Homogalacturonan synthesis in Arabidopsis thaliana requires a Golgi-localized protein with a putative methyltransferase domain

Authors


(fax + 33 1 30 83 30 99; e-mail gmouille@versailles.inra.fr).

Summary

Pectins are a family of complex cell-wall polysaccharides, the biosynthesis of which remains poorly understood. We identified dwarf mutants with reduced cell adhesion at a novel locus, QUASIMODO2 (QUA2). qua2-1 showed a 50% reduction in homogalacturonan (HG) content compared with the wild type, without affecting other cell-wall polysaccharides. The remaining HG in qua2-1 showed an unaltered degree of methylesterification. Positional cloning and GFP fusions showed that QUA2, consistent with a role in HG synthesis, encodes a Golgi-localized protein. In contrast to QUA1, another Golgi-localized protein required for HG-synthesis, QUA2 does not show sequence similarity to glycosyltransferases, but instead contains a putative methyltransferase (MT) domain. The Arabidopsis genome encodes 29 QUA2-related proteins. Interestingly, the transcript profiles of QUA1 and QUA2 are correlated and other pairs of QUA1 and QUA2 homologues with correlated transcript profiles can be identified. Together, the results lead to the hypothesis that QUA2 is a pectin MT, and that polymerization and methylation of homogalacturonan are interdependent reactions.

Introduction

Pectins are extremely complex polysaccharides, which play key roles in plant growth, development and defence (for review see Willats et al., 2001). They can be envisioned as multi-block co-polymers. The simplest of these blocks is homogalacturonan (HG), an unbranched polymer of (1→4)-α-d-GalpA. A second major block, rhamnogalacturonan I (RG-I), is composed of a scaffold of repeating disaccharide units [→2)-α-l-Rhap-(1→4)-α-d-GalpA-(1→]n, which is in most cases decorated with arabinan and (arabino)-galactan side chains. A third minor building block, referred to as RG-II, has a highly conserved and complex structure based on an HG backbone decorated with different monosaccharides, and forms dimers in the wall through boron di-ester bonds, which may play a role in the control of wall porosity. Immunocytological studies suggest that pectin is synthesized in the Golgi apparatus in a highly methyl-esterified form (Zhang and Staehelin, 1992) and transported in vesicles to the cell surface (Northcote and Pickett-Heaps, 1966). Upon incorporation in the cell wall, pectins can be selectively demethylated by wall-bound pectin methylesterases (Mohnen, 1999; Zhang and Staehelin, 1992). The degree of methyl-esterification and the distribution pattern of methyl groups on the HG backbone determine, to a large extent, the ability of pectins to interact with calcium ions in vitro (Kohn et al., 1983; Ralet et al., 2003; Thibault and Rinaudo, 1985) and most likely in vivo.

Despite the biological and industrial importance of pectin, very little is known about its biosynthesis. This is due in part to the complexity of the pectin synthesis machinery, which involves at least 53 enzymes (Mohnen, 1999), and the difficulty in studying the biochemistry of Golgi-localized membrane-bound enzymes, which may be part of high molecular-weight complexes (Doong and Mohnen, 1998). Recently the first pectin synthesis mutants in a putative glycosyltransferase (GT) (QUA1) in Arabidopsis (Bouton et al., 2002), and a putative RG-II glucuronosyltransferase (NpGUT1) inNicotiana plumbaginifolia callus, were described (Iwai et al., 2002). qua1 mutants are dwarfed with reduced cell adhesion and show a specific reduction in GalA in the cell walls, suggesting that the encoded putative GT is involved in HG synthesis (Bouton et al., 2002). Fifteen QUA1-related proteins, all members of CAZY family 8 (Henrissat et al., 2001), are encoded by the Arabidopsis genome. One of these members (galacturonosyltransferase1 or GAUT1) was recently identified as a galacturonosyltransferase (GalAT) following co-purification of HG:GalAT activity and GAUT1 protein from Arabidopsis suspension cells; by recovery of HG:GalAT activity upon immunoprecipitation using an anti-GAUT1 antibody; and following transient expression of GAUT1 in human embryonic kidney cells (Sterling et al., 2006). Pectin methyltransferase (MT) activity, which is able to methylate HG oligosaccharides (oligogalacturonides) and pectin with a low level of methylation, has been detected in microsomes of tobacco cells (Goubet et al., 1998) and soybean hypocotyls (Ishikawa et al., 2000). The enzyme activity has been solubilized and partially purified, but the protein remains to be identified.

In this study, we identified a novel class of allelic mutants with a cell-wall phenotype very similar to that of qua1. Mutants were also dwarfed, with reduced cell adhesion. Cell-wall analysis showed that the qua2-1 mutant was specifically affected in pectin, with a 50% reduction of HG content without observable changes in other polysaccharides, including RG-I. The remaining acid-extractable HG in qua2-1 showed a degree of methylation comparable with that of the wild type (WT). The gene was isolated and encodes a Golgi-localized, predicted type II membrane protein, which comprises a putative MT domain. Transcript profiles of QUA1 and QUA2 were correlated, and other correlated pairs of QUA1 and QUA2 homologues could be identified. Our results support the hypothesis that QUA2 is a pectin MT and that HG synthesis involves a coupled GalAT and pectin MT reaction carried out by the QUA1/QUA2 pair.

Results

FT–IR microspectroscopy-based clustering identified novel putative pectin mutants

Fourier transform–infrared (FT–IR) microspectroscopy is a powerful method for the identification of mutants with cell-wall alterations (Mouille et al., 2003). Information about cell-wall changes in unknown mutants can be obtained through their clustering with already characterized mutants. Dark-grown hypocotyls of a collection of dwarf mutants obtained in our laboratory were screened by FT–IR microspectroscopy. Two ethyl methane sulfonate (EMS)-induced mutants clustered with two quasimodo1 (qua1) alleles (Figure 1a). This name was chosen for the deformed dwarf phenotype of the mutant seedlings, reminiscent of Quasimodo in the novel Notre-Dame de Paris by Victor Hugo. Complementation analysis identified one novel allele of qua1 (qua1-3), which was confirmed by sequencing (G1220A in the coding sequence, causing G407E). Interestingly, the second mutant was not allelic to qua1. The phenotype of this mutant was very similar to that of qua1 alleles, including short, dark-grown hypocotyls, reduced cell adhesion (Figure 1b) and a dwarfed mature plant (data not shown). This mutant was referred to as quasimodo2-1 (qua2-1).

Figure 1.

 Defects in cell elongation, cell adhesion and cell-wall composition observed in quasimodo2 (qua2) mutants.
(a) Dendrogram obtained by hierarchical cluster analysis of Arabidopsis lines based on FT–IR spectra sampled from dark-grown hypocotyls. Reference mutants and WT treated with inhibitors have been described by Mouille et al. (2003). qua2-1 and qua2-2 cluster with qua1 alleles, indicating that they have comparable cell-wall changes in the hypocotyl. qua2-1 seedlings transformed with a construct expressing QUA2::GFP from its native promoter showed a WT phenotype (Figure 4a) and clustered with Col0 and WS, indicating the complementation of the cell-wall defect by the transgene.
(b) Phenotypes of 4-day-old dark-grown seedlings.
(c) Confocal laser scanning microscopy of propidium iodide-stained 4-day-old hypocotyls of dark-grown seedlings. Note the irregular surface of the hypocotyl due to the detached epidermal cells.

Pectins in qua2-1 showed a 50% reduction in HG content compared with WT, without a change in the degree of methylation of the remaining H + -extracted HG

The monosaccharide composition of WT and qua2-1 cell-wall material obtained from mature leaves of glasshouse-grown plants is shown in Figure 2(a). The GalA content was significantly lower in qua2 mutants (13% less) compared with WT, with a compensatory small increase in the amount of all other monosaccharides. This suggests a specific reduction in pectin content in the mutant, without a change in other polysaccharides. To compare the fine structure of mutant and WT pectin, pectic polysaccharides were sequentially extracted from the cell-wall material with hot dilute acid (H+ fraction) and cold dilute alkali (OH fraction). This procedure recovered around 80% of all GalA and Rha, both in WT and qua2 (data not shown). From these extracts, highly enriched pectic fractions were obtained by anion-exchange chromatography. To assess whether the qua2 mutation specifically affected HG or RG-I regions, we isolated HG domains exploiting the differences in susceptibility to acid hydrolysis of the glycosidic linkages. Indeed, under mild acid conditions, the linkages between GalA and Rha in the RG-I domains are more labile than those between adjacent GalA residues in HG domains (Thibault et al., 1993). Under the conditions used, resistant HG domains were recovered as acid-insoluble material, while hydrolyzed RG-I domains were solubilized (Thibault et al., 1993). Knowing the amount of pectin extracted per g cell-wall material, the GalA content of the purified H+- and OH-extracted pectins and the distribution of GalA between HG and RG-I, we could calculate the amount of RG-I-derived and HG-derived GalA per g cell-wall material (Figure 2b). The amount of RG-I-derived GalA did not differ significantly between WT and qua2. Instead, HG-derived GalA showed a 50% reduction in qua2 compared with WT. These results demonstrate unambiguously that qua2 is specifically deficient in HG, with no change in the amount of the RG-I domains.

Figure 2.

 Reduced homogalacturonan content in qua2-1.
(a) Monosaccharide composition of cell-wall material from leaves of 4-week-old plants. The galacturonic acid (GalA) content in qua2-1 is 13% lower than in the WT. The monosaccharide composition was done in triplicate on an extract consisting of 24 individual plants. Error bars, SD.
(b) Distribution of GalA between purified rhamnogalacturonan-I (RG-I) and homogalacturonan (HG) fractions of the same cell-wall material. The GalA content is 50% reduced in qua2-1 compared with the WT in purified HG, but no significant differences are observed between RG-I fractions from mutant and WT. Pectin purification was done in triplicate. Error bars, SD.
(c) Localization of HG using PAM1 antibodies. PAM1, which recognizes unesterified HG domains, labels cell walls lining intercellular spaces between parenchyma cells in WT inflorescence stems. This labelling is absent in qua2-1 stems.

Next, we determined the degree of methyl-esterification of WT and mutant pectic fractions. As the OH fraction was completely de-esterified during the extraction, only the purified H+ fraction could be analyzed. Methyl-esterification is thought to be restricted to HG domains (Komalavilas and Mort, 1989;Perrone et al., 2002; Ralet et al., 2005). Based on this premise and the average GalA distribution between HG and RG-I domains for WT and qua2, we estimated the degree of methyl-esterification of the HG domains at 77 (±6) and 85 (±5) for WT and qua2, respectively. To confirm this, we studied the digestibility of the purified H+ fractions from WT and qua2 with an endopolygalacturonase (endo-PG). Endo-PGs hydrolyse the 1–4 linkages between adjacent α-d-GalA residues, and are known to act specifically on HG. The digestibility of both WT and qua2 pectins was poor (data not shown), and comparable with that of the highly methylated lime pectin (degree of methyl-esterification >75), which confirms the elevated degree of methyl-esterification of their HG domains. In conclusion, these results demonstrate that qua2 specifically causes a 50% reduction in the amount of HG without significantly affecting the RG-I content and the degree of methyl-esterification of the remaining HG.

Using specific anti-pectin antibodies, we further investigated the distribution of pectins in WT and qua2. The monoclonal antibody PAM1 recognizes the unesterified HG component of the pectin backbone (Willats et al., 1999). This antibody strongly labelled walls lining the intercellular spaces between cortical cells of Col0 inflorescence stems. Interestingly, qua2 stems showed a much weaker labelling (Figure 2c), which confirms the reduced amount of HG in the mutant.

QUA2 encodes a novel protein with a putative MT domain

As a first step in a map-based cloning strategy, we generated an F2 mapping population from a cross between qua2-1 (in a Col0 background) and WT accession WS. The genotyping of 40 homozygote mutant seedlings located the mutation to chromosome 1, between positions 27.300.000 and 30.300.000. Using two flanking markers, we selected additional recombinants within this interval among a population of 1000 F2 plants. Additional markers further restricted the interval to 116 kb. This interval contained 32 annotated ORFs. Given the comparable phenotype of qua1 and qua2 alleles, we reasoned that QUA2 could interact with QUA1, and hence both genes might be co-regulated at the transcript level. Using the genesis program (http://genome.tugraz.at), we performed a hierarchical clustering, using the NASCArray data set (Nottingham Stock Center Transcriptomics Service; http://affymetrix.arabidopsis.info/narrays/help/psp-wbubn.html), of 26 ORFs (six ORFs were not represented on the Affymetrix chips) within the mapping interval plus QUA1 (see in Supplementary material). Sequencing of the most co-regulated At1 g78240 gene in qua2-1 revealed a mutation at position + 1879 of the coding sequence, causing an amino-acid change to a stop codon eliminating 59 amino acids of the predicted protein (Figure 3). One additional allele for gene At1 g78240 was isolated from the GABI-Kat collection (Rosso et al., 2003; Figure 3). The T-DNA was inserted at position + 1808 of the coding sequence, and homozygous lines for the insertion showed a morphological and FT–IR phenotype comparable with that of qua2-1 (data not shown; Figure 1a). The T-DNA insertion allele was further referred to as qua2-2.

Figure 3.

QUA2 encodes a putative type II membrane protein with a methyltransferase domain.
Structure of QUA2 gene (above) and QUA2 protein (below). Exon and intron positions, and the position of the T-DNA insertion in qua2-2 and the point mutation in qua2-1, are shown. Predicted membrane anchor and methyltransferase domains in QUA2 are also shown.

The final proof that the mutation in At1 g78240 was responsible for the qua2 phenotype was obtained by the complementation of all aspects of the phenotype by a QUA2–green fluorescent protein (GFP) fusion protein expressed from its native promoter (Figure 4a; see below). The ORF At1 g78240 encodes a predicted type II membrane protein, with the predicted membrane-spanning domain close to the amino terminus, separating a short cytoplasmic tail from a predicted extra-cytoplasmic domain (Figure 3). The latter domain shows two conserved motifs. First, the sequence from aa 155 to 676 can be aligned over its entire length (Figure 3; for alignment) with the pfam03141 motif corresponding to the ‘Domain of Unknown Function (DUF) 248’, which characterizes a family of putative MTs. Second, within DUF248, a SAM-dependent MT motif can be identified from aa 230 to 372.

Figure 4.

 QUA2–GFP accumulates in the Golgi apparatus.
(a) Phenotype of 5-day-old light-grown seedlings of WT, qua2-1 and qua2-1 expressing QUA2::GFP from its native promoter (qua2QUA2::GFP). Note that expression of QUA2::GFP restores a WT phenotype in a qua2-1 background.
(b) Subcellular localization of QUA2–GFP in the hypocotyl epidermis of a stable Arabidopsis transformant. QUA2–GFP accumulates in intracellular compartments (labelled in green). Red objects are chloroplasts.
(c) Co-localization of QUA2–GFP with Golgi-marker sialyltransferase–YFP, both transiently expressed in epidermal cells of Nicotiana benthamiana. QUA2–GFP co-localizes with ST–YFP.

QUA2 and QUA1 are localized in the Golgi apparatus

To determine the intracellular localization of QUA2, we transformed the genomic sequence of QUA2 fused at its C-terminus to GFP behind the native promoter (2 kb) into qua2-1. This construct (QUA2::GFP) entirely complemented the mutant growth phenotype, as shown by the WT dark-grown and light-grown hypocotyl length and the absence of the cell adhesion defect (Figure 4a; data not shown) in the transformant as well as the normally sized adult plants (data not shown). In addition, the qua2 cell-wall defect was complemented by the construct, as shown by the presence of the qua2QUA2::GFP transformant within the FT–IR cluster containing the WT controls Col0 and WS (Figure 1a). The fusion protein was therefore functional, and the localization of the GFP fluorescence should reflect the localization of the native QUA2 protein (Figure 4b). GFP fluorescence was detected in hypocotyls and leaves of these transgenic plants, and accumulated in discrete intracellular compartments. These compartments were highly motile, as shown in time-lapse movies (see qua2QUA2GFP.avi, in Supplementary material). This motility was abolished by a 1-h treatment with the actin inhibitor latrunculin B (0.1 μm) but not with 10 μm of the microtubule inhibitor oryzalin (qua2QUA2GFPlatrunculin.avi and qua2QUA2GFPoryzalin.avi,). We previously showed that 10 μm oryzalin completely disorganized the microtubule network within 10 min, and abolished the motility of other intracellular compartments that contain GFP-KOR1, a protein involved in cellulose synthesis (Robert et al., 2005). Transient co-expression studies in Nicotiana benthamiana epidermal cells showed that QUA2–GFP co-localized with the Golgi marker sialyltransferase–yellow fluorescent protein (YFP) (Figure 4c; Brandizzi et al., 2002). Similar observations were made with a QUA1–GFP fusion construct expressed under the control of the CaMV 35S promoter in N. benthamiana (data not shown). The Golgi localization of both QUA1 and QUA2 was confirmed by proteomics studies on microsomes from Arabidopsis suspension-cultured cells (Dunkley et al., 2006).

QUA2 mRNA shows a ubiquitous expression pattern in Arabidopsis

Microarray data show that QUA2 mRNA was present in all tissues investigated (http://www.genevestigator.ethz.ch/at; Zimmermann et al., 2004). This was confirmed by QUA2 promoter–GUS fusions in transgenic Arabidopsis, which revealed GUS activity throughout seedlings and adult plants (Figure 5; data not shown). This is consistent with a generic role of QUA2 in pectin synthesis.

Figure 5.

 The uidA reporter gene under control of the QUA2 promoter sequence is expressed throughout dark-grown and light-grown seedlings.
(a) 4-day-old dark-grown seedlings; (b) 5-day-old light-grown seedlings.

QUA2 is a member of a gene family at least four members of which are co-regulated with distinct QUA1 family members

The Arabidopsis genome contains 29 QUA2 family members, each of which contains a conserved MT domain. Rice contains 25 QUA2 homologues. The phylogenetic tree showed at least 14 conserved isoforms with orthologues in rice, suggesting specialized functions within this gene family. Interestingly, based on the transcript-profiling data, other correlated Family 8 GT/QUA2 homologue pairs could be identified. For instance, probing the expression angler software (Toufighi et al., 2005; http://bbc.botany.utoronto.ca/ntools/cgi-bin/ntools_expression_angler.cgi; NASC392 data set) with QUA1 identified a second putative MT in addition to QUA2 (Table 1). The transcript profile of the closest QUA1 homologue (GAUT9, At3 g02350) also correlated with that of two MTs, one of which (At3 g13300) also correlated with transcript profile of QUA1. Another family 8 GT member (GAUT1, At3 g61130), the only member for which in vitro GalAT activity has been demonstrated (Sterling et al., 2006), also correlated with two MTs, one of which also correlated with GAUT9 (Table 1). This raises the interesting possibility that pairs of interacting QUA1-like and QUA2-like isoforms exist.

Table 1.   Co-expressed pairs of putative galacturonosyltransferases (GalATs) and methyltransferases
Putative GalAT QUA2 homologuesPearson coefficient
  1. Pairs were identified using expression angler software (Toufighi et al., 2005; http://bbc.botany.utoronto.ca/ntools/cgi-bin/ntools_expression_angler.cgi, NASC392 data set).

At3 g25140 (QUA1/GAUT8)At3 g233000.551
At1 g78240 (QUA2)0.541
At3 g02350 (GAUT9)At3 g233000.554
At1 g268500.520
At3 g61130 (GAUT1)At4 g143600.667
At1 g268500.577

Discussion

In this study, we demonstrated first that QUA2 is required for normal synthesis of HG in Arabidopsis, and second that QUA2 is a predicted type II membrane protein with a lumenal putative MT domain, which, consistent with a role in the synthesis of HG, accumulates in the Golgi apparatus.

qua1 and qua2 mutants have similar phenotypes: both mutants are dwarfed, with reduced cell adhesion, and show a reduced GalA content and very similar FT–IR profiles. QUA1 – also referred to as GAUT8 (Sterling et al., 2006) – is a member of GT family 8. So far, attempts to demonstrate the enzyme activity for QUA1 expressed in a heterologous system have failed. Recently, however, GalAT activity has been observed for a protein of the same family, GAUT1, expressed in human embryonic kidney cells (Sterling et al., 2006). Given the mutant phenotype and the similarity to GAUT1, it is likely that QUA1 also encodes a GalAT. Concerning QUA2, despite several attempts, we have not been able to produce a soluble version of the protein in a heterologous system, so it remains to be shown whether QUA2 indeed has MT activity. Methyltransferases constitute a large protein family with a wide array of substrates: DNA, proteins, lipids, small metabolites and oligo- and polysaccharides. Pectic polysaccharides are a likely substrate for QUA2, given its localization in the Golgi apparatus, which was also observed for pectin methyltransferase activity (Baydon et al., 1999; Bourlard et al., 1997; Goubet and Mohnen, 1999a,b), the co-regulation with QUA1 (suggesting a possible interaction between the two proteins) and, most importantly, the specific effects on HG synthesis of mutations in QUA2.

The requirement of both QUA1 and QUA2 for HG synthesis can be explained in at least two non-exclusive ways. First, the two proteins may be part of a protein complex, which may be destabilized in the absence of one of the partners, leading to the partial loss of GalAT activity. Second, it is conceivable that QUA1 GalAT requires a QUA2-methylated GalA as an acceptor. The requirement of both proteins for HG synthesis is consistent with the observation that GalAT activity could be solubilized from tobacco microsomes, but the enzyme was not processive on solubilization (Doong and Mohnen, 1998; Guillaumie et al., 2003). This suggests that the processivity of the enzyme requires co-factors, which may include pectin MT. Recently, Sterling et al. (2006) showed that GAUT1, expressed in embryonic kidney cells, catalyzed the incorporation of 14C-GalA from UDP-14C -GalA onto HG OGA acceptors, but did not investigate the processivity of this enzyme preparation. It will be interesting to see whether co-expression of both QUA1 and QUA2 will allow the reconstitution of HG synthesis in a heterologous system.

The qua2-1 mutation led to a 50% reduction in the amount of the HG. Either QUA2 is required for the synthesis of the entire HG pool and qua2-1 is a leaky allele, or other QUA2 family members are responsible for synthesis of the remaining HG in this mutant. Nevertheless, alternative, more elaborate explanations cannot be excluded at this stage: for instance, QUA2 could have a role in the methylation/synthesis of some of the methylated sugars in RG-II (such as methyl xylose or fucose), and indirectly, through some unknown mechanism, affect HG synthesis. At least 29 other expressed putative membrane-bound MTs are encoded by the Arabidopsis genome. Interestingly, putative MTs are among the most common plant Golgi membrane proteins as shown by proteomics studies (Dunkley et al., 2006), which identified 14 different predicted MTs in Golgi fractions from suspension cultured Arabidopsis cells. At least three of these putative MTs were co-expressed at the transcript level with QUA1/GAUT8, GAUT9 and GAUT1, respectively. It is therefore conceivable that different MT isoforms specifically interact with different GTs to carry out coupled reactions either to synthesize specific pectic polysaccharides (HG, RG-I, RG-II) or different steps in HG biosynthesis, perhaps in different cell types and/or environmental conditions.

Experimental procedures

Plant material and growth conditions

Arabidopsis seedlings were grown at 20°C on solid medium, as described by Estelle and Somerville (1987), without sucrose. Seeds were cold-treated for 48 h to synchronize germination, and plants were grown in a 16 h light/8 h dark cycle. For dark growth conditions, seeds were exposed to fluorescent white light (200 μmol m−2 sec−1) for 4 h to induce germination, after which plates were wrapped in three layers of aluminum foil. Seedling age was counted starting from the light exposure.

The qua2-1 mutant was identified among a collection of dark-grown, short-hypocotyl mutants isolated from an EMS-mutagenized M2 population in Col0. qua2-1 mutants were back-crossed twice before phenotypic analysis.

WT and qua2-1 cell wall material was prepared from 4-week-old glasshouse plants as alcohol-insoluble residue, as reported previously (Bouton et al., 2002). Lime pectins were prepared as described previously (Limberg et al., 2000). Fusarium moniliforme endopolygalacturonase was purified to homogeneity from a transformed Saccharomyces cerevisiae strain (Caprari et al., 1996) as described previously (Bonnin et al., 2001).

Sampling of FT–IR spectra

Four-day-old seedlings were squashed between two BaF2 windows and rinsed abundantly in distilled water for 2 min. The samples were then dried on the windows at 37°C for 20 min. For each mutant, 20 spectra were collected from individual hypocotyls of seedlings from four independent cultures (five seedlings from each culture), as described by Mouille et al. (2003). Normalization of the data and the discriminant variable selection method were performed as described by Mouille et al. (2003). Based on the Mahalanobis distances calculated using the 28 selected wavenumbers, a dendrogram was constructed using the Ward clustering algorithm (Mouille et al., 2003; Robin et al., 2003).

Monosaccharide composition of cell-wall material and soluble polysaccharides

The uronic acid content was determined by the automated m-hydroxybiphenyl (mhbp) method (Thibault, 1979) and by the method of Ahmed and Labavitch (1977). Quantification of GlcA versus GalA was assessed by the differences in response of these uronic acids in the mhbp method, depending on the presence of tetraborate (Filisetti-Cozzi and Carpita, 1991; Renard et al., 1999). Individual neutral sugars were analyzed as their alditol acetate derivatives (Blakeney et al., 1983) by gas liquid chromatography after hydrolysis with 2 m trifluoroacetic acid at 121°C for 2 h in the soluble materials. For analysis of insoluble materials, a prehydrolysis step in 72% H2SO4 (30 min, 25°C) followed by a hydrolysis step in 1 m H2SO4 (2 h, 100°C) was included (Saeman et al., 1954). The degree of methyl-esterification was calculated after HPLC determination of methanol released by alkaline de-esterification of pectins (Levigne et al., 2002). Isopropanol was added as internal standard.

Pectin sequential extraction

Cell-wall material (approximately 350 mg) was first extracted three times with 14 ml 0.05 n HCl at 85°C for 30 min. The slurry was filtered on G4-sintered glass and the residue was carefully rinsed with distilled water. The supernatant (H+ extract) was neutralized to pH 4.5 using 1 m NaOH, extensively dialyzed against distilled water and freeze-dried. The wet residue was resuspended in distilled water at 4°C and brought to pH 12.4 with NaOH. Three extraction steps of 30 min were carried out at 4°C. The slurry was filtered on G4-sintered glass and the residue was rinsed carefully with distilled water. The supernatant (OH extract) was neutralized to pH 4.5 using 1 m HCl, extensively dialyzed against distilled water and freeze-dried. The remaining residue was rinsed three times with EtOH/water 70:30 (v/v), dried by solvent exchange and left overnight at 40°C. Four independent sequential extractions were performed on Col0 and qua2-1 cell-wall material.

Pectin purification by anion-exchange chromatography

Low-pressure anion-exchange chromatograhy was performed at room temperature on a DEAE-Sepharose CL-6B column (32 × 2.6 cm). For the purification of H+ extracts, the column was equilibrated with degassed 0.05 m sodium succinate buffer pH 4.5. The freeze-dried H+ extracts were solubilized in distilled water (approximately 50 mg in 15 ml), loaded onto the column and eluted at a flow rate of 90 ml h−1. The gel was washed with 300 ml 0.05 m sodium succinate buffer. The bound material was then eluted with a linear NaCl gradient (0–0.6 m NaCl in 0.05 m sodium succinate buffer, 600 ml). Sodium succinate buffer containing 0.6 m NaCl (300 ml) was then applied. Fractions (9 ml) were collected, and GalA and total neutral sugars were quantified colorimetrically (Thibault, 1979; Tollier and Robin, 1979). The bound material (purified H+-extracted pectins) was dialyzed extensively against water and freeze-dried for further analysis. For OH extracts, the column was equilibrated with degassed 0.05 m NH3-succinate buffer pH 4.8. The freeze-dried OH extracts were suspended in distilled water (approximately 150 mg in 30 ml), brought to pH 7, and hydrolyzed by P5380, subtilisin A from Bacillus sp. protease (Sigma, http://www.sigmaaldrich.com) (1.5 ml at 5 mg ml−1) for 17 h at 40°C. After centrifugation with a bench-top centrifuge (10 000 g, 10 min), the supernatant was loaded onto the column and eluted at flow rate of 90 ml h−1. The gel was first washed with 300 ml 0.05 m NH3-succinate buffer. The bound material was then eluted with an NH3-succinate buffer gradient (0.05–1 m, 450 ml then 1–2 m, 150 ml). 2 m NH3-succinate buffer (300 ml) was then applied. Fractions (9 ml) were collected and analyzed as described above. The bound material (purified OH-extracted pectins) was extensively dialyzed against water and freeze-dried for further analysis.

Homogalacturonan isolation

Homogalacturonans were isolated as described previously (Thibault et al., 1993). Purified pectins were first de-esterified by 0.1 m NaOH at 4°C. De-esterified pectins (30–60 mg) were hydrolyzed by 0.1 m HCl (6 mg ml−1) at 80°C for 72 h in sealed tubes. After cooling, the acid-soluble (RG-I) and acid-insoluble (HG) fractions were separated by centrifugation of the reaction mixture at 15 000 g for 20 min. The insoluble fraction was washed and resuspended in distilled water (10 ml). The suspension was brought to pH 7 by 0.1 m NaOH and dialyzed against distilled water. The GalA content of the acid-soluble and acid-insoluble fractions was determined colorimetrically (Thibault, 1979).

Enzymatic hydrolysis by endopolygalacturonase and analysis of reaction products

Pectins (2 mg) were dissolved in 0.05 m sodium acetate buffer pH 4 by overnight shaking at room temperature. Pectin solutions were incubated at 30°C with 35 pkat of F. moniliforme endopolygalacturonase per mg substrate added at t = 0 and 25 pkat/mg substrate added at t = 26 h. At t = 28 h, hydrolysates were filtered on 0.45 μm Minisart RC15 Sartorius membranes and analyzed by high-performance anion-exchange chromatography (HPAEC) pH13. HPAEC was performed using a Waters (http://www.waters.com) 626 pump equipped with a Waters 600S controller and a Waters 717 plus autosampler. Oligomers were monitored using a pulse amperometric detector (EC 2000, Thermo Separation Products, http://www.thermo.com). borwin software (JMBS Developments, Grenoble, France) was used for data acquisition and processing. Filtered hydrolysates were applied (20 μl) on an analytical Carbopac PA-1 column (2 × 250 mm) equipped with a Carbopac PA-1 guard column at 0.25 ml min−1 (pH 13). The elution was carried out with 250 mm sodium acetate containing 100 mm NaOH (0–20 min), followed by two linear gradient phases of 500 to 700 mm sodium acetate in 100 mm NaOH (20–60 min) and 700 to 800 mm sodium acetate in 100 mm NaOH (60–65 min). Oligogalacturonates were quantified using response factors obtained from oligogalacturonate ‘home-made’ standards (Ralet et al., 2005).

Immunofluorescence detection of homogalacturonan

Plant tissue was fixed and sectioned as described by Willats et al. (1999). Sections (10 μm) were blocked with 5% milk powder in phosphate-buffered saline (MPBS) for 30 min and washed with PBS. PAM1 Phage antibodies (Willats et al., 1999) were applied for 1 h at a titer of approximately 1012 phage ml−1 diluted in 5% MPBS. After washing with PBS, anti-M13 antibody (Pharmacia, http://www.pfizer.com) was applied for 1 h as a 1/50 dilution in 5% MPBS. Following washing with PBS, anti-mouse antibody conjugated to fluorescein isothiocyanate (anti-mouse/FITC, Sigma) was applied for 1 h at a 1/100 dilution in 5% MPBS. Finally, sections were washed with PBS, mounted in PBS/glycerol-based anti-fade solution (Citifluor, http://www.citifluor.co.uk) and examined with a microscope equipped with epifluorescence (Willats et al., 1999).

Cloning procedures

The QUA2 2.1-kb promoter including sequence up to 12 bases 5′ to the ATG was PCR-amplified from WT Col0 using the QUA2promaatB1 and QUA2promaatB2 primers. The PCR fragment was transferred into the pDONR201 Gateway vector (Invitrogen, http://www.invitrogen.com). The promoter was then transferred to the pGWB3 binary vector placing the uidA reporter gene under the control of the QUA2 2.1-kb promoter. The pQUA2::uidA construct was transformed into Agrobacterium C58 (GV3580) (Wen-jun and Forde, 1989).

To construct the QUA2::GFP fusion, the QUA2 genomic fragment including 2.1 kb of sequence upstream of the ATG was PCR-amplified from WT Col0 using the QUA2promaatB1 and QUA2cDNAaatB2 primer pair. The resulting PCR fragment was transferred into the pDONR201Gateway vector (Invitrogen) and the clone was sequenced. The GFP fusion construct was made by transferring the QUA2 genomic sequence from the pDONR201 entry clone to pGWB4, creating a C-terminal fusion with the GFP gene.

Primer sequences used are: QUA2promaatB1 GGGGACAAGTTTGTACAAAAAAGCAGGCTGTGTTACATCAATCATCC QUA2promaatB2 GGGGACCACTTTGTACAAGAAAGCTGGGTCGCTATAAATCTAAGATCC QUA2cDNAaatB2 GGGGACCACTTTGTACAAGAAAGCTGGGTCGATTGATTGTCGCTTGGTG

Transient expression in epidermal cells of N. benthamiana

Four-week-old N. benthamiana glasshouse plants grown at 21°C were used for Agrobacterium-mediated transient expression (adapted from Kapila et al., 1997). Agrobacterium strain C58 (GV3580) was transformed by electroporation. A single colony from each transformant was inoculated into 1 ml LB medium. The bacterial culture was incubated at 28°C with agitation for 20 h. Bacterial culture (1 ml) was then pelleted in an Eppendorf tube by centrifugation at 2200 g for 5 min at room temperature. The pellet was washed once with 1 ml buffer (13 g l−1 S-medium (Duchefa, http://www.duchefa.com) and 40 g l−1 saccharose pH 5.7), and diluted to OD0.1 (at 600 nm) in the same buffer prior to infiltration. For experiments requiring co-expression of two different constructs, 0.5 ml of each diluted bacterial culture was mixed before infiltration. The bacterial suspension was inoculated using a 1-ml syringe without a needle by gentle pressure through the stomata on the lower epidermal surface. Transformed plants then were incubated under normal growth conditions. Transformed leaves were analyzed 48–72 h after infection of lower epidermal cells.

Laser scanning confocal microscopy

Images were collected with a spectral Leica SP2 AOBS confocal microscope (Leica Microsystems, http://www.leica-microsystems.com) equipped with an argon laser and an HeNe laser.

For cell-surface staining, seedlings were incubated in a propidium iodide solution (0.1 mg ml−1) for 5 min and rinsed with water prior to observation. The excitation line of an argon ion laser of 488 nm was used, and fluorescence was detected within 560–620 nm.

For imaging expression of GFP constructs, the excitation line of an argon ion laser of 488 nm was used. Fluorescence was detected within 495–540 nm.

For imaging of a combination of GFP and YFP, a dual-laser excitation with alternate 458 nm (for GFP) and 514 (for YFP) were used under line-switching mode. Fluorescence was detected at 475–527 nm (GFP) and 527–596 nm (YFP). The images were coded green (GFP) and red (YFP), giving yellow colours in merged images. Images were acquired using a 63 ×  water-immersion objective (pinhole 1 Airy unit). Each image shown represents either a single focal plane or a projection of individual images taken as a z-series.

Acknowledgements

The authors gratefully acknowledge the skilful technical assistance of Marie-Jeanne Crépeau Many thanks to Kian Hematy for phylogenetic analysis and critical reading of the results. The GWB3 and GWB4 Gateway binary vectors were kindly supplied by Tsuyoshi Nakagawa (Shimane University, Japan). Chris Hawes (Oxford Brookes University, UK) is thanked for providing the sialyltransferase–YFP construct, Paul Knox (Centre for Plant Sciences, University of Leeds) for the PAM1 antibody and Jorn Mikkelson at Danisco (http://www.danisco.com) for providing model lime pectins. Olivier Grandjean, Lionel Gissot and Jorunn Johansen are thanked for skilful tehnical advice on confocal microscopy. Owrda Saidi (MSc student) is thanked for technical help. Catherine Lilley and Paul Dupree are thanked for allowing us to cite their unpublished data. Funding was provided by The Swedish Foundation for Strategic Research to A.M., Formas grant 22.0/2003-1246 (C.E.) and the EEC, FP5 grant ‘EUROPECTIN’ to H.H. Malcolm Bennett is thanked for support to C.E. during her PhD research.

Ancillary