br See also Figure S and Table S br C
See also Figure S3 and Table S3.
gene Bortezomib (PS-341) in each individual KD cell line. We could only evaluate transcriptional effects mediated by siRNA corepressor depletion within the background of a specific shRNA line (i.e., shGFP, shARv7, and shARfl), because the shRNA-mediated ef-fects were much stronger than the transient siRNA effects. When this was done, we observed that the transcriptional re-sponses to siNCOR1 or siNCOR2 were strongly attenuated within the shARv7 and shARfl cell lines, whereas this was not the case in the control cell line (shGFP) containing both ARv7 and ARfl (Figures 4C, S4F, and S4G). This suggests that NCOR-mediated transcription is, at least in part, dependent on the presence of ARv7 and ARfl. Consistently, KD of NRIP1, which could not be validated convincingly as an inter-actor of ARv7, did not produce this effect (Figure S4H). Taken together, these findings suggest that ARv7 mediates its repres-sive function by preferentially interacting with corepressors NCOR1 and NCOR2.
ARv7 Negatively Regulates H3K27ac
To further characterize the mechanism whereby ARv7 nega-tively affects transcription, we utilized ChIP-seq to assess levels of histone H3K27 acetylation (H3K27ac), a mark of active enhancers and transcriptional activity (Nord et al., 2013; Rada-Iglesias et al., 2011; Wang et al., 2016) H3K27ac cistromes were determined in the AR KD cell lines, and signals centered at ‘‘high-confidence’’ AR-binding sites (union of ARN, ARv7, and ARfl peaks) were analyzed (Figure 5A). To correlate H3K27ac levels with AR function, we stratified the averaged H3K27ac signals at the AR sites using Pearson correlation. This allowed us to distinguish two clusters with distinct H3K27ac signals (Figures 5A and S5A). Cluster I (n = 3,284) showed decreased H3K27ac levels after ARv7 and ARfl KD, indicative of ARfl and ARv7 positively regulating H3K27ac at these sites. In contrast, H3K27ac levels in cluster II (n = 4,268) were significantly increased upon ARv7 KD, yet decreased upon ARfl KD. This suggests that genes in cluster II sites are mostly differentially regulated by the two AR iso-
Figure 4. ARv7 Binds to Transcriptional Co-repressors NCOR1, NCOR2, and NRIP1
(A) MARCoNI assay (using a pan-AR antibody) of select corepressor peptides and cell lysates from indicated LNCaP95 cells. Results are the mean of three experiments ±SD. *p % 0.05, ***p % 0.001, Student’s t test.
(C) Heatmap of the union of dysregulated genes (adjusted p value <0.05) in response to NCOR1 or NCOR2 KD (siNCOR1 or siNCOR2) relative to con-trol (siCtrl) in the background of shGFP, shARv7, and shARfl cells.
See also Figure S4.
forms. To test this hypothesis, we exam-ined the ARv7- and ARfl-binding sites (Figures 5B and 5C). Although no sub-
stantial difference in ARv7 or ARfl binding was apparent be-tween the two clusters, we observed diminished signal intensities in response to KD of either AR isoform. This indi-cates an interdependent binding of ARv7 and ARfl, in agree-ment with our previous finding (Figure 3E). We next correlated the cluster-specific AR cistromes with the previously deter-mined AR transcriptomes (Figures 2A and 2B). As this analysis depends on a stringent peak-to-gene association, we only considered targets significantly dysregulated (DEseq; p < 0.05) upon AR KD, localized within 10 kb of an AR-binding site (Table S4). For cluster I targets, the fold changes for the majority of genes was <0 (Figure 5D), consistent with these genes being activated by either AR isoform. In contrast, cluster
II was biased toward shARv7 upregulated (fold change >0) and shARfl downregulated genes (fold change <0) (Figure 5D), which suggests that cluster II is predominantly associated with ARv7-repressed genes. These findings indicate that ARv7-dependent gene repression is a consequence of ARv7-mediated inhibition of H3K27ac.
To identify additional genomic factors involved in ARv7-dependent repression, we examined the underlying DNA se-quences in each cluster. Although we identified the AR-binding motif as the top sequence in both clusters, its enrichment was much higher in cluster I than in cluster II (Figure 5E). This sug-gests that the differences in the isoform transcriptomes may be influenced by the strength of the AR-binding motif. The sec-ond most common binding motif identified was that of FOXA1, a known determinant of AR action (Jin et al., 2014; Sahu et al., 2011). In contrast to the AR motif, enrichment of the FOXA1 motif was identical across the two clusters (Figure 5E). FOXA1 binding increased across both clusters in response to AR isoform deple-tion, albeit significantly more for ARv7 compared with ARfl KD (Figure 5F), with no apparent changes in FOXA1 level (Fig-ure S5B). In addition, significantly higher FOXA1 levels were also observed at ARfl and ARv7 shared sites (n = 2,629) compared with ARfl-only sites (n = 4,737) (Figures S3D, S3E, and S5C). Combined, these results suggest that ARv7 (and to