Analysis of the cartilage proteome from three different mouse models of genetic skeletal diseases reveals common and discrete disease signatures

The chondrodysplasias are a diverse group of many different phenotypes that arise when endochondral bone growth is disrupted and include two clinically related phenotypes; multiple epiphyseal dysplasia MED and pseudoachondroplasia PSACH Warman et al., 2011. MED is genetically heterogeneous and can be caused by mutations in genes encoding the extracellular matrix ECM proteins matrilin-3, type IX collagen and cartilage oligomeric matrix protein COMP Briggs and Chapman, 2002; Jackson et al., 2012; PSACH results exclusively from COMP mutations Jackson et al., 2012 and is more severe than MED, but both phenotypes comprise a disease spectrum with symptoms that can include joint pain and stiffness, lower-limb deformities and early onset osteoarthritis Briggs and Chapman, 2002.

We have previously generated targeted mouse models of PSACH-MED with mutations in matrilin-3 moderate MED: Matn3 V194D and COMP mild PSACH: Comp T585M and severe PSACH: Comp DelD469 and described the resulting phenotypes in detail Leighton et al., 2007; Pirg-Garcia et al., 2007; Suleman et al., 2012. Briefly, all three mice models exhibit disproportionate short stature due to decreased chondrocyte proliferation and increased and/or spatially dysregulated apoptosis in the cartilage growth plate.

Whilst the cellular response to mutant matrilin-3 and COMP expression has been documented in detail Bell et al., 2012; Leighton et al., 2007; Pirg-Garcia et al., 2007; Suleman et al., 2012, the effects of these mutant proteins on the organisation and composition of the cartilage ECM are not clearly defined. Immunohistochemical IHC analysis has previously identified differences in the staining pattern for matrilin-3, COMP and type IX collagen in the ECM of mutant growth plates Leighton et al., 2007; Pirg-Garcia et al., 2007; Suleman et al., 2012, whilst the appearance of the ECM ultra-structure is also different in all three mutant mice. For example, the collagen fibrils in the inter-territorial matrix were more clearly visible by electron microscopy EM suggesting that lower levels of fibril surface-associated proteins were decorating individual collagen fibrils Leighton et al., 2007; Pirg-Garcia et al., 2007; Suleman et al., 2012. Similar changes to the organisation of the cartilage pericellular matrix in mice expressing a COMP DelD469 transgene were also noted by EM and IHC Schmitz et al., 2008. Furthermore, recent studies demonstrated that a secreted variant of COMP carrying a MED-mutation in the C-terminal domain p.H587R disrupted collagen fibrillogenesis in vitro and also in a cell culture model Hansen et al., 2011. In contrast, COMP-null mice show no differences in collagen fibril diameter Svensson et al., 2002 and these data therefore suggest that PSACH and MED-causing mutations in matrilin-3 and COMP can cause changes in the ultra-structure of the cartilage ECM through antimorphic mechanisms. However, the extent to which these changes compromise the organisation and function of the ECM, and therefore contribute to disease pathology and cartilage degradation, remain to be determined.

The unique biochemical properties of cartilage have made the study of its ECM technically difficult relative to other tissues Wilson et al., 2009. Recently, however, methods have been described that allow the reproducible analysis of cartilage proteins by Western blotting, 2-D gel electrophoresis and mass spectrometry MS Wilson et al., 2009. However, these methods have not been employed to study cartilage preparations from a heterogeneous series of gene-targeted mouse models of genetic skeletal diseases.

We hypothesised that the expression of mutant matrilin-3 and COMP would induce changes to the extractability of other cartilage components, which would be indicative of alterations to the anchoring or associations of individual molecules and the overall functional properties of the tissue. We tested this hypothesis by analysing the proteins isolated from sequential extractions of cartilage from mutant mouse models with different genotypes. This study describes for the first time the proteomic characterisation of cartilage from Matn3 V194D, Comp T585M and Comp DelD469 mutant mice using a combination of Western blotting and semi-quantitative spectral counting mass spectrometry and identifies both common and unique disease signatures that also provide insight into detrimental changes that may predispose to cartilage degeneration.

In the first instance we used a candidate approach to determine the extractability of a number of ECM molecules including members of the matrilin protein family, type IX collagen and decorin, which are all known to interact with matrilin-3 and COMP Budde et al., 2005; Fresquet et al., 2010; Fresquet et al., 2007; Holden et al., 2001; Mann et al., 2004; Wiberg et al., 2003. Articular and epiphyseal cartilage from 3-week-old mouse knee joints was sequentially extracted in a series of three buffers and the extracted proteins were separated by SDS-PAGE and visualised by Western blotting. The extraction of decorin appeared consistent between all genotypes and was present primarily in buffer 3 Fig.1; in contrast, the extraction of matrilins 14, type IX collagen and COMP all showed obvious genotype-specific differences in one or more of the disease models Figs1, 2. Whilst there were numerous minor changes in protein extractability, particularly slight differences in oligomeric forms, we chose to concentrate on the more obvious changes, which we considered would have the greatest influence on ECM organization.

Matrilin-1 was slightly less extractable from Matn3 V194D and Comp DelD469 cartilage with buffer 2 and several higher-order oligomeric forms of matrilin-1 were not extractable from Matn3 V194D cartilage, but were readily extracted from the cartilage of all other genotypes Fig.1. There was also a reduction in the amount of matrilin-2 oligomers extracted in buffers 2 and 3 from Comp T585M and Comp DelD469 cartilage relative to the control and Matn3 V194D cartilages Fig.1. A decrease in the extraction of matrilin-4 dimers and trimers from Matn3 V194D cartilage buffer 2 samples was observed compared to the other genotypes, and an increase in the extraction of some matrilin-4 oligomers that were slightly smaller in size than full-length trimers was also noted in all three mutant genotypes relative to the wild type control Fig.1.

The most striking difference was that it was not possible to extract type IX collagen from either wild type or Matn3 V194D cartilage using any of the buffer conditions, whereas a single molecular form corresponding in size to full length 1IX chain was readily extractable from both Comp DelD469 and T585M cartilages using buffers 2 and 3 Fig.1. Significantly, these changes in the extractability of type IX collagen were not the result of differences in the total levels of this protein since IHC and microarray analyses have consistently demonstrated comparable levels in mice of all three mutant genotypes that are similar to wild type controls supplementary material Figs S1, S2Leighton et al., 2007; Pirg-Garcia et al., 2007; Suleman et al., 2012.

In agreement with previous studies showing the intracellular aggregation of mutant matrilin-3 Bell et al., 2012, we observed that matrilin-3 was extracted with buffers 13 from Matn3 V194D cartilage as a high molecular weight aggregate under non-reducing conditions Fig.2A. We also noted that a small amount of matrilin-3 was also extracted with buffer 1 in Comp DelD469 and Comp T585M cartilages compared to wild type Fig.2A, insert. Furthermore, there were also differences in the molecular forms of matrilin-3 that were extracted from Comp DelD469 and Comp T585M cartilages. In Comp T585M cartilage two oligomeric forms of approximately 200kDa were extracted in near equal quantities, whilst the smaller of these two forms was entirely absent in Comp DelD469 cartilage Fig.2A, insert. The sizes of these oligomers suggested that they may represent both intact and a smaller proteolytically processed matrilin-3 tetramer. There were no discrete oligomers present in Matn3 V194D cartilage.

Analysis of cartilage sequential extractions revealed differences in the extractability of COMP between mice of all three genotypes Fig.2B. For example, when analyzed under non-reducing conditions a greater quantity of COMP was extracted in buffers 1 and 2 from Matn3 V194D cartilage compared to both Comp mutant mice and the wild type control Fig.2B, top panel. The extraction of COMP from Comp T585M and Comp DelD469 mice was also increased in all buffers compared to the wild type control, although the majority of COMP from both Comp T585M and Comp DelD469 cartilage was extracted with buffer 3, which was in contrast to Matn3 V194D cartilage. Finally, COMP extracted from all mutant genotypes appeared larger in apparent molecular weight compared to COMP extracted from wild type cartilage under non-reducing conditions.

Interestingly, when protein samples were resolved under reducing conditions Western blotting consistently showed the complete absence of a single COMP fragment 90kDa from both Matn3 V194D and Comp DelD469 buffer 2 and buffer 3 extractions, which might represent full-length monomeric COMP Fig.2B, bottom panel. Surprisingly, a number of higher migrating bands were also detected by the COMP antibody between 120150kDa in all three genotypes and these proteins were most prevalent in buffer 3 samples; however, the nature of these bands remains unclear.

We hypothesized that the wild type matrilin-3 extracted in buffer 1 from Comp DelD469 and Comp T585M samples might represent intracellular retained protein Fig.2B, top panel. We tested this hypothesis by performing SDS-PAGE and Western blotting of proteins isolated from the chondrocytes of 5-day-old mutant Comp mice. This analysis confirmed that wild type matrilin-3 was co-retained with mutant COMP in Comp DelD469 chondrocytes as a tetramer Fig.3A, DelD469 in left panel. However, the level of co-retention of wild type matrilin-3 in Comp DelD469 chondrocytes was low in comparison to the retention of mutant matrilin-3 in Matn3 V194D chondrocytes Fig.3A, compare V194D with DelD469 in right panel. Furthermore, the retained wild type matrilin-3 was not present as high molecular weight aggregates in the Comp DelD469 chondrocytes not shown. In contrast, we were not able to demonstrate the significant intracellular retention of matrilin-3 in isolated Comp T585M chondrocytes not shown, which is consistent with the T585M form of mutant COMP being efficiently secreted Pirg-Garcia et al., 2007. This latter finding therefore suggests that increased matrilin-3 extraction with buffer 1 from Comp T585M is most likely due to disrupted anchoring/interactions directly due to the T585M mutation in COMP.

Sequential protein extractions using buffers 1 and 2 indicated that COMP was more readily extracted from Matn3 V194D cartilage than wild type cartilage Fig.2B, bottom panel and we hypothesized that at least some of this COMP could represent intracellular co-retained protein. We therefore analyzed proteins extracted from isolated chondrocytes of Matn3 V194D cartilage and free from any contaminating ECM. Wild type COMP was detected within the chondrocytes of Matn3 V194D mice but at a lower concentration than the mutant COMP detected in Comp DelD469 chondrocytes Fig.3B. Under non-reducing conditions the retained wild type COMP within Matn3 V194D chondrocytes appeared as a single high molecular weight oligomeric form >250kDa, which was consistent with a tetrameric form of COMP Fig.3B, V194D in left panel. In contrast, several oligomeric forms of mutant COMP were detected in Comp DelD469 chondrocytes Fig.3B, DelD469 in left panel, which could be resolved to monomers under reducing conditions Fig.3B, right panel.

Given that it was not practical to analyze the extraction profiles for every cartilage ECM component using our candidate approach, we used a semi-quantitative proteomic method to compare cartilage protein extractions in a global and non-biased manner. For this investigation we used a method of in-gel trypsin digestion of total proteins sequentially extracted from mouse cartilage followed by LC-MS/MS analysis. In order to quantify the relative levels of each protein identified we used spectral counting, which has been established as a valid method for this approach Liu et al., 2004; Lundgren et al., 2010.

We compared the relative quantities of peptide spectra that were detected by LC-MS/MS in wild type, Matn3 V194D, Comp DelD469 and Comp T585M cartilage extracted by buffers 1, 2 and 3. The number of spectra identified for each protein were compared between wild type and mutant genotypes using the beta binomial test Pham et al., 2010. Table1 summarizes the total number of proteins identified and the number of proteins that were significantly changed in quantity between mutant and wild type cartilage samples. The individual genes that encode the relevant proteins were also subjected to DAVID functional analysis in order to group proteins sharing common molecular functions, cellular compartment or biological processes into annotation clusters see supplementary material Tables S1S3 for a list of all proteins significantly changed between mutant and wild type cartilage. In the first instance we chose to focus on changes in the extraction profiles of cartilage ECM proteins.

In order to identify both commonalities and differences in the extraction profiles of structural ECM components from the three different mouse models of PSACH-MED, we compared the different MS-derived peptide profiles unweighted spectra number supplementary material Tables S1S3 from mice of all three genotypes Fig.4.

We noted common extraction profiles for two FACIT fibril associated collagens with interrupted triple helices collagen polypeptide chains; 1XII had decreased extractability in Matn3 V194D and Comp T585M cartilage with buffer 2, whilst 1XIV had decreased extractability in Matn3 V194D and Comp DelD469 cartilage with buffers 1 and 2 Fig.4B. In contrast 1XIV had increased extractability with buffer 3 from Comp T585M cartilage Fig.4A.

Interestingly, in Comp T585M cartilage 1IX had marginally decreased extractability in buffer 1, but considerably increased extractability in buffer 3, which partially validated the observations from the SDS-PAGE analysis Fig.1. Surprisingly, the only significant change in type IX collagen extractability in cartilage from Comp DelD469 mice was a slight decrease in buffer 1; whilst a trend towards increased extractability from buffer 3 samples was observed P0.072, even though there had been clear differences in SDS-PAGE analysis of buffer 3 samples. Overall these observations suggest that important interactions between COMP/matrilin-3 and various FACIT collagens types IX, XII and XIV are disrupted by the PSACH-MED mutations.

Tenascin X was noticeably less extractable from Comp DelD469 cartilage buffer 3, whilst Tenascin C was less extractable from both Comp DelD469 and Matn3 V194D cartilage buffer 3 Fig.4B, suggesting an important role for tenascin proteins in disrupted cellmatrix interactions within these mutant cartilages. Finally, alpha-2-HS-glycoprotein fetuin-A, which has anti-calcification properties was less extractable from both Matn3 V194D and Comp T585M cartilage buffer 3 Fig.4B.

In addition to the common profiles, we also noted unique ECM extraction profiles for Matn3 V194D matrilin-3 and decorin, Comp T585M 1XI and fibronectin and Comp DelD469 1VI, nidogen, thrombospondin-3 and tenascin X mice, suggesting that genotype specific differences in cartilage ECM organization may also contribute to disease pathogenesis Fig.4B.

In contrast to ECM proteins with decreased extractability, the only common profile for increased extractability was for laminin 2 in Matn3 V194D and Comp T585M cartilage buffer 3, whilst unique extraction profiles were noted for Matn3 V194D the SLRPs biglycan and osteomodulin and also 3VI, Comp T585M 6VI, 1IX and 1XIV and Comp DelD469 1I and thrombospondin-4 cartilage samples Fig.4A.

The spectral count data were verified using SDS-PAGE and Western blot analyses of the cartilage protein extractions Fig.5. For this validation of the LC-MS/MS approach we chose to focus only on those ECM proteins that were commonly decreased in extractability in more than one mouse model e.g. types XIV and XII collagen and tenascin C, or highly decreased in extractability in a single mouse model tenascin X in DelD469 COMP Fig.2; supplementary material Tables S1S3.

Following SDS-PAGE and Western blotting the band intensities of proteins largely mirrored the spectral count data derived from mass spectrometry analyses. For example, there was a clear reduction in type XIV in protein samples from Matn3 V194D and Comp DelD469 cartilage buffers 1 and 2 Fig.5A, top panel that closely correlated with the LC-MS/MS data supplementary material Tables S1S3. Type XII collagen reactivity was also correspondingly reduced in Matn3 V194D samples buffer 2, but it was also clear that there were other changes in extractability using buffers 1 and 3 that had not been detected using LC-MS/MS Fig.5A, bottom panel. In particular, type XII collagen was less extractable in buffer 1 from the cartilage of all mutant mice.

Analysis of tenascin C indicated that it was specifically the largest protein isoform I2: 250kDa that was reduced in Matn3 V194D and Comp DelD469 buffer 3 extractions when compared with wild type and Comp T585M cartilages Fig.5B, top panel. In addition, reduced extractability of tenascin X was detected in cartilage from both the Comp T585M buffer 3 only and Comp DelD469 buffers 2 and 3 mice Fig.5B, bottom panel that was slightly different to spectral count data, which had only revealed a significant decrease in extractability of tenascin X in Comp DelD469 buffers 3 samples. There was no difference in the extraction of tenascin X from wild type and Matn3 V194D cartilages, which was consistent with LC-MS/MS data.

In addition to identifying changes in the extractability of cartilage ECM proteins we were also interested to know if the proteomic-derived data correlated with the disease specific signatures previously defined using gene expression studies. For this comparison we compared the proteome from Matn3 V194D cartilage supplementary material Table S1 with previously published gene array data for this mouse model Nundlall et al., 2010. The LC-MS/MS approach identified significantly increased protein levels of several chaperones and foldases including Grp94, Grp78 BiP and calreticulin Calr, members of the protein disulphide isomerase family PDIA 3, 4 and 6 and two novel ER-stress induced proteins, CRELD2 and Armet/Manf. Significantly, all these proteins were previously shown to be transcriptionally upregulated by microarray analysis Nundlall et al., 2010. Conversely, the extraction of proteins associated with intracellular protein trafficking were significantly reduced in Matn3 V194D samples e.g. buffer 1: coatomer subunit beta COPB2, coatomer subunit delta COPD and GTPase Ran supplementary material Table S1, which is consistent with the intracellular retention of mutant matrilin-3 Fig.3. Furthermore, the amount of proliferating nuclear cell antigen PNCA was decreased in Matn3 V194D samples, concomitant with the decreased rate of chondrocyte proliferation previously observed in these mutant mice Leighton et al., 2007.

In this study we investigated the cartilage protein extraction profile of three gene targeted mouse models of phenotypically-related chondrodysplasias using both candidate and global approaches in order to gain insight into ECM disorganization and derive new perspectives on disease mechanisms and cartilage degradation. This is the first time that such a dual approach has been used to analyse cartilage from genetic models of human ECM diseases.

SDS-PAGE and Western blot analysis of sequentially extracted knee cartilage revealed genotype-specific differences in the extraction of a number of proteins that included matrilin-1 to -4, COMP and type IX collagen, which are all known to interact with each other Budde et al., 2005; Fresquet et al., 2010; Fresquet et al., 2007; Holden et al., 2001; Mann et al., 2004; Wiberg et al., 2003. While we were unable to analyse the insoluble material that remained following extraction of cartilage tissue and therefore could not quantify total differences in protein abundance, our method was especially useful in allowing us to identify subtle differences in the extraction of individual protein oligomers that may not have been detected using other methods. For example, two specific heterooligomeric forms of matrilin-1 were clearly absent from Matn3 V194D cartilage extractions and given our previous data confirming the intracellular co-retention of a proportion of wild type matrilin-1 with mutant matrilin-3 in the Matn3 V194D model Bell et al., 2012, we hypothesize that these matrilin-1 oligomers missing from Matn3 V194D cartilage extractions represent matrilin-1/-3 hetero-oligomers, which we propose are present within the heterogeneous disulphide bonded aggregates retained within the ER of Matn3 V194D chondrocytes.

We detected a form of type IX collagen in sequential extractions from Comp DelD469 and Comp T585M cartilage that contained the 1IX NC4 domain and which corresponded in size to the predicted molecular weight of the 1IX chain. This form of type IX collagen was not extractable using the same conditions from wild type or Matn3 V194D cartilage. Robust microarray studies and immunohistochemical analyses have previously demonstrated that there are comparable levels of type IX collagen gene expression Nundlall et al., 2010; Suleman et al., 2012 and protein between the cartilage of wild type and all three mutant mice Leighton et al., 2007; Pirg-Garcia et al., 2007; Suleman et al., 2012 confirming that we had indeed detected differences in extractability, rather than genotype specific differences in the relative amounts of type IX collagen. The results from LC-MS/MS analysis partially validated our finding of increased 1IX collagen extraction from Comp T585M and Comp DelD469 cartilage. For example, a significant increase in the quantity of 1IX collagen was detected in Comp T585M buffer 3 samples relative to wild type controls and a trend towards an increase in 1IX collagen extraction was observed in Comp DelD469 buffer 3 samples, but this did not reach statistical significance WT spectra2 vs DelD469 spectra5; P0.072; data not shown. However, in contrast to Western blot analysis which detected no differences in the extraction of type IX collagen from Comp DelD469 buffer 1 samples, LC-MS/MS revealed a small decrease from similar extractions WT spectra2 vs DelD469 spectra0; P0.019. The small number of matched spectra implies that this could potentially represent a false positive, and therefore demonstrates the importance of using complementary methods to gain an accurate understanding of the cartilage proteome.

Types II, IX and XI collagen are highly cross-linked in mouse cartilage by 3 weeks of age Mendler et al., 1989; therefore, the identification of extractable type IX collagen from Comp DelD469 and Comp T585M cartilage suggests that it may be less tightly integrated into the ECM of this abnormal cartilage. This would have an impact upon the integrity of the cartilage ECM and given the known function of type IX collagen in cartilage and its influence on cartilage degradation when knocked out Fssler et al., 1994, may ultimately contribute to the early onset OA seen in PSACH and MED patients.

We also identified differences in the molecular forms of matrilin-3 that were extracted from Comp T585M and Comp DelD469 cartilage by buffer 1. The matrilin-3 form that was absent from Comp DelD469 samples was identified in wild type following longer exposure, data not shown and Comp T585M buffer 1 extractions and we hypothesize that this oligomer may correspond to an N-terminally-cleaved form of tetrameric matrilin-3 Hills et al., 2007. This observation suggests a possible difference in matrilin-3 processing in Comp DelD469 mice, the pathological consequence of which is not known. Interestingly, matrilins are cleaved by ADAMT4 and -5 Ehlen et al., 2009 and it was postulated that the release of matrilin-3 from the matrix could in turn lead to further induction of pro-inflammatory cytokines and proteases in chondrocytes Klatt et al., 2009 which could influence the occurrence of OA at later stages of PSACH-MED.

The finding of an overall increase in the extractability of matrilin-3 and COMP by buffer 1 in all three PSACH-MED mouse models led us to consider that some of this protein could represent intracellular retained material. Analysis of protein lysates from isolated chondrocytes confirmed this hypothesis for the Matn3 V194D and Comp DelD469 models, but more importantly, it also confirmed that whilst both mutant proteins are retained, only mutant V194D matrilin-3 is retained as a high molecular weight covalently-linked aggregate Bell et al., 2012. This important observation demonstrates that the consequences of the individual mutations on protein misfolding and retention/accumulation differs between Matn3 V194D and Comp DelD469 and perhaps such differences begin to provide a molecular explanation for why, in contrast to Matn3 V194D, Comp DelD469 does not elicit a classical unfolded protein response Suleman et al., 2012.

We also used a proteomic method LC-MS/MS spectral counting to compare cartilage sequential protein extractions in a global non-biased manner. Our investigation made use of the same cartilage extraction protocol used for Western blot analysis and we detected a total of 957 unique proteins, which was comparable with previous murine cartilage proteomic investigations Belluoccio et al., 2006; Pecora et al., 2007. LC-MS/MS analysis of femoral cartilage extractions identified significant changes in the extraction profiles of many proteins including structural and non-structural ECM proteins, in addition to proteins with known roles in protein folding and trafficking.

Among the most consistent findings was a relative decrease in the amount of type XIV collagen extracted from Matn3 V194D and Comp DelD469 cartilage compared with control and Comp T585M samples. This decrease was evident through both spectral counting and Western blotting approaches. Type XIV collagen has recently been identified as a binding partner of COMP in vitroAgarwal et al., 2012, which suggests the possibility that a disruption to the ECM by Matn3 V194D and Comp DelD469 mutations could affect type XIV collagen stability in the ECM by altering direct binding between these proteins. This finding is especially important given the role of type XIV in collagen fibrillogenesis Ansorge et al., 2009, the known disruption to collagen fibrillogenesis caused by an MED-causing COMP mutation Hansen et al., 2011 and the previous finding of altered collagen fibril morphology in Matn3 V194D, Comp DelD469 and Comp T585M cartilages Leighton et al., 2007; Pirg-Garcia et al., 2007; Suleman et al., 2012 supplementary material Fig. S3.

Interestingly, and somewhat in contrast to type XIV collagen, there was decreased extraction of type XII collagen from Matn3 V194D and Comp T585M cartilages based on LC-MS/MS analysis of buffer 2 samples. This observation was partially validated by Western blotting, although the latter technique identified additional differences that had not been detected by LC-MS/MS. Recently, type XII collagen has been shown in vitro to bind COMP via its collagenous domains Agarwal et al., 2012. Taken together these findings suggest a close functional relationship between glycoproteins matrilin-3 and COMP and FACIT collagens Types IX, XII and XIV in the chondrocyte pericellular matrix and that disruptions to this network might be a key disease trigger in cartilage degradation. Indeed, electron microscopy of cartilage from the three mouse models has previously demonstrated changes in the morphology of the ECM Leighton et al., 2007; Pirg-Garcia et al., 2007; Suleman et al., 2012. In particular, the collagen fibrils were more clearly visible, suggesting that lower levels of fibril surface-associated proteins were decorating individual collagen fibrils supplementary material Fig. S3. The common and discrete disease signatures that we have identified in this study might therefore explain these differences in the cartilage phenotype of these genetic mouse models of PSACH-MED.

A further difference in the extraction of a collagen-interacting protein was observed in Comp DelD469 cartilage extractions, where there was a significant decrease in the detection of tenascin X compared with other genotypes. Tenascin X has only recently been detected in cartilage Wilson et al., 2012 and its knockout in mice has been shown to cause a reduction in the collagen content of skin Mao et al., 2002. Furthermore, a wide variety of gene deletion and truncating point mutations lead to tenascin X deficiency in skin and recessive EhlersDanlos syndrome Schalkwijk et al., 2001; a disease characterized by hyperextensible skin, hypermobile joints and tissue fragility. The knockout of tenascin X in mice, and its deletion in humans, also causes alterations in muscle function Huijing et al., 2010; Voermans et al., 2011 that induces histological features of a mild myopathy, which is also observed in Comp mutant mice Pirg et al., 2010. Our finding provides the first evidence of changes in the extractability of tenascin X from the cartilage of a chondrodysplasia disease model. If the reduced detection of tenascin X in cartilage were to extrapolate to other tissues where mutant COMP is expressed this could potentially have pathological consequences and contribute to the myotendonopathy associated with PSACH and MED Pirg and Briggs, 2010.

A notable and potentially related observation was the decreased detection of tenascin C in Matn3 V194D and Comp DelD469 cartilage extractions. Tenascin C is an extracellular glycoprotein involved in tissue injury/repair and is upregulated in osteoarthritic cartilage, where tenascin C protein fragments are believed to induce inflammatory mediators and ECM degradation Patel et al., 2011; Sofat et al., 2012. We found no evidence of tenascin C fragmentation in Matn3 V194D and Comp DelD469 samples since the examination of matched peptides did not reveal any genotype specific differences in sequence coverage data not shown. Interestingly, the decrease in tenascin C in Comp DelD469 samples was accompanied by a corresponding reduction in transforming growth factor beta induced protein ig-h3 Tgfbi supplementary material Table S3. Both of these proteins have recently been proposed to contribute to TGF1 signaling in early cartilage development Wilson et al., 2012 and alterations within TGF1 signaling have been linked to osteoarthritis van der Kraan et al., 2010, whilst COMP itself has recently been identified as a TGF-responsive gene Li et al., 2011. The detrimental effects of chondrodysplasia-causing mutations on the TGF1 signaling pathway could therefore yield pathological consequences for cartilage development and degeneration and potentially contribute to PSACH and MED disease mechanisms. Alternatively, a decrease in TGF1 signaling in Comp DelD469 cartilage might form part of a feedback mechanism to reduce the expression of mutant COMP.

The LC-MS/MS approach also validated the intracellular disease signature of Matn3 V194D chondrocytes by confirming increased protein levels of specific chaperones Armet, Creld2, Grp78, Grp94 and Calr and foldases PDIA 3, 4 and 6, that have been previously shown to be highly upregulated at the transcriptional level Nundlall et al., 2010. These transcriptional changes therefore support the proteomic analyses described in this investigation.

Finally, recent studies have shown that numerous myogenic genes including MYH4 are expressed in articular cartilage, and that changes in their expression constitutes a novel disease signature in the STR/Ort murine model of OA Poulet et al., 2012. Interestingly, increased levels of myosin heavy polypeptide 4 MYH4 were detected in Matn3 V194D, Comp T585M and Comp DelD469 cartilage buffer 2 compared with controls supplementary material Tables S1S3. These data therefore provide new insight into the cartilage degeneration that is associated with PSACH and MED and identifies areas for further investigation.

To conclude, we have completed a proteomic analysis of Matn3 V194D, Comp T585M and Comp DelD469 models of chondrodysplasia and demonstrated that the mutation of matrilin-3 or COMP can induce changes to the extractability of other cartilage proteins confirming that there are disruptions to the organization of the cartilage ECM. Whilst many of these changes in ECM organization are genotype specific, there is also likely to be a generalized disruption to cartilage integrity that ultimately increases susceptibility to osteoarthritis, which is common to both PSACH and MED. This semi-quantitative proteomic approach, in combination with gene expression studies, also provides a powerful tool to generate reference -omics profiles that will help establish disease signatures in chondrodysplasias that are also relevant to more common forms of joint degeneration such as osteoarthritis. Defining these diseases signatures is a prerequisite for the identification and validation of potential biomarkers and future therapeutic applications.

Matn3 V194D, Comp DelD469 and Comp T585M mice were generated as previously described Leighton et al., 2007; Pirg-Garcia et al., 2007; Suleman et al., 2012 and control mice were of an equivalent mixed genetic background 50 C57BL/6 and 50 129Sv to all mutant lines. Wild type and mutant mice were used to provide cartilage for sequential protein extractions and chondrocytes for Western blot analysis of intracellular proteins.

Cartilage from knee joints and femoral heads were dissected from the hind limbs of 3-week-old mice and frozen at 80C. Knee joints were dissected by removal of the patella tendon and other connective tissue surrounding the joint using a scalpel, followed by excision of all tissue proximal/distal to tibial/femoral epiphyseal growth plates, respectively. Femoral heads were dissected by careful removal from the acetabulum using a scalpel, followed by removal of all tissue distal to the femoral head. Cartilage samples were thawed and sequentially extracted in buffer 1 0.15M NaCl, 50mM Tris pH7.4, buffer 2 1M NaCl 10mM EDTA, 50mM Tris pH7.4 and buffer 3 4M GuHCl, 10mM EDTA, 50mM Tris pH7.4 as previously described Nicolae et al., 2007. All buffers contained 2mM phenylmethylsulfonyl fluoride and 2mM N-ethylmaleimide protease inhibitors. The supernatants from the extractions were ethanol precipitated, air dried and resuspended as previously described Nicolae et al., 2007. 20l aliquots of each extraction were prepared and frozen at 80C prior to analysis by SDS-PAGE. Costal chondrocytes were isolated as previously described Nundlall et al., 2010.

Aliquots from three biological replicates per genotype i.e. tissue extracts from 3 different mice of knee or femoral head cartilage extractions and rib chondrocyte preparations were thawed and separated on 413.5 and 412 SDS-polyacrylamide gels respectively. When protein samples were reduced, dithiothreitol DTT was added to a final concentration of 0.1M and samples were boiled at 90C for 5min prior to loading. Proteins were transferred to nitrocellulose membranes and uniform transfer was evaluated by ponceau staining supplementary material Fig. S4. The primary antibodies used were Matrilin-1 Hauser and Paulsson, 1994, -2 Piecha et al., 1999, -3 reduced samples, R&D; non-reduced samples Klatt et al., 2000, -4 Klatt et al., 2001, COMP reduced samples, Genetex; non-reduced samples DiCesare et al., 1994, type IX collagen NC4 domain Budde et al., 2005, type XII collagen Veit et al., 2006, type XIV collagen Ansorge et al., 2009, tenascin C Sigma, tenascin X and decorin Wiberg et al., 2003. Membranes were incubated with either peroxidase-conjugated or near-infrared fluorescently labeled secondary antibodies for 1h. Secondary antibodies used were peroxidase-conjugated swine anti-rabbit IgG Dako, donkey anti-mouse, anti-rabbit, anti-guinea pig and anti-goat IgGs Licor.

Those blots incubated with fluorescently labeled antibodies were imaged using the Odyssey system Licor. Peroxidase-conjugated antibodies were detected using 3-amin-opthalhydrazide 1.25mM, p-coumaric acid 225M and 0.01 H₂O₂ then exposed to photographic film.

20l aliquots of femoral head cartilage extracts were run on 412 SDS-polyacrylamide gels for 4mins at 200V. Total protein pools were excised from the gel before being dehydrated, reduced, alkylated and washed. Samples were then digested with trypsin overnight at 37C and analysed by LC-MS/MS using a NanoAcquity LC Waters, Manchester, UK coupled to a LTQ Velos Thermo Fisher Scientific, Waltham, MA mass spectrometer. Peptides were concentrated on a pre-column 20mm180m i.d, Waters. The peptides were then separated using a gradient from 99 A 0.1 formic acid in water and 1 B 0.1 formic acid in acetonitrile to 25 B, in 45min at 200nL min¹, using a 75mm250m i.d. 1.7M BEH C18, analytical column Waters. Peptides were selected for fragmentation automatically by data dependent analysis.

Data were interrogated using Mascot version 2.2 Matrix Science, UK against the UniProt database version 2011-05 with taxonomy of Mus musculus and the following search parameters selected: fragment tolerance: 0.6Da; parent tolerance: 0.5Da; fixed modifications allowed: 57 on C carbamidomethyl, 16 on M oxidation; max missed cleavages: 1. Mascot search results were validated using Scaffold version 3.3.1 Proteome Software, Portland, USA to assign confidence values to peptide/protein matches, where Peptide/Protein Prophet algorithm confidence values of 0.7 and 0.99 were used respectively. Identified proteins were defined as having a number of matched peptide spectra 2, and the unweighted spectral count was used as a measure of quantification. These parameters constrained the protein false discovery rate FDR to 0.2 in all analyses. Three biological replicates were used in all experiments except for the analysis of buffer 3 wild type samples, where two biological replicates were used as a result of sample failure in one replicate. The number of spectra identified for each protein were compared between wild type and mutant genotypes using the beta-binomial test in R version 2.14.2, BetaBinomial package Pham et al., 2010. A P-value<0.05 was considered significant. Lists containing the Uniprot IDs of proteins that were significantly different in quantity between wild type/mutant samples were analysed using The Database for Annotation, Visualization and Integrated Discovery DAVID bioinformatics resource 6.7 http://david.abcc.ncifcrf.govDennis et al., 2003.

We gratefully acknowledge the support of the Wellcome Trust M.D.B. is the recipient of a Wellcome Trust Senior Research Fellowship in Basic Biomedical Science; Grant 084353/Z/07/Z, the BBSRC P.A.B. was the recipient of a BBSRC-Case studentship with industrial support generously provided by Certus Technology Associates Limited and Deutsche Forschungsgemeinschaft DFG: ZA56/2-1 to F.Z. and WA 1338/2-6 to R.W.. The research was undertaken in the Wellcome Trust Centre for CellMatrix Research Grant 088785/Z/09/Z and the Biomolecular Analysis Facility of the Faculty of Life Sciences at the University of Manchester.