Transduction of antigen-presenting cells in the brain by AAV9 warrants caution in preclinical studies

To the editor:

Recently, Hinderer et al.¹ published the results of a study in cynomolgus macaques that we believe is deserving of comment. In this study, adeno-associated virus serotype 9 expressing green fluorescent protein (AAV9-GFP) was injected intrathecally into these animals at either the lumbar region of the spine or into cisterna magna. The macaques were then analyzed 2 weeks after injection for expression of GFP in the brain and spinal cord. The authors nicely document significantly stronger transduction of primate brain after cisternal as compared to lumbar infusion, in agreement with our unpublished findings. For this reason, we have reported only cisternal delivery of AAV7 and AAV9 (refs. 2, 3), as the difference in central nervous system expression is so strikingly in favor of cisternal delivery.

In our studies we have documented in both rodents and nonhuman primates a robust immune response against foreign (but not self) antigens like GFP, which was originally derived from the jellyfish Aequoria victoria. AAV9 transduces antigen-presenting cells in nonhuman primate brain³ and liver.⁴ Lest this be seen as a phenomenon specific to GFP, we also showed that human aromatic L-amino acid decarboxylase (AADC) in an AAV9 vector triggers exactly the same kind of cytotoxic responses in rats as GFP does.⁵ To the rodent immune system, human AADC is just as foreign as GFP. In contrast, encoding either human AADC or GFP in the highly neuron-specific AAV2 yields no such cytotoxic response, because neurons are not professional antigen-presenting cells, although they clearly—like nearly all mammalian cells—present antigen via the major histocompatibility complex class I.

In experiments in which AAV9 (and other broader specificity serotypes like AAV5 and AAV7) triggers such immunotoxicity, we see brisk upregulation of the major histocompatibility complex class II on astrocytes and microglia within 2 weeks after vector administration. However, the full immunotoxic effect is not visible for at least a month. Thus, the apparent discrepancy between our published data and the present study can be explained entirely in terms of the acute nature of the experiment performed by Hinderer et al. We agree that 2 weeks would not reveal significant signs of a classic immune response against a foreign antigen, and we caution other investigators that safety studies designed to reveal possible immunotoxicity should extend well beyond 6 weeks. We have settled on a 90-day experimental period for such studies. It should also be noted that this potential problem with central nervous system transduction exists only for the expression of non-self proteins with AAV serotypes that are not neuron-specific. Bacterial proteins such as tetracycline transactivator,⁶ CRISPR,⁷ or channel rhodopsins,⁸ and similar foreign genes must be presumed immunologically guilty before proven innocent.