Differential Reactivities of Carcinoembryonic Antigen (CEA) and CEA-Related Monoclonal and Polyclonal Antibodies in Common Epithelial Malignancies KIERAN SHEAHAN, M.B., M.R.C.P.I., MICHAEL J . O'BRIEN, M.D., BOHDANA BURKE, M.D., PETER A. DERVAN, M.D., J . CONOR O'KEANE, M.B., M.R.C.P.I., LEONARD S. GOTTLIEB, M.D., AND NORMAN ZAMCHECK, M.D.
Received September 27, 1989; accepted for publication November 17, 1989. This work was presented in part at the 78th Annual Meeting of the United States/Canadian Academy of Pathology, San Francisco, March 1989. Address reprint requests to Dr. O'Brien: Mallory Institute of Pathology, 784 Massachusetts Avenue, Boston, Massachusetts 02118.
Mallory Institute of Pathology and Boston City Hospital, Pathology Department, Boston University Medical School, Boston, Massachusetts, and Mater Misericordiae Hospital, Dublin, Ireland
SINCE THE INITIAL study by Primus and associates in 197526 of carcinoembryonic antigen (CEA) in paraffin tissue using immunoperoxidase techniques, it has become one of the most widely used immunohistochemical markers in surgical pathology. Despite this, the use of immunostaining for CEA applied to solving problems of tumor histogenesis in surgical pathology is poorly understood, and a CEA antibody is often included in an antibody panel because "it is there" rather than with a specific goal of differential diagnosis in mind. Polyclonal antisera are still commonly used in immunohistology laboratories, although cross-reactivity with antigenic epitopes of CEArelated molecules such as nonspecific cross-reacting antigen (NCA) and biliary glycoprotein (BGP) is widely recognized and is the presumed reason for the considerable variation in staining reported in some tissues.28 Although absorption against cross-reacting antigens can be used for routine studies, this is technically demanding and may be beyond the scope of many small laboratories. The availability now of monoclonal antibodies to CEA20 warrants a fresh look at the role of CEA in immunohistologic diagnosis. Toward accomplishing this, we have performed an immunohistologic survey of a large series of common epithelial malignancies, using four selected monoclonal antibodies to CEA and a polyclonal antibody. The monoclonal antibodies are from a larger panel produced and extensively characterized by Haggarty and colleagues14 and on which preliminary studies of tissue localization have been reported by Jothy and associates.16 The specific objectives of the study that we report here were threefold: (1) to seek evidence that different species of the CEA molecule exist in carcinomas of different histogenesis; (2) to determine how variation in the capacity to produce CEA (of any species) among the common epithelial malignancies can be used to identify their site of origin; and (3) to
157
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 6, 2016
To evaluate the role of carcinoembryonic antigen (CEA) in solving problems of tumor histogenesis in surgical pathology, monoclonal antibodies to four distinct epitopes of CEA (E-Z-EM) were applied to paraffin sections of 303 epithelial neoplasms from multiple sites. Two epitopes were CEA specific (D14 and B7.1), one was shared with nonspecific cross-reacting antigen (NCA) (B7.8), and the fourth (B18) was common to CEA, NCA, and biliary glycoprotein antigen (BGP). A sample of the tumors (n = 110) was also stained with a polyclonal anti-CEA (DAKO). Gastrointestinal adenocarcinomas, including esophageal and gastric (n = 19), small intestinal (n = 8), colorectal (n = 56), biliary tract (n = 8), and pancreatic adenocarcinomas (n = 14), were consistently positive with all five antibodies. Other predominantly gland-forming carcinomas tested, comprising lung (n = 22), ovary (n = 18), and endometrium (n = 12), were either invariably negative with all five antibodies (endometrial adenocarcinoma, nonmucinous ovarian adenocarcinoma) or demonstrated selective and variable positivity (lung: D14, 50%; ovarian mucinous: D14, 50%). Among large polygonal cell carcinomas (hepatocellular carcinoma, renal cell carcinoma, melanoma, and adrenal carcinoma), only hepatomas stained positively, showing a distinctive canalicular staining pattern with the B18 (BGP epitope) (55%) and polyclonal antibody (50%). In the small polygonal cell carcinoma category, true CEA positivity was rare in breast (D14, 10% and B7.1,14%) and never seen in prostatic carcinomas and carcinoid tumors. A subset of these breast (8 of 42), prostate (4 of 22), and carcinoids (4 of 7) showed exclusive positivity for the B18 antibody (NCA/BGP epitope). Ovarian serous papillary carcinomas (n = 14), papillary carcinomas of thyroid (n = 12), transitional cell carcinomas of the bladder (n = 11), and mesotheliomas (n = 3) were negative with all monoclonal antibodies. Metastatic carcinomas (n = 74) showed a similar pattern of reactivity to primary tumors. The authors conclude that CEA immunostaining may assist in identifying the histogenesis of epithelial tumors in several morphologic categories; that differential reactivities of the CEA monoclonal antibody panel exceed those of the polyclonal antibody; and that the discriminating power of the monoclonal panel is related to whether (1) CEA is or is not produced or (2) NCA or BGP is produced without concomitant CEA production. There is little evidence to support a concept of site-specific CEA species. (Key words: CEA; Monoclonal antibodies; Polyclonal antibodies; Carcinoma; Metastases; Immunohistology) Am J Clin Pathol 1990;94:157-164
158
SHEAHAN ET AL.
determine whether the differential expression of CEA, NCA, and biliary glycoprotein has useful applications in investigating tumor histogenesis. Materials and Methods
Antibody Specificity The CEA molecule has at least 10-15 distinct epitopes, some of which are CEA specific, whereas others are not, being shared by one or more CEA-related proteins.33 The epitopes recognized by the four monoclonal antibodies used in this study are as follows: D14 recognizes an epitope specific for CEA; B18 recognizes an epitope shared by CEA, biliary glycoprotein (BGP), and nonspecific crossreacting antigen 1 (NCA 1); the expression of the latter in granulocytes serves as a useful built-in control on most
tissue sections. BGP is expressed by bile canalicular structures in the liver.23,34 B7.8 recognizes an epitope shared by CEA and a nonspecific cross-reacting antigen (NCA) different from NCA 1 that is not expressed on granulocytes. Finally B7.1 also recognizes a CEA-specific epitope, although different than that recognized by D14. We defined our staining grade by the percentage of positive tumor cells: 1% or less staining was considered as grade 0; 1-50% as grade 1; and 50-100% as grade 2. Because the main focus of our study was to evaluate CEA in epithelial tumors as a marker of histogenesis, we devised the following simplified morphologic classification of tumor categories in which to present and compare CEA staining data. Gland-forming carcinomas (gastrointestinal [GI] and non-GI); small polygonal cell carcinomas (e.g., breast, prostate, carcinoid); large polygonal cell carcinomas (e.g., hepatoma, renal and adrenal carcinoma, malignant melanoma); and papillary carcinomas (thyroid, ovary, bladder). Several tumor types could obviously be placed in more than one category (e.g., gland forming and/or papillary). Results Gland-Forming Carcinomas (GI tract, pancreas, and biliary tract) The reactivities of the antibodies studied with adenocarcinomas of different GI sites, pancreas, and biliary tract are given in Table 1. All four monoclonal antibodies stained the great majority of colorectal and pancreatic adenocarcinomas, with a frequency ranging from 92 to 100%. Within this group of tumors, a negative stain was never seen in colorectal carcinoma. A small number of adenocarcinomas (Table 1) of esophagus, stomach, small intestine, and pancreas stained negatively with one or more antibodies. Staining of less than 50% of the tumor was seen most frequently in gastric and least frequently in colorectal carcinoma (Fig. 1). This group of tumors showed no significant staining differences between CEA-
Table 1. CEA Immunostaining Profile of Gland-Forming Carcinomas, Adenocarcinomas of Gastrointestinal Tract, Pancreas, and Biliary Tract No. (%) Positive Site
No. Tested
D14
B7.1
B18
B7.8
DAKO
Esophagus and stomach Small intestine Colorectum Gastrointestinal (all sites) Pancreas Biliary tract
19 8 56 83 14 8
17(89) 6(75) 56(100) 79 (95) 13(92) 6(75)
18(95) 7(88) 56(100) 81 (98) 13(92) 7(88)
18(95) 7(88) 56(100) 81 (98) 14(100) 7(88)
18(95) 7(88) 56(100) 81 (98) 14(100) 6(75)
10(100)*
* 10 tested. t 13 tested.
% 3 tested. § 6 tested.
13 (100)t 23(100) 2(66)* 5 (83)§
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 6, 2016
Three hundred three epithelial tumors (comprising primary and metastatic carcinomas of the colon [n = 56], esophagus, and stomach [n = 19], small intestine [n = 8], pancreas [n = 14], liver [n = 22], biliary tract [n = 8], breast [n = 42], lung [n = 36], ovary [n = 18], endometrium [n = 12], prostate [n = 22], kidney [n = 15], bladder [n = 12], thyroid [n = 12], and adrenal [n = 7]), 7 melanomas, 3 mesotheliomas, and 7 carcinoid tumors were obtained from the surgical pathology files of the Mallory Institute of Pathology and Mater Misericordiae Hospital, Dublin, Ireland. All tissue had beenfixedin 10% formalin and embedded in paraffin. With the use of an indirect immunoperoxidase technique,31 representative sections from each case were stained with four CEA monoclonal antibodies, namely D14, B18, B7.8, and B7.1. (Histosystem®, E-Z-EM, Inc.). A proportion of tumors (n = 110) was also stained with polyclonal anti-CEA (DAKO). Sections of colorectal carcinoma were used as positive controls. Replacement of the primary antibody with normal serum was used as a negative control. The chromogen used was aminoethyl carbazole and the counterstain, Mayer's hematoxylin, or methyl green.
A.J.C.P. • August 1990
Vol. 94 • No. 2
CEA IN EPITHELIAL MALIGNANCIES
/>
I A': '
«*
;/l IS?!
ft/VS
. 's\.(*
)fif-
FIG. 1 (upper, left). Colorectal adenocarcinoma showing diffuse, grade 2 staining with D14 monoclonal antibody. One hundred percent of such cases were positive with all CEA antibodies. Methyl green counterstain (X100). FIG. 2 (upper, right). CEA monoclonal antibody B18 showing a distinctive canalicular staining pattern, which is demonstrable in approximately 50% of hepatomas. Hematoxylin counterstain (X400). FIG. 3 (lower). CEA monoclonal antibody B18 outlining areas of squamous differentiation in a metastatic poorly differentiated carcinoma. Positive staining is confined to cells showing keratinization. Methyl green counterstain (X250).
specific antibodies (D14 and B7.1), between CEA-NCABGP antibodies (B18 and B7.8), or between CEA-specific and CEA-NCA-BGP groups. Polyclonal CEA antibody showed comparable sensitivity in staining 100% of adenocarcinomas of the colorectum, esophagus, and stomach tested.
Gland-Forming Carcinomas—Nongastrointestinal (lung, endometrium, and ovary) The B18 epitope was most frequently positive (68%) and the CEA-specific D14 epitope most often negative (50%) in lung adenocarcinomas (Table 2). Polyclonal anti-
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 6, 2016
.4
Received September 27, 1989; accepted for publication November 17, 1989. This work was presented in part at the 78th Annual Meeting of the United States/Canadian Academy of Pathology, San Francisco, March 1989. Address reprint requests to Dr. O'Brien: Mallory Institute of Pathology, 784 Massachusetts Avenue, Boston, Massachusetts 02118.
Mallory Institute of Pathology and Boston City Hospital, Pathology Department, Boston University Medical School, Boston, Massachusetts, and Mater Misericordiae Hospital, Dublin, Ireland
SINCE THE INITIAL study by Primus and associates in 197526 of carcinoembryonic antigen (CEA) in paraffin tissue using immunoperoxidase techniques, it has become one of the most widely used immunohistochemical markers in surgical pathology. Despite this, the use of immunostaining for CEA applied to solving problems of tumor histogenesis in surgical pathology is poorly understood, and a CEA antibody is often included in an antibody panel because "it is there" rather than with a specific goal of differential diagnosis in mind. Polyclonal antisera are still commonly used in immunohistology laboratories, although cross-reactivity with antigenic epitopes of CEArelated molecules such as nonspecific cross-reacting antigen (NCA) and biliary glycoprotein (BGP) is widely recognized and is the presumed reason for the considerable variation in staining reported in some tissues.28 Although absorption against cross-reacting antigens can be used for routine studies, this is technically demanding and may be beyond the scope of many small laboratories. The availability now of monoclonal antibodies to CEA20 warrants a fresh look at the role of CEA in immunohistologic diagnosis. Toward accomplishing this, we have performed an immunohistologic survey of a large series of common epithelial malignancies, using four selected monoclonal antibodies to CEA and a polyclonal antibody. The monoclonal antibodies are from a larger panel produced and extensively characterized by Haggarty and colleagues14 and on which preliminary studies of tissue localization have been reported by Jothy and associates.16 The specific objectives of the study that we report here were threefold: (1) to seek evidence that different species of the CEA molecule exist in carcinomas of different histogenesis; (2) to determine how variation in the capacity to produce CEA (of any species) among the common epithelial malignancies can be used to identify their site of origin; and (3) to
157
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 6, 2016
To evaluate the role of carcinoembryonic antigen (CEA) in solving problems of tumor histogenesis in surgical pathology, monoclonal antibodies to four distinct epitopes of CEA (E-Z-EM) were applied to paraffin sections of 303 epithelial neoplasms from multiple sites. Two epitopes were CEA specific (D14 and B7.1), one was shared with nonspecific cross-reacting antigen (NCA) (B7.8), and the fourth (B18) was common to CEA, NCA, and biliary glycoprotein antigen (BGP). A sample of the tumors (n = 110) was also stained with a polyclonal anti-CEA (DAKO). Gastrointestinal adenocarcinomas, including esophageal and gastric (n = 19), small intestinal (n = 8), colorectal (n = 56), biliary tract (n = 8), and pancreatic adenocarcinomas (n = 14), were consistently positive with all five antibodies. Other predominantly gland-forming carcinomas tested, comprising lung (n = 22), ovary (n = 18), and endometrium (n = 12), were either invariably negative with all five antibodies (endometrial adenocarcinoma, nonmucinous ovarian adenocarcinoma) or demonstrated selective and variable positivity (lung: D14, 50%; ovarian mucinous: D14, 50%). Among large polygonal cell carcinomas (hepatocellular carcinoma, renal cell carcinoma, melanoma, and adrenal carcinoma), only hepatomas stained positively, showing a distinctive canalicular staining pattern with the B18 (BGP epitope) (55%) and polyclonal antibody (50%). In the small polygonal cell carcinoma category, true CEA positivity was rare in breast (D14, 10% and B7.1,14%) and never seen in prostatic carcinomas and carcinoid tumors. A subset of these breast (8 of 42), prostate (4 of 22), and carcinoids (4 of 7) showed exclusive positivity for the B18 antibody (NCA/BGP epitope). Ovarian serous papillary carcinomas (n = 14), papillary carcinomas of thyroid (n = 12), transitional cell carcinomas of the bladder (n = 11), and mesotheliomas (n = 3) were negative with all monoclonal antibodies. Metastatic carcinomas (n = 74) showed a similar pattern of reactivity to primary tumors. The authors conclude that CEA immunostaining may assist in identifying the histogenesis of epithelial tumors in several morphologic categories; that differential reactivities of the CEA monoclonal antibody panel exceed those of the polyclonal antibody; and that the discriminating power of the monoclonal panel is related to whether (1) CEA is or is not produced or (2) NCA or BGP is produced without concomitant CEA production. There is little evidence to support a concept of site-specific CEA species. (Key words: CEA; Monoclonal antibodies; Polyclonal antibodies; Carcinoma; Metastases; Immunohistology) Am J Clin Pathol 1990;94:157-164
158
SHEAHAN ET AL.
determine whether the differential expression of CEA, NCA, and biliary glycoprotein has useful applications in investigating tumor histogenesis. Materials and Methods
Antibody Specificity The CEA molecule has at least 10-15 distinct epitopes, some of which are CEA specific, whereas others are not, being shared by one or more CEA-related proteins.33 The epitopes recognized by the four monoclonal antibodies used in this study are as follows: D14 recognizes an epitope specific for CEA; B18 recognizes an epitope shared by CEA, biliary glycoprotein (BGP), and nonspecific crossreacting antigen 1 (NCA 1); the expression of the latter in granulocytes serves as a useful built-in control on most
tissue sections. BGP is expressed by bile canalicular structures in the liver.23,34 B7.8 recognizes an epitope shared by CEA and a nonspecific cross-reacting antigen (NCA) different from NCA 1 that is not expressed on granulocytes. Finally B7.1 also recognizes a CEA-specific epitope, although different than that recognized by D14. We defined our staining grade by the percentage of positive tumor cells: 1% or less staining was considered as grade 0; 1-50% as grade 1; and 50-100% as grade 2. Because the main focus of our study was to evaluate CEA in epithelial tumors as a marker of histogenesis, we devised the following simplified morphologic classification of tumor categories in which to present and compare CEA staining data. Gland-forming carcinomas (gastrointestinal [GI] and non-GI); small polygonal cell carcinomas (e.g., breast, prostate, carcinoid); large polygonal cell carcinomas (e.g., hepatoma, renal and adrenal carcinoma, malignant melanoma); and papillary carcinomas (thyroid, ovary, bladder). Several tumor types could obviously be placed in more than one category (e.g., gland forming and/or papillary). Results Gland-Forming Carcinomas (GI tract, pancreas, and biliary tract) The reactivities of the antibodies studied with adenocarcinomas of different GI sites, pancreas, and biliary tract are given in Table 1. All four monoclonal antibodies stained the great majority of colorectal and pancreatic adenocarcinomas, with a frequency ranging from 92 to 100%. Within this group of tumors, a negative stain was never seen in colorectal carcinoma. A small number of adenocarcinomas (Table 1) of esophagus, stomach, small intestine, and pancreas stained negatively with one or more antibodies. Staining of less than 50% of the tumor was seen most frequently in gastric and least frequently in colorectal carcinoma (Fig. 1). This group of tumors showed no significant staining differences between CEA-
Table 1. CEA Immunostaining Profile of Gland-Forming Carcinomas, Adenocarcinomas of Gastrointestinal Tract, Pancreas, and Biliary Tract No. (%) Positive Site
No. Tested
D14
B7.1
B18
B7.8
DAKO
Esophagus and stomach Small intestine Colorectum Gastrointestinal (all sites) Pancreas Biliary tract
19 8 56 83 14 8
17(89) 6(75) 56(100) 79 (95) 13(92) 6(75)
18(95) 7(88) 56(100) 81 (98) 13(92) 7(88)
18(95) 7(88) 56(100) 81 (98) 14(100) 7(88)
18(95) 7(88) 56(100) 81 (98) 14(100) 6(75)
10(100)*
* 10 tested. t 13 tested.
% 3 tested. § 6 tested.
13 (100)t 23(100) 2(66)* 5 (83)§
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 6, 2016
Three hundred three epithelial tumors (comprising primary and metastatic carcinomas of the colon [n = 56], esophagus, and stomach [n = 19], small intestine [n = 8], pancreas [n = 14], liver [n = 22], biliary tract [n = 8], breast [n = 42], lung [n = 36], ovary [n = 18], endometrium [n = 12], prostate [n = 22], kidney [n = 15], bladder [n = 12], thyroid [n = 12], and adrenal [n = 7]), 7 melanomas, 3 mesotheliomas, and 7 carcinoid tumors were obtained from the surgical pathology files of the Mallory Institute of Pathology and Mater Misericordiae Hospital, Dublin, Ireland. All tissue had beenfixedin 10% formalin and embedded in paraffin. With the use of an indirect immunoperoxidase technique,31 representative sections from each case were stained with four CEA monoclonal antibodies, namely D14, B18, B7.8, and B7.1. (Histosystem®, E-Z-EM, Inc.). A proportion of tumors (n = 110) was also stained with polyclonal anti-CEA (DAKO). Sections of colorectal carcinoma were used as positive controls. Replacement of the primary antibody with normal serum was used as a negative control. The chromogen used was aminoethyl carbazole and the counterstain, Mayer's hematoxylin, or methyl green.
A.J.C.P. • August 1990
Vol. 94 • No. 2
CEA IN EPITHELIAL MALIGNANCIES
/>
I A': '
«*
;/l IS?!
ft/VS
. 's\.(*
)fif-
FIG. 1 (upper, left). Colorectal adenocarcinoma showing diffuse, grade 2 staining with D14 monoclonal antibody. One hundred percent of such cases were positive with all CEA antibodies. Methyl green counterstain (X100). FIG. 2 (upper, right). CEA monoclonal antibody B18 showing a distinctive canalicular staining pattern, which is demonstrable in approximately 50% of hepatomas. Hematoxylin counterstain (X400). FIG. 3 (lower). CEA monoclonal antibody B18 outlining areas of squamous differentiation in a metastatic poorly differentiated carcinoma. Positive staining is confined to cells showing keratinization. Methyl green counterstain (X250).
specific antibodies (D14 and B7.1), between CEA-NCABGP antibodies (B18 and B7.8), or between CEA-specific and CEA-NCA-BGP groups. Polyclonal CEA antibody showed comparable sensitivity in staining 100% of adenocarcinomas of the colorectum, esophagus, and stomach tested.
Gland-Forming Carcinomas—Nongastrointestinal (lung, endometrium, and ovary) The B18 epitope was most frequently positive (68%) and the CEA-specific D14 epitope most often negative (50%) in lung adenocarcinomas (Table 2). Polyclonal anti-
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 6, 2016
.4
