Recombinant Expression, Purification, and Biochemical Characterization of Chondroitinase ABC II from Proteus vulgaris*

Chondroitin lyases (or chondroitinases) are a family of enzymes that depolymerize chondroitin sulfate (CS) and dermatan sulfate (DS) galactosaminoglycans, which have gained prominence as important players in central nervous system biology. Two distinct chondroitinase ABC enzymes, cABCI and cABCII, were identified in Proteus vulgaris. Recently, cABCI was cloned, recombinantly expressed, and extensively characterized structurally and biochemically. This study focuses on recombinant expression, purification, biochemical characterization, and understanding the structure-function relationship of cABCII. The biochemical parameters for optimal activity and kinetic parameters associated with processing of various CS and DS substrates were determined. The profile of products formed by action of cABCII on different substrates was compared with product profile of cABCI. A homology-based structural model of cABCII and its complexes with CS oligosaccharides was constructed. This structural model provided molecular insights into the experimentally observed differences in the product profile of cABCII as compared with that of cABCI. The critical active site residues involved in the catalytic activity of cABCII identified based on the structural model were validated using site-directed mutagenesis and kinetic characterization of the mutants. The development of such a contaminant-free cABCII enzyme provides additional tools to decode the biologically important structure-function relationship of CS and DS galactosaminoglycans and offers novel therapeutic strategies for recovery after central nervous system injury.

GalAG-depolymerizing chondroitinases have been classified broadly into three subfamilies. Chondroitinase AC depolymerizes chondroitin-4-sulfate (C4S) and chondroitin-6-sulfate (C6S), whereas chondroitinase B depolymerizes dermatan sulfate as its sole substrate (2)(3)(4)(5)(6)(7)(8)(9). Chondroitinase ABC has the broadest substrate specificity in that it depolymerizes both CS and DS substrates (10 -12). Chondroitinases have been employed in attempts to promote functional locomotor recovery following trauma to the central nervous system (13)(14)(15). The application of cABCI at the site of central nervous system injury is believed to prune CS chains from proteoglycans localized to the glial scar. The absence of these CS chains, inhibitory to axon regeneration, facilitates neural outgrowth and reconstruction of damaged tissue. However, the use of chondroitinases as therapeutics is limited because of the lack of availability of pure and contaminant-free enzyme. Further, chondroitinase enzymes are often difficult to handle, because of thermal instability and spontaneous proteolysis, as reported by various groups (12,16,17).
Chondroitinase AC (cAC) and chondroitinase B from Pedobacter heparinus have been characterized extensively in terms of their enzymatic activity and substrate specificity. The crystal structure and co-crystal structure of chondroitinase B with its DS substrate together with site-directed mutagenesis of its putative active site residues provided detailed insights into its substrate processing and also revealed a calcium-dependent catalytic activity (3,6,8). The co-crystal structures of cAC with different CS and DS oligosaccharide substrate complexes led to the proposal of multiple scenarios in which the active site residues contributed to the catalytic activity of the enzyme (7). The crystal structure of another cAC from Artrhobacter aurescens and its co-crystal structure with CS substrates provided molecular insights into the active site of this enzyme and also its exolytic mode of action compared with the endolytic mode of cAC from P. heparinus (4).
Two distinct broad substrate specificity GalAG-degrading chondroitinase ABC lyases, cABCI and cABCII, have been identified in Proteus vulgaris (12). In fact, the ability of the conventional enzyme known as "chondroitinase ABC" to catalyze the complete depolymerization of GalAG substrates to disaccharides is actually the result of the joint action of cABCI and cABCII. Recently, cABCI from P. vulgaris was cloned, recombinantly expressed, and characterized biochemically in terms of its active site and the role of divalent cations in processing CS and DS substrates (10,18).
Building on our previous efforts, the present study describes the cloning, recombinant expression, and biochemical characterization of cABCII from P. vulgaris. Using an efficient system of Escherichia coli-mediated expression and purification, recombinant cABCII was obtained, and the conditions for its optimal activity were examined. The kinetic parameters and mode of action of the enzyme were characterized using CS and DS substrates. In contrast to cABCI, which is predominantly an endolytic enzyme, cABCII is an exolytic enzyme that cleaves the substrate only at the nonreducing end of the polysaccharide. Furthermore, cABCII was able to efficiently cleave a DS tetrasaccharide that was resistant to cleavage by cABCI (10). The structural basis for the role of the active site residues in enzymatic activity and exolytic processing of cABCII was further elucidated using homology-based structural models of the enzyme-substrate complexes constructed using the cABC I and cAC crystal structures (4,7,11) as templates. The establishment of a contaminant-free recombinant cABCII and detailed characterization of its structure-function relationship enables using this enzyme in a variety of biotechnological applications that include sequencing biologically important GalAG motifs and augmenting therapeutic strategies to reverse central nervous system deficit.

EXPERIMENTAL PROCEDURES
Isolating Chondroitinase ABC II from P. vulgaris-Genomic DNA was isolated from cultures of P. vulgaris (ATCC 6896) using a DNeasy purification kit (Qiagen). The primers were designed based on the available sequence of the gene for both the full-length and mature versions (19). Forward primers were designed so as to incorporate an NdeI restriction site; the reverse primer was designed to incorporate BamHI and XhoI restriction sites. This allowed cloning into a pET-28a vector (Novagen). The primers for cloning cABCII had the sequences: 5Ј-CATATGCTAATAAAAAACCCTTTAGCCC-3Ј (forward primer for the full-length gene), 5Ј-CATATGTTACCCACTC-TGTCTCATGAAGC-3Ј (forward primer for the truncated gene-excluding signal sequence), and 5Ј-GGATCCTCG-AGTTACTTAACTAAATTAATAACAGTAGG-3Ј (reverse primer). It should be noted that for the truncated gene an additional methionine was introduced into the primer sequence to allow for translation of the protein product. This causes an increment in the numbering of the residues by one for the final protein product thus produced. PCR was run using genomic DNA as template with an extension time of 3 min. The PCR product was ligated into the pCR 4-TOPO vector using the TOPO TA cloning kit (Invitrogen) and transformed into TOP10 E. coli cells. Plasmid DNA was isolated, and the cABCII gene was excised by exploiting the NdeI and XhoI restriction sites. The excised gene was ligated into similarly digested pET28a. These ligation products were transformed into DH5␣ E. coli cells. Plasmid DNA isolated from the colonies was screened by restriction digestion for incorporation of the cABCII gene. Sequencing was also undertaken to confirm incorporation of the gene. E. coli cells (BL21(DE3)) were transformed with plasmid DNA for expression.
Protein Preparation-Recombinant cABCII was expressed in E. coli using an adapted version of a previous approach (5,18). The pET28a expression system contains an inducible T7 promoter, as well as an N-terminal six-histidine tag for facile purification. Cultures of Luria-Bertani broth containing kanamycin were inoculated, induced with 1 mM isopropyl-␤-D-thiogalactopyranoside in mid-log phase (A 600 ϭ ϳ0. 8), and incubated at room temperature overnight. Centrifugation was used to harvest the cells, and the supernatant was discarded. The cell pellet was kept on ice and resuspended in 50 mM Tris, 250 mM NaCl, 10 mM imidazole, pH 7.9 (binding buffer), and then lysed by sonication. Soluble protein was collected by centrifugation at 15,000 ϫ g for 15 min at 4°C. The soluble lysate was sequentially filtered through a 0.8-m membrane and then a 0.45-m membrane. A 5-ml Hi-Trap Metal Chelate column (GE Healthcare) was prepared by charging with 200 mM NiSO 4 and treatment with binding buffer. The protein was loaded onto the column, washed with a buffer containing 100 mM Tris, 250 mM NaCl, and 50 mM imidazole, and eluted into a similar buffer with increased imidazole (250 mM). The six-histidine tag was removed using a thrombin capture kit (Novagen) as previously described (20). The presence and purity of the proteins was assessed by standard methods using SDS-polyacrylamide gel electrophoresis. Protein concentration was measured using the Bradford assay (Bio-Rad) with bovine serum albumin (Sigma) as a standard.
Site-directed Mutagenesis-A QuikChange site-directed mutagenesis kit (Stratagene) was used with plasmid DNA template to induce mutations in the cABCII clone. As previously described (18), plasmid denaturation and annealing of customcrafted complementary oligonucleotide primers were used to introduce mutations. The primers were designed as follows (all primers are read in the 5Ј 3 3Ј orientation): the 5Ј I23T primer had the sequence GAA GGT GAA TTA CCC AAT ACC CTT ACC ACT TC and the 3Ј I23T primer had the sequence GA AGT GGT AAG GGT ATT GGG TAA TTC ACC TTC; the 5Ј G247V primer had the sequence TGG GAA AAA TTG GTG TTA ACA CAA CAC GCT GAT GGC and the 3Ј G247V primer had the sequence GCC ATC AGC GTG TTG TGT TAA CAC CAA TTT TTC CCA; the 5Ј R356W primer had the sequence CC AGA GAA CTT TTT GAT GCA TGG TTT ATT GGC CGT C and the 3Ј R356W primer had the sequence G ACG GCC AAT AAA CCA TGC ATC AAA AAG TTC TCT GG; the 5Ј D389N primer had the sequence GGA CGT ATT TTT GAA AAA AAT AAT GAA ATT GTT GAT GCA AAT GTC and the 3Ј D389N primer had the sequence GAC ATT TGC ATC AAC AAT TTC ATT ATT TTT TTC AAA AAT ACG TCC; and the 5Ј H490R primer had the sequence CGC TTA TCT ACT TCA GCA CAT GAG CGT TTA AAA GAT G and the 3Ј H490R primer had the sequence C ATC TTT TAA ACG CTC ATG TGC TGA AGT AGA TAA GCG. Extension of the primers in a PCR machine with Pfu Turbo DNA polymerase followed. The mutated plasmids were transformed into XL1-Blue supercompetent cells. The plasmids were prepared using a Qiagen miniprep kit. Each clone was sequenced to confirm the presence of the desired mutation. Plasmid DNA was used to transform BL21 (DE3) E. coli. In addition to the mutants described above, other mutations in residues believed to be important for enzyme activity were also made. The primer sequences for each of the mutants are listed below. The Composition Analysis of Products from cABCII Processing of CS and DS Substrates-To investigate the composition of the final products of cABCII digestion, capillary electrophoresis was performed as previously described (6). Substrates included C6S from shark cartilage (Sigma), DS from porcine intestinal mucosa (Sigma), and C4S from sturgeon notochord (Seikagaku). Exhaustive overnight digestions of substrate by cABCII were analyzed using a Hewlett Packard three-dimensional capillary electrophoresis instrument with an extended path length cell. A voltage of 30 kV was applied using reverse polarity. Oligosaccharides were injected into the capillary using hydrodynamic pressure. They were detected using an ultraviolet detector set to 232 nm.
Biochemical Characterization of Chondroitinase ABC II Activity-Substrates (C6S and DS) were dissolved at 1 mg/ml concentration in various buffers to determine the relative effects of pH, temperature, ionic strength, and sodium acetate concentration on enzyme activity, as previously described (18). Chondroitin from shark cartilage (Seikagaku), hyaluronan from human umbilical cord (Sigma), heparin (Celsus), heparan sulfate (Celsus), and keratan sulfate (Sigma) were also used in these studies. For activity experiments, 2 l of enzyme was placed in 248 l of 50 mM Tris/HCl, pH 8.0, and reacted with 1 mg/ml substrate at 37°C (0.25 mg/ml for hyaluronan). Product formation was monitored as an increase in absorbance at 232 nm as a function of time in a SpectraMax 190 (Molecular Devices) 96-well quartz format. For kinetic assays, 1 l of enzyme (0.2-1.0 g/l) was added to 249 l of a solution containing a GalAG substrate. Substrate concentration ranged from 0.1 to 5 mg/ml. Product formation was monitored by measuring the absorbance at 232 nm every 2 s. Kinetic parameters were determined using the initial rate of product formation and calculated as previously described based on Michaelis-Menten and Hanes techniques (18).
Modeling the Theoretical cABCII-Substrate Structural Complex-The crystal structure of cABCI from P. vulgaris (Protein Data Bank code 1HN0) was used as a template to obtain the model of cABCII. Initial inspection of the sequence alignment between cABCI and cABCII revealed that cABCII had multiple insertions of large loop regions as compared with cABCI. In addition to cABCI, the crystal structures of distinct cACs (which share the same structural fold with cABCI) from P. heparinus (PhcAC; Protein Data Bank code 1CB8) and from A. aurescens (AacAC; Protein Data Bank code 1RW9) were also used to model the loop regions in cABCII that aligned with either of the cACs. The structural superimposition of cABCI and these distinct cACs was obtained using combinatorial extension-Monte Carlo (CE-MC) multiple structural alignment tool (21) (supplemental Fig. S1). A homology-based structural model of cABCII was generated using the homology module of InsightII v2005 (Accelrys, San Diego, CA). The deletions in the modeled structure were closed by constrained minimization upon holding most of the structure rigid, except for regions close to the deletion site. This was followed by 300 iterations of steepest descent and 400 iterations of conjugate gradient minimization without including charges. The loops and side chains of all of the residues were then allowed to move freely by performing 500 iterations of steepest descent minimization. The refined structure was then subjected to 500 iterations of steepest descent minimization without including charges and 500 iterations of conjugate gradient minimization including charges to obtain the final predicted model of the cABCII enzyme. The final model of cABCII was validated using Whatif web-based interfaces and the Ramachandran plot explorer (supplemental Fig. S2).
A C4S Tetrasaccharide Substrate, (GlcA-GalNAc,4S) 2 , was docked into the putative active site of cABCII using the following approach. The SuperPose version 1.0 server (22) was used to superimpose the co-crystal structure of cAC-CS tetrasaccharide complex (Protein Data Bank code 1HMW) with the modeled cABCII structure. The CS tetrasaccharide in this cAC complex had a uronic acid with a ⌬4,5 unsaturated linkage at the nonreducing end. The starting model of the C4S tetrasaccharide in the cABCII active site was therefore derived from the coordinates of a C4S hexasaccharide (Protein Data Bank code 1C4S), which was superimposed on the CS substrate in the cAC co-crystal structure. The enzyme-substrate complex was subject to minimization without charges with 400 steps of steepest descent and 600 steps of conjugate-gradient methods. This was followed by another 500 steps, each of steepest descent and conjugate-gradient methods with charges. To constrain the ring torsion angles to maintain the ring conformation of the tetrasaccharides during the process of energy minimization, a force constant of 7000 kcal/mol was utilized. To evaluate the exolytic versus endolytic propensity of cABCII in comparison with that of cABCI, an octasaccharide C4S substrate, (GlcA-GalNAc,4S) 4 , was also docked in a similar fashion (described above) into putative active sites of cABCI and cABCII, respectively. The octasaccharide was generated from the coordinates of a C4S hexasaccharide (Protein Data Bank code 1C4S) by adding another, GlcA-GalNAc,4S, to the reducing end of the hexasaccharide.
The viewer, builder, and discover modules of InsightII v2005 (Accelrys) were used for the visualization, structure building, and energy minimization, respectively. The AMBER force field (Amber95) provided with the Discover module was used to assign the potentials for both the enzyme and substrate. The parameters for sulfates and sulfamate groups in glycosaminoglycans described previously (23) were incorporated into this force field to assign potentials for the C4S substrates. A distancedependent dielectric constant of 4*r and scaling of 0.5 for the p1-4 cross terms were used in the discover module for the AMBER force field-based simulations according to the specifications in the InsightII manual.

RESULTS
Cloning, Expression, and Purification of cABCII-Two versions of the cABCII gene were cloned from P. vulgaris DNA: a full-length version and a truncated version that corresponds to the mature form of the enzyme, that is, without its putative leader sequence. These cABCII transcripts are ϳ3 kb in length-rather large, but still at an appropriate size to tolerate amplification via standard polymerase chain reaction techniques. Following cloning into a TOPO vector, the PCR product was subcloned into the pET28a expression vector. This facilitates E. coli-mediated uptake of the transcript and, ultimately, expression of the protein. Chondroitinase ABC II was expressed in E. coli as described under "Experimental Procedures." Purification over a charged Ni 2ϩ resin was possible because of the incorporation of an N-terminal His 6 tag.
As described in our earlier studies on recombinant cABCI (18), we faced similar challenges because of the discrepancies between our cABCII transcript and the sequence reported previously (19). The codons of concern occupy positions 23 (our clone contains isoleucine instead of threonine), 247 (glycine instead of valine), 356 (arginine instead of tryptophan), 389 (aspartic acid instead of asparagine), and 490 (histidine instead of arginine). After unsuccessfully trying out different measures to minimize transcript discrepancies (including the use of various high fidelity polymerases), sequential site-directed mutagenesis was successfully employed to achieve the desired DNA sequence.
Chondroitinase ABC II was expressed as a fusion protein with an N-terminal His 6 tag. Expression was induced in the log phase by the addition of isopropyl-␤-Dthiogalactopyranoside. Purification of cABCII generated in excess of 50 mg of protein/500 ml of culture. SDS/PAGE analysis (Fig. 1A) confirmed the presence of highly pure cABCII at ϳ100 kDa, consistent with previously reported masses of the enzyme (12). Expression of the full-length cABCII clone generated a protein largely present in the insoluble fraction. The yield of soluble enzyme was greatly improved by the engineered removal of the hydrophobic N-terminal signal sequence. We then turned our attention to the truncated clones, both the original sequence and the transcript that underwent our sequential mutagenesis resolution. The recombinant protein with the sequence discrepancies was unable to effectively process GalAG substrates (Fig. 1B). The expression and purification of the modified cABCII gene, on the other hand, generated a protein product with reinvigorated functionality as demonstrated by the observed cleavage of chondroitin-6-sulfate (Fig.  1B).
Biochemical Conditions for Optimal Enzyme Activity-After establishing the active recombinant cABCII, the reaction conditions for optimal cleavage of GalAGs were investigated. These reaction parameters included temperature, pH, ionic strength, and buffer system. For C6S substrate, cABCII demonstrated maximal processing at 37°C, with a greater than 50% decline in activity at 42°C (Fig. 2A). Chondroitinase ABC II similarly acted on DS substrate maximally at 37°C, with a 50% drop at 42°C (Fig. 2A). Activity against both substrates fell dramatically in excess of 45°C. For both C6S and DS substrates, 37°C was chosen as the optimal temperature for biochemical experiments. A Tris buffer system was chosen for biochemical experiments because it permitted greater activity relative to phosphate buffer (data not shown). The recombinant enzyme demonstrated maximal activity at pH 8.0 for C6S. For DS, maximal activity occurred in the range from pH 8.0 to 8.5 (Fig. 2B). Chondroitinase ABC II processing of GalAG substrates is greatly curtailed below pH 7.5 and above pH 8.8. For purposes of simplicity, pH 8.0 was taken as the optimal pH for biochemical studies for both C6S and DS in a Tris buffer system. For cABCI, ionic strength proved to be an important determinant in the processing of GalAG substrates (18). In the case of cABCII, ϳ100 mM NaCl reduced processing by 50% for both C6S and DS substrates (Fig. 2C). Processing of C6S and DS is virtually eliminated in excess of 250 mM NaCl. Furthermore, the addition of ϳ50 mM sodium acetate for C6S and 100 mM sodium acetate for DS activates cABCII-mediated processing (Fig. 2D). Depolymerization of GalAG substrate by cABCII is nearly completely inhibited at 500 mM sodium acetate. Therefore, unlike cABCI, which required ϳ50 mM of NaCl for C6S processing (and 125 mM for DS) (18), for cABCII the presence of NaCl in the buffer actually decreases the activity.
Galactosaminoglycan Processing by cABCII-The specific activity of the recombinant cABCII was tested against a full panel of glycosaminoglycan substrates ( Table 1). The results suggest that cABCII is most proficient in degrading C6S, C4S, and DS substrates. It was not possible to detect reaction progression for hyaluronan, possibly because of low cleavage rate and nonoptimal pH for this substrate. As expected, other glycosaminoglycan families, like heparin and heparan sulfate (which contain glucosamine instead of galactosamine), were not processed by cABCII. Comparison of cABCII activity with that of cABCI (18) shows that cABCI more efficiently processes CS (ϳ10-fold) and DS (more than 7-fold) than cABCII. Kinetic parameters for recombinant cABCII acting on C6S, DS, and C4S are summarized in Table 2. These measurements were obtained from the initial reaction rates against each substrate using both Michaelis-Menten and Hanes analysis. The turnover numbers for cABCII were experimentally determined to be 1300, 1150, and 1000 min Ϫ1 for C6S, DS, and C4S, respectively. The catalytic efficiency was highest against C6S substrate (132 M Ϫ1 min Ϫ1 ) and comparable for DS and C4S (both 60 M Ϫ1 min Ϫ1 ).
The depolymerization of GalAG substrates by cABCII was further scrutinized with capillary electrophoresis. These studies allow for the characterization of the final products of cABCII digestion following a 20-h exhaustive reaction at 37°C and therefore represent an end point assay for cABCII activity. For all of the substrates examined, the product profile contains an overwhelming proportion of disaccharide products (Fig. 3). For C6S, the dominant product is a 6-O-sulfated disaccharide. For C4S, the major product is a 4-O-sulfated disaccharide. With DS, a mixture of disaccharide products includes two monosulfated species, the 4-O-sulfated DS disaccharide and the 6-Osulfated DS disaccharide, and a doubly sulfated disaccharide, the 4-O-and 6-O-di-sulfated disaccharide. It should also be noted that the amount of disaccharide released in each of these end point assays is considerably less (as measured by the double bonds generated, i.e. A 232 nm ) when compared with those released by cABCI processing of these substrates. Comparison of   the processing of DS by cABCI and cABCII (Fig. 3C, inset) indicated that although cABCI processing resulted in a resistant tetrasaccharide ⌬UA-GalNAc,4S-⌱doA-GalNAc,4S (10), cABCII action on DS substrates forms only disaccharide products. Structural Investigation of the Active Site Residues and Mode of Action of cABCII-The crystal structures of cABCI, PhcAC, and AacAC were used as templates to obtain a homology-based structural model of cABCII and its complex with the substrate (Fig. 4). The resulting model therefore adopted a multi-domain structure (similar to that of cABCI and the cACs) comprising an N-terminal ␤-domain with a jellyroll fold, the catalytic ␣-helix domain (incomplete toroid (␣/␣)5-fold), and a C-terminal antiparallel ␤-sheet domain. To contrast the active site and positioning of the substrate within cABCII, the obtained model was superimposed on the enzyme-substrate co-crystal structures of the cAC enzymes (see "Experimental Procedures"). This superimposition showed that the critical active site residues were conserved and aligned spatially (supplemental Fig. S1). Comparison of cABC structures with that of cACs showed that the substrate binding and the active site of the cABCs formed a wider groove to accommodate a broad range of CS and DS substrates. The notable differences in the structures of cABCI and cABCII were in two specific loop regions Ile 224 -Thr 231 and Gly 914 -Leu 937 that were present in cABCII and not cABCI (supplemental Fig. S1). The presence of extra loop regions in AacAC compared with that of PhcAC has been implicated to account for the predominant exolytic activity of AacAC as against the endolytic activity of PhcAC (4). To better understand the role of the loop regions in cABCII in substrate processing, an octasaccharide of C4S was docked into the putative active site of cABCII. Assuming that the GalNAc,4S-GlcA cleavable linkage occupied subsites Ϫ1 and ϩ 1, respectively, the sugars on the nonreducing end of the cleavable linkage occupied Ϫ2, Ϫ3, and Ϫ4 positions, whereas the sugars on the reducing end occupied ϩ2, ϩ3, and ϩ4 positions.
In the docked model of the cABCII-octasaccharide complex, the sugars at the nonreducing end Ϫ3 and Ϫ4 positions had unfavorable steric contacts with the Ile 224 -Thr 231 loop region (Fig. 5). Also, it appeared that for a long oligosaccharide (octa and higher) to be accommodated into the active site of cABCII, the chain would have to bend out of the active site groove. The cABCI active site groove, on the other hand, is open at both reducing and nonreducing ends of the substrate and hence is readily able to accommodate longer oligosaccharide substrates in the active site. Together the above observations offer an explanation based on the structural model for the ability of cABCII to process smaller oligosaccharides such as tetrasaccharides more efficiently than cABCI. Further, these observations also support the notion that cABCII processes its substrates in a predominantly exolytic fashion starting from the nonreducing end.
To understand the interactions of cABCII with its substrates, a C4S tetrasaccharide was docked into the putative active site of cABCII (see "Experimental Procedures"). It has been shown previously that a tetrad of residues including Arg, His, Tyr, and Glu play a critical role in the lyase activity of cABCI and the cACs. The analogous tetrad of residues in cABCII (that struc-  oxygen, red; nitrogen, blue) superimposed on the cABCI template (C␣ trace colored pink), which was used in the homology modeling. The active site groove of the enzyme is zoomed in for clarity. The distinct additional loop regions in the active site groove of cABCII, Ile 224 -Thr 231 and Gly 914 -Leu 937 , are colored in brick red. The critical active site tetrad residues that structurally coincide are labeled using their numbering in cABCII and the single alphabet code for clarity (Arg and His are colored blue, Glu is colored red, and Tyr is colored magenta). This figure was generated using PyMol software.
turally aligned with cABCI and cACs) are Arg 513 , His 453 , Tyr 460 , and Glu 608 (Fig. 6). Although Glu 608 does appear to be directly involved in the contact with the sugar, it is positioned to make critical hydrogen bonding interactions with His 453 and Arg 513 , thus positioning these residues for their activity (Fig. 6). In addition to these critical active site residues, there are several other residues that are proximal to the CS substrate and hence could play a key role in positioning the substrate for enzyme action. The residue His 452 proximal to His 453 is on the opposite side of Arg 513 in the base of the groove and is positioned to interact with carboxyl group of the GlcA in the cleavage site. Residues His 343 and Tyr 346 on the top side of the grove (opposite to the catalytic tetrad at the base) are positioned to interact with GlcA and GalNAc,4S in the ϩ1 and ϩ2 subsites and hence could play a critical role in anchoring the substrate in the active site. The residues His 514 and His 519 are proximal to the 4-sulfate group of the GlcNAc4S residue at the Ϫ1 subsite.
Site-directed Mutagenesis and Kinetic Characterization of the Putative Active Site Residues-Based on the above structural model of cABCII-CS tetrasaccharide complex, a putative catalytic tetrad, His 453 , Tyr 460 , Arg 513 , and Glu 608 , and the other residues in the active site that are positioned to interact with the substrate such as His 343 , His 452 , and His 456 were identified. To probe the contribution of these residues to enzymatic activity, they were mutated to alanines, and the resulting enzyme products were assayed for activity on chondrotin-6-sulfate and dermatan sulfate substrates ( Table 3). The results show that mutants H453A, Y460A, R513A, and E608A all show no detectable activity on either C6S or DS, consistent with their designation as the critical catalytic tetrad required for enzyme activity. H452A shows similar K m values to the wild type enzyme but a greatly (100-fold) reduced catalytic efficiency. On the other hand, mutation of His 456 to alanine has limited effect on overall catalytic efficiency (ϳ3-fold reduction), indicating that this residue likely does not play an important role in catalysis. Surprisingly, mutation of His 343 to alanine also yields an enzyme that shows no detectable activity on C6S and DS, suggesting that His 343 plays a critical role in the enzymatic activity in addition to the catalytic tetrad.

DISCUSSION
This report is the first to describe the expression and characterization of a stable, highly active, contaminant-free recombinant chondroitinase ABC II from P. vulgaris. The sequence anomalies in the original clone resulted in a catalytically inactive enzyme, although these anomalies were not in the critical active site residues (Fig. 1B). It is therefore likely for these sequence differences to affect the overall stability of the enzyme, which in turn results in an inactive enzyme. Fixing these anomalies resulted in a fully active recombinant enzyme that efficiently processed both CS and DS substrates. This recombinant cABCII was examined structurally and biochemically, including reaction conditions to maximize enzyme efficacy, the product profile following digestion of GalAG substrates, kinetic analysis, substrate specificity, mode of action analysis, and structural insights into the active site-substrate interactions. Comparison of substrate processing of cABCII with that of cABCI showed that cABCI does cleave GalAG substrates at superior rates to cABCII; however, depolymerization mediated by cABCII proceeds by a course distinct from cABCI. Chondroitinase ABC I seems to prefer longer chain substrates  ) and cABCII (bottom in gray) with a C4S octasaccharide (shown as a stick model) docked into their putative active sites (generated using PyMol). Shown on the left and right is the active site groove seen from the nonreducing and reducing end of the octasaccharide respectively. Note that the octasaccharide is readily accommodated in the active site of cABCI. On the other hand, the access for the octasaccharide appears to be constricted toward the nonreducing end in the active site groove of cABCII.

TABLE 3 Kinetic analysis of chondroitinase ABC II site-directed mutants with different substrates
The values, where measurable, are the means of at least three experiments Ϯ standard deviation. NDA, no detectable activity. MDA, minimally detectable activity (activity was too low to allow for reliable determination of kinetic parameters). and cleaves in a predominantly endolytic fashion (10,12). Chondroitinase ABC II, on the other hand, appears to cleave shorter oligosaccharide substrates more efficiently in a predominantly exolytic fashion, resulting only in disaccharide products. The presence of these distinct cABCs offers the bacteria the ability to rapidly process GalAG substrates where cABCI would cleave the naturally occurring long GalAG polysaccharides and cABCII would then act on the smaller oli-gosaccharide fragments from cABCI processing to generate disaccharides that can be readily utilized for bacterial metabolism. The proposed exolytic mechanism of cABCII and its ability to efficiently process smaller GalAG substrates is also supported by the structural model of the cABCII-C4S octasaccharide complex. In this model the two extra loop regions, Ile 224 -Thr 231 and Gly 914 -Leu 937 , in cAB-CII appear to constrict the active site groove specifically on the nonreducing side of the substrate. This groove constriction restricts the access of internal cleavable linkages in a long GalAG chain and thus points to a predominant exolytic mode of action for cABCII. The structural basis for differences in the activity of cABCI and cABCII is similar to the framework proposed for endolytic action of PhcAC versus exolytic action of AacAC (4,9).
A combination of the theoretical structural model of the enzyme-C4S tetrasaccharide substrate complex and site-directed mutagenesis enabled the identification of critical residues involved in the enzymatic activity. This combined analysis further suggests the likely role of these active site residues in catalytic action of the enzyme. A tetrad of residues including His 453 , Tyr 460 , Arg 513 , and Glu 608 in cABCII was structurally conserved in cABCI and the cACs. The corresponding tetrad in cABCI has been shown to be critical for enzyme activity (10). The mutation of any of these four critical residues to Ala resulted in a completely inactive enzyme (Table  3), which strongly supports the central role of this tetrad in the catalytic activity of cABCII. The proximity of His 453 to the C5 proton of the GlcA in the structural model suggests the role of His 453 as a general base for abstraction of the C5 proton. The proximity of Tyr 460 to the glycosidic linkage between GalNAc,4S-GlcA makes it an ideal candidate for protonating the GalNAc sugar after cleavage. The neutralization of the carboxylate group to increase the lability of the C5 proton can potentially be accomplished by the Arg 513 residue. Furthermore, Arg 513 , His 452 , and His 453 are all equally likely candidates for the stabilization of the uronic acid C5 carbanion intermediate formed after proton abstraction. As men- . Critical residues involved in catalytic action of cABCII. A shows stereo view of a C4S tetrasaccharide (carbon, cyan; oxygen, red; nitrogen, blue; sulfur, yellow) substrate docked into the active site groove of the structural model of cABCII. The groove is shown as a cartoon model generated using PyMol, and the residues that are positioned to interact with the substrate are also shown in the following colors: His, Arg, Lys (blue); Tyr (purple); and Asp, Glu (red). B shows a two-dimensional schematic of the chemical structure of C4S tetrasaccharide and its interactions with the critical residues in the active site. The sugars are numbered Ϫ2, Ϫ1, ϩ1, and ϩ2 from nonreducing to reducing end of the C4S substrate where the cleavage occurs between GalNAc,4S and GlcA in the Ϫ1 and ϩ1 sites, respectively. The tetrad of residues His 453 , Tyr 460 , Arg 513 , and Glu 605 are conserved in cABCI, PhcAC, and AacAC and have been shown to play a critical role in the catalytic activity of these enzymes. tioned earlier, Glu 608 is not directly involved in interactions with the substrate, but it is positioned to interact with both His 453 and Arg 513 via hydrogen bonding. Hence mutation of Glu 608 to Ala disrupts these interactions and is therefore unfavorable for the optimal positioning of these critical residues for catalytic activity. Earlier studies (2,10,24) implicate distinct interactions of cABCI with CS and DS substrates. The same scenario holds good for cABCII because of the structurally conserved active site tetrad. Therefore, the proposed roles of the tetrad in catalytic activity of cABCII could potentially be interchanged to accommodate a broad range of CS and DS substrates.
In addition to the tetrad, it was surprising to note that the H343A mutant completely lost the catalytic activity toward both CS and DS substrates ( Table 3). This residue is positioned on the top side of the active site groove opposite the tetrad, which is at the base of the groove (Fig. 6). In the structural model of cABCII described in this study, His 343 is not as proximal as the tetrad to the substrate. However, the model developed in this study is based primarily on the uncomplexed cABCI enzyme. Therefore, it is possible that in the presence of the substrate the active site groove could become more "closed," which would position His 343 proximal to the C5 atom of the GlcA where it could play a critical role in either neutralizing the carboxylate group or stabilizing the C5 carbanion transition state.
This understanding of the structure and mechanism of action of cABCII extends our understanding of precisely how these lyases function and how various structural features contribute to the depolymerization process. The distinct substrate processing ability of cABCII enables its use as an additional valuable resource in technologies directed at determining the fine structural elements of biologically relevant GalAGs. These enzymes may further be useful directly in strategies to interfere with GalAG function in vivo, for example, in neural regeneration therapeutics and other such biomedical applications.