Psoriatic arthritis is a chronic and debilitating inflammatory arthropathy. It accounts for 15% of referrals to early arthritis clinics, and has considerable morbidity [¹]. The Outcome Measures in Rheumatology Clinical Trials (OMERACT) PsA working group has identified a hierarchy of domains to be included in PsA clinical trials [²], which includes tissue analysis and magnetic resonance imaging (MRI) in the outer domain, on the research agenda. Utilizing these two domains, we have sought a potential synovial biomarker of treatment response in PsA. A biomarker is defined as a characteristic that is objectively measured and evaluated as an indicator of a normal biologic process, a pathophysiologic process, or a pharmacological response to therapeutic intervention [³].

It has already been established in rheumatoid arthritis (RA) that the mean change in DAS28 correlates with the mean change in synovial sublining CD68 expression across several RA patient cohorts receiving different therapeutic agents [^4-7]. Few studies have correlated clinical composite scores with changes in PsA synovial cell populations however. One of the reasons for this is that no composite score has yet been validated in PsA, although such work is currently in progress [⁸]. DAS28, a score validated in RA [⁹], has proven to be a highly effective tool in previous studies of PsA and biologic agents [^10-12] and is suitable for studies involving synovial tissue analysis as it focuses on articular involvement. In the synovial tissue of our patient cohort, we measured the expression of CD68, a macrophage marker, given the clinical correlations found in RA; FVIII, an endothelial cell marker, due to the hypervascularity and vessel tortuosity evident in inflammed PsA synovium compared to that of RA [^13-16]; and CD3, a T-cell marker. Importantly, a previously published study which utilized DAS28 found a significant correlation between ΔDAS28 and ΔCD3 in the synovium of patients with PsA after adalimumab treatment [¹²]. Should this finding prove reproducible, particularly if different therapeutic agents are used, ΔCD3 may meet the discrimination criterion of the OMERACT biomarker validation filter [¹⁷] and the exploration of ΔCD3 as a predictive biomarker of future treatment response in PsA would be supported. ΔCD3 could be used to determine the potential efficacy of new therapeutic agents in PsA at an early stage, as is already happening in RA clinical trials of novel therapeutic compounds, where synovial sublining ΔCD68 measurements are being observed to reflect clinical response [¹⁸,¹⁹].

While MRI has been used to highlight the importance of bone marrow oedema and entheseal sites of inflammation in PsA [²⁰,²¹], to date there have been no studies comparing histological change with quantified synovitis by dynamic or static MRI. In this study we examine the relationship between clinical scores and both immunohistochemical (IHC) and MRI measures of synovitis following biologic treatment in PsA to help identify a potential biomarker of treatment response.

Twenty-five patients completed at least one of the IHC or MRI components of these studies (19 completed all components), and were, therefore, included for clinical analysis (10 anakinra, 15 etanercept). Patients were 50% and 66.6% female, had a mean age (range) of 43.2 (27 to 60) and 48.7 (26 to 64) years and a mean disease duration of 9.1 (1 to 42) and 7.5 (1 to 29) years in the anakinra and etanercept treated cohorts, respectively. Four patients had oligoarthritis (≤4 involved joints) at enrolment (all etanercept treated); the remaining 21 patients had polyarticular disease (mean SJC66 17, SD 9.2, range 6 to 43).

Five of the anakinra patients (50%) and 11 of the etanercept patients (73.3%) were taking a non-steroidal anti-inflammatory drug (NSAID), 2 and 4 of whom were also on a stable dose of prednisolone ≤10 mg.

This study has demonstrated that change in synovial CD3+ T-cell expression correlates with both ΔDAS28 and ΔMRI synovitis scores in a cohort of patients with PsA treated with either anakinra or etanercept.

Over the last 15 years, some fundamental features of the spondyloarthropathy (SpA) synovium have been elucidated. First, the inflamed synovium of SpA appears to differ histologically from that of RA [^13-16,²⁹]. Second, the synovial histology of subtypes of SpA, including oligoarticular versus polyarticular PsA, have been shown to be similar [¹⁶,²⁹,³⁰]. Third, there are histological changes in the synovium when patients with SpA respond to effective treatment. Two studies of anti-TNF therapy in SpA have shown a significant reduction in polymorphonuclear (PMN) cells, CD4+ and CD8+ T cells and macrophage subsets after 12 weeks, plus a trend toward reduced CD3+ T cell numbers [³¹,³²]. Exclusively in PsA, reduction in T cell and sublining macrophage infiltration has been observed as early as 48 hours after an infliximab infusion [³³] and also following treatment with alefacept and methotrexate [³⁴,³⁵]. Correlations with clinical outcomes were not performed in these studies. Consistent with our etanercept responders, in a cohort of PsA patients treated with adalimumab, a significant reduction in the number of CD3 positive cells was observed after four weeks [¹²]. The number of CD68+ cells in the synovial sublining did decrease in the responders of both latter studies, but not significantly, while CD68sl expression in this current study's non-responders significantly increased. Thus, in PsA, change in synovial CD3 cell infiltration, and not CD68, appears to be a superior biomarker of treatment response. Reduction in angiogenesis has been demonstrated in PsA patients after infliximab treatment [³⁶,³⁷], but was not found after etanercept treatment in the present study, and may be related to the difference in mechanism of action between the anti-TNF antibodies compared to etanercept [³⁸].

Only two previously published trials of SpA synovium have measured DAS28 scores. Of 52 SpA patients who may have received infliximab, etanercept or no biologic treatment, DAS28 scores were calculated for 28 patients who had polyarticular disease [²⁸]. These scores correlated only with change in CD163 expression, a macrophage subset marker, in the lining layer, and not with change in CD3 or CD68 expression in the sublining layer. The patients with PsA were not evaluated as a distinct group. Only one other trial, which used adalimumab or placebo, has exclusively enrolled PsA patients and used DAS28 as a primary clinical outcome measure [¹²]. Consistent with our results, that study also demonstrated a significant correlation between ΔCD3 and ΔDAS28. Taking these two papers together, four different agents have now been used in two PsA cohorts (anakinra/etanercept and adalimumab/placebo) and both studies have found this proportional relationship between ΔCD3 and ΔDAS28.

While not conclusive, the findings in this study are also consistent with the hypothesis that T-cells play an active role in PsA pathogenesis. Large numbers of T cells are present in the PsA synovium, synovial fluid and subchondral bone beneath the entheses [¹³,²⁹,³⁹], where bone oedema and erosions can occur. Th1 derived cytokines dominate in the PsA synovium [³⁵,⁴⁰] and the CD8+ cell population contains T cell repertoires which are oligoclonally expanded [⁴¹]. The association of PsA with human leukocyte antigen (HLA) Class 1 [⁴²], the development of Ps and PsA as a manifestation of the Acquired Immunodeficiency Syndrome (AIDS) and the transmission of PsA following bone marrow transplantation [⁴³,⁴⁴] all suggest T-cells take part in disease expression. Lastly, the fact that cyclosporin and ustekinumab, which impair T-cell activation, and alefacept, which specifically targets activated T cells, are effective in PsA, support T-cell involvement further [^45-47].

The use of MRI in PsA research has been reviewed in detail [⁴⁸], and an MRI scoring system for hands in PsA has recently been developed by the OMERACT imaging group [⁴⁹,⁵⁰]. We opted for our scoring method as it focuses on knee synovitis and is semi-quantitative. MRI synovitis has been shown to reduce significantly following anti-TNF therapy in PsA [⁵¹,⁵²], and we found this to be the case in our combined group of clinical knee responders. These former studies assessed mostly hand joints as opposed to exclusively knees and used quantitative analysis. Correlations of MRI findings and histopathology in inflammatory arthritis are emerging in the literature. Bollow et al. compared dynamic MRI and sacroiliac joint immunohistochemistry [⁵³] and found T cells and macrophages to be the most common inflammatory cells in active SpA sacroiliitis, although >95% of the tissue obtained was cartilage and bone. In severe AS, MRI-detected bone oedema has correlated well with histological bone marrow oedema of zygoapophyseal joints, but less so with actual inflammatory cell infiltrates [⁵⁴]. Other studies correlating MRI findings with synovial and bone oedema histology have been undertaken in RA [⁵⁵,⁵⁶], but not yet in PsA. This is the first study, therefore, to demonstrate a relationship between MRI synovitis and CD3 expression in PsA, both at all time points in the study and when comparing the changes with treatment.

No relationship was found in this cohort between ΔMRI synovitis and ΔDAS28. The fact that this MRI data reflects change in a single joint only, in contrast to DAS28, and that improvement in DAS28 may or may not involve the knee, will contribute to this. The strongest correlation being between ΔCD3 expression and ΔMRI synovitis scores is not surprising, as these originate from the same single joint and are objectively measured, excluding any subjectivity of clinical scoring and additional pathologies that could influence single joint clinical scores.

As two patients did not undergo follow-up MRI and three different patients did not have adequate synovial tissue for analysis, our study is limited by some uncoupling of the patient groups included in IHC and MRI analyses. Also, in three pre-treatment scans (two etanercept, one anakinra), insufficient fat suppression may have led to an underestimation of the degree of synovitis.

This study demonstrates a significant correlation between synovial ΔCD3 expression and ΔDAS28, and synovial ΔCD3 expression and ΔMRI synovitis scores in a cohort of 25 patients with PsA treated with either anakinra or etanercept. The establishment of ΔCD3 as a candidate biomarker of treatment response in PsA should prompt other studies using different therapeutic agents to reinforce this concept, and also to determine its ability to predict future clinical outcomes. Further work focusing on changes in peripheral blood T-cell subsets for a more easily accessible biomarker could prove useful.

Δ: change in; AEC: amino-ethylcarbazole; CASPAR: Classification of psoriatic arthritis; CRP: C-reactive protein; d VAS: disease visual analogue scale; DAS28: disease activity score assessing 28 joints; DIA: digital image analysis; EULAR: European League Against Rheumatism; FVIII: Factor VIII; HAQ: health assessment questionnaire; IHC: immunohistochemistry; IL-1: interleukin 1; IOD: integrated optical density; ll: synovial lining layer; MRI: magnetic resonance imaging; n/a: not applicable; NR: non-responder; OMERACT: outcome measures in rheumatology clinical trials; pVAS: pain visual analogue scale; PsA: psoriatic arthritis; R: responder; RA: rheumatoid arthritis; SJC: swollen joint count; sl: synovial sublining layer; SpA: spondyloarthropathy; TJC: tender joint count; TNF: tumour necrosis factor; VAS: visual analogue scale.

UF had grant research support from GlaxoSmithKline and is a consultant for and received grant research support from Wyeth. PPT is a consultant for Abbott, Amgen, Schering-Plough, and Wyeth. DJV received grant research support from GlaxoSmithKline, is a consultant for and received grant research support from Schering-Plough, is a site primary investigator for Roche, and is a consultant for and received grant research support from Wyeth. OF received grant research support from Abbott, is a primary investigator for Bristol Myers Squibb, and received grant research support from Wyeth.

The other authors declare that they have no competing interests.

EKP cut and stored the etanercept slides and performed the initial IHC staining, performed the digital image analysis, collated all data, performed all statistical analysis and wrote the manuscript. DMG participated in the autostainer immunohistochemistry. MG cut and stored all anakinra slides and performed initial IHC staining. MV made a substantial contribution to the DIA part of this work and arranged DIA data ready for analysis. AG was involved in patient recruitment and clinical assessment. IB participated in the autostainer immunohistochemistry. UF made a substantial contribution to the statistical analysis. BB made a substantial contribution to the conception and design of the study. PPT has been involved in the DIA aspect of this work and revising the manuscript critically for content. RG arranged the MRI scanning and performed the MRI synovitis scoring. DV made a substantial contribution to the conception and design of the study and coordinated the arthroscopies. OF conceived of the study, participated in its design and coordination and has been involved in the revision of the manuscript critically for content. All authors have read and approved the final version of this manuscript.

The authors wish to acknowledge the assistance of Drs. Ceara Walsh, Ronan Mullan and Tom Smeets.

The digital image analysis aspect of this work was supported by the Dutch Arthritis Association and the European Community's F6 (Autocure) funding. The investigator originated protocols were supported by Amgen and Wyeth. Funders had no involvement in the design, production, results or presentation of this research.