Antiviral protection by antibodies targeting the glycan cap of Ebola virus glycoprotein requires activation of the complement system

Antibodies to Ebola virus glycoprotein (EBOV GP) represent an important correlate of the vaccine efficiency and infection survival. Both neutralization and some of the Fc-mediated effects are known to contribute the protection conferred by antibodies of various epitope specificities. At the same time, the role of the complement system in antibody-mediated protection remains unclear. In this study, we compared complement activation by two groups of representative monoclonal antibodies (mAbs) interacting with the glycan cap (GC) or the membrane-proximal external region (MPER) of the viral sole glycoprotein GP. Binding of GC-specific mAbs to GP induced complement-dependent cytotoxicity (CDC) in the GP-expressing cell line via C3 deposition on GP in contrast to MPER-specific mAbs that did not. Moreover, treatment of cells with a glycosylation inhibitor increased the CDC activity, suggesting that N-linked glycans downregulate CDC. In the mouse model of EBOV infection, depletion of the complement system by cobra venom factor led to an impairment of protection exerted by GC-specific but not MPER-specific mAbs. Our data suggest that activation of the complement system is an essential component of antiviral protection by antibodies targeting GC of EBOV GP.


Introduction
Filoviruses include one of the deadliest human pathogens known to date. Ebolavirus genus of the Filoviridae family includes Ebola virus (EBOV), Sudan virus (SUDV), Bundibugyo virus (BDBV), Taï Forest virus (TAFV), Reston (RESTV) and Bombali virus (BOMV) [1]. EBOV, SUDV and BDBV are known to cause outbreaks and epidemics of highly lethal disease, which is often accompanied by hemorrhagic manifestations and systemic multiorgan dysfunction, with unpredictable periodicity, location, and scale potential in vivo by introduction of mutations regulating the activation of complement. Speci cally, it was shown that KWES set of amino acid mutations in an Fc-fragment of VIC16 mAb, which upregulates complement activation, improves the protection of mice from EBOV infection compared to unmodi ed mAb [25]. However, the direct requirement of complement for mAb-mediated protection against lovirus infections has not been demonstrated.
In the present study, we compared complement activation by ebolavirus GP-speci c mAbs with different epitope speci cities. First, using an antibody-dependent complement deposition (ADCD) assay, GCspeci c mAbs were shown to better induce C3 deposition compared to MPER mAbs. Second, we developed a complement-dependent cytotoxicity (CDC) assay and demonstrated that GC-speci c mAbs stimulate killing of the target antigen-expressing cells by complement, an activity that can be inhibited by mAbs recognizing the other parts of GP, such as the MPER or base region. Using the chemical inhibitor of N-linked glycosylation, we further showed that N-linked glycans on the GP surface, while serving as part of the mAb epitope, can nevertheless downregulate mAb-mediated CDC activity. This nding represents a previously unknown mechanism of evasion of antiviral complement activity employed by EBOV. Finally, the depletion of complement in mice by injection of the cobra venom factor (CVF) impaired the survival of EBOV-challenged animals upon treatment with some mAbs speci c to GC, but not to the MPER, indicating requirement of the functional complement system for effective protection by GC-speci c mAbs. These results contribute understanding the mechanisms of virus-complement interplay and highlight an important role of the complement system in anti-EBOV activity of GP-speci c mAbs. The obtained data can inform the selection of GC-speci c mAbs for improved therapeutic antibody combinations.

Results
Glycan cap mAbs are more potent complement activators compared to MPER mAbs. From our previously published studies on human mAbs isolated from survivors of ebolavirus infection [26,34,37,[44][45][46], we selected a panel of neutralizing GC-or MPER-speci c mAbs to determine a possible difference in complement activation between these two groups of antibodies. A well-characterized 13C6 mouse mAb [42], which is a component of MB-003 and ZMapp combinations against EBOV and known to neutralize virus only in the presence of complement [20,27], and mAbs ADI-15820 [35,36] and KZ52 [47] isolated from human survivors of EBOV infection, also were included in this panel. For some antibodies, the recombinant versions bearing K322A (KA) mutation in the Fc fragment were produced, since this mutation greatly reduces binding to C1q resulting in the lack of e cient activation of complement [48].
First, we measured a dose-dependent mAb ability to induce C3 deposition onto GP-coated beads in ADCD assay (Fig. 1A, B). All the mAbs tested, with the exception of rBDBV223-IgG3 and -IgG4, belonged to the IgG1 subclass, allowing us to dissect the role of the epitope speci city in the activity. High levels of C3 deposition were observed for the GC-speci c mAbs, whereas most of the tested mAbs speci c to MPER, EBOV GP base region (KZ52), or irrelevant target (mAb (2D22; speci c to dengue virus envelope protein [49]) did not show activity. We next asked whether the observed mAb effects on C3 deposition translated into the complement-mediated killing of antigen-expressing cells. For this work, we developed a CDC assay using the EBOV GPkik-293FS EGFP CCR5-SNAP cell line [50]. This cell line constitutively expresses EBOV GP on the plasma membrane, EGFP in the cytoplasm and the SNAP-tag CCR5, which can be speci cally labeled with SNAP-Surface Alexa Fluor 647, on the cell surface. Fluorophore-labeled cells were incubated consecutively with the mAbs and complement, and the cytotoxicity was Quanti ed as a percentage of EGFP -AF647 + cells [51,52] by analytical ow cytometry (Fig. 1C). Consistent with the data in Fig. 1A, B, CDC activity was observed for some of the GC-speci c mAbs and with the hybridomaproduced and recombinant (rBDBV223-IgG3) versions of BDBV223 mAb, both belonging to the IgG3 subclass.
The speci city of the CDC assay was further validated in separate experiments using selected mAbs with high CDC activities. When complement was pre-treated with zymosan A, which was expected to consume the complement system activity [53], the cytotoxicity was signi cantly reduced for all antibodies regardless of the epitope speci city or IgG subclass (Fig. 1D, left). These results suggest that the cell killing activities observed for mAbs in the developed assay depends speci cally on the presence of intact complement. Conversely, when complement was pre-treated with antibody 1E2 against the mannosebinding lectin (MBL), changes in mAb activity were not detected (Fig. 1D, middle). These data also were con rmed when tissue factor pathway inhibitor (TFPI) was added to cells along with the mAbs (Fig. 1D, right). TFPI is known as a selective inhibitor of MASP-2 serine protease of the lectin pathway, which does not affect the classical pathway proteases C1s or C1r [54]. Altogether, our data demonstrate that the observed mAb-driven cytotoxicity results from activation of the classical complement pathway.
MPER-and base-region-speci c mAbs block cytotoxicity induced by GC or MPER mAbs. We next tested if antibodies of various epitope speci cities could interact to block or synergize each other's CDC activities. In the rst experiment, mixtures of GC-speci c EBOV90 mAb with different concentrations of MPER (rBDBV223-IgG1), base (KZ52) or an irrelevant isotype-control antibody 2D22 were added to target cells ( Fig. 2A). The CDC activity (the percentage of EGFP − AF647 + cells) of EBOV90 was dose-dependently inhibited by BDBV223 and KZ52, but not by 2D22, indicating that inactive MPER and base mAbs can reduce activities of GC-speci c mAbs. At the same time, when BDBV223 was added to the GC mAb with low CDC activity (BDBV270), it did not change the CDC activity ( Fig. 2A, controls).
In the second experiment, mixtures of the hybridoma-derived BDBV223 MPER mAb (IgG3 subclass) with different concentrations of a GC-speci c (BDBV270), base-speci c (KZ52) or irrelevant 2D22 mAb were tested (Fig. 2B). The CDC activity of BDBV223 was inhibited by KZ52, but not by BDBV270, in a dosedependent manner. Surprisingly, the percentage of EGFP − AF647 + cells following incubation witth BDBV223 (IgG3) also was decreased when 50 µg/mL of 2D22 was added, suggesting some non-speci c inhibition of the activity of BDBV223 by a maximal tested dose of an irrelevant antibody. When we mixed two highly active GC-speci c mAbs of different epitope speci cities, EBOV90 and MPER-speci c BDBV223 (IgG3). no change in the CDC activity over the GC-speci c antibody mixed with the 2D22 control mAb was observed (Fig. 2B, controls). The results of these experiments show that: 1) MPER-speci c mAbs can inhibit CDC activity of the GC-speci c mAbs, but not vice versa; 2) base-region-speci c mAbs can inhibit the activity induced by both GC-and MPER-speci c mAbs; and 3) mAbs targeting different EBOV GP epitopes do not show synergistic CDC activity.
N-linked glycans on EBOV GP prevent CDC. EBOV GP is a heavily glycosylated protein [8,55], which can affect multiple biological properties of the virus [56]. We next explored a possible role of N-glycans on EBOV GP in modulating mAb-induced CDC activity. First, we tested if pre-treatment of cells with tunicamycin, a chemical inhibitor of N-linked glycosylation [57], would alter mAb binding to EBOV GP. Tunicamycin treatment resulted in a signi cant decrease of mAb binding to 293F cells expressing EBOV GP, except for the BDBV317 MPER-speci c mAb ( Fig. 3A). At the same time, this treatment did not have a detectable effect on GP expression on the surface of target cells (Fig. 3B), suggesting that an impairment of mAb binding was not due to a reduction of GP expression level caused by tunicamycin. Interestingly, when the CDC assay was run for mAbs following tunicamycin treatment of cells, the opposite effect was observed: an increase, rather than decrease of activity, for some of the tested mAbs, regardless of the epitope speci city (Fig. 3C). Therefore, even though N-deglycosylation of GP disfavors mAb binding, it nevertheless results in hyperactivation of the complement-mediated lysis of target cells induced by mAbs. Overall, these data show that N-linked glycans on EBOV GP protect cells from CDC. GC-speci c, but not MPER-speci c mAbs, require complement for in vivo protection against EBOV. Finally, we addressed the role of the complement system in mAb-mediated protection against EBOV in vivo. Groups of BALB/c mice were treated with CVF to deplete their complement system, or mock-treated and next day exposed to a lethal (1,000 PFU) dose of mouse-adapted EBOV. On day 1 after infection mice were treated with individual mAbs at 100 µg (∼5 mg/kg) or mock-treated, and on day 3 after infection, treatment or mock-treatment with CVF was repeated. For two out of three GC-speci c mAbs tested, administration of CVF caused a signi cant impairment of protection (Fig. 4). Notably, for one of these mAbs, BDBV270, the protection was completely abrogated by CVF, as no animals in BDBV270/CVF group survived the infection. In contrast, full protection of mice against EBOV challenge was achieved by all MPER-speci c mAbs tested, regardless of the CVF treatment. These data suggest that activation of the complement system is an important antiviral mechanism, which is required for in vivo protection conferred by mAbs targeting the GC but not the MPER of EBOV GP.

Discussion
Using a combination of in vitro and in vivo approaches, we investigated the role of the complement system in antiviral mechanisms employed by antibodies directed to EBOV GP. First, using a bead-based ADCD assay, we compared the ability of two mAb groups with different epitope speci cities to induce the C3 deposition on GP-coated surface. GC-speci c mAbs were shown to be superior to their MPER-speci c counterparts in ADCD activity (Fig. 1A, B). These data are in line with the results of a previous study, which analyzed multiple functional activities for 168 EBOV GP mAbs. Notably, antibodies targeting the most exposed GP regions, such as the head, GC and MLD, demonstrated stronger engagement of Fceffector functions compared to mAbs against the conformationally obscured, "hidden" epitopes (i.e., HR2/MPER, IFL) [18,23]. This observation was hypothesized to result from a greater accessibility of Fc fragments of mAbs bound to outer GP regions for the interaction with Fc receptors at the surface of immune cells, or with the complement system components [58].
Deposition of complement can lead to formation of MAC and lysis of lipid membranes of enveloped viruses [59] or infected host cells expressing viral antigens [60]. Deposition complement on virion particles may contribute to direct elimination of viral particles but probably is not critical for protection by mAbs that neutralize virus without complement [34,37]. Elimination of infected cells by complement enhanced mechanisms, however, is more likely to reduce total viral burden. To test if the observed mAb ability to induce the C3 deposition on cells would result in an increased mAb-mediated cytotoxicity, we developed a CDC assay using a human-origin (human embryonic kidney 293F) cell line constitutively expressing EBOV GP [50] (Fig. 1C). The relative activity of individual mAbs in the CDC assay was similar to that determined by the ADCD assay, con rming the functional relevance of C3 deposition. High activity also was detected for an IgG3 form of the BDBV223 MPER mAb, the only antibody for which other subclasses in addition to IgG1 were tested. These results also were con rmed by ELISA (Fig. 1E). It is known that Fc-mediated activities vary greatly among IgG subclasses. The amino acid sequence of the C H 2 region [61] and the antibody hinge region length [62] determines the complement-xing potential of antibodies. IgG3 has the most potent a nity for binding to C1q, followed by IgG1, with a very weak association for IgG2 and no detectable interaction for IgG4 [40,63]. From that perspective, BDBV223-IgG3 serves as a positive control in the tested panel.
The speci city of the 293F-cell-based CDC assay we developed was validated with complement-depleting or pathway-inhibiting compounds. First, using zymosan A, we showed that the mAb-mediated cytotoxicity requires the presence of intact complement. Zymosan is a carbohydrate substance extracted from yeast cell walls and is a potent activator of the alternative complement pathway. Zymosan can directly interact with properdin, the regulatory plasma glycoprotein produced by neutrophils which forms the stabilizing complex with C3bBb convertase (C3bBbP). After non-covalent attachment to the surface of zymosan particles, properdin binds C3b and initiates assembly of C3bBbP complexes, facilitating the prompt depletion of C3 complement component by the ampli cation convertase [64]. Second, using TFPI and anti-MBL antibody, we demonstrated that the observed CDC activity results from activation of the classical, but not lectin, complement pathway and, therefore, speci cally requires the presence of GPspeci c antibodies (Fig. 1D).
Next, considering the difference in complement activation by GC-and MPER-speci c mAbs, we questioned the possible biological outcome of interaction between these two antibody groups, as should normally happen in a context of a polyclonal antibody response to EBOV infection and/or vaccination. We found that the KZ52 base-speci c mAb dose-dependently inhibited CDC activity of the GC-speci c mAb EBOV90 and the MPER mAb BDBV223 (IgG3) mAbs, and that BDBV223 (IgG1) mAb inhibited the activity of EBOV90 (Fig. 2). GC-speci c antibodies enhanced binding of base mAbs to EBOV GP [31,32,37,65]. Presumably, the EBOV90/KZ52 combination favors binding of the complement-inactive KZ52 mAb, therefore shifting the balance towards low/no complement activation. Regarding the BDBV223/KZ52 and EBOV90/BDBV223 pairs, it is possible that binding of one mAb can interfere with the binding of another one by a partial stochiometric hindrance of its epitope and/or by inducing conformational changes in the GP. Another possibility is that an inactive mAb interacts more strongly with GP, simply because of higher a nity for binding compared to that of the second mAb, thus creating unfavorable conditions for complement activation. The latter scenario can be especially true for the EBOV90/BDBV223 combination, considering the high reactivity of BDBV223 to EBOV GP (EC 50  . GP1 Nglycans are suggested to participate in an immune evasion by shielding the epitopes from antibody recognition [55]. The GP2 subunit contains two N-linked glycosylation sites that contribute GP expression, stability, and cell entry [70,71]. Given the important role of N-linked glycans in virus structure and life cycle, we addressed their possible effect on the mAb CDC activity using tunicamycin. Initially identi ed as a natural antibiotic, tunicamycin is now widely used for blocking N-linked glycosylation by inhibiting the transfer of UDP-N-acetylglucosamine to dolichol phosphate in the endoplasmic reticulum of eukaryotic cells [72][73][74]. For most of the tested mAbs, tunicamycin treatment reduced binding to GP (Fig. 3A). This nding was unexpected, since removal of glycans by mutagenesis was shown to enhance sensitivity of vesicular stomatitis virus (VSV) pseudotyped with EBOV GP (VSV/EBOV-GP) to neutralization by whole IgG puri ed from the serum of vaccinated or convalescent cynomolgus macaques [55,70]. However, our data suggest that, for certain mAbs, N-linked glycans may be a part of their epitopes, rather than shielding the epitopes from immune recognition. In particular, the epitopes for mAbs 13C6 and BDBV289 contain N238 and N268 glycans, and the EBOV293 mAb epitope contains an N268 glycan [69]. Interestingly, removal of the N563 glycan site by mutagenesis enhanced VSV/EBOV-GP neutralization for some mAbs, while impairing it for the other mAbs [36]. In our panel, the only mAb that demonstrated an increased GP binding in the presence of tunicamycin was BDBV317. Its epitope can be shielded by a N618 glycan, which is located close to the escape mutation site identi ed for this mAb [44]. It should be noted that tunicamycin treatment does not allow dissection of the role of speci c glycans, which can be studied in part by site-directed mutagenesis approaches [36, 55,70,71]. However, tunicamycin treatment has the advantage that it does not change the amino acid residue at the site of glycosylation, minimizing a possible impact on GP expression (Fig. 3B). Surprisingly, tunicamycin treatment not only did not reduce the CDC activity, as one could expect based on mAb binding data (Fig. 3A), but, instead, caused an increase of the cytotoxicity for some of the tested mAbs (Fig. 3C). To our knowledge, the phenomenon of speci c downregulation of the classical complement pathway by N-linked glycans at the viral surface has not been described. A possibility is that EBOV employs GP glycosylation to reduce the antiviral complement activity.
Finally, we selected a few available GC-and MPER-speci c mAbs, for which we have previously reported protection in vivo [26, 34,37,45], and addressed the role of the complement system using CVF treatment in the mouse model of EBOV infection. CVF shares structural and functional properties with C3. It also has C3b-like activity in forming the extremely stable CVF-dependent convertase, CVF,Bb, which cleaves C3 and C5 components [75]. Treatment of BALB/c mice with CVF depletes complement [76,77]. We showed that CVF treatment signi cantly impaired the protection conferred by the GC-speci c mAbs EBOV293 and BDBV270 but did not impair the protection by BDBV289 mAb or any MPER-speci c mAb (Fig. 4). Interestingly, although EBOV293 and BDBV270 initiated C3 deposition (Fig. 1A, B), they did not demonstrate CDC activity (Fig. 1C). These data suggest that C3 opsonization mediated by these mAbs can potentially trigger complement-dependent antiviral mechanisms other than MAC assembly and lysis of antigen-expressing cells. Such alternative mechanisms can include phagocytosis of opsonized targets by complement receptor-bearing cells [39,78]. BDBV289 is more potently neutralizing compared to mAbs BDBV270 and EBOV293 [34,37], and was shown protect in mouse, guinea pig [34] and rhesus macaque [79] models of ebolavirus infection. It is therefore possible that BDBV289-mediated protection relies mainly on Fab-dependent virus neutralization and does not require complement activation. Similarly, MPER-speci c antibodies can protect through direct virus neutralization, likely by interfering with the viral fusion machinery [80].
Ebolaviruses continue to pose a signi cant threat to public health by inducing outbreaks and epidemics of a highly lethal disease. Passive immunotherapy remains the most reliable therapeutic option for prophylaxis and post-exposure treatment. Therefore, understanding of protective mechanisms used by antibodies is critical to inform development of the most effective immunotherapeutic regimens and design of vaccines. In the present study, we addressed the antiviral mechanism for GC-binding mAbs.
Using in vitro assays, we showed that 1) GC mAbs are superior to MPER mAbs in complement activation; 2) CDC activity can be dose-dependently inhibited by complement-inactive mAbs of different epitope speci city; 3) N-linked glycans can serve as a part of a mAb epitope and 4) N-linked glycans greatly downregulate CDC activity. Moreover, we were able to directly demonstrate the requirement of intact complement for in vivo protection conferred by GC-speci c mAbs. Altogether, our results highlight the previously underappreciated role for activation of the complement system as an important mechanism of antibody-mediated protection against EBOV.

Materials And Methods
Cell lines. 293F cells expressing EBOV GP (strain Kikwit) on the plasma membrane, EGFP in the cytoplasm and the SNAP-tag CCR5 on the cell surface [50] were kindly provided by Dr. George K. Lewis (University of Maryland). The cell suspension was maintained in FreeStyle™ 293 expression medium (Gibco) containing 1 µg/mL puromycin (InvivoGen) at 37°C in 8% CO 2 shaken at 130 rpm. Vero-E6 cells (green monkey kidney epithelial) were obtained from ATCC (CRL-1586). Cells were maintained in minimum essential medium supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin solution (Gibco) at 37°C in 5% CO 2 . Production of hybridoma-derived and recombinant mAbs. Hybridoma mAbs EBOV90 (IgG1 isotype), BDBV270 (IgG1 isotype), EBOV293 (IgG1 isotype), BDBV317 (IgG1 isotype), BDBV223 (IgG3 isotype), and BDBV289 (IgG1 isotype) were isolated from a human survivor of a natural EBOV or BDBV infection as described previously [34,37]. MAb 2D22 (IgG1 isotype) that is speci c to dengue virus envelope (E) protein was described previously [82]. MAbs EBOV293, BDBV43, BDBV270, EBOV402, BDBV223, BDBV317 and mAbs ADI-15820 and KZ52 were produced in mammalian Expi293F or ExpiCHO cells (Gibco). ADI-15820 and KZ52 were produced based on known heavy-and light-chain variable region genes for these mAbs. Antibody heavy-and light-chain variable region genes were sequenced from hybridoma lines that had been cloned biologically by ow cytometric sorting. Brie y, total RNA was extracted using the RNeasy Mini kit (QIAGEN) and reverse-transcriptase PCR (RT-PCR) ampli cation of the antibody gene cDNAs was performed using the PrimeScript One Step RT-PCR kit (Takara Bio Inc.) according to the manufacturer's protocol with gene-speci c primers [83]. The thermal cycling conditions were as follows: 50°C for 30 min, 94°C for 2 min, 40 cycles of (94°C for 30 s, 58°C for 30 s and 72°C for 1 min). PCR products were puri ed using Agencourt AMPure XP magnetic beads (Beckman Coulter) and sequenced directly using an ABI3700 automated DNA analyzer. For recombinant mAb production, cDNA encoding the genes of heavy and light chains were cloned into DNA plasmid monocistronic expression vectors for mammalian cell culture mAb secretion encoding IgG1-, IgG3, IgG4, or IgG1-KA -heavy chain [84] and transformed into Escherichia coli cells. This vector contains an enhanced 2A sequence and GSG linker that enables simultaneous expression of mAb heavy-and light-chain genes from a single construct after transfection. MAb proteins were produced following transiently transfection of Expi293F or ExpiCHO cells following the manufacturer's protocol and were puri ed from ltered culture supernatants by fast protein liquid chromatography on an ÄKTA instrument using HiTrap MabSelect Sure or HiTrap Protein G columns (GE Healthcare). Puri ed mAbs were buffer exchanged into phosphate buffered saline (PBS), ltered using sterile 0.45-µm pore size lter devices (Millipore), concentrated, and stored in aliquots at -80°C until use. Puri cation of hybridoma-produced mAbs is described elsewhere [85].
Analysis of mAb IgG subclass speci city. The isotype and subclass of secreted antibodies were con rmed by ELISA using murine anti-human IgG1, IgG3 or IgG4 mouse antibodies conjugated with alkaline phosphatase (Southern Biotech).  to verify viral titer. Mice (n = 5) were treated i.p. with 20 µg (or approximately 1 unit) of CVF (Sigma-Aldrich) in 500 µL PBS or mock-treated at one day prior to and three days after the challenge, and with 100 µg (~ 5 mg/kg) of individual mAb in 100 µL PBS on day 1 post-challenge. Mice were monitored twice daily from day 0 to day 14 post-challenge for illness, survival, and weight loss, followed by once daily monitoring from day 15 to the end of the study at day 28, as described elsewhere [86]. Moribund mice were euthanized as per the approved protocol (see Ethics statement). All mice were euthanized on day 28 after EBOV challenge.

Antibody
Statistical analysis. Statistical analyses and generation of graphs were performed using GraphPad Prism version 6.07 (GraphPad Software). One-way ANOVA with multiple comparisons (Tukey's test) or a T-test were used for statistical data analysis. Animal survival data were analyzed by log-rank (Mantel-Cox) test.   were treated as in (A), and CDC assay was performed as in Fig. 1C. The percentages of EGFP-AF647+ cells in samples treated with the vehicle control or tunicamycin and incubated with 2D22 mAb were used for background signal subtraction. Mean ± SEM of triplicate samples are shown. *p 0.01; **p 0.001; ***p 0.0001; ns, not signi cant (unpaired t-test). Representative ow cytometry dot plots are shown. Figure 4