Arabidopsis irregular xylem8 and irregular xylem9: Implications for the Complexity of Glucuronoxylan Biosynthesis

Abstract

Mutations of Arabidopsis thaliana IRREGULAR XYLEM8 (IRX8) and IRX9 were previously shown to cause a collapsed xylem phenotype and decreases in xylose and cellulose in cell walls. In this study, we characterized IRX8 and IRX9 and performed chemical and structural analyses of glucuronoxylan (GX) from irx8 and irx9 plants. IRX8 and IRX9 are expressed specifically in cells undergoing secondary wall thickening, and their encoded proteins are targeted to the Golgi, where GX is synthesized. ¹H-NMR spectroscopy showed that the reducing end of Arabidopsis GX contains the glycosyl sequence 4-β-d-Xylp-(1→4)-β-d-Xylp-(1→3)-α-l-Rhap-(1→2)-α-d-GalpA-(1→4)-d-Xylp, which was previously identified in birch (Betula verrucosa) and spruce (Picea abies) GX. This indicates that the reducing end structure of GXs is evolutionarily conserved in woody and herbaceous plants. This sequence is more abundant in irx9 GX than in the wild type, whereas irx8 and fragile fiber8 (fra8) plants are nearly devoid of it. The number of GX chains increased and the GX chain length decreased in irx9 plants. Conversely, the number of GX chains decreased and the chain length heterodispersity increased in irx8 and fra8 plants. Our results suggest that IRX9 is required for normal GX elongation and indicate roles for IRX8 and FRA8 in the synthesis of the glycosyl sequence at the GX reducing end.

INTRODUCTION

Glucuronoxylans (GXs) together with cellulose and lignin are the three major components of secondary cell walls in woody plant tissues, which constitute the bulk of terrestrial biomass. The biosynthesis of cellulose and lignin has been studied intensively (Boerjan et al., 2003; Scheible and Pauly, 2004; Lerouxel et al., 2006; Somerville, 2006), whereas the mechanisms of GX biosynthesis are poorly understood. GX is composed of a linear backbone of β-(1-4)-linked d-xylosyl (Xyl) residues, some of which bear a single α-d-glucuronic acid (GlcA) or 4-O-methyl-α-d-glucuronic acid (MeGlcA) residue at O2 (Figure 1). The Xyl residues can also be substituted with arabinosyl and acetyl residues (Ebringerová and Heinze, 2000). Early work established that GX isolated from birch (Betula verrucosa) and spruce (Picea abies) wood contains the glycosyl sequence 4-β-d-Xylp-(1→4)-β-d-Xylp-(1→3)-α-l-Rhap-(1→2)-α-d-GalpA-(1→4)-d-Xylp at the reducing end (Shimizu et al., 1976; Johansson and Samuelson, 1977; Andersson et al., 1983).

Figure 1.

Structures of the Xylo-Oligosaccharides Generated by Endoxylanase Treatment of Arabidopsis GX.

(A) 1,4-Linked β-d-xylo-oligosaccharides.

(B) 1,4-Linked β-d-xylo-oligosaccharides partially substituted at O2 with glucuronic acid.

(C) 1,4-Linked β-d-xylo-oligosaccharides partially substituted at O2 with 4-O-methyl glucuronic acid.

(D) The glycosyl sequence (1) at the reducing end of Arabidopsis GXs.

(E) The glycosyl sequence (2) of the tetraglycosyl-xylitol.

The xylo-oligosaccharides are generated by treating the 1 and 4 n KOH–soluble materials with endoxylanase. The xylitol of glycosyl sequence 2 is formed from the reducing xylose of glycosyl sequence 1 when glucuronoxylans are solubilized from cell walls using alkali-containing NaBH₄. NaBH₄ converts the reducing xylose to xylitol. Glycosyl sequence 2 was isolated from the endoxylanase digests by reverse-phase HPLC.

Based on GX structure, it is likely that a number of glycosyltransferases (GTs) are required for the initiation, elongation, and termination of the xylan backbone, together with enzymes that add and modify side chains. Xylosyltransferase and glucuronyltransferase activities have been detected in numerous plants (Dalessandro and Northcote, 1981a, 1981b; Waldron and Brett, 1983; Baydoun et al., 1989; Suzuki et al., 1991; Porchia and Scheller, 2000; Kuroyama and Tsumuraya, 2001; Gregory et al., 2002). However, none of the genes encoding these enzymes has been identified, nor have any of the enzymes been purified to homogeneity and biochemically characterized.

Recent genomic analysis of wood formation in poplar (Populus species) revealed 25 putative GTs whose expression is associated with secondary wall synthesis (Aspeborg et al., 2005). Arabidopsis thaliana homologs of three poplar wood–associated GTs, FRAGILE FIBER8 (FRA8), IRREGULAR XYLEM8 (IRX8), and IRX9, have been shown to be required for normal vessel morphology and wall thickness and for normal amounts of xylose and cellulose in cell walls (Brown et al., 2005; Persson et al., 2005; Zhong et al., 2005). The FRA8 gene, which encodes a putative GT in family GT47 (Coutinho et al., 2003), is specifically expressed in cells undergoing secondary wall thickening. Plants carrying mutations in this gene have reduced amounts of wall GX and a decreased ratio of GlcA to MeGlcA residues in the GX (Zhong et al., 2005). Expression of the poplar (Populus alba × tremula) GT47C gene in fra8 plants rescues the defects in secondary wall thickness and GX synthesis, suggesting that GT47C is a functional homolog of FRA8 (Zhou et al., 2006).

The IRX8 gene encodes a putative GT in family GT8 (Brown et al., 2005; Persson et al., 2005). The expression of several genes encoding GTs in family GT8 (Coutinho et al., 2003) is associated with wood formation in poplar (Aspeborg et al., 2005). Homologs of these wood-associated GTs are present in Arabidopsis, and several of them, including IRX8, At4g33330, At3g18660, and At1g19300, are highly expressed in Arabidopsis stems (Brown et al., 2005; Persson et al., 2005; Ye et al., 2006). Several members of the GT8 family catalyze the transfer of uronic acids to glycans. For example, three Arabidopsis GT8 proteins, QUASIMODO1 (QUA1) (Bouton et al., 2002), PARVUS (Lao et al., 2003), and GALACTURONOSYL TRANSFERASE1 (GAUT1) (Sterling et al., 2006), have been identified and are believed to have a role in pectin biosynthesis. Of these three, only GAUT1 has been biochemically characterized and shown to have galacturonosyltransferase activity (Sterling et al., 2006).

Mutation of the IRX9 gene, which encodes a putative GT in family GT43 (Coutinho et al., 2003), was shown to result in plants with decreased amounts of wall GX, suggesting that this gene is required for GX synthesis (Bauer et al., 2006). The poplar (Populus tremula × tremuloides) GT43A and Ptt GT43B genes, which are homologs of IRX9, have been shown to be highly expressed during wood formation (Aspeborg et al., 2005). In addition, a cotton (Gossypium hirsutum) gene, which resides in the same phylogenetic subgroup as Ptt GT43A, Ptt GT43B, and IRX9, is highly expressed during cotton fiber development (Wu and Liu, 2005). Together, these findings suggest that family 43 GTs have an important role in secondary wall synthesis.

In this report, we show that IRX8 and IRX9 are specifically expressed in fibers and vessels and that their encoded proteins are localized in the Golgi. We also show that glycosyl sequence 1 (Figure 1) is located at the reducing end of Arabidopsis GX. We further demonstrate that the irx8, irx9, and fra8 mutations result in changes in the abundance of this reducing end sequence and lead to alterations in the number and length of GX chains compared with wild-type GX. Our results provide evidence for the possible involvement of IRX8 and FRA8 in the synthesis of the reducing end sequence of GXs and suggest that IRX9 has an essential role in the elongation of the xylan backbone.

RESULTS

IRX8 and IRX9 Are Specifically Expressed in Fibers and Vessels

The interfascicular fibers of Arabidopsis inflorescence stems deposit a massive amount of secondary walls and provide a model system for studying the mechanisms of secondary wall synthesis (Zhong et al., 2001). During a search for xylan biosynthetic genes that are highly expressed in interfascicular fibers, we found that two GT genes, At5g54690 (IRX8) and At2g37090 (IRX9), exhibited an organ expression pattern similar to that of FRA8, based on expression data from the AtGenExpress project (Schmid et al., 2005; http://www.weigelworld.org/resources/microarray/AtGenExpress/) and RT-PCR analysis (data not shown). Further expression analysis showed that IRX8 and IRX9 were expressed in developing interfascicular fibers but not in parenchymatous pith cells (Figure 2A).

Figure 2.

Expression Patterns of the IRX8 and IRX9 Genes in Arabidopsis Stems and Roots.

Cross sections of stems were hybridized with digoxigenin-labeled antisense RNA probes of IRX8 and IRX9. The hybridization signals were detected using alkaline phosphatase–conjugated antibodies. Transgenic plants expressing the GUS reporter gene fused with the IRX8 and IRX9 genes were examined for GUS activity. if, interfascicular fiber; sx, secondary xylem; xy, xylem. Bar in (B) = 145 μm for (B) to (L).

(A) RT-PCR analysis of laser-microdissected cells from stems showing the expression of IRX8 and IRX9 together with several other secondary wall biosynthetic genes in fiber cells but not in pith cells. The expression of a ubiquitin gene was used as an internal control.

(B) to (D) In situ hybridization of stem sections showing the expression of IRX8 (B) and IRX9 (C) genes in interfascicular fibers and developing xylem cells. A stem section hybridized with the IRX9 sense probe is shown as a control (D).

(E) to (H) Cross sections of stems ([E] to [G]) and roots (H) of transgenic IRX8∷GUS plants showing intense GUS staining in interfascicular fibers and xylem cells undergoing secondary wall synthesis. The stem sections were from a young elongating internode (E), an internode near cessation of elongation (F), and a nonelongating internode (G).

(I) to (L) Cross sections of stems ([I] to [K]) and roots (L) of transgenic IRX9∷GUS plants showing strong GUS staining in interfascicular fibers and xylem cells undergoing secondary wall synthesis. The stem sections were from a young elongating internode (I), an internode near cessation of elongation (J), and a nonelongating internode (K).

Mutation of the IRX8 or IRX9 gene causes a strong dwarf phenotype (Brown et al., 2005); thus, it is important to know whether these genes function specifically in secondary wall synthesis by examining when and where they are expressed (Brown et al., 2005). To investigate their expression patterns, we fused the wild-type IRX8 and IRX9 genes with the β-glucuronidase (GUS) reporter gene and transformed the constructs into wild-type Arabidopsis plants. Analysis of GUS activity in transgenic plants revealed that IRX8 and IRX9 were specifically expressed in cells undergoing secondary wall thickening, including interfascicular fibers and primary and secondary xylem in stems and developing secondary xylem in roots (Figures 2E to 2L). These expression patterns were confirmed by in situ mRNA hybridization (Figures 2B to 2D). Our results demonstrate that the expression of IRX8 and IRX9 is specifically associated with secondary wall thickening in fibers and vessels.

IRX8 and IRX9 Are Targeted to the Golgi

The irx8 and irx9 mutations result in a reduction in both xylose and cellulose in cell walls (Brown et al., 2005; Bauer et al., 2006). Thus, we investigated whether the IRX8 and IRX9 proteins are located in the Golgi, where GX is synthesized, or at the plasma membrane, where cellulose is synthesized. Transgenic Arabidopsis plants expressing green fluorescent protein (GFP)–tagged IRX8 or IRX9 exhibited punctate fluorescence signals in the cytoplasm of root epidermal cells (Figures 3A and 3B), indicating that IRX8 and IRX9 are located in subcellular organelles. Cotransfection of yellow fluorescent protein (YFP)–tagged IRX8 or IRX9 with cyan fluorescent protein (CFP)–tagged Golgi-localized MUR4 in carrot (Daucus carota) protoplasts revealed their colocalization (Figures 3E to 3L), suggesting that IRX8 and IRX9 are located in the Golgi. Sequence analysis using the TMHMM2.0 program (http://www.cbs.dtu.dk/services/tmhmm-2.0/) predicted that IRX8 and IRX9 are type II membrane proteins with a single transmembrane helix at the N terminus (data not shown). These results are consistent with a role of IRX8 and IRX9 in GX biosynthesis, which is known to occur in the Golgi (Gregory et al., 2002).

Figure 3.

Subcellular Localization of Fluorescent Protein–Tagged IRX8 and IRX9 Proteins.

Fluorescent protein–tagged IRX8 and IRX9 were expressed in Arabidopsis plants and carrot protoplasts, and their subcellular locations were examined with a laser confocal microscope. Bar in (A) = 11 μm for (A) to (C); bar in (D) = 21 μm for (D) to (L).

(A) and (B) Confocal imaging of Arabidopsis root epidermal cells expressing GFP-tagged IRX8 (A) and GFP-tagged IRX9 (B) showing punctate fluorescence signals.

(C) Confocal imaging of Arabidopsis root epidermal cells expressing GFP alone.

(D) Confocal imaging of a carrot cell expressing YFP alone.

(E) to (H) Differential interference contrast image (E) of a carrot cell expressing IRX8-YFP and MUR4-CFP and the corresponding IRX8-YFP signals (F), Golgi-localized MUR4-CFP signals (G), and the merged image of IRX8-YFP and MUR4-CFP signals (H).

(I) to (L) Differential interference contrast image (I) of a carrot cell expressing IRX9-YFP and MUR4-CFP and the corresponding IRX9-YFP signals (J), MUR4-CFP signals (K), and the merged image of IRX9-YFP and MUR4-CFP signals (L).

irx8 Results in a Reduction in LM10 Antibody Labeling of Secondary Walls of Fibers and Vessels

The irx8 and irx9 mutations were shown to result in plants with abnormal vessel morphology (Brown et al., 2005; Persson et al., 2005). To extend these findings, we examined the thickness of secondary walls in two independent T-DNA insertion lines for the IRX8 (irx8-1, SALK_008642; irx8-2, SALK_014026) and IRX9 (irx9-1, SALK_058238; irx9-2, SALK_057033) genes. We found that the wall thickness of interfascicular fibers (Figures 4A to 4F) and vessels (data not shown) in the stems of irx8 and irx9 mutants was decreased by ∼60% compared with the wild type, which is most likely the cause of their reduced stem strength (Figure 4K). In addition, the wall thickness of xylary fibers and vessels in the secondary xylem of roots was also reduced (Figures 4G to 4I). The reduced secondary wall thickness phenotype caused by mutations of the IRX8 and IRX9 genes is consistent with their expression patterns. The defects in secondary wall thickness, stem strength, and plant growth in both mutants were rescued by expression of their corresponding wild-type genes (Figures 4J and 4K), confirming that these phenotypes were caused directly by mutations of the IRX8 and IRX9 genes.

Figure 4.

Reduced Secondary Wall Thickness in Fibers of irx8 and irx9 Plants.

Roots and bottom internodes of inflorescence stems from 10-week-old plants were sectioned for examination of fibers and vessels. co, cortex; if, interfascicular fiber; ve, vessel; xf, xylary fiber; xy, xylem. Bar in (A) = 120 μm for (A) to (C); bar in (D) = 2.8 μm for (D) to (F); bar in (G) = 67 μm for (G) to (I).

(A) to (C) Cross sections of interfascicular regions of stems showing thin-walled fibers in irx8 (B) and irx9 (C) compared with the wild type (A).

(D) to (F) Transmission electron micrographs of walls of interfascicular fibers in the wild type (D), irx8 (E), and irx9 (F).

(G) to (I) Cross sections of secondary xylem regions of roots showing thin xylary fibers and deformed vessels in irx8 (H) and irx9 (I) compared with the wild type (G).

(J) Growth defects in irx8 and irx9 plants were rescued by expression of the corresponding wild-type genes.

(K) Breaking force measurements showing that expression of the wild-type IRX8 and IRX9 genes restored the stem strength of irx8 and irx9, respectively, to a level comparable with that of the wild type. Each bar represents the breaking force of the inflorescence stem of an individual plant.

To investigate whether the irx8 mutation affects GX deposition in secondary walls, we performed immunolocalization of GX using monoclonal antibodies LM10 and LM11. LM10 has been reported to bind to 4-O-methylglucuronoxylan but not to arabinoxylan and glucuronoarabinoxylan, whereas LM11 interacts with both 4-O-methylglucuronoxylan and arabinoxylan (McCartney et al., 2005). Only LM10 labeling is shown because similar results were observed with the LM11 antibody. Immunolabeling of stem and root sections revealed strong fluorescence signals in the walls of interfascicular fibers and xylem cells in wild-type Arabidopsis (Figures 5A and 5D), but only weak signals were detected in the corresponding tissues of irx8 plants (Figures 5B and 5E). Transmission electron microscopy demonstrated that the density of immunogold labeling was reduced drastically in irx8 compared with the wild type (Figures 5G and 5H). By contrast, the fluorescence labeling of walls in irx9 plants remained strong (Figures 4C and 4F). However, the overall intensity was reduced, most likely as a result of the reduced wall thickness. The density of the immunogold labeling of walls in irx9 plants (Figure 5I) and wild-type plants (Figure 5G) was comparable. Our data suggest that even though both the irx8 and irx9 mutations caused a nearly 60% reduction in secondary wall thickness, these mutations have different effects on GX deposition in Arabidopsis secondary walls.

Figure 5.

Immunolocalization of Xylan in the Stems and Roots of irx8 and irx9 Mutants.

Stem and root sections from 10-week-old Arabidopsis plants were used for immunolocalization of xylan with the monoclonal antibody LM10 that was generated against plant cell wall (1,4)-β-d-xylan. Xylan signals were detected with fluorescein isothiocyanate–conjugated secondary antibodies and visualized with a laser confocal microscope, or detected with gold-conjugated secondary antibodies and visualized with a transmission electron microscope. if, interfascicular fiber; sx, secondary xylem; xy, xylem. Bar in (A) = 154 μm for (A) to (F); bar in (G) = 0.52 μm for (G) to (I).

(A) to (C) Xylan immunofluorescence signals in stem sections of the wild type (A), irx8 (B), and irx9 (C).

(D) to (F) Xylan immunofluorescence signals in root sections of the wild type (D), irx8 (E), and irx9 (F).

(G) to (I) Immunogold labeling of xylan in fiber walls of the wild type (G), irx8 (H), and irx9 (I).

The GX Content of Cell Walls of Inflorescence Stems Is Reduced in irx8 and irx9 Plants

Previous studies have shown that the irx8 and irx9 mutations lead to a reduction in cell wall xylose, which has led to the suggestion that GX synthesis is affected in these mutants (Brown et al., 2005; Persson et al., 2005; Bauer et al., 2006). To confirm and extend this notion, we isolated and characterized GXs from the cell walls of inflorescence stems of wild-type, irx8, and irx9 plants. We also isolated walls from fra8 stems so that we could compare the affects on GX structure of each of the three mutations. We found that most of the xylose-rich polysaccharides are solubilized by treating the walls with 1 and 4 n KOH. At least 80% of this material remained soluble after neutralization. The irx8, irx9, and fra8 mutations led to 57, 70, and 67% reductions, respectively, in the xylose content of these extracts (Table 1). We established that these changes are attributable to a decrease in GX content by showing similar reductions in the amounts of xylo-oligosaccharides that are generated by treating the KOH extracts with endoxylanase (Table 2). NMR analysis indicated that the high-molecular-weight material remaining after endoxylanase treatment contained little if any GX and that xyloglucan and pectin are the predominant polysaccharide components of this fraction. Together, these results indicate that the recovered xylo-oligosaccharides represent the majority of the KOH-solubilized GX.

Table 1.

Glycosyl Residue Compositions of the 1 and 4 n KOH–Soluble Materials from Wild-Type and Mutant Arabidopsis Stems

Cell Wall Fraction	Rha	Fucose	Ara	Xyl	Man	Gal	Glc
1 n KOH–soluble
Wild type	2	<1	2	66	1	3	5
fra8	1	1	2	13	3	4	10
irx8	1	1	4	12	2	6	10
irx9	2	<1	3	31	2	4	12
4 n KOH–soluble
Wild type	1	1	2	26	3	5	8
fra8	1	1	2	17	7	4	11
irx8	1	1	3	16	7	6	12
irx9	1	1	3	9	6	6	10

Data shown are means (mg sugar/g AIRs) of two duplications of two independent analyses. AIRs were treated sequentially with endopolygalacturonase and pectin methyl esterase, xyloglucan-specific endoglucanse, and 1 and 4 n KOH. The 1 and 4 n KOH–soluble materials were hydrolyzed with trifluoroacetic acid, and the released monosaccharides were converted to their alditol acetate derivatives. Gas chromatography–mass spectrometry was used for the identification and quantitation of the monosaccharides. The variance is not shown but was <10%.

Table 2.

Amounts of Wild-Type and Mutant Arabidopsis GXs

Plant	1 n KOH (mg/g AIRs)	4 n KOH (mg/g AIRs)	Ratio 1 n:4 n	Total (mg/g AIRs)
Wild type	81	43	1.9	124
fra8	27	25	1.1	52
irx8	18	13	1.4	31
irx9	52	4	13.0	56

The GX contents of the alkali-soluble fractions were obtained by treating the fractions with endoxylanase and measuring the amounts of xylo-oligosaccharide formed by size-exclusion chromatography with refractive index detection.

irx8, irx9, and fra8 Affect the Abundance of the Sequence 4-β-d-Xylp-(1→4)-β-d-Xylp-(1→3)-α-l-Rhap-(1→2)-α-d-GalpA-(1→4)-d-Xylp That Is Present at the Reducing End of Arabidopsis GX

To further investigate changes in GX structure and establish GX chemotypes for each mutant, we used size-exclusion chromatography (SEC) to separate the neutral and acidic xylo-oligosaccharides generated by endoxylanase treatment of the 1 n KOH–soluble material from wild-type, irx8, and irx9 plants (data not shown). All six fractions were analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). The MALDI-TOF spectra of the fractions contained [M + Na]⁺ ions, indicating the presence of the expected products (Figures 1A to 1C): Xyl_3-4 (m/z 437 and 569) in the neutral fractions and Xyl_4-6GlcA (m/z 745, 877, and 1009) and Xyl_4-6MeGlcA (m/z 759, 891, and 1023) in the acidic fractions. The ¹H-NMR spectra of the acidic fractions (Figure 6, Table 3) showed that in the wild type, irx8, and irx9, the GlcA:MeGlcA ratios were 0.76, 0.06, and 0.03, respectively. A similar decrease in the proportion of GlcA was observed in fra8 GX (Zhong et al., 2005), suggesting that all three mutations lead to an alteration in GX structure.

Figure 6.

Anomeric Region of the 600-MHz ¹H-NMR Spectra of the Xylo-Oligosaccharides Generated by Endoxylanase Treatment of the Material Solubilized by 1 n KOH–Containing NaBH₄.

(A) Wild-type xylo-oligosaccharides.

(B) fra8 xylo-oligosaccharides.

(C) irx8 xylo-oligosaccharides.

(D) irx9 xylo-oligosaccharides.

(E) Tetraglycosyl-xylitol (glycosyl sequence 2; see Figure 1) isolated from the products generated by endoxylanase treatment of irx9 GX.

The anomeric region shown contains diagnostic well-resolved signals for each residue in the complex. These include the following: U1, α-d-GlcpA H1; M1, 4-O-methyl α-d-GlcpA H1; G1, α-d-GalpA H1; R1, α-l-Rhap H1; X1, 3-linked β-d-Xylp H1; G4, α-d-GalpA H4; R2, α-l-Rhap H2. α and β indicate the anomeric signals of α and β reducing xylose H1, respectively.

Table 3.

¹H-NMR Assignments of the Acidic Xylo-Oligosaccharides Generated by Endoxylanase Treatment of the 1 and 4 n KOH–Soluble Materials

Residue	H1	H2	H3	H4	H5	H5ax
α-Xylp (reducing)	5.183	3.543	3.807
β-Xylp (reducing)	4.583	3.249	3.546	3.779	4.055	3.377
β-Xylp (internal)	4.474	3.278	3.573	3.787	4.146	3.436
β-Xylp (internal)	4.478	3.297	3.548	3.791	4.108	3.373
β-Xylp (terminal)	4.459	3.256	3.425	3.621	3.971	3.299
β-Xylp (4Me-α-GlcA)	4.624	3.438	3.618	3.803	4.103	3.381
β-Xylp (α-GlcA)	4.637	3.482	3.633	3.808	4.110	3.392
4-O-Methyl-α-GlcpA	5.285	3.572	3.758	3.218	4.327
α-GlcpA	5.302	3.551	3.739	3.472	4.360

Chemical shifts are reported in ppm relative to internal acetone, δ 2.225. β-Xyl (4Me-α-GlcA) is a β-linked xylosyl residue that is substituted at O2 with 4-O-methyl GlcA (see Figure 1 for details). β-Xyl (α-GlcA) is a β-linked xylosyl residue that is substituted at O2 with GlcA (see Figure 1 for details). H4 and H5 of the reducing α xylose were not assigned.

The ¹H-NMR spectrum of the irx9 acidic xylo-oligosaccharides contained several unexpected signals that were much stronger than in the wild-type spectrum and that were nearly absent in the irx8 and fra8 spectra (Figures 6A to 6D). To investigate the origin of these signals, we separated the acidic oligosaccharides by reverse-phase HPLC. MALDI-TOF MS of one of the fractions was enriched in an ion at m/z 761, consistent with the presence of an oligoglycosyl alditol. Such alditols are generated when the residue at the reducing end of the polysaccharide reacts with NaBH₄, which is added to the alkaline extraction medium to avoid degradation by base-catalyzed peeling reactions (Johansson and Samuelson, 1977). To improve the separation of the xylo-oligosaccharides by reverse-phase HPLC, the reducing oligosaccharides were converted to their 2-aminobenzamide derivatives. This resulted in the separation of the oligosaccharide alditol (m/z 761), which binds weakly to the column, from the 2-aminobenzamide derivatives, which bind strongly. The ¹H-NMR spectrum of the purified oligoglycosyl alditol was dominated by the unexpected signals initially detected in the spectrum of the acidic xylo-oligosaccharide fraction from irx9. These signals were assigned to protons of α-GalpA, α-Rhap, and β-Xylp residues (Figure 6E, Table 4) on the basis of their scalar coupling patterns and chemical shifts in the gCOSY and TOCSY spectra. The linkages between each glycosyl residue were deduced from the interglycosidic scalar couplings observed in the high-resolution gCOSY spectrum (Jia et al., 2005) and from the interglycosidic dipolar couplings observed in NOESY and ROESY spectra (Table 4). Together, the NMR and MALDI-TOF MS analyses provide sufficient evidence to unambiguously assign the linkage and sequence of residues in the oligoglycosyl alditol as β-d-Xylp-(1→4)-β-d-Xylp-(1→3)-α-l-Rhap-(1→2)-α-d-GalpA-(1→4)-d-xylitol (sequence 2; Figure 1E). Previous studies have shown that β-d-Xylp-(1→4)-β-d-Xylp-(1→3)-α-l-Rhap-(1→2)-α-d-GalpA-(1→4)-d-Xyl (sequence 1; Figure 1D) is present at the reducing end of birch and spruce GXs (Shimizu et al., 1976; Johansson and Samuelson, 1977; Andersson et al., 1983). The reducing xylose is converted to xylitol when the GX is solubilized with KOH containing NaBH₄. This led us to believe that glycosyl sequence 1 is present at the reducing end of Arabidopsis GX and that the accumulation of this glycosyl sequence is affected in irx8, irx9, and fra8 plants.

Table 4.

¹H-NMR Assignments of Glycosyl Sequence 2

Residue	H1	H2	H3	H4	H5	H5ax	H6
β-Xylp (terminal)	4.464 (7.9)	3.260	3.431	3.628	3.973	3.313
β-Xylp (internal)	4.678 (7.9)	3.392	3.584	3.790	4.088	3.407
α-Rhap	5.080 (1.5)	4.241	3.946	3.622	3.796		1.332
α-GalpA	5.219 (3.8)	3.959	4.103	4.306	4.448
Interglycosidic cross-peaks
H1 β-Xylp to H4 β-Xylp			4.464	3.790	ROESY
H1 β-Xylp to H3 α-Rhap			4.678	3.946	ROESY, NOESY, and gCOSY
H1 α-Rhap to H2 α-GalpA			5.08	3.959	ROESY, NOESY, and gCOSY
H1 α-GalpA to xylitol			5.219	3.870	ROESY

Chemical shifts are reported in ppm relative to internal acetone, δ 2.225. Scalar coupling constants are given in parentheses. Interglycosidic cross-peaks allow the linkage between two glycosyl residues to be determined and were obtained from the high-resolution gCOSY, ROESY, and NOESY NMR spectra.

To confirm that glycosyl sequence 1 is indeed present at the reducing end of Arabidopsis GX, we solubilized GXs from irx9 cell walls in the absence and presence of NaBH₄. The ¹H-NMR spectra of both GXs contained resonances of the internal α-GalpA, α-Rhap, and β-Xylp residues of glycosyl sequence 1 (Figures 7A and 7B). Signals for H-1 of the α and β forms of the reducing xylose were present only in the GX extracted in the absence of NaBH₄. The presence of this reducing sugar resulted in discernible chemical shift and anomerization effects (Sakaguchi et al., 2001) on several of the resonances of the α-GalpA, α-Rhap, and 3-linked β-Xylp residues (Figure 7A). In particular, H-5 of the α-d-GalpA was split into two fully resolved signals (a and b in Figure 7A) whose intensities are comparable to those of the H-1 signals of the α and β forms of the reducing xylose (α and β in Figure 7A). This analysis verified that the 3-linked β-Xylp, α-Rhap, and α-GalpA residues are in close proximity to the xylose at the reducing end of the GX. We conclude that glycosyl sequence 1 is located at the reducing ends of GXs synthesized by Arabidopsis and that the abundance of this sequence is altered in irx8, irx9, and fra8 GXs.

Figure 7.

Anomeric Region of the 600-MHz ¹H-NMR Spectra of Alkali-Soluble GX from the Stems of irx9 Plants.

(A) ¹H-NMR spectrum of GX extracted from irx9 cell walls with 0.25 n KOH in the absence of NaBH_4.

(B) ¹H-NMR spectrum of GX extracted from irx9 cell walls with 1.0 n KOH in the presence NaBH₄.

(C) ¹H-NMR spectrum of the purified glycosyl sequence 2 (see Figure 1).

H1 resonances of reducing xylose residues are labeled with the anomeric configuration (α or β) of the residue. Changes in the positions of the resonance of specific nonreducing residues are indicated by double-headed arrows. G1, α-d-GalpA H1; R1, α-l-Rhap H1; X1, 3-linked β-d-Xylp H1; G5, α-d-GalpA H5; G4, α-d-GalpA H4; R2, α-l-Rhap H2. All of these resonances exhibit distinct anomerization effects in the spectrum of the GX solubilized without NaBH₄, especially α-d-GalpA H5, which is split into two fully resolved signals labeled a and b, corresponding to the α and β forms, respectively, of the reducing xylose residue.

The Molecular Weight Distribution of GX Is Altered in irx8, irx9, and fra8 Plants

Our data have clearly shown that the irx8, irx9, and fra8 mutations lead to a discernible reduction in the amounts of cell wall GX and to alterations in GX structure. We then decided to determine whether the size distribution of the GX was also affected by these mutations. Toward this goal, we analyzed the 1 and 4 n KOH–soluble materials by SEC. These SEC profiles correspond to the sum of the profiles of all of the components present in the extract. Selectively removing one component and then subtracting the resulting SEC profile from the total allows a profile corresponding to that of the component that was removed to be obtained. Thus, we treated the alkali extracts with a homogenous endoxylanase and analyzed the products by SEC. The profile of the endoxylanase-treated material was then subtracted from the untreated profile (Figure 8). The resulting profile represents the material that is susceptible to endoxylanase and thus corresponds to the size distribution of the GX. We found an increase in the heterodispersity of the GXs isolated from irx8 (Figures 8A and 8B) and fra8 (Figures 8C and 8D) plants. Moreover, some of the GXs from these plants have a higher molecular weight than do wild-type GX (Figures 8G and 8H). By contrast, the irx9 mutation resulted in a decrease in the molecular weight of the GX (Figures 8E and 8F). This decrease in molecular weight may account for the fact that ∼90% of the irx9 GX was solubilized by 1 n KOH, whereas the same treatment solubilized only 65% of the GX in wild-type plants (Table 2).

Figure 8.

SEC Profiles Showing the Size Distribution of GX from irx8, irx9, fra8, and Wild-Type Plants.

The 1 and 4 n KOH–soluble materials from wild-type and mutant plants were treated separately with endoxylanase and analyzed by SEC. Each SEC profile was then subtracted from its corresponding SEC profile of the untreated material. This resulting profile corresponds to the size distribution profile of the GX.

(A) and (B) The reconstructed profiles of irx8 GX.

(C) and (D) The reconstructed profiles of fra8 GX.

(E) and (F) The reconstructed profiles of irx9 GX.

(G) and (H) The reconstructed profiles of wild-type GX.

(A), (C), (E), and (G) show data for the 1 n KOH–soluble materials. (B), (D), (F), and (H) show data for the 4 n KOH–soluble materials. The column excluded volume (Vo) and the elution positions of dextran molecular mass markers of 500, 70, 40, and 10 kD are shown.

The Increased Size Heterogeneity of irx8 and fra8 GX Is Correlated with a Decrease in the Proportion of GX Molecules with Glycosyl Sequence 1 at Their Reducing Ends

Our data had provided evidence that wild-type, irx8, irx9, and fra8 plants synthesize GXs with different size distributions (Figure 8). We initially assumed that all GX molecules solubilized by alkali-containing NaBH₄ would contain glycosyl sequence 2 (Figure 1E) at their reducing ends. An estimate of the average degree of polymerization of these GXs could then be obtained by integrating selected resonances of reducing oligosaccharides and glycosyl sequence 2 in the NMR spectra of the endoxylanase-generated xylo-oligosaccharides. However, during this analysis, we discovered that some of these oligomers had Xylp, rather than GalpA, attached to xylitol (Table 5). Our data suggest that at least 80% of the GXs from irx8 and fra8 plants had 4-β-d-Xylp-(1→4)-d-Xylp at their reducing ends (Table 6). By contrast, 98% of the GXs from irx9 plants and 84% of wild-type GXs contain glycosyl sequence 1 at their reducing ends (Table 6). Thus, the increased size heterogeneity of irx8 and fra8 GXs is correlated with a decrease in the proportion of GX molecules that have glycosyl sequence 1 at their reducing ends. Furthermore, the fra8 and irx8 mutations result in a decrease in the number of GX molecules synthesized, whereas the primary effect of the irx9 mutation is a decrease in the GX molecular weight (Figure 8, Table 6).

Table 5.

¹H-NMR Assignments of the Xylp Residues Linked to Xylitol in Oligoglycosyl-Xylitols Containing Xyl, GlcA, or 4-O-Methyl-GlcA

	β-Xylosyl Residues Linked to Xylitol
Oligosaccharide	H1	H2	H3	H4	H5	H5ax
Xyl2-xylitol	4.580	3.352	3.574	3.782	4.101	3.379
Xyl-xylitol	4.554	3.32	3.447	3.622	3.968	3.310
(Me)GlcA-Xyl4-xylitol	4.573	3.34	3.595	3.793	4.104	3.379
(Me)GlcAXyl3-xylitol	4.578	3.357	3.573	3.787	4.104	3.379

These oligosaccharides were generated by NaBH₄ reduction of the products generated by endoxylanase treatment of wild-type GX. Chemical shifts are reported in ppm relative to internal acetone, δ 2.225.

Table 6.

Selected Properties of Wild-Type and Mutant Arabidopsis GXs

	GX
Plant	Average DP	Percentage with Glycosyl Sequence 1 at End	μmol/g AIRs
Wild type	93	84	10
fra8	80	22	5
irx8	125	4	2
irx9	28	98	14

The degree of polymerization (DP) of the GXs and the percentage of GX with glycosyl sequence 1 at the reducing end were determined from the ratios of total residues to Xyl-xylitol and GalA-xylitol estimated by ¹H-NMR spectroscopy. The average DP allows the molecular weight of the GX to be estimated. Dividing the amount of GX (mg/g AIRs; see Table 2) by its molecular weight allows the amount of GX (μmol/g AIRs) to be calculated (see text for details).

DISCUSSION

The data we have obtained in this study provide compelling evidence that glycosyl sequence 1 is present at the reducing end of Arabidopsis GX. This structure was first identified as a component of birch and spruce GXs (Shimizu et al., 1976; Johansson and Samuelson, 1977; Andersson et al., 1983). Our study sheds light on its biological significance.

IRX8 and FRA8 Are Required for the Synthesis of Glycosyl Sequence 1 at the Reducing End of Arabidopsis GX and a Normal Amount of GX Chains

Our results indicate that functional FRA8 and IRX8 genes (or homologous genes) are necessary for the biosynthesis of glycosyl sequence 1, which is located at the reducing end of the GX. The presence of small amounts of this structure in these mutants may be attributable to the expression of homologs of these genes. This is in agreement with our recent finding that overexpression of a FRA8 homolog (At5g22940) that is normally expressed at low levels in Arabidopsis stems can functionally complement the fra8 mutation (our unpublished data).

We have shown that the irx8 mutation leads to a reduction in the number of GX chains and that most of the small amount of GX that is synthesized by these plants lacks glycosyl sequence 1 at its reducing end. These findings indicate that IRX8 is involved in the synthesis of glycosyl sequence 1, whose presence is correlated with normal GX biosynthesis. IRX8 encodes a putative GT in family GT8 (Brown et al., 2005; Persson et al., 2005) with homology with GUAT1, an α-GalA transferase (Sterling et al., 2006), and QUA1, a putative α-GalA transferase (Bouton et al., 2002), whose expression is correlated with xylan and homogalacturonan biosynthesis (Orfila et al., 2005). Family GT8 enzymes are retaining glycosyl transferases that catalyze the formation of α-glycosidic bonds when using α-linked donor substrates such as UDP-α-d-GalA. Thus, it is possible that IRX8 catalyzes the addition of an α-d-GalA residue to O4 of the reducing Xyl residue of glycosyl sequence 1.

FRA8 was shown previously to be required for the synthesis of normal amounts of GX (Zhong et al., 2005). The FRA8 gene encodes a putative GT in family GT47 (Zhong et al., 2005). This family includes enzymes with an inverting mechanism, which usually leads to β-glycosidic linkages (when typical α-linked donor substrates are used). Thus, it is possible that FRA8 catalyzes the formation of the β-linkage of xylose to either O3 of the rhamnose or O4 of the penultimate xylose of glycosyl sequence 1 (Figure 1) if UDP-α-d-Xyl is the donor substrate. However, in plants, the addition of α-Rha residues is catalyzed by inverting GTs that use UDP-β-l-Rha as the donor substrate (Watt et al., 2004). Therefore, it is also possible that FRA8 catalyzes the addition of rhamnose during the biosynthesis of glycosyl sequence 1.

The ratio of GlcA to MeGlcA in GXs from irx8, irx9, and fra8 plants is lower than in wild-type GX, such that MeGlcA predominates in the mutant plants. We previously described several mechanisms that are consistent with this observation (Zhong et al., 2005). Among these was the idea that the proportion of GlcA residues methylated depends on a balance between the rate of GX synthesis and the rate at which GlcA side chains of the nascent GX are methylated (Kauss and Hassid, 1967). In this scenario, the GlcA methylation rate, which is likely to be controlled by the availability of the methyltransferase and donor substrate (S-adenosyl-l-Met) (Kauss and Hassid, 1967), is lower than the rate at which GlcA side chains are added. Thus, in wild-type plants, only a portion of the GlcA side chains are methylated. Conversely, methylation may keep pace with GX synthesis if the rate of GX synthesis is decreased, thereby decreasing the rate at which GlcA side chains become available as acceptor substrates. Thus, in mutant plants that produce less GX, a higher proportion of GlcA is methylated. The observation of this effect in all three mutant plants described here is consistent with this saturation mechanism.

IRX9 Is Essential for the Normal Elongation of GX Chains

The elongation of the GX chain appears to require the cooperation of the transferases that catalyze the addition of xylose to the main chain and the addition of GlcA as a side chain (Baydoun et al., 1989). Thus, our observation that irx9 plants synthesize GX with reduced chain length could result from the disruption of either of these two processes. Our data indicate that IRX9 is not involved in the synthesis of glycosyl sequence 1 (Table 6). IRX9 is a putative GT in family GT43. Enzymes in this family are distinguished by an inverting mechanism, typically catalyzing the formation of β-glycosidic bonds using α-linked glycosyl donors. Our demonstration that the irx9 mutation leads to a decrease in the chain length of GX suggests that IRX9 encodes a xylan synthase responsible for adding β-xylosyl residues to the nascent GX. This hypothesis is consistent with our results indicating that IRX9 is highly expressed in cells undergoing secondary wall biogenesis and that IRX9 is localized in the Golgi, where GX synthesis occurs (Baydoun and Brett, 1997; Gregory et al., 2002). However, it is also possible, as we discussed previously (Zhong et al., 2005), that an inverting GT can catalyze the formation of an α-linkage when a β-linked substrate (such as glycosyl phospholipids) is used as the donor. Such inverting enzymes can also catalyze the formation of high-energy β-linked glycosides (such as glycosyl phospholipids) that are subsequently used as glycosyl donors. Thus, an alternative interpretation is that IRX9 is directly or indirectly involved in the transfer of α-linked GlcA residues to the GX backbone. Additional studies are required to determine whether IRX9 catalyzes the addition of xylose or GlcA to the GX backbone.

Glycosyl Sequence 1 May Function as a Link to Other Wall Polymers

There is evidence suggesting that nascent GX is covalently attached to protein (Crosthwaite et al., 1994). We obtained no evidence that GX is attached to protein, or any other cell wall polymer, via glycosyl sequence 1. If such a link exists, it is cleaved by alkali during GX extraction. Clearly, demonstration that glycosyl sequence 1 constitutes a covalent connection between GX and other wall polymers will require structural characterization of the intact link.

Models of GX Biosynthesis during Secondary Wall Formation

Our data, together with the results of previous studies (Shimizu et al., 1976; Andersson et al., 1983), show that glycosyl sequence 1 is located at the reducing ends of GXs in angiosperms and gymnosperms, suggesting that evolutionarily diverse plants synthesize GX in a similar manner. Two possible biosynthetic mechanisms that differ in the direction of xylan chain elongation and the function of glycosyl sequence 1 are consistent with our experimental observations. The first mechanism requires that glycosyl sequence 1 is a primer for GX synthesis and that the backbone is extended by adding xylose to the nonreducing end. There is precedent for this mechanism, as animal glycosaminoglycans are formed by adding monosaccharides to the nonreducing end of a linker oligosaccharide that is attached to protein via a xylose residue at the reducing end (Prydz and Dalen, 2000; Sugahara and Kitagawa, 2002). If this mechanism occurs in plants, different enzymes are likely to initiate primer synthesis, as BLAST searches of the Arabidopsis genomic database revealed no close homologs of the xylosyltransferase (EC 2.4.2.26) that attaches the xylosyl residue of the linker oligosaccharide in glycosaminoglycans to protein (Prydz and Dalen, 2000; Sugahara and Kitagawa, 2002). The second mechanism requires that GX is synthesized by adding monosaccharides to the reducing end and that transfer of glycosyl sequence 1 (either one residue at a time or en bloc) to the reducing end terminates chain elongation. The synthesis of hyaluronan (Bodevin-Authelet et al., 2005; Tlapak-Simmons et al., 2005) and starch (Mukerjea and Robyt, 2005), which have been reported to occur by the addition of monosaccharides to the reducing end, serve as precedents for this mechanism. Furthermore, the presence of GX chains that do not have glycosyl sequence 1 at the reducing end suggests either that glycosyl sequence 1 is cleaved from some of the GX chains during wall biogenesis or that GX chains can be synthesized without involving glycosyl sequence 1 as a primer. The correlation of decreased synthesis of glycosyl sequence 1 with a lower proportion of GX molecules having this structure at their reducing end supports a model in which glycosyl sequence 1 acts as a chain terminator. However, chain termination may still occur at relatively low frequency in the absence of glycosyl sequence 1 (e.g., by transfer of the nascent GX chain to water). Regardless of the actual mechanism operating in vivo, obtaining unambiguous evidence for the direction of xylan backbone elongation will be technically challenging, although a general approach to address this question has been proposed (Mukerjea and Robyt, 2005).

Several explicit models that include specific roles for the IRX8, IRX9, and FRA8 gene products in GX biosynthesis are consistent with the available data. A common characteristic of many of these models is the requirement for multiple-subunit enzyme complexes. If this notion is correct, then it is unlikely that biochemical activity can be shown using in vitro assays based on individual gene products. The apparent complexity of the process and the products formed suggests that simple oligosaccharides may not function as acceptor substrates. Nevertheless, we expressed the recombinant IRX9 protein in yeast and assayed for xylosyltransferase activity as described (Zhong et al., 2005). However, we did not detect transferase activity (data not shown). This may be attributable to improper folding or incorrect posttranslational modification of the protein in a heterologous expression system or to the use of inappropriate acceptor substrates. In any case, our observations provide a framework for the rational design of experiments to understand GX synthesis and to assign biochemical functions to specific gene products involved in this process.

In summary, our data suggest that the coordinated action of FRA8, IRX8, and IRX9 is necessary for the regulation of GX chain length and for the normal assembly and morphology of secondary cell walls in Arabidopsis. FRA8 and IRX8 (or their homologs) are likely to be required for the synthesis of glycosyl sequence 1 at the reducing end of the GX, whereas IRX9 is involved in elongation of the GX chain. Our results indicate that mutating IRX8, IRX9, or FRA8 affects the amount and extractability of GX. Understanding the functions of the proteins encoded by these genes will facilitate the targeted modification of secondary cell walls by genetic manipulation of GX biosynthesis. Using these techniques to modify the physical properties of GX will almost certainly lead to the more efficient use of plant secondary walls in energy production, wood and paper production, and nutrition.

METHODS

Plant Growth

The T-DNA insertion lines SALK_008642, SALK_014026, SALK_058238, and SALK_057033 were obtained from the ABRC. Both wild-type Arabidopsis thaliana (ecotype Columbia) and mutant plants were grown in a greenhouse under the same conditions.

Expression Analysis

Interfascicular fiber cells and pith cells from inflorescence stems of 5-week-old Arabidopsis plants were isolated using the PALM microlaser system (PALM Microlaser Technologies), and the microdissected cells were used for RNA isolation and amplification as described (Zhong et al., 2006). The amplified RNA was used in RT-PCR analysis of the expression levels of genes. PCR was performed for various numbers of cycles to determine the logarithmic phase of amplifications. The experiments were repeated three times, and identical results were obtained. RNA without reverse transcription was used as a negative control for PCR amplification, and no PCR product was detected. The expression of a ubiquitin gene was used as an internal control.

Complementation of Mutants and GUS Reporter Gene Analysis in Arabidopsis

A 6.4-kb genomic DNA fragment containing the IRX8 gene (including a 2.7-kb 5′ upstream region, the entire exon and intron sequence, and a 1.4-kb 3′ downstream region) and a 6.8-kb DNA fragment containing the IRX9 gene (including a 3-kb 5′ upstream region, the entire exon and intron sequence, and a 1.5-kb 3′ downstream region) were amplified by PCR using a high-fidelity DNA polymerase with gene-specific primers and cloned into the binary vector pGPTV-HPT (Becker et al., 1992). The constructs were introduced into irx8-1 or irx9-2 plants, respectively, by Agrobacterium tumefaciens–mediated transformation (Bechtold and Bouchez, 1994). Transgenic plants were selected by growth on nutrient agar containing hygromycin and used for analysis.

The cell type and developmental expression patterns of the IRX8 and IRX9 genes were examined using the GUS reporter gene. The genomic DNA fragments that complemented the mutant phenotypes were used for GUS reporter gene analysis. The GUS gene was inserted in-frame right before the stop codon of the IRX8 or IRX9 gene and then cloned into the binary vector pBI101 to create the IRX9∷GUS and IRX8∷GUS constructs. The constructs were transformed into wild-type Arabidopsis for the generation of transgenic plants, which were used for expression analysis of the GUS reporter gene (Zhong et al., 2005).

In Situ Localization of mRNAs

Wild-type Arabidopsis stems were used for in situ mRNA localization as described (McAbee et al., 2005). Tissues were fixed in 2.5% formaldehyde and 0.5% glutaraldehyde and then embedded in paraffin. Ten-micrometer-thick sections were cut, mounted onto slides, and hybridized with digoxigenin-labeled IRX8 or IRX9 antisense RNA probe synthesized using the DIG RNA labeling mix (Roche). The hybridization signals were detected with alkaline phosphatase–conjugated antibodies against digoxigenin and subsequent color development with alkaline phosphatase substrates.

Subcellular Localization of Fluorescent Protein–Tagged Proteins

The full-length IRX8 and IRX9 cDNAs were amplified by PCR using a high-fidelity DNA polymerase and fused in-frame with GFP cDNA (ABRC). The IRX8-GFP and IRX9-GFP fusion cDNAs were cloned downstream of the cauliflower mosaic virus 35S promoter in the binary vector pBI121. The constructs were transformed into wild-type Arabidopsis for the generation of transgenic plants. The GFP signals from roots of 3-d-old transgenic seedlings were observed using a Leica TCs SP2 spectral confocal microscope (Leica Microsystems). Images were processed with Adobe Photoshop version 7.0 (Adobe Systems).

The colocalization of fluorescent protein–tagged IRX8 and IRX9 with a Golgi marker was performed in carrot (Daucus carota) protoplasts as described (Zhong et al., 2005). The IRX8 and IRX9 cDNAs were fused in-frame with YFP cDNA and ligated between the cauliflower mosaic virus 35S promoter and the nopaline synthase terminator in a high-copy vector. The IRX8-YFP or IRX9-YFP construct together with the MUR4-CFP construct (Burget et al., 2003) were cotransfected into carrot protoplasts. After 20 h of incubation, the transfected protoplasts were examined for yellow and cyan fluorescent signals using a Leica TCs SP2 spectral confocal microscope.

Immunolocalization of Xylan in the Cell Walls of Arabidopsis Stems

Basal internodes of stems from 10-week-old plants were fixed at 4°C overnight with 2% glutaraldehyde in PBS (33 mM Na₂HPO₄, 1.8 mM NaH₂PO₄, and 140 mM NaCl, pH 7.2). After fixation, tissues were dehydrated through a gradient of ethanol and embedded in LR White resin (Electron Microscopy Sciences). One-micrometer-thick sections were cut with a microtome and incubated with the LM10 or LM11 monoclonal antibody, which recognizes GX (Plantprobes) (McCartney et al., 2005), and then with fluorescein isothiocyanate–conjugated secondary antibodies. The fluorescence-labeled sections were observed using a Leica TCs SP2 spectral confocal microscope. Images from single optical sections were collected and processed with Adobe Photoshop. For transmission electron microscopy, 85-nm-thick sections were incubated with the LM10 monoclonal antibody and then with gold (10 nm)-conjugated secondary antibodies. The immunogold-labeled sections were visualized with a Zeiss EM 902A transmission electron microscope (Carl Zeiss).

Breaking Force Measurements

The basal parts of the main inflorescences of 10-week-old Arabidopsis plants were measured for their breaking force using a digital force/length tester (model DHT4-50; Larson System). Individual plants of wild type, irx8-1, irx9-2, and complemented irx8-1 and irx9-2 were used for breaking force measurements. The breaking force (g) was calculated as the force needed to break apart a stem segment (Zhong et al., 1997).

Histology of Arabidopsis Stems

Basal internodes of the inflorescence stems of 10-week-old plants were fixed at 4°C overnight with 2% glutaraldehyde in PBS (33 mM Na₂HPO₄, 1.8 mM NaH₂PO₄, and 140 mM NaCl, pH 7.2). After fixation, tissues were postfixed in 1% (v/v) OsO₄ and then dehydrated through a gradient of ethanol and embedded in Spurr's resin (Electron Microscopy Sciences). One-micrometer-thick sections were stained with toluidine blue for light microscopy. For transmission electron microscopy, 85-nm-thick sections were stained with uranyl acetate and lead citrate and visualized using a Zeiss EM 902A transmission electron microscope.

Isolation and Extraction of Cell Walls

Cell walls were prepared as alcohol-insoluble residues (AIRs) from inflorescence stems from ∼200 wild-type or mutant plants. Similar yields of AIRs (80 to 100 mg AIRs/g fresh weight of stems) were obtained for wild-type and mutant plants. Polysaccharides were solubilized (Zhong et al., 2005) by sequential extraction of the AIRs with (1) endopolygalacturonase (10 units; prepared from Aspergillus niger) and pectin methylesterase (10 units; prepared from Aspergillus aryzae; Novozymes); (2) a xyloglucan-specific endoglucanase (XEG; 7.8 units; Novozymes) purified as described (Pauly et al., 1999); (3) 1 n KOH containing 1% (w/v) NaBH₄; and (4) 4 n KOH containing 1% (w/v) NaBH_4. The 1 and 4 n KOH extracts were neutralized with glacial acetic acid and then dialyzed against water. After dialysis, >80% of the material remained soluble. GXs with a reducing residue at one end were solubilized by treating irx9 AIRs (previously extracted with pectin methylesterase/endopolygalacturonase and XEG) with 0.25 n KOH in the absence of NaBH₄. A low concentration of KOH was used to minimize base-catalyzed degradation of the reducing end of the polysaccharide (Johansson and Samuelson, 1977). GX in the 0.25 n KOH extract was precipitated by the addition of ethanol (to 80%) containing acetic acid. The precipitated GX was dialyzed and lyophilized.

Glycosyl Residue Composition Analysis

The 1 and 4 n KOH–soluble polysaccharides were converted to alditol acetate derivatives as described (Carpita and Shea, 1989). The derivatives were analyzed by gas chromatography–electron-impact mass spectrometry.

Enzyme Treatment of Cell Wall Extracts and Generation of Xylan Oligosaccharides

The 1 and 4 n KOH–soluble materials were treated for 24 h at 37°C with Trichoderma viride endoxylanase (0.04 units/10 mg fraction; Megazyme). The endoxylanase digestion products were separated into neutral and acidic fractions by SEC using a Sephadex G-25 column (90 × 2.5 cm) eluted with water. Because of their larger size and ionic nature, acidic oligosaccharides were eluted from this matrix before the neutral oligosaccharides. Fractions containing the acidic oligosaccharides were pooled and analyzed by MALDI-TOF MS.

Size-Exclusion Chromatographic Analyses of Xylans

The materials solubilized by treating Arabidopsis cell walls with 1 and 4 n KOH were treated separately with endoxylanase. The reaction mixtures were then separated by SEC using a Superose 6 HR10/30 column. The molecular weight distribution of the GX in these alkali extracts was evaluated by comparing the SEC profiles of untreated and endoxylanase-treated materials (see Supplemental Figure 1 online). Polymeric GX was identified in these profiles by its susceptibility to endoxylanase and resistance to XEG treatment (0.78 units/10 mg; Novozymes) (Pauly et al., 1999).

Calculation of the Molecular Properties of the Xylans

The cross-peaks in the COSY NMR spectra were integrated to determine the relative abundances of the tetraglycosyl-xylitol (sequence 2; Figure 1), the xylosyl-xylitol, and all residues in the reducing oligosaccharides in samples prepared by endoxylanase treatment of KOH-solubilized materials. This enables the average number of residues in the polymer (DP) and the ratio of tetraglycosyl-xylitol to xylosyl-xylitol to be calculated. This gives the percentage of GXs with glycosyl sequence 2 at the end and allows the average molecular weight of the GXs to be calculated, which, in turn, allows the amounts of GX (mg/g AIRs) to be converted to a molar value (μmol/g AIRs).

MALDI-TOF MS

Positive-ion MALDI-TOF mass spectra were recorded using an Applied Biosystems Voyager-DE biospectrometry workstation. Aqueous samples (1 μL of a 1-mg/mL solution) were mixed with an equal volume of matrix solution (0.1 M 2,5-dihydroxbenzoic acid and 0.03 M 1-hydroxyisoquinoline in aqueous 50% MeCN) and dried on the MALDI target plate. Typically, spectra from 200 laser shots were summed to generate a mass spectrum.

¹H-NMR Spectroscopy

GX and xylo-oligosaccharides (4 to 8 mg) were dissolved in D₂O (0.7 mL, 99.9%; Cambridge Isotope Laboratories). ¹H-NMR spectra were recorded with a Varian Inova NMR spectrometer operating at 600 MHz and with a sample temperature of 298 K. All two-dimensional spectra were recorded using standard Varian pulse programs. Chemical shifts were measured relative to internal acetone (δ 2.225). Resonances were assigned to specific glycosyl residues on the basis of their scalar coupling patterns and chemical shifts. Intraglycosidic scalar couplings and long-range (interglycosidic) scalar couplings were simultaneously detected in gCOSY spectra of the oligosaccharides recorded at high resolution (Jia et al., 2005). Interglycosidic dipolar couplings for oligosaccharides were observed in NOESY spectra recorded using a mixing time of 200 ms. ROESY spectra were recorded with a mixing time of 200 ms and a spin-lock field (γB₁) of 1488 Hz.

Isolation of β-d-Xylp-(1→4)-β-d-Xylp-(1→3)-α-l-Rhap-(1→2)-α-d-GalpA-(1→4)-d-Xylitol

Endoxylanase-generated GX oligosaccharides were reacted for 2 h at 65°C and pH 5.5 with 0.2 M 2-aminobenzamide in the presence of 1 M NaBH₃CN to quantitatively convert the reducing oligosaccharides to their fluorescent derivatives (Ishii et al., 2002). The β-d-Xylp-(1→4)-β-d-Xylp-(1→3)-α-l-Rhap-(1→2)-α-d-GalpA-(1→4)-d-xylitol was readily separated from the relatively nonpolar 2-aminobenzamide–derivatived xylo-oligosaccharides by reverse-phase HPLC on an octadecyl silica column (Zorbax; RP-18; 25 × 10 cm). Oligosaccharides were eluted with aqueous methanol (0 to 25% [v/v] at 1 mL/min) and detected using a Sedex 55 evaporative light scattering detector (Sedere).

Accession Numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries under the following accession numbers: FRA8 (DQ182567), IRX8 (NM_124850), and IRX9 (NM_129265).

Supplemental Data

The following material is available in the online version of this article.

• Supplemental Figure 1. SEC on Superose 6 HR10/30 of the Materials Solubilized by Treating Arabidopsis Cell Walls with 1 and 4 n KOH before and after Treatment with Endoxylanase.