Recent success stories of treating cancer with immune checkpoint inhibitors (ICIs) illustrate the important roles that programmed cell death protein-1 (PD-1) and cytotoxic T-lymphocyte-associated protein-4 (CTLA-4) play in regulating T-cell specific anti-tumor responses.1 Immunotherapy with monoclonal antibodies such as nivolumab or pembrolizumab (targeting PD-1), ipilimumab (targeting CTLA-4) or atezolizumab (targeting the ligand for PD-1, PD-L1) aims to attenuate ICI signaling in order to unleash potent T-cell mediated antitumor activity. However, ICI therapy is associated with inflammation that can recapitulate many features of autoimmunity.2 3 Approximately 50% of patients receiving ICIs experience immune-related adverse events (irAEs) that are most often reported to be gastrointestinal, dermatological or endocrine in nature.2 3 Rheumatological irAEs occur in 5%–20% of cases.4–6

Mild cases of rheumatic irAEs can be managed with corticosteroids and/or conventional disease-modifying antirheumatic drugs (cDMARDs), while refractory cases are often treated with biological DMARDs (bDMARDs) targeting tumor necrosis factor α (TNFα)7 and interleukin-6 (IL-6),8 although without any informed histopathological rationale. Critically, no clinical trials are currently in progress that will provide evidence for the preferential use of one bDMARD over another in corticosteroid and/or cDMARD-refractory irAEs. Here in, we describe a case of nivolumab-induced severe, cDMARD-refractory polyarthritis, successfully treated with synovial-biopsy informed bDMARD therapy. This is the first report characterizing parallel peripheral blood, synovial fluid (SF) and synovial tissue (ST) inflammatory infiltrates in a rheumatic irAE with comparison to equivalent samples from patients with early rheumatoid arthritis (RA).

Checkpoint inhibitors have changed the therapeutic landscape for the treatment of metastatic cancers.1 However, irAEs result in reduced quality of life2 3 and in severe cases, cessation of a potentially lifesaving therapy.

While case series have reported success in treating rheumatologic irAEs with both corticosteroids and bDMARDs, both are thought to compromise ICI-mediated antitumor effects with prolonged use16 and long-term usage of corticosteroids can result in significant systemic toxicity.2 A better understanding of the immunopathogenesis of these events will allow greater confidence in selecting appropriate therapies. Furthermore, the synovial biopsy-guided approach we propose could be extrapolated to other irAEs including the more commonly seen colitis and dermatological complications.

ST research has increased our understanding of RA pathogenesis,11 12 17 resulted in the identification of new therapeutic targets18 19 and has the potential to stratify treatment.9 10 In this study, ST sampling provided a histopathological rationale for the choice of bDMARD to treat a severe treatment-refractory rheumatic irAE. irAE ST IHC was reminiscent of the RA synovium, demonstrating differentially expressed cytokines. This invited the option of targeting therapy to the dominant inflammatory cytokine, including TNFα and IL-67 8 for which antibodies have already been trialed empirically in irAEs. Abatacept, a CTLA-4 fusion protein targeting T-cells, may result in a loss of anti-tumor response20 and as a result is not routinely considered a treatment option.

Abundant macrophages and CD45RO+memory T-cell infiltration were observed in the ST of this patient with irAE synovitis. Importantly, while these features recapitulated those of RA, we also observed a noticeable increase in TNFα expression and absence of IL-6. Critically, the predominance of TNFα suggested that a TNFα inhibitor would be a more suitable bDMARD therapy than tocilizumab. Two (of the three) RA patients’ ST biopsies were positive for IL6, serving as positive control. Interestingly, the RA patient whose biopsy had more significantly positive IL6 also had lymphoid (follicular) pattern of infiltrate, similar to our patient with irAE.

Histological examinaton of the ST also indicated an absence of PD-1 labeling in the nivolumab-treated irAE patient compared with the ST from patients with early RA, despite significant positive labeling for PDL-1 expressing cells (figure 2). PD-1 expressing T-cells are considered key players in classic chronic autoimmune diseases like RA.21 Interestingly, flow cytometry of the abundant T-cells in the inflammatory synovium demonstrated PD-1 labeling was markedly lower in CD45RO+memory T-cells from irAE compared with RA. We hypothesize that this is due to sustained occupancy of the PD-1 receptor by nivolumab, even at the time of arthroscopy (~200 days after cessasion). This is considerably longer than the reported serum half-life of 12–20 days for nivolumab,22 23 and while prolonged occupancy on CD4 +T cells has been reported for nivolumab,23 24 this is the first indication of prolonged occupancy of PD-1 on CD4 +T cells in inflamed tissues in rheumatic irAEs and may influence ST T-cell infiltration and chronic inflammation. Indeed, it is plausible that the half-life of ICIs varies in different tissues, which may underpin both their antitumor effect as well as explaining the variation seen in the time of onset of organ-specific irAE.25

Using arthroscopic synovial biopsy, we were able to target a bDMARD to the dominant inflammatory cytokine resulting in resolution of severe synovitis in this patient. The presence of PD-1 staining in memory T-cells in both the peripheral blood and SF compartments following infliximab treatment is interesting and may indicate the natural attrition of nivolumab from the surface of T-cells, or indeed, the emergence of a new non-occupied T-cell population. It is interesting to speculate what the implications may be, of the re-emergence of this population of cells with respect to possible relapse of his SCC (this has not occurred thus far, 30 months following cessation of Nivolumab). A recently published26 case report of a synovial biopsy of a patient with irAE presenting as asymmetric inflammatory oligoarthritis and enthesitis also revealed a lymphoid pattern of infiltration. This patient responded only partially to adalimumab (in combination with methotrexate and corticosteroids), presumably owing to the fact that treatment was not directed by cytokine/cellular dominance; cytokine staining was not performed and the authors raise the possibility as to whether the patient may have responded better had an IL-6R inhibitor may have been used as reported in the study in RA by Dennis et al.10 In contrast, we believe the marked benefit to infliximab in our patient was as a result of dominance of TNF in his ST that underpinned this benefit.

A limitation of our study was the lack of interrogation of ST using molecular methods; it would have been ideal to further evaluate TNF and IL6/IL6R pathways, in particular. In addition, the reasons for selecting the cytokine/cellular staining that we did was also pragmatic: biological disease-modifying therapies are available (in the context of the inflammatory arthritides) against TNF, T-cell co-stimulation, IL6 and B-cells (anti-CD20) and hence we wished to use the information to specifically direct therapeutic choice.

In conclusion, our data indicate absence of PD1 positivity despite cessation of nivolumab 200 days prior, indicating a sustained tissue response much longer than the serum half-life of this medication. We demonstrate that a synovial biopsy-guided therapy approach using ST biopsy to inform biologic therapy is feasible, results in rapid resolution of disease and provides important information about the immunopathogenesis of the process. This synovial biopsy-guided therapy could be reasonably extended to other irAEs and may improve response rates over empirical therapy. Future investigations on irAE patients should aim to collect cells for single cell RNAseq, which could yield cellular information about irAE initiation, confirm PD-1 mRNA expression in infiltrating T-cells, and identify patterns in T-cell clonality that may help elucidate the mechanisms of disease and ultimately improve clinical outcomes; it is also probable that delineation of inflammatory pathways by such methods in future analyses will be productive with regards to more accurately identifying involved pathways and directing therapeutic choice.