Silencing of OsGRXS17 in rice improves drought stress tolerance by modulating ROS accumulation and stomatal closure

Glutaredoxins (GRXs) modulate redox-dependent signaling pathways and have emerged as key mediators in plant responses to environmental stimuli. Here we report that RNAi-mediated suppression of Oryza sativa GRXS17 (OsGRXS17) improved drought tolerance in rice. Gene expression studies showed that OsGRXS17 was present throughout the plant and that transcript abundance increased in response to drought stress and abscisic acid (ABA) treatment. Localization studies, utilizing GFP-OsGRXS17 fusion proteins, indicated that OsGRXS17 resides in both the cytoplasm and the nuclear envelope. Under drought stress conditions, rice plants with reduced OsGRXS17 expression showed lower rates of water loss and stomatal conductance, higher relative water content, and enhanced survival compared to wild-type controls. Further characterization of the OsGRXS17 down-regulated plants revealed an elevation in H2O2 production within the guard cells, increased sensitivity to ABA, and a reduction in stomatal apertures. The findings demonstrate a critical link between OsGRXS17, the modulation of guard cell H2O2 concentrations, and stomatal closure, expanding our understanding of the mechanisms governing plant responses to drought.

To mitigate cellular damage, plants have evolved a versatile ROS scavenging system that functions in conjunction with the mechanisms regulating ROS production to control cellular ROS concentrations [14][15][16] . Glutaredoxins (GRXs) are ubiquitous oxidoreductases in the thioredoxin (TRX) family and are involved in maintaining cellular redox homeostasis and regulating the redox-dependent signal pathway. GRXs utilize the reducing power of glutathione (GSH) to catalyze reversible reduction of disulfide bonds of the cognate target proteins [19][20][21] . GRXs also act as redox regulators in different aspects of plant growth, such as iron homeostasis, heavy metal detoxification, plant development, and plant-pathogen interaction 19,21,22 . Genomic sequence data have revealed families of GRX genes in higher plants, including Arabidopsis thaliana with fifty predicted GRX genes, Populus trichocarpa with thirty-six, and Oryza sativa with twenty-seven [23][24][25] . Based on the predicted active sites, plant GRXs can be subdivided into four groups 26,27 . GRXs of class I and class II have so-called CxxC/S and CGFS active sites, respectively, and are conserved in all photosynthetic organisms. Class III GRXs are specific to higher plants and have a peculiar CCxx active site. GRXs of class IV harbor a CxDC/S active site. Many studies have demonstrated Class II GRXs involvement in stress adaptation of plants [28][29][30][31][32] .
GRXS17 is a Class II GRX that plays an essential role in chilling stress, heat stress, and photoperiod responses [33][34][35][36] . However, a function for GRXS17 in drought stress response has not been investigated. In this study, we examined the spatial expression of OsGRXS17 in rice plants under normal and drought stress conditions and used OsGRXS17 silenced rice plants to investigate a role for OsGRXS17 in drought stress responses.

OsGRXS17 is a Functional Homolog of AtGRXS17. Examination of the Rice Functional Genomic
Express Database (http://signal.salk.edu/cgi-bin/RiceGE) revealed a rice homolog, OsGRXS17 (Os10g35720), of the Arabidopsis AtGRXS17 with 67.8% identity at the amino acid level (Fig. S1). OsGRXS17 (Os10g35720.1, long isoform) consists of 491 amino acid residues with an N-terminal TRX-like homology domain (HD) and three tandem GRX-HDs (Fig. 1a,b). Similar homologs were detected in maize, tomato, and potato (Fig. S1). Evidence was found in rice for an alternatively spliced short isoform (Os10g35720.2), which encodes a protein of 384 amino acids and consists of the TRX-like HD and only two monothiol GRX-HDs (Fig. 1a,b). In yeast, grx3grx4 double mutants, when exposed to H 2 O 2 , show markedly reduced cell growth, which can be recovered by expression of the Arabidopsis AtGRXS17 34 . The long isoform of OsGRXS17 showed similar properties by restoring growth in the presence of H 2 O 2 when expressed in the double-mutant strain (Fig. 2a). Arabidopsis T-DNA insertion mutant (atgrxs17) lines are defective in vegetative growth and development and sensitive to elevated temperature stress 35 . OsGRXS17 complements the loss of AtGRXS17 in the atgrxs17 line, rescuing defective mutant phenotypes under heat stress (Fig. 2b). When expressed as a GFP fusion protein GFP-OsGRXS17 is detected in both the cytoplasm and nuclei of tobacco leaf epidermal cells and rice protoplasts (Fig. 2c), similar to the results that were seen with AtGRXS17 33,34 .
Expression Pattern of OsGRXS17 in Response to Stress and Plant Hormones. Although the gene was annotated to have two alternative splicing variants, we were only able to detect the long isoform in the shoots of the 14-day-old seedlings that we studied. Thus, all the quantitative RT-PCR (qRT-PCR) data presented here are about the long isoform. The expression pattern of the OsGRXS17 gene, as determined by qRT-PCR, indicated that leaves and roots had higher expression levels than either stems or young panicles (Fig. 3a). The expression pattern of OsGRXS17 in rice shoots was also examined under reduced watering (Fig. 3b), in the presence of polyethylene glycol (PEG) (Fig. 3c), after application of ABA (Fig. 3d), indole-3-acetic acid (IAA) (Fig. 3e), and high salinity treatments (Fig. 3f). OsGRXS17 transcript levels were elevated 2 h after the start of drought stress treatment and continued to increase until 8 h, at which time the transcript levels plateaued (Fig. 3b). PEG or ABA treatments led OsGRXS17 rescues the hypersensitivity of yeast grx3grx4 mutant and defective phenotypes of Arabidopsis atgrxs17 mutant. (a) Vector control-, AtGRXS17-and OsGRXS17-expressing grx3/grx4 cells were grown on YPD, SC-Ura or SC-Ura +1 mM H 2 O 2 media for 3 days at 30 °C. OsGRXS17 was able to suppress the sensitivity of grx3grx4 cells to oxidative stress. AtGRXS17 was used as a positive control here. (b) Rice OsGRXS17 suppressed atgrxs17 KO mutant phenotypes grown under elevated temperature (28 °C), while atgrxs17 KO seedlings displayed short primary roots and the growth of seedlings was inhibited when grown under 28 °C. (c) Transient expression of GFP-OsGRXS17 and free GFP in tobacco epidermal cells and rice protoplast. Scale bars = 50 µm. As expected the free GFP control was localized in both the cytoplasm and the nuclei. A vector harboring 35S::mCherry::NLS (the mCherry red fluorescent protein linked to a nuclear localization signal) was used as a control for nuclear localization in transient co-expression assays (Fig. 2c to high levels of expression of OsGRXS17 after 24 h (Fig. 3c,d). OsGRXS17 levels transiently increased 2 h after treatments with IAA or salt (Fig. 3e,f).

Generation of OsGRXS17
Silenced RNAi Rice Plants. Expression of OsGRXS17 was altered by expression of inhibitory short hairpin RNAs derived from OsGRXS17 gene under the control of the maize ubiquitin (Ubi) promoter. The construct was transformed into Oryza sativa L. Japonica cv. Nipponbare, self-pollinated, and the progeny lines were genotyped for the presence of T-DNA (Fig. 4a). The copy number of T-DNA insertions of those transgenic lines was determined by Southern blot analysis using a hygromycin (hpt) gene-specific probe (Fig. S2). Lines OsGRXS17 RNAi-6 and OsGRXS17 RNAi-7 contain a single transgene insertion, while OsGRXS17 RNAi-8 and OsGRXS17 RNAi-13 had multiple integration events (Fig. S2). Expression levels of OsGRXS17 in lines OsGRXS17 RNAi-6, -7, -8 and -13 fell into two groups as measured by qRT-PCR (Fig. 4b). Lines OsGRXS17 RNAi-7, -8 and -13 had reduced levels of OsGRXS17 expression at ~80-90% compared to the wild-type plants, while OsGRXS17 RNAi-6 had reduced levels of OsGRXS17 expression at ~25% compared to the wild-type plants.
To examine if the expression of other endogenous Oryza sativa GRX genes, which are the most closely related to OsGRXS17, is altered in the OsGRXS17 RNAi lines, the expression levels of four endogenous OsGRX genes (subgroup II OsGRXS14, OsGRXS15.1, OsGRXS15.2, and OsGRXS16) identified by phylogenetic analysis (Fig. S3) and amino acid sequence alignment (Fig. S4a) were measured by qRT-PCR. No significant differences in the expression level were found between wild-type and OsGRXS17 RNAi rice plants (Fig. S4b). The phenotypes of the OsGRXS17 silenced plants were indistinguishable from wild-type plants at both vegetative and reproductive stages under normal growth conditions (Fig. 4c,d). The agronomic traits analyses, including panicle number per plant, main panicle length, grain number of main panicle, seed set percentage of main panicle, 100-grain weight of main panicle and panicle weight per plant, indicated no differences between OsGRXS17 silenced and wild-type rice plants (Fig. S5).

Silenced Expression of OsGRXS17
Enhances Drought Stress Tolerance in Rice. T2 homozygous plants of the four OsGRXS17 silenced lines and wild-type plants were subjected to drought stress treatment. The lines were visually indistinguishable before withholding water (Fig. 5a, 0 d). After withholding water for 9 days, leaves of OsGRXS17 RNAi-7, -8 and -13 rice plants remained green and turgid, while wild-type and OsGRXS17 RNAi-6 rice plants wilted (Fig. 5a, 9 and 10 d). At 11 days, all plants displayed wilting (Fig. 5a, 11 d). Wild-type and OsGRXS17 silenced rice plants were re-watered after 11 days to test recovery from drought stress treatment (Fig. 5a, 17 d with 6 days of re-watering). Six days after the new watering regimen, the percentage of plants that visibly recovered on the basis of new green leaf production was measured (Fig. 5a, 17 d with 6 days of re-watering and Fig. 5b). Silenced plants showed recovery (survival) rates in a range of 23-75%, while 8% of the wild-type plants survived (Fig. 5b). Further, lines OsGRXS17 RNAi-7, -8 and -13 had higher survival rates than OsGRXS17 RNAi-6. To evaluate the physiological responses associated with drought tolerance in OsGRXS17 silenced plants, water loss in leaves of wild-type and OsGRXS17 silenced plants was measured. Under water deficit conditions, all silenced plant lines showed lower water loss as compared to wild-type plants at 3 h (Fig. 5c). By 4 h, water loss rates of OsGRXS17 RNAi-7, -8, and -13 were lower than wild-type and OsGRXS17 RNAi-6 ( Fig. 5c). The relative water content of all silenced lines was greater at 5 days of water deprivation in comparison to wild-type, although the water content of OsGRXS17 RNAi-6 was intermediate to wild-type and the other silenced lines (Fig. 5d). At the same time, stomatal conductance was lower in OsGRXS17 RNAi-7, -8, and -13 compared to wild-type and OsGRXS17 RNAi-6 plants under both normal (Fig. 5e, 0 d) and water-deficit conditions (Fig. 5e). These results indicated that the enhanced drought tolerance of OsGRXS17 silenced plants is caused by an increased ability of retaining water.

Reduced OsGRXS17 Expression Promotes Greater Stomatal Closure in Both the Presence and
Absence of Exogenous ABA. Because OsGRXS17 silenced plants showed reduced water loss rates and stomatal conductance, the effect of OsGRXS17 expression on stomatal aperture was investigated. Stomatal aperture status, upon examination by scanning electron microscopy, was classified into three groups: closed, partially open, and open (Fig. 6a). The results indicated that reduced OsGRXS17 expression led to a greater degree of stomata closure under normal growth conditions (Fig. 6b, left panel). In the extremes, wild-type plants had 2% closed stomata, 28% partially opened stomata, and 75% open stomata, while OsGRXS17 RNAi-13 plants had 20% closed, 42% partially opened, and 38% open stomata (Fig. 6b, Control). Stomatal response to the application of ABA was then evaluated. All of the silenced lines showed greater stomatal closure compared to wild-type plants in the presence of ABA (Fig. 6b, ABA). No differences in stomatal density were found between OsGRXS17 silenced lines and wild-type plants (Fig. S6). The results indicated that ABA treatment resulted in a higher percentage of closed . Asterisks (*, **, ***) represent statistically significant differences between wild-type and OsGRXS17 silenced lines (Student's t-test, *P < 0.05, **P < 0.01, ***P < 0.001). stomata in OsGRXS17 silenced rice plants in comparison to wild-type plants. To determine if the ABA hypersensitivity of the silenced lines was caused by over-accumulation of endogenous ABA, the levels of ABA were measured by metabolite profiling. ABA content and accumulation patterns were similar between the wild-type and silenced plants over the entire period of drought stress treatment (Fig. 6c), indicating that the increased stomatal closure in OsGRXS17 silenced rice plants was not the result of ABA over-accumulation.

Reduced Expression of OsGRXS17 is Associated with the H 2 O 2 Accumulation in Guard
Cells. ABA-induced accumulation of H 2 O 2 , which is synthesized in guard cells, is essential for stomata closure by activating plasma membrane Ca 2+ channels 5,37,38 . H 2 O 2 accumulation was measured by staining the leaves of wild-type and OsGRXS17 silenced plants using 3,3′-diaminobenzidine (DAB) and image analysis. In the absence of ABA, the leaves of OsGRXS17 RNAi-7, -8 and -13 plants displayed slightly higher H 2 O 2 levels on the basis of DAB staining than those of OsGRXS17 RNAi-6 and wild-type plants (Fig. 7a, Control). In the presence of 100 µmol/L ABA, H 2 O 2 accumulation was visibly increased in both wild-type and silenced leaves (Fig. 7a, ABA). However, the OsGRXS17 RNAi-7, -8 and -13 plants displayed much stronger staining in comparison to leaves of The four-week-old plants (0d) were treated by withholding water for 11 days and re-watered for 6 days to recover. (b) The survival rate of the wild-type and OsGRXS17 silenced rice plants after 11 days of drought stress treatment and 6 days of re-watering. (c) Water loss rate in the leaves cut from four-week-old wild-type and OsGRXS17 silenced rice plants (n = 3 repeats). (d) The relative water content of four-week-old wild-type and OsGRXS17 silenced rice plants after 5 days of drought stress treatment (n = 3 repeats). (e) Stomatal conductance of four-week-old wild-type and OsGRXS17 silenced rice plants under drought stress for 11 days (n = 3 repeats). Error bars represent the means ± SD (n = 3). Asterisks (*, **, ***) represent statistically significant differences between wild-type and OsGRXS17 silenced lines (Student's t-test, *P < 0.05, **P < 0.01, ***P < 0.001).
SCieNtifiC REPORtS | 7: 15950 | DOI:10.1038/s41598-017-16230-7 OsGRXS17 RNAi-6 and wild-type plants (Fig. 7a, ABA). Quantitative image analysis revealed higher H 2 O 2 accumulation in OsGRXS17 RNAi-7, -8 and -13 compared to that of OsGRXS17 RNAi-6 and wild-type plants in both the absence of exogenous ABA (Fig. 7b, control) and presence of exogenous ABA (Fig. 7b, ABA). H 2 O 2 accumulation was also monitored in guard cells using the fluorescent dye, 2′,7′-dichlorodihydrofluorescein diacetate (H 2 DCFDA). The results showed that OsGRXS17 RNAi-7, -8 and -13 lines had stronger fluorescence signal in the guard cell as compared to OsGRXS17 RNAi-6 and wild-type plants in the absence of ABA addition, indicating that more accumulation of H 2 O 2 in the guard cell of OsGRXS17 RNAi-7, -8 and -13 (Fig. 7c, Control). In the presence of ABA addition, the fluorescence signal in the guard cells was enhanced in both wild-type and OsGRXS17 silenced rice plants, with, again, OsGRXS17 RNAi-7, -8 and -13 having stronger fluorescence signals in the guard cells as compared to OsGRXS17 RNAi-6 and wild-type plants, indicating that higher accumulation of H 2 O 2 in the guard cells of OsGRXS17 RNAi-7, -8 and -13 lines than that of OsGRXS17 RNAi-6 and wild-type plants under ABA treatment (Fig. 7c, ABA). The cytoplasm and nucleus of guard cells are dumbbell shaped and the two terminal masses are connected through a very thin central canal 39 . It is worth noting that the two terminals of rice guard cell have a fluorescence signal that is more easily detected while the central region is less noticeable due to the thickness of the guard cell wall in the central region, which is consistent with previous studies 40,41 . Quantitative analysis of fluorescence intensity also showed that H 2 O 2 accumulation in the guard cell was higher in OsGRXS17 RNAi-7, -8 and -13 compared with that of OsGRXS17 RNAi-6 and wild-type plants with or without ABA treatment (Fig. 7d). A negative correlation between the OsGRXS17 expression level (Fig. 4b) and the H 2 O 2 accumulation in guard cells indicates that the increased ABA-induced stomatal closure in the OsGRXS17 silenced lines was associated with higher H 2 O 2 accumulation.

Expression of ABA-Responsive Genes Is Elevated in OsGRXS17 Silenced Plants. To exam-
ine if the ABA hypersensitivity and increased stomatal closure of the silenced lines associated with, at least in part, changes in ABA-responsive gene expression, the expression levels and patterns of four ABA-dependent drought-responsive genes (RAB16A, LEA3, LIP9, and SalT) and two ABA-independent genes (DREB1A and DREB1E) 42 were analyzed in ABA-treated leaf samples by qRT-PCR. Inconsistent expression levels and patterns were detected for LIP9, SalT, DREB1A and DREB1E in both wild-type and OsGRXS17 silenced rice plants. Expression levels of two ABA-responsive genes LEA3 and RAB16A [belongs to LEA (late embryogenesis  (Figs 8 and S7). To evaluate if drought stress responses and tolerance in the silenced lines is associated with altered regulation of the six selected ABA-dependent and/ or ABA-independent drought-responsive genes, the expression levels and patterns of the six genes were also analyzed in drought stress treated-leaf samples. All six genes were highly upregulated after 24 h of drought stress treatment regardless of genotype (Fig. S8). These results suggest that repression of OsGRXS17 selectively affects some, but not all ABA-responsive genes.

Discussion
Previous studies have shown that the monothiol CGFS-type glutaredoxin GRXS17 plays important roles in plant abiotic stress adaptation, and that ectopic over-expression of GRXS17 in plants enables enhanced heat and chilling stress tolerance [33][34][35][36]43 . Based on these reports, we hypothesized that reduction in GRXS17 expression would result in plants that are more sensitive to abiotic stresses, including drought, compared to wild-type controls. Contrary to this hypothesis, here, suppression of OsGRXS17 expression resulted in enhanced tolerance to the drought stress conditions in rice. Previous work had also shown that ROS, H 2 O 2 in particular, are important signaling molecules in the regulation of stomatal aperture 44 . OsGRXS17 silenced rice plants were found to have higher steady state H 2 O 2 concentrations in guard cells in the absence of drought stress or exogenous ABA treatment. The increase in H 2 O 2 accumulation is consistent with the observed increase in stomatal closure and increased sensitivity to the Error bars represent the means ± SD (n = 3 repeats, 6 plants in each repeat). Asterisks (*) represent statistically significant differences between wild-type and OsGRXS17 silenced lines (Student's t-test, *P < 0.05).
endogenous ABA compared to wild-type controls. Earlier studies have shown that antioxidant enzymes function in the stomatal closure via modulating ROS levels in guard cells. For example, a rice dst (drought and salt tolerance) mutant, lacking a zinc finger transcription factor, showed drought tolerance compared to wild-type plants. Further analyses indicated that DST bound directly to the promoter of genes encoding antioxidant enzymes that were negative regulators of H 2 O 2 homeostasis 40 . Tobacco plants engineered to express higher level of dehydroascorbate reductase (DHAR) had an increase in ascorbic acid redox state, reduction in guard cell H 2 O 2 concentrations, a reduction in stomatal closure, and an increase in susceptibility to drought while tobacco plants engineered to suppress expression of DHAR had a decrease in ascorbic acid redox state, an increase in guard cell H 2 O 2 level, and increase in drought tolerance 45 . In addition, the H 2 O 2 -induced stomatal closure could be reversed by exogenous ascorbic acid treatment in a concentration-dependent manner 5 . Consistent with these reported studies, we found that guard cell H 2 O 2 concentrations inversely correlated with OsGRXS17 expression. The lower the expression of OsGRXS17, the more H 2 O 2 accumulated in guard cells, indicating that OsGRXS17 plays an important role in modulating H 2 O 2 homeostasis in guard cells.
Another factor modulating cellular H 2 O 2 concentrations is ABA. ABA has been shown to induce H 2 O 2 production in guard cells via the activation of plasma membrane NADPH oxidases 5,38,46 . H 2 O 2 then acts as an essential second messenger in the ABA signaling pathway, mediating stomatal closure in response to drought stress 37,38,47,48 . In Vicia faba, ABA-induced stomatal closure was abolished by the presence of catalase (CAT), an H 2 O 2 scavenger, and diphenylene iodonium (DPI), an inhibitor of NADPH oxidases, which remove H 2 O 2 and reduce the production of H 2 O 2 , respectively 5 . Similarly, DPI has been shown to inhibit ABA-induced stomatal closure in Arabidopsis 38 . In this study, OsGRXS17 silenced rice plants had increased H 2 O 2 in guard cells even without exogenous ABA treatment, possibly due to 1) the loss of a functional role of OsGRXS17 as a ROS scavenger or 2) the hypersensitivity of OsGRXS17 silenced lines to endogenous ABA. Further, we did find that OsGRXS17 silenced rice plants accumulated more H 2 O 2 after ABA treatment compared to wild-type rice plants. Since H 2 O 2 plays an important role in mediating ABA-induced stomatal closure, it is likely that enhanced H 2 O 2 accumulation is associated with increased stomatal closure in OsGRXS17 silenced rice plants in the presence of ABA. However, we cannot still exclude the possibility that another parallel pathway may involve in stomatal closure via H 2 O 2 production.
Application of exogenous ABA was also found to upregulate two ABA-dependent drought stress-responsive genes, LEA3 and RAB16A, in the OsGRXS17 silenced RNAi-7, -8 and -13 rice plants in comparison to OsGRXS17 RNAi-6 and wild-type plants. These two genes; however, did not show different expression levels and patterns between wild-type and OsGRXS17 silenced rice plants under drought stress treatment (Fig. S8). LEA proteins are a family of highly hydrophilic proteins that accumulate in mature seeds and vegetative tissues under dehydrated conditions and play an important role in protection of proteins and membrane stabilization. Overexpression of OsLEA3 has been shown to enhance drought tolerance in rice 49 . Thus, the reduction of OsGRXS17 expression may enhance drought tolerance by mediating ABA-regulated but not drought-regulated mRNA accumulation of LEA genes. Expression of other ABA-responsive and drought-responsive genes were not affected by OsGRXS17 silencing under drought stress and exogenous ABA treatments, suggesting that OsGRXS17 might be involved in one portion of the ABA signaling pathways. OsGRXS17 could regulate ABA signaling through a post-translational Error bars represent the means ± SD (n = 3). Asterisks (**, ***) represent statistically significant differences between wild-type and OsGRXS17 silenced lines (Student's t-test, **P < 0.01, ***P < 0.001).
SCieNtifiC REPORtS | 7: 15950 | DOI:10.1038/s41598-017-16230-7 modification mechanism. Measurements of endogenous ABA content revealed no significant difference between OsGRXS17 silenced and the wild-type rice plants under drought stress, suggesting that OsGRXS17 might not be involved in the ABA biosynthesis pathways. Taken together, these results suggested that OsGRXS17 may play dual roles in a parallel way. It regulates H 2 O 2 homeostasis, acting as a ROS scavenger, and involves in H 2 O 2 -mediated stomata closure. It may also participates in downstream signaling pathways of ABA by regulating the gene expression of LEA3 and RAB16A.
As plants lose over 95% of water through transpiration, engineering of stomata activity is an important approach to enhance drought tolerance in plants 50 . To further characterize the drought tolerance of silenced OsGRXS17 rice plants, we measured the water maintaining capacity and found the transgenic rice plants indeed showed reduced water loss rate, higher relative water content and reduced stomatal conductance compared to wild-type plants under drought stress, suggesting that increased ROS accumulation in OsGRXS17 silenced rice plants might play an important role in drought stress response through regulating the redox-dependent signaling pathway. Although the OsGRXS17 silenced plants exhibit reducing stomatal opening, no effect on rice grain yield was found (Fig. S2). Thus under both normal and drought stress conditions, CO 2 influx is assumed to be sufficient to support normal photosynthesis and plant growth.

Methods
Yeast Assays. The full-length cDNA of AtGRXS17 and OsGRXS17 was subcloned into yeast expression vector piUGpd. Yeast grx3grx4 double mutant (MATa ura3-52 leu2Δ1 his3Δ200 grx3::kanMX4 grx4::kanMX4) were provided by Dr. Enrique Herrero (Universitat de Lleida, Lleida, Spain). Yeast growth assays were performed as previously described 29 . OsGRXS17 complementation assays in atgrxs17 mutants. To understand the effect of OsGRXS17 on plant growth in atgrxs17 mutants, an Arabidopsis expression cassette was developed. Full-length OsGRXS17 was inserted into the entry vector (pENTR/D-TOPO vector, Invitrogen, Carlsbad, CA). The 35S::GFP-OsGRXS17 construct was made by LR clonase reaction between the entry vector and pB7WGF2 destination vector 51 and was introduced into A. tumefaciens strain LBA4404 using the freeze-thaw method 52 . The atgrxs17 mutant plants were transformed using the floral dip method 53 . Plants were screened by spraying a 1% BASTA solution on cotyledons. Seeds of wild-type, atgrxs17 mutant and T2 transgenic atgrxs17 mutants expressing OsGRXS17 were placed on half-strength Murashige and Skoog (MS) media supplemented with 0.5% sucrose 54 . Seeds were allowed to germinate for three days at control temperatures followed by control (22 °C) or heat stress (28 °C) conditions for ten days supplemented with 150 µmol/m 2 /s light intensity.

Subcellular localization of OsGRXS17 in plant cells. To study the subcellular localization of
OsGRXS17 in plant cells, an Agrobacterium-mediated transient expression assay was conducted in tobacco leaves (N. tabacum) and polyethylene glycol (PEG)-mediated protoplast transformation was performed in protoplasts derived from rice leaf sheaths as described previously 55,56 . Full length OsGRXS17 was inserted into the pEN-TR/D-TOPO cloning vector. The GFP-OsGRXS17 construct was produced by an LR clonase reaction between the entry vector and pB7WGF2 51 . As a control, a modified green fluorescent protein construct (Free GFP construct) was made by the Cre-loxP recombination system 57 . To function as a nuclear marker, pSK001 was generated by inserting a 1.9 kb SacI-HindIII fragment from pBV579 (containing 35S::mCherry::NLS::Tnos) into the unique SacI and HindIII sites of pCAMBIA1300. These three constructs were introduced into A. tumefaciens LBA4404 for transient expression in tobacco leaves. A. tumefaciens cells were cultivated overnight, and 5 mL of the culture was pelleted and re-suspended with infiltration medium (250 mg D-glucose, 5 mL MES stock solution, 5 mL Na 3 PO 4 •12H 2 O stock solution, 5 mL 1 M acetosyringone stock solution; made up to 50 mL final volume with ddH 2 O.) to an optical density of 0.1. A. tumefaciens cells were then infiltrated into tobacco leaves, and the infiltrated tobacco plants were kept under constant light for 1.5-2 days. The fluorescence signals were detected after 1.5-2 days of inoculation.
For PEG-mediated protoplast transformation, protoplasts derived from rice sheath tissues were used 56 . Rice seedlings were grown in a growth chamber with a 16 h light (28 °C)/8 h dark (22 °C) photoperiod. Sheath tissues from fifty 2-week-old seedlings were sliced into 1-mm strips with a razor blade and placed into a flask with the digestion solution (0.15 M sorbitol, 0.25 M sucrose, 35 mM CaCl 2 , 20 mM KCl, 1.5% Cellulase R10 (From Trichoderma viride, 7.5 U/mg), 0.75% Macerozyme (R10 Macerating enzyme from Rhizopus sp. RPI) and 10 mM MES-KOH (pH 5.7). Vacuum was applied to the samples in the dark for 30 mins and then incubated at room temperature for 2 hours with gentle shaking at 20-30 rpm. The digested tissues were filtered into a centrifuge tube using a 40 µm nylon mesh and then the mesh was rinsed with 20 mL W5 solution (0.1% glucose, 0.08% KCl, 0.9% NaCl, 1.84% CaCl 2 •2H 2 O, 2 mM MES-KOH, pH 5.7). After centrifuging at 100 g for 7 mins at room temperature, the protoplasts were collected from the interface between the digestion solution and W5. The protoplasts were washed with W5 solution twice, resuspended in 3 mL W5 solution and incubated on ice for 30 min. Then, W5 solution was removed and the protoplasts were resuspended in MMG solution (0.4 M mannitol, 15 mM MgCl 2 , 4 mM MES-KOH, pH 5.7). Ten micrograms of plasmid DNA and 100 µL protoplasts (adjusted to 10 6 protoplasts/ mL) were gently mixed with 130 µL PEG-calcium transfection solution (40% PEG4000, 0.2 M mannitol, 100 mM CaCl 2 ). After incubation for 30 mins, the transfection mix was diluted with 500 µL W5 solution, centrifuged at 100 g for 2 mins, and then resuspended in 1 mL W5 solution. The fluorescence signals were detected after 18 hrs of incubation at room temperature.
Images were captured with a confocal laser scanning system (Leica, SP5 X, Leica Microsystems Inc., Buffalo Grove, IL, USA) and fluorescence microscope (Zeiss Axio-Plan, Carl Zeiss Microscopy, Thornwood, NY, USA). The fluorescence signals were detected at 510 nm (excitation at 488 nm) for GFP and at 610 nm (excitation at 587 nm) for mCherry. RNAi Plasmid Construction and Rice Transformation. Two isoforms of OsGRXS17 (Os10g35720.1 and Os10g35720.2) were identified using Rice Functional Genomic Express Database (http://signal.salk.edu/cgi-bin/ RiceGE). To knock down both isoforms of OsGRXS17 gene, a 398-bp of the OsGRXS17 gene at the 3′ ends of the coding sequence and 3′UTRs was amplified using a forward primer: 5′-CACCAGGGATCGTTGCGAAAGAAA-3′ and reverse primer: 5′-AGCAAACTCGATGGTCGACGGATG-3′ as the silenced-targeted region and subcloned into the pENTR/D-TOPO vector (Fig. S9). This 389 bp of knockdown target sequence covers 316 bp of coding sequence of the long isoform of OsGRXS17 mRNA and 73 bp of 3′ UTR region of both isoforms (long and short isoforms) of OsGRXS17 mRNA (Fig. S9), which was aligned against Rice Functional Genomic Express Database to avoid off-target problems. After verification by DNA sequencing, the Gateway ™ cassette was introduced the pANDA vector 58 by the LR recombination reaction. Recombination between pENTR vectors and destination vectors were performed according to the manufacturer's instructions (Invitrogen, Carlsbad, CA).
The verified plasmid DNAs were introduced into Agrobacterium tumefaciens LBA 4404 using the freeze-thaw method 52 . Mature seed-derived callus from rice (Oryza sativa L. Japonica) cv. Nipponbare was used for Agrobacterium-mediated transformation 59 . After inoculating with A. tumefaciens, callus was transferred to regeneration medium for 4-10 weeks at 25 °C under a 16-h photoperiod. The regenerated shoots were transferred to rooting medium for four more weeks, then established in soil.

Plant Materials and Growth Conditions.
To analyze the expression pattern of OsGRXS17 in response to stress and hormones, the 2-week-old wild-type seedlings were placed on dry filter paper or transferred to filter paper saturated with MS solution supplemented with 25% PEG, 100 μM ABA, 1 μM IAA or 200mM NaCl, respectively, as described previously 60 . T2-generation OsGRXS17 silenced or wild-type rice seeds were surface-sterilized and geminated on MS medium with or without 40 mg/L hygromycin for 7 days, and the 7-d-old seedlings were transferred into small pots containing Baccto premium potting soil (Michigan Peat Company, Houston, Texas, USA) in growth chamber with a 16 h light (28 °C)/8 h dark (22 °C) photoperiod. The pots were kept in the flat-bottom trays containing water. For drought treatment, 3-week-old rice seedlings with three leaves appeared were exposed to drought stress treatments. Each pot was filled with the same amount of soil to provide similar soil humidity. The water was withheld from the trays for 11 days and then the stressed plants were re-watered to allow the wilted rice plants to recover. After 14 days of recovery, the survival rates (%) were calculated from the numbers of surviving plants per total tested plants. The plants were scored as viable if one or more new leaves appeared.
Physiological Measurements. The measurement of relative water content (RWC) was performed as described previously 61 . Briefly, on the fifth day of withholding water treatment, the first fully expanded leaves were detached and the fresh weight (FW) was measured immediately. Then the leaves were completely immerged in the double distilled water overnight for rehydration. The turgid weight (TW) was measured after blotting the rehydrated leaves. Finally, the dry weight (DW) was measured after drying in an oven at 80 °C overnight. The RWC was calculated as follows: RWC = (FW−DW)/(TW−DW). The measurement of water loss rate was performed by placing the detached first fully expanded leaves on a laboratory bench and recording their fresh weight for 0, 1, 2, 3, 4, and 5h. It is expressed as percentage of initial fresh weight. The stomatal conductance (mmol m −2 s −1 ) was measured in the first fully expanded leaf of 4-week-old wild-type and OsGRXS17 silenced plants using a portable porometer (SC-1 Leaf porometer, Decagon Devices, Pullman, WA, USA) in the auto mode for 30 seconds. The stomatal conductance was followed after being treated under drought stress for 0, 1, 3, 5, 7, 9 and 11 days. To measure the stomata density, the first fully expanded leaves of 4-week-old wild-type and OsGRXS17 silenced rice seedlings grown in the growth chamber were sampled. Imprints were made by coating the adaxial surface with clear nail polish. After a few minutes, the clear tape was used to peel off the nail polish and this was mounted onto microscope slides. The stomata density (number of stomata per unit area) was counted from three random areas on the leaf under a light microscope (Olympus CH30; Olympus, Tokyo, Japan).  43 . The first fully expanded leaves detached from 4-week-old wild-type and OsGRXS17 silenced rice plants were vacuum-infiltrated in 0.01% Tween 20 for 5 mins and then treated with 100 µM ABA for 3 h. The sampled leaves were collected and incubated in DAB solution (1 mg/ mL, pH 3.8; Sigma-Aldrich) for 24 h at room temperature in darkness. The leaves were then de-colorized in boiling ethanol (96%) for 10 min before photographing. For the root tips staining, the 7-day-old rice seedlings were incubated in DAB solution (0.1 mg/mL, pH 3.8; Sigma-Aldrich) for 2 h at room temperature in darkness. Then the root tips were rinsed by ddH 2 O for three times and mounted on microscope slides for photographing. Quantitative analyses of DAB staining were performed using image J analysis 43 .
The H 2 DCFDA staining assay was performed as previously to detect H 2 O 2 production in the guard cell 40 . The first fully expanded leaves from 4-week-old wild-type and OsGRXS17 silenced rice plants were vacuum-infiltrated in 0.01% Tween 20 for 5 min and then incubated in 2% (w/v) cellulose Onozuka RS (Sigma-Aldrich) at 37 °C for 5 h to facilitate peeling off the epidermal layers. The peeled epidermal strips were incubated in the loading buffer (10 mM Tris-HCl, 50 mM KCl at pH 7.2) and then transferred to the staining buffer (loading buffer containing 50 mM H 2 DCFDA) for 20 min. The stained epidermal strips were washed in the ddH 2 O for three times to remove the excess H 2 DCFDA and mounted on microscope slides to detect with a Zeiss LSM 780 laser-scanning confocal microscope (Carl Zeiss SAS, Jena, Germany) using following parameter settings: excitation 488 nm, emission 546 nm, 2% laser power percent, 16 Bit depth, image size 1024 × 1024 pixels, scanning speed 7. Fluorescence was analyzed using ImageJ software. Thirty to fifty guard cells were observed per treatment for three independent replicates.
SCieNtifiC REPORtS | 7: 15950 | DOI:10.1038/s41598-017-16230-7 Scanning Electron Microscopy. The first fully expanded leaves detached from 4-week-old wild-type and OsGRXS17 silenced rice plants treated with 100 µM ABA or drought for 3 h were used. The leaf segment (0.5 cm) were cut from the middle of the leaf, fixed by 2.5% glutaraldehyde in 0.1 M phosphate buffer at pH7.2. The samples were then rinsed 3 times in distilled water, dehydrated in ethanol series (30,50,70, 80, 95%), and rinsed 3 times in 100% ethanol. The samples were finally critical point dried using hexamethyldisilazane (HMDS) 62,63 . The dehydrated samples were then sputter-coated with gold and used for stomatal observation by using a Hitachi S-3500N scanning electron microscope (Hitachi, Tokyo, Japan). Thirty to fifty guard cells were observed per treatment for three independent replicates. RNA Extraction and qRT-PCR. Total RNA was isolated from leaves, stems, roots and panicles using the Qiagen Plant RNeasy kit (Qiagen, Valencia, CA) and on-column DNA digestion according to the manufacturer's instructions. The cDNA was transcripted from 1 µg quantity of total RNA using iScript ™ Select cDNA Synthesis Kit (Biorad, Hercules, CA). The qRT-PCR was carried out in a total volume of 10 µL containing 4.2 µL the reverse-transcribed product, 0.4 µL 10 mM of each primer, and 5 µL SYBR Green PCR Master Mix (Bio-Rad). The PCR was performed with a Bio-Rad CFX-96 real time system (BioRad). Primer efficiencies were measured and relative expression level was calculated using the comparative Ct method. TFIIAγ5 was used as the internal control to normalize the samples. The primers used for qRT-PCR were listed in the supplemental Table 1. DNA Gel-blot Analysis. Rice genomic DNA was extracted from 4-week-old rice seedlings using CTAB method as described previously 64 . HindIII-digested genomic DNA (30 µg) was separated by electrophoresis on 0.8% (w/v) agarose gel, and blotted onto a nylon membrane (Zeta-probe GT membrane, Bio-Rad, Hercules, CA) according to the manufacturer's instructions. The probe for the hpt gene was isolated from a pIPKB007 vector by PCR amplification. The membranes were pre-hybridized at 65 °C in 7% SDS and 0.25 M Na 2 HPO 4 for 2 h and then hybridized overnight at 65 °C in the same solution containing the probe labeled with the enzyme horseradish peroxidase (ECL Direct ™ Nucleic Acid Labeling and Detection System, Amersham Biosciences, Piscataway, NJ, USA) for 10-12 h at 42 °C. Membranes were washed twice for 40 min each with 20 mm Na 2 HPO 4 and 5% SDS at 65 °C and then washed twice again for 30 min each with 20 mm Na 2 HPO 4 and 1% SDS at 65 °C. Finally, the membrane was wrapped in Saran Wrap and exposed to X-ray film (Fuji Film Medical Systems, Stamford, CT) for 1-2 h. Measurement of ABA content by metabolite profiling analysis. First fully expanded leaves from 4-week-old wild type and OsGRXS17 silenced rice plants were collected after 0 (0d), 5 (5d) and 8 (8d) days of drought stress treatment. Four biological replicates from each line were collected at each time point, lyophilized and kept at −80 °C until extraction. One hundred mg of lyophilized samples was submitted to Metabolon, Inc. (Durham, NC), for sample extraction and metabolite profiling analysis. The ABA content is provided by metabolite profiling analysis 62 . In brief, each sample was thawed on ice and extracted using an automated MicroLab STAR system in 400 µL of methanol containing recovery standards. A series of organic and aqueous extractions were performed to remove the protein fraction and also allow maximum recovery of small molecules. The extract was divided for two analyses: one for HPLC/MS analysis and the other for GC/MS analysis. Compounds were identified by comparison to library entries of purified standards or recurrent unknown entities.