A semi-dominant mutation in OsCESA9 improves biomass enzymatic digestibility and salt tolerance by relatively remodeling cell walls in rice

Yafeng Ye Hefei Institutes of Physical Science Shuoxun Wang Institute of Genetics and Developmental Biology Chinese Academy of Sciences Kun Wu Institute of Genetics and Developmental Biology Chinese Academy of Sciences Yan Ren Hefei Institutes of Physical Science Hongrui Jiang Hefei Institutes of Physical Science Jianfeng Chen Institute of Genetics and Developmental Biology Chinese Academy of Sciences Liangzhi Tao Hefei Institutes of Physical Science Xiangdong Fu Institute of Genetics and Developmental Biology Chinese Academy of Sciences Binmei Liu (  liubm@ipp.ac.cn ) https://orcid.org/0000-0001-9337-1246 Yuejin Wu Hefei Institutes of Physical Science


Background
As one of the most important staple food crops, rice produces large quantities of lignocellulose biomass, which is a key component for the production of bio-energy and bio-based products. However, due to the lignocellulose recalcitrance of the secondary cell wall, the utilization of rice straw for bio-energy is very limited, and the processing of rice straw is very di cult (Himmel et al., 2007;Sathitsuksanoh et al., 2013). Biomass sacchari cation is an important indicator for enzymatic digestion to release soluble sugars, which is one of the major steps for conversion of lignocellulose to ethanol (Wang et al., 2016b). Cellulose is a large molecular polysaccharide composed of β-1, 4-D-glucan chains, which are synthesized individually and form micro brils by intra-and intermolecular hydrogen bonds and Van der Waals forces (McFarlane et al., 2014). The cellulose micro brils are the framework of the plant cell wall, and the hemi-celluloses and pectins or lignins ll in the gaps between the cellulose micro brils and can cross-link cellulose micro brils to form two different types of plant cell wall, namely, the primary cell wall and secondary cell wall (Hofte and Voxeur, 2017). Hence, the cellulose content and crystallinity are key parameters negatively affecting biomass digestibility. Therefore, the genetic modi cation of cellulose is very useful for reducing lignocellulose recalcitrance for enhancing biomass sacchari cation and improving rice straw decomposition rates (Vermerris and Abril, 2015).
In higher plants, cellulose synthesis is catalyzed by the plasma membrane-localized cellulose synthase complexes (CSCs) using uridine diphosphate (UDP)-glucose as substrate (McFarlane et al., 2014). The functional CSC requires the cooperation of at least three distinct cellulose synthase (CESA) isoforms, which are encoded by different CESA genes (Taylor et al., 2003). Since the rst higher plant CESA gene was cloned from cotton in 1996 (Pear et al., 1996), many CESA genes from different plant species have become publicly available (Joshi and Mans eld, 2007). All of the CESA proteins in different plant species are 986-1088 amino acids in length and share common domains and motifs, such as the zinc nger domains at the N-terminus, eight transmembrane domains (TMDs) and a central cytoplasmic domain with D,D,D,QXXRW motif between TMD2 and TMD3. The current model of cellulose synthesis predicts that the rst two conserved aspartic acid residues coordinate UDP, the third plays a role in providing the catalytic base for the glucan extension, and the QXXRW residues function as a binding site for the terminal glucan residues of the chain (Somerville, 2006). In rice, OsCESA4, OsCESA7 and OsCESA9 comprise the CSCs necessary for secondary cell wall cellulose synthesis (Tanaka et al., 2003). To date, numerous mutants in these three genes have been identi ed, all of these mutants show the brittle culm (bc) phenotype due to decreases in cellulose content (Kotake et  able to maintain normal plant growth and grain production in plants. For example, the bc13 mutant with one amino acid alteration (G101K) in the N-terminus of OsCESA9 shows normal plant growth and cadmium tolerance, despite a reduction in cellulose (Song et al., 2013). The Osfc16 mutant with substitutions (W481C, P482S) at the P-CR conserved site in OsCESA9 shows slightly affected plant growth and maintains grain yields similar to its wild type, but not only that, it also enhances biomass sacchari cation due to much improved lignocellulose features such as cellulose DP & CrI and nanocellulose bers assembly (Li et al., 2017). Bc6, a semi-dominant brittle culm mutant harbors a substitution (R588G), which is located 203 amino acids before the catalytic motif QXXRW, and also shows a brittle culm without any pleiotropic phenotype such as dwar sm and withering (Kotake et al., 2011). Several OsCESA9 mutants have been identi ed, but only the Bc6 is a semi-dominant brittle culm mutant (Kotake et al., 2011). To further interpret why only OsCESA9 mutants maintain normal plant growth and grain yield, more OsCESA9 mutants with conserved-site mutations need to be discovered, particularly semi-dominant mutants, in order to elucidate the function of the OsCESA9 conserved-site domain.
Soil salinity is one of the major stresses adversely affecting plant growth and crop productivity (Munns and Tester, 2008). The  (Chen et al., 2005). In this study, a novel semi-dominant brittle cum (sdbc) mutant, which harbors a substitution (D387N) at the rst conserved aspartic acid residue of the CESA9 protein, was obtained by heavy ion beam treatment of the japonica cultivar Wuyunjing7. The WT/sdbc1 heterozygotes show a mild brittle culm phenotype without any the appearance of pleiotropic phenotypes and also exhibit improved salt tolerance. Further evidence revealed that the mutation OsCESA9 D387N can compete with wild type OsCESA9 to form an abnormal CSC with OsCESA4 and OsCESA7.

Plant materials and growth conditions
The sdbc1 mutant was obtained from a japonica cultivar Wuyunjing7 by heavy ion beam treatment (the energy of carbon ions is 80 MeV/nucleon, irradiation dose is 120Gy). The F 1 plants and genetic analysis population were generated by the cross between sdbc1 and its wild type (WT) Wuyunjing7. An F 2 mapping population was generated from the cross between sdbc1 and an indica cultivar, 93-11. All rice plants used in this research were grown in the experimental elds at the Institute of Technical Biology and Agriculture Engineering, Hefei Institute of Physical Science, Chinese Academy of Sciences (Hefei, China) and Sanya (Hainan province, China) in the natural growing season.

Measurement of extension force and microscopy
Extension force of the second internodes and ag leaves of WT, sdbc1 and F1 plants were determined according to (Ye et al., 2015). The maximum force required to break apart the internodes and leaves were considered as the mechanical strength of WT, sdbc1 and F1 plants at heading stage.
For transmission electron microscopy, the second internodes of WT, sdbc1 and F1 plants were cut with a razor and immediately postxed in 70% ethanol (V/V), 5% acetic acid (V/V), and 3.7% formaldehyde (V/V) mixture for at least 2 hours. Samples were dried to the critical point, sputter-coated with gold, and observed with a scanning electron microscope (S570; Hitachi, Tokyo, Japan). The thickness of sclerenchyma cell wall was measured by Image J software.

Phenotypic evaluation
Several agronomic traits, including plant height, number of tillers per plant, number of grains per panicle, panicle length, setting rate, 1,000-grain weight and lodging index were chosen for production evaluation. The plant height was measured from the ground surface to the tallest panicle. The number of tillers per plant was the number of effective tillers with corresponding panicle has 10 or more grains. The number of grains per panicle was the total number of grains per panicle. The panicle length was the distance from length top to the rst internode top. The setting rate was scored as the number of full grains per panicle divided by the number of grains per panicle. The 1,000-grain weight was calculated based on random number of grains (> 100) and then converted to 1,000 grain weight.
The plant lodging index was analyzed according to (Li et al., 2017). All traits were measured at full senescence. The data from fty plants in each background in 2017 and 2018 was used for analysis. The independent sample t-test program SPSS 10.0 for Windows was used to compare the mean values.
Cell wall composition analysis and sacchari cation assays The second internodes of WT, sdbc1 and F1 plants at mature stage were used to prepare alcohol-insoluble residues (AIRs) of the cell walls. De-starched AIRs were produced as previously described (Zhou et al., 2009 (Updegraff, 1969).
For the sacchari cation assay, 1mg of AIR was boiled to deactivated endogenous enzymes. After cooling to room temperature, the digestion was performed according to (Gao et al., 2017). The released sugars in the supernatant were measured by reading the A 540 on an ELISA reader (Tecan) as described previously (Gao et al., 2017).

Map-based cloning
The sdbc1 locus was mapped and cloned using 5014 homozygous mutant F 2 plants, which were selected from the population of sdbc1 and 93-11. Molecular markers distributed throughout the whole rice genome were used for sdbc1 locus rough mapping. Molecular markers for ne mapping were developed to narrow down the mutated locus to a 45 kb region on chromosome 9. The corresponding DNA fragments in this mapping region were ampli ed from mutants and WT plants using KOD DNA polymerase (TOYOBO, http://www.toyobo.co.jp/e/bio) and sequenced by BioSune company (http://www.biosune.com). For complementation of sdbc1 mutant, the full-length sdbc1 mutant coding sequence, a 2.5-kb fragment of upstream sequence from ATG and 0.2-kb fragment of downstream sequence from TGA were inserted into binary vector pCAMBIA2300 between the EcoRI and BamHI sites to generate the construct pSDBC1::sdbc1 (pSdbc1F), which was introduced into WT by the Agrobacterium-mediated transformation procedure as described previously (Ye et al., 2018). De ne CAPS marker for identi cation of sdbc1 mutation background was also developed by using enzyme AhdI (NEB, www.neb.com)

Expression analysis
Total RNA was extracted from various rice tissues using TRIzol reagent (Invitrogen), as described previously (Wadsworth et al., 1988). The complementary DNA (cDNA) was synthesized from total RNA using a reverse transcriptional kit (TransGen, http://www.transgen.com.cn/). Quantitative RT-PCR was performed using relevant primers and qRT-PCR kit (TransGen, http://www.transgen.com.cn/) on a quantitative 7500 PCR system (ABI). All assays were repeated at least three times, the OsActin1 (LOC_Os03g50885) gene was used as an internal control. The data were analyzed according to (Wong and Medrano, 2005). All the primers used in this study are listed in Supplementary data at RICE online.

Subcellular localization analysis
To observe OsCESA9 and OsCESA9 D387N subcellular localization, GFP was fused to their C-terminus and inserted into the pCAMBIA1300 driving by CaMV (Cauli ower mosaic virus) 35S promoter. The N. benthamiana leaves expressing GFP positive OsCESA9 and OsCESA9 D387N were observed using confocal laser scanning microscope (Leica TCS SP5).

Yeast two-hybrid assays
Yeast two-hybrid assays were performed by mating-based split ubiquitin system (mbSUS) according to (Obrdlik et al., 2004). The OsCESA9 and OsCESA9 D387N were cloned into pX-NubWTgate (182) vector, and transformed into yeast strain AP5. Transformants were selected on synthetic dropout (SD) media lacking tryptophan (T). OsCESA4 and OsCESA7 were cloned into pMetYCgate (184) vector, and transformed into yeast strain AP4. Transformants were selected on SD lacking leucine (L). The AP4 and AP5 suspensions were mixed and plated on YPD. After 12 h at 30°C, cells were selected for diploids by replica plating on SD lacking tryptophan (T) and leucine (L), and incubated at 30°C for 2-3 days. For growth assays, diploid cells were spotted on an SD medium lacking tryptophan (T), leucine (L), histidine (H) and adenine (A). Growth was monitored for 3-5 days. The β-galactosidase activity assays were performed for detecting the interaction intensity according to the manufacturer's protocol (Takara Bio Inc.).

BiFC assays
For the BiFC assays, the full length cDNAs of OsCESA4,OsCESA7, OsCESA9 and OsCESA9 D387N were cloned into serial pSPY vectors containing either N-or C-terminal enhanced yellow uorescent protein fragments via Gateway cloning technology. These constructs were then introduced into A. tumefaciens strain GV3101 and co-in ltrated into the N. benthamiana leaves. The injected leaves were detached after 2 days to observe uorescence with a confocal laser-scanning microscope (TCS SP5; Leica).

Western blot
The extraction of total membrane protein and separation of the plasma membrane (PM) and endomembrane fractions were Determination of the MDA content and relative electrical conductivity The content of MDA was measured as previously described (Ouyang et al., 2010). In brief, about 1 g leaves was homogenized in 10 mL of 10% trichloroacetic (TCA) followed by centrifugation at 5000 g for 10 min. Subsequently, 2 mL of the supernatant was mixed with an equal volume of thiobarbituric acid (TBA) and boiled for 15 min and the quickly cooled on ice. The absorbance valus at wavelength of 450, 532 and 600 nm were measured on a plate reader (Tecan). The nal quantity of MDA (μmol/L) was calculated by using the following formula: 6.45 * (A 532 -A 600 ) -0.56 * A 450 .
The relative electrical conductivity (REC) was measured according to (Cao et al., 2007). The washed leaves were cut into 0.5 cm pieces and immersed in a 50 mL test tube containing 25 mL deionized water for 30 min, and then the conductivity of the solution was measured using a conductivity meter (DDS-11A). After boiling the samples for 10 min, their conductivity was measured again when the solution was cooled to room temperature. The relative electrical conductivity (REC) was calculated by using the following formula: REC = C 1 /C 2 *100%. Where C 1 and C 2 are the electrolyte conductivities measured before and after boiling, respectively.

Measurements of ion contents
Measurements of ion concentration in rice tissues were performed as described elsewhere (Fang et al., 2018). The samples of various tissues were treated at 105 °C for 1 h. Powered samples were digested with ultrapure nitric acid for 1 day, and the digested samples were boiled at 95 °C for 10 min three times. Ion contents were determined using an inductively coupled plasma optical emission spectrometer (ICP-MS, Agilent 7700 series, USA).

Statistical analysis
All data were analyzed in SPSS version 20.0. Student's t-test was used to compare pairs of means, whilst comparisons between multiple groups were performed using ANOVA followed by Duncan's multiple range test.

Accession numbers
Sequence data used in this manuscript can be found in the rice genome annotation database (http://rice.plantbiology.msu.edu) under the following accession numbers:

Results
Sdbc1 is a semi-dominant brittle culm mutant and shows a reduction in mechanical strength and cellulose content  (Table S1). The homozygous sdbc1 mutant (termed as sdbc1) plants showed easily broken culms and leaves, but the WT/sdbc1 heterozygous F 1 (termed as F1) plants only exhibited mild brittle culm phenotypes. Except for the brittle culm phenotype, both were morphologically indistinguishable from the WT during the mature stage (Figure 1a, b, c). In the seedling stage, sdbc1 plants showed an easily broken and slender phenotype, but the F1 plants only showed a slender but not easily broken phenotype ( Figure S1). We quantitatively measured the force required to break the second internode of the culms and the ag leaves in the WT, sdbc1, and F1 plants. The force to break the culms of sdbc1 and F1 plants was reduced to about 30% and 60% of that to break the culms of the WT, respectively (Figure 1d), and the force to break the leaves of sdbc1 was only about a quarter of that required to break the leaves of WT, but the force to break the leaves of the To understand the underlying cause of the observed brittleness, scanning electron microscopy (SEM) was performed to observe the second internode cross-sections of the WT, sdbc1 and F1 plants. The observations revealed that the sclerenchyma cell walls, the thickness of which is highly correlated with mechanical strength, were obviously thickened in the WT plants (Figure 2a, b, g). In contrast, the thickness of sclerenchyma cell walls in the sdbc1 plants was reduced obviously (Figure 2c, d, g). The thickness of the sclerenchyma cell walls in the F1 plants was between the WT and sdbc1 (Figure 2e, f, g). No obvious differences were observed in the parenchyma cell walls (Figure 2a, c, e, h). The defects in mechanical strength and wall structure suggested that the cell wall composition in the sdbc1 and F1 plants may be altered. We therefore examined the cell wall composition in the second internodes of the WT, sdbc1 and F1 plants at the mature stage. As shown in Table 1, the cellulose content of the sdbc1 and F1 internodes was decreased by about 30% and 14% that of the WT, respectively, whereas the neutral sugar content derived from non-cellulosic polysaccharides was generally increased in sdbc1 and F1 samples, especially for arabinose (Ara) and xylose (Xyl), the two major sugars of hemi-cellulose arabinoxylan at a signi cantly high level (Table 1), we also detected the total hemi-cellulose content of the sdbc1 and F1 internodes higher than that of WT ( Figure S2a), but pectin and starch content were not signi cant different ( Figure S2b, c).
Taken together, the Sdbc1 is a semi-dominant brittle culm mutant, and its defects in mechanical strength are correlated with thin sclerenchyma cell walls and lower cellulose content and higher hemi-cellulose content in the Sdbc1 mutant.
A conserved-site mutation in OsCESA9 results in Sdbc1 phenotypes To investigate the molecular basis of the above phenotypes, a map-based cloning approach was performed to isolate the SDBC1 gene. We used the F 2 population generated by crossing sdbc1 with 93-11, a WT polymorphic indica variety. The sdbc1 locus was located between molecular markers ISR14 and ISR15 on chromosome 9, and further pinpointed within an approximate 45-kb region between markers A7 and A8 (Figure 3a). The 45-kb region contains 6 putative open-reading frames (ORFs), in which the most possible candidate ORF Os09g25490 that encodes OsCESA9 involved in secondary cell wall cellulose biosynthesis was selected for sequencing ( Figure 3b). Sequencing of this ORF in sdbc1 revealed one base pair mutation at position 1629, which changes GAC to AAC and causes a substitution at the 387th amino acid residue from aspartic acid (Asp, D) to asparagine (Asn, N) (Figure 3b). The sequencing atlas showed that mild brittle culm plants were bimodal at the mutation site ( Figure S3), which further suggested that Sdbc1 was a semi-dominant brittle culm mutant. This mutation is at the rst conserved aspartic acid residue domain (Figure 3c), which is fully conserved in all of the CESA family proteins ( Figure S4). To con rm that Os09g25490 corresponds to the sdbc1 locus, a 5.6-kb DNA fragment containing the 2.5-kb putative promoter and the coding region from sdbc1 was cloned into vector pCAMBIA2300 to generate the plasmid pSdbc1F (Figure 3d), which was introduced into the WT plants. All 10 of the transgenic lines showed the mild brittle culm phenotype (Figure 3e). The mutated site also was con rmed by a CAPS marker, digested by AhdI (Figure 3f). We therefore concluded that Os09g25490 is SDBC1.
The CESA9 D387N mutation enhances biomass sacchari cation In recent years, rice straw has been highlighted as an important material for biofuel production, the high cellulose content and crystallinity determine lignocellulose recalcitrance, leading to costly biomass processing ( improve enzymatic sacchari cation. We detected whether the CESA9 D387N mutation also enhances biomass sacchari cation. We examined the sacchari cation e ciency of lignocellulosic material derived from WT, sdbc1, and F1 plants. The sugar yields were signi cantly higher in sdbc1 and F1 than that of WT ( Figure 4). These results suggest that the OsCESA9 D387N mutation can enhance biomass enzymatic sacchari cation.
The CESA9 D387N mutation does not affect the expression pattern of OsCESA9 or its subcellular localization The missense mutation (D387N) in sdbc1 occurs at the rst conserved aspartic acid residue. Among the various mutations in OsCESA isoforms, the mutation of this aspartic acid has not been reported. To determine the effect of this important amino acid residue at the molecular level, we rst detected whether the expression level and pattern of OsCESA9 had been affected in sdbc1 and F1 plants. Quantitative real-time polymerase chain reaction (qRT-PCR) analysis revealed that the expression level and pattern of OsCESA9 had not been affected in the WT, sdbc1 and F1 plants (Figure 5a). Given that OsCESA9 needs to interact with OsCESA4 and OsCESA7 to form functional CSCs for secondary cell wall cellulose synthesis, the expression levels and patterns of OsCESA4 and OsCESA7 were also detected by qRT-PCR. There were no obvious differences in the WT, sdbc1 and F1 plants ( Figure S5).
The correct subcellular localization of a protein is important for it to function normally. To determine whether the conserved-site mutation in OsCESA9 affects its subcellular localization, we expressed the WT form OsCESA9-GFP and mutation form OsCESA9 D387N -GFP in Nicotiana benthamiana leaves. Confocal laser scanning microscope observation revealed that uorescent signals of OsCESA9-GFP and OsCESA9 D387N -GFP were detected in the plasma membrane (Figure 5b), which was consistent with the subcellular localization of OsCESA9.
Taken together, the CESA9 D387N mutation does not affect the expression pattern of OsCESA9 or change its protein localization.
OsCESA9 D387N can compete with OsCESA9 to interact with OsCESA4 and OsCESA7 As mentioned above, OsCESA9 needs to interact with OsCESA4 and OsCESA7 to form functional CSCs. To determine whether OsCESA9 D387N affects the interaction with OsCESA4 and OsCESA7, the mating-based split ubiquitin system (mbSUS) for detecting membrane protein interaction was performed. The yeast two-hybrid (Y2H) assay showed that OsCESA9 and OsCESA9 D387N could directly interact with OsCESA4 and OsCESA7 in vitro (Figure 6a), and the β-galactosidase activity assay indicated that there was no signi cant change in the interaction intensity of OsCESA9 D387N with OsCESA4 and OsCESA7, compared with OsCESA9 ( Figure S6).
Thus, OsCESA9 D387N can also interact with OsCESA4 and OsCESA7 to form CSCs.
Given that the OsCESA9 D387N is a semi-dominant mutation and OsCESA9 D387N can interact with OsCESA4 and OsCESA7, we speculated that the mutation OsCESA9 D387N will compete with WT OsCESA9 for physical interaction with OsCESA4 and OsCESA9. To verify this hypothesis, the split re y luciferase complementation (SFLC) assays were performed, the results showed that the interaction intensity of OsCESA4 or OsCESA7 with OsCESA9 is stronger in the absence of OsCESA9 D387N than that of in the presence of OsCESA9 D387N , even though they all have the same protein level (Figure 6b). We further quanti ed their interaction intensity by analyzing LUC activity, the results showed that the LUC activity signi cantly decrease in the presence of OsCESA9 D387N (Figure 6c).
These suggest that the interactions of OsCESA9 with OsCESA4 and OsCESA7 were weakened in the presence of OsCESA9 D387N (Figure 6b, c). Thus, we reasoned that OsCESA9 D387N is able to compete with OsCESA9 for interaction with OsCESA4 and OsCESA7.
The CSCs are located in the plasma membrane where they perform their function. We further tested whether OsCESA9 D387N altered the subcellular localization of the protein complex with OsCESA4 and OsCESA7. Bimolecular uorescence complementation (BiFC) assays were performed to verify that OsCESA9 D387N directly interacts with OsCESA4 and OsCESA7 on the plasma membrane, consistent with its WT form OsCESA9 (Figure 6d).
In combination, these results suggest that OsCESA9 D387N can compete with OsCESA9 for interaction with OsCESA4 and OsCESA7, but it does not affect the subcellular localization of the OsCESA4/7/9 D387N complex.
The OsCESA9 D387N mutation does not affect secondary wall CSC tra cking Our current understanding of cellulose synthesis suggests that CESAs are assembled into CSCs in either the endoplasmic reticulum (ER) or the Golgi apparatus and tra cked by vesicles to the plasma membrane (PM) (McFarlane et al., 2014). We aimed to determine whether the missense mutation in sdbc1 affects the intracellular tra cking of OsCESA4, OsCESA7 and OsCESA9. First, the amount of OsCESA4, OsCESA7 and OsCESA9 proteins detected by western blot analysis did not differ much (Figure 7a). We then examined the distribution and/or abundance of OsCESA4, OsCESA7 and OsCESA9 between the PM and endo-membrane systems in WT and sdbc1 plants. We therefore separated proteins of the plasma membrane (PEG fraction) and endo-membranes (DEX fraction) and probed them with corresponding polyclonal antibodies. The western blot results showed that there was no signi cant difference in the protein levels of OsCESA4, OsCESA7 and OsCESA9 between the WT and sdbc1 plants, wherever in the endomembrane (Figure 7b) or PM ( Figure 7c). Thus, the OsCESA9 D387N mutation does not affect secondary cell wall CSC protein intracellular tra cking or abundance at the PM.

OsCESA9/OsCESA9 D387N heterozygous plants have better salt tolerance
Soil salinity is one of the major stresses adversely affecting plant growth and crop productivity (Munns and Gilliham, 2015). As previously reported, modulating cell wall cellulose synthesis may be one of the main adaptations of plants to osmotic stresses, such as salinity stress (Chen et al., 2005). To determine whether the OsCESA9 D387N mutation improves plant tolerance to salt, salinity stress was applied to sdbc1 and F1 plants. One-month-old WT, sabc1 and F1 plants were treated with either 0 or 200 mM NaCl for 1 week. These plants grew normally under normal hydroponic culture (Figure 8a). However, the F1 plants grown in 200 mM NaCl demonstrated signi cant differences in salt sensitivity in terms of seedling growth (Figure 8b). Moreover, the F1 plants exhibited a smaller decline in seedling height and biomass under salt stress than the WT and sabc1 plants (Figure 8c, d). Additionally, the survival rates of the F1 plants 1 week after 200 mM NaCl treatment were signi cantly higher than those of the WT and sdbc1 plants (Figure 8e). To further con rm phenomenon, three-week-old WT, sdbc1 and F1 plants were subjected to 100 mM NaCl for one week, the survival rates of the F1 plants were signi cantly higher than those of the WT and sdbc1 plants ( Figure S7). Together, these results suggested that the OsCESA9/OsCESA9 D387N heterozygous genotype might enhance salt stress tolerance.
Physiological analyses were performed to reveal some mechanism insights into the salt tolerance phenotype of the  Figure S8a). However, after 200 mM NaCl treatment, the F1 plants had very few brown H 2 O 2 spots within the total leaf area, whereas the WT and sdbc1 exhibited more brown areas than those of the F1 plants ( Figure S8a). Relative electrical conductivity (REC) is an indicator of cell membrane injury, and malondialdehyde (MDA) is an indicator of oxidative attack on membrane lipids. Leaves from F1 plants accumulated signi cantly lower MDA content and REC than those of WT and sdbc1 plants under salt stress ( Figure S8b, c). Proline accumulation is linked with stress tolerance, therefore we detected proline content among these plants.
Although proline content increased dramatically under salt stress, there was more in F1 plants compared with that in WT and sdbc1 plants ( Figure S8d). These results suggested that the OsCESA9/OsCESA9 D387N heterozygous plants show more salt tolerance.
Salinity stress mainly in uences Na + and K + distribution and homeostasis and high Na + accumulation directly reduces carbon xation and biomass production in plants (Munns and Tester, 2008). To test if the OsCESA9/OsCESA9 D387N heterozygous genotype contributed to maintaining the ion balance in plants, we determined the content of Na + and K + in the shoots and roots of WT, sdbc1 and F1 plants with and without salt treatment. There were lower Na + levels in the shoots and roots of F1 plants under salt treatment, compared with the WT and sdbc1 plants (Figure 9a, b). The K + content exhibited no signi cant differences in the shoots and roots of these plants under salt stress. Moreover, the K + /Na + ratio in the shoots and roots of the F1 plants was higher than that of the WT and sdbc1 plants under salt stress. Thus, the OsCESA9/OsCESA9 D387N heterozygous genotype affected K + /Na + homeostasis and altered Na + and K + distribution between the shoots and roots. These resulted in the OsCESA9/OsCESA9 D387N heterozygous plants demonstrating greater salt tolerance.
To determine the reason for the enhanced salt tolerance in OsCESA9/OsCESA9 D387N heterozygous plants, we examined the expression of related genes in WT, sdbc1 and F1 organs in the control and salt treatments by qRT-PCR. OsSOS1 encodes a rice plasma membrane Na + /H + exchanger protein and is induced by salt stress (Shi et al., 2002). As shown in Figure 9c, the expression level of OsSOS1 was signi cantly higher in the sdbc1 and F1 plants than the WT under both conditions. OsHKT1;5, a major gene contributing to Na + removal from the xylem and salt tolerance (Ren et al., 2005), was up-regulated in sdbc1 and F1 plants under both conditions ( Figure   9c). These results are consistent with the decrease in Na + content in shoots and roots of F1 plants. Although no obvious differences were observed in the shoots and roots of these plants under salt stress, the expression of OsKAT1, encoding the K1 channel protein (Obata et al., 2007), was up-regulated in F1 plants (Figure 9c). Under control conditions, the expression of OsKAT1 was signi cantly up-regulated in sdbc1 and F1 plants (Figure 9c). This result is consistent with the higher K + content in sdbc1 and F1 plants without salt stress. Together, the expression of the above genes is consistent or in con ict with the expectation according to their functions reported previously. The diverse alterations in gene expression indicated that the examined genes are indirectly responsible for the increased salt tolerance in OsCESA9/OsCESA9 D387N heterozygous plants.
We also investigated OsCESA9 expression in WT plants with different concentration NaCl supply. The qRT-PCR assay revealed that the expression of OsCESA9 in the shoots and roots was remarkably down-regulated with the increase in salt concentration ( Figure S10). These results suggest an association between OsCESA9 and salt stress. To determine whether salt stress affect integrity of the cell wall, SEM was performed to observe the leaf sheath cross-sections of WT, sdbc1 and F1 plants before and after salt stress. The observations revealed that the thickness of sclerenchyma cell wall among all plants were obviously decreased after salt stress ( Figure   S11). These results suggest an association between cell wall biosynthesis and salt stress.
The OsCESA9/OsCESA9 D387N heterozygous plants are suitable for use in straw return to the eld It was previously reported that brittle culm mutants have potential prospect for straw treatment due to their lower cellulose contents and culm easily broken during harvest (Ye et al., 2015). Given that the OsCESA9/OsCESA9 D387N heterozygous plants also show lower cellulose contents and easily broken culms without any morphological abnormalities, including lodging resistance (Figure 1f-k and S12), we considered that they may be suitable for use in the return of straw to the eld. To assess this possibility, we carried out eld harvesting experiments. The results showed that the internodes of OsCESA9/OsCESA9 D387N heterozygous plants were easily crushed by a rice combine harvester and evenly distributed into the eld ( Figure S12b, d, f), compared to WT ( Figure S13a, c, e). Together, the OsCESA9/OsCESA9 D387N heterozygous plants are suitable for use in the return of straw to the eld.

Discussion
OsCESA9 D387N causes a dominant negative effect on inferior mechanical strength In rice, OsCESA4, OsCESA7 and OsCESA9 are required for secondary cell wall cellulose biosynthesis (Tanaka et al., 2003). Mutations in any of these CESA genes lead to a brittle culm phenotype and a reduction in cellulose content (Song et al., 2013;Wang et al., 2016a;Zhang et al., 2009). In this study, we isolated a novel OsCESA9 allele, sdbc1, with a brittle culm and lower cellulose content. To date, many oscesa9 mutants have been identi ed (Figure 3c), but only the Bc6 mutant caused by a missense mutation, R588G, shows a semi-dominant brittle phenotype. Here, sdbc1 was caused by a missense mutation, D387N, in the rst conserved aspartic acid residue of OsCESA9, which also resulted in the semi-dominant brittle phenotype. As the other OsCESA9 alleles are recessive, this suggests that the semi-dominant phenotypes of Bc6 and Sdbc1 are not caused by a dosage effect. We expressed the sdbc1 mutant cDNA in WT plants resulted in a mild brittle culm phenotype (Figure 3e). The introduction of the bc6 mutant gene into WT plants previously caused decreased cellulose content and brittle phenotype (Kotake et al., 2011). Therefore, mutation of the two sites can interfere with the function of WT OsCESA9.
OsCESA9 shares common domains and motifs, such as zinc nger domains at the N-terminus, the D, D, D, QXXRW motif and eight TMDs, with all plant CESAs. Missense mutations located at distinct sites of OsCESA9 exhibit different phenotypes, indicating that these may play special roles in assigning functions to speci c domains or motifs of the CESA proteins. The sdbc1 mutation mapped to the rst conserved aspartic acid residue, which is critical for substrate binding. We also veri ed that the sdbc1 mutation does not affect its gene expression pattern, subcellular localization and interaction with OsCESA4 and OsCESA7. The OsCESA9 D387N mutation also does not affect the tra cking of secondary cell wall CSCs. Therefore, we speculated that the possible mechanism for the dominant negative effect of the sdbc1 mutation is OsCESA9 D387N competing with the WT OsCESA9 for interaction with OsCESA4 and OsCESA7, and then forming nonfunctional or partially functional CSCs. In the sdbc1 homozygous plants, only in the presence of OsCESA9 D387N , the OsCESA9 D387N interacts with OsCESA4 and OsCESA7 for polymerizing nonfunctional or partially functional CESA4/7/9 D387N CSCs, affecting cellulose synthesis ( Figure 10). The normal CESA4/7/9 and abnormal CESA4/7/9 D387N CSCs simultaneously exist in the Sdbc1 heterozygous plants (Figure 10), so these exhibit the semi-dominant brittle phenotype. Unlike the Bc6 mutation, OsCESA9 R588G , which is located near to P586 corresponding to P557 of AtCESA7 that is altered in the semi-dominant fra5 mutant of Arabidopsis (Kotake et al., 2011;Zhong et al., 2003). A study on the Arabidopsis fra5 mutant suggested that the missense P557T mutation of AtCesA7 affects the interaction between CESA proteins or between CESA and other cellular components. Hence, Bc6 may share the common mechanism for the dominant negative effect, with the fra5 mutant (Kotake et al., 2011). Therefore, although the Sdbc1 and Bc6 mutants exhibited a semi-dominant brittle phenotype, the mechanism for the dominant negative effect differs. The sdbc1 and bc6 mutation occurs in the region between TMD2 and TMD3, but not all mutations that occur in this region exhibit semi-dominant phenotype, such as the osfc16 and bc88 mutations located between sdbc1 and bc6 mutation sites (Figure 3c), However, these two mutants are recessive and show obviously different phenotypes (Li et al., 2017, Rao et al., 2013. The osfc16 mutant with substitutions (W481C, P482S) in OsCESA9 shows slightly affected plant growth and signi cantly reduced cellulose crystallinity and thinner secondary cell walls (Li et al., 2017). The bc88 mutant, which harbors the substitution (P421L) in OsCESA9 exhibits a diversity of pleiotropic phenotypes, including a brittle culm phenotype, dwar sm, withered leaf tips at the seedling stage, and an 18-d delay in heading date at the mature stage (Rao et al., 2013). Studies of these mutants suggest that the different conserved sites in the central cytoplasmic domain play a distinct role in cellulose synthesis. Therefore, more novel OsCESA9 alleles have been identi ed, which will provide insight into how CESAs function and what biological functions cellulose plays in plants.

The OsCESA9 D387N mutation confers salt tolerance in heterozygous plants
The plant cell wall plays a critical role in development and in adaptation to abiotic and biotic stresses (Keegstra, 2010). Soil salinity is a key abiotic stress affecting crop productivity worldwide (Munns and Gilliham, 2015).  (Cao et al., 2011). We speculate the ROS may be the indirect and nal responses, and it may be another possible that the altered cell wall property improves cell osmic ability for salt tolerance. As a cellulose synthase mutant, both sdbc1 homozygous and Sdbc1 heterozygous plants show increased soluble sugar contents, but only Sdbc1 plants exhibit salt tolerance. These may indicate that a proper reduction in cellulose content is also very important for salt tolerance. The severe reduction in cellulose content in the sdbc1 homozygous plants possibly explains why they show higher level Na + accumulation, although the expressions of OsSOS1 and OsHKT1;5 were also up-regulated.
Previous studies have identi ed many types of cesa mutants in response in salt, drought and other stresses. For instance, Arabidopsis cesa8 lew2-1 (L802F) (Chen et al., 2005) and cesa3 ixr1-1 (G998D) (Scheible et al., 2001) confer drought and isoxaben resistance to the mutant plants. Rice bc13, another OsCESA9 allele, with a mutation at the 101 st residue (G101K) in the N-terminal region conserved with acidic amino acids just after the zinc ngers, confers cadmium (Cd) tolerance and lower Cd accumulation in the grain (Song et al., 2013). In our research, we also found that the retention of some heavy metals, such as Cd, lead (Pb), copper (Cu) in the grain of sdbc1 homozygous and Sdbc1 heterozygous plants was lower than in the WT plants ( Figure S14). Hence, the sdbc1 mutant may also have heavy metal resistance, but these assumptions need further study. This mutation could be introduced into other rice varieties for breeding low heavy metal accumulating rice, especially for the hybrid rice, as the Sdbc1 heterozygous plants simultaneously have slat tolerance and better agronomic traits. We designed the CAPS marker for distinguishing the WT and sdbc1 background, which can be applied for the identi cation of the mutation point for marker-assisted selection.
The OsCESA9 D387N mutation has potential prospect for residue management As one of the most important staple food crops, rice produces signi cant quantities of agronomic biomass residue every year. The processing of these residues has been a challenge due to the lignocellulose recalcitrance of the secondary cell wall (Alam et al., 2019). Straw burning is preferred by farmers, as it is economical and convenient, but it causes environmental problems. The high cellulose content of the cell wall of rice straw directly leads to slow decomposition (Tian et al. 1992 (Figure 4). Hybrid rice not only has higher grain yield, but it also produces greater biomass than that of conventional rice varieties (Jiang et al., 2016). Given that the OsCESA9 D387N mutation results in a dominant negative effect on cellulose biosynthesis, the OsCESA9 D387N mutation can be introduced into elite sterile lines or restorer lines to further resolve the straw processing issues of hybrid rice.

Conclusions
In this study, we isolated a semi-dominant brittle culm (sdbc) phenotype. Map-based cloning revealed a substitution (D387N) at the rst conserved aspartic acid residues of OsCESA9 corresponding to the mutant phenotypes. Further experiments indicate that mutation OsCESA9 D387N may compete with WT OsCESA9 for interaction with OsCESA4 and OsCESA7 and further to form nonfunctional or partially functional CSCs. The sdbc1 homozygous and Sdbc1 heterozygous plants show lower cellulose content and reduction in secondary wall thickness and enhances biomass enzymatic sacchari cation compared with WT. Not only that, we found that the OsCESA9/OsCESA9 D387N heterozygous plants increase salt tolerance through indirectly affecting related genes expression and improve straw returning to eld due to its easily broken culms and lower cellulose content without any negative effects in plant normal growth and lodging. Hence, manipulation of OsCESA9 D387N can provide the perspective of the rice straw for biofuels and bioproducts due to its improved enzymatic sacchari cation.           Figure S9 The expression levels of genes that encode for Na + and K + transporters in WT, sdbc1 and F1 plants. The Actin1 gene was used as an internal control. CK, control check. Error bars indicate the SE of three biological repeats. Figure S10 Relative expression levels of OsCESA9 in roots and shoots of WT with or without salt treatments. The Actin1 gene was used as an internal control. Error bars indicate the SE of three biological repeats.   Figure S14 The content of Cd, Pb and Cu in WT, sdbc1 and F1 grains. Error bars indicate SE from the mean of three replicates. Different letters denote signi cant differences (P < 0.05, Duncan's multiple range test).