Crystal Structure of the Sulfotransferase Domain of Human Heparan Sulfate N-Deacetylase/ N-Sulfotransferase 1*

Heparan sulfateN-deacetylase/N-sulfotransferase (HSNST) catalyzes the first and obligatory step in the biosynthesis of heparan sulfates and heparin. The crystal structure of the sulfotransferase domain (NST1) of human HSNST-1 has been determined at 2.3-Å resolution in a binary complex with 3′-phosphoadenosine 5′-phosphate (PAP). NST1 is approximately spherical with an open cleft, and consists of a single α/β fold with a central five-stranded parallel β-sheet and a three-stranded anti-parallel β-sheet bearing an interstrand disulfide bond. The structural regions α1, α6, β1, β7, 5′-phosphosulfate binding loop (between β1 and α1), and a random coil (between β8 and α13) constitute the PAP binding site of NST1. The α6 and random coil (between β2 and α2), which form an open cleft near the 5′-phosphate of the PAP molecule, may provide interactions for substrate binding. The conserved residue Lys-614 is in position to form a hydrogen bond with the bridge oxygen of the 5′-phosphate.


Heparan sulfate N-deacetylase/N-sulfotransferase (HSNST) catalyzes the first and obligatory step in the biosynthesis of heparan sulfates and heparin. The crystal structure of the sulfotransferase domain (NST1) of human HSNST-1 has been determined at 2.3-Å resolution in a binary complex with 3-phosphoadenosine 5phosphate (PAP). NST1 is approximately spherical with an open cleft, and consists of a single
Heparan sulfate chains are ubiquitous as proteoglycans on cell surfaces and in the extracellular matrix. They have been increasingly implicated in various biological processes including cell growth, cell differentiation, blood coagulation, and viral and bacterial infections (1,2). Reduced biosynthesis of heparan sulfates, for instance, results in defective WINGLESS/WNT signaling in Drosophila (3,4). Mice lacking the heparan sulfate 2-O-sulfotransferase gene die neonatally from defective kidney development (5). The recent crystal structures of the fibroblast growth factor-heparin and antithrombin-heparin complexes have shown specific protein-sulfate interactions (6,7). Modifi-cation by sulfation thus, can alter functional specificity and diversity of heparans and heparins.
Heparan sulfation is catalyzed by a group of the Golgi-membrane enzymes called heparan sulfate sulfotransferases. The superfamily includes also a large number of cytosolic sulfotransferases that sulfate low molecular weight substrates such as steroids, bioamines, pharmaceutical drugs, and environmental chemicals. The membrane and cytosolic sulfotransferases share little overall sequence similarity, whereas all sulfotransferases use 3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS) 1 as the ubiquitous sulfate donor. The crystal structure of the estrogen sulfotransferase (EST)-PAP-estradiol (E2) complex has revealed the structural motifs for the 5Ј-and 3Ј-phosphate binding of PAP (8,9). It remains to be structurally determined whether these motifs are also conserved in heparan sulfate sulfotransferases. Multiple sequence alignments have suggested that these motifs may be conserved (9). The reaction mechanisms and specific substrate binding that lead to the diverse heparan sulfations are poorly understood.
The bi-functional enzyme HSNST sequentially deacetylates and sulfates the amino group of the disaccharide glucuronic acid-N-acetylglucosamine (GlcA-GlcNAc) unit of heparan sulfate (10,11). Amino acid sequence alignment of the human HSNST1 with EST identified the N-sulfotransferase domain (NST1) (12). Human NST1 is similar to the corresponding domain of mouse HSNST reported by Berninsone and Hirschberg (13). Subsequently, site-directed mutagenesis has shown that Lys-614 is a critical residue for NST1 catalysis (12). Using NST1 crystals grown as previously reported (12), we now describe herein the crystallographic structure NST1. This structure displays the conserved nature of the structure of the PAP-binding site and identifies possible catalytic residues. In addition, a substrate binding site is suggested from the structure.

EXPERIMENTAL PROCEDURES
Protein Expression, Purification, Crystallization, and Enzyme Assay-Selenomethionyl NST1, using a pGEX-4T3-NST1 plasmid, was expressed in the methionine auxotrophic Escherichia coli strain B834 (DE3) with a defined minimal essential medium (without methionine) containing 50 mg of selenomethionine per 1 liter of culture. The NST1 was then purified, and crystals (P2 1 2 1 2 or P2 1 ) were grown under the same conditions as described previously (14). Heparan sulfate sulfotransferase activity of NST1 was also measured according to the previously described procedure (12).
Crystallographic Data Collection and Processing-Two MAD data sets of selenomethionyl NST1 were collected at Ϫ180°C from two separate single crystals (both P2 1 2 1 2) on a MAR detector at beamline X9B of the NSLS, Brookhaven National Laboratory. Three wavelengths were selected from the fluorescence spectra: f1 (0.97163 Å: remote), f2 (0.97907 Å: peak), and f3 (0.97940 Å: edge) (Table I). Native data of an NST1 crystal (P2 1 ) were collected at Ϫ180°C on an R-axis IV with an RU300 rotating anode generator.
Structure Determination and Refinement-All data were processed using SCALEPACK and DENZO (14). Because of heavy ice rings in data set 1 and data set 2 being weak, F(A)s were calculated with CCP4 (15) using data between 20 and 4 Å of data set 1. Positions for four of the six selenium atoms were determined using SHELX96 (16). Data set 1 was * 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.
The reprocessed to eliminate all reflection near the ice rings between 3.95 and 3.1 Å. Subsequently, the reprocessed data set 1 was merged with data set 2 to obtain a complete data set to 2.85-Å resolution. SHARP (17) was then used for refinement of the selenium sites. Solvent flattening and histogram matching were carried out using DM and Solomon from CCP4 (15). In the model building process using O (18), SigmaA maps were generated by combining the phases from polyalanine fragments with the MAD phases (15). After multiple cycles of positional, torsion angle, and b factor refinements using X-PLOR (19), the R-factor and R free were 23.8 and 31.9%, respectively. Because some of the loop regions still lacked interpretable density, molecular replacement was employed to determine phases for the data from the P2 1 crystal. Multiple cycles of manual rebuilding and refinement using the P2 1 data at 2.3 Å reduced the R and R free factors to 20.9 and 25.8%, respectively. The stereochemistry of the refined model was verified using PROCHECK (15). Data collection and refinement statistics are summarized in Table I. The coordinates have been deposited in the Protein Data Bank with code 1NST.

RESULTS AND DISCUSSIONS
The overall structure of NST1 is roughly spherical with an open cleft (Fig. 1A). This structure is composed of a fivestranded parallel ␤-sheet (␤1, ␤2, ␤3, ␤4, and ␤5) with ␣ helices on both sides of the ␤-sheet (Fig. 1B). This fold is similar to EST (the 1.9-Å r.m.s. deviation for 97 C␣s in ␤1, ␤3, ␤4, ␤5, ␣1, ␣6, ␣11, ␣12, and ␣13) as well as to the nucleotide binding motif observed in nucleotide kinases (20). The loop between ␤1 and ␣1 adopts the same PSB-loop configuration as the 5Ј-phosphate binding site of PAPS. A cavity formed between the PSB-loop, and ␣6 defines the PAP binding site. Three ␤ strands (␤6, ␤7, and ␤8) near the C terminus form an anti-parallel ␤-sheet with a single disulfide bond between ␤7 and ␤8. An open cleft that runs perpendicular to the PAP binding cavity is large enough to contain a hexasaccharide chain. The ␣6 and random coil between ␤2 and ␣2 constitute the cleft near the 5Ј-phosphate of PAP and thus may constitute part of the substrate binding site.
The secondary structural elements that comprise the PAP binding site in NST1 and residues forming specific interactions to the PAP molecule are depicted (Fig. 2), respectively. The PSB-loop (residues 612-617) and ␣1 of NST1 constitute the 5ЈPSB motif and provide the major binding sites for the 5Јphosphate of the PAP molecule. Backbone amide nitrogens from PSB-loop residues 614 -618 are all within hydrogen bonding distance of the 5Ј-phosphate. The side-chain N of Lys-614 and the O␥s of both Thr-617 and Thr-618 are also hydrogenbonded to the 5Ј-phosphate. ␣6 and ␤4 are the key elements of the 3ЈPB motif, and the O␥ of Ser-712 from this helix forms a hydrogen bond to the 3Ј-phosphate of the PAP molecule. The PAP molecule in NST1 is bound in the same orientation (relative to the PSB-loop) as seen in EST. The PAP binding site, found in the EST structure determined previously, is conserved. The r.m.s. deviation (with EST) for 47 C␣s in ␣1, ␣6, ␤1, ␤4, and PSB-loop is 1.16 Å.
The anti-parallel ␤-sheet (␤6, ␤7, and ␤8) and the following random coil provide the remaining interactions for the PAP binding site (Fig. 2). These interactions reveal diversity in the binding site of NST1. The side-chains of Lys-833 and Tyr-837 from this random coil are within hydrogen bonding distance to two oxygen atoms of the 5Ј-phosphate and the oxygen atom of the 3Ј-phosphate, respectively. Besides these side-chain interactions, the backbone nitrogens of Gly-834 and Arg-835 are also within hydrogen bonding distance of a 3Ј-phosphate oxygen of the PAP molecule. The adenine ring from the PAP molecule is in position to form a parallel ring stacking interaction with Phe-816 of ␤7. Moreover, the backbone oxygen of Trp-817 is within hydrogen bonding distance to the N-6 of the adenine. The interactions of these residues with the PAP molecule are unique features in NST1 that are not present in the crystal structure of the EST-PAP complex (8).
Lys-614 of NST1 is known to be conserved in other heparan sulfate sulfotransferases as well as in all cytosolic sulfotransferase (9,12). Although this residue plays a critical role in NST1 activity (12), the structural basis of its role in catalysis has remained unresolved. The crystal structure of the EST-PAP-vanadate complex (Fig. 3) has recently been solved and has provided a possible transition state template for the sulfuryl transfer reaction (21). Superimposition of NST1 on this structure indicates that the side-chains of Lys-614 (in NST1) and Lys-48 (in EST) exhibit a similar orientation and conformation (Fig. 3). N of Lys-614 is found to be directly coordinated to an oxygen of the PAP molecule in NST1. This oxygen is also coordinated to Lys-48 (N) in EST and is implicated as the bridge oxygen of the leaving phosphate group of PAP (21). Moreover, the mutation of Lys-614 to Arg gives a variant with a significant level of NST1 activity (63 Ϯ 7.0 and 9.4 Ϯ 2.4 nmol of sulfate/min/mg of protein in the wild-type and K614R mutant, respectively), whereas the K614A mutation abolishes activity completely (12). These structural and mutational data suggest that Lys-614 may act as a possible proton donor in catalysis, similar to Lys-48 in EST (21). Lys-833 of NST1 is also coordinated with the bridge oxygen (Fig. 3). Lys-833 is, in fact, conserved not only in Caenorhabditis elegans HSNST but also in human heparan sulfate 3-O-sulfotransferase (see the sequence alignments in Shworak et al. (22)). Thus, Lys-833 and its counterparts may play a significant role in catalysis.
In sharp contrast to the hydrophobic pocket of estrogen binding site in EST, the putative substrate binding site of NST1 appears to be a large open cleft with a hydrophilic surface, with a random coil (residues 640 -647, approximately 12 Å in length) and ␣6 forming the center of the cleft near the 5Јphosphate of the PAP molecule. This amphipathic random coil positions negatively charged side-chains (Glu-641, Glu-642, Gln-644, and Asn-647) toward the center, whereas the hydrophobic side-chains (Ile-643, Phe-645, and Phe-656) are buried in the hydrophobic core of NST1. The side-chains of residues (Trp-713, His-716, Gln-717, and His-720) in ␣6 constitute the opposing face of the cleft. The center of this cleft (approximate dimensions: 12 Å in length, 8 Å in width, and 8 Å in depth) is large enough to accommodate a trisaccharide unit of polysaccharide chain. Further studies, such as the determination of the complex structure of NST1 complexed with polysaccharide, are needed to conclude whether this center portion of the cleft is, in fact, the substrate binding site. CONCLUSION The striking similarities between the PAP binding orientation in NST1 and EST provide structural evidence that the Golgi membrane and cytosolic enzymes belong to the same family of enzymes. The similar topology and function of Lys-614 to Lys-48 of EST suggest a common reaction mechanism in all sulfotransferases. Lys-833 may be an additional catalytic residue not present in the cytosolic enzymes. The NST1 structure provides an excellent model for investigating the substrate specificity of heparan sulfate sulfotransferases so that we may better understand sulfation at specific positions of glucuronic acid-N-acetylglucosamine.
Acknowledgments-We thank Rick Moore for excellent technical assistance, Dr. M. Duffel for supply of PAP, Dr. Z. Dauter for assistance with data collection at NSLS, and Drs. T. Hall and I. Tanaka for helping with MAD phasing. Our sincere appreciation is also acknowledged to Drs. Lee Pedersen, W. Beard, and T. Hall for comments on this manuscript. FIG. 1. Overall structure and topology of NST1. A, ribbon representation of the NST1 structure in complex with PAP. Helices are in yellow, ␤-strands in green, random coil in blue, disulfide bond in light blue, and PAP molecule in red. B, topological representation of the polypeptide fold of NST1. The residue numbers define the secondary structural elements. The N-terminal 18 residues (587-600) and 4 residues from 665 to 669 are disordered in the NST1 crystal. These figures were prepared with SETOR (23).

FIG. 2. Ribbon representation of the PAP binding site in NST1.
The side-chains that can interact with PAP are also included. The disulfide bond between Cys-818 and Cys-828 is shown in blue. The sidechains of key residues involved in the binding of PAP are also shown in this picture prepared using SETOR (23).