Lack of both androgen receptor and forkhead box A1 (FOXA1) expression is a poor prognostic factor in estrogen receptor-positive breast cancers

The present study aimed to examine the associations between androgen receptor (AR) and forkhead box A1 (FOXA1) and to investigate clinicopathological features and survival according to both biomarker status in estrogen receptor (ER)-positive breast cancers using in vitro study, patient cohort data, and the cBioPortal for Cancer Genomics and Kaplan-Meier Plotter websites. Experiments using T47D and ZR75-1 demonstrated AR-overexpressing cell lines decreased in cell proliferation through downregulation of ER, but FOXA1 did not change. Knockdown of FOXA1 resulted in a significantly reduced cell viability. Patients with immunohistochemically AR(-)/FOXA1(-) tumor frequently showed node metastasis, high grade, and high Ki-67 proliferation, therefore, significantly worse survival in ER-positive disease. AR and FOXA1 mRNA levels were significantly higher in ER-positive than in ER-negative tumors and AR-low/FOXA1-low tumors showed high grade, frequent basal-like subtype and worse disease-free survival in ER-positive cancers of public gene dataset, similarly to patient cohort results. The Kaplan-Meier Plotter analysis independently validated patients with both low AR/FOXA1 tumor were significantly associated with worse relapse-free survival in ER-positive cancers. This study suggests that distinctive clinicopathological features according to AR and FOXA1 are determined and a lack of both biomarkers is an independent poor prognostic factor in ER-positive tumors.


I. INTRODUCTION
Androgen receptor (AR) is a mainly ligand-dependent transcription factor that regulates target gene expression. The AR gene is located on the X chromosome.
The 110-kDa AR phosphoprotein mediates diverse biological actions in the development and maintenance of the reproductive, musculoskeletal, cardiovascular, immune, neural, and hematopoietic systems and is involved in the development of malignancies in the prostate, bladder, liver, kidney, and lung. 1,2 Recent attention has focused on the emerging roles of AR not only as a prognostic and predictive factor, but also as a therapeutic target in breast cancer patients. [3][4][5] A systematic review and meta-analysis showed that positive AR expression was significantly associated with better survival of patients with early breast cancer irrespective of estrogen receptor (ER) status. 6 However, in vitro evidence partly supported clinical studies and AR showed antiproliferative activity in only ER-positive breast cancers but rather AR signaling promoted tumor growth in ER-negative and human epidermal growth factor receptor 2 (HER2)-positive breast tumors. 3,7 Furthermore, Lehmann et al. 8 identified six subtypes of triple-negative breast cancer (TNBC), one of them being a luminal androgen receptor (LAR) subtype with distinct features among diversely heterogeneous TNBCs. 9 Many studies consistently suggest that AR is a favorable biomarker in hormone receptor-positive tumors, but the clinical or biological impact of AR has not been clearly defined. Therefore, additional approaches are necessary to clarify the various roles of AR and its control mechanisms according to ER status.
Forkhead box A1 (FOXA1), initially discovered as hepatocyte nuclear factor 3α (HNF3α), is a member of the FOX family transcription factors. 10 Because of a lack of the basic amino acids in FOXAs for chromatin compaction, binding of FOXAs to nucleosomes creates an open chromatin configuration that can recruit other transcriptional regulators. 11,12 Thus, FOXA1 belongs to a 'pioneering factor'. 13 Recent meta-analyses of breast cancers demonstrated that high FOXA1 levels were positively correlated with ER-positive and progesterone receptor (PR)-positive tumors. 14 Patients with high FOXA1 expression showed better disease-free survival (DFS) and overall survival (OS). 15 A study by Hurtado et al. 16 supported that FOXA1 played a key role in differentially influencing interactions between ER and chromatin. It was required for almost binding events and transcriptional activities of ER in breast cancer cells. Genetic analysis of invasive lobular carcinomas, which were predominantly categorized as the luminal A subtype, exhibited recurrent FOXA1 mutations and correlation with high FOXA1 activity. 17 The data confirmed that FOXA1 was closely associated with the ER signaling pathway and suggested that FOXA1 may explain heterogeneous features of hormone receptor-positive tumors.
In contrast, molecular apocrine breast tumors are characterized by apocrine features upon histopathological examination, ER-negativity, AR-positivity, and a high association with HER2 amplification. 18,19 Even though most molecular apocrine breast cancers are ER-negative, they have AR-driven, hormonally regulated transcriptional activities mediated by FOXA1, similar to ER-mediated transcription in luminal subtype breast cancers. 20 Immunohistochemically, AR and FOXA1 were expressed in 100% (54/54) and 93% (50/54) of histologically diagnosed, almost ER-negative apocrine breast cancers, respectively. 21 In hormone-dependent prostate cancers, FOXA1 is a global mediator of AR action and facilitates prostate cancer growth. 22 A recent study suggested that FOXA1 promotes cell proliferation through AR by activation of the Notch pathway in endometrial cancers. 23 An ancillary immunohistochemistry (IHC) study of AR and FOXA1 in 592 TNBCs from the UNICANCER PACS08 adjuvant multicenter trial suggested that co-expression of both markers seems to be associated with distinct clinicopathological features of luminal tumors compared to other TNBCs. 24 These findings implied a close molecular connection between AR and FOXA1 and the important clinical roles of both biomarkers in various cancer types, including breast malignancy.
However, the clear genetic or clinical implications of these biomarkers on tumor biology and patient prognosis have not been fully explained according to ER status of breast cancer, especially in ER-positive tumors.
The purpose of the present study was to explore the genetic expression patterns and associations between AR and FOXA1 by ER status determined from web-based breast cancer genetic datasets. Next, it was to examine the influence among biomarkers through an in vitro ER-positive cell lines study.
Finally, the present study aimed to investigate and validate clinicopathological characteristics and survival outcomes according to combined AR and FOXA1 protein and mRNA status in mainly ER-positive patients using clinical data of a single institution and public datasets.

Web-based bioinformatics analysis
A. The cBioPortal for Cancer Genomics Genomic analysis was performed for investigating the associations between ESR1, AR, and FOXA1 through the cBioPortal for Cancer Genomics The probability of relapse-free survival according to AR and FOXA1 status including subgroup analyses was calculated using the KM Plotter (http://kmplot.com/analysis). 28 It is an online tool that allows analysis of the effects of 54,675 genes on survival by using 10,188 cancer samples, which includes 4,142 breast cancer patients with a mean follow-up duration of 69 months (June, 2016). Survival and gene expression data were derived from the Gene Expression Omnibus (Affymetrix microarray only), European Genomephenome Atlas, and TCGA. The Affymetrix probe set IDs selected were 226197_at for AR and 204667_at for FOXA1 in the present study. Multiple genes were entered through a multigene classifier using the mean expression of selected biomarkers. To analyze the prognostic value of combined AR and FOXA1, the patient samples were split into two groups using the lower quartile as a cutoff value. Hazard ratio (HR) with 95% confidence interval (CI) and log-rank p-value were calculated, and survival curves were displayed on the webpage.

In vitro cell lines study
A. Cell culture Human breast cancer cell lines (T47D and ZR75-1) were obtained from the American Type Culture Collection (Manassas, VA, USA). All reagents related to animal cell culture were purchased from Life Technologies (Big Cabin, OK, USA). Cells were cultured in Dulbecco's modified Eagle's medium.
Among them, Figure 3 shows 24 neighbor genes and 5 interaction types with known gene function in tumor cells or a close connection between ESR1, AR, and FOXA1. and frequently altered neighbor genes from the TCGA Provisional dataset.
Copy number alteration and mutation of the query genes are not entered into the network analysis. Only interactions closely connected among the query genes are presented.

In vitro cell lines study
To investigate the association between AR and FOXA1 in ER-positive tumors, an in vitro study was performed using T47D and ZR75-1 breast cancer cell lines.
As shown in Figures 4A and 4B expression level was not changed by overexpression of AR (Fig. 5C).  Next, the effects of FOXA1 overexpression on ER activity were compared in mock-and AR-overexpressing T47D cell lines. As shown in Figure 6A, overexpression of FOXA1 in these cell lines had no effect on ER activity. However, knockdown of FOXA1 resulted in a significant reduction of cellular viability on day 5 (Fig. 6C), suggesting that FOXA1 has essential roles for viability of the ER-positive tumor cell lines, although there were no direct effects on ER and AR activities (Fig. 6). (E) Western blot analysis was performed to detect protein level of FOXA1.
Expression of β-actin was analyzed as a loading control.      When examining the clinicopathological characteristics of ER-positive tumors based on AR and FOXA1 status, similar trends were observed between the AR(+)/ FOXA1(+) group and histopathological parameters, including small tumor size, node-negative disease, lower TNM stage, histologic grade I/II, and PR-positive expression ( Table 3). The AR(-)/FOXA1(-) group frequently showed node metastasis, high grade, PR-negative expression, and high Ki-67 proliferation.    (Fig. 7).   (Table 4.) Node metastasis, HER2-positivity, and absence of chemotherapy were also significantly associated with increased risk of poor DFS and OS in the TMA study.    Figure 9. The distribution of mRNA expression was unimodal for AR, but bimodal for FOXA1 in both datasets.   (Table 5). In ER-positive cases, AR-low/ FOXA1-low tumors were significantly associated with age > 50 years and PR-negativity. Although AR-high/FOXA-high cases demonstrated higher frequencies of lower TNM stage, there was no statistical difference. In the whole population of the METABRIC dataset, which provides more information regarding histopathology and treatment patterns than the TCGA Provisional dataset, AR-low/FOXA1-low tumors were significantly associated with age at diagnosis ≤ 50 years (p < 0.001), high grade (p < 0.001), ER-negativity (p < 0.001), PR-negativity (p < 0.001), HER2-negativity (p < 0.001), and the basal-like subtype (p < 0.001). AR-high/FOXA1-low subgroup showed the highest frequency of stage III disease (p = 0.004), HER2-positive tumors (p < 0.001), and the HER2-enriched subtype (p < 0.001). The clinicopathological characteristics determined by to AR and FOXA1 status in ER-positive breast cancer patients of the METABRIC dataset are presented in Table 6. Similarly, in ER-positive tumors, AR-low/FOXA1-low cases were significantly associated with high grade, PR-negativity, the basal-like subtype, and chemotherapy administration. AR-low/FOXA1-low tumors also showed higher advanced stage and HER2-negativity, but without statistical significance.     Table 7). The AR-low/FOXA1-low group was determined to be a significantly poor prognostic factor than the AR-low/FOXA1-high and AR-high/FOXA1-high groups for DFS and the AR-low/FOXA1-high group for OS when age, stage, grade, HER2, and use of chemotherapy and hormone therapy were adjusted. log-rank p = 0.392).

IV. DISCUSSION
An exploration of the web-based genetic analysis in the present study showed that approximately 10% of breast cancers were altered in ESR1, AR, or FOXA1 genes and generally changes in genes concurrently occurred even though the frequency was low. It has been suggested that long length of CAG repeats in the exon 1 of the AR gene is associated with decreased efficacy of androgenic activity and the increased risk of breast cancer in women. 33 16,38 These findings suggest that AR along with FOXA1 exerts its function in specific cellular situations.
Unfortunately, the present study exhibited no changes in FOXA1 mRNA levels by AR overexpression. In addition, FOXA1 overexpression did not affect ER activity in mock-and AR-overexpression cell lines but knockdown of FOXA1 induced marked loss of ER-positive cell viability as shown in Figure 6. These findings suggested that FOXA1, as a pioneering factor in downstream of the ER signaling pathway, played an important, complex role in tumor survival, as shown as a lineage-specific oncogene in luminal cancer cell lines. 16 dataset. This association was also validated in the analyses of the KM Plotter.
The TCGA Provisional dataset did not present statistically significant associations in the present analysis, but a mean follow-up duration of the TCGA Provisional dataset was shorter than that of other datasets. Therefore, future analysis with longer follow-up periods might show different findings and should be conducted to confirm the hypothesis.
Interestingly, according to analyses of ER-negative tumors using the