Residues on Adeno-associated Virus Capsid Lumen Dictate Interactions and Compatibility with the Assembly-Activating Protein

Efforts to engineer the AAV capsid to gain desirable properties for gene therapy (e.g., tropism, reduced immunogenicity, and higher potency) require that capsid modifications do not affect particle assembly. The relationship between VP and the cofactor that facilitates its assembly, AAP, is central to both assembly preservation and vector production. Understanding the requirements for this compatibility can inform manufacturing strategies to maximize production and reduce costs. Additionally, library-based approaches that simultaneously examine a large number of capsid variants would benefit from a universally functional AAP, which could hedge against overlooking variants with potentially valuable phenotypes that were lost during vector library production due to incompatibility with the cognate AAP. Studying interactions between the structural and nonstructural components of AAV enhances our fundamental knowledge of capsid assembly mechanisms and the protein-protein interactions required for productive assembly of the icosahedral capsid.

ABSTRACT

The adeno-associated virus (AAV) serves as a broadly used vector system for in vivo gene delivery. The process of AAV capsid assembly remains poorly understood. The viral cofactor assembly-activating protein (AAP) is required for maximum AAV production and has multiple roles in capsid assembly, namely, trafficking of the structural proteins (VP) to the nuclear site of assembly, promoting the stability of VP against multiple degradation pathways, and facilitating stable interactions between VP monomers. The N-terminal 60 amino acids of AAP (AAPN) are essential for these functions. Presumably, AAP must physically interact with VP to execute its multiple functions, but the molecular nature of the AAP-VP interaction is not well understood. Here, we query how structurally related AAVs functionally engage AAP from AAV serotype 2 (AAP2) toward virion assembly. These studies led to the identification of key residues on the lumenal capsid surface that are important for AAP-VP and for VP-VP interactions. Replacing a cluster of glutamic acid residues with a glutamine-rich motif on the conserved VP beta-barrel structure of variants incompatible with AAP2 creates a gain-of-function mutant compatible with AAP2. Conversely, mutating positively charged residues within the hydrophobic region of AAP2 and conserved core domains within AAPN creates a gain-of-function AAP2 mutant that rescues assembly of the incompatible variant. Our results suggest a model for capsid assembly where surface charge/neutrality dictates an interaction between AAPN and the lumenal VP surface to nucleate capsid assembly.

IMPORTANCE Efforts to engineer the AAV capsid to gain desirable properties for gene therapy (e.g., tropism, reduced immunogenicity, and higher potency) require that capsid modifications do not affect particle assembly. The relationship between VP and the cofactor that facilitates its assembly, AAP, is central to both assembly preservation and vector production. Understanding the requirements for this compatibility can inform manufacturing strategies to maximize production and reduce costs. Additionally, library-based approaches that simultaneously examine a large number of capsid variants would benefit from a universally functional AAP, which could hedge against overlooking variants with potentially valuable phenotypes that were lost during vector library production due to incompatibility with the cognate AAP. Studying interactions between the structural and nonstructural components of AAV enhances our fundamental knowledge of capsid assembly mechanisms and the protein-protein interactions required for productive assembly of the icosahedral capsid.

INTRODUCTION

The nonautonomous adeno-associated virus (AAV) is not associated with disease, and thus there has historically been less incentive to study basic AAV biology in as great depth as that of human pathogens. Clinical success with AAV-based gene therapy vectors have elevated the interest in the study of AAV biology to inform vector engineering efforts. Within the family Parvoviridae, AAV represents one of the smallest and simplest mammalian viruses, AAV is a nonenveloped, 25-nm, T=1 icosahedral capsid containing only the 4.7-kb genome. AAV The single-stranded DNA (ssDNA) genome includes two genes, rep, which encodes four nonstructural proteins, and cap, which encodes the three structural proteins (VP). The different protein products are generated through alternative splicing (1) and noncanonical translation start codons (2, 3), such that the products of the respective gene share portions of overlapping identity yet remain functionally distinct. For example, VP3 is ∼530 amino acids in length and takes on a characteristic fold that forms the basic structural capsid monomer (4); VP2 and VP1 contain the entire VP3 sequence and same structural fold, but extend N-terminally by ∼65 and ∼202 residues, respectively. These N-terminal “tails” are disordered and situated in the interior of a newly assembled capsid (5); during the entry phase of infection, the VP1 tail is extruded (6) through a pore at the 5-fold axis of symmetry (7). This VP1 unique region contains a phospholipase domain critical for infectivity (8), whereas VP2 has been demonstrated to be dispensable for capsid assembly and virion function (9).

The native-state capsid is assembled from approximately 5 VP1, 5 VP2, and 50 VP3 proteins. The genome is inserted into preformed capsids through the 5-fold pore by rep gene products (10), but neither Rep nor the genome are involved in assembling the capsid (11). As a dependoparvovirus, AAV requires a helper virus, such as an adenovirus or herpes simplex virus (HSV) coinfection, to achieve gene expression, but capsid assembly and genome packaging can occur independently of the helper virus (11). Noninfectious virus-like particles (VLPs) devoid of the genome can be assembled solely of VP3 monomers, demonstrating VP3 as the minimal structural requirement for an AAV capsid. Despite the seemingly negligible requirements for AAV capsid assembly, the VP proteins cannot self-assemble into the icosahedron (12). Many processes required for assembly are orchestrated by a multifunctional ∼28-kDa viral cofactor whose discovery transformed our understanding of AAV virion production, the assembly-activating protein (AAP). AAP is encoded within the cap gene but translated from a +1 shifted reading frame from the VP open reading frames (ORFs) (13). The AAP of AAV serotype 2 (AAP2) is necessary to transport VP proteins to the nucleolus, the site of AAV2 capsid assembly, and it does so via a nucleolar localization signal (NoLS) within a basic region toward the C terminus of the AAP protein (14). However, nucleolar localization of VPs by fusing a NoLS to their N termini is not sufficient to assemble capsids (13), indicating other critical roles for AAP in addition to trafficking.

First demonstrated as essential for assembly of AAV2 capsids, AAP has now been investigated across a wide array of AAV serotypes in the context of native VP1/2/3 capsids and VLPs. Universally, AAP is required to achieve maximal production titers, but recent studies have shown that the absolute requirement for AAP across AAV variants ranges broadly. Some serotypes, such as AAV4 and AAV5, can assemble appreciable capsid quantities in the absence of any AAP, and many serotypes, such as AAV2 and AAV8, require the near-full-length AAP protein, but several serotypes only need the N-terminal 60 amino acids of AAP (AAPN) to assemble capsids, including AAV3 and AAV9 (15–17). A set of 9 putative ancestral capsid variants (AncAAVs), inferred by ancestral sequence reconstruction (18), show a similar range in requirement for AAP, as well as a range in functionality of their cognate AAP, although preserving the AAP reading frame was not taken into account in the design of these variants. For example, when AAP expression is truncated with an early stop codon in the AAP reading frame of one of these variants, Anc113, vector titers are increased, demonstrating a dysfunctional Anc113 cognate AAP (17). We harnessed the phenotypic range of AAP requirements across natural and AncAAVs to identify residues on VP at the trimer interface and a domain of AAP (residues 20 to 40) that are important for assembly functions (17). This phenotype-to-phylogeny mapping strategy helped elucidate that AAP functions to stabilize VP proteins from multiple degradation pathways as well as promote VP-VP interactions, and that AAP dependency stems from the various abilities of VP proteins to perform these roles on their own. The role of AAP2 in AAV2 VP protein stability was also shown to be critical for both mammalian and insect cell vector production systems (16).

AAP must interact with VP in a specific way to execute its numerous functions, but the nature of the specific AAP-VP interaction is not well understood. The AAPs of most serotypes are compatible with VPs of heterologous serotypes to assemble capsids in the absence of their cognate AAP, albeit to various degrees (15, 19); this promiscuity of AAP-VP compatibility across evolutionarily divergent AAV variants implies that the nature of the AAP-VP interaction is conserved. Dissection of individual linear domains of AAP revealed that the hydrophobic region (HR; residues ∼20 to 40) and the conserved core (CC; residues ∼40 to 60) regions of AAPN are critical for AAP-VP interaction and cross-serotype compatibility (20). Conversely, the remarkable tolerance of AAP to withstand recombination events and retain functionality (21) suggests equally critical roles for VP residues in AAP-VP cooperation. Here, we aimed to identify the potential AAP binding site on VP and delineate specific residues on both VP and AAP that are important for a functional AAP-VP interaction. In addition to deepening our understanding of AAV assembly from a basic virology standpoint, understanding how AAP and VP interact to promote maximal assembly efficiency can inform vector engineering strategies to produce higher-titer vectors for clinical use—treating a larger patient base while minimizing manufacturing costs. Moreover, considering the popularity of library-based discovery approaches where a large number of VP perturbations are examined at once, a universally functional AAP could hedge against overlooking VP variants that may otherwise be lost due to assembly or AAP incompatibility issues.

RESULTS

Compatibility with AAP2 ranges broadly across 14 capsid variants.

In previous work, we truncated AAP expression from rep-cap constructs and observed three assembly phenotypes across 21 AAV variants, AAP independent (able to assemble without any AAP), AAP dependent (requires a full-length AAP to assemble), and AAPC (the C-terminal 120 amino acids of AAP) independent (able to assemble with only AAPN) (17). These phenotypes, when correlated to the reconstructed phylogeny, revealed an AAPC-independent branch diverging from an entirely AAP-dependent branch. This pattern of phylogenetically contiguous variants among which there is a phenotypic switch informed us which serotypes to compare in an alignment to identify VP residues that share identity strictly within their phenotypic group. Mutating the 12 residues identified by this alignment resulted in a gain of AAPC independence, revealing a minimal motif that enables VP to assemble using only AAPN. For simplicity, we named this entire process phenotype-to-phylogeny mapping. To begin investigating the nature of the VP-AAP interaction, we applied the phenotype-to-phylogeny mapping strategy, using AAP2 trans complementation of 14 AAP-dependent AAV variants as a phenotypic readout to map residues on VP important for AAP compatibility. HEK293 cells were quadruply transfected with a helper plasmid, an inverted terminal repeat (ITR)-flanked reporter genome, rep-cap plasmids in which AAP expression was truncated with an early stop codon (AAPstop60, a mutation that is silent in the VP reading frame), and with either an empty plasmid or a cytomegalovirus (CMV)-driven AAP2 expression construct (pAAP2) (Fig. 1A). The vector was quantified by quantitative PCR (qPCR) on DNase-resistant particles (DRP) and is reported as a percentage of each capsid variant’s wild-type (WT) AAP titer, produced in parallel from rep-cap plasmids without the AAPstop60 mutation (WT AAP) (Fig. 1A and 1B). Results indicate a broad range in ability for AAP2 to assemble heterologous AAP-dependent serotypes. Variants like Anc80, Anc82, and Anc83 function as well or better with pAAP2 than their endogenous AAPs, whereas Anc81 and Anc127 show no rescue by pAAP2. The relative rescue by pAAP2 was determined as the difference between AAPstop60 titers in the presence or absence of pAAP2 (Fig. 2A, difference between black and gray bars in Fig. 1B). Note that Anc83 does not conserve the predicted CTG start codon for AAP, so the noncanonical mechanisms through which AAP is typically expressed likely function suboptimally. Anc83 has very low WT AAP titers, typically around 1e9 genome copies (GC)/ml. Truncating expression of AAP, as in Anc83AAPstop60, makes little difference in this already low titer and would otherwise fit the definition for an AAPC-independent serotype (producing at or above 10% WT titer in the AAPstop60 context). However, addition of AAP2 in trans rescues Anc83 titers by an average of 100-fold. Reporting this as a % WT AAP titer suggests a hyper-productive functionality for AAP2 and Anc83, yet titers are in the normal range for crude AAV vector preparations (albeit at the higher end, around 1e11 GC/ml). Thus, Anc83’s apparent AAP independence and hyperassembly by pAAP2—while clearly demonstrating high functionality between Anc83 VPs and AAP2—is a result of normalization and is likely indicative of low endogenous AAP83 expression.

FIG 1.

Compatibility with AAP2 ranges broadly across 14 capsid variants. (A) Schematic of rep-cap and AAP2 expression constructs used in vector production. Black arrows, transcription start sites at p5, p19, and p40 viral promoters; gray arrows, cap gene product translation start codons. The early stop codon (red) was introduced by site-directed mutagenesis (∼60 amino acids [aa]) into the AAP ORF. (B) Crude vector produced from WT or AAPstop60 rep-cap constructs with empty plasmid or pAAP2 in trans was titrated by quantitative PCR (qPCR) to quantify DNase-resistant particles (DRP). AAPstop60 titers (gray bars) and AAPstop60 + pAAP2 titers (black bars) are reported as percentages of their respective WT AAP titers and represent the average of at least 3 independent experiments ± standard error of the mean (SEM). Blue lettering indicates consistently low WT AAP titer. †, not detectable. A paired one-tailed t test was conducted comparing the AAPstop60 titers to their own AAPstop60 + AAP2 (*, P < 0.05; **, P < 0.01; ***, P < 0.005).

FIG 2.

Residue 413 contributes to AAP2 compatibility and VP-VP interactions. (A) Relative rescue of AAPstop60 constructs by pAAP2, as determined by the given formula and using values in Fig. 1B. Blue boxes, highly AAP2 compatible (relative rescue, ≥100%); Red boxes, AAP2 incompatible (relative rescue, <1%). (B) AAP2 compatibility phenotypes correlated to previously reconstructed phylogeny. Variants in bold indicate AAP-dependent serotypes, which comprise the set evaluated in this study. Nodes and branches are colored according to AAP2 compatibility phenotype. (C) Partial alignment of Anc80, 81, 82, and 83 VP1 sequences. Sequence highlight color corresponds to AAP2 compatibility phenotype. Arrowhead indicates the residue of interest. (D) Crude vector preps produced from the indicated AAPstop60 constructs and either pAAP2 or FLAG-AAP2 added in trans were quantified by qPCR on DRP. Titers are reported as a percentage of AAV2 WT AAP or Anc81 WT AAP vector titer (indicated on y axis), produced in parallel. Graphs represent the average of 3 independent experiments ± SEM. †, not detectable. A paired two-tailed t test was conducted comparing AAV2AAPstop60+pAAP2 rescue titers to FLAG-AAP2 rescue titers (black bars) and comparing Anc81 to Anc81E413Q titers under each condition (red and blue bars) (*, P < 0.05; **, P < 0.01; ***, P < 0.005). (E) Schematic of expression constructs for HA-VP1 and FLAG-AAP2. (F) HEK293 cells were transfected with FLAG-AAP2 and either WT Anc81 HA-VP1 or E413Q HA-VP1, as indicated above each lane. Immunoprecipitation (IP) was performed on cell lysates using an anti-HA antibody. VP were detected by Western blot analysis using the B1 antibody and AAP was detected with anti-FLAG (anti-FLAG blots are long exposures).

Identifying a candidate VP residue important for AAP2 compatibility.

We next correlated the AAP2 compatibility phenotypes to the reconstructed phylogeny we generated previously using Ancestral Sequence Reconstruction (18), First, the relative rescue by pAAP2 was determined for each serotype as the difference between each variant’s AAPstop60 titer in the presence or absence of pAAP2 (Fig. 2A). Serotypes were further categorized based on their relative rescue by AAP2 (blue, above 100% of WT AAP titers; gray, between 1 to 100% of WT AAP titers; red, below 1% of WT AAP titers) (Fig. 2A). Considering the high percentage in sequence identity shared between four variants with the highest or lowest AAP2 compatibility, a striking differential emerges along the Anc80 to Anc83 lineage, where Anc81 is uniquely nonfunctional among highly AAP2-compatible variants (Fig. 2B). Alignment of these four variants’ capsid sequences reveals a single residue whose identity is conserved only among the three variants that function with AAP2—a glutamine in Anc80, Anc82, and Anc83 is a glutamic acid in Anc81 at position 413 (Fig. 2C, arrowhead). Note that Anc127 is not phylogenetically contiguous with the other variants displaying extreme high or low AAP2 compatibility and thus was not included in the alignment. We hypothesized that mutating Anc81E413 to glutamine (Q) would restore rescue of AAPstop60 constructs by pAAP2. Indeed, pAAP2 rescues Anc81E413QAAPstop60 to 5.6-fold higher titers than Anc81AAPstop60 (Fig. 2D), but this mutation alone is not sufficient to reach pAAP2 rescue levels comparable to those of Anc80, Anc82, and Anc83 AAPstop60, which rescue to near or above 100% WT AAP titers with pAAP2.

Considering that pAAP2 did not complement its homologous VP, AAV2AAPstop60, to 100% of WT AAP titer (Fig. 1B), yet other groups have reported ∼100% rescue in their hands (15, 16), we wanted to address whether the disparity is due to the pAAP2 expression construct. Moreover, biochemical studies on the AAP-VP interaction require a tagged AAP in lieu of commercially available antibodies. An N-terminally FLAG-tagged AAP2 (FLAG-AAP2) rescued AAV2 AAPstop60 production to 82% of AAV2 WT AAP titer, a marked improvement over pAAP2 (Fig. 2D, black bars). FLAG-AAP2 rescues Anc81AAPstop60 production to a higher degree than pAAP2, and rescued Anc81E413Q to yet a higher degree (Fig. 2D). This mirrors the Anc81 versus Anc81E413Q trend observed with pAAP2, albeit to a lesser fold difference (3.3-fold with FLAG-AAP2 versus 5.6-fold with pAAP2) but a nonetheless significant increase. The nature of the improved trans complementation by FLAG-AAP2 over pAAP2 is unclear, as both are CMV-driven but within different vector backbones [pIRES-Ago2 versus pCDNA3.1(−), respectively]. One explanation is that the tag contributes to an increase in AAP stability so that more AAP is available to stabilize, transport, and oligomerize VP proteins, leading to a greater vector yield. Further studies are required to elucidate the mechanisms of improved FLAG-AAP function that are beyond the scope of the present study.

VP residue 413 is important for VP-VP interactions.

To assess whether the gain of function in Anc81E413Q versus WT Anc81 trans complementation by AAP2 is a result of increased interactions between AAP2 and the VP proteins, coimmunoprecipitations were performed using human influenza hemagglutinin (HA)-tagged VP1 as bait to pull down FLAG-AAP2. N-terminal HA-tagged VP1 constructs were generated, including the AAPstop60 mutation and mutated VP2 and VP3 start codons to silence their expression (Fig. 2E), although leaky expression of VP3 and some VP2 persisted at a low level (Fig. 2F, input lanes). Both WT Anc81 and E413Q mutant HA-VP1 coprecipitated an equivalently low quantity of FLAG-AAP2, but E413Q was able to coprecipitate an appreciably larger amount of VP3 protein despite lower VP3 abundance in the input (Fig. 2F, IP lanes). In line with the increase in AAP2 rescue titers observed for Anc81E413Q, these observations suggest that in the presence of AAP2, E413Q mutant VP proteins form more stable and/or more numerous VP-VP interactions than those of WT Anc81 VP proteins,

Capsid lumenal surface charge influences AAP-VP assembly compatibility.

To examine the role of residue 413 in AAP-VP cooperation, we modeled Anc81 residue identities onto the structure of the AAV8 trimer (PDB identifier [ID] 2QA0 [22]) (Fig. 3A). E413 is located on the highly conserved beta-barrel structure and points toward the capsid lumen (Fig. 3B). This glutamic acid presumably contributes to a wide patch of negative charge predicted on the Anc81 capsid lumenal surface (Fig. 3C, green arrow). The E413Q mutation is predicted to neutralize a significant portion of this negative charge (Fig. 3D, green arrow), and, when combined with mutation of another nearby glutamic acid, E684, to glutamine, this negative patch is greatly diminished (Fig. 3D and E, green arrowheads), and a positively charged pocket is predicted to appear adjacent to this site (Fig. 3D and E, brown arrowheads). To test the effect of a less negative Anc81 lumenal surface on AAP2 compatibility for capsid assembly, a panel of mutants was generated to target E413, E684, and an additional neighboring glutamic acid at residue 686 (Fig. 3B). These sites on Anc81 WT AAP and AAPstop60 constructs were mutated to glutamine to test neutral charge, or to lysine to test increasingly positive surface charges, and vector was produced in the presence or absence of FLAG-AAP2 (Fig. 3F). WT AAP titers obtained for most of these mutants were very similar to those of Anc81 (within 20%) (Fig. 3F, white bars). In addition, all mutants remained highly dependent on AAP (Fig. 3F, gray bars). Among the producing mutants, various degrees of rescue by FLAG-AAP2 could be observed (Fig. 3F, black bars). While FLAG-AAP2 rescues the E413Q mutant 3.3-fold better than WT Anc81, the E413/684Q double mutant rescues 6.4-fold better than the WT. Conversely, the E413Q/E686K double mutant produces nearly a full order of magnitude lower titer in the WT AAP context, and AAP2 fails to rescue this mutant. Even more dramatic is the loss of vector in all AAP contexts of the E413Q/E684K/E686K triple mutant.

FIG 3.

Capsid lumenal surface charge influences AAP-VP assembly compatibility. (A) Approximate model of an Anc 81 VP trimer viewed from lumenal surface with each monomer differentially colored. Inset shows boundaries of zoomed in panels B through E. (B) Ribbon representation depicting the conserved VP beta-barrel structure. Glutamic acids of interest are listed and highlighted the corresponding color on the structure. (C to E) Surface charge predictions for (C) Anc81, (D) the Anc81 E413Q single mutant, and (E) the Anc81 E413Q/E684Q double mutant. Charges correspond to scale at left and were generated by the Adaptive Poisson-Boltzmann Solver (APBS) plugin using PyMol. Green arrows and arrowheads indicate negative patches predicted to disappear with mutagenesis. Tan arrowheads indicate a positive pocket predicted to appear with mutagenesis. (F) Vector produced from quadruply transfected HEK 293 cells (adenovirus helper, inverted terminal repeat [ITR]-flanked reporter genome, rep-cap [with appropriate mutations], and either FLAG-AAP2 or empty plasmid) was quantified by qPCR on DRP and is reported as a percentage of Anc81 WT AAP titers produced in parallel. Graph represents the average of at least four independent experiments ± SEM. †, not detectable. A paired two-tailed t test was conducted comparing the mutant rescue titers to that of the wild type (*, P < 0.05; **, P < 0.01; ***, P < 0.005; ‡, at least one titer used in analysis is below background levels). (G) Whole-cell lysates from HEK293 cells transfected as in (F) were harvested at 24 h and electrophoresed (25 μg total protein per lane), and VP protein detected by Western blotting with the B1 antibody. VP mutants are listed below lanes and AAP context is above; WT, WT AAP; s60, AAPstop60; rF2, AAPstop60+FLAG-AAP2.

We have shown previously that in the absence of AAP, rescuing VP protein by pharmacologically inhibiting its degradation does not result in a concomitant rescue in vector titer, demonstrating that AAP is required for both stability and promoting VP-VP interactions (17). To examine whether the beta-barrel mutants’ gain or loss in AAP2 rescue titers correlates with increased or decreased VP protein, we interrogated VP protein levels in whole-cell lysates 36 h after transfection with helper, rep-cap, and empty or FLAG-AAP2 plasmids (Fig. 3G). VP levels mirror vector titers, highlighting the importance of the beta-barrel for AAP to exert its multiple functions. These results support a hypothesis that VP-AAP cooperation to promote VP stability and to assemble the capsid relies on a tight balance of charges and neutrality in the VP regions forming the capsid lumen.

Neutralizing a lumenal negative patch robustly increases AAP-VP and VP-VP coprecipitation.

One explanation for to the gain of assembly function observed in the glutamine mutants is that the VP beta-barrel is a binding site for AAP, and neutral (yet polar) surface residues allow for favorable interactions with AAP2. To test this directly, we performed immunoprecipitations (IP) in both directions, pulling down on the glutamine mutant panel of Anc81 HA-tagged VP1 proteins and interrogating FLAG-AAP2 coprecipitation and pulling down on FLAG-AAP2 and interrogating VP coprecipitation (Fig. 4). Constructs used were the same as those in in Fig. 2E, incorporating the appropriate mutation(s). In the absence of any AAP, input VP3 levels are lower than when AAP2 is present across all mutants (Fig. 4A); this is consistent with AAP’s previously demonstrated role in VP stability. Additionally, VP1 input levels for the E413/684Q double mutant and the E413/684/686Q triple mutant are significantly less than those for E413Q (Fig. 4A), suggesting that E684Q affects VP stability. Despite variation in input VP1 levels across the panel of mutants, comparable quantities of HA-VP1 are precipitated in all eight anti-HA pulldowns (Fig. 4B). None of the VP1 mutants were able to coprecipitate VP3 proteins in the absence of AAP (Fig. 4B), consistent with previously demonstrated characteristics of AAP-dependent serotypes (17). The E413/684Q double mutant is the only VP1 that was able to robustly coprecipitate FLAG-AAP2, and it also coprecipitates appreciably more VP3 than other mutants (Fig. 4B), consistent with this mutant’s high level of compatibility with AAP2 (Fig. 3B). In the reciprocal experiment, FLAG-AAP2 coprecipitates E413Q and the E413/684Q double mutant the most robustly and E413/686Q nominally but the triple mutant not at all (Fig. 4C); low input levels of the triple mutant corroborate an inability for AAP2-VP interactions, assuming that AAP must bind VP (directly or indirectly) to perform its stabilizing function. These findings suggest that a precise charge on lumenal surface residues is critical for AAP-VP cooperation toward VP stability, oligomerization, and, ultimately, efficient capsid assembly.

FIG 4.

Effect of E to Q mutations on VP-VP and VP-AAP coprecipitation. HEK293 cells were transfected with mutant Anc81 HA-tagged VP1 constructs, with or without FLAG-AAP2, as indicated above lanes. Lysates were diluted such that total protein was equal across samples. Input (5%) (A) was removed, lysate divided in two, and IPs performed with rabbit anti-HA (B) or mouse anti-FLAG (C) antibodies to precipitate VP1 and AAP2, respectively. VP proteins were detected with the B1 antibody (top blots) and AAP2 with anti-FLAG antibody (bottom blots). All gels were electrophoresed, blotted to PVDF membranes, incubated with detection antibodies/reagents, and exposed in parallel.

Positive residues in AAPN impede AAP2-Anc81 compatibility.

We next aimed to identify what residues of AAP are important for AAP-VP cooperation. In addition to AAP2, we tested AAPs from AAV8 (pAAP8), Anc80 (pAAP80), and Anc82 (pAAP82) for their ability to trans complement production from Anc81AAPstop60 constructs. pAAP80 cooperates with Anc81 to produce greater than 10% of Anc81 WT AAP titers, but pAAP8 and pAAP82 are incompatible with Anc81, demonstrated by their <1% rescue (Fig. 5A). Comparing the AAP protein sequences of AAP80 and AAP81 (Anc81VP-compatible AAPs, blue bars) to those of AAP2, AAP8, and AAP82 (Anc81VP-incompatible AAPs, red bars), six residues emerge that share identity only within their respective phenotypic group (Fig. 5B, arrowheads). Two of these residues, C74 and C167, were previously examined by Naumer and colleagues, who demonstrated that mutating these sites to alanine does not affect AAP2 function (23), so we excluded these residues from further analysis here. For the remaining four sites of interest, we individually mutated pAAP2 residue identities to the following AAP80 and AAP81 identities: L12H, Q17P, H34L, and R50Q. In addition, we generated the double mutant H34L/R50Q because only these two residues fall within the hydrophobic region (Fig. 5B, orange bar) or conserved core (Fig. 5B, green bar) domains of AAP previously shown to be critical for capsid assembly in most serotypes (17, 20). We tested the ability of this panel of AAP2 mutants to rescue Anc81AAPstop60 production (Fig. 5C, gray bars). Individually, neither L12H nor Q17P have a significant effect on AAP2-Anc81 compatibility. However, H34L and R50Q rescue vector production by 3.9-fold and 4.1-fold over WT pAAP2, respectively, and the H34L/R50Q double mutant provides a 13-fold better rescue than WT pAAP2, demonstrating the importance of both residues in conjunction for VP-AAP compatibility. Additionally, we tested a FLAG-tagged version of this double mutant to rescue Anc81AAPstop60 (Fig. 5C, black bars), which exhibits a full order of magnitude increase in compatibility compared to FLAG-AAP2. To examine whether these mutations affect homologous compatibility, we rescued AAV2AAPstop60 production with the tagged and untagged versions of AAP2H34L/R50Q, and observed no significant difference in rescue over WT AAP2 (Fig. 5D). These results suggest that residues 34 and 50 of AAP have important functional roles in coordinating the assembly process and are highly influential for compatibility with heterologous capsids while also mutationally robust against decrease in function. Moreover, the positive-to-neutral nature of the mutations made at these sites mirrors the importance of charged versus neutral residues for assembly we demonstrate for the VP beta-barrel.

FIG 5.

Positive residues in AAPN impede AAP2-Anc81 compatibility. (A) Vector produced from quadruply transfected HEK 293 cells (adenovirus helper, ITR-flanked reporter genome, Anc81 WT AAP or AAPstop60 rep-cap constructs, and the appropriate pAAP or empty plasmid; x axis) was quantified by qPCR on DRP and is reported as a percentage of Anc81 WT AAP titers produced in parallel. Graph represents the average of at least three independent experiments ± SEM. A paired two-tailed t test was conducted comparing WTAAP2 titers to rescue with other AAPs (*, P < 0.05; **, P < 0.01; ***. P < 0.005). (B) Alignment of compatible (blue) and incompatible (red) AAP sequences. Sites of interest are marked with arrowheads. Colored bars above alignment denote the previously defined domains AAPN (purple), hydrophobic region (orange), and conserved core (green). (C and D) Anc81AAPstop60 vector trans complemented by AAP constructs listed along the x axis was quantified by qPCR on DRP and is reported as a percentage of (C) Anc81 WT AAP titers or (D) AAV2 WT AAP titers produced in parallel. Graph represents the average of three independent experiments ± SEM. †, not detected. A paired two-tailed t test was conducted comparing rescue with pAAP2 to rescue with pAAP2 mutants (gray bars) or comparing rescue with FLAG-AAP2 to rescue with FLAG-AAP2 mutants (*, P < 0.05; **, P < 0.01; ***, P < 0.005; N.S., not significant).

DISCUSSION

Although AAV is the simplest form of an icosahedral capsid, the intermonomeric interactions occurring within the assembly are considerably complex and contribute to a remarkably stable virion that can withstand appreciably high temperatures (24–26), a wide range of pH (27), and long shelf-lives (27), while remaining infectious. There is a reasonably high energy barrier associated with assembling this highly stable product, corroborated by the inability of VP monomers to self-assemble (12) and the requirement for AAP, which promotes VP oligomerization to overcome such energy barriers. AAP also defends VP proteins from host attempts to degrade them (17), and transports them to the required cellular compartment to be assembled into the full icosahedron (13, 14). It is unlikely that orchestration of such myriad events occurs without a physical interaction between the molecules involved, raising our interest in elucidating the nature of the VP-AAP interaction. In addition to basic interest in AAV biology, capsid assembly is a primary requirement for any therapeutic vector. Understanding the VP-AAP interaction is fundamental to rational design, directed evolution, and mass mutagenesis or library-based approaches to capsid engineering aim to improve clinically desirable properties of AAV vectors. We previously postulated that AAP evolved to allow VP a greater degree of mutational freedom and still accomplish capsid assembly. Capsid engineering efforts exploring VP permutations would benefit greatly from a “universal” AAP capable of promoting assembly of a broad range of VP compositions.

AAP is generally resilient against loss of function imbued by domain swapping, capsid shuffling, error-prone PCR, or excluding it as a variable when deriving the AncAAVs by ancestral sequence reconstruction, where the +1 reading frame was not considered (18, 21, 28–31). Nonetheless, most of the AncAAVs (1) require AAP and (2) express a functional cognate AAP (17). Important exceptions are Anc113, which produces slightly better in the AAPstop60 context than in the WT AAP context (17), and Anc83, which does not seem to express its own AAP (Fig. 1B). Despite the resilience of AAP, it is not universally functional. For example, both AAP4 and AAP5 can trans complement VLP production of both AAV4 and AAV5 but are incompatible with nearly all other serotypes (14, 16). Here, we leverage the incomplete universality of AAP2 (Fig. 1) and extreme assembly rescue phenotypes among closely related capsid variants (Fig. 2) to systematically demonstrate that the lumenal surface of the VP beta-barrel is a site of VP-AAP interaction (Fig. 4), which depends on finely tuned charge and neutrality for productive assembly of the AAV capsid (Fig. 3). Naumer and colleagues identified an AAV2 beta-barrel mutant, I682S, that abolished assembly (23), indicating that not only neutrality but hydrophobicity is important in this region for AAP-VP cooperation. In turn, we show that neutralizing charged residues in AAP2 increases heterologous VP compatibility (Fig. 5). Importantly, these mutations do not affect homologous AAP-VP compatibility, thus implicating neutrality at residues 34 and 50 as elemental for a universally functional AAP.

The wealth of capsid structural data across AAV serotypes demonstrates the high conservation of the VP beta-barrel structure that lines the capsid lumen, and this structure is indeed well conserved among all parvoviruses. Mutations in the beta-barrel often disrupt capsid assembly (32), but the nature of the loss of assembly is not always clear (i.e., whether VP is misfolded or whether intermonomeric interactions are disrupted) particularly because such studies examine purified viral preps that eliminates unassembled monomers or oligomers. By examining total VP protein levels in the crude lysates (Fig. 3G), if VP proteins were still abundant but we observed a large loss in titer of a particular mutant, this would have suggested that AAP is able to perform its protein stabilizing role (implying a functional AAP-VP interaction) but that the VP proteins themselves are defective for capsid assembly, capsid stability, or packaging, independent of any role of AAP. Since total VP protein loss was always observed in conjunction with decreased titers in the mutant panel (Fig. 3F and 3G), one explanation that fits our observations well is that the AAP-VP interaction is disrupted by the mutation and AAP cannot perform its protein-stabilizing function, an obvious prerequisite for capsid assembly. This does not, however, rule out the possibility that mutations like E684K and E686K in conjunction exhibit a total loss of titer and VP protein stability as a function of a folding defect, which would not only abolish capsid assembly but would likely be subject to a misfolded protein response whether or not AAP is present; this is also consistent with our observations. Overall, our observations add to knowledge of the importance of the beta-barrel for assembly by adding the AAP context.

Beta-sheets and beta-barrels are well-known docking sites for interaction partners, but considering the evidence against AAP remaining part of the assembled capsid (it is not detected in any solved capsid structures nor in purified AAV preparations), a lumenal interaction site implies that AAP must either scaffold the icosahedron around itself and then escape or that it dissociates from oligomerized VPs at earlier assembly steps. Either scenario would require the AAP-VP interaction to be transient yet sufficiently strong for AAP to mediate nuclear/nucleolar trafficking of VP, supporting the importance of finely tuned surface charge for VP-AAP cooperation. Anc81’s glutamic acid at site 413 is conserved in the Anc81-AAV7 branch of the reconstructed phylogeny, and it is also conserved within the outgroup consisting of AAV4, rh.32.33, and AAV5. Of these 6 variants, only Anc81 and AAV7 are AAP dependent, and it is difficult to draw conclusions about drivers of AAP-VP compatibility based on capsids that do not strictly require AAP. It is worth noting, however, that AAV7 is rescued by AAP2 to only 5.9% of its WT AAP titer (Fig. 1), and while the outgroup serotypes can assemble capsids in the absence of AAP, AAV4 and AAV5 VLP titers are increased when their cognate AAP, but not AAP2, is present (15). The similar phenotypes for AAP2 compatibility among variants with negative glutamic acid at residue 413 further supports the beta-barrel as critical for AAP interaction and cooperation.

Modeling how AAP interacts with VP’s beta-barrel is limited by the lack of empirical three-dimensional (3D) structural data for AAP. Within AAPN, the HR is predicted to take on a helical structure, whereas the CC is predicted to be disordered, as is the rest of AAP (20). The HR and CC domains of AAP are stretches of hydrophobic and neutral polar residues interspersed with an occasional charged residue, such as H34 and R50, which are conserved in AAP2, AAP8, and AAP82. The full-magnitude gain in Anc81 rescue by AAP2 observed when these positive residues are neutralized indicates a reciprocated importance of charge and neutrality on AAP for its cooperation with VP (Fig. 5). Tse and colleagues demonstrated that the HR and CC of AAP are critical for coprecipitation of VP (20), but further studies are required to determine whether residues 34 and 50 in this region contribute to AAP compatibility by increasing AAP-VP interactions or by enhancing the ability of AAP to promote VP stability and/or oligomerization.

One interpretation for these observations collectively is that AAPN binds the VP beta-barrel directly. However, an indirect interaction mediated by a host cellular factor(s) has not been ruled out. While several host factors have been implicated in AAV transduction, the host factors involved in assembly are unknown. Oligomerization of crudely purified VP proteins is enhanced when whole-cell lysate is added to in vitro reactions (12), highlighting the importance of host factors for assembly. Whether the AAP-VP interaction is direct or indirect, we present evidence that the VP beta-barrel and AAPN are involved. Further studies are required to determine if the VP residues we examine here directly influence productive assembly by stabilizing VP-VP interactions, or indirectly influence it through increased AAP-VP interactions that lead indirectly to increased VP-VP interactions (Fig. 4). Future studies examining the host factors involved will be crucial for understanding of AAV assembly mechanisms.

Our data support a model for capsid assembly where AAP directly or indirectly binds the lumenal surface of VP to perform its many roles, and a productive interaction requires a precise tuning of charge and neutrality on both proteins. We identify specific residues of both proteins that are involved in productive assembly and characterize chemical properties of these residues that increase or decrease vector titer when VP is trans complemented by a heterologous AAP, shedding light on what drives VP-AAP compatibility at the molecular level. A universally compatible AAP has direct implications for large-scale and library-based capsid engineering efforts, allowing examination of variants with potentially interesting phenotypes (e.g., increased transduction or decreased immune response) that would otherwise be absent due to concomitant negative effects on assembly or endogenous AAP interactions. Our findings not only describe interactions that need to be preserved during vector engineering efforts but how these sites could be modulated to increase assembly efficiency and ultimately increase vector yield for clinical applications.

MATERIALS AND METHODS

Vectors and sequences.

Adeno-associated viral vectors (AAV2 ITRs) were packaged into either extant or ancestral viral capsids. Extant capsids include AAV1 (GenBank accession number AAD27757.1), AAV2 (accession number AAC03780.1), AAV6 (accession number AF028704.1), AAV7 (accession number NC_006260.1), Rh.10 (accession number AAO88201.1), and AAV8 (accession number AAN03857.1). Ancestral AAV capsids (WT and WT AAP) include Anc80L65, Anc81, Anc82, Anc83, Anc84, Anc126, and Anc127 (accession numbers KT235804 to KT235812). Sequence alignments were generated using ClustalW with BioEdit software.

Crude virus preparations/titration.

Virus preparations to assay production in all serotypes and mutants were prepared as follows. Polyethylenimine (PEI) transfections of AAV cis ITR-CMV-EGFPT2A-Luc-ITR (2 mg), AAV trans rep-cap (2 mg), and adenovirus helper plasmid (4 mg) were performed on HEK293 cells at 90% confluence in 6-well dishes.

The Polyethylenimine Max (Polysciences)/DNA ratio was maintained at 1.375:1 (wt/wt) in serum-free medium. Virus was harvested after 72 h by three freeze/thaw cycles, followed by centrifugation at 15,000 × g.

For DRP titers, crude preparations were DNase I treated, and resistant (packaged) vector genome copies were used to titrate preparations by TaqMan qPCR amplification (Applied Biosystems 7500; Life Technologies) with primers and probes detecting cytomegalovirus (CMV) promoter regions of the transgene cassette.

Immunoprecipitations.

Transfections were performed with 8 μg each of CMV-HA-VP1 and either pCDNA3.1 (empty) or FLAG-AAP2 plasmid on 10 cm dishes of HEK 293 cells at ∼80% confluence. The PEI Max (Polysciences)/DNA ratio was maintained at 1.375:1 (wt/wt) in serum-free media. At 24 h posttransfection, medium was aspirated and 1 ml lysis buffer (0.1% Triton X-100, 150 mM NaCl, 50 mM Tris [pH 8], plus cOmplete Mini protease inhibitor) added directly to plate, which was then subjected to one freeze-thaw cycle. Lysate was collected, clarified by centrifugation, and total protein measured and diluted (as appropriate) to ensure equivalent starting material across samples. Immunoprecipitation was performed on 500 μl lysate with rabbit anti-HA antibody (Abcam 9110) or anti-FLAG M2 (Sigma F1804) and Pierce Protein A/G Plus agarose beads. Precipitated proteins were washed and eluted in 4× NuPAGE lithium dodecyl sulfate (LDS) sample buffer +0.5% β-mercaptoethanol (BME) at 90°C for 10 min. Aliquots of 10 μl (IP) or 30 μl (input) were loaded and electrophoresed on NuPAGE 4 to 12% bis-Tris gels and detected by Western blotting.

Western blotting.

Electrophoresed proteins were transferred to polyvinylidene difluoride (PVDF) membranes, incubated with primary antibody (B1, 1:250, ARP product no. 03-65158; anti-FLAG M2, 1:5,000, Sigma F1804) overnight, and detected with anti-mouse (GE Healthcare LNXA931/AE) or anti-rabbit (Sigma A0545) horseradish peroxidase (HRP)-conjugated secondary antibody and Thermo Super Signal West Pico.

Cloning and site-directed mutagenesis.

AAPstop60 constructs were generated previously (17). pAAP2 was generated by ligating a PCR product amplified from an AAV2 rep-cap plasmid with primers adding restriction enzyme recognition sites for directional cloning into pCDNA3.1(−). For FLAG-AAP2, complementary oligonucleotides encoding the FLAG tag were designed such that the bottom strand created a GATC overhang immediately downstream of FLAG, annealed in T4 ligase buffer ramping from 95°C to 25°C at 5°/min, T4 polynucleotide kinase treated, and ligated into EcoRV- and NheI-digested HA-FLAG-AAP2 expression plasmids (16), a generous gift from Dirk Grimm. Further mutations of the aforementioned plasmids were performed using a QuikChange II site-directed mutagenesis kit (Agilent) according to the manufacturer’s instructions and using the primer pairs in Table 1.

TABLE 1.

Primer pairs used for site-directed mutagenesis

Primer name	Sequence
Anc81E413Q_F	CGGGCAACAACTTTCAGTTCAGCTACACG
Anc81E413Q_R	CGTGTAGCTGAACTGAAAGTTGTTGCCCG
Anc81E684Q/E686Q_F	GGTCAGCGTGCAAATTCAATGGGAGCTGC
Anc81 E684K/E686K_F	GGTCAGCGTGAAAATTAAATGGGAGCTGC
Anc81E684K_F	GGACAGGTCAGCGTGAAAATTGAATGGG
Anc81E684Q_F	GGACAGGTCAGCGTGCAAATTGAATGGG
Anc81E686Q_F	CAGCGTGGAAATTCAATGGGAGCTGC
Anc81E686K_F	CAGCGTGGAAATTAAATGGGAGCTGC
Anc81 E684Q/E686Q_R	GCAGCTCCCATTGAATTTGCACGCTGACC
Anc81 E684K/E686K_R	GCAGCTCCCATTTAATTTTCACGCTGACC
Anc81E684K_R	CCCATTCAATTTTCACGCTGACCTGTCC
Anc81E684Q_R	CCCATTCAATTTGCACGCTGACCTGTCC
Anc81E686Q_R	GCAGCTCCCATTGAATTTCCACGCTG
Anc81E686K_R	GCAGCTCCCATTTAATTTCCACGCTG
AAP2H34L_F	CAGGCAGTGGCGCTCCAATGGCAGAC
AAP2H34L_R	GTCTGCCATTGGAGCGCCACTGCCTG
AAP2R50Q_F	GGGTAATTCCTCAGGAAATTGGCATTG
AAP2R50Q_R	CAATGCCAATTTCCTGAGGAATTACCC
AAP2L12H_F	CTGACCCCCAGCCACTCGGACAGCCAC
AAP2L12H_R	GTGGCTGTCCGAGTGGCTGGGGGTCAG
AAP2Q17P_F	CGGACAGCCACCCACAGCCCCCTCTGG
AAP2Q17P_R	CCAGAGGGGGCTGTGGGTGGCTGTCCG

Statistical analysis.

Viral titers were compared assuming a log-normal distribution of replicates. A one-tailed or two-tailed Student's t test was performed where appropriate.

Capsid structures and surface charge approximations.

Anc81 was modeled in PyMol using the mutagenesis wizard to change residues of AAV8 (PDB ID 2QA0) trimers obtained from VIPERdb (http://viperdb.scripps.edu) (33). The Adaptive Poisson-Boltzmann Solver (APBS) plugin (34) was used to generate solvent-corrected surface charge approximations with maximum/minimum ramp values of 3.5 to −3.5 k_bT/e_c (where (k_b is Boltzmann's constant, T is temperature in K, and e_c is the charge of an electron).