The cell adhesion molecule, E-cadherin, distinguishes mesothelial cells from carcinoma cells in fluids

293CANCERCYTOPATHOLOGY

The Cell Adhesion Molecule, E-Cadherin,Distinguishes Mesothelial Cells from Carcinoma Cellsin Fluids

BACKGROUND. The distinction between benign reactive mesothelial cells and wellKevin Schofield, B.S. C.T. (ASCP)

differentiated carcinoma can be difficult in pleural, peritoneal, and especially peri-Thomas D’Aquila, B.S.cardial fluids. E-cadherin is an adhesion protein that is specifically expressed inDavid L. Rimm, M.D., Ph.D.cells of epithelial lineage. In this study, anti-E-cadherin antibodies were used to

identify and distinguish carcinoma cells from reactive mesothelial cells.Department of Pathology, Yale UniversityMETHODS. Pleural, peritoneal, and pericardial fluids were prepared using the CytycSchool of Medicine, New Haven, Connecticut.Thin Prepy processor. The specimens were comprised of a mix of 45 cases that

were diagnosed as carcinoma, suspicious, or reactive by Papanicolaou staining of

routine material seen by the authors’ service. Routine immunologic techniques

were used with a commercially available E-cadherin antibody.

RESULTS. In most cases of carcinoma, tumor cells showed a strong positive mem-

branous reaction product (32 of 37). This included four cases that were not cyto-

morphologically diagnosed as malignant, but subsequently proved to be malignant.

E-cadherin staining was not observed in five tumors, two of which were not ex-

pected to express this protein. One benign case showed cells staining for E-cad-

herin, although the cells were not malignant by morphologic criteria. Because

this case was a surgical pelvic washing, these cells more likely were epithelial

contaminants than true false-positives.

CONCLUSIONS. The epithelial specific cell-cell adhesion marker E-cadherin reliably

distinguishes reactive mesothelial cells from carcinoma and is a useful adjunctive

test to distinguish benign reactive mesothelial cells from well differentiated carci-

noma cells in fluid specimens. Cancer (Cancer Cytopathol) 1997;81:293–8.

q 1997 American Cancer Society.

KEYWORDS: adhesion, cadherin, effusion, cytopathology, Thin Prepy.

APresented as a poster at the American Societylthough cytologic distinction between metastatic carcinoma andreactive mesothelial cells in pleural, peritoneal, and pericardial

of Cytopathology Meeting, Denver, Colorado, fluids can be very difficult, the staging of tumors and subsequentNovember 5–9, 1996. treatment of patients depends on it. One reason for the difficulty is

that mesothelial cells can take on a wide range of so-called ‘‘atypical’’Supported by grants from the William and Cath-and/or ‘‘reactive’’ changes in response to many different forms oferine Weldon Donaghue Foundation for Medicalstimuli, including infection, cirrhosis, and pneumonia. The goal ofResearch and the U.S. Army (DAMD17-94-

J4366). this study was to find a marker that could reliably and reproduciblydistinguish reactive mesothelial cells from neoplastic cells.

Address for reprints: David L. Rimm M.D., One method for distinguishing between carcinoma and benignPh.D., Dept. of Pathology, Yale University mesothelial cells is immunostaining. Typically, a panel including anti-School of Medicine, 310 Cedar St., New Haven,

bodies to keratin, CD15 (LeuM1), carcinoembryonic antigen (CEA),CT 06510.and occasionally BerEP4 or B72.3 is used. Nearly all carcinomas arepositive for keratin, but so are most mesothelial cells, both reactiveReceived May 29, 1997; accepted June 26,

1997. and malignant. CEA is more helpful because nearly all mesothelial

q 1997 American Cancer Society

/ 730a$$78dy 10-09-97 09:38:57 ccyta W: Can Cyto

294 CANCER (CANCER CYTOPATHOLOGY) October 25, 1997 / Volume 81 / Number 5

cells are negative for CEA. The main weakness of this included cytologic diagnoses of carcinoma, suspi-marker is that only 50–67% of the adenocarcinomas cious, atypical, and reactive by Papanicolaou (Pap)express it.1,2 CD15 (LeuM1) also is helpful because stain. Each case had surgical follow-up or a previousmany carcinomas stain with this marker but mesothe- specimen in the surgical pathology files. All cases werelial cells do not.3 Similarly, both BerEP44 and B72.35 obtained from the Yale Pathology Department’s Criti-have been shown to selectively recognize adenocarci- cal Technologies Cytology Collection with permissionnomas. Unfortunately, these methods are not much from the Yale Human Investigation Committee, proto-better than the simple, old-fashioned method of peri- col #8219. The cases are summarized in Table 1.odic acid–Schiff diastase staining, which reveals ap- All pleural, pericardial, and pelvic/peritoneal flu-proximately 50% of the adenocarcinomas, and when ids were submitted for routine cytologic examinationdone properly, is negative in mesothelial cells. As a in the usual manner. Specimens were spun down andresult, cocktails of multiple markers have been tried the supernatant fluid was discarded. The pellet wasand even logistic regression analysis has been used to then resuspended in CytoLyty (Cytyc Corp., Boxbor-select the best panel.6 To the authors’ knowledge, to ough, MA) and centrifuged again. The supernatantdate, there is no single marker in common usage that fluid was decanted again and 2 drops of the pelletdefinitively separates cells of epithelial lineage from were placed in PreservCyty (Cytyc Corp., Boxborough,those of mesenchymal lineage. MA) solution. After 15 minutes, the specimen was pro-

E-cadherin, an epithelial specific homotypic adhe- cessed using the Cytyc Thin Prepy (Cytyc Corp., Box-sion protein, has the potential to be a marker with borough, MA) processor. Finally, the specimens werehigh sensitivity and specificity for the detection of car- stained using the Pap staining technique for routinecinoma cells. It is a 120-kilodalton transmembrane screening by the cytotechnologist and sign-out by theglycoprotein whose calcium sensitive homotypic ad- pathologist.hesion is the primary stabilizing interaction in cell- To evaluate E-cadherin expression, at least twocell adhesion and a signal for polarization and cell

Thin-Prep slides were made from each case. One wasdifferentiation.7 It has been shown that interruption of

treated with E-cadherin antibody and the other wasthe function of any of the components of the cadherin-

used as a negative control. Thin-Prep slides werebased transmembrane complex leads to loss of epithe-

rinsed for 5 minutes in tap water followed by 5 minuteslial cell-cell adhesion.8 Cells of mesenchymal origin,

in Tris-buffered saline (TBS) (150 mM NaCl and 20like mesothelial cells, express a related cadherin called

mM Tris [pH Å 8]). Slides then were blocked for 20N-cadherin, and do not express E-cadherin.9,10 E-cad-minutes with diluted normal serum from the Vectas-herin was used in this study to mark tumors of epithe-tain ABC-AP Kit (Vector Laboratories, Burlingame,lial lineage. Although E-cadherin is lost in some tu-CA). After blocking, slides were washed once in 1Xmors (e.g., lobular carcinoma of the breast11 and gas-TBS. E-cadherin monoclonal antibody (Transductiontric signet ring cell carcinoma12), nearly all well differ-Laboratories, Lexington, KY) was diluted 1:250 in 1Xentiated epithelial tumors maintain expression of thisTBS, 3–4 drops applied per slide, and then incubatedmarker. Antibodies to the extracellular domain of E-for 30 minutes. The slides then were washed againcadherin do not cross react with N-cadherin. In thisin 1X TBS for 5 minutes. Then 3–4 drops of dilutedstudy, the authors tested the hypothesis that well dif-biotinylated antibody from the Vectastain ABC-AP kitferentiated carcinoma can be distinguished fromwere applied per slide and incubated for 30 minutes,mesothelial cells by antigenic recognition of E-cadh-followed by a 5-minute wash in 1X TBS. Vectastainerin expression.ABC-AP Reagent was added (3–4 drops) and the slideswere incubated for an additional 30 minutes followedMETHODSby a 5-minute wash in 1X TBS. Vector Red alkalineForty-five specimens were selected from the study ser-phosphatase substrate (Vector Laboratories, Burlin-vice between May 1995 and April 1996. The cases in-game, CA) was applied (3–4 drops) and the slides werecluded 13 pleural fluids, 28 peritoneal/pelvic wash-incubated for 10 minutes. The slides then were washedings, and 4 pericardial fluids. The 45 cases were se-for 5 minutes in tap water, counterstained with hema-lected at random from the routine cytologic materialtoxylin, and coverslipped.seen by the study service over the last year. The cases

Stained slides were scored by the cytotechnologistwere chosen to test the new reagent on a broad rangeand the pathologist. All the cases were scored indepen-of specimens and were not representative of the aver-dently using a binary system: present (positive) or ab-age distribution of cases received during that period.sent (negative). There were no discrepancies in scoringThe cytologic diagnoses used were those assigned at

‘‘sign-out’’ by the attending physician on service. They between the pathologist and the cytotechnologist. Chi-


E-Cadherin Staining of Carcinoma Cells in Fluids/Schofield et al. 295

TABLE 1Summary of Cases in Current Study

Case no. Site Cyto diagnosis E-cad Surgical

1 Peritoneal Adenocarcinoma (/) Adenocarcinoma2 Peritoneal Suspicious (/) Adenocarcinoma3 Pleural Adenocarcinoma (/) Adenocarcinoma4 Pleural Adenocarcinoma (/) Adenocarcinoma5 Peritoneal Adenocarcinoma (/) Adenocarcinoma6 Pleural Adenocarcinoma (/) Adenocarcinoma7 Pericardial Nonsmall cell carcinoma (/) Adenocarcinoma8 Peritoneal Negative (0) Negative9 Pleural Adenocarcinoma-large cell (0) Adenocarcinoma-large cell10 Pericardial Adenocarcinoma (/) Adenocarcinoma11 Pericardial Negative (0) Negative12 Pericardial Negative (0) Negative13 Peritoneal Adenocarcinoma (/) History of adenocarcinoma14 Peritoneal Adenocarcinoma (/) Adenocarcinoma15 Peritoneal Adenocarcinoma (/) Adenocarcinoma16 Pleural Adenocarcinoma (/) History of adenocarcinoma17 Pleural Negative (/) Adenocarcinoma18 Pleural Melanoma (0) Melanoma19 Pleural Negative (0) Negative20 Peritoneal Adenocarcinoma (/) Adenocarcinoma21 Peritoneal Negative (0) Negative22 Pleural Adenocarcinoma (/) Adenocarcinoma23 Pleural Negative (0) Negative24 Peritoneal Adenocarcinoma (/) History of adenocarcinoma25 Pleural Adenocarcinoma-signet ring (/) Adenocarcinoma-signet ring26 Peritoneal Adenocarcinoma (/) Adenocarcinoma27 Pleural Adenocarcinoma (/) Adenocarcinoma28 Peritoneal Adenocarcinoma (/) Adenocarcinoma29 Peritoneal Negative (/) Negative30 Peritoneal Suspicious (0) Adenocarcinoma31 Pleural Atypical (/) Adenocarcinoma32 Peritoneal Dysgerminoma (/) Dysgerminoma33 Pelvic wash Adenocarcinoma (/) Adenocarcinoma34 Pelvic wash Adenocarcinoma (/) Adenocarcinoma35 Pelvic wash Negative (/) Adenocarcinoma36 Pelvic wash Adenocarcinoma (/) Adenocarcinoma37 Pelvic wash Adenocarcinoma (/) Adenocarcinoma38 Pelvic wash Adenocarcinoma (/) Adenocarcinoma39 Pelvic wash Adenocarcinoma (/) Adenocarcinoma40 Pelvic wash Adenocarcinoma (0) Adenocarcinoma41 Pelvic wash Negative (0) Negative42 Pelvic wash Adenocarcinoma (/) Adenocarcinoma43 Pelvic wash Hepatocellular Ca (0) Hepatocellular Ca44 Peritoneal Adenocarcinoma (/) Adenocarcinoma45 Peritoneal Adenocarcinoma (/) Adenocarcinoma

Cyto: cytologic; E-cad: E-cadherin; Ca: carcinoma; /: positive; 0: negative.

square statistics were performed using the StatView membrane, with the minority in biosynthetic pathwaysin the endoplasmic reticulum or Golgi’s complex (not4.5 program for the Apple Macintosh computer.shown). In tumor cells the distribution was more hetero-geneous. Both benign and malignant epithelial cellsRESULTS

Cells treated with E-cadherin antibody are shown in Fig- stained but mesothelial cells, lymphocytes, and in-flammatory cells did not. In each case, accompanyingures 1 and 2. The staining pattern was predominantly

membranous, but in many cases also was cytoplasmic. Pap stained cells and negative controls counterstainedwith hematoxylin showed the specificity of the antibodyIn benign cells the majority of the cadherin was at the



FIGURE 1. Examples of E-cadherin staining at low power (A–C: 1200) and high power (D–I: 1600). The left column (A, D, and G) shows Papanicolaou(Pap) stained images. Equivalent areas from a second Thin Prep are shown in the middle column after E-cadherin staining and detection with VectorRedy chromogen (Vector Laboratories, Burlingame, CA) (B, E, and H). The column on the far right shows a negative control (neg control) for eachcase (C, F, and I).

staining. Figure 1 shows three classic example cases in- by cytologic diagnosis. Although most cases (28 of 32)that were malignant had E-cadherin staining, 4 cases didcluding one with single malignant cells (Fig. 1A-C) and

others in clusters at higher power. Figure 2 shows an not. Thirteen cases were not determined to be malignantby cytology. Of these, three cases that were termed ‘‘suspi-example of a ‘‘difficult’’ case that was determined to

be atypical after the Pap stain. Suspicious clusters, as cious’’ showed malignant cells on a subsequent surgicalprocedure. E-cadherin staining was positive in two ofobserved in the lower Pap stained frame, were rare and

not well preserved. A duplicate Thin Prepy overlaid with these three cases. Of the ten cytologically negative cases,three showed E-cadherin staining. Two of these threeanti-E-cadherin antibody showed rare (two to three per

slide) groups of unambiguous positively stained clusters, cases were found to be positive for ovarian adenocarci-noma on a subsequent surgical specimen, suggesting theyincluding a membranous pattern (Fig. 2B). Some atypi-

cal-appearing cells did not stain, which suggests that were not false-positive cases. On reexamination of thethird case, a pelvic wash specimen from a patient withthey were not epithelial (Fig. 2E), but rather reactive

mesothelial cells. an ovarian abscess, the cells that stained did not appearto be morphologically malignant and may have beenThe results are summarized in Table 1. E-cadherin

staining was not subtle, and easily scored as either posi- squamous cells or other cells of epithelial origin that arti-factually entered the wound during the surgical proce-tive or negative, with no disagreements between the cyto-

technologist and the pathologist. The summary of E- dures. The final tabulation of staining, split by final ob-tainable diagnosis, is shown in Table 3. The correlationcadherin scoring of all the cases is shown in Table 2, split


E-Cadherin Staining of Carcinoma Cells in Fluids/Schofield et al. 297

FIGURE 2. An example of a difficult case. A morphologic diagnosis of ‘‘atypical’’ was made on this case. A single cluster shows bright staining (B)and ‘‘atypical’’ morphology corresponding with a positive follow-up surgical specimen. Papanicolaou (Pap) stained examples (A and D), E-cadherin (E-cad) staining (B and E) and negative controls (neg control) (C and F) are as indicated. Note that some atypical-appearing mesothelial cells were negativefor E-cadherin expression (E) suggesting they were reactive mesothelial cells (original magnification 1400).

TABLE 3TABLE 2Summary of E-Cadherin Scoring for All Cases Final Tabulation of Staining, Based on Final Obtainable Diagnosis

Final diagnosis E-cadherin positive E-cadherin negativeCytologic diagnosis E-cadherin positive E-cadherin negative

Positive for malignancy 28 4 Benign (8) 1 7Malignant (37) 32 5Negative for malignancy 3a 7

Atypical/suspicious 2b 1

a Two of these three cases had borderline ovarian tumors after surgery.b Both of these cases had malignant tumors found during surgery.

common. Specifically, lobular carcinoma of thebreast11,15 and poorly differentiated gastric carcino-mas12 show mutations in the E-cadherin gene. Themajority of other tumors show a reduction or alter-of malignancy with staining for E-cadherin was testedation rather than a complete loss. The results of theusing the chi-square method and had a value of 18.4,current study confirm this expression pattern, in thatwhich suggests a highly significant association (P õmost cases show retained E-cadherin expression, al-0.0001).though its localization sometimes is altered. Otherwork in progress in the study laboratory also confirmsDISCUSSIONthis expression pattern. The authors found that onlyE-cadherin is highly specific for epithelial cells, but1 of 81 ductal carcinomas of the breast truly lost ex-has been termed a tumor (or invasion) suppressor pro-pression, although ú80% of the cases showed sometein13 because it is lost in some malignancies.14 Morealteration in staining (Dillon et al. unpublished data).than 50 articles have been published showing alter-In spite of many articles describing a reduction or al-ations in E-cadherin expression in many types of can-

cer, but true documented mutations are relatively less teration of expression, true loss of expression is rare.



2. Nance KV, Silverman JF. Immunocytochemical panel for theIts expression, albeit altered or reduced, is useful foridentification of malignant cells in serous effusions. Am Jthe diagnosis of carcinoma in fluids.Clin Pathol 1991;95:867–74.

In this study there were four tumors that did not 3. Lauritzen AF. Distinction between cells in serous effusionsstain with the E-cadherin antibody. One of those was using a panel of antibodies. Virchows Arch A Pathol Anat

Histopathol 1987;411:299–304.a melanoma, which was not expected to express E-4. Bailey ME, Brown RW, Mody DR, Cagle P, Ramzy I. Ber-cadherin. A second was a hepatocellular carcinoma.

EP4 for differentiating adenocarcinoma from reactive andThis may be due to true mutation because there isneoplastic mesothelial cells in serous effusions. Comparison

evidence in the literature of E-cadherin gene muta- with carcinoembryonic antigen, B72.3 and Leu-M1. Acta Cy-tions in some cases of hepatocellular carcinoma.16 Fi- tol 1996;40:1212–6.

5. Lauritzen AF. Diagnostic value of monoclonal antibodynally, two other cases of metastatic adenocarcinomaB72.3 in detecting adenocarcinoma cells in serous effusions.failed to stain with E-cadherin antibody. These resultsAPMIS 1989;97:761–6.most likely are simply examples of poorly differenti-

6. Frisman DM, McCarthy WF, Schleiff P, Buckner SB, Nocitoated tumors that have lost or have dramatically re- JD Jr., O’Leary TJ. Immunocytochemistry in the differentialduced expression of E-cadherin such that it is not de- diagnosis of effusions: use of logistic regression to select a

panel of antibodies to distinguish adenocarcinomas fromtectable in the assay used by the authors.mesothelial proliferations. Mod Pathol 1993;6:179–84.Conversely, the authors were able to detect stain-

7. Takeichi M. Morphogenetic roles of classic cadherins. Curring in four cases that were not clearly cytomorphologi-Opin Cell Biol 1995;7:619–27.

cally malignant. In each of these cases, the patients 8. Takeichi M. Cadherin cell adhesion receptors as a morpho-had ovarian tumors. This is consistent with the litera- genetic regulator. Science 1991;251:1451–5.

9. Larue L, Antos C, Butz S, Huber O, Delmas V, Dominis M,ture suggesting that serous ovarian tumors are amonget al. A role for cadherins in tissue formation. Developmentthe most difficult to distinguish from reactive meso-1996;122:3185–94.thelial cells. Notably, these are well differentiated tu-

10. Peralta-Soler A, Knudsen KA, Jaurand MC, Johnson KR,mors and E-cadherin loss is very rare in these tumors.17

Wheelock MJ, Klein S, et al. The differential expression ofGiven these findings, it would be desirable to test N-cadherin and E-cadherin distinguishes pleural mesotheli-

omas from lung adenocarcinomas. Hum Pathol 1995;the sensitivity and specificity of E-cadherin staining.26:1363–9.In this primary study, the authors focused on the col-

11. Berx G, Cletonjansen AM, Strumane K, Deleeuw W, Nolletlection of a range of specimens to find cases in whichF, Vanroy F, et al. E-cadherin is inactivated in a majority of

E-cadherin did not stain the tumor cells. As a result, invasive human lobular breast cancers by truncation muta-the cases chosen are not representative of a popula- tions throughout its extracellular domain. Oncogene

1996;13:1919–25.tion of expected cases, and thus sensitivity and speci-12. Becker KF, Atkinson MJ, Reich U, Becker I, Nekarda H, Siew-ficity statistics may not be valid. A new study is cur-

ert JR, et al. E-cadherin gene mutations provide clues torently underway using sequential cases to assess sensi-diffuse type gastric carcinomas. Cancer Res 1994;54:3845–

tivity and specificity and compare this method with 52.conventional cytomorphologic diagnoses. 13. Vleminckx K, Vakaet L Jr., Mareel M, Fiers W, Van Roy F.

Genetic manipulation of E-cadherin expression by epithelialUltimately, other cadherins, especially N-cadherintumor cells reveals an invasion suppressor role. Cellmay be useful in distinguishing mesothelial cells from1991;66:107–19.tumors in effusions. N-cadherin has been shown to be

14. Behrens J. Cell contacts, differentiation, and invasiveness ofuseful in distinguishing adenocarcinomas from meso- epithelial cells. Invasion Metastasis 1994;14:61–70.theliomas because it is expressed in mesothelial cells, 15. Berx G, Cleton-Jansen AM, Nollet F, de Leeuw WJF, van de

Vijver MJ, Cornelisse C, et al. E-cadherin is a tumour/inva-both benign and malignant, and not in epithelialsion suppressor gene mutated in human lobular breast can-cells.10

cers. EMBO J 1995;14:6107–15.16. Slagle BL, Zhou Y-Z, Birchmeier W, Scorsone KA. Deletion

REFERENCES of the E-cadherin gene in hepatitis B virus positive chinese1. Silverman JF, Nance K, Phillips B, Norris HT. The use of hepatocellular carcinomas. Hepatology 1993;18:757–62.

immunoperoxidase panels for the cytologic diagnosis of ma- 17. Risinger JI, Berchuck A, Kohler MF, Boyd J. Mutations of thelignancy in serous effusions. Diagn Cytopathol 1987;3:134– E-cadherin gene in human gynecologic cancers. Nat Genet

1994;7:98–102.40.


Documents

The cell adhesion molecule, E-cadherin, distinguishes mesothelial cells from carcinoma cells in fluids