Abnormal bone quality in cartilage oligomeric matrix protein and matrilin 3 double-deficient mice caused by increased tissue inhibitor of metalloproteinases 3 deposition and delayed aggrecan degradation
Cartilage oligomeric matrix protein (COMP) and matrilin 3 are extracellular matrix proteins that are abundant in cartilage. As adaptor molecules, both proteins bridge and stabilize macromolecular networks consisting of fibrillar collagens and proteoglycans. Mutations in the genes coding for COMP and matrilin 3 have been linked to human chondrodysplasias, while in mice, deficiency in COMP or matrilin 3 does not cause any pronounced skeletal abnormalities. Given the similar functions of COMP and matrilin 3 in the assembly and stabilization of the extracellular matrix, our aim was to determine whether these proteins could functionally compensate for each other.
To assess this putative redundancy of COMP and matrilin 3, we generated COMP/matrilin 3 double-deficient mice and performed an in-depth analysis of their skeletal development.
At the newborn stage, the overall skeletal morphology of the double mutants was normal, but at 1 month of age, the long bones were shortened and the total body length reduced. Peripheral quantitative computed tomography revealed increased metaphyseal trabecular bone mineral density in the femora. Moreover, the degradation of aggrecan in the cartilage remnants in the metaphyseal trabecular bone was delayed, paralleled by increased deposition of tissue inhibitor of metalloproteinases 3 (TIMP-3). The structure and morphology of the growth plate were grossly normal, but in the center, focal closures were observed, a phenotype very similar to that described in matrix metalloproteinase 13 (MMP-13)–deficient mice.
We propose that a lack of COMP and matrilin 3 leads to increased deposition of TIMP-3, which causes partial inactivation of MMPs, including MMP-13, a mechanism that would explain the similarities in phenotype between COMP/matrilin 3 double-deficient and MMP-13–deficient mice.
The 2 major components of the cartilage extracellular matrix are the fibrillar collagens and the large hyaluronan-binding proteoglycan aggrecan. A variety of different proteins, including cartilage oligomeric matrix protein (COMP) and matrilin 3, modulate the assembly and structure of the extracellular matrix, generating complex functional networks. Matrilins are noncollagenous oligomeric extracellular matrix proteins with similar domain structure and function. Matrilin 1 and matrilin 3 are found almost exclusively in cartilage, while matrilin 2 and matrilin 4 have a broader tissue distribution. Subunits of matrilins 1 and 4 mainly form homotrimers, whereas subunits of matrilins 2 and 3 are found in homotetramers (1, 2). In addition, heterotrimers consisting of matrilin 1 and matrilin 3 subunits can form. Matrilins function as adaptor proteins in the extracellular matrix and connect collagen fibrils with each other and with aggrecan (2). Via their von Willebrand type A–like domain, matrilins interact with several other matrix components, including COMP (3), type IX collagen (4), type II collagen (5–7), and decorin, biglycan, and aggrecan (7).
COMP belongs to the thrombospondin family, and similar to matrilins, it is a multisubunit modular protein. COMP is predominantly expressed not only in all types of cartilage, but also in tendons (8). COMP binds in a zinc-dependent manner to triple-helical types I and II collagen and facilitates their fibrillogenesis. After the collagen fibrils have formed, COMP dissociates from the mature fibril (9, 10). It also interacts with a number of other extracellular matrix proteins, including matrilins (3), type IX collagen (11–13), and aggrecan (14).
Mutations in both matrilin 3 and COMP have been linked to multiple epiphyseal dysplasia in humans, the pathogenesis of which includes both intracellular and extracellular alterations (15–17). Despite the large number of studies on matrilins and COMP, their roles in the function, assembly, and remodeling of the cartilage extracellular matrix are not fully understood. To shed light on their in vivo function, both COMP-null (18) and matrilin 3–null (19, 20) mice were generated. COMP-deficient mice are indistinguishable from their wild-type littermates and display no obvious defects in skeletal development (18). Matrilin 3–knockout mice show only minor changes (19, 20), having wider collagen fibrils and an increased cartilage collagen volume density, but no up-regulation of other matrilin family members is observed.
COMP and matrilin 3 not only bind to each other, but they also interact with a similar repertoire of extracellular matrix components, such as types II and IX collagen (9–13). Given the similar adaptor function, it is possible that they can functionally compensate for each other. To assess the function of COMP and matrilin 3 in matrix assembly and remodeling and to elucidate their potential compensatory roles in skeletal development, we generated mice deficient in both COMP and matrilin 3 and studied their phenotypes at the morphologic and molecular levels.
MATERIALS AND METHODS
Generation of COMP/matrilin 3 double-deficient mice.
Mice deficient for COMP (18) or matrilin 3 (19) were generated as described previously. Both lines were bred for at least 6 generations onto a C57BL/6 background before COMP/matrilin 3 double-deficient mice were obtained by interbreeding of compound-heterozygous mice. Genotyping was performed by polymerase chain reaction analysis of tail genomic DNA (18, 19). The phenotype of double-deficient animals was compared with those of both of the single-deficient lines; C57BL/6 mice served as controls.
Newborn mice were euthanized, skinned, and eviscerated. After dehydration and fixation in 96% ethanol for 48 hours, cartilage was stained with Alcian blue (0.015% in 80% ethanol and 20% acetic acid) for 24 hours followed by incubation in 96% ethanol for 2 days. Samples were cleared by immersion in 1% KOH. Bones were counterstained with alizarin red for 3 hours. The stained skeletons were transferred to 2% KOH for 2 days to dissolve the soft tissue, and the cleared skeletons were then preserved in glycerol. Alizarin red–stained parts were measured to determine the bone length.
Femora from 1-month-old male mice were scanned by peripheral QCT using an XCT Research M scanner and accompanying software (version 5.50; Stratec Medizintechnik) as described previously (21). Briefly, bone parameters were determined as the mean values obtained for the 3 slices at the distal femoral metaphysis. Cortical parameters and the periosteal and endosteal circumferences were evaluated at the midshaft. The femur length was also determined by peripheral QCT.
Histologic and immunohistochemical analyses.
Hind limbs were isolated and fixed overnight in 4% paraformaldehyde (PFA) in phosphate buffered saline, pH 7.4. Limbs from 1-month-old mice were decalcified in 0.5% EDTA for 3 weeks. After decalcification, these samples were processed in exactly the same way as samples from newborn animals. After embedding in paraffin, 5-μm sections were cut and deparaffinized. For analysis of tissue morphology, hematoxylin and eosin staining was performed, and cartilage was counterstained for 20 minutes with Alcian blue as described above. ImageJ software (National Institutes of Health; online at http://rsbweb.nih.gov/ij/), was used to quantify the Alcian blue–stained area in relation to the total area in comparable regions below the growth plate on 3 sections per genotype. Von Kossa's staining was used to visualize calcified tissue, as described elsewhere (21). Hematoxylin and Safranin O were used for counterstaining.
For immunohistochemical analysis, sections were digested for 30 minutes at 37°C with 5 mg/ml of bovine testicular hyaluronidase (Sigma). Digested samples were postfixed for 10 minutes with 4% PFA, permeabilized for 10 minutes with Triton X-100, and blocked for 1 hour at room temperature with 10% fetal calf serum (FCS) and 5% normal goat serum (NGS) in Tris buffered saline (TBS). Primary antibodies in TBS containing 5% FCS were cultured overnight at 4°C. After washing with TBS, sections were incubated for 1 hour in TBS containing 1% FCS and 5% NGS with Alexa 488–labeled or Alexa 546–labeled antibodies (1:1,000 dilution; Molecular Probes) against mouse or rabbit immunoglobulins. The slides were mounted in Dako fluorescent mounting medium and evaluated using an Axiophot fluorescence microscope (Zeiss).
Cartilage extraction, sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), and immunoblotting.
Knee joints were dissected, weighed, and frozen at −80°C. On the day of extraction, the specimens were thawed and cut into 1-mm3 pieces. Ten volumes of chilled buffer I (0.15M NaCl, 50 mM Tris, pH 7.4) was added, and the tissue was extracted overnight at 4°C with continuous stirring. Extracts were clarified by centrifugation, and the supernatants were stored at –20°C. The pellets were reextracted in an identical manner with buffer II (1M NaCl, 10 mM EDTA, 50 mM Tris, pH 7.4) and the remaining insoluble material was extracted with buffer III (4M guanidine hydrochloride, 10 mM EDTA, 50 mM Tris, pH 7.4). All extraction buffers contained 2 mM phenylmethylsulfonyl fluoride and 2 mM N-ethylmaleimide. Aliquots of all extracts were precipitated with ethanol, and the pellets were processed and resuspended in SDS-PAGE sample buffer as described previously (19).
To determine the total amount of type II collagen in the tissue, the pellets obtained after extraction with buffer III were digested with pepsin as described elsewhere (22). Samples were applied to 4–12% SDS-PAGE gels, and electrophoresis was performed as described by Laemmli (23). For immunoblotting, the proteins were transferred to a nitrocellulose membrane and incubated with appropriate antibodies diluted in TBS. Bound antibodies were detected by luminescence using peroxidase-conjugated secondary antibodies (Dako), 3-aminophthalhydrazide (1.25 mM), p-coumaric acid (225 μM), and 0.01% H2O2.
The following primary antibodies were used: rabbit anti–type I collagen (1:25 dilution; Quartett item no. 2031503505), mouse monoclonal anti–type II collagen (1:500 dilution; Calbiochem item no. II-4C11), rabbit anti-NC4 domain of type IX collagen (1:2,000 dilution; see ref.4), a monoclonal mouse anti–type X collagen (clone X53; provided by Dr. Klaus von der Mark, University of Erlangen–Nuremberg, Erlangen, Germany), rabbit anti-COMP (1:1,000 dilution; see ref.24), rabbit anti–matrilin 3, rabbit anti–matrilin 1, and rabbit anti–matrilin 4 (1:1,000 dilution; see ref.25), rabbit antiaggrecan (1:1,000 dilution; Chemicon item no. AB1031), rabbit anti–G1-DIPEN (1:4,000 dilution; gift from Dr. Attila Aszódi, Max Planck Institute for Biochemistry, Martinsried, Germany, and Dr. Amanda Fosang, University of Melbourne, Melbourne, Victoria, Australia), rabbit anti–MMP-9 (1:400 dilution; Chemicon item no. AB19016), rabbit anti–MMP-13 (1:400 dilution; Chemicon item no. AB8120), and rabbit anti–tissue inhibitor of metalloproteinases 3 (anti–TIMP-3) (1:250 dilution; Biomol item no. T3866a).
Results are presented as the mean ± SD. Statistical analysis was performed using IBM SPSS software. Student's unpaired t-test was conducted to test for differences. P values less than 0.05 were considered significant.
Long bones in newborn COMP/matrilin 3 double-deficient mice.
COMP/matrilin 3 double-deficient mice obtained by breeding of compound-heterozygous mice were of the expected Mendelian ratio, with no obvious abnormalities. Adult animals were fertile and had a normal lifespan. To study the early stages of development, whole-mount skeletal staining of newborn animals was performed. Length measurement of the long bones revealed no significant differences between the double-mutant animals and the controls (Figure 1A).
To elucidate whether the lack of COMP and matrilin 3 affects the growth plate structure or the distribution of proteoglycans and/or mineral, we performed combined Safranin O and von Kossa's staining. The morphology and cellular arrangement of the double-mutant growth plate were normal, with a columnar organization of proliferative cells, an arrangement of hypertrophic cells comparable to that in controls, and with a similar distribution of proteoglycans (Figure 1B). Von Kossa's staining showed mineralized trabecular structures in the metaphysis, with no alterations as compared to controls (Figure 1B). The lack of COMP and matrilin 3 was confirmed by immunohistochemical staining.
Immunostaining for matrilin 1, matrilin 4, and types II, X, and IX collagen were performed in order to analyze the effect of COMP and matrilin 3 deficiency on the distribution of other cartilage matrix proteins. All investigated matrix components showed a similar distribution in mutant animals as in controls, and no alterations in staining intensity were observed except for type II collagen, which showed slightly decreased signals in double-deficient mice (Figure 1C).
Biochemical analysis of knee joint cartilage was performed after sequential protein extraction in order to investigate a potential up-regulation or redistribution of the other relevant cartilage matrilins. Extraction with TBS was followed by extraction with high-salt and EDTA and, finally, with guanidine hydrochloride. Immunoblot analysis confirmed the absence of COMP and/or matrilin 3 in mutant animals and showed no major differences in the amount or solubility of matrilin 1 and matrilin 4 (Figure 1D). The fact that slightly lower amounts of higher oligomers containing matrilin 4 was detected in double-deficient animals indicates that the loss of both COMP and matrilin 3 may influence the oligomerization behavior of the other matrilins.
Physical parameters of long bones in COMP/matrilin 3 double-deficient mice.
Skeletal development at later stages was studied in 1-month-old animals. At this age, a significant reduction in total body length and femur length was detected in double-deficient animals (Table 1). To address the question of whether an increased rate of apoptosis is involved in body and femur length reduction, we performed TUNEL staining. We did not detect any significant differences in the proportion of apoptotic cells in wild-type mice (mean ± SD 3.00 ± 0.80%) as compared with double-deficient mice (2.75 ± 0.96%; P = 0.32). The analysis of long bones by peripheral QCT revealed a significantly elevated trabecular bone mineral content and density at the distal femoral metaphysis (Table 1). In contrast, all measured bone parameters in the middiaphysis, including the cortical bone mineral content, cortical thickness, and cortical bone mineral density were similar. Both the periosteal and endosteal circumferences of the cortical bone at the middiaphysis were slightly, but not significantly, decreased (Table 1).
Table 1. Body length and bone parameters at the distal metaphysis and middiaphysis of the femur, as determined by peripheral quantitative computed tomography*
Control mice (n = 12)
COMP−/−/ matrilin 3−/− mice (n = 12)
Values are the mean ± SD. CSA = cross-sectional area; BMC = bone mineral content; BMD = bone mineral density.
Growth plates and endochondral bone formation in COMP/matrilin 3 double-deficient mice.
To elucidate the molecular basis for the altered mineralization in mutant mice, sections of femora from 1-month-old mice were stained for proteoglycans and collagen. The proteoglycan distribution in the growth plate, as detected by Alcian blue staining, was comparable in all genotypes. However, the staining in the lower trabecular regions of the metaphysis revealed clear differences (Figure 2A). The Alcian blue–positive area occupied 7.0 ± 2.9% (mean ± SD) of the total area in wild-type mice, 10.1 ± 1.1% in COMP-deficient mice, 9.7 ±1.1% in matrilin 3–deficient mice, and 16.4 ± 2.9% in double-deficient mice. The difference between wild-type and double-deficient animals was significant (P = 0.0079). The single-deficient animals displayed an intermediate phenotype, in that the area was slightly increased to between the values in wild-type and double-deficient animals.
Immunohistochemical staining for type IX collagen revealed identical intensity and localization in samples from all mouse genotypes (Figure 2A). Aggrecan immunoreactivity was normally distributed in the growth plate of all genotypes, but as with Alcian blue, the remaining aggrecan was detected in the more distal regions of the metaphysis of double-mutant animals, but not in controls, indicating a delay in aggrecan degradation (Figure 2A). Therefore, we monitored the MMP-mediated degradation of aggrecan by immunohistochemical staining for the N-terminal aggrecan fragment obtained after MMP cleavage, using a neoepitope antibody against the sequence DIPEN (26). In control animals, the chondro-osseous junction and the trabecular region of the metaphysis were positive for DIPEN, while the signal was reduced at the chondro-osseous junction in double-mutant animals and was almost completely missing from the trabecular region. Single-null animals again showed an intermediate intensity and distribution of the DIPEN epitope (Figure 2A).
Since we found slight differences in the staining intensity for type II collagen as early as the newborn stage, we also analyzed the expression in 1-month-old animals. With identical exposure times, it appeared as if type II collagen was reduced in COMP-deficient mice and almost absent in double-deficient mice (Figure 2B). This finding was unexpected, and considering that the complete lack of type II collagen is not compatible with life (27), we determined the total amount of type II collagen in the tissue. We found that the total amount of type II collagen was unchanged, but the extractability was different. In wild-type animals, most type II collagen could be extracted with guanidine hydrochloride, whereas for the other genotypes, pepsin digestion was necessary. In double-deficient animals, almost all type II collagen needed pepsin digestion (Figure 2B). It appears that in this case, decreased immunofluorescence staining does not necessarily mean less protein. We conclude that the collagenous matrix is organized differently; most likely, the collagen fibrils are affected, and therefore, the epitopes are less accessible to the antibody.
Cartilage and bone remodeling during endochondral bone formation in COMP/matrilin 3 double-deficient mice.
During endochondral ossification, the cartilage model for the newly forming trabecular bone is degraded by proteinases. To determine whether the delayed aggrecan degradation is due to decreased osteoclast numbers, we performed tartrate-resistant alkaline phosphatase (TRAP) staining. No apparent alterations in the number or distribution of osteoclasts were found in any of the 3 mutant genotypes as compared to the wild-type controls (Figure 3A). We quantified the percentage of TRAP-positive cells versus the total number of cells at the vascular invasion front, evaluating 3 sections and counting >2,000 cells for each genotype. We did not detect significant differences in TRAP-positive cells between sections from wild-type mice and those from double-deficient mice (14.63 ± 0.6% versus 14.49 ± 3.2%; P = 0.48).
To test for changes in the distribution of the major matrix-degrading proteinases, we performed immunohistochemical staining for several members of the MMP and ADAMTS families. The distribution of the collagenase MMP-13 and the gelatinase MMP-9 was similar in all 4 genotypes (Figure 3A). Signals for ADAMTS-4, ADAMTS-5, MMP-1, and MMP-3 were below the detection limit in the mice we examined (data not shown). However, we performed immunoblotting to quantify MMPs at the protein level (Figure 3B). The MMP-9 antibody reacts with the latent pro form of MMP-9, having a molecular weight of ∼100 kd, which was detected as a single band in all samples. Interestingly, all 3 deficient mouse lines showed reduced signals as compared to controls, with matrilin 3–deficient animals having the lowest MMP-9 levels. This result is consistent with the immunostaining results. In general, the levels of MMP-13 seem to be less affected. However, here also, the matrilin 3–deficient mice had the lowest levels of both the latent and active forms, with molecular weights of 60 kd and 48 kd, respectively.
TIMP-3 deposition in COMP/matrilin 3 double-deficient animals.
The tissue distribution of the major matrix-degrading proteinases was not affected in double-mutant mice; however, alterations in the activity of these enzymes can lead to impaired aggrecan degradation. TIMP-3 is a matrix-bound inhibitor of both ADAMTS and MMP molecules (28, 29), and changes in the amount or deposition of TIMP-3 could lead to an abnormal cartilage matrix turnover and degradation. We therefore performed immunostaining for this inhibitor.
TIMP-3 was barely detectable in the articular cartilage, in the growth plate, and in the metaphysis of control mice, with some staining seen in prehypertrophic chondrocytes. In contrast, almost no intracellular immunostaining, but intense extracellular immunostaining, was detected throughout the articular cartilage and growth plate of double-mutant animals, especially at the upper border and in the cores of the trabecular bone (Figure 4). This suggests that the delayed aggrecan degradation is the consequence of increased TIMP-3 deposition in the extracellular matrix of double-mutant mice. An intermediate degree of TIMP-3 deposition was observed in the articular cartilage of single-mutant mice, with some intracellular staining found in prehypertrophic chondrocytes and with less-intense extracellular staining in the growth plate than in double-deficient animals. Interestingly, the type II collagen staining intensities in the articular cartilage (Figure 4A) of the different mouse genotypes were similar to those observed in the growth plate (Figure 2B). Again, the weakest signal was detected on sections from double-deficient animals and was complementary to the strongest TIMP-3 signal.
Focal closure of the growth plate in COMP/matrilin 3 double-deficient animals.
The overall morphology of the double-mutant articular cartilage and growth plate was normal, but interestingly, in the central region of the epiphysis, a focal closure was observed in 90% of the COMP/matrilin 3 double-deficient animals examined (n = 10) (Figure 5A). The frequency of such closures in mice deficient in matrilin 3 was similar to that in wild-type controls (13% [n = 8] and 22% [n = 9], respectively), but was moderately elevated in the COMP-null animals (50% [n = 8]). Furthermore, a mislocalization of hypertrophic chondrocytes in the resting and proliferative zone was seen in regions surrounding the closure. At the center of the closure, a blood vessel was found, which was surrounded by mineralized tissue, leading to the formation of tube-like structures oriented vertically in the center of the growth plate (Figures 5A and B). To further analyze these abnormal structures with regard to chondrocyte differentiation into hypertrophic cells and bone formation, we performed immunostaining using antibodies directed against types I, II, and X collagen. Using longer exposure times, we could see that type II collagen is present in unaffected regions of the growth plate but absent in the focal closure itself. In contrast, we detected type X collagen throughout the closure, indicating dysregulated chondrocyte differentiation and the presence of hypertrophic cells. The lack of type II collagen but positive staining for type I collagen implies that in this area, cartilage is converted into bone (Figure 5C).
COMP and matrilin 3 are two extracellular matrix adaptor proteins with a similar ligand repertoire (7, 9–14). Mutations in either COMP or matrilin 3 are associated with human diseases that affect skeletal development (15–17), but development proceeds normally in both COMP and matrilin 3 single-deficient mice, without any pronounced skeletal abnormalities (18, 19). To elucidate the role of COMP and matrilin 3 in osteogenesis and to investigate putative compensatory roles of the two adaptor proteins, we generated mice deficient in both molecules. We report herein that COMP/matrilin 3 double-deficient mice showed normal skeletal development as newborns (Figure 1), but displayed altered cartilage and bone remodeling at later stages of development (Figure 2), which was initially observed as a reduction in body length and long bone length. Interestingly, knockin mice for either COMP or matrilin 3 carrying mutations that cause chondrodysplasia in humans showed a similar reduction in bone length (30, 31). The phenotype of such knockin mice is similar to that seen in multiple epiphyseal dysplasia in humans, but only in animals homozygous for the mutation, which is different from the autosomal-dominant inheritance in humans.
The size reduction seen in our double-mutant animals may indicate that this phenotype occurs at least partly as a result of the lack of the protein and not only as a result of the increased apoptosis of growth plate chondrocytes, as has been reported in animals carrying mutations that cause multiple epiphyseal dysplasia (30, 31). The reduction in bone length in COMP/matrilin 3 double-deficient animals was associated with abnormal bone quality due to delayed aggrecan degradation and altered replacement of cartilage by bone in the metaphyseal trabeculae (Figure 2A).
In addition to the altered aggrecan degradation, we detected a reorganization of the collagen network. Even though the total amount of type II collagen seemed to be unchanged, the findings of both the antibody staining and the extractability experiments were markedly changed in the absence of COMP and matrilin 3. We speculate that the fibril structure is altered, since both COMP and matrilin 3 bind to collagens and are known to influence fibrillogenesis and collagen volume density, respectively (4, 6, 9, 10, 12). Moreover, consistent with the histologic findings, the physical bone parameters obtained from the peripheral QCT measurements showed alterations in the metaphysis, with an elevation of trabecular bone mineral content and density (Table 1).
Osteoclasts are responsible for the breakdown of the cartilaginous core, allowing osteoblasts to generate trabecular bone (32, 33). Alterations in the number or distribution of osteoclasts have been shown to lead to abnormalities in cartilage degradation (34). However, using TRAP staining, we found no differences in the number or distribution of osteoclasts in mutant animals as compared to controls (Figure 3A).
Several members of the MMP family are known to degrade aggrecan (35–39). MMP-mediated cleavage of aggrecan results in the formation of the neoepitope DIPEN at the C-terminus of the N-terminal fragment. In cartilage, this fragment remains bound to hyaluronan (36). The reduction in aggrecan cleavage in double-null mice could be due to a reduced amount of MMPs or to a decrease in their activity. We therefore used specific antibodies to detect different members of the MMP family, but often, the level of these MMPs was very low or even below the detection limit. The distribution of the collagenase MMP-13 and the gelatinase MMP-9, however, both reported to have aggrecanase activity (35, 36, 39, 40), was comparable in all 4 genotypes (Figure 3A). Interestingly, all 3 deficient mouse lines showed reduced total amounts of MMP-9 as compared to control animals, with matrilin 3–deficient animals having the lowest MMP-9 level. The levels of MMP-13 seemed to be generally less affected. These results are consistent with the immunostaining intensities. However, it is uncertain how well the MMP amount correlates with enzyme activity.
TIMP-3 is known to be an endogenous matrix-bound broad-spectrum inhibitor of both ADAMTS and MMP molecules (41, 42). Altered amounts or distribution of TIMP-3 could lead to changes in MMP activity. TIMP-3–deficient mice have been reported to have increased collagen and aggrecan degradation in joints (43). In our animal model, in which more TIMP-3 is deposited, we saw the opposite effects: delayed cartilage degradation and higher trabecular content. Osteoclast activation is likely to be independent of TIMP-3, since the number of osteoclasts is not affected when the protein is either lacking (43, 44) or elevated, as in our model. Both COMP and TIMP-3 can bind to the glycosaminoglycan side chains of aggrecan (14, 28), and matrilins interact with the aggrecan protein core (45, 46). TIMP-3 and COMP and/or matrilin 3 may compete for binding sites on aggrecan, and the presence of COMP and matrilin 3 could restrict TIMP-3 integration into the cartilage extracellular matrix, which would explain the broader distribution of TIMP-3 in the double-mutant mice. It is also possible that the lack of adaptor proteins, such as COMP and matrilin 3, destabilizes the extracellular matrix, which could trigger an increase in TIMP-3 deposition.
The overall appearance of the growth plate from 1-month-old double-mutant mice was normal, but in the central region of the bone, focal closures of the growth plate were observed in almost all double-null animals (Figure 5). Such focal closures have also been reported in mice deficient for MMP-13 and in mice carrying a mutant type II collagen allele (47). In double-deficient mice, MMP-13 is present in normal amounts, but the increased deposition of TIMP-3 appears to cause a similar phenotype as that seen when MMP-13 is absent. In the type II collagen–knockin mice, a single-point mutation makes the protein resistant to MMP cleavage, again leading to a phenotype similar to that seen in our double-knockout model.
The mechanisms that lead to closure of the growth plate are largely unknown. In vertebrates, where closure occurs naturally during maturation, this is initiated from the primary ossification center. In contrast, as seen in growth plates from COMP/matrilin 3 double-mutant animals, where closure is incomplete, the focal closure in our mouse model is initiated from the secondary ossification center, with cells showing hypertrophy in the reserve and proliferation zone. Cells in the upper zones of the growth plate are larger and are type X collagen positive, similar to those in the hypertrophic zone (Figures 5B and C). MMP-13–deficient mice have a higher trabecular proteoglycan content and trabecular bone mineral density (47–49), whereas mice lacking both COMP and matrilin 3 showed similar, but less pronounced, changes, indicating that MMP inhibition by TIMP-3 is incomplete. Both MMP-9– and MMP-13–deficient mice were reported to show massive enlargement of the hypertrophic zone of the growth plate (47–49). In contrast, our COMP/matrilin 3 double-deficient mice did not show these alterations, further suggesting that the MMP inhibition is only partial.
Even though COMP and matrilin 3 are structurally unrelated, they appear to share similar functions and the one can at least partly compensate for the lack of the other. Taken together, our results show that the matrix deposition of TIMP-3 is hindered by COMP and matrilin 3 and that the lack of these adaptor proteins impairs cartilage resorption upon bone formation.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Zaucke had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Groma, Xin, Grskovic, Niehoff, Brachvogel, Paulsson, Zaucke.
Acquisition of data. Groma, Xin, Grskovic, Niehoff.
Analysis and interpretation of data. Groma, Xin, Grskovic, Niehoff, Brachvogel, Paulsson, Zaucke.
We are grateful to Attila Aszódi for advice and generous sharing of antibodies.