Engineering and Evolution of Synthetic Adeno-Associated Virus (AAV) Gene Therapy Vectors via DNA Family Shuffling

Adeno-associated viral (AAV) vectors represent some of the most potent and promising vehicles for therapeutic human gene transfer due to a unique combination of beneficial properties¹. These include the apathogenicity of the underlying wildtype viruses and the highly advanced methodologies for production of high-titer, high-purity and clinical-grade recombinant vectors². A further particular advantage of the AAV system over other viruses is the availability of a wealth of naturally occurring serotypes which differ in essential properties yet can all be easily engineered as vectors using a common protocol^1,2. Moreover, a number of groups including our own have recently devised strategies to use these natural viruses as templates for the creation of synthetic vectors which either combine the assets of multiple input serotypes, or which enhance the properties of a single isolate. The respective technologies to achieve these goals are either DNA family shuffling³, i.e. fragmentation of various AAV capsid genes followed by their re-assembly based on partial homologies (typically >80% for most AAV serotypes), or peptide display^4,5, i.e. insertion of usually seven amino acids into an exposed loop of the viral capsid where the peptide ideally mediates re-targeting to a desired cell type. For maximum success, both methods are applied in a high-throughput fashion whereby the protocols are up-scaled to yield libraries of around one million distinct capsid variants. Each clone is then comprised of a unique combination of numerous parental viruses (DNA shuffling approach) or contains a distinctive peptide within the same viral backbone (peptide display approach). The subsequent final step is iterative selection of such a library on target cells in order to enrich for individual capsids fulfilling most or ideally all requirements of the selection process. The latter preferably combines positive pressure, such as growth on a certain cell type of interest, with negative selection, for instance elimination of all capsids reacting with anti-AAV antibodies. This combination increases chances that synthetic capsids surviving the selection match the needs of the given application in a manner that would probably not have been found in any naturally occurring AAV isolate. Here, we focus on the DNA family shuffling method as the theoretically and experimentally more challenging of the two technologies. We describe and demonstrate all essential steps for the generation and selection of shuffled AAV libraries (Fig. 1), and then discuss the pitfalls and critical aspects of the protocols that one needs to be aware of in order to succeed with molecular AAV evolution.     

Directed evolution of adeno-associated virus yields enhanced gene delivery vectors

Adeno-associated viral vectors are highly safe and efficient gene delivery vehicles. However, numerous challenges in vector design remain, including neutralizing antibody responses, tissue transport and infection of resistant cell types. Changes must be made to the viral capsid to overcome these problems; however, very often insufficient information is available for rational design of improvements. We therefore applied a directed evolution approach involving the generation of large mutant capsid libraries and selection of adeno-associated virus (AAV) 2 variants with enhanced properties. High-throughput selection processes were designed to isolate mutants within the library with altered affinities for heparin or the ability to evade antibody neutralization and deliver genes more efficiently than wild-type capsid in the presence of anti-AAV serum. This approach, which can be extended to additional gene delivery challenges and serotypes, directs viral evolution to generate 'designer' gene delivery vectors with specified, enhanced properties.     

Combinatorial engineering of a gene therapy vector: directed evolution of adeno-associated virus

BACKGROUND: Viruses are being exploited as vectors to deliver therapeutic genetic information into target cells. The success of this approach will depend on the ability to overcome current limitations, especially in terms of safety and efficiency, through molecular engineering of the viral particles. METHODS: Here we show that in vitro directed evolution can be successfully performed to randomize the viral capsid by error prone PCR and to obtain mutants with improved phenotype. RESULTS: To demonstrate the potential of this technology we selected several adeno-associated virus (AAV) capsid variants that are less efficiently neutralized by human antibodies. These mutations can be used to generate novel vectors for the treatment of patients with pre-existing immunity to AAV. CONCLUSIONS: Our results demonstrate that combinatorial engineering overcomes the limitations of rational design approaches posed by incomplete understanding of the infectious process and at the same time offers a powerful tool to dissect basic viral biology by reverse genetics.     

Improving clinical efficacy of adeno associated vectors by rational capsid bioengineering

Adeno associated vectors (AAV) have shown considerable promise to treat various genetic disorders in both preclinical and clinical settings mainly because of its safety profile. However, efficient use of AAV to deliver genes in immune-competent sites like muscles and liver requires very high doses which are associated with concomitant cellular immune response against the viral capsids leading to destruction of the transduced cells. Coupled with that, there are enough evidences that at high doses, AAV particles are subjected to increased cellular phosphorylation/uniquitination leading to proteasome mediated degradation and loss of the viral particles. The presence of preexisting immunity against AAV further adds on to the problem which is acting as a major roadblock to efficiently use it as a gene therapy vector in the clinics. To overcome this, rational bioengineering of AAV capsid becomes a prime tool by which specific amino acid residue(s) can be suitably modified/replaced by compatible residue(s) to create vectors having lower host immune response and higher intracellular trafficking rate. This article reviews the various aspects of rationally designing AAV capsids like by site-directed mutagenesis, directed evolution and combinatorial libraries which can create vectors having not only immune evasive property but also enhanced gene expression and transduction capability. One or more combinations of these strategies have strong potential to create novel vectors which will have suitable clinical efficiency even at a low dose.     

Peptide ligands incorporated into the threefold spike capsid domain to re-direct gene transduction of AAV8 and AAV9 in vivo

Efficiency and specificity of viral vectors are vital issues in gene therapy. Insertion of peptide ligands into the adeno-associated viral (AAV) capsid at receptor binding sites can re-target AAV2-derived vectors to alternative cell types. Also, the use of serotypes AAV8 and -9 is more efficient than AAV2 for gene transfer to certain tissues in vivo. Consequently, re-targeting of these serotypes by ligand insertion could be a promising approach but has not been explored so far. Here, we generated AAV8 and -9 vectors displaying peptides in the threefold spike capsid domain. These peptides had been selected from peptide libraries displayed on capsids of AAV serotype 2 to optimize systemic gene delivery to murine lung tissue and to breast cancer tissue in PymT transgenic mice (PymT). Such peptide insertions at position 590 of the AAV8 capsid and position 589 of the AAV9 capsid changed the transduction properties of both serotypes. However, both peptides inserted in AAV8 did not result in the same changes of tissue tropism as they did in AAV2. While the AAV2 peptides selected on murine lung tissue did not alter tropism of serotypes 8 and -9, insertion of the AAV2-derived peptide selected on breast cancer tissue augmented tumor gene delivery in both serotypes. Further, this peptide mediated a strong but unspecific in vivo gene transfer for AAV8 and abrogated transduction of various control tissues for AAV9. Our findings indicate that peptide insertion into defined sites of AAV8 and -9 capsids can change and improve their efficiency and specificity compared to their wild type variants and to AAV2, making these insertion sites attractive for the generation of novel targeted vectors in these serotypes.     

Production,Processing, and Characterization of Synthetic AAV Gene Therapy Vectors

Over the last two decades, gene therapy vectors based on wild-type Adeno-associated viruses (AAV) are safe and efficacious in numerous clinical trials and are translated into three approved gene therapy products. Concomitantly, a large body of preclinical work has illustrated the power and potential of engineered synthetic AAV capsids that often excel in terms of an organ or cell specificity, the efficiency of in vitro or in vivo gene transfer, and/or reactivity with anti-AAV immune responses. In turn, this has created a demand for new, scalable, easy-to-implement, and plug-and-play platform processes that are compatible with the rapidly increasing range of AAV capsid variants. Here, the focus is on recent advances in methodologies for downstream processing and characterization of natural or synthetic AAV vectors, comprising different chromatography techniques and thermostability measurements. To illustrate the breadth of this portfolio, two chimeric capsids are used as representative examples that are derived through forward- or backwards-directed molecular evolution, namely, AAV-DJ and Anc80. Collectively, this ever-expanding arsenal of technologies promises to facilitate the development of the next AAV vector generation derived from synthetic capsids and to accelerate their manufacturing, and to thus boost the field of human gene therapy.     

Efficient mouse airway transduction following recombination between AAV vectors carrying parts of a larger gene

The small packaging capacity of adeno-associated virus (AAV) vectors limits the utility of this promising vector system for transfer of large genes. We explored the possibility that larger genes could be reconstituted following homologous recombination between AAV vectors carrying overlapping gene fragments. An alkaline phosphatase (AP) gene was split between two such AAV vectors (rec vectors) and packaged using AAV2 or AAV6 capsid proteins. Rec vectors having either capsid protein recombined to express AP in cultured cells at about 1-2% of the rate observed for an intact vector. Surprisingly, the AAV6 rec vectors transduced lung cells in mice almost as efficiently as did an intact vector, with 10% of airway epithelial cells, the target for treatment of cystic fibrosis (CF), being positive. Thus AAV rec vectors may be useful for diseases such as CF that require transfer of large genes.     

High-throughput screens and selections of enzyme-encoding genes

The availability of vast gene repertoires from both natural sources (genomic and cDNA libraries) and artificial sources (gene libraries) demands the development and application of novel technologies that enable the screening or selection of large libraries for a variety of enzymatic activities. We describe recent developments in the selection of enzyme-coding genes for directed evolution and functional genomics. We focus on HTS approaches that enable selection from large libraries (>10(6) gene variants) with relatively humble means (i.e. non-robotic systems), and on in vitro compartmentalization in particular.     

Adeno-associated virus type 2 (AAV2) capsid-specific cytotoxic T lymphocytes eliminate only vector-transduced cells coexpressing the AAV2 capsid in vivo

A recent clinical trial has suggested that recombinant adeno-associated virus (rAAV) vector transduction in humans induces a cytotoxic T-lymphocyte (CTL) response against the AAV2 capsid. To directly address the ability of AAV capsid-specific CTLs to eliminate rAAV-transduced cells in vitro and in vivo in mice, we first demonstrated that AAV2 capsid-specific CTLs could be induced by dendritic cells with endogenous AAV2 capsid expression or pulsed with AAV2 vectors. These CTLs were able to kill a cell line stable for capsid expression in vitro and also in a mouse tumor xenograft model in vivo. Parent colon carcinoma (CT26) cells transduced with a large amount of AAV2 vectors in vitro were also destroyed by these CTLs. To determine the effect of CTLs on the elimination of target cells transduced by AAV2 vectors in vivo, we carried out adoptive transfer experiments. CTLs eliminated liver cells with endogenous AAV2 capsid expression but not liver cells transduced by AAV2 vectors, regardless of the reporter genes. Similar results were obtained for rAAV2 transduction in muscle. Our data strongly suggest that AAV vector-transduced cells are rarely eliminated by AAV2 capsid-specific CTLs in vivo, even though the AAV capsid can induce a CTL response. In conclusion, AAV capsid-specific CTLs do not appear to play a role in elimination of rAAV-transduced cells in a mouse model. In addition, our data suggest that the mouse model may not mimic the immune response noted in humans and additional modification to AAV vectors may be required for further study in order to elicit a similar cellular immune response.     

Construction of diverse adeno-associated viral libraries for directed evolution of enhanced gene delivery vehicles

Rational design of improved gene delivery vehicles is a challenging and potentially time-consuming process. As an alternative approach, directed evolution can provide a rapid and efficient means for identifying novel proteins with improved function. Here we describe a methodology for generating very large, random adeno-associated viral (AAV) libraries that can be selected for a desired function. First, the AAV2 cap gene is amplified in an error-prone PCR reaction and further diversified through a staggered extension process. The resulting PCR product is then cloned into pSub2 to generate a diverse (>10(6)) AAV2 plasmid library. Finally, the AAV2 plasmid library is used to package a diverse pool of mutant AAV2 virions, such that particles are composed of a mutant AAV genome surrounded by the capsid proteins encoded in that genome, which can be used for functional screening and evolution. This procedure can be performed in approximately 2 weeks.     

A Novel Adeno-Associated Viral Variant for Efficient and Selective Intravitreal Transduction of Rat Müller Cells

Background

The pathologies of numerous retinal degenerative diseases can be attributed to a multitude of genetic factors, and individualized treatment options for afflicted patients are limited and cost-inefficient. In light of the shared neurodegenerative phenotype among these disorders, a safe and broad-based neuroprotective approach would be desirable to overcome these obstacles. As a result, gene delivery of secretable-neuroprotective factors to Müller cells, a type of retinal glia that contacts all classes of retinal neurons, represents an ideal approach to mediate protection of the entire retina through a simple and innocuous intraocular, or intravitreal, injection of an efficient vehicle such as an adeno-associated viral vector (AAV). Although several naturally occurring AAV variants have been isolated with a variety of tropisms, or cellular specificities, these vectors inefficiently infect Müller cells via intravitreal injection.

Methodology/Principal Findings

We have previously applied directed evolution to create several novel AAV variants capable of efficient infection of both rat and human astrocytes through iterative selection of a panel of highly diverse AAV libraries. Here, in vivo and in vitro characterization of these isolated variants identifies a previously unreported AAV variant ShH10, closely related to AAV serotype 6 (AAV6), capable of efficient, selective Müller cell infection through intravitreal injection. Importantly, this new variant shows significantly improved transduction relative to AAV2 (>60%) and AAV6.

Conclusions/Significance

Our findings demonstrate that AAV is a highly versatile vector capable of powerful shifts in tropism from minor sequence changes. This isolated variant represents a new therapeutic vector to treat retinal degenerative diseases through secretion of neuroprotective factors from Müller cells as well as provides new opportunities to study their biological functions in the retina.     

Frustration and Direct-Coupling Analyses to Predict Formation and Function of Adeno-Associated Virus

Adeno-associated virus (AAV) is a promising gene therapy vector because of its efficient gene delivery and relatively mild immunogenicity. To improve delivery target specificity, researchers use combinatorial and rational library design strategies to generate novel AAV capsid variants. These approaches frequently propose high proportions of nonforming or noninfective capsid protein sequences that reduce the effective depth of synthesized vector DNA libraries, thereby raising the discovery cost of novel vectors. We evaluated two computational techniques for their ability to estimate the impact of residue mutations on AAV capsid protein-protein interactions and thus predict changes in vector fitness, reasoning that these approaches might inform the design of functionally enriched AAV libraries and accelerate therapeutic candidate identification. The Frustratometer computes an energy function derived from the energy landscape theory of protein folding. Direct-coupling analysis (DCA) is a statistical framework that captures residue coevolution within proteins. We applied the Frustratometer to select candidate protein residues predicted to favor assembled or disassembled capsid states, then predicted mutation effects at these sites using the Frustratometer and DCA. Capsid mutants were experimentally assessed for changes in virus formation, stability, and transduction ability. The Frustratometer-based metric showed a counterintuitive correlation with viral stability, whereas a DCA-derived metric was highly correlated with virus transduction ability in the small population of residues studied. Our results suggest that coevolutionary models may be able to elucidate complex capsid residue-residue interaction networks essential for viral function, but further study is needed to understand the relationship between protein energy simulations and viral capsid metastability.     

Recombinant adeno-associated viral vectors are deficient in provoking a DNA damage response

Adeno-associated virus type 2 (AAV2) provokes a DNA damage response that mimics a stalled replication fork. We have previously shown that this response is dependent on ataxia telangiectasia-mutated and Rad3-related kinase and involves recruitment of DNA repair proteins into foci associated with AAV2 DNA. Here, we investigated whether recombinant AAV2 (rAAV2) vectors are able to produce a similar response. Surprisingly, the results show that both single-stranded and double-stranded green fluorescent protein-expressing rAAV2 vectors are defective in producing such a response. We show that the DNA damage signaling initiated by AAV2 was not due to the virus-encoded Rep or viral capsid proteins. UV-inactivated AAV2 induced a response similar to that of untreated AAV2. This type of DNA damage response was not provoked by other DNA molecules, such as single-stranded bacteriophage M13 or plasmid DNAs. Rather, the results indicate that the ability of AAV2 to produce a DNA damage response can be attributed to the presence of cis-acting AAV2 DNA sequences, which are absent in rAAV2 vectors and could function as origins of replication creating stalled replication complexes. This hypothesis was tested by using a single-stranded rAAV2 vector containing the p5 AAV2 sequence that has previously been shown to enhance AAV2 replication. This vector was indeed able to trigger DNA damage signaling. These findings support the conclusion that efficient formation of AAV2 replication complexes is required for this AAV2-induced DNA damage response and provide an explanation for the poor response in rAAV2-infected cells.     

Tools for genome-wide strain design and construction

Advances in DNA sequencing and synthesis technologies concurrent with the development of new recombinant DNA approaches have enabled the extension of directed evolution algorithms to the genome-scale. It is now possible to simultaneously map the effect of mutation(s) in each and every gene in the genome onto almost any screenable or selectable phenotype in less than a week. Such maps can be used to direct the design and construction of libraries containing billions of rationally designed combinatorial mutations. Such combinatorial libraries can now also be created and evaluated in less than a week. The review presents and discusses these new technologies within the context of directed evolution and inverse metabolic engineering.     

Infectious Clones and Vectors Derived from Adeno-Associated Virus (AAV) Serotypes Other Than AAV Type 2

Adeno-associated viruses (AAVs) are single-stranded dependent parvoviruses being developed as transducing vectors. Although at least five serotypes exist (AAV types 1 to 5 [AAV1 to -5]), only AAV2, AAV3, and AAV4 have been sequenced, and the vectors in use were almost all derived from AAV2. Here we report the cloning and sequencing of a second AAV3 genome and a new AAV serotype designated AAV6 that is related to AAV1. AAV2, AAV3, and AAV6 were 82% identical at the nucleotide sequence level, and AAV4 was 75 to 78% identical to these AAVs. Significant sequence variation was noted in portions of the capsid proteins that presumably are responsible for serotype-specific functions. Vectors produced from AAV3 and AAV6 differed from AAV2 vectors in host range and serologic reactivity. The AAV3 and AAV6 vector serotypes were able to transduce cells in the presence of serum from animals previously exposed to AAV2 vectors. Our results suggest that vectors based on alternative AAV serotypes will have advantages over existing AAV2 vectors, including the transduction of different cell types, and resistance to neutralizing antibodies against AAV2. This could be especially important for gene therapy, as significant immunity against AAV2 exists in human populations and many protocols will likely require multiple vector doses.     

Mutational analysis of the adeno-associated virus type 2 (AAV2) capsid gene and construction of AAV2 vectors with altered tropism

Adeno-associated virus type 2 (AAV2) has proven to be a valuable vector for gene therapy. Characterization of the functional domains of the AAV capsid proteins can facilitate our understanding of viral tissue tropism, immunoreactivity, viral entry, and DNA packaging, all of which are important issues for generating improved vectors. To obtain a comprehensive genetic map of the AAV capsid gene, we have constructed 93 mutants at 59 different positions in the AAV capsid gene by site-directed mutagenesis. Several types of mutants were studied, including epitope tag or ligand insertion mutants, alanine scanning mutants, and epitope substitution mutants. Analysis of these mutants revealed eight separate phenotypes. Infectious titers of the mutants revealed four classes. Class 1 mutants were viable, class 2 mutants were partially defective, class 3 mutants were temperature sensitive, and class 4 mutants were noninfectious. Further analysis revealed some of the defects in the class 2, 3, and 4 mutants. Among the class 4 mutants, a subset completely abolished capsid formation. These mutants were located predominantly, but not exclusively, in what are likely to be beta-barrel structures in the capsid protein VP3. Two of these mutants were insertions at the N and C termini of VP3, suggesting that both ends of VP3 play a role that is important for capsid assembly or stability. Several class 2 and 3 mutants produced capsids that were unstable during purification of viral particles. One mutant, R432A, made only empty capsids, presumably due to a defect in packaging viral DNA. Additionally, five mutants were defective in heparan binding, a step that is believed to be essential for viral entry. These were distributed into two amino acid clusters in what is likely to be a cell surface loop in the capsid protein VP3. The first cluster spanned amino acids 509 to 522; the second was between amino acids 561 and 591. In addition to the heparan binding clusters, hemagglutinin epitope tag insertions identified several other regions that were on the surface of the capsid. These included insertions at amino acids 1, 34, 138, 266, 447, 591, and 664. Positions 1 and 138 were the N termini of VP1 and VP2, respectively; position 34 was exclusively in VP1; the remaining surface positions were located in putative loop regions of VP3. The remaining mutants, most of them partially defective, were presumably defective in steps of viral entry that were not tested in the preliminary screening, including intracellular trafficking, viral uncoating, or coreceptor binding. Finally, in vitro experiments showed that insertion of the serpin receptor ligand in the N-terminal regions of VP1 or VP2 can change the tropism of AAV. Our results provide information on AAV capsid functional domains and are useful for future design of AAV vectors for targeting of specific tissues.     

Novel caprine adeno-associated virus (AAV) capsid (AAV-Go.1) is closely related to the primate AAV-5 and has unique tropism and neutralization properties

Preexisting humoral immunity to adeno-associated virus (AAV) vectors may limit their clinical utility in gene delivery. We describe a novel caprine AAV (AAV-Go.1) capsid with unique biological properties. AAV-Go.1 capsid was cloned from goat-derived adenovirus preparations. Surprisingly, AAV-Go.1 capsid was 94% identical to the human AAV-5, with differences predicted to be largely on the surface and on or under the spike-like protrusions. In an in vitro neutralization assay using human immunoglobulin G (IgG) (intravenous immune globulin [IVIG]), AAV-Go.1 had higher resistance than AAV-5 (100-fold) and resistance similar to that of AAV-4 or AAV-8. In an in vivo model, SCID mice were pretreated with IVIG to generate normal human IgG plasma levels prior to the administration of AAV human factor IX vectors. Protein expression after intramuscular administration of AAV-Go.1 was unaffected in IVIG-pretreated mice, while it was reduced 5- and 10-fold after administration of AAV-1 and AAV-8, respectively. In contrast, protein expression after intravenous administration of AAV-Go.1 was reduced 7.1-fold, similar to the 3.8-fold reduction observed after AAV-8 administration in IVIG-pretreated mice, and protein expression was essentially extinguished after AAV-2 administration in mice pretreated with much less IVIG (15-fold). AAV-Go.1 vectors also demonstrated a marked tropism for lung when administered intravenously in SCID mice. The pulmonary tropism and high neutralization resistance to human preexisting antibodies suggest novel therapeutic uses for AAV-Go.1 vectors, including targeting diseases such as cystic fibrosis. Nonprimate sources of AAVs may be useful to identify additional capsids with distinct tropisms and high resistance to neutralization by human preexisting antibodies.     

Incremental truncation as a strategy in the engineering of novel biocatalysts

The application and success of combinatorial approaches to protein engineering problems have increased dramatically. However, current directed evolution strategies lack a combinatorial methodology for creating libraries of hybrid enzymes which lack high homology or for creating libraries of highly homologous genes with fusions at regions of non-identity. To create such hybrid enzyme libraries, we have developed a series of combinatorial approaches that utilize the incremental truncation of genes, gene fragments or gene libraries. For incremental truncation, Exonuclease III is used to create a library of all possible single base-pair deletions of a given piece of DNA. Incremental truncation libraries (ITLs) have applications in protein engineering as well as protein folding, enzyme evolution, and the chemical synthesis of proteins. In addition, we are developing a methodology of DNA shuffling which is independent of DNA sequence homology.     

Heparin binding directs activation of T cells against adeno-associated virus serotype 2 capsid

Activation of T cells to the capsid of adeno-associated virus (AAV) serotype 2 vectors has been implicated in liver toxicity in a recent human gene therapy trial of hemophilia B. To further investigate this kind of toxicity, we evaluated T-cell responses to AAV capsids after intramuscular injection of vectors into mice and nonhuman primates. High levels of T cells specific to capsids of vectors based on AAV2 and a phylogenetically related AAV variant were detected. Vectors from other AAV clades such as AAV8 (ref. 3), however, did not lead to activation of capsid-specific T cells. Through the generation of AAV2-AAV8 hybrids and the creation of site-directed mutations, we mapped the domain that directs the activation of T cells to the RXXR motif on VP3, which was previously shown to confer binding of the virion to heparan sulfate proteoglycan (HSPG). Evaluation of natural and engineered AAV variants showed direct correlations between heparin binding, uptake into human dendritic cells (DCs) and activation of capsid-specific T cells. The role of heparin binding in the activation of CD8(+) T cells may be useful in modulating the immunogenicity of antigens and improving the safety profile of existing AAV vectors for gene therapy.