Surface-Matrix Screening Identifies Semi-specific Interactions that Improve Potency of a Near Pan-reactive HIV-1-Neutralizing Antibody

Abstract

Highly effective HIV-1-neutralizing antibodies could have utility in the prevention or treatment of HIV-1 infection. To improve the potency of 10E8, an antibody capable of near pan-HIV-1 neutralization, we engineered 10E8-surface mutants and screened for improved neutralization. Variants with the largest functional enhancements involved the addition of hydrophobic or positively charged residues, which were positioned to interact with viral membrane lipids or viral glycan-sialic acids, respectively. In both cases, the site of improvement was spatially separated from the region of antibody mediating molecular contact with the protein component of the antigen, thereby improving peripheral semi-specific interactions while maintaining unmodified dominant contacts responsible for broad recognition. The optimized 10E8 antibody, with mutations to phenylalanine and arginine, retained the extraordinary breadth of 10E8 but with ∼10-fold increased potency. We propose surface-matrix screening as a general method to improve antibodies, with improved semi-specific interactions between antibody and antigen enabling increased potency without compromising breadth.

Keywords: 10E8; HIV-1; MPER; antibody improvement; broadly neutralizing antibody; membrane-proximal external region; surface-matrix screening.

Figure 1. Surface-Matrix Screening to Improve Neutralization Potency of Antibody 10E8

(A) Hypothesis-driven deductive approach (left) versus hypothesis-independent surface-matrix approach (right). The broad sampling of the matrix approach provides extensive information linking surface chemistry to biological functionalities of interest. Shown is the 10E8 Fab with heavy chain in light blue and light chain in white, and with gp41 peptide in magenta. (B) Schematic flow diagram of surface-matrix approach. (C) Selection of a 9-virus panel (shown as red in dendrogram) from the 208-isolate panel for neutralization assessment. The clade of each selected strain is displayed in parentheses. (D) Experimental variation in neutralization. Strains with non-sigmoidal neutralization curves and high assay-replication error are shown in italics. (E) Minimum observable signal as a function of assay variability and number of variants screened (see Figure S2). (F) Neutralization IC₅₀ by 10E8 variants, arranged by geometric mean IC₅₀-fold improvement for all viruses shown. “>50” was considered as “50” in geometric mean calculations. (G) Crystal structure of gp41 peptide (magenta) complexed with improved antibody 10E8v4-5R+100cF (light blue, heavy chain; white, light chain). See also Figures S1 and S2 and Tables S1C and S2.

Figure 2. Interactive Surfaces Identified by Surface-Matrix Screening

(A) Crystal structure of Fab 10E8 is shown with residues (cyan) that, when altered by surface-matrix screening, decreased neutralization by over 3 SDs (3.4-fold). Residue alteration types are noted above each structure, with all surface-matrix screening results are shown in Figure S1. Notably, two surfaces, one facing the expected position of the Env trimer and the other facing membrane, are identified by clusters of inhibitory N-glycan mutations. These are not seen in Arg mutations, but the putative membrane-interactive surface is identified by both Phe/Trp and 7-glycine mutations. Note that orientations shown in (A) allow better labeling and are slightly different from 10E8 orientation in (B). (B) Orientation of 10E8 Fab bound to detergent-solubilized Env trimer from a detergent-solubilized complex with PGT151 (PDB: 5FUU and EMDB: 3312). See also Figures S1 and S2 and Table S1C.

Figure 3. Chemical Preferences of Functional Hotspots Identified by Surface-Matrix Screening

(A) Effects of residue substitutions at heavy chain position 5. (B) Location of R at position 5 and F at position 100c alteration relative to structure of 10E8 bound to detergent-solubilized Env trimer (insets show residue environments surrounding the two hotspots). (C) Effects of residue substitutions at heavy chain position 100c.

Figure 4. MPER-Directed Antibody 4E10 Shares the Membrane-Binding Functional Hotspot of 10E8

(A) Structural model of antibody 10E8v4 recognizing lipid headgroups, as defined by Irimia et al. (2017), with hotspot position 100c shown in green. (B) Structural model of antibody 4E10 recognizing lipid headgroups, as defined by Irimia et al. (2016). (C) Virus neutralization by 4E10 and variants showing 100aW with improved activity. (D) Superposition gp41 peptide as recognized by 4E10 and 10E8 identifies 4E10 CDR H3 as a hot-spot of functional enhancement.

Figure 5. Enhanced Binding to gp41 Peptide in a Membrane Context by 10E8v4-100cW and 4E10-100aW IgGs Modeled Membrane-Epitope Co-recognition and Polyreactivity for Several HIV-1-Neutralizing Antibodies

(A) Binding of 10E8v4, 4E10, 10E8v4-100cW, and 4E10-100aW variants to MPER peptide. IgG, immunoglobulin G. K_D, equilibrium dissociation constant. (B) Binding of 10E8v4, 4E10, 10E8v4-100cW, and 4E10-100aW variants to lipid bilayers containing MPER peptide. (C) Membrane proximal residues (shown in spheres) for 10E8, 4E10, CAP248-2B, DH511.2, and 2F5. Positions of viral membrane surface approximated by dashed line (Supplemental Experimental Procedures). (D) Membrane interaction propensity of membrane-proximal residues. (E) HEp2 cell binding versus lipophilicity. Error bars represent SD. (F) Cardiolipin binding versus lipophilicity. See also Figure S3.

Figure 6. Semi-specific Interactions with Glycan Shield Observed with Hotspot around Heavy Chain Position 5

(A) Modeled glycan shield co-recognition of Arg hits identified by surface-matrix screening of 10E8 variants. (B) Impact of Arg additions on the neutralization (and polyreactivity) of 10E8 variants. (C) Average improvement in 10E8 variant neutralization upon the addition of 1 or 2 Arg. Error bars represent SD. See also Figure S4 and Table S1D.

Figure 7. Manufacturing Characteristics, Serum Half-Life, and Neutralization Potency of 10E8 Variants

(A) Manufacturing properties of optimized 10E8 variants. CD, circular dichroism; DLS, dynamic light scattering; DSC, differential scanning calorimetry; OD₃₅₀, optical density 350; T_m, melting temperature. (B) Pharmacokinetics of 10E8 variants in rhesus macaque. Error bars represent SD. (C) Potency of optimized 10E8v4-5R+100cF on a panel of 208 Env pseudoviruses: antibodies shown are being developed for clinical evaluation. Note that the physical combination of N6 with 10E8v4-5R+100cF neutralizes all strains in the 208-isolate panel at less than 1 µg/mL. The median IC₅₀ based on all 208 viruses (including all resistant strains) for each antibody is displayed. (D) Completeness of neutralization by single antibodies (10 µg/mL) and antibody combinations with 5 µg/mL of each antibody (predicted using the Bliss Hill model). The numbers on top indicate the percent viruses that were predicted to be neutralized at >95%. See also Figure S5 and Tables S1E and S1F.