AAV capsid variants with brain-wide transgene expression and decreased liver targeting after intravenous delivery in mouse and marmoset

Abstract

Genetic intervention is increasingly being explored as a therapeutic option for debilitating disorders of the central nervous system. The safety and efficacy of gene therapies rely upon expressing a transgene in affected cells while minimizing off-target expression. Here we show organ-specific targeting of adeno-associated virus (AAV) capsids after intravenous delivery, which we achieved by employing a Cre-transgenic-based screening platform and sequential engineering of AAV-PHP.eB between the surface-exposed AA452 and AA460 of VP3. From this selection, we identified capsid variants that were enriched in the brain and targeted away from the liver in C57BL/6J mice. This tropism extends to marmoset (Callithrix jacchus), enabling robust, non-invasive gene delivery to the marmoset brain after intravenous administration. Notably, the capsids identified result in distinct transgene expression profiles within the brain, with one exhibiting high specificity to neurons. The ability to cross the blood–brain barrier with neuronal specificity in rodents and non-human primates enables new avenues for basic research and therapeutic possibilities unattainable with naturally occurring serotypes.

Main

The wide array of debilitating disorders affecting the central nervous system (CNS) make it a primary target for the application of novel therapeutic modalities, as evidenced by the rapid development and implementation of gene therapies targeting the brain. AAVs are increasingly used as delivery vehicles owing to their strong clinical safety record, low pathogenicity and stable, non-integrating expression in vivo¹. AAVs were first approved for gene therapy in humans in 2012 to treat lipoprotein lipase deficiency² and have, more recently, been approved to treat spinal muscular atrophy³, retinal dystrophy and hemophilia^4,5. Many more advanced-stage clinical trials using AAVs are underway⁶. Currently, most AAV-based gene therapies rely on naturally occurring serotypes with highly overlapping tropisms⁷, limiting the applicability, efficacy and safety of novel gene therapies. At the same time, the inability to broadly and efficiently target many therapeutically relevant cell populations within organs that have traditionally been refractory to AAV delivery has hindered novel exploratory and therapeutic efforts. These constraints, which are especially prevalent for the CNS, motivated us to enhance AAV efficiency and specificity through directed evolution.

Naturally occurring AAVs have evolved to broadly infect cells⁸, which is a desirable characteristic for the survival of the virus but undesirable for targeting specific cell types. This limitation has typically been addressed by injecting viral vectors directly into the area of interest in the CNS⁹, through either intracranial or intrathecal injection. Although direct injection is a valuable technique for targeting focal cell populations, it requires considerable surgical expertise, and this approach cannot address applications that require broad and uniform area coverage (for example, all cortex or all striatum) or applications that are surgically difficult to access owing to the invasiveness of direct injections (for example, cerebellum and dorsal raphe)^9,10,11. Systemic administration into the bloodstream is an appealing solution to the limitations of direct injections, as gene therapy vectors can be non-invasively delivered throughout the body. However, naturally occurring AAV serotypes tend to target non-CNS tissues, notably the liver, at high levels^7,8. The liver is an immunologically active organ, with large populations of phagocytic cells that play a critical role in immune activation^12,13. Viral targeting to these tissues can trigger immune response, such as liver toxicity^14,15, reducing the safety and limiting the efficacy of systemic injection. Naturally occurring serotypes also have severely limited transduction efficiency in the brain owing to the stringency of the blood–brain barrier, consequently requiring the production of large, high-quality vector titers. Enabling and refining both direct and systemic injections will best equip the community with flexible options to pursue a broader range of neuroscience applications.

To target a precise anatomical region and/or cell type in conjunction with systemic injection, specificity can be obtained or refined through the inclusion of cell-type-specific promoters^16,17,18, enhancer elements^{19,20,21,22,23} and microRNA target sites^24,25 into AAV viral genomes, or viral capsids can be engineered to alter their tissue tropism. Previously, we harnessed the power and specificity of Cre transgenic mice to apply increased selective pressure to viral engineering, leading to variants AAV-PHP.B and AAV-PHP.eB, which cross the blood–brain barrier after intravenous administration and broadly transduce cells throughout the CNS^22,26. These variants transformed the way that the CNS is studied in rodents, opening up completely novel approaches for measuring and affecting brain activity, in both exploratory and translational contexts^18,27,28,29. However, these variants also transduce cells in the liver and other off-target organs. Therefore, the development of AAV variants with decreased liver targeting is particularly important to avoid strong, systemic immune responses^30,31,32.

To achieve specificity with the viral capsid, it is imperative that one applies both positive and negative selective pressure to engineered capsid libraries. The M-CREATE method³³ applies next-generation sequencing (NGS) of synthetic libraries with built-in controls to screen viral variants across multiple Cre transgenic lines for both positive and negative features. Using the M-CREATE method, we selected a panel of novel variants that are both highly enriched in the CNS and targeted away from varying peripheral organs. Transgene expression after delivery with AAV.CAP-B10, described herein, was found to be highly specific for neurons in the CNS, significantly decreased in all peripheral organs assayed and targeted away from the liver in mice. Notably, although AAV-PHP.B failed to translate to non-human primates (NHPs)^34,35, here we show robust transgene expression after intravenous administration of newly engineered variants in the adult marmoset CNS with minimal liver expression compared to AAV9 and AAV-PHP.eB.

Results

Engineering AAV capsids at the three-fold point of symmetry

The most commonly altered position within the AAV capsid is the surface-exposed loop containing amino acid (AA) 588, because it is the site of heparan sulfate binding in AAV2 (ref. ³⁶) and is amenable to peptide display^37,38. The only known receptors for AAV9 are N-linked terminal galactose³⁹ and the AAV receptor^40,41, but the possibility of co-receptors is still unexplored. Binding interactions with cell surface receptors occur near the three-fold axis of symmetry of the viral capsid (Fig. 1a) where the surface-exposed loop containing AA455 of AAV9 is the farthest protruding⁴². Having already engineered AAV-PHP.B and AAV-PHP.eB at the AA588 loop for enhanced CNS transduction^22,26, and with the additional goal of decreasing viral transduction of peripheral organs, we theorized that introducing diversity into the AA455 loop would enhance the interaction with existing mutations of the AA588 loop and refine transduction. Thus, we engineered and performed two rounds of selection with a 7-AA substitution library of the AA455 loop, between AA452 and AA460 (Fig. 1a) in AAV-PHP.eB….