AAV's Golden Jubilee

The paper recently published in Nature by Gardner et al.¹ marks yet another milestone in the development of adeno-associated virus (AAV) as an effective gene delivery vehicle, and is an appropriate way to highlight the fiftieth anniversary of the discovery of AAV. Using simple intramuscular injections in primates, Gardner et al. demonstrated that they can use AAV to express a synthetic hybrid antibody-receptor decoy that has neutralizing capacity for a broad range of currently circulating HIV strains. The injected primates were protected against HIV challenge for up to 10 months. They suggest that this approach may be a novel and effective way of providing immunity to HIV.

AAV was discovered 50 years ago as a contaminant of adenovirus preparations grown on monkey kidney cell cultures.² The viral particles were ~26 nm in diameter (Figure 1), and it was apparent from the start that it was a distinct virus that could be propagated only in the presence of a helper virus, either adenovirus or herpesvirus. Infection of cells with AAV alone produced no cytopathic effect, and no human disease has ever been associated with AAV infections. Early work in cell culture also suggested that AAV could cause a latent infection that could be rescued by infection with the helper virus. It was commonly assumed at the time that latency was achieved by viral integration. Because it caused no disease, AAV received relatively little attention by the medical community and only a handful of laboratories pursued the unusual biology of this single-stranded DNA virus. However, when the discovery of restriction enzymes made cloning possible, the finding that AAV could persist in a latent state and caused no disease made it a viable candidate for development as a gene therapy vector. AAV's requirement for a helper virus for propagation provided an additional layer of safety against potential spread of recombinants in the human population, a major concern at the time.

Figure 1.

Depth-cued atomic structure of adeno-associated virus 1, showing high peaks (red), intermediate elevations (pink) and valleys (yellow) on the capsid surface.

Because AAV did not form viral plaques on its own, the first problem was to develop a genetic system. This was solved by Muzyczka's group,³ who cloned the intact wild-type viral genome into a bacterial plasmid and showed that the plasmid itself was infectious and would produce virus. The infectious plasmid, coupled with the complete DNA sequence of AAV,⁴ allowed investigators to do a detailed genetic mapping of the AAV coding regions and the inverted terminal repeats (ITRs). Two major open reading frames were identified, rep and cap—rep encoding proteins needed for DNA replication and packaging, cap encoding the three capsid proteins. Earlier work had already demonstrated that the ITRs were the origins for DNA replication,⁵ but it was not yet clear what other elements were essential for recombinant virus production, e.g., packaging signals.

The first AAV vectors for long-term transduction were constructed by Hermonat and Muzyczka⁶ and deleted only the capsid-coding region. They demonstrated that they could produce reasonably pure recombinant virus and that the virus could transfer and permanently express a foreign gene in cell culture. A few years later, McLaughlin et al.⁷ went one step further and showed that all of the AAV coding sequences could be replaced with a transgene cassette that contained appropriate transcriptional control signals. These vectors are the so-called single-stranded AAV vectors that are commonly used today for gene therapy. This work also demonstrated that the only elements needed in cis for vector replication, packaging, and chromosome integration were the ITRs. Several groups then analyzed the integration sites, and Berns and his colleagues discovered that when the AAV rep gene is expressed, AAV genomes preferentially integrate into a unique region of human chromosome 19 (ref. 8). However, it was also clear that the frequency of transduction was often low in cell culture.

The picture changed dramatically in the 1990s. Several groups demonstrated that they could achieve long-term (6–12 months) gene expression following injection into airway, liver, muscle, brain, and eye.⁹ This was unheard of at the time. All of the other vector systems appeared to shut off gene expression either through innate or adaptive immune responses or by epigenetic mechanisms. For reasons still not clear, AAV vectors achieved a constant level of expression that persisted. It also became clear that the recombinant genome persisted in animal tissues largely as an episome and had a very low (often undetectable) frequency of integration.^10,11 This meant that the potential for insertional mutagenesis and tumor induction was likely to be low for AAV vectors.

The first clinical trial for AAV was reported in 1996. Flotte et al. transduced the cystic fibrosis transmembrane gene (CFTR) into human airway.¹² They demonstrated expression of CFTR and, more importantly, they saw no adverse events. Since then, AAV vectors have been tested in the clinic for a variety of genetic and idiopathic diseases, using both systemic delivery and delivery to specific organs, including eye, brain, liver, muscle, and lung. To date, most of these trials have shown evidence of gene expression and there have been no adverse events attributed to AAV vectors in human trials. Three recent clinical applications are particularly noteworthy. Several groups have shown significant therapeutic efficacy for the correction of RPE65 deficiency, a type of congenital blindness.¹³ Nathwani et al.¹⁴ showed therapeutic efficacy for the correction of factor IX deficiency. And uniQure, a Dutch company, has recently been awarded approval from the European Commission for an AAV-based therapy for lipoprotein lipase deficiency.¹⁵ This is the first gene therapy drug for a genetic disease that has been approved for human use.

These recent developments have captured the attention of the biotech industry, which relies heavily on delivery of drugs that consist of peptides, antibodies, siRNA, or antisense DNA. All of these can be engineered to be expressed long term via AAV, thus obviating the need for repeated drug applications, something that could be particularly useful in undeveloped countries. Thus, AAV gene therapy has the potential to change drug delivery in a fundamental way. It promises not only drugs for single genetic diseases that were previously intractable, but also new strategies for the treatment of diseases of all kinds. And as shown by the work of Gardner et al., it may also solve difficult vaccine problems.

A major reason for the current optimism is the simplicity of the vector structure. A nonenveloped, T=1 icosahedron, the viral particle consists of just one protein sequence on its surface repeated 60 times to produce a nanoparticle that is smaller than most synthetic nanoparticles. The X-ray crystal structures of most of the currently used serotypes have been solved (Figure 1), and many laboratories have demonstrated that the capsid can be decorated with novel targeting ligands that improve tissue specificity.¹⁶ Moreover, more than 100 new serotypes of AAV have been isolated from primate tissues.¹⁷ Understanding the capsid structure has also allowed many groups to create libraries of AAV-containing surface regions that have been shuffled or randomized.¹⁸ These in turn can be used to select the best surface properties for targeting a particular tissue by a process called directed evolution.

Another major factor in the success of AAV is that it can be made in large quantities and easily purified. A major breakthrough was the development of DNA transfection methods that allowed investigators to make small quantities of pure rAAV that are used today for research purposes.^19,20,21 In addition, scalable production methods for growing rAAV have been developed, using baculovirus or herpesvirus.^22,23 These methods are capable of producing ~10⁵ AAV vector genomes per cell, a level achieved by few other mammalian viruses. And since the particle is essentially a protein complex, all of the conventional high-throughput purification methods can be used to manufacture the final clinical product. Moreover, AAV DNA replication has been completed reconstituted in vitro with purified components,²⁴ and the key genes involved in capsid assembly have been identified.²⁵ It may soon be possible, therefore, to commercially synthesize rAAV in a completely cell-free system.

Finally, it is also worth noting that AAV is also creating a quiet revolution in the basic sciences. Because AAV can be used to quantitatively alter the genotype of somatic cells in a permanent way, it can be used to engineer new animal models. The first example of this was by Kirik et al.,²⁶ who overexpressed α-synuclein in the substantia nigra of rats. Their work produced for the first time a model of Parkinson's disease that showed progressive neurodegeneration of dopamine neurons in the brain, something that was not possible using conventional transgenic animal models. This has led more recently to the use of AAV-based expression of channel rhodopsins in the brain to study optogenetics. This approach promises to be an important tool for unraveling the connections in the brain and perhaps telling us how the most sophisticated organ in the body works.²⁷ By the end of this century, when other types of therapies supplant viral-based gene therapy, the value of AAV as a basic research tool may prove to be its most important contribution.