A phase 1/2 trial of an immune-modulatory vaccine against IDO/PD-L1 in combination with nivolumab in metastatic melanoma

Abstract

Anti-programmed death (PD)-1 (aPD1) therapy is an effective treatment for metastatic melanoma (MM); however, over 50% of patients progress due to resistance. We tested a first-in-class immune-modulatory vaccine (IO102/IO103) against indoleamine 2,3-dioxygenase (IDO) and PD ligand 1 (PD-L1), targeting immunosuppressive cells and tumor cells expressing IDO and/or PD-L1 (IDO/PD-L1), combined with nivolumab. Thirty aPD1 therapy-naive patients with MM were treated in a phase 1/2 study ( https://clinicaltrials.gov/ , NCT03047928). The primary endpoint was feasibility and safety; the systemic toxicity profile was comparable to that of nivolumab monotherapy. Secondary endpoints were efficacy and immunogenicity; an objective response rate (ORR) of 80% (confidence interval (CI), 62.7-90.5%) was reached, with 43% (CI, 27.4-60.8%) complete responses. After a median follow-up of 22.9 months, the median progression-free survival (PFS) was 26 months (CI, 15.4-69 months). Median overall survival (OS) was not reached. Vaccine-specific responses assessed in vitro were detected in the blood of >93% of patients during vaccination. Vaccine-reactive T cells comprised CD4⁺ and CD8⁺ T cells with activity against IDO- and PD-L1-expressing cancer and immune cells. T cell influx of peripherally expanded T cells into tumor sites was observed in responding patients, and general enrichment of IDO- and PD-L1-specific clones after treatment was documented. These clinical efficacy and favorable safety data support further validation in a larger randomized trial to confirm the clinical potential of this immunomodulating approach.

Fig. 1. Clinical response.

a, Pie charts with percent ORR, CR, PR and PD according to RECIST 1.1 by investigator review of all patients (n = 30), PD-L1⁺ patients (>1%, (n = 17)) and PD-L1⁻ patients (<1%, n = 13)), respectively. Two-sided CIs (95%) were constructed using the Clopper–Pearson method. b, Treatment effect in MM1636 compared with a matched historical control group from the DAMMED database (n = 74). Patients in MM1636 (n = 29) were matched with the exact same combination variable according to age (≤70 years, >70 years), sex, LDH levels (normal, elevated), M stage (M1a, M1b, M1c), BRAF status (wild type, mutated) and PD-L1 status (<1%, ≥1%). Estimates for treatment effects were calculated by weighted logistic regression analyses and weighted Cox proportional hazard model. Bar height indicates the estimated response rate; tops of bars are centers for error bars. Odds ratios (OR), response rates and their corresponding 95% CIs were extracted from the regression model. All P values were two sided, and P values below 0.05 were considered statistically significant. c, Best change in the sum of target lesion size compared with that at baseline (n = 30). The horizontal line at −30 shows the threshold for defining an objective response in the absence of non-target disease progression or new lesions according to RECIST 1.1. Two patients with 100% reduction in target lesion size had non-target lesions present. White stars, six patients had normalization (<10 mm) of fluor-18-deoxyglucose (FDG)-negative lymph nodes (at baseline, lymph nodes were >1.5 cm and FDG⁺) and 100% reduction of non-lymph node lesions and are considered to have had a CR (green bar). Black star, one patient (MM29) was considered to have had a PR (blue bar), although he had not reached a −30% change in target lesion size. The patient had a single measurable 13-mm lung metastasis at baseline and multiple biopsy-verified cutaneous metastases on the left crus (not detectable by positron-emission tomography–computed tomography (PET–CT) at baseline). The best change in target lesion size was 10 mm, and a post-treatment biopsy from the cutaneous metastases showed no sign of malignancy. Thus, overall, the patient was classified as having a PR. d, Kaplan–Meier curve of the response duration in the 24 patients with an objective response. e, Kaplan–Meier curve of PFS in all 30 treated patients. f, Kaplan–Meier curve of OS in all 30 treated patients.

Fig. 2. Clinical response.

a, Swimmer plot showing response duration and time to response according to RECIST 1.1 for all treated patients (n = 30). Triangles indicate first evidence of PR, while squares indicate first evidence of CR. Closed circles indicate time of progression. Arrows indicate ongoing responses. Patient MM18 died due to nivolumab-induced side effects. b, Spider plot showing response kinetics in all treated patients (n = 30). Red squares indicate time of progression. c, PET–CT images of patient MM42 before and after treatment (after 12 series of treatment) showing FDG metabolism in target lesions. SC, subcutaneous.

Fig. 3. Vaccine-specific responses in blood.

a, IDO- and PD-L1-specific T cell responses in PBMCs at baseline and on vaccination as measured by the IFN-γ ELISPOT assay (n = 30). *Responses were calculated as the difference between the average numbers of spots in wells stimulated with IDO or PD-L1 peptide (triplicates) and those from the corresponding control (DMSO), and statistical analyses of ELISPOT responses were performed using a distribution-free resampling method (Moodie et al.). DR (double response), response was not statistically confirmed due to replicate number, but the number of spots in peptide wells was two times higher than that in control wells (DMSO). NS, no significant response and no DR. For a detailed overview of responses at serial time points on vaccination, see Extended Data Fig. 4a. b, IDO- and PD-L1-specific T cell response in PBMCs in all treated patients measured by the IFN-γ ELISPOT assay at baseline and on vaccination. On-vaccination responses were selected from the ‘best’ ELISPOT response at different time points for each patient (series 3, 6, 12, 18 or 24) during vaccination (n = 30). Wilcoxon matched-pairs signed-rank test was used to compare responses to IDO or PD-L1 peptides in the vaccine between baseline and later time points. c, Representative example of ELISPOT wells with response in patient MM23 in serial PBMCs before and on treatment. d, IDO-specific CD4⁺ and CD8⁺ T cells were isolated and expanded from PBMCs stimulated in vitro with the IDO peptide and a low dose of IL-2 for 14–15 d before sorting using the Miltenyi Cytokine Secretion Assay—Cell Enrichment and Detection kit. To assess their cytolytic potential, IDO-specific T cells were stimulated with IDO peptide, and expression of CD107a, IFN-γ and TNF-α was assessed by flow cytometry (the example is from patient MM14).

Fig. 4. PD-L1- and IDO-specific T cells from vaccinated patients react against PD-L1- and IDO-expressing target cells.

a, Left, PD-L1-specific T cell culture (MM1636.05) reactivity against PD-L1 peptide or autologous tumor cells in the IFN-γ ELISPOT assay. Tumor cells were either not treated or pretreated with 200 U ml⁻¹ IFN-γ for 48 h before the assay. Effector:target (E:T) ratio of 10:1 was used. Right, PD-L1 and HLA-II surface expression on melanoma cells with (green) or without (yellow) pretreatment with IFN-γ compared to an isotype control (gray) as assessed by flow cytometry. b, Left, PD-L1-specific T cell (MM1636.05) reactivity in the IFN-γ ELISPOT assay against autologous tumor cells pretreated with IFN-γ (500 U ml⁻¹) and transfected with mock or PD-L1 small interfering (si)RNA 24 h after transfection. E:T ratio, 10:1. Right, PD-L1 surface expression on melanoma tumor cells (MM1636.05) assessed by flow cytometry 24 h after transfection with mock (blue) or PD-L1 (red) siRNA compared to the isotype control (gray). c, Reactivity of the CD4⁺ PD-L1-specific T cell clone (MM1636.14) against PD-L1 peptide or autologous CD14⁺ cells; E:T ratio, 10:1. CD14⁺ cells were isolated using magnetic bead sorting and used as targets in an ELISPOT assay directly or after pretreatment for 2 d with TCM derived from the autologous tumor cell line. d, Quantitative PCR with reverse transcription (RT–qPCR) analysis of PD-L1 (CD274) expression in sorted CD14⁺ cells before and after treatment with autologous TCM for 48 h. e, Reactivity of the IDO-specific CD4⁺ T cell clone (MM1636.23) against IDO peptide combined with HLA-DR (L243)-, HLA-DQ (SPV-L3)- or HLA-DP (B7/21)-blocking antibodies (aHLA-DR, aHLA-DQ and aHLA-DP) in an intracellular staining assay (ICS) for IFN-γ and TNF-α production. T cells were incubated with individual blocking antibodies (2 μg ml⁻¹) for 30 min before adding IDO peptide. f, Reactivity of the IDO-specific CD4⁺ T cell clone (MM1636.23) against the HLA-DR-matched IDO-expressing cell line MonoMac1 transfected with mock or IDO siRNA in an ICS assay, E.T ratio, 4:1. siRNA transfection was performed 48 h before the experiment. g, RT–qPCR analysis of IDO1 expression in MonoMac1 cells 48 h after siRNA transfection. h, Reactivity of the CD4⁺ IDO-specific T cell clone (MM1636.14) against IDO peptide or autologous CD14⁺ cells; E:T ratio, 20:1. CD14⁺ cells were isolated using magnetic bead sorting and used as targets in an ELISPOT assay directly or after pretreatment with TCM derived from the autologous tumor cell line. i, RT–qPCR analysis of IDO1 expression in sorted CD14⁺ cells before and after treatment with autologous TCM for 48 h. Bars in RT–qPCR data (d,g,i) represent the mean of three (d,i) or six (g) technical replicates ±s.d.; P values were determined by two-tailed parametric t-tests. ELISPOT counts (a,b,c,h) represent the mean value of three technical replicates ±s.e.m.; response P values were determined using the distribution-free resampling (DFR) method. TNTC, too numerous to count.

Fig. 5. Changes in the TME after treatment. Number of CD3⁺ and CD8⁺ T cells, TCR fraction, TCR clonality, TCR repertoire, biopsy expanded TCR clones and enrichment of IDO- and PD-L1-specific T cells at the tumor site.

a, Number of CD3⁺ and CD8⁺ T cells per mm² (sum in the validated area) at the tumor site detected by IHC of paired biopsies from four patients. b, Example of IHC of CD3⁺ and CD8⁺ T cells at the tumor site before and after treatment (cells per mm²) in one patient (MM01). c, T cell fraction at the tumor site at baseline and cycle 6 by TCR sequencing. The T cell fraction was calculated by taking the total number of T cell templates and dividing by the total number of nucleated cells. d, Tracking of vaccine-associated clones at baseline and cycle 6 in tumor biopsies. Cumulative frequencies of IDO and PD-L1 vaccine-specific TCR rearrangements are represented. e,f, TCR clonality and TCR repertoire richness in five patients at the tumor site at baseline and cycle 6. Simpson clonality measures how evenly TCR sequences are distributed among a set of T cells, where 0 indicates an even distribution of frequencies and 1 indicates an asymmetric distribution in which a few clones dominate. TCR repertoire richness reports the mean number of unique rearrangements. g, Bar chart representing baseline expanded biopsy clones from five patients (colored bars) and the detection of biopsy expanded clones also found in the blood at baseline and series 3, 6 and 12 (white and gray bars) by TCR sequencing.

Extended Data Fig. 1. Mode of Action and treatment plan.

a) Anticipated mechanism of action of the combination therapy of an IDO/PD-L1 derived peptide vaccine and nivolumab (anti-PD1). 1) The IDO/PD-L1 peptide vaccine is administered subcutaneously (s.c) and nivolumab is administered intravenous (IV). 2) The vaccines peptides are phagocytosed by an antigen presenting cell and presented to IDO and PD-L1 specific T cells, which are activated. 3) The activated T cells migrates to the tumor site where they attack both immune-suppressive cells and tumour cells expressing IDO and/or PD-L1 leading to cytokine production and a pro-inflammatory tumour microenvironment. 4/5) Enhanced tumour killing by both IDO/PD-L1 specific T cells and tumor specific cytotoxic T cells due to PD-1 blockade. Created with BioRender.com b) Treatment plan. After written informed content patients were screened. Before treatment start a baseline PET/CT scan was performed, baseline blood sample for research use and if assessable a baseline needle biopsy. Patients were treated with the IDO/PD-L1 peptide vaccine subcutaneously biweekly for the first 6 injections and thereafter every fourth week for a maximum of 15 vaccines. Nivolumab was administered in parallel biweekly (3mg/kg) up to 24 series. If patients needed subsequent nivolumab after ended vaccination regiment they were treated with 6 mg/kg every fourth week up to two years. Needle biopsy and delayed type hypersensitivity (DTH) was performed after 6 series of treatment if assessable. PET/CT scans was performed every third month.

Extended Data Fig. 2. Progression free survival and overall survival in matched historical control group and vaccine injection site reaction.

a) Kaplan-Meier curve of progression free survival with corresponding 95% confidence intervals in the matched historical control group (n=74). Patients were matched on BRAF-status, PD-L1-status, age, gender, M-stage and LDH level). b) Kaplan-Meier curve of overall survival with corresponding 95% confidence intervals in the matched historical control group (n=74). c) Images of injection site reaction in patient MM20 (CR) after 11 vaccination show redness, rash and granuloma at injection site.

Extended Data Fig. 3. Vaccine specific responses in blood.

a) Heatmap of specific (background has been subtracted) IDO and PD-L1 responses in PBMCs at baseline, series 3, 6, 12, 18 and 24 measured by in vitro IFN-γ Elispot assay show fluctuations in the blood during treatment (n=30). b) Heatmap of specific (background has been subtracted) IDO and PD-L1 responses in PBMCs at baseline and 3 and 6 months after last vaccine measured by in vitro IFN- γ Elispot assay (n=30). 2.5–3.2x10⁵ cells per well were used. c) Vaccine associated clones were tracked in the blood by summing the frequency of each rearrangement enriched in either IDO or PD-L1 T cells.

Extended Data Fig. 4. Ex vivo vaccine specific responses in blood.

Heatmaps of detected specific (background has been subtracted) IDO (a) and PD-L1 (b) responses in PBMCs at baseline and on/after treatment as measured by IFNγ ELISPOT (n=25). C) Example of well images of ex vivo ELISPOT wells for three different patients (n=3). 6-9x10⁵ cells per well were used.

Extended Data Fig. 5. CD4 and CD8 vaccine specific T cell responses in blood.

Top: heatmaps of IDO specific CD4 (left) and CD8 (right) T cell responses in PBMCs at baseline and on/post treatment. Bottom: heatmaps of PD-L1 specific CD4 (left) and CD8 (right) T cell responses. Responses quantified by flow cytometry by an increased expression of CD107α, CD137 and TNFα after 8h peptide stimulation. Values represent specific responses after background values have been subtracted (n=21). Statistical analysis were performed using two-sided Wilcoxon matched-paired rank test.

Extended Data Fig. 6. Pro-inflammatory profiles of sorted CD4 and CD8 T cells from blood.

a/b/c/d) Percentage of in vitro stimulated and sorted PD-L1 and IDO specific CD4 and CD8 T cells secreting cytokines.

Extended Data Fig. 7. Vaccine specific responses in skin.

a) IDO and PD-L1 specific T cell responses in SKILs after 6 series of treatment measured by IFN-γ Elispot assay (n=13). SKILs were grown from DTH injection with either IDO peptide, PD-L1 peptide or a mix as presented by different blue colours. * Responses were calculated as the difference between average numbers of spots in wells stimulated with IDO or PD-L1 peptide (triplicates) and corresponding control (DMSO) and statistical analyses of Elispot responses were performed using distribution-free resampling method (Moodie et al.). DR: Not statistically confirmed response due to replicate number but number of spots in peptide wells are two times higher than control wells (DMSO). NS: No significant response and no DR. b/c/d) Percentage of cytokine secreting/CD107a+ CD4+ and CD8+ IDO and PD-L1 specific T cells in response to in vitro peptide stimulation by flow cytometry. e) Skin infiltrating PD-L1 specific T cell clones also found in biopsy in patient MM01. TCR sequencing was performed on a PD-L1 specific T cell culture generated from DTH area on the lower back injected with PD-L1 peptide on patient MM01. Bars show the frequency of top 25 clones in the culture with which indicates a high Simpson clonality of 0.43. f) Tracking the frequency of the top five skin infiltrating PD-L1 specific clones in tumour before and after treatment.

Extended Data Fig. 8. T cell changes in blood after treatment. T cell fraction, TCR clonality and repertoire richness in blood and peripherally expanded TCR clones in both responding and non-responding patients.

a) T cell fraction in peripheral blood of five patients at baseline, series 3, 6 and 12 by TCR sequencing. T cell fraction was calculated by taking the total number of T cell templates and dividing by the total number of nucleated cells (n=5.) b) TCR clonality in peripheral blood of five patients at baseline, series 3, 6 and 12 by TCR sequencing (n=5). Simpson clonality measures how evenly TCR sequences are distributed amongst a set of T cells where 0 indicate even distribution of frequencies and 1 indicate an asymmetric distribution where a few clones dominate. c) TCR repertoire richness in peripheral blood of five patients at baseline, series 3, 6 and 12 by TCR sequencing (n=5). TCR repertoire richness report the mean number of unique rearrangements. d) Bar-chart chart representing dynamics of the expanded T cell clones. Coloured bars represent peripherally expanded clones in five patients at series 3, series 6 and series 12 as compared to the baseline PBMC samples (n=5). Light gray bar represents peripherally expanded clones that are present in baseline biopsy samples while dark gray bar represents peripherally expanded clones that are present in the post-treatment biopsy (after 6 series). e) Frequency of the dominant TCR β chain in clonal PD-L1 and IDO specific cultures as determined by CDR3 sequencing.

Extended Data Fig. 9. Profiling of genes relevant to T cell activation, cytokines and exhaustion markers on pre- and post-treatment biopsies in two patients.

a) RNA expression profiling of genes related to T cell activation was performed using NanoString nCounter (n=2). b) RNA expression profiling of genes related to cytokine activity was performed using NanoString nCounter (n=2). c) RNA expression profiling of genes related to checkpoint inhibitors was performed using NanoString nCounter (n=2).

Extended Data Fig. 10. Treatment induced upregulation of PD-L1, IDO, MHC I and MHC II and distance between CD8 T cells and PD-L1 expressing cells.

a) IHC on 4 paired biopsies stained for CD3 and CD8 T cells, PD-L1, MHC I and MHC II on tumor cells/mm² (sum in the validated area) and IDO H-score (from 0 to 300). H-score is the expression of IDO on both immune and tumor cells: The score is obtained by the formula: 3 x percentage of cells with a strong staining + 2 x percentage of cells with a moderate staining + percentage of cells with a weak staining (n=4). b) Distance in µm between CD8+ T cells and PD-L1+ stained cells on five baseline biopsies detected by IHC (n=5).