Enhanced potency of a broadly neutralizing HIV-1 antibody in vitro improves protection against lentiviral infection in vivo

Abstract

Over the past 5 years, a new generation of highly potent and broadly neutralizing HIV-1 antibodies has been identified. These antibodies can protect against lentiviral infection in nonhuman primates (NHPs), suggesting that passive antibody transfer would prevent HIV-1 transmission in humans. To increase the protective efficacy of such monoclonal antibodies, we employed next-generation sequencing, computational bioinformatics, and structure-guided design to enhance the neutralization potency and breadth of VRC01, an antibody that targets the CD4 binding site of the HIV-1 envelope. One variant, VRC07-523, was 5- to 8-fold more potent than VRC01, neutralized 96% of viruses tested, and displayed minimal autoreactivity. To compare its protective efficacy to that of VRC01 in vivo, we performed a series of simian-human immunodeficiency virus (SHIV) challenge experiments in nonhuman primates and calculated the doses of VRC07-523 and VRC01 that provide 50% protection (EC50). VRC07-523 prevented infection in NHPs at a 5-fold lower concentration than VRC01. These results suggest that increased neutralization potency in vitro correlates with improved protection against infection in vivo, documenting the improved functional efficacy of VRC07-523 and its potential clinical relevance for protecting against HIV-1 infection in humans.

Importance: In the absence of an effective HIV-1 vaccine, alternative strategies are needed to block HIV-1 transmission. Direct administration of HIV-1-neutralizing antibodies may be able to prevent HIV-1 infections in humans. This approach could be especially useful in individuals at high risk for contracting HIV-1 and could be used together with antiretroviral drugs to prevent infection. To optimize the chance of success, such antibodies can be modified to improve their potency, breadth, and in vivo half-life. Here, knowledge of the structure of a potent neutralizing antibody, VRC01, that targets the CD4-binding site of the HIV-1 envelope protein was used to engineer a next-generation antibody with 5- to 8-fold increased potency in vitro. When administered to nonhuman primates, this antibody conferred protection at a 5-fold lower concentration than the original antibody. Our studies demonstrate an important correlation between in vitro assays used to evaluate the therapeutic potential of antibodies and their in vivo effectiveness.

FIG 1

VRC07 is a more potent clonal relative of VRC01. (A) An alignment of VRC01, NIH45-46, and VRC07 heavy-chain variable regions highlights the four-amino-acid insertion in the CDR H3 (red). Additional amino acid changes from VRC01 are highlighted in green (NIH45-46) and blue (VRC07). (B) VRC07 neutralization was assessed on a panel of 179 viral strains. For comparison, b12, VRC01, and NIH45-46 also were analyzed. Potency-breadth curves show the fraction of viruses neutralized at increasing IC₅₀ cutoff values. (C) The percentage of viruses neutralized with an IC₅₀ of less than 50 μg/ml and less than 1 μg/ml are listed for VRC01, NIH45-46, and VRC07. (D) Superposition of the VRC01 and VRC07 antibody-bound gp120 structures. The VRC07 heavy chain is in purple, the VRC01 heavy chain is shown in green, and both light chains are shown in light purple. VRC01 and VRC07 adopt the same modes of gp120 recognition, while the four-amino-acid insertion (highlighted in red) in the CDR H3 of VRC07 makes additional contacts with gp120 (gray). (E) Closeup view of interactions (marked as a square box in panel D) between gp120 and the CDR H3 loop of VRC07. Hydrogen bonds and electrostatic interactions (dashed line in red) were found between gp120 and R100a, D100b, and Y100c of VRC07.

FIG 2

Structure-guided optimizations of VRC07 increase potency and breadth. (A) The structure of VRC07 (heavy chain, purple; light chain, light purple) bound to gp120 (gray; bridging sheet, blue; V5 loop, cyan). The insets show the series of mutations that were engineered into VRC07 to increase neutralization potency and breadth while maintaining minimal autoreactivity. Top left, alteration of the heavy-chain G54 to H increases binding to gp120 (gray) by mimicking the F43_CD4 binding to gp120. The crystal structure confirmed that the VRC07–G54H MAb and gp120 interaction resembles the CD4-gp120 interaction. Top right, modification of the N terminus of the light chain. Poor resolution (shown by a lack of electron density [dark blue mesh]) in the crystal structure of the N-terminal end of the VRC07 light chain suggests that this region is disordered. Modeling the N-terminal residues (yellow) reveals a potential clash with the V5 loop (cyan). Deletion of light-chain residues E1 and I2 along with a V3S mutation minimized the potential clash with the V5 loop. Bottom left, two somatically mutated residues (I37 and T93, green) were reverted to their germ line amino acids (V and A, respectively), resulting in improved potency. Bottom right, replacement of selected hydrophobic residues (highlighted in yellow and labeled) in the light chain with hydrophilic residues were designed to enhance solubility and resulted in increased potency. An additional mutation (N72T) removed the N-linked glycosylation site in the light chain. (B) Mutations included in each of the optimized VRC07 MAb are summarized. (C) Antibodies were tested against a panel of 179 HIV-1 strains. The fraction of viruses neutralized at the given IC₅₀ is shown. The optimized VRC07 variants (VRC07-501, VRC07-508, VRC07-523, and VRC07-544) were tested in an IgG1 format containing the FcRn binding site LS mutation. The constant region LS mutation has no effect on neutralization.

FIG 3

Autoreactivity of optimized VRC07 MAbs. (A) The reactivity of VRC07 variants to HEp-2 human epithelial cells was analyzed with an immunofluorescence cell staining assay. The autoreactive antibody 2F5 also was included in the analysis. Antibodies were tested at 50 μg/ml and 25 μg/ml, and scores are summarized in panel B. neg, negative. (C) MAb binding to cardiolipin, a mitochondrial membrane phospholipid, was assessed by ELISA. VRC07 variants and 2F5 were tested at 100 μg/ml and 3-fold serial dilutions. In both assays, the optimized VRC07 variants (VRC07-501, VRC07-508, VRC07-523, and VRC07-544) were tested in an IgG1 format containing the FcRn binding site LS mutation (described in the legend to Fig. 4). The constant region LS mutation does not affect autoreactivity results. OD, optical density.

FIG 4

Optimized VRC07 MAbs with the LS mutation have extended plasma half-lives. Pharmacokinetic studies were performed in male rhesus macaques. All MAbs were administered at 10 mg/kg intravenously, and plasma levels were monitored by a gp120 core (RSC3) ELISA. The number of NHPs is indicated, and means ± standard deviations are plotted. (A) The pharmacokinetics of VRC07 and VRC07-G54W are shown. (B) An FcRn binding site mutation (LS) was added to VRC07 and improved plasma half-life by ∼2-fold. (C) VRC07-508-LS and VRC07-523-LS also had increased plasma levels and extended half-lives compared to wild-type VRC07, while VRC07-501-LS and VRC07-544-LS had pharmacokinetics similar to those of VRC07. (D) Terminal (β) phase half-life was calculated through day 21 using a noncompartmental model. Mean areas under the curve and clearance rates also are shown for each MAb. All calculations were performed with WinNonlin software (Pharsight).

FIG 5

Improved protective efficacy of VRC07-523-LS compared to VRC01-LS. (A) Neutralization assays were performed with the SHIV-BaLP4 challenge stock and antibody VRC01-LS or VRC07-523-LS using TZM-bl target cells. (B) VRC07-523-LS is about 5-fold more potent than VRC01-LS against the SHIV-BaLP4 stock in vitro. (C) Rhesus macaques were administered either 0.2 mg/kg or 0.05 mg/kg of VRC07-523-LS (n = 4 each) or 0.3 mg/kg of VRC01-LS (n = 12) and challenged intrarectally with SHIV-BaLP4 5 days later. Plasma concentrations of VRC07-523-LS and VRC01-LS at the time of challenge were determined by RSC3 ELISA. Plasma MAb levels of animals that became infected are graphed with closed icons, and plasma levels of animals that were protected are graphed with open icons. (D) The probability of protection from SHIV challenge can be predicted using a probit regression model based on the plasma MAb concentration at the time of challenge. Curves for VRC01-LS (green) and VRC07-523-LS (blue) graph the calculated probability of protection (y axis) at increasing plasma MAb levels (x axis). The EC₅₀ for VRC07-523-LS was 0.47 μg/ml (90% confidence interval [CI], 0.31 to 0.96 μg/ml), and the VRC01-LS EC₅₀ was 2.54 μg/ml (90% CI, 1.86 to 2.96 μg/ml). Ninety percent confidence intervals are shown for each curve with horizontal lines, and these do not overlap.