The Mammalian Chitinase-like Lectin, YKL-40, Binds Specifically to Type I Collagen and Modulates the Rate of Type I Collagen Fibril Formation*

YKL-40 is expressed in arthritic cartilage and produced in large amounts by cultured chondrocytes, but its exact role is unclear, and the identities of its physiological ligands remain unknown. Purification of YKL-40 from resorbing bovine nasal cartilage and chondrocyte monolayers demonstrated the existence of three isoforms, a major and minor form from resorbing cartilage and a third species from chondrocytes. Affinity chromatography experiments with purified YKL-40 demonstrated specific binding of all three forms to collagen types I, II, and III, thus identifying collagens as potential YKL-40 ligands. Binding to immobilized type I collagen was inhibited by soluble native ligand, but not heat-denatured ligand, confirming a specific interaction. Binding of the chondrocyte-derived species to type I collagen was also demonstrated by surface plasmon resonance analysis, and the dissociation rate constant was calculated (3.42 × 10-3 to 4.50 × 10-3 s-1). The chondrocyte-derived species was found to prevent collagenolytic cleavage of type I collagen and to stimulate the rate of type I collagen fibril formation in a concentration-dependent manner. By contrast, the cartilage major form had an inhibitory effect on type I collagen fibrillogenesis. Digestion with N-glycosidase F, endoglycosidase H and lectin blotting did not reveal any difference in the carbohydrate component of these two YKL-40 species, indicating that this does not account for the opposing effects on fibril formation rate.

Immunohistochemical studies have shown that YKL-40 is present in the cytoplasm of arthritic chondrocytes, but not in the surrounding cartilage matrix (36,37), except where the cartilage has been invaded by synovium (37). In osteoarthritic cartilage, YKL-40 positive chondrocytes are more abundant when histopathological changes are evident and they are also particularly prominent in areas with a considerable biomechanical load (36). In early OA, YKL-40 expression is confined primarily to the superficial zone, but spreads to the remaining middle and deep zones as disease progresses (48). It is absent from healthy cartilage from young adults, but sparse expression is sometimes found in healthy cartilage from older adults (36), possibly reflecting deterioration of the tissue with age. Chondrocytes appear to rapidly synthesize large amounts of YKL-40 in response to an altered biochemical and/or biomechanical environment, because newly established explant cultures of healthy human cartilage do not produce YKL-40, but secretion as a major protein occurs after a few days of culture (4,11). It is also a prominent secretory product of cultured human articular chondrocytes (4,11,19), accounting for 33% of the conditioned medium protein (19).
The exact functions of YKL-40 in joint destruction and connective tissue turnover in general, remain unclear. In particular, there is a lack of knowledge regarding the identity of physiological ligands. A vertebrate enzyme (DG42) has been identified which may catalyze the synthesis of chito-oligosaccharides (49 -51) and it is also possible that short chito-oligosaccharides are used as primers for hyaluronan synthesis (49,51). If these primers exist, they may be retained at the reducing ends of the molecules, allowing them to be recognized by YKL-40. YKL-40 also binds heparin (2,3,5,12), suggesting that it could interact with heparan sulfate proteoglycans (HSPGs) in vivo. Recently, YKL-40 has been shown to stimulate the proliferation of connective tissue cells through activation of mitogen-activated protein kinase and protein kinase B-mediated signaling pathways (52,53). The cellular receptors responsible for mediating these effects have not yet been identified and cell surface HS moieties are potential candidates for this role. YKL-40 also stimulates the directional migration and tubulogenesis of endothelial cells (54) and the adhesion and migration of vascular smooth muscle cells (55) and dampens the response of chondrocytes and synovial cells to inflammatory cytokines (56). However, the molecules with which it interacts to elicit these effects are also unknown. A clearer understanding of the exact physiological ligands with which YKL-40 interacts is therefore critical to uncovering its precise role and functions in vivo. Although carbohydrate structures are implicated by the existing evidence, it has not yet been conclusively demonstrated that these actually serve as YKL-40 ligands, either in cell culture models, or in vivo and the existence of protein ligands has not been reported. In this study, we describe the ability of YKL-40 to bind collagen, thus identifying this important extracellular matrix (ECM) macromolecule as a novel, non-carbohydrate ligand. The physiological role of this interaction is also investigated.
NaCl was added to a final concentration of 1 M to crude medium and 20 ml was applied to a phenyl-Sepharose column (V t ϭ 20 ml) equilibrated in 1 M NaCl at 21°C. The column was washed with equilibration buffer and eluted at 4°C with 50 mM NaCl. The elution fractions were pooled and NaCl added to 0.5 M. A concanavalin A-Sepharose column (V t ϭ 3 ml) was equilibrated in 0.5 M NaCl; the sample was applied, washed with equilibration buffer and eluted with 0.5 M NaCl/20% methyl ␣-D-glucopyranoside. Pooled eluted fractions were dialyzed into 50 mM NaCl and applied to DEAE-Sephacel (V t ϭ 2 ml). The flow-through and wash were applied to heparin-Sepharose (V t ϭ 1 ml), washed with 50 mM NaCl, and eluted with 1 M NaCl. Pooled eluted fractions were chromatographed on an Ultrogel AcA44 column (V t ϭ 170 ml, 85 ϫ 1.6 cm) in 1 M NaCl at 6.0 ml h Ϫ1 collecting 1.1-ml fractions. Two YKL-40-containing peaks were eluted, which were pooled separately, dialyzed into 50 mM NaCl and concentrated over heparin-Sepharose. The final heparin-Sepharose elution pools were dialyzed into cacodylate buffer with 50 mM NaCl and quantitated by BCA assay (Pierce). Purity was demonstrated by SDS-PAGE silver-staining, and the identity of both pools (major and minor) was confirmed by high pressure liquid chromatography tandem mass spectrometry (HPLC MS/MS) analysis.
Bovine Nasal Chondrocyte Extraction and Culture-Bovine nasal cartilage was cut into small pieces, washed with Dulbecco's phosphate-buffered saline (PBS) and digested on an orbital shaker with hyaluronidase (Sigma, 1 mg/ml in PBS, 15 min, 37°C), trypsin (Sigma, 0.25% w/v in PBS, 30 min, 37°C) and bacterial collagenase (from Clostridium histolyticum, Sigma, 3 mg/ml in growth medium, see below, 15-20 h, 37°C). Cells were harvested from the digested tissue supernatant by centrifugation (217 ϫ g, 10 min) and seeded into 8 T162 flasks (Corning Inc.) at 6 ϫ 10 6 cells/flask in growth medium (Dulbecco's modified Eagle's medium containing 25 mM HEPES, 10% (v/v) heat-inactivated fetal bovine serum (Perbio Science), 2 mM L-glutamine, 200 units ml Ϫ1 penicillin, 200 g ml Ϫ1 streptomycin, 40 units ml Ϫ1 nystatin). Confluent cultures were passaged 1:2, regrown to confluence, and incubated in serum-free medium (other additions as above except 100 units ml Ϫ1 penicillin, 100 g ml Ϫ1 streptomycin) for 24 h after washing three times with Dulbecco's PBS to remove serum proteins. Cells were then stimulated with IL-1␣ (1 ng ml Ϫ1 ) and OSM (10 ng ml Ϫ1 ) in serum-free medium for 72 h, followed by collection of the conditioned medium. Three further 72-h re-stimulations of the same cells and collection of the conditioned medium were also performed. Sodium azide and Brij-35 (0.02%/0.05% w/v final concentrations, respectively) were added to the conditioned medium, which was filtered to remove cell debris. All tissue culture reagents were from Invitrogen, unless otherwise stated.
Purification of YKL-40 from Cytokine-stimulated Bovine Nasal Chondrocytes-1140 ml of conditioned medium was mixed overnight at 4°C with 10 ml of heparin-Sepharose equilibrated in cacodylate buffer with 50 mM NaCl followed by collection of the flow-through. The resin was washed with equilibration buffer and eluted with 1 M NaCl. The first two elution fractions were pooled (20 ml) and applied to an Ultrogel AcA44 column (V t ϭ 1300 ml, 80 ϫ 3.2 cm) equilibrated in 1 M NaCl and run at 40 ml h Ϫ1 collecting 10-ml fractions. SDS-PAGE silver stains of the fractions revealed a single YKL-40-containing peak, which was pooled and desalted to 50 mM NaCl by dialysis followed by concentration on a heparin-Sepharose column (V t ϭ 5 ml). The final heparin-Sepharose elution pool was dialyzed into cacodylate buffer with 50 mM NaCl and quantitated by BCA assay. Purity was demonstrated by SDS-PAGE silver staining and identity confirmed by HPLC MS/MS analysis.
Production of Recombinant Matrix Metalloproteinase-1 (MMP-1) and Tissue Inhibitor of Metalloproteinases-2 (TIMP-2)-The cDNA for human proMMP-1 was generously provided by Dr. Alan Galloway (British Biotech) as 0.3-kb and 1.7-kb EcoRI fragments in pUC19. Plasmid DNA containing the 0.3-kb fragment was partially digested with EcoRI, dephosphorylated, and purified. This linearized plasmid was then ligated to the 1.7-kb EcoRI fragment to generate pUC19/ proMMP-1. Correct orientation was verified by restriction digest with XmnI. PCR was performed on pGEX2T/MMP-1 (58) to generate a DNA fragment representing the stabilized form of active MMP-1 (58). This PCR fragment was XmnIdigested, gel-purified, and the larger 3Ј-fragment ligated to AatII/XmnI-digested pUC19/proMMP-1. The ligation product was purified, and PCR performed with Pfu DNA polymerase and T4 PNK-treated primers to generate a fragment representing full-length, stabilized proMMP-1. The resulting PCR product was purified and ligated with dephosphorylated SmaI-digested pUC18 (Amersham Biosciences) to generate pUC18/ proMMP-1. Correct orientation was assessed by digestion with NdeI, presence of the internal mutation to improve proteinase stability by BglII digestion (58), and the DNA sequence verified by dideoxy sequencing. The proMMP-1 sequence was then subcloned into the expression vector pRSETA (Invitrogen), which was subsequently transferred to HMS174(DE3)pLysS or BL21(DE3)pLysS Escherichia coli cells (Invitrogen) for expression purposes. Inclusion bodies were prepared from isopropyl-1-thio-␤-D-galactopyranoside-induced 500-ml cultures using Bugbuster reagents (Novagen) and resuspended in 20 mM Tris-HCl, pH 8.0, 6 M guanidine-HCl, 0.02% (w/v) sodium azide, 5 mM dithiothreitol followed by refolding of denatured protein essentially as previously described (59). Refolding caused activation of proMMP-1 to the fully active form or conversion to an intermediate lacking the polyhistidine tag and the first 4 amino acids of the proenzyme. Activation with p-aminophenylmercuric acetate (APMA) resulted in conversion of the intermediate form to the fully active enzyme. Refolded MMP-1 was desalted by dialysis and concentrated by heparin-Sepharose chromatography. Enzyme assays (see below) were performed without APMA activation and therefore measured only the fully active form present in the refolded sample. Recombinant TIMP-2 was purified by Ultrogel AcA44 gel filtration and heparin-Sepharose chromatography from COS cell conditioned medium (supplied by British Biotech).
Protein Identification by HPLC MS/MS Analysis-In-gel digestion with trypsin was performed in a robotic digestion system (Investigator ProGest, Genomic Solutions) as previously described (60). Tandem electrospray mass spectra were recorded using a Q-Tof hybrid quadrupole/orthogonal acceleration time of flight spectrometer (Micromass) interfaced to a Micromass CapLC chromatograph. Samples were dissolved in 0.1% aqueous formic acid and introduced into the spectrometer via a Pepmap C18 column (300 m ϫ 0.5 cm, LC Packings) and were eluted with an acetonitrile/0.1% formic acid gradient (5-70% acetonitrile over 20 min).
The capillary voltage was set to 3,500 V and data-dependent MS/MS acquisitions were performed on precursors with charge states of 2, 3, or 4 over a survey mass range 400 -1300. Proteins were identified by correlation of uninterpreted tandem mass spectra to entries in Swiss-Prot/TrEMBL, using Pro-teinLynx Global Server (Version 1.1, Micromass) (61). The data base was created by merging the FASTA format files of Swiss-Prot, TrEMBL, and their associated splice variants. No taxonomic, mass or pI constraints were applied. One missed cleavage per peptide was allowed, and the fragment ion mass tolerance window was set to 100 ppm. Cystines were assumed to be carbamidomethylated, but other potential modifications were not considered in the first pass search. All matching spectra were reviewed by an expert and where necessary, sequences were verified manually using the MassLynx program Pepseq (Micromass). Additional data were also acquired at a commercial facility (University of York) using an Applied Biosystems 4700 Proteomics Analyzer (MALDI-TOF/TOF).
Collagen-Sepharose Chromatography-Highly pure bovine skin type I collagen was prepared as previously described (62,63); bovine articular type II collagen and bovine type III collagen were generous gifts from collaborators. Type I collagen (1 or 2.5 mg ml Ϫ1 ), type II collagen (0.5 mg ml Ϫ1 ), and type III collagen (1 mg ml Ϫ1 ) were coupled to cyanogen bromide-activated Sepharose, essentially as previously described (64). A type I collagen-Sepharose minicolumn (V t ϭ 500 l) was equilibrated in 25 mM sodium cacodylate, pH 7.2, 10 mM CaCl 2 , 0.02% (w/v) sodium azide, 0.05% (w/v) Brij-35, followed by application of resorbing cartilage conditioned medium (500 l). The column was washed with equilibration buffer (3ϫ V t ) and eluted with equilibration buffer ϩ 1 M NaCl (4ϫ V t ). For type II and type III collagen, 50-l minicolumns were equilibrated in the above buffer followed by application of 50 l of resorbing cartilage conditioned medium. The columns were washed (5 ϫ V t ) and eluted (5 ϫ V t ) as above.
To demonstrate the purity of the type I collagen used for coupling, ϳ25 g (in 50 mM Tris-HCl, pH 7.6, 200 mM NaCl, 0.02% w/v sodium azide, 1 M glucose) was digested for 72 h at 37°C with buffer alone, bacterial collagenase (0.5 g), or recombinant human MMP-1 (1.6 g) in the additional presence of 1 mM CaCl 2 . The reactions were applied to 6.5% SDS-PAGE gels and visualized by staining with Coomassie Brilliant Blue.
Collagen-Sepharose Binding Assays-For type I collagen, 400 ng of YKL-40 was applied to minicolumns (V t ϭ 40 l) equilibrated as for "Collagen-Sepharose Chromatography" followed by washes with equilibration buffer (5 x V t ) and elution with equilibration buffer ϩ 1 M NaCl (5 ϫ V t ). Fraction samples (12 l) were run on 12.5% SDS-PAGE gels and silver-stained. Identical control experiments using cyanogen bromide-activated Sepharose without a ligand attached and gelatin-Sepharose (Amersham Biosciences) were also performed. For type II and type III collagen, 40 ng of YKL-40 was applied to 50-l columns, and 36-l fraction samples were analyzed on gels. Identical control experiments using ligand-free Sepharose were again performed.
In control experiments, YKL-40 was mixed with denatured type I collagen (gelatin, 4 g) or buffer alone (50 mM Tris-HCl, pH 7.6, 200 mM NaCl, 0.02% w/v sodium azide). The collagen was denatured by incubation at 56°C for 30 min. As an additional control, YKL-40 was also applied to ligand-free Sepharose. The columns were equilibrated, washed (3 ϫ V t ), and eluted (4 ϫ V t ) as before, and 12-l fraction samples were analyzed on gels.
To demonstrate specific binding of chondrocyte-derived YKL-40 to the collagen ligand, type I collagen-Sepharose minicolumns (V t ϭ 100 l) were incubated for 2 h at 37°C with buffer alone (25 mM sodium cacodylate, pH 7.5, 0.02% w/v sodium azide , 0.05% w/v Brij-35) or bacterial collagenase (20 g), prior to application of YKL-40. Columns were equilibrated, washed, and eluted, followed by analysis of the fractions, as for the competition assays.
Surface Plasmon Resonance-Assays were performed using a Biacore 2000 instrument. Type I collagen was coupled to a CM5 sensor chip via primary amine groups in 20 mM sodium acetate, pH 4.0. To activate the surface, a 7-min pulse of 50 mM N-hydroxysuccinimide/200 mM N-ethyl-NЈ-dimethylaminopropyl carbodiimide was used. Type I collagen (100 g ml Ϫ1 ) was immobilized to give 765 resonance units of coupled protein, and excess reactive groups were then blocked with 1 M ethanolamine. A control flow cell was activated and blocked in the same manner, without collagen injection. Measurements were carried out in cacodylate buffer with 50 mM NaCl at 25°C with the chondrocyte-derived YKL-40 species. YKL-40 (1.00, 0.875, 0.75, 0.50, 0.25, and 0.10 M) was injected at 10 or 30 l min Ϫ1 for 2 min, and dissociation was initiated by replacing the YKL-40 with binding buffer. Regeneration of the chip was attempted with six different conditions: 1) cacodylate buffer with 1 M NaCl, 2) 20 mM sodium acetate, pH 4.0, 3) 20 mM sodium acetate/1 M NaCl, pH 3.8, 4) 50 mM glycine-HCl, pH 2.0, 5) 0.2% (w/v) SDS, or 6) 10 mM HCl. The first four conditions were unsuccessful because of lack of analyte dissociation; 10 mM HCl appeared to cause partial hydrolysis of the collagen ligand, and 0.2% (w/v) SDS was abandoned because of possible ligand denaturation. Dissociation was therefore allowed to proceed until the resonance units returned to baseline; the next cycle of binding and dissociation was then performed. Data from the control flow cell were subtracted from the collagencoupled flow cell for each cycle. Eight binding interactions were performed per experiment using five YKL-40 concentrations. The data were fit to the simple 1:1 Langmuir isotherm, and the kinetic parameters (dissociation rate constant k d , association rate constant, k a and affinity constant K D ) were calculated using the BIAevaluation system 3.0.
Diffuse Fibril Assay for Collagenase Activity-The assay is a modification of that previously described by Cawston and Barrett (62) and was performed using 3 H-acetylated bovine type I collagen from calf skin (62,63). The concentration of the collagen was estimated by hydroxyproline assay, using a modification of existing methods (65,66). Ice-cold collagen (400 g ml Ϫ1 in 50 mM Tris-HCl, pH 7.6, 200 mM NaCl, 0.02% w/v sodium azide) was diluted to 20 g ml Ϫ1 in a final volume of 100 l with ice-cold cacodylate buffer with 50 mM NaCl containing recombinant human MMP-1 (diluted to 3.25 or 1.625 g ml Ϫ1 ) and/or purified YKL-40 from chondrocyte cultures (25, 50, or 100 g ml Ϫ1 ). Samples were digested at 37°C for 90 min followed by centrifugation (10,000 ϫ g, 15 min, 4°C). 33-l supernatant samples were counted for 1 min in a 1450 Microbeta Trilux liquid scintillation counter (Wallac) with 200 l of Optiphase Supermix scintillation fluid (Wallac). Counts in the supernatants above the level of the blanks (buffer only) represent soluble, enzyme-digested collagen. One unit of activity degrades 1 g of fibrillar collagen per minute at 37°C, and the assay range is ϳ0.10 -0.40 units ml Ϫ1 . Activity values were corrected with respect to the initial dilution of the MMP-1 sample, and all conditions were examined in triplicate. YKL-40 alone had no collagenolytic activity, but caused the counts in the supernatants to drop slightly compared with the buffer blanks (minus 5-15%). This effect was taken into account when calculating MMP-1 activity in the presence of YKL-40.
Fibril Formation Assays-Bovine type I collagen was prepared from calf skin, radiolabeled, and quantitated as for the diffuse fibril assay. Ice-cold collagen was diluted to 20 g ml Ϫ1 in a final volume of 100 l with ice-cold cacodylate buffer containing 50 mM NaCl or with the same buffer containing native (25, 50, or 100 g ml Ϫ1 ) YKL-40 from chondrocyte cultures or heat-denatured bovine serum albumin (BSA) (100 g ml Ϫ1 ). BSA was heat-denatured by boiling for 5 min. Fibril formation was performed at 37°C for 0, 5, 10, 20, 30, and 40 min, followed by centrifugation of the samples (10, 000 ϫ g, 15 min, 4°C). The supernatants were counted as for the diffuse fibril assay; counts in the supernatants represent soluble collagen which has not formed fibrils. All conditions were examined in triplicate.
A slightly modified procedure was used for assays with the purified major pool from cartilage. Ice-cold collagen was diluted to 19 g ml Ϫ1 in a final volume of 105 l with ice-cold cacodylate buffer containing 50 mM NaCl or with the same buffer containing native YKL-40 (5 or 50 g ml Ϫ1 ) or native BSA (50 g ml Ϫ1 ). The samples with 50 g ml Ϫ1 YKL-40 also contained 10 g ml Ϫ1 TIMP-2, to inhibit the very low level of collagenase present in this YKL-40 preparation and thereby prevent any collagenolytic cleavage of the collagen during the assay. Fibril formation was performed at 37°C for 0, 5, 10, 20, 40, and 60 min followed by addition of 50 l of ice-cold 100 mM Tris-HCl, pH 8.5, 15 mM CaCl 2, 0.02% w/v sodium azide and centrifugation (10, 000 ϫ g, 15 min, 4°C). 50 l of the supernatants were counted as above.
Purified, native YKL-40 samples were digested with N-glycosidase F using a kit (Roche Applied Science GmbH). The major and minor pools from cartilage (0.16 g and 0.13 g, respectively) and the chondrocyte pool (0.15 g), each in 10 l of cacodylate buffer with 50 mM NaCl were added to 10 l of the kit reaction buffer. Recombinant N-glycosidase F was reconstituted in nanopure water according to the manufacturer's instructions, and 10 l was added per reaction followed by digestion for 24 h at 37°C. Identical reactions without enzyme or without protein substrate were also performed. Digestions were terminated and analyzed as for endoglycosidase H.

Screening of Resorbing Cartilage Conditioned Medium for
Collagen-binding Proteins-Cartilage conditioned medium was screened for collagen-binding proteins using collagen-Sepharose chromatography and HPLC MS/MS analyses. Most of the cartilage proteins did not show any significant interaction with the type I collagen ligand and were therefore not found in the eluted material. The eluted proteins included bone/cartilage proteoglycan I precursor, fibronectin, MMP-1, MMP-3, YKL-40, and two further unidentified species (Fig. 1, A and B). Detection of YKL-40 in the column elute therefore indicates binding to type I collagen. Similar binding to type II and type III collagen-Sepharose was also seen (Fig. 1C ).
YKL-40 was identified from fourteen tryptic peptides, corresponding to 176 amino acid residues, which were sequenced by tandem mass spectrometry (Fig. 1B). Amino acid sequence coverage was thus obtained for 49% of the sequence, giving a secure identification. The most statistically significant hits retrieved by data base searching of the raw MS/MS data were bovine, ovine, and caprine YKL-40. The two best bovine matches were O18949, an N-and C-terminally truncated sequence from TrEMBL, and CH3L1_Bovin, the SwissProt entry for the fulllength protein. There were 12 amino acid differences between these two sequences in the 332 residue region of overlap. Nine of these conflicts occurred in peptides sequenced here (Fig. 1B), and in all cases, the experimental data supported the TrEMBL sequence (O18949), suggesting that this entry is more representative of protein purified from nasal cartilage.
Purification of YKL-40-In a crude mixture of proteins, binding to collagens could reflect an indirect interaction via a third protein; further investigations were therefore undertaken using purified YKL-40. YKL-40 was isolated from resorbing cartilage conditioned medium by a five-step procedure as described under "Experimental Procedures." Purification from resorbing cartilage has not been described before, although YKL-40 has been isolated from cultured chondrocytes (4,5,7). Previous purifications from whey proteins or cell culture medium used heparin affinity chromatography as the main or only separation step (2, 3, 5, 10, 12), but this approach is inadequate for cartilage medium, which contains a much more complex mixture of proteins, several of which bind heparin. In line with previous reports, DEAE ion exchange chromatography (4, 7) and gel filtration (2, 4, 7) were found to be useful, and two additional steps (phenyl-Sepharose and concanavalin A-Sepharose) were also necessary. Most of the contaminants were removed by the first four columns, leaving YKL-40 as the major protein in the AcA44 starting material. This gel filtration step removed nearly all the remaining impurities and also resolved the YKL-40 into two distinct elution peaks ( Fig. 2A). Most of the YKL-40 was in the second (major) peak, which was preceded by a smaller (minor) peak. These peaks were pooled separately and concentrated on heparin-Sepharose. Rechromatographing of the minor pool alone on the same gel filtration column confirmed that this eluted as a distinct peak and was not simply the leading edge of the major peak.
YKL-40 was also purified from the conditioned medium of IL-1␣/OSM-stimulated bovine nasal chondrocytes in monolayer culture by heparin-Sepharose and gel filtration chroma-tography. Gel filtration of the final purified pool indicated a similar elution profile to that of the cartilage major form (data not shown). However, unlike the cartilage major form, the chondrocyte-derived protein formed a precipitate under low salt conditions (50 mM). Precipitation was reversed by dialysis into 1 M NaCl, indicating that it involved electrostatic interactions and did not result from protein denaturation. This indicates a structural difference between the cartilage major form and the species from chondrocyte cultures. The chondrocyte-FIGURE 1. Identification of collagen-binding proteins from resorbing bovine nasal cartilage medium. A, conditioned medium (500 l) was applied to a type I collagen-Sepharose minicolumn (V t ϭ 500 l), followed by column washes and elution with 1 M NaCl. Eluted proteins (third and fourth column volumes) were concentrated 8.5-fold by trichloroacetic acid precipitation and run reduced on 6.5% and 12.5% SDS-PAGE gels. Protein bands were silver-stained and excised, followed by tryptic digestion and HPLC MS/MS analyses. Gels of the eluted proteins prior to trichloroacetic acid precipitation are shown, with identifications; M r , molecular mass markers in kDa. B, trypsin-digested fragments obtained from the YKL-40 identified in A are highlighted on the full-length sequence of bovine YKL-40 (without the signal peptide, O18949 combined with the first three and last 27 amino acids of CH3L1_Bovin). C, conditioned medium (50 l) was applied to type II and type III collagen-Sepharose minicolumns (V t ϭ 50 l), followed by column washes and elution with 1 M NaCl. Eluted proteins were separated on 12.5% SDS-PAGE gels under reducing conditions and silver-stained. The positions of molecular mass markers (M r ) in kDa and YKL-40 are indicated. derived YKL-40 is not identical to the cartilage minor form either, because these proteins do not comigrate during gel filtration. The purification of YKL-40 from resorbing cartilage therefore identified two YKL-40 isoforms, which were distinct to the species produced by cultured chondrocytes. The chondrocyte form is presumably the bovine equivalent of the single YKL-40 species isolated by others from these cells (4,5,7).
Each purified YKL-40 pool ran as a single band on SDS-PAGE silver stains, with an apparent molecular mass of 41.8 kDa (Fig. 2B, reducing conditions), consistent with published values (39 or 42 kDa, Refs. [2][3][4][5]. For each species, the mobility was similar under non-reducing and reducing conditions, as was previously shown for the porcine orthologue (1), even though structural data for the human protein indicate the presence of two disulfide bonds (6, 7). Co-migration of the two cartilage forms under non-reducing conditions shows that the minor form is not a disulfide-bonded multimer of the major form. Instead, a small proportion of each of the three pools appears to consist of a disulfide-bonded dimer (Fig. 2B). HPLC MS/MS analyses were used to confirm the identity of all three forms as YKL-40 (Fig. 2C). They are referred to as the cartilage major form, cartilage minor form and chondrocyte-derived species in the remainder of this report.
Purified YKL-40 Binds to Collagen Types I, II, and III-Affinity chromatography using type I, type II, and type III collagen-Sepharose was used to demonstrate a direct binding interaction between purified YKL-40 and these fibril-forming collagens. For type I collagen, similar results were obtained with both of the cartilage forms and the chondrocyte-derived species, in which YKL-40 protein was recovered essentially only in the elution fractions (Fig. 3). A weaker interaction was seen with ligand-free Sepharose, in which a large proportion of protein was also found in the flow-through and washes. Some nonspecific binding to the resin probably reflects the hydrophobic nature of YKL-40, which is evident from its ability to bind phenyl-Sepharose (see "Experimental Procedures"). Similar binding to type II and type III collagen-Sepharose was also observed (Fig.  3A). Data for the cartilage major form are shown, which are representative of all three species. Recognition of the triple helix is required for the interactions with collagen, since binding to gelatin (denatured collagen)-Sepharose was weak and similar to that seen to ligand-free Sepharose (Fig. 3). Lack of binding to gelatin has been previously demonstrated for the porcine orthologue (1,67). Electrostatic affinities appear to be involved, since binding was disrupted by 1 M NaCl. These data therefore demonstrate specific interactions between YKL-40 and the major fibril-forming collagens, thus identifying potential new ligands for this protein.
To further demonstrate the specificity of the interaction with type I collagen, the purity of the sample used as the coupling ligand is shown (Fig. 4A). The only proteins visible as Coomassie Blue-stained bands are the type I collagen ␤ and ␣ chains and some higher molecular weight multimers, the identities of which are confirmed by digestion with both bacterial collagenase and the MMP collagenase, MMP-1. Furthermore, the binding of the chondrocyte-derived YKL-40 species to type I collagen-Sepharose could be partially inhibited by an excess of soluble type I collagen, but not type I gelatin (Fig. 4B, competition). Competition for binding in the presence of soluble collagen caused some of the YKL-40 to appear in the flow through and first wash, whereas all the YKL-40 was found in the elution in the presence of type I gelatin or buffer only. A 6.25-fold increase in the quantity of competing type I collagen still resulted in partial inhibition of binding only (data not shown). The proportion of the binding which cannot be competed by soluble ligand is nonspecific and the binding of YKL-40 to type I collagen-Sepharose therefore results from both a specific interaction with the ligand and some additional, nonspecific binding to the resin. The existence of a nonspecific interaction is also demonstrated by the ability of YKL-40 to partially bind ligand-free Sepharose (Figs. 3 and 4B). As a final test for specificity, type I collagen-Sepharose was digested with bacterial collagenase prior to YKL-40 application (Fig. 4B). This resulted in the recovery of some YKL-40 in the flow-through and washes, in contrast to recovery in the elution only without digestion. The residual binding after digestion of the collagen ligand is nonspecific. These data therefore again demonstrate that FIGURE 3. YKL-40 binds to fibril-forming collagens. Purified YKL-40 major and minor forms from resorbing cartilage (A and B, respectively) and the species derived from chondrocyte cultures (C) were each applied to type I collagen-Sepharose minicolumns, followed by SDS-PAGE on 12.5% gels and silver staining of the fractions as described under "Experimental Procedures." The cartilage major form (A) was also applied to type II and type III collagen-Sepharose. Control experiments using gelatin-Sepharose and cyanogen bromide-activated Sepharose without a ligand attached (none) were also performed. SM, starting material for columns, FT, flowthrough, M r , molecular mass markers in kDa.
YKL-40 binds specifically to type I collagen and also nonspecifically to the Sepharose resin.
The binding of chondrocyte-derived YKL-40 to type I collagen was examined further by surface plasmon resonance analysis, with type I collagen as the coupled ligand. Fig. 5 confirms a specific interaction between these two proteins: an overlay plot consisting of eight binding curves derived from five YKL-40 concentrations is shown, with data from the control (uncoupled) flow cell subtracted. Kinetic parameters (k d , k a , and K D ) were also calculated by fitting the data to the Langmuir model for 1:1 binding. In two separate experiments, each with eight binding curves, k d values of 4.50 ϫ 10 Ϫ3 s Ϫ1 (Ϯ6.67 ϫ 10 Ϫ4 s Ϫ1 ) and 3.42 ϫ 10 Ϫ3 s Ϫ1 (Ϯ2.38 ϫ 10 Ϫ4 s Ϫ1 ) were obtained ( Table   1). The dissociation phase followed exponential decay kinetics, with 2 values of 0.33 and 2.29 from global fits of all eight curves. In contrast, inconsistent k a values were observed (2.08 ϫ 10 2 M Ϫ1 s Ϫ1 to 4.83 ϫ 10 3 M Ϫ1 s Ϫ1 ), resulting in similarly variable affinity constants (K D ϭ 9.70 ϫ 10 Ϫ7 to 2.11 ϫ 10 Ϫ5 M). This lack of consistency was observed at both the intra-and inter-experimental level and is probably related to the ability of chondrocyte-derived YKL-40 to self-aggregate under low salt conditions. This is likely to result in variable concentrations of monomers and multimers in different binding cycles, making an accurate calculation of the free monomer concentration impossible. However, a reliable k d value can still be obtained under these conditions because this parameter is independent of both analyte concentration and quantity of analyte bound.
The Chondrocyte-derived YKL-40 Species Prevents Collagenolytic Cleavage of Type I Collagen Fibrils-The possible physiological role of the YKL-40/type I collagen interaction was

TABLE 1 Calculation of dissociation rate constant (k d ) for the binding of chondrocyte-derived YKL-40 to type I collagen by surface plasmon resonance analysis
Type I collagen was coupled to a Biacore CM5 sensor chip via primary amine groups, as described under "Experimental Procedures." Chondrocyte-derived YKL-40 (1.00, 0.875, 0.75, 0.50, 0.25, and 0.10 M) was injected at 10 or 30 l min Ϫ1 for 2 min. Dissociation was then initiated by replacing the YKL-40 with binding buffer and allowed to proceed until the resonance units returned to baseline. Eight binding interactions were performed per experiment using five YKL-40 concentrations. Data from a control (uncoupled) flow cell were subtracted from the collagencoupled flow cell; the resulting binding curves were then analyzed using the Langmuir model for 1:1 binding, and the dissociation rate constant, k d , was calculated.  A, type I collagen used for coupling was digested with buffer alone (Buffer), recombinant human MMP-1 (MMP-1), or bacterial collagenase (BC) followed by application to a 6. 5% SDS-PAGE gel and Coomassie Blue staining. The positions of the undigested collagen ␤ and ␣ chains (␤ 11 , ␤ 12 , ␣ 1 , and ␣ 2 ) are shown. B, in competition assays, binding of chondrocyte-derived YKL-40 to type I collagen-Sepharose was competed with buffer alone (None), type I collagen or type I gelatin, as indicated. Binding to ligand-free-Sepharose is also shown for comparison. Type I collagen-Sepharose was additionally digested with buffer alone (None) or bacterial collagenase, as indicated, prior to YKL-40 application, to demonstrate specific binding of YKL-40 to the collagen ligand. SM, starting material for columns, FT, flow-through, M r , molecular mass markers in kDa.

Antagonistic Effects of YKL-40 on Collagen Fibrillogenesis
explored by investigating the capacity of YKL-40 to influence collagenolytic cleavage of type I collagen fibrils. The ability of the MMP collagenase, MMP-1 (interstitial/fibroblast collagenase), to cleave native, reconstituted, type I collagen fibrils (20 g ml Ϫ1 ) was measured in the absence and presence of purified, chondrocyte-derived YKL-40 (25,50, and 100 g ml Ϫ1 ) in a 90-min assay. YKL-40 caused a concentration-dependent reduction in the level of cleavage in all experiments ( Table 2). A 60 -70% reduction was observed with 100 g ml Ϫ1 YKL-40; there was some variability in the exact magnitude of effect seen with 25 or 50 g ml Ϫ1 . These data indicated that the binding of YKL-40 to collagen either hindered enzyme accessibility and/or altered collagen structure and behavior in a way which affected susceptibility to cleavage. One of several possible explanations is that YKL-40 influences aspects of collagen fibril formation, because this occurs during the assay, and soluble (non-fibrillar) collagen is considerably more susceptible to cleavage than fibrils. We therefore investigated the effect of YKL-40 on the rate of type I collagen fibril formation.

YKL-40 Modulates the Rate of Type I Collagen Fibril
Formation-In vitro fibril formation using soluble collagen extracted from tissues occurs spontaneously in neutral solutions above 20°C and follows a characteristic, near-sigmoidal plot, with a lag phase, exponential phase, and plateau phase (reviewed in Ref. 68). Under the assay conditions just described above (20 g ml Ϫ1 collagen, 37°C), attainment of the plateau phase is seen after 20 -40 min in the absence of YKL-40. Fibril formation was therefore measured under these same conditions, at time points ranging from 0 to 40 min, in the presence of chondrocyte-derived YKL-40. Increasing concentrations (25,50, and 100 g ml Ϫ1 ) stimulated the rate of fibril formation in a concentration-dependent manner (Fig. 6A), such that the gradient of the exponential phase became progressively steeper, and the plateau phase was reached at successively earlier time points. YKL-40 at 100 g ml Ϫ1 caused attainment of the plateau phase after 5-10 min, which corresponds to the start of the exponential phase for fibrils formed in its absence. In addition, the lag phase, which was ϳ5 min with buffer alone, was undetectable with 50 or 100 g ml Ϫ1 YKL-40.
These effects were compared with those of BSA (100 g ml Ϫ1 ), which was heat-denatured to form a control/irrelevant protein precipitate, because chondrocyte-derived YKL-40 forms a precipitate under the assay conditions used. BSA had a negligible effect on the rate of fibril formation (Fig. 6B), thus demonstrating the specificity of the effects seen with YKL-40. Because fibrillar collagen is considerably more resistant to cleavage than soluble substrate, this effect must contribute to the reduction of collagenolytic cleavage seen in the presence of YKL-40 and may fully account for it. YKL-40 may also coat the collagen molecules and thereby deny the enzyme access to the cleavage sites; in addition, we cannot rule out a direct interaction between YKL-40 and MMP-1, although this seems unlikely.
Because of its ability to precipitate, the chondrocyte-derived YKL-40 may stimulate the rate of fibrillogenesis by binding to the collagen monomers and increasing their local concentration. However, the major form of YKL-40 from resorbing cartilage does not form a precipitate and may therefore have a different effect on fibril formation rate. Using similar conditions to those employed for the chondrocyte-derived species, we found that the cartilage major form had an inhibitory rather than stimulatory effect on fibrillogenesis (Fig. 6C ). 50 g ml Ϫ1 of the cartilage major form delayed fibril formation, such that after 10 or 20 min, a large proportion of the YKL-40-exposed collagen was still soluble, whereas most of the material without YKL-40 had formed fibrils. Native, soluble BSA, also at 50 g ml Ϫ1 , was used as a negative control and did not inhibit fibrillogenesis; in addition, no effect was seen with 5 g ml Ϫ1 YKL-40. The consistency of the antagonistic effects of the two YKL-40 isoforms on fibril formation rate was confirmed in four further experiments. These demonstrated stimulation of fibril formation rate by the chondrocyte-derived species or inhibition by the cartilage major form (data not shown).
The cartilage-derived major form may inhibit fibril formation by binding individual triple helices and competing with their aggregation into fibrils, in a mechanism which requires the YKL-40 to be soluble. The structural difference between the chondrocyte-derived species and the cartilage major form which results in a differential ability to precipitate is not immediately apparent, since both species co-migrate during gel electrophoresis (Fig. 2B) and gel filtration chromatography. One possible explanation is that the two forms are glycosylation variants; the carbohydrate component of each was therefore investigated.
Glycosylation of the Chondrocyte-derived YKL-40 Species and the Cartilage Major Form-The bovine YKL-40 sequence (CH3L1_Bovin) contains two consensus sites for N-linked glycosylation, and a previous study (2) indicates that at least one is occupied. In line with these observations, all three YKL-40 species (chondrocyte-derived and cartilage major and minor forms) were cleaved by N-glycosidase F (Fig. 7A) to give deglycosylated proteins of reduced molecular mass (Ϫ1.5 kDa), as shown by SDS-PAGE with silver staining. Incomplete digestion of the chondrocyte-derived and cartilage major forms (data not shown) did not reveal any intermediates, suggesting that only one consensus site is occupied. Endoglycosidase H, which cleaves N-linked, high mannose, or hybrid oligosaccharides, but not N-linked, complex carbohydrate, did not cleave either

YKL-40 from cultured chondrocytes prevents collagenolytic cleavage of type I collagen fibrils
The collagenolytic activity of recombinant human MMP-1 in the absence and presence of chondrocyte-derived YKL-40 (25,50, and 100 g/ml) was measured using the diffuse fibril assay, as described under "Experimental Procedures." Activities (bold Ϯ S.D.) are shown in units ml Ϫ1 where 1 unit degrades 1 g of fibrillar collagen per minute at 37°C. The reduction of cleavage seen in the presence of YKL-40 is expressed as a percentage, in which 0% is the activity in the absence of YKL-40.

Experiment
No YKL-40 YKL-40 (25 g/ml) YKL-40 (50 g/ml) YKL-40 (100 g/ml) the chondrocyte-derived species or the cartilage major form (Fig. 7B). Under similar conditions, bovine pancreatic ribonuclease B, which contains one N-linked high mannose chain was cleaved (data not shown). These data therefore indicate that the chondrocyte-derived species and cartilage major form each contain a single, similarly-sized chain of complex, N-linked carbohydrate. The terminal residues of complex, N-linked carbohydrate chains consist of galactose ␤-(1,4)-linked to N-acetylglucosamine, in which either the galactose is the terminal sugar, or it is further linked to a terminal sialic acid residue. Sialic acid is negatively charged and the ability of the chondrocyte-derived species to precipitate in low salt could be caused by electrostatic interactions between these residues and the positively charged, surface-exposed, putative heparin-binding site (6). Conversely, the absence of sialic acid in the cartilage major form may prevent precipitation. As there is currently no published information regarding the structure of the terminal sugar residues of FIGURE 6. YKL-40 modulates the rate of type I collagen fibril formation. A, 3 H-acetylated bovine type I collagen was diluted to 20 g/ml with buffer only (छ) or purified, chondrocyte-derived YKL-40 precipitate at 25 (᭜), 50 (Ⅺ), and 100 g/ml (f). Fibril formation was performed at 37°C for 0, 5, 10, 20, 30, and 40 min followed by collection of the fibril pellet by centrifugation. The supernatants were counted as described under "Experimental Procedures" to quantify the soluble (non-fibrillar) collagen remaining at each time point. All data points are derived from triplicate analyses from which the mean average is plotted; error bars indicate the standard deviation. B, bovine type I collagen fibril formation was measured as in A with buffer only (छ) or heat-denatured BSA precipitate at 100 g/ml (F). C, 3 H-acetylated bovine type I collagen was diluted to 19 g/ml in buffer only (छ) or buffer containing either purified YKL-40 (cartilage major form, 5 and 50 g/ml, ‚ and OE, respectively) or native BSA (50 g/ml, E). Fibril formation was performed at 37°C for the indicated times and measured as in A.

Antagonistic Effects of YKL-40 on Collagen Fibrillogenesis
YKL-40, these were investigated in both isoforms by lectin blotting (Fig. 7C ). Neither species reacted with GNA, confirming the absence of high mannose carbohydrate; by contrast, a strong reaction was seen with the high mannose-containing protein carboxypeptidase Y. Both YKL-40 species contain sialic acid, terminally linked ␣- (2,6) or ␣- (2,3) to galactose, as shown by positive reactions with SNA and MAA respectively. Fetuin contains both types of sialic acid linkage and therefore reacts with both lectins, whereas transferrin contains only the ␣- (2,6) linkage and is detected only with SNA, as shown. The similar level of reaction of the two YKL-40 forms to SNA and MAA suggests that the quantity of sialic acid and its manner of linkage are identical in each case. Neither species reacted with DSA, demonstrating the absence of terminal galactose residues ␤ (1,4)-linked to N-acetylglucosamine and therefore indicating that all chain branches in both forms terminate with sialic acid. This also applies to transferrin, which therefore does not react with DSA, whereas strong reactions are seen with fetuin and desialylated fetuin, both of which contain unsialylated chains. Taken together, these data indicate that the ability of the chondrocyte form to precipitate is not caused by a differential presence of sialic acid, since lectin blotting did not reveal any differences in the terminal carbohydrate structures of the two YKL-40 species; furthermore these data represent the first analysis of these structures in YKL-40.

DISCUSSION
Screening of resorbing cartilage conditioned medium for collagen-binding proteins resulted in the identification of collagen as a new potential ligand for YKL-40. The collagen binding ability of YKL-40 was confirmed using purified protein, which was shown to bind directly and specifically to types I, II, and III collagen, in an interaction which required recognition of the collagen triple helix. Specific binding to type I collagen was also demonstrated by surface plasmon resonance analysis, and the dissociation rate constant was calculated. Because the binding kinetics were found to fit the 1:1 Langmuir isotherm, this indicates that either a 1:1 molar stoichiometric interaction takes place, or that there are multiple, identical, non-interacting binding sites for YKL-40 on type I collagen. Either of these binding modes are possible, given that type I collagen is a long, rod-shaped molecule with a molecular mass ϳ7.5 times higher than that of YKL-40 and has a repeating structure in the triple helical region consisting of the proline and hydroxyproline-rich sequence, Gly-Xaa-Yaa. If the calculated k d is combined with a reciprocal k a (i.e. ϳ3-4 ϫ 10 3 M Ϫ1 s Ϫ1 ) then a K D value of 1 M can be predicted. Reciprocal k d and k a values are quite common, but not exclusive, and the true k a of monomeric YKL-40 for type I collagen may be higher, giving a correspondingly lower K D .
Several cell types produce YKL-40 as a major protein (3,4,9,10,15,16,67) and cultured human articular chondrocytes, in particular, synthesize large amounts (4,11,19), accounting for 33% of the conditioned medium protein (19). Extended culture of these cells results in the accumulation of ϳ50 g ml Ϫ1 (ϳ1.25 M) YKL-40 in the conditioned medium (11). Hence, it seems likely that chondrocytes are capable of producing sufficient YKL-40 to reproduce the effects on fibril formation reported here, even if the binding to collagen is of relatively low affinity (e.g. K D ϭ 1 M). YKL-40 is also found in the low g ml Ϫ1 range in synovial fluid (10,35,37), for example, a median value of ϳ5 g ml Ϫ1 (ϳ0.125 M) in RA samples (37). In vivo, the fibril-forming collagens are synthesized as soluble procollagens, which are specifically cleaved by the procollagen metalloproteinases, resulting in removal of the terminal propeptides (see Refs. 68 and 69 for reviews). Propeptide cleavage is a prerequisite for fibril formation, in which removal of the C-propeptides appears to be the critical step (70), and is sufficient alone to initiate fibrillogenesis (69). Early fibril formation in vivo takes place in pericellular crypts formed between cellular protrusions (68) and may also occur prior to secretion, because of intracellular activity of the procollagen C-proteinase, bone morphogenetic protein-1 (69). The concentration of YKL-40 in these compartments may be higher than that seen in body fluids, again allowing a low affinity interaction (e.g. K D ϭ 1 M) to be effective in the modulation of fibril formation rate.
We therefore conclude that binding of YKL-40 to fibrilforming collagens and the modulation of fibril formation rate are likely to be important in the role of YKL-40 in vivo. The ability to regulate fibrillogenesis suggests a physiological role in the laying down of new collagen fibrils during development and connective tissue remodeling. This concurs with the association of YKL-40 with processes such as mammary gland involution (2) and fetal cartilage development (48). The putative YKL-40 ligands are thought to be carbohydrate structures, since YKL-40 is capable of binding chitin, chito-oligosaccharides, and heparin. Our data do not preclude recognition of carbohydrate ligands by YKL-40, and simultaneous binding to both collagen and a glycan structure might in fact be possible, via the hydrophobic substrate cleft or putative heparin binding site (6, 7) and a separate collagen binding sequence. The previously described cell signaling effects of YKL-40 (52-56) might therefore involve the simultaneous stimulation of cell surface HSPGs and integrins, since an interaction with collagen would provide a molecular bridge between YKL-40 and the integrin.
Key questions regarding the interaction of YKL-40 with collagen are whether the YKL-40 remains associated with the collagen once fibril formation is complete and also whether YKL-40 recognizes fibrillar collagen (as opposed to soluble triple helices) as a ligand. Preliminary data show that when fibril formation proceeds to completion in the presence of YKL-40 (cartilage major form) and the fibril pellet is collected by centrifugation, then most or all of the YKL-40 is found in the supernatant, i.e. not associated with the fibril pellet. 3 Other studies show that when porcine vascular smooth muscle cells are seeded onto collagen gels, the YKL-40 produced cannot be detected in the gel (12). In addition, YKL-40 secreted by arthritic cartilage in situ does not appear to localize with the type II collagen-containing cartilage matrix (36,37), even when the cartilage is treated with trypsin or hyaluronidase prior to staining. Furthermore, in cultured chondrocytes and cartilage explant cultures, YKL-40 is detected only in the cells and not in the pericellular matrix (11). Although it is possible that its presence is actually masked by the type II collagen itself, this seems unlikely, since metabolic labeling studies with cartilage explants also indicate a lack of strong association with matrix components (4). The calculated k d (3.42 ϫ 10 Ϫ3 to 4.50 ϫ 10 Ϫ3 s Ϫ1 ) and a possible K D in the low micromolar range also make a permanent association with collagen unlikely. Taken together, these findings indicate that YKL-40 does not remain permanently associated with collagen once it has formed fibrils; neither does it recognize fibrillar collagen (as opposed to soluble triple helices) as a ligand. A transient association with newly formed soluble collagen, the function of which is to regulate the laying down of new fibrils seems, instead, to be a much more likely scenario. Given this potential role in early fibrillogenesis, YKL-40 may also be involved in aspects of procollagen processing by influencing the functions of the procollagen C proteinase enhancer proteins, PCOLCE1 and PCOLCE2, because these are known to bind to the triple helical portions of fibrillar collagens (71).
The findings of this report indicate that cartilage tissue is capable of producing multiple YKL-40 isoforms, two of which have antagonistic effects on the rate of fibril formation. The exact structural basis for these opposing effects and how they relate to a differential ability to precipitate require further study. Lectin blotting did not reveal any difference in the structure of the terminal carbohydrate residues of the cartilage major form and chondrocyte-derived species, indicating that this is not the source of the variation. The explanation that we favor is that the two forms are conformational variants, in which one conformation allows the juxtaposition of oppositely charged groups on adjacent molecules, such as, for example, the negatively charged sialic acid residues and the positively charged, putative, heparin binding site (6). The published crystal structure of human YKL-40 (7) showed that binding of oligosaccharide ligand to the carbohydrate binding cleft induced a large conformational change in the molecule, although this finding was not confirmed by another structural report (6). The possibility, however, remains that the chondrocyte-derived species and the cartilage major form are conformational variants, produced by differential interactions with glycan structures in situ and that this accounts for the observed differences in behavior. Because the cartilage minor form elutes prior to the other two species on gel filtration, it may be a non-covalent, multimeric complex of the major species or a conformational variant of the latter, in which the variation alters the elution characteristics. Further studies will differentiate between these explanations. The properties of all three YKL-40 species are summarized in Table 3.
Because YKL-40 is produced by arthritic human cartilage in situ, particularly when degenerative changes are apparent, but not by healthy tissue (4,36,37,48), it seems likely that expression either causes or results from pathology. The stimulus for expression seems to be a change in the ECM environment of the chondrocytes, because YKL-40 is induced in healthy cartilage in response to the trauma caused by tissue resection (4,11,36). Stress to cartilage is known to increase the onset of arthritis, especially in the case of OA, which has been linked with sports injuries and the alteration of mechanical forces across the joint. An initial insult to the cartilage, such as an injury, and the resulting stress to the chondrocytes may induce the expression of a YKL-40 isoform with an inhibitory effect on collagen fibril formation rate, i.e. analogous to the cartilage major form identified in this report. Newly synthesized collagen produced during attempted repair will then be more susceptible to collagenolytic cleavage and the incorporation of new fibrils into the damaged tissue architecture will therefore be compromised. The inability to adequately repair the ECM will maintain the chondrocytes in an abnormal, stressful environment, resulting in continued YKL-40 expression and the emergence of a self-perpetuating, downward spiral, which, along with other factors, will eventually cause gross, irreversible damage to the cartilage collagen network.
Alternatively, YKL-40 may help the chondrocytes to respond to changes in their environment and may prevent or repair damage caused by injury or disease. The ability of YKL-40 to dampen the responses of chondrocytes and synovial cells to inflammatory cytokines, decreasing the production of matrix metalloproteinases and chemokines (56) suggests a protective role in inflammatory environments, limiting degradation of the ECM and thus controlling tissue damage. It has also been shown that inflammatory cytokines can induce YKL-40 production from articular chondrocytes and YKL-40 therefore functions in a negative feedback loop to limit the destructive effects of these signaling molecules (72,73). Excessive production of YKL-40, particularly of isoforms, which stimulate collagen fibril formation rate, may therefore cause matrix accumulation, leading to tissue fibrosis. Future work will address these important issues.  Yes  Collagen binding  Types I, II, and III  Types I, II, and III  Types I, II, and III Effect on type I fibril formation rate Inhibitory Unknown Stimulatory a Established from purified samples run on 12.5% SDS-PAGE gels under reducing conditions. b The elution volume of the peak fraction obtained from AcA44 gel filtration is shown as the mean average from 3 runs Ϯ S.D. For the cartilage minor form, 2 runs only were performed, and the mean average and range (in parentheses) are shown.