Structural Analysis of the Sulfotransferase (3- O -Sulfotransferase Isoform 3) Involved in the Biosynthesis of an Entry Receptor for Herpes Simplex Virus 1,*

Heparan sulfate (HS) plays essential roles in assisting herpes simplex virus infection and other biological processes. The biosynthesis of HS includes numerous specialized sulfotransferases that generate a variety of sulfated saccharide sequences, conferring the selectivity of biological functions of HS. We report a structural study of human HS 3- O -sulfotransferase isoform 3 (3-OST-3), a key sulfotransferase that transfers a sulfuryl group to a specific glucosamine in HS generating an entry receptor for herpes simplex virus 1. We have obtained the crystal structure of 3-OST-3 at 1.95 Å in a ternary complex with 3 ′ -phosphoadenosine 5 ′ -phosphate and a tetrasaccharide substrate. Mutational analyses were also performed on the residues involved in the binding of the substrate. Residues Gln 255 and Lys 368 are essential for the sulfotransferase activity and lie within hydrogen bonding distances to the carboxyl and sulfo groups of the uronic acid unit. These residues participate in the substrate recognition of 3-OST-3. This structure provides atomic level evidence for delineating the substrate recognition and catalytic mechanism for 3-OST-3. PAPS, 3 ′ -phosphoadenosine 5 ′ -phosphosulfate; NST, N -sulfotransferase; PAP, 3 ′ -phosphoadenosine 5 ′ -phosphate; Tricine, N -[2-hydroxy-1,1-bis(hydroxymethyl)ethyl]glycine; MES, 4-morpholineethanesulfonic acid; HPLC, high pressure liquid chromatography; GlcUA, glucuronic acid.

Heparan sulfate (HS) 1 is ubiquitous on the cell surface and in the extracellular matrix. It has diverse roles in regulating embryonic development and homeostasis as well as in assisting viral infections (1,2). HS is a highly sulfated polysaccharide consisting of 1→4-linked sulfated glucosamine and sulfated glucuronic/iduronic acid units. HS is initially synthesized as a copolymer of glucuronic acid and N-acetylated glucosamine by D-glucuronyl and Nacetyl-D-glucosaminyltransferase, followed by various modifications in the Golgi apparatus (3). These modifications include N-deacetylation and N-sulfonation of glucosamine, C 5 epimerization of glucuronic acid to form iduronic acid units, 2-O-sulfonation of iduronic and glucuronic acid units, and 6-O-sulfonation and 3-O-sulfonation of glucosamine units. The specific sulfated saccharide sequences of HS determine its biological function (4).
Infections caused by herpes simplex virus type 1 (HSV-1) are highly prevalent in human and result in localized mucocutaneous lesions and, in rare cases, encephalitis (5). HS plays critical roles in assisting HSV-1 infections in both viral attachment and entry steps (6). In the attachment step, HSV-1 binds to the cells through the interactions of envelope glycoproteins gC or/and gB with cell surface HS (7). Viral entry requires the interaction of a third viral glycoprotein, gD, with a specific cell surface entry receptor to induce fusion of the viral envelope with the cell membrane (in the presence of gB, gH, and gL) to establish an infection (8). The 3-O-sulfated HS, representing the culmination of a series of specific sulfated saccharide sequences, binds to gD and serves as an entry receptor of HSV-1 (9,10). The 3-O-sulfated HS is synthesized by heparan sulfate 3-O-sulfotransferase isoform 3 (3-OST-3) and by 3-OST-5 (9,10). A gD-binding HS octasaccharide was isolated and characterized, confirming that HSV-1 utilizes a unique HS sequence for its entry (11). Understanding the biosynthesis of 3-O-sulfated HS could potentially lead to a new strategy for developing therapeutic agents against HSV infection. HS 3-O-sulfotransferase is present in at least six different isoforms with unique expression patterns in human tissues (10,12). The amino acid sequences of the different isoforms have greater than 60% homology in the sulfotransferase domains (12). These different 3-OST isoforms transfer the sulfuryl group from 3′-phosphoadenosine 5′-phosphosulfate (PAPS) to the 3-OH position of glucosamine units residing within the context of different saccharide sequences. As a result, the HS products generated by these different isoforms exhibit unique and distinctive biological activities (Fig. 1A). It is known that the HS modified by 3-OST-1 and 3-OST-5 displays anticoagulant activity. The HS modified by 3-OST-3 and 3-OST-5 serves as an entry receptor for HSV-1 (9,10,13). The molecular mechanism used by 3-OSTs to distinguish substrates with unique saccharide sequences remains largely unknown.
Previously, x-ray crystal structures have been obtained for the sulfotransferase domains of human N-deacetylase/N-sulfotransferase (NST-1) and mouse 3-OST-1 in the complex with 3′-phosphoadenosine 5′-phosphate (PAP) (14,15). The structural information combined with the results of mutational analysis provided us with preliminary insight into the HS sulfotransferase mechanism (14,16). Furthermore, potential roles of specific amino acid residues from NST-1 and 3-OST-1 for directing the transfer of the sulfuryl group were suggested (14). However, because these structures lacked acceptor substrate, they have provided limited knowledge to our understanding of what dictates HS binding, in particular, among the 3-OST isoforms.
In this study, we report the crystal structure of a ternary complex of human 3-OST-3, PAP, and a tetrasaccharide substrate with the structure of δUA2S-GlcNS6S-IdoUA2S-GlcNS6S (for structure see Fig. 1B). This ternary complex structure, combined with the mutagenesis data presented here, reveals the residues involved in catalysis and in substrate binding. The data provide structural evidence for which residues confer substrate specificity between the 3-OST isoforms. These results present new insight into the biosynthesis of the HS with various biological functions ranging from viral infection to blood coagulation.

EXPERIMENTAL PROCEDURES Expression, Purification, and Crystallization of 3-OST-3
Preparation of 3-OST-3 Bacterial Expression Plasmid-The cDNA fragment encoding the catalytic domain of 3-OST-3 (Gly 139 -Gly 406 ) was amplified from plasmid h3-OST-3A-pcDNA3 (17) with a 5′ overhang containing an EcoRI site and a 3′ overhang containing an NotI site. This construct was inserted into the pGEX4T3 vector (Amersham Biosciences) using the EcoRI and NotI restriction sites to produce a glutathione Stransferase fusion protein. The resultant plasmid was sequenced to confirm the reading frame and the lack of mutations within the coding region (University of North Carolina, DNA sequencing core facility). The plasmid was transformed into BL21(DE3)RIL cells (Stratagene).
Protein Expression and Purification-Cells containing the expression plasmid were grown in twelve 2.8-liter Fernbach flasks containing 1 liter of LB medium with 100 μg/ml of ampicillin at 37 °C. When the A 600 reached 0.6-0.8, the temperature was lowered to 22 °C for 15 min. Isopropyl-β-D-thiogalactopyranoside was then added to a final concentration of 200 μM, and the cells were allowed to shake overnight. The cells were pelleted and resuspended in 120 ml of sonication buffer (25 mM sodium/potassium phosphate, pH 7.4, and 625 mM NaCl). The cells were sonicated, and the cell debris was pelleted by centrifugation. The soluble fraction was loaded onto 20 ml of glutathione-Sepharose 4B (Amersham Biosciences) resin in batch and washed repeatedly with sonication buffer. The catalytic domain of 3-OST-3 was cleaved from the fusion protein on the resin at 4 °C overnight using 400 units of thrombin in a total volume of 50 ml. PAP (Sigma) was added to the solution for a final concentration of 0.5 mM. The protein was concentrated to ~35 mg/ml. Protein was further purified by a Superdex 75 16/60 (Amersham Biosciences) column. The eluted fractions containing pure 3-OST-3 were pooled. The buffer was exchanged with the crystallization buffer (20 mM Tris, pH 7.5, 100 mM NaCl, and 1 mM PAP) by repeated concentration and dilution. The final concentration of the protein was 10 mg/ml, and additional PAP was added for a final concentration of 4 mM.
Protein Crystallization, Data Collection, and Structure Solution-Crystals of the catalytic domain of 3-OST-3 were obtained using the hanging drop method by mixing 2 μl of the above protein solution with 2 μl of the reservoir solution consisting of 12-13% polyethylene glycol 4000, 200 mM ammonium acetate, and 100 mM sodium citrate, pH 5.5. For data collection, the crystals were transferred in four steps into a cryo-protectant consisting of 15% polyethylene glycol 4000, 12.5% ethylene glycol, 4 mM PAP, 200 mM ammonium acetate, 100 mM NaCl, and 100 mM sodium citrate, pH 5.5. To obtain the ternary complex, the crystal was soaked overnight in cryo-protectant containing 20 mM of the tetrasaccharide. The crystals were frozen in liquid nitrogen and then placed in a stream of nitrogen gas cooled to −180 °C. The data were collected on a Rigaku RU3H generator fitted with Osmic mirrors and a Raxis IV area detector. The data were indexed, integrated, and scaled using HKL2000 (18). Coordinates from the 3-OST-1 isoform (Protein Data Bank code 1S6T) were used to find a solution for the molecular replacement problem using the program Molrep from CCP4 (19,20). The model of 3-OST-3 was refined by iterative cycles of model building in O (21) and refinement in CNS (22). The starting model for the tetrasaccharide was obtained from the Protein Data Bank coordinates 1BFB (23).

Mutational Analysis of 3-OST-3
Preparation of 3-OST-3 Mutant Plasmids-The cDNA fragment encoding the catalytic domain of 3-OST-3 (Gly 139 -Gly 406 ) was amplified from plasmid h3-OST-3A-pcDNA3 (17) with a 5′ overhang containing an NdeI site and a 3′ overhang containing an EcoRI site. This construct was inserted into the pET28a vector (Novagen) using the NdeI and EcoRI restriction sites to produce a His 6 -tagged protein. The mutants were prepared using Gene Tailor site-directed mutagenesis kit from Invitrogen as described elsewhere (14).

Expression and Characterization of 3-OST-3-
The procedures for the expression and purification of 3-OST-3 were essentially identical to those for 3-OST-1 as described in a prior publication (14). To confirm that bacterial 3-OST-3 has similar activity to the protein expressed in SF9 cells, we conducted a kinetic analysis to determine the K m and V max toward PAPS and HS, respectively. The K m and V max toward PAPS of the 3-OST-3 expressed in Escherichia coli is 40 μM and 27 pmol of sulfate/min, respectively, which are close to the K m value (16 μM) and V max value (27 pmol of sulfate/min) of the 3-OST-3 expressed in SF9 cells. The K m and V max toward HS of the 3-OST-3 expressed in E. coli is 2.5 μM and 24 pmol of sulfate/min, respectively, which are close to the K m (1 μM) and V max (21 pmol of sulfate/min) of the 3-OST-3 expressed in SF9 cells.

Expression and Purification of 3-OST-3 Mutants-
The expression plasmids for various 3-OST-3 mutants were transformed individually into BL21(DE3)RIL cells (Stratagene). Each mutant construct was grown in 500 ml of LB broth and induced by isopropyl-β-D-thiogalactopyranoside. The bacteria were harvested and sonicated. The lysate was subjected to a 400-μl nitrilotriacetic acid-agarose column (0.75 cm × 1 cm), followed by 5 ml of wash with 10 mM imidazole, 25 mM Tris, and 500 mM NaCl, pH 7.5. Mutant proteins were eluted with 1 ml of elution buffer containing 25 mM Tris, 500 mM NaCl, and 250 mM imidazole, pH 7.5. Approximately 25 μl of the eluent was subjected to the analysis on a 16.5% Tris-Tricine-PAGE gel (Bio-Rad), and the gel was stained by Coomassie Blue. The expression level of the mutant protein was estimated by determining the intensity of the Coomassie-stained protein band near 30 kDa. Wild type 3-OST-3 was expressed along with the mutants. The expression levels and the sulfotransferase activity of mutant proteins were normalized to those of wild type protein.

Disaccharide Analysis of the HS Modified by Wild Type 3-OST-3 and 3-OST-3
Mutants-3-OST-3-modified HS was degraded with nitrous acid at pH 1.5, followed by reduction with sodium borohydride (24). The resultant 35 S-labeled disaccharides were resolved by a C 18 reversed phase column (0.46 × 25 cm) (Vydac) under reversed phase ion pairing HPLC conditions. The identities of the disaccharides were determined by coeluting with appropriate 35 S-labeled disaccharide standards (25).

Determination of 3-O-[ 35 S]Sulfonation Site on the Tetrasaccharide
We utilized a combination of the digestions of heparin lyase and δ 4,5 -glycuronate-2-sulfatase to determine the 3-O-[ 35  The conditions for δ 4,5 -glycuronate-2-sulfatase digestion are described elsewhere (11). Digestion with heparin lyase of the tetrasaccharide was carried out using a mixture of heparin lyase I, II, and IV. The digestion condition for the mixture of heparin lyases was described in a prior publication (11). The anion exchange HPLC was performed on a silical based amino-bound anion exchange HPLC (0.46 × 24 cm; Waters) eluted with a linear gradient of KH 2 PO 4 (300-1000 mM in 60 min) followed by 1 M KH 2 PO 4 for an additional 20 min at a flow rate of 0.5 ml/min. We noted that this 3-O-sulfated tetrasaccharide is resistant to nitrous acid (at pH 1.5) degradation.

Crystal Analysis
Human 3-OST-3 crystallizes in space group C222 1 and diffracts to 1.85 Å resolution. Two molecules were present in the asymmetric unit (molecules A and B), and each protein molecule contains one molecule of PAP. The crystals of 3-OST-3 are densely packed, generating a unit cell with a Matthews coefficient of 2.0 and an overall solvent content of 38.5% (28). Both molecules in the asymmetric unit are well ordered and have similar Bfactors (Table I). The data were collected on a similar crystal soaked in the tetrasaccharide at 1.95 Å. One of the two molecules in the asymmetric unit has clear electron density for the tetrasaccharide that allows for precise positioning. The other molecule has some electron density in the binding pocket region, but it is not clear enough to model.

Overall Fold
The catalytic domain of human 3-OST-3 (Gly 139 -Gly 406 ) is roughly spherical with a large open cleft running across the surface of the sphere (Fig. 2). At the core of the sphere is an α/β motif. This α/β motif is comprised of a five-stranded parallel β-sheet (β7-β6-β2-β5-β3), laterally flanked by α-helices. Central to this structural motif is the phosphosulfate-binding loop (K162-R166), contained within a strand-loop-helix comprised of β2 and α2 (29). The phosphosulfate-binding loop allows for several strong hydrogen bonding interactions with the 5′-phosphate of PAP ( Fig. 3 and Table II). These interactions are believed to help position the donor substrate, PAPS, in the precise location for catalysis (30). This binding site is well conserved with the previously reported 3-OST-1 and NST-1 (14,15). The 5′phosphate of PAP is found at the bottom of the large open cleft that provides the binding pocket for HS substrate (Figs. 2 and 3). A conserved α-helix (α5) composed of residues Ala 244 -Lys 259 is positioned in close proximity to the PAP-binding site and protrudes outward into the open cleft. The C-terminal region of the protein contains a small threestranded anti-parallel β-sheet (β8-β9-β10). The second and third strands of this sheet are structurally reinforced by a disulfide bond between residues Cys 351 and Cys 363 . Strand 10 of this sheet is connected to the C-terminal β-helix (β11) via a long coil (Lys 366 -Arg 377 ) that lies along one side of the open cleft, partially obscuring the PAP-binding site.
The structure of 3-OST-3 is remarkably similar to that of 3-OST-1, with a root mean square deviation of 1.1 Å for 246 equivalent Cα atoms as calculated by the software program O (14,21), although some minor differences present within looped regions were notable (Fig.  2). In contrast to 3-OST-1, 3-OST-3 has an insertion of three amino acids (Ala 355 -Gly 357 ) into the loop formed by the disulfide bond between Cys 351 and Cys 363 in strands 9 and 10 of the C-terminal β-sheet. These additional residues cause the loop to swing in the direction of the open cleft. Replacing the disulfide bond loop regions of 3-OST-1 (Cys 260 -Cys 269 ) with the sequence of 3-OST-3 (Cys 351 -Cys 363 ) has no effect on the substrate specificity of 3-OSTs (data not shown). This finding suggests that this loop structure does not directly contribute to the substrate specificity of 3-OST-1 and 3-OST-3. Another notable difference between 3-OST-1 and 3-OST-3 is the presence of a N-terminal α-helix, Pro 136 -Glu 147 (residues Pro 136 -Ser 138 are cloning artifacts from the expression vector), found in 3-OST-3. This helix packs along the side of the protein in a shallow groove formed by the C-terminal three-stranded β-sheet of a neighboring molecule in the crystal. However, it is external to the conserved sulfotransferase domain.

Tetrasaccharide-binding Site
The structure of the protein with bound substrate is virtually identical to that of the apoprotein with respect to the backbone positions. A small number of side chains were altered upon substrate binding, most of which are found in or near the HS-binding site. The tetrasaccharide (δUA2S-GlcNS6S-IdoUA2S-GlcNS6S; Fig. 1B) binds in an extended conformation along the open cleft. This position places the 3-OH group of the glucosamine unit (G3) in close proximity to the PAP-binding site within the open cleft. It is interesting to note that the tetrasaccharide appears to behave rather like two distinct disaccharide units (Fig. 4). The two units on the reducing end (I2-G1) have fewer potential hydrogen bonding partners than those (U4-G3) on the nonreducing end and thus are more disordered and consequently have higher B-factors (Table I). In addition, the sugar rings of U4-G3 lie in the same plane that is rotated almost 90° around the α(1→4) glycosidic bond, with respect to the plane of the sugar rings of I2-G1. A similar twist has been previously observed in the basic fibroblast growth factor/HS structure (23).
The saccharide that forms the most extensive interactions with the protein is the uronic acid (U4) (Fig. 4, a and b, Table IV). This observation suggests that this sugar unit contributes significantly to the substrate specificity for 3-OST-3. Residues Gln 255 , Lys 259 , and Arg 370 are all within hydrogen bonding distance to the 2-O-sulfo group. Residues Arg 166 , Lys 215 , and Lys 368 are all within hydrogen bonding distance to the carboxyl group (Table IV). In addition to interactions with the charged groups, the hydroxyl from residue Thr 367 is within hydrogen bonding distance to the 3-OH of U4. It should be noted that δ 4,5 unsaturated 2-Osulfo uronic acid (δUA2S or U4 in this tetrasaccharide), a product of heparin lyase digestion, is a saturated saccharide unit (in the form of 2-O-sulfo iduronic acid, or IdoUA2S) in HS polysaccharide. The 2 So conformation of 2-O-sulfo iduronic acid and the 2 H 1 (31) conformation of δUA2S (U4) unit allow similar positioning of the 2-O-sulfo, 3-OH, and 5carboxyl groups. Thus, the δUA2S mimics IdoUA2S sufficiently well to be recognized by 3-OST-3. Furthermore, the solution conformation of this tetrasaccharide indicates that neither the I2 or U4 units substantially modify the conformations of the adjacent G1 and G3 residue (31). Indeed, this tetrasaccharide is a substrate for 3-OST-3 as described below.
The G3 unit appears to be the acceptor site of sulfonation. This unit is present in a 4 C 1 conformation, and its 3-OH is positioned within hydrogen bonding distance to the glutamate residue (Glu 184 ) (Table IV). It has been demonstrated that the equivalent residue in similar enzymes functions as the catalytic base (14,16). A less ordered sulfo group at the C-6 position was observed, suggesting that it is not essential for substrate specificity. This observation is consistent with results from biochemical studies of 3-OST-3 (11,25). The Nsulfo group is in a position to form a hydrogen bond with the backbone amide of Ser 218 . It has been determined that 3-OST-3 sulfonates an N-unsubstituted glucosamine unit at the polysaccharide level (Fig. 1A) (25). Both glucosamine units (G1 and G3) are N-sulfonated (Fig. 1B) in the tetrasaccharide used in this experiment. Furthermore, the N-sulfo group was well ordered in a confined pocket, although no positively charged amino acid residues are found nearby. The restricted location of the N-sulfo group is perhaps due to the fact that the acceptor 3-OH group is highly oriented, consequently reducing the flexibility of N-sulfo group. In light of these results, we are currently unable to conclusively determine whether this N-sulfo group is required for substrate recognition.
Numerous amino acid residues have contacts with the carboxyl group of the iduronic acid unit (I2). The amino acid residues include Lys 161 , Gln 255 , and Lys 259 (Fig. 4c). In addition, the 3-OH position lies well within hydrogen bonding distance of Arg 190 . Interestingly, the 2-O-sulfo group is not positioned near any hydrogen bonding partners, yet it is reasonably well ordered. The I2 unit is present in the 2 S 0 skew boat conformation, which differs from the conformation of this saccharide bound to basic fibroblast growth factor (23). Recent work has demonstrated the importance of the iduronic acid skew boat conformation in exhibiting the anticoagulant activity (32).
The G1 unit is also present in a 4 C 1 conformation (Fig. 4c). The N-sulfo group interacts with the side chain of Thr 256 as well as with the side chains of Asp 252 , Trp 283 , and Ser 284 (Table  IV). The 6-O-sulfo group makes no hydrogen bonding contacts with the protein.
Consequently, it is the most disordered group in the tetrasaccharide. Interestingly, a metal ion modeled as a sodium ion appears to bind to the I2 and G1 units. The backbone carbonyl and side chain of Asp 252 , as well as the side chain of Thr 256 , provide three of the ligands coordinated to the metal (Fig. 4c, Table III). The other three ligand interactions, which complete the octahedral coordination, are from the N-sulfo and 3-OH of the G1 unit as well as a carboxylate oxygen atom of the I2 unit.

Determination of 3-O-Sulfonation Site on the Tetrasaccharide
To confirm the conclusions from the structural data, we proved that 3-OST-3 transfers the [ 35 S]sulfuryl group to the G3 unit (rather than to the G1 unit) when the tetrasaccharide is used as a substrate. We prepared the 3-O-[ 35 (Fig. 5C, dashed line), whereas the product (δUA-[3-35 S]GlcNS3S6S) of the sulfatasepretreated sample was eluted at 25 min (Fig. 5C, solid line). Therefore, our data demonstrated that the 3-O-[ 35 S]sulfonation site is indeed on the G3 unit as predicted by the crystal structure.

Catalytic Mechanism
The ternary structure of 3-OST-3/PAP/tetrasaccharide provides atomic details about the reaction mechanism of HS sulfotransferases. When PAPS is superimposed onto PAP, the acceptor hydroxyl group of G3 is located 2.9 Å from the sulfur atom of PAPS and on the opposite of the sulfur from the leaving group PAP (Fig. 6). The sulfonate group is easily accommodated by the enzyme without rearrangement of side chain residues. The location of the sulfonate group is consistent with a S N 2-like in-line displacement mechanism. Given that Glu 184 is 2.8 Å from the 3-OH of the G3 unit (Fig. 4), it is reasonable to suppose that this residue likely functions as a catalytic base, deprotonating the 3-OH for nucleophilic attack on the sulfonate group. This glutamate residue is conserved among 3-OSTs and NSTs, thereby suggesting an essential role for the catalytic function of HS sulfotransferases (NSTs transfer the sulfuryl group to the amino position of a glucosamine unit within a HS polysaccharide.). Indeed, mutations of this glutamate residue result in a complete loss of sulfotransferase activities for NST-1 and 3-OST-1 (14, 16) and a 99.9% loss for 3-OST-3 (Table V).
A hydrogen bonding network that involves Glu 184 , His 186 , and Asp 189 is observed in the ternary complex (Fig. 6). The OE2 oxygen of Glu 184 is 2.9 Å from the NE2 atom of His 186 . The ND1 atom of His 186 is 2.8 Å from OD2 of Asp 189 . This hydrogen bonding pattern is reminiscent of a catalytic triad. The results of the mutational analysis demonstrated that His 186 and Asp 189 are essential for the sulfotransferase activity (Table V). This hydrogen bond network could either help properly position the glutamate for catalysis or serve as a charge relay system to regulate the pK a of the glutamate. A similar hydrogen bond network was also observed in 3-OST-1 but not in NST-1 (14,15). Thus, the interactions within the "triad" appear to be specific for 3-OSTs.

Mutational analysis
Site-directed mutagenesis experiments were performed to decipher the roles of the amino acid residues in performing the catalytic functions and determining substrate specificity. The results are summarized in Table V. We conducted the disaccharide analysis of the HS modified by the mutants that maintain the sulfotransferase activity (catalytically active mutants). Our data indicate that wild type protein and all of the catalytically active mutants sulfonate identical disaccharide sequences (IdoUA2S-GlcNH 2 and IdoUA2S-GlcNH 2 6S; see Fig. 1A for structures), suggesting that these mutants maintain wild type 3-OST-3 substrate specificity (data not shown). This conclusion is further strengthened by our findings that none of the HS modified by the active mutants bind to antithrombin (data not shown). This is consistent with our understanding that 3-OST-1-modified HS binds to antithrombin, whereas 3-OST-3-modified HS does not (Fig. 1A).
Several mutations resulted in the loss of enzymatic activity (catalytically inactive mutants). Among these, K162A, E184Q, H186F, and D189N are probably due to defects in catalytic function. These conclusions are based on the crystal structural data as described above, as well as our previous results of the characterization of corresponding 3-OST-1 mutants (14). The roles of Lys 215 and Arg 370 are more complex. These two residues could be involved in both binding to PAPS and substrate. Like Glu 184 and Lys 162 , Lys 215 and Arg 370 are conserved in the 3-OSTs and NST-1.
Two mutants, Q255A and K368A, are of special interest. Mutations at Gln 255 and Lys 368 nearly abolish sulfotransferase activity. However, mutations at the corresponding residues of 3-OST-1, Q166A and K274A, maintain 34.0 and 17.4% of the sulfotransferase activity, respectively (14). The data suggest that Gln 255 and Lys 368 contribute to the substrate specificity of 3-OST-3. Indeed, the structural data indicate that the side chains of Gln 255 and Lys 368 are within the hydrogen bonding distance to the carboxyl and 2-O-sulfo groups of the U4 unit and the carboxyl group of the I2 unit as described above. This is consistent with the fact that 3-OST-1 recognizes a disaccharide substrate (GlcUA-GlcNS±6S) that does not contain the 2-O-sulfo group and the orientation of the carboxyl group of the GlcUA unit is likely distinct from that of an iduronic acid unit (Fig. 1A).
Presently, we are unable to explain the mutation results for Lys 194 and Lys 366 because the side chains of both residues do not interact with the tetrasaccharide. Residue Lys 194 is located over 14 Å from the tetrasaccharide. Residue Lys 366 is located in a position where it could interact with the nonreducing end of the oligosaccharide if one more sugar (a glucosamine) were present, as is the case in the polymeric HS substrate used in the mutation studies. Interestingly, in the x-ray crystal structure, Lys 366 forms a lattice contact with the 6-O-sufo group of G3 from a tetrasaccharide binding in a crystallographically related molecule. This interaction does not appear to be bio-chemically relevant because the 6-Osulfo group is not required for 3-OST-3 activity (11), and the buried surface area between the crystallographic interface of the two protein molecules is less than 3% of the total surface area.
In summary, the structure of the 3-OST-3/PAP/tetrasaccharide has provided new insights into the requirements for the substrate recognition by 3-OSTs and a more complete description of the HS sulfotransferase mechanism. It appears that 3-OST-3 recognizes the carboxyl and sulfo groups of the iduronic acid (U4) at the nonreducing end of the glucosamine unit (G3) being sulfonated. The amino acid residues involved in binding to these structural motifs are essential for the activity of 3-OST-3. This conclusion is consistent with those from previous biochemical studies of 3-OSTs; 3-OST-1 sulfates a glucosamine linked to a glucuronic acid unit at the nonreducing end, whereas 3-OST-3 sulfates the glucosamine unit that is linked to a 2-O-sulfo iduronic acid unit at the nonreducing end ( Fig.  1A) (9,11,17,25). It is very important to note that the 3-OST-3 modification site is flanked by two iduronic acid units in a skew boat like conformation. In contrast, the conformation of the I2 unit is in the chair form in the fibroblast growth factor/tetrasaccharide complex (23). It is likely that the conformation of the flanking iduronic acid residues contribute to substrate recognition by 3-OST-3.
The 3-O-sulfonation of HS has been directly linked to the activities for regulating blood coagulation and for promoting infection by HSV-1. Consequently elucidation of the general mechanism and discovery of the atomic level requirements for specificity provide important insight for understanding the biosynthesis of biologically active forms of HS.  β-Strands and α-helices are colored purple and cyan, respectively. This figure was created using Molscript and Raster3d (33,34). Structural alignment of 3-OST-3 with 3-OST-1 is shown besides the ribbon diagram. Cyan barrels represent residues in helices, and purple represents regions in β-strands. Residues that are structurally dissimilar are shaded with a yellow background. Residues that form interactions with the tetrasaccharide are colored green, and the cysteines that form a disulfide are colored orange. Red asterisks represent the three residues that may form a catalytic triad in the 3-OSTs. Residues that are within hydrogen bonding distance or that line the binding pocket are pictured. Possible hydrogen bonds between the protein and the PAP are displayed as black dashed lines. This figure was created using Molscript and Raster3d (33,34).    Residues in allowed (>99.8%) regions (%) 100 100 a R sym = Σ (I i -⟨I⟩/Σ(I j ) where I j is the intensity of the ith observation and ⟨I⟩ is the mean intensity of the reflection.
b R cryst = Σ||F o -F c calculated from working data set.
c R free was calculated from 5% of data randomly chosen not to be included in refinement.
d The Ramachandran results were determined by MolProbity (35).   c The expression level of the proteins was determined by the intensity of the Coomassie Blue-stained protein band migrated at 30 kDa on SDS-PAGE.