Two Xyloglucan Xylosyltransferases Catalyze the Addition of Multiple Xylosyl Residues to Cellohexaose*

Xyloglucan (XyG) is the principal hemicellulose found in the primary cell walls of most plants. XyG is composed of a β-(1,4)-glucan backbone that is substituted in a regular pattern with xylosyl residues, which are added by at least one and likely two or three xylosyltransferase (XT) enzymes. Previous work identified seven Arabidopsis thaliana candidate genes, one of which (AtXT1) was shown to encode a protein with XT activity (Faik, A., Price, N. J., Raikhel, N. V., and Keegstra, K. (2002) Proc. Natl. Acad. Sci. U. S. A. 99, 7797-7802). We expressed both AtXT1 and a second closely related gene, now called AtXT2, in insect cells and demonstrated that both have XT activity for cellopentaose and cellohexaose acceptor substrates. Moreover, we showed that cellohexaose was a significantly better acceptor substrate than cellopentaose. Product structural characterization showed that AtXT1 and AtXT2 preferentially added the first xylosyl residue to the fourth glucosyl residue from the reducing end of both acceptors. Furthermore, when the ratio of UDP-xylose to cellohexaose and the reaction time were increased, both AtXT1 and AtXT2 added a second xylosyl residue adjacent to the first, which generated dixylosylated cellohexaose. On the basis of these results, we concluded that AtXT1 and AtXT2 have the same acceptor specificities and generate the same products in vitro. The implications of these results for understanding in vivo XyG biosynthesis are considered.

Xyloglucan (XyG) is the principal hemicellulose found in the primary cell walls of most plants. XyG is composed of a ␤-(1,4)glucan backbone that is substituted in a regular pattern with xylosyl residues, which are added by at least one and likely two or three xylosyltransferase (XT) enzymes. Previous work identified seven Arabidopsis thaliana candidate genes, one of which (AtXT1) was shown to encode a protein with XT activity (Faik, A., Price, N. J., Raikhel, N. V., and Keegstra, K. (2002) Proc. Natl. Acad. Sci. U. S. A. 99, 7797-7802). We expressed both AtXT1 and a second closely related gene, now called AtXT2, in insect cells and demonstrated that both have XT activity for cellopentaose and cellohexaose acceptor substrates. Moreover, we showed that cellohexaose was a significantly better acceptor substrate than cellopentaose. Product structural characterization showed that AtXT1 and AtXT2 preferentially added the first xylosyl residue to the fourth glucosyl residue from the reducing end of both acceptors. Furthermore, when the ratio of UDP-xylose to cellohexaose and the reaction time were increased, both AtXT1 and AtXT2 added a second xylosyl residue adjacent to the first, which generated dixylosylated cellohexaose. On the basis of these results, we concluded that AtXT1 and AtXT2 have the same acceptor specificities and generate the same products in vitro. The implications of these results for understanding in vivo XyG biosynthesis are considered.
The primary wall surrounding plant cells is composed of cellulose, hemicellulose, pectin, and protein. Xyloglucan (XyG) 2 is the principal hemicellulose found in the primary cell walls of non-graminaceous plants, where it composes up to 25% of the cell wall. XyG is thought to link via hydrogen bonds to the surfaces of adjacent cellulose microfibrils, thereby forming three-dimensional cellulose-XyG networks that function as the principal load-bearing structure of the primary cell wall (1).
XyG is composed of a backbone of 1,4-linked ␤-D-Glcp residues that are substituted in a regular pattern at O-6 with an ␣-D-Xylp residue (where "Xyl" is xylose) to form repeating subunits with specific xylosylation patterns that are highly conserved (2). Most plants have XyG with an XXXG-type repeating subunit (where "G" denotes an unsubstituted glucosyl residue, and "X" denotes a glucosyl residue substituted at O-6 with an ␣-D-Xylp residue; see Ref. 3 for XyG nomenclature). However, plants in Poaceae and Solanaceae have an XXGG-type repeating subunit and perhaps also an XXGGG-type repeating subunit (for reviews, see Refs. 4 -6). Further sugar substitution occurs primarily at specific xylosyl residues within the repeating subunit. These xylosyl residues are substituted at O-2 with a variety of glycosyl moieties, with the most common being ␤-D-Galp (represented by "L" and found in many plant species), ␣-L-Araf (where "Ara" is arabinose; represented by "S" and found only in Poaceae and Solanaceae), and the disaccharide ␣-L-Fucp-(1,2)-␤-D-Galp (where "Fuc" is fucose; represented by "F"). Although the structure of the XyG repeating subunit is generally highly conserved within a plant species, the frequency and distribution of further carbohydrate substitutions may vary significantly. For example, endoglucanase digestion of Arabidopsis XyG produces a diverse array of oligosaccharides, including XXXG, XXLG, XXFG, XLFG, XLLG, and XLXG, all containing XXXG as the core XyG repeating subunit (7).
The biosynthesis of XyG requires ␣-fucosyltransferase, ␤-galactosyltransferase, ␣-xylosyltransferase (XT), and ␤-(1,4)glucan synthase activities. Recently, much progress has been made in the identification and characterization of the genes and proteins involved in XyG biosynthesis. An XyG ␣-fucosyltransferase (FUT1) was purified from pea epicotyls using traditional biochemical purification techniques (8). The amino acid sequence was used to identify XyG fucosyltransferase genes in Arabidopsis thaliana (AtFUT1) (8) and Pisum sativum (PsFUT1) (9). Alternatively, a genetic strategy was used to identify one of the XyG galactosyltransferases. A screen of ethylmethane sulfonate-mutagenized Arabidopsis plants identified a mutant line (mur3) with a reduction in cell wall fucose content (10); further investigation determined that MUR3 encodes an XyG galactosyltransferase (7).
Recently, Faik et al. (11) identified seven putative XyG XT genes that are related to the previously characterized fenugreek galactomannan galactosyltransferase (GMGT) (12) and that belong to CAZy family GT34 (afmb.cnrs-mrs.fr/CAZY/) (13). The results of heterologous expression of the seven candidate XT genes in Pichia pastoris indicate that AtXT1 (At3g62720) encodes a protein with XT activity when cellopentaose is the acceptor substrate (11). However, attempts to express other members of this gene family in Pichia cells did not result in detectable XT activity, so the functions of the AtGT2-7 glycosyltransferases remain to be determined. The lack of activity was particularly surprising for At4g02500 because it has high sequence similarity (83% identical and 91% similar at the amino acid level) to AtXT1. Therefore, we revisited the heterologous expression of AtXT1 and At4g02500 in insect cells. In this study, we further characterized AtXT1 activity and present evidence that At4g02500 (henceforth referred to as AtXT2) encodes a protein with XT activity. Furthermore, through product characterization, we show that both proteins can add multiple xylosyl residues to cellohexaose.
Cloning of AtXT1 and AtXT2-AtXT1 (At3g62720) and AtXT2 (At4g02500) were PCR-amplified from full-length cDNA clones with gene-specific primers containing a CACC adapter on the 5Ј-end of the forward primer, which is required for directional cloning using the pENTR/D-TOPO vector. The primer sequences are as follows: AtXT1, CACCATGATAGA-GAAGTCTATAGGAGCGCA (forward) and CGCAAAATT-AAAAGATACAAAACAA (reverse); and AtXT2, CACCATG-ATTGAGAGGTGTTTAGGAGCTTA (forward) and CCTA-AACGCAAAACCGATTC (reverse). PCR products were separated on a 1.5% agarose gel, and the DNA corresponding to the size of the expected PCR product was purified using the GENECLEAN SPIN kit, directionally cloned into the pENTR/ D-TOPO donor vector, and transformed into E. coli TOP10 competent cells. Proper donor constructs were identified from isolated plasmids that had been sequenced to ensure that no mutations had been introduced.
Expression of AtXT1 and AtXT2-The Drosophila expression system and the Bac-to-Bac baculoviral expression system were used for heterologous expression of AtXT1 and AtXT2. Constructs of AtXT1 and AtXT2 in pENTR/D-TOPO donor vectors were recombined into pMT-DEST48 and pDEST8 destination vectors via 18-h Clonase reactions.
Drosophila S2 (Schneider 2) cells were cotransfected with the pCoBlast vector and pMT-DEST48 containing either the fulllength AtXT1 or AtXT2 cDNA sequence. Selection of cell lines stably expressing AtXT1 and AtXT2 was done with blasticidin following the supplier's instructions (Invitrogen). Expression of AtXT1 and AtXT2 was induced by the addition of copper sulfate (500 M) to the cell cultures 24 h prior to cell harvest. Control S2 cell lines that were not transfected were cultured following the supplier's instructions. S2 cell cultures (20 -30 ml each) were centrifuged at 500 ϫ g for 10 min; the supernatant was removed; and the cells were solubilized with extraction buffer (100 mM HEPES (pH 7.0) containing 1% Triton X-100, 5 mM MnCl 2 , and one protease inhibitor tablet/20 ml).
The pDEST8-AtXT1 and pDEST8-AtXT2 constructs were electroporated into E. coli DH10Bac cells for transposition of the respective XT gene into Autographa californica bacmids, which were subsequently isolated from positive clones, and the inserts were sequenced. Positive recombinant bacmids were transfected into Sf21 (S podoptera frugiperda 21) cells following the supplier's instructions (Invitrogen). The Sf21 cell cultures were divided, placed into 50-ml conical tubes, and centrifuged at 5000 ϫ g for 15 min. The supernatant was decanted, and the pelleted Sf21 cells were frozen in liquid nitrogen and stored at Ϫ80°C.
The S2 or Sf21 pellets were suspended in ice-cold extraction buffer. The cells were suspended with a pipette and disrupted by 3 ϫ 15 s sonication with a 45-s pause, with storage on ice between sonication treatments. Cell disruption was verified by microscopy. Protein assays were conducted, and the protein concentration was adjusted to 6.25 mg/ml with extraction buffer. The cell extracts were either placed on ice for immediate use in enzyme assays or frozen with liquid nitrogen and stored at Ϫ80°C.
Immunoblot Analysis-Cell extracts (either 40 g of S2 extract total protein or 20 g of Sf21 extract total protein) were separated by SDS-PAGE as described previously (14), and the proteins were transferred to polyvinylidene difluoride membranes (15). Membranes were blocked with a 5% solution of milk proteins and probed with a polyclonal antibody (1:1000 dilution) raised in rabbits against a fragment of AtXT1 (amino acids . This antibody was a gift of Ahmed Faik, and the details of its preparation and characterization will be reported elsewhere. 3 Membranes were probed with horseradish peroxidase-conjugated goat anti-rabbit antibodies and developed using the Super Signal West Pico chemiluminescence system. Membranes were stained with Coomassie Blue to ensure uniform protein loading and transfer. XT Assay-The XT assay was based upon conditions described previously (11). Cell extracts from S2 and Sf21 cells expressing AtXT1 and AtXT2 were initially surveyed for XT activity with 0.4 mM UDP-xylose, 44,400 dpm UDP-[ 14 C]xylose, and 4.0 mM acceptor substrate (0.1:1.0 substrate ratio); all other XT reactions were conducted in the presence of either 1.0 mM UDP-xylose, 44,400 dpm UDP-[ 14 C]xylose, and 1.0 mM acceptor substrate (1.0:1.0 substrate ratio) or 2.5 mM UDP-xylose, 111,000 dpm UDP-[ 14 C]xylose, and 1.0 mM acceptor substrate (2.5:1.0 substrate ratio). For each XT assay, 40 l of cell extract was mixed with 10 l of substrate solution containing the appropriate amounts of UDP-xylose, UDP-[ 14 C]xylose, and acceptor substrate dissolved in extraction buffer. XT reactions were incubated at 25°C for either 1 or 18 h and terminated by the addition of 450 l of Dowex 1-X8 ion-exchange resin suspended in water (1:2 resin/water). Each reaction slurry was loaded onto a Micro Bio-Spin chromatography column and spun in a microcentrifuge at 1300 ϫ g for 2 min. The column effluent (either 50 or 300 l) was mixed with 2.0 ml of scintillation counting fluid; radioactivity was then measured on a Beckman Coulter LS-5000 scintillation counter. XT reactions that were destined to be analyzed by matrix-assisted laser desorption ionization time-offlight mass spectrometry (MALDI-TOF-MS) did not contain UDP-[ 14 C]xylose and were carried out in parallel with XT reactions containing UDP-[ 14 C]xylose. XT Product Characterization Strategy-To structurally characterize XT products, we developed a strategy that used a combination of high performance anion-exchange chromatography (HPAEC) and MALDI-TOF-MS analysis of XT products sequentially digested with ␤-glucosidase and EGII. The structural characterization strategy consisted of three major steps, each of which involved the parallel analysis of radiolabeled and unlabeled products generated by enzymes with a combination of HPAEC and MALDI-TOF-MS.
Radiolabeled reaction products were partitioned by HPAEC and fractionated into 250-l aliquots. The radioactivity of each fraction was determined and plotted against time to generate a radioactivity profile that was used to monitor the distribution and amount of radioactivity before and after each enzyme treatment. Unlabeled reaction products were partitioned by HPAEC, desalted, and fractionated into 500-l  aliquots. Each fraction was analyzed by MALDI-TOF-MS to determine the presence and proportion of each reaction species present at a specific elution time interval.
In the first step, five XT reactions, with or without radiolabel, were pooled (total volume of ϳ1.25 ml), dried down, and suspended in 100 l of water; 25 l of each sample was then analyzed by either HPAEC or HPAEC followed by MALDI-TOF-MS. In the second step, 75-l aliquots of radioactive or nonradioactive XT products were spiked with 10 mol of cellohexaose, dried down, suspended in 75 l of hydrolytic enzyme buffer (25 mM sodium acetate and 0.002% thimerosal (pH 5.0)) containing 50.0 milliunits of ␤-glucosidase, and incubated overnight at 37°C. Cellohexaose was added as an internal control to verify complete ␤-glucosidase digestion. Reactions were terminated by heating at 100°C for 10 min, and insoluble material was pelleted by centrifugation at 14,000 ϫ g for 5 min. The supernatant was removed, and a 25-l aliquot was analyzed by HPAEC followed by either liquid scintillation counting or MALDI-TOF-MS. In the third step, 10 mol of cellohexaose and XXXG were added to the remaining 50-l aliquot of radioactive products as an internal control to verify com-plete EGII digestion and as an internal standard for the purpose of calculating the R F value of each radioactive peak, respectively. For the remaining 50 l of nonradioactive products, 10 mol of cellohexaose was added as an internal standard to verify complete EGII digestion. Each aliquot was dried down and suspended in 50 l of hydrolytic enzyme buffer containing 5.0 milliunits of EGII and incubated overnight at 37°C. Reactions were terminated by heating at 100°C for 10 min, and insoluble material was pelleted by centrifugation at 14,000 ϫ g for 5 min. The supernatant was removed, and a 25-l aliquot was analyzed by HPAEC followed by either liquid scintillation counting or MALDI-TOF-MS.
High pH Anion-exchange Chromatography Instrumentation and Conditions-The XT reaction products were analyzed using a Dionex DX300 carbohydrate analysis system equipped with a CarboPac PA10 column (4 ϫ 250 mm) and an ED 50 for pulsed amperometric detection following methods that were adapted from previously published studies (16,17). Reaction products (25 l/injection) were eluted at a flow rate of 1 ml/min with a gradient of 50 mM sodium acetate and 100 mM sodium hydroxide (0 min) to 100 mM sodium acetate and 100 mM sodium hydroxide (20 min). The column was washed with 300 mM sodium hydroxide (15 min) and re-equilibrated with 50 mM sodium acetate and 100 mM sodium hydroxide (15 min) after each injection. For HPAEC analysis of radiolabeled XT products, each injection was fractionated into 250-l aliquots and mixed with 2.0 ml of scintillation counting fluid, and the radioactivity was determined by scintillation counting. Non-radiolabeled XT reaction products destined for MALDI-TOF-MS analysis were partitioned by HPAEC, desalted with a Dionex in-line desalting module following the manufacturer's instructions, fractionated into 500-l aliquots, and dried down.
MALDI-TOF-MS-XT reaction products were analyzed by MALDI-TOF-MS based upon previously published methods (18). Briefly, desalted fractions from the HPAEC-partitioned reaction products were dried down, suspended in 2 l of 2,5dihydroxybenzoic acid dissolved in acetonitrile (10 mg/ml), loaded onto the MALDI-TOF-MS plate, and allowed to crystallize. XT reaction products were analyzed with a Voyager-DE STR MALDI-TOF mass spectrometer (Applied Biosystems, Foster City, CA) in reflector mode with a 100-ns extraction time and a 20-kV accelerating voltage. Each spectrum consisted of data accumulated over 50 laser shots.

Heterologous Expression of AtXT1 and AtXT2-AtXT1 and
AtXT2 were expressed in two insect cell lines (Drosophila S2 and Sf21) as an alternative to heterologous expression in P. pastoris. The expression of AtXT1 or AtXT2 was verified by immunoblot analyses of S2 and Sf21 cell extracts with an antibody raised against a peptide fragment of AtXT1 (Fig. 1). Because immunoblot analysis verified the presence of AtXT1 and AtXT2 proteins in both expression systems, detergent-solubilized extracts were assayed for XT activity.
In the presence of either cellopentaose or cellohexaose as the acceptor substrate, detergent-solubilized extracts from both S2 and Sf21 cells expressing either AtXT1 or AtXT2 showed significant levels of XT activity (Fig. 2, A and B). Although both enzymes utilized either cellopentaose or cellohexaose as an acceptor, the activities of both enzymes were 4-fold higher with cellohexaose as the acceptor compared with cellopentaose. Therefore, further biochemical studies on AtXT1 and AtXT2 activities were conducted using cellohexaose as the acceptor substrate. Furthermore, because we wanted large quantities of material for biochemical characterization of XT activity, we utilized Sf21 cells expressing either AtXT1 or AtXT2 for the experiments described below.
Preliminary Characterization of AtXT1 and AtXT2 Enzymatic Activities Produced in Sf21 Cells-In initial studies, we used an XT assay with a 1:10 donor/acceptor substrate ratio (0.4 mM UDP-xylose and 4.0 mM cellopentaose or cellohexaose). However, HPAEC analysis of AtXT1 and AtXT2 products conducted at these substrate ratios showed that a significant amount of cellohexaose was not xylosylated (data not shown). Experiments to increase the amount of xylose incorporated into products demonstrated that a 1:1 substrate ratio (1.0 mM UDP-xylose and 1.0 mM cellohexaose) increased the amount of xylose incorporated and decreased the amount of cellohexaose remaining after 1 h compared with the previously published assay conditions (11) (data not shown). Because we were more interested in defining the products of these enzymes than in analyzing the kinetic parameters of each enzyme, the latter reaction conditions were adopted as the standard XT assay conditions and used for further characterization of AtXT1 and AtXT2 substrate specificities and the products generated. The specificity of AtXT1 and AtXT2 was investigated using various sugar nucleotide donors and oligosaccharide acceptors. AtXT1 and AtXT2 are specific for cellooligosaccharides with a degree of polymerization of Ͼ4 (Fig. 3, A  and B). Cello-oligosaccharides    (20 -24). However, the previous studies were done with enzyme preparations in which both XyG ␤-glucan synthase and XT activities were present, so it was unclear whether the ␤-glucan synthase, XT, or both required Mn 2ϩ . Using AtXT1 and AtXT2 produced in Sf21 cells, we determined that AtXT1 and AtXT2 activities required Mn 2ϩ (supplemental Fig. S1A); the maximum amount of xylose incorporation was achieved over a broad concentration range of 3-15 mM Mn 2ϩ (supplemental Fig. S1B). Finally, like many glycosyltransferases, both AtXT1 and AtXT2 activities were inhibited by UDP (supplemental Fig. S2).
We further investigated the effects of varying the time of incubation and substrate ratio on xylose incorporation. Although the amount of xylose incorporated into products was nearly linear during a 1-h assay at a 1:1 substrate ratio (data not shown), time course experiments demonstrated that xylose incorporation increased between 2-and 3-fold after 18 h (Table 1 and supplemental Fig. S3). Xylose incorporation was affected not only by reaction time, but also by an increase in the substrate ratio from 1:1 to 2.5:1 (2.5 mM UDP-xylose and 1.0 mM cellohexaose) ( Table 1). Because incubation time and substrate ratio had significant effects on xylose incorporation, we sought to determine whether the variation of these conditions had an affect on the structure of the products generated.
Biochemical Characterization of AtXT1 and AtXT2 Reaction Products-Previous work utilized a combination of carbohydrate-specific hydrolases and size-exclusion chromatography to show that AtXT1 catalyzes the addition of a xylosyl residue to cellopentaose to produce primarily GXGGG (11). In this study, we used HPAEC and MALDI-TOF-MS in conjunction with the substrate specificities of ␤-glucosidase and EGII to determine 1) the structure of the XT products generated when cellohexaose was used as the acceptor and 2) whether the product structure varied when the substrate ratio and reaction time were varied. Although product characterization is shown for only one of the two enzymes (AtXT1 in Figs. 4 and 6 -8 and AtXT2 in Fig. 5), all analyses were performed on products generated with both enzymes, and similar results were obtained regardless of which enzyme was used.
The first question we sought to answer was whether AtXT1 and AtXT2 have the ability to add more than one xylosyl residue to cellohexaose. HPAEC analysis of AtXT1 and AtXT2 products generated at the substrate ratios and incubation times presented in Table 1 revealed the presence of radiolabeled peaks that eluted after cellohexaose (Fig. 4). The total amount of radioactivity incorporated and the proportion of radioactivity associated with each peak were dependent upon the time of incubation and substrate ratio. Parallel XT reactions containing only unlabeled UDP-xylose were performed, and the products were separated by HPAEC, fractionated, and analyzed by MALDI-TOF-MS to determine the masses of the oligosaccharides present in each peak. MALDI-TOF-MS analysis of prod- ucts generated during a 1-h reaction at a 1:1 substrate ratio (the same as the reaction shown Fig. 4A) contained mainly monoxylosylated cellohexaose along with smaller amounts of unreacted cellohexaose and dixylosylated cellohexaose (data not shown).
Analysis of both labeled (Fig. 4B) and unlabeled (Fig. 5A) reactions conducted for 18 h at a 2.5:1 substrate ratio revealed a more complex mixture of products. MALDI-TOF-MS of HPAEC-partitioned unlabeled products demonstrated that the first peak contained mainly monoxylosylated cellohexaose (Fig.  5B); that the second peak contained mainly dixylosylated cellohexaose (Fig. 5C); and that the small shoulder on the second peak contained the same products as the second peak, but with a significantly higher proportion of trixylosylated cellohexaose (Fig. 5D). These results show that AtXT1 and AtXT2 are capable of adding more than one xylosyl residue to cellohexaose and that the three products produced at a 2.5:1 substrate ratio during 18-h reactions (Figs. 4B and 5A) are mono-, di-, and trixylosylated cellohexaose, respectively.
We next sought to determine the xylosylation patterns of XT reaction products. We first investigated the position of the xylosyl residue closest to the nonreducing end by digestion of the reaction products with ␤-glucosidase. This enzyme will hydrolyze unsubstituted glucosyl residues from the nonreducing end of ␤-(1,4)-linked oligosaccharides, but cannot remove a substituted glucosyl residue. In reactions containing mainly monoxylosylated products (e.g. as shown in Fig. 4A at a 1:1 substrate ratio and a 1-h incubation), ␤-glucosidase treatment produced predominantly XGGG (data not shown), thereby leading to the conclusion that the primary product generated under these reaction conditions was GGXGGG. When a more complex mixture of products generated during an 18-h reaction at a 2.5:1 substrate ratio (as shown in Figs. 4B and 6A) was treated with ␤-glucosidase and analyzed by HPAEC, a significant shift in the distribution of radioactivity was observed (Fig.  6, compare A and B). MALDI-TOF-MS analyses of a parallel experiment lacking radiolabel indicated that the main peak consisted primarily of XGGG (Fig. 7, A and B), whereas the shoulder contained mainly dixylosylated cellotetraose (Fig. 7, B and C). The second peak of ␤-glucosidase-digested products (Fig. 6B, peak B-II) contained XGGGG (Fig. 7, D and E), XGGGGG (Fig. 7F), dixylosylated cellopentaose (Fig. 7, D-F), and trixylosylated cellopentaose (Fig. 7, E and F). Although the amount of radioactivity associated with the elution time interval of 7.0 to 9.0 min is very small (Fig. 6B), MALDI-TOF-MS analyses indicated the presence of ions that corresponded to XGG and dixylosylated cellotriose (data not shown).
The presence of ions corresponding to the monoxylosylated products XGG, XGGG, XGGGG, and XGGGGG from ␤-glucosidase-digested XT products indicated that AtXT1 and AtXT2 were capable of generating GGGXGG, GGXGGG, GXGGGG, and XGGGGG, respectively. Furthermore, the HPAEC analysis of ␤-glucosidase-treated XT products presented in Fig. 6B demonstrated that the amount of radioactivity associated with XGGG was significantly greater compared with all other species of ␤-glucosidase products combined, thereby leading to the conclusion that the primary product generated under all reaction conditions was GGXGGG. Although this scheme identified the first xylose-substituted glucosyl residue from the nonreducing end of the XT products, it had a significant drawback in that it could not resolve the structures of reaction products containing more than one xylose substitution.
To solve this problem, ␤-glucosidase-digested reaction products were subsequently treated with EGII, an endoglucanase that will hydrolyze the glycosidic linkage of unsubstituted ␤-(1,4)-linked glucosyl residues. The products resulting from sequential digestion with both enzymes were analyzed by HPAEC (Fig. 6C) and MALDI-TOF-MS (Fig. 8). HPAEC anal-ysis revealed three radioactive peaks with R F values (Fig. 6C) that corresponded to previously published values for XG, XXG, and XXXG (17). MALDI-TOF-MS analyses of separated unlabeled products showed the presence of ions with m/z ratios consistent with these peak assignments (Fig. 8).
These results allowed us to deduce the structures of the multixylosylated reaction products identified by MALDI-TOF-MS in Figs. 5 and 6. The structure of the dixylosylated cellotetraose that composed the shoulder of XGGG in Fig. 6B was XXGG; the structures of the di-and trixylosylated cellopentaose that were identified as components of the second peak in Fig. 6B were XXGGG and XXXGG, respectively. On the basis of these deductions, we concluded that the original XT multixylosylated products consisted primarily of GGXXGG, with minor amounts of GXXGGG and GXXXGG. Although we could not completely discount the presence of minor amounts of XT products with alternating xylose substitution patterns (such as GGXGXG and GXGXGG), significant amounts of these reaction products could not have been present given the proportion of radioactivity associated with XG after EGII digestion (Fig. 6C) compared with the proportion of radioactivity associated with either XGGG after ␤-glucosidase digestion or XT-generated GGXGGG (Fig. 6A). Therefore, these results provided strong evidence that AtXT1 and AtXT2 both added the first xylosyl residue primarily to the fourth glucose from the reducing end to produce GGXGGG and then added the second xylosyl residue to the adjacent glucosyl residue to form predominately GGXXGG.

DISCUSSION
Early biochemical studies demonstrated that microsomal preparations from peas (20,(23)(24)(25)(26) and soybeans (21,(27)(28)(29) have the ability to synthesize XyG in vitro when a mixture of UDP-glucose and UDP-xylose are added as a substrate. Under these conditions, the microsomal membrane preparations are capable of producing nascent XyG fragments that have the proper substitution patterns (21,23,26). However, solubilization with detergents and further biochemical purification of either the XyG ␤-glucan synthase or XyG XT were not successful, suggesting that XT and ␤-glucan synthase formed a complex in the Golgi membranes, where their combined activities synthesized the core XyG molecule consisting of repeating subunits.
Recently, Faik et al. (11) demonstrated that XT activity can be detected in detergent-solubilized pea microsomes if a cellooligosaccharide acceptor is added. Using this in vitro assay, Faik et al. screened a series of P. pastoris lines transformed with one of seven Arabidopsis candidate XT genes and determined that one gene, AtXT1 (At3g62720), encodes a protein with XT activity. In this study, we have shown that another member of the gene family, AtXT2 (At4g02500; formerly AtGT2), encodes a protein with XT activity. We demonstrated that, in vitro, AtXT1 and AtXT2 exhibited similar preferences for the location of xylose addition to cellohexaose. Furthermore, we have shown that AtXT1 and AtXT2 could catalyze the addition of multiple xylosyl residues to cellohexaose.
In the previous study (11), it was unclear whether the inability to detect XT activity in P. pastoris cell lines transformed with the other members of this gene family was due to the lack of protein production, the generation of inactive proteins, or an inadequate enzyme assay. Because we were able to demonstrate that AtXT2 encodes a protein with XT activity and exhibits similar substrate specificity and preferences for the location of xylose addition to cellohexaose as AtXT1, we believe that the most likely explanation for the earlier results (11) is that AtXT2 was produced in P. pastoris.
In vitro XT assays indicated that either AtXT1 or AtXT2 used either cellopentaose or cellohexaose as an acceptor, with cellohexaose acting as the superior substrate. In this study, we could not detect the low level of activity seen previously when cellotetraose was used as the acceptor for AtXT1 (11). However, it is interesting to note that cellotetraose was also not an acceptor for the enzyme solubilized from pea Golgi membranes (11).
Previous work has shown that GXGGG is the predominant reaction product formed by AtXT1 when cellopentaose is used as the acceptor substrate (11). A comparison of the reaction products generated by either AtXT1 or AtXT2 showed that both enzymes preferentially added a xylosyl residue to the fourth glucosyl residue from the reducing end of the acceptor, producing GXGGG from cellopentaose and GGXGGG from cellohexaose.
In contrast to the high specificity of XyG biosynthetic enzymes AtFUT1 (8,9) and MUR3 (7), which add only a single sugar to their acceptor substrates, both AtXT1 and AtXT2 are capable of adding multiple xylosyl residues to cellohexaose. Further xylosylation of GGXGGG could be achieved when longer reaction times and/or higher substrate ratios were used with either AtXT1 or AtXT2; both enzymes preferentially FIGURE 8. MALDI-TOF-MS analysis of reaction products sequentially digested with ␤-glucosidase and EGII. AtXT1 reaction products synthesized during an 18-h incubation in the presence of 5 mM MnCl 2 , 2.5 mM unla-beled UDP-xylose, and 1.0 mM cellohexaose were sequentially digested with ␤-glucosidase and EGII. Digested reaction products were separated by HPAEC, desalted with an in-line desalting module, and fractionated into 500-l aliquots. Each aliquot was analyzed by MALDI-TOF-MS.