Physiological and Pathological Secretion of Cartilage Oligomeric Matrix Protein by Cells in Culture*

Abnormalities in cartilage oligomeric matrix protein (COMP), a pentameric structural protein of the cartilage extracellular matrix, have been identified in pseudoachondroplasia and multiple epiphyseal dysplasia, two human autosomal dominant osteochondrodysplasias. However, the function of the protein remains unknown. With the goal of establishing a model to study the mechanisms by which COMP mutations cause disease, we have analyzed synthesis and secretion of COMP in cultured chondrocytes, tendon, and ligament cells. Pentameric protein detected inside of control cells suggested that pentamerization is an intracellular process. Patient cells expressed mutant and normal RNA and secreted COMP at levels similar to controls, suggesting that abnormal pentamers are likely to be found in the extracellular matrix. Inclusions within patient cartilage stained with anti-COMP antibodies, and cultured cells presented similar inclusions, indicating that presumably abnormal COMP pentamers are less efficiently secreted than normal molecules. We conclude that the COMP disorders are likely to result from a combination of a decreased amount of COMP in the matrix and a dominant negative effect due to the presence of abnormal pentamers in cartilage.

Cartilage oligomeric matrix protein (COMP) 1 is a 524-kDa homopentameric extracellular matrix glycoprotein, which belongs to the thrombospondin family of proteins (1)(2)(3)(4). Each monomer is composed of an amino-terminal cysteine-rich domain, four EGF-like domains, eight calmodulin-like repeats, and a C-terminal globular domain. The cysteine-rich domain is responsible for assembly of the monomers into pentamers by interchain disulfide bonds (5). The COOH-terminal domain may be involved in binding cells, such as chondrocytes (6), and proteins in the extracellular matrix (7,8).
COMP has recently been identified as the abnormal protein in two human autosomal dominant skeletal dysplasias, pseudoachondroplasia (PSACH) and multiple epiphyseal dysplasia (MED) (9,10). In addition to mutations in COMP, MED can also result from mutations in the gene encoding the ␣2 chain of type IX procollagen (11). The vast majority of COMP mutations identified in PSACH and MED are located in the calmodulin-like repeats (9,10,(12)(13)(14). Since these repeats are potential calcium-binding domains, the mutations are postulated to affect calcium binding and to alter the structure of this region of the protein. The level of COMP in serum or synovial fluid has also been used as a marker of cartilage degeneration in patients with osteoarthritis or rheumatoid arthritis (15,16).
Consistent with the involvement of the protein in chondrodysplasias and arthritis, cartilage is the major site of expression of COMP. In mouse limb bud cultures, COMP expression was shown to be a late marker of chondrogenesis, appearing slightly after type II procollagen and aggrecan (17). In human adult articular cartilage, COMP is expressed predominantly in the interterritorial matrix (18). In contrast, strong immunostaining was found in the territorial matrix around the chondrocytes in human fetal cartilage growth plate (18). Although enzyme-linked immunosorbent assays have not detected COMP in non-cartilaginous tissues (1), COMP has also been purified from bovine tendon (19).
The tissue pathology of cartilage in patients with PSACH is striking. Electron microscopy has shown marked dilation of the rough endoplasmic reticulum (RER) in chondrocytes (20,21). Similar but smaller RER inclusions are found in MED (22). These inclusions have been postulated to result from the impaired secretion of structurally abnormal matrix proteins. Before COMP was identified as the mutant protein in these disorders, immunohistochemistry had been performed with antibodies against various extracellular matrix proteins. These experiments showed that inclusions stained strongly with antibodies against aggrecan, weakly with antibodies against link protein, but not with anti-type II collagen antibodies (21,22). Very recently, immuno-electron microscopy of cartilage from a patient with PSACH has identified COMP and collagen IX in the inclusions (23). MED patients with the COL9A2 mutation are phenotypically similar to patients with a COMP mutation but, in two families, no inclusions were found in the chondrocyte RER (24).
In an effort to understand the normal biosynthesis of COMP, as well as the pathophysiology of PSACH and MED, we have used cultured cells as a model to study the synthesis and secretion of normal and abnormal COMP. These studies have demonstrated that COMP pentamers are assembled intracellularly, suggesting that abnormal monomers are incorporated into COMP pentamers. Immunohistochemical staining demonstrated the presence of COMP within RER inclusions of patient cartilage. Cultured cells from patients also contained RER inclusions but did not show a net decrease in COMP secretion. These data support the hypothesis that the COMP disorders result from a combination of a qualitative defect due to the secretion of abnormal COMP in the extracellular matrix and a quantitative defect resulting from intracellular retention of structurally abnormal COMP.

EXPERIMENTAL PROCEDURES
Cell Lines and Patients-Cultures were derived from control and patient tissues referred to the International Skeletal Dysplasia Registry at the Cedars-Sinai Research Institute. Primary cell cultures were established from knee ligament, Achilles' tendon, and proximal and distal femoral cartilage from a 28-week-old control human male fetus (International Skeletal Dysplasia Registry reference number R96-373). Control chondrocytes were also obtained from the iliac crest of an adult (R93-084) who underwent surgery for a problem unrelated to a COMP disorder.
Patient R91-213A has a radiographic diagnosis of mild pseudoachondroplasia (25). In this family, the disease has been linked to the COMP locus on chromosome 19 (26), and the molecular defect is a single nucleotide change leading to the G427E substitution in the sixth calmodulin repeat. 2 Patients R95-204 and R85-160A have diagnoses of unclassified MED and severe PSACH, respectively. The mutations in both patients are duplication of one or two codons in the stretch of five aspartic acid codons (469 -473), respectively, in the seventh calmodulin domain of COMP and will be described in detail elsewhere. 3 In all three cases, chondrocytes or tendon cells were grown from the proximal femur, which was obtained at hip replacement surgery. Patient R68-29 has a diagnosis of PSACH and a point mutation leading to a single amino acid change (D446N) in the sixth calmodulin domain, identical to that of the patient described by Maddox et al. (23). Two unrelated patients, R94-291A and R92-58, have typical PSACH and a missense mutation (G309R) in the second calmodulin domain. 2 Steady-state Cell Labeling-Cells were plated at near-confluence in a 25-cm 2 flask and incubated overnight. They were then starved for 4 h in serum-free Dulbecco's modified Eagle's medium high-glucose medium before labeling with 40 Ci of [ 35 S]cysteine (NEN Life Science Products) in 400 l of fresh serum-free medium. After 48 h, the medium was collected. After rinsing with ice-cold phosphate-buffered saline, the cell layer was harvested in 400 l of prechilled lysis buffer (10 mM Tris-HCl, pH 7.4, 0.15 M NaCl, 0.1 g/liter Triton X-100, 1 mM phenylmethylsulfonyl fluoride).
Purified human transforming growth factor-␤1 (TGF-␤, R & D Systems, Minneapolis, MN) was, when indicated, added at the same time as [ 35 S]cysteine, at a final concentration of 4 ng/ml, which was determined in preliminary experiments to yield maximum induction of COMP (not shown). Ascorbic acid was used at 50 g/ml.
Identification of COMP; Purification, Western Blotting, and Immunoprecipitation-To identify COMP in our cell samples, we partially purified COMP protein from human fetal cartilage according to published procedures (1, 6, 18) by salt-EDTA extraction and DEAE-cellulose chromatography. Cartilage and cell extracts, and cell supernatants, were separated by electrophoresis through 7% SDS-polyacrylamide gels, each with a 5% polyacrylamide stacking gel. Kaleidoscope molecular weight markers were from Bio-Rad. COMP was further identified by Western blotting and immunoprecipitation, using polyclonal antibodies against purified bovine COMP, a gift from Dr. D. Heinegård (1). For Western blotting, the proteins were transferred from the gel to a Hybond-C (Amersham Pharmacia Biotech) membrane by overnight electric transfer. COMP was detected by chemiluminescence with the Amersham Pharmacia Biotech ECL kit, following the manufacturer's instructions. For immunoprecipitation, 1 l of anti-COMP antibody was typically used for 100 -200 l of sample. After a 1.5-h incubation on ice, protein A-Sepharose (Amersham Pharmacia Biotech; 30 l of a 50:50 slurry in 10 mM Tris, pH 7.4, 0.15 M NaCl, 0.1% bovine serum albumin, and 0.1% Triton X-100) was added, and the sample was gently mixed in the cold room for 1 h. The beads were washed three times with the same buffer before addition of sample buffer for electrophoresis.
Immunohistochemistry and Electron Microscopy-Electron microscopy was performed according to a modification of described techniques that will be published in detail elsewhere (27). Briefly, fresh chondroosseous tissue or cultured cells were fixed in glutaraldehyde, post-fixed with 3% osmium tetraoxide, and dehydrated with an acetone gradient. After dehydration, the tissue was infiltrated with acetone/Spurr resin with several changes of resin. The resin was polymerized and sectioned (500 -600 Å). Sections were stained with uranyl acetate and lead citrate and examined with a Zeiss 902 transmission electron microscope.
For immunohistochemistry, frozen sections (6 m) of cartilage were fixed with 4% paraformaldehyde. Endogenous peroxidase activity was blocked using Endo/Blocker (Biomeda, Foster City, CA). Proteoglycans were partially digested with Auto/Zyme solution (Biomeda). The sec-tions were blocked with 0.1% bovine serum albumin, 1% non-fat dry milk, and 2% serum, then incubated with the primary antibody. After incubation with the biotinylated secondary antibody (at a concentration of 7.5 mg/liter) and ABC solution (Vector Laboratories, Burlingame, CA), sections were developed with a peroxidase substrate 3,3Ј-diaminobenzidene solution (Biomeda) and mounted. Anti-type IX collagen (Development Studies Hybridoma Bank, Iowa City, IA), anti-type II collagen (a gift from Dr. Robin Poole, Montreal, Canada), and anti-aggrecan (a gift from Dr. D. Heinegård, Lund, Sweden) polyclonal antibodies were used at the dilutions of 1:1000, 1:1500, and 1:1000, respectively.
RNA Purification, Reverse Transcription-Polymerase Chain Reaction-RNA from cultured cells was purified with Trizol (Life Technologies, Inc.). Reverse transcription was performed with a COMP-specific primer (5Ј-TTTGTCCTCTCTGAGCCCTTC-3Ј) complementary to a portion of the 3Ј-untranslated region of the message, using the Superscript Preamplification System (Life Technologies, Inc.). The cDNA was amplified using gene-specific primers surrounding the mutation, and sequence was determined by cycle sequencing with ThermoSequenase (Amersham Pharmacia Biotech).

COMP Synthesis by Cultured Control
Cells-COMP biosynthesis was examined in cultured cells derived from cartilage and tendon, tissues known to contain COMP. Since joint laxity is a major aspect of the phenotype of PSACH, we also examined cultured ligament cells. All three cell types were grown in monolayer and subjected to steady-state labeling with [ 35 S]cysteine. The supernatants and cell layers were collected separately and analyzed by polyacrylamide gel electrophoresis, Western blotting, and immunoprecipitation. In the supernatants of the cultured ligament cells, we identified a protein that co-migrated with partially purified COMP from fetal cartilage (Fig. 1, lane 1) and that could be immunoprecipitated with anti-COMP antibodies (Fig. 1, lane 2). Reduction of the immunoprecipitated material yielded a protein that co-migrated with monomeric COMP (Fig. 1, lane 5). In the absence of antibody (Fig. 1, lane 6), this protein was not observed. Comparable results were also obtained with cultured tendon cells and chondrocytes (data not shown). In the immunoprecipitation experiments, a second protein, with a molecular mass of 130 -140 kDa after reduction, was also apparent (Fig. 1, lane 5). It is possible that this molecule is monomeric thrombospondin-4, the closest member to COMP in the thrombospondin family of proteins, but we did not test this hypothesis. This protein was not detected in purified cartilage extracts.
We tested the effect of transforming growth factor-␤1, a growth factor known to affect bone and cartilage growth and/or differentiation and that stimulates COMP synthesis by cultured synovial cells (28) on COMP biosynthesis by the cell lines. Treatment with soluble purified TGF-␤1 strongly stimulated COMP secretion in cultured ligament (Fig. 1, lanes 7 and  8), tendon (Fig. 1, lanes 10 and 11), and chondrocytes (Fig. 1,  lanes 15 and 16). The general metabolism of the cells was enhanced by this treatment, but a few proteins, including COMP, were remarkably more stimulated than others (e.g. Fig.  1, lanes 15 and 16). Ascorbic acid, which is known to enhance collagen secretion (29), did not affect COMP synthesis (data not shown).
For the cultured ligament cells, only the pentameric form of COMP could be detected in the supernatant and the cell layer samples, suggesting that the pentamer is assembled intracellularly (Fig. 2, a and b). Identical results were also obtained with cultured chondrocytes (Fig. 2c) and tendon cells (data not shown).
Synthesis and Secretion of COMP by PSACH and MED Patient Cells-We examined COMP biosynthesis by cultured cells from PSACH and MED patients. The three patients studied had distinct mutations in one of the calmodulin-like domains of COMP. In addition, for all three patients, direct sequence analysis of reverse transcription-polymerase chain reaction products containing the mutations demonstrated approximately equal levels of expression of the normal and mutant alleles (e.g. Fig. 3). This apparent stability of the transcript derived from the mutant allele suggests that normal and abnormal monomers are approximately equally translated in the patient cells and are available for incorporation into pentamers. At equal levels of expression of monomers derived from the normal and abnormal alleles, by random association, only 1 out of 32 pentameric molecules should consist of five normal monomers, and more than 97% of the pentamers should contain at least one abnormal monomer. Therefore, under the hypothesis that structurally abnormal pentamers are retained in the RER, only about 3% of the normal amount of COMP would be expected to be secreted into the supernatant by patient cells.
To test this hypothesis, primary cultures of chondrocytes (mild PSACH patient R91-213A, typical PSACH patient R85-160, and MED patient R95-204) or tendon cells (R95-204) were labeled with [ 35 S]cysteine, in parallel with control adult chondrocyte or fetal tendon cell cultures. The COMP protein synthesized by the patient cells was indistinguishable, both quantitatively and qualitatively, from that synthesized by control cells. After a 48-h labeling, there was neither a dramatic decrease in secretion (Fig. 1, lanes 12, 13, and 14) nor an increased amount of COMP in the cell layer (Fig. 2c) of patient cells as compared with controls. As in the control cell cultures, COMP was found as a pentamer both within the cells and in the secreted compartment, and no free monomer was detected (Fig. 2c).
Electron Microscopy of Tissue and Cultured Cells-Cartilage from all three patients showed inclusions in the RER of chondrocytes. The inclusions were freckled in the case of the patient with mild PSACH (R91-213) (25), lamellar or freckled in the patient with typical PSACH (R85-160), and granular in the MED patient (R95-204) (Fig. 4, A-D).
To determine whether the cultured cell model replicated the phenotype observed in vivo, we performed electron microscopy on cultured chondrocytes. After 4 days in culture in monolayer, neither chondrocytes nor tendon cells exhibited inclusions (not shown). However, longer term chondrocyte cultures of 4 -5 weeks at confluence, without passage, showed dilated RER in cells from all three patients (Fig. 4, E-H). The inclusions were all granular in aspect. Control adult chondrocytes cultured and analyzed in parallel did not show RER inclusions (not shown).
RER Inclusions Contain COMP-In control fetal cartilage, polyclonal antibodies against COMP, used at a 1/100 dilution, revealed a pericellular expression of COMP, as has been described (Refs. 1, 18, and 30; data not shown). Under these conditions, staining of the articular cartilage from a patient with PSACH (R68-29) showed intense intracellular staining of the cells, suggesting that COMP protein is contained in the RER inclusions (not shown). However, at this high antiserum concentration, we and others (1, 18) observed background intracellular staining in control cells. We therefore repeated the experiment at a lower concentration of antiserum (1/1000). Under these conditions too, strong intracellular staining was detected inside patient chondrocytes, but not controls, indicating that the RER inclusions contain COMP (Fig. 5). A similar result was obtained with cartilage from two other PSACH patients (R68-29 and R92-58) (not shown). Immunostaining with antibodies against other extracellular matrix proteins showed strong intracellular staining of patient cartilage with anti-type IX collagen antibodies, but anti-aggrecan or anti-type II collagen antibodies did not stain above control levels (Fig. 5). DISCUSSION COMP is a pentameric glycoprotein of the cartilage extracellular matrix responsible, when abnormal, for two autosomal dominant human skeletal dysplasias, PSACH and MED. COMP expression is limited to few, poorly accessible, tissues. To get insight in the function of COMP and to understand the cellular pathology of PSACH and MED, we have established an in vitro cell culture model to study the metabolism of the normal and mutant proteins.
Pentameric COMP Is Secreted by Control Cultured Cells-Chondrocytes isolated from adult or fetal cartilage, grown in monolayer, synthesized and secreted pentameric COMP. Since chondrocytes are known to dedifferentiate quickly when grown in monolayer, results obtained in such cell cultures need to be cautiously interpreted. We therefore examined two other cell types, tendon and ligament cells, which are fibroblast-type cells. Both cell types expressed COMP, confirming, in humans, tendon as a source of this protein (19,31). Expression of COMP by ligament cells may be relevant to the joint laxity observed in PSACH patients. Synthesis of COMP was dramatically increased by treatment of the cells with a growth factor known to affect bone and cartilage growth, TGF-␤1 (28).
In all three cell types, the pentameric form of COMP was detected in the supernatant as well as inside the cells. The much greater affinity of the available antibody for the pentamer than for the monomer doesn't allow us to rule out the presence of low levels of free monomer. However, the intracellular presence of pentamer suggests that pentamerization of COMP is an intracellular process. This result is similar to what has been observed for another protein of the thrombospondin family, thrombospondin-1, which co-translationally assembles into trimers within the lumen of the RER (32). In patients with PSACH and MED, all COMP mutations are found in the calmodulin-like or the C-terminal domain. One can therefore predict that, if patients express abnormal monomers, these may be incorporated into pentamers before the mutant region is translated and before the molecule could be detected as abnormal by the cell. This further implies that a secretion defect in patients would operate at the level of the pentamer.
COMP Is Secreted by Patient Cells-We subsequently tested patient cells of similar tissue origin with the goal of studying how the mutations affect the metabolism and function of COMP. Since MED and PSACH are inherited in an autosomal dominant mode, the pathologies could be caused by either haploinsufficiency or a dominant negative effect. Because both the normal and mutant alleles were expressed at similar levels in the patient cell cultures, we expect that the mutant RNA is stably expressed and likely to be translated. This suggests that the cells are not functionally haploinsufficient for COMP and that the mutations are pathologic, at least in part, through a dominant negative effect, due to incorporation of one or more abnormal monomers into COMP pentamers. In addition, we observed no dramatic quantitative secretion defect of COMP in cultured cells from patients with COMP mutations. This suggests that the cells in culture are able to secrete abnormal pentamers, supporting the hypothesis that COMP mutations in PSACH and MED exert their phenotypic effect through a dominant negative mechanism. If this also proves true in vivo, it implies that abnormal COMP is present in the matrix, but is not able to perform its normal function.
The biochemical data supporting a dominant negative model are also compatible with the molecular genetic data. A priori, haploinsufficiency is an unlikely mechanism, since it would be difficult to understand how different point mutations in the COMP gene could cause two distinct phenotypes (9). Also, truly genetically haploinsufficient patients have never been described. In a panel of PSACH patients in which all mutations have been identified, none had either nonsense mutations predicted to lead to a truncated protein or genomic deletion of COMP. 4 It is entirely plausible that the different mutations could affect different weeks in culture. These data suggest that, in addition to the dominant negative effects of COMP mutations, there is also impaired secretion of the abnormal pentamers and therefore a reduction of the amount of COMP in the extracellular matrix. Staining of the retained material with antibodies against COMP and type IX collagen has been demonstrated in one recently reported PSACH patient (23) and in three PSACH patients with COMP mutations described here. Thus intracellular retention of both COMP and type IX collagen is a general characteristic of PSACH, and reduced amounts of these two proteins in the matrix may contribute to the disease. Absence of inclusion staining with anti-aggrecan and anti-type II procollagen antibodies in our patients suggests that retention is specific and that COMP retention does not lead to a general deficit in protein secretion by chondrocytes.
Although our data, both in vitro and in vivo, identify RER inclusions in chondrocytes, there also appears to be some anti-COMP antibody staining of the cartilage matrix of patients. This suggests that the abnormal COMP allele contributes to the matrix in vivo. While this finding may be in contrast to data suggesting absence of COMP in the cartilage matrix of one PSACH patient (23), differences in antibodies and techniques may provide an explanation for the discrepancy. Another interesting finding of the paper by Maddox et al. (23) is the fact that, in contrast to chondrocytes, patient tendon cells seem to be able to secrete COMP into the extracellular matrix. It will be critical, when more tendon tissue from patients becomes available, to determine by electron microscopy whether tendon shows RER inclusions similar to those in cartilage.
In conclusion, our current model for the pathology of the COMP disorders is the combined effects of abnormal COMP in the matrix and poor secretion of abnormal pentamers by chondrocytes and other cell types that express COMP. At least in vitro, and possibly in vivo, inclusions in the RER appear to develop slowly over time. We hypothesize that retention of structurally abnormal COMP leads to secondary retention of other matrix proteins, including type IX collagen, perhaps because they normally interact with COMP. This possibility was also suggested by the observation that MED can also result from mutations in COL9A2 (11,33). Determining the relative importance of the qualitative and quantitative defects in COMP biosynthesis on disease pathology will require a better understanding of the role of COMP in the matrix and how its specific functions are disrupted by mutations. The availability of cultured cells as an in vitro model can now be used to explore the various functions of COMP, including calcium binding, interaction with chondrocytes, and binding to other extracellular matrix components, as well as the effect of mutations on these properties.