Epithelial Membrane Antigen-A Diagnostic Discriminant in Surgical Pathology:
Immunohistochemical Profile in Epithelial, Mesenchymal, and Hematopoietic Neoplasms Using Paraffin Sections and Monoclonal Antibodies
GERALDINE S. PINKUS, MD, AND PAUL J. KURTIN, MD
Glycoproteins isolated from human milk fat globule membranes, designated epithelial membrane antigen (EMA), have been de- tected immunohistochemically in most nonneoplastic epithelia and are potentially a highly effective marker for establishing the epithelial nature of neoplastic cells. With commercially available monoclonal antibodies and an indirect immunoperoxidase tech- nique, EMA localization was evaluated in paraffin-embedded tis- sues from a wide variety of neoplasms (320 specimens). Adeno- carcinomas from various primary sites (breast, lung, colon, stomach, pancreas, gallbladder, prostate, endocrine glands, ovary, kidney, thyroid) were immunoreactive for EMA in 88 of 97 cases (91 per cent). Cytoplasmic and apical luminal membrane staining were the most common patterns of immunoreactivity, with peripheral membrane staining or other patterns also seen in some neoplasms. Squamous cell (13 of 13 cases) and transi- tional cell (12 of 12 cases) carcinomas, small cell anaplastic car- cinomas (12 of 12 cases), and mesotheliomas (six of six cases) were also uniformly EMA-positive. Malignant lymphomas of the Hodgkin’s (15 cases) and non-Hodgkin’s types (74 cases), except for the true histiocytic lymphomas and occasional T-cell lym- phomas, were nonreactive for EMA. Neoplastic and nonneo- plastic plasma cells showed variable EMA positivity. Endocrine neoplasms (17 cases), including carcinoid tumors, medullary car- cinoma of thyroid, adrenocortical carcinomas and pheochro- mocytomas, and germ cell tumors (eight cases, embryonal car- cinoma and seminoma), and a wide variety of soft tissue tumors (27 cases) generally lacked immunoreactivity for EMA; the ex- ceptions to this finding were synovial sarcomas and an epithe- lioid sarcoma. Malignant melanomas (eight cases) were typically nonreactive. Based on the observations in this large series of neoplasms, EMA is an excellent marker of epithelial differentia- tion, appears to be highly reliable for discriminating between poorly differentiated carcinomas and malignant lymphomas, and is especially helpful in characterizing small cell anaplastic car- cinomas. Epithelial membrane antigen immunoreactivity is well preserved in paraffin sections of routinely processed tissues, fa- cilitating application of this technique in diagnostic surgical pa- thology. HUM PATHOL 16:929-940, 1985.
In their most functionally differentiated state, human mammary epithelial cells are capable of the synthesis and secretion of milk. As milk fat globules are secreted by a process of reverse pinocytosis, they are surrounded by a membrane acquired from the apical portion of the mammary cells.1 Human milk fat globule membranes, isolated from delipidized
milk, have been found to be a highly effective im- munogen for the production of polyclonal and monoclonal antibodies.2-+ Initially, these polyspecific antisera were thought to react uniquely with breast epithelium.2 However, subsequent studies demon- strated immunoreactivity with normal simple epi- thelia of many sites, including pancreas, stomach, in- testine, salivary gland, bile ducts, endometrium, fal- lopian tube, urinary tract, respiratory tract, and sweat glands.3,5 By contrast, normal squamous epithelium, hepatocytes, neuroectodermally and mesenchymally derived tissues, gonads, and hematopoietic and lym- phoid tissues were uniformly nonreactive. Based on the broad staining profile for a wide variety of epi- thelia, with immunohistochemical localization pre- dominantly to luminal surface membranes, the im- munogen was designated epithelial membrane an- tigen (EMA).
The chemical nature of EMA has not been com- pletely defined. Available data suggest that EMA is not a single well-defined molecule, but rather a group of molecules with high molecular weights, low protein content, high carbohydrate content (about 50 per cent), and a high content of inorganic material of neither the sulfate nor the phosphate-type.6,7 The an- tigenic determinant apparently rests in the carbohy- drate moiety,8 with galactose and N-acetylglucosa- mine being the two major sugars.6 The antigen rec- ognized by anti-EMA antibodies appears to be distinct from that identified by antibodies to carcinoem- bryonic antigen, alpha-lactalbumin, lactoferrin, casein, cystic disease fluid protein, the secretory com- ponent of IgA, and muramidase.3
Initial studies of human normal and neoplastic tissues, using polyclonal anti-EMA antibodies, dem- onstrated patterns of immunoreactivity related to his- togenesis and degree of differentiation.5,9,10 In gen- eral, epithelial and mesothelial neoplasms exhibited immunoreactivity for EMA, while lymphomas, mel- anomas, and most sarcomas were nonreactive. Mono- clonal antibodies to EMA, some of which differ in their staining patterns in breast tissue and extramam- mary tissues, 11-13 have also been described. Anti- bodies to EMA were not generally available until the recent introduction of a commercial monoclonal re- agent, purportedly with a broad spectrum of epithe- lial immunoreactivity. However, definitive studies with this reagent have not been reported. Since EMA immunoreactivity is well preserved in fixed, paraffin-
Received from the Department of Pathology, Brigham and Women’s Hospital, Boston, Massachusetts. Revision accepted for publication February 4, 1985.
Address correspondence and reprint requests to Dr. Pinkus; Department of Pathology, Brigham and Women’s Hospital, 75 Francis Street, Boston, MA 02115.
embedded tissues,3 this reagent is potentially a highly effective diagnostic tool for establishing an epithelial derivation for poorly differentiated neoplasms, per- mitting their distinction from hematopoietic and mesenchymal tumors ..
The purpose of this study was to determine the profile of immunoreactivity for EMA in large series of neoplasms of various types (carcinomas, including the small cell anaplastic type; sarcomas; mesothe- liomas; and lymphomas) with commercially available monoclonal antibodies to EMA, and to assess the po- tential value of EMA as a marker in diagnostic sur- gical pathology.
MATERIALS AND METHODS
Tissues were retrieved from the surgical pa- thology files of the Brigham and Women’s Hospital. Most specimens were fixed in 10 per cent neutral buffered formalin; B-5 solution, Zenker’s solution, and/or Bouin’s solution were also used in some cases. Tissues were processed routinely and embedded in paraffin.
For immunoperoxidase studies, paraffin sections (3 to 4 pm) were deparaffinized, placed in methanolic peroxide (5 parts methanol to 1 part 3 per cent aqueous hydrogen peroxide, v/v) for 30 minutes to consume endogenous peroxidase, washed well, and placed in 0.05 M TRIS buffer, pH 7.6, supplemented with either 2 per cent horse serum (for the avidin- biotin-peroxidase complex [ABC] technique) or 2 per cent swine serum (for other techniques de- scribed). Sections were then incubated for one hour with monoclonal antibodies to EMA (1:40 to 1:50 di- lution; Dako Corporation, Santa Barbara, California). Anti-EMA antibodies were raised against human milk fat globule membranes; they are of the IgG class and represent tissue culture supernatant from murine hy- bridoma cells (clone E 29/EP 1). Following the incu- bation, slides were washed well with TRIS-saline (1 part 0.5 M TRIS buffer, pH 7.6, and 9 parts normal saline, v/v), and placed in TRIS buffer supplemented with appropriate serum. For the ABC technique, the Vectastain ABC kit for mouse immunoglobulin (PK 4005; Vector Labs, Burlingame, California) was used, with sequential incubation with biotinylated horse an- timouse immunoglobulin antibodies and avidin- biotin-peroxidase complexes. An alternate tech- nique that also provided excellent results consisted of sequential incubation (45 to 60 minutes) with per- oxidase-conjugated rabbit antimouse immunoglob- ulin antibodies (1:30 to 1:40 dilution; Dako), followed by peroxidase-conjugated swine antirabbit immuno- globulin antibodies (1:60 dilution, Dako). The latter two reagents were diluted with 0.1 M TRIS buffer supplemented with 2 to 3 per cent human AB serum (to eliminate any cross-reactivity with human immu- noglobulin). For both techniques, antibody localiza- tion was determined by a peroxidase reaction with 3,3’-diaminobenzidine tetrahydrochloride (Aldrich Chemical Co, Milwaukee, Wisconsin) as the chro-
mogen. Sections were counterstained with hema- toxylin or methyl green solution, dehydrated, and mounted with Permount.
Control sections included lactating breast (posi- tive controls) and serial sections of the tumors, with either diluted culture medium (used for preparation of monoclonal antibody) or an irrelevant monoclonal antibody (e.g., anti-Factor VIII-associated antigen) substituted for anti-EMA antibodies (negative con- trols).
In 20 cases formalin-fixed tissues were processed with and without preliminary trypsinization with type II porcine trypsin (T-8128, Sigma Chemical Co, St. Louis, Missouri; 25 mg/dl in 0.05 M TRIS buffer, pH 7.8, containing 0.134 g/dl CaClo dihydrate, for 20 minutes at 37℃), to determine whether immuno- reactivity for EMA was augmented by proteolytic digestion.
RESULTS
Preliminary studies of normal tissues revealed immunoreactivity for EMA in glandular and ductal epithelia as well as in transitional epithelium. Renal parenchymal staining was confined to the distal and collecting tubules. Normal squamous epithelium, he- patocytes, soft tissue, synovium, lymphoid cells, and hematopoietic elements were EMA-negative. Skin ap- pendages were immunoreactive for EMA.
Adenocarcinomas
Table 1 summarizes the EMA staining patterns for 97 adenocarcinomas (84 primary tumors, 11 me- tastases, and two cases in which both primary tumors and metastases were evaluated) from various sites, including breast (15 infiltrating ductal, three infil- trating lobular, and one medullary type), stomach (five), pancreas (four), colon (14), gallbladder (one), lung (nine), prostate (ten), endometrium (four), ovary (six serous, two mucinous, one endometrioid), kidney (16), and thyroid (six). In 88 cases (91 per cent), neoplastic cells were immunoreactive for EMA. Nonreactive tumors were of prostatic (six cases) or thyroid (three cases) origin.
The proportions of neoplastic cells that were im- munoreactive for EMA varied from case to case and from area to area in a given section. Tumor cells usually stained more intensely than the adjacent non- neoplastic epithelium. The staining patterns of EMA reactivity in the adenocarcinomas had common fea- tures, independent of the site of origin of the tumor, with some features unique to certain primary sites (table 1) or to tumor growth patterns. The most common pattern of EMA positivity consisted of finely granular cytoplasmic staining, observed in 78 cases (80 per cent); this staining was diffuse in most of the tumors (figs. 1 and 2) and focal, usually immediately subjacent to the apical membrane, in a minority of cases. In most of the EMA-positive tumors with a glandular growth pattern (58 of 97; 60 per cent),
EPITHELIAL MEMBRANE ANTIGEN (Pinkus & Kurtin)
| Primary Site | No. Evaluated | No. Immunoreactive for EMA | Staining Pattern | |||
|---|---|---|---|---|---|---|
| Apical Membrane | Peripheral Membrane | Cytoplasm | Intracellular Lumina | |||
| Breast | ||||||
| Ductal, infiltrative | 15 | 15 | 13 | 9 | 15 | 4 |
| Lobular | 3 | 3 | 0 | 2 | 3 | 1 |
| Medullary | 1 | 1 | 1 | 1 | I | 0 |
| Stomach | 5 | 5 | 1 | 1 | 5 | 0 |
| Pancreas | 4 | 4 | 3 | 1 | 4 | 0 |
| Colon | 14 | 14 | 12 | 0 | 12 | 0 |
| Gallbladder | 1 | 1 | 1 | 0 | 1 | 0 |
| Lung | 9 | 9 | 7 | 4 | 9 | 0 |
| Prostate | 10 | 4 | 3 | 0 | 4 | 1 |
| Endometrium | 4 | 4 | 3 | 0 | 4 | 0 |
| Ovary | ||||||
| Serous | 6 | 6 | 5 | 1 | 6 | 0 |
| Mucinous | 2 | 2 | 0 | 0 | 2 | 0 |
| Endometrioid | 1 | I | 0 | 0 | 1 | 0 |
| Kidney | 16 | 16 | 7 | 14 | 9 | 0 |
| Thyroid | ||||||
| Papillary | 3 | 2 | 2 | 0 | 2 | 0 |
| Follicular | 3 | 1 | 0 | 1 | 0 | 0 |
| Totals | 97 | 88 (91%) | 58 (60%) | 34 (35%) | 78 (80%) | 6 (6%) |
* Tissues were fixed in formalin (87); formalin and Bouin’s solution (2); formalin, Zenker’s, and Bouin’s solutions (3); formalin and Zenker’s solution (1); formalin and B5 solution (1); Zenker’s solution (2); or B5 solution (1).
linear apical membrane staining that outlined and ac- centuated all or part of the luminal circumference was also apparent. This pattern was particularly prominent in tumors primary in breast (fig. 2), lung (fig. 3), colon, endometrium, and ovary (serous type). In many cases luminal contents were also immuno- reactive for EMA (fig. 3). Apical membrane staining for EMA was generally accompanied by cytoplasmic immunoreactivity, except in three cases. Poorly dif- ferentiated carcinomas in which glandular lumina were not apparent on routine sections occasionally showed a luminal staining pattern for EMA.
Distinct peripheral membrane staining for EMA was seen more often in adenocarcinomas (primary and metastatic) of breast, lung (fig. 4), and renal or- igin (fig. 5) than in tumors in other sites. Immuno- reactivity was present both in areas of solid growth and between occasional cells forming acini. In pul- monary adenocarcinomas and renal cell carcinomas, the staining pattern appeared to correlate with ar- chitectural features. In tumors that formed papillae, linear staining of the plasma membrane of the su- perficial surfaces of the cells lining the papillae was seen (fig. 6). In seven renal cell carcinomas with gland formation, linear apical membrane staining of neo- plastic cells was seen. In one spindle cell (sarcoma- toid) and one clear cell tumor only focal cytoplasmic staining was observed.
In five adenocarcinomas, immunoreactivity for EMA distinctly outlined intracytoplasmic lumina (fig. 2, inset). With the exception of one case of prostatic carcinoma, this pattern occurred uniquely in tumors of breast origin, although in a minor proportion of the cases (five of 19; 26 per cent).
In EMA-positive adenocarcinomas of thyroid or- igin (two of three papillary; one of three follicular),
only a minor proportion of cells (10 per cent or less) stained. The papillary tumors showed apical mem- brane and/or diffuse granular cytoplasmic staining. In the follicular carcinoma, small circles of EMA reac- tivity occurred between cells.
Other Neoplasms Immunoreactive for EMA
Results for neoplasms other than adenocarci- nomas with immunoreactivity for EMA are summa- rized in table 2.
Squamous cell carcinomas. Thirteen squamous cell carcinomas (all formalin-fixed) from various sites were evaluated. The staining patterns for EMA in infiltrating squamous cell carcinomas of lung (four), uvula (one), and skin (three) were similar. Basaloid cells were weakly stained or negative. In neoplastic cells with squamous differentiation, diffuse granular cytoplasmic staining was seen in all but one case, in which staining was focal. Staining intensity appeared to be greatest at the superficial aspect of the epithe- lium (fig. 7). Peripheral membrane staining was also observed in all but one skin tumor. In the skin, neo- plastic cells consistently stained more strongly than the adjacent uninvolved epidermis. Five in situ car- cinomas of the skin stained to some extent for EMA, with the same topographic variation observed in the invasive tumors.
Transitional cell carcinomas. Twelve transitional cell carcinomas (all formalin-fixed), including one in situ carcinoma of the ureter, four papillary tumors, five invasive carcinomas, and two metastases, were evaluated. In all four papillary transitional cell car- cinomas (three bladder, one renal pelvis) and in one in situ carcinoma of the ureter, the surface of the plasma membrane of the most superficial cells
FIGURE 4 (center right). Adenocarcinoma, primary in lung: formalin fixation. Immunoreactivity for EMA is localized predominantly at the periphery of the neoplastic cells and appears to delineate cell membranes. Weak cytoplasmic staining is also visible in some cells. (Immunoperoxidase technique, methyl green counterstain. x120:)
FIGURE 5 (bottom left). Renal cell carcinoma, clear cell type, pulmonary metastasis; formalin fixation. Strong membrane staining for EMA, circumscribing completely or nearly completely the individual tumor cells, was typically observed in renal cell carcinomas, both primary tumors and metastases. (Immunoperoxidase technique, methyl green counterstain. x150.)
FIGURE 6 (bottom right). Papillary adenocarcinoma, primary in lung; formalin fixation. Immunoreactivity for EMA is localized predominantly to the superficial aspect of the papillae, with strong apical membrane and apical cytoplasmic staining. Nuclei and connective tissue cores of papillae are nonreactive. (Immunoperoxidase technique, methyl green counterstain. x260.)
showed strong immunoreactivity for EMA (fig. 8). In these cases the finely granular cytoplasmic staining that was also present appeared less intense than the surface membrane staining and was frequently po- larized toward the epithelial surface. Cytoplasmic staining for EMA also varied topographically, being most prominent in superficial cells and least apparent in the most basally oriented cells. Diffuse intracyto- plasmic staining was seen in all five invasive bladder tumors, which, like the papillary tumors, exhibited topographic variability. Cells that appeared to be more differentiated tended to have more cytoplasmic staining than intermediate cells, and only rare basa- loid cells stained. In one case, half of the positive cells had an eccentric punctate intracytoplasmic staining pattern.
Two transitional cell carcinomas (both in the bladder), one metastatic to the epidural region and the other to lymph node, showed strong cytoplasmic staining. In the lymph node metastasis, linear mem- brane staining demarcated clusters of tumor cells from the surrounding stroma (fig. 9).
Small cell anaplastic carcinomas. Twelve small cell anaplastic carcinomas of pulmonary origin-three primary and nine metastatic (six nodal and one each in liver, breast, and bone marrow)-and one Merkel cell tumor metastatic to bone marrow were evaluated. Specimens were fixed in formalin (eight), formalin and B-5 (one), B-5 (one), or Zenker’s solution (two). In seven of the small cell carcinomas and the Merkel cell tumor, diffuse granular cytoplasmic staining for EMA was present in 20 to 90 per cent of the neo-
| Tumor Type | No. Evaluated | No. Immunoreactive for EMA |
|---|---|---|
| Small cell anaplastict | 12 | 12 |
| Merkel cell tumor | 1 | 1 |
| Transitional cell carcinoma# | 12 | 12 |
| Squamous cell carcinoma§ | 13 | 13 |
| Paget's disease, extramammary | 1 | 1 |
| Salivary gland | ||
| Mixed tumor | 3 | 2 |
| Warthin's tumor | 2 | 2 |
| Acinic cell tumor | 1 | 0 |
| Hepatoma | ||
| Hepatocellular | 1 | 0 |
| Cholangiocarcinoma | 2 | 2 |
| Mesothelioma | 6 | 6 |
| Synovial sarcoma | 3 | 3 |
| Epithelioid sarcoma | 1 | 1 |
| Neuroblastoma[ | 2 | 1 |
| Chordoma | 1 | 1 |
| Totals | 61 | 57 |
* Specimens were fixed in formalin (51), B5 (2), Zenker’s (2), both B5 and formalin (4), or both formalin and Zenker’s solu- tion (2).
t Lung, primary (3) and metastatic (9).
# Bladder, primary (8) and metastatic (2); kidney (1); ure- ter (1).
§ Skin, in situ (5) and invasive (3); lung (4); uvula (1).
” Bone marrow metastasis, fixed in Zenker’s.
plastic cells (figs. 10 and 11). In two cases focal accen- tuation of staining between cells was also observed. In five cases multiple small punctate granules or small globules of cytoplasmic immunoreactivity for EMA were present (fig. 12). In one tumor pseudorosettes formed, with strong granular staining in the center of the Tosettes (fig. 13).
Extramammary Paget’s disease. One tumor (for- malin-fixed) involving the perineal tissues was ex- amined. Strong EMA staining was observed in the cytoplasm of neoplastic cells within the epithelium (fig. 14).
Neuroblastoma. One of the two neuroblastomas examined (both metastatic to bone marrow and fixed ‘in Zenker’s solution) showed granular cytoplasmic staining for EMA (fig. 15).
Salivary gland tumors. Six salivary gland tumors (three mixed tumors, two Warthin’s tumors, and one acinic cell tumor; all formalin-fixed) were evaluated.
FIGURE 12 (bottom left). Small cell anaplastic carcinoma of lung; formalin fixation. Punctate juxtanuclear immunoreactivity for EMA is observed in many tumor cells. (Immunoperoxidase technique, hematoxylin counterstain. x 650.)
FIGURE 13 (bottom right). Small cell anaplastic carcinoma of lung, metastatic to lymph node; formalin fixation. An infrequent staining pattern observed in this type of neoplasm consisted of localized cytoplasmic immunoreactivity for EMA, polarized toward the center of the pseudorosettes formed by tumor cells. Focal punctate cytoplasmic staining is also apparent. (Immunoperoxidase technique, methyl green counterstain. x230.)
Myoepithelial cells and the myxoid matrix were uni- formly nonreactive in the mixed tumors. In the glan- dular elements of the mixed tumors and in the co- lumnar epithelial cells of the Warthin’s tumors, gran- ular apical cytoplasmic staining was associated with linear apical membrane staining. The acinic cell tumor was nonreactive for EMA.
Hepatomas. Immunoreactivity for EMA was ap- parent in both of the cholangiocarcinomas analyzed, with cytoplasmic and apical membrane staining. The single hepatocellular hepatoma evaluated was non- reactive. In uninvolved hepatic tissue, only bile duct epithelium was EMA-positive (apical membrane staining).
Mesotheliomalsynovial sarcoma. The patterns of EMA reactivity in these two biphasic tumor types were similar. In the epithelioid components of all three synovial sarcomas (formalin-fixed), strong linear apical membrane staining for EMA in cells
forming gland-like structures was accompanied by weak, finely granular cytoplasmic staining. Spindle cells also revealed cytoplasmic staining for EMA, with accentuation at the cell membrane, although the numbers of immunoreactive cells varied considerably from case to case (from 1-2 per cent to 30-40 per cent).
In five of six pleural mesotheliomas (four fixed in formalin and two in formalin and Zenker’s solu- tion), most of the tumor cells revealed linear apical membrane staining for EMA, with associated diffuse cytoplasmic staining in four cases (fig. 16) and focal membrane staining. In the remaining case, only oc- casional cells had granular cytoplasmic EMA posi- tivity. Spindle cells were immunoreactive in two cases. The proportion of cells positive for EMA was greater in the tumors fixed in Zenker’s solution than in those fixed in formalin.
Epithelioid sarcoma. In the single specimen ex-
| No. Evaluated | No. Immunoreactive for EMA | |
|---|---|---|
| T-cell neoplasmst | 35 | 6¢ |
| B-cell lymphomas§ | 33 | 0 |
| True histiocytic lymphomas | 3 | 3 |
| Hodgkin's disease" | 15 | 0# |
| Lymphoblastic lymphomas ** | 3 | 0 |
| Acute lymphoblastic leukemia | 1 | 0 |
| Acute monocytic leukemia | 1 | 0 |
| Acute myelogenous leukemia | 1 | 0 |
| Polycythemia vera | 1 | 0 |
| Myeloma/plasmacytoma | 7 | 7tt |
| Macroglobulinemia | 3 | 2## |
* Specimens were fixed in B5 solution (69), Zenker’s solution (19), formalin (11), Bouin’s (1), Bayley’s fixative (2), or multiple fixatives (1).
t All T-cell neoplasms were immunologically phenotyped; there were 22 immunoblastic sarcomas, 7 lymphocpithelioid cell lymphomas, 2 lymphocytic leukemias, one lymphoblastic lym- phoma, I mycosis fungoides (nodal), 2 unclassified T-cell lym- phomas.
# Immunoreactivity was observed in immunoblastic sarcomas, with strong membrane staining of most neoplastic cells in only I case. Punctate cytoplasmic staining was evident in about 40-50% of the cells in another case, with only rare cells showing membrane and punctate cytoplasmic staining in the remaining 4 cases.
§ The B-cell lymphomas included 23 follicular center cell lym- phomas of small cleaved (5), large cleaved (3), mixed follicular center cell (1), small noncleaved (2), and large noncleaved (12) types; 9 B-cell immunoblastic sarcomas; 1 small lymphocytic lym- phoma. The B-cell derivation was confirmed by immunologic studies in 24 cases.
” Tumors included 8 nodular sclerosis, 5 mixed cellularity, and 2 with lymphocyte predominance (nodular).
# Reed-Sternberg cells and variants were nonreactive, al- though occasional plasma cells exhibited EMA positivity.
** Immunologic studies unavailable.
tt Variable numbers of plasmacytic cells were reactive for EMA; in most cases minor populations were positive, although a majority of cells were immunoreactive in one gastric plasmacy- toma.
EMA staining observed for occasional plasmacytic cells.
amined (formalin-fixed), nearly all of the neoplastic cells were immunoreactive for EMA, with strong membrane staining and weak cytoplasmic staining (fig. 17).
Hematopoietic Neoplasms Reactive for EMA
Non-Hodgkin’s lymphomas. A total of 74 non- Hodgkin’s lymphomas of a wide variety of cell types were evaluated (table 3). B-cell lymphomas (33 cases) were EMA-negative. However, all three of the true histiocytic lymphomas and six of the 35 T-cell neo- plasms were immunoreactive for EMA, with mem- brane and/or cytoplasmic staining (diffuse and/or punctate; figs. 18 and 19). Although most of the neo- plastic cells in the true histiocytic lymphomas were immunoreactive for EMA (fig. 18), the majority of large lymphoid cells were positive in only one of the six immunoreactive T-cell immunoblastic sarcomas (fig. 19, top left and right). In one of the other EMA- positive T-cell lymphomas, about 40 to 50 per cent
of the large neoplastic cells revealed punctate cyto- plasmic staining (fig. 19, bottom right). In the re- maining four cases membrane and/or cytoplasmic staining was seen only in rare cells. A definite histio- cytic origin was established for the three true histio- cytic lymphomas by histochemical, immunologic, and ultrastructural techniques. Further studies to assess the immunoreactivity of histiocytes in other disorders indicated that the staining pattern observed in the histiocytic lymphomas was unusual. Histiocytes of various types (sinusoidal, interdigitating, dendritic, and tingible body-type) in reactive nodes (two speci- mens) or tonsil (one specimen) were nonreactive. In one of three nodes involved by sarcoidosis, occasional histiocytes (less than 10 per cent) constituting gran- ulomas or in the sinusoidal network of the node showed focal cytoplasmic staining. Membrane staining was not observed. In two cases of toxoplasmic lymphadenitis, epithelioid histiocytes, as well as those in normal structures of the node, were negative.
Hematopoietic elements involved in myelopro- liferative disorders were nonreactive for EMA (table
.
3), although some plasma cells present in these spec- imens were positive.
Hodgkin’s disease. Reed-Sternberg cells and var- iants in 15 cases of Hodgkin’s disease (eight nodular sclerosis, five mixed cellularity, two nodular lympho- cyte predominance) were EMA-negative (table 3). Oc- casional plasma cells exhibited membrane staining for EMA.
Myelomalplasmacytomalmacroglobulinemia. Plas- macytic cells in six myelomas (bone marrow biopsy specimens; Zenker’s fixation), one gastric plasmacy- toma (B-5 fixation), and two of three marrow speci- mens (Zenker’s fixation) involved by macroglobuli- nemia were immunoreactive for EMA (table 3). The numbers of positive cells were highly variable. In some cases only occasional plasmacytic cells showed membrane (the predominant pattern) or focal punc- tate cytoplasmic staining. However, in the gastric plasmacytoma, the majority of cells were reactive for EMA (fig. 20). Immunoreactivity of plasma cells ap- peared to be independent of light or heavy chain im- munoglobulin determinants; cases evaluated in- cluded IgA, K (three cases); IgA, À (one case); IgG, K (one case); IgG, À (one case), and IgM, K (three cases). Residual hematopoietic cells were nonreactive.
Neoplasms Lacking Immunoreactivity for EMA
Fifty-nine neoplasms that fell into several general categories (table 4) were not immunoreactive for
EMA. Certain endocrine tumors (17 evaluated), i.e., pheochromocytomas, adrenocortical carcinomas, car- cinoid tumors, and medullary carcinoma of the thy- roid, and a wide variety of soft tissue tumors (24 eval- uated), with the exception of synovial sarcomas and an epithelioid sarcoma (table 2), lacked immu- noreactivity for EMA. Germ cell tumors, except for glandular or squamoid foci in a teratoma, were EMA- negative. Malignant melanomas (predominantly ame- lanotic; six fixed in formalin, one in B-5, and one in formalin and Zenker’s solution) were typically non- reactive. However, in two cases (Zenker’s fixation), focal weak perinuclear staining was observed in oc- casional cells.
Other Observations
Inspindle cells surrounding nerve fiber bundles, i.e., perineural cells, weak to moderate granular cy- toplasmic staining for EMA was also seen. This pat- tern was most prominent in a traumatic neuroma but was also observed in myelinated and nonmyelinated nerve, regardless of tissue site.
Occasional nonneoplastic plasma cells (variable numbers; generally minor populations) also showed membrane staining for EMA, as in the myelomas (table 3). Focal weak cytoplasmic staining was also ob- served. In studies in which irrelevant monoclonal an- tibodies, e.g., anti-Factor VIII, or TRIS buffer were substituted for anti-EMA antibodies, plasma cell
membranes did not stain, verifying that staining was related to monoclonal antibodies to EMA (rather than to other reagents in the procedure). However, since the scattered EMA-positive plasma cells were readily identifiable as to cell type, this finding did not affect interpretation.
In 20 cases EMA immunoreactivity was evaluated in specimens that had been fixed in 10 per cent neu- tral buffered formalin and one or more other fixa- tives (B-5, Zenker’s, or Bouin’s solution). In general, but not in all cases, formalin appeared to be less ef- fective than the other fixatives for EMA preservation. However, the results obtained with formalin-fixed tis- sues were acceptable and readily interpretable. Pre- liminary trypsinization of formalin-fixed sections was also evaluated in an attempt to optimize staining. These results were variable, with enhanced staining in some tumors, diminished staining in some tumors, and no discernible effect in others.
Background staining with the techniques used was minimal to absent, providing sections of excellent quality for the detection of even focal staining. Con- trol sections, with TRIS buffer substituted for the monoclonal antibody, did not stain. Substitution of an irrelevant monoclonal antibody, i.e., anti-Factor VIII-associated antigen, for the EMA antibodies re- sulted in only the anticipated staining of endothelial cells or megakaryocytes.
DISCUSSION
Monoclonal antibodies to epithelial membrane antigen, glycoproteins isolated from human milk fat globule membranes, are a highly effective reagent for determining the epithelial nature of neoplastic cells. In general, our study of EMA immunoreactivity in 320 neoplasms with monoclonal antibodies parallels the EMA distribution in various tumors observed by Sloanne and Ormerod5 with polyclonal (rabbit) anti-
bodies. Most adenocarcinomas were strongly immu- noreactive for EMA, with cytoplasmic and apical membrane staining being the most common patterns of reactivity (table 1). Adenocarcinomas of prostatic or thyroid derivation appeared to be the least reactive for this marker. However, more specific antibodies, e.g., against prostatic acid phosphatase or prostate- specific epithelial antigen or against thyroglobulin, are available for the identification of neoplasms from these sites. Other epithelial neoplasms of squamous cell or transitional cell types were also immunoreac- tive for EMA, as were small cell anaplastic carcinomas and neuroblastomas (table 2). The basis for staining in the latter two tumor types is obscure. However, Sloane and Ormerod5 previously described EMA reactivity in most of these tumor types, including small cell anaplastic tumors, although staining of neu- roblastomas was not observed in their study.
Epithelial membrane antigen is potentially an ex- tremely useful diagnostic discriminant for defining the epithelial nature of neoplasms, including those with spindle cell growth patterns, and for distin- guishing between malignant lymphomas and poorly differentiated epithelial neoplasms. Although an ab- sence of EMA immunoreactivity would only support a diagnosis of lymphoma, a combined approach to anaplastic tumors that also includes monoclonal an- tibodies to leukocyte common antigen (LCA), a
| Tumor Type | No. of Tumors |
|---|---|
| Soft tissue tumors | |
| Fibromatosis | 3 |
| Ewing's sarcoma | 1 |
| Leiomyoma | 1 |
| Leiomyosarcoma | 4 |
| Rhabdomyosarcoma (embryonal) | 4 |
| Lipoma | 1 |
| Lymphangioma | 1 |
| Malignant fibrous histiocytoma (plcomorphic) | 1 |
| Traumatic neuroma | 1 |
| Neurofibroma | 4 |
| Schwannoma | 2 |
| Granular cell tumor | 1 |
| Endocrine tumors | |
| Adrenocortical carcinoma | 7 |
| Pheochromocytoma, adrenal | 1 |
| Medullary carcinoma, thyroid | 1 |
| Carcinoid tumort | 8 |
| Germ cell tumors | |
| Embryonal carcinoma | 7# |
| Seminoma | 1 |
| Others | |
| Malignant melanoma | 8§ |
| Eosinophilic granuloma | 2 |
* Tissues were fixed in formalin (56), B5 solution (2), or for- malin and Zenker’s solution (1).
+ Bronchus (4), small intestine (3), colon (1).
# In 2 tumors with components of mature teratoma, foci of squamoid and glandular differentiation were EMA-positive.
§ Focal weak perinuclear staining in 2 cases; only tissue fixed in Zenker’s solution was reactive in 1 case.
marker predominantly for lymphoid cells, 14 should permit distinction between these neoplasms. In ad- dition, since occasional T-cell lymphomas and true histiocytic lymphomas may be immunoreactive for EMA, evaluation of LCA is recommended as a com- plementary technique. In our previous study, in which monoclonal antibodies to ICA (combined PD7/ 26 and 2B11) were used, most non-Hodgkin’s lym- phomas were LCA-positive, while epithelial and mes- enchymal neoplasms were LCA-negative.14 Gatter et al.15 also suggested that a battery of monoclonal an- tibodies may be helpful in the assessment of ana- plastic tumors.
The EMA immunoreactivity of small cell ana- plastic carcinomas is a particularly noteworthy finding. Our 12 small cell anaplastic tumors were uni- formly EMA-positive, as were 25 of 47 similar tumors evaluated by Sloane and Ormerod.3 A marker that is capable of characterizing these tumors, particularly in paraffin sections, has definite relevance in diag- nostic pathology. Alternatively, small cell neoplasms of lymphoid derivation may be readily distinguished from these tumors with antibodies to LCA.14
Certain tumors, on the basis of our studies and the previous report of Sloane and Ormerod,5 warrant comment with regard to EMA immunoreactivity. Ma- lignant melanomas were typically negative for EMA, although focal weak staining was apparent in two of our cases. The lack of EMA staining, complemented by immunoreactivity for S-100 protein and an ab- sence of keratin proteins, would strongly support a diagnosis of melanoma. Immunoreactivity for EMA in neoplasms of possible germ cell origin would ex- clude embryonal carcinoma and seminoma. Since soft tissue tumors, with the exception of synovial sarcomas and an epithelioid sarcoma, were uniformly nonreac- tive for EMA, EMA positivity for a spindle cell neo- plasm would suggest epithelial rather than soft tissue derivation, unless morphologic or clinical features raised the possibility of either of these sarcomas. In- terestingly, synovial sarcomas and epithelioid sar- comas are also the soft tissue tumors that are uniquely immunoreactive for keratin proteins.16,17 A further diagnostic application of EMA staining patterns re- lates to renal cell carcinomas, which may show spindle cell growth patterns. Although these tumors are gen- erally nonreactive for keratin proteins (with callus- derived keratin antibodies), strong immunoreactivity for EMA is typically observed (fig. 5). In addition, for tumors in which renal or adrenal origins cannot be ascertained, EMA positivity would support renal de- rivation. Staining patterns for EMA in primary he- patic tumors of the hepatocellular type differ from those in the cholangiocarcinomatous type, in that only tumors of the latter type are EMA-positive, as found in our hepatic tumors and in a larger series evaluated by Bonetti et al. 18
Within the group of non-Hodgkin’s lymphomas, the unusual cases of true histiocytic lymphoma are of special interest. These tumors were EMA-positive (table 2) and also co-expressed LCA (unpublished ob- servation). Using three different monoclonal anti-
bodies to EMA (including E29), Delsol et al.19 simi- larly observed EMA positivity in neoplastic histiocytes in two cases of malignant histiocytosis. In the latter study, seven of 13 T-cell lymphomas were also EMA- positive. The numbers of immunoreactive cells in their cases varied considerably (5 to 80 per cent), with the majority of large neoplastic cells showing immu- noreactivity in only two T-cell lymphomas. Fewer T- cell neoplasms (six of 35) were positive in our study, with most of the cells showing immunoreactivity in only one case. The basis for this difference is not entirely clear. Delsol et al.19 did indicate that EMA immunoreactivity was not identical for all trans- formed mononuclear cells in peripheral blood. Only 2 to 3 per cent of phytohemagglutin-stimulated cells were EMA-positive, while 80 per cent expressed EMA after transformation with human T-cell leukemia virus (HTLV)-II. Since the immunohistochemical staining techniques were relatively similar in the two studies, discrepancies may be related to differences inherent in the tumor cells, to the different fixatives used, or, possibly, to the use of cryostat sections (rather than paraffin sections) in some of the previ- ously reported cases. Most of our tumors were fixed in B-5, which, in our experience, optimizes EMA reactivity in paraffin sections, as compared with for- malin. Some differences are also apparent for our EMA studies of Hodgkin’s disease, as compared with the previously reported study. 19 Reed-Sternberg cells and variants were nonreactive for EMA in our cases, including two of the lymphocyte predominance type. In their larger series (46 cases), EMA immunoreac- tivity was observed for Reed-Sternberg cells in 14 of 20 cases of Hodgkin’s disease of the lymphocyte pre- dominance type. However, in both studies the infil- trates in tumors of the mixed cellularity and nodular sclerosis types were EMA-negative, except for staining of plasma cells.
The EMA staining pattern observed for neo- plastic (fig. 20) and nonneoplastic plasma cells was unanticipated, though Delsol et al.19 reported similar findings. Unlike the B-cell neoplasms, including im- munoblastic sarcomas, which were clearly nonreactive for EMA (table 3), some plasma cells obviously con- tained antigenic determinants, mainly related to the cell membrane, that were recognized by monoclonal antibodies to EMA. This staining pattern could not be correlated with specific immunoglobulin heavy or light chains.
Although normal and reactive mesothelial cells are either reportedly weakly positive or nonreactive for EMA,5 neoplastic cells of mesotheliomas are EMA-positive5 (table 2). The relative lack of EMA staining in nomneoplastic mesothelial cells has been used to advantage in the identification of malignant epithelial cells in serous effusions, since the latter cells typically exhibit strong EMA immunoreactivity. 20,21 Further assessment of EMA as a diagnostic aid in cytologic analysis appears to be warranted. Immu- nohistochemical detection of EMA-positive cells has also been used to advantage in characterizing small numbers of tumor cells in bone marrow smears.22
However, since occasional plasma cells in our study were EMA-positive, cytologic features of imunoreac- tive cells must be evaluated carefully.
Epithelial membrane antigen immunoreactivity, as detected by monoclonal antibodies in this study and by rabbit antiserum in earlier reports,5 is well preserved in routinely processed, formalin-fixed, paraffin-embedded tissues. Immunoreactivity in for- malin sections is occasionally, but not consistently, en- hanced by preliminary trypsinization. Comparative studies in which other fixatives were used demon- strated relatively stronger staining following fixation in B-5, Zenker’s, or Bouin’s solution. However, the results for routinely processed material were gener- ally satisfactory to excellent. The ability to apply this commercially available reagent with a broad epithelial staining profile to paraffin sections of routine mate- rial makes this diagnostic approach particularly at- tractive.
In summary, monoclonal antibodies to EMA are a highly effective diagnostic tool for the detection of poorly differentiated neoplasms of epithelial deriva- tion, as well as small cell anaplastic tumors. With monoclonal antibodies to LCA used as a complemen- tary technique, it is possible to discriminate accurately between lymphoid and epithelial derivations for an- aplastic tumors in paraffin-embedded tissues.
Acknowledgments. The authors acknowledge Jonathan Said, MD (Cedars-Sinai Medical Center, Los Angeles, Cal- ifornia) for contributing material on T-cell lymphomas, and Constance Etheridge and Erin O’Connor for expert tech- nical assistance.
REFERENCES
1. Dowben RM, Brunner JR, Philpott DE: Studies on milk fat globule membranes. Biochim Biophys Acta 135:1, 1967
2. Ceriani RL, Thompson K, Peterson JA, et al: Surface differ- entiation antigens of human mammary epithelial cells car- ried on the human milk fat globule. Proc Natl Acad Sci USA 74:582, 1977
3. Heyderman E, Steele K, Ormerod MG: A new antigen on the epithelial membrane: its immunoperoxidase localization in normal and neoplastic tissue. J Clin Pathol 32:35, 1979
4. Taylor-Papadimitriou J, Peterson JA, Arklie J, et al: Mono- clonal antibodies to epithelium-specific components of the human milk fat globule membrane: production and reaction with cells in culture. Int J Cancer 28:17, 1981
5. Sloane JP, Ormerod MG: Distribution of epithelial membrane antigen in normal and neoplastic tissues and its value in diagnostic tumor pathology. Cancer 47:1786, 1981
6. Ormerod MG, Steele K, Westwood JH, et al: Epithelial mem- brane antigen: partial purification, assay and properties. Br J Cancer 48:533, 1983
7. Shimizu M, Yamauchi K: Isolation and characterization of mucin-like glycoprotein in human milk fat globule mem- brane. J Biochem 91:515, 1982
8. Ormerod MG, Bussolati G, Sloane JP, et al: Similarities of antisera to casein and epithelial membrane antigen. Vir- chows Arch [A] 397:327, 1982
9. Sloane JP, Ormerod MG, Imrie SF, et al: The use of antisera to epithelial membrane antigen in detecting micrometas- tases in histological sections. Br J Cancer 42:392, 1980
10. Sloane JP, Ormerod MG, Carter RL, et al: An immunocyto- chemical study of the distribution of epithelial membrane antigen in normal and disordered squamous epithelium. Diagn Histopathol 5:11, 1982
11. Foster CS, Edward PAW, Dinsdale EA, et al: Monoclonal an- tibodies to the human mammary gland. I. Distribution of determinants in non-neoplastic mammary and extra mam- mary tissues. Virchows Arch [A] 394:279, 1982
12. Foster CS, Dinsdale EA, Edwards PAW, et al: Monoclonal an- tibodies to the human mammary gland. II. Distribution of determinants in breast carcinomas. Virchows Arch [A] 394:295, 1982
13. Arklie J, Taylor-Papadimitriou J, Bodmer W, et al: Differen- tiation antigens expressed by epithelial cells in the lactating breast are also detectable in breast cancers. Int J Cancer 28:23, 1981
14. Kurtin PJ, Pinkus GS: Leukocyte common antigen-a diag- nostic discriminant between hematopoietic and nonhema- topoietic neoplasms in paraffin sections using monoclonal antibodies: correlation with immunologic studies and ultra- structural localization. HUM PATHOL 16:353, 1985
15. Gatter KC, Alcock C, Heryet A, et al: The differential diag- nosis of routinely processed anaplastic tumors using mono- clonal antibodies. Am J Clin Pathol 82:33, 1984
16. Corson JM, Weiss LM, Banks-Schlegel SP, et al: Keratin pro- teins and carcinoembryonic antigen in synovial sarcomas: an immunohistochemical study of 24 cases. HUM PATHOL 15:615, 1984
17. Chase DR, Weiss SW, Enzinger FM, et al: Keratin in epithe- lioid sarcoma-an immunohistochemical study. Am J Surg Pathol 8:435, 1984
18. Bonetti F, Chilosi M, Pisa R, et al: Epithelial membrane an- tigen expression in cholangiocarcinoma. A useful immu- nohistochemical tool for differential diagnosis with hepa- tocarcinoma. Virchows Arch [A] 401:307, 1983
19. Delsol G, Stein H, Pulford KAF, et al: Human lymphoid cells express epithelial membrane antigens. Lancet 2:1124, 1984
20. To A, Coleman DV, Dearnaley DP, et al: Use of antisera to epithelial membrane antigen for the cytodiagnosis of malig- nancy in serous effusions. J Clin Pathol 34:1326, 1981
21. To A, Dearnaley DP, Ormerod MG, et al: Epithelial mem- brane antigen. Its use in the cytodiagnosis of malignancy in serous effusions. Am J Clin Pathol 78:214, 1982
22. Dearnaley DP, Sloane JP, Ormerod MG, et al: Increased de- tection of mammary carcinoma cells in marrow smears using antisera to epithelial membrane antigen. Br J Cancer 44:85, 1981