Immunomodulatory effects and improved outcomes with cisplatin- versus carboplatin-based chemotherapy plus atezolizumab in urothelial cancer

Abstract

In metastatic urothelial cancer (mUC), cisplatin versus carboplatin leads to durable disease control in a subset of patients. The IMvigor130 trial reveals more favorable effects with atezolizumab combined with gemcitabine and cisplatin (GemCis) versus gemcitabine and carboplatin (GemCarbo). This study investigates the immunomodulatory effects of cisplatin as a potential explanation for these observations. Our findings indicate that improved outcomes with GemCis versus GemCarbo are primarily observed in patients with pretreatment tumors exhibiting features of restrained adaptive immunity. In addition, GemCis versus GemCarbo ± atezolizumab induces transcriptional changes in circulating immune cells, including upregulation of antigen presentation and T cell activation programs. In vitro experiments demonstrate that cisplatin, compared with carboplatin, exerts direct immunomodulatory effects on cancer cells, promoting dendritic cell activation and antigen-specific T cell killing. These results underscore the key role of immune modulation in cisplatin's efficacy in mUC and highlight the importance of specific chemotherapy backbones in immunotherapy combination regimens.

Trial registration: ClinicalTrials.gov NCT02807636.

Keywords: PD-1 blockade; PD-L1 blockade; bladder cancer; carboplatin; chemotherapy; cisplatin; immune checkpoint blockade; immunogenic cell death; single cell RNA sequencing; urothelial cancer.

Graphical abstract

Figure 1

Survival outcomes with GemCis but not with GemCarbo are dependent on pretreatment tumor PD-L1 expression in an exploratory analysis of the IMvigor130 study (A) Study design of phase 3 IMvigor130. Patients in arms A and C received atezolizumab and placebo, respectively, in combination with the investigators’ choice of platinum drugs (cisplatin versus carboplatin) and gemcitabine. Arm A, n = 451; arm B, n = 362; arm C, n = 400. (B) Forest plot showing overall survival in patients in arms A and C by use of GemCis versus GemCarbo. Hazard ratios, 95% confidence intervals, and p values were calculated using a univariate Cox model. The diamonds represent the hazard ratios, and the horizontal bars their 95% confidence intervals. (C) Kaplan-Meier curves showing overall survival in patients in arm C stratified by PD-L1 status (IC2/3 versus IC0/1) and use of GemCis (left) or GemCarbo (right). (D) Kaplan-Meier curves showing overall survival in patients in arm A stratified by PD-L1 status (IC2/3 versus IC0/1) and use of GemCis plus atezolizumab (left) or GemCarbo plus atezolizumab (right). In (C) and (D), p values were estimated using the log rank test. Hazard ratios and 95% confidence intervals were calculated using a univariate Cox model. (E) Kaplan-Meier curves showing overall survival in patients in arm C stratified by actual receipt of GemCis or GemCarbo and tumor PD-L1 status (IC2/3 versus IC0/1). Patients classified as “cisplatin-eligible” according to standard criteria were included. p values were estimated using the log rank test. See also Figures S1 and S2 and Tables S1 and S2.

Figure 2

PD-L1 IC2/3 versus IC0/1 tumors are enriched in immune-related gene signatures (A) Heatmap detailing immune-related gene signatures based on bulk RNA sequencing of pretreatment archival tumor specimens from the IMvigor130 study according to PD-L1 status (IC2/3 and IC0/1) and treatment arm. (B) Volcano plot detailing differentially expressed genes based on bulk RNA sequencing of pretreatment archival tumor specimens from the IMvigor130 study according to PD-L1 status (IC2/3 versus IC0/1). Differential expression analysis was conducted with limma-based statistical methods and the Benjamini-Hochberg correction. (C) Gene set enrichment analysis based on bulk RNA sequencing of pretreatment archival tumor specimens from the IMvigor130 study according to PD-L1 status (IC2/3 versus IC0/1) and treatment arm or specific platinum drug received. The hue represents the false discovery rate (FDR) significance derived from the fgsea package. Black asterisks represent FDR < 0.05.

Figure 3

GemCis ± atezolizumab versus GemCarbo ± atezolizumab induces proinflammatory transcriptional programs across multiple immune cell subsets in PBMCs (A) Single-cell CITE-seq experimental design and analysis workflow. (B) Uniform Manifold Approximation and Projection (UMAP) visualization of single cells captured, colored by major cell types (n = 865,922) (left), and heatmap showing scaled expression of the canonical marker genes across cell types (right). (C and D) Heatmaps showing pathway enrichment on-treatment (C3D1) versus at baseline (C1D1) across multiple immune cell types. Treatments received were GemCis (left, n = 17 pairs) or GemCarbo (right, n = 16 pairs) in arm C (C) and atezolizumab (n = 33 pairs) in arm B (D). (E) Comparison of MHC class gene expression on-treatment (C3D1) versus at baseline (C1D1) in monocytes collected from patients receiving GemCis or GemCarbo in arm C. (F) Heatmaps showing pathway enrichment on-treatment (C3D1) versus at baseline (C1D1) across multiple immune cell types. Treatments received were GemCis + atezolizumab (left, n = 14 pairs) or GemCarbo + atezolizumab (right, n = 24 pairs) in arm A. In (C), (D), and (F), red indicates enrichment in on-treatment samples and blue indicates enrichment at baseline. The hue represents the false discovery rate (FDR) significance derived from the fgsea package. Black asterisks represent FDR < 0.05. See also Figures S3 and S4.

Figure 4

Atezolizumab added to GemCis versus GemCarbo leads to distinct transcriptional states of circulating immune cells including changes reflective of T cell activation states (A) UMAP visualization of total T and NK cells captured, colored by major cell types (n = 707,614). (B) UMAP visualization of total CD8 T cells captured, colored by different cell types (n = 171,729) (left), and heatmap showing scaled expression of the top 10 cell markers ranked by fold change in each cell type (right). Black asterisks indicate cell types not included in subsequent analysis due to low cell numbers. (C) Heatmaps showing pathway enrichment on-treatment (C3D1) versus at baseline (C1D1) across T and NK cell types. Treatments received were GemCis (first panel, n = 17 pairs) or GemCarbo (second panel, n = 16 pairs) in arm C and GemCis + atezolizumab (third panel, n = 14 pairs) or GemCarbo + atezolizumab (fourth panel, n = 24 pairs) in arm A. (D) Heatmaps showing pathway enrichment with GemCis on-treatment (C3D1) versus baseline (C1D1) across T and NK cell types in responders (left) and nonresponders (right). In (C) and (D), red indicates enrichment in on-treatment samples and blue indicates enrichment at baseline. The hue represents the false discovery rate (FDR) significance derived from the fgsea package. Black asterisks represent FDR < 0.05.

Figure 5

The direct effects of cisplatin versus carboplatin on downstream immune-related programs in cancer cells are mediated by the DNA damage transducer ATR (A) Enrichment of Hallmark gene sets with cisplatin (red) versus carboplatin (gray) treatment in the 5637 human bladder cancer cell line (top), RT112 human bladder cancer cell line (middle), and MC38 murine colon cancer cell line (bottom). The cisplatin and carboplatin concentrations used were 5 and 35 μM, respectively, and cells were collected 24 h after treatment. Only pathways that were significantly changed with cisplatin versus carboplatin (false discovery rate [FDR] < 0.05) are shown. n = 3 per treatment group for the 5637 and RT112 cell lines; n = 2 per treatment group for the MC38 cell line. (B) Protein expression of PD-L1 as determined by the median fluorescence intensity in the three cell lines treated with increasing concentrations of cisplatin (red) or carboplatin (black) for 24 h. Data depict the aggregate of three independent experiments (mean ± SEM). (C) Representative immunoblots showing p-ATM S1981, p-ATR T1989, p-Chk1 S317, p-Chk2 T68, and GAPDH levels in lysates of the 5637 cell line treated with increasing concentrations of cisplatin or carboplatin for 6 h. One representative experiment out of four independent experiments is shown. (D) Protein expression of PD-L1 as determined by median fluorescence intensity in the 5637 cell line treated with increasing concentrations of cisplatin or carboplatin for 24 h, in the presence or absence of an ATM inhibitor (KU-55933, 1 μM) or ATR inhibitor (VE-821, 1 μM). Data depict the aggregate of three independent experiments (mean ± SEM). (E and F) Bulk RNA-seq analysis of 5637 cells treated with 5 μM cisplatin or 35 μM carboplatin, in the absence (gold) or presence of 1 μM ATR inhibitor (navy blue) for 24 h. (E) Enrichment of Hallmark gene sets. n = 4 per treatment group. (F) Heatmap displaying the scaled transcriptional expression of genes involved in immune-related transcriptional programs. Heatmap shows an average score of n = 4 individual samples per treatment group. See also Figures S5–S8.

Figure 6

Co-culture of cisplatin- versus carboplatin-pretreated MC38 cancer cells with bone marrow-derived DCs and OT-I T cells induces DC activation, OT-I T cell proliferation, and antigen-specific T cell-mediated MC38 killing (A) HMGB1 secretion as measured in the supernatant collected from MC38 cancer cells treated with increasing concentrations of cisplatin (red) or carboplatin (black). Supernatant was collected 3 days after treatment. Data depict one representative experiment of two independent experiments; duplicate conditions for each experiment. Data are mean ± SEM. (B) Protein expression of MHC-I, PD-L1, CD40, and CD86 as measured by median fluorescence intensity in monocyte-derived DCs after 24 h of co-culture with MC38 cancer cells that were primed with dose titrations of cisplatin (red) or carboplatin (black) for 24 h. Data depict one representative experiment of three independent experiments; duplicate conditions for each experiment. Data are mean ± SEM. (C) Flow cytometry histograms depicting OT-I T cell proliferation measured by CellTrace Blue dilution assay. MC38-OVA cells were pretreated with dose titrations of cisplatin or carboplatin for 24 h. OT-I T cells were co-cultured with pretreated MC38-OVA cells for 3 days in the presence or absence of monocyte-derived DCs before the proliferation assay. Data depict one representative experiment of two independent experiments. (D and E) Representative flow cytogram showing OT-I T cell-mediated MC38-OVA cancer cell killing (D) and the summarized percentage of OT-I T cell-mediated tumor cell death (E) as measured by 7-AAD and annexin V staining. MC38-OVA tumor cells were pretreated with increasing concentrations of cisplatin or carboplatin for 24 h, followed by co-culture with OT-I T cells for 5 h before the assay. Data depict one representative experiment of three independent experiments; duplicate conditions for each experiment. Data are mean ± SEM. See also Figure S9.