The AAV Vector Toolkit: Poised at the Clinical Crossroads

Liver gene transfer: AAV8

Several studies have validated preferential liver transduction in nonhuman primates and dog models by AAV8 vectors. While the absolute transgene expression levels in primate and dog liver have been shown to be lower than mouse models, promising therapeutic indices have been achieved in large animals.^14,15 For instance, intravenous administration of self-complementary AAV8 vectors in the nonhuman primate liver can mediate expression levels of factor IX sufficient for phenotypic correction in hemophilia patients^16,17,18 and sustained correction of disease in a canine model of hemophilia.^19,20,21 Effects of transient immunosuppression on AAV8 liver transduction in nonhuman primates and the ability of proteasomal inhibitors to enhance AAV8 liver transduction in canine models have also been evaluated.^20,22 Moreover, a recent study evaluating intrauterine gene transfer with AAV8 vectors in rhesus macaques yielded robust liver-specific expression of factor IX in injected offspring for up to 2 years.²³ In another recent study, AAV8-mediated hepatic gene transfer in neonatal rhesus macaques was found to be less stable when compared to adolescent animals. Although infant monkeys displayed transgene expression in nearly 98% of hepatocytes within 1 week of administration, a significant loss of transgene expression was observed ~1 month postinjection.²⁴

Taken together with the successful completion of St Jude's phase I clinical trial in hemophilia patients,²⁵ the aforementioned studies corroborate the choice of AAV8 vectors as the lead candidate for factor IX gene transfer in the human liver. These studies also provide a roadmap for evaluating AAV8 vectors in clinical liver gene transfer protocols for treatment of conditions such as hemophilia B, α-1 antitrypsin deficiency, phenylketonuria and lysosomal storage disorders to name a few.

A common observation in clinical and preclinical studies is that pre-existing low levels of NAbs profoundly impact gene transfer efficiency. A recent mechanistic study demonstrated that AAV8 capsid-specific NAb diminished liver deposition of genomes and increased genome distribution to the spleen.^26,27 Three separate studies have established the prevalence of NAbs to AAV8 and other serotypes in the human population. The NAb prevalence for AAV8 was noted to be lowest amongst other serotypes in case of pediatric hemophilia patients (22.6%),²⁸ healthy human subjects (19%)²⁹ and a worldwide epidemiology study.³⁰ Despite this relatively favorable antigenic profile for AAV8 in humans, development of strategies to evade NAbs would make a diverse patient cohort eligible for enrollment in clinical trials. Within this framework, generating Nab-escape AAV8 variants and reengineering antigenic domains on the AAV8 capsid³¹ are likely to yield next generation vector candidates for liver gene transfer. Furthermore, the availability of the AAV8 crystal structure³² should facilitate engineering new lab-derived AAV strains.

Cardiac and musculoskeletal gene transfer: AAV1, AAV6, and AAV9

Intracoronary delivery of AAV1 vectors carrying the SERCA2a gene has been shown to prevent cardiac dysfunction, improve ventricular remodeling and vascular reactivity in a porcine model of heart failure.^33,34 These studies provided the foundation for the current phase I/II clinical trial of gene therapy for cardiac failure with AAV1/SERCA2a vectors.³⁵ In addition to cardiac muscle, regional intravenous delivery of AAV1 vectors to the hind limb of nonhuman primates has also been reported to mediate robust transgene expression similar to AAV8 in different skeletal muscle groups without any signs of immunotoxicity.^36,37,38

Similar to AAV1, coronary infusion of the closely related serotype AAV6 mediates efficient and long-term myocardial gene expression.³⁹ Effective cardiac delivery of shRNA encoding vector genomes by AAV6 has also been demonstrated in canine models.⁴⁰ Furthermore, a recent report utilizing a novel technique for myocardial gene delivery demonstrates robust cardiac gene transfer using AAV6 vectors in a sheep model.⁴¹ Likewise, high transduction efficiencies have also been demonstrated in skeletal muscle of nonhuman primates and a canine muscular dystrophy model following intramuscular administration of AAV1 and AAV6 vectors, respectively.^42,43 An open label, dose-escalation study in hemophilia patients involving intramuscular administration of AAV2 vectors encoding factor IX revealed the need for higher transduction efficiency in muscle to attain a sustained therapeutic effect. More recently, phase I and II clinical trials for gene therapy of α-1 antitrypsin deficiency involving intramuscular administration of AAV1 vectors have been successfully completed thereby supporting the safety and feasibility of this approach.^44,45 Similar application of AAV1 vectors for correction of lipoprotein lipase deficiency through intramuscular administration in human subjects has also been demonstrated.⁴⁶

Early studies with AAV9 vectors in infant rhesus macaques have demonstrated preferential cardiac transduction following intravenous administration in nonhuman primates similar to mouse models.⁴⁷ These studies have now been corroborated by a preclinical study demonstrating efficient cardiac-specific gene transfer following intracoronary administration of AAV9 vectors in a porcine model of post-ischemic heart failure.⁴⁸ However, another study in adult human primates comparing different AAV vectors has shown that AAV9 vectors mediate less efficient cardiac gene transfer than AAV6 in adult nonhuman primates.⁴⁹ That said, an interesting related observation is that in contrast to liver studies, cardiac gene transfer efficiency of different AAV serotypes generally appears to display a lesser extent of cross-species variation. In addition to delivery to the heart, widespread skeletal muscle expression of human mini-dystrophin in a neonatal golden retriever muscular dystrophy model following intravenous administration of AAV9 vectors has been reported.⁵⁰ Taken together, preclinical studies in large animal models strongly argue for the evaluation of AAV serotypes 1, 6, and 9 as lead vector candidates in gene therapy of cardiac and musculoskeletal disease.

CNS gene transfer: AAV5 and AAV9

In the past 5 years, several studies evaluating the transduction efficiency of different AAV serotypes in the primate brain have been reported against the backdrop of phase I clinical trials using AAV2 vectors for gene therapy of Canavan's, Parkinson's, and Batten's disease. For instance, convection-enhanced delivery,^51,52 a dose–response study and long-term assessment have demonstrated clinical improvement in Parkinsonian monkeys with AAV2 vectors.^53,54,55 Convection-enhanced delivery of AAV1 vectors in the nonhuman primate brain revealed transduction of oligodendrocytes and astrocytes in addition to the neuronal population.⁵⁶ In another recent study in adult cynomolgus monkeys involving striatal injection of AAV1, AAV5 and AAV8, AAV5 and AAV1 were superior to AAV8 in transducing neurons in the nonhuman primate striatum.⁵⁷ Such efficient intracerebral gene transfer using AAV5 vectors following intracerebral injection in nonhuman primates has now been corroborated in other preclinical studies.^58,59 Another emerging candidate for central nervous system (CNS) gene delivery is AAVrh.10, a rhesus macaque isolate belonging to clade E (AAV8 family^60,61). This AAV isolate has shown promising results in the rat brain and found suitable for therapeutic CNS gene transfer in a mouse model of Batten disease.⁶²

Recent studies have also focused on cross-species comparison of the ability of AAV9 vectors to traverse the blood–brain barrier in mouse models and nonhuman primates following intravenous administration.^63,64,65 In addition to decreased transduction efficiency in comparison with the mouse brain, a shift in AAV9 tropism from neuronal to glial cells was observed in the monkey brain following IV administration.⁶⁴ These findings underscore the importance of cross-species variation in transduction efficiency and tissue tropism of different AAV serotypes. Furthermore, systemically injected AAV9 in cynomolgus macaques was efficient at crossing the blood–brain barrier, with transgene expression being detected in glial cells throughout the brain and dorsal root ganglia neurons and motor neurons within the spinal cord.⁶⁵ Systemic injection of AAV9 vectors in macaques also results in robust transduction of skeletal muscle and other peripheral organs. As an alternative strategy, restricted gene expression in the primate CNS has be achieved by AAV9 delivery to cerebrospinal fluid, which efficiently targets motor neurons. These strategies provide the rationale for translation of AAV9-mediated gene transfer to patients with CNS-related disorders. Lastly, a recent study comparing the ability of different AAV strains to traverse the blood–brain barrier in mice demonstrates that AAVrh.10 is at least as efficient as AAV9 vectors in CNS gene transfer following systemic administration.⁶⁶

Gene transfer to the eye: AAV4 and AAV8

Over the past 5 years, AAV vectors have clearly emerged as lead candidates for therapeutic gene transfer in a wide range of eye diseases. The safety and efficacy of AAV2 vectors delivering the RPE65 transgene in clinical gene therapy of Leber's congenital amaurosis has been unequivocally established.^{3,4,5,6,7,8,9} An earlier comparison of AAV serotypes 1, 2, and 5 yielded similar transgene expression levels within retinal pigmented epithelium (RPE) as well as photoreceptors upon subretinal administration in a canine model of Leber's congenital amaurosis.⁶⁷ Another study has demonstrated successful restoration of vision in RPE65-deficient Briard dogs using AAV4 vectors, which display selective tropism for the RPE.⁶⁸ Transduction efficiency of AAV4 was shown to be similar to AAV2 vectors in this canine model.

Subretinal administration of AAV8 vectors in canine models has been shown to mediate transgene expression in the RPE, photoreceptors as well as cells of the inner nuclear layer and ganglion cells.⁶⁹ These results suggest that AAV8 vectors might undergo transport along neurons of the visual pathway. Moreover, while no species differences have been noted between rats, dogs and primates with AAV1, AAV2, or AAV5 vectors, AAV8 vectors appear to display greater spread in the canine model. A recent dosage threshold study comparing AAV2 and AAV8 vectors was carried out in a nonhuman primate model.⁷⁰ Although both serotypes demonstrated comparable transduction levels, AAV8 vectors transduced photoreceptors with greater efficiency when compared to AAV2.

In general, the pre-existing NAb response is thought to minimally affect AAV transduction efficiency in immune-privileged sites such as the brain or the eye. Nevertheless, it is noteworthy to mention that a transient increase in anti-AAV2 NAbs was seen in two of three subjects in the Leber's congenital amaurosis clinical trial.⁹ Whether this transient increase in NAbs will impact vector readministration in the same or contralateral eye remains to be seen. The availability of other AAV serotypes with similar transduction efficiency or broader tropism within the eye might serve as an advantage in such a scenario.

Tyrosine-mutant AAV vectors

In an effort to reduce dose-related toxicity of AAV vectors, Zhong, Srivastava and others⁸⁴ have developed a series of next generation tyrosine-mutant vectors that display improved gene transfer efficiency. Briefly, phosphorylation of surface-exposed tyrosine residues on AAV2 capsids is correlated with decreased transduction efficiency and thought to occur due to increased ubiquitination resulting in proteasomal degradation. Mutagenesis of tyrosines into phenylalanines on the AAV2 capsid was demonstrated to bypass this critical barrier to transduction and improved gene transfer efficiency by as much as 30-fold at a log lower vector dose in mice. Subsequent studies in mice have shown that this strategy can be broadly applied to different tissues such as liver, retina, and skeletal muscle and has been expanded to include AAV serotypes such as AAV3, AAV6, AAV8, and AAV9.^{85,86,87,88,89} A recent report has also demonstrated that mutation of capsid surface-exposed serines prone to phosphorylation can enhance transduction in a similar fashion.⁹⁰ These exciting new reagents are poised for preclinical dose-variation studies in large animal models. Within this framework, it is noteworthy to mention that enhanced gene transfer by AAV vectors has been achieved following concurrent administration of the proteasome inhibitor, bortezomib, in a canine model.²² Successful translational studies with tyrosine-mutant vectors would not only corroborate the aforementioned studies, but also enable reduction in vector dose needed for clinical trials.

AAV2.5

The first example of a hybrid AAV vector to proceed to clinical trials is AAV2.5, a rationally engineered AAV strain designed to graft the muscle tropism determinants of AAV1 onto parental AAV2.⁹¹ As shown in preclinical studies and in a phase I clinical trial of Duchenne muscular dystrophy, AAV2.5 is capable of robust gene transfer in skeletal muscle. Consistent with an antigenically distinct profile, this hybrid vector has also been shown to evade NAbs against both AAV1 and AAV2 capsids. While this clinical study highlighted the need to consider T-cell immunity to self and foreign dystrophin epitopes in Duchenne muscular dystrophy patients,⁹² the AAV2.5 vector demonstrated an excellent safety profile and remains a promising vector candidate for clinical gene transfer in musculoskeletal diseases.

AAV6.2

Another interesting subset of hybrid AAV vectors are the singleton vectors developed by Vandenberghe, Wilson and others.⁹³ For instance, the engineered AAV6.2 strain contains a single F129L mutation in the phospholipase A2 domain originally present in the closely related AAV1. The AAV6.2 vector has also been shown to mediate efficient gene transfer in comparison with other related strains within the same clade following intravenous administration in a mouse model.⁹³ More importantly, this hybrid vector outperformed several AAV serotypes in the mouse conducting and nasal airways as well as cultured human airway epithelia.⁹⁴ Functional correction of cystic fibrosis transmembrane regulator expression cultured cystic fibrosis human airway epithelia has also been demonstrated.⁹⁵ Evaluation of AAV6.2 in large animal models is likely to shed more light on the translational potential of this hybrid vector for gene therapy of cystic fibrosis and other lung diseases.

AAV2i8

The UNC Gene Therapy Center has recently developed reengineered AAV vectors demonstrating attenuated liver sequestration following intravenous administration.⁹⁶ Liver-detargeting potential of mutant AAV2 vectors was originally observed as a consequence of mutating heparin-binding arginine residues on the AAV capsid by Kern, Kleinschmidt and others.⁹⁷ While reengineering the heparin-binding footprint of AAV2 with corresponding domains from other AAV serotypes, we generated the hybrid AAV2i8 vector harboring a linear epitope from AAV8. In addition to being detargeted from the murine liver, this hybrid strain displayed the ability to traverse the blood vessel barrier and transduce cardiac and skeletal muscle tissue with high efficiency. Since then, we have successfully carried out isolated limb infusion studies⁹⁸ as well as intravenous administration resulting in robust gene expression in primates (McPhee SW, Asokan A, Tarantal A, Samulski, RJ, unpublished results). Notably, low volume injections of AAV2i8 are significantly more efficient in transducing primate skeletal muscle during isolated limb infusion studies, in contrast to AAV8, which only performs well at high injection volumes. Also noteworthy against the backdrop of AAV2i8 vector development studies are recent clinical studies of transvenous limb perfusion with saline⁹⁹ in muscular dystrophy patients at the UNC Wellstone Center. These studies suggest that high-pressure retrograde transvenous limb perfusion with saline up to 20% of limb volume at above infusion parameters is safe and feasible in adult human muscular dystrophy. Taken together with the aforementioned results, AAV2i8 is a promising lead candidate for correction of musculoskeletal diseases through isolated transvenous limb infusions.

NAb-escape mutants

Amongst the earliest examples of promising vector candidates developed using such combinatorial library approaches are the AAV2-derived mutants, AAV2.15 and AAV2.4.¹⁰⁸ Both these strains harbor mutations at critical antigenic sites, thereby efficiently evading NAbs in human serum. These strains have the potential for immediate translation as candidates for vector readministration in ongoing clinical trials. Another approach involves randomization of previously mapped immunogenic epitopes on the AAV2 capsid to evolve NAb-escape mutants.¹⁰⁹ More recently, a structure-based approach to reengineer surface-exposed antigenic epitopes previously identified using cryo-EM has been proposed.³¹ This strategy was used to engineer novel AAV8 NAb-escape variants suitable for administration in the presence of pre-existing humoral immunity to the AAV8 capsid.

Airway-tropic AAV vectors

As outlined earlier, airway gene transfer faces major hurdles such as the lack of availability of appropriate animal models, poor transduction efficiency due to various physiological barriers and cross-species variation. Keeping these challenges in mind, novel vector candidates for airway gene transfer have been developed by subjecting combinatorial AAV libraries to directed evolution in cultured human airway epithelia. One study yielded an AAV2/AAV5 chimeric harboring a single point mutation (AAV2.5T)¹¹⁰ that was 100-fold more efficient than AAV5 vectors in mediating gene transfer to human airway cultures and rescued chloride ion transport following cystic fibrosis transmembrane regulator gene transfer in diseased human airway cultures. These results are currently being evaluated for translation to a porcine model of cystic fibrosis.^75,76 Another contemporary study yielded two chimeric strains, AAV-HAE1 and AAV-HAE2, containing capsid components derived from AAV1, AAV6 and AAV9 that were more efficient than AAV6 and efficiently corrected the cystic fibrosis defect in diseased human airway cultures.¹¹¹ Both studies corroborate the notion that AAV5, AAV6, and mutants thereof constitute lead candidates for evaluation in large animal models of lung disease.

Muscle-tropic AAV vectors

Directed evolution of the myocardium-tropic AAVM41, a chimeric capsid derived from AAV1, AAV6, AAV7 and AAV8, was recently achieved by Yang, Xiao and others¹¹² at the UNC Gene Therapy Center. AAVM41 was shown to be more efficient than AAV6 in transducing the murine heart. In addition, the mutant also demonstrated remarkably attenuated tropism for liver, skeletal muscle amongst other organs. Efficient rescue of cardiac functions were demonstrated following administration of AAVM41 vectors delivering delta-sarcoglycan in a hamster cardiomyopathy model. The UNC Gene Therapy Center also recently reported a subset of liver-detargeted AAV9 mutant vectors, which allowed efficient gene expression in cardiac and skeletal muscle similar to parental AAV9.¹¹³ The aforementioned vectors, when combined with other transcriptional and translational regulation strategies are likely to yield optimal vector candidates for selective cardiac or musculoskeletal gene transfer. Evaluation of these vectors in canine and primate models is forthcoming.

We would like to thank Drs Arun Srivastava (U Florida), Guangping Gao (U Mass), Krys Bankiewicz (UCSF), Steve Gray, and Nagesh Pulicherla (UNC) for helpful comments. We would also like to acknowledge funding support from NIH for A.A. (HL089221, AI072176); D.V.S. (HL081527), and R.J.S. (AI072176, AI080726, DK084033, U54AR056953).