Serum and synovial fluid cytokine profiling in hip osteoarthritis: distinct from knee osteoarthritis and correlated with pain

Approximately 1 in 8 individuals are afflicted with Osteoarthritis (OA) and although it is more common in older populations, it is becoming a serious health and economical concern in young, active individuals [1, 2]. Therefore, it is essential to further understand the factors involved in the onset and progression of the disease, so that more efficient diagnostics and treatments can be developed.

While much of the research focus in OA has been directed towards the knee, it is necessary to examine if the pathways and mechanisms in knee OA are conserved in other joints. For example, in the hip, recent literature have provided insight into mechanical causes of hip OA including impingement and dysplasia [3], while others have focused on potential genetic predispositions with or without mechanical risk factors [4]. This suggests that, as in knee OA, there could be potentially many avenues of hip OA onset that eventually result in patients with diverse etiology converging later in the disease trajectory. Our previous work in knee OA looking at systemic and local inflammatory profiles suggests that inflammatory profiles are distinct in patients with knee OA from those without OA [5] and therefore that this approach may be able to discriminate systemic inflammatory differences between patients with knee vs. hip OA.

More generally, there have been a number of studies and many biochemical markers that have been identified in knee OA [6‐10]. Historically groups have focused on cartilage metabolism markers including which demonstrate changes in concentration and fragment species with the onset and progression of OA [11‐14]. Inflammation is known to be present in OA and the changes in individual inflammatory markers have been correlated with both severity and progression and OA [15, 16]. Regardless of the type of biochemical marker examined however, knee OA and hip OA are almost always grouped together or only one joint is examined within a given study [17]. Most of the common OA biochemical markers (collagen fragments, cartilage oligomeric matrix protein (COMP), etc.) have been tested on both knee and hip cohorts, but only a very few studies looked at both joints and run comparisons [17]. The uncontrolled variances (such as different definition of OA, cohort characteristic, experimental method) across studies make the results almost incomparable. Among the studies that contain both cohorts, the comparisons of biochemical markers between knees and hips are largely ignored [18, 19]. It is well known that mechanical properties differ between joints (knee vs. hip) [20, 21], but as far as we are aware only two studies have directly compared between cohorts of hip and knee OA patients: the effectiveness of bone metabolism for OA prognostic [22] and the difference of association between COMP and OA symptoms [23]. However, neither of them compared the absolute concentration of biochemical markers between two phenotypes.

In addition to biochemical markers, it is also important to gain a better understanding of pain in OA. Pain is the leading disabling symptom of OA [24]. According to the American College of Rheumatology (ACR) criteria for OA, pain is a necessary condition for clinical diagnosis of OA, with the support by other sufficient conditions such as age, radiographic evidences or physical examinations. OA pain is hard to quantify from patient to patient based on a number of confounding factors [25] and its inconsistency with other symptoms such as radiographic findings is not yet well understood. Recent studies have greatly improved our understanding of pain at the molecular level [26], and new biochemical markers have also been reported to be correlated with severity of OA pain in the past decade (Additional file 1: Table S1). Cytokines play an important role in the pain signal pathway from lesions to higher brain processing centers [27]. While cytokines can change the sensitivity of peripheral receptors to nociceptive input by a variety of mechanisms in experimental animal models [28‐30], very few human studies have reported correlations between cytokines and OA pain. The chemokine sub-family of cytokines is also of specific interest in patients with arthritis since these small molecules are capable of recruiting additional immune and/or stem/progenitor cells to sites of injury and/or increased inflammation.

With our incomplete understanding of the role of any specific cytokine in the inflammatory process, understanding the ‘profile’ of inflammation in a given patient population is an essential starting point. Therefore, the primary aim of this study was to examine system and local inflammatory profiles in hip OA patients, using knee OA and non-OA control as comparator groups to identify if differences exist between hip and knee OA. Furthermore in the hip OA cohort, it was attempted to determine if there were any cytokine (s) that correlated with pain symptoms, as well as other clinical measurements.

To our knowledge, few papers have discriminated which joint the OA is affecting when comparing OA with control groups. In this study, we have found that 4 cytokines (EGF, FGF2, MCP3, IL8) are expressed differently in knee OA patients versus hip OA patients after controlling for age, gender and K/L grade. Interestingly, when we compared these two OA cohorts with the control cohort, the picture became even more complex. EGF levels were similar between the hip OA and control cohorts but both differed from the knee OA cohort. MCP3 was similar between the knee OA and control cohorts but differed from the hip OA cohort. The scatter plot of first 3 components of PCA of the 3 cohorts (hip OA, knee OA and control) showed that, neither Hip OA nor Knee OA is “closer” to the control. The center of gravity of the 3 cohorts represents a triangle shape (Fig. 1). This infers that the three cohorts have their own characteristic cytokine profiles (though knee OA and hip OA cohorts are more similar to each other than either is to the control cohort). This may suggest that, at least for serum cytokines, OA may have distinct inflammatory profiles with subtypes that vary according to the affected joint. Failing to treat them as separate groups may be one of the reasons that the accuracy of biochemical markers in OA varies among studies [37]. This may partially explain why none of the biochemical markers that have been studied to date have been approved for diagnosis or prognosis of OA. Another possible reason might be the heterogenetic and multifactorial nature of OA itself. The existence of subgroups in OA might invalidate any statistical single biochemical marker comparison which considers OA as one group-one biochemical marker is not enough to classify all of these subgroups. Previous studies have looked into subgrouping OA before statistical analyses [38]. For instance, by using serum metabolomics data of OA patients, Zhang et al. identified at least 3 distinct subgroups which were not associated with any known confounders including age, sex, BMI and comorbidities [39]. The current study suggests that hip and knee OA cohorts may not be as similar in terms of the disease as previously thought and this has implications that OA in other joints (hands, spine, etc.) may also cloud diagnostic efforts, however, this hypothesis will require further study and validation.

In specific regards to the inflammatory cytokines examined in the current study, it is believed that aging brings about changes in the expression of many cytokines. Aged-related changes in the immune system may elevate the concentrations of cytokines which some have postulated may ultimately lead to chronic inflammation [40]. This being said, in our dataset none of the cytokines are statistically significantly correlated with age. This might be due to the fact that most patients have passed their middle-age and the age distribution of these two cohorts are relatively concentrated (mean = 59.9, SD = 10.1). Although the control cohort is significant younger than the two OA cohorts (mean = 40.0, SD = 9.5). A limitation of the current study is that we unfortunately do not have the linked age information for each patient sample in the control cohort (only the cohort level details) and therefore we are not able to run a correlation analysis. Additionally, many factors can confound cytokine concentrations, such as statistical errors and mismatched cohort characteristics (such as races, disease history, diet and etc.). While, it can be unrealistic to find two exactly matched cohorts, these mismatch factors may lead to changes in serum cytokine concentrations in an unpredictable way and highly controlled cohort studies will be required to address these discrepancies.

TNFα, IL1, IL6 and IL17 have been reported to be correlated with development of neuropathic pain in various animal models [27, 41]; however, except for IL6, which had weak correlation with PCA factor 2 of 10 HOOS pain questions, none of these proteins were significantly correlated with pain level scores in our dataset. We have tested the correlation between our two different pain level scores-SF36 body pain and HOOS pain sub-score, they are significantly correlated (correlation = 0.569, p-value < 0.0001). This suggests the method that we chose to evaluate osteoarthritis pain levels was reliable between questionnaires. Among all 36 cytokines, we found that MDC and IP10 were correlated with pain in the hip OA cohort. IP10 (CXCL10) belongs to CXC chemokine family. It is expressed by a variety of cell types. When IP10 binds to CXCR3, this can result in activation and recruitment of leukocytes, and also regulate cell growth and apoptosis [42]. In specific regard to pain, IP10 has been reported to be involved in breast cancer-induced bone pain by activation of microglia in rat models [43]. In our study, IP10 was positively correlated with pain in patients with hip OA. It is possible that IP10 could be inducing OA pain through a similar pathway by activation of microglia in the joint environment. MDC (CCL22) belongs to the CC chemokine family and it is expressed mainly by macrophages and dendritic cells. MDC recruits Th2, Th17 and regulatory T cells through the binding of trafficking receptor CCR4 [44, 45]. To our knowledge, no publication has directly looked at MDC and pain.

The concentrations of these cytokines in synovial fluid were consistent with the immuno-histochemistry results; with MDC and IP10 higher in synovial fluid than in serum, and were higher in hip OA than knee OA or control cohorts. This strongly suggests that within the hip OA cohort, MDC and IP10 are being produced within the joint environment and then may diffuse into serum through the synovial fluid. In the knee OA cohort however, MDC and IP10 were not present within the joint (synovium or synovial fluid) and therefore any detectable levels of these in the serum are most likely being generated in other tissues in the body. While, the concentration of IL6 was higher in hip and knee OA cohorts than controls, IL6 was not found in hip synovium. This suggests that there might be other tissues in the joint (cartilage, ligaments, etc) that produce IL6 in the hip OA cohort. The relationship between serum and synovial fluid cytokine levels are still poorly understood with previous studies demonstrating little correlation [46] or high correlation [47] between systemic vs. local levels. However, the biological basis behind these correlations (or lack thereof) remain unknown, and further study will be required on specific markers to determine if the choice of sample local will impact the usefulness of the result obtained.

These findings validate previous studies that cytokines are differently expressed in OA patients [15]; and suggest that while cytokine profiles are generally similar between hip and knee OA, that there are specific differences that may be related to differential disease processes within a given joint. Furthermore, 3 cytokines (IL6, MDC and IP10) were identified that correlated with hip OA. These results set the ground work for future studies that will be required to understand: if cytokines play distinct roles in in the onset/pathogenesis of OA in different joints; and what is the pathway and/or mechanism by which the identified cytokines regulate OA pain.