Introduction
Rheumatoid arthritis (RA) is a chronic inflammatory disease characterised by progressive joint damage. The aetiology of RA is, to a large extent, unknown, and current therapies are aimed at reducing the inflammatory disease burden. Recent data support the practice of frequent monitoring of disease activity in RA patients to achieve remission faster [
1,
2]. However, patients may still experience radiographic progression of bone erosions, despite being judged to be in clinical remission [
1,
3,
4]. The emergence of imaging techniques such as ultrasound (US) and magnetic resonance imaging (MRI) have paved the way for detailed descriptions of joint inflammation and erosion. These imaging modalities have recently been accepted as part of the diagnostic criteria in RA [
1,
5]. Because inflammatory activity detected by MRI or Doppler ultrasound (DUS) has shown to be a good predictor of disease course and response to treatment, the expansion of the use of these detection modalities in clinical assessments of RA has been suggested [
6‐
8].
The invasive front of the synovium—the pannus—is believed to be the main driver of cartilage degradation and bone erosion; however, to the best of our knowledge, no researchers to date have investigated the association of histological features of synovitis with the extent of bone marrow oedema (BME) and erosions visualised by MRI. This information would be interesting because the presence of cluster of differentiation–positive (CD68+) macrophages in RA synovium has been linked to the development of erosions detected on radiographs [
9]. Investigators who conducted MRI-based studies focused mainly on, and demonstrated, associations of MRI-detected synovitis and synovial histopathology [
10‐
12] with MRI-detected BME and osseous histopathology [
13‐
15]. Furthermore, previously published data have generally been based on small sample sizes. Also, the synovial material examined in the vast majority of cases was obtained from knee joints of patients undergoing knee joint replacement surgery in cases where inflammation due to concomitant osteoarthritis may have confounded the results. Current evidence is based mainly on studies in which the investigators assessed pathology on the basis of whole-joint imaging, which may have decreased sensitivity for the detection of local activity in the different areas of the synovium.
Studies of RA synovial explants have demonstrated the value of the capacity they provide for evaluating inflammatory output and the response to anti-inflammatory treatment, as well as for visualising the morphologic features of the RA synovium [
16‐
24]. Several methods of establishing synovial cultures have been described, but the use of assays based on whole synovial tissue has been recommended in order to maintain synovial architecture and cell-to-cell contact [
25]. Researchers in prior biopsy-based explant studies might not have taken the heterogeneous synovial tissue distribution into account, however, thus compromising the explant assays’ ability to detect overall joint inflammation [
18,
26]. In recent years, novel immunoassay techniques, such as multiplex technology, have paved the way for simultaneous analysis of multiple cytokines and chemokines in a small sample. However, it has been shown that synovial production of heterophilic antibodies (HAs), such as rheumatoid factor (RF), can have a great impact on assay reliability if they are not blocked [
27]. In contrast to synovial biopsies, explants consist of live cells, and the production of inflammatory mediators from these intact tissue cultures may thus provide novel information about RA pathogenesis.
Our aim in this study was to compare the association of synovial inflammatory mediator production with the colour Doppler ultrasound (CDUS) fraction [
28] and/or the rheumatoid arthritis magnetic resonance imaging score (RAMRIS) [
7] using 3-T MRI scans of the corresponding anatomical sites in the target hand. Obtaining synovial tissue by synovectomy under direct visual control enabled us to harvest as much of the synovium as possible from the different anatomical sites of the hand joints. The results of this study should thus provide novel insights into the association between synovial inflammatory mediators and RA hallmarks.
Discussion
In this study of the hand joints of RA patients, the degree of synovitis—estimated by CDUS using the CFmax and by MRI using the RAMRIS criteria—was associated with synovial explant production of key inflammatory mediators. This finding supports previously published observations that DUS and MRI are sensitive imaging modalities for the detection of local inflammation. Among the four mediators investigated in our present study, we observed an interesting polarisation. Whereas IL-6, MCP-1 and MIP-1β were associated with synovitis detected by MRI or CFmax in combination, only MCP-1 and IL-6 were associated with the extent of BME and bone erosions visualised by MRI. Furthermore IL-6, IL-8 and MCP-1 production were correlated with DAS-28-CRP score, indicating that the local synovial mediator production reflects systemic disease.
This polarisation in synovial mediator expression is in accord with results produced in biopsy studies in which investigators examined differences in cellular populations and mediator and chemokine receptor expression at the cartilage–pannus junction (CPJ), which is believed to be the site driving the erosive process in the bones of RA patients. In biopsies, IL-6 expression was found exclusively at the CPJ, in contrast to IL-8-positive cells (primarily CD68+ macrophages), the vast majority of which were located at non-CPJ locations [
36,
37]. The picture is not completely clear, however, because CD68+ macrophages have been linked to radiographically visualised RA progression [
38]. MCP-1 and its receptor, chemokine (C-C motif) receptor 2, was located with the highest density near the synovial lining, in contrast to MIP-1β, which had the most pronounced staining in synovial endothelium. MCP-1- and MIP-1β-positive cells, however, were located throughout the synovium [
39]. The lack of correlation with DAS-28-CRP score and MIP-1β may indicate that synovial MIP-1β production has a role in RA pathogenesis that is more marginal than that of IL-8, which also had a strong signal (approximate ρ = 0.58) to MRI-detected synovitis. In contrast to MIP-1β, IL-8 was statistically significantly associated with DAS-28-CRP score. However, further studies are needed to verify the clinical significance of the cytokine expression pattern in the synovium.
The lack of association between BME and bone erosion with synovial IL-8 and MIP-1β release can be explained by the main production of these mediators’ originating from non-CPJ areas in the synovium. Researchers in two large clinical studies reported a lack of association of radiographic outcome and synovitis in anti–tumour necrosis factor α (anti-TNF-α)–treated cohorts [
40,
41]. These findings support our observations of different implications of these mediators in RA pathogenesis and suggest that changes in mediator profiles upon treatment may be able to predict and stratify disease outcomes with regard to synovitis and bone destruction.
Some limitations should be taken into account concerning day-to-day variations in our study. The RA patients in our study had DUS and MRI performed 24 hours prior to surgery, and day-to-day variations may have affected the reported associations [
42]. Furthermore, the experimental setup did not include synovectomies from healthy controls. Therefore, baseline information on mediator release from ‘normal’ synovium was not possible. However, cytokine profiling of synovial fluid from healthy controls using a similar Bio-Plex assay setup (Bio-Rad Laboratories, Hercules, CA, USA) showed that the synovial fluid quantities of the four mediators were extremely low, with MCP-1 having the highest mean concentration at 0.5 ± 0.4 pg/ml [
30]. We therefore believe that cytokine levels above assay detection limits (lowest detection limit of IL-8 = 26 pg/ml) should reflect synovial pathology. Taking the very low synovial fluid cytokine concentrations into account, the relatively high cutoff values for the assay’s lowest detection limits may have resulted in some overestimation of synovial explant cytokine release. Considering the generally high concentration of synovial explant mediator release, an assay imprecision in the range of 26 to 360 pg/ml, depending on the cytokine, should not affect the overall associations with regard to imaging.
The results of this study show a noteworthy variation in synovial cytokine production and corresponding imaging pathology. The absence of signalling on DUS scans does not rule out synovial cytokine production. These findings are in accord with the results of our previous study in which DUS-detected activity was compared to synovial histopathology in a similar group of RA patients [
35]. That previous study did not include grading of the synovial hypertrophy, because the images were obtained only for evaluation of colour Doppler. Further research on the association of synovial hypertrophy with synovial inflammation may provide important insights into RA disease pathogenesis.
Cellular stress inflicted by the in vitro circumstances could perhaps have skewed the mediator release from the explant cultures, resulting in higher mediator concentrations. This phenomenon did not generally occur, however, as low concentrations of synovial mediators were observed in some of the explants cultured from sites with moderate to high RAMRIS scores and in the presence of DUS-visualised activity. The observed variations may in fact have been caused by different kinetics controlling synovial perfusion, cellularity and mediator production. These factors may also account for the differences in the cytokines associated with synovial perfusion detected by CFmax and with synovitis detected by MRI. Use of biological DMARDs (bDMARDs) may result in changes in cytokine production, but, because of the limited population size in this study, it was not possible to adjust for the different treatment modalities. Further research focused on explants and imaging in a prospective study may clarify the robustness of this method for prognostic use in bDMARD treatment. We made a great effort in designing the study to guide the area of the synovectomy on the basis of the US data; however, a completely accurate match was not possible. MRI mapping of the RAMRIS synovitis score was performed at a regional level in the wrist (RC or MC). Furthermore, BME and bone erosion data were determined as averaged values derived from the synovectomised area. This may represent a source of bias because the synovium was removed only from the dorsal part of the joint. Unfortunately, because of insufficient matching imaging data, a comparison of MRI and CDUS regarding their utility in detecting synovitis was not possible.
The reason IL-8 did not reach a statistically significant association with the MRI synovitis score is likely due to the large number of wells (52%) above the assay’s upper detection limit. None of the other cytokine levels were above the assay detection limit. Among the wells that were above detection levels, an MRI-based synovitis score of 3 was found in 72%, whereas the remaining wells had a score of 2. Budget considerations and the high dilution factors limited the possibilities of including several key inflammatory mediator candidates for measurement, despite their presence in supernatants. These candidate mediators included TNF-α, vascular endothelial growth factor, matrix metalloprotease 3, tissue inhibitor of metalloprotease 1, IL-10 and interferon γ. The synovial explant model therefore offers significant information regarding synovial inflammatory activity. Interestingly, IL-1β levels were very low and under the detection limit (0.57 pg/ml) in five of eight patients in the initial screening. A recent publication also described low levels of IL-1β and TNF-α based on a whole-tissue synovial explant system [
43]. This information is in contrast to reports of studies in which enzyme-digested synovial tissue was used, possibly due to changes induced by the digestion process (for example, inhibition of formation of three-dimensional cell layers or lipopolysaccharides in the collagenase, which are potent inducers of IL-1β and TNF-α) [
44].
All patients included in our present study had synovial pathology defined by synovial hypertrophy visualised on US scans. Several sites were without DUS-visualised activity, and one case (three synovial explant positions) was judged to have no synovitis on the basis of MRI. Our findings emphasise the importance of considering the degree of greyscale synovitis when evaluating RA patients on the basis of US in the clinic.
Authors’ contributors
MA was responsible for the study conception and design, manuscript writing, establishment of the synovial explant assay, data collection and analysis and critical revision of the manuscript. MB was responsible for the study conception and design, manuscript writing, data collection and analysis and critical revision of the manuscript. KE and STP were responsible for the study conception and design, data collection and analysis and critical revision of the manuscript. RC was responsible for the study conception and design, manuscript writing, statistical analyses and critical revision of the manuscript. KS, EMB, NV was responsible for the study conception and design and critical revision of the manuscript. NS was responsible for the study conception and design, manuscript writing, data collection and critical revision of the manuscript. PS, UM, BDS, LK and HB were responsible for the study conception and design, securing funding and critical revision of the manuscript. All authors read and approved the final manuscript.
Acknowledgments
The authors would like to thank the study participants as well as Inger Wätjen, Eva Littrup Andersen, Mette Okkels, Jette Møller Frøsig and Suzi Høeg Madsen. The study was supported by unrestricted grants from Novo Nordisk; The Danish Agency for Science, Technology and Innovation; The Oak Foundation; and all the centres involved in the study, which were The Parker Institute; the Department of Rheumatology, Copenhagen University Hospital, Bispebjerg and Frederiksberg, Copenhagen; Translational Immunology, Biopharmaceutical Research Unit, Måløv, Denmark; Novo Nordisk, Copenhagen; the Department of Radiology, Copenhagen University Hospital, Bispebjerg and Frederiksberg, Copenhagen; the Section of Hand Surgery, Department of Orthopedics, Gentofte University Hospital, Hellerup, Denmark; Biomarkers, Novo Nordisk, Søborg, Denmark; and the Department of Anesthesiology, Intensive Care and Operations, Gentofte University Hospital. The Parker Institute is supported by grants from the Oak Foundation, a nonprofit funding source that had no role in the design or conduct of the study; the collection, analysis and interpretation of the data; or the preparation, review and approval of the version of the manuscript submitted for publication.
Competing interests
During the course of this study Martin Andersen, Kalle Söderstöm, Pieter Spee, Ulrik GW Mørch, and Lars Karlsson were employed at Novo Nordisk. Kalle Söderstöm, Pieter Spee, Ulrik GW Mørch, and Lars Karlsson owned stocks in Novo Nordisk. MA, KS, UGWM, PS and LK were all employed by Novo Nordisk during the patient recruitment period. KS, PS, UGWM and LK owned stocks in Novo Nordisk.