Combinations of Single Chain Variable Fragments From HIV Broadly Neutralizing Antibodies Demonstrate High Potency and Breadth

Abstract

Broadly neutralizing antibodies (bNAbs) are currently being assessed in clinical trials for their ability to prevent HIV infection. Single chain variable fragments (scFv) of bNAbs have advantages over full antibodies as their smaller size permits improved diffusion into mucosal tissues and facilitates vector-driven gene expression. We have previously shown that scFv of bNAbs individually retain significant breadth and potency. Here we tested combinations of five scFv derived from bNAbs CAP256-VRC26.25 (V2-apex), PGT121 (N332-supersite), 3BNC117 (CD4bs), 8ANC195 (gp120-gp41 interface) and 10E8v4 (MPER). Either two or three scFv were combined in equimolar amounts and tested in the TZM-bl neutralization assay against a multiclade panel of 17 viruses. Experimental IC₅₀ and IC₈₀ data were compared to predicted neutralization titers based on single scFv titers using the Loewe additive and the Bliss-Hill model. Like full-sized antibodies, combinations of scFv showed significantly improved potency and breadth compared to single scFv. Combinations of two or three scFv generally followed an independent action model for breadth and potency with no significant synergy or antagonism observed overall although some exceptions were noted. The Loewe model underestimated potency for some dual and triple combinations while the Bliss-Hill model was better at predicting IC₈₀ titers of triple combinations. Given this, we used the Bliss-Hill model to predict the coverage of scFv against a 45-virus panel at concentrations that correlated with protection in the AMP trials. Using IC₈₀ titers and concentrations of 1μg/mL, there was 93% coverage for one dual scFv combination (3BNC117+10E8v4), and 96% coverage for two of the triple combinations (CAP256.25+3BNC117+10E8v4 and PGT121+3BNC117+10E8v4). Combinations of scFv, therefore, show significantly improved breadth and potency over individual scFv and given their size advantage, have potential for use in passive immunization.

Keywords: HIV; HIV prevention; broadly neutralizing antibodies; combinations of scFv; single chain variable fragments.

Figure 1

Neutralization titers of single, dual, and triple combinations of scFv. Heat map showing IC₅₀ neutralization titers in μg/mL for single (A), dual (B), and triple (C) combinations of scFv against a panel of 17 subtype A, B, and C viruses. Viruses insensitive to individual bNAbs are shown as >30μg/mL. The fold improvement in IC₅₀ titers of the dual and triple combinations relative to the IC₅₀ of the best scFv in the combination is included in (B, C). Values with >3-fold increase or decrease in neutralization are shown in bold. Geometric mean potency is included at the bottom of each table. Each scFv is present in the combination at the titer indicated.

Figure 2

Combinations of scFv that show improved neutralization potency over single scFv. Neutralization curves showing potential additive, synergistic and antagonistic potency effects of combinations of antibodies. Combinations of scFv compared to single scFv neutralization curves with CAP256.25 in purple, PGT121 in light blue, 3BNC117 in orange, 8ANC195 in green and 10E8v4 in pink. The dual or triple combinations are represented in black. Geometric mean values are used for each data point with error bars representing repeat experiments. (A) Examples of combinations where neither synergy nor antagonism is observed for three viruses tested against dual combinations of scFv. Additive potency is represented by the combination curve (black) overlapping with the best scFv in the mixture. (B) Potential synergy as observed in dual combinations of scFv against three different virus strains. Synergy was represented by a left shift of the combination curve (black) relative to the most potent scFv in the mixture, or by steeper neutralization curves resulting in improved IC80. (C) Potential antagonism as represented by a right shift of the combination curve relative to the most potent scFv in the mixture, observed in combinations of two or three scFv tested against ZM249, which was sensitive to all scFv in the combinations.

Figure 3

Comparison of experimental and predicted combinations of dual and triple scFv. (A, B) IC₅₀ and IC₈₀ titers of dual antibody combinations plotted against the predicted IC₅₀ and IC₈₀ titers according to the Loewe Additive (left) and Bliss-Hill Independence (right) models. (C, D) Predicted IC₅₀ and IC₈₀ titers of triple combinations versus the experimental IC₅₀ and IC₈₀ titers. Values where both or all titers of single scFv >30μg/mL are excluded. (E, F) Comparison of the absolute Log(IC₅₀) and Log(IC₈₀) difference between the experimental titers and predicted titers according to the Loewe Additive (green) and Bliss-Hill independence (purple) for the dual combinations. (G, H) Comparison of the absolute Log(IC₅₀) and Log(IC₈₀) difference between the experimental titers and predicted titers according to the Loewe Additive (green) and Bliss-Hill independence (purple) models for the triple combinations. For (A, B, E, F), a nonlinear robust regression log-log line (red) and an equity line (black dotted) as well as r, R2, are shown and p<0.0001. For (C, D, G, H), mean values are given with a standard error of the mean. A paired, Wilcoxon t-test was performed on each pair with significant p values above the graphs (p < 0.05 = * p < 0.01 = ** p < 0.001= ***).

Figure 4

No synergy observed using fold difference between experimental and Bliss-Hill predictions of dual and triple scFv combinations. Log-fold differences of double and triple IC₅₀ (A, C) and IC₈₀ (B, D) experimental combinations to the Bliss-Hill Independence model prediction. Mean and the 95% confidence interval are shown with a dotted line indicating no fold change. Log (2.5) fold difference dotted lines (0.4 or -0.4) are also shown. Red and blue are used to indicate more than a log difference of 0.4 (~2.5x) with red dots indicating potential synergy and blue dots indicating potential antagonism.

Figure 5

Predicted breadth and potency of scFv combinations against large virus panel. (A) Scatter plot showing predicted IC₅₀ titers of dual combinations based on single scFv data for a panel of 45 viruses using the Bliss-Hill Independence model. The IC₅₀ data is plotted in μg/mL recalculated from nM based on the formula given in the methods section. (B) Predicted IC₈₀ titers of dual combinations based on the Bliss-Hill Independence model ranked by geometric mean. (C) Breadth-potency curves of IC₅₀ titers of most potent and broad dual combinations were plotted for the IC₅₀.with the others in grey. (D) Breadth-potency curves from IC₈₀ data for the most potent and broad dual combinations. (E) IC₅₀ titers for triple combinations based on the Bliss-Hill Independence model. (F) Predicted IC₈₀ titers for the triple combinations ranked based on geometric mean. (G) Breadth potency curves for the IC₅₀ of the most potent and broad triple combinations are plotted. (H) The breadth-potency curves are based on IC₈₀ titers of the most potent and broad combinations. The maximum neutralization for the IC₅₀ and IC₈₀ data is set at 50μg/mL for all plots. Geometric mean is indicated by a black line in A, C, E, and G Dotted lines are shown at 0.01μg/mL and 1μg/mL for breadth-potency plots IC₅₀ and the IC₈₀ (B, D, F, H). Each scFv is present in the combination at the titer indicated.

Figure 6

Neutralization breadth for single and combinations of scFv against a 45-virus panel at four concentrations. Percentage neutralization at 0.1, 1, 10, and 50μg/mL was calculated from single scFv titers and from predicted scFv dual and triple combination titers for the IC₅₀ and IC₈₀. Colors indicate percentage neutralization with 1-19% in green, 20-49% yellow, 50-79% yellow-orange, 80-89% orange, 90-99% red and 100% dark red. Geometric mean at 50μg/mL is also given. Each scFv is present in the combinations at the titers indicated.

Figure 7

Predicted coverage and estimated efficacy of best scFv combinations. (A, B) Active coverage by one, two, three or four scFv was calculated at 2 concentrations (1μg/mL (A) and 10μg/mL (B) at IC₈₀ based on the predicted titers according to the Bliss-Hill Independence model. The broadest triple and quadruple combinations are shown. (C, D) Breadth Potency curves of active coverage of 1 and 2 scFv are given for the best combinations (3BNC17+10E8v4 and 3BNC17+10E8v4+PGT121, and CAP256.25+3BNC17+10E8v4+PGT121) at 1 μg/mL (C) and 10μg/mL (D). 50% and 90% breadth are indicated by the dotted lines in (C, D).