Rice cellulose synthase-like D4 is essential for normal cell-wall biosynthesis and plant growth


  • Ming Li,

    1. National Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Guangyan Xiong,

    1. National Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Rui Li,

    1. National Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Jiajun Cui,

    1. National Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Ding Tang,

    1. National Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
  • Baocai Zhang,

    1. National Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
  • Markus Pauly,

    1. DOE Plant Research Lab and Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, MI 48824, USA
    Search for more papers by this author
  • Zhukuan Cheng,

    Corresponding author
    1. National Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
  • Yihua Zhou

    Corresponding author
    1. National Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author

For correspondence (fax +86 10 64873428; e-mail yhzhou@genetics.ac.cn or zkcheng@genetics.ac.cn).


Cellulose synthase-like (CSL) proteins of glycosyltransferase family 2 (GT2) are believed to be involved in the biosynthesis of cell-wall polymers. The CSL D sub-family (CSLD) is common to all plants, but the functions of CSLDs remain to be elucidated. We report here an in-depth characterization of a narrow leaf and dwarf1 (nd1) rice mutant that shows significant reduction in plant growth due to retarded cell division. Map-based cloning revealed that ND1 encodes OsCSLD4, one of five members of the CSLD sub-family in rice. OsCSLD4 is mainly expressed in tissues undergoing rapid growth. Expression of OsCSLD4 fluorescently tagged at the C- or N-terminus in rice protoplast cells or Nicotiana benthamiana leaves showed that the protein is located in the endoplasmic reticulum or Golgi vesicles. Golgi localization was verified using phenotype-rescued transgenic plants expressing OsCSLD4–GUS under the control of its own promoter. Two phenotype-altered tissues, culms and root tips, were used to investigate the specific wall defects. Immunological studies and monosaccharide compositional and glycosyl linkage analyses explored several wall compositional effects caused by disruption of OsCSLD4, including alterations in the structure of arabinoxylan and the content of cellulose and homogalacturonan, which are distinct in the monocot grass species Oryza sativa (rice). The inconsistent alterations in the two tissues and the observable structural defects in primary walls indicate that OsCSLD4 plays important roles in cell-wall formation and plant growth.


Plant height is an important agronomic trait of crops. In the 1960s, breeding of semi-dwarf wheat and rice gave rise to the ‘green revolution’ and increased grain yields to a new level. Genetic studies using dwarf mutants of Arabidopsis and other species have revealed that this important phenotype is controlled by various factors such as gibberellins (Ashikari et al., 1999; Ueguchi-Tanaka et al., 2000), brassinosteroids (Li and Chory, 1997; Azpiroz et al., 1998; Hong et al., 2003) and cell-wall biogenesis (Reiter et al., 1993). However, our knowledge about the relationship between dwarfism and cell-wall formation at the molecular level is very limited.

Dwarfism is often correlated with problems in cell processes such as morphogenesis and propagation. Because the establishment of cell size and cell shape is associated with cell-wall deposition and metabolism, and progress through the cell cycle includes crucial steps in biogenesis of a new cell plate, any attempt to understand plant morphogenesis will ultimately face the challenge of understanding cell-wall synthesis and expansion (Somerville et al., 2004). Of the thousands of genes involved in cell-wall biogenesis and remodeling, the cellulose synthase super-family that contains cellulose synthase active subunits (CESA) and CESA-like genes (CSL) is very important. CESA and CSL proteins are large (approximately 850–1000 amino acid residues), and share structural similarities, including three to six transmembrane domains toward the C-terminus and one or two toward the N-terminus, and a common D_D_D_QxxRW motif, which is believed to be involved in catalytic activity (Richmond and Somerville, 2000, 2001). CESAs are responsible for biosynthesis of cellulose, the core component of cell walls (Somerville, 2006), while CSL proteins are thought to catalyze biosynthesis of the various β-linked glycan backbones (Lerouxel et al., 2006). Plant genomes contain tens of CSLs, which are divided into nine sub-families, CSLs A–H (Keegstra and Walton, 2006) and CSL J (Fincher, 2009). Many attempts have been made to understand the specific functions of CSLs. Studies using transcriptional profiling of guar seeds (Dhugga et al., 2004) and in vitro enzymatic activity analysis (Liepman et al., 2005) have shown that the CSLA sub-family is involved in β-1,4-mannan synthesis. Using the nasturtium (Tropaeolum majus) seed system, expression of a CSLC gene was found to correlate with the deposition of xyloglucan (XyG). Heterologous expression of this gene and the Arabidopsis homolog, AtCSLC4, in the yeast Pichia pastoris resulted in production of a β-1,4-glucan, demonstrating their activity in XyG backbone biosynthesis (Cocuron et al., 2007). Burton et al. (2006) identified six rice CSLF genes that control the synthesis of mixed-linkage glucan (β-1,3;1,4-glucan, MLG) based on a barley quantitative trait locus (QTL) using a comparative genomics approach. Testing for the presence of MLG in the transformed Arabidopsis plants (which lack MLG per se) by expressing OsCSLF2 and OsCSLF4 confirmed its function in catalyzing the backbone biosynthesis of MLG. More recently, Doblin et al. (2009) reported that the CSLH sub-family in barley also mediates MLG synthesis. However, the enzymatic activities of the other five CSL sub-families (CSLB, CSLD, CSLE, CSLG and CSLJ) are less well understood.

Owing to their high sequence similarity to CESAs, it has been suggested that CSLDs function as cellulose synthases (Doblin et al., 2001); however, there is inadequate evidence to support this hypothesis. Several mutants that have altered levels of AtCSLD transcripts have been described. For example, kojak and csld3 in Arabidopsis display defects in root hair growth (Favery et al., 2001; Wang et al., 2001). Studies on Arabidopsis insertion mutants of CSLD1, CSLD2 and CSLD4 revealed that disruption of these genes results in defects in the production of normal tip-growing cells, including root hair and pollen tubes (Bernal et al., 2008). The functional ortholog of AtCSLD3 in rice is OsCSLD1, which is also required for root hair elongation (Kim et al., 2007). AtCSLD5 appears to play a role in cellulose synthesis, because its expression was significantly increased in an Arabidopsis suspension cell line habituating to the cellulose synthase inhibitor isoxaben (Manfield et al., 2004). Four T-DNA insertion lines of AtCSLD5 were generated. The knockout and atcsld5-1 plants exhibited reduced growth and altered xylan levels, indicating that AtCSLD5 functions in xylan production and plant development (Bernal et al., 2007). These lines of genetic evidence help us to understand the roles of the CSLD sub-family in plant development and cell-wall biogenesis.

Significant compositional differences in cell walls are well known between monocot grasses and dicots (Carpita and Gibeaut, 1993). These cell walls differ considerably in the types, relative abundance and cross-linking of their polysaccharides (Vogel, 2008). Therefore, the walls of higher plants can be divided into two major categories according to their composition and structure (Carpita and Gibeaut, 1993; Carpita, 1996). In type I cell walls, which are often found in dicots, non-commelinoid monocots and gymnosperms, cellulose fibers are embedded in a network of XyG, pectin and structural proteins. In type II cell walls, which exist only in the commelinoid monocots (Poaceae and related families), the network of non-cellulosic polysaccharides is enriched in glucuronoarabinoxylans (GAX) and MLG, with very low levels of galactomannan, pectin and structural proteins (Vogel, 2008). These differences are due to the presence or otherwise of genes responsible for their biosynthesis and remodeling. Therefore, studying the role and behavior of specific genes in dicot and monocot species will provide new insights into understanding cell-wall biogenesis at the global level.

Among nine sub-families of CSLs, some appear to be common in all plants, whereas others are present only in specific species (Farrokhi et al., 2006; Keegstra and Walton, 2006; Vogel, 2008). CSLD is present in both grasses and dicots, and its product may thus occur in all plants. As mentioned above, although five csld mutants have been reported in Arabidopsis and one in rice, most of these studies focused on identification of their developmental roles. Here, we report the functional characterization of rice CSLD4. Detailed wall compositional analyses revealed the distinct effects of oscsld4 on arabinoxylan (AX) structure, the major hemicellulose in grass cell walls, and also revealed an inconsistent alteration in the amounts of homogalacturonan (HG) and xylan, which were different from those commonly found in Arabidopsis. Thus, the data presented here highlight the functions of OsCSLD4 in the monocot grass species rice.


nd1 plants have reduced plant height

To understand the mechanism that controls rice height, we isolated a narrow leaf and dwarf1 (nd1) mutant from the γ-ray-irradiated rice indica cultivar Zhongxian 3037, and performed in-depth characterization. The nd1 plants exhibited dwarfism throughout growth and development. Two-week-old mutant seedlings had reduced length in both the aerial part and the roots (Figure 1a). At the mature stage, the height of mutant plants was only 50% of that of the wild-type plants (Figure 1b). The reduced height resulted from uniformly shortened internodes in the mutant culms which was confirmed by comparison of the length of each internode between the mutant and wild-type (Figure 1c–f). In addition to dwarfism, nd1 plants showed narrow leaves (Figure 1d) and thin culms, as well as additional morphological abnormalities, including significantly reduced panicle size and fertility, small grains, and slightly increased tiller number (data not shown). Therefore, the nd1 mutation results in reduced plant height and overall growth abnormalities.

Figure 1.

 Phenotypic characterization.
(a) Two-week-old seedlings of wild-type and nd1.
(b) Mature plant of wild-type and nd1.
(c) Internodes of wild-type and nd1.
(d) Leaves of wild-type and nd1.
(e) Quantification of the wild-type and nd1 internode length (mean of 15 culms ± SD).
(f) Percentage contribution of each internode to the total plant height of wild-type and nd1 plants (mean of 15 culms ± SD).
I–V, the 1st to 5th internodes. Scale bars = 1.5 cm (a) and 10 cm (b).

The small size of nd1 results from reduced cell number

The generally small size of nd1 prompted us to examine the anatomical features of the mutant and wild-type plants. Internode sectioning revealed that the cell size in the transverse and longitudinal directions was not significantly different in the mutant and wild-type plants (Figure 2a–f), but there were one or two fewer layers of sclerenchyma cells in nd1 plants (Figure 2d) compared with wild-type plants (Figure 2c). We also compared the cell number in the longitudinal direction, which contributed to the length of each internode, between nd1 and wild-type plants. The cell number in each internode of nd1 was only 40–70% of that in the wild-type (Figure 2g). Cross-sectioning of the leaves (Figure 2h,i) and leaf sheaths (Figure 2j,k) further demonstrated that the thin culms and narrow leaves resulted from decreased cell number, as determined by the number of vascular bundles (Figure 2l,m). Therefore, reduced cell number causes the small size of the mutant plants.

Figure 2.

 Anatomical features of the wild-type and nd1 plants.
(a–d) Cross-sections of wild-type (a, c) and nd1 (b, d) culms, showing the abnormally thickened parenchyma cells [inset in (b)] and reduced number of layers of sclerenchyma cells (SC) (d) in nd1.
(e, f) Longitudinal sections of wild-type (e) and nd1 (f) culms.
(g) Number of parenchyma cells in each internode (mean of four culms ± SD).
(h, i) Cross-sections of wild-type (h) and nd1 (i) leaves.
(j, k) Cross-sections of wild-type (j) and nd1 (k) leaf sheaths.
(l) Quantification of the large (LV) and small veins (SV) in leaves of wild-type and nd1 plants (mean of five leaves ± SD).
(m) Quantification of the vascular bundles in leaf sheaths (LS) of wild-type and nd1 (mean of three leaf sheaths of the1st and 2nd internodes ± SD).
I–V, mean of the 1st and 2nd leaf sheaths in three seedlings ± SD. Sacle bars = 100 μm (a, b, e, f), 40 μm (c, d), 300 μm (h, i) and 150 μm (j, k).

To investigate the cause of the reduced cell number, we compared the progression of cell division in the root tips of nd1 and wild-type plants. As shown in Figure S1, the four microtubule arrays identified using anti-α-tubulin antibody in a typical dividing phase during mitosis showed neither structural nor morphological deficiencies in the mutant. However, the proportion of cells involved in division was only 50% of that in wild-type plants, based on a quantitative assay (Table S1). Taken together, these results suggest that the reduced growth in mutant plants, including dwarfism and narrow leaves, may result from altered cell-cycle progression.

Positional cloning of the nd1 locus

To investigate the molecular basis of nd1, we performed positional cloning of this gene. From the F2 population generated by crossing nd1 mutants with Wuyunjing 8, a wild-type polymorphic japonica variety, 7024 homozygous mutant plants were obtained and used for genetic analysis. The nd1 locus was mapped between the molecular markers R887 and S15552 on chromosome 12 (Figure 3a). Further mapping using adjacent molecular markers (Table S2) indicated that the nd1 locus is in BAC clone AL845342, and its position was gradually narrowed to a 24 kb DNA region (Figure 3a). Based on the annotations of the rice genome database, this DNA region comprises two putative open reading frames (ORFs). We identified one of these genes, Os12g36890, which encodes OsCSLD4, as the most likely candidate. We sequenced this ORF and compared it with that of the wild-type. A single base pair substitution was found in the second exon at nucleotide 2894, which changed GCC to GTC, resulting in a change from alanine to valine at the 965th amino acid residue (Figure 3a).

Figure 3.

 Map-based cloning of ND1.
(a) The nd1 locus was mapped to a 24 kb DNA region on chromosome 12. The ND1 gene comprises two exons (boxes) and one intron (line).
(b) Constructs for complementary analyses. pCSLD4 and pCSLD4T were used for transformation of nd1 mutant plants, and pCSLD4RNi was used for transformation of wild-type plants. The dotted line in pCSLD4T indicates the deleted genomic region.

To confirm that Os12g36890 corresponds to the nd1 locus, 10.27 or 3.5 kb genomic DNAs with or without the entire ORF were inserted into vector pCAMBIA 1300 to generate plasmids pCSLD4 and pCSLD4T, respectively (Figure 3b). These plasmids were then transformed into nd1 mutant plants, and 132 and 78 independent transgenic lines were obtained, respectively. The mutant phenotypes were fully rescued in 130 lines carrying the Os12g36890 ORF (Figure S2a), but not in the 78 lines without the ORF (Figure S2b). Moreover, OsCSLD4 knockdown transgenic plants generated by transforming an RNAi construct (Figure 3b) into wild-type plants (Figure S2c) mimicked the nd1 phenotypes (Figure S2d). Therefore, Os12g36890 represents the nd1 gene.

ND1 encodes a CSLD member

Sequencing analysis revealed that the ND1 gene has only one intron and its cDNA is 4123 bp in length, with an ORF of 3648 nucleotides encoding a protein of 1215 amino acids. A Pfam search indicated that ND1 is a CSLD4, as it has features that are characteristic of CSLs, such as 6–8 transmembrane domains and a D_D_D_QxxRW motif (Figure 4a). The mutated alanine in nd1 is highly conserved in all CSLD members of higher plants, indicating the importance of this residue (Figure 4a). A BLASTX search found five CSLD members in the rice genome and six in Arabidopsis. To determine the role of OsCSLD4, an unrooted tree was constructed for the CSLDs of Arabidopsis and rice using the neighbor-joining method. OsCSLD4 was placed in a monophyletic clade with AtCSLD5, with 100% bootstrap support (Figure 4b), which may provide important clues for understanding the function of OsCSLD4 in rice.

Figure 4.

ND1 encodes OsCSLD4.
(a) The mutated residue (alanine at position 965, indicated by an arrowhead) is conserved in the CSLD sub-family of rice, Arabidopsis and other plant species. The D_D_QxxRW conserved motif is also indicated.
(b) Phylogenetic analysis of the CSLD members in rice and Arabidopsis. The scale bar is an indicator of genetic distance based on branch length.
The accession numbers of CSLDs used in this analysis are At2g33100 (AtCSLD1), At5g16910 (AtCSLD2), At3g03050 (AtCSLD3), At4g38190 (AtCSLD4), At1g02730 (AtCSLD5), At1g32180 (AtCSLD6), Os10g42750 (OsCSLD1), Os06g02180 (OsCSLD2), Os08g25710 (OsCSLD3), Os12g36890 (OsCSLD4), Os06g22980 (OsCSLD5), XP_001769140 (PpCSLD3), AAO03579.1 (PtCSLD4), AAK49455.1 (NaCSLD) and CAN70317.1 (VvCSLD). Na, Nicotiana alata; Pp, Physcomitrella patens; Pt, Populus tomentosa; Vv, Vitis vinifera.

OsCSLD4 is mainly expressed in rapidly growing tissues

To define the expression pattern of OsCSLD4, we surveyed its expression in various tissues of rice using RT-PCR. OsCSLD4 was highly expressed in the callus derived from seeds, root tips, the stem apical meristem and young panicles, but showed low levels in mature tissues such as culms, leaves and leaf sheaths (Figure 5a). We also generated transgenic plants harboring the OsCSLD4 putative promoter region (2315 bp before ATG) and the GUS gene to verify expression at the cellular level. As shown in Figure 5(b–k), GUS signals were mainly observed in root tips, including those of primary (Figure 5b,h), lateral (Figure 5c,d) and adventitious (Figure 5c,i) roots, even at their initiation stage (Figure 5e). Wax sectioning further showed GUS signals in rapidly growing and undifferentiated cells (Figure 5f–h). In the aerial part, GUS staining was detected in the tiller buds, intercalary meristem (Figure 5j), and the callus developed from T1 transgenic seeds (Figure 5k), consistent with the results of RT-PCR (Figure 5a). RNA in situ hybridization further showed OsCSLD4 signals in initiating tiller buds and the intercalary meristem (Figure 5l–n). All these lines of evidence suggest that OsCSLD4 is expressed in rapidly growing tissues, explaining the observed phenotype of reduced growth in mutant plants.

Figure 5.

 Expression pattern of OsCSLD4.
(a) RT-PCR amplification of OsCSLD4 in various rice tissues, using the ubiquitin gene as the internal control. SAM, stem apical meristem.
(b–k) GUS patterns in various organs of pOsCSLD4::GUS transgenic plants, showing GUS signals in a rice seedling (b), meristematic cells of adventitious roots (c), lateral roots (d), initiating lateral roots (e), wax sections of lateral roots (f), an enlargement of the lateral roots (g), primary root (h), the initiating adventitious roots at the internodes (i), a hand-cut section of the heading shoot (j), and a callus from T1 seeds (k).
(l, m) RNA in situ hybridization of the longitudinal sections of young stem.
(n) Background control of in situ hybridization using a sense probe.
AR, adventitious roots; TB; tiller buds; IM, intercalary meristem; PV, peripheral vascular; NV, net vascular. Scale bars = 0.5 cm (b), 0.4 cm (c–e), 40 μm (f, h), 100 μm (g), 0.4 cm (i–k) and 70 μm (l–n).

OsCSLD4 is a Golgi-localized protein

Several reported CSLDs from Arabidopsis are located in the endoplasmic reticulum (ER) and/or Golgi apparatus (Favery et al., 2001; Bernal et al., 2007). To determine the localization of OsCSLD4, we transiently expressed C/N-terminal GFP fusions of OsCSLD4 in rice protoplast cells or Nicotiana benthamiana leaves. Expressing GFP alone resulted in generally distributed signals in the cytoplasm, plasma membrane and nucleus (Figure 6a). OsCSLD4–GFP was limited to the fibril network within the cytoplasm (Figure 6b), whereas GFP–OsCSLD4 displayed a punctate pattern (Figure 6c), suggesting localization in the ER and Golgi apparatus. We verified this localization by co-transformation of OsCSLD4–GFP with the ER marker protein mCherry–HDEL in rice protoplast cells. The almost completely merged signals of the two fluorescent proteins suggest that OsCSLD4 is localized to the ER (Figure 6d–f). However, for unknown reasons, very few GFP–OsCSLD4-transformed rice protoplast cells were obtained. We therefore also expressed it in leaves of Nicotiana benthamiana. As shown in Figure 6(g–i), GFP–OsCSLD4 showed a punctate distribution that almost overlapped with that of the Golgi marker Man49–mCherry, indicating that OsCSLD4 is a Golgi-localized protein. We further generated transgenic plants expressing OsCSLD4–GUS driven by its own promoter in the nd1 background. The wild-type appearance of transgenic plants indicated that the OsCSLD4–GUS fusion protein functions as a native protein. When probed using anti-GUS antibodies, punctate signals were detected in the cells of culms and roots of T1 plants (Figure 6j–o), in a pattern that was almost identical to obtained using the Golgi tracker, BODIPY®-TR C5-ceramide. Therefore, OsCSLD4 is very likely localized in the Golgi apparatus. The fact that the C-terminal fusion driven by 35S or its own promoter showed a diverse subcellular localization suggests that OsCSLD4 may have multiple locations in cells. The ER targeting may also result from disruption of normal targeting due to the constitutive expression.

Figure 6.

 Subcellular localization of OsCSLD4.
(a–c) Rice protoplast cell expressing GFP alone (a), OsCSLD4–GFP (b) and GFP–OsCSLD4 (c).
(d–f) Rice protoplast cell co-expressing OsCSLD4–GFP (d) and ER marker mCherry–HDEL (e), with a merged image (f).
(g–i) Nicotiana benthamiana leaf cell co-expressing GFP–OsCSLD4 (g) and Golgi marker Man49–mCherry (h), with a merged image (i).
(j–m) Immunochemical staining of the OsCSLD4–GUS transgenic culm section using the Golgi tracker BODIPY®-TR C5-ceramide (j), anti-GUS antibodies (k); (l) merged image and (m) DIC image.
(n, o) Immunochemical staining of OsCSLD4–GUS transgenic root cells with anti-GUS antibodies (n); (o) DIC image.
Scale bars = 10 μm (a–f), 5 μm (g–i), 20 μm (j–m) and 10 μm (n, o).

The OsCSLD4 mutation alters wall formation

CSLDs are believed to be directly involved in wall polysaccharide synthesis. To investigate whether the mutation in OsCSLD4 results in observable wall defects, we compared the wall structure of culms and roots between the mutant and wild-type under transmission electron microscopy. As shown in Figure 7(a–f), some parenchyma cell walls in the mutant culms were abnormally thickened (Figure 7b,d–f) compared with those in the wild-type plants (Figure 7a,c). This result is consistent with those shown in Figure 2(b). In addition, some walls were severely folded and filled with electron-dense material (Figure 7d,f). Because OsCSLD4 is expressed in root tips and affects cell propagation, we analyzed the walls at the root division zone (Figure 7g–j). Some walls of the mutant plants were folded and knotted (Figure 7h–j), similar to culm walls. Therefore, the polymer produced by OsCSLD4 is essential for normal primary wall structure.

Figure 7.

 Micrographs of culms and root tips in wild-type and nd1 plants.
(a, b) Culm cross-sections, showing the parenchyma cell walls of wild-type (a) and nd1 (b) plants. Arrows show the abnormally thickened walls.
(c–f) Transmission electron micrographs of wild-type (c) and mutant (d–f) parenchyma cell walls. Arrows show the abnormally thickened and folded walls.
(g, h) Longitudinal sections of wild-type (g) and nd1 (h) root tips.
(i, j) Transmission electron micrographs of the nd1 root walls. Arrows show the folded and knotted walls.
Scale bars = 100 μm (a, b), 2 μm (c–f), 5 μm (g, h) and 500 nm (i, j).

To confirm these results, we used various antibodies that recognize specific polysaccharide epitopes to explore the compositional differences between wild-type and nd1 walls. JIM7 and JIM5 are monoclonal antibodies that label homogalacturonans (HG) that have high and low degrees of esterification, respectively (Knox et al., 1990). The culms and roots of nd1 exhibited increased labeling with the JIM7 probe (Figure 8a–d), in contrast to the almost equal signals in the walls of wild-type and mutant cells probed with JIM5 (Figure 8e,f). LM6, which recognizes an epitope in α-1,5-arabinan (Willats et al., 1998), labeled the walls of culms similarly in mutant and wild-type (Figure 8g,h), but labeled mutant roots with lower intensity than wild-type (Figure 8i,j). However, in cells labeled with LM10, which is specific to β-1,4-xylan (McCartney et al., 2005), the reduction in fluorescent signals was pronounced in mutant culms (Figure 8k,l) but not mutant roots (data not shown). These results indicate that the wall composition is altered in mutant plants.

Figure 8.

 Immunochemical labeling of sections of wild-type and mutant culms and roots with various sugar antibodies.
(a, b) Culm cross-sections probed with JIM7.
(c, d) Root cross-sections probed with JIM7.
(e, f) Culm cross-sections probed with JIM5.
(g, h) Culm cross-sections probed with LM6.
(i, j) Root cross-sections probed with LM6.
(k, l) Culm cross-sections probed with LM10.
Scale bars = 10 μm (culm sections) and 5 μm (root sections).

The nd1 culms have a reduced xylan and cellulose content, but an increased amount of homogalacturonan

The above results prompted us to examine the wall components in the mutant and wild-type plants. Because dwarfism is the major phenotype, the first target of compositional analysis was the culms, which contain both primary and secondary walls. As shown in Table 1, a number of wall compositional changes were observed. The cellulose content and xylose content in nd1 were decreased by approximately 14 and 17%, respectively, whereas that of arabinose was increased by 15%. In addition, the galacturonic acid (GalA) content, representing pectin, was increased by 54%. To identify the affected polymer in the mutant plants, we solubilized a pectic fraction from wall materials using ammonium oxalate and subsequently a hemicellulose fraction using 4 N KOH. We determined the monosaccharide composition of both fractions (Table 2). A major increase in GalA was found in the pectic fraction (over 160%). The almost unchanged rhamnose content in the pectin fraction indicated that the increased GalA level may largely result from the increased HG content (Table 2), in agreement with the increased labeling with JIM7 (Figure 8b,d). A major decrease (approximately 16% reduction) was found for xylose in the nd1 hemicellulose fraction. The concomitant increase in arabinose content in the same fraction suggests that the ndl culms contain a more highly substituted arabinoxylan (AX). A glycosidic linkage analysis of these fractions further confirmed reduction in the non-substituted xylosyl residue (4-Xyl) of AX in the 4 N KOH fraction (Table 3), with a 22% reduction in the ratio of 4-Xyl compared to 3,4- and 2,4-Xyl (substituted AX chain residue) in nd1. Consistently, a significant increase in Terminal Arabinose (T-Ara) suggested that the nd1 culms contain less xylan, but have a higher degree of arabinosylation.

Table 1.   Monosaccharide compositional analysis of wall residues from mature culms and 4-day-old root tips of wild-type and nd1 plants
  1. Alcohol-insoluble residues (AIR) were prepared from the 2nd internodes and 4-day-old root tips of wild-type and nd1 plants as described in Experimental procedures. Tetramethylsilane (TMS) derivatives were analyzed by GC-MS for glycosyl residue composition. The cellulose content was determined by the Updegraff method (see Experimental procedures). The results are given as means (mg g−1 of AIR) of three independent assays. The variance is not shown but is <15%.

  2. *Significance between the wild-type and mutant determined by the least-significant difference (t test at < 0.05).

4-day-old root tipsWT5.718.9127.3*110.8*6.631.516.036.18.1*292.3
Table 2.   Monosaccharide compositional analysis of wall fractions derived from mature culms and 4-day-old root tips of wild-type and nd1 plants
  1. aTotal alcohol-insoluble residue (AIR) was generated from culms and 4-day-old root tips of wild-type and nd1 plants. After extensive enzymatic de-starching, the AIR was sequentially fractionated by 0.5% hot ammonium oxalate followed by 4 N potassium hydroxide (4 N KOH). Soluble components were neutralized and dialyzed against water. The amount of sugar residues was determined by GC-MS analysis of TMS derivatives. For further details, see Experimental procedures. Data are the means (mg g−1 AIR) of four independent assays. The variance is not shown but is <15%.

  2. bThe residue remaining after 4 N KOH extraction was thoroughly rinsed with water. Samples were hydrolyzed in 2 M TFA. The amount of sugar residues was determined by GC-MS analysis of alditol acetate derivatives. The variance was <15%.

  3. cND, not determined.

  4. *Statistically significant difference with respect to the wild-type (t test at < 0.05).

CulmsAmmonium oxalateaWT3.**12.7*4.045.0
4 N KOH-solubleaWT1.8*1.960.3*271.9*1.33.9*4.52.9*3.7352.2
Remaining residuebWT1.0ND6.4*11.22.0*2.141.7NDcND64.4
4-day-old root tipsAmmonium oxalateaWT3.35.416.011.15.723.24.433.19.1111.3
4 N KOH-solubleaWT3.416.4131.3*117.2**4.5*334.8
Remaining residuebWT0.63.213.9*7.6*
Table 3.   Glycosyl linkage analysis of fractionated wall residues from mature culms and 4-day-old root tips of wild-type and nd1 plants
FractionsaAmmonium oxalate4 N KOH
CulmsRoot tipsCulmsRoot tips
  1. aThe fractions correspond to those analyzed in Table 2. Glycosyl residue linkage analysis was performed using a modification of the Hakamori method (1964).

  2. bSugar residues are expressed as a percentage of total peak areas.

  3. cND, not detected.

  4. *Statistically significant difference with respect to the wild-type (t test at < 0.05).


Taken together, these results lead us to suggest that the OsCSLD4 mutation decreases xylan and cellulose content but increases HG content in the walls of culms.

Mutation in OsCSLD4 decreases the terminal arabinose and increases branching xylose abundance in root tips

We also determined the wall composition of root tips, which mainly comprise primary walls. Unlike the monosaccharide composition in culms, the mutant root tips showed a reduction in arabinose content rather than in xylose and cellulose, whereas the glucuronic acid (GlcA) content was increased by approximately 28% compared with the wild-type (Table 1). Fractional compositional analysis revealed no changes in the pectic fraction, but a reduction in arabinose and a concomitant increase in xylose content was observed in the 4 N KOH fraction (Table 2). Glycosidic linkage analysis of the hemicellulosic fraction revealed a slight but significant increase in substituted xylan (3,4-linked), with a small but significant decrease in T-Ara (Table 3), in the mutant root tips. That a higher degree of substitution of xylan did not result in the expected increase in its substituents (T-Ara) might be due to the increased amount of GlcA in nd1 or contamination of this fraction with other polymers. In contrast to our observation of the nd1 culms, these results provide evidence that mutation in OsCSLD4 causes alterations in AX structure in the root tips.

The extractability of polysaccharides in nd1 and wild-type plants was also analyzed by examining the sugar composition of the wall residues after KOH extraction. As shown in Table 2, the amount of arabinose was increased in both the remaining residues of culms and 4-day-old root tips. This tightly linked arabinose is usually associated with highly cross-linked extensins (Egelund et al., 2007), and thus might indicate an increase in this glycoprotein in the nd1 walls. However, as the xylose content of this fraction of 4-day-old root tips was also increased, it is possible that the increase in arabinose might also arise from a more tightly linked hemicellulose fraction in nd1.


Although the overall architecture of the cellulose networks is similar, the walls of grasses and dicots differ with respect to their non-cellulosic polysaccharide contents, which comprise pectin and hemicellulose (Vogel, 2008). In contrast to dicot hemicellulose, which consists mainly of XyG and galactomannans, MLG and arabinoxylan (AX) are the major components of the walls of grasses (Carpita, 1996; Burton et al., 2006). CSL genes have been postulated to be responsible for the biosynthesis of the β-1,4 linkages found in the backbones of hemicellulose (Keegstra and Walton, 2006; Vogel, 2008). CSLD is found in both monocots and dicots, suggesting that the glycans that it synthesizes are common to all plant species. Therefore, studying the role of specific genes in both grass and dicot species will provide a better understanding of their functions in cell-wall biosynthesis. We have described here in-depth characterization of the nd1 mutant, map-based cloning of CSLD4, gene transcriptional and protein localization assays, as well as detailed wall structure and composition analyses. All the data indicate that OsCSLD4 plays an important role in rice wall biogenesis and plant development.

Many attempts have been made to determine the enzymatic activities of CSLs because those provide direct evidence of their functions. The heterologous expression of CSLs in yeast Pichia and S2 insect cells has shown that CSLA and CSLC function in galactomannan and XyG backbone elongation (Dhugga et al., 2004; Cocuron et al., 2007). However, dueto problems in correct folding of recombinant proteins and lack of co-factors, substrates or oligosaccharide primers in the heterologous expression systems, the activities of other CSLs remain elusive (Liepman et al., 2005). Although several mutants in the CSLD sub-family have been reported in Arabidopsis and rice, most of these studies were limited to identification of their roles in plant growth and development (Favery et al., 2001; Bernal et al., 2007, 2008; Kim et al., 2007). Because little is known about their biochemical activity, another means to identify CSLD function is to determine the exact deficiency in the walls of the mutants through wall compositional analyses. Until now, none of the csld mutants have been subjected to detailed wall compositional analyses to reveal their functions in cell-wall biogenesis.

Here, we isolated a csld4 mutant from rice and performed detailed wall compositional analyses, including monosaccharide composition and glycosyl linkage assays, combined with wall fractioning. The two organs used, culms and root tips, represent secondary and primary walls, respectively, and are major phenotype-changed tissues in the mutant. One of the observed wall structural changes is a reduction of xylan content and a higher degree of arabinosyl substituents in the culms. However, the xylan content is not significantly altered in root tips, although its substitution is increased. Other major changes were observed: notably, a decrease in β-glucan content (cellulose) and an increase in pectic HG content, suggesting that disruption of CSLD4 causes complex wall polysaccharide modifications. The wall alterations identified may be a consequence of a different cell type composition in the mutant or aberrant cell-wall formation. Bernal et al. (2007) have reported on a mutant of AtCSLD5, an ortholog of OsCSLD4 in Arabidopsis. Enzymatic and immunochemical staining analyses of atcsld5 revealed an apparent reduction in xylan (Bernal et al., 2007). However, the alterations in xylose and other sugars were not evaluated in atclsd5. AX is present in low amounts in Arabidopsis cell walls (Vogel, 2008; York and O’Neill, 2008). Thus, our study of the role of OsCSLD4 in regulating AX structure is unique in that is considers side-chain length and substitution pattern in monocot plants. Given the abundance of AX and the diverse sugar alterations in the two tissues of the rice nd1 mutant, it is possible that OsCSLD4 may not function as a xylan synthase. Another interesting finding from the chemical analyses is the inconsistent alteration of AX and HG in grass walls. Co-modulation of HG and xylan synthesis is common in Arabidopsis. For example, the activities and quantities of HG and xylan synthase were correlatively reduced in the quasimodo (Orfila et al., 2005), irregular xylem8 (Persson et al., 2007) and atcsld5 (Bernal et al., 2007) mutants, which may reflect a physical or functional association of these two polymers in Arabidopsis. Such inconsistency in the rice nd1 mutant suggests that the HG and AX of grasses may have different cross-linking patterns than those of dicots. Therefore, new insights into the role of CSLD4 in cell-wall biogenesis are provided in terms of detailed wall composition analyses of nd1 in the monocot grass species rice.

The pleiotropic polysaccharide modifications in nd1 indicate that OsCSLD4 is important for normal wall biosynthesis; the specific wall polymer whose biogenesis is catalyzed by OsCSLD4 remains to be elucidated. Given the difficulties in enzymatic and chemical approaches, use of transcript analysis is a means to explore a link between a gene and a specific polysaccharide product, because clear expression patterns may be linked to the progressive changes occurring in the cell wall (Farrokhi et al., 2006). Successful studies have been performed using some seed systems. Seeds of guar (Cyamopsis tetragonoloba) and nasturtium (Tropaeolum majus) accumulate large amounts of galactomannan and XyG during maturation. Transcriptional profiling of these seeds has uncovered evidence of CSLA and CSLC involvement in mannan and XyG backbone biosynthesis (Dhugga et al., 2004; Liepman et al., 2005; Cocuron et al., 2007). Increasing lines of evidence have shown that a number of genes in the CSL super-family are developmentally regulated (Favery et al., 2001; Goubet et al., 2003; Persson et al., 2005; Bernal et al., 2008), and are expressed at consistently lower levels than CESA genes (Hamann et al., 2004). Six CSLD genes have been annotated in the Arabidopsis genome, and five of them have been genetically characterized (Favery et al., 2001; Wang et al., 2001; Bernal et al., 2007, 2008). AtCSLD1 and AtCSLD4 are expressed exclusively in pollen tubes (Bernal et al., 2008), whereas the remaining CSLD genes are universally expressed, with the highest expression of AtCSLD2 and AtCSLD3 in roots (Favery et al., 2001; Wang et al., 2001; Bernal et al., 2008) and that of AtCSLD5 in the shoot apex and hypocotyls (Bernal et al., 2007). Consistent with the characteristics of Arabidopsis orthologs, OsCSLD1, which is required for root hair growth, is highly expressed in root hair cells (Kim et al., 2007), whereas OsCSLD4 is expressed in the apex of many organs. This expression pattern of OsCSLD4 hints at the difficulties of harvesting functioning tissues/cells for chemical analysis; any defect may be undetectable against a background of high levels of other wall polymers, resulting in failure to find specific deficiencies in the nd1 walls. This notion is further supported by our observation of wall structural abnormalities only in some primary cell walls of culms and root tip cells in nd1 plants. The polymer synthesized by OsCSLD4 is thus essential for the physical properties of the primary cell wall. Moreover, the wall is a dynamic polysaccharide network. Its structure and composition vary among cell types and at various developmental stages (Carpita and Gibeaut, 1993). Thus, it is also possible that the product of CSLD4 is temporarily incorporated into walls at an early stage of cellular development, but is then substituted by or required for the synthesis of other wall polymers, such as AX.

Of all the CSL gene families, the CSLD genes are the most closely related in sequence to the CESA genes that are required for β-1,4-glucan synthesis (Richmond and Somerville, 2001) and the OsCSLF genes, which catalyze the accumulation of β-1,3;1,4-glucan (Burton et al., 2006). CSLDs have therefore been postulated to produce another β-linked glucan (Favery et al., 2001; Wang et al., 2001). Our finding that OsCSLD4 is a Golgi protein is consistent with previous reports (Bernal et al., 2007, 2008), suggesting that OsCSLD4 is unlikely to participate in the synthesis of crystalline cellulose, which occurs at the plasma membrane. However, the rice proteomic analysis identified OsCSLD2 as a plasma membrane protein (Natera et al., 2008), indicating a potential role in cellulose biosynthesis. The highly up-regulated expression of AtCSLD5 in isoxaben-habituated cells, which accumulated a less crystalline form of cellulose, implies that CSLDs are involved in the synthesis of non-crystalline cellulose (Manfield et al., 2004). Although significant decreases in mannose or its β-linked sugars were not detected in the nd1 mutant plants, the possibility that OsCSLD4 acts as a mannosyltransferase or glucosyltransferase cannot be excluded.

Experimental procedures

Plant materials and growth conditions

The rice (Oryza sativa) narrow leaf and dwarf mutant nd1 was isolated from the γ-ray-irradiated indica cultivar Zhongxian 3037. The F2 mapping population was generated from a cross between the nd1 mutant and Wuyunjing 8, a polymorphic japonica cultivar. In the F2 population, plants exhibiting dwarfism and narrow leaves were selected for gene mapping. All plants were grown in the paddy fields of the Institute of Genetics and Developmental Biology in Beijing (China) or Sanya (Hainan Province, China) during the natural growing season.

Positional cloning

ND1 was mapped and cloned for 7024 F2 mutant plants using molecular markers (Table S2). The corresponding DNA fragments were amplified from mutants and wild-type plants using LA-Taq (TaKaRa, http://www.takara-bio.com/) and sequenced using an Applied Biosystems 3730 sequencer (http://www.appliedbiosystemscom/).

For functional complementation, a 10.27 kb genomic DNA fragment, containing the entire OsCSLD4 coding region, the 3005 bp upstream sequences (including an approximately 2310 bp upstream sequence and a 690 bp fragment of the upstream gene) and the 3363 bp downstream sequence, was digested from BAC clone OSJNBa0027H05 (Arizona Genomics Institute, http://www.genome.arizona.edu) using XbaI and HpaI and inserted into the pCAMBIA 1300 vector (Cambia, http://www.cambia.org/) to generate construct pCSLD4. A control construct, pCSLD4T, was generated by digesting pCSLD4 with BamHI to remove the whole ORF of ND1. The two binary plasmids were introduced into Agrobacterium tumefaciens EHA105 by electroporation and transformed into the nd1 mutant plants. For RNAi analysis, a 600 bp fragment comprising 200 bp of the 5′ UTR and 400 bp of the ORF was amplified using primers ATTCACATGAGAGATCCTC and AGAAGTTGAGCACGTGGCCG. After sequencing confirmation, the PCR products were inserted into vector pCAMBIA23A in both the forward and reverse directions to generate the construct pCSLD4RNi. Then this binary plasmid was introduced into the wild-type plants (Nipponbare background) by Agrobacterium infection. Rice transformation was performed as described previously (Hiei et al., 1994).

Phylogenetic analysis included rice and Arabidopsis CSLD members, as well as those from other species, such as poplar (Populus tomentosa) and grapevine (Vitis vinifera). Multiple sequence alignments and bioinformatic analyses were performed as described by Li et al. (2003). A neighbor-joining tree was constructed using MEGA version 2 (Kumar et al., 2001), using the Poisson correction distance. The number of bootstrap replicates was 1000. Similar topology was obtained to estimate maximum parsimony, and the PROML program in the Phylip package (http://evolution.genetics.washington.edu/phylip/doc/proml.html) was used to estimate maximum likelihood.

Anatomical analyses

Fresh hand-cut sections (approximately 20 μm) of the 2nd internode of culms, and 2nd leaves and young stems of 14-day-old seedlings, were prepared and stained with toluidine blue to evaluate cell anatomical features under a light microscope (Leica, http://www.leica.com). For transmission electron microscopy, the 2nd internodes of culms and 4-day-old root tips were fixed in 2.5% v/v glutaraldehyde in PBS buffer (4 mm sodium phosphate, pH 7.2). After thorough rinsing in the same buffer, the samples were post-fixed in 2% w/v OsO4 for 0.5 h, and dehydrated through an ethanol gradient (30, 50, 70, 90, 100 and 100%, each for 30 min). Then, the samples were infiltrated and embedded using a Spurr kit (Sigma, http://www.sigmaaldrich.com/). Ultra-thin 80 nm sections were prepared using an Ultracut E ultramicrotome (Leica), and picked up on formvar-coated copper grids. The sections were stained with uranyl acetate and lead citrate, and viewed under a Hitachi H7500 transmission electron microscope (http://www.hitachi.co.jp).

Gene expression

Total RNAs were extracted from various tissues of wild-type rice plants using an RNeasy kit (Qiagen, http://www.qiagen.com/). The tissues included callus developed from seeds, 0.5–1 cm root tips, culms of the third internode, 2–3 mm stem shoots (stem apical meristem), fully extended leaves, leaf sheaths and 10 cm young panicles. Two micrograms of total RNA were digested with RNase-free DNase I (Invitrogen, http://www.invitrogen.com/), and used to synthesize cDNA using a reverse transcription kit (Promega, http://www.promega.com/). RT-PCR was performed using primers CAGTAGTGTAGTGGTGTCGA (forward) and CAATCGCTTAATCACACCTA (reverse) for OsCSLD4, and CAAGATGATCTGCCGCAAATG (forward) and primers TTTAACCAGTCCATGAACCCG (reverse) for ubiquitin, which was used as an internal control for normalization of each RNA sample.

To generate pOsCSLD4::GUS transgenic plants, the 2315 bp genomic fragment upstream of the OsCSLD4 putative translation start codon was amplified by PCR using primers CCATCCATCATTTCCTCTGT and CCACAAATCACCTCAAAACC. After sequencing confirmation, the DNA fragment was inserted into the vector pCAMBIA 1391Z between the PstI and BamHI sites in-frame with the GUS reporter gene. The construct was transformed into wild-type plants by the Agrobacterium-mediated transformation procedure. Transgenic plants were selected on hygromycin, and T1 transgenic plants were used for analysis of GUS activity. The GUS staining assay was performed as previously described (Scarpella et al., 2003).

Subcellular localization

EGFP (kindly provided by Dr Yasuo Niwa, School of Food and Nutritional Sciences, University of Shizuoka, Shizuoka City, Japan) (Niwa et al., 1999) was fused to the N- or C-terminus of OsCSLD4 and inserted between the CaMV 35S promoter and the nopaline synthase terminator in the PUC19 or pCAMBIA 2300 vectors (Cambia) to generate constructs for transformation of the rice protoplasts and N. benthamiana leaves, respectively. Co-transfection of rice protoplasts with GFP-tagged OsCSLD4 and mCherry-tagged HDEL (ER marker) or Man49 (Golgi marker) (Nelson et al., 2007) was performed as described by Zhou et al. (2009). Co-infiltration of N. benthamiana leaves was performed as described by Bernal et al. (2007), with some modifications. A. tumefaciens strain C58C1 harboring vectors bearing GFP–OsCSLD4 and mCherry-tagged HDEL (ER marker) or Man49 (Golgi marker) were cultured overnight, then sub-cultured in YEP medium containing 10 mm MES (pH 5.7) and 20 μm acetosyringone until the OD600 reached 0.8–1.5. After harvesting by centrifugation (2000 g for 5 min at room temperature), the bacterial cells were suspended in buffer containing 10 mm MES (pH 5.7), 10 mm MgCl2 and 200 μm acetosyringone, and allowed to stand at room temperature for at least 3 h. N. benthamiana leaves (3–4 weeks old) were then infiltrated with the bacterial suspensions using 1 ml syringes. Forty-eight hours after infiltration, the infiltrated leaves were observed by confocal laser scanning microscopy (Leica TCS SP5). For immunochemical staining of OsCSLD4 protein, we generated transgenic plants by expressing OsCSLD4–GUS under the control of its native promoter fragment (2315 bp) in the nd1 mutant background. Fresh hand-cut sections (approximately 20 μm) of young culms and 2-day-old roots of T1 plants were stained using BODIPY®-TR C5-ceramide as described by the manufacturer (Invitrogen) and fixed in 4% paraformaldehyde (Sigma) in PBS buffer (pH 7.2) for 30 min. After thorough rinsing in PBS buffer, the samples were blocked in 1% BSA (in PBS buffer, pH 7.2) for 2 h, and then incubated with primary antibodies (anti-GUS antibodies, Sigma) at 1:500 dilution (in 0.1% BSA in PBS buffer). The secondary antibodies, fluorescein isothiocyanate-conjugated anti-rabbit IgG or Cy3-conjugated anti-rabbit IgG (Sigma), were used at the same dilution. The treated samples were observed by confocal laser scanning microscopy (Leica TCS SP5).


Culms from developmentally matched wild-type and nd1 plants, as well as the age-matched roots, were collected and fixed in 4% paraformaldehyde (Sigma). After dehydration using an ethanol gradient (30, 50, 70, 90, 100 and 100%, each for 30 min), the samples were embedded in butyl methyl methacrylate (Sigma) and sectioned (2–3 μm) using an Ultracut E ultramicrotome (Leica). Immunochemical staining was performed as described previously (Zhou et al., 2009).

Chemical analysis of wall components

Cell-wall residues of culms and root tips from developmentally matched wild-type and mutant plants were prepared into alcohol-insoluble residue (AIR) as previously described (Harholt et al., 2006). De-starching was performed by treating AIR with pullulanase M3 (0.5 U mg−1, Megazyme, http://www.megazyme.com) and α-amylase (0.75 U mg−1, Sigma) in 0.1 m NaOAc buffer (pH 5.0) overnight. For methanolysis, 0.7 mg of de-starched AIR was heated in 1 m methanolic HCl at 80°C overnight. The mixture was evaporated and treated with TRI-SIL reagent (Sigma). The silylated sugar was extracted in hexane and analyzed by GC-MS (Agilent,http://www.agilent.com.cn) 6890N and 5973I) (York et al., 1985). The crystalline cellulose content was measured using a modified method as described by Updegraff (1969). In detail, after the above treatment, the pellets were hydrolyzed with Updegraff reagent at 100°C for 30 min. After cooling, the pellets were washed with acetone and hydrolyzed with 72% sulfuric acid. Then an anthrone assay was performed to quantify cellulose content. For fractional analysis, the wall residues of culms and root tips were sequentially fractionated using 0.5% hot ammonium oxalate solution and 4 N KOH. Soluble components were neutralized, dialyzed against water, and freeze-dried. The insoluble components were collected and thoroughly rinsed in water. Then the samples were hydrolyzed in 2 m trifluoroacetic acid (TFA) at 121°C for 90 min and reduced with sodium borohydride in 1 m ammonium hydroxide. The alditol acetates produced were analyzed by GC-MS. The recovery rates for AIR, de-starched AIR and ammonium oxalate and 4 N KOH fractions are shown in Table S3. Glycosyl linkage analysis was performed using a modification of the Hakamori method (1964): 0.5 mg of de-starched AIR was methylated by incubating the residues in dimethyl sulfoxide dissolved in iodomethane and NaOH for 16 h (Ciucanu and Kerek, 1984). Then the samples were hydrolyzed in 2 m TFA and reduced using sodium borodeuteride in 1 m ammonium hydroxide. The partially methylated alditol acetate was analyzed by GC-MS, as previously described (York et al., 1985).


We thank Yinhong Zhang (Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing, China) for confocal microscope examination, Hongjing Hao (Institute of Atomic Energy, Chinese Academy of Agricultural Sciences, Beijing, China) for transmission electron microscopy, and Jianhua Wei (Beijing Academy of Agriculture and Forestry Sciences) for GC-MS analysis. Karen Bird (US Department of Energy Plant Research Laboratory) is thanked for text editing. This work was supported by grants from the National Natural Science Foundation of China (90717117), the Ministry of Sciences and Technology of China (2005CB120805 and 2006AA10A101), and the Knowledge Innovation Program of the Chinese Academy of Sciences (KSCX2-YW-G-033).