Translational randomized phase II trial of cabozantinib in combination with nivolumab in advanced, recurrent, or metastatic endometrial cancer

Abstract

Background: Combining immunotherapy and antiangiogenic agents is a promising treatment strategy in endometrial cancer. To date, no biomarkers for response have been identified and data on post-immunotherapy progression are lacking. We explored the combination of a checkpoint inhibitor (nivolumab) and an antiangiogenic agent (cabozantinib) in immunotherapy-naïve endometrial cancer and in patients whose disease progressed on previous immunotherapy with baseline biopsy for immune profiling.

Patients and methods: In this phase II trial (ClinicalTrials.gov NCT03367741, registered December 11, 2017), women with recurrent endometrial cancer were randomized 2:1 to nivolumab with cabozantinib (Arm A) or nivolumab alone (Arm B). The primary endpoint was Response Evaluation Criteria in Solid Tumors-defined progression-free survival (PFS). Patients with carcinosarcoma or prior immune checkpoint inhibitor received combination treatment (Arm C). Baseline biopsy and serial peripheral blood mononuclear cell (PBMC) samples were analyzed and associations between patient outcome and immune data from cytometry by time of flight (CyTOF) and PBMCs were explored.

Results: Median PFS was 5.3 (90% CI 3.5 to 9.2) months in Arm A (n=36) and 1.9 (90% CI 1.6 to 3.4) months in Arm B (n=18) (HR=0.59, 90% CI 0.35 to 0.98; log-rank p=0.09, meeting the prespecified statistical significance criteria). The most common treatment-related adverse events in Arm A were diarrhea (50%) and elevated liver enzymes (aspartate aminotransferase 47%, alanine aminotransferase 42%). In-depth baseline CyTOF analysis across treatment arms (n=40) identified 35 immune-cell subsets. Among immunotherapy-pretreated patients in Arm C, non-progressors had significantly higher proportions of activated tissue-resident (CD103+CD69+) ɣδ T cells than progressors (adjusted p=0.009).

Conclusions: Adding cabozantinib to nivolumab significantly improved outcomes in heavily pretreated endometrial cancer. A subgroup of immunotherapy-pretreated patients identified by baseline immune profile and potentially benefiting from combination with antiangiogenics requires further investigation.

Keywords: biomarkers; clinical trials; combination; drug therapy; female; genital neoplasms; immunotherapy; phase II as topic; tumor.

Conflict of interest statement

Competing interests: SL has received honoraria from AstraZeneca, Merck, Eisai, GSK, and Roche. PAK has participated in Advisory Boards/Scientific Advisory Committees for Alkermes, AstraZeneca, Bayer, GSK, Merck, Pfizer, Tesaro, Vertex, and Repare; and has received institutional funding as Principal Investigator from AstraZeneca, Bayer, Eli Lilly, GSK, Merck, Merck KGaA, Pfizer, and Tesaro/GSK. BXW has no conflicts of interest related to this manuscript; financial disclosures that are not related: he has received honoraria from Tessa Therapeutics and AstraZeneca. MSB has no conflicts of interest related to this manuscript; financial disclosures that are not related: he has received institutional research support from Merck, Transgene, Pharmacyclics, Immune Design, Bristol Myers Squibb, Marker Therapeutics, Sorrento, Viewpoint Molecular Targeting, and Genentech; and is an Advisory Board member (unpaid) for TILT Biotherapeutics, Viewpoint Molecular Targeting, and Sorrento. SLG has received personal fees from AstraZeneca, Immunogen, Sermonix, Elvar Therapeutics, and GSK; and has received grants from AstraZeneca, AbbVie, Pfizer, Rigel, Iovance, Tesaro, Genentech/Roche, PharmaMar, and GSK; and has patents for Sermonix (US patent no. 10,905,659 and 10,258.604). FJB has participated in Advisory Boards for Merck, Eisai, and Agenus; and has received research funding from Eisai, Clovis, ImmunoGen, Merck, and Beigene (all outside the submitted work). TLW has no significant conflicts of interest related to this manuscript; financial disclosures that are not related: she has received research support to the institution for clinical trials from AbbVie, AstraZeneca, Clovis Oncology, Mersana, Mirati, Novartis, Roche Genentech, and Tesaro-GSK. LD has received personal fees from AstraZeneca, Genentech/Roche, MorphoTek, Merck, Inovio, Advance Medical, UpToDate, Cue Biopharma, British Journal of Obstetrics and Gynaecology, Parexel, State of California, Elsevier, ASCO, Expert review, ClearView Heath Care, National Cancer Institute, and JB Learning; and has received grants from Genentech/Roche, Cerulean/NextGen, AbbVie, Tesaro, Pfizer, GSK/Novartis, Morab, MorphoTek, Merck, Aduro BioTech, Syndax, Ludwig, LEAP Therapeutics, Eisai, Lycera, Inovio, and Advaxis; she reports other disclosures from Merck, GSK/Novartis and Genentech/Roche. IC has received travel grants from Tesaro; and is an advisor for AstraZeneca and GSK. PSO has no conflicts of interest related to this manuscript; financial disclosures that are not related to the current work: EMD Serono, Symphogen, Providence, and Tessa Therapeutics. GFF participates in an Advisory Board for GSK; has received honoraria from UpToDate; has received reviewer compensation from Journal of Clinical Oncology and Lancet Oncology; and has received payments to institution for clinical trial conduct from Roche, Syros, GSK, Iovance, Sermonix, Comugen, Cellex, Corcept, and Plexxikon. No disclosures were reported by the other authors.

© Author(s) (or their employer(s)) 2022. Re-use permitted under CC BY. Published by BMJ.

Figures

Figure 1

Patient flow. IO, immuno-oncology.

Figure 2

Clinical outcomes. (A) Progression-free survival in Arms A and B. (B) Best response and treatment duration by patient in Arms A and B. (C) Best response and treatment duration in Arm C (carcinosarcoma cohort and post-IO cohort). Carcino, carcinosarcoma; IO, immuno-oncology; MS, microsatellite; PFS, progression-free survival.

Figure 3

Overview of immune cell populations present in baseline biopsies. CD45+EpCAM– cells from all 40 patients (16 from Arm A (two with MSI-H tumors), 8 from Arm B, and 16 from Arm C (two with MSI-H tumors and five with carcinosarcoma) were pooled for unsupervised clustering using phenograph. (A) UMAP visualization shows the distribution of 35 phenograph-defined clusters across the major immune cell populations: CD4+ and CD8+ T cells, ɣδ T cells, regulatory T cells (Tregs), B cells, ILCs/NK cells, and myeloid subsets, as shown by the expression of lineage markers. The 35 phenograph-defined clusters can be separated into 7 CD4+ T-cell clusters (including one Treg cluster), 10 CD8+ T-cell clusters, two ɣδ T-cell clusters, seven B-cell clusters, three ILC/NK cell clusters, and six myeloid clusters. Each cluster has a unique pattern of marker expression, including markers of immune activation and suppression, cellular adhesion, trafficking, and proliferation. (B) UMAPs show the mean signal intensity and high-dimensional localization of the markers used to define the major immune cell populations. (C) Single-cell heatmap shows the (hierarchical) marker expression profiles that define each of the 35 unique immune cell clusters, grouped by immune cell subsets. (D) The baseline immune composition as a proportion of CD45+ EpCAM– cells is shown for each patient across all treatment Arms. Patients are ordered by best response, MSI status, histology, and study identifier. All major immune populations are present in varying proportions in the baseline biopsies of most patients. CTLA-4, cytotoxic T-lymphocyte antigen-4; EpCAM, epithelial cell-adhesion molecule; HLA, human leukocyte antigen; ILC, innate lymphoid cell; IO, immuno-oncology; MSI, microsatellite instable; MSI-H, microsatellite instable-high; NK, natural killer; PD, progressive disease; PD-1, programmed cell death 1; PD-L1, programmed cell death ligand-1; PR, partial response; SD, stable disease; TCR, T-cell receptor; UMAP, uniform manifold approximation and projection.

Figure 4

Comparison of baseline biopsies from non-progressors (best response, PR/SD ≥3 months) and progressors (best response, PD/SD phenograph-defined immune cell clusters present in the baseline biopsies of non-progressors and progressors in Arm A (n=7 per group) and Arm C (n=5 per group). (B) Bar graphs show the differential abundance of each immune cell cluster between non-progressors and progressors within Arm A and Arm C. Data are presented as the log fold-change (logFC). LogFC >0 indicates more abundant in non-progressors; logFC <0 indicates less abundant in non-progressors. In Arm A (upper panel), there was no statistically significant difference between non-progressors and progressors, although non-progressors showed a non-significant trend toward a higher proportion (>2 logFC) of cluster 3 ɣδ T cells (3865 cells). In Arm C (lower panel), prior-IO non-progressors had a significantly higher proportion (6 logFC; adjusted p=0.009) of the same cluster 3 ɣδ T cells vs progressors. Cluster 30 CD45RA+CD27+ CD8 T cells and cluster 32 CD45RA+CD27+ CD4 T cells were also significantly higher in non-progressors than in progressors; however, we detected only 18 and 30 cells in each cluster, respectively. There were no significant differences between progressors and non-progressors in the proportion of cluster 26 ɣδ T cells. Tissue-resident (CD103+CD69+) GzmBlow CD8 T cells (cluster 13; 3894 cells) and CD11c+CD31+ myeloid cells (cluster 33; 149 cells) were significantly more abundant in progressors than non-progressors. (C) Bar graph showing the differential abundance of each immune-cell cluster between Arm A (n=14) and Arm C (n=10). Data are presented as logFC. LogFC >0 indicates more abundant in Arm A; logFC <0 indicates less abundant in Arm A. Several CD45RA+ clusters (clusters 28, 32, 24, and 27) were significantly more abundant in IO-naïve (Arm A) than IO-pretreated (Arm C) patients, whereas activated, tissue-resident (CD103+CD69+) cluster 3 ɣδ T cells were significantly more abundant in patients whose disease progressed on prior IO (see online supplemental figure S5A for CD45RA and CD45RO expression in these clusters; see online supplemental figure S5B for selected marker expression in tissue-resident CD8 T-cell clusters). (D) Histograms show the signal intensity of selected markers on cells from cluster 3 (red) and cluster 26 (blue). Most cluster 3 ɣδ T cells were tissue resident (CD103+, CD69high) and PD-1high, TIGIThigh, CD39high, and Helioshigh, whereas cluster 26 ɣδ T cells had lower and more variable expression of these T-cell activation/checkpoint markers. (E) Scatter dot plots showing the expression of CD3 and TCR-ɣδ by cluster 3 (red) and cluster 26 (blue) cells in non-progressors (n=12) and progressors (n=20) from Arms A, B, and C. The number of events in each cluster is indicated. Overall, cluster 3 ɣδ T cells were more abundant in baseline biopsies from non-progressors than progressors (4058 vs 150 events), whereas the total number of cluster 26 ɣδ T cells was similar in the two groups (3636 vs 4227 events), suggesting that cluster 3 ɣδ T cells may be important for response to cabozantinib–nivolumab combination therapy. (F) UMAPs show the 35 phenograph-defined immune cell clusters present in the baseline biopsies of Arm B crossover patients before treatment with nivolumab (pre-nivo), and at the time of crossover (post-nivo) to Arm C before the start of combination treatment with cabozantinib–nivolumab (n=5). Differential abundance analysis of the phenograph-defined clusters revealed no significant differences between pre-nivolumab and post-nivolumab biopsies (online supplemental figure S6A). (G) Graphs depicting the proportion of cells from clusters 18, 3, and 26 pre-nivo and post-nivo among CD45+EpCAM– cells. These paired biopsy data further suggest that an increase in activated tissue-resident cluster 3 ɣδ T cells before cabozantinib–nivolumab combination therapy is potentially associated with a more favorable response. Patient A whose disease progressed on Arm B but responded on Arm C (highlighted in blue) exhibited a 3.8-fold increase in the percentage of activated, tissue-resident (CD103+CD69+) GzmBhigh CD8+ T cells (cluster 18), and >5-fold increases in the percentage of ɣδ T cells (clusters 3 and 26) following nivolumab monotherapy (see online supplemental figure S6B for all clusters). Compared with the crossover patients whose disease progressed in Arm C, the proportion of clusters 3 and 18 were more than 2-fold higher in patient A post-nivolumab before initiation of combination therapy. Patient B whose disease progressed on Arm B but was stable for ≥3 months on Arm C (highlighted in yellow) had a 25-fold increase in cluster 3 ɣδ T cells following nivolumab monotherapy. *Adjusted p<0.05 by Benjamini-Hochberg method. EpCAM, epithelial cell-adhesion molecule; ILC, innate lymphoid cell; IO, immuno-oncology; nivo, nivolumab; MSS, microsatellite stable; NK, natural killer; PD, progressive disease; PD-1, programmed cell death 1; PR, partial response; SD, stable disease; TCR, T-cell receptor; Treg, regulatory T cell; UMAP, uniform manifold approximation and projection.

Figure 5

Comparison of ɣδ T cells from baseline biopsies and serial PBMC samples. (A) UMAP shows the 13 phenograph-defined ɣδ T-cell clusters present in the baseline biopsies (n=32, Arms A, B, and C) and serial PBMCs (n=28 samples, Arms B and C). (B) UMAP shows the contribution of cells from baseline biopsies (blue) and PBMCs (multiple colors) to each of the 13 phenograph-defined ɣδ T-cell clusters. (C) Single-cell heatmap shows the (hierarchical) marker expression profiles that define each of the 13 ɣδ T-cell clusters, grouped by whether the majority of cells in each cluster are from PBMCs, a mix of PBMCs and biopsies, or biopsies. (D) Signal intensity of select markers on ɣδ T cells from each cluster. (E) Schematic shows the time points that serial PBMCs were collected from Arm B and Arm C patients (n=8 patients) for CyTOF analysis. PBMCs were isolated at baseline, C1D15, and at progression. (F) Graphs show the proportions of total ɣδ T cells (left) and cluster 11 ɣδ T cells (right) at each time point. Patients with a best response while on Arm C of PR (blue) and SD ≥3 months (yellow) are highlighted. (G) The top panel shows the diversity profiles (n=14) reconstructed for each patient (n=5 patients) at different time points (Baseline T0, Arm B T1, Arm B T2, Arm C T3, time of progression T4). The first three points on each profile demonstrate diversity indices: richness, Shannon diversity, and Gini-Simpson diversity, respectively. The rest of the indices depict the effective number of species when higher weights are given to the most abundant clonotypes. While more linear profiles illustrate more even distribution of clonotypes, high drops in the profile demonstrate skewness of the frequency distribution of clonotypes in the repertoire. Bottom panels add more resolution to the diversity profiles by showing the frequency distribution of each repertoire. Each point demonstrates a unique clonotype with its frequency shown on the y-axis. Repertoires with dots spanning over a wider spectrum of frequencies tend to have larger drops in their diversity profiles. (H) Area above the curve (AAC) of diversity profiles. In the progression window there is either no shift or a decrease in AAC in time points associated with Arm B compared with baseline T0. Comparing AACs of each point inside the progression window with a point in Arm C for each patient shows an increased value of AAC for responders versus a decreased value of AAC for progressors. Patients with microsatellite instability-high disease and those with carcinosarcoma were excluded. C1D15, cycle 1 day 15; Cabo, cabozantinib; CDR3, complementarity-determining region 3; CTLA-4, cytotoxic T-lymphocyte antigen-4; CXCR, C-X-C chemokine receptor; CyTOF, cytometry by time of flight; HLA-DR, human leukocyte antigen – DR isotype; IO, immuno-oncology; MSS, microsatellite stable; nivo, nivolumab; PBMC, peripheral blood mononuclear cell; PD, progressive disease; PD-1, programmed cell death 1; PD-L1, programmed cell death ligand-1; PR, partial response; SD, stable disease; UMAP, uniform manifold approximation and projection.