*3.1. Structural Biology to Reveal Epitopes on GP Targeted by Antibodies*

Antibodies isolated from immunized animals or patients infected with ebolavirus have been shown to target several different regions on the surface of the GP trimer. Epitope mapping can be achieved rapidly using competition binding assays or negative stain electron microscopy (EM). However, to definitively understand the antibody interactions with glycoprotein, a high-resolution structure by X-ray crystallography or cryo-EM is required. Structural characterizations of these antibodies in complex with full-length GP ectodomain or GP peptides provide detailed information of the neutralizing mechanism and inform new approaches for broad immunotherapy and vaccine design. The recent development of single-particle cryo-EM capabilities facilitated the determination of more structures of antibody-glycoprotein complexes, including those that involve asymmetric interactions that are difficult to assess by crystallography. As a whole, the structural analyses illustrate how antibodies target the various epitopes on the GP surface, particularly those epitopes in highly conserved regions, to achieve high potency and/or cross-reactivity.

#### *3.2. mAbs Targeting the Glycan Cap*

In EBOV, the glycan cap spans between residues 227 and 312, and the majority of residues (amino acids 227–295) are present in both GP and sGP (Figure 2). Thus, antibodies targeting the GP glycan cap typically also react with the abundant, non-structural sGP. If these antibodies were elicited by natural infection, they may, in fact, have been elicited against sGP, which is at least five-fold more abundant than membrane-bound GP. One component of the therapeutic cocktail ZMapp [13,72,73], 13C6 [74], targets the glycan cap and offers protection in in vivo models of infection despite having low neutralizing potency [29,72].

Antibodies against the glycan cap, including 13C6, can be characterized by higher levels of immune effector functions [69]. Some neutralize as well, and several have been characterized functionally and structurally, such as EBOV-548 and EBOV-296 (Figure 3A) [60,75]. Approaching the glycan cap via different angles, anti-glycan cap mAbs with GP largely involve CDRH3 or CDRH2 that mimic and displace the β18-18 hairpin, which acts as an

anchor for the MLD. The proposed mechanism of neutralization for these anti-glycan cap antibodies is blockage of the cathepsin cleavage event that is required for RBS exposure and viral entry [75]. Higher numbers of contacts between a mAb and the MLD cradle are shown to introduce instability in the GP trimer, and thus these antibodies can synergize with those that target the fusion loop. Other glycan-cap targeting mAbs include the Q206, Q314, and Q411 antibodies identified in immunized macaques, which provide partial protection in a mouse model of EBOV challenge [76]. Overall, mAbs in this group are usually potent but rarely have broad neutralizing activity. However, a combination of both neutralizing and effector functions and the demonstrated synergistic effect when pairing with mAbs that target the fusion loop, make some of the more potent glycan cap mAbs good candidates for inclusion in therapeutic cocktails [54,60,75].

**Figure 3.** Structural models of neutralizing antibody recognition against key epitopes on the GP surface. GP epitopes are colored as in Figure 2. The variable regions of the antibodies targeting different epitopes are shown in cartoon representation. (**A**) Glycan cap-targeting antibody EBOV-296 (PDB: 7KF9) (**B**) Head-region targeting antibody 5T0180 (PDB:6S8J) (**C**) IFL-targeting antibody ADI-15878 and ADI-15946 (PDB: 6EA5, 6MAM) (**D**) Stalk region-targeting antibody BDBV 223 (PDB: 6N7J, 5JQ3) (**E**) MLD-targeting antibody 14G7 (PDB: 2Y6S. 5JQ3) (**F**) An example of a broad neutralizing antibody cocktail 1C3 and 1C11, with 2 antibodies targeting the head and IFL, respectively. (PDB: 7SWD).

#### *3.3. mAbs Targeting the Apex/Head/Receptor Binding Region of GP*

The GP1 Head epitope lies under the glycan cap and contains residues that are part of the RBS. In the late endosome, primed GPCL exhibits a fully exposed RBS that is competent for binding to domain C of NPC1 (NPC1-C) [42]. In contrast to ebolaviruses, the glycan cap of marburgviruses provides a less complete shield and the RBS is more exposed prior to cleavage, such that several antibodies including MR78 and MR191 can target the RBS directly [77–79]. MR78 and MR191 somewhat mimic interactions made by a loop of NPC1-C, which contains aromatic residues that can reach into a hydrophobic cavity on GPCL [42].

The monotherapy mAb114, which was isolated from a survivor of the 1995 Kikwit EVD outbreak, and is approved to treat EVD, targets the head epitope in the RBS via a near-vertical angle to block receptor binding [80]. FVM04, isolated from immunized macaques [50], binds to the inner chalice of the GP trimer near the glycan cap with a tilted angle, so that from low-resolution negative stain EM, only one Fab bound to GP trimer can be visualized [81]. FVM04 can bind and neutralize both EBOV and SUDV but has lower activity toward BDBV [81]. Other head-targeting antibodies (5T0180, 1T0227, and 3T0265), isolated from rVSV-ZEBOV vaccine recipients, also bind the RBS and block NPC1-C, while avoiding contact with the glycan cap region [82] (Figure 3B). Although the residues within the footprint of these 3 antibodies are mostly conserved, a key difference between EBOV and BDBV/SUDV at residue 224 (G in EBOV, N in BDBV/SUDV) sterically prevents their binding to BDBV/SUDV GP, thus limiting the breadth of antibody potency [82]. Overall, previously discovered head targeting neutralizing antibodies are potently neutralizing and protective, but limited in breadth.

A recently published apex-targeting antibody 1C3 is unique among currently characterized EBOV antibodies in that it targets the center of the GP chalice and binds one Fab to one GP trimer (Figure 3F) [62]. 1C3 potently neutralizes both EBOV and SUDV. Although 1C3 does not neutralize BDBV, it does bind to recombinant BDBV GP in ELISA, suggesting that this antibody could potentially contribute to protection against BDBV infection through Fc-dependent effector functions [62]. The asymmetric binding with 1:1 stoichiometry of 1C3 allows more variability within its footprint [83]. Notably, the quaternary recognition of 1C3 is specific for GP and not shed sGP, which may provide an advantage for mAb therapeutic candidates.

#### *3.4. mAbs Targeting Internal Fusion Loop (IFL)*

The IFL region (residues 511–554) plays a critical role in viral-host membrane fusion. This important role translates to a high degree of sequence conservation (60–70%), which makes the IFL an ideal epitope for cross-reactive antibodies. The IFL region can be further divided into stem/base (IFLstem; residues 511–520 and 543–554) and the loop/paddle (IFLloop; residues 521–542) [84]. MAbs targeting the IFL usually also contact parts of the base epitope, which is immediately adjacent to the IFL and forms the base of the "bowl" of the GP chalice. The central span of GP, including both the IFL and base, has been termed the "waist" and is targeted by antibodies via a continuum of antibody epitopes [85]. Several cross-reactive antibodies identified thus far have been categorized into the IFL targeting group, including CA45 [86,87], 6D6 [88], ADI-15946 [89], ADI-15878 [85], 2G1 [90], EBOV-520 [60], EBOV-515 [61], and 1C3 [62]. Among these mAbs, 6D6, ADI-15878, and 1C3 have similar footprints that overlap and include the IFLloop region and part of the base region (although the 6D6 complex structure has been resolved only at low-resolution with negative stain EM). Meanwhile, CA45, ADI-15946, EBOV-520, and EBOV-515 have similar footprints that include both the IFLstem and other parts of the base region (Figure 3C). 2G1, isolated from a vaccinated donor and which cross-neutralizes pseudotyped EBOV, SUDV, and BDBV, was determined to bind GP2 by competition assays and was further mapped to the fusion loop by computational modeling [90].

ADI-15878 contacts the IFLloop and the portion of the base termed the N-terminal pocket, which is occupied by the flexible N-terminal tail of GP2 in the GP apo-structure. The residues that line the N-terminal pocket are highly conserved, but those in the flexible N-terminal tail are not. Therefore, the ability of the ADI-15878 CDRs to reach the conserved pocket region underneath the N-terminal tail provides the cross-reactivity against different ebolaviruses [85]. CA45 targets both the IFLstem and a region termed the DFF cavity [87,91], which is occupied by a short flexible loop that includes residues 192–194 near the cathepsin cleavage site (residues DFF) in the apo GP structure. The footprint of 1C11 partially overlaps that of ADI-15878, but is shifted upwards [62].

ADI-15946, EBOV-520, and EBOV-515 all contact the IFLstem and a region termed the 310 pocket, the core of which encompasses residues 71–75 of GP1. The pocket extends to surrounding residues, including 76–78 of GP1 and 510–516 of GP2 [60,89]. This region is occupied by the β17-β18 loop (residues 287–291), which is part of the glycan cap in the uncleaved GP apo-structure and is exposed in GPCL following removal of the glycan cap. Similar to the crystal structure of the ADI-15946-GPCL complex, the cryo-EM structure of EBOV-520 in complex with uncleaved GP ectodomain shows that the CDRH3 loop of the antibody contacts the 310 pocket [60]. Although the glycan cap region is intact in the EBOV-520 complex structure, the β17-β18 loop cannot be visualized, suggesting that this flexible loop is displaced upon contact with the mAb [60].

ADI-15946 potently neutralizes EBOV and BDBV, but not SUDV, whereas EBOV-520 neutralizes all three viruses. Comparing the footprints of the two mAbs, EBOV-520 is shifted slightly upward to avoid non-conserved position 506 (N506 in EBOV and R506 in SUDV), which may allow the extra reactivity towards SUDV GP [60]. The study on ADI-15946 also provided a good example of structure-based rational engineering. Three residues were substituted to reduce steric and charge clashes with the non-conserved residues in SUDV (R100A in CDRH3) or to improve binding by generating a double tyrosine binding motif (S65Y and F67Y in the FRL3). The designed mAb variant successfully expanded the breadth of ADI-15946 to enhance its binding and neutralization against SUDV [89].

Overall, mAbs that target the IFL region enlist a variety of approaches to contact the most conserved region on the GP surface and showcase the largest number of broadly neutralizing antibodies that have been discovered and characterized to date. The identification of the flexible loops in GP that potentially compete with the mAbs from this group suggests that removal of such regions would contribute to improved antigens and facilitate the development of more antibodies that target these ideal epitopes.

#### *3.5. mAbs Targeting GP Stalk and MPER Region*

The stalk/MPER region lies near the C terminus of GP2 above the transmembrane domain. As part of the fusion machinery, the stalk/MPER region has high sequence conservation across filovirus species and is a prime target for mAbs that have broad potency. Negative stain EM was used to map the binding footprint of several crossreactive antibodies that target the stalk region (BDBV 223, BDBV 317, BDBV 340, and ADI-16061) [54,55]. A high-resolution X-ray crystal structure of the Fab from BDBV 223, isolated from a survivor of BDBV infection, was determined in complex with a synthetic peptide of the epitope region was determined [92] (Figure 3D). Interestingly, the alignment of the complex structure to the GP trimer structure and tomographic reconstruction of the GP trimer on the virus membrane [93] revealed that BDBV 223 binding interferes with the trimeric bundle assembly and anchoring of the GP spike in the viral membrane. Thus, interference with the six-helix bundle formation needed to drive membrane fusion could be a key mechanism by which BDBV 223 neutralizes infection [92].
