A direct comparison of ARfl and
A direct comparison of ARfl- and ARv7-dependent transcrip-tomes in the absence of hormone revealed significant differ-ences between the two AR isoform-specific transcriptomes (Table S2), with no significant correlation between them (R [Pearson] = 0.195) (Figure 2D), although expression of some genes, including the canonical AR targets KLK2, KLK3, and IGF1, was activated by ARv7 and ARfl (Figure 2D). This observa-tion conflicts with the hypothesis that ARv7 simply acts as a constitutively active form of ARfl (Li et al., 2013), but instead sug-gests that ARv7 and ARfl have different transcriptional roles in CRPC (Hu et al., 2012).
Figure 1. LNCaP95 Cell Growth Is Dependent
on both ARfl and ARv7
(A) Top: schematic of the full-length AR (ARfl) and
the AR variant 7 (ARv7), with the N-terminal domain
ligand-binding domain (LBD), and cryptic exon
shARfl LNCaP cells using an N-terminal, pan-AR
antibody. Actin signals serve as a loading control.
(B) Proliferation assay of indicated cells grown in 2D
culture. Data are the mean of three independent
(C) Representative scanning electron microscopy
images of LNCaP95 cells after 7 days of growth in
(D) Quantification of 3D cell growth data in (C). Data
are the mean of four independent experiments
relative to day ±SEM.
dent’s t test. See also Figure S1.
ARfl and ARv7 Bind to the Same Sites in Chromatin and Heterodimerize
To further identify the direct targets of ARfl and ARv7, we exam-ined their respective cistromes using chromatin immunoprecip-itation sequencing (ChIP-seq) in LNCaP95 cells treated with and without DHT (10 nM) for 4 h. We utilized AZD 2281 specific to ARfl or ARv7 and, as a control, an antibody that recognized both ARfl and ARv7 (Figure S3A). To date, the number of pub-lished AR-V cistromes has been limited (Chan et al., 2015; Chen et al., 2018; He et al., 2018; Lu et al., 2015) and there have been no direct comparisons between ARfl and ARv7 cistromes in an endogenous setting. We first confirmed the specificity and ChIP suitability of the ARv7 antibody using coim-munoprecipitation (coIP; Figure S3B) and ChIP-seq following ARv7 KD (Figure S3C). We then carried out ChIP-seq for both AR isoforms and observed 3,497 binding sites for ARv7 and 12,389 binding sites for ARfl, in the absence of hormone (Figures 3A and S3C). DHT treatment increased the ARv7 cistrome 2-fold (n = 6,149), and the ARfl cistrome 5-fold (>60,000 sites). Although most ARv7 sites were contained within the ARfl cistrome, a small number of sites (n = 794 for vehicle and n = 465 for DHT) were exclusive to the ARv7 cistrome, suggesting that ARv7 might function independently of ARfl. To test this hypothesis, we compared the two AR cistromes (without DHT) with a cistrome using an N-terminal AR antibody that recognizes both AR iso-forms (Figures S3A and S3B). Peaks that were common between at least two of the three cistromes were defined as ‘‘high-confi-dence’’ AR-binding sites (n = 2,828) (Figures 3B and 3C). The remaining ARv7-unique peaks (n = 595), unlike the ‘‘high-confi-dence’’ ARv7 peaks, were only minimally affected by silencing of ARv7 (Figure 3C), despite efficient protein reduction (Figures 1A and S1A). Taken together, these results suggest that most of the exclusive ARv7 peaks are in fact artifacts rather than bona fide ARv7 chromatin-binding sites.
Given that the majority of ARv7-binding sites overlap with the ARfl cistrome (Figure 3A), we compared sites shared by ARfl and ARv7 (n = 2629) and sites occupied by ARfl only (n = 4,737) (Table S3). This revealed that the AR motif was more prevalent at ARfl/ ARv7 sites than at ARfl-only sites (Figure S3D). Although we did
not observe any differences in gene expression in response to ARfl KD for targets in the vicinity of ARfl/ARv7 or ARfl-only sites, shARv7-regulated genes associated with the two sites were significantly differentially expressed (Figure S3E). To investigate whether ARv7 and ARfl co-occupy the same genomic loci, we carried out sequential ARfl/ARv7 ChIP-re-ChIP experiments at select target genes. We detected positive signal enrichment (Fig-ure S3F), suggesting a potential functional interaction (i.e., heter-odimerization) of the two receptors. To investigate this further, we next employed acceptor photobleaching fluorescence reso-nance energy transfer (FRET) (Figure 3D). We observed strong FRET signals for ARfl/ARfl and ARfl/ARv7 or ARv7/ARfl interac-tions, but not for ARv7/ARv7 homotypic interaction. Although these results reinforce a model of ARfl and ARv7 heterodimeriza-tion, they do not establish codependent binding on chromatin, as this is not required for the FRET signal. To further investigate chromatin binding, we performed ChIP-seq of both AR variants in the shGFP, shARv7, and shARfl LNCaP95 cells. Here, KD of ARv7 significantly reduced ARfl chromatin binding in both the vehicle and DHT condition (Figure 3E). Concordantly, loss of ARfl also reduced ARv7 binding in both treatment conditions (Figure 3E). Similar results were obtained in 22Rv1 cells, where codependent binding of ARfl and ARv7 was observed at ARE-containing sites with high levels of both factors (Figures S3G– S3I). Taken together, these results suggest that ARfl and ARv7 form heterodimers and can modulate their respective DNA-bind-ing affinities.