Uveal melanoma (UM) comprises < 3% of all melanomas with an incidence of 5–10 cases/million [1] and underlying biology that is distinct from cutaneous melanoma (CM). In the last decade, the interrogation of the genetic landscape [2] and advances in immuno-oncology [3] have led to a remarkable improved survival rate of 40–60% [4] in patients with metastatic CM. In contrast, patients with UM rarely (ORR 0–2.6%) [5, 6] respond to ICIs, including anti-CTLA-4 and anti-PD1 monotherapies, and show low response rate (15.8%) to the combination [7]. Intrinsic resistance to ICIs in UM may be related to various mechanisms, including a low somatic mutation rate [8] and paucity of tumor infiltrating lymphocytes [9]. In CM, ICI-related skin toxicities, such as rash and vitiligo, correlate with increased tumor response and prolonged survival [10]. The immunological underpinnings for this phenomenon in patients remain poorly understood. Delineating underlying mechanisms may help to identify approaches for dissociating treatment benefits and risks.

Here we report a patient with metastatic UM who experienced an exceptional response to dual blockade of PD-1 and CTLA-4. This response was accompanied by severe and unique immune-related adverse events (irAEs). Integrated analysis of several tissues, including primary tumor, a liver metastasis, inflamed duodenum and peripheral blood using whole-exome, transcriptome and T cell receptor (TCR) sequencing, and multiplexed immunofluorescence identified a dominant T cell clone. This report suggests that tumor-reactive T cell clones may play a role in mediating toxicity in healthy tissues.

A 60-year-old woman was diagnosed with 18 × 14 mm UM of the right eye and underwent enucleation in 2009. Pathology confirmed UM with monosomy 3 and 8q amplification. She developed a solitary hepatic metastasis in 2014 and underwent right hepatectomy. A multi-gene panel analysis of the tumor showed somatic BAP-1 and GNA11 mutations. She developed extensive metastases 9 months later with multiple hepatic, bone and lung lesions, and elevation of lactate dehydrogenase (LDH) > 1300 U/L. She received combination nivolumab and ipilimumab therapy. After two infusions, she developed central serous retinopathy of the left eye with retinal detachment, tinnitus and vitiligo resembling Vogt-Koyanagi-Harada (VKH) disease, an ocular autoimmune syndrome (Fig. 1c). CT scan at 12 weeks demonstrated significant reduction in hepatic metastases (Fig. 1a and b), and disappearance of lung and bone metastases. LDH level initially rose and then normalized (Fig. 1f). She continued on nivolumab monotherapy and experienced a near-complete response, but developed grade 3 duodenitis (Fig. 1d and e) requiring prolonged high-dose immunosuppressive therapy, including high-dose prednisone, followed by infliximab, and vedolizumab with eventual resolution. The clinical antitumor response persisted for over 1 year from treatment initiation and over 9 months from the last dose of immunotherapy. Unfortunately, she developed progressive brain and liver metastases after 1.5 year. Nivolumab monotherapy was resumed resulting in a mixed response and additional skin and eye toxicity, preventing further treatment. Due to overall declining health, the patient decided for supportive care and died 6 months after reinitiating original systemic therapy.Fig1Fig. 1

Clinical Characteristics. Panel a and b depict pre- and post-treatment computed tomography of the liver with complete resolution of liver metastases. Panel c depicts central serous retinopathy (arrow) on fundoscopic examination and Optical Coherence Tomography (OCT). Panel d and e shows endoscopic and pathologic findings of post-treatment duodenitis (arrow) with marked acute inflammatory cell infiltrate involving most of the glandular epithelium. The infiltrate is predominantly within deep crypt spaces (arrowhead). Panel f shows decline of serum LDH shortly after immunotherapy were initiated

This exceptional response of the patient’s UM to combined ICI illustrates the potential critical role of clonal expansion of TILs in achieving clinical benefit from immunotherapy. That the same CD8+ TIL clones in the primary tumor expanded in the metastases suggests that these TILs were directed against the same antigen on the tumor. Detection of this clone in PBMC and within the duodenum, a site of irAE, strongly suggests that the antitumor effects and toxicity were mediated by the same T cell clone. The involvement of barrier tissues (skin, gastrointestinal mucosa) by irAEs beyond drug exposure supports the possibility that adverse effects may be mediated by tissue-resident memory T cells (TRM) [22] Further, the presence of a specific TCR clone in the primary eye tumor, liver metastasis and site of enteritis and the occurrence of VKH and vitiligo suggest that the clone was directed against a shared antigen/epitope rather than a tumor neoantigen. In the absence of neoantigens that strongly bind class-I MHC, the irAEs may stem from enhanced TCR interactions with self-antigens arising from ICI therapy.

When anti-tumor reactivity is directed against a shared antigen the therapeutic index may be narrow, with toxicity occurring nearly simultaneously with antitumor response, as in this case. This observation has been recently reported in another case of metastatic UM that achieved a durable major response to the combination ICI in association with autoimmune hepatitis, uveitis and exocrine pancreatic insufficiency [23].

While activation of quiescent immune-reactive T cells resident in non-tumor tissue is purported to be the mechanism of autoimmune thyroiditis, colitis, and other typical toxicities of ICI, the irAEs in both cases of advanced UM suggest that there was cross-reactivity of T cell immunity against an antigen that was shared between the UM and the contralateral uvea, the skin and non-tumor organs. That the duodenitis was unresponsive to steroids and infliximab, but did respond promptly to vedolizumab [24], a monoclonal antibody to integrin α4β7 (LPAM-1), which blocks T cell infiltration into the GI tract, suggests that her GI toxicity derived from circulating T cells that trafficked into the GI tract following expansion in a distant site, such as the tumor. A similar pathophysiology may occur in the cases of extreme cardiotoxicity, where expanded T cell clones that were present in the tumor were also identified in the inflamed heart musculature [25].

Thus, the value of a high mutational load may be to induce immunoreactivity against multiple neoantigens and simultaneously overwhelm any shared antigen response, thereby eliminating the tumor before significant shared antigen expansion and related organ toxicity develops. Such biology might explain the enhanced durable antitumor benefit observed in patients with irAEs, despite the use of often-prolonged immunosuppressive therapy.

Ultimately, it may be possible, albeit difficult, to identify at baseline patients with expanded immunity against antigens shared by the tumor and vital host organs, so that one could take measures to minimize the organ toxicity. Alternatively one could use pre-treatment vaccination against neo-epitopes in order the steer the immune response towards tumor restricted rather than shared antigens.

This report emphasizes that while the response to ICI is significantly lower in patients with UM compared to those with CM, a major response can be achieved. Additional work is required to select the appropriate patients with UM who are destined to benefit from ICI and to determine a means of dissociating anti-tumor responses against shared antigens from toxicity to organs expressing the same antigens.