This study was conducted with two separate populations comprising chronic HBV carriers. To investigate the association between PTEN polymorphism and HCC, a total of 698 patients (development population, 417 chronic HBV carriers without HCC and 281 chronic HBV carriers with HCC) were enrolled in an observational case-control study. These patients had visited the outpatient clinics of the Gastroenterology Department of Ajou University Hospital between March 2002 and December 2008. Chronic HBV carriers were diagnosed as per the positive results for HBsAg and HBV DNA for more than 6 months. The multiphase computed tomography or magnetic resonance imaging of patients diagnosed with HCC showed key imaging features, including size ≥1 cm, arterial phase hyperenhancement, and depending on exact size, a combination of washout, threshold growth, and capsule appearance. Liver biopsy was performed for diagnosis in cases wherein these criteria were absent but HCC or other malignancy was considered probable [13]. In addition, to validate the association between PTEN SNPs and HCC development, we enrolled a retrospective population comprising 200 patients (validation population) with HBV-associated liver cirrhosis. Liver cirrhosis was defined with histological information or ultrasonographic findings associated with cirrhosis (spleen size, >12 cm; portal vein, >16 mm; nodules within the hepatic parenchyma) [14]. These patients had started 0.5 mg entecavir between September 2006 and November 2012 and received entecavir treatment for at least 6 months. Patients were monitored every 3 to 6 months for the development of HCC until February 2015. Patients that developed HCC within 6 months were excluded from the study. All investigations performed in the present study were conducted in accordance with the guidelines of the 1975 Declaration of Helsinki. Informed consent was obtained from all participants, and the study protocol was approved by the Institutional Review Board of Human Research of Ajou University Hospital (GN3-08-030).

Blood samples were stored at -80℃ until analysis. Genomic DNA was purified with G-DEXTM blood genomic DNA purification kit (Intron Biotechnology Inc., Seongnam, South Korea) and quantified using PicoGreen double-stranded DNA quantification reagent as per the standard protocol (Molecular Probes, Eugene, OR, USA). PicoGreen working solution was prepared (1:200 dilution) in Tris-EDTA buffer (pH 8.0). The λ DNA standard and samples were mixed with PicoGreen working solution and incubated at 25℃ for 5 minutes. Plates were read using Victor™ 3 multi-label counter (excitation 480 nm and emission 520 nm; PerkinElmer Inc., Waltham, MA, USA). A standard DNA concentration curve was generated using known λ DNA concentrations.

The validated SNPs in PTEN were identified using a public SNP database (http://www.ncbi.nlm.nih.gov/projects/SNP/). For the development population, we chose six SNPs with minor allele frequency ≥5% and r²≥0.8, as per the literature review. Thus, we analyzed the SNPs at six polymorphic sites, namely, positions rs1234221 (A to C substitution), rs1903860 (T to C substitution), rs1234220 (A to G substitution), rs1903858 (A to G substitution), rs2299941 (A to G substitution), and rs17431184 (T to C substitution) (Table 1). Genotyping was performed using GoldenGate genotyping assay kit according to standard protocols (Illumina Inc., San Diego, CA, USA) as previously described [15]. Signal intensities were extracted and analyzed using BEADSTUDIO software ver. 3.0.22 (Illumina Inc.). For the validation population, we selected three SNPs (rs1903860, rs1903858, and rs2299941) based on the results from the development population.

Statistical analysis was performed using IBM SPSS ver. 22.0 (IBM Corp., Armonk, NY, USA). Maintenance of Hardy–Weinberg equilibrium for each SNP was investigated to determine any deviation in genotype frequency using χ² tests. Genetic models used in association testing were the co-dominant, dominant, and recessive models.

We analyzed the association between genotype and HCC presence using a multiple logistic regression model by controlling for covariates such as age and gender. In this analysis, chronic HBV carriers without HCC were regarded as the control group, while their analysis results were compared with those of HBV-associated patients with HCC. For the validation population, the association between HCC development and genotype was analyzed using multivariate Cox regression analysis.

The presence of a linkage disequilibrium (LD) block of SNPs was confirmed using HAPLOVIEW software (version 4.0; http://www.broad.mit.edu/mpg/haploview). Individual haplotypes were inferred using the EM algorithm with the SAS haplotype procedure. Haplotype analysis was performed using multiple logistic regression. All statistical tests were two-tailed and statistical significance was set at P <0.05. For more precise estimates, statistical computation was used to determine P values of Fisher’s exact test with Monte Carlo simulation (number of iteration=100,000).

Baseline demographic and clinical parameters of development population were shown as chronic HBV carriers without HCC and patients with HBV-associated HCC (Table 2). Patients with HCC were older than chronic HBV carriers without HCC, and men predominated both groups. Clinicopathologic characteristics that estimated the severity of cancer among the enrolled patients with HCC are depicted in Table 2. Serum levels of alpha-fetoprotein greater than 400 ng/mL were detected in approximately one-third of patients. In total, 108 patients (44.5%) had HCC larger than 5 cm in diameter, and 77 patients (31.8%) exhibited vascular invasion. About half of patients (54.2%) presented advanced tumor stage (modified Union for International Cancer Control stage III + IV).

Baseline characteristics of validation population are shown in Table 3. About one-third of the validation population was detected positive for hepatitis B e-antigen, and 20% patients had decompensated cirrhosis. In total, 41 patients from the validation population developed HCC during the course of entecavir therapy. The mean HCC size was 2.8±2.4 cm, while the mean number of HCC was 1.3±0.5.

The observed genotype frequencies of all six SNPs of PTEN in patients with HCC as well as chronic HBV carriers without HCC were in agreement with the Hardy–Weinberg equilibrium (P >0.05). These SNPs were analyzed for LD, and their haplotypes were constructed. From EM algorithm, four common haplotypes, namely, ht1 (A-C-A-G-G), ht2 (A-TA-A-A), ht3 (C-T-A-A-A), ht4 (A-T-G-G-A), and another rare haplotype (ht5) at rs1234221, rs1903860, rs1234220, rs1903858, and rs2299941 positions of PTEN were estimated. The values of LD coefficient (D') ranged from 0.859 to 1.000, and those of r² ranged from 0.046 to 0.811. The most common haplotype was ht1 (A-C-A-G-G) (Fig. 1A). For the validation population, four common haplotypes T-A-A, C-G-G, T-G-A, and G-G-A at rs1903860, rs1903858, and rs2299941 positions were estimated. The values of (D') ranged from 0.986 to 1.000, while those of r² ranged from 0.553 to 0.833 (Fig. 1B).

The relationship between development of HCC and genotype/haplotype frequencies of the six SNPs in PTEN was investigated among chronic HBV carriers (Table 4). PTEN rs1903860 C allele was associated with decreased risk of HCC development (odds ratio [OR]=0.829, 95% confidence interval [CI]=0.428-0.980, P =0.040) in the co-dominant model. PTEN rs1903858 G allele was associated with decreased risk of HCC development in the co-dominant and recessive model (OR=0.534, 95% CI=0.342-0.833, P =0.006 and OR=0.613, 95% CI=0.406-0.927, P =0.020, respectively). An association was also reported between PTEN rs2299941 G allele and decreased risk of HCC development in the co-dominant model (OR=0.658, 95% CI=0.436-0.992, P =0.045). Ht2 (A-T-A-A-A) was associated with increased risk of HCC (OR=2.277, 95% CI=1.144-4.532, P =0.019).

We evaluated the relationship between PTEN rs1903860, rs1903858, and rs2299941 genotypes/haplotypes and HCC development using the validation population including patients with HBV-related liver cirrhosis. Among 200 patients, 41 (20.5%) developed HCC during the follow-up period (median, 58.8 months; range, 6-100 months). One SNP and one haplotype were associated with low risk of HCC development (Table 5). Kaplan-Meier survival estimates revealed the association between patients with PTEN rs2299941 G allele and low risk of HCC development (P =0.069, Fig. 2). Furthermore, multivariate analysis showed that patients with rs2299941 AG or GG genotype exhibited a lower risk of HCC development than those with AA genotype (hazard ratio [HR]=0.582, 95% CI=0.353-0.962, P =0.035). The haplotype containing rs2299941 A allele (haplotype TGA) was also associated with increased risk of HCC development in the multivariate recessive model (HR=8.969, 95% CI=1.146-70.188, P =0.037). However, the number TGA homozygote carriers was too small (only one).