We here identify a novel role of NuclErbB-2 as a biomarker of poor clinical outcome in MembErbB-2-positive breast tumors.
First, we developed an IF protocol using the rabbit polyclonal antibody C-18, raised against the ErbB-2 carboxy (C)-terminal region [8
], for the detection of ErbB-2 in paraffin-embedded tumor sections. We chose C-18 because we had already revealed its exquisite sensitivity to detect NuclErbB-2 and MembErbB-2 in breast cancer cells by IF [8
]. Since seminal findings [6
] and our own previous work [8
] demonstrated that full-length ErbB-2 is present in the nuclear compartment of breast cancer cells, we also tested the mouse monoclonal 9G6 antibody directed against the ErbB-2 amino (N) terminus, which we previously found recognized NuclErbB-2 and MembErbB-2 by IF in breast cancer cells [8
]. In addition, we used another ErbB-2 monoclonal antibody directed against the ErbB-2 C-terminus (clone e2-4001) as well as two polyclonal antibodies that recognize ErbB-2 phosphorylated at tyrosine 1222 or 877, two sites we have already revealed are phosphorylated in NuclErbB-2 [8
]. We found that none of these antibodies detected NuclErbB-2 or MembErbB-2 by IF staining of the tissue arrays. This failure could be explained by the fact that we could not appropriately retrieve the single site or epitope on the antigen recognized by the monoclonal antibodies, or the site specifically containing the phosphorylated residue in the case of the two anti-phospho ErbB-2 polyclonal antibodies, due to the harsh conditions used in the preparation of the tissue blocks. On the contrary, given that polyclonal antibodies, such as C-18, recognize a broad range of epitopes on the antigen, including denaturation-resistant epitopes, we successfully detected ErbB-2 by IF in the tissue arrays with C-18.
Our present findings using the IF protocol with the C-18 antibody showed 33.6% NuclErbB-2 positivity and 14.2% MembErbB-2 overexpression in our Chilean cohort. This percentage of NuclErbB-2 staining is comparable to the one (38%) reported in breast tumors for nuclear expression of the epidermal growth factor receptor (EGF-R), another member of the ErbBs family [27
]. Excellent to substantial levels of concordance were found between MembErbB-2 overexpression determined by IF with the C-18 antibody and by IHC using C-18, BT-HER2 or A0485 antibodies, indicating that the IF protocol we developed is as sensitive and specific as routinely used IHC staining procedures, including those using the Food and Drug Administration (FDA)-approved A0485 antibody, the same clone used in the commercial HercepTest immunostaining kit, also approved by the FDA. We found no correlation between NuclErbB-2 and MembErbB-2 positivity, in contrast to two previous works which reported a direct correlation between MembErbB-2 and NuclErbB-2 presence in breast tumors [6
]. In said works, ErbB-2 localization was studied by either IHC [6
] or IF [7
] using the A0485 antibody. This discrepancy is unlikely due to differences in how NuclErbB-2 positivity was defined in those studies and in ours since we found the same results after we re-evaluated the correlation applying a less stringent criterion for NuclErbB-2 positivity. No correlation was found either when we extended the cohort under study. Another difference between said study exploring NuclErbB-2 prevalence in breast cancer by IF with A0485 [7
] and ours, is the lower incidence of NuclErbB-2 detected by A0485 (12.3%) as compared to the one we found with C-18. A probable explanation for these disparities is that C-18 has much greater sensitivity for detecting NuclErbB-2 than A0485. Therefore, a significant number of samples which scored negative for NuclErbB-2 IF staining with A0485 are in fact detected by C-18, accounting for both the higher NuclErbB-2 positivity and the lack of correlation between MembErbB-2 and NuclErbB-2 presence. Several of our data support this hypothesis. First, our results of NuclErbB-2 incidence studied by IHC using A0485 and C-18 demonstrated that A0485 showed a significantly lower percentage of NuclErbB-2 presence (4%) as compared to C-18 (12%). Second, in accordance with the previous work mentioned above [6
], we found a good correlation between MembErbB-2 overexpression and NuclErbB-2-positivity in the IHC staining with A0485, but no correlation between NuclErbB-2 and MembErbB-2 positivity in the IHC staining with C-18. Our recent findings demonstrating that in the nucleus of breast cancer cells ErbB-2 is associated with Stat3 and PR [8
], could help to explain the different sensitivity of both antibodies. A larger number of epitopes recognized by A0485 could be blocked by NuclErbB-2 association with other proteins e.g. Stat3 and PR, than those recognized by C-18. On the other hand, the prevalence of NuclErbB-2 was lower when we used C-18 in IHC than when we used it in IF. This is likely due to the fact that while in the IHC staining with C-18 we used the standard conditions for membrane ErbB-2 detection [13
], we introduced modifications in our IF protocol to enhance the sensitivity of NuclErbB-2 detection, including optimization of the antigen retrieval protocol and improvement of the conditions of interaction between antibodies and antigens, as described in a pioneering work assessing the clinical value of EGF-R nuclear presence [27
]. The fact that the discrepancy in C-18 sensitivity when used in IF or in IHC is observed for the detection of NuclErbB-2 but not for MembErbB-2 argues for the possibility that differences in the conformation of MembErbB-2 and NuclErbB-2 and/or the association of NuclErbB-2 with other proteins may restrict antibody access to NuclErbB-2.
Although there are no extensive studies assessing the frequency of MembErbB-2 positivity in LA women in their own native countries [28
], a series of startling works have shown the role of MembErbB-2 overexpression as predictor of poor outcome in breast cancer cohorts from Brasil [29
]. Here we found 13% of MembErbB-2 overexpression by IHC and 14.2% by IF in our LA cohort. MembErbB-2-positive rates among Caucasian and Asian women, in whom most studies have been done to date, currently tend to be below 20%, with reports by most investigators stating that the real positive rate ranges between 15%-20% [31
]. Therefore, our present findings for the first time show that the incidence of MembErbB-2 positivity in LA women living in their own countries is not significantly different from that of women living in developed countries.
MembErbB-2 overexpression has long been found to be associated with poor clinical outcome [2
]. We here demonstrate that patients whose tumors expressed MembErbB-2 and NuclErbB-2 have worse OS compared with those with tumors showing only MembErbB-2. Moreover, NuclErbB-2 positivity was a significant independent predictor of poor survival in patients with MembErbB-2 overexpression. NuclErbB-2 also resulted a marker of lower overall survival in the subgroup of patients with tumors MembErbB-2+/ER-PR-, included in the ErbB-2-positive molecular subtype [23
]. This latter finding for the first time highlights NuclErbB-2 presence as a molecular signature that might help to define two biologically distinct subsets of tumors within the ErbB-2-positive molecular subtype. The role of NuclErbB-2 as a TF [6
], as well as our recent discoveries showing that ErbB-2 acts also as a coactivator [8
] may underlie the poor outcome of tumors with NuclErbB-2. The combined capacity of ErbB-2 to activate mitogenic signaling pathways when located in the membrane and to act straight in the nucleus as a transcriptional regulator might drive the assembly of a gene network different from the one assembled in tumors with exclusive MembErbB-2 presence. This network would be involved in growth, metastasis and/or response to anti-ErbB-2 therapies, such as trastuzumab. Blockage of MembErbB-2 capacity to activate cytoplasmic signaling cascades is one of the mechanisms of trastuzumab action [5
]. Therefore, a likely explanation to trastuzumab resistance could be ErbB-2 nuclear presence and function as a transcriptional regulator. Strong support to this possibility is provided by our findings that abrogation of ErbB-2 nuclear localization inhibits in vitro
and in vivo
growth of breast tumors expressing both NuclErbB-2 and MembErbB-2 [8
The significant number of early stage tumors with no lymph node metastasis we have in our total cohort may account for the lack of NuclErbB-2 prognosis value in our overall population. Indeed although MembErbB-2 positivity is a poor prognostic factor in axillary node positive and negative tumors, the clinical significance of MembErbB-2 in early stage cancer is not yet completely understood [32
]. On the other hand, our recent findings on PR and ErbB-2 interaction [8
] provide a most exciting explanation for the comparable OS in MembErbB-2-/ER+PR+tumors with and without NuclErbB-2. We revealed that in HR+breast tumors, PR co-opts ErbB-2 function not only as membrane tyrosine kinase but also as transcriptional regulator [8
]. In such scenario, endocrine therapies targeting ER, which in turn modulate the effects of PR, an ER-target gene, would also abolish ErbB-2 action at both membrane and nucleus.
Several key features of our cohort, such as levels of MembErbB-2 overexpression, the inverse correlation between HR presence and MembErbB-2 overexpression, and the fact that most of the tumors were in early stages, are comparable to those of the North American population [2
] supporting extrapolation of our conclusions onto North American women. We hope our present findings will encourage further studies of NuclErbB-2 role as biomarker in larger populations stratified according to their treatment with endocrine therapy or chemotherapy and, in the case of MembErbB-2 overexpressing tumors, with trastuzumab or more recently with lapatinib.