Wild-type AAV Insertions in Hepatocellular Carcinoma Do Not Inform Debate Over Genotoxicity Risk of Vectorized AAV

To the editor:

In a recent editorial in Molecular Therapy, Büning and Schmidt¹ discussed the role of gene therapy vectors based on adeno-associated viruses (AAVs) in the development of liver tumors following the identification by Nault et al.² of recurrent AAV2 clonal insertions in 11 cases of hepatocellular carcinoma (HCC). In this editorial, the authors appear to confound the issue of AAV2 insertion in HCC development with the risk of AAV2-based vector genotoxicity. Here, we aim to clarify our interpretation of the occurrence of AAV2 insertions in spontaneous human HCC, which are clearly unrelated to vectorized AAV used in gene therapy.

Büning and Schmidt asked whether AAV2 insertions observed in human HCC could be passenger or cancer driver events. By analyzing 193 specimens of HCC, we identified clonal AAV2 insertions, all occurring in genes (TERT, CCNE1, KMT2B, TNFSF10, and CCNA2) known to be involved in cancer development. Moreover, each of these insertions was associated with overexpression of the corresponding targeted gene. These results showed that the AAV2 insertions are recurrent clonal genomic alterations selected during the natural history of HCC development. In agreement with the general multistep molecular mechanism observed in solid tumor development, other mutations accumulated in additional cancer driver genes. In addition, we showed by functional analyses that AAV2 insertions in either the TERT promoter or the TNFSF10 3ʹ-UTR increased the promoter activity and luciferase expression in cellulo. All of these results showed that AAV2 insertion met the criteria of a cancer driver event selected during tumor development in humans as defined by molecular oncologists.³

The authors suggested that the small size of the inserted viral sequences and the occurrence of these insertions in either intronic or exonic coding or noncoding human sequences argued against a genotoxic effect of AAV2. However, similar viral oncogenic insertions have been identified in HCC involving small parts of the hepatitis B virus (HBV) genome. Both viruses have the ability to insert into the hepatocyte genome and to induce insertional mutagenesis distributed over the human genome, introducing enhancer or promoter sequences; they could also stabilize targeted messenger RNA or disrupt coding sequences. Interestingly, recurrent insertions of AAV2 and HBV target the same loci in the promoters of the telomerase (TERT), cyclins (CCNA2 and CCNE1) and MLL4 genes.^4,5 Clonal AAV2 insertion in human HCC is a very rare event, and oncogenic insertional mutagenesis by the latter is significantly less frequent than that induced by insertions of HBV, taking into account the large fraction of the population who are infected by AAV2 during their lifetime.

Büning and Schmidt were surprised by the lack of recurrent AAV2 insertions observed at the AAVS1 region, a 4-kilobase locus at chromosome 19q13.42 previously described as a hot spot for AAV insertions.⁶ However, recurrent insertions at AAVS1 were identified in cell culture and mainly in HeLa cells, a highly proliferative cervical carcinoma–derived cell line presenting HPV insertions and a large number of chromosome rearrangements, including chromothripsis at chromosome 19. Moreover, this hot spot of insertion was identified using PCR-based techniques, whereas we used a DNA capture strategy unbiased by DNA amplification to isolate AAV2 insertions in liver tissues. In 27 human liver tissues, clonal and nonclonal AAV2 insertion events were distributed all over the human genome without clustering within AAVS1. This is the result of the natural history of AAV2 infection in vivo in adult human hepatocytes. Altogether, these results suggest that artificial infection of proliferative cell lines induces a different insertional profile compared to what occurs during natural AAV infection in humans. Moreover, a recent analysis of integration profiles of wild-type AAV2 (AAVwt) in diploid human fibroblasts showed that genomic hot spots differed from those described previously in aneuploid HeLa cells. In particular, the AAVS1 hot spot on chromosome 19q13.42 was targeted by 2.5% of all AAVwt integrations in diploid human fibroblasts, whereas up to 45% of all integrations were detected there in HeLa cells. Genomic hot spot patterns of AAVwt integration are cell type-dependent and correlate with variations in chromatin accessibility.⁷

A significant portion of the native AAV genome is deleted during construction of AAV-based vectors for gene therapy. However, AAV-derived vectors are also able to insert into the human genome. Our study did not aim to challenge the safety of AAV vectors in gene therapy. Carcinogenesis is a slow process that involves many driver genes and takes place years after an exposure to toxins or viruses. Because no liver tumors have been observed to date in patients treated using AAV vectors for gene therapy, the risk of HCC is probably very low, if it exists at all. However, we must consider the two independent mouse models of AAV vectorization that developed HCC through clonal oncogenic insertion.^7,8 Finally, as for other viral vectors with potential for insertion into the human genome, patients should be followed longitudinally to monitor long-term effects that take into account the slow evolution of hepatocytes initiated for malignant transformation and the latency before tumor diagnosis.