Molecular pathogenesis of hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is the fifth most common cancer and the third leading cause of cancer death worldwide. Hepatocarcinogenesis is a multistep process evolving from normal through chronic hepatitis/cirrhosis and dysplastic nodules to HCC. With advances in molecular methods, there is a growing understanding of the molecular mechanisms in hepatocarcinogenesis. Hepatocarcinogenesis is strongly linked to increases in allelic losses, chromosomal changes, gene mutations, epigenetic alterations and alterations in molecular cellular pathways. Some of these alterations are accompanied by a stepwise increase in the different pathological disease stages in hepatocarcinogenesis. Overall, a detailed understanding of the underlying molecular mechanisms involved in the progression of HCC is of fundamental importance to the development of effective prevention and treatment regimes for HCC.

Hepatocellular carcinoma (HCC) is one of the most common malignancies in the world. Globally, it ranks fifth among the most common cancers but is the third leading cause of cancer death, with an estimate of more than 500 000 new cases each year (1,2). There is a striking geographical difference in the incidence of HCC. Eighty per cent of new cases occur in developing countries. High-incidence areas include the sub-Saharan Africa, east and southeast Asia, whereas lowincidence areas include northern and western Europe and North America. Better control of the risk factors has resulted in a recent decline in HCC in some places such as Taiwan and China. For instance, the mortality rates because of HCC in male and female children younger than 15 years of age have decreased by up to 70 and 62%, respectively, in Taiwan because of vaccination against hepatitis B virus (HBV) (3). However, there is a trend of rising rates of HCC in developed countries in Europe and North America recently. In USA, the age-adjusted incidence has doubled over the past two decades (4). In southeast Asia, HCC is the second most common fatal cancer. It has been the second most common fatal cancer and its ranking among the common causes of fatal cancers has not changed since the 1970s. A male predominance is obvious, with a male to female ratio of 2-4:1 (5).
The risk factors of HCC are well established and include chronic HBV and hepatitis C viral (HCV) infection (6,7), cirrhosis (8) and aflatoxin B1 (9). Alcohol abuse leading to cirrhosis increases HCC risk appreciably by promoting liver cirrhosis; only severe, but not moderate, alcohol consumption is related to HCC (10,11). The role of tobacco smoking in the causation of HCC is controversial. Other risk factors include inherited metabolic diseases such as hereditary haemochromatosis (12), a-1-antitrypsin deficiency and hereditary tyrosinaemia. Obesity and diabetes can lead to non-alcoholic steatohepatitis, which is also an established risk factor for HCC, most likely via progression of the steatohepatitic disease to cirrhosis and HCC (13).

Multistep hepatocarcinogenesis
Hepatocarcinogenesis is believed to be a multistep process (Fig. 1). In the majority of the cases HCC arises in a background liver disease of either chronic hepatitis or cirrhosis, although HCC can arise in normal liver. In autopsy series worldwide, there are about 10-20% of HCC patients having no cirrhosis (14). However, only a small proportion of HCC arise in absolutely normal or normal-looking livers and the majority of non-cirrhotic livers show fibrosis ranging from mild fibrosis to septal or bridging fibrosis. There are also other histological changes including acinar necro-inflammation, steatosis or liver cell dysplasia (15).
Cirrhosis is a definite, established risk factor for HCC. Chronic HBV and HCV infection, hereditary haemochromatosis and alcoholism are leading causes of cirrhosis. Cohort studies have revealed trends of decreasing mortality rates owing to cirrhosis and increasing mortality rates owing to HCC in USA and in European countries (16,17). This is related to the better management of non-HCC complications of cirrhosis, resulting in the longer survival of patients with cirrhosis. Among the risk factors causing cirrhosis and HCC, patients with cirrhosis owing to HCV infection are associated with the highest HCC incidence. The 5-year cumulative incidence in Japan and in the West is 30 and 17% respectively (8). This is followed by hereditary haemochromatosis with a 5year cumulative incidence of 21%. In patients with HBV-related cirrhosis, the 5-year cumulative HCC risk is 15% in high-endemic areas and 10% in the West. In viral-related cirrhosis, HBV/HCV and HBV/hepatitis D virus (HDV) co-infections increase the HCC risk by two-to six-fold relative to each infection alone.
In humans, liver cell dysplasia (LCD) consists of large and small cell dysplasia (SCD), and they usually occur in a background of cirrhosis or chronic hepatitis. LCD and SCD can be observed in up to 30 and 25%, respectively, in liver biopsies from patients with chronic liver diseases because of HBV or HCV infection (18,19). The risk of LCD in developing into HCC is controversial (18,19). On the contrary, SCD is an important independent risk factor for developing HCC in virus-associated cirrhosis (19,20). In another study on LCD and SCD adjacent to HCC using microdissection and chromosomal genomic hybridization (CGH), no genetic abnormalities were detected in LCD foci (21). In contrast, the adjacent SCD showed a subset of chromosomal alterations present in HCCs. To date, there seems to be more evidence supporting that SCD is a risk factor for HCC than LCD.
Dysplastic nodules (DN) are precancerous lesions in hepatocarcinogenesis and arise from a cirrhotic background (22,23). These nodules are increasingly detected by radiographical techniques in cirrhotic livers or are removed during transplantation procedures. They have created a new and challenging histological spectrum of liver pathology. They are divided into high-and low-grade DNs, based on the histological features including cellular architecture, presence or absence of portal tracts and cytological features (22). The number of dysplastic nodules detected clinically has increased, because patients with HBV-or HCVassociated cirrhosis, at increased risk for HCC, undergo regular cancer surveillance. The natural outcome of dysplastic nodules in cirrhosis detected by ultrasonography is still poorly understood. Seki et al. (24) found that most of the DNs disappeared or remained unchanged and 12.1% of the nodules progressed to HCC, in a study using fine needle aspiration (FNA) coupled with ultrasound features. In another study, high-grade DN and LCD around the nodules were found to be independent predictors of malignant transformation (25).
In the classification of the Liver Cancer Study Group of Japan, early HCC is defined as well-differentiated HCC with an obscure tumour margin (26). Essentially, early HCC has no substantial destruction of the underlying hepatic structure and the presence of tumour cell invasion into the intratumoural portal tracts (stromal invasion) has been suggested to be useful in distinguishing it from high-grade DN (26). The lesions experienced are tiny ( 1.2 cm) and well differentiated. They carry a good prognosis. In a recent Japanese study, the 5-and 10-year survival rates of patients with early HCC were 85 and 61% respectively (27). The prognosis was significantly better than in patients with 'small advanced' HCC with expansive growth and capsule formation in the same study and early HCC is likely to be a favourable prognostic factor. Similar results were also obtained in another Japanese study on single HCC smaller than 2 cm in diameter (28). The 5-year overall survival and recurrence-free survival rates were 93 and 47% in the small HCC group, whereas they were 54 and 16%, respectively, in the overt HCC group (28). It is to be noted, however, that there is a discrepancy in the diagnostic criteria for well-differentiated HCC and high-grade DN. Many of the vaguely nodular, welldifferentiated HCC diagnosed by Japanese pathologists tend to be interpreted as high-grade DNs by Western pathologists (26,29).

Expression profiling of hepatocellular carcinoma
It is commonly believed that development and progression of cancers are accompanied by complex changes in the patterns of gene expression. cDNA micro-array has been used to compare the gene expression profiles between human HCCs and the non-tumourous liver tissues from the same patient or other normal livers. As expected, the gene expression patterns were found to vary significantly among the HCC and non-tumourous liver samples (30)(31)(32)(33)(34)(35)(36). Interestingly, 'proliferation cluster', which comprised of genes associated with cell proliferation and mitosis, was found to have increased expression in HCC samples. On the other hand, most genes that were expressed at lower levels in HCC than in nontumourous liver tissues belonged to 'liver-specific cluster', which comprised of genes specifically expressed in differentiated hepatocytes (30,33). These observations suggest that accelerated cell proliferation accomplished by the upregulation of proliferation and mitotic-associated genes are generally involved in hepatocarcinogenesis. However, downregulation of liver-specific genes is possibly associated with dedifferentiation of cancer cells during tumour progression.
Recently, cDNA micro-array has been successfully used to compare the gene expression profilers in different aetiological and clinicopathological features of HCCs. HBV and HCV are the most well-documented risk factors for HCC. Some recent studies compared the gene expression profiles between HBV-and HCV-infected HCC and demonstrated significant differences in gene expression profiles between HBV-and HCV-associated HCC (31,33,37). These findings add further support to the hypothesis that HBV-and HCV-associated HCCs may result from distinct mechanisms. Vascular invasion is a major factor affecting tumour metastasis and patients' prognosis. Identification of the genes involved in vascular invasion may have potential benefits to improve the diagnosis, treatment and patient management of HCC. Genes regulating extracellular matrix or cell motility, such as MMP14 and RhoC, were found to have increased expression in HCCs with the vascular invasion pheno-type (30,33). Matrix metalloproteinases (MMPs) play an important role in cancer cell invasion by degrading extracellular matrices (38). HCC cells constitutively expressing MMP14 mRNA can promote cells to invade through matrix gel in vitro and this MMP14-dependent invasion is increased in response to hepatocyte growth factor (HGF) and is blocked by MMP inhibitors (39). Clinicopathologically, expression of MMP14 mRNA shows a tendency to be associated with poorer differentiation in HCC and has a strong statistical association with a poor outcome of patient survival (40). Rho proteins are small GTPase, function in regulating the cytoskeletal re-organization and thereby cell morphology and motility. RhoC is playing an indispensable role in cancer metastasis (41,42). Overexpression of RhoC is common in human HCC and was associated with increase of tumour invasiveness and metastasis (43,44). In addition, high-RhoC expression also links to the shorter survival of HCC patients (43). These findings suggest that the breakdown of extracellular matrix and increase in cell motility may play crucial roles in cancer invasion and metastasis.

Chromosomal and genetic alterations in hepatocarcinogenesis
Carcinogenesis is a stepwise process of accumulation of chromosomal, genetic and epigenetic abnormalities that can lead to cellular dysfunction. Recent evidence suggests that chromosomal instability (CIN) emerges at an early stage and accumulates during hepatocarcinogenesis, resulting in the acquisition of malignant phenotype.
Loss of heterozygosity (LOH) is uncommon in cirrhotic livers. In a previous study analysing chromosomal gains and losses using CGH, the frequency and pattern of genetic alterations in DNs highly resembled those in HCCs (45). Gains of DNA were found to cluster in chromosome arms 1p, 1q, 7q, 15q, 16p, 17q and 20q and losses of DNA at 3p, 4q, 9p and 11q (45). Such frequency and pattern of genetic alterations are not seen in other hepatocellular nodules such as focal nodular hyperplasia and hepatocellular adenomas. This implies that DN may be a precancerous lesion. Altogether, the genetic results in previous and present studies show that there is a stepwise increase in the genetic abnormalities from cirrhosis through DN to HCC, giving support to the notion of multistep hepatocarcinogenesis. Major progress in the classification of liver nodules and understanding has been achieved through image analysis techniques and careful histological dissection of explanted native livers. In Liver International (2007) a study on explanted livers, which allow examining the hepatocellular nodules and confirming their nature, it was found that HCC nodules were significantly associated with the presence of high-grade DNs (46). Moreover, low-grade DNs and macroregenerative nodules did not show chromosomal imbalances of allelic losses on 8p and of gains of 1q, as in high-grade DN and HCC (47). Recently, genome-wide micro-array analysis has been employed to study the molecular expression profiles of the different stages in hepatocarcinogenesis from cirrhosis through dysplastic nodules to HCC (48)(49)(50). These have been useful tools in identifying a possible 'molecular signature' in distinguishing dysplastic nodules from HCC.
In HCC, recurrent chromosome alterations, including loss of 1p, 4q, 8p, 16q and 17p and gain of 1q, 8q and 20q have been revealed by CGH, allelotyping (51)(52)(53)(54)(55) and by other technology (56). These findings reflect a high degree of CIN in HCC (57), contributing to hepatocarcinogenesis. LOH has been widely used to define chromosomal aberrations in HCC. Comprehensive genome-wide allelotyping with a large number of highly informative microsatellite markers on multiple chromosomes is useful in detecting recurrent allelic losses of cancers (58). Coupled with detailed clinicopathological correlation and gene mutation analysis (e.g. p53 or b-catenin mutations), different genetic pathways can be delineated (59,60). In HCC, using comprehensive genome-wide allelotyping, it has shown that HCC can be divided into two groups, one having chromosome stability with b-catenin mutation and chromosome 8p losses, and the other demonstrating chromosomal instability, with frequent allelic losses on chromosomes 1p, 4q, 6q, 9p, 13q, 16p, 16q and 17p and p53 and Axin1 mutations (60). A recent study, using comprehensive genome-wide allelotyping with more than 400 markers and coupled with detailed clinicopathological correlation and p53 mutation, has shown that HCC can be stratified into low-stage and advanced-stage tumours with different survival prognostication (59). Apart from genome-wide allelotyping, chromosome-specific high-density allelotyping is particularly useful for detecting minimal deleted regions on the cancer genome and narrowing down the location of existing tumour suppressor genes to facilitate positional candidate cloning of tumour suppressor genes (61,62). For instance, chromosomes 1p, 8p and 13q are among the most frequently affected chromosome arms in HCC and other cancers. Previous reports in HCC have shown that allelic losses on these chromosomes range from 17.5 to 53%, and they may harbour putative tumour suppressor genes (61,63,64).
With CGH analysis, HCC has been shown to harbour multiple chromosomal abnormalities, predominant losses, with increased chromosomal instability. Recurrent aberrant gains have been found on 1q, 8q, 16p and 20q, and recurrent chromosomal losses on 1p, 4q, 8p, 13q, 16q and 17p (65)(66)(67)(68). Recently, arraybased CGH has been used to provide high-resolution mapping of chromosomal aberrations in HCC. Although the chromosomal abnormalities reported are similar to those obtained with CGH, coupled with correlation with gene expression data, it may be feasible to identify novel oncogenes and tumour suppressor genes (69).
Aberrations have also been found to differ in HCC with different aetiological backgrounds. Chromosomal aberrations were more frequent in HBV-related HCCs than in HCV-associated tumours (70). Another study has shown that HBV-associated HCCs had more frequent (40% on average) losses at 4q, 16q and 17p (including the p53 region) than in non-viral HCC samples, suggesting that these abnormalities are much associated with HBV infection (70). With regard to other chromosomes, a gain of 10q (7/41, 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 HBV-and HCV-associated HCCs (71).
Cirrhotic livers with high liver cell proliferative activity have a higher risk of developing HCC (72). Aberrant DNA methylation (hypermethylation) has also been observed in livers with chronic hepatitis or cirrhosis and is implicated as an early event in hepatocarcinogenesis (73). Another chromosomal abnormality that is seen in cirrhotic livers is telomere shortening. Telomere shortening limits the number of cell divisions of cells and may affect the regenerative capacity of organ systems during ageing and chronic disease. A previous study was performed on cirrhotic livers and compared with non-cirrhotic samples using quantitative fluorescence in situ hybridization (74). Telomere shortening was seen only in hepatocytes and hepatocyte telomere shortening correlated with progression of fibrosis in cirrhosis samples. It was suggested that fibrotic scarring at the cirrhosis stage is a consequence of hepatocyte telomere shortening and senescence. In another study on telomere shortening, it was found that there was a gradual shortening of telomere during hepatocarcinogenesis, and the telomere lengths inversely correlated with the mRNA level of genes involved in telomere maintenance, i.e. the telomeric repeat-binding factor 1 (TRF1), TRF2 and the TRF1-interacting nuclear protein 2 (TIN2) (75).
Recently, telomere shortening has been shown to play a role in liver DN. Telomere shortening and telomerase activity in high-grade DNs were found to be comparable to those of HCC (75). Moreover, there were significant differences between low-grade and high-grade DNs in that most low-grade DNs had similar levels of telomere shortening and telomerase activity to those of the chronic hepatitis and cirrhosis. In another study, Oh et al. (76), further investigated the expression of telomere-binding proteins and telomere lengths in cirrhosis, DN and HCC. They observed that the expression of the telomere-binding proteins (TRF1, TRF2, TIN2 mRNA and TRF1 protein) had a stepwise increase with regard to chronic hepatitis and cirrhosis, DN and HCC. There was a marked increase in high-grade DNs and DNs with HCC foci. The increase in such expression was even more significant in HCCs. Moreover, there was a gradual shortening of telomere with a significant reduction in length in DNs. Most (77.6%) of the DNs had shorter telomeres when compared with their adjacent chronic hepatitis or cirrhosis and the telomere lengths were inversely associated with the mRNA level of these genes. The results suggest that telomere shortening is an important mechanism in hepatocarcinogenesis. All these data support a stepwise molecular change in hepatocarcinogenesis. They also support the notion that DN is precancerous.

Epigenetic alterations in hepatocarcinogenesis
Recently, in addition to genetic alterations, many lines of evidence have indicated that epigenetic alterations also play a very important role in human carcinogenesis. The term 'epigenetic' generally refers to heritable changes in DNA methylation and histone modifications that stably modify gene transcription but do not involve changes of the DNA sequences (77). DNA methylation, a covalent addition of methyl group (-CH 3 ) to the 5-position of cytosine, is the most wellcharacterized epigenetic event. In mammalian cells, DNA methylation is mainly found in the cytosine residues of CpG dinucleotides within the repetitive elements or promoter-related CpG islands. DNA methylation is essential for development and differentiation, but research over the past decade has linked epigenetic alterations to human carcinogenesis (78).
Among all, the most emphasized epigenetic alteration in cancers is the aberrant hypermethylation of CpG islands on gene promoter regions. Hypermethy-lated promoters are almost always transcriptionally silent. Such hypermethylation is essential for normal development (79), X-chromosome inactivation (80) and imprinting (81). However, aberrant CpG island methylation results in inactivation of tumour suppressor genes. Retinoblastoma 1 (RB1) gene is the first classical tumour suppressor gene reported to be inactivated in human cancers by promoter hypermethylation (82). Despite the fact that RB1 promoter methylation is frequently detected in other solid tumours, there is no evidence to support that RB1 promoter methylation also contributes to hepatocarcinogenesis (83). However, it is found that another tumour suppressor gene in this pathway, p16/INK4A, is frequently inactivated in HCC by promoter methylation (66,84). p16 [also known as cyclin-dependent kinase inhibitor 2A (CDKN2A)], a cyclin-dependent kinase inhibitor, regulates the phosphorylation status of the RB1 gene product. Loss of p16 function may limit the tumour suppressor function of the wild-type RB1 gene product and lead to unregulated cellular proliferation (85). We have previously reported that deleted in liver cancer 1 (DLC1), a negative regulator of Rho family GTPases, was frequently silenced in human HCCs via DNA hypermethylation (86). Ectopic overexpression of DLC1 in hypermethylated HCC cells suppressed Rho-mediated cytoskeletal re-arrangement and cell motility and, in turn, substantially abolished their invasiveness (87). In addition, DNA hypermethylation has also been found in other tumuor suppressor genes that regulate various cellular pathways in human HCC and well-characterized examples include E-cadherin (88), RAS-association domain family (RASSF1A) (89), Glutathione S-transferase pi-1 (GTSP1) (90), suppressor of cytokine signalling (SOSC-1) (91), soluble frizzled related protein-1 (SFRP1) (92) and phosphate and tensin homolog (PTEN) (93). Recently, we and others have conducted genome-wide screening for DNA methylation-silenced genes that are related to hepatocarcinogenesis and some hypermethylated tumour suppressor genes have been identified and characterized (94)(95)(96). With advancement in micro-array technology, the number of hypermethylated genes in human HCC will further increase and may lead to a better understanding in hepatocarcinogenesis.
Similar to chromosome abnormalities discussed above, aberrant DNA methylation in human HCC also exhibits an increasing trend in the multisteps of hepatocarcinogenesis (Fig. 2). Several independent studies have revealed that DNA methylation in tumour suppressor genes is an early event and can be found in non-cancerous liver tissues (97)(98)(99) (100). These data suggest that p16 hypermethylation occurs in the early stages (cirrhotic nodules and DNs) and may predispose to HCC. Similarly, two recent studies have shown a stepwise increase of aberrant DNA hypermethylation of a panel of well-characterized tumour suppressor genes, from normal liver, chronic hepatitis, cirrhosis, dysplastic nodule to HCC (101,102). All these findings suggest that aberrant hypermethylation of tumour suppressor genes is an early event and may accumulate leading to the development of HCC.
Cancer cells often simultaneously exhibit global DNA hypomethylation and gene-specific DNA hypermethylation. In mammalian cells, DNA methylation is established and maintained by the co-operative functions of de novo DNA methyltransferases (DNMTs), DNMT3A and DNMT3B and 'maintenance' DNA methyltransferase, DNMT1 (103,104). Aberrant expression of DNMTs can give rise to methylation errors or to de novo methylation events at normally unmethy-lated CpG sites, which will then be copied by DNMT1 after cell division (105). In vitro studies have demonstrated that overexpression of exogenous DNMT1 can induce DNA hypermethylation and result in transformation phenotype in NIH3T3 cells (106,107). Consistently, inactivation of DNMT1 by RNA interference (RNAi) or anti-sense oligonucleotide results in demethylation and re-expression of silenced tumour suppressor gene in cancer cells (108,109). Several independent studies have revealed significant overexpression of DNMT1, DNMT3A and DNMT3B in human HCC (110,111). The dysregulation of DNMT proteins significantly correlated with advanced tumour stages and poorer patients' survival rates (110). Although the exact mechanisms of DNMTs dysregulation leading to DNA methylation abnormalities in cancer genome remain elusive, these data have provided support to the hypothesis that increased DNA methyltransferase activity is implicated in human carcinogenesis and hepatocarcinogenesis.
Studies are now beginning to move towards investigating the initiating factors for the frequent DNA methylation abnormalities in HCC. Several recent studies have uncovered the role of viral oncoproteins in cancer epigenetic dysregulation, in that viral oncoproteins (e.g. the simian virus 40T antigen and the latent membrane protein 1 of Epstein-Barr virus) are capable of activating DNMT1 and DNMT3B in human cancer cell lines (112)(113)(114)(115). In HCC, reported data have shown that DNA hypermethylation on GSTP1 and E-cadherin promoters are more frequently found in HCC patients with HBV infection (116). Very recently, Park et al. (117) have reported that the HBV x protein (HBx) might contribute to the aberrant DNA methylation during hepatocarcinogenesis through regulating the expression of DNMTs. Evidence accumulated over these years also suggests that nutrition plays an important role in epigenetic alterations. It has been reported that methyl-deficient diet was sufficient to induce hepatocarcinogenesis in a rodent model. Deficiency of the major dietary sources of methyl groups, including methionine, choline, folic acid and vitamin B-12, could result in depletion of intracellular methyl group pools and increased genome-wide DNA hypomethylation (118,119). Recently studies have further shown that DNA damage can alter the DNA methylation pattern. Oxidation of 5-methylcytosine to 5-hydroxymethylcytosine prevents DNMT1 methylation of cytosine in the opposite strand and would result in loss of methylation at that CpG site. On the other hand, inflammation-mediated damage products, 5-chlorocytosine and 5-bromocytosine, can mimic 5-methylcytosine in directing DNMT1 activity, resulting in aberrant de novo DNA methylation of previously unmethylated sites (120).

Mutational analysis in hepatocellular carcinoma
In human cancers, mutations have been found in a large number of genes, including p53, adenomatous polyposis coli (APC), breast cancer 1 (BRAC1), breast cancer 2 (BRAC2), Rb, Ras and b-catenin, which are involved in the regulation of cell proliferation, cellcycle progression, apoptosis and metastasis. When compared with other cancers such as colon and breast cancers, gene mutations in HCC are not as frequent.
Of these genes, p53 and b-catenin probably are the most frequently mutated genes in HCC. Interestingly, recent data suggesting HCC harbouring p53 and bcatenin mutation may arise from two independent pathways, in one of which p53 mutation is associated with a high level of chromosomal instability whereas in the other, b-catenin mutation is associated with frequently epigenetic alterations (121). We will briefly discuss p53 in this section and mutations of b-catenin and other genes will be discussed later in the next section. p53 gene is the most well characterized tumour suppressor gene known to be mutated at very high frequencies in tumours of different cellular origins (122). The frequency of p53 mutation in HCC ranged from 13 to 33% in Asian populations (123)(124)(125). Mutant p53 has a much longer half-life than the wildtype protein. Hence, p53 mutation is closely associated with overexpression of protein. Pathologically, p53 overexpression was more frequently seen in tumours with poor cellular differentiation and of larger tumour size (126), suggestive of a late event in development. In general, mutations of p53 cluster in exons 5-9 in a random fashion. However, in geographical regions where aflatoxin is prevalent, a specific hot-spot mutation at codon 249 of exon 7 of the gene was found. Previous evidence has revealed that 249 mutant promotes cell proliferation and inhibits the wild-type p53-mediated apoptosis and, as a result, confers growth advantage to cancer cells (127)(128)(129). Another line of evidence linking p53 to human HCC arose from the finding that HBx bound directly to p53 (130,131). HBx inhibits p53 transcriptional activity and represses p53-mediated apoptosis (129,130,132). Indeed, p53 has a critical role in controlling apoptosis and loss of p53 function often coincides with overexpression of oncogenes and underexpression other tumour suppressor genes. Recently, mechanistic links between p53 and Kruppel-like factor 6 (KLF6), Growth arrest and DNA damage inducible protein (GADD45), Inhibitor of growth 1b (ING1b) and stathmin have been reported in HCC (133)(134)(135)(136). In addition, the simultaneous loss of p53 and p14/ARF has been implicated in HCC metastasis (137). Identification of the collaborative effects between dysregulation of various cellular pathways and loss of p53 function unquestionably is a major focus in cancer biology.

Molecular cellular pathways in hepatocarcinogenesis
As a consequence of the activation of cellular oncogenes or the inactivation of tumour suppressor genes, deregulation of various signalling pathways has been reported in HCC subsets, such as Wnt/b-catenin, Ras, p14ARF/p53, p16INK4A/Rb, transforming growth factor-b (TGF-b) and PTEN/Akt pathways (5). In the following part, we briefly summarize two of the most common molecular cellular pathways, Wnt/b-catenin and Ras, in human HCC.

Wnt signalling pathway
Wnt/b-catenin signalling pathways had a critical role in the control of cellular proliferation, motility, morphology and embryonic development (138). Activation of Wnt/b-catenin is also commonly associated with the development of HCC and other human cancers (139). b-catenin is the central player in the Wnt signalling pathway (Fig. 3). In the absence of Wnt signals, b-catenin is recruited to the degradation complex consisting of APC, Axin and glycogen synthase kinase (GSK)-3b. When bound to this complex, b-catenin is phosphorylated by GSK-3b at the serine and threonine residues (i.e. S33, S37, T41 and S45 residues). Hyperphosphorylated b-catenin is then targeted for degradation through the ubiquitin-proteasome system. This continuous degradation maintains the free cytoplasmic b-catenin at a low level in the normal steady state. When Wnt ligands bind to the transmembrane co-receptors, frizzled and low density lipoprotein (LDL) receptor-related protein 5/6 (LRP5/ 6), Wnt/b-catenin signalling transduction is triggered. The activation signal is then transmitted to the cytoplasm through phosphorylation of dishevelled (Dvl). Activated Dvl binds to and suppresses the kinase activity of GSK-3b, thereby stabilizing b-catenin. Accumulation of the free, hypophosphorylated bcatenin in the cytoplasmic pool facilitates its nuclear translocation, where b-catenin forms a complex with members of T-cell factor (Tcf)/lymphoid enhancing factor (LEF) transcription factors and upregulates the expression of target genes including the proto-oncogene c-Myc (140) and cyclin D1 (141,142).
Activation of the Wnt/b-catenin signalling pathway is mainly attributed to the mutations of the APC gene in colon cancer (143). However, in human HCC, APC mutations are rare; instead, promoter methylation plays a major role in APC inactivation (102,144). Mutations of b-catenin and Axin1 have also been frequently reported (145)(146)(147)(148)(149)(150). Almost all b-catenin mutations have been found within exon 3 of the gene, predominantly at codon 32-37, 41 and 45. These mutations protect the protein from degradation and hence stabilize the protein (145,146,(149)(150)(151). We and others have shown that b-catenin gene mutation and nuclear accumulation of the protein in HCCs ranged from 13 to 34% and from 11 to 43% respectively (145-147, 149, 150). HCV-associated HCCs tend to have higher frequencies of both b-catenin mutations and nuclear accumulation than HBV-associated ones. In addition, mutation of Axin1 has been found in approximately 5-10% of human HCCs. Axin1 mutations including point mutations and small deletion are more frequently observed at the N-terminal half of the protein, which might stabilize bcatenin by impeding the formation of the APC/GSK-3b/b-catenin complex (148,149). Besides Axin and bcatenin mutations, inappropriate activation of the Wnt/b-catenin signalling pathway can be brought about when the upstream mediators are dysregulated. Overexpression of the Wnt ligand and the Frizzled receptor has been demonstrated in human HCC (152,153). On the other hand, the Wnt antagonist, sFRP1 has been found to be epigenetically silenced in human cancers including HCC (92,154). Likewise, a recent study has shown that Dvl-1 and Dvl-3 are overexpressed in human HCC (155). Overexpression of Dvl is associated with b-catenin accumulation and Wnt/bcatenin signalling activation (156,157). In addition, we have recently demonstrated that two Dvl inhibitors, HDPR1 and Prickle-1, are also frequently underexpressed in human HCC (155,158). PIN1 is another Wnt signalling regulator found to be dysregulated in HCC. Overexpression of PIN1 stabilizes b-catenin by inhibiting its interaction with APC (159). Interestingly, overexpression of PIN1 and mutation of bcatenin appear to be mutually exclusive events in Wnt signalling activation in HCC (160). Apart from the above-mentioned genes, other members of the Wnt/bcatenin signalling pathways, such as LPR5/6 and their antagonists Dickkopf (DKK)1/3, are also subjects of investigation (161). It would not be surprising that more Wnt/b-catenin regulators would be identified and implicated in carcinogenesis (162).

Ras signalling pathway
Ras proteins, consisting of N-Ras, H-Ras and K-Ras, are the prototype of 21 kDa GTPases and perhaps the most well-characterized proto-oncogenes in human cancers. Ras signalling is initiated when a ligand, for example, epidermal growth factor (EGF), binds to its receptor on the cell membrane and induces phosphorylation of tyrosine residues on the cytoplasmic domain of the receptor (Fig. 4). This phosphorylation produces binding sites for proteins with src homology 2 (SH2) domains, such as growth factor receptor-bound protein 2 (GRB2). The Ras protein is then recruited to the activated receptor and converted to its active GTP-bound form by Son of sevenless protein (SOS), or other Ras guanine nucleotide exchange factors (RasGEF). Activation of Ras facilitates the binding of RAF and transmits the signalling cascade from the cell membrane to cytoplasm through phosphorylation of methyl ethyl ketone (MEK) and its downstream effectors extracellular signal-regulated kinase (ERK)1/2. Subsequently, hyperphosphorylated ERK1/2 translocate into nuclear and regulate the expression of target genes. In addition to the RAF/MEK/ERK pathway, activated Ras can also stimulate many other effector proteins including p120 GAP, RalGDS, phosphatidylinositol 3-kinase (PI3K), protein kinase C (PKC). Indeed, Ras signalling is known to regulate diverse cellular functions including cell growth, survival and migration (163). Previous in vitro studies have clearly demonstrated that the ectopic expression of Ras could induce transformation in immortalized woodchuck hepatic cell line and enhance metastatic phenotype in human HCC cell lines (164,165). Unlike other solid tumours, mutations of Ras genes in HCC are rare. However, overexpression of Ras proteins has been found in human HCC and cirrhotic livers (166). Conversely, downregulation of the physiological inhibitors of Ras/Raf/MEK/ERK pathways has been frequently observed in human HCC. For instance, Raf-1 kinase inhibitory protein (RKIP) is downregulated in human HCC. Downregulation of RKIP leads to increased ERK1/2 activity. In contrast, forced expression of RKIP suppresses the Raf kinase pathway and reduces the activity of ERK1/2 in HCC cells (167,168). Similarly, Spred-1, another Ras/Raf/MEK/ERK pathway inhibitor, was found to be frequently underexpressed in human HCCs and its expression level inversely correlated with cancer metastasis (169). It is important to note that, although activation of Ras signalling is usually associated with cellular transformation, studies have also indicated that Ras signalling may induce cell senescence and apoptosis, particularly in untransformed cells (170,171). The pro-apoptotic function of Ras is dependent on its downstream effector RAS-association domain family (RASSF1) proteins. RASSF1 directly interacts with active Ras, which then consequentially activates mammalian sterile 20-like kinase-1 (MST1) and induces apoptosis (172). In this regard, RASSF1A may function as a gatekeeper against the oncogenic activity of Ras. In human HCC as well as other solid tumours, RASSF1A is frequently epigenetically silenced (89,173,174). Loss of RASSF1 expression in HCC may thereby shift the balance of RAS activities towards a growth-promoting effect.

Conclusion
In summary, hepatocarcinogenesis is closely associated with accumulation of chromosomal, genetic and epigenetic alterations. Some of these alterations occur at different stages of the hepatocarcinogenesis and result in deregulation of important molecular cellular pathways. Some show a stepwise increase in disease progression. A detailed understanding of the underlying molecular mechanisms involved in the progression of HCC is of fundamental importance and helps in the development of effective prevention and treatment regimes for HCC.