Triple-negative breast cancer (TNBC) continues to be the most difficult-to-treat BC subtype. TNBCs do not express the conventional BC targets currently exploited for therapeutic intervention, including the estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) [1]. TNBCs are inherently more aggressive, with an increased risk of death within 5 years post-diagnosis and a higher rate of distant metastasis and recurrence following surgery compared to other breast cancer subtypes ^[2][3][2,3]. Precision medicine approaches to targeting TNBC are complicated by inter- and intra-tumoral heterogeneity.

Over the past few decades, the androgen receptor (AR) has been highlighted as an alternate endocrine target of interest for the subset of TNBC patients positive for AR expression. AR expression is assessed by immunohistochemistry (IHC); about 20%–40% of TNBC tumors express AR [4]. AR is most highly expressed in the “luminal AR (LAR)” molecular subtype of TNBC; however, in vitro and in vivo experiments involving AR-inhibitors and AR knockdown have demonstrated that non-LAR molecular subtypes of TNBC may also be critically dependent on AR protein for viability, proliferation, migration, and invasion [5]. These results led to the notion that AR-targeted therapies could exploit this AR dependence and benefit multiple molecular subtypes of AR-positive TNBC. Although several AR inhibitors (such as bicalutamide, abiraterone acetate, and enzalutamide) have elicited significant antitumor activity in advanced clinical trials with AR-positive TNBC patients, approximately only one in five experienced sustained clinical benefit. This limited success suggests the development of therapeutic resistance among the majority of patients ^{[4][5][6][7][8]}[4,5,6,7,8]. However, the mechanisms of resistance to AR inhibition in TNBC remain underexplored. Moving the needle on our understanding of this resistance to AR-targeted therapies is imperative for improving therapeutic efficacy and for the advancement of AR inhibitory drugs to the FDA-approval stage.

Hypoxia, or limited oxygen availability due to insufficient blood supply, is a unique feature of most solid tumors [9]. Rapid and uncontrolled tumor growth can lead to tumors outgrowing their surrounding vasculature, which leads to a significant drop in normal oxygen levels (from 9%–2%) and the development of hypoxia (<2%). Oxygen is necessary for tumor cell metabolism and proliferation. Thus, slow-dividing cells located in hypoxic areas of tumors can escape most anticancer agents that kill rapidly dividing cells [10]. Furthermore, hypoxia induces profound gene expression changes that promote tumor cells’ survival, the growth of new vasculature, energy metabolism, lineage plasticity, and heterogeneity, often conferring an aggressive and metastatic phenotype ^[11][12][11,12]. Thus, hypoxia has been significantly implicated in therapeutic resistance to multiple anticancer agents and underlies poor patient prognosis ^[13][14][15][13,14,15].

The adaptive response to hypoxia is regulated by hypoxia-inducible factors (HIFs), which modulate hypoxic gene expression and regulate oxygen homeostasis. Heterodimeric HIF is composed of an oxygen-labile α subunit and a stable β subunit; HIF α/β heterodimers are thus comprised of one of three α subunits (HIF-1α, HIF-2α, or HIF-3α) and one of two β subunits (HIF-β and aryl hydrocarbon receptor nuclear translocator or ARNT). The HIF-α subunit undergoes proteasomal degradation under normoxic conditions; in contrast, HIF-β and ARNT are constitutively expressed and reside in the nucleus. Under hypoxic conditions, stabilized HIF-α translocates to the nucleus, where the active HIF-α/ARNT or HIF-α/HIF-β complex activates the transcription of a large group of target genes after binding to cognate hypoxia-responsive elements (HREs) ^[16][17][18][16,17,18].

2. Tried and Tested: Androgen Synthesis and AR Signaling as Therapeutic Targets for TNBC

Since AR-driven/-dependent TNBC is a potentially actionable subtype, androgen- and AR-targeting therapies have been explored for their ability to stymie disease progression in AR-positive TNBC ^[19][28]. Clinical trials involving TNBC patients with both early-stage and advanced disease have evaluated the clinical benefit rates (CBRs) for patients under treatment with anti-androgen drugs, including bicalutamide or enzalutamide, which were administered as single agents as well as in combination therapies. Drugs that reduce androgen production are also being evaluated for AR-positive TNBC and have shown varying degrees of efficacy but were largely supportive of targeting the androgen/AR axis in AR-positive TNBC. Bicalutamide is FDA-approved for the treatment of advanced PC in combination with a luteinizing hormone-releasing hormone analog. Bicalutamide is orally available and is a competitive inhibitor of AR that permits AR nuclear localization and binding to chromatin but preferentially recruits AR corepressors rather than coactivators ^[20][29]. A clinical study conducted by investigators at the Memorial Sloan Kettering Cancer Center (NCT00468715/TBCRC011) showed that daily treatment with 150 mg bicalutamide led to 19% CBR lasting longer than 6 months and a median progression-free survival (PFS) of 12 weeks in ER-negative/PR-negative/AR-positive metastatic BC patients [6]. These proof-of-concept data in AR-positive BC provided an impetus to pursue next-generation AR antagonists because bicalutamide can have partial agonist effects ^[21][30]. PC patients who develop resistance to bicalutamide are oftentimes responsive to the next-generation antagonist enzalutamide, which (a) has an eightfold higher affinity for AR than bicalutamide and (b) is a robust inhibitor of AR signaling that blocks AR nuclear translocation and coactivator interactions and attenuates the DNA binding capacity of AR ^[22][31]. In a phase I dose-escalation study in patients with metastatic BC in 2013, enzalutamide was well tolerated at 160 mg daily. In a 2015 phase II study, among the AR-positive patients evaluable for response, the CBR (defined as complete response + partial response + stable disease) at 16 weeks was 42% with patients receiving 160 mg enzalutamide daily. About 34% of the participants in the study continued to show clinical benefits after 24 weeks. Furthermore, when a novel, binary, gene signature-based predictive biomarker that was reflective of AR signaling was used to stratify patients, the outcomes for diagnostic-positive patients were more favorable in all response measures, including CBR16 (39% vs 11%), CBR24 (36% vs 6%), and median PFS (16.1 weeks vs 8.1 weeks), when compared with those of patients who tested negative for this companion diagnostic. Interestingly, this companion diagnostic was a stronger predictor of response to enzalutamide than AR expression (evaluated by IHC) alone. These responses to an AR antagonist supported (a) the development of anti-AR therapy for the treatment of a subgroup of patients with TNBC [8], (b) the development of companion diagnostics that better identify the androgen/AR-dependence of tumors, and (c) the pre-planned incorporation of such predictive biomarkers (that are superior to AR positivity alone) and novel biomarker discovery into clinical trials. The ENDEAR trial, which was scheduled to be an international, double-blind, randomized, placebo-controlled, three-arm phase III study in TNBC patients who tested positive for the companion diagnostic developed during the phase II trial previously mentioned, was unfortunately abandoned in 2017 based on a joint decision by the partnering pharmaceutical companies, citing a need to improve the clarity on the role of androgen signaling in BC. Based on the promising data available regarding the metastatic setting, a phase II study is currently evaluating the feasibility of enzalutamide as endocrine therapy in the adjuvant setting (1 year of enzalutamide 160 mg orally daily) for patients with early-stage (Stages I–III) TNBC of the LAR subtype (NCT02750358). This single-arm trial has met its primary endpoint of feasibility; however, the median overall survival (OS) has not yet been reached. Apalutamide and darolutamide are two promising new-generation AR inhibitors under evaluation in phase III clinical trials in patients with non-metastatic CRPC (NCT01946204 and NCT02200614, respectively). Apalutamide, structurally similar to enzalutamide, has exhibited a similar success rate as enzalutamide at a lower dosage in an LNCaP xenograft mouse model but does not induce AR nuclear translocation or DNA binding ^[23][32]. Darolutamide is particularly interesting owing to its ability to also antagonize the AR mutants F876 L, W741 L, and T877A ^[24][33]. The applicability and efficacy of these drugs in the context of AR-positive TNBC are yet to be extensively studied. Agents that target steroidogenic enzymes to impair intracrine and adrenal androgen biosynthesis and concomitantly reduce downstream estrogen synthesis have also been tested in clinical trials for TNBC; abiraterone acetate, a CYP17 inhibitor, is one such therapeutic, which is a robust, orally available, selective inhibitor of both 17α-hydroxylase and C17, 20-lyase. In a phase II multicenter trial that aimed to assess the safety and efficacy of abiraterone acetate in combination with prednisone (which has to be included to offset the increase in aldosterone production that results from reduced cortisol levels) in women with metastatic or inoperable locally advanced AR-positive TNBC, the CBR at 6 months was 20.0% and median PFS was 2.8 months. This study suggested that some TNBC patients with molecular apocrine-like tumors may benefit from the coadministration of abiraterone acetate and prednisone. Seviteronel, an oral, selective CYP17-lyase inhibitor and AR antagonist, is also currently being evaluated as a novel option for the treatment of AR-positive TNBC because this agent does not cause a decrease in cortisol and no steroid supplementation is required. Preliminary pharmacokinetic data from a large phase I/II trial for seviteronel in AR-positive TNBC and ER-positive metastatic BC suggested that the bioavailability of seviteronel may be sex-specific ^[25][34]. The phase II trial in this population established the promising CYP17-lyase inhibitory activity of seviteronel, as evidenced by an early and potent reduction in both estradiol and testosterone ^[26][35]. Pre-clinical data also advocate for the administration of seviteronel as a radiotherapy-sensitizing agent in AR-positive TNBC ^[27][36]. Since other oncogenic aberrations may co-exist with AR dependency, combination regimens involving AR inhibitors and other targeted therapies are also underway for AR-positive TNBC. Palbociclib, an orally administered and highly selective and reversible CDK4/6 inhibitor, prevents the onset of the S phase, resulting in cytotoxicity and growth arrest. Studies have shown that luminal phenotypes, elevated cyclin D1 and Rb protein expression, and reduced p16 expression are all associated with sensitivity to Palbociclib ^[28][37]. AR-positive TNBC often exhibits a luminal gene expression profile and has intact Rb protein; these observations provide a rationale for combining AR inhibitors with Palbociclib for the treatment of AR-positive TNBC. A phase I/II trial (NCT02605486) in patients with metastatic BC ^[26][35] met its pre-specified endpoint of PFS of at least 6 months. A trial testing the combination of bicalutamide and ribociclib (another CDK4/6 inhibitor) is also underway (NCT03090165). AR antagonists have also been paired with drugs that target the PI3K/AKT/mTOR pathways because AR-positive BCs are often associated with activating PIK3CA mutations and pAKT ^[29][30][31][38,39,40]. Pre-clinical studies showed that LAR cell lines are significantly more sensitive to NVP-BEZ235 (a dual PI3 K/mTOR inhibitor) when compared to cell lines of basal-like TNBC molecular subtypes ^[31][40]. These findings catalyzed an investigator-initiated, multi-institutional phase I/II study (TBCRC032) that evaluated the safety and efficacy of enzalutamide alone or in combination with the PI3K inhibitor, taselisib, in patients with metastatic AR-positive (≥10%) BC ^[32][41]. The combination of enzalutamide and taselisib significantly increased the CBR in the AR-positive TNBC patients. Importantly, analyses confirmed earlier findings that AR protein expression alone is insufficient for identifying patients with AR-dependent tumors, and a greater understanding of TNBC molecular subtypes and AR splice variants may identify patients more or less likely to respond to AR antagonists. Since the LAR subtype is generally resistant to conventional multidrug neoadjuvant chemotherapy (NAC) regimens and exhibits low rates of pathologic complete response, or pCR ^[33][42], a phase II clinical trial was carried out to assess whether combining AR inhibition with NAC would improve pCR or near-pCR in early-stage AR-positive (10%) TNBC patients treated with enzalutamide and weekly paclitaxel (NCT02689427). The results of this trial showed that 33.3% of TNBC patients who did not respond to conventional NAC showed a pCR with the enzalutamide and paclitaxel combination; notably, all responders showed an upregulated androgen response pathway as measured by transcriptomic analysis in pre-treatment biopsies ^[34][43].

3. The Arch Nemesis: Hypoxia as a Barrier in Androgen/AR Signaling Inhibition in Cancer

A significant body of evidence shows that tumor hypoxia interferes with therapies that disrupt AR signaling or reduce androgen levels in PC. A recent study showed that enzalutamide induces hypoxia and microenvironment adaptation and that two cytokines—interleukin-8 (IL8) and vascular endothelial growth factor A (VEGF-A)—upregulated by this treatment-induced hypoxia may contribute to this treatment resistance ^[35][44]. This study also showed that the concurrent inhibition of both IL8 and VEGF-A in PC pre-clinical models prolonged tumor sensitivity to enzalutamide. One primary mechanism of this hypoxia-mediated resistance to AR-targeted therapies is via upregulating AR signaling. Several studies have shown that overexpressing HIF-1α in PC cells stimulated AR signaling with the androgen dihydrotestosterone (DHT), enhanced AR transcriptional activity, and subsequently increased secretion of VEGF ^{[19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42][43]}[21,22,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46]. Specifically, hypoxia increased AR translocation to the nucleus and recruitment to the prostate-specific antigen (PSA) promoter ^[36][44][19,21]. It was also shown via chromatin immunoprecipitation that HIF-1α interacts with AR on the human PSA gene promoter in DHT-stimulated PC cells, suggesting that HIF-1α and AR may be cooperating to activate AR target genes ^[45][47]. Collectively, these findings suggest that hypoxia enhances the AR signal transduction pathway. As a result, hypoxia has been reported to drive insensitivity to androgen/AR-targeted therapies in PC. In PC cells, DHT-mediated AR signaling induced HIF-1α expression, and the dual treatment of enzalutamide and HIF-1α inhibition abrogated cell growth, induced apoptosis, and reduced VEGF levels more effectively than the administration of each treatment alone ^[39][25]. These studies thus provide a strong impetus for combining therapeutic targeting of the HIF-1α and AR pathways as a strategy to combat enzalutamide resistance. It was also discovered that HIF-1α coordinates AR translocation to the nucleus via binding to β-catenin and AR to form a ternary complex that binds to the androgen response elements ^[46][48]. Under hypoxic conditions, the dietary polyphenol, resveratrol, reduced HIF-1α levels and inhibited β-catenin nuclear accumulation to suppress hypoxia-induced AR transactivation in prostate tumors possibly in a proteasome-independent manner ^[40][26]. These findings suggest that targeting the Wnt/β-catenin signaling axis in combination with androgen/AR signaling may also circumvent hypoxia-mediated resistance to AR-targeted therapies. Furthermore, the AR-signaling-mediated induction of HIF-1 was shown to occur via the phosphatidylinositol 3-kinase (PI3K) signaling as the inhibition of this pathway blocked HIF-1 activation ^[43][46]. Lastly, it was recently shown that hypoxia simultaneously upregulates HIF-1α and transient receptor potential melastatin subfamily member 7 (TRPM7) in androgen-independent PC cells and that knockdown of TRPM7 inhibited hypoxia-induced migration and invasion via the increased RACK1-mediated degradation of HIF-1α ^[47][49]. Some groups have begun to unravel other mechanisms of intra-tumoral hypoxia-mediated resistance to androgen/AR signaling inhibition. Hypoxia-mediated resistance to the potent second-generation AR inhibitor, enzalutamide, was shown to mechanistically occur in PC via restoration of glucose-6-phosphate isomerase (GPI), which is transcriptionally repressed by AR in hypoxia and maintains glucose metabolism and energy homeostasis under hypoxic conditions ^[37][22]. AR inhibition restores GPI, which was shown to promote resistance to this inhibition via redirecting glucose flux from the androgen/AR-dependent pentose phosphate pathway to the hypoxia-induced glycolysis pathway. Thus, GPI inhibition was shown to circumvent enzalutamide resistance in vivo. Cancer stem cells have also been implicated in treatment resistance to AR-targeted therapies in PC ^[48][49][50,51]. ADT has been shown to increase cancer stem cell numbers in prostate tumors and HIF signaling in response to hypoxia and induce the expression of stemness and EMT genes that promote the emergence of these cancer stem cells ^[50][51][52][52,53,54]. Furthermore, HIF-1α was shown to regulate CD44, which is expressed on stem-like BC cells, by increasing the presence of CD44 molecules and the percentage of CD44 positive cells under hypoxic conditions ^[53][55]. Thus, the expansion of the cancer stem cell subpopulation upon the induction of adaptive HIF signaling may also be conferring resistance to AR-targeted therapies.

4. Double Trouble: Hypoxia in Triple-Negative Breast Cancer

Hypoxic regions have been observed in over 50% of breast tumors but more frequently in TNBC. Genomic profiling revealed high HIF-1α signaling activity in basal-like/TNBCs ^[54][56]. Specifically, HIF-1α expression was observed to be highly upregulated in TNBCs to withstand a hypoxic tumor microenvironment. TNBCs have also been shown to upregulate bicarbonate transporters (NDBTs) such as SLCA4 and SLC4A5, which regulate tumor pH levels when adapting to hypoxic conditions ^[55][57]. The enhanced hypoxic signaling observed in TNBC has been suggested to underlie advanced progression and treatment resistance in TNBC, which may also be undermining the efficacy of AR inhibition. Specifically, HIF-1α upregulates Snail expression, which increases the migration and invasiveness of BC cells by downregulating E-cadherin. Farnesyltransferase treatment of TNBC cells to inhibit HIF-1α expression decreased mRNA levels of HIF-1α pathway genes such as Snail, glucose transporter 1, pyruvate dehydrogenase kinase 1, and lactate dehydrogenase A ^[56][58]. HIF-1α and HIF-2α silencing in TNBC xenografts significantly reduced tumor growth ^[57][59]. In TNBC, HIF-1α upregulated complement 1 q binding protein (C1QBP), which blocked metastasis of TNBC cells and increased their sensitivity to paclitaxel ^[58][60]. The depletion of C1QBP in TNBC cells also decreased VCAM-1 expression and suppressed the activation of hypoxia-induced protein kinase C-nuclear factor-κappa B signaling ^[58][60]. The nano-treatment of TNBC in vitro and in vivo with the hypoxia-targeting drug tirapazamine, effectively abrogated tumor cell growth and progression in hypoxic regions ^[41][27]. The inhibition of TNBC hypoxia-induced NDBTs was shown to notably suppress migration and invasion in vitro and in vivo via attenuating the NDBT-mediated hypoxic phospho-signaling activation and modulating the expression of critical EMT-related genes, such as vimentin, which prevents downregulation of E-cadherin ^[55][57]. The inhibition of the hypoxia-induced carbonic anhydrase IX in TNBC cell lines impaired their ability to form new vasculature and mammospheres as well as metastasize ^[59][61]. Hence, targeting hypoxia in TNBC has been shown to increase tumor sensitivity to chemotherapies such as cisplatin, doxorubicin, and 5-fluorouracil ^{[60][61][62][63]}[62,63,64,65]. Hypoxia has also been shown to predict prognosis in TNBC. Positive HIF-1α IHC expression significantly correlated with greater tumor size, higher histological grade, positive lymph node status, and higher tumor TNM stage as well as poorer postoperative survival ^[64][66]. The enrichment of a hypoxia-related three-gene signature model was shown to predict poorer overall survival (OS) in TNBC patients ^[65][67]. Furthermore, a high combined hypoxia and immune-base gene signature score predicted a poorer prognosis among TNBC patients ^[66][68].