Assessment of synovitis to predict bone erosions in rheumatoid arthritis

Abstract

Although rheumatoid arthritis (RA) is traditionally considered as the prototype of destructive arthritis, the course of the disease varies considerably, with some patients experiencing more rapid progression of joint damage and disability than others. Given the increasing availability of treatment targets and options, timely recognition of individual’s outcomes could allow therapeutic allocation according to personalized benefit–risk profiles. Research efforts are thus increasingly focused at discovering predictive markers that could identify patients with aggressive, rapidly progressive disease and poor prognosis. As joint destruction in RA is the result of the cumulative burden of inflammation, variables reflecting the severity of synovitis and its persistence over time might refine our ability to build early prognostic algorithms. The goal of this article is to review the clinical implications of the assessment of synovitis in relation to radiographic outcomes. Traditional and novel assessment tools will be discussed, including clinical measures, imaging techniques and tissue biomarkers. Achievements in the field of synovial tissue analysis and peripheral blood biomarkers of synovitis represent only the first steps of ongoing progress, which still need to be integrated into the phenotypic heterogeneity of RA.

Introduction

Rheumatoid arthritis (RA) is traditionally considered as the prototype of destructive arthritis. Chronic inflammation of the synovial membrane in RA indeed results in bone and cartilage resorption through the production of molecules that support the differentiation and activation of osteoclasts as well as enzymes which degrade articular cartilage and bone [Schett, 2007; Walsh and Gravallese, 2010]. The central importance of joint remodeling processes in RA pathology is highlighted by the fact that the assessment of structural damage using imaging techniques is a major diagnostic, monitoring, and outcome parameter in both clinical trials and routine clinical practice [Lillegraven et al. 2012]. Joint destruction often begins very early and progresses rapidly within the first years from symptoms onset [Lindqvist et al. 2003; Machold et al. 2007]. The severity of bone and cartilage damage does have functional implications at least in the later stages of the disease, as suggested by multiple studies demonstrating the association between increasing joint destruction over time and progressive decline in physical function [Drossaers-Bakker et al. 1999; Scott et al. 2000; Welsing et al. 2001]. Preventing joint damage is therefore a key measure of current and novel therapies and treatment strategies in RA.

Despite common clinical and pathophysiological features, RA appears to consist of a variety of different phenotypes in terms of clinical expression and long-term course, including the extent and rate of progression of joint damage. Studies on very early arthritis (≤3 months from symptom onset) have shown that up to 20% of RA patients already present with erosions at baseline despite early referral [Nell et al. 2004; Machold et al. 2007], and substantial structural damage further develops even during disease-modifying antirheumatic drug (DMARD) therapy [Machold et al. 2007; Rezaei et al. 2012]. Such a proportion may be even higher in selected cohorts of patients with aggressive RA eligible for randomized clinical trials, such as in the BeSt study, where erosive disease is observed in up to 70% of the cases [Visser et al. 2010]. On the other hand, rheumatologists in daily clinical practice well appreciate that a considerable number of patients does not experience any erosion even after long-term disease. Following the greater availability of treatment targets and options, the prospect of distinguishing which patient with early RA is to run a severe disease course and which is not is, at present, one of the main challenges in the management of RA. Early predictors of disease outcome would indeed ideally channel aggressive treatments to patients most at risk, thus optimizing efficacy, adverse events and socioeconomic costs.

A number of demographic, immunologic and genetic factors have been variably associated with more severe outcomes in RA at the population level, including age, gender, autoantibody status and human leukocyte antigen (HLA) genotyping [Markatseli et al. 2010]. In particular, positivity for anticitrullinated protein antibodies (ACPAs) is currently recognized as one of the strongest independent predictors of erosive disease [Meyer et al. 2003; Forslind et al. 2004; van der Helm-van Mil et al. 2005; Syversen et al. 2008]. In the study by Syversen and colleagues [Syversen et al. 2008], a positive ACPA test predicted radiographic progression after 10 years of follow up with an odds ratio (OR) of 4 in a multiple regression model also including female gender, high erythrocyte sedimentation rate (ESR) and IgM rheumatoid factor (RF) positivity. Combined together, the probability of radiographic progression was 92.5% in females with a positive ACPA and IgM RF test and a high ESR [Syversen et al. 2008]. Despite excellent sensitivity, however, specificity may be unsatisfactory. A recent study by Liao and colleagues [Liao et al. 2011] failed to demonstrate a significant value of the ACPA test in predicting erosion-free status after 2 years of follow up. Many efforts are thus being made in the search for additional markers able to identify different radiographic outcomes since the earliest stages of the disease. As joint destruction in RA is the result of the cumulative burden of inflammation, predictive factors of radiographic progression may be searched within variables associated with the severity of inflammation and its persistence over time. Here we focus on new and renewed data which make the assessment of synovitis through clinical, imaging, histopathologic and serologic examinations, a valuable tool to look into disease heterogeneity in RA.

Clinical assessment

Clinical evaluation of involved joints through formal joint counts remains the cornerstone of the quantitative assessment of inflammation in RA. Physical examination per se is thus expected to provide important information on different radiological outcomes of the disease. Accordingly, cumulative swollen joint counts (SJCs) have repeatedly been shown to be associated with joint damage progression over time at both the patient and the individual joint level [Van Leeuwen et al. 1994; Smolen et al. 2006; Boers et al. 2001; Klarenbeek et al. 2010]. A proof of concept of the tight relationship between clinically active joints and structural changes is provided by the demonstration that repair (the opposite of progression), although it remains an extremely rare feature in RA [van der Linden et al. 2010], may only occur in association with improvement or cessation of clinical swelling [Lukas et al. 2010]. The crucial importance of joint assessment to predict radiographic outcomes in patients with RA is further highlighted by recent evidences showing that joint damage progression in remission is driven by residual swollen joints [Aletaha and Smolen, 2011], which appear to be more predictive compared with other variables of inflammation such as acute phase reactants [Aletaha et al. 2011]. The association between clinical inflammation and joint destruction has strong foundation in patients treated synthetic DMARDs. It is, however, worthwhile to underline that the relationship is not as clearly defined in patients treated with biologic agents, in which an apparent dissociation between clinical and radiographic outcomes can be observed. Patients with clinical disease activity treated with anti-TNF (tumor necrosis factor) agents may indeed have no evidence of erosive progression [Smolen et al. 2005]. Accordingly, SJCs averaged over time appear associated with radiographic outcomes during methotrexate monotherapy but not during combination therapy with infliximab [Smolen et al. 2006]. The fact that treatment choice per se is a major determinant of radiographic progression is further highlighted by recent evidence demonstrating that other predictors (autoantibody status, acute phase reactants, baseline erosion scores), alone or in combination, perform differently in patients receiving different treatment strategies [Visser et al. 2010].

If structural outcomes can collectively be determined by persistent joint swelling, a clinically relevant issue is whether the severity of joint involvement at baseline may be equally predictive of the rate of radiographic progression over time. This would imply that patients with high levels of disease activity at presentation will continue to do badly. The relationship between the amount of initial joint inflammation and the persistence of inflammation itself, however, is not so obvious [Drouin and Haraoui, 2010]. Accordingly, there are conflicting results from the literature concerning the predictive value of a single assessment of SJCs at baseline for radiographic damage. In the long-term study by Kaarela [Kaarela, 1985], the number of swollen joints was an independent predictive factor of radiological progression. This association was either weaker [Smolen et al. 2006] or not confirmed in more recent studies [Combe et al. 2001; Machold et al. 2007; Courvoisier et al. 2008]. Such a discrepancy is also reflected by the variable inclusion of SJC in the recently developed matrix risk models for the prediction of rapid radiographic progression in RA [Vastesaeger et al. 2009; Visser et al. 2010]. In contrast, re-assessment of joint involvement soon after initiation of therapy seems more informative on radiographic outcomes [Smolen et al. 2006], in line with the notion that the strongest determinant of persistent disease activity is the level of disease activity itself after the first few months of treatment [Aletaha et al. 2007].

Imaging assessment

Modern ultrasonography (US) enables clinicians to sensitively assess soft-tissue structures, and the use of power Doppler (PD) further extends the analysis to synovial vascularity. Being more sensitive than clinical examination in detecting synovial inflammation [Naredo et al. 2005] and in discriminating between ‘active’ and ‘inactive’ joints, US-PD was expected to provide additional value compared with manual joint counts in predicting the evolution of RA. Using radiographic progression as a final measure of disease outcome, results have been however conflicting. If on the one hand time-integrated values of US-PD correlate well with the amount of joint damage progression [Naredo et al. 2007], a single time assessment at baseline has not been consistently associated with radiographic outcomes [Taylor et al. 2004; Naredo et al. 2007; Reynolds et al. 2009; Fukae et al. 2010]. Thus, similarly to SJCs, measuring cumulative synovial inflammation over time is more informative compared with single time point measurements, and highest levels of synovitis at baseline do not appear to predict per se worst disease evolution, irrespective of how ‘sensitive’ synovial inflammation is measured. Also, it is important to emphasize that the very high sensitivity of advanced PD equipments may to some extent compromise specificity. PD activity is indeed observed in the hand and finger joints of a number of healthy individuals [Terslev et al. 2008]. Similarly, US-PD measurements may be affected by physiological or pharmaceutical stimuli also in RA patients [Ellegaard et al. 2009; Zayat et al. 2011]. Altogether, these limitations should be taken into account when considering the prognostic value of US-PD assessment of disease activity. A particular setting in which US-PD may provide valuable information for disease prognostication is the state of clinical remission or low disease activity. Here indeed inflammatory findings on US-PD are commonly seen [Brown et al. 2006, 2008; Scirè et al. 2009], and both grayscale synovial hypertrophy and PD synovitis have been demonstrated to be independent predictors of structural progression on radiographs [Brown et al. 2008]. Owing to the lack of comparison between the predictive value of manual SJCs and US-PD scores in these patients, however, the superiority of imaging techniques compared with physical examination remains to be demonstrated.

Similarly to US-PD, synovitis detected through magnetic resonance imaging (MRI) does not appear predictive of radiographic erosions when assessed at baseline in active disease [McQueen et al. 2003; Bøyesen et al. 2011], again confirming the weak prognostic value of single time measures. Accordingly, a recent systematic review reported a variable ability of MRI findings to predict radiographic progression (range 18–100% for sensitivity and 5.9–97% for specificity), thus not recommending its routine clinical use [Suter et al. 2011]. MRI, however, can uniquely visualize another relevant inflammatory lesion in RA, the bone marrow edema (BME), which appears as an increased signal intensity on fat-suppressed T2-weighted, proton density or short tau inversion recovery (STIR) pulse sequences, consistent with increased water content [Østergaard et al. 2003]. Although BME is not specific for RA, its histopathological characterization in long-standing disease has consistently shown that BME truly corresponds to intramarrow inflammation [Jimenez-Boj et al. 2007; McQueen et al. 2007]. Differently from synovitis, BME appears as a strong predictor of radiographic progression in both early and established disease even when single baseline assessments are performed [McQueen et al. 1999; Hetland et al. 2009, 2010; Bøyesen et al. 2011]. In particular, in the study by Hetland and colleagues, the predictive value of MRI was investigated in a standardized treatment setting against a number of other relevant prognostic factors, such as immunologic (ACPA, IgM RF, IgA RF), environmental (smoking, educational level), genetic (shared epitope) and disease activity markers [Hetland et al. 2009]. In this comprehensive model, MRI BME at presentation was the strongest independent predictor of radiographic progression 2 years later in early RA patients as determined in both multivariable linear and logistic regression analyses. The 5-year extension study in the same cohort confirmed the predictive ability of BME in the long term [Hetland et al. 2010]. The different prognostic significance of the assessment of BME at baseline compared with synovitis is currently unexplained. Whether BME represents a more sustained inflammatory process remains to be demonstrated, and different accessibility of the subchondral bone marrow compartment to treatment and/or different cellular composition of the inflammatory infiltrates are only speculative possibilities.

Synovial tissue analysis

Histological and molecular analyses of the synovial membrane ideally represent the gold standard for the assessment of joint inflammation in RA. The synovial tissue indeed undergoes early inflammatory changes, which include resident cell activation and infiltration by hematopoietic cells [Hitchon and El-Gabalawy, 2011], that can be clearly recognized even when clinical signs of arthritis are lacking [Kraan et al. 1998]. Compared with clinical and imaging assessments, examination of the synovial tissue offers the potential advantage of capturing both quantitative and qualitative features of the inflammatory process of RA. The rheumatoid lesion is indeed characterized by a high degree of cellular and molecular heterogeneity across different disease phases and clinical subsets [Bugatti et al. 2011a]. The pathophysiological background and the clinical correlates of such heterogeneity have only in part started to be understood. Synovial tissue analysis is therefore at present mainly a research area, with limited clinical value in daily practice. However, as synovitis is the primary event underlying signs and symptoms of arthritis in RA, increasing efforts are being made in order to identify synovial prognostic biomarkers that could be used in individual patients [de Hair et al. 2011]. The widespread diffusion of mini-invasive techniques for synovial sampling has recently made the clinical use of synovial tissue analysis a more realistic goal. Major drawbacks that need to be considered in light of the potential for clinical translation include the fact that synovial biopsy, although ‘mini’, remains an invasive procedure and requires some technical skills and facilities that might ultimately influence the outcome. In addition, the synovial tissue is more easily accessible in larger joints such as the knee, which is generally affected in patients with a more aggressive disease course. This could account for selection bias. Finally, although multiple sampling from different locations within the joint is currently recommended [van de Sande et al. 2011], overestimation or underestimation of inflammation due to the wide heterogeneity of the synovium cannot be excluded completely.

A number of independent studies have tried to identify the microscopic and molecular factors possibly associated with joint damage progression in RA. One of the cell populations which has historically attracted great interest are synovial marcophages. It is widely accepted that the number of CD68+ macrophages in the sublining strongly correlates with local disease activity [Tak et al. 1997] and systemic inflammation [Baeten et al. 2005], and is sensitive to change following active antirheumatic therapies [Haringman et al. 2005]. The dynamic correlation of tissue macrophages with the levels of ongoing inflammation, however, intrinsically limits their possible long-term prognostic value as a single time measure. The association between the degree of synovial macrophage infiltration and the development of bone erosions at follow up appears indeed highly controversial [Mulherin et al. 1996; Kraan et al. 2004]. The same applies to macrophage-derived cytokines, none of which has been shown to accurately predict joint damage progression in RA [Rooney et al. 2010].

Another relevant aspect of synovial tissue heterogeneity with possible pathophysiologic and clinical value is the degree and local organization of infiltrating T and B lymphocytes, which have been classically described as either randomly distributed within the sublining or spatially grouped into discrete lymphoid microenvironments [Takemura et al. 2001]. The functional properties associated with the development of T–B cell niches in the context of the chronically inflamed synovium are beyond the scope of this review and have been extensively summarized elsewhere [Manzo et al. 2010]. Different lymphoid patterns are also been intensively investigated for their possible translational value. Depending on the histological criteria adopted to identify lymphocytic aggregates as well as the cutoff values set for defining lymphoid aggregation, different clinical studies yielded partly conflicting conclusions on whether these structures are associated with more severe RA subsets in terms of radiographic damage [Klimiuk et al. 2003; Thurlings et al. 2008; Cañete et al. 2009]. However, it appears that the more selectively we look at specific B-cell-centered processes the more unfavorable outcomes we capture. When B cells are analyzed as core elements of the process of lymphocytic aggregation, indeed, increasing degrees of B-cell infiltration appear to be associated with a higher prevalence of erosive disease cross-sectionally, in parallel with an unbalanced molecular milieu favoring bone remodeling [Bugatti et al. 2011b]. These same findings are true when other cell populations of the B-cell lineage are analyzed, such as CD79a+ B cells [Mo et al. 2011]. Whether B-cell-rich synovitis and bone erosions are parallel manifestations of a more severe disease subset or are temporarily distinct phenomena remains to be demonstrated. Furthermore, as B cells are precursors of antibody-producing cells and autoantibodies are recognized as relevant prognostic factors in erosive RA, the clinical value of synovial B-cell infiltration theoretically may not be independent from the autoantibody status. The relationship between synovial lymphoid aggregates and RA-specific autoantibodies is however not unequivocal. Whilst previous studies have suggested that the process of lymphocytic aggregation is neither restricted nor preferentially associated with autoantibody positivity [Thurlings et al. 2008; Cantaert et al. 2008], recent data obtained in murine chimeric systems demonstrated a significant relationship between synovial B-cell niches and ongoing production of ACPA [Humby et al. 2009]. Overall, the independent prognostic value of synovial B-cell infiltration is at present nondeterminable, and prospective studies in early RA are awaited.

Synovial tissue biomarkers in peripheral blood

Overall, the specificity of clinical and imaging assessments of inflammation at disease onset is too low to make such variables fully predictive of joint damage progression in RA. On the other hand, synovial tissue analysis is still in its infancy in terms of clinical translation. Great efforts are thus being made in order to identify novel variables which reliably reflect the pathophysiological cascade of RA and which can be used as prognostic markers of disease severity. In this field, serological biomarkers reflecting bone and cartilage destruction are emerging as valid tools and have been the subject of a recent review [Karsdal et al. 2011]. As bone and cartilage turnover represent the downstream effect of joint inflammation, equal attention is in parallel being paid to possible biomarkers of synovitis.

C-reactive protein (CRP) and ESR are only indirectly linked to synovitis. The acute phase proteins are indeed primarily produced by hepatocytes in response to interleukins (IL) released during inflammation [Gabay and Kushner, 1999]. Although the predictive value of ESR and, to a lesser extent, of CRP for radiographic progression has been proposed in several independent studies [Combe et al. 2001; Lindqvist et al. 2003, 2005], their value compared with physical examination appears overall poor [Aletaha et al. 2011]. Disappointingly, however, none of the cytokines that drive the acute systemic response and that can be produced at least in part at the synovial tissue level, such as IL-1, IL-6 and TNF-α, has been convincingly shown to be associated with progression of joint damage [Roux-Lombard et al. 2001; Rooney et al. 2010], thus making the crude assessment of classical inflammatory biomarkers hardly advisable.

Additional value could be provided by serological markers which reflect more specific aspects of synovial pathology, differently regulated in different patients, such as angiogenesis (and vascular biology in general) and mononuclear cell infiltration. The degree of activation of the vascular endothelium, as assessed through E-selectin serum levels, has been shown to be associated with progression of joint destruction in early RA [Kuuliala et al. 2002], and similar data have been also provided for vascular endothelial growth factor (VEGF) [Ballara et al. 2001]. More recently, markers of different T-cell subsets, such as granzyme B and IL-22, have been shown to be elevated in the serum of RA patients in association with radiographic progression [Goldbach-Mansky et al. 2005; Leipe et al. 2011]. Overall, the specificity and predictive value of such markers would however need confirmation in further studies.

An emerging peripheral biomarker of synovitis is the chemokine CXCL13. CXCL13 is involved in the positioning, cooperation and activation of B and T cells within lymphoid and extralymphoid sites, including RA synovium [Loetscher and Moser, 2002; Manzo et al. 2010]. Peripheral blood and serum expression of CXCL13 are increased in RA patients compared with healthy donors [Manzo et al. 2008; Rioja et al. 2008; Rosengren et al. 2011], and preliminary results indicate a significant correlation between serum and synovial tissue levels of CXCL13 [Rosengren et al. 2011]. CXCL13 appears as a marker of severity in RA. A large prospective study has indeed shown that early RA patients with the highest levels of serum CXCL13 are those with the highest rate of progression of joint damage over long-term follow up [Meeuwisse et al. 2011]. In line with these data, we could demonstrate that serum levels of CXCL13 in untreated patients with disease duration <12 months are associated with clinical and US synovitis and predict US-PD signal persistence [Bugatti et al. 2012]. Importantly, both in the study by Meeuwisse and colleagues and in our previous study, the predictive value of CXCL13 was independent of the level of inflammation indicated by the CRP level as well as the ACPA status [Meeuwisse et al. 2011; Bugatti et al. 2012]. Thus, although larger, prospective, confirmatory studies are needed, there may be opportunities to identify peripheral blood biomarkers capable of reflecting synovial pathology and predicting clinical outcomes in RA.

Conclusions

Patients with different outcomes are still difficult to identify in the earliest stages of the disease, and RA phenotypic heterogeneity remains largely unpredictable. As joint damage progression is the result of the cumulative burden of inflammation over time, the identification of disease variables associated with persistent, refractory joint inflammation could provide valuable tools to determine which patient is to run a more severe and rapid radiographic course. Joint inflammation can now be assessed through a variety of different techniques able to provide increasing information on tissue pathobiology. Achievements in the field of synovial tissue analysis and biomarkers of synovitis in the peripheral blood represent only the first steps of the ongoing progress and cannot currently guide clinical decision making. Further research is needed to integrate clinical, imaging and biologic studies into comprehensive models able to accurately predict disease severity and stratify patients into prognostic subgroups.