This is the first study to demonstrate serum protein signatures associated with osteoarthritis. Using a case-control design nested in a large longitudinal study of normative aging, we detected 16 proteins that were different in participants with radiographic knee and hand OA compared to age-, gender-, and BMI-matched controls free of OA. Four of these proteins (MMP-7, IL-15, PAI-1 and sVAP-1) associated with OA presence at the time of the classifying x-ray were also differentially expressed between cases and controls, at a time when OA was not detectable by conventional x-ray. Six additional proteins were different between OA cases and controls at a time when radiographic evidence of OA was not present, but not ten years later when radiographic features had developed. We have also clearly demonstrated that these signatures more accurately distinguished cases from controls at both time points than all proteins assayed, and that there is overlap and consistency in the protein signatures associated with OA at the two points in time.
It is plausible that the protein signatures identified in association with OA are serum markers of disease initiating events (initial OA signature) or disease sustaining events (prevalent OA signature). Accordingly, our data suggest that altered extracellular matrix metabolism plays a central role in OA initiation. The pattern includes a shift away from constitutively expressed MMP-2 towards higher MMP-7 expression and propensity for enhanced plasmin activation to render extracellular matrix vulnerable to degeneration and injury is consistent with previous reports of MMP-7 over expression in human OA chondrocytes(22
) and synovial fluid samples(23
), and may also explain the lower PAI-1 concentrations observed in OA samples that has not previously been reported. The observations of higher IL-15, vascular adhesion protein-1 (VAP-1) and adhesion molecule concentrations in OA samples, although not previously reported, suggest that cellular and immune mechanisms contribute to OA initiation. The sustained profile comprised of high MMP-7, IL-15, VAP and low PAI-1 associated with prevalent OA suggest that these same mechanisms also contribute to OA perpetuation. However, in contrast to the earlier time point, the protein signature associated with prevalent OA suggests that reparative mechanisms are operative, as evidenced by over-expressed TIMPs and growth factors (FGF, IGF binding protein, GM-CSF, NT-4). The increased expression of inflammatory cytokines and chemokines associated with prevalent OA lend additional support of the notion of OA as an inflammatory disease.
The protein signatures associated with OA in this study are not specific to cartilage, and therefore understandably distinct from those identified in microarray studies of cartilage(24
). These proteins cannot be presumed to cartilage-specific events although collagen turnover products are also associated with OA in our study. Interestingly, some of the proteins identified that were associated with older age (e.g. IL-6) are also associated with frailty and sarcopenia. With rare exceptions, the proteins associated with OA were distinct from the expression pattern associated with and predictive of OA development.
Although these results advance our understanding of the molecular markers of OA, there are several limitations inherent in the data. First, this study is based upon microarray data and hence vulnerable to the criticisms implicit to this technology. The simultaneous quantification of 169 proteins might be regarded as “fishing.” However the specific proteins selected a priori for this chip include mediators of inflammation, cell growth, activation and metabolism that are relevant to OA pathogenesis. Additionally, a number of quality control features such as target specificity and use of a 4-fold change as the threshold for inclusion were implemented to reduce the chance of false positive results. Despite the considerable improvement, the sensitivity of this RCA-based assay remains inferior to conventional ELISA assays. We also acknowledge the possibility that the length of storage prior to testing may have compromised sample integrity. Therefore, we cannot exclude that some proteins important to the pathogenesis of OA but with levels below the RCA limits of detection could not be identified. Second, the study sample is small due to our adherence to very strict inclusion and exclusion criteria with careful matching of controls to cases on the most likely confounders of age group, sex and BMI. While this approach increased the precision of our measures and reduced the chances of outcome misclassification, the presence of a few outliers may have influenced the analysis results. Due to the small sample size, we were not able to examine signatures according to specific OA sub-type or clinical pattern. Given the selection criteria for this study, the OA-associated protein signatures identified in this study are most relevant to OA of the knee that and might be less relevant to OA of other locations, or that develops secondary to other processes. Third, this study relied on radiographic definitions of OA to classify individuals as cases or controls. Given the limited sensitivity of conventional x-rays for early changes of OA, it remains possible that participants classified as “normal” may have been affected by early OA that escaped radiographic detection.
Despite the above limitations, this study offers new insight that can be best gleaned through a prospective, long lasting, longitudinal study such as the BLSA. We conclude that osteoarthritis of the knee and hand is associated with serum protein signatures that include mediators of cellular activation, inflammation and matrix degradation, and that some of these proteins were predictive of OA development years prior to radiographic detection. Additionally, except for four proteins, the set of proteins associated with OA at the time of the initial x-ray was not different at the time of the second x-ray and vice-versa. These findings support the notion that these protein signatures are responsive to change over time - and that initiating events and their mediators are distinct from those that sustain the disease. However, additional studies are necessary to establish the utility of these protein profiles for the diagnosis and monitoring of OA. The protein signatures identified in this study may have utility as targets of future intervention studies and/or as prognostic indicators or predictors of therapeutic response.