Global Identification of Multiple OsGH9 Family Members and Their Involvement in Cellulose Crystallinity Modification in Rice

Plant glycoside hydrolase family 9 (GH9) comprises typical endo-β-1,4-glucanase (EGases, EC3.2.1.4). Although GH9A (KORRIGAN) family genes have been reported to be involved in cellulose biosynthesis in plants, much remains unknown about other GH9 subclasses. In this study, we observed a global gene co-expression profiling and conducted a correlation analysis between OsGH9 and OsCESA among 66 tissues covering most periods of life cycles in 2 rice varieties. Our results showed that OsGH9A3 and B5 possessed an extremely high co-expression with OsCESA1, 3, and 8 typical for cellulose biosynthesis in rice. Using two distinct rice non-GH9 mutants and wild type, we performed integrative analysis of gene expression level by qRT-PCR, cellulase activities in situ and in vitro, and lignocellulose crystallinity index (CrI) in four internodes of stem tissues. For the first time, OsGH9B1, 3, and 16 were characterized with the potential role in lignocellulose crystallinity alteration in rice, whereas OsGH9A3 and B5 were suggested for cellulose biosynthesis. In addition, phylogenetic analysis and gene co-expression comparison revealed GH9 function similarity in Arabidopsis and rice. Hence, the data can provide insights into GH9 function in plants and offer the potential strategy for genetic manipulation of plant cell wall using the five aforementioned novel OsGH9 genes.


Introduction
Cellulose is the major wall polysaccharide in plants and has a wide application for biofuel, paper, and other chemical products [1,2]. Due to their crystalline property, cellulose microfibrils are highly recalcitrant to biomass saccharification [3]. Hence, understanding cellulose biosynthesis and crystallization is essential.
Cellulose is a fibrous polymer of glucose units linked by b-1, 4glucosidic bonds. It can self-associate into non-crystalline and crystalline microfibrils in a plant cell wall, providing mechanical strength and flexibility during plant growth and development [4]. Over the past years, the crystallinity index (CrI) has been applied to account for lignocellulose crystallinity by characteristic X-ray diffraction (XRD) patterns and solid-state 13 C nuclear magnetic resonance (NMR) spectra [5,6]. In higher plants, cellulose is synthesized at the plasma membrane by a symmetrical rosette of six global protein complexes, with each complex containing several structurally similar cellulose synthase (CESA) subunits [7]. AtCESA1, 3, and 6 in Arabidopsis and OsCESA1, 3, and 8 in rice have been identified for primary cell wall formation, whereas AtCESA4, 7, and 8 and OsCESA4, 7, and 9 are responsible for cellulose biosynthesis in the secondary cell walls, respectively [8][9][10][11]. Furthermore, AtCESA3 ixr1-2 and AtCESA6 ixr2-1 mutants show increased biomass saccharification efficiency and reduced cellulose crystallinity index (CrI) [12,13]. CESA1 A903V and CESA3 T942I mutants also display reduced cellulose microfibril crystallinity [14]. In addition, other genes, such as COBRA and KORRIGAN, have been reported to contribute to cellulose biosynthesis [15,16].
A number of KORRIGAN (kor) mutants of GH9A family genes show reduced crystalline cellulose in plants [15,28,29]. For instance, ectopic overexpression of PttKOR1 in Arabidopsis kor1-1 mutant leads to a higher cellulose crystallinity [30], whereas RNA interference (RNAi) of PaxgKOR could reduce cellulose level and increase cellulose crystallinity [31]. In rice, silencing of OsGH9A3 results in reduction of cell elongation and cellulose content; it also causes an increase of pectin content in leaves [32]. In addition, KORRIGAN protein can either cleave sterol-cellodextrin substrate [33], or remove glucan chains incorrectly assembled in the growing microfibrils [34]. Although both GH9B and GH9C have been reported with activities for cello-oligosaccharide release and xyloglucan cleavage in plants [25,[35][36][37][38], little is known about their functions in cellulose biosynthesis and crystallization in rice [39,40].
Rice, an important food crop worldwide, is a model for gene function analysis in monocotyledonous plants. The completion of the rice genome sequencing may likely identify the potential function of the entire GH9 family genes in rice, based on bioinformatics analysis and related biological characterization. In this study, we initially observed a gene co-expression profiling between OsGH9 and OsCESA, and then performed a comparative analysis of the multiple OsGH9 genes' functions in cellulose synthesis and assembly, especially in their involvement in modification of lignocellulose crystallinity in rice.

Identification of OsGH9 family in rice
Searching local rice genome database yielded a total of 25 OsGH9 family genes located in 9 chromosomes in rice, except 1 gene (LOC_Os01g64140) with a partial DNA sequence (Table  S1). Based on the nomenclature of GH9 family in Arabidopsis [26], GH9 family proteins can be divided into three subclasses (A, B, and C) ( Figure 1A). GH9A subclass (OsGH9A1, 2 and 3) contained the charged and hydrophilic amino acid-rich N-termini, transmembrane domain (TM), and proline-rich C-termini. The GH9A proteins had large tails with 71 to 73 amino acid residues (Table S1). In addition, all GH9A proteins contained polarized targeting signals (LL and YXXW) and six predicted glycosylation sites (data not shown). Particularly, OsGH9B18 showed the sequence homology and motif similar to OsGH9B13 (Figure 1), despite that its orthologue in Arabidopsis has been grouped as AtGH9A4 ( Figure S4). By comparison, all GH9B and GH9C proteins did not contain any TM, but had a catalytic domain (CD) and an N-terminal cleavable signal peptide (SP) without significant sequence conservation in the extracellular secreted proteins, as shown by PSORT prediction tool (data not shown). GH9B proteins displayed diversity in the transmembrane helix and pI value, and GH9C proteins showed a putative cellulose-binding domain (CBD) as the bacterial cellulase did (Table S1).
Co-expression profiling between OsGH9 and OsCESA family As KORRIGAN genes have been identified to show co-expression with cellulose synthase gene CESA in Arabidopsis [41], we initially observed the co-expression profiling between entire OsGH9 and OsCESAs family genes among 66 tissues (Table S9) from the most periods of life cycles of 2 rice varieties (ZS97 and MH63) using cDNA chip's CREP data (http://crep.ncpgr.cn) [42]. As a result, 25 OsGH9 family genes can be classified into 3 expression clusters (I, II, III) with 6 subunits (Ia, Ib, Ic, IIa, IIb, IIc) ( Figure 2). In general, OsGH9B8, 9, 10, and 11 genes in Cluster Ia were highly co-expressed with OsCESA4, 7, and 9 typical for secondary cell wall biosynthesis [10], whereas OsGH9A3 and OsGH9B5 genes in Cluster Ib showed strong co-expression with OsCESA1, 3, 8, 5, and 6 for primary cell wall formation [11]. In particular, Clusters Ic, IIa, and IIb presented a tissue-specific expression in the radicle, panicle, and calli, respectively, and Clusters IIc and III showed a weak co-expression pattern. Correlation analysis further confirmed that OsGH9A3 and OsGH9B5 had significant co-expression with OsCESA1, 3, and 8 (p,0.01), suggesting that OsGH9A3 and OsGH9B5 may have a role in cellulose biosynthesis (Tables 1, S2, and S3). Notably, a significant correlation was also found among OsGH9B1, 2, 3, and 16 (Tables 2, S3), confirming their tight coexpression observed in Cluster IIa.
Furthermore, 18 GH9 family gene expressions were detected by qRT-PCR ( Figure S1). Both OsGH9A3, B5, and OsCESA1, 3, 8 were highly expressed in radicle, plumule, and internodes tissues. Although OsGH9B1, 3, and 16 genes were determined with high expression levels, OsGH9B2 gene was undetectable by qRT-PCR in most tissues. In addition, OsGH9A2 and OsGH9C1 showed the specifically high expression in the radicle tissues. OsGH9B6 was highly expressed in stamen tissues, and other OsGH9 family genes were expressed in several tissues in rice.

Analysis of two rice mutants
After large-scale screening and identification of rice mutants in morphological phenotypes and cell wall characteristics including cell wall components, compositions and cell wall degradability, we selected two distinct T-DNA (Osfc4 and Osfc11) mutants with genetic backgrounds of Nipponbare (NPB, wild type) in order to identify GH9 family's potential function on cellulose biosynthesis and crystallization ( Figure 3). Both homozygous mutants exhibited a normal growth phenotype during most of their life cycles. However, both were detected with fragile culms compared with the wild type ( Figure 3A). Carbohydrate analysis indicated that the two mutants had much lower cellulose and relatively higher hemicellulose levels than wild type in their mature stem tissues ( Figure 3B). The stem tissues from 1 st to 4 th internodes at booting stage of rice presented a consistent increasing course of cellulose biosynthesis, so we observed a typical alteration of cellulose content among mutants and wild type ( Figures 3C and 3D). However, the two mutants and the wild type showed a large difference in cellulose production in 1 st and 4 th internodes, providing the experimental materials excellent for functional analysis of GH9 family genes [43]. In addition, both mutants were genetically identified to be the non-GH9 genes involved in cell wall biosynthesis and modification (data not shown), which allowed the functional analysis of entire OsGH9 family genes in rice.
Correlation among cellulase activity, lignocellulose crystallinity, and OsGH9 RNA transcripts Using two rice mutants and wild type, we initially detected cellulase activity in situ and in vitro in the stem internode tissues at  Table S8 using the interProScan search program http://www.ebi.ac.uk/Tools/InterProScan/, and then identified using the MEME program (version 4.0) http://meme.sdsc. edu/meme/cgi-bin/meme.cgi. doi:10.1371/journal.pone.0050171.g001 booting stage. As shown in Figure 4A, the cellulase activity in situ could be observed in the cell walls of vascular bundles of internodes in the stem tissues. The cellulase activity in vitro with substrate resorufin cellobioside was quantitatively determined with a dynamic alteration during stem internode growth and development ( Figures 4B and S2). As a result, both mutants and wild type displayed much higher cellulase activities in young internode (1 st and 2 nd ) tissues than in the old ones (3 rd and 4 th ), indicating cellulase predominant activity in primary cell wall biosynthesis. Meanwhile, we detected lignocellulose crystallinity in all tissues with a consistent increase CrI value from 1 st to 4 th internodes in both mutants and wild type ( Figure 4C). Furthermore, a correlation analysis was conducted between cellulase activity and lignocellulose CrI with R 2 value at 0.44 ( Figure 4D). Despite the R 2 value being less than 0.5, the correlation coefficient value significantly reached 20.70 (p,0.05), suggesting that cellulase may modify cellulose crystallinity.
We then analyzed the representative gene expression levels of 3 clusters (Ia, Ib, and IIa) in 4 internode tissues of mutants and wild type by qRT-PCR analysis ( Figure S1, Table S4) to test the GH9 family genes' involved roles in the alteration of lignocellulose crystallinity. GH9B1 and GH9B16 in Cluster IIa showed extremely high correlation coefficient values (0.813 to 0.902, at p,0.01, respectively), either positively with cellulase specific activity or negatively with lignocellulose CrI ( Figure 4E and Table S5). Although GH9B3 in Cluster IIa did not display a significant correlation with cellulase activity, its coefficient value related to lignocellulose CrI reached 20.716 at p,0.01. Hence, our findings suggest that OsGH9B1, B3, and B16 may have enzymatic  activities for lignocellulose crystallinity modification in rice in terms of stem internode growth and development. By contrast, other GH9 genes in Clusters Ia and Ib did not show any significant correlation with lignocellulose CrI or cellulase activity. Furthermore, OsGH9A3 and OsGH9B5 exhibited significantly positive co-expression (p,0.01) with OsCESA1, 3, and 8 among the four stem internodes of mutants and wild type, but other OsGH9B genes did not show any positive co-expression (Tables S4, S6). Notably, OsGH9A1 was highly co-expressed with OsCESA3, 4, 8 (p,0.05), whereas OsGH9B1 and OsGH9B16 were negative with OsCESA7 (p,0.01). It was confirmed that OsGH9A3 and B5, other than OsGH9B1, 3, 16, may play a role in cellulose biosynthesis, although additional evidences need to provide. It also suggests that OsGH9A1 may have an effect on cellulose biosynthesis at least in the rice stem internode growth and development.

Comparison with AtGH9 family in Arabidopsis
According to the catalytic domain (CD) analysis, 25 putative AtGH9 family genes with 3 subclasses (A, B, and C) can be found in Arabidopsis [26] Based on the co-expression profiling among the 63 tissues in Arabidopsis, 25 AtGH9 genes can also be classified into 3 clusters ( Figure S3, Table S10). By comparison, AtGH9A1 and B7 in Cluster Ic, like OsGH9A3 and B5 in Cluster Ib, were highly co-expressed with AtCESA1, 3, and 6 typical for primary cell wall biosynthesis in Arabidopsis (Table 3). Furthermore, AtGH9B1 and B2 showed high co-expression in the flower and carpel tissues, similar to OsGH9B1, B2, B3, and B16 expression in panicle tissue. In addition, phylogenetic analysis was consistent with the coexpression patterns in rice and Arabidopsis ( Figure S4). Due to the close relationship in the phylogenetic tree, AtGH9A1 and B7 were suggested to have a role in cellulose biosynthesis, whereas AtGH9B1 and B2, like OsGH9B1, 3, 16 in rice, may have enzymatic activities for lignocellulose crystallinity modification in Arabidopsis.

Discussion
Large-scale co-expression has been performed to identify the gene functions in plant cell wall formation across plant species [44]. In Arabidopsis, KORRIGAN mutants have been characterized as GH9A family genes involved in cellulose biosynthesis [15,[28][29][30]. The regression analysis of the 408 publicly available microarray data sets could even reveal the AtKOR1 (AtGH9A1) co-expression with AtCESA1, 3, and 6 typical for primary cell wall biosynthesis [41]. In aspen tree, RT-PCR, in situ hybridization, and tissue-print assays demonstrated the co-expression of PtrKOR with PtrCESA1, 2, and 3 genes associated with the secondary cell wall synthesis in xylem cells [45]. Although several GH9A (KOR) genes have been identified in Arabidopsis, rice, and other plants [32,46], little remains known about other OsGH9 subclass functions. In this study, global co-expression profiling and correlation analysis indicated that two subclasses of OsGH9 family genes (OsGH9A3 and OsGH9B5) were highly co-expressed with OsCESA1, 3, and 8 genes typical for cellulose synthesis in rice.
Similarly, two subclasses of AtGH9 family genes (AtGH9A1 and AtGH9B7) were also identified for cellulose biosynthesis in Arabidopsis. It suggests that genetic silencing of OsGH9A3 or OsGH9B5 or both genes in the transgenic plants may clarify their potential role, like KORRIGAN, in cellulose biosynthesis in rice and other plants. More importantly, co-expression patterns of the genes could somewhat suggest the potential interaction or coordination of their proteins. For instance, based on the high co-expression pattern, OsGH9B1, 3, and 16 were initially suggested to be involved in lignocellulose crystallinity modification, other than in cellulose biosynthesis, which was sequentially confirmed by  Table 3. Comparison of gene functional patterns in rice and Arabidopsis based on the co-expression profiling data in Figure 2 and Figure S3. The co-expression data of rice and Arabidopsis GH9 family genes derived from Figure 2 and Figure S3. doi:10.1371/journal.pone.0050171.t003 integrative analysis among gene expression level, cellulase activity, and lignocellulose CrI. Obviously, gene co-expression profiling and correlation analysis are powerful tools in the identification of the entire GH9 family genes' functions in rice and other plants.
The in situ hybridization of endo-b-1, 4-glucanase in differential tissues has been reported in pine [47], Brassica napus [48], and aspen trees [45]. A novel, real-time fluorogenic assay with resorufin-b-cellobioside as a substrate has been used for observing glycoside hydrolase activity in planta. Recently, the resorufin glycosides have been detected with high sensitivity in muro cellulase enzyme activity assay due to the significant resorufin ionization at typical apoplastic pH values [49]. By this means, KOR1overexpressing Arabidopsis plants have been found to show increased cellulase activity in stem tissues compared with wildtype plants [30]. In this study, the cellulase activity in situ was observed specifically in the cell wall of the four internode stems in rice, and the cellulase activities in vitro were quantified and compared between mutants and wild type. Notably, both wild type and mutants displayed a consistent increase at cellulose levels and CrI values from 1 st to 4 th internode, but showed a constant decrease at cellulase activity in vitro except at 2 nd internode of wild type. It suggests that the four internodes would be model materials accounting for cellulase effect on cellulose biosynthesis and lignocellulose crystalline feature in rice and other plants. Despite two mutants and wild type showed a difference at each internode, we could conduct a correlation analysis using all four internodes of wild type and mutants. The correlation calculations among OsGH9 expression level, cellulase activity, and lignocellulose crystallinity, could indicate that OsGH9B1, 3, and 16 have specific cellulase activities for lignocellulose crystallinity modification, other than for cellulose biosynthesis. By contrast, OsGH9A3 and B5 did not show any significant correlation either with cellulase activity or lignocellulose CrI. Hence, we could interpret that the changed lignocellulose crystallinity of AtGH9A1 (Kor) may be a consequence of cellulose biosynthesis, because recent report has indicated that CesA mutant could result in the lignocellulose crystallinity alteration [58].
Mutant selection and reverse genetic analysis have been broadly applied to identify the target gene function in plants. However, both approaches have their typical limitations and disadvantages, especially if the target gene is lethal to plant growth, or genetically redundant, or functionally dependent on isoform coordination [13,49,50]. Alternatively, we found that analysis of non-GH9 rice mutants is a useful approach for identification of the multiple OsGH9 family genes. Because the two rice mutants used in this study were genetically identified as the non-GH9 mutants (data not shown), we could conduct a correlation analysis among all detectable OsGH9 gene expressions, cellulase activity, and lignocellulose crystallinity in the wild type and two mutants, suggesting that OsGH9B1, 3, and 16 have a coordinate function on lignocellulose crystallinity modification. To our knowledge, the GH9B subclass gene function in plants has not yet been discovered. Although the GH9B function could be investigated by reverse genetic analysis, the extremely high co-expression and coordination among OsGH9B1, 3, and 16 suggest that silencing of the individual OsGH9B isoform gene may not result in any significant alteration of lignocellulose crystallinity. In this case, cosilencing of OsGH9B1, 3, and 16 in a mutant may be essential for functional analysis in the future. Hence, non-GH9 mutants are advantageous for function analysis of the entire OsGH9 family genes.
A total of 25 GH9 family genes with 3 subclasses (A, B, and C) were identified in both rice and Arabidopsis. GH9A (KORRIGAN) proteins, such as OsGH9A3 and AtGH9A1, containing the transmembrane domain, could be co-localized with CESA complex [51], indicating their involvement in cellulose biosynthesis. However, GH9B subclass, such as OsGH9B5 and AtGH9B7, comprised the secreted proteins, suggesting that GH9B should be distinguished from GH9A for cellulose biosynthesis. OsGH9B1/ 2/3/16 and AtGH9B1/2 were proposed to have the enzymatic activity for lignocellulose crystallinity alteration, other than for cellulose biosynthesis, indicating that these proteins may have specific activities for post-modification of cellulose microfibers in the cell wall. In addition, GH9C subclass comprised the secreted proteins, but had a C-terminal CBD to crystalline cellulose, suggesting a specific role in the turnover of crystalline cellulose [23,26]. Although GH9C function was not identified in this study, OsGH9C1 and AtGH9C1/3 were highly and specifically expressed in the radicle tissues, providing potential for further investigation in the future.

Conclusions
Global gene co-expression profiling and correlation analysis based on microarray data from 66 tissues of 2 rice varieties indicated the OsGHA3 and B5 potential role in cellulose biosynthesis. Integrative analysis of OsGH9 gene expression level, cellulase specific activities in situ and in vitro, and lignocellulose crystallinity (CrI) in distinct two rice mutants and wild type revealed that OsGH9B1, 3, and 16 may have enzymatic activities for lignocellulose crystallinity modification. The results can provide new insights into OsGH9 function in plants and offer a strategy for genetic manipulation of OsGH9 genes toward bioenergy crop breeding in rice.

Plant materials and growth condition
A japonica rice var. Nipponbare (NPB) and T-DNA insertion homozygous mutant lines (Osfc4, Osfc11) were grown in the field of Huazhong Agricultural University, Wuhan, China. When the tip of panicle just protruded out of the flag leaf at booting stage, four different internodes of stems were collected to analyze the OsGH9 family gene expression, in situ and in vitro cellulase activities and cell wall component determination. When the panicles protruded out 2 cm above the top of the flag leaf, stems were sampled for the analysis of qRT-PCR and cell wall component. All tissues were obtained from 3-6 plants and pooled for each biological replicate in the biological triplicate.

Phylogenetic and structural analysis
The Hidden Markov Model (HMM) profile of the glycoside hydrolase family 9 domain (PF00759.1) was downloaded from PFam http://pfam.sanger.ac.uk/. We employed a name search and the protein family ID PF00759.1 for the identification of OsGH9 genes from the rice genome. Information about the chromosomal localization, coding sequence (CDS), amino acid (AA) and full length cDNA accessions was obtained from TIGR http://www.tigr.org and KOME http://cdna01.dna.affrc.go.jp/ cDNA. The multiple alignment analysis with counterparts in Arabidopsis was performed using the Clustal X program (version 1.83) [52], the unrooted phylogenetic trees were constructed with the MEGA3.1 program [53] and the neighbor joining method with 1,000 bootstrap replicates.

Co-expression profiling analysis
The transcriptional profile data of GH9 family and CESA family genes in 33 tissues (Table S9) of Zhenshan 97 (ZS97) and Minghui 63 (MH63), Z1-Z33 and M1-M33 (Figure 2) was respectively obtained from the CREP database http://crep. ncpgr.cn. Massively parallel signature sequencing (MPSS) data website: http://mpss.udel.edu/rice/mpss_ index. php [56] was used to determine the transcriptional profiles of the genes with conflicting probe set signals. The expression values were logtransformed, and cluster analysis was performed using a software cluster with Euclidean distances and the hierarchical cluster method of complete linkage clustering as described in [11,42]. Correlation coefficients of these gene expressions were also calculated to determine whether they are significantly different at 0.01 or 0.05 percent levels, respectively The gene expression profiling of AtGH9 and AtCESA families in the 66 tissues of Arabidopsis (Table S10) was based on the Gene Expression Omnibus database http://www.ncbi.nlm. nih.gov/ geo/ using the GSE series accession numbers GSE5629, GSE5630, GSE5631, GSE5632, GSE5633 and GSE5634. The raw data were processed with the Affymetrix Microarray Analysis Suite (MAS Version 5, Affymetrix) [57]. Subsequent analysis of the gene expression data was performed in the statistical computing language R http://www.r-project.org using packages available from the Bioconductor project http://www. bioconductor.org as described in [11].

qRT-PCR analysis
Total RNA was isolated from samples using RNAprep pure Plant Kit (DP432, TIANGEN BIOTECH), and 5 mg total RNA was reverse transcribed with an oligo(dT) 18 primer in a 50 ml reaction using an M-MLV Reverse Transcriptase (Promega, USA) according to the manufacturer's instructions. The qRT-PCR was performed in a 20 ml reaction system (cDNA template 2.0 ml, 26SYBR Green1 Mix10 ml, primer-F 0.5 ml, primer-R 0.5 ml, MilliQ 7.0 ml) with SYBR Green qPCR kit (ZOMANBIO, China) on Two Color Real-time PCR Detection System (MyiQ2, BIO-RAD) using the following program: 2 min at 95uC followed by 40 cycles of 15 sec at 95uC, 15 sec at 60uC, 25 sec at 72uC. Ubiquitin gene (AK059011) was used as an internal standard in the qRT-PCR. The gene expression unit was subjective to the percentage of the target gene expression value relative to the internal standard (Ubiquitin gene). All quantitative PCR experiments were performed in biological triplicate. All the gene-specific primers used were listed in Table S7.
In situ and in vitro cellulase activity assay In situ cellulase assay was performed as described previously by Takahashi et al [30] with minor modifications. The 3 rd internodes of rice stems were hand-sectioned (<100 mm) and placed in 0. In vitro cellulase activity was also detected by Markergene TM Fluorescent Cellulase Assay Kit. Four internodes of rice stems with 0.1 g each were ground to powder in liquid nitrogen, and suspended in 300 ml reaction buffer (100 mM sodium acetate buffer, pH 6.0) at 4uC for 5 min. The supernatant was collected after centrifuge twice at 18,000 g at 4uC for 15 min (Eppendorf Centrifuge 5417 R). 50 ml of supernatant was incubated with 50 ml of 0.5 mM Resorufin Cellobioside (substrate reagent) and reacted for 5 min in a black flat-bottomed 96-well microtiter plate (Greiner Microlon).
Fluorescence intensity of resorufin released was measured with excitation (550 nm) and emission (595 nm), at 35uC for 45 cycles with a cycle time of 1 min using Multimode microplate reader (TECAN Infinite M200) according to the method described by Takahashi J et al [30]. Fluorescence values of blank (50 ml substrate reagent was added to 50 ml reaction buffer) were subtracted at each time point. A standard curve of Resorufin ranging from 0 to 50 mM was prepared to determine concentration of Resorufin in rice internode tissues extracts reacted with Resorufin Cellobioside. The protein concentration of samples was measured by Bradford method in triplicate. All the reactions were performed with biological triplicates.

Plant cell wall fractionation and wall polysaccharide analysis
The plant tissues, including internodes or stems of different stages, were first heated at 105uC for 20 min, dried to constant weight at 60uC for about 7 d and kept dry until use. The extraction and fractionation of cell wall polysaccharides were performed as previously described by Peng et al with minor modification [58]. The crude cell wall material was extracted with 0.5% (w/v) ammonium oxalate and heated for 1 h in a boiling water bath. The remaining pellet was suspended in 4 M KOH containing 1.0 mg.ml 21 sodium borohydride for 1 h at 25uC, and the combined supernatant was neutralized, dialyzed and lyophilized for total hemicelluloses analysis. The KOH non-extractable residue was further extracted with acetic-nitric acids for 1 h at 100uC and the remaining pellet was used for cellulose determination. All samples were carried out in biological triplicate.
Colorimetric assay of total hexoses and pentoses: UV-VIS Spectrometer (V-1100D, Shanghai MAPADA Instruments Co., Ltd. Shanghai, China) was used for the absorbance reading. Hexoses were detected using the anthrone/H 2 SO 4 method, and pentoses were detected using the orcinol/HCl method. For cellulose determinations, the cellulose sample was dissolved in 67% (v/v) H 2 SO 4 (1.0 ml) with shaking at 25uC for 1 h, and then 10.0 ml aliquot was used for the anthrone/H 2 SO 4 method. Total hemicelluloses level was subject to the sum total of hexoses and pentoses. Considering that high pentoses level can affect the absorbance reading at 620 nm for hexoses content by anthrone/ H 2 SO 4 method, the deduction from pentoses reading at 660 nm was carried out for final hexoses calculation. A series of xylose concentrations were analyzed for plotting the standard curve referred for the deduction, which was verified by GC-MS analysis. All experiments were carried out in biological triplicate.
Total lignin content was determined by two-step acid hydrolysis method. The solution was filtered with membrane filter (0.22 mm). 20 ml solution was injected into a HPLC (Waters 1525 HPLC) column Kromat Universil C18 (4.6 mm6250 mm, 5 mm) operating at 28uC with CH 3 OH:H 2 O:HAc (25:74:1, v/v/v) carrier liquid at the flow rate of 1 ml.min 21 . All experiments were carried out in biological triplicate.

Detection of the crystallinity index
Detection of crystallinity index of lignocellulose (CrI) was described by Xu et al [59] with a minor modification. The internodes of rice stem tissues at booting stages were cut into small pieces through 40 mesh sieve. The fine raw biomass powder of the plant tissue was laid on the glass sample holder (3565065 mm), analyzed under plateau conditions: Ni-filtered Cu Ka radiation (l = 0.154056 nm) generated at 40 kV and 18 mA, and scanned at speed (0.0197u/s ) from 10u to 45u. CrI was calculated as 1006(I2002Iam)/I200, where I200 is the intensity of the 200 peak (h = 22.5u), Iam is the intensity at the minimum between the 200 and 110 peaks (h = 18.5u). Standard error was detected at 6 0.05,0.15 using five samples in triplicate.

Statistic calculation
The SPSS 17.0 was used for statistical analysis. Correlation coefficients were generated by performing Spearman rank correlation analysis for all pairs of measured traits across the whole population. This analysis used average values calculated from all original determinations for a given traits pair.   Table S5 Correlation coefficients between OsGH9 genes expression level and cellulase specific activity or lignocellulose CrI in 12 internodes of mutants (fc4 and fc11) and wild type (NPB) at booting stages (n = 12). (DOCX)