Structure-guided evolution of antigenically distinct adeno-associated virus variants for immune evasion

We analyzed previously resolved, cryoreconstructed structures of AAV1 capsids complexed with four different Fabs of mouse anti-AAV1 mAbs (9, 10). Annotation of the density for the Fabs onto the AAV1 capsid surface revealed that this subset of antibodies masks nearly the entire capsid surface (Fig. 1A). We then identified a subset of capsid surface residues (through the construction of 2D roadmap images, which pinpoint the Fab contacts on the AAV1 surface) that lie within these antigenic footprints and are implicated in direct contact with the different antibodies (Fig. 1B). Further analysis and comparison with different AAV serotypes revealed a prominent clustering of capsid antigenic footprints at the threefold symmetry axis on the capsid surface. Specifically, amino acid residues within three surface regions—region IV (456-AQNK-459), region V (492-TKTDNNNS-499), and region VIII (588-STDPATGDVH-597)—were selected for saturation mutagenesis and generation of different AAV libraries. It is important to note that the different capsid antigenic motifs (CAMs) listed above are subsets of VRs IV, V, and VIII (8). Each AAV capsid library was then subjected to five rounds of directed evolution in vascular endothelial cells, which are highly permissive to the parental AAV1 strain (Fig. 1C). AAV variants were identified, and combination AAV libraries were engineered using the latter as templates. Iterative rounds of evolution and capsid engineering yielded antibody-evading AAV strains characterized in the present study (Fig. 1D).

As outlined above, the structural determined AAV region IV, V, and VIII libraries were subjected to five rounds of directed evolution. Libraries were then sequenced using the Illumina MiSeq system, with each unselected (parental) library sequenced at ∼2 × 10⁶ reads and selected (evolved) libraries sequenced at ∼2 × 10⁵ reads. Demultiplexed reads were probed for mutagenized regions of interest with a custom Perl script, with a high percentage of reads mapping to these regions for all libraries (Fig. S1A). At both the nucleotide and amino acid levels, all unselected libraries demonstrated high diversity and minimal bias toward any particular sequence, whereas the evolved libraries showed dramatically increased representation of one or more lead variants (Fig. 2 A–C and E–G and Fig. S1B). Furthermore, within the top-10 selected variants for each library, many amino acid sequences showed similarities at multiple residues (Fig. 2 E–G). For instance, in the evolved region V library, 97.55% of sequences spanning the mutagenized region of interest read TPGGNATR, whereas minor variants largely mimicked this sequence (Fig. 2F). In the case of region VIII, we observed significant enrichment (86.6%) for a variant with amino acid residues TADHDTKGV (Fig. 2G). The evolved region IV library demonstrated higher plasticity (QVRG, 69.57%; ERPR, 14.05%; SGGR, 3.62%), as evidenced by the range of amino acid residues tolerated within that antigenic region (Fig. 2A).

We then generated a combination AAV library (regions V and VIII; Fig. 2D), which carries the lead epitope from the evolved region V library and a randomized region VIII. Interestingly, subjecting this library to directed evolution yielded the wild-type (WT) AAV1 sequence in region VIII (92.27%), i.e., STDPATGDVH (Fig. 2H). Although a second variant with the sequence DLDPKATEVE was also enriched (by 1.4%) (Fig. 2D), this observation demonstrates the evolutionary and structural constraints imposed by the interaction between region V and regions VIII to maintain capsid integrity. Nevertheless, these results corroborate the notion that antigenic footprints on the AAV capsid surface are mutable and can be evolved into distinct footprints while maintaining infectivity.

Multiple evolved AAV variants were selected from each library (annotated with the abbreviation CAM) for subsequent characterization. Specifically, the different variants were CAM101-107 (region IV), CAM108 (region V), and CAM109-116 (region VIII). All CAM variants packaging the ssCBA-Luc genome were produced and their transduction efficiencies assessed in vascular endothelial cells (Fig. S2J). A single CAM variant from each evolved library that displayed the highest transduction efficiency was shortlisted for further characterization. Specifically, CAM106 (456-SERR-459), CAM108 (492-TPGGNATR-499), and CAM109 (588-TADHDTKGVL-597) showed similar to modestly improved transduction efficiency compared with parental AAV1 on vascular endothelial cells. These observations support the notion that antigenic footprints can be reengineered and evolved while maintaining or improving on the endogenous attributes of the corresponding parental AAV strain. Further evaluation of the physical properties of these lead CAM variants confirmed that yield (vector genome titers), capsid morphology (EM), and packaging efficiency (proportion of full to empty particles) were comparable to those of parental AAV1 vectors (Fig. S2 A–D).

We first evaluated the ability of single-region CAM variants to avoid neutralization by mouse mAbs ADK1a, 4E4, and 5H7 described earlier (9, 10). The three mAbs that we selected for these studies were generated from mice immunized with AAV1 (6, 17). All three mAbs bind to threefold protrusions on the capsid surface (8, 9). An earlier study showed that 4E4 and 5H7 reduce AAV1 transduction by affecting both the binding and postattachment steps essential for host cell transduction (6). As shown in Fig. 3 A–C, each CAM variant shows a distinct NAb evasion profile. As expected, parental AAV1 was neutralized by all mAbs tested at different dilutions. The CAM106 and CAM108 variants were resistant to neutralization by 4E4, whereas CAM109 was completely neutralized, similar to AAV1 (Fig. 3A). These observations are consistent with the observation that 4E4 maps over regions IV and V, whereas region VIII is located farther away from the capsid surface. We next determined that CAM108 and CAM109 both avoided neutralization by 5H7 (which maps over regions V and VIII), whereas CAM106 was significantly affected by 5H7, similar to AAV1 (Fig. 3B). With ADK1a, which partially recognizes residues within region IV at the top of the threefold protrusions, CAM106 was significantly resistant to neutralization, whereas both CAM108 and CAM109 were effectively neutralized (Fig. 3C).

To further test whether the ability of CAM variants to avoid neutralization can be reproduced in vivo, AAV1 and CAM variants packaging ssCBA-Luc were mixed with the corresponding mAbs and injected i.m. into mice. In the absence of mAbs, all CAM variants and AAV1 showed similar luciferase transgene expression in mouse muscle (Fig. 3E). In the presence of antibodies, the neutralization profiles of the CAM variants corroborated the results from in vitro studies. In brief, CAM106 was resistant to ADK1a and 4E4, whereas CAM108 efficiently transduced mouse muscle in the presence of 4E4 or 5H7, and CAM109 evaded 5H7 with high efficiency. Importantly, AAV1 transduction of mouse muscle was completely abolished when coadministered with any of these antibodies (Fig. 3 F–H). Quantitative analysis of luciferase transgene expression by CAM variants normalized to AAV1 confirmed these observations (Fig. 3I).

Based on the promising results from our mAb neutralization studies, we hypothesized that combining different, evolved antigenic footprints will alter the capsid surface sufficiently to prevent recognition and neutralization by antibodies. To test this hypothesis, we generated three variants through a combination of rational mutagenesis, library generation, and iterative evolution. First, we observed that rational combination of antigenic footprints from CAM106 (region IV) and CAM108 (region V) yielded a functional and stable AAV variant, dubbed CAM117 (region IV + V) (Fig. 4A), but specific amino acid residues on CAM108 (region V) and CAM109 (region VIII) were not structurally compatible when combined on the same capsid. This observation is consistent with findings reported in other AAV libraries (18), in which simultaneous mutagenesis of these epitopes on the same capsid reduced viral titers, presumably owing to assembly defects. However, it is important to note that this apparent drawback can be addressed effectively by our iterative approach. Specifically, to facilitate structural compatibility, we generated an AAV capsid library on region VIII using CAM108 (region V) as a template. After three iterative cycles of directed evolution on vascular endothelial cells as described earlier, several viable variants were generated (Fig. 4A). After initial characterization, CAM125 (region V, 492-TPGGNATR-499; region VIII, 588-DLDPKATEVE-597) was selected for further analysis. We then iteratively engineered a third variant (CAM130) by grafting the evolved antigenic footprint from CAM106 onto CAM125. The CAM130 variant contains the following amino acid residues in three distinct antigenic footprints: region IV, 456-SERR-459; region V, 492-TPGGNATR-499; and region VIII, 588-DLDPKATEVE-597 (Fig. 4A). All three iteratively engineered variants—CAM117, CAM125, and CAM130—show physical attributes comparable to those of parental AAV1 with regard to vector titers and ratio of full to empty particles (Fig. S2 D–I).

To test whether antigenically distinct CAM variants can evade polyclonal NAbs found in serum, we seroconverted mice by immunization with WT AAV1 capsids. Overall, antisera obtained from individual mice efficiently neutralized parental AAV1, whereas CAM117, CAM125, and CAM130 displayed increased resistance to neutralization (Fig. 4 B–D). In brief, we tested antisera dilutions ranging over two orders of magnitude (1:3,200–1:50) to generate sigmoidal neutralization curves. The serum concentration required for 50% neutralization of transduction (ND₅₀) was 2- to 16-fold higher for each CAM variant compared with parental AAV1 in each individual subject (Figs. 4 B–D). Furthermore, we observed an incremental ability to evade NAbs with each iterative engineering/evolution step. Specifically, the most antigenically distinct variant, CAM130, showed an 8- to 16-fold improvement in resistance to neutralization (ND₅₀ >1:200; Fig. 4 B–D). These results corroborate the idea that antigenic footprints on AAV capsids are modular and cumulative in their ability to mediate NAb evasion. A similar, but less robust trend was observed with regard to the neutralizing potential of serum obtained from naïve mice used as a control (Fig. 4E).

To validate whether our approach can be translated in larger animal models, we tested the ability of AAV1 and the lead variant, CAM130, to evade NAbs generated in NHPs (rhesus macaques). In brief, we subjected AAV vectors to neutralization assays using sera collected at three different time points: preimmunization (naïve) and 4 wk and 9 wk postimmunization. All macaques seroconverted after immunization with NAb titers at the highest levels in week 4, with declines at week 9 in subjects 1 and 2 and increased potency at week 9 in subject 3 (Fig. 5). Moreover, naïve sera from subjects 1 and 3 before immunization could neutralize AAV1 effectively (Fig. 5 A and G). Because AAV1 was first identified as a contaminant in simian adenovirus stocks isolated from NHP tissue, we expected the natural host (i.e., rhesus macaques) to harbor antibodies against AAV1 capsids (19). We tested antisera dilutions ranging over two orders of magnitude (1:320–1:5) to generate neutralization curves as described earlier. Antisera obtained at 4 wk after immunization neutralized AAV1 effectively at ND₅₀ >1:320. In contrast, CAM130 displayed a significant shift consistent with improved resistance to neutralization (by approximately eightfold; ND₅₀ >1:40) compared with AAV1 (Fig. 5 B, E, and H). A similar trend toward enhanced resistance to NAbs was observed for CAM130 when evaluating antisera obtained at 9 wk postimmunization (Fig. 5 C, F, and I). These results strongly support the idea that antigenicity of AAV capsids can be engineered to broadly evade NAbs from different animal species on the basis of structural cues obtained from mouse mAb footprints.

To test whether CAM130 can evade NAbs and display a better profile compared with AAV1 in the general NHP and human populations, we tested serum samples obtained from cohorts of 10 subjects each. We evaluated a fixed serum dilution of 1:5 to reflect the currently mandated exclusion criterion used in ongoing clinical trials for hemophilia and other indications requiring systemic AAV administration. As shown in Fig. 6A, serum from NHP subjects p-A and p-B displayed high NAb titers that completely neutralized AAV1 and CAM130. At the other end of the spectrum, serum samples p-I and p-J neutralized neither AAV1 nor CAM130. Serum samples for subjects p-C–p-H efficiently neutralized AAV1 and reduced transduction efficiency to below 50% of that of untreated controls. None of these serum samples neutralized the antigenically distinct CAM130 variant. Thus, CAM130 showed exceptional NAb evasion in this cohort by evading 8 out of 10 serum samples (Fig. 6A).

We then used a similar approach to test serum from 10 human subjects using exclusion criteria (from 1:5 dilution to any detectable NAbs) mandated by several clinical gene therapy trials (e.g., ClinicalTrials.gov NCT01620801, NCT02618915, NCT01687608). We segregated human sera into two high-titer (h-A and h-B), six intermediate-titer (h-C–h-H) and two modest-titer subgroups. CAM130 was able to evade polyclonal NAbs in human sera in 8 of the 10 samples tested, whereas AAV1 did so in only 2 of the 10 samples (Fig. 6B). Taken together, our findings studies strongly support the idea that the three murine epitopes engineered in this study are conserved across species, and that the antigenically distinct CAM130 variant has the potential to evade antisera in a species-independent manner. Thus, our approach provides a roadmap for clinical translation and has the potential to significantly expand the patient cohort.

We compared the in vivo tissue tropism, transduction efficiency, and biodistribution of CAM130 and the parental AAV1 strain in mice. A dose of 1 × 10¹¹ vector genomes (vg)/mouse of AAV vectors packaging scCBh-GFP was injected i.v. into 6- to 8-wk-old female BALB/c mice via the tail vein. At 3 wk postinjection, CAM130 had an apparently enhanced cardiac GFP expression profile compared with AAV1 (Fig. S3 A and B). A modest yet significant (approximately twofold) increase in GFP-positive cardiac myofibers was found in CAM130-treated animals compared with the AAV1 cohort (Fig. S3G). However, when we administered 1 × 10¹¹ vg of AAV vectors packaging ssCBA-Luc genomes i.v., the increase in luciferase activity within the heart compared with that seen in AAV1-treated mice was marginal and not statistically significant (Fig. S3H). Thus, we conclude that the CAM130 variant is as effective, if not slightly more effective, than the parental AAV1 strain. We found no major differences in transduction efficiency in other organs, including liver, lung, brain, kidney, and spleen (Fig. S4A). Importantly, we found no differences in the systemic biodistribution of CAM130 and AAV1 vectors in major tissues, such as heart and liver (Fig. S4B). In general, higher vector genome copy numbers (∼10-fold) were recovered from the liver compared with cardiac tissue for both CAM130 and AAV1 vectors (Fig. S4B).

To further compare the tropism of CAM130 and AAV1, we evaluated the transduction profiles of these two strains after CNS administration. A dose of 3 × 10⁹ vg of AAV1 or CAM130 packaging scCBh-GFP genomes were injected by intra-CSF administration in neonatal mice. Both AAV1 and CAM130 spread well within the brain, with a general preference for transducing the ipsilateral side more readily than the contralateral side (Fig. S3 C and D). As seen with cardiac tissue, a similar number of GFP-positive neurons was observed for CAM130 compared with AAV1 (Fig. S3I). In particular, CAM130 appeared to selectively transduce neurons particularly within the motor cortex, cortex, and, most prominently, in the hippocampus, similar to AAV1 (Figs. S3 E and F). Taken together, these in vivo results confirm that antigenic footprints on AAV capsids can be engineered to effectively evade NAbs without compromising vector yield, cellular/tissue tropism, or transduction and biodistribution profiles. We expect that this approach can be readily extended to other AAV serotypes as well.

HEK293 and MB114 cells (a kind gift from Linda Van Dyk, University of Colorado) were maintained in DMEM supplemented with 10% FBS (Thermo Fisher Scientific), 100 U/mL penicillin, and 10 µg/mL streptomycin (Thermo Fisher Scientific) in 5% CO₂ at 37 °C. Murine adenovirus 1 (MAV-1) was purchased from American Type Culture Collection and amplified by infecting MB114 cells at a multiplicity of infection (MOI) of 1. At day 6 postinfection (∼50% cytopathic effect), media containing progeny MAV-1 viruses were harvested and centrifuged at 3,000 × g for 5 min, and the supernatant was stored at −80 °C for subsequent evolution studies. Mouse anti-AAV1 mAbs ADK1a, 4E4, and 5H7 have been described previously (6, 9, 17). Naïve human serum samples were purchased from Valley Biomedical. Naïve serum from rhesus macaques was a kind gift from Yoland Smith and Adriana Galvan (Yerkes National Primate Center, Emory University). Antisera against AAV1 capsids, generated by immunizing rhesus macaques i.m. with AAV1 capsids, was a kind gift from Jonah Sacha (Oregon National Primate Center). All mouse, human, and NHP sera used in this study were heat-inactivated at 55 °C for 15 min before use.

Recombinant AAV vectors were produced by transfecting four 150-mm dishes containing HEK293 cells at 70–80% confluence with polyethylenimine using the triple-plasmid protocol. Recombinant vectors packaging single-stranded genomes encoding firefly luciferase driven by the chicken β-actin promoter (ssCBA-Luc) or self-complementary green fluorescence protein driven by a hybrid chicken β-actin promoter (scCBh-GFP) were generated using this method. Subsequent steps involving the harvesting of recombinant AAV vectors and downstream purification were carried out using updated methods (28). Recombinant AAV vector titers were determined by quantitative PCR with primers amplifying AAV2 inverted terminal repeat regions, 5′-AACATGCTACGCAGAGAGGGAGTGG-3′ and 5′-CATGAGACAAGGAACCCCTAGTGATGGAG-3′.

Antigenic footprints of AAV serotypes 1/6, AAV2, AAV5, and AAV8 were determined using previously resolved structures of AAV capsids complexed with different mouse mAbs (8, 9). To restrict diversity and maximize the efficiency of AAV library generation, only amino acid residues directly in contact with antibodies were included for analysis. Contact surface residues on each serotype were either aligned by Clustal Omega or structurally superimposed using PyMOL (Schrödinger). Structural alignment revealed that antibody footprints from multiple serotypes overlap in close proximity to the threefold symmetry axis or around the fivefold pore and close to the twofold depression, especially at the twofold or fivefold wall (8, 9). Of these CAMs, we determined that the majority of the antibodies analyzed have direct contact with the threefold symmetry or the twofold or fivefold wall, supporting the notion that these regions are critical antigenic determinants on the AAV capsid (8, 9). For the present study, antigenic footprints for three distinct mAbs (4E4, 5H7, and ADK1a) were visualized on the AAV1 capsid (Protein Data Bank ID code 3NG9), and roadmap images were generated using RIVEM (29).

AAV libraries were engineered through saturation mutagenesis of amino acid residues within different antigenic footprints associated with distinct mAbs described above. In brief, for Gibson assembly, 12 oligos with an average length of 70 nt were ordered from IDT. Each oligo contained at least 15–20 nt overlapping homology to the neighboring oligos. Three oligos contained degenerate nucleotides within genomic regions coding for different antigenic footprints. Plasmid libraries were then generated by in vitro assembly of multiple oligos using High-Fidelity Gibson Assembly Mix (New England BioLabs) according to the manufacturer’s instructions. The assembled fragments were either PCR-amplified for 10 cycles using Phusion HF (New England BioLabs) or directly cloned into pTR-AAV1** plasmid between the BspEI and SbfI restriction sites. Plasmid pTR-AAV1** contains genes encoding AAV2 Rep and AAV1 Cap with stop codons at positions 490 and 491 (AAV1 VP1 numbering) introduced by site-directed mutagenesis (Agilent). The entire construct is flanked by AAV2 inverted terminal repeats to enable packaging and replication of pseudotyped AAV1 libraries on helper virus coinfection. Of note, the AAV1** capsid gene was incorporated before library cloning, to reduce WT AAV1 plasmid contamination during cloning. Ligation reactions were then concentrated and purified by ethanol precipitation. Purified ligation products were electroporated into DH10B ElectroMax cells (Invitrogen) and directly plated on multiple 5,245-mm² bioassay dishes (Corning) to avoid bias from bacterial suspension cultures. Plasmid DNA from pTR-AAV1CAM libraries were purified from pooled colonies grown on LB agar plates using the Invitrogen Maxiprep Kit.

Equal amounts (15 µg) of each pTR-AAV1CAM library and the Ad helper plasmid pXX680 were transfected onto HEK293 cells at 70–80% confluency on each 150-mm dish using polyethylenimine to generate CAM viral libraries. AAV CAM libraries were purified using standard procedures described earlier. MB114 cells were seeded on a 150-mm tissue culture dish overnight to reach 60–70% confluence before inoculation with AAV CAM libraries at an MOI ranging from 1,000 to 10,000. At 24 h posttransduction, MAV-1 was added as a helper virus to promote AAV replication. At 6 d postinfection with MAV-1 (50% cytopathic effect), the supernatant was harvested, and DNase I-resistant vector genomes were quantified on day 7. Media containing replicating AAV strains and MAV-1 obtained from each round of infection were then used as inoculum for each subsequent cycle, for a total of five rounds of evolution. Subsequent iterative rounds of evolution were carried out in a similar fashion with AAV capsid libraries containing different permutations and combinations of newly evolved antigenic footprints.

To analyze the sequence diversity of the parental and evolved AAV CAM libraries, DNase I resistant vector genomes were isolated from media and amplified by Q5 polymerase for 10–18 cycles (New England BioLabs) using primers 5′-CCCTACACGACGCTCTTCCGATCTNNNNNcagaactcaaaatcagtccggaagt-3′ and 5′-GACTGGAGTTCAGACGTGTGCTCTTCCGATCTNNNNNgccaggtaatgctcccatagc-3′. Illumina MiSeq sequencing adaptor for multiplexing was added in a second round of PCR using Q5 Polymerase with P5 and P7 primers. After each round of PCR, the products were purified using the PureLink PCR Micro Kit (Thermo Fisher Scientific). The quality of the amplicons was verified using a Bioanalyzer (Agilent), and concentrations were quantified using a Qubit spectrometer (Thermo Fisher Scientific). Libraries were then prepared for sequencing with the Illumina MiSeq 300 Reagent Kit v2, following the manufacturer’s instructions, and sequenced on the MiSeq system.

Demultiplexed reads were subjected to a quality control check using FastQC (v.0.11.5), with no sequences flagged for poor quality, and analyzed via a custom Perl script (Dataset S1). In brief, raw sequencing files were probed for mutagenized regions of interest, and the frequencies of different nucleotide sequences in this region were counted and ranked for each library. Nucleotide sequences were also translated, and these amino acid sequences were similarly counted and ranked. Amino acid sequence frequencies across libraries were then plotted in the R graphics package v3.2.4.

To characterize selected clones from each library, DNase I-resistant vector genomes were isolated from media and amplified by Phusion HF (New England BioLabs) using primers flanking the BspEI and SbfI sites. The PCR products were gel-purified, subcloned into TOPO cloning vectors (Thermo Fisher Scientific), and sent out for standard Sanger sequencing (Eton Bioscience). Unique sequences were subcloned into an AAV helper plasmid backbone using BspEI and SbfI sites. Unique recombinant AAV CAM variants were produced following an updated AAV vector production protocol (28).

Here 25 μL of antibody or antiserum (as specified for individual experiments) was mixed with an equal volume containing recombinant AAV vectors (1,000-10,000 vg/cell) in tissue culture-treated, black, glass- bottom 96-well plates (Corning) and then incubated at room temperature for 30 min. Assay parameters were carefully controlled as outlined previously (30). A total of 5 × 10⁴ HEK293 cells in 50 µL of medium was then added to each well, and the plates were incubated in 5% CO₂ at 37 °C for 48 h. Cells were then lysed with 25 µL of 1× passive lysis buffer (Promega) for 30 min at room temperature. Luciferase activity was measured on a Victor 3 multilabel plate reader (PerkinElmer) immediately after the addition of 25 µL of luciferin (Promega). All readouts were normalized to controls with no antibody/antiserum treatment. Recombinant AAV vectors packaging ssCBA-Luc transgenes were prediluted in DMEM + 5% FBS + penicillin-streptomycin and used in this assay.

Each hind limb of 6- to 8-wk-old female BALB/c mice (Jackson Laboratory) was injected i.m. with 2 × 10¹⁰ AAV packaging CBA-Luc premixed with three different mAbs—4E4, 5H7, and ADK1a—at 1:500, 1:50, and 1:5 dilution, respectively, in a final volume of 20 µL. At 4 wk postinjection, luciferase activity was measured using a Xenogen IVIS Lumina system (PerkinElmer) at 5 min after i.p. injection of 175 µL of in vivo D-luciferin (120 mg/kg; Nanolight) per mouse. Luciferase activity was measured as photons/s/cm²/sr and analyzed using Living Image 3.2 software (Caliper Life Sciences).

Statistical analysis was carried out with GraphPad Prism software using an unpaired, two-tailed Student t test. In the figures, *P < 0.5, **P < 0.05, ***P < 0.005, ****P < 0.0005, n.s. means not significant, and error bars represent 1 SD.

A total of 5 × 10⁹ vg was resuspended in NuPAGE LDS sample buffer (Invitrogen). + 50 mM DTT. Samples were run on NuPAGE 4–12% Bis-Tris Gel and transferred onto PVDF membrane (Invitrogen). AAV capsid proteins were detected using mouse mAb B1 (1:50) and secondary HRP-conjugated goat anti-mouse (Jackson ImmunoResearch Laboratories). For EM studies, 1 × 10⁹ vg/µL of the virus was prepared in PBS and absorbed on a Formvar/Carbon 400 mesh, Cu grid (Ted Pella). Samples were negatively stained with 2% uranyl acetate and analyzed using a Zeiss Supra 25 field emission scanning electron microscope.

Here 1 × 10¹⁰ vg of wtAAV1 in 20 µL of PBS was injected intramuscularly into each hind leg of 6- to 8-wk-old female BALB/c mice. Whole blood was collected by cardiac puncture at 4 wk postinjection, and serum was isolated using standard coagulation and centrifugation protocols. In brief, mouse blood was coagulated at room temperature for 30 min and then centrifuged at 2,000 × g for 10 min at 4 °C. All serum was heat-inactivated at 55 °C for 15 min and stored at −80 °C.

A dose of 1 × 10¹¹ vg of AAV vectors packaging the scCBh-GFP transgene cassette in 200 µL of PBS was injected into C57/Bl6 mice i.v. via the tail vein. Mice were killed at 3 wk postinjection and perfused with 4% paraformaldehyde in PBS. Multiple organs, including heart, brain, liver, and kidney, were harvested. Tissues were sectioned to 50-µm- thin slices with a vibratome (VT1200S; Leica) and stained for GFP with standard immunohistochemistry 3,3′-diaminobenzidine as described previously (28). At least three sections per organ from three different mice were submitted for slide scanning by University of North Carolina Translational Pathology Lab services.

For biodistribution analysis, 1 × 10¹¹ vg of AAV vectors packaging ssCBA-Luc were injected i.v. as described for BALB/c mice. At 2 wk postinjection, mice were killed and perfused with 1× PBS. Multiple organs, including heart, brain, lung, liver, spleen, kidney, and muscle, were harvested. DNA was harvested using the Qiagen DNeasy Kit, according to the manufacturer’s instructions. Vector genome copy numbers were determined by quantitative PCR using as described previously using luciferase transgene primers, 5′-CCTTCGCTTCAAAAAATGGAAC-3′ and 5′-AAAAGCACTCTGATTGACAAATAC-3′. Viral genome copy numbers were normalized to mouse genomic DNA in each sample. Tissue samples were also processed for luciferase activity assays by homogenization in 1x passive lysis buffer (Promega) using a Qiagen TissueLyserII at a frequency of 20 Hz for three 45-s pulses. The homogenate was spun down, and 20 µL of supernatant was mixed with 50 µL of luciferin (Promega) and immediately measured using a Victor 3 multilabel plate reader (PerkinElmer).

Postnatal day 0 (P0) C57/Bl6 pups that had been anesthetized on ice for 2 min, then given by stereotaxic intracerebroventricular injections with AAV vectors packaging the scCBh-GFP transgene cassette. A dose of 3 × 10⁹ vg in 3 µL of PBS was injected into the left lateral ventricle using a Hamilton 700 series syringe with a 26s-gauge needle (Sigma-Aldrich), attached to a KOPF-900 small animal stereotaxic instrument (KOPF Instruments). All neonatal injections were performed 0.5 mm relative to the sagittal sinus, 2 mm rostral to transverse sinus, and 1.5 mm deep. After vector administration, mice were revived under a heat lamp and rubbed in the bedding before being placed back with the dam. Mouse brains were harvested at 2 wk after vector administration (P14). Brains were postfixed and immunostained as described previously (28). All work with animals in this study was carried out in accordance with National Institutes of Health guidelines and approved by the University of North Carolina’s Institutional Animal Care and Use Committee.