Identification of Phosphatase That Dephosphorylates Xylose in the Glycosaminoglycan-Protein Linkage Region of Proteoglycans*

Background: The enzyme responsible for the dephosphorylation of xylose in glycosaminoglycans remains unknown. Results: A protein that removed the phosphate from the phosphorylated linkage trisaccharide and localized in the Golgi was identified. Inhibition of this protein phosphatase resulted in decreased glycosaminoglycan levels in cells. Conclusion: We identified the phosphoxylose phosphatase that regulates glycosaminoglycan synthesis. Significance: Transient phosphorylation of xylose residues controls glycosaminoglycan synthesis. Recently, we demonstrated that FAM20B is a kinase that phosphorylates the xylose (Xyl) residue in the glycosaminoglycan-protein linkage region of proteoglycans. The phosphorylation of Xyl residues by FAM20B enhances the formation of the linkage region. Rapid dephosphorylation is probably induced just after synthesis of the linker and just before polymerization initiates. Indeed, in vitro chondroitin or heparan sulfate polymerization does not occur when the Xyl residue of the tetrasaccharide linkage region is phosphorylated. However, the enzyme responsible for the dephosphorylation of Xyl remains unknown. Here, we identified a novel protein that dephosphorylates the Xyl residue and designated it 2-phosphoxylose phosphatase. The phosphatase efficiently removed the phosphate from the phosphorylated trisaccharide, Galβ1–3Galβ1–4Xyl(2-O-phosphate), but not from phosphorylated tetrasaccharide, GlcUAβ1–3Galβ1–3Galβ1–4Xyl(2-O-phosphate). Additionally, RNA interference-mediated inhibition of 2-phosphoxylose phosphatase resulted in increased amounts of GlcNAcα1–4GlcUAβ1–3Galβ1–3Galβ1–4Xyl(2-O-phosphate), Galβ1–3Galβ1–4Xyl(2-O-phosphate), and Galβ1–4Xyl(2-O-phosphate) in the cells. Gel filtration analysis of the glycosaminoglycan chains synthesized in the knockdown cells revealed that these cells produced decreased amounts of glycosaminoglycan chains and that the chains had similar lengths to those in the mock-transfected cells. Transcripts encoding this phosphatase were ubiquitously, but differentially, expressed in human tissues. Moreover, the phosphatase localized to the Golgi and interacted with the glucuronyltransferase-I involved in the completion of the glycosaminoglycan-protein linkage region. Based on these findings, we conclude that transient phosphorylation of the Xyl residue in the glycosaminoglycan-protein linkage region controls the formation of glycosaminoglycan chains of proteoglycans.

Previous structural analysis of the GAG-protein linkage region revealed the presence of several modifications. One such modification is the phosphorylation of the Xyl residue at position 2 (1,6). This modification occurs in both HS and CS derived from cell-rich tissues (7,8) and appears to affect the transfer of Gal and GlcUA residues by GalT-I and GlcAT-I, respectively (9,10). Findings from in vitro experiments with authentic substrates indicate that phosphorylation of the Xyl residue prevents the transfer of a Gal residue by GalT-I (9). In contrast, GlcAT-I efficiently transfers a GlcUA residue to the phosphorylated trisaccharide in vitro (10). These results sug-gest that phosphorylation of the Xyl residue takes place after transfer of the first Gal residue by GalT-I and before transfer of the GlcUA residue by GlcAT-I. Indeed, phosphorylation of the Xyl residue is most prominent after the addition of two Gal residues (11,12). Transient phosphorylation of Xyl residues seems to enhance the formation of the linkage region (13)(14)(15). Subsequently, rapid dephosphorylation is probably induced just before initiation of polymerization. In fact, the formation of the disaccharide repeat regions of the HS and CS chains is initiated on dephosphorylated tetrasaccharide linkage structures (4,16). Therefore, dephosphorylation of the Xyl residue may be required for biosynthetic maturation of the linkage region sequence, which may be a prerequisite for the initiation and efficient elongation of the repeating disaccharide region of GAG chains.
Recently, we demonstrated that the FAM20B kinase phosphorylates the Xyl residue in the linkage region (14). Overexpression of FAM20B increased the amount of both CS and HS in HeLa cells, whereas RNA interference of FAM20B resulted in a reduction of both GAGs in the cells (14). Gel filtration analysis of the GAG chains synthesized in cells overexpressing FAM20B revealed that the GAG chains had similar lengths to those in the mock-transfected cells. These results indicate that FAM20B regulates the number of GAG chains by phosphorylating the Xyl residue in the GAG-protein linkage region of proteoglycans. However, the enzyme responsible for the dephosphorylation of Xyl remains unknown. Here, we describe the cloning of a human cDNA encoding a novel protein capable of dephosphorylating this Xyl residue. We designated this enzyme 2-phosphoxylose phosphatase (XYLP).
Construction of a Soluble Form of 2-Phosphoxylose Phosphatase-PCR was used to amplify a cDNA fragment predicted to encode a truncated form of the putative phosphatase (XYLP) lacking the first 37 N-terminal amino acids; the template cDNA was obtained from Open Biosystems (MHS1010-7508558, corresponding to the human acid phosphatase-like-2 (ACPL2) cDNA in GenBank TM BC035834); the 5Ј-primer (5Ј-GAAGATCTGGAATGAGTAGCAAGAGTCGA-3Ј) contained an in-frame BglII site, and the 3Ј-primer (5Ј-GAAGAT-CTGTGGACCTTTCCCTATGCTCT-3Ј) contained a BglII site located 38 bp downstream of the predicted stop codon. PCR was performed using KOD-Plus DNA polymerase (TOYOBO, Tokyo, Japan) for 30 cycles at 94°C for 30 s, 58°C for 30 s, and 68°C for 150 s in 5% (v/v) dimethyl sulfoxide. The PCR fragment was then subcloned into the BamHI site of pGIR201protA (18), thereby resulting in the fusion of the insulin signal sequence and the protein A sequence present in the vector. The sequences of the plasmid construct (pEF-BOS/IP-ACPL2) were confirmed by DNA sequencing.
Site-directed Mutagenesis-A soluble form of GlcAT-I was produced as described previously (10). A two-stage PCR mutagenesis method was used to construct an expression plasmid that encoded a soluble form of the GlcAT-I mutant. Two separate PCRs were performed to generate two overlapping gene fragments using a cDNA that encoded the soluble form of GlcAT-I as a template. In the first PCR, a sense 5Ј-primer (5Ј-GGAAGATCTCTACGGCAGGAAGGATCTGAGGAT-3Ј) containing a BglII site and an antisense internal mutagenic (E281A) primer (5Ј-GGCCACCTGGCGAGCAGTCTTCTG-3Ј, the mutated nucleotide is underlined) were used. In the second round of PCR, a sense internal mutagenic primer (complementary to the antisense internal mutagenic primer) and an antisense 3Ј-primer (5Ј-GGAAGATCTGATGGTG-GAGAAAAGTCCGTG-3Ј) containing a BglII site, which was located 18 bp downstream of the stop codon, were used. These two PCR products were gel-purified and then used as the template for a third PCR amplification with the sense 5Ј-primer and the antisense 3Ј-primer described previously. The final PCR fragment was subcloned into the BamHI site of pGIR201protA (18).
Expression of a Soluble Form of Phosphatase-FuGENE 6 (Roche Diagnostics) was used according to the manufacturer's instructions to transfect pEF-BOS/IP-ACPL2 (6.0 g) into COS-1 cells that were growing on 100-mm plates. For co-transfection experiments, the XYLP and GlcAT-I or GlcAT-I E281A mutant expression plasmids (3.0 g each) were co-transfected into COS-1 cells on 100-mm plates using FuGENE 6, as above. Two days after transfection, 4 ml of the culture medium was collected and incubated with 10 l of IgG-Sepharose (Amersham Biosciences) for 12 h at 4°C. The beads were recovered by centrifugation and washed with the assay buffer. The beads were then resuspended in the same assay buffer and tested for phosphatase activity as described below. To measure the amount of protein absorbed onto IgG-Sepharose beads, the bound protein was eluted with 1 M acetic acid, and BCA protein assay reagent (enhanced protocol, Pierce) was then used to measure protein content of the eluant.
Phosphatase Assays and Identification of Reaction Products-First, a phosphate transfer reaction was conducted. ␣-TM with a truncated linkage region tetrasaccharide (GlcUA␤1-3Gal␤1-3Gal␤1-4Xyl) (1 nmol) was used as an acceptor in each 20-l incubation mixture, which contained 10 l of beads bound to the soluble form of FAM20B (14) and 10 M [␥-32 P]ATP (1.11 ϫ 10 5 dpm), 50 mM Tris buffer, pH 7.0, 10 mM MnCl 2 , 10 mM CaCl 2 , and 0.1% BSA. Next, ␤-glucuronidase was used to digest the phosphorylated tetrasaccharides at 37°C in a reaction buffer containing 50 mM NaOAc and 1 mM MgCl 2 . Additionally, phosphorylated tetrasaccharides were used as acceptors in individual 20-l incubation mixtures, which contained 10 l of beads bound to a soluble form of chondroitin GalNAc transferase-1 (19) or EXTL2 (20). Then, the products of each reaction were separated by gel filtration chromatography on a Superdex peptide column that had been equilibrated with elution buffer (0.25 M NH 4 HCO 3 , 7% 1-propanol). The fractions containing the enzyme reaction products were pooled and dehydrated. The isolated reaction products were used as substrates for the phosphatase reactions. , recombinant alkaline phosphatase from bovine intestine expressed in Pichia pastoris, which was used as a control. Each of these mixtures was incubated at 37°C for 4 h, and the products of each reaction were then separated by gel filtration chromatography on a Superdex peptide column equilibrated with the elution buffer (0.25 M NH 4 HCO 3 , 7% 1-propanol). Fractions (0.4 ml each) were collected at a flow rate of 0.4 ml/min, and a liquid scintillation counter was used to measure the radioactivity.
Construction of an Expression Vector Encoding the Phosphatase and Preparation of Stably Transfected Cells-PCR was used to amplify a cDNA fragment encoding the phosphatase (XYLP); the ACPL2 cDNA (GenBank TM BC035834) was the template, and the 5Ј-primer (5Ј-GAAGATCTGGACATGTTC-CCGAT-3Ј) and 3Ј-primer (5Ј-GAAGATCTGTGGACTTTC-CCTA-3Ј) each contained a BglII site. KOD-Plus DNA polymerase (TOYOBO) was used to perform PCR for 30 cycles, each at 94°C for 30 s, 53°C for 42 s, and 68°C for 180 s in 5% (v/v) dimethyl sulfoxide. The PCR fragments were subcloned into the BamHI site of the pCMV expression vector (Invitrogen). The sequence of each plasmid construct was confirmed by DNA sequencing. FuGENE 6 (Roche Diagnostics) was used according to the manufacturer's instructions to transfect the expression plasmid (6.0 g) into HeLa cells that were grown on 100-mm plates. Transfectants were cultured in the presence of 300 g/ml G418. The surviving colonies were then picked and propagated for experiments.
MISSION shRNA (Sigma) was used to inhibit XYLP expression in HeLa cells. Specifically, two constructs, each encoding hairpin RNA and isolated by the RNAi Consortium (clone number TRCN0000052898 or TRCN0000052899), were used. Each shRNA plasmid (6.0 g) was transfected into HeLa cells on 100-mm plates by using FuGENE 6 according to the manufacturer's instructions. Transfectants were cultured in the presence of 0.4 g/ml puromycin. Surviving colonies were then picked and propagated for experiments.
Isolation and Purification of GAGs-Cells were homogenized in acetone and thoroughly air-dried. The dried materials were digested at 55°C for 24 h with heat-pretreated (60°C for 30 min) actinase E (10% by weight of dried materials) in 250 l of 0.1 M borate sodium, pH 8.0, containing 10 mM calcium acetate. Following this digestion, each sample was adjusted to 5% trichloroacetic acid, and the resultant precipitate was removed by centrifugation. The soluble fraction was then mixed with ether. The aqueous phase was neutralized with 1.0 M sodium carbonate and adjusted to contain 80% ethanol. The resultant precipitate was dissolved in 50 mM pyridine acetate and subjected to gel filtration through a PD-10 column (Amersham Bioscience) with 50 mM pyridine acetate as an eluent. The flowthrough fractions were collected and dried. Each dried sample was subsequently dissolved in water.
Analysis of Disaccharide Composition of CS and HS-Purified GAGs were digested with 5 mIU of chondroitinase ABC or a mixture of 0.5 mIU of heparinase and 0.5 mIU of heparitinase at 37°C for 3 h. Reactions were terminated by boiling each mixture for 1 min. The digests were derivatized with the fluorophore 2-aminobenzamide, and high performance liquid chromatography (HPLC) was then used to analyze the samples as reported previously (22).
Gel Filtration Chromatography of GAGs-To measure GAG chain lengths, the purified GAGs were subjected to reductive ␤-elimination using NaBH 4 /NaOH and then analyzed by gel filtration chromatography on Superdex-200 columns (10 ϫ 300 mm) eluted with 0.2 M ammonium bicarbonate at a flow rate of 0.4 ml/min. Fractions were collected at 3-min intervals, lyophilized, and digested with chondroitinase ABC or a mixture of heparinase and heparitinase. The digests were derivatized with 2-aminobenzamide and then analyzed by HPLC on an aminebound PA-03 column as described previously (23,24).
Metabolic Labeling-shRNA control cells, shRNA XYLP cells, FAM20B-overexpressing cells, shRNA FAM20B cells (14), wild-type ESCs, or GlcAT-I Ϫ/Ϫ ESCs (21) were metabolically labeled with [ 32 P]NaH 2 PO 4 (285.2 mCi/mmol) in sodiumand phosphate-free DMEM (Invitrogen) containing 10% dialyzed fetal bovine serum at 37°C for 30 h. Each cell layer was treated with 1 mM LiOH at 4°C for 12 h to liberate the O-linked saccharides from core proteins (16,23). Each sample was subsequently neutralized and then applied to a column (1-ml bed volume) of AG 50W-X2 (H ϩ form, Bio-Rad). Flow-through fractions containing O-linked oligosaccharide components were pooled and neutralized with 1 mM NH 4 HCO 3 . 2AB was used to derivatize the oligosaccharide component of the linkage region as described previously (17). The 2AB-derivatized putative linkage regions were analyzed by gel filtration chromatography on a Superdex peptide column (10 ϫ 300 mm) eluted with 0.2 M ammonium bicarbonate at a flow rate of 0.4 ml/min. The pooled fractions were further analyzed by HPLC on an amine-bound PA-03 column in conjunction with enzymatic digestion as described previously (20,25).
Determination of Expression Levels of XYLP and FAM20B mRNA by Real Time PCR-MTC Multiple Tissue cDNA panels were purchased from Clontech. Each panel contained first strand cDNAs from a specific human tissue, and the cDNA has been normalized against the G3PDH transcript. The primer pairs for XYLP and real time-PCR conditions were described above. The primer pair for FAM20B was described previously (14).
Pulldown Assays-The cDNA fragment predicted to encode a truncated form of GlcAT-I that lacks the first 43 N-terminal amino acids of GlcAT-I was amplified with a 5Ј-primer (5Ј-GCTCTAGACTACGGCAGAAGGATCTGAGGA-3Ј) containing an in-frame XbaI site and a 3Ј-primer (5Ј-GGAAGAT-CTGTGCCTGAAAAGAGGTGGTAG-3Ј) containing a BglII site as described previously (27). The cDNA fragment predicted to encode a truncated form of GalT-II that lacks the first 34 N-terminal amino acids of GalT-II was amplified with a 5Ј-primer (5Ј-CGGAATTCGCCGAGCCCGGGGACCCCA-GG-3Ј) containing an in-frame EcoRI site and a 3Ј-primer (5Ј-CGGGATCCTCAGGGGATGCCCTCCCTTCT-3Ј) containing a BamHI site. This fragment was inserted into a pcDNA3Ins-His expression vector to encode a fusion protein with the insulin signal sequence and a His 6 sequence tag. The His-tagged and the protein A-tagged expression vectors were co-transfected into COS-1 cells grown on 100-mm plates. FuGENE TM 6 (Roche Diagnostics) was used according to the manufacturer's instructions. Two days after transfection, 1 ml of the culture medium was collected and incubated with 10 l of Ni-NTA-agarose (Qiagen) overnight at 4°C. The beads were recovered by centrifugation, washed with TBS buffer containing Tween 20 three times, and subjected to SDS-PAGE (7% gel); the separated proteins were transferred to a polyvinylidene difluoride (PVDF) membrane. The membrane was incubated in PBS containing 2% skim milk and 0.1% Tween 20, then incubated with IgG antibody, and then treated with anti-mouse IgG conjugated with horseradish peroxidase (Amersham Bioscience). An enhanced chemiluminescence (ECL) select detection reagent (Amersham Bioscience) was used to visualize antibodylabeled protein bands.
Statistical Analysis-Data are expressed as mean Ϯ S.D. The Student's t test was used to assess statistical significance. Differences were considered significant with p Ͻ 0.05.

RESULTS
Phosphorylated Linkage Structure Did Not Serve as a Substrate for Chondroitin Polymerization-Xyl phosphorylation increases during Gal addition onto the linkage disaccharide, Gal␤1-4Xyl, and is nearly stoichiometric at the trisaccharide level; however, GlcUA addition is accompanied by rapid Xyl dephosphorylation during decorin biosynthesis in rat skin fibroblasts (11). These results indicate that the phosphorylated linkage structure, GlcUA␤1-3Gal␤1-3Gal␤1-4Xyl(2-O-phosphate), might markedly affect the chondroitin polymerase-mediated production of CS chains. Previously, we demonstrated that chondroitin polymerization occurs when ChSy-1 and ChPF are co-expressed, and the acceptor substrate is ␣-TM bearing a truncated linkage region tetrasaccharide, GlcUA␤1-3Gal␤1-3Gal␤1-4Xyl (5). Therefore, we investigated whether polymerization could occur on the phosphorylated linkage structure, GlcUA-Gal-Gal-Xyl(2-O-phosphate). We used ␣-TM bearing a tetrasaccharide (GlcUA-Gal-Gal-Xyl) as a primer and recombinant FAM20B as an enzyme source to generate a phosphorylated linkage structure GlcUA-Gal-Gal-[ 32 P]-Xyl(2-O-phosphate) attached to ␣-TM. This phosphorylated structure GlcUA-Gal-Gal-[ 32 P]Xyl(2-O-phosphate)-TM was incubated with the ChSy-1 and ChPF proteins. Polymerization did not occur when GlcUA-Gal-Gal-[ 32 P]Xyl-(2-O-phosphate)-TM was the acceptor substrate ( Fig. 1). In addition, because we demonstrated that chondroitin polymerization is achieved by any two combinations of ChSy-1, ChSy-2, ChSy-3, and ChPF, we measured the polymerization activity using co-expression of ChSy-1/ChSy-2, ChSy-1/ChSy3, ChSy-2/ChSy-3, ChSy-2/ ChPF, or ChSy-3/ChPF as the enzyme source. Polymerization did not occur with any enzyme subunit combination when GlcUA-Gal-Gal-[ 32 P]Xyl-(2-O-phosphate)-TM was the acceptor substrate (data not shown). These results indicated that the phosphorylated linkage structure did not serve as a substrate for chondroitin polymerization. Therefore, based on these results and the finding that the phosphorylated linkage structure is not used by HS polymerases as a substrate (20), we reasoned that dephosphorylation of the Xyl residue may be required for biosynthetic maturation of the linkage region sequence and that this maturation may be a prerequisite for the initiation and the efficient elongation of the repeating disaccharide region of GAG chains.
Identification of the Putative Phosphatase as 2-Phosphoxylose Phosphatase-Based on these findings, we propose the existence of an enzyme responsible for the dephosphorylation of Xyl. We used the query term "unidentified phosphatase-like" to search the database at the National Center for Biotechnology Information (National Institutes of Health, Bethesda) and identified a candidate protein with a type II transmembrane protein topology. The predicted gene product, designated 2-phosphoxylose phosphatase (XYLP) here (originally acid phosphatase like-2; ACPL2; GenBank TM accession number AB827640), comprised 480 amino acids; this protein was predicted to have the type II transmembrane protein topology that is characteristic of all other enzymes involved in GAG biosynthesis ( Fig.  2A). Database searches indicated that the amino acid sequence displayed weak sequence similarity to human acidic phosphatase gene family members (8.3, 22.3, and 20.1% identical to ACP1, ACP2, and ACP3, respectively) ( Fig. 2A).
To examine whether XYLP could dephosphorylate 2-phosphoxylose within the GAG-protein linkage region, we generated a soluble form of XYLP by replacing the first 37 amino acids of XYLP with a cleavable insulin signal sequence and a protein A IgG binding domain. This soluble recombinant XYLP fusion protein was expressed in COS-1 cells. Based on Western blot analysis of culture medium, these transgenic COS-1 cells secreted a protein of ϳ90 kDa (Fig. 2B). The putative phosphatase fusion protein in the medium was purified with IgG-Sepharose beads to eliminate endogenous phosphatase; the protein-bound beads were then used as an enzyme source. We used six different substrates with variable linkage regions to assess 2-phosphoxylose phosphatase activity of the bound fusion protein. As shown in Table 1 Additionally, no phosphatase activity was detected using 32 Posteopontin or 32 P-labeled matrix extracellular phosphoglycoprotein as substrates; each of these glycoproteins is phosphorylated by the Golgi kinase FAM20C (28) homologous to FAM20B (a Xyl kinase) (14). In sharp contrast, recombinant alkaline phosphatase from bovine intestine utilized all substrates examined (Table 1).
Because XYLP showed weak sequence similarity to human acidic phosphatase gene family members, the effects of pH on the enzymatic activity of XYLP were examined; Gal-Gal-[ 32 P]Xyl(2-O-phosphate) was used as the substrate for these assays. XYLP exhibited optimum activity at pH 5.8 (Fig. 2C). These findings indicated the protein secreted by the transgenic COS-1 cells was a phosphatase that dephosphorylates Xyl residues in the GAG-protein linkage region.
Subcellular Localization of XYLP-To examine the intracellular localization of Xyl phosphatase, we generated a full-length form of XYLP that carried EGFP C termini (XYLP-EGFP). XYLP-EGFP was then co-expressed with a Golgi marker  MARCH 7, 2014 • VOLUME 289 • NUMBER 10

JOURNAL OF BIOLOGICAL CHEMISTRY 6699
(Golgi-DsRed) or an ER marker (ER-DsRed) in HeLa cells; we used confocal microscopy to examine the subcellular localization of the fluorescent proteins. XYLP-EGFP (Fig. 3A) co-localized with the Golgi-DsRed marker (Fig. 3C), but XYLP-EGFP (Fig. 3D) did not completely co-localize with the ER-DsRed marker (Fig. 3F). These results indicate that XYLP acted as a 2-phosphoxylose phosphatase in the Golgi apparatus. Involvement of XYLP in GAG Biosynthesis-To further examine the physiological relevance of XYLP, we investigated whether overexpression of XYLP changes the amount of GAGs  Table 2, the composition of the disaccharides and the amount of CS and HS isolated from clones stably transfected with transgenic XYLP were determined by HPLC; these clones were designated XYLP-1 cells. The disaccharide composition of CS and of HS in the XYLP-1 cells was similar to that in the control HeLa cells; however, the total amount of CS and of HS was greater in XYLP-1 cells than in control cells. This augmentation was small probably because endogenous levels of XYLP are high in HeLa cells.
Thus, we examined whether knockdown of XYLP decreased the amount of CS and HS in HeLa cells. The efficiency of XYLP gene silencing was determined by quantitative real time RT-PCR. Transfection of shRNA XYLP-1 or shRNA XYLP-2 (two different shRNAs directed against XYLP) resulted in 83 or 87% knockdown, respectively, of XYLP mRNA. shRNA XYLP-1 and shRNA XYLP-2 resulted in a 33 or 54% decrease, respectively, in the amount of CS. Similarly, shRNA XYLP-1 and shRNA XYLP-2 resulted in a 27 or 47% decrease, respectively, in the amount of HS. In each assay, shRNA XYLP-1 cells or shRNA XYLP-2 cells were compared with shRNA control HeLa cells. Taken together, these results indicated that XYLP regulates the total amount of GAG synthesized in cells and plays an important role in the biosynthesis of GAGs.
Involvement of XLYP in the Increased Number of CS and HS Chains-We next compared the length and number of CS and HS chains obtained from the shRNA XYLP-2 with those obtained from shRNA control HeLa cells. Gel filtration analysis using a Superdex 200 column revealed that the length of the CS and HS chains in shRNA XYLP-2 cells was similar to that in the shRNA control cells; however, the number of short and long HS chains in xylp-2 cells was smaller than that in shRNA control cells (Fig. 4, A and B). These results indicated that XYLP might regulate the number of CS and HS chains.
Real Time PCR Analysis of XYLP in Several Human Tissues-Real time PCR and first strand cDNAs from different human tissues were used to assess the tissue-specific expression pattern of XYLP and FAM20B. Both enzymes were expressed ubiquitously, but somewhat differentially, in all tissues examined. The strong XYLP expression was evident in the spleen and fetal liver, and moderate expression was evident in the placenta, pancreas, kidney, thymus, and colon. Notably, the expression pattern was similar to that of FAM20B in the heart, pancreas, prostate, ovary, and small intestine and different from that of FAM20B in the placenta, lung, liver, kidney, spleen, thymus, colon, and fetal liver (Fig. 6).
Interactions of XYLP and GlcAT-I-To evaluate interactions between XLYP and GlcAT-I, co-expression of XLYP and GlcAT-I was carried out. First, to confirm that the co-expression of two proteins augments the dephosphorylation activity, the culture medium from each transfection experiment was incubated with IgG-Sepharose, and the IgG-Sepharose-bound proteins were then evaluated for enzyme activity. Dephosphorylation activity was evident in medium from each transfectant when Gal-Gal-[ 32 P]Xyl(2-O-phosphate)-TM was used as a substrate (Table 4). Remarkably, in the presence of UDP-GlcUA, co-expression of XYLP and GlcAT-I augmented the dephosphorylation activity of XYLP over 16-fold (Table 4), although such an augmentation was not observed in the absence of UDP-GlcUA. These results suggest that the addition of the first GlcUA residue by GlcAT-I was accompanied by rapid dephosphorylation by XYLP. Then, to clarify the involvement of the transfer of GlcUA in a marked augmentation of the dephosphorylation activity, we constructed a GlcAT-I mutant, which is expected to lack GlcAT-I activity. Based on the crystal structure of GlcAT-I, the residue Glu-281 is in position to function as a catalytic base (29,30). In fact, the GlcAT-I E281A mutant  would not possess GlcAT-I activity. Notably, co-expression of XYLP and the GlcAT-I E281A mutant did not result in a marked augmentation of the dephosphorylation activity even in the presence of UDP-GlcUA.
Next, we used pulldown assays to determine whether interactions between XYLP and GlcAT-I occurred. For this analysis, soluble forms of protein A-tagged XYLP fusion proteins (XYLP-ProA) and a soluble form of a His 6 -tagged GlcAT-I, GalT-II, and Fam20B fusion protein (GlcAT-I-His, GalT-II-His, and Fam20B-His, respectively) were generated. To evaluate associations among these proteins, pulldown assays were performed. In addition, to ensure specificity, we conducted these assays with GalT-II or FAM20B (a Xyl kinase). For this analysis, soluble forms of a protein A-tagged XYLP fusion protein (XYLP-ProA) and His 6 -tagged GlcAT-I, GalT-II, and Fam20B fusion proteins (GlcAT-I-His, GalT-II-His, and Fam20B-His, respectively) were generated. Ni-NTA-agarose was added to the culture medium to pull down His-tagged proteins. Then, the proteins were subjected to SDS-PAGE followed by Western blotting, and the resulting blots were probed with IgG antibody; antibody signal was detected with an ECL select detection reagent. No band was detected with samples from the co-transfectant expressing XYLP-ProA and GalT-II-His or XYLP-ProA and Fam20B-His (Fig. 7A). However, proteins with a molecular mass of ϳ90 kDa and corresponding to XYLP-ProA were detected with samples from the co-transfectant expressing XYLP-ProA and GlcAT-I-His (Fig. 7A).
Then, to investigate the intracellular localization of XYLP in the presence or absence of GlcAT-I, XYLP-EGFP was co-ex- pressed with a Golgi-DsRed or an ER-DsRed in wild-type or GlcAT-I Ϫ/Ϫ ESCs (21), respectively. In the presence of GlcAT-I (wild-type ESCs), XYLP-EGFP co-localized with the Golgi-DsRed marker, but XYLP-EGFP did not completely co-localize with the ER-DsRed marker (Fig. 7, B and C). These results were consistent with those obtained in HeLa cells (see Fig. 3, A and C). Remarkably, in the absence of GlcAT-I (GlcAT-I Ϫ/Ϫ ESCs), XYLP-EGFP co-localized predominantly with the ER-DsRed marker, whereas XYLP-EGFP did not completely co-localize with the Golgi-DsRed marker (Fig. 7, B and C). These results indicate that XYLP is transported to the Golgi with interaction of GlcAT-I and that expression of GlcAT-I is indispensable for accumulation of XYLP in the Golgi complex.
To further examine whether XYLP-dependent dephosphorylation is influenced by knock-out of GlcAT-I, the linkage oligosaccharides obtained from the GlcAT-I Ϫ/Ϫ ESCs were compared with those obtained from wild-type ESCs. The 2-phosphoxylose in each cell line was radiolabeled with [ 32 P]NaH 2 PO 4 and isolated after mild alkaline treatment with LiOH as described above. The isolated oligosaccharides were derivatized with the fluorophore 2AB and separated by gel filtration chromatography using a Superdex peptide column.  (Table 5), despite the fact that XYLP was expressed in both wild-type and GlcAT-I Ϫ/Ϫ ESCs. These results also suggest that GlcAT-I-mediated addition of GlcUA might facilitate the XYLP-dependent dephosphorylation. Taken together, these results indicated that XYLP and GlcAT-I interacted with each other and that GlcAT-I-mediated addition of GlcUA can be accompanied by rapid XYLP-dependent dephosphorylation during completion of linkage region formation.

DISCUSSION
Previously, we demonstrated that FAM20B is a Xyl kinase that phosphorylates C2 of the Xyl residue in the GAG-protein linkage region of nonphosphorylated trisaccharide serine structures (Gal␤1-3Gal␤1-4Xyl␤1-O-Ser) (14). Additionally, Moses et al. (11,12) demonstrated that phosphorylation gradually increases as the linkage region forms and that Xyl phosphorylation was completed at the Gal␤1-3Gal␤1-4Xyl stage. Only the trisaccharide contained an almost fully phosphorylated Xyl. Addition of the first GlcUA residue was accompanied by rapid dephosphorylation, suggesting that the phosphoryla-   tion of Xyl is a transient phenomenon (11,12). Thus, it is suggested that GlcAT-I-mediated addition of GlcUA can be accompanied by rapid XYLP-dependent dephosphorylation during completion of linkage region formation. In fact, the phosphorylated linkage tetrasaccharide was not detected in cells (Tables 3 and 5). In addition, XYLP and GlcAT-I can form heterooligomers apparently required for maximal dephosphorylation activity (Table 4) and translocation to the Golgi in vitro (Fig. 7, B and C). These findings are similar to those in the case of co-transfection of EXT1 and EXT2, which forms an enzyme complex for the polymerization of HS (31). In view of these results, both XYLP and GlcAT-I seem to be indispensable for efficient maturation of the linkage region tetrasaccharide. Thus, we concluded that the phosphorylation and dephosphorylation of Xyl residues were tightly regulated by FAM20B, XYLP, and GlcAT-I.  Interactions of XYLP and GlcAT-I. A, culture medium from cells co-expressing XYLP-ProA and GlcAT-I-His, XYLP-ProA and GalT-II-His, or XYLP-ProA and FAM20B-His was incubated with Ni-NTA-agarose to purify His 6 -tagged proteins and any associated proteins; the purified proteins were subjected to SDS-PAGE. The separated proteins were transferred to PVDF membrane and allowed to react with IgG as a primary antibody; ECL select detection reagent was used to visualize antibody-labeled proteins. B and C, XYLP-EGFP was co-expressed with a Golgi-DsRed (B) or an ER-DsRed (C) in wild-type or GlcAT-I Ϫ/Ϫ ESCs, respectively. B, XYLP-EGFP co-localized with the Golgi-DsRed marker in wild-type ESCs, but XYLP-EGFP did not completely co-localize with the Golgi-DsRed marker in GlcAT-I Ϫ/Ϫ ESCs. C, XYLP-EGFP did not completely co-localize with the ER-DsRed marker in wild-type ESCs, but XYLP-EGFP co-localized with the ER-DsRed marker in GlcAT-I Ϫ/Ϫ ESCs. Seph, Sepharose; WB, Western blot.

GlcAT-I ؊/؊ ESCs
Gal-͓ 32 P͔Xyl(2P)-2AB a 1.0 Ϯ 0. Here, we found that RNA interference of XYLP resulted in a reduction in the amounts of CS and HS in cells. In addition, gel filtration analysis of the GAG chains synthesized in the knockdown cells revealed that the CS and HS chains had similar lengths to CS and HS chains, respectively, in shRNA control cells. These results indicated that XYLP regulated the number of CS and HS chains. These findings are in agreement with the results that the phosphorylated linkage tetrasaccharide did not serve as a substrate for chondroitin or HS polymerization. Therefore, we reasoned that dephosphorylation of the Xyl residue by XYLP may be required for biosynthetic maturation of the linkage region sequence and that this maturation may be a prerequisite for the initiation and the efficient elongation of the repeating disaccharide region of GAG chains. Similarly, overexpression of FAM20B increased the amount of both CS and HS in HeLa cells, and RNA interference of FAM20B resulted in a reduction in the amounts of both CS and HS in the cells (14). Gel filtration analysis of the GAG chains synthesized in cells that overexpressed FAM20B revealed that they had lengths similar to those in the mock-transfected cells (14). These results also indicate that FAM20B regulates the number of GAG chains by phosphorylating the Xyl residue in the GAG-protein linkage region of proteoglycans. In this regard, it should be noted that GlcAT-I could efficiently transfer a GlcUA residue to the phosphorylated trisaccharide serine, Gal␤1-3Gal␤1-4Xyl(2-O-phosphate)␤1-O-Ser, than to the nonphosphorylated counterpart Gal␤1-3Gal␤1-4Xyl␤1-O-Ser (10). Moreover, in rat articular cartilage explants, the introduction of GlcAT-I enhanced GAG synthesis via an increase in the abundance rather than the length of GAG chains, whereas antisense

2P
Core protein FIGURE 8. Phosphorylation and dephosphorylation of Xyl residues regulate the formation of the linkage region and GAG biosynthesis. Synthesis of the linkage region is initiated by the addition of a Xyl residue to a specific serine residue on a core protein; this event is followed by the sequential transfer of two Gal residues, and synthesis of the linkage region is completed by transfer of a GlcUA residue. During synthesis of the linkage region, transient FAM20Bcatalyzed phosphorylation of the Xyl residue occurs and enhances the activity of GalT-II and of GlcAT-I. After synthesis of the phosphorylated linkage region trisaccharide, GlcAT-I transfers GlcUA to the phosphorylated trisaccharide serine structure, Gal␤1-3Gal␤1-4Xyl(2-O-phosphate)␤1-O-Ser. Concomitant with this process, dephosphorylation of Xyl is immediately induced by XLYP. Then, chondroitin (Chn) or HS polymerases induce polymerization of disaccharide chains onto the linkage region tetrasaccharide. If formation of linkage region is excessively accelerated by FAM20B and/or dephosphorylation of the Xyl by XYLP is attenuated, biosynthetic intermediates (phosphorylated linkage tetrasaccharides) could accumulate; under this condition, EXTL2 may immediately transfer a GlcNAc to the phosphorylated linkage tetrasaccharide and thereby induce chain termination (20). 2P represents 2-O-phosphate.