Structure and Function of HNK-1 Sulfotransferase

HNK-1 glycan, sulfo→3GlcAβ1→3Galβ1→4GlcNAc→R, is uniquely enriched in neural cells and natural killer cells and is thought to play important roles in cell-cell interaction. HNK-1 glycan synthesis is dependent on HNK-1 sulfotransferase (HNK-1ST), and cDNAs encoding human and rat HNK-1ST have been recently cloned. HNK-1ST belongs to the sulfotransferase gene family, which shares two homologous sequences in their catalytic domains. In the present study, we have individually mutated amino acid residues in these conserved sequences and determined how such mutations affect the binding to the donor substrate, adenosine 3′-phosphate 5′-phosphosulfate, and an acceptor. Mutations of Lys128, Arg189, Asp190, Pro191, and Ser197 to Ala all abolished the enzymatic activity. When Lys128 and Asp190 were conservatively mutated to Arg and Glu, respectively, however, the mutated enzymes still maintained residual activity, and both mutant enzymes still bound to adenosine 3′,5′-diphosphate-agarose. K128R and D190E mutant enzymes, on the other hand, exhibited reduced affinity to the acceptor as demonstrated by kinetic studies. These findings, together with those on the crystal structure of estrogen sulfotransferase and heparan sulfateN-deacetylase/sulfotransferase, suggest that Lys128 may be close to the 3-hydroxyl group of β-glucuronic acid in a HNK-1 acceptor. In contrast, the effect by mutation at Asp190 may be due to conformational change because this amino acid and Pro191 reside in a transition of the secondary structure of the enzyme. These results indicate that conserved amino acid residues in HNK-1ST play roles in maintaining a functional conformation and are directly involved in binding to donor and acceptor substrates.

To elucidate the roles of HNK-1 glycan in molecular details, we and others have cloned a sulfotransferase HNK-1ST that forms HNK-1 glycan on neural cell adhesion molecule and glycolipids (18,19). This was possible because ␤1,3-glucuronosyltransferase that forms a HNK-1 precursor structure was cloned before (20). When we compared the amino acid sequence of HNK-1ST to those of Golgi-associated sulfotransferases cloned before (21)(22)(23)(24), we noticed that amino acid sequence motifs, including the sequence RDP, are shared by those enzymes (18). Although we did not initially detect its homology with cytosolic sulfotransferases, subsequent studies revealed that there are two regions in the catalytic domain that are shared by all sulfotransferases cloned so far (25,26). This homology was also proposed because the crystal structure of a mouse estrogen sulfotransferase became available (27), and the amino acid residues involved in the interaction with the 5Јphosphosulfate or 3Ј-phosphate group of PAPS were identified (Fig. 1). The studies thus revealed that a lysine or arginine residue is shared by all sulfotransferases cloned to date in the 5Ј-phosphosulfate binding (5Ј-PSB) 1 site and an arginine residue is shared by almost all sulfotransferases at the 3Ј-phosphate binding (3Ј-PB) motif (25)(26)(27). Moreover, the entirely conserved serine residue at the 3Ј-PB loop apparently provides a hydrogen bond with a hydroxyl group of 3Ј-phosphate (27). These studies, however, did not reveal how those or other amino acids are involved in the interaction with an acceptor substrate.
In the present study, we individually mutated five different amino acid residues in the 5Ј-PSB and 3Ј-PB sites of HNK-1ST by site-directed mutagenesis. We first assayed those mutant enzymes for transfer of [ 35  catalytic properties were obtained. Those mutants were also assayed for their binding to a PAP analogue of PAPS. Finally, these results were interpreted in relation to a three-dimensional structure inferred from the crystal structures determined for estrogen sulfotransferase and the human heparan sulfate N-deacetylase/sulfotransferase, HSNST (27,28). Our results strongly suggest that these two regions of the catalytic domain, in particular those amino acids in 5Ј-PSB site, are involved in the interaction with both acceptor and donor substrates.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis-The cDNA encoding HNK-1ST was cloned as described previously (18). A cDNA encoding a catalytic domain of HNK-1ST fused with an IgG-binding domain of protein A was constructed as described before (18). The plasmid, pcDNAI-A⅐HNK-1ST (18), encoding a soluble protein A-HNK-1ST fusion protein was digested with BamHI and SpeI to release a 1.6-kilobase pair cDNA fragment. pBluescript II KS ϩ vector (Stratagene, La Jolla, CA) was digested with the same restriction enzymes and ligated to the HNK-1ST fragment to give plasmid pBSII KS ϩ -HNK-1ST. Site-directed mutagenesis was performed on this plasmid using a Chameleon TM double-stranded, sitedirected mutagenesis kit (Stratagene) (29,30). Phosphorylated oligonucleotides covering conservative or nonconservative amino acid replacement were synthesized and purified by denaturing polyacrylamide gel electrophoresis before use. Following site-directed mutagenesis of the required amino acid residue on pBSII KS ϩ -HNK-1ST, a 0.53-kilobase pair fragment flanked by Bsu36I and BamHI and containing the specific codon replacement was sequenced to confirm the correct sequence. This 0.53-kilobase pair DNA fragment was released by digestion with Bsu36I and BamHI and was used to replace the corresponding sequence in the wild type DNA sequence in pcDNAI-A⅐HNK-1ST.
Production of HNK-1ST-The plasmid pcDNAI-A⅐HNK-1ST containing the appropriate mutation was transfected into COS-1 cells using LipofectAMINE ® Plus reagents (Life Technologies, Inc.) (18). Following 24 h of growth in a complete medium at 37°C, the medium was replaced with serum-free OptiMEM solution (Life Technologies, Inc.). After a 48-h incubation at 37°C, the culture medium containing the fusion protein was collected and filtered through a 0.22-m membrane. The fusion protein was isolated by adsorption to IgG-Sepharose 6FF beads (Amersham Pharmacia Biotech) (31). After washing, the fusion protein-containing beads were resuspended in 100 mM Tris-HCl buffer, pH 7, at 50% suspension volume. This was used as the enzyme source. Appropriate controls (both wild type and mock-transfected reactions) were performed.
HNK-1ST and the mutant enzymes released were separated by SDSpolyacrylamide gel electrophoresis, and their amount was estimated by Western blot analysis using peroxidase-conjugated rabbit anti-goat IgG antibodies (30) purchased from ICN.
Sulfotransferase Assay-The acceptor, GlcA␤133Gal␤134GlcNAc-␤13octyl was synthesized as described previously (18,32). Appropriate amounts of the acceptor and the donor, adenosine 3Ј-phosphate 5Јphospho[ 35 S]sulfate, were mixed and dried down under vacuum. The material was then redissolved in 40 l of a solution containing 100 mM Tris-HCl, pH 7, 0.1% Triton X-100, 2.5 mM ATP, and 10 mM MnCl 2 . To this, 10 l of the 50% suspension of the enzyme solution, prepared as described above, was added. The mixture was incubated at 37°C for 1 h with constant, gentle agitation. The reaction was stopped by adding 450 l of 250 mM ammonium formate, pH 4. The labeled products were applied to C18 reverse-phase chromatography columns (Alltech) and eluted with 70% methanol. A portion of the eluted product was subjected to scintillation counting. Relative activity was measured and standardized against the values obtained for the wild type after subtraction of the background activity for the mock samples. The K m for wild type HNK-1ST and mutant enzymes exhibiting detectable enzymatic activity was determined using a Lineweaver-Burk plot at various concentrations of the acceptor (2.5 M to 2000 M) and donor (1 M to 2000 M).
Binding of HNK-1ST to Immobilized Adenosine Diphosphate-Following the transfection protocol as described above, the culture medium (20 ml) from the transfected cells was concentrated approximately 20-fold using Centriprep-10 spin columns (Amicon). The concentrated medium containing the enzyme (100 l) was mixed with 900 l of 100 mM Tris-HCl, pH 7, containing 5 mM MnCl 2 , applied to 1 ml of adenosine-3Ј, 5Ј-diphosphate-agarose (Sigma), washed, and equilibrated in 100 mM Tris-HCl, pH 7, as described previously (33). After washing the beads with 4 ml of 100 mM Tris-HCl, pH 7, the bound enzyme was eluted with a NaCl linear gradient (0 -500 mM) with increasing NaCl at 25 mM/ml. The eluted sample (200 l) from each fraction (1 ml) was coated onto a 96-well microtiter plate and incubated at 4°C overnight. After blocking the wells with 1% bovine serum albumin for 2 h at room temperature, a 1:1000 dilution of rabbit anti-goat IgG-horseradish peroxidase conjugate (200 l) was added and incubated for 1 h at room temperature. Following three washes with Tris-buffered saline containing 0.2% Tween 80, 200 l of a chromogenic substrate (0.1 mg of o-phenylenediamine/ml of 50 mM phosphate-citrate buffer, pH 5, containing 0.003% H 2 O 2 ) was added to each well. After incubation at room temperature to allow sufficient color to develop, the reaction was stopped with 50 l of 2 N sulfuric acid. The absorbance at 490 nm was read using a microtiter plate reader (Bio-Rad).
Modeling of HNK-1ST Active Sites-The active site of HNK-1ST was modeled using the estrogen sulfotransferase and the HSNST coordinates (27,28). The program "O" was utilized to make the changes to the sequence using the program Molscript (34).

Isolation of Mutant HNK-1ST and Its Assay-Previously
we have shown that HNK-1ST can be expressed as a soluble protein fused with protein A (18). Such a soluble chimeric protein has a significant activity and is judged to reflect approximately the catalytic activity of the enzyme present in Golgi (18). In the present study, we took advantage of this property, and a soluble chimeric HNK-1ST was used throughout the studies.
As shown in Fig. 2, the wild type enzyme exhibits an activity profile typical for a transferase, and substrate inhibition or acceptor inhibition was not observed. We then individually mutated amino acid residues Lys 128 in the 5Ј-PSB site and Arg 189 , Asp 190 , Pro 191 , and Ser 197 in the 3Ј-PB motif (see Fig. 1 also). After these mutated enzymes and the wild type enzyme were expressed, the expression of each enzyme was estimated by Western blot analysis using rabbit IgG as a probe because the protein A portion of the chimeric protein can bind to rabbit IgG (Fig. 3). After adjusting the amount of the enzyme, the catalytic activity of each enzyme preparation was assayed using standard incubation conditions, as described under "Experimental Procedures." As shown in Fig. 4, almost no activity remained when Lys 128 , Arg 189 , Asp 190 , Pro 191 , and Ser 197 were individually mutated to an alanine residue. In contrast, the conservative mutation of K128R and D190E maintained 15 and 60% of the activity compared with that of wild type. Mutation of R189K or P191G, on the other hand, was not tolerated, and the mutant enzymes completely lost the activity.
Amino Acid Mutations Affect Binding to Donor-To determine how these mutations cause a change in the binding to PAPS, the wild type HNK-1ST and each mutated enzyme were individually applied to a column of PAP-agarose, and the bound enzyme was eluted by a linear gradient of NaCl. The eluted enzymes were immunochemically detected using rabbit IgGperoxidase conjugate. As shown in Fig. 5A, the wild type enzyme bound to the column in a biphasic way, and the majority of the bound enzyme was eluted after the gradient elution of NaCl started. Similar results were reported in a flavonol sulfotransferase (33). In contrast, K128A mutant barely bound to PAP-agarose, whereas the K128R mutant enzyme bound to the same column as much as the wild type enzyme.
In a similar assay, the R189K, R189A, S197T, and S197A mutant enzymes did not bind to the PAP-agarose column (Fig.  5, B and D), whereas D190E bound well to the same column (Fig. 5C). These results demonstrated that K128R and D190E mutations did not significantly affect the binding to PAPS, whereas the conservative mutation of R189K and S197T dramatically reduced the binding to PAPS.
Mutation on Lys 128 and Asp 190 Affects Binding to Acceptor Substrate-The above results indicate that the conservative mutation at Lys 128 and Asp 190 did not significantly affect the binding to PAPS but reduced the catalytic activity. To deter- mine whether such a decrease in activity was due to impaired binding to the acceptor, the enzyme activity was determined using varying concentrations of donor substrate PAPS or the acceptor, GlcA␤133Gal␤134GlcNAc␤13octyl.
The results shown in Fig. 2, A and B, were replotted for the Lineweaver-Burk plot as seen in Fig. 2, C and D. The K m for the donor substrate PAPS is almost the same between the wild type (66 M) and the D190E mutant (70 M) and is even better for the K128R mutant (12 M) than for the wild type enzyme. In contrast, the K m value for the acceptor was 316 M for the wild type enzyme and increased 5.6-fold for K128R mutant (1776 M) and 2.4-fold for D190E mutant (744 M), respectively. These results suggest that Lys 128 and Asp 190 may be involved in the binding to the acceptor or very close to the acceptor binding site (see also under "Discussion").

DISCUSSION
Comparison of sulfotransferases cloned to date revealed that there are two amino acid sequences that are highly conserved among different sulfotransferases (Fig. 1). In the present study, we first demonstrated that mutations of Arg 189 , Pro 191 , and Ser 197 inactivate the enzyme, indicating that these amino acid residues are important for either binding to the donor PAPS or maintaining the conformation of the enzyme for transfer of the sulfate group to the acceptor molecule. All mutations of these amino acids, including conservative replacement, resulted in the loss of PAPS binding in these mutant enzymes (Fig. 5). These results confirm the previous conclusions that 5Ј-phosphosulfate and 3Ј-phosphate binding sites are critical for the binding to PAPS (25)(26)(27)(28).
More importantly, the present study also revealed that the mutations at Lys 128 and Asp 190 can be tolerated well for the binding to PAPS, and such a binding became even better after mutation of Lys 128 to Arg. The excellent binding of these mu-tant enzymes to PAPS can be seen in both affinity chromatography using PAP-agarose (Fig. 4) and in kinetic studies (Fig. 2). In contrast, the same kinetic studies showed that these mutant have lower affinity to the acceptor, GlcA␤133Gal␤134GlcNA-c␤13octyl (Fig. 2). These results suggest that these two amino acids may be more intimately involved in the binding of an acceptor than in the binding of a donor substrate.
No studies are available on the mutation of Asn 131 in the estrogen sulfotransferase, which corresponds to Asp 190 in HNK-1ST, although crystal structural studies showed that Arg 130 , which is next to Asn 131 , interacts with the 3Ј-phosphate group of PAP (27). Our mutation experiments demonstrated that the mutation of Arg 189 to either Lys or Ala resulted in the complete loss of the enzymatic activity (Fig. 2). The loss of the activity was accompanied by the loss in the binding to PAP (Fig. 5).
The present study also demonstrated that mutation of Asp 190 to Glu did not inactivate the enzyme, whereas mutation of Arg 189 to Lys completely inactivated the enzyme. These results may suggest that Arg 189 is critical in the binding to a 3Јphosphate group of PAPS, whereas Asp 190 is involved in the binding to an acceptor. However, these results need to be evaluated in relation to the three-dimensional structure of HNK-1ST. The crystal structures of the estrogen sulfotransferase and HSNST demonstrated that the structure of the 5Ј-PSB and 3Ј-PB sites is well conserved in these two different enzymes (27,28). These results, together with the results obtained in the present study, allow us to propose a possible three-dimensional structural model of HNK-1ST, as shown in Fig. 6.
Concerning the 3Ј-PB site, Arg 189 and Ser 197 can interact directly with the 3Ј-phosphate group of PAP in the model, thus, mutations of Arg 189 and Ser 197 can profoundly alter PAPS binding, affecting HNK-1ST activity. Asp 190 and Pro 191 are not in a position to interact directly with the PAP molecule. They reside in a transition of secondary structure from a ␤-sheet to an ␣-helix, making for a very tight turn of about 90° (Fig. 6). FIG. 5. Binding of wild type and mutant HNK-1ST enzymes to PAP-agarose. The chimeric enzymes fused with protein A were concentrated from culture medium and applied to a column of adenosine 3Ј,5Ј-diphosphate-agarose. After washing the column with 100 mM Tris-HCl, pH 7, the bound enzyme was eluted with a linear gradient of NaCl (0 -500 mM) (OE) as described under "Experimental Procedures." The eluted enzyme was detected by incubating with rabbit anti-goat IgGhorseradish peroxidase conjugate. The elution profiles are shown for A, wild type (E), K128R (Ⅺ), K128A (f); B, R189K (छ), R189A (ࡗ); C, D190E (‚); D, S197T (ƒ) and S197A () mutant enzymes. Through its interactions with backbone N atoms of residues 192 and 193 in the following ␣-helix, Asp 190 may help to stabilize this turn. Thus, mutation of Asp 190 may alter the core structure of the PAP binding site, affecting indirectly HNK-1ST activity. Pro 191 may also be involved in the stabilization because its mutation abolished HNK-1ST activity. These residues (Arg 189 , Ser 197 , Asp 190 , and Pro 191 ) are located more than 20 Å from the nearest atom in the glucuronic acid molecule and the PAP molecule than the glucuronic acid molecule. They are not in a position to interact directly with the acceptor molecule for HNK-1ST. It would be more of an indirect effect, if any mutations of these residues affected the binding of the acceptor. The R(D/N)P sequence may thus be conserved to maintain the structure of the PAPS binding site.
Lys 128 (or Arg in some sulfotransferases) is conserved in the 5Ј-PSB site of all known sulfotransferases (Fig. 1). It can form hydrogen bonds with the 5Ј-phosphate group of PAP and is involved in catalysis. Mutations of this residue inactivated HNK-1ST, except that K128R still maintains some of the original activity. Lys 128 is 9 Å away from the 3-hydroxyl group of the glucuronic acid molecule and is unlikely to interact directly with the acceptor of HNK-1ST. In the mouse estrogen sulfotransferase, the K48R mutation in the 5Ј-PSB site did not change K m of the enzyme for either estrogen or PAP (35). In contrast, the K128R mutation of HNK-1ST increased K m toward glucuronic acid in the acceptor molecule more than 5-fold. In the model, the end of the side chain of Arg (at the position of Lys 128 ) can be moved to within 3.2 Å of the 3-hydroxyl group of the glucuronic acid molecule (Fig. 6). A bulkier nonplaner substrate of HNK-1ST may actually come closer to the position of the mutated Arg. Thus the K128R mutation could have an effect on the binding of the acceptor in HNK-1ST.
In the mouse estrogen sulfotransferase, Lys 48 corresponds to Lys 128 in HNK-1ST. In contrast to the results obtained in the present study, the mutation of Lys 48 to Arg did not change the K m toward estrogen or PAPS in a previous study (34). The importance of Lys 48 was, however, demonstrated because its mutation to methionine completely inactivated the enzyme in the same study. These results indicate that Lys 128 in the HNK-1ST is closer to an acceptor than Lys 48 in the estrogen sulfotransferase. This difference is probably due to the difference in acceptor structures (Fig. 6).
The present study demonstrated that HNK-1ST and estrogen sulfotransferase share many of the structural requirements for the enzymatic activity. The study, on the other hand, also indicates that acceptor binding sites of different enzymes are distinct from one another. Further studies on the threedimensional structure of HNK-1ST will be of significance to determine how the acceptor is recognized by the enzyme and provide critical information on the three-dimensional structure of the acceptor oligosaccharide.