Background

Forkhead box A1 (FOXA1), as a member of FOX family proteins, is a master regulator in hormone-sensitive tissues [1]. In breast cancer, FOXA1 plays multi-functional roles. It is involved in the interactions of estrogen receptor (ER) with chromatin and promotes the development of breast cancer [1, 2] while it positively regulates the expression of tumor suppressor gene such as E-cadherin and p27 [3,4,5] and prevents tumor invasion and metastasis. In addition, FOXA1 is also a major determinant of endocrine therapy effects by mediating different ER binding profiles which were associated with endocrine therapy response [6,7,8].

The associations between FOXA1 expression in breast cancer tissues and the prognosis have been explored in a large amount of previous studies, but the results were quite different with positive (high expression of FOXA1 related to a better prognosis) [9,10,11,12], negative [13, 14], and null associations [15], particularly for ER-positive breast cancer. Intriguingly, we noticed that the positive associations mostly occurred in the studies with a shorter follow-up time (5–12 years) [9,10,11,12], whereas the null associations occurred mainly in the studies with a longer follow-up time (up to 20 years) [15, 16]. Particularly, one study reported a null association between FOXA1 expression level and distant-metastasis-free-survival during the overall follow-up period (12.5 years) while there was an increased distant metastasis risk among patients with high expression level in the later period of follow-up [17]. These previous findings strongly suggested that the prognostic effects of FOXA1 on breast cancer may change with time.

In the present study, therefore, we investigated whether FOXA1 in breast cancer tissues have an independent time-varying effect on the prognosis, so as to provide an insight into the role of FOXA1 in breast cancer and help us further understand the timing and duration of endocrine therapy.

Materials and methods

Study population

A total of 1063 females with pathologically diagnosed primary invasive breast cancer and > 1 cm of tumor size in diameter were recruited between January 2008 and December 2015 from the Cancer Center of Sun Yat‐sen University in Guangzhou, China. Patients who lacked information of FOXA1 (N = 13) were excluded from the study and 98.6% (N = 1041) of the included patients were successfully followed up until Dec 31, 2019.

Collections of demographic and clinicopathologic information

Demographic characteristics including age and menopausal status were collected in face-to-face interview by trained investigators using structured questionnaires. BMI and clinicopathologic characteristics including clinical stage, histological grade, ER, progesterone receptor (PR), Human epidermal growth factor receptor 2 (HER-2) status and proliferation index factor Ki-67 (Ki-67) etc. were collected from medical records. Detailed definitions of ER, PR, and HER2 status were previously described in detail [18].

Construction of tissue microarray (TMA)

Formalin-fixed and paraffin-embedded tissues of included patients were retrieved. Hematoxylin and eosin (HE)-stained sections of tissue specimens were reviewed by two experienced pathologists, followed by re-slicing and re-staining with HE. Representative tumor tissue regions and adjacent normal tissue regions (If available) were marked on the re-stained HE sections. From the marked regions, two tumor tissue cylinders and one adjacent normal tissue cylinder (If unavailable, it would be replaced with the tumor tissue) with a diameter of 1 mm were punched out of the corresponding paraffin block as donor block and placed into the TMA paraffin block using an automatic tissue arrayer (MiniCore®, Mitogen, UK). The layout of the cores was determined in advanced by TMA Designer 2 Software. Sections of 4-μm cut from TMA blocks were pasted on the coded glass slides and then placed in the oven at 65 °C for 30 min and finally sealed the tissue surface with paraffin.

Immunohistochemistry (IHC)

The TMAs were baked at 60 °C for 2 h and then dewaxed with xylene and ethanol. Next, antigen retrieval was performed in super-pressure kettle using EDTA (PH 9.0) and then endogenous peroxide was blocked using 3% H2O2. After the preparations, slides were incubated in rabbit monoclonal to FOXA1 (EPR10881)-ChIP Grade (ab170933, diluted 1:100, Abcam) overnight at 4 °C and labeled with the EnVision Detection System (Peroxidase/DAB, Rabbit/Mouse) (Dako K5007). Then slides were developed by diaminobenzidine (DAB) and were counterstained by hematoxylin. Finally, slides were dehydrated and mounted.

IHC stained sections were digitally imaged using Pannoramic Scanner and CaseViewer software. IHC staining was analyzed by an experienced pathologist and scored for staining intensity (0-no staining, 1-weak, 2-moderate and 3-strong) and percentage of tumor cell nuclear staining (0–100). Multiplying staining intensity with percentages yields an H score ranging from 0 to 300. Mean value of H-score from duplicate cores was taken.

Follow up

Patients were followed up by phone calls or out-patient visits every 3 months in the first year, every 6 months in the second and third year after diagnosis and annually thereafter. The endpoints of this study were overall survival (OS) and progression free survival (PFS), which were defined as the time from diagnosis to death and the time from diagnosis to disease progression including recurrence, metastasis, and death, respectively. Survival status was censored at the latest follow-up date or Dec 31, 2019.

Statistical analysis

Kruskal–Wallis test and Mann–Whitney U test were used to test the associations of FOXA1 H-scores (defined as a continuous variable) with age, BMI, clinical stage, menopausal status, histological grade and expressions of HER-2, ER and PR. FOXA1 H-score was modeled as continuous variable and fitted in a Cox proportional hazard model using restricted cubic splines with knots at the 5th, 35th, 65th, and 95th percentiles to estimate the hazard ratios (HRs) and 95% confidence bands assuming proportional hazard (PH). Then FOXA1 H-score was categorized according to the results of restricted cubic splines and the distribution of FOXA1 H-score. Univariate survival analyses of FOXA1 (defined as categorical variable) were perform using Kaplan–Meier method. Log-rank test was used to estimate the differences in survival curves of FOXA1 and to estimate the associations between demographic and clinicopathologic characteristics and breast cancer prognosis to control the potential confounders.

Flexible parametric models were used to perform time-varying effect analysis. The logarithm of the baseline hazard function was modeled as a natural cubic spline function of log time using a 2 degrees-of-freedom according to Akaike information criterion, where FOXA1 and ER were separately treated as variables with time-dependent effect. We also adjusted age at diagnosis, clinical stage, histological grade, and ER status in the models to estimate HRs and 95% confidence intervals (CI) over time. To confirm the independent time-varying effect of FOXA1, we further treated both FOXA1 and ER as covariates with time-dependent effect in the model, in which HRs and 95% CIs of FOXA1 were calculated separately over time for patients with ER-positive or ER-negative tumors. All analyses were conducted using R 3.6.2 and a two-sided P-value below 0.05 was considered as statistical significance.

Results

Demographic and clinicopathological characteristics and the associations with FOXA1 expression

The median age at diagnosis was 48 years (interquartile range: 41‐56 years). More than half of the women had a BMI under 23.0 (51.3%) and 58.7% of them were premenopausal. A great part of the women were diagnosed with low histological grade (grade I/II: 73.3%), early stage (stage I/II: 69.6%), ER-positive (73.5%), PR-positive (72.8%), or HER-2 negative (61.3%) (Table 1).

Table 1 Demographic and clinicopathological characteristics and the associations with FOXA1
Full size table

FOXA1 expression was evaluated in the nucleus of breast cancer cells. The H-score of FOXA1 ranged from 0 to 300 with a median (interquartile range) of 280 (270–285). The median (P25, P75) of FOXA1 H-score for ER-positive patients [285.0 (270.0, 285.0)] was significantly higher than that for ER-negative patients [270.0 (0.0, 285.0)] (P < 0.01). In addition, FOXA1 expression was also lower in tumors with higher grade, higher Ki-67, PR negative, or HER-2 negative (all P < 0.05). No marked differences in FOXA1 expression were observed between different age, BMI, menopausal status, clinical stage, tumor size, nodal status and metastasis (Table 1).

Prognostic effects of FOXA1 on breast cancer

Of the 1041 women, 125 died and 217 experienced disease progression with a median follow-up time of 69.5 months. For OS, the risk of death was relatively flat until around 270 of FOXA1 H-score and then slightly increased from 270 to 285 of FOXA1 H-score, but for FOXA1 H-scores > 285, the risk of death was decreased (Fig. 1A), although the association between FOXA1 and OS was not significant (Fig. 1A, P = 0.59, Pnonlinear = 0.49). For PFS, similar pattern was observed (Fig. 1B, P = 0.075, Pnonlinear = 0.59). Based on the results of restricted cubic splines and the distribution of FOXA1 H-score, we categorized FOXA1 H-score into three levels using tertiles as the cut-off points: ≤ 270, 271–285 and > 285.

Fig. 1
figure 1

Restricted cubic splines of FOXA1 with breast cancer (A) OS and (B) PFS

Full size image

In Kaplan–Meier analysis, no statistically significant differences were observed both for OS and PFS (log rank P = 0.19 and 0.88, respectively). Whereas, the survival probability was the highest for those with FOXA1 > 285 in the early period of follow-up while it changed to the lowest and the curves crossed over in the later period (Fig. 2).

Fig. 2
figure 2

Kaplan-Meier analyses of (A) OS and (B) PFS of FOXA1

Full size image

Time-varying effect of FOXA1 on breast cancer prognosis

Significant time-varying effects of FOXA1 on breast cancer prognosis (both OS and PFS) were observed (all P < 0.01). Compared to FOXA1 ≤ 270, the HRs of death for the patients with H-score = 271–285 of FOXA1 expression increased from 0.35 (95% CI 0.14–0.86) at 6 months after diagnosis to 2.88 (95% CI 1.35–6.15) at 120 months with a crossover at around 36 months, and for those with the highest FOXA1 expression (H-score > 285), the HRs increased from 0.18 (95% CI 0.06–0.53) to 1.25 (95% CI 0.66–2.81) with a crossover at around 84 months in the adjusted model. Similar patterns were also observed for PFS. (Fig. 3; Table 2).

Fig. 3
figure 3

Estimated hazard ratio of OS (A) and PFS (B) by FOXA1. FOXA1 ≤270 is the reference. Shading indicates the 95% CI

Full size image
Table 2 Time-varying HRs and 95% CIs for FOXA1 in association with breast cancer prognosis
Full size table

We further found that there was also a time-varying effect for ER on the survival (P = 0.02 for OS and P < 0.01 for PFS, Additional Table 1) and ER was associated with FOXA1 level, which suggested a potential confounding effect of ER on FOXA1 for the time-varying effect. Therefore, we allowed for the effects of both FOXA1 and ER to change over time in the model and let them adjust each other. It turned out that the time-varying effects of FOXA1 on both OS (P < 0.01) and PFS (P = 0.01) remained but the effect of ER disappeared on OS (P = 0.13). In both ER strata, the time-varying patterns of FOXA1 were similar to that in the whole population (Figs. 4, 5; Additional Tables 2, 3).

Fig. 4
figure 4

Estimated HR of OS by FOXA1 with (A) ER-negative or (B) ER-positive tumors. FOXA1 ≤270 is the reference. Shading indicates the 95% CI

Full size image
Fig. 5
figure 5

Estimated HR of PFS by FOXA1 with (A) ER-negative or (B) ER-positive tumors. FOXA1 ≤270 is the reference. Shading indicates the 95% CI

Full size image

Discussion

In this study, we found that a higher FOXA1 expression level was associated with less aggressive characteristics of breast cancer, such as lower histological grade, lower Ki-67 expression, positive ER or PR. Overall, FOXA1 expression was not significantly associated with the prognosis of breast cancer patients, while there was a marked time-varying effect of FOXA1 on the prognosis. Compared with the low level of FOXA1, a high level of FOXA1 was associated with a protective effect on the survival of breast cancer patients in the early years after diagnosis, but this protective effect gradually diminished with time and an adverse effect occurred in the later years. These time-varying effects of FOXA1 were independent of ER status.

In consistent with our study, a lot of previous population studies have also found that FOXA1 expression level was higher in tumors with less aggressive characteristics [16, 19, 20]; Cellular experiments revealed that upregulation of FOXA1 inhibited epithelial to mesenchymal transition (EMT), migration and invasion in breast cancer cells [21, 22]. Moreover, another finding of the present study that the significant association of a high FOXA1 expression level with a better prognosis of breast cancer in the early stage after diagnosis also supported this result.

It was a quite interesting phenomenon that the protective effect on breast cancer survival of a high FOXA1 expression level gradually diminished and shifted to a detrimental effect in later years after diagnosis. One of the possible reasons for this time-varying effect was the altered FOXA1 expression level. It has been found that FOXA1 expression level decreased in long-term tamoxifen-treated MCF7 cells [23] and ER-negative cells exposed to bisphenol A [22]. Another reason was the termination of endocrine therapy, which may result in the diminished protective effect of high FOXA1 expression level because FOXA1 plays its role depending on endocrine therapy [6]; the Chinese Anti-Cancer Association (CSCA) Diagnosis and Treatment Guidelines for breast cancer (version 2008) [24] recommended ER positive patients to receive endocrine therapy for 5 years after surgery and most of the patients complied with the guideline [25], which was also consistent with our results that the protective effect of highest FOXA1 expression on cancer progression shifted to an detrimental effect at 60 months after diagnosis. The third reason was the acquisition of endocrine resistance caused by FOXA1 through transcription reprogramming in breast cancer cells with the extension of endocrine therapy time [8, 14, 26,27,28].

We found that the time-varying effects of FOXA1 on breast cancer were independent on ER status while the same effects of ER were affected by FOXA1 to some extent. This phenomenon may be explained by that FOXA1 was essential for sustained ER expression [29] and was the upstream of ER-chromatin interactions, regulating more than 90% of ER binding events [6, 30]. In addition, this result likely suggested that the time-varying effects of FOXA1 may also be mediated through other pathways, such as androgen receptor [31] and AGR2 [32, 33]. The exact mechanisms remained to be explored.

There were some limitations in this study. First, only patients with tumor > 1 cm were included and that may lead to selective bias. However, FOXA1 expression was independent of tumor size in this study, causing non-differential bias on the associations between FOXA1 and prognosis. Second, we were unable to collect the information on the changes of FOXA1 over time that made us fail to make sure whether the time-varying effect was due to changes in FOXA1 expression. Third, we didn’t collect the information about treatment which was associated with prognosis. However, since the treatment was determined according to the clinicopathological characteristics such as ER status and tumor stage, adjustment of these characteristics in the statistic models largely controlled the confounding effects of the treatment. Finally, a follow-up time up to 10 years may lead a bias estimation in the later stage of follow-up. However, the crossover time-points of time-varying effects occurred before the median follow-up time (72.2 months), indicating that the results of the time-varying effect were reliable.

Conclusion

This study firstly revealed the time-varying effect of FOXA1 on breast cancer prognosis: a higher expression of FOXA1 was associated with a better survival in the early stage after diagnosis while it associated with a poor survival in the late stage. Similar results were observed when further treated ER as covariates with time-dependent effect in the model. These findings provided an insight into the roles of FOXA1 as a marker of breast cancer prognosis and argued in favor of an extended endocrine therapy rather than 5 years.