The crystal structure of the signature domain of cartilage oligomeric matrix protein: implications for collagen, glycosaminoglycan and integrin binding

Expression of recombinant protein

A recombinant COMP protein, designated E4T3C5, which includes sequences from the last EGF-like repeat, the type 3 repeats, and the CTD (A225 to A757), was prepared by PCR using full-length cDNA for human COMP as a template with the forward primer 697hcompf (TGGAATTCTGCACAGCGCTTCTGCCCC) and the reverse primer 698hcompr (GATACCGGTGGCTTGCCGCAGCTGATG) (Supplemental Fig. 1). The fidelity of the PCR product was verified by nucleotide sequencing. The E4T3C5 PCR product was cloned into the EcoRI and AgeI sites of the vector pMT/BiP/V5-HisA (Invitrogen, Carlsbad, CA, USA), resulting in inclusion of only the polyhistidine sequence at the C terminus. The expression vector was cotransfected into Drosophila S2 cells with the selection vector pCoHYGRO according to the manufacturer’s protocols (Invitrogen). Transfected cells were selected with hygromycin B, and expression of the recombinant protein was monitored by SDS-PAGE and Western blot analysis. For large-scale preparation of recombinant protein, S2 cells were grown in serum-free medium for 5 d. The culture supernatant was centrifuged to remove the cells and dialyzed against 20 mM HEPES (pH 7.8), 2 mM CaCl₂, and 500 mM NaCl. The dialysate was applied to a column of ProBond resin (Invitrogen). The column was eluted with 20 mM HEPES (pH 6.0), 2 mM CaCl₂, 500 mM NaCl, and 500 mM imidazole. The protein eluted with 500 mM imidazole was dialyzed against 20 mM HEPES (pH 7.0), 2 mM CaCl₂, and 500 mM NaCl, and 1% sucrose was added prior to storage.

Protein crystallization and diffraction data collection

The E4T3C5 protein was crystallized using the sitting drop vapor diffusion technique. Crystals appeared from the buffer, including 50 mM ammonium sulfate, 0.05M Bis-Tris (pH 6.5), and 30% (v/v) pentaerythreitol ethoxylate (15/4 EO/OH). The protein crystallization buffer includes 250 mM NaCl, 0.27 mM CaCl₂, and 20 mM HEPES (pH 8.0). Before data collection, the E4T3C5 crystals were treated with a cryoprotectant composed of the crystallization buffer with the addition of either 25% (v/v) glycerol or 30% (v/v) PEG400.

X-ray diffraction data were collected at 100 K at the 19ID beamline of the Structural Biology Center at the Advanced Photon Source at Argonne National Laboratory using the program SBCcollect. The diffraction patterns from E4T3C5 crystals were all sharp, with mosaicities <0.5°. However, the crystals systematically diffracted to the resolution limits of 3.2–3.5 Å, with the best data set processed to 3.15 Å with HKL3000 suite (Table 1) (21).

Structure determination and refinement

The structure of E4T3C5 was solved using the molecular replacement method. The structures of the CTD with a portion of the type 3 repeats of human TSP-1 (N841-P1152, PDB code: 1UX6) and the CTD with all of the type 3 repeats of human TSP-2 (E693-N1171, PDB code: 1YO8) were used separately as structural templates for searching COMP molecules in the crystal structure. They both resolved three E4T3C5 monomers in one asymmetric unit. The estimated solvent content of the E4T3C5 is as high as 71.2%, which is consistent with the low-resolution limits of their X-ray diffraction data sets. Initial refinements and difference Fourier maps indicated that all three CTDs were well positioned, while most of type 3 repeats were out of densities and needed to be rebuilt. After cycles of refinement using the program REFMAC (22) and manual model building using the program COOT (23), the structure of three E4T3C5 monomers was completed from their uniquely positioned EGF4 domain to their unique C-terminal helices (E747-A757). Although the resolution limit of the data set is as low as 3.15 Å, the electron density maps are of high quality and are very instructive for building N-linked glycans to N472 up to the branched mannose moieties. Several sulfate groups associated with clustered arginine residues were also added into the structural model. The final structural model includes all amino residues from A225 to A757 in the constructs for three monomers plus two N-terminal residues from the expression vector. The electron densities for the three E4T3C5 monomers are continuous at the 1σ contour level without breaks.

Overall structure of COMP and its comparison to TSP-1 and TSP-2

There are three E4T3C5 monomers in each asymmetric unit. They form a pseudo-3-fold trimeric assembly with their EGF4 domains contacting each other (Fig. 1). The monomer-monomer interface areas of the assembly are small (<600 Ų), and interactions across their interfaces are largely nonspecific. Thus, the trimer is not likely to be of biological significance, and it is probably a result of molecular packing of E4T3C5 in the crystal. This is consistent with the data from size-exclusion chromatography that indicate that E4T3C5 runs as a monomer in solution (Supplemental Fig. 2).

There are no significant conformational differences between the three E4T3C5 monomers that comprise an asymmetric unit. Pairwise structural alignments (Cα only) performed using the least-square-fit method resulted in root-mean-squared deviation (RMSD) values from 0.92 to 0.96 Å. There is also no major local conformational variation between the monomers.

Whereas the E4T3C5 region of COMP shows overall structural similarity to the equivalent regions of TSP-1 and TSP-2, only their CTDs, which are members of the LNS (Laminin A G-domain/neurexin/sex hormone-binding globulin [SHBG]) domain family, can be nearly fully aligned, with an RMSD value of 0.86 Å (Fig. 2A) (24). The most significant conformational difference within the CTDs of COMP and TSP-2 occurs in the β1_β2 loop. In COMP, the β1_β2 loop bends toward the β7_β8 loop, with D544 on the β1_β2 loop forming two hydrogen bonds with Y621 and W622 on the β7_β8 loop (Figs. 2A and 3C). In TSP2, the residue corresponding to D544 is T972. This residue and its neighboring residues on the β1_β2 loop do not bend toward, or make contact with, the β7_β8 loop.

Significant conformational differences between COMP and TSP-2 are also observed in the type 3 repeats. Many of the type 3 repeats, especially those in the middle (from repeat 4C to 8N), cannot be superimposed (Fig. 2A). Like TSP-2, type 3 repeat 1C and the first residue of type 3 repeat 2N have direct contacts to the CTD in COMP. The type 3 repeats do not associate with the CTD again until repeat 8N in TSP-2 or repeat 9C in COMP. The detachment of type 3 repeat 8N from the CTD is another significant conformational difference between COMP and TSP-2. In TSP-2, residue N851 of type 3 repeat 8N forms a specific side-chain–side-chain interaction to S1034 of the CTD, as well as a water molecule-mediated hydrogen bond network to Q963 and M964 of the CTD (Fig. 2B). There are also two hydrogen bonds formed between type 3 repeat 8N main-chain hydroxyl groups from D850 and N851 to Y1031(Oη) and S1034(N) of the CTD, respectively (Fig. 2B). The association of type 3 repeat 8N with the CTD is also observed in TSP-1 (20). Whereas the side-chain–side-chain interaction is lost because a serine residue (S831) in TSP-1 replaces the asparagine residue in TSP-2 (N851), other hydrogen bonds between repeat 8N and the CTD (involving Y1011 and S1014) are conserved (20). In COMP, however, although there are tyrosine and serine residues in the equivalent positions of the CTD (Y603 and S607), there are no hydrogen bonds between type 3 repeat 8N and the CTD because of the increased separation of the type 3 repeats and the CTD in this region. Moreover, the side chain of H423 of COMP (equivalent to residue of N851 in TSP-2) faces away from the CTD. The lack of attachment of type 3 repeat 8N to the CTD in COMP reduces the overall interaction between the type 3 repeats and CTD. The earlier and later type 3 repeats are still closely associated with the CTD, as they are in TSP-2.

Whereas the type 3 repeats and the CTD region are the most highly conserved regions of the TSPs, TSP-3, TSP-4, and COMP have 13- to 18-aa residues at their C termini that are not present in TSP-1 and TSP-2. A site for N-linked glycosylation is present at the beginning of this C-terminal sequence in human COMP, as well as human TSP-3 and TSP-4 (see below). In COMP, removal of these C-terminal extra residues can cause a mild form of EDM1 (25). In the E4T3C5 structure, these extra residues (from P746 to A757) form a short helix that runs across the exposed β-sheet of the CTD (Figs. 1B and 2C). This helix forms multiple interactions to other parts of the CTD, blocking the access of almost half of the convex β-sheet that lies beneath it. The residue D748 on the helix forms a hydrogen bond and a salt bridge to residues Y694 and R696 on the β12 strand of the sheet, respectively. Residues P746, Y749, and H752 on the helix contribute hydrophobic interaction to CTD, including the residue V711 on the β-bulge in the middle of the edge β13 strand. This bulge divides the β13 into two parts, β13a and β13b (Fig. 2C). The tail helix represents a unique feature of the CTD of COMP that may have its own functions and may affect the accessibility and functions of the exposed β-sheet of the CTD.

The C-terminal part of the last EGF repeat of COMP (EGF4) is similar to that of the last EGF repeat of TSP-2 (EGF3) in that these regions interact with a turn between the β12 and β13a strands of the CTD (Fig. 1B). However, EGF3 of TSP-2 also binds to the type 3 repeats, with its N-terminal region residue D673 forming interactions with R928 and K933 of type 3 repeat 12N and D943 of type 3 repeat 13C. Part of EGF2 of TSP-2 also packs closely to type 3 repeat 12N. By contrast, EGF4 of COMP has no direct contact to type 3 repeats.

Ca²⁺-binding sites and potential MIDAS

Multiple Ca²⁺ binding sites are the hallmark of the signature domain of the TSP family members. The features of these binding sites, especially those associated with the structures of the 13 type 3 repeats, have been discussed in the reports of other TSP structures (19, 20). There are eight C-type 3 repeats, with each containing two Ca²⁺ ions (Figs. 1B and 3A). One of the Ca²⁺ ions is completely coordinated by protein atoms, while the other has one coordinating water molecule that is largely buried (Fig. 3A). The 16 Ca²⁺-binding sites associated with the C-type repeats are essentially “closed” in that they have limited accessibility to the solvent or other macromolecules.

The five N-type 3 repeats also usually have two Ca²⁺-binding sites. The first one has one coordinating water molecule, similar to the second Ca²⁺-binding site in the C-type 3 repeats (Fig. 3B). It can also be regarded as a closed Ca²⁺-binding site. The second site in each N-type 3 repeat is solvent exposed. The coordinating water molecule, when it is present, is open to solvent. These “open” sites can lose Ca²⁺ ions due to unfavorable residue substitution. For example, in TSP-2, type 3 repeats 5N and 12N do not bind a second Ca²⁺ ion. In COMP, only type 3 repeat 12N loses its second Ca²⁺ ion (Fig. 1B).

There are also Ca²⁺-binding sites between some of the type 3 repeats. In TSP-2, there are two such extra Ca²⁺-binding sites, between type 3 repeats 2N and 3C, and 10N and 11C. In COMP, there is only one of these Ca²⁺-binding sites, which is between type 3 repeats 10N and 11C (Fig. 1B). The sites that are between repeats are all open. In total, both TSP-2 and COMP have 21 closed and 5 open Ca²⁺-binding sites associated with their 13 type 3 repeats. The closed Ca-bindings are more conserved than the open sites and probably exist in all TSP family members. TSP-3 and TSP-4 have an inserted sequence (PPGP) in type 3 repeat 11C (Supplemental Fig. 1). The COMP structure indicates that this sequence is probably exposed on the surface of these proteins.

In addition to the Ca²⁺-binding sites in the type 3 repeats, there are also three very well conserved Ca²⁺-binding sites in the CTD (Figs. 1B, C and 3C). For the convenience of the following discussion, we will refer to these three sites as CTD_Ca1, CTD_Ca2, and CTD_ Ca3. CTD_Ca1 and CTD_Ca2 are only ∼4.0 Å apart in the E4T3C5 structure (Fig. 3). The highly conserved aspartic acid triplet (D593DD of COMP) on the β5_β6 loop and residues N565, Q619, and S728 contribute to the formation of these two Ca²⁺-binding sites on the top of the CTD (Fig. 3C). The residues Q616 and D678 and two main-chain carboxyl groups from the β5_β6 loop form CTD_Ca3. All three Ca²⁺-binding sites in the CTD are solvent exposed or open in the absence of ligands, requiring water molecules to complete the ions’ coordination (19, 20). In TSP-1, there is an extra Ca²⁺-binding site primarily associated with the β1_β2 loop, which is not present in either COMP or TSP-2, even though the residues forming the metal binding sites in TSP-1 (D956 and D975) are conserved in both TSP-2 and COMP. The absence of this Ca²⁺-binding site is likely caused by a local conformational change of the β1_β2 loop.

One of the most remarkable features revealed from the E4T3C5 structure is that each of the three E4T3C5 monomers in the asymmetric unit (Fig. 1A) contacts its closest E4T3C5 monomer from a neighboring asymmetric unit in a dimeric conformation (Fig. 3C, right panel). For example, monomer A in one asymmetric unit forms a pseudo-2-fold “dimer” with a symmetry-related monomer C in an adjacent asymmetric unit. The primary monomer-monomer contact is formed from the interaction of E341 from the type 3 repeat 4C of one monomer with the two solvent exposed Ca²⁺-binding sites, CTD_Ca1 and CTD_Ca2, of the other monomer. E341 completes the coordinations of CTD_Ca1 and weakly coordinates to CTD_Ca2 (2.8–2.9 Å) (Fig. 3C). The pair of Ca²⁺-mediated interactions apparently dominate the dimer interface, which is rather small. Because the E4T3C5 runs as a monomer in solution, the dimer seems to result from the molecular packing of the E4T3C5 molecule and is unlikely to be of biological significance in solution. However, we can not preclude the possibility that dimers form in regions of the extracellular matrix where the protein is immobilized and at high concentrations. Metal ion coordination from a glutamic acid residue of a neighboring molecule has been observed in the crystal structures of the αM integrin and the von Willebrand factor A domain of human capillary morphogenesis protein 2 (CMG2) (26, 27). The similar ligand-binding modes have been observed in the interactions of collagen with the MIDAS of α2β1 integrin and ICAM-1 and -3 with αLβ2 (14, 28, 29). Interestingly, in the crystal structure of the CTD and part of type 3 repeats of TSP-1, the side chain of D825 from a neighboring molecule also chelates the equivalent CTD_Ca1 and CTD_Ca2, completing the metals coordination (20). The potential MIDAS of COMP, as well as TSP-1 and TSP-2 is positioned on the outer surface of the CTD, rendering it readily accessible to a wide range of ligands (Figs. 1 and 3).

RGD motif, unpaired cysteine, and N-linked glycan

The sequence motif R367GD that is a part of symmetric motif D365GRGD found in type 3 repeat 5N has been one of the focuses in the study of the potential interaction of COMP with integrins (Fig. 4A) (1). In the COMP structure, R367 is solvent exposed, however, G368 is fully buried and the side chain of D369 chelates the closed Ca²⁺ ion indirectly via a bridging water molecule in type 3 repeat 5N (Fig. 4B). D369 also forms a hydrogen bond to the main-chain amide group of T362. These interactions are also conserved for the D369 equivalents in the structures of TSP-1 and TSP-2, and they seem to be critical for the proper conformation of type 3 repeat 5N (19, 20). Therefore, the residue D369 is not available for direct ligand binding unless type 3 repeat 5N unfolds. However, because the type 3 repeats are rich in aspartic acid, including some that are solvent exposed and not involved in Ca²⁺ binding, such as D369’s neighbor D373, it is possible that analogous RGD motifs can form (see Fig. 4A and Discussion).

The CTDs of COMP, TSP-4, and TSP-1 contain an unpaired cysteine residue. The unpaired cysteine residue (C726) of COMP is in the C-terminal portion of the β14 strand of the CTD, with its free sulfhydryl group being part of the CTD β-sandwich core. As illustrated in Fig. 4C, C726 is completely buried under the β3_β4 and β14_β15 loops. Thus, the sulfhydryl group of C726 should be inert unless the CTD is denatured. Whereas the unpaired cysteine residue (C974) of TSP-1 was mutated to serine in the studies of Kvansakul et al. (20) to avoid disulfide bond exchange, the structural data indicate that C974 of TSP-1 is on the β3_β4 loop, pointing toward the core of CTD. It is remarkable that even though the unpaired cysteines are far apart in the sequences of TSP-1 and COMP, the free sulfhydryl groups have similar locations in the structures. However, C974 of TSP-1 is less buried than C726 of COMP.

Whereas N-linked glycans are added to proteins that are expressed in Drosophila S2 cells, they are not fully processed. In the E4T3C5 structure, an N-linked glycan that includes the first branched mannose moieties, is added to N742 on the C-terminal α-helix (Figs. 1B and 2C). This glycan lies between the N-terminal region of EGF4 and the CTD and may prevent these two regions from interacting with each other, as they do in TSP-2 (19). There are two N-linked glycosylation sites in TSP-2. One is on the insertion within type 3 repeat 1C, and the other is at the end of the β8-strand in the CTD (Fig. 2A). These post-translational modifications would not be expected to affect the structure of TSP-2.

Molecular surface electrostatic potential of COMP

The type 3 repeats are rich in negatively charged aspartic acid residues. A construct, including the type 3 repeats and the CTD by themselves, has a calculated pI value of 4.1 with a net charge as negative as −62 at pH 7. Therefore, the 29 Ca²⁺ ions (total charge +58) per monomer that have been identified in the E4T3C5 structure are not only critical for the proper folding of the type 3 repeats, as discussed earlier but also have significant effect on the local charge balance and overall charge distribution. Charged amino acid residues and metal ions determine a protein’s surface electrostatic property, which is important for protein-protein and protein-ligand binding. This is especially true for the more ionic interactions, such as the binding of glycosaminoglycans (GAG) to proteins. We have previously reported that a recombinant protein that contains the type 3 repeats and the CTD of COMP is retained on a heparin-Sepharose column and can be eluted with buffer containing 0.25 M NaCl (5). In this study, we have found that E4T3C5 behaves similarly (data not shown). A cluster of positively charged residues on a protein surface is an indication of a potential GAG-binding site. A survey of COMP surface electrostatic potential shows that there is one major positive surface patch associated with the type 3 repeat 1C and the inserted sequence motif (E278-C287) (Fig. 5). The clustered basic residues that contribute to the positively charged patch include K279, R281, R285, R288, and K289. Next to this patch is a shallow groove with R558 and R738 at its bottom. A SO₄²⁻ group from the crystallization buffer is bound to R558 and R738 in all three E4T3C5 monomers in the asymmetric unit of the COMP structure. Because the anionic group and clustered positively charged residues provide the anchoring site for protein-heparin interaction, the presence of the sulfate group bound to R558 and R738 may indicate a potential heparin-binding site of COMP (30). In monomer A, there are two additional SO₄²⁻ groups located between R288 and R305. Another sulfate group is found to associate with R352 and R367 only in monomer B (Fig. 5A). R367 is from the RGD sequence motif in type 3 repeat 5N. There are 14 arginine residues in the 13 type 3 repeats (Fig. 4A). Interestingly, 11 of them are in the first 6 repeats, from 1C to 6C. With the exception of R388, all of the arginines are on one side of these repeats and are highly exposed, suggesting that the region between the two sulfate groups found in monomer B may define an extended site for COMP to bind proteoglycans such as aggrecan (5) (Figs. 4A and 5A). An additional positively charged patch on the CTD that could potentially contribute to heparin binding is formed from R672, K677, K679, and K680 (Fig. 5B). Together with the first patch (K279, R281, R285, R288, and K289), it may define a second extended GAG binding site.