Epigenomic charting and functional annotation of risk loci in renal cell carcinoma

Abstract

While the mutational and transcriptional landscapes of renal cell carcinoma (RCC) are well-known, the epigenome is poorly understood. We characterize the epigenome of clear cell (ccRCC), papillary (pRCC), and chromophobe RCC (chRCC) by using ChIP-seq, ATAC-Seq, RNA-seq, and SNP arrays. We integrate 153 individual data sets from 42 patients and nominate 50 histology-specific master transcription factors (MTF) to define RCC histologic subtypes, including EPAS1 and ETS-1 in ccRCC, HNF1B in pRCC, and FOXI1 in chRCC. We confirm histology-specific MTFs via immunohistochemistry including a ccRCC-specific TF, BHLHE41. FOXI1 overexpression with knock-down of EPAS1 in the 786-O ccRCC cell line induces transcriptional upregulation of chRCC-specific genes, TFCP2L1, ATP6V0D2, KIT, and INSRR, implicating FOXI1 as a MTF for chRCC. Integrating RCC GWAS risk SNPs with H3K27ac ChIP-seq and ATAC-seq data reveals that risk-variants are significantly enriched in allelically-imbalanced peaks. This epigenomic atlas in primary human samples provides a resource for future investigation.

Fig. 1. Landscape of H3K27ac signals across RCC histologies.

A Distribution of RCC H3K27ac peaks according to genomic region for 30 fresh frozen RCC tumor samples (12 chRCC, 6 pRCC, 12 ccRCC). B Numbers of histology-specific and common H3K27ac peaks. C H3K27ac profiles at PAX8 in six representative samples from each RCC histology. D Hierarchical clustering of chRCC, ccRCC, and pRCC based on sample-to-sample pairwise correlation of the H3K27ac ChIP-seq peaks. E Distribution of histology-specific H3K27ac peaks among RCC subtypes. F–H Volcano plots with the log change of gene expression (FPKM) in one histology compared to the other two histologies (F ccRCC vs. others, G pRCC vs. others, H chRCC vs. others). Two-sided P values were used and corrected for multiple comparison testing (FDR-adjusted P value <0.05). RCC renal cell carcinoma, chRCC chromophobe RCC, ccRCC clear cell RCC, pRCC papillary RCC.

Fig. 2. Epigenetic annotation of regulatory elements identifies enrichment of histology-specific pathways and TFs.

A GREAT analysis of chromophobe-enriched peaks (n = 8939). Two-sided P values are shown. B GREAT analysis of papillary-enriched peaks (n = 3653). Two-sided P values are shown. C GREAT analysis of clear cell-enriched peaks in clear cell vs. papillary only comparison (n = 1265). A–C GREAT calculates statistical enrichments for association between genomic regions and annotations. Two-sided P values are shown. D Density map of correlation between H3K27ac versus H3K4me2 ChIP-seq peaks across subtypes. E Four most significantly enriched nucleotide motifs present in chRCC-specific sites by de novo motif analysis, limited by ATAC peaks. F, G H3K27ac profiles near FOXI1 and TFCP2L1, respectively, in two representative samples of each histology (chRCC, ccRCC, pRCC). H Two most significantly enriched nucleotide motifs present inccRCC specific sites by de novo motif analysis. P values are two-sided and unadjusted for multiple comparisons. I, J H3K27ac profiles near ETS-1, and HNF1B, respectively, in two representative samples of RCC histology. K Three most significantly enriched nucleotide motifs present in pRCC-specific sites by de novo motif analysis. P values are two-sided and unadjusted for multiple comparisons. RCC renal cell carcinoma, chRCC chromophobe RCC, ccRCC clear cell RCC, pRCC papillary RCC.

Fig. 3. Multi-dimensional integrative analysis identifies histology-specific master TFs.

A Overview of the approach used to nominate histology-specific master TFs participating in CRC. B Heatmap integrating the 50 histology-specific TFs identified by the meta-analysis approach (CES, differential expression, SE rank analysis, and CaCTS). C Representative immunohistochemical stainings of indicated antibodies in samples from ccRCC, chRCC, and pRCC tumors. Four histology-specific master TFs are shown. The number of positive tumors/number of tumors examined are indicated below each image (for ccRCC, n = 4 for TF BHLHE41, n = 3 for TFs HNF1B, NKX6.1, and ZNF395. For chRCC, n = 3 for TFs BHLHE41, HNF1B, and ZNF395, n = 2 for NKX6.1. For pRCC, n = 3 for BHLHE41, NKX6.1, and ZNF395, n = 2 for HNF1B). Scale bar is 50 μm. D ChIP-seq binding profiles of EPAS1 across 6 ccRCC and 2 chRCC human tissue samples. E GREAT analysis of chRCC- enriched EPAS1 peaks relative to ccRCC. Two-sided P values are shown. F GREAT analysis of ccRCC-enriched EPAS1 peaks relative to chRCC. Two-sided P values are shown. E, F GREAT calculates statistical enrichments for the association between genomic regions and annotations. ChIP-seq chromatin immunoprecipitation sequencing, ATAC-seq assay for transposase-accessible chromatin sequencing, DEG differentially expressed genes, SE superenhancer, CaCTS cancer core transcription-factor specificity, CRC core regulatory circuitries, RCC renal cell carcinoma, chRCC chromophobe RCC, ccRCC clear cell RCC, pRCC papillary RCC. Source data are provided as a Source Data file.

Fig. 4. Functional perturbation of two master TFs in a ccRCC cell line yields a more chRCC-like transcriptional profile.

A Downregulated GO biological terms in the cell line 786-O FOXI1 OE/EPAS1 KD. Two-sided adjusted P value corrected for multiple comparisons. B Rank order of differentially expressed TFs between 786-O CTRL and 786-O FOXI1 OE/EPAS1 KD by expression levels. Each TF dot represents one TF. Wald test from DESeq2 was used. Purple dots indicate adjusted P value <0.001 by DESeq Expression. C Gene expression levels of CAIX, CD117, and CK7 across different 786-O cell line conditions. D Heatmap showing the relative expression of the top 20 upregulated genes in ccRCC vs. chRCC and vice versa in the different cell line conditions. E GSEA analysis showing the top 100 upregulated genes from chRCC vs. ccRCC dataset in 786-O FOXI1 OE/EPAS1 KD vs. 786-O CTRL comparison. F GSEA analysis of the top 100 upregulated genes in ccRCC vs. chRCC in 786-O CTRL vs. 786-O FOXI1 OE/EPAS1 KD. CTRL control, KD knockdown, OE overexpression, TF transcription factor, GSEA gene set enrichment analysis, 786Oc1 CTRL1, 786Oc2 CTRL2, 786OD1 786-O FOXI1 OE/EPAS1 KD1, 786OD2 786-O FOXI1 OE/EPAS1 KD2, NES Normalized Enrichment Score, FDR false discovery rate, FC fold change, RCC renal cell carcinoma, chRCC chromophobe RCC, ccRCC clear cell RCC, pRCC papillary RCC.

Fig. 5. Allelically imbalanced H3K27ac peaks in ccRCC and pRCC.

A Schematic of allelic imbalance at heterozygous SNPs. B Schematic showing subset of allelically imbalanced H3K27ac peaks from total H3K27ac peaks in RCC (pRCC and ccRCC) samples. C AI at rs4903064 (chr14:73279420; DPF3) in RCC. D rs4903064 C-allele sequence context creates a HIF-2a binding site. E RPKM values for the EPAS1 ChIP signal in the peaks around rs4903064 in 43 samples. Genotype is shown on the X axis. Box boundaries correspond to 1st and 3rd quartiles; whiskers extend to a maximum of 1.5× the interquartile range. Two-sided unadjusted p-values are shown. Statistical analysis was performed using Kruskal–Wallis test. F Overlayed EPAS1 and H3K27ac ChIP-Seq tracks for seven samples within ~10 Kb of genomic coordinates of rs4903064. Individual sample genotypes are shown. G Venn diagram showing overlap of imbalanced H3K27ac peaks with ATAC-seq peaks in RCC. H Enrichment of risk SNPs from GWAS for RCC and other diseases in H3K27ac peaks with differential allelic imbalance, compared to all H3K27ac peaks. Empiric one-sided P value is indicated (unadjusted for multiple comparison). Statistical analysis was performed using Kruskal–Wallis test. GWAS risk SNPs rs7132434 and rs4733579 demonstrating allelic imbalance in RCC. I AI at Chr12 SNP rs7132423 and Chr8 SNP rs4733579 shown. J AI at Chr2 SNPs within EPAS1. rs46552601 and rs46541176 shown. K Chr2 SNPs within EPAS1, with imbalance plots split by histology. Adjusted Q values for imbalance are indicated. ChIP-seq chromatin immunoprecipitation sequencing, Ca cancer, HTN hypertension, Dx diagnosis, T2D type 2 diabetes mellitus, AID autoimmune disease, C0 reference allele, C1 alternate allele, AI allelic imbalance, RCC renal cell carcinoma, chRCC chromophobe RCC, ccRCC clear cell RCC, pRCC papillary RCC, SNP single-nucleotide polymorphism.