Adeno-Associated Virus Type 2 and Hepatocellular Carcinoma?

In a recent Nature Genetics letter, entitled “Recurrent AAV2-related insertional mutagenesis in human hepatocellular carcinomas,” Nault and colleagues¹ document that of 193 patients with hepatocellular carcinoma (HCC), 11 contained an integrated genome sequence of the wild-type adeno-associated virus 2 (AAV2), and suggest that AAV2 is associated with oncogenic insertional mutagenesis in human HCC.

Because AAV2 has long been known to be a nonpathogenic human parvovirus² and, in fact, has been shown to possess antitumor activity,^3–6 it is critical that the scientific and clinical implications of these studies be rigorously assessed to justify their conclusions. We have carefully analyzed the data presented by Nault and colleagues¹ and reached a conclusion that is at variance with that of the authors.

The majority of the insertions contained the AAV2 poly(A) site. In the TNFSF10 (TRAIL [tumor necrosis factor-related apoptosis-inducing ligand] gene) cases, AAV2 poly(A) increased TRAIL expression, which may be desirable because TRAIL, which selectively kills cancer cells, has been extensively investigated as a cancer therapy drug. For CCNA2, insertion in clones 2, 3, and 4 in fact generated transcripts that encode nonfunctional CCNA2. For the KMT2B case, the insertion of a 386-bp AAV2 3′-DNA fragment effectively introduced 5 in-frame stop codons in exon 3 of the gene's 37 exons, and thus inactivated this allele. Given that in HCC, this gene is expressed at very high levels, its inactivation might be desirable as well. For one of the CCNA2 insertions (CHC2128T), the authors did not present any RNA data, which makes it difficult to draw any definitive conclusion in this case.

The single case of TERT promoter insertion is also unconvincing. The promoter is taken out of the sequence context. The TERT promoter is usually extremely weak in normal cells, and the insertion is unlikely the cause of TERT activation. Also, the authors did not show how high the TERT activities (RNA counts) are in HCC cells without the AAV2 insertion; they may be on the same level regardless. The authors use DNA-mediated transfection assays in two established human hepatoma cell lines to document a modest 3- to 6-fold increase in TERT promoter activity to justify their claim that insertion of AAV2 sequences was responsible for the genesis of HCC. A more appropriate control would have been to use primary human hepatocytes to establish a causal relationship between AAV integration and TERT gene overexpression. The transfection experiments also did not include a reference reporter gene as an internal control for transfection errors.

PCR positivity of AAV2 sequences in only 7% of HCC tumors, while there was 21% in the adjacent normal liver tissues, further argues against the authors' conclusion, and can be further questioned by their deep sequencing data in only 7 tumors versus 20 normal tissues. Overall, our conclusion is that in the majority of the cases, AAV2 insertion may have either slowed tumor progression or had no effect.

It is important to emphasize that up to 90% of the human population is seropositive for AAV2,⁷ and this stands in stark contrast to an HCC prevalence of fewer than 10 cases of HCC per 100,000 in the United States, a rate that appears to be increasing coincident with the prevalence of hepatitis C virus infection.⁸ It is difficult to explain this if AAV infection truly represents a significant risk for HCC development. Moreover, one of the major limitations of the study by Nault and colleagues¹ is the complete absence of any serological data with reference to AAV2 antibody positivity among case subjects and control subjects. The final limitation is the absence of any information on the possible coinfection by adenovirus and/or herpesvirus in patients with HCC, both because these viruses serve as common natural helper viruses for AAV2 replication, and because both are known to be associated with tumorigenesis themselves. Furthermore, in contrast to established replication-competent oncogenic viruses (e.g. HPV, HCV), there is no evidence of insertion of full-length AAV genomes in AAV-positive HCC. Thus, the study by Nault and colleagues also fails to establish a causal relationship between replication-competent, fully functional AAV and HCC.

In general, AAV has been demonstrated to be an inhibitor rather than an enhancer of carcinogenesis caused by coinfecting viruses. AAV has been shown not only to induce selective apoptosis in cells that lack active p53, but also to inhibit tumor growth in mice.⁹ Furthermore, epidemiological data suggest a protective role for infection by AAV in patients with cervical carcinoma because only 14% of these patients were seropositive for AAV,^10,11 and AAV has been shown to inhibit human papillomavirus type 16, a virus well known to be the etiologic agent of cervical cancer.¹²

Nault and colleagues¹ also mention a few examples of rodent models in which recombinant AAV vectors have been shown to induce HCC.^13–15 However, none of the insertions described in their studies was found in previous studies with mice in which HCC was reported after AAV gene therapy with prohibitively high vector doses. Although this may be a reflection on interspecies differences, it could also be suggested that the risk of hitting these HCC driver genes is reduced with rAAV vectors as opposed to wild-type AAV. Furthermore, data from hundreds of normal mice treated with rAAV vectors showed no evidence of malignancy.^16,17 Similarly, no evidence of cancer has been detected in large animals, such as dogs¹⁸ and nonhuman primates,¹⁹ administered relatively high doses of rAAV vectors. Perhaps most importantly, in 117 phase I/II/III clinical trials carried out to date with rAAV vectors,²⁰ no adverse event, including cancer of any type, has ever been reported.

The conclusions drawn by Nault and colleagues¹ represent an enormous leap from their data. No comment is made about the possible influence of Rep, a key viral element not only for integration but also for its antioncogenic activity, on their findings. Also, it is important to learn where the integrations in normal tissues occurred relative to those seen in tumors. It would also be useful to know which other changes were seen in the sequences of the genes (introns) of the patients with HCC but without predisposing factors. Furthermore, natural infections with wild-type AAV2 have no demonstrable connection with administration of current rAAV vectors. The ∼5% incidence of HCC associated with AAV2 integration appears to be a secondary event after the initiation of tumorigenesis.

In summary, it is difficult to draw any definitive conclusions from the data presented by Nault and colleagues,¹ given the complexity of the cause-and-effect relationship. Perhaps, time will tell. On the contrary, viewed in a more positive light, much like cervical carcinoma, AAV infection might indeed be a key factor in preventing HCC in humans.

Note Added in Proof

While our article was being reviewed, Büning and Manfred similarly challenged the conclusions by Nault and colleagues in an Editorial entitled: “Adeno-associated Vector Toxicity—To Be or Not to Be?”, which was published in the November 25, 2015 issue of Molecular Therapy (Molecular Therapy, 23: 1673–1675, 2015).

Author Disclosure

B.J.B., W.W.H., and N.M. own equity in, and W.W.H. is a consultant for, Applied Genetic Technologies Corporation; T.R.F. is advisor to Dimension Therapeutics and Editas Medicine; G.G. is a founder of, and holds equity in, Voyager Therapeutics; and X.X. owns equity in Asklepios BioPharmaceutical, Inc. These companies may in the future commercialize some aspects of AAV vector-mediated gene therapy. T.V. serves as key opinion leader for various companies with a vested interest in hemophilia gene therapy. The other authors declare no competing financial interests.