In healthy adult livers, hepatocytes are quiescent cells, being renewed slowly, approximately once a year. However, the liver has an extremely powerful regenerative capacity, as demonstrated experimentally in rodents, and as observed in patients with chronic liver diseases [26]. This regenerative capacity is due mostly to the ability of mature hepatocytes to proliferate in response to a diminution of the total liver mass either experimentally, or following exposure to viral and nonviral hepatotoxic agents. In addition, the adult liver seems to harbor hepatocyte-progenitor cells (< 0.1% of total hepatocyte mass) that are able to restore liver hepatocyte populations [27]. However, hepatocytes do not have unlimited replicative capacity, due to the lack of telomerase activity that is needed to avoid telomere shortening during successive cell divisions. This is best exemplified by decreased hepatocyte proliferation in the cirrhosis stage of chronic liver diseases, providing in vivo evidence for the exhaustion of hepatocyte proliferation capacity [28].

Senescence mechanisms in hepatocytes and in liver tissue are not well known. However, a limited number of in vitro studies with hepatocytes, as well as numerous descriptive in vivo studies in liver tissue provide sufficient evidence that hepatocytes can undergo senescence type changes. Limited proliferative capacity of somatic cells is controlled by replicative senescence. By contrast to primary hepatocytes which do not proliferate in culture, fetal hepatocytes display better proliferation capacity and can enter replicative senescence [29]. This is accompanied by progressive shortening of telomeres in a context of telomerase-free activity. In contrast to in vitro studies, in vivo senescence of human hepatocytes is better known. Replicative senescence displays a gradual increase from 10% in normal liver, to more than 80% in cirrhosis, being detected in 60% HCCs [30]. Telomere shortening during aging is slow and stabilizes at mid age in healthy liver, so that the loss of telomeric DNA does not reach a level to induce telomere dysfunction and DNA damage response (DDR). On the other hand, telomere loss is accelerated in chronic liver disease to reach lowest levels in the cirrhotic liver. Therefore, one plausible mechanism involved in cirrhosis is probably telomere-dependent senescence, the so-called replicative senescence [31].

Hepatocyte senescence that is observed in severe chronic liver diseases may also be induced by telomere-independent mechanisms, such as ROS-induced senescence (RIS) and oncogene-induced senescence (OIS). RIS and OIS are rare events under normal physiological conditions, but could more commonly occur in context of ROS overproduction and/or oncogenic signals activation, both leading to DDR and activation of the ATM/Chk/p53 pathway and, by alternative mechanisms, the p16^INK4a/pRb pathway. OIS is a DNA damage response triggered by DNA hyper-replication [32], and several oncogenic pathways – i.e. activated Ras, c-myc or Wnt/β-catenin - have been shown as potentially involved in OIS in mammary cells or fibroblasts, but nothing is known about liver cell types [33-34]. As witness of DDR occurrence in chronic liver diseases, upregulation of DNA repair enzymes which may reflect increased DNA damages, has been reported in cirrhosis [35]. ROS production can be favored by the context of chronic liver injury with inflammation, cell death, oxidative stress as well as some of the etiological factors such as HCV and alcohol induce mitochondrial dysfunction [36]. Later on, we will describe the main oncogenic pathways found as activated in HCC and that might lead to OIS although this phenomenon of senescence-induction/senescence-escape has not been clearly demonstrated so far in liver cells during hepatocarcinogenesis.

The abrogation of DNA damage checkpoints could represent a selective advantage allowing clonal expansion of genetically altered hepatocytes at the stage of senescence (Figure 3). As stated earlier, the p16^INK4a/pRb and p53/p21^Cip1 pathways play crucial roles in senescence arrest as observed in different in vitro and in vivo models [37]. Thus their invalidation could help the initiated cells to escape the senescent process and go ahead to genetic instability and cancerous transformation. Strikingly, etiologic factors per se of HCC, such the viral HBV X protein, have been shown to help cells to bypass OIS [38].

The p53 pathway is a major tumor-suppressor pathway that: i) limits cell survival and proliferation (replicative senescence) in response to telomere shortening; ii) induces cell-cycle arrest in response to oncogene activation (OIS); and iii) protects genome integrity. The p53 pathway is affected at multiple levels in human HCC as follows: i) p53 mutations occur in aflatoxin-induced HCC (50%) and with lower frequency (20%-30%) in HCC not associated with aflatoxin; ii) microdeletions of p14^ARF occur in 15%-20% of human HCC but rarely in p53 mutant HCC; iii) increased Mdm2 expression has been observed in human HCC; iv) the vast majority of human HCC overexpresses gankyrin, which inhibits both the pRb and p53-checkpoint functions [39]. It seems likely that loss of p53 checkpoint function at the cirrhosis stage would lead to an expansion of hepatocytes with dysfunctional telomeres, chromosomal instability (CIN), and initiation of HCC. The p21^Cip1 gene, downstream target of p53, is accumulated in cirrhosis by comparison to normal liver tissues [40]. Although p21^Cip1 gene has not been found mutated in HCCs, its promoter is highly methylated in HCCs as compared to cirrhosis, showing that p21^Cip1 is repressed in HCCs [41]. In addition to its effect on p53 checkpoint function, deletion of p14^ARF down-regulates expression of the p27 tumor suppressor [42].

The p16/pRb checkpoint is another major pathway limiting cell proliferation in response to telomere shortening, DNA damage, and oncogene activation. In human HCC, retinoblastoma pathway alterations (p16^INK4a, p15^INK4b or RB1 genes) – i.e. mutation/deletion of RB1 and/or methylation of p16^INK4a, p15^INK4b promoters - are observed in more than 80% of cases, with repression of p16^INK4a by promoter methylation being the most frequent alteration [43], as well as gankyrin expression [39], indicating that the pRb checkpoint is dysfunctional in the vast majority of human HCCs. It is conceivable that an impairment of the pRb checkpoint would allow an expansion of hepatocytes with dysfunctional telomeres at the cirrhosis stage. In agreement with this assumption, p16^INK4a accumulates in cirrhosis by comparison to normal liver tissues, but p16^INK4a expression is diminished in premalignant liver tumors (small cell changes) and HCC [40].

Activation of telomerase occurs during the transition from premalignant lesions to HCC. More than 90% of human HCC show an activation of telomerase, which is the rate-limiting step for initiation of cell immortalization [44]. Experimental data from mice have shown that telomere shortening increases the initiation of liver tumors but telomerase deficiency limits the progression of early lesions toward macroscopic tumors [45]. A current hypothesis indicates that telomerase activation is necessary to limit telomere dysfunction and to prevent the accumulation of excessive levels of chromosomal instability and DNA damage that would impair tumor growth independent of p53 checkpoint function [46-47]. The molecular mechanisms involved in TERT suppression in somatic cells and its reactivation in cancer cells are poorly known. The integration of HBV DNA sequences into TERT gene provides evidence for a virus-induced deregulation of TERT expression, but this appears to rarely occur [48]. In addition, the viral X and PreS2 HBV proteins as well as the core HCV protein may upregulate telomerase activity [49]. Another hypothesis would be that HCC arise from stem/progenitor cell-like root cells that may already express TERT at sufficient levels to maintain telomere integrity. Experimentally, telomerase deletion limits the progression of p53-mutant HCCs with short telomeres [47]. These observations suggest that the aberrations affecting telomerase activity and senescence controlling genes such as p53 may cooperate during hepatocarcinogenesis.

In summary, HCC is characterized by mutational inactivation of p53, a major player in DNA damage-induced senescence. In addition, p15^INK4b, p16^INK4a, p21^Cip1 CDKIs are often inactivated in this cancer mostly by epigenetic mechanisms involving promoter methylation. These changes may play a critical role in the bypass of senescence that is observed in most cirrhosis cases, allowing some initiated cells to escape senescence control and proliferate. Furthermore, it has been shown that Twist-1 and Twist-2 proteins can switch pre-senescent mammary cells towards epithelio-mesenchymal transition (EMT) and senescence-bypassing rather than towards senescence [34,50-52]. Both proteins override oncogene-induced premature senescence by abrogating the p53 and pRb pathways, and this phenomenon remains to be confirmed in hepatocarcinogenesis. In the absence of telomerase activity such cells would probably not survive due to telomere loss. However, since more than 80% of HCCs display telomerase activity, it is highly likely that the telomerase reactivation, together with the inactivation of major CDKIs, plays a critical role in HCC development by conferring premalignant or malignant cells the ability to proliferate indefinitely. However, cellular immortality is not sufficient for full malignancy. Thus, senescence-related aberrations that are observed in HCC cells, may confer a partial survival advantage that would need to be complemented by other genetic or epigenetic alterations to reach the cancerous phenotype.

Proto-oncogenes of the ras family (H-ras, K-ras et N-ras) play a key role in transduction of the mitogenic signal linked to mitogen-activated protein kinases (MAPK). In humans, they can be activated by point mutation at codons 12, 13 and 61. In hepatocarcinogenesis, these mutations are uncommon and, when present, found at codon 12 for K-ras and H-ras, and at codon 61 for K-ras and N-ras [54]. In contrast, the p21-Ras protein is frequently expressed in cirrhosis and HCCs [55]. In vitro, ectopic expression of mutated ras is a strong oncogenic stress factor and leads frequently to cancerous transformation when the p53/p21^Cip1 and p16^INK4a/pRb checkpoints are inhibited and telomerase reactivated in mammary cell models [33]. However, few are known concerning liver cells.

The c-myc proto-oncogene stimulates a pattern of cellular gene expression by its regulatory elements and is involved in gene expression during cell growth and differentiation [56]. It has been shown to be important and acts as a driver oncogene in the process of experimental hepatocarcinogenesis since it is able to initiate and promote hepatocarcinogenesis in transgenic mice by giving rise sequentially to liver dysplasia and HCC [5,57-59]. However, its role in human hepatocarcinogenesis remains unclear. Previous studies have shown that c-myc is barely expressed in normal liver tissues [60]. By contrast, c-myc has been found overexpressed in most of the human hepatoma cell lines, and its repression by ribozyme or antisense therapy induces differentiation and growth inhibition of these cells [61-62]. Also, in vivo studies have shown that c-myc expression may gradually increase from normal liver to chronic hepatitis, cirrhosis and HCC [63].

Mechanisms of c-myc overexpression during human hepatocarcinogenesis are poorly understood, but might be related to amplification of the gene (40%-60% of HCCs, but rare in liver dysplasia) or hypomethylation of its regulatory sequences [64-65]. Speculation of c-myc overexpression as being an early event in the premalignant steps of human hepatocarcinogenesis may be of major importance since it has been found in more than 50% of human HCCs and their corresponding adjacent nontumor liver. Recent studies have shown that activation of c-myc is strongly associated with the malignant conversion of preneoplastic high grade dysplastic liver nodules in HCC [66]. A cytogenetic tumor progression model constructed by Poon and colleagues have determined that gains of 8q22-24 (bearing the c-myc allele) are among the earliest genomic events associated with HCC development [67]. Furthermore, it has experimentally been shown that c-myc is able to induce oncogenic stress in mouse embryonic fibroblasts [34], although it remains to be determined in normal human liver cells. Several observations have initially let hypothesize that, although c-myc triggering cellular growth may play a role in the premalignant steps during the process of malignant transformation (as reflected by its high mRNA steady state level in peritumor liver), it might not play a significant role in sustaining the growth of the tumor cells (as reflected by its lower mRNA steady state level in the corresponding tumor tissue) [68-69]. However, more recent data strengthen the role of c-MYC in maintenance of the cancerous phenotype since, although c-myc gene mRNA steady state level might not change or decrease in HCCs by comparison to the matched adjacent preneoplastic tissues, c-MYC activity can be enhanced by posttranscriptional modifications affecting the half-life of the protein. Phosphorylation of the Thr-58 residue in the Myc box I domain targets MYC for ubiquitination and consequent proteasosome-mediated degradation [70]. Although mutations of this region are frequently observed in Burkitt lymphomas and lead to stabilization of MYC, as well as to disruption of its pro-apoptotic function, these mutations have not been found in human HCC. A member of the COP9 signalosome – i.e. CSN5 - located at 8q13 locus, regulates activity of the ubiquitin ligase complex. It has recently been shown that CSN5 overexpression occurs at the early stages of hepatocarcinogenesis and shows a significant association with the presence of the c-MYC-regulated expression signature. These results are consistent with the notion that CSN5 plays an important role in liver cancer progression by a mechanism involving stabilization of the c-MYC protein and enhancement of its activity [66]. Another mechanism of increased c-MYC-activity could be its stabilizing interaction with HIF (hypoxia inducible factors) [71]. Finally, that transcriptional activity of c-MYC itself could be regulated by multiple pathways, including RAS/RAF/MAPK, JAK/STAT, and Wnt/β-catenin signaling, which may result in a significant overlap between the c-MYC and other oncogenic pathways [72-75]. Thus it appears that c-MYC could be a central regulator of malignant transformation in early hepatocarcinogenesis.

The Wnt/β-catenin pathway is regulated tightly during early liver development [76]. The β-catenin molecule is an important multifactorial protein which is involved in cell-cell adhesion by strengthening the linkage between cadherin and α-catenin to the actin cytoskeleton. Its soluble form can translocate from cytosol to nucleus where it transactivates genes involved in cell fate during physiological homeostasis as well as for the setting of cancer properties. The Wnt/Frizzled signaling network controls activation of the canonical Wnt/β-catenin signalling cascade and/or the noncanonical c-Jun N-terminal kinase (JNK) and protein kinase C (PKC) [77].

Although evidence is accumulating that alterations of the Wnt/β-catenin pathway, due to or unrelated to β-catenin gene mutation, is a common event in hepatocarcinogenesis [78-84], little is known concerning the noncanonical elements. The previous finding of inappropriate activation of the Wnt/β-catenin pathway in hepatocarcinogenesis resulting from CTNNB1 gene mutations has provided clues toward understanding this process [78]. Different cohort studies of human HCC tissues have shown that activating mutations hitting the Wnt/β-catenin pathways have been observed within the heterogeneous “end product” tumor bulk : the CTNNB1 gene encoding for the β-catenin protein in 10% to 30% of HCCs, the AXIN-1 gene in 7% to 9%, whereas APC gene mutations are exceptional [85-87]. However, the meaning of these mutations is questionable due to their absence in liver dysplasia, cirrhotic nodules or chronic hepatitis tissues, and due to their heterogeneity in HCC bulks since they concern only a fraction of tumor cells. Additionally, it appears that aberrant activation of the Wnt/β-catenin pathway as manifested by cellular and nuclear accumulation of the β-catenin protein occurs in a higher percentage of HCCs - i.e. 35% to 85% - while the Wnt/β-catenin pathway gene mutations are absent [81-83]. More recently, it has been shown that, in absence of Wnt/β-catenin pathway gene mutation, activation of the Wnt/β-catenin pathway can occur by enhancement of the Wnt/Frizzled-mediated signalling. Indeed, it has been shown that binding of the WNT3 ligand on the FZD7 receptor can activate the canonical Wnt/β-catenin pathway in human and rodent HCCs, enhancing the cancerous phenotype of cancerous human hepatoma cell lines. Interestingly, WNT3 and/or FZD7 are overexpressed not only in 60% to 90% of human HCCs but also in 35-60% of the surrounding preneoplastic liver tissues, letting hypothesize that activation of the WNT3/FZD7-mediated signalling might be an early event in hepatocarcinogenesis [79,80,88-90]. Additionally, down-regulation of the Wnt/Frizzled-mediated signaling per se has recently been shown as being an early key signal for senescence of primary human cells, suggesting at opposite that activation of Wnt/Frizzled-mediated signaling might force the presenescent cell to bypass the senescence program [91]. Thus it remains to be confirmed that activation of the Wnt/Frizzled-mediated signaling might be involved in the early steps of hepatocarcinogeneis through induction of cellular oncogenic stress and allowing to bypass senescence.