The traditional histological classification of invasive breast cancer tumours has been debated by results from gene expression arrays leading to the molecular categorisation of breast cancer into luminal and nonluminal tumours (Perou et al, 2000
). As invasive ductal breast cancers develop via the noninvasive precursor DCIS, these lesions may be categorised in the same way. Most similarities were found among the different stages of breast tumour progression, and it was suggested that gene expression alterations conferring the potential for invasive growth are already present in the preinvasive stadium of breast cancer (Ma et al, 2003
). It has been shown that the molecular subgroups of invasive carcinoma can be distinguished using a set of IHC markers (Makretsov et al, 2004
). In addition, response to treatment is subgroup dependent (Carey et al, 2007
). Using IHC, our analysis focused on the identification of subgroups in pure DCIS, to improve insight into the different pathways of tumour development, and to produce a classification of DCIS based on marker expression. Sixteen markers were selected to distinguish luminal and nonluminal cell differentiation, or because of their reported value to differentiate DCIS. Unsupervised hierarchical analysis, like in the evaluation of gene expression arrays, was performed to categorise DCIS. Ten markers were either positive or negative in nearly all DCISs and therefore not useful for classification.
Oestrogen receptor, PR, AR, Her2, Bcl-2, and p53 were used for IHC classification of DCIS. Oestrogen receptor, PR, and AR were positive in 68, 46 and 37% of the patients in our series, respectively. Others found ER, PR, and AR expression in 54–73, 49–61, and 33–44%, respectively (Selim et al, 2002
; Baqai and Shousha, 2003
; Barnes et al, 2005
; Rody et al, 2005
). We further found that well- and intermediately differentiated DCISs were predominantly ER positive and PR positive, while poorly differentiated DCIS usually lacked steroid receptor expression and was correlated with Her2 overexpression. This finding became further evident by the unsupervised hierarchical clustering results that clearly divided DCIS into luminal and nonluminal lesions. The luminal-type DCIS was further divided into an AR-positive and AR-negative subtype.
Bcl-2, involved in apoptosis, was present in 64% of all DCISs, while p53 was expressed in 26% of the cases in our series. These findings are in correspondence with results from others who reported Bcl-2 and p53 expression in 76 and 24% of DCIS cases, respectively (Siziopikou et al, 1996
). The Bcl-2-positive/p53-negative phenotype, which is similar to normal epithelium and benign lesions, was observed in 95 cases originating from the luminal clusters. This phenotype might reflect a more favourable group of lesions.
Androgen receptor expression was most frequently seen in intermediately and well-differentiated DCIS (P
=0.018) in our series of patients. Not many studies investigated AR in DCIS. Moinfar et al (2003)
reported a higher rate of AR expression in especially low-grade DCIS as opposed to high-grade DCIS, although others did not find a correlation between AR expression and grade (Selim et al, 2002
). Androgen receptor-positive breast cancer patients have prolonged survival and a better response to hormonal treatment than AR-negative patients.
Within the nonluminal type, a Her2-positive/ER-negative subtype with 91% poorly differentiated DCISs could be identified. Her2 is known to be amplified and/or overexpressed in invasive breast cancer in 10–30% of cases and associated with poor outcome (Ravdin and Chamness, 1995
; Tsuda, 2001
). The absence of Her2 overexpression in normal ducts and atypical ductal hyperplasia, and the frequent of Her2 amplification found in DCIS suggests that Her2 alterations are an early event in the pathway of development of Her2-positive invasive carcinomas. In our study, 39% of the cases were positive for Her2. The higher frequency of Her2-positive lesions in DCIS compared with invasive breast cancer has been argued to occur due to loss of expression; however, it might indicate that in the breast cancer progression model, there may be lesions that do not frequently evolve into invasive breast cancers, including Her2-positive DCIS lesions. Moreover, the mammographic detection of poorly differentiated Her2-positive DCIS often occurs at an early stage due to the conspicuous microcalcifications.
Basal-like carcinomas have been identified in gene expression profiling studies as a subtype of invasive breast cancer. These lesions are ER negative, PR negative, and Her2 negative (triple negative). We found eight (5%) triple-negative lesions. Four of them were poorly differentiated. Bryan et al (2006)
studied 66 cases of high nuclear grade DCIS to determine the frequency of the triple-negative phenotype and showed that only four cases (6%) exhibited the triple-negative phenotype. In contrast with our results, they found EGFR expressed in all four triple-negative lesions and also in a selection of nontriple-negative lesions, while we found negative staining in all lesions. In addition, they observed more frequent expression of keratins 5/6 and 14 compared with our series. This could be a result of the interobserver variability, since our cutoff point was more than 10% strong membraneous staining, while Bryan et al (2006)
considered any staining as positive. Recently, (Livasy et al (2007)
found 8% basal-like subtypes in a population-based series of 245 patients. Given that invasive breast cancers typically share immunophenotypic features with the DCIS lesion from which they arise, these findings corroborate the possibility that the triple-negative DCIS lesions represent a precursor lesion to invasive basal-like carcinomas. In these (medullary-like and metaplastic) carcinomas, in situ
components are usually minor or absent, suggesting a rapid progression from in situ
to the invasive stage. This is in keeping with the absence of basal-like in situ
lesions in preventive mastectomy specimens of BRCA1 carriers, which are prone to develop basal-like tumours (Hoogerbrugge et al, 2003
Clustering analysis showed that the well-differentiated DCIS and intermediately differentiated DCIS share IHC features among the different clusters. It seems that intermediately differentiated DCIS shows more resemblance with well-differentiated DCIS as compared with poorly differentiated DCIS. A recent study from our institute investigating classification of DCIS by gene expression profiling confirms this finding and identified luminal, Her2-, and basal-like tumours in a series of 40 DCIS lesions (Hannemann et al, 2006
). A classification of DCIS by IHC might identify identical groups of luminal and nonluminal tumours, which can be further subdivided, reflecting the heterogeneous nature of DCIS. Therefore, IHC can assist in objectivation of variations in morphologic tumour classification of DCIS.