Predicting Hepatocellular Carcinoma by Detection of Aberrant Promoter Methylation in Serum DNA Free

Hepatocellular carcinoma (HCC) is one of the most common and rapidly fatal human malignancies. The almost 500,000 new cases and nearly equivalent number of fatalities show the lack of effective therapeutic alternatives for this disease that is largely diagnosed at an advanced stage; most patients die within 1 year of diagnosis (1). Chronic hepatitis B and C virus infections are well-documented risk factors for the development of HCC. Several environmental factors, including aflatoxin B₁, a dietary mold contaminant, and polycyclic aromatic hydrocarbons, ubiquitous environmental contaminants, are also associated with the development of HCC (2–4). Although HCC incidence is highest in East Asia and sub-Saharan Africa (1), it is also increasing in United States (5) Currently available screening tests to detect smaller and more frequently unifocal (early stage) HCC combine α-fetoprotein analysis and ultrasound. However, although screening for early detection of HCC has become quite common in clinical practice, its effectiveness remains controversial (6).

As with other cancers, the development of HCC is a complex, multistep process. The molecular pathogenesis of HCC seems to involve multiple genetic aberrations in the molecular control of hepatocyte proliferation, differentiation, and death and the maintenance of genomic integrity. This process is influenced by the cumulative activation and inactivation of oncogenes, tumor suppressor genes, and other genes. Epigenetic alterations are also involved in cancer development and progression (7–9). Methylation of promoter CpG islands is known to inhibit transcriptional initiation and cause permanent silencing of downstream genes. Hypermethylation of p16, a cyclin-dependent kinase inhibitor gene that regulates the cell cycle, has been detected frequently in human cancers (10). p15, another cyclin-dependent kinase inhibitor gene adjacent to p16 on chromosome 9p21, has been postulated to be a tumor suppressor modulating pRb phosphorylation. It is also aberrantly methylated in several human neoplasms, including HCC (11, 12). The ras association domain family 1A (RASSF1A) gene is located in chromosome 3p21.3 and, from initial studies in lung and breast cancer, was suggested to be a tumor suppressor gene (13). We and others have consistently reported a high frequency, in the range of 80% to 90%, of RASSF1A promoter hypermethylation in HCC tissues (14, 15).

Detection of methylated DNA has been suggested as a potential biomarker for early detection of cancer (16). Because an ideal biomarker should appear early in the course of disease and should be detectable in biological samples that can be obtained noninvasively, many studies have focused on the detection of genetic and epigenetic abnormalities in exfoliated cells from sputum, bronchoalveolar lavage, or cervical smears as well as in the circulating DNA found in serum or plasma. In our recent study, p16 was methylated in 24 of 40 (62%) tissues and 12 of 39 (32%) plasma DNAs from blood collected at the time of diagnosis (17). Other studies of HCC patients have also found that methylated DNA can be frequently detected in blood collected at the time of diagnosis (12, 18, 19). These results suggest that biomarkers in plasma or serum may help in estimating the risk for the development of HCC; however, their sensitivity and specificity for HCC detection and their clinical utility remain uncertain at the present time. No studies of HCC have assayed blood samples collected years before diagnosis for aberrant methylation.

In the present study, we explored the possible diagnostic value of aberrant promoter hypermethylation using a panel of three tumor suppressor genes in serum DNA for early detection of HCC. We took advantage of a sample bank collected for a cancer screening program in Taiwan, in which repeat samples before diagnosis were available.

Human subjects and sample collection. This study was approved by Columbia University's Institutional Review Board as well as the research ethics committee of the College of Public Health, National Taiwan University (Taipei, Taiwan); written informed consent was obtained from all subjects, and strict quality controls and safeguards were used to protect confidentiality. Fifty subjects with HCC were randomly chosen from cases identified in the Cancer Screening Program study, a community-based cohort recruited in Taiwan; 50 controls were selected by matching by age (within 3 years) and sex. Blood samples for controls were collected from 1991 to 1992. The cohort characteristics and methods of screening and follow-up have been described in more detail previously (20). Briefly, individuals were between 30 to 64 years old and lived in seven townships in Taiwan, three located on Penghu islets with the highest HCC incidence in Taiwan and the other four from Taiwan island. A total of 12,020 males and 11,923 females were recruited between July 1990 and June 1992. Participants were personally interviewed based on a structured questionnaire and donated a 20 mL blood sample at recruitment. Aliquots of serum were separated from other components in blood and stored at −70°C. Specimens were transported on dry ice to a central laboratory at the National Taiwan University and were kept at −70°C until shipment to Columbia for analysis.

Blood samples were screened in Taiwan for serologic markers, including alanine transaminase, aspartate transaminase, α-fetoprotein, HBsAg, and anti–hepatitis C virus (HCV), using commercial kits (HBsAg, anti-HCV, and α-fetoprotein, Abbott Laboratories, North Chicago, IL) or a serum chemistry autoanalyzer (alanine transaminase and aspartate transaminase, Hitachi Model 736, Hitachi Co., Tokyo, Japan). Any participant who had an elevated level of α-fetoprotein (≥20 ng/mL) was positive for HBsAg or anti-HCV or any participant who had a family history of HCC or liver cirrhosis among first-degree relatives was referred for upper abdominal ultrasonographic examination. They were also followed with additional blood collections. Suspected HCC cases were referred to teaching medical centers for confirmatory diagnosis by computerized tomography, digital subtracted angiogram, aspiration cytology, and pathologic examination. The criteria for HCC diagnosis included a histopathologic examination or a positive lesion detected by at least two different imaging techniques. Through 2003, a total of 162 HCC patients provided a baseline and at least one follow-up blood sample. Because the annual follow-up in high-risk subjects was voluntary, each case had multiple samples collected before diagnosis.

Methylation-specific PCR. DNA was extracted from 200 μL serum using QIAamp UltraSens Virus kits (Qiagen, Valencia, CA) following the viral RNA and DNA purification protocol. Bisulfite modification was conducted using a CpGenome DNA Modification kit (Chemicon International, Temecula, CA) following the manufacturer's recommendations. PCR was conducted with the CpGWIZ p16 and p15 Amplification kits (Chemicon International) and AmpliTaq Gold Polymerase (Perkin-Elmer, Norwalk, CT) and a total of 40 cycles. The thermal profile consisted of an initial denaturation step of 95°C for 10 min followed by repetitions of 95°C for 45 s, 60°C for 45 s, and 72°C for 60 s, with a final extension step of 72°C for 10 min. For detection of RASSF1A methylation, primers and amplification conditions were as described previously (19). PCR products were analyzed by agarose gel electrophoresis and Vista Green (Amersham Biosciences, Piscataway, NJ) staining. The methylated DNA control from the CpGenome Amplification kit and universal methylated DNA (Chemicon International) were used as positive controls and distilled water as a negative control. As a quality control for the bisulfite modification process, all bisulfite-treated DNAs were also amplified with primers specific for the unmethylated p16, p15, and RASSF1A (12, 18, 19).

Statistical analysis. The associations of methylation status with clinical factors were analyzed by Fisher's exact test, using the data from the blood collected closest to diagnosis. Differences in the means of continuous variables (i.e., alanine transaminase, aspartate transaminase, and α-fetoprotein) between the methylation status of genes were analyzed using Mann-Whitney test. Differences at P < 0.05 were considered significant. Conditional logistic regression was used to construct receiver operating characteristic curves using clinical risk factors and methylation biomarkers (21).

Subject characteristics. The demographic data are presented in Table 1. There were a total of 11 females and 39 males, 50% were smokers, 24% had habitual alcohol consumption, 51% were HBsAg positive and 24% were anti-HCV positive.

p16, p15, and RASSF1A promoter methylation in serum DNA. The promoter methylation status for p16, p15, and RASSF1A of DNA isolated from serum collected at different time points before diagnosis was assayed by methylation-specific PCR (MSP). Figure 1 shows representative MSP analyses for RASSF1A and p16 in serum DNA from five HCC cases with blood collected at different time point before diagnosis. In Fig. 2, representative MSP analysis using methylated and unmethylated primers for p15 are shown for three cases. In the 50 serum samples from HCC cases collected closest to diagnosis (0-9 years prior), 22 (44%) were positive for methylation of p16, 12 (22%) for p15, and 35 (70%) for RASSF1A (Table 1). Six of the 50 cases had hypermethylation of all three genes, 13 cases for two genes and 25 cases for one gene; six subjects were hypermethylation negative for all three genes.

A total of 14 samples were available that had been collected 1 to 3 years earlier than the sample collected closest to diagnosis, and of these, hypermethylation was positive for nine (64%) for p16, two (14%) for p15, and four (29%) for RASSF1A (Table 2). In the three available serum samples that were collected another 2 years earlier, two (67%) were positive for promoter hypermethylation of RASSF1A (Table 2) but none for p16 and p15 (Table 2). The methylation status of p16, p15, or RASSF1A did not differ by gender. The frequency of p16 promoter hypermethylation was significantly higher in HBsAg-positive (60.0%) than HBsAg-negative (25.0%) HCC cases (P = 0.01) The association of p15 promoter hypermethylation with hepatitis B virus infection was also statistically significant. (HBsAg-positive cases, 36.0%; HBsAg-negative cases, 12.5%; P = 0.04). The association between p16 and p15 promoter hypermethylation was weak and not statistically significant (P = 0.05). There was no association of RASSF1A promoter hypermethylation with hepatitis B virus infection (HBsAg-positive case, 72.0%; HBsAg-negative cases, 70.8%; P = 0.24). Methylation status of any one of three genes did not differ based on anti-HCV status, smoking, or habitual alcohol consumption. We also looked at the association of methylation and subjects' alanine transaminase, aspartate transaminase, and α-fetoprotein status, but no significant relationships were found.

Among the 50 samples from controls, promoter hypermethylation was detected in three subjects for RASSF1A and two for p16 (Table 3). No subjects were positive for p15 methylation.

Two receiver operating characteristic curves were constructed by separately, including only clinical risk factors (age, HBsAg status, anti-HCV status, smoking, and alcohol status) and these factors plus the three hypermethylation biomarkers (p16, p15, and RASSF1 methylation; Fig. 3A and B, respectively). The overall predictive accuracy is relatively low (67%) for the model that includes only the clinical risk factors (with a sensitivity of 66% and a specificity of 68%). The overall predictive accuracy is much better (89%) for the model that includes not only clinical factors but also hypermethylation biomarkers. The sensitivity and specificity were 84% and 94%, respectively, under the probability cutpoint of 0.50.

In the present study, we investigated serum DNA methylation for p16, p15, and RASSF1A, three tumor suppressor genes frequently hypermethylated in HCC. Blood samples were collected from 50 HCC cases 0 to 9 years before diagnosis. The frequencies of detection of gene methylation in the available samples collected closest to diagnosis are consistent with previous studies of serum DNA from HCC patients using blood collected at the time of diagnosis: p16, 44.% versus 48% (22); p15, 22% versus 25% (12); and RASSF1A, 70% versus 43% (19). The detection frequencies for p16 and RASSF1A are also similar to our previous findings in HCC tissue DNAs (14). Hypermethylation was detected 1 to 8 years before clinical diagnosis for p16, 1 to 5 years for p15, and 1 to 9 years for RASSF1A. These findings show that p16, p15, and RASSF1A hypermethylation are early events in the development of HCC.

A specific missense mutation in the p53 tumor suppressor gene at codon 249 has been reported in >50% of HCC tumors and in paired blood samples from areas of high dietary exposure to aflatoxin B₁ (23). Jackson et al. (24) detected this mutation in DNA from plasma collected 1 to 5 years before diagnosis, suggesting that it could be used as a biomarker for aflatoxin exposure and HCC development, but it only can be detected in cases from areas with high aflatoxin B₁ exposure.

Several studies have indicated that epigenetic changes might ‘addict’ cancer cells to altered signal transduction pathways during the early stages of tumor development (16, 25). Hypermethylation of CpG islands in gene promoters can appear early in the progression of lung and colon cancer or can be characteristic of premalignant lesions at these sites (26). Belinsky et al. (27) reported that the frequency of aberrant methylation of p16 increased during disease progression from bronchial basal cell hyperplasia (17%) to squamous metaplasia (24%) to carcinoma in situ (50%). Aberrant promoter methylation of p16 and MGMT was also detected by others in sputum DNA in 100% of patients with squamous cell carcinoma of the lung up to 3 years before clinical diagnosis (28) and by Belinsky et al. (29) in multiple other genes in sputum from patients with lung cancer several months to 3 years before clinical diagnosis. These findings show the promise of gene promoter hypermethylation in sputum as a molecular marker for identifying people at high risk for cancer.

Epigenetic alterations, including methylation of p16 (30–32), p15 (12), RASSF1A (14), MGMT (33), GSTP1 (34, 35), and other genes, are prevalent in HCC tissue samples (36). By using MSP, methylation changes of p16, p15, and RASSF1A were also detected in the plasma and serum of HCC patients (13, 19, 20). Wong et al. (18) reported a good correlation between p16 hypermethylation in HCC tissues and plasma/serum DNA (72%) and that p16 hypermethylation was not detected in the plasma/serum of patients with either liver cirrhosis or hepatitis. However, in another study, 17% of cirrhosis patients had serum DNA with aberrant p16 methylation (22). These differing results may be due to the lack of standardized processing of blood samples and methods of analysis; the relatively small sample size and diversity in the clinical courses of patients may also contribute to the variation. Thus far, no studies on the relationship between methylation status of p15 and RASSF1A in serum DNA and cirrhosis have been reported. Thus, the relationship between methylation status of different tumor suppressor genes and precancerous lesions like cirrhosis needs further study and the significance of epigenetic changes in serum DNA from cirrhosis patients is currently unclear.

p16 and p15 methylation were associated with hepatitis B virus infection in this study, implying an environment-epigenetic interaction in the development of HCC. A recent study with similar results suggested that hepatitis viruses might induce p16 methylation in liver tissues with chronic inflammation, before the appearance of HCC (37), but this correlation is still controversial (14, 18, 38). No correlation between p15 methylation and hepatitis B virus infection was found in the previous study (12). In the present study, p15 methylation correlated with p16 methylation in serum DNA (P = 0.05). Dual p15 and p16 methylation has been found almost exclusively in hematologic malignancies, such as Burkitt's lymphoma and acute T-cell leukemia (11, 39). In terms of clinical relevance, p16 and p15 methylation were significantly associated with the development of a recurrence or metastasis (12). Thus, p16 and p15 methylation may be implicated in tumor progression. Although it is recognized that malignant tumors harbor dense methylation in normally unmethylated promoter CpG islands (8), our previous study and those of others showed that hypermethylation of tumor suppressor genes, including p16 and RASSF1A, were absent or very low in normal tissues DNA (14, 40, 41). In a study of breast cancer, a few healthy controls (n = 10) were tested and RASSF1A was methylated in the plasma DNA from one subject (10%; ref. 42).

In the present study, 50 matched serum DNAs from normal controls were also investigated for methylation status. Promoter hypermethylation of p16 and RASSF1A was detected in two and three normal controls, respectively. Compared with the 50 cases, these detection levels are very low (2 controls/22 cases for p16 and 3 controls/35 cases for RASSF1A). Three of four positive controls (one subject had hypermethylation in both p16 and RASSF1A) had either hepatitis B virus or/and HCV infections; one subject had a history of smoking and alcohol drinking. Although it is controversial, some risk factors have been reported to correlate with gene methylation (14, 29). Hypermethylation in serum DNA from controls was perhaps due to hepatitis virus infection and chemical carcinogen exposure. Another possibility was that some “normal controls” have cryptogenic hepatic cirrhosis. A small percentage of patients with liver cirrhosis were reported to be positive for p16 methylation in serum DNA (22), but these changes still need further study. In the present study, the 50 cases were randomly selected. Thus, although the results may not represent the methylation status of all the HCC cases in this cohort, there should be no selection bias.

In conclusion, this is the first study to prospectively examine epigenetic changes in tumor suppressor genes for predicting HCC development in a cohort of high-risk subjects. p16, p15, and RASSF1A promoter hypermethylation were detected in DNA from serum samples collected up to 9 years before clinical diagnosis. Compared with controls, detection of promoter hypermethylation on these three genes was much more frequent in HCC patients before diagnosis. These molecular changes may be a valuable biomarker for early detection, risk assessment in high-risk populations, and monitoring the clinical course of HCC.