FOXC1: A Specific Biomarker for Triple-Negative Breast Cancer Diagnosis and Classification

A cohort of 2443 patients diagnosed with invasive breast cancer without distant metastasis between July 2019 and April 2020 at Fudan University Shanghai Cancer Center (Shanghai, China) was enrolled in this study. All experiments were performed and data were generated in accordance with the ethics standards of the relevant national and international rules and regulations (Good Clinical Practice, Declaration of Helsinki). This study was approved by the Ethics Committee of Fudan University Shanghai Cancer Center, and each participant signed an informed consent form before enrollment in this study. The clinicopathologic characteristics were retrieved from the medical records, including age, tumor size, tumor grade, and lymph node metastasis. Patients' tumors were staged according to the 8th edition of the American Joint Committee on Cancer (AJCC) staging manual.¹²  Pathologic staging including tumor size (T) and lymph node (N) status was determined by surgical resection. Tumor histologic grading was performed according to the Nottingham grading system by assessing the following 3 morphologic features: tubule formation, nuclear pleomorphism, and mitotic counts. Hematoxylin-eosin–stained slides from each case were reviewed by the authors, and the diagnoses were confirmed according to the updated 5th edition of the World Health Organization (WHO) classification.¹³  According to the 2019 5th edition of the WHO classification, invasive carcinomas with apocrine differentiation are defined by strict morphologic and IHC criteria as follows: apocrine morphology in greater than 90% of tumor cells; and ER⁻, PR⁻, gross cystic disease fluid protein 15–positive (GCDFP15⁺), and AR⁺ immune profile, which was defined as pure apocrine carcinoma in a previous study.¹⁴ 

The details of the IHC antibodies used are shown in Table 1. All staining was performed with a Ventana BenchMark Ultra autostainer (Ventana Medical System Inc, Roche, Tucson, Arizona) according to the BenchMark ULTRA advanced staining system operator guide. Antigen retrieval was conducted with ULTRA Cell Conditioning Solution (ULTRA CC1, pH = 8.5) at 90° to 100°C. Nuclear ER/PR staining in 1% or more was defined as ER/PR-positive according to the 2020 American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guideline.¹⁵  Samples containing 1% to 10% ER⁺ cells were defined as ER low-positive. HER2 status was assessed as proposed in the 2018 ASCO/CAP guideline.¹⁶  HER2 was considered positive with an IHC score of 3+ or when gene amplification was detected by fluorescence in situ hybridization (FISH). Ki-67 staining was assessed according to the 2011 recommendations from the International Ki-67 in Breast Cancer Working Group.¹⁷  Ki-67 values of 20% or above were considered high, while those less than 20% were considered low.¹⁸  Only nuclear staining was considered for positive AR, FOXC1, and GATA3 expression, whereas cytoplasmic staining characterized positive cytokeratin (CK) 5/6, CK14, and GCDFP-15 expression. Androgen receptor was assessed semiquantitatively with the Allred scoring method, which incorporates the intensity and distribution of reactivity. Immunohistochemistry staining with AR, GCDFP15, CK5/6, and CK14 was considered positive when more than 1% tumor cells were stained. FOXC1 was assessed semiquantitatively according to the distribution of reactivity.

Patients were divided into the following immunohistochemical surrogate subtypes according to the definitions adopted by the 2013 St. Gallen Consensus Panel: (1) luminal A–like subtype (ER⁺, PR⁺ [≥20%], HER2⁻, and Ki-67 <20%), (2) luminal B–like (HER2⁻ subtype: ER⁺, HER2⁻, Ki-67 >20% or PR⁻/⁺; HER2⁺ subtype: ER⁺, HER2 overexpression with any PR or Ki-67 index), (3) HER2⁺ (nonluminal) (HER2 gene amplification or HER2 IHC 3+ ), and (4) TNBC (negative for both hormone receptors and HER2).

SPSS software version 19 (SPSS Inc, Chicago, Illinois) was used to perform all statistical analyses. Clinicopathologic parameters were compared between different subgroups by using χ² analysis. A P value ≤ .05 was considered statistically significant.

A total of 2443 patients with invasive breast cancer were included in this study (Table 2), and of these, 762 (31.19%) were diagnosed with the luminal A–like subtype, 1010 (41.34%) were diagnosed with the luminal B–like subtype, 301 (12.32%) were diagnosed with the HER2⁺ subtype, and 370 (15.15%) were diagnosed with the TNBC subtype. Only 6 patients (0.24%) were male. The median age at diagnosis was 51 years (range, 21–97 years). Tumor size ranged from 0.1 to 11 cm in diameter, with a median diameter of 2.3 cm. Except for 2288 patients (93.66%) with IBC-NST, 155 patients (6.34%) with the following special histologic types were included: lobular, mucinous, metaplastic, apocrine, adenoid cystic carcinoma (AdCC), and micropapillary carcinomas.

FOXC1 expression was scored according to the percentage of positively stained cancer cells with nuclear staining. Because the cutoff value of FOXC1⁺ expression has not been consistently confirmed, we first investigated the most appropriate cutoff value. The area under the receiver operating characteristic (ROC) curve (AUC) confirmed the high diagnostic value of FOXC1 in distinguishing the triple-negative phenotype from the other subtypes (AUC = 0.861, P < .001; Figure 1, A). Additionally, the ROC curve indicated that 1% or higher was the best cutoff value, as it displayed a maximized sum of sensitivity (77.27%) and specificity (90.54%), while a cutoff value of 10% or higher displayed the highest specificity value (94.43%) (Figure 1, A). Thus, in this study, FOXC1 was considered positive when at least 1% of tumor cells were stained.

In breast cancer, FOXC1 positivity was observed most frequently in TNBC (Figure 1, B through J). FOXC1 was positive in 77.84% (288 of 370) of patients with TNBC, with predominantly intermediate and high expression (scores ≥10%) in 91.67% (264 of 288) of cases. FOXC1 demonstrated a low expression rate and intensity in other subtypes. FOXC1 was expressed in only 4.86% (37 of 762) of luminal A–like breast cancers and exhibited weak expression (scores <10%) in 59.46% (22 of 37) of cases. FOXC1 was expressed in 8.02% (81 of 1010) of luminal B–like breast cancers, and of these, weak expression was observed in 36 cases (44.44%, 36 of 81). Moreover, FOXC1 was expressed in 25.91% (78 of 301) of HER2⁺ breast cancers with weak expression observed in 56 cases (71.79%, 56 of 78). Moreover, of the non-TNBC patients with positive FOXC1 expression, only a few cases demonstrated strong staining (scores higher than 80%) as follows: luminal A–like (4 of 37), luminal B–like (17 of 81), and HER2⁺ (4 of 78).

In the 118 cases of luminal A–like and B–like subtypes with positive FOXC1 expression, patients with high FOXC1 expression (scores higher than 40%) showed significantly low ER expression (1%–10%, P = .004; Figure 2, A through I) and negative PR expression (P < .001; Figure 2, A through H and J) when compared to those with low FOXC1 expression.

Thus, FOXC1 is mainly expressed in triple-negative phenotype breast cancer and is a useful marker in distinguishing TNBC from other molecular subtypes of breast cancer. FOXC1 expression is negatively correlated with ER and PR expression.

As FOXC1 was mainly expressed in the TNBC subtype, we next investigated its association with the clinicopathologic features of TNBC. Patients with TNBC and FOXC1 expression were significantly more likely to be younger (P < .001) than FOXC1⁻ patients, and this result was also observed in the cohort of 327 IBC-NST cases only (P = .007). Moreover, the histologic grade (P < .001) was significantly higher in patients with positive FOXC1 expression, and this result was also observed in patients with IBC-NST only (P = .004). No significant association was found between FOXC1 expression and tumor size or nodal status (Table 3; Supplemental Table 1, see supplemental digital content at https://meridian.allenpress.com/aplm in the August 2022 table of contents).

Morphologically, among the 370 patients with TNBC, in addition to IBC-NST (327 patients, 88.38%), the following special histologic subtypes were included: metaplastic carcinoma (6 patients; Supplemental Table 2), invasive carcinoma with apocrine differentiation (29 patients), invasive lobular carcinoma (5 patients), and AdCC (3 patients). Interestingly, we observed that FOXC1 expression was significantly different in patients with different histologic subtypes (Figure 3, A through H; Table 3). TNBC with FOXC1 expression was more likely to be IBC-NST (85.02%, 278 of 327), metaplastic carcinoma (5 of 6, 83.33%), and AdCC (3 of 3, 100%). In the 6 cases of metaplastic carcinoma, the case with negative FOXC1 expression was metaplastic carcinoma with heterologous mesenchymal differentiation accompanying a conventional invasive ductal carcinoma component (Supplemental Table 2). Nevertheless, of the 29 patients with invasive carcinoma with apocrine differentiation, only 1 patient showed FOXC1 expression, and the staining was weak (<10%).

In patients with TNBC, the expression rate of FOXC1 was higher than the rates of AR (173 of 370, 46.76%), GCDFP15 (124 of 370, 33.51%), CK5/6 (235 of 370, 63.51%), and CK14 (178 of 370, 48.11%) expression. In addition, 82.70% (306 of 370) of patients demonstrated high Ki-67 index (Table 4). FOXC1 expression showed significantly positive correlation with the expression of Ki-67 (P < .001) and basal markers (CK5/6, P < .001; CK14, P < .001) (Table 4). FOXC1 expression was negatively associated with the apocrine differentiation–related markers: AR (P < .001) and GCDFP15 (P < .001; Table 4, Figure 4, A through J).

Taken together, the findings demonstrated that FOXC1 expression is positively correlated with unfavorable pathologic features in TNBC and that FOXC1 positive expression is associated with the basal-like phenotype of TNBC. Moreover, FOXC1 expression is negatively correlated with apocrine morphology and apocrine-related marker expression. These findings suggest that the inclusion of FOXC1 in an immunopanel may be relatively useful for the classification of TNBC.

To further investigate the diagnostic value of FOXC1 mRNA (messenger RNA) expression in breast cancer, we analyzed FOXC1 mRNA expression in 731 invasive breast cancer cases in breast cancer datasets from The Cancer Genome Atlas. As compared to normal breast tissues, FOXC1 was upregulated in BLBC and downregulated in the luminal and HER2 subtypes (Figure 5, A). The AUC confirmed the high diagnostic value of FOXC1 mRNA levels in distinguishing BLBC from the other subtypes (AUC = 0.975, P < .001; Figure 5, B). In BLBC, 96.21% (127 of 132) of patients had high FOXC1 expression, especially those with IBC-NST (117 of 120) and metaplastic carcinoma (6 of 6; Supplemental Table 3). Moreover, FOXC1 mRNA expression was negatively correlated with AR and GCDFP15 expression in BLBC (Figure 5, C and D), which was consistent with the findings in our cohort. These results demonstrate that FOXC1 mRNA expression is also a specific marker of BLBC, confirming that FOXC1 is valuable in breast cancer subtyping.

FOXC1 has been demonstrated to be of critical and central importance in the progression of BLBC.⁶  FOXC1 mRNA and protein levels have been reported to be specific prognostic and diagnostic biomarkers for BLBC.⁹  The expression of FOXC1 across the spectrum of morphologic and IHC-based breast cancer subtypes has not been well defined. In this study, FOXC1 protein expression was evaluated by IHC in a large cohort of patients with invasive breast cancer to systematically profile its expression in breast cancer, particularly in TNBC.

In previous studies, FOXC1 expression has been mainly used to identify BLBC. Furthermore, FOXC1 is an important marker that can be used in the subclassification of TNBCs. A recent study by Zhao et al,¹⁰  involving the surrogate classification of TNBC, defined FOXC1 expression in at least 10% positive tumor cells as positivity. A subsequent phase Ib/II FUTURE trial has demonstrated the clinical benefit of subtyping-based targeted therapy for refractory metastatic TNBC.¹⁹  These studies have provided further evidence suggesting that FOXC1 is of great importance in TNBC diagnosis and subtyping. However, studies on the cutoff value of positive FOXC1 expression in breast cancer are limited, and the cutoff value of positive FOXC1 expression varies. Some studies have defined positive FOXC1 expression as any nuclear staining in tumor cells.⁹  Jensen et al²⁰  reported that cutoff values of 4 or 5 (as assessed by a modified Allred score, from 0 to 8) show a maximized sum of the sensitivity and specificity values to predict the basal-like subtype defined by PAM50. Kim et al¹¹  used an Allred score of 4 or above as the cutoff value for FOXC1 expression according to the study by Jensen et al,²⁰  and further divided the basal-like type of TNBC into basal-like immune-activated and basal-like immunosuppressed types, based on the IDO1 and FOXC1 expression status. Thus, identifying a consistently used cutoff value of FOXC1 positive expression is highly significant for establishing an internationally accepted surrogate molecular subtyping strategy for TNBC.

In our study, the results also demonstrated that FOXC1 was primarily expressed in TNBC but rarely expressed in the luminal-like and HER2⁺ subtypes, which was consistent with previous studies. ROC curves confirmed the high diagnostic value of FOXC1 in differentiating the triple-negative phenotype from other subtypes. The threshold of 1% positive staining was established as the best cutoff value. The mechanism of BLBC-specific expression of FOXC1 has been widely investigated in several studies. It has been reported that breast cancer susceptibility gene 1 (BRCA1) and GATA3 interact at the FOXC1 promoter to corepress FOXC1 transcription in BLBC.²¹  Taken together, our findings support the diagnostic specificity and utility of FOXC1 in identifying the triple-negative phenotype based on IHC.

Notably, FOXC1 was rarely expressed in luminal-like subtypes. In luminal-like subtypes, patients with high FOXC1 expression were more likely to demonstrate low ER expression and negative PR expression. These findings were consistent with a previous study reporting that breast cancer patients with positive ER expression have higher methylation levels of the FOXC1 promoter, leading to lower levels of FOXC1 expression.²²  Another study has demonstrated that FOXC1 competes with GATA3 for binding to the ER gene and contributes to ER downregulation.²³  These studies provide possible mechanisms for the low expression of FOXC1 in the luminal-like subtypes, but the underlying mechanism of low FOXC1 in HER2⁺ subtypes has rarely been addressed.

Furthermore, FOXC1 expression has been reported to be positively associated with brain metastasis and lung metastasis as well as to be significantly correlated with poor outcome in patients with breast cancer.^8,24  FOXC1 expression has also been demonstrated to be correlated with sensitivity to olaparib in BRCA1 mutant cell lines.²⁵  In TNBC, we found that positive FOXC1 expression was associated with younger age and more aggressive biological behaviors, including high histologic grade and high Ki-67 expression, suggesting that FOXC1 expression is an unfavorable biomarker in TNBC. Unfortunately, the prognosis of the patients in the present study could not be analyzed; therefore, we were unable to confirm the prognostic value of FOXC1, which requires further investigation.

Morphologically, we found that positive FOXC1 expression was mainly observed in IBC-NST, metaplastic carcinoma, and AdCC. In addition, FOXC1 expression was significantly associated with the expression of the basal cytokeratin proteins CK14 and CK5/6. Other studies have defined TNBC with AR⁻, CD8⁻, and FOXC1⁺ expression as the basal-like immunosuppressed subtype.¹⁰  All these results support that FOXC1 is a useful and specific marker associated with the basal-like phenotype. Similar to FOXC1, SOX10 is another marker that is predominantly overexpressed in TNBC and is significantly associated with basal-like or myoepithelial differentiation.²⁶  It has also been demonstrated that SOX10 is a sensitive diagnostic marker for breast AdCC.²⁷  Given the increased use of SOX10 and FOXC1 in breast cancer, the correlation of SOX10 and FOXC1 expression in breast cancer should be tested in future studies to optimize the clinical utility of these markers. In contrast, we found that FOXC1 expression was negatively correlated with AR and GCDFP15 expression in TNBC. Correspondingly, FOXC1 was rarely expressed in carcinomas with apocrine differentiation, indicating that FOXC1 testing may be able to distinguish apocrine differentiation in patients with TNBC. Furthermore, the mRNA expression levels of FOXC1 were also significantly negatively correlated with AR and GCDFP15 mRNA expression in BLBC. These data implied that both FOXC1 mRNA and protein levels are useful markers in predicting the triple-negative phenotype. These results should be interpreted with caution owing to the limited number of special tumor subtypes in our study.

At present, FOXC1 overexpression has been reported in at least 16 types of cancer, including hepatocellular carcinoma, endometrial cancer, gastric cancer, and lymphoma.²⁸  Among these tumors, FOXC1 evaluation in breast cancer, specifically in BLBC, is the most studied, but whether FOXC1 can be used in the differential diagnosis of cancers of unknown origin has rarely been studied. In a meta-analysis on the likelihood of FOXC1 expression in early- and late-stage tumors, the results demonstrated that the expression rate of FOXC1 is similar in different cancers, including gastric carcinoma, non–small cell lung cancer, breast cancer, and hepatocellular carcinoma.²⁹  However, considering the high expression rate of FOXC1 in TNBC, it may play a role in the differential diagnosis when used in combination with other biomarkers. Therefore, it would be valuable to investigate whether FOXC1 staining could be used to distinguish TNBC metastasis from primary cancer.

In conclusion, FOXC1 is a relatively specific marker that is most frequently expressed in TNBC compared with other molecular subtypes. The cutoff value of 1% shows preferable diagnostic ability in predicting the triple-negative phenotype from other subtypes. In TNBC, FOXC1 is more likely to be expressed in certain histologic subtypes, most notably IBC-NST, metaplastic breast cancer, and AdCC, but is rarely expressed in invasive carcinoma with apocrine differentiation and in invasive lobular carcinoma. In addition, FOXC1 expression in TNBC is negatively correlated with the expression of apocrine-related markers (AR and GCDFP15) but is positively correlated with the expression of basal markers (CK5/6 and CK14), suggesting the concept that FOXC1 may serve as an additional diagnostic marker in the histologic and molecular subclassification of TNBC in future studies.

We are grateful to all patients who participated in this research.