Structural and Biophysical Analysis of Adeno-Associated Virus Serotype 2 Capsid Assembly Variants

ABSTRACT

Adeno-associated virus (AAV) is a nonenveloped single-stranded DNA (ssDNA) icosahedral T=1 virus being developed as a vector for clinical gene delivery systems. Currently, there are approximately 160 AAV clinical trials, with AAV2 being the most widely studied serotype. To further understand the AAV gene delivery system, this study investigates the role of viral protein (VP) symmetry interactions on capsid assembly, genome packaging, stability, and infectivity. A total of 25 (seven 2-fold, nine 3-fold, and nine 5-fold symmetry interface) AAV2 VP variants were studied. Six 2-fold and two 5-fold variants did not assemble capsids based on native immunoblots and anti-AAV2 enzyme-linked immunosorbent assays (ELISAs). Seven of the 3-fold and seven of the 5-fold variants that assembled capsids were less stable, while the only 2-fold variant that assembled had ~2°C higher thermal stability (T_m) than recombinant wild-type AAV2 (wtAAV2). Three of the 3-fold variants (AAV2-R432A, AAV2-L510A, and N511R) had an approximately 3-log defect in genome packaging. Consistent with previous reports of the 5-fold axes, the region of the capsid is important for VP1u externalization and genome ejection, and one 5-fold variant (R404A) had a significant defect in viral infectivity. The structures of wtAAV2 packaged with a transgene (AAV2-full) and without a transgene (AAV2-empty) and one 5-fold variant (AAV2-R404A) were determined by cryo-electron microscopy and three dimensional (3D)-image reconstruction to 2.8, 2.9, and 3.6 Å resolution, respectively. These structures revealed the role of stabilizing interactions on the assembly, stability, packaging, and infectivity of the virus capsid. This study provides insight into the structural characterization and functional implications of the rational design of AAV vectors.

IMPORTANCE Adeno-associated viruses (AAVs) have been shown to be useful vectors for gene therapy applications. Consequently, AAV has been approved as a biologic for the treatment of several monogenic disorders, and many additional clinical trials are ongoing. These successes have generated significant interest in all aspects of the basic biology of AAV. However, to date, there are limited data available on the importance of the capsid viral protein (VP) symmetry-related interactions required to assemble and maintain the stability of the AAV capsids and the infectivity of the AAV capsids. Characterizing the residue type and interactions at these symmetry-driven assembly interfaces of AAV2 has provided the foundation for understanding their role in AAV vectors (serotypes and engineered chimeras) and has determined the residues or regions of the capsid that can or cannot tolerate alterations.

INTRODUCTION

Adeno-associated viruses (AAV) are nonpathogenic, nonenveloped, single-stranded DNA (ssDNA) viruses. The AAVs belong to the Parvovirinae subfamily within the Parvoviridae and are members of the Dependoparvovirus genus, as they require a helper virus for replicative infection (1). The AAV capsid is a T=1 icosahedron, ~260 Å diameter in size, that packages a linear ssDNA genome of ~4.7 kb (2). The genome contains 2 open reading frames (ORFs), rep and cap (1). The rep ORF encodes four nonstructural overlapping proteins, Rep78, Rep68, Rep52, and Rep40, that are involved in viral DNA replication and packaging (3). The cap ORF encodes three overlapping structural viral proteins (VPs), VP1 (87 kDa), VP2 (72 kDa), and VP3 (62 kDa), that are generated from alternate start codons and alternative splicing (4, 5). The VPs have a common C-terminal region (VP3) with VP2 and VP1, having N-terminal extensions of ~65 and ~200 amino acids, respectively. The AAV capsid is assembled from 60 copies (in total) of VP1, VP2, and VP3 which are stochastically incorporated with an overall ratio of 1:1:10, respectively (6, 7). The VP3 common region forms the contiguous shell of the virus capsid and has roles in receptor binding, cellular trafficking, antibody recognition, and genomic DNA packaging (8–10). Both VP1 and VP2 contain nuclear localization sequences and have been suggested to play a role in transporting the capsid to the nucleus, and VP1 has a unique N-terminal region (~140 amino acids) which displays a phospholipase A2 activity required for endosomal escape during cellular trafficking (11, 12). Two additional proteins encoded in the cap ORF are the assembly-activating protein (AAP) and the membrane-associated protein (MAAP). The AAP (23 kDa) is encoded in an alternate reading frame within the cap ORF with a +1 frameshift, and a nonconventional CTG start site and has been shown to stabilize unassembled VPs, aid transport to the nucleolus, and facilitate capsid assembly (13–15). The MAAP also uses a cryptic start site in the cap ORF and has been shown to be important for capsid egress (16).

To date, 13 human and nonhuman primate AAV serotypes (AAV1 to -13) and over 100 genome isolates (17, 18) have been identified. These AAVs exhibit between 50 and 99% amino acid sequence identity, with AAV4 and AAV5 being the most divergent (18). The AAV serotype 2 (AAV2) is the best-characterized member, because of its role in the discovery of AAV and development as a vector for gene delivery applications and, as such, serves as the model system for this study (19).

Five biologic AAVs have been approved for clinical treatments: Glybera, approved in 2012 as an AAV1 vector for the treatment of lipoprotein lipase deficiency (20), Luxturna, approved in 2017 as an AAV2 vector for the treatment of retinal dystrophy (21), Zolgensma, approved in 2019 as an AAV9 vector for the treatment of spinal muscular atrophy (22), Upstaza, approved in 2022 as an AAV2 vector for the treatment of aromatic l-amino acid decarboxylase (AADC) deficiency (23), and Roctavian, also approved in 2022, as an AAV5 vector approved for the treatment of hemophilia A (24); there are also several other AAV FDA approvals in the pipeline (www.clinicaltrials.gov).

Understanding the role of the AAV capsid structure and its function as a therapeutic agent has prompted the determination of high-resolution structures of AAV1 to -13, AAVs from nonhuman primates, and engineered AAV capsid variants. These structures have been determined either by X-ray crystallography and/or cryo-electron microscopy (cryo-EM) (reviewed in references 25 and 26). The capsid structures have provided information on the VP3 common regions, including AAV2 from residues 217 to 735; although VP1 and VP2 are present in the samples studied, they are not structurally ordered (27). A main structural motif of the AAV VP3 are a conserved eight-stranded antiparallel β-barrel motif (βB-βI) and βA, which forms the core of the capsid and an α-helix (αA), all of which are superposable between the AAV serotype structures (28). There are large inter-β-strand loops which vary in amino acid length, composition, and structure between the AAV serotypes that are defined as variable regions (VR I-IX), based on the two structurally diverse serotypes, AAV2 and AAV4 (28). The inter-β-strand loops constitute 70% of the VP3 fold; several of them colocalize at the icosahedral axes of the capsid and interact with each other at the 2-, 3- and 5-fold axes. These loops create the surface topology of the capsid (2), which is characterized by several prominent features; these include a depression at the 2-fold axes, three protrusions radiating around the 3-fold axes, and a channel at the 5-fold axes. The protrusions at the 3-fold axes have been shown to play important roles in receptor and coreceptor binding, antigenicity, and transduction efficiency. For example, the AAV2 3-fold protrusions bind heparan sulfate proteoglycan, which is the primary glycan receptor, and αvβ1 and αvβ5 integrin and the glycosylated protein AAVR, which are coreceptors (29–33). The βD-βE loop (VR II) forms the channel at the 5-fold axes, which provides solvent access to the interior of the capsid and is described to be important for genome packaging and VP1u externalization (34, 35). While there is a considerable amount of data available about the capsid structure and the role it plays in the AAV life cycle, there is limited information available on the symmetry-related VP interactions and their involvement in the assembly and stability of the capsid.

The AAV capsid assembly occurs in a two-step process, a fast step, involving the interaction of the VPs to form the 60mer T=1 shell, followed by a slow step, culminating with the packaging of the viral genome aided by the Rep proteins (36). Only VP3 is required for capsid assembly when the AAP is provided in trans (15). Previous alanine amino acid replacement scanning by random mutagenesis of several of the AAV VPs indicated that conserved residues in the 2-fold, 3-fold, and 5-fold interfaces are important for capsid assembly and genome packaging (34, 37, 38). In comparison, machine-guided comprehensive mutagenesis allowing for the screening of all VP residues to the 20 canonical amino acids and has provided important information on the effect of the amino acid type on vector production, specifically, capsid assembly and genome packaging (16). For example, mutations in the buried surface area and the 5-fold axes were more detrimental, and those in the variable regions and 3-fold axes were more accommodated, especially those variants that were amino acid substitutions found in existing serotypes. It was also proposed that mutations to positive-charge amino acids were deleterious at all positions, and negative charge-amino acids, especially at the 3-fold protrusions, were beneficial for viral fitness (16).

This study provides a structural view of symmetry interface alanine variants to decipher the role of side chain interactions on VP:VP and VP:genome interactions during capsid assembly and packaging. Relatedly, AAP has been shown to be critical for AAV capsid assembly; to date, the mechanism by which this process is facilitated is not known. It has been shown, however, that AAP uses multiple basic regions on its C terminus to translocate the VPs to the nucleolus, where capsid assembly occurs (39), and acts as a scaffold for AAV capsid assembly (13). Deciphering the functional components of AAP reveals two regions, the hydrophobic core and the conserved region, which interacts with VP on its luminal surface (40); it has been previously proposed that AAV8 VP residues Q413, E684, and E686 on the luminal side of the VP create an important charge balance required for AAP interaction (40). In our study, we also identified two additional residues, V221 and S224, that adversely affect capsid assembly. These residues are located in the AAP ORF and are the only assembly variants rescued by the addition of AAP in trans. Therefore, V221 and S224 are variants of AAP as well as AAV VP assembly variants.

RESULTS AND DISCUSSION

AAV2 VP capsid symmetry interfaces.

A total of 25 amino acids were selected and targeted based on their VP capsid interface location for study using site-directed mutagenesis of the cap ORF for AAV2 (Table 1). The selections were based on regions of the capsid that were highly conserved within the Parvovirus family and the Dependoparvovirus genus, specifically, AAV1 to -13, and on previously reported assembly variants (28, 34, 37, 38).

TABLE 1.

Comparative assessment of the biophysical properties of AAV2 variants^a

Variantb	Capsid/mL (cell lysate)	Capsid/mL (purified)	Titer (genome/mL)	Assembly ratio (cell lysates)c	Packaging efficiencyd	Normalized RLU	Tm (°C)
WT	5.25E+11	5.77E+12	4.39E+11	1.00E+00	1.79E+01	1.00E+00	66.8 ± 1.3
R294A	3.79E+07	BD	BD	7.22E-05	BD	BD	BD
Q297A	3.79E+07	BD	BD	7.22E-05	BD	BD	BD
R298A	3.79E+07	BD	BD	7.22E-05	BD	BD	BD
R294/R298A	3.79E+07	BD	BD	7.22E-05	BD	BD	BD
K692A	6.14E+10	1.63E+12	5.50E+10	1.17E-01	1.09E+02	6.65E-01	68.0 ± 2.3
W694A	3.79E+07	BD	BD	7.22E-05	BD	BD	BD
P696A	3.79E+07	BD	BD	7.22E-05	BD	BD	BD
R389A	6.04E+11	8.85E+12	3.09E+11	1.15E+00	3.50E+01	1.26E+00	68.4 ± 0.9
R432A	2.53E+11	3.90E+12	7.47E+9	4.82E-01	1.03E+04	2.92E-01	54.8 ± 3.3
Y441A	1.03E+11	1.80E+11	2.14E+10	1.96E-01	2.36E+02	1.88E-01	49.2 ± 4.6
L510A	1.83E+11	8.99E+12	2.16E+11	3.49E-01	6.06E+03	8.15E-02	58.7 ± 1.3
N511R	2.26E+11	7.97E+12	2.11E+10	4.30E-01	3.59E+03	6.29E-02	56.4 ± 2.6
E564A	2.32E+11	4.38E+12	1.74E+11	4.42E-01	2.80E+01	9.78E-03	69.0 ± 1.3
P602A	5.42E+11	7.14E+12	9.57E+10	1.03E+00	1.26E+02	4.28E-01	63.1 ± 1.0
P622A	1.86E+11	9.20E+11	5.82E+10	3.54E-01	7.06E+00	1.87E-01	56.0 ± 0.0
D625A	2.49E+11	1.89E+13	4.51E+11	4.74E-01	6.45E+01	2.18E-01	64.3 ± 2.3
V221/S224A	3.79E+07	BD	BD	7.22E-05	BD	BD	BD
H255/K258A	4.01E+10	2.74E+11	3.70E+10	7.64E-02	8.68E+00	7.18E-01	52.5 ± 3.5
T331A	2.10E+11	8.05E+11	1.34E+10	4.00E-01	7.73E+01	2.60E-01	65.5 ± 0.4
Y397A	4.28E+11	3.84E+12	1.65E+11	8.15E-01	6.30E+01	5.93E-01	66.3 ± 0.3
M402A	3.79E+07	BD	BD	7.22E-05	BD	BD	BD
R404A	4.65E+11	4.94E+12	3.49E+10	8.86E-01	1.48E+02	2.28E-03	65.3 ± 0.3
Y413A	3.28E+10	2.26E+11	1.19E+10	6.25E-02	1.78E+01	2.25E-01	57.5 ± 2.2
F415A	2.52E+10	8.40E+10	1.54E+10	4.80E-02	1.32E+01	6.98E-02	56.0 ± 1.0
F661A	1.59E+11	5.40E+12	4.46E+10	3.03E-01	3.45E+01	1.56E-01	65.7 ± 0.3

^aBD, below detection; RLU, relative light unit.

^bAssembly ratio, A20 ELISA of variant (capsid/mL)/A20 ELISA of WT AAV2 (capsid/mL).

^cPackaging efficiency, (capsid/mL)/(genome/mL).

^dUnderlining indicates variants screened for defect in AAP.

Seven of the VP variants were located within the 2-fold symmetry-related interface. Four of them (R294A, Q297A, R298A, and R294A/R298A) formed the wall lining of the 2-fold depression within the αA helix, and 3 of them (K692A, W694A, and P696A) formed the C-terminal loop which forms the 2-fold floor. It should be noted that R294A had been previously studied as part of AAV2 VP mut24 (R294A and R295A) and characterized as an assembly mutant (37), while K692A could be considered an outlier of the group because it is also involved in a 3-fold interaction (Fig. 1A and B). The buried surface area of the 2-fold VP interface is 3,060 Ų and represents the energetically weakest interface, and as such, amino acid changes in this region could potentially have significant effects on the stability of the interface and therefore the capsid integrity. It is also important to note that point mutations/deletions of the C terminus of the VPs also reduce interactions with AAP and abolish capsid assembly (13).

FIG 1.

Ribbon diagram of wtAAV2 capsid icosahedral symmetry interfaces. (A) Viewed down the 2-fold axis. (B) Closeup indicating the location of VP variants (yellow spheres). (C) Viewed down the 3-fold axis. (D) Closeup indicating the location of VP variants (green spheres). (E) Viewed down the 5-fold axis. (F) Closeup indicating the location of VP variants (orange spheres). The VP variants were also labeled with the same color as the interacting monomer.

Nine of the VP amino acid substitutions were involved in 3-fold interactions. The buried VP surface area of this interface is 10,350 Ų and therefore represents the energetically strongest interface interactions. The large, buried surface area is caused by the extensive inter-VP interactions generated predominantly by the EF and GH loops that interlock to form most of the 3-fold interactions. Compared to the amino acids located at the 2-fold and 5-fold interfaces, which are structurally conserved between the AAV serotypes, those at the 3-fold interface are predominantly within the VRs. Five of the selected variants, R389A, Y441A, L510A and N511R, and P602A are located in VR III, IV, V, and VIII, respectively. The other variants, R432A, P622A, and D625A, are in the GH, loop which is more structurally conserved between serotypes than the VRs mentioned (Fig. 1C and D). Furthermore, to date, no 3-fold symmetry variant of the AAV2 VP has been reported to adversely affect capsid assembly. These observations taken together would indicate that the 3-fold symmetry-related interactions are the most tolerant to adaptation and represent a region of the capsid that can be engineered with the smallest adverse effect on capsid assembly.

The remaining 9 VP variants studied are involved in 5-fold interactions, 2 of which are double variants, V221A/S224A and H255/K258A (mut 21) (34, 37). Of note, V221A, located at the interior entrance of the capsid 5-fold pore, has been previously substituted to tryptophan, tyrosine, and cysteine and in each case was shown to assemble capsids. In this study it was substituted to an alanine (34, 41). The capsid 5-fold pore of AAV2 formed by VP interactions of the conserved β-ribbon between βD and βE (324 to 338) creates a “turret-like” structure interacting with the N terminus of the adjacent VP (219 to 229). The remaining 5-fold variants were located within the capsid body and within the conserved β-strands. The buried VP surface area of the 5-fold interface is 4,950 Ų, which greater than that of the 2-fold interface but less than those of the 3-fold VP interactions. A distinguishing feature of the capsid is the HI loop, which sits on the depression around the 5-fold interface. The F661A VP variant did not form a symmetry interface interaction, as it was positioned on the corner of the HI loop (Fig. 1E and F). The HI loop also has no effect on capsid assembly and can be modified without any adverse effects on AAV capsid assembly, packaging, and infectivity. The final two VP variants studied were Y413A and F415A, which are located in the βG strand. The VP F415 position was previously studied and characterized as a part of the assembly mutant mut30 (F415A and D416A) (37).

Capsid 2- and 5-fold interfaces are important for capsid assembly.

The AAV2 VPs and Rep proteins were expressed and observed for all 25 variants, based on the cell lysate results with B1 and 1F antibody, respectively. B1 antibody recognizes a linear sequence of the C terminal of VP1, VP2, and VP3, and 1F antibody recognizes a linear sequence within the common central domain of the replication proteins Rep40, Rep52, Rep68, and Rep78 (42). The immunoblots confirmed the production of the VPs, which are required to facilitate the assembly of the virus capsid (Fig. 2A). The immunoblot analysis further identified AAV2 VP variants that disrupt and prevent capsid assembly. Six of the seven 2-fold variants (R294A/R298A, Q297A, R298A, W694A, and P696A) were A20 and C37 negative, whereas all of the nine 3-fold variants were positive for both antibodies, and three of the nine 5-fold variants (V221A/S224A, M402A, and E415A) were also A20 and C37 negative (Fig. 2A, rows 3 and 4). Both C37 and A20 recognize different conformational epitopes on the AAV2 capsid (43). A quantitative comparison of the wild-type AAV2 (wtAAV2) and variant capsid confirmed that the assembly defective variants identified in Fig. 2A were significant, with approximately 4 log decreases compared to wtAAV2 (Fig. 2B). It also illustrated that three variants, K692A, H255A/K258A, and Y413A, had lower assembly efficiencies, producing 1 to 2 logs fewer capsids than wtAAV2 (Fig. 2B). These data are consistent with previously studied 5-fold assembly variants, including H229/D231A (mut 19), K321A/E322A (mut26), N334W, V221W, V221C, and V221Y (34, 37). Although none of the 3-fold variants were found to be severely defective for capsid assembly, it is worth noting that several of them displayed a trend toward poorer assembly efficiency (2- to 7-fold) than wtAAV2 (Fig. 2B; Table 1).

FIG 2.

wtAAV2 and variant cell lysate protein expression in HEK 293 cells. (A) Cell lysate denatured dot blots: B1 (row 1) for capsid proteins (VP1-3), 1F (row 2) for replication proteins (Rep 42, Rep 52, Rep 68, and Rep 78, and native dot blots A20 (row 3) and C37 (row 4) to identify intact capsids. (B) Histogram of the ratio of total AAV2 variant capsid to total AAV2 WT capsid (assembly ratio) determined by A20 ELISA (row 3). The variant capsids are grouped into three categories: 2-, 3- and 5-fold symmetry-related interfaces (refer to Fig. 1). Experiments were done in triplicates, and statistics were done via t test analysis versus the WT; *, P < 0.05; **, P < 0.001; ***, P < 0.0001. (C) Complementation assay of capsid assembly variants. The virus titer of variants supplied with AAP in trans were determined by anti-AAV2 ELISA and A20 native immunoblots. The variants screened are listed in Table 1 as potential AAP variants. The top row shows variants without AAP, and the bottom row shows variants supplemented with AAP. (D) SDS-PAGE and negative-stain electron micrograph of the AAV2-V221A/S224A variant complemented with AAP.

Western blot analysis confirmed the expression of VP1, VP2, and VP3 for wtAAV2 and variant capsids that were not defective in assembly (data not shown). All the assembly-deficient variants demonstrated a reduction in VP expression (data not shown). As stated before, AAP has been shown to be required for the assembly of the AAV capsid, and it is encoded by a cryptic start site present in VP2 with its ORF extending to VP3. Of the 25 variants generated, 7 (V221A/S224A, R294A, Q297A, R298A, R294A/R298A, H255A/K258A, andT331A) have mutations within the AAP ORF. The translation of all the AAV2 mutants generated using AAPs’ reading frame indicates that V221A/S224A, R294A, Q297A, R298A, R294A/R298A, H255A/K258A, and T331A were also potential AAP variants, with R298A and R294A/R298A generating stop codons. Taking this into consideration, complementation assays utilizing AAP supplied to the expression system in trans were performed and indicated that the only VP variant rescued was V221A/S224A, and therefore it was the only AAP variant generated by the VP mutagenesis (Fig. 2C). Purification of V221A/S224A rescued with AAP demonstrated VP1, VP2, and VP3 expression in the expected 1:1:10 ratio, respectively, and the assembly of the T=1 capsid (Fig. 2D).

Capsid symmetry-related interfaces play an important role in genome packaging and infectivity.

The genome packaging ratio of wtAAV2 compared to that of the VP variants was determined by quantitative PCR (qPCR) and confirmed by alkaline agarose gel electrophoresis. Several 3- and 5-fold variants were shown to be defective for packaging compared to wtAAV2 (Table 1; Fig. 3A). For example, R432A, L510A, N511R, and P602A, which are all involved in 3-fold symmetry-related interactions were severely defective for genome packaging compared to wtAAV2 (Table 1; Fig. 3A). This is significant, as the 5-fold axis and not the 3-fold axis has been previously shown to be important for genome packaging and VP1 externalization (8, 34), although a previous cryo-EM reconstruction of AAV2-R432A capsid revealed the loss of both intra- and intermolecular hydrogen-bonding interactions, along the 3-fold axis, which was propagated toward the 5-fold axis by the arginine to alanine substitution. The loss of these interactions was shown to cause destabilization of the capsid, which was confirmed by differential scanning calorimetry (44). It was proposed that the destabilization of AAV2-R432A capsid was a major factor in the loss of the virus’s ability to stably package genome without being compromised (44). This concept also seems to hold true for the 3-fold symmetry-related variants, which are also significantly less stable than wtAAV2.

FIG 3.

Comparative analysis of purified wtAAV2 and variant packaging and infectivity phenotype. (Top) Histogram of DNA packaging ratio obtained for A20-positive variants. The packaging ratio is determined by dividing the A20 value of capsid/mL with quantitative PCR or (qPCR) values of genome/mL. (Bottom) Histogram of the relative light units (RLU) for each A20-positive variant compared to WT AAV2. Experiments were done in triplicates, and statistics were done via t test analysis versus the WT; *, P < 0.05; **, P < 0.001; ***, P < 0.0001.

To further determine the phenotype of the variants generated compared to wtAAV2, their infectivity was measured by luminescence of the packaged luciferase transgene. Normalized luminescence of the variants compared to wtAAV2 showed that 7 of the variants were 1 to 2 logs defective for infectivity (Fig. 3B; Table 1). These VP variants are spatially located over the entire symmetry-related interfaces. It has previously been shown that L510A and N511R, which are important for the interaction with the secondary receptor integrin α_v β₁ binding site, are also important for genome packaging as well as infectivity defects (45). In addition, the amino acid NGR (511 to 513) sequence has also been predicted to be involved in the α_v β₁ integrin interaction (46). There is no clear correlation between the symmetry-related position of these variants and their effect on infectivity, and therefore, infectivity may be more closely related to receptor binding, viral trafficking, or genome release (uncoating) than a particular capsid assembly symmetry axis.

Symmetry-related interactions effect capsid morphology and stability.

To confirm the presence of assembled capsids, purified variants were visualized by negative-stain EM (Fig. 4). The VP content and purity were verified by SDS-PAGE. All assembly variants expressed VP1 to VP3 in the expected ratio. Negative-stain EM confirmed the presence of assembled capsids for each purified variant. The variant capsids all appeared to have different morphologies. Some capsids were stain penetrated, for example, L510A, while others were not, for example H255A/K258A, and others contained a mixture of stain and not stain penetrated, for example, D625A (Fig. 4). The AAV variants that were defective for assembly had at least 1 log fewer total number of particles than wtAAV2. These variants did not appear to have a T=1 capsid morphology, for example, AAV2-F415A. There is no direct correlation between capsid gross morphology and assembly; the micrographs, however, do confirm the presence of viral capsids and eliminate the possibility of purified VP oligomers/intermediates.

FIG 4.

Qualification of purified wtAAV2 and variant capsids. For each variant panel, on the left is the Coomassie-stained SDS-PAGE and on the right the corresponding negative-stain EM.

To ascertain the role of capsid stability on the packaging and infectivity of the capsid variants, the thermal profiles and melting temperature (T_m) were determined using differential scanning fluorimetry (DSF). For a control reference, the T_m of wtAAV2 was determined to be 67°C, which is consistent with the previously determined value (47). Somewhat surprisingly, the T_m of the 2-fold symmetry-related K692A variant was ~2°C more stable than that of wtAAV2. At the same time, the T_m of L510A (3-fold symmetry variant) was ~8°C, and the R404A (5-fold symmetry variant) T_m of ~3°C was less stable than that of wtAAV2 (Fig. 5). The 2-fold symmetry variant K692A, with the least buried surface and therefore expected to have the least effect on capsid stability, was slightly more stable than wtAAV2. Also, it was previously shown that the melting temperature of the capsid can be very sensitive to amino acid substitutions; in AAV6 a lysine at position 531 is destabilizing, whereas in AAV1 a glutamic acid at the same position is stabilizing (47). Similarly, both AAV2-TT and AAVv66 compared to AAV2 have serine and threonine instead of arginine at positions 585 and 588. AAV-TT and AAVv66 are approximately 10°C more stable than AAV2 at physiological pH (48, 49). These data imply that maybe charged residues may have a stabilizing, effect while positive residues would have a destabilizing effect on the capsid, and therefore K692A would be predicted to be less stable than wtAAV2. However, in AAV2 the K692A variant also interacts with F398, which may facilitate a slight stabilization when the lysine is substituted by an alanine, which is more hydrophobic (Fig. 5A, right). Similar to the R432A variant, L510A and N511R are located at a 3-fold symmetry interface, and both destabilize the capsid by ~8°C compared to wtAAV2. The change of an arginine to an alanine, if charge alone was a factor, should have a stabilizing effect on the virus capsid. However, charge is probably only important in the context of surrounding amino acids, and this study would imply that it is the loss of both inter- and intramolecular interactions at the symmetry interfaces that caused the destabilization of the capsid. There is no structure available for the L510A capsid variant; however, this amino acid is at the 3-fold symmetry-related interface, which has numerous interactions that are disrupted when the leucine is changed to an alanine and could cause the destabilization of the capsid. The DSF data showed that all the 5-fold variants were less stable than the wtAAV2, each amino acid substitution being associated with several stabilizing interactions that are disrupted when replaced with an alanine (Fig. 5C).

FIG 5.

Thermal profile of wtAAV2 and variant capsids. (A) wtAAV2 (black) and K692A (2-fold symmetry interface variant; gray) (refer to Fig. 1A and B). (B) wtAAV2 (black) and 3-fold symmetry interface variants (gray) (refer to Fig. 1C and D). (C) wtAAV2 (black) and 5-fold symmetry interface variants (gray). The normalized RFU is plotted against temperature (°C), and the peak temperature represents the melting temperature (T_m) of the capsid.

Structure function correlation of AAV2-fulls, AAV2-empties, and AAV2-R404A capsids.

To determine the effects of the individual amino changes on the structure of the variant and the correlation to the capsid physical characterization/function, the structure of wtAAV2-fulls (packaging DNA) (PDB ID: 8FYW), wtAAV2-empties (void of DNA) (PDB ID: 8FZ0), AAV2 variant (AAV2-R404A [5-fold variant] [PDB ID: 8FZN]), and AAV2-N511R (3-fold variant), which were severely defective for infectivity and genome packaging respectively, were determined by cryo-EM reconstruction. The AAV2-N511R fitted model was structurally identical to the wtAAV2 model, and therefore it is not included in this analysis. The samples were purified by AVB column chromatography, which does not separate DNA containing (fulls) from DNA void of (empties) capsids. The full and empty capsids were separated at the particle extraction and classification steps of the micrograph data processing. The fulls had dark and the empties had light density centers, with both exhibiting a dark outer ring indicating the protein capsid shell (Fig. 6A and B). There were significantly fewer full capsids in the AAV2-R404A sample, so the structural analysis was only done on the empties. The resolutions of the final maps of wtAAV2-fulls, wtAAV2-empties, and AAV2-R404A were 2.85, 2.95, and 3.62 Å, respectively (Fig. 6C and Table 2). All 3 capsid structures displayed a morphology common to all AAVs determined to date (reviewed in reference 25). In brief, the capsids surface topology has protrusions surrounding the 3-fold axis, depressions at the 2-fold axis, and turrets surrounding the 5-fold axis or 5-fold pore (Fig. 7A).

FIG 6.

Cryo-EM of wtAAV2-full, wtAAV2-empty, and AAV2-R404A. (A) Representative micrograph showing wtAAV2-full labeled with a white arrow and empty capsids labeled with a gray arrow; scale bar = 633 Å. On the right are the corresponding 2D classes generated for the wtAAV2-fulls and -empties. (B) Representative micrograph showing AAV2-R404A capsids, which are predominantly empties on the left, and the corresponding 2D classes on the right. (C) FSC plots of wtAAV2-full (gray), -empty (cyan), and AAV2-R404A (orange), with resolutions of 2.84, 2.94, and 3.45 Å, respectively, at an FSC threshold of 0.143.

TABLE 2.

Summary of cryo-EM structures

Characteristic	wtAAV2		AA2-R404A
	Fulls	Empties	Empties
PDB ID	8FYW	8FZ0	8FZN
EMDB ID	29598	29600	29636
Total no. of micrographs	1,674	1,674	1,384
Defocus range (μm)	1.0–2.5	1.0–2.5	0.8–3.5
Electron dose (e-/Å2)	60.0	60.0	52.4
Frames/micrograph	50	50	32
Pixel size (Å/pixel)	1.09	1.08	1.01
Particles used for final map	3,838	6,782	23,783
B-factor used for final map (Å2)	–1/25	–1/25	–1/25
Resolution of final map (Å)	2.84	2.94	3.62
Residue range	217–735	217–735	237–735
Map CC	0.833	0.854	0.825
RMSD (bonds) (Å)	0.009	0.009	0.009
RMSD (angles) (°)	0.926	0.968	0.760
All-atom clash score	7.47	8.94	8.10
Ramachandran Plot
Favored (%)	96.2	94.6	92.8
Allowed (%)	3.8	4.9	6.4
Outliers (%)	0.0	0.5	0.8
Rotamer outliers (%)	0.3	0.2	0.0
C-β deviations	0	4	0

FIG 7.

Cryo-EM model and map of wtAAV2-full, wtAAV2-empty, and AAV2-R404A. (A) Radially colored surface map showing icosahedral 2-, 3-, and 5-fold axes, 2/5-fold wall, and 3-fold protrusions of the capsids. (B) Color key and cross section of the capsids. Figures were generated using Chimera (75). The atomic model of amino acids (aa) 408 to 412 of wtAAV2-full (gray, left) and -empty (cyan, right) AAV2-R404A fitted into their corresponding EM density maps (black mesh). The map is contoured at a sigma (α) threshold of 3.0. (C) The atomic model of the amino acids (aa) 408 - 412 of wtAAV2-full (gray, left), wtAAV2 -empty (cyan, middle), and AAV2-R404A (orange, right) fitted into their corresponding EM density maps (black mesh). (D) The atomic model and corresponding EM density map of aa 419-421 and 628-632 of wtAAV2-full (gray, left) with the dinucleotide (dCMP and dAMP), wtAAV2-empty (cyan, middle), and AAV2-R404A (orange, right). (E) The atomic model and EM density map of aa 403-405 and 336-339 of wtAAV2 -empty (cyan, left), and AAV2-R404A (orange, right). The maps are contoured at a sigma (α) threshold of 3.0.

This contrasts with the interior of the capsids, where the wtAAV2-fulls had electron density consistent with the presence of the packaged genome. There was also additional density in the wtAAV2-full and -empty capsids within the 5-fold channel, which was absent from the interior of the AAV2-R404A capsids. This density within the 5-fold channel is consistent with previous observations and assigned to amino acids 204 to 210, the N terminus of VP3 (50–52). The resolution of the EM maps was not sufficient to unambiguously assign these residues. However, the resolution of the EM maps was suitable for the accurate fitting of the remainder of the VP3 common region (Fig. 7C) and the confirmation of the arginine to alanine substitution at position 404 (Fig. 7E), consistent with previously determined AAV structures that compared full and empty capsids. The wtAAV2-fulls have electron density in the capsid interior that extends into a conserved nucleotide binding pocket close to the icosahedral 3-fold axis. This ordered nucleotide density is inserted between P419 and S422 and H629 and P630 and is flanked on either side by weaker unassigned nucleotide density, which has also been observed in other genomes containing AAVs (51, 53). A nucleotide has been observed in the same DNA binding pocket in AAV1, AAV3b, AAV4, AAV6, AAV7, AAVrh.10, AAV11, AAV12, AAV13, and AAVrh32.33 structures but not in AAV2, AAV5, and AAV9. The VP residues involved in the interactions are conserved in the AAV serotypes, and it has been proposed that bound nucleotide may play a role in the capsid/DNA stabilization, genome packaging, or capsid assembly (50). Though it is not clear whether this pocket is essential for these roles, the structures that were determined by cryo-EM in which the fulls were separated from the empties do not have an ordered nucleotide in the empty capsid; this is consistent with our data (51, 54). There is no ordered nucleotide density seen in the wtAAV2 empties or AAV2-R404A empties (Fig. 7D). This would imply that the nucleotide is not important for capsid assembly but may have a role in genome packaging.

The atomic models, generated from the structures, were built into their corresponding electron density maps based on wtAAV2 model (PDB ID: 6U0V). The VP1u, the VP1/VP2 common region, and the N terminus of VP3 are disordered and were not observed in any of the three structures, similar to all previously determined AAV structures. The N-terminal residue modeled for wtAAV2-fulls and -empties and AAV2-R404A are 217, 217, and 236, respectively. The Cα position of the monomers of wtAAV2 fulls and empties superposes with a root mean square deviation (RMSD) of 0.35 Å with differences observed for several residues facing the lumen of the capsid; for example, H229 adopts a dual confirmation in the wtAAV2-empties and a single conformation in the wtAAV-fulls. Comparatively, the superposition of the Cα position of AAV2-empties with AAV2-R404A VP has an RMSD of 0.87 Å. This variability observed between wtAAV2 empties and AAV2-R404A is based largely on the flexibility of additional amino acids 217 to 237 in the N terminus of AAV2-R404A (Fig. 8A and B). The consequence of the disorder of the additional N terminus of AAV2-R404A is the opening of the base of the 5-fold pore with an 11.6-Å distance away from the symmetry axis (Fig. 8C and D). In addition, the absence of the side chain for amino acid 404 in the AAV2-R404A capsid (Fig. 7E) resulted in the movement of residues that previously interacted with R404 in wtAAV2-empties, namely, amino acids 336 to 338 (Fig. 7E). One of the major side chain movements observed in the R404A variant capsid was L336, which interacts with V221, which has moved away from the 5-fold axis and therefore facilitates the widening of the channel. Disordering of the N terminus of VP3, which is located under the base of the 5-fold channel, has also been observed for the acidic pH structures of snake AAV and two AAV2 capsid variants, AAV2-L336C and AAV2-R432A (44, 55). The first ordered residue of AAV2-L336C is G226, and the first ordered residue of AAV-R432A is S232. Both variants are defective for genome packaging, and AAV2-L336C has altered VP1u externalization. In both, AAV2-L336C and AAV2-R432A residue R404A shift radially outward, which causes an equivalent shift of the chain containing L336. L336 is conserved in all parvovriuses and is involved in a hydrophobic interaction with V221, which is believed to affect genome retention (56). This loss of order of the N-terminal loop has been previously proposed to block the externalization of VP1u, which contains the phospholipase A domain and therefore is important for endosomal escape and is the likely explanation for the loss of infectivity of the R404A variant (34).

FIG 8.

Comparative structures of wtAAV2-full, wtAAV2-empty, and AAV2-R404A. (A) Ribbon diagram of the monomers with wtAAV2-full (gray) on the left, -empty (cyan) in the middle, and AAV2-R404A on the right. The conserved eight-stranded anti-parallel β-barrel core with β-BIDG and β-CHEF sheet and αA helix are labeled. The icosahedral 2-, 3-, and 5-fold axes are denoted by an oval, a triangle, and a pentagon, respectively. The N-terminal and C-terminal residues and DE and HI loops are also labeled. (B) Zoom-in of the N-terminal G217 to V239, which are missing amino acids 217 to 235. (C) Stick diagrams of a section of two opposing 5-fold symmetry-related monomers. The distance between the top and bottom of the 5-fold pore is shown in Angstroms. (D) Surface diagram of a pentamer of wtAAV2-empties and AAV2-R404A viewed from the inner surface of the capsid. The N-terminal of wtAAV2-empties is colored cyan and the AAV2-R404A is colored orange. Figures were generated in Chimera (75).

In conclusion, this study suggests that the 2- and 5-fold interface interactions are critical for AAV capsid assembly (Fig. 9A). Compilation of previously identified amino acids crucial for AAV capsid assembly also aligns at the 2- and 5-fold interface except for 2 amino acids close to the 3-fold axis (Fig. 9B), and overwhelmingly points to these interfaces as critical to the assembly of the capsid (12, 13, 34, 38, 57). Rational design approaches for AAV capsid engineering utilize strategies such as capsid point mutations, domain swapping, and chemical modification in order to modulate, improve, optimize, and enhance the therapeutic efficiency of the vector (reviewed in reference 58). Systematic mutational analysis of the symmetry-related interactions of the AAV2 capsid has provided information on the regions of the capsid that should not be modified, for example, the 2-fold interface and some regions of the 5-fold symmetry-related interface. Conversely, the 3-fold symmetry-related interactions can tolerate modifications to improve the efficacy of the vector with a minimal effect on capsid assembly. Furthermore, this study also indicates that the 3-fold symmetry-related interactions may also play a role in genome packaging or retention.

FIG 9.

Determinants of wtAAV2 capsid assembly. (A) Surface representation of the wtAAV2 monomer shown on the left and the 60mer on the right, with the amino acid positions that affect capsid assembly shown as spheres. Amino acid positions that were generated in this study and are located in the 2-fold symmetry-related interface are colored yellow, and the 5-fold symmetry-related interactions are colored orange. (B) Surface representation of the wtAAV2 monomer shown on the left and the 60mer on the right, with all amino acid positions that affect capsid assembly shown as spheres. The amino acids described in panel A are also shown. The previously reported amino acids that are important for the structural integrity of the capsid are colored blue. The amino acids that are not in the 2-fold or 5-fold symmetry are shown as a red broken circle. The icosahedral 2-, 3-, and 5-fold axes are denoted by an oval, a triangle, and a pentagon, respectively. Figures were generated in PyMOL (81).

Previous studies show that the key capsid assembly intermediate for other icosahedral ssDNA viruses, for example, the Microviridae, with the prototype member ФX174, and the Geminiviridae, maize streak virus, are pentamers (59, 60). In fact, sedimentation analysis of soluble cytoplasmic fractions of newly synthesized AAV2 VP1, VP2, and VP3 showed oligomerization consistent with VP monomers and pentamers but no assembled capsids (61). The data presented here, however, confirm the importance of the 2-fold and 5-fold symmetry-related interface on capsid assembly, with the redundancy of the 3-fold axis, which makes it more tolerant of amino acid substitutions, and peptide insertions and deletions. Interestingly, trimeric subassemblies have been reported to be important for capsid formation of the autonomous parvoviruses minute virus of mice (MVM) and canine parvovirus, which belong to the Protoparvovirus genus. Structural analysis of all AAV assembly mutants to date compared with those of MVM, in the context of a capsid composed of trimers or pentamers, indicates that disruption of the 5- and 2-fold interfaces of AAV is critically important for capsid assembly, while MVM appears to be affected by the 3- and 5-fold interfaces (62, 63).

MATERIALS AND METHODS

Identification of wtAAV2 capsid symmetry interactions.

The coordinates for wtAAV2 VPs were accessed from the Protein Data Bank (PDB) (PDB ID: 6U0V). The coordinates for the 2-, 3-, and 5-fold symmetry-related VPs were generated in the database VIPER and subroutine oligomer generator (64). The program PyMOL was used to generate the images of the dimer, trimer, and pentamer. A list of 21 variants was generated based on residues that were structurally conserved between AAV2 and other parvoviruses, including MVM, and another 4 residues were selected based on residues that were identified in previous mutagenesis studies that were believed to be important for AAV2 assembly (Table 1).

Recombinant wtAAV2 (rAAV2) and variant production.

The AAV2 plasmid pXRAAV2, which contains the AAV2 VP1, -2, and -3 sequences, and AAV2 replication proteins were used to generate the selected rAAV2 mutants. The mutations were generated by site-directed mutagenesis according to the kit protocols (Stratagene, catalog [cat.] no. 200524). HEK 293 cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with penicillin and streptomycin at 100 U/mL and 10% fetal bovine serum (FBS) in a 15 cm³-petri dish at 37°C and 5% CO₂. To produce wild-type (WT) and variant rAAVs, HEK 293 cells were triple transfected with pXRAAV2 or mutation (13 μg), pHelper (20.5 μg), which contains the adenovirus helper genes, and pTR-UF3-luciferase (18 μg), which contains both the reporter gene and the AAV inverted terminal repeats required for genome packaging. The transfection protocol was previously described (47), and the transfected HEK 293 cells were incubated for 72 h at 37°C. The HEK 293 cells were then harvested by centrifugation at 1,140 × g for 20 min. The supernatant was precipitated in 10% polyethylene glycol (PEG) and resuspended in 1 mL lysis buffer per 15-cm plate. The cell pellet was resuspended in 1 mL lysis buffer (1× phosphate-buffered saline [PBS], 1 mM MgCl₂, 2.5 mM KCl [TD buffer], and 0.5 M NaCl) for each 15-cm plate. The virus in the cell pellet was then released by three freeze/thaw cycles and treated with Benzonase to remove all the nucleic acid that was not encapsidated. The cell lysate and PEG precipitate resuspension were combined and clarified by centrifugation at 3,700 × g for 20 min. The clarified supernatants were diluted 1:5 with TD buffer and loaded onto a prepacked 1-mL AVB Sepharose HP column (GE Healthcare) at a flow rate of 0.5 mL/min. After loading, the column was washed with 20 mL of TD buffer. The virus was eluted with 10 mL elution buffer (0.1 M glycine-HCl, pH 2.7) and neutralized with 1 mL 1 M Tris-HCl at pH 10 (neutralization buffer). The elution fractions were concentrated, and buffer was exchanged in 1× TD buffer and concentrated to a final volume of 250 μL. The capsid integrity and purity were verified by SDS-PAGE and negative-stain EM, respectively.

Immunoblots and quantification of wtAAV2 and variant capsids.

The crude cell lysates were analyzed with immunoblots using a panel of antibodies designed to recognize the AAV2 denatured VPs (VP1 to -3) as well as the native capsid. The B1 antibody recognizes C-terminal residues of the AAV VPs (except AAV4), and based on the structure as well as epitope mapping of the capsid, these residues are only accessible when the capsid is denatured. The A20 and C37 antibodies were used to identify epitopes of the AAV2 which are only available when the capsid is in its native conformation. The 1F antibody was used as a protein expression control, as it recognizes all the AAV2 replication proteins in both native and denatured conformations. All primary antibodies were diluted 1:3,000 in 1% milk, and the secondary antibody used to detect the 1F, B1, C37, and A20 antibodies was the anti-mouse antibody diluted 1:5,000. The AAV2-antibody complex was detected with the ECL Western blot substrate (Amersham, cat. no. RPN2109). The immunoblots were treated as both denatured and native dot blots, and as the names suggest, the native samples were not boiled, and the denatured samples were boiled for 10 min at 100°C. The total capsid titer was determined by an A20 enzyme-linked immunosorbent assay (ELISA) according to the manufacturer’s instructions (American Research Products, cat. no. PRATV). The clarified cell lysates were serially diluted, and 100 μL was added to the kit well. Readings that were within the detection limit of the kit were used to calculate the quantity of all capsids, both empties and fulls.

Determination of the packaging and particle-infectivity ratio of wtAAV2 and variant capsids.

The total packaged genome or genome copy number was determined by quantitative PCR (qPCR). A total of 5 μL of crude lysate was treated with DNase I for 1 h at 37°C to degrade any DNA that was not encapsidated, and each sample was digested with 4 μL of proteinase K (Roche, cat. no. 1373196) and 20 μL of 10× proteinase K buffer (10 mM Tris HCl, pH 8.0, 10 mM EDTA, 10% SDS), and the solution was diluted with sterile distilled water to a total volume of 200 μL. The mixture was incubated for 1 h in a water bath at 37°C and treated twice with an equal volume of phenol-chloroform (Roche, cat. no. 1373196), and the upper aqueous layer was transferred to a clean Eppendorf container after each extraction. The aqueous fraction was then treated with chloroform (Fisher cat. no. C298-500), and the aqueous layer was transferred to a sterile Eppendorf container. The DNA was precipitated overnight at −20°C, after adding 10% NaOAc (sodium acetate) pH 5.2, 1 μL glycogen, and a 3× volume of 95% ethanol. The sample was then pelleted for 20 min at 16,000 × g. The pellet was air dried and resuspended with 20 μL water. Then, 2 μL of the viral DNA, 1 μL of primers (forward and reverse) to the luciferase gene, 12.5 μL of iQ SYBR green supermix, which contains Taq DNA polymerase (Bio-Rad, cat. no. 170-8882), were combined with sterile water to a total volume of 25 μL. The genome titer of the cell lysates was determined by quantitative PCR in an mYiQ2 thermocycler (Bio-Rad) (65).

The genome-containing titer of rAAV2-luc and mutant vectors, with a packaged luciferase transgene, was determined by qPCR as previously described (65). For the transduction assay, ~2.5 × 10⁴ HEK 293 cells were seeded in a 96-well plate 1 day prior to infection. On the day of infection, the rAAV2 and variant vectors were added to cells at a multiplicity of infection (MOI) of 10⁵ and were added to 90 μL DMEM supplemented with 2% FBS and 1% antibiotic-antimycotic (ABAM). The old medium on the cells was discarded, and the vector-medium mixture was added to the cells. After the cells were incubated at 37°C and 5% CO₂ for 48 h, the medium was removed, and the cells were washed with PBS. For the luciferase assay, the cells were lysed, and the luminescence was determined using the luciferase assay kit (Promega), according to the manufacturer’s protocol, on a Synergy HTX plate reader (BioTek).

Negative-stain electron microscopy.

To determine the gross morphology of the purified (wtAAV2 and variant) capsids, 5 μL of each was loaded onto a glow-discharged carbon-coated copper EM grid (Ted Pella, Inc., cat. no. 01754-f) for 2 min and negatively stained with 5 μL of 2% uranyl acetate for 20 s. The grids were visualized on a Tecnai G2 Spirit TWIN device equipped with a charge-coupled device (CCD) camera at a magnification of 40,000×.

Differential scanning fluorimetry.

To compare the stability of the wtAAV2 and variants, DSF was used. This method monitors binding of the dye SYPRO orange to exposed hydrophobic regions of the protein during denaturation and unfolding. For the analysis, 2.5 μL of 1% SYPRO orange (Invitrogen) was added to 22.5 μL of rAAV and mt capsids at ~0.1 mg/mL to make a total reaction volume of 25 μL. The assay was run using a Bio-Rad MyiQ2 thermocycler instrument, and the experimental temperature ranged from 30 to 99°C with temperature ramping of 0.5°C per step. The rate of change of fluorescence with temperature was recorded as an inverse thermal profile namely, –dRFU/dT versus temperature. The –dRFU/dT values were multiplied by −1 and normalized to 1 by dividing the raw values with the peak value for evaluation. The peak value recorded on the thermogram is the T_m. All experiments were conducted in triplicates.

Cryo-EM and data collection.

A total of 3 μL of purified AAV2 and AAV2-R404A capsids, at 0.5 mg/mL, was pipetted onto glow-discharged copper grids containing 2 nm continuous carbon support over 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 in-house for ice quality and particle distribution suitability for data collection on an FEI Tecnai G2 F20 -TWIN microscope operated at 200 kV and −20 e-/Ų dosage (low dose). The grids determined to be suitable for high-resolution data collection were used to collect micrograph movie frames using the application Leginon (66) on a Titan Krios electron microscope (FEI Co.). The data were collected at 300 kV with a DE20 (AAV2-R404A) and with a DE64 (AAV2) direct electron detector. Each movie frame was aligned with the DE_process_frames_software package (Direct Electron) with corresponding dark and bright reference images without radiation dose compensation as previously described (67, 68) for the data collected on the DE20. MotionCor2 was used to align movie frames collected on the DE64 detector with dose weighting (69); the data collection parameters are listed in Table 2.

wtAAV2 and AAV2-R404A structure determination, model building, and refinement.

Three-dimensional image reconstruction was done using the cisTEM software package (70). Aligned micrographs were imported, and their contrast transfer function (CTF) parameters were estimated (71). Micrographs that were astigmatic and of low quality were removed. Capsids were extracted using an automated particle selection program and with particle radius of 130 Å (72). The capsids were 2D classified using a box size of 1.5 times the diameter of the capsid, ideal classes were selected, and ice particles and other contaminants were discarded (73). Empty and genome-containing capsids, were separated on a second round of 2D classification using a smaller box size to classify based on the internal content. The genomes containing and empty capsids were then processed separately. An ab initio 3D model was generated, and auto refinement was used to generate the 3D density map. The resolution of the maps was estimated based on a Fourier shell correlation (FSC) of 0.143. The density map was sharpened with a precutoff B-factor value of −90 Ų and postcutoff B-factor value of 25 Ų. The sharpened density maps were visually inspected in the programs Coot and Chimera and were used for the assignment of the amino acid sidechains and main chains (74–76). Figures were generated using the UCSF-Chimera and PyMol programs (75–77).

The cryo-EM maps were converted from the MRC to XPlor map format using the e2proc3D.py subroutine in EMAN2 (78). The XPlor map was normalized and converted to a CCP4 format using the program MAPMAN (79). The maps were fitted with the model of AAV2 VP3 60mer) PDB ID: 6U0V) using the fit-in-map function in Chimera (75, 76). The voxel size of the maps at a maximum correlation coefficient was determined, and the models were saved relative to the maps. The VP models were manually built into the maps using the real-space-refinement subroutine in Coot (74). The maps were further refined using the subroutines rigid body, real space, and B-factor refinements in the program PHENIX (80), and the final refinement statistics are listed in Table 2.