Patients with melanoma treated with ICI at the Oncology division of the University Hospital in Basel were invited to participate in this cohort study. Inclusion criteria for cohort 1 were: monotherapy with PD-1 blocking antibody (pembrolizumab or nivolumab), for cohort 2 combination ICI treatment with PD-1 and CTLA-4 blocking antibodies (nivolumab and ipilimumab). Clinical data was collected from the electronic hospital data base. Severe irAEs were defined as AEs that led to permanent discontinuation of ICI therapy, all other irAEs were deemed “mild”.

Ethics approval was obtained from the local ethical committee to analyze blood samples (Northwest and Central Switzerland Ethics Committee EK321/10). Peripheral blood mononuclear cells (PBMCs) were collected, isolated by a Ficoll density gradient and frozen for later analysis. Blood was obtained before the start of the ICB and 4-6 weeks after start of ICB. Isolated PBMCs were cryopreserved by adding 10% DMSO containing fetal calf serum.

We established and validated flow cytometry panel (surface marker: CD45RO, CD14, CD11c, CD127, CD11b, CD335, CD25, HLA-DR, CD8, CD16, CD56, CD3, live/dead, CD123, CD19, CD4, PD-1, and intracellular marker: FoxP3, Ki-67). A single-stain control was made using VersaComp Antibody Capture Beads for each fluorochrome (Beckman Coulter, USA). Cells were heated at 95 degrees for 5min for single-stain control of dead cells. Dead cells were stained with live/dead blue. Fluorophore-conjugated antibodies were diluted in FACS buffer (PBS, 2% FBS, 0.5 mM EDTA). Patient samples were thawed at 37 degrees and washed with FACS buffer. All cells were incubated for 20min on ice with Live/dead Blue in PBS, followed by 20min incubation with anti-human Fc block (Invitrogen, USA) to prevent unspecific antibody binding. Afterwards the cell suspension was stained with a surface fluorophore-conjugated antibody solution for 30min on ice. After incubation cells were treated with a fix/perm solution (eBioscience, USA) before staining with the an intracellular antibody mix. Cells were resuspended in FACS buffer and fluorescence measured with a Cytek Aurora machine (Cytek Biosciences Inc., USA). The acquired data was analyzed using FlowJo 10.3 (TreeStar Inc, USA) after singlet- (FSC-A vs. FSC-H and SSC-A vs. SSC-W) and live/dead cell discrimination (gating strategy: Supplementary Figures 1, 2).

Patient characteristics are presented in a descriptive manner. For the statistical analysis of the acquired Aurora data GraphPad Prism was used (Version 9). To compare the two patient groups (severe and mild irAEs) we used a non-parametric T test (Mann-Whitney t-test, p < 0.05 was considered as statistically significant).

We collected blood samples of two cohorts of patients. In the first cohort, we included melanoma patients receiving PD-1 blocking monotherapy with nivolumab or pembrolizumab (Table 1). In the second cohort, we included melanoma patients receiving dual immune checkpoint blockade with nivolumab and ipilimumab targeting PD-1 and CTLA-4 (Table 2). In each of the two cohorts we divided the patients into patients that developed severe irAEs leading to cessation of the ted ICI therapy and patients with only mild side effects. In the monotherapy group, we observed 3 patients with irAEs that led to the discontinuation of the ICI therapy mainly due to the type of irAE including myocarditis, colitis and arthritis (Table 1). In cohort 2, six patients had to stop the treatment with dual ICI therapy with ipilimumab and nivolumab. Colitis and hepatitis were the most prevalent irAE observed in this cohort that led to discontinuation (Table 2). In general, patients receiving monotherapy had less advanced disease and were often treated in an adjuvant setting (Table 1). Patients in the cohort receiving dual ICI therapy had more often advanced stage IV disease (Table 2). No direct association was seen between relapse and irAE (Table 1). This is probably due to the low number of patients involved in this correlative study.

To analyze the immune cell composition of patients with mild and severe irAEs after ICB, we stained PBMCs before treatment start and at time of irAE development and analyzed them by flow cytometry. In patients receiving PD-1 blocking monotherapy, no differences in CD3⁺ T cells were observed between patients with mild or severe irAEs (Figure 1A). In addition, no or only minimal changes were seen in T cell subsets including CD4⁺, CD8⁺ T cells, CD56⁺ T cells and FoxP3⁺ regulatory T cells (Figures 1B–E). CD56⁺ T cells were slightly increased in patients with only mild side effects before starting ICI therapy (Figure 1C). No differences were observed over time (time point 1 before the start and time point 2 taken 4-6 weeks after start of ICI therapy). Further characterization of T cell subsets showed no significant difference in both CD4 and CD8 compartments between patients with severe irAEs or mild irAEs (Figures 2A–H). In addition, no differences were observed in peripheral C11c⁺ dendritic cells, CD123⁺ plasmacytoid dendritic cells, CD19⁺ B cells, or CD14⁺ myeloid cells (Supplementary Figure 3). Moreover, no differential expression of Ki67, CD25, and PD-1 was demonstrated (Supplementary Figures 4, 5). Taken together, no clear differences of T cell subsets in the peripheral blood of patients treated with PD-1 blocking monotherapy were found.

During PD-1 blocking monotherapy, only few patients developed severe irAEs. In contrast, patients treated with dual ICI therapy develop severe irAEs more often. Therefore, we further analyzed immune cell changes in the peripheral blood of these patients, comparing severe irAEs that led to the discontinuation of dual ICI therapy with mild irAE patients. No changes in percentage of T cells were observed in the two groups and also over time (Figure 3A). However, a decreased frequency of CD4⁺ T cells and Increased frequency of CD56⁺ T cells and CD8+ T cells before and after the start of the treatment was found (Figures 3B–D). No changes and differences were observed in FoxP3⁺ regulatory T cells (Figure 3E). Additional analysis of T cell subsets in this cohort of patients under dual ICI therapy showed no changes (Figures 4A–H), except for the CD8⁺ effector memory cells re-expressing CD45RA (EMRA, Figure 4E). An increased frequency of EMRA CD8⁺ T cells was seen before and after start of dual ICB with PD-1 and CTLA-4 blocking antibodies (Figure 4E). No major changes were seen between patients with mild and severe irAEs in other cell types (Supplementary Figure 3). We observed an increase in Ki67⁺ CD8 T cells after starting the treatment in patients experiencing severe irAE (Supplementary Figures 4, 5).

In patients with dual ICI therapy targeting PD-1 and CTLA-4, changes in peripheral T cells in patients with severe irAE could be detected that where already present before start of the treatment and remained during the course of treatment.