Epigenomic signatures of sarcomatoid differentiation to guide the treatment of renal cell carcinoma

Abstract

Renal cell carcinoma with sarcomatoid differentiation (sRCC) is associated with poor survival and a heightened response to immune checkpoint inhibitors (ICIs). Two major barriers to improving outcomes for sRCC are the limited understanding of its gene regulatory programs and the low diagnostic yield of tumor biopsies due to spatial heterogeneity. Herein, we characterized the epigenomic landscape of sRCC by profiling 107 epigenomic libraries from tissue and plasma samples from 50 patients with RCC and healthy volunteers. By profiling histone modifications and DNA methylation, we identified highly recurrent epigenomic reprogramming enriched in sRCC. Furthermore, CRISPRa experiments implicated the transcription factor FOSL1 in activating sRCC-associated gene regulatory programs, and FOSL1 expression was associated with the response to ICIs in RCC in two randomized clinical trials. Finally, we established a blood-based diagnostic approach using detectable sRCC epigenomic signatures in patient plasma, providing a framework for discovering epigenomic correlates of tumor histology via liquid biopsy.

Keywords: CP: Cancer; CP: Genomics; FOSL1; epigenomics; histone modifications; immune checkpoint inhibitors; immunotherapy; kidney cancer; liquid biopsy; renal cell carcinoma; sarcomatoid; transcription factors.

Figure 1. Tissue- and plasma-based epigenomic characterization of sRCC.

(A) Epigenomic profiling of frozen tissue samples from patients with epithelioid or sarcomatoid ccRCC. (B) Stratification of tissue samples by subtype-specific epigenomic features and their correlation with clinical outcomes in three cohorts. (C) Detection of sarcomatoid-like gene expression patterns following activation of expression of a candidate transcription factor in an RCC cell line model. (D) Plasma-based epigenomic profiling of patients with ccRCC and healthy volunteers. ccRCC: clear cell renal cell carcinoma, TF: transcription factor: ICI: immune checkpoint inhibitors, TKI: tyrosine kinase inhibitors, LP-WGS: low pass whole genome sequencing, cfDNA: cell-free DNA.

Figure 2.. H3K27ac, H3K4me2, and methylated CpG dinucleotide regions between sarcomatoid and epithelioid ccRCC reflect distinct epigenomic programs.

(A) Epigenomic datasets generated from tissue samples. (B) Distribution of H3K27ac, H3K4me2, and MeDIP peaks by genomic annotations (C-E) Principal component analysis (PCA) plots of the H3K27ac, H3K4me2, and MeDIP peaks in epithelioid and sarcomatoid ccRCC. (F) Venn-Diagrams of the subtype-enriched H3K27ac, H3K4me2, and MeDIP peaks.

Figure 3.. Sarcomatoid-enriched CREs are associated with immune system activation and drivers of EMT.

(A) Heatmaps of normalized H3K27ac tag densities at differential H3K27ac cis-regulatory elements (CREs) between epithelioid and sarcomatoid ccRCC samples located ±2 kb from peak center. (B) GREAT analysis of sarcomatoid-enriched H3K27ac peaks (n = 1,621). (C) GREAT analysis of epithelioid-enriched H3K27ac peaks (n = 3,386). (D) Three top non-redundant significantly enriched nucleotide motifs present in sarcomatoid-specific sites by de novo motif analysis. (E) H3K27ac profiles near FOSL1 in five representative samples of each ccRCC subtype, normalized to signal at GAPDH. (F) Three significantly enriched nucleotide motifs present in epithelioid-specific sites by de novo motif analysis. (G) H3K27ac profiles near EPAS1 normalized to GAPDH in five representative samples of each ccRCC subtype (epithelioid and sarcomatoid).

Figure 4.. FOSL1 is upregulated in sRCC and is associated with clinical outcomes.

(A) Boxplots of FOSL1 expression levels in three clinical cohorts (TCGA, JR101, and IM151). (B) Progression-free survival in patients in JR101 by FOSL1 levels. (C) Overall survival in patients with sRCC and epithelioid ccRCC divided by expression of FOSL1 (High vs. Low) in the TCGA cohort. (D) Progression-free survival in patients with sRCC and epithelioid ccRCC divided by expression of FOSL1 (High vs. Low) in the sunitinib arms of JR101 and IM151. TPM: Transcripts per million, TCGA: The Cancer Genome Atlas, JR101: Javelin Renal 101, IM151: Immotion151, Epi: epithelioid ccRCC, AveAxi: avelumab plus axitinib, AtezoBev: atezolizumab plus bevacizumab, SUN: sunitinib.

Figure 5.. Expression of FOSL1 in an epithelioid ccRCC cell line activates sRCC transcriptional programs.

(A) Principal Component Analysis (PCA) plot of RNA-sequencing data from two conditions (FOSL1 CRISPRa and negative control Caki-1 cells, n=3 independent biological replicates). (B) Volcano plot showing upregulated and downregulated genes in FOSL1 CRISPRa vs. negative control Caki-1 cells (n=3 independent biological replicates). (C) Representative immunoblot image of FOSL1 and HIF2α protein levels in Caki-1 cell lines (negative control and FOSL1 CRISPRa). (D-E) Quantification of FOSL1 and HIF2α protein levels represented as mean ± SD band intensity normalized to GAPDH levels (n=3 independent biological replicates). Protein expression level differences between negative control and FOSL1 CRISPRa Caki-1 cells were analyzed by an unpaired t-test. Statistical significance p values are: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Error bars represent the standard deviation from the mean. (F) Normalized enrichment scores of pathways upregulated in FOSL1 CRISPRa Caki-1 cells. (G) Gene set enrichment analysis (GSEA) of genes upregulated in sarcomatoid ccRCC from JR101 in FOSL1 CRISPRa Caki-1 cells. ccRCC: clear cell renal cell carcinoma, NES: normalized enrichment score, FDR: false-detection rate. JR101: Javelin Renal 101.

Figure 6.. Tissue-informed epigenomic signatures enable the detection of sRCC in plasma.

(A) Schematic demonstrating the measurement of cfChIP and cfMeDIP signals at TF binding sites differentially methylated regions, respectively. (B) Epigenomic datasets generated from plasma. (C) H3K4me3 cfChIP profiles at PAX8 in plasma of patients with sRCC (n=2), epithelioid ccRCC (n=2) and healthy volunteers (n=4). (D) Aggregated H3K27ac cfChIP-seq signals in kidney cancer and healthy plasma at REs identified by ATAC-seq in ccRCC. (E) Aggregated cfMeDIP-seq signals at tissue-informed upregulated sarcomatoid-specific DMRs and comparison between sRCC, epithelioid ccRCC, and healthy plasma, respectively. (F) Aggregated H3K27ac cfChIP-seq signals at tissue-informed upregulated sarcomatoid-specific H3K27ac sites and comparison between sRCC, epithelioid ccRCC, and healthy plasma, respectively. (G) Aggregated H3K27ac cfChIP-seq signal at HIF2α binding sites for ccRCC and comparison between sRCC, epithelioid ccRCC, and healthy plasma, respectively. (D-G) Dark lines show the mean signal across all samples in the indicated class. Gray shades represent the standard errors of the means. Boxplots indicate the area under the curve for the aggregate H3K27ac or MeDIP profiles for each sample. Wilcoxon test p-values are indicated for comparison across the groups: ccRCC (red), healthy (gray), epithelioid ccRCC (blue), and sRCC (orange). ccRCC: clear cell renal cell carcinoma, Sarc: sarcomatoid ccRCC (sRCC), Epi: epithelioid ccRCC, RE: regulatory elements, DMR: differentially methylated regions.