Fig. 1

STARRseq identifies AR-dependent enhancers. a Schematic representation of AR STARRseq. In this, high confidence ARBS (n = 4139), non-AR enhancers (positive control; n = 500), and regions with ARE motifs but no AR binding (n = 2783) were captured from normal human DNA and cloned into hSTARR-ORI plasmid. The resulting plasmid library was transfected into LNCaP cells by electroporation. Following DHT/EtOH treatment, STARR mRNA was extracted and sequenced to quantify the enhancer-mediated rate of self-transcription at each region. b Strong androgen-dependent enhancer activity (blue) was observed at known AR binding sites (red; GSE83860) proximal to KLK3. c Enhancer activity of AR CREs with varying levels of STARRseq signal (n = 42) was validated with a luciferase assay (4 biological replicates ± SEM). A strong correlation is observed between luciferase and STARRseq signals. d Volcano plot of androgen-dependent changes in STARRseq enhancer activity for clinical ARBS, ARE motif alone, and non-AR enhancers. Significantly induced enhancers (LFC > 1, p-adj < 0.05) are highlighted in red. e Schematic representation of the different classes of AR enhancers. f Heatmap of STARRseq (blue) represented as LFC over input plasmid library. Publicly available ChIPseq of AR (GSE83860, pink), Pol2 (GSE28126, purple), and H3K27ac (GSE51621, pink) in EtOH or DHT-treated LNCaP cells is shown as reads per kilobase of transcript, per million mapped reads (RPKM). GROseq (GSE83860) shows the normalized LFC of either the positive (pink) or the negative (cyan) RNA strands. The heatmap is divided based on the functional classes of each enhancer class identified by STARRseq. g Density map of androgen-induced changes to H3K27ac ChIPseq and STARRseq at inactive and inducible AR enhancers