Alterations in AR-FOXA1 signaling in prostate cancer progression and therapeutic resistance

Abstract

The androgen receptor (AR) is instrumental in the onset and progression of prostate cancer (PCa), establishing androgen deprivation therapy (ADT) as the first-line treatment for metastatic disease. However, the effectiveness of ADT is commonly short-lived. Many patients eventually relapse and develop castration-resistant prostate cancer (CRPC), commonly marked by reactivated AR signaling. Although next-generation AR signaling inhibitors (ARSi) provide temporary control, resistance inevitably emerges. While a small subset of CRPC cases may evolve through AR-independent pathways, most regain partial AR function through multiple mechanisms. A key regulator of AR activity is the pioneer transcription factor FOXA1, which governs AR binding to chromatin. The AR-FOXA1 axis is essential for prostate luminal epithelial cell lineage determination and drives the development of prostate adenocarcinoma. Emerging evidence shows profound alterations in this axis in CRPC and in tumors resistant to ARSi therapies. In this review, we highlight the genetic, epigenetic, transcriptional, and posttranscriptional changes within the AR-FOXA1 axis in PCa following ADT and ARSi treatments.

1. Introduction

The androgen receptor (AR) is a key driver of prostate cancer (PCa) development and progression.¹ For more than eighty years, the cornerstone of systemic therapy has been androgen deprivation therapy (ADT) - achieved through surgical or medical castration - to inhibit AR transcriptional activity.² Although ADT can effectively suppress tumor growth initially, most patients eventually relapse within several years and develop a more aggressive, treatment-resistant disease state termed castration-resistant prostate cancer (CRPC).³ In an effort to improve outcomes, the past decade has seen the introduction and the FDA approval of next-generation AR signaling inhibitors (ARSi), including the CYP17 inhibitor abiraterone⁴ and the AR antagonists enzalutamide,⁵ apalutamide,⁶ and darolutamide.⁷ These agents delay disease progression for one to two years, yet resistance inevitably emerges. Recent efforts have been focused on optimizing treatment by either combining ARSi with other agents such as PARP inhibitors, immune checkpoint inhibitors, or taxanes,8, 9, 10, 11, 12, 13, 14 or treating them earlier in the disease course, such as the neoadjuvant setting.15, 16, 17, 18 Other treatment options targeting AR pathway in ongoing clinical trials include PROTAC AR degraders, such as ARV-110¹⁹ and ARV-766 (NCT05067140), CYP11 inhibitor (ODM-208) ,²⁰ and Bipolar Androgen Therapy (BAT) .21, 22, 23, 24 While a small subset of CRPC tumors evolves into highly aggressive, AR-indifferent phenotypes - such as neuroendocrine PCa (NEPC)25, 26, 27 or double-negative PCa (DNPC)^28,29 - characterized by either the expression or absence of neuroendocrine markers, most CRPC cases eventually exhibit reactivated AR signaling via multiple molecular alterations.

The AR gene, located on the X chromosome, encodes a nuclear steroid hormone receptor with three structural domains: the N-terminal domain (NTD), the DNA-binding domain (DBD), and the C-terminal ligand-binding domain (LBD).³ The DBD and LBD are connected by a hinge region. AR transcript includes eight exons: exon 1 encodes the NTD; exons 2–3 encode the DBD; and exons 4–8 encode the hinge region and LBD.^30,31 Most AR antagonists bind to LBD and function by competing with androgens. Upon androgen stimulation, AR translocates from the cytoplasm to the nucleus and binds to specific chromatin regions known as androgen-responsive elements (AREs).³² This chromatin interaction is a complex process, tightly regulated by additional transcription factors and epigenetic modulators.³³ Among the critical regulators of AR is FOXA1, a member of the FOXA family of pioneer transcription factors.³⁴ These factors facilitate tissue-specific gene expression by loosening tightly packed chromatin.³⁵ In AR-driven PCa, FOXA1 serves as a pivotal pioneer factor for AR, essential for sustaining the luminal adenocarcinoma subtype.^34,36,37 The FOXA1 gene resides on chromosome 14 and contains three exons. Structurally, the FOXA1 protein features a highly conserved Forehead DNA-binding domain (FKHD) and intrinsically disordered N- and C-terminal regions.^38,39 Because the Forkhead domain closely resembles the linker histone, FOXA1 can displace repressive linker histone and thereby loosen the chromatin architecture, facilitating interactions with additional transcription factors.^40,41

Recent high-throughput sequencing studies of tumor chromatin, DNA, and RNA have revealed numerous alterations in this AR-FOXA1 axis - at genetic, epigenetic, transcriptional, and post-transcriptional levels - emerging in CRPC tumors in response to ADT or ARSi.^25,42, 43, 44, 45, 46 The following sections will discuss key findings about these alterations and how they contribute to resistance, thereby informing future therapeutic strategies aimed at improving patient outcomes.

2. Alterations in AR

2.1. AR mutations

It has been well studied that AR coding mutations, many of which cluster within LBD, play pivotal roles in the development of therapeutic resistance in PCa and lead to altered sensitivity or activation profiles under low-androgen or antagonist-treated conditions. Mutations in the LBD are rare in primary PCa but frequently observed in CRPC, with L702H, H875Y, and T878A/S among the most common mutations, and W742C and F877L occurring less frequently.^43,47

There are two major resistance mechanisms - (i) Antagonist-to-agonist conversion and (ii) Altered ligand specificity. Some mutations can convert antagonist into agonist. For example, the W742C mutation disrupts a key hydrophobic interaction between W742 residue and AR antagonists, conferring resistance to bicalutamide and enzalutamide48, 49, 50 but not to hydroxyflutamide48, 49, 50 or darolutamide.⁵¹ Notably, the W742C mutant converts bicalutamide into an agonist, while hydroxyflutamide retains its antagonist function. The H875Y and T878A mutations can convert hydroxyflutamide into an agonist,⁴⁸ leading to therapeutic resistance. The F877L mutation is highly associated with resistance to enzalutamide and apalutamide and often co-occurs with T878A. This mutant or the combination can be activated by these two structurally related ligands.^52,53

Other mutations can alter the ligand binding affinity of AR, enabling activation by non-canonical steroid hormones beyond testosterone and dihydrotestosterone (DHT). The L702H mutation induces structural changes in the LBD, allowing AR to be activated by glucocorticoids at physiological levels.^54,55 This has significant therapeutic implications, as glucocorticoids are frequently used in CRPC treatment to mitigate the side effect of CYP17 inhibitors and manage cancer-related inflammation. Interestingly, although hydroxyflutamide is no longer used in treating PCa, the T878A and H875Y mutations remain strongly associated with CYP17 inhibitor resistance in patients.^56,57 These mutations, such as T878A, enable AR activation by progesterones, which serve as precursors for adrenal androgens.^58,59 After CYP17 inhibitor treatment, progesterone can be cumulated and thus leads to sustained AR signaling despite decreased adrenal androgen levels, ultimately promoting resistance. Overall, these common AR LBD mutations are presented in ∼15–20 % of CRPC cases.⁶⁰

While other LBD mutations are generally rare, some function through distinct mechanisms. For example, a Q784* nonsense mutation has been identified in CRPC patients resistant to ADT and androgen synthesis inhibition.⁶¹ This mutation results in a C-terminal truncation of the LBD, resembling the structure and function of AR splice variants. Although the truncated AR protein is transcriptionally inactive on its own, it localizes constitutively to the nucleus, dimerizes with full-length AR, and enhances its transcriptional activity even in low-androgen conditions. Together, these studies underscore the complex interplay between AR LBD mutations and therapeutic resistance, demonstrating how certain mutations contribute to the persistence of AR signaling in CRPC progression.

2.2. AR gene structural alterations

Increased AR expression is a critical driver of the therapy resistance and PCa progression. Previous studies have shown that overexpression of AR mRNA alone is sufficient to induce resistance to ADTs.62, 63, 64 Moreover, elevated expression of AR, in collaboration with other factors, could also alter the AR cistrome and drive the reprogramming of AR signaling toward cell cycle regulation and DNA damage repair pathways in CRPC cells.^64,65 Reduced androgen levels is another factor that contributes to AR reprogramming in CRPC. A recent study provided molecular evidence that low androgen levels promote AR monomer-driven non-genomic AR activity to activate mTOR signaling and support tumor proliferation under ADT conditions. In contrast, high androgen levels favor AR dimerization, which exerts growth-suppressive effects.⁶⁶ Therefore, the transition from low AR expression under high androgen levels (as seen in primary PCa) to high AR expression under low androgen conditions (typical of CRPC) may induce a substantial shift in AR cistrome and transcriptome profiles. Indeed, Labaf et al.⁶⁴ used an inducible AR expression model to show that in the CRPC context, AR binding becomes reprogrammed with reduced FOXA1 enrichment and increased dependence on EZH2. This shift resulted in enhanced transcriptional output at DNA damage repair gene loci.

While this increased AR expression after ADTs could be resulted from epigenetic and transcription processes, such as activation through an intronic enhancer,⁶⁷ or post-transcriptional regulation via 5′ and 3′ UTR-mediated effects on mRNA stability,68, 69, 70 the major mechanism driving AR overexpression in CRPC is AR gene amplification and structure alterations at AR enhancers, which are rarely detected in primary PCa.^43,45,71, 72, 73, 74, 75, 76 AR gene amplification is a hallmark genetic event in CRPC cells, occurring in up to 60 % of tumors post-ARSi treatment.^43,45,56

Recently, structural alterations affecting 5′ and 3′ AR enhancers have also been identified.74, 75, 76 In 2018, Takeda et al.⁷⁴ reported the identification of an enhancer region ∼650 kb upstream of the AR gene, characterized by a cluster of highly active chromatin marks resembling a super-enhancer. This site is selectively amplified at a much higher rate than the AR gene body, occurring in over 80 % of CRPC cases. Chromatin looping analysis confirmed that this ∼9 kb region physically interacts with the AR promoter, facilitating AR gene transcription. This AR enhancer is enriched for the binding of HOXB13 and FOXA1. Notably, introducing extra copies of this enhancer into PCa cell lines confers resistance to castration and reduces sensitivity to enzalutamide treatment. This AR enhancer is believed to have been active during prostate development, subsequently silenced and epigenetically bookmarked in adult tissues, and eventually reactivated in CRPC cells.

At the similar time, Viswanathan et al.⁷⁵ examined mCRPC biopsies and cell-free DNA and uncovered multiple structural alterations in CRPC, including complex rearrangements at the AR locus. Notably, they identified highly recurrent tandem duplications involving the upstream AR enhancer in up to 87 % of CRPC cases, providing further evidence of strong selective pressure on this enhancer region. This genome-wide tandem duplicator phenotype appears to be associated with CDK12 inactivation, suggesting a potential mechanism driving these events. Importantly, their findings suggest that even a modest increase in enhancer copy number can drive AR expression to levels comparable to high-level AR gene amplification. Similarly, another study by Quigley et al. also reported the structural variations of AR enhancer, primarily tandem duplications.⁷³ They found evidence of selective amplification of the AR enhancer over the AR gene body in 13 % of cases, further supporting the role of enhancer-driven AR overexpression.

A more recent study published last year by Zhu et al. analyzed paired CRPC samples collected before ARSi treatment and after progression on ARSi, revealing genetic and transcriptional changes associated with ARSi resistance.⁷⁶ Notably, they identified amplifications involving another enhancer approximately 60 kb downstream of the AR gene body (C4 enhancer). This site exhibits significant FOXA1 binding and is associated with chromatin accessibility and enhancer activation marks. Additionally, this study also confirmed the gain of additional copies of the AR gene body and its upstream enhancer following ARSi treatment, accompanied by increased AR mRNA expression.

Overall, these recent discoveries on AR enhancers and their structural variations (summarized in Fig. 1) clearly indicate the strong evolutionary pressure on AR signaling, revealing key genetic mechanisms that drive AR overexpression to overcome ADT and ARSi treatments. However, it remains unclear whether resistance primarily arises from AR overexpression of itself or from the increased levels of LBD-truncated AR splice variants, which are inherently resistant to LBD-targeting ARSi treatments.

Fig. 1.

AR gene amplification and enhancer tandem duplication in CRPC. A schematic illustrating a copy number gain of AR gene (two copies), each exhibiting tandem duplications of distal 5′ or 3′ enhancers, which can cause AR overexpression. AR, androgen receptor; AR-FL, AR-full length; AR-V7, AR splice variant 7; TD, tandem duplication; TF, transcription factor.

2.3. AR mRNA alternative splicing

Another key mechanism underlying CRPC progression and the resistance to ADT and ARSi is the elevated expression of LBD-truncated AR splice variants (AR-Vs). These variants can be classified into four distinct groups (summarized in Fig. 2): (a) Class I variants are constitutively active LBD-truncated isoforms that function independently of ligand binding and are active when expressed alone (e.g., AR-V7,⁷⁷ ARv567es)⁷⁸; (b) Class II variants are LBD-truncated isoforms with conditional activity that rely on the presence of full-length AR (AR-FL) or AR-V7 for transcriptional output and have limited activity when expressed alone (e.g., AR-V1,⁷⁷ AR-V3,⁷⁹ AR-V4,⁷⁷ AR-V9)⁸⁰; (c) Class III variants are LBD-truncated isoforms with minimal or uncharacterized activity, and their functional significance remains unclear (e.g., AR-V2,⁷⁷ AR-V5,⁷⁷ AR-V6,⁷⁷ AR-V8,⁸¹ AR-V10,⁸¹ AR-V11)⁸¹; (d) Class IV variants are non-truncated AR isoforms that retain a full-length or near full-length AR structure but are functionally altered due to aberrant splicing events (e.g., AR-23,⁸² AR-45)⁸³.

Fig. 2.

Classification of AR splice variants. A diagram showing the structure of the AR gene, accompanied by a table summarizing AR splice variants, categorized based on their structural features and functional properties. AR, androgen receptor; AR-FL, AR-full length; NLS, nuclear localization signal.

Amongst these AR-V isoforms, AR-V7 is the most prevalent,84, 85, 86, 87 whereas other variants, such as ARv567es, occur less frequently.^78,88 Prior studies have shown that functionally active AR-Vs, particularly AR-V7, play crucial roles in driving resistance to AR-targeted therapies.^78,88, 89, 90, 91, 92, 93, 94, 95 The class I variants typically retain the NTD and DBD but lack the LBD, leading to a constitutively active receptor that functions independently of androgens and drives gene transcription in CRPC cells.⁹⁶ AR-V7 lacks exons 4–8, which encode the hinge region and LBD, but contains a unique C-terminal extension derived from a cryptic exon 3.⁸¹ Interestingly, unlike Class II and III AR-Vs that lose their nuclear localization signal (NLS) due to truncation of the hinge region, AR-V7 protein is constitutively located in nucleus, possibly attributed to a new NLS within its unique C-terminal sequence or an alternative mechanism facilitating nuclear import.^97,98 AR-V7 is largely undetectable in primary PCa cases (<1 %), but shows a dramatical increase in expression after ADT, with its expression detected in ∼75 % of CRPC cases.⁹⁵ In most AR-V7 positive CRPC cases, its expression level is highly correlated with the expression of the AR-FL. ARv567es is another constitutively active AR variant that retains the full sequences of exons 1–4 and exon 8, while skipping exons 5, 6, and 7.⁷⁸ A recent study has shown that this variant can cause a reprogramming of AR, including loss of canonical AR-regulated genes and increased expression of AR-repressed genes.⁹⁹ Although ARv567es can be detected in CRPC tumor tissues and circulating tumor cells, its expression appears to be significantly less frequent than AR-V7.^100,101

Following ADT or ARSi treatment, AR-V7 expression rapidly and significantly increases.^87,95 Three mechanisms have been involved in driving this increase. First, this upregulation is largely attributed to the overall amplification of AR gene.⁷¹ Second, the loss of liganded AR-FL to repress its own transcription is another key regulatory mechanism. Specifically, AR-FL suppresses an intron 2 enhancer that regulate the expression of both AR-FL and AR-V7.⁶⁷ Under normal conditions, this repression mechanism helps eliminating AR isoform expression in prostate and primary PCa. However, when ADT or ARSi disrupts this regulatory feedback, the suppression is impaired, leading to a concurrent increase in both AR-FL and AR-V7 expression.^67,87 Third, alterations in mRNA splicing machinery also contribute to the increased production of AR-V7.^86,102

The functional consequences of AR-V7 expression have been the subject of extensive research. Early mechanistic studies suggested that AR-V7 primarily functions by heterodimerizing with AR-FL, enhancing its transcriptional activity under low-androgen conditions.^81,93 This model supports the idea that AR-V7 mainly acts as a co-regulator of AR-FL rather than as an independent transcription factor. However, more recent studies have provided compelling evidence that AR-V7 can function independently to regulate its own transcriptional programs.91, 92, 93^,103 The question of whether AR-V7 and other AR-Vs can establish distinct transcriptional networks by binding to chromatin sites not occupied by AR-FL remains an ongoing debate.^81,90, 91, 92, 93, 94^,104 Some studies suggest that AR-Vs may drive unique gene expression profiles that are independent of AR-FL, which could contribute to therapeutic resistance and disease progression. Nonetheless, emerging evidence indicate that AR-V7 interacts with unique cofactors that may further distinguish its function from AR-FL. For example, one study identified ZFX as a unique transcription factor that specifically associates with AR-V7, suggesting a potential role in modulating AR-V7-driven gene expression.⁹¹ Another study demonstrated that the absence of LBD in AR-V7 allows it to preferentially interact with corepressors, such as NCoRs, which could alter its transcriptional activity in ways distinct from AR-FL.⁹⁴ A very recent study also demonstrated that AR-V7 chromatin activity is supported by nuclear factor I-mediated chromatin access.¹⁰⁵ These findings suggest that AR-V7 may not only enhance AR-FL activity but also establish independent transcriptional programs through interactions with distinct chromatin-associated factors.

While AR-Vs are well recognized for driving resistance to ADT and ARSi, it remains unclear whether these variants are inherently pathogenic and contribute to more aggressive tumor progression through transcriptional programs distinct from those regulated by AR-FL. Notably, prostate-specific overexpression of AR-V7 or ARv567es in transgenic mouse models has been shown to induce prostatic intraepithelial neoplasia (PIN) or invasive carcinoma by activating oncogenic transcriptional programs.^88,106 This contrasts sharply with earlier findings in which AR-FL overexpression alone failed to induce prostate neoplasia,¹⁰⁷ suggesting that AR-Vs possess distinct biological functions beyond merely sustaining AR signaling in the absence of androgens. Clinical evidence further supports this notion, as increased AR-V expression in CRPC bone metastases, compared to castration-sensitive prostate cancer samples, correlates with poorer prognosis.¹⁰⁸ A recently published study by Han et al. provides additional mechanistic insight, demonstrating that AR-V7, but not AR-FL, actively promotes tumor cell invasion and bone metastasis in vivo.¹⁰⁹ This study identified a unique AR-V7-dependent transcriptional program enriched for genes involved in epithelial-mesenchymal transition (EMT) and metastasis, functioning entirely independently of AR-FL (Fig. 3). Among these targets is SOX9, a key regulator of stemness and metastatic progression in PCa.^110,111 These findings suggest that AR-V7 not only sustain AR signaling in CRPC but also drive aggressive tumor phenotypes through distinct regulatory networks.

Fig. 3.

Functional difference between AR-FL and AR-V7 in CRPC. An illustration showing differential pathways regulated by AR-FL and AR-V7: AR-FL predominantly targets proliferation pathways, while lipid metabolism is co-regulated by both AR-FL and AR-V7. In contrast, AR-V7 uniquely drives epithelial-mesenchymal transition -related gene expression. AR, androgen receptor; AR-FL, AR-full length; AR-V7, AR splice variant 7; CRPC, castration-resistant prostate cancer.

Overall, given their functional independence from AR-FL and their role in advanced disease, AR-Vs represent a compelling therapeutic target. Further research into the unique regulatory mechanisms and transcriptional networks of AR-Vs will be essential for developing targeted therapies aimed at AR-Vs specifically, which may provide a more effective treatment strategy for patients with AR-V-driven CRPC.

2.4. AR post-translational modifications

AR protein undergoes various post-translational modifications, including phosphorylation, methylation, acetylation, ubiquitination, and others. In response to ADT or ARSi, some of these modifications may play crucial roles in restoring AR activity and promote cellular adaption to the therapies.

2.4.1. Phosphorylation

Phosphorylation is the most well-studied post-translational modification of AR. It can occur at over 20 serine, threonine, and tyrosine residues.112, 113, 114 Many of these sites are constitutively phosphorylated, such as S94, while others are induced by androgens or other stimuli, including S81, S213, S308, S256, Y534, and S650.^113,114 These phosphorylation events are essential for modulating AR stability, subcellular localization, and transcriptional activity. Early evidence suggested that S650 phosphorylation is important for AR transcriptional activity.¹¹² Further studies indicate that S650 is regulated by stress kinase signaling, possibly phosphorylated by p38 and JNK1,¹¹⁵ and dephosphorylated by PP1.¹¹⁶ This phosphorylation event is crucial for AR nuclear export, particularly in response to stress conditions.¹¹⁵ Several studies have demonstrated that AKT directly binds to AR and phosphorylates the S213 residue, enhancing androgen-mediated nuclear localization of AR,^117,118 although other evidence suggests that this phosphorylation event may not occur in vivo under physiological conditions.¹¹³ Another phosphorylation site, S256, is potentially phosphorylated by casein kinase II, though further validation is needed.^113,116,119 AR can also undergo phosphorylation at tyrosine residues by tyrosine kinases. Increased tyrosine phosphorylation has been observed in CRPC xenograft models.¹²⁰ Among these sites, Y534 phosphorylation has been detected in CRPC patient samples and is believed to be mediated by Src kinase.¹²⁰ This phosphorylation event facilitates EGF-induced AR transcriptional activity, promotes androgen-independent AR activation, and enhances nuclear translocation. Consequently, Y534 phosphorylation may contribute to sustained AR activity in CRPC, even in the absence of androgens. This aligns with findings that CRPC exhibits an overall increase in tyrosine phosphorylation networks and that tyrosine kinase activity can restore AR function under CRPC conditions.^121,122

One of the most well-studied AR phosphorylation sites in PCa is S81, which is located within the polyglutamine tract region of the NTD.^113,123, 124, 125, 126 S81 can be phosphorylated by CDK1 or CDK9, though under different conditions. CDK1 phosphorylates S81 in the absence of androgens and during mitotic progression,¹²³ while CDK9 phosphorylates S81 when liganded AR is bound to chromatin and actively engaged in transcriptional regulation.¹²⁷ Studies suggest that the full phosphorylation status of S81 may be sustained by both kinases, ensuring continued AR activity.^126,128 The function of S81 phosphorylation has been linked to enhancing AR chromatin binding, coactivator recruitment, chromatin looping, stability, and transcriptional activation.^123,124,126, 127, 128 Importantly, S81 phosphorylation has been detected in both primary PCa and CRPC patient samples, underscoring its potential role in disease progression.¹²⁵ In CRPC xenograft models, phosphorylated S81 is strongly associated with the reactivation of AR signaling, suggesting that S81 phosphorylation may play a pivotal role in CRPC progression and therapeutic resistance.¹²⁵

Interestingly, AR phosphorylation events, such as S81 and S213, occur not only in full-length AR but also in AR splice variants.^109,129,130 A recent study by Han et al.¹⁰⁹ reported that S81 phosphorylation enhances the pro-metastatic function of AR-V7 by selectively altering its transcriptional program, rather than directly influencing its chromatin binding. Unlike AR-FL, AR-V7 appears to be specifically phosphorylated by CDK9 but not CDK1. This suggests a distinct regulatory mechanism between AR-FL and AR-V7, potentially driven by structural differences, such as the absence of N- to C-terminal interaction in AR-V7. Targeting this phosphorylation with CDK9 inhibitors effectively prevented S81 phosphorylation in AR-V7 and successfully blocked AR-V7-induced SOX9 expression and its associated metastasis-promoting function using in vivo models. This highlights a potential therapeutic strategy to specifically inhibit AR-V7 function in CRPC. Moreover, this study also suggests that the high levels of S81 phosphorylation detected in patient samples¹²⁵ may, in part, be a consequence of AR-V7 expression.

2.4.2. Methylation and acetylation

AR can be methylated at K630 and K632 within the KLKK motif of its NLS in the hinge region by SETD7 (also known as SET9).^131,132 These methylation events are thought to enhance AR transcriptional activity by promoting nuclear translocation, coactivator recruitment, N-to-C interaction, and protein stability. Interestingly, a recent study by Wang et al. suggests that SETD7 may function as a tumor suppressor rather than a tumor promoter in CRPC by methylating the K270 residue of FOXA1, which potentially destabilizes FOXA1 chromatin binding.¹³³ Given that loss of SETD7 expression is common in CRPC patient datasets,¹³³ this may indicate a decrease in AR methylation in CRPC, which could in turn reduce AR activity. However, the overall effect is likely balanced by increased FOXA1 chromatin binding, which enhances AR chromatin binding and transcriptional activation. Notably, AR-V7 lacks these methylation sites, yet it still relies on FOXA1 for its canonical transcriptional activity.134, 135, 136, 137 Therefore, the loss of SETD7 expression may paradoxically enhance AR-V7 activity by stabilizing FOXA1 binding, further driving androgen-independent AR signaling in CRPC. AR also interacts with various other methyltransferases and demethylases, such as EZH2 and LSD1.138, 139, 140, 141 However, whether these enzymes can directly modulate AR methylation remains undetermined.

Interestingly, AR can also be acetylated at K632 and K633 within the same KLKK motif by p300/CBP.¹⁴² Mutations at these lysine residues prevent their acetylation, leading to impaired DHT-induced transcriptional activation of AR. In addition to p300/CBP, Tip60 has also been identified as an AR acetyltransferase, directly modifying AR through the KLKK motif.¹⁴³ This acetylation event may be dynamically regulated by HDAC1, which represses Tip60-mediated AR transcription activity by removing the acetylation marks. These findings add another layer of complexity to AR regulation, suggesting that coactivators such as p300/CBP and Tip60 can enhance AR activity through direct acetylation. However, these observations may contradict studies on AR methylation,^131,132 as the lysine methylation may antagonize the acetylation. Given the potential interplay between these modifications, future studies are needed to delineate the functional crosstalk between AR acetylation and methylation and to determine how these modifications cooperatively or antagonistically influence AR-driven transcription and prostate cancer progression. Beyond the hot spot site of KLKK motif, a recent mass spectrometry study identified additional post-translational modifications, including methylation and acetylation at K313 (previously referred to as K311) and methylation at K318. Functional analysis demonstrated that a double mutant at these sites reduces AR activity, although the specific enzymes responsible for these modifications remain unidentified.¹⁴⁴

2.4.3. Ubiquitination

Ubiquitination is one of the most prevalent post-translational modifications in eukaryotic cells and serves as a critical mediator in various cellular signaling pathways, including proteasomal degradation, endocytosis and membrane trafficking, kinase activation, DNA damage repair, and transcriptional regulation.¹⁴⁵ In the context of AR regulation in PCa cells, multiple E3 ubiquitin ligases have been reported to modulate AR stability and activity, including MDM2, CHIP (also called STUB1), and SPOP, which facilitate AR degradation through the ubiquitin-proteasome pathway.146, 147, 148, 149 Several lysine residues have been identified as key ubiquitination sites on AR, including K311, K845/K847, and K911, each with distinct regulatory roles.^144,150,151 K311 is a potential target for MDM2-mediated ubiquitination, though MDM2 is known to ubiquitinate AR at multiple sites.¹⁵¹ Mutation of K311 increases AR protein stability but paradoxically decreases its transcriptional activity, potentially due to reduced interaction with the transcriptional coactivator p300. Interestingly, this ubiquitination event is dependent on the presence of the LBD, suggesting a role for C-terminal interactions with E2/E3 ligases in AR ubiquitination.¹⁵¹ The K911 site was identified from a mCRPC cell line model and the mutation of this site has been shown to increase AR stability, enhance chromatin binding, and promote transcriptional activity.¹⁴⁴

In addition to proteolytic K48-linked ubiquitination, non-degradative ubiquitination has also been observed on AR. RNF6 facilitates K6- and K27-linked polyubiquitination at K845 and K847, which does not lead to AR degradation but instead enhances AR recruitment of coactivators and modulates its transcriptional activity.¹⁵⁰ Notably, RNF6 is upregulated in CRPC patient samples, suggesting that this modification may play a role in restoring AR signaling in CRPC. Under castration conditions, TRAF4 also mediates K27-linked ubiquitination at the AR C-terminus.¹⁵² This modification does not target AR for degradation but instead alters AR activity, suggesting a potential role in PCa progression under low-androgen conditions.

Notably, some of these ubiquitination mechanisms may also apply to AR-V7, particularly MDM2- and CHIP-mediated ubiquitination and degradation.^153,154 However, the full extent of ubiquitination-mediated regulation of AR-V7 remains an area for further study. Overall, these studies indicate that ubiquitination serves as a key regulatory mechanism in modulating AR function, with implications for CRPC progression and potential therapeutic strategies to target AR activity.

2.5. AR reprogramming by other genetic changes

The AR signaling in PCa can be also altered and reprogrammed through changes in other transcription factors, pioneer factors, and epigenetic regulators. Notably, mutations and alterations in specific driver genes can profoundly influence AR activity and contribute to PCa progression. In this section, we highlight several key genetic alterations that can reprogram AR signaling (summarized in Fig. 4).

Fig. 4.

Driver genetic alterations inducing AR reprogramming. The schematic highlights key genetic mutations and alterations that contribute to AR pathway reprogramming in PCa. The effects of SPOP mutations, TP53 alterations, and FOXA1 mutations on AR signaling are not discussed in this section. AR, androgen receptor.

2.5.1. TMPRSS2-ERG fusion gene

The TMPRSS2-ERG gene fusion, present in approximately 40–50 % of PCa,^155,156 results from the fusion of the androgen-responsive TMPRSS2 promoter with the ERG oncogene. This rearrangement drives androgen-dependent overexpression of the ERG transcription factor, which may disrupt normal AR signaling in PCa cells. Mechanistically, it has been shown that ERG can repress AR expression and directly bind to AR target loci, thereby inhibiting AR activity at specific downstream genes.¹⁵⁷ However, in several other studies, ERG is believed to drive the maintenance and reprogramming of AR signaling. Using the same VCaP PCa model, Cai et al.¹¹⁰ has shown that ERG can activate a subset of cryptic AR regulatory elements, reprogramming AR activity and promoting the expression of oncogenic targets such as SOX9, ultimately enhancing EMT and metastasis. In addition to modulating AR, ERG overexpression has also been shown to transcriptionally activate NO-cGMP signaling axis, which is associated with endothelial differentiation.¹⁵⁸ In ERG-overexpressing transgenic mouse models, ERG appears to maintain AR activity and reinforce the luminal epithelial lineage. When ERG overexpression is combined with Pten loss, AR transcriptional activity is paradoxically restored, alongside the upregulation of genes involved in cell death, migration, inflammation, and angiogenesis, indicating that ERG may prime prostate epithelial cells to upstream oncogenic alterations such as PTEN loss.¹⁵⁹ Further studies in mouse models show that in a Pten/Trp53-deficient background, ERG similarly reactivates AR signaling, inducing luminal epithelial gene expression and restricting cellular plasticity.¹⁶⁰ This suggests that ERG may constrain lineage switching and preserve tumor dependency on AR signaling, thereby maintaining sensitivity to ARSIs. Recent lineage-tracing studies in mouse prostates reveal that ERG expression suppresses Trp63, a master regulator of basal cell identity, thereby inhibiting the ability of luminal cells to transdifferentiate into basal-like lineages.¹⁶¹ Interestingly, data from a recent preprint suggest that ERG activation in basal cells may drive transition into a highly proliferative intermediate state exhibiting mixed luminal and basal markers.¹⁶² Taken together, these findings highlight the multifaceted and context-dependent role of TMPRSS2-ERG in modulating AR signaling and lineage determination throughout prostate cancer progression.

2.5.2. MYC and MYCN amplifications

Amplification of MYC (∼20–35 %) and MYCN (5–20 %) is frequently observed in advanced PCa and plays a pivotal role in reprogramming AR signaling.^45,163 MYC has been shown to share chromatin binding sites with AR and can functionally repress AR transcriptional activity at canonical target genes,¹⁶⁴ although other studies suggest that MYC may also enhance the expression and activity of both AR-FL and AR-V7, particularly in the context of therapy resistance.¹⁶⁵ Recent work has confirmed that MYC overexpression disrupts the AR transcriptional program not by displacing AR from chromatin, but rather by increasing promoter-proximal pausing of RNA polymerase II at AR-regulated genes.¹⁶⁶ This transcriptional interference contributes to loss of AR-dependent differentiation signaling and promotes tumor initiation and progression. Additionally, MYC cooperates with pioneer factors such as FOXA1 to access non-canonical regulatory regions, further reprogramming the AR cistrome and reshaping downstream transcriptional output.¹⁶⁶ Interestingly, AR also plays a suppressive role on MYC expression by disrupting a super-enhancer near the MYC locus, suggesting a finely tuned reciprocal regulatory relationship between AR and MYC.¹⁶⁷ AR inhibition may disrupt this balance, leading to elevated MYC expression in CRPC. In turn, MYC upregulation has been shown to promote AR-independent progression and contribute to resistance to AR pathway inhibitors. MYC also mediates metabolic reprogramming in response to AR signaling inhibition, driving a shift from glycolysis to increased mitochondrial metabolism.^168,169

In contrast, MYCN amplification is strongly associated with NEPC and facilitates lineage plasticity. MYCN directly suppresses AR expression and activity, contributing to the emergence of AR-independent tumor phenotypes.¹⁶³ Furthermore, MYCN collaborates with epigenetic regulators such as EZH2 to repress AR target genes and activate neuroendocrine transcriptional programs, reinforcing the AR-null state. Together, MYC and MYCN amplification contribute to PCa progression and therapeutic resistance through distinct mechanisms - MYC by reprogramming AR-dependent transcription and sustaining tumor growth under AR inhibition, and MYCN by driving transdifferentiation toward AR-independent, neuroendocrine states.

2.5.3. RB1 loss

Loss of the RB1 tumor suppressor gene is associated with aggressive PCa phenotypes,⁴⁵ and when combined with TP53 loss may drive lineage plasticity and transition to NEPC.^27,170 RB1 loss leads to hyperactivation of E2F transcription factors, which can regulate AR activity. While some studies suggest that E2F1, in collaboration with DNMT1, can inhibit AR signaling,^171,172 other work has shown that Rb deletion leads to unchecked AR activity via the activation of E2F1, driving therapeutic bypass of ADTs and tumor progression.¹⁷³ Mechanistically, a more recent study reported that RB1 loss promotes chromatin co-occupancy of AR and E2F1, forming transcriptional collaboration that reprograms AR activity and induces gene networks that protect cells from apoptosis.¹⁷⁴ Interestingly, another study using PCa models with RB1 and TP53 knockout showed that while RB1 loss alone has limited impact on canonical AR signaling, concurrent loss of both RB1 and TP53 attenuates AR signaling and promotes a stem-cell-like transcriptional program.¹⁷⁵

In addition to modulating the canonical transcriptional activation activity of AR, one of the most significant roles of Rb is perhaps as a corepressor of AR. Rb mediates the anti-tumor effects of high-dose androgens, facilitating the transcriptional repression activity of AR on DNA replication and cell cycle genes in CRPC cells.^67,176 Rb-like proteins, such as RBL1 and RBL2, may also participate in this repression and potentially compensate in the context of RB1 loss.^177,178 Consequently, the deficiency of Rb and Rb-like proteins weakens AR’s ability to repress these cell cycle and E2F target genes, and alters the chromatin binding landscape of FOXA1 and AR.¹⁷⁷ This results in reprogramming of AR transcriptional repression functions toward alternate gene pathways, including those involved in cell polarity and oxidative phosphorylation.^177,179 Together, these findings not only highlight the critical role of RB1 loss in altering classic AR signaling but also underscore its involvement in mediating AR activity under high-dose androgen treatment, such as in bipolar androgen therapy in CRPC.

2.5.4. ZNF397 deletions

Deletions of ZNF397 gene have been recently identified in PCa and particularly in CRPC. ZNF397 functions as a novel coactivator of AR and functions to maintain AR-driven luminal lineage in PCa cells. Loss of ZNF397 leads to a transcriptional reprogramming from an AR-dependent state to a TET2-driven lineage plastic state, contributing to resistance against AR-targeted therapies.¹⁸⁰ Notably, the study also suggests that inhibiting TET2 may reverse this resistance in ZNF397-deficient tumors, highlighting a potential therapeutic strategy to overcome lineage plasticity-associated treatment resistance.

2.5.5. CHD1 loss

CHD1 is a chromatin remodeler gene frequently deleted or mutated in PCa.¹⁸¹ CHD1 loss occurs in 10–15 % of PCa cases and tends to be mutually exclusive with common ETS-family gene fusions such as TMPRSS2-ERG.^43,72,182 Clinically, CHD1-deficient PCa tends to show poor responses to AR-targeted therapies.^182,183 Notably, recent data also indicate that CHD1 deletions are more frequent in African American PCa and are associated with more aggressive disease progression.¹⁸⁴ CHD1 has tumor-suppressive functions in the prostate. In transgenic mouse models, co-deletion of Chd1 and Pten leads to more aggressive development of invasive carcinoma, highlighting the role of CHD1 in restraining tumor progression.¹⁸⁵ Mechanistically, CHD1 acts as a cofactor of AR with an enhancer-specific role. It regulates nucleosomal positioning flanking AR binding sites and maintains open chromatin architecture at AR target loci enriched with additional AR cofactors.¹⁸⁵ Loss of CHD1 leads to a redistribution of the AR cistrome toward HOXB13-enriched sites, disrupting the canonical, “normal-like” AR transcriptional program associated with growth suppression and instead activating pro-oncogenic transcriptional networks.¹⁸⁵

In the context of CRPC progression, CHD1 plays a critical role in preserving a luminal AR-driven transcriptional program that supports prostate epithelial differentiation.^182,183 CHD1 loss disrupts this regulation, redirecting AR activity toward non-luminal lineage gene programs, thereby contributing to resistance to AR-targeted therapies.¹⁸³ For example, CHD1-deficient tumors can exhibit a switch to glucocorticoid receptor signaling, which can partially compensate for loss of AR activity.¹⁸³ Furthermore, CHD1 deletions frequently co-occur with MAP3K7 loss, forming a distinct molecular subclass of PCa that exhibits elevated AR-V7 expression under ADT conditions.¹⁸⁶ This suggests that CHD1 loss may facilitate AR-V-driven signaling, further promoting castration resistance. In addition to its impact on AR signaling, CHD1 loss also drives significant tumor microenvironment remodeling, including reduced myeloid-derived suppressor cells and increased CD8⁺ T cell infiltration, potentially altering immune responsiveness.¹⁸⁷ Overall, CHD1 loss defines a distinct molecular subtype of PCa characterized by AR cistrome reprogramming, lineage plasticity, therapy resistance, and tumor microenvironment reshaping.

3. Alterations in FOXA1

The unique "winged helix" structure of FOXA1, which resembles that of linker histone H1, facilitates chromatin loosening by displacing nucleosomes and enabling transcription factors, such as AR, to access DNA.188, 189, 190 Additionally, the intrinsically disordered regions (IDRs) in its N- and C-termini confer liquid-liquid phase separation (LLPS) properties, which may enhance its ability to dissolve heterochromatin through condensate formation (Fig. 5A).¹⁹¹ This pioneering activity establishes FOXA1 as a critical regulator of oncogenic pathways driving PCa progression. One of the most prominent roles of FOXA1 in PCa is the reprogramming of the AR cistrome during disease progression.^192,193 By redistributing AR binding sites, FOXA1 orchestrates transcriptional changes that support tumor evolution and therapy resistance. In mCRPC, FOXA1 overexpression occurs in ∼25 % of cases due to gene amplification and structural rearrangements including translocations and tandem duplications within the FOXA1 locus.¹⁹⁴ Additionally, FOXA1 coding mutations, prevalent in ∼10–15 % of cases, cluster predominantly within the Wing2 region of the FKHD domain.^72,194,195 These mutations, comprising missense and insertion/deletion (indel) alterations, account for ∼60 % of all FOXA1 mutations. This section focuses on FOXA1 aberrations, including genetic alterations and post-translational modifications, and their effects on PCa development. We also provide insights on the potential therapeutic approaches to target FOXA1 alterations in PCa.

Fig. 5.

FOXA1 alterations and functions. (A) Schematic of FOXA1 protein structure. (B-D) Summary of FOXA1 alterations and their molecular functions. AR, androgen receptor; LLPS, liquid-liquid phase separation; PTMs, post-translational modifications.

3.1. Interplay between AR and FOXA1

FOXA1 preferentially bind to histone 3 lysine 4 di-methylation (H3K4me2)-enriched enhancers in PCa cells³⁶ and this binding pattern further reinforces the lineage-specifying role of FOXA1, where it collaborates closely with AR. In primary PCa patient samples, FOXA1 and AR protein levels exhibit a significant positive correlation, with high FOXA1 and AR expression predicting poorer patient survival.¹⁹⁶ Functionally, overexpression of FOXA1 in androgen-responsive LNCaP cells leads to a substantial increase in AR genomic binding at novel regions characterized by increased chromatin accessibility.¹⁹⁷ Conversely, androgen-induced AR chromatin binding also promotes FOXA1 recruitment, highlighting its role as an AR cofactor.¹⁹⁸ This is supported by the single-molecule tracking studies that revealed dynamic interactions between hormone receptors and FOXA1, suggesting an “initiator” role of AR for FOXA1 binding in PCa cells.¹⁹⁹ Interestingly, while the canonical activity of AR splice variants appears to depend on FOXA1,¹³⁵ the distinct cistrome of AR-V7 is dependent on ZFX or other unique cofactors.⁹¹ Supporting this, recent studies by Han et al. demonstrated that AR-V7-specific binding may occur at chromatin sites with low FOXA1 enrichment.¹⁰⁹ Nevertheless, the intricate interplay between FOXA1 and AR remains a crucial driver of PCa progression and a promising therapeutic target.

Although FOXA1 is widely considered an essential pioneer factor for AR recruitment, its depletion does not abolish global AR binding. Instead, it triggers a dramatic reprogramming of AR binding sites.^196,200 This AR reprogramming in PCa was first described by the groups of Rosenfeld and Fu, where they classified AR binding events into “lost,” “conserved,” and “gained” groups following FOXA1 knockdown.²⁰¹ By integrating multiple epigenetic profiles with global nuclear run-on sequencing (GRO-seq) analyses, they demonstrated that “gained” AR binding events drive cell proliferation programs without requiring nucleosome remodeling. Around the same time, Sahu et al. defined three distinct classes of AREs based on their FOXA1 dependence: (1) pioneered by FOXA1, (2) independent of FOXA1 (3) masked by FOXA1.¹⁹⁶ Interestingly, the majority (∼70 %) of AREs either pioneered by or independent of FOXA1 are marked by H3K4me2, whereas ∼50 % of newly established AREs following FOXA1 depletion lack this modification. Further supporting this model, Jin et al. investigated the AR cistrome and transcriptome in both FOXA1-depleted LNCaP cells and FOXA1-overexpressing AR/FOXA1-negative Du145 cells.²⁰⁰ Their findings have revealed a finely tuned cooperative program between FOXA1 and AR. Notably, FOXA1 loss allows AR to bind chromatin regions containing AREs even in the absence of androgen, whereas FOXA1 redirects AR to composite FKHD-ARE motifs. However, excessive FOXA1 levels dilute AR-specific transcriptional programs by creating an overabundance of open chromatin regions. Additionally, that study demonstrated that FOXA1 directly represses AR expression, highlighting its tumor-suppressive role in restricting AR signaling.

In clinical samples, AR-targeted therapies can also profoundly reprogram the FOXA1 chromatin landscape. A neoadjuvant clinical trial (DARANA cohort), in which high-risk PCa patients received short-term (∼3 months) enzalutamide monotherapy, provided paired pre- and post-treatment biopsies for epigenomic and transcriptomic analysis.²⁰² As expected, AR binding followed the same trend as FOXA1, with newly acquired FOXA1 binding sites marked by enhancer activation (H3K27ac enrichment). Interestingly, FOXA1 binding sites present in pre-treatment samples appeared non-functional but may be reactivated during metastatic progression to regulate developmental programs, consistent with the team’s earlier report.¹⁹³ Moreover, FOXA1 binding post-therapy shifted toward cooperation with the circadian clock component ARNTL, which compensated for AR inhibition. While ARNTL’s role in long-term drug-resistant PCa remains unclear, these findings suggest that additional chromatin factors support PCa survival programs after AR blockade. Given that FOXA1 and other lineage pioneer factors, such as HOXB13, are expressed in normal epithelial cells, it is likely that specific chromatin regulators may drive AR reprogramming in PCa. To explore this, Parolia et al. conducted an epigenetics-focused CRISPR screen, identifying NSD2, a H3K36 mono- and di-methyltransferase, as selectively expressed in transformed PCa cells.²⁰³ Notably, NSD2-dependent AR binding sites were enriched for a composite motif comprising a FOXA1 element juxtaposed to an AR half-site, suggesting that NSD2 assists FOXA1 and HOXB13 in expanding the AR enhancer network. Together, these studies highlight the multifaceted role of FOXA1 in shaping AR signaling and chromatin dynamics in PCa. A deeper understanding of this complex regulatory interplay will be critical for developing novel therapeutic strategies to combat treatment resistance and disease progression in prostate cancer.

3.2. FOXA1 mutations

FOXA1 genetic mutation rate increases from ∼4 %−9 % in primary stage to ∼10–15 % in metastatic CRPC stage in patient cohorts predominantly from Western countries.^72,204 Strikingly, recent reports in Chinese primary PCa cohorts revealed that FOXA1 is the most frequently mutated gene, with mutation rates ranging from ∼30 %−40 %.²⁰⁵ This striking difference highlights the ethnic and racial background as critical risk factor in PCa. Additionally, while most FOXA1 mutations in the Chinese cohort cluster in the Wing2 region, mutations in Western cohorts occur across the entire coding region, albeit with a notable enrichment in the Wing2 region.

Mutations within the FKHD are generally missense mutations (FKHD-ms) and in-frame insertions (FKHD-ins) and deletions (FKHD-del). These mutations directly impact FOXA1 DNA binding properties. The effects of FKHD-ms on AR activity have been described in various models but no consensus has been reached. These mutants can increase AR transcriptional output through enhancing FOXA1 chromatin mobility and binding frequency.¹⁹⁴ Conversely, in PCa cell lines stably overexpressing hotspot FKHD-ms (e.g., H247Y, M253K), the mutants rendered decreased chromatin binding on the AR-regulated enhancers and impaired AR transcriptional activity.²⁰⁶ However, these mutants showed gained binding at enhancers of genes involved in EMT pathway, enhancing metastatic potential.^206,207 These hotspot mutations appear to preferentially bind enhancers containing non-canonical FKHD motifs. Unlike the canonical core FKHD-binding motifs (GTAAAC/T), mutants favor AT-rich elements in EMT gene enhancers. Additionally, FOXA1 mutants in mouse prostate organoids showed increased binding to GTAAAG/A motifs,¹⁹⁵ underscoring context-dependent variations in DNA-binding preferences.

A recent large-scale sequencing study of 5014 PCa cases from the CarisLifeSciences Precision Oncology Alliance database introduced a refined classification schema for FOXA1 alterations, delineating four main classes and seven subclasses.²⁰⁸ Comparing these to wild-type FOXA1 cases (comprising 84 % of samples) revealed distinct class-specific molecular and immunologic features. All missense or in-frame insertion/deletion mutations were denoted as class 1, with subclasses assigned to mutations located in the non-Wing2 region (AA 168–246) as class 1A, and mutations located in the Wing2 region (AA 247–269) as class 1B The rest of missense or in-frame were assigned as class 1C, except for R219, which was denoted as a separate class 2 because of its enrichment in NEPC. All class 1 mutations exhibited similar pathway patterns, including increased E2F signaling and androgen response pathway. However, class 1C indicated significantly improved clinical outcomes, whereas class 2 has the opposing effects. It should be noted that FOXA1 binds to nucleosomal DNA, with Wing1 and Wing2 regions contacting the DNA minor groove, while α-helix 3 interacts with the major groove.¹⁹⁰ Therefore, it is likely that different mutations within each subdomain may exhibit distinct molecular functions. In silico mutagenesis studies suggested that R219 (class 2) mutations located in the α-helix 3 and Wing2 mutations (class 1B) have significantly higher free-energy changes compared to WT.²⁰⁸ Further biophysical analysis may provide a better understanding of the molecular basis of these mutations (summarized in Fig. 5B).

FOXA1 truncation mutants lacking the C-terminus demonstrate molecular distinctions from FKHD mutations. These mutants exhibit dominant chromatin binding and displace WT FOXA1 and its co-repressor TLE3 from chromatin, leading to the activation of WNT signaling.¹⁹⁴ These actions confer the mutant cells more invasive capacity, similar to some FKHD-ms cell line models. It should be noted that an α-helical structure composed of nine amino acids around FOXA1-K414 has been shown to play critical roles in histone interactions and chromatin opening.²⁰⁹ A majority of the truncation mutants will lose this α-helical structure, suggesting additional chromatin-associated mechanisms underlie their oncogenic activity. Interestingly, findings from the CarisLifeSciences database suggest that truncation location correlates with opposing clinical outcomes.²⁰⁸ Specifically, mutations occurring before K270 (class 3A) are significantly associated with improved overall survival (OS), whereas mutations occurring after K270 (class 3B) are significantly correlated with worse OS, potentially due to stronger activation of E2F signaling. However, it remains unclear how compromised FKHD in these mutants impacts their transcriptional activity (summarized in Fig. 5C).

3.3. FOXA1 non-coding alterations

The rate of FOXA1 gene structural rearrangements significantly increase from approximately 8 % in primary stage to 25 % in mCRPC.¹⁹⁴ Previous reports identified a super-enhancer region upstream of the FOXA1 gene body, termed FOXMIND, which is frequently “hijacked” by other key oncogenic factors, such as ETV1 and MYC, leading to aberrant overexpression. Additionally, oncogenes such as CCNA1 and HOXA1 can also be inserted upstream of FOXA1. Focal Tandem duplications involving both FOXMIND and FOXA1 loci drive FOXA1 overexpression. This type of genetic alteration, initially categorized as Class 3, has been recently revised to Class 4.²⁰⁸ As expected, FOXA1 amplification is significantly associated with poor clinical outcomes in metastatic samples, likely through its role in enhancing AR signaling. Interestingly, despite their divergent features, mutant FOXA1 broadly show significantly elevated transcript levels comparing to the WT, but similar levels to the samples with FOXA1 amplification. However, whether FKHD mutants have distinct functional consequences compared to FOXA1 amplification remains unclear. In addition to these coding and structural alterations that result in aberrant FOXA1 protein synthesis, non-coding mutations within the FOXA1 gene loci have also been reported. For instance, a sequencing study of the plasma cell-free DNA from mCRPC patients revealed that approximately 12 % harbored indel mutations in the 3′-UTR of FOXA1.²¹⁰ Although these noncoding mutations have been proposed as passenger events, their biological functions remain unclear. Nevertheless, given their specificity to PCa, these mutations may serve as valuable biomarkers for liquid biopsy applications.

3.4. FOXA1 post-translational modifications

Similar to AR protein, the FOXA1 activity is also profoundly influenced by various post-translational modifications, including methylation, acetylation, phosphorylation and SUMOylation, which primarily modulate FOXA1 chromatin binding and protein stability. For instance, the histone acetyltransferase (HAT) p300 can acetylate lysine residues within the Wing1 and Wing2 regions of FOXA1, attenuating its DNA-binding affinity and reducing its capacity to remodel chromatin.²¹¹ Similarly, SUMOylation of FOXA1 within the N-terminal and C-terminal transactivation domains promotes protein turnover of FOXA1 and enhances AR nuclear mobility while dampening the transcriptional activity of FOXA1.²¹² Furthermore, a recent proteogenomic profiling study identified phosphorylation at S331 in high-risk PCa samples from Chinese patients.²¹³ This modification enhances FOXA1-regulated chromatin accessibility and AR cistrome reprogramming.

Recent studies have highlighted the pivotal roles of FOXA1 methylation in regulating AR activity during CRPC progression. FOXA1 K270 has been identified as a direct substrate of LSD1,¹³⁷ an enzyme known to demethylate mono- or di-methylated H3K4.^138,214 Located immediately adjacent to the Wing 2 region of the FKHD domain, K270 plays a crucial role in FOXA1 function. Genome-wide analyses and biochemical studies have demonstrated that LSD1-mediated demethylation of FOXA1 enhances its chromatin binding and activation of the AR transcriptome. The unmethylated FOXA1 was further shown to stabilize BRD4 residency on chromatin and collaborate with LSD1 in activating oncogenic super-enhancer programs, which are specifically prominent in CRPC.²¹⁵ Conversely, SETD7, a histone methyltransferase functioning as a tumor suppressor in PCa, has been identified as the methyltransferase of the methylation of FOXA1 K270.¹³³ These findings support a model where SETD7 and LSD1 maintain a balanced regulation of FOXA1-mediated enhancer accessibility during early PCa stages. However, loss of SETD7 expression in advanced stages disrupts this equilibrium, tipping FOXA1 binding toward oncogenic enhancer activation. Given the pivotal role of K270 in determining functional outcomes, future research should also focus on elucidating the regulation of K270 methylation in FOXA1-mutant cells.

Additionally, FOXA1 methylation at K295, mediated by EZH2 - the catalytic subunit of the polycomb repressive complex (PRC2) - has been shown to enhance FOXA1 protein stability. K295 methylation promotes interaction with BUB3, which recruits USP7 to remove ubiquitination from FOXA1.²¹⁶ However, the demethylase responsible for FOXA1 K295 remains unidentified. Collectively, these findings underscore the therapeutic potential of targeting FOXA1 methylation/demethylation processes in CRPC (summarized in Fig. 5D).

Moreover, the spectrum of FOXA1 post-translational modifications in PCa may be extended based on recent findings in breast cancer (BCa), where FOXA1 is a key regulator of estrogen receptor (ER) activity. O-linked β-N-acetylglucosamine modification (O-GlcNAcylation) at the C-terminus of FOXA1 has been associated with promoting BCa metastasis.²¹⁷ This modification reduces FOXA1 protein stability and alters its interactome, leading to global redistribution of FOXA1 chromatin binding and transcriptional silencing of the adhesion-related genes. Furthermore, T429 and T464 phosphorylation enhance FOXA1 interaction with histone proteins, promoting ER signaling activation.²¹⁸ Whether these or additional FOXA1 modifications occur during CRPC progression remains to be determined. Future studies should aim to elucidate how these modifications influence FOXA1 activity and its intricate interplay with AR signaling.

3.5. AR-independent activity of FOXA1

Emerging evidence highlights FOXA1 functions beyond its role in regulating AR transcriptional activity. Comparative transcriptomic analyses of PCa patients at various stages have shown that FOXA1 is strongly associated with splicing-related genes.²¹⁹ Unlike AR, ERG, and MYC - well-established regulators of oncogenic alternative splicing - FOXA1 appears to act as a master regulator by fine-tuning the expression of splicing-related genes, many of which are robust biomarkers for disease recurrence. More recently, integrative analyses, including nucleolar proteomics, large-scale immunofluorescence, RNA-seq, and ChIP-seq, revealed that FOXA1 binds to the rDNA repeats in various PCa models and patient samples, promoting ribosome RNA biogenesis and protein translation.²²⁰

As a luminal epithelial lineage determining factor, FOXA1 plays a critical role in maintaining cell identity. While FOXA1 expression is high in adenocarcinoma, its levels decline in NEPC.²²¹ FOXA1 directly represses interleukin 8, a mediator of NE differentiation upon FOXA1 loss.²²² Surprisingly, a comprehensive epigenomic profiling study using mCRPC PDX models has suggested that FOXA1 binding may shift to NE regulatory elements during lineage plasticity,²²³ indicating a potential alteration of its transcriptional programs. However, in another study aiming to explore the transcriptional dynamics underlying lineage transitions, Han et al.²²⁴ employed single-cell transcriptomic and chromatin accessibility profiling in a murine lineage-tracing model following Pten, Trp53, and Rb1 deletions and identified a shift from FOXA1 to FOXA2 as a critical pioneering transcription factor for the NEPC lineage transition from prostate luminal cells, while FOXA1 still governs the luminal lineage. The study is consistent with the reported activity that FOXA2 can cooperate with HIF1α to activate hypoxia-associated NE programs.²²⁵ In a very recent study, Wang et al. demonstrated that the lineage-defining role of FOXA2 in NEPC extends to other distinct AR-independent CRPC subtypes and FOXA2 can collaborate with the AP-1 complex to initiate a transition to a AR-indifferent, multi-lineage stage, which subsequently facilitates tumor adaption and lineage determination.²²⁶ Notably, LSD1-mediated demethylation of FOXA2 K265, a residue analogous to FOXA1 K270, similarly regulates FOXA2 chromatin binding.²²⁶ Despite these insights, the mechanisms driving the transition between FOXA1- and FOXA2-dominated transcriptional programs during lineage switching remain poorly understood. Overall, these recent studies indicate that FOXA2 but not FOXA1 is the major driver of lineage plasticity in CRPC. However, whether FOXA1 chromatin binding can shift alongside FOXA2 in lineage-plastic PCa cells remains an open question that warrants further investigation.

FOXA1 also exhibits transcription-independent roles in immune regulation. A meta-analysis of PCa and BCa primary tumors revealed that FOXA1 negatively correlates with interferon activity.²²⁷ Overexpression of FOXA1 suppresses interferon signaling and immune responses through a transactivation-independent mechanism.²²⁷ Mechanistically, FOXA1 directly binds to the DNA-binding domain of STAT2, blocking its DNA-binding ability - an effect that persists regardless of clinically relevant FOXA1 mutations, including C-terminal truncations. In contrast, FOXA1 loss induces an immunosuppressive tumor microenvironment in PCa by repressing hypoxia signaling.²²⁸ Chemokine CCL2 was identified as a key downstream target of the FOXA1-HIF1α axis. Co-culture experiments demonstrated that FOXA1 loss promotes tumor-associated macrophage infiltration in a HIF1α-dependent manner. Future studies using animal models that recapitulate clinical features will be crucial to fully elucidate the regulatory role of FOXA1 in immune response.

3.6. Therapeutic targeting of FOXA1

Directly targeting FOXA1 remains to be challenging due to the absence of a well-defined druggable domain. However, FOXA1 collaborates with multiple actionable epigenetic modifiers to regulate downstream gene transcription. For example, FOXA1 works with TET1, a 5-methylcytosine dioxygenase, to promote local DNA demethylation.²²⁹ It also interacts with MLL complexes, facilitating the deposition of H3K4me1/2.²³⁰ FOXA1 is tightly associated with LSD1 in regulating AR-regulated enhancers.¹³⁹ We have previously shown that the levels of FOXA1 may determine the efficacy of LSD1 inhibitors in treating FOXA1-positive CRPC tumors.¹³⁷ However, it remains unclear whether LSD1 inhibition would be effective in tumors harboring FOXA1 mutations. Given the close proximity of Wing2 domain to K270, it is uncertain how hotspot mutations in FOXA1 affect its interaction with LSD1. Notably, the contrasting clinical outcomes associated with truncation mutations occurring before versus after K270 suggest that LSD1 may still modulate the transcriptional consequences of these mutations. Additionally, a recent study indicate that a subset of FKHD-ms can disrupt the LLPS properties of FOXA1, potentially leading to dramatic alterations in its transcriptional activity.¹⁹¹ Further investigation of how FOXA1 alterations reshape its epigenetic interactions could uncover novel vulnerabilities in mutant cells and inform innovative therapeutic strategies.

The recent revised classification schema underscores the importance of managing CRPC based on specific classes of FOXA1 alteration, adding complexity to the development of mutation-specific therapeutic strategies. One potential approach is to exploit the pathways selectively activated by these mutants. For example, MET, an EMT-associated gene upregulated by several FKHD-ms, can be targeted by the FDA-approved drug crizotinib,²⁰⁶ which is currently used to treat metastatic non-small cell lung cancer.^231,232 Studies using stable cell lines and xenograft models expressing FOXA1-M253K demonstrated that crizotinib effectively inhibits mutant cell growth.²⁰⁶ Similarly, the WNT inhibitor XAV939 has been shown to abrogate the invasive phenotype associated with truncating mutations.¹⁹⁴ In silico analysis of transcriptomic data from primary PCa tumors suggests that FOXA1 mutant cells might also be vulnerable to agents targeting LSD1, MAOA, PDGFB, and HSP90AB1.²³³ However, the direct actionable downstream targets of different of FOXA1 mutant classes remain largely unknown. Future studies should focus on identifying and characterizing these targets to develop more precise therapeutic strategies.

4. Conclusions

In summary, the AR-FOXA1 axis remains the central driver and a key therapeutic target in CRPC. Genetic and epigenetic alterations in the AR gene, its locus, and its coded protein primarily function to restore AR signaling under conditions of low-androgen availability or ARSi treatments. However, AR splice variants, such as AR-V7, may acquire additional oncogenic functions, driving tumor-promoting pathways that are distinct from AR-FL. Future next-generation AR-targeted therapies should focus on suppressing AR variant function or selectively targeting post-translational modifications of AR-FL/AR-V7 proteins. In particular, epigenetic cofactors that colocalize with AR at chromatin - and in some cases directly modify AR protein at chromatin sites - have emerged as critical targets for therapeutic development. Beyond AR, FOXA1 plays a multifaceted role in chromatin regulation and therapy resistance in prostate cancer. With its well-established pioneer factor function in facilitating AR chromatin binding, FOXA1 also serves as a key regulatory node during the evolution of therapy resistance. However, the complexity of FOXA1 interactions with chromatin-associated factors, including other transcription factors, epigenetic modifiers, and co-regulators, remains incompletely understood. Recent advances in AI-assisted structural analysis tools, such as AlphaFold, in combination with multi-omics technologies (e.g., single-cell transcriptomics, chromatin accessibility profiling, and proteomics), provide promising avenues to further elucidate the structural dynamics and regulatory networks of FOXA1. These insights hold substantial promise for developing novel, mutation-informed therapeutic strategies targeting FOXA1-driven prostate cancer, ultimately improving outcomes for patients with advanced disease.

Author contributions

All authors contributed to writing the manuscript and designed and prepared the figures and legends.