Why does AAV transgene expression decline over time?

One of the leading vectors used for gene therapy remains Adeno Associated Virus (AAV). When AAV transgene expression declines over one to two years, the vector genome is usually still present and the loss is transcriptional. Understanding the mechanisms behind AAV transgene silencing enables more rational vector design and improves the likelihood of achieving durable therapeutic expression.

In NHP liver, expression is biphasic and the late decline tracks loss of active chromatin rather than copy number due to cell division or DNA degradation (Greig et al., Nat Biotechnol 2024). Durability is therefore, in large part, a property of the promoter used, which we noticed surprises people when we tell this.

Populations that divide silence predominantly through CpG methylation of the promoter, while quiescent tissue silences mostly through repressive chromatin, with epigenetic repressors depositing H3K9me3 on histone-associated rAAV genomes. The dominant mechanism, and hence the mitigation, depends on the target tissue, which is why a single durability claim across tissues should be treated with caution.

Although the field is changing, many advanced therapies in the clinic are still driven by viral promoters, but these components have their downsides, too. The dense transcription factor-binding architecture that makes CMV strong, for example, also recruits silencing machinery. Cellular promoters such as PGK and EF1a, and defined human-derived elements, resist silencing better in comparable settings.

When we get approached with a request to develop ‘unsilencable’ promoters, the most common is getting asked to deplete CpGs. Depleting CpG reduces TLR9-driven immune clearance of transduced cells, but CpG is also the methylation substrate. In quiescent tissue, where chromatin silencing dominates, CpG depletion is necessary but not sufficient, and codon optimisation that inflates CpG can work against both durability and immunogenicity, so we always look for a balance between these.

We normally select for architectures less prone to silencing, can incorporate anti-silencing contexts, and build longer-term or multi-passage readouts where durability is the endpoint. No element is silencing-proof, but SuRE™ empirically can test for prolonged activity where our customer needs us to.

Talk to us. If late expression loss is on your risk list, we can make durability a selection criterion.