Table 1.
Detection of sites subject to positive selection through AAV capsid genes.
Table 2.
Detection of episodic diversifying selection among capsid genes of AAV lineages.
Fig 1.
Distribution of positively selected sites of AAV lineages on their capsid structures.
The positively selected residues for the AAV2, AAV1/6, AAV5, AAV8 and AAV9 lineages are highlighted on their respective capsid surface images (left side) and a stereographic roadmap projection (right side) in (A), (B), (C), (D) and (E), respectively. (A) AAV2, residue E548 (blue) was detected by the MEME method. (B) AAV1/6 lineage, residues M599 (red) and V699 (pink) (detected by the branch-site test) and T326 (salmon), K493 (light-magenta), and D495 (hot-pink) (detected by MEME) are displayed on the AAV6 capsid. (C) AAV5, residue R710 (cyan) was detected by MEME. (D) AAV8, residues A551 (light-green) and Q594 (forest-green) were both detected by MEME. (E) AAV9, residues A273 (orange), D327 (deep-salmon) and N598 (chocolate) were detected by MEME. The black triangle depicts a viral asymmetric unit, with the solid pentagon (5-fold), triangle (3-fold), and oval (2-fold) indicating the icosahedral symmetry axes respectively. In the stereographic Roadmap projections, the boundary for each residue is in black, and the residues are as labeled.
Table 3.
Putative functional implications of positively selected sites in AAV capsids.
Fig 2.
Selection analysis of AAV capsid coding sequences and the evolutionary pathway of AAV2 capsid site 410.
The full-length AAV capsid coding sequences were subjected to recombination analysis and then partitioned into segments without recombination. The segment 1735–1773 was used for reconstruction of a phylogenetic tree with the global MG94 x REV model by the MEME program. The sequence isolates were labeled with reference to the source species (HU, human; RH, rhesus macaque; CY, cynomolgus macaque; BB, baboon; CH, chimpanzee) or further to the related human tissues (T, tonsil/adenoid; LG, lung; S, spleen) [5,6]. Partial AAV isolates were removed from the tree because of their duplicated sequences. Site 410 was identified as experiencing positive diversifying selection on a number of branches, which are shown in red. To further illustrate the evolutionary pathway of this site, the amino acid residues corresponding to each branch or cluster were annotated, and the replacement events relevant to positive selection are shown above or adjacent to the branches. The scale for genetic distance is indicated at the top.
Fig 3.
Selection analysis of AAV capsid coding sequences and the evolutionary pathway of AAV2 capsid site 548.
After recombination analysis of the full-length AAV capsid coding sequences, segment 2572–3468 was used for phylogenetic reconstruction with the global MG94xREV model by the MEME program. The sequence labeling was the same as that in Fig 2, with the redundant sequences removed for clear observation. The major AAV clades are indicated to the right of the taxa, as previously described [5]. Site 548 was identified as experiencing positive diversifying selection on some branches, which are colored in red. To further illustrate the evolutionary pathway of this site, the amino acid residues corresponding to each branch or cluster were annotated, and the replacement events inferring positive selection are displayed below or adjacent to the branches. The scale for genetic distance is shown at the top.
Fig 4.
Exploration of the immunological drive for the evolution of site 410 of the AAV2 capsid gene lineage.
(A) The sequence of the 15-mer peptide 82, with site 410 being variable among AAV lineages. This T-cell epitope was originally identified in a human subject by the IFN-γ ELISPOT assay [15]. Peptide 82, with glutamine (Q) and threonine (T) at site 410, was hypothesized to represent cognate or variant epitopes with variable affinities recognized by human or mouse MHCII molecules (S2 Table). (B) ELISPOT assay to examine the immunological effects of a mutation at site 410 from Q to T in the AAV2 capsid in a human subject. Approximately 100 human plasma samples were prescreened by ELISA for hu.1 capsid binding, which contains the peptide 82-Q410 epitope with a high prevalence in human populations [5]. PBMCs were then isolated from the positive blood samples to detect IFN-γ secretion by ELISPOT assay after incubation with peptide 82-Q410. Finally, human PBMC samples that were positive according to both ELISA and ELISPOT prescreening were subjected to IFN-γ ELISPOT assays to compare the mutation effects of the peptides 82-Q410 and 82-T410 using a paired t-test. (C), (D) Immunological effects of replacement at site 410 of the AAV2 capsid in BALB/c mice. An MHCII-binding peptide analogous to peptide 82 was predicted by RANKPEP analysis in BALB/c mice. Four BALB/c mice from each group were intramuscularly immunized with Ad-AAV2Cap or PBS. Splenocytes were harvested 9 days after immunization for incubation with the peptides 82-T410 or 82-Q410 before intracellular cytokine staining to detect IFN-γ secretion in CD4+ T cells. Representative flow cytometry images are shown in (C), and an unpaired t-test was performed to compare the reactivity of CD4+ T cells to AAV2 epitopes in (D). (E), (F) Localization of site 410 in the AAV2 structure. A ribbon model of an AAV2 VP3 pentamer is shown from the inside of the capsid (E) and in a side view (F). The 410 sites are represented as sphere models in yellow.
Fig 5.
Mutational effects of AAV2 capsid site 548 on the binding and neutralization of mouse monoclonal antibodies.
(A) Preparation of mouse C11 monoclonal antibody. The AAV2-G548 vector containing the lacZ transgene was used for intramuscular immunization of BALB/c mice, and splenocytes were subsequently isolated for fusion with the SP2/0 myeloma cell line. The generated hybridomas were diluted for culture and ELISAs using a panel of engineered AAV vectors, including AAV2-G548, AAV2.1 and AAV1.2, to screen the antibody with the cognate epitope with that of mouse A20 monoclonal antibody [33]. Created by BioRender.com. (B) Binding of the AAV2-G548 and AAV2-E548 capsids to a mouse C11 monoclonal antibody, as determined by by ELISA. (C) Neutralization of the AAV2-G548 and AAV2-E548 vectors by the C11 antibody. The recombinant AAVs harnessing the lacZ gene were incubated with a serially diluted antibody at 37°C for 1 hour before being inoculated into 293 cells. Transgene expression was monitored 24 hours later by a beta-galactosidase assay. (D) Binding of the AAV2-G548 and AAV2-E548 capsids with a mouse A20 monoclonal antibody. (E) Neutralization of the AAV2-G548 and AAV2-E548 vectors by the A20 antibody. The data in (B, D, E) were further used for calculation of EC50 values from ELISAs (F, G) and IC50 values from neutralization experiments (H) for comparison between the two AAV2 variants using paired t-tests. Each experiment was performed in duplicate, and the data are shown as the mean values ± SDs.
Fig 6.
Immunological effects of AAV2 capsid mutations on the binding and neutralization of human antibodies.
(A) Binding of the AAV2-E548 and AAV2-G548 capsids to individual human plasma samples by ELISA. The 49 human samples were roughly categorized into two groups based on their distinct OD450 readings in the preliminary experiment. The calculation and comparison of their EC50 values for AAV2-G548 and AAV2-E548 further supported the significant differences between these groups according to the Mann-Whitney test. (B, C, D) Neutralization experiments on the 14 human plasma samples selected by their different antibody titers against AAV2-G548 and AAV2-E548 (the different OD450 groups in (A)). Recombinant AAVs expressing the lacZ gene were incubated with serially diluted antibodies at 37°C for 1 hour before being inoculated into 293 cells. Transgene expression was monitored 24 hours later by a β-galactosidase assay. Two of the plasma samples did not neutralize AAV2 capsids (indicated by NA), and eleven of the remaining twelve plasma samples showed higher IC50 values for AAV2-E548 than for AAV2-G548 (B). Furthermore, the IC50 values from the four human plasma samples with greater differences between the two AAV2 variants are displayed in (C), and a representative neutralization curve is shown for plasma sample number 40# in (D). The indicated p values were obtained by paired t-tests in (C), with the significant difference between IC50 values at the 0.05 level shown by asterisks in (B). (E) Binding of the AAV2-E548 and AAV2-G548 capsids to pooled human plasma (IVIG) by ELISA. (F) Neutralization of the AAV2-E548 and AAV2-G548 vectors by IVIG. Paired t-tests were performed to compare the EC50 values of the AAV2-G548 and AAV2-E548 capsids for binding (G) and the IC50 values for neutralization (H) by IVIG. The data are shown as the mean values ± SDs.
Fig 7.
Evolution of AAV2 variants in a mouse model in the presence of human neutralizing antibodies.
(A) Distribution of AAV2 variants in mouse liver after one round of screening. An AAV2 library was constructed with saturation mutagenesis at the capsid site 548. A total of 1012 v.g. of this library was intravenously injected into SCID mice pretreated with a physiological concentration of human IVIG 24 hours prior. Mouse livers were collected for genomic DNA extraction and capsid gene sequencing 2 weeks later. (B) Characterization of AAV2 variants on their tissue tropism in the mouse liver. A total of 1010 v.g. of pooled AAV2 variants of the same ratio was injected intravenously into the SCID mice without IVIG pretreatment. Mouse livers were collected for examination of gene frequencies of AAV2 capsid genes 2 weeks later. (C) Neutralization effects on AAV2 variants by individual human plasma samples exemplified by the plasma sample number 40#. AAV2 variants harnessing the lacZ reporter gene were incubated with serially diluted human plasma samples for 1 hour at 37°C before inoculation into 293 cells. The cells were harvested for the β-galactosidase assay 24 hours later. (D) Neutralization effects of human IVIG on AAV2 variants. The results were collected from four mice in (B) and from two cell monolayers in (C) and (D). One-way ANOVA followed by Dunnett’s multiple comparisons test was performed to compare the other AAV2 variants to AAV2-E548 in (B), (C) and (D), and significant differences at the 0.05 level are indicated by asterisks. The data are shown as the mean values ± SDs.