Introduction

According to figures from Cancer Research UK, there were 45,695 cases of female breast cancer (FBC) and 277 cases of male breast cancer (MBC) diagnosed in the UK in 2007 [1]. In the US it was estimated that 1,970 men and 207,090 women would be diagnosed with breast cancer in 2010 [2]. Whilst MBC accounts for less than 1% of breast cancer diagnoses worldwide, the overall improvements in survival and mortality observed in FBC has not been seen to the same extent in MBC, as demonstrated in a recent interrogation of the Surveillance Epidemiology and End Results (SEER) database [3]. Moreover, the incidence rate of MBC is rising steadily [4–7].

The etiology of MBC is poorly understood with most of our current knowledge regarding its biology, natural history, and treatment extrapolated from FBC. Retrospective studies are generally weakened by the small numbers of cases available from any one centre with studies published on as few as 15 cases [8], making it hard to draw biologically meaningful conclusions. It is, therefore, a challenge to accrue sufficiently large numbers to allow comparative analysis of possible prognostic or predictive biomarkers. Many articles imply a general similarity of MBC to FBC and this has resulted in MBC patients being treated in exactly the same way as females in the clinic, which may not be optimal. Survival rates for MBC are generally assumed to be lower than FBC, probably as a result of later diagnosis and the assumption that treatments which are proven in FBC through clinical trials will have the same impact in men [9].

A 40-year review of records of 759 cases from invasive MBC from the US Armed Forces Institute of Pathology database showed that the frequency of histological subtypes in men was comparable to that of FBC, with the exception of papillary carcinoma which was twice more common in MBC [10]. To date, modern molecular subtyping has been reported in a single study of MBC where only luminal A (35/42) and luminal B (7/42) subtypes were observed [11].

Using a co-ordinated multi-centre approach, the aim of this study was to conduct the first large scale study to address and compare the expression profile of hormone receptors and their effect on survival in FBC and MBC.

Methods

Patient cohorts

Following ethical approval from the Leeds (West) Research Ethics Committee (06/Q1205/156), 514 formalin-fixed paraffin-embedded blocks of male (251) and matched female (263) breast cancers were obtained retrospectively. The latter were all from Europe and the former from Europe (n = 196) and Canada (n = 55). Informed consent was not required as the anonymised material pre-dated September 2006, came from a Tissue Bank approved by the UK Human Tissue Authority (or equivalent) or were from non-UK patients. Patients had not received any therapy before surgery. Details on adjuvant therapy were not extensively available; where available this was predominantly endocrine therapy (tamoxifen). Patient characteristics are presented in Table 1. Cases were reviewed by specialised breast consultant histopathologists (AMH, AMS, RAB) to confirm histology and marked up for assembly into tissue microarrays (TMAs) using 3 × 0.6 mm tissue cores per case taken from formalin-fixed paraffin-embedded material as previously described [10].

Table 1 Patient characteristics of male and female breast cancers
Full size table

Immunohistochemistry (IHC)

Antibodies, dilution, and retrieval methods are listed in Table 1. The antibody panel was focused on hormone receptors oestrogen receptor (ER)α, ERβ isoforms, progesterone receptor (PR) isoforms and androgen receptor (AR) and additional biomarkers selected to distinguish molecular subtypes of breast cancer (CK5/6, 14, 18, HER2). Each marker was run as a batch with appropriate positive (tissue known to express the biomarker of interest) and negative (no primary antibody) controls. Scoring was overseen by AMH, AMS, and RAB. Following visualisation of the signal with 3-3′diaminobenzidine chromogen, TMAs were digitised (Aperio Technologies), and hormone receptor immunoreactivity was scored using the Allred system with the following cut offs: ERα > 2, ERβ (and isoforms) > 3, AR > 2, PR (and isoforms) > 2, as validated in previous studies [12–14].

Hierachical clustering and principal components analysis (PCA)

For hierarchical cluster analysis, IHC measurements were used as inputs for all cases in each of the male and female cohorts. A Euclidian distance measure was employed with complete linkage of clusters. Clustering was conducted for the data structure of cases with the immunohistochemical parameters and for the immunohistochemical parameters as they were expressed through the population. Cluster dendrograms were plotted for both of these analyses for each gender and compared. PCA was applied to the same dataset. Analysis was based on covariances between parameters and cases in the data. Variances were computed based on the sum of squares/n − 1. Plots of the influences of variables in the factor plane of the first and second and the second and third principal components for both the male and female cases were plotted separately. As with the hierarchical clustering, both the male and female cases were combined into a single data set. The influence of variables in the factor plane of the combined cases were plotted for the first and second and the second and third principal components. The distributions of cases within the combined sets were also plotted.

Statistical analysis

Patient and disease characteristics were compared between male and FBCs using the χ 2 test (GraphPad). Associations with disease-free and overall survival (DFS and OS, respectively) were analysed by Kaplan–Meier plots and log rank test. P-values were two-sided, and P < 0.05 was considered significant.

Results

A total of 514 cases of breast carcinoma were studied, including 251 males and 263 females. The median age for the male cohort was 66 years (range 30–94) and 59 years (range 27–92) for females. Patient characteristics are shown in Table 2. As this was a matched cohort, no significant differences were observed in grade, or lymph node status between genders. Significant differences were observed in the distribution of histopathological subtypes (P < 0.0001). There was an even distribution of ductal phenotype whilst lobular carcinomas found in 9% of the female cohort was only seen in a single male case. Papillary and mucinous phenotypes were restricted to males. A significantly higher proportion of males expressed ERα compared to females (80 and 68%, respectively), although no differences in the frequency of PR was observed (71 and 72%, respectively). Follow up data was available on 183 (73%) male and 237 (90%) female cases.

Table 2 Details of antibodies used for immunohistochemical analysis
Full size table

Both cohorts were classified into molecular subtypes by IHC: luminal A (ERα+, and/or PR+, HER2−), luminal B (ERα+, and/or PR+, HER2+), HER2 (ERα, PR−, HER2+) and basal-like (ERα−, PR−, HER2−, CK5/6+) according to previous studies [15–17]. Representative immunoprofiles for each subgroup are shown in Fig. 1. Significant differences were observed between molecular subtypes (P = 0.0004). Luminal A was seen in 98% of males and 90% of females. Luminal B or HER2 subgroups were not observed in males but found in 6 and 2% of females, respectively. Basal-like tumours (ERα−, PR−, HER2−, CK5/6+) were infrequent in both cohorts (2% in each).

Fig. 1
figure 1

Semi-serial sections from male (a, c) and female (b, d) breast carcinoma TMAs showing immunoprofiles for each molecular subgroup. a luminal A (ERα+, PR+, HER2−), b luminal B (ERα+, PR+, HER2+), c basal (ERα−, PR−, HER2−, CK5/6+), d HER2 (ERα−, PR−, HER2+). Original magnification = ×10 (TMA core) and ×40 (insets)

Full size image

We then examined the frequencies of expression between genders of other hormone receptors including AR, nuclear and cytoplasmic ERβ1 and ERβ2, nuclear ERβ5, plus the PR isoforms A and B (Table 3). AR immunoreactivity was expressed in 64% of males and 93% females, respectively, (P < 0.0001). For ERβ1 and -2, both nuclear and cytoplasmic immunoreactivity was assessed [12]. ERβ1 nuclear immunoreactivity was significantly expressed in FBC whilst cytoplasmic ERβ1 and ERβ2 immunoreactivity were associated with MBC. No associations were observed for ERβ5. Of the PR isoforms, only PRA was significantly expressed in MBC. As the male cohort contained cases of European and Canadian origin, we tested if there were differences between these; none were found.

Table 3 Expression of AR, ERβ, and PR isoforms in male and female luminal A breast carcinoma
Full size table

In luminal A carcinomas, no differences in overall survival were observed between genders at either 5 or 10 years (Fig. 2a, b). This was also reflected in the unselected cohorts (data not shown). When hormone receptor expression was considered, only AR significantly associated with survival. AR-positive luminal A MBC had significantly improved overall survival over the equivalent FBC at 5 (P = 0.01, HR = 0.39, 95% CI = 0.26–0.87) but not 10 (P = 0.29, HR = 0.75, 95% CI = 0.46–1.26) years (Fig. 2c, d). In the unselected group, ERα and AR-positive MBC had significantly improved overall survival over ERα and AR-negative cases at both 5 (P = 0.04, HR = 0.37, 95% CI = 0.07–0.97) and 10 (P = 0.04, HR = 0.43, 95% CI = 0.12–0.97) years (Fig. 2e, f) with ERα and AR-positive MBC also having significantly improved overall survival over the equivalent FBCs at 5 (P = 0.05, HR = 0.48, 95% CI = 0.29–1.00) but not 10 (P = 0.37, HR = 0.79, 95% CI = 0.48–1.32) years (Fig. 2g, h).

Fig. 2
figure 2

Kaplan–Meier survival curves comparing survival between genders according to different tumour classifications. No gender-related differences were observed comparing OS of MBC (blue line) and FBC (pink line) at 5 (a) and 10 (b) years. Comparison of OS of Luminal A AR-positive MBC (blue line) and FBC (pink line) showed MBC had significantly improved OS at 5 (c) but not 10 (d) years. In unselected ERα-positive MBC those which were AR positive (black line) had significantly improved over those who were AR negative (orange line) at both 5 (e) and 10 (f) years. Comparison of OS of unselected ERα and AR-positive carcinomas showed MBC (blue line) had significantly improved survival over FBC (pink line) at 5 (g) but not 10 (h) years

Full size image

Hierarchical clustering based on hormone receptor profiles classified MBC and FBC into three distinct groups (Fig. 3). The cytoplasmic ERβ cluster was common to both genders. In FBC an ERα/PR cluster was observed, grouping ERα and PR isoforms, whilst ERβ isoforms clustered with AR (ERβ/AR cluster). In MBC, there were striking changes in the position of ERα; AR and ERα clustered with ERβ (ERα/β AR cluster) whilst PR isoforms formed an independent cluster (PR cluster). This was also reflected in a PCA-based plot of variable factor co-ordinates (data not shown).

Fig. 3
figure 3

Hierarchical cluster analysis showing relatedness of hormone receptor profiles between genders. Three clusters were observed in each gender (dotted lines). The cERβ cluster was common to both but there was a shift in the position of ERα in MBC

Full size image

Discussion

Currently, MBC is treated based on the assumption that it is essentially the same disease as FBC. In this the largest comparative study to date, directly comparing the immunohistochemical profile of matched MBCs and FBCs has revealed that whilst superficially there is similarity between genders, when probed more deeply, subtle differences are uncovered.

The histological breakdown of our cohort is in line with previous reports with the papillary phenotype and variants thereof, which are twice more common in males, only seen in the male cohort [10]. Although we did not observe any mucinous carcinomas in our female cohort the expected frequency of this phenotype is only 0.9% [18]. We observed only a single case of lobular carcinoma in males; this is to be expected given the rarity of this phenotype in men [19]. Thus, our cohort can be regarded as representative.

We used IHC to classify our breast tumours into molecular subtypes. Whilst gene array analysis is still considered the ‘gold standard’, we were unable to apply this to our 514-case cohort due to high cost. Nevertheless, molecular profiling of breast cancer based on immunohistochemical typing has now gained widespread acceptance as a surrogate method and is arguably more robust as it overcomes the limitations of gene array in that it directly identifies the cells expressing the marker of interest [15–17, 20]. When applied to TMAs, a limitation is core loss, which we experienced in this study and which may have impacted on the higher than anticipated levels of luminal A phenotype in the female population. Nevertheless, luminal A phenotype was the most common in both sexes with basal-like tumours infrequent in both. Sporadic expression of basal cytokeratins has been previously reported in MBC [21]. Interestingly, the luminal B phenotype was not seen in males. This contrasts a recent study of 42 MBCs where luminal B was seen in 17% of cases [11]. The lack of luminal B carcinomas in our male cohort reflects the absence of HER2 expression, which has been variably reported in MBC ranging from 0 to 95% [22]. It is notable that some of the earlier studies on HER2 relied solely on IHC to determine positivity, considering any degree of membrane immunoreactivity positive [8, 22–28]. The validity of studies relying on HER2 IHC without recognition of gene expression is questionable. According to ASCO/ACP and NEQAS guidelines only those scoring 3+ or above are considered HER2 positive. Equivocal cases are scored 2+ and go forward for FISH analysis and only those with HER2 gene amplification are considered positive. In studies using both IHC and FISH to detect HER2, protein expression was always higher than gene amplification [29–31]. Although we relied on IHC to detect HER2, we are confident our data is robust; we are a regional HER2 testing centre and our HER2 IHC was conducted via this service using two different antibodies and running the test according to clinical standards. Whilst we did not observe any cases which scored >2+ in our MBC cohort, we observed scores of 1+ in 22/251 cases (8%). These would be considered negative in clinical practice. Nevertheless, this provides confidence that the high frequency of HER2 negativity we observed was not simply due to antigen degradation in archival material.

It has been suggested that separation of luminal A and luminal B breast tumours should be based on the expression of proliferation markers such as Ki67, not on HER2 expression as is currently the case [32]. However, this has yet to gain widespread acceptance and we believe there are several issues that still need to be standardised before this can be implemented. These include choice of Ki67 antibody, e.g. MIB1 or SP6 [33] and how to optimally distinguish between low and high proliferation scores. Once these issues are resolved it will be interesting to determine if the difference between MBC and FBC in this series is purely the result of lack of HER2 expression, or if MBCs have a lower proliferative index as well. Contrary to the general impression, one of the most significant findings from this, the most authoritative study to date, was the observation of no differences in overall survival at 5 or 10 years between genders in either our unselected or the luminal A cohorts. Whilst there have been a number of case–control and population-based studies addressing survival in MBC using data from cancer registries [3, 4, 34, 35], direct comparative studies between genders are scarce. A Chinese study of 35 MBC and 70 matched FBC showed the latter had significantly better overall survival at 5 and 10 years, but when the comparison was restricted to female postmenopausal, outcomes were similar [36]. Despite a small number of cases in the male arm and unbalanced cohort size, a Japanese study of 14 MBC and 140 FBC showed no difference in overall survival [37]. This was also reflected in a UK study comparing outcome in 41 MBC and 123 FBC which showed that when matched for key prognostic factors (size, grade and lymph node status), outcome was similar between genders [38], agreeing with our study. Of note was the observation that when MBC was compared with an unselected FBC group, males had worse outcome [38], which may explain some of the earlier studies inferring a worse prognosis in men [39, 40]. Gender comparative information obtained from 1988 to 2003 SEER data showed worse breast cancer-specific survival in males diagnosed with stage I disease; however, the authors attributed this to in-stage migration rather than being of clinical relevance [41].

A common finding in MBC is the higher frequency of hormone receptor expression in particular ERα (reviewed in [42]), which was also reflected in this study. We have explored this further using hierarchical clustering where one of the striking features was the occurrence of ERα on distinct clusters in males and females. In FBC, ERα clustered with PR and its isoforms; as PR is oestrogen-regulated [43], this is unsurprising. In MBC, the position of ERα changed, clustering instead with ERβ isoforms and AR. The potential role of ERβ in breast cancer has been the subject of much debate and recent work in FBC by us and others shows this depends on the cell location and the type of isoforms expressed [9, 44]. However, ERβ is subject to complex regulation involving 3′UTRs [45] and microRNAs [46], and we need to further understand its biology before speculating on any role it may play in MBC.

Regarding the potential relationship of AR with ERα, at a functional level, AR transfection into ERα-positive breast cancer cells inhibited ERα transactivation and oestrogen-stimulated growth through interaction with oestrogen response elements [47]. Other work indicates that oestrogen activation via ERα and ERβ can mediate AR signalling [48]. Given the recognised pro-proliferative effects of ERα and PR and the anti-proliferative effects of ERβ and AR [49] this suggests the coordinated expression of these receptors could influence survival. This was demonstrated in the current study where AR-positive luminal A MBC had improved overall survival than the equivalent FBC and was also borne out in the unselected group. This supports other recent work showing AR is an important prognostic factor in ERα-positive FBC [12]. Previous studies examining the impact of AR in MBC have been compounded by the high expression frequency of AR in some studies [22], the low number of cases in others [50], or combinations of both [51, 52] precluding meaningful survival analysis. Studies examining the effect of AR on survival in MBC are contradictory, possibly as a result of limited numbers available from single centre studies. In a series of 43 MBC Kwiatowska et al. [53] showed AR expression correlated with reduced survival. A similar-sized study (n = 47) showed no effects of AR expression in MBC on survival, however, it is noteworthy that in the study, MIB-1 expression, which detects proliferating cells, was higher in AR negative compared to AR-positive cases [26]. The association of AR positivity with better outcome in MBC in this study indicates this is a potentially important prognostic factor, paralleling observations in FBC where the potential prognostic role for AR in FBC is receiving increased attention [54] following the observation that AR is an independent prognostic marker in a large series of FBCs (n = 953; [12]). Moreover, AR expression in MBC could turn out to have both prognostic and predictive value; its presence suggests that anti-androgen therapy could be explored as a therapeutic approach. Androgen blockade is commonly used in prostate cancer treatment but so far remains inadequately tested in breast cancer [55]. The complete absence of cell line models derived from MBCs presents a challenge in being able to model this in vitro.

To our knowledge, this is the largest retrospective biomarker study directly comparing matched male and female breast carcinoma; however, we acknowledge there are some limitations. Despite our best efforts, these include lack of availability of follow up data and missing/incomplete clinicopathological data. Of note is the absence of data on germline mutations, particularly BRCA2 which is involved in MBC development and associated with reduced survival [53, 56, 57]. Nevertheless, our study has confirmed that whilst superficially similar to FBC, when studied more rigorously MBC is biologically different, echoing a hypothesis proposed by the Multidisciplinary Meeting on MBC [58] and supporting recent work at the transcriptional [59], microRNA [60, 61] and genomic [62] levels. With the incidence of MBC rising [4–7], collectively these studies provide a strong impetus for further study of this rare cancer type.