Identification and Biochemical Characterization of Four Wood-Associated Glucuronoxylan Methyltransferases in Populus

Wood is one of the promising bioenergy feedstocks for lignocellulosic biofuel production. Understanding how wood components are synthesized will help us design strategies for better utilization of wood for biofuel production. One of the major wood components is xylan, in which about 10% of xylosyl residues are substituted with glucuronic acid (GlcA) side chains. All the GlcA side chains of xylan in wood of Populus trichocarpa are methylated, which is different from Arabidopsis xylan in which about 60% of GlcA side chains are methylated. Genes responsible for methylation of GlcA side chains in Populus xylan have not been identified. Here, we report genetic and biochemical analyses of four DUF579 domain-containing proteins, PtrGXM1, PtrGXM2, PtrGXM3 and PtrGXM4, from Populus trichocarpa and their roles in GlcA methylation in xylan. The PtrGXM genes were found to be highly expressed in wood-forming cells and their encoded proteins were shown to be localized in the Golgi. When overexpressed in the Arabidopsis gxm1/2/3 triple mutant, PtrGXMs were able to partially complement the mutant phenotypes including defects in glucuronoxylan methyltransferase activity and GlcA methylation in xylan, indicating that PtrGXMs most likely function as glucuronoxylan methyltransferases. Direct evidence was provided by enzymatic analysis of recombinant PtrGXM proteins showing that they possessed a methyltransferase activity capable of transferring the methyl group onto GlcA-substituted xylooligomers. Kinetic analysis showed that PtrGXMs exhibited differential affinities toward the GlcA-substituted xylooligomer acceptor with PtrGXM3 and PtrGXM4 having 10 times higher K m values than PtrGXM1 and PtrGXM2. Together, these findings indicate that PtrGXMs are methyltransferases mediating GlcA methylation in Populus xylan during wood formation.


Introduction
Terrestrial plants fix about 56 billion metric tons of carbon annually, of which nearly half is stored in wood, the most abundant plant biomass [1]. Therefore, wood is an important reservoir for fixed carbon and plays a significant role in the regulation of atmospheric CO 2 level. Furthermore, wood is a raw material vital for numerous applications, such as burning for energy, pulping and paper-making, construction, and furnituremaking, and it is a renewable source for biofuel production [2]. Because of the immense role wood plays in our daily life, tremendous efforts have been devoted into understanding how wood is synthesized in order to develop molecular and genetic tools for custom-designing wood composition tailored for diverse end uses [3].
Wood is mainly composed of three wall polymers, i.e., cellulose, hemicelluloses, and lignin, the proportion of which varies among different tree species. Cellulose, consisting of linear chains of b-1,4linked glucosyl residues, is the predominant constituent ranging from 41% to 51% in softwood and hardwood from gymnosperms and angiosperms, respectively [4]. Genes encoding cellulose synthase catalytic subunits involved in wood formation have been identified and functionally characterized in tree species [5]. Lignin is a complex polyphenolic polymer, the content of which varies from 25 to 35% in softwood and 18 to 25% in hardwood [4]. Genes encoding enzymes in the phenylpropanoid pathway leading to monolignol biosynthesis have been identified in tree species and transgenic trees with altered expression of lignin biosynthetic genes exhibit reduced lignin content and/or altered lignin composition [6].
The third major wood component is hemicellulose, which consists mainly of xylan and glucomannan. In softwood, glucomannan is the predominant hemicellulose (about 20% of wood) and xylan is only half as abundant, whereas in hardwood, xylan is the predominant hemicellulose (ranging from 20 to 35% of wood) and glucomannan is a minor component (about 3% of wood) [4]. Glucomannan is composed of b-1,4-linked glucosyl and mannosyl residues, and genes encoding mannan synthases responsible for glucomannan biosynthesis have been identified and biochemically characterized in several tree species, such as Populus and pine [7,8]. Xylan from hardwood consists of a linear chain of b-1,4-linked xylosyl residues, to which 4-O-methylglucuronic acid (MeGlcA) side chains are attached at O-2 [4]. Hardwood xylan is often heavily acetylated; 40 to 60% of xylosyl residues contain acetyl groups at O-2 and/or O-3, depending on the tree species [9]. In addition, the reducing end of hardwood xylan from birch and Populus contains a tetrasaccharide sequence, b-D-Xyl-(1R3)-a-L-Rha-(1R2)-a-D-GalA-(1R4)-D-Xyl, that is distinct from the xylosyl backbone [10,11]. The biosynthesis of xylan requires a suite of enzymes that are responsible for the xylan backbone elongation, side chain addition and modification, and the synthesis of the tetrasaccharide reducing end sequence [12][13][14][15][16][17][18]. Our understanding of genes involved in xylan biosynthesis during wood formation in tree species is still limited.
Early genomic studies of wood formation in poplar led to the first identification of a number of glycosyltransferase genes, belonging to families GT2, GT8, GT43 and GT47, that are potentially involved in the biosynthesis of wood components [12]. Further molecular and biochemical analyses of some of these wood-associated glycosyltransferase genes have demonstrated that family GT43 genes in Populus are functional orthologs of Arabidopsis GT43 genes and they form two functionally nonredundant groups responsible for the elongation of xylan backbone [13][14][15]. Several other Populus wood-associated glycosyltransferase genes, PoGT8D, PoGT8E/PoGT8F and PoGT47C, have been shown to be functional orthologs of Arabidopsis IRX8, PARVUS and FRA8, respectively, that are involved in the biosynthesis of the xylan tetrasaccharide reducing end sequence but their exact biochemical functions have not been revealed [11,14,[16][17][18]. Genes responsible for glucuronic acid (GlcA) substitution, GlcA methylation, and acetylation of xylan in tree species have not been characterized. In Arabidopsis, three GUX genes encode glucuronyltransferases mediating GlcA substitution of xylan [19,20] and three GXM genes encode methyltransferases catalyzing GlcA methylation of xylan [21,22]. Acetylation of Arabidopsis xylan involves RWAs and ESK1 [23,24,25].
In this study, we performed functional analysis of four Populus genes encoding DUF579 domain-containing proteins that are induced by PtrWND2B, a wood-associated master transcriptional regulator controlling secondary wall biosynthesis [26,27]. These PtrWND2B-induced genes, namely PtrGXM1, PtrGXM2, PtrGXM3, and PtrGXM4, are highly expressed in developing wood in Populus stems and their encoded proteins are localized in the Golgi. PtrGXMs are close homologs of Arabidopsis GXMs that have recently been demonstrated to encode methyltransferases catalyzing GlcA methylation of xylan [21,22] and they were able to partially restore the 4-O-methylation of GlcA in xylan when overexpressed in the Arabidopsis gxm1/2/3 triple mutant. Moreover, recombinant PtrGXMs exhibit a methyltransferase activity transferring the methyl group from the methyl donor Sadenosylmethionine to GlcA-substituted xylooligomer acceptors. These results demonstrate that PtrGXMs are wood-associated methyltransferases responsible for GlcA methylation of Populus xylan.

Results
Microsomes from Populus developing wood exhibit glucuronoxylan methyltransferase activity As a first step toward molecular and biochemical characterization of methyltransferases responsible for GlcA methytlation in xylan in Populus, we examined glucuronoxylan methyltransferase activity in developing wood of Populus. Microsomes isolated from Populus stems were incubated with S-adenosyl-L-methionine, radiolabeled on the methyl group, as the methyl donor and GlcA-substituted Xyl 4 as the acceptor. While the microsomes exhibited no methyltransferase activity in the absence of the exogenous acceptor, a high level of methyltransferase activity was detected in the presence of the GlcA-substituted Xyl 4 acceptor (Fig. 1A). This methyltransferase activity was both time-and protein concentration-dependent (Fig. 1B, C). The reaction products were verified using matrix-assisted laser desorption ionization-time-of-flight mass spectrometry (MALDI-TOF-MS). The control reaction with the addition of the acceptor but no methyl donor exhibited an ion peak [M+Na] + at mass-to-charge ratio (m/z) 745 that corresponds to the acceptor, GlcA-substituted Xyl 4 (Fig. 1D). In the reaction with the addition of both the methyl donor and the acceptor, a signal peak at m/z 759 corresponding to MeGlcA-substituted Xyl 4 was evident in addition to the GlcAsubstituted Xyl 4 acceptor peak (m/z 745) (Fig. 1D). These results demonstrate that the Populus stem microsomes possess a methyltransferase activity capable of transferring the methyl group from S-adenosyl-L-methionine onto the GlcA-substituted xylooligosaccharide acceptor.

The expression of PtrGXM genes is PtrWND2-induced and wood-associated
To search for genes that are responsible for the above observed glucuronoxylan methyltransferase activity in Populus stems, we first analyzed the expression patterns of Populus DUF579 genes that are close homologs of Arabidopsis GXM genes. There are 11 DUF579 genes ( Fig. 2A) in the Populus genome, of which four are closely grouped together with the Arabidopsis GXM1, GXM2 and GXM3 genes known to encode glucuronoxylan methyltransferases. These four Populus DUF579 genes (namely PtrGXM1/2/3/ 4) apparently fall into two pairs (PtrGXM1/PtrGXM2 and PtrGXM3/PtrGXM4), each of which was likely originated from genome duplication [28]. All four PtrGXM genes were found to be highly expressed in Populus stems undergoing secondary growth (Fig. 2B). In addition, the expression of PtrGXM1, PtrGXM2 and PtrGXM3 was shown to be significantly upregulated (Fig. 2C) in transgenic Populus plants overexpressing PtrWND2B, which is a transcriptional master switch activating secondary wall biosynthesis during wood formation [26,27]. In situ hybridization further demonstrated that the PtrGXM genes were predominantly expressed in cells undergoing secondary wall thickening, including developing secondary xylem and phloem fibers (Fig. 3).

PtrGXM proteins are targeted to the Golgi
To further substantiate their involvement in xylan biosynthesis, we exmined whether PtrGXM proteins are targeted to the Golgi, in which xylan biosynthesis occurs. PtrGXMs are typical type II membrane proteins containing one transmembrane helix as predicted by the TMHMM2.0 program for prediction of transmembrane helices in proteins (http://www.cbs.dtu.dk/ service/TMHMM-2.0) (Fig. 4A). To examine the subcellular localization of PtrGXMs, we coexpressed in Arabidopsis protoplasts yellow fluorescence protein-tagged PtrGXMs together with cyan fluorescence protein-tagged FRA8, a family GT47 glycosyltransferase known to be located in the Golgi [29]. While the control protoplasts expressing YFP alone showed fluorescent signals throughout the cytoplasm (Fig. 4B, C), the protoplasts expressing PtrGXM-YFP exhibited punctate signals, which were colocalized with the FRA8-CFP signals ( Fig. 4D-S). These results demonstrate that PtrGXMs are localized in the Golgi, where xylan biosynthesis occurs.

PtrGXMs are functional orthologs of Arabidopsis GXMs
The wood-associated expression of PtrGXMs and their Golgilocalization indicate that PtrGXMs most likely play a role in the biosynthesis of xylan. Considering the fact that the four PtrGXMs are phylogenetically closely grouped together with the Arabidopsis glucuronoxylan methyltransferases GXM1/2/3, we hypothesized that PtrGXMs are also glucuronoxylan methyltransferases involved in the methylation of GlcA side chains of xylan. Therefore, we tested the ability of PtrGXMs to complement the Arabidopsis gxm1/2/3 triple mutant. The full-length PtrGXM cDNAs driven by the cauliflower mosaic virus (CaMV) 35S promoter were transformed into the gxm1/2/3 mutant, and more than 64 independently transgenic lines for each construct were generated. The expression of PtrGXMs in 8 representative transgenic lines for each construct was confirmed by reverse transcription PCR analysis of PtrGXM transcripts (data not shown). Microsomes were isolated from stems of six-week-old transgenic plants and subsequently used for glucuronoxylan methyltransferase activity assays. The glucuronoxylan methyltransferase activities from 8 representative transgenic lines for each expression construct of PtrGXMs were shown in Fig. 5. It was evident that although the gxm1/2/3 mutant exhibited a loss of glucuronoxylan methyltransferase activity, the gxm1/2/3 lines expressing PtrGXMs showed a partial restoration of glucuronoxylan methyltransferase activity.
To find out whether expression of PtrGXMs restored the GlcA methylation of xylan in the gxm1/2/3 mutant, we isolated xylan from the pooled stems of the 8 transgenic lines as shown in Fig. 5 for each expression contruct of PtrGXMs. The isolated xylan was digested with xylanase to generate xylooligosaccharides, which were subsequently subjected to MALDI-TOF-MS (Fig. 6). Xylooligosaccharides from wild-type xylan exhibited a prominent ion peak at m/z 759, which corresponds to MeGlcA-substituted Xyl 4 , and a less prominent ion peak at m/z 745, which corresponds to GlcA-substituted Xyl 4 . Simultaneous mutations of the Arabidopsis GXM1, GXM2 and GXM3 genes in the gxm1/2/3 triple mutant resulted in a complete loss of the ion peak (m/z 759) attributed to MeGlcA-substituted Xyl 4 . As a result, the GlcA-substituted Xyl 4 peak at m/z 745 was significantly elevated in the gxm1/2/3 mutant compared with that in the wild type. Expression of PtrGXMs in gxm1/2/3 led to a partial restoration of the ion peak at m/z 759 corresponding to MeGlcA-substituted Xyl 4 (Fig. 6). The partial restoration of the GlcA methylation was further verified by 1 H nuclear magnetic resonance (NMR) spectroscopy (Fig. 7). The wild-type xylan displayed resonances characteristic of H1 of GlcA and MeGlcA side chains at 5.31 and 5.29 ppm, respectively, and H5 of GlcA and MeGlcA side chains at 4.42 and 4.39 ppm, respectively. It also exhibited resonances at 4.63 and 4.46 ppm that are attributed to H1 of branched and unbranched backbone xylosyl residues, respectively. Xylan from the gxm1/2/3 mutant lacked all the resonances attributed to MeGlcA side chains, including H1 of MeGlcA, H5 of MeGlcA, and H1 of MeGlcAbranched b-Xyl. Analysis of xylan from gxm1/2/3 lines expressing PtrGXMs revealed a partial restoration of all the resonances attributed to MeGlcA side chains (Fig. 7). Integration analysis of the GlcA and MeGlcA signal peaks revealed that although the gxm1/2/3 mutant completely lacks GlcA methylation in xylan, 8% to 21% of the GlcA side chains in xylan were methylated in gxm1/ 2/3 expressing PtrGXMs ( Table 1). The resonance intensities for the xylan reducing end tetrasaccharide sequence, b-D-Xyl-(1R3)a-L-Rha-(1R2)-a-D-GalA-(1R4)-D-Xyl (H1 of 3-linked b-D-Xyl, H1 of a-L-Rha, H1 of a-D-GalA, H2 of a-L-Rha, and H4 of a-D-GalA), were unaltered in gxm1/2/3 expressing PtrGXMs compared with the gxm1/2/3 mutant and the wild type. Together, these results provide genetic evidence demonstrating that expression of each of the four PtrGXM genes in the gxm1/2/3 mutant is able to partially restore the glucuronoxylan methyltransferase activity and the GlcA methylation in xylan, indicating that PtrGXMs are functional orthologs of the Arabidopsis GXMs catalyzing the methylation of GlcA side chains in xylan.

Recombinant PtrGXM proteins exhibit glucuronoxylan methyltransferase activity
To provide definitive proof that PtrGXMs are glucuronoxylan methyltransferases, we examined the glucuronoxylan methyltransferase activity of purified recombinant PtrGXM proteins. PtrGXM1, PtrGXM2, PtrGXM3, and PtrGXM4 fused with maltose binding protein were successfully expressed in and purified from E. coli (Fig. 8A). Recombinant PtrGXMs were incubated with radiolabeled S-adenosyl-L-methionine as the methyl donor and GlcA-substituted Xyl 4 as the acceptor. It was found that all four recombinant PtrGXMs exhibited a methyltransferase activity capable of transferring the radiolabeled methyl group onto the GlcA-substituted Xyl 4 acceptor (Fig. 8B). In contrast, they were unable to transfer the radiolabeled methyl group onto unsubstituted Xyl 4 , GlcA, or UDP-GlcA. Purified maltose binding protein alone did not show any methyltransferase activity toward any of the substrates tested. The PtrGXM-catalyzed reaction products were further analyzed by MALDI-TOF-MS. It was evident that in addition to the prominent ion peak at m/z 745 attributed to the GlcA-substituted Xyl 4 acceptor, the PtrGXMcatalyzed reaction products had another noticeable ion peak at m/ z 759 corresponding to MeGlcA-substituted Xyl 4 (Fig. 9). The control reaction incubated with maltose binding protein contained only the ion peak for the acceptor (m/z 745). These results unequivocally demonstrate that PtrGXM1, PtrGXM2, PtrGXM3, and PtrGXM4 are glucuronoxylan methyltransferases catalyzing the transfer of the methyl group onto GlcA side chains in xylan.
We further analyzed the biochemical properties of the recombinant PtrGXMs. The glucuronoxylan methyltransferase activities of PtrGXMs were both time-and protein concentrationdependent (Fig. 10A, B). The optimal temperature for the reactions was 37uC for PtrGXM1 and PtrGXM2 and 45uC for PtrGXM3 and PtrGXM4 (Fig. 10C). To examine the substrate binding affinities of PtrGXMs, we investigated their kinetic properties using different concentrations of the GlcA-substituted Xyl 4 acceptor. The K m and V max values were calculated by Lineweaver-Burk analysis. It was found that the transfer of the methyl group onto GlcA-substituted Xyl 4 by PtrGXMs was dependent on the acceptor concentration ( Fig. 11) with apparent K m values of 3.25, 3.98, 0.30 and 0.29 mM and V max of 145, 123, 353 and 737 pmol/min/mg protein for PtrGXM1, PtrGXM2, PtrGXM3, and PtrGXM4, respectively. These results indicate that PtrGXM3 and PtrGXM4 exhibit much higher substrate affinities for the GlcA-substituted Xyl 4 acceptor than PtrGXM1 and PtrGXM2.

Discussion
Populus is one of the many plant species that have been considered for use as lignocellulosic biofuel feedstocks [2]. The bulk of Populus feedstock is wood, which is mainly composed of cellulose (43-48%), xylan (18-28%) and lignin (19-21%) [30]. Therefore, a complete understanding of how these wood components are synthesized will be instrumental for a better utilization of this abundant feedstock for biofuel production. Previous studies have demonstrated that a number of genes, belonging to GT43, GT47, and GT8 families, are involved in the biosynthesis of the backbone and the reducing end sequence of xylan in Populus [11,[13][14][15][16][17][18]. Our current finding that four Populus DUF579 domain-containing proteins, PtrGXM1, PtrGXM2, PtrGXM3 and PtrGXM4, are methyltransferases that mediate GlcA methylation in xylan further enriches our understanding of the biosynthesis of xylan, the second most abundant polysaccharide in dicot wood.
Several lines of genetic and biochemical evidence demonstrate that the four PtrGXMs are wood-associated glucuronoxylan methyltransferases. First, they are expressed in developing wood undergoing secondary wall thickening. Second, the expression of PtrGXMs except PtrGXM4 is activated by PtrWND2B, a master transcriptional switch regulating the secondary wall biosynthetic program during wood formation. Since there are 11 other   wood-associated PtrWND homologs in Populus [27], it is possible that PtrGXM4 expression is regulated by other PtrWNDs. Third, PtrGXMs are able to partially restore the deficiency in GlcA methylation in xylan when expressed in Arabidopsis gxm1/2/3 triple mutant. Fourth, their recombinant proteins exhibit a methyltransferase activity capable of transferring the methyl group from S-adenosyl-L-methionine onto the GlcA-substituted xylooligomer acceptor. As PtrGXMs are unable to transfer the methyl group onto GlcA or UDP-GlcA, it is clear that the methylation of GlcA occurs after it is added onto the xylan backbone. Further biochemical analysis of PtrGXMs revealed that while PtrGXM1 and PtrGXM2 share similar biochemical properties in the temperature optimum, K m and V max , PtrGXM3 is more similar to PtrGXM4, which is consistent with the phylogenetic analysis showing that PtrGXM1 and PtrGXM2 are a pair of genes likely originated from genome duplication, and so are PtrGXM3 and PtrGXM4. It is interesting to note that PtrGXM3 and PtrGXM4 have much higher substrate affinities and reaction rates than PtrGXM1 and PtrGXM2, indicating that the GlcA methylation in Populus xylan might be mainly mediated by PtrGXM3 and PtrGXM4.

(D) to (G) An Arabidopsis protoplast (D) co-expressing PtrGXM1-YFP (E) and the Golgi-localized FRA8-CFP (F). (H) to (K) An Arabidopsis protoplast (H) co-expressing PtrGXM2-YFP (I) and FRA8-CFP (J). (L) to (O) An Arabidopsis protoplast (L) co-expressing PtrGXM3-YFP (M) and FRA8-CFP (N). (P) to (S) An Arabidopsis protoplast (P) co-expressing
Although the structure of xylan, including the xylan backbone, the reducing end sequence, the degree of acetylation and the frequency of GlcA substitution, between Populus and Arabidopsis is similar [11,14], the degree of methylation of GlcA side chains differs significantly. The GlcA side chains in Populus xylan are all methylated, whereas only about 60% of the GlcA side chains in Arabidopsis xylan are methylated. It is currently unknown what factors account for the difference of the degree of GlcA methylation in xylan between Populus and Arabidopsis. Enzyme kinetic analysis of PtrGXMs revealed that although PtrGXM1 and PtrGXM2 showed K m values (3.25 and 3.98 mM, respectively) similar to that of the Arabidopsis glucuronoxylan methyltransferase GXM3/GXMT1 (K m = 3.85 mM) that is responsible for the bulk of the GlcA methylation of xylan [22], the K m values of PtrGXM3 and PtrGXM4 (0.30 and 0.29 mM, respectively) were about 8 times lower. This finding indicates that PtrGXM3 and PtrGXM4 have a much higher substrate affinity than the Arabidopsis glucuronoxylan methyltransferase, which may enable a complete methylation of the GlcA side chains in Populus xylan during wood formation. It is also possible that the amount of GXM enzymes for GlcA methylation is much higher in Populus wood than in Arabidopsis secondary wall-forming cells, and therefore sufficient glucuronoxylan methyltransferase activities are available for a complete methylation of GlcA side chains in xylan in Populus but not in Arabidopsis.
Although expression of PtrGXMs in the Arabidopsis gxm1/2/3 triple mutant restored the level of glucuronoxylan methyltransferase activities close to that of the wild type, the PtrGXMcomplemented transgenic plants only resulted in a partial restoration of GlcA methylation in xylan. This partial complementation could be due to several possible reasons. Since multiple GXMs are expressed in cells undergoing xylan biosynthesis in both Populus and Arabidopsis, it is possible that the expression of one PtrGXM in the gxm1/2/3 mutant may not be sufficient to catch up with the overall rate of xylan biosynthesis so that only a fraction of GlcA side chains are methylated. It is also possible that GlcA methylation is carried out by a protein complex containing GXMs and other xylan biosynthetic enzymes. PtrGXM sequences may have diverged from Arabidopsis GXMs to such a degree that they may not be efficiently incorporated into the Arabidopsis xylan biosynthetic enzyme complex and hence there is not sufficient GXM activity available in the complex to restore GlcA methylation during xylan biosynthesis in planta. This possibility appears to be in agreement with the observation that although expression of PtrGXMs in the gxm1/2/3 mutant only led to low levels of GlcA methylation in xylan, the glucuronoxylan methyltransferase activity as assayed in vitro, which does not require the xylan biosynthetic enzyme complex, was restored to a much higher level.
The potential use of Populus as a bioenergy feedstock for biofuel production largely depends on whether it is sufficiently economical to convert cell wall polysaccharides into fermentable sugars [2]. One of the major obstacles for efficient utilization of plant biomass for biofuel production is biomass recalcitrance, i.e., resistance of   cell walls to enzymatic digestion into sugars. Lignin and xylan, both of which form cross-linking networks in cell walls, contribute to biomass recalcitrance by blocking the access of cellulolytic enzymes to cellulose [31]. A reduction in the degree of GlcA methylation in xylan has been shown to enhance xylan release during hydrothermal treatment of cell walls, indicating that decreasing GlcA methylation in xylan could facilitate the conversion of biomass into fermentable sugars [22]. Our  identification of four glucuronoxylan methyltransferase genes in Populus provides molecular tools to modify the cell wall composition of wood so that it would be more suitable for conversion into biofuels.
There exist 11 genes encoding DUF579 domain-containing proteins in the Populus genome (Fig. 2). Our study has established that four of them, PtrGXM1, PtrGXM2, PtrGXM3, and PtrGXM4, are glucuronoxylan methyltransferases. Four other Populus DUF579 genes (Potri.005G141300, Potri.007G047000, Potri.001G302600, and Potri.009G098800) fall into the same phylogenentic subgroup as Arabidopsis IRX15 and IRX15L that have also been implicated in xylan biosythesis [32,33]. Further functional characterization of Populus DUF579 genes will undoubtedly provide new insights into our understanding of the biochemical mechanisms of xylan biosynthesis during wood formation in tree species.

Microsome isolation and methyltransferase activity assay
Microsomes were isolated from Populus trichocarpa stems or Arabidopsis inflorescence stems according to Kuroyama and Tsumuraya [38] and used for the methyltransferase activity assay [21]. The protein concentration in microsome preparations was measured using the Bio-Rad protein assay kit. Each reaction mixture (50 ml) contained 100 mM Tris-HCl, pH 8.0, 1 mM CoCl 2 , 0.5% Triton X-100, 2 mg/ml GlcA-substituted Xyl 4 isolated from the gxm2/3 double mutant, S-[ 14 C-methyl] adenosylmethionine (0.1 mCi; American Radiolabeled Chemical), and 200 mg microsomes or 20 mg recombinant PtrGXM proteins. For MALDI analysis of GXM-catalyzed reaction products, nonradiolabeled S-adenosylmethionine (1 mM) was used. After 1-h incubation, the reaction mixture was passed through AG1-X4 anion exchange resin [34]. The bound MeGlcA-substituted xylooligomers were eluted with 3N acetic acid and the eluate was counted for the amount of radioactivity with a PerkinElmer scintillation counter (for radiolabeled S-adenosylmethionine) or subjected to MALDI analysis (for non-radiolabeled S-adenosylmethionine).

MALDI-TOF MS
The reaction products from the methyltransferase activity assay or xylooligosaccharides isolated from cell walls were used for MALDI-TOF-MS according to Zhong et al. [29]. The spectrometer was operated in the positive-ion mode with an accelerating voltage of 30 kV, an extractor voltage of 9 kV, and a source pressure of approximately 8610 27 torr. The aqueous sample was mixed (1:1, v/v) with the MALDI matrix (0.2 M 2,5-dihydroxbenzoic acid and 0.06 M 1-hydroxyisoquinoline in 50% acetonitrile) and dried on a stainless steel target plate. Spectra are the average of 100 laser shots.

Gene expression analysis
Total RNA was isolated from various organs of Populus or leaves of transgenic PtrWND2B-overexpressing Populus lines [27] with a Qiagen RNA isolation kit. They were converted into first strand cDNAs, which were used as templates for real-time quantitative PCR analysis with the QuantiTect SYBR Green PCR kit (Clontech). The PCR primers for PtrGXM1 are 59-actgccatatttacagctggaatg-39 and 59-ctactctgggcaaaagggtctgtc-39; those for PtrGXM2 are 59-agaatgactgccatatatactgcc-39 and 59-ctactctgggcaaaaaggtctgtc-39; those for PtrGXM3 are 59-tgacgaggcaccagggaggatgac-39 and 59-ctaaggacaaaaaggtctgcccga-39; and those for PtrGXM4 are 59-tgatgaggctccagggagaatgaa-39 and 59-tggacaaaaaggttttcccgaact-39. The relative expression level was calculated by normalizing the PCR threshold cycle number of each gene with that of a Populus actin reference gene [11]. The data were the average of three biological replicates.

In situ mRNA hybridization
Populus stems were fixed in 2.5% formaldehyde and 0.5% glutaraldehyde, embedded in paraffin, and sectioned for in situ mRNA localization [13]. The 300-bp 39 noncoding regions of PtrGXMs were used for generation of digoxigenin-labeled antisense and sense RNA probes with the DIG RNA labeling mix (Roche). The 39 noncoding regions of PtrGXM1 and PtrGXM2 are highly conserved with 70% identity and those of PtrGXM3 and PtrGXM4 share 65% identity. There is no significant sequence identity between PtrGXM1/2 and PtrGXM3/4. Dot-blot analysis showed that the PtrGXM1 probe had a low degree of cross hybridization with PtrGXM2 RNA and the PtrGXM3 probe had a low degree of cross hybridization with PtrGXM4 RNA. However, no cross hybridization was detected between the PtrGXM1/2 probes and PtrGXM3/4 RNA. After incubation of stem sections with the antisense or sense PtrGXM probes, the hybridization signals were detected by incubating with alkaline phosphatase-conjugated antibodies against digoxygenin and subsequent color development with alkaline phosphatase substrates.

Subcellular protein localization
For subcellular localization of PtrGXMs, their full-length cDNAs were fused in frame with the yellow fluorescent protein (YFP) cDNA and ligated between the CaMV 35S promoter and the nopaline synthase terminator in pBI221 (Clontech). The YFPtagged PtrGXMs together with cyan fluorescent protein (CFP)tagged Golgi marker (FRA8) [29] were co-transfected into Arabidopsis protoplasts [35], and the fluorescence signals in transfected protoplasts were visualized using a Leica TCs SP2 spectral confocal microscope (Leica Microsystems). At least 10 fluorescence-positive protoplasts were examined and the same subcellular localization pattern was observed for each construct.

Generation of PtrGXM overexpression lines
The full-length PtrGXM cDNAs were PCR-amplified, sequenceverified and then ligated into the Xba1 site between the cauliflower mosaic virus (CaMV) 35S promoter and the nopaline synthase terminator in the pGPTV binary vector [36]. The primers for PtrGXM1 are 59-tccaatatgaggcctaaaggccaa-39 and 59ctactctgggcaaaagggtctgtc-39; those for PtrGXM2 are 59-atgaggcctaaaagccagcaa-39 and 59-ctactctgggcaaaaaggtctgtc-39; those for PtrGXM3 are 59-accatgagacccaacaaaaaccaa -39 and 59-ctaaggacaaaaaggtctgcccga-39; and those for PtrGXM4 are 59-accatgaggtccaaaaaccaatct-39 and 59-ctatggacaaaaaggttttcccga-39. An Nhe1 site was engineered at the 59 end of each primer. The resulting PtrGXM overexpression constructs were transformed into the Arabidopsis gxm1/2/3 triple mutant. Transgenic plants were selected on hygromycin, and their inflorescence stems were used for microsome isolation and cell wall extraction. For each construct, at least 64 transgenic lines were generated and stems of representative lines were analyzed for the glucuronoxylan methyltransferase activity.

Cell wall extraction
Inflorescence stems of mature plants were extracted for cell walls according to Zhong et al. [29]. Cell walls were incubated with 1N KOH to release xylan, and the isolated xylan was further digested with b-xylanase M6 (Megazyme) to produce xylooligosaccharides for structural analyses [29].

H-NMR spectroscopy
NMR spectra of the xylooligosaccharides from b-xylanase digestion were acquired at 20uC on a 600 MHz spectrometer (599.7 MHz, 1 H) using a 3 mm cryogenic triple resonance probe [37]. All NMR samples were prepared with 100% D 2 O. For all experiments, 128 transients were collected using a spectral width of 6,000 Hz and an acquisition time of 5-seconds. The residual water resonance was suppressed by a 1-second presaturation pulse at a field strength of 40 Hz. All spectra were processed with 0.2-Hz apodization followed by zero-filling to 128 k points. The 1 H NMR assignments were done by comparing them to the NMR spectra data for xylan structure [29,39].

Generation of recombinant PtrGXM proteins
The cDNAs of PtrGXM1, PtrGXM2, PtrGXM3 and PtrGXM4 without the transmembrane domain sequences were PCRamplified, sequence-verified, and then ligated in frame into the BamH1 site of the bacterial expression vector pMAL, which is tagged with a maltose binding protein (MBP) (New England BioLabs). The primers for PtrGXM1 are 59-ccgtcttcccaggaaagtcca-39 and 59-ctactctgggcaaaagggtctgtc-39; those for PtrGXM2 are 59tcatcttcccaggaaagttca-39 and 59-ctactctgggcaaaaaggtctgtc-39; those for PtrGXM3 are 59-tcatctccgagatcaaatcca-39 and 59-ctaaggacaaaaaggtctgcccga-39; and those for PtrGXM4 are 59-tctcccagaccaaattcatcc-39 and 59-ctatggacaaaaaggttttcccga-39. A BamH1 site was engineered at the 59 end of each primer. MBP-tagged recombinant PtrGXM proteins expressed in E. coli were purified using amylose resin according to the manufacturer's protocol and confirmed by SDS polyacrylamide gel electrophoresis.

Statistical Analysis
The experimental data of the quantitative PCR analysis and enzymatic activity assay were subjected to statistical analysis using the Student's t test program (http://www.graphpad.com/ quickcalcs/ttest1.cfm), and the quantitative difference between the two groups of data for comparison in each experiment was found to be statistically significant (p,0.001).