Arabidopsis GT34 family contains five xyloglucan α-1,6-xylosyltransferases


Author for correspondence:
Simon McQueen-Mason
Tel: +44 1904328775


  • The Arabidopsis genome includes seven family 34 glycosyltransferase (GT34) encoding genes. XXT1 and XXT2 have previously been shown to encode XyG α-1,6-xylosyltransferases, while knockout mutants of a third, XXT5, exhibit decreased XyG content, suggesting a similar activity. Here, we extend the study to the rest of the Arabidopsis GT34 genes in terms of biochemical activity and their roles in XyG biosynthesis.
  • The enzyme activity of XXTs was investigated using recombinant protein expressed in E. coli. XyG analysis of single and double T-DNA insertion knockouts, together with overexpression of GT34s in selected mutant lines, provided detailed function of each gene.
  • We reveal the activity of the third member of the GT34 gene family (XXT4) that exhibits xylosyltransferase activity. Double mutants for either xxt2 or xxt5 had a large impact on XyG content, structure and size distribution. Overexpression of the remaining member, XXT3, was able to restore XyG epitopes in xxt2, xxt5 and xxt2 xxt5 double knockouts, suggesting that it also encodes a protein with XXT activity.
  • Our work demonstrates that five of the seven Arabidopsis GT34 genes encode XXT enzymes.


Xyloglucans comprise the major hemicellulose polymers in the primary walls of most angiosperms with the exception of the grasses, where they are less abundant (Carpita & Gibeaut, 1993). Various observations in the past have led to the hypothesis that XyGs play a key role in growing plant cell walls by providing major crosslinking tethers between neighbouring cellulose microfibrils (Albersheim et al., 2010; Fry, 1989; Hayashi, 1989; Cosgrove, 1999). This model has recently been challenged by Cavalier et al. (2008), who suggested that the double insertion knockout plants lacking XyGs could grow fairly normally and the role of XyGs in plant growth and development is more subtle than previously thought. In support of this, Dick-Pérez et al. (2011) demonstrated that cellulose-XyG crosslinks do not occur as much as previously expected, and analysis of triple knockout mutants suggested that the load-bearing function of the plant cell wall relies on a single network of cellulose, hemicelluloses and pectin rather than a cellulose-XyG network, and that the lack of XyGs in the cell walls caused weakening of this network. Nevertheless, follow-up work by Park & Cosgrove (2012) demonstrated that, although not essential for survival, XyGs are indeed important for wall loosening by α-expansin or through acid-induced growth, in association with xylan and pectin polymers in the walls. Together with a number of enzymes that have been shown to modulate cell wall extensibility and thereby cell expansion by modifying XyG molecules (including expansins, McQueen-Mason et al., 1992; XyG endo-transglucosylases/hydrolases, Fry et al., 1992; and endo-glucanases, Park et al., 2003), this supports a biological role for XyGs in cell wall extension.

XyGs are complex polysaccharides comprising a β-1,4-d-glucan backbone decorated with α-1,6-xylose (Xyl) side chains. In most dicots, including Arabidopsis, these xylose side chains typically occur on contiguous blocks of three glucose residues separated by an unsubstituted glucose. Generally, seed storage XyGs are not further substituted, whereas those from growing plant cell walls are typically substituted with β-1,2-d-galactose linked to the second or third Xyl residues, and these galactosyl residues may be further substituted with α-1,2-l-fucose (Zablackis et al., 1995).

In Arabidopsis the characterization of XyG fucosyl and galactosyltransferases has been the subject of extensive study (Perrin et al., 1999; Vanzin et al., 2002; Madson et al., 2003). In addition, two XyG xylosyltransferases genes (XXT1 and XXT2) were first identified in Arabidopsis through the enzymatic activity of recombinant proteins (Faik et al., 2002; Cavalier & Keegstra, 2006). More recently, Cavalier et al. (2008) demonstrated that xxt1 knockout mutants showed barely detectable alterations in XyG content, and xxt2 mutants showed only modest decreases in XyG content and composition. These two genes exhibited very similar, generally ubiquitous expression patterns, which indicate the potential for functional redundancy. This is supported by dramatic reductions in XyG content in double mutants. Cavalier et al. (2008) reported that XXT1 and XXT2 are the prime enzymes required for XyG biosynthesis as double xxt1-1 xxt2-1 mutants contained no detectable XyGs. Mutations in a third XXT gene (XXT5) were also shown to have reduced XyG content (Zabotina et al., 2008).

Arabidopsis glycosyltransferase family GT34 comprises seven glycoyltransferases genes including XXT1, XXT2 and XXT5. The GT34 family also includes two genes with a high sequence similarity to the well-characterized α-1,6-galactosyltransferases (GMGT), which are involved in the synthesis of the side chains of galactomannan in fenugreek, lotus and coffee (Edwards et al., 1999, 2004; Faik et al., 2002; Préet al., 2008). Here we study this family as a whole and we demonstrate, using a combination of approaches, that five of them (XXT1-5) encode XXTs, and that the loss of function of these genes, alone or in combinations, has a range of impacts on XyG composition.

Materials and Methods

Plant materials and growth conditions

Arabidopsis thaliana (L.) Heynh seeds were germinated for 7 d on 1% (w/v) agar plates containing ½ strength Murashige and Skoog medium, and then grown in Levington’s compost under the same conditions: 16 h light (115 μmol m−2 s−1) at 25°C. Plant material for cell wall extractions and RNA preparation were harvested in the same way. Roots were collected from seedlings grown vertically on plates for 10 d. Hypocotyls were harvested from seedlings grown in the dark. Rosette leaves were collected from 3-wk-old plants. Cauline leaves, stems, siliques, buds, flowers and pedicels were collected from 4–5 wk-old plants. For seed and embryo samples, the seed coats were removed before further preparation.

Recombinant protein expression and purification in E. coli

Coding sequences without transmembrane domain were amplified by PCR and inserted into the pGEX-4T-3 vector (Amersham Bioscience) in-frame with GST tags. The numbers of 5′ nucleotides removed from the start codon were 147, 147, 213, 180 and 216 for XXT1-5 and 144 for both AtGT6 and 7 (see primers for truncated sequence amplifications in Supporting Information Table S1). Constructs were expressed in E. coli (BL21) using 75 ml 2 ×-YT media supplemented with ampicillin (50 μg ml−1) at 15°C with shaking at 150 rpm for 2 d. For induction, IPTG was added to a final concentration of 0.5 mM and incubated for a further 24 h. Cells were harvested and resuspended in 20 ml spheroblast buffer (750 mM sucrose, 200 mM Tris-HCl and 0.5 mM EDTA, pH 8.0). Lysozyme was added to a concentration of 1 mg ml−1 and incubated at 4°C with shaking for 30 min. Buffer was discarded and cells were disrupted by resuspension in 5 ml PBS buffer containing 0.2 mM PMSF. Cell debris was removed by centrifugation at 16 000 g for 10 min, and supernatants were incubated with 100 μl of 50% slurry of Glutathione Sepharose 4B (Amersham Bioscience) with shaking for 30 min. Sepharose beads were collected and washed three times with 10 ml PBS. The beads were resuspended in 50 μl elution buffer (20 mM reduced-form glutathione, 100 mM Tris-HCl, pH 8.0, and 120 mM NaCl) by shaking at 100 g for 10 min before collecting the supernatant by brief centrifugation. This elution step was repeated three times. Purified enzymes were quantified by Bradford assays with BSA standards and SDS-PAGE.

Recombinant GST-fusion protein activity assays and assay product detection using TLC

Expressed proteins were used in a form of immobilized enzymes with the glutathione beads. Approximately 0.3 μg of proteins were incubated at 30°C for 18 h in a reaction mixture containing: 50 mM PIPES, pH 6.9, 0.5 mM UDP-sugar (Xyl, Gal or Glu), 20 mM β-mercaptoethanol and 150 μg of oligo/polysaccharide acceptors; or cellohexaose, 5 mM MgCl2, 5 mM MnCl2 and 0.37 KBq of UDP-[14C] sugar (Xyl, Gal or Glu), made up to a final volume of 25 μl. The reactions were stopped by boiling at 100°C for 10 min. TLC analysis was performed using a method described by Robyt & Mukerjea (1994) and Atichokudomchai et al. (2006). Whatman TLC K5 chromatography (silica-gel) plates and a solvent system containing acetonitrile, 1-propanol, water and ethyl acetate at the ratio 85 : 50 : 50 : 25 by volume were used for this analysis. Reactions (5 μl) were spotted on the plates 1.5 cm from the bottom and were developed for 2 h before being dried. For radioactive detection, TLC plates were exposed to a phosphor screen overnight and read using a Phosphorimager (Bio-Rad). Signal intensity was converted to quantity by calculating against UDP-[14C]xylose standard curves. For colorimetric detection, TLC plates were dipped into ethanol solution containing 0.1% (w/v) α-napthol and 5% (v/v) sulphuric acid, and dried before incubation at 120°C for 10 min.

Promoter:GUS fusion analysis and RT-PCR

Promoter:GUS fusion lines were generated using c. 1.5 kb of the 5′-upstream regions of XXT genes. The region was amplified from BAC clones by PCR (see Table S2 for primers), and inserted into the pHGWFS7 vector (Karimi et al., 2002). Arabidopsis (Col-0) was transformed by floral dipping with Agrobacterium tumefaciens (GV3101) harbouring the constructs. Transformants were selected for hygromycin resistance (35 μg ml−1) and at least 10 transformants for each construct were examined. GUS activity was detected by infiltrating plant tissues with staining solution (0.5 mM K3Fe(CN)6, 0.5 mM K4Fe(CN)6, 60 mM Na2HPO4, 30 mM NaH2PO4, 0.1% (w/v) Triton-X100 and 0.05% (w/v) X-Gluc) and incubation at 37°C overnight. For RT-PCR, RNA from various Arabidopsis tissues was treated with DNase I, and cDNA was synthesized from 1 μg of total RNA using Superscript™ II RNase H Reverse Transcriptase (Invitrogen) with oligo-dT primer. PCR amplifications of the purified RNA were performed to ensure no gDNA contamination. PCR reactions were performed with 1 μl of cDNA, Taq DNA polymerase with supplied buffer, 500 μM of dNTPs and 5 μM of each primer pair (Table S3), under conditions as follows: 94°C for 2 min, 30 cycles of (94°C for 30 s, 64–55°C for 30 s and 72°C for 30 s) and 72°C for 2 min. The annealing temperatures for XXT1-5 and AtGT6-7 were 64, 60, 55, 57, 65, respectively, and 55, 55°C, respectively. Genomic DNA was used as a positive control, and 18s RNA primers were used to indicate cDNA quality.

Selection of homozygous T-DNA insertion lines

T-DNA insertion lines were obtained from two sources: The Salk Institute Genomic Analysis Laboratory and GABI-KAT. Homozygous plants of each line were identified by PCR with gene-specific primers and border primers (see primer locations in Fig. S2 and primer sequences in Table S4). All homozygous lines were tested by RT-PCR using cDNA synthesized from RNA extracts of 7 d-old seedlings to confirm the complete disruption of the genes (see primers in Table S5). Crosses were made between homozygous lines and successful crosses identified by PCR using DNA from F1 plants before screening for double homozygous plants in the F2 population.

Immunolabelling studies

Stem tissue was hand-sectioned and fixed in 4% (w/v) paraformaldehyde (PFA) in 100 mM PIPES, pH 6.9, before further labelling. Seed samples were fixed with 1% (w/v) glutaraldehyde, 2% (w/v) PFA in 100 mM sodium cacodylate, pH 7.0, before embedding using a progressive low temperature method (VandenBosch, 1991). Sections were blocked with 3% (w/v) BSA in 100 mM PIPES buffer, pH 6.9, before incubation in primary antibody: CCRC-M1 (University of Georgia, Athens) or LM15 (XyG backbone; gift from Professor Paul Knox). Antibodies were used at 1 : 10 concentration diluted in 1% (w/v) BSA in buffer and incubated for 1 h at room temperature. After washing three times briefly and 3 × 5 min with PIPES buffer, sections were incubated in secondary antibody for 1 h at room temperature, in the dark. Secondary antibodies (anti-mouse and anti-rat conjugated to FITC; Sigma) were used at a 1 : 40 dilution in 1% BSA in PIPES buffer. Sections were washed again before viewing using a Nikon FXA microscope. For seeds, labelling was visualized by confocal microscopy (Zeiss) and lambda studies were performed to identify autofluorescence.

Xyloglucan oligosaccharide analysis

Cell walls were extracted from stems (with siliques removed) that were ground in liquid nitrogen and washed with absolute ethanol at least three times. To fractionate cell wall materials, dried samples were resuspended with 50 mM CDTA, pH 7.0, overnight at room temperature to extract a pectin-rich fraction. For hemicellulose extraction, the remaining insoluble cell wall material was incubated further in 4 M NaOH containing 1% sodium borohydride overnight, and then neutralized with glacial acetic acid, dialysed against water and dried. Hemicellulose extracts were dissolved to a concentration of 1.5 mg ml−1 in 100 mM NH4OAc, pH 5.0, and incubated with 10 μg of Trichoderma XG5 xyloglucan endo-glucanase (a gift from Novozymes) overnight at room temperature. The reactions were stopped by boiling, and unhydrolysed polysaccharides were precipitated with 80% ethanol. Oligosaccharides were separated by HPAEC using a Dionex CarboPac PA-20 column with a linear gradient over 40 min starting with 100 mM NaOH and ending with 100 mM NaOAc and 100 mM NaOH at a flow rate of 0.5 ml min−1, and detected using a pulsed amperometric detector.

Gel permeation chromatography and iodine quantification

One milligram of hemicellulose extracts was separated on a Superose 6L column using a Smart chromatography system (Pharmarcia, Stockholm, Sweden) using water as the mobile phase at 40 μl min−1 flow rate, collecting 80 μl fractions. Colorimetric detection with iodine was employed to quantify XyGs (Kooiman, 1960). For XyG quantification, cell wall extracts were de-starched by hydrolysis with 10 units of thermo-stable α-amylase (Sigma) at 100°C for 15 min. Following this, the extracts were precipitated and washed using 80% ethanol before resuspension and assay with iodine. Hemicellulose extracts from stems were assayed (40 μl of 300 μg extracts in 112 μl of 20% Na2SO4 and 28 μl of Lugol’s solution) and OD600 nm measured and quantified against standard curves generated with tamarind XyGs (Dainippon Pharmaceutical Co, Osaka, Japan). The quantities obtained were expressed on mg g−1 DW stem and then normalized in comparison to values from wild-type stems.

Complementation studies

Overexpression cassettes for each of the seven GT34 genes were constructed with the pH2GW7 vector (Karimi et al., 2002). xxt2-1, xxt5 and xxt2-1 xxt5 double mutant plants were transformed by floral dipping. Transformants were selected for hygromycin resistance (35 μg ml−1). At least 10 selected transformants from each transformation were examined by immunolabelling using CCRC-M1 in stem sections.

Gene numbers

The gene numbers of the GT34 genes are: At3g62720 (XXT1), At4g02500 (XXT2), At1g74380 (XXT5), At1g18690 (XXT4), At5g07720 (XXT3), At4g37690 (AtGT6), At2g22900 (AtGT7).


Recombinant proteins encoded by three Arabidopsis GT34 genes can xylosylate cello-oligosaccharides

Previous work using recombinant protein produced in Pichia pastoris and insect cells revealed that XXT1 and XXT2 can catalyse the transfer of xylose (Xyl) from UDP-Xyl to cellopentaose and cellohexaose (Faik et al., 2002; Cavalier & Keegstra, 2006). In our hands, it proved difficult to produce detectable levels of active protein in either P. pastoris or in insect cells, with activity having to be determined in whole cell lysates. By contrast, we found that active protein was more reliably produced by expressing truncated proteins (lacking the N-terminal transmembrane domain) in E. coli as GST fusion proteins. All seven Arabidopsis GT34 proteins expressed in this way were purified as shown in Fig. 1(a). A 62 kDa protein band that co-purifies is likely to be the bacterial chaperonin GroEL (Hou et al., 2004), which is frequently observed after GST purification. Quantification using Bradford assays revealed that the yield of the purified soluble protein was in the range of 3–7 μg per 75 ml culture.

Figure 1.

Expression of Arabidopsis GT34 family members as GST-fusion proteins in Escherichia coli and analysis of enzyme reaction products. (a) SDS-PAGE separation of GST-Sepharose purified GT34 family proteins-GST fusion produced in E. coli cells (upper band) and a co-purifying bacterial chaperonin (lower band). (b) α-napthol staining of TLC separation of enzyme reaction products (left panel). Autoradiograph obtained from the same TLC separation (right panel) The asterisk indicates the position of products generated by incubation with UDP-[14C]xylose (two sharp wavy lines) and cellohexaose as acceptor for GST-XXT1, GST-XXT2 and GST-XXT4.

The activity of the purified enzymes was assayed using UDP-[14C]Xyl and cellohexaose as substrates, as previously reported (Cavalier & Keegstra, 2006), except that the proteins were immobilized on glutathione beads, and the radioactive products were analysed by TLC and autoradiography. From the seven proteins, xylosyltransferase activity was observed for XXT1 and XXT2 as reported previously (Cavalier & Keegstra, 2006), and also for AtGT4 (Fig. 1b), now referred to as XXT4. The measurement of the radiolabelled product by densitometry of the TLC plates indicated that XXT1 and XXT2 showed strong activity (c. 88% and 66% of radiolabelled substrate incorporated into cellohexaose, respectively), with XXT1 consistently higher than XXT2, while XXT4 activity was lower (c. 40% incorporation). Product analysis of XXT4 activity using nonradiolabelled UDP-Xyl by mass spectroscopy showed that XXT4 could incorporate up to three Xyl moieties to cellohexaose acceptors (Fig. S1).

No activity with UDP-Xyl was observed from the other four enzymes although a range of acceptors, including cellopentaose, cellohexaose, tamarind-XyGs, de-galactosylated tamarind-XyGs and hemicellulose extracts from xxt2-1 and xxt5 mutants were tried. We also tested these enzymes with various UDP-[14C]sugars and monosaccharide acceptors (single sugar assays) as suggested by Egelund et al. (2006), in all combinations based on cell wall polysaccharide linkages, but no incorporated products were observed (Table S6).

Expression analysis indicates the potential for functional redundancy amongst XXT genes

With at least four GT34s encoding XXTs, there is a clear potential for functional redundancy amongst family members. To examine this, we characterized the gene expression patterns of family members using promoter:GUS fusions and RT-PCR. The GUS expression results presented in Fig. 2 are representative of those seen in at least 10 transformants obtained from each gene. We provide both an overall view of the expression pattern of whole seedlings and images of specific tissues where GUS expression was observed for each gene. XXT1, XXT2 and XXT5 showed similar expression patterns (Fig. 2a,b,d). All three promoters drove GUS expression in most tissues including leaves, roots and stems. GUS expression was particularly strong in actively growing tissues. The three genes showed subtle variations in expression in flowers and developing seed and these are presented in Fig. 2. During embryo development, XXT1:GUS expression was very low at early stages and localized to the seed coat and endosperm, but became more apparent as the embryo reached maturity, while XXT2:GUS expression was strong in early developing embryos, declined with maturation, and was absent from the seed coat and endosperm. By contrast, XXT5:GUS expression was strong throughout seed development and was present in the seed coat and embryo. RT-PCR confirmed the rather ubiquitous expression patterns of these three genes (Fig. 2e). By contrast, XXT4:GUS expression was observed only in the rosette, pedicel and the vascular tissues of the stem (Fig. 2c). Conversely, no XXT3:GUS expression was detected, even though this GUS expression line was generated in the same way as others. RT-PCR indicated that XXT3 is expressed in several specific tissues, including stem 2 and floral buds (Fig. 2e).

Figure 2.

Expression of XXT genes revealed by staining of Arabidopsis plants transformed with promoter-GUS fusions and by RT-PCR studies. (a) XXT1: i, 14-d-old seedling; ii, seed coat; iii, mature embryo; iv, silique; v, flower; vi, stem; vii, root. Bars: i, 4 mm; ii, iii, vi, 0.1 mm; iv, v, vii, 1 mm. (b) XXT2: i, 14-d-old seedling; ii, globular stage embryo; iii, mature embryo; iv, leaf; v, flower; vi, stem; vii, silique; viii, root; ix, root. Bars: i, 4mm; ii, iii, vi–viii, 0.1 mm; iv, 10 mm; v, ix, 1 mm. (c) XXT4: i, 14-d-old seedling; ii, rosette; iii, stem. Bars: i, 4 mm; ii, 2 mm; iii, 0.2 mm. (d) XXT5: i, 14-d-old seedling; ii, globular stage embryo; iii, torpedo stage embryo; iv, mature embryo; v, seed coat; vi, flower; vii, silique; viii, stem; ix, root; x, root tip. Bars: i, 4 mm; ii, 50 μm; iii–v, 0.2 mm; vi, 1 mm; vii–ix, 0.1 mm; x, 0.3 mm. (e) Reverse transcriptase-PCR analysis of transcript abundance for XXT genes in developing seeds and mature tissues of Arabidopsis. DAF, days after flowering.

Characterization of T-DNA insertion mutants of the Arabidopsis XXT genes

Although the analysis of XXT1, XXT2 and XXT5 insertion lines has been reported previously by Cavalier et al. (2008) and Zabotina et al. (2008), we characterized additional single and double mutants generated in this study to allow extensive comparisons of the effects of gene knockout of XXT genes. The T-DNA insertion positions of these lines are shown in Fig. S2. Homozygous insertion lines of the five XXT genes were identified by PCR, with RT-PCR to confirm them as null alleles (Fig. S3). We were unable to obtain a true null allele of XXT3 as the only available line with an insertion targeting this gene appeared to produce a normal XXT3 transcript, as determined by RT-PCR. Note that the xxt1 mutants used here, designated xxt1-2 and xxt1-3, are allelic to the xxt1-1 mutant used by Cavalier et al. (2008). No obvious alterations in growth or development were observed in any of the confirmed mutants. Further observation by profiling the monosaccharide content of cell walls revealed that xxt2-1, xxt2-2 and xxt5 had reduced levels of Xyl in cell walls in a range of tissues, with the most dramatic reductions in stems, embryos and seedlings (data not shown). These, however, agreed with previous work reported by Cavalier et al. (2008) and Zabotina et al. (2008). Because of the potential for genetic redundancy among XXT genes, we generated double mutants by crossing these lines (Fig. 3). Double homozygous lines were readily recovered from F2 generations at the 1 : 3 expected ratio (see Fig. S4 for genotypic identifications).

Figure 3.

Double T-DNA insertion mutants generated by crossing. Phylogenetic trees were constructed using predicted protein sequences generated in ClustalW (European Bioinformatics Initiative, Hinxton, UK). Double mutants are listed on the right.

xxt knockout lines show alteration in XyG immunolabelling

Immunolabelling of the xxt mutants using antibodies recognising XyGs and galactomannans revealed that both the xxt2-1 and xxt5 mutants showed a reduction in the labelling of stem and embryo tissues with CCRC-M1 antibody (which recognizes fucosylated-XyGs; Puhlmann et al., 1994) and with LM15 antibody (which binds to the XyG backbone; Marcus et al., 2008), when compared to the wild-type (Fig. 4). In the wild-type, both antibodies show a similar labelling pattern (albeit with slightly less intensity for the LM15) where specific labelling is observed at the cell walls of phloem, vascular cambium and primary xylem tissues in stems and throughout the cell walls of embryos (Fig. 4). By contrast, in stem sections of xxt2-1 and xxt5 mutants, neither antibody recognized the phloem nor the vascular cambium, and labelling in the primary xylem was markedly reduced (Fig. 4). Immunolabelling of xxt1-2 xxt5 and xxt2-1 xxt5 double mutants revealed a further reduction in XyG epitopes in both stem and embryo sections, with a complete absence of detectable labelling in primary xylem, as compared to the reduced level seen in the xxt2 and xxt5 single mutants. Similarly, little or no labelling was detectable in the cell walls of the double xxt1-2 xxt5 and xxt2-1 xxt5 mutant embryos. In the xxt1 and xxt5 mutants, immunolabelling with either antibody resulted, predominantly, in cytoplasmic fluorescence, rather than labelling at the cell wall. When comparing the fluorescence between xxt5 and double xxt1-2 xxt5 mutants, the marker was most evident at the cell walls of xxt5 in certain areas of the embryo sections in a patchy fashion for both antibodies. The labelling apparent in the cytoplasm using both antibodies could indicate the presence of nonsecreted XyGs in the embryos or could be a nonspecific signal. The xxt4 mutants were indistinguishable from wild-type. Similarly, xxt4 xxt2, and xxt4 xxt5 double mutants were indistinguishable from xxt2 and xxt5 single mutants (data not shown).

Figure 4.

Immunofluorescence labelling of stem and embryo sections of Arabidopsisxxt mutants using xyloglucan-specific antibodies (CCRC-M1 and LM15). Labelling is evident in phloem (Ph), vascular cambium (VC) and primary xylem (1˚ xy), whilst autofluorescence (probably due to lignin) is present in the secondary xylem. Negative controls (no primary antibody) of a transverse stem section (bar, 50 μm) and a longitudinal section through a mature embryo (bar, 100 μm) incubated with secondary antibodies are shown.

Extractable XyG content in xxt mutants

Iodine staining has previously been shown to be specific for XyG in cell wall extracts (Kooiman, 1960) and has been employed for quantifications of XyGs from a range of plant species (Hsu & Reeves, 1967; Gould et al., 1971). Thus, we used this method to quantify the amount of XyG in de-starched alkaline extracts from mutant and wild-type stem cell walls. This revealed that extractable XyG content is reduced by 32% and 50% in xxt2-1 and xxt5 respectively, compared to wild-type (Fig. 5). In double mutants for xxt2-1 × xxt5 and xxt1-2 × xxt5, XyG content was reduced further to levels < 40% of the wild-type. Although a slight reduction was observed in the xxt1-2 mutant, no detectable decrease was observed in xxt4, and the crossing of xxt4 into the other xxt backgrounds had no apparent additive effect on XyG content.

Figure 5.

Extractable xyloglucan (XyG) content of Arabidopsis stem xxt mutants quantified by iodine staining. XyG was extracted from stem cell wall material by 4 M NaOH, neutralized and treated with α-amylase, before being measured colorimetrically by iodine staining. Data are averages from three separate extracts, each extract measured in duplicate, and are presented as percentage of the content of Col-0, with SE bars shown.

Analysis of XyG oligosaccharides from cell walls of xxt mutants

Profiling of the oligosaccharides generated by XyG-endoglucanase digestion and HPAEC-PAD revealed a complexity of changes in the side-chain distributions of XyG in the various xxt mutants. The result showed that the ratios of oligosaccharide products of digestion were altered in most mutants (Fig. 6), but as in other analyses, the xxt1-2 was indistinguishable from wild-type. By contrast single mutants of xxt2-1 and xxt4 both exhibited a similar change in the ratio of oligosaccharides compared to wild-type, in which the relative levels of XXXG were reduced. Concomitant increases in levels of XXFG and XLXG are observed in these lines. A similar pattern was apparent in the xxt2-1 xxt4 double mutant. By contrast, the changes in oligosaccharide profiles from the cell walls of the xxt1-2 xxt5 double mutants revealed a dramatic alteration whereby the relative levels of all galactosylated and fucosylated oligosaccharides were considerably lower with much higher levels of XXXG.

Figure 6.

Xyloglucan oligosaccharides generated from stem hemicellulose extracts of xxt mutants with xyloglucan-specific endoglucanase. The relative abundance of each oligosaccharide (in percentage) is derived from integrated fragment signals of the pulsed amperometric detection of oligosaccharide peaks separated by HPAEC. The data represent averages obtained from three separate extracts and digests, each of which was analysed twice, with SE bars shown. Abbreviations in the model structure represent: G, β-d-Glcp-(1→4); X, α-d-Xylp-(1→6); L, β-d-Galp-(1→2); F, α-l-Fucp-(1→2).

Molecular mass distribution of XyG in xxt mutants

We used gel permeation chromatography (GPC) of de-starched alkaline extracts detected by iodine staining to look for changes in the molecular mass distribution of XyGs from stem cell walls of the various mutants (Fig. 7). As well as changes in XyG -epitopes and content, there are also changes in the sizes of extractable XyGs in some of the mutants. Although the XyG profile from xxt1-2 mutants was indistinguishable from wild-type, the profiles from xxt2-1 and xxt5 were clearly altered from the wild-type and both showed a marked decrease in higher molecular mass components, compared to the wild-type profile. The profile for xxt4 was also notably different from wild-type with a substantial decrease in high molecular mass XyGs and a peak in the mid-range rather than at the low molecular mass end as seen in xxt2-1 and xxt5. The profile of XyGs from xxt2-1 xxt5 double mutants showed a further loss of mid-range and high molecular mass XyGs compared to either single mutant. In spite of the xxt1-2 profile being unchanged, the profile from the xxt1-2 xxt5 double mutant revealed even more dramatic changes than the xxt5 single mutant profile, and generally showed the most dramatic differences compared to wild-type, with an even stronger effect than that seen in the xxt2-1 xxt5 double mutant.

Figure 7.

Size exclusion chromatography profile of alkali-extracted xyloglucans (XyGs) from the stem of Arabidopsisxxt mutants. Stem alkali extracted XyGs were separated by size exclusion chromatography and detected by iodine staining of the eluted fractions. Results are the averages of samples from three separate extractions, with duplicate analysis.

Complementation of immunolabelling phenotypes

The decrease in labelling with XyG antibodies in xxt2-1 and xxt5 mutants could be fully complemented by overexpressing wild-type copies of the respective genes, as illustrated in Fig. 8(a). As well as confirming the role of these two genes, this enabled us to examine the capacity of other members of the GT34s (including AtGT6 and AtGT7) to complement these two mutants. We classified the observed complementation relative to that of XXT2 and XXT5 expressed in their respective mutant background (Fig. 8). XXT1 overexpression demonstrated a strong complementation in xxt2-1, xxt5 and xxt2-1 xxt5 mutant backgrounds, whereas complementation by XXT2 was strong in the xxt2-1 background but weak in xxt5 and the xxt2-1 xxt5 double mutant backgrounds (Table 1). Interestingly, XXT3 strongly complemented in all three mutant backgrounds, suggesting that this gene does indeed encode an active protein for XyG biosynthesis. By contrast, complementation of the two single mutants by XXT4 was weak and did not generate any detectable epitope in the xxt2-1 xxt5 double mutant background. The overexpression of XXT5 gave strong complementation in the two single mutants but only weak complementation in the double mutant. The complementary effects of two other GT34 family proteins, AtGT6 and AtGT7, were also tested against the single and double mutants, but neither protein could restore any detectable XyG epitopes. This suggests that the biochemical activity of the gene products of AtGT6 and AtGT7 is unlikely to be XXT.

Figure 8.

Immunofluoresence labelling of stem sections using xyloglucan (XyG)-specific antibodies of Arabidopsisxxt mutants expressing 35S:GT34 genes for complementation. Overexpression constructs for GT34 genes were transformed into various xxt2 or xxt5 mutant backgrounds to examine their ability to compensate for the loss of XyG epitopes in sections from these plants. The three panels show representative immunofluorescence level used to rank for (a) high, (b) low, or (c) no complementation as presented in Table 1. Bar, 100 μm.

Table 1.   Levels of compensation in relative abundance of XyG epitopes in stem sections of Arabidopsis mutant lines transformed with seven 35S:GT34 constructs compared to corresponding untransformed line
ConstructsLevel of compensation in background genotype
xxt2-1xxt5xxt2-1 xxt5
  1. ND, not determined.



Previous work has shown that three of the seven members of Arabidopsis GT34s encode XXTs (Cavalier & Keegstra, 2006; Cavalier et al., 2008; Zabotina et al., 2008). Here we demonstrate, using biochemical assays, that a fourth member (XXT4) encodes an XXT. Overexpression studies indicate that the final member, XXT3, is also an XXT. Expression studies revealed that XXT1, XXT2 and XXT5 are all expressed in most organs and at most stages of Arabidopsis development, while XXT3 and XXT4 show more specialized expression patterns. Examination of knockout lines suggests that the loss of XXT1 action has little discernable impact on XyG content, whereas xxt2-1 and xxt5 mutants show clear reductions in XyG content and changes in extractable XyG polymer size, as well as changes in side-chain patterns. In line with this, double xxt2-1 xxt5 mutants have even lower XyG levels than the single mutants, but they still contain detectable XyG and show relatively normal growth. We obtained viable xxt1-2 xxt5 double mutants, which showed marked decreases in XyG content compared to the xxt5 single mutant, despite the fact that there was little difference between xxt1-2 single mutant and wild-type. This study indicates that XXT1, XXT2 and XXT5 are the major genes responsible for XyG biosynthesis in Arabidopsis with XXT3 and XXT4 playing more restricted roles.

Previous work using P. pastoris and insect cells as expression systems revealed that recombinant XXT1 and XXT2 had XXT activity, but no activity was detected for other family members (Faik et al., 2002; Cavalier & Keegstra, 2006). For our studies we used E. coli as an expression host, which enabled the demonstration of XXT activity for XXT1, 2 and 4. Using this system, we successfully produced soluble protein for all seven GT34 members by removing the coding sequence for the predicted transmembrane domain from the N-terminus, and we suggest that this may prove a fruitful system for others working with similar enzymes. However, it is not clear why activity was not obtained for XXT3 and 5. This may reflect the complexity of XyG biosynthesis, with these enzymes perhaps only transferring Xyl to partially xylosylated substrates or requiring other specific patterns of substrate. Alternatively, these proteins may function in complexes in plant cells, which are not present in the purified recombinant proteins. Nevertheless, the reduction in XyG content in xxt5 mutants and the complementation by XXT3 and XXT5 overexpression of the loss of XyG in xxt2-1, xxt5 and xxt2-1 xxt5 mutant cell walls supports the case that these genes encode XXTs.

Our results demonstrate that, of seven Arabidopsis GT34 genes, there are at least five genes that encode XXT enzymes and a further two encoding enzymes of unknown function (AtGT6 and AtGT7). This is in accord with the proposal by Faik et al. (2002) who compared the protein sequences of Arabidopsis GT34s and found them to fall into two clusters: five were grouped with those with demonstrated XXT activity and other two grouped with known galactomannan galactosyltransferase (GMGT) from fenugreek (Edwards et al., 1999). We examined whether AtGT6 and AtGT7 could encode XXT enzymes by assaying the activity of recombinant enzymes, and by overexpressing the genes in xxt mutant lines to see if they could complement these lesions. Neither of these approaches indicated that AtGT6 and AtGT7 encode XXTs, and this, taken in conjunction with their closer similarity to known GMGTs, suggests that they are more likely to encode GMGTs but does not preclude the possibility that they might encode XXT enzymes.

The physiological role of XXT1 appears complex. Loss of function xxt1-2 mutants exhibited no detectable changes in XyG content despite the fact that this gene encodes the most active of the assayed recombinant XXT proteins and is expressed at high levels in most tissues of the plant. This is in line with the report by Cavalier et al. (2008) and analyses of xxt1-3 mutants (data not shown). Moreover, XXT1 can complement the loss of XXT2, XXT5 and both genes in double xxt2-1 xxt5 mutants. In addition, the loss of XXT1 function in either an xxt2, or xxt5 background has a clear effect on XyG content. In contrast, the xxt1-2 xxt5 double mutants grow relatively normally and show a similar reduction in XyG content to that seen in the xxt2-1 xxt5 double mutant. Interestingly, the extractable XyG Mr distribution from the xxt1-2 xxt5 double mutant is substantially different to that of the xxt2-1 xxt5 or any of the other mutants. In all of the mutants with decreased XyG content, the most substantial decreases are associated with the higher Mr polymers, with the lower Mr end of the chromatograms remaining relatively unaffected, whereas the xxt1-2 xxt5 double mutant shows reduced amounts of XyG across all the different sizes of polymer.

It is important to note that the Mr distribution of XyGs is modulated by the action of XyG modifying enzymes such as XyG transglycosylase/hydrolases (XTHs), endoglucanases (Catala et al., 1997) and α-xylosidases (Sampedro et al., 2010; Günl & Pauly, 2011) in the cell walls. Thompson et al. (1997) reported that XyGs are first synthesized as relatively low Mr polymers before being incorporated into much higher Mr polymers in the walls. Therefore, the alterations of XyG Mr observed in the mutants may not be purely the direct result of altered XyG biosynthesis, but may also be a consequence of changes in the susceptibility of the XyG to these activities. In this context it is interesting to note that altered extractable XyG Mr profiles were the only obvious change that we could find in the xxt4 mutant, suggesting perhaps that a subtle change in XyG structure might have implications for integration of the polymer into cell walls.

Changes in XyG side-chain distributions also reflect the effect of each XXT knockout on xylose side-chain substitution. The mutants all showed a similar trend in that XXXG oligosaccharides were greatly reduced, while the relative abundance of oligosaccharides with galactosyl and fucosyl residues was increased. The reduced amount and the decreased relative molecular mass of XyG in the mutants suggest that the action of the xylosyltransferases is required for glucan chain elongation. It is also possible that the increased representation of galactosylated and fucosylated oligosaccharides in the XyG profiles indicates slow xylosylation in the xxt mutants. The reduction in galactosyl and fucosyl residues in XyG from the xxt1-2 xxt5 could indicate that the early forms of XyG produced after xylosylation are not appropriate for further galacto- or fucosylation.

Both XXT3 and XXT4 appear to play restricted roles in XyG biosynthesis, and thus it seems likely that XXT1, 2 and 5 are responsible for producing most of the XyG in the major tissues in Arabidopsis. If we accept this general premise, then the presence of XyG in xxt1-2 xxt5 and xxt2-1 xxt5 double mutants is likely to be produced by XXT2 or XXT1 working alone in the major tissues of the plants. Given that XyG in the mutants contains blocks of 3 Xyl residues both XXT1 and XXT2 may, individually, be capable of completing the pattern of blocks of three xylose side chains in XyGs. By contrast, it is striking that XXT5, which is expressed fairly ubiquitously and apparently at relatively high levels and is able to compensate for the loss of the epitopes in all the tested mutants, is unable to complement the loss of function of both XXT1 and XXT2, as shown by Cavalier et al. (2008). This agrees with the suggestion by Cavalier et al. (2008) that XXT1 and XXT2 may be required for the chain initiation of XyG biosynthesis and that XXT5 may lack this ability or have it at very low efficiency.

A possible explanation for XXT3 and XXT5 lacking XXT enzymatic activity but being able to complement the loss of XyG epitopes in the mutants is that these two enzymes may function as a part of an enzyme complex and thereby enhance the transferase activity. In a similar report Rautengarten et al. (2011), demonstrated that the UDP-Ara mutase protein family includes proteins with and without mutase activity, interacting in a protein complex.

In vivo, XyG xylosylation seems likely to occur in close spatial and temporal proximity to elongating glucan chains. This is because of the insolubility of ß-1,4-glucans, which require xylosylation before more than five or six contiguous naked glucose residues are produced. Thus, it seems likely that the XXTs will work in close proximity to the glucan synthase in the Golgi. Partially xylosylated XyG could be the substrate for some XXTs, hence explaining the low activity when short cello-oligomers are used. Studies of the interactions between the glucan synthase and side-chain glycosyltransferases may enable us to unravel the apparent complexity of this process.


This work was supported by funds from The Garfield Weston Foundation, Kasetsart University Research and Development Institute, and The Thailand Research Fund (MRG5480289). L.D.G. was supported by funds from the European Commission Framework Programme 7 through Grant Agreement Number 211982, project RENEWALL. We are grateful to Professor Paul Knox (University of Leeds) for the donation of antibodies and Dr Kirk Schnoor (Novozymes) for the XyG-specific endoglucanase.