Engineering Liver-detargeted AAV9 Vectors for Cardiac and Musculoskeletal Gene Transfer

Introduction

Clinical gene transfer with adeno-associated viral (AAV) vectors has rapidly gained momentum in recent years. The ongoing translation from bench-to-bedside has been spurred in part by the availability of a versatile AAV toolkit displaying diverse tissue tropisms across multiple species.^1,2 Among numerous AAV isolates, AAV9 vectors display a systemic, multiorgan transduction profile following intravenous administration.³ Rapid onset of gene expression and high transgene expression levels mediated by AAV9 vectors in heart and liver have been reported.^4,5,6 In addition, efficient transduction of neurons in neonatal mice as well as skeletal muscle in neonatal dogs following intravascular administration has been observed.^7,8 These attributes make AAV9 a viable candidate for therapeutic gene transfer in systemic diseases such as lysosomal storage disorders.

Paradoxically, a wide range of clinically relevant applications require vector targeting to specific tissues rather than multiorgan gene expression. For instance, gene therapy of cardiac disease or muscular dystrophies requires vectors capable of efficient and selective gene transfer to heart and/or skeletal muscle.^6,8,9 Therapeutic approaches targeting the liver or skeletal muscle are preferred for treatment of hemophilic disorders,¹⁰ while the lung is considered a target organ for gene therapy of α-1 antitrypsin deficiency.^11,12 Tissue-specific transgene expression has been achieved through different approaches. For instance, at the vector genome level, transcriptional targeting elements such as tissue-specific promoters^13,14 and microRNA-mediated gene regulation¹⁵ have been developed. When used in conjunction with the appropriate AAV serotype or variant, the latter strategies can provide exquisite control over gene expression patterns in different tissues following systemic injection.

Attempts to develop tissue-targeted AAV vectors have included insertion of targeting motifs into the GH loop of VP3 subunits of different AAV capsid templates or directed evolution of tissue-tropic AAV variants through DNA shuffling.^16,17,18,19 These efforts, in conjunction with alanine scanning mutagenesis studies have been instrumental in implicating the icosahedral threefold symmetry axis of AAV as a key determinant in the recognition of host cell surface receptors.^2,20,21,22 This highly variable trimer region, formed by interdigitation of GH loops from three VP3 subunits is thought to influence diverse tissue tropisms displayed by different AAV strains.^{23,24,25,26,27} The current study hinges on re-engineering GH loop residues on the AAV9 capsid surface. Using a combination of sequence analysis, structural models, and in vivo screening, we identified several functionally diverse AAV9 capsid variants. Specifically, we report the discovery of novel AAV9 variants displaying attenuated liver tropism and with significant potential for therapeutic gene transfer in the treatment of musculoskeletal and cardiac diseases.

Results

Discussion

Robust systemic transduction following intravenous administration is a hallmark of AAV9 vectors.^3,4,6 Several studies have highlighted the ability of this serotype for efficient cardiac gene transfer as well as the ability to cross the blood brain barrier.^5,28 These attributes make AAV9 a viable candidate for a “top-down” capsid engineering approach that would not only separate individual traits outlined above, but also potentially yield variants with restricted tissue tropisms. Based on aforementioned rationale and guided by structural cues available for other AAV serotypes,^{23,24,26,29,30} we generated a randomized AAV9 capsid library with mutations clustered within the VP3 GH loop region which assembles the protrusions that surround the icosahedral threefold axis. We then developed a comprehensive screening approach involving sequence analysis, molecular modeling, and in vivo studies to identify variants with novel functional phenotypes.

A critical finding in the current study is the discovery of mutations that result in defective phenotypes. Subtype I includes two variants, AAV9.11 and 9.47 with distinct defects. Mutant 9.11 carries two mutations, T568P and Q590L, which results in a transduction-deficient phenotype. The T568P mutation is buried within the AAV9 trimer (model not shown) and does not appear to impact capsid assembly or packaging efficiency as indicated by viral titers (Supplementary Table S2). The Q590L mutation is located within the variable region VIII described by Govindasamy et al.²⁴ Other variants carrying mutations in the AAV9 inner loop region (residues 590–595, VP1 numbering) include AAV9.9 (W595C) and AAV9.16 (Q592L), which display a modest decrease in transduction efficiency within the liver in comparison with the parental AAV9. Residues within the corresponding inner loop region on AAV2 have previously been implicated in heparan sulfate binding.^20,21 In corollary, it is likely that mutation of one or more residues within the 590–595 region might alter interaction of AAV9 with its cognate receptor. Variant AAV9.47 displayed a defective biodistribution profile, which in turn adversely affected transduction efficiency. Consistent with this defective profile, AAV9.47 is rapidly eliminated from blood circulation when compared to AAV9, AAV9.45, and AAV9.68 vectors (Supplementary Figure S1). One plausible explanation for the latter phenotype is the contribution of G453D and K557E mutations to negatively charged clusters on the AAV9.47 surface (Figure 1e), thereby making the capsid prone to rapid blood clearance. A similar decrease in circulation half-life has been observed upon masking positively charged residues or addition of negatively charged moieties to the surface of Qbeta and M13 bacteriophage particles.^31,32

In the current study, we have identified novel liver-detargeted variants of AAV9 using comprehensive structural analysis and in vivo screening. Similar variants of other AAV serotypes developed using rational or combinatorial mutagenesis have been reported in the literature. For instance, altering residues within the heparan sulfate footprint on the AAV2 capsid has yielded liver-detargeted variants.^27,33 Yang et al. (2009)⁶ recently identified a myocardium-tropic AAVM41 that showed gene transfer efficiency similar to AAV9 in the heart, but 80-fold decrease in gene expression as well as tenfold lower vector genome copy numbers in the liver. Although no contributing domains from AAV9 were observed, the AAVM41 capsid subunit sequence revealed critical domains derived from AAV serotypes 1, 6, and 7 that appear to confer the cardiotropic phenotype. The current study revealed the critical role played by specific amino acid residues in conferring liver tropism to AAV9 vectors. Specifically, subtype II variants 9.24, 9.45, and 9.61 appear to possess mutations (at positions N498 and W503) that cluster within variable region V previously described by Govindasamy et al.²⁴ Further, AAV9.68, a subtype III variant containing a P504T mutation showed preferential liver tropism as demonstrated by a decrease in heart-to-liver transgene expression ratio when compared to AAV9. This region (residues 498–504) is located behind the inner loop residues 590–595 at the threefold symmetry axis (Figure 1e) and has previously been shown to play a critical role in transduction for AAV2.³⁴ In addition, a region including these amino acid residues was also implicated in laminin receptor recognition by AAV8 capsids.³⁵ Further optimization of AAV9-derived vectors by varying amino acid residues at different positions and/or combining multiple point mutations (e.g., N498, W503, and W595) onto a single AAV9 capsid template could improve liver-detargeting efficiency.

The precise mechanism behind the altered liver tropism of AAV9.45 and other subtype II mutants remains to be determined. Analysis of structural features of different variants in this study suggests that residues 498–504 taken together with the adjacent 590–595 cluster contain key residues (N498, W503, P504, Q590L) that might constitute a partial receptor footprint on the AAV9 capsid (Figure 7). Consequently, one potential explanation is that altered affinity for an AAV9 receptor could, in turn, affect liver tropism. It is noteworthy to mention that similar examples have been reported in case of other viruses. For instance, variants of mouse parvovirus (minute virus of mice)^36,37 carrying mutations that alter receptor binding affinity have also been shown to display altered infectivity and tissue tropism. Another example hinges on the key role played by interactions between the adenovirus capsid hexon protein and vitamin K-dependent coagulation factors in mediating hepatocyte transduction. In this case, alteration of hexon-factor X interactions resulted in liver detargeting of adenoviral vectors.³⁸ Identification of similar factors such as glycans and/or glycoproteins/glycolipids that dictate AAV9 tissue tropism and analysis of mutant capsid-receptor interactions could provide further insight into the liver-detargeting mechanism.

Liver-detargeted AAV9 variants demonstrate significant potential for gene transfer in the treatment of cardiac and musculoskeletal diseases. Further optimization of liver-detargeted AAV9 vectors with transcriptional targeting elements such as cardiac- or muscle-specific promoters^39,14 or microRNA-122 target sequences^40,41 would allow selective delivery of therapeutic transgenes to heart and/or skeletal muscle.

Materials and Methods

Generation of AAV9 capsid library. The AAV9 helper plasmid, pXR9, containing AAV2 Rep and AAV9 Cap genes was obtained from the University of North Carolina Vector Core. A random plasmid library was generated by subjecting the capsid region encoding amino acids 390 to 627 (VP1 numbering; Genbank: AY530579.1)⁴² to error-prone PCR using forward 5′-GGT CGT TCG TCC TTT TAC TGC CTG GAA-3′ and reverse 5′-GCC GTC CGT GTG AGG AAT TTT GGC CCA-3′ primers (Integrated DNA Technologies, Coralville, IA). Cycling was carried out as per manufacturer instructions outlined in the GeneMorph II EZ clone domain mutagenesis kit (Agilent Technologies, Santa Clara, CA). Sequencing of individual clones was carried out by the University of North Carolina Genome Analysis facility and capsid sequences analyzed using VectorNTI software (Invitrogen, Carlsbad, CA).

Molecular modeling studies. Homology models of the VP3 monomer of AAV9 and different variants were generated using the SWISS-MODEL online 3D model building server⁴³ (http://swissmodel.expasy.org/) with the crystal structure of AAV8 as template (PDB ID: 2QA0).²⁶ VP3 trimer models were obtained using the online oligomer generator tool in the VIPERdb2 database⁴⁴ (http://viperdb.scripps.edu/). Surface rendered depictions of amino acid positions and cartoon models were generated using the program Pymol (The PyMOL Molecular Graphics System, Schrödinger LLC, http://www.pymol.org/). Lastly, “roadmap” projections of the AAV9 capsid surface highlighting different amino acid residues were constructed using the Radial Interpretation of Viral Electron Density Maps program.⁴⁵

Cell lines, plasmids, and viruses. HEK 293 cells were maintained at 37 °C in 5% CO₂ in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and penicillin-streptomycin-amphotericin B. Parental and variant pXR9 plasmid stocks were obtained from the randomized library described earlier. The plasmid pXX6-80, containing adenoviral helper genes was obtained from the University of North Carolina vector core. Vector cassettes, pTR-CBA-Luc, containing the CBA promoter-driven luciferase transgene and pTR-CBA-Tom, encoding tdTomato—a red fluorescent protein, were generated by ligating Luc/Tom inserts flanked by BamHI-NotI sites into the pTR-CBA backbone. Parental and variant AAV9 vectors were produced by the triple-transfection method followed by cesium chloride gradient ultracentrifugation and dialysis as described elsewhere.⁴⁶ Viral titers were determined by qPCR using a Roche Lightcycler with primers specific for the CBA promoter (forward 5′-CGT CAA TGG GTG GAG TAT TT-3′ reverse 5′-GCG ATG ACT AAT ACG TAG ATG-3′) or Luc transgene region (forward 5′-AAA AGC ACT CTG ATT GAC AAA TAC-3′ reverse 5′-CCT TCG CTT CAA AAA ATG GAA C-3′).

Animal studies. Housing and handling of BALB/c mice used in the current study were carried out in compliance with National Institutes of Health guidelines and approved by Institutional Animal Care and Use Committee at University of North Carolina–Chapel Hill (#09-117). At 8–10 weeks of age, animals were injected via the tail vein with a dose of 5 × 10¹⁰ vector genome-containing particles of AAV9 and related variants packaging CBA-luc or CBA-tom vector cassettes. Luciferase expression in animals was imaged at different time intervals using a Xenogen IVIS Lumina imaging system (Caliper Lifesciences, Hopkinton, MA) following intraperitoneal injection of D-luciferin substrate (120 mg/kg; Nanolight, Pinetop, AZ). Bioluminescent image analysis was carried out using the Living Image software.

Quantitation of luciferase expression. The same group of animals utilized for imaging studies, were killed at 4 weeks postinjection and the following organs were collected: brain, heart, lung, liver, and skeletal muscle (gastrocnemius). Approximately 50 mg of each tissue was homogenized in 150 µl of passive lysis buffer (Promega, Madison, WI) using a Tissue lyser II system (Qiagen, Valencia, CA). Tissue lysates were centrifuged at 8,000 rpm for 2 minutes to pellet debris and 50 µl of the supernatant transferred to 96-well plates for luminometric analysis (Promega) using a Victor2 luminometer (Perkin Elmer, Waltham, MA). Total protein concentration in tissue lysates was determined using the Bradford assay (BioRad, Hercules, CA). For monitoring time course of gene expression, AAV9 and AAV9.45 vectors were administered at a dose of 5 × 10¹⁰ vg/mouse to three different groups of animals. After being killed at 1, 2, and 4 weeks, heart and liver tissue from each group were processed for quantitation of luciferase transgene expression as described earlier. In order to determine the effect of vector dose on transgene expression level, AAV9 and AAV9.45 vectors were administered at three different doses: low (1 × 10¹⁰ vg per mouse), medium (5 × 10¹⁰ vg per mouse) and high (1 × 10¹¹ vg per mouse). Animals were killed at 2 weeks postadministration, following which heart and liver tissue were processed further for determination of luciferase transgene expression levels.

Quantitation of vector genomes. Approximately 100 µl of supernatant from tissue lysates obtained as mentioned above was processed using a DNeasy kit (Qiagen) to extract host and vector genomic DNA. Vector genome (Luc) and mouse lamin gene (internal standard) copy numbers were determined from 100 ng of total extracted DNA using qPCR. Vector genome copy numbers in blood were determined at different time intervals following intravenous administration of 1 × 10¹⁰ particles of AAV9 and related variants packaging the CBA-luc cassette. At 1, 24, and 48 hours postinjection, 10 µl of whole blood was collected from the tail vein in heparinized capillary tubes (Fisherbrand Hematocrit) and viral DNA quantified by qPCR.

Histological analysis. Two weeks after intravenous administration of 5 × 10¹⁰ particles of AAV9 and related variants packaging the CBA-tom cassette, mice were overdosed with intraperitoneal avertin (0.2 ml/10 g of a 1.25% solution) and perfused transcardially with ice-cold phosphate-buffered saline, then freshly prepared 4% paraformaldehyde in phosphate-buffered saline. Heart and liver tissues were then fixed overnight at 4 °C and 40 µm thick sections cut using a Leica vibrating blade microtome. Tissue sections were then imaged using an Olympus fluorescence microscope equipped with a rhodamine filter (emission max: 580 nm) and images collected using a Hamamatsu digital camera.

SUPPLEMENTARY MATERIAL Figure S1. Blood circulation profiles of adeno-associated virus serotype 9 (AAV9) and representative variants from each functional subtype; AAV9.47 (I), AAV9.45 (II) and AAV9.68 (III). Vector genome copy numbers in blood were determined at different time intervals following intravenous administration of 1×10¹⁰ particles of AAV9 and related variants packaging the CBA-luc cassette. At 1, 24 and 48hrs post-injection, 10μl of whole blood was collected from the tail vein in heparinized capillary tubes and viral DNA quantified by qPCR. Vector genome copy numbers are normalized per mL of whole blood. All experiments were carried out in triplicate. Error bars represent standard deviation. Table S1. List of variants. Table S2. Vector genome titers.

Supplementary Material