Loss of matrilin 1 does not exacerbate the skeletal phenotype in a mouse model of multiple epiphyseal dysplasia caused by a Matn3 V194D mutation

Objective Mutations in matrilin 3 can result in multiple epiphyseal dysplasia (MED), a disease characterized by delayed and irregular bone growth and early-onset osteoarthritis. Although intracellular retention of the majority of mutant matrilin 3 was previously observed in a murine model of MED caused by a Matn3 V194D mutation, some mutant protein was secreted into the extracellular matrix. Thus, it was proposed that secretion of mutant matrilin 3 may be dependent on the formation of hetero-oligomers with matrilin 1. The aim of this study was to investigate the hypothesis that deletion of matrilin 1 would abolish the formation of matrilin 1/matrilin 3 hetero-oligomers, eliminate the secretion of mutant matrilin 3, and influence disease severity. Methods Mice with a Matn3 V194D mutation were crossed with Matn1-null mice, generating mice that were homozygous for V194D and null for matrilin 1. This novel mouse was used for in-depth phenotyping, while cartilage and chondrocytes were studied both histochemically and biochemically. Results Endochondral ossification was not disrupted any further in mice with a double V194D mutation compared with mice with a single mutation. A similar proportion of mutant matrilin 3 was present in the extracellular matrix, and the amount of retained mutant matrilin 3 was not noticeably increased. Retained mutant matrilin 3 formed disulfide-bonded aggregates and caused the co-retention of matrilin 1. Conclusion We showed that secretion of matrilin 3 V194D mutant protein is not dependent on hetero-oligomerization with matrilin 1, and that the total ablation of matrilin 1 expression has no impact on disease severity in mice with MED. Mutant matrilin 3 oligomers form non-native disulfide-bonded aggregates through the misfolded A domain.

More than 20 different missense mutations within the matrilin 3 gene have been shown to cause autosomal-dominant MED, and MATN3 mutations are believed to account for ϳ20% of all cases of autosomaldominant MED (5). The majority of MATN3 mutations affect residues that are located within the central ␤-sheet of the single A domain of matrilin 3 (5,(15)(16)(17), while a smaller proportion of mutations that affect residues in the ␣-helices of the same domain have recently been described (5,18,19).
Several different cell culture model systems have been used in the past to study the effect of matrilin 3 mutations in vitro (15,19,20). The most striking effect of MATN3 mutations was the intracellular retention of mutant protein in cells in culture and also in patient cartilage. This distinctive finding was recently verified in a murine model of MED caused by a Matn3 V194D mutation, which replicated the human phenotype by exhibiting mild short-limbed dwarfism (21,22). In this murine model, mutant matrilin 3 was retained within the endoplasmic reticulum (ER) of chondrocytes from the growth plates of endochondral bones. The retention of mutant matrilin 3 elicited a conventional unfolded protein response characterized by the up-regulation of classic chaperone proteins such as BiP/glucose-regulated protein 78-kd, calnexin, and calreticulin, as well as less-characterized ER stress-associated genes such as Creld2 and Armet/Manf (22).
Despite intracellular retention of the majority of mutant matrilin 3 in vivo, a small proportion of mutant protein was secreted into the ECM, even in mice that were homozygous for the V194D mutation (21). The structural and/or functional effect of this mutant matrilin 3 in the ECM is not known, nor is the extent to which its presence in the ECM might contribute to the chondrodysplasia phenotype. Given the widespread literature describing the hetero-oligomerization of matrilin 3 with matrilin 1 (23)(24)(25)(26), we considered the possibility that the secretion of mutant matrilin 3 may be dependent on its oligomerization into heterotetramers containing wildtype matrilin 1.
We hypothesized that deletion of matrilin 1, which would abolish the formation of matrilin 1/matrilin 3 heterotetramers, would also eliminate the secretion of mutant matrilin 3 into the ECM and increase the severity of chondrodysplasia in mice with the Matn3 V194D mutation. A study testing this hypothesis would provide new insight into the disease mechanisms of MED and the role of hetero-oligomerization on the trafficking and secretion of matrilin 1/matrilin 3 proteins. We therefore crossed mice with the Matn3 V194D (Matn3 m/m ) mutation with Matn1 Ϫ/Ϫ mice and generated a novel mouse line that was homozygous for the Matn3 V194D mutation and null for matrilin 1 (Matn1 Ϫ/Ϫ Matn3 m/m ). We investigated the phenotypic, morphologic, and biochemical characterization of this novel murine model to determine the interdependency of matrilin 1 and matrilin 3 oligomerization on the trafficking and secretion of these proteins and their effect on disease severity.

Generation of mice.
Mice with the Matn3 V194D mutation and Matn1 Ϫ/Ϫ mice were generated as previously described (21,27). These 2 transgenic lines were crossed to produce mice with the following genotypes: Measurement of bone lengths and monitoring of growth rates. Whole-mouse radiographs were obtained to measure the bone lengths of mice at ages 3, 6, and 9 weeks, using proprietary software (Certus Technology Associates Ltd.). Growth curves were produced by determining the body weights of the mice at 3, 6, and 9 weeks of age. One-way analysis of variance (ANOVA) was used to determine statistically significant (P Ͻ 0.05) differences between the different genotypes.
Cell proliferation (bromodeoxyuridine [BrdU]) and apoptosis (TUNEL) assays on tibial growth plates. BrdU and TUNEL assays were performed as described previously (21,28). BrdU-positive cells were counted and expressed as a percentage of the total number of cells within the proliferating zone. TUNEL-positive cells from each zone were counted and expressed as a percentage of the total number of cells within the different zones of the growth plate. Cross-sections from at least 3 different matched areas across the tibia growth plate for 3 mice of each genotype were used. Significant differences in chondrocyte proliferation and apoptosis between genotypes were tested using t-tests for independent samples. P values less than 0.05 were considered significant.
Western blot analysis of cartilage and chondrocyte intracellular proteins. Chondrocytes were isolated from pooled rib cartilage samples obtained from 7-day-old mice, as described previously (21,28). Chondrocyte numbers were measured using a hemocytometer, and aliquots of 1.5 ϫ 10 5 cells were prepared. Aliquots were spun in a microfuge for 10 minutes, supernatants were discarded, and the cell pellets were resuspended in 5ϫ sodium dodecyl sulfate (SDS) loading buffer. Ponceau staining was used to confirm equal loading of total protein prior to blotting with antibodies specific for matrilin 3 (supplied by R&D Systems for reduced samples and a kind gift from Raimund Wagener, PhD [University of Cologne] for nonreduced samples) and matrilin 1 (a gift from Raimund Wagener). Aliquots from 2 different biologic replicates were analyzed and normalized against a ␤-actin control.
Sequential cartilage extractions. Knee joints were dissected from the hind limbs of 3-week-old wild-type mice and flash frozen in liquid nitrogen. The joints were subsequently thawed and cut into ϳ1-mm 3 pieces. Tissue was extracted in 10 volumes (ml/mg wet tissue) of extraction buffer A (0.15M NaCl, 50 mM Tris [pH 7.4]) overnight at 4°C with continuous mixing. Extracts were spun in a microfuge for 10 minutes, and 100-l aliquots of the supernatants were stored at Ϫ20°C. Insoluble material was extracted in buffer B (4M GuHCl, 10 mM EDTA, 50 mM Tris [pH 7.4]) as previously described, and 100-l aliquots of the supernatants were stored at Ϫ20°C. All cartilage extractions took place in the presence of 2 mM phenylmethylsulfonyl fluoride and 2 mM N-ethylmaleimide. Supernatants from the extractions were ethanol precipitated with 96% ethanol, and the precipitated materials were spun in a microfuge for 10 minutes at 4°C. Pellets were washed with a mixture of 9 volumes of 96% ethanol and 1 volume of Tris buffered saline for 2 hours at 4°C. Precipitates were centrifuged for 10 minutes at 4°C, and the pellets were air-dried and resuspended in nonreducing SDS-polyacrylamide gel electrophoresis (PAGE) sample buffer. Twenty-microliter aliquots were analyzed by SDS-PAGE/Western blotting.

Infrared detection of Western blots.
Following Western blotting of wild-type cartilage extracts (buffer B), primary antibodies against matrilin 1 (polyclonal rabbit; a kind gift from Raimund Wagener) and matrilin 3 (polyclonal goat; R&D Systems) were applied simultaneously to the same nitrocellulose membrane. Anti-goat and anti-rabbit secondary antibodies (Li-Cor IRDye 680 and Li-Cor IRDye 800CW, respectively) were subsequently applied together to detect both matrilin 1 and matrilin 3 in the same lane of the same membrane. The Western blot was then imaged using an Odyssey Infrared Imaging System (Li-Cor).
Cell culture primary chondrocytes. Costal chondrocytes from 5-day-old mice were extracted as described above, and the washed cell pellet was resuspended in 5 ml of cell culture medium (Dulbecco's modified Eagle's medium-Ham's F-12 supplemented with 10% fetal calf serum, 140 M ascorbate, 2 mM L-glutamine, 100 units/ml penicillin, and 100 units/ml streptomycin). Chondrocytes were seeded onto a 12-well plate or an 8-well Permanox Chamber Slide (Nunc) at densities of 2.5 ϫ 10 5 and 2 ϫ 10 4 cells/cm 2 , respectively, and the culture medium was replaced after 2 days. Cell lysate and culture media were collected from 12-well plate cultures after 5 days, while chamber slide cultures were fixed in 4% paraformaldehyde/phosphate buffered saline (PBS) for 10 minutes prior to immunocytochemical analysis.
Immunocytochemical analysis of primary chondrocytes. Fixed chondrocytes were permeabilized with 0.2% Triton-PBS for 8 minutes, washed in PBS, blocked with 2% donkey serum-PBS, and incubated with primary antibodies against either matrilin 1 (Atlas Antibodies) or matrilin 3 (R&D Systems) for 1 hour. Chondrocytes were then washed with PBS, incubated with secondary antibody (Alexa Fluor 555-conjugated donkey anti-rabbit or Alexa Fluor 488conjugated donkey anti-goat; Invitrogen), and washed in PBS before mounting in Vectashield medium with DAPI (Vector) and imaging.

RESULTS
Effect of matrilin 1 deletion on disease severity in mice with the Matn3 V194D mutation. To determine the phenotypic consequences of matrilin 1 knockout on the skeletal development of mice with the Matn3 V194D mutation, radiographs were obtained at 3, 6, and 9 weeks of age and used to measure bone lengths. The tibia and femur were used to assess endochondral bone growth, while the inner canthal distance and skull length were used to monitor intramembranous ossification ( Figure  1A and data not shown). The bone length measurements in mice of all 3 genotypes, Matn1 ϩ/ϩ Matn3 m/m , Matn1 ϩ/Ϫ Matn3 m/m , and Matn1 Ϫ/Ϫ Matn3 m/m , showed no significant differences at any time point. The analyses of growth curves also revealed no phenotypic effects due to the deletion of matrilin 1 ( Figure 1B).
Effect of matrilin 1 deletion on growth plate disorganization in mice with the Matn3 V194D mutation. Hematoxylin and eosin staining was performed to determine the effect of matrilin 1 deletion on organization of the tibial growth plates of Matn3 m/m mice ( Figure 2A). The appearance and morphology of the growth plates from Matn1 Ϫ/Ϫ mice were consistent with those described in previous studies (31). Furthermore, the growth plates of 3-week-old Matn1 ϩ/ϩ Matn3 m/m and Matn1 Ϫ/Ϫ Matn3 m/m mice showed disruption to the morphology and organization of chondrocytes, which was comparable with our previous findings in Matn3 m/m mice (21). For example, chondrocytes in the proliferative zone in both genotypes appeared misshapen, and their arrangement into chondrons was disturbed when compared with wild-type and Matn1 Ϫ/Ϫ control mice ( Figure 2A).

Secretion of mutant matrilin 3 into the ECM in
the absence of matrilin 1. Given the fact that matrilin 1 and matrilin 3 monomers form hetero-oligomers, we used immunohistochemistry to analyze the tibial growth plates of 3-week-old mice to determine whether the deletion of matrilin 1 might abrogate the secretion of mutant matrilin 3. In wild-type and Matn1 Ϫ/Ϫ mice, normal matrilin 3 was secreted into the ECM, as expected. In contrast, intracellular retention of mutant matrilin 3 was observed in mice with the Matn3 V194D mutation that were either null or wild-type for matrilin 1 (i.e., Matn1 ϩ/ϩ Matn3 m/m and Matn1 Ϫ/Ϫ Matn3 m/m mice, respectively) ( Figure 2B). However, it was also obvious that a proportion of mutant matrilin 3 was still secreted into the ECM independent of matrilin 1. In wild-type and Matn1 Ϫ/Ϫ mice, staining for matrilin 3 appeared more pronounced in the pericellular matrix surrounding the chondrons. In contrast, staining for mutant matrilin 3 was more evenly distributed in the interterritorial regions between chondrocyte columns of growth plate cartilage in Matn1 ϩ/ϩ Matn3 m/m and Matn1 Ϫ/Ϫ Matn3 m/m mice. Effect of matrilin 1 deletion on the localization of other ECM proteins in the growth plates of mice with the Matn3 V194D mutation. Given the predicted role of the matrilin proteins as adaptor molecules, we used immunohistochemistry to probe the localization of key ECM molecules in Matn1 Ϫ/Ϫ Matn3 m/m mice. At 3 weeks of age, the spatial localization of collagen types II, VI, IX, and X, COMP, and fibronectin were comparable between Matn1 Ϫ/Ϫ Matn3 m/m and Matn1 ϩ/ϩ Matn3 m/m mice ( Figure 2C and data not shown). Furthermore, there were no apparent differences in the localization of these key proteins due to the deletion of Matn1 alone, with the exception of type IX collagen, the staining for which was consistently higher at the boundary between the resting zone and secondary center of ossification in wild-type and Matn1 ϩ/ϩ Matn3 m/m mice. This observation mirrors the increased staining for matrilin 1 at the same location ( Figure 2B) and suggests that these proteins interact in vivo.
Effect of matrilin 1 deletion on the reduced rate of chondrocyte proliferation in mice with the Matn3 V194D mutation. We previously showed that reduced chondrocyte proliferation in the growth plate was a key disease mechanism in a murine model of MED caused by a Matn3 V194D mutation (21). We therefore performed BrdU labeling to determine whether the deletion of matrilin 1 would further reduce the rate of chondrocyte proliferation ( Figure 3A). One-way ANOVA revealed no statistically significant differences between the proportions of proliferating chondrocytes in mice of all 3 genotypes.
Influence of matrilin 1 deletion on spatially dysregulated chondrocyte apoptosis in mice with the Matn3 V194D mutation. We previously showed that dysregulated chondrocyte apoptosis in the growth plate is another key disease mechanism in mice with the Matn3 V194D mutation (21). We therefore calculated the relative rates of chondrocyte apoptosis within the tibial growth plates of 3-week-old mice to determine whether the deletion of matrilin 1 influences chondrocyte apoptosis ( Figure 3B). Based on these data, it was evident that the relative levels of apoptosis within the different growth plate zones were not significantly different between Matn1 ϩ/ϩ Matn3 m/m and Matn1 Ϫ/Ϫ Matn3 m/m mice, when using t-tests for independent samples.
Molecular characterization of retained mutant matrilin 3. It has previously been demonstrated that (matrilin 1) 2 (matrilin 3) 2 heterotetramers can be isolated from human and bovine cartilage (23,24,26), but the existence of this molecular form in mouse cartilage has not been convincingly demonstrated to date (32). We therefore performed Western blot analysis of GuHCl-extracted mouse cartilage proteins, using direct infrared fluorescence detection with a Li-Cor Odyssey Imaging System, to confirm the existence of matrilin 1/matrilin 3 heterotetramers in mouse cartilage ( Figure  4A). However, it was clear from this experiment that these heterotetramers were less abundant than the respective homotrimers and homotetramers of matrilin 1 and matrilin 3.
We hypothesized that the deletion of matrilin 1 might increase the relative amount of mutant matrilin 3 retained within chondrocytes by preventing the formation and secretion of these matrilin 1 (wild-type)/ matrilin 3 (mutant) hetero-oligomers. We tested this hypothesis by isolating chondrocytes from the pooled rib cartilage specimens of 7-day-old littermates (ϳ5-7 pups per litter) and analyzing the intracellular proteins by SDS-PAGE and Western blotting. The complete absence of any detectable matrilin 3 in the wild-type cell extracts demonstrated that the cellular protein preparations were not contaminated with any secreted matrilin 3 protein derived from the ECM ( Figure 4B). Subsequent experiments reproducibly showed that the relative amounts of intracellular mutant matrilin 3 protein did not differ between Matn1 ϩ/ϩ Matn3 m/m mice and Matn1 Ϫ/Ϫ Matn3 m/m mice ( Figure 4C, part i).
In order to determine the oligomeric state of the retained mutant matrilin 3, we also analyzed nonreduced protein samples, using SDS-PAGE. This analysis indicated that although some of the retained mutant matrilin 3 was present as distinct oligomers, the vast majority of mutant protein existed as high molecular weight disulfide-bonded aggregates ( Figure 4C, part ii) independent of the presence or absence of matrilin 1. This aggregated form of mutant matrilin 3 was confirmed by examining sequential protein extracts from total mice cartilage ( Figure 4C, part iii), which showed that mutant matrilin 3 was readily extracted by Tris-NaCl (buffer A), while GuHCl (buffer B) was required to extract wildtype matrilin 3. In addition, wild-type matrilin 3 appeared to exist in 2 oligomeric forms that corresponded in size to trimers and tetramers; however, tetramers were clearly the more prevalent form.
A proportion of matrilin 1 was co-retained with mutant matrilin 3 in Matn1 ϩ/ϩ Matn3 m/m mouse chondrocytes ( Figure 4D, part i), which appeared consistent in size with the putative (matrilin 1) 2 (matrilin 3) 2 heterotetramer extracted from wild-type cartilage when analyzed under nonreducing conditions and the image exposure was increased ( Figure 4D, part ii). The obser-vation of apparently high molecular weight matrilin 1-containing aggregates most likely arose from the increased exposure of the image, and such aggregates were not observed upon reduction.
Molecular characterization of secreted mutant matrilin 3. To determine the molecular organization of secreted mutant matrilin 3, we analyzed the cell/celllayer and media proteins of cultured mouse chondrocytes ( Figure 5A). Western blot analysis of wild-type samples showed the presence of trimers and tetramers, which was consistent with the molecular forms observed following GuHCl extraction of cartilage; in both cases, tetramers were the predominant form. Fluorescence microscopy confirmed that wild-type matrilin 3 was present in pericellular networks, and that there was no intracellular retention ( Figure 5B). In contrast, mutant matrilin 3 existed predominantly in a tetrameric form in both the cell/cell-layer and media of cultured Matn1 ϩ/ϩ Matn3 m/m chondrocytes ( Figure 5A), while the presence in the cell/cell-layer fraction of larger aggregates was comparable with that observed in isolated chondrocytes. Furthermore, intracellular retention of mutant matrilin 3 was verified by fluorescence microscopy, which demonstrated a punctate appearance and also confirmed the absence of matrilin 3 from the pericellular networks ( Figure 5B).
Matrilin 1 was present in a tetrameric form in both the cell/cell-layer and media samples of cultured Matn1 ϩ/ϩ Matn3 ϩ/ϩ chondrocytes ( Figure 5C), and the pericellular localization of matrilin 1 was confirmed by fluorescence microscopy ( Figure 5D). Finally, these studies confirmed the retention of considerable amounts of matrilin 1 in the cell/cell-layer fraction of cultured Matn1 ϩ/ϩ Matn3 m/m chondrocytes ( Figure 5C), which also appeared to have an intracellular punctate appearance ( Figure 5D).
Possible mechanism of aggregation of mutant matrilin 3 oligomers. We previously demonstrated the intracellular retention of recombinant mutant V194D matrilin 3 A domains following reducing SDS-PAGE (15). Here, we showed that the retained V194D mutant A domains formed a regular pattern of disulfide-bonded aggregates ( Figure 5E), which could be resolved to a single band following reduction. This observation was consistent with the disulfide-bonded aggregates of full-length mutant matrilin isolated from Matn1 ϩ/ϩ Matn3 m/m chondrocytes and cartilage ( Figure 4C) and points to the misfolded mutant A domain as the nucleation site for the formation of the non-native disulfide-bonded fulllength mutant matrilin 3 aggregates.

DISCUSSION
In this study, we explored the effect of Matn1 deletion on the phenotype of mice that were homozygous for a MED-causing matrilin 3 mutation (Matn3 V194D) (21). We hypothesized that the relatively low levels of mutant matrilin 3 in the cartilage ECM of mice with the Matn3 V194D mutation might be secreted through the formation of (matrilin 1 ϩ/ϩ ) n (matrilin 3 m/m ) n hetero-oligomers. The formation of a hetero-oligomeric form of matrilin 1/matrilin 3 in vivo has previously been demonstrated for human and bovine cartilage (23)(24)(25)(26) and is now confirmed for mice in this study, but its physiologic relevance and the interdependency of matrilin 1 and matrilin 3 monomers in oligomerization, protein trafficking, and secretion remain unknown.
We evaluated the phenotype of Matn1 Ϫ/Ϫ Matn3 m/m mice and demonstrated that the deletion of Matn1 had no effect on the relative growth rate and long bone lengths of mice with the Matn3 V194D mutation. These observations are in accordance with findings for Matn1-knockout mice, which showed no overt skeletal phenotype (31). Our analyses also confirmed that there were no further differences in the relative levels of chondrocyte proliferation and apoptosis within the growth plates of mice with the Matn3 V194D mutation due to the deletion of Matn1. We previously proposed that these detrimental changes in the chondrocyte phenotype are key disease mechanisms in MED caused by the Matn3 V194D mutation (21,22) and also in pseudoachondroplasia-MED caused by Comp p.T585M (28) and Comp p.D469del (33) mutations. Overall, these data suggest that changes in the levels of matrilin 1 expression would not modify disease severity in patients with MED and therefore do not account for the interfamilial variability seen in some families with MED and MATN3 mutations (34).
The analysis of growth plate morphology and the spatial distribution of key structural proteins such as type IX collagen and COMP showed some variation between mice of different genotypes, possibly because matrilin 1/matrilin 3, COMP, and type IX collagen have been shown to interact in vitro (7,13). For example, there was reduced staining for type IX collagen at the boundary between the resting zone and the secondary center of ossification in the growth plates of mice that lacked matrilin 1. This reduced staining for type IX collagen was inversely proportional to the higher levels of staining for matrilin 1 in the same region in wild-type and Matn1 ϩ/ϩ Matn3 m/m mice, suggesting that these 2 molecules interact in vivo, and that matrilin 1 is important for anchoring type IX collagen into the ECM. This observation confirms recent in vitro binding studies showing interactions between type IX collagen and the A domains of matrilin 1 (13). Type IX collagendeficient mice also showed altered integration of matri-lin 3 into the ECM (10), confirming that these proteins interact in vivo.
Interestingly, immunohistochemical analysis revealed that although staining for wild-type matrilin 3 was localized primarily to the pericellular matrix surrounding chondrocytes, mutant matrilin 3 staining was spread evenly throughout the territorial matrix between chondrocyte columns, irrespective of the matrilin 1 genotype. Similar observations were made in the original characterization of mice with the Matn3 V194D mutation and may suggest that the secreted mutant matrilin 3 is less able to incorporate correctly into the pericellular matrix. Sequential extractions of mutant mouse cartilage showed increased extractability of mutant matrilin 3, which was present predominantly as disulfide-bonded high molecular weight aggregates; however, this approach is limited, because it is likely that extracted intracellular mutant proteins will mask ECM-derived proteins. In contrast, wild-type matrilin 3 was extracted only with GuHCl, suggesting that it is tightly integrated into the ECM. Indeed, fluorescence microscopy of cultured wild-type mouse chondrocytes showed that matrilin 3 was present in filamentous networks surrounding the cells.
We also used this novel murine model to probe the nature of matrilin 1/matrilin 3 hetero-oligomers in vivo. Immunohistochemical analysis and Western blotting of intracellular proteins were used to test the hypothesis that the ablation of matrilin 1 would alter the secretion of matrilin 3 V194D and result in increased levels of retained mutant matrilin 3. However, immunohistochemistry showed that the deletion of matrilin 1 did not abrogate the secretion of mutant matrilin 3, because ECM staining for mutant matrilin 3 was similar in Matn1 ϩ/ϩ Matn3 m/m and Matn1 Ϫ/Ϫ Matn3 m/m mice. Western blotting of extracted intracellular proteins from Matn1 ϩ/ϩ Matn3 m/m and Matn1 Ϫ/Ϫ Matn3 m/m chondrocytes confirmed that the intracellular levels of retained mutant matrilin 3 protein were not altered. Furthermore, the analysis of cultured Matn1 ϩ/ϩ Matn3 m/m chondrocytes confirmed that a small proportion of mutant matrilin 3 was secreted as a putative tetramer, although the majority still existed in punctate intracellular inclusions.
Intracellular matrilin 1 was also detected at low levels by Western blotting of total chondrocyte protein following reducing SDS-PAGE. The results of SDS-PAGE run under nonreducing conditions suggested that these oligomers might represent (matrilin 1) 2 (matrilin 3) 2 heterotetramers. This observation suggests that matrilin 1 monomers are being retained due to their association with mutant matrilin 3 monomers in heterotetramers. Although we could not demonstrate any phenotypic differences between Matn1 ϩ/ϩ Matn3 m/m and Matn1 Ϫ/Ϫ Matn3 m/m mice, we cannot exclude the possibility that the retention of these hetero-oligomers, along with the more prevalent matrilin 3 homo-oligomers, contributes to the etiology of MED. The apparent differences in the levels of retained matrilin 1 between isolated and cultured chondrocytes are attributable to the use of collagenase for the isolation of ECM-free chondrocytes. In contrast, both intracellular and celllayer proteins (pericellular) are collected from cultured chondrocytes by cell scraping (i.e., "cell-layer" samples). Overall, these findings suggest that the majority of intracellular and extracellular mutant matrilin 3 appears to be in the form of (matrilin 3) 4 , and that (matrilin 1) wild-type (matrilin 3) mutant hetero-oligomers represent only a small fraction, which is consistent with the relative proportion of wild-type heterotetramers compared with the respective homotrimers and homotetramers of matrilin 1 and matrilin 3.
Finally, using nonreducing SDS-PAGE and Western blotting, we showed that the retained mutant matrilin 3 forms disulfide-bonded aggregates in chondrocytes, and that the mutant A domain alone (V194D) also formed similar disulfide-bonded aggregates in vitro. We previously demonstrated that various missense mutations in the ␤-strands of the A domain delayed folding and prevented correct intramolecular disulfide bond formation (15). In this study, we extended these findings and demonstrated that the mutant A domains form aggregates through non-native intermolecular disulfide bonds. It is likely that the increased expression of several members of the protein disulfide isomerase-associated protein family (i.e., Pdia4, Pdia5, and Pdia6) in chondrocytes from mice with the Matn3 V194D mutation (22) may be the direct result of incorrect disulfide bond formation between mutant matrilin 3 monomers nucleated at the A domain.
In conclusion, we explored the effect of matrilin 1 deletion in mice homozygous for a MED-causing matrilin 3 mutation. We observed no alteration to the MEDlike phenotype of mutant mice and no relative increase in the quantity of intracellular mutant matrilin 3, suggesting that mutant matrilin 3 exists primarily as a homo-oligomer. The observation that chondrocytes can potentially fold and secrete a proportion of mutant matrilin 3 oligomers supports the idea that treatment with small molecular chaperones might aid in the secretion of mutant matrilin 3 and pave the way for the development of suitable therapies (22).