Coordinated Activation of Cellulose and Repression of Lignin Biosynthesis Pathways in Rice

Cellulose from plant biomass is the largest renewable energy resource of carbon fixed from the atmosphere, which can be converted into fermentable sugars for production into ethanol. However, the cellulose present as lignocellulosic biomass is embedded in a hemicellulose and lignin matrix from which it needs to be extracted for efficient processing. Here, we show that expression of an Arabidopsis ( Arabidopsis thaliana ) transcription factor SHINE (SHN) in rice ( Oryza sativa ), a model for the grasses, causes a 34% increase in cellulose and a 45% reduction in lignin content. The rice AtSHN lines also exhibit an altered lignin composition correlated with improved digestibility, with no compromise in plant strength and performance. Using a detailed systems-level analysis of global gene expression in rice, we reveal the SHN regulatory network coordinating down-regulation of lignin biosynthesis and up-regulation of cellulose and other cell wall biosynthesis pathway genes. The results thus support the development of non-food crops and crop wastes with increased cellulose, and low lignin with good agronomic performance that could improve the economic viability of lignocellulosic crop utilization for biofuels.


INTRODUCTION
Crop residues are a vast resource of lignocellulose feedstock available for conversion to biofuels, and their utilization does not compete with food supplies unlike grain-based feedstocks (Haigler et al., 2001). Rice (Oryza sativa) straw itself constitutes half the crop-waste worldwide, which is either burnt or wasted (Sticklen, 2006). Non-food perennial grasses such as switchgrass (Panicum virgatum) and Miscanthus (Miscanthus giganteus) as well as fast growing woody crops make up the bulk of lignocellulosic resources. In either case, plant lignocellulosic cell walls are quite resistant to digestion of the complex polysaccharides (cellulose) into simple sugars before fermentation due to the presence of heavily cross-linked lignin. Methods to lower lignin and improve the availability and levels of cellulose are therefore important to make the conversion into biofuels economically feasible.
Cellulose is the most abundant biopolymer on earth, comprising 25-50% of plant biomass with an estimated 100 billion tons synthesized annually as a result of photosynthesis (Haigler et al., 2001;Sticklen, 2006). Cellulose is made up of glucose units and is synthesized at the plasma membrane by the cellulose synthase complex, comprising multiple CESA proteins that belong to multigene families in plants (Somerville, 2006). Long chain cellulose polymers are organized into microfibrils that make up the core content of plant cell walls, contributing to the strength, structure and development of plants (Sticklen, 2006). Hemicelluloses are polysaccharides in plant cell walls that include xyloglucans, xylans, mannans and glucomannans, and β -(1→3, 1→4)-glucans, and are synthesized by glycosyltransferases (GTs) located in the Golgi membranes. The most important biological role of hemicelluloses is their contribution to strengthening the cell wall by interaction with cellulose and, in some cell walls, with lignin (Scheller and Ulvskov, 2010). Despite its importance, the details regarding the synthesis of hemicelluloses remain very elusive and very little is known about the regulation of the cellulose biosynthesis pathway.
Lignin, the second most abundant polymer, is a complex comprised of guaiacyl (G), syringyl (S) and p-hydroxylphenyl (H) phenylpropanoid units (Supplemental Fig. S1), contributing to lignin heterogeneity (Boerjan et al., 2003). Angiosperm dicot lignin is primarily composed of G and S units, and monocot lignin is a mixture of G, S and H units (Supplemental gymnosperms like pines) are more resistant to chemical degradation, making the composition of lignin (the relative ratio of G to S units), along with its quantity, crucial for the digestibility of crops for conversion into biofuels and cellulosic products. The monolignol biosynthetic genes have therefore been used in engineering lignin content and composition in several plants (Vanholme et al., 2008). Many of these studies were first reported in non-feedstock model dicot plants such as tobacco and Arabidopsis (Zhou et al., 2009), and the expectation is that similar approaches can be applied to cellulosic feedstock crops, but very few detailed engineering studies have been reported in the grasses, which are a major lignocellulosic resource.
In the grasses, the maize and sorghum brown midrib mutations (Li et al., 2008) show alterations in lignin content and digestibility; the maize brittle stalk2 (bk2) and rice brittle culm1 (bc1) mutations of a similar gene have a brittle phenotype due to reduction in cellulose and cell wall composition with no compensatory changes in lignin (Li et al., 2003b); and, the rice flexible culm 1 (fc1) mutant has reduced lignin, H and G residues (Li et al., 2009). However, significant reductions in lignin or digestibility in the monocot crops, including the brown midrib and other mutants, are also accompanied by reductions in plant growth, biomass, stalk strength, or pathogen resistance (Li et al., 2008).
Several transcription factors (TFs) have also been shown to affect cellulose and lignin content and composition (Mele et al., 2003;Kubo et al., 2005;Zhong et al., 2006;Zhong and Ye, 2009). Transcriptional regulation is achieved by top-level NAC TFs SND1 (SECONDARY WALL-ASSOCIATED NAC DOMAIN 1), NST1/2 (NAC SECONDARY WALL THICKENING PROMOTING 1/2), VND6/7 (VASCULAR-RELATED NAC-DOMAIN 6/7) that activate a nexus of intermediate TFs, mostly MYBs. These intermediate TFs in turn activate low-level MYB TFs that bind to and activate target cell wall biosynthetic genes, presumably, achieving high levels of specificity. Such a multi-layered regulatory network that affects multiple target genes offers the cell a robust mechanism to achieve coherent changes in the flux through the pathways. Additionally, the network also points to the possibility of existence of multiple knobs and switches that can be tuned to execute specific regulation of different cell wall pathways in order to optimize secondary cell wall composition. Specifically, cell wall with increased cellulose content in combination with reduced lignin for enhanced sugar and ethanol would yield valuable cellulosic feedstock (Jakob et al., 2009). 8 distinguish 35,161 rice genes, and thus reliably characterize the differential expression of many cell wall pathway genes.
In addition, several TFs were observed to be regulated by SHN in the microarray experiment (Supplemental Table S1). These included rice homologs of several NAC and MYB transcriptional activators of the cell wall biosynthetic pathways uncovered in Arabidopsis and other plant species (Zhong and Ye, 2009).
To verify SHN-regulation of the monolignol and cell wall biosynthesis pathways as observed in the microarray, we carried out rigorous quantitative real-time PCR (qRT-PCR) experiments to determine the abundance and tissue-specificity of biosynthetic genes in two different tissue types -leaf and culm -in SHN and WT plants. We found that transcript levels of seven CAD's, and four 4CL's were significantly repressed in one of the tissues ( Fig. 2A top), confirming their regulation in leaf and culm by SHN. Two genes CAD4 (Os11g40690) and CAD7 (Os04g52280) were either not responsive or slightly induced in the culm or leaf tissue, respectively. Pathway-wide repression of lignin biosynthesis including the 4CL and CAD genes that catalyze specific branches of lignin biosynthesis leading to different lignin monomers, suggested possible alterations in lignin composition along with overall reduction in response to SHN expression.
Expressions of a battery of cellulose and other cell wall biosynthesis genes, on the other hand, were induced in SHN leaf and culm compared to WT (Fig. 2A bottom;Supplemental Fig. S2). The varying levels of expression of individual members of cell wall (cellulose or hemicelluloses) biosynthetic gene families in culm and leaf points to selective expression in specific tissues or organs. Nevertheless, five out of eleven CESA genes in rice are significantly up-regulated by SHN in culm tissue ( Fig. 2A bottom). Induction of three out of the five CESA genes is consistent across leaf and culm. Together, these suggest a role of SHN in inducing the expression of certain cellulose and other cell wall biosynthetic genes in rice.
In addition, seven putative rice cell wall biosynthesis TFs -three NAC genes and four MYB genes -were confirmed to be down-regulated in the leaf tissue (Fig. 2B). Six of these TFs were also repressed in culm. One TF, MYB20/43 (Os02g49986), was however up-regulated in both tissue types (Fig. 2B). These modes of differential gene expression of TFs may account for the coordinate expression of their biosynthetic targets in new spatial or temporal patterns as required to generate functional integrity of the pathway. Taken altogether, these data indicates that AtSHN is a key regulator of monolignol and other cell wall biosynthesis. These gene expression results urged us to test for associated phenotypic and biochemical changes.

Rice AtSHN Lines Have Altered Cell Walls in Mechanical Tissues
To assay for changes in lignin and cellulose, transverse sections of the culms of WT and rice AtSHN lines were histochemically stained with phloroglucinol and calcofluor solutions. Electron microscopy analysis demonstrated that the rice AtSHN lines also showed heavily thickened sclerenchyma cell walls (Fig. 3G), and a thickened folded cell wall structure ( Fig. 3I and J) compared with that of the WT ( Fig. 3F and H). These analyses together suggest a thickening of secondary walls due to deposition of cellulose in place of lignin in SHN lines.
Further quantitative analysis showed that the wall thickness of sclerenchyma and bundle sheath fiber cells was increased by approximately 45% and 48%, respectively, in the SHN overexpressors compared to WT (Table I).
Since the rice AtSHN plants are morphologically indistinguishable from WT plants, we tested if the reduced lignin had an effect on its physical properties. To quantitatively compare the strength of the tissues, we determined the breaking forces of rice AtSHN and WT plants (Supplemental Table S2), which showed that the force required to break the leaves and culms of rice AtSHN lines was almost identical to that of WT.

Rice AtSHN Lines Have Enhanced Cellulose and Reduced Lignin
Expression analysis and cellular phenotypes implied that the cellulose/cell-wall biochemical composition in the rice AtSHN lines may be altered. We therefore analyzed the major secondary cell wall constituents like cellulose and lignin contents in rice AtSHN and WT plants. Chemical analysis of rice culms showed a significant 34% increase in cellulose content (Fig. 4A) and a 45% reduction in lignin (Fig. 4B), in the rice AtSHN lines compared to WT. Furthermore, analysis of neutral cell wall associated carbohydrates (arabinose, fucose, galactose, glucose, mannose, rhamnose and xylose), showed that the amount of glucose and xylose, the main sugars in cellulose and hemicellulosic polysaccharides (Li et al., 2003b), respectively, were significantly increased in the SHN lines (Table II). For the above analyses we used equal amounts of dry weight, because the cell wall preparation calculated as the ratio of dry weight to fresh weight, was not significantly different between WT and SHN lines. Hence, we suggest that the lower amount of lignin is mass-balanced by hemicelluloses and/or cellulose, without change in total dry weight. Taken together, the results support the hypothesis that SHN causes a decrease in lignin and a compensatory increase in hemicellulose and cellulose.
The digestibility of lignocellulosic feedstock for feed and biofuel is dependent, in addition to lignin content, on the composition, manifested as the ratio of G and S monomers (Vanholme et al., 2008). Since enzymes involved in specific parallel branches of lignin biosynthesis were altered in expression, we conceived an effect of SHN expression on the composition of lignin in addition to its content. GC-MS analysis (Lu and Ralph, 1997) revealed that the rice AtSHN lines, compared to WT, have a 54% reduction in G content and consequent reduction in G:S ratio to two-third (Fig. 4C). The S content is not significantly changed, possibly due to a selective reduction in activity of the 4CL, CCR and CAD genes involved in G synthesis, as observed in the gene expression analysis ( Fig. 2; Supplemental Table S1). Alternatively, AtSHN might cause equal inhibition of the individual steps, but under such circumstances, the branch leading to coniferyl alcohol might be more efficient at channeling limited precursors than the branch leading to sinapyl alcohol. Selective inhibition in monolignol biosynthesis has been observed previously in tobacco by expressing o-methyl transferase (Atanassova et al., 1995) and Antirrhinum MYB genes (Tamagnone et al., 1998).

Coexpression Network Underlying Regulation of Lignin and Cellulose Biosynthesis
In order to evaluate if the rice SHN gene Os06g40150 (OsSHN), homolog of Arabidopsis SHN2, also has an intrinsic association with the cell wall pathways, an extensive analysis of coexpression in rice was undertaken. A global coexpression network of rice genes was constructed based on public gene expression datasets in rice. Raw Affymetrix rice expression profiles pertaining 'response to environmental conditions' were collected from GEO (Barrett et al., 2009) andArrayExpress (Parkinson et al., 2009). Gene expression values across the diverse conditions were extracted using the custom gene-centric probeset definitions, and pairwise genegene correlations were calculated to create the global network (see Methods). From this global network, we explored the connectivity between i) AtSHN-regulated lignin and other cell wallrelated genes, ii) TFs associated with these pathways, and other TF genes differentially expressed in the AtSHN microarray, and iii) the OsSHN gene. In addition to finding that the biosynthesis genes were intimately connected to the TFs including the NACs and MYBs, remarkably, the OsSHN gene was also found to be connected to these TFs, together forming a dense coexpression network (Fig. 5).
We then cross-referenced this network with our independent SHN expression profile of rice genes to gauge their correspondence. Overlaying the differential expression of the genes (from microarray and qRT-PCR) onto the nodes of the network showed that almost all the gene regulation observed due to the expression of Arabidopsis SHN in rice was explained by the signs of the rice network coexpression edges: positively correlated gene pairs were regulated in the same direction (both up-or both down-regulated), and negatively correlated gene pairs were regulated in opposite directions. OsSHN was directly connected to the homolog of VND6 through a negative edge, supporting the down-regulation of the VND6 homolog in the expression study. Here, for simplicity in transferring functional information between species, the rice TF

DISCUSSION
Cellulose is the most abundant biopolymer and a major renewable energy resource, which can be processed into bioethanol as liquid fuel. In spite of its significance, very little is known about the regulation of cellulose biosynthesis and its co-regulation with the synthesis of other cell-wall biopolymers. We present a systems-level analysis on the complex regulation of cellulose and lignin biosynthesis, showing an unprecedented significant increase (1/3rd) in cellulose and decrease in lignin using the grass/crop model rice. These results offer novel ways for engineering non-food grasses and crop wastes for the production of lignocellulosic feedstocks that can be efficiently processed into biofuels. Expression of SHN leads to coordinate regulation of multiple steps in the pathways for monolignol and cellulose biosynthesis. Repression of the lignin pathway was exhibited as a moderate reduction in expression levels of many enzymes spread across the pathway rather than a drastic reduction of a few enzymes (Supplemental Fig.   S3). The moderate reduction in terms of the absolute expression levels indicate that gene expression is not completely shut off, but allows a background flux through the pathway.
Moreover, repression of 4CL and CAD genes that catalyze specific branches of lignin biosynthesis leading to different lignin monomers suggests possible alterations in lignin composition along with overall reduction in response to SHN expression. Differential expression of genes was observed in the microarray (Supplemental Table S1), and was confirmed using qRT-PCR analysis using gene-specific primers of multi-gene family members at two different tissues of the plant (Fig. 2). SHN overexpressors had significantly repressed transcript levels of lignification genes such as CAD and 4CL family members ( Fig addition, three rice CESA genes (OsCesA4, OsCesA7 and OsCesA9) known to be involved in the synthesis of cellulose in the secondary cell walls (Tanaka et al., 2003) and responsible for the overall strength of the plant are up-regulated in SHN lines in both leaf and culm ( Fig. 2A bottom). While expression changes estimated by microarrays and qRT-PCR were concordant for several genes, more genes, including CESA genes, whose changes were not statistically significant (indicating either noisy or no change) using microarray were detected by qRT-PCR as being differentially expressed. WT plants (Fig. 4).
The change in cellulose and lignin did not adversely affect the rice AtSHN lines, as they displayed normal plant phenotypes, maturity and seed yield under greenhouse conditions (data not shown). In addition the strength of the stem/culm of rice AtSHN lines was unaltered. The tensile or bending strength of grass tissue, such as that of maize, has been shown to correlate with the cellulose content, whereas lignin is thought to play a role in resistance to compression (Dhugga, 2007). The increase in cellulose in secondary walls of SHN lines also probably offsets any reduction in mechanical strength due to reduced lignin. Such a compensatory increase in cellulose with reduced lignin has also been observed due to inhibition of a 4CL gene in aspen trees that leads to better growth (Hu et al., 1999). Similarly, plants expressing both antisense 4CL and sense CALd5H have been found to have 52% less lignin and 30% more cellulose (Li et al., 2003a). Maize brown midrib mutants have also been shown to harbor reduced lignin and increased hemicellulose (with no change in cellulose content (Vermerris et al., 2010). On the other hand, the rice bc1 mutant and maize bk2 mutant, which have reduced cellulose content, have been shown to contain more lignin (Li et al., 2003b;Ching et al., 2006;Sindhu et al., 2007).
These studies, along with ours presented here, provide evidence for complex interdependent regulation of the different cell wall pathways. However, we provide novel evidence for a TF -SHN -coordinating such a compensatory regulatory mechanism.
Coexpression network analysis was performed as an independent route to validating the function of SHN (Fig. 5). Using a large gene expression compendium in rice, correlations between the expression profiles of all the cell wall regulatory and biosynthetic genes along with OsSHN were calculated. Coexpression network analysis has found use in dissecting transcriptional regulation in Arabidopsis (Usadel et al., 2009) and rice (Wang et al., 2009;Fu and Xue, 2010) to gain biological insights into general and case-specific regulation of gene expression. This approach has also been used to identify novel members of biochemical processes, including cellulose synthesis (Persson et al., 2005). In the current study, we have used the rice coexpression network as a predictive tool to first assess the expected associations between TFs and their targets, as in the case of NAC and MYB TFs and their putative lignin and other cell wall biosynthetic targets. Second, it was used to discover novel genes that might have a role in a pathway/process of interest using guilt-by-association, as in the case of OsSHN. And, finally, in conjunction with an independent gene expression dataset (here SHN microarray), it was used to demarcate positive and negative interactions and propose a regulatory model for genes of interest.
This analysis shows that OsSHN, the rice homolog of AtSHN, has a strong association with the cell wall regulatory and biosynthetic machinery in rice. The nature of association between the cell wall TFs and the biosynthetic genes is strongly positive as expected. The intriguing finding is the differential association of OsSHN with the TFs -via a negative connection to VND6 and a positive connection to MYB20/43 (and several other TFs), which are positively correlated with lignin and cellulose/other-cell-wall biosynthetic genes, respectively. In addition, most of the coexpression associations in the rice network agree with, and hence, corroborate the expression changes of the TFs and biosynthetic genes in response to AtSHN expression. Finally, AtSHN expressed in rice functions in the context of rice genes -established by the coexpression network -to perform its functions; a context in which OsSHN is expected to function in. Here, it is relevant to note that the OsSHN gene is not regulated upon expression of AtSHN in rice, and hence, all the observed differential expression is solely due to AtSHN. Based on this analysis we, therefore, hypothesized that OsSHN has a native association with cell wall regulatory and biosynthetic pathways, with an ability to coordinately regulate the lignin and cellulose pathways by shutting down the main switches (NACs), and intervening by directly regulating downstream MYBs: repressing MYBs specific to lignin biosynthesis (in addition to release of NAC activation), and activating MYBs and other TFs specific to cellulose/other cell wall biosynthesis.
To further pursue this hypothesis, we were interested in finding if there were any clues for potential SHN regulation of the NAC and MYB TFs. We performed de novo motif discovery on the upstream regions of all the TFs in the network. However, the analysis showed no significant motifs, which we reasoned to be because SHN could bind to and regulate just a few major TFs that could then regulate other TFs and biosynthetic genes. Therefore, we restricted ourselves to the few TF candidates that showed confirmed gene expression changes and were homologous to Arabidopsis TFs associated with the secondary cell wall biosynthetic pathways. We then searched the upstream regions of these TF genes for the presence of GCC-box motif, a putative binding site of AP2-ERF TFs (Ohme-Takagi and Shinshi, 1995). Identification of several GCCbox motifs in these sequences (Fig. 6A) led us to postulate that SHN could directly bind to and regulate these TFs. We then experimentally confirmed SHN binding to these promoter regions using mobility-shift assay (Fig. 6B).
With evidence for SHN regulating the putative NAC and MYB TFs, we gleaned support for MYB-mediated transcriptional regulation of the cell wall pathway genes in rice to that in other species. To this end, we performed de novo sequence motif discovery on the promoter regions of and other coherent (type 2; right); the relevance of this feature in a dynamical sense remains to be determined. We also believe that TFs other than those tested here could be involved in mediating SHN up-regulation of cellulose and other cell wall genes, candidates for which have been identified as those up-regulated in response to SHN expression and positively correlated with the up-regulated cell wall genes.
Since several insights about cell wall biosynthesis have been gleaned in Arabidopsis it is also important to put our findings in rice in that context and deliberate the probable role of AtSHN in Arabidopsis. To gain some understanding about AtSHN function in relation to secondary cell wall biosynthesis, we re-analyzed the gene expression profiles of Arabidopsis AtSHN (WIN1) overexpression lines compared to WT controls from Broun et al. (2004). Since the design lacks replication, we quantified the relative expression levels of ~390 (present in the 8K Affymetrix Arabidopsis genome array) cell-wall-related genes (out of ~930 total in the genome) in a 'strong' WIN1 overexpressor compared to WT (Supplemental Table S3  Therefore, an overall similar association with secondary cell wall regulation and biosynthesis is conserved across Arabidopsis and rice, with different details: we show that AtSHN in rice causes an inverse regulation of the pathways by differently regulating various TFs of the lignin and cellulose biosynthetic pathways. Furthermore, to explore the expression pattern of OsSHN in different rice organs and tissues, we used the rice eFP browser (Winter et al., 2007) to identify that the gene is expressed in the inflorescence stages P3, P4, P5 and P6, with the strongest expression at the P5 stage.
Using the rice expression atlas (Jiao et al., 2009) resource, we observed that OsSHN is expressed in the coleoptile (0hr) and fresh whole leaf and to a lesser extent in epiblast (12hr) and seedling blade (Supplemental Table S4). These cell types represent young growing tissue where lower lignin deposition is expected, consistent with SHN's proposed role in up-regulating cellulose synthesis and down-regulating lignin synthesis pathways in rice.

Plant Genotyping and Statistical Analysis
Out of fifteen transgenic rice AtSHN lines generated (Trijatmiko, 2005), T 3 progeny of different transgenic SHN lines segregating for single inserts were tested for stable expression pattern using qRT-PCR and genotyped for presence of the Arabidopsis SHN2 (AtSHN) transgene locus.
Based on the results of qRT-PCR amplification of AtSHN, it was evident that three lines showed significant and stable expression, and were used for further analysis described here.
Fifty progeny of three independent lines expressing AtSHN were grown under controlled growth chambers. For all analyses, six plants were used for each of the three transgenic lines and WT. The data in the experiments of quantitative real-time PCR, cell wall composition analysis, breaking-force measurements and measurement of cell wall thickness were analyzed using the Student's t test. All the t-tests performed were two-sided.

Histochemical Staining and Microscopy
Histochemical localization of the lignin and cellulose was done using phloroglucinol and calcofluor staining as described (Li et al., 2003b), using ~20 µm thick hand-cut sections from rice culms. The stained sections were examined and photographed under a light microscope washed twice in 0.1M Na cacodylate buffer for 15 min each and post-fixed in 1% OsO4 for 1hr, dehydrated through an ethanol gradient and infiltrated. Samples were critical point dried, sputtercoated with gold in an E-100 ion sputter, and viewed under a scanning electron microscope (Carl Zeiss-EV040). For transmission electron microscopy, ultrathin sections were made using an ultra microtome (MT-X, RMC, USA), and the sections were thoroughly stained with aqueous 2 % uranyl acetate for 10 min, followed by lead citrate for 2 min. The sections were viewed under a JEM-1010 electron microscope (JEOL, Tokyo, Japan) operating at 60kV. The wall thickness of sclerenchyma and bundle sheath fiber cells was measured using the ImageJ program (Rasband, 1997).

Measurement of Breaking Force
The breaking force of rice culms and leaves, defined as the force required to break the segment, were measured using a force-testing device (Kokubo et al., 1989) assembled in-house with parts obtained from Instron (www.instron.com) and Fisher Scientific (www.fishersci.com). The first internodes of culms and flag leaves were used for immediate fresh tissue measurements.

Cell wall Composition Analysis
Rice culms from WT and SHN lines were used for acid detergent lignin (ADL) analysis (Van Soest, 1967) and cellulose content as described (Scott and Melvin, 1953;Updegraff, 1969). For ADL analysis, samples were dried at 55 0 C for 48h, and ground to pass a 1-mm screen in a cyclone mill. After that 0.5g (±0.05g) of air dried sample was directly placed into the filter bags for ADF (Acid Detergent Fiber) determinations (ANKOM Technology, Macedon, NY) using a Fiber analyzer and followed by ADL analysis by treating with 72% sulphuric acid for 3h. The data from WT and SHN lines were determined in replicates through gravimetric analysis and are calculated as percent on dry matter basis. For cellulose, the dried powder was hydrolyzed with acetic-nitric reagent for 1 h at room temperature and centrifuged at 8,000 rpm for 5 min. The resulting pellet was extracted twice with acetone, and dried under vacuum at 45 0 C. The resulting precipitate was resuspended in 67 % (v/v) H 2 SO 4 for 1 h at room temperature, and cellulose content was determined at absorbance 625 nm on an appropriate dilution of the sulfuric acid solution by the anthrone method (Scott and Melvin, 1953 was carried out essentially as described (Albersheim et al., 1967;Foster et al., 2010). Briefly, dried cell wall material from culms of both WT and SHN lines were treated with 2M trifluoroacetic acid for 90 min at 121 0 C, derivatizing the resulting solubilized monosaccharide to their alditol acetates (Albersheim et al., 1967), and analyzed on an GC-MS equipped with a SP-2380 (30 mm X 0.25mm X 0.25 µm df; Supelco) column. The peaks are identified by mass profiles and/or retention times of standards. Monosaccharides are quantified based on standard curves.

DFRC (Derivatization Followed by Reductive Cleavage) is a method for lignin compositional
analysis that produces analyzable monomers and dimers by cleaving α -and β -ethers in lignins (Lu andRalph, 1997, 1998). Dry plant samples (~20 mg) were suspended in acetyl bromide/acetic acid (20 % v/v) in a 10-ml round-bottom flask and gently stirred at 50 0 C for 3.5 h. Solid residues were filtered off, and the filtrate was evaporated by a rotary evaporator at 50 0 C under reduced pressure. After evaporation, the residue was dissolved in 2.5 ml of dioxane/acetic acid/water (5:4:1 v/v/v), and zinc dust (50 mg) was added to the solution as it was stirred.
Stirring was maintained for 30 min. Before gas chromatography-mass spectrometry (GC-MS) analysis, a solid-phase extraction was applied for collection of monomers while dimers and oligomers were removed. Thus, crude acetylated DFRC products in 50-100 μ l dichloromethane were loaded onto a preconditioned SPE column (3 ml silica, normal phase) and eluted with 12 ml cyclohexane/ethyl acetate (5/1, v/v). The eluant was concentrated to about 0.1 ml for GC-MS analysis.

Gene Expression Analysis
Total RNA was isolated from the rice leaf and culm tissue of WT and SHN lines using the RNeasy plant kit (Qiagen, USA), RNA quantity/quality measured by the Agilent 2100 Bioanalyzer (Agilent Technolgies, USA). For each sample 4 µg total RNA was used to generate first-strand cDNA with a T7-Oligo(dT) primer. Following second-strand synthesis, in vitro transcription was performed using the GeneChip® IVT Labeling Kit. The preparation and processing of labelled and fragmented cRNA targets, as well as hybridization to rice Affymetrix GeneChips, washing, staining, and scanning were carried out according to manufacturer's instructions (http://www.affymetrix.com). The RNA samples used for the microarray experiments were also used to synthesize cDNA templates for qRT-PCR analysis. The comparative Ct method of quantitation was used with the Actin gene as a reference (Ambavaram and Pereira, 2010). The relative fold-change for each of the selected genes was detected from both the WT and SHN transgenic lines. Three independent biological replicates of each sample and two technical replicates of each biological replicate were used for expression analysis. For each sample, 1 μ g total RNA from one of the biological replicates was converted into cDNA using oligo-dT 15-mer (Promega) and Superscript III reverse transcriptase (Invitrogen Life Technologies, Carlsbad, CA). This cDNA was diluted to 250 μ L in sterile water. Validation experiments were performed on 5 to 6 log dilutions of each of the target genes together with the Actin reference to determine if the amplification efficiencies were equal. Triplicate qRT-PCR reactions were performed using iQ™SYBR® Green Supermix in a Bio-Rad iQ5™ thermo cycler (Bio-Rad, Hercules, CA). Melting curve analysis, by applying increasing temperature from 55°C to 95°C (0.5°C/10 s), and gel electrophoresis of the final product confirmed single amplicons.
Negative control reactions using untranscribed RNA were also run to confirm absence of genomic DNA. To determine relative fold differences for each sample in each experiment, the Ct value for each gene was normalized to the Ct value for Actin and was calculated relative to a calibrator using the equation 2 -∆∆Ct (Livak and Schmittgen, 2001).

Reannotation of Rice Genechip Probe-Gene Mapping
A high-quality custom chip definition file (CDF) was built for the rice GeneChip array by uniquely mapping 442,810 probe sequences (http://www.affymetrix.com/analysis/downloads/data/) to 35,161 rice gene-based probesets in the following manner: (i) probes that have perfect sequence identity with a single target gene were selected, (ii) probes mapping to reverse complements of genes were annotated separately as antisense probes (not used in the above counts), and finally, (iii) probes were grouped into probe sets, each corresponding to a single gene, and probe sets with at least 3 probes were retained (>98% probe sets have >=5 probes). Note that these stringent criteria used to construct the CDF make it possible to reliably measure expression values of members of multigene families (free from cross-hybridization between paralogs showing high sequence similarity) and to get around 'one gene to multiple probesets' ambiguities. Moreover, improving probe-gene mapping also considerably aids in better quantification of coexpression by filtering out false-positive correlations (Casneuf et al., 2007). This new custom CDF is available from NCBI with the GEO accession number GPL11322.

Analysis of Differential Gene Expression
Raw data from the SHN overexpression experiment were background corrected, normalized and summarized according to the custom CDF using RMA (Ihaka and Gentleman, 1996;Irizarry et al., 2003;Gentleman et al., 2004), followed by non-specific filtering of genes that do not have enough variation (interquartile range (IQR) across samples < IQR median ) to allow reliable detection of differential expression. A linear model was then used to detect differential expression of the remaining genes (Smyth, 2004). The p-values from the moderated t-tests were converted to q-values to correct for multiple hypothesis testing (Storey and Tibshirani, 2003), and genes with q-value <0.1 were declared as differentially expressed in response to AtSHN expression. WIN1 overepxression data (Broun et al., 2004) was obtained from NCBI GEO accession GSE1071 and raw data was similarly pre-preprocessed using RMA based on a custom CDF for the Arabidopsis Genome array obtained from http://brainarray.mbni.med.umich.edu/Brainarray/ (Dai et al., 2005).

Curation of Lignin and Cellulose Biosynthetic Genes and Putative Regulators
Rice genes involved in cell wall biosynthesis and regulation were identified as homologs of Arabidopsis cell wall-related genes (Remm et al., 2001;Yokoyama and Nishitani, 2004;Zhong and Ye, 2009) Table S5).
Raw data were background corrected, normalized and summarized according to the custom CDF using justRMA (Irizarry et al., 2003), and expression values were averaged across replicates. Pearson correlations were first calculated between every pair of genes (Huttenhower et al., 2008) (see Note at the end of this subsection), which were then Fisher Z-transformed (David, 1949) and standardized to get coexpression scores (z cs ) with a N(0,1) distribution. This formulation was robust and highly interpretable as deviations from the expected value, and even by level of significance where |z cs | values greater than 1.645, 1.96 and 2.58 correspond to 10%, 5% and 1% extremes of the distribution of z cs scores.
TFs of interest were connected to each other when they had a strong correlation (|z cs | >1.645). TFs were connected to pathway genes based on a more rigorous procedure: For each set of 'pathway' genes, 'lignin' and 'other cell wall', a pathway correlation matrix was first created taking the pathway genes along the columns, and the pathway genes plus the putative TFs along the rows. Here, cell (i, j) contained the coexpression score z cs (i, j) between genes i and j. This was same as the adjacency matrix of the pathway genes, only with extra rows of TFs. Thus, each row i contained the vector of z cs 's of gene i to all the pathway genes, measuring its 'association' with the entire pathway. Pearson correlations were then calculated between rows, and TFpathway-gene pairs with absolute correlation >0.8 were selected. This correlation measures how well the genes agree with the regulatory program of the pathway. TF-TF edges were left undirected while TF-pathway-gene edges were directed from TF to putative target. Among the pathway genes, only those regulated by SHN (from the expression studies) were included in the final network. The network was visualized using Cytoscape v2.7.0 (Shannon et al., 2003), and isavailable as Supplemental File S1, which can be imported into Cytoscape. Note: There are popular methods to derive regulatory networks from gene expression data based on calculation of mutual information (MI) -for example, ARACNE (Basso et al., 2005) and CLR (Faith et al., 2007) -that perform better than simple correlation-based methods. But, these methods require very large amounts of gene expression data to contain expression of the genes across a large dynamic range to calculate MI reliably. Hence, with relatively less data in rice, especially when considering those similar in biological context, using MI-based methods would not be possible. Then, for measuring simple correlations, the Spearman rank correlation metric is a good choice. But, given we have carefully chosen datasets similar in biological context, identical in experimental platform, and resolved ambiguity in hybridization using a redefinition of probe-gene mapping in the array, we sought to using a metric more sensitive to the actual expression values, like the Pearson correlation coefficient, rather than one that works on the relative ranks of the values, like the Spearman rank correlation coefficient.

Promoter Analysis
For promoter analysis in rice, FIRE (Elemento et al., 2007) was used to discover motifs specific to the cell wall pathway genes by comparing the motif content of 1 Kb upstream sequences of these genes to that of the rest of the genome. Briefly, FIRE seeks to discover motifs whose patterns of presence/absence across all considered regulatory regions (motif profile) are most informative about the expression of the corresponding genes (expression profile). The sole reported motif 'CACCA[ACG]NC[AC]' was compared to known cis-elements in public databases (Higo et al., 1999;Crooks et al., 2004;Mahony and Benos, 2007) to identify it as one similar to the Arabidopsis AC II element. This de novo approach was taken since cis-regulatory motifs could diverge quickly across species making them hard to find simply by searching. A Perl script was used to search for GCC-box motifs '[AG]CCGNC' in the 1 Kb upstream sequences of SHN-regulated TFs.

Assays
The AtSHN coding region was PCR amplified using proofreading DNA polymerase with gene- was incubated for 10 min on ice before adding 10 or 100-fold excess of unlabeled competitor DNA and the reaction mixture was further incubated for 20 min at room temperature before loading onto 5% native polyacrylamide gel. The resolved DNA-protein complexes were electroblotted onto nylon membrane and subsequently detected using the chemiluminescent Detection Kit.

Accession Numbers
The Rice Genome Annotation or the locus ID numbers for the genes investigated in this study are OsSHN (Os06g40150), OsCAD1 (Os10g11810), OsCAD2 (Os02g09490), OsCAD3         Table S1 for IDs, names and annotations of all the genes in the network. This network has been provided in Supplemental File S1 and can be imported into Cytoscape.  42 Table I. Quantitative comparison of the wall thickness of sclerenchyma and bundle sheath cells in the culms of WT and SHN lines.
www.plantphysiol.org on August 30, 2017 -Published by Downloaded from Copyright © 2010 American Society of Plant Biologists. All rights reserved.