|Home | About | Journals | Submit | Contact Us | Français|
Endocrine therapies for breast cancer that target the estrogen receptor (ER) are ineffective in the 25-30% of cases that are ER negative (ER−). Androgen receptor (AR) is expressed in 60-70% of breast tumors, independent of ER status. How androgens and AR regulate breast cancer growth remains largely unknown. We find that AR is enriched in ER−breast tumors that over-express HER2. Through analysis of the AR cistrome and androgen-regulated gene expression in ER−/HER2+ breast cancers we find that AR mediates ligand-dependent activation of Wnt and HER2 signaling pathways through direct transcriptional induction of WNT7B and HER3. Specific targeting of AR, Wnt or HER2 signaling impairs androgen-stimulated tumor cell growth suggesting potential therapeutic approaches for ER−/HER2+ breast cancers.
Sex steroid hormones are critical for the development and progression of cancers of reproductive organs including the uterus and breast in women and the prostate in men. While estrogen is considered to be a female-selective sex steroid and androgen a male one, both are present and play important developmental functions in both sexes. While estrogen is widely recognized for its role in breast cancer, little is known concerning a potential role for androgen in this disease. Steroid hormone receptors are crucial components of steroid hormone signaling that act as transcription factors to regulate gene expression, however their function contributes to hormonal carcinogenesis. Studies of estrogen and estrogen receptor (ER) have led to significant progress in the development of endocrine therapies targeting estrogen production or ER for both breast cancer treatment and prevention. However, there are currently no effective endocrine therapies for the 25-30% of breast cancers that are ER-negative (ER−).
In addition to defining breast cancers based on their expression of ER, the recognition of the role played by amplification of the HER2 oncogene has had a major impact on the disease. Therapies targeting HER2 such as trastuzumab are becoming increasingly important in the treatment of HER2+ tumors of all stages. Nonetheless, despite endocrine therapies for ER+ tumors and HER2-targeted therapies for HER2+ tumors, significant numbers of breast tumors fail to respond. Recent studies have found that the androgen receptor (AR) is expressed in 60-70% of breast cancers regardless of ER status (Agoff et al., 2003; Kuenen-Boumeester et al., 1996; Niemeier et al., 2009). It has been postulated that AR functions as an anti-proliferative effecter in ER+ breast cancer by antagonizing ER (Peters et al., 2009), whereas it facilitates tumor cell growth in an androgen-dependent manner in an ER−/AR+ breast cancer cell line model (Doane et al., 2006). The ER−/AR+ subclass was previously identified as a group of breast tumors with histological apocrine features and termed the molecular apocrine subtype (Farmer et al., 2005). Microarray analyses reveal intact and active AR signaling in ER−/AR+ breast tumors (Doane et al., 2006). However, the mechanism by which AR contributes to breast cancer development and progression is unknown. To comprehensively decipher the role of AR in breast cancer, we used an integrative analysis of the AR cistrome together with androgen-regulated gene expression profiles to define the important AR regulated pathways that are involved in stimulating the growth of ER−/HER2+ breast cancers.
Gene expression microarray studies provide an important tool for tumor classification, prognosis and as a guide to therapy. We began by analyzing published microarray datasets (Hess et al., 2006; Ivshina et al., 2006; Wang et al., 2005a) of breast tumors to gain an overall view on AR gene expression across different molecular subtypes in breast cancer. Using a 50-gene signature (PAM50)-based subtype classification (Hu et al., 2006; Parker et al., 2009; Sorlie et al., 2001; Sorlie et al., 2003), we assigned the tumor samples into five subclasses (Luminal A, Luminal B, normal-like, basal-like and HER2-enriched) and the gene expression level of AR together with the important biomarkers, ER, FOXA1 and HER2, are summarized in a heatmap (Figure 1A). Interestingly, we observed a high level of AR expression in HER2-enriched tumors. Importantly, within ER-negative subgroup that contains both basal-like and ER−/HER2+ tumors, high AR expression is correlated with HER2 amplification and overexpression (rPearson>0.69, p<6.6×10-11). Consistent with this observation, recent immunohistochemical studies also suggested a significant correlation of AR with HER2 overexpression in ER-negative breast tumors (Agoff et al., 2003; Niemeier et al., 2009; Park et al., 2009). Notably AR is not highly expressed in the basal-like subtype.
To understand the genomic action of androgens and AR in ER−/HER2+ breast cancer, we investigated the gene expression profiles of MDA-MB-453 human breast cancer cells treated with 5α-dihydrotestosterone (DHT). This ER−/HER2+ breast cancer cell line belongs to HER2-enriched subtype based on the PAM50 gene signature. With abundant expression of AR (Doane et al., 2006), MDA-MB-453 cells showed increased growth responsiveness to androgens, such as DHT (Figure 1B) and the synthetic androgen R1881 (Doane et al., 2006). Addition of bicalutamide, a non-steroidal AR antagonist, abolished the DHT-stimulated cell growth (Figure 1B) and also induced apoptosis (Figure 1C) in MDA-MB-453 breast cancer cells. Gene expression microarray analyses identified a total of 1341 genes that are differentially regulated by DHT (q-value ≤0.05 and fold change ≥1.5).
To identify the direct targets of AR transcriptional regulation, we performed AR chromatin immunoprecipitation followed by next-generation sequencing (ChIP-seq) to define the AR cistrome in MDA-MB-453 breast cancer cells. We identified a total of 2406 high-confidence AR binding sites following 4h of androgen stimulation (p≤1×10-5). These AR binding sites were found predominantly at distal intergenic and intronic regions (Figure S1A) and possessed highly conserved genomic sequences (Figure S1B). Interestingly, while this AR cistrome overlaps to a small degree with the AR cistrome in LNCaP prostate cancer cells or the ER cistrome in MCF7 ER+ breast cancer cells, the majority of the binding sites are cell-specific (Figure 1D). Thus, our study reveals a unique hormone-regulated AR cistrome in ER−/HER2+ breast cancer cells. Importantly, motif enrichment analysis of the AR cistrome revealed the presence of not only AR and other nuclear receptor binding motifs, but also forkhead factor binding motifs including FOXA1 (Figure 1E). Among all forkhead family transcription factors, FOXA1 is highly expressed in HER2+ breast tumors and associated with AR expression (rPearson>0.76, p<2×10-10), as compared to basal-like tumors of the ER-negative subgroup (Figure 1A and and2A).2A). It is worth noting that high levels of FOXA1 have been observed in AR+ prostate cancers (Mirosevich et al., 2006) and ER+ breast cancers (Lacroix and Leclercq, 2004). In addition to its expression in breast cancer, FOXA1 has been shown to maintain ER expression during normal mammary ductal morphogenesis (Bernardo et al., 2010) and to act as a cell-type specific pioneer factor that facilitates recruitment of ER and AR to their cis-regulatory elements across the genome (Lupien et al., 2008). To gain further insight into the potential association between FOXA1 and AR in ER−/HER2+ breast cancer, we determined the FOXA1 cistrome in MDA-MB-453 cells and identified 7591 high-confidence FOXA1 binding sites (p≤1×10-5). We observed a significant overlap (~37%) between the AR and FOXA1 cistromes in MDA-MB-453 breast cancer cells (Figure 2B). Analysis of ChIP-seq signals further verified a strong correlation and mutual enrichment between AR and FOXA1 binding (Figure 2C). Similar to the FOXA1 cistromes previously identified in MCF7 and LNCaP cells, the FOXA1 cistrome of MDA-MB-453 cells showed predominant enrichment in distal intergenic and intronic regulatory regions in genome (Figure S1C). These FOXA1 binding sites were highly conserved (Figure S1D) and significantly enriched with the FOXA1 binding motif (Figure 2C, inserted panel). Interestingly, we observed a significant overlap of FOXA1 cistromes between MDA-MB-453 and MCF7 breast cancer cells (Figure 2D), however the overlapped AR/FOXA1 sites in MDA-MB-453 are largely distinct from the ER/FOXA1 sites in MCF7.
To determine the functional significance of AR and FOXA1 chromatin co-localization, we examined the correlation of transcription factor binding to DHT-regulated gene expression in MDA-MB-453 breast cancer cells. The sequences of overlapped AR/FOXA1 sites are highly conserved (Figure S2A) and overall either AR or FOXA1 binding sites are significantly enriched within 50 kb of transcriptional start sites (TSS) of the DHT-upregulated genes (Figure S2B-D). This was especially true for sites where AR and FOXA1 bound near to one another (Figure 2E). In contrast, the genes downregulated by DHT have only FOXA1 and not AR binding sites nearby (Figure S2C and D), suggesting a role of FOXA1 in maintaining basal expression of these genes and predominantly indirect mechanisms of androgen-mediated repression in this cell type (Lupien et al., 2008). These data indicate an important collaboration between AR and FOXA1 in transcriptional activation of AR target genes and androgen signaling in ER−/HER2+ breast cancer cells.
In keeping with the finding that AR showed cell-type-specific chromatin occupancy in breast cancer and prostate cancer cells in response to androgen stimulation, it also regulated distinct transcriptional programs. By defining the DHT-upregulated transcriptional program in MDA-MB-453 breast cancer cells, we found this gene signature was significantly correlated with the published overexpression gene signatures of HER2+ breast tumors and as well as ER+ breast tumors (Figure 3A) using Oncomine Concepts Map analysis (Rhodes et al., 2007). In order to determine the identity of androgen-modulated biological pathways in MDA-MB-453 cells we performed gene set enrichment analysis (GSEA) (Subramanian et al., 2005) of the androgen-regulated genes. Among the highest scoring categories for the DHT-stimulated genes (Figure S3), we observed significant enrichment of genes in the Wnt signaling pathway (Figure 3B). Screening of the genes in the Wnt pathway overexpressed in MDA-MB-453 cells revealed the significant upregulation of the canonical Wnt ligand WNT7B (Figure 3C). Analysis of the AR cistrome identified an AR binding site within the WNT7B gene locus (+1.5 kb of TSS) (Figure 4A). We validated using ChIP-qPCR that the occupancy of AR at this site was enriched only after DHT stimulation in MDA-MB-453 cells (Figure 4B). We also confirmed by RT-PCR that DHT treatment substantially elevated the mRNA level of WNT7B (Figure 4C), and silencing of AR effectively impeded this induction (Figure S4A). The androgen-dependent upregulation of WNT7B by AR was further validated in two other ER−breast cancer cell lines HCC202 and SUM185PE, both of which express AR and showed increased growth responsiveness to androgen stimulation (Figure S4). These data suggest that AR directly activates WNT7B transcription in a ligand-dependent manner.
Wnt ligands are involved in the development of normal mammary gland and increased expression of WNT1 is oncogenic in breast cancer models in mice (Turashvili et al., 2006). WNT7B acts in the canonical Wnt pathway and is expressed during the stages of ductal formation and involution in mammary gland development (Gavin and McMahon, 1992). WNT7B has the ability to maintain mammary epithelial cells in an uncommitted state (Weber-Hall et al., 1994) and its elevated expression has been found in 10% of all breast carcinomas (Huguet et al., 1994). We found by analyzing published breast tumor microarray datasets (Hess et al., 2006; Ivshina et al., 2006) that WNT7B expression is highly correlated with AR status (rPearson>0.44, p<9×10-3). As a canonical Wnt ligand, WNT7B activates TCF-regulated gene transcription in a cell-type-specific manner through the stabilization of β-catenin (Wang et al., 2005b). Interestingly, we found that DHT stimulation led to nuclear AR, increased WNT7B protein levels and the activation of β-catenin in MDA-MB-453 breast cancer cells, as demonstrated by the increased protein levels of β-catenin and its active form in the nuclear fraction (Figure 4D). Consistent with canonical Wnt signaling β-catenin mRNA levels were not affected (Figure S4B). To confirm whether the nuclear translocation of β-catenin is dependent on AR-mediated WNT7B upregulation, we transduced the MDA-MB-453 cells with lentiviral shRNA targeting either WNT7B or AR and the subcellular localization of β-catenin was determined by confocal microscopy. As shown in Figure 4E, DHT treatment induced the nuclear accumulation of both β-catenin and AR. In contrast, silencing of WNT7B or AR significantly abrogated β-catenin nuclear localization (Figure 4E), indicating that AR and WNT7B mediate Wnt/β-catenin activation in response to DHT.
In the canonical Wnt pathway, nuclear β-catenin is a pivotal regulator of TCF-regulated gene transcription. β-catenin has also been shown to act as a co-regulator of AR in prostate cancer cells to enhance the transcriptional activity of AR (Yang et al., 2002). We detected the physical interaction between AR and β-catenin in the nuclear extracts of MDA-MB-453 breast cancer cells by co-immunoprecipitation (Figure 5A). However, neither DHT nor recombinant WNT7B protein was able to activate the TCF/LEF TOP-Flash reporter; instead they suppressed the luciferase activity and the co-treatment displayed a synergistic inhibitory effect (Figure S5A). This inhibition is likely due to a previously described mechanism in prostate cancer in which AR competes with TCFs for limiting amounts of nuclear β-catenin, thus suppressing β-catenin/TCF transcriptional activity (Chesire and Isaacs, 2002; Mulholland et al., 2003). Moreover, DHT treatment led to decreased expression of the four TCF/LEF family genes (Figure S5B) in MDA-MB-453 breast cancer cells, suggesting an additional mechanism by which androgen and AR inhibit TCF-regulated transcription.
Physical interaction between AR and β-catenin represents a secondary response to androgen stimulation. Their interaction has been previously implicated in prostate cancer progression (Mulholland et al., 2005; Terry et al., 2006). However, the genes co-regulated by AR and β-catenin are not known. To investigate the regulatory role of AR in collaboration with β-catenin, we defined the AR cistrome at a late time point (16h) following prolonged DHT stimulation. AR ChIP-seq revealed a total of 3431 high-confidence sites in MDA-MB-453 cells (p≤1×10-5). Similar to the 4h-AR cistrome, these 16h-AR sites are mainly located in distal intergenic and intronic regions and are highly enriched near the DHT-upregulated genes (Figure 5B). Analysis of the 4h-AR, 16h-AR and FOXA1 cistromes revealed significant numbers of sites that were differentially occupied by AR at 4h and 16h of DHT treatment (Figure 5C). There was significant overlap between AR at both time points with FOXA1. The differential androgen-stimulated AR binding patterns at early and late time points suggested a hierarchical regulatory network involving AR, FOXA1 and other AR cofactors. To investigate the potential targets regulated by AR and β-catenin, we first examined the overlapped sites between 16h-AR and FOXA1 (744 sites as shown in Figure 5C) and the nearby genes. Interestingly, gene ontology (GO) analysis revealed that genes with uniquely overlapped 16h-AR/FOXA1 binding sites within 50 kb of their TSS were significantly associated with several important biological processes including most notably “regulation of cell proliferation” (Figure 5D). Through screening of the genes involved in these GO categories, we found HER3 as a gene involved in several of the enriched biological processes (Figure 5D, highlighted in red). Inspection of the FOXA1 and 16h-AR cistromes identified a site bound by both close to the TSS of the HER3 gene (Figure 6A). We focused on HER3 as it encodes an important co-receptor of HER2 in HER2+ breast cancers. We demonstrated that FOXA1 binding at the HER3 gene was dramatically increased following DHT treatment, and concomitantly, a substantially enhanced occupancy of both AR and β-catenin was also verified at this same region in MDA-MB-453 (Figure 6B) and as well SUM185PE breast cancer cells (Figure S5E).
Sequence analysis of this overlapped AR/FOXA1 region revealed two FOXA1 binding motifs and one ARE half-site motif nearby (Figure S5C). Additionally, three TCF4/TCF7L2 motifs also occur within this 950-bp region (Figure S5C, blue boxes). Given that TCF factors recruit β-catenin to chromatin as part of canonical Wnt signaling and DHT downregulates TCF expression in MDA-MB-453 breast cancer cells, we speculated that TCF4 might initially reside at this regulatory region of the HER3 gene before AR activation, and may recruit β-catenin together with AR upon DHT stimulation. Consistent with this hypothesis, TCF4 ChIP showed that TCF4 was able to occupy the same region in the absence of DHT, whereas the recruitment of TCF4 to HER3 gradually declined following DHT stimulation (Figure S5D), likely due to the DHT-mediated downregulation of TCFs in MDA-MB-453 cells as described above. Androgen stimulation led to increased expression of HER3, which was abrogated by silencing of AR or β-catenin (Figure 6C and D). As expected, AR silencing caused a decrease in β-catenin at protein levels; interestingly, β-catenin knockdown also diminished AR expression in MDA-MB-453 cells (Figure 6D). This observation implicated a feedback mechanism between Wnt/β-catenin and androgen/AR pathways. In prostate cancer cells, it has been suggested that Wnt signaling modulates AR transcriptional activity and protein stability by multiple pathways (reviewed in (Terry et al., 2006)), including direct regulation of AR transcription and protein stability by β-catenin/TCF (Yang et al., 2006). These findings indicate a regulatory mechanism by which androgen induces HER3 expression in ER−/HER2+ breast cancer cells through collaboration between AR and β-catenin. HER3 together with EGFR/HER1, HER2 and HER4 constitutes the HER family of tyrosine kinases. It is through dimerization and transphosphorylation that the HER members trigger signaling cascades. Importantly, HER2 activation requires heterodimerization with another HER receptor due to its lack of ligand binding activity (Baselga and Swain, 2009). The significance of HER3 in HER2+ breast tumors is well recognized for its ability to form heterodimers with HER2 and modulate the PI3K/AKT pathway (reviewed in (Baselga and Swain, 2009)). HER3 is thought to be an important determinant of the oncogenic activity of HER2 in breast cancer (Hsieh and Moasser, 2007; Lee-Hoeflich et al., 2008). It has been shown that increased HER3 expression substantially augments HER2 signaling, and contributes to the incomplete inhibition of HER2 activity by tyrosine kinase inhibitors (Sergina et al., 2007). Similarly, we observed that the phosphorylation of both HER3 and HER2 was elevated as an indication of the active HER2 signaling pathway during DHT treatment (Figure 6E). As a downstream event of HER2/HER3 signaling, the PI3K/AKT pathway was also significantly activated in response to androgen, demonstrated by the increased phosphorylation of AKT (Figure 6E) in MDA-MB-453 cells. Activation of this signaling cascade was also observed in DHT-stimulated SUM185PE breast cancer cells (Figure S5F). These observations indicate that the elevated expression of HER3 in response to androgen leads to enforced HER2/HER3 signaling in ER−/HER2+ breast cancer cells.
HER3-mediated PI3K/AKT activation is crucial for maintaining cell proliferation and survival. Thus, we hypothesized that AR/β-catenin-mediated upregulation of HER3 and activation of HER2/HER3 signaling might account for increased cell growth following androgen stimulation. Consistent with this hypothesis, silencing of AR, WNT7B or β-catenin effectively impeded the DHT-induced cell growth (Figure 6F), confirming the important function of AR and WNT7B/β-catenin in ER−/HER2+ breast cancer. To gain further insights into the effects of antagonizing AR and the AR-regulated signaling cascade in ER−/HER2+ breast cancer, we tested the pharmacological inhibitors targeting AR (bicalutamide), Wnt/β-catenin (IWP2 and IWR1) (Chen et al., 2009), or HER2 (tyrosine kinase inhibitors lapatinib and GW583340) in cell growth assays. Inhibitors targeting AR and Wnt/β-catenin pathways were able to abolish the DHT-induced recruitment of AR and β-catenin to the HER3 gene (Figure S5H), which blocked induction of HER3 at mRNA level (Figure S5I). We further demonstrated that these inhibitors targeting AR, Wnt or HER2 signaling greatly diminished DHT-induced HER3 activation, shown as a decrease of both total protein level and phosphorylation of HER3 in the presence of DHT (Figure 6G). In accordance with attenuated HER3 activity, DHT-induced cell growth of MDA-MB-453 was abrogated by addition of these inhibitors (Figure 6H). Similar effects were observed in SUM185PE breast cancer cells (Figure S5G). Notably, the tyrosine kinase inhibitors, lapatinib and GW583340, were able to inhibit tumor cell growth significantly only with DHT co-treatment. This suggests that androgen stimulation increases the dependence on HER2/HER3 signaling for growth in ER−/HER2+ breast tumors that express AR.
Using breast cancer cell lines, we identified HER3 as a downstream target of AR. To determine whether HER3 is also highly expressed in clinical ER−/HER2+ breast tumors, we first analyzed the published gene expression microarrays of breast tumor samples. Analysis of two independent data sets revealed significantly higher levels of HER3 expression in ER−/HER2+ breast tumors compared to basal-like ones (Figure 7A) and importantly HER3 expression is highly correlated with AR expression across all ER-negative breast tumors (rPearson>0.41, p<2.8×10-4). To further confirm this correlation, we performed immunohistochemical staining of AR and HER3 in breast carcinoma tissue microarrays and quantified the levels of AR and HER3 proteins in human breast cancer tissues by automated quantitative analysis (AQUA) (Camp et al., 2002). Figure 7B shows representative images from immunofluorescent labeling of AR and HER3 in four subtypes of ER−/PR− breast tumors classified by the expression of HER2 and AR. Each panel shows the AQUA score indicating the protein level of AR or HER3. AR expression is predominantly nuclear and HER3 expression was both membranous and cytoplasmic, notably excluded from the nuclear compartment (non-nuclear). The quantitative analysis demonstrated significantly higher HER3 expression in HER2+/AR+ breast tumors as compared to the other ER−/PR−subtypes (Figure 7C). Thus these analyses in breast cancer patient samples demonstrate that HER3 expression is correlated with AR and HER2 status in ER-negative breast cancers and support our finding that HER3 is a direct AR target.
To determine the anti-tumor effect of targeting AR in vivo, we established an MDA-MB-453 xenograft model. The MDA-MB-453 cells were implanted in the mammary gland of NOD/SCID mice, and one group of the mice received the concomitant implantation of a DHT slow-release pellet. DHT dramatically promoted the growth of MDA-MB-453 tumors in vivo (p=0.0003) (Figure 8A). After tumors reached a volume of 400mm3, daily oral administration of bicalutamide at 10 mg/kg was delivered to half of the mice carrying DHT pellets (n=8). As expected bicalutamide treatment resulted in a significant inhibition of DHT-stimulated growth of MDA-MB-453 xenograft tumors (p=0.0007) and tumors regressed to control levels following three weeks of treatment (Figure 8A).
Immunoblot analysis of MDA-MB-453 tumor extracts confirmed that AR levels were elevated by DHT treatment in vivo. Consistent with our cell culture findings, the AR-regulated Wnt/β-catenin and HER2/HER3 signaling pathways were substantially induced in the tumors from the mice carrying DHT pellets, whereas bicalutamide treatment abrogated the DHT-induced activation of these pathways (Figure 8B). Hemotoxylin and eosin staining of paraffin-embedded tumor sections revealed that while MDA-MB-453 xenografts exhibited well-organized sheets of round and cohesive undifferentiated tumor cells, the tumors treated with bicalutamide showed more discohesive cells within a disordered structure (Figure 8C). Tumor sections from each treatment group were further analyzed by immunohistochemical staining of AR and its targets to determine the effects of DHT and bicalutamide on their expression in vivo. Importantly, the DHT-treated xenograft tumors showed significant nuclear AR and membranous HER2 staining, and elevated HER3 and p-AKT levels as compared to the tumors from the mice without DHT pellets. Significantly, bicalutamide treatment led to substantial reduction in expression of HER3 and p-AKT, and enhanced cell death as detected by TUNEL assay (Figure 8C). Taken together, these results demonstrate that inhibiting AR with bicalutamide blocked androgen-stimulated oncogenic HER2/HER3 signaling and inhibited the growth of ER−/HER2+/AR+ breast tumors in vivo.
Endocrine therapies that target estrogen and ER signaling pathways play a critical role in the treatment of the majority of breast cancer patients. However, over a quarter of breast tumors fail to express ER and are thus resistant to these therapies. Our findings suggest that there are a significant number of ER−/HER2+ breast tumors that express AR and are growth stimulated by androgens. We have shown that androgens and AR stimulate oncogenic Wnt and HER2 signaling pathways by transcriptional upregulation of WNT7B and HER3 in ER−/HER2+ breast cancer (as summarized in Figure 8D), by which androgen stimulates tumor cell growth. This regulatory network indicates an intrinsic link between AR and growth factor pathways in ER-negative breast cancer.
Hormonal regulation of Wnt expression has been suggested in mammary gland development, such as WNT4 by progesterone (Brisken et al., 2000), and WNT5B and WNT7B by human chorionic gonadotropin (hCG) (Kuorelahti et al., 2007), whereas the direct effect of androgen and AR on WNT expression was previously unknown in the normal mammary gland or breast tumors. Overexpression of Wnts causes aberrant activation of the Wnt signaling pathway, which is a major driving force in a broad spectrum of human cancers. In this study, we have shown that AR-stimulated elevation of WNT7B leads to activated nuclear translocation of β-catenin. This in turn leads to β-catenin interaction with AR to induce HER3 gene expression. These findings highlight a crucial androgen-dependent signaling cascade that is regulated by AR in ER−/HER2+ breast cancer. The nuclear localization of β-catenin is an important predictor of poor prognosis and outcome in colorectal cancer. However, the subcellular location of β-catenin has not been well investigated in breast cancer. We find that the collaboration between β-catenin and AR together with FOXA1 plays an important role in ER−/HER2+ breast tumors through binding to regulatory regions of the HER3 gene.
HER3 is receiving increased attention as a therapeutic target in HER2+ breast cancers. Clinical studies of EGFR and HER2 antagonists suggest that HER2+ breast cancer is primarily driven by HER2/HER3 heterodimers rather than HER2/EGFR heterodimers (reviewed in (Baselga and Swain, 2009)). Interfering with HER2/HER3 heterodimerization has been suggested as an option to inhibit HER3 activation with the development of agents such as pertuzumab. Our work indicates that targeting AR with antagonists such as bicalutamide, may be another way to target HER3. Clinical trials of the anti-androgen bicalutamide in ER−/AR+ metastatic breast cancer are ongoing (NCT00468715). Our study also suggests that pharmacological inhibitors targeting AR-regulated pathways will only be effective in the presence of active androgen signaling and may be limited to ER−/HER2+ tumors. Expression of androgen/AR-regulated genes in ER−/HER2+ breast tumors could be predictors of a therapeutic response to AR target therapies. It has been shown that the signaling pathways including Wnt, HER2 and PI3K/AKT are able to actively enforce AR activity, constituting positive feed-forward circuits (Mellinghoff et al., 2004; Mulholland et al., 2006; Terry et al., 2006). This may potentially lead to androgen-independent AR activation in breast tumors, as has been demonstrated in castration-resistant prostate cancer (Wang et al., 2009). Collectively, this work provides not only novel insights into androgen-dependent AR function in breast cancer, but also reveals the mechanistic basis for targeting AR as a therapeutic opportunity for patients with invasive ER−/HER2+ breast tumors.
Hormone-depleted MDA-MB-453 cells were treated with vehicle or 10 nM DHT for 4, 16 or 48 hours, and the total RNA was isolated and hybridized to Affymetrix human U133 plus 2.0 expression array. All Gene expression microarray data were normalized and summarized using RMA (Irizarry et al., 2003). The differentially expressed genes between DHT and vehicle conditions were identified using SAM algorithm (Tusher et al., 2001) that is implemented in the MeV program (Saeed et al., 2006) (q≤0.05, fold change≥1.5). The raw data of breast tumor microarray were obtained from previous publication (Hess et al., 2006; Ivshina et al., 2006; Wang et al., 2005a) and were normalized using RMA (Irizarry et al., 2003). The intrinsic subtype of individual breast tumor sample in clinical datasets was assigned using a 50-gene signature (PAM50) based on a supervised approach (Parker et al., 2009). The co-expression between AR and HER2, FOXA1, HER3 or WNT7B across breast tumor subtypes was characterized using Pearson correlation coefficient. The differentially expressed genes between different subtypes were identified using Welch's t test (q≤0.05).
ChIP was performed as previously described (Carroll et al., 2005) and the AR and FOXA1 ChIP-seq experiments were performed in hormone-depleted MDA-MB-453 cells with 10 nM DHT exposure. Briefly, 5-10×106 cells were cross-linked with 1% formaldehyde for 10 min at room temperature. Chromatin was sonicated in SDS lysis buffer to 300-500 bp. Antibodies were incubated with sonicated chromatin overnight, followed by incubation with Dynabeads Protein A (Invitrogen) for 4 h for chromatin collection. Concentrations of the ChIP DNA were quantified by Qubit Fluorometer (Invitrogen). 15 ng of the input (total sonicated chromatin DNA) or 6 ng of the ChIP samples were subjected to poly(A) tailing process (see Supplemental Information for details) and direct sequencing using Helicos HeliScope sequencer at Dana-Farber Cancer Institute. The antibodies used in ChIP experiments are AR (N-20, Santa Cruz), FOXA1 (ab-23738 from Abcam), β-catenin (H-102, Santa Cruz), and TCF4/TCF7L2 (N-20, Santa Cruz and clone6H5-3, Millipore).
The endogenous co-IP experiments were performed using nuclear extracts as described previously (Xu et al., 2010). Briefly, 5 mg of nuclear extracts were incubated with 5 μg of AR or β-catenin antibody on a rotator overnight at 4°C. The protein complexes were precipitated by addition of protein G/A-agarose beads (Roche) with incubation for 4 h at 4°C. After wash for four times for 15 min, the beads were boiled for 5 min in protein sample buffer (Bio-Rad). The antibodies used for IP include AR (N-20, Santa Cruz) and β-catenin (H-102, Santa Cruz). The immunoprecipitated products or the cell lysates were processed for immunoblot as described previously (Wang et al., 2007) using antibodies against AR, HER3, TBP and GAPDH from Santa Cruz, β-catenin and active β-catenin from Millipore, WNT7B from Novus, p-HER3, p-HER2, p-AKT, HER2 and AKT from Cell Signaling.
The staining of AR and HER3 was performed on the archival Yale breast carcinoma tissue microarray cohort by a modified indirect immunofluorescence method as described previously (Camp et al., 2002) (Rubin et al., 2004). The tumor compartment of each histospot was defined using anti-cytokeratin antibodies and the primary antibodies are AR (Dako AR441), and HER3 (noncommercial antibody from CST, D11E5). Secondary labeling of targets was performed by signal amplification using horseradish peroxidase-labeled secondary reagents (species-specific Dako Envision) followed by Cy-5 tyramide incubation. DAPI in an anti-fading mounting medium was used for nuclear staining. Automated quantitative analysis (AQUA®) is a method that allows exact and objective measurement of protein levels within a defined tumor area, as described previously (Camp et al., 2002). All statistical analysis was performed using Statview software and p values were based on two-sided testing, and p<0.05 was considered as statistically significant. The tissue assessed in this study was obtained from the Yale Pathology Archives based on Yale Human Investigation Committee protocols #9505008219, #0304025173 and #0003011706. These protocols, to Dr. David Rimm, allow retrieval of tissue from archives that was consented or has been approved for use with waiver of consent.
Six-week-old female NOD-SCID-IL2Rγc–/– mice (Jackson Lab) were used for xenograft studies. Approximately 6×106 viable MDA-MB-453 cells were resuspended in 30 μl 50% growth factor-reduced Matrigel (BD Biosciences) and injected orthotopically into the fourth inguinal gland as previously described (Torres-Arzayus et al., 2006). The DHT release pellet (60 day release pellet, Innovative Research of America) was implanted subcutaneously at the time of surgery. The mice bearing tumors > 400 mm3 were randomly grouped and treated by daily oral administration with vehicle or bicalutamide (10 mg/kg). Tumor volume was determined by measuring the minimum and maximum tumor diameters using the formula: (minimum diameter)2 × (maximum diameter)/2. All animal protocols were conducted with the approval of the IACUC at Dana-Farber Cancer Institute.
Endocrine therapies that target the estrogen receptor (ER) are the cornerstone of breast cancer treatment for the majority of patients. However, 25-30% of breast tumors do not express ER and are not responsive to existing endocrine therapies. The androgen receptor (AR) is also expressed in a majority of breast tumors, even in those that lack ER. By coupling ChIP-seq technology with genome-wide analysis of gene expression, we have identified a previously unappreciated function of AR in regulating oncogenic Wnt and HER2 signaling in breast cancer leading to androgen-stimulated growth of ER−/HER2+ breast cancers. Our findings provide evidence that drugs targeting AR and the AR-regulated signaling cascade are potential therapies for ER-negative breast cancers harboring HER2 overexpression.
We are grateful to Cheng Fan, Joel Parker, Philip Bernard and Charles Perou for kindly sharing their datasets. We thank Yi Zheng and Xi Chen for technical support. We thank Michael Verzi, Thomas Westerling, Mathieu Lupien and Jian Xu for helpful advice and discussions. This work was partly supported by a grant from the National Cancer Institute (P01CA080111 to M.B.), the National Institutes of Health (R01HG004069 to X.S.L.), and a Department of Defense Award (W81XWH-10-1-0037 to M.N.).
Accession Number: Gene expression microarray and ChIP-seq data have been deposited in the GEO database with accession number GSE28789.
Supplemental Information: Supplemental Information includes five figures and Supplemental Experimental Procedures.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.