Vector design Tour de Force: integrating combinatorial and rational approaches to derive novel adeno-associated virus variants

Abstract

Methodologies to improve existing adeno-associated virus (AAV) vectors for gene therapy include either rational approaches or directed evolution to derive capsid variants characterized by superior transduction efficiencies in targeted tissues. Here, we integrated both approaches in one unified design strategy of "virtual family shuffling" to derive a combinatorial capsid library whereby only variable regions on the surface of the capsid are modified. Individual sublibraries were first assembled in order to preselect compatible amino acid residues within restricted surface-exposed regions to minimize the generation of dead-end variants. Subsequently, the successful families were interbred to derive a combined library of ~8 × 10(5) complexity. Next-generation sequencing of the packaged viral DNA revealed capsid surface areas susceptible to directed evolution, thus providing guidance for future designs. We demonstrated the utility of the library by deriving an AAV2-based vector characterized by a 20-fold higher transduction efficiency in murine liver, now equivalent to that of AAV8.

Figure 1

Nucleotide sequence of CapLib variable regions (VRs) and corresponding amino acid diversity. Each VR region shows a nucleotide sequence (lowercase letters) as designed and encoded by synthetic oligonucleotides, including degenerate positions (IUPAC nucleotide code) highlighted in yellow. The corresponding amino acid sequence (UPPERCASE letters above nucleotide sequence) encoded by all possible codon combinations is shown in the line/s above. Amino acids highlighted in blue were selected for the inclusion into the library sequence because they were found in one or more naturally occurring 150 variants used for the library design (Supplementary Figure S1), whereas those that are not highlighted represent an additional amino acid diversity encoded by the degenerate nucleotides. Highlighted in orange are amino acid residues not encoded by the wild-type AAV2 sequence that were introduced during design to increase transduction efficiency (Y/F) or eliminate heparin binding (R/A). The amino acid residues in logo style shows the alignment of the sequences deduced from the next-generation sequencing of the library viral DNA. The x-axis designates residue position (VP1 numbering) and the y-axis the relative frequency of each amino acid at that position. Amino acids are colored according to their chemical properties: polar amino acids are green; basic, blue; acidic, red; and hydrophobic amino acids are black.

Figure 2

Schematic representation of the CapLib design and construction steps. Adeno-associated virus (AAV)-2 capsid is shown at the top of the chart whereby numbers represent the respective amino acid residues. Variable region (VR) sublibraries were packaged and purified (indicated by differentially colored VRs). DNAs from these individual sublibraries were purified and used to construct a final combined library.

Figure 3

Variable regions (VRs) on the AAV2 capsid surface. (a) Adeno-associated virus (AAV2) capsid full capsid (assembled from 60 VP3 monomers shown to the right), displaying the location of the eight VRs, colored as described below the image of the monomer. (b) LiA and LiC full capsids variants and their respective monomers. The mutated residues and their positions are shown by color and by number, respectively.

Figure 4

Analysis of packaged viral library complexity. (a) Shannon entropy of the viral library computed from 840 individual protein sequences. The x-axis designates residue position (VP1 numbering) and the y-axis computed value of the entropy. (b) Distribution of the distinct capsid sequences and their copy numbers. (c) Distribution of the number of mutations in the sample of analyzed protein sequences. (d) Distribution of the number of mutant VRs. The most frequently encountered mutant VRs combinations are marked by the Roman numerals over the respective graph areas. (e) Comparison of expected and observed percentages of mutant residues. The x-axis designates residue position (VP1 numbering); the respective VR boundaries are shown below amino acid numbers for better orientation. The y-axis designates the percentage of the expected (introduced by design) mutation (blue) versus experimentally deduced from the sequencing of the plasmid library (red) versus viral library (green). The similitude of the graphs height at each position indicates a relative neutrality of this position for the capsid assembly/structure, whereas the difference between a theoretical (blue) and plasmid-derived sequences (red) indicates a selective pressure against this particular mutated residue within one single VR library or within a context of all mutated VRs in the combined viral library (green). Because the sample of the sequenced plasmid library included only 21 individually sequenced random clones, due to the low sampling representation, the number of observed mutants (red) sometimes exceeds the expected frequency (blue) which always (with the exception of residues 444, 500, and 588) includes wild-type AAV2 residue.

Figure 5

Selection of new variants from murine liver. (a) Amino acid sequences of the variable regions (VRs) of the variants selected from the library after three rounds of directed evolution in the mouse liver. Wild-type AAV2 sequences are shown in bold for comparison. (b) Expression of luciferase transgene in the liver of mice at day 40 after tail-vein administration of rAAV vectors packaged into capsids as indicated. (c) Time course of rAAV-Luc expression in the liver as assessed by luciferase imaging in vivo. The y-axis designates the level of bioluminescence (n = 3 mice per group). (d) Systemic expression of human F.IX in hemophilia B (C3H/HeJ/F9^−/−) mice over 3 months following peripheral vein delivery of ApoE/hAAT-hF.IX transgene cassette packaged in AAV8 (red), LiC (green), or AAV2-M3 (triple AAV2 mutant Y444F, Y500F, and Y730F, blue) capsid (10¹⁰ vg/mouse, n = 4/group). (e) Systemic expression of human F.IX in hemophilia B (C3H/HeJ/F9^−/−) mice at 4 weeks following peripheral vein delivery of ApoE/hAAT-hF.IX transgene cassette packaged either in AAV8 (red) or LiA (yellow) capsid (10¹⁰ vg/mouse, n = 4/group). (f) Biodistribution study of LiA and LiC variants (tabulated as a number of viral genomes per nanogram of genomic DNA, y-axis) conducted 23 days after peripheral vein delivery in BALB/c male mice (n = 3), as compared to AAV2 and AAV8 (x-axis). Note predominantly liver-specific targeting of the selected variants LiA and LiC.