A Nck‐associated protein 1‐like protein affects drought sensitivity by its involvement in leaf epidermal development and stomatal closure in rice

Summary Water deficit is a major environmental threat affecting crop yields worldwide. In this study, a drought stress‐sensitive mutant drought sensitive 8 (ds8) was identified in rice (Oryza sativa L.). The DS8 gene was cloned using a map‐based approach. Further analysis revealed that DS8 encoded a Nck‐associated protein 1 (NAP1)‐like protein, a component of the SCAR/WAVE complex, which played a vital role in actin filament nucleation activity. The mutant exhibited changes in leaf cuticle development. Functional analysis revealed that the mutation of DS8 increased stomatal density and impaired stomatal closure activity. The distorted actin filaments in the mutant led to a defect in abscisic acid (ABA)‐mediated stomatal closure and increased ABA accumulation. All these resulted in excessive water loss in ds8 leaves. Notably, antisense transgenic lines also exhibited increased drought sensitivity, along with impaired stomatal closure and elevated ABA levels. These findings suggest that DS8 affects drought sensitivity by influencing actin filament activity.


INTRODUCTION
Adverse climatic conditions, such as drought, threaten rice growth and production. To adapt to drought stress, plants have evolved various mechanisms involving adaptive changes at the morphological, physiological and molecular levels. Characteristics of the leaf epidermis are important indicators of plant drought tolerance (Fang and Xiong, 2015).
In rice, the epidermal cell wall is covered by a hydrophobic cuticle on the outside and embedded with silica beneath the cuticle (Kunst and Samuels, 2003;Ma and Yamaji, 2008). The cuticle, composed of cutin and wax, is an indispensable barrier that prevents non-stomatal water loss (Mao et al., 2012). Mutants with changes in cuticle wax components, such as wsl1, wsl3, wsl4, cfl1 and dwa1, have increased sensitivity to drought stress (Yu et al., 2008;Wu et al., 2011;Zhu and Xiong, 2013;Gan et al., 2016;Wang et al., 2017). The silica-embedded cuticular papillae (CP) also help reduce cuticular transpiration in rice (Yoshida et al., 1962;Agarie et al., 1998;Epstein, 1999;Gao et al., 2005). Stomata, which play a role in water and gas exchange, are responsible for approximately 90% of the water loss that occurs in the leaf epidermis (Buckley, 2005). Stomatal density and stomatal aperture are therefore important indices used to evaluate drought resistance in plants. Overexpression lines of OsDT11 and SDD1 display dramatically enhanced drought tolerance due to their reduced stomatal densities (Yoo et al., 2010;Li et al., 2016). Stomata generally close in response to low environmental humidity to prevent excessive water loss (Pantin and Blatt, 2018). Genes that positively modulate stomatal closure can improve the survival rates of plants during drought stress (Huang et al., 2009;Jin et al., 2013;Dey et al., 2016;Gao et al., 2018).
Abscisic acid (ABA) induces stomatal closure through a complex signaling network (Munemasa et al., 2015;Matsuda et al., 2016;Sussmilch and McAdam, 2017). The perception of the ABA signal by ABA receptors activates downstream plasma membrane Ca 2+ channels and K + efflux channels, resulting in stomatal closure (Munemasa et al., 2015). A decrease in the ABA content in guard cells dampens the response of leaf stomata to ABA stimulation, leading to reduced stomatal closure, followed by increased sensitivity to water loss (Kang et al., 2010;Matsuda et al., 2016). ABA also regulates stomatal aperture by driving reactive oxygen species (ROS) biosynthesis (Mittler and Blumwald, 2015;Sierla et al., 2016). Rice plants with elevated H 2 O 2 levels in guard cells exhibit reduced stomatal aperture and increased tolerance to water deficit Hu et al., 2017).
Most studies on ABA-mediated stomatal closure have focused on the functions of ABA receptors or ion channels in guard cells (Park et al., 2009;Geiger et al., 2010;Guo et al., 2011;Zhao et al., 2016;M€ uller et al., 2017). Some studies have pointed to a relationship between the ABAsignaling network and actin filaments (Li et al., 2014). Actin filaments are major components of the complex cytoskeleton, and participate in various cellular events affecting plant growth and development (McCurdy et al., 2001;Thomas and Staiger, 2014). During stomatal closure, the actin filaments in guard cells undergo a depolymerization-reintegration process that functions upstream of stomatal movement (Eun and Lee, 2000;Gao et al., 2009). To ensure successful stomatal closure, the radially oriented actin filaments in open stomata must depolymerize, become randomly distributed, and reintegrate in longitudinal actin cables Lemichez et al., 2001;Zhang et al., 2007;Higaki et al., 2010;Zhao et al., 2011). Suppressed actin filaments dynamics in Arabidopsis thaliana guard cell delays ABA-mediated stomatal closure and results in hypersensitivity to drought stress (Zhao et al., 2011). The Arp2/3 complex, comprising seven elements (ARP2, ARP3, ARPC1, ARPC2, ARPC3, ARPC4 and ARPC5), is activated by the SCAR/WAVE complex to increase the efficiency of actin filaments nucleation and to initiate new actin filaments branching (Pollard, 2007;Yanagisawa et al., 2015). The SCAR/WAVE complex consists of five evolutionarily conserved subunits, namely ABI, BRICK1/ HSPC300, NAP1/NAP125, PIR/SRA1 and SCAR (Le et al., 2006;Mendoza, 2013;Zhou et al., 2016). The SCAR/WAVE complex members can enhance the Arp2/3 complexmediated nucleation efficiency of actin filaments in vitro and in vivo (Welch and Mullins, 2002;Bai et al., 2015). All those results indicated that the SCAR/WAVE-Arp2/3 complex may be involved in ABA-mediated stomatal closure.
Enormous effort has been devoted to elucidating the role of the SCAR/WAVE-Arp2/3 complex in the formation and development of leaf epidermal cells by involvement in actin filaments modeling over the past decades (Frank and Smith, 2002;Djakovic et al., 2006;Facette et al., 2015); however, few studies have focused on the function of SCAR/WAVE complex in plant responses to environmental stress, especially drought stress. In our study, the Nck-associated protein 1 (NAP1)-like protein coding gene DROUGHT SENSITIVE 8 (DS8) was hypothesized to affect drought sensitivity by influencing actin filament activity. The study was designed to find out how DS8 affects drought sensitivity in rice plant, focusing on the impacts of DS8 on indicators related to drought sensitivity, including cuticle and stoma. Cytochalasin B (CB) and sodium tungstate treatments were performed to preliminarily analyze the relationship between ABA and actin filaments activity during ABA-induced stomatal closure.

RESULTS
ds8 is sensitive to drought stress The ds8 mutant was screened under drought stress conditions from a Nipponbare (NPB, japonica) rice mutant pool generated by ethyl methanesulfonate mutagenesis. A drought stress test indicated that ds8 seedlings were more sensitive to drought stress than the wild-type ( Figure 1a). After recovery from drought treatment, the survival rate of ds8 was only 9.4% that of wild-type ( Figure 1b). In addition, when excised flag leaves were placed in an incubator at 60% humidity, ds8 leaves curled quickly ( Figure 1c) and showed a higher water loss rate than wild-type leaves (Figure 1d). We also investigated major agronomic traits related to grain yield in dry environmental conditions. Under drought conditions, ds8 suffered higher decline rates in agronomic traits such as tiller number, grain number per panicle, filled grains per panicle and grain yield per plant than the wild-type ( Figure S1; Table S1).

Map-based cloning of DS8
Reciprocal crosses between ds8 and three varieties (9311, ZF802 and CJ06) of wild-type rice plants were carried out. Plants from all six F 1 populations exhibited normal phenotypes. In all F 2 populations, the ratio of wild-type to ds8 mutant phenotypes in every combination was~3:1 (Table S2).
We performed map-based cloning, and mapped DS8 to chromosome 8 between markers RM1345 and RM3120 ( Figure 2a). The primer sequences used are listed (Table S3). Furthermore, using 2319 homozygous mutant plants, DS8 was narrowed down to a 53.7-kb genomic region between markers E1 and E2. There were seven predicted open reading frames (ORFs) in this region, according to the Rice Genome Annotation Project (RGAP). We sequenced these predicted ORFs and identified a singlebase substitution (G to A) at the last nucleotide of the second intron in LOC_Os08g43130 in ds8. This mutation caused a premature stop codon due to the deletion of the entire third exon (81 bp) and part of the 17th exon (8 bp; Figure 2b).
A genetic complementation test was arranged to ds8 mutant and 22 positive proDS8:DS8 transformants were  Table S4).
We carried out quantitative reverse-transcription polymerase chain reaction analysis of DS8 transcript levels in various tissues at the booting stage. DS8 was widely expressed in panicle, culm, sheath, root and leaf tissue ( Figure S3a). We also detected GUS (b-glucuronidase) activity in these tissues in proDS8:GUS transgenic rice plants ( Figure S3b-f). Analysis of the subcellular localization of DS8-eGFP in rice protoplasts and pro35S:DS8-eGFP transgenic plants indicated that DS8 was localized to the cytoplasm ( Figure S3g,h).

DS8 encodes a NAP1-like protein involved in actin filaments function
Sequence analysis showed that DS8 comprised 23 exons and 22 introns, and encoded a 1359 amino-acid protein ( Figure 2b). A protein-protein BLAST search against the NCBI database using the deduced DS8 protein sequence as a query revealed that it was a NAP1-like protein. We selected 25 DS8 homologs from various species to determine the evolutionary relationship between DS8 and these proteins, finding that the DS8 homologs were conserved in both monocots and dicots ( Figure S4).
Protein sequence alignment revealed that DS8 shares 86.28% amino acid identity with homologs from Hordeum vulgare, Sorghum bicolor, Zea mays and A. thaliana ( Figure S5).
The organization of actin filaments in the seedling roots of ds8 and wild-type plants was observed. The actin filaments were complete and regularly arranged in wild-type. In ds8, disordered actin filaments were observed at the meristematic zone, became more severe during cell growth from the meristematic zone to maturation zone, and appeared incomplete and cluttered in the maturation zone (Figure 2f-h).

DS8 is involved in leaf cuticle development
We stained the leaves of wild-type and ds8 plants with toluidine blue, which can roughly reflect the integrity of the cuticle on the leaf surface. Wild-type leaves successfully resisted toluidine blue staining, whereas ds8 leaves were partially stained by the blue dye ( Figure 3a). Transmission electron microscopy (TEM) analysis showed that the cuticular layer of the leaf epidermis in ds8 was significantly thicker (nearly 1.85-fold) than that of wild-type (Figure 3b,c). However, the less osmiophilic membrane of ds8 exhibited reduced electron density, as revealed by its lighter coloring (Figure 3b). We also compared the wax components of the leaf cuticles of wild-type and ds8 plants. The levels of very-long-chain fatty acids, such as C20:0, C24:0 and C26:0, were reduced in ds8, whereas the levels of alkanes such as C33:0 and C35:0 were significantly higher in ds8 compared with the wildtype ( Figure 3d). The cell wall components in wild-type and ds8 leaves were also measured. Pectin, hemicellulose I and cellulose levels were significantly lower in ds8 leaves than in wild-type leaves ( Figure S6a). The levels of monosaccharides such as arabinose, mannose, fructose, glucose and galactose were also significantly reduced in ds8 leaves ( Figure S6b).

DS8 affects stomatal density
Observation of leaf ultrastructure revealed that the bulliform cell band on the upper leaf epidermis of ds8 was more wrinkled than that of the wild-type ( Figure 4a). Crosssections of ds8 leaves exhibited shrunken bulliform cells, leaving a groove between two vascular bundles (Figure 4b). The width of the S-ST-BC unit in the leaf epidermis of wild-type and ds8 was measured. The width of the S-ST-BC unit in ds8 leaf was reduced to approximately (d) Transgenic complementation seedlings subjected to drought stress. Scale bar: 5 cm. (e) Water loss rates of excised leaves harvested from transgenic complementation lines at the heading stage. Data are represented as mean AE SD (n = 3). The least significance difference test was applied at 0.05 probability level (Table S4).

DS8 affects stomatal closure and abscisic acid accumulation
The stomatal aperture status on the upper leaf epidermis was examined. In general, the stomatal aperture status was categorized into three types ( Figure 5a): completely open (CO); partially open (PO); and completely closed (CC). The proportion of CO stomata in seedling leaf was significantly higher in ds8 than in the wild-type (approximately 86 and 54%, respectively) under field conditions. Meanwhile, the proportions of PO and CC stomata were significantly lower in ds8 than in the wild-type ( Figure 5b). We then treated wild-type and ds8 leaves with ABA and H 2 O 2 , which can promote stomatal closure. After being opened to a maximum level, the proportion of CO-type stomata was significantly reduced in wild-type leaves during ABA or H 2 O 2 treatment, whereas the degree of COtype stomata in ds8 leaves exhibited an unexpected increase ( Figure 5c). In addition, the ABA levels in wildtype and ds8 leaves during various growth phases in the paddy field were measured. The endogenous ABA content was significantly higher in ds8 than in the wild-type. The ABA content remained high in ds8 leaves throughout the growing season ( Figure 5d). Furthermore, the ds8 mutant exhibited withered leaf tips under natural field conditions ( Figure S7a,b).
During the filling stage, elevated ROS contents were detected in ds8 flag leaves, as revealed by 3,3ʹ-diaminobenzidine (DAB) and nitrotetrazolium blue chloride (NBT) staining (Figure 5e,f). We also detected higher H 2 O 2 production and significantly reduced catalase (CAT) activity in ds8 flag leaves compared with wild-type (Figure 5g,h). Significantly higher rates of electrolyte leakage were detected in ds8 leaves than in wild-type leaves ( Figure 5i). TEM analysis showed that the number of dark osmiophilic globules and starch grains, which usually increased during leaf senescence, significantly increased in ds8 mesophyll cells at the filling stage compared with the wild-type (Figure S7c). The chlorophyll content and photosynthesis rate was significantly reduced in ds8 leaves compared with the wild-type ( Figure S7d,e).

DS8-antisense transgenic lines exhibit increased drought sensitivity
The DS8-antisense transgenic lines were generated. Two antisense transgenic lines with clearly reduced DS8 transcript levels showed increased sensitivity to drought stress (Figure 6a,b). Compared with the wild-type, the proportion of CO-type stomata and ABA content were both higher in the DS8-antisense transgenic seedlings under normal field conditions (Figure 6c Table S5). These results confirmed that DS8 was involved in leaf water loss and foliar ABA accumulation.

Suppressed actin filaments activity causes increased abscisic acid accumulation
We treated 13-day-old wild-type seedlings with 5 lM CB, an inhibitor of actin polymerization; 0.05% dimethylsulfoxide (DMSO) treatment was used for the control. Although 0.05% DMSO treatment led to disordered actin filaments to a certain extent, the number of actin filaments in CB-treated root tips was clearly reduced compared with the control (Figure 7a). The ABA content of the seedling leaves was 33.97 ng g À1 on the 4th day after 0.05% DMSO treatment, whereas it was approximately 141.73 ng g À1 after CB treatment ( Figure 7b). The foliar ABA content increased when the activity of actin filaments was suppressed.
On the other hand, ds8 seedlings were treated with various concentrations of sodium tungstate, an ABA biosynthesis inhibitor (Figure 7c). The ABA content in ds8 seedling leaves decreased with the increase of sodium tungstate concentration. Unexpectedly, the degree of stomatal closure in ds8 leaves increased in response to decreased ABA content, reaching a maximum when the ABA content was similar to that of wild-type (when the concentrations of sodium tungstate was 1.0 mM). With further decline of ABA content in response to a higher concentration of sodium tungstate (1.5 mM and 2.0 mM), the degree of stomatal closure in ds8 leaves turned back to a decrease (Figure 7d).

DISCUSSION
In plants, SCAR/WAVE complex subunits were generally reported to affect cell morphogenesis by regulating actin filaments nucleation and branching. Their functions in drought stress were rarely mentioned (Brembu et al., 2004;Salah et al., 2004;Le et al., 2006;Panteris et al., 2009). In this study, we isolated and characterized the rice mutant ds8, which exhibited increased sensitivity to drought stress. DS8 encoded a NAP1-like protein, a component of the SCAR/WAVE complex. DS8 is reported to encode an interacting protein of LPL2, a PIROGI/Specifically Rac1associated protein 1 (PIR/SRA1)-like protein involved in

DS8 is involved in actin filaments modeling
In Arabidopsis, GNARLED encodes a NAP1 homolog that is reported to positively regulate the function of SCAR/ WAVE-Arp2/3 complex in the process of actin filaments nucleation and branching (Machesky et al., 1999;Deeks et al., 2004;Salah et al., 2004). In this study, DS8 shared high sequence similarity with GNARLED, according to the multiple sequence alignment ( Figure S5). Like most SCAR/ WAVE subunit mutants, ds8 exhibited an abnormal actin cytoskeleton (Figure 2f-h). These results indicated that DS8 encoded a NAP1 homolog that was involved in the nucleation of actin filaments by contributing to SCAR/WAVE-Arp2/3 complex activity.

DS8 is involved in cuticle development
The successful transition of extant land plants from aquatic to terrestrial habitats has largely depended on the evolutionary acquisition of unique adaptations to the terrestrial environment, such as the distinctive epidermal components (Chen et al., 2017). Disorganized actin filaments in plants with SCAR/WAVE complex deficiency results in changes in epidermal trichome (in Arabidopsis)/CP (in rice) and pavement cell morphology (Basu et al., , 2005Deeks et al., 2004;Djakovic et al., 2006;Bai et al., 2015;Rao et al., 2015;Zhou et al., 2016). In our study, we also found epidermal changes in ds8 plants.  (Figure 4d). So, the increased stomatal density in ds8 due to altered bulliform cell morphology was a byproduct of abnormal epidermal cell morphogenesis. Unlike other SCAR/WAVE subunit mutants, the cuticle of ds8 leaves was impaired according to the result of toluidine blue staining (Figure 3a). ds8 exhibited altered cuticle wax components ( Figure 3d) and cell wall deposition ( Figure S6) in leaves. The changes in proportion of cuticular wax components (Figure 3d), which could cause  (Yu et al., 2008;Qin et al., 2011;Mao et al., 2012;Zhu and Xiong, 2013;Gan et al., 2016;Wang et al., 2017), accounted for the damaged cuticle layer in ds8. Some rice mutants with abnormal cuticles have thicker but less dense cuticular layers when compared with wild-type plants (Mao et al., 2012;Gan et al., 2016), which could also be observed in our study (Figure 3b,c). All these results suggested that DS8 not only functioned in epidermal cell morphogenesis like other SCAR/WAVE subunits, but also influenced the development of the cuticle and cell wall. Therefore, DS8 was important for building complete leaf epidermis to protect plants against non-stomatal water loss.

DS8 is required for abscisic acid-induced stomatal closure
Mechanistic studies of stomatal movement have demonstrated that actin filaments dynamics functions downstream of ABA-regulated stomatal movement, and the inhibition of actin filaments nucleation or distribution reduces stomatal closure in response to ABA or H 2 O 2 stimulation (Pei et al., 2000;Zhang et al., 2001;Gudesblat et al., 2007;Hardham et al., 2007;MacRobbie and Kurup, 2007;Gao et al., 2009;Kim et al., 2010;Zhao et al., 2011;Jiang et al., 2012;Li et al., 2014;Munemasa et al., 2015;Cao et al., 2016). So, the presence of abnormal actin filaments in ds8 explained why the mutant leaves exhibited a reduced degree of stomatal closure during normal growth (Figure 5b). The stomatal closure defect in ds8 could not be recovered by exogenous ABA or H 2 O 2 treatment (Figure 5c) because the actin filaments activity downstream of ABA-induced stomatal closure was damaged. Besides, this defect unexpectedly recovered marginally in response to a decline in ABA content and subsequently decreased in response to a further decrease in ABA content (Figure 7c,d). This revealed that the ABA content in ds8 could even enhance the severity of the stomatal closure deficiency when it exceeds a certain content. All these results suggested that DS8 was indispensable for ABA-induced stomatal closure. When DS8 was invalid, the impaired actin filaments activity interfered with the ABA-mediated stomatal closure.
Although the SCAR/WAVE complex and Arp2/3 complex were reported to work together to regulate actin filaments modeling (Pollard, 2007;Yanagisawa et al., 2015), there are still some doubts about their functions. In Arabidopsis, mutants of Arp2/3 complex subunits show reduced sensitivity to ABA-regulated stomatal closure (Jiang et al., 2012;Li et al., 2014). However, the mutant of PIR, a subunit of the SCAR/WAVE complex, shows normal sensitivity to ABA-regulated stomatal closure. This contradiction leaves two explanations. Firstly, the residual SCAR/WAVE complex activity in the PIR mutant is sufficient for the activation of Arp2/3 complex, so the activity of actin filaments is not affected. Secondly, the SCAR/WAVE complex is not involved in ABA-mediated stomatal closure (Isner et al., 2017). In our study, the insensitivity of ds8 to ABA treatment (Figure 5c) prompted us to support the former possibility. Different subunits of the SCAR/WAVE complex may affect the actin filaments activity at different degrees. During ABA-mediated stomatal closure, there is a feedback path. The depolymerization of actin filaments is thought to activate Ca 2+ channels in the plasma membrane, which directly activates NADPH oxidases, leading to H 2 O 2 production (Sagi and Fluhr, 2001;Zhang et al., 2007;Li et al., 2014). We suspected that ABA may be the first factor activated by actin filaments in this feedback path, for ABA could also activate Ca 2+ channels (Munemasa et al., 2015) and the ABA content in ds8 leaf was higher compared with that in wild-type (Figure 5d). To investigate whether the increased ABA content in ds8 leaves was due to restricted actin filaments activity, we treated ds8 plants with CB. The ABA content of the CB treatment group was almost four times that of the control after the treatment (Figure 7b). These results indicated that the increase in ABA content in ds8 was caused by suppressed actin filaments activity. Thus, we propose that the feedback path may be firstly activated by the increase in ABA levels induced by actin filaments dynamics. On the other hand, ABA was reported to promote leaf senescence (Takasaki et al., 2015;Zhao et al., 2016;Miao et al., 2018). A series of alterations detected in ds8 leaves, such as elevated H 2 O 2 content, higher rates of electrolyte leakage, reduced chlorophyll content, decreased photosynthetic rate, etc. indicated that leaf senescence was promoted in ds8 due to the elevated ABA content (Figures 5e-i and S7).
In conclusion, DS8 encoded a NAP1-like protein, a component of SCAR/WAVE involved in actin filaments activity. The dysfunction of DS8 resulted in abnormal development of the leaf epidermis, including impaired cuticles and increased stomatal density. Moreover, the impaired actin cytoskeleton in ds8 led to defective ABA-mediated stomatal closure. All of these defects in leaf traits rendered ds8 suffering excessive water loss and increased sensitivity to drought. The impaired actin filaments in ds8 leaves promoted the accumulation of endogenous ABA, leading to accelerated leaf senescence. In summary, DS8 affects the sensitivity of rice plants to drought stress. Our findings reveal a genetic and molecular mechanism for the involvement of DS8 in the response of rice to drought stress.

Plant materials and culture conditions
Plants were grown in a paddy field at the China National Rice Research Institute in Hangzhou under normal irrigation throughout the growth period (5-8 cm shallow water depth during early vegetative stage; saturated soil condition during heading stage and early filling stage). For ABA content measurement, leaves of wild-type and mutant plants were harvested at the seedling stage, heading stage and 15 days after flowering. Flag leaves of wildtype and mutant plants were collected during the heading stage and filling stage for specific analysis.

Drought tolerance test and leaf water loss rate measurements
For the drought tolerance test in seedlings, germinated seeds were sown in containers of soil, cultured in an incubator (12 h light at 30°C and 12 h dark at 25°C, 70% humidity) and watered regularly. Watering was stopped for 7-10 days during the threeleaf stage, followed by regular watering for 10 days. The test was repeated at least three times and only one picture was taken. The survival rate was scored after recovery. For the drought tolerance test in the field, irrigation was stopped 30 days after the seedlings were transplanted into large outdoor cement-bound plots. Traits were examined 40 days after flowering. For leaf water loss rate analysis, six flag leaves were harvested from six individual plants during the heading stage and incubated in water. The excised leaves were cultured in an incubator at 28°C and 60% humidity after removing the excess water. The leaves were weighed every 10 min. The test was repeated three times, six leaves were used each time.

Morphological observation of actin filaments
Alexa Fluor 488-phalloidin staining of actin filaments was performed as described previously (Olyslaegers and Verbelen, 1998). To observe actin filaments, root tips were excised from seedlings and washed twice with PEM buffer (0.1 M PIPES, 10 mM EGTA, 0.3 M mannitol, 5 mM MgSO 4 , pH 6.9). The tissues were fixed in 4% paraformaldehyde in PEM buffer for 1 h, and washed twice with PEM buffer. The samples were stained with 0.66 mM Alexa Fluor 488-phalloidin (Life Technologies, Eugene, OR, USA) in PEM buffer for 1 h and visualized under a confocal laser-scanning microscope (LSM 700, Zeiss, Oberkochen, Germany).

Observation of cuticular layer and quantification of wax and cell wall components
For TEM observations of the cuticular layer, the middle parts of flag leaves were harvested during the heading stage, fixed in precooled fixation buffer (0.1 M phosphate buffer containing 2.5% glutaraldehyde, pH 7.0) for at least 8 h, washed using phosphate buffer (0.1 M, pH 7.0) and dehydrated using a graded ethanol series. The tissue samples were then transferred to acetone for 20 min, embedded with resin, sectioned, stained with uranyl acetate and lead citrate, and visualized by TEM Hitachi,Tokyo,Japan). For wax components quantification, flag leaves were harvested during the heading stage. Wax components were quantified by the Metabolomics Facility of the Institute of Genetics and Developmental Biology, Chinese Academy of Sciences. The test was repeated four times, and flag leaves collected from six plants were used each time. The content of cell wall components was analyzed according to Zhong and L€ auchli (1993). The test was repeated three times, and flag leaves collected from six plants were used each time. section analysis, the middle parts of flag leaves were collected from wild-type and mutant plants, and fixed in 50% FAA (0.9 M glacial acetic acid, 3.7% formaldehyde and 50% ethanol) overnight at 4°C. The paraffin section analysis was performed according to Ren et al. (2016).

Analysis of stomatal aperture status and abscisic acid treatment
To analyse the stomatal aperture status of wild-type and mutant leaves, the seedlings cultured under field conditions were used. The middle part of the upper leaf epidermis was covered with transparent nail polish and allowed to dry for about 5 min. The nail polish cast, which could provide an impression of the leaf surface, was removed from the leaf and attached to a microscope slide with transparent adhesive tape (Peel et al., 2017). The stomatal apertures status was observed under a 90i light microscope (Nikon, Tokyo, Japan) at 10 9 40 amplification. Stomatal types were defined and analyzed according to Matsuda et al. (2016).
For ABA and H 2 O 2 treatment, germinated seeds were sown in reconstructive 96-well plates and cultivated in distilled water for 3 days (12 h light at 30°C and 12 h dark at 25°C, 70% humidity). The seedlings were transferred to 1/2 Kimura B nutrient solution (Ueno et al., 2009) and cultivated for 10 days without changing the environmental conditions. Leaves were excised from 13-day-old seedlings and incubated in MES buffer (30 mM KCl, 0.1 mM CaCl 2 , 10 mM MES-Tris, pH 6.1) for 90 min in the light (400 lmol m À2 sec À1 ) to induce the maximum level of stomatal opening. Leaves with open stomata (used as the control) were transferred to MES buffer containing 30 lM ABA or 10 À4 M H 2 O 2 for 1 h (Li et al., 2014). The stomatal aperture status was analyzed according to the method mentioned in the previous paragraph.

Measurement of abscisic acid, reactive oxygen species content and electrolyte leakage
To measure ABA content, fresh leaf samples were weighed and ground in liquid nitrogen. The powdered samples were combined with isopropanol-hydrochloric acid extraction buffer for 30 min at 4°C. After the addition of methylene chloride, the suspension was mixed for 30 min at 4°C and centrifuged to isolate the organic phase. The organic phase was dried under liquid nitrogen and dissolved in methyl alcohol, which contained 0.1% formic acid. The final lysate was filtered and used for HPLC-MS/MS. For the analysis of ROS content, leaf tissues were harvested during the filling stage. A rough estimate of superoxide anion and hyperoxide content was performed by NBT and DAB staining . The H 2 O 2 content, CAT activity and electrolyte leakage were determined according to methods mentioned in previous studies (Moradi and Ismail, 2007;Morita et al., 2009).

Measurement of net photosynthetic rate and pigment content
During filling stage, the photosynthetic rates of the flag leaves were measured using a Li-6400 portable photosynthesis system (Li-Cor, Lincoln, NE, USA) under the following conditions: 400 lmol mol À1 ambient CO 2 ; 1200 lmol m À2 sec À1 photosynthetic photon flux density and 500 lmol sec À1 air flow rate. The chlorophyll content was determined by spectrophotometry (Ye et al., 2016). The middle parts of flag leaves were collected from wild-type and mutant plants cut into segments of about 1 cm each. The leaf samples were immersed in 80% acetone for 24 h (26°C in dark) and measured by DU800 ultraviolet spectrophotometer (Beckman, Fullerton, CA, USA).

Sodium tungstate and cytochalasin B treatment
For sodium tungstate treatment, germinated seeds were sown in reconstructive 96-well plates and placed in an incubator (12 h light at 30°C and 12 h dark at 25°C, 70% humidity), cultivated in distilled water for 3 days and in 1/2 Kimura B nutrient solution for 10 days. Then, the 13-day-old mutant seedlings were transferred to 1/2 Kimura B nutrient solution containing 0, 0.5, 1.0, 1.5 or 2.0 mM sodium tungstate for 7 days. Thirteen-day-old wild-type seedlings were transferred to 1/2 Kimura B nutrient solution containing 0 mM sodium tungstate as a control. For CB treatment, 13-day-old wild-type seedlings were transferred to 1/2 Kimura B nutrient solution containing 0.05% DMSO or 5 lM CB (dissolved in 0.05% DMSO). The ABA content and stomatal aperture status were analyzed according to the before-mentioned methods in 'Measurement of ABA, ROS content and electrolyte leakage' and 'Analysis of stomatal aperture status and ABA treatment'.

Construct generation and transformation
For the complementation test, a 14.6-kb genomic fragment containing the entire ORF of LOC_Os08g43130 and its native promoter were amplified from NPB genomic DNA and introduced into binary vector pCAMBIA1300 to generate the proDS8:DS8 plasmid. The plasmid was used to transform ds8 calli by Agrobacteriummediated transformation. A GUS reporter construct containing the native DS8 promoter was introduced into wild-type rice plants to examine the expression pattern of DS8. To determine the subcellular localization of DS8, a p35S:DS8-eGFP construct containing the full-length DS8 CDS without a stop codon was generated. Both the empty p35S:eGFP vector (as a control) and the p35S:DS8-eGFP construct were introduced into wild-type rice protoplasts and calli. A DS8-antisense construct containing the reverse DS8 CDS fragment under the control of the 35S promoter was generated and transformed into the wild-type to produce DS8 antisense transgenic lines. The primer sequences used are listed in Table S3.

Accession numbers
Sequence data from this article can be found in the NCBI website (http:www.ncbi.nlm.nih.gov) under the following accession numbers: Os06g0130900 (Histone H3), Os08g0544500 (DS8). Figure S3. Expression pattern of DS8 and subcellular localization of DS8. Figure S4. DS8 encodes a putative NAP1-like protein. Figure S5. Protein sequence alignment of DS8 and its homologs from several species. Figure S6. Analysis of cell wall components and monosaccharide content. Figure S7. ds8 exhibits withered leaf tips and reduced chlorophyll content. Table S1. Phenotypic data of wild-type and ds8 under normal and dry environmental conditions. Table S2. Reciprocal crosses between ds8 and wild-type indica cultivars. Table S3. Primers used in this study. Table S4. Water loss rates of excised leaves harvested from transgenic complementation lines at the heading stage. Table S5. Water loss rates of excised leaves harvested from WT, ds8 and antisense transgenic plants at the heading stage.