Ebola Virus Glycoprotein with Increased Infectivity Dominated the 2013-2016 Epidemic

Abstract

The magnitude of the 2013-2016 Ebola virus disease (EVD) epidemic enabled an unprecedented number of viral mutations to occur over successive human-to-human transmission events, increasing the probability that adaptation to the human host occurred during the outbreak. We investigated one nonsynonymous mutation, Ebola virus (EBOV) glycoprotein (GP) mutant A82V, for its effect on viral infectivity. This mutation, located at the NPC1-binding site on EBOV GP, occurred early in the 2013-2016 outbreak and rose to high frequency. We found that GP-A82V had heightened ability to infect primate cells, including human dendritic cells. The increased infectivity was restricted to cells that have primate-specific NPC1 sequences at the EBOV interface, suggesting that this mutation was indeed an adaptation to the human host. GP-A82V was associated with increased mortality, consistent with the hypothesis that the heightened intrinsic infectivity of GP-A82V contributed to disease severity during the EVD epidemic.

Keywords: Ebola virus; Filovirus; NPC1; RNA virus; adaptation; epidemic; glycoprotein; infection; mutation; outbreak.

Graphical abstract

Figure 1

Evolution of GP-A82V Coincides with Dramatic Increase in EBOV Infections (A) Maximum likelihood phylogeny of 1,489 full-length Makona EBOV sequences. Branches are color coded according to country of sampling. Bars to the right side indicate GP sequence and arrowheads indicate when individual residue changes occurred. Scale bar indicates nucleotide substitutions/site. (B–E) Plots of the temporal sampling of EBOV GP genotypes during 2014 (left axis) and the cumulative number of EVD cases (right axis). Shown are data for all Mano River Union (MRU) countries (B), Guinea (C), Sierra Leone (D), and Liberia (E). See also Tables S1 and S2 and Data S1.

Figure 2

GP-A82V Enhances EBOV Infectivity in Human Cells (A–C) Lentiviral virions bearing a GFP transgene and pseudotyped with ancestral EBOV Makona GP or the indicated GP variants were produced by transfection of HEK293 cells and used to transduce U2OS cells (A), HEK293 cells (B), or MDDCs (C). GFP-positive cells resulting from transduction with variant GPs were quantified by flow cytometry and normalized to the ancestral GP using lentivirions produced in parallel. Shown are means ± SEM. (D) Western blots (top) of enriched lentiviral particles pseudotyped with C-terminally V5-tagged EBOV GPs probed with anti-V5 and anti-p24 antibodies. Bar graph showing EBOV GP1 + 2 signal intensity relative to that observed for the corresponding lentiviral capsid (p24). In (A) and (B), each data point represents a normalized transduction using lentiviral stocks derived from independent transfections. In (C), data points represent independent experiments with four independent viral stocks and eight different human donors. ^∗p < 0.05; ^∗∗p < 0.001; ^∗∗∗p < 0.001; repeated-measures ANOVA with Dunnett’s post-test comparing to ancestral EBOV GP.

Figure 3

EBOV GP-A82V Enhances the Ability of EBOV Virion Cores to Fuse with Target Cell Cytoplasm Virus-like particles (VLPs) were generated with EBOV VP40-β-lactamase fusion protein and either the ancestral EBOV GP or one of the indicated mutants. U2OS cells were incubated with VLP-containing supernatant for 2 hr at 4°C, then for 2 hr at 37°C, and then loaded with CCF4-AM overnight at 11°C. Cleavage of CCF4-AM was measured using a fluorescent plate reader with 400/30 excitation and 460/40 emission filters. Signal for cleaved CCF4-AM signal for all EBOV GPs was compared to that observed with the ancestral EBOV GP. Data are means ± SEM (n = 6 viral infections) from a representative experiment. ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; N.S. p > 0.05; one-way ANOVA with Burnett’s post-test comparing with ancestral EBOV GP.

Figure 4

GP-A82V Substitution Specifically Enhances Infectivity in Primate Cells (A) Plot of the amino acid conservation in mammals across the region of NPC1 that interacts with EBOV GP. Shaded regions indicate EBOV-interacting residues of NPC1. (B) Alignment of NPC1 sequences in and around the second NPC1 interacting loop from a subset of mammalian species used for (A). (C) Relative infectivity data for EBOV GP containing the A82V substitution in relation to the ancestral GP in four primate cell lines and five cell lines from other mammalian species. Data are means ± SEM (n = 3), with data points representing infections using independent viral stocks. ^∗∗p < 0.01; repeated-measures ANOVA with Dunnett’s post-test comparing to ancestral EBOV GP. See also Table S3.

Figure 5

Comparative Structural Analysis of the EBOV GP-A82V Substitution and NPC1 Orthologs (A) Structural overview of EBOV GP2 (yellow) and cleaved GP (GPcl, light blue) bound to human NPC1 (orange), as experimentally derived by Wang et al. (2016). The EBOV GP-A82V substitution (red) occurs in the α1 helix of proteolytically cleaved GP that interacts with NPC1 loop 2. (B and C) Zoom of the (B) ancestral GP-A82 (blue) and (C) derived GP-V82 (red) variants in relation to proximal amino acid side chains. The distance between R85 and E178 is close enough (i.e., <4 Å) where a salt bridge is possible. (D and E) NPC1 loop 2 amino acid (D) sequence alignments and (E) target-template homology models of cell species used in this study (rodents, green; other mammals, cyan; Figure 4) and bats (purple) compared to primates (orange). Highlighted in (C) are the amino acid differences in loop 2 from the target species compared to the template, humans, and their location compared to the GP-A82V variant.

Figure 6

Association between GP-A82V and Increased Rate of Lethal Infection (A and B) Spatial distribution of GP genotypes for all available EBOV Makona sequencing data (A) or Guinean isolates linked to information regarding clinical outcome and viral load (B). The data included in (B) were used in subsequent modeling analyses. (C) Association between patient viral load (as determined by C(t) values) and EVD-associated mortality. This analysis used all observations for which C(t) values were measured (n = 313). (D) Viral load information (as determined by C(t) values) in individuals infected with EBOV encoding either ancestral or A82V GP. This analysis used all observations for which C(t) values were measured (A82: n = 97; V82: n = 216). Plots in (C) and (D) show mean with the box covering from the second to third quartile (25%–75%) of samples, and the bars marking the 5% and 95% quantiles. Dots represent samples outside of the 95% probability region. (E) Mortality data in individuals infected with EBOV encoding either ancestral or A82V GP. (F) Depiction of the correlation between GP genotype and mortality, based on results of a binomial generalized linear model using C(t) values and GP genotype as covariates to predict case fatality rates over a range of viral loads (depicted by transformed C(t) values). C(t) values were transformed by subtracting the mean, dividing by two standard deviations and flipping the sign such that the value 0 in the graph corresponds to the average C(t) value and the transformed variable reflects viral load. See also Figure S1.

Figure S1

Varying-Intercept Model for Fatality Rates, Related to Figure 6 We fit a model that allowed each location to have its own base fatality rate (i.e., its own intercept) and also its own slope for the cumulative cases (cumCases) variable, in order to account for heterogeneity in how locations responded to demand for healthcare. We find that the variability in the data and possibly the relatively small number of observations lead to very broad estimates of the model parameters. While there seem to be differences in base fatality rate between Forecariah and Nzerekore for example, the overlap of confidence intervals precludes definitive conclusions.