Crystal Structure of (cid:1) 1,3-Glucuronyltransferase I in Complex with Active Donor Substrate UDP-GlcUA*

(cid:1) 1,3-Glucuronyltransferase (GlcAT-I) is an essential enzyme involved in heparan sulfate and chondroitin sulfate biosynthesis. GlcAT-I is an inverting glycosyltransferase that catalyzes the transfer of glucuronic acid (GlcUA) to the common growing linker region Gal (cid:1) 1– 3Gal (cid:1) 1–4Xyl that is attached to a serine side chain of a core protein. Previously the structure of GlcAT-I has been solved in the presence of the donor product UDP and an acceptor analog Gal (cid:1) 1–3Gal (cid:1) 1–4Xyl (Pedersen, L. C., Tsuchida, K., Kitagawa, H., Sugahara, K., Darden, T. A. & Negishi, M. (2000) J. Biol. Chem. 275, 34580– 34585). Here we report the x-ray crystal structure of GlcAT-I in complex with the complete donor UDP-Glc-UA, thereby providing structures of an inverting glycosyltransferase in which both the complete donor and acceptor substrates are present in the active site. This structure supports the in-line displacement reaction mechanism previously proposed. It also provides infor-mation on the essential amino acid residues that deter-mine donor substrate specificity. In this scheme, Glu acts as a catalytic base that deprotonates the 3-hydroxyl of the terminal Gal moiety, thereby increasing its nucleophilic character for an in-line attack on the C-1 carbon of the GlcUA of the UDP-GlcUA molecule. The transition state resembles an oxo-carbenium ion-like transition state. The charge build-up on group manganese

Glycosaminoglycans such as heparan sulfates and chondroitin sulfates are attached to proteoglycans that are distributed on the cell surface and extracellular matrix (2). These molecules have been implicated in a variety of biological processes such as cell growth and differentiation, blood coagulation, and viral and bacterial infection (3,4). Both heparan sulfate and chondroitin sulfate utilize the common linker region GlcUA␤1-3Gal␤1-3Gal␤1-4Xyl-Ser to attach to the core protein at specific serine residues. GlcAT-I 1 catalyzes the transfer of the GlcUA from the donor substrate UDP-GlcUA to the linker region Gal␤1-3Gal␤1-4Xyl-Ser (5,6). In heparan biosynthesis this step is followed by the addition of N-acetylglucosamine (GlcNAc) by ␣1,4-N-acetylglucosaminyltransferase (EXTL-2) to the GlcUA of the linker (7). The heparan polymerases (EXTs) elongate the polysaccharide chain by alternating the addition of GlcUA and GlcNAc (8). The bifunctional N-deacetylase/N-sulfotransferases (NDSTs) initiate sulfation of heparan by first deacetylation of certain GlcNAc moieties at the N2 position followed by sulfation of the nitrogen (9). This in turn allows for epimerization of certain GlcUA residues to iduronic acid by C5-epimerase followed by sulfation by specific sulfotransferases such as the 3-O-sulfotransferase, the 2-O-sulfotransferase, and the 6-O-sulfotransferase (3,4).
Disruption of the enzymes involved in glycosaminoglycan biosynthesis has severe biological consequences in mammals. In humans, mutations in the genes EXT1 and EXT2 encoding for heparan polymerases lead to a disease known as hereditary multiple exostoses, which is characterized by cartilage-capped tumors (8). Mice deficient in NDST-1 die in the neonatal state and exhibit phenotypes for pulmonary hypoplasia (10). Mice deficient in NDST-2 exhibit abnormal mast cells (11,12). In addition mice deficient of another heparan sulfotransferase, heparan sulfate 2-O-ST, die in the neonatal state from renal agenesis (13). Because GlcAT-I lies before these enzymes in the biosynthesis cascade, it appears that its function is essential for proper growth and development as well. To better understand the mechanistic properties of enzymes involved in glycosaminoglycan biosynthesis we have used x-ray crystallography to examine amino acid residues required for substrate recognition as well as catalysis for the GlcAT-I enzyme. The enzyme GlcAT-I can be divided into three regions: a membrane-spanning region (residues 8 -25), a proline-rich stem region (residues 26 -74), and a catalytic domain (residues 75-335). We previously solved the structure of the catalytic domain GlcAT-I in complex with the donor product UDP and an acceptor analog Gal␤1-3Gal␤1-4Xyl (1).
We now report the structure of the enzyme in the presence of the active donor UDP-GlcUA. This structure not only reveals the residues involved in recognizing the GlcUA but also supports the mechanism previously proposed for inverting glycosyltransferases. This provides insight for glycosaminoglycan biosynthesis as well as for many processes that utilize UDPsugar-dependent glycosyltransferases.

MATERIALS AND METHODS
Protein of the recombinant human GlcAT-I catalytic domain was cloned, expressed, and purified as previously described (1). Crystals of the enzyme were obtained by the vapor diffusion hanging drop method. 4 l of 15 mg/ml GlcAT-I in 25 mM HEPES pH 7.5, 50 mM NaCl, 10 mM MnCl 2 , and 10 mM UDP-GlcUA was mixed with 4 l of the reservoir solution containing 21% monomethyl ether polyethylene glycol 2000 (MME-PEG-2000) and 100 mM MES, pH 6.0. Crystals were grown at 23°C.
Although the protein was crystallized in the presence of UDP-GlcUA, only the UDP portion of the molecule was visible in the electron density when data were collected on these crystals (1). Two possibilities exist for the lack of electron density. The first is that the GlcUA portion of the molecule is highly mobile and not fixed in a single conformation. The other possibility is that on the time scale of the crystallization the bond in UDP-GlcUA is hydrolyzed releasing the GlcUA. The latter case has been demonstrated for previous crystallization attempts of T4 phage ␤-glucosyltransferase with UDP-Glc (14).
To overcome this problem, crystals were transferred to 4°C. After 30 min crystals were transferred over 1 h in four steps from a solution containing 21% MME-PEG-2000, 100 mM MES, pH 6.0, 10 mM MnCl 2 , and 10 mM UDP-GlcUA to a solution containing 23% MME-PEG-2000, * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
100 mM MES, pH 6.0, 10 mM MnCl 2 , 10 mM UDP-GlcUA, and 10% ethylene glycol. Crystals were then immediately flash frozen in a nitrogen gas stream at Ϫ170°C. This technique proved successful as we obtained strong interpretable electron density for the entire UDP-GlcUA molecule (Fig. 1).
Diffraction data of GlcAT-1 crystals in the presence of UDP-GlcUA were collected using a RUH3R rotating anode generator and a RAXIS IV area detector system. Reflection data were processed using DENZO and SCALEPACK (15). The model of GlcAT-I from the ternary complex minus the UDP and acceptor substrate atoms was used as the starting model for refinement. Coordinates were refined using multiple cycles of manual model building with the program "O" (16) and refinement in CNS (17). The final model contains residues Met 75 -Pro 140 , Trp 152 -Val 335 in molecule A and residues Met 75 -Pro 140 , Ala 154 -Glu 244 , Arg 247 -Val 335 in molecule B (Table I). Each unit contains one molecule of UDP-GlcUA bound in the donor binding site. The geometry of the model was analyzed using PROCHECK (18).

RESULTS AND DISCUSSION
UDP-GlcUA Binding-The catalytic domain of GlcAT-I crystallizes with two molecules in the asymmetric unit. These two molecules have been suggested to represent the physiological dimer (1). The catalytic domain of GlcAT-I is made up of two subdomains. Residues 75-197 comprise the NTP-sugar donor substrate binding subdomain, whereas residues 198 -308 define the acceptor substrate binding subdomain. The UDP-GlcUA binding subdomain consists of an ␣/␤ Rossmann-like motif. The position of the uridine base is fixed through a hydrogen bond with conserved residue Asp 113 (3.0 Å) and a ring stacking interaction with Tyr 84 (Fig. 1, left and right). The diagnostic feature of the donor binding subdomain is a DXD motif (DDD for GlcAT-I) found at residues Asp 194 -Asp 196 . These residues are involved in hydrogen bonding to the UDP portion of the molecule as well as interacting directly with the metal ion (Mn 2ϩ ) required for catalysis. Asp 194 forms a hydrogen bond (2.9 Å) with one of the two water molecules coordinating the Mn 2ϩ ion. Asp 195 forms a hydrogen bond with the O 2 oxygen of the ribose ring (2.9 Å). Both oxygen atoms from the carboxylate group of Asp 196 interact with the Mn 2ϩ metal ion (2.2 Å). The other atoms involved in the octahedral coordination of the Mn 2ϩ ion are oxygen atoms from the ␣ and ␤ phosphates (2.1 and 2.1 Å) of the UDP-GlcUA as well as the two water molecules mentioned above (2.1 and 2.1 Å).
The position of the GlcUA is determined though extensive interactions with conserved residues (Figs. 1, left, and 2). The 2-hydroxyl forms a hydrogen bond with the NE2 nitrogen of His 308 (2.7 Å) and with OD2 of Asp 252 (2.5 Å). Asp 252 also forms a hydrogen bond with the 4-hydroxyl on the acceptor Gal. This is the only residue that forms hydrogen bonds with both the donor and acceptor molecules (1). The 3-hydroxyl interacts with the OD1 atom of Asp 194 (2.9 Å) from the DXD motif. The 4-hydroxyl forms hydrogen bonds with both the NH2 nitrogen of Arg 161 (3.0 Å) and the carbonyl oxygen of Arg 156 (3.0 Å). Finally, an oxygen from the carboxylate group of GlcUA is within hydrogen bonding distance of the backbone amide of Arg 156 (2.8 Å).
Implications for Catalysis-Previously it has been proposed that the catalytic reaction for the inverting glycosyltransferases proceeded through an in-line displacement reaction mechanism (1,19,20). In this scheme a base deprotonates the 3-hydroxyl of the terminal Gal to form a strong nucleophile. The nucleophile then attacks the C1 carbon of the GlcUA from the UDP-GlcUA donor and forms an oxo-carbenium ion-like transition state (Fig. 3). Finally the UDP then dissociates from the GlcUA␤1-3Gal␤1-3Gal␤1-4Xyl-Ser product. This would result in an inversion of the stereochemistry with respect to the C-1 bond converting it from an ␣ linkage with the UDP to a ␤ linkage with respect to the Gal␤1-3Gal␤1-4Xyl. Previously it was assumed that the C1 carbon would be in-line with the leaving oxygen on the ␤ phosphate of UDP and the 3-oxygen of the terminal Gal. However, there were no data to indicate this was true. Therefore it became imperative to obtain the crystal structure of GlcAT-I not only of the UDP/Gal␤1-3Gal␤1-4Xyl complex but also with the entire UDP-GlcUA donor bound.
Superposition of the ternary complex with UDP/Gal␤1-3Gal␤1-4Xyl and the UDP-GlcUA structures reveals the relative orientation of the donor to the acceptor substrate (Fig. 2). We find that the acceptor substrate binds in a groove in the acceptor binding domain with the Xyl portions of the Gal␤1-3Gal␤1-4Xyl analog extending away from the protein and with the 3-hydroxyl of the terminal Gal 5.1 Å from the ␤ phosphate of the UDP molecule. Conserved residues Glu 227 and Arg 247 form hydrogen bonds with the 6-hydroxyl of the terminal Gal whereas Asp 252 forms a hydrogen bond with the 4-hydroxyl. Glu 281 forms a hydrogen bond with the 3-hydroxyl of the terminal Gal moiety, and therefore it has been proposed that this residue may function as the catalytic base.
In the superposition the 3-hydroxyl on the acceptor Gal is located 3.8 Å from the C1 carbon of the GlcUA moiety. The C-1 carbon is located between the leaving oxygen of the ␤ phosphate and the incoming 3-hydroxyl of the Gal. This orientation is highly suggestive of an in-line displacement reaction mechanism that supports the previously proposed mechanism for GlcAT-I. The catalytic region of GlcAT-I shows a great deal of similarity to other inverting glycosyltransferases despite little to no sequence identity. The enzymes N-acetylglucosaminyltransferase I from rabbit (GnT1) (21) and SpsA from Bacillus subtilis (19) both share a great deal of structural similarity not only with the nucleotide binding subdomain of GlcAT-I but the overall fold as well. ␤1,4-Galactosyltransferase from bovine (␤4GalT1) (22-24) also shows some similarities in overall fold to GlcAT-I. Because of the similarities in the fold and the chemical reaction these enzymes catalyze, these enzymes and calculated from working data set. R free is calculated from 5% of data randomly chosen not to be included in refinement.
other NTP-sugar-dependent inverting glycosyltransferases with a similar fold may share a similar mechanism.
Comparison to Retaining Glycosyltransferases-The donor binding subdomain of GlcAT-I shows a great deal of similarity to other glycosyltransferases such as galactosyltransferase LgtC from Neisseria meningitis (25) and bovine ␣1,3-galactosyltransferase (␣3GalT) (26) despite little apparent sequence identity. In addition to the present GlcAT-I structure, four of these structures (GnT1, ␣3GalT, LgtC, and ␤4GalT1) have been solved with a complete donor substrate in place (21, 24 -26). Three of these glycosyltransferase structures contain an acceptor substrate: GlcAT-I, ␤4GalT1, and LgtC. The first two glycosyltransferases catalyze an inverting transfer reaction and therefore are often called inverting glycosyltransferase, whereas the last one is a retaining glycosyltransferase that catalyzes retaining transfer reaction.
To better understand the differences between retaining and inverting transfer reactions we have superimposed the ternary complex structure of LgtC to the previously described ternary complex of GlcAT-I (Fig. 4). Despite the fact that GlcAT-I and LgtC are inverting and retaining enzymes, respectively, the UDP-sugar molecules superimpose well. Interactions of the donor substrates with the enzymes are conserved. For example, the corresponding aspartic acids in the DXD motifs superimpose well. The first position of the DXD motif (Asp 194 and Asp 103 of GlcAT-I and LgtC, respectively) interacts with the donor sugar, and the third position (Asp 196 and Asp 105 , respectively) coordinates with the Mn 2ϩ ion. Arg 161 forms a hydrogen bond with the side chain of position 1 in the DXD motif as well as with the donor sugar in the GlcAT-I structure. Arg 86 is the equivalent residue found in the LgtC structure, thereby indicating that these interactions are conserved in both structures. Asn 153 and Asp 252 are also superimposed in the LgtC and GlcAT-I structures, respectively, and both residues are in position to form a hydrogen bond with the O-2 oxygen of the donor sugar substrate (Fig. 4). In addition, His 244 of LgtC is in a similar location as His 308 of GlcAT1 with respect to the donor sugar, although it does not appear to form an interaction with the sugar like His 308 but rather interacts with the Mn 2ϩ ion. As a result of these similarities, the orientation and conformation of the donor substrate in the active site are conserved in both the inverting GlcAT-1 and retaining LgtC enzymes.
Interestingly, the superposition of the retaining enzyme LgtC with the inverting enzyme GlcAT-I reveals the position of the acceptor substrates to be different with respect to the C-1 carbon of the donor sugar. As a result, the acceptor 4-hydroxyl for LgtC would not be in-line to the leaving UDP (Fig. 4). It has been suggested that the retaining enzymes proceed by a The acceptor oxygen of the 3-hydroxyl is labeled O3. An arrow shows the direction of attack on the C-1 carbon of the GlcUA molecule for the proposed in-line displacement reaction. The root mean square deviation between the two structures is 0.184 Å over 250 C␣ atoms (figure created using Molscript (27) and Raster3D (28)).
suggesting it as a catalytic nucleophile in the first transfer (data not shown). Gln 189 has been mutated to examine its role as a catalytic residue in the transfer reaction (25). Gln 189 mutations decreased but did not abolish activity of LgtC, indicating another mechanism other than a double displacement mechanism may be involved for the retaining enzymes (25). From this analysis it is clear that striking similarities exist between the retaining and inverting glycosyltransferases in the UDP-sugar binding subdomain. However these similarities do not necessarily extend to the acceptor binding subdomain.
Conclusion-We have obtained the crystal structure of GlcAT-I in the presence of the donor substrate UDP-GlcUA. This structure has revealed that side chains from conserved residues Arg 161 , Asp 194 , Asp 252 , and His 308 as well as backbone interactions from conserved residue Arg 156 orient the donor substrate for catalysis. This position of the donor with respect  (27) and Raster3D (28)).