Proteomics is increasingly applied in basic and clinical rheumatology and orthopaedics to help understand the biology of joint tissues, responses to pro-inflammatory cytokines and growth factors as well as mechanical signals [23,27,29,30]
. Proteomic techniques also have the potential to identify novel biomarkers of joint diseases. In this study we set out to refine a serum-free explant model of articular cartilage inflammation and tested the feasibility of using it for studying the major proteins present in the secretome. The explant model was originally established in our laboratory with the intention of using it as a screening system for testing the effects of putative anti-inflammatory compounds from plant extracts 
as well as identifying biomarkers of early OA. High throughput proteomics and cross species peptide matching were then used to identify the main proteins present in the secretome. This approach has a key advantage: it does not require gel electrophoresis and is more rapid than traditional 2-D gel MALDI-TOF MS analysis.
Using comparative proteomic analysis, we identified several constitutively expressed ECM molecules in the supernatants of cartilage explants, namely aggrecan, fibromodulin, decorin, biglycan and COMP. These proteins have been identified in proteomic studies in human 
and mouse 
cartilage, thus facilitating the validation of our equine cartilage model. One of the most abundant proteins in the explant media was COMP, a non-collagenous matrix protein that mediates chondrocyte attachment to the ECM 
and organises ECM assembly 
. COMP has been suggested as a potential biomarker for OA and it has been detected in synovial fluid, serum and urine from human and equine OA patients [17,40–44]
. Moreover, its presence correlates with disease severity and progression. The presence of COMP in this model, concurs with that of other in vitro
studies that have found COMP abundant in media exposed to cartilage explants [24,25,28]
Western blotting confirmed that IL-1β-treatment increases the abundance of MMP-3 in the cartilage secretome. MMP-3 is a proteolytic enzyme that degrades ECM components such as fibronectin, collagens and cartilage proteoglycans. Its levels are increased in clinical OA cases and the gene encoding this protein has been shown to be up-regulated in IL-1β-stimulated chondrocytes in vitro
Immunohistochemical studies have reported MMP-3 in the synovium, cartilage and medium of cultured cartilage explants from OA patients 
. Our qualitative LC–MS/MS data suggested that MMP-3 was higher in IL-1β-treated samples compared to controls and this observation was confirmed by western blotting. NSAID-treatment actually reduced MMP-3 levels, an observation that has been found previously for bovine MMP-3 mRNA and protein 
. The weak band(s) in the tissue homogenate controls showed that MMP-3 was present in cartilage, but in lower concentrations than in the IL-1β-treated samples and thus may have been below the detection threshold used to analyse the samples.
Western blotting for MMP-1 detected a band of roughly 102 kDa in both the cartilage control and all the secretome samples (present as a double band in IL-1β-treated samples corresponding to active and inactive forms), which was nearly double the predicted molecular weight (54 kDa). Although the samples were run under reducing conditions, it may be that the observed band corresponds to a dimer form of MMP-1 that was resistant to the reducing agent.
The equine sequence for TSP-1 was not present in the Swiss-Prot database at the time of this study but cross species matching with bovine, human and mouse confirmed the presence of TSP-1. The fragments matched the protein motifs of TSP-1 from multiple other species, thus increasing our confidence in the detection of this protein in the horse cartilage secretome. The presence of TSP-1 was also confirmed by western blotting using two different commercial antibodies. TSP-1 is an adhesive glycoprotein that mediates cell-to-cell and cell-to-matrix interactions [48,49]
. Human cartilage studies have shown that TSP-1 is increased in early OA cartilage lesions 
and both the proteomic and western blot results from our study, suggest that this is reflected in our cartilage explant model of early OA.
The abundance of cartilage ECM-associated proteins in the top scoring entries highlights the specificity of this in vitro
model and the suitability of high throughput protein discovery techniques for interrogating it. However, despite using a targeted approach and an explant culture system, a small number of non-matrix associated proteins were also identified in the media of several samples, for example, vimentin, dynein heavy chain, alpha enolase and the membrane bound sodium hydrogen exchanger 2 (NHE2) (see ). There are a number of possible explanations and interpretations for the presence of these proteins. The most obvious explanation for the presence of these non-ECM proteins is the high sensitivity of the high-throughput approach for detecting proteins released from the cytoplasm following cell death; the detection of these proteins may be attributable to a small number of dead or dying chondrocytes. However, it is known that there are frequently observed proteins in many proteomic experiments regardless of species or tissue being studied, such as alpha enolase and vimentin [51,52]
In addition to the cartilage ECM and chondrocyte-associated proteins, the database search also detected some apparently unlikely candidates for biomarkers of arthritis, for example Bullous pemphigoid antigen or breast cancer type 2 protein susceptibility homologue. However, it is highly likely that some of these less abundant proteins with lower Mowse scores could be proteins of interest in their own right or modifications of the abundant proteins, which are relevant to OA. Ideal biomarkers for OA are those that can indicate the very early stages of disease before too much destruction has occurred 
. Therefore the best markers of early OA are unlikely to be fragments or degraded forms of major ECM proteins such as fibrillar collagens 
. Notably, we did not identify any fibrillar collagens in this study. This adds weight to the argument that incubating cartilage explants with IL-1β creates an environment that could mimic the early stages of OA. Although we used a Mowse score of 40 as the cut-off point, it is possible that proteins under the detection threshold could be potentially useful biomarkers. For example, the protein arginine deiminase type-4 (PADI4), which had a Mowse score of 39 (see ), is known to catalyse citrullination of arginine residues of proteins. Citrullinated peptides can act as autoantigens thereby increasing the risk of rheumatoid arthritis (RA) 
, as such, PADI4 gene expression has been detected in the synovial tissue from RA patients 
. In addition, autoantibodies against PADI4 are associated with increased RA severity 
. Although RA is an auto-immune disease, unlike OA, the protein may have other functions and potential involvement in OA. The presence of PADI4 in the supernatant of the cartilage explants in this study may partly explain why this enzyme could be a target of autoantibodies to citrullinated proteins, in fact this antigen may not be as specific for rheumatoid arthritis (RA) as previously suggested [58,59]
, the enzyme may also be released in OA. PADI4 is one of the many less abundant proteins identified in the model that may be potential biomarkers. More studies with larger sample numbers may provide more information in this area. However, greater potential lies in using methods that facilitate mining below the abundant proteins in the secretome to allow robust and clear detection of the lower abundance components. The most obvious of these methods would be the depletion of the major/high abundance proteins, which is commonly used when studying biological fluids such as plasma and cerebrospinal fluid [60,61]
. Another approach would be the use of metabolic labelling during culture to allow the newly synthesised proteins to be studied 
. Adopting these approaches will aid identification of these less abundant but potentially more interesting proteins in terms of the molecular changes accompanying early OA.
Proteomics has been used to study soluble extracts of cartilage [24,63]
, chondrocyte whole cell lysates 
, synovial fluid 
and synovial fibroblasts 
, from both normal and OA patients. More recently proteomics has been used to characterise the extracellular space in tracheal in the human aorta 
. The cartilage explant model employed in our study can reduce some of the confounding variables associated with clinical samples such as individual variation, differing disease aetiology and the stage of disease progression. In effect, using the cartilage explant system provides a more homogenous model with appropriate controls. In addition, explants can be cultured in serum-free media, which address the technical issue of contaminating serum albumins and other high abundance proteins, which can prevent the identification of the lower abundance proteins.
The advantage of a proteomic approach over a genomic approach to studying potential disease indicators are that gene expression levels do not necessarily correlate with protein levels 
. For example, CLU expression is increased at the mRNA level in early OA cartilage and decreased in advanced OA 
. IL-1α-stimulated cartilage explants have been shown to produce decreased levels of CLU compared to untreated cartilage 
. This could reflect the differences between clinical cases and in vitro
models of OA and the various stimuli used to initiate cartilage degradation. Clearly, further work is needed to look at levels of these proteins in the synovial fluid and serum of clinical cases.