Ebola virus disease (EVD) causes severe illness characterized by rapid onset of fever, vomiting, diarrhea and bleeding diathesis (1, 2). The challenges of a large outbreak and the failure of traditional quarantine and contact tracing measures (3, 4) to control the 2014 West Africa outbreak highlights the urgency for therapies. The success in non-human primates (NHP) of ZMapp, a cocktail of three chimeric monoclonal antibodies (mAbs) derived from immunized mice (5–7), illustrated the potential of mAb therapies against EVD, and it is currently being evaluated in human trials. To date, efforts in NHP to simplify the ZMapp regimen to contain fewer mAbs have not been successful (7). We sought to isolate mAbs from human EVD survivors, with the goal of identifying antibodies that confer clinical protection either as single or dual-combination agents.
We obtained blood from two survivors of the 1995 Kikwit EVD outbreak (8) eleven years after infection. To determine if the subjects retained circulating antibodies against Ebola virus (EBOV) glycoprotein (GP), we assessed GP-specific antibodies by ELISA (Fig. 1A) (9). The reciprocal EC90 titer for subject 1 (S1) was 2,326, greater than ten-fold higher than control sera. Moreover, serum from the more severely ill subject, S1, displayed potent virus neutralizing activity (Fig. 1B), indicating that S1 maintained serologic memory against EBOV GP more than a decade following infection and suggesting the potential to clone immunoglobulins with potent neutralizing activity from S1’s memory B cells.
Therefore, we sorted memory B-cells from S1’s peripheral blood mononuclear cells, and immortalized individual clones with Epstein-Barr virus (10). Forty clone supernatants displayed a range of GP-binding (Fig. 1C), and two, 100 and 114, expressed antibodies with markedly higher neutralizing activity than all others (Fig. 1D). A second immortalization yielded 21 clones, from which the GP-specific clones 165 and 166 were rescued (fig. S1).
mAb100, mAb114, mAb165 and mAb166 sequences were PCR-amplified and antibodies produced by transient transfection. We assessed ELISA binding to EBOV GP and observed that mAb114, mAb165 and mAb166 displayed nearly 50% higher maximal binding than KZ52, a prototypic EBOV GP-specific human mAb (11), and 25% higher than 13C6, a component of the ZMapp cocktail (6, 7) (Fig. 2A). The binding curve of mAb100 plateaus similarly to KZ52 (Fig. 2A). mAb100 and mAb114 achieved half maximal binding (EC50) at a concentration of 0.02 μg/mL, which is 7-19 fold lower than the other mAbs. mAb165 and mAb166 had binding profiles similar to each other, with EC50s of 0.38 μg/mL and 0.35 μg/mL respectively, while EBOV control mAbs KZ52 and 13C6 had EC50s of 0.33 μg/mL and 0.14 μg/mL (Fig. 2A).
We next evaluated S1 mAbs capacity for neutralization using lentiviral particles pseudotyped with EBOV GP, a BSL-2 method that has been demonstrated to recapitulate wild type EBOV (BSL-4) results (12) (Fig. 2B and fig. S2A). mAb165 and mAb166 exhibited similar half-maximal inhibition (IC50) concentrations of 1.77 and 0.86 μg/ml, respectively. mAb100 and mAb114 were the most potent, with IC50s about one-log lower (0.06 and 0.09 μg/ml, respectively) than mAb165 and mAb166. Notably, all four of the isolated mAbs inhibited 100% of the virus unlike KZ52, which consistently displayed only 80-90% maximum inhibition, and 13C6, which neutralized < 20% at 10 μg/mL. Importantly, S1 mAbs also potently neutralized the 2014 West African Makona-variant (fig. S2B). Neutralization of wild type EBOV particles by each of the isolated antibodies was confirmed by plaque reduction assay (fig. S3).
Sequence analysis revealed that S1 mAbs displayed between 85-95% and 89-97% germline identity for heavy and light chains, respectively (Fig. 2C). Analyses of germline gene usage and V(D)J recombination indicate that they originate from different B cell lineages. The role of somatic hypermutations for the two most potent antibodies, mAb100 and mAb114, were analyzed using variants that were partially or completely reverted to the unmutated common ancestors (UCAs) (Fig. 2D and fig. S4, A and B). The fully reverted mAb100 (UCA/UCA), as well as a variant with a single change from the UCA VL (UCA/gL, A89T), recognized cells expressing GP with 2-4 fold weaker binding compared with the fully matured antibody (Fig. 2Eand fig. S4, A and B). When three HCDR3 mutations (A96V/V103Y/Y114S) were introduced in the reverted germline antibody (gH/UCA), binding was comparable to the fully matured mAb100 (sH/sL) suggesting that three mutations are sufficient to confer the binding observed for the fully matured mAb100. In the case of the mAb114, the fully reverted version (UCA/UCA) demonstrated negligible binding to GP (Fig. 2F and fig. S4, C and D). Introduction of two mutations (A96V and Y108S) in the HCDR3 of mAb114 germline (gH/UCA) was sufficient to confer an increase in binding, although still not to the degree seen with the fully matured mAb. Since these mutations are located at the base of the HCDR3 loop, they likely do not make direct contact with GP and thus may have a stabilizing effect on the whole HCDR3. The fully matured light chain and the two HCDR3 mutations (gH/sL) were sufficient to confer binding equivalent to the fully matured mAb (sH/sL). Importantly, the fully mutated light chain gene (UCA/sL), can partially compensate for a lack of somatic mutation in the heavy chain (Fig. 2F and fig. S4, C and D). The presence of additional mutations on either VH or VL is required to achieve the level of the fully matured mAb114 binding. These results suggest a rapid pathway of mAb114 affinity maturation through one or two somatic mutations, which became redundant as further mutations accumulated, a finding that is reminiscent of what was recently observed for the generation of broadly neutralizing influenza antibodies (13).
Since mAb100 and mAb114 were the most potently neutralizing antibodies, they were considered optimal candidates for further evaluation in NHP. In order to assess the potential for combination therapy, we wished to first rule out antagonistic binding to GP. We found that each antibody bound to GP in the presence of the other, suggesting that they recognize distinct regions on GP (Fig. 3A) and therefore could be used as combination immunotherapy to maximize efficacy (14). To define the regions targeted by mAb100 and mAb114 we employed biolayer interferometry to assess GP binding in competition with mAbs KZ52 and 13C6, which have epitopes in the GP base and glycan cap, respectively (15, 16). We found that mAb100 competes with KZ52 for binding at the base of GP, while mAb114 recognizes at least in part the glycan cap region, as demonstrated by competition with 13C6 (Fig. 3, B and C).
Since some EBOV GP antibodies have been suggested to mediate antibody-dependent cell-mediated cytotoxicity (ADCC) (17) the ADCC activity of mAb100 and mAb114 were determined in a flow cytometric assay (Fig. 3D). We found that mAb100 and mAb114 mediated ADCC with maximal activity observed at a mAb concentration of 0.03 μg/ml. Target cell killing was mediated through Fc receptors since mAbs containing Fc LALA mutations (18) abrogated ADCC activity. Therefore, these mAbs have the potential to induce direct killing of infected cells in vivo, a key viral clearance mechanism.
The presence of potent neutralizing and ADCC activity, and the absence of cross competition, supported testing mAb100 and mAb114 in vivo protective efficacy. We challenged four rhesus macaques with a lethal dose of EBOV. One day post-challenge, the treatment group (n = 3) received an intravenous injection with a mixture of mAb100 and mAb114 at a combined dose of 50 mg/kg, and the treatment was repeated twice at 24-hour intervals. Circulating GP-specific antibody titers in mAb recipients peaked after the second injection, and reciprocal ELISA titers remained above 105 throughout the study, suggesting minimal clearance of the mAbs (Fig. 4A). The untreated macaque succumbed to EVD on day 10 with a circulating viral load exceeding 108 ge/ml (Fig. 4, B and C). In contrast, all three mAb-treated macaques survived challenge without detectable viremia. Consistent with historic controls, the untreated animal displayed hallmark indicators of EVD including thrombocytopenia and elevations in alanine transaminase and creatinine from day 6 through the time of death (Fig. 4D and figs. S5 to S8). In contrast, the treatment group remained within normal ranges for these parameters, and free of all EVD symptoms.
We next asked whether monotherapy is sufficient for protection, and focused on mAb114 since it showed higher maximal binding than mAb100. We exposed four macaques to a lethal dose of EBOV and administered 50 mg/kg of mAb114 (n = 3) to the treatment group after a one-day delay, followed by two doses at 24-hour intervals. All treated macaques survived, whereas the control animal succumbed to EVD on day 6 with a peak viral load of 1010 ge/ml (Fig. 4, E to G). In contrast to the previous experiment, transient viremia was observed in the treated animals (Fig. 4F), but remained at levels less than 0.1% of the untreated control animal. Despite transient viremia, treated animals remained free of clinical and laboratory abnormalities (Fig. 4H and figs. S9 to S12).
Since a delay in treatment is a distinct possibility in an outbreak setting, we evaluated the therapeutic potential for mAb114 when treatment was delayed until 5 days after lethal EBOV challenge. All three animals in the treatment group survived, while the control animal succumbed to EVD on day 9. Moreover, animals in the treatment group remained symptom-free and protected against thrombocytopenia, transaminitis and renal dysfunction (Fig. 4Iand figs. S13 to S16).
mAb114 has several characteristics that may contribute to protection as a monotherapy compared to KZ52 and 13C6, which were non-protective in NHPs (7, 19). First, while KZ52 and mAb114 potently neutralize EBOV in vitro, only mAb114 completely neutralizes input virus. Secondly, mAb114 does not require complement for neutralizing activity (Fig. 2B) in contrast to 13C6 (6). Based on these observations, one hypothesis is that protective monotherapy may require both potent binding and complete complement-independent neutralization. In addition, mAb114’s specific mechanism of neutralization, which targets an essential step in virus entry (20), and the observed in vitro ADCC activity may contribute to mAb114’s ability to protect against lethal EVD in macaques.
In these studies, we showed that antibodies as well as memory B cells specific to EBOV are maintained in a survivor more than a decade following infection. For the mAbs isolated from this survivor, a potential role for antibody-dependent killing is suggested by in vitro ADCC activity that in vivo may be mediated by multiple effector cells such as natural killer cells, macrophages or neutrophils. In addition, these mAbs showed potent neutralizing activity against Ebola GP variants that have evolved over a 40-year period. Together, these data demonstrate the therapeutic potential of these mAbs as dual- and/or mono-therapy and contribute to understanding the mechanisms of antibody-mediated protection against Ebola virus disease.
Materials and Methods
Figs. S1 to S16