Antibody-directed evolution reveals a mechanism for enhanced neutralization at the HIV-1 fusion peptide site

Abstract

The HIV-1 fusion peptide (FP) represents a promising vaccine target, but global FP sequence diversity among circulating strains has limited anti-FP antibodies to ~60% neutralization breadth. Here we evolve the FP-targeting antibody VRC34.01 in vitro to enhance FP-neutralization using site saturation mutagenesis and yeast display. Successive rounds of directed evolution by iterative selection of antibodies for binding to resistant HIV-1 strains establish a variant, VRC34.01_mm28, as a best-in-class antibody with 10-fold enhanced potency compared to the template antibody and ~80% breadth on a cross-clade 208-strain neutralization panel. Structural analyses demonstrate that the improved paratope expands the FP binding groove to accommodate diverse FP sequences of different lengths while also recognizing the HIV-1 Env backbone. These data reveal critical antibody features for enhanced neutralization breadth and potency against the FP site of vulnerability and accelerate clinical development of broad HIV-1 FP-targeting vaccines and therapeutics.

Fig. 1. Precision library generation and yeast display screening were applied to the template antibody VRC34.01 to enhance recognition of HIV-1 Envelope (Env)-displayed fusion peptide (FP).

A Workflow for precision anti-HIV-1 antibody engineering via yeast display. The VRC34.01 template antibody variable region genes were mutated and selected for improved HIV-1 affinity via successive rounds of site-saturated mutagenesis (SSM) and DNA shuffling. Antibody mutant libraries were screened using FACS to fractionate mutant populations by antigen binding affinity phenotypes. Next-generation sequencing (NGS) was used to mine antibody sequences and bin antibody variants for trimer binding function based on quantitative variant prevalence analysis across sort groups. Biophysical characterization of engineered single- and multi-mutation antibody variants revealed anti-FP antibody sequence-structure-function relationships and defined potent gain-of-function mechanisms that enhanced FP-targeted HIV-1 neutralization. B Yeast libraries expressing antibody in a surface-bound fragment antigen binding (Fab) format were stained with fluorescence markers to measure Fab-surface expression (Y-axis) versus antigen binding (X-axis) and screened via FACS. Template VRC34.01 (top row) and single-mutant amino-acid substitution libraries were generated via SSM across the entire VRC34.01 variable region heavy (VH) and variable region light (VL) genes (middle row: pre-sorting populations; bottom row: round 3 high-affinity enriched populations). Libraries are shown bound to diverse FP sequences displayed on HIV-1 trimer probes. Single-mutant libraries were fractionated in three sequential rounds into high-, medium-, and low-affinity performance bins using FACS, resulting in sorted libraries with phenotypically observable differences in trimer binding. Medium-, and low-affinity plots are also provided in Supplemental Fig. 1A.

Fig. 2. Bioinformatic mining of single-mutation NGS data from SSM library screens revealed multiple mutations that provided enhanced HIV-1 neutralization potency and breadth.

A Enrichment ratios (ER) are plotted for single mutant antibody sequences derived from Round 3 high-affinity sorted libraries. Mutations were defined by determining the percent identity match to the template gene and denoting the substituted amino acid residue relative to the template sequence, with Base 1 corresponding to the start of the antibody variable region. NGS analysis of single mutant library screens highlighted multiple amino acid substitutions across VRC34.01 antibody variable regions that could enhance diverse HIV-1 FP recognition. B A 20-virus panel was optimized for predictive correlation with neutralization breadth on a larger 208-virus panel as a pre-screening tool to down-select promising candidates prior to 208-virus panel neutralization analysis. Template antibody VRC34.01 recognized 60% of strains in the 20-virus panel and ~50% in the 208-virus panel. PGT151, the broadest previously reported antibody that interacts with FP, recognized 70% of strains in the 20-virus panel and ~60% of the broader 208 strains. One-sided p-value for significance of data points deviating from a straight line is shown. No adjustments were made for multiple comparisons. C Four single mutations discovered by SSM screening (VH_T59Y, VH_T59F, VH_E2K, VH_E2P) were expressed as soluble IgG and revealed neutralization to 75% of the 20-virus panel. Rational combinations of these top single-mutant variants further improved neutralization breadth, with a maximum 85% breadth on the 20-virus panel achieved for VH_E2K_T59F (also referred to as VRC34.01-Combo1, Kabat numbering VH_E2K_T59F). Antibodies were considered neutralizing if their IC50 potency was <50 μg/mL. D 208-virus panel data revealed enhanced breadth for Combo1 (VH_E2K_T59F) to 68% of all strains, with strong neutralization of FP_v1 strains (98%), moderate neutralization of FP_Thai strains (61%), and gain-of-function for neutralization of FP_v3 and FP_v4 strains that were not neutralized at all by VRC34.01 (43% and 11%, respectively). Combo1 showed improvements compared to the template VRC34.01 antibody against all HIV-1 FP subclasses.

Fig. 3. Structural basis for VRC34.01-Combo1’s recognition of diverse FP.

A Crystal structure of VRC34.01-Combo1 Fab in complex with fusion peptide variant FP8_v1. B CryoEM structure of BG505 DS-SOSIP HIV trimer in complex with the VRC34.01_Combo1 Fab. C, D Details of interactions between VRC34.01-Combo1 (cyan/orange/magenta) bound to peptide FP8v1 (AVGIGAVF) that provided improved neutralization breadth compared to parental VRC34.01 (green/yellow/violet). E Representative electron density around critical residues W50_HC, F58_HC, Y94_LC, V513_FP, G514_FP, and I515_FP.

Fig. 4. Targeted multi-mutation screening further enhanced neutralization of FP8_v3 and FP8_v4 strains, enabling best-in-class anti-FP neutralization breadth and potency.

A Multi-mutation yeast libraries expressing antibody as surface-bound Fab were stained with fluorescence markers to measure Fab-surface expression (Y-axis) versus antigen binding (X-axis) and screened via FACS. Multi-mutant yeast display libraries (red) were compared to surface-expressed VRC34.01 template Fab (blue). Multi-mutant library generation and FACS screening provided phenotypically apparent affinity improvements against BG505 SOSIP trimers encoding FP_Thai, FP_v3, and FP_v4 relative to template mAb VRC34.01, with the goal of further improving HIV-1 neutralization by Combo1 (VH_E2K_T59F, Fig. 2D). B Heat maps of enriched mutant sequences from screening multi-mutation libraries 1, 2, 3, 4, and 5 against trimer probes displaying four different FP variants. Bioinformatic mining of multi-mutation yeast display library screening data revealed several mutations enriched against diverse FP sequences. C When expressed as soluble IgG, numerous multi-mutation variants achieved 95% neutralization in the 20-virus panel, with one variant (VRC34.01_mm28, or mm28, VH_E2K_A33P_T59F) achieving 100% neutralization of HIV-1 strains in the panel. Antibodies were considered neutralizing with an IC50 potency <50 μg/mL.

Fig. 5. Top VRC34.01 mutant variant characterization out performs template antibody via stepwise improvements in potency and breadth via engineered mutations.

A 208-virus panel neutralization data showed enhanced breadth compared with the template VRC34.01 antibody against all HIV-1 FP subclasses for antibody variant mm28 (VH_E2K_A33P_T59F). When binning neutralization by FP sequences screened in yeast display, mm28 showed 100% neutralization breadth of FP_v1 strains (vs. 91% breadth for the VRC34.01 template antibody), 83% neutralization of FP_Thai strains (vs. 11%), 86% neutralization of FP_v3 (vs. 0%), 22% neutralization FP_v4 strains, (vs. 0%), and 72% neutralization of other diverse FP sequences (vs. 43%). B Dendrograms reveal that the engineered antibodies Combo1 (68% total breadth) and mm28 (79% total breath) increased both cross-clade recognition and neutralization against VRC34.01-resistant and sensitive strains. C Titration curves show that Combo1 and mm28 gained stepwise improvements in neutralization potency and breadth compared to template VRC34.01. D Recognition of VRC34.01-resistant and sensitive HIV-1 strains displaying diverse FP sequences were compared for several antibodies in the study.

Fig. 6. Structural basis for VRC34.01-mm28 recognition of diverse FP.

A CryoEM reconstruction of VRC34.01_mm28 bound to BG505-DSSOSIP. False colors are shown for a single copy each of gp120, gp41 and the heavy and light chains of VRC34.01_mm28. B Superposition of VRC34.01 and VRC34.01_mm28 bound to BG505, showing no discernable differences in binding. C Crystal structure of VRC34.01_mm28 bound to FP8v4 (AVGTIGAM) with (D) an enlarged surface view of the peptide binding region. E Mutation of Ala33 to Pro in VRC34.01_mm28 causes a cascading shift in Tyr97 and Asn100, resulting in the enlarged binding cleft. F Superimposition of VRC-34.01_Combo1 bound to FP8v1 and VRC34.01-mm28 bound to FP8v4, demonstrating the broader binding cleft of mm28.

Fig. 7. Biophysical and genetic correlates for best-in-class FP-directed antibody neutralization.

A Rare mutations identified through gene-specific substitution profiles (GSSPs) critical for the recognition of FP and HIV−1 Env trimer. GSSPs for the VRC34.01_Combo1 heavy chain IGHV1-2 gene are shown. Mutations identified in yeast display were highlighted with cyan background. A substitution frequency <0.5% was defined as a rare mutation and colored with a red square. CDR1 and CDR2 were shown with green and blue boxes, respectively. B Binding of optimized antibodies to His-tagged FP Ala/Gly mutants. Binding of alanine (gray bars) and glycine (white bars) mutants within FP, normalized by binding to the wild-type sequence, are shown, with FP amino acids on the y-axis. Experiments were performed in triplicate (n = 3 independent experiments), error bars indicate the mean with SD plotted for each residue position in the overlaid scatter dot plots. C ITC-derived affinity to diverse FP sequences for various antibody variants. D Residue-Residue Pair Energy Analysis by 100 ns Molecular Dynamics Simulation to understand the roles of E2K, T59F, and A33P mutations in improved neutralization performance. Mean is represented as red dashed lines, and red arrows show the energy difference between mutants and the VRC34.01 (WT) antibody. Gray dashed lines represent VRC34.01 (WT) performance for comparison.