CRISPR screening identifies regulators of enhancer-mediated androgen receptor transcription in advanced prostate cancer

Abstract

Amplification of the androgen receptor (AR) locus is the most frequent alteration in metastatic castration-resistant prostate cancer (CRPC). Recently, it was discovered that an enhancer of the AR is co-amplified with the AR gene body and contributes to increased AR transcription and resistance to androgen deprivation therapy. However, the mechanism of enhancer activation in advanced disease is unknown. Here, we used CRISPR-Cas9 screening to identify transcription factors that bind to the AR enhancer and modulate enhancer-mediated AR transcription. We demonstrate that HOXB13, GATA2, and TFAP2C bind the AR enhancer in patient-derived xenografts and directly impact features associated with an active chromatin state. Interestingly, the AR enhancer belongs to a set of regulatory elements that require HOXB13 to maintain FOXA1 binding, further delineating the role of HOXB13 in CRPC. This work provides a framework to functionally identify trans-acting factors required for the activation of disease-related noncoding regulatory elements.

Keywords: CP: Cancer; CP: Molecular biology; CRISPR; CRPC; androgen receptor; enhancer; epigenetics; functional genomics; prostate cancer.

Figure 1.. CRISPR-Cas9 screen identifies transcription factors that modulate AR transcription

(A) Schematic of FACS-based CRISPR-Cas9 screen targeting human transcription factors using the LNCaP cell line with GFP knocked in in the 3′ UTR of endogenous AR. (B) Results of the screen indicating the top hits and corresponding log(false discovery rate) (log(FDR)) and STARS score. Red dashed lines indicate FDRs of 0.01 and 0.001. (C) AR expression as determined by RT-qPCR following suppression of indicated transcription factors by siRNA. Expression is shown as log fold change relative to nontargeting control siRNA after normalization to GAPDH. Data represent the average ± standard deviation of at least 3 biological replicates, and significance was determined by a Student’s t test. **p < 0.01 and ***p < 0.001. (D) AR protein expression as determined by immunoblotting following suppression of indicated factors by siRNA. Actin is used as a loading control. See also Figure S1.

Figure 2.. GATA2, HOXB13, and TFAP2C bind to the AR enhancer in cell line and patient-derived models of prostate cancer

(A) Schematic of AR enhancer and promoter (top) with locations of transcription factor binding motifs (bottom). (B) H3K27ac and transcription factor profiles from ChIP-seq performed in LNCaP cells. The LNCaP DNaseI hypersensitivity (DHS) track was downloaded from the ENCODE project. (C) H3K27ac and transcription factor profiles from ChIP-seq performed in indicated PDXs of metastatic prostate cancer. See also Figure S2.

Figure 3.. Candidate transcription factors affect AR enhancer acetylation, chromatin accessibility, and interaction with AR promoter

(A) ChIP-seq H3K27ac profiles in LNCaP cells at the AR enhancer (left) and promoter (right) following siRNA of indicated transcription factors or nontargeting control (siCTRL). (B) DNA accessibility as determined by ATAC-seq in LNCaP cells at the AR enhancer (left) and promoter (right) following siRNA of indicated transcription factors or nontargeting control (siCTRL). (C) Enhancer-promoter interactions as determined by 4C-seq anchored at the AR enhancer (VP: Enhancer) or promoter (VP: Promoter). 4C profiles represent overlays of nontargeting control siRNA (siCTRL) with indicated transcription factor siRNA. VP: 4C-seq viewpoint. The LNCaP DNaseI hypersensitivity (DHS) track was downloaded from the ENCODE project. See also Figure S3.

Figure 4.. HOXB13 loss causes genome-wide enhancer and FOXA1 reprogramming

(A) Scatterplot of H3K27ac ChIP-seq peaks after treatment with indicated siRNAs. Differential peaks with FDRs <0.05 are indicated in magenta. (B) Correlation heatmap of H3K27ac peaks following suppression of indicated transcription factors by siRNA or a nontargeting (NT) control. Subscript indicates the replicate number. (C) Correlation heatmap of gene expression as determined by RNA-seq following suppression by the indicated siRNA. (D) Profile plots and heatmaps of H3K27ac signal at regions that were significantly decreased or increased by HOXB13 suppression. (E) Profile plot of average HOXB13 ChIP-seq signal (top) and top enriched motifs (bottom) at H3K27ac sites that show significant decreased or increased signals. (F) Profile plot and heatmap of FOXA1 ChIP-seq signal at H3K27ac decreased and increased sites. See also Figure S4.