|Home | About | Journals | Submit | Contact Us | Français|
To explore the utility of cytogenetic abnormalities in the distinction of hepatic adenoma (HA) and well-differentiated hepatocellular carcinoma (HCC).
Array-based comparative genomic hybridization (CGH) was used to determine chromosomal abnormalities in 39 hepatocellular neoplasms: 12 HA, 15 atypical hepatocellular neoplasms (AHN) and 12 well-differentiated HCC. The designation of AHN was used in two situations: (i) adenoma-like neoplasms (n = 8) in male patients (any age) and women >50 years and <15 years old; (ii) adenoma-like neoplasms with focal atypical features (n = 7). CGH abnormalities were seen in none of the HAs (0/12), eight (53%) AHNs and 11 (92%) HCCs. The number and nature of abnormalities in AHN was similar to HCC with gains in 1q, 8q and 7q being the most common. Although follow-up information was limited, recurrence and/or metastasis were observed in three AHNs (two with abnormal, one with normal CGH).
Adenoma-like neoplasms with focal atypical morphological features or unusual clinical settings such as male gender or women outside the 15–50 year age group can show chromosomal abnormalities similar to well-differentiated HCC. Even though these tumours morphologically mimic adenoma, they can recur and metastasize. Determination of chromosomal abnormalities can be useful in the diagnosis of AHN.
It can be extremely difficult to distinguish hepatic adenoma from well-differentiated hepatocellular carcinoma (HCC) arising in non-cirrhotic liver on morphological grounds. The presence of macrotrabecular or acinar architecture, small cell change, cytological atypia, prominent nucleoli, mitotic activity, vascular invasion and absence of Kupffer cells favour HCC. Reticulin staining can sometimes be helpful by demonstrating the intact reticulin framework in adenoma compared with fragmentation and patchy loss of reticulin fibres in HCC. However, none of these features can be relied upon with certainty, as most of them may not be present in well-differentiated HCC. On the other hand, occasional nuclear atypia and acinar architecture can be observed in hepatic adenomas (HA).1 The distinction can be even more daunting when a small amount of tissue is available for review in the form of a needle biopsy specimen.2,3 The natural history and management of HA and HCC are different and hence it is crucial to make this distinction.4,5
Comparative genomic hybridization (CGH) has revealed a consistent pattern of chromosomal gains and losses in HCC. The most prominent changes are gains of part or entire chromosome arms 8q (49–81%), 1q (60–79%) and 7q (40–64%), and loss of 16q (36–65%).6–12 Other common abnormalities include overrepresentation at sites Xq and 5p, and losses at 4q, 8p, 13q, 16q and 17p. 6–12 Certain clinicopathological associations have been noted with specific abnormalities. Loss of 4q has been associated with elevated α-fetoprotein levels, p53 mutations,10 tumour size and vascular invasion.12 Gain of 8q and 20q are associated with large tumour size.7,11 Gain of 8q and loss of 13q are seen more often in HCC arising in non-cirrhotic liver.11,12 Chromosome 9p and 6q losses have been reported to be independent predictors of poor out-come. 8 In contrast to HCC, genomic instability has not been observed in adenomas.
These results suggest that cytogenetic studies may be helpful in distinguishing between adenomas and HCC. However, several issues remain to be addressed before cytogenetic analysis can be used in clinical practice for this purpose. The total number of adenomas studied so far is limited and have largely been studied by one group.13 Most of the adenomas studied in the literature have had typical clinicopathological characteristics, i.e. they arose in young women and lacked atypical morphological features. Typical HA is less likely to pose diagnostic problems and determination of genetic abnormalities will not be required for its diagnosis in clinical practice. On the other hand, ‘adenomas’ with atypical features like those arising in men, older women or those with focal atypical morphological features can be impossible to distinguish from HCC at the microscopic level. The evaluation of chromosomal abnormalities will be most helpful in these clinical situations.
In this study, array-based CGH was used for the delineation of chromosomal abnormalities in HA, well-differentiated HCC and a group of atypical hepatocellular neoplasms (AHN) where a definite diagnosis could not be established on clinical and histological grounds.
The study group consisted of 39 resected hepatocellular neoplasms from the pathology files of the University of California San Francisco, Mayo Clinic, University of Toronto and Yale New Haven Medical Centers. The study set comprised 12 HA, 15 AHN and 12 well-differentiated HCC. The slides were reviewed to confirm the diagnosis in each case. Clinicopathological parameters including age, gender and date of surgery were recorded from the pathology report.
The designation of AHN was used in two situations: (i) tumours with adenoma-like morphology (n = 8) that occurred in men (any age) or in women outside the 15–50 years age range; (ii) tumours with adenoma-like morphology (n = 7) with focal atypical areas showing small cell change, prominent acinar architecture or abnormal trabeculae. The atypical features were randomly distributed and did not form a definite nodule within the main tumour. In all cases, atypia was present in <5% of the tumour and was not considered sufficient for an unequivocal diagnosis of HCC.
The arrays were prepared and hybridized as previously described.14 In brief, human 1.14 arrays were obtained from the University College San Francisco (UCSF) Cancer Center Array Core (http://cancer.ucsf.edu/array/index.php). The arrays comprised 2433 bacterial artificial chromosome (BAC) clones covering the human genome at 1.5 Mb resolution. For hybridization, 1 mg each of tumour-and gender-matched reference DNA isolated from normal donor lymphocytes was labelled by random priming using Cy3-dCTP and Cy5-dCTP, respectively, using the Bioprime Kit (Invitrogen, Carlsbad, CA, USA). Sephadex G-50 column (Amersham, Piscataway, NJ, USA) was used to remove unincorporated fluorescent nucleotides. Sample and reference DNA were mixed with 100 mg Cot-1, precipitated and resuspended in hybridization solution, followed by denaturation for 10 min at 72°C and incubated for 1 h at 37°C to block repetitive sequences. Hybridization was performed for 48–72 h in a moist chamber on a slow rocking table. The arrays were washed for 10 min in 50% formamide, 2× standard saline citrate at 45°C and for 10 min in phosphate buffer at room temperature. Slides were mounted in mounting solution containing 0.3 mg/ml 4′-6-diamidino-2-phenylindole (DAPI). Three single-colour intensity images (DAPI, Cy3 and Cy5) were collected for each array using a charge-coupled device camera.
The UCSF SPOT software (http://jainlab.ucsf.edu/Downloads.html) was used 15 to segment automatically the spots based on the DAPI images, perform local background correction and calculate various measurement parameters, including log2 ratios of the total integrated Cy3 and Cy5 intensities for each spot. A second custom program SPROC (http://jainlab.ucsf.edu/Downloads.html) was used to associate clone identities and a mapping information file with each spot, so that the data could be plotted relative to the position of the BACs. Chromosomal aberrations were classified as a gain when the normalized log2 Cy3/Cy5 ratio was >0.225 and as a loss when the ratio was <−0.225. This number was determined as threefold of the average standard deviation of normal versus normal array CGH hybridization.
The differences in chromosomal abnormalities in atypical hepatocellular neoplasms and HCC were compared using Fisher’s exact test and unpaired Student’s t-test.
All 12 cases of HA occurred in women (age range 27–50 years). History of oral contraceptive use was present in three cases, absent in two cases and not known in the others. All cases had typical morphological features: one to two cell thick plates, unpaired arteries, no bile ducts, absence of cytological atypia and intact reticulin network (Figure 1).
All 15 cases of AHN occurred in non-cirrhotic livers (age range 3–57 years, eight men, seven women). Hepatic adenoma-like morphology was seen in eight cases (Table 1). These cases did not show any significant cytological atypia, and the reticulin framework was intact (Figures 2 and and3).3). These cases were placed in the atypical category since they occurred in men (n = 4) or in women >50 years or <15 years old (n = 3). The other seven cases of AHN also showed adenoma-like morphology, but in addition had focal atypia in the form of small cell change, prominent acinar architecture or thick cell plates (Figures 4 and and5).5). These changes were not considered sufficient for an unequivocal diagnosis of HCC.
Of the 12 patients with HCC, four had cirrhosis and the remainder arose in non-cirrhotic livers. The cause for cirrhosis was chronic hepatitis C in three cases and chronic hepatitis B in one case. All cases of HCC were histologically classified as well-differentiated and showed uniformly thick cell plates (three or more cells), small cell change, high nuclear–cytoplasmic ratio and an abnormal reticulin network.
None of the 12 HAs showed gains or losses at any chromosomal locus. There were no recurrences in three cases for whom follow-up information was available.
Abnormalities by array-based CGH were observed in 11 (92%) cases (Table 2). The most common abnormalities were gains at 1q (55%), 8q (33%) and 7q (25%). There were no differences in changes between tumours arising in cirrhotic versus non-cirrhotic liver.
Of the 15 AHNs, abnormalities were observed in eight (53%) cases (Table 2).
The number of cases with chromosomal aberrations was significantly lower in AHNs compared with HCCs (53% versus 92%, P = 0.04). The mean number of chromosomal aberrations in AHNs was also lower compared with well-differentiated HCCs, but this difference was not significant (Table 2). Gains of 1q, 8q and 7q were the most common abnormalities in both AHN and HCC, and there was no difference among the two groups (Table 2). Some abnormalities that were observed only in HCC were chromosome 3 gain (n = 2), 13q loss (n = 2), 16q loss (n = 3) and chromosome 19 abnormalities (n = 3).
Well-differentiated HCCs may have minimal cytological and architectural atypia and can closely resemble hepatic adenomas in non-cirrhotic liver. It may be difficult to identify conclusively small cell change and thick cell plates (three or more cells thick), especially on needle biopsy specimens, to meet the diagnostic criteria of HCC as recommended by the International Working Party.16 This distinction is critical, as HCC is a malignant neoplasm with potential for recurrence and metastasis as opposed to the benign outcome in adenomas. Several studies have shown characteristic chromosomal abnormalities in HCC by CGH 6–12 and fluorescence in situ hybridization (FISH).13,17,18 Gains at chromosomal regions 1q, 7q, 8p, 8q and X, and losses at 4p, 11q and 16q are characteristic of HCC.6–12
Our results show that a majority of AHNs show chromosomal aberrations that are typical of well-differentiated HCC, even though the morphology closely resembles adenoma. Gains of 1q, 8q and 7q were the most frequent aberrations in both AHN and well-differentiated HCC. Similar chromosomal changes have been reported in well-differentiated HCC in the literature. Gains of 1q and 8q have been identified in 66–86% and 30–80%, respectively, in two series of well-differentiated HCC.3,19 Gains of 7q occurred in up to 50% of cases.3 A large CGH-based study utilizing hierarchical clustering and a recent meta-analysis study have identified gains of 1q and 8q as the earliest CGH changes in hepatic carcinogenesis.20,21 These data strongly suggest that some AHNs represent the earliest form of well-differentiated HCC. Although follow-up of AHN cases was limited, the presence of recurrence and/or metastasis in three cases further supports this contention.
Some chromosomal changes, including loss of chromosome 3, 4q, 13q and 16q, were observed only in HCC, but not in AHNs. These changes have been associated with intermediate or late stages of hepatic carcinogenesis.21 Cytogenetic abnormalities were fewer in AHN compared with HCC (not significant) and were seen in a smaller number of cases (53% versus 92%, P = 0.04). Hence, AHN bears a resemblance to well-differentiated HCC in terms of cytogenetic aberrations and clinical course, but appears to be cytogenetically less advanced. This is somewhat similar to the increasing aneuploidy that parallels progression of well- to poorly-differentiated HCC.22
None of the HAs in our series showed chromosomal gains or losses at any locus, which is in accordance with the published literature. In a study of 10 HAs by Wilkens et al.,3 eight did not show any abnormalities on CGH. One case showed gain of 7p, and another case showed gains of 17q, 20p and 20q. None of the adenomas showed the early changes observed in HCC, such as gains of 1q and 8q or other characteristic changes such as gains of 6q, 7q, 8q and Xq or losses of 16q, 8p or 4q. In another study using FISH analysis by five centromeric probes (chromosomes 1, 6, 7, 8 and X), the same group has shown that none of 14 adenomas had any abnormality, whereas 13/14 HCCs showed gain of at least one locus.13 In two other studies, most of the HCCs but none of the HAs showed gains of chromosomes 1 or 8.17,18 The only reported exception was 1q gain in two HAs in the setting of type 1 glycogenosis.23 One case with HA-like morphology in our series also occurred in the setting of type 1 glycogenosis and showed gains of 7q, 18q and 19p. Another case in our series with HA-like morphology occurred in a patient with hereditary haemochromatosis and showed CGH abnormalities characteristic of HCC. HA-like tumours and HCC have been reported in hereditary haemochromatosis in the absence of cirrhosis, 24–26 and patients homozygous for C282Y mutations have been shown to be at increased risk for HCC.27 Hence, adenoma-like neoplasms in the setting of liver diseases such as type 1 glycogenosis or haemochromatosis may represent well-differentiated HCCs or preneoplastic lesions akin to high-grade dysplastic nodules in cirrhotic liver.
Even though most of the AHNs in this study showed chromosomal changes typical of HCC, the sample size in this study was small and may not be an accurate indicator of the proportion of AHNs that are HCC. Several AHNs in our series did not have any chromosomal abnormality. It is possible that some of these may be uncommon instances of HAs in men or women >50 years old. The incidence of HAs in men has reportedly increased in a dramatic fashion and has recently been reported as 1:1 (male:female) in the last few years in one series.28 However, great caution needs to be exercised before a benign diagnosis is rendered in these situations, as AHNs may represent HCC without cytogenetic abnormalities. No cytogenetic changes on CGH have been observed in up to 7.6% of HCC.6 Indeed, metastatic HCC occurred in one case of AHN with normal cytogenetics in our study. There have been recent attempts to classify HAs based on molecular abnormalities.29,30 Adenomas in men with atypical histological features (cytological atypia, acinar architecture) frequently have β-catenin mutations.29 These tumours are likely to be interpreted as borderline lesions between adenoma and HCC corresponding to the category of AHN defined in this study. It would be interesting to determine if adenoma-like neoplasms with atypical features and β-catenin mutations show chromosomal abnormalities characteristic of HCC. If the latter is true, some AHNs may be a distinct form of well-differentiated HCC that arise in non-cirrhotic liver and are characterized by β-catenin mutations and a favourable outcome. Interestingly, good prognosis and occurrence in non-cirrhotic liver have been noted in HCC with β-catenin mutations.8
There have been rare case reports in the literature of ‘transformation’ of HA to HCC. The HCC was present concurrently with the adenoma 31–36 or developed several years after resection of adenoma.37–39 Most of these tumours would have been categorized as AHN in our study. In our series, several cases of adenoma-like AHN showed multiple chromosomal gains and losses, and one of them recurred after 3 years. The latter case could easily have been interpreted as HA recurring as HCC if the CGH changes had not been taken into account. Although malignant transformation in adenomas may rarely occur, it is likely that the ‘adenoma’ portion of many of the reported cases has been an extremely well-differentiated form of HCC similar to AHNs in our study. Mouse studies suggest that HA and HCC follow different genetic pathways, making it less likely that true HAs develop into HCC.40
CGH is a good screening tool for chromosomal abnormalities and has played a key role in characterizing abnormalities that can discriminate between adenomas and HCC. However, it may be too cumber-some for routine clinical use. An assay based on FISH using locus-specific or centromeric probes especially directed at chromosomes 1 and 8 can be valuable for the definite diagnosis of AHNs. The FISH assay can also be performed on core biopsy specimens. Like CGH, several reports have tried FISH in HAs and HCC,13,17,18,41 but its utility in morphologically borderline lesions designated as AHNs in this study has not been examined. Recently, immunohistochemical expression of glypican-3, an oncofetal antigen, has been described in HCC and is not expressed in adenomas. 42–44 However, glypican-3 has low sensitivity (approximately 50%) for the diagnosis of well-differentiated HCC that bears a close histological resemblance to adenoma45 or may be negative in needle biopsy due to focal staining.46 Hence, positive results favour HCC, but negative results are inconclusive. The combination of reticulin stain, glypican-3 immunohistochemistry and FISH assay for 1q and 8q may be able to establish the correct diagnosis in most cases.
In summary, the majority of AHNs that resemble HA, but occur in an unusual age/gender setting or have focal atypical cytoarchitectural features show chromosomal abnormalities typical of HCC. Since these abnormalities are absent in HAs, at least some of the AHNs are likely to represent well-differentiated HCC. Despite the well-differentiated appearance and histological resemblance to adenoma, these tumours can recur and metastasize. Delineation of chromosomal abnormalities can be helpful in distinguishing HA from HCC when clinicopathological features such as clinical setting and morphological findings yield equivocal results.
This study was funded by REAC grant and the Liver Center, University of California, San Francisco, CA, USA.