rAAV Integration and Cancer Risk: What Can We Learn from a Recent Study?

Recombinant adeno-associated virus (rAAV) is generally considered a non-integrating vector. Most vector DNA remains episomal, but rare integration into host DNA does happen. The safety concern is not simply whether integration occurs, but where it occurs and whether it can affect genes involved in cancer.

This question has received particular attention in the liver. Earlier studies in neonatal mice treated with high-dose, liver-directed rAAV found an association between integration at specific sites and hepatocellular carcinoma (HCC). The effect was dose-dependent and raised questions about the long-term consequences of rare rAAV integration.

A recent study in Molecular Therapy looked at this question using a much larger number of integration sites.

What did the researchers do?

The study used AAVLK03.hOTC, a recombinant liver-directed vector developed for ornithine transcarbamylase (OTC) deficiency.

The vector was given intravenously to cynomolgus macaques and to mice with livers repopulated with primary human hepatocytes. Several doses were tested, with 2 × 10¹³ vg/kg as the shared high dose in both models. Most of the integration sites analyzed in the study came from this dose.

The researchers mapped approximately 1.5 million integration sites in human hepatocytes and 1.3 million in macaque hepatocytes. This large dataset allowed them to look more closely at where integration was occurring and whether certain regions were being targeted more often than expected.

What did they find?

Integration was not completely random.

rAAV was found more often in active genes and regions of open chromatin. This is worth paying attention to because these are regions of DNA that are actively being used by the cell.

But the main finding was reassuring. Even with roughly 2.8 million integration sites, the researchers did not see preferential integration near genes associated with HCC. The common integration sites they identified also did not involve known HCC driver genes.

For this particular vector, the results support a low risk of insertional oncogenesis. That conclusion, however, applies to the vector and experimental conditions studied and should not automatically be extended to other rAAV vectors.

There are limitations to both models

The comparison between human and macaque hepatocytes was also informative. The integration patterns were not identical, showing that results from one species may not fully predict what happens in another.

The mice with humanized livers provide actual human hepatocytes, which is a major advantage. But those cells are still growing inside a mouse liver rather than a human liver. The authors point out that this environment can change the behavior of the human cells.

Using both models gives a better picture, but neither can completely reproduce what happens in a treated person.

What about ocular rAAV gene therapy?

This is where the study raises an interesting question for our field.

Ocular gene therapies generally use local delivery and different AAV capsids, so the liver findings cannot simply be applied to the eye. The biology of the target cells is also different. Most retinal and RPE target cells are long-lived and post-mitotic, although they still have regions of open chromatin.

So the question for ocular rAAV is more specific:

Where do rare integrations occur in retinal and RPE cells, and could any of these integrations affect genes involved in abnormal cell growth?

Answering this will likely require looking at integration in the actual target tissues and considering the specific capsid, construct, dose, and route of administration.

The liver study provides reassuring data for AAVLK03.hOTC. It also shows why integration risk needs to be considered in the context of the specific vector and the cells being targeted, rather than treating rAAV integration as the same across different gene therapies.

Reference: Alexander et al., Molecular Therapy (2026)

Read the study