Hepatitis B Virus–associated Multistep Hepatocarcinogenesis: a Stepwise Increase in Allelic Alterations

Hepatocarcinogenesis is a multistep process, but systematic analysis using a genetic or molecular approach to accurately delineate the different stages of hepatocellular carcinoma (HCC) development is scarce. In this study, we used genome-wide allelotyping to systematically evaluate the allelic alterations in the multisteps of hepatitis B virus–associated hepatocarcinogenesis. The overall fractional allelic loss (FAL) indices of cirrhosis, dysplastic nodules (DN), and HCC were significantly different, with a clear stepwise increase (P < 0.001).


Introduction
Hepatocellular carcinoma (HCC) is one of the most common malignancies worldwide and also locally in Southeast Asia and Hong Kong.Similar to other cancers, hepatocarcinogenesis is believed to be a multistep process, largely based on observations from follow-up of patients with liver diseases who later develop HCC.This is particularly true for patients with chronic hepatitis B virus (HBV) and hepatitis C virus (HCV) infections.During the long preneoplastic stage, in which the liver often shows chronic hepatitis or cirrhosis, hepatocyte proliferation is accelerated by up-regulation of mitogenic pathways.This leads to the production of monoclonal populations of aberrant and dysplastic hepatocytes, which can result in large cell changes, dysplastic foci, and dysplastic nodules (DN; ref. 1).They can further progress to HCC and are thus considered premalignant lesions of HCC.However, systematic analysis using a genetic or molecular approach to accurately delineate the different steps/stages of HCC development is scarce.
Cirrhosis, irrespective of causes, is a definite, established risk factor of HCC.In patients with cirrhosis, cirrhotic livers with high liver cell proliferative activity have a higher risk of developing cancer (2).In autopsy series worldwide, 80% to 90% of HCC patients had cirrhosis (3); only f10% to 20% of HCC patients had no cirrhosis, although the proportion of HCC without cirrhosis may vary due to different methodologies.
As a result of regular cancer surveillance of patients with cirrhosis, the number of DNs detected clinically has increased notably.Thus, DNs have become a unique clinical entity requiring better understanding of its pathogenesis, diagnosis, and clearer treatment guidelines.Because DNs arise from cirrhotic livers, their detection is thus more difficult in a background of regeneration nodules.Very likely, DNs proceed to develop into HCC asymptomatically.The recognition of DN provides an important link regarding the pathogenesis from cirrhosis to early HCC and finally to overt HCC during multistep hepatocarcinogenesis.The development of DN has been the subject of study with regard to its role in hepatocarcinogenesis (4,5).With advances in technology, genetic studies offer a more direct evaluation on the genetic nature or clonality of DN.Previous studies on clonal analysis using the pattern of inactivation of X-linked genes (6,7) showed that DNs were monoclonal in origin.Allelic imbalance of insulin-like growth factor II gene was also observed in 42% of DNs, compared with 83% of HCCs (8).However, p53 mutations were rare in DNs (and cirrhotic livers), in contrast to HCC (9).In general, the genetic profile of liver DNs is poorly understood.There are only a few reports on the allelic or chromosomal changes of DNs in the literature (10)(11)(12)(13).However, in most of these reports, only a small number of chromosomes were examined.Genome-wide analysis on DNs is warranted.
Systematic analysis using a genetic or molecular approach to accurately delineate the different stages of HCC development is scarce.The present study consists of a genome-wide analysis using loss of heterozygosity (LOH) assay to examine 19 loci located on various chromosomes (1, 4, 8, 9, 13, 16, and 17) previously shown to have frequent chromosomal losses in HCC.Allelotyping using LOH assay is a frequently used means to study genetic losses.The present study is unique as it analyzed cirrhotic DN and HCC samples from a single institution and with HBV as the predominant etiology.In previous studies, the etiology was due to a mixture of chronic HBV and HBC viral infection and alcohol consumption, and there is no previous study on DNs with predominantly HBV infection as the etiologic factor.Our present study with HBV as the predominant etiology may help us delineate the allelic changes in DNs in the multistep hepatocarcinogenesis in relation to HBV infection.

Materials and Methods
Patients and samples.For cirrhosis cases, they were randomly selected from patients who underwent liver transplantation in 2003 and 2004.None of the patients had HCC in the explanted livers.Twenty patients were men and four were women.Their ages ranged from 31 to 57 y (mean, 47.6 y).Most (21 of 24, 87.5%) of the patients were HBsAg positive.The remaining three cases were related to Wilson's disease, chronic alcoholism, and autoimmune hepatitis.Normal DNA was obtained from normal tissues (gallbladder or porta hepatic lymph nodes) from the same operation for comparison of allelic changes.
For DNs, the patient record file from the Department of Pathology between 1993 and 2005 was reviewed.Altogether, there were 38 cases with the diagnosis of ''DN'' in liver specimens.After review, 10 cases were excluded because they were actually HCC.There were another 12 cases with only needle biopsy specimens, and they were excluded because the tissues were not adequate to make a definitive distinction whether they were DNs or well-differentiated HCCs.Another two cases were excluded because there were no normal tissues available from these two patients either in the same operation or in the record list (e.g., gastric biopsy) for comparison of allelic changes.A case of tyrosinemia in a 2-y-old girl was also excluded as it may due to different pathogenesis and age group.Finally, there were altogether 18 DNs from 13 patients who underwent either surgical resection or liver transplantation between June 1998 and June 2005 for further study.The patients' characteristics are listed in Supplementary Table S1.All patients were men, and their ages ranged from 46 to 69 y (mean, 56.7 y).Eight cases were liver explants, and the remaining five were liver resection specimens for HCCs or hepatic nodular lesions.Most of the DNs arose from cirrhotic livers, except for 2 cases of DNs arising in livers with chronic hepatitis with frequent porto-portal bridging fibrosis.Six (46.2%) of these 13 patients had concomitant HCC in their livers.All patients had chronic HBV infection, except for a man who had HCV-associated cirrhosis.The size of the DNs ranged from 1 to 2.8 cm in diameter, with mean size of 1.8 cm.Grossly, they were yellow, brown, and sometimes green due to bile production.The DNs were single in nine patients and more than one in four patients.Of the 18 DNs, 6 (33.3%) were low grade (LG) and 12 (66.7%)were high grade (HG), as classified according to the criteria proposed by the International Working Group in 1995 (1,14).
For HCC, a total of 72 HCC tumor nodules from 33 patients were studied (Table 1).This cohort of patients with HCC was different from those with  Kong.More than one tumor nodules were available in each case.Twentyeight patients were male and 5 were female.Their ages ranged from 16 to 78 y (mean, 54.5 y).Serum HBsAg was positive in 28 (84.8%)patients and none of them were anti-HCV positive.The tumor size ranged from 0.7 to 27 cm (mean, 4.7 cm).Fourteen of the 33 patients had chronic hepatitis, 18 had cirrhosis, and 1 had nonspecific changes as background liver disease.None of the patients had treatment for HCC before surgical resection.Twenty-five (cases H-201-H-211) of the 72 HCC nodules were included in our previous report (15).Cases of metastatic HCC were searched from the Pathology record file in the Department of Pathology between 1986 and 2005.To ensure that the HCCs represented true metastatic tumors and not multicentric or multifocal HCCs, only those recurrent HCCs found in extrahepatic organs or sites with the following criteria were included for study.The inclusion criteria were that (a) the recurrent or metastatic HCC were in extrahepatic sites, (b) there was no evidence of recurrent tumor in the liver at the time of diagnosis of extrahepatic metastasis, and (c) the patients had surgical samples of primary HCC and surgical or biopsy samples metastatic lesions so that tissues were available for evaluation.As a result of the stringent criteria and the rarity of HCC patients with operable metastases, there were altogether four patients only.The patients' demographic data and clinicopathologic features are listed in Supplementary Table S2.
Paraffin sections of 4-Am thick were cut, dewaxed, rehydrated, and lightly stained with Hematoxylin.All sections of cirrhosis, DNs, and HCCs were examined and microdissected with a 25G needle under a dissecting microscope (Nikon).Normal tissues such as the gallbladder or porta hepatic lymph nodes resected at the time of operation were used as normal DNA for LOH assay.
DNA extraction.Genomic DNA was extracted from target tissues with the phenol/chloroform method.Paraffin sections were lysed in buffer with proteinase K and incubated at 42jC overnight with constant gentle agitation.DNA was isolated by phenol-chloroform, precipitated, air dried, and resuspended in TE buffer.
LOH analysis.PCR was performed in a 20 AL reaction volume, which contained 100 ng of genomic DNA, 125 Amol/L deoxynucleotide triphosphate, 0.1 Amol/L each of forward and reverse primers, 0.2U of Taq polymerase, 2 AL 10Â buffer (10 mmol/L Tris-HCl, 50 mmol/L KCl, and 0.1% TritonX-100), and 1.5 mmol/L magnesium chloride.The PCR comprised an initial denaturation step at 94jC for 5 min, followed by four cycles of touch-down PCR denaturation at 94jC for 1 min, annealing at 60jC, 59jC, 58jC, and 57jC for 1 min, respectively, and extension at 72jC for 1 min.It was followed by 30 cycles of denaturation at 94jC for 1 min, annealing at 56jC for 1 min, extension at 72jC for 1 min, and a final extension at 72jC for 7 min.
The 19 polymorphic microsatellite markers chosen consisted of D1S199, D1S2892, D1S214, D4S424, D4S402, D8S258, D8S264, D8S277, D9S171, D9S269, D9S288, D13S170, D13S285, D13S159, D16S514, D16S505, D16S3066, D17S1828, and D17S831 (ABI Prism Linkage Mapping set; Applied Biosystems), with loci located on chromosomes 1, 4, 8, 9, 13, 16, and 17.These markers were chosen for their higher frequencies of heterozygosity and better success rates of results (16).The PCR products were run on an ABI Prism 377 automated sequencer (Applied Biosystems).The data were collected and analyzed by Genescan software (Applied Biosystems) and Genotyper software (Applied Biosystems), respectively.The peak heights were compared and LOH index was defined as (T 2 /T 1 )/(N 2 /N 1 ), where ''T'' was the cirrhosis/DN/HCC, ''N'' was the normal DNA (or nontumorous livers for HCC cases), and 1 and 2 were the intensities of smaller and larger alleles, respectively.LOH was defined as present if the LOH index was <0.5 or z2.0.Each of the experiments was repeated at least twice independently.The fractional allelic loss (FAL) index was calculated as the number of loci showing LOH per total number of informative loci.
Immunohistochemistry. Immunohistochemical staining for glypican 3 (GPC3) was performed on formalin-fixed, paraffin-embedded sections using the Ultravision LP Value Detection System (Lab vision) according to manufacturer's instructions.Briefly, antigen retrieval was performed by boiling the sections at 100jC with 1 mmol/L EDTA buffer (pH 8.0) for 15 min.Endogenous peroxidase was blocked with 3% hydrogen peroxide (BDH; VWR International Ltd.) for 20 min.Anti-human GPC3 (B0025R; BioMosaics) was used at 1:100 dilution and incubated at room temperature for an hour.
Statistical analysis.Fisher's exact and m 2 tests were used for the analysis of categorical data.For continuous data, Mann-Whitney U and Kruskall-Wallis tests were used for unpaired data and Wilcoxon rank sum test for paired data as appropriate.Data were analyzed using the SPSS for Windows 14.0 (SPSS, Inc.).Tests were considered significant when their P values were <0.05.

Results
LOH results of cirrhotic livers.The LOH results were based on comparison with the normal DNA ( from gallbladder or lymph node) in each case.LOH was observed at one or more of the 19 loci examined in 17 (70.8%) of the 24 cirrhotic livers (Fig. 1).The range of FAL index was 0 to 0.33, and the mean FAL index F SD and median were low at 0.09 F 0.09 and 0.07, respectively.
LOH results of DN.The results of LOH analysis on the 18 DNs are shown in Fig. 2B.The LOH results were based on comparison with the normal DNA ( from gallbladder or lymph node) in each case.All DNs showed LOH at one or more loci on the chromosomes tested.The overall FAL index of the 18 DNs ranged from 0.08 to 0.71 (mean F SD, 0.27 F 0.19; median, 0.20).
The FAL index of the 6 LGDNs ranged from 0.08 to 0.29 (mean F SD, 0.16 F 0.07; median, 0.14), whereas the FAL index of the 12 HGDNs ranged from 0.14 to 0.71 (mean F SD, 0.33 F 0.21; median, 0.27).The FAL indices of the HGDNs were significantly higher than those of the LGDNs (P = 0.033).In three of the four cases (cases 3, 8, 10, and 13) having more than one DN, the FAL indices of the DNs were similar among those in the same livers.Only in case 10, the FAL indices were quite different, with 0.29 and 0.71 for the two nodules.
LOH results of HCC nodules.LOH was detected at one or more of the loci examined in all the 33 cases, although 6 (8.1%) of the 72 HCC nodules showed no LOH using the 20 polymorphic markers (Fig. 3).The FAL index of the 72 nodules ranged from 0 to 0.85 (mean F SD, 0.40 F 0.23; median, 0.38).
Comparison of LOH results among cirrhotic livers, DNs, and HCCs.The cirrhotic livers, DNs, and HCC showed increasing FAL indices in a stepwise manner (P < 0.001, Kruskall-Wallis test; Fig. 4A).A statistically significant difference in the FAL index was documented from cirrhotic livers to DNs (P < 0.001) and from DNs to HCCs (P = 0.018).The FAL index of HCCs was higher than that of cirrhotic livers (P < 0.001; Fig. 4B).
In addition, the FAL index of the HGDNs was significantly higher than that of the LGDNs (P = 0.031; Fig. 4B).Also, the FAL index of HCC was significantly higher than that of LGDNs (P = 0.007), but the FAL index of the HGDNs was close to that of the HCCs (P = 0.235; Fig. 4B).The FAL index of the cirrhotic livers was lower than that of the LGDNs but the difference did not reach statistical significance (P = 0.057).On the other hand, the FAL index of the cirrhotic livers was significantly lower than that of the HGDNs (P < 0.001).
LOH results of metastatic HCC.In the four cases with extrahepatic HCC metastasis but no clinical evidence of HCC recurrence in the remnant liver, the FAL index of the primary HCCs ranged from 0.08 to 0.50 (mean, 0.32; median, 0.35), whereas that of the corresponding extrahepatic metastatic HCCs ranged from 0.08 to 0.75 (mean, 0.42; median, 0.44), which was slightly higher than that of the primary tumors (Fig. 4C).However, the number of cases was too small for meaningful statistical analysis.
Because of the difficulty to obtain cases of extrahepatic metastatic HCCs with tissue available for analysis, we then analyzed multinodular HCCs consisting of cases with intrahepatic metastasis, which were more readily available, for further study as below.
Identification of the group of intrahepatic metastasis in multinodular HCC.In our previous finding, multinodular HCCs having z30% discrepancy in LOH patterns were likely to be multicentric occurrence, and those with <30% discrepancy were likely to be intrahepatic metastasis (15).The LOH pattern was regarded as identical when the same markers showed loss of the same allele and different when the same markers showed loss of one allele in one tumor nodule but no loss in the other.In the present study, the LOH patterns among the multiple HCC tumor nodules of individual patients were compared.We used our previously reported finding of 30% discrepancy in LOH pattern between HCC nodules per informative markers as a cutoff value to distinguish multicentric occurrence of multinodular HCCs from intrahepatic metastasis.Twelve (36.4%) of the 33 cases had z30% discrepancy in LOH patterns, thus indicating multicentric occurrence of HCCs, and 21 (63.6%)cases had <30% discrepancy, thus indicating intrahepatic metastasis (Fig. 3).The frequencies of multicentric occurrence and intrahepatic metastasis results were consistent with our previous reported findings.Indeed, those multinodular HCCs with <30% discrepancy of LOH patterns between the nodules, i.e., multinodular HCCs of intrahepatic metastasis, had significantly more frequent venous invasion (P = 0.005) and tumor microsatellite formation (P = 0.005; Table 2).
We then examined the FAL indices of the two groups of multinodular HCC.The multicentric occurrence group had FAL index ranging from 0 to 0.85 (mean F SD, 0.42 F 0.23; median, 0.41), which was not significantly different from that of the intrahepatic metastasis group (range, 0 to 0.77; mean F SD, 0.39 F 0.22; median, 0.38; P = 0.460).
LOH results of intrahepatic metastatic HCC.We further examined the group of multinodular HCCs of intrahepatic metastasis (21 cases, 46 nodules) and compared the primary tumors and the corresponding tumor nodules of intrahepatic metastasis.The primary tumors had significantly large tumor sizes than the corresponding intrahepatic metastases (mean F SD, 8.63 F 5.34 and 1.89 F 1.31, respectively; P = 0.002).As shown in Fig. 4D, there was no significant difference between the FAL indices of the primary tumors and their corresponding tumors of intrahepatic metastasis (P = 0.497, Wilcoxon rank test).

Discussion
Using genome-wide allelotyping, our results have clearly documented a stepwise increase in allelic losses from cirrhosis, through DN (LG to HG), to primary and metastatic HCC.The study is unique as it analyzed cirrhotic, DN, and HCC samples from a single institution.In addition, the cases were predominantly associated with HBV infection and the findings thus give insight on HBV-related hepatocarcinogenesis.We had previously reported the allelic losses on a different, separate series of cirrhotic livers with concomitant HCC (17).Although it would be interesting to follow the different steps of hepatocarcinogenesis in the same patients and study on a series of cases having cirrhosis, DNs, and HCCs from the same patients, the number of such cases will be small and have limitations for a meaningful analysis.
Cirrhosis is an established risk factor of HCC, irrespective of causes.Indeed, cohort studies indicate that HCC is currently the major cause of liver-related death in patients with compensated cirrhosis.Although cirrhosis has often been viewed as a premalignant condition, studies of its genetic alterations are few.T nod, tumor nodules from the same patients.Column as T1, T2, and T3, respectively.Microsatellite markers are listed in the first row.Gray boxes, retention; white boxes, noninformative or data not available; black boxes, LOH; lined boxes, microsatellite instability.LOH occurring at different alleles among the tumor nodules from same patient are indicated as A1 and A2, respectively.Twelve (36.4%) of the 33 cases had z30% discrepancy in LOH patterns, thus indicating multicentric occurrence of HCCs, and 21 (63.6%)cases had <30% discrepancy, thus indicating intrahepatic metastasis.The overall LOH frequency of informative loci in the 72 HCC samples was 0.39 (348 of 827).FAL index of HCC ranged from 0.00 to 0.85, with mean F SD being 0.40 F 0.23.The FAL indices of HCCs of multicentric occurrence and intrahepatic metastasis were 0.42 (10 of 68) and 0.39 (28 of 101), respectively.
In this study, we have shown that LOH was uncommon in cirrhotic livers, which in this study were predominantly due to chronic HBV infection.Although LOH was observed at z1 of the 19 loci in 70.8% of the cirrhotic livers, the mean FAL index was low, at 0.09.This low incidence of allelic losses was consistent with those in previous reports (11,13,18) and also our previous report on mostly HBVassociated cirrhotic livers but with concomitant HCC (17).From the previous reported series, the FAL indices in cirrhosis vary from 0.05 to 0.15 (11,13,18).The etiologies in the previous studies consisted of a mixture of both chronic HBV and HCV infection, with HBV-associated cirrhosis ranging from 21% to 44% of their cases and HCV-associated cirrhosis ranging from 33% to 55%.From the data, it seems that the frequencies of allelic losses in cirrhosis due to different viral etiologies are similar.
The development of DN has been the subject of study with regard to its role in hepatocarcinogenesis, but the genetic profile of liver DNs is poorly understood.In this study, we selected the DNs for study after stringent histologic examination, supported with immunohistochemistry for GPC3 (Supplementary Table S1; refs.19,20).Absence of GPC3 expression supported the histologic classification of DN.Importantly, all the DNs in the present study were from either resected liver specimens or liver explants and not needle biopsy specimens, therefore allowing comprehensive histologic examination and confirmation.In the present study, all of the DNs, both LG and HG, showed LOH at one or more loci on these chromosomes tested.In addition, the frequency of allelic alterations in terms of FAL index was significantly higher than that of cirrhotic livers.In particular, the FAL index of the HGDNs was significantly higher than that of the LGDNs.The FAL index of HGDNs was very close to that of HCCs in this study.The findings have clinical relevance; because HGDN is genetically close to HCC in terms of allelic losses, its biological behavior may likely resemble that of (early) HCC.Hence, liver resection should be considered for HGDN as a curative therapeutic measure (21).
Furthermore, the LOH results of DNs in the present study are in accordance with those of the few previous reports.In a previous report studying LOH mainly at 1p36-p32, an increasing frequency of LOH was observed from cirrhotic nodules, through LGDNs and HGDNs, to HCC (11).In another study examining chromosomes 8p (8p21-22) and 11p (11p13), a high frequency of LOH was observed in DNs (10).LOH was found to be in high frequency (up to 41%) in DNs at certain loci on particular chromosomes, comparable with that (42.1%) in HCC in the same study.However, there was no statistically significant difference in the incidence and FAL index between LG and HGDNs.Another study analyzing LOH on chromosome 1 only showed a stepwise increase from cirrhosis through DN to HCC (13).Additionally, in a previous study analyzing chromosomal gains and losses using comparative genomic hybridization, the frequency and pattern of genetic alterations in DNs highly resembled those in HCCs (12).Apart from allelic losses, other molecular alterations have also been found in DN.There is also a stepwise change in hepatocarcinogenesis, including increasing methylation of CpG islands of genes (22,23), telomere shortening (24), and expression of the telomere-binding proteins

Cancer Research
Cancer Res 2008; 68: (14).July 15, 2008 (25).Recently, the molecular expression profiles of the different stages in hepatocarcinogenesis from cirrhosis through DNs to HCC have been examined with microarray analysis (26)(27)(28).These have been useful tools in identifying a possible ''molecular signature'' in distinguishing DNs from HCC.All these data give supportive evidence for a stepwise molecular change in hepatocarcinogenesis and support the notion that DN is precancerous from a genetic approach.
LOH was a common occurrence in our predominantly HBVassociated HCCs, with all HCC cases exhibiting LOH at one or more loci and an overall FAL index of 0.40, which was significantly higher than DNs and cirrhotic livers.In addition, the 4 cases of metastatic HCC showed a slightly higher FAL index (0.42) than their corresponding primary HCCs (0.32).
Aberrations have been found to differ in HCC with different etiologic background.Chromosomal aberrations were more frequent in HBV-related HCCs than in HCV-associated tumors (29).Another study has shown that HBV-associated HCCs had a significantly more frequent (40% on average) losses at 4q, 16q, and 17p than in nonviral HCC samples, suggesting that these abnormalities are much associated with HBV infection (30).Similarly, more prevalent loss on 17p, including the p53 region, was also seen in HBV-associated HCCs in another study (29).On other chromosomes, a gain of 10q (7 of 41, 17%) was detected exclusively in cases with HCV infection, whereas an amplification of 11q13 was more frequently seen in HBV-positive HCCs.It is to be noted though that in some other studies, no significant difference in chromosomal aberrations was found between HBVand HCV-associated HCCs (31).
In addition, multinodular form of HCC is common (32).Multiple HCC nodules within a liver may reflect either the dissemination of cancer cells from a single primary HCC to form satellite tumor nodules (intrahepatic metastasis), or synchronous development of several independent tumors (multicentric occurrence).The two types of HCC nodules carry important differences in pathogenesis and can have an effect on its treatment and prognosis.Intrahepatic metastasis obviously indicates tumor progression.To address this question, we previously determined the molecular relationship between multiple tumor nodules in HCC within individual patients in 11 patients (15).Using LOH and comparative genomic hybridization analyses, 36% of the patients had multiple HCCs with different clonalities, hence of multicentric origin, whereas the remaining 64% patients had multiple HCCs with similar clonal relationship, hence of intrahepatic metastasis.The results were further confirmed by molecular method using HBV DNA integration pattern.Importantly, pathologic features alone could not accurately predict their clonal origins in every case.Therefore, assessment of DNA alterations can provide precise determination of the clonality of the multiple HCCs within a patient.From this previous study using the genome-wide LOH assay, a cutoff value of 30% discrepancy in LOH pattern was found and could be applied to distinguish intrahepatic metastases and multicentric occurrence.This cutoff value of 30% discrepancy rate of the LOH status of the analyzed loci was also identified as a criterion by a recent study from Japan (33).
In the present study, we applied this criterion to distinguish the group of multinodular HCC with intrahepatic metastasis.Upon pathologic correlation, those multinodular HCCs with <30% discrepancy of LOH patterns between the nodules had significantly more frequent venous invasion and tumor microsatellite formation.As venous invasion and tumor microsatellite formation are established features of intrahepatic metastasis in HCC, the finding was consistent with the notion derived from LOH analysis in this study that these multiple HCC nodules represented intrahepatic metastasis.We made use of this criterion to delineate tumors of intrahepatic metastasis for further analysis of allelic losses but found no significant difference between the FAL indices of the primary tumors and their corresponding tumors of intrahepatic metastasis.This indicates that, in contrast to cirrhosis progressing to DN and HCC, the transformation of primary HCC to a metastatic phenotype is not accompanied by major significant allelic losses.However, such absence of major allelic losses in this transformation to a metastatic phenotype does not exclude smallscale chromosomal losses or gene deletions occurring in such process.In fact, small-scale chromosomal losses or gene deletions have been reported in metastatic HCC compared with the corresponding primary HCC (34).Overall, we have shown that there was a stepwise increase in the frequencies of allelic losses from cirrhosis, LGDN, and HGDN to HCC.We conclude that HBV-associated hepatocarcinogenesis is a multistep process accompanied by stepwise increase in allelic losses.Consistent with this notion, losses of chromosomal materials, which may harbor tumor-or metastasis-suppressor genes, can promote tumor development and progression.
cirrhosis or DN in this study.The resection was done between May 1993 and August 2000 in the Department of Surgery at Queen Mary Hospital, Hong

Figure 1 .
Figure 1.Results of LOH analysis on the 24 cirrhosis cases.Case numbers are indicated in the left column.Microsatellite markers and their chromosomal positions are listed in the first two rows.Gray boxes, retention; white boxes, noninformative or data not available; black boxes, LOH.The overall LOH frequency of informative loci was 0.09 (31 of 340); FAL index of cirrhosis cases ranged from 0.00 to 0.33, with a mean F SD of 0.09 F 0.09.

Figure 2 .
Figure 2. A, histologic features of two representative LGDNs (top ) and two HGDNs (bottom ).B, results of LOH analysis on the 18 liver DNs.Case numbers are indicated in the left column.Six nodules were LGDNs and 12 were HGDNs.Microsatellite markers and their chromosomal positions are listed in the first two rows.Gray boxes, retention; white boxes, noninformative or data not available; black boxes, LOH; lined boxes, microsatellite instability.The overall LOH frequencies of informative loci in the 18 DNs was 0.19 (36 of 185).The overall LOH frequencies in LG and HGDNs were 0.15 (10 of 68) and 0.29 (26 of 90), respectively.FAL index of DNs ranged from 0.08 to 0.71; mean F SD, 0.27 F 0.19.Mean FAL indices for LG and HGDNs were 0.16 and 0.33, respectively.

Figure 3 .
Figure 3. Results of LOH analysis on the 72 HCCs.Case numbers are indicated in the left column.T nod, tumor nodules from the same patients.Column as T1, T2, andT3, respectively.Microsatellite markers are listed in the first row.Gray boxes, retention; white boxes, noninformative or data not available; black boxes, LOH; lined boxes, microsatellite instability.LOH occurring at different alleles among the tumor nodules from same patient are indicated as A1 and A2, respectively.Twelve (36.4%) of the 33 cases had z30% discrepancy in LOH patterns, thus indicating multicentric occurrence of HCCs, and 21 (63.6%)cases had <30% discrepancy, thus indicating intrahepatic metastasis.The overall LOH frequency of informative loci in the 72 HCC samples was 0.39 (348 of 827).FAL index of HCC ranged from 0.00 to 0.85, with mean F SD being 0.40 F 0.23.The FAL indices of HCCs of multicentric occurrence and intrahepatic metastasis were 0.42 (10 of 68) and 0.39 (28 of 101), respectively.

Figure 4 .
Figure 4. A, scatter plot showing stepwise increase of FAL index, from cirrhotic liver through DN to HCC.Horizontal lines, median FAL indices.B, comparison of the median FAL indices between different histologic groups by Mann-Whitney U test.Tests were considered significant when their P values were <0.05.C, results of LOH analysis on the four HCC cases with extrahepatic metastasis.Case numbers are indicated in the left column.Microsatellite markers are listed in the first row.Gray boxes, retention; white boxes, noninformative or data not available; black boxes, LOH.D, comparison of the FAL indices between primary HCC and corresponding intrahepatic metastasis in the 21 cases of multinodular HCC (46 nodules).The box plot shows the median, quartile, and extreme values.Horizontal lines, median FAL indices.There was no significant difference between the FAL indices of the primary tumors and their corresponding tumors of intrahepatic metastasis (P = 0.497, Wilcoxon rank test).