In silico improvement of affinity for highly protective anti-malarial antibodies

Abstract

The monoclonal antibody CIS43 preferentially binds the junctional region of Plasmodium falciparum circumsporozoite protein (PfCSP) and is highly protective in humans. Here, we develop an in silico pipeline to improve antigen-antibody interaction energies and apply it to CIS43 variants elicited in CIS43-germline knock-in mice. Improved binding of CIS43 variants to the CIS43 junctional epitope (PfCSP peptide 21) was achieved by introducing single and double amino acid substitutions in the peptide 21-proximal heavy- and light-chain-variable regions. The best in silico designed variant, antibody P3-43-LS, was 2- to 3-fold more protective than antibody CIS43-LS, the clinical version of CIS43 with half-life extending leucine-serine (LS) mutations, and had comparable protection to the current best-in-class antibody (iGL-CIS43.D3-LS) to this region. Crystal structures of the improved antibodies revealed atomic-level interactions accounting for gains in binding affinity. This in silico approach to improve antibody affinity can thus be used to enhance potency of PfCSP monoclonal antibodies.

Keywords: Biological sciences; Immunology; Natural sciences; Structural biology.

Graphical abstract

Figure 1

Antibody improvement pipeline identifies CIS43 variants with improved in silico non-bonded interaction energies to junctional peptide 21 (A) Schematic of PfCSP sequence and junctional peptides. (B) Antibody improvement pipeline is composed of three sequential steps. First, in silico interaction energies are calculated. Second, antigenic filtering down-selects most promising antibody variants. Lastly, functional assessment compares antibody variants against current best-in-class antibodies. (C) Pareto optimal solutions from in silico energies for all single mutants based on m42.127 and m43.151 templates. Mutations within the top pareto fronts are depicted in black and red circles. Variants satisfying all stability constraints are highlighted in red. (D) Allowable region for mutations within 12 Å of peptide 21 (green) for m42.127 (left) and m43.151 (right) is shown in gray. Residue positions with pareto-optimal energies identified in silico are shown in red. See also Figures S1 and S3.

Figure 2

In silico designed CIS43 variants show improved binding to peptide 21 (A) Heavy chain sequences for the top mutants as characterized by purified AlphaLISA signals. (B) Biolayer interferometry (BLI) affinity of selected CIS43 variants to peptide 21. Variants are grouped depending on their template antibody. Error bars correspond to the standard error in affinity measurement. (C) In silico interface energy strongly correlates to peptide 21 BLI Kd measurements for m42.127-based variants. See also Figure S2.

Figure 3

Functional characterization reveals best in silico-designed antibody P3-43 is superior to clinical template antibody CIS43 (A) Schematic of mouse malaria challenge model. (B) Liver burden protection and parasitemia at 50 μg and 100 μg dosage were assessed for iGL-CIS43.D3-LS and P3-43-LS along with clinical antibody CIS43-LS at dosage of 100 μg. Relative to clinical antibody CIS43-LS, both iGL-CIS43.D3-LS and P3-43-LS are significantly more protective as assessed by uncorrected Dunn test. (C) Repeat protection experiments, conducted by a different research team, showed an almost identical outcome with iGL-CIS43.D3-LS and P3-43-LS being significantly more superior than CIS43-LS, as assessed by uncorrected Dunn test.

Figure 4

In silico energies and crystal structure of P3-21 in complex with peptide 21 depict additional atomic interactions explaining increase in affinity (A) In silico pairwise energy analysis. For each pair of residues between peptide 21 (y axis) and the antibody (x axis), the total interface energy for the antibody is subtracted from the corresponding template antibody. Lower values indicate more favorable interaction energies. In the bar plots below, values are summed across the peptide 21 positions to further examine the CDRH3 region. (B) Crystal structures of P3-21 (top left panel in brown) and P3-43 (top right panel in red) in complex with peptide21 (green) is shown in cartoon illustration. Mutated residues are depicted in cyan stick conformation. Critical binding regions (bottom dotted panels), including heavy chain position 99 and 100 of template antibody m42.127 and P3-21 in complex with peptide 21. Amino acid Arg99 in P3-21 introduces new electrostatic interactions with peptide 21 backbone atoms. Interatomic interactions from residue 99 in heavy chain and residue 9 in peptide 21 is displayed and compared with that of template m42.127 crystal structure. P3-43 mutation Lys100 introduces new interactions with peptide 21 Asp11. See also Figure S4.