Functional Characterization of Adaptive Mutations during the West African Ebola Virus Outbreak

ABSTRACT

The Ebola virus (EBOV) outbreak in West Africa started in December 2013, claimed more than 11,000 lives, threatened to destabilize a whole region, and showed how easily health crises can turn into humanitarian disasters. EBOV genomic sequences of the West African outbreak revealed nonsynonymous mutations, which induced considerable public attention, but their role in virus spread and disease remains obscure. In this study, we investigated the functional significance of three nonsynonymous mutations that emerged early during the West African EBOV outbreak. Almost 90% of more than 1,000 EBOV genomes sequenced during the outbreak carried the signature of three mutations: a D759G substitution in the active center of the L polymerase, an A82V substitution in the receptor binding domain of surface glycoprotein GP, and an R111C substitution in the self-assembly domain of RNA-encapsidating nucleoprotein NP. Using a newly developed virus-like particle system and reverse genetics, we found that the mutations have an impact on the functions of the respective viral proteins and on the growth of recombinant EBOVs. The mutation in L increased viral transcription and replication, whereas the mutation in NP decreased viral transcription and replication. The mutation in the receptor binding domain of the glycoprotein GP improved the efficiency of GP-mediated viral entry into target cells. Recombinant EBOVs with combinations of the three mutations showed a growth advantage over the prototype isolate Makona C7 lacking the mutations. This study showed that virus variants with improved fitness emerged early during the West African EBOV outbreak.

IMPORTANCE The dimension of the Ebola virus outbreak in West Africa was unprecedented. Amino acid substitutions in the viral L polymerase, surface glycoprotein GP, and nucleocapsid protein NP emerged, were fixed early in the outbreak, and were found in almost 90% of the sequences. Here we showed that these mutations affected the functional activity of viral proteins and improved viral growth in cell culture. Our results demonstrate emergence of adaptive changes in the Ebola virus genome during virus circulation in humans and prompt further studies on the potential role of these changes in virus transmissibility and pathogenicity.

INTRODUCTION

Ebola virus (EBOV) belongs to the family Filoviridae in the order Mononegavirales. The Ebola virus disease is characterized by severe fever accompanied by systemic inflammation and damage to the endothelial cell barrier leading to shock and multiorgan failure with high case/fatality rates (1). EBOV particles are filamentous (1 μm in length, 80 nm in diameter) and composed of seven viral proteins. The nucleocapsid complex contains the viral RNA encapsidated by nucleoprotein NP and four additional viral proteins. The polymerase complex of VP35 and L is associated with NP via interaction with VP35, and viral proteins VP24 and VP30 complete the mature nucleocapsid (2). A shell of matrix protein VP40 surrounds the nucleocapsids, and the whole particle is enveloped by a lipid membrane into which surface glycoprotein GP is incorporated. GP recognizes the cellular receptor and mediates fusion.

The EBOV variant which caused the outbreak in West Africa was called Makona for a river at the border of Guinea and Sierra Leone (3). The West African outbreak, most likely caused by a single zoonotic introduction event, started in December 2013, and over 28,000 cases led to more than 11,000 deaths (4, 5). Next-generation sequencing (NGS) of patient isolates permitted an early assessment of EBOV molecular epidemiology (6–13). Importantly, these NGS analyses showed the emergence of a number of nonsynonymous mutations in the EBOV genome. Because mutations in the EBOV genome had not been consistently observed in previous EBOV outbreaks due to the relatively short chains of transmission at that time (14), this result induced considerable public attention. In particular, it seemed important to understand whether substitutions in the EBOV proteins improved viral fitness. So far, only a few studies have addressed the functional significance of point mutations detected in few EBOV isolates (15, 16). In order to understand the impact of the nonsynonymous mutations that occurred frequently during the West African outbreak, we analyzed the evolution of EBOV, established an EBOV Makona-specific infectious virus-like particle (iVLP) assay, and constructed and rescued recombinant EBOV Makona variants for comparative analyses. We showed that three mutations in L, GP, and NP emerged early during the West African outbreak, were fixed rapidly, and were found in 90% of the sequenced EBOV patient isolates. The mutations improved the function of L and GP, and recombinant viruses containing these mutations showed enhanced fitness in cell culture.

RESULTS

Three point mutations were fixed in the Ebola virus genome during the West African outbreak.

While the West African EBOV outbreak started in December 2013, the first viral sequences were obtained when isolates C5, C7, and C15 from three patients in Guinea (March 2014) were analyzed (6). C7 and C5 were closer than C15 to the root of phylogenetic trees of the EBOV Makona viruses (Fig. 1A) (8, 11, 12). We therefore used isolate C7 as the reference sequence in our analysis of 1,020 sequences. Three amino acid substitutions in the EBOV proteins with respect to Makona C7 attracted our particular attention as they (i) were located in functional domains, (ii) emerged early in the outbreak, and (iii) were fixed in all descendant sequences (Fig. 1A and 2; see also Table S1 in the supplemental material) (6). The first substitution, an exchange from aspartate to glycine, was located at amino acid position 759 of the L polymerase (L_D759G) in close vicinity to the highly conserved GDN motif of the enzymatically active center (amino acid 741 to 743). Substitution from alanine to valine at amino acid position 82 of GP (GP_A82V) took place in the receptor binding domain (17, 18). This position was found by Ladner et al. to be under positive selective pressure during the West African outbreak (10). The third substitution, arginine to cysteine, occurred at amino acid position 111 of NP (NP_R111C) in a multifunctional region of NP important for homo-oligomerization, the formation of nucleocapsids, genome replication, the interaction with VP40, and the incorporation of nucleocapsids into VLPs (19, 20). These three specific amino acid substitutions had not been detected in any published EBOV sequence before the West African outbreak (Fig. 1B; see also Table S2).

FIG 1.

Ebola virus sequence variation at amino acid positions NP₁₁₁, GP₈₂, and L₇₅₉. (A) Phylogenetic tree of the EBOV sequences obtained during the West African outbreak. The sequences are labeled with the country of sampling, the date of sampling, and the GenBank accession number. Sequences of Makona isolates C5, C7, and C15 are shown in purple font. Amino acid signatures at positions NP₁₁₁, GP₈₂, and L₇₅₉ are depicted next to the sequence name. Black circles, NP_111R, GP_82A, and L_759D (RAD = C07 signature); blue circles, NP_111R, GP_82A, and L_D759G (RAG = point mutation in L); red circles, NP_111R, GP_A82V, and L_D759G (RVG = point mutations in L and GP). The branch containing 910 sequences with the triple mutation NP_R111C, GP_A82V, and L_D759G (CVG) is collapsed for clarity (green). The scale bar represents 0.0001 units of nucleotide substitutions per site. Percentages of replicate trees in which the associated taxa clustered together in more than 50% of 1,000 bootstrap replicates are shown next to the branches. (B) Sequence variations at positions NP₁₁₁, GP₈₂, and L₇₅₉ in all sequenced Ebola virus genes available in the database (from 1976 to 2013) before the West African outbreak. aa, amino acids.

FIG 2.

Emergence and fixation of three amino acid mutations at positions NP₁₁₁, GP₈₂, and L₇₅₉ of Ebola Makona virus during the West African outbreak. (A) Vertical red bars in the schematic of the EBOV genome illustrate the location of the three amino acid mutations in NP, GP, and L open reading frames. Row 1, amino acid mutations at NP₁₁₁, GP₈₂, and L₇₅₉; row 2, incidence of the three different mutations in 1,011 full viral genome sequences from EBOV cases that occurred between March 2014 and October 2015; row 3, date of the first appearance of mutations at positions NP₁₁₁, GP₈₂, and L₇₅₉ and the respective GenBank accession numbers (in parentheses). (B) Chronological appearance of EBOV Makona mutants carrying mutations at positions NP₁₁₁, GP₈₂, and L₇₅₉. From March until June, the majority of sequences had the signature of the prototype Makona C7 (69%). Sequences with the single mutation L_759D were reported for a short time period (until April 2014) (see Table S1). Until end of May 2014, the double mutant GP_A82V and L_D75 and the triple mutant NP_R111C, GP_A82V, and L_D759G coexisted before the triple mutation became the dominant signature in the following months until the end of the outbreak.

The three mutations accumulated sequentially in the order L_D759G > GP_A82V > NP_R111C (Fig. 1A and 2a; see also Table S1). L_D759G was first detected in sequences from samples collected in March 2014 in Guinea (Fig. 2B). This mutation was then supplemented by the GP_A82V substitution. The double mutant L+GP was first detected at the end of March in Guinea and was later replaced by a triple mutant containing the additional substitution NP_R111C (L+GP+NP). The triple signature was first detected in Sierra Leone (end of May 2014) and was detected in more than 90% of the viruses isolated after June 2014 (Fig. 2B).

The mutations GP_A82V, L_D759G, and NP_R111C influence viral transcription and replication.

To characterize potential phenotypic effects of the mutations L_D759G, GP_A82V, and NP_R111C, we established and used an EBOV Makona-specific virus-like particle (VLP) assay (21, 22). This assay permits investigation of the function of individual EBOV proteins in replication and transcription of an EBOV-specific minigenome as well as infectivity of released virus-like particles (VLPs) (Fig. 3A). Briefly, all EBOV proteins were expressed intracellularly together with an EBOV Makona-specific minigenome serving as the template for transcription and replication by the viral polymerase (Fig. 3A, producer cells). The minigenome encodes Renilla luciferase, and its enzymatic activity allows monitoring of viral transcription and replication. Replicated minigenomes are encapsidated by the viral nucleocapsid proteins forming mininucleocapsids, which serve as additional transcription templates and are enveloped and released as VLPs containing the mininucleocapsids, VP40, and GP. The VLPs are used to infect new target cells (Fig. 3A, naive indicator cells), in which the minigenome is transcribed by VLP-associated viral proteins, allowing assessment of their transcriptional activity. If the indicator cells expressed the EBOV nucleocapsid proteins in trans (Fig. 3A, pretransfected indicator cells), VLP infection allowed monitoring of the efficiency of both transcription and replication.

FIG 3.

Functional analyses of the three mutations NP₁₁₁, GP₈₂, and L₇₅₉ using reporter assays. (A) A schematic of the virus-like particle (VLP) assay used. HEK293 cells (producer cells) were transfected with plasmids expressing all of the viral structural proteins and a Makona-specific minigenome encoding Renilla luciferase. The minigenome is encapsidated, transcribed, and replicated by the viral nucleocapsid proteins. Viral replication and transcription are monitored by measuring Renilla luciferase activity. Replication results in the formation of mininucleocapsids which are released from cells in the form of virus-like particles (VLPs). VLPs are used to infect either naive Huh7 cells (naive indicator cells) or Huh7 cells pretransfected with plasmids encoding the EBOV nucleocapsid proteins (pretransfected indicator cells). Naive indicator cells support only primary transcription of the incoming mininucleocapsids, whereas pretransfected indicator cells support both transcription and replication of the incoming mininucleocapsids. Luc, luciferase activity; MNCs, mininucleocapsids; MG, minigenome. (B) Functional analyses of mutant EBOV proteins in a VLP assay. The VLP assay was performed by using plasmids encoding the Makona C7 structural proteins (column 1), Makona C7 plasmids without the L plasmid (column 2, negative control), and Makona C7 structural proteins with single or combinations of plasmids replaced by mutant plasmids (columns 3 to 7). The reporter gene activity was measured (using Renilla luciferase). Values were normalized to Makona C7 (value set as 1). Error bars indicate standard deviations of the results of at least seven independent experiments. P values: *, ≤0.05; **, ≤0.01. (C) Expression control of viral proteins in the cell lysates and in VLPs. Cell lysates and VLPs from the supernatants were subjected to Western blot analysis. The viral proteins were stained with monospecific antibodies against NP, VP40, and GP. Numbering of the lanes is identical to that described for panel B. GP_ER, G protein-coupled estrogen receptor.

We conducted the VLP assays using L, GP, or NP of Makona C7 and their mutated counterparts (Fig. 3B). The individual mutant proteins L_D759G and GP_A82V increased the levels of the reporter gene signals in the producer cells (Fig. 3B, left panel, columns 4 and 5), whereas the replacement of NP by NP_R111C decreased the reporter gene activity (Fig. 3B, left panel, column 3). When we replaced two or three of the Makona C7 proteins (L and GP or L, GP, and NP) by the respective mutants, the reporter activity was also higher than that seen with the wild-type proteins (Fig. 3B, left panel, columns 6 and 7). This pattern became more distinct in the cells infected with VLPs generated in the presence of individual mutant proteins (Fig. 3B, middle and right panels). VLPs containing either GP_A82V or L_D759G induced increased levels of reporter signals either 3.4-fold or 2.0-fold, respectively (Fig. 3B, middle panel, columns 4 and 5). The presence of NP_R111C decreased the reporter gene activity in comparison with that seen with Makona C7 NP (0.5-fold) (Fig. 3B, middle panel, column 3). The simultaneous exchange of GP and L resulted in VLPs with 2.7-fold-enhanced levels of reporter gene (Fig. 3B, middle panel, column 7). Infection with VLPs formed in the presence of three mutant proteins (NP, GP, and L) resulted in levels of reporter signals which were increased by 1.5-fold and 1.7-fold in comparison with Makona C7 levels (Fig. 3B, middle and right panels, column 6). NP, GP, and VP40 protein levels in producer cells and VLPs were unaltered (Fig. 3C).

Transcription and replication of EBOV are mediated by the nucleocapsid proteins NP, VP35, and VP30 and the L polymerase. It is therefore conceivable that mutations in the functionally important domains of the nucleocapsid components affected transcription and replication directly. However, it was surprising that a mutation in GP affected the transcription/replication signal in the producer cells as well (Fig. 3B and 4). To analyze whether the presence of GP has an influence in producer cells, we determined reporter gene activity in producer cells in the presence or in the absence of GP. Reporter gene activity was set to 100% in the presence of GP. When GP was omitted, the luciferase signal was reduced to 40% (Fig. 4A). This result indicated that GP-decorated VLPs released from producer cells are able to infect fresh target cells within the same cell culture, leading to increased reporter activity. This implied that GP_A82V mediates entry into target cells more efficiently than GP. To investigate this hypothesis, a VLP-based entry assay was employed (23). The assay measures how efficiently VLPs enter target cells by monitoring the release of the mininucleocapsids from the endosome/lysosome into the cytoplasm of the infected cell. We observed that GP_A82V mediated the entry of VLPs into target cells two times more efficiently than did GP (Fig. 4B). Analysis of the crystal structure of GP in complex with its receptor Niemann-Pick C1 (NPC1) (24) shows that amino acid 82 is located exactly in the interface between the two molecules (Fig. 4C). Thus, mutations at this position could have a direct effect on the receptor binding activity of GP and by these means could influence viral cell entry.

FIG 4.

Functional analysis of GP- and GP_A82V-mediated entry of VLPs. (A) Influence of GP on reporter gene activity in a VLP setting in producer cells as described in the Fig. 3A legend. Makona C7 structural proteins were expressed in Huh7 cells either with GP (+ GP) or without GP (Ø GP). Reporter gene activity was measured and normalized (presence of GP = 100%). (B) GP- and GP_A82V-mediated entry of VLPs. EBOV VLPs containing a VP30/Renilla luciferase fusion protein carrying either Makona GP or Makona GP_A82V were produced and purified. HEK293 cells were transduced with the purified VLPs, and the entry was monitored using the appearance of the VP30/Renilla luciferase protein in the cytosol of the transduced cells as the readout of infection (23). The reporter gene activity of cytosolic nucleocapsid-associated VP30/Renilla luciferase in the transduced cells was measured. (A and B) The error bars represent the standard deviations of the results of three independent experiments. P values: **, ≤0.01; ***, ≤0.001. (C) Location of the A82V mutation in the receptor binding domain of EBOV GP. The model is based on the X-ray structure of the GP complexed with its endosomal receptor NPC1 (Protein Data Bank [PDB] identifier 5F18). The interacting parts of GP1 and NPC1 are shown in green and purple, respectively; GP2 is shown in cyan. The red ball depicts the C-beta atom of alanine in position 82. Amino acid residues located within 4 Å of residue 82 are highlighted in yellow.

Recombinant mutant Ebola viruses show a growth advantage in cell culture.

On the basis of the results of the VLP assays, we employed reverse genetics to construct and rescue recombinant EBOV Makona C7 virus and two mutant viruses, namely, the double mutant containing GP_A82V and L_D759G (EBOV_GP+L) and a triple mutant containing NP_R111C, GP_A82V, and L_D759G (EBOV_NP+GP+L). Analyses of the supernatants of infected cells by SDS-PAGE and following Western blotting or silver staining confirmed that all viruses had been rescued successfully (Fig. 5A and B). The growth kinetics of recombinant Makona C7 and the two mutant viruses was studied in cells of monkey (VeroE6) and human (Huh7) origin (Fig. 5C). EBOV_GP+L and EBOV_NP+GP+L showed accelerated growth in comparison with Makona C7 in both cell lines (Fig. 5C). To corroborate these results, we compared the growth rates of the recombinant Makona C7 virus and its mutants in a pairwise competition assay (Fig. 5D) (25). Mixtures containing either equivalent or excess amounts of Makona C7 and each mutant virus were used to infect Huh7 cells. After day 2 and day 3 postinfection (p.i.) or after passaging the supernatant of the infected cells to fresh cells, at day 3 p.i., samples of the cell supernatants were harvested, and the viral genomic RNA was subjected to reverse transcription (RT) and Sanger sequencing. Sequence chromatographs showed that Makona C7 was outcompeted by EBOV_GP+L and EBOV_GP+L+NP 3 days after infection and even with excess Makona C7 after two to three passages, confirming that replication of the mutants was more efficient than that of EBOV Makona C7 (Fig. 5D).

FIG 5.

Characterization of rEBOV_GP+L and rEBOV_NP+GP+L. Mutations GP_A82V and L_D759G (rEBOV_GP+L) or NP_R111C, GP_A82V, and L_D759G (rEBOV_NP+GP+L) were introduced into the genome of EBOV Makona by reverse genetics, and recombinant EBOVs were rescued. (A and B) Purified viruses were subjected to SDS-PAGE followed by Western blotting (A) and silver staining (B). Western blots were developed with chicken anti-VP40 and anti-GP antibodies and secondary IRDye680-conjugated anti-chicken antibodies. ⭑, nonviral protein band (albumin) copurified with the virions. (C) rEBOV Makona C7, rEBOV_GP+L, or rEBOV_NP+GP+L was used to infectVeroE6 or Huh7 cells. Samples from the supernatants were harvested at the indicated times and used to perform TCID₅₀ assays. Error bars indicate standard deviations. d p.i., day postinfection. P values: *, ≤0.05; **, ≤0.01. (D) For the experiments whose results are depicted in the upper part of the panel, equal amounts (1:1) of PFU of rEBOV Makona and either rEBOV_GP+L or rEBOV_NP+GP+L were mixed and used to infect Huh7 cells. Samples were removed from the supernatants at day 2 (d2) and day 3 (d3) p.i., and viral RNA was purified. RT-PCR was performed, amplifying fragments that covered GP position 82, and amplicons were sequenced by Sanger sequencing. For the experiments whose results are depicted in the lower part of the panel, Huh7 cells were infected with a 9:1 mixture of rEBOV Makona and either rEBOV_GP+L or rEBOV_NP+GP+L as described above. Supernatants of infected cells were passaged at day 3 p.i. (p3) onto fresh Huh7 cells. Viral RNA was harvested and sequenced at day 3 p.i. of each passage.

DISCUSSION

The mutations analyzed in this study facilitated viral replication in cell culture, suggesting that they increase the fitness of the virus in human cells. To our knowledge, this is the first evidence of adaptive evolution of Ebola virus after its zoonotic transmission to humans.

Our VLP-based analyses indicated that the L_D759G mutation enhances viral replication/transcription and that the GP_A82V mutation increases the efficiency of virus entry into target cells. The combination of the two effects could explain the success of viruses containing these two mutations. A bioinformatics analysis suggested that the GP_A82V mutation was under positive selection (10), which was clearly supported by our data. The mutation A82V in GP is located in the receptor binding region (17, 18). The importance of this particular region of GP for entry is indicated by the fact that the mutation of F at position 88 to A is deleterious for entry of VLPs in human cell lines (26). Also, additional hydrophobic residues (L111, I113, L122, and F225) located in the three-dimensional (3D) model of EBOV GP in close proximity to F88 have been described to be important for interaction with the cellular receptor. A structure of the complex of GP with its endosomal receptor Niemann-Pick C1 (NPC1) indicates that two protruding loops of NPC1 interact with a hydrophobic cavity on the surface of GP (27). This cavity is formed by several amino acid residues, among them V79, P80, T83, W86, G87, and F88. Modeling of the GP_A82V mutation suggests that it would affect the fine structure of the cavity and thus the interaction between GP and NPC1 (Fig. 4C) and thus supports our data on the effects of the A82V mutation on efficiency of cell entry.

The D759G mutation in the L polymerase is located in close proximity to the highly conserved ₇₄₀MGDNQ motif inside the catalytically active site of filovirus polymerases (28, 29). Indeed, the central “GDN” motif is conserved in RNA-dependent RNA polymerases of all mononegaviruses (30). Interestingly, although most of the previously studied mutations in this region have negative effects on polymerase activity, D759G enhanced transcription/replication (Fig. 3B). Since the enhancing effect was also detectable in naive indicator cells, where only primary transcription is supported, it is likely that this mutation affects both transcription and replication. Mutations in the flanking regions of the MGDNQ motif of Marburg virus have been shown to play a role in the adaptation of Marburg virus to guinea pigs (31, 32). Although the crystal structure of the L protein of EBOV remains to be determined, the structure of the highly homologous L protein of VSV was recently resolved (28). On the basis of this structure, several potential phenotypic effects of the D759G mutation in EBOV L can be predicted, namely, allosteric effects on the structure of the catalytic site of the RNA-dependent RNA polymerase and alteration of the atomic interactions of L protein with the RNA and/or cellular factors.

The biological significance, if any, of the fixed substitution NP_R111C in the NP remains obscure and deserves further studies. We do not see obvious advantages of this mutation since GP_A82V and L_D759G were sufficient for improved fitness, and it remains to be determined whether the NP_R111C mutation was adaptive and if so, what the mechanism was.

Presently, it is unclear whether the investigated mutations enhanced the pathogenicity of EBOV Makona for humans. Preliminary data from epidemiological analyses related to a large NGS project seem to suggest that the mortality is not altered by the described mutations, since the overall case fatality rate was not significantly changed during the course of the outbreak (7). However, in this particular study, only isolates from severely ill patients with an unusually high mortality rate of more than 80% were selected, whereas the numbers published by WHO suggest an average case fatality rate of approximately 40% (5). This bias might have obscured the effects of the adaptive mutations. Albariño et al. (16) compared two closely related EBOV isolates which differed by two nonsynonymous point mutations in L (G882S) and VP30 (W282stop). The viruses were isolated from two Ebola virus disease patients. One of the patients underwent a fatal disease, while the other patient survived. Using a minigenome system and recombinant viruses, it was shown that the detected point mutations in L and VP30 slightly enhanced viral fitness in cell culture. Whether the mutations were decisive for the outcome of the infections is enigmatic.

A retrospective analysis of a hospitalized EBOV-positive patient cohort in Conakry, Guinea, revealed that, starting from August 2014, the mean viremia in these patients increased by an order of magnitude which was accompanied by an 14% increase in the case fatality ratio (33). These changes coincided with the emergence of the triple mutant at the end of July 2014 in Guinea after having been established in Sierra Leone (see Table S1 in the supplemental material). Thus, the study by Faye et al. seems to agree with our data on the improved fitness of the triple mutant in human cells. However, because a complex pattern of parameters, which includes infection dose, comorbidities, and quality of treatment, influences the mortality of Ebola virus disease, it is currently not possible to correlate putative effects of the investigated mutations with pathogenicity. To address this issue, it will be necessary to test the recombinant mutant viruses in suitable animal models.

The appearance of fitness-enhancing adaptive mutations that were selected and fixed during the West African EBOV outbreak adds strong support to the argument that it is necessary to stop outbreaks of emerging infections at the earliest possible time point.

MATERIALS AND METHODS

Cell culture and viruses.

Human embryonic kidney cells (HEK293; purchased from the American Type Culture Collection [ATCC], Manassas, VA, USA), human hepatoma cells (Huh7; kindly provided by K. Conzelmann, Munich, Germany), and African green monkey kidney cells (Vero E6; from ATCC) were maintained in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal calf serum (FCS), l-glutamine, and penicillin-streptomycin (Pen/Strep) solution (Gibco, Karlsruhe, Germany).

All cell lines were regularly authenticated by PCR, light microscopy, and DNA barcoding. Mycoplasma contamination was excluded by using a MycoAlert Plus mycoplasma detection kit (Lonza, Basel, Switzerland). Recombinant viruses were propagated in Huh-7 cells. All work with infectious EBOV Makona was performed at the biosafety level 4 (BSL4) facility of the Institute of Virology, Philipps-University Marburg, according to national regulations.

EBOV Makona C7 (GenBank accession number KJ660347.2), isolated from a female Guinean patient in March 2014, was kindly provided by Stephan Günther, Bernhard-Nocht-Institute for Tropical Medicine, Hamburg, Germany.

Plasmids and mutagenesis.

Cloning of full-length EBOV Makona C7 cDNA was performed by amplification of EBOV Makona genomic viral RNA isolated from supernatants of infected cell cultures by RT-PCR. PCR primers added a T7 promoter at the 3′ end and a hepatitis δ ribozyme at the 5′ end. The 3′ and 5′ sequences of the viral genome were amplified using oligonucleotides designed on the basis of the published C7 sequence (GenBank accession number KJ660347.2). The antigenomic sequence of EBOV Makona was cloned in three parts (fragment 1 [FR1], T7 leader-NP-VP35-VP40-GP-; FR2, GP-VP30-VP24-L-; FR3, L-trailer-ribozyme) flanked by unique restriction sites into three individual plasmids carrying a kanamycin resistance gene (pKan). Assembly of the full-length plasmid containing the whole antigenome of EBOV Makona was performed by standard ligation of the three DNA fragments into a minimal vector carrying an ampicillin resistance gene (pAmp).

EBOV Makona C7 open reading frames (ORFs) for viral proteins NP, VP35, VP40, VP30, VP24, and L were amplified from viral RNA by reverse transcription-PCR using specific oligonucleotides with integrated restriction sites. Resulting cDNA was digested, purified, and ligated into pCAGGS vector. Sequencing revealed only one mismatch in L compared to published sequence at position 6393 of the ORF of L. This mutation from C to T was silent (TAT and TAC = tyrosine). The cloning of GP was described elsewhere (34).

Multisite-directed mutagenesis was used according to the instructions of the manufacturer (Agilent) to introduce mutations into the NP gene, GP gene, or L gene. Mutagenesis was performed in pCAGGS, pKanFr1, pKanFr2, or pKanFr3 vector. Mutation of the nucleotide at position 331 (C>T) in the NP gene resulted in an amino acid substitution at position 111 (R>C), mutation of the nucleotide at position 245 (C>T) resulted in an amino acid substitution at position 82 (A>V) in the GP gene, and mutation of the nucleotide at position 2276 (A>G) in the L gene resulted in an amino acid substitution at position 759 (D>G). Mutated plasmids were combined to obtain two or three mutations per single genome. Introduced mutations were confirmed by sequencing.

To obtain the Makona C7 minigenome, the leader sequence was amplified from pKanFr1 via PCR. The Renilla luciferase gene flanked by a NotI restriction site and a part of the leader sequence conserved between Mayinga and Makona as well as the fragment containing the hepatitis δ ribozyme flanked by a homologous part of the leader and a SacI restriction site were amplified via PCR from the ZEBOV Mayinga minigenome 3E-5E (35, 36). The Renilla luciferase gene, the leader, and the ribozyme were combined via overlap extension PCR to obtain a product flanked by NotI and SacI restriction sites (fragment A). The trailer sequence flanked by XmaI and NotI restriction sites was amplified via PCR from pKanFr3 (fragment B). Fragments A and B were combined via the NotI restriction site and cloned into the target vector via the XmaI and SacI restriction sites. A T7 promoter was inserted via Q5 site-directed mutagenesis (NEB).

Primer sequences and detailed cloning strategies are available upon request.

Virus-like particle (VLP) assay.

The VLP assay was performed as described earlier (21, 22). Briefly, HEK293 cells (producer cells) were transfected with plasmids encoding Makona C7 proteins (or their respective mutants as indicated), the Makona C7 minigenome, pCAGGS-T7, and pGL4 as a reference for transfection efficiency. Three days later, cells were lysed and the luciferase activity was determined. Released VLPs were used to infect naive Huh7 cells or Huh7 cells pretransfected with plasmids coding for Makona C7 NP, VP30, L, or the respective mutants. Three days later, Huh7 cells were lysed and used for a luciferase assay. Statistical analysis was performed in GraphPad PRISM using an unpaired t test.

For Western blot analysis, aliquots of the cell lysates and VLP suspensions were separated by SDS-PAGE, and the proteins were blotted to nitrocellulose membranes. For detection of NP, a specific chicken anti-NP antibody (kindly provided by R. Schade, Institut für Pharmakologie, Charité-Universitätsmedizin Berlin [37]) was used. For detection of VP40, a specific mouse monoclonal anti-VP40 antibody was used (38). For detection of GP, a specific chicken anti-GP antibody was used (38). Anti-chicken and anti-mouse IRDye 680- or 800-conjugated antibodies from goat were used as secondary antibodies. Labeled proteins were detected using an Odyssey Imager (Li-COR).

Entry assay.

Purification of filamentous VLPs from cell supernatants and entry assays were carried out as described previously (23, 39). In brief, HEK293 cells were transfected with 1 μg pCAGGS-Mayinga-NP, 1 μg pCAGGS-Mayinga-VP35, 8.3 μg pCAGGS-Mayinga-L, 2 μg pCAGGS-Mayinga-VP40, 12.5 μg pCAGGS-Mayinga-VP30-Luc, and either 2 μg pCAGGS-Makona C7-GP or 2 μg pCAGGS-Makona GP_A82V per T75 flask using TransIT-LT1 and were incubated for 72 h. Then, filamentous VLPs were harvested by ultracentrifugation as described previously (39). The final pellets of filamentous VLPs were used to perform an entry assay as described in reference 23. Briefly, HEK293 cells were incubated with purified filamentous iVLPs to allow GP-mediated entry of nucleocapsid-associated VP30-Luc into the cytoplasm. The luciferase activity of cytoplasmic VP30-Luc was determined at the indicated time points. For statistical analysis, an unpaired t test was performed.

Recovery of recombinant viruses.

Huh7 cells (2 × 10⁵) were transfected with an EBOV full-length genome plasmid (1 μg), a T7 polymerase expressing construct (250 ng), and plasmids coding for EBOV nucleocapsid proteins NP (125 ng), VP35 (125 ng), VP30 (100 ng), and L (1 μg) under BSL4 conditions. Transfection procedure was carried out by using TransIT-LT1 (Mirus Bio LCC) at a ratio to the DNA of 3 to 1 in 2 ml Opti-MEM according to the manufacturer's instructions. Medium was changed at 3 to 6 h posttransfection (p.t.) to DMEM supplemented with 3% FCS, Pen/Strep, and glutamine. At 5 to 7 days p.t., a blind passage of the supernatant of transfected cells was performed in 2 × 10⁵ Huh7 cells. Passage 1 cells were monitored until cytopathic effect (CPE) was detected. In general, CPE was visible 3 to 5 days after inoculation, and the supernatants were harvested for analysis and virus stock production. Full-length sequencing of mutated EBOV genomes after two passages in cell culture confirmed that only the desired mutations were inserted.

Infections and competitive growth analysis.

All work with infectious EBOV was performed at the BSL4 laboratory of the Institute of Virology, Philipps-University Marburg, in compliance with national regulations.

VeroE6 and Huh7 cells were infected with recombinant EBOV at a multiplicity of infection (MOI) of 0.01. After virus adsorption for 1 h at 37°C, the inoculum was removed and cells were incubated in DMEM–3% FCS. Samples from the supernatant were taken at days 1, 2, 3, and 7 for VeroE6 cells and at days 1, 2, and 3 for Huh7 cells. Samples were clarified by centrifugation for 5 min at 5,000 rpm and titrated on VeroE6 cells. For titration, 30% confluent VeroE6 cells in 96-well plates were infected with 10-fold serial dilutions of viral suspensions. The cytopathic effect was analyzed after 10 days by light microscopy, and the 50% tissue culture infective dose (TCID₅₀) per milliliter was calculated using the Spearman-Kärber method (40). For statistical analysis, an unpaired t test was performed.

For competitive growth assays, Huh7 cells were infected with a mixture of two viruses at a total MOI of 0.01 PFU per cell. Input mixtures were prepared with a ratio of recombinant EBOV (rEBOV) C7 to mutant virus of either 1:1 or 9:1. Samples of input mixtures and samples of the supernatants of the cultures coinfected with the 1:1 mixture were taken at day 2 and day 3 postinfection (p.i.). Samples of the supernatant of the cultures coinfected with the 9:1 mixture were taken from each passage at day 3 p.i. and were used for infecting fresh Huh7 cells and downstream analysis. All samples were clarified by centrifugation for 5 min at 5,000 rpm and were used for RNA isolation and subsequent DNase digestion. To amplify the fragments covering the respective mutation in the viral glycoprotein GP or polymerase L genes, respective cDNAs were synthesized using an Omniscript RT kit (Qiagen) and amplified using Taq-PCR (Peqlab) or a Transcriptor One-Step RT-PCR kit (Roche). PCR products were sequenced by the Sanger method (Seqlab, Germany) using reverse primers. The chromatograms showing reverse complement sequences were analyzed.

Analysis of sequences.

All full-length sequences of the EBOV genomes from samples collected in 2014 and 2015 (1,020 sequences) were downloaded from GenBank through the NCBI Virus Variation Resource (41) on 5 January 2016. The sequences were aligned using the MAFFT method available through Unipro UGENE package version 1.20.0 (42). Analysis of amino acid polymorphisms in translated sequences was performed using BioEdit version 7.1.11 (43). The evolutionary history was inferred using MEGA7 (44) and the minimum-evolution method. A total of 18,729 nucleotide positions were analyzed in the final data set. All ambiguous positions were removed for each sequence pair.

All sequences were compared with that of EBOV Makona isolate C7 (GenBank accession number KJ660347.2). The amino acid sequences of NP at position 111, of GP at position 82, and of L at position 759 were checked for polymorphism, and percentages of different signatures at these positions were calculated for Fig. 2B (see Table S1 in the supplemental material). Amino acid sequences of EBOVs isolated prior to the West African outbreak of 2014 to 2015 were obtained from the NCBI Virus Variation Resource (41) and analyzed for polymorphism in NP, GP, and L as described above for Fig. 2B (see Table S2). For Fig. 2, sequences with incomplete data for positions NP₁₁₁, GP₈₂, and L₇₅₉ were not taken into account (9 sequences). In addition, sequences without an exact date were not considered for Fig. 2B (15 sequences).

Molecular modeling.

Locations of amino acid substitutions on the 3D structure of GP were analyzed with Pymol, version 1.7.4.1 (Schrödinger, LLC), using atomic coordinates of GP complexed with Niemann-Pick C1 protein (PDB 5F18) (27).