Expression of the ERBB3 gene product in breast cancer.

Abnormalities of the EGF receptor and/or the related ERBB2 receptor occur in a significant proportion of cases of human breast cancer and are important influences in the behaviour of this tumour type. In this report we demonstrate by nucleic acid analysis and immunohistochemistry that the recently recognised third member of this gene family, ERBB3, shows a wide range of expression in breast cancer, and shows stronger immunoreactivity than that observed in normal tissue in 43 out of 195 cases (22%) of primary breast cancer. Overexpression of ERBB3 appears to result from increased levels of gene transcription since in none of the cell lines or primary cancers analysed did we find evidence of gene amplification. High expression of ERBB3 is positively associated with the presence of lymph node metastases, but there was no demonstrable relationship with patient survival in this series. ImagesFigure 1Figure 2Figure 3

A great deal of progress has been made in understanding the oncogenes and growth factors which are involved in the molecular pathogenesis of breast cancer, allowing the development of new markers for patient selection and prognosis and potentially the identification of new targets for therapy. The type 1 family of receptor tyrosine kinases including the epidermal growth factor receptor (EGFR) and the ERBB2 (or HER2) protein have been recognised as being particularly important influences in breast cancer (Gullick 1990a). Both of these receptor proteins are expressed at low levels on normal breast epithelial and myoepithelial cells and are often overexpressed in breast cancer. Patients with tumours in which there is elevated EGFR expression have a relatively poor prognosis with shorter relapse-free survival and overall survival (Sainsbury et al., 1987;Costa et al., 1988). Overexpression of the ERBB2 protein is also a marker of poor prognosis for node-positive disease (Slamon et al., 1987) and probably also for node-negative disease (Perren, 1991). More exciting, however, is the accumulating evidence that the ERBB2 receptor protein has potential as a target for antibody-directed therapy (Shepard et al., 1991) and other approaches such as inhibition of receptor dimerisation and tyrosine kinase function (Gullick 1990b).
A new member of the type 1 receptor family, ERBB3, was cloned by two groups (Kraus et al., 1989;Plowman et al., 1990) and found to be expressed at the RNA level in a number of tumour cell lines, including 6 out of 17 breast cancer cell lines (Plowman et al., 1990). Because amplification and/or overexpression of the other members of this receptor family (EGFR and ERBB2) are clearly important factors in neoplastic transformation of breast epithelium, we were interested to determine whether ERBB3 was activated by similar mechanisms. The aim of our present study was to investigate the possibility that abnormalities of structure and/or expression of this receptor occur in human breast cancer. We have therefore examined in detail a panel of breast cell lines (both malignant and non-malignant) and a large series of tumour samples well-characterised for clinicopathological parameters and activation of oncogenes. We have used Southern and Northern blot analysis with a specific ERBB3 probe to determine gene structure and mRNA expression and also immunohistochemistry with two new antibodies raised against ERBB3-specific peptides for protein expression . The data were analysed for any associations between ERBB3 expression and factors such as tumour size, grade and stage, expression of oncogenes and patient survival.

Materials and methods
Tumour and normal tissue samples Immunohistochemical studies were carried out on formalinfixed, paraffin-embedded sections from 195 patients with primary infiltrating breast cancer of various types managed at Guy's Hospital, London. One hundred and sixty eight patients had infiltrating ductal carcinoma, 24 had infiltrating lobular carcinoma and in three patients carcinoma was of special type. One hundred and forty two of the patients were treated by modified radical mastectomy and 53 patients had conservation treatment and radiotherapy. Twenty nine patients received adjuvant therapy: in eight it was endocrine therapy and in 21 it was CMF chemotherapy (cyclophosphamide, methotrexate and 5-fluorouracil). The patients were treated at Guy's Hospital between 1979 and 1982 and have a medium follow-up of 10.04 years. Patients selected for this study all had verified follow-up data and the presence and date of recurrence was determined in a standard manner according to the criteria of Hayward et al. (1978).
There was sufficient formalin-fixed, paraffin-embedded material available for further immunohistochemical investigations. The overall survival of these patients is representative of the survival of all patients with breast cancer managed at Guy's Hospital during the same period. A large volume of information is available for these patients including analysis of expression of EGF receptors (immunohistochemistry with antibody 12E, Gullick et al., 1991a), ERBB2 (immunohistochemistry with antibody 21N, Venter et al., 1987), oestrogen receptors (as described by King et al., 1979) and p53 (immunohistochemistry with antibody CM1, Midgley et al., 1992) together with clinical and pathological parameters.
High molecular weight DNA was prepared from 36 human breast cancer samples which had been snap-frozen in liquid nitrogen immediately after surgical resection and subsequently stored at -70°C and also from normal peripheral blood leucocytes of these patients. In a pilot study immunohistochemical staining with two antibodies (49.3 and 61.3, see below) for ERBB3 protein was performed on paraffin sections of the breast tumours from these cases.
Thirty six biopsy samples of non-neoplastic breast tissue from women in different phases of the menstrual cycle were also examined for immunoreactivity with antibodies to ERBB3.

Cell lines
The MTSV-1.7, MRSV-2. 1 and MRSV-2.4 immortalized, non-malignant breast cell lines (Bartek et al., 1991) were kindly donated by Dr Joyce Taylor-Papadimitrou, ICRF London. All breast cancer cell lines were obtained from the American Type Culture Collection. Cell pellets in agarose plugs were fixed in formalin and embedded in paraffin blocks.
The transfected cell line 293/HER3 (cotransfected with the HER3x expression vector containing full-length ERBB3 cDNA in pCDM8 and the pMClneo plasmid) and the parent line 293 were obtained from Dr Greg Plowman Southern blot analysis 10 jig DNA from each of the paired normal/tumour samples was digested with restriction enzymes EcoRI or PstI, electrophoresed through 0.8% agarose gel and blotted onto nylon membranes (Hybond-N+, Amersham). Hybridisation was performed with a cDNA probe for ERBB3 labelled with 32P-ATP by the random primer method, and filters washed at high stringency in 0.1 x SSC/0. 1% SDS at 65°C for 10 min. The ERBB3 probe was generated by PCR from a placental cDNA library and represents positions 3625 to 4138 of the sequence reported by Kraus et al. (1989), including sequences encoding the C-terminal 170 amino acids of the ERBB3 protein. The 513 base pair fragment was cloned in Bluescript and double-stranded sequence analysis performed to confirm that it conformed in every respect to the ERBB3 sequence published by Kraus et al. (1989). It is completely specific for ERBB3 in hybridisation analysis of human DNA and RNA.
After autoradiography the filters were stripped and reprobed with a probe for ,-actin (Cleveland et al., 1980). RNA analysis RNA was extracted from cultured cells and Northern blots prepared by standard techniques (Ausubel et al., 1991). Blots were hybridised with the ERBB3 cDNA probe as described above.  After autoradiography the filters were stripped and reprobed with for ,-actin.

Immunohistochemistry
The immunohistochemical staining of paraffin sections was carried out using the affinity-purified rabbit polyclonal antisera to peptide sequences of ERBB3 as previously described . The antibodies were shown to specifically recognise ERBB3 protein in 293 cells cotransfected with the HER 3 x expression vector (full length ERBB3/ HER3 cDNA in pCDM8) and the pMClneo plasmid ( Figure  1). All cases were immunostained with the antibody 49.3 while 36 cases (those examined at the DNA level by Southern blot, see above) were examined both with antibody 49.3 and another specific anti-ERBB3 antibody 61.3 . Briefly, the primary antibody (49.3 at a concentration of 5 jig ml-' or 61.3 at a concentration of 1 .tg ml-, in phosphate-buffered saline (PBS) with 0.5% bovine serum albumin) was incubated with the section for 1 h at room temperature. The second layer was biotinylated anti-rabbit antibody (Dako, Copenhagen, Denmark) employed at a dilu-

Statistical analysis
Relationships between variables were analysed using the Chisquared test. Survival curves were calculated using the method of Kaplan & Meier (Peto et al., 1977) and differences between curves were analysed by the logrank test.

Results
We observed a wide range in the level of expression of ERBB3 mRNA and protein in the various breast cell lines examined (Table I and Figure 2). The immortalised, nonmalignant cell lines, such as MTSV-1.7, showed barely detec-  aStaining was assessed quantitatively according to the intensity of the colour reaction and graded as less than normal, within the normal range or more intense than normal. In general, staining was fairly homogeneous within a particular tumour.  (Figure 2). There was no abnormality of structure or gene copy number observed on Southern blot analysis of DNA from any of the cell lines.
In the normal breast tissue samples examined we found that there was finely granular cytoplasmic immunoreactivity of weak to moderate intensity in the luminal epithelial cells of the ducts and acini throughout all phases of the menstrual cycle, and in occasional cases there was also weak immunoreactivity in the myoepithelial cells.
In the 36 tumours analysed by Southern blot we detected no abnormality of gene structure or copy number (data not shown), but immunohistochemistry showed that there was wide variation in the level of ERBB3 expression in the primary breast cancer specimens. Immunohistochemistry with the two anti-ERBB3 antibodies gave concordant results but since the level of background staining of stromal tissues was found to be less with 49.3 than with 61.3 in pilot studies on 36 cases, this antibody was used for the larger series of 195 cases (Table II). Twenty six (13%) of these 195 cases showed no detectable immunoreactivity in tumour cells (i.e., less than normal), while at the other end of the spectrum 42 cases (22%) showed cytoplasmic staining of much greater intensity than normal breast epithelium (Figure 3). The frequency of high expression was higher in infiltrating ductal carcinomas (39/168 cases, 23%) than in invasive lobular carcinomas (3/24 cases, 12.5%), but this difference is not statistically significant. We have seen only one case of breast cancer displaying membrane staining with the anti-ERBB3 antiserum 49.3. In the remaining cases the level of immunoreactivity in the tumour cells was similar to that observed in normal breast epithelium.
There was a statistically significant association between strong ERBB3 immunoreactivity (greater than normal breast epithelium) and the presence of lymph node metastases (P = 0.02, Table III). Those cases with overexpression of ERBB3 showed no difference in survival compared to those cases expressing normal or low levels, regardless of nodal status (Figure 3). There was no association between the level of expression of ERBB3 and tumour size or histological grade of invasive ductal carcinomas (Table III). We found no correlation between ERBB3 immunoreactivity and immunoreactivity for EGF receptor, ERBB2 and p53, nor with steroid receptor status.

Discussion
Our studies have shown that while the ERBB3 protein is expressed in the majority of human breast cancers at levels similar to those of the normal breast epithelium it is overexpressed in approximately 22% of cases. This overexpression appears to be due to upregulation of ERBB3 gene transcription rather than gene amplification.
Examination of the panel of breast cell lines showed that there was close correlation between the level of ERBB3 mRNA expression and intensity of immunoreactivity with the specific anti-ERBB3 antibodies. It is interesting that only or low 153) Time (years) Figure 4 Survival for patients with operable breast cancer according to tumour immunoreactivity with antibody 49.3. very low levels of ERBB3 expression were detected in the nonmalignant immortal cell lines, while higher levels were found in the cell lines derived from tumours. None of these cell lines showed amplification or rearrangement of the ERBB3 gene, and nor did any of the primary breast cancers which were examined by Southern blot.
Expression of ERBB3 showed a wide variation in the tumour biopsies, some samples being completely negative by immunohistochemistry while others were stained much more strongly than normal breast epithelium. The immunoreactivity was nearly always cytoplasmic and finely granular and, in contrast to the usual patterns of ERBB2 and EGF receptor overexpression in breast cancer, we saw only one case of membrane staining for ERBB3. However, we have observed more frequent membrane immunoreactivity for ERBB3 in pancreatic cancers and gastric cancers . Membrane immunoreactivity for the related receptor protein ERBB2 is characteristically seen in tumour cases with very high levels of expression associated with gene amplification while cytoplasmic staining is usually indicative of lower levels of expression (Venter et al., 1987). Our finding of cytoplasmic immunoreactivity may imply that the level of overexpression of the ERBB3 protein is relatively modest in breast cancer. We found no evidence of ERBB3 gene amplification in these tumours and cell lines, and similar findings were reported in 17 breast cancer lines by Kraus et al. (1989). However, the fact that immunoreactivity for this protein, postulated to be a transmembrane receptor, is generally confined to the cytoplasm is intriguing but not necessarily surprising. There is evidence that substantial proportions of EGFR and ERBB2 proteins may be in cellular pools other than the plasma membrane, suggesting that cytoplasmic immunoreactivity may be significant. Indeed, Kumar et al. (1991) have recently shown that even in a cell line expressing high levels of ERBB2 from multiple copies of the gene only about 19% of the receptor protein was expressed on the cell membrane, the remainder being in other cell fractions.
Overexpression of the ERBB3 protein in these breast cancers was independent of the level of expression of either the EGF receptor or the ERBB2 receptor. Both of the latter have been reported to represent prognostic markers when found overexpressed in breast cancer (reviewed in Gullick, 1990a), but in this present series ERBB3 overexpression was not found to correlate with disease outcome. However, it was associated with the presence of lymph node metastases. Our previous experience with ERBB2 (Gullick et al., 1991b) suggests that substantially larger series of cases will be required to reliably confirm or exclude the utility of ERBB3 overexpression as a prognostic marker in breast cancer. Our report shows that while high expression of this protein occurs in a substantial proportion of breast cancers the majority of cases show normal or low levels of expression. Whether ERBB3 represents an oncogene involved in breast cancer remains uncertain.