A great challenge in breast cancer management is to identify patients who will not benefit from systemic adjuvant chemotherapy. Current approaches are hindered by a dearth of clinically useful biomarkers, with the exception of ER and possibly HER2/neu. Although promising, the readiness of gene expression signatures to further stratify patients awaits the results of prospective studies. In contrast, immunohistochemical analysis of BCL2 protein expression is a simple, well-validated, inexpensive and widely available test (used routinely in diagnostic pathology of low-grade lymphoproliferative disorders). The prospective analysis in over 11
000 women with early-stage breast cancer reported here demonstrates for the first time the robust prognostic significance of BCL2 protein expression independent of ER, as well as all the other traditional prognostic markers used in clinical practice (see ) (Hayes et al, 1996
Significant temporal variation exists in the relative contribution of individual prognostic markers during prolonged follow-up, further limiting their clinical value. Our analysis confirms the previously described time dependence of hormone receptors (ER and PR) as prognostic markers in breast cancer (Hilsenbeck et al, 1998
; Anderson et al, 2006
). This time dependence has also been noted in gene expression signatures that appear to be better predictors of relapse in the first 5 years (Desmedt et al, 2007
; Cardoso et al, 2008
). In contrast, our study reveals that the prognostic effect of BCL2 protein expression is time independent and BCL2 continues to be associated with favourable outcome over time, increasing its potential clinical value given the frequent occurrence of late relapses (particularly in ER+ breast cancer).
BCL2 belongs to a group of related proteins that are key regulators of apoptosis or programmed cell death (Cory et al, 2003
). The tumourigenic potential of inappropriate BCL2 protein expression was first described as a result of the chromosomal translocation (t
(14,18)) seen in subsets of non-Hodgkin's lymphoma, in which it is associated with adverse outcome (Tsujimoto et al, 1984
). Since this discovery, overexpression of BCL2 protein has been identified in a variety of solid organ malignancies, including breast cancer. In contrast to non-Hodgkin's lymphoma, BCL2 protein expression in breast cancer is associated with an indolent phenotype of low-grade, slowly proliferating, ER+ breast tumours (Silvestrini et al, 1994
; Lipponen et al, 1995
). This ‘paradoxical' favourable prognostic effect of BCL2 in breast cancer could be related to its non-apoptotic functions (Pietenpol et al, 1994
; O'Reilly et al, 1996
). Increased expression of BCL2 protein may also disrupt the balance with other members of the BCL2 family, including the expression of pro-apoptotic proteins (Cory et al, 2003
The exact mechanism of differential BCL2 protein expression in breast cancer is complex. BCL2 is expressed in normal breast glandular epithelium and is known to be upregulated by oestrogen, possibly as a direct result of transcriptional induction (Wang and Phang, 1995
; Leung and Wang, 1999
). We show that, in cancers, BCL2 positivity is not simply a surrogate for ER positivity: 14% of BCL2+ tumours were ER− and 31% of BCL2− tumours were ER+. BCL2 amplification/copy number gain is rare and correlation between transcript and protein levels in breast cancer is not linear (unpublished observations), suggesting post-transcriptional regulation. Therefore, although in Oncotype DX (Paik et al, 2004
), BCL2 is one of the 21 genes in the prognostic signature, measurement of BCL2 protein expression may provide prognostic information that is not identical.
The prognostic value of BCL2 was present across molecular subtypes (ER+/luminal, HER2+, HER2− and triple negative), an important new observation, and was independent of tumour size, grade and stage. Women with ER+/BCL2− disease were found to have a worse prognosis than those with ER−/BCL2+ disease. The interaction between treatment and the prognostic role of BCL2 was also addressed, showing that the prognostic impact of BCL2 is independent of adjuvant therapy received. Finally, BCL2 is an important prognostic variable in a risk prediction setting and improves the performance of prediction algorithms such as AOL. Novel markers that could be used to save women from unnecessary cytotoxic adjuvant therapy are urgently needed and BCL2 provides valuable additional prognostic information to guide clinical decision making in this setting.
In summary, this large analysis establishes BCL2 as an independent and powerful prognostic protein marker in early-stage breast cancer. BCL2 provides prognostic information in all subgroups defined by other prognostic factors. Thus, it could be used with prediction algorithms such as AOL to improve the ability to discriminate risk groups, by simply using the HRs and the prevalence of BCL2 positivity in each subgroup we report. Indeed, the assignment of patients to risk groups is significantly altered by BCL2 expression (see , , ). The exact way to apply BCL2 testing for risk stratification and the approach to standardise BCL2 immunohistochemistry for this application will now require rigorous prospective assessment, but, given our findings reported here, this is feasible.