Sitravatinib in combination with nivolumab plus ipilimumab in patients with advanced clear cell renal cell carcinoma: a phase 1 trial

Abstract

We conducted a phase I trial to determine the optimal dose of triplet therapy with the tyrosine kinase inhibitor sitravatinib plus nivolumab plus ipilimumab in 22 previously untreated patients with advanced clear cell renal cell carcinoma. The primary endpoint was safety. Secondary endpoints were objective response rate (ORR), disease control rate (DCR), duration of response (DOR), progression-free survival (PFS), overall survival (OS), 1-year survival probability, and sitravatinib pharmacokinetics. Sitravatinib dose of 35 mg daily plus nivolumab 3 mg/kg and ipilimumab 1 mg/kg resulted in high frequency of immune-related adverse events. Subsequent dose reduction of ipilimumab to 0.7 mg/kg allowed safe escalation of sitravatinib up to 100 mg daily. Overall, the triplet combination achieved ORR 45.5%, DCR 86.4%, median PFS 14.5 months, and 1-year survival 80.8%. Median OS and DOR were not reached. Sitravatinib exposure increased dose-dependently. Single-cell RNA-seq of longitudinally collected tumor biopsies from 12 patients identified a tumor cell-specific epithelial-mesenchymal transition-like program associated with treatment resistance and poor outcomes. Treatment resistance was characterized by a transition from cytotoxic to exhausted T cell state and enrichment for M2-like myeloid cells. The observed hypothesis-generating changes in gene expression dynamics and cellular states may help inform future strategies to optimize immunotherapy efficacy. Clinical Trials.gov identifier: NCT04518046.

Fig. 1. Trial design and conduct.

a Study schema. Tumor biopsy samples were collected at baseline, prior to the second infusion of the triplet therapy (approximately 3 weeks after treatment initiation), and at disease progression. b CONSORT diagram. c Treatment course timeline for the seven patients enrolled in cohort 1. C2, cycle #2; ccRCC, clear cell renal cell carcinoma; CR complete response, DLT dose-limiting toxicity, EOT end of treatment, PD progressive disease, PR partial response, SD stable disease, Q3W every 3 weeks, Q4W every 4 weeks.

Fig. 2. Efficacy outcomes of sitravatinib in combination with nivolumab and ipilimumab.

a Waterfall plot showing the maximum change in the sum of the longest dimensions in each of the 22 patients treated with sitravatinib + nivolumab + ipilimumab. The IMDC risk category, presence or absence of sarcomatoid and/or rhabdoid dedifferentiation, as well as the baseline peripheral blood neutrophil-to-lymphocyte ratio (NLR) for each patient are also shown. b Swimmer plot showing the durability of responses to therapy. The IMDC risk category, presence or absence of sarcomatoid and/or rhabdoid dedifferentiation, and NLR for each patient are also shown. c, d Kaplan-Meier survival curves for progression-free survival (c) and overall survival (d) in all 22 patients treated with sitravatinib + nivolumab + ipilimumab. Dotted lines intersect the survival curve at median (also printed). Shaded blue areas show 95% Hall-Wellner confidence bands. Source data are provided in the Source Data file.

Fig. 3. Single-cell landscape of tumor microenvironment (TME) cells from longitudinally collected tumor samples at baseline and during trial therapy.

a Uniform manifold approximation and projection (UMAP) of scRNA-seq data from TME cells. Colors and numbers indicate scRNA-seq clusters. b Expression levels and frequencies of selected markers across TME cell clusters. Bubble size is proportional to the percentage of cells expressing a gene and color intensity is proportional to average scaled gene expression. c UMAP view of TME cell clusters (top) and cell density (bottom) displaying TME cell distribution across four groups. Pre_R (n = 3; baseline samples from patients that demonstrated an objective response to trial therapy), Pre_NR (n = 9; baseline samples from patients that did not demonstrate an objective response to trial therapy), Post (n = 4; samples collected at the C2 timepoint), and EOT (n = 3; samples collected at the EOT timepoint). Downsampling was applied and 8,346 cells were included for each group. High relative cell density is shown as bright magma. d Distribution of TME cell clusters across groups. Left bar plot showing the relative proportion of cells from all patients for each TME cell subset. Heat map showing tissue prevalence estimated by the ratio of observed to expected cell number (Ro/e). e Therapeutic index of major immune cell types in ccRCC tumors treated with sitravatinib + nivolumab + ipilimumab. Dot size represents the significance evaluated by −Log10 (p-value). The two-sided p-value is obtained from the Pearson correlation analysis between cell type proportions and objective response rates without adjustment for multiple comparisons. Source data are provided in the Source Data file.

Fig. 4. EMT high tumor cell clusters associated with treatment resistance.

a Uniform manifold approximation and projection (UMAP) of scRNA-seq data from malignant cells. Colors and numbers indicate scRNA-seq clusters. Left bar plot showing the relative proportion of cells from all patients and biopsy sites for each malignant cell subset. b UMAP view of malignant cell clusters (top) and cell density (bottom) displaying malignant cell distribution across three groups. Pre_R (n = 3; baseline samples from patients that demonstrated an objective response to trial therapy), Pre_NR (n = 9; baseline samples from patients that did not demonstrate an objective response to trial therapy), and EOT (n = 3; samples collected at the EOT timepoint). Downsampling was applied and 2597 cells were included for each group. High relative cell density is shown as bright magma. c Distribution of malignant cell clusters across groups. Left bar plot showing the relative proportion of cells from all patients for each malignant cell subset. Heat map showing tissue prevalence estimated by the ratio of observed to expected cell number (Ro/e). d The therapeutic index of major immune cell types in ccRCC tumors treated with sitravatinib + nivolumab + ipilimumab. Dot size represents the significance evaluated by −Log10 (p-value). The two-sided p-value is obtained from the Pearson correlation analysis between cell type proportions and objective response rates without adjustment for multiple comparisons. e Pathway enrichment analysis for signature genes in C8 cell clusters. Barplot shows selected top ranked pathways. The p-value was calculated using a two-sided Fisher’s exact test without adjustment for multiple comparisons. f Expression levels of selected EMT related meta-program signatures across malignant cell clusters. N = 2011 cells for C1, 1437 cells for C2, 1113 cells for C3, 1101 cells for C4, 1355 cells for C5, 904 cells for C6, 860 cells for C7, 806 cells for C8, 1313 cells for C9, and 446 cells for C10. Box-and-whisker plots show all values with range (whiskers), interquartile range (box) and median (center line). g Selected EMT transcription factor activities between different malignant cell clusters. h Bubble plot showing key marker gene expression between the C2 and C8 cell clusters. i Disease-specific survival (DSS) differences based on C8 signature expression level in patients with clear cell renal cell carcinoma (KIRC) in the TCGA dataset. The p-value was calculated using the log-rank test to compare survival distributions between the defined groups. A two-sided test without adjustment for multiple compariasons was applied to assess whether there were significant differences in DSS between the groups. Source data are provided in the Source Data file.

Fig. 5. Longitudinal T cell changes during trial therapy.

a Uniform manifold approximation and projection (UMAP) of scRNA-seq data from CD4 + T cells. Colors and numbers indicate scRNA-seq clusters. b Distribution of CD4 + T cell clusters across groups. Left bar plot showing the relative proportion of cells from all patients for each CD4 + T cell subset. Heat map showing tissue prevalence estimated by the ratio of observed to expected cell number (Ro/e). Pre_R (n = 3; baseline samples from patients that demonstrated an objective response to trial therapy), Pre_NR (n = 9; baseline samples from patients that did not demonstrate an objective response to trial therapy), Post (n = 4; samples collected at the C2 timepoint), and EOT (n = 3; samples collected at the EOT timepoint). c UMAP view of CD4 + T cell clusters (top) and cell density (bottom) displaying CD4 + T cell distribution across four groups. Downsampling was applied and 1,345 cells were included for each group. High relative cell density is shown as bright magma. d Heat map illustrating expression of 17 curated gene signatures across CD4 + T cell clusters. Heat map was generated based on the scaled gene signature scores. e Radar plot showing enrichment of selected five CD4 + T cell states. f UMAP of scRNA-seq data from CD8 + T cells. Colors and numbers indicate scRNA-seq clusters. g Distribution of CD8 + T cell clusters across groups. Left bar plot showing the relative proportion of cells from all patients for each CD8 + T cell subset. Heat map showing tissue prevalence estimated by Ro/e. h Heat map illustrating expression of 20curated gene signatures across CD8 + T cell clusters. Heat map was generated based on the scaled gene signature scores. i Radar plot showing enrichment of selected five CD8 + T cell states. Source data are provided in the Source Data file.

Fig. 6. Longitudinal myeloid cell changes during trial therapy.

a Uniform manifold approximation and projection (UMAP) of scRNA-seq data from myeloid cells. Colors and numbers indicate scRNA-seq clusters. b Distribution of myeloid cell clusters across groups. Left bar plot showing the relative proportion of cells from all patients for each myeloid cell subset. Heat map showing tissue prevalence estimated by the ratio of observed to expected cell number (Ro/e). Pre_R (n = 3; baseline samples from patients that demonstrated an objective response to trial therapy), Pre_NR (n = 9; baseline samples from patients that did not demonstrate an objective response to trial therapy), Post (n = 4; samples collected at the C2 timepoint), and EOT (n = 3; samples collected at the EOT timepoint). c UMAP view of myeloid T cell clusters (left) and cell density (right) displaying myeloid cell distribution across four groups. Downsampling was applied and 1601 cells were included for each group. High relative cell density is shown as bright magma. d Expression levels of pathway signatures across myeloid cell clusters. N = 1950 cells for TAM_C0_APOC1, 1914 cells for Mono_C1_S100A8, 1725 cells for TAM_C2_C1QA, 1253 cells for TAM_C3_MT + , 1047 cells for TAM_C4_CD163_MERTK, 898 cells for Myeloid_C5_OLR1, 593 cells for cDC2_C6, 529 cells for Mast_C7, 497 cells for TAM_C8_Prolif, 208 cells for pDC_C9, 171 cells for cDC1_C10, and 84 cells for DC_C11_LAMP3. Box-and-whisker plots show all values with range (whiskers), interquartile range (box) and median (center line). e Cell communication analysis reveals strong CSF1-CSF1R cell communication between malignant cell populations leading to treatment resistance and TAMs. The p-values represent the likelihood of observing the identified or more extreme interactions if the null hypothesis of no interactions was correct, with Benjamini-Hochberg adjusted p-values for multiple testing correction. f Expression levels of M2 macrophage related signature in macrophage and monocytes cell clusters across groups. N = 2517 cells for Pre_R, 4074 cells for Pre_NR, 1601 cells for Post, and 2677 cells for the EOT timepoint. Box-and-whisker plots show all values with range (whiskers), interquartile range (box) and median (center line). ****p-value < 0.0001. The p-value was calculated by using the two-sided Wilcoxon rank-sum test without adjustment for multiple comparisons. g Expression levels of CSF1R gene in macrophage and monocytes cell clusters across groups. N = 2517 cells for Pre_R, 4074 cells for Pre_NR, 1601 cells for Post, and 2677 cells for the EOT timepoint. Box-and-whisker plots show all values with range (whiskers), interquartile range (box) and median (center line). ****p-value < 0.0001. The p-value was calculated by using the two-sided Wilcoxon rank-sum test without adjustment for multiple comparisons. h Heat map illustrating expression of top 5 selected hallmark signatures across groups in macrophage and monocytes cell clusters. Heat map was generated based on the scaled gene signature scores. Source data are provided in the Source Data file.