Synovitis in osteoarthritis: current understanding with therapeutic implications

The synovium is a specialized connective tissue that lines diarthrodial joints, surrounds tendons and forms the lining of bursae and fat pads. In synovial joints, the synovium seals the synovial cavity and fluid from surrounding tissues. The synovium is responsible for the maintenance of synovial fluid volume and composition, mainly by producing lubricin and hyaluronic acid. Through the synovial fluid, the synovium also aids in chondrocyte nutrition (together with subchondral bone), as articular cartilage has no intrinsic vascular or lymphatic supply [6].

The normal synovium has two layers. The outer layer, or subintima, is up to 5 mm thick and consists of multiple types of connective tissues: fibrous (dense collagenous type), adipose (found mainly in fat pads) or areolar (loose collagenous type). This layer is rich in type I collagen and microvascular blood supply, accompanied by lymphatic vessels and nerve fibres, but is relatively acellular [7]. The inner layer, or intima, lies next to the joint cavity and consists of a layer of 1–4 cells, only 20–40 μm thick. These synoviocytes have been identified by immunohistochemical and cytochemical methods as macrophages and fibroblasts; the latter is the dominant cell population in healthy synovium [7].

The histological pattern of synovium in OA patients is characterized by synovial lining hyperplasia, sublining fibrosis and stromal vascularization [8]. There is an abundant influx of leukocytes from the vascular compartment in response to cytokines and cell adhesion molecules [6], and several studies have shown macrophages and T-cell lymphocytes to be the most predominant immune cells in OA synovium, whereas mast cells, B cells and plasma cells are also found but to a lesser extent [9‐11].

Macrophage infiltration in the synovium is common in both OA and RA [8]. These macrophages can cluster and form multinucleated giant cells (MGCs) for improved phagocytosis, and are increased in similar numbers in inflamed OA and RA synovia compared with non-inflamed OA and post-mortem controls [8]. However, OA and RA demonstrate slightly different subgroups of MGCs, suggesting different drivers for these clusters, such as more cartilage debris in OA.

Much of the innate immune activation and cytokine production in the OA joint is attributed to synovial macrophages, but other cells including synoviocytes and chondrocytes also play a role [12]. The underlying mechanisms are complex and beyond the scope of this review. In short, molecules from degraded hyaline cartilage released into the synovial cavity are likely to initiate synovial inflammation in OA (Fig. 1). Early in knee OA, damage to the meniscus may also release tissue debris, although molecules released from subchondral bone may also play a role. Synoviocytes react by producing pro-inflammatory mediators, which in turn attract immune cells, increase angiogenesis and induce a phenotypic shift in chondrocytes [13]. A vicious cycle follows, as chondrocytes produce additional cytokines and proteolytic enzymes that eventually increase cartilage degradation and induce further synovial inflammation [4].

The processes driving inflammation in OA are complex. Given that OA is age related, immunosenescence may play a role in the immune response to tissue damage. A recent report analysed immune cell composition of the blood of OA patients and found compromised immune function of T cells and B cells beyond what appeared directly related to ageing, and this could reflect both inflammation and autoreactivity [14]. Also, trauma can trigger release of local inflammatory mediators, and there is increasing evidence that metabolic syndrome and obesity increase systemic low-grade inflammatory mediators which may synergize with other inflammatory mechanisms in OA [15].

Wojdasiewicz et al. [16] recently described in detail the mediating cytokines and their signalling pathways that are up-regulated in OA and most often have catabolic (i.e. degrading) effects, including interleukin-1 beta (IL-1β), tumour necrosis factor (TNF) alpha, IL-6, IL-15, IL-17 and IL-18. IL-1β and TNF-α have been the most extensively studied cytokines. They are elevated in OA synovial fluid, synovial membrane, cartilage and subchondral bone and, in the course of OA, have synergistic effects on signalling pathways that increase inflammation and cartilage degradation [16]. Despite these known actions, anti-TNF treatment has not provided symptomatic benefits in OA randomized trials, although it has reduced structural progression in individual joints with high degrees of inflammation, whereas IL-1β inhibition has so far been disappointing in terms of symptom and structural benefits.

Another group of ‘haematopoietic’ cytokines, or colony-stimulating factors (CSFs), is known to activate and mature myeloid cells both systemically and locally [17], and there is an increased interest in targeting these CSFs in inflammatory and autoimmune diseases. During an inflammatory response, they act more distinctly and restricted on cell receptors than other cytokines, causing increased cell numbers of selected myeloid populations and enhancing their survival or modifying tissue destinations. Preliminary data show beneficial effects of anti-CSF therapies in OA models, and a phase II trial in inflammatory hand OA has been initiated [17].

Some cytokines (such as IL-4, IL-10 and IL-13) have anti-inflammatory and sometimes anabolic effects, and may modulate an inflammatory response and slow progression of OA [16]. In a healthy joint, the balance between anabolic and catabolic cytokines contributes to stable turnover of cartilage, whereas an imbalance is found in OA. Although the literature is limited, there is emerging evidence that an up-regulation of these cytokines, either individually or combined, may induce cartilage repair in OA. A novel fusion protein of IL-4 and IL-10 has induced both structural repair and reduced inflammation ex vivo, and inhibited pain in animal models of OA [18].

OA synovial tissue typically displays a mild to moderate degree of inflammation on standard histological staining [20]. In smaller histological studies, the prevalence of inflamed synovium ranges greatly from approximately half to nearly all tissue samples depending on patient pre-selection and OA severity [11]. When comparing microscopic synovial changes in early and late OA, the literature has shown conflicting results [11], probably due to different definitions of ‘early’ and ‘late’ OA.

At the ‘macroscopic’ level, MRI provides valuable insights into synovitis and can visualize synovial hypertrophy, synovial fluid volume and level of synovial enhancement after intravenous injection of a contrast agent (Fig. 2). In terms of validity, MRI inflammation measures correlate well with histological inflammation and with effusion volume on arthrocentesis [21, 22]. In over a thousand knee OA patients, non-contrast-enhanced (non-CE)-MRI synovitis was present in 60% and effusion in 66% [23]. On CE-MRI, synovitis has been strongly correlated with radiographic OA severity and was reported in 74% of 404 patients having all grades of knee OA [24], and in 95% of 125 patients with mainly moderate to severe radiographic OA [25]. With respect to meniscal damage, only severe lesions have shown borderline association with synovitis [24]. MRI scans of hand OA patients have demonstrated synovitis in a median (interquartile range) 6 (4–7) interphalangeal joints of the dominant hand, with moderate to severe synovitis (grade 2–3) being infrequent [26].

Although ultrasound cannot assess cartilage deep in the joint or intra-bone lesions, it is well suited to assess inflammatory changes. A large meta-analysis of ultrasound-detected synovial changes found that people with knee OA or knee pain had high prevalence of effusion, synovial hypertrophy and power Doppler signals (ranging from 33 to 52% of the knees), and that these measures correlated well with histological findings [27]. A large multicentre study using ultrasound in 600 people with symptomatic knee OA demonstrated synovial inflammation or effusion in 46% of the population, despite using a strict definition of synovial hypertrophy [28]. In two large hand OA cohorts, nearly all patients (94% and 96%) had evidence of sonographic synovitis by grey scale in at least one finger joint [29, 30] and by power Doppler signals in 42% of the patients [29]. At the joint level, 28.7% of 1078 finger joints with Kellgren & Lawrence grade ≥ 2 demonstrated synovial effusion and/or hypertrophy by ultrasound [29].

With MRI and US studies demonstrating a high frequency of synovitis, it remains unclear whether inflammatory OA is a distinct subtype or just part of the spectrum of OA pathology severity. The group with more severe inflammation can be identified with imaging and serum biomarkers (such as the matrix metalloproteinase-dependent degradation of C-reactive protein (CRPM), connective tissue type I collagen turnover (C1M) and matrix metalloproteinase 3 (MMP-3)) [15]. To add further complexity, the expression of a given phenotype also depends on person-specific factors such as local muscle strength and obesity. A study followed two selected groups of knee OA patients for 4 years: one with metabolic syndrome and one lean group with frequent physical activity [31]. These groups differed significantly in the MRI expression of joint pathology, in which Hoffa synovitis was more prevalent in the active lean group, and prepatellar bursa signal, osteophyte scores and cartilage damage scores were higher and more prevalent in the metabolic syndrome group [31]. Inflammatory OA is a term often applied to a subset of hand OA with a greater degree of synovial inflammation and associated with radiographic erosive disease [32].

A systematic review from 2011 found an increased risk of MRI-detected synovitis and effusion in patients with symptomatic knee OA [35], but none of the included papers employed CE-MRI, without which effusion and synovitis cannot be readily differentiated. There are, however, emerging data for an even stronger association with pain severity when contrast enhancement is used [36].

Several cross-sectional studies on hand OA have shown similar associations between sonographic or MRI-detected synovitis and concurrent pain. A hand OA cohort applying ultrasound found inflammatory features (i.e. grey-scale synovial effusion or hypertrophy and power Doppler signals) to be significantly and dose-dependently associated with pain upon palpation and with the Australian/Canadian Osteoarthritis Hand Index (AUSCAN) pain and stiffness subscales and the Short Form (SF)-36 [30]. These results were replicated with MRI in another cohort [26], suggesting that synovial inflammation can cause pain also in hand OA.

The effects of synovitis on longitudinal measurements of pain are more complex to measure and results are ambiguous. Most studies have been performed on knee OA cohorts with 2–5 years of follow-up, often assessing VAS pain, WOMAC, KOOS or ICOAP. Some studies find an improvement in pain when some (but not all) aspects of synovial inflammation diminish [37] and a worsening of pain with increased inflammation over time [38], whereas others do not replicate these associations [39]. In hand OA patients, MRI-detected synovitis has been found to predict incident joint tenderness but not AUSCAN pain [40].

The effect of synovitis on OA is confounded by concurrent pathologies, raising an important question: does synovitis have an independent effect on OA progression? Felson et al. [41] recently examined the risk for incident radiographic knee OA after adjusting for structural pathology known to cause synovitis. The authors compared 239 cases and 731 control knees in the MOST cohort and found that cartilage lesions, meniscal damage, synovitis and BMLs were all risk factors for OA. Furthermore, when adjusting for confounding pathologies, synovitis remained associated with incident radiographic OA when the total synovitis score was 3 or higher on a 0–9 scale (OR 1.6, 95% CI 1.2–2.1). Another paper from the same cohort showed that knees without OA (i.e. having neither MRI-defined cartilage damage nor tibiofemoral radiographic OA) had a significant increased risk of cartilage loss whenever effusion synovitis was present (OR 2.7, 95% CI 1.4–5.1), independent of confounders for inflammation and cartilage loss [42]. These results suggest that synovitis effusion assessed on non-CE-MRI is an independent predictor of incident radiographic OA.

A series of recent papers from the OA Initiative (OAI) also support the importance of synovitis prior to incident radiographic OA. Atukorala et al. [43] performed nested case–control analyses on 133 knee joints that developed radiographic knee OA and an equal number of joints that did not develop knee OA. The authors found effusion synovitis and Hoffa synovitis 1 year prior to diagnosis to be significantly associated with subsequent OA development (OR 3.23, 95% CI 1.72–6.06 and OR 2.40, 1.43–4.04). A similar design was applied by Roemer et al. [44], who looked at repeated MRI scans up to 4 years prior to evident radiographic knee OA. They found that presence of Hoffa synovitis, effusion synovitis, medial BMLs and medial meniscal damage increased the risk of OA 2 years prior to incident radiographic OA, and that the number of features present increased the risk more than the presence of any single feature. More studies examining the incidence of the earliest detected MRI changes of OA will be required to understand more definitely the role of synovitis in the early stages of OA.

There is strong evidence that synovitis is associated with further worsening of OA structure. Longitudinal analyses of 531 knee OA patients demonstrated that ultrasound-detected effusion predicted pain, radiographic progression and also joint replacement [45]. There are also cohorts reporting higher risk of radiographic progression and cartilage deterioration over 2 years in knee joints with increased MRI synovitis score over time than those with a decrease in synovitis [39, 46].

Studies of hand OA patients report similar associations. Using ultrasound, Mathiessen et al. [29] found grey-scale synovitis and power Doppler signals at baseline to strongly predict radiographic progression after 5 years, and even more so when inflammation persisted [47]. Similar results were found in two cohorts using MRI on hand OA patients, as gadolinium-enhanced synovitis was associated with both onset and progression of radiographic hand OA, including development of erosions [48, 49]. The total amount of synovial inflammation also increased the risk of progression [49].

While symptoms are the usual primary outcome of an analgesic OA trial, imaging or serum biomarkers of synovial inflammation may provide evidence for proof of mechanism [15]. Synovial tissue volume was used for this purpose in a recent open-label study of 120 patients with knee OA receiving intra-articular (IA) steroid injection with subsequent reduction in CE-MRI synovial volume that correlated with improvement in knee pain [50]. Dynamic CE-MRI-derived measurements of synovial enhancement, where images are acquired every few seconds after contrast injection, were shown to be even more sensitive to the response of treatment and more strongly associated with changes in pain than synovial tissue volume [51, 52].

Part of the evidence to support therapeutic targeting of inflammation in OA is that the two pharmacological therapies with consistently good effect sizes in relieving OA pain are NSAIDs and IA corticosteroid injections [53]; both have anti-inflammatory mechanisms of action.

Further support for the concept of treating inflammation is that the analgesic benefits from IA steroids also seem dose dependent. In a hip OA study where 120 patients received either 40 or 80 mg of IA corticosteroids, both doses had a beneficial effect at week 6 (in terms of pain, stiffness and disability), while the 80 mg dose demonstrated a statistically significant benefit at week 12 [54]. This stands in contrast to low-dose oral prednisolone (5 mg) which had no analgesic effect on 70 hand OA patients in a short-term (4 weeks) randomized controlled trial [55]. Intramuscular depot corticosteroid demonstrated somewhat better results with significant short-term effects on knee pain [56], although no statistically significant reduction in ultrasound-detected synovial inflammation was found [56]. A preliminary report of a novel slow-release, microsphere formulation of triamcinolone in a large OA knee randomized controlled trial demonstrated important analgesic benefits to 13 weeks, offering hope of long-lasting IA steroid therapy [57].

Methotrexate (MTX) has an anti-inflammatory effect by suppressing the inflammatory functions of neutrophils, macrophages and monocytes, dendritic cells and lymphocytes through adenosine release, and thereby reducing secretion of inflammatory cytokines, including TNF-α and IL-6. A recent open-label pilot study of 30 patients with knee OA who took a MTX dose of 15–20 mg/week for 6 months suggested an analgesic benefit, with 43% achieving the Osteoarthritis Research Society International (OARSI) responder criteria [58]. A subsequent randomized controlled trial is underway.

A recent review has summarized the role of biologic therapies that modify inflammatory components, including anti-TNF agents, growth factors and IL-1 antagonists, and their reported effects in OA trials [59]. Anti-TNF agents have to date had minimal effects on symptoms, although the trial designs are heterogeneous and there may be a possible structural-modifying effect in joints with the most inflammation [60]. Also, there are emerging and promising data on anti-inflammatory cytokines such as IL-4 and IL-10, as well as anti-CSF therapies, as described earlier.