High-Resolution Crystal Structure of Arthrobacter Aurescens Chondroitin AC Lyase: an Enzyme–Substrate Complex Deﬁnes the Catalytic Mechanism

doi:10.1016/j.jmb.2003.12.071 J. Mol. Biol. (2004) 337, 367–386

High-resolution Crystal Structure of Arthrobacter aurescens Chondroitin AC Lyase: An Enzyme–Substrate Complex Deﬁnes the Catalytic Mechanism

Vladimir V. Lunin1, Yunge Li1, Robert J. Linhardt2, Hirofumi Miyazono3 Mamoru Kyogashima3, Takuji Kaneko3, Alexander W. Bell4 and Miroslaw Cygler1*

1Biotechnology Research Chondroitin lyases (EC 4.2.2.4 and EC 4.2.2.5) are glycosaminoglycan- Institute, National Research degrading enzymes that act as eliminases. Chondroitin lyase AC from Council of Canada, and Arthrobacter aurescens (ArthroAC) is known to act on chondroitin 4-sulfate Montreál Joint Centre for and chondroitin 6-sulfate but not on dermatan sulfate. Like other chon- Structural Biology, Montreál droitin AC lyases, it is capable of cleaving hyaluronan. Que´bec, 6100 Royalmount Ave. We have determined the three-dimensional crystal structure of Montreál, Que´bec, Canada ArthroAC in its native form as well as in complex with its substrates H4P 2R2 (chondroitin 4-sulfate tetrasaccharide, CStetra and hyaluronan tetrasaccharide) at resolution varying from 1.25 A˚ to 1.9 A˚ . The primary sequence of 2Department of Chemistry ArthroAC has not been previously determined but it was possible to Division of Medicinal determine the amino acid sequence of this enzyme from the high- Chemistry and Department of resolution electron density maps and to confirm it by mass spectrometry. Chemical and Biochemical The enzyme–substrate complexes were obtained by soaking the substrate Engineering, The University of into the crystals for varying lengths of time (30 seconds to ten hours) and Iowa, 115 S. Grand Ave, PHAR flash-cooling the crystals. The electron density map for crystals soaked in S328, Iowa City, IA the substrate for as short as 30 seconds showed the substrate clearly and 52242-1112, USA indicated that the ring of central glucuronic acid assumes a distorted 3Central Research Laboratories boat conformation. This structure strongly supports the lytic mechanism Seikagaku Corporation, Tateno where Tyr242 acts as a general base that abstracts the proton from the C5 3-1253, Higashiyamato-shi position of glucuronic acid while Asn183 and His233 neutralize the charge Tokyo 207-0021, Japan on the glucuronate acidic group. Comparison of this structure with that of

4 chondroitinase AC from Flavobacterium heparinum (FlavoAC) provides an Montreál Proteomics Network explanation for the exolytic and endolytic mode of action of ArthroAC 740 Dr Penfield Ave., Montreál and FlavoAC, respectively. Que´bec, Canada H3A 1A4 Crown Copyright q 2004 Published by Elsevier Ltd. All rights reserved. Keywords: chondroitin lyase; chondroitinase AC; Arthrobacter aurescens; *Corresponding author substrate binding; catalytic mechanism

Introduction ides, composed of disaccharide repeating units of a substituted glucosamine or galactosamine Glycosaminoglycans (GAGs) are the carbo- attached through (1,4) linkage to a uronic acid hydrate components of proteoglycans, which are molecule. These disaccharide units are linked (1,3) a major component of the extracellular matrix.1 or (1,4) into a polysaccharide chain.2 The glucosa- They are highly negatively charged polysacchar- mine/galactosamine units are sulfated extensively, and their synthesis requires the concerted action of a large number of enzymes.3,4 Supplementary data associated with this article can be Glycosaminoglycans are degraded enzymatically found at doi: 10.1016/j.jmb.2003.12.071 5 Abbreviations used: GAG, glycosaminoglycan; MS, by two types of enzymes, hydrolases and lyases. mass spectrometry. Hydrolases catalyze cleavage of the glycosyl- E-mail address of the corresponding author: oxygen bond by addition of water, producing [email protected] a saturated disaccharide. Lyases cleave the

0022-2836/$ - see front matter Crown Copyright q 2004 Published by Elsevier Ltd. All rights reserved. 368 Chondroitinase AC Crystal Structure and Mechanism oxygen–aglycone linkage through proton abstrac- as the complexes with chondroitin tetrasaccharide tion, producing an unsaturated disaccharide pro- (DUAp (1 ! 3)-b-D-GalpNAc4S (1 ! 4)-b-D-GlcAp duct with a double bond between C4 and C5. The (1 ! 3)-a,b-D-GalpNAc4S, where DUAp is the enzymatic mechanisms of hydrolases are well unsaturated sugar residue, 4-deoxy-a-L-threo-hex- understood and reactions proceed either according 4-enopyranosyluronic acid; GlcAp, glucopyrano- to the retaining or inverting mechanism.6 On the syluronic acid; GalpN, 2-deoxy-2-aminogalacto- other hand, the molecular details of the enzymatic pyranose; S, sulfate; and Ac, acetate) and mechanism of GAG lyases are still poorly under- hyaluronan tetrasaccharide substrates. stood. A chemically plausible mechanism for the b Despite the purification and characterization of elimination reaction has been proposed;7 however, ArthroAC many years ago13,17,18 and its extensive the constitution of the active site and the roles of use as an analytical tool in glycosaminoglycan individual amino acids are not clear. A number of analysis, this enzyme has not been cloned and its bacterial species synthesize GAG lyases, enzymes amino acid sequence has not been determined, used to degrade and utilize glycosaminoglycans although its amino acid composition and carbo- as a source of carbon in the bacterium’s natural hydrate content were reported.18 We obtained environment.5,8 Polysaccharide lyases with known crystals that diffract up to 1.25 A˚ resolution. The three-dimensional structures fall into two architec- high-resolution data led to high-quality electron tures: the right-handed parallel b-helix (pectate/ density maps of native enzyme and several com- pectin lyases, chondroitinase B, rhamnoglucuronan plexes, and allowed us to deduce confidently the lyase) and (a/a)n toroid (n ¼ 5 for Flavobacterium amino acid sequence for 99% of the amino acid heparinum chondroitin AC and chondroitin ABC residues and to propose the molecular details of lyases, bacterial hyaluronate lyases, xanthan lyase, the catalytic mechanism, which is common to and n ¼ 6 for alginate lyases). A catalytic mechan- FlavoAC and hyaluronate lyases. Here, we follow ism has been proposed for pectate lyases, Ca2þ- the nomenclature introduced by Davies et al.19 and dependent enzymes,9 but it remains to be seen if it designate the sugars on the reducing end of the applies to other lyases having the b-helix topology. break with a þ sign and the sugars on the non- Several plausible mechanisms have been proposed reducing end with a 2 sign. In this nomenclature, for the lyases with the (a/a)5 toroidal fold with the enzymes break the bond between sugars 21 a histidine or a tyrosine residue in the role of a and þ1, the latter being an uronic acid. general base abstracting the proton from the C5 atom of glucuronic acid, and a tyrosine or an arginine residue acting as a general acid donating a pro- Results and Discussion ton to the bridging O4 atom.10,11 However, there is insufficient evidence to indicate which of these Amino acid sequence and its conservation proposed mechanisms is utilized by the enzymes. The nature of the group presumed to be necessary The molecular mass of the entire molecule was to neutralize the charge of the glucuronic acid measured by ion spray mass spectrometry. Two carboxylic group is not clear, since these enzymes species were present with molecular masses of do not require Ca2þ and there is no positively 79,502 Da and 79,840 Da. The greater mass corre- charged group in the vicinity of the uronic acid, as sponds well to the molecular mass of 79,785 Da expected from the accepted chemical mechanism.7 (average mass) calculated from the amino acid Glycosaminoglycan-degrading enzymes with sequence. Based on the analysis of MS/MS spectra, defined specificity have found widespread appli- we believe that the smaller mass corresponds to cations as analytical tools for the analysis of the the fragment missing three N-terminal amino acid structure of glycosaminoglycans and other poly- residues. Such a good agreement between the pre- saccharides.12 Chondroitin AC lyases are used fre- dicted and measured molecular mass indicated quently for this purpose. These enzymes cleave that (1) no major assignment errors were made the glycosidic bond on the non-reducing end of an and (2) no glycosylation or other modification uronic acid and use as a substrate either chondroi- were present (FlavoAC is glycosylated).16 tin 4-sulfate or chondroitin 6-sulfate but not Peptides extracted from an in-gel trypsin digest dermatan sulfate. They display a varied degree of purified A. aurescens chondroitin AC lyase were of activity toward hyaluronan.13 Enzymes from analyzed by LC-QToF mass spectrometry as two sources, chondroitin AC lyase from Arthrobacter described in Materials and Methods. The resulting aurescens (ArthroAC) and from F. heparinum peaklist of fragmentation spectra was matched in- (FlavoAC), are commercially available (Seikagaku house against the sequence deduced from the crys- Corporation) and used frequently. The latter tal structure employing Mascot (MatrixScience) enzyme has been cloned and overexpressed in software.20 Matched tandem MS spectra were con- F. heparinum14 and in Escherichia coli.15 We pre- firmed manually and the remaining spectra were viously determined the three-dimensional struc- interpreted manually. This process was repeated ture of this enzyme on its own16 and in complex several times, using the two sets of experimental with several dermatan sulfate oligosaccharides.10 data iteratively to determine the optimal sequence Here, we present the three-dimensional structure of the lyase (Supplementary Material). of chondroitin AC lyase from A. aurescens as well The results of MS/MS analysis of 202 tandem Chondroitinase AC Crystal Structure and Mechanism 369 mass spectra covering 88.5% of the entire lyase 1F1S22 and Bacillus sp. xanthan lyase 1J0M.23 ArthroAC sequence (see Materials and Methods) Structure-based alignment of their sequences is confirmed the amino acid sequence identified shown in Figure 1. The residues important for the from electron density maps. Moreover, this analy- integrity of the active site (see below) include sis of individual MS/MS fragmentation data Asn183, His233, Tyr242, Arg296 and Glu407 of allowed us to make side-chain assignments for ArthroAC and are conserved in all of the related several residues located in flexible loops for which enzymes, and Glu412 is replaced by an aspartate electron density was not easily interpretable residue in one case. (shown in small letters in Figure 1). The enzyme chondroitin ABC lyase I shows The agreement for 18 (4–21) residues between good sequence similarity to the above-mentioned the residue type derived from the electron density enzymes only for the C-terminal domain. How- maps and mass spectrometry, and that determined ever, its three-dimensional structure (PDB code by Edman degradation provides an independent 1HN0) showed that the catalytic domain has (a/a)5 measure underscoring the low level of errors to be topology and can be structurally aligned with the expected in our assignment. While the sequence of catalytic domains of the other lyases.24 While the ArthroAC has not been determined previously, its substrate-binding site shows little sequence con- amino acid composition was reported nearly 30 servation, the active-site residues are conserved, years ago.18 These data showed good correlation suggesting the same enzymatic mechanism. There with the amino acid sequence of ArthroAC in this is no equivalent, however, to the Asn183 of work, supporting our assignments (Table 1). ArthroAC, and there are differences in the local The derived sequence of ArthroAC was used to structure in this region. Huang et al. suggested identify homologous sequences in the NCBI data- that this enzyme utilizes as a replacement an argin- base with the program BLAST.21 There are ,50 ine residue remote in the linear sequence.24 The such sequences for which the similarity extends structure-based alignment of chondroitin ABC along the entire protein. They include hyaluronate, lyase I with the other lyases is included in Figure 1. xanthan and chondroitin AC lyases. ArthroAC shows the highest level of sequence identity with Overall fold various hyaluronan and xanthan lyases (38%) and a lower level of identity with chondroitin AC The ArthroAC molecule has an overall a þ b lyase from F. heparinum (24%). architecture and consists of two domains (Figure 2). The structures of four proteins representative of The N-terminal a-helical domain contains 13 these sequences are known; namely, chondroitin AC a-helices, ten of which form an incomplete 16 lyase 1CB8, Streptococcus pneumoniae hyaluronate double-layered (a/a)5 toroid as classified within lyase 1EGU,11 Streptococcus agalactiae hyaluronate the SCOP database.25 There is a long, deep groove on one side of the toroid that forms the location of the active site and substrate-binding site. Three a-helices at the N terminus precede the (a/a) Table 1. Comparison of percentage distribution of amino 5 acids between chemical amino acid analysis and crystal- toroid and constrict the cleft on one side. Residues lographic assignment conserved across the sequences of proteins homologous to ArthroAC cluster in the area of this cleft. Residue Sequenced from the map From Hiyama & 18 The C-terminal domain is composed almost type (%) Okada entirely of antiparallel b-strands arranged into Ala 14.1 13.4 four b-sheets. The first two sheets contain nine Arg 5.2 4.7 b-strands, some of them rather long. The third Asx 9.1 9.0 sheet has seven b-strands and the last one five Glx 6.35 6.5 Trp 2.5 2.4 b-strands. There is only one short a-helix within Val 7.0 8.2 this domain (Figure 2). The second domain can be Ser 6.7 6.6 subdivided into two subdomains; the first encom- Thr 8.85 8.4 passes the first two large b-sheets and one short His 1.85 1.7 a-helix, while the second subdomain is composed Phe 2.9 2.9 Gly 10.6 11.9 of the third and fourth b-sheet. Pro 4.0 3.3 Ile 3.6 3.3 Substrate-binding site Leu 9.1 8.8 Lys 3.3 3.3 Tyr 2.5 2.6 Initial experiments of soaking native crystals Met 1.45 1.1 of ArthroAC in a 5 mM solution of chondroitin Cysa 0.9 2.3 4-sulfate tetrasaccharide substrate for prolonged a The significant difference in the number of cysteine residues times before collecting diffraction data showed between our data and those reported by Hiyama & Okada while clear density for only a disaccharide product, indi- there is very good agreement for the other amino acids, is likely cating that, like other GAG lyases,10 ArthroAC related to the fact that hydrolysis was used for the determi- retains enzymatic activity in the crystals. Therefore, nation of most amino acids while sulfhydryl titration was used we decided to investigate by X-ray diffraction for cysteine. the enzyme–substrate complex as a function of Figure 1 (legend opposite) Chondroitinase AC Crystal Structure and Mechanism 371

Figure 2. Stereo drawing of the ribbon representation of ArthroAC showing the bound tetrasaccharide. The individual a-helical hairpins of the N-terminal (a/a)5 toroid are in different colors. The individual b-sheets of the C-terminal domain are in discrete colors. Insertions in ArthroAC blocking the cleft are in gray. The substrate is shown in stick representation and colored in magenta. The Figure was prepared with the programs MOLSCRIPT52 and Raster3d.53 soaking time. The soaking time ranged from 30 22) subsites suggests a partial presence of the seconds to ten hours (Tables 2 and 3). At the end disaccharide product in the 2 sites. For the ten of each soak, the crystal was immediately flash- hour soak, the sugar units in subsites (21, 22) are frozen and diffraction data collected (resolution clearly visible in the difference electron density varying between 1.25 A˚ and 1.6 A˚ , Table 3). Hya- map and refine with an occupancy of 1.0, while luronan tetrasaccharide was also used as a sub- the derived occupancy for the sugars at the (þ1, strate with a soaking time of two minutes and þ2) subsites were 0.25. In the complex of diffraction data were collected to 1.9 A˚ resolution. ArthroAC with hyaluronan tetrasaccharide, only The structures were refined independently. In the sugar units in subsites (21, 22) were visible each case, the ArthroAC molecule was refined in the electron density, corresponding to a disac- first, then the difference electron density map was charide reaction product and in accord with higher inspected and interpreted appropriately. The activity of ArthroAC toward hyaluronan.13 The modeled substrate/product was included in the location and orientation of the hyaluronan sugars refinement. Several sugar units were clearly visible is the same as the corresponding sugars of the in each refined model. The relative occupancy of chondroitin sulfate tetrasaccharide substrate. In the 2 and þ sites along the timed snapshots were the following discussion, we refer to the ArthroAC– evaluated as described in Materials and Methods. tetrasaccharide complex after a 30 seconds soak. The reaction in the present crystals occurs on the The oligosaccharide is bound within the groove minute timescale, indicated by well-defined elec- in the N-terminal domain (Figure 2) and makes tron density for the entire tetrasaccharide substrate contacts with residues Asn124, Trp125, Trp126, after 30 seconds and even longer soaks (Figure 3(a) Arg134, Gln169, Arg174, Asn183, His233, Tyr242, and (b); and Table 4). Indeed, the substrate is best Arg296, Arg300, Asn303, Asn410 and Trp465 defined in this dataset, with assigned occupancies (Figure 3(c)). Tryptophan residues play an essential of ,0.6. Somewhat higher occupancy for the (21, role in substrate binding. Two of these residues,

Figure 1. Structure-based sequence alignment for ArthroAC, FlavoAC (1CB8), S. pneumoniae hyaluronate lyase (1EGU), S. alagalactiae hyaluronate lyase (1F1S), Bacillus sp. xanthan lyase (1J0M) and P. vulgaris chondroitin ABC lyase I (1HN0). Insertions in the ArthroAC sequence that close off one end of the substrate-binding site are boxed, a-helices are marked in white letters on black background and b-stands are marked by black letters on gray background. The secondary structure assignments follow sPDBv.51 Residues conserved in all six proteins are marked by an asterisk (*) above the sequence. Arrows mark residues essential for catalysis, Asn183, His233, Tyr242 and Arg296 and Glu407. Several residues that were assigned on the basis of the MS/MS data alone (disordered side-chain) are shown in small letters. Letter ‘x’ indicates (Glu/Gln) or (Asp/Asn), which we cannot distinguish, and question marks (?) indicate residues for which we are less certain of their amino acid type. 372 Chondroitinase AC Crystal Structure and Mechanism

Table 2. Data collection statistics for various substrate soaking times Substrate Soaking time CStetra HAtetra Two Ten Two 30 seconds minutes minutes 35 minutes Two hours Four hours Ten hours minutes

Wavelength (A˚ ) 0.9798 0.9798 0.9798 0.9798 0.9798 0.9798 0.9798 0.9798 a (A˚ 57.9 57.6 57.7 57.6 57.6 57.6 57.6 57.6 b (A˚ ) 86.9 86.5 86.4 86.4 86.5 86.5 86.3 86.3 c (A˚ ) 81.5 80.7 80.6 80.5 80.6 80.6 80.5 80.6 b (deg.) 107.0 106.8 106.9 106.9 106.9 106.9 107.0 106.9 Resolution range 50–1.41 50–1.6 50–1.5 50–1.35 50–1.3 50–1.35 50–1.25 50–1.9 (last shell) (1.46–1.41) (1.66–1.6) (1.55–1.5) (1.4–1.35) (1.35–1.3) (1.4–1.35) (1.29–1.25) (1.97–1.9) Rsym (last shell) 0.081 0.077 0.077 0.055 0.057 0.059 0.058 0.075 (0.794) (0.739) (0.483) (0.567) (0.517) (0.530) (0.430) (0.652) Completeness (%) 99.9 (99.9) 100 (100) 99.9 (99.9) 98.3 (96.3) 96.5 (89.1) 96.1 (93.5) 90.8 (56.4) 100 (99.9) (last shell) I=sðIÞ (last shell) 8.1 (2.0) 8.3 (2.5) 9.2 (3.4) 8.4 (2.0) 11.1 (3.5) 8.9 (2.9) 9.8 (2.8) 7.7 (2.8) Total reﬂections 706,542 501,855 460,480 452,905 1,338,199 610,947 939,404 235,444 Unique reﬂections 146,507 100,158 120,769 162,044 178,855 158,853 188,554 60,174 Redundancy 4.8 5.0 3.8 2.8 7.5 3.8 5.0 3.9

Table 3. Reﬁnement statistics Model Hg CStetra CStetra CStetra HAtetra derivative Native 30 seconds 10 minutes 10 hours Two minutes

Resolution range 50–1.3 50–1.35 50–1.45 50–1.5 50–1.25 50–1.9

R-factor (Rfree) 0.134 (0.155) 0.130 (0.175) 0.138 (0.177) 0.136 (0.170) 0.113 (0.142) 0.190 (0.251) No. non-hydrogen protein atoms 5687 5623 5617 5627 5646 5629 No. of water molecules 1103 1025 1049 1061 1107 837 Average B-factor (A˚ 2) Protein main-chain atoms 14.9 14.1 15.5 15.4 13.6 21.8 Side-chain atoms 16.5 16.3 17.9 16.7 15.4 22.4 Water molecules 28.7 29.4 31.4 30.5 29.6 31.6 Substrate atoms 24.8 21.2 15.5 22.5 r.m.s.d. bond length (A˚ ) 0.019 0.022 0.023 0.021 0.021 0.018 r.m.s.d bond angle (deg.) 1.84 1.97 1.84 1.86 2.01 1.72 Ramachandran plot. Residues in: Most favorable region (%) 90.8 90.7 90.5 89.7 89.9 88.3 Disallowed regions (%) 0.3 0.3 0.3 0.3 0.3 0.5

Trp126 and Trp465, provide stacking interactions through a bridging water molecule, to Asp222 and with the sugar units occupying positions 21 and Gln232. The 21 sugar 4-O-sulfo group is posi- þ2, respectively, while Trp125 is aligned edge-on tioned just above the guanidinium group of and forms a hydrogen bond with the bridging Arg300 and, in addition, forms H-bonds to Glu412 oxygen atom between the þ2andþ1units.The and Asn598 through a bridging water molecule. 4-O-sulfo groups of the substrate form several Both 4-O-sulfo groups contribute to substrate bind- interactions with the protein. The 4-O-sulfo group ing but do not add signiﬁcantly to the speciﬁcity of of the þ2 sugar makes H-bonds to Gln169 and, substrate recognition. The chondroitin sulfate tetrasaccharide used in these studies was obtained by the action of GAG Table 4. Relative occupancies of sugars in positions 22, lyases and contained an unsaturated ring at the 21, þ1, þ2 for different soaking times non-reducing end, with a C4vC5 double bond.10 Site occupancy The electron density for the 22 sugar corresponds Soak time very well to this unsaturated ring in E confor- 2 2 þ þ 1 2 1 1 2 Phosphate mation, with the C5 having sp2 hybridization Native 1.0 (Figure 3a). A list of hydrogen bonds between the CStetra 30 seconds 0.7 0.7 0.6 0.6 0.4 tetrasaccharide and the protein side-chains is Two minutes 0.7 0.7 0.5 0.5 0.5 given in Table 5. Ten minutes 0.7 0.7 0.4 0.4 0.6 35 minutes 0.7 0.7 0.4 0.4 0.6 Most interesting from the viewpoint of the Two hours 1.0 1.0 0.4 0.4 0.6 mechanism of catalysis is the conformation and Four hours 1.0 1.0 0.3 0.3 0.7 interactions with the enzyme of the þ1 glucuronic Ten hours 1.0 1.0 0.25 0.25 0.7 tetra acid. The electron density shows that this ring HA Two minutes 1.0 1.0 – – 1.0 assumes a distorted boat conformation O,3B with Chondroitinase AC Crystal Structure and Mechanism 373

Table 5. Close contacts between the substrate and the group is also only 2.8 A˚ from the C5 of the glucu- protein in CStetra complex for the 30 second soak ronic acid, with the O atom nearly along the experiment derived direction of the C5–H5 bond (Figure Sugar number Atom Protein atom Distance (A˚ ) 3(d)). The Arg296 side-chain also forms a hydrogen bond to the bridging oxygen atom between þ1 and 22 O2 OD2 Asn303 2.85 21 sugars and is 3.0 A˚ from the Tyr242 hydroxyl O2 NH1 Arg300 3.07 group. On the other hand, the distance from the O3 OD1 Asn303 3.03 O6A NH1 Arg134 3.01 potential base His233 NE2 to the C5 of the þ1 O6B NH2 Arg134 2.90 glucuronic acid is relatively long at 4.0 A˚ . The electron density corresponding to the 2 1 O7 NH2 Arg300 2.92 O7 NH1 Arg300 3.01 carboxylate group of the glucuronic acid in the þ1 SO43 NE Arg300 3.27 site showed not two but three bulges, with the SO44 NH2 Arg300 3.46 third one having somewhat lower density. This þ 1 O2 ND1 Asn410 3.20 shape was common to the density observed in all O2 OD1 Asn124 2.77 datasets and its position coincided with the phos- O3 ND2 Asn124 2.91 phate group in the native structure. We have O4 NH2 Arg296 3.02 modeled a phosphate group with partial occu- O4 OH Tyr242 2.88 C5 OH Tyr242 2.75 pancy in the same location (Table 4), assuming O6A NE2 His233 2.76 that it is present there when the glucuronic acid O6A ND2 Asn183 3.11 does not occupy this site. The total occupancy of O6B OD1 Asn183 2.62 the þ1 site, that is glucuronic acid plus phosphate, þ 2 O1 NH2 Arg174 3.22 equals 1. O3 NE1 Trp125 3.32 A strong peak in the electron density map was O7 NH1 Arg174 3.07 found in the proximity of the tetrasaccharide. It is SO41 NE2 His233 2.85 surrounded by six oxygen atoms in tetragonal SO43 NE2 Gln169 2.34 bipyramidal coordination, with distances to the equatorial oxygen atoms of 2.2–2.4 A˚ and to the axial oxygen atoms of 2.8 A˚ . This peak was inter- the hydroxyl groups equatorial and the C5 preted as a sodium ion. The equatorial ligands are carboxylate group pseudoaxial (Figure 3(b)). This the carbonyl groups of His233 and Trp465 and carboxylate group is placed exactly opposite the two water molecules, while the axial ligands are conserved side-chain of Asn183, whose OD1 and the OG1 of Thr235 and a water molecule. ND2 atoms are clearly distinguished by their peak height in the electron density map (Supplementary Substrate specificity Material). The distance between the amide ND2 atom and the carboxyl O6A is 3.1 A˚ , and that The initial characterization of ArthroAC showed between carbonyl OD1 and carboxyl O6B is 2.6 A˚ . that the enzyme degrades chondroitin 4-sulfate, This short O6B· · ·OD1 distance suggests the exist- chondroitin 6-sulfate and hyaluronan.13 Our results ence of a hydrogen bond between them,26 which with soaking various oligosaccharides indicated in turn would indicate that the glucuronic acid that ArthroAC displays higher activity toward carboxylate group is protonated and therefore in a hyaluronan than to chondroitin sulfate. We have neutral state. The O6A, in addition to accepting a determined the kinetic parameters of ArthroAC hydrogen bond from ND2 of Asn183, also forms a with GAG obtained from whale cartilage (CS-A, second, 2.8 A˚ long hydrogen bond with the NE2 predominantly chondroitin 4-sulfate), shark carti- atom of His233. The geometry of hydrogen bonds lage (CS-C, predominantly chondroitin 6-sulfate), involving the carboxylate group is very close to shark fin (CS-D, a mixture of chondroitin 4-sulfate, ideal, the C6–O6A/B-donor angles are in the 6-sulfate and -4,6-disulfate), low molecular mass range 115–1248 and the –COO group and the hyaluronan and high molecular mass hyaluronan three hydrogen bond donor atoms are nearly (Table 6). Indeed, the Vmax for hyaluronan is twice coplanar (Figure 3(d)). His233 donates, in addition, as high as that for chondroitin sulfate, while the a second hydrogen bond from the ND1 atom to KM calculated on a per monomer basis is about the side-chain of the conserved Glu407; therefore, three times lower. These values point to a rather this histidine residue must be protonated. The small contribution made by the sulfate groups to hydroxyl groups at C2 and C3 of the þ1 unit are the total binding energy of the substrate. held firmly through several hydrogen bonds. Atoms O2 and O3 are H-bonded to OD1 and ND2 Comparison with other GAG lyases of Asn124, respectively, while O2 is H-bonded also to ND2 of Asn410 (Figure 3(c)). ArthroAC is very similar in overall structure to Two other residues make crucial contacts with other GAG lyases with the (a/a)5 fold. The closest the substrate. The hydroxyl O atom of Tyr242 is similarity is with S. pneumonia hyaluronate lyase within H-bonding distance of the bridging oxygen (SpHL, PDB code 1EGU). These two structures atom between þ1 and 21 sugars (2.9 A˚ ), and is superimpose with root-mean-squares (rms) devi- 3.3 A˚ from O5 of the þ1 sugar ring. This hydroxyl ation of 1.3 A˚ for 632 Ca atoms out of 750 residues 374 Chondroitinase AC Crystal Structure and Mechanism

Figure 3 (legend opposite)

(Figure 4). The superposition with FlavoAC results Y408F mutant (inactive, equivalent to Y242 of in an rms deviation of 1.4 A˚ for 468 Ca atoms. The ArthroAC) complexed with hyaluronan oligo- similarity extends as well to the general mode of saccharide,27 FlavoAC complexed with the derma- substrate binding. Comparison with the SpHL tan sulfate hexasaccharide, and its inactive Y234F Chondroitinase AC Crystal Structure and Mechanism 375

Figure 3. (a) Electron density for the tetrasaccharide of chondroitin 4-sulfate substrate in the omit map calculated without the substrate present with the data for the 30 seconds soak of ArthroAC in 5 mM tetrasaccharide solution and contoured at the 3s level. The phosphate group with partial occupancy is shown. An arrow marks the bond that is cleaved by the enzyme. (b) Close-up of the same map for the þ1 glucuronic acid showing the distorted boat conformation; (c) stereo view of the substrate-binding site. The tetrasaccharide is shown in thick semitransparent lines, hydrogen bonds are shown with broken lines; (d) close-up of the active site showing residues involved in catalysis and the hydrogen bonding network connecting these residues. Magenta dotted line marks the close contact between the Tyr hydroxyl group and the C5 atom of glucuronic acid. mutant (ArthroAC Y242 equivalent) complexed That differs from the conformation of the glucuron- with chondroitin tetrasaccharide10 shows that the ate sugar observed in the wild-type ArthroAC– mode of oligosaccharide binding is nearly identi- substrate complex, where the ring forms a dis- cal, with the largest difference restricted to the torted boat with a pseudoaxial carboxylate group. glucuronic acid at the þ1 site (Figure 4(b)). The A detailed comparison of these structures side-chains making crucial contacts with the reveals subtle differences in the hydrogen bonding oligosaccharide are conserved in their type and network involving active-site residues. In ArthroAC, position. In the previously observed complexes the hydrogen bonds between the C5 carboxylic with Tyr-to-Phe mutant enzymes, the þ1 glucuro- group of the þ1 sugar and Asn183 and His233 nic acid ring is in a chair conformation, with the have nearly ideal geometry. The corresponding carboxylic group at C5 in an equatorial position. H-bonds in the FlavoAC(Y234F) and SpHL(Y408F) 376 Chondroitinase AC Crystal Structure and Mechanism

Table 6. Kinetic parameters of ArthroAC CS-A CS-C CS-D LMHA (50K) HA (1000K)

V (DABS/s) 1.42 £ 10 2 3 1.74 £ 10 2 3 6.10 £ 10 2 4 2.57 £ 10 2 3 2.78 £ 10 2 3 Km (mg/ml) 0.196 0.196 0.188 0.052 0.082 Ma (Da) 505.2313 511.1607 532.166 401.3 401.3 Km (mM) 0.387 0.383 0.353 0.130 0.205 Mb (Da) 19,000 43,000 30,000 50,000 1,000,000 Km (nM) 10.304 4.554 6.256 1.046 0.082 Disaccharide composition (%)c GAG source Whale cartilage Shark cartilage Shark ﬁn DDi-0S 1.6 1.7 0.6 DDi-6S 19.3 72.9 43.9 DDi-4S 76.2 15.4 26.9

DDi-diSD 2.7 9.3 21.3 DDi-diSE 0.3 0.6 7 DDi-triS – – 0.3 a Disaccharides composition. b GPC-HPLC (CS STD). c Analysis of unsaturated disaccharides from glycosaminoglycan by HPLC.

complexes have less favorable geometry (Figure forms a H-bond to the carboxylate group of uronic 4(b)). This asparagine residue in SpHL is modeled acid and is protonated in the complex, as judged with the opposite orientation of the amide group from it being hydrogen bonded to two acidic to that in the other two enzymes. In the FlavoAC groups (Figure 3(d)). (Y234F)-chondroitin sulfate tetrasaccharide com- On the basis of structural evidence from GAG plex, the glucuronate sugar remains in a chair con- lyase–oligosaccharide complexes, several propo- formation with all substituents to the ring being sals have been put forward as to the identity of equatorial, but rotates so that the carboxylic group the general base and general acid participating in occupies the space vacated by the missing the reaction. Jedrzejas and co-workers proposed hydroxyl group of Tyr234.10 Thus, even small that the proton is abstracted by a nearby histidine changes in the active site affect the mode of bind- residue (His233 in the present structure), and that ing of the substrate and may result in a non- another proton is donated to O4 by a tyrosine resi- productive binding. due (Tyr242 here).11,27,31 They support this proposed role of the histidine residue as the general Catalytic mechanism base by the close distance between the histidine NE1 and C5 in their structures. Huang et al. con- The enzymatic reaction carried out by GAG sidered three possible mechanistic scenarios, ulti- lyases is thought to proceed via abstraction of the mately favoring one in which a tyrosine residue C5 proton by a general base followed by proton initially functioned as a general base and subse- donation by a general acid or a water molecule to quently as a general acid.10 In a structurally related the bridging O4, with concomitant b-elimination but sequence-distant alginate lyase also, the central of the leaving group.7 Recent kinetic analysis of catalytic role was assigned to a tyrosine residue.32,33 the FlavoAC using a well defined synthetic sub- The kinetic characterization of FlavoAC with a strate agrees with the predicted stepwise, as synthetic substrate strongly favors tyrosine as the opposed to concerted, mechanism.28 A proposal proton acceptor.28 for the mechanism of polysaccharide lyases formu- We have carefully re-analyzed the available lated by Gacesa included the neutralization of the structural data to assess the role of histidine acidic group by a positively charged group to shift (His233 in ArthroAC and its equivalent) in cataly- the equilibrium toward the enolate tautomeric sis. The first suggestion for the role of histidine as form.7 Unlike polysaccharide lyases that adopt a general base was derived from the structure of the b-helix fold, the structures of FlavoAC and hyaluronan lyase and its complexes with a disac- hyaluronate lyases showed an absence of such a charide product,11,34 in which the þ1 site was not positively charged group in the vicinity of the occupied. These authors modeled the þ1 sugar acidic group of uronate. Instead, an asparagine and estimated the NE1· · ·C5 distance to be ,4A˚ . side-chain was found to face and form a H-bond This value corresponds to the N· · ·C van der with the acidic group, leaving the issue of neutral- Waals distance and seems to be too long for the ization of the acidic group an open question. The proposed proton-abstracting role of the histidine. structures of the complexes presented here suggest A direct observation of the enzyme–substrate that Asn183 OD1 forms a strong hydrogen bond complex was accomplished for FlavoAC(Y234F)10 with this carboxylic group, substantially increasing and for SpHL(Y408F).27 In the case of FlavoAC 29,30 its pKa and promoting its protonation. This (Y234F), it was concluded that the mode of sub- asparagine residue is aided by His233, which also strate binding in the þ1 site is influenced by the Chondroitinase AC Crystal Structure and Mechanism 377

Figure 4. (a) Stereo view of the superposition of the Ca traces of ArthroAC (blue) and SpHL (1EGU) (red); (b) overlay of oligosaccharide substrates from ArthroAC (blue), FlavoAC(Y234F) (1HMW, magenta) and SpHL(Y408F) (1LXK, green) based on the superposition of the backbone of active site Asn, His, Tyr (Phe) and Arg residues. The Figure was prepared with programs sPDBv51 and POV-Raye (http://www.povray.org/). mutation and does not reflect the reaction inter- A structure of the crystals of S. agalactiae hya- mediate (the C5 carboxylic group occupies the luronate lyase soaked for several days in 10– volume vacated by the missing Tyr hydroxyl 50 mM hexasaccharide substrate was reported group in the Phe mutant) therefore precluding the recently.31 In the model, the enzyme assumes a distinction between the possible mechanisms. In more open conformation and Tyr408 and His399 the case of SpHL(Y408F), the distance between are further away from the carbohydrate. Specifi- His399 and the C5 of the uronic acid is 3.73 A˚ , but cally, the distance of 5.7 A˚ between the NE1 atom the C5 proton lies almost along the line of CD2– of His399 and the C5 atom of the þ1 sugar does NE2 bond, very poor geometry indeed for proton not allow the conclusion to be drawn that the His abstraction. If the Phe408 is replaced in this model plays the role of general base. As well, the temper- by the original Tyr, its hydroxyl O atom would be ature factors in this structure (PDB code 1LXM) only ,3.0 A˚ from C5 and would have a much for most of the atoms of the modeled substrate are better geometry for interacting with the C5 proton. equal to 100 A˚ 2, significantly higher than the aver- Hence, we find these results equally inconclusive age value of ,35 A˚ 2 for the surrounding atoms, concerning the assignment of His as the general suggesting poor order of the substrate. The differ- base. ence electron density map (not reported in the 378 Chondroitinase AC Crystal Structure and Mechanism

Figure 5. Proposed catalytic mechanism of (a/a)5 GAG lyases. Panel 1, Tetrasaccharide bound in the substrate-binding site. The hydrogen bonding network involving the active site and substrate-binding residues is shown schemati- cally. The tryptophan residues stack against sugars in 21 and þ2 subsites. His233 is protonated and OD1 of Asn183 forms a strong hydrogen bond with the protonated carboxylic group of glucuronic acid. Deprotonated Tyr242 abstracts the C5 proton. Panel 2, Tyr242 accepts the proton from C5 atom leading to carbanion formation. Now Tyr242 forms hydrogen bond with bridging O4. Panel 3, The proton from Tyr242 is transferred to the O1 of galactosamine in the 21 subsite with concomitant break of C4–O4 bond and formation of unsaturated ring in þ1 subsite. Panel 4, Move- ment of Trp465 triggers the release of disaccharide product from (þ1, þ 2) subsites, reorganization of the active site and release of product from (22, 21) subsites. The enzyme is ready for the next catalytic cycle.

original paper) calculated from the deposited than His233, playing the role of a general base in structure factors and coordinates, with the sub- the first step of catalysis. The proposed reaction strate excluded from calculations, showed weak mechanism is shown in Figure 5. Substrate binding and scattered density that in our view cannot be to ArthroAC is associated with a deformation of modeled reliably as a single carbohydrate molecule the glucuronate sugar ring at the þ1 site from a (Supplementary Material). The role of His399 as a chair to a distorted boat and protonation of its general base is further put in doubt by the fact acidic group. Such a change to a higher-energy con- that His399Ala mutant of SpHL retains a signifi- formation of the sugar ring is not unusual and has cant fraction (6%) of the wild-type enzyme been observed in other carbohydrate-processing activity.11 enzymes.35 – 40 The distortion of the ring brings the The present series of timed snapshots of pseudoaxial acidic group coplanar with the amide structures of complexes of the wild-type ArthroAC group of Asn183, and the presence of an additional with chondroitin 4-sulfate tetrasaccharide substrate proton leads to the formation of two hydrogen provides very strong support for Tyr242, rather bonds, one of them being a strong O· · ·H· · ·O Chondroitinase AC Crystal Structure and Mechanism 379 hydrogen bond. The high-resolution structure of actions to the sugar unit of the substrate bound in the complex shows that the Oh atom of Tyr242 is the þ2 site (Figure 3(c)). This loop in ArthroAC- positioned 2.8 A˚ from the C5 atom of the þ1 glu- Hg assumes an open conformation, indicating its curonic acid, along the direction of the C5–H intrinsic mobility even in the crystal environment. bond. This Oh atom is at the same time within When Trp465 is sequestered from the substrate- hydrogen bonding distance of the O atom bridging binding cleft, the side-chains of Arg296 and þ1 and 21 sugars and of Arg296. We postulate His233 extend into the volume previously occu- that Tyr242 becomes deprotonated upon substrate pied by Trp465 (Figure 6). His233 moves away binding and that the Oh takes up the proton from from Glu407, forming hydrogen bonds with water the C5 atom of glucuronic acid (Figure 5, panel 2), molecules, and is most likely neutral. The salt- which is then transferred to the bridging O atom bridge between Glu412 and Arg296 is broken, but 2.9 A˚ away, with a concomitant break of the O4– the side-chain of Glu407 follows Arg296 and C4 bond (Figure 5, panel 3). A similar mechanism forms a stronger salt-bridge, with two H-bonds. utilizing a deprotonated tyrosine residue was pro- Movement of this loop after the reaction is composed for alginate lyase A1-III,33 which is structu- pleted would substantially decrease the binding of rally similar to ArthroAC despite little sequence the product in (þ1, þ2) sites and aid in its release similarity. from the enzyme. A rearrangement of residues in What then is the role of His233? This side-chain, the active site together with their hydrogen bond- and its equivalents in other lyases, is in the ing network and neutralization of His233 would proximity of the þ1 glucuronic acid, but with the lead to reprotonation of Tyr242 making it ready NE2 atom at a distance of ,4A˚ or more from C5 for the next catalytic cycle (Figure 5, panel 4). We of the þ1 sugar, typical for van der Waals inter- propose that the deprotonation of Tyr242 and pro- actions. The geometry of the H· · ·NE2 is also such tonation of His233 is concomitant with substrate that the direction of the expected lone pair on NE2 binding and triggered by the interaction of the is ,608 away from the direction to the C5 proton. carboxylate group on the þ1 sugar with Asn183 This geometry is observed consistently in the struc- and His233. tures of the FlavoAC and SpH carbohydrate com- The access of the substrate to the active site plexes. Finally, NE2 of His233 is hydrogen bonded necessitates either local movements of two to to the carboxylate group of the þ1 glucuronic acid three loops closing off the binding site,10 or a and, thus, must be protonated. Considering these hinge movement between the domains leading to facts collectively, it is unlikely that His233 plays a global movement of the N and C-terminal the role of general base to remove the C5 proton. domains creating an opening to the active site.27,33 We postulate that this side-chain has two functions: (1) it helps properly orient the carboxylate group of Structural rationale for the exolytic versus the glucuronic acid through the NE2· · ·O6A hydro- endolytic mode of action gen bond; (2) being protonated in the complex and in conjunction with Arg296, it lowers the pKa of Enzymatic characterization of ArthroAC showed Tyr242 leading to the deprotonation of its hydroxyl that it acts as an exolyase, releasing disaccharides group and priming it for the role as a general base from the glycosaminoglycan substrate,41 while (Figure 5, panel 1). This second function of His233 FlavoAC acts as an endolyase.42 Comparison of is directly related to the nearby presence of the structures of these two enzymes provides a Glu407, which engages the histidine ND2 proton rationale for the observed differences in the in a hydrogen bond and through electrostatic inter- mode of action. ArthroAC has two insertions in actions aids in histidine protonation. Glu407 forms the N-terminal domain relative to the FlavoAC: ,15 at the same time a salt-bridge with Arg296, com- residues (Arg23–Ser38) and 25 residues (Thr343– pleting a tetrad of hydrogen bonded residues Gly366). These two segments form a-helices- (Tyr242, His233, Arg296, Glu407) forming the containing loops near the N and C termini of the active site. domain, which come together, closing off a large The conservation in FlavoAC and hyaluronate part of the cleft running along the side of the (a/a)5 lyases of all Asn183, His233, Tyr242, Arg300 and toroidal N-terminal domain. These loops form a Glu407, residues critical for the above described wall that converts the open cleft into a deep cavity mechanism supports the view that this is a com- and precludes binding of an extended oligosac- mon catalytic mechanism for this class of enzymes. charide. The size of the cavity restricts the binding A possible mechanism of product release was of the carbohydrate on the non-reducing end and illuminated by the structure of thimerosal-soaked can accommodate approximately, two to three ArthroAC solved here. Of the three bound heavy- sugar molecules (Figure 7(a)), which correlates atoms, two bound to surface-exposed cysteine resi- with the exolytic activity of this enzyme. Binding dues. The third bound to Cys408, which in the of a longer carbohydrate would require substantial native structure is covered by the 460–469 loop rearrangement of these two loops and apparently and inaccessible to the solvent, suggesting that does not occur frequently. The open cleft of the this loop is flexible enough to allow access for a FlavoAC does not impose such a constraint and relatively large thimerosal molecule. At the tip of allows binding to the middle of a long carbo- this loop is Trp465, which provides stacking inter- hydrate (Figure 7(b)). 380 Chondroitinase AC Crystal Structure and Mechanism

Figure 6. Stereo view of the conformation of the 460–469 loop in native and complexed ArthroAC (blue) and in ArthroAC-Hg (magenta). The location of the thimerosal Hg atom near the Cys408 in the open conformation is shown as a magenta ball. The side-chains of His233, Arg296, Glu407 and Trp465 are shown explicitly with the hydrogen bonds marked in broken lines. The þ2 and þ1 sugars are also shown.

Materials and Methods ultrafiltration (Ultrafilter Type P0200, cut-off 20,000 Da, Advantec, Tokyo Japan) under N2: Finally, the enzyme Protein purification was loaded onto a Sephacryl S-200 HR gel-filtration column (Amersham Bioscience Corp, Piscataway, NJ) The protein was purified from its natural host at equilibrated with 0.01 M sodium acetate buffer (pH 5.6) Seikagaku Corporation. To stimulate expression of chon- and eluted with the same buffer. The fractions showing droitin AC lyase II, A. aurescens was grown under the the highest specific activity and high purity by SDS- following culture conditions: medium (32.5 l) containing PAGE were pooled and used in crystallization experiments. 0.4% (w/v) peptone (Kyokuto Pharmaceutical Industry Co. Ltd, Tokyo), 0.4% (w/v) Ehrlich’s fish extract (Kyokuto Pharmaceutical Industry Co. Ltd, Tokyo) and Oligosaccharide preparation 0.75% (w/v) chondroitin sulfate C (Seikagaku Co., Tokyo); initial pH 6.2; aeration rate 1 vessel volume per Chondrotin 4-sulfate tetrasaccharide (CStetra) and hya- minute; agitation speed 220 rpm; cultivation time 24 luronan tetrasaccharide were prepared and characterized hours. The enzyme was purified essentially as as described.10 Briefly, chondroitin 4-sulfate from bovine described13 but with some modifications. Briefly, after trachea and dermatan sulfate from porcine intestinal bacterial cells were pelleted by centrifugation at 15,000g mucosa were subjected to controlled depolymerization for 15 minutes, solid ammonium sulfate was added to using chondroitin ABC lyase and the reactions were ter- the supernatant fluid up to 75% saturation. At 4 8C, 75 g minated prior to completion by boiling for five minutes. of the protein precipitate (50,000 units) was dialyzed Each oligosaccharide mixture was separated on a Bio- against 20 mM sodium acetate (pH 5.2) and loaded onto Gel P6 column and fractions consisting of tetrasacchar- an SP-Sepharose (Amersham Bioscience Corp, Piscat- ides and hexasaccharides were collected. These mixtures away, NJ) column (2.6 cm £ 70 cm) pre-equilibrated were further fractionated by strong anion-exchange with the same buffer. The enzyme was eluted with a HPLC, single oligosaccharides were obtained, and their linear gradient from 20 mM to 300 mM sodium acetate purity confirmed by capillary electrophoresis and their buffer (pH 5.2). Enzyme activity and protein amounts structures by MS and NMR analyses. The structure of were monitored as described.13 This chromatography CStetra was DUAp (1 ! 3)-b-D-GalpNAc4S (1 ! 4)-b-D- procedure was repeated three times and the enzyme- GlcAp (1 ! 3)-a,b-D-GalpNAc4S, where DUAp is the containing fractions collected and concentrated by unsaturated sugar residue, 4-deoxy-a-L-threo-hex-4-is Chondroitinase AC Crystal Structure and Mechanism 381

Figure 7. Stereo view of the molecular surface within the extended substrate binding site of (a) ArthroAC–tetrasaccharide complex. The insertions in the sequence capping the substrate binding site are shown in green. (b) FlavoAC, with bound tetrasaccharides. The Figure was prepared with the program GRASP.54 enopyranosyluronic acid; GlcAp is glucopyranosyl- We obtained crystals by the hanging-drop, vapor-diffu- uronic acid; GalpN is 2-deoxy-2-aminogalactopyranose; sion method in drops containing 2 ml of protein (10 mg/ S is sulfate; and Ac is acetate. ml) and 2 ml of reservoir solution (23% (w/v) PEG8000, 0.1 M sodium phosphate buffer (pH 6.4), 0.4 M ammonium acetate, 10% (v/v) glycerol) suspended Protein crystallization and data collection over 1 ml of reservoir solution. Small crystals appeared within a few days. Larger crystals were obtained by Needle-shaped crystals of ArthroAC were reported macroseeding with precipitant concentration lowered to many years ago;13 however, they were not characterized. 21% (w/v). 382 Chondroitinase AC Crystal Structure and Mechanism

The crystals belong to the monoclinic space group P21 containing thimerosal derivative of ArthroAC. Analysis with cell dimensions a ¼ 57:6A˚ , b ¼ 85:5A˚ , c ¼ 80:5A˚ , of the MAD datasets using the program SOLVE44 b ¼ 106.98 and contain one molecule in the asymmetric revealed three Hg sites in the asymmetric unit. These unit. Prior to data collection the crystals were immersed sites were used to calculate experimental phases to a for ten seconds in a cryoprotectant solution containing resolution of 1.3 A˚ and resulted in an overall figure of 22.5% (w/v) PEG8000, 0.1 M sodium phosphate buffer merit (FOM) of 0.33–1.3 A˚ . Electron density modification (pH 6.4), 0.4 M ammonium acetate and 20% (v/v) performed with the program RESOLVE45 assuming a glycerol, mounted in a nylon loop and flash-cooled in a solvent content of 0.4 led to a significant increase of ˚ cold stream of N2 gas to 100 K. These crystals diffracted the FOM (0.45 at 1.3 A resolution) and substantially to 1.3 A˚ resolution at the X8C beamline at Brookhaven improved the electron density map. Approximately, National Laboratory. 80% of the protein main chain was built automatically Crystals of the protein complexed with thimerosal using the program RESOLVE. Additional fragments of were obtained by an overnight soak of native crystals in the main chain and many side-chains were built manu- the cryoprotectant solution containing 2 mM Hg salt. ally using the program O.46 Since the primary sequence These crystals were non-isomorphous with the native of the protein was unknown, the amino acid type for crystals and had cell dimensions: a ¼ 57:4A˚ , b ¼ 85:3A˚ , each residue was selected to fit the experimental electron c ¼ 82:2A˚ , b ¼ 105.88. Multiwavelength anomalous dif- density map. At 1.3 A˚ resolution, most of the assign- fraction (MAD) data at three wavelengths were collected ments were unambiguous and nearly the entire chain (Table 7). was traced in the initial map. This initial sequence assignment was adjusted during the progress of refinement as the electron density features improved. Initial Enzyme–substrate complexes refinement was performed with the program CNS, version 1.147 and the model was rebuilt manually using the To obtain complexes of ArthroAC with its substrates program O. Subsequently, refinement was continued we resorted to a series of soaks of the native crystals in using the program REFMAC5, version 5.1.08.48 During cryoprotectant solution containing the substrate for refinement of this model, 1% of the reflections were set times ranging from 30 seconds to ten hours (Table 2). aside for the calculation of Rfree: Water molecules were Native crystals were soaked in the cryo-protecting solu- initially added automatically with the program CNS tion containing 5 mM CStetra or HAtetra for a specified and subsequently updated and corrected by visual length of time, then flash-frozen in a cold N2 gas stream inspection of the difference map. The final model of the (100 K) on the detector and used immediately for data thimerosal-soaked crystal has been refined to an R-factor ˚ collection. There was no significant change in cell dimen- of 0.134 and Rfree of 0.156 at 1.3 A resolution (Table 3). sions as compared to the crystals of native ArthroAC The current model contains 754 residues (Pro4–Arg757). (Table 2). The final amino acid sequence of the model was derived Diffraction data were collected at the X8C beamline, from the combination of electron density maps and mass NSLS, Brookhaven National Laboratory, using the spectrometry data. Quantum-4 CCD detector. The highest-resolution data, 1.25 A˚ , were obtained from the native crystal soaked for ten hours in 5 mM CStetra. Data processing and scaling Refinement of native protein and enzyme– was performed with HKL2000.43 Data collection statistics substrate complexes are shown in Tables 2 and 7. The model of Hg-bound ArthroAC was taken as a Structure determination and refinement starting point for refinement of the crystal structure of the native protein. Refinement was performed with the The structure of ArthroAC was solved from the Hg- program REFMAC5. For monitoring of Rfree during refinement, 1% of reflections were set aside. The electron density map showed that one loop had a substantially different conformation and was rebuilt manually. This Table 7. MAD data collection statistics model was refined at 1.35 A˚ resolution to a final R-factor of 0.130 and Rfree of 0.175. The model contains residues Hg— Hg— Hg— 4–757, 1025 water molecules, one sodium ion and one peak inflection remote Native phosphate ion (present at 0.1 M concentration in the Wavelength (A˚ ) 1.005133 1.009078 0.997068 0.950000 mother liquor). a (A˚ ) 57.3 57.6 This model, in turn, was used to determine the struc- b (A˚ ) 85.2 86.5 tures of all the complexes of chondroitin AC lyase with c (A˚ ) 82.1 80.5 chondroitin 4-sulfate tetrasaccharide (CStetra) with differ- b (deg.) 105.8 106.9 ent soaking times (30 seconds, two minutes, ten minutes, Resolution 40–1.3 40–1.3 50–1.4 50–1.3 ˚ 35 minutes, two hours, four hours, ten hours) and of the range (A) (last (1.35– (1.35– (1.45– (1.35– HAtetra complex (two minutes soaking time). It was clear shell) 1.30) 1.30) 1.40) 1.30) from the height of peaks in the difference electron dens- Rsym (last shell) 0.066 0.067 0.067 0.086 (0.274) (0.325) (0.221) (0.540) ity map that the occupancies of sugars in sites (22, 21) Completeness 98.4 98.8 99.0 96.4 and (þ1, þ2) differed systematically from structure to (%) (last shell) (85.0) (89.0) (90.7) (93.4) structure, in a manner consistent with the enzyme being I=sðIÞ (last 14.9 14.7 (6.8) 13.7 8.6 (3.4) active in the crystal. In such a case, the initial concen- shell) (5.3) (4.5) tration of tetrasaccharide would decrease with time, Total reflections 876,240 884,015 721,975 1,294,307 while that of a disaccharide product increases. At any Unique reflec- 182,880 183,424 147,941 178,066 time, active sites of some enzyme molecules might tions be temporarily empty. Previous structural studies on Redundancy 4.8 4.8 4.9 7.3 chondroitin lyase showed that disaccharides were found Chondroitinase AC Crystal Structure and Mechanism 383 predominantly in subsites (22, 21), and not in (þ1, by the presence of hydrogen bond donors or acceptors þ2).10 Therefore, at any given time there is a mixture of within 3 A˚ . The assignment of Phe was rather straight- tetrasaccharides bound to sites (22, 21, þ1, þ2) and forward. The remaining amino acid types, Ser, Ala and disaccharides bound predominantly to sites (22, 21). Gly, were assigned based on the shape of the electron For that reason, the occupancies of sites (22, 21) are density. expected to be higher than for sites (þ1, þ2). To obtain The electron density was convincing for Asx and Glx. an estimate of the relative occupancies at the 2 and þ Initially, all Asx were refined as Asp and all Glx as Glu. sites, we assumed that, since the substrate is bound When the R-factor dropped below 0.2 we reassigned tightly to the enzyme through numerous hydrogen these residues to either Gln (Asn) or Glu (Asp) based bonds, stacking and van der Waals interactions, the tem- on: (1) the height of peaks corresponding to the side- perature factors of the substrate are similar to those of chain atoms; (2) the hydrogen bonding network and the the residues with which it interacts. We have adjusted directionality of hydrogen bonds; (3) sequence conserva- independently the occupancies of sites (22, 21) and tion in related protein sequences. All these assignments (þ1, þ2) in steps of 0.1 until the average temperature were re-evaluated during each rebuilding cycle and cross- factor of the substrate was close to that of the surround- checked between all independently refined structures. ing atoms (Table 4). The derived occupancy values provide information about the relative rather than absolute occupancies. In the HAtetra complex, only two sugar rings were visible in the difference electron density Mass spectrometry-based assignments map, in positions (22, 21) (reaction product). Full Mass spectrometry was applied to determine the con- statistics of refinements are summarized in Table 3. sistency between the measured molecular mass of the PROCHECK49 showed that all models have good geo- entire molecule and trypsin-derived peptides and the metry with no outliers. predicted molecular mass based on the established amino acid sequence. The molecular mass of the entire Protein Data Bank accession numbers protein was determined on an LC/MS Agilent 1100 mass spectrometer. For MS/MS analysis the protein was Coordinates of the native chondroitinase AC in-gel digested with trypsin. Peptides were separated on (ArthroACnat), the mercury derivative (ArthroACHg), a CapLC HPLC (Waters, Milford, USA) with 0.1% (v/v) the two minutes soaked complex with hyaluronan aqueous formic acid and 0.21% (v/v) formic acid in (ArthroACHA), and 30 seconds, ten minutes and ten acetonitrile used for the gradient composition. A volume hours soaked complexes with CStetra (ArthroACCStetra) are of 20 ml of in-gel digest sample was injected onto the deposited in the Protein Data Bank, RCSB, with acces- trapping column at a flow-rate of 15 ml/minute and sion codes 1RW9, 1RWA, 1RWC, 1RWF, 1RWG, 1RWH. then washed for five minutes with 0.1% (v/v) aqueous formic acid. The peptides from the trapping column e Amino acid sequence assignment were directed to the PicoFrit analytical column (New Objective, MA) filled with 10 cm of C18 BioBasice pack- ing (5 mm, 300 A˚ ,75mmID£ 10 cm). The spraying tip As mentioned above, the amino acid sequence of the of the PicoFrite column was positioned near the protein has not been previously determined. The high sampling cone of the mass spectrometer and the capil- resolution of the diffraction data and the parallel lary voltage adjusted to achieve the best plume possible. independent refinement of several structures (native, thi- The analysis was done on a QTOF-2 mass spectrometer merosal derivative and several enzyme–substrate com- (Micromass, UK) upgraded with EPCAS electronics. The plexes) allowed us to deduce the sequence directly from instrument was set in data directed analysis (DDA) the electron density maps. This sequence was confirmed mode, where an MS survey scan from 350 to 1600 m/z and corrected in several places by mass spectrometry was recorded in one second, then the strongest ion was analysis. For each of the five best datasets at resolutions ˚ selected for MS/MS (50–2000 m/z) for duration of one 1.25–1.5 A, the refinements converged at very low R- second. The DDA was set to select doubly and triply factors of 0.114–0.138 (R-free of 0.142–0.178), indicating charged ions. The instrument then switched back to the highly refined and reliable structures. MS survey to select the second most intense ion for the next MS/MS spectrum. The mass spectrometric analysis Electron density-based assignments was designed to collect a maximum of different ions to get a maximum of protein coverage. The interscan time At 1.3 A˚ resolution the residue types Trp, Tyr, His, was set to 0.1 second. Arg, Lys, Pro, Ile, Leu are defined unambiguously by The optimized lyase sequence of 757 amino acid resi- their shape (Supplementary Material). The three highest dues when matched with the experimental 317 tandem peaks in the MAD-derived experimental map corre- mass spectra resulted in the assignment of 157 of the sponded to Hg atoms and each was associated with a 317 tandem MS spectra that identified 37 tryptic frag- cysteine residue. The subsequent 12 highest peaks occur ments (645 residues) that covered 85.2% of the sequence. in the middle of the side-chain electron densities and Mascot search parameters were restricted to oxidation of were associated either with a cysteine or a methionine methionine, histidine and/or tryptophan, alkylation of residue, based on the shape of the electron density. The cysteine by iodoacetamide or in-gel during electrophor- next group of peaks corresponded to main chain or esis by acrylamide, and the deamidation of asparagine/ side-chain oxygen atoms. During the refinement, three glutamine or the carbamidomethyl derivative of cys- additional residues on the protein surface were assigned teine, with a parent ion mass tolerance of 0.5 Da and as methionine based on the refined shape of the electron allowing for one missed cleavage. Relaxing the strin- density. A total of seven cysteine and 11 methionine gency of the search to allow for non-specific proteolysis residues were identified. The distinction between Val by trypsin (and maintaining the parameters above) and Thr could be made based on the height of the elec- resulted in the identification of a further 28 (non-tryptic) tron density for CG2 versus CG1/OG1, often supported fragments defined by 35 tandem MS spectra with a 384 Chondroitinase AC Crystal Structure and Mechanism concomitant increase in coverage by 25 amino acid resi- seconds, 100 ml of solution A was added to solution B dues to 88.5%, while further decreasing the stringency and the increased absorbance (ABS) at 232 nm due to of the matching by increasing the tolerance on the parent the generation of unsaturated bonds in disaccharides peptide mass to 2 Da resulted in the assignment of caused by Arthro AC was monitored at every 30 seconds another ten spectra to the sequence. In total, 202 of the by spectrophotometer (UV 160A, Shimadzu, Kyoto, tandem MS spectra were matched to the best sequence Japan). Reaction velocities (v) at different concentration in this manner. The quality of the remaining MS/MS of the GAGs were plotted as DABS=Dt: Since all GAGs spectra was insufficient for confirmation of sequence used in the experiment were catalyzed to disaccharide although sequence tags50 of low confidence could be gen- units, we calculated substrate concentrations of the erated and about 42 of the spectra were weak, sparse and GAGs as molar basis of disaccharide units contained in possibly correspond to background noise. An accounting each GAG. For example, in the case of the CS-A, it was of the peptides as predicted from the X-ray data but not composed of 95.5% of mono-sulfated disaccharide unit observed by tandem MS indicated that several short (503.3 Da), 3.0% of di-sulfated disaccharide unit (605.3 Da) tryptic peptides (residue numbers 44–50, 51–57, 175– and 1.6% of non-sulfated saccharide unit (401.3 Da). 176, 295–296, 297–300, 342–343, 356–360, 385–388, Therefore, mean molecular mass of the CS-A as disac- 496–501, 502, 586–588, 639, 640–645, 646–650, 662–664, charide unit can be calculated to be 505.2. Accordingly, 752–754, 755–756, 757) that account for 64 residues mean molecular mass of the CS-C, the CS-D and both were most likely lost during the loading/washing of the hyaluronic acid from the disaccharide-unit basis were reversed-phase LC column, whereas the rather large calculated as 511.2 Da, 532.2 Da and 401.3 Da, respec- tryptic fragment (residues 503–525) that is well defined tively. By use of substrate concentration based on this by the X-ray data may have remained in the gel during calculation, a Lineweaver–Burke plot was obtained, and the robotic in-gel digestion/extraction and/or remained V and KM of every GAG were determined. bound to the reversed-phase column throughout the LC-QToF MS analysis. Results of Mascot analysis and samples of MS/MS spectra for several peptides are deposited as Supplementary Material. An example of the experimental and the refined elec- Acknowledgements tron density map for the region around Asn183 is shown in Supplementary Material. Finally, the sequence We thank Drs Allan Matte, Joseph D. Schrag and of ArthroAC derived by combining information from J. Sivaraman for help with data collection, electron density maps and mass spectrometry was Ms France Dumas for amino acid sequencing, aligned with related enzymes based on their structural superposition (Figure 1). Ms Christine Munger, Dr Daniel Boismenu, At completion of this process we were confident of the Mr Sajid Karsan and Montreal Proteomics Network identity of 746 out of 754 modeled residues (98.9%). For for mass spectrometry analysis. This work was three additional residues (Asn/Asp and Gln/Glu) we supported, in part, by CIHR grant 200009MOP- are less certain of the assignment and it is based primar- 84373-M-CFAA-26164 to M.C. and NIH grants ily on differences in the B-factors and potential formation GM38060 and HL62244 to R.J.L. of hydrogen bonds. Finally, five residues located in flexible loops exposed to solvent and with B-factors of more than 30 A˚ 2 (average B for the protein is ,15 A˚ 2) and for which no fragment was found in MS/MS data could References not be identified unambiguously. These residues are dis- 1. Scott, J. E. (1995). Extracellular matrix, supramolecu- tant from the substrate binding site (marked in Figure 1). lar organisation and shape. J. Anat. 187, 259–269. At the end of the structure determination process we 2. Iozzo, R. V. (1998). Matrix proteoglycans: from mol- have determined the N-terminal amino acid sequence of ecular design to cellular function. Annu. Rev. Biochem. 21 residues of a dissolved crystal by Edman degradation. 67, 609–652. The first three N-terminal residues are disordered in the 3. Lidholt, K. (1997). Biosynthesis of glycosaminogly- structure, the identity of the subsequent 18 residues cans in mammalian cells and in bacteria. Biochem. agrees with the electron-density-based assignments. Soc. Trans. 25, 866–870. 4. Prydz, K. & Dalen, K. T. (2000). Synthesis and sorting of proteoglycans. J. Cell Sci. 113, 193–205. Kinetic analysis 5. Ernst, S., Langer, R., Cooney, C. L. & Sasisekharan, R. (1995). Enzymatic degradation of glycosaminogly- Glycosaminoglycans CS-A (19,000 Da), CS-C (43,000 Da), cans. Crit. Rev. Biochem. Mol. Biol. 30, 387–444. CS-D (30,000 Da), low molecular mass hyaluronic acid 6. McCarter, J. D. & Withers, S. G. (1994). Mechanisms (50,000 Da) and hyaluronic acid (1,000,000 Da) were of enzymatic glycoside hydrolysis. Curr. Opin. Struct. obtained from Seikagaku Corporation (Tokyo, Japan). Biol. 4, 885–892. Arthro AC was dissolved at concentration of 50 mU/ml 7. Gacesa, P. (1987). Alginate-modifying enzymes. A in 0.1% (w/v) BSA (solution A). Each chondroitin sulfate proposed unified mechanism of action for the lyases and lower molecular mass of hyaluronic acid was dis- and epimerases. FEBS Letters, 212, 199–202. solved in distilled water at concentration of 10 mg/ml, 8. Sutherland, I. W. (1995). Polysaccharide lyases. FEMS whereas hyaluronic acid was dissolved in it at concen- Microbiol. Rev. 16, 323–347. tration of 2 mg/ml, respectively. From 10 ml to 200 mlof 9. Scavetta, R. D., Herron, S. R., Hotchkiss, A. T., Kita, solution containing each GAG thus prepared was added N., Keen, N. T., Benen, J. A. et al. (1999). Structure of to 160 ml of 0.4 M sodium acetate buffer (pH 6.0) and a plant cell wall fragment complexed to pectate the appropriate amount of distilled water was added to lyase C. Plant Cell, 11, 1081–1092. each solution to adjust the final volume to 700 ml (solu- 10. Huang, W., Boju, L., Tkalec, L., Su, H., Yang, H. O., tion B). Each solution B was incubated at 37 8C. After 30 Gunay, N. S. et al. (2001). Active site of chondroitin Chondroitinase AC Crystal Structure and Mechanism 385

AC lyase revealed by the structure of enzyme–oligo- (2002). Mechanism of hyaluronan degradation by saccharide complexes and mutagenesis. Biochemistry, Streptococcus pneumoniae hyaluronate lyase: struc- 40, 2359–2372. tures of complexes with the substrate. J. Biol. Chem. 11. Li, S., Kelly, S. J., Lamani, E., Ferraroni, M. & Jedrzejas, 277, 28287–28297. M. J. (2000). Structural basis of hyaluronan degra- 28. Rye, C. S. & Withers, S. G. (2002). Elucidation of the dation by Streptococcus pneumoniae hyaluronate mechanism of polysaccharide cleavage by chondroi- lyase. EMBO J. 19, 1228–1240. tin AC lyase from Flavobacterium heparinum. J. Am. 12. Linhardt, R. J., Galliher, P. M. & Cooney, C. L. (1986). Chem. Soc. 124, 9756–9767. Polysaccharide lyases. Appl. Biochem. Biotechnol. 12, 29. Cleland, W. W. (2000). Low-barrier hydrogen bonds 135–176. and enzymatic catalysis. Arch. Biochem. Biophys. 382, 13. Hiyama, K. & Okada, S. (1975). Crystallization and 1–5. some properties of chondroitinase from Arthrobacter 30. Gerlt, J. A. & Gassman, P. G. (1993). Understanding aurescens. J. Biol. Chem. 250, 1824–1828. the rates of certain enzyme-catalyzed reactions: 14. Tkalec, A. L., Fink, D., Blain, F., Zhang-Sun, G., proton abstraction from carbon acids, acyl-transfer Laliberte, M., Bennett, D. C. et al. (2000). Isolation reactions, and displacement reactions of phosphodi- and expression in Escherichia coli of cslA and cslB, esters. Biochemistry, 32, 11943–11952. genes coding for the chondroitin sulfate-degrading 31. Mello, L. V., De Groot, B. L., Li, S. & Jedrzejas, M. J. enzymes chondroitinase AC and chondroitinase B, (2002). Structure and flexibility of Streptococcus aga- respectively, from Flavobacterium heparinum. Appl. lactiae hyaluronate lyase complex with its substrate. Environ. Microbiol. 66, 29–35. Insights into the mechanism of processive degra- 15. Pojasek, K., Shriver, Z., Kiley, P., Venkataraman, G. & dation of hyaluronan. J. Biol. Chem. 277, Sasisekharan, R. (2001). Recombinant expression, 36678–36688. purification, and kinetic characterization of chon- 32. Yoon, H. J., Hashimoto, W., Miyake, O., Murata, K. & droitinase AC and chondroitinase B from Flavobacter- Mikami, B. (2001). Crystal structure of alginate lyase ium heparinum. Biochem. Biophys. Res. Commun. 286, A1-III complexed with trisaccharide product at 343–351. 2.0 A˚ resolution. J. Mol. Biol. 307, 9–16. 16. Fe´thie`re, J., Eggimann, B. & Cygler, M. (1999). Crystal 33. Mikami, B., Suzuki, S., Yoon, H. J., Miyake, O., structure of chondroitin AC lyase, a representative of Hashimoto, W., Murata, K. (2002). X-ray structural a family of glycosaminoglycan degrading enzymes. analysis of algininatelyase A1-III mutants/substrate J. Mol. Biol. 288, 635–647. complexes: activation of a catalytic tyrosine residue 17. Hiyama, K. & Okada, S. (1977). Action of chondroiti- by a flexible lid loop. Acta Crystallog. sect. A, 58 nases. III. Ionic strength effects and kinetics in the (Suppl), C271. action of chondroitinase AC. J. Biochem. (Tokyo), 82, 34. Ponnuraj, K. & Jedrzejas, M. J. (2000). Mechanism of 429–436. hyaluronan binding and degradation: structure of 18. Hiyama, K. & Okada, S. (1975). Amino acid compo- Streptococcus pneumoniae hyaluronate lyase in com- sition and physiochemical characterization of chon- plex with hyaluronic acid disaccharide at 1.7 A reso- droitinase from Arthrobacter aurescens. J. Biochem. lution. J. Mol. Biol. 299, 885–895. (Tokyo), 78, 1183–1190. 35. Sidhu, G., Withers, S. G., Nguyen, N. T., McIntosh, 19. Davies, G. J., Wilson, K. S. & Henrissat, B. (1997). L. P., Ziser, L. & Brayer, G. D. (1999). Sugar ring dis- Nomenclature for sugar-binding subsites in glycosyl tortion in the glycosyl-enzyme intermediate of a hydrolases. Biochem. J. 321, 557–559. family G/11 xylanase. Biochemistry, 38, 5346–5354. 20. Perkins, D. N., Pappin, D. J., Creasy, D. M. & Cottrell, 36. White, A. & Rose, D. R. (1997). Mechanism of cataly- J. S. (1999). Probability-based protein identification sis by retaining beta-glycosyl hydrolases. Curr. Opin. by searching sequence databases using mass spec- Struct. Biol. 7, 645–651. trometry data. Electrophoresis, 20, 3551–3567. 37. Sulzenbacher, G., Driguez, H., Henrissat, B., 21. Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, Schulein, M. & Davies, G. J. (1996). Structure of the J., Zhang, Z., Miller, W. & Lipman, D. J. (1997). Fusarium oxysporum endoglucanase I with a non- Gapped BLAST and PSI-BLAST: a new generation of hydrolyzable substrate analogue: substrate distortion protein database search programs. Nucl. Acids Res. gives rise to the preferred axial orientation for the 25, 3389–3402. leaving group. Biochemistry, 35, 15280–15287. 22. Li, S. & Jedrzejas, M. J. (2001). Hyaluronan binding 38. Kuroki, R., Weaver, L. H. & Matthews, B. W. (1993). and degradation by Streptococcus agalactiae hyaluron- A covalent enzyme–substrate intermediate with sac- ate lyase. J. Biol. Chem. 276, 41407–41416. charide distortion in a mutant T4 lysozyme. Science, 23. Hashimoto, W., Nankai, H., Mikami, B. & Murata, K. 262, 2030–2033. (2003). Crystal structure of Bacillus sp GL1 xanthan 39. Strynadka, N. C. & James, M. N. (1991). Lysozyme lyase, which acts on the side chains of xanthan. revisited: crystallographic evidence for distortion of J. Biol. Chem. 278, 7663–7673. an N-acetylmuramic acid residue bound in site D. 24. Huang, W., Lunin, V. V., Li, Y., Suzuki, S., Sugiura, J. Mol. Biol. 220, 401–424. N., Miyazono, H. & Cygler, M. (2003). Crystal struc- 40. Strynadka, N. C. & James, M. N. (1996). Lysozyme: a ture of Proteus vulgaris chondroitin sulfate ABC model enzyme in protein crystallography. EXS, 75, lyase I at 1.9 A˚ resolution. J. Mol. Biol. 328, 623–634. 185–222. 25. Lo, C. L., Ailey, B., Hubbard, T. J., Brenner, S. E., 41. Jandik, K. A., Gu, K. & Linhardt, R. J. (1994). Action Murzin, A. G. & Chothia, C. (2000). SCOP: a struc- pattern of polysaccharide lyases on glycosaminogly- tural classification of proteins database. Nucl. Acids cans. Glycobiology, 4, 289–296. Res. 28, 257–259. 42. Capila, I., Wu, Y., Rethwisch, D. W., Matte, A., 26. Cleland, W. W., Frey, P. A. & Gerlt, J. A. (1998). The Cygler, M. & Linhardt, R. J. (2002). Role of arginine low barrier hydrogen bond in enzymatic catalysis. 292 in the catalytic activity of chondroitin AC lyase J. Biol. Chem. 273, 25529–25532. from Flavobacterium heparinum. Biochim. Biophys. 27. Jedrzejas, M. J., Mello, L. V., De Groot, B. L. & Li, S. Acta, 1597, 260–270. 386 Chondroitinase AC Crystal Structure and Mechanism

43. Otwinowski, Z. & Minor, W. (1997). Processing of 52. Kraulis, P. J. (1991). MOLSCRIPT: a program to pro- X-ray diffraction data collected in oscillation mode. duce both detailed and schematic plots of protein Methods Enzymol. 276, 307–326. structures. J. Appl. Crystallog. 24, 946–950. 44. Terwilliger, T. C. & Berendzen, J. (1999). Automated 53. Merritt, E. A. & Bacon, D. J. (1997). Raster3D: photo- MAD and MIR structure solution. Acta Crystallog. realistic molecular graphics. Methods Enzymol. 277, sect. D, 55, 849–861. 505–524. 45. Terwilliger, T. C. (2000). Maximum-likelihood dens- 54. Nicholls, A., Sharp, K. A. & Honig, B. (1991). Protein ity modification. Acta Crystallog. sect. D, 56, 965–972. folding and association: insights from the interfacial 46. Jones, T. A., Zhou, J. Y., Cowan, S. W. & Kjeldgaard, and thermodynamic properties of hydrocarbons. Pro- M. (1991). Improved methods for building protein teins: Struct. Funct. Genet. 11, 281–296. models in electron density maps and the location of errors in these models. Acta Crystallog. sect. A, 47, 110y119. Edited by I. Wilson 47. Bru¨ nger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998). Crystallography and NMR system: a new software suite for macromolecular structure determination. (Received 23 September 2003; received in revised form Acta Crystallog. sect. D, 54, 905–921. 19 December 2003; accepted 29 December 2003) 48. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallog. sect. D, 53, 240–255. 49. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283–291. 50. Mann, M. & Wilm, M. (1994). Error-tolerant identifi- Supplementary Material comprising Figures cation of peptides in sequence databases by peptide showing electron density maps at various contour sequence tags. Anal. Chem. 66, 4390–4399. levels and Tables and plots displaying peptide 51. Guex, N. & Peitsch, M. C. (1997). SWISS-MODEL identification by mass spectrometry is available on and the Swiss-PdbViewer: an environment for com- Science Direct parative protein modeling. Electrophoresis, 18, 2714–2723.