|Home | About | Journals | Submit | Contact Us | Français|
Through exomic sequencing of ten hepatitis C virus (HCV)-associated hepatocellular carcinomas (HCC) and subsequent evaluation of additional affected individuals, we discovered novel inactivating mutations of ARID2 in four major subtypes of HCC (HCV-associated HCC, hepatitis B virus (HBV)-associated HCC, alcohol-associated HCC and HCC with no known etiology). Notably, 1 8.2% of individuals with HCV-associated HCC in the United States and Europe harbored ARID2 inactivation mutations, suggesting that ARID2 is a tumor suppressor gene that is relatively commonly mutated in this tumor subtype.
With an estimated 748,000 newly diagnosed cases per year, HCC is the third leading cause of cancer-related deaths worldwide1. In the United States, the five-year survival rate of individuals with liver cancer is 11.7%, making it one of the most lethal forms of neoplasia2. Epidemiologic studies have conclusively linked HBV and HCV infections as well as alcohol consumption and aflatoxin B exposure to the development of HCC3. However, whether these etiologic factors are associated with distinct molecular pathways involved in HCC development is largely unknown.
To gain additional insights into the genetic basis of HCC, we determined the sequences of ~18,000 protein-coding genes (the ‘exome’) in the cancers and normal tissues of ten individuals with HCV-associated HCC (Supplementary Table 1 and Supplementary Methods). Massively parallel sequencing of captured DNA resulted in an average depth of coverage of 98-fold of each base in the targeted regions; 94.7% of the bases were represented by at least ten reads (Supplementary Table 2).
Using stringent criteria (Supplementary Methods), we identified 689 potential somatic mutations. Through visual confirmation of the mutant tag sequences of each potential mutation, we discovered 429 non-synonymous somatic mutations in 411 genes in ten HCV-associated HCCs (Supplementary Table 3). The number of mutations per tumor ranged from 17 to 85, with a mean of 42.9 mutations per tumor (Supplementary Fig. 1). We selected five genes that we found to be somatically mutated in more than one tumor in the discovery set for further analysis: CTNNB1 was mutated in four tumors, TP53 was mutated in three tumors, and ARID2, DMXL1 and NLRP1 were each mutated in two tumors. These mutations were all non-synonymous, and we confirmed each by Sanger sequencing (Supplementary Fig. 2a and Supplementary Table 4). CTNNB1 and TP53 mutations have been previously observed in HCC3, but recurrent mutations in the other three genes identified here have not been previously observed in any tumor type.
We evaluated all coding exons of each of these five genes (Supplementary Table 5) in an additional 23 HCV-associated HCCs (Supplementary Table 6). We found that CTNNB1, TP53, ARID2, DMXL1 and NLRP1 were mutated in 8 (24.2%), 4 (12.1%), 6 (18.2%), 2 (6.1%) and 2 (6.1%) of the total 33 HCCs, respectively (Table 1 and Supplementary Table 4).
The nature of the somatic mutations in tumors can generally be used to classify them as oncogenes or as tumor suppressor genes4. All bona fide oncogenes are mutated recurrently either at the same codon or are clustered at a few functionally important codons. Moreover, the mutations in oncogenes are nearly always missense and confer a gain of function, such as constitutive activity of the encoded protein. In contrast, all bona fide tumor suppressor genes are mutated at a variety of positions throughout the coding region. The mutations in tumor suppressor genes result in loss of function and many truncate the encoded proteins through out-of-frame insertions or deletions (indels), nonsense mutations or splice site alterations.
Based on these genetic criteria, the ARID2 mutations were the simplest to interpret: all were predicted to inactivate the encoded protein, unequivocally establishing ARID2 as an HCC tumor suppressor gene.
ARID2 is a subunit of the polybromo- and BRG1-associated factor (PBAF) chromatin remodeling complex, which facilitates ligand-dependent transcriptional activation by nuclear receptors5. ARID2 contains a conservative N-terminal AT-rich DNA interaction (ARID) domain, followed by three LLxxLL motifs and two conservative C-terminal C2H2 Zn-finger motifs, which directly bind to DNA or interact with proteins6. All ARID2 mutations in HCC were predicted to result in polypeptides lacking these intact Zn finger motifs (Table 1 and Supplementary Fig. 2b).
To determine the prevalence of ARID2 mutations in HBV-associated HCC and nonviral HCC, we evaluated ARID2 and the other four genes described above in an additional 106 tumor samples (Supplementary Tables 4 and 7). Several previously unknown mutational patterns emerged from the analyses of these tumors. First, ARID2 mutations were significantly enriched in HCV-associated HCC (14.0%, 6 out of 43 tumors) compared with HBV-associated HCC (2.0%, 1 out of 50 tumors; P = 0.046; Table 2). It is likely that HCV or HBV infections are the major contributor to this difference. However, we cannot rule out the possibility that individuals’ ethnicities, viral subtypes or other environmental factors play a role in determining the selective advantage afforded by ARID2 mutations.
Second, ARID2 mutations were correlated with CTNNB1 mutations and tended to be mutually exclusive with TP53 mutations: of the nine HCC samples with mutations of ARID2, six contained CTNNB1 mutations (P = 0.0022) but none contained TP53 mutations (P = 0.21). Although not statistically significant, mutations of the related chromatin remodeling gene ARID1A and mutations of TP53 have been observed to be mutually exclusive in ovarian carcinomas7,8. Notably, distinct gene expression patterns have been reported in TP53 mutant compared to TP53 wild-type HCC9. Third, the prevalence of TP53 mutations was significantly higher in HCC tumors in individuals from China than in tumors from individuals in the United States or Europe (P < 0.0001; Supplementary Table 8). This difference was unlikely to be caused by aflatoxin B1 exposure10, as the HCC samples were not from individuals living in rural areas of China, and their tumors did not contain the mutation at TP53 codon 249 that is characteristic of aflatoxin B1–induced tumors11. Finally, mutations of CTNNB1 occurred more often in HCV-associated HCC than in HBV-associated HCC (P = 0.018; Table 2), consistent with previous observations12.
Why is ARID2 mutated in HCV-associated HCC? The exact mechanism is unknown. Functional studies have shown that suppression of ARID2 by small interfering RNA reduced both basal and interferon-α–induced IFITM1 (interferon-induced transmembrane protein 1) expression5. It also has been suggested that subverting the function of interferon-α–induced Jak-STAT signaling is important for the lifelong persistence of HCV infection. The HCV core protein can directly bind to the SH2 domain of STAT1 and inhibit interferon-α–induced nuclear import13,14. Thus we hypothesize that the inactivating mutations in ARID2 could repress interferon-α–induced Jak-STAT signaling and thereby provide a selective advantage for chronic HCV propagation during HCC development.
We thank N. Silliman, J. Ptak, L. Dobbyn, J. Schaeffer, M. Whalen, Z. Khan, J. Ma, Z. Wang and R. Mi for expert technical assistance. This work was supported by The Virginia and D.K. Ludwig Fund for Cancer Research and US National Institutes of Health grants CA43460, CA57345, CA62924, CA121113, DK078686, DK080736, DK081417, AACR Stand Up to Cancer Dream Team Translational Cancer Research Grant and National Science and Technology Major Project Grant 2008ZX10002-025.
Accession codes. The ARID2 coding sequence is available in the CCDS database under the accession number CCDS31783.1.
Note: Supplementary information is available on the Nature Genetics website.
AUTHOR CONTRIBUTIONSM.L., B.V., K.W.K., S.Z., M.S.T. and R.H.H. designed the study. M.S.T., J.C., H.Z., S.Z., M.L., L.W., X.Z., L.D.W., R.A.A., M.A.C., T.M.P., H.D.D., R.K., G.J.A.O., R.H.H., V.E.V. and B.V. collected and analyzed the HCC samples. M.L., N.P. and K.W.K. performed genomic sequencing. M.L., K.W.K., B.V. and N.P. analyzed the genetic data. M.L., B.V. and K.W.K. wrote draft manuscripts. All authors contributed to the final version of the paper.
COMPETING FINANCIAL INTERESTS
The authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/naturegenetics/.