Catalytic Mechanism of Heparinase II Investigated by Site-directed Mutagenesis and the Crystal Structure with Its Substrate*

Heparinase II (HepII) is an 85-kDa dimeric enzyme that depolymerizes both heparin and heparan sulfate glycosaminoglycans through a β-elimination mechanism. Recently, we determined the crystal structure of HepII from Pedobacter heparinus (previously known as Flavobacterium heparinum) in complex with a heparin disaccharide product, and identified the location of its active site. Here we present the structure of HepII complexed with a heparan sulfate disaccharide product, proving that the same binding/active site is responsible for the degradation of both uronic acid epimers containing substrates. The key enzymatic step involves removal of a proton from the C5 carbon (a chiral center) of the uronic acid, posing a topological challenge to abstract the proton from either side of the ring in a single active site. We have identified three potential active site residues equidistant from C5 and located on both sides of the uronate product and determined their role in catalysis using a set of defined tetrasaccharide substrates. HepII H202A/Y257A mutant lost activity for both substrates and we determined its crystal structure complexed with a heparan sulfate-derived tetrasaccharide. Based on kinetic characterization of various mutants and the structure of the enzyme-substrate complex we propose residues participating in catalysis and their specific roles.

Heparin and heparan sulfate (HS) 3 glycosaminoglycans (GAGs) are negatively charged, linear polysaccharides com-posed of repeating disaccharide units of uronic acid and glucosamine residues (GlcN, 2-amino-2-deoxy-␣-D-glucopyranose) (1). Heparin typically contains ϳ90% iduronic acid (IdoA, ␣-Lidopyranosyluronic acid) and 10% glucuronic acid (GlcA, ␤-Dglucopyranosyluronic acid), with a high content of 2-O-sulfo groups on the IdoA residue. The glucosamine residue in heparin is predominantly substituted with N-sulfo groups (GlcNS, where S is sulfo) and 6-O-sulfo groups with a small number of N-acetyl groups and much less frequently with 3-O-sulfo groups. In contrast, HS is somewhat more diverse in its primary structure and characterized by a higher percentage of the GlcA epimer, N-acetyl-substituted GlcN (GlcNAc) and a lower percentage of 2-O-sulfo, 6-O-sulfo, and N-sulfo groups. The modifications in HS are not uniform; rather, they are concentrated within specific regions of the polysaccharide, giving rise to a short, sulfo group containing sequence motifs responsible for the interactions between HS and a diverse repertoire of proteins leading to its multiple biological roles. These complex polysaccharides provide docking sites for numerous protein ligands involved in diverse biological processes, ranging from cancer and angiogenesis, anticoagulation, inflammatory processes, viral and microbial pathogenesis to multiple aspects of development (2)(3)(4)(5)(6)(7)(8)(9). Moreover, HS-GAGs are abundant at the cell surface as part of the proteoglycan cell surface receptors (3,4).
Specialized microorganisms express GAG-degrading lyases serving nutritional purposes of both themselves and their vertebrate hosts. The lyases depolymerize GAGs through a ␤-elimination mechanism characterized by the release of an unsaturated product (10). The efficiency of the elimination relies on the stereospecificity of the active site amino acids participating in the various mechanistic steps of elimination, namely charge neutralization of the uronic acid carboxyl group, proton abstraction by a general base, and protonation of the leaving group by a general acid. Heparinase II (heparin lyase II; heparitinase II, HepII; EC number 4.2.2.7) from Pedobacter heparinus (previously known as Flavobacterium heparinum) with a molecular mass of 84.5 kDa and a pI of 8.9, is unique in its ability to cleave both heparin and HS-like regions of GAGs ( Fig.  1) regardless of their sulfation patterns (11). The mature protein consists of 747 residues starting with pyroglutamate 26 and ending in Arg 772 , with a molecular mass of 84,545 Da (12) and is the prototype of a new polysaccharide lyases family PL21 (CAZy (37)).
HepII acts in an endolytic manner (13) and displays broad selectivity, catalyzing the cleavage of linkages adjacent not only to IdoA and GlcA but also to the rare ␣-L-galacturonic acid residues (14,15). At the glucosamine positions HepII will degrade either GlcNAc or GlcNS and even the rare GlcN, provided that uronic acid contains a 2-O-sulfo group. Although HepII has greater affinity for heparin, its turnover rate for HS is higher (16). Moreover, the enzyme displays preference toward degradation of glycosidic bonds containing GlcA over ones containing IdoA. HepII shows greater catalytic efficiency for longer rather than shorter oligosaccharide substrates (13). Previously, we have solved the structure of native HepII (overexpressed in its original host P. heparinus; PDB accession code 2FUQ (12,17,18) and the structure of rHepII (overexpressed in Escherichia coli) in complex with a disaccharide product (PDB accession code 2FUT (17)).
The crystal structure combined with solution studies showed that the biological unit of the enzyme is a dimer with two active sites located at opposite sides of the dimer and separated by ϳ80 Å (17). Each monomer of HepII has an ␣ϩ␤ lyase-fold similar to that of the enzymes from the PL8 family (CAZy). Marked differences in the relative disposition of the structural subdomains in HepII, as compared with the other PL8 family proteins with known structures, lead to a significantly larger participation of the central ␤-sheet domain in substrate binding.
The HepII-disaccharide complex provided the first insight into the active site and allowed us to speculate on possible roles of several amino acids in performing the ␤-elimination. Based on the structure, three residues, Tyr 257 , His 202 , and His 406 , were proposed to form the putative active site. Here we investigate further the role of these residues in the catalytic mechanism of HepII employing site-directed mutagenesis, enzymatic charac-terization of the mutants, crystallography, and molecular simulations.

Preparation of the HS Tetrasaccharide Substrates
Heparin and HS sodium salts were obtained from porcine intestinal mucosa (Celsus Laboratories, Cincinnati, OH). Recombinant P. heparinus heparin lyases I and III expressed in E. coli and used in depolymerization were a gift from Dr. J. Liu. HS tetrasaccharides were prepared as previously described (19). Briefly, the HS was partially digested (30% reaction completion) with heparinase III and fractionated by gel-permeation chromatography using a Bio-Gel P-10 column to obtain uniform sized oligosaccharides. The fraction containing tetrasaccharides was desalted on a Bio-Gel P-2 column (Sigma). The tetrasaccharide mixture was then concentrated and fractionated on a semi-preparative HPLC column (Shimadzu, Columbia, MD). After collection of individual peaks from the HPLC, the peaks were re-separated (Spherisorb S5 SAX 20 ϫ 250 mm semi-preparative column, Waters, Milford, MA) to obtain high purity oligosaccharides. Fifteen peaks, corresponding primarily to tetrasaccharides, were obtained in the initial semi-preparative HPLC. Chromatographic purification of these peaks were analyzed by LC-MS (Agilent 1100 LC/MSD, Agilent Technologies, Inc., Wilmington, DE) under conditions previously described (20). Generally, 5 g of each sample was loaded through a 5-m Agilent Zorbax SB-C18 (0.5 ϫ 250 mm) column. Eluent A was water/acetonitrile (85:15, v/v) and eluent B was water/acetonitrile (35:65, v/v). Both eluents contained 12 mM tributylamine and 38 mM ammonium acetate and their pH was adjusted to 6.5 with acetic acid. A gradient of 0% B for 15 min, and 0% to 100% B over 85 min was used at a flow rate of 10 l/min. The electrospray interface was set in negative ionization mode with the skimmer potential at Ϫ40.0 V, capillary exit Ϫ120 0.5 V, and source temperature of 325°C to obtain maximum abundance of the ions in a full scan spectra (150 -1500 Da, 10 full scans/s). Nitrogen was used as a drying (5 liter/min) and nebulizing gas (20 p.s.i.). Auto MS/MS was turned on in these experiments using an estimated cycle time of 0.07 min. Total ion chromatograms and mass spectra were processed using Data Analysis 2.0 (Bruker Software, Billerica, MA). Tetrasaccharide 5 (Fig. 2) was obtained from heparin digestion by heparinase I and purified as described above. Structures 1, 3, 4, and 5 ( Fig. 2) were confirmed by 1 H NMR and two-dimensional COSY NMR (600 MHz NMR spectrometer, Bruker). Before data acquisition, the active hydrogen atoms of each sample were exchanged three times, each time with 0.3 ml of deuterium oxide (99.96%) followed by lyophylization. Structure 2 was prepared from structure 4 by N-sulfonation using the protocol as previously described (19). Basically, 0.9 mg of tetrasaccharide 4 was dissolved in 0.2 ml of solution containing 10 mg/ml of sodium bicarbonate and 10 mg/ml of trimethylamine-sulfurtrioxide complex and incubated at 50°C for 12 h. Equal portions of sodium bicarbonate and trimethylamine-sulfurtrioxide complex were added two more times at 12-h intervals. The reaction mixture was concentrated and then desalted by a Bio-Gel P-2 column (1 ϫ 50 cm) and the product was lyophilized and its structure was confirmed by LC-MS.

Site-directed Mutagenesis
HepII was cloned as described earlier (17). HepII mutants were constructed using the QuikChange Site-directed Mutagenesis Kit (Stratagene), according to the manufacturer's protocol and their sequences confirmed by DNA sequencing. The mutants were expressed in E. coli BL21(DE3).

Activity Analysis of Mutant Heparinase II on Polysaccharides and Structure-defined Tetrasaccharides as Substrate
Activity studies were performed on both polysaccharide and structurally defined tetrasaccharide substrates using WT and mutated heparinase II. In the polysaccharide assays, each reaction used 2 g/l of heparin or HS treated with 0.5 g/l of either WT or mutated enzyme in 50 mM sodium phosphate buffer, pH 7.0, at 37°C. Reactions were terminated after 20 h by adding 30 mM hydrochloric acid until the pH reached 3.0. Aliquots were taken from the mixture for the measurement of absorbance change at 232 nm. PAGE has also been performed to qualitatively analyze the degree of reaction completed in addition to photometric measurement (see supplementary data for details).
Activity assay of WT and mutated heparinase II on tetrasaccharide substrate were performed each using 0.5 g/l of enzyme reacting with 0.1 g/l of tetrasaccharide in 20 mM sodium phosphate buffer containing 0.1% bovine serum albumin, pH 7.0, at 37°C. Control reactions were prepared using the same amount of substrate with no enzyme. The reaction was run for 12 h and was terminated by boiling the mixtures for 10 min followed by centrifugation at 10,000 ϫ g to remove the inactivated enzymes. The supernatants containing the reaction products were separated and analyzed by LC-MS.
The native protein was crystallized and diffraction data were collected as described previously (17,18). Briefly, initial crystals were obtained from 6.5 mg/ml of protein in 10 mM sodium phosphate, pH 7.5, 100 mM NaCl, 5 mM dithiothreitol, and a reservoir solution containing 17% (w/v) PEG 3350, 200 mM sodium phosphate, pH 5. The initial, needle-like crystals were crushed manually and used for microseeding over the same rHepII Mutants-Expression, purification, and crystallization of rHepII from P. heparinus were preformed as described earlier (17). Briefly, recombinant, histidine-tagged rHepII was expressed in BL21(DE3), grown in room temperature (20°C) in LB medium supplemented with ampicillin. The protein was purified by immobilized metal-chelate chromatography (Qiagen) and the histidine tag was removed using thrombin (Sigma). The enzyme was further purified on a Source 15S cation exchange column (GE Healthcare). The purified protein was concentrated to ϳ7 mg/ml by ultrafiltration using a Centricon YM-100 concentrator (Millipore Corp.) in 25 mM HEPES, pH 6.9, 150 mM NaCl, 5 mM dithiothreitol. The protein was crystallized at room temperature using the hanging drop vapor diffusion method in 24-well Linbro plates (Hampton Research). One microliter of protein was mixed with 1 l of reservoir solution containing 25-26% PEG 3350, 200 mM ammonium acetate or 200 mM MgCl 2 or 200 mM Mg-formate, 100 mM Tris, pH 8 -8.5, set over 1 ml of reservoir solution.

Data Collection Structure Determination and Refinement
Native HepII crystals were soaked briefly in reservoir solution supplemented with 20% glycerol and 3 mM HS tetrasaccharide substrate and flash frozen under N 2 stream at 100 K (Oxford Cryosystems, Oxford, UK). Diffraction data were collected on a HTC area detector using Micromax 007 generator with Osmic mirrors (Rigaku). The data were processed to 2.35-Å resolution using d*TREK (Rigaku). Data collection statistics are given in Table 1. rHepII mutant crystals were soaked for 20 min in 25% (w/v) PEG 3350, 200 mM ammonium acetate or 200 mM MgCl 2 or 200 mM Mg-formate, 12% (v/v) glycerol, supplemented with the 3 mM heparin or HS tetrasaccharide substrate and frozen in the nitrogen stream at 100 K (Oxford Cryosystems). Diffraction data extending to 2.1-Å resolution were collected at beamline X8C (NSLS, Brookhaven National Laboratory), using a Quantum-4 CCD area detector (ADSC, San Diego, CA) and processed using HKL2000 (22). Data collection statistics are summarized in Table 1.
The structures were determined by molecular replacement using the program MolRep (23) with the PDB deposited coordinates of nHepII and rHepII as a search models (PDB codes 2FUQ and 2FUT) (17). The structures were refined using the program REFMAC5 (24). The refined structure of the nHepII-HS product complex extends from PCA26 to Arg 772 and contains 3 protein molecules in the asymmetric unit. The model also contains 3 Zn 2ϩ ions, 511 water molecules, 10 phosphate molecules, 3 glycosylation sites containing 4-membered glycans, and 3 disaccharide products bound to each monomer of HepII. The refined structure of the rHepII-HS tetrasaccharide complex contains two protein molecules in the asymmetric unit, each comprised of Ala 29 -Arg 772 , 2 Zn 2ϩ ions, 784 water molecules, 4 acetate molecules, and two tetrasaccharide molecules. The refinement statistics for both structures are summarized in Table 1.
Molecular Dynamics of HS and Heparin Tetrasaccharides Bound to HepII-The starting model for calculations was the WT HepII with superimposed tetrasaccharide 3 determined here. An IdoA-containing tetrasaccharide analog was constructed with IdoA in a 1 C 4 chair or in the 2 S 0 twist-boat conformation. The protonation state at physiological pH was adopted, with the exception of several His residues for which tautomeric and ionization states were decided upon visual examination of their structural contexts. The Glu 205 residue

Activity of HepII Mutants
Heparin and HS as Substrates-Three amino acids, His 202 , Tyr 257 , and His 406 , in the vicinity of the uronic acid C5 atom, the chiral center of the reaction, were previously identified as potential catalytic residues (17). The side chains of Tyr 257 and His 406 stack parallel to each other on one side of the sugar ring (a general direction of the C5 proton in GlcA in HS substrate). The third residue, His 202 , approaches the C5 from the opposite side of the ring than His 406 (a general direction of C5 proton in IdoA in heparin substrate). The Tyr 257 -His 406 arrangement is similar to that found in the active site of other PL8 enzymes, namely ChonAC (27,28), ChonABC (29), hyaluronan lyases (30), and alginate lyase (31). These enzymes utilize the tyrosine side chain in a deprotonated state as a general base to abstract the C5 proton, triggering the elimination (27). To investigate the roles of these residues in catalysis we constructed Y257F, Y257A, H406A, and H202A mutants. They were expressed and purified to homogeneity. The mutants showed no measurable activity in a real time assay at 37°C that measured the initial velocity of double bond formation (monitoring the absorbance at 232 nm).
To evaluate if these mutants displayed a low level of residual activity, we mixed each single mutant with either heparin or HS for a prolonged period of time (ϳ12 h) and analyzed the banding patterns of the oligosaccharides distribution by PAGE (supplemental Fig. S1). UV detection was used to quantify roughly the amount of oligosaccharide degradation products. Parallel experiments were done with all the mutants using WT HepII as control (Table 2). In these assays the Y257F mutant showed the highest residual activity, whereas the H406A mutant showed the lowest residual activity toward both heparin and HS substrates. The products were measurable only when a high concentration of protein and prolonged incubation period were used. H202A also displays low residual activity, in particular toward heparin.
To exclude the possibility that the diminished activity is due to effect of these mutations on substrate binding rather than catalysis we measured the binding kinetics of WT and mutant HepII to heparin using surface plasmon resonance (see supplemental data for experimental details). The obtained k on , k off , and K D showed that the mutations had only a small effect on binding, with K D changing by no more than a factor of 3 (Table 3).
Defined Tetrasaccharide Substrates-To better characterize the specificity of these mutants and their effects on catalysis we have prepared small amounts of defined tetrasaccharides as shown in Fig. 2. Pairs of epimeric tetrasaccharides were prepared containing either iduronic acid or glucuronic acid at their ϩ1 site and differing in their sulfation pattern. The structures of tetrasaccharide 1 (0.9 mg), tetrasaccharide 2 (1.0 mg), tetrasac-  (Fig. 2).
The activity of HepII and its mutants was assessed over a prolonged incubation time. The degree of tetrasaccharide digestion using WT HepII is presented in Table 4. The first two entries in Table 4 and Fig. 3 show the results on 1 and 2, substrates that only differ in the chirality at the C5 position in their internal uronic acid residue. Tetrasaccharide 2 has an internal GlcA, the most common uronic acid unit in HS and tetrasaccharide 1 has an internal IdoA, the most common uronic acid in heparin. Y257F reached 100% digestion on tetrasaccharide 1 but only ϳ90% conversion on tetrasaccharide 2. H406A had trouble digesting both tetrasaccharides 1 and 2, giving an  The mutants were next examined for their activities on tetrasaccharides 3, 4, and 5 (the last three entries in Table 4 and supplemental Fig. S2). Tetrasaccharide 3 is the least digested, whereas tetrasaccharide 5 showed full conversion. Structural differences between tetrasaccharides 3, 4, and 5 can be seen in the substitution of the GlcN residue at the Ϫ1 site. Tetrasaccharide 3 contains a GlcNAc, tetrasaccharide 4 contains an unsubstituted GlcN, and tetrasaccharide 5 a GlcNS. These data suggests that the activities of rHepII H202A, H406A, and WT HepII on these tetrasaccharide substrates are significantly impacted by the polarity of the C2 amino group in the GlcN residue at the Ϫ1 subsite. The highly polar GlcNS residue in tetrasaccharide 5 makes it the best substrate, whereas the low polarity GlcNAc residue in tetrasaccharide 3 makes it the poorest substrate.
We crystallized HepII(Y257F), HepII(Y257A), and HepII(H202A) in the presence of tetrasaccharide substrate 3 and solved the structures of these complexes (data not shown). The structure of each complex showed only the presence of a disaccharide degradation product bound in the active site, confirming residual activity of these mutants.
The observation that all single mutants showed some residual activity made us look for other residues with a potential role in catalysis. Only one such residue was apparent, Tyr 429 . Therefore, we constructed Y429A and Y429F mutants and assayed its properties (Table 4). Whereas this mutation affects specific activity, the effect was much less dramatic, indicating that this side chain is likely involved in substrate binding but not directly in catalysis.
Next, we prepared three double mutants, H202A/Y257A, Y257A/H406A, and H202A/H406A, to test if they still display any residual activity. Overnight incubation with excess heparin or HS substrates showed no appearance of disaccharides indicating that these double mutants are completely inactive. We have crystallized one of these mutants, H202A/Y257A, with HS tetrasaccharide 3 and determined the structure of the enzymesubstrate complex. The entire tetrasaccharide was observed in the electron density map.

Structure of HepII-HS Disaccharide Degradation Product
Previously we have determined the structure of HepII complexed with a heparin disaccharide degradation product (⌬UA2S-GlcNS6S-GlcA2S-GlcNS6S) and postulated that both heparin and HS bind HepII in the same site and with only one active site present in the enzyme (17). To validate this hypothesis we soaked native crystals of HepII with a representative HS tetrasaccharide 3 (⌬UA-GlcNAc-GlcA-GlcNAc), and solved the structure of this complex (Table 1). Only the density for a disaccharide degradation product (⌬UA-GlcNAc) was observed in the crystal indicating that the enzyme retained activity in the crystal. The disaccharide was bound in "ϩ1" and "ϩ2" subsites (Fig. 4a), in a very similar way as observed for a disaccharide product of heparin degradation in HepII co-   100  92  0  95  96  Tetra 2  100  93  92  100  95  94  Tetra 3  10  38  53  3  ND  ND  Tetra 4  55  ND b  ND  18  ND  ND  Tetra 5  100  ND  100  100 ND ND a Conversion percentage ϭ ͓product peak area/(product peak area ϩ substrate peak area)͔ ϫ 100 (from Fig. 3 and supplemental Fig. S2). b ND, not determined. crystallized with heparin (17). Indeed, the superposition of the two complexes based on the C␣ atoms of HepII shows that the two disaccharides overlay each other. The disaccharide was found in a deep, elongated, positively charged cleft formed between the N-terminal and the central subdomains of HepII. Both sugar rings were stacked parallel to the rims of the cleft and the N-acetyl/N-sulfo group points toward the bottom of the cleft. The binding site is clearly divided into two parts, a narrow and less deep part where the "ϩ" subsites are located (toward the reducing end of the substrate (32)) and a wider and deeper part containing the "Ϫ" subsites, in which the sugars on the non-reducing side of the cleaved bond are bound. The products occupy ϩ1 and ϩ2 subsites differentiating HepII from other GAG lyases, where the disaccharide product binds tighter to the Ϫ subsites (31,33). Elaborate hydrogen-bonding networks maintain the position and correct orientation of side chains lining the binding site. The disaccharide forms intimate contacts with the protein through hydrogen bonds, aromatic stacking, and van der Waals interactions (Fig. 4a).

Structure of HepII(H202A/Y257A)-HS Tetrasaccharide
The recombinant HepII H202A/Y257A double mutant was crystallized and the crystals were soaked with the HS representative tetrasaccharide 3, ⌬UA-GlcNAc-GlcA-GlcNAc. The omit electron density map showed clearly the entire substrate (Fig. 4b). This structure allowed us to map the full binding/ catalytic subsite from "Ϫ2" to ϩ2 (PDB accession number 3E7J). No major change in the structure as compared with the WT enzyme was observed. The Ϫ subsites displayed fewer direct contacts between the protein and the oligosaccharide than the "ϩ" sites, representing weaker and less specific binding (Fig. 4c).
The GlcNAc in ϩ2 subsite adopts a similar conformation and keeps the same interactions with Asn 405 , Gly 470 , Asn 437 , and His 202 as GlcNS6S (Fig. 4d). Moreover, Tyr 436 side chain stacks parallel against the GlcNAc sugar ring in the ϩ2 subsite increasing its stability and contributing to the binding. At the ϩ1 subsite the tetrasaccharide contains a GlcA rather than a ⌬UA, present in the product, which is present in a different conformation and shows slightly different binding to the protein. This is in part due to the replacement of Tyr 257 by an alanine in the H202A/Y257A double mutant, which created a void that is occupied by the carboxylate group of GlcA. The GlcA moiety assumes the energetically favorable chair conformation. Indeed, modeling the side chain of Tyr 257 showed that the carboxylic oxygen of GlcA would be only 0.8 Å from the hydroxyl group of Tyr 257 . The carboxylic oxygen atoms in the GlcA of the substrate are hydrogen bonded to NE His-406 and OH Tyr-429 and through a water molecule to NH2 Arg-261 , whereas the carboxylic oxygen atoms of the product bind directly to OE2 Glu-205 , NE His-406 , and NH2 Arg-261 .
As expected, the Ϫ subsites maintain fewer direct interactions with HepII. The GlcNAc in the Ϫ1 subsite binds with its acetyl group pointing toward the opening of the cleft. Only two arginine side chains, Arg 96 and Arg 148 , are involved in direct interactions with GlcNAc in Ϫ1 subsite (Fig. 4c). The NH2 Arg-96 group is hydrogen bound to the O3 GlcNAc and the NH1 Arg-148 to O5 GlcNAc and O6 GlcNAc . The interactions are completed by two water molecules linking O3 GlcNAc to OD2 Asp-309 and the N-acetyl O atom to the N Tyr-436 atom. The GlcA at Ϫ2 subsite forms one direct hydrogen bond between the carboxylic group of GlcA and NH2 Arg-96 , and indirect bonds via water molecules connecting O3 GlcA with NZ Lys-316 and ND2 Asn-259 , and O2 GlcA with O Asp-307 .
The bound tetrasaccharide is bent between the sugar rings bound in the Ϫ subsites and the ϩ subsites (Fig. 4c). A similar bending of the polysaccharide backbone around the glycosidic sessile bond was shown in the enzyme-substrate complex of chondroitin lyase AC from Arthrobacter aurescens (27).
The structures of the representative HS tetrasaccharide (⌬UA-GlcNAc-GlcA-GlcNAc) and heparin or HS disaccharide product (⌬UA-GlcNS6S/GlcNAc) provide the basis for understanding the broad substrate specificity of this enzyme. As mentioned above, sulfation of the ϩ2 GlcA does not affect substrate binding as extrapolated from the ϩ1, ϩ2 subsites occupied by the products (Fig. 4a). The structures also provide an explanation for the inability of HepII to cleave the polysaccharide when GlcNAc in the ϩ2 subsite is sulfated at position 3. The 3-OH group points toward the bottom of the substrate binding cavity and is in proximity to Arg 261 , Asn 405 , and His 406 . The sulfate attached at this position would collide with these side chains preventing proper positioning of the ϩ1 uronic acid for catalysis. The Ϫ subsites are located in the more open part of the HepII binding cleft. Only few interactions were found and substrates seem to have some degree of freedom with and do not discriminate against any specific substitutions. The observed dependence of catalytic efficiency on the polarity of the C2 amino group in the GlcN residue at the Ϫ1 subsite is likely related to subtle effects as it cannot be simply explained by the structure alone.

Molecular Dynamics of HS and Heparin Tetrasaccharides Bound to HepII
Comparison of all the HepII structures we have determined here and previously (17) showed that oligosaccharide binding causes only small changes in the enzyme, localized to the substrate binding site. This behavior parallels what we have observed for chondroitinase AC (27,34) and chondroitinase B (35). Therefore we asked the question: what conformation of the tetrasaccharide is compatible with the structure of the wild type HepII in the ground state? To answer this question for the GlcA-and IdoA-containing substrates we resorted to constrained molecular dynamics (MD) simulations. During the simulations only the substrate molecule and select protein side chains lining the binding site were allowed to move in the explicit solvent. This conservative substrate modeling assured that the protein structure was prevented from opening or distortion of the active site during the simulation and that the substrate had to adapt to a relatively rigid binding site. The starting conformation of tetrasaccharide 3 (⌬UA-GlcNAc-GlcA-GlcNAc) in the WT HepII was modeled based on its structure bound to the (H202A/Y257A) HepII mutant, with GlcA in the ϩ1 subsite. An analog of tetrasaccharide 3 with the GlcA substituted by IdoA was modeled in WT HepII with IdoA in the ϩ1 subsite. The sugar residues in the ϩ1 subsite were built in their respective stable conformations. In case of GlcA the 4 C 1 chair was modeled, whereas for IdoA two conformations were investigated, 1 C 4 chair and 2 S 0 twist-boat. The initial enzyme-substrate binding mode and conformation of sugar units in ϩ2, Ϫ1, and Ϫ2 subsites were taken from the crystal structure of tetrasaccharide 3 bound to the mutated HepII (supplemental data). Each of these models was subjected to 5 ns of constrained molecular dynamics simulation.
The obtained MD models for GlcA-and IdoA-containing tetrasaccharides bound to WT HepII are depicted in Fig. 5, a and b, respectively. In both cases, the modeled position of the uronic acid carboxylate group corresponds to that observed for the bound product (Fig. 5c) as determined previously (17) and in this study. Accordingly, the carboxylate group is in a different position than observed in the tetrasaccharide 3 bound to the HepII(H202A/Y257A) mutant (Fig. 4c) and does not collide with Tyr 257 . Other enzyme-substrate interactions in the models are largely preserved as in the crystal structure of tetrasaccharide 3 bound to the mutant enzyme, particularly in the ϩ2 and Ϫ1 subsites.
In the modeled substrates bound to WT HepII, neutralization of the carboxylate moiety of GlcA or IdoA is predicted to be achieved primarily by three enzyme side chains: Glu 205 , Arg 261 , and His 406 , including a short H-bond to the Glu 205 carboxylate group, similarly to that determined experimentally for the bound product (Fig. 5c). Protonation of the Glu 205 side chain was critical to preserve this mode of uronic acid neutralization during the MD simulations consistent with that observed in all HepII crystal structures, i.e. there was a requirement for a proton sharing between the carboxylate groups of the uronic acid and of the Glu 205 side chain, supporting our proposition for the crucial role of Glu 205 in charge neutralization of the carboxylate. During the MD simulation of the tetrasaccharide substrate, the GlcA sugar ring in the ϩ1 subsite transitioned to a boat conformation (Fig. 5a). This is a less stable conformation than the 4 C 1 chair conformation observed for the GlcA ring in the tetrasaccharide 3 bound in a more relaxed state to the HepII(H202A/Y257A) mutant due to the space created by the Y257A mutation. Apparently, the steric hindrance in the WT enzyme and the complementary interactions for neutralization of the carboxylate group enforced the more strained boat conformation of GlcA. In support for this putative GlcA bound conformation, a binding mode featuring the same boat conformation adopted by GlcA in the ϩ1 subsite was previously observed for a tetrasaccharide substrate bound to chondroitin lyase AC from A. aurescens with no conformational changes in the protein caused by substrate binding (27) (Fig. 5d). The neutralization of the GlcA carboxylate group is also accomplished in a similar fashion for the two enzymes, the role of His 406 from HepII being taken by ChonAC residue His 233 (not related in the primary sequence but occupying the same spatial position in the three-dimensional structure), whereas the combined role of Glu 205 and Arg 261 in HepII is mimicked by a single Asn 183 residue in ChonAC (Fig. 5d). Moreover, the position occupied by Tyr 256 in HepII is also preserved in ChonAC and taken by Tyr 242 .
In the case of the IdoA residue bound in the ϩ1 subsite, both the stable 1 C 4 chair and 2 S 0 twist-boat ring conformations of this sugar allowed carboxylate neutralization modes consistent with that observed in the enzyme-product complex. However, an analysis of MD trajectories indicate that only in the model of the tetrasaccharide with the IdoA resi-due in the twist-boat conformation is the phenolic hydroxyl of the Tyr 257 residue within H-bonding distance from the leaving group oxygen in the Ϫ1 subsite able to act as a general acid (2.9 Å in 2 S 0 versus 3.4 Å in 1 C 4 ). We thus propose that in a catalytically competent binding mode, the IdoA residue will be accommodated in the ϩ1 subsite in a twistboat conformation (Fig. 5, b and c).
The substrate-enzyme ground state models obtained by constrained MD simulation indicate the conformation of the tetrasaccharide substrates compatible with the reasonably rigid substrate binding site of HepII. They are informative for surveying the position of putative catalytic residues around the C5 atom of the uronic acid in the ϩ1 subsite. Comparison of GlcA and IdoA in substrates with ⌬UA in the product shows that the sp 3 C5 atoms of GlcA and IdoA are located on the opposite sides of the plane defined by the substituents of the sp2 C5 atom of the ⌬UA product (Fig. 5c). The H5 proton of GlcA points toward the Tyr 257 hydroxyl group (C5 GlcA -OH Tyr-257 distance of 3.6 Å, Fig. 5a), whereas the H5 proton of IdoA points toward the His 202 ring (C5 IdoA -NE His-202 distance of 3.7 Å) (Fig. 5b). Tyr 257 is within H-bonding to the O1 atom of the residue in the Ϫ1 subsite (O1 GlcNAc -OH Tyr-257 distances of 2.8 and 2.9 Å for GlcA-and IdoA-containing substrates, respectively).

Proposed Catalytic Mechanism
In the context of the structures of HepII product and substrate complexes and MD simulations we propose that Glu 205 -His 406 -Arg 261 is responsible for charge neutralization of the uronic acid. The neutralization of the negative charge on the carboxylic group at position ϩ1 is achieved by formation of a low energy barrier hydrogen bond with a protonated Glu 205 . Glu 205 is assisted by His 406 and Arg 261 , which form hydrogen bonds to the second oxygen atom of the carboxylate group (Fig.  6). This charge neutralization reduces the pK a of the hydrogen bound to C5, priming it for abstraction by a basic side chain during the second step of the ␤-elimination. In GlcA-containing substrate (HS) we propose that Tyr 257 serves a dual function: in a deprotonated form it could serve as the catalytic base abstracting the C5 proton from GlcA (Fig. 6, top). In the MD model the OH Tyr-257 -H5 GlcA distance is ϳ2.6 Å. Following the elimination of the glycosidic bond, Tyr 257 could serve as a general acid, donating a proton to the non-reducing end of the GlcNAc leaving group (OH Tyr-257 -O1 GlcNAc distance of 2.8 Å), restoring the -OH functional group prior to product release. This mechanism is similar to that proposed for chondroitinase AC (27). During degradation of IdoA-containing substrate (heparin), His 202 , located on the opposite side of C5 to Tyr 257 , is proposed to serve as a general base removing the proton from the IdoA epimer (NE His-202 -H5 IdoA ϭ ϳ2.6 Å) (Fig. 6, bottom). Because the intermediates from heparin and HS degradation are the same (after the formation of the planar C4-C5 double bond at the uronic acid) Tyr 257 is proposed to serve as the general acid during both heparin and HS degradation.
The proposed role of His 406 is the neutralization of the uronic acid carboxylate, and not a general base for abstraction of the H5 proton from the GlcA residue. Mutation of His 202 into alanine severely impaired degradation of the IdoA-containing sulfated substrate relative to the GlcA-containing sulfated substrate. Accordingly, the structural model is consistent with a role of His 202 as a general base for abstraction of the H5 proton FIGURE 6. Proposed mechanism of HepII-catalyzed hydrolysis of heparin and HS substrates based on structural and mutagenesis studies. In the HS hydrolysis (top), Tyr 257 is proposed to function both as a general base to abstract the H5 proton from the GlcA bound in the ϩ1 subsite, and as a general acid to protonate the living group in the Ϫ1 subsite. In the heparin hydrolysis (bottom), His 202 is proposed to function as a general base to abstract the H5 proton from the IdoA bound in the ϩ1 subsite, whereas Tyr 257 would fulfill the general acid role to protonate the living group in the Ϫ1 subsite. The side chains of Glu 205 , Arg 261 , and His 406 are proposed to serve to neutralize the carboxylate group of the uronic acid in both cases. Distances associated with double-headed dotted arrows are in Å units.
only from the IdoA residue. In the case of degradation of GlcAcontaining substrates, the structural model is consistent with Tyr 257 fulfilling roles as both general base and acid successively.