Structural Studies of Adeno-Associated Virus Serotype 8 Capsid Transitions Associated with Endosomal Trafficking

Abstract

The single-stranded DNA (ssDNA) parvoviruses enter host cells through receptor-mediated endocytosis, and infection depends on processing in the early to late endosome as well as in the lysosome prior to nuclear entry for replication. However, the mechanisms of capsid endosomal processing, including the effects of low pH, are poorly understood. To gain insight into the structural transitions required for this essential step in infection, the crystal structures of empty and green fluorescent protein (GFP) gene-packaged adeno-associated virus serotype 8 (AAV8) have been determined at pH values of 6.0, 5.5, and 4.0 and then at pH 7.5 after incubation at pH 4.0, mimicking the conditions encountered during endocytic trafficking. While the capsid viral protein (VP) topologies of all the structures were similar, significant amino acid side chain conformational rearrangements were observed on (i) the interior surface of the capsid under the icosahedral 3-fold axis near ordered nucleic acid density that was lost concomitant with the conformational change as pH was reduced and (ii) the exterior capsid surface close to the icosahedral 2-fold depression. The 3-fold change is consistent with DNA release from an ordering interaction on the inside surface of the capsid at low pH values and suggests transitions that likely trigger the capsid for genome uncoating. The surface change results in disruption of VP-VP interface interactions and a decrease in buried surface area between VP monomers. This disruption points to capsid destabilization which may (i) release VP1 amino acids for its phospholipase A2 function for endosomal escape and nuclear localization signals for nuclear targeting and (ii) trigger genome uncoating.

INTRODUCTION

Viral infection is initiated by entry into a host cell, and most viruses take advantage of existing host cell entry mechanisms, such as endocytosis, for internalization (22). However, while studies of enveloped viruses have yielded detailed information on their entry mechanisms (reviewed in references 8, 15, and 44), the understanding of this process for the nonenveloped viruses is still very limited. The single-stranded DNA (ssDNA)-packaging Parvoviridae, small nonenveloped viruses with a T=1 capsid, have been shown to enter host cells using surface receptors/coreceptors and the endocytotic pathway in a process requiring acidification (reviewed in reference 10). However, capsid transitions associated with endosomal processing during trafficking to the nucleus for genome replication, including mechanisms that lead to escape from endosomal compartments into the cytoplasm for nuclear localization and genome uncoating, are poorly understood.

The adeno-associated viruses (AAVs), belonging to the Dependovirus genus of the Parvoviridae, are being investigated for use as therapeutic gene delivery vectors (e.g., see reviews in references 7, 11, 12, and 17). Among the 12 distinct human and nonhuman primate serotypes of AAVs characterized to date (18, 34), AAV8 shows enhanced liver transduction efficiency compared to those of the other serotypes (18) and has been employed as a liver-directed gene therapy vector (25, 30, 42, 46). The improved transduction by AAV8 compared to that of AAV2 has been reported to be due to faster capsid uncoating by the former serotype and hence a faster onset of gene expression (2, 47).

The AAV capsid is assembled from 60 copies (in total) of the overlapping capsid viral proteins (VPs), VP1, VP2, and VP3, coded from the cap open reading frame, in a proposed ratio of 1:1:10, respectively. VP1 has a unique N-terminal region of 137 amino acids in AAV8 and shares another 66 residues with only VP2. The overlapping VP1-VP3 region (∼530 C-terminal residues) is multifunctional, facilitates receptor attachment, cellular transduction, capsid assembly, and genome packaging, and is the target of the host immune response against the capsid (24, 26, 33, 38, 43, 48–50). The unique N-terminal region of VP1 (VP1u) has a phospholipase A2 (PLA2) activity required for acidification-associated viral escape from endosomes during trafficking to the nucleus for DNA replication (19, 45). This VP1u, which also contains nuclear localization signals (NLS), is predicted to become externalized, from an internal disposition, while the capsid remains intact. The low-pH-associated capsid transitions that facilitate this event or that trigger the capsid “ready” for genome uncoating following endosomal processing are unknown and require further investigation toward a fuller understanding of the biology of the AAV vectors and of parvoviruses in general.

To gain insight into the effect of endosomal processing on parvoviruses, we have determined crystal structures of AAV8 empty (no DNA) virus-like particles (VLPs) as well as green fluorescent protein (GFP) gene-packaged (DNA full; recombinant AAV8 [rAAV8]-GFP) capsids at pH 6.0 and pH 5.5, which represent the pH values of early to late endosomal compartments, and at pH 4.0 of lysosomes in which AAV viruses have been previously observed during infection (3, 14, 23). In addition, the capsid structures of AAV8 VLPs and rAAV8-GFP crystals incubated at pH 4.0 for 24 h and then transferred to pH 7.5 (pH 4.0/7.5) were determined. This experiment was designed to mimic the pH changes experienced by the capsids during the predicted endosomal escape to the cytosol prior to nuclear entry. Capsid VP side chain rearrangements, observed with decreasing pH values, highlight capsid dynamics consistent with destabilization that likely facilitates the VP1u release for the PLA2 activity required for endosomal escape and NLS function and capsid readiness for genome release.

MATERIALS AND METHODS

Capsid production and purification.

AAV8 VLPs without packaged DNA (empty) were produced and purified by following previously described procedures (29). Capsids with packaged DNA, rAAV8 capsids packaging a GFP transgene (rAAV8-GFP), were produced via calcium phosphate-based cotransfection of plasmid DNA containing AAV8 cap, AAV2 rep, and adenoviral helper (E2A, E4, and VA) genes (pDG8) and the GFP gene flanked by the AAV2 inverted terminal repeats (pTR-UF11) into HEK293 cells. Following transfection, the cells were incubated for ∼72 h, harvested, and lysed by three cycles of freeze-thawing, with the addition of magnesium and Benzonase (Merck KGaA, Germany) in the last cycle. The clarified cell lysate was loaded onto an iodixanol step gradient (15, 25, 40, and 60%) (Nycomed) and centrifuged at 350,000 × g at 18°C for 1 h. The 40 to 60% gradient fraction was collected, diluted 1:1 with 20 mM Tris-HCl (pH 7.5) and 150 mM NaCl (the wash buffer), and loaded onto a HiTrap 5-ml Q column. After being washed, the virus-containing fraction was eluted by increasing the NaCl concentration to 1 M. The eluted sample was buffer exchanged into 20 mM Tris-HCl (pH 7.5) and 150 mM NaCl and concentrated to ∼5 mg/ml for crystallization using the hanging drop vapor diffusion method as previously described for AAV8 VLPs (29).

AAV8 crystallization, data collection, and reduction.

Crystallization, data collection, and structure determination were as previously described (29, 36). To obtain structures of the capsid at various pH values, crystals of AAV8 VLPs and rAAV8-GFP grown in 20 mM Tris-HCl (pH 7.5), 4% polyethylene glycol 8000 (PEG 8000), and 1 M NaCl were soaked in cryosolutions at the desired pH values. All cryosolutions contained 4% PEG 8000, 1 M NaCl, and 24% glycerol in 50 mM buffer at the appropriate pH values. For pH 7.5, pH 6.0, pH 5.5, and pH 4.0, the buffers were Tris-HCl, bis-Tris-HCl, sodium citrate, and sodium acetate, respectively. For endosomal escape pH experiments, pH 4.0/7.5, the preformed crystals were soaked in cryosolution at pH 4.0 for ∼24 h and then transferred to the pH 7.5 cryosolution and soaked for another 24 h prior to cryocooling for X-ray diffraction data collection. The X-ray diffraction data sets were collected at the Advance Photon Source (APS; 22-ID) at the Argonne National Laboratory, National Synchrotron Light Source (NSLS; X29) at the Brookhaven National Laboratory, and Cornell High Energy Synchrotron Source (CHESS; F1) at Cornell University. The data sets were collected at crystal-to-detector distances of 300 to 400 Å and at an oscillation angle of 0.3°. The data were indexed and processed with DENZO and scaled and reduced with SCALEPACK (39). The crystals of AAV8 VLPs belonged to the P6₃22 space group with unit cell dimensions of a = b = 254.8 Å and c = 445.4 Å (29, 36). The rAAV8-GFP capsid crystals had slightly larger unit cell parameters, with a = b = 257.9 Å and c = 448.7 Å, which remained unchanged at pH 6.0, pH 5.5, and pH 4.0 but decreased slightly for the pH 4.0/7.5 crystals to a = b = 256.4 Å and c = 447.4 Å. This minor change (less than 1%) is likely due to the prolonged soaking in the cryoprotectant solutions. The VLP and rAAV8-GFP crystals diffracted X-rays to at least a resolution of 2.7 Å, with the exception of the pH 4.0/7.5 structures, which diffracted to a resolution of 3.2 Å (Table 1).

Table 1.

Data processing and refinement statistics for the rAAV8-GFP structures

Parameter	Result
	P6322; a = 257.9 Å, c = 448.7 Å				P6322; a = 256.4 Å, c = 447.4 Å
	pH 7.5	pH 6.0	pH 5.5	pH 4.0	pH 7.5 from pH 4
X-ray source	APS(22-ID), CHESS(F1), BNL(X29)	APS(22-ID), CHESS(F1), BNL(X29)	APS(22-ID), CHESS(F1), BNL(X29)	APS(22-ID), CHESS(F1), BNL(X29)	APS(22-ID), CHESS(F1), BNL(X29)
Resolution range (Å)	40–2.7	40–2.7	40–2.7	40–2.7	40–3.2
No. of observations	938,161	304,563	559,536	567,265	295,903
No. of unique reflections	221,840	175,411	185,894	297,183	126,615
% completeness	92.9	73.5	77.8	86.8	88.2
Rsyma	0.083	0.118	0.115	0.114	0.139
Rcryst (workb/freec)	0.21/0.21	0.24/0.25	0.25/0.25	0.24/0.24	0.29/0.30
Molecular averaging NCS correlation coefficient	0.95	0.92	0.91	0.91	0.85
No. of protein atoms	4,136	4,136	4,136	4,136	4,136
No. of DNA atoms	37	43	23	0	19
No. of solvent atoms	76	54	85	87	0
RMSD, bond length (Å)	0.008	0.009	0.034	0.035	0.009
RMSD, bond angles (°)	1.54	1.55	1.50	1.54	1.60
Avg B factor, main chain (Å2)	44.3	32.1	37.7	36.2	44.3
Avg B factor, side chain (Å2)	44.6	32.2	38.2	36.9	44.4
Residues in the most favored/additional/generously allowed regions (%)	0.87/0.13/0.0	0.85/0.15/0.0	0.93/0.07/0.0	0.92/0.08/0.0	0.81/0.19/0.0

^aCalculated as ∑_h∑_i|I_ih − <I_h>|/∑_h∑_i <I_h>, where <I_h> is the mean of the observations I_ih of reflection h.

^bCalculated as ∑||F_o| − |F_c||/∑|F_o|, where F_o and F_c are the observed and calculated structure factors, respectively.

^cSame as R_cryst, but calculated with a 5% randomly selected fraction of the reflection data not included in the refinement.

Structure determination and refinement.

The AAV8 VLP and rAAV8-GFP structures were determined as previously described for AAV8 VLPs solved at pH 7.5 and by using this structure as a phasing model (36). After a cycle of rigid-body refinement, several iterative cycles of model refinement were performed using the simulated annealing, energy minimization, atomic position, and anisotropic temperature factor (B-factor) refinement options in the Crystallography & NMR (nuclear magnetic resonance) System (CNS) program (6). The refinement cycles were alternated with averaged electron density calculations in the CNS program (6) and manual model rebuilding by using the COOT program (16). Strict 10-fold noncrystallographic symmetry (NCS) operator constraints were applied throughout the refinement process and for electron density averaging as dictated by the capsid orientation and packing in the hexagonal space group (29, 36). The data collection, processing, and refinement statistics are given in Table 1. The refinement statistics were comparable to or better than those reported for virus structures determined from data sets of comparable completeness and resolutions for other parvoviruses or members of other virus families as detailed on the VIPERdb website, http://viperdb.scripps.edu/ (41). The association energy and buried surface area for icosahedral symmetry-related monomers, 2-fold, 3-fold, and 5-fold, as calculated on the VIPERdb website (41), are given in Table 2. Figures were generated using the PyMOL program (13).

Table 2.

Association energy and buried surface area for AAV8 VP interfaces under physiological and acidic conditions

pH	Association energy (kcal/mol)			Buried surface area (Å2)
	I–2	I–3	I–5	I–2	I–3	I–5
7.5	−67.4	−213.1	−102.1	3,235	10,373	5,058
4	−63.8	−213.7	−101.6	3,082	10,411	5,033

Protein Data Bank accession numbers.

The refined coordinates of rAAV8-GFP VP3 at pH 7.5, pH 6, pH 5.5, pH 4.0, and pH 4.0/7.5 have been deposited in the Protein Data Bank (PDB) with the accession numbers 3RA4, 3RA9, 3RA8, 3RA2, and 3RAA, respectively. The AAV8 VLP VP3 structures were identical to those of the VP3 structures of the rAAV8-GFP capsids at comparable pH values, and thus these coordinates have not been deposited.

RESULTS AND DISCUSSION

The structures of AAV8 VLPs and rAAV8-GFP capsids have been determined at pH values which mimic the conditions encountered by AAVs during trafficking through the endocytic pathway and the low acidic pH of the lysosome. The main chain structures are identical at each given pH value, with differences occurring only inside the capsid region, where densities consistent with those of ordered DNA nucleotides were observed. The structural similarity between VLPs and capsids produced in mammalian cells validates efforts using heterologous expression systems, such as the baculovirus/Sf9 system, for large-scale AAV vector production (27). As reported for the empty VLP structure of AAV8 determined at pH 7.5 (36), the overlapping VP polypeptide region from residue 220 to 738 (VP1 numbering; here referred to as VP3) was ordered in all the crystal structures (Fig. 1 A). The first 219 N-terminal amino acids, which include the VP1 unique region (VP1u), the VP1-VP2 overlap, and the first 15 residues of VP3, are not ordered in the electron density map. This disorder is believed to be due to the low copy number of the VP1u (∼5 chains), VP1-VP2 overlap (∼5 chains), and/or a differential disposition of the N-terminal regions of VP1-VP3, which is inconsistent with the icosahedral symmetry imposed during structure determination. Equivalent N-terminal VP regions are disordered in all parvovirus structures determined by X-ray crystallography to date (reviewed in reference 9). Given the equivalence of the VLP and AAV8-GFP capsid structures, the GFP genome-packaged structures will be described and discussed in detail below.

Fig. 1.

Comparisons of AAV8 pH structures. (A) Superimposition of the backbone of the rAAV8-GFP structures determined at the four different pH values: pH 7.5 (green), 6.0 (cyan), 5.5 (yellow), 4.0 (magenta). The VR regions, II, IV, and VI (defined in reference 20) and residues 565 to 567, showing minor main chain differences between the structures at the different pHs, are indicated. The conserved helix (αA), eight-stranded β-barrel (BIDG and CHEF), β-strand A (letter A), DE loop between β-strands D and E, HI loop between β-strands H and I, first ordered N-terminal residue (220), and C-terminal residue (738) are labeled. (B) A superimposition of the backbone of the rAAV8-GFP structures as is described for panel A, showing the location and conformation of the residues in the pH quartet (R392, H529, E566, and Y707), F631 and H632 in the rAAV8-GFP pH 7.5 (green) and pH 4.0 (magenta) structures in stick representation. The dinucleotide ordered in the pH 7.5 structure is also shown in orange and blue and is indicated by the open-headed arrow. Outside (C) and interior (D; a 180° rotation relative to panel C) surface view (down an icosahedral 2-fold axis) of the AAV8 VP3s related to the reference monomer (Ref; backbone superimposition of all four pH structures as described for panels A and B) by icosahedral 5-fold (5f1 to 5f4 in purple, lime green, firebrick red, and cyan, respectively), 3-fold (3f1 to 3f3 in salmon, yellow, and orange, respectively), and 2-fold (2f in deep blue) symmetry relationships. The symmetry-related VPs are for the pH 7.5 structure. In panel C, the pH quartet residues (located close to the icosahedral 2-fold axis) contributed from the pH 7.5 Ref, 2f1, 3f2, and 5f1 VP3s are shown in CPK, with Cα atoms colored according to the contributing VP3. In panel D, F631 and H632 from the pH 7.5 VP3s are shown, at the icosahedral 3-fold axes, in CPK, with Cα atoms colored according to the contributing VP3. In panels C and D, the black triangle depicts a viral asymmetric unit bounded by 5-, 3-, and 2-fold icosahedral symmetry axes. The approximate icosahedral 2-, 3-, and 5-fold axes are indicated by the filled oval, triangle, and pentagon, respectively, in panels A to D.

pH-mediated transitions are localized to two specific AAV8 capsid regions.

Comparison of the AAV8 VP3 crystal structure previously reported at pH 7.5 (36), and that of the AAV8-GFP capsid VP3 at pH 7.5, to those determined at pH 6.0, 5.5, 4.0 and 4.0/7.5 showed no significant main chain differences but identified two capsid regions with amino acid side chain conformational switches (Fig. 1 and 2). Superimposition of the Cα atoms showed a root mean square deviation (RMSD) of 0.12, 0.19, and 0.21 Å, respectively, between the AAV8-GFP structures determined at pH 7.5 and pH 6.0, pH 7.5 and pH 5.5, and pH 7.5 and pH 4.0. Slightly larger main chain differences were observed at previously defined AAV VP variable regions (VR) (20, 36), including VRII (AAV8 residues 329 to 332; ∼0.5 to 0.96 Å with all the low pH structures), and VRIV (residues 455 to 458; ∼0.5 to 0.75 Å with the pH 5.5 and pH 4.0 structures), VRVI (residues 534 and 535; ∼0.4 Å with the pH 4.0 structure), as well as residues 565 to 567 (∼0.4 to 0.7 Å with the pH 4.0 structure), (Fig. 1A). The largest difference in VRII, located at the top of the loop between the βD and βE strands (the DE loop) (Fig. 1A), was between the pH 7.5 and pH 4.0 structures. Five icosahedral symmetry-related DE loops line the channel at the 5-fold axes of the AAV capsid. Thus, although these residues have high temperature factor values (∼75 to 86 Ų), it is possible that the minor capsid changes observed are related to priming of the icosahedral 5-fold channel for the VP1u externalization for its PLA2 function postulated to occur through this axis (5, 21, 28, 45). Variable loop region IV, which is also flexible in nature, with high temperature factor values (∼70 Ų), forms part of three protrusions that surround the icosahedral 3-fold axis close to residues reported to be involved in receptor binding (1). This loop region was also observed to exhibit a minor conformation difference between AAV1 and AAV6 despite a lack of sequence difference (37), suggesting that the minor structural variation observed in the AAV8 structures might also be due to flexibility rather than a pH-associated change. Variable loop region VI and residues 565 to 567 are located on the wall of the depression at the icosahedral 2-fold axis/base of the 3-fold protrusions of the viral capsid (Fig. 1A and B and Fig. 2), and the difference in structure is due to local conformational side chain switches resulting in alteration of interactions associated with acidification, as described below.

Fig. 2.

pH-mediated AAV8 capsid surface transitions. (A) Conformational changes in AAV8 VP amino acid side chains of the pH quartet residues (392, 529, 566, and 707) at pH 7.5, 6.0, 5.5, and 4.0 close to the icosahedral 2-fold axes are shown within a 2F_o-F_c density map (light-gray mesh, contoured at a threshold of ∼1.0 σ, where F_o are the observed structure factors and F_c are structure factors calculated from the model). The reference (Ref), 2-fold (2f), and 3-fold (3f) symmetry-related VP monomers (in coil representation) are shown in green (Ref), red (2f), and blue (3f). The amino acids are shown as a stick model and colored according to atom type, with C in cyan. (B) Conformational differences in the pH quartet residues for the pH 7.5 and pH 4.0 structures. The symmetry-related VPs are colored as described for panel A; the amino acids are shown as described for panel A for the pH 7.5 structure and colored according to atom type, with C in pink in the pH 4.0 structure. Predicted hydrogen bond interactions are indicated with the dashed lines. The approximate positions of AAV variable regions III, VI, and IX (defined in reference 20) are indicated.

Significant amino acid side chain conformational rearrangements were observed on two regions of the AAV8 capsid associated with the pH changes. The first is on the AAV8 capsid surface at amino acids R392, Y707, E566, and H529, which we call the “pH quartet” (Fig. 1C and 2). At pH 7.5, the side chain of E566 is in position to form double hydrogen bonds with R392 from a 3-fold-related monomer and a hydrogen bond with the hydroxyl group of Y707 from a 2-fold-related molecule (Fig. 2 A and B). However, as the pH is lowered to 6.0, 5.5, and 4.0, the side chain of E566 adopts an alternative conformation that facilitates the formation of a hydrogen bond with the side chain of H529 (Fig. 2A and B). This new interaction is likely due to the protonation of H529, with a pKa of 6.2, which is now able to provide a proton for this interaction. The conformational change of E566 at a lower pH results in the loss of interactions with the 3-fold-related R392 and the 2-fold-related Y707 (Fig. 2B) and reduces the intersubunit contacts between the VP monomers (Fig. 2B and Table 2). Concomitant with the change at E566, the side chain of Y707 becomes oriented toward the 2-fold axis (Fig. 1B and 2). In the pH 4.0/7.5 structure, the transitions are reversed to the conformations observed at pH 7.5 (not shown).

The most significant outcome of the pH quartet transition is the weakening of the icosahedral 2-fold interface, with an ∼5% reduction in association energy and buried surface area (Table 2). Interestingly, the 2-fold interface is already the thinnest region of the parvovirus capsid, being only one polypeptide chain thick, with VP monomers overlapping above and below the 2-fold axis (Fig. 1C) (20, 36). However, the 2-fold axis has not been previously implicated in the events associated with endosomal processing of the parvovirus capsid, such as the externalization of VP1u and VP2 N termini or genome uncoating, since the 5-fold pore is predicted to be a portal for these exposures (5, 21, 28, 45). Nonetheless, in the crystal structure of AAV8 determined at pH 7.5 (36), weakly ordered density inside the capsid, under the 2-fold axes, running between 5-fold-related VP monomers, and within the channel that runs from the inside to the outside of the capsid at the icosahedral 5-fold axis was postulated to be due to VP1/VP2 N-terminal sequences. And in the cryoelectron microscopy (cryo-EM) and image reconstruction structure reported for AAV2 and AAV4, globules of density inside the capsid, under the icosahedral 2-fold axes, and running between 5-fold-related VP monomers have also been postulated to be due to VP1/VP2 N-terminal regions (28, 40). The reported structural observations, combined with the transitions observed in reduced-pH AAV8 structures directly adjacent to the 2-fold axis (Fig. 1C), suggest that this capsid contact region may serve as the disassembly point or capsid destabilization site for VP1/VP2 N-terminal externalization.

The AAV8 residues involved in the pH quartet transitions, R392, H529, E566, and Y707, are highly conserved in the AAV1-AAV12 (40), although the interactions between the residues differ in some of the available crystal structures determined at physiological pH (Fig. 3) (L. Govindasamy and M. Agbandje-McKenna, personal communication) (20, 36, 51). The sequence conservation suggests a role in AAV infection, and the differences in side chain orientations and interactions may reflect differences which dictate (i) host cell tropisms or (ii) cellular interaction requirements during trafficking. Thus, it would be interesting to analyze the effect of pH on these residues on other serotypes and to characterize their effect on infection by mutagenesis.

Fig. 3.

Structurally equivalent residues to the AAV8 pH quartet amino acids in other AAV crystal structures. The amino acid orientations and interactions are shown for AAV1, AAV2, AAV8, and AAV4. For each virus, the main chain from the reference (Ref) monomer is shown in green, 2-fold (2f) in cyan, and 3-fold (3f) in blue. The side chains are colored according to atom type, with the C atom the same as the main chain color. These structures were determined under physiological pH (pH 7.3 to pH 7.5). The R390-E564 interaction, equivalent to AAV8 R392-E566, is conserved in AAV1 and AAV4 (R383-E562); the equivalent residues are too far apart in AAV2. Residues equivalent to Y707 adopt a different conformation in AAV1 (Y705) and AAV4 (Y703), but this residue is in the same conformation in AAV2 (Y704).

In reviewing the available AAV mutagenesis literature, while there is no data available for AAV8, residues equivalent or adjacent to the AAV8 pH quartet amino acids have been mutated in AAV2. An R389A AAV2 mutant (R392 in AAV8) showed a wild-type phenotype (49). This observation could be due to a difference in capsid pH transitions between AAV2 and AAV8 since in the published crystal structure of AAV2 determined at physiological pH the interaction between this residue and E563 (equivalent to E566 in AAV8) was not observed (Fig. 3) (51). However, mutations of AAV2 residues 561 to 564 (equivalent to 564 to 567 in AAV8), which form a conserved acidic patch on the capsid surface of AAVs (discussed in references 20 and 32), to alanines had no effect on capsid assembly and packaging but resulted in a noninfectious phenotype (49). This observation suggests an essential role (yet to be defined) for these residues, which includes the E566 that changes conformation with decreasing pH values, in the AAV life cycle. While there is no specific mutation for AAV2 residue H526 (equivalent to AAV8 H529), a charge to alanine mutations at adjacent residues 527 to 532, which also had no effect on capsid assembly and packaging, led to a defect in receptor attachment and a reduction in infectivity and transduction (33, 49), again pointing to an essential role for these residues in the AAV2 life cycle. However, the functions of this stretch of amino acids, in addition to receptor attachment, remain to be characterized. Interestingly, AAV2 residue Y704 (equivalent to Y707 in AAV8) (Fig. 3), is one of seven capsid surface-exposed tyrosine residues reported to be phosphorylated for subsequent ubiquitination and proteasomal degradation during AAV2 trafficking, resulting in reduced transduction of certain target cells which can be overcome by mutation to phenylalanine (52, 53). This observation points to a functional importance for this residue, and the conformational transition observed could be priming for cellular interactions during trafficking in addition to the weakening of the 2-fold interface and destabilization of the capsid.

The second region of the AAV8 VP3 exhibiting a significant structural difference with pH was located in the interior capsid surface at the 3-fold axis resulting from a conformational change in the side chain of residue H632 adjacent to F631 (Fig. 1B, D, and 4). Residues F631 and H632 are highly conserved among AAVs, and the loop containing these residues assembles the innermost surface region of the 3-fold axis (Fig. 1D and 4). The side chain of H632 becomes oriented toward F631 as the pH decreases. At pH 7.5 and pH 6.0, the H632 side chain is oriented away from F631 and toward residue P633, as was observed for the equivalent residues in other AAV crystal structures (20, 31, 36, 37, 51) (Fig. 4, top). At pH 5.5, however, a dual conformation was observed for the H632 side chain, with the alternate conformation for the H632 side chain pointing toward F631 (not shown). At pH 4.0, the alternate conformation is the dominant form (Fig. 4, bottom). At pH 7.5, the H632 side chain forms a hydrogen bond with the side chain of N611 from a 3-fold symmetry-related VP3 (Fig. 4, top), and at acidic pH, H632 points toward F631, breaking the bond (Fig. 4, bottom). The altered conformation of H632 could result in a cation-aromatic interaction between the positively charged H632 side chain and aromatic F631 (red in Fig. 4, bottom; distance of 4.1 Å) and hydrogen bonding interactions with the main chain atoms of N630 from a 3-fold-related VP3 (Fig. 4, bottom). The cation-aromatic interaction is commonly observed for protein structures, frequently with arginine/lysine and aromatic residues. However, histidine has also been shown to interact with aromatic side chains and more so when it is positively charged (4). The perpendicular angle observed between the planes of H632 and F631 is also frequently observed in acidic environments (Fig. 4, bottom). The altered conformation of the side chain of the pH-sensitive H632 residue likely plays an important role in endosomal transition of the viral capsid given that acidification has been shown to be essential for trafficking and escape from the endocytic pathway (19, 45). This could be in the readiness of the capsid for packaged DNA uncoating or also part of the proposed capsid destabilization transition (in addition to the pH quartet changes) required to enable deployment of the VP1u for its PLA2 function for endosomal escape via the abrogation of VP-DNA interactions (see below).

Fig. 4.

pH-mediated transitions in the AAV8 capsid interior at the 3-fold region. (Top) Stereo image of the rAAV8-GFP pH 7.5 structure for residues 630 to 633. These residues are shown within a 2F_o-F_c electron density map (light-gray mesh, contoured at a 1.0 σ threshold) for the reference (Ref) VP3. At pH 7.5, the side chain of H632 is in an “out” conformation, pointing away from F631 (located at the icosahedral 3-fold axis). (Bottom) A stereo image of the rAAV8-GFP pH 4.0 structure for the same residues as those shown for the pH 7.5 structure but with the H632 residue in an “in” conformation, pointing toward F631. The 2F_oF_c electron density map is contoured at the same sigma threshold as that for the pH 7.5 structure. Predicted hydrogen bond interactions and other atomic distances are indicated with the dashed lines. A potential cation-aromatic interaction between the positively charged H632 side chain and aromatic F631 is indicated with the solid red line. In both images, the residues from the reference VP3 (Ref) are shown, with C in green (and labeled), and the two icosahedral 3-fold-related monomers are shown with C in pink (3f1) and cyan (3f2). Labels for some of the 3-fold-related residues have been omitted for clarity. The approximate icosahedral 3-fold axis is indicated in the filled triangle.

A pH-induced change in the AAV8 capsid interior alters DNA ordering.

As was previously observed in the AAV8 VLP structure determined at pH 7.5 (36), density consistent with a single dAMP nucleotide was ordered in the interior of the VLP structure determined at pH 6.0 (not shown) in a conserved DNA binding pocket (20, 31, 36, 37). For the rAAV8-GFP structures, densities consistent with two connected nucleotides were ordered inside the capsid at pH 7.5 (Fig. 5 A, left) and pH 6.0 (not shown), which could be modeled as a dAMP and dCMP dinucleotide. The nucleotide assignments were based on lack of density for an amino group at the C-2 position of the purine base and for a methyl group at the C-5 position of the pyrimidine base as previously discussed (36, 37). The DNA densities in both the VLP and rAAV8-GFP structures became less ordered as the crystal condition was decreased to pH 5.5 and were lost at pH 4.0 (Fig. 5A, middle). The dAMP density was restored in the pH 4.0/7.5 structure (Fig. 5A, right), suggesting that the lack of ordering at pH 4.0 was likely due to disrupted interactions with VP3 amino acids with the decreasing pH.

Fig. 5.

pH-induced DNA disordering in the interior capsid surface of AAV8 at the 3-fold region. (A) Electron density ordered close to H632 and modeled as a dinucleotide. The dinucleotide model is shown within an F_o-F_c map (light-gray mesh, contoured at a 2.5 σ threshold) for the rAAV8-GFP pH 7.5, pH 4.0, and pH 4.0/7.5 structures. The amino acids forming the DNA binding pocket, which is located in the capsid interior close to the icosahedral 3-fold axis, are shown as a stick model and colored according to atom type, with C in green and labeled for one VP3 monomer. The dinucleotide model is shown as a stick model and colored according to atom type, with C in pink and labeled. (B) Same as described for Fig. 1D but showing the location of the dinucleotide (in blue and red CPK) for each VP3 monomer inside the capsid. The black triangle depicts a viral asymmetric unit bound by 5-, 3-, and 2-fold icosahedral symmetry axes. The approximate icosahedral 2-, 3-, and 5-fold axes are indicated by the filled oval, triangle, and pentagon, respectively. (C) A close-up view of the interior surface of the rAAV8-GFP structure at pH 7.5, viewed down the icosahedral 3-fold axis with interiorly located His residues H626, H632, and H644 shown as cyan, with their O and N colored red and blue, respectively. The dinucleotide (colored according to atom type) is shown in its binding pocket, where the base of the dAMP nucleotide is positioned mostly inside the capsid, with only its sugar and phosphate groups visible in this view.

A dAMP nucleotide has also been observed in the crystal structure of wild-type AAV4 (20) and AAV3B (31). For AAV6, two nucleotides were ordered in VLPs, with a dAMP at the position equivalent to that observed for the other AAVs and with a dCMP that was not connected but was rather located close to the 3-fold axis (37). These nucleotides are ordered in a binding pocket similar to that of the AAV8 structure, with residues equivalent to AAV8's P633 and P422 interacting with the adenine of the structurally ordered dAMP nucleotide via base stacking, and the pocket also contains a residue equivalent to H632 (Fig. 5A). The disappearance of the densities for the dAMP and dinucleotide in the structures of AAV8 VLPs and rAAV8-GFP, respectively, with decreasing pH values (Fig. 5A, pH 7.5 and pH 4.0, left and middle for the AAV8-GFP structures) was concomitant with the H632 side chain shift described above. The closest atom of the adenine base of the dAMP in the ordered dinucleotide to the capsid is located ∼3.9 Å away from H632 at pH 7.5, and the closest sugar atom is ∼4.0 Å away. At pH 4.0, there was no ordered electron density for the nucleotides. The H632 side chain conformational shift at pH 4.0 eliminates its proximity to the adenine base and positions it close to the sugar in the pH 7.5 structure (Fig. 5A). The dCMP is stabilized by interactions with the side chain carboxyl group and main chain carbonyl oxygen atoms of D420 (not shown) in addition to base stacking with P422 (Fig. 5A, left). There are no significant changes in the conformation of these residues with decreasing pH values, although the ionization state of D420 is expected to change with the decrease in pH to 4.0 and may result in loss of interaction.

The observation of an ordered dAMP within a conserved binding site in the majority of the crystal structures of AAVs determined to date (20, 31, 36, 37) suggests a role for the interacting residues, particularly P422 and P333, which stack with the base of this nucleotide and H632 (in AAV8 and the equivalent residues in the other serotypes), in DNA stabilization inside the capsid close to the 3-fold axes (Fig. 5B and C). The change in the side chain conformation of H632 with decreasing pH values, which is predicted to result from a change in protonation state, appears to serve as one of the steps leading to release of the genome from capsid interaction. Thus, the H632 side chain shift at low pH values is presumably an essential step in DNA uncoating and/or VP1 externalization, since capsid stability is likely to be decreased following release from DNA interactions, or both. However, acidification has been reported to be insufficient for genome uncoating for the AAVs (45). Consistently, in the pH 4.0/7.5 endosomal escape mimic structure, the density for the dAMP portion of dinucleotide observed at pH 7.5, which disappeared at pH 4.0, is partially restored when the crystals are reincubated at pH 7.5 (Fig. 5A, right). Thus, it appears that while at pH 4.0 specific interactions between the capsid and packaged DNA are weakened, this is not sufficient for genomic uncoating. Rather, the packaged genome may be becoming compacted. Consistently, small-angle X-ray and neutron solution scattering studies conducted on AAV8 at the same pH values as those used in these structural studies show a reduction in the radius of gyration of the packaged DNA as the pH is decreased (unpublished data). This compaction could be a step required for genome release or release of the VP1u for externalization, although these possibilities remain to be tested. The current postulation is that the 5-fold pore is the externalization portal for the VP1 and VP2 N termini and packaged DNA in the AAVs and other parvoviruses. However, due to the high flexibility of the DE loop which assembles this portal (Fig. 1A), the small changes observed at the 5-fold pore at the lower pH values in the rAAV8-GFP structures are difficult to interpret, and the slight difference in diameter at the top of the channel between the pH 7.5 and pH 4.0 structures (∼1.0 Å) is a decrease for the pH 4.0 structure. Thus, in addition to low pH, other cellular factors, including proteolytic enzymes (2), likely function to facilitate additional essential capsid transition/dynamic events to externalize these VP1u and VP1/VP2 N-terminal regions.

In conclusion, the crystal structures of AAV8 determined under physiological conditions and the low endosomal/lysosomal pH values presented here provide clues to how the acidic environment triggers changes to its capsid during transition into the cytoplasm for subsequent nuclear entry. Histidine residues, strategically placed in the capsid, are observed to act as switches that sense pH changes which initiate capsid structural transitions. Of the 12 His residues within the common AAV8 VP3 sequence (residues 220 to 738) ordered in the crystal structure, four are located in the interior surface of the capsid, with H230 close to the 5-fold axis and H626, H632, and H644 (along with H424) concentrated at the 3-fold axis adjacent to the ordered density modeled as DNA (H424 and H632 are shown in Fig. 5A; 626, 632, and 644 are shown in Fig. 5C). Three other His residues, H361, H429, and H512, are involved in 3-fold symmetry-related interactions (not shown), and H529 facilitates the disruption of the interaction between E566 and 3-fold related R392 as discussed above (Fig. 2). H256, H291, and H293 are involved in intramonomer interactions (not shown). Below its pKa of 6.2, His residues are expected to be positively charged, which would render the inner surface of the capsid (Fig. 5C) more basic. This transition could weaken intra- and intermonomer interactions, which may be important for the VP1u PLA2 externalization as well as DNA release from capsid amino acid interactions. However, the exact role of a positively charged capsid interior in promoting endosomal processing events remains to be elucidated through mutagenesis approaches. Significantly, the structural transitions induced by low pH on the capsid surface and in the interior are reversible. Whether or not the parvovirus capsids experience pH 7.5 following endosomal/lysosomal escape is not known. However, the restoration of capsid amino acid interactions as observed would help to stabilize the capsid following the deployment of the VP1u, and restored capsid-DNA interactions would protect the genome until uncoating occurs in the nucleus.