Potential conflict of interest: Nothing to report.
Hepatitis B virus (HBV) is a major etiological factor of hepatocellular carcinoma (HCC). However, the postoperative prognostic value of the virological factors assayed directly from liver tissue has never been investigated. To address this issue, 185 liver samples obtained from the noncancerous part of surgically removed HBV-associated HCC tissues were subjected to virological analysis. Assayed factors included the amount of HBV-DNA in the liver tissues; genotype; and the presence of the HBV precore stop codon G1896A mutation, basal core promoter A1762T/G1764A mutation, and pre-S deletions/stop codon mutation. All virological factors and clinicopathological factors were subjected to Cox proportional hazard model analysis to estimate postoperative survival. It was found that an HBV-DNA level >3.0 × 107 copies/g of liver tissue and the presence of the basal core promoter mutation independently predicted disease-free (adjusted hazard ratio 1.641 [95% confidence interval (CI) 1.010-2.667] and 2.075 [95% CI 1.203-3.579], respectively) and overall (adjusted hazard ratio 2.807 [95% CI 1.000-7.880] and 5.697 [95% CI 1.678-19.342], respectively) survival. Kaplan-Meier survival analysis indicated that in-frame, short stretch (<100 bp) pre-S deletions, but not large fragment (>100 bp) pre-S deletions, were significantly associated with poorer disease-free (P = 0.005) and overall (P = 0.020) survival. A hot deletion region located between codons 107 and 141 of the pre-S sequence was identified for the short stretch pre-S deletion mutants. Conclusion: The amount of HBV-DNA in liver tissue and the presence of the basal core promoter mutation were two independent predictors for postoperative survival in HCC. A short stretch pre-S deletion located between codons 107 and 141 was strongly associated with a poorer postoperative prognosis. (Hepatology 2010)
Worldwide, hepatocellular carcinoma (HCC) is the fifth most commonly diagnosed solid cancer and the third most common cause of cancer-related death.1 HCC is multifactorial in origin. The three most important causes are chronic hepatitis B virus (HBV) infection, chronic hepatitis C virus infection, and alcoholic liver disease.2, 3 Other risk factors include old age, male sex, underlying chronic liver diseases, and most importantly, liver cirrhosis.4-6 Aflatoxin exposure and diabetes have also been linked to the development of HCC.7
In areas such as Southeast Asia, where chronic hepatitis B continues to be highly prevalent, a correspondingly high incidence of HCC is found.4 Furthermore, despite a successful vaccination program in Taiwan, HBV remains the major etiological factor of HCC in patients over 25 years of age. Consequently, much research has been conducted using HBV as an inroad into understanding HCC itself. To date, several studies have demonstrated that serum HBV-DNA levels as well as other virological factors are closely associated with the development of HCC.8, 9 However, whereas there have been a few studies examining the prognostic value of these serological viral factors in HCC patients after surgical removal of the cancer,10-12 no study has examined the prognostic value of the virological factors assayed directly from liver tissue.
Because of the multifactorial etiology, the key molecular pathways leading to hepatocarcinogenesis remain elusive. Owing to the advance of genomic medicine, it has been found that hepatocarcinogenesis involves not only multiple cascades of molecular events but also heterogeneous cellular pathways.13 To understand the molecular processes associated with tumor cell growth, invasion, and metastasis, researchers have searched for prognostic molecular markers in patients undergoing total resection of HCC. Because no grossly detectable tumor remains after surgical resection, these patients form a relatively homogeneous group. Presumably, the time to recurrence (disease-free survival) or death (overall survival) in these patients reflects the growth behavior of the HCC cells. With the help of effective statistical methods, several molecules capable of predicting postoperative survival have been identified, such as proline-directed protein kinase F(A), MKP-1 (a mitogen-activated protein kinase), vascular endothelial growth factor, proliferating cell nuclear antigen, p53, TA (tissue factor), cytokeratin-19, telomerase activity, and interleukin-10.14-22 Supposedly, these molecules are tightly linked to hepatocarcinogenesis and are therefore candidates for targeted therapy. In HBV-associated HCC, several virological factors have also been demonstrated to regulate cancer cell growth in both cell-based and animal-based studies.23 However, the prognostic value of these virological factors assayed directly from liver tissue has never been examined. In this study, we aim to address this important issue.
The patients included, the methods of HBV-DNA extraction from noncancerous liver tissues, and the methods of statistical analysis are described in the Supporting Information.
The HBV-DNA concentration was quantified using Roche Taqman HBV Monitor (Roche Diagnostics, Basel, Switzerland). The detection limit of this test was 69 copies/mL. In this test, 5.82 copies/mL were equal to 1 IU/mL. The HBV-DNA levels in the noncancerous liver tissue were calculated as copies per gram. HBV genotypes were determined using the restriction fragment length polymorphism method or phylogenetic sequence analysis.24 The methods to detect HBV basal core promoter (BCP) A1762T/G1764A mutations and precore stop codon G1896A mutations have been described.25 To identify pre-S mutations, the pre-S region flanked by P1, 5′-GCGGGTCACCATATTCTTGGGAAC-3′ (nucleotides 2821-2844, sense) and P2, 5′-GAGCAGGGGTCCTAGGAATC-3′ (nucleotides 196-177, antisense) was amplified by way of polymerase chain reaction (PCR). The expected size was 596 bp. The PCR products were subjected to Southern blot analysis to identify positively hybridized bands <500 bp (pre-S deletion mutants with large deleted regions). On the other hand, all PCR products 500-596 bp were gel-purified and subjected to direct sequencing to identify pre-S deletion mutants with small deleted regions (<100 bp). If a mixture of wild-type and deletion mutants was found by way of direct sequencing, the gel-purified PCR product was cloned to pCR2.1-TOPO vector (Invitrogen, Carlsbad, CA) and 10-15 clones were sequenced to identify the deletion mutants. Based on these results, the pre-S sequences were categorized as either (1) pre-S sequences carrying neither deletions nor stop codon mutations (wild-type), (2) pre-S deletion mutants carrying deleted regions >100 bp, (3) pre-S deletion mutants carrying deleted regions <100 bp, or (4) pre-S mutants carrying stop codon mutations.
Of the 185 patients included, 24 were women and 161 were men. Their basic clinical characteristics are listed in Table 1. Notably, the men were significantly older than the women (52.3 ± 13.4 versus 44.6 ± 13.1 years; P = 0.012), and the men had lower alpha-fetoprotein levels than the women (median 56.3 versus 218.8 ng/mL; P = 0.043). Otherwise, no significant differences were observed between men and women with HCC in terms of cirrhosis, tumor number, largest tumor size, ascites, albumin, bilirubin, prothrombin time, creatinine, aspartate aminotransferase (AST), alanine aminotransferase, and alcohol use.
Table 1. Clinical Characteristics of Patients Included in the Study
Female (n = 24)
Male (n = 161)
Data are presented as the mean ± SD unless noted otherwise.
Virological Analysis of Noncancerous Liver Tissues in HCC.
HBV-DNA was extracted from noncancerous liver tissue for virological analysis. It was found that the HBV-DNA level was lower than the detection limit (0.345 × 106 copies per gram of liver tissue) of our quantitative test in 15 patients. Of these 15 patients, seven were also negative for HBV-DNA on PCR assay, whereas eight were positive. Of all patients, 79 had HBV-DNA levels >30 × 106 copies/g.
For the 178 patients in whom HBV-DNA could be assessed by way of PCR, genotype and the presence of the precore stop codon G1896A mutation, BCP A1762T/G1764A mutation, and pre-S mutation were assayed. It was found that 61 (34.3%) patients were genotype C, 119 (66.9%) patients had precore stop codon G1896A mutations, and 103 (57.9%) patients had BCP A1762T/G1764A mutations. Mutation analysis for the pre-S region revealed that 120 (67.4%) patients had no deletions or stop codon mutations in the pre-S region, whereas 42 (23.6%), 11 (6.2%), and 5 (2.8%) patients had large fragment (>100 bp) pre-S deletions, short fragment (<100 bp) pre-S deletions, and stop codon mutations in the pre-S region, respectively.
Clinicopathological and Virological Parameters Associated with Postoperative Survival in HCC.
The Cox proportional hazard model was used to examine the association between clinicopathological and virological factors and disease-free survival after surgical resection of HBV-related HCC (Table 2). Univariate analysis revealed that microvascular invasion, tumor size >3 cm, alpha-fetoprotein >10 ng/mL, ascites, albumin >4 g/dL, prothrombin time >12 seconds, AST >30 U/L, intrahepatic HBV-DNA >30 × 106 copies/g, genotype C, BCP A1762T/G1764A mutation, and pre-S short fragment (<100 bp) deletion were associated with a shorter disease-free survival. After adjusting for other confounding factors, multivariate analysis revealed that alpha-fetoprotein > 10 ng/mL, ascites, prothrombin time > 12 seconds, AST >30 U/L, intrahepatic HBV-DNA >30 × 106 copies/g, and BCP A1762T/G1764A mutation were associated with a shorter disease-free survival, whereas stop codon mutations in the pre-S region were associated with a longer survival.
Table 2. Univariate and Multivariate Analysis of Clinicopathologial and Virological Parameters for Disease-Free Survival in HBV-Associated HCC Patients
Similarly, the Cox proportional hazard model was used to examine the association between clinicopathological and virological factors and overall survival after surgical resection of HBV-related HCC (Table 3). Of all patients included, the cause of death was documented in 29 patients. Liver failure with hepatoencephalopathy and subsequent multiorgan failure occurred in 19 patients (sepsis was documented in 11 of these 19 patients); progressively enlarged HCC with tumor rupture and shock occurred in three patients; lung metastasis and respiratory failure developed in five patients; esophageal varices bleeding with hypovolemic shock occurred in one patient; and severe intracranial hemorrhage occurred in one patient. Univariate analysis revealed that tumor size >3 cm, ascites, bilirubin >1.4 mg/dL, prothrombin time >12 seconds, AST >30 U/L, intrahepatic HBV-DNA >30 × 106 copies/g, and BCP A1762T/G1764A mutation were associated with a shorter overall survival. After adjusting for other confounding factors, multivariate analysis revealed that age >50 years, bilirubin >1.4 mg/dL, intrahepatic HBV-DNA >30 × 106 copies/g, and BCP A1762T/G1764A mutation were associated with a shorter overall survival.
Table 3. Univariate and Multivariate Analysis of Clinicopathologial and Virological Parameters for Overall Survival in HBV-Associated HCC Patients
Combination of Intrahepatic HBV-DNA Level and BCP Mutation to Predict Postoperative Survival.
The Cox proportional hazard model indicated that the amount of HBV-DNA in noncancerous liver tissue and the presence of the BCP A1762T/G1764A mutation were important independent predictors for both disease-free and overall survival. Kaplan-Meier survival analysis was thus performed to understand the predictive value of these two factors when combined. As expected, intrahepatic HBV-DNA levels significantly predicted disease-free (P = 0.002) and overall (P = 0.045) survival. The presence of the BCP A1762T/G1764A mutation also significantly predicted disease-free (P = 0.001) and overall survival (P = 0.012) (Fig. 1).
When the two factors were combined, it was found that 43 (24.2%) patients had lower HBV-DNA levels with no BCP mutations, 32 (18.0%) had higher HBV-DNA levels with no BCP mutations, 56 (31.5%) had lower HBV-DNA levels with BCP mutations, and 47 (26.4%) had higher HBV-DNA levels with BCP mutations. In addition, patients with higher HBV-DNA levels with BCP mutations had significantly shorter postoperative disease-free and overall survival (Fig. 2).
Short Fragment Pre-S Deletion Mutations Predict Poor Postoperative Prognosis.
In this study, pre-S mutations were categorized into three types. It was found that all mutants coexisted with wild-type pre-S sequences (which were without stop codon mutations or deletions). Furthermore, in all patients carrying large fragment (>100 bp) pre-S deletions, multiple forms of the mutation were detected (Fig. 3A). In contrast, in patients carrying short fragment (<100 bp) pre-S deletions, only one form was detected in each patient, though the region of deletion varied among different patients. The Cox proportional hazard model indicated that the 11 patients carrying short fragment (<100 bp) pre-S deletions had very poor disease-free survival (mean 13.1 months [95% confidence interval [CI] 8.0-18.1]). Kaplan-Meier survival analysis was thus performed to understand the predictive value of this form of pre-S deletion (Fig. 3B,C). It was discovered that this group of patients had significantly shorter disease-free (P = 0.005) and overall (P = 0.020) survival.
Codon 107-141 of the Pre-S Sequence Is a Hot Mutation Region in the Short Fragment Pre-S Deletion Mutants.
The sequences of the short fragment pre-S deletion mutants were examined. All deletions were in-frame (Fig. 4). Although all mutants were unique and no two were identical, in eight of them (PSMUT 1-8), the deletions were discovered to be located between codons 107 and 141 of the pre-S region. Furthermore, in PSMUT 9 and 10, there were rearrangements of the pre-S sequences with duplications of codons 86-121 followed by a short stretch deletion, also located between codons 107 and 141. In each patient, the percentages of the short fragment pre-S deletion mutants in the viral population were calculated by sequencing 15 clones derived from each liver sample (93% in patients PSMUT 1, 2, and 5; 87% in patients PSMUT 3 and 4; 33% in patients PSMUT 6-8; 20% in patients PSMUT 9 and 10; and 60% in patient PSMUT 11, respectively).
Comparison of the Virological Parameters Derived from Serum and Liver Samples.
Of the 185 patients, 89 had serum samples stored in the serum bank. These samples were retrieved for virological analysis. Regression analysis indicated that the HBV-DNA levels derived from liver samples were significantly correlated with those from serum samples (regression coefficient 0.0439; 95% CI 0.0391-0.0487; P < 0.001) (Fig. 5). In a subgroup of patients, a large amount of HBV-DNA was found in the liver tissue, but a relatively lower HBV-DNA concentration was detected in the serum sample (Fig. 5, squares [n = 8]), suggesting a defect of viral secretion. In another subgroup of patients, a small amount of HBV-DNA was found in the liver tissue, but a high HBV-DNA concentration was detected in the serum sample (Fig. 5, circles [n = 5]), indicating an extraordinarily high efficiency of viral secretion. Using cases without abnormal secretion efficiency, a regression equation was obtained to convert tissue to serum HBV-DNA levels for practical application (Fig. 5). Sequence analysis for precore stop codon mutation, BCP mutation, and genotype revealed discrepancies between serum-derived and tissue-derived data in 11 patients, including eight and seven discrepancies for precore stop codon mutation and BCP mutation, respectively. Interestingly, of these 11 patients, six were in either of the two abnormal subgroups (Fig. 5, solid squares and circles).
Discrepancies in the detection of large fragment pre-S mutants occurred in 18 patients. The mutants were detected in the serum but not the tissue samples in 11 of them, whereas in the remaining seven patients, the mutants were detected in the tissue but not the serum samples. Five of these seven patients were in the subgroup with secretion defect (Fig. 5, asterisks). Of the 89 patients with available serum samples, short fragment pre-S deletion mutants were detected in the liver tissues in nine of them (PSMUT 1-8 and 11). Of these mutants, seven were also detected in the serum samples (PSMUT 1-5, 7, 8). The remaining two were in the subgroup with secretion defect (Fig. 5, pound symbols).
Despite HCC's multifactorial etiology, HBV remains the major causative factor in Southeast Asia, where HBV is highly prevalent.26 By characterizing the HBV genome sequences derived from patients' serum, several studies have illustrated that various virological factors are closely associated with hepatocarcinogenesis, including serum HBV-DNA concentration, genotype, and the presence of a basal core promoter mutation.27 Furthermore, by performing cell-based and animal-based experiments, several viral proteins have been shown to be implicated in the oncogenesis of liver cancer, including X proteins, large surface proteins, and pre-S deletion mutants.23, 28, 29 Despite convincing evidence provided by these studies, the link between HBV and carcinogenesis has not been carefully investigated by examining the prognostic value of these virological factors in a postoperative model. The connection between a growth-regulating protein and carcinogenesis can be illustrated by demonstrating the prognostic value of its expression level or functional mutation in surgically removed cancer tissues. Candidates for targeted anticancer therapy have been identified with the help of such methods. However, in HBV-associated HCC, only a limited number of studies have focused on this purpose, and only the HBV in the serum samples were used for correlation.10-12 In this study, we assayed the virological factors directly from the noncancerous liver tissue adjacent to surgically removed HCCs. Our data clearly indicate that the viral load of HBV and the presence of BCP mutations were independently associated with postoperative prognosis. Therefore, these two virological factors were not only involved in hepatocarcinogenesis as reported but also affected postoperative prognosis. It was known that HBV-related HCCs could have multiple clonal origins. In such patients, after surgical removal of HCC, the remaining noncancerous part of the liver could experience multiple events of de novo oncogenesis. Therefore, the virological factors were still involved. Additionally, high intrahepatic HBV-DNA levels led to continuous hepatitis activities, resulting in deterioration of liver function and thus poorer overall survival. The present data strongly advocate antiviral therapy in HBV-associated HCC patients after surgical removal of the cancer, especially in the subgroup of patients with the aforementioned prognostic factors.
Comparison of the virological parameters derived from the serum and tissue samples revealed significant variations of viral secretion efficiency in the liver tissues among different patients. Differential secretion efficiency led to alterations of the compositions of viral mutants when they were secreted from hepatocytes to serum. In particular, some pre-S deletion mutants were detected in only the liver tissues, and these patients tended to have a secretion defect (Fig. 5). It is likely that development of pre-S deletion mutants resulted in retention of a large proportion of viruses in the hepatocytes, interfering with the detection of the pre-S mutants in the serum samples. Of the eight patients with secretion defect (Fig. 5, squares), recurrence of HCC was documented in six of them (medium time to recurrence, 10.5 months). In five patients (Fig. 5, circles), high HBV-DNA levels were detected in the serum samples, whereas low levels were found in the liver tissues, suggesting an extraordinarily high efficiency of viral secretion. It remained possible that this observation resulted from local heterogeneity of viral loads in the liver. Clinical analysis revealed that only one of these five patients (serum 65.7 × 106 copies/mL, tissue 51.6 × 106 copies/g) experienced recurrence of HCC (5.1 months after surgery). Taken together, tissue HBV-DNA levels seem to be more reliable for prediction of prognosis. However, clinically it is much easier to obtain serum samples for evaluation and this can be performed before surgery. Thus, they serve as an alternative for prognosis prediction. Furthermore, most patients died of non–tumor-related events, likely related to increased viral replication and progression of liver disease. In this view, serum HBV-DNA might offer a better prediction value, because liver tissue sampling bias could be avoided.
Of note, more patients in this study had genotype B than genotype C, which appears to contradict what previous studies have demonstrated: that genotype C, not genotype B, is associated with HCC. However, among younger patients, genotype B has also been found to be associated with HCC.30 In this study, most of the patients included were under 60 years of age, because patients receiving surgical resections were more likely to be younger. Furthermore, the present data as well as a previous study indicate that genotype C–related HCC is associated with a poorer prognosis and is likely to be more invasive11; this makes it less likely to be associated with resectable HCC, because resectable HCC tends to be less aggressive and is therefore diagnosed at earlier stages. As such, it was not surprising that more of genotype B but not genotype C was found in the patients of this study, seeing as they had resectable HCCs.
In this study, consistent with previous studies, the HBV-DNA level was closely associated with postoperative prognosis, albeit intrahepatic and not serum viral load was measured in this study. Additionally, univariate analysis indicated that genotype C was associated with a poorer prognosis. However, this factor did not appear to be significant in multivariate analysis. It was likely that the contribution of genotype C to the prognosis of HCC was masked by the presence of the BCP mutation, because genotype C was closely associated with the BCP mutation in HBV infection.31 The reason why the BCP mutation was independently linked to hepatocarcinogenesis is not well understood. This could not be explained by better replication efficiency, because the mutation did not appear to be associated with changes in HBV-DNA levels, and one study revealed that HBV with BCP mutations actually had lower promoter activities.32 A possible explanation is that the reduced promoter activities in BCP mutants help with evasion of host immunity during development of liver cancer. Alternatively, the concurrent amino acid substitutions in the X protein might enhance its oncogenicity. However, the latter possibility has not been supported by experimental evidence to date.
Recent studies have shown that pre-S deletion mutants are highly implicated in hepatocarcinogenesis.28, 29, 33 Other studies, however, have demonstrated that pre-S deletions occur often in the stage of chronic active hepatitis, with almost the same frequency as that seen in HCC.34, 35 Furthermore, the deleted regions were distributed randomly in the pre-S region, and no particular patterns could be identified when comparing chronic hepatitis and HCC patients.34, 35 It was therefore questioned whether the pre-S deletion mutants were indeed causatively associated with liver cancer. In the present study, through the use of a postoperative prognostic model, we discovered that only patients with short in-frame deletion/rearrangement pre-S deletion mutants (n = 11) showed a poorer disease-free survival as well as overall survival. However, no significant prognostic value was observed in pre-S sequences harboring deletions >100 bp in length. Furthermore, five patients who carried stop codons in the pre-S region (thus interrupting the expression of the pre-S protein) had a better disease-free prognosis on multivariate analysis, suggesting a functional role of the pre-S proteins in liver cancer. Perhaps the most striking finding of this study was that the short pre-S deletion mutations were all located in a designated region between codons 107 and 141, strongly suggesting that this area was functionally related to the development of cancer. At present, it is unclear whether this is associated with immunological evading mechanisms or perturbations of cellular signaling pathways.
In conclusion, by adopting a postoperative prognostic model, we discovered that HBV viral load and the presence of the BCP A1762T/G1764A mutation were two independent predictors of postoperative survival in HCC. Additionally, a short stretch of pre-S deletion located between codons 107 and 141 was strongly implicated in postoperative prognosis.