Given the importance of AR signaling in CRPC, there has been a dedicated interest in dissecting the mechanisms of AR function after androgen deprivation. Many lines of evidence suggest that androgen-dependent AR signaling remains functional in CRPC. It is known that the serum in clinical CRPC is never totally androgen free, that residual androgens are present within the prostate at levels capable of activating the AR despite castration and that enhanced intratumoral androgen synthesis has been commonly observed in CRPC (25
). Furthermore, >50% of CRPC patients showing disease progression on initial lines of hormonal therapies remain responsive to further hormone manipulation (75
), suggesting that androgen-dependent AR function remains in CRPC. As a result, AR activity in CRPC has been assessed largely based on androgen-responsive reporters or prostate specific androgen (PSA) production. Next-generation drugs have targeted androgen-dependent AR signaling by inhibition of androgen synthesis (abiraterone) and block of AR ligand-binding (MDV3100) (5
). However, the heterogeneous and often transient response to these new anti-androgen therapies raises the question of whether and how AR-mediated gene transcription occurs in the absence of ligand binding.
Prostate cancer is a molecularly heterogeneous disease even within a single patient, and multiple mechanisms may co-ordinately contribute to CRPC progression. While ligand-dependent AR signaling continues to play an important role in the early stages of CRPC when residual androgen-mediated AR signaling is active, ligand-independent activation of AR may occur in an environment where androgen levels are below castrate levels following severe ligand-depriving therapies. Such therapies have been associated with complete elimination of testosterone in the tumor microenvironment (such as bone) and in some cases a loss of CYP17 (a key enzyme in androgen biosynthesis) in prostate cancer cells (78–80
). More importantly, the fact that all anti-androgen approaches eventually fail strongly demonstrates the need to identify and target alternative androgen-independent AR signaling pathways. We reason that androgen-dependent and androgen-independent AR signaling can coexist, and that the relative importance of these two pathways depends on local androgen levels, AR expression and other cellular contexts such as co-regulators (8
). The androgen-independent AR binding described here occurs at extremely low levels of androgen, which may provide a mechanism for CRPC to develop and survive in a truly androgen-free milieu.
Previous studies have identified AR binding events in the presence of androgen in CRPC cells (15
). In this study, we performed AR ChIP-seq in CRPC cells cultured in hormone-depleted media and identified a large number of robust androgen-independent AR binding events. Taken together, these results show that both androgen-dependent and -independent AR signaling play a role in CRPC. The identification of androgen-independent AR binding events does not diminish the importance of androgen-dependent AR signaling. In fact, the androgen-dependent pathway is still intact in CRPC cells and can be rapidly reactivated by androgen stimulation. The fact that androgen-dependent CRPC growth can be inhibited by blocking ligand binding using an AR antagonist further supports the role of androgen-dependent AR signaling in CRPC. In the absence of ligand, however, the AR is no longer directed to canonical AD-ORs, but persistently occupies genomic loci characterized by a pre-existing accessible chromatin structure. These open chromatin structures are often associated with constitutively active genes whose expression is unaffected by AR binding. Instead, AI-ORs interact with neighboring genes and regulate their expression through DNA looping. Androgen-independent AR binding activates a distinct set of cell-cycle genes that can drive cancer cell proliferation after androgen depletion. Although androgen stimulation does not diminish AR occupancies at AI-ORs, expression of AI-OR-associated genes may decrease, likely due to transcription squelching. Inhibition of androgen-independent pathways is accompanied by activation of androgen-dependent pathways, enabling cancer cell survival in the absence or presence of androgen.
Recent studies show that promoter–promoter interactions are widespread in human cells (71
), with many chromatin complexes spanning 150–200 kb. Our results suggest that AR-bound promoters interact with distal genes through a similar promoter-centered interaction. The AR may function as a bridge between two promoters and bring general transcription machinery from a highly active promoter to a distal target gene. An important question is how the AR is recruited to AI-ORs independent of androgen stimulation. Previous studies showed that AR protein is more active and stable in LNCaP-derived CRPC cells (C4-2 or C4-2B cells) compared with parental cells (38
). AR in C4-2B cells is also predominately localized to the nucleus, suggestive of intrinsic transcriptional activity. There is a growing body of evidence suggesting that the AR can be activated through a range of post-translational modifications (87
), which may provide an explanation for higher AR activity and ligand-independent DNA binding in C4-2B cells.
We conclude that androgen-dependent and androgen-independent AR signaling can coexist in CRPC, with their relative importance dependent on AR activity and androgen levels in tumor microenvironment. Androgen deprivation results in a dramatic alteration of genome-wide AR occupancies and reprogramming of AR-mediated gene expression. The androgen-independent AR signaling described here may be an important therapeutic target when androgen-deprivation therapy and anti-androgen treatments fail. More importantly, these results suggest a general mechanism whereby, hormone deprivation reprograms genome-wide hormone receptor binding and gene regulation.