The androgen receptor (AR) is a member of the nuclear hormone receptor family of transcription factors,1 necessary for the normal growth, terminal differentiation and function of male urogenital structures, including the prostate gland. During development, the AR is expressed in the urogenital sinus mesenchyme, thereby facilitating androgen action to induce epithelial differentiation and morphogenesis.2, 3, 4, 5, 6, 7 Although AR activity is not required for epithelial cell differentiation, it is subsequently required for normal epithelial cell function.7, 8
The AR plays an important role in both early and advanced stages of prostate cancer etiology. Prostate cancer is the most commonly diagnosed invasive cancer in men in the United States and other Western countries,9 generally arising within the prostate luminal epithelial cell compartment. While 80% of men with this disease will not require treatment or will be cured by surgery and/or radiation, approximately 20% of patients will develop metastatic cancer necessitating systemic therapy. Currently, most therapies target the production or action of testicular androgens (i.e., androgen ablation therapy, AAT) that act via the AR to provide the critical growth and survival signals to prostate cells.
The AR is expressed to some degree in nearly all primary prostate cancers,10, 11, 12 and studies in both humans and animal models suggest a relationship between the cellular AR level in both primary and metastatic lesions and subsequent disease progression.13, 14, 15 The levels of AR in prostate cancer tissues from patients with progressive disease following treatment with AAT [i.e., androgen depletion independent (ADI) disease] are often comparable with the levels observed in the primary tumor, or even increased, thereby contributing to enhanced sensitivity to available ligand.16, 17 While the duration of response to AAT is variable and dependent on the stage of disease at treatment, prior treatments and the individual tumor characteristics, unfortunately progression to ADI disease is almost inevitable and has been attributed to aberrant AR signaling.15, 18, 19, 20 Importantly, it should be emphasized that while AAT usually results in rapid epithelial cell death, tumor regression and favorable initial clinical response,21, 22 loss of androgen action in the stromal compartment is also an important consequence of AAT and results in perturbations in stromal–epithelial communications that dictate prostate glandular homeostasis which could facilitate disease progression.
Somatic mutation in the AR gene is associated with and likely contributes functionally to prostate cancer progression. Although mutations in AR are not commonly found in early stage prostate cancer,23, 24 they are indeed present prior to hormone therapy in a proportion of the tumor cells,25 and have been identified with greater frequency in patients with ADI disease.25, 26, 27 This supports the hypothesis that somatic mutations in AR that arise with stochastic frequency in early disease may provide a selective growth advantage that facilitates progression. Typically, mutations in the AR open reading frame broaden the specificity of the receptor to other ligands or alter coregulator interactions, thereby enhancing the receptor's ability to maintain sufficient function in a low androgen environment for maintenance of prostate cancer cell growth and survival (reviewed in Ref.28). A similar response can be achieved through AR gene amplification, increased protein stability, altered levels and function of AR coregulators and activation of AR by growth factors or other kinase pathways.15, 28, 29, 30, 31, 32 Unfortunately, it has been difficult using clinical specimens alone to determine the exact mechanism whereby AR function contributes to the pathogenesis of spontaneous primary and metastatic prostate cancer.
One important concept emerging from work with genetically engineered mouse models is that neoplastic growth of the prostate may require abrogation of normal AR function, either directly through changes in AR structure and function or indirectly via changes in the AR signaling axis.33 Key to understanding this hypothesis is to appreciate that AR does not act independently to mediate tissue-specific responses to androgen, but requires interaction with a distinct profile of cell-specific as well as ubiquitous cofactors.34, 35, 36 Either autonomous changes in AR itself or nonautonomous changes in AR cofactors could lead to aberrant AR signaling. While the ability of many cofactors to modulate AR action has been studied extensively in both mesenchymal and epithelial compartments of the benign prostate,37, 38, 39, 40, 41, 42 their role in prostate cancer has largely been unexplored. Hence the goal of this article will be to review the evidence that AR coregulators can act as contributing or causative factors in the genesis and/or progression of prostate cancer.
Androgen receptor structure
The human AR gene is composed of 8 exons encoding a modular protein of 919 amino acids that can be divided into 4 structurally and functionally distinct domains: an amino-terminal transactivation domain (NTD), a DNA-binding domain (DBD), a hinge region and a ligand-binding domain (LBD). Whereas the DBD and LBD of the AR have a highly ordered canonical structure conserved amongst species, there has been significant divergence during evolution of the NTD and hinge regions allowing for different homeostatic control and complexity of AR signaling between organisms.43
The AR NTD represents approximately half of the receptor coding sequence. In contrast to other steroid receptors, which use the activation function 2 (AF2) surface in the LBD, the AR-NTD mediates the majority of AR transcriptional activity and is the most active coregulator interaction surface.44, 45 The AR-NTD is composed of 2 transcription activation units (TAU): TAU-1 (amino acids 1–370) identified as being required for full agonist-stimulated AR activity, and TAU-5 (amino acids 360–528) that confers constitutive AR activity in the absence of a LBD.46 More recent evidence indicates that an interplay between a core TAU-1 domain (amino acids 173–203) and TAU-5 is necessary and sufficient for AR transcriptional activity in some cell lines.47
Within the first 30 amino acids of the AR-NTD resides a highly conserved 23FQNLF27 peptide that confers interaction with the AF2 surface in the LBD following agonist but not antagonist binding, the so-called N/C interaction.48, 49, 50, 51 The human AR-NTD also contains long polyglutamine (amino acids 58–78) and polyglycine (amino acids 465–488) repeats, which are polymorphic and influence the function of the receptor.52 The mouse AR, by contrast, contains a polymorphic CAG repeat further downstream between amino acids 193–213, which is interspersed with basic, hydrophilic histidine residues. Although the AR NTD is described as a relatively unstructured domain, recent evidence indicates that the structure of the NTD changes upon binding proteins or to DNA.53, 54 This raises the possibility that the AR-NTD serves as a flexible platform for the recruitment and assembly of coregulators and members of the transcriptional machinery, and may serve as the primary mediator of the cell and gene specific effects of androgens.
The cysteine-rich DBD contains two zinc finger motifs and a short C-terminal extension that forms part of the hinge.55, 56 The first zinc finger mediates DNA recognition through interaction with specific base-pairs in response elements, facilitating binding of the receptor in the major groove of DNA.57 Conserved amino acids in the second zinc finger stabilize the DNA bound receptor complex and mediate dimerization between steroid receptor monomers.56, 58, 59 The AR binds as a head to head homodimer on either the consensus hormone response element, which comprises inverted hexameric half sites (5′-AGAACA-3′) separated by 3 spacer nucleotides, or on selective AREs consisting of 2 hexanucleotide half sites (5′-AGAACA-3′) arranged as direct repeats in promoters of androgen regulated genes such as probasin (PB; PB-ARE-2),60 the secretory component (SC; ARE1.2),61 and the mouse sex-limited protein.62 The AR head-to-head homodimer arrangement has been postulated to stabilize the AR on the DNA, allowing additional contacts between homodimer members.63, 64, 65 The AR hinge region situated between the DBD and LBD is encoded by the proximal region of exon 4 and contains a bipartite nuclear localization signal and important sites for phosphorylation, acetylation and degradation.66, 67, 68, 69, 70
The LBD, encoded by the distal part of exon 4 and exons 5–8, mediates high affinity binding of the AR to androgenic ligands. The recently solved crystal structure of the AR-LBD revealed a canonical ligand binding pocket formed by the ordered arrangement of 12 conserved alpha-helices.71, 72 Agonist binding induces a conformational change in the LBD and the folding of helix 12 back across the ligand-binding pocket. These changes result in the formation of the AF2 surface that for the AR, unlike other steroid receptors, predominantly mediates interaction with the FQNLF peptide of the AR-NTD in addition to binding coregulators containing similar sequences.73, 74 In this manner, the AR-NTD competes for the AF2 surface with certain coregulators, providing a mechanism for divergent determinants of AR transcriptional activity and the dominance of the TAU-1 and TAU-5 activation domains.75
The human AR upstream promoter lacks a classical TATA box but contains Sp1-binding sites, a cAMP-response element and an NF1 (nuclear factor 1) site.76, 77, 78 The AR can regulate its own expression via four exonic AREs located in exons D and E,78, 79 although this is in part controlled by consensus response elements for the transcription factors Myc and Max.78 Recently, the Forkhead transcription factor, FOXO3a, was identified as a positive regulator of AR gene expression via a consensus DNA-binding sequence in the AR promoter. Importantly, phosphorylation by the PI3K/Akt pathway leads to inactivation of FOXO3a that would in turn antagonize AR expression. The AR is also post-transcriptionally regulated by the PI3K/Akt target mTOR that regulates mRNA translation initiation through the ribosomal S6 kinase and initiation factor eIF4.80 Hence it would be expected that strategies designed to inhibit the Akt pathway would lead to increased nuclear accumulation of FOXO3a and enhancement of AR gene transcription.78 While FOXO3a is a positive regulator of AR, it is interesting that another member of the FOXO family, FOXO1A (FKHR), can antagonize the ability of AR to regulate the transcription of target gene expression or stimulate proliferation.81 Conversely, activated AR can inhibit FKHR binding to DNA and expression of FAS ligand, which is known to cause prostate cancer cell apoptosis and cell cycle arrest.82 These findings are indicative of reciprocal antagonism between AR and FOXO1A.
Once transcribed, unliganded AR is maintained primarily in the cytoplasm in a complex with heat-shock proteins and cochaperone molecules. The minimal chaperone complex required for the efficient folding and stabilization of steroid hormone receptors consists of Hsp70 (hsc70), Hsp40 (Ydj1), Hop (p60), Hsp90 and p23.83, 84, 85, 86 These interactions probably maintain the AR in a stable, partially unfolded state primed for high affinity interaction with androgenic ligands, including the most potent natural androgen, 5α-dihydrotestosterone (DHT).87 While the intrinsic activity of the cytoplasmic AR complex is not yet understood, upon ligand binding, the chaperone heterocomplex is known to mediate receptor trafficking to the nucleus, possibly by facilitating interaction with the cytoplasmic protein dynein that drives rapid and active transport of the receptor to the nucleus along the cytoskeleton.88, 89, 90 The Hsp70 proteins and potentially Ydj1, a putative component of the matrix lamina pore complex, may also facilitate the translocation of the receptor across intranuclear membranes by mediating the unfolding and refolding of AR.89, 91, 92 Once in the nucleus the chaperone complex dissociates from the receptor,88 but continues to play a role in receptor movement and may mediate the cyclic assembly/disassembly of the AR-transcription factor complex by promoting the dissociation of hormone from the receptor and returning it to a primed state ready for reactivation should hormone become available. Without continued presence of hormone, AR action is rapidly terminated.86, 93, 94
A cascade of ligand-loaded and AR initiated protein–protein interactions likely causes the stepwise remodeling of chromatin structure at target promoters, recruitment of basal transcription machinery and initiation of RNA polymerase activity.95, 96 These events can be modulated by a subset of the 130-plus putative AR interacting coregulators that can promote (coactivators) or inhibit (corepressors) AR function. These coregulators are broadly divided into 4 main types: (i) molecular chaperones that coordinate AR maturation and movement, (ii) histone modifiers (e.g., CBP/p300, NCoR), (iii) coordinators of transcription (e.g., TRAP/DRIP/ARC) and (iv) DNA structural modifiers (e.g., SWI/SNF/BRG1). It should be noted that many of the putative AR coregulators were identified using physical interaction-dependent yeast-two-hybrid screens with discrete AR receptor domains as bait, and there is a paucity of data concerning their ability to influence AR regulated gene-specific expression. This is largely due to the fact that the transcriptional space for any AR-coregulator pair (i.e., the set of gene targets available to and potentially regulated by a given AR-cofactor complex in any given cell context) is essentially unknown. However, once the transcriptional space is defined for AR and each coregulator, it should be possible to more closely define the precise role of a given cofactor in mediating AR function in different physiologic states including development, growth, aging and diseases (e.g., prostate cancer). A partial list of the AR coregulators cited within this review and their corresponding AR interaction domains is provided in Figure 1 and Table I.
Another possible mechanism whereby AR signaling can be corrupted involves genetic recombination events that would place a structural gene (e.g., for a transcription factor or cofactor) not normally under androgen control under direct transcriptional control of the AR. Indeed, a recent report describes how members of the ETS family can become AR regulated through fusion with the androgen regulated TMPRSS2 gene.132 While the causal relationship between this event and prostate cancer remains to be defined, such translocations underscore how hijacking AR signaling might lead to deregulated cellular growth. A comprehensive discussion of spontaneous mutations in AR targets remains the topic for another review.
Role of coregulators in AR transactivation and prostate cancer
The formation of an active AR-directed preinitiation complex most likely occurs via the sequential recruitment of coregulators with distinct activities to the ligand bound nuclear AR. The first identified and most widely understood of these coregulators are the p160 coactivators. This small family consists of 3 160-kDa proteins, namely steroid receptor coactivator 1 (SRC1); transcriptional intermediary factor 2 (TIF2) and its mouse homolog glucocorticoid receptor interacting protein 1 (GRIP1); and amplified in breast cancer 1 (AIB1/SRC3).133, 134, 135, 136, 137 (Note: All cofactor abbreviations and synonyms are listed in full in Table II). The p160 coactivators have been shown to bind to TAU-5 of the AR NTD, an interaction that is influenced by TAU-1,47 and to the AF-2 surface.138, 139, 140 Their recruitment directly influences AR transactivation capacity via intrinsic histone acetyltransferase activity,141 and indirectly by acting as platforms for the recruitment of secondary coactivators possessing chromatin remodeling and protein acetyltransferase capabilities, such as CBP/p300142 and pCAF,143 or protein methyltransferase activity like CARM1 or PRMT1.144, 145 CBP/p300 has been shown to contribute to transcriptional activation by remodeling chromatin and recruiting the basal transcription factors, TFIIB and TBP.134, 145, 146 Nucleosome remodeling complexes such as SWI/SNF are subsequently recruited147 followed by a multisubunit Mediator complex that serves to bridge DNA-bound transcription factors with the general transcriptional machinery, particularly RNA polymerase II.148 In addition, the AR has the capacity to interact directly with members of the transcription machinery, including components of the RNA polymerase II complex107 and TFIIF,101 the latter promoting conformational changes in the AR-NTD.149
Table II. List of AR Cofactors
Amino-terminal enhancer of split
Amino-terminal enhancer of split
Amplified in breast cancer-1/Cbp-interacting protein/ ACTR/ receptor-associated coactivator 3/thyroid hormone receptor activator molecule 1/steroid receptor coactivator protein 3
Nuclear receptor coactivator 3
Amplified in breast cancer-3 protein/activating signal cointegrator-2
Nuclear receptor coactivator 6
Androgen receptor-associated protein 24/Ran, member ras oncogene family
thyroid hormone receptor associated protein mediator complex
Zinc finger-containing, miz1, Pias-like protein on chromosome 10
Retinoic acid induced 17
Alterations in cofactor levels and function have been proposed to contribute to the emergence of ADI disease.15 Immunohistochemical evaluation of the three p160 coactivators in benign and cancerous prostate tissues demonstrated over expression of SRC1 in approximately half of prostate tumors grown in an intact androgen environment, and high levels of both TIF2 and SRC1 in the majority of recurrent ADI tumors.150 Moreover, a correlation between increased levels of AIB1 and prostate tumor grade and stage has been identified in clinically localized disease.151, 152 In tissue samples from patients with biopsy-proven prostate cancer who underwent prostatectomy, p300 levels not only correlated with in vivo proliferation, but predicted larger tumor volumes, likelihood of extra-prostatic extension, seminal vesicle involvement at prostatectomy as well as progression after surgery.153 Levels of CBP were highly expressed in advanced prostate cancer and, in particular, in tissues from patients that failed endocrine therapy.154 These two latter observations are of particular interest given a report that p300 and CBP can acetylate AR at highly conserved lysine-rich motifs required for maximal DHT-induced transcription.118
When coactivator levels increase, AR function most likely will change as a consequence. For example, an environment enriched for coactivators could make AR more sensitive or responsive to low levels of agonist,155 or allow promiscuous activation of AR by abundant yet low-affinity androgenic ligands such as the adrenal androgens androstenedione and dehydroepiandrosterone.150 It still remains to be determined, however, if or how changes in p160 family or CBP/p300 influence the expression of specific AR target genes, whether this is AR-specific or AR-dependent, and if a causal relationship exists between cofactor expression levels and the initiation or progression of disease.
Members of the Cdc25 family of dual-specificity phosphatases that activate cyclin-dependent kinases to enable cell cycle progression are differentially expressed in prostate cancer. Cdc25B has been shown to interact directly with AR in a hormone dependent manner but independent of its cell cycle function.156 Over expression of Cdc25B has been observed in cancer cell lines and human cancers157, 158, 159 and correlated with histological prostate tumor grade and frequently with more poorly differentiated tumors. However, as the Cdc25 family has also been implicated in cancer through cooperation with known oncogenic factors such as Ras and loss of Rb1,160 it is currently unclear whether Cdc25B acts dependent or independent of AR in prostate cancer. Other cofactors such as cyclin G-associated kinase (GAK) interact with the AR NTD independent of hormone and can enhance both intrinsic NTD transcriptional activity and ligand-dependent AR activation. Levels of GAK can increase with prolonged AAT and in patients with ADI disease.161 The Tat-interactive protein 60 kDa (Tip60) has also been implicated in prostate cancer progression and the development of ADI disease following long-term AAT. Originally identified as a coactivator for the human immunodeficiency virus type I-encoded TAT protein, this ligand-dependent coactivator acts on the AR, PR (progesterone receptor) and ER (estrogen receptor).123 Tip60 has been found concentrated in the nucleus in biopsies of ADI disease and increased levels and nuclear accumulation of Tip60 have been observed following androgen withdrawal in CWR22 xenografts and LNCaP cells. While levels of Tip60 decrease with androgen treatment and localize to the cytoplasm,162 a recent independent study indicated that Tip60 is downregulated in metastatic cancer cells at both RNA and protein levels,163 which may represent a stochastic tumor-specific change, or indicate different roles of this coregulator in AR transcriptional activity at different stages of prostate cancer progression.
Gelsolin (GSN) is an actin severing protein and another AR cofactor implicated in cancer.164 GSN binds the AR DBD and LBD during nuclear translocation and, despite the lack of an intrinsic nuclear localization signal, colocalizes with the receptor in the nucleus in a ligand dependent manner.114 Overexpression of GSN enhances AR transcriptional activity in the presence of either androgen or the nonsteroidal AR antagonist hydroxyflutamide.114 Immunohistochemical analysis of tissues representing prostate adenocarcinoma, prostatic intraepithelial neoplasia (PIN), and benign prostatic hyperplasia (BPH) has demonstrated decreased expression of GSN when compared with levels in nonproliferative tissues.165 GSN levels are also decreased in breast, lung and bladder cancers.166, 167, 168 Interestingly, a study by Nishimura et al. found that expression of GSN was up regulated in LNCaP cells, LNCaP xenografts and human prostate tumors following androgen depletion.114 Therefore, it is tempting to speculate that changing levels of GSN in response to AAT might differentially enhance nuclear translocation and transactivation of antagonist-bound AR. However, because GSN also regulates cytoskeleton reorganization, changes in cell morphology and motility,169 it may play a role in actin-dependent cellular processes including growth and differentiation. It should be mentioned that GSN is cleaved by caspase-3 during apoptosis170 and results in an amino-terminal GSN fragment that has an unregulated actin filament severing activity capable of mediating both morphologic changes and nuclear fragmentation associated with apoptosis.171
The recently identified AR coactivator AR trapped clone-27 (ART-27) is expressed in normal differentiated prostate epithelial cells and benign and premalignant (HG-PIN) epithelium. ART-27 is a nuclear protein that interacts with the AR NTD and enhances its transactivation when overexpressed in a variety of cultured mammalian cells. While ART-27 was shown to enhance androgen-mediated transcription of PSA, enforced expression in LNCaP cells inhibited proliferation in response to androgen.39 Levels of ART-27 drastically decreased in prostate cancer specimens compared with nonmalignant tissue, possibly reflecting the state of glandular differentiation.39 Indeed, the functions of ART-27 are consistent with tumor suppressor/prodifferentiation genes and may explain why the expression of ART-27 decreases dramatically during prostate cancer progression.
ARA70/ELE1, a non-AR specific coregulator, can enhance hydroxyflutamide-stimulated AR activity via an FXXLF domain. In one report, levels of ARA70 protein were found to increase in high grade prostate carcinomas, prostate cancer cell lines and xenografts.172 In contrast, other studies have shown expression of ARA70 mRNA to decrease in prostate tumor tissue,40 increase in response to hormone deprivation173 or not to change between normal and prostate cancer in the same tissue.174 Because of an ability to interact with other nuclear hormone receptors, however, the significance of the AR-ARA70 interaction in prostate cancer progression remains to be clearly defined. It is interesting to note that another AR coregulator, ARA160/TMF1, interacts with the AR NTD and cooperates with ARA70 to enhance AR activation by nonclassical ligands,129, 173 but levels of ARA160 mRNA expression are similar in normal and prostate tumor tissue.40
Ran/ARA24 interacts with the TAU-1 region of the AR-NTD and such binding decreases with increasing poly-Q length.97 The involvement of ARA24 in prostate cancer progression is inconclusive since ARA24 RNA expression has been found elevated in the early stages of primary prostate cancer in one study40 but at similar levels in BPH as in primary and hormone refractory carcinomas in another study.175
ARA55 (Hic-5) is a member of the group 3 subfamily of LIM domain proteins, and is preferentially expressed in prostate stromal cells.40 ARA55 binds to the AR in a hormone-dependent manner through its C-terminal LIM domains, and results in increased AR activity and altered specificity of receptor binding to alternate ligands.121, 176 Expression of ARA55 mRNA is reportedly lower in nonmalignant prostate tissue compared with adjacent tumoral tissue,174 and in hormone refractory prostate cancer compared with untreated tumors or BPH. Increased expression of ARA55 has been associated with both shorter recurrence free survival and overall survival in hormone refractory prostate cancer patients.177
Perhaps two of the best characterized AR corepressors are the nuclear receptor corepressor (NCoR1) and the silencing mediator of retinoid and thyroid hormone receptor (SMRT/NCoR2). Recruitment of NCoR1 and SMRT by agonist and antagonist-bound AR has been shown to suppress agonist-induced activation of androgen regulated genes including PSA, TSC22, NKX3.1 and B2M. Because these corepressors compete for the same AR binding surfaces as key coactivators including the p160 family, their relative expression is an important factor that can determine the level and consequence of AR signaling in prostate cancer cells.178, 179 Indeed, recruitment of corepressors is believed to be a key mediator of antagonist mediated inhibition of steroid receptors including the AR,64, 99, 178, 180, 181 suggesting that loss of expression of corepressors could facilitate tumor growth in the presence of antagonists used in conjunction with AAT.182
A number of studies have attempted to compare the steady state levels of coregulator transcripts from a variety of clinical samples, prostate cancer cell lines and xenograft models.37, 40, 183 Given the remarkable heterogeneity of prostate cancer samples examined it was not surprising that there was back of concordance between data sets. For example, when the expression of 8 cofactors known to interact with AR (ARA160, ARA70, ARA55, ARA54, ARA24, PIAS1, SRC1 and TRAP220) were examined by quantitative in situ hybridization in 43 primary human prostate cancer tissues, only the levels of PIAS1 and ARA24 were found to be increased while levels of ARA70 actually decreased.40 In an independent study, expression of 16 coactivators and corepressors known to interact with AR (SRC1, β-catenin, TIF2, PIAS1, PIASx, ARIP4, BRCA1, AIB1, AIB3, CBP, STAT1, NCoR1, AES, cyclin D1, p300 and ARA24) were examined in prostate cancer cell lines, xenografts and clinical specimens and found to be expressed at similar levels in BPH and in untreated and hormone-refractory carcinoma by quantitative reverse transcriptase-polymerase chain reaction. In one report the levels of expression of PIAS1 and SRC1 were significantly lower (approximately 2-fold) in samples representing ADI disease than in untreated primary disease,175 while in another study ARA55 and SRC1 were found to be elevated in higher grade prostate tissues and in patients with poor response to endocrine therapy, while levels of TIF2 and AIB1 were consistently low in all samples tested.183
To determine whether any coregulators were consistently upregulated in prostate cancer we examined Oncomine datasets in which patterns of gene-expression alterations were compared between prostate tumor and normal prostate samples by oligonucleotide array analysis184, 185, 186, 187, 188, 189, 190, 191 (see Table III). None of the p160 family members (i.e., SRC1, TIF2 or AIB1) were consistently altered in cancer. Indeed, of 8 studies investigating levels of SRC1, 2 identified a significant increase in expression levels in tumor samples vs. normal (p = 9.4E-17 and 0.019, respectively),184, 185 and one revealed a significant decrease186 (p = 0.044). Moreover, 2 of 7 studies found TIF2 increased in cancer (p = 0.031 and 0.002, respectively)189, 191 while one reported a significant decrease (p = 6.5E-04).185 Only one study found AIB1 to be upregulated (p = 3.7E-05).190
Table III. Evidence that Steady State Coregulator Transcript Levels Differ Between Cancer and Normal Prostate
We next investigated levels of transcripts encoding AR coregulators deregulated in primary prostate cancer (PIAS1, ARA24, ARA70). With few exceptions, these coregulators were not significantly altered in cancer compared with nonmalignant tissue. ARA55 was significantly down regulated in the majority of studies; however, as ARA55 is preferentially expressed in stromal cells,40 this difference may be attributed to varying proportions of epithelial and stromal cells in samples of normal and malignant prostate, a fact that is not often considered in the data analysis. Additionally, Tip60 and GSN were downregulated in the majority of Oncomine datasets despite having been shown to be elevated in specimens of castration resistant cancer,114, 162 possibly reflecting the differential deregulation of these proteins following hormone withdrawal therapy.
In general, the transactivation potential of steroid hormone receptors derives from the capacity of agonist ligands to induce the correct formation of the AF-2 surface within the LBD, which in turn serves as a recruitment site for LxxLL motifs found in multiple coregulators. The AR actually diverges from this model since the majority of the AR transactivation capacity is believed to act through cofactor interactions in the NTD, a switch most likely driven by the evolutionarily conserved capacity of the AR to adopt the high-affinity N/C interaction upon agonist binding. Although occupancy of the AF-2 surface by the FQNLF peptide of the AR-NTD normally excludes cofactor binding at this site192 and serves to stabilize the receptor-ligand complex,193 the N/C itself appears to be dispensable for AR transactivation capacity in vitro,52 but perhaps not in vivo.194 Importantly, if the cellular levels of LxxLL/FxxLF-containing coregulators (e.g., p160s, ARA70, ARA55) increase, they would be expected to compete with N/C interaction for AF-2.75, 195 Although the consequence of this competition is not yet clear, shifting the responsibility for transactivation from the NTD to AF-2 potentially influences AR structure, composition of the transcription complex and target gene selectivity and specificity.
AR mutations and coregulator interactions
The incidence of somatic mutation in the AR in prostate cancer samples ranges from 20 to 40%,196 and over 60 mutations have been identified.197 As discussed above, somatic mutations in AR occur with low frequency (<10%) in primary prostate cancer but can be detected with greater frequency in tumor cells of distant metastases and recurrent prostate cancer following AAT198, 199 and may impart altered sensitivity and specificity.200 It is interesting that the majority of AR point mutations identified in clinical prostate cancer (79%) map to just 3 discreet regions within the LBD: amino acids 670 to 678 (at the boundary of the hinge and LBD), 701 to 730 (“signature sequence” loop between helices 3 and 4 of nuclear receptors involved in ligand recognition and specificity) and 874 to 910 (which flank AF-2).200, 201 Moreover, C-terminal mutations are more common to tumors identified before androgen ablation, while NTD mutations appear to occur with higher frequency following androgen ablation and predominantly collocate to either a region incorporating the polyQ tract or a small region of the transactivation unit TAU-5.52, 200, 201
The biological consequence of a specific somatic mutation in AR likely relates to the physical location of the mutation within the structural protein and the ability of that variant to influence higher order structure, macromolecular assembly and function. For example, naturally occurring mutations encompassed by the LBD signature sequence, AF-2, and the boundary of the hinge and LBD (670QPIF673, codons 670–678) result in AR variants that exhibit increased transactivation responses to both androgenic and nonandrogenic ligands.27, 28, 202, 203
Only a few of the AR mutations identified in prostate cancer have been investigated with regard to their ability to influence the N/C interaction. For instance, the A748T mutation identified in metastatic prostate cancer is located in helix 5 of the LBD and correlates with reduced N/C terminal interaction at lower concentrations of DHT (10−10, and 10−9 M), while normal N/C-terminal interactions are observed at high hormone concentrations.204 Additionally, a somatic mutation detected in a human prostate cancer that interrupts the AR polyglutamine tract by 2 nonconsecutive leucine residues (AR-polyQ2L) exhibits disrupted interdomain communication (N/C interaction) at all concentrations of DHT tested.52 Underscoring the finding by Hsu et al. that the strength of N/C interactions does not always correlate with effects on AR-mediated transactivation,205 the hormone-stimulated transactivation activity of the A748T mutant was substantially lower than wtAR, whereas that of AR-polyQ2L was substantially greater.52, 204 While the implications of these findings are not yet fully understood, they could indicate the relative importance of the N/C interaction at different promoters and in different cellular contexts. An important question is whether AR gene mutations alter the transcriptional space of the AR; or whether threshold levels of cofactors can influence the N/C interaction.
In addition to enhancing the affinity for different ligands, a somatic mutation in AR such as T877A may alter coregulator interactions and transactivation. The T877A variant was first identified in the LNCaP prostate cancer cell line206 and subsequently found at a high frequency in clinical prostate cancer. The T877A mutation broadens the capacity of AR to adopt activation-permissive conformational changes following binding of nonclassical ligands.72 The T877A variant is considerably less responsive to repression by the corepressor NCoR1 compared with wtAR,179 and responds better to potentiation by the coactivator, SRC1.207 These results underscore independent mechanisms by which a single mutation can lead to enhanced transcriptional activity of the AR.
A recent study identified a novel set of AR somatic mutations (G142V, D221H, E872Q and M886I) in prostate cancer samples.208 Although none of the receptor variants differed in their ability to bind the synthetic androgen, methyltrienolone, two of the variants, E872Q and M886I, displayed significantly increased transactivation capacity compared with wtAR in the presence of the coactivators TIF2 and CBP (but not p300), and were more efficiently repressed by NCoR1. The H874Y mutation in the highly conserved LBD motif (873LHQFTFDL880) was originally identified in a patient with hormone refractory prostate cancer. The mutation imparts AR responsiveness to adrenal androgens, nonandrogenic steroids and anti-androgens,209 and enhances the transcriptional response of the receptor to all of the p160 coactivators.209
A study by Li et al. examined the ability of coactivators to enhance AR transcriptional activity of AR NTD variants identified in prostate cancer and androgen insensitivity syndrome. Interestingly, two variants, E2K from androgen insensitivity syndrome and P340L from prostate cancer, displayed reduced transcriptional responses to the coactivator ART-27 compared with wtAR, even though the response to p160 coactivators TIF2 and SRC1 was unchanged. Coimmunoprecipitation studies demonstrated reduced interactions between ART-27 and AR-E2K compared with wtAR, but increased interactions between ART-27 and AR-P340L.210 These findings suggest a complex interplay between different regions of the AR-NTD and distinct coregulators, which influences the recruitment of and response to individual cofactors.
Spontaneous somatic mutations in AR have been identified in primary prostate tumors in the TRAMP (transgenic adenocarcinoma of the mouse prostate) model of prostate cancer200 that display differential activation in the presence of coregulators. For example, the E231G somatic variation in the evolutionarily conserved NTD signature motif that is unique to AR (in mouse, residues 229–242, in human, residues 234–247), displays increased transcriptional activity compared with wtAR in the presence of ARA70 or ARA160. Furthermore, unlike the wtAR the E231G variant responds to estradiol in the presence of ARA70.200 It is of critical importance to note that mice expressing a prostate-specific AR-E231G variant transgene developed PIN by 12 weeks of age, with progression to invasive and metastatic disease in 100% of mice by 50 weeks of age.33 In contrast, enforced expression of similar transgenes encoding wtAR or the AR-T877A variant did not induce any significant prostate abnormalities or cancer over the same time frame. This finding supports the theory that abrogation of the classical AR signaling axis can contribute to prostate cancer progression by altering the transactivation potential of AR at particular gene targets and/or influencing promoter specificity. In support of this hypothesis, a recent report has shown that AR-E231G exerts a negative effect on the recruitment of the COOH terminus of the Hsp70-interacting protein (CHIP) corepressor known to promote AR degradation.98 The interaction between AR and CHIP was found to be decreased by both the AR-E231G mutation and another mutation, AR-A229T, localized to the same conserved short N-terminal signature motif and isolated from another tumor arising spontaneously in TRAMP. Interaction between AR variants harboring the AR-E231G and AR-A229T mutations (human E236G and A234T) displayed reduced interaction with CHIP by 16 and 43%, respectively, when compared with wild type.98 These findings suggest that some AR variants may act as oncogenes to promote the initiation and/or progression of prostate cancer by virtue of their altered ability to interact with cofactors. Conversely, it is conceivable that mutated cofactors or altered cellular levels of cofactors may well be able to “unmask” the oncogenic potential of a wild type AR.
Somatic mutations in coregulators
The Wnt signaling pathway component and TCF/Lef-1 partner β-catenin was recently identified as interacting directly with AR and the transcriptional coactivator TIF2.211 Exon 3 of the β-catenin gene is a mutational hotspot in 5–7% of advanced prostate cancers that may act by stimulating TCF-dependent and/or AR-dependent transcription.212, 213 One particular variant, β-catenin S33F, has been shown to contribute to broadened ligand specificity of AR in cell-based assays resulting in increased transcriptional activation by androstenedione and estradiol and diminished inhibition by the AR antagonist bicalutamide.214 Despite the paucity of β-catenin mutations in prostate cancer,213, 215 there is evidence for increased β-catenin/TCF transcriptional activity owing to the upregulation of β-catenin target genes in prostate tumor samples and tumor-derived cell lines.216, 217, 218 In addition, increased cytoplasmic and nuclear β-catenin expression has been observed in 20–30% of prostate cancer specimens, with higher levels corresponding with higher grade cancers.219, 220 Upregulation of β-catenin/TCF target genes may be mediated by calpain which is reportedly capable of generating a stable NH2-terminal truncated form of β-catenin in both metastatic prostate cancer samples as well as prostate and cancer cell lines.221 Alternatively, stabilization of β-catenin may result from upregulation of the peptidyl–prolyl cis/trans isomerase PIN1 that binds to β-catenin and inhibits its association with adenomatous polyposis coli (APC) gene product and subsequent phosphorylation and glycogen synthase kinase 3β (GSK-3β)-mediated degradation.222 Increased expression of PIN1 correlates positively with clinical stage of prostate cancer and increased chance of disease recurrence223 and is elevated in metastatic disease.222
It has been known for some time that mutations in the two major hereditary breast cancer susceptibility genes, BRCA1 and BRCA2, that predispose women to early-onset breast and/or ovarian cancer224, 225, 226, 227, 228 are also associated with increased risk of developing prostate cancer.229, 230 Both BRCA1 and BRCA2 have been shown to enhance transcriptional activation of AR in a manner that is potentiated by the presence of TIF2.109, 112 It is unclear whether the ability of BRCA1 and BRCA2 to coexist in a holoenzyme complex and participate in a common DNA damage response pathway associated with the activation of homologous recombination and double-strand break repair is altered in prostate cancers.231, 232 Androgen signaling has been implicated as playing a role in the development and progression of breast cancer through the findings that androgen inhibits the proliferation of AR-positive breast cancer cell lines in culture.233, 234, 235, 236 Moreover, the AR level in primary tumors is a critical determinant of the response of breast cancer to hormonal manipulations,236, 237 and reduced or impaired AR signaling has been reported in cases of hereditary male breast cancer.238 Inactivating mutations in either BRCA1 or BRCA2 may then reduce AR signaling, resulting in altered proliferative potential of affected cells. Studies regarding the incidence of BRCA1 and BRCA2 mutations in prostate cancer more consistently implicate BRCA2 mutations as playing a role in prostate cancer progression. Mutations in BRCA2 are significantly associated with higher prostate cancer risk239 and have been identified in Ashkenazi Jewish, Icelandic populations and in prostate cancers identified at a younger age,240 whereas very few BRCA1 mutations are associated with high risk prostate cancer families.239, 241 Further studies should reveal whether BRCA2 mutations found in prostate cancer are associated with impaired homology-directed DNA repair and thus increased chromosomal instability and/or if the BRCA2 mutation is associated with loss of AR transactivation and decreased AR signaling potency.
Polypeptide growth factor signaling modifies AR action
There is considerable evidence supporting the ability of polypeptide growth hormones and cytokines to stimulate the transactivation potential of AR, a process known as signaling cross-talk; however, little is known about how AR or cofactor phosphorylation mediates this effect. Initial studies in LNCaP cells revealed that interleukin-6 (IL-6), insulin like growth factor (IGF-1), keratinocyte growth factor and epidermal growth factor (EGF) can each stimulate transcription of prototype AR target genes in the absence of androgens (Ref.242 and references therein). Indeed, the mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3 kinase/Akt pathways have both been implicated as mediating the effects of cytokine and growth factor stimulation on AR transactivation.30, 243, 244, 245, 246 Additionally, it has been shown that protein kinase A (PKA) and protein kinase C both augment ligand-dependent and ligand-independent AR transactivation.244, 247, 248 One potential mechanism by which AR transactivation may be increased is by enhanced cofactor recruitment in response to AR phosphorylation, although this still needs to be rigorously tested.
Phosphorylation of AR: MAPK, Akt, PKA
The AR has potential phosphorylation sites for casein kinase II, PKA, calcium calmodulin II kinase, protein kinase C, MAP kinases and Akt/PKB. Indeed, 6 serines are phosphorylated on AR in response to treatment with the synthetic androgen R1881 (serines 16, 81, 256, 308, 424 and 650) while one residue (serine 94) is constitutively phosphorylated and phosphorylation of Ser-650 is upregulated in response to treatment with hormone, forskolin, EGF and phorbol-12-myristate13-acetate.69
Phosphorylation of AR by MAPK or Akt can result in the sensitization of AR to low circulating levels of DHT such as those present during AAT.249, 250 Direct phosphorylation of AR by Erk-2 (at serine 514, not detected by Gioeli et al.) is associated with increased transactivation, increased prostate cell growth, increased interaction with the AR coregulator ARA70 and increased sensitivity to hormone levels.245, 250, 251 Direct phosphorylation by Akt has been shown to occur at residues Ser213 and Ser791, resulting in modulation of transcriptional activity in prostate cancer cell lines. In fact, Her-2/neu stimulation of AR signaling and prostate cancer cell survival and growth was shown to be dependent on activation of Akt.252, 253 Although one study implicates phosphorylation of AR by Akt at Ser-213 and Ser-791 in causing increased AR transactivation,252 another report suggests that agonist-dependent phosphorylation of AR at Ser-213 abrogates AR-mediated transcription.254 Additionally, there is data demonstrating that phosphorylation of AR by Akt suppresses AR activity in androgen-dependent LNCaP cells at low passage while enhancing AR activity in cells at high passage.255 It will be important to determine if the basis for the pleitropic effects of phosphorylation at Ser-213 in different cell types and at various stages of prostate cancer progression involves alternate recruitment of AR coregulators.
Several studies have addressed the effects on transactivation of AR in the presence of an activated PKA. It was previously shown that AR could be activated by the PKA activator, forskolin, in the absence of androgen.247 Furthermore, activation by forskolin could induce expression of PSA in androgen-depleted LNCaP cells in a manner dependent on AR that was blocked by the antiandrogen, bicalutamide.256 Although direct in vivo phosphorylation of AR by PKA has yet to be demonstrated, there is evidence for dephosphorylation of AR in response to forskolin stimulation on residues Ser-641 and -653.257
Phosphorylation of coregulators and AR transactivation
Phosphorylation of coregulators can increase recruitment of other coregulators and members of the basal transcription machinery to the transcription complex.258, 259 Ueda et al. found that SRC1 phosphorylation by MAPK was required for optimal ligand-independent activation of AR by IL-6.260 Similarly, the transcriptional activity of the coregulator AIB1 was enhanced by MAPK phosphorylation, which also enhanced recruitment of p300.261
Wu et al. identified 6 functional in vivo AIB1 phosphorylation sites, all of which were required for maximal coactivation of the AR.262 Interestingly, not all sites were required for coactivation of NFκB indicating differential phosphorylation for activation of distinct signaling pathways.262
Growth factor induced phosphorylation of TIF2 occurs through ERKs at Ser-736, which is required for full growth factor induction of TIF2 transcriptional activation and coactivator function.259 Additionally, Gregory et al. also demonstrated that EGF-mediated increases in androgen-dependent transactivation were linked to phosphorylation of TIF2 and that EGF signaling increased the coimmunoprecipitation of TIF2 and AR.263 Therefore post-translational modification of AR coregulators may play an important role in assembly of the transcriptional complex and growth factor-induced enhancement of AR transactivation.
The proline-rich tyrosine kinase 2(PYK2)/focal adhesion kinase is activated in response to a number of stimuli including integrin stimulation, growth factor treatment, PI3K activation and increases in intracellular calcium. PYK2 has been shown to bind to and phosphorylate the AR coregulator ARA55. Interaction of ARA55 with PYK2 decreases ARA55 coactivator activity with AR. Inhibition of transactivation could be due to either reduced transactivation activity by phosphorylation or by the sequestering ARA55 from the AR through its interaction with PYK2.264
Acetylation of AR and coregulator recruitment
Post-translational modification of the AR by acetylation can modulate cofactor recruitment. Previous studies indicate that AR is acetylated in vitro by the histone acetyltransferase p300, P/CAF and Tip6070 at an evolutionarily conserved motif 630KLKK633.265 Acetylation of the AR in vivo is induced by ligand (DHT) and by histone deacetylase inhibitors in living cells.70, 118 AR acetylation mutants were selectively defective in DHT-induced transactivation of androgen-responsive reporter genes and coactivation by SRC1, Ubc9, Tip60 and p300, but showed 10-fold increased binding of the NCoR1 corepressor compared with wildtype AR in the presence of ligand. Furthermore, AR acetylation-defective mutants also showed decreased regulation by Akt, PKA and JNK (c-Jun N-terminal kinase) whereas MAPK signaling and sumoylation of the AR were unaffected.265 Therefore post-translational modification of the AR by acetylation can not only regulate assembly of the transcription complex through coactivator/corepressor complex binding, but also modulate receptor transactivation by other signaling pathways. Overexpression of the acetylase p300 has been reported in prostate cancer tissues with higher expression correlating with a higher Gleason score, larger tumor volume and extraprostatic extension of prostate cancer.153 Collectively, these findings suggest that increased acetylation of AR in the prostate may play a pleitropic but nevertheless a significant role in the progression of prostate cancer.
AR sumoylation and coregulator recruitment
The AR can be modified by the small ubiquitin-like modifier (SUMO) at lysines 385 and 511 in an androgen-enhanced fashion.266 The AR interacting protein Ubc-9 catalyzes attachment of SUMO-1 to AR. Such modification likely plays a role in protein targeting and/or stability, and recently has been shown to play a role modifying activity of transcription factors such as c-jun and p53.267, 268 SUMO-1 modification of the AR negatively regulates AR transactivation capacity in some promoter-based assays, and mutation of the SUMO sites can enhance AR transcriptional activity without influencing its transrepressing activity.266, 269
Sumoylation may play a role in cooperativity of the receptor on multiple hormone response elements.266 Transrepression via sumoylation could occur through the selective recruitment of coregulators that depend on AR sumoylation for interaction. The AR corepressor Daxx, originally identified as a cytoplasmic signaling molecule linking Fas receptor to JNK in Fas-mediated apoptosis, binds AR in a sumoylation-dependent manner and inhibits the DNA-binding activity of AR through interaction with the AR DBD.111 Conversely, hZimp10, a novel PIAS (protein inhibitor of activated STAT)-like protein, has been shown to augment the transcriptional activity of AR in an AR sumoylation-dependent fashion.103 Just as sumoylation can mediate the interaction of AR with coregulators, modification by SUMO-1 may also mediate the interaction of coregulators with transcription factors. The AR coregulator TIF2 can be modified by sumoylation at three lysine residues, two of which are localized in the nuclear receptor interaction domain. Mutation of these lysines to arginines impairs the ability of TIF2 to colocalize with AR in nuclei and results in attenuated ability to enhance AR-dependent transcription. The conservation of sumoylation sites in AIB1 and SRC1 may implicate modification by sumoylation as a general mechanism in the regulation of SRC protein activity.270
The AR signaling axis has a critical role in prostate biology and deregulation of the axis can take place at a number of levels. Although it is difficult to implicate altered cofactor expression, structure or function as being consistently associated with prostate cancer, it has become clear that we need to determine the molecular mechanism through which AR coregulators are able to differentially respond to a changing microenvironment to regulate specific gene targets involved in such processes as cell survival, apoptosis, invasion, metastasis and the emergence of the ADI phenotype. The daunting task ahead will be to identify, catalog and validate the “transcriptional space” for each AR/AR coregulator pair and correlate these data sets with the transcriptional profiles of clinical and experimental cancers. Given the number of possible cofactor-AR permutations we will likely need to develop new methodologies to explore gene-specific interactions at the functional level for each AR regulated gene, and determine how this is influenced by the cancer microenvironment.
GB is a recipient of an NHMRC CJ Martin Biomedical Fellowship, and EFN is supported by the Florey Adelaide Male Aging Study.