A Fc-Engineering approach to define functional humoral correlates of immunity against Ebola virus

Summary

Protective Ebola virus (EBOV) antibodies have neutralizing activity and induction of Fc-mediated innate immune effector functions. Efforts to enhance Fc-effector functionality often focus on maximizing antibody-dependent cellular cytotoxicity, yet distinct combinations of functions may be critical for antibody-mediated protection. As neutralizing antibodies have been cloned from EBOV disease survivors, we sought to identify survivor Fc-effector profiles to help guide Fc-optimization strategies. Survivors developed a range of functional antibody responses, and we therefore applied a rapid, high-throughput Fc-engineering platform to define the most protective profiles. We generated a library of Fc-variants with identical Fabs from an EBOV neutralizing antibody. Fc-variants with antibody-mediated complement deposition and moderate NK cell activity demonstrated complete protective activity in a stringent in vivo mouse model. Our findings highlight the importance of specific effector functions in antibody-mediated protection and the experimental platform presents a generalizable resource for identifying correlates of immunity to guide therapeutic antibody design.

eTOC Blurb.

Gunn et al. profile Ebola virus disease survivors and apply a platform for engineering antibody Fc domains to define protective profiles. Fc-variants with complement deposition, yet moderate NK cell activity completely protected infected mice from disease. This experimental platform can be used for identifying correlates of immunity to other pathogens, and to guide therapeutic antibody design.

Graphical Abstract

Introduction

Monoclonal antibodies (mAbs) are among the fastest growing class of drugs for the treatment of cancer, autoimmunity, and infectious diseases (Walker and Burton, 2018; Weiner et al., 2009). The critical role of the antibody constant domain (Fc), both in extending half-life as well as in driving immune killing, has inspired engineering efforts that fully exploit the biological potential of these unique therapeutics. Engineering efforts including modification of Fc-glycosylation and the introduction of point mutations that collectively shape the affinity of Fc binding to a wide array of canonical and non-canonical Fc-receptors or complement provide a means to drive targeted immune clearing activities. For example, alterations in the Fc have been developed to selectively increase antibody-dependent cellular cytotoxicity (ADCC) (Lazar et al., 2006; Shields et al., 2001), antibody-dependent cellular phagocytosis (ADCP) (Lazar et al., 2006; Richards et al., 2008), and antibody-dependent complement deposition (ADCD) (Diebolder et al., 2014; Steurer et al., 1995). Although Fc modifications have been successfully exploited in the context of infectious diseases (Qiu et al., 2014), in some instances antibody Fc-effector function has been linked to both pathological and protective functions, indicating that delicate selection of Fc modifications is needed to maintain balanced Fc-activity. For example, antibody-dependent enhancement (ADE) has been linked to susceptibility to disease caused by dengue virus (Dejnirattisai et al., 2010; Wang et al., 2017), respiratory syncytial virus (Polack et al., 2002), and coronaviruses (Wan et al., 2020). Conversely, immuno-protective functions, including ADCC, have also been noted for these pathogens (Sun et al., 2019; van Erp et al., 2019), again highlighting the need for precise manipulation of Fc-effector function. Moreover, emerging correlate analyses across infectious diseases now provide a unique opportunity to define mechanistic antibody correlates of immunity indicating that discrete profiles of innate immune effector function, rather than pan-functional activity (Ackerman et al., 2018; Gunn et al., 2018), may be needed to guide therapeutic engineering to maximize protection.

In the context of Ebola virus infection, both neutralizing activity and Fc-mediated effector functions are linked to mAb-mediated protection (Gunn et al., 2018; Saphire et al., 2018a; Zeitlin et al., 2011). However, the precise effector functions that track with protection have yet to be defined. Promising mAbs with potent neutralizing activity have been isolated from human survivors of Ebola virus disease (EVD) (Bornholdt et al., 2016; Corti et al., 2016; Flyak et al., 2016; Maruyama et al., 1999), but the characterization of unique Fc-functional profiles of these antibodies and how these functions are linked to protection is only just starting. In this study we sought to link EVD survivor antibody profiling to Fc-engineering to accelerate our mechanistic understanding of protective immune responses that can guide therapeutic design.

The polyfunctional antibody profiles of nearly all EVD survivors are marked by detectable levels of at least one and often many antibody effector functions. To selectively map the precise antibody effector profiles associated with maximal protection from disease, we used a high-throughput platform for rapid, parallel generation of a panel of Fc-engineered mAbs that represent the diversity of functional profiles observed in human survivors. Using the mAb VIC16, which is neutralizing in vitro but offered only modest protection in an animal model Ebola virus infection (Saphire et al., 2018a), a subset of Fc-modifications were selected to represent EVD survivor profiles and tested in a stringent in vivo mouse model for the ability to enhance protection against Ebola virus, allowing the precise identification of Fc-effector functions - that coupled to neutralization- are required to achieve protective immunity against Ebola virus. Our findings reveal that antibodies having complement activity and moderate, rather than robust, NK cell activation were most protective, conferring complete protection from death and eliminating symptoms of infection. These data highlight the utility of coupling unbiased Fc-engineering approaches to complement correlate analyses and accelerate the development of highly protective antibodies to Ebola virus. This platform can likely be extended to engineer antibodies against other deadly pathogens.

Results

Diversity in antibody-mediated innate immune effector functions across human survivors of EVD.

Emerging correlates analyses using libraries of therapeutic mAbs pointed to the importance of phagocytic functions and NK cell activation as key functional properties that can confer protection in mice (Gunn et al., 2018; Saphire et al., 2018a). Given that limited evidence exists for reinfection after EVD survival, we hypothesized that the persisting humoral immune response in survivors points to specific functional humoral immune responses associated with protection. Thus, we profiled the functional humoral immune response in a group of 40 confirmed human survivors of EVD and 26 household contacts collected in Sierra Leone between 2018 and 2019, or ~3–4 years after acute Ebola virus infection to represent the long-lived humoral response. Complete humoral profiling was performed for each sample to measure 66 different features (Figure S1). These measurements captured: induction of six different innate immune effector functions by Ebola glycoprotein (GP)-specific antibodies (ADCP, ADCD, ADNP, ADNKA CD107a, IFNγ, and MIP-1β), levels of IgG, IgA, and IgM binding to Ebola GP, soluble GP (sGP), nucleoprotein (NP), VP40, and VP24, and binding of Ebola GP, sGP, NP, VP40, and VP24-specific IgG to FcγRs (FcγR3A, FcγR2A) and FcαR.

Similar to the profiles previously observed across EVD survivors at 6 months following infection (Gunn et al., 2020), heterogeneous functional profiles were observed for antibodies isolated from survivors and their household contacts at 36–48 months after infection (Figure S1). Unsupervised hierarchical clustering of survivors and contacts based on effector function alone demonstrated that a majority of survivors clustered separately from contacts, with only 3 misclassified survivors and contacts, which highlights the durable presence of antibody effector functions in nearly all survivors (Figure 1A). Among these durable humoral immune responses, antibody-dependent complement activation was least common among the survivors, whereas antibody dependent phagocytosis (ADCP) and NK cell activating antibodies (ADNKA) were present in nearly all survivors. Using K-means clustering, ten major functional clusters were defined within the survivor profiles (Figure 1B; left) that represented unique combinations of functional antibody profiles (Figure 1B; right). A subset of antibodies from survivors induced high levels of all effector functions (groups 1, 6, and 10). Conversely, the majority of survivors exhibited more variable combinations that were marked by low-to-intermediate NK cell function accompanied by at least one type of function related to phagocytosis or complement. Importantly, although emerging data point to a critical role for NK cells in control of EBOV infection (Bornholdt et al., 2019; Corti et al., 2016; Gunn et al., 2018; Warfield et al., 2004), the results of this study suggest that NK cell function does not occur alone.

Figure 1. Diversity in antibody-mediated innate immune effector functions across human survivors of EVD.

A. Unsupervised hierarchical clustering of cluster EVD survivors (S-value; black text); household contacts (C-value; red text) and a USA-based healthy control (blue text) based on induction of antibody-mediated innate immune effector functions. The heatmap shows the relative value of the indicated functions with purple and green representing the maximum and minimum values, respectively, across the samples as indicated in the legend. A heatmap of all measured antibody features are shown in Figure S1.

B. K-means clustering of 40 EVD survivors into 10 distinct clusters. Principal component analysis of mAbs within each cluster is shown on the left. Composite functional flower plots using the mean values for each cluster were overlaid onto the mirrored PCA on the right graph. Each segment represents a different function, and the size of the segment represents the magnitude of the function relative to the entire panel as indicated in the inset legend. Cluster 9 has no detectable functional activity.

Fc-engineering to define mechanisms of protection.

Given the importance of Fc-effector function for mAb-mediated protection against Ebola virus challenge in animal models (Saphire et al., 2018b), we next performed Fc-engineering of mAbs to gain a more granular understanding of which antibody effector functions are most critical for protection. For Fc-engineering we selected VIC16, a mAb that was part of a 171-antibody panel for which antibody features that correlated with protection were extensively characterized as part of the Viral Hemorrhagic Fever Immunotherapeutics Consortium (VIC) (Saphire et al., 2018a). VIC16 conferred moderate protection in a mouse model of infection with the Zaire strain of Ebola virus (EBOV) (Figure S2A), despite having good neutralization of EBOV in cell culture (Saphire et al., 2018a) through the ability to bind to the EBOV GP internal fusion loop, a key neutralizing epitope targeted by some of the most protective mAbs known to date (Wec et al., 2017; Zhao et al., 2017). Compared to other VIC panel mAbs that bound the internal fusion loop, VIC16 alone exhibited limited to no antibody effector function (Figure S2B). Thus, we hypothesized that generating a library of VIC16 Fc-variants, able to broadly confer distinct Fc-effector functions found in EVD survivors, would provide an opportunity to rescue, improve, and define the minimal requisite antibody effector functions required to achieve protection against lethal challenge with Ebola virus, providing critical clues related to the correlates of humoral immunity to Ebola virus.

A high-throughput cloning approach to rapidly generate Fc-variants: REFORM

Structure-guided mutational screens have identified mutations in both the CH2 and CH3 domains of the antibody Fc domain that alter binding to Fcγ–receptors (FcγRs) (Chu et al., 2008; Lazar et al., 2006; Moldt et al., 2011; Moore et al., 2010; Richards et al., 2008; Shields et al., 2001; Smith et al., 2012; Stavenhagen et al., 2007; Steurer et al., 1995; Xu et al., 2000), complement (Diebolder et al., 2014; Idusogie et al., 2000; Idusogie et al., 2001; Moore et al., 2010; Steurer et al., 1995), and/or the neonatal Fc-receptor (FcRn) (Dall’Acqua et al., 2002; Datta-Mannan et al., 2007; Hinton et al., 2004; Yeung et al., 2009; Zalevsky et al., 2010), resulting in either enhancement or ablation of different innate immune effector functions or antibody half-life. To generate a library of distinct IgG1 Fc domains encoding different Fc mutations, we used a high-throughput cloning platform, Golden Gate (Engler et al., 2008), which takes advantage of unique overhangs generated by digestion with the Type IIS endonuclease BsaI, to ligate all antibody fragments into a pUC vector in a single reaction that avoids the need for sequential cloning of individual antibody domains (variable heavy chain, variable light chain, hinge, constant heavy chain) required in traditional cloning approaches. This method, termed Rationally Engineered and Functionally Optimized Monoclonal antibodies (REFORM), allowed us to rapidly generate antibody plasmids encoding each of the 61 different Fc variants with the same Fab binding domain. The VIC16 variable heavy and light chains were used to generate a panel of VIC16 Fc variants (Table 1). The antibody plasmids were transfected into suspensions of 293F mammalian cells and supernatants were collected 7 days post-transfection for purification of mAbs using purified using protein A/G beads. The expression of each modified mAb was confirmed by SDS-PAGE, which showed that the heavy and light chains had the expected size (Figure S3A–B). Production of some mAbs was poor and these were subsequently excluded from downstream analysis. FPLC analysis of the mutated mAbs produced a single peak indicating that monomeric IgG antibody was obtained (Figure S3C).

Table 1. List of REFORM variants generated for this study.

The specific amino acid mutations are shown, and the published impact on antibody-dependent cellular cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP), antibody-dependent complement deposition (ADCD) or binding to FcRn is indicated.

Fc Variant	AA Mutations	Description	Ref.
IgG1		Human IgG1
LALA	L234A/L235A	no FcγR binding
LALALS	L234A/L235A/M428L/N434S	no FcγR binding ↑FcRn	(Zalevsky et al., 2010)
LALAPG	L234A/L235A/P329G	no FcγR binding
N297Q	N297Q	Aglycosylated
N297QLS	N297Q/M428L/N434S	Aglycosylated; ↑FcRn	(Zalevsky et al., 2010)
SAEAKA	S298A/E333A/K334A	↑ ADCC	(Lazar et al., 2006; Shields et al., 2001)
SAEAKALS	S298A/E333A/K334A/M428L/N434S	↑ ADCC ↑FcRn	(Lazar et al., 2006; Shields et al., 2001; Zalevsky et al., 2010)
LPLIL	F243L/R292P/Y300L/V305I/P396L	↑ ADCC	(Stavenhagen et al., 2007)
LPLILLS	F243L/R292P/Y300L/V305I/P396L; M428L/N434S	↑ ADCC ↑ FcRn	(Stavenhagen et al., 2007; Zalevsky et al., 2010)
E345R	E345R	↑ ADCD	(Diebolder et al., 2014)
E345RLS	E345R/M428L/N434S	↑ ADCD ↑ FcRn	(Diebolder et al., 2014; Zalevsky et al., 2010)
HFST	H268F/S324T	↑ ADCD	(Moore et al., 2010)
HFSTLS	H268F/S324T/M428L/N434S	↑ ADCD ↑FcRN	(Moore et al., 2010; Zalevsky et al., 2010)
S324T	S324T	↑ ADCD	(Moore et al., 2010)
S324TLS	S324T/M428L/N434S	↑ ADCD ↑ FcRn	(Moore et al., 2010; Zalevsky et al., 2010)
RGY	E345R/E430G/S440Y	↑ ADCD	(Diebolder et al., 2014)
RGYLS	E345R/E430G/S440Y/M428L/N434S	↑ ADCD ↑ FcRn	(Diebolder et al., 2014; Zalevsky et al., 2010)
SEHFST	S267E/H268F/S324T	↑ ADCD	(Moore et al., 2010)
SEHFSTLS	S267E/H268F/S324T/M428L/N434S	↑ ADCD ↑ FcRn	(Moore et al., 2010; Zalevsky et al., 2010)
KWES	K236W/E333S	↑ ADCD	(Steurer et al., 1995)
KWESLS	K236W/E333S/M428L/N434S	↑ ADCD ↑ FcRn	(Steurer et al., 1995; Zalevsky et al., 2010)
E333A	E333A	↑ ADCD, ↑ADCC	(Idusogie et al., 2000; Idusogie et al., 2001)
I332E	I332E	↑ ADCC ↑ADCP	(Lazar et al., 2006)
I332ELS	I332E/M428L/N434S	↑ ADCC ↑ADCP ↑ FcRn	(Lazar et al., 2006; Zalevsky et al., 2010)
IEGA	G236A/I332E	↑ ADCC ↑ADCP	(Richards et al., 2008)
IEGALS	G236A/I332E/M428L/N434S	↑ ADCC, ↑ADCP, ↑ FcRn	(Richards et al., 2008; Zalevsky et al., 2010)
SDIE	S239D/I332E	↑ ADCC ↑ADCP	(Lazar et al., 2006; Richards et al., 2008)
SDIELS	S239D/I332E/M428L/N434S	↑ ADCC, ↑ADCP ↑ FcRn	(Lazar et al., 2006; Richards et al., 2008; Zalevsky et al., 2010)
SDIEAL	S239D/A330L/I332E	↑ ADCC, ↑ADCP	(Lazar et al., 2006; Moldt et al., 2011)
SDIEALLS	S239D/A330L/I332E/M428L/N434S	↑ ADCC, ↑ADCP, ↑ FcRn	(Lazar et al., 2006; Moldt et al., 2011; Zalevsky et al., 2010)
SDIEALYTE	S239D/A330L/I332E/M252Y/S254T/T256E	↑ ADCC, ↑ADCP	(Dall’Acqua et al., 2002; Lazar et al., 2006)
SDIEALYTELS	S239D/A330L/I332E/M252Y/S254T/T256E/M428L/N434S	↑ADCP, ↑ FcRn	(Dall’Acqua et al., 2002; Zalevsky et al., 2010)
SDIEGA	G236A/S239D/I332E	↑ ADCC, ↑ADCP	(Moldt et al., 2011; Richards et al., 2008)
SDIEGALS	G236A/S239D/I332E/M428L/N434S	↑ ADCC, ↑ADCP, ↑ FcRn	(Moldt et al., 2011; Richards et al., 2008; Zalevsky et al., 2010)
SDIESA	S239D/S298A/I332E	↑ ADCC, ↑ADCP
SDIESALS	S239D/S298A/I332E/M428L/N434S	↑ ADCC, ↑ADCP, ↑ FcRn	(Zalevsky et al., 2010)
SDIEALGA	G236A/S239D/A330L/I332E	↑ ADCC, ↑ADCP	(Smith et al., 2012)
SDIEALGALS	G236A/S239D/A330L/I332E/M428L/N434S	↑ ADCC, ↑ADCP, ↑ FcRn	(Smith et al., 2012; Zalevsky et al., 2010)
K326W	K236W	↑ ADCD, ↑ADCC	(Idusogie et al., 2000; Idusogie et al., 2001)
K326WLS	K236W/M428L/N434S	↑ ADCD, ↑ADCC	(Idusogie et al., 2000; Idusogie et al., 2001; Zalevsky et al., 2010)
EFTEA	G236A/S267E/H268F/S324T/I332E	↑ ADCC, ↑ADCP, ↑ADCD	(Moore et al., 2010)
EFTEALS	G236A/S267E/H268F/S324T/I332E/M428L/N434S	↑ ADCC, ↑ADCP, ↑ADCD ↑ FcRn	(Moore et al., 2010; Zalevsky et al., 2010)
AAA	T307A/E380A/N434A	↑ FcRn	(Shields et al., 2001)
EANA	E380A/N434A	↑ FcRn	(Shields et al., 2001)
N434W	N434W	↑ FcRn	(Yeung et al., 2009)
LS	M428L/N434S	↑ FcRn	(Zalevsky et al., 2010)
YTE	M252Y/S254T/T256E	↓ADCC, ↑ FcRn	(Dall’Acqua et al., 2002)
YTELS	M252Y/S254T/T256E/M428L/N434S	↓ADCC, ↑ FcRn	(Dall’Acqua et al., 2002; Zalevsky et al., 2010)
DVNH	D376V/N434H	↑ FcRn	(Datta-Mannan et al., 2007)
N434A	N434A	↑ FcRn	(Shields et al., 2001)
PINH	P257I/N434H	↑ FcRn	(Datta-Mannan et al., 2007)
PIQI	P257I/Q311I	↑ FcRn	(Datta-Mannan et al., 2007)
PIQILS	P257I/Q311I/M428L/N434S	↑ FcRn	(Zalevsky et al., 2010)
QL	T250Q/M428L	↑ FcRn	(Hinton et al., 2004)
AALS	T307A/E380A/M428L/N434S	↑ FcRn	(Zalevsky et al., 2010)
P257ILS	P257I/M428L/N434S	↑ FcRn	(Zalevsky et al., 2010)
QLS	T250Q/M428L/N434S	↑ FcRn	(Zalevsky et al., 2010)
ALS	E333A/M428L/N434S	↑ ADCD, ↑ADCC, ↑ FcRn	(Zalevsky et al., 2010)
SELF	S267E/L328F	↑ FcγRIIb binding	(Chu et al., 2008)
SELFLS	S267E/L328F/M428L/N434S	↑ FcγRIIb binding	(Chu et al., 2008; Zalevsky et al., 2010)
D376VLS	D376V/M428L/N434S	↑ FcRn	(Datta-Mannan et al., 2007; Zalevsky et al., 2010)
E380ALS	E380A/M428L/N434S	↑ FcRn	(Shields et al., 2001; Zalevsky et al., 2010)

Functional characterization of the REFORM antibody panel.

To assess whether the Fc mutations affected the Fab domain, we first examined binding of each of the VIC16 REFORM antibodies (REFORMabs) to EBOV GP by ELISA (Table S1). Although the original VIC16 mAb was a murine IgG1 antibody, chimerization of VIC16 with human IgG Fc domains was previously shown not to affect its ability to bind to EBOV GP antigen (Zeitlin et al., 2011). Similarly, nearly all of the REFORMabs exhibited EBOV GP binding that was comparable to wild type. However, among the panel of 61 mAbs, 8 had significantly decreased (<50%) GP binding relative to the other mAbs in the panel, despite no apparent issues during production of these mAbs. Thus, 53 Fc-variants had binding capacities equivalent to the original IgG1 construct and could be used to examine the role of Fc-effector activity in in vivo protection.

To define the unique functional profiles across the VIC16 REFORMab panel, the ability of each Fc-variant to drive phagocytosis of GP-coated beads by monocytes (ADCP) and neutrophils (ADNP), as well as complement deposition (ADCD) onto GP-coated beads was profiled, as was the ability to induce NK cell degranulation (ADNKA: CD107a) and NK cell secretion of IFNγ (ADNKA: IFNγ) and MIP-1β (ADNKA: MIP-1β) (Figure 2A). All REFORMabs were compared to the original murine IgG1 VIC16 mAb. Interestingly, wild type VIC16-human IgG1 induce only ADCP activity. In contrast, a range of functional activity was clearly observed across the REFORMab panel (Figure 2A). In particular, REFORMabs carrying Fc-variants shown to enhance ADCC activity, such as SDIE (Lazar et al., 2006), showed significantly elevated NK cell activation compared to the original VIC16 and the humanized IgG1 variant. Similarly, increased ADCP activity and ADCD activity were observed for variants with corresponding mutations that are predicted to enhance these functions. Although activation of neutrophils has not been routinely screened across Fc-variants, several mutants, including SDIE, exhibited enhanced ADNP and binding to FcγR3A (Lazar et al., 2006), highlighting additional effector functions that are present in this panel of Fc-variants.

Figure 2. Functional characterization of the REFORMab panel.

A. VIC16 REFORMabs (5 μg/ml) were evaluated for induction of antibody-dependent monocyte phagocytosis (ADCP), neutrophil phagocytosis (ADNP), complement deposition (ADCD), NK cell degranulation (NK: CD107a), NK cell secretion of IFNγ (NK: IFNγ) and NK secretion of MIP-1β (NK: MIP-1β) following antibody incubation with recombinant Ebola GP. The raw values for each effector function are plotted. The comparison of protection and function between internal fusion loop binding antibodies is shown in Figure S2. The quality control analysis of produced REFORMabs is shown in Figure S3, and representative flow cytometry plots are shown in Figure S4.

B. Plot of IC₅₀ (μg/ml)⁻¹ calculated from an 8-point, 2-fold dilution series of VIC16 REFORMabs in a neutralization assay with VSV pseudovirions expressing EBOV GP.

C. Functional profile of each REFORMab shown as a flower plot, in which each petal represents a different function, and the size of the petal represents the magnitude of the function relative to the entire panel as indicated in the legend.

Given that Fc is known to influence Fab activity (Kunert et al., 2004), we next screened the neutralizing activity of each VIC16 REFORMab using vesicular stomatitis virus (VSV) pseudotyped to display EBOV GP. The neutralizing activity of each REFORMab was compared to the original murine IgG1 VIC16 mAb and the antibody KZ52 that shows strong neutralization activity and was used as a positive control (Parren et al., 2002).The IC₅₀ of the original VIC16 and KZ52 was 0.34μg/ml and 0.08μg/ml, respectively, which is consistent with previously published IC₅₀ values for both mAbs (Parren et al., 2002; Saphire et al., 2018a). The majority of REFORM antibodies showed potent neutralizing activity comparable to the wildtype VIC16 activity (IC₅₀ <1μg/ml), yet 8 mAbs lost neutralization activity while maintaining EBOV GP binding and induction of effector function (Figure 2B). Of these, 7 of 8 (87.5%) were combinations of functional variants paired with the half-life extension mutation, LS (Zalevsky et al., 2010). An additional 20 REFORMabs had an IC₅₀ >1μg/ml and were considered to have lost neutralization potency compared to the rest of the panel. Interestingly, many of these antibodies still induced Fc-effector functions, albeit to a reduced level in some instances, indicating that specific combinations of changes in the Fc domain impacted Fab-mediated neutralizing activity as has been previously observed for engineered HIV-specific mAbs (Kunert et al., 2004).

For an overview of the polyfunctional profile of individual antibodies, for each REFORMab we generated composite flower plots wherein each effector function was represented by a color-coded petal that vary in size according to the relative activity of each function (Figure 2C). We observed robust variation across all the Fc-variants having varying degrees of neutralization (dark blue petal) and addition of individual or combinations of Fc-effector functions. Thus, this panel of REFORMabs provides a heterogeneous source of Fc-variation on a single antibody specificity, providing an opportunity to define specific protective correlates of immunity against EBOV.

Down-selection of REFORM antibodies for in vivo evaluation.

Among the 30 REFORMabs that retained neutralizing activity of IC₅₀ <1μg/ml, heterogeneity was observed in antibody functional profiles (Figure 3A, left). To define classes of functional variants, we used K-means clustering to identify unique functional clusters (Figure 3A, right). Across these 30 REFORMabs, 6 different groups emerged that captured distinct functional profiles including non-functional mAbs (cluster 1), predominantly monofunctional mAbs (clusters 2, 3), and polyfunctional antibodies (clusters 4–6). The monofunctional antibodies in cluster 2 induced only monocyte phagocytosis (ADCP). The sole mAb in cluster 3, KWES, exhibited high levels of complement deposition. Within the polyfunctional clusters, clusters 4 and 5 differed mainly by the magnitude of NK cell-mediated activity, with mAbs in clusters 4 and 5 exhibiting moderate and high levels of NK cell activation, respectively. Cluster 6 contained only one mAb, EFTEA, that induced all functions tested, including high levels of complement activity, yet moderate levels of NK cell activation.

Figure 3. Selection of REFORMabs for in vivo evaluation.

A. Unsupervised hierarchical clustering with two-way color coding was used to cluster antibodies according to induction of innate immune effector functions (left graph), and K-means clustering was used to group IgG1 VIC16 REFORMabs having a neutralization IC₅₀ of <1μg/ml into 6 distinct clusters (right graph). Results for principal component analysis of mAbs within each cluster are shown, and the inset bi-ray plot indicates the loadings within the plot. The K-means clustering defined-cluster for each antibody is indicated below the heatmap.

B. The five down-selected REFORMabs were produced in larger quantities and re-evaluated in triplicate in dilution curves for neutralization of VSV-EBOV GP pseudovirions, and induction of antibody-dependent complement deposition (ADCD), neutrophil phagocytosis (ADNP), monocyte phagocytosis (ADCP), NK cell degranulation (NK: CD107a), NK cell secretion of IFNγ (NK: IFNγ), and NK secretion of MIP-1β (NK: MIP-1β). Antibodies were assayed for neutralization in a 3-fold dilution curve (3 μg/ml – 0.0048 μg/ml); ADCD in a 2-fold dilution curve (10 μg/ml – 0.16 μg/ml); and ADNP, ADCP, and ADNKA in a 5-fold dilution curve (5 μg/ml – 0.00064 μg/ml).

C. Selected REFORMabs were re-evaluated for induction of effector function, and the AUC for each function across the down-selected in vivo panel is shown in a flower plot

D. Binding affinity for human FcγRs (FcγR3a, FcγR2A, FcγR2B, and FcγR3B) and mouse FcγRs (FcγRI FcγRIV, FcγR3, and FcγR2b) was measured for the selected antibodies by surface plasmon resonance (human FcγR) or biolayer inferometry (mouse FcγR) and K_D (M⁻¹) was plotted. The correlation matrix across effector functions and FcγR binding is shown in Figure S5.

E. Binding of low-affinity murine FcγRs (mFcγRIIb and mFcγRIII) to mAb:Ebola GP immune complexes was determined using fluorescently labeled recombinant murine FcγRs at the indicated antibody concentrations. Median Fluorescent Intensity (MFI) of FcγR binding is plotted.

Comparing the functional profiles of the different clusters of VIC16 REFORMabs to the profiles induced in EVD survivors (Figure 1B), we noted that EVD survivor cluster 5, which exhibited moderate levels of NK cell activation, but high complement and phagocytic activity, was similar to REFORM cluster 6 (EFTEA). Higher levels of complement activity in EVD cluster 4 were mimicked by REFORM cluster 3 (KWES), and the high levels of NK cell activation and phagocytic activity yet low complement activation in EVD cluster 2 were represented in REFORM cluster 5. The non-functional EVD cluster 9 was represented by REFORM cluster 1. Thus, the functional diversity identified in EVD survivors was mimicked by the functional diversity in the REFORM panel, and five REFORMabs were selected for in vivo analysis.

Distinct functional profiles track with in vivo antibody-mediated protection.

To define the specific Fc-effector profiles linked to protection, five representative REFORMabs representing the different functional profiles were produced. Binding affinity for Ebola GP, induction of each effector function, and neutralization and was re-measured for the large-scale batches of these five antibodies (Figure 3B). IgG1 was monofunctional, and exhibited only monocyte phagocytosis; KWES induced high levels of complement deposition and low levels of ADCP and NK cell activation; LALA represented a non-functional Fc domain; SDIEALGA captured polyfunctional Fc domains with high NK cell activation and neutrophil phagocytosis; and EFTEA captured all effector functions with high levels of complement deposition, ADCP, and ADNP but moderate levels of NK cell activation (Figure 3C).

The binding affinity of each mAb for both human and mouse FcγRs was measured by surface plasmon resonance (Figure 3D; Table 2). As expected, the SDIEALGA variant known to increase antibody affinity for the activating receptors FcγR3A and FcγR2A (Smith et al., 2012) exhibited a 9-fold and 18-fold increase in K_D for FcγR3A and FcγR2A, respectively, compared to the VIC16 human IgG1 (Table 2). The K_D of the EFTEA variant for FcγR2A and the inhibitory FcγR2B was 120-fold and 19-fold higher compared to IgG1, respectively, consistent with increases described in previous reports (Moore et al., 2010). Only the SDIEALGA variant showed elevated binding to FcγR3B, a neutrophil-specific FcγR (Bruhns, 2012). The LALA variant showed significantly abrogated binding to the human FcγRs, as expected (Hessell et al., 2007). KWES demonstrated modest enhancement in binding to FcγR3A and FcγR2A, consistent with the moderate levels of NK cell activation and phagocytic activities observed, and no binding to FcγR3B was detected, which likely explains the limited ADNP activity seen for this variant.

Table 2. FcγR binding affinities for the downselected VIC16 REFORMabs.

The K_D (μM) values for the indicated antibodies are shown and the fold-change relative to IgG1 is shown in brackets.

Variant	huFcγR3A	huFcγR2A	huFcγR2B	huFcγR3B	mFcγRI	mFcγRIV	mFcγR3	mFcγR2b
IgG1	6.1 [1]	30 [1]	23 [1]	25 [1]	0.08 [1]	0.24 [1]	2.83 [1]	14.9 [1]
LALA	1700 [0.003]	892 [0.03]	ND	ND	ND	ND	820.3 [0.04]	ND
SDIEALGA	0.675 [9]	1.6 [19]	21 [1.1]	3.2 [7.8]	0.1 [0.8]	0.001 [235]	0.01 [0.16]	0.13 [114]
EFTEA	6.1 [1]	0.252 [119]	1.2 [19]	475 [0.05]	0.06 [1.3]	0.12 [2]	1.1 [9]	0.49 [30]
KWES	1.6 [3.8]	8.6 [3.4]	224 [0.12]	ND	0.07 [1.1]	0.16 [1.5]	0.02 [0.94]	46 [0.3]

ND = not detected

Mounting evidence points to significant parallels in human IgG1:humanFcγR (hFcγR) and human IgG1:mouse FcγRs (mFcγR) binding profiles (Dekkers et al., 2017; Overdijk et al., 2012), with the exception of mFcγRI. However, whether the REFORMab hFcγR binding profiles translated across species was uncertain, although this feature is key to in vivo efficacy testing in mice. Similar to differences in hFcγR binding profiles, the SDIEALGA variant (14-fold increase) exhibited enhanced binding to mFcγRIV (Figure 3D; Table 2), which in mice is expressed on monocytes/macrophages and neutrophils and is associated with ADCC (Nimmerjahn et al., 2005). EFTEA showed increased affinity for mFcγRIII (9-fold increase), present on all murine myeloid cells and the sole mFcγR expressed on resting murine NK cells (Bruhns, 2012), whereas SDIEALGA exhibited a 6-fold decrease in binding to mFcγRIII compared to IgG1, thus highlighting differential capacities of the variants to engage resting murine NK cells. The SDIEALGA and EFTEA variants showed a modest (4-fold and 2-fold, respectively) increase in binding to mFcγRIIB, the inhibitory receptor present on murine myeloid cells and B cells. Although the KWES variant exhibited an enhanced ability to activate complement, a pathway that is highly conserved across species (Nonaka and Kimura, 2006), its binding to other mFcγRs did not increase. The LALA variant, which showed the expected abrogation of binding to human FcγRs, exhibited abolished binding to mFcγRI, mFcγRIV and mFcγRIIb, and weak binding to mFcγRIII. All other antibodies (IgG1, EFTEA, SDIEALGA, and KWES) demonstrated comparable binding to the high affinity mFcγRI, present on murine dendritic cells. As immune complexes increase avidity of antibodies to low-affinity FcγRs, we also performed analysis of murine FcγRIIb and FcγRIII binding to Ebola GP:mAb immune complexes. The low affinity murine FcγR bound to all of the immune complexes with the exception of the LALA variant immune complexes (Figure 3E). The mFcγRIIb bound highest to the SDIEALGA variant immune complex, consistent with the high binding affinity observed by SPR, whereas KWES showed elevated binding to mFcγRIII compared with the other mAbs (Figure 3E).

To begin to start developing comparative rules sets to translate binding between human and mouse FcRs, we determined the FcγR binding affinities across a subset of the VIC16 REFORM panel to both human and mouse FcγRs (Figure S5A). Strong concordance in binding affinity was observed between mFcγRIV and human huFcγR3A and huFcγR3B, indicating similarity between these receptors and highlighting that Fc mutations aimed at enhancing binding to huFcγR3A will likely also enhance binding to mFcγRIV. Accordingly, affinity to mFcγRIV was strongly associated with human NK cell and neutrophil effector functions (Figure S5B). Moderate concordance was observed between mFcγRIV and huFcγR2A high-affinity variant H131, highlighting parallels in across the non-FcγRI high affinity activating FcγRs between species. Together, these data highlight the unique binding potential of the REFORMabs to both human and mouse FcRs and provide the foundation to build a comparative rule set to translate FcR-meditated activities across species that will help provide critical insights for potential in vivo differential therapeutic benefit.

Dissecting Fc-correlates of immunity in vivo

To define Fc-correlates that contribute to antiviral immunity in vivo, we tested the performance of REFORMabs in a stringent murine model of Ebola virus infection (Bornholdt et al., 2016; Bray et al., 1998; Saphire et al., 2018a; Wec et al., 2017; Zeitlin et al., 2011). For each REFORMab, groups of 5 BALB/c mice were infected with 1,000 plaque-forming units (pfu) of mouse-adapted Ebola virus, 1 day before administration of 100 μg of antibody per mouse. Mice were monitored for survival, clinical disease score, and weight loss for 28 days (Figure 4A–B). As expected, PBS-treated mice succumbed to infection by day 5 post-infection, developing clinical disease and weight loss by day 2 post-infection (Figure 4A), highlighting the very aggressive and acute nature of the infection. The LALA variant, which showed loss of hFcγR binding but retained weak binding to mFcγRIII, protected 80% (4/5) of the mice, suggesting that neutralization activity with some mFcγR recruitment was sufficient to protect a majority of animals. However, animals given LALA REFORMabs lost weight and developed disease signs through day 12 (Figure 4B). The animals ultimately survived infection, but the treatment with these REFORMAbs could not prevent disease. Similarly, wild type IgG1, which induced a monofunctional monocyte phagocytic response, also protected 80% of animals, but did not prevent development of disease signs or weight loss. In contrast, the polyfunctional variant SDIEALGA that exhibited robust NK cell activation, but lacked complement activation, protected only 60% of the mice from death, and did not prevent development of disease. The EFTEA and KWES variants conferred 100% survival in treated animals, and, remarkably, the animals showed no evidence of development of disease. Only one animal treated with the EFTEA variant showed mild clinical signs at day 5 post-infection (Figure 4A). Both EFTEA and KWES induced high levels of complement activation and exhibited polyfunctional activity, with moderate induction of NK cell-mediated activity and monocyte phagocytosis.

Figure 4. Distinct functional profiles track with in vivo antibody-mediated protection.

A. Selected REFORMAbs were administered to mice (100 μg/mouse) 24 hours after infection with 1,000 pfu of mouse-adapted Ebola virus. Mice were monitored for 28 days post-infection, and survival of mice is plotted.

B. Clinical disease score (top row) and percentage of starting weight (bottom row) of mice is shown with different colors representing individual mice. The percentage survival is indicated in the top right corner panels showing clinical disease score for each mAb.

C. Selected mAbs were administered to mice at a reduced dose (30 μg/mouse) 24 hours after infection with 1,000 pfu of mouse-adapted Ebola virus. Mice were monitored for 28 days post-infection, and the survival is plotted.

D. Clinical disease score (top row) and percentage of starting weight (bottom row) of mice treated with 30 μg of the indicated antibody is shown with different colors representing individual mice.

To further confirm the protective effects of EFTEA and KWES, a second group of mice were dosed with approximately one-third (30 μg/mouse) of the original antibody dose (Figure 4C–D). Animals treated with wild type, monofunctional IgG1 at this reduced dose all succumbed to Ebola virus infection. In contrast, animals treated with reduced doses of either EFTEA and KWES variants rapidly recovered from symptoms and disease and all survived (Figure 4C–D). These data suggest that even at low doses of antibodies carrying these modifications, animals could recover and survive a lethal viral challenge, which points to the critical importance of complement and low-level ADCC as mechanistic correlates of protection.

Discussion

In-depth characterization of antibodies isolated from human survivors of EVD identified critical neutralizing epitopes that can be harnessed in therapeutic treatment of infection (Bornholdt et al., 2016; Corti et al., 2016; Flyak et al., 2016; Maruyama et al., 1999). Given our emerging appreciation for the role of Fc-effector function in the control/clearance of EBOV infection (Gunn et al., 2018; Misasi et al., 2016; Olinger et al., 2012; Zeitlin et al., 2011), we hypothesized that a similar analysis of human EVD survivors could help identify specific Fc-effector profiles that are most critical for control and resolution of disease. Nearly all EVD survivors possessed functional EBOV-specific humoral immunity, albeit with variable levels and combinations of antibody effector functions. Thus, to precisely define the mechanistic Fc-effector correlates of protection, we performed Fc-engineering. While Fc-modifications that potentiate NK cell mediated ADCC have largely been explored in EBOV-monoclonal mediated engineering (Bornholdt et al., 2019; Zeitlin et al., 2011), here we exploited an agnostic Fc-engineering approach to generate a minimal set of Fc-variants on a single Fab able to trigger antibody profiles similar to those observed in EVD survivors. Surprisingly, passive transfer of Fc-engineered variants highlighted the critical role of robust complement activation and moderate NK cell activity as key mechanistic correlates of protection against EBOV. These data point to a critical role not only for combinations of effector functions, but also for the levels of induction, which may be key to antibody-mediated protection, wherein the most protective mAbs, KWES and EFTEA, completely abrogated development of clinical disease at a 100μg dose, and survival even at low antibody doses. These data highlight the disease-modifying effects of Fc-engineering, yet also highlight the critical need to customize Fc-design to maximize the recruitment of the functional correlates of immunity to the pathogen of interest.

Accumulating data points to similarities in mFcγR:IgG1 and hFcγR:IgG1 responses (Overdijk et al., 2012), although translation of findings for Fc engineering of a human IgG1 requires careful dissection of binding profiles across species. Recent analysis of the binding affinity of human IgG1 to both human and murine FcRs suggests that hIgG1 binding profiles are more comparable to mouse IgG2a than previously thought (Dekkers et al., 2017). Along these lines, our analysis with human IgG1:mFcγRs support and extend these findings, suggesting that engineered hIgG1 variants can skew binding to mouse FcRs. The Fc-silencing LALA variant, for example, has been used in mouse models to investigate the requirement for FcγR-mediated effector function in antibody-mediated protection (Gilchuk et al., 2018; Ilinykh et al., 2020). As expected, the LALA variant displayed abrogated binding to both human and mouse FcγRs, clearly demonstrating comparable activity across species, and offering a simple means to investigate the role of Fc-effector function as a correlate of protection against Ebola virus. Similarly, given the highly conserved nature of complement activity across species, we observed strong concordance in antibody-mediated complement deposition between guinea pig-derived and human-derived complement (Fischinger et al., 2019), pointing to translatability across species.

However, antibodies do not bind to individual FcRs in vivo, but instead bind to collections of FcRs present on distinct innate immune cells in vivo in the setting of immune-complexed pathogens, infected cells, or debris. We chose our down-selected panel to reflect the effector profiles in human survivors of EVD yet as protective efficacy must be tested in animal models, a complete picture is required to fully capture the cross-species translation of antibodies from animal models to humans. Thus here, each REFORMab was tested in human and murine functional assays as well as in their capacity to interact with mouse and human FcγRs. The data demonstrate remarkable concordance in expected FcR-binding:functional activity, particularly between murine FcγRIV and human FcγR3A/B binding and activation of human NK cells and neutrophils (Figure S5). Future cross-species translation must take both antibody:FcR binding profiles into consideration, but also cell type specific FcR expression profile differences across species to fully translate in vivo observations across the species. Specifically, differences in expression of FcγRs on murine and human innate immune cells, their localization, and their differential innate immune effector functions can help resolve mechanisms across the species. In the context of our data presented here, the Fc engineered mAbs, except for LALA, showed similar binding to mFcγRIII, the only FcγR expressed on murine NK cells. However, the SDIEALGA variant induced the highest levels of human NK cell degranulation, representative of ADCC, and displayed the highest affinity for mFcγRIV, the FcγR that drives ADCC in mice via macrophage and neutrophils rather than NK cells. Thus, these data suggest that SDIEALGA likely drives elevated levels of ADCC in mice but through different cell types in mice and humans, suggesting that identifying cross-species functional-matches that track with protection may be more informative than simply defining Fc-receptor binding similarities. Thus, developing cross-species maps using modified monoclonal antibodies with the same specificities, matching half-lives/biodistributions, but differential binding properties to FcRs across species may provide the critical means to help define mechanistic translatability across the models.

Immunopathology is a hallmark of EVD (Baseler et al., 2017), and although ADCC has been proposed to be a critical mechanism of protection against EBOV (Corti et al., 2016), the SDIEALGA variant, which exhibited the highest binding to mFcγRIV and human NK cell activation of the panel, showed the lowest protection activity among the VIC16 variants, indicating that maximal induction ADCC may be detrimental to antibody-mediated protection. Previous studies suggest that SDIEALGA exhibits similar pharmacokinetics, suggesting that reduced protection is unlikely related to enhanced clearance of the antibody variant (DiLillo et al., 2014). Thus, the ability to induce just enough activation without overactivation of innate immune cells may be a critical parameter of antibody-mediated protection. Of note, antibody-mediated activation of human NK cells was observed in most survivors, and the most protective REFORMabs retained moderate NK cell functions, further supporting the hypothesis that a balanced, but potentially non-overactive ADCC response in the setting of complement activation may be critical to fully contain and control this lethal pathogen.

The rapid emergence of SARS-CoV-2 and resulting COVID-19 pandemic has highlighted the urgent need for the rapid development of highly effective and potent therapeutics to treat infected patients. The unprecedented speed at which neutralizing antibodies against SARS-CoV-2 have been identified allows for the rapid development of antibody-based therapeutics to help slow and end the pandemic (Ju et al., 2020; Pinto et al., 2020). However, humoral immune responses have been linked to both protection (Agnihothram et al., 2013; Corti et al., 2015; Deming et al., 2006) and pathology (Jaume et al., 2011; Liu et al., 2019; Wang et al., 2014) following coronavirus infections, rendering Fc-engineering efforts quite contentious. Nonetheless, the REFORM platform can accelerate the development of highly potent therapeutics against SARS-CoV-2 through the rapid generation of libraries of Fc-variants that are tailored to have different Fc-effector functions. These variants can be tested in animal models to define the precise Fc-effector profiles associated with pathogenesis or protection. Moreover, linked to intense pharmacokinetic and pharmacodynamic profiling of each variant, therapeutically effective antibodies with ideal therapeutic activities may be defined. Ultimately, modeling binding profiles across FcγRs:Fc effector functions and biodistribution in animal models and humans may provide bridging information to guide the accelerated development of highly effective therapeutics as countermeasures for emerging diseases.

Importantly, the Fc-effector mechanistic correlates of immunity likely differ across pathogens as well as across epitopes even within the same pathogen, with the latter related to antibody angle-of-attack or stoichiometry (Bournazos et al., 2019; He et al., 2016; Saphire et al., 2018b). Notably, the antibody panel examined in this study does not include all published IgG1 Fc-point mutants, nor does it include modifications of other IgG subclasses, IgA, and IgM, which represent a much larger explorable immune space that is naturally exploited during an immune response to drive protective immunity. However, the approach detailed here offers an unbiased systems biology approach to broadly control Fc-effector function and develop large libraries of “functionally unique” reagents that can be tested for in vivo or in vitro efficacy. Thus, the REFORM platform offers a unique approach to rapidly screen Fc-engineered mAbs to define the mechanism of antibody action against any target pathogen to accelerate monoclonal therapeutic or vaccine design against EBOV and emerging pathogens like SARS-CoV-2.

Limitations of the Study

The REFORMAb platform utilizes a rapid cloning platform to generate a panel of functionally diverse human antibodies with identical Fab domains. As the ultimate goal of this work is to guide the development of new antibody-based therapies for human Ebola virus infection, we performed effector function analysis using human innate immune cells. However, we used a murine challenge model to determine whether the protective efficacy of a neutralizing Ebola-specific antibody, VIC16, is enhanced by the incorporation of specific Fc mutations, thus creating a complex scenario in interpreting the mechanism of protection of human antibodies. Some effector functions, such as complement, are readily translated across species, and thus the finding that the REFORMabs with high complement activity protected mice from disease support the hypothesis that antibody-mediated complement activation may also be protective in humans. In contrast, ADCC activity in mice and humans are mediated by different immune cells and different FcγRs – macrophage and neutrophils mediate ADCC in mice via FcγRIV whereas NK cells are primarily responsible for ADCC in humans via FcγR3A. We observed a strong association between activation of human NK cells and antibody binding to FcγR3A across the REFORMabs and between binding of the REFORMab panel to murine FcγRIV and human FcγR3A, suggesting that activation of human NK cells may serve as a surrogate for ADCC in mice, albeit mediated by a different cell type in the mouse challenge model. Concordance between antibody-dependent phagocytosis by mouse and human innate immune cells have been previously observed (Gunn et al., 2018), yet were not compared in this study. The development of bridging analyses to translate findings across species is ongoing, and the identification of the effector function that tracks with protection rather than the cell type that mediates it may enable better translation between species. Dissecting the similarities and differences between animal models and humans will enable an improved framework towards effective translation of antibody efficacy across species to ultimately leverage the immune system to fight the infection in humans inspired by functional correlates of immunity in animal models.

STAR Methods and Materials.

RESOURCE AVAILABILITY

Lead Contact.

Further information and requests for resources and reagents should be directed to and will be fulfilled by Lead Contact, Dr. Galit Alter (galter@mgh.harvard.edu)

Materials Availability.

Plasmids and antibodies generated in this study may be requested with a material transfer agreement.

Data and Code Availability.

The datasets generated in this study are available on Mendeley (doi: 10.17632/9b2h8p9gtc.1)

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Patient samples.

Plasma samples were collected from individuals with documented clinical history of EVD approximately 3–4 years after infection and from household contacts of EVD survivors. At the time of acute disease in 2014, EVD patients received standard of care at Kenema Government Hospital (e.g., intravenous fluid replacement and administration of antibiotics as determined by the treating physician). Peripheral blood mononuclear cells were collected from MGH blood bank donations for use in innate immune effector assays. All subjects provided written consent. This study was approved by the Institutional Review Board of the Human Subjects Committee of MGH, Tulane University Institutional Review Board, and the Sierra Leone Ethics and Scientific Review Committee.

Primary human innate immune cells.

Innate immune effector cells were isolated from fresh peripheral blood samples collected by the Ragon Institute or the MGH Blood bank from healthy human volunteers. All subjects signed informed consent and the study was approved by the MGH Institutional Review Board.

Cell lines.

Freestyle 293F cells were purchased through Thermo Fisher and maintained in Freestyle 293F medium (Thermo Fisher) in a HT Infors shaking incubator at 125 RPM, 37C 8% CO₂. The human monocyte cell line, THP-1, was purchased through ATCC. Cells were maintained in RPMI1640 supplemented with 10% fetal bovine serum, L-Glutamine, penicillin/streptomycin, and 55μM beta-mercaptoethanol in a humidified incubator at 5% CO₂.

Animals.

Naïve seven-week-old BALB/c mice used for challenge studies were purchased from Charles River Laboratory) were housed in microisolator cages in the ABSL-4 facility at the Galveston National Laboratory.

METHOD DETAILS.

Plasmid design and construction.

Four donor pUC plasmids encoding the variable heavy domain, the Fc domain, a furin P2A sequence, or the variable light domain of an antibody surrounded by BsaI sites were designed. Two destination vectors were designed that contained an IL-2 secretion signal, the suicide gene ccdB flanked by BsaI sites, and either the kappa or lambda light chain sequences. A library of 63 Fc variants were synthesized and cloned into the pUC donor plasmids for the Fc domain (Table 1). The 5 plasmids were combined in a single digestion-ligation reaction to generate an expression plasmid encoding both the heavy chain and light chain of a monoclonal antibody as follows: 50 ng of each donor and destination plasmid, 20 units BsaI-HF (NEB), 200 units T4 ligase, and 0.67 mM ATP in 1X T4 ligase buffer (NEB) supplemented with BSA were combined and incubated at 37 °C for 1 h followed by incubation at 50 °C for 15 min and 80 °C for 15 min. Ligation was performed by adding 200 units T4 ligase followed by incubation at 16°C for 30 min. Inclusion of a ccdB gene that is removed during the digestion-ligation process increases the likelihood that the recombinant plasmid is selected as opposed to the parental vector. Two alternate destination vectors were created for expression of only the light chain to optimize the heavy chain to light chain ratio during antibody production. Ligation products (2 μl) were transformed into Stellar competent cells (Clontech) and plated onto agar plates with kanamycin. The resulting colonies were screened and sequenced to verify the identity of each new plasmid.

Production of antibodies in mammalian cells.

Plasmids were expanded and transfected into 293F suspension cells grown in Freestyle™ 293 Expression media (Gibco). Cells were seeded two days pre-transfection at a density of 0.4×10⁶ cells/mL in 50 mL in a 125 mL baffled Erlenmeyer flask with a vented closure. On the day of transfection, cells were counted again and the cell density was adjusted to 1.2×10⁶ cells/mL in 50 mL media. Total DNA (25 μg) was transfected into cells using Polyethylenimine (PEI) (Polysciences) at 1 μg/μl in a ratio of 3 μg PEI to 1 μg DNA. Cell culture supernatants were harvested 5 days post-transfection. Supernatants were incubated with Protein G magnetic beads (Millipore) overnight at 4 °C. The Protein G Beads were washed 4 times with 1x PBS before antibodies were eluted using Pierce™ IgG Elution Buffer (Thermo Fisher Scientific) and neutralized with Tris-HCl pH 8.0 at a ratio of 1:10. For 500 mL production volumes, cells were transfected with 250 μg total DNA with PEI at the same ratio as that used for smaller production. Cell culture supernatants were harvested 5 days post-transfection and incubated overnight at 4 °C with Pierce™ Protein A/G Plus Agarose. Agarose resin was collected by pouring the supernatant over a BioRad Econo-Column Chromatography Column. Resin was washed with PBS before antibody was eluted with Pierce™ IgG Elution Buffer directly into Tris-HCl pH 8.0 at a ratio of 1:10. Antibody eluates were concentrated using Amicon Ultra-15 Centrifugal Filter Units with a 50 kDa molecular weight cut off to a final volume of approximately 1mL.

Neutralizing activity.

Three-fold dilutions of antibodies (3 μg/ml – 0.0048 μg/ml) were assayed in technical duplicates and incubated with 1×10⁴ RLU/well of Ebola GP-pseudotyped vesicular stomatitis virus expressing luciferase (IBT Bioservices) for 1 hour at room temperature before addition to VeroE6 monolayers (6×10⁵ cells/well) and incubation at 37 °C for 14–16h. Cells were lysed using Passive Lysis Buffer (Promega) and luciferase activity was determined using a luciferase activating reagent (Promega). IC₅₀ titers were determined using Prism 8.0.

Antigen-binding ELISA.

Recombinant Ebola GP antigen (IBT Bioservices) was coated onto MaxiSorp 384-well plates (Nunc) at 62.5 ng/well at 4 °C overnight. Wells were washed with PBS and blocked with 5% BSA prior to addition of 2-fold dilutions of antibodies (10–0.08 μg/ml) for 2 h at room temperature. Wells were washed 6 times with PBS + 0.01% Tween 20 and incubated for 1 h with an HRP-conjugated anti-human IgG Fc antibody (1:5000; Jackson ImmunoResearch). Wells were again washed 6 times with PBS + 0.01% Tween 20, incubated with 1X TMB substrate for 5 minutes and the reaction was stopped with 1N H₂SO₄ before absorbance at 405 nm was measured. The area under the curve was determined using Prism 8.0.

Antigen affinity measurement by biolayer inferometry.

0.3 μg/ml of each mAb was immobilized to anti-human IgG Fc (AHC) kinetic biosensors (Sartorius/ForteBio) in 1X kinetics buffer for 180s followed by association and dissociation measurements to two-fold dilutions of recombinant Ebola GP starting at 500nM on an Octet Red96 instrument (Forte Bio). Association and dissociation were measured for 120s and 600s, respectively. Following subtraction of the reference control, the KD was calculated using a 1:1 binding model in the Octet Data Analysis software version 8.1.

Antibody-dependent NK cell degranulation.

Recombinant Ebola GP antigen (IBT Bioservices) was coated onto MaxiSorp 96-well plates (Nunc) at 300 ng/well at 4 °C overnight. Wells were washed with PBS and blocked with 5% BSA prior to addition of antibodies (5 μg/ml) and incubation for 2 h at 37 °C. Unbound antibodies were removed by washing with PBS, and NK cells enriched from the peripheral blood of human donors were added at 5×10⁴ cells/well in the presence of 4 μg/ml brefeldin A (Sigma-Aldrich), 5 μg/ml GolgiStop (Life Technologies) and anti-CD107a antibody (Clone H4A3, BD Biosciences) for 5 h. Cells were fixed and permeabilized with Fix/Perm (Life Technologies) according to the manufacturer’s instructions to stain for intracellular IFNγ (Clone B27, BD Biosciences), and MIP-1β (Clone D21–1351, BD Biosciences). Cells were analyzed on a BD LSRII flow cytometer. Gating strategy and representative flow cytometry plots are shown in Figure S4.

Antibody-dependent neutrophil phagocytosis (ADNP).

Biotinylated Ebola GP was coupled to yellow-green Neutravidin beads (Life Technologies). Antibodies were diluted in culture medium to 5 μg/ml and incubated with GP-coated beads for 2 h at 37 °C. Freshly isolated white blood cells from human donor peripheral blood (5×10⁴ cells/well) were incubated for 1h at 37 °C. Cells were then stained for CD66b (Clone G10F5; BioLegend), CD3 (Clone UCHT1; BD Biosciences), and CD14 (Clone MP9; BD Biosciences), fixed with 4% paraformaldehyde, and analyzed by flow cytometry. Neutrophils were defined as SSC-A^high CD66b⁺, CD3⁻, CD14⁻. A phagocytic score was determined using the following formula: (percentage of FITC+ cells)*(geometric mean fluorescent intensity (gMFI) of the FITC+ cells)/10,000. Gating strategy and representative flow cytometry plots are shown in Figure S4.

Antibody-dependent cellular phagocytosis by human monocytes (ADCP).

GP-coated beads were generated as described for ADNP. Antibodies were diluted in culture medium to 5μg/ml and incubated with GP-coated beads for 2 h at 37 °C. Unbound antibodies were removed by centrifugation prior to the addition of THP-1 cells at 2.5×10⁴ cells/well. Cells were fixed with 4% paraformaldehyde and analyzed by flow cytometry. A phagocytic score was determined as described above. Gating strategy and representative flow cytometry plots are shown in Figure S4.

Antibody-mediated complement deposition (ADCD).

GP-coated beads were generated as described for ADNP, but with substitution of red-fluorescent Neutravidin beads (Life Technologies). Antibodies were diluted in culture medium to 5μg/ml and incubated with GP-coated beads for 2 h at 37 °C. Unbound antibodies were removed by centrifugation prior to the addition of reconstituted guinea pig complement (Cedarlane Labs) diluted in veronal buffer supplemented with calcium and magnesium (Boston Bioproducts) for 20 min at 37 °C. Beads were washed with PBS containing 15 mM EDTA, and stained with an FITC-conjugated anti-guinea pig C3 antibody (MP Biomedicals). C3 deposition onto beads was analyzed by flow cytometry. The gMFI of FITC of all beads was measured. Gating strategy and representative flow cytometry plots are shown in Figure S4.

Measurement of FcR binding kinetics by SPR.

Each mAb (12.5 μg/ml) was printed onto a MX96 200M chip (Xantec Bioanalytics) using a continuous flow microspotter (CFM) (Wasatch Microfluidics) with sulfo-NHS and EDC coupling chemistry. Measurement of FcγR binding kinetics was performed using an image-based array reader (MX96, IBIS Technologies). The chip was quenched with 1 M ethanolamine and 3-fold dilutions of recombinant FcγR (Duke University; 10 μM to 0.005 μM) were sequentially injected. To account for non-specific signals, the signals for each antibody were double-referenced using signals from both blank injections and from uncoupled inter-spots between antibodies. Data was processed using Sprint software (IBIS Technologies), and the kinetics of binding (K_D) was determined using Kinetics software (Carterra).

Measurement of FcR binding kinetics by biolayer inferometry.

100nM of each mAb was immobilized to anti-human IgG Fab-CH1 kinetic biosensors (Sartorius/ForteBio) in 1X kinetics buffer for 300s followed by association and dissociation measurements to two-fold dilutions of recombinant murine FcRs starting at 200nM (mFcγRI and mFcγRIV) or 600nM (mFcγRIIb and mFcγRIII) on an Octet Red96 instrument (Forte Bio). Association and dissociation were measured for 120s and 600s, respectively. Following subtraction of the reference control, the K_D was calculated using a 2:1 analyte binding model in the Octet Data Analysis software version 8.1.

Measurement of murine FcγR binding to immune complexes.

Recombinant EBOV GP (IBT Bioservices) were coupled to MagPlex beads (Luminex) via sulfo-NHS coupling chemistry. Murine FcγR receptors (FcγRIII, FcγRIV, FcγRI) and one inhibitory receptor (FcγR2B). Serial dilutions of the indicated antibodies samples were diluted in 1X PBS + 0.1% bovine serum albumin (BSA) + 0.05% Tween20 and incubated with antigen-coupled beads for 2 hours. Beads were washed and incubated with recombinant biotinylated Fc-receptors that were tetramerized via Streptavidin-PE for 1 hour at room temperature. Beads were washed analyzed on a Sartorius iQue screener. The median fluorescent intensity of 30 beads/region was recorded. A human IgG1 isotype control was used to establish antigen-specificity and background binding.

In vivo protection.

The animal protocol for testing of mAbs in mice was approved by the Institutional Animal Care and Use Committee of the UTMB. Seven-week-old BALB/c mice (Charles River Laboratories) were placed in the ABSL-4 facility at the Galveston National Laboratory. Groups of five animals were injected intraperitoneally with 1,000 PFU of mouse-adapted EBOV, strain Mayinga (Bray et al., 1998). The animals were injected with mAbs by the intraperitoneal route using 0.1 mg or 0.03 mg per treatment 24 h after virus challenge. Animals treated with PBS served as controls. Animal observation procedures were performed as described elsewhere (Ilinykh et al., 2018). The overall observation period was 28 days.

QUANTIFICATION AND STATISTICAL ANALYSIS

Correlation analysis.

Univariate spearman correlation r values were calculated using GraphPad Prism, and the p values were corrected for multiple comparisons using Benjamini-Hochberg correction.

K-means clustering analysis.

K-means clustering analysis was performed in JMP Pro 15, and elbow joint analysis was used to determine the optimal number of clusters in R.

Hierarchical clustering analysis.

Unsupervised hierarchical clustering analysis was performed in JMP Pro 15, using the Ward method.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
KZ52	IBT Bioservices	Cat #: 0260-001
VIC16	Takada et al., 2003	n/a
Mouse anti-human CD66b (clone G10F5), Pacific Blue	BioLegend	RRID:AB_2563294
Mouse anti-human CD3 (clone UCHT1), Alexa Fluor 700	BD Biosciences	RRID:AB_396952
Mouse anti-human CD107a (clone H4A3), PE-Cy5	BD Biosciences	RRID:AB_396136
Mouse anti-human IFNgamma (clone B27), FITC	BD Biosciences	RRID: AB_398580
Mouse anti-human MIP-1beta (clone D21-1351), PE	BD Biosciences	RRID: AB_393549
Mouse anti-human CD56 (clone B159), PE-Cy7	BD Biosciences	RRID:AB_396853
Mouse anti-human CD14 (clone MφP9), APC-Cy7	BD Biosciences	RRID:AB_396889
Mouse anti-human CD16 (clone 3G8), APC-Cy7	BD Biosciences	RRID:AB_396864
Anti-guinea pig complement C3 goat IgG fraction, FITC	MP Biomedicals	Cat # 0855385
anti-human IgG Fc antibody	Jackson ImmunoResearch	Cat # 109-035-190
Mouse Anti-Human IgG1-Fc PE	Southern Biotech	Cat # 9054-09
Mouse Anti-Human IgG2-Fc PE	Southern Biotech	Cat # 9060-09
Mouse Anti-Human IgG3-Hinge PE	Southern Biotech	Cat # 9210-09
Mouse Anti-Human IgG4-Fc PE	Southern Biotech	Cat # 9200-09
Mouse Anti-Human IgA1-Fc PE	Southern Biotech	Cat # 9130-09
Mouse Anti-Human IgM-Fc PE	Southern Biotech	Cat # 9020-09
VIC16 REFORMabs	This manuscript	n/a
Bacterial and Virus Strains
Mouse-adapted EBOV/Mayinga (EBOV/M.mus-tc/COD/76/Yambuku-Mayinga	Bray et al., 1998	n/a
Stellar competent cells	Takara	Cat # 636766
Biological Samples
Recombinant Vesicular Stomatitis Virus pseudotyped Ebola glycoprotein (rVSV pseudotyped EbOV gP)	IBT Bioservices	Cat # 1001-001
LowTox Guinea Pig Complement	CedarLane Labs	Cat # CL4051
Chemicals, Peptides, and Recombinant Proteins
Brefeldin A	Sigma Aldrich	Cat # B7651
GolgiStop	BD Biosciences	Cat # 554724
EBOV GPΔTM	IBT Bioservices	Cat # 0501-015
Polyethylenimine	Polysciences	Cat # 23966-1
Recombinant FcγRs (mouse and human)	Duke University Human Vaccine Institute	n/a
Recombinant murine FcγRI	R&D Systems	Cat # 2074-FC-050
Streptavidin-R-Phycoerythrin	Prozyme	Cat # PJ31S
Critical Commercial Assays
RosetteSep NK cell enrichment kit	Stem Cell Technologies	Cat # 15025
Luciferase Assay System	Promega	Cat # E1500
Deposited Data
The datasets generated from this study are available through Mendeley.	doi: 10.17632/9b2h8p9gtc.1
Experimental Models: Cell Lines
THP-1 monocytes	ATCC	RRID: CVCL_0006
FreeStyle 293F cells	ThermoFisher	Cat # R79007
VeroE6	ATCC	RRID: CVCL_0574
Experimental Models: Organisms/Strains
Mouse: Female BALB/c	Charles River; The Jackson Laboratories	Strain Code 028 (CR); RRID:IMSR_JAX:000651
Recombinant DNA
pUC19 vector	New England Biolabs	Cat # N3041S
Software and Algorithms
GraphPad Prism 8	GraphPad Software, Inc.	RRID:SCR_002798
R Studio	R Project for Statistical Computing	RRID:SCR_000432
Octet Data Analysis software version 8.1	Forte Bio/ Sartorius	https://www.sartorius.com/en/products/protein-analysis/octet-systems-software
Sprint Software	IBIS Technologies	https://www.ibis-spr.nl/product/ibis-triangle-software/
Flow Jo	BD Bioscience	RRID:SCR_008520
JMP Pro 15	JMP	RRID:SCR_014242
Kinetics Data Analysis software	Carterra	https://carterra-bio.com/applications/kinetics-software/
Other
anti-human IgG Fc (AHC) kinetic biosensors	Sartorius	18-5060
FluoSpheres® NeutrAvidin®-Labeled Microspheres, 1.0 μm, yellow-green fluorescent (505/515), 1% solids	Life Technologies	Cat #: F-8776
FluoSpheres® NeutrAvidin®-Labeled Microspheres, 1.0 μm, red fluorescent (580/605), 1% solids	Life Technologies	Cat #: F8775
PureProteome Protein G magnetic beads	Millipore	Cat # LSKMAGG10
MagPlex microspheres	Luminex corporation	Cat # MC12001-01, MCI12040-01, MCI10077-01

• Ebola virus disease survivors develop diverse profiles of antibody responses.

• Fc-engineering of monoclonal antibodies can replicate human functional profiles.

• A Fc-variant panel was generated using an anti-Ebola virus neutralizing Fab domain.

• Fc-variants with high complement yet moderate ADCC activity were protective in vivo.