AAV2 and Hepatocellular Carcinoma

Kenneth I. Berns, Arun Srivastava, and colleagues are unconvinced by our study describing adeno-associated virus type 2 (AAV2) insertional mutagenesis in cancer driver genes in 11 cases of hepatocellular carcinoma (HCC), published recently in Nature Genetics. ¹ From our results, we concluded that AAV2 insertions contribute to a subset of HCC development in rare patients, mainly in the absence of cirrhosis and other risk factors. In contrast, Berns and his colleagues, deeply involved in the development of AAV vectors in gene therapy, aimed to launch a debate on the potential antioncogenic role of AAV2 infection in humans. Overall, their interpretations of our results are against the general understanding of genomic alterations identified in human cancers that are the result of a positive selection during carcinogenesis and tumor progression. ² Accordingly, AAV2 insertions identified in our study were genomic events selected during the natural history of liver tumor development in humans. We are not dealing with “clones” (as suggested by Berns et al.) but we are describing recurrent alterations selected during clonal expansion of tumor hepatocytes in patients not treated with AAV vector but from the general population.

In contrast to the assessment of Berns and colleagues, antitumor activity of AAV has only been marginally described in studies performed between 1971 and 1991, in nonhuman cell lines. ^{3 –5} No robust evidence has been published and validated in vivo without introducing a “killer gene” in an AAV recombinant vector. ⁶ Whereas some studies have suggested that AAV2 could counteract protumorigenic function of HPV in cellulo, low AAV2 seroprevalence in human cervical carcinomas published in 1976 by Mayor et al. suggesting a protective effect of AAV2 remains to be externally confirmed in a well-defined case–control study with HPV serology. ⁴ Moreover, analyses of AAV2 sequences in cervical preneoplastic lesions are controversial in their interpretation and insertions within the human genome not analyzed with efficient techniques. ⁷ Consequently, the tumor suppressive role of AAV2 in human cervical carcinoma is not supported by strong evidence. In contrast, we observed an increased number of viral chimeric and nonchimeric AAV2 reads in the tumors compared with their corresponding nontumor tissues, suggesting a positive clonal selection of AAV2-infected hepatocytes during malignant transformation in vivo in human liver and arguing against an antitumor effect of AAV2. ¹

The 11 clonal AAV2 insertions identified in HCC targeted 5 different genes, all with an overexpression of the corresponding transcript. Among these genes, four were also previously identified as cancer driver genes with recurrent somatic alterations and similar predicted consequences in HCC without AAV insertion (see Table 1): (1) In TERT promoter, somatic mutations, gene amplifications, and HBV insertions are frequent genomic alterations leading to telomerase overexpression that is identified in 90% of HCC. ^{8 –10} We constructed a functional assay by cloning Firefly luciferase reporter under the control of a wild type or AAV2-inserted TERT promoter, and our experimental results were classically normalized with Renilla-luciferase control, and also validated in an additional cell line. We demonstrated that AAV2 insertion induced an increased promoter activity at a similar level to that induced by hot spot of somatic mutations (at positions −124 and −146), the most frequent oncogenic alterations identified in a large number of solid tumors, including HCC. ^8,11,12 Moreover, this increased TERT promoter activity was specific to the AAV2 sequence per se because it was not observed with insertion of a scramble sequence of equal length and base content as the viral insertion. Consequently, our in cellulo model provided evidence for the role of AAV2 integration in oncogenic TERT activation. (2) Recurrent HBV insertions have been previously identified in CCNA2, CCNE1, and MLL4 strikingly at the same location compared with AAV2 insertions (Nault et al., ¹ Supplementary Fig. 7). ⁹ In our study, we showed a dramatic overexpression of the full-length mature transcript of CCNA2 targeted by AAV2 together with overexpression of immature transcripts. These immature transcripts have a premature ending induced by the use of the viral polyA instead of the human polyA. (3) Finally, MLL4 somatic inactivating mutations were also recurrently observed in HCC, revealing MLL4 as a pivotal driver gene in liver carcinogenesis targeted by mutations and viral integrations. ^9,10 To further our understanding of all the functional consequences of AAV2 insertions but also of spontaneous nucleotide mutations and HBV insertions in these genes, additional functional experiments are required in future projects of research.

Since AAV2 clonal insertions were observed in a rare subtype of HCC, that is, tumors developed in normal liver without known etiology, it is clearly a rare mechanism of carcinogenesis contrasting with the high frequency of AAV infection. However, a high prevalence of infection in the general population associated with a low incidence of specific cancer is a common observation in infections with oncogenic virus. ¹³ Epstein–Barr virus (EBV), the first proven associated virus to cancer, demonstrates a seroprevalence among adults higher than 80% in most populations. Its relation with lymphoma occurrence was identified epidemiologically and demonstrated with the identification of the EBV viral proteins in a large number of different subtypes of lymphoma. ¹⁴ For Merkel cell polyomavirus infection, seroprevalence is ranging from 40% to 80% in the general population, and its relation with Merkel cell carcinoma, a very rare but aggressive skin malignancy, was identified with Merkel cell polyomavirus DNA insertion occurring in 80% of the tumors. ¹⁵ Also, as stated by Moore and Chang, “Another difficulty to understanding viruses in human cancer has been the slow realization that virus infection alone is never sufficient for tumorigenesis, an unsurprising fact that is also true for non-neoplastic viral diseases.” ¹³ Accordingly, other alterations that could cooperate with AAV2 insertion in HCC development remain to be identified.

Finally, we fully disagree with Berns and colleagues, who claimed a protective role for AAV infection after re-interpreting our results, mainly because (1) no robust and recent data in the literature reported evidence of a tumor-suppressive role of AAV2 in human hepatocytes and in human cancers; (2) in our study, AAV2 insertions are clonaly selected in the human tumors, and because they are identified in the tumors, per definition, they have no efficient antitumor effect; (3) somatic AAV2 insertions are very similar to HBV insertions, a well-known cancer-associated virus in the liver; (4) two different mouse models of recombinant AAV-mediated gene therapy developed HCC because of similar mechanisms (insertional mutagenesis of the inverse tandem repeat region, in flip and flop conformation) ^16,17 ; and (5) most of HCC with AAV2 insertion are developed on noncirrhotic liver without etiology.

Concerning potential consequences in gene therapy based on AAV vectors, we recognize that these recombinant viruses differ significantly from their wild-type counterpart; however, as concluded by David Russell and Markus Grompe in their editorial to our article, “Close follow-ups of patients treated with AAV vectors will shed light on some of these issues, and renewed research into the potential oncogenicity of AAV vectors is now more important than ever.” ¹⁸