FOXA1: a transcription factor with parallel functions in development and cancer

Pancreas and kidney

Mice that are homozygous null for Foxa1 lack any overt morphological abnormalities [21,22]. However, they are postnatally lethal due to a combined phenotype of severe hypoglycaemia and dehydration. Interestingly, loss of Foxa1 also dramatically decreases circulating glucagon levels. This is paradoxical to the observed hypoglycaemic state because the pancreas normally responds to hypoglycaemia by releasing glucagon, which then induces conversion of glycogen stores into glucose in the liver. Further investigation revealed that FOXA1 is normally expressed in glucagon-expressing pancreatic α-islet cells, where it is necessary to transcriptionally activate the proglucagon promoter (Gcg) [21]. Loss of Foxa1 reduces glucagon transcription, subsequently decreasing glucagon secretion into the circulation. These results confirm a role for FOXA1 in regulating glucose homoeostasis through directing α-islet cell function. In addition to the loss of proglucagon gene expression, Foxa1 null mice also fail to secrete insulin in response to glucose administration [22]. This defect is attributed to the up-regulation of mitochondrial Ucp2 (uncoupling protein 2) that occurs in the pancreatic β-cells of Foxal null mice [26]. Increased UCP2 expression uncouples oxidative phosphorylation, which decreases glucose-mediated ATP synthesis and insulin secretion from β-islet cells. Together, these studies revealed that FOXA1 modulates glucose homoeostasis through multiple mechanisms: transcriptional activation of the proglucagon gene and repression of Ucp2 expression.

To investigate the role of FOXA1 during pancreatic development without the physiological complexity of a global knockout, mice were generated with loxP sites flanking exon 2 of Foxa1. These mice were then crossed with Pdx1-Cre^Early transgenics mice [27]. PDX-1 (pancreas and duodenal homeobox protein-1) expressing cells generate each cell type (exocrine, endocrine and duct) of the pancreas allowing for pancreas-selective expression of Cre DNA recombinase and disruption of floxed transgenes [28]. Surprisingly, Pdx1-Cre^Early; Foxa1^loxP/loxP mice are viable and fertile indicating that the postnatal lethality observed in the germ-line knockouts [21] was not solely due to an isolated defect in the pancreas, but likely also involved severe dehydration [27]. Compound conditional knockout mouse studies subsequently revealed that FOXA1 and FOXA2 co-operatively control pancreatic acinar and islet morphogenesis [27]. The presence of at least one wild-type allele of Foxa2 can compensate for complete loss of Foxa1 in the pancreas, resulting in viable mice. Interestingly, the absence of one or two wild-type alleles of Foxa1 dictates the level of specification of exocrine and endocrine cell lineages in the pancreas of mice completely lacking Foxa2 [27,29]. In other words, mice that are wild-type, heterozygous null and homozygous null for Foxa1 in the pancreas have the most normal, intermediary and severe disruption respectively of pancreatic lineage specification when combined with the homozygous null Foxa2 allele. For a detailed description of FOXA1/2 compound mutants that have been generated to investigate their functions in the pancreas and other tissues, see [3,19]. Notably, mice null for Foxa3 are hypoglycaemic when fasted due to decreased expression of the GLUT2 (glucose transporter) [24], and thus each FOXA family member is necessary for maintaining glucose homoeostasis.

In addition to the deregulation of pancreatic proglucagon and Ucp2 expression, pups lacking FOXA1 are severely dehydrated [21], prompting studies to investigate FOXA1 function in the kidney. Foxa1 null mice develop nephrogenic diabetes insipidus as shown by the inability to respond to arginine-vasopressin [ADH (anti-diuretic hormone)] [30]. Interestingly, the development of this disease could not be explained by a decrease in the expression of genes encoding the vasopressin 2 receptor, aquaporins or other proteins involved in water reabsorption. Although the precise mechanism by which FOXA1 modulates kidney function remains to be determined, it has been postulated that the lethality of the Foxa1 null animals occurs as a result of the combined impact of pancreatic insufficiency and kidney defects. This is supported by the viability of mice with a pancreatic specific deletion of Foxa1 (Pdx1-Cre^Early; Foxa1^loxP/oxP) as discussed above. It is unclear whether the kidney defect in Foxa1 null mice is due to a direct effect on this organ or changes that occur in response to an abnormal developmental programme. Use of a kidney-specific inducible knockout would facilitate addressing this question, as well as provide a resource for identifying the specific genes which are controlled by FOXA1 that ultimately establish arginine-vasopressin responsiveness. In addition, such animals would generate a potential model for nephrogenic diabetes insipidus in humans. Identifying the specific role FOXA1 plays in water reuptake should ultimately provide important insights for clinical management of this disease.

Liver

Given the role for FOXA proteins in regulating liver-specific gene transcription [2], it was hypothesized that FOXA1 would be necessary for normal liver development. However, while Foxa1 null mice die postnatally [21], they have a morphologically normal liver. Similar to the maintenance of normal pancreatic morphology in Foxa1 knockouts due to FOXA2 compensation, FOXA2 also offsets loss of Foxa1 in the liver. When investigated combinatorially, genomic loss of Foxa1, within the context of an endoderm-specific conditional deletion of Foxa2 (Foxa3-Cre; Foxa2^loxP/loxP), completely blocked the onset of liver specification whereas conditional loss of Foxa2 alone had no effect [31]. The endodermal disruption of Foxa2 occurs at E8.5 (embryonic day 8.5) when using Foxa3-Cre to induce recombination, and this precedes liver specification. To investigate liver development after the onset of specification, Alf-pCre; Foxa1^loxp/loxp; Foxa2^loxp/loxp mice were generated [32]. In these mice, Cre expression occurs at E10.5 in the liver primordium, but complete loss of Foxa1/2 was not observed until P2 (postnatal day 2). These mice undergo normal hepatocyte differentiation, but during adulthood develop bile duct hyperplasia and fibrosis as a result of increased IL-6 (interleukin-6) expression. Under normal conditions, FOXA1/2 and the GR (glucocorticoid receptor) co-operate to repress Il6 transcription. In the absence of FOXA1/2, this repression is ablated, allowing NF-κB (nuclear factor κB)-induced up-regulation of Il6. The excessive IL-6 then induces proliferation of bile duct epithelial cells (cholangiocytes) [32]. These results demonstrate the combined importance of FOXA1 and FOXA2 in liver cell specification and differentiation, and suggest a potential role for FOXA1/2 in human diseases of the liver. Along these lines, FOXA2 expression is decreased in the livers of humans with cholestatic disease [33], but a role for FOXA1 in biliary diseases remains to be established.

Lung

FOXA1 is observed in the lung bud at E10.5, and is maintained in mature conducting airway epithelial and secretory alveolar type II cells [18]. Mice null for Foxa1 have delayed lung development [34]. Normally, the fetal lung develops through three distinct morphological phases: pseudoglandular, canalicular and saccular. While lungs of wild-type mice reach the saccular stage by E16.5, those lacking FOXA1 are delayed in the canalicular stage until E17.5. Similarly, lung septation in postnatal Foxa1 knockout mice (P5) was significantly delayed. Loss of Foxa1 also decreased the expression of lung differentiation markers [e.g. CCSP (clara cell secretory protein), SP-B (pro-surfactant protein-B) and SP-C (pro-surfactant protein-C)] during embryogenesis. This is at least partly due to the dependency of both CCSP (Scbglal) [35] and SP-B (Sftpb) [36,37] promoters on FOXA1 for activation in lung epithelium. How FOXA1 modulates SP-C expression is not known. Although FOXA1 deficiency causes a developmental delay during embryogenesis, histology of the knockout lungs was indistinguishable from controls by P13. That said, cuboidal type II cells had higher glycogen levels, a decreased number of lamellar bodies and less surfactant than those in wild-type lungs. As in other tissues (e.g. pancreas and liver), FOXA2 eventually compensates for loss of Foxa1 in the lung. Complete loss of Foxa1, combined with doxycycline-induced deletion of Foxa2 driven by Cre expression under control of the SP-C lung-specific promoter [Foxa1^−/− /SPC-rtTA^−ltg/(tetO)₇Cre^−ltg/Foxa2^loxP/loxP] during embryogenesis, drastically impairs lung branching morphogenesis and epithelial differentiation [38]. Similar to Foxa1 null lungs, CCSP expression is absent in the compound mutants. In addition, loss of Foxa1 and Foxa2 decreases mRNA expression of sonic hedgehog (Shh), a necessary activator of lung epithelial specification [39,40]. FOXA1 expression is increased in a rat model of ALI (acute lung injury), in which it is suggested to function in alveolar type II epithelial cell apoptosis [41], and is similarly necessary for H₂O₂-induced apoptosis in A549 cells [42]. Together, these results imply FOXA1 may participate in the onset or progression of lung diseases not limited to cancer as discussed below. It will be important to generate mice with lung specific conditional deletion or transgenic overexpression of FOXA1 to gain mechanistic insight into the precise role FOXA1 may play in these diseases.

Brain

FOXA1 and FOXA2 are expressed in the notochord and floor-plate of the fetal brain [14–16], and Foxa2 null mice are embryonic lethal as a result of failed node and notochord development [20,43]. Although Foxa1 null pups do not exhibit overt CNS (central nervous system) architectural defects [21,22], they have delayed maturation of mature mDA (midbrain dopaminergic) neurons [44]. Conditional loss of Foxa2 using a Nestin-Cre transgene dose-dependently exacerbates the Foxa1 null phenotype, indicating that even one copy of Foxa2 compensates for Foxa1 loss and vice versa. The phenotypic response to FOXA loss has been attributed to the ability of these proteins to control expression of neurogenin 2, Nurr1 and EN1 (engrailed 1), three factors essential for neuronal differentiation [45–47]. Complimenting these studies, conditional loss of both Foxal and Foxa2, through breeding with En1-klCre transgenic mice, led to disruption of each FOXA protein earlier in development (E9.75) [48] compared with Nestin-Cre (E10.5)-mediated disruption [44]. Through these two approaches, Lin et al. [48] showed that FOXA1/2 positively regulate expression of the LIM homeodomain transcription factors Lmx1a/1b and negatively regulate the morphogenically associated homeodomain protein Nkx2.2 earlier in development. Intriguingly, FOXA1 and FOXA2 are redundantly necessary for the initiation of SHH expression in the midbrain [48,49], paralleling the regulation of Shh by FOXA1 observed in the lung that was discussed previously. Further demonstrating the importance of FOXA in SHH signalling, at least in this context, FOXA1/2 are also responsible for restricting the expression of Gli1 and Gli2, transcriptional mediators of SHH signalling, and patched-1, the receptor for SHH, in the ventral midbrain [49].

Gastrointestinal tract

FOXA1/2 are also expressed in the epithelia of the gastrointestinal tract [18]. To investigate how the combined loss of Foxa1 and Foxa2 specifically affects intestinal development, floxed alleles were conditionally disrupted using the villin promoter to direct expression of Cre (Vil-Cre) to the intestinal epithelium [50]. The compound-conditional mutants are viable, but grow slower, and have decreased body weight due to a reduction in lean muscle mass and less body fat. Although overall intestinal morphology is normal, the number, secretory capacity and differentiation of goblet cells are reduced in the combined mutants. FOXA1, but not FOXA2, was found to bind to the promoter of the Muc2 gene that encodes mucin 2. These results compliment previous in vitro studies in which FOXA1 and FOXA2 can both transactivate the Muc2 promoter, with FOXA1 being more efficient [51]. In addition to diminished goblet cell formation, loss of Foxa1/2 leads to decreased differentiation of L- and D-cell enteroendocrine lineages [51], which produce hormones involved in intestinal function. Paralleling in vitro regulation of the proglucagon promoter by FOXA1/2 [21,52], intestines of the combined conditional mutants lack D-cells, as shown by loss of cells expressing GLP (glucagon-like peptides)-1 and -2. In addition, cell populations that express somatostatin and PYY (peptide YY) (D- and L-cells respectively) were decreased. Furthermore, expression of Islet-1 and Pax6, two transcription factors involved in enteroendocrine cell differentiation, are decreased, providing evidence that FOXA1/2 precede Islet-1 and Pax6 expression in the enteroendocrine transcriptional hierarchy. These results indicate that FOXA proteins are essential for development of both mucin-producing goblet and enteroendocrine cellular populations. The loss of these cell types probably results in insufficient nutrient absorption required for normal growth, thus providing an explanation for the decreased growth rate of these mice during early life.

Prostate

Morphogenesis of the prostate is dependent on stromal AR expression, whereas prostatic secretory function relies on epithelial AR expression [53,54]. FOXA1 is an established regulator of AR transcriptional activity in prostate cancer cells (reviewed below) leading to an investigation of FOXA1 during prostate development. Expression analyses revealed that FOXA1 is present in prostate tissue from the onset of development at E18 in the urogenital sinus epithelium throughout adulthood, with no detectable expression in the stroma [55–57]. Considering the postnatal lethality of the Foxa1 null mice, Matusik and co-workers [58] investigated prostate morphogenesis in the absence of FOXA1 through utilization of two distinct rescue strategies: renal capsule grafting and tissue recombination. While loss of Foxa1 does not alter rudiment formation, luminal epithelial lineage differentiation is blocked as shown by luminal CK8 (cytokeratin 8) staining being present only in concert with the basal epithelial lineage marker p63. Furthermore, CK5-positive basal cell and α-SMA (α-smooth muscle actin)-positive smooth muscle cell populations were aberrantly expanded. The epithelium within the Foxa1 null prostate also fails to polarize and undergo proper lumen formation, and, as expected, results in a secretion defect. While prostatic tissue lacking epithelial AR has a similar secretory phenotype [54], Foxa1 null prostates maintain epithelial AR expression. Since FOXA1 facilitates AR transcriptional activity as discussed below, it is possible that while FOXA1 is not required for AR expression, it is necessary for AR function in the developing prostate, thus resulting in the observed secretory defect. This is supported by the loss of Pbsn (probasin) and Sbp (spermine-binding protein) mRNA, which are normally stimulated by AR activation, in the Foxa1 null prostate glands [58].

Interestingly, FOXA2 is aberrantly up-regulated in Foxa1 knockout prostate epithelium [58]. Normally, FOXA2 is only found in prostatic epithelia during embryogenesis, and in a small population of mature basal epithelial cells that co-express syn-aptophysin, an NE (neuroendocrine) marker [57,59]. Thus, unlike in other tissues (e.g. lung, liver and pancreas), FOXA2 does not compensate for FOXA1 loss in the prostate. As discussed below with regard to prostate cancer, the up-regulation of FOXA2 in the Foxa1 null prostates may enable AR function in an androgen-independent manner, resulting in the observed hyper-proliferative phenotype. Of note, loss of Foxa1 also decreases expression of the homeobox gene Nkx3.1, a putative tumour suppressor whose deficiency also leads to prostatic hyperplasia [60]. Prostates lacking FOXA1 also have increased SHH, an established contributor to prostate carcinogenesis [61–63]. Interestingly, the negative regulation of SHH by FOXA1 in the prostate differs from the positive regulation observed in lung and brain. This may be due to the absence of FOXA1/FOXA2 collaboration in the prostate, but this possibility has not yet been tested.

Mammary gland

In breast cancer, FOXA1 expression positively correlates with that of ER and another transcription factor, GATA3 [64–66], both of which are necessary mediators of normal mammary gland development [67–70]. This strong correlation, the known up-regulation of FOXA1 as a result of GATA3 overexpression in HEK-293T cells [HEK-293 cells expressing the large T-antigen of SV40 (simian virus 40)] [71], and the co-expression of FOXA1 and GATA3 in mouse mammary epithelia, prompted Kouros-Mehr et al. [70] to investigate whether GATA3 could regulate FOXA1 expression in the normal mammary gland. Using ChIP (chromatin immunoprecipitation), the authors showed that GATA3 binds to the Foxal promoter in mouse primary mammary epithelial cells. Further investigation into a mammary-specific role for the GATA3-interacting protein, FOG2 (friend of GATA2), revealed that conditional disruption of Fog2 by WAP (whey acidic protein)-Cre-mediated excision resulted in decreased Foxal mRNA expression during involution [72]. However, limitations in the technical design of this study, specifically the lack of epithelial-specific normalization approaches, leave open the possibility that FOG2 does not directly regulate Foxal expression. A more recent study investigating FOXA1 protein expression in virgin adult mammary glands lacking Gata3 following MMTV (murine mammary tumour virus)-Cre-induced conditional deletion failed to detect a change in the percentage of FOXA1-positive cells in response to Gata3 disruption compared with wild-type controls [73]. These results indicate that GATA3 is not required for FOXA1 expression in the normal mammary gland. However, they do not negate GATA3 binding to the Foxal promoter in primary mammary cells [70] and leave open the possibility that GATA3 may regulate FOXA1 expression in other tissue types [71].

The observed correlation between FOXA1 and ER in human breast cancer also led to in vitro functional studies that revealed an essential role for FOXA1 in mediating transcriptional regulation by ER, as reviewed below. These results suggested that FOXA1 may be necessary during mammary morphogenesis simply due to its requirement for full ER activity. Supporting this hypothesis, FOXA1 and ER are co-expressed in mammary luminal epithelium during development [73]. To directly evaluate the role for FOXA1 in this process, we rescued Foxa1 null mammary tissue via renal capsule grafting and orthotopic transplantation [73]. Loss of Foxa1 completely blocks epithelial outgrowth in the orthotopic transplant model, and inhibits ductal invasion in renally transplanted glands. Unlike in the prostate, Foxal knockout mammary epithelium exhibits no observable alterations in the lineage markers α-SMA, CK8 and E-cadherin. Another difference between mammary and prostate responses to FOXA1 loss involves expression of their cognate steroid hormone receptors. While AR expression is maintained in the Foxal null prostate, loss of Foxal in the mammary gland is accompanied by a lack of epithelial ER expression. In support of these in vivo findings, FOXA1 binds to the ESRl promoter and is required for ER mRNA and protein expression in breast cancer cells [73]. Combined, these results revealed that FOXA1 is not only required for ER activity but also for its expression. The transcriptional hierarchy of FOXA1, ER and GATA3 has also been further refined. Mammary glands deficient in ER maintain expression of FOXA1 and Gata3 [73]. In contrast, mammary glands deficient in GATA3 maintain FOXA1 [73], but lose ER [69,70]. Together, these results place both GATA-3 and FOXA1 upstream of ER, and also reveal that the expression of GATA-3 and FOXA1 are not interdependent. Interestingly, unlike the secretary defects seen in the pancreas, lung, intestine and prostate discussed above, loss of Foxal in the mammary gland does not affect lobulo-alveolar maturation and milk production [73]. This result was unexpected because epithelial ER is required for lobulo-alveologenesis [67,74,75]. One explanation is the possibility that FOXA1 actively represses maturation of the alveolar lineage, and its loss in virgin epithelia leads to premature alveologenesis independently of ER. To test this possibility, a novel mouse model must be developed wherein FOXA1 expression is enforced during pregnancy. If alveologenesis fails to occur in the presence of pregnancy-associated hormones, this would suggest that FOXA1 plays two, somewhat opposing, roles in the mammary gland. Specifically, it may promote morphogenesis of ducts, but inhibit terminal differentiation of alveoli. Lastly, it is interesting to note that like in the prostate, FOXA2 is not expressed in the mature mammary gland [18]. Thus the mammary gland represents a second tissue where FOXA1 acts independently of its related FOXA member to control morphogenesis.

FOXA1 in prostate cancer

As previously discussed, FOXA1 is critical for prostate development and loss of Foxa1 causes epithelial hyper-proliferation, activation of oncogenic SHH and decreased expression of the tumour suppressor Nkx3.1 [58]. These results suggest a tumour suppressive role for FOXA1 in prostate epithelia, and pose several questions regarding if, and how, FOXA1 regulates prostatic tumour initiation, progression and/or metastasis. While Foxa1 null prostates maintain epithelial AR expression, they lack AR-dependent secretory differentiation, implying that FOXA1 may modulate AR activation of this process [58]. Given the majority of prostate cancers are driven by aberrant AR signalling, much work has been done to define a role for FOXA1 in this process.

FOXA1 expression in prostate cancer

The LPB-Tag (T antigen) LADY mouse model of prostate cancer expresses SV40 large Tag under control of a large (12 kb) fragment of the prostate-specific probasin promoter (LPB), giving rise to prostate tumours that recapitulate human disease [96]. LADY tumours have increased nuclear expression of FOXA1 and AR that accompanies Tag [59]. Interestingly, FOXA1 is retained in both androgen-dependent (12T-7f) and androgen-independent (12T-10) models. Furthermore, 12T-10-associated liver metastases maintain FOXA1. Together, these results argue against a tumour suppressive role for FOXA1 and suggest that it may function independent of androgens. FOXA1 has also been examined in the TRAMP (transgenic adenocarcinoma of mouse prostate) mouse model that differs from LADY by expressing both SV40 large T- and small t-antigens downstream of a minimal probasin promoter [97,98]. Chiaverotti et al. [99] examined murine strain-specific differences on TRAMP-induced tumour formation and found that expression of the transgene in the FVB/N background decreases survival and increases the incidence of NE carcinomas compared with TRAMP-C57BL/6 animals. NE carcinoma, although rare in humans, is unresponsive to androgen ablation therapy [100]. Interestingly, TRAMP-induced NE carcinomas co-express FOXA1 and FOXA2. Normal prostate epithelia and atypical hyperplasia have high levels of FOXA1 and AR, whereas FOXA2 is only expressed within a small percentage of mature prostate basal epithelial cells [57,59]. This, along with co-expression of epithelial (E-cadherin) and NE (syn-aptophysin) lineage markers, support the possibility that NE carcinomas initiate within a bipotential epithelial population that is more susceptible to tumorigenic processes in FVB/N compared with C57BL/6 mice. The combined expression of FOXA1 and FOXA2 in NE tumours may provide a growth advantage in the absence of androgens. Indeed, in vitro analyses revealed that FOXA2 activates PSA (prostate-specific antigen) promoter activity in an androgen-independent fashion, and suggested that the conversion from FOXA2-negativity into FOXA2-positivity may be responsible for the transition from androgen-dependence to androgen-independence [59].

FOXA1 mRNA positively correlates with human prostate cancer cell line and xenograft models that have been defined as androgen dependent [101]. Examination of FOXA1 and FOXA2 expression in a small cohort of human prostate adenocarcinomas indicated that FOXA1 is present in all samples independent of Gleason score, with the level of expression being indistinguishable from benign tissue [59]. In contrast, FOXA2 is undetectable in low-grade prostate adenocarcinomas, but is found in NE small cell carcinomas similar to the TRAMP mouse model discussed above. The expression of FOXA1 was not reported for these human NE samples [59]. Recent studies investigating FOXA1 protein levels in primary and metastatic human prostate tumours, revealed high FOXA1 expression positively correlates with metastatic disease and poorer outcomes [102,103]. In addition, an amplicon at 14q21, which includes FOXA1, has been identified in metastatic prostate tumours [104]. Lastly, examination of FOXA1 and AR expression revealed poorer outcomes from prostate cancer when these two proteins are co-expressed compared with expression of AR alone [103]. Overall, these studies confirm that FOXA1 expression is maintained in mouse models of prostate cancer as well as human tumours, but more robust expression analyses in human prostate tumours are necessary to delineate its association with patient outcomes and anti-androgen treatment response. In addition, the specific functions that FOXA1 may have in androgen-dependent compared with androgen-independent disease require further exploration. Determining if FOXA1 can be used as a predictor of anti-androgen therapeutic success could provide a valuable diagnostic in the management of prostate cancer.

FOXA1 and the androgen receptor

Most men diagnosed with prostate cancer have AR-positive, hormone-dependent disease that is initially responsive to androgen-ablation therapy. Unfortunately, the majority of these cases become resistant to this therapeutic approach due to acquisition of androgen-independence [105]. To better understand AR signalling within both androgen-dependent and -independent contexts, the co-operation of FOXA1 with AR transactivation has been investigated, albeit with some discrepancies in the resulting data [59,106–109]. Initial reports indicated that FOXA1 is necessary for AR activation of target gene expression [106]. These studies involved blockade of FOXA1 binding to the probasin and PSA promoters by mutating the FOXA1 consensus sequences within the promoters. Such disruption blocked androgen (R1881)-induced promoter activity [106], suggesting that FOXA1 is necessary for androgen activation of the expression of these genes (Figure 1). In contrast with these results, overexpression of FOXA1 in other studies inhibited DHT (dihydrotestosterone)-induced PSA promoter activity [59] and AR activation of transcription [109]. While the overexpression of FOXA1 in vitro could non-specifically squelch the transcriptional machinery, and explain these discrepant results, additional studies have shown that reducing FOXA1 expression does not alter DHT-induced expression of PSA/KLK3 (kallikrein-related peptidase 3) [107,108], TMPRSS2 (transmembrane protease serine 2) or PDE9A (phosphodiesterase 9A) [107], nor is FOXA1 necessary for androgen-induced cell cycle progression [107]. However, loss of FOXA1 increases expression of a subset of DHT-induced genes, including PSA/KLK3, and intensifies androgen-induced cell cycle progression [110]. Interestingly, Jia et al. [108] observed a reduction in DHT-induced expression of TMPRSS2 following FOXA1 silencing in an LNCaP variant (C4-2B) that is AR-positive, but androgen-independent, suggesting that FOXA1 facilitates AR-stimulated transcription of at least some genes. Supporting this finding, FOXA1 is also necessary for expression of UBE2C (ubiquitin-conjugating enzyme E2C) in an AR-positive, androgen-independent LNCaP variant (abl), but not in parental cells [111]. FOXA1 also binds to, and may regulate the homeobox transcription factor Hoxbl3 [112], which is a uniquely androgen-independent gene [113]. Overall, these results suggest that FOXA1 modulates AR target gene transcriptional activity, but through various mechanisms that depend on the level of responsiveness of the cell or gene to androgens. They also indicate that FOXA1 can modulate ligand-dependent and -independent actions of AR in prostate cancer cells.

Analyses of the binding of FOXA1 to AR target genes and its role in AR transactivation have suggested varied functions for FOXA1 that are gene-specific. FOXA1 binds adjacent to ARE (androgen response elements) in the probasin, PSA [106,114] and Sbp [114] promoters similarly in the presence and absence of DHT. This is similarly seen on a genome-wide level [110]. Others have also reported that FOXA1 binds to AR bound regions, but recruitment of FOXA1 to these sites is enhanced by DHT [107,108]. Importantly, while FOXA1 was necessary for androgen-induced transcriptional activity of the probasin and PSA promoters, blocking direct binding of FOXA1 to these promoters did not impact AR binding [106], suggesting that FOXA1 is not necessary for recruitment of AR. In contrast, AR binding to enhancer regions of several other genes [UBE2C, CDKl (cyclin-dependent kinase 1) and CDC20 (cell division cycle 20)] was reduced in response to FOXA1 silencing in parental (LNCaP) and androgen-independent (LNCaP-abl) prostate cancer cell lines [111]. Hence, the dependency of AR binding on preoccupancy of FOXA1 at adjacent sites appears to be highly dependent on gene context [111]. Supporting the requirement for FOXA1 binding as an essential modulator of AR activity, a germ line SNP (single nucleotide polymorphism) identified in prostate cancer patients within the risk locus 8q24 increases its affinity for FOXA1 binding and potentiates the ability of this chromosomal region to mediate androgen responsiveness of a heterologous promoter compared with the wild-type sequence [115]. Although these results suggest a tight link between FOXA1 binding and androgen responsiveness, the authors did not directly investigate whether FOXA1 is necessary for either AR binding or enhancer activity in its native chromosomal context. As a mechanism for how FOXA1 dictates AR binding at specific genes, studies in both prostate and breast cancer cells revealed that H3K4 dimethylation occurs proximal to FOXA1-binding sites, and can precede FOXA1 binding to DNA [116]. The overexpression of the histone demethylase KDM1 (lysine-specific demethylase 1) reduced AR binding, probably due to loss of bound FOXA1 [111]. The presence of FOXA1 in LNCaP cells is also associated with DNA hypomethylation [13], and an open chromatin state as measured by FAIRE (formaldehyde-assisted isolation of regulatory elements), where high FAIRE FOXA1-binding sites are more likely to bind AR [117]. Overall, it is quite evident that FOXA1 participates in AR target-gene recognition, but it seems likely that this regulation is dependent on additional co-factors, androgen dependence and/or the distance from the transcriptional start site (i.e. promoter or enhancer).

As another mode of regulation, the winged helix/forkhead domain of FOXA1 has been found to complex directly with the DNA-binding domain/hinge region of AR [106,110,118]. This complex permits AR binding to promoter DNA even in the absence of AR consensus elements and vice versa for FOXA1 binding [118]. Formation of a FOXA1/AR multimeric transcriptional complex that requires only one of the two protein’s consensus elements could explain why mutagenesis of FOXA1-binding sites does not alter AR binding [106], whereas decreasing FOXA1 expression does [111] as discussed above. The FOXA1/AR complex may also be necessary for AR nuclear localization. In this case, decreasing FOXA1 expression would greatly impede the access of AR to its target genes, whereas mutating the FOXA1 consensus element would only directly affect FOXA1 binding. Most recently, it has been demonstrated by two distinct groups that the FOXA1/AR interaction is required for maintaining a particular hormonal transcriptional programme [103,110]. Loss of this interaction, either through experimental knockdown of FOXA1, or via somatic mutation of the AR gene, results in an AR-induced transcriptional programme that is distinct from that induced by AR when co-expressed with FOXA1. This is due to the differential binding of AR to targets that are occupied by FOXA1 compared with those that lack a FOXA1-binding site. FOXA1 can also form complexes with USF2 (upstream stimulatory factor 2) in prostate cancer cells, but USF2 does not bind to AR, nor is it required for androgen-induced PSA/KLK3 expression [114]. Since USF2 binds to a similar DNA consensus element as FOXA1, it is also possible that a complex of FOXA1 and USF2 may modulate prostate-specific promoter activity, but this remains to be directly examined.

Additional roles of FOXA1 in prostate cancer

Although the majority of work investigating FOXA1 in prostate cancer has been aimed at understanding how it is involved in modulating AR transcriptional activity, FOXA1 has also been implicated in other signalling pathways. FOXA1/2 positively regulate transcription of the gene AGR2 (anterior gradient protein 2), whose expression increases aggressiveness of prostate cancer cells. This regulation can be abrogated by the presence of EBP1 (ERBB3-binding protein 1) [119]. AGR2 is an established metastasis inducer [120], and its activation by FOXA1 was first discovered in goblet cells [121]. Placement of FOXA1 in the EBP1/FOXA1/AGR2 pathway indicates that it may play an important modulatory role in metastatic progression. Interestingly, EBP1 has also been shown to negatively regulate AR-mediated transcription [122]. While not specifically tested, it is possible that EBP1 represses AR transactivation by inhibiting FOXA1 function in a similar manner as seen for the transcriptional regulation of AGR2. In contrast to its pro-metastatic positive regulation of AGR2, expression of FOXA1 is also down-regulated by SOX4 (sex determining region Y-box 4), a prostate cancer oncogene [123]. This contradiction exemplifies the broad spectrum of activity for FOXA1 in the cancerous prostate, and at least partially explains how loss of Foxa1 in the normal prostate could lead to both tumorigenic (i.e. increased SHH) and anti-tumorigenic (i.e. loss of AR function) phenotypes. Lastly, it is intriguing to speculate a potential pro-carcinogenic functional interaction between FOXA1, Nkx3.1 and SHH in the prostate given the aberrant regulation of these three factors in the Foxa1 null prostate. Future in vitro work should aim to delineate the connection between these pathways and their interdependent contributions to disease phenotypes.

FOXA1 in breast cancer

The molecular subtypes of breast cancer were identified via unsupervised hierarchical clustering of cDNA microarray data from a cohort of human breast tumours, in which each subtype was distinguished by the differential expression of numerous genes. In particular, FOXA1 mRNA is expressed in luminal subtype tumours along with several other discriminatory genes, including ESR1 and GATA3 [124,125]. While the six distinct subtypes [luminal A, luminal B, luminal C, HER2/ERBB2, basal and normal-like] are classified by genome-wide expression profiling [125], they are often identified using the surrogate markers: ER, PR (progesterone receptor) and the epidermal growth factor receptor 2 (HER2/ERBB2), a proto-oncogene amplified in ~ 20–30% of breast cancer patients. Luminal A subtype tumours are ER⁺ /PR⁺ /HER2⁻ and confer a favourable prognostic outcome that is at least partially due to the efficacy of anti-hormone therapies [126]. In contrast, basal subtype tumours tend to be triple negative (ER⁻ /PR⁻ /HER2⁻) and confer a poor prognosis due to the inherent aggressiveness of these tumours and the absence of subtype-specific targeted therapies [127].

FOXA1 expression in breast cancer

Since being identified as a novel luminal subtype gene, the clinical significance of FOXA1 protein expression in breast tumours has been investigated by multiple groups. Wolf et al. [128] were the first to investigate a breast tumour TMA (tissue microarray) for FOXA1 expression, revealing that FOXA1 is associated with low tumour grade, and, not surprisingly, ER expression [129]. A number of publications soon followed with various TMA compositions, and, when combined, provide expression data for FOXA1 on over 5000 human breast cancers [64,65,129–132]. FOXA1 significantly associates with ER expression in each study. Moreover, FOXA1 positively correlates with the luminal subtype as defined by ER and/or PR positivity, HER2 negativity [64,129], and luminal CK expression [64,65,129]. Negative correlations have been observed for FOXA1 and the basal subtype when defined as ER-negative, HER2-negative with CK5/6 and/or EGFR positivity [129, 132] or by basal CK expression [65,129]. Importantly, several groups have proposed utilizing FOXA1 as a prognostic tool to stratify patients within the ER-positive luminal subtype [64,129,130]. Results from these studies suggest that FOXA1 can predict survival within this population, and thus aid in therapeutic decision-making. However, TMA data generated by one group argues that FOXA1 is unable to further stratify prognosis within the subgroup of patients with ER-positive disease [65]. To directly test the independent prognostic and predictive ability of FOXA1, Ademuyiwa et al. [131] investigated the correlation between FOXA1 expression and Oncotype DX recurrence scores in patients with ER-positive node-negative disease. The Oncotype DX is a molecular diagnostic used to predict recurrence and response to chemotherapy. FOXA1 negatively correlated with recurrence score leading the authors to suggest FOXA1 immunostaining could function as a more cost-effective pathological marker than the Oncotype DX. Of note, FOXA1 is not a component of Oncotype DX. Lastly, increased copy number of chromosomal locus 14q13, where FOXA1 is located, is associated with the ER- and PR-positive breast cancer subtype [133], providing further evidence that FOXA1 may promote tumour differentiation.

Although FOXA1 statistically correlates with ER, it has been noted that some ER-negative tumours also have high FOXA1 expression [65,128]. In fact, a subset of ER- and PR-negative tumours is molecularly more similar to ER- and PR-positive than to triple negative tumours. Most importantly, FOXA1 was identified as one of the differentially expressed genes within this ‘ER-positive-like’ subgroup [134]. These tumours are more likely to possess apocrine features, express AR and have an androgen responsive molecular signature, and a breast cancer cell line (MDA-MB-453) that recapitulates this phenotype is growth stimulated by androgens and inhibited by anti-androgens [135]. Given the role of FOXA1 in modulating AR transactivation in prostate cancer, it is possible that the presence of FOXA1 in these ER-negative tumours is required for androgen-dependent responsiveness mediated by AR. Supporting this hypothesis, testosterone increases FOXA1 mRNA expression in another AR-positive breast cancer cell line (SUM190) [136]. Indeed, recent studies provide evidence that FOXA1 mediates AR binding and transcriptional regulation in the MDA-MB-453 cells [137,138]. FOXA1/AR signalling in these cells is similar to FOXA1/ER signalling in MCF7 cells, and is necessary for maintenance of the apocrine gene expression signature [137]. Further evidence revealed that AR transcriptionally regulates WNT7B, and subsequent induction of HER3 gene expression via an AR/FOXA1/β-catenin complex [138]. Enhanced HER3/HER2 signalling likely drives tumorigenesis of apocrine tumors. Hence, FOXA1/AR induces a set of genes that convey the phenotype, and perhaps induces the formation, of apocrine of breast cancers.

FOXA1 and the oestrogen receptor

Functioning as a chromatin remodeller, FOXA1 was discovered to co-operate with ER in activating the liver-specific vitellogenin promoter [139], and had been shown to regulate the well-established ER target, TFF1 (trefoil factor 1; pS2) [140]. In 2005, two groups independently extended these findings to breast cancer, demonstrating that FOXA1 was necessary for oestrogen-induced ER binding, and the up-regulation of several target genes, including TFF1 [141,142]. Carrolletal. [141] identified forkhead consensus motifs in close proximity to approximately half of the ER-binding sites spanning chromosomes 21 and 22 of a breast cancer cell line. Likewise, Laganière et al. [142] analysed ER-binding sites via a human promoter array, and found 12% of ER-bound promoters also contained forkhead motifs [143]. Several genome-wide analyses have confirmed these data [116,143–146], although one group has reported much less (<10%) commonality in the location of forkhead and ER-binding motifs in the genome [147]. That said, considerable co-operativity between FOXA1 and ER has been repeatedly observed in comparisons of full genome ER and FOXA1 ChlP-chip [116] and ChlP-seq [146] data, showing that over half of ER-binding sites are also occupied by FOXA1. This co-operation is observed for both ER up-regulated and down-regulated genes [116,146] (Figure 2A). For example, FOXA1 is necessary for ER repression of Reprimo (RPRM) transcription [148]. FOXA1 and ER can also have opposing roles in regulating gene expression as demonstrated for BASE (breast cancer and salivary gland expression); FOXA1 is required for BASE expression, whereas ER represses transcription of this gene [149].

Functionally, FOXA1 is necessary for cell growth [150] and specifically for oestrogen-mediated cell-cycle progression of breast cancer cells [142,151]. These results are at least partly explained by the requirement for FOXA1 to mediate ER stimulation of CCND1, the gene encoding cyclin D1, through a downstream enhancer [151]. The presence of FOXA1 correlates with bound RNA polymerase II, active chromatin and histone H4 acetylation [151] at ER-activated genes, where each is a hallmark of proliferating cells. Similar to its role in prostate cancer cells, FOXA1 binding is associated with an open chromatin state as measured by FAIRE, where loss of FOXA1 decreases the level of open chromatin on a global scale [117,146]. Specifically, FOXA1 binding at high FAIRE regions is accompanied by increased histone H3, lysine 9 (H3K9) acetylation, decreased H3K9 mono- and dimethylation, and increased H3K4 dimethylation [117]. The presence of FOXA1 is also associated with DNA hypomethylation [13]. Together, these results support a pioneering function for FOXA1 in modulating chromatin structure for subsequent ER binding and activity (Figure 2A), similar to that predicted for AR, although the specific timing of events is not well established in cancer cell lines. In addition to modulating ER activity, we recently reported that FOXA1 also directly binds to the ESR1 (oestrogen receptor 1) promoter and is required for expression of ER mRNA and protein in breast cancer cells [73]. Thus, unlike its relationship with AR in the prostate, FOXA1 controls both the expression and activity of ER in breast cancer (Figure 2B).

In response to oestrogen stimulation, FOXA1 is preferentially bound to intergenic enhancers and introns, mirroring that of ER [141]. It is recruited to DNA in the presence and absence of estradiol, but unlike with androgen stimulation, binding of FOXA1 generally decreases with oestrogen treatment [141,143]. A direct comparison of FOXA1 in ER and AR signalling using breast and prostate cells, respectively, revealed that while there are common binding sites, FOXA1 has a distinct binding profile depending on whether it is co-expressed with ER or AR [116]. This tissue specificity is dependent on differential recruitment of FOXA1 to ER compared with AR target genes due to lineage-specific H3K4 dimethylation [116]. Moreover, the presence of FOXA1 at the TFF1 gene enhancer is also dependent on binding of the histone variant H2A.Z by p400 [152], providing further evidence that tissue-specific FOXA1 activity is dictated by histone modifications. Another factor has recently been identified that also delineates FOXA1-specific recruitment and subsequent ER transcriptional regulation. CTCF (CCCTC-binding factor) is an insulator binding protein [153,154] whose loss was shown to impede FOXA1 binding to known target genes [e.g. TFF1 and PGR (progesterone receptor)], while loss of FOXA1 had no impact on CTCF binding [154]. In contrast, Hurtado et al. [146] found that loss of CTCF increases FOXA1-binding capacity. Despite the discrepancy between these two reports, it is apparent that CTCF can modulate FOXA1 binding. The context dependency of this interaction must be further explored to unravel the molecular mechanisms by which CTCF enhances or impedes binding of FOXA1 to DNA.

The interplay between FOXA1 and ER may extend beyond the ability of FOXA1 to control ER expression and activity. Using a different experimental approach than Carroll et al. [141] who identified FOXA1-binding sites in ER target genes through motif analysis, Laganière et al. [142] first identified FOXA1 as an oestrogen-induced ER target gene and then subsequently observed FOXA1-binding sites within ER responsive genes. Providing support for FOXA1 being an ER target, a more recent analysis of the ER chromatin interaction network revealed that ER-binding sites flank the FOXA1 gene [155]. Although these results suggest a positive regulatory loop between ER and FOXA1 (Figure 2B), others have described FOXA1 to be down-regulated with oestrogen treatment [128,156–158]. Moreover, mice-lacking Esr1 maintain expression of FOXA1 in the mammary epithelium [73]. These disparate observations are further complicated by the varying roles of FOXA1 in mediating ER signalling in tissues other than the breast [146,159,160]. Osteosarcoma cells that have been engineered to express the ER (U2OS-ER) have undetectable levels of FOXA1 protein, suggesting that the ER is insufficient to induce expression of FOXA1. Given the absence of FOXA1 in U2OS-ER cells, it is not surprising that forkhead consensus motifs do not correlate with the ER cistrome, nor is FOXA1 required for ER-regulated transcription in this context [160]. These results suggest that the role of FOXA1 in modulating ER activity is context dependent, and are consistent with the observation that FOXA1 is necessary for ER expression only in the epithelium and not in the stroma of the murine mammary gland [73]. In support of an indispensable role for FOXA1 in ER activity in breast epithelial-specific gene regulation, exogenous expression of FOXA1 in U2OS-ER cells induces ER binding to otherwise breast-specific ER targets [e.g. TFF1 and XBP1 (X-box binding protein 1)] that is accompanied by increased expression of these genes [146]. Furthermore, FOXA1 overexpressing U2OS-ER cells also become sensitive to the growth inhibitory effects of tamoxifen, a SERM (selective ER modulator). In contrast, FOXA1 is endogenously expressed in, and is necessary for, resveratrol-induced ER transactivation of BMP2 (bone morphogenetic protein-2) gene expression in the osteoblast cell line MC3T3-E1 [159]. It is possible that the distinct observations seen with FOXA1 and ER in these studies can be attributed to the differences between the U2OS cells that were derived from a human osteosarcoma compared with the non-transformed MC3T3-E1 cells that were isolated from newborn mouse calvaria [161]. FOXA1 has been proposed as a putative breast cancer therapeutic target based on the concept that ER dependency on FOXA1 will translate to an increase in the efficacy of tamoxifen in ER-dependent breast cancer [162,163]. Functioning as an SERM, tamoxifen has antagonistic properties in the breast, but is agonistic in the bone, where it maintains oestrogen signalling and is protective against osteoporosis. The differing roles of FOXA1 in cancerous and normal bone tissue should be further defined before considering targeting FOXA1 in breast cancer patients. If FOXA1 is expressed in normal bone, as seen in the MC3T3-E1 cells, but loss of FOXA1 correlates with tumorigenic progression, as shown by the lack of FOXA1 in transformed U2OS cells, then the systemic reduction of FOXA1 expression could inadvertently induce bone tumours.

Additional roles of FOXA1 in breast cancer

In addition to the well-described ability of FOXA1 to control ER activity, there have been other facets of FOXA1 described in breast cancer. One of these is a possible role in HER2 signalling [136,150]. Treatment of the breast cancer cell line SUM190 with heregulin, an indirect activator of HER2, induces FOXA1 expression [136]. This suggests that FOXA1 may be downstream of HER2 signalling and mediate some of its actions. Indeed, FOXA1 silencing potentiates the cellular toxicity of herceptin, a humanized monoclonal antibody that inactivates HER2 in HER2-amplified breast cancer cell lines [150]. While the specific mechanisms remain unknown, these results support a role for FOXA1 in HER2-induced cell survival. However, it is important to note that the presence or absence of FOXA1 does not correlate with clinical outcome of patients with HER2-positive disease [164], indicating that FOXA1 may not play a critical role in HER2-induced signalling in vivo.

In addition to many other genes, FOXA1 binds to, and activates transcription of, the gene encoding HSP72 (heat-shock protein 72; HSPA1A), the stress-inducible cytosolic form of HSP70 [165]. HSP70 independently associates with poorer outcome in breast cancer patients with node-negative disease [166]. Thus it is possible that FOXA1 and HSP72 co-operatively dictate the aggressive behaviours of these cancers, but this has not been tested. As mentioned earlier, FOXA1 also participates in ER-dependent transactivation of CCND1 (cyclin D1) [151]. Intriguingly, FOXA1, ER and cyclin D1 are co-ordinately up-regulated in tumours arising in mice that are bi-transgenic for mammary gland-specific expression of ILK (integrin-linked kinase) and Wnt1 (MMTV-Wnt1/ILK) [167]. Forced co-expression of ILK and Wnt1 leads to increased epithelial proliferation, a greater percentage of luminal progenitor cells and increased tumour formation as opposed to the expression of either alone. The authors of this work proposed that the convergence of Wnt1 and ILK on β-catenin activation may result in up-regulation of FOXA1, which is downstream of SOX17/β-catenin during endodermal development in Xenopus [168]. Notably we have shown that FOXA1 is expressed in the mammary epithelial luminal progenitor population [73]. Whether FOXA1 is necessary for expansion of luminal progenitors and the formation of mammary tumours, or is simply expressed as a bystander in this bi-transgenic or other mouse models of breast cancer, remains to be seen.

Given its correlation with the less aggressive luminal subtype of breast cancer, it is not surprising that FOXA1 activates the transcription of genes [e.g. CDH1 (t-cadherin) and CDKN1B (cyclin-dependent kinase inhibitor 1B)] implicated in decreased breast cancer tumorigenicity independently of ER [169,170]. Exogenous overexpression of FOXA1 in a triple negative, basal breast cancer cell line (MDA-MB-231) induces expression of the cell-cell adhesion molecule, E-cadherin [171]. In this context, FOXA1 directly stimulates transcription of the E-cadherin gene (CDH1), and the associated induction of E-cadherin expression decreases the migratory capacity of these cells, suggesting that FOXA1 may promote their differentiation. Activation of CDH1 occurs in the absence of ER, supporting the notion that FOXA1 has ER-independent roles in dictating a more differentiated luminal cell phenotype. Another example of FOXA1 functioning in a tumour suppressive role is its regulation of the cyclin-dependent kinase inhibitor p27^Kip1 [172]. FOXA1 binds to the promoter of p27^Kip1 (CDKNlB) in breast cancer cells and synergizes with BRCA1 (breast-cancer susceptibility gene 1) to induce its transactivation in a colon cancer cell line. In addition, breast cancer cells overexpressing FOXA1 have increased p27^Kip1 promoter activity and decreased cell number [128]. When combined, these studies paint a complicated picture of FOXA1 as a participant in multiple signalling pathways in breast cancer, which are both oncogenic and tumour suppressive. It is likely that the specific molecular functions of FOXA1 are dictated by many components, including, but not limited to, the epithelial origin of the tumour initiating cell and the co-expression of ER, HER2 and other regulatory factors that have not yet been identified.