Phosphorylation and Sulfation of Oligosaccharide Substrates Critically Influence the Activity of Human β1,4-Galactosyltransferase 7 (GalT-I) and β1,3-Glucuronosyltransferase I (GlcAT-I) Involved in the Biosynthesis of the Glycosaminoglycan-Protein Linkage Region of Proteoglycans*

We determined whether the two major structural modifications, i.e. phosphorylation and sulfation of the glycosaminoglycan-protein linkage region (GlcAβ1–3Galβ1–3Galβ1–4Xylβ1), govern the specificity of the glycosyltransferases responsible for the biosynthesis of the tetrasaccharide primer. We analyzed the influence of C-2 phosphorylation of Xyl residue on human β1,4-galactosyltransferase 7 (GalT-I), which catalyzes the transfer of Gal onto Xyl, and we evaluated the consequences of C-4/C-6 sulfation of Galβ1–3Gal (Gal2-Gal1) on the activity and specificity of β1,3-glucuronosyltransferase I (GlcAT-I) responsible for the completion of the glycosaminoglycan primer sequence. For this purpose, a series of phosphorylated xylosides and sulfated C-4 and C-6 analogs of Galβ1–3Gal was synthesized and tested as potential substrates for the recombinant enzymes. Our results revealed that the phosphorylation of Xyl on the C-2 position prevents GalT-I activity, suggesting that this modification may occur once Gal is attached to the Xyl residue of the nascent oligosaccharide linkage. On the other hand, we showed that sulfation on C-6 position of Gal1 of the Galβ1–3Gal analog markedly enhanced GlcAT-I catalytic efficiency and we demonstrated the importance of Trp243 and Lys317 residues of Gal1 binding site for enzyme activity. In contrast, we found that GlcAT-I was unable to use digalactosides as acceptor substrates when Gal1 was sulfated on C-4 position or when Gal2 was sulfated on both C-4 and C-6 positions. Altogether, we demonstrated that oligosaccharide modifications of the linkage region control the specificity of the glycosyltransferases, a process that may regulate maturation and processing of glycosaminoglycan chains.

Proteoglycans (PGs), 1 which are located at cell surfaces and in extracellular matrix play vital functions in many biological processes such as cell proliferation, cell adhesion, blood coagulation, and wound repair (1). In most cases, biological activities of PGs are governed by interactions of their glycosaminoglycan (GAG) chains with growth factors, cytokines, morphogens, and a variety of protein ligands. Accordingly, the disruption of the glycosyltransferases involved in either initiation or elongation of GAG chains has severe biological consequences. In human, mutations of the gene encoding ␤1,4-galactosyltransferase 7 (referred to as GalT-I in GAG synthesis pathway) cause a progeroid variant of Ehlers-Danlos syndrome characterized by aged appearance, developmental delay, dwarfism, and various connective tissue abnormalities (2). Likewise, it has been shown that the mutations of EXT1 and EXT2 genes coding for glycosyltransferases responsible for heparan sulfate (HS) chain polymerization lead to the hereditary multiple exostosis disorders characterized by tumors of bony outgrowth (3).
The major GAGs found on PGs are HS/heparin and chondroitin sulfate (CS)/dermatan sulfate, which are linear polysaccharides consisting of a repetition of [GlcNAc␣1-4GlcA␤1 -4] and [GalNAc␤1-4GlcA␤1-3] disaccharide units, respectively. The assembly of GAG chains is initiated by the synthesis of a common GAG-protein linkage structure, GlcA␤1-3Gal␤1-3Gal␤1-4Xyl-O-, which is attached to specific serine residues of different core proteins. This linkage tetrasaccharide is formed by the stepwise addition of each sugar residue from the corresponding UDP-sugar catalyzed by O-xylosyltransferase I (4), GalT-I (5), ␤1,3-galactosyltransferase 6 (GalT-II) (6), and ␤1,3-glucuronosyltransferase I (GlcAT-I) (7) (see Fig. 1A). The transfer of either a ␣GlcNAc or ␤GalNAc residue on the terminal GlcA of the linkage region initiates the polymerization of the HS or CS chains, respectively. HS and CS are further modified by the cooperative action of epimerases and sulfo-transferases that adds considerable complexity and functionality to the polysaccharide GAG chains.
Recently, structural studies have shown the presence of various modifications of the GAG-protein linkage region (for review see Ref 8). Sulfate substitutions are primarily found at the C-4 and/or C-6 positions of the outer Gal (referred to as Gal2) and of the inner Gal (referred to as Gal1) of the linkage tetrasaccharide, whereas the presence of a phosphate has been exclusively demonstrated on the C-2 position of Xyl (see Fig.  1A). C-6 sulfation of Gal2 has been identified as the predominant modification in human aggrecan CS chains (9), and C-4 sulfation of the same Gal residue has been found in urinary trypsin inhibitor (10) and inter-␣-trypsin inhibitor (11). Sulfation at the 4-and 6-positions of Gal1 and/or Gal2 is also present in CS from aggrecan of shark cartilage (12), bovine nasal septum (13), articular bovine cartilage (14), and from mouse syndecan-1 (15). So far, sulfated Gal residues of the linkage region have been demonstrated for CS and dermatan sulfate but not in HS or heparin, whereas a C-2 phosphorylated Xyl residue has been found in both HS/heparin and CS/dermatan sulfate chains. Although the role of these substitutions is not fully understood, it has been suggested that phosphorylation and sulfation may regulate maturation and processing of growing GAG chains (16,17). A prerequisite step in the evaluation of this hypothesis is the determination of the substrate specificity of the glycosyltransferases responsible for the biosynthesis of the linkage region with regard to phosphorylation and sulfation. Our group has been deeply involved in the structurefunction studies of GlcAT-I, which catalyzes the transfer of GlcA onto Gal2 of Gal␤1-3Gal, thus completing the final step of GAG-protein linkage synthesis (18 -20). This enzyme has attracted much attention in our laboratory and others (21), because it appears to play a rate-limiting role in GAG biosynthesis (22,23). GalT-I, which is responsible for the transfer of Gal onto the Xyl residue of the glycopeptide primer of PGs, has recently been cloned but has not been deeply characterized (5).
In this study, we determined whether the phosphorylation of Xyl and sulfation of Gal1 and/or Gal2 affect GalT-I and GlcAT-I activity, respectively. For this purpose, we designed and synthesized C-2-phosphorylated xylosides as well as C-4 and/or C-6-sulfated digalactose analogs of the GAG-protein linkage region. These compounds have been tested as potential sub- A, structure of the tetrasaccharide sequence attached to specific serine of consensus serine-glycine residues of the core protein. Potential sulfation and phosphorylation sites are marked by arrows. The glycosyltransferases involved are indicated, and the enzymes used in this study are underlined. B, chemical structure of the unphosphorylated and phosphorylated compounds tested as substrates of GalT-I (numbers 1 and 2) and of the corresponding reaction products (numbers 3 and 4). C, chemical structure of the unsulfated and sulfated compounds used as substrates of GlcAT-I (numbers 5-10) and of the corresponding reaction products (numbers 11-16).
strates of the two recombinant human enzymes expressed in the yeast Pichia pastoris. We showed that GalT-I efficiently catalyzed the transfer of Gal onto the nonphosphorylated xyloside, whereas the phosphorylated analog was not the substrate, suggesting that C-2 phosphorylation precludes the transfer of the first Gal on Xyl of the glycopeptide primer. By contrast, we demonstrated that the Gal␤1-3Gal(6-sulfate) derivative serves as a much better acceptor substrate for the recombinant human GlcAT-I than its unsulfated counterpart, raising the possibility that sulfation at the C-6 position of Gal1 may contribute to the efficient completion of the linkage region tetrasaccharide. Our data show that phosphorylation and sulfation critically influence the specificity of the glycosyltransferases involved in the formation of the GAG-protein linkage region.

EXPERIMENTAL PROCEDURES
Materials-UDP-Gal, UDP-GlcA, and anti-rabbit alkaline phosphatase-conjugated immunoglobulins were from Sigma. The products used for chemical synthesis were provided from Aldrich. UDP[ 14 C]-GlcA and UDP[ 14 C]-Gal were purchased from Amersham Biosciences. Trifluoroacetic acid and dimethyl sulfoxide were obtained from Merck (Darmstadt, Germany), and HPLC grade acetonitrile was from Carlo Erba (Val de Reuil, France). Bacterial and yeast culture media were provided by Difco. Protein assay reagent was obtained from Bio-Rad (Hercules, CA). Restriction enzymes and T4 DNA ligase were provided by New England Biolabs (Hitchin, United Kingdom). The P. pastoris expression system and competent Escherichia coli cells were purchased from Invitrogen. The QuikChange site-directed mutagenesis kit was from Stratagene (La Jolla, CA).
Chemical Synthesis-The 7-methoxy-2-naphthyl (MN)-derived xylosides in which the methoxy group served as a marker for NMR characterization and UV detection were prepared as described below. The structure of the following compounds is presented in  (24) in toluene at 0°C under the catalysis of trimethylsilyl triflate gave the ␤-linked disaccharide derivative at a 73% yield. Mild hydrolysis of the isopropylidene acetal with aqueous acetic acid afforded the corresponding diol at a 90% yield. This later compound was trans-esterified to give Gal␤1-4Xyl␤1-O-MN (m.p. 238 -240°C) at a 94% yield. Tin-mediated regioselective acetylation at O-3 of the Xyl residue of the disaccharide diol, as previously reported (25) for the synthesis of phosphorylated glycopeptides followed by phosphorylation at O-2 as described above, gave the corresponding fully protected disaccharide derivative at a 60% overall yield. Subsequent hydrogenolysis followed by hydrazinolysis of the esters and salification with NaOH gave Gal␤1-4Xyl(2-phosphate)␤1-O-MN as its sodium salt, [␣] D Ϫ41°( c1, water), at an 81% overall yield. All of the synthetic compounds had analytical data (NMR, mass, and elemental analyses) fully consistent with the expected structures.
The following methoxyphenyl (MP)-digalactosides, analogs of the GAG-protein linkage region, were prepared as previously described (26). The structure of these following compounds is presented in were synthesized as previously reported (27).
cDNA Cloning and Plasmid Construction-The human GalT-I sequence was cloned by PCR from a placenta cDNA library (Clontech, Palo Alto, CA) using a sense primer (5Ј-ATGTTCCCCTCGCGGAG-GAAAGCGGCGCAGC-3Ј) together with an antisense primer (5Ј-TCAGCTGAATGTGCACCAGGGTGTGGCGGTCTTG-3Ј) corresponding to the 5Ј end and the 3Ј end of the coding region of GalT-I as described by Almeida et al. (5). The fragment amplified using Advantaq polymerase (Clontech) was subcloned into PCR2.1 (TA-cloning kit (Invitrogen)) and sequenced on both strands. The cDNA sequence was 100% identical to that previously described (5). For the heterologous expression of the human GalT-I in P. pastoris, the full-length cDNA sequence was modified by PCR to include a SacII site and a Kozak consensus sequence at the 5Ј end and a XbaI site at the 3Ј end using appropriate oligonucleotides. The modified cDNA was then subcloned into the SacII-XbaI sites of the yeast expression vector pPICZB to produce pPICZ-GalT-I. Cloning of the full-length GlcAT-I cDNA and construction of the recombinant pPICZ-GlcAT-I were performed as previously described (19). Construction of amino acid substituted mutants of GlcAT-I was carried out using pPICZ-GlcAT-I as template with the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's recommendations. Mutagenic primers are presented in Table I. Mutants were systematically checked by sequencing, and the various mutants were individually expressed in P. pastoris as described below. For the expression of the soluble catalytic domain of GlcAT-I (21), the sequence encoding Thr 76 -Val 335 was amplified and subcloned into the SacII-XbaI sites of pPICZB. The active protein was expressed intracellularly in P. pastoris and purified from the cytosolic fraction as described below.

W243A
Sense Heterologous Expression in the Yeast P. pastoris and Purification-Each recombinant pPICZ vector was individually transformed into the P. pastoris SMD1168 yeast strain using the P. pastoris Easy Comp Transformation kit (Invitrogen). Transformants were selected on YPD plates (1% (w/v) yeast extract, 2% (w/v) peptone, 2% (w/v) dextrose) containing 100 g/ml Zeocin (Invitrogen). The cells were grown in BMGY medium (1% (w/v) yeast extract, 2% (w/v) peptone, 100 mM potassium phosphate, pH 6.0, 1.34% (w/v) yeast nitrogen base, and 1% (v/v) glycerol). Expression was induced by methanol (2%, v/v) for 48 h at 30°C in a rotary shaker (180 rpm). Yeast cells were broken with glass beads and further submitted to differential centrifugation as previously described (19). The 100,000 ϫ g pellet corresponding to the membrane fraction was resuspended by Dounce homogenization in sucrose-HEPES buffer (0.25 M sucrose, 5 mM HEPES, pH 7.4). Protein concentration was evaluated by the method of Bradford (28), and membrane fractions were analyzed by SDS-PAGE and immunoblotting and used for kinetic analyses.
For purification of GlcAT-I, the supernatant from the 100,000 ϫ g centrifugation containing the soluble GlcAT-I form (Thr 76 -Val 335 ) was dialyzed against buffer A (20 mM sodium phosphate, pH 7.0) at 4°C overnight and applied to a DEAE-Sephacel column (6 ϫ 1.5 cm, Sigma), which had been equilibrated with the same buffer at a flow rate of 0.5 ml/min. The flow-through was applied onto a cation exchange support (Macro-Prep High-S, 11 ϫ 21 cm, Bio-Rad) equilibrated with buffer A at a flow rate of 0.5 ml/min. After washing with buffer A and then with the same buffer containing 100 mM NaCl (5-column volumes), the protein was eluted with a linear gradient of 0.02-0.2 M sodium phosphate buffer, pH 7.0, and analyzed by SDS-PAGE, Coomassie Blue staining, and immunoblotting. The protein was purified to apparent homogeneity, and 0.05-0.1 g protein/assay was used for kinetic analyses.
SDS-PAGE was performed under reducing conditions according to Laemmli (29). Immunoblot analysis was performed using a polyclonal anti-peptide antibody and alkaline phosphatase-conjugated anti-rabbit immunoglobulins as secondary antibodies as previously described (19). The amount of recombinant wild type and mutant GlcAT-I protein expressed in membrane fractions of yeast cells was evaluated from Western blot analysis using a calibration curve established with 0.01-0.10 g of purified protein run on the same gel.
Glycosyltransferase Activity Assays and Kinetics-For GalT-I assays, 50 -100 g of membrane proteins were incubated in 100 mM acetate buffer, pH 6.5, containing 5 mM MnCl 2 , 1 mM UDP-Gal, 1 mM xyloside (compounds 1 and 2) at 37°C for 60 min in a total volume of 100 l. After incubation, the reaction was stopped by the addition of 10 l of 6 M HCl. The samples were centrifuged for 10 min at 100,000 ϫ g to precipitate proteins, and an aliquot of the supernatant was analyzed by reverse-phase HPLC using a C 18 column (4.6 ϫ 150 mm, 5 m, Alltech Associates, Deerfield, IL) at a detection wavelength of 285 nm. The mobile phase was composed of 20% (v/v) acetonitrile and 0.05% (v/v) trifluoroacetic acid in water for the nonphosphorylated acceptor substrate (Xyl␤1-O-MN) and 17% (v/v) acetonitrile and 0.025% (v/v) trifluoroacetic acid in water for the phosphorylated xyloside (Xyl(2-phosphate)␤1-O-MN) at flow rate of 1 ml/min. Activity of recombinant human GlcAT-I was evaluated using Gal␤1-3Gal␤1-O-MP (compound 5) and each of the monosulfated (compounds 6 -9) or disulfated digalactosides (compound 10) as acceptor substrate. Incubations were performed in 100 mM acetate buffer with 50 mM MnCl 2 , 50 -100 g of membrane proteins (or 0.05-0.1 g of purified enzyme), 1-25 mM substrate, and 1 mM UDP-GlcA in a total volume of 100 l at 37°C for 60 min in a pH range from 5.0 to 7.5. The mixtures were applied to a Sep-Pak C 18 cartridge (Waters, Milford, MA) before chromatographic analysis to remove UDP-GlcA. After the cartridge was washed with 2 ml of water, the reaction product and the acceptor substrate were eluted by 1 ml of acetonitrile and further analyzed by gel filtration HPLC using a Superdex Peptide HR10/30 column (10 ϫ 300 mm, Amersham Biosciences) with 0.02 M dipotassium phosphate buffer, pH 8.0, containing 0.2 M NaCl as eluent at a flow rate of 0.3 ml/min. The reaction product was detected at a wavelength of 285 nm. We verified that no loss of reaction product occurred upon Sep-Pak extraction.
For each set of experiments, control assays in which either the donor or the acceptor substrate was omitted were systematically run under the same conditions. The addition of a GlcA or Gal residue on the nonreducing end of acceptor substrates was verified by comparison of retention time with that of the corresponding chemically synthesized standards (see Fig. 1, B and C) and by the associated radioactivity monitored by a LB 507A Berthold radioactivity detector when incubations were carried out in the presence of UDP[ 14 C]-GlcA or UDP[ 14 C]-Gal (0.1 Ci) as donor substrate. Quantitation of the reaction products was performed with calibration curves drawn with increasing concen-trations of the standards resolved under similar conditions. A time course of the glucuronosyltransferase and galactosyltransferase activities in membrane fractions showed that the product formation was linear with the time over 120 min. Initial rate data were subsequently taken after a 60-min incubation time.
Apparent kinetic parameters (K m and V max ) of GalT-I were determined using linear least-squares regression analysis of double-reciprocal plots of initial velocity versus UDP-Gal (0 -2 mM) at a constant concentration of Xyl␤1-O-MN (10 mM) and of initial velocity versus acceptor substrate (0 -25 mM) at a constant concentration of UDP-Gal (1 mM). Kinetic parameters (K m and V max ) of GlcAT-I were determined using linear least-squares regression analysis of double-reciprocal plots of (1 mM) at an optimum of pH 6.5. Kinetic parameters of GlcAT-I toward Gal␤1-3Gal(6-sulfate)␤1-O-MP were determined in a 0 -2.5 mM range of substrate at a constant concentration of UDP-GlcA (1 mM) and an optimum of pH 5.5.
Molecular Modeling-The model structure of the GlcAT-I complex with the unsulfated and sulfated digalactosides was built by energy minimization and molecular dynamics calculations based on the initial structure of GlcAT-I (Protein Data Bank code 1FGG) (21). The conformation with the lowest estimated free energy of binding for each substrate was minimized using AMBER 7.0 program (30). To avoid the missing residue regions in the Protein Data Bank file, all of the minimizations were done in a sphere of a 12-Å radius from the sulfated molecule and all of the residues outside this sphere were kept fixed. Charges of Gal␤1-3Gal(6-sulfate) were calculated using GAUSSIAN94 package (31) and the HF/6 -31 G* basis set. Atom-centered charges were fitted with Antechamber of AMBER 7 software package (30). An automated flexible docking of sulfated molecules into the active site of GlcAT-I was done using AutoDock 3.0 program using the genetic algorithm feature (32).

Xyl(2-phosphate)␤1-O-MN Is Not a Substrate for GalT-I-
The activity of the recombinant human GalT-I expressed in P. pastoris was tested toward the unphosphorylated xyloside (Xyl␤1-O-MN) by reverse-phase HPLC. A chromatogram of the HPLC resolution of the substrate and the reaction product (Gal␤1-4Xyl␤1-O-MN) formed in the presence of UDP-Gal as co-substrate is illustrated in Fig. 2A. The reaction product was identified by the following: 1) the comparison of retention time with that of the chemically synthesized standard; 2) the associated radioactivity when UDP[ 14 C]Gal was used as donor substrate; and 3) the absence of detectable peak when substrate or co-substrate was omitted from the assay. As expected, the results showed that the recombinant GalT-I was active toward Xyl␤1-O-MN. Determination of the kinetic parameters of the recombinant enzyme led to apparent V max and K m values toward the acceptor substrate (Xyl␤1-O-MN) of 47.2 pmol⅐min Ϫ1 ⅐mg protein Ϫ1 and 2.0 mM, respectively. Apparent V max and K m constants toward the donor substrate (UDP-Gal) were 46.3 pmol⅐min Ϫ1 ⅐mg protein Ϫ1 and 0.14 mM, respectively.
The phosphorylated derivative (Xyl(2-phosphate)␤1-O-MN was next tested as acceptor substrate for GalT-I. Chromatographic analysis of the chemically synthesized standards (i.e. substrate and reaction product) showed that Xyl(2-phosphate)␤1-O-MN) and Gal␤1-4Xyl(2-phosphate)␤1-O-MN were resolved under our analytical conditions (Fig. 2B). The enzymatic assay indicated that, in contrast to the unphosphorylated xyloside, no activity could be observed when using the phosphorylated analog as a potential substrate for the recombinant enzyme. The same result was obtained in various conditions of assay: pH 5.0 -7.5; substrate concentration range of 0 -50 mM; and Mn 2ϩ concentration of 0 -100 mM. This clearly indicated that GalT-I was unable to catalyze the transfer of Gal from UDP-Gal onto Xyl when this residue was phosphorylated at the C-2 position.
Gal␤1-3Gal(6-sulfate)␤1-O-MP Is a Preferred Substrate for the Human GlcAT-I-GlcAT-I catalyzed the transfer of GlcA provided by UDP-GlcA onto Gal2 of the GAG-protein linker region and was previously shown to recognize Gal␤1-3Gal (Gal2-Gal1) disaccharide as a minimum substrate (21). We analyzed the specificity of the human recombinant enzyme expressed in the yeast P. pastoris toward various sulfated analogs of Gal␤1-3Gal␤1-O-MP (Fig. 1B). A gel filtration HPLC assay was developed for the separation of the substrates and reaction products (Fig. 3). As expected, the recombinant GlcAT-I exhibited high activity toward the unsulfated Gal␤1-3Gal␤1-O-MP substrate as shown in Table II. Interestingly, GlcAT-I efficiently catalyzed the transfer of GlcA onto the monosulfated derivative Gal␤1-3Gal(6-sulfate)␤1-O-MP (see HPLC chromatogram in Fig. 3 and Table II). By contrast, the recombinant enzyme exhibited no significant activity toward any of the other monosulfated or disulfated substrates tested, i.e. Gal(4-sulfate)␤1-3Gal␤1-O-MP, Gal(6-sulfate)␤1-3Gal␤1-O-MP, Gal␤1-3Gal(4-sulfate)␤1-O-MP, or the di-C-6-sulfated derivative. These data clearly indicated that the active site of the enzyme could accommodate a sulfate group only at the C-6 position of Gal1. The V max values of the membrane-bound GlcAT-I for Gal␤1-3Gal␤1-O-MP and its C-6 sulfated analog were similar (345 and 303 nmol⅐min Ϫ1 ⅐mg GlcAT-I protein Ϫ1 , respectively). By contrast, the K m of the membrane-bound GlcAT-I for Gal␤1-3Gal(6-sulfate)␤1-O-MP was 6-fold lower than that of its unsulfated derivative (Table II). Kinetic analyses performed on purified GlcAT-I confirmed that the enzyme exhibited a much higher affinity toward the sulfated derivative than toward the unsulfated compound (ϳ10-fold). Thus, our results showed that GlcAT-I exhibited a markedly higher efficiency toward the sulfated substrate compared with its unsubstituted counterpart due to lower the K m value, suggesting that the 6-sulfated digalactoside was a better substrate than the unsulfated compound.
Identification of Active Site Residues Involved in Recognition of Sulfated and Unsulfated Derivatives-In an attempt to further delineate the structural determinants governing GlcAT-I specificity, we built a computer-aided model of the substrateenzyme complex based on the x-ray structure of the catalytic domain of GlcAT-I bound to UDP (21). Fig. 4 shows the superimposition of unsulfated and C-6-sulfated substrate in the acceptor binding site, pointing out residues in the vicinity of Gal1. This figure shows the dominant stacking interactions between the side chain of Trp 243 and Gal1 ring of the digalactoside acceptor substrate. Docking Gal␤1-3Gal(6-sulfate) with GlcAT-I structure revealed that this compound adopted a position similar to that of the unsulfated substrate with a possible stacking interaction with Trp 243 . The structural analysis of GlcAT-I had previously provided evidence for a dimeric organization of this protein with residues Glu 312 -Gln 318 of one monomer extending into the active site of the other (21). In the related ␤1,3-glucuronosyltransferase GlcAT-P, the residues Val 320 and Asn 321 , which correspond to Lys 317 and Gln 318 in GlcAT-I, had also been suggested to contribute to acceptor substrate specificity (33). In GlcAT-I, Gln 318 of the second monomer was found in a position to form a hydrogen bond with the O-6 of Gal1. Molecular modeling suggested that this residue could also interact with the sulfate substituent of Gal␤1-3Gal(6-sulfate) (Fig. 4). In addition, computational modeling pointed out the presence of a conserved diglycine motif Gly 222 -Gly 223 in the vicinity of O-4 and O-6 of Gal1 moiety but not at hydrogen bond distance. These residues formed a cavity that could apparently accommodate the sulfated molecule. Based on this information, we constructed a series of mutants to determine the importance of the different residues in the vicinity of Gal1 ring. Both conservative and nonconservative mutations were carried out, and the consequences of these mutations were evaluated on GlcAT-I activity.
Upon expression in P. pastoris, immunoblot analysis showed that each mutant was expressed at a similar level or even higher than the wild-type protein (Fig. 5A). Specific activities of wild-type GlcAT-I and mutants were normalized relative to the amount of expressed protein and were evaluated using Gal␤1-3Gal␤1-O-MP and its analog sulfated on C-6 of the Gal1 unit as substrate (Fig. 5B). The replacement of Trp 243 by alanine led to a complete loss of GlcAT-I activity toward both the unsulfated and sulfated digalactoside, indicating that this residue was critical for glucuronosyltransferase activity, in agreement with molecular modeling prediction. Substitution of this amino acid by a conservative phenylalanine residue slightly restored the enzyme activity toward the unsulfated and C-6 sulfated analog to ϳ8 and 3% of the initial activity, respectively (Fig. 5B). The importance of the conserved Gly 222 -Gly 223 diglycine motif was analyzed by alanine replacement. The G222A mutant exhibited strongly reduced activity toward Gal␤1-3Gal␤1-O-MP and Gal␤1-3Gal(6-sulfate)␤1-O-MP (3.7-and 7.1-fold, respectively) (Fig. 5B). Substitution of Gly 223 to alanine led to a near complete loss of GlcAT-I activity toward both substrates (Fig. 5B). The double Gly 222 -Gly 223 to alanine mutation yielded a totally inactive enzyme, indicating that increasing the side chain of the residues at these positions prevents GlcAT-I activity (Fig.  5B). Finally, the effects of the mutation of Lys 317 and of the adjacent Glu 318 (belonging to the second GlcAT-I monomer) were also evaluated. Unexpectedly, the results indicated that alanine substitution of Glu 318 only slightly reduced enzyme activity toward the unsulfated analog and did not change that of Gal␤1-3Gal(6-sulfate)␤1-O-MP. By contrast, the mutation of Lys 317 to alanine led to a complete loss of GlcAT-I activity toward both substrates tested (Fig. 5B). The replacement of Lys 317 by an arginine residue restored GlcAT-I activity to some extent (to ϳ15% of the initial activity), emphasizing the importance of a positive charge at this position. DISCUSSION The biosynthesis of the common GAG-protein linkage region is a key step in the assembly of PGs, because completion of this tetrasaccharide sequence is essential for the conversion of core proteins to functional PGs. Attempts to elucidate signal elements that regulate GAG initiation and elongation have recently focused on the structure of the linkage region. In this study, we investigated, for the first time, the influence of the two major modifications of the tetrasaccharide sequence, i.e. the phosphorylation of Xyl and sulfation of Gal1/Gal2 on the activity of human GalT-I and GlcAT-I enzymes, respectively. This could be achieved by the development of a stereo-controlled high-yielding synthesis of phosphorylated and sulfated analogs that have been tested as substrates of the recombinant GalT-I and GlcAT-I expressed in yeast P. pastoris. This strategy has been proven as highly valuable for probing the specificity of other glycosyltransferases such as mannosyltransferases (34).
A major finding of this study was that GalT-I did not show activity toward the C-2-phosphorylated xyloside, suggesting that the presence of this modification on the acceptor substrate precludes recognition and/or transfer of Gal onto xyloside derivatives. This observation was consistent with a biosynthetic mechanism in which phosphorylation would occur once Gal was attached to the Xyl residue of the nascent oligosaccharide linkage. In agreement, the studies of the biosynthesis of human decorin expressed in rat fibroblasts by pulse-chase experiments showed that phosphorylation was prominent after the addition of the two Gal monosaccharides onto the Xyl residue (35). On   the other hand, these authors found that a proportion of the xylosylated decorin intermediate was phosphorylated. However, this work did not indicate whether or not the phosphorylated xylosylated decorin was further engaged in the GAG biosynthetic pathway. Since we showed that phosphorylated xyloside cannot serve as a substrate for GalT-I, it can be suggested that phosphorylation may arrest the biosynthesis of some GAG chains and thus may represent a regulatory mechanism in the biosynthesis rate of PGs. In a similar way, ␣-Gal-NAc capping of the core tetrasaccharide primer (36) and 4,6disulfatation of GalNAc residues of CS chains (37) have been postulated to represent chain termination mechanisms. However, this issue awaits further investigation.
In addition to phosphorylation, sulfation has been shown to occur on the Gal residues of the linkage tetrasaccharide region of numerous CS-PGs (see . Systematic synthesis of variously sulfated analogs of (Gal␤1-3Gal)␤1-O-MP offered a unique opportunity to examine whether sulfation was important with regard to the specificity of GlcAT-I. Interestingly, we demonstrated here that the efficiency of GlcAT-I was markedly enhanced when Gal1 of the acceptor substrate was sulfated at the C-6 position. The affinity of the full-length membranebound protein and purified enzyme was substantially higher for the sulfated acceptor substrate than for the nonsulfated analog, suggesting that the transfer of GlcA on the sulfated species would be favored. On the other hand, no GlcAT-I activity was observed when Gal1 was sulfated at the C-4 position or when Gal2 was sulfated at either C-4 or C-6 position or when Gal1 and Gal2 were sulfated on C-6 position. The structure of GlcAT-I indicated that the majority of the hydrogen bond interactions between GlcAT-I and the acceptor substrate were through the O-6-, O-4-, and O-3-hydroxyl groups of Gal2, consistent with the idea that the presence of a sulfate substituent at any of these positions may prevent substrate binding. In agreement, no satisfactory docking of the sulfated digalactosides in GlcAT-I active site with the exception of Gal␤1-3Gal(6sulfate)␤1-O-MP could be achieved. Because the catalytic efficiency of GlcAT-I was significantly higher with Gal␤1-3Gal(6sulfate) derivative than with the nonsulfated analog, we further investigated how the enzyme could accommodate this structure within the active site. For this purpose, the systematic mutations of amino acid residues that were in the vicinity of Gal1 based on structural and molecular modeling were carried out. The mutation of Trp 243 residue highlighted the importance of the aromatic side chain of tryptophan and that of phenylalanine in the W243F mutant. This residue appeared essential in substrate positioning via stacking interaction with Gal1 in the case of both sulfated and unsulfated digalactosides. This result corroborated the studies carried out on the related ␤1,3-glucuronosyltransferase GlcAT-P involved in the synthesis of HNK1 epitope emphasizing the importance of stacking interactions between Phe 245 of GlcAT-P (corresponding to Trp 243 in GlcAT-I) and the GlcNAc ring of the acceptor substrate, N-acetyllactosamine (Gal␤1-4GlcNAc) (33). Accordingly, several studies indicated that aromatic residues were main determinants for specificity of glycosyltransferases such as cyclodextrin-glycosyltransferase (38). In addition, we showed that increasing the size of the side chain of the two sequential glycine residues, Gly 222 and Gly 223 (either individually or in combination), by alanine replacement strongly impaired GlcAT-I activity. Because these residues were not in hydrogen bond distance to the acceptor substrate, it can be suggested that they play a role in the organization of the active site rather than being directly involved in interactions with the digalactoside acceptor substrate. In a model of GlcAT-P complexed to Lewis X, the fucose moiety of this saccharide structure clashed with Gly 223 and Gly 224 (corresponding to Gly 222 and Gly 223 in GlcAT-I), supporting the notion that these residues were important active site residues of the ␤1,3-glucuronosyltransferases (33).
Furthermore, the structural analysis of GlcAT-I and GlcAT-P suggested that the C terminus extending from the neighboring monomer may be involved in the interaction with Gal1 or with GalNAc of the disaccharide acceptor substrate of these enzymes, respectively. In GlcAT-P, it has been shown that Val 320 and Asn 321 , which corresponded to Lys 317 and Gln 318 in the GlcAT-I sequence, played an important role in the recognition of GlcNAc moiety. Unexpectedly, we showed that the mutation of Gln 318 , predicted to interact with O-6 of Gal1, into alanine had little effect on catalytic activity of GlcAT-I toward either sulfated or unsulfated digalactoside, indicating that the functional role of this residue was not critical for binding or specificity. By contrast, the mutation of Lys 317 produced a deleterious effect on GlcAT-I activity, emphasizing the contribution of this residue of the neighbor monomer in substrate recognition. Altogether, our results favor the idea that the stacking interactions between both unsulfated and C-6sulfated digalactosides and Trp 243 together with the presence of the sequential Gly 222 -Gly 223 and Lys 317 are essential for substrate recognition and GlcAT-I activity.
Our data suggest that sulfation is a key element influencing GlcAT-I activity, which may regulate the processing of PG biosynthesis. In a similar way, it has been demonstrated that sulfation is important with regard to the specificity of the glycosyltransferases involved in CS chains synthesis. Early studies from Kitagawa et al. (39) on a bovine serum ␣-GalNActransferase using various CS oligosaccharides analogs ((GlcA␤1-3GalNAc␤4-) n ) as substrates show that C-4 sulfation of the penultimate GalNAc unit markedly inhibits the enzyme activity. On the other hand, Sato et al. (40) show strong activity of the recombinant CSGalNAcT-2 toward sulfated oligosaccharide and polysaccharide CS substrates, suggesting that sulfation stimulates this CS-synthase and possibly elongation of CS chains. In a recent study, Seko et al. (41) demonstrate that the ␤1,4-galactosyltransferase 4 preferred keratan sulfate-related oligosaccharides to nonsulfated GlcNAc residues as acceptor substrates. Altogether, these studies and our data support the idea that sulfation represents an important regulatory mecha-nism of GAG processing and assembly.
In conclusion, we demonstrate for the first time that phosphorylation and sulfation critically determine the specificity of two glycosyltransferases involved in the biosynthesis of the GAG-protein linkage region. This study emphasizes the potential role of selective modification of the tetrasaccharide primer sequence in the regulation of biosynthesis of the growing GAGlinkage region and therefore in the conversion of nonglycanated to glycanated PGs. The elucidation of the structure and function of glycosyltransferases involved in the biosynthetic pathway of GAGs recently became a major challenge because of their implication in several pathologies and their potential as pharmacological targets. Glycosaminoglycan precursors are currently proposed as anti-amyloid or anti-thrombotic agents (42,43) and in anti-cancer therapy (44). A better understanding of the specificity of the glycosyltransferases responsible for priming GAG biosynthesis achieved in this study thus represents an important step toward the development of GAG primers as potential therapeutic agents.