Predictive value of CpG island methylator phenotype for tumor recurrence in hepatitis B virus-associated hepatocellular carcinoma following liver transplantation

A total of sixty-five patients (59 men, 6 women; mean age 48.8 years; range, 29-67 years) who underwent LT in our institution between 2003 and 2005 were enrolled in this retrospective study according to the same eligibility criteria as our recent study [5]: (a) The diagnosis of HCC were confirmed by histopathologic examination either before or after transplantation (as an incidental finding); (b) all patients included in this retrospective study were HBV-positive; (c) The clinical and laboratory data was obtained for all 65 patients, including portal vein tumor thrombi (PVTT), preoperative alpha-fetoprotein (AFP) level, histopathologic grading, tumor size, and tumor number; (d) all these patients were Han Chinese; (e) none of these patients received preoperative adjuvant antineoplastic therapy. The follow-up course and diagnosis criteria of recurrence have been shown in the previous study [7]. This study was approved by local ethic committee, and informed consent was obtained according to the Declaration of Helsinki. All the tissue samples from primary tumors were stored at -80°C tissue banks.

Aberrant promoter methylation of these genes was determined by method of methylation-specific polymerase chain reaction (MSP) as reported by Herman et al [32]. Genomic DNA was isolated from frozen liver tissue using DNeasy Tissue Kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. DNA samples were modified using EZ DNA Methylation Golden Kit (Zymo Research, Orange, CA). MSP distinguishes unmethylated alleles of a given gene based on DNA sequence alterations after bisulfite treatment of DNA, which converts unmethylated but not methylated cytosines to uracils. Subsequent polymerase chain reaction using primers specific to sequences corresponding to either methylated or unmethylated DNA sequences was then performed. DNA methylation of CpG islands was then determined by PCR using specific primers for either methylated or unmethylated DNA. Two sets of primers were used to amplify each region of interest: one pair recognized a sequence in which CpG sites were unmethylated (bisulfite-modified to UpG), and the other recognized a sequence in which CpG sites were methylated (unmodified by bisulfite treatment). The MSP primer sequences of each gene for the unmethylated and methylated reactions were determined as described previously [12, 22‐31]. The PCR amplifications were carried out with treated DNA as template in a total volume of 25 μl containing 25 pM of each primers, 25 μM deoxynucleoside triphosphates, 30 ng of bisulfate-treated DNA, 1 U of hot-start Taq polymerase (Takara, Shiga, Japan) and the respective buffers. Hot start polymerase chain reaction was performed at 95°C hot start for 10 minutes followed by 30 repetitive cycles consisting of denaturation at 95°C for 30 seconds, annealing at specific temperature for 30 seconds, and extension at 72°C for 30 seconds, then finished with a final 10-minute extension. To prepare the positive methylation control, 1 μg of genomic DNA from normal human liver was treated in vitro with SssI methyltransferase (NEB, Beverly, MA), yielding completely methylated DNA at all CpG rich regions. Bisulfite-modified DNA from normal human liver served as a positive control for the unmethylated alleles. Water blanks were also included with each assay. The PCR products were analyzed on ethidium bromide-stained 2% agarose gel. No clinicopathological or follow-up data were revealed to the bench researchers until the MSP results were finalized.

Statistical comparisons were performed using either the Pearson's chi-square test or two sided Fisher's exact test, as appropriate. Recurrence-free survival (RFS) was defined as the interval from the operation day to date of documented tumor recurrence or date of the most recent follow-up visit if recurrence did not occur. Log-rank tests were used to identify the number of methylated genes that was most predictive of RFS. Survival curves of the patients were compared using the Kaplan-Meier method and analyzed using the Log-rank test. For a given gene, a univariate Cox proportional hazard regression model was used to estimate the effect of methylation status (partially or completely methylated versus unmethylated) on the relative hazard of risk for tumor recurrence. The same method was also used to evaluate the univariate correlations between CIMP status or clinicalpathological variables, and tumor recurrence. Only variables with significance in univariate analysis were performed in the multivariate analysis using Cox proportional hazard model.

The frequency of promoter methylated of the gene included in the panel was 51% (33 of 65) for P16, 43% (28 of 65) for CDH1, 55% (36 of 65) for SOCS1, 60% (39 of 65) for GSTP1, 52% (34 of 65) for SYK, 66% (43 of 65) for XAF1, 52% (34 of 65) for DAPK1. The other five genes with the methylation frequency less than 40% were not analyzed in the CIMP panel (5% for MGMT, 15% for TMS1, 20% for SFRP1, 25% for TIMP3, 15% for DKK1). Representative examples of MSP assay results are presented in Fig. 1. Overall, only two cases showed no methylation of any of these genes.

Bild vergrößern

Study of correlations between promoter methylation status of each gene and clinicopathological characteristics are summarized in Additional file 1. Overall, the hypermethylation of GSTP1 was present more frequently in patients with poor histopathologic grading (P = 0.003) and multiple tumor numbers (P = 0.011). The hypermethylaton of XAF1 was present more frequently in HCC, with AFP > 400 ng/ml than those with AFP ≤ 400 ng/ml (P = 0.007). Likewise, patients with methylated DAPK were suffered from larger tumor size than those with unmethylated DAPK promoter (P = 0.016). However, we found no significant associations between the methylation status of P16, CDH1, SOCS1, SYK and the following clinicopathological features: patient's age and gender, PVTT, AFP level, histopathologic grading, tumor size and tumor numbers.

The relative risk for tumor recurrence of methylation status of an individual gene was shown in Additional file 2. Only methylaton of SYK was found to be prone to recurrence (P = 0.046), which was similar to the previous study in HCC [12]. No significant correlation was observed between tumor recurrence and methylation status of any other individual gene.

Originally, CIMP-positive gastric cancer was defined as a tumor with methylation at more than three genes [33]. CIMP was also defined by the average number of methylated genes per tumor in several recent studies [34, 35]. However, a recent new study offered potential markers to define a CIMP concept, in which the thredshold distinguishing CIMP+ from CIMP-samples was chosen by minimizing the within-group sum of squared errors [36]. In the present study, a methylation phenotype was defined on the basis of results of log-rank tests of the effect of number of methylated genes on RFS as reported by Yang [37]. The lowest P value in Table 2 was used to identify an optimal cutoff of the number of methylated genes. Accordingly, CIMP status was classified as CIMP+ samples (with three or more methylated genes) and CIMP-samples (with two or fewer methylated genes). Of the HCCs, 72% (47 of 65) were CIMP+.

Using statistical analysis, we examined CIMP with regard to HCC patient clinicopathologic parameters (Table 3). A significant difference between CIMP status and preoperative AFP level was being found in HCC with LT (P = 0.017). Meanwhile, we found that CIMP+ was more frequent in HCC, with AFP > 400 ng/ml (87%, 26/30) than those with AFP ≤ 400 ng/ml (60%, 21/35; P = 0.017). Likewise, patients with CIMP+ were prone to have multiple tumor numbers than those with CIMP- (P = 0.007). However, we found no significant associations between the CIMP status with other clinicopathological features like patient's age and gender, PVTT, histopathologic grading and tumor size. In addition, the results showed a correlation between CIMP and Hangzhou criteria, which consider tumor size, histological grading, and serum AFP level [38]. Patients that exceeded the Hangzhou criteria had a higher CIMP+ frequency compare to those matched Hangzhou criteria (84% vs. 57%, P = 0.017) (Table 3).

Recently, we used the Hangzhou criteria as a new criteria for prognosis prediction of HCC patients post-LT, and found those matched Hangzhou criteria had better RFS than the others. In this study, when the 65 patients were grouped according to the Hangzhou criteria, similar results were observed (P < 0.001, data not shown). Consistently, the cumulative 3-year RFS rate in patients with CIMP- is 64%, compared to that of 25% in patients with CIMP+. The mean RFS for CIMP- patients was significantly longer than that for CIMP+ patients by Kaplan-Meier estimates (30.3 vs. 15.3 months, P = 0.004) (Fig. 2). Considering the association of CIMP and Hangzhou criteria, the prognostic value of CIMP might overlap with that of Hangzhou criteria. Therefore, further segregation based on CIMP was applied to analyze the RFS in those who matched the Hangzhou criteria or not, respectively. Resultantly, a trend toward shorter RFS in CIMP+ than CIMP- was seen in both patients who matched the Hangzhou criteria and those exceeded the criteria, although there was no significant difference(P = 0.101,0.113, respectively, data not shown). Furthermore, Cox univariate analysis also revealed that the clinicopathological variables could provide significant predictive values for recurrence including preoperative AFP level, tumor size and CIMP status (data not shown). Multivariate analysis revealed that a larger tumor size and CIMP+ could be independent prognostic factors for RFS (Table 4).

Bild vergrößern