A β–glucuronosyltransferase from Arabidopsis thaliana involved in biosynthesis of type II arabinogalactan has a role in cell elongation during seedling growth



We have characterized a β–glucuronosyltransferase (AtGlcAT14A) from Arabidopsis thaliana that is involved in the biosynthesis of type II arabinogalactan (AG). This enzyme belongs to the Carbohydrate Active Enzyme database glycosyltransferase family 14 (GT14). The protein was localized to the Golgi apparatus when transiently expressed in Nicotiana benthamiana. The soluble catalytic domain expressed in Pichia pastoris transferred glucuronic acid (GlcA) to β–1,6–galactooligosaccharides with degrees of polymerization (DP) ranging from 3–11, and to β–1,3–galactooligosaccharides of DP5 and 7, indicating that the enzyme is a glucuronosyltransferase that modifies both the β–1,6- and β–1,3-galactan present in type II AG. Two allelic T–DNA insertion mutant lines showed 20–35% enhanced cell elongation during seedling growth compared to wild-type. Analyses of AG isolated from the mutants revealed a reduction of GlcA substitution on Gal–β–1,6–Gal and β–1,3–Gal, indicating an in vivo role of AtGlcAT14A in synthesis of those structures in type II AG. Moreover, a relative increase in the levels of 3-, 6- and 3,6-linked galactose (Gal) and reduced levels of 3-, 2- and 2,5-linked arabinose (Ara) were seen, suggesting that the mutation in AtGlcAT14A results in a relative increase of the longer and branched β–1,3- and β–1,6-galactans. This increase of galactosylation in the mutants is most likely caused by increased availability of the O6 position of Gal, which is a shared acceptor site for AtGlcAT14A and galactosyltransferases in synthesis of type II AG, and thus addition of GlcA may terminate Gal chain extension. We discuss a role for the glucuronosyltransferase in the biosynthesis of type II AG, with a biological role during seedling growth.


The arabinogalactan proteins (AGPs, AG proteins) belong to a highly diverse class of glycoproteins present on cell surfaces of plants (Seifert and Roberts, 2007; Ellis et al., 2010; Tan et al., 2012). AGPs consist mainly of glycans (>90% w/w), and are synthesized by post-translational modification in the secretory pathway. The importance of the carbohydrate moieties of AGPs in various cellular processes has been reported, including somatic embryogenesis, cell–cell interactions and cell elongation (Seifert and Roberts, 2007). Most of these studies involved use of monoclonal antibodies raised against the AG polysaccharides, and a temporal and spatial appearance of specific AG epitopes during development has been reported. However, as the precise epitope structure for most of the antibodies is not known, the molecular function of the AG glycan structures remains to be determined. The appearance of different AG glycans is probably controlled by a glycosylation process rather than redundancy of protein cores, as a developmentally controlled pattern of glycosylation was observed on a single synthetic peptide expressed in Arabidopsis (Estevez et al., 2006). Therefore, elucidation of the enzymes involved in the biosynthesis of AG glycans is expected to facilitate our understanding of the function of AGPs.

The structure of AG polysaccharides is very heterogeneous even on a single peptide (Estevez et al., 2006; Tan et al., 2010), but commonly consists of a β–1,3-galactan backbone with substitution at the O6 position with β–1,6-galactan side chains (type II AG, Tan et al., 2012); a model structure is shown in Figure 3(a). The side chains are typically further substituted by arabinose (Ara) and less frequently with other sugars such as glucuronic acid or 4–O–methyl glucuronic acid (collectively referred to as GlcA), rhamnose (Rha) and fucose (Fuc) (Tsumuraya et al., 1988; Tan et al., 2010; Tryfona et al., 2010, 2012). The glycosylation of AGPs is catalyzed by glycosyltransferases (GTs) that are located mainly in the Golgi apparatus. GTs act in a regio- and stereo-specific manner (Lairson et al., 2008), and it is expected that at least ten functionally distinct GTs are required for the biosynthesis of type II AG. So far, two fucosyltransferases (AtFUT4 and AtFUT6; Wu et al., 2010), two galactosyltransferases (AtGALT2; Basu et al., 2013; AtGALT31A; Geshi et al., 2013) and a putative arabinosyltransferase (AtRAY1; Gille et al., 2013) from Arabidopsis have been reported in the AG glycosylation pathway. AG fucosyltransferase activity was demonstrated by gain of function of fucosylated AGs by heterologous expression of Arabidopsis AtFUT4 and AtFUT6 in tobacco BY2 cells (Wu et al., 2010). Fucose on AGPs is important for root development (Van Hengel and Roberts, 2002), but a role for AtFUT4 and AtFUT6 in vivo remains to be elucidated. Galactosyltransferase (GalT) activity towards hydroxyproline in the synthetic peptide was demonstrated for Arabidopsis AtGALT2 expressed in Pichia pastoris (Basu et al., 2013). The atgalt2 mutants demonstrated lower GalT activity and a reduced level of β–galactosyl Yariv-precipited AGPs, but no apparent morphological phenotype was reported (Basu et al., 2013). The GalT activity in elongation of the β1,6-galactan side chains of AG was demonstrated by Arabidopsis AtGALT31A expressed in Escherichia coli and Nicotiana benthamiana (Geshi et al., 2013). A mutation in AtGALT31A caused aberrant asymmetric formative divisions in the hypophysis during embryogenesis, and embryo development was arrested at the globular stage, indicating an essential role for AG glycan in the normal development of embryo (Geshi et al., 2013). Although GT activity of RAY1 has not yet been demonstrated, mutations in this GT family 77 gene led to an extensive reduction in Ara moieties in beta-glucosyl Yariv-precipited AGPs (Gille et al., 2013). This alteration in arabinosylation leads to shortened primary roots, but lateral root growth is not affected.

We have characterized an Arabidopsis GT encoded by At5g39990 that belongs to the GT14 family in the Carbohydrate Active Enzyme database (CAZy, www.cazy.org; Cantarel et al., 2009). The GT14 family contains several mammalian GTs involved in protein glycosylation, e.g. β–1,6-N–acetyl glucosaminyltransferases catalyzing β–1,6-linked N–acetylglucosaminylation in core 2 or I–branched O–glycosylation and protein O–β-xylosyltransferases (Bierhuizen et al., 1993; Yu et al., 2001; Wilson, 2002). In contrast, none of the putative plant GT14s [e.g. 11 from Arabidopsis, 12 from rice (Oryza sativa)] have been characterized. Plant GT14s are phylogenetically related to the protein family containing Domain of Unknown Function 266 (DUF266; Ye et al., 2011), and a mutation in a DUF266 protein in rice (brittle culm 10, BC10) caused a severe alteration in the mechanical strength of the stem and the AG quantity and structure (Zhou et al., 2009). The authors concluded that BC10 is probably a GT, but involvement in the AG glycosylation pathway is not clear.

In this paper, we provide evidence for GlcAT activity of At5g39990 and its role in the biosynthesis of type II AG structures and cell elongation during seedling growth.


CAZy family GT14

Little is known about plant GTs in CAZy family GT14. As plants do not have the same type of glycoconjugates produced by mammalian GT14 enzymes (Bierhuizen et al., 1993; Yu et al., 2001; Wilson, 2002) and the plant GT14 enzymes are phylogenetically distantly related to their mammalian counterparts (Figure 1), a distinct activity is expected for the plant enzymes. We hypothesized that plant GT14 enzymes may be involved in the AG glycosylation pathway for three reasons: (i) plant GT14 members are related to DUF266 (Ye et al., 2011), and mutation in one of the DUF266 proteins in rice (BC10) caused severe alteration in AG quantity and structure (Zhou et al., 2009), (ii) some GT14s are co-expressed with genes encoding the protein backbone of AGPs (Showalter et al., 2010), and (iii) some GT14 genes are co-expressed with AtGALT31A (At1g32930), whose heterologously expressed protein demonstrated GT activity in elongation of the β–1,6-galactan side chains of AG (Geshi et al., 2013). An Arabidopsis gene, At5g39990 (indicated by an asterisk in Figure 1), is co-expressed with AtGALT31A during stem elongation (Figure S1), thus we selected the protein encoded by this gene for further characterization regarding its involvement in AG biosynthesis.

Figure 1.

Phylogenetic relationship among selected plant proteins and characterized GT14 proteins.

The tree includes plant sequences from Arabidopsis thaliana (At), Oryza sativa japonica group (Os/OJ), Triticum aestivum (Ta) and Populus tremula. The phylogeny analysis was performed and the phylogenetic tree was created using MEGA5 software (Tamura et al., 2011). The asterisk indicates At5g39990 (AtGlcAT14A) described in this paper. The sequence relationship within Arabidopsis GT14 enzymes is shown in Table S1.

Recombinant expression of the catalytic domain encoded by At5g39990 in Pichia pastoris

For biochemical characterization, the catalytic domain of At5g39990 was expressed as a soluble secreted construct with an N–terminal FLAG tag in Pichia pastoris. The recombinant protein was purified by immunoprecipitation and analyzed by SDS–PAGE (Figure 2a). The purified protein appeared as a smear, with the lowest apparent molecular size corresponding to 52 kDa, which was slightly higher than the expected size (47101.4 Da; Figure 2a, lane 8). Treatment with endoglycosidase H resulted in a shift of the protein size from 52 to 48 kDa, suggesting that the recombinant protein was N–glycosylated (Figure 2a, lanes 9 and 10).

Figure 2.

Enzyme activity of recombinant At5g39990 (AtGlcAT14A).

(a) SDS–PAGE analysis of the recombinant At5g39990 catalytic domain expressed in Pichia pastoris; lane M, markers. Lanes 1–6, 10-fold concentrated Pichia culture broth (12 μl each) expressing empty pPICZαA vector (2.8 μg total protein, lanes 1 and 4), RGXT1 (6.8 μg total protein, lanes 2 and 5) or At5g39990 (3.9 μg total protein, lanes 3 and 6); proteins were stained using Coomassie brilliant blue (lanes 1–3) or analyzed by Western blotting (lanes 4–6). Lanes 7 and 8, Western blot of affinity-purified samples (10 μl of bead slurry each) for RGXT1 (114 ng protein, lane 7) or At5g39990 (77 ng protein, lane 8). Lanes 9 and 10, deglycosylation of the purified At5g39990 by endoglycosidase H: untreated samples (lane 9) and treated samples (lane 10).

(b) Donor substrate identification for recombinant At5g39990. Seven NDP-[14C]-sugars were tested for GT activity on GAGP8–GFP acceptor using tenfold concentrated Pichia culture broth expressing the empty pPICZαA vector [black bars, lane 4 in (a)] or At5g39990 [gray bars, lane 6 in (a)]. Incorporation of [14C]-sugar onto the acceptor was analyzed by immunoprecipitation using anti-GFP antibody followed by scintillation counting (= 3). At5g39990 encodes a GlcAT that is involved in AG glycosylation, and was thus named AtGlcAT14A (Arabidopsis thaliana GlcA transferase from family GT14).

(c) Glucuronosyltransferase activity using purified At5g39990 [38.5 ng protein, gray bar, lane 8 in (a)] in comparison with the empty pPICZαA vector (black bar) and purified RGXT1 [white bar, lane 7 in (a)]. The assay method was the same as described in (b) (= 3). Purified RGXT1 is active as a xylosyltransferase (Figure S2).

(d) Effect of divalent metal ions and EDTA on the catalytic activity of purified AtGlcAT14A. The assay method was the same as described in (b) (= 2).

Characterization of donor substrate specificity of the recombinant protein

First, we tried to identify the donor substrate for the recombinant protein by testing the GT activity towards seven different NDP-[14C]-sugars. We used microsomes prepared from N. benthamiana after expression of a synthetic peptide encoding a glycomodule for AG glycosylation as acceptor for the assay (Figure 2b; GAGP8–GFP; Xu et al., 2005). Structural characterization of this GAGP8–GFP acceptor indicated the presence of various type II AG polysaccharides (Geshi et al., 2013), and thus an appropriate mixture of AG acceptors for various AG GTs. We assessed the transfer of [14C]-sugar to the GAGP8–GFP acceptor by immunoprecipitation of GAGP8–GFP and scintillation counting. Pichia culture broth containing the secreted At5g39990 protein (Figure 2a, lane 6) showed GT activity specifically towards UDP-[14C]-glucuronic acid (UDP-[14C]-GlcA) (Figure 2b), but no activity was observed for the Pichia culture broth harboring an empty pPICZαA vector as a control. This result indicates that the At5g39990 catalytic domain encodes a glucuronosyltransferase (GlcAT). We confirmed this finding using purified recombinant protein (Figure 2a, lane 8), which showed the same activity, i.e. transfer of [14C]-GlcA onto GAGP8–GFP acceptor (Figure 2c), while the empty pPICZαA vector and Arabidopsis α–1,3-d–xylosyltransferase, which is involved in biosynthesis of pectic rhamnogalacturonan II (AtRGXT1, Figure S2; Petersen et al., 2009) subjected to the same purification procedure did not show such activity (Figure 2c). Thus, we confirmed that At5g39990 encodes a GlcAT involved in AG glycosylation, and we named this protein AtGlcAT14A (Arabidopsis thaliana GlcA transferase from family GT14).

Many GTs require divalent metal ions, commonly Mn2+ or Mg2+, for catalytic activity (Lairson et al., 2008). However, the mammalian GT14 enzymes do not have such a requirement, because their sequences lack the metal ion binding DXD motif (Williams et al., 1980; Bierhuizen and Fukuda, 1992; Yeh et al., 1999; Schwientek et al., 2000). AtGlcAT14A also lacks the DXD motif and is fully active in the absence of Mn2+ or Mg2+, while high concentrations of Mn2+ (>1 mm) or Mg2+ (>10 mm) had an inhibitory effect (Figure 2d). EDTA at >5 mm concentrations also inhibited the enzyme activity, thus other divalent metal ions may be involved in the enzyme activity. Apparently, family GT14 enzymes including AtGlcAT14A have a different mode of substrate binding than most of the other GTs, which bind the UDP moiety of the substrate by the DXD motif via Mn2+ or Mg2+ ions.

Characterization of GlcA incorporation sites for AtGlcAT14A

Next, we investigated the site of GlcA incorporation in the AG acceptors. β–GlcA residues are present at the O6 position of Gal in β–1,6-galactan side chains as well as in the β–1,3-galactan main chain of type II AG (Tryfona et al., 2010, 2012; Nie et al., 2013), thus we tested β–1,6- and β–1,3-galactooligosaccharides of various lengths labeled with 2–aminobenzoic acid (2AA) and analyzed the products using normal-phase HPLC (Figure 3). Using an acceptor mixture of β–1,6-galactooligosaccharides of DP3–11, incubation with empty pPICZαA vector (Figure 3a,e,g,i) and AtRGXT1 (Figure 3b) controls did not result any changes, and the acceptor mixture was detected as it is (indicated by black arrowheads in Figure 3a,b). In contrast, incubation with AtGlcAT14A resulted in additional peaks detected between β–1,6-galactooligosaccharides acceptors (indicated by white arrowheads and marked as GP in Figure 3c). Further treatment using exo-β–glucuronidase reduced the level of those additional peaks and increased the level of unmodified 2AA-labeled β–1,6-galactooligosaccharide peaks (compare black and white arrowheads in Figure 3d). The results indicate that AtGlcAT14A transfers GlcA to 2AA-labeled β–1,6-galactooligosaccharides of DP3–11 via a β–linkage. Likewise, when 2AA-labeled β–1,3-galactooligosaccharide acceptors of DP3, 5 and 7 were tested, incubation with AtGlcAT14A resulted in no modification onto β–1,3-galactooligosaccharide of DP3 (Figure 3f), but an additional peak appeared when β–1,3-galactooligosaccharides of DP5 and 7 were used as acceptors (Figure 3h,j). This result indicates that AtGlcAT14A transfers GlcA to 2AA-labeled β–1,3-galactooligosaccharides of longer than DP5. Taken together, AtGlcAT14A is probably responsible for adding GlcA to both the β–1,6-galactan side chain and the β–1,3-galactan main chain of type II AGs (sites of action indicated in Figure 4).

Figure 3.

Glucuronosyltransferase activity using β-1,6- or β-1,3-galactooligosaccharides (β-1,6-Galmix-2AA or β-1,3-Galn-2AA) as acceptor.

(a–d) Assay of GlcAT activity towards β-1,6-Galmix-2AA (DP3–11) performed using affinity-purified materials from empty pPICZαA vector (a), RGXT1 (b) and AtGlcAT14A (c), and the reaction mixture of (c) treated with β-glucuronidase (d). For comparison, the chromatogram from (c) is shown as a gray line in (d).

(e–j) Assay using affinity-purified materials from empty pPICZαA vector (e,g,i) and AtGlcAT14A (f,h, j) of GlcAT activity towards β-1,3-Gal3-2AA (e,f), β-1,3-Gal5-2AA (g,h) and β-1,3-Gal7-2AA (i,j).

DP, degree of polymerization; GP, glycosylated product; black arrowheads indicate β-1,6-Galmix-2AA or β-1,3-Galn-2AA used as acceptor; white arrowheads indicate glycosylated product on β-1,6-Galmix-2AA or β-1,3-Galn-2AA.

Figure 4.

Model structure of type II AG based on AG analyzed from Arabidopsis leaf (modified from Figure 7 in Tryfona et al., 2012;).

Expected cleavage sites by hydrolases used in this study are indicated. The boxed GlcAp indicates the sites of GlcA transfer by AtGlcAT14A based on the in vitro enzyme assay shown in Figure 3.

Subcellular localization of AtGlcAT14A

To investigate the subcellular localization of AtGlcAT14A, full-length AtGlcAT14A conjugated with monomeric CFP (AtGlcAT14A–mCer3, Figure 5a) and STtmd–YFP (a Golgi marker comprising the sialyltransferase short cytoplasmic tail and single transmembrane domain fused to YFP, Figure 5b; Boevink et al., 1998) were transiently co-expressed in N. benthamiana leaves. AtGlcAT14A–mCer3 fluorescence was detected in punctate vesicles that co-localized with STtmd–YFP (Figure 5c), indicating localization of AtGlcAT14A in the Golgi apparatus.

Figure 5.

Subcellular localization of AtGlcAT14A–mCer3 transiently expressed in N. benthamiana leaves.

(a) AtGlcAT14A–mCer3, (b) STtmd–YFP, and (c) overlaid image of (a) and (b). The results indicate co-localization of AtGlcAT14A–mCer3 and STtmd–YFP in the Golgi apparatus. Scale bars = 10 μm.

In silico transcriptomic database analysis using GeneCAT (Mutwil et al., 2008) indicated that AtGlcAT14A is co-expressed with AtGALT31A, an enzyme that is involved in the synthesis of β–1,6-galactan side chains in type II AGs (AtGALT31A; Figure S1). To determine whether the AtGlcAT14A and AtGALT31A proteins interact, we investigated plausible protein–protein interaction using the acceptor photobleaching FRET technique (Poulsen et al., 2013), but the result did not indicate molecular interactions between the two proteins (Figure S3).

Characterization of T–DNA insertion lines of AtGlcAT14A

To investigate a role for AtGlcAT14A in vivo, we analyzed two allelic homozygous T–DNA insertion lines in Arabidopsis (SALK_064313 and SALK_043905, designated atglcat14a–1 and atglcat14a–2, respectively). The T–DNA insertions are in the 4th and 2nd exons of atglcat14a–1 and atglcat14a–2, respectively (Figure 6a). RT–PCR analysis using two sets of primers (annealing sites on the atglcat14a sequence indicated in Figure 6a) indicated the presence of a truncated transcript in atglcat14a–1 and a lack of transcript in atglcat14a–2 (Figure 6b).

Figure 6.

Analysis of AtGlcAT14A T-DNA insertion lines by RT-PCR.

(a) Schematic drawing of the AtGlcAT14A (At5g39990) gene structure and predicted sites of T-DNA insertions in two independent lines: atglcat14a-1 (SALK_064313) and atglcat14a-2 (SALK_043905). Exons are represented by thick gray boxes; introns and non-coding regions are represented by thin gray lines. The expected T-DNA insertion sites for each mutant are indicated. Black arrowheads indicate primer binding sites.

(b) Semi-quantitative RT-PCR to analyze the presence of transcript in the mutant lines. UBC indicates ubiquitin, which was used as the reference gene.

According to transcriptomics analysis using GeneCAT (http://genecat.mpg.de; Mutwil et al., 2008), Genevestigator (http://www.genevestigator.com/gv/) and the Arabidopsis eFP browser (Winter et al., 2007), AtGlcAT14A is highly expressed in root, the shoot apex and the shoot apex inflorescence (Figure S1). Hence, we first investigated roots and hypocotyls of the mutant seedlings grown in the dark. Interestingly, 5-day-old seedlings of atglcat14a–1 and -2 showed significantly increased elongation rates compared to wild-type in both roots (30 and 18%, respectively) and hypocotyls (33 and 23%, respectively) (Figure 7a). The monosaccharide composition of AG extracts from both atglcat14a–1 and -2 did not indicate significant changes in GlcA content, but showed an approximately 12% increase in Gal and an approximately 12% decrease in Ara compared to wild-type (Figure 7b). We further performed glycosidic linkage analysis for the AG extracts, which revealed an increase in 3-, 6- and 3,6-linked Gal and a decrease in 2-, 3-, 5- and 2,5-linked Ara in the mutant AG compared to wild-type (Table 1). The 3-, 6- and 3,6-linked Gal are specific to the type II AG structure, demonstrating little contamination by pectic rhamnogalacturonan I in the AG extracts. The structures of 2-, 3- and 5-linked Ara are commonly found as a part of type II AGs (Tan et al., 2004; Tan et al., 2010; Tryfona et al., 2010, 2012), but the 2,5-linked Ara detected in this study has not been reported and may appear specifically in type II AG during seedling growth. The increase in 3,6- and 6-linkages of Gal in the mutant AG may be the result of an increase in chain length in both β–1,6- and β–1,3-galactan. The relative decreases detected in multiple linkages for Ara are most likely the result of the relative increase in Gal linkages, or GlcA may be required for the addition of Ara. These results indicate that the mutation in AtGlcAT14A facilitates elongation of both β–1,6- and β–1,3-galactans, and initiation of 6–branching of β–1,3-galactan main chains in type II AG.

Table 1. Glycosidic linkage analysis (% total peak area) for AG extracts from wild-type and atglcat14a mutant plants
  1. Values are means ± standard deviation.

  2. Values that are significantly different between wild-type and atglcat14a shown in bold (< 0.01, Student's t test, = 4).

  3. T-Araf: terminal arabinofuranose, T-Fucp: terminal fucopyranose, 2-Araf: 2-linked arabinofuranose, 3-Araf: 3-linked arabinofuranose, T-Galp: terminal galactopyranose, 5-Araf: 5-linked arabinofuranose, 3-Galp: 3-linked galactopyranose, 2,5-Araf: 2,5-linked arabinofuranose. 6-Galp: 6-linked galactopyranose, 3,6-Galp: 3,6-linked galactopyranose.

Wild-type9.0 ± 1.20.7 ± 0.1 11.1 ± 0.6 5.9 ± 0.4 5.0 ± 0.3 6.0 ± 0.1 5.3 ± 0.3 9.2 ± 0.9 9.4 ± 0.5 19.4 ± 0.5 80.9
atglcat14a 9.3 ± 0.50.8 ± 0.1 7.2 ± 0.1 3.5 ± 0.1 5.7 ± 0.3 5.1 ± 0.1 8.3 ± 0.1 5.4 ± 0.1 14.1 ± 0.3 24.1 ± 0.3 83.5
Figure 7.

Growth phenotype and monosaccharide composition of wild-type and mutant seedlings.

(a) Length of hypocotyls and roots from seedlings grown in the dark and in the right for 5 days, respectively, was compared to the wild-type as 100% (> 45). Asterisks indicate significant differences compared with wild-type (Student's t test, < 0.01).

(b) Monosaccharide composition (mol%) of AGP extracts from 5-day-old etiolated seedlings analyzed by HPAEC-PAD (= 6 for wild-type, = 5 for atglcat14a-1 and -2). X/M, mixture of Xyl and Man. Asterisks indicate significant differences compared with wild-type (Student's t test, < 0.01).

We anticipated the reduced level of GlcA in the mutants, but the level of GlcA in the isolated AG extracts was low, and we did not detect significant differences by HPAEC-PAD (high-performance anion exchange chromatography with pulsed amperometric detection) monosaccharide analysis (Figure 7b) and immune labeling on the root surface using LM2 antibody (Smallwood et al., 1996) (Figure S4). Therefore, we performed a detailed analysis of AG side-chain structures from roots growing hydroponically for 3 weeks, by treatment of AG extracts with specific hydrolases followed by polysaccharide analysis using carbohydrate gel electrophoresis (PACE). We chose mature roots for the analysis to obtain enough material and also to minimize the effects of redundant activities, because AtGlcAT14A is highly expressed in roots while its closest homolog (At5g15050, 72% amino acid sequence identity, Figure 1) is expressed at a low level. No visible growth changes were observed at this stage for the mutants used in the analysis. The HPAEC-PAD monosaccharide composition analysis of mature roots did not indicate significant differences between wild-type and atglcat14a–1 and -2 (Figure S5), while PACE using treatments with exo-β–1,3-galactanase (Tsumuraya et al., 1990) followed by α–arabinofuranosidase (Takata et al., 2010) (cleavage sites indicated in Figure 4) revealed the presence of diverse oligosaccharides co-migrating with Ara, Gal, β–1,6-linked galactan and their substitutes at the O6 position with GlcA (GlcA-Gal, GlcA-Gal2, GlcA-Gal3) as these structures defined by Tryfona et al. (2010) (Figure 8a). Quantification of the oligosaccharides indicated that the ratios of GlcA-Gal/Gal and GlcA-Gal2/Gal2 were approximately 25 and 30% less, respectively, in atglcat14a-1 and -2, respectively (Figure 8b,c). There may also be differences in the GlcA-Gal3/Gal3 ratio, but the intensity of the corresponding bands was too low to perform statistical analysis. These results indicate reduced GlcA substitution on the β–1,3-linked Gal in the main chain and β–1,6-Gal2 in the side chains of the mutants compared to wild-type, and demonstrates the in vivo role of AtGlcAT14A in synthesis of those structures in type II AG (Figure 4).

Figure 8.

Structural analysis of AG side chains.

(a) AG side chains were released from wild-type, atglcat14a-1 and atglcat14a-2 AG extracts of mature roots by treatment with specific exo-β-1,3-galactanase followed by α-arabinofuranosidase (cleavage sites indicated in Figure 4), and separated by PACE. The migration positions of Ara, Gal, GlcA-(1→6)-β-Gal, GlcA-(1→6)-β-Gal2, GlcA-(1→6)-β-Gal3 and β-1,6-galactooligosaccharides of DP 2–8 are shown.

(b,c) Ratio of GlcA-substituted versus non-substituted Gal (GlcAGal/Gal) (b) and β-1,6-galactobiose (GlcAGal2/Gal2) (c) based on quantified intensity of corresponding spots in (a) (the positions indicated by arrows). Values are means ± SD of three biological replicates. Asterisks indicate significant differences compared with wild-type (Student's t test, < 0.05).


A novel β–GlcAT from the GT14 family is involved in biosynthesis of type II AG

The AG polysaccharides are very complex structures and therefore require a large number of GT enzymes for their biosynthesis; however, little is known about the AG glycosylation pathway. Here, we demonstrated that recombinantly expressed AtGlcAT14A possesses GlcAT activity, adding GlcA to both β–1,3- and β–1,6-galactooligosaccharides (Figure 3). Consistent with the enzyme activity shown in vitro, the level of GlcA in the corresponding structures was reduced in the two allelic T–DNA insertion mutant lines, indicating an in vivo role for AtGlcAT14A in adding GlcA to both the β–1,3-galactan main chain and the β–1,6-galactan side chain of type II AG.

GTs that are classified in the same CAZy family often share the same retaining or inverting reaction mechanism, and often transfer a donor sugar to the same position on the acceptor substrate (Amado et al., 1999). Therefore, it is reasonable that AtGlcAT14A (β–1,6-GlcAT) and mammalian β–1,6-GlcNAc transferases belong to the same CAZy family (GT14) as both enzymes catalyze an inverting reaction and transfer a sugar residue onto the O6 position of the acceptor sugar. In plants, GUX from family GT8 has been characterized biochemically as a GlcAT that adds GlcA to a β–1,4-xylan backbone via α–1,2 linkage (Mortimer et al., 2010; Lee et al., 2012; Rennie et al., 2012). As the GUX enzyme has a retaining activity, AtGlcAT14A described in this paper is the first example of a GlcAT with inverting activity in plants.

AtGlcAT14A required neither Mn2+ nor Mg2+ for its catalytic activity (Figure 2d), which is also observed for several mammalian GT14 enzymes (Williams et al., 1980; Bierhuizen and Fukuda, 1992; Yeh et al., 1999; Schwientek et al., 2000), and thus may be a common feature in GT14 family members. They all lack the DXD motif that is present in most other GT families and binds the nucleotide diphosphate of the substrate via a divalent metal ion. The crystal structure of a GT14 family member, murine β–1,6-GlcNAc transferase (C2GnT–L), indicated that R378 and K401 play a role in electrostatically stabilizing the nucleotide diphosphate leaving group instead of divalent metal ions (Pak et al., 2006). Plant GT14 sequences including AtGlcAT14A appear to lack the corresponding residue for R378 (Figure S6), thus plant GT14 enzymes appear to interact with the donor substrate in a slightly different way than mammalian GT14 enzymes. Elucidation of the crystal structure of AtGlcAT14A together with the substrate will provide further insights regarding evolution of GT14 enzymes between mammals and plants.

Physiological role of AtGlcAT14A and GlcA substitution on AGPs in vivo

We analyzed two independent T–DNA insertion mutant lines, and the results suggested that AtGlcAT14A has significant influence on cell elongation during seedling growth (Figure 7a). In both AtGlcAT14A mutant lines, we detected a relative increase of Gal and a reduction of Ara in AG extracts (Figure 7b). Further detailed structural analysis by determining the glycosidic linkage showed a relative increase of Gal in 3-, 6- and 3,6-linkages (Table 1), indicating the presence of longer β–1,3- and β–1,6-galactans as well as 3,6-branched Gal in the mutants. The increase in the number of 3,6- and 6-linkages for Gal is probably the result of increased acceptor sites (O6 site of β-1,3-linked Gal and O6 site of the terminal Gal of β–1,6-linked galactan) at which the GalT(s) may initiate and elongate β–1,6-galactan side chains, as addition of GlcA and Gal occurs at the same acceptor sites (Figure 4). Nevertheless, addition of GlcA appears to terminate elongation of β–1,3- and β–1,6-galactans, as well as initiating 3,6-Gal branches in type II AG in vivo.

The level of GlcA in type II AG is very low, and we were unable to detect an altered level of GlcA in the mutants using HPAEC-PAD monosaccharide composition analysis or LM2 immunolabeling on the root surface in seedlings (Figure 7b and Figures S4 and S5). Recent technical advances using PACE coupled with AG-specific hydrolases allowed us to obtain more detailed information for the side-chain structures of type II AG. As shown in Figure 8, the GlcA-Gal/Gal and GlcA-Gal2/Gal2 ratios were approximately 25–30% less in the two AtGlcAT14A mutants, indicating a role for AtGlcAT14A in synthesis of these structures in vivo. We did not observe a visible alteration of growth in the 3-week-old mutants analyzed, but we are working on generation of the double mutant atglcat14a at5g15050 in order to investigate further physiological roles of AG GlcAT in plants.

Evidently, the mutation in AtGlcAT14A caused substantial changes in type II AG structure in the mutants. Eudes et al. (2008) investigated an Arabidopsis β–glucuronidase (AtGUS2), whose endogenous substrate is considered mainly to be type II AG, and reported increased cell elongation in seedlings of over-expressing AtGUS2 and the opposite phenotype in a T–DNA insertion line (Eudes et al., 2008). Together with our results, these results indicate a role for GlcAT(s) and GlcA hydrolase(s) towards type II AG in cell elongation. However, the cell elongation phenotype was not directly related to a change in the particular chemical composition of type II AG; we observed an increase in Gal and a decrease in Ara in the AG extracts of atglcat14a mutants due to altering AG glycosylation in the Golgi during biosynthesis, while Eudes et al. (2008) observed a decrease of both Gal and Ara in the AG extracts of plants over-expressing AtGUS2, which most likely does not interfere with the biosynthesis of AGs but affects AG modification after deposition in the apoplast. However, both types of mutant plants exhibited increased cell elongation in seedlings. In both cases, the changes in GlcA level were very minor, and Eudes et al. (2008) further reported a substantial increase in xylose in AG extracts from the AtGUS2 over-expressing line, which was not the case for AG extracts from atglcat14a mutants (Figure 7b). The cell-wall extensibility results from a complex and poorly understood network of interactions between various cell-wall polymers (Wolf et al., 2012) and an AGP covalently linked to pectin and arabinoxylan has been described (Tan et al., 2013). We do not know to what extent such structures are present in the cell walls, and little is known regarding the interactions of type II AG with other polymers, but the cell elongation results reported here may be a result of overall changes in cell-wall architecture and integrity caused by large structural changes in AG and other interacting polymers. Alternatively, the process may be more specifically mediated by AG and the intracellular Ca2+ signaling pathway as interruption of cell-wall integrity, such as binding of the β–glucosyl Yariv reagent to AG, triggers a rapid increase in intracellular Ca2+ (Roy et al., 1999; Pickard and Fujiki, 2005), and AG directly binds and releases Ca2+ in a pH-dependent way (Lamport and Varnai, 2013). The binding of Ca2+ is considered to occur via GlcA residues in the side chains of AG, thus family GT14 enzymes may play a role in Ca2+ signaling via biosynthesis of GlcA in type II AG.

Our future research will focus on characterization of the GT14 enzymes with regard to catalytic mechanisms and their relationship to other GTs in the biosynthesis of type II AG in order to understand the roles of biosynthetic enzymes in the final structures and properties of AG, such as a role of GlcA in regulating the intracellular Ca2+ signaling pathway.

Experimental procedures


UDP-α–d–glucose (UDP-Glc), UDP-α–d–xylose (UDP-Xyl), GDP-α–d–mannose (GDP-Man), GDP-α–d–fucose (GDP-Fuc), UDP-α–dN–acetyl-d–glucosamine (UDP-GlcNAc), UDP-α–d–glucuronic acid (UDP-GlcA), UDP-α–d–galactose (UDP-Gal) and β–glucuronidase type II from Helix pomatia were purchased from Sigma Aldrich (www.sigmaaldrich.com). UDP-α–d–[14C]-Glc, UDP-α–d–[14C]-Xyl, GDP-α–d–[14C]-Man, GDP-α–d–[14C]-Fuc, UDP-α–d–[14C]-GlcNAc, UDP-α–d–[14C]-GlcA and UDP-α–d–[14C]-Gal were purchased from PerkinElmer (http://www.perkinelmer.com). The synthetic glycomodule GAGP8–GFP construct, placed under the control of the tobacco signal sequence in the pBI121 vector (Xu et al., 2005), was kindly provided by Marcia Kielieszewski (Department of Chemistry and Biochemistry, Ohio University, Athens, OH). The method for preparation of the microsomes after expression of GAGP8–GFP has been described previously (Geshi et al., 2013). The mixture of β–1,6-galactooligosaccharides was prepared by sequential digestion of leaf AGP extract from the mur1 mutant using α–arabinofuranosidase, exo-β–1,3-galactanase and β–glucuronidase as described previously (Tryfona et al., 2010). β–1,3-galactooligosaccharides of DP3, 5 and 7 were prepared by partial acid hydrolysis with 40 mm trifluoroacetic acid at 100°C for 1 h (Kitazawa et al., 2013). Labeling of oligosaccharides with 2AA was performed as described previously (Tryfona et al., 2010).

Chemically synthesized 4–O-methyl GlcA (Sixta et al., 2009) was a kind gift from Paul Kosma (Division of Organic Chemistry, University of Natural Resources and Applied Life Sciences, Vienna, Austria). Monomeric Cerulean3 (mCer3) in the pmCer3–C1 vector (Markwardt et al., 2011) was a kind gift from Mark A. Rizzo (Department of Physiology, University of Maryland, Baltimore, MD). Termamyl SC was a kind gift from Novozymes A/S (http://www.novozymes.com).

Plant materials

Arabidopsis T–DNA insertion lines atglcat14a–1 (SALK_064313) and atglcat14a–2 (SALK_043905) were obtained from the Salk Institute (Alonso et al., 2003;). Arabidopsis ecotype Col–0 was used as the wild-type for comparison. Plants were grown in the greenhouse under an 8 h photoperiod at 20°C and 70% relative humidity. Homozygous lines were identified by PCR using primer sets suggested by the Salk Institute (gene-specific primers for SALK_064313: 5′-ACCTTAAGGCATGTTGTGTGG-3′ and 5′-CCAACAGCATTCAAGCTTTTC-3′; gene-specific primers for SALK_043905: 5′-ATTGGTTCAATCTTCGCTTTG-3′ and 5′-TCAACCAATGAGAAATGGAGC-3′; T–DNA left border primer: 5′-ATTTTGCCGATTTCGGAAC-3′).

Plants were grown hydroponically as described by Lehmann et al. (2009). Plants were grown for 3 weeks under short-day conditions (8 h photoperiod at 20°C), and then transferred to long-day conditions (16 h photoperiod at 20°C).

To determine the cell elongation rate in mutant lines, seeds were surface-sterilized and plated onto solidified half-strength MS medium in culture plates. After 4 days of stratification at 4°C in the dark, seeds were exposed to 4 h fluorescent white light at 20°C to synchronize germination before wrapping plates in aluminum foil for growth in darkness for 5 days at 20°C.

Wild-type N. benthamiana was grown in the greenhouse at 28°C (day) and 18°C (night) with a 16 h photoperiod and used for transient expression.


Transmembrane helices were predicted using the TMHMM Server version 2.0 (http://www.cbs.dtu.dk/services/TMHMM-2.0/; Sonnhammer et al., 1998), and N–glycosylation sites were predicted using NetNGlyc 1.0 (http://www.cbs.dtu.dk/services/NetNGlyc/; Blom et al., 2004). In silico expression analysis was performed using GeneCAT (http://genecat.mpg.de; Mutwil et al., 2008), Genevestigator (http://www.genevestigator.com/gv/) and the Arabidopsis eFP browser (Winter et al., 2007).

For generation of the phylogenetic tree, the full-length amino acid sequence of model plants (see Figure 1) and characterized GT14 proteins were aligned using the T–Coffee multiple sequence alignment tool (http://www.tcoffee.org/Projects/tcoffee/; Notredame et al., 2000), and the phylogenetic analysis was performed by neighbor-joining implemented in the Molecular Evolution and Genetic Analysis software package version 5 (MEGA5; Tamura et al., 2011) with complete deletion of gaps and the Poisson correction distance of substitution rates. Statistical support for phylogenetic grouping was estimated by 1000 bootstrap re-samplings.

Cloning of the AtGlcAT14A catalytic domain for expression in Pichia pastoris

Total RNA was extracted from the shoot apex of A. thaliana (Col–0) using the Spectrum™ Plant Total RNA kit (Sigma-Aldrich), and converted to cDNAs using the iScript cDNA synthesis kit (Bio–Rad, www.bio-rad.com). The full-length At5g39990 (AtGlcAT14A) sequence was amplified by PCR from cDNA using the gene-specific forward primer 5′-ATGAAGAAATTGAGAAGCTATTAC-3′ and reverse primer 5′-TCACTTACACTGTTTTGATCGG-3′ with partial Gateway sites added to the 5′ end of each primer. The resulting PCR product was re-amplified using Gateway site-specific primers, and cloned into a Gateway entry pDONR221 vector (Life Technologies, www.lifetechnologies.com). This full-length cDNA was used as a template to amplify the soluble catalytic domain (amino acid residues 58-447) by PCR using forward primer 5′GGATCCTCACGTGACTTCACCAACCGGAGGAGT-3′ and reverse primer 5′GCGGCCGCGAGCTCTCACTTACACTGTTTTGATC-3′ (BamHI, PmlI, NotI and SacI restriction sites are underlined). The amplified DNA was cloned into the modified pPICZαA vector (Life Technologies) containing an introduced N–terminal FLAG tag (MDYKDDDDK) (Petersen et al., 2009) using PmlI and NotI. All constructs were verified by sequencing at each cloning step.

Expression of secreted soluble AtGlcAT14A in Pichia pastoris

The soluble catalytic domain of At5g39990 cloned into the modified pPICZαA vector was transformed into P. pastoris KM71 strain (Petersen et al., 2009), which was grown as described by Dilokpimol et al. (2011). Each 100 ml of culture medium contained approximately 2.5 ± 0.9 mg recombinant protein. The culture medium was harvested by centrifugation at 12 000 g at 4 °C for 1 h and concentrated tenfold using Vivaspin 20 ultrafiltration devices (10 000 molecular weight cut-off, Polyethersulfore; GE Healthcare, www.gelifesciences.com). The culture medium was replaced with PBS by repeating the concentration step in the presence of PBS. For purification of the recombinant AtGlcAT14A, 1 ml of buffer-exchanged material was treated with 50 μl anti-Flag M2 affinity slurry containing 50% gel (Sigma-Aldrich) in the presence of 5 mm n–dodecyl β–d–maltoside, and incubated for 16 h at 4°C with rotation. The gel was collected by centrifugation at 500 g for 30 sec at 4°C followed by washing three times with PBS, pH 7.0, by centrifugation at 500 g for 30 sec at 4 °C, prior to further characterization and enzymatic assay. Each 1 μl anti-Flag M2 affinity slurry contained 7.7 ± 1.5 ng protein.

Protein analyses

Protein concentration was determined by the Bradford assay using Bio–Rad protein assay dye reagent. SDS–PAGE was performed using Criterion XT pre-cast gels (12% Bis/Tris; Bio–Rad) and Western blotting was performed using the Criterion system (Bio–Rad). After blocking the membrane using PBS containing 5% w/v skimmed milk, FLAG-conjugated peptides were detected using monoclonal anti-Flag M2 antibody (Sigma-Aldrich) followed by horseradish peroxidise-conjugated anti-mouse immunoglobulins (Dako Cytomation, www.dako.com), and the signal was developed by chemiluminesence (Supersignal West Dura extended duration substrate, Thermo Scientific, www.thermoscientific.com). Endoglycosidase H treatment was performed according to the manufacturer's instructions (New England BioLabs, www.neb.com).

Enzyme assays

For the assay using N. benthamiana microsome-expressed GAGP8–GFP as the acceptor, the reaction was performed in the presence of 0.1 mM NDP-sugar (containing 277.5 Bq of NDP-[14C-]-sugar), 28 mM HEPES, 10 mM MnCl2, pH 7.0, and 5 μl of either tenfold concentrated AtGlcAT14A in PBS or affinity-purified AtGlcAT14A on bead slurry as the enzyme source and 5 μl of GAGP8–GFP (5 μg μl−1) as the acceptor. Incubation was performed at 22°C for 16 h, and samples were stored at −20°C prior to analysis. The product formed on the GAGP8–GFP acceptor was collected by immunoprecipitation by incubating with 0.4 μg GFP antibody (Sigma-Aldrich) in the presence of 0.15 m NaCl and 5 mm n–dodecyl β–d–maltoside at 4°C for 16 h, followed by incubation with 10 μl protein G/agarose (approximately 50% v/v slurry in H2O, Sigma-Aldrich) at 4 °C for 1 h. The presence of 14C-sugar in the immunoprecipitated material was analyzed by scintillation counting.

For the assay using 2AA-labeled β–1,6-galactooligosaccharides as acceptor, the reaction was performed in the presence of 0.1 mm NDP-sugar, 28 mm HEPES, 10 mm MnCl2, pH 7.0, 5 μl affinity-purified AtGlcAT14A as the enzyme source and 5 μl 2AA-labeled β–1,6-galactooligosaccharides (10 μg containing approximately 1900 fluorescence signal per μl) as the acceptor. Treatment of the products with 0.6 units of β–glucuronidase (type II, Sigma-Aldrich) was performed at 37°C for 1 h. The reaction mixtures were filtered (Ultrafree–MC poly(vinylidene difluoride) membrane, pore size of 0.22 μm) and desalted using Superdex peptide HR 10/30 (GE Healthcare, www.gehealthcare.com) prior to analysis by HPLC (Dionex, www.dionex.com) using a TSKgel Amide–80 column (3 μm spherical silica particle, 4.6 mm inside diameter, 150 mm length of the column; 3.2 mm inside diameter, 15 mm length of the guard column; TOSOH, www.tosohbioscience.com) using a gradient of acetonitrile (ACN)/50 mm ammonium formate, pH 4.3 (80% ACN, 12 min; 80–45% ACN, 45 min; 45% ACN, 15 min; 45–80% ACN, 5 min; 80% ACN, 20 min for re-equilibrating) at a constant flow rate (0.5 ml per minute) at 25°C.

The products were detected using a fluorescence detector (RF2000, Dionex). 2AA-labeled β–1,6-galactose, β–1,6-galactobiose and β–1,6-galactotriose were used as standards to identify the 2AA-labeled β–1,6-galactooligosaccharides elution positions.

For the assay using β–1,3-galactooligosaccharides as acceptor, the reaction was performed in the presence of 0.1 mm NDP-sugar, 16 mm McIlvaine buffer (McIlvaine, 1921), pH 5.0, 5 μl affinity-purified AtGlcAT14A as the enzyme source, and 2.5 μl of 2 mm β–1,3-Gal3-, β–1,3-Gal5- or β–1,3-Gal7-2AA as the acceptor. The reaction mixtures were filtered through Ultrafree–MC poly(vinylidene difluoride) membrane (pore size of 0.22 μm) prior to analysis using a TSKgel Amide–80 column by HPLC as described above with a modified ACN gradient (80% ACN, 12 min; 80–45% ACN, 55 min; 45% ACN, 15 min; 45–80% ACN, 5 min; 80% ACN).

Subcellular localization

Monomeric Cerulean3 (mCer3; Markwardt et al., 2011) was PCR-amplified using forward primer 5′-GGTGCCTAGGGTGGTGAGCAAGGGCGAGGAG-3′) and reverse primer 5′-CTCTAGGGACTAGTTAATTAAGCGTAATCTGGAACATCGTATGGGTACTTGTACAGCTCGTCCATGCC-3′) (restriction sites AvrII and SpeI are underlined, the HA tag is in italics) and replaced with cyan fluorescent protein (CFP) in a pEarleyGate 102 Gateway vector (Earley et al., 2006) using AvrII and SpeI. Full-length At5g39990 cDNA cloned into pDONR221 (Life Technologies) was moved to the modified pEarleyGate 102 vector by LR reaction (Life Technologies) and transformed into Agrobacterium tumefaciens strain C58C1 pGV3850. Determination of the subcellular localization was performed as described by Harholt et al. (2012). The sialyltransferase short cytoplasmic tail and single transmembrane domain fused to YFP (STtmd–YFP; Boevink et al., 1998) was used as a Golgi marker. After 2 days of infiltration, leaves were analyzed using a Leica TCS SP5 inverted confocal laser scanning microscope (www.leica.com) with a 63 x water immersion objective (Numerical Sperture of 1.2). Excitation with an argon laser was performed at 458 and 514 nm, and emission was detected at 475–505 and 525–600 nm, for mCer3 and YFP, respectively.


RNA was extracted from plant material ground in liquid nitrogen using the Spectrum™ plant total RNA kit (Sigma-Aldrich). RT–PCR was performed using cDNA synthesized using the iScript cDNA synthesis kit (Bio–Rad) and primers 5′-AGAGAGGAGCTTCATGGAT-3′ (a), 5′-ACCGGTTTTGCTCTTTGTGATGCTT-3′ (b), 5′-CATCCCAACAGCATTCAAGC-3′ (c) and 5′-AACTTTCTAGCAAACGGGGC-3′ (d) (position of primers indicated in Figure 6a). Ubiquitin-specific primers were used as control: LP, 5′-TCAAATGGACCGCTCTTATC-3′; RP, 5′-CACAGACTGAAGCGTCCAAG-3′.

Structural analysis of AG

An AG-enriched fraction was prepared as described by Tryfona et al. (2012) with modifications. Fresh frozen plant material was ground to a fine powder using a Retsch tissue lyzer (Qiagen, www.qiagen.com). Four volumes of PBS were added to samples and heated to 100°C for 30 min. During boiling, samples were treated with 10 μg μl−1 (1.2 kilo Novo units per μg) Termamyl SC (Novozymes) to remove starch. One kilo Novo unit is defined as the amount of enzyme that dextrinizes 5.26 g of starch dry substance (Merck Amylum soluble, Merck Millopore, www.merckmillopore.com) per hour under standard conditions (as defined by Novo Nordisk, www.novonordisk.com) for α–amylase determination (37 ± 0.05°C, 0.3 mm Ca2+, pH 5.6) (Wu et al., 2012). After the reaction, samples were cooled to room temperature, cell debris was removed by centrifugation at 4000 g, and the supernatant was mixed with 2.5 volumes of 96% v/v ethanol and kept at 4°C overnight. Precipitates were isolated by centrifugation (12 000 g, 20 min, 4°C), and resuspended in 10 volumes of water by constant shaking at 4°C overnight. Monosaccharide composition analysis by HPAEC-PAD was performed as described by Obro et al. (2004). The PACE analysis was performed as described by Tryfona et al. (2010). Glycosidic linkage analysis was performed by GC/MS analysis of their partially methylated alditol acetates derivatized as described by Ciucanu and Kerek (1984) with modifications by Ciucanu (2006). The resulting partially methylated alditol acetates were separated using a gas chromatograph (Agilent 7890A, Agilent Technologies, www.agilent.com) equipped with a Supelco SP2380 column (Sigma-Aldrich) and a mass spectrometer (Agilent 5975C). Peaks were identified based on their retention time and ion fragmentation pattern, and assigned according to the Complex Carbohydrate Research Center's partially methylated alditol acetate database (http://www.ccrc.uga.edu/specdb/ms/pmaa/pframe.html).


We would like to thank Jack Egelund and Peter B. Jørgensen for their support with the cloning work. This work was supported by the Faculty of Life Sciences, University of Copenhagen (E.K.), the Danish Council for Strategic Research, Food, Health and Welfare and the Danish Council for Independent Research, Technology and Production Sciences (A.D., C.P.P. and N.G.), the Villum Foundation's Young Investigator Program (J.H.), and the US Department of Energy (M.H.). Work performed by G.X. and M.P. was funded by the Energy Biosciences Institute. Imaging data were collected at the Center for Advanced Bioimaging of Denmark, University of Copenhagen, Denmark.