsis, the RGA protein promotes the expression of XERICO, an E3 ligase, which promotes destabilization of the ABA catabolic gene to antagonize GA effects (Zentella et al., 2007). After the drought treatment, the expression of BnaXERICO was 6.1 to 18.2 times induced in bnaa6.rga-D (Supplementary Figure S10A), which was much greater than in WT. Under the same conditions, the expression of BnaXERICO was not significantly induced in bnarga (Supplementary Figure S10A). Similarly, BnaA10.ABF2 and BnaC6.ABF2 (a paralog of BnaABF2) were also significantly up-regulated in bnaa6.rga-D and up-regulated to a lesser extent in bnarga (Supplementary Figure S10B). Thus, DELLA proteins appear to positively regulate ABA metabolic- and signaling-related genes.

In plants, DELLA proteins are degraded by GIBBERELLIN INSENSITIVE DWARF1 (GID1)-mediated pathways (Ueguchi-Tanaka et al., 2007). Therefore, we speculated that these GA signaling genes would be responsive to drought. To investigate this possibility, the expression levels of GA signaling genes were examined in bnaA6.rga-D, bnarga, and WT after the drought treatment. At 1 h after the drought treatment, the expression levels of BnaGID1a, and BnaGID1c, and BnaSLY1 were greatly reduced, and then increased (Supplementary Figure S11), suggesting that GA negatively regulates plant drought tolerance by decreasing the stabilities of DELLA proteins. The expression of BnaGID1a was about 1.5–6.9 times induced in bnaa6.rga-D after drought treatment (Supplementary Figure S11A), which was much greater than in WT. Under the same conditions, the expression of BnaGID1a was repressed in bnarga (Supplementary Figure 11A). Similarly, BnaGID1b were also significantly up-regulated in bnaa6.rga-D and suppressed in bnarga compared with WT (Supplementary Figure S11B). However, the expression pattern of BnaSLY1 was no difference among bnaA6.rga-D, bnarga, and WT after the drought treatment (Supplementary Figure S11C). The expression levels of BnaGID1a and BnaGID1c were suppressed, and then increased, indicating that the feedback regulation between DELLA and GID1 could promote the latter’s transcription.

Under water-deficit conditions, ABA and other small molecules rapidly accumulate, which confers drought tolerance in B. napus (Zhu et al., 2010). Sequence analyses have identified all the ABA biosynthesis and signaling components in B. napus, indicating that this pathway is conserved in this species (Zhu et al., 2016). Although ABA pathway genes, as well as some other stress responsive genes, are up-regulated by drought stress in B. napus (Li et al., 2005; Zhu et al., 2010), whether these genes play important roles during drought tolerance remains obscure owing to the lack of genetic evidences. Here, our results indicated that BnaRGA proteins, key repressors of GA signaling, promoted drought tolerance by interacting with BnaABF2 in B. napus. The identification of these GA genes provided us insights into the regulatory mechanisms of drought resistance in B. napus.

Gibberellin mediates various developmental processes throughout the life cycle of the plant, and a GA deficiency results in severe dwarfism (Ueguchi-Tanaka et al., 2007). In addition to promoting plant growth and development, GA biosynthesis or signaling is involved in modulating plant abiotic resistance to stresses, such as drought, salinity, and other environmental stimuli (Magome et al., 2008; Colebrook et al., 2014; Nir et al., 2014, 2017). Specifically, the GA deficient mutants ga20ox1/2 and ga3ox1/2 are more resistant to drought in Arabidopsis (Colebrook et al., 2014). Overexpressing the GAMT gene, which encodes a GA methyltransferase, enhances tolerance to water-deficit stress in tomato (Nir et al., 2014). The loss of DELLA encoding genes, such as the mutant procera in tomato (Nir et al., 2017), results in rapid water loss under water-deficit conditions, and the daily transpiration level is decreased in the tomato GID1 double mutant (Illouz-Eliaz et al., 2019). Although GA is critical for stress responses in Arabidopsis and tomato, there are no reports on its roles in B. napus. We found that gain-of-function of BnaA6.RGA and BnaC7.RGA mutants (bnaa6.rga-D and ds-3) were more tolerant to drought. However, the loss of the four BnaRGA genes (bnarga) led to a hypersensitivity to drought (Figure 3), suggesting that the functions of GA pathway genes in plant drought tolerance are conserved.

DELLA family proteins are key repressors in GA signal transduction (Van De Velde et al., 2017). When the GA level is increased, DELLA proteins interact with GID1 and are degraded by the ubiquitin-dependent proteasome pathway to promote the expression of GA-responsive genes (Ueguchi-Tanaka et al., 2007). The expression levels of the positive regulators of GA signaling, such as GA receptor (GID1a/c) and F-box (SCF^SLY1/2) genes, were greatly reduced by drought treatments, consistent with the results in Arabidopsis (Kilian et al., 2007), which might stimulate DELLA protein accumulation. However, expression levels of GID1 were first reduced and then increased, which was depended on BnaRGA (Supplementary Figure S11), indicating that the feedback regulation between DELLA and GID1 could promote the latter’s transcription, as reported in Arabidopsis (Zentella et al., 2007). Therefore, we speculate that GA negatively regulates plant drought tolerance by suppressing the stability levels of DELLA proteins.

As an allotetraploid species, a majority of genes in B. napus are present in multiple copies that share high sequence similarities. BnaA6.RGA, BnaA9.RGA, and BnaC9.RGA were induced by drought and ABA (Figure 1), suggesting that other DELLA proteins regulate drought responses together with BnaA6.RGA. Consistent with this hypothesis, no significant differences were found between the single mutants of the four BnaRGAs and WT after the drought treatment, while the quadruple mutant bnarga significantly decreased the sensitivity of stomatal closure. These findings indicate that RGA proteins function redundantly in enhancing drought tolerance in B. napus. Although the expression pattern was quite different among the four paralogs of BnaRGA after the drought treatment, protein interaction assay indicated that all four BnaRGAs could interact with BnaA10.ABF2, and ds-3 also displayed more drought tolerance than WT, suggesting different BnaRGA-BnaA10.ABF2 complex might co-activate drought related genes in different tissues or at the different time point.

Abscisic acid plays important roles in drought tolerance. ABA-mediated stomatal closure is involved in drought tolerance (Xiong et al., 2002). The ABA-induced stomatal closure rate was accelerated in bnaa6.rga-D, suggesting that ABA’s response was enhanced. However, the effects of DELLA on stomatal aperture are strongly suppressed in ABA-deficient mutants (Nir et al., 2017), indicating that these functions are ABA-dependent. DELLA proteins belong to a subgroup of the GRAS transcription factor family, which lack a DNA-binding domain (Yoshida et al., 2014). Thus, DELLA proteins usually function by interacting with other DNA-binding proteins to regulate the transcriptional activity of downstream genes (Van De Velde et al., 2017). In this way, DELLA proteins not only promote ABA biosynthesis (Zentella et al., 2007; Piskurewicz et al., 2008) but also the signaling components (Lim et al., 2013). Here, BnaA10.ABF2 interacted directly with each of the four BnaRGAs (Figure 4), providing a direct link between the ABA and GA signaling pathways during drought tolerance. Moreover, this protein interaction enhanced the expression of downstream drought-responsive genes (Figure 7). These results revealed a regulatory mechanism underlying drought tolerance in B. napus.

In Arabidopsis, the RGA protein promotes the expression of XERICO which destabilized the ABA catabolic genes (Zentella et al., 2007). Similarly, we found that the expression of BnaXERICO was up-regulated in bnaa6.rga-D and down-regulated in bnarga (Supplementary Figure S10A). Additionally, the ABA signaling genes BnaA10.ABF2 and BnaC6.ABF2 were also induced in bnaa6.rga-D (Supplementary Figure S10B). These results suggested that DELLA proteins positively regulate stomatal movement by up-regulating ABA signaling-related genes and down-regulating ABA catabolic genes. After the long-term drought treatment, drought induces similar levels of ABA accumulation in bnaa6.rga-D and bnarga lines, indicating that drought induced ABA accumulation might include BnaRGA-dependent and -independent way.

The N-terminal DELLA domain is important for the degradation of these proteins. Plants constitutively expressing a truncated DELLA, lacking the DELLA or TVHYNP motif, have growth defects that mimic a GA shortage, such as severe dwarfism, which is significantly reduced the crop yield (Fleet and Sun, 2005; Ueguchi-Tanaka et al., 2007; Zhao et al., 2017). There are several strategies have been developed for generating transgenic plants which is improved one agronomic trait without affecting others. For example, expression levels of PRO driven by guard cell-specific promoters greatly increase tomato drought tolerance but have no obvious effects on plant growth (Nir et al., 2017). The other strategy uses CRISPR/Cas9 genome editing toolkits to generate diverse cis-regulatory alleles that provide beneficial quantitative variation for breeding (Rodriguez-Leal et al., 2017). These reports provide new strategies to generate drought-tolerant crops and speed up molecular breeding in the future.

In summary, our findings indicate that BnaRGA proteins play important roles in plant adaptation to water-deficit stress. We proposed a working model for BnaA6.RGA-mediated drought tolerance in B. napus (Figure 8). Drought leads to an ABA accumulation and a GA reduction, which induces the expression of ABF2 and promotes the RGA accumulation. The BnaA6.RGA-BnaABF2 complex enhances the expression of drought responsive genes (such as BnaRAB18), resulting in enhanced ABA signaling, which then increases the plant’s drought resistance. In addition, the accumulated BnaA6.RGA might enhance guard cell sensitivity to ABA by an unknown mechanism, which leads to rapid stomatal closure. Both long term and short term of ABA responses enhance plant drought resistance (Figure 8). Our findings provide novel insights into the crosstalk between GA and ABA signaling pathways, and the editing resources obtained in our study provide desirable germplasm for further breeding of drought tolerance in rapeseed.

The single mutants, bnaa6.rga, bnac7.rga, bnaa9.rga, bnac9.rga, quadruple mutant bnarga, and gain-of-function mutant bnaa6.rga-D were generated by the CRISPR/Cas9 technology in B. napus (Westar) (Yang et al., 2017). The gain-of-function mutants of BnaC7.RGA, ds-3, was obtained by screening rapeseed EMS library (Zhao et al., 2017). To germinate, the seeds were soaked in water for 7 days and then placed in soil. All plants were cultivated in a growth room under a light intensity of 120 μmol m^–2 s^–1 with a 16/8 h light/dark photoperiod (23 ± 1°C and 60–70% RH).

Each gene was amplified from the B. napus cDNA library by PCR using gene-specific primers. For the yeast two hybrid assay, the BnaA10.ABF2 coding region was cloned into the vector pGBKT7 as the bait (Clontech, United States), while the constructs of pGADT7-BnaA6/A9/C7/C9.RGA serving as the prey were obtained from the previous report (Zhao et al., 2017). For the BiFC assay, the full-length coding regions of BnaRGAs, and the N terminus of BnaA6.RGA (1–148 amino acid)/BnaC7.RGA (1–132 amino acid), and the C terminus of BnaA6.RGA (149–572 amino acid)/BnaC7.RGA (133–556 amino acid) were cloned into the vector pFGC-YN173 fused with the N terminus of YFP, and BnaA10.ABF2 was inserted into the vector pFGC-YC155 fused with the C terminus of YFP. For the dual luciferase assay, the full-length coding regions of BnaA6.RGA, BnaC7.RGA, BnaA10.ABF2, and the N terminus of BnaA6.RGA (1–148 amino acid)/BnaC7.RGA (1–132 amino acid) were inserted into the vector pRI101-GFP as the effectors, while the 1,496 bp promoter of BnaC9.RAB18 was cloned into the vector pGreenII 0800-LUC as the reporter. All primers are listed in Supplementary Table S1.

To measure the water loss, the leaves were detached from 3-week-old plants and placed in a petri dish on a laboratory bench (23 ± 1°C and 30–40% RH), and the weight of the detached leaves was then measured every 0.5 h for a period of 3 h. The experiment was repeated three times. For each repeat, at least five leaves from different plants of each genotype were used. Water loss was presented as the percentage of the fresh weight (FW) loss.

One-week-old plants were transferred from the Hoagland medium to the same weight of water-saturated soil for growing in the greenhouse for 2 weeks, and then deprived of water for 20 days. Each pot contained 130 ± 0.1 g (dry weight, DW) of soil and one plant. The pots were irrigated with water to saturation, allowed to drain, and then weighted to obtain the initial weight, after which they were subjected to drought for different time periods. The relative soil water content (SWC) was calculated as: SWC = (final FW − DW)/(initial weight − DW) × 100. Subsequently, the survival rates of plants were determined after three days of re-watering (rehydration).

The RWC of fully expanded leaves from 3-week-old plants grown in pots was measured after 3 weeks of drought treatment. Frist, leaves were removed and immediately weighted to obtain the FW. The leaves were then placed into petri dishes filled with ddH₂O. After 24 h, the leaves were blotted to remove external water, and then weighted to obtain the leaf turgid weight (TW). Finally, leaves were dried to a constant weight at 60°C and then weighted to obtain the leaf DW. The RWC of leaves was calculated as: RWC = (FW − DW)/(TW − DW) × 100.

Stomatal assays were performed as previously described (Desikan et al., 2002). In brief, five fully expanded leaves were harvested from different 3-week-old plants for each genotype and incubated in the MES-KCl buffer (50 mM KCl, 10 mM CaCl₂, 10 mM MES-KOH, pH 6.15) under light at 22°C for 3 h. Once the stomata were fully open, leaves were incubated in the MES-KCl buffer with 1 μM of ABA or DMSO (control). After 1–2 h of treatment, the epidermal strips were immediately peeled from the abaxial side of leaves. After that, 30–40 stomata from one leaf were measured, adding up to 150–200 stomata for each genotype. The pictures were taken by microscope equipped with a digital camera (AxioCam ICc5, Zeiss). Stomatal aperture was analyzed using the software Image J. Three independent repeats were performed.

Abscisic acid contents were measured as previously described (Ding et al., 2011). In brief, leaves were harvested from plants that were irrigated or non-irrigated for 10 days. The leaves (about 200 mg) were ground in liquid N₂ and homogenized in 90% (v/v) methanol containing 200 mg L^–1 of diethydithiocarbamic acid sodium salt. The extracts were then incubated overnight in darkness at 4°C, followed by a 8,000 × g centrifugation at 4°C. The methanolic supernatant was vacuumed centrifuge at 4°C to evaporate the supernatant, and the residue was dissolved by methanolic Tris buffer (10% methanol, 50 mM Tris, pH 8.0, 1 mM MgCl₂, and 150 mM NaCl). An ELISA kit was used for the determination of ABA following the manufacturer’s instructions (Agdia¹). Three independent repeats were performed.

The membrane ion leakage was measured as described previously with slight modifications (Lee et al., 2011). In brief, five fully expanded leaves, that were irrigated or non-irrigated for 10 or 20 days, were harvested from bnaa6.rga-D, ds-3, bnarga, and their relative wild-type plants. Then, a hole puncher was used to obtain 60 leaf disks from each leaf. The leaf disks were transferred to glass test tubes filled with 10 mL of ddH₂O, and vacuumed until the leaf disks were submerged by water. After that, the tubes were shaken on an oscillator for 1 h. Electrical conductivity was measured in accordance with the manufacturer’s instructions (BANTE, China). The initial conductivity was measured as R1. Total conductivity was determined after boiling for 10 min as R2. The relative ion leakage was calculated by using the following formula: R1/R2 × 100%.

Malondialdehyde content, expressed as units/mg protein, was measured using analytical kits (Cat#A003-2, Nanjing Jiancheng Bioengineering Institute, Nanjing, China). In all experiments, three independent repeats were performed.

For transient ABA or drought treatment, the 3-week-old seedlings were transferred into Hoagland medium supplemented with 100 μM ABA or placed in a petri dish on a laboratory bench, respectively. Then the samples were collected at indicated time point, and were frozen immediately by liquid nitrogen. Total RNA was extracted using a Plant Total RNA Isolation Kit (Sangon Biotech, Shanghai, China, No. SK8631) following the manufacturer’s instructions. Approximately 1 μg of total RNA was used for cDNA synthesis using a PrimeScript^TM RT reagent kit (TaKaRa, Japan, Cat#RR047A). For qPCR, a total volume of 10 μL reaction mixture was used containing 5 μL of 2 × SYBR Green Master Mix (BioRad, United States), 0.5 μL of 5 × diluted cDNA, 0.25 μL of each primer, and 4 μL of ddH₂O. Amplification was performed using a CFX Connect^TM system (Bio-rad, United States). The amplification program consisted of one cycle of 95°C for 5 min, followed by 50 cycles of 95°C for 15 s, 60°C for 20 s, and 72°C for 20 s. The fluorescent product was detected at the third step of each cycle. The expression level of each gene was calculated using the 2^–ΔΔCT method (Livak and Schmittgen, 2001). All analyses were repeated three times using biological replicates. The gene BnaGAPDH (BnaC05g12400D) was used as the internal control. All primers are listed in Supplementary Table S2.

The yeast Gal4 system was employed for two-hybrid analysis of BnaRGAs and BnaA10.ABF2 protein interactions following the yeast transformation handbook (Yeast Transformation System 2; Clontech, United States). In brief, a single colony of yeast AH109 was incubated at 30°C overnight. The cells were harvested by centrifugation and then resuspended in 25 mL ddH₂O. Then the cell pellets were dissolved into 1.5-mL sterile 1 × Tris/LiAc solution to make the competent cells. For the yeast two-hybrid assay, the bait (0.5 μg) and/or prey (0.5 μg) plasmids with 0.1 mg of carrier DNA were co-transformed into the yeast-competent cells using polyethylene glycol/LiAc solution. After incubation at 30°C for 30 min, 70 μL DMSO (dimethyl sulfoxide) was added and incubation continued at 42°C for 15 min. The cells were centrifuged and washed using ddH₂O. The presence of transgenes in yeast cells was confirmed by growing these cells on plates containing solid synthetic defined (SD) medium lacking Leu and Trp (SD/−2). To assess protein–protein interactions, transformed yeast cells were re-suspended in ddH₂O to an optical density at OD₆₀₀ of 1.0. Samples (5 μL) of suspended yeast cells were spread on plates containing SD medium lacking Ade, His, Leu, and Trp (SD/−4). To detect protein–protein interactions, plates were examined after 3 days of incubation at 30°C. For each experiment, a total of 10 clones were selected and tested.

The BiFC assay was performed as previously described (Koo et al., 2017). In brief, the Agrobacterium tumefaciens (strain GV3101) cells containing the desired constructs were injected into the 4- to 5-week-old Nicotiana benthamiana leaves with the buffer (10 mM MES, pH = 5.7, 10 mM MgCl₂, and 150 μM acetosyringone). The final concentrations of the bacteria were adjusted to an OD₆₀₀ = 0.2 for each construct. The transiently transformed leaves were analyzed at 48 h after injection. The YFP fluorescence (excitation/emission wavelength: 514 nm/527 nm) was observed under a fluorescence microscope (SPX8, Leica).

Dual-luciferase assays were performed by using the Dual-Luciferase_® Reporter Assay System (Promega, Madison, WI, United States). All reagents were prepared as described by the manufacturer. Briefly, the agrobacteria harboring the reporter and effector vectors were injected into the tobacco leaves. After 3 days, leaf disks in a diameter of 2 cm were harvested and ground, and dissolved into 100 μL of Passive Lysis Buffer. After 30 s, a 50 μL aliquot was used for luminescence measurements with the SPARK^® MULTIMODE MICROPLATE (TECAN, Swiss). The following steps were used for luminescence measurements: 50 μL of the firefly luciferase reagent (LARII) was added to the test sample, with a 10 s equilibration time and measurement of luminescence with a 10 s integration time, followed by addition of 50 μl of the REN reagent and firefly quenching (Stop and Glow^TM buffer), 10 s equilibration time, and measurement of luminescence with a 10 s integration time. The data are represented as the ratio of firefly to Renilla luciferase activity (Fluc/Rluc). Each data point consisted of at least three biological replicates, and 10 repeats were performed for each assay.

The protein sequences were obtained from the website² (Jin et al., 2017). The sequence alignment was performed using Clustal Omega.³ An unrooted phylogenetic tree was constructed using MEGA7⁴ (Kumar et al., 2016) with the neighbor-joining statistical method and bootstrap analysis (1,000 replicates).

Statistical analyses were performed to determine significant differences between genotypes, using Duncan’s multiple range tests in the ANOVA program of SPSS (IBM SPSS 22). at P values <0.05 or <0.01.