Isolation of a Potently Neutralizing and Protective Human Monoclonal Antibody Targeting Yellow Fever Virus

Abstract

Yellow fever virus (YFV) causes sporadic outbreaks of infection in South America and sub-Saharan Africa. While live-attenuated yellow fever virus vaccines based on three substrains of 17D are considered some of the most effective vaccines in use, problems with production and distribution have created large populations of unvaccinated, vulnerable individuals in areas of endemicity. To date, specific antiviral therapeutics have not been licensed for human use against YFV or any other related flavivirus. Recent advances in monoclonal antibody (mAb) technology have allowed the identification of numerous candidate therapeutics targeting highly pathogenic viruses, including many flaviviruses. Here, we sought to identify a highly neutralizing antibody targeting the YFV envelope (E) protein as a therapeutic candidate. We used human B cell hybridoma technology to isolate mAbs from circulating memory B cells from human YFV vaccine recipients. These antibodies bound to recombinant YFV E protein and recognized at least five major antigenic sites on E. Two mAbs (designated YFV-136 and YFV-121) recognized a shared antigenic site and neutralized the YFV-17D vaccine strain in vitro. YFV-136 also potently inhibited infection by multiple wild-type YFV strains, in part, at a postattachment step in the virus replication cycle. YFV-136 showed therapeutic protection in two animal models of YFV challenge, including hamsters and immunocompromised mice engrafted with human hepatocytes. These studies define features of the antigenic landscape of the YFV E protein recognized by the human B cell response and identify a therapeutic antibody candidate that inhibits infection and disease caused by highly virulent strains of YFV. IMPORTANCE Yellow fever virus (YFV) is a mosquito-borne virus that occasionally causes outbreaks of severe infection and disease in South America and sub-Saharan Africa. There are very effective live-attenuated (weakened) yellow fever virus vaccines, but recent problems with their production and distribution have left many people in affected areas vulnerable. Here, we sought to isolate an antibody targeting the surface of the virus for possible use in the future as a biologic drug to prevent or treat YFV infection. We isolated naturally occurring antibodies from individuals who had received a YFV vaccine. We created antibodies and tested them. We found that the antibody with the most powerful antiviral activity was a beneficial treatment in two different small-animal models of human infection. These studies identified features of the virus that are recognized by the human immune system and generated a therapeutic antibody candidate that inhibits infection caused by highly virulent strains of YFV.

Keywords: monoclonal antibodies; mouse model; neutralization; vaccine; yellow fever virus.

FIG 1

ELISA binding and FRNT neutralization by human mAbs targeting YFV E protein. (A) Half-maximal effective concentrations (EC₅₀s) of antibody binding to YFV E as determined by ELISAs. Values were interpolated using a nonlinear regression model in Prism software. Data from a single experiment are shown, representative of results from three independent experiments. (B) Focus reduction neutralization test (FRNT) to assess the neutralization of YFV-17D by YFV-121 and YFV-136. Neutralization values were fit to a nonlinear regression model. Data from a single experiment are shown, representing results from at least two independent experiments.

FIG 2

Competition binding and neutralization by human mAbs targeting YFV E protein. (A) Octet biolayer interferometry competition binding of human mAbs against YFV E. Antibodies listed from top to bottom were associated with immobilized YFV E protein, with antibodies shown from left to right tested for their ability to bind in the presence of the first antibody. Binding is expressed as a percentage of residual binding, with black boxes indicating complete competition, gray boxes indicating partial competition, and white boxes indicating no competition. Antibodies were clustered based on their competition profiles and labeled A to F. (B to D) Neutralization of diverse YFV strains assessed by a focus reduction neutralization test (FRNT). Values were fit to a nonlinear regression model using Prism software. Three independent experiments were performed in technical triplicate, with data from a single representative experiment shown.

FIG 3

Epitope mapping for Fab YFV-136 by HDX-MS. (A) Representative kinetic plots for the 12 different peptides showing the effects of Fab binding by HDX. Black or green lines are for YFV E protein in the absence or presence of the mAb YFV-136, respectively. At the top of each panel are the residue numbers and charge states of the peptide. (B) Sequences and positions of each peptide. (C) Woods plots showing accumulated differences in percent deuteration (bound state − unbound state) across all time points for each analyzed peptide. The propagated error for the cumulative difference was calculated for each peptide, and 99% confidence intervals were calculated. Peptides whose differential exchange exceeds the 99% confidence interval are considered to show significant differences between the bound and unbound states and to be involved in binding. Peptides that do not show any significant differences between the bound and unbound states are in gray, whereas protected peptides are in green. The blue vertical line shows the location of the H67Y escape mutation identified in the studies shown in Fig. 4. (D) Protected peptides are shown in green on a ribbon representation of the YFV E dimer. The domains are indicated in red (DI), yellow (DII), and blue (DIII).

FIG 4

Critical residue for neutralization escape and mechanism-of-action studies for YFV-136. (A) Cell impedance measurements during the first round of YFV-17D escape mutant virus selection. Each box represents the cell impedance within a single well of a 96-well plate as a function of time. * indicates wells that exhibit a drop in cell impedance at late time points, suggesting viral escape in the presence of 5 μg/mL YFV-136. A single well marked with # was used as a control for cell culture adaptation. * and # wells were propagated once more in culture on the device and finally in 6-well culture dishes in the presence of 10 μg/mL YFV-136 (or no antibody for #) to allow viral outgrowth. Viral RNA was isolated, and the prM and E genes were amplified by RT-PCR using primers flanking the prM and E genes. These same primers, and two other primers targeting internal prM and/or E sequences, were used to sequence virus isolated from * and # wells by Sanger sequencing. These sequences were aligned in Geneious software to identify point mutations. (B) Escape mutant identified in panel A mapped onto the crystal structure of YFV E (PDB accession number 6IW5). Colors denote domain I (red), domain II (yellow), and domain III (blue). (C) Focus reduction neutralization test of YFV-136 before or after attachment of virus to host cells. Neutralization values were assessed using a nonlinear regression model in Prism software. Two independent experiments were performed in technical triplicate, with data from a single representative experiment shown.

FIG 5

Syrian golden hamster challenge studies to assess YFV-136 therapeutic efficacy. (A) Kaplan-Meier survival curves of animals (YFV-136, n = 10; DENV-2D22 control, n = 15; uninfected controls, n = 5) treated with 50 mg/kg of YFV-136 or 20 mg/kg of the isotype control 3 days after inoculation with 200 50% cell culture infectious doses (CCID₅₀) of the hamster-adapted YFV Jimenez strain. Statistical analysis was performed using a Wilcoxon log rank test. (B) Weights of YFV-infected animals treated with YFV-136 or isotype control mAb or uninoculated animals throughout the course of the study. (C) Serum virus titers (CCID₅₀) were assessed 6 days after virus inoculation. One-way analysis of variance (ANOVA) with Dunnett’s multiple-comparison posttest was used to assess statistical significance. (D) Serum alanine aminotransferase (ALT) levels 6 days after inoculation were assessed as a proxy for liver damage. One-way ANOVA with Dunnett’s multiple-comparison posttest was used to assess statistical significance; *** indicates p < 0.001, ** indicates. p < 0.01.

FIG 6

hFRG challenge studies to assess YFV-136 therapeutic efficacy. FRG mice engrafted with human hepatocytes (hFRG) were inoculated with 2 × 10⁵ focus-forming units (FFU) of wt YFV-DakH1279. Eight hours later, mice were administered a single 10-mg/kg dose of YFV-136 (n = 5) (blue points) or the isotype control (n = 5) (black points). Note that two of the control mAb-treated YFV-infected mice succumbed to infection at 4 dpi (denoted by open circles), and thus, some serum-based measurements were not available. (A and B) Weight loss showing mean values (A) and individual animal profiles (B). (C and D) Viral burdens at 4 dpi in the liver (C) and serum (D) as measured by RT-quantitative PCR (qPCR). (E to H) Serum biomarkers of hepatic injury at 4 dpi. Samples were tested for prothrombin time (E) and alanine aminotransferase (ALT) (F), total bilirubin (G), and ammonia (H) levels. For panels C to H, a Mann-Whitney test was performed (*, P < 0.05; **, P < 0.01). Bars denote median values. Dashed lines in panels C and D indicate the (lower and/or upper) limit of detection of the assay. Gray boxes indicate the reference range for each parameter in mice; for ALT, a dashed line is shown to denote the upper limit of “normal” for healthy hFRG mice at baseline. CI, confidence interval.