Hepatocellular carcinoma (HCC), is the fifth most frequent human cancer, with the highest frequency in sub-Saharan Africa and far eastern Asia, where hepatitis B virus and hepatitis C virus infections are endemic and food is contaminated by Aflatoxin B1.1 HCC incidence is rising in Europe and United States, presumably due to increased incidence of hepatitis C virus infection, cirrhosis related to Type II diabetes, and alcoholic hepatitis.2 HCC is rapidly fatal, since most patients at risk are not diagnosed in time and amenable to potentially curative treatments, i.e., partial liver resection or transplantation.1, 2
Hepatocarcinogenesis is characterized by the accumulation of various alterations in oncogenes and oncosuppressor genes.3 This event is presumably due to the increased tendency of initiated cells to acquire mutations following dysregulation of the mechanisms preserving genome integrity, a condition known as genomic instability4 (GI). Indeed, a number of studies have demonstrated the progressive appearance of GI, such as point mutations, loss of heterozygosity and chromosomal alterations from preneoplastic liver to HCC.5 In addition, microsatellite instability occurs in HCC, although large variations have been reported in different studies.5
A body of evidence indicates that labile methyl groups play a pivotal role in liver carcinogenesis. Depletion of methyl groups in rats fed a lipotrope-deficient diet causes cancer or enhances liver tumor development induced by various carcinogens.6, 7 Marked decrease in S-adenosyl-L-methionine (SAM) and SAM:S-adenosylhomocysteine (SAH) ratio, associated with global DNA hypomethylation and overexpression of c-jun, c-myc and c-H-ras, was detected in preneoplastic and neoplastic liver from rats fed this diet.7 Similar changes also occur during chemically-induced hepatocarcinogenesis, in rats fed diets with adequate content of labile methyl donors,8, 9, 10 as well as in human cirrhosis and HCC.11 Taken together, these data suggest that alterations in methionine metabolism and DNA methylation are common and essential features of liver tumor development across species.
Decrease in SAM content in preneoplastic and neoplastic liver may depend on changes in methionine adenosyltransferase (MAT) isozyme pattern. In mammals, MAT1A gene, expressed only in the liver, encodes MATI/III isozymes, whereas MAT2A encodes MATII isozyme, which is widely expressed.12 Fall in MAT1A expression and MATI/III activity with concomitant upregulation of MAT2A occurs in hepatoma cell lines and rodent HCC as well as in human liver cirrhosis and HCC.13, 14 MATII isoform is inhibited by its reaction product and its upregulation does not lead to increase in SAM liver content.12 Differential expression of MAT1A and Mat2A genes influences DNA methylation and growth of human HCC.12, 15 DNA hypomethylation may generate GI during carcinogenesis.16, 17, 18, 19, 20, 21, 22 However, whether deregulation of methionine/SAM metabolism and DNA methylation are prerequisite for HCC development, and the prognostic role of these changes, remain open questions. Interestingly, increase in SAM: SAH ratio, DNA methylation, growth inhibition and prevention of HCC development were observed in preneoplastic and neoplastic liver of rats subjected to SAM administration.8, 9 SAM effects are reversed by 5-azacytidine, a hypomethylating agent and inhibitor of DNA methyltransferases.23In vitro growth of human HCC is inhibited by either the transfection of MAT1A gene or SAM addition to the culture medium.16 Both these treatments result in increased SAM level in tumor cells. Thus, the study of the relationships between the deregulation of methionine/SAM metabolism, changes in DNA methylation, GI and HCC development could be of primary importance to improve our knowledge on HCC pathogenesis as well as to identify new diagnostic and prognostic markers and therapeutic targets.
Two distinct categories of human HCC have been hypothesized. Central features of the first category are better prognosis, activation of the canonical Wnt/β-catenin pathway and low GI, whereas poorer prognosis and extensive chromosomal instability characterize the second category.24 Intriguingly, c-Myc and c-Myc/Tgf-α transgenic mouse models recapitulate the main pathogenetic mechanisms of human HCC, with c-Myc tumors resembling human HCC characterized by activated β-catenin and better prognosis, and c-Myc/Tgf-α HCC those characterized by shorter survival.25 Thus, c-Myc and c-Myc/Tgf-α transgenic mouse models as well as human HCC with different prognosis, as defined by patient's survival length, represent valid systems to investigate the pathogenetic role of alterations in methionine metabolism and DNA methylation, and their correlation with GI and HCC prognosis. In the present paper, we addressed these issues and demonstrated the influence of decrease in MAT1A levels with concomitant MAT2A upregulation as well as reduced SAM:SAH ratio on generation of DNA hypomethylation and GI in c-Myc transgenic mice and human preneoplastic liver. These parameters progressively increased in HCC, reaching the highest values in liver tumors associated with patients' shortest survival length. Furthermore, a small human HCC subgroup was identified in which, as detected in c-Myc/Tgf-α double transgenic mice, HCC develop in the presence of normal SAM, Mat1A and global DNA methylation levels, indicating that global DNA hypomethylation is not always required for HCC development.
Material and methods
Generation of Alb/c-myc (c-Myc) in CBA/JxC56BL/6J, and Alb/c-myc/MT/TGF-α (c-Myc/TGF-α) transgenic mice in (CBA/JxC56BL/6J) × CD1 genetic background has been previously described.26 Intact animals of both strains were used as normal controls. In the absence of interstrain differences for all parameters studied in normal liver, the results found in both strains are presented together. Animal study protocols were conducted according to the National Institutes of Health guidelines for animal care. Normal, dysplastic (16 weeks) and neoplastic (32 and 68–70 weeks for c-Myc/Tgf-α and c-Myc transgenics, respectively) liver was collected from male mice. Some c-Myc/TGF-α mice, starting from weaning, were divided in 2 groups fed, respectively, a NIH-31 based chow without (untreated mice), and with vitamin E (2,000 units/kg of diet) up to being killed (at 9 and 26 weeks).
Five normal human livers, 52 HCCs and corresponding surrounding tissues were used. Liver samples were kindly provided by Dr. Snorri S. Thorgeirsson (Laboratory of Experimental Carcinogenesis, National Cancer Institute, Bethesda, MD) and were obtained from a previous study.25 Clinicopathological features of the patients are shown in Supplementary Table I. Institutional Review Board approval was obtained at participating hospitals and the National Institutes of Health.
Samples were processed for hematoxylin and eosin staining, and proliferating cell nuclear antigen (PCNA) immunohistochemistry as published,27 using a mouse anti-PCNA monoclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Nuclear changes of apoptotic cells were evaluated in sections stained with the ApoTag peroxidase in situ apoptosis detection kit (Millipore, Billerica, MA). The percentage of PCNA-positive nuclei and apoptotic cells (PCNA and apoptotic index, respectively) was calculated by counting at least 2,000 hepatocytes per sample.
S-Adenosylmethionine and S-Adenosylhomocysteine assay
Liver MatI/III and MatII activities were determined as published.13, 28 Liver SAM and SAH content in perchloric acid extracts was determined by isocratic high-performance liquid chromatography as reported.29 For quantitative analysis, the area of the peaks in the tissue extracts was compared with that of standard solutions.
DNA and RNA isolation
DNA was isolated as described.27 Total RNA was isolated, purified and treated with DnaseI (Amersham Biosciences, Piscataway, NJ), as published27 and cDNA was synthesized by the High Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA).
Primers for S-Adenosylhomocysteine hydrolase (Ahcy), Cystathionine beta-synthase (Cbs), Betaine-homocysteine methyltransferase (Bhmt), Methionineadenosyltransferases 1A and 2A (Mat1A, Mat2A), Methylthioadenosinephosphorylase (Mtap), 5-Methyltetrahydrofolate-homocysteine methyltransferase (Mtr), Ornithine antizyme 1 (Oaz1), Ornithine decarboxylase (Odc), SAM decarboxylase (Amd), Spermidine synthase (Smr), Spermine synthase (Sms) genes and RNR-18 were chosen with the assistance of the “Assay-on-Demand™ Products” (Applied Biosystems, Foster City, CA). PCR reactions were performed with 75–300 ng of cDNA, using an ABI Prism 7000 Sequence Detection System and the TaqMan Universal PCR Master Mix (Applied Biosystems) as described.27 Quantitative values were calculated by using the PE Biosystems Analysis software.
Hepatic tissue samples were homogenized in lysis buffer as published.27 Protein concentrations were determined with the Bio-Rad Protein Assay Kit (Bio-Rad, Hercules, CA), using bovine serum albumin as standard. Aliquots of 100 μg were denatured, separatedby SDS-PAGE, and transferred onto nitrocellulose membranes by electroblotting.27 Membranes were probed with anti-ODC, MTAP and OAZ1 goat polyclonal antibodies (Santa Cruz Biotechnology), followed by incubation with HRP-secondary antibody, and revealed with Luminol Reagent (Santa Cruz Biotechnology). Densities of the protein bands were normalized to actin levels and calculated by ImageQuant Software. For co-immunoprecipitation studies, 350 mg of wild-type, preneoplastic and neoplastic liver from c-Myc and c-Myc/TGF-α transgenic mice were processed as reported.30 Immunoprecipitations were carried out with 6 μg of rabbit polyclonal anti-ODC (Santa Cruz Biotechnology) antibody. Immunoprecipitated proteins were subjected to western blotting and membranes were probed with anti-OAZ1 goat polyclonal antibody.
Global DNA methylation
Two μg of genomic DNA samples were pre-treated with 10-fold excess of HpaII endonuclease (New England Biolabs, Beverly, MA) and subjected to the cytosine-extension assay as described.31Msp I, an isoschizomer of HpaII insensitive to methylation status, was used as a control for efficiency of enzyme digestion. Relative incorporation of [3H]dCTP was compared between the samples. Results of the cytosine-extension assay were validated by comparing them with those obtained with the methyl acceptance assay in the same collection of samples. Methyl acceptance assay was performed as previously described.32 The percentage of methyl cytosine per μg of DNA was also determined in acid extracts of some normal and preneoplastic liver as well as HCC from transgenic mice by HPLC, as previously reported.9 No significant differences in the methylation pattern of the tissues evaluated by the 3 methods were found, and only the results of the cytosine-extension assay are shown.
Random amplified polymorphic DNA analysis
Twenty-two GC-rich arbitrary primers33 (Supplementary Table II) were used to score genomic alterations in preneoplastic and neoplastic mouse lesions and human HCCs. Random amplified polymorphic DNA (RAPD) reactions were performed as previously described.34 Every experiment included fingerprinting for 3 different concentrations of genomic DNA to evaluate the method reproducibility. The frequency of altered RAPD profiles was calculated for each liver lesion.34 RAPD analysis represents a sensitive and reliable method allowing the quantitative evaluation of a wide number of genetic alterations, ranging from large deletions or genomic rearrangements to single nucleotide mutations.35
Data are means ± standard deviation (SD). Student's t and Tukey–Kramer tests were used to evaluate statistical significance. Fisher's exact test was used for comparative analysis of the survival of HCC patient subgroups. Values of p < 0.05 were considered to be significant.
Human tumors were classified as HCC with better (HCCB) and poorer (HCCP) prognosis on the basis of the survival time, which ranges higher and lower than 36 months for HCCB and HCCP, respectively. Mean values ± SD of survival length were: 75.27 ± 20.85 (n = 15) for HCCB, and 22.58 ± 6.29 (n = 28) for HCCP (HCCB vs. HCCP, p < 0.001; Supplementary Table I). On the basis of the length of patients' survival, PCNA and apoptosis indices and DNA methylation features (see later) a 3rd human HCC subgroup was identified, with mean survival of 43 ± 27.07 months (n = 9; HCCI vs. HCCB and HCCP, p < 0.001 and < 0.01, respectively). Fisher's exact test comparison of means, and comparison of Kaplan–Meyer survival curves (Supplementary Fig. 1) by log-rank test showed significant differences among the 3 survival subgroups (p < 0.0001). No significant differences among groups were observed for average age at the time of diagnosis, sex, tumor size and differentiation and alpha-fetoprotein serum levels. Cirrhosis was present in about 87 and 65% of HCCB and HCCP, respectively, whereas all HCCI developed in the absence of cirrhosis. No significant differences were detected in any of the parameters investigated between noncirrhotic and cirrhotic surrounding liver tissues.
S-Adenosylmethionine/S-adenosylhomocysteine ratio and methionine adenosyltransferase isozymes
SAM content underwent 44% decrease in c-Myc HCC, when compared with normal liver, but slightly increased in neoplastic liver of c-Myc/Tgf-α transgenic mice (Fig. 1). Consequently, SAH level increased 70% in dysplastic and neoplastic lesions of c-Myc transgenics, whereas it did not change and decreased 59% in corresponding lesions from c-Myc/Tgf-α transgenics. SAM/SAH ratio sharply decreased in dysplastic and neoplastic liver of c-Myc transgenics, and rose 207% in HCC of double transgenics.
Decrease in SAM/SAH ratio was associated with reduced MatI/III and increased MATII activity in HCC of c-Myc transgenic mice, with consequent 32-fold decrease in MatI/III:MatII activity ratio (Fig. 1). Less pronounced changes were detected in c-Myc dysplastic liver. In dysplastic and neoplastic lesions of c-Myc/Tgf-α transgenics, the levels of MatI/III did not change when compared with wild-type livers, whereas a progressive induction of MatII activity (although at lower extent than in c-Myc corresponding lesions) occurred. As a consequence, MatI/III:MatII ratio, although lower in dysplastic and neoplastic liver of c-Myc/Tgf-α transgenic mice than in normal liver, was still 2- to 11-fold higher than in c-Myc corresponding lesions.
In human liver samples, a significant decrease in SAM content was detected in HCCB (Fig. 2) and, at a lower extent, in surrounding livers (SLB and SLP), with respect to control liver. SAM decrease was associated with increased SAH levels and consequent fall in SAM:SAH ratio in HCCB, SLB and SLP. An even more pronounced fall in SAM content occurred in HCCP with a concomitant rise in SAH levels, leading to a marked decrease in SAM:SAH ratio, up to values 22-fold lower than those of HCCB (p < 0.0001). No change in these parameters occurred in SLI and HCCI, with respect to control. The alterations in SAM and SAH content were paralleled by decrease in MATI/III and rise in MATII activities in the lesions with better and poorer prognosis, with consequent sharp reduction of MATI/III:MATII ratio. No significant changes in these parameters occurred in SLI and HCCI when compared with normal livers.
The changes in MATs activity were paralleled by decrease in Mat1A and increase in Mat2A gene expression (Fig. 3) in HCC of c-Myc transgenic mice, and increase in Mat1A and, at a lower level, in Mat2A in HCC of double transgenics. Analogous but lower changes occurred in dysplastic liver of the 2 transgenic lines. In human liver lesions (Fig. 4), a sharp decrease in MAT1A expression occurred in HCC and surrounding liver with better and poorer prognosis when compared with normal livers, with the lowest levels being detected in HCCP. MAT1A decline was associated with an increase in MAT2A expression in HCC, particularly in HCCP. In contrast, these parameters did not change in SLI and HCCI.
Proliferation, genomic instability and DNA methylation
RAPD analysis (Fig. 5) showed highest GI in the lesions of double transgenic mice, with significantly higher values in HCC than dysplastic liver of both transgenic lines. Similarly, PCNA index was ∼6 times higher in dysplastic lesions and 11–15 times higher in HCC of both transgenic mice than in control liver, with a relatively small difference between HCCs and no difference between dysplastic lesions of c-Myc and c-Myc/Tgf-α transgenics. A significant correlation occurred between RAPD and PCNA values in the tissues tested (r = 0.909; p < 0.0001).
A progressive rise in DNA hypomethylation was observed in dysplastic and neoplastic liver of c-Myc transgenic mice, with no changes detected in liver lesions from c-Myc/Tgf-α transgenics (Fig. 5). Therefore, evaluation of the correlation between DNA methylation and GI was only possible in c-Myc transgenics. Increase in dCTP incorporation into DNA, inversely proportional to global DNA methylation, exhibited a strong and highly significant correlation with GI (r = 0.957; p < 0.0001).
Since c-Myc/Tgf-α transgenic mice exhibit strong oxidative damage,36 we evaluated the effect of vitamin E on GI and DNA methylation of early (9 weeks) and late (16 weeks) dysplastic lesions in these mice. Vitamin E treatment had no effect on DNA methylation levels in c-Myc/TGF-α liver lesions ([3H]dCTP incorporation in untreated animals: 5.6 ± 0.3, 9.8 ± 0.3 and 16.8 ± 1.6, in normal liver, dysplastic liver at 9 and 26 weeks, respectively; [3H]dCTP incorporation following vitamin E administration: 6.3 ± 0.4, 10.1 ± 0.7 and 18.0 ± 0.9, respectively; means ± SD, n = 8). However, about 70% decrease in GI occurred at 9 and 26 weeks in dysplastic liver of vitamin E-treated mice (RAPD values of early and late dysplasia: 19.62 ± 2.56 and 21.12 ± 3.33, respectively, without vitamin E, and 5.81 ± 1.1 and 6.37 ± 1.53, following vitamin E treatment, respectively; means ± SD, n = 8; without versus with vitamin E, p < 0.001).
In the absence of normal liver controls corresponding to single HCCs, RAPD analysis of human lesions was performed by comparing each HCC with corresponding non-neoplastic surrounding liver (Fig. 6). GI was 2 times higher in HCCP and 2 times lower in HCCI than in HCCB. Similarly, PCNA index was highest in HCCP and lowest in HCCI, whereas no significant differences among surrounding liver subgroups were detected. A strong correlation between RAPD and PCNA was found (r = 0.92; p < 0.0001). DNA hypomethylation progressively increased from nontumorous surrounding liver tissues to HCC, with the highest values detected in HCCP. No changes in global DNA methylation occurred in HCCI and SLI. A strong correlation occurred between GI and global DNA methylation (r = 0.95; p < 0.0001). The results in Figure 6 also show significant negative correlations of RAPD and PCNA with survival rate (r = −0.453 and −0.582, respectively; p < 0.0001). A negative correlation of labeled cytosine incorporation into DNA and survival rate (r = −0.463; p < 0.0005) was also found, indicating that global DNA hypomethylation negatively correlates with survival length.
In HCCI, the presence of GI and PCNA values lower than those of HCCB and HCCP apparently contrasts with the observation that the survival of HCCI patients is intermediate between that of HCCB and HCCP patients. Thus, we evaluated whether changes in apoptosis were responsible for the reduced survival of HCCI patients when compared with HCCB patients. For comparison, apoptosis was also determined in dysplastic lesions and HCCs of c-Myc and c-Myc/Tgf-α transgenic mice. As shown in Figure 7(a), the apoptotic index increased more than 200-folds in dysplastic and neoplastic lesions from c-Myc transgenic mice with respect to normal liver. A lower apoptosis increase occurred in c-Myc/Tgf-α dysplastic liver, whereas it declined sharply in HCC, reaching values ∼10-fold lower than those of c-Myc HCC. Rise in apoptosis was also found in all human HCC subgroups and corresponding SL when compared with normal livers [Fig. 7(b)]. Importantly, HCCI and HCCP showed apoptotic indices ∼4-fold lower than HCCB. A positive correlation between apoptosis and survival rate was also found (r = 0.743; p < 0.0001). Taking into account the PCNA data (Figures 5 and 6), these results assign a role to apoptosis in the determination of HCC aggressiveness and prognosis.
S-Adenosylmethionine-related metabolic pathways
SAM-related pathways are controlled by genes involved in reduced glutathione, methionine and polyamine synthesis. We comparatively evaluated the expression of some key genes of these pathways. In transgenic mice (Fig. 3), Ahcy levels increased in HCC of c-Myc transgenic mice as well as in dysplastic liver and HCC of the double transgenics, suggesting a relatively high production of homocysteine. Presumably, this was not associated with rise in reduced glutathione synthesis because of decreased Cbs levels in dysplastic and neoplastic tissues of the 2 transgenic models. Bhmt and Mtr control methionine synthesis. Bhmt expression decreased in dysplastic and neoplastic liver of both transgenic lines. Mtr expression did not change in c-Myc lesions, showing a slight increase in dysplastic liver and a 400% rise in HCC of double transgenics. As concerns polyamine synthesis, a progressive increase in Amd, Odc, Smr and Sms mRNAs was detected in dysplastic and neoplastic lesions of both transgenic models, with highest levels being detected in the lesions of c-Myc/Tgf-α transgenics. An increase in Oaz1, much more evident in the lesions of c-Myc transgenics, occurred in dysplastic and neoplastic lesions of both mouse models. Finally, Mtap, encoding a key enzyme for methionine re-synthesis through the salvage pathway, was overexpressed only in the lesions of double transgenics.
Some genes related to polyamine synthesis and salvage pathway were also evaluated at the protein level in transgenic mice (Fig. 8). Odc and Oaz1 expression in the different tissues closely resembled that of specific mRNAs. Relatively low increase in Odc and high rise in Oaz1 was associated with the formation of relative high amounts of Odc/Oaz1 complexes in the lesions of c-Myc transgenics. In contrast, high Odc levels and low production of Oaz1 led to the formation of relatively low amounts of Odc/Oaz1 complexes in double transgenic mice. These changes were more evident in HCC than dysplastic lesions. The high expression of Mtap gene, in the lesions of double transgenics, was also confirmed at the protein level.
A slight increase or no change in AHCY expression (Fig. 4) in human HCCs with better and poorer prognosis and their corresponding surrounding liver, with respect to control liver, contrasted with AHCY sharp increase in HCC with intermediate prognosis (HCCI) and surrounding liver (SLI). Cbs mRNA underwent a marked decrease in HCC and SL of all subgroups. As concerns the genes encoding key enzymes of methionine synthesis, BHMT gene expression decreased in liver lesions of all subgroups, MTR showed a consistent increase only in SLI and HCCI, and MTAP, a key gene for methionine synthesis through the salvage pathway, decreased in HCCI and in HCCP. Finally, the levels of genes encoding key enzymes for polyamine synthesis, including AMD, ODC, SMR and SMS, were progressively increased from non-neoplastic surrounding liver tissues to HCC, with the highest values being detected in HCCP. OAZ1 was overexpressed both in HCC and surrounding liver from all subgroups, with HCCP showing the lowest values.
Several researches showed the role of low SAM levels in the pathogenesis of oxidative damage, inflammation, fibrosis and carcinogenesis of rodent and human liver.7, 8, 9, 10, 11, 12, 13, 14, 37 Knockout mice lacking Mat1A gene exhibit low hepatic levels of SAM, spontaneous oxidative stress, hepatic hyperplasia, steatohepatitis and HCC.37 Furthermore, transfection of Hu-H7 cells with MAT1A and antisense oligonucleotides against MAT2A results in sizable decrease in DNA synthesis and increase in SAM level, SAM:SAH ratio, and DNA methylation.15 Analogous changes occur after supplementation of nontransfected Hu-H7 cells with SAM. In preneoplastic and neoplastic rat liver lesions, DNA synthesis inversely correlates with MATI/III activity, SAM content, SAM:SAH ratio and DNA methylation.9, 10 These observations suggest a causative role of decreased Mat1A expression and MATI/III activity in liver carcinogenesis. Here, we show similar changes of MATs mRNAs and activity associated with Mat1A:Mat2A switch in dysplastic and neoplastic liver of c-Myc transgenic mice and human HCCs with better and poorer prognosis. Most pronounced fall in SAM:SAH, MATI/III:MATII ratio and MAT1A and MAT2A expression ratio was detected in Myc HCC and HCCP, with lower changes being present in mouse dysplastic liver and human nontumorous surrounding liver. The alterations of MAT1A:MAT2A ratio, associated with the reduction of methionine synthesis through the pathway controlled by BHMT gene and, in the case of human HCC, also through the salvage pathway, may contribute to the decrease in SAM synthesis, high SAH content, and decrease in SAM:SAH ratio and DNA methylation in c-Myc transgenics and human HCCs (Fig. 9). A fall in SAM liver content may enhance tumorigenesis through multiple mechanisms, including spontaneous oxidative stress, eventually enhanced by a fall in GSH synthesis, and DNA hypomethylation.7, 9, 37 Attention has been focused on the role of DNA hypomethylation in generation of chromosome instability.16, 17, 18, 19 GI and DNA hypomethylation are often early events in carcinogenesis, and a link between DNA hypomethylation and chromosome instability is shown by the finding that the demethylating agents 5-azacytidine and 5-aza-2-deoxycytidine induce pericentromeric rearrangement.18, 19 Our findings are in agreement with these observations, and the presence of abnormal altered RAPD bands in 45% of the primers used in this study indicates a relatively high number of genomic abnormalities in murine and human liver lesions, implying a relatively high GI. Notably, global DNA hypomethylation correlates with GI and DNA synthesis both in c-Myc transgenic mice and human HCCs. In the latter, DNA hypomethylation and GI are higher in HCCP than HCCB and inversely correlated with the length of patients' survival. These observations strongly suggest that changes in methionine and SAM metabolism and in DNA hypomethylation may have an important prognostic value in human hepatocarcinogenesis. A correlation of global DNA hypomethylation with poor prognosis has been reported for ovary38 and squamous cell lung39 cancers, whereas the impact of aberrant DNA methylation on breast cancer is controversial.40, 41
A completely different situation occurred in c-Myc/Tgf-α double transgenics, where absence of Mat1A:Mat2A switch was associated with relatively high SAM content and SAM:SAH ratio, and elevated expression of genes regulating methionine synthesis (at least through Mtr), and methionine recovery through the salvage pathway. These observations are in agreement with the absence or very low decrease of DNA methylation in c-Myc/Tgf-α double transgenics. Thus, these findings indicate that DNA hypomethylation does not play a role in the generation of high GI and proliferation, associated with low apoptosis, in dysplastic lesions and HCCs from c-Myc/Tgf-α transgenic mice. In this respect, liver lesions of the double transgenics differ from HCCB and HCCP, which instead exhibit decrease in MATI/III:MATII and SAM:SAH ratios leading to DNA hypomethylation, high GI and elevated growth rate. The observation that all these changes are significantly higher in HCCP than in HCCB strongly suggests their prognostic role. In contrast, the behavior of c-Myc/Tgf-α double transgenics closely corresponds to that of a small human HCC subgroup, representing about 17% of total HCCs examined, with intermediate survival length between HCCB and HCCP. HCCI are characterized by the absence of both MAT1A:MAT2A switch and DNA hypomethylation, and the presence of relatively low GI. Similar to HCCB, HCCI exhibit relatively low DNA synthesis, but an apoptotic index as low as that of HCCP. These parameters categorize HCCI. The low apoptosis, in the presence of relatively low proliferation, might be responsible in HCCI of a progression rate intermediate between that of HCCB and HCCP. Recent observations indicate a proapoptotic effect of the inhibition of AHCY.42 This suggests a link between high AHCY gene expression and low apoptosis, both in HCCI and c-Myc/Tgf-α HCC. The molecular mechanisms whereby AHCY suppresses apoptosis are still unclear. To the best of our knowledge, this is the first report of a HCC subgroup, both in rodents and humans, showing barely detectable changes in DNA methylation, a situation that excludes a role played by DNA hypomethylation in the high GI characteristic of c-Myc/Tgf-α double transgenic mice. In these mice, GI and HCC development are strongly inhibited by vitamin E.36, 43 High production of reactive oxygen species, associated with lipid peroxidation, increased nitric oxide synthase and nitric oxide-mediated oxidative damage, and disruption of DNA repair mechanisms36, 44 may concur to elevated GI, proliferative activity, and HCC aggressiveness in double transgenics. Overproduction of nitric oxide also occurs in human HCC.45 However, different from c-Myc/Tgf-α mice, human HCCI exhibit relatively low GI, implying that diverse molecular mechanisms underlie the phenotypic behavior of c-Myc/Tgf-α HCC and HCCI. Further studies are needed to analyze these differences.
An intriguing finding of this study is the absence of cirrhosis in all HCCI patients (Supplementary Table I). Cirrhosis was instead present in the majority of HCCB and HCCP cases. Since various alterations present in human HCC may also be found, although at a lower extent, in cirrhotic liver,3, 5 it cannot be excluded the influence of the lesions present in precancerous liver on HCC subtypes. However, the absence of cirrhosis in 45% of HCCP patients, excludes a major role of this alteration in the phenotypic behavior of HCCI.
The expression features of key enzymes in polyamine synthesis indicate a rise in polyamine production in dysplastic and neoplastic lesions of c-Myc and c-Myc/Tgf-α transgenic mice, as well as in human HCC and nontumorous surrounding liver. OAZ1 expression may control polyamine synthesis by counterbalancing ODC activity. We found that OAZ1 expression is inversely correlated to HCC aggressiveness in transgenic mice models, and to HCC prognosis in humans. SAM decarboxylation is a key reaction of polyamine synthesis, and in the presence of low SAM liver levels, polyamine synthesis is metabolically conserved at the expense of methylation reactions,3 This may explain the apparent scarce influence of the alterations of methionine metabolism on polyamine synthesis, both in transgenic mice and human HCC.
The molecular mechanisms underlying the changes in expression of genes involved in methionine and SAM metabolism in c-Myc and c-Myc/Tgf-α transgenic mice remain unknown. Promoter methylation has been reported as a mechanism responsible for downregulation of the MAT1A gene in HCCs.46 However, the relationships between the overexpression of the c-Myc transgene either alone or in combination with Tgf-α and MAT1A expression levels are unclear. Both c-Myc and TGF-α are upregulated in human HCCs,3, 5 but at a lower extent than in double transgenic mice. Further work is required to address the molecular mechanisms underlying different dysregulation of methionine metabolism in different HCC subtypes.
Collectively, our results indicate that early changes in methionine/SAM metabolism and global DNA methylation may have a prognostic value for hepatocarcinogenesis in the majority of individuals, probably acting through a modulation of GI. Furthermore, we identified a HCC subtype that may develop both in human and mouse liver independent of changes in DNA methylation. Interestingly, global DNA hypomethylation was found to promote early lesions in the colon and liver in a mouse model of intestinal carcinogenesis, while suppressing later stages of tumorigenesis.21 Conversely, DNA hypomethylation in ApcMin/+ mice suppresses tumor formation,47, 48, 49 implying that DNA hypomethylation may exert paradoxically opposite roles on carcinogenesis depending on the cell type, tumor stage or the model used. These findings are in keeping with our observation in double transgenic mice and HCCI, in which tumor development can also occur in the absence of global DNA hypomethylation. This indicates, different from previous evidence,9, 13, 15 that molecular alterations linked to SAM metabolism and DNA methylation are necessary for the development of the majority, but not all, human HCCs. These observations may have some relevance to chemoprevention of liver tumors by SAM,8, 9, 10 at least in HCC subtypes in which DNA hypomethylation plays a pathogenetic role. Furthermore, our finding that a rise in DNA hypomethylation is associated with increased aggressiveness of HCC and decreased patient survival calls for caution in using demethylating agents as anticancer drugs.
We thank Dr. Snorri S. Thorgeirsson (Laboratory of Experimental Carcinogenesis, National Cancer Institute, Bethesda, MD, USA) for providing the c-Myc and c-Myc/Tgf-α transgenic mice and the human liver tissue samples.