The Heparin/Heparan Sulfate 2-O-Sulfatase from Flavobacterium heparinum

In the previous paper (Myette, J. R., Shriver, Z., Claycamp, C., McLean, M. W., Venkataraman, G., and Sasisekharan, R. (2003) J. Biol. Chem. 278, 12157–12166), we described the molecular cloning, recombinant expression, and preliminary biochemical characterization of the heparin/heparan sulfate 2-O-sulfatase fromFlavobacterium heparinum. In this paper, we extend our structure-function investigation of the 2-O-sulfatase. First, we have constructed a homology-based structural model of the enzyme active site, using as a framework the available crystallographic data for three highly related arylsulfatases. In this model, we have identified important structural parameters within the enzyme active site relevant to enzyme function, especially as they relate to its substrate specificity. By docking various disaccharide substrates, we identified potential structural determinants present within these substrates that would complement this unique active site architecture. These determinants included the position and number of sulfates present on the glucosamine, oligosaccharide chain length, the presence of a Δ4,5-unsaturated double bond, and the exolytic versusendolytic potential of the enzyme. The predictions made from our model provided a structural basis of substrate specificity originally interpreted from the biochemical and kinetic data. Our modeling approach was further complemented experimentally using peptide mapping in tandem with mass spectrometry and site-directed mutagenesis to physically demonstrate the presence of a covalently modified cysteine (formylglycine) within the active site. This combinatorial approach of structure modeling and biochemical studies provides insight into the molecular basis of enzyme function.

It is a highly variable sulfation pattern, in particular, which ascribes to each GAG chain a unique structural signature. In turn, this signature dictates specific GAG-protein interactions underlying critical biological processes related to cell and tissue function. Given this critical structure-function relationship of GAG sulfation, enzymes that can hydrolyze these sulfates in a structurally specific manner are important tools for the determination of GAG fine structure to better ascertain these structure-function relationships. In the previous paper (21), we described the cloning, recombinant expression, and biochemical characterization of one such sulfatase, the 2-O-sulfatase from Flavobacterium heparinum.
As members of a large enzyme family, the sulfatases hydrolyze a wide array of sulfate esters (for a review, see Refs. 3 and 4). Their respective substrates include sulfated complex carbohydrates such as the glycosaminoglycans (GAGs), steroids, sphingolipids, xenobiotic compounds, and amino acids such as tyrosine. Additionally, many of these enzymes are able to hydrolyze in vitro smaller synthetic substrates (e.g. 4-nitrophenyl sulfate and catechol sulfate). It is for this reason that these enzymes are often generically described as "arylsulfatases" (even when their preferred in vivo substrate is ill defined). Despite their disparate substrate specificities, the members of this enzyme family share both considerable structural homology and a common catalytic mechanism with one another (5). Each sulfatase possesses a signature catalytic domain toward its amino terminus, which is readily identified by the consensus sequence (C/S)XPXRXXXX(S/T)G. The conserved cysteine (or less commonly serine) within this sulfatase motif is of particular functional importance, since it is covalently modified to an L-C␣-formylglycine (L-2-amino-3-oxopropionic acid) (4,6,7). The ubiquitous importance of this chemical modification was first functionally identified by its relationship to the etiology of multiple sulfatase deficiency, a genetically recessive disorder in which there is a complete loss of sulfatase activity due to a lack of this critical aldehyde (FGly) within the active site of all expressed sulfatases (8).
The crystal structures of two human lysosomal sulfatases, arylsulfatase A (cerebroside-3-sulfate 3-sulfohydrolase) (9,10) and arylsulfatase B (N-acetylgalactosamine-4-sulfate 4-sulfohydrolase) (11), and a bacterial arylsulfatase from Pseudomonas aeruginosa (12) have been solved. These three sulfatases share an identical alkaline phosphatase-like structural fold (according to the Structural Classification of Proteins Database) 2 composed of a series of mixed parallel and antiparallel ␤-strands flanked by long and short ␣-helices on either side (9 -12). In addition to their common structural fold, these sulfatase structures also possess a high degree of homology within their respective active sites, especially in the region localized around the modified cysteine (FGly). Taken together, these crystal structures present a clear and consistent description of conserved active site residues and their potential catalytic role in sulfate ester hydrolysis. At the same time, this strong structural homology is somewhat surprising, considering that at least two of these sulfatases act on notably different substrates (e.g. sulfated sphingolipid versus sulfated GAG). The structural basis for substrate specificity, therefore, remains to be determined. The question of this relationship of enzyme active site structure to substrate specificity would appear to be especially relevant to GAGs where multiple sulfate esters are present.
Using the strong structural and functional similarities that have been observed among the members of the sulfatase family of enzymes, we extend our characterization of the recombinant 2-O-sulfatase in this paper to include a structure-function description of the enzyme's substrate specificity. To do so, we first constructed a homology-based structural model for the wild type 2-O-sulfatase using the three crystal structures (cited above) as a molecular scaffold. From this model, we present a picture of the enzyme active site from which we have identified specific residues likely to be involved in substrate binding and catalysis. Predictions concerning substrate discrimination made from this model are supported by the biochemical and kinetic data presented in the previous paper. In addition, a specific prediction regarding the sulfatase's exolytic mode of action is made; this prediction is subsequently validated by the biochemical data. Finally, we use a combination of aldehydespecific chemical labeling and peptide mapping methods in parallel with site-directed mutagenesis to identify the predicted active site cysteine, which is covalently modified to a formylglycine. The unsaturated heparin decasaccharide ⌬U 2S  H NS,6S I 2S H NS,6S I 2S H NS,6S IH NAc,6S GH NS,3S,6S 3 (AT-10) was generated by a partial heparinase digestion and purified as described (13). Modified trypsin (sequencing grade) was purchased from Roche Molecular Biochemicals. Texas Red hydrazine was purchased from Molecular Probes, Inc. (Eugene, OR). Ellman's reagent was purchased from Pierce. Enzymes for molecular cloning and PCR were obtained from New England Biolabs (Beverly, MA). All other reagents were purchased from Sigma unless otherwise noted.

Reagents-
Homology Modeling of 2-O-Sulfatase-The crystal structures of human arylsulfatase A, human arylsulfatase B, and the P. aeruginosa arylsulfatase 4 were used to obtain a structural model for the 2-Osulfatase enzyme. A multiple sequence alignment of the 2-O-sulfatase with different bacterial and lysosomal sulfatases was performed using ClustalW (14) (Fig. 1). Based on this multiple sequence alignment, three model structures of 2-O-sulfatase were obtained corresponding to its alignment with the other three sulfatases. The models were constructed using the Homology module of the Insight II molecular simulations package (Accelrys, San Diego, CA). The side chain of the critical Cys-82, which is shown to undergo posttranslational modification in the active enzyme, was replaced by the geminal diol (C␤(OH) 2 ). The potentials for the model structures were assigned using the AMBER force field (15). The deletions in the modeled structures were closed using 200 steps of steepest descent minimization without including charges by keeping most of the structure rigid and allowing the regions close to the deletion move freely. The final refined structure was subjected to 400 steps of steepest descent minimization without including charges and 400 steps of conjugate gradient minimization including charges.
Molecular Docking of Disaccharide Substrates into the Active Site of the Modeled 2-O-Sulfatase-Heparin-derived disaccharides with a ⌬U at the nonreducing end were modeled as follows. The coordinates of the trisulfated ⌬U containing disaccharide (⌬U 2S H NS,6S ) were obtained from the co-crystal structure of a heparinase-derived hexasaccharide with fibroblast growth factor 2 (Protein Data Bank code 1BFC). This trisulfated disaccharide structure was used as a reference to generate the structural models for other disaccharides including ⌬U 2S H NS , ⌬U 2S H NAc , and ⌬U 2S H NAc,6S . The coordinates of trisulfated disaccharides (I 2S H NS,6S ) containing iduronic acids in the 1 C 4 and 2 S 0 conformations were also obtained from 1BFC. Similarly, chondroitin sulfatederived disaccharides ⌬U 2S Gal NAc,4S and ⌬U 2S Gal NAc,6S were modeled using a reference structure of a chondroitin-4 sulfate disaccharide ⌬UGal NAc,4S whose coordinates were obtained from its co-crystal structure with the chondroitinase B enzyme (Protein Data Bank code 1DBO). The potentials for these disaccharides were assigned using the AMBER force field modified to include carbohydrates (15) with sulfate and sulfamate groups (16).
The orientation of the cleavable sulfate group relative to O␥1 of the geminal diol in the active site of human arylsulfatase A and the bacterial arylsulfatase was identical as observed in their respective crystal structures. This orientation was such that one of the faces of the tetrahedral formed by the three oxygen atoms of SO 3 Ϫ was oriented toward O␥1, facilitating the nucleophilic attack of the sulfur atom and the transfer of the SO 3 Ϫ group to O␥1. This highly specific orientation of the sulfate group helped in positioning the disaccharide substrates relative to the active site of the 2-O-sulfatase. After fixing the orientation of the 2-O-sulfate group, the glycosidic torsion angles and exocyclic torsion angles were adjusted manually to remove unfavorable steric contacts with the amino acids in the active site. The enzyme substrate complexes were minimized using 200 steps of steepest descent followed by 400 steps of Newton-Raphson minimization including charges. Most of the enzyme was kept rigid, and only the loop regions constituting the active site were allowed to move freely. On the other hand, a force constant of 7000 kcal/mol was applied to constrain the ring torsion angles to fix the ring conformation of the disaccharides during the energy minimization. The manual positioning of the substrates was done using the Viewer module, building of the disaccharide structures from the reference structures was done using the Builder module, and the energy minimization was done using the Discover module of Insight II.
Recombinant Expression and Protein Purification of the Flavobacterial 2-O-Sulfatase-The molecular cloning, recombinant expression in Escherichia coli, and subsequent purification of the 6ϫ histidine-tagged 2-O-sulfatase by nickel affinity chromatography is as described in the preceding paper. The sulfatase was expressed as an amino-terminal truncation lacking its putative signal sequence (commonly referred to in both papers as 2-O ⌬ 1-24 ).
AT-10 Compositional Analyses by Capillary Electrophoresis and MALDI-MS-Approximately 10 g of the AT-10 oligosaccharide were incubated with 100 pmol of 2-O ⌬N 1-24 in a 40-l reaction volume at 30°C. 15-l aliquots were removed at 4 and 17 h and heat-inactivated at 95°C. The oligosaccharide reaction products (along with 15 l of a minus sulfatase control) were subjected to an exhaustive heparinase I and III digestion prior to capillary electrophoresis-based compositional analysis. Desulfation of the decasaccharide was assayed in parallel by MALDI-MS using established methods (17).

2-O-Sulfatase Active Site Labeling and Peptide
Mapping-Approximately 500 g of 6ϫ histidine-tagged 2-O ⌬N 1-24 (wild type enzyme and C82A site-directed mutant) were lyophilized by Speed-Vac centrifugation and vigorously resuspended in 90 l of denaturation buffer containing 6 M guanidinium hydrochloride, 0.1 M Tris-HCl, pH 7.5. Active site aldehydes were fluorescently labeled by adding 25 l of Texas Red hydrazine made up as a 10 mM stock in dimethyl formamide. Labeling was carried out for 3 h at room temperature with gentle mixing on a rotating platform. The hydrazone linkage was stabilized by the addition of 10 l of a fresh 5 M sodium cyanoborohydride stock made up in 1 N NaOH. Reduction was carried out for 1 h at room temperature. Unreacted fluorophore was removed by repeated acetone precipitation (added 5:1 (v/v)). Acetone was prechilled at Ϫ20°C. Samples were chilled at Ϫ85°C for 20 min prior to spinning in a microcentrifuge for 10 min, maximum speed, at 4°C. Pellets were dried by Speed-Vac centrifugation.
The labeled sulfatase (and unlabeled control) were proteolyzed with sequence grade-modified trypsin for 20 h at 37°C in digestion buffer that contained 0.1 M Tris-HCl, pH 8.5, 1 mM EDTA, 1 mM dithiothreitol, and 10% acetonitrile (v/v) in a 30-l reaction volume. Trypsin was first reconstituted as a 2.5 mg/ml stock in 1% acetic acid and added at a 1:5 ratio (w/w) relative to the target protein. Following trypsin digestion, peptide cysteines were reduced by the addition of 50 mM dithiothreitol (50°C under argon, 1 h). Reduced cysteines were subsequently alkylated for 30 min at 37°C (in the dark) by the addition of 150 mM iodoacetamide, added from a 2 M stock made up in 0.1 M Tris-HCl, pH 8.5. This reduction-alkylation cycle was repeated one more time. Molecular masses of select peptides were determined by MALDI-MS as described (18) using 1 l of ␣-cyano-4-hydroxycinnamic acid in 50% acetonitrile, 0.3% trifluoroacetic acid as a matrix.
Site-directed Mutagenesis of the C82A Active Site Mutant-The sitedirected mutant C82A was cloned by recombinant PCR using outside primers 5Ј-TCT AGA CAT ATG CAA ACC TCA AAA GTA GCA GCT-3Ј (forward) and (5Ј-GT CTC GAG GAT CCT TAT TTT TTT AAT GCA TAA AAC GAA TCC-3Ј (reverse) in addition to the following mutagenic primer pair: 5Ј-C CAG CCG CTC GCT ACA CCT TCA CG-3Ј (forward) and 5Ј-CG TGA AGG TGT AGC GAG CGG CTG G-3Ј (reverse). The engineered codon change for each DNA strand is underlined. Subcloning into pET28a, recombinant expression in the E. coli strain BL21 (DE3), and subsequent purification by nickel chelation chromatography using the N-terminal 6ϫ histidine purification tag are as described for 2-O ⌬ 1-24 (previous paper (21)).

Structure-based Homology Modeling of the 2-O-Sulfatase
Active Site-In comparing the structures of human arylsulfatase A (9, 10), arylsulfatase B (N-acetylgalactosamine-4-sulfatase) (11), and a bacterial arylsulfatase from Pseudomonas aeruginosa (12), we observed an obvious structural homology between each of them, especially as it pertained to a conservation of critical active site residues and their spatial arrangement. By extension, most of these amino acids were likewise conserved in the flavobacterial 2-O-sulfatase as evident by a direct alignment of their primary sequences (Fig. 1). We used this close structural relationship to construct three homology-based models for the flavobacterial 2-O-sulfatase, each one based on one of the three crystal structures examined. We ultimately chose as our representative 2-O-sulfatase structure the homology model constructed using the N-acetylgalactosamine-4-sulfatase (arylsulfatase B) (Fig. 2). This decision was largely based on arylsulfatase B also being a GAG desulfating enzyme. In this model, we replaced cysteine 82 with a formylglycine (FGly-82). We chose to represent FGly-82 in the hydrated state as a geminal diol (-C␤(OH) 2 ), consistent with the proposed resting state (before catalysis) of the enzyme (5,9).
In the structural model, the surface of the active site pocket is composed of many amino acids that can potentially interact with a disaccharide substrate (Fig. 3). These include Lys-107, Lys-175, Lys-238, Gln-237 and -309, Thr-104, Glu-106, and Asp-159. Lysines and glutamines are commonly occurring amino acids in heparin binding sites that interact with the sulfate and carboxylate groups of the GAG chain. Unlike the amino acids proximal to the FGly-82, these residues are not conserved in the other sulfatases that we examined (Table I, denoted in normal type), suggesting a potentially unique role of these amino acids in dictating oligosaccharide substrate specificity. This disparity is particularly true when directly comparing the 2-O-sulfatase and arylsulfatase A; many of the nonconserved amino acids in the 2-O-sulfatase are charged, whereas those in arylsulfatase A are predominantly hydrophobic. This observation is consistent with the obvious structural distinction of their respective substrates (i.e. the highly sulfated HSGAG substrates of the 2-O-sulfatase versus the long hydrophobic alkyl chains of cerebroside 3-sulfate substrate of arylsulfatase A).
Active Site Topology and the Exolytic Action of the 2-O-Sulfatase-Upon inspection of the 2-O-sulfatase structure, several amino acids that potentially constitute the active site were identified (Table I). There are several structurally conserved basic amino acids in the proximity of FGly-82 including Arg-86, Lys-134, His-136, and Lys-308. The topology of the active site as observed in our structural model indicated that the critical FGly-82 and the basic amino acid cluster are located at the bottom of a deep pocket (Fig. 2B). Such a restrictive access to the active site would appear to impose a clear structural constraint on the substrate as it relates to the position of the 2-O-sulfate group within the oligosaccharide chain (i.e. externally versus internally positioned) upon which the enzyme acts. We would predict from this topology that only a sulfate group present at the nonreducing end of the oligosaccharide will be favorably positioned for catalysis; the juxtaposition of an internal sulfate into the active site would require a substantial bending of the oligosaccharide chain. Such chain distortion would be sterically unfavorable. Based on these constraints, therefore, we predict the sulfatase to hydrolyze 2-O-sulfates in an exclusively exolytic fashion. This exclusivity for the nonreducing end does not necessarily preclude, however, the enzyme acting on longer chain oligosaccharides (i.e. those exceeding a disaccharide in length), provided that they in fact possess sulfates at the terminal 2-OH-position. The model does suggest a likely kinetic preference for disaccharide substrates, since they would most readily diffuse into and out of this narrow active site (see enzyme-substrate structural modeling below).
We addressed this important question using as a substrate the purified heparin-derived AT-10 decasaccharide  (Fig. 4). Loss of a single sulfate was independently determined by mass spectrometry. Sequencing analysis of the singularly desulfated product allowed us to assign the original position of the 2-Osulfate exclusively to the uronic acid present at the nonreducing end. This result was confirmed using other oligosaccharide substrates, including tetra-and hexasaccharides containing both nonreducing end 2-O-sulfates and internal 2-O-sulfates (data not shown). Taken together, these results strongly suggest a sulfatase whose hydrolytic activity proceeds in an exolytic fashion. This conclusion is supported by our model, which, given the narrow topology of the enzyme active site, predicts a strong preference for the cleavage of sulfates positioned at the nonreducing end.
Enzyme-Substrate Structural Complex: Interactions between 2-O-Sulfatase and Disaccharides-Since the active site can readily accommodate disaccharide substrates, we modeled several unsaturated glycosaminoglycan disaccharides as described under "Experimental Procedures." Our choice of ⌬4,5-unsaturated substrates was logical for two reasons: 1) ␤-eliminative cleavage of a HS polysaccharide by the flavobacterial lyases that naturally occurs in vivo results in the formation of disaccharides and other small oligosaccharides, all possessing a ⌬4,5-unsaturated bond at the nonreducing end uronic acid and 2) the obligatory substrate-product relationship between the 2-O-sulfatase and the ⌬4,5-glycuronidase that exists both in vitro (as described in the previous paper (21)) and in vivo. In addition, we directly tested the requirement of a ⌬4,5 uronic acid directly in experiments described below. A representative structural complex involving the trisulfated disaccharide ⌬U 2S H NS,6S (Fig. 3) was used to describe the molecular interactions between the enzyme and the substrate. This choice was made based on the biochemical data (shown in the accompanying manuscript (21)) that indicated the trisulfated disaccharide as the kinetically preferred substrate. A description of these interactions and their proposed functional roles is shown in Table II. The functional roles of the conserved active site amino acids (listed in boldface type in Table I) (Table I).
The Requirement for an Unsaturated ⌬4,5 Nonreducing Terminus-A closer inspection of the modeled enzyme-substrate complex revealed some interesting possibilities pertaining to the role of the nonconserved amino acids in substrate recognition and binding. The planar carboxylate group attached to the C-5 atom of the ⌬4,5 uronic acid is oriented in such a manner as to potentially interact with Lys-175 and Lys-238. These amino acids could play an important role, therefore, in favor-ably orienting the 2-O-sulfate within the active site. We were further interested in this arrangement given the additional constraint imposed upon the planar carboxyl group of the uronic acid by the presence of the C-4 -C-5 double bond. This constraint may further influence substrate orientation within the active site. As such, we would predict a substrate discrimination exhibited by the 2-O-sulfatase that is based on the presence of the ⌬4,5 double bond at the oligosaccharide nonreducing terminus. In the absence of this double bond, the favorable orientation of the C-5 carboxylate afforded by charge interactions with Lys-175 and Lys-238, respectively, would not Basic residues are colored blue, acidic amino acids are colored red, Gln is colored light blue, Thr is colored orange, Leu is colored brown, and FGly-82 is colored purple. The docked disaccharide (colored in green) is also shown using a stick representation. The sulfur atom of 2-O-sulfate group is colored yellow, and oxygen atoms of the 2-O-sulfate group and the planar carboxyl group are colored red. B, schematic representation of the amino acids shown in A using the same color scheme. Potential metal ion coordination is also shown, with the divalent cation (M 2ϩ ) depicted as a gray circle.
occur. The change in the orientation of the C-5 carboxylate due to the lack of the double bond would, in turn, disrupt the favorable positioning of the 2-O-sulfate group for catalysis.
To better understand this likely structural constraint, we superimposed substrate disaccharides containing a nonreducing end iduronic acid in either the 1 C 4 or 2 S 0 conformation (data not shown). The superimposition was such that the S-O-C-2-C-1 atoms of all of the uronic acids coincided, thereby fixing the orientation of the 2-O-sulfate group. In this model, the carboxylate groups of the iduronic acid containing disaccharide substrates were, in fact, pointing away from the active site pocket and were not positioned to interact as favorably with Lys-175 or Lys-238 as compared with the original disaccharide substrate possessing a planar C-5 carboxylate.
We assessed this prediction experimentally by testing the ability of the 2-O-sulfatase to hydrolyze size-fractionated hexasaccharides derived from the nitrous acid treatment of heparin. Unlike enzymatic cleavage, these chemically derived heparin saccharides do not possess a ⌬4,5-unsaturated bond at their respective nonreducing ends. A majority of the resultant tetrasaccharides, however, do contain an I 2S at this end. Using MALDI-MS, we were unable to detect any enzyme-dependent desulfation of treated hexasaccharides under conditions that resulted in quantitative elimination of 2-O-sulfate from ⌬4,5containing oligosaccharides. This result strongly suggests a structural requirement for the ⌬4,5 bond in addition to the exolytic preference of the enzyme. As we have already pointed out, our model provides a rationale for this requirement (i.e. the obligatory physical connection between this bond and the planar carboxylate at the uronic acid C-5 position). The constraint imposed by a planar C-4 -C-5 double bond on the orientation of this carboxylate in turn confers upon the same substrate an orientation of the 2-O-sulfate within the enzyme active site that is favorable for catalysis.
Enzyme-Sulfate Interactions in the 2-O-Sulfatase Active Site-In the previous paper (21), we present a biochemical data demonstrating a clear kinetic preference of the enzyme for highly sulfated disaccharide substrates, namely those possessing a glucosamine sulfated at the 6-OH-and 2-N-positions. Our structural model of the sulfatase-trisulfated disaccharide complex also points out key interactions involving these additional sulfates present on the adjoining glucosamine. In particular, the 6-O-sulfate group interacts with the basic side chain of Lys-107 within the enzyme active site (Fig. 3). This putative charge interaction would probably play an important role in stabilizing the orientation of the substrate in the active site. In contrast, the N-sulfate group of the disaccharide glucosamine is proximal to a contiguous stretch of leucines (positions 390 -392). In such an arrangement, it is the methyl group of an N-acetylated glucosamine rather than a sulfate at this position that is more likely to make favorable hydrophobic contacts with these residues. This prediction was borne out in one of our models docking the ⌬U 2S H NAc,6S substrate in the active site (data not shown).
We also modeled enzyme-substrate complexes containing two unsaturated chondroitin sulfate disaccharides (⌬U 2S Gal NAc,4S and ⌬U 2S Gal NAc,6S ). In comparison with our original model using the heparin disaccharide substrate, we found interactions with the 2-O-sulfate and carboxyl group of the ⌬U monosaccharide that were identical to that of ⌬U 2S H NS,6S . There were fewer interactions involving the 4-sulfate and 6-sulfate groups, however. This particular model, therefore, is consistent with our kinetic data describing the TABLE I Structure-based comparison of sulfatase active site residues Highly conserved amino acids are listed in boldface type. Nonconserved amino acids are listed in normal type. Amino acids in the 2-O-sulfatase that could be potentially involved in substrate binding are noted by an asterisk. Structural alignment of the modeled 2-Osulfatase structure with the other sulfatases was obtained based on superposition of their C-␣ traces using the combinatorial extension algorithm (20). Regions of deletion in the structural alignment are noted with a minus sign.  ability of the so-called "heparin/heparan sulfate" 2-O-sulfatase to hydrolyze 2-O-sulfated chondroitin disaccharides. Given a lack of additional favorable contacts between the enzyme and substrate (e.g. with either the 4-O-or 6-O-sulfates), our model also predicts a lower catalytic efficiency for the chondroitin disaccharides relative to the structurally corresponding heparin disaccharides.
Metal Ion Coordination-In discussing this model, we must briefly consider the potential role of divalent metal ions. We decided not to include any such metal ions in our theoretical model of the 2-O-sulfatase, since we could find no divalent metal requirement for enzymatic activity (see preceding paper (21)). A divalent metal ion is present, however, in all three sulfatase crystal structures that we examined. In each case, the metal ion coordinates with the oxygen atoms of the sulfate group of the respective substrate. Additionally, a cluster of four highly conserved acidic amino acids has been observed to coordinate this divalent metal ion. In the case of human arylsulfatase B, for example, the oxygen atoms of Asp-53, Asp-54, Asp-300, and Asn-301 are coordinated with a Ca 2ϩ ion. Three of the four corresponding amino acids in the flavobacterial sulfatase model that we have identified as potentially coordinating with a metal ion are Asp-42, Gln-43, and Asp-295 (Table I). The fourth amino acid in the 2-O-sulfatase corresponding spatially to Asn-301 of arylsulfatase B is His-296. The positive charge of this position, however, does not favor the proximal location of a divalent metal cation. It is perhaps this unfavorable charge interaction that interferes with proper metal ion coordination.
Enzyme-Substrate Model: Mechanism for Catalysis-Nearly identical mechanisms for the hydrolysis of the sulfate ester bond involving the conserved active site amino acids have been proposed for human arylsulfatases A and B and the bacterial sulfatase from P. aeruginosa. Thus, the mechanism proposed for these other sulfatases should logically apply to the 2-Osulfatase. The resting state of the active sulfatase in each of the crystal structures is proposed to contain the geminal diol, which is stabilized by interactions with basic residues. His-136 and Arg-86 of the flavobacterial 2-O-sulfatase enzyme are positioned appropriately in the active site to do so (Fig. 2B). A critical step in catalysis involves the correct positioning the 2-O-sulfate group such that the sulfur atom is accessible to the O-␥1 of the geminal diol. We have already described how interactions of specific active site amino acids with the planar carboxyl group of the uronic acid (Lys-175 and -238), with the 6-O-sulfate of the glucosamine (e.g. Lys-107 and possibly Thr-104) and with the 2-O-sulfate itself (Lys-134 and -308) are likely to serve in this capacity (Table II). At the same time, interaction of the 2-O-sulfate group with charged amino acids would also enhance any electron density withdrawal from the oxygen atoms, thereby increasing the electrophilicity of the sulfur center. It has also been suggested that the nucleophilicity of the O-␥1 atom is enhanced by a possible proton donation to a neighboring aspartic acid residue. In our structural model of the 2-O-sulfatase, this residue would correspond to Asp-295.
An S N 2 mechanism has been proposed to follow the above steps and eventually lead to the cleavage of the sulfate ester bond (10,12). In this mechanism, the exocyclic oxygen atom on the leaving substrate may be protonated by water or potentially by neighboring amino acids. In the 2-O-sulfatase active site model, Lys-308 is juxtaposed to protonate the leaving group (Fig. 2). The resulting sulfate group on the geminal diol is subsequently eliminated by abstraction of a proton from O-␥2, regenerating the formylglycine. His-136 is positioned to abstract this proton. Our functional assignment of these lysine  and histidine residues is based entirely on homologous positions found in the other sulfatases examined and their proposal to catalytically function in the fashion described (5).

2-O-Sulfatase Peptide Mapping and Chemical
Modification of Active Site Formylglycine-Finally, in describing the structure-function relationship of the 2-O-sulfatase active site, we come to the central catalytic player itself, the formylglycine at position 82. The recombinant expression of catalytically active 2-O-sulfatase in E. coli functionally argues for this covalent modification of the active site in vivo. We further established the catalytic function of Cys-82 by site-directed mutagenesis. The mutant (C82A) was recombinantly expressed and purified as a histidine-tagged protein in the same manner employed for the wild type enzyme. Comparable expression levels of soluble protein were achieved. The C82A mutant, however, was completely inactive (data not shown). Both the wild type and mutant possessed the same secondary structure as exhibited by their virtually superimposible CD spectra (data not shown), arguing against any adverse global conformational changes induced by the molecular replacement of the cysteine by alanine.
We also set out to demonstrate the physical presence of the FGly at position 82 by the tandem use of protein chemistry and mass spectrometry. 10 nmol of wild-type sulfatase (2-O ⌬N 1-24 ) and the C82A mutant were reacted with Texas Red hydrazide (620.74 Da) as described under "Experimental Procedures." The two sulfatase fractions were subsequently trypsinized under mildly denaturing conditions followed by reductive methylation of the unmodified cysteines. The molecular masses of the resultant peptides were determined by MALDI-MS (Fig. 5). In this experiment, we identified a single ionized species uniquely present in the labeled sulfatase experiment (Fig. 5B) but absent in the active site mutant (Fig. 5C) or in the unlabeled control (Fig. 5A). The empirical mass of this species corresponded most closely to the peptide sequence FTRAYCAQ-PLCTPSR, resulting from a partial trypsin cleavage. This peptide contains the sulfatase consensus sequence CXPXR, which includes the critical active site cysteine (underlined) at position 82. The mass of this peptide is consistent with first the conversion of this cysteine to a formylglycine (FGly-82) followed by the covalent hydrazone linkage of the aldehyde-reactive fluorophore at this position. It also takes into account the carbamidomethylation of the second (unmodified) cysteine present in this peptide. These data, taken together with the loss of function observed for the C82A mutant, establish the important structure-function relationship for this active site modification.
To confirm and extend these findings, we also completed chemical modification experiments on the wild type enzyme using the cysteine-specific reagent DTNB (Ellman's reagent). Apart from the catalytic Cys-82 that is converted to a formylglycine, none of the remaining seven cysteines appear to be highly conserved among other members of the sulfatase family (Fig. 1). We were unable to effectively inhibit enzyme activity with the addition of DTNB or dithiothreitol (data not shown). This general lack of inhibition by these two cysteine-reactive agents suggests at least two probabilities. First, the 2-O-sulfatase does not require intramolecular disulfide linkages to critically stabilize a catalytically active conformation. This presumption is not surprising, given the fact the native 2-O-sulfatase is a bacterially derived protein. Second, free sulfhydryls are not directly participating in catalysis. It is possible, however, that a few of these cysteines are buried and are therefore not accessible to sulfhydryl exchange. At least five of the eight cysteines, however, do react with DTNB under nondenaturing conditions (data not shown). This latter fact may suggest an alternate role for these solvent-accessible cysteines (along with specific histidines), perhaps as metal-coordinating thiolates. This possibility is a reasonable one when one makes the comparison between the 2-O-sulfatase and alkaline phosphatase. Both of these enzymes are esterases with similar catalytic mechanisms, including the presumptive formation of a covalent intermediate. The two hydrolytic enzymes also possess structurally related domains, in particular a highly superimposible active site that includes a divalent metal binding pocket. In the case of alkaline phosphatase, it is zinc rather than calcium (or Mg 2ϩ ) that is coordinated within this pocket.
In the previous paper, we described the molecular cloning of a HSGAG 2-O-sulfatase from F. heparinum and its recombinant expression in E. coli as a soluble, highly active enzyme in milligram quantities. We also presented a characterization of the enzyme's biochemical properties as it pertains to optimal in vitro reaction conditions and substrate kinetics. In this paper, we build upon this biochemical characterization of the enzyme by providing an invaluable structural framework to address the molecular basis of enzyme function. To the best of our knowledge, this paper represents one of the first structural descriptions of an HSGAG desulfating enzyme. As stated before, our homology model of the 2-O-sulfatase has already provided valuable insight into likely structural determinants for substrate specificity and catalysis. Many of these insights are validated by our biochemical and kinetic evaluation of this specificity. Importantly, we now have in place a meaningful structural framework in which to address additional questions pertaining to oligosaccharide-sulfatase interactions.