Comparative immunohistochemical profile of hepatocellular carcinoma, cholangiocarcinoma, and metastatic adenocarcinoma

Comparative Immunohistochemical Profile of Hepatocellular Carcinoma, Cholangiocarcinoma, and Metastatic Adenocarcinoma SEAN K. LAU, MD, SONAM PRAKASH, MD, STEPHEN A. GELLER, MD, AND RANDA ALSABEH, MD Distinguishing hepatocellular carcinoma (HCC) from cholangiocarcinoma (CC) and metastatic adenocarcinoma (MA) involving the liver can be problematic, often requiring the use of immunohistochemistry to facilitate diagnosis. Hep Par 1, a monoclonal antibody with expression confined primarily to benign and malignant hepatocytes, has recently become commercially available. We evaluated Hep Par 1 along with other immunohistochemical markers used to differentiate HCC, CC, and MA, including AE1/AE3, CAM 5.2, B72.3, monoclonal carcinoembryonic antigen (mCEA), polyclonal CEA (pCEA), alpha-fetoprotein (AFP), factor XIIIa, inhibin, CD10, villin, MOC-31, cytokeratin (CK) 7, CK 19, and CK 20, to determine the markers most useful in differentiating these entities. Forty-two cases of HCC, 9 cases of CC, and 56 cases of MA (24 colon, 15 pancreas, 8 ovary, 5 breast, and 4 stomach) were studied. Hep Par 1 was sensitive and specific for HCC, with 38 of 42 (90%) cases staining positively, whereas reactivity was observed in only 8 of 56 (14%) MAs and 0 of 9 CCs. Though limited somewhat by poor sensitivity, a bile

canalicular pattern of staining with pCEA, CD10, and villin was specific for HCC and was not observed in the other tumors. Lack of mCEA and MOC-31 immunoreactivity was also characteristic of HCCs. CK 19 positivity favored CC over HCC, but was not useful in differentiating CC from MA. Expression of AFP, although observed in only about one third of the cases, favored HCC over CC and MA. CK 7 and CK 20 were also useful in this differential diagnosis, particularly when dealing with MA of colonic origin. AE1/AE3, CAM 5.2, B72.3, inhibin, and factor XIIIa were noncontributory in differentiating these entities. HUM PATHOL 33:1175-1181. Copyright 2002, Elsevier Science (USA). All rights reserved. Key words: hepatocellular carcinoma, cholangiocarcinoma, metastatic adenocarcinoma. Abbreviations: HCC, hepatocellular carcinoma; CC, cholangiocarcinoma; MA, metastatic adenocarcinoma; mCEA, monoclonal carcinoembryonic antigen; pCEA, polyclonal carcinoembryonic antigen.

The histologic diagnosis of hepatocellular carcinoma (HCC) is relatively straightforward when the tumor recapitulates the cytoarchitectural appearance of the normal liver. However, HCC exhibiting a pseudoglandular or poorly differentiated morphology may be difficult to distinguish from cholangiocarcinoma (CC) or metastatic adenocarcinoma (MA) involving the liver. In this setting, immunohistochemistry is often used to assess histologically problematic liver tumors. Because a highly specific marker for HCC has not been identified, the immunohistochemical diagnosis has been based on the use of multiple antibodies with differing sensitivities and specificities in reacting with HCC, CC, or MA.1-17 Hepatocyte paraffin 1 (Hep Par 1), a monoclonal antibody that reacts with a hepatocyte-specific epitope, has recently become commercially available (Dako, Glostrup, Denmark). Hep Par 1 expression appears to be largely confined to HCCs, suggesting that this may be a useful marker for evaluating histologically problematic neoplasms in the liver.12,13,18,19 We evaluated

Hep Par 1, along with other commonly used immunohistochemical stains used to distinguish HCC from CC and MA, in an effort to develop an effective panel of antibodies that could facilitate differentiation of these entities.

From the Department of Pathology and Laboratory Medicine, Cedars-Sinai Medical Center, Los Angeles, CA. Accepted for publication September 4, 2002. Address correspondence and reprint requests to: Randa Alsabeh, MD, Cedars-Sinai Medical Center, Department of Pathology and Laboratory Medicine, 8700 Beverly Boulevard, Room #8725, Los Angeles, CA 90048. Copyright 2002, Elsevier Science (USA). All rights reserved. 0046-8177/02/3312-0005$35.00/0 doi:10.1053/hupa.2002.130104

MATERIALS AND METHODS Forty-two cases of HCC, 9 cases of CC, and 56 cases of MA involving the liver (24 colon, 15 pancreas, 8 ovary, 5 breast, and 4 stomach) were obtained from the archives of the Department of Pathology and Laboratory Medicine, CedarsSinai Medical Center, Los Angeles, CA. The histologic diagnosis of HCC was made on surgically resected (n⫽37) and needle biopsy (n⫽5) specimens. HCCs were classified as well differentiated (n⫽13), moderately differentiated (n⫽20), or poorly differentiated (n⫽9), corresponding to Edmondson and Steiner20 grades I/II, III, and IV, respectively. The HCCs included 1 sclerosing/scirrhous variant and 1 fibrolamellar variant. Cases of CC were characterized histologically by proliferating glands or tubules with an associated fibrous stroma, and were confirmed by clinical exclusion of an extrahepatic primary tumor. In all cases of MA, the primary site was well established histologically by a previously resected extrahepatic primary tumor and/or a clinical history of a known primary tumor outside of the liver. A representative formalinfixed, paraffin-embedded tissue block from each case was identified for immunohistochemical studies. The primary antibodies used are listed in Table 1. Sections were cut at 4 ␮, deparaffinized in xylene, and rehydrated in graded ethanols. Pretreatment procedures to enhance immunostaining were performed for particular antibodies, as indicated in Table 1. Sections were incubated with primary antibodies in a Dako

1175

HUMAN PATHOLOGY

Volume 33, No. 12 (December 2002)

TABLE 1. Antibodies Used for Immunohistochemical Staining Antibody

Source

Type

Dilution

Hepatocyte AFP pCEA mCEA Factor XIIIa Inhibin CD10 Villin MOC-31 CK 7 CK 19 CK 20 B72.3 CAM 5.2 AE1/AE3

Dako Dako Dako Dako Calbiochem Serotech Novocastra Biocare Dako Dako Dako Dako Biogenex Becton Dickinson Boehringer-Mannheim

Monoclonal Polyclonal Polyclonal Monoclonal Polyclonal Monoclonal Monoclonal Monoclonal Monoclonal Monoclonal Monoclonal Monoclonal Monoclonal Monoclonal Monoclonal

Auto-Stainer, with antibody localization done through a standardized 2-step method using the Dako EnVision⫹ System and 3,3⬘-diaminobenzidine as a chromogen. A blocking procedure for endogenous biotin was performed following the manufacturer’s instructions (Oxford Bio-Innovation, Oxfordshire, UK). Appropriate positive and negative tissue controls were used throughout. Immunoreactivity was evaluated according to the intensity of tumor cell staining (0 to 3⫹), as well as the percentage of tumor cells that were stained. A particular tumor was considered unequivocally positive if ⬎10% of the tumor cells reacted with any intensity.

RESULTS The immunohistochemical results are summarized in Tables 2, 3, and 4. Thirty-eight (90%) of the 42 HCCs reacted with Hep Par 1 (Fig 1). Of the 4 HCCs negative for Hep Par 1, 2 were moderately differentiated tumors and 2 were poorly differentiated tumors. Most of the MAs were negative for this marker (Fig 2), although positive staining was noted in 8 (14%) MAs at various TABLE 2. Immunohistochemical Findings in Hepatocellular Carcinoma, Cholangiocarcinoma, and Metastatic Adenocarcinoma Type

Antibody

Hepatocellular carcinoma (n ⫽ 42)

Cholangiocarcinoma (n ⫽ 9)

Metastatic adenocarcinoma (n ⫽ 56)

Hep Par1 AE1/AE3 CAM 5.2 CK 7 CK 19 CK 20 B72.3 Factor XIIIa AFP pCEA mCEA Villin CD10 MOC-31 Inhibin

38 (90%) 19 (45%) 41 (98%) 9 (21%) 4 (10%) 2 (5%) 1 (2%) 37 (88%) 12 (29%) 39 (93%) 0 (0%) 13 (31%) 21 (50%) 5 (12%) 0 (0%)

0 (0%) 8 (89%) 8 (89%) 7 (78%) 4 (44%) 1 (11%) 3 (33%) 5 (56%) 0 (0%) 9 (100%) 2 (22%) 2 (22%) 1 (11%) 6 (67%) 0 (0%)

8 (14%) 53 (95%) 48 (89%) 20 (36%) 16 (29%) 17 (30%) 7 (13%) 22 (43%) 3 (5%) 54 (96%) 35 (62%) 22 (39%) 3 (5%) 37 (66%) 0 (0%)

1:100 1:500 1:60 1:400 1:60 1:30 1:5 1:10 1:10 1:30 1:100 1:300 1:300 1:5 1:400

Pretreatment Citrate buffer, pH 6.0 Citrate buffer, pH 6.0 Trypsin 5 minutes, citrate buffer, pH 6.0 Pressure cooker, citrate buffer, pH 6.0 Citrate buffer, pH 6.0 Citrate buffer, pH 6.0 Citrate buffer, pH 6.0 Pressure cooker, pH 9.5 Citrate buffer, pH 6.0 Citrate buffer, pH 6.0 Citrate buffer, pH 6.0 Citrate buffer, pH 6.0 Citrate buffer, pH 6.0 Citrate buffer, pH 6.0 Citrate buffer, pH 6.0

primary sites. None of the 9 CCs were positive for Hep Par 1. Polyclonal carcinoembryonic antigen (pCEA) expression was observed in 39 (93%) HCCs. Twenty-nine (69%) of these pCEA- positive cases exhibited a bile canalicular staining pattern (Fig 3A), with the remainder demonstrating cytoplasmic and/or membranous staining. Cytoplasmic pCEA positivity was present in 54 (96%) MAs and in all 9 CCs studied. A bile canalicular pattern of pCEA was not observed in any of these tumors. Thirty-five (62%) MAs and 2 (22%) CCs reacted with the monoclonal carcinoembryonic antigen (mCEA) antibody. In contrast, mCEA staining was absent in all 42 HCCs evaluated (Fig 4). Thirty-seven (66%) MAs and 6 (67%) CCs were positive for MOC-31, whereas only 5 (12%) HCCs reacted with this antibody. Factor XIIIa positivity was observed in 37 (88%) HCCs, 5 (56%) CCs, and 22 (43%) MAs. Thirteen (31%) HCCs were villin positive. Most displayed a cytoplasmic staining pattern, although a bile canalicular pattern of expression was noted in 7 cases (Fig 3B). Cytoplasmic and apical staining for villin was identified in 22 (39%) MAs and 2 (22%) CCs. CD10 expression was present in 21 (50%) HCCs. Although staining was predominantly cytoplasmic, 17 cases showed areas of canalicular staining (Fig 3C). In contrast, only 3 (5%) MAs and 1 (11%) CC were CD10 positive, displaying cytoplasmic reactivity only. Alpha-fetoprotein (AFP)-positive cells were identified in 3 (5%) MAs and 12 (29%) HCCs. All 9 CCs were uniformly AFP-negative. AE1/AE3 expression was present in most of the tumors studied, staining 53 (95%) of MAs, 19 (45%) of HCCs, and 8 (89%) of CCs. Similarly, CAM 5.2 immunoreactivity was noted in 48 (89%) MAs, 41 (98%) HCCs, and 8 (89%) CCs. In contrast, expression of B72.3 was observed in only 7 (13%) MAs, 1 (2%) HCC, and 3 (33%) CCs. Four (44%) CCs expressed cytokeratin (CK) 19, whereas 16 (29%) MAs and 4 (10%) HCCs were posi-

1176

IMMUNOHISTOCHEMISTRY OF LIVER TUMORS (Lau et al)

TABLE 3. Immunohistochemical Findings in Metastatic Adenocarcinomas Primary Site

Antibody

Colon (n ⫽ 24)

Breast (n ⫽ 5)

Stomach (n ⫽ 4)

Pancreas (n ⫽ 15)

Ovary (n ⫽ 8)

Total (n ⫽ 56)

Hep Par1 AE1/AE3 CAM 5.2 CK 7 CK 19 CK 20 B72.3 Factor XIIIa AFP pCEA mCEA Villin CD10 MOC-31 Inhibin

1 24 22 2 7 17 2 3 1 23 23 16 1 23 0

2 4 5 3 2 0 0 2 1 5 0 0 0 2 0

2 3 4 0 0 0 0 1 0 4 2 3 0 2 0

3 14 15 12 7 0 4 11 0 15 9 3 1 4 0

0 8 4 3 0 0 1 5 1 7 1 0 1 6 0

8 (14%) 53 (95%) 48 (89%) 20 (36%) 16 (29%) 17 (30%) 7 (13%) 22 (43%) 3 (5%) 54 (96%) 35 (62%) 22 (39%) 3 (5%) 37 (66%) 0 (0%)

tive for this antibody. CK 7 immunoreactivity was noted in 7 (78%) CCs, 20 (36%) MAs, and 9 (21%) HCCs. Seventeen (30%) MAs expressed CK 20, all of which were of colonic origin. CK 20 positivity was present in only 2 (5%) HCCs and 1 (11%) CCs. Inhibin reactivity was absent in all HCCs, CCs, and MAs studied. DISCUSSION Hep Par 1 is a mouse monoclonal antibody generated using failed allograft liver tissue as the immunogen.18 The antibody reacts with an as-yet uncharacterized antigen present in hepatocytes, although some reports have suggested that the epitope is localized to mitochondria, based on the granular intracytoplasmic pattern of immunostaining.12,18,21 Hep Par 1 has been established as a fairly sensitive marker for HCCs. In the initial investigation on the use of this antibody, Wennerberg et al18 reported expression in 37 of 38 HCCs, the single exception being an example of sclerosing HCC. Similar results were obtained by Leong et al,13 who observed positive labeling in 30 of 32 HCCs. The 2 cases that failed to stain for Hep Par 1 included a poorly differentiated variant and a sclerosing variant of HCC. Minervini et al12 reported an 82% sensitivity of Hep Par

1 for the detection of hepatocellular neoplasms, with a tendency toward a lack of strong staining observed in 50% of poorly differentiated HCCs. The results of the present investigation are consistent with previous reports. Here, 90% of HCCs studied reacted with Hep Par 1, substantiating the high sensitivity of this marker. Similar to previous studies that have demonstrated a trend toward a lack of expression in poorly differentiated HCCs,12,13 the 4 cases that did not mark for Hep Par 1 in the present series were morphologically moderate or poorly differentiated tumors. Despite its high sensitivity, Hep Par 1 is not entirely specific for HCC. Although normal bile ducts are negative for Hep Par 1,18 reactivity of this antibody has been noted in occasional CCs.12,13,18 Hep Par 1 immunoreactivity was not observed in any of the CCs evaluated in this study. In addition, rare extrahepatic neoplasms of colonic, gastric, and pancreatic origin have also been shown to have focal reactivity,18 an observation confirmed by the present study, which demonstrated Hep Par 1 positivity in 1 colonic, 2 gastric, and 3 pancreatic carcinomas, as well as in 2 breast carcinomas metastatic to the liver. More recently, Hep Par 1 expression has been demonstrated in 6 of 7 gastrointestinal adenocarcinomas with hepatoid morphologic features.22

TABLE 4. Immunohistochemical Staining Patterns of Positivity for pCEA, CD10, and villin in HCC, CC, and MA Antibody pCEA

CD10

Villin

Tumor Type

Canalicular

Noncanalicular*

Canalicular

Noncanalicular†

Canalicular

Noncanalicular§

HCC (n ⫽ 42) CC (n ⫽ 9) MA (n ⫽ 56)

29 (69%) 0 (0%) 0 (0%)

10 (24%) 9 (100%) 54 (96%)

17 (40%) 0 (0%) 0 (0%)

4 (10%) 1 (11%) 3 (5%)

7 (17%) 0 (0%) 0 (0%)

6 (14%) 2 (22%) 22 (39%)

*Cytoplasmic and/or membranous staining. †Cytoplasmic staining. §Cytoplasmic and/or apical staining.

1177

HUMAN PATHOLOGY

Volume 33, No. 12 (December 2002)

FIGURE 1. Hep Par 1 expression in HCC. Staining is characteristically cytoplasmic and granular. (Original magnification ⫻100.)

Normal hepatocytes express CK 8 and CK 18, whereas bile duct epithelial cells express CK 7, CK 8, CK 18, and CK 19.3,15,23 The differential expression of cytokeratin subtypes has been used as a potential means of separating HCC, CC, and MA. This has been based on the premise that tumors in general continue to express particular cytokeratins characteristic of their cell of origin. The pooled AE1/AE3 antibody recognizes a variety of cytokeratins, although not CK 8 or CK 18. Several studies have demonstrated AE1/AE3 immunoreactivity in most CCs but in only a minority of HCCs.2,3,24 In contrast, our study, as well as that of Ma et al,7 documented AE1/AE3 positivity in a large percentage of HCCs. Based on the overlapping staining patterns of this antibody, AE1/AE3 appears to have little value in distinguishing HCC from CC or MA. A similar problem limits the diagnostic utility of CAM 5.2, which recognizes CK 8, CK 18, and CK 19. CAM 5.2 staining was observed in most of the HCCs, CCs, and MAs studied, which is compatible with results reported by other investigators,2,7,8,12 indicating that the immunoreactivity of this antibody does not differ between these tumors. Comparatively more success has been found with the use of antibody to CK 19. CK 19 expression appears relatively specific for CC, though HCCs sometimes exhibit CK 19 positivity.3,6,10,13,23 Similar to previous reports, we documented only 4 of 42 (10%) HCCs to be CK 19 positive. Our observed sensitivity of CK 19 for CC (44%) differs somewhat from the experience of other investigators, who have reported CK 19 positivity in 85% to 100% of CCs.3,6,8,10,13,15 Because CK 19 stains most CCs and only rarely stains HCCs, positivity for this antibody is an effective marker for differentiating CC from HCC. More recently, differential expression of CK 7 and CK 20 has been investigated as a means of differentiating HCC from CC. As mentioned previously, CK 7 is present in bile duct epithelium, but not in normal hepatocytes.15 As such, CK 7 has been observed in most CCs6,10,11,15,17 and only occasionally in HCCs.6,10,11,17,23

In contrast, both CCs and HCCs tend to be CK 20 negative.11,13,17 In keeping with these previous reports, in our study most CCs were CK 7 positive and CK 20 negative. CK 20 expression was also absent from most HCCs, although CK 7 immunoreactivity was detected in 9 (21%) of these tumors. Therefore, CK 7 positivity does not preclude a diagnosis of HCC. Observed immunoreactivity of MAs involving the liver for different cytokeratins has varied, depending on the particular antibodies used and on the primary sites of the metastatic neoplasms studied.3,4,6-8,11-13,15-17 The broad specificity of these keratin antibodies in general does not appear to allow for meaningful differentiation of MA from CC and HCC. However, evaluation of CK 7 and CK 20 expression in this context may be of some value. MAs of colonic origin, in particular, consistently exhibit a CK 20 –positive, CK 7–negative immunophenotype11,16,17,25 that differs significantly from the immunohistochemical profile of CCs, which tend to be CK 20 negative and CK 7 positive,11,16,17,25 and HCCs, which are normally negative for both markers.11,17 We found that these general staining patterns were retained, with the colonic MAs exhibiting a CK 20 –positive, CK 7–negative phenotype, in contrast to the CCs, which were positive for CK 7 in 78% of cases and for CK 20 in 11% of cases, and HCCs, which exhibited CK 7 immunoreactivity in 21% of cases and CK 20 immunoreactivity in 5% of cases studied. This observed differential pattern of CK 7 and CK 20 expression indicates that these markers may have some discriminative value in distinguishing HCC, CC, and MA, specifically of colonic origin. Ma et al7 investigated the use of B72.3, which reacts with a variety of adenocarcinomas, in the context of distinguishing primary and metastatic carcinomas in the liver. They reported reactivity with this antibody in 4 of 8 (50%) CCs and 12 of 24 (50%) MAs, but only in 6 of 56 (11%) HCCs. These observations were subsequently confirmed by Fucich et al,9 who observed B72.3 staining in 6 of 10 CCs, 7 of 7 MAs, and 1 of 14 HCCs. Similarly, in the present study, 2% of HCCs stained with

FIGURE 2. Absence of Hep Par 1 expression in MA. Note the staining of nonneoplastic hepatocytes. (Original magnification ⫻200.)

1178

IMMUNOHISTOCHEMISTRY OF LIVER TUMORS (Lau et al)

FIGURE 3. Bile canalicular immunohistochemical staining pattern in HCC: (A) pCEA, (B) villin, (C) CD10. (Original magnification ⫻200.)

the B72.3 antibody. In contrast, 3 (33%) of 9 CCs and 7 (13%) of 56 MAs were B72.3 positive. Thus, although there has been an observed trend toward B72.3 negativity in HCCs, the fact that a large percentage of CCs and MAs often fail to stain limits the discriminative value of this marker. Several investigators have identified factor XIIIa as an effective marker for HCCs.6,9 In contrast, Leong et al13 and Porcell et al17 found no positive staining for factor XIIIa in 32 and 13 HCCs, respectively. In the present study, factor XIIIa immunoreactivity was observed in 88% of HCCs, but was also present in 56% of CCs and 43% of MAs. This lack of specificity, also been demonstrated in other studies,6,9 serves to limit the value of factor XIIIa in differentiating these particular tumors.

FIGURE 4. Lack of mCEA staining in HCC. (Original magnification ⫻200.)

We found AFP positivity in 29% of HCCs studied, a frequency consistent with previous reports1,5-7,9,12,17,26,27 which have identified AFP staining in 2%26 to 61.5%6 of such tumors. Despite the poor sensitivity of this marker, AFP is relatively specific for HCC. This has been documented in various studies that noted a general absence of AFP expression in CC and MA.5-7,9,12,17,27 These observations are confirmed by the present study, which demonstrated AFP expression in only 3 MAs and no CCs. The discriminative value of AFP thus appears limited by poor sensitivity, although AFP positivity is highly suggestive of HCC. In our study, a bile canalicular pattern of pCEA expression was observed in 29 of 42 (69%) cases of HCC, sometimes in association with cytoplasmic and/or membranous staining. This canalicular pattern was not observed in any of the CCs or MAs examined. The bile canalicular pattern of pCEA immunoreactivity present in HCCs has been described previously3,4,6,7,12,17,27 and is attributed to cross-reactivity with biliary glycoprotein I, which is present in normal bile ducts and bile canaliculi. This pattern of staining is highly specific for HCC and is not typically observed in other types of neoplasms.3,4,6,7,12,17,27 The reported sensitivity of pCEA for HCC has ranged from 60%3 to 95%,27 with some investigators noting decreased immunoreactivity in poorly differentiated tumors.3,4,27 Cytoplasmic expression of pCEA is not uncommonly observed in HCC,4,7,12,27 as supported in the present study, although this pattern of staining, in contrast to the bile canalicular pattern, is nonspecific and is seen in both CC and MA. Unlike pCEA, mCEA is typically not expressed by HCC,4-7,12 although expression has been observed in up

1179

HUMAN PATHOLOGY

Volume 33, No. 12 (December 2002)

to 11% of cases in 1 study.7 Because mCEA is positive in a wide variety of adenocarcinomas, including CC, lack of staining has been used to support a diagnosis of HCC. In the present investigation, although 62% of the MAs were positive for mCEA, none of the 42 HCCs were noted to express this marker. These observations are in agreement with previous investigations4-7,12 and suggest that this differential staining pattern of mCEA is of use in differentiating HCC from other adenocarcinomas. In addition to pCEA, a similar bile canalicular pattern of staining has been observed in HCC using antibodies to villin28 and CD10.27,29 Villin expression has been shown to be a sensitive marker for adenocarcinomas of the gastrointestinal tract, including those of colonic, gastric, pancreaticobiliary, and hepatic origin.28 The pattern of staining in these tumors is primarily apical (brush border) or cytoplasmic. HCCs typically share this cytoplasmic pattern of immunoreactivity, but can also exhibit focal areas of bile canalicular accentuation.28 These observations are confirmed by the present study, in which 7 of the 13 HCCs expressing villin were noted to have a bile canalicular pattern, either in isolation or accompanying areas of cytoplasmic reactivity. Of note, all 7 cases also demonstrated canalicular staining for pCEA. As reported in a previous study,28 this bile canalicular pattern of villin expression was unique to HCCs and was not demonstrated in other villin-positive MAs or CCs. Despite the high specificity of the bile canalicular pattern of staining for HCC, the utility of this marker appears limited by poor sensitivity, with only a minority of HCCs in this study exhibiting this pattern of villin expression. CD10, or the common acute lymphoblastic leukemia antigen, has been used predominantly as a marker for categorizing and subclassifying hematolymphoid neoplasms. In the context of nonhematopoietic tumors, CD10 is a relatively nonspecific marker, with positive staining occurring in a variety of neoplasms.29 Wereas most tumors exhibit luminal, cytoplasmic, or membranous staining patterns, a canalicular pattern of expression similar to that of pCEA has been demonstrated in normal liver tissue and HCC.27,29 In the present study, 21 of 42 (50%) HCCs expressed CD10, 17 (40%) of which showed areas of canalicular staining. In contrast, CD10 immunoreactivity was observed in 11% of CCs and 5% of MAs, with none of these positive cases exhibiting a canalicular pattern of staining. Most of the CD10-positive cases also demonstrated bile canalicular staining for pCEA and/or villin. However, 2 cases of HCC exhibited bile canalicular positivity for CD10 in the absence of bile canalicular staining for pCEA and villin. Previous studies have also demonstrated the high specificity of CD10 canalicular staining for HCC.27,29 Borscheri et al27 documented this pattern of CD10 reactivity in 68.3% of HCCs and in none of the metastatic liver tumors examined. Similarly, Chu and Arber29 observed CD10 expression in 7 of 13 HCCs. A canalicular pattern was present in 6 of these cases and was not noted in any of the other neoplasms studied. Interestingly, as in the present study, these investigators also observed a lack of CD10 expression in CCs, with all

12 cases negative in their report. Based on these observations, the bile canalicular pattern of CD10 immunoreactivity appears to be specific for HCC, although poor sensitivity may limit its usefulness. The differential pattern of CD10 expression noted between HCC and CC needs to be examined further in a larger number of cases. Recently, MOC-31 has been investigated in the context of differentiating HCC, CC, and MA.14,17 Proca et al14 observed MOC-31 staining in 14 of 14 CCs and 33 of 33 MAs, but in 0 of 15 HCCs. These findings were replicated by Porcell et al,17 who noted MOC-31 expression in 13 of 14 CCs and 27 of 27 MAs, but in 0 of 13 HCCs. Based on these observations, MOC-31 has been suggested to be a useful marker for evaluating hepatic tumors when these entities enter the differential diagnosis. In contrast to these previous reports, however, we noted MOC-31 expression in 5 of 42 (12%) HCCs and in 6 of 9 (67%) CCs. In addition, MOC-31 was identified in only 66% of MAs. The reasons for these differences in observed staining sensitivity are unclear, but they may result from differences in the number of cases examined. Although we were unable to fully substantiate the results of the previous studies, we did find a similar trend toward MOC-31 negativity in HCCs and MOC-31 positivity in CCs and MAs, suggesting that this marker is of potential value in the differential diagnosis. The initial report detailing inhibin immunoreactivity in HCCs described positive expression of this antibody in 89% of these tumors.30 Subsequently, the observed inhibin reactivity in HCC was shown to be a false-positive result secondary to the presence of endogenous biotin in hepatocytes.31 When a biotin- blocking step was used, HCCs were shown to be largely negative for inhibin.31,32 Our results confirm the low value of inhibin immunostaining in differentiating HCC from other neoplasms involving the liver when a step to block endogenous biotin activity is used. All 42 HCCs in this study, as well as all of the CCs and MAs, were negative for inhibin. The results presented in this study confirm previous observations regarding the high specificity and sensitivity of Hep Par 1 for HCC. As such, this marker should be included in any antibody panel used to distinguish HCC from CC and MA. mCEA and MOC-31 also appear to be useful in distinguishing these entities, because positive staining for either of these markers would favor a diagnosis of CC or MA rather than HCC. pCEA positivity, when observed in a bile canalicular pattern, is very specific and relatively sensitive for HCC. Although limited somewhat by poor sensitivity, CD10 and villin also appear to have diagnostic value; the presence of a bile canalicular pattern of staining strongly supports a diagnosis of HCC, because this pattern of immunoreactivity is almost never observed in other tumors. Similarly, AFP, although reactive in a minority of HCCs, appears to be relatively specific for this entity and provides additional evidence supporting this diagnosis. Antibodies to CK 7, CK 19, and CK 20 are also considered worthwhile, and inclusion in an

1180

IMMUNOHISTOCHEMISTRY OF LIVER TUMORS (Lau et al)

extended panel will further facilitate differentiation of HCC, CC, and MA, particularly when the differential diagnosis includes MA of colonic origin. The other antibodies evaluated in this study do not significantly contribute to this differentiation. REFERENCES 1. Goodman ZD, Ishak KG, Langloss, JM, et al: Combined hepatocellular-colangiocarcinoma: A histologic and immunohistochemical study. Cancer 55:1124-1135, 1985 2. Johnson DE, Herndier BG, Medeiros LJ, et al: The diagnostic utility of the keratin profiles of hepatocellular carcinoma and cholangiocarcinoma. Am J Surg Pathol 12:187-197, 1988 3. Balaton AJ, Nehama-Sibony M, Gotheil C, et al: Distinction between hepatocellular carcinoma, cholangiocarcinoma, and metastatic carcinoma based on immunohistochemical staining for carcinoembryonic antigen and for cytokeratin 19 on paraffin sections. J Pathol 156:305-310, 1988 4. Christensen WN, Boitnott JK, Kuhajda FP: Immunoperoxidase staining as a diagnostic aid for hepatocellular carcinoma. Mod Pathol 2:8-12, 1989 5. Brumm C, Schulze C, Charels K, et al: The significance of alpha-fetoprotein and other tumor markers in differential immunocytochemistry of primary liver tumors. Histopathology 14:503-513, 1989 6. Hurlimann J, Gardiol D: Immunohistochemistry in the differential diagnosis of liver carcinomas. Am J Surg Pathol 15:280-288, 1991 7. Ma CK, Zarbo RJ, Frierson HF, et al: Comparative Immunohistochemical study of primary and metastatic carcinomas of the liver. Am J Clin Pathol 99:551-557, 1993 8. Akasofu M, Kawahara E, Kurumaya H, et al: Immunohistochemical detection of breast-specific antigens and cytokeratins in metastatic breast carcinoma in the liver. Acta Pathologica Japonica 43:736-744, 1993 9. Fucich LF, Cheles MK, Thung SN, et al: Primary vs. metastatic hepatic carcinoma: An immunohistochemical study of 34 cases. Arch Pathol Lab Med 118:927-930, 1994 10. D’Errico A, Baccarini P, Fiorentino M, et al: Histogenesis of primary liver carcinomas: Strengths and weaknesses of cytokeratin profile and albumin mRNA detection. HUM PATHOL 27:599-604, 1996 11. Maeda T, Kajiyama K, Adachi E, et al: The expression of cytokeratins 7, 19, and 20 in primary and metastatic carcinomas of the liver. Mod Pathol 9:901-909, 1996 12. Minervini MI, Demetris AJ, Lee RG, et al: Utilization of hepatocyte-specific antibody in the immunocytochemical evaluation of liver tumors. Mod Pathol 10:686-692, 1997 13. Leong AS, Sormunen RT, Tsui WM, et al: Hep Par 1 and selected antibodies in the immunohistochemical distinction of hepatocellular carcinoma from cholangiocarcinoma, combined tumors, and metastatic carcinoma. Histopathology 33:318-324, 1998 14. Proca DM, Niemann TH, Porcell AI, et al: MOC31 immunoreactivity in primary and metastatic carcinoma of the liver: Report of findings and review of other utilized markers. Appl Immunohistochem Mol Morph 82:120-125, 2000 15. Shimonishi T, Miyazaki K, Nakanuma Y: Cytokeratin profile relates to histological subtypes and intrahepatic location of intrahe-

patic cholangiocarcinoma and primary sites of metastatic adenocarcinoma of liver. Histopathology 37:55-63, 2000 16. Rullier A, Le Bail B, Fawaz R, et al: Cytokeratin 7 and 20 expression in cholangiocarcinomas varies along the biliary tract but still differs from that in colorectal carcinoma metastasis. Am J Surg Pathol 24:870-876, 2000 17. Porcell AI, De Young BR, Proca DM, et al: Immunohistochemical analysis of hepatocellular and adenocarcinoma in the liver: MOC31 compares favorably with other putative markers. Mod Pathol 13:773-778, 2000 18. Wennerberg AE, Nalesnik MA, Coleman WB: Hepatocyte paraffin 1: A monoclonal antibody that reacts with hepatocytes and can be used for differential diagnosis of hepatic tumors. Am J Pathol 143:1050-1054, 1993 19. Murakata LA, Ishak KG, Nzeako UC: Clear cell carcinoma of the liver: A comparative immunohistochemical study with renal clear cell carcinoma. Mod Pathol 13:874-881, 2000 20. Edmondson HA, Steiner PE: Primary carcinoma of the liver: A study of 100 cases among 48,900 necropsies. Cancer 7:462-503, 1954 21. Fasano M, Theise ND, Nalesnik M, et al: Immunohistochemical evaluation of hepatoblastomas with use of the hepatocyte-specific marker, hepatocyte paraffin 1, and the polyclonal anti-carcinoembryonic antigen. Mod Pathol 11:934-938, 1998 22. Maitra A, Murakata LA, Albore-Saavedra J: Immunoreactivity for hepatocyte paraffin 1 antibody in hepatoid adenocarcinomas of the gastrointestinal tract. Am J Clin Pathol 115:689-694, 2001 23. Van Eyken P, Sciot R, Paterson A, et al: Cytokeratin expression in hepatocellular carcinoma: An immunohistochemical study. HUM PATHOL 19:562-568, 1988 24. Lai YS, Thung SN, Gerber MA, et al: Expression of cytokeratins in normal and diseased livers and in primary liver carcinomas. Arch Pathol Lab Med 113:134-138, 1989 25. Tot T: Adenocarcinomas metastatic to the liver: The value of cytokeratins 20 and 7 in the search for unknown primary tumors. Cancer 85:171-177, 1999 26. Ferrandez-Izquiedro A, Llombart-Bosch A: Immunohistochemical characterization of 130 cases of primary hepatic carcinomas. Pathol Res Pract 182:783-791, 1987 27. Borscheri N, Roessner A, Rocken C: Canalicular immunostaining of neprilysin (CD10) as a diagnostic marker for hepatocellular carcinomas. Am J Surg Pathol 25:1297-1303, 2001 28. Bacchi CE, Gown AM: Distribution and pattern of expression of villin, a gastrointestinal-associated cytoskeletal protein, in human carcinomas: A study employing paraffin-embedded tissue. Lab Invest 64:418-424, 1991 29. Chu PG, Arber DA: Paraffin-section detection of CD10 in 505 nonhematopoietic neoplasms: Frequent expression in renal cell carcinoma and endometrial stromal sarcoma. Am J Clin Pathol 113: 374-382, 2000 30. McCluggage WG, Maxwell P, Patterson A, et al: Immunohistochemical staining of hepatocellular carcinoma with monoclonal antibody against inhibin. Histopathology 30:518-522, 1997 31. Iezzoni JC, Mills SE, Pelkey TJ, et al: Inhibin is not an immunohistochemical marker for hepatocellular carcinoma: An example of the potential pitfall in diagnostic immunohistochemistry caused by endogenous biotin. Am J Clin Pathol 111:229-234, 1999 32. Renshaw AA, Granter SR: A comparison of A103 and inhibin reactivity in adrenal cortical tumors: Distinction from hepatocellular carcinoma and renal tumors. Mod Pathol 11:1160-1164, 1998

1181