Abstract
CD66a, also known as biliary glycoprotein (BGP), is a member of the carcinoembryonic antigen (CEA) family and the human homologue of the rat cell-CAM. There is evidence that aberrant expression or loss of CD66a in tumor tissue is of biological significance. No data about its expression in breast carcinoma cells and only sparse information about the expression of CD66a in normal breast are available thus far. In this study we used monoclonal antibodies to analyze the expression of CD66a and CEA in normal tissue, benign lesions, and in noninvasive and invasive carcinomas of the mammary gland. In normal tissue and benign lesions, CD66a was consistently expressed at the apical sites of epithelial cells and in myoepithelia, whereas CEA was absent or was restricted only to some apical membranes within the ductal tree. The specific staining of myoepithelia was most evident in pseudoinfiltrative radial scars and sclerosing adenosis. However, the apical expression of CD66a disappeared with the development of the malignant phenotype in noninvasive and invasive carcinomas, and changed gradually from low- to high-grade noninvasive carcinomas into a predominant uniform membrane staining all around the atypical cells. CEA expression was irregular in intensity and distribution. The native apical CD66a staining was partially preserved in some highly differentiated invasive carcinomas with a better prognosis, such as tubular and papillary carcinomas. These findings indicate that loss of CD66a expression rather than a change in staining patterns coincides with the development of the malignant phenotype.
CELL ADHESION plays a key role in the establishment of tissue architecture and differentiation. To date, many cell–cell adhesion molecules (CAMs) and cell–matrix or –substrate adhesion molecules (SAMs) have been identified. Compared to normal tissues, malignant tumors are characterized by a disruption of tissue architecture and a derangement in differentiation. It has been postulated that changes in cell–cell and cell–matrix interactions account for the ability of cancer cells to overcome normal tissue boundaries and disperse to distant sites. The loss of cell–cell binding that closely correlates with differentiation and invasive potential of malignant tumors is accompanied by loss or altered expression of cell adhesion molecules (Pignatelli and Vessey 1994).
CD66a (BGP) is the human homologue of cell CAM, an established cell adhesion molecule of the rat (Ocklind et al. 1983; Hinoda et al. 1988; Odin et al. 1988; Lin and Guidotti 1989; Aurivillius et al. 1990). The designation “biliary glycoprotein” refers to a glycoprotein first described in human bile (Svenberg 1976; Svenberg et al. 1979). Considering internal amino acid sequences of this glycoprotein, a BGP cDNA clone in a normal colon library was identified (Hinoda et al. 1988). CD66a is a member of the carcinoembryonic antigen (CEA) family, which belongs to the immunoglobulin superfamily. However, CD66a combines structural features of the immunoglobulin superfamily with functional properties of cadherins (Rojas et al. 1990; Turbide et al. 1991). In addition, BGP is the major antigen of the CD66 cluster of granulocyte differentiation antigens (Stoffel et al. 1993).
The expression of CD66a in human epithelial, myeloid, and endothelial cells was described by Prall et al. (1996). Furthermore, the expression of CD66a was studied in specimens of colorectal carcinomas and in the corresponding normal mucosa. Reduced expression was found in the majority of the carcinomas, indicating that loss or the reduction of CD66a expression is a major event in colorectal carcinogenesis (Neumaier et al. 1993).
No data on the expression of CD66a in noninvasive and invasive breast carcinomas have been published thus far, and only sparse information about its expression in normal cells of the breast is available (Prall et al. 1996). The aim of our study was to analyze the expression of CD66a and CEA in invasive carcinomas of the breast and to compare it with that in normal tissue, benign proliferative lesions, and premalignant lesions of the mammary gland.
Materials and Methods
The isolation and characterization of the CD66a MAb4D1/C2 and the CEA MAbT84.66 were described previously in detail (Wagener et al. 1983a,b; Neumaier et al. 1985; Drzeniek et al. 1991; Stoffel et al. 1993).
Tissue Samples
A total of 146 tissue samples were collected from routinely processed, paraffin-embedded material, including normal mature breast tissue (five), fibrocystic lesions (five), adenosis and radial scar (nine), ductal and lobular hyperplasia (six), atypical ductal and lobular hyperplasia (six), noninvasive ductal and lobular carcinomas (15), and invasive carcinomas of different type and grading (100) (see Table 1). The tissue material was selected after histological review from the files of the Department of Gynecological Histopathology, Clinic of Gynecology and Obstetrics, University of Hamburg. All lesions were classified according to the most recent WHO criteria (Rosen and Oberman 1992), with the exception of intraductal carcinomas, which were graded as proposed by Holland et al. (1994).
Immunohistochemistry
The surgical specimens had been routinely fixed in 4% buffered formalin. Serial sections 4–6 μm thick were cut from the paraffin blocks and mounted on APES-coated slides, deparaffinized in xylene, and rehydrated in graded alcohol to TBS (50 mM Tris, 150 mM NaCl, pH 7.4). Slides were microwaved five times for 2 min in 10 mM citrate, pH 6.0, in a commercial kitchen microwave oven. After cooling for 20 min, slides were washed in TBS, blocked for 30 min at room temperature with normal goat serum (Dako; Glostrup, Denmark) diluted 1:20 in TBS, and incubated overnight at 4C with 4D1/C2 at 4 μg/ml or T84.66 at 2 μg/ml in TBS. Nonimmune murine immunoglobulins (Dako) at equal concentrations were used as negative controls. Alkaline phosphatase–anti-alkaline phosphatase (APAAP) detection (Dako) and color development with new fuchsin was carried out according to standard procedures. Finally, the slides were counter-stained with hemalum and mounted with glycerin/gelatin.
Immunohistochemical detection of CD66a and CEA in invasive carcinomas
Immunocytochemistry
Imprints of three normal mature breast tissues, three ductal hyperplasias, three lobular hyperplasias, and five invasive ductal carcinomas were fixed in 0.05% glutaraldehyde solution in 0.1 mol/liter phosphate buffer, pH 7.4, for five min. The immunocytochemical staining with MAb 4D1/C2 was performed as described above without microwave pretreatment.
Microscopic Evaluation
Evaluation of conventional histology, immunohistochemistry, and immunocytochemistry was performed independently by two pathologists. Questionable cases were reviewed in conference and discrepancies were resolved by discussion. Immunohistochemical and immunocytochemical stains were assessed on an arbitrary scale for the density of the immunohistochemical staining. Cases were classified as: labeled cells absent (−), sparse (+), moderate (++), or strong (+++). In addition, the pattern of immunohistochemical labeling was noted.
Statistical Methods
The χ-square test was used for testing significance of relationships. Because the small frequencies observed, Yates correction was applied to improve the approximation of the χ-square statistics.
Results
Immunohistochemistry
The specificity of MAb4D1/C2 for CD66a and of MAb T84.66 for CEA has been established in previous studies (Drzeniek et al. 1991; Stoffel et al. 1993; Prall et al. 1996). The staining results are described below and summarized in Table 1. Granulocytes always showed strong cytoplasmic staining with MAb 4D1/C2, and endothelial cells in vessels of the mammary gland were negative.
Normal Mature Breast
The columnar epithelium of the extralobular ducts and the epithelia of terminal duct lobular units, consisting of terminal intralobular ducts and multiple lobular ductules, were strongly or moderately marked at their apical surfaces with MAb 4D1/C2. The well-defined myoepithelial layer displayed strong or moderate membrane staining (Figure 1a).
MAb T84.66 showed no or only a focal and sparse reaction in some apical membranes. Myoepithelia were always negative.
The intra- and interlobular connective tissue and the vessel walls did not show staining reactions either with MAb 4D1/C2 or with MAb T84.66.
Benign Proliferative Lesions
In duct ectasia and cysts as well as in blunt duct adenosis, the epithelial layer of the affected ducts and ductules showed strong to moderate apical membrane staining with MAb 4D1/C2 (Figure 1c), but only in a few of the cases a focal and sparse staining with MAb T84.66 appeared. The outer layer, the myoepithelium, displayed distinct membrane staining with MAb 4D1/C2. In sclerosing adenosis and radial scar, the pseudoinfiltrative outgrowths were moderately or strongly decorated (Figure 1d), whereas the MAb T84.66 was negative. Both MAbs did not show a reaction in apocrine metaplasia.
Duct Hyperplasia and Atypical Duct Hyperplasia
Both epithelial and myoepithelial cells participating in the varied growth patterns of intraduct hyperplasias showed moderate to strong staining with the MAb 4D1/C2, giving rise to apical membrane labeling. Conversely, the staining intensity of T84.66 clearly decreased. Similar results were obtained in the cases of atypical duct hyperplasia.
Intraductal Carcinoma
The group of intraductal carcinoma consisted of well-differentiated (four) and poorly differentiated types (six). Both MAbs showed moderate to sparse membrane or apical membrane staining of the atypical duct cells, but the staining of MAb T84.66 was only focal. Apical membrane staining was observed in differentiated zones, at the luminal surface, and around intercellular spaces of micropapillary, clinging, and cribriform growth patterns of well-differentiated types (Figure 1e). In contrast, membrane staining was striking in solid growth areas of poorly differentiated cases (Figure 1f). In 2/6 cases of the poorly differentiated type, the myoepithelia showed only a partially positive reaction with weak staining intensity.
Lobular Hyperplasia and Atypical Lobular Hyperplasia
In the epithelia of terminal intralobular ducts and lobular ductules, an immunohistochemical reaction at their apical surface could be demonstrated with both MAbs, but the reaction was clearly stronger with MAb 4D1/C2. The myoepithelium showed membrane staining only with MAb 4D1/C2.
Lobular Carcinoma In Situ (LCIS)
The atypical cells of LCIS showed strong or moderate membrane staining with MAb 4D1/C2 and a focal moderate to sparse reaction with MAb T84.66.
Invasive Carcinoma
In 77 of 100 invasive carcinomas (77%), the cells showed staining with MAb 4D1/C2 (Table 1), whereas in 23 tumors expression was absent. In 66 of the 77 positive cases (85, 71%), uniform membrane staining was detectable (Figure 1g). A native apical membrane staining was preserved in 11 cases (14, 29%). Staining with MAb T84.66 was frequently only focal.
No relationship was found between expression of CD66a and grading, age, tumor size, menopausal status, or hormone receptor status. However, the apical membrane staining was strongly correlated with well-differentiated invasive duct carcinomas (29, 17%) especially of the papillary (2/2 cases) and tubular type (2/3 cases; see Figure 1i). This correlation was not seen with the CEA-specific MAb T84.66.
Immunocytochemistry
Imprints of normal mature breast, duct, and lobular hyperplasias, and of invasive duct carcinomas showed a positive reaction with MAbs 4D1/C2 and T84.66. Sparse or moderate uniform staining on the surface of normal and hyperplastic epithelial cells and carcinoma cells was seen (Figures 1b and 1h).
Discussion
In the present study, the expression of CD66a was analyzed by immunohistochemistry and immunocytochemistry in a variety of benign and malignant mammary lesions. An MAb was used that does not crossreact with other members of the CEA family. The specificity of this MAb for CD66a has been described in detail (Drzeniek et al. 1991).
Immunohistochemistry and -cytochemistry with MAb 4D1/C2.
It has been shown previously that CD66a is expressed in a broad range of human epithelia, in endothelial cells, and in cells of the myeloid lineage. Well-defined CD66a staining of apical membranes could be observed in duct epithelia of the prostate, in the brush border cells of the colon, and in epithelia of large and small pancreatic ducts. Furthermore, normal mammary, endometrial, and endocervical glands were positive for CD66a but were negative for CEA (Hansson et al. 1989; Prall et al. 1996).
The immunohistochemical data of the present study support the previous findings that CD66a is expressed in exocrine glands. The columnar epithelium of the extralobular ducts and the epithelia of terminal duct lobular units in normal mature breast were labeled exclusively at their apical surfaces. Similar findings could also be demonstrated in benign lesions of the mammary gland such as duct ectasia, blunt duct adenosis, duct hyperplasia, and lobular hyperplasia, as well as atypical duct hyperplasia. In contrast to the previous findings (Prall et al. 1996), some of these cases showed a focal expression of CEA in apical membranes of epithelia.
However, the distribution of CD66a in a wide variety of epithelia does not provide an immediate clue to its biological function. Because of transfection studies, the homology between CD66a and cell CAM, and the putative intereaction between CD66a with E-selectin, it has been postulated that CD66a may function as a cell adhesion molecule in vivo.
Consistent CD66a expression would be expected in the lateral membrane domain, where adhesive interactions between epithelial cells predominate. However, immunohistochemical analysis of tissue sections showed an interesting phenomenon of exclusive strong apical reaction of CD66a in glandular epithelia. This phenomenon does not exclude a possible function in cell adhesion, because the CD66a epitope recognized by MAb 4D1/C2 may be masked at adjacent lateral or basal membranes by possible interaction with its binding molecules.
In an attempt to resolve this apparent paradox, we made imprints of normal mature breast, duct and lobular hyperplasias, and carcinomas. Surprisingly, the mechanically dissociated cells stained by immunocytochemistry showed an exclusive uniform membrane staining independent of histological diagnosis, thus supporting the theory of epitope masking. Moreover, Mowery and Hixson (1991) reported similar findings. In rat hepatocytes, cell CAM expression was absent in the lateral and basal membranes but became positive when the tissue was dissociated.
In early stages of malignant transformation, the apical membrane expression of CD66a changes to a distinct uniform membrane staining, similar to the immunocytochemical analysis of imprints. This uniform membrane expression was still mixed with the native apical staining in intraduct carcinomas of well-differentiated micropapillary, clinging, and cribriform types, and in some well-differentiated tubule and papillary carcinomas. The uniform membrane staining predominated in atypical lobular hyperplasia, LCIS, poorly differentiated intraduct carcinoma, and the vast majority of invasive carcinomas. If the occurrence of the uniform membrane staining is due to a loss or reduction of interaction of the adhesion molecule with its binding molecules, this indicates an important shift towards the malignant phenotype.
In a portion of the invasive carcinomas, expression of CD66a was absent. The loss or downregulation of CD66a expression in these cases is in accordance with similar findings in other malignant tumors. Several experiments indicate that the loss of CD66a in tumor tissues is of great functional importance. Expression of CD66a is downregulated in the majority of human colorectal carcinomas (Neumaier et al. 1993). Similarly, cell CAM or its murine homologue has been reported to be downregulated in hepatocellular carcinomas in rats (Hixson et al. 1985; McEntire et al. 1989) and in colorectal carcinomas in mice (Rosenberg et al. 1993). Recently, Hsie et al. (1995) have transfected cell CAM cDNA into a tumorigenic prostate cancer cell line. The transfected clones showed significantly lower growth rates and less tumorigenicity in vivo than control cells.
In contrast, the staining with MAb T84.66 was frequently only focal, faint, or absent, predominantly in poorly differentiated carcinomas (Table 1) with a well-known doubtful prognosis.
Both MAbs did not show a reaction in epithelia with apocrine differentiation.
Surprisingly, the myoepithelial layer displayed distinct staining with the CD66a-specific MAb 4D1/C2. The specific decoration of myoepithelia was most evident in pseudoinfiltrative radial scars and sclerosing adenosis. On the other hand, myoepithelia were always negative with MAb T84.66. This phenomenon supports the differential specificity of both MAbs, even though there is no explanation for this feature.
In conclusion, the change from apical to membrane staining patterns in noninvasive carcinomas and the partial reduction or loss of CD66a expression in invasive carcinomas of the breast compared with normal tissue and benign proliferative lesions of mammary gland appear to be features of malignancy and loss of differentiation.
Footnotes
Acknowledgements
Supported by the Deutsche Krebshilfe/Dr. Mildred Scheel Stiftung (M 74/92/St1) and by the Hamburger Stiftung zur Förderung der Krebsbekämpfung.
The excellent technical assistance of Ms Gabriele Becker and Ms Sonja Petersen is gratefully acknowledged.