mRNA design, generation, and vaccination schedule
mRNA vaccines encoding the GP of MARV or RAVV were synthesized in vitro from linearized DNA templates of the mRNA by T7 polymerase-mediated transcription in which the UTP was substituted with 1-methylpseudo-UTP. mRNA constructs were then encapsulated in LNP formulations for subsequent delivery in vivo as previously (24). Two groups of Hartley guinea pigs (n = 5) were prime vaccinated via the intramuscular route on day 0 and boosted on day 27 with mRNA-LNP vaccine against MARV or against RAVV (Fig. 1). Two control groups of guinea pigs (n = 5) were mock vaccinated with PBS. Over the vaccination phase of the study, serum was collected, and the antibody binding and functional profiles were characterized.
Both mRNA vaccines generate autologous and asymmetric heterologous virus-neutralizing antibody responses
Following each vaccination dose, we monitored virus-specific antibody responses in serum. High anti-MARV or -RAVV IgG titers were detected by ELISA 27 days after the prime dose of the respective mRNA vaccine (Fig. 2A), which was further elevated by the booster dose. MARV- and RAVV-specific IgG titers were comparable after both the prime and boost vaccinations. Both vaccines also induced neutralizing antibodies against the respective viruses, which somewhat mirrored the IgG titers in that titers after the prime vaccination were further elevated by the booster (Fig. 2B).
Given that MARV and RAVV are genetically distinct but share 78% GP sequence identity at the amino acid level, we determined the ability of the serum collected after the boost vaccination to neutralize heterologous virus (Fig. 2C). The MARV vaccine autologous virus-neutralizing antibody titer was lower compared to the autologous virus-neutralizing titer generated by the RAVV vaccine (Fig. 2B). However, the MARV vaccine induced a higher cross-neutralizing titers against RAVV (reciprocal neutralizing titer 50 [NT50] of 89.0) compared to the RAVV vaccine neutralizing titers against MARV (NT50 of 8.2) (Fig. 2C).
These data demonstrate that MARV and RAVV mRNA vaccines elicit comparable binding antibody responses after prime and boost doses against their respective viruses. The RAVV vaccine yielded higher autologous neutralizing antibodies than MARV, but the MARV response appeared to be more cross-neutralizing.
The mRNA vaccines differ in their response towards the proteolytically cleaved form of GP
We further assessed the level of homologous binding to truncated forms of GP (Fig. 3A); MARV or RAVV GP ectodomains (GPΔTM), mucin-deleted ectodomains (GPΔmuc, Δ257–425) and proteolytically-cleaved GPs (GPcl), and the wing-deleted RAVV GPΔmuc (GPΔmucΔw, additional deletion of residues Δ436–483). All GP forms were immobilized on Octet biolayer interferometry (BLI) sensors at comparable levels and allowed to bind antibodies in serum. In general, similar antibody binding levels were observed between MARV and RAVV vaccine-derived serum to immobilized GPΔTM or GPΔmuc from the respective virus (Fig. 3B). MARV-specific serum antibodies appeared to have a lower binding capacity to GPcl than RAVV-specific serum antibodies. Therefore, the response generated from MARV mRNA vaccination may target the glycan cap (GC) which is absent on the proteolytically cleaved form of GP more so than RAVV-vaccination.
The MARV vaccine induces a greater antibody response to the GP glycan cap compared to the RAVV vaccine
The proportion of the vaccine-induced antibody response directed toward regions on MARV or RAVV GP [MLD, GC or receptor binding domain (RBD) of the GP1 subunit, and the wing and base regions of the GP2 subunit] was determined to identify the regions predominantly responsible for the binding antibody response to vaccination. GP region-specific responses were measured using BLI competition assays. Serum antibodies from vaccinated animals were allowed to bind a GP protein immobilized on the BLI sensor after pre-adsorption treatment with a GP variant (Fig. 3A) to remove antibodies targeting regions shared between the competing and immobilized GP. The proportion of MLD-specific antibodies was inferred from the percent of serum antibody binding to GPΔTM not removed by GPΔmuc pre-adsorption (Fig. 3C). MLD antibodies in the MARV vaccine recipient comprised approximately 40% of the response (Fig. 3D). The proportion of MLD antibodies in the RAVV vaccine recipients was similar to that of MARV recipients. GPcl, the protease cleaved form of GP, lacks the GC which is present in GPΔmuc and GPΔTM. We could therefore deduce the proportion of antibodies binding to the GC by subtracting the level of binding to GPΔTM removed by GPcl pre-adsorption from the level of binding to GPΔTM removed by GPΔmuc pre-adsorption (Fig. 3C,D). Approximately 55% of the MARV-directed response towards GPΔTM targeted the GC (Fig. 3D). However, the GC antibody proportion of the RAVV-directed response at ~ 17% was substantially lower than the MARV-directed response. When GPΔmuc was used as the capture ligand instead of GPΔTM (Fig. 3C), GC antibody proportions were augmented to 70% or 30% for the MARV or RAVV response (Fig. 3D), respectively, given the ratio of the surface areas of GC to GPΔmuc is greater than the ratio of GC to the full GPΔTM.
The fraction of the response towards the combined RBD, wing and GP2 regions was determined by the level of binding to GPΔTM or GPΔmuc inhibited by the presence of GPcl (Fig. 3C). The MARV-vaccine response towards this combined region was 30% less than the RAVV response. The greater proportion of GC antibodies relative to the total amount of binding antibodies in MARV recipients may have offset the response towards the combined RBD, wing and GP2 regions. However, the higher frequency of RAVV antibodies towards the combined RBD, wing and GP2 regions was confirmed using the reverse setting to calculate the level of binding to GPcl inhibited by GPΔmuc or GPΔTM pre-adsorption.
The proportion of the response directed to the wing domain could only be determined for RAVV due to the availability of RAVV-derived GPΔmucΔw. The level of binding to GPΔmuc following pre-adsorption of serum with GPΔmucΔw was used to calculate the frequency of the response towards the wing domain. Interestingly, GPΔmucΔw (436–483) had minimal effect on blocking antibody binding to all GP forms in the competition assays indicating that most of the response targeted the wing domain (Suppl. Figure 1A). However, in the reverse setting, serum antibody binding to immobilized GPΔmucΔw in the presence competing GP forms showed a substantial proportion of the response targeted shared regions, RBD and GP2 (Suppl. Figure 1B). Therefore, the actual wing domain antibody proportions in serum antibodies may be misrepresented in this assay system. The total response binding to GPΔmucΔw was poor (Suppl. Figure 1C). Furthermore, binding to the wing domain facilitates the structural rearrangement of GP to enhance binding of RBD antibodies (25). The absent wing domain on GPΔmucΔw may have prevented the sequestering of serum RBD antibodies by thwarting the cooperative recognition of RBD domains that occurs upon engagement of the wing domain.
Overall, these data demonstrate that both vaccines comparably target the MLD, but the MARV vaccine induces a greater antibody response to GC, while the RAVV vaccine induces a greater response towards the combined RBD, wing and GP2 regions.
MARV and RAVV vaccines induce virus-neutralizing antibodies specific for different regions of GP
We next determined the regions on GP targeted by neutralizing antibody responses generated by the MARV or RAVV mRNA vaccine. Day 56 sera from MARV or RAVV-vaccinated guinea pigs were diluted to a concentration required to achieve at least 80% of neutralization. Diluted sera generated were then pre-absorbed with increasing concentrations of truncated GP proteins derived from the specific viruses targeted by each mRNA vaccine (MARV GPΔmuc and GPcl or RAVV GPΔmuc, GPΔmucΔw and GPcl). The presence of the truncated GP proteins sequestered antibodies that bound to regions shared with the full-length GP on the virus.
The ability of serum from MARV mRNA vaccinated animals to neutralize the virus was nearly abolished with increasing concentration of MARV GPΔmuc, indicating that non-MLD-specific antibodies are major contributors to the neutralizing capacity of serum antibodies (Fig. 4A). The ability of non-MLD antibodies to neutralize virus in the presence of MARV GPcl was diminished to a lesser extent compared to neutralization in the presence of MARV GPΔmuc, indicating that the antibodies targeting the GC structure absent on MARV GPcl, contribute to virus neutralization.
For RAVV-vaccinated animals, non-MLD-binding antibodies also contributed to neutralization activity, although seemingly to a lesser extent than the MARV-derived antibodies since infectivity was not fully restored in the presence of RAVV GPΔmuc (Fig. 4B). While this finding may suggest that a substantial proportion of the RAVV-vaccine derived neutralizing antibodies targets the MLD, the presence of increasing concentrations of GPcl, the furin cleaved form of GP which lacks both the GC and MLD domains, restored virus infectivity to a better extent than did GPΔmuc. Accessibility to the RBD on the cleaved structure may enable improved sequestering of RBD-specific neutralizing antibodies. This finding also indicates that RAVV vaccine-derived neutralizing antibodies targeted the GP in its cleaved form better than the MARV-vaccine derived neutralizing antibody response.
Interestingly, RAVV virus infectivity was not restored in the presence of GPΔmucΔw, indicating that the RAVV wing domain is important for neutralization activity. Taken together, these data demonstrate that the GP regions targeted by neutralizing antibodies diverge between the two vaccines: the GC region was heavily involved in the neutralization response after MARV vaccination, while the RBD, wing and GP2 regions of the GPcl structure contributed substantially to the neutralization response following RAVV vaccination.
The vaccines induce antibodies binding to protective epitopes in RBD and wing domain
We quantified the prevalence of the response directed to known protective epitopes in the wing and RBD domains of GP. Representative monoclonal antibodies (mAbs) isolated in our previous studies from human survivors with epitopes in the RBD (MR72, MR78, MR82, MR111, MR191 and MR198), and wing domain (MR228 and MR235) (25, 26) were selected to compete for binding to GP on the BLI platform with serum antibodies in samples collected after the booster dose (Suppl. Figure 2). MR111 was specific for RAVV GP, and all RBD-specific mAbs are neutralizing antibodies (26). The two mAbs specific to the wing domain do not neutralize virus but have Fc-mediated effector functions (25). MR228 only bound the wing domain of MARV GP, limiting the number of wing domain antibodies that could be used in competition for RAVV GP binding. Among these antibodies, MR72, MR78, MR82 and MR228 protected small animal models (25–27) while MR191 protected NHPs (27). Each of the representative RBD epitopes was targeted by a similar quantity of serum from recipients of the MARV mRNA vaccine (Fig. 5A). The response toward all RBD mAb epitopes, except for MR72, was lower in frequency than the response towards the wing domain epitope for MR228. The antibody frequency towards the wing domain epitope for MR235 was comparable with the RBD-targeted response. Conversely, the frequency of antibodies in RAVV mRNA recipients was similar towards most RBD epitopes, except the epitopes for MR82 and MR198. The frequency of antibodies targeting RBD epitopes was generally higher than the frequency directed towards the wing domain epitope for MR235 (Fig. 5B). Overall, the response recognition frequency to known epitopes in the RBD and wing domains appeared to differ between the MARV and RAVV mRNA vaccines.
The vaccines induce antibodies specific for both cross-reactive and unique linear epitopes
The MARV and RAVV vaccine antibody response profiles were further scrutinized for any parallels or uniqueness in their linear epitope recognition. Linear epitopes of GP targeted by antibodies were characterized using peptide arrays designed with overlapping 15-mer peptides spanning the entire GP of MARV (strain Angola) or RAVV (strain Ravn), offset by 4 amino acids. Serum antibodies in samples collected after boost vaccination dose were allowed to bind each of the GP proteins to identify homologous or heterologous linear epitope recognition.
Linear epitopes located within the RBD (peptides 15 to 18, 23 to 30) and wing domains (peptides 109 to 116) of MARV and RAVV GP were recognized by both homologous and heterologous vaccine-induced antibodies. Moreover, the magnitude of vaccine-induced antibody binding to these epitopes within the homologous virus somewhat mirrored the magnitude of binding observed towards the same epitope in the heterologous virus. The RBD is highly conserved between the marburgviruses. Peptides 15 to 18, 23 to 25 and 29 to 30 encompass part of the engagement site for the host receptor Niemann-Pick C1 and the footprint for mAbs MR78 and MR191, which are neutralizing mAbs isolated from human survivors (28) that provide post-exposure protection to NHPs (23, 27). Peptides in the MARV GP2 wing domain recognized by both MARV and RAVV mRNA vaccine-specific antibodies encompass the epitopes for three mAbs, the two human mAbs MR235 and MR228 and the murine mAb 30G4 (25, 29). Peptides in the RAVV GP2 wing domain corresponding to the MR228 epitope were not recognized by MARV and RAVV mRNA vaccine-specific antibodies (peptide 112). MR228 failed to bind RAVV GP due to a two amino acid difference (aa 454T-455E) in the epitope compared to MARV GP (aa 454A-455P). This amino acid divergence appears to diminish the recognition potential of the humoral response towards RAVV.
Antibodies targeted to linear epitopes in the MLD were unique to the respective viruses, with MARV-MLD antibodies unable to recognize the RAVV-MLD and vice versa. RAVV-vaccine antibodies had a greater breadth of binding to the MLD than MARV-derived antibodies. MLD is a poorly conserved region between the two viruses, and therefore the lack of recognition of MLD in heterologous viruses was not unexpected.
Interestingly, MARV mRNA vaccination induced antibodies with a greater capacity to bind the internal fusion loop (IFL) region of both MARV and RAVV, compared to the RAVV mRNA-vaccination. Weak recognition of the IFL by RAVV-vaccine antibodies indicates they have lower affinity than MARV vaccine antibodies, the RAVV IFL is poorly ranked amongst the immunogenic B-cell epitope hierarchy or the RAVV GP2 stem is somewhat obstructed (29). MARV mRNA vaccination also induced antibodies with a greater capacity to bind the GP2 stem region of both MARV and RAVV GP compared to RAVV mRNA vaccination. Antibody recognition of the stem was generally weak, with detection of peptide 149 in the heptad repeat-1 (HR1) of both MARV and RAVV GPs being strongest in the region. While protective mAbs targeting the stem have not been identified for marburgviruses thus far, an indication of rarity, mAbs specific for the GP stem of ebolaviruses, have been isolated from human survivors (30).
The peptide arrays highlight regions within GP that are virus-specific and regions that are cross-reactive. While the homologous antibody response following vaccination was greater in breadth and magnitude toward the linear epitopes from the respective virus, a comparable magnitude of recognition was observed for heterologous binding at cross-reactive epitopes. The strongest binding was observed for the RBD of GP1 and the wing of GP2. The linear epitopes recognized by cross-reactive vaccine-induced antibody populations may be important contributors to their ability to cross-neutralize the two viruses.
The vaccines induce multiple Fc mediated effector functions
In addition to mechanical neutralization by antibodies, their Fc-mediated effector functions have been implicated in contributing to protection in vaccinated and natural infection survivors (31). We examined the ability of the MARV and RAVV vaccine-induced immune sera to activate phagocytosis mediated by neutrophils (antibody-dependent neutrophil phagocytosis; ADNP) and monocytes (antibody-dependent cellular phagocytosis; ADCP). Antibody responses produced after MARV or RAVV mRNA vaccination activated virus strain-specific ADNP and ADCP functions in vitro. ADNP and ADCP activities were higher after prime MARV vaccination compared to after RAVV vaccination, but activities increased after the boost dose such that discernable differences were not observed between the two vaccines (Fig. 7A,B).
We also examined the ability of serum antibodies to facilitate antibody-dependent natural killer (ADNK) cellular cytotoxicity by measuring their markers for degranulation (CD107a) and activation (macrophage inflammatory protein-1β [MIP-1β] and interferon-γ [IFNγ]). Activation of the NK cellular activity was achieved by both vaccines (Fig. 7C-E). The MARV and RAVV boosters were required to activate similar levels of degranulation. The levels of NK cells positive for MIP-1β were comparably increased after prime vaccinations with both vaccines and further elevated by the booster vaccinations. Almost no NK cells positive for IFNγ were detected after prime vaccination; unexpectedly, the levels increased after a booster of the RAVV but not the MARV vaccine. NK cells may control infection directly by their cytolytic functions and only partially by relying on their recruitment of other immune cells through MIP-1β, but not IFNγ production.
mRNA vaccines protect against MARV and RAVV infections
At day 56, guinea pigs were challenged with 1,000 plaque-forming units (PFU) of guinea pig-adapted MARV strain Angola (32) or guinea pig-adapted RAVV strain RAVV (33), respectively (Fig. 1). All guinea pigs vaccinated against MARV or RAVV survived infection (Fig. 8A). Over the 28-day infection phase, serum was collected at 3-day intervals for the first 12 days to measure for viremia (Fig. 1). Guinea pigs were also monitored for clinical signs of disease including lethargy, neurologic signs and weight loss. Vaccinated guinea pigs maintained steady weight over the infection phase (Fig. 8D), did not have detectable viremia (Fig. 8B), and displayed no signs of disease (Fig. 8C). One RAVV-vaccinated guinea pig sustained a physical injury unrelated to infection and was euthanized at day 23 (Fig. 8A). No virus was detected in the blood of this animal collected at the time of euthanasia. Control RAVV-infected guinea pigs developed severe disease and exhibited weight loss before succumbing to infection by day 9. Four out of 5 MARV-infected control guinea pigs succumbed by day 8, and exhibited clinical disease and weight loss over the infection course (Fig. 8C,D). The shorter time to lethality with MARV infection compared to RAVV is consistent with its greater virulence observed in the guinea pig model (33). The lack of detectable circulating virus in the surviving RAVV control guinea pig (Fig. 8B) may indicate imprecise administration of the infectious inoculum by intraperitoneal injection, a strict requirement to achieve uniform lethality.