Crystal structure and carbohydrate-binding properties of the human cartilage glycoprotein-39

The human cartilage glycoprotein-39 (HCgp-39 or YKL40) is expressed by synovial cells and macrophages during inflammation. Its precise physiological role is unknown. However, it has been proposed that HCgp-39 acts as an autoantigen in rheumatoid arthritis, and high expression levels have been associated with cancer development. HCgp-39 shares high sequence homology with family 18 chitinases, and although it binds to chitin it lacks enzymatic activity. The crystal structure of HCgp-39 shows that the protein displays a ( (cid:1) / (cid:2) ) 8 -barrel fold with an insertion of an (cid:2) (cid:3) (cid:1) domain. A 43-Å long carbohydrate-binding cleft is present at the C-terminal side of the (cid:1) -strands in the ( (cid:1) / (cid:2) ) 8 barrel. Binding of chitin fragments of different lengths identified nine sug-ar-binding subsites in the groove. Protein-carbohydrate interactions are mainly mediated by stacking of side chains of aromatic amino acid residues. Surprisingly, the specificity of chitin binding to HCgp-39 depends on the length of the oligosaccharide. Although chitin disac-charides tend to occupy the distal subsites, longer chains bind preferably to the central subsites in the groove. Despite the absence of enzymatic activity, long

Rheumatoid arthritis (RA) 1 is a chronic, systemic, and extremely disabling disease that affects mainly women. It is characterized by inflammation of synovial tissues causing joint pain, swelling, and stiffness. At the moment, there is no cure for rheumatoid arthritis, and therefore current treatments focus on relieving the pain and inflammation and on stopping joint destruction. A better understanding of how rheumatoid arthritis affects the body may help to design and develop efficient treatments. Some of the most promising areas of research for the abatement of RA include oral tolerance therapy and the use of monoclonal antibodies and drugs to block the activity of interleukins, neutrophils, and T-cells (1).
The human cartilage glycoprotein-39 (HCgp-39 or YKL40) is a candidate autoantigen in RA having the capability to induce T-cell-mediated autoimmune response (2)(3)(4). Moreover, recent studies have shown that HCgp-39-derived peptides can induce immunologic tolerance in patients with persistent RA (5). HCgp-39 is secreted by articular chondrocytes and synovial cells (6), and it is expressed in peripheral blood-derived macrophages in association with their differentiation from monocytes to macrophages (7). Increased levels of HCgp-39 have been detected in the blood of RA patients (8,9), in inflamed tissues (7,10,11), and recently also in spinal fluids following neural damage (12). Additionally, abnormal expression of HCgp-39 is related to the extent of liver fibrosis (13) and to the development of cancer (14 -16). Based on these findings it has been suggested that HCgp-39 could be used as a clinical marker for early detection of disease activity and for testing the effectiveness of therapy (17).
The precise physiological function of HCgp-39 is unknown. However, it has been demonstrated recently that HCgp-39 stimulates cell proliferation of connective tissue acting synergistically with the insulin-like growth factor 1 (18), consistent with a proposed role in tissue remodeling and cellular differentiation (7,19). A similar function has been suggested for the homologous imaginal disc growth factors that promote cell proliferation in Drosophila (20).
HCgp-39 is closely related to the mammalian proteins YM1 (21) and YKL39 (22), oviduct-specific glycoproteins (23), and to the human chitinolytic enzyme chitotriosidase (24). Based on amino acid sequence and structure relationships all of these proteins have been grouped in glycoside hydrolase family 18, which also includes bacterial and plant chitinases (25). 2 These mammalian chitinase-like proteins (CLPs), which lack catalytic activity, could function as lectins able to recognize specific glycan structures present in mammalian tissues. Crystal structures of CLPs have confirmed a high degree of structure similarity to the family 18 chitinases (27)(28)(29), but none of those crystal structures has clearly revealed how CLPs bind oligosaccharides.
Here we present the structure of HCgp-39 and for the first time for a mammalian CLP, the crystal structures of complexes with intact chitin fragments bound in the carbohydrate-binding groove. Surprisingly, the HCgp-39 interacts distinctly differently with short and long chitin fragments. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The

Crystallization and Heavy Atom Derivative Search-Recombinant
HCgp-39, produced by a Chinese hamster ovary transfectant and purified to homogeneity via cation exchange chromatography, immunochromatography, and size exclusion chromatography, was kindly provided by NV Organon, Oss, The Netherlands.
HCgp-39 was crystallized using the hanging drop vapor-diffusion method. The protein solution contained 4 mg/ml HCgp-39 in 1 M NaCl, 10 mM BES buffer, pH 7.5. The precipitant consisted of 10% polyethylene glycol 8000, 0.5 M NaCl in 0.1 M sodium citrate buffer, pH 5.1. The crystals grew to a maximum size of 0.8 ϫ 0.15 ϫ 0.2 mm 3 . They belong to space group P4 3 with 4 molecules in the asymmetric unit and a solvent content of ϳ50%. HCgp-39 crystals, which were briefly soaked in a solution of 10% glycerol in mother liquor and then flash-frozen, diffracted to a resolution of 2.6 Å using synchrotron radiation. Subsequently, improved diffraction could be obtained by avoiding glycerol in the cryoprotectant solution. Instead, the HCgp-39 crystals were stabilized in 35% polyethylene glycol 8000, 0.2 M NaCl in 20 mM sodium citrate buffer, pH 5.5, or 20 mM BES buffer, pH 7.5, for a few hours and then flash-frozen in liquid nitrogen.
Heavy atom derivatives were obtained by soaking the crystals for 3 days in solutions of mother liquor containing 3 mM heavy atom compound. Among the over 20 different heavy atom compounds tested, 10 gave acceptable diffraction but only 4 could be used as isomorphous derivatives. The complex structures were obtained by soaking HCgp-39 crystals in the presence of 5-10 mM chitin oligosaccharides (GlcNAc 2 , GlcNAc 4 , GlcNAc 5 ) in 25% (w/v) polyethylene glycol 8000, 0.2 M NaCl, 20 mM BES-NaOH, pH 8.0, for about 15 h at room temperature.
Data Collection and Processing-Diffraction data for the native data set were collected at the ELETTRA Synchrotron, Trieste, Italy. To attempt structure determination by the MAD method, data on the (C 2 H 5 ) 3 HgPO 4 derivative was collected at beamline BM14 at the European Synchrotron Radiation Facility, Grenoble, France, with the wavelengths tuned to 1 ϭ 1.005 Å and 2 ϭ 0.775 Å ( Table I). The other derivative data sets were collected at beamline X31 at the European Molecular Biology Laboratory outstation at Deutsches Elektronen-Synchrotron, Hamburg, Germany, with the wavelength tuned to ϭ 0.99 Å, which, although not optimal, allowed the detection of the anomalous signal of the cis-Pt(NH 3 ) 2 Cl 2 derivative (Table I). X-ray diffraction data on HCgp-39 crystals soaked with (GlcNAc) 5 were collected in-house on an ENRAF-Nonius CuK␣ rotating anode generator equipped with a MacScience DIP2030 image plate detector. Data on (GlcNAc) 4 -and (GlcNAc) 2 -complexed HCgp-39 were collected at the ESRF, Grenoble, at beamlines ID14 -1 and ID14 -2, respectively. All data were integrated and merged using the programs DENZO and SCALEPACK (30). Programs from the CCP4 suite (31) were used for various calculations. Details on data collections and processing statistics are given in Tables  I and II.
Structure Determination and Refinement-Heavy atom position search, parameter refinement, and phases were calculated with programs from the PHASES package (32). The anomalous data were included in the calculations. The initial phases had a figure of merit of 0.40. Next, the non-crystallographic symmetry operators were determined by manual superposition of the heavy atom sites and refined with the program IMP of the O package (33). The program DM (34) was used to improve the initial phases by applying solvent flattening, histogram matching, and 4-fold non-crystallographic symmetry averaging.  (35). 10% of the reflections, selected in very thin resolution shells, was set aside for R free factor calculations. Non-crystallographic symmetry restraints were imposed throughout the whole refinement procedure. Bulk solvent correction and overall anisotropic B-factor corrections were included from the beginning resulting in an R factor of 23.5% and R free of 25.8%. Incorporation of experimental phase information during refinement resulted in improved R factor and R free values of 21.6 and 24.0%, respectively, for data to 2.6 Å. The final model consists of 4 ϫ 362 residues, 4 ϫ 2 N-acetylglucosamine residues (N-linked glycosylation), and a total of 117 water molecules. The stereochemical quality was assessed with the program PROCHECK (36). Refinement statistics are given in Table II.
The structures of HCgp-39 in complex with chitin fragments were solved by standard methods starting from the coordinates of the native HCgp-39 structure determined at 2.6 Å resolution. To optimize the position of the 4 molecules in the asymmetric unit a three-dimensional rigid body refinement of the native HCgp-39 model was performed using data up to 3.5 Å. Subsequently the model was adjusted using a round of simulated annealing. At this stage the electron density corresponding to the soaked ligands was visible in the chitin-binding groove from both 2F o Ϫ F c and F o Ϫ F c Fourier maps. The refinement, carried out with the program CNS (Crystallography & NMR System) (35), was followed by monitoring the behavior of the free R factor calculated from 10% of the data. The ␤(1, 4)-linked N-acetylglucosamine residues were fitted in the electron density using the program O (33). After a few rounds of x-ray-restrained energy minimization followed by B-factor refinement, water molecules were added. Non-crystallographic symmetry restraints were kept for the protein core throughout the whole refinement procedure. Refinement statistics are reported in Table II. Coordinates and structure factors have been deposited in the protein data bank with entry codes 1NWR, 1NWS, 1NWT, and 1NWU.

RESULTS AND DISCUSSION
A recombinant glycosylated form of the human cartilage glycoprotein-39 was crystallized. The structure was determined using the multiple isomorphous replacement method supplemented with anomalous scattering from three heavy atom derivatives (Table II). The native structure of HCgp-39 was determined at 2.6 Å. The final model comprises 4 HCgp-39 monomers (residues 22-383) and 117 water molecules. The 4 crystallographically independent molecules are essentially identical, having a root mean square difference of 0.02 Å for 361 equivalent C␣ atoms. In a Ramachandran plot 91.4% of the 1252 non-glycine and non-proline residues are in the most favored regions, 8.0% lie in the additional allowed sections, and the remaining 0.6% in the generously allowed parts.
In the crystal structure of HCgp-39 the first 21 N-terminal residues that form the leader sequence are not present. Disulfide bonds are formed between Cys 26 and Cys 51 and between Cys 300 and Cys 364 . Three cis peptide bonds are present (resi-  Carbohydrate Binding-A distinguishing feature of HCgp-39 is its ability to bind to insoluble chitin (40). By soaking HCgp-39 crystals in the presence of chitin fragments we could determine the structures of HCgp-39 in complex with (Glc-NAc) 2 , (GlcNAc) 4 , and (GlcNAc) 5 (Table II, Fig. 2). These structures allowed the identification of 9 sugar-binding subsites in the 43 Å groove, across the C-terminal ends of the ␤-strands of the (␤/␣) 8 barrel. They were numbered from Ϫ6 to ϩ3 from the non-reducing end, as is the current convention (41), with the Ϫ1 and ϩ1 subsites corresponding to the ones immediately upstream and downstream from the scissile bond in chitinases (Figs. 2 and 3). The carbohydrate-binding groove of HCgp-39 is lined with aromatic residues, which have hydrophobic stacking interactions with the hydrophobic sides of the bound sugar rings (Fig. 3). Among them, Trp 99 and Trp 352 additionally participate in ligand recognition by making hydrogen bonds with the C3-OH of the Ϫ1 GlcNAc and the C7-O of the Ϫ2 GlcNAc, respectively. The ␣ ϩ ␤ domain also takes part in sugar binding, with Arg 263 making hydrogen bonding interactions with the C6-OH of the Ϫ1 GlcNAc. Specific recognition of GlcNAc is achieved at subsites Ϫ5, Ϫ2, and ϩ2. Binding specificity at subsite Ϫ5 is determined by a hydrogen-bonding interaction between the Glu 70 -O⑀ and the nitrogen atom of the N-acetyl group of the bound GlcNAc. At subsite Ϫ2 the binding of GlcNAc is stabilized by a complex hydrogen-bonding network involving Trp 352 , Asn 100 , and water mediated interactions with Arg 263 , Glu 290 , and Gly 97 . Finally, the N-acetyl group of ϩ2 GlcNAc is hydrogen-bonded, via a water molecule, to the side chain hydroxyl groups of Thr 184 and Ser 179 . A more detailed description of protein-carbohydrate interactions is given in Fig.  3.
The presence of chitin fragments in the binding groove does not cause drastic conformational changes in the protein. The only exception is the rotation of the side chain of Trp 99 to make a parallel stacking interaction with the ϩ1 sugar ring (Figs. 3  and 4). This interaction is critical for stabilizing the structure of the chitin oligosaccharide, which becomes bent and twisted upon binding (Fig. 3). In family 18 chitinases, including the human chitotriosidase (42), the residue corresponding to Trp 99 is already in this chitin-binding orientation in the uncomplexed enzyme. Because of the difference in orientation of Trp 99 , the chitin-binding groove is wider in HCgp-39 than in human chitotriosidase. This could explain why the 39-kDa form of human chitotriosidase, unlike HCgp-39, cannot bind to insoluble chitin (40). Interestingly, in the plant chitinase hevamine, which acts as a defense endochitinase, the residue corresponding to Trp 99 is a glycine, a substitution that ensures a completely open chitin-binding groove and full accessibility of the central subsites located near the catalytic residues.
HCgp-39 is a lectin, and bound carbohydrates are not hydrolyzed. Interestingly, the binding mode of chitin to HCgp-39 is similar to that observed in productive chitin-chitinase complexes. For instance, we see that the Ϫ1 N-acetylglucosamine residue has a boat conformation as occurs in chitin when bound to bacterial chitinases (43,44). The distorted conformation, visible in the (GlcNAc) 4 and (GlcNAc) 5 complex structures, is stabilized by multiple interactions involving Trp 99 , Trp 352 , Arg 263 , and Tyr 141 (Fig. 3). Intriguingly, although 6 subsites, from Ϫ1 to Ϫ6, are available for chitin oligosaccharides to bind in a linear fashion this distorted conformation is preferred. Apparently, binding of GlcNAc to residues Ϫ2 to ϩ2 is energetically more favorable than binding to sites from Ϫ1 to Ϫ4. Indeed, the protein-carbohydrate interactions at subsites Ϫ3 a R free corresponds to R factor calculated using 10% of the total reflections selected randomly in thin shells and excluded during refinement. b The numbers in parentheses refer to the highest resolution bin. and Ϫ4 are not very strong (see below). This is in contrast to what has been observed in hen egg white lysozyme where a ␦-lactone derived from (GlcNAc) 4 occupies subsites Ϫ1 to Ϫ4 (45). Furthermore, the N-acetyl carbonyl oxygen atom of the Ϫ1 GlcNAc bound to HCgp-39 is oriented away from the C1 atom, whereas in chitinase-substrate complexes the oxygen atom is close to the C1 carbon as expected for the substrate-assisted catalytic mechanism of family 18 chitinases (39,46). This difference results from the E140L and D138A substitutions that in HCgp-39 eliminate the essential catalytic residues and create a large hydrophobic cavity where the N-acetyl group can be accommodated in extended conformation, causing total impairment of the chitinase activity. The structure of HCgp-39 in complex with (GlcNAc) 5 allowed unambiguous identification of subsites Ϫ3 to ϩ2. However, F o ϪF c and 2F o ϪF c Fourier maps, calculated with the refined model including 5 GlcNAc residues bound to HCgp-39 showed residual density features at both extremities of the built-in chitopentaose (Fig. 2). 1 N-acetylglucosamine residue was added to the oligosaccharide at the ϩ3 site resulting in a slightly better R free factor and a better appearance of the electron density map. Consequently the occupancy values were corrected to 0.5 for the residues in the Ϫ3 and ϩ3 sites. These results suggest that in HCgp-39 a chitopentaose molecule binds in subsites Ϫ2 to ϩ3 and subsites Ϫ3 to ϩ2 with comparable frequency. Subsequent examination of the electron density maps revealed more uninterpreted features beyond the identified six sites, indicative of additional binding sites for the chitopentaose in HCgp-39 (Fig. 3). Similar results were obtained from soaking studies with (GlcNAc) 4 , where the initial maps showed the strongest electron density in subsites Ϫ2 and Ϫ1. Nevertheless, residual electron density, although not clearly interpretable, was visible along the whole binding groove especially upstream subsite Ϫ3 (Fig. 3).
Compelling experimental evidence for occupation of subsites Ϫ5 and Ϫ6 came from the structure of HCgp-39 with chitobiose ((GlcNAc) 2 ) (Fig. 3). In these subsites Trp 71 , Glu 70 , and Tyr 34 form a flat area on the protein surface on which the disaccharide can be accommodated. Initial electron density maps showed that binding of (GlcNAc) 2 also occurs in the middle part of the groove, although with lower occupancy (Fig. 2). Because of the poor quality of the electron density we have refrained from building any ligand in those subsites. Soaking of HCgp-39 crystals in the presence of (GlcNAc) 3 produced similar results (data not shown). Finally, HCgp-39 is glycosylated and although Asn 60 is in the vicinity of the groove, the N-linked glycan, which was left intact during the crystallization, does not interfere with chitin binding in the crystal.
The described HCgp-39 crystallographic complexes clearly identify subsites Ϫ3 to ϩ3, Ϫ5, and Ϫ6 in the carbohydratebinding groove. Combining the experimental structures yielded a model in which an oligosaccharide made up of 9 GlcNAc residues is bound to HCgp-39 ( Figs. 2 and 3). The model was obtained by inserting a ␤(1, 4)-linked GlcNAc at position Ϫ4. At subsite Ϫ4, no aromatic side chains are available to stabilize the binding, but the side chain of Glu 70 could form a hydrogen bond interaction with the C6-OH of the GlcNAc residue.
Our results show that binding of either long or short oligosaccharides to HCgp-39 is possible. Remarkably, whereas chitin fragments of 4 or more GlcNAc residues tend to occupy the central part of the groove, shorter oligosaccharides bind preferentially at the more distant subsites on the protein surface. The existence of two distinct binding sites with selective affinity for long and short oligosaccharides has never been documented for chitinases or for chitinase-like proteins and could be unique for HCgp-39. This suggests the possibility that HCgp-39 like many other lectins could function by cross-linking two targets. Although, because the 2 binding sites are only about 10 Å apart, the simultaneous binding of 2 glycoproteins is questionable. Alternatively, ligand binding to the two sites could be either mutually exclusive and thus provoke different effects or synergic, with the surface site recruiting ligands that are subsequently fixed in the central part of the groove. However, the binding of a chito-oligosaccharide from subsites Ϫ6 to ϩ3 may also occur, although the occurrence of such a long chito-oligosaccharide in mammals has not yet been demonstrated.
The observation that chitin oligosaccharides bind to HCgp-39 suggests interesting speculations. It has been found in vertebrates that short chito-oligosaccharides are used as primers for the synthesis of hyaluronic acid (HA) and that they are kept at the reducing ends of the nascent glycans (47,48). HA is composed of repeating disaccharide units of N-acetylglucosamine and D-glucuronic acid linked together by alternat- ing ␤(1, 4) and ␤(1, 3) glycosidic bonds. It is a structural component of hyaline cartilage, synovial joint fluid, and skin tissues. HA plays an important biological role in embryogenesis, cell proliferation, and tissue remodeling, and it is involved in acute and chronic inflammation processes (49). The expression of HCgp-39 in vivo has been related to similar events (7,50,51). Thus the function of HCgp-39 could be linked to the functions of HA. Because of its chitin-binding ability HCgp-39 could participate in specific signaling processes by perceiving the presence of newly synthesized HA chains that still contain the chito-oligosaccharide.
Furthermore it has been reported that both the bovine and porcine homologs of HCgp-39 (52,53) are able to bind to heparin. HCgp-39 presents a cluster of basic amino acid residues, GRRDKQH, at position 143-149. Similar consensus sequences have been found in heparin-binding proteins (54). In HCgp-39 this putative heparin-binding site is located in a surface loop (Fig. 4). However, soaking of HCgp-39 crystals or co-crystallization in the presence of fully sulfated heparin fragments and heparin-like compounds did not result in conclusive evidence of binding at this site. Likewise, attempts to model the trisulfated disaccharide unit, -IdoA(2-OSO 3 )-GlcNSO 3 (6-OSO 3 )-, abundant in heparin, into the HCgp-39 structure proved unsuccessful. Instead, the structurally related heparan sulfate, which bears less sulfate groups could be a more likely ligand of HCgp-39. Heparan sulfate proteoglycans are present at the cell surface and in the extracellular matrix where they play a role in growth factor/cytokine action, cell adhesion, organogenesis, and wound healing. Heparan sulfate is structurally heterogeneous depending on the producing cell type, and it contains GlcNAc and N-unsubstituted glucosamine units (55), the presence of which has also been implicated in the ability to bind to L-selectins (56). Unlike heparin, heparan sulfate unsulfated fragments can be accommodated in the binding groove of HCgp-39. A similar function has been proposed for the related protein YM1, which showed specificity for glucosamine oligosaccharides (21).
Relationships with Other Chitinase-like Proteins-The crystal structures of HCgp-39 in complex with chitin also help to evaluate the carbohydrate-binding properties of other mammalian CLPs. A multiple sequence alignment indicates extensive sequence similarity among the mammalian CLPs. However, of all residues participating in chitin binding in HCgp-39 only Trp 31 and Trp 352 are highly conserved. All residues in HCgp-39 that are involved in making hydrogen-bonding interactions with the bound chitin are conserved in chitotriosidase (24), as well as in MGP40 (27), in bovine CLP1 (52), and in porcine gp38k (53) indicative of a similar ligand specificity for these proteins.
Comparison of the available three-dimensional structures of mammalian CLPs shows high structure similarity apart from small differences in the proximity of the binding cleft. In particular the crystal structure of MGP40 (27) can be superimposed onto HCgp-39 with a root mean square difference of 0.75 Å for 358 equivalent C␣ pairs. All chitin-binding residues, including Trp 99 , are conserved in both structures at the same position. Based on structure and sequence homology (83% identity) it can be concluded that MGP40 is the equivalent of HCgp-39 expressed by goat cells.
The structure of the imaginal disc growth factor-2 from Drosophila (29) (15.7% sequence identity) can be superimposed onto HCgp-39 with a root mean square difference of 1.57 Å for 327 equivalent C␣ atoms. Although it has been proposed that both imaginal disc growth factor-2 and HCgp-39 could function by stimulating cell proliferation through a similar signaling pathway involving also the binding to an insulin-like receptor (18,29), the sequence and structural features indicates that these proteins are likely to display different carbohydrate specificities. In fact, neither the tryptophan residues lining the carbohydrate-binding groove nor the polar residues determining chitin specificity in HCgp-39 and chitotriosidase are conserved in imaginal disc growth factor-2.
Conclusions-In the present study we determined the structure of HCgp-39 in complex with chitin fragments of different length by means of protein x-ray crystallography. Our structures show that HCgp-39 is able to bind to chitin in a similar fashion as seen in family 18 chitinases. Sugar binding occurs via van der Waals interactions with the side chains of aromatic amino acid residues and several hydrogen-bonding interactions involving the sugar-hydroxyl groups. Chitin oligosaccharides fit strikingly well in the binding groove of HCgp-39. However, the N-acetyl group, which is characteristic of GlcNAc, is only specifically recognized at subsites Ϫ2 and Ϫ5. Furthermore, we show that in HCgp-39 binding specificity also depends on the length of the chito-oligosaccharide. Although short chitin fragments bind at one end of the groove longer oligosaccharides bind in a distorted manner to the central subsites. The functional significance of this unique property needs further investigation.
HCgp-39 stimulates cell proliferation, and it is secreted during inflammation and in response to tissue injury. However the exact role of HCgp-39 and the nature of its physiological ligand remain unknown at present. Because of its chitin-binding ability HCgp-39 could function in host defense mechanisms against chitin-containing pathogens. Alternatively we propose that HCgp-39 could bind to chitin-like oligosaccharides present in nascent chains of hyaluronan, which is also involved in tissuerepairing events and inflammation. Modeling of alternative candidate ligands revealed that although binding of heparin in the HCgp-39 groove appears unlikely, fragments of unsulfated heparan sulfate can be modeled in place of chitin. The implications of carbohydrate binding for the biological function of HCgp-39 await further study.
Finally, HCgp-39 is a candidate autoantigen in RA. Although no evidence has yet been found of a direct involvement of anti-HCgp-39 antibodies in RA onset, it has been shown that immune response to the HCgp-39 epitopes 259 -271 and 263-275 correlates with disease activity (2,4,5). The structure of HCgp-39 shows that although region 266 -275 maps onto the protein surface, residues 259 -265 are buried in the vicinity of the chitin-binding groove (Fig. 4).