FGFR inhibition augments anti-PD-1 efficacy in murine FGFR3-mutant bladder cancer by abrogating immunosuppression

Abstract

The combination of targeted therapy with immune checkpoint inhibition (ICI) is an area of intense interest. We studied the interaction of fibroblast growth factor receptor (FGFR) inhibition with ICI in urothelial carcinoma (UC) of the bladder, in which FGFR3 is altered in 50% of cases. Using an FGFR3-driven, Trp53-mutant genetically engineered murine model (UPFL), we demonstrate that UPFL tumors recapitulate the histology and molecular subtype of their FGFR3-altered human counterparts. Additionally, UPFL1 allografts exhibit hyperprogression to ICI associated with an expansion of T regulatory cells (Tregs). Erdafitinib blocked Treg proliferation in vitro, while in vivo ICI-induced Treg expansion was fully abrogated by FGFR inhibition. Combined erdafitinib and ICI resulted in high therapeutic efficacy. In aggregate, our work establishes that, in mice, co-alteration of FGFR3 and Trp53 results in high-grade, non-muscle-invasive UC and presents a previously underappreciated role for FGFR inhibition in blocking ICI-induced Treg expansion.

Keywords: Cancer immunotherapy; Mouse models; Oncology.

Figure 1. UPFL mice develop papillary, high-grade, non–muscle-invasive bladder cancers.

(A) Schematic of the targeted and recombined Col1a1-LSL-FGFR3^S249C locus. (B) Proportion of TCGA-BLCA project tumors within each consensus molecular subtype with FGFR3 mutations. Inset numbers annotate actual patient numbers. (C) Violin plots of log₂-normalized Upk2 and Upk3a RNA expression in basal, intermediate, and luminal cells of mouse urothelium. Two-sided t test–derived P values were calculated between Upk2 and Upk3a within each cell type. Tumor-free survival and tumor growth were monitored by ultrasound imaging. (D) Representative serial ultrasound images of the bladder in a UPFL mouse, demonstrating bladder tumor growth (n = 51). (E) Kaplan-Meier curve of tumor-free survival of 51 total mice. (F) Photomicrographs of a bladder tumor and an upper tract urothelial carcinoma (UTUC) of the proximal ureter. Scale bars: 1 mm (left) and 100 μm (right). (G) Percentage of tumors of indicated pathologic T stage and grade (n = 9). (H) Stacked bar plots representing the co-occurrence of FGFR3 and p53 pathway alterations (TP53, ATM, RB1, MDM2, E2F3, ATR) within MIBC (TCGA and MSK [2014]) and NMIBC (UROMOL and MSK [2017]). Two-sided Fisher’s exact test was performed to calculate significance and log₂-transformed odds ratios were generated to determine association, with log₂(OR) < 0 indicating mutual exclusivity.

Figure 2. UPFL tumors are associated with luminal expression patterns.

(A) Proportion of BBN, UPPL, and UPFL primary tumors of the indicated RNA consensus MIBC molecular subtype. (B) Proportion of BBN, UPPL, and UPFL primary tumors of the indicated RNA UROMOL NMIBC molecular subtype. (C) Heatmap of unsupervised clustering of BBN, UPPL, and UPFL primary tumors by canonical basal and luminal genes. (D) Box-and-whisker plots of normalized expression of basal and luminal score as well as luminal to basal score ratio of BBN, UPPL, and UPFL primary tumors. (E) Box-and-whisker plots of normalized expression of Krt6a and Upk1a of BBN, UPPL, and UPFL primary tumors. (F) Box-and-whisker plots of PPARG and PPARG gene signature (17, 25) of BBN, UPPL, and UPFL primary tumors. All box-and-whisker plots show the IQR and midline at the median. Error bars represent Q1/Q3 ± (1.5 × IQR). Two-sided t tests followed by Bonferroni’s correction to account for multiple comparisons were performed; the P values are shown above the given comparison. (G) Volcano plot of regulon activity between UPFL versus BBN primary tumors. (H) Volcano plot of regulon activity between UPFL versus UPPL primary tumors. The x axis represents the log₂(fold change) between UPFL and BBN (G) or UPPL (H) and the y axis is the Benjamini-Hochberg false discovery rate for the given gene.

Figure 3. FGFR3 promotes a luminal expression pattern and suppresses the basal transcriptional program across all urothelial cell types, but most prominently in luminal cells.

(A) tSNE plot of scRNA-seq data of combined control (n = 2) and erdafitinib-treated (n = 2) tumors. (B) tSNE plot of scRNA-seq data of epithelial cells of combined control (n = 2) and erdafitinib-treated (n = 2) tumors clustered on basal and luminal genes, demonstrating 3 clusters. (C) Percentage and level of RNA expression of differentially expressed genes across the indicated clusters. (D) Histograms of basal and luminal scores of the indicated clusters from B. (E) Proportion of epithelial cell type (basal, intermediate, luminal) of control or erdafitinib-treated tumors demonstrates expansion of intermediate cells in erdafitinib-treated tumors. (F) Back-to-back violin plots of basal and luminal scores by cell type (basal, intermediate, luminal). (G) Back-to-back violin plots of Krt5 and Upk3a RNA expression by cell type (basal, intermediate, luminal). Two-sided t tests were performed, with the P values shown above the given comparison.

Figure 4. UPFL tumors have an intermediate T cell–inflamed immune contexture.

(A) GSEA plots of the indicated gene signatures. (B) Heatmap of unsupervised clustering of BBN, UPPL, and UPFL primary tumors by a panel of immune gene signatures. Box-and-whisker plots of the indicated immune gene signatures of BBN, UPPL, and UPFL primary tumors: (C) CD8⁺ T cells, (D) cytotoxic cells, (E) T follicular helper cells (TFH), (F) γδ T cells (Tgd), (G) central memory (Tcm) and effector memory T cells (Tem), and (H) fibroblast TGF-β response (FTBRS) and EMT Stroma signature. All box-and-whisker plots show the IQR and midline at the median. Error bars represent Q1/Q3 ± (1.5 × IQR). Two-sided t tests followed by Bonferroni’s correction to account for multiple comparisons were performed, with the P values shown above the given comparison.

Figure 5. UPFL1 cell lines are sensitive to FGFR inhibition with erdafitinib.

(A) Representative bright-field and GFP images of an epithelial island surrounded by GFP-expressing irradiated fibroblasts. (B) PCR demonstrating evidence of LoxP recombination of the LSL-FGFR3 S249C allele in UPFL1 and UPFL3 cell lines, while liver tissue shows only the targeted allele. (C) IC₅₀ curves were generated in biologic triplicate for UPFL1 and UPFL3 cells’ response to erdafitinib. UPPL1541 cells serve as a control for non–FGFR3-mutated murine bladder cancer cell line. (D) Immunoblots with the indicated antibodies in whole-cell extracts of UPFL1 cells treated with erdafitinib for the indicated dose and time.

Figure 6. FGFR inhibition enhances the effect of PD-1 blockade.

(A) Tumor growth curves of UPFL1 subcutaneous syngeneic tumors treated with the indicated treatments when tumor reached 150 to 300 mm³ in volume. Significance testing was performed by 1-way ANOVA with post hoc Tukey’s HSD. *P < 0.05, **P < 0.001. (B) FFPE tumors were sectioned and dual stained for CD8 and with Masson’s trichrome. Immune phenotyping was performed and the phenotype/CD8 intensity call was plotted as the proportion of each treatment group. Flow cytometry was performed on UPFL1 syngeneic tumors following 1 week of treatment. CTRL, control. (C) Box-and-whisker plots of percentage of cells in each treatment group of CD45⁺, CD3⁺, and CD8⁺ cytotoxic T cells after 1 week of treatment. (D) Box-and-whisker plot of percentage of cells in each treatment group of CD4⁺ cytotoxic T cells and CTLA-4⁺ cells after 1 week of treatment. (E) Cell types were assigned to scRNA-seq data from either control or erdafitinib-treated UPFL GEMM tumors using SingleR. The frequency of each immune cell type was plotted as a proportion of all immune cells. The inset percentage represents the percentage of T cells as a proportion of the total immune population. (F) The T cell subset of cells were plotted by the scRNA expression values for Icos, a marker of active/proliferative Tregs. (G) FoxP3⁺GFP⁺ cells were isolated from murine spleens and cocultured in the presence or absence of APCs and increasing doses of erdafitinib (n = 3 for each group). (H) Bulk RNA-seq data (GSE135390) from flow-sorted T cells (naive, Th, and Treg) were plotted for FOXP3, FGFR1, FGFR2, and FGFR4. Plots show the IQR and midline at the median. Error bars represent Q1/Q3 ± (1.5 × IQR). Two-sided t tests followed by Bonferroni’s correction to account for multiple comparisons were preformed, with the P values shown above the given comparison.