Introduction

The role of estrogen receptor (ER) and the progesterone receptor (PR) in human breast carcinomas is well established, but little is known about the significance of androgen receptor (AR). The androgen receptor is expressed in 70% to 90% of invasive breast cancers: a frequency comparable to, or higher than, that reported for ER (60–80%) and PR (50–70%) [1–11]. Although ER, PR, and AR are frequently co-expressed in breast cancers, a considerable proportion of AR-positive carcinomas, ranging from 9% to 50%, are negative for both ER and PR [2, 3, 5, 7, 9, 10, 12–15]. In addition, AR was identified as the only sex steroid receptor expressed in approximately 25% of breast cancer metastases [16].

The influence of AR expression on survival of breast cancer patients is unclear. AR seemed to predict a response to hormone therapy and a better overall survival, but AR expression was not predictive of disease-free survival in ER-positive neoplasms, while ER status was an independent prognostic factor for disease-free survival [5, 13, 17]. For ER-negative tumors, AR positivity was associated with significantly better disease-free survival and with better prognosis for metastatic disease [8] and with significantly better survival after disease recurrence compared with AR-negative tumors [18].

Gene expression profiles allow breast carcinoma (BC) classification into five main groups, two of them ER+ (luminal A and luminal B) and three ER− (human epidermal growth factor receptor 2 (HER)+, basal like, and normal breast like). The main immunophenotype of the basal-like group as defined by gene expression analysis is the lack of expression of ER, PR, and HER2 and the expression of basal-associated markers such as basal cytokeratins (CK5, CK17, and CK14) and EGFR. These are currently used as surrogate immunohistochemical markers for the identification of basal-like BCs in routine practice [15, 19–23]. The normal-like gene expression pattern was typified by the high expression of gene characteristic of adipose tissue and other non-epithelial cell types. This molecular subtype is not easily defined by immunohistochemical surrogates (ER−, PR−, HER2−, CK5/6−), and thus these tumors were designated as an “unclassified group.” Farmer et al. [24] have shown that AR expression level divides ER-negative tumors into two major gene expression clusters, including ER−/AR− (basal) and ER−/AR+ (molecular apocrine subtype).

HER2 is a well-known prognostic and predictive parameter [25] and represents a biologically and clinically important feature. HER2 protein overexpression has been detected in 20–30% of invasive breast carcinoma [26, 27], and it has been found both in ER+ and in ER− breast cancer [28]. Some studies have investigated the relationship between AR and HER2 immunohistochemical status in breast carcinomas with conflicting results. Narita et al. [11], in agreement with other workers [6, 14], did not find any significant correlation, but other studies reported HER2 overexpression in grade 3, AR-positive carcinomas [8, 9, 29]. In addition, a functionally significant crosstalk between androgen receptor and the HER2 pathway has been identified in ER-negative breast cancer [30].

The prevalence and role of AR expression in triple-negative ER−/PR−/HER2 have not been clearly elucidated. Rakha et al. [31], investigating 282 triple-negative consecutive invasive breast cancers, showed AR expression in 13% of the cases.

The aim of this study was to evaluate the immunohistochemical expression of AR in a series of 232 ER− and PR− breast cancer and to correlate it with HER2 expression. In addition, the AR expression has been correlated with clinicopathological parameters, histopathological features, and follow-up data.

Materials and methods

Histological study

A total of 232 cases of ER- and PR-negative invasive BCs were selected among 1,301 consecutive invasive ductal BCs collected at the Department of Pathology, Ospedale di Circolo, and University of Insubria, Varese, Italy, from 1995 to 2000. Criteria for selection included the following: the cases were primary breast cancer, the histological type was ductal invasive carcinoma, NOS, excluding the apocrine and adenoid-cystic cancer types, and the immunohistochemical expression of ER and PR was considered negative (carcinomas were completely negative in 208 cases or present in less than 10% of tumor cells in 24 cases included in our series because clinical treatment followed guidelines of completely negative BCs). A complete and intense membrane staining for HER2 in >10% of tumor cells was considered as positive (94 cases). Of these BCs, 89 cases showed HER2 positivity in >30% of tumor cells (score 3+) while the remaining five cases were scored 2+. No FISH test for HER2 amplification was done for these cases. The tumor ER, PR, and HER2 status was available from pathology reports. Clinicopathological information concerning age, tumor site, size, and local extension were obtained from patient records. The tumor pathological stage was assessed according to the TNM sixth edition recommendations [32].

Formalin-fixed paraffin-embedded sections of tissues obtained from surgical samples (226 cases) and trucut core biopsies (six cases) were routinely stained with hematoxylin and eosin (H&E) for morphological assessment. Microscopic slides of all cases were reviewed by two senior pathologists (CR and FS) in order to assess the histological type, according to the World Health Organization’s recommendations [33], and the grade, applying the Elston–Ellis criteria [34].

Immunohistochemical study

For each case, immunohistochemical expression of AR was evaluated along with p53 and proliferative index (Ki-67, antibody MIB-1). In addition, basal and luminal differentiation of breast carcinomas was investigated using antibodies against CK5, CK14, CK8, and CK18 [35, 36]. In all cases, 4-μm-thick paraffin sections were deparaffinized in bioclear and rehydrated through a graded series of ethanol. Endogenous peroxidase activity was inhibited with 10% hydrogen peroxide (H2O2) in distilled water for 10 min. Antigen retrieval was provided using microwave for 10 min in 0.01 M sodium citrate buffer pH 6 and protease digestion. Sections were incubated overnight at 4°C with the primary antibodies and subsequently for 1 h with a 1:200 dilution of biotin-labeled secondary anti-mouse or anti-rabbit antibody followed by ABC complex (avidin, 1:100, and biotin-labeled peroxidase, 1:100) (Dako). Sections were then stained with a solution of 3,3′diaminobenzidine (Sigma) with H2O2 in Tris–HCl buffer, pH 7.4, lightly counterstained with Mayer’s hematoxylin, dehydrated, and mounted. (The source and dilution of primary antibodies used are listed in Table 1.)

Table 1 Antibodies used for the immunohistochemical study
Full size table

In all cases, immunoreactivity was semiquantitatively evaluated and recorded as the percentage of positive cells. BCs were classified as basal like when at least 5% of cancer cells were immunoreactive for CK5 and/or CK14, independent of the intensity of the immunostaining [35, 36]. For AR, we considered nuclear labeling in more than 10% of neoplastic cells as the cutoff point for positivity, similar to standardized criteria used for other steroid hormone receptors [9].

Positive controls included normal breast tissue surrounding the neoplastic lesion (for all antibodies) and normal prostatic tissue (for AR). Negative controls included the omission of primary antibodies. All the immunoreactions were separately evaluated by two senior pathologists.

Statistical analysis

The statistical analysis was performed using the χ 2 test with Yates’ correction (SPSS 7.5 software). The survival analysis was performed by employing the Kaplan–Meier product limit estimate of probability of survival against time, producing a product limit survival curve for each of the following variables: age, stage, presence of lymph node metastases, and immunohistochemical expression of basal (CK5, CK14) and luminal (CK8, CK18) immunophenotype and AR.

Multivariate analysis was carried out using the Cox proportional hazard model. Variables with p < 0.15 in the univariate analysis were entered into the multivariate analysis. For all the survival analyses, the Survan XL program, version 1.14 (copyright © 1995–1997), was employed.

Results

Clinicopathological data and immunoprofile

The mean age of the 232 patients at diagnosis was 58.7 years (range 24–97 years). The most frequent location of tumors was the upper-external quadrant (37%), and the diameter ranged from 0.3 to 10.5 cm (mean 2.37 cm). The pathological stage at diagnosis was known in 224 patients: 112 tumors were in stage pT1 (50%), 91 in pT2 (40.6%), five in pT3 (2.2%), and 16 in pT4 (7.1%). Lymph node status was available for 175 cases in which axillary dissection was performed, and lymph node involvement was present in 83 cases (47.4%). The series of 232 cases included 216 (93.1%) poorly differentiated (G3) and 16 (6.8%) moderately differentiated (G2) ductal BCs.

The mean percentage of Ki-67-positive nuclei was 55.7% (range 14–95). Overexpression of p53 in more than 10% of tumor cells was present in 168/231 (72.7%) cases with a mean percentage of p53-positive nuclei of 63.6%. The comprehensive results of the immunohistochemical study are reported in Table 2. According to luminal and basal cytokeratin expression, tumors were classified into two groups: 100 (44.2%) luminal-type BCs, expressing only CK8 and/or CK18, and 126 (55.8%) basal-like-type BCs positive for CK5 and/or CK14 in more than 5% of cancer cells, often also showing luminal cytokeratin expression; six cases failed. In addition, according to the expression of HER2, 137 (59.3%) cases were considered as triple-negative BCs (when ER, PR, and HER2 were negative) (Fig. 1) and 94 (40.7%) as HER2-positive BCs; one case failed.

Table 2 Clinicopathological profile of ER- and PR-negative breast carcinoma
Full size table
Fig. 1
figure 1

Grade 3 triple-negative BC (a H&E staining) showing negative ER (b), HER2 (c), and AR (d) immunostaining

Full size image

AR immunohistochemistry

AR expression was evaluable in 226 cases, and AR positivity was observed in 128 BCs (56.6%). One hundred and ten cases (48.8%) had >10% positive cells, with an average of 59% of AR nuclear immunoreactivity (range 10–100%), and 18 (7.8%) showed <10% of positively stained cells, and 98 tumors (43.4%) were AR negative (Table 3).

Table 3 Expression of AR in 226 ER- and PR-negative breast cancer
Full size table

Comparison of AR expression with other tumor clinicopathological features is shown in Table 4. While the pT did not show any statistically significant difference between T1 and T2–T3–T4 (45.8% of cases versus 51.8%; p = 0.38), the correlation of lymph node status with AR expression showed statistically significant results (p < 0.034). In fact, axillary lymph node metastases were present in 63.4% (52/82) of patients with AR-positive tumors, compared with 36.6% (30/82) of those with AR-negative lesions. On the contrary, 41 out 89 (46.1%) pN0 patients were AR positive and 48 out 89 (53.9%) were AR negative. Regarding tumor grade, AR positivity was seen in eight (50%) out of 16 G2 cases and in 102 (48.6%) out of 210 G3 cases.

Table 4 Correlation between AR immunoexpression and other clinical and histopathological features
Full size table

The basal-like-type BCs were AR positive in 31.2% (39/125) of cases, while in the luminal group 71.1% of cases (69/97) were positive, showing highly significant correlation (p < 10−8). Regarding AR immunoexpression and HER2 status, statistical analysis showed a strongly significant correlation: 76.7% (69/90) of HER2 positive BCs cases were AR positive (Fig. 2) compared with only 30.4% (41/135) of triple-negative BC types (p < 10−10).

Fig. 2
figure 2

Grade 3 BC (a H&E staining) showing negative ER immunostaining (b) and strong AR immunostaining (c) together with HER2, score 3 positivity (d)

Full size image

Survival analysis

Information about overall and disease-free survival was available for 224 patients: 138 (61.6%) patients were alive after a median follow-up time of 116.4 months (range 83–173 months) while 86 (38.4%) out of 224 were dead of disease after a median follow-up time of 36.3 months (range 2–99 months). Regarding clinical evolution, 130 of the 138 alive patients (94.2%) were disease free after a median follow-up time of 116.2 months (range 83–173 months), while only eight (5.8%) experienced disease relapse after a median follow-up time of 68.3 months (range 6–119 months). For the group of the 86 dead women, 54 (62.8%) patients died with disease after a median follow-up of 33.63 months (range 2–99 months) while 32 (37.2%) developed a disease relapse after a median of 34.3 months (range 0–71 months).

No relationship between overall survival and AR expression was found (p = 0.16). In fact, among the still-living patients, 45.3% (62/137) harbored AR-positive BCs and 54.7% (75/137) harbored AR-negative BCs, while among the 84 deceased patients, 47 (56%) harbored AR-positive BCs and 37 (44%) harbored AR-negative ones.

Also, there was no significant correlation between AR status and disease-free survival (p = 0.44) since 22 (56.4%) of 39 patients with disease relapse harbored AR-positive BCs and 17 (43.6%) harbored AR-negative BCs. On the other hand, of the 183 patients with no evidence of distant metastasis, 86 (46.9%) women harbored AR-positive BCs while 93 (50.8%) harbored AR-negative BCs.

Survival analysis has been performed to determine a possible relationship between patient outcome and age at diagnosis, stage, presence of lymph node metastases, and immunohistochemical expression of basal (CK5, CK14) and luminal (CK8, CK18) markers and AR. Factors influencing survival in the univariate analysis were age at diagnosis (the probability of survival after a follow-up time of 144 months was 87.4% for patients ≤40 years old vs 57.5% for patients >40 years old, p = 0.006), pT stage (median survival time of 100 months for T1 vs 84 months for T2–T3–T4, p < 0.0001), and axillary lymph node status (median survival time of 101 months for lymph-node-negative tumors vs 70 months for lymph-node-positive tumors, p < 0.0001) (Figs. 3 and 4).

Fig. 3
figure 3

Kaplan–Meier survival curves illustrating disease-free survival for patients stratified by AR status and AR/HER2 status. No significant differences was observed

Full size image
Fig. 4
figure 4

Kaplan–Meier survival curves illustrating overall survival for patients stratified by age at diagnosis (a), pT stage (b), and lymph node status (b). Significantly worse survival rates were noticed for patients with tumors occurred after the age of 40 years, advanced pT stage and presence of lymph node metastases

Full size image

Multivariate analysis was performed using the Cox proportional hazard model including the three variables which were statistically significant in the univariate analysis. Results showed that stage and lymph node status are independent prognostic factors (p = 0.0261 and p = 0.0017 respectively, Table 5).

Table 5 Results of multivariate analysis
Full size table

Discussion

AR as a target for the treatment of hormone-receptor-negative breast cancer is an uncharacterized territory and targeting AR with AR androgen-based hormonal therapy in ER and PR breast cancer should be taken into consideration as potential interactions between AR and other tumor-growth-related factors [37]. Recently, there has been renewed interest in alternative hormonal treatment, including progestin medroxyprogesterone acetate, for use both in early disease and in the advanced setting when conventional therapies fail [38–40]. These studies suggest AR as a marker for the efficiency of endocrine therapy and for new hormonal therapeutic strategies in women with estrogen-negative cancer.

Other studies have examined the significance of AR in breast carcinomas, usually in comparison with ER and PR status, but few immunohistochemical studies have investigated the expression of AR as the only sex steroid receptor expressed in ductal invasive breast carcinomas. In this context, our study focused on a large series of ER- and PR-negative ductal carcinomas, since most ER/PR-positive tumors are AR positive. It is clear from this study that AR expression is a frequent feature of ER- and PR-negative ductal breast carcinomas. Our results show a 56.6% incidence of AR positivity in ER- and PR-negative ductal invasive breast carcinomas, which is similar to the earlier reported percentage in previous investigations [8, 9, 41].

As is commonly expected in ER- and PR-negative ductal breast carcinomas, in the present investigation, almost all tumors (93.1%) were G3, showing a higher percentage of G3 cases than those (74%) reported by Scawn and Shousha [41]. AR positivity has been shown in 48.6% of poorly differentiated carcinomas (G3), in agreement with previous immunohistochemical studies examining a smaller series of cases of ER/PR-negative breast carcinomas [5, 8, 9, 11, 14].

Bryan et al. [17] did not find a significant correlation between AR and menopausal status in their patients, in contrast with Agoff and colleagues [8] who found that AR expression correlated with both increasing age and menopausal status. For our study, no information was available about menopausal status of the patients, but we could not find any tumor correlation of AR expression of BCs and the age of the patients.

Regarding the nodal status, which represents a reliable prognostic factor in breast carcinoma, there was a significant correlation between lymph node status and overall survival. In our study, we found that the percentage of axillary lymph node disease was 47.4%, in agreement with other investigations [41, 42] but higher than the general incidence of 20–30% commonly observed in the breast cancer [43]. One notable finding of the current study was the positive association between AR expression and lymph node metastases, in contrast with the results of other studies [6, 11]. On the other hand, we did not find a correlation of AR expression with the pT stage of the tumors, in agreement with recent studies [10].

An interesting finding of our study was the significant correlation between AR expression and HER2 status. This result is in agreement with the observations of other workers that have detected HER2 overexpression in G3-AR-positive carcinomas [8, 9, 29] but is in contrast with some other studies [6, 11, 14]. These data suggest that AR can be involved in the pathogenesis of breast carcinomas, indicating that HER2 may control the transcriptional activity of AR, the latter regulating its own transcription. The interaction of AR with the HER2-related pathway has been investigated in human prostate cancer cell lines, suggesting a crosstalk between AR and HER2 pathway and demonstrating that HER2 induces AR transactivation through the MAP kinase pathway [44–48]. These findings have potential clinical relevance since anti-androgens did not inhibit AR in the presence of HER2.

Doane et al. [28] point out the heterogeneity of AR expression as well as the complexity of AR signaling in breast cancer, showing that androgens inhibited the growth of some cell lines but stimulated the growth of other cell lines. Because of the clinical heterogeneity of ER-negative, PR-negative tumors, Doane et al. [28] speculated that molecular subsets exist in ER- and PR-negative tumors and postulated that AR may act in concert with other signal transduction pathway, and it is known that receptor tyrosine kinase pathway functions as a modulator of nuclear hormone receptor activity [49]. Very recently, similar results have been reported by Niemeier et al. [50] who, studying the AR expression in 189 consecutive BCs, outlined the close association between AR positivity and HER2 status in five out of eight HER2 BCs while Park et al. [51] reported the AR positivity in 32 out of 43 HER2 BCs. The authors suggest a potential target role for AR both in AR+ HER2 BC molecular class and in AR+ triple-negative BCs.

Carcinoma of the breast is recognized to be a heterogeneous disease consisting of phenotypically diverse population of cancer cells. Previous studies have demonstrated that BCs with basal-like phenotype is a special group of tumors that appear to be predominantly grade 3 ER and PR negative [52] and is characterized by a specific immunohistochemical profile and poor prognosis [51–53]. In our study, basal-like carcinomas represented more than half of the cases (55.8%), and they expressed AR receptors less frequently (31.2% of the cases) than luminal carcinomas (71.1% of the cases), in agreement with other investigations [15, 31, 54].

The prognostic significance of AR expression in ER- and PR-negative breast cancer has been poorly investigated, while most studies have emphasized the prognostic value of AR status in ER+/PR+ BCs [5, 7, 13, 14, 17]. Our study focused on ER- and PR-negative tumors and did not find any association between AR expression with both the overall survival and the disease-free survival (Fig. 3). Our findings are only partially in agreement with previous investigations [8] that have found an association between AR expression and disease-free survival in the subset of patients with ER-negative BCs in univariate analysis but not in multivariate.

Our investigation has shown that, with multivariate analysis, only lymph node status and tumor size were independent prognostic variables, while AR status was not a significant predictor of disease-free survival and overall survival as well as the luminal, basal-like, and HER2 immunohistochemical categories. These results may be explained by the fact that our study was concentrated on ER- and PR-negative tumors, which tend to be high-grade ductal carcinomas with more advanced stage at diagnosis and a high rate of proliferation. On the other hand, this study was a retrospective investigation, and our survival curves may reflect innate characteristics of the tumors since that treatment has not been influenced by AR expression, and all patients had received standard therapies. In fact, the therapeutic approach to BCs in the middle of the 1990s was non-tailored as suggested by the more recent treatment guidelines. Although AR is expressed in almost 50% of ER-negative tumors, the functional role of AR in BC remains unclear; further exploration of this area could expand the repertoire of potential treatment for patient with ER-negative breast cancer [55].

We can conclude that AR is frequently expressed in ER- and PR-negative poorly differentiated (G3) BCs and that expression of HER2 and AR is highly correlated (p < 0.005). In addition, a notable finding was that the majority of AR-positive BCs were lymph node positive. The frequent expression of AR and common coexpression of AR and HER2 oncoprotein in ER- and PR-negative BCs raise the important question of the role of AR and HER2 in the pathogenesis and treatment of ER- and PR-negative BCs. Further investigations of AR in ER- and PR-negative BCs could provide more information concerning the involvement of AR in the pathogenesis of breast cancer and the potential role in clinical management of women with ER- and PR-negative breast carcinomas.