Molecular Changes in Hepatocellular Dysplastic Nodules on Microdissected Liver Biopsies

Transcription

1 Molecular Changes in Hepatocellular Dysplastic Nodules on Microdissected Liver Biopsies MARCO MAGGIONI, 1 GUIDO COGGI, 1 BARBARA CASSANI, 1 PAOLO BIANCHI, 2 SOLANGE ROMAGNOLI, 1 ALESSANDRA MANDELLI, 1 MAURO BORZIO, 3 PIERGIUSEPPE COLOMBO, 2 AND MASSIMO RONCALLI 2 The genetic profile of dysplastic hepatocellular nodules arising in cirrhosis is poorly understood. We assessed loss of heterozygosity (LOH) and microsatellite instability (MI) in 10 dysplastic nodules (4 low-grade and 6 high-grade) with surrounding cirrhosis and in 10 hepatocellular carcinomas (HCC). Six microsatellite loci were selected and investigated on microdissected needle biopsies. Twenty-four (24.4%) informative loci showed allelic loss, while MI was seen in 3 loci only (3%). The most involved sites were located on chromosomes 4q (54.5%) and 8p (50%). LOH was documented in 16.6%, cirrhotic, 50% low-grade dysplastic nodules (LGDN), 83% high-grade dysplastic nodules (HGDN), and 70% malignant nodules. LOH at multiple loci was increasingly seen from cirrhotic to HGDN, but not from the latter to HCC. The fractional allelic loss (FAL) was significantly increased in dysplastic and neoplastic nodules as compared with cirrhosis (P <.01). The progressive accumulation of genetic changes in cirrhotic, dysplastic, and malignant hepatocellular nodules is in keeping with a multistep process of carcinogenesis; within this spectrum, HGDN can be considered advanced precursors of HCC. (HEPATOLOGY 2000;32: ) Hepatocellular carcinoma (HCC) is one of the most diffuse cancers worldwide, but the mechanism of liver carcinogenesis remains largely unknown. Hepatocellular dysplastic nodules are considered HCC precursors and are currently classified into low-grade (LGDN) and high-grade (HGDN) categories. 1 Their premalignant nature was inferred by a number of features such as morphology, 2-4 proliferative activity, 5,6 vascular pattern, 7 DNA content, 8 and clonality However, their biological properties are still unknown, and it has been repeatedly emphasized that the classification of hepatocellular nodules should be supported by genetic data to be diagnostically and prognostically helpful. 1 Abbreviations: HCC, hepatocellular carcinoma; LGDN, low-grade dysplastic nodule; HGDN, high-grade dysplastic nodule; LOH, loss of heterozygosity; MI, microsatellite instability; FAL, fractional allelic loss. From the Departments of Pathology at 1 San Paolo Hospital of Milan, and 2 Humanitas Clinical Institute, Rozzano, University of Milan, Milan; and 3 Department of Medicine, Fatebenefratelli Hospital of Milan, Italy. Received March 13, 2000; accepted July 20, Address reprint requests to: Marco Maggioni M.D., II Cattedra di Anatomia Patologica, Azienda Ospedaliera Ospedale San Paolo, Via di Rudin 8, 20146, Milan, Italy. anapat2@mailserver.unimi.it; fax: Copyright 2000 by the American Association for the Study of Liver Diseases /00/ $3.00/0 doi: /jhep p53 mutations have been mainly demonstrated in aflatoxinassociated HCC 12,13 but not in dysplastic nodules. 14 Interestingly, loss of heterozygosity (LOH) at multiple loci was shown in most hepatocellular malignancies, particularly on chromosomes 1p, 1q, 4q, 6q, 8p, 8q, 13q, 16p, and 17p Some of these loci are the site of well-known tumor suppressor genes, such as the retinoblastoma or the insulin-like growth factor II receptor gene. Allelic loss and microsatellite instability (MI) have also been reported in the cirrhotic parenchyma adjacent to HCC, suggesting that molecular abnormalities, possibly leading to HCC development, are already present before the onset of dysplastic or malignant nodules However, genetic changes have been very rarely documented in hepatocellular dysplastic nodules. Yamada et al. 23 found the same insulinlike growth factor receptor II gene allelic loss in both dysplastic nodules and the adjacent cirrhotic tissue. The allelic imbalance of insulin-like growth factor II was also demonstrated by Aihara et al. 24 in 3 of 7 dysplastic nodules. The study of the molecular profile of nonmalignant hepatocellular nodules has been insofar hindered by their low prevalence and by the fact that these lesions are actually destroyed by treatment (i.e., percutaneous alcohol injection, chemoembolization). End-stage posttransplanted cirrhotic livers harboring malignant and nonmalignant nodules have been used in the past for molecular studies. However, these specimens do not represent the habitual condition of well-compensated and routinely followed-up cirrhotic livers. By contrast, liver biopsies performed in vivo on ultrasound-detected nodules in cirrhotic patients would be optimal for these purposes. It is indeed this group of patients who could potentially benefit from the availability of genetic information. However, a molecular assay conducted on very small specimens such as liver biopsies may be limited by the DNA yield. Tissue microdissection is therefore mandatory to precisely delimitate hepatocellular populations from mesenchymal, inflammatory, and/or biliary cells, which, in turn, may be used as internal control at least on retrospective material. In this study, we selected a retrospective series of biopsy samples from both dysplastic nodules with cirrhotic surroundings and HCCs. Microdissected liver specimens were used to assess allelic loss and MI at 6 microsatellite loci, which were previously found to be abnormal in HCC. The aims were to document on liver biopsies molecular abnormalities of hepatocellular dysplastic nodules and to ascertain whether LGDNs and HGDNSs have distinct genetic profiles, as also compared with cirrhotic and malignant nodules.

2 HEPATOLOGY Vol. 32, No. 5, 2000 MAGGIONI ET AL. 943 MATERIALS AND METHODS Formalin-fixed, paraffin-embedded liver needle biopsies were screened from the files of the II Department of Pathology, University of Milan, between 1995 and Ten dysplastic nodules were identified, reviewed by 2 expert pathologists (M.M. and M.R.), and classified into low- and high-grade categories according to criteria recently proposed. 1 Only cirrhosis-associated nodules with available extranodular parenchyma were selected. In all cases, the diagnosis was made after a careful comparison between the intranodular and the extranodular liver. Dysplastic nodules were recognized on the basis of an expansive hepatocellular growth containing at least one isolated portal tract. Hepatic cords were usually of 2-cell thickness; in LGDNs, large or small cell changes were focally documented in intralesional samples. In HGDNs, architectural abnormalities such as nuclear crowding, small cell changes, and occasional gland-like structures alternated in an uneven distribution with bland-looking areas. Ten additional needle biopsies of HCC (2 with available nonneoplastic tissue) in cirrhosis were also selected. In 2 cases (no. 1 and 6), these lesions appeared during the follow-up of dysplastic nodules. Overt HCC was recognized when the above-mentioned architectural abnormalities and/or more than 3-cell-thick hepatocellular trabeculae were evenly distributed throughout the sample. Overall, 32 needle specimens from 18 patients (13 males and 5 females; mean age, 63.8 years) were collected and classified as follows: 4 LGDNs, 6 HGDNs, 10 HCCs (8 well- and 2 moderately differentiated), and 12 cirrhotic nodules. The etiology of cirrhosis was hepatitis B virus related (8 cases), hepatitis C virus related (6 cases), and alcohol-related (4 cases). From each paraffin block, 5- m-thick consecutive sections were obtained. The first section was stained with hematoxylin-eosin and used as a reference slide. The second and third sections were stained with hematoxylin-eosin, uncovered, and microdissected. Dysplastic, malignant, and cirrhotic hepatocellular subpopulations were carefully separated from fibrous septa containing bile ducts, vessels, and inflammatory cells (Fig. 1). The latter was used as an internal control for allelotype analysis. Microdissected tissue was collected in Eppendorf tubes, dehydrated, resuspended in 50 L extraction mixture (50 mmol/l Tris-HCl [ph 8], 5 mmol/l ethylenediaminetetraacetic acid, 0.5% Tween 20, 400 ng/ml Proteinase K 400), and digested overnight at 55 C. The loci investigated and the markers used were the following: 2p16 (D2S123) and 3p24 (D3S1611), 2 intronic regions of the mismatch repair genes hmsh2 and hmlh1, 6q27 (D6S281) locus, the site of the insulin-like growth factor receptor 2; 13q14, a CA repeat within intron 2 of the Rb gene; 4q35 (D4S426) and 8p23 (D8S277), which have been found to be abnormal in most cases of HCC. 16 All FIG. 1. Liver biopsy showing a HCC with 2 fibrous tracts before (A) and after (B) microdissection. *Microdissected areas. (Original magnification 100.) primer sequences were downloaded from GDB database ( DNA was amplified in a mixture containing 10 pmol of each primer (one fluorescent), 0.2 mmol/l dntps, 1.5 mmol/l MgCl 2, and 1.25 UTaq Gold polymerase (Perkin Elmer, Foster City, CA). Amplification was performed during 45 cycles of denaturation (94 C 1 minute), annealing (50 C 1 minute), and extension (72 C 1 minute). For 13q14, the annealing temperature was optimized at 55 C 1 minute. One microliter of fluorescent amplified product was electrophoresed on the Fluorescent Capillary System ABI Prism 310 (Applied Biosystem, Foster City, CA) and analyzed by Gene Scan Analysis software (PE Applied Biosystem). Each locus was assessed for LOH and MI. LOH was defined by at least 50% reduction of 1 allele in the intralesional sample as compared with the internal nonhepatocyte control and calculated as previously described. 25,26 MI was defined by the presence of additional shifts in the intralesional sample, as compared with the internal nonhepatocyte control. All positive results were repeated twice. The fractional allelic loss (FAL) was separately calculated in cirrhosis, LGDNs, HGDNs, and HCCs by dividing the number of LOH by the total number of informative loci. 27 Statistical differences were assessed by the Student s t test. RESULTS In the 18 patients, 147 of 192 samples (32 lesions analyzed with 6 markers) were successfully amplified (76.5%); of these 147 samples, 98 (66.6 %) resulted informative and were submitted to genetic analysis. Genetic data for individual patients, lesions, and markers are detailed in Table 1. Overall, 27 microsatellite abnormalities occurred in 14 patients: 24 were allelic losses (2 in cirrhotic, 3 in LGDN, 9 in HGDN, and 10 in malignant nodules), and 3 MI/dinucleotide shifts (1 LGDN, 1 HGDN, and 1 malignant nodule). No association was seen among the number and distribution of genetic abnormalities with age, sex, and etiology of cirrhosis. At least 1 individual allelic loss occurred in 2 cirrhotic (16.6%), 2 LGDNs (50%), 5 HGDNs (83%), and 7 malignant nodules (70%). The progression of these genetic abnormalities from cirrhosis to HCC is shown in Table 2. Individual but not concurrent losses were detected in cirrhosis in 2 cases; in these cases, the same clone-like genetic abnormality was documented in the associated LGDNs (Fig. 2). When considering cases with at least 2 informative loci, concurrent losses were a feature of genetically abnormal LGDN (1 of 3 cases [33%]), HGDNs (3 of 6 cases [50%]; Fig. 3), and malignant nodules (3 of 7 cases [43%]) (Table 2). Increasing FAL values were seen from cirrhotic (0.05) to LGDNs (0.21), HGDNs (0.47), and malignant nodules (0.38). A statistically significant difference in FAL values was documented between cirrhotic and LGDNs (P.007), HGDNs (P.004), and malignant nodules (P.002). No statistical difference was documented among dysplastic and malignant nodules. As regards the distribution of genetic abnormalities in the different types of hepatocellular nodules, 4q and 8p LOH were documented in cirrhotic, dysplastic, and malignant lesions; 6q LOH was shown in both dysplastic and malignant nodules, but not in cirrhosis; while 2p, 3p, and 13q LOH were restricted to HGDN and malignant nodules. The highest rate of LOH was found at 4q35 (54.5%) and 8p23 (50%), followed by 2p16 (42.8%), 3p24 (40%), 13q14 (38.4%), and 6q27 (33.3%). Dinucleotide shifts were demonstrated as individual genetic abnormalities at 8p23 (2 cases) and 13q14 (1 case) loci. Concurrent dinucleotide shifts were not detected.

3 944 MAGGIONI ET AL. HEPATOLOGY November 2000 TABLE 1. Detail of Allelotype Analysis Case No. Lesion Rb D2S123 D3S1611 D4S426 D6S281 D8S277 1 Cirrhosis E E E E E HGDN E E F E F HCC (G1) E E E E F 2 Cirrhosis E / / E / / HGDN E / / E / / 3 HCC (G1) / E / F 4 Cirrhosis E / E / HGDNS F / E / 5 HCC (G1) / / / / E 6 Cirrhosis / E / LGDN / E E HCC (G1) / E E 7 Cirrhosis E E / / HCC (G2) F E E 8 Cirrhosis E E E F E E LGDN E E E F F } 9 Cirrhosis E / E E E HGDN E F F E } 10 HCC (G1) E F / E F E 11 Cirrhosis / / / / E LGDN / / / / E 12 Cirrhosis E E E E E HCC (G2) F E F E E 13 Cirrhosis E / E E F LGDN E E E E F 14 HCC (G1) } 15 HCC (G1) / / / / / F 16 Cirrhosis E E HGDN E F 17 Cirrhosis E E E / HGDN F F F / 18 HCC (G1) F / F / / E NOTE. F, loss of heterozygosity; }, microsatellite instability; E, heterozygous;, homozygous; /, not amplified. DISCUSSION In the present study, we assumed that genetic information obtained from hepatocellular nodules that arose from cirrhosis could be valuable for a better understanding of their biological properties and for potential diagnostic applications. Careful microdissection was used to select appropriate and homogeneous hepatocyte subpopulations and control tissue for allelotype analysis. More than 75% microdissected samples resulted suitable for genetic analysis, yielding a proportion of informative cases in keeping with that from other authors, in series of surgically resected HCCs. 16 We consistently found allelic loss at different chromosomal loci in both dysplastic and malignant hepatocellular nodules. The rate of LOH detected in malignant nodules is within the range reported by other authors Nagai et al. 16 showed allelic loss in almost all their hepatocellular malignancies, suggesting that in HCC, the rate of LOH is dependent on the number of genetic markers used for analysis. TABLE 2. Distribution of Allelic Loss in Individual Lesions and Patients Case No Cirrhosis (12 cases) 4q35 8p23 LGDNs (4 cases) 4q35 8p23 6q27 HGDNs (6 cases) 4q35 13q14 2p16 4q35 2p16 8p23 3p24 4q35 13q14 HCC (10 cases) 8p23 8p23 13q14 2p16 4q35 8p23 3p24 6q27 13q14 13q14 NOTE. (white boxes), tissue available; (gray boxes), no tissue available.

4 HEPATOLOGY Vol. 32, No. 5, 2000 MAGGIONI ET AL. 945 In the present series, chromosomes 4q and 8p were the most common site of allelic loss. This result is consistent with the observations made by other authors, showing that these chromosomes are the elective sites of LOH in HCC. 16,17 Interestingly, we also found these losses in LGDNs and cirrhotic nodules, suggesting that these changes take place earlier than previously supposed during the process of hepatocarcinogenesis. We also found increasing FAL values from cirrhotic to LGDNs, to HGDNs and malignant nodules. This result supports a multistep process of carcinogenesis and suggests that dysplastic nodules are genuine HCC precursors, particularly those of high-grade type. This result is in keeping with previous reports 5-8 showing that premalignant changes such as increased proliferation index and abnormal ploidy and vascularization are features of HGDNs. Disruption of the cell cycle by allelic loss of key genes such as the Rb gene and other putative oncosuppressors located on 4q and 8p might well contribute to this abnormal phenotype. Chromosome 4q contains genes encoding growth factors and others predominantly expressed in the liver (albumin, fibrinogen, alcohol dehydrogenase, and others) so that the deletion of this region might alter hepatocyte growth condition and microenvironmental homeostasis. Interestingly, the FAL value of LGDNs was significantly increased as compared with the adjacent cirrhosis but resulted lower, although not statistically significant, as compared with HGDNs. This result suggests that LGDNs also stand on an abnormal genetic background that is not, however, as solid as that of HGDNs. Accordingly, for current clinical management, it seems wise to strictly follow-up LGDNs by ultrasound and serologic markers, while an aggressive treatment (ablative therapy by surgery, ethanol injection, or thermo-ablation) should be reserved for HGDNs, as previously proposed. 28 Interestingly, we also found clone-like genetic changes in sporadic cases of cirrhosis and in the associated LGDNs. This FIG. 2. Case 8. Left: morphology of cirrhosis (extralesional sample, top) as compared with the low-grade dysplastic nodule (intralesional sample, bottom). Right: Electropherogram for 4q35 of nonhepatocyte control (upper), cirrhotic (middle), and low-grade dysplastic nodule (bottom): allelic loss is detectable in both cirrhotic and dysplastic nodules. (Original magnification 250.) FIG. 3. Case 17. Left: morphology of cirrhosis (extralesional sample, top) as compared with the HGDN (intralesional sample, bottom). Right: Electropherogram for 4q35 and 13q14 of nonhepatocyte control (upper), cirrhotic (middle), and high-grade dysplastic nodule (bottom): concurrent allelic losses are detectable only in the latter lesion. (Original magnification 250.) result corroborates the hypothesis that discrete clone-like hepatocyte subpopulations take place very early in chronic liver disease,20,29 and then are likely to expand in morphologically recognizable dysplastic nodules. This feature is quite consistent with a model of liver carcinogenesis put forward by Theise et al. 30 MI was detected infrequently in dysplastic and malignant lesions, and never in cirrhosis. This result suggests that a replication error phenotype does not play a key role in the process of neoplastic transformation of the human liver, as previously suggested. 29,31 In conclusion, we showed that both the cirrhotic background and the associated dysplastic nodules harbor genetic changes under the form of LOH of putative tumor suppressor genes. This genetic damage progressively increases from cirrhosis to dysplastic and malignant nodules, suggesting a continuum spectrum of lesions, in agreement with a multistep model of hepatic carcinogenesis. Notably, the genetic profile of HGDNs is similar to that detected in malignant nodules. The above molecular data seem to support the actual classification of hepatocellular nodules in cirrhosis, which was recently established on simple morphologic grounds. 1 Whether genetic information can also be used for diagnostic purposes to support morphologic features is not yet settled, given the restricted number of cases investigated. Additional information is expected to come from studies investigating the genetic phenotype of a larger number of prospectically followed-up dysplastic hepatocellular nodules, particularly of low-grade type. REFERENCES 1. International Working Party. Terminology of nodular hepatocellular lesions. HEPATOLOGY 1995;22: Nakanuma Y, Terada T, Terasaki S, Ueda K, Nonomura A, Kawahara E, Matsui O. Atypical adenomatous hyperplasia in liver cirrhosis: low-grade hepatocellular carcinoma or borderline lesion? Histopathology 1990;17: Nakanuma Y, Terada T, Ueda K, Terasaki S, Nonomura A, Matsui O. Adenomatous hyperplasia of the liver as a precancerous lesion. Liver 1993;13:1-9.

5 946 MAGGIONI ET AL. HEPATOLOGY November Borzio M, Borzio F, Croce A, Sala M, Salmi A, Leandro G, Bruno S, Roncalli M. Ultrasonography detected macroregenerative nodules in cirrhosis: a prospective study. Gastroenterology 1997;112: Terada T, Nakanuma Y. Cell proliferative activity in adenomatous hyperplasia of the liver and small hepatocellular carcinoma. An immunohistochemical study demonstrating proliferating cell nuclear antigen. Cancer 1992;70: Le Bail B, Belleannee G, Bernard PH, Saric J, Balabaud C, Bioulac-Sage P. Adenomatous hyperplasia in cirrhotic livers: histological evaluation, cellular density, and proliferative activity of 35 macronodular lesions in the cirrhotic explants of 10 adult French patients. Hum Pathol 1995;26: Roncalli M, Roz E, Coggi G, Di Rocco MG, Bossi P, Minola E, Gambacorta M, et al. The vascular profile of regenerative and dysplastic nodules of the cirrhotic liver: implications for diagnosis and classification. HEPATOLOGY 1999;30: Orsatti G, Theise ND, Thung SN, Paronetto F. DNA cytometric analysis of macroregenerative nodules (adenomatous hyperplasia) of the liver: evidence in support of their preneoplastic nature. HEPATOLOGY 1993;17: Tsuda H, Hirohashi S, Shimosato Y, Terada M, Hasegawa H. Clonal origin of atypical adenomatous hyperplasia of the liver and clonal identity with hepatocellular carcinoma. Gastroenterology 1988;95: Aihara T, Noguchi S, Sasaki Y, Nakano H, Monden M, Imaoka S. Clonal analysis of precancerous lesion of hepatocellular carcinoma. Gastroenterology 1996;111: Paradis V, Laurendau I, Vidaud M, Bedossa P. Clonal analysis of macronodules in cirrhosis. HEPATOLOGY 1998;28: Bressac B, Kew M, Wands J, Oztur M. Selected G to T mutations of p53 gene in hepatocellular carcinoma from southern Africa. Nature 1991; 350: Hsu IC, Metcalf RA, Sun T, Welsh JA, Wang NJ, Harris CC. Mutational hotspot in the p53 gene in human hepatocellular carcinomas. Nature 1991;350: Kang YK, Kim CJ, Kim WO, Kim HO, Kang GH, Kim YI. p53 mutation and overexpression in hepatocellular carcinoma and dysplastic nodules in the liver. Virchows Arch 1998;432: Ding SF, Habib NA. Loss of heterozygosity in liver tumors. J Hepatol 1995;22: Nagai H, Pineau P, Tiollais P, Buendia MA, Dejean A. Comprehensive allelotyping of human hepatocellular carcinoma. Oncogene 1997;14: Piao Z, Park C, Park JH, Kim H. Allelotype analysis of hepatocellular carcinoma. Int J Cancer 1998;75: Boige V, Laurent-Puig P, Fouchet P, Flejou JF, Monges G, Bedossa P, Bioulac-Sage P, et al. Concerted nonsyntenic allelic losses in hyperploid hepatocellular carcinoma as determined by high-resolution allelotype. Cancer Res 1997;57: Salvucci M, Lemoine A, Azoulay D, Sebagh M, Bismuth MR, May E, Debuire B. Frequent microsatellite instability in post hepatitis B viral cirrhosis. Oncogene 1996;13: Ashida K, Kishimoto Y, Nakamoto K, Wada K, Shiota G, Hirooka Y, Kamisaki Y, et al. Loss of heterozygosity of the retinoblastoma gene in liver cirrhosis accompanying hepatocellular carcinoma. J Cancer Res Clin Oncol 1997;123: MacDonald GA, Greenson JK, Saito K, Cherian SP, Appelman HD, Boland R. Microsatellite instability and loss of heterozygosity at DNA mismatch repair gene loci occurs during hepatic carcinogenesis. HEPATOLOGY 1998;28: Tsopanomichalou M, Kouromalis E, Ergazaki M, Spandidos DA. Loss of heterozygosity and microsatellite instability in human non-neoplastic hepatic lesions. Liver 1999;19: Yamada T, de Sousa AT, Finkelstein S, Jirtle RL. Loss of the gene encoding mannose 6-phosphate/insulin-like growth factor II receptor is an early event in liver carcinogenesis. PNAS 1997;94: Aihara T, Noguchi S, Miyoshi Y, Nakano H, Sasaki Y, Nakamura Y, Monden M, et al. Allelic imbalance of insulin-like growth factor II gene expression in cancerous and precancerous lesions of the liver. HEPATOL- OGY 1998;28: Canzian F, Salovaara R, Hemminki A, Chadwick RB, Aaltonen LA, de la Chapelle A. Semiautomated assessment of loss of heterozygosity and replication error in tumors. Cancer Res 1996;56: Roncalli M, Bosari S, Marchetti A, Buttitta F, Bossi P, Graziani D, Peracchia A, et al. Cell-cycle related gene abnormalities and product expression in esophageal carcinoma. Lab Invest 1998;78: Vogelstein B, Fearon ER, Kern SE, Hamilton SR, Preisinger AC, Nakamura Y, White R. Allelotype of colorectal carcinomas. Science 1989;244: Ferrel LD, Crawford JM, Dhillon AP, Scheuer PJ, Nakanuma Y. Proposal for standardized criteria for the diagnosis of benign, borderline, and malignant hepatocellular lesion arising in chronic advanced liver disease. Am J Surg Pathol 1993;17: Roncalli M, Bianchi P, Grimaldi GC, Ricci D, Laghi L, Maggioni M, Opocher E, et al. Fractional allelic loss in non end-stage cirrhosis: correlations with hepatocellular carcinoma development during follow up. HEPATOLOGY 2000;31: Theise ND, Marcelin K, Goldfischer M, Hytiroglu P, Ferrell L, Thung SN. Low proliferative activity in macroregenerative nodules: evidence for an alternate hypothesis concerning human hepatocarcinogenesis. Liver 1996;16: Takagi K, Esumi M, Takano S, Iwai S. Replication error frequencies in primary hepatocellular carcinoma: a comparison of solitary versus multiple primary cancers. Liver 1998;18: