FOXA1 and AR in invasive breast cancer: new findings on their co-expression and impact on prognosis in ER-positive patients

We collected a series of 479 female patients that underwent surgery for BC from June 1994 to December 2012 at the Breast Unit of the Città della Salute e della Scienza Hospital of Turin, Italy. All patients were treated with surgery, either mastectomy or wide local excision, followed by radiotherapy.

Clinical-pathological data such as age at time of diagnosis, surgery (conserving surgery vs radical mastectomy), type of therapy (hormonal therapy, chemotherapy), type and site of recurrences, histological types, tumor size (< 15 mm vs ≥ 15 mm), nodal involvement, histologic grade and vascular invasion were collected. Medical charts of all patients were reviewed to confirm accuracy of previously recorded data. Tumor slides were re-evaluated to select representative blocks that were used to construct multicore tissue microarrays (TMAs, tissue arrayer Galileo TMA CK 3500, Integrated Systems Engineering Srl, Milan, Italy), as previously described [21].

To confirm the results of the diagnostic reports, immunohistochemistry (IHC) was performed on TMA sections using an automated slide processing platform (Ventana BenchMark AutoStainer, Ventana Medical Systems, Tucson, AZ, USA) and the following primary antibodies were used: prediluted anti-ER rabbit monoclonal antibody (SP1, Ventana-Roche, Tucson, AZ, USA); prediluted anti-Progesterone receptor (PgR) rabbit monoclonal antibody (1E2, Ventana-Roche); anti-Ki67 monoclonal antibody (MIB1, diluted 1:100 Dako); anti-human c-erbB2 oncoprotein (Ventana Pathway HER-2/Neu-4B5). In addition, AR and FOXA1 expression were tested using anti-AR mouse monoclonal antibody (AR441, diluted 1:50, Dako, Glostrup, Denmark) and prediluted anti-FOXA1 mouse monoclonal antibody (2F83, Ventana-Roche). Positive and negative controls (omission of the primary antibody and IgG-matched serum) were included for each IHC run.

The cut-off value for ER and PgR expression was set at 1%, as suggested by St Gallen Consensus meeting [22], and the same cut-off was also adopted for AR and FOXA1 expression [5]. The percentage of Ki67-positive cells was recorded and the cut-off for dichotomizing tumors with low and high proliferative fraction was established at 20% according to 2013 St Gallen Consensus meeting [23] and also on the basis of the median Ki67 value of our local laboratory [24, 25]. HER2 status was classified as negative (score 0, 1+ and 2+ not amplified) or positive (when scored 3+ by IHC or HER2 amplified by FISH) according to the recommended guidelines for invasive carcinoma [26].

To determine the specificity of AR and FOXA1 antibodies, we compared gene expression levels (using qPCR) with IHC results. The relationship between AR and FOXA1 was validated using relative quantification mRNA analyses.

qPCR for AR and FOXA1 mRNA was performed on 65 fresh-frozen BC samples (Fig. 1). Total RNA was extracted from tissues using TRIzol Reagent (Invitrogen Ltd., Paisley, UK) following manufacturer’s instructions. DNase I was added to remove remaining genomic DNA. 1 μg of total RNA was reverse-transcribed with iScript cDNA Synthesis Kit (Bio-Rad Laboratories Inc., Hercules, CA, USA), following manufacturer protocol. Primers (Additional file 1: Table S1) were designed using Beacon Designer 5.0 software according to parameters outlined in the Bio-Rad iCycler Manual. Specificity of primers was confirmed by BLAST analysis. qPCR was performed using a BioRad iQ iCycler Detection System (Bio-Rad Laboratories Inc., Hercules, CA, USA) with SYBR green fluorophore. Reactions were performed in a total volume of 25 μl containing 12.5 μl of IQ SYBR Green Supermix (Bio-Rad Laboratories Inc., Hercules, CA, USA), 1 μl of each primer at 10 μM concentration, and 5 μl of the previously reverse-transcribed cDNA template. The protocol used was as follows: denaturation (95 °C for 5 min) and amplification repeated 40 times (95 °C for 15 s, 60 °C for 30 s). At each run, a melting curve analysis was performed to ensure a single specific amplified product for every reaction. Results were normalized using the Delta-Ct (Δct) method, using β-actin as housekeeping gene. Samples with a Δct ≤ 6 were defined as positive.

Pearson’s Chi square test and Student’s t-test were preliminary performed to compare respectively categorical and continuous variables, and to evaluate potential differences in the variable distribution among groups. Test for median and means (Analysis of Variance-ANOVA) were performed. For more than two groups Tukey HSD post-hoc test was performed. Disease-Free Interval (DFI) was calculated from the date of surgical excision of the primary tumor to the date of first disease relapse or last check-up. Disease-specific survival (DSS) was calculated from the surgical excision date of the primary tumor to the date of BC death or last check-up [24, 27]. Survival distribution curves were plotted using the Kaplan-Meier method and the statistical comparisons were performed using the log-rank test. Cox regression analyses were carried out on DFI and DSS to calculate crude and adjusted HRs and 95% confidence intervals (CIs) for the different study group. Cases lost to follow up and cases with a non-BC related cause of death were censored at the last follow up control. The step-wise model selection method was used to determine the final Cox regression model. Akaike Information Criterion test (AIC) and likelihood ratio test (LRT) were carried out to measure how selected variables improve parsimony and goodness of fit of the selected model. The proportional hazard assumption was assessed with the Schoenfeld residuals. This did not give reasons to suspect violation of this assumption. All statistical tests were two sided. P-values < 0.05 were considered significant. Statistical analyses were performed using Stata/SE12.0 Statistical Software (STATA, College Station, TX).

Clinical and histopathological characteristics of the whole population are reported in Table 1. The median follow up was 10.1 years (7,7-12,7). The majority of patients was over 50 years (> 80%) of age and underwent conservative surgery. Positive expression of ER, AR and FOXA1 was observed in 78, 60 and 85% of cases respectively. As previously reported [13, 15], in our cohort FOXA1 positivity was associated with small tumor size (< 15 mm), absence of lymph node metastases, low histological grade, no special type (NST) histotype, low level of Ki67, as well as, with ER+ and PgR+ tumors (Additional file 1: Table S2). In the consecutive series of patients, 58% of cases showed AR+/FOXA1+ (Table 2), while 14% presented AR-/FOXA1- immunophenotype and only 1.7% of cases were AR+/FOXA1-. This latter subgroup, did not show specific features. Compared to the other subgroups, FOXA1-/AR- BC phenotype was more frequently associated with high histological grade, large tumor size, no expression of ER and PgR and high proliferation index (P < 0.001) (Additional file 1: Table S3).

We found a strict correlation of FOXA1 and AR mRNA and protein expression (Fig. 2). To correlate the expression of ER, AR and FOXA1, we decided to use qPCR results, because this procedure allows quantifying more precisely the level of expression of each molecule. As shown in Fig. 3, there was a linear correlation (Spearman’s correlation) of the level of FOXA1 mRNA with the level of AR (r = 0.8975; P < 0.001) (Fig. 3a) and ER (r = 0.7326; P < 0.001) mRNA (Fig. 3b).

Furthermore, FOXA1 mRNA was closely related to AR mRNA expression, regardless of ER status. Indeed, FOXA1 mRNA was expressed in all samples with ER+/AR+ (27 cases) and ER−/AR+ (3 cases) (Low delta-Ct. Fig. 4b and d), in 8/25 ER+/AR- cases and in only 1/10 ER−/AR- cases (High Delta-Ct. Fig. 4c and e).

At univariate analysis performed on whole cohort, metastatic lymph nodes, histological grade, vascular invasion, ER and PR positivity, high Ki67 and HER2 overexpression were confirmed as significant prognostic factors. Additionally, the expression of AR and FOXA1 were associated with a better DFI and DSS (Table 3, Additional file 1: Figure S1).

To analyze the impact of FOXA1 and AR in patients with BC (ER+ or ER-), we created three BC subgroups (FOXA1+/AR+; FOXA1+/AR-; FOXA1−/AR-). We were unable to perform any analyses on the FOXA1−/AR+ BC since only 8 patients carried this phenotype (Table 2). As shown in Fig. 5, in the consecutive series of patients, the lack of expression of both, FOXA1 and AR (FOXA1−/AR-), was related to a worse DFI and DSS compared to the other groups.

Finally, we investigated the relationship between FOXA1, AR and prognosis in BC patients stratified for ER expression. As shown in Fig. 6, in ER+ BC, FOXA1 expression was closely related to good prognosis independently of AR expression.

Multivariate analyses (Table 4) performed on ER+ BC confirmed that FOXA1 may provide more information than AR on DFI, but not on DSS. In the subset of patients with ER- BC, FOXA1, alone or in association with AR, did not show any relationship with outcome (data not shown).