Birth of a New Therapeutic Platform: 47 Years of Adeno-associated Virus Biology From Virus Discovery to Licensed Gene Therapy

Within American biomedical research, as supported primarily by the National Institutes of Health (NIH) since the post–World War II era, it has been considered axiomatic that investments in discovery research will ultimately lead to the development of novel approaches to diagnostics and therapeutics that will improve the health and well-being of citizens of the United States and the broader world. I reflect here in a historical context on how one such series of investments in discovery has resulted in the emergence of the first licensed human gene therapy product in Europe, namely Glybera.^1,2,3 Glybera (alipogene tiparvovec) is a recombinant adeno-associated virus (rAAV) vector with AAV2 inverted terminal repeats (ITRs) encapsidated into AAV1 capsids (so-called rAAV2/1, or, for this purpose, rAAV1), which, when administered intramuscularly, can result in expression of sufficient levels of lipoprotein lipase (LPL) to be considered a safe and effective treatment of LPL deficiency. It was licensed in the European Union in November 2012, 47 years after the near simultaneous discovery of AAV in laboratories at the NIH Bethesda campus and at the University of Pittsburgh.^4,5,6,7,8 I trace key milestones in the progress from the discovery of the virus, to its use as a platform for an approved therapeutic product (Figure 1).

Figure 1.

Time line of major milestones in AAV biology and vectorology. The 47 years between the discovery of adeno-associated virus (AAV) and the licensure of Glybera are depicted in three intervals, each approximately 15 years long (1965–1982, 1982–1997, and 1997–2012), with each individually marked segment representing approximately 5 years. IM, intramuscular; ITR, inverted terminal repeat.

The discovery

The emergence of electron microscopy (EM) in the 1960s revolutionized many aspects of biomedical research but perhaps none more so than the discipline of virology. The identification of a simian virus contaminating poliovirus vaccines led to widespread use of EM to examine preparations of partially purified viruses of all types.^9,10 This practice led to the discovery of a number of “satellite viruses” that had the ability to replicate efficiently in the presence of other viruses. The application of this process to adenovirus preparations led two groups—one led by M. David Hoggan at the Laboratory of Infectious Diseases of the National Cancer Institute of the NIH Bethesda campus, the other by Robert W. Atchison of the Department of Epidemiology and Microbiology in the Graduate School of Public Health at the University of Pittsburgh—to identify small DNA viruses (20–25 nm in diameter) contaminating cultures of simian and human adenoviruses.^7,8 These particles were proven to be antigenically distinct from adenoviruses and therefore not simply degradation products of the latter. Furthermore, Atchison deduced that “replication of the particles in cell culture was obtained only when they were inoculated simultaneously with adenoviruses. This suggests that these adenovirus-associated particles behave as defective viruses.”⁸

From that point on, the terms “adenovirus-associated virus” and “adeno-associated virus” were both used for a period of time, with the shorter version ultimately being accepted by the International Committee on Taxonomy of Viruses (ICTV) (http://ictvonline.org). Shortly after their discovery by EM, Neil Blacklow, an infectious disease physician at the NIH working with Hoggan, identified AAV in fresh human tissues from infants and children with adenovirus-induced diarrheal illnesses.^4,5,6 The coinfection by AAV did not substantially alter the clinical course of these infections but showed a trend toward less severe symptoms in coinfected patients. Thus, the concept of AAV being a nonpathogenic virus began to emerge. In fact, a series of subsequent studies showed a tendency of AAV to inhibit the tumorigenicity of other viruses, such as oncogenic adenoviruses, papillomaviruses, and herpesviruses, a property that has yet to be fully explained or explored. Needless to say, the innocuousness of a virus can diminish somewhat the public health demand for its study.^{11,12,13,14,15} Subsequent study of AAV thus focused not on its clinical importance but rather on the fascinating question of how a virus expressing only two major genes could possibly function.

Growth in understanding of the AAV life cycle

In 1969, a team at the National Institute of Allergy and Infectious Diseases demonstrated that AAV is a single-stranded DNA virus, utilizing an isotopic labeling method (Figure 2).¹⁶ One of the authors of that article, Kenneth I. Berns, subsequently made a series of pivotal discoveries with regard to the AAV life cycle, including description of its persistent infection in cultured cells, a hairpin self-priming mechanism for AAV DNA replication, and, later, the ability of AAV2 to preferentially integrate within what is now termed the AAVS1 locus on human chromosome 19 (refs. 17–24). Interest in AAV biology increased as the novel mechanisms of DNA replication (including replication of the terminal repeats), messenger RNA (mRNA) splicing, and translation of the various versions of the AAV Rep and Cap proteins were elucidated.^{25,26,27,28,29,30,31,32,32,33,34,35,36,37,38,39,40,41,42} During this time, the laboratories of Berns at the University of Florida and Cornell Weill Medical College, Nicholas Muzyczka at the University of Florida and Stony Brook University, and Barrie Carter at the National Institutes of Diabetes and Digestive and Kidney Diseases (NIDDK) published many of the notable findings.

Figure 2.

Figure from a 1969 publication on adeno-associated virus (AAV) biology. The illustration depicts an AAV's unusual property of being packaged as single-stranded DNA with equal proportions of plus and minus sense strands. Such studies spurred interest in AAV long before the advent of human gene therapy. Reprinted with permission from ref. 16.

From “not quite a virus” to “quite an amazing vector”

The secondary structure of the AAV ITRs made it technically difficult to clone AAV sequences into a bacterial plasmid. This hurdle was overcome in parallel efforts by the teams of Muzyczka (including R. Jude Samulski) at the University of Florida and Barrie Carter at the NIDDK.^43,44,45 This facilitated the complete DNA sequencing of the AAV2 genome in 1983 by Arun Srivastava,⁴⁶ working in the Berns laboratory at the University of Florida. In addition to making studies of the molecular biology of AAV2 more tractable, the cloning of AAV DNA facilitated the engineering of recombinant AAV2 (rAAV2) vectors, which were likewise nearly simultaneously proven to be viable for gene transfer in mammalian cells in culture by the Muzyczka and Carter laboratories in 1984.^14,47 Suddenly AAV was no longer a novelty but rather a potentially useful tool in the emerging field of gene therapy. Early rAAV vectors were hampered by the contamination of preparations with wild-type AAV, a problem that was overcome by Samulski and Thomas Shenk, with their development of non-overlapping plasmid constructs, including a proviral vector plasmid comprising AAV2-ITRs flanking the gene cargo and a helper plasmid expressing Rep and Cap, in trans.^48,49 The development of this relatively wild-type free system enabled rAAV2 production that was sufficiently efficient for testing in vivo gene transfer in animal models.

Early in vitro use of rAAV and the first use in humans

In the late 1980s and early 1990s, efforts by the Cystic Fibrosis Foundation (CFF) pushed forward the first in vivo gene therapy experiments in animals and humans for both recombinant adenovirus (rAd) and rAAV vectors. In the case of rAAV2, the packaging limit of the vector (approximately 5 kb) limited the size of the promoter cassette that could be used for expression of the cystic fibrosis transmembrane regulator (CFTR), which has a coding sequence of 4.4 kb (ref. 50). The use of the AAV2 p5 promoter and the subsequent discovery of cryptic promoter activity from elements within the AAV2-ITR enabled the packaging of CFTR vectors.⁵¹ These vectors were initially used to overcome the CF defect in airway epithelial cells in culture, and then to express CFTR mRNA and protein in vivo in the rabbit airway after endobronchial instillation with a fiber-optic bronchoscope.^51,52,53 Studies in both cultured airway cells and in the airways of rabbits and rhesus monkeys indicated that rAAV2-CFTR persisted for at least six months in airway cells as an episomal element.^54,55,56 Based on the observed safety in these studies, rAAV2-CFTR was used for the first human gene therapy in November 1995 in the nose and bronchus of an adult with CF.⁵⁷ A series of extensions of that original trial, and follow-up trials in the sinuses of CF patients were also completed.^{58,59,60,61,62,63,64}

Shortly after the report of rAAV2-CFTR gene transfer in the airways, a rAAV2–tyrosine hydroxylase vector was used for in vivo expression in the brains of nonhuman primates as a preclinical approach to therapy for Parkinson's disease, and rAAV2–epo and rAAV2–factor IX vectors were demonstrated to be effective for sustained expression of erythropoietin in muscle to stimulate red blood cell production in anemia and factor IX in liver for hemophilia.^65,66,67,68 The ability of rAAV2 vectors to express factor IX was subsequently demonstrated in patients as well.⁶⁹

Early evidence of bioactivity in humans

Within the first 10 years of AAV use in humans, trials of rAAV2 and rAAV1 vectors were undertaken for CF, hemophilia B, Batten's disease, Canavan's disease, and α₁-antitrypsin (AAT) deficiency.^{57,60,61,62,63,,69,70,71,72,73,74,75,76,77,78,79} In vivo expression at low levels was demonstrated in each of these trials. Vector persistence and expression were directly related to the life span of the transduced cells. Trials in the airways generally showed a duration of expression in the range of 60 days, consistent with the life span of airway epithelial cells, whereas trials in the central nervous system and in the uninjured muscle and liver showed longer-term duration: up to several years in some cases. Throughout this period, data on the safety of rAAV vectors began to accumulate.

Documentation of clinical efficacy in four diseases

The first true breakthrough in clinical efficacy in humans was made in 2008, with the emergence of data from three parallel trials of rAAV2-RPE65 gene therapy in patients with Leber's congenital amaurosis (LCA; Figure 3), an autosomal recessive disease resulting in functional blindness due to an inability to recycle retinoids in the visual cycle.^{80,81,82,83,84,85,86,87,88,89,90} Each of these studies showed that a single subretinal injection of rAAV2-RPE65 resulted in long-term objective improvements in sensitivity to light and subjective improvements in vision. The gene therapy also was safe. Follow-up studies of patients in these early cohorts showed persistent improvements for several years.

Figure 3.

Evidence of efficacy in a human gene therapy trial. The trial used a rAAV2-RPE65 vector to treat patients with the autosomal recessive blinding disorder Leber's congenital amaurosis (LCA). The several-log10-fold change in light sensitivity in the injected portions of the treated retinas (indicated by stars) of three patients (P1, P2, P3) indicates benefit to visual function in these patients. On the opposite side of the panel are the light-sensitivity results from the contralateral eyes (P1, P2, P3) and from an untreated patient (RPE65-LCA). Reprinted with permission from ref. 80.

Shortly after the publication of the LCA data, additional examples of clinical efficacy in rAAV gene therapy trials began to emerge. Hemophilia B trials progressed with the use of an rAAV–factor IX vector delivered systemically to achieve liver delivery.⁹¹ Patients in this trial were treated with oral prednisolone because of an elevation of liver enzymes, but after this treatment, factor IX levels stabilized and patients demonstrated a marked reduction in the need for recombinant protein replacement. Several trials were performed with an rAAV2–aromatic amino acid decarboxylase (rAAV2-AADC) vector in both patients with Parkinson's disease and those with congenital AADC deficiency.^92,93 The clinical effects in infants with AADC deficiency were most remarkable, showing significant improvements in gross motor development, including the ability to walk by 16 months in one patient, a milestone not previously observed in AADC-deficient infants without treatment.

The emergence of Glybera

AAV serotype 2 was the best studied of the AAV serotypes in the preclinical gene therapy period, and it was the first serotype developed into a vector. The ability to cross-package or pseudotype vector genomes with AAV2-ITRs into capsids of other serotypes was demonstrated in the late 1990s.^94,95,96,97 As long as the AAV2-Rep gene was supplied in trans, an AAV2-ITR–flanked genome could be packaged into any other AAV serotype capsid. Studies of the first six AAV serotypes demonstrated that AAV1 capsids were more effective than AAV2 for gene transfer into skeletal muscle, whereas AAV5 and AAV6 capsid vectors were more effective for gene transfer into the murine airway.^95,96,97 This property would later be expanded to encompass more than 100 serotypes and genomic variants of AAV that were found to be naturally occurring in humans and nonhuman primates (Figure 4).^98,99,100

Figure 4.

The biodiversity of primate-derived adeno-associated virus (AAV) genomic variants. The variants (shown within clades here), which were identified in in the landmark 2004 article by Gao et al.,98 may provide many new potential capsid vehicles for therapeutic gene transfer in the future. Reprinted with permission from ref. 98.

The ability of rAAV vectors to express gene products from skeletal muscle had initially been exploited with rAAV2 vectors expressing factor IX or AAT in humans. In the early 2000s, trials of intramuscular rAAV1-AAT and rAAV1-LPL were initiated.^1,2,3,77 The AAT trials were an attempt to replace the second most abundant serum protein in patients with a defect in this gene, and, although persistent transgene expression was readily demonstrated, the dose was insufficient to reach a therapeutic threshold. Meanwhile the rAAV1-LPL vector resulted in a lowering of the serum cholesterol levels in patients with LPL deficiency and ultimately in a reduction in the recurrent pancreatitis that is observed as a consequence of hyperlipidemia in these patients. On the basis of these data, rAAV1-LPL was licensed under the trade name Glybera in November 2012, 47 years after the discovery of AAV and 17 years after the first use of rAAV2-CFTR in humans.

Lessons learned from the AAV story

Apart from being the vector class to become the first licensed gene therapy product in the Western world, the story of AAV provides a number of potentially useful lessons. The first is that the study of AAV throughout more than half of its history was purely for the sake of scientific curiosity. As has been the case with basic science generally, and the stance of NIH more specifically, the early study of AAV was hypothesis-driven basic science rather than goal-directed therapeutic development. A practical use for AAV was not envisioned in any way until the early 1980s. Had curiosity not driven the scientists of the time to study the peculiar life cycle of this nonpathogenic virus, the world might not have learned of this potentially important platform for treatment of human diseases.

The second lesson is more difficult to interpret. It seems ironic that a vector based on a virus discovered on the NIH campus and first made into a vector on the same campus supported by NIH funding, as well as the first one used in animals and humans on NIH-funded projects, would emerge as a licensed therapy in the Netherlands rather than in the United States through the use of a production protocol also developed at the NIH. Did this occur because the regulatory environment in Europe was more open to innovations in gene therapy? Was it because of a more favorable investment environment there? Or was it the determined will of investigators at the sponsoring company, UniQure, or its predecessor, Amsterdam Molecular Therapeutics, that resulted in this outcome? The ongoing history of this field may provide clearer answers to these questions. In the meantime, gene therapy researchers and investors alike should be invigorated by this development and continue to strive to exploit this platform to benefit more patients, with a greater diversity of diseases, who have been waiting so long for this field to mature. In addition, the NIH, as well as the investigators funded by them, should be lauded for their prescience and commitment to basic scientific research.