Profiling HBV integrations in hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is the third leading cause of cancer death worldwide. In this context, chronic viral hepatitis B (HBV) infection represents the most common etiology of HCC. Notably, although other common causes of HCC including chronic viral hepatitis C and chronic alcoholic liver disease are mediated by progression through cirrhosis, the pathogenesis of HCC in HBV infection does not entirely depend on this mechanism. A major proposed pathway by which HCC may arise from chronic HBV infection is the integration of HBV DNA into the host genome, resulting in oncogene activation, tumor-suppressor gene inactivation, or other predisposition to chromosomal instability (1).

Since the 1980’s, many studies have detected evidence of HBV DNA integrated not only into normal hepatocytes, but specifically into HCC cells, supporting the hypothesized role of HBV DNA integration in hepatocarcinogenesis. However, early studies observed viral DNA integration in a non-recurrent pattern, meaning that the integration was not found to occur at predictable and reproducible sites in the host genome. It is important to note that these initial studies were limited by technological constraints to the analysis of only certain portions of the host genome, and also constrained by small sample size.

Over the past several years, the new technology of next-generation sequencing (NGS) has resulted in an explosion of studies in which somatic aberrations in cancer cells have been explored with much greater efficiency and accuracy (2). Specifically, NGS has been increasingly used as a powerful means to help pinpoint the genomic events implicated in hepatocarcinogenesis, and to design preventive and therapeutic interventions tailored to target these events (3).

In a recent study, Sung and colleagues used NGS to construct a genome-wide HBV DNA integration map at single-base resolution by analyzing tumor and paired non-tumor liver tissues from 81 HBV-positive and 7 HBV-negative Chinese HCC patients (4). The authors identified 399 integration breakpoints in 75 HBV-positive and 1 HBV-negative patients. They found that viral integration occurred more frequently in tumors (344 events) than in paired non-tumor tissues (55 events). Of the 344 integrations into tumor tissue, 179 were found in known coding genes and were over-represented in the exons and promoter regions of these genes. Furthermore, they identified specific sites of recurrent integration, or “hotspots”, of HBV DNA at breakpoints in the host genome within particular genes including TERT, MLL4, CCNE1, SENP5, and ROCK1. The findings are in line with previous studies using PCR- or hybridization-based techniques showing that HBV integrations commonly reside in TERT on chromosome 5, and MLL4 on chromosome 19 (5-9).

Sung et al. went on to further characterize the consequences of HBV DNA integrations at these breakpoints, and found that mRNA expression levels of TERT, MLL4 and CCNE1 genes were significantly higher in tumor tissues carrying recurrent HBV DNA integrations at these breakpoints. In addition, the authors noted integrations at these sites were associated with dramatically increased copy number variations of the host genes, substantiating the link between HBV DNA integration and chromosomal instability. Finally, the authors reported an association between the total number of all detected integration sites and important clinical characteristics of the patient population including AFP levels, age at HCC diagnosis, and overall survival.

Collectively, these findings confirmed TERT and MLL4 genes as HBV integration hotspots in HCC patients. Moreover, they suggested that recurrent HBV DNA integration into the host genome is an important driving event in HBV-related hepatocarcinogenesis, as well as a potential target for individualized therapies.

Another key finding of the study by Sung et al. is that approximately 40% of the integration breakpoints on the HBV genome occurred within a 1,800-bp range, an insertion hotspot region harboring several essential viral genes. This is consistent with the results of another recent study by Jiang and colleagues who applied high-depth whole genome NGS to sequence tumor and adjacent normal liver tissues from 3 HBV-positive HCC patients and 1 HBV-negative HCC patient (10). Interestingly, although Jiang et al. also confirmed the link between HBV integration and elevated expression of host genes, there was only one shared recurrent integration site (MLL4) between that study and the one by Sung et al. This discrepancy may be partially explained by sample size limitations and individual variations. It may also be accounted for by potential intra-tumor heterogeneity, a major challenge in the design and analysis of current NGS studies, despite the commonly recognized theory of clonal expansion in cancer development (11,12). Neither study reported a detailed procedure of HCC sampling (4,10), and therefore it is possible that differences in sampling could have introduced bias into the findings.

Another difference between the two studies relates to sequencing depth. Sung et al. identified a total of 344 integration sites from 76 HCC tumors through a >30× depth of coverage, whereas Jiang et al. observed 148 integration sites from 3 HCC tumors by a >80× deeper sequencing of samples containing >80% tumor content. Jiang et al. further compared the results from the >80× depth to an even higher coverage of >240×, and found the number of detected integration sites were proportional to sequencing depth. Assuming the actual average numbers of integrations per sample were comparable between the tumors in these two studies, the fact that Jiang et al. identified a much higher number of integrations per sample might be due to the different stringencies in detection criteria used in the two studies, as well as differences in potential sequencing errors, analysis biases and artifacts. Furthermore, it could have reflected the presence of tumor heterogeneity that resulted from the inclusion of normal tissues in the tumor samples and thus reduced the sequencing sensitivity in the study of Sung et al. Further in-depth assessments of allelic bias and number of integration sites that were shared across samples in each study may provide additional clues to these observations.

In addition to constructing HBV integration maps, NGS has been widely applied to survey HCC genomes to identify other aberrations such as mutations, small insertions and deletions, copy number variations, and structural variations. As a result, many frequently mutated genes have been discovered and linked to HCC development, including TP53, AXIN1, ARID1A, ARID2, CTNNB1, as well as others (13-17). However, a causative link between HBV integration sites and specific gene mutations remains to be demonstrated. Moreover, although Sung et al. reported a significant association between the overall numbers of integration sites and patient survival, it is unknown whether this association was particularly prominent in patients with recurrent integrations in hotspot genes. Sung et al. (4) reported sequencing data on the largest HBV-positive HCC patient population to date, which may offer pivotal insights into the mechanism of malignant transformation from HBV infection to HCC. A thorough understanding of HBV DNA integration into the host genome, and the resulting aberrant gene expression and cancer development will likely necessitate even higher-depth sequencing coverage in larger patient populations, in order to tackle the heterogeneity issue and yield statistically robust findings. Due to the rapid decrease in NGS cost and increase in computational capacity, it will not be long before future studies likes Sung’s and Jiang’s pave new roads for researchers to establish a comprehensive understanding of HCC genomes, eventually facilitating the development of novel targeted and personalized therapeutic options to prevent and treat this devastating disease.