Engineering of HIV-1 neutralizing antibody CAP256V2LS for manufacturability and improved half life

Abstract

The broadly neutralizing antibody (bNAb) CAP256-VRC26.25 has exceptional potency against HIV-1 and has been considered for clinical use. During the characterization and production of this bNAb, we observed several unusual features. First, the antibody appeared to adhere to pipette tips, requiring tips to be changed during serial dilution to accurately measure potency. Second, during production scale-up, proteolytic cleavage was discovered to target an extended heavy chain loop, which was attributed to a protease in spent medium from 2-week culture. To enable large scale production, we altered the site of cleavage via a single amino acid change, K100mA. The resultant antibody retained potency and breadth while avoiding protease cleavage. We also added the half-life extending mutation LS, which improved the in vivo persistence in animal models, but did not impact neutralization activity; we observed the same preservation of neutralization for bNAbs VRC01, N6, and PGDM1400 with LS on a 208-virus panel. The final engineered antibody, CAP256V2LS, retained the extraordinary neutralization potency of the parental antibody, had a favorable pharmacokinetic profile in animal models, and was negative in in vitro assessment of autoreactivity. CAP256V2LS has the requisite potency, developability and suitability for scale-up, allowing its advancement as a clinical candidate.

Figure 1

CAP256-VRC26.25 mutations at the proteolytic cleavage site have a variable effect on neutralization potency. (a) CDRH3 sequence of wild-type CAP256-VRC26.25. The arrow and slash indicate the site of cleavage. The positions are labeled according to the Kabat numbering scheme. Cysteines involved in a disulfide bond are highlighted in green. (b) The wild-type CAP256-VRC26.25 Fab is shown (PDB ID 5DT1) with the CDRH3 highlighted to show the location of the cleavage between K100m and Q100n and its relation to the disulfide bond (green) and sulfated tyrosines (stick representation). (c) Neutralization of Env-pseudovirus panel by CAP256.25 variants. Each Ab was expressed using the indicated CAP256-VRC26.25 heavy chain paired with wild type light chain. nd not done. Data are IC50 in μg/ml.

Figure 2

CAP256-VRC26.25 variants resist cleavage. (a) Samples were treated with day 19 CHO supernatant flow-through containing protease (Day 19FT) for 0 or 72 h. Treated and untreated samples were separated by size using reducing capillary gel electrophoresis. Peaks labelled HCC1, HCC2 contain clipped product; HC contains unclipped heavy chain, LC contains light chain. (b) % Cleaved heavy chain showing 0 and 72 h treatments. % HCC was calculated as area of each peak divided by total area of all peaks.

Figure 3

K100mA mutant maintains breadth and potency as assessed on large virus panels. (a) Neutralization by CAP256-VRC26.25 wild type, K100mA, and CAP256V2LS was measured against a multi-clade 208 Env-pseudovirus panel (left) and 100 acute Clade C Env-pseudoviruses (right). Each dot represents one pseudovirus. (b) Dendrogram shows phylogenetic relatedness of Env sequences in multi-clade panel, color-coded by neutralization sensitivity.

Figure 4

Position 100m in the CAP256-VRC26.25 CDRH3 does not contact the HIV-1 Env. (a) Cryo-EM structure of CAP256-VRC26.25 in complex with the CAP256.wk34.c80 SOSIP.RnS2 Env trimer. Trimer is shown in surface rendering, with protomers in grey, pink, and light blue; glycans in green; antibody heavy chain in yellow ribbon, light chain in slate ribbon. (b) Close-up of the unliganded CDRH3 conformation, highlighting the K100m contact with D100g (top) in contrast to the bound Fab where D100g is making electrostatic contacts with K169 of the Env. (c) (left) K100m is located above the central cavity in the Env apex facing the solvent. (right) modeled alanine at position 100m indicates no addition or loss of Env contacts. (d) The central cavity relative to position 100m is shown at a 90° rotation from (c).

Figure 5

CAP256V2LS has a favorable pharmacokinetic (PK) profile in animal models and no autoreactivity in vitro. (a) PK in rhesus macaques (non-human primates, NHP). Naïve rhesus macaques were infused with 10 mg/kg of antibody, and the antibody concentrations were measured by an anti-idiotype based ELISA. Each line represents one animal. (b) PK in human FcRn mice. Naïve human FcRn mice were infused with 5 mg/kg of CAP256V2LS or VRC07-523LS and antibody concentrations were measured by an anti-idiotype based ELISA. (c) Hep-2 cell staining. (d) Cardiolipin binding at 100 and 33 μg/ml inputs; values shown as GPL units. VRC01-LS, 4E10, VRC07-523-LS, and VRC07-G54W are included as controls in (c,d).