Expression of c-erbB3 protein in primary breast carcinomas.

Expression of c-erbB3 protein was investigated in 104 primary breast carcinomas comprising nine comedo ductal carcinoma in situ (DCIS), 91 invasive ductal carcinomas and four invasive lobular carcinomas using two monoclonal antibodies, RTJ1 and RTJ2. Of the 91 invasive ductal carcinomas, seven contained the comedo DCIS component adjacent to the invasive component. An immunohistochemical technique was used to evaluate the association between expression of c-erbB3 and clinical parameters and tumour markers such as epidermal growth factor receptor (EGFR), c-erbB2, cathepsin-D and p53 in archival formalin-fixed paraffin-embedded tumour tissues. Our results indicated that RTJ1 and RTJ2 gave identical staining patterns and concordant results. It was found that the overexpression of c-erbB3 protein was observed in 67% (6/9) of comedo DCIS, 52% (44/84) of invasive ductal carcinomas, 71% (5/7) of carcinomas containing both the in situ and invasive lesions and 25% (1/4) of invasive lobular carcinomas. A significant relationship (P < 0.05) was observed between strong immunoreactivity of c-erbB3 protein and histological grade, EGFR and cathepsin-D, but not with expression of c-erbB2, p53, oestrogen receptor status, lymph node metastases or age of patient. However, we noted that a high percentage of oestrogen receptor-negative tumours (59%), lymph node-positive tumours (63%) and c-erbB2 (63%) were strongly positive for c-erbB3 protein. We have also documented that a high percentage of EGFR (67%), c-erbB2 (67%), p53 (75%) and cathepsin-D-positive DCIS (60%) were strongly positive for c-erbB3. These observations suggest that overexpression of c-erbB3 protein could play an important role in tumour progression from non-invasive to invasive and, also, that it may have the potential to be used as a marker for poor prognosis of breast cancer.

C-erbB3. a new member of the erbB type I family of tyrosine kinase receptors. has been mapped to human chromosome 12qll-13 ). However. the related epidermal growth factor receptor (EGFR) is located on chromosome 7pl2-13 (Spurr et al. 1984) and the c-erbB2 gene on 17pl'-2 1.3 (Coussens et al. 1985. The gene for the c-erbB3 receptor is transcribed into a 6.2-kb mRNA. which is translated into the protein and modified by glycosvlation to give a mature protein of about 160 kDa (Plowman et al. 1990: Prigent and. Recently. it has been shown that heregulilnneu differentiation factor (NDF) binds to and stimulates the tyrosine kinase activity of c-erbB3 (Carrawav et al. 1994).
Expression of c-erbB3 has been detected in normal human adult and fetal tissues using polvclonal antibodies ). Rajkumar et al (1993) have described the production of a monoclonal antibody (RTJI) that is specific for c-erbB3 protein and reported positive immunostaining in a series of gastrointestinal tract tumours. Other studies have also demonstrated overexpression of c-erbB3 protein in breast carcinomas (Lemoine et al. 1992a: Gasparini et al. 1994: Quinn et al. 1994: Travis et al. 1996. pancreatic cancers (Lemorne et al. 1992b). gastric cancers (Sanidas et al. 1993). cenical cancer (Hunt et al. 1995. prostate cancer (Poller et al. 1992). bladder cancer (Rajkumar et al. 1996) and ox-arian carcinomas (Mandai et al. 1994: Simpson et al. 1995. Expression of EGFR and c-erbB2 receptors and the other members of the erbB familv has been extensivelv studied in many different tumours. and it has been found that their oxerexpression is associated w-ith poor prognosis (Lofts andGullick. 1991: Gullick. 1991). Elevated levels of EGFR and c-erbB2 has-e been found in breast carcinomas and are often found to be associated with poor prognosis (Sainsburx et al. 1987: Costa et al. 1988. Currently. there is little information axailable regarding the expression of c-erbB3 protein in formalin-fixed and paraffin-embedded primary breast carcinomas. As breast cancer is an important disease in Malaysia. especially in younger women. we used this study to evaluate the c-erbB3 expression in Malavsian breast cancer patients. The indirect immunoperoxidase method wvas used to examine the expression of c-erbB3 protein in breast carcinoma tissues fixed in formalin. We used two different types of monoclonal antibodies. RTJI (IgM) and RTJ2 (IgG1). which were raised against the same 49.3-kDa synthetic peptide to detect the c-erbB3 protein. Here. we report the expression of c-erbB3 protein in different histological types of pnrmary breast carcinomas and relate it to several clinical parameters. such as oestrogen receptor status. lymph node status. histological grade and patients' age. and tumour markers such as EGFR. c-erbB2. cathepsin-D (oestrogen-regulated protein) and p53 (tumour-suppressor gene).

MATERIALS AND METHODS Patients and tissues
A consecuti-e series of 29 and 75 primary breast carcinomas were obtained. respectively. from Subang Jaya Medical Centre. Subano Jaya. Selangor. Malaysia. and Hospital Kuala Lumpur. Kuala Lumpur. Malaysia Tumours in most of the patients were detected symptomatically and only a small proportion were diagnosed mammographically. with further histological confirmation in both the hospitals. Patients were treated by wide local excision with axillary node dissection if the tumour size was less than 4 cm. Mastectomy was performed when the tumour was more than 4 cm in size. Tle median tumour size was 2 cm. No treatment had been given before the tumour was excised. The median number of nodes retrieved in axillary dissection was 12 nodes for Hospital Kuala Lumpur and ten nodes for Subang Jaya Medical Centre. In our study. lymphatic invasion was seen in almost all cases.
Vascular invasion was observed in five cases. The tumours were fixed in buffered formalin for 12-18 h after surgery. Paraffim sections of tissues were cut at 5 gm thickness for histological and immunohistochemical evaluation. Full clinical information was available for each patient. Of the 104 cases of primary breast carcinomas, nine were ductal carcinoma in situ (DCIS) of comedo type, 84 were invasive ductal carcinomas, seven contained both the in situ and the invasive component and four were invasive lobular carcinomas. Forty-one of the patients showed lymph node metastases and 54 showed no lymph node involvement. Patients were grouped into two groups based on age at the time of diagnosis: age < 50 years and age > 50 years. Sixty-six patients were diagnosed at age below 50 and 38 were diagnosed at the age of 50 and above. All the invasive ductal carcinomas were graded using the method described by Elston (1987). Eighteen of the invasive ductal carcinomas were well differentiated (grade I), 43 were moderately differentiated (grade H) and 30 were poorly differentiated (grade III). Invasive lobular carcinomas were graded according to the modified criteria (Elston. 1987) of Bloom-Richardson (Bloom and Richardson. 1957). All four invasive lobular carcinomas were poorly differentiated.
The sections were deparaffinized in xylene and rehydrated in descending concentrations of ethanol. After rinsing in distilled water. endogenous peroxidase was blocked with 3% hydrogen peroxide (in water) for 10 min at 37°C. The sections were washed briefly in phosphate-buffered saline (PBS) pH 7.6 for 5 min. The sections were incubated with the c-erbB3 primary antibodies (RTJ1 and RTJ2) (dilution 1:20) for 1.5 h at room temperature.
After the incubation. the slides were washed in PBS for 15 min and incubated with biotinylated secondary antibody (LSAB/HRP, Duet. Dako) for 10 min at 37°C. After further washing for 5 mi.
the sections were incubated with streptavidin-conjugated peroxidase complex (LSAB/HRP. Duet. Dako) for another 10 min at 37°C. The tissue sections were washed again in PBS for S min.
Initially, the sections were microwaved twice in 10 mM sodium citrate buffer (pH 6.0) for 5 min. The slides were cooled at room temperature for 20 min and washed in PBS before incubating with the primary antibody. T'he rest of the staining procedures were followed as above. Breast carcinoma tissues known to react with the marker were used as positive controls and incubated in each batch of staining. As for negative control, the primary antibody was omitted and replaced with PBS and negative control antibodies mouse IgGl (clone DAK-GOI, Dako) and IgM (clone DAK-G08, Dako). The antibodies were adjusted to the same concentration as the primary antibody.

Scoring and statistical analysis
The staining for c-erbB3 was cytoplasmic and was assessed according to the method descnbed by Quinn et al (1994). Negative staining or equivocal staining was considered as less than normal expression, weak but defmnite positive staining was considered to represent normal expression and strong, granular staining indicated overexpression. However, for the purpose of statistical analysis. only the intensity of staining was considered. Strong membrane staining for c-erbB2 and epidermal growth factor receptor (EGFR) and strong cytoplasmic staining for cathepsin-D were considered to represent overexpression of the gene product. Tumours that exhibited strong positivity for c-erbB3. c-erbB2, EGFR and cathepsin-D were included in statistical analysis. For p53, the tumour scores were based on the method described by Isola et al (1992). Tumours were scored as strongly positive if more than 20% of the nuclei were stained. hImmunostaining in more than 20% of the tumour cells was taken to represent p53 protein overexpression. If only a small proportion of the nuclei were stained (1-20%), the tumour was scored weakly positive.
Weakly positive and negative tumours were not included in statistical analysis. Oestrogen receptor expression was considered positive when more than 10% of the tumour cells demonstrated positive nuclear staining and negative when less than 10% stained (Pertschuk et al, 1990). This method was used because it has been demonstrated to be more predictive of patients' prognosis.
The chi-squared analysis (contingency tables) was performed to assess the significance of association between expression of c-erbB3 and clinicopathological parameters and other tumour markers. Statistical Graphics System Version 5.0 was used to conduct the correlation tests. The level of significance used throughout the statistical test was 0.05 (5%).

RESULTS
In the present study. expression of c-erbB3 protein was detected using two monoclonal antibodies, RTJ1 and RTJ2. We observed identical staining for both the antibodies. In general. the staining was homogeneous in the majority of the tumour cells with mild C-erbB3 in breast cancer 1387 Figure 1 Immunohistochemical analysis of c-erbB3 expression detected by RTJ1 in paraffin-embedded invasive ductal carcinoma tissue. The intense staining for the c-erbB3 protein was observed as cytoplasmic staining in the malignant epithelial cells (magnification x 133) Figure 2 Expression of c-erbB3 protein was detected by RTJ2 immunohistochemically in paraffin-embedded invasive ductal carcinoma tissue. Strong staining for c-erbB3 protein was noted in the cytoplasm of the malignant cells (magnification x 66) heterogeneous reactivity. The staining for c-erbB3 detected by both the antibodies was predominantly cytoplasmic, with none of the tumours exhibiting membrane immunoreactivity. The staining was finely granular throughout the cytoplasm of the malignant  (71) S, significance at P < 0.05. NS, not significant. aNine DCIS cases were not included in statistical analysis. epithelial cells, which are shown in Figure 1 (RTJ 1) and Figure 2 (RTJ2). Immunopositivity of c-erbB3 protein was found in 65% of the 104 primary breast carcinomas. Strong staining, which denotes overexpression, was seen in 54% (56/104) and weak staining in 11% (12/104) of the total breast carcinomas studied. No immunoreactivity was observed in non-malignant cells. Antibody reactivity was not observed in the stromal cells. No staining was noted in negative control tissue.
Immunopositivity of c-erbB3 was evaluated in comedo DCIS, invasive ductal carcinomas, carcinomas with combined lesions (in situ and invasive ductal component) and invasive lobular carcinomas. Our data indicated that six (67%) of the nine comedo DCIS were strongly positive for c-erbB3 protein. No weak staining was seen in the in situ carcinomas. Of the 84 invasive ductal carcinomas, 44 (52%) showed strong immunopositivity, whereas seven (8%) showed weak staining. Among the four invasive lobular carcinomas, one (25%) was strongly stained and three (75%) were weakly stained. Immunoreactivity of c-erbB3 protein was also observed in five (71 %) of seven cases in which comedo DCIS was adjacent to the invasive ductal component. In these cases, both the in situ and the invasive component were strongly positive. No weak staining was observed. Table 1 summarizes the relationship between c-erbB3 expression (detected by RTJ 1 and RTJ2) and prognostic factors including oestrogen receptor status, lymph node involvement, histological grade and patients' age. Nine comedo DCIS were analysed separately. We did not observe any differences between the antibodies. Overexpression of c-erbB3 was significantly associated with poorly differentiated tumours (P = 0.004). We demonstrated that 71% of the grade Ill tumours exhibited strong c-erbB3 positivity. No significant difference was observed with oestrogen receptor status (P > 0.05), lymph node involvement (P > 0.05) and patients' age (P > 0.05). Although the associations between c-erbB3 and oestrogen receptor or lymph node status were not supported statistically, we showed that a high percentage of oestrogen receptornegative tumours (59%) and lymph node-positive tumours (63%) were c-erbB3 positive. Coexpression with other tumour markers, such as EGFR, c-erbB2, p53 and cathepsin-D, was analysed ( Table 2). Our results showed a significant relationship between overexpression of c-erbB3 protein and EGFR (P = 0.003) and cathepsin-D (P = 0.006), but not with c-erbB2 (P > 0.05) or p53 (P > 0.05). We found that high percentage of tumours strongly positive for EGFR (68%) and cathepsin-D (64%) were also strongly positive for c-erbB3. Sixty three per cent of c-erbB2positive tumours were strongly positive for c-erbB3, but the relationship was not supported statistically. All nine comedo DCIS obtained from patients below age 50 were negative for oestrogen receptor. Six of these were strongly positive for c-erbB3. Our data on comedo DCIS indicated that 67% (4/6) of EGFR-positive tumours, 67% (6/9) of c-erbB2-positive tumours, 75% (3/4) of p53-positive tumours and 60% (3/5) of cathepsin-D-positive tumours exhibited strong c-erbB3 positivity.

DISCUSSION
Our data showed that over half (54%) of 104 primary breast carcinomas of various histological types overexpressed c-erbB3 protein, which was detected by RTJ1 and RTJ2. Both the antibodies gave an identical staining pattem, which suggests the specificity of the antibodies. The staining was predominantly cytoplasmic, with none of the tumours demonstrating membrane staining. RTJ1, an IgM monoclonal antibody, was shown to give a better staining pattem than the polyclonal antibody when tested on formalin-fixed paraffin-embedded tissues or breast cancer cell lines using the immunohistochemical technique, although it was shown to detect other proteins of higher molecular weight in Western blots (Rajkumar et al, 1993). The superior perfonnance of RTJ 1 could be due to it being an IgM antibody with ten potential combining sites and a high avidity for multiple copies of the c-erbB3 protein in tissue sections (Rajkumar et al, 1993). RTJ2, an IgGl monoclonal antibody directed against the same peptide (49.3 kDa) , gave an identical immunohistochemical staining pattern with tissues which had earlier been stained with RTJ1 (Rajkumar et al, 1993) and rabbit polyclonal antibody . Furthermore, RTJ2 has been shown to be specific for c-erbB3 protein as determined by Western blotting, immunoprecipitation and immunocytochemistry. Lemoine et al (1992a) showed good agreement between the intensity of signal on Norhern blot analysis and the intensity of immunohistochemical staining with the 49.3-kDa polyclonal antibody using breast cancer and non-breast cancer cell lines. In addition, Rajkumar et al (1993) had shown that breast carcinoma cell lines and non-malignant breast carcinoma cell lines that had been previously shown to express high, intermediate and low levels of c-erbB3 mRNA (Lemoine et al, 1992a) demonstrated an identical staining reaction when stained with the polyclonal antibody (Lemoine et al, 1992a) and RTJ1. As RTJI and RTJ2 gave concordant results and identical staining pattems which were also indicated by the present study, the cytoplasmic staining of RTJ1 could not be due to nonspecific reaction or cross-reaction. In a recent paper on a larger study with RTJ1, Travis et al (1996) reported that 15% of 346 primary breast cancer cases and 35% of 145 advanced breast cancer cases showed strong staining for c-erbB3. Previously, Lemoine et al (1992a) had reported increased expression of c-erbB3 in 22% of 195 primary infiltraing breast carcinomas. In another study by Quinn et al (1994), 29% of 97 primary breast cancer cases demonstrated c-erbB3 overexpression. Poller et al (1992), using polyclonal antibody and a smaller sample size, demonstrated that 13 of 14 primary breast carcinomas were positive for c-erbB3 immunoreactivity but failed to describe the staining intensity of the positive cells. This makes it difficult to interpret their data Using RTJ1, these authors had shown that the staining was observed to be predominantly cytoplasmic.  has suggested that there is a possibility that the majority of c-erbB3 protein at any one time is present in intracellular pools. Initially, it had been thought that the cytoplasmic immunoreactivity of c-erbB2 was non-specific (de Potter et al, 1989). However, Kumar et al (1991) had reported that only about 20% of the total c-erbB2 protein expressed in SKBR-3 breast cancer cells was located on the cell surface, the remainder being in other intracellular fractions. In addition, Coombs et al (1993) had shown that the cytoplasmic staining for c-erbB2 can be blocked by the immunizing peptide, suggesting that this represents expression of the c-erbB2 protein. This could also be one of the possibilities for c-erbB3. In contrast, Gasparini et al (1994), using the same antibody (RTJ1), has reported that 65% of the node-negative breast carcinomas showed staining of cell membranes, but 13% showed strong positivity and a much higher percentage (89%) showed cytoplasmic staining. In addition, the authors have indicated that there is a highly significant association between membrane and cytoplasmic staining. In common with other findings, the present data showed that these antibodies (RTJ1 and RTJ2) locate the c-erbB3 protein mainly in the cytoplasm, suggesting that cytoplasmic immunoreactivity may be significant. The high proportion of breast carcinomas with strongly expressed c-erbB3 indicates that the gene may have an important role in breast tumorigenesis.
Of the nine ductal carcinoma in situ of comedo subtype, six (67%) showed overexpression of c-erbB3 gene product.
Interestingly, ductal cacinoma in situ has been shown to be a precursor of invasive carcinomas, because it is frequently present in tissues adjacent to breast cancer and is associated with the presence of invasive carcinoma in the same region of the same breast where the DCIS was found (Page et al, 1982;Betsill et al, 1987). Comedo DCIS, a subtype of DCIS, has been associated with a higher rate of proliferation, as determined by thymidine labelling index (TLI), compared with other subtyps of DCIS (Meyer, 1986). Furthermore, features such as comedo subtype have been demonstrated to be associated with local recurrence and progression to invasive breast cancer (Lagios et al. 1989;Lagios, 1990). This implies that comedo DCIS are biologically aggressive neoplasms.
Based on these observations, we suggest that expression of c-erbB3 may be up-regulated during the preinvasive stages and could be an important pathogenic factor in early events of breast malignancies.
In addition, we have demonstae that five ( Poller et al ( 1992) have shown that expression of c-erbB3 was detected in the invasive and in the in situ component present in the same tumour. This clearlv indicates that expression of c-erbB3 is not only present in non-invasive staaes. but also maintained in the invasive stages. We also noted that a high frequency of invasive ductal carcinomas (52%-c ) showed elevated levels of c-erbB3 protein. The above information further strengthens our findings that c-erbB3 may be involved in the progression from preinvasive to invasive stage. The above observations have an important implication regarding the role of c-erbB3 in the initiation and progression of breast cancer. Elevated levels of c-erbB3 were seen in one (25%7) invasive lobular carcinoma. but the role of c-erbB3 in this type of cancer cannot be determined because of an insufficient number of samples.
In addition. we have investigated the relationship between elevated levels of c-erbB3 protein and various clinicopathological parameters. Our data indicated that overexpression of c-erbB3 was significantly associated with high histological grade tumours.
Expression of c-erbB3 was independent of oestrogen receptor status. nodal involvement and patients' age. although a hiah percentage of oestrogen receptor-negative tumours (59%7c) and lymph node-positive tumours (63%7) were strongly positive for c-erbB3 protein. Lemoine et al (1992a) reported that overexpression of c-erbB3 was significantly correlated with the presence of lymph node metastases. but not with oestrogen receptor status and tumour grade. In a larger study. Travis et al (1996) were unable to demonstrate significant association between strong c-erbB3 positivitvand oestrogen receptor status. histological grade. lymph node status or patients age. Similar observation was noted in a smaller series of 97 malignant breast tumours by Quinn et al (1994). In lymph node-negative breast cancer patients. Gaspanini et al (1994) could not show any association between c-erbB3 positivity and patients age. histological grading or steroid hormone receptors. However. we were not able to compare our results with those of last study because the patient groups were different. Furthermore. Gasparini et al (1994) used membrane staining to interpret the data but did not determine the association between cvtoplasmic staining and pathological or other clinical characteristics in order to observ-e the difference between membrane and cytoplasmic stainingy. It has been well documented that tumours that lack oestrogen receptors demonstrate lymph node involvement and poorly differentiated tumours were associated with poor prognosis. However. our results showed that expression of c-erbB3 protein was significantlv associated with histological grade which has not been documented previously. These observations indicate c-erbB3. a new member of the type 1 growth factor receptor family. may have potential to be an indicator for poor prognosis.
Overexpression of c-erbB3 protein was significantly associated with EGFR and cathepsin-D. but independent of p53 and c-erbB2. Our data were in agreement with Lemoine et al (I1992a) for p53 and c-erbB2. but not for EGFR. Travis et al ( 1996) and Quinn et al ( 1994) did not find any association between c-erbB3 and c-erbB2. Gasparini et al (1994) found a significant association with c-erbB2 in nodenegative breast carcinoma patients. which is different from the group of patients in the present study. We have documented that a high percentage of EGFR-positive (67%7c). c-erbB2-positive (67%). pS3positive (75%c) and cathepsin-D-positive (6(%)%) DCIS were strongly positive for c-erbB3. Recent studies have shown that transfection of cells with c-erbB3 and c-erbB2 reconstitutes a higher affinity binding, receptor. which is capable of goenerating a tyrosine phosphory-lation signal in response to heregulim. Furthermore. in cells expressing c-erbB2 and c-erbB3. both proteins become tyrosine phosphorylated upon interaction with heregulin (Sliwkowski et al. 1994). This heregulin-stimulated phosphorvlation of c-erbB3 is likely mediated bv cross-phosphorylation of the c-erbB3 protein by the c-erbB2 receptor tvTosine kinase (Kim et al. 1994: Sliw-kowski et al. 1994). It also has been shown that the bindinc of epidermal growth factor (EGF) to the EGFR results in the activation of its protein tyrosine kinase activity and the phosphorylation of the c-erbB3 protein on tyTosine residues (Kim et al. 1994). From various studies. it has been noted that overexpression of EGFR and cathepsin-D has been associated with poor prognosis (Lewis et al. 1990: Nicholson et al. 1990: Tandon et al. 1990: Winstanley et al. 1993. suggesting that coexpression of c-erbB3 protein and these tumour markers in the same tumours mav be useful to predict the outcome or status of the disease.
Presence of high levels of c-erbB3 in non-invasive. non-invasive and invasive in the same lesions and invasive carcinomas may indicate that c-erbB3 could be involved in the progression of tumours from preinvasive to invasive stare. Furthermore. its association with established prognostic factors such as histologrical arade and with tumour markers includinc EGFR and cathepsin-D may suggest that c-erbB3 could have potential to be a prognostic indicator for breast cancer. However. other studies on a greater number of patients. with clinical follow-up. are needed to demonstrate the usefulness of this new marker.