Comparative structural, biophysical, and receptor binding study of true type and wild type AAV2

Abstract

Adeno-associated viruses (AAV) are utilized as gene transfer vectors in the treatment of monogenic disorders. A variant, rationally engineered based on natural AAV2 isolates, designated AAV-True Type (AAV-TT), is highly neurotropic compared to wild type AAV2 in vivo, and vectors based on it, are currently being evaluated for central nervous system applications. AAV-TT differs from AAV2 by 14 amino acids, including R585S and R588T, two residues previously shown to be essential for heparan sulfate binding of AAV2. The capsid structures of AAV-TT and AAV2 visualized by cryo-electron microscopy at 3.4 and 3.0 Å resolution, respectively, highlighted structural perturbations at specific amino acid differences. Differential scanning fluorimetry (DSF) performed at different pH conditions demonstrated that the melting temperature (T_m) of AAV2 was consistently ~5 °C lower than AAV-TT, but both showed maximal stability at pH 5.5, corresponding to the pH in the late endosome, proposed as required for VP1u externalization to facilitate endosomal escape. Reintroduction of arginines at positions 585 and 588 in AAV-TT caused a reduction in T_m, demonstrating that the lack of basic amino acids at these positions are associated with capsid stability. These results provide structural and thermal annotation of AAV2/AAV-TT residue differences, that account for divergent cell binding, transduction, antigenic reactivity, and transduction of permissive tissues between the two viruses. Specifically, these data indicate that AAV-TT may not utilize a glycan receptor mediated pathway to enter cells and may have lower antigenic properties as compared to AAV2.

1. Introduction

Adeno-associated viruses (AAV) are small (~260 Å), non-enveloped ssDNA viruses, not associated with any known disease (Cotmore et al., 2019). They package a linear 4.7 kb ssDNA genome into a T = 1 icosahedral capsid. The genome contains 2 open reading frames (ORFs), rep and cap, which encode the replication (Rep) and capsid viral proteins (VP), respectively. In addition, frameshift reads in the cap ORF lead to the expression of the assembly activating protein (AAP) (Sonntag et al., 2010; Sonntag et al., 2011) and membrane associated accessory protein (MAAP) (Ogden et al., 2019). The capsid is stochastically assembled from 60 copies of three VPs, VP1, VP2, and VP3 in a ratio of 1:1:10, respectively (Buller and Rose, 1978; Snijder et al., 2014; Wörner et al., 2021). The sequence of VP3 is contained in VP2 and the sequence of VP2 is contained in VP1, the N-terminus of VP1 is therefore referred to as the VP1 unique region (VP1u). The overlapping region between VP1 and VP2 is termed the VP1/2 common region. AAP is important for translocation of translated VPs from the cytoplasm to the nucleus and facilitates assembly of the virus capsid (Sonntag et al., 2010). Presently, there are 13 AAV serotypes (AAV1-AAV13) identified and over 150 genotypes isolated from human and non-human primates (Gao et al., 2004). Specifically, AAV1, AAV2, AAV3, AAV5, AAV6, and AAV9 were isolated from human hosts (Atchison et al., 1965; Bantel-Schaal and zur Hausen, 1984; Gao et al., 2004; Georg-Fries et al., 1984; Hoggan et al., 1966; Melnick et al., 1965; Muramatsu et al., 1996; Rutledge et al., 1998). AAV4, AAV7, AAV8, AAV10, and AAV11 were found in non-human primates (Blacklow et al., 1968; Gao et al., 2002; Mori et al., 2004; Mori et al., 2008), and two serotypes, AAV12 and AAV13, were found as contaminants in adenovirus stocks (Schmidt et al., 2008a; Schmidt et al., 2008b). The amino acid sequence identity among the defined serotypes is ~60–99%, with AAV4 and AAV5 showing the greatest divergence (Gao et al., 2004).

Viral vectors based on AAV have shown promise in therapeutic gene transfer and are the most frequently used viral gene delivery method, with over 248 clinical trials currently in progress (https://www.clinicaltrials.gov). AAV2 is the most widely used AAV serotype in clinical gene therapy trials (https://www.clinicaltrials.gov) (Daya and Berns, 2008; Mueller and Flotte, 2008). However, one major limitation to the use of AAV2 for gene therapy purposes is its broad tissue tropism. Methods to improve tissue tropism include the use of other AAV serotypes (AAV1, 3–13) and viruses isolated from non-human primates (reviewed in (Wang et al., 2019); rational vector design based on available AAV structures (reviewed in (Mietzsch et al., 2019); directed evolution (Asokan et al., 2010; Kotterman and Schaffer, 2014; Shen et al., 2013; Tse et al., 2017); and in-silico construction (Zinn et al., 2015). Efforts have also been directed to the isolation of new genotypes from human samples (Bartel et al., 2012; Hsu et al., 2020; Limberis et al., 2009), including the natural AAV2 isolates from which AAV true type (AAV-TT) was derived (Chen et al., 2005).

To date three AAV vectors have been approved as biologics: Glybera^®, based on AAV1, for the treatment of lipoprotein lipase deficiency (Bryant et al., 2013); Luxturna^®, based on AAV2, for the treatment of Leber’s Congenital Amaurosis (Russell et al., 2017); and Zolgensma^®, based on AAV9, for the treatment of spinal muscular atrophy (Pattali et al., 2019). Advantages of AAV vectors include the non-pathogenic character of the virus (Carter, 2004); persistence of the delivered gene of interest as an episome (Afione et al., 1996; Schnepp et al., 2009); infection of both dividing and non-dividing cells (Hallek et al., 1998); long-term gene expression (Batty and Lillicrap, 2019); and the ability to produce and purify in large quantities (Kohlbrenner et al., 2005; Kotin and Snyder, 2017; Mietzsch et al., 2014; Smith et al., 2009; Zeltner et al., 2010).

AAV infection/transduction is initiated by the interaction of the capsid with a receptor and co-receptor to facilitate cellular entry (Huang et al., 2014; Nonnenmacher and Weber, 2012). The tropism of different AAVs is based largely on the receptor and co-receptor utilized (reviewed in (Nonnenmacher and Weber, 2012). The primary receptor for AAV2 is heparan sulfate proteoglycan (HSPG) and the key capsid residues for this interaction have been identified as R585 and R588 (Kern et al., 2003). However, circulating AAV2-like viruses isolated from human tissue have been shown not to encode for arginine’s at these positions (Chen et al., 2005; Hsu et al., 2020). Recently, a newly identified AAV receptor, AAVR, was demonstrated as a universal post-entry trafficking receptor (Pillay et al., 2016; Pillay et al., 2017; Summerford et al., 2016).

Following attachment, AAV is trafficked in clathrin-coated pits acidified to pH 6.0 (early endosome), pH 5.5 (late endosome), en route to the nucleus for genome replication, and has been shown to undergo pH-induced conformational changes that are essential for endosomal escape and transduction (Nam et al., 2007; Venkatakrishnan et al., 2013). Some virions may also be exposed to pH 4.0 in the lysosome. Alternative entry pathways mediated via CLIC/GEEC and caveolar endocytosis have been described as well (Nonnenmacher and Weber, 2012), highlighting that the processes involved in intracellular trafficking are complex. It is thought that AAV capsids which remain stable throughout intracellular trafficking and nuclear translocation have a stronger transduction profile.

Here we present the structural and biophysical characterization of AAV-TT, rationally designed from natural AAV2 isolates (Chen et al., 2005; Tordo et al., 2018). AAV-TT has been shown to have an improved in vivo transduction and spread in the CNS compared to AAV2 and is currently being evaluated for the treatment of diseases with global neuropathology (Tordo et al., 2018). AAV-TT differs from AAV2 by 14 amino acids, with one, two, and eleven amino acids substitutions distributed within VP1, VP2, and VP3, respectively. These include residues 585 and 588 in VP3, which are arginines in AAV2 and a serine and threonine in AAV-TT, respectively. AAV1 and AAV-TT both have a serine at 585 and a threonine at 588. AAV1 utilizes sialic acid as its primary receptor whereas AAV2 engages with HSPG, for which the aforementioned arginines have been shown to be critical for receptor binding (Huang et al., 2016; Wu et al., 2000; Wu et al., 2006).

To understand the differences in transduction ability between AAV2 and AAV-TT, we determined the structure of AAV-TT using cryo-electron microscopy (cryo-EM) to 3.3 Å resolution, and compared it to the previously determined structure of AAV2 to 3.0 Å resolution (PDB ID 6U0V). Superposition of the VP3 structures showed no significant differences in the main-chain Cα positions. However, structural perturbations were observed at the sites of amino acid differences. Heparin binding assays confirmed the role of residues 585 and 588 in HSPG recognition by showing that the lack of binding of AAV-TT could be restored by reintroduction of an arginine at position 585 and 588 (AAV-TT S585R/T588R). In vitro cell binding, infectivity assays and glycan arrays further demonstrate that AAV-TT does not use galactose, n-acetyl glucosamine or sialic acid as its primary receptor. Furthermore, high transduction levels previously seen in vivo in different rodent models could not be observed in neuronal cell lines or primary cortical neurons without restoration of the ability to bind HSPG. These data highlight that in vitro systems, including human iPSC derived neurons, do not serve as good models for in vivo transduction. Comparative stability analysis showed that the higher thermal melting temperatures (T_m) for AAV-TT is associated with the lack of basic amino acids at positions 585 and 588. Functional roles have previously been assigned to a number of the residues that differ between AAV2 and AAV-TT and this report provides a structural annotated guide, which maybe be useful for further vector development.

2. Materials and methods

Cell lines.

HeLa cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM), Neuro2A (derived from mouse neuroblastomas) and, U-87 (derived from human malignant gliomas) cells were cultured in DMEM/20 mM HEPES pH7.4, Lec2 (expressing terminal galactose), Lec8 (expressing terminal N-acetyl glucosamine), and Pro5 (expressing terminal sialic acid) cells were cultured in minimum essential medium (MEM). All media were supplemented with 10 % fetal bovine serum (Gibco^™) and 2 mM penicillin–streptomycin (Sigma). SH-SY5Y cells were cultured in DMEM:F12 (1:1) supplemented with 15% calf serum (Gibco^™) and 2 mM penicillin–streptomycin (Sigma). Rat primary cortical neurons (Gibco^™, A10840) were cultured in Neurobasal^™ medium (Gibco^™) supplemented with 2% B-27 (Gibco^™) and 0.5 mM GlutaMAX-I (Gibco^™) and fed every third day by aspirating half of the medium from each well and replacing it with fresh medium. Cells were infected at day 7 of culture. Human iPS cells (Thermo Fisher Scientific, A18945) were cultured in Essential 8 flex medium (Thermo Fisher Scientific, A2858501) and passaged using PBS/0.5 mM EDTA. Human Cortical progenitors were derived from iPS cells using a published protocol (Serio et al., 2013). They were cultured in base medium (50% Neurobasal, Thermo Fischer, 21103049) with 50% advanced DMEM (Thermo Fisher Scientific, 12634010), 5 Units/mL Penicillin Streptomycin (Thermo Fisher Scientific, 10378016), GlutaMAX (Thermo Fisher Scientific, 35050061), B-27 Supplement (Thermo Fisher Scientific, 17504044), and N-2 supplement (Thermo Fisher Scientific, 17502048) with 20 ng/mL FGF (Preprotech, 100-18B). Cells were passaged using PBS/0.5 mM EDTA. All mammalian cells were incubated at 37 °C in a humidified 5% CO₂ incubator and Sf9 cells were incubated at 28 °C.

Virus-like particle production and purification.

The AAV-TT cap gene was synthesized by GeneArt (Thermo Fisher Scientific) into a transfer vector pFastbac. The transfer vector was used to generate the bacmid containing the AAV-TT cap gene by homologous recombination in DH10 cells. The recombinant bacmid was used to generate AAV-TT virus-like particles (VLPs) in Sf9 cells according to the Bac-to-Bac expression system (Invitrogen) and as previously reported for AAV9 VLPs (Mitchell et al., 2009). A baculovirus expressing AAV2 VLPs was a generous gift from Sergei Zolotukhin (University of Florida), while those expressing AAV1, AAV5, and AAV9 were created as previously reported (DiMattia et al., 2005; Miller et al., 2006; Mitchell et al., 2009). Infected Sf9 cells expressing AAV-TT and AAV2 were harvested by centrifugation at 1.5 K rpm for 20 min in a JA10 rotor 72 hr post infection. The VLPs were released by three freeze thaws in an ethanol/dry ice slurry and a 37 °C water bath. The VLPs in the supernatant were precipitated by 10% PEG 8000. The PEG pellet was resuspended, combined with the cell lysate sample, treated with Benzonase at 37 °C for 1 hr, and clarified by centrifugation at 10 K rpm in a JA-20 rotor for 20 min. The clarified supernatant was further purified by a discontinuous step iodixanol gradient and ion exchange chromatography, according to a previously established protocol (Zolotukhin et al., 2002). The purity of the VLPs were determined by Coomassie stained SDS-PAGE. To determine capsid integrity, 5 μl of purified sample was loaded on to a glow discharged carbon coated copper electron microscope (EM) grid for 30 sec, blotted and stained with 2% uranyl acetate. The grids were examined with an FEI Spirit Transmission EM. The images were collected at x46,000 magnification and 120 kV with a Gatan 2Kx2K CCD camera.

Recombinant AAV production and purification.

Recombinant AAV2 (rAAV2) and AAV-TT virions were produced by co-transfecting HEK293T cells, grown in 15-cm dishes, with 2 plasmids: (i) the pDG plasmid encoding AAV2 rep and either the AAV2 or AAV-TT cap genes and adenovirus 5 helper functions and (ii) one plasmid encoding the transgene cassette flanked by the ITRs of AAV2 (pAAV-CAG-eGFP or pAAV-CMV-fLuc-eGFP). HEK293T cells were transfected using poly-ethylenimine (PEImax, Polysciences Inc.) and harvested 72 hr posttransfection. The cells were centrifuged and the pellet was resuspended in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 2 mM MgCl₂, pH8). Crude lysate was produced by three sequential freeze–thaw cycles and the supernatant was precipitated with ammonium sulfate. The clarified cell lysate and precipitated supernatant were treated with benzonase (Sigma) to remove cellular and non-encapsidated DNA. Both were then clarified by centrifugation and filtered at 0.22 μm before purification using the AKTA purifier chromatography system (GE Healthcare) and an AVB sepharose affinity column (GE Healthcare). The purified vector fractions were dialyzed against 1X PBS overnight using a dialysis cassette (Slide-A-Lyzer^™ 10 K MWCO, Thermo Fisher Scientific). The vectors were titred by qPCR using GFP primers; alkaline gel electrophoresis was performed to confirm the titre and assess the genome integrity (Kohlbrenner and Weber, 2017). The capsid titre was determined by SDS-PAGE electrophoresis and SYPRO-RUBY staining (Invitrogen) as previously described (Kohlbrenner et al., 2012).

Generation and purification of AAV capsid variants.

AAV-TT and AAV2 capsid variants were generated by site-directed mutagenesis of the pFastbac transfer vectors, using the site directed mutagenesis kit (Thermo Fisher Scientific) and the Bac-to-Bac expression system (Invitrogen) according to the manufacturer’s protocol. The AAV-TT variants generated were substitutions to the corresponding AAV2 sequence and vice versa for the AAV2 variants. The transfer vector was used to generate the bacmid containing the AAV-TT and AAV2 mutant cap genes by homologous recombination in DH10 cells. The recombinant bacmid was used to express AAV-TT and AAV2 variants in Sf9 cells according to the Bac-to-Bac expression system (Invitrogen) and purified by iodixanol step gradient and ion exchange chromatography, according to a previously established protocol (Zolotukhin et al., 2002). The AAV-TT variant is, AAV-TT-S585R/T588R (AAV-TT-HB⁺) and the AAV2 variant is AAV2-R585S/R588T (AAV2-HB⁻).

Cryo-EM and data collection of AAV-TT.

Three microliters of purified AAV-TT, at 1 mg/ml, was pipetted onto glow-discharged copper grids containing 2 nm carbon support holes (Quantifoil R 2/4 200 mesh, Electron Microscopy Sciences). The grids were immediately vitrified with a Mark IV Vitrobot (FEI Co.). The grids were screened for suitability (particle distribution and ice thickness) for high resolution data collection using a Tecnai G2 F20 -Twin transmission electron microscope operated at 200 kV and −20e⁻/Ų dosage on a 4Kx4K camera (Gatan). Movie frame micrographs were collected for suitable grids, with 2 exposures per hole, using a Titan Krios electron microscope (FEI Co.) operated at 300 kV with a DE64 detector (Direct Electron) as part of the Southeastern Center for Microscopy of MacroMolecular Machines (SECM4). The data collection parameters are listed in Table 2. Each movie was aligned using MotionCor2 with corresponding dark and bright reference images and applying radiation dose compensation (Zheng et al., 2017).

Table 2.

List of AAV2 and AAV-TT VP Amino Acid Differences.

AAV-TT Residue	Residue #	AAV2 Residue	VP Location
I	125	V	VP1
A	151	V	VP2
S	162	A	VP2
S	205	T	VP3
S	312	N	VP3
M	457	Q	VP3
A	492	S	VP3
D	499	E	VP3
Y	533	F	VP3
D	546	G	VP3
G	548	E	VP3
S	585	R	VP3
T	588	R	VP3
S	593	A	VP3

Structure determination, model building, and refinement of AAV-TT.

The micrographs were also processed with the cisTEM software package and the final map generated was used in the model building, and the parameters are listed in Table 1 (Grant et al., 2018). In brief, the aligned particles were imported and the CTF parameters estimated to allow for the elimination of micrographs of poor quality. Automatic particle picking was done using a capsid radius of 130 Å. The capsids were 2D classified using a box size of approximately 1.5x diameter. The classes without ice or debris were selected and used for ab initio 3D model generation, auto refinement and density map sharpening. The map was sharpened with a pre-cut off B-factor value of −90 Ų and a post-cut off B-factor value of 25 Ų. The structure determination procedure followed the “gold standard” protocol, and the resolution was estimated with a Fourier shell correlation (FSC) of 0.143. The B-factor sharpened map was used for the side- and main-chains manual model building and refinement in the Coot and Phenix programs (Afonine et al., 2013; Emsley et al., 2010). The AAV-TT map was interpreted with a monomer from AAV2 (PDB ID 6U0V). The AAV-TT specific residues were substituted from the corresponding AAV2 residues using a subroutine ‘Simple Mutate’ in the program Coot. The fitted AAV-TT VP3 monomer was converted to a 60mer using the subroutine Oligomer generator in the online program VIPERdb (viperdb.scripps.edu) (Carrillo-Tripp et al., 2009). The reconstructed map was converted from the MRC format to an XPlor format using the program e2proc3D.py subroutine in EMAN2 (Tang et al., 2007) and converted to the CCP4 fomat using the program MAPMAN (Kleywegt and Jones, 1996). The 60mer model was fitted into the normalized CCP4 map using the ‘Fit-in-map’ function in the program Chimera (Pettersen et al., 2004). The voxel size was screened for the best fit between the model and the map. A monomer was extracted, it was saved relative to the map and used for manual fitting using the program Coot (Emsley et al., 2010). The fitted monomer was converted to a 60mer using the Oligomer Generator in ViperDB (Carrillo-Tripp et al., 2009). The model was further refined in the Phenix application using real space, and B-factor refinement subroutines (Adams et al., 2010) and the final refinement statistics are listed in Table 1. Figures were generated using the UCSF-Chimera (Yang et al., 2012) and Pymol programs (DeLano, 2002).

Table 1.

AAV-TT summary of data collection and refinement.

Data Collection Parameters
Total number of micrographs	1444
Defocus range (μm)	0.8–3.91
Electron dose (e− / Å2)	59
Frames / micrograph	30
Pixel size (Å / pixel)	0.98
Starting number of particles	14,778
Particles used for final map	9,241
B-factor used for final map (Å2)	25
Resolution of final map (Å)	3.4
PHENIX model Refinement Statistics
Residue range	216–735
Map CC	0.851
RMSD [bonds] (Å)	0.010
RMSD [angles] (Å)	0.823
All-atom clashscore	8.31
Ramachandran plot
Favored (%)	96.31
Allowed (%)	3.69
Outliers (%)	0.00
Rotamer outliers (%)	0.00
C-β deviations	0.00

Heparin binding assay.

The heparan sulfate binding abilities of AAV2 (positive control), AAV-TT, AAV2-HB⁻ variant (AAV2-R585S/R588T), AAV-TT-HB⁺ variant (AAV-TT-S585R/T588R), and AAV9 (Tseng et al., 2016) (negative control) were tested using heparin agarose columns. VLPs, at 100 ng, in PBS-MK buffer (1xPBS with 1 mM MgCl2, and 2.5 mM KCl), were loaded onto an equilibrated 500 μl heparin-conjugated agarose type 1 resin column (H6508, Sigma) and the flow through (FT) collected by gravity flow. The column was washed with 5 column volumes of PBS-MK buffer and bound virus was eluted with increasing concentration (100 mM increments) of NaCl in PBS-MK buffer (Boye et al., 2016). The load (L), flow-through (FT), wash (W), and elution fraction (EF) samples were denatured by boiling at 100 °C for 10 min and analyzed by dot immunoblot. The samples were loaded on a nitrocellulose membrane and immunoblotted with primary antibody B1 (American Research Product), which recognizes a linear epitope at the C-terminus of the AAV VP (Wobus et al., 2000), and an anti-mouse monoclonal secondary antibody conjugated to horseradish peroxidase (HRP).

Cell binding assays.

Cell binding assays were used to determine potential glycan receptor(s) utilized by AAV-TT compared to AAV2 and HB^−/+ (R585S/R588T and S585R/T588R) variants. Three additional AAVs, AAV1, AAV5, and AAV9, were included as controls. The assay used fluorescently labeled purified virus. VLPs, at 1 mg/ml, were dialyzed extensively into PBS and labeled with DyLight 488 (Thermo Fisher Scientific) according to the manufacturer’s protocol. The completed reaction was further (3X) dialyzed into PBS. To confirm that the labeling reaction worked, the VLPs were examined on a 10% SDS-PAGE, using UV light imaging to detect the fluorophore, and an equivalent amount of VLPs were detected by staining with Coomassie blue using 0.5 mg/ml, 0.75 mg/ml and 1 mg/ml BSA standards. CHO cell lines, Pro-5, Lec2, and Lec8, displaying terminal sialic acid, galactose, and N-acetyl glucosamine, respectively, and neuronal cell lines, Neuro2A, U-87, and SH-SY5Y, were grown to 70% confluency in a 15-cm tissue culture dish. The cells were harvested by adding 2 ml of 0.5 M EDTA pH8 to the cell media. The harvested cells were pelleted by centrifugation at 500 rpm for 5 min (Beckman centrifuge). The cells were then incubated in pre-chilled media (without FBS) and counted with a hemocytometer. The cells were diluted to 5x10⁵ cells/ml in 10 ml of media, pre-chilled on ice for 30 min and aliquoted at 500 μl per tube. Each labelled VLP (AAV-TT, AAV2, AAV-TT-HB⁺, AAV2-HB⁻, AAV1, AAV5, and AAV9) was added at a multiplicity of infection (MOI) of 5x10⁴ per tube of cells. The cells and VLPs were incubated at 4°C for 4 hr with continual rocking. Unbound VLPs were removed by washing 3X with pre-chilled PBS with centrifugation in a Beckman centrifuge at 1000 rpm at 4°C for 5 min. The cells were resuspended in 500 μl of FBS and the number of labelled cells determined by fluorescent activating cell sorting (FACs) utilizing a FACS Calibur (Becton Dickinson). The experiments were done in triplicate and a paired samples student’s t-test was conducted.

Glycan Microarray.

Two different microarrays were used towards identifying potential glycans utilized by AAV-TT for infection. Labeled AAV2 and AAV-TT VLPs, at approximately 1 mg/ml, were submitted to a customized glycan microarray manufactured at the Max Planck Institute of Colloids and Interfaces, Potsdam, Germany that contains N-acetyllactosamine, dermatan sulfate, keratan sulfate and heparan sulfate probes (de Paz et al., 2007; Geissner et al., 2019; Hahm et al., 2017). The procedure, in short, is as follows: 250 μM synthetic oligosaccharides or natural heparin (5 kDa; Santa Cruz Biotechnology) were spotted onto N-hydroxysuccinimide-activated CodeLink slides (SurModics). The glycans were allowed to link to the slides in a humid chamber for 24 hr and then quenched for 1 hr at 50 °C with ethanolamine as described (Geissner et al., 2019). The slides were blocked with 1% (w/v) BSA in PBS-MK for 1 hr at RT to prevent non-specific binding, washed 3X with PBS-MK, and dried by centrifugation in a slide centrifuge. Fluorescein isothiocyanate (FITC)-labeled recombinant AAV (rAAV) vectors (according to the manufacturer’s protocol) were incubated on BSA blocked glycan slides at different concentrations in PBS-MK with 1% (w/v) BSA and 0.01% (v/v) Tween 20 for 24 hr at 4 °C. In parallel, fluorescence-labeled lectin from Bandeiraea simplicifolia FITC conjugate (Sigma-Aldrich) was incubated at 40 μg/mL in 50 mM HEPES buffer (pH 7.4) with 5 mM MgCl₂ and 5 mM CaCl₂, to verify successful glycan immobilization. After washing three times with PBS-MK, 0.1% (v/v) Tween 20, and once with deionized water, slides were dried by centrifugation and scanned with a GenePix 4300A microarray scanner (Molecular Devices). Fluorescence was excited at 488 nm. The photomultiplier tube gain was adjusted to reveal scans without saturated signals. Background-subtracted mean fluorescence intensity signals were determined with GenePix Pro 7 software (Molecular Devices) and exported to Microsoft Excel for further analysis (Hahm et al., 2017).

A second microarray screen was performed by the Consortium of functional glycomics (CFG) https://www.functionalglycomics.org/fg/. Labeled AAV-TT and AAV2 VLPs, at 50 and 200 μg, were dissolved in binding buffer (TSM: 0.02 M Tris–HCl, 0.15 M NaCl, 0.002 M CaCl₂, 0.002 M MgCl₂, 0.05% Tween 20, 1% BSA, pH 7.4) and analyzed. The microarray slides were rehydrated for 5 min in a solution of TSM plus 0.05% Tween (TSMW) at RT and allowed to drain by gently touching the end to a paper towel. Approximately 50 – 70 μl of each VLP was applied to the printed surface of the slide and a cover slip (24 by 50 mm) was applied to spread the sample over the entire microarray. The slide was incubated in a humidified chamber at an appropriate temperature for at least 1 hr. After incubation, the cover slip was removed and the slide washed by dipping four times (~3–5 s each) into a Coplin jar containing 100 ml of wash buffer I (TSMW: 0.02 M Tris–HCl, 0.15 M NaCl, 0.002 M CaCl₂, 0.002 M MgCl₂, 0.05% Tween 20, pH 7.4) followed by washing with wash buffer II (TSM, 0.02 M Tris–HCl, 0.15 M NaCl, 0.002 M CaCl₂, 0.002 M MgCl₂, pH 7.4). The fluorescently labeled VLPs applied to the slides were washed using wash buffer I and II as detailed above. The slides were then allowed to air dry for ~ 5 min and analyzed in a fluorescence reader, ProScanArray (PerkinElmer, Waltham, MA), equipped with multiple lasers. The software provided by the scanner manufacturer (ImaGene from BioDiscovery, El Segundo, CA) was used to process the image and to assign the glycan structures to each spot. The average relative fluorescence unit (RFU) value for each spot was determined as the average of 4 RFU values after removing the highest and lowest values from six data points. The final data were presented as histograms, with the glycan identification number related to the glycan structure plotted against the average RFU value, with the standard deviation shown as error bars (Smith et al., 2010).

In vitro transduction assay.

HeLa (cervical carcinoma-derived cells), SH-SY5Y (human neuroblast cell line), rat primary cortical neurons (Gibco^™, A10840), iPS cells, and human cortical neurons were used. HeLa cells were seeded in a 48-well plate at 4x10⁴ cells/well 3 hr before treatment. The media was replaced by 150 μl of complete media containing rAAV vectors expressing eGFP at MOI 10³, 10⁴ or 10⁵ vg/cell. Two hours post infection, an additional 150 μl of complete DMEM was added to each well. SH-SY5Y cells were seeded in a 48-well plate at a density of 6x10⁴ cells/well one day before transduction in complete media. The cells were then transduced as described above. Rat primary cortical neurons were seeded in a 48-well plate at a density of 5x10⁴ cells/well and cultured for one week before treatment. On the day of infection, 300 μl of complete media was slowly removed from each well and rAAV vectors were added at a MOI 5x10⁴ vg/cell. Two hours post infection, 300 μl of complete media was added to each well. The cells were then analyzed by FACS one week post-infection. IPS cells were passaged 1:6 and seeded into a 24-well plate. 24 hr later the transduction was performed using Essential 8 medium with Revita cell supplement (Thermo Fischer Scientific, A2644501). Human cortical progenitor cells were differentiated in base medium with 0.1 μM compound E (Enzo Life Science, ALX-270–415) and seeded into a 24-well plate. Imaging of transduced iPS cells and cortical neurons was performed on a Leica Dmi8 using a 20x (0.4NA) objective with a CoolLED pE-300 series light source. Live cell imaging was performed with 5% CO₂ and 37 °C humidified air. Transduction efficiency was manually analyzed by counting all cells per image using brightfield and transduced cells using the GFP filter. Transduction data for all cell types were collected from three biological replicates.

Thermal Stability.

Differential Scanning Fluorimetry (DSF) was used to determine the melting temperature (T_m) of AAV-TT, AAV2, AAV-TT-HB⁺, and AAV2-HB⁻ This method monitors the binding of the SYPRO orange dye to hydrophobic regions of proteins that are exposed during unfolding. For the comparative analysis of VLPs at different pH, 2.5 μl of each virus at 1 mg/ml, were added to 20 μl citrate phosphate buffer [CiPO₄ (0.2 M Na₂HPO₄, and 0.1 M citric acid pH 7.4, pH 6.0, pH 5.5, and pH 4.0)] and incubated at 4°C for 30 min. In addition, 2.5 μl of 1%-SYPRO-orange dye (Invitrogen Inc.) was added to each mixture for a total reaction volume of 25 μl. The assays were run in a Bio-Rad CFX thermocycler instrument with temperature ranging from 30 to 99 °C and ramping at 0.5 °C per step. A qPCR instrument with filters FAM (carboxyfluorescein, wavelength 485 nm) for excitation and ROX (carboxy-X-rhodamine, wavelength 625 nm) for emission, were used. The rate of change of fluorescence with temperature was recorded and the thermal profile was output as −dRFU/dT versus temperature. For evaluation, the thermal profile was inverted by multiplying with −1 and normalized by dividing the raw values of the profile by the peak dRFU/dT value. The peak value recorded on the thermogram is the T_m. A negative control consisting of 22.5 μl of each buffer and 2.5 μl of SYPRO-orange was included for each run. All experiments were conducted in triplicate as previously reported (Bennett et al., 2017).

Native dot Immunoblot.

To determine the antigenic properties of AAV-TT, AAV2, AAV-TT-HB⁺, and AAV2-HB⁻ 100 ng of each purified VLP was loaded onto nitrocellulose membranes for a native dot immunoblot or boiled for 10 min at 100 °C for a denatured dot immunoblot. The protocol was as previously described in (Boye et al., 2016; Gurda et al., 2012). The membrane was blocked with 5% milk in T-PBS (PBS and 0.05% Tween) for 1 hr. The native dot immunoblots were probed with monoclonal antibodies (MAbs) that detect conformational epitopes on the virus capsid, e.g., C37 and A20 detect AAV2 (Wobus et al., 2000); ADK9 (negative control) detects AAV9 (Tseng et al., 2016). The denatured dot immunoblot was probed with MAb B1 (Wobus et al., 2000). The MAbs were used at a dilution of 1:3000.

2.1. Structure accession number

The cryo-EM reconstructed density map and atomic model built for AAV-TT were deposited with accession numbers EMD-24266 and PDB 7NA6, respectively, in the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB).

3. Results and discussion

Cryo-EM reveals that the AAV-TT structure is similar to AAV2, with minor perturbations in sidechains.

The yield of VLPs for AAV-TT, AAV2, and their variants was ~2 mg from a 1 L infection. The purity of the samples was verified by Coomassie stained SDS-PAGE which showed VP1, VP2, and VP3 in the expected ratio (Fig. 1A, left, example shown for AAV-TT). Negative stain EM confirmed assembled capsids of the expected size (Fig. 1A, middle, example shown for AAV-TT). Frozen micrographs collected from vitrified AAV-TT particles were used for structure determination by cryo-EM. A total of 9241 (of 14,778) particles were extracted from the aligned micrographs (Fig. 1A, right) were utilized for the 3D-image reconstruction resulting in a reconstructed map at a resolution of 3.4 Å (FSC 0.143) (Fig. 1B and Fig. 2A). The high resolution of the AAV-TT map allowed for accurate fitting of the main- and side-chains of individual amino acid residues for residues 216–735 (C-terminal residue) and confirmed the amino acid sequence of AAV-TT when compared to an AAV2 VP3 model built into a reconstructed map at comparable resolution (Bennett et al., 2020) (Fig. 2B and 3). The AAVT-TT model building was done using a map contoured at a threshold of 2.0σ. The statistics of the model refinement is given in Table 1.

Fig. 1.

Cryo-EM data collection and refinement of AAV-TT. A) SDS-PAGE of purified AAV-TT (left) negative stain EM (middle) and cryo-EM micrograph (right). B) Fourier Shell Correlation plotted against inverse resolution (FSC plot).

Fig. 2.

Capsid surface comparison of AAV-TT (top) and AAV2 (bottom). Structures are colored according to the radial distance from the innermost region of the capsid, which is blue. The outermost region is red and a range of intermediate colors are listed according to the color key. The icosahedral 2-, 3-, and 5-fold axes are represented as an oblong, a triangle and a pentagon respectively.

Fig. 3.

Cryo-EM image reconstruction of AAV-TT capsid to 3.35 Å and AAV2 to 3.0 Å respectively, for comparison. A) Superposition of AAV-TT (orange) and AAV2 (blue) model fitted to the AAV-TT (orange mesh) and AAV2 (blue mesh) density maps contoured to 1σ with an enlarged view of residues 311, 312, 313, 547, 548, 584, and 585. Residues colored blue are AAV2 and orange are AAV-TT specific, respectively. B) AAV-TT (orange) model βB strand fitted to the density map contoured to 1σ. C) Superposition of AAV-TT (orange) and AAV2 (blue) model, icosahedral 2-fold, 3-fold, and 5-fold axes represented as an oblong, a triangle and a pentagon respectively. The C-terminal residue is labeled, and the N-terminal residue fitted into AAV2 density is 217 and that of AAV-TT is 216. The beta strands BIDG and CHEF facing the interior of the capsid are as labeled.

The surface morphology of the AAV-TT capsid is similar to that of AAV2 with minor changes due to differences in the side-chains (Fig. 2). The surface consists of shallow depressions at the icosahedral 2-fold axes with walls formed by αA and a conserved loop region at the C-termini of the VP3 (residues 696–704), located after βI; three protrusions surrounding the 3-fold axes, each assembled from three subloops within the GH loop with apexes formed by VR-1V, -V, and -VIII; and depressions, positioned above the βCHEF sheet on which lies the HI loop, surrounding a channel, assembled from 5 DE loops, at the 5-fold axes (Figs. 2 and 3C). There is a raised capsid region between the depressions at the 2- and 5-fold axes termed the 2/5-fold wall, which contains antigenic regions and amino acids thought to play a role in cellular trafficking and genome transcription (Nam et al., 2011; Salganik et al., 2014; Wobus et al., 2000). The 3-fold protrusions contain important determinants of receptor binding and immunogenicity (reviewed in (Halder et al., 2015; Mietzsch et al., 2019) (Gurda et al., 2012; Gurda et al., 2013; Pulicherla et al., 2011; Raupp et al., 2012; Shen et al., 2013)) and are the location of the majority of the amino acid differences between AAV2 and AAV-TT (Fig. 4). The HI and DE loops assembling the 5-fold region play a role in AAV genome packaging and uncoating, and the 5-fold channel is reported to be involved in the externalization of the VP1u region for its PLA2 function (Bleker et al., 2005; Wu et al., 2000).

Fig. 4.

Models of AAV-TT based on Cryo-EM structure. A) Monomer B) Surface representation of AAV-TT capsid map colored orange, with residues colored according to the table. The icosahedral 2-fold, 3-fold, and 5-fold axes are represented as an oblong, a triangle and a pentagon respectively. Listed AAV-TT/AAV2 residues located in VP3 and modeled in the AAV-TT density map.

Although the capsid is composed of VP1, VP2, and VP3, due to the low copy number, flexibility and predicted disorder of VP1u and the VP2/VP3 common regions, only the VP3 common region is observed in the AAV-TT map. This lack of N-terminal region ordering is consistent with observations for all other parvovirus structures determined to date (reviewed in (Mietzsch et al., 2017) except B19 for which the VP2 N-terminus could be observed on the outside of the capsid (Kaufmann et al., 2008). The VP3 monomers of AAV-TT and AAV2 superpose with a Cα root-mean-square deviation (RMSD) of 0.61 Å. With the exception of residues 216–223 where the chains deviate (Fig. 3C). The VP of AAV-TT maintains the core secondary structures of other parvoviruses, an eight-stranded β-barrel core (βBIDG and βCHEF) motif and α-helix A (αA) (Fig. 3C). An additional β-strand, βA, interacts with the βB of the βBIDG sheet (Fig. 3C). Large loops connect the secondary structure elements with regions of variability at their apex defined as variable regions (VRs, VR-I to -IX) based on the comparison of AAV2 and AAV4 (defined by (Govindasamy et al., 2006)). AAV2 and AAV4 represent the range of sequence and structure diversity among the AAVs discovered to date. A recently characterized naturally occurring AAV variant (AAVv66; PDB ID 6U3Q), share nine amino acid in common with AAV-TT (Hsu et al., 2020). Both capsids share approximately 98% sequence identity to AAV2 but are biophysically distinct, and both exhibit superior neuronal spread when compared to AAV2. Superposition of the monomer of AAV-TT and AAVv66 show a Cα RMSD of 0.49 Å, which is similar to value obtained for their superposition with AAV2. The residues that are different between AAV-TT, AAVv66 and AAV2 are located predominantly on the VRs of the capsid surface. VRs of the AAV capsid are important for antigenicity, trafficking and receptor binding.

AAV-TT does not interact with glycan receptor ligands for cellular entry in vitro.

The amino acid VP sequences utilized in the development of AAV-TT were derived from multiple AAV genomes isolated from human pediatric tonsils, spleen and lungs. The sequences isolated from tonsils and lungs had 98% amino acid identity to AAV2, yet none contained the HSPG binding residues R585 and R588 (Chen et al., 2005). This is consistent with the theory that the AAV2 serotype, that is presently utilized in labs, has been tissue culture adapted (Cabanes-Creus et al., 2020) and points to the potential use of another receptor or mechanism of cellular uptake (reviewed in (Nonnenmacher and Weber, 2012)) for AAV2 variants currently circulating in the human population . The receptors and AAV residues that are important for the engagement with target cells/tissues are known for AAV1, AAV2, AAV3b, AAV4, AAV5, AAV6 and AAV9 (reviewed in (Huang et al., 2014; Mietzsch et al., 2019; Nonnenmacher and Weber, 2012)). Structural and sequence alignment of the ordered VP3 monomer of AAV-TT with AAV1 – 6, and 9, for which glycan receptor attachment residues are known, showed that AAV-TT most closely resembles AAV9, with the same amino acids at positions D271, N272, Y446, and W503, but lacked N470 reported to be critical for galactose binding (Bell et al., 2012).

The primary receptor for AAV2 is HSPG and the residues important for this interaction are R484, R487, K532, R585, and R588, with the latter two considered critical (Kern et al., 2003; Opie et al., 2003; Summerford and Samulski, 1998). HSPGs are ubiquitously expressed on many cell surfaces and in the extracellular matrix of most animal tissues, where they interact with many ligands (Li and Kusche-Gullberg, 2016). A major difference between AAV-TT and AAV2 are residues 585 and 588 (Table 2, Fig. 4) that alter their HSPG binding phenotype (Fig. 5). A heparin binding assay and glycan microarray screening (Fig. 6) confirmed that AAV-TT does not bind HSPG. There is an excess of virus used for the heparin binding assay which leads to an excess of virus in the FT and the columns are washed until no virus is detected, the virus is then eluted with increasing NaCl concentration. In addition, R585S and R588T substitutions into AAV2 (AAV2-HB⁻) resulted in loss of heparin binding, while introducing S585R and T588R changes in AAV-TT (AAV-TT-HB⁺) conferred heparin binding (Fig. 5).

Fig. 5.

Heparin binding assay of AAV-TT, AAV-TT-HB⁺, AAV2, AAV2-HB⁻. AAV2 is the positive control and AAV9 is the negative control. The load (L), flow through (FT), initial wash (Wi) and final wash (Wf), and elution fractions (Ef) were denatured and detected using immunoblot with B1 as the primary antibody.

Fig. 6.

Glycan array analysis of AAV2 and AAV-TT. Glycans utilized in the array are Dermatan sulfates (S10-S14), Sialic Acid (S18-S23), Natural and Synthetic Heparan sulfate (S24-S36) and Sialyl Tn (S17).

Two independent glycan microarrays were performed for AAV2 and AAV-TT, using fluorescently labeled VLPs. While AAV2 bound to different variants on the glycan array from the Max Planck Institute of Colloids and Interfaces (Fig. 6). The CFG array identified a few weak glycan binders for AAV-TT with the top 5 sharing a terminal Gal-β1-4GlcNAc-β1-3-Gal (Chart ID# 539 and 561) or Neuro5Ac-α2-6/3-Gal-β1-4GlcNAc linkage (Chart ID# 388, 592, and 561) (data not shown). These linkages are common on many naturally existing glycans suggesting that AAV-TT may not utilize a glycan receptor mediated pathway to enter specific cells, at least in vitro.

Residues 585 and 588 determine cellular recognition in vitro for AAV2 and AAV-TT.

In addition to the glycan microarrays described above, cell-binding assays, with cell lines expressing different terminal glycans, were conducted towards identifying an AAV-TT glycan receptor. CHO cell lines Pro5, Lec2, and Lec8 displaying terminal glycans sialic acid, galactose, and N-acetyl glucosamine, respectively, were analyzed along with three neuronal cell lines U-87 (glioma cells), Neuro 2a (neuroblast), and SH-SY5Y (neuroblast). These latter neuronal cells were included as AAV-TT has been shown to have neurotropic properties in several rodent models. Neuronal cell lines display a variety of glycans, including multiple terminal sialic acids and chondroitin sulfate glycosaminoglycans (Hayes and Melrose, 2018). The purity, fluorescence labelling, and integrity of all the VLPs used were verified by Coomassie and UV analyzed SDS PAGE, and negative-stain EM, respectively, prior to use (Fig. 7A). The cell binding analysis included AAV2, AAV-TT, their variants, and AAV1, 5, and 9 were used as controls. AAV2 binds all CHO cells and SH-SY5Y cells with greater than 90% efficiency, U-87 with 70% efficiency, and Neuro2a with 20% efficiency (Fig. 7B). AAV-TT does not bind any of the cell lines tested. AAV-TT-HB⁺ has the same binding profile as AAV2 while AAV2-HB⁻ exhibits the same binding profile as AAV-TT (Fig. 7B). These observations indicate that both AAV2 and AAV-TT-HB⁺ are using residues R585 and R588 to interact with HSPG on the cell lines tested. Positive controls AAV1 and 5 bound Pro5 cells that display terminal sialic acid as expected, and AAV9 bound Lec2 cells displaying terminal galactose (Fig. 7B). Interestingly, AAV5 bound the U-87 cells with higher efficiency than the Pro5 cells suggesting a higher abundance of sialic acid.

Fig. 7.

Cell binding assay. A) Coomassie stained (left) and UV (middle) detected SDS PAGE and negative stain EM (right) of purified, labeled AAV1, AAV2, AAV-TT, AAV2-HB⁻, AAV-TT-HB⁺, AAV5 and AAV9 VLPs. B) The percentage of cells labeled compared to the total number of cells (10,000) detected by FACS and analyzed by CS Express⁶. The cell lines used were Lec2 (terminal galactose), Pro5 (terminal sialic acid), Lec8 (terminal N-acetyl-glucosamine), Neuro-2a, SH-SY5Y, and U87. The experiments were done in triplicate.

HeLa and SH-SY5Y cells, infected at different MOI with AAV2, AAV-TT, AAV2-HB⁻ and AAV-TT-HB⁺ expressing eGFP, also showed that HSPG binding is important for cellular transduction (Fig. 8A and 8B), as previously described (Tordo et al, 2018). As expected, AAV2-HB⁻ loses its in vitro transduction capability, and AAV-TT-HB+, gaining HSPG binding ability, results in successful transduction. Interestingly, the transduction efficiency achieved with AAV-TT-HB⁺ in HeLa and SH-SY5Y was markedly higher than observed for AAV2, indicating that AAV-TT may have an improved ability to navigate the post-entry trafficking process. In addition, transduction was tested in rat primary cortical neurons, human iPS cells, and human iPSC-derived cortical neurons (Fig. 8C, 8D and 8E). Although AAV2 showed high transduction efficiency, with up to 70% in rat cortical neurons (MOI 5 × 10⁴) and ~80% in iPS cells (MOI 10⁵), and to a lesser extent (~10%) in human cortical neurons, AAV-TT and AAV2-HB⁻ failed to transduce any of these cells (Fig. 8). The ability to transduce these primary cells was restored by introducing the ability to bind HSPG. These results support a requirement for residues R585 and R588 in these infections requiring HSPG binding residues and highlight the discrepancy between in vivo and in vitro transduction.

Fig. 8.

Transduction in immortalized cell lines and rat primary cortical neurons. HeLa (A) and SH-SY5Y (B) cells were transduced with AAV2, AAV2-HB⁻, AAV-TT and AAV-TT-HB⁺ expressing eGFP at the indicated MOIs and analyzed 48 h post-infection by flow cytometry. (A) and (B) The data represent the mean of three independent experiments. (C) Transduction in rat primary cortical neurons infected at a MOI of 5x10⁴ vg/cell, the cells were analyzed 7 days post-infection by flow cytometry. Transduction in human iPS cells and iPS cell-derived human cortical progenitors. Cells were transduced with AAV2, AAV2-HB⁻, AAV-TT and AAV-TT-HB⁺ expressing eGFP at a MOI 10⁵ vg/cell and analyzed 48hrs post-infection. The data represent the mean of three independent experiments.

Thermal Stability as a function of the endo/lysosomal pathway differentiates AAV2 and AAV-TT.

To probe the role of the residue differences between AAV2 and AAV-TT on their stability, the T_m was determined by DSF at cytoplasmic pH (pH 7.4) as well as the pH encountered in the endo/lysosomal pathway (pH 6.0, 5.5, and 4.0). AAV-TT is approximately 3 to 6 °C more stable than AAV2 consistent with the previous report (Tordo et al., 2018). In addition, the pH 7.4 study showed that residues R585 and R588 cause a destabilization of the AAV2 capsid, confirmed by a corresponding loss in T_m when these residues are added back to AAV-TT (AAV-TT-HB⁺) (Fig. 9A and B). This is similar to the phenotype observed for the Tm of AAV1 and AAV6. Conversion of E531 in AAV1 to K531 (AAV1-E531K) causes a reduction in Tm; this is the reverse of what is observed for AAV6-K531E which causes an increase in the Tm of AAV6 (Bennett et al., 2017).

Fig. 9.

Characterization of AAV2, AAV-TT (VLPs) and mutants by Differential Scanning Fluorimetry (DSF). A) Thermograms of AAV2 (blue), AAV-TT (orange), AAV2-HB⁻ (broken blue) and AAV-TT-HB⁺ (broken orange). B) Thermograms of AAV2 and AAV-TT at pH7.4, 6.0, 5.5 and 4.0.

AAV amino acid side-chains have been shown to undergo conformational changes when exposed to the decreasing pHs of the early (pH 6.0) and late (pH 5.5) endosome, and lysosome (pH 4.0) (Bartlett et al., 2000; Douar et al., 2001; Nam et al., 2007). The change in stability of different AAV serotypes at the pH encountered in the endosome and lysosome has been shown to be serotype specific (Lins-Austin et al., 2020). The T_m of AAV-TT is ~5°C greater than AAV2 at all the pH conditions tested. The T_m of AAV-TT VLPs at pH 7.4, 72.6°C, increases to 82.5°C at pH 6.0 while the capsid is most stable at pH 5.5, with T_m at 85°C, and decreased to 77.9°C at pH 4.0 (Fig. 9B). This profile is consistent with the stabilization of the virus capsid in preparation for the release of VP1u to facilitate escape into the cytoplasm and transport to the nucleus where the genome is replicated (Lins-Austin et al., 2020). The increase in stability of AAV-TT compared to AAV2 at physiological pH as well as the endo/lysosomal pH may suggest that AAV-TT is more suited to withstand potential pressure associated with genome packaging, VP1 externalization, and capsid transport.

AAV-TT maybe less antigenically reactive than AAV2.

To assess the recognition by neutralizing antibodies, AAV2, AAV-TT, AAV-TT-HB⁺ and AAV2-HB⁻ were probed with anti-AAV2 capsid MAbs A20 and C37 (Fig. 10). In this study, MAb B1 was used as a positive control to detect denatured samples while the ADK9 antibody, specific for AAV9, was used as a negative control in a native dot immunoblot (Tseng et al., 2016; Wobus et al., 2000). The binding sites for the two anti-AAV2 capsid MAbs are known from cryo-EM and mutagenesis studies (Gurda et al., 2013; Lochrie et al., 2006; McCraw et al., 2012; Wobus et al., 2000). The binding site for A20 covers a large region of the capsid and includes residues in VR-I, -III, -VII, -IX forming the 2/5-fold wall and the HI loop (Fig. 2B and 4) (Lochrie et al., 2006; McCraw et al., 2012). The binding site for C37-B includes residues in VR-IV, VR-V, and VR-VIII forming the 3-fold protrusions (Fig. 2B and 4) (Gurda et al., 2013).

Fig. 10.

Native Immunoblots of AAV-TT, AAV2, AAVTT-HB⁺, AAV2-HB⁻, and AAV9. Blots were probed with AAV2 native antibodies A20 and C37; AAV9 was also probed with the native antibody ADK9. Denatured blots were probed with B1.

Native dot immunoblot analysis confirmed the binding of A20 and C37-B to AAV2, while only A20 bound to AAV-TT (Fig. 10). Significantly, residues within the A20 footprint include those in VR-VII, for example, E548 (Lochrie et al., 2006), thus binding should be improved for AAV-TT when substituted with AAV2 residues. The AAV-TT-HB⁺, with residues 585 and 588 reverted to arginines, restores a low level of binding to C37-B, confirming VR-VIII, containing R585 and R588, as a determinant of C37-B binding to AAV2 (Gurda et al., 2013). This is further supported by the lack of C37-B binding by the AAV2-HB⁻ variant (Fig. 10). This observation may indicate that AAV-TT has the potential to escape a subset of neutralizing antibodies that affect AAV2 transduction.

4. Summary

This study utilized cryo-EM to determine the structure of AAV-TT, a rationally designed capsid based on the sequence of naturally circulating AAV isolates close to AAV2 which is being investigated for CNS applications. Structurally, the VP topology and capsid morphology of AAV-TT is homologous to AAV2, but amino acid position differences were clearly delineated. AAV-TT differs by 14 amino acids from AAV2, two of which, S585R and T588R are essential for HSPG binding. These two AAV-TT residues are similar to those in equivalent positions in AAV1. However, they do not play a role in AAV1′s sialic acid binding. Although various approaches were utilized, potential glycan receptors were not identified. Consistently, AAV-TT is unable to infect cells in vitro unless residue positions 585 and 588 are substituted to arginine, as present in AAV2. AAV-TT is significantly more stable than AAV2 throughout the endo/lysosomal pathway. A role for this phenotype in infection is yet to be determined but suggests a potential to withstand destabilizing conditions during infection. The data presented identify residues that can be engineered, do not affect the capsid structure, and result in vectors with improved function such as escape of neutralizing antibodies.