A “Trojan horse” bispecific-antibody strategy for broad protection against ebolaviruses

Treating Ebola with a Trojan horse The recent major Ebola virus outbreak in West Africa high-lighted the need for effective therapeutics against this and other filoviruses. Neutralizing ebolaviruses with antibodies is a challenge because the viruses bind their entry receptor, NPC1, inside the cell within endosomes rather than on the cell surface. Furthermore, enzymes in endosomes cleave the Ebola virus surface glycoprotein (GP) to reveal its receptor binding site. Wec et al. now report a bispecific antibody strategy targeting all known ebolaviruses that overcomes this problem (see the Perspective by Labrijn and Parren). They coupled an antibody specific for a conserved, surface-exposed epitope of GP to antibodies that recognize either NPC1 or the NPC1 binding site on GP. Treating mice therapeutically with these antibodies allowed them to survive otherwise lethal ebolavirus infection. Science, this issue p. 350; see also p. 284 Bispecific antibodies show therapeutic efficacy against ebolaviruses in mice. There is an urgent need for monoclonal antibody (mAb) therapies that broadly protect against Ebola virus and other filoviruses. The conserved, essential interaction between the filovirus glycoprotein, GP, and its entry receptor Niemann-Pick C1 (NPC1) provides an attractive target for such mAbs but is shielded by multiple mechanisms, including physical sequestration in late endosomes. Here, we describe a bispecific-antibody strategy to target this interaction, in which mAbs specific for NPC1 or the GP receptor–binding site are coupled to a mAb against a conserved, surface-exposed GP epitope. Bispecific antibodies, but not parent mAbs, neutralized all known ebolaviruses by coopting viral particles themselves for endosomal delivery and conferred postexposure protection against multiple ebolaviruses in mice. Such “Trojan horse” bispecific antibodies have potential as broad antifilovirus immunotherapeutics.

There is an urgent need for monoclonal antibody (mAb) therapies that broadly protect against Ebola virus and other filoviruses. The conserved, essential interaction between the filovirus glycoprotein, GP, and its entry receptor Niemann-Pick C1 (NPC1) provides an attractive target for such mAbs but is shielded by multiple mechanisms, including physical sequestration in late endosomes. Here, we describe a bispecific-antibody strategy to target this interaction, in which mAbs specific for NPC1 or the GP receptor-binding site are coupled to a mAb against a conserved, surface-exposed GP epitope. Bispecific antibodies, but not parent mAbs, neutralized all known ebolaviruses by coopting viral particles themselves for endosomal delivery and conferred postexposure protection against multiple ebolaviruses in mice. Such "Trojan horse" bispecific antibodies have potential as broad antifilovirus immunotherapeutics.
T he development of therapeutics targeting Ebola virus (EBOV) and other filoviruses is a global health priority. The success of ZMapp-a cocktail of three monoclonal antibodies (mAbs) targeting the EBOV surface glycoprotein GP-in reversing Ebola virus disease in nonhuman primates (NHPs) has underscored the promise of antiviral immunotherapy (1). However, most available mAbs have a narrow antiviral spectrum, because they recognize variable surface-exposed GP epitopes (2). ZMapp protects against EBOV but not against other filoviruses with known epidemic potential, including the ebolaviruses Bundibugyo virus (BDBV) and Sudan virus (SUDV) and the more divergent marburgviruses. Given the scientific and logistical challenges inherent in developing a separate mAb cocktail for each filovirus, as well as the need for preparedness against newly emerging or engineered viral variants, broadly protective antifilovirus immunotherapies are highly desirable. A few mAbs have shown cross-neutralization and protection in rodents, indicating that cross-species protection by a single molecule is possible; however, such antibodies are rare (3)(4)(5)(6)(7)(8).
We envisioned a bispecific antibody (bsAb)engineering strategy to block intracellular GP CL -NPC1 interaction by a "Trojan horse" mechanism. We reasoned that, by coupling receptor or RBStargeting mAbs to a delivery mAb directed against a broadly conserved epitope in uncleaved GP, virions themselves could be coopted to transport bsAbs to the appropriate endosomal compartments (Fig. 1, A and B). To block the filovirusreceptor interaction, we chose mAbs targeting both its viral and host facets: MR72, a human mAb that recognizes the GP CL RBS (above), and mAb-548, a novel murine mAb that engages human NPC1-C. mAb-548 bound with picomolar affinity to an NPC1-C epitope that overlaps the GP CLbinding interface and blocked GP CL -NPC1-C association in vitro at pH 5.5, the presumptive pH of late endosomes (figs. S1 and S2). mAb-548 resembled MR72 in its lack of neutralizing activity against uncleaved viruses (Fig. 2, A and B, and fig. S6), likely because NPC1 is absent from the cell surface (11,24). To deliver mAb-548 and MR72 to endosomes, we selected the macaque mAb FVM09, which recognizes a conserved linear epitope in the GP glycan cap of all known ebolaviruses ( Fig. 1A and fig. S4) (8). FVM09 does not neutralize infection and confers limited in vivo protection against EBOV (8).
The heavy-and light-chain variable domains (V H and V L , respectively) of FVM09 were fused to mAb-548 and MR72 by using the dual-variable domain immunoglobulin (DVD-Ig) design strategy (25). The DVD-Ig format was chosen as a test case because it allows bivalent binding of both combining sites but does not use long polypeptide linkers that may be susceptible to proteolysis or immunogenic presentation. The FVM09~548 and FVM09~MR72 DVD-Igs could be readily isolated from transiently transfected human embryonic kidney 293 (HEK293) cells ( fig. S3A). Size-exclusion chromatography-multiangle light scattering indicated a monodisperse population of monomers, with some higher aggregate present ( fig. S3, B and C). Each DVD-Ig could bind to EBOV GP via the FVM09 "outer" variable domains, with no loss of affinity relative to the parent FVM09 immunoglobulin G (IgG), as determined by biolayer interferometry (BLI) ( Fig.  1C and table S1). FVM09~548 could recognize human NPC1-C, by means of its "inner" variable domains, with a subpicomolar equilibrium dissociation constant (K D ). The MR72 variable domains also retained subnanomolar affinity toward GP CL in the DVD-Ig format . Two-phase binding studies, in which each DVD-Ig was first exposed to EBOV GP and then to NPC1-C or GP CL (Fig. 1D), indicated that there were no steric restrictions to engagement of both combining sites.
We tested the DVD-Igs for their capacity to neutralize infection in human cells by recombinant vesicular stomatitis viruses bearing EBOV GP (rVSV-EBOV GP) or control nonfilovirus glycoproteins derived from VSV and Andes hantavirus (Fig. 2, A and B, and fig. S5) (26). Both FVM09~548 and FVM09~MR72 specifically and potently neutralized rVSV-EBOV GP, whereas the parental mAbs FVM09, mAb-548, and MR72 had little or no neutralizing activity. Equimolar mixtures of the "delivery" IgG, FVM09, with each receptor-RBS-targeting IgG (mAb-548 or MR72) also did not neutralize infection (Fig. 2, A and B); this indicated that DVD-Ig antiviral activity requires the physical linkage of delivery and receptor-RBSbinding specificities. Overall, the DVD-Ig halfmaximal inhibitory concentration (IC 50 ) values were in the nanomolar range, similar to the measured K D of the FVM09-GP complex but higher than those of the mAb-548-NPC1-C and MR72-GP CL complexes.
The GP CL -NPC1 interaction is conserved among filoviruses (11,15,21,27,28), and thus, we postulated that the DVD-Igs would exhibit broad neutralizing activity. rVSVs bearing GP proteins from the four other ebolaviruses were sensitive to neutralization by both DVD-Igs, whereas rVSV-MARV GP was resistant ( Fig. 2, C, D, and G), consistent with the known specificity of FVM09 toward ebolaviruses (8). We next tested the DVD-Igs against authentic EBOV, BDBV, and SUDV (Fig. 2, E and F). Each ebolavirus was neutralized by both receptor-and RBS-targeting DVD-Igs but not by the individual parent IgGs ( Fig. 2G and fig. S6).
The success of antibody therapeutics has fueled the development of a panoply of optimized bsAb architectures, several of which (including the DVD-Ig) are in clinical trials (29). To explore the generality of our strategy to other formats, we generated an FVM09*MR72 "asymmetric IgG" using the DuoBody platform (figs. S7 and S8) (30). FVM09*MR72 broadly neutralized rVSVs bearing ebolavirus GPs, albeit with reduced potency against SUDV GP, possibly because of its loss of bivalent recognition of GP and/or GP CL . Nonetheless, these results illustrate that endosomal targeting of the ebolavirus-receptor interaction is amenable to other bispecific-antibody formats.
Our observation that bsAbs combining two nonneutralizing antibodies could confer potent neutralization implied critical roles for both binding specificities. This hypothesis is supported by three pieces of evidence. First, the activity of the DVD-Igs against rVSV-EBOV GP particles containing two point mutations in the FVM09 epitope was greatly reduced ( Fig. 3A and fig. S9).   shown at the left. Parent IgGs and DVD-Igs covalently labeled with the aciddependent fluorophore pHrodo Red were incubated with rVSV-EBOV GP particles and exposed to cells. Virus -Ab + and virus + Ab + populations were measured by flow cytometry. Averages ± SD for four technical replicates pooled from two independent experiments are shown. Group means for the percentage of Ab + cells were compared by two-factor analysis of variance (ANOVA) (see fig. S11). Šídák's post hoc test was used to compare the capacity of each Ab to internalize into virusversus virus + cell populations ( •••• P < 0.0001; ns, not significant). Dunnett's post hoc test was used to compare the internalization of each Ab to that of the "no Ab" control in virus + cell populations (****P < 0.0001; all other Ab versus no Ab comparisons were not significant). (E) Delivery of Abs to NPC1 + endosomes. FVM09~548 was incubated with rVSV-EBOV GP particles and exposed to cells expressing an NPC1-enhanced blue fluorescent protein-2 fusion protein. Viral particles, Ab, and NPC1 were visualized by fluorescence microscopy (also see figs. S12 and S13). Representative images from two independent experiments are shown. Scale bar, 20 mm.
We postulated that the bsAbs harness extracellular virions for their delivery to endosomal sites of filovirus-receptor interaction in the context of natural infection. Accordingly, we evaluated the internalization of DVD-Igs and their parent IgGs into cellular endosomes ( Fig. 3D and figs. S10 and S11). Each Ab was covalently labeled with the acid-dependent fluorescent probe pHrodo Red and exposed to cells, either alone or following preincubation with fluorescent rVSV-EBOV GP particles (19,26). Cells were measured for both virus-and Ab-associated fluorescence by flow cytometry. Virus-negative cells displayed little Ab signal; this indicated that neither the DVD-Igs nor their parent IgGs could internalize into cells without virions. By contrast, viruspositive cells were strongly positive for the DVD-Igs but not for the parent mAb-548 and MR72 IgGs. Concordantly, only FVM09 and the DVD-Igs could efficiently colocalize with virions ( Fig. 3E and fig. S12) or Ebola virus-like particles (VLPs) (fig. S13) in NPC1 + late endosomes, where viral membrane fusion takes place (19,31). These results, together with the capacity of FVM09 to bind EBOV GP with high affinity between pH 5.5 and 7.5 ( fig. S14 and table S1), suggest that virion-bsAb complexes remain associated in early endosomes and traffic together to late endosomes, where proteolytic removal of the GP glycan cap dislodges FVM09, and where mAb-548 and MR72 can engage their respective cellular and viral targets. Collectively, our findings support a twostep "deliver-and-block" mechanism for bsAb neutralization.
Finally, we evaluated the protective efficacy of the DVD-Igs in two murine models of lethal ebolavirus challenge. Because our prior experiments were conducted in human cells, we first tested DVD-Ig neutralization activity in murine NIH/3T3 cells ( fig. S15). Whereas FVM09~MR72 retained full activity, FVM09~548 exhibited poor neutralization in murine cells (fig. S15, A and B). This could be readily explained by FVM09~548's reduced binding affinity for the murine NPC1 ortholog (fig. S15C) and consequent reduced capacity to block the GP CL -NPC1 interaction (fig. S15E). The discrepancy between binding of mAb-548 to human and mouse NPC1 likely arises from species-dependent amino acid sequence differences in the mAb-548-binding region of NPC1-C ( fig. S16). By contrast, this region in human NPC1-C is identical to those of rhesus macaques and crabeating (cynomolgus) macaques ( fig. S16), which provide the two NHP models of filovirus challenge currently in use. mAb-548 bound strongly to, and inhibited GP CL interaction with, an NHP NPC1 ortholog derived from the mantled guereza, which also shares an identical mAb-548-binding region ( fig. S15, D and F). Therefore, although host species-specific differences in NPC1 binding may affect FVM09~548's efficacy in rodents, they are unlikely to do so in NHPs and humans.
Both DVD-Igs were tested for their capacity to protect BALB/c mice when administered 2 days after a lethal challenge with EBOV-MA (Fig. 4A) (32). FVM09~MR72 afforded a high level of protection (70%) relative to the untreated group, whereas no significant survival was recorded for FVM09~548 and parent IgG mixtures. We also evaluated the DVD-Igs for postexposure protection against a lethal human SUDV isolate in the immunocompromised, type 1 interferon a/b receptor-deficient (IFNa/b R −/− ) mouse model (Fig. 4B) (7, 33). FVM09~MR72 was fully protective, and FVM09~548 provided partial protection, relative to the untreated group. The limited in vivo efficacy of FVM09~548 was consistent with its reduced capacity to inhibit the GP CLmurine NPC1 interaction (fig. S15). These findings provide evidence that a bsAb targeting the critical intracellular virus-receptor interaction can confer broad protection against lethal ebolavirus challenge, even under stringent conditions of postexposure treatment.
Recent antibody discovery efforts have demonstrated the existence of conserved GP surface epitopes that can elicit broadly reactive mAbs with cross-protective potential (3)(4)(5)(6)(7)34). Herein, we describe a complementary strategy to generate broadly protective Abs that target highly conserved epitopes at the intracellular filovirusreceptor interface, which are normally shielded from GP-specific mAbs. Because the cryptic epitopetargeting components of the bsAbs engineered in this study block endosomal receptor binding by all known filoviruses (15) (this study), nextgeneration molecules combining them with appropriate delivery mAbs of viral or cellular origin may afford coverage against all filoviruses, including newly emerging and engineered variants. This Trojan horse bispecific-antibody approach may also find utility against other viral pathogens known to use intracellular receptors [e.g., Lassa virus (35)], or more generally, to target entry-related virus structural rearrangements that occur only in the endolysosomal pathway.  The proteasome generates the epitopes presented on human leukocyte antigen (HLA) class I molecules that elicit CD8 + T cell responses. Reports of proteasome-generated spliced epitopes exist, but they have been regarded as rare events. Here, however, we show that the proteasome-generated spliced peptide pool accounts for one-third of the entire HLA class I immunopeptidome in terms of diversity and one-fourth in terms of abundance. This pool also represents a unique set of antigens, possessing particular and distinguishing features. We validated this observation using a range of complementary experimental and bioinformatics approaches, as well as multiple cell types. The widespread appearance and abundance of proteasome-catalyzed peptide splicing events has implications for immunobiology and autoimmunity theories and may provide a previously untapped source of epitopes for use in vaccines and cancer immunotherapy.
T he presentation of epitopes on the cell surface is a key mechanism by which organisms identify the presence of pathogens, metabolic malfunctioning, or tumors. The HLA class I (HLA-I) immunopeptidome-the set of epitopes allocated onto the HLA-I moleculesimpinges on the CD8 + T cell repertoire and the cell-mediated immune response (1). HLA-I immunopeptidomes are usually investigated by sequence identification of peptides eluted from HLA-I molecules by means of tandem liquid chromatographymass spectrometry (LC-MS/MS) (fig. S1). The key step for the transformation of a protein into HLA-I-restricted epitopes is usually processing by the proteasome (1), which cuts proteins into peptides; alternatively, the proteasome can also cut and paste peptide sequences, thereby releasing peptide antigens that do not correspond to the original protein sequence (2) (fig. S2). This proteasomecatalyzed peptide splicing (PCPS) has long been considered to occur only rarely; partly this has been because the screening of the HLA-I immunopeptidome for proteasome-generated spliced peptides was impeded by methodological challenges.
To overcome these problems, we developed an analytical strategy that accounts for recent discoveries underpinning the PCPS mechanism and can handle the vast proteome-wide human spliced peptide database ( fig. S3). With this strategy, we initially analyzed the HLA-I-eluted im-munopeptidome of the GR lymphoblastoid cell line (GR-LCL); for a deeper coverage of the immunopeptidome, we adopted a two-dimensional (2D) peptide prefractionation strategy followed by a hybrid peptide fragmentation method [electrontransfer higher-energy collision dissociation (EThcD)] for peptide identification (3, 4) (fig. S1), supplemented by an adapted target-decoy approach ( fig. S4). Our analysis led to the identification of 6592 nonspliced and 3417 spliced peptides 9 to 12 amino acid residues in length (9-to 12-mer peptides) (table S1). The latter number represents 34% of the total of identified antigenic peptides (Fig. 1A), thereby increasing the number of identified HLA-I ligands by some 50%. We confirmed the authenticity of the identified spliced antigenic peptides by comparing the LC-MS/MS spectra of 98 exemplary spliced peptides with their corresponding synthetic peptides and computing their correlation score (table S2 and fig. S5). In addition, we verified the proteasomedependent generation of the spliced antigenic peptides in vitro for three examples by digestions of synthetic polypeptides harboring the corresponding antigenic peptides by purified 20S proteasome ( fig. S6).
We queried the HLA-I immunopeptidome mass spectrometry data of GR-LCL against the standard Swissprot human proteome database, which does not account for spliced peptides; this revealed that 655 peptides (i.e., 9% of the total nonspliced antigenic peptides) would be erroneously matched against this incomplete database as nonspliced peptides, because they have much better hits as spliced peptides in our search against the combined spliced and nonspliced peptide database (Fig. 1A). The correct sequences for a set of these antigenic spliced peptides were verified by comparing the LC-MS/MS spectra of the synthetic spliced and nonspliced candidates with the corresponding LC-MS/MS of the GR-LCL HLA-I immunopeptidome ( fig. S7, A to I). Our identifications are further supported by the ion score distributions (see fig. S4) of the nonspliced peptides and of those spliced peptides that were wrongly assigned as nonspliced peptides using the standard Swissprot human proteome database, which differ only slightly in their median but not in the overall shape ( fig. S7J).