Recent history has seen regular deadly zoonotic coronavirus spillover events with the emergence of severe acute respiratory syndrome coronavirus (SARS-CoV) in 2002 (1), Middle East respiratory syndrome (MERS) coronavirus in 2012 (2) and SARS-CoV-2 in December 2019 (3). SARS related coronaviruses are found in bats throughout Southeast Asia (4) and the serology of people living in proximity to a Rhinolophus spp bat colony suggests these zoonotic infections are not uncommon (5). Since the publication of the SARS-CoV-2 genome in January 2020 (3) there has been rapid development and deployment of vaccines for SARS-CoV-2 (6) and the clinical development of multiple SARS-CoV-2 spike specific neutralizing monoclonal antibodies (mAbs) from convalescent patients or animals, reviewed in (7).

Evolution of the SARS-COV-2 spike protein has selected for increased transmissibility, for example by increased affinity for host cells (8), with the emergence and then dominance of many new variants of concern (VOC) including Alpha, B.1.1.7; Beta, B.1.351; Gamma, P.1; Delta, B.1.617.2 (9) and most recently, Omicron, B.1.1.529 (WHO) that impact the neutralization efficacy of antibodies generated against the spike antigen of earlier strains (10). This includes profound escape from neutralization by some mAbs (11–15) and significant loss of neutralization activity of convalescent sera (13, 16, 17) and of humoral responses to first generation vaccines (12, 18, 19) reviewed in (10). Reinfection by neutralization-escape variants (20, 21) and break-through infection in vaccinees is now a feature of the pandemic (13, 19). Furthermore, protective antibody responses in humans are largely restricted to specific coronavirus species since few Abs to SARS-CoV-2 receptor binding domain (RBD) cross-neutralize SARS-CoV or MERS-CoV (22) but the recent identification of spike-specific broadly neutralizing mAbs may be a key to future pan-beta-coronavirus pandemic preparedness (23). Overall, despite increased surveillance and biosecurity (24) and the development of SARS-CoV-2 vaccines and mAbs, a critical vulnerability to variants of concern (VOC) and future pandemic novel coronaviruses persists. There is a need for prophylactic and therapeutic approaches that are more broadly effective against Sarbecoviruses.

Decoy proteins, based on the host entry receptor, inhibit viral entry, and achieve cross neutralization of multiple virus species or strains (25–31). Angiotensin-converting enzyme 2 (ACE2), is the principal entry receptor for the major human pathogenic coronaviruses, SARS-CoV and SARS-CoV-2 (32, 33), as well as the human endemic coronavirus NL63 (34). ACE2 is a transmembrane carboxypeptidase (35) with the ectodomain comprised of a catalytic domain, as well as a collectrin domain likely involved in dimerization (36). It normally plays a role in cardiovascular homeostasis by cleaving angiotensin II, the key agonist of the renin–angiotensin–aldosterone system (RAAS) that regulates blood pressure and electrolytes (37, 38). SARS-CoV-2 entry into host cells is blocked by the recombinant soluble ACE2 catalytic ectodomain in its native form (39, 40) or when engineered for higher avidity (31, 41, 42) or affinity (28, 29, 31, 41–45) for the ancestral spike, which has thus far been retained against later VOC (28, 29, 31, 41). Enzyme inactive forms of antiviral ACE2 decoy proteins have also been developed (26, 46, 47). However, ACE2 enzymatic activity, by cleaving angiotensin II, is protective in lung injury models and may therefore be beneficial to retain in an ACE2-based biological for COVID19 (38, 39, 48, 49).

To improve virus neutralization potency or pharmacokinetic properties of SARS-CoV-2 decoys, the ACE2 ectodomain, with or without the collectrin domain, has been fused to the Fc portion of IgG (26, 28, 29, 42–47, 50–55), which results in increased neutralisation potency by bivalency and increased serum half-life (56). In mAb studies using in vivo SARS-CoV-2 challenge models, Fc-dependent immune effector functions, which include antibody-dependent cellular cytotoxicity (ADCC), phagocytosis and clearance of viruses, bolstered protection against infection and pathology above that provided by neutralization alone (57–61). Similarly, ACE-Fc decoys have demonstrated protective activity in human ACE2 transgenic mouse models (29, 49, 62) and in hamsters (28) challenged with SARS-CoV-2.

We report the development of multifunctional antiviral proteins by applying novel mutations of the Fc and glycan modification to manipulate the Fc component of ACE2-Fc which resulted variously in increased virus neutralization, complement directed killing and activation of FcγRIIIa.

A series of ACE2-Fc fusion proteins (Table 1) were produced and analyzed for improved capacity to neutralize SARS-CoV-2 infection and to enhance or transform Fc-dependent effector functions attributed normally to the mechanisms of action of antibodies. Three versions of the ACE2 ectodomain were fused to the human IgG1 Fc portion. The first ACE2 fusion comprised the full length ACE2 ectodomain (flACE2-Fc, aa 19-740), including both the catalytic and collectrin domains (Figure 1A) and the second, an enhanced full length ACE2 ectodomain, EflACE2-Fc, with enhanced binding to SARS-CoV-2 S protein-RBD resulting from three amino acid mutations in the RBD binding site of the ACE2 protein (T27Y, L79T and N330Y) (43) and the third comprised a truncated ectodomain (trACE2-Fc, aa 19-615) containing the ACE2 catalytic domain but lacking the collectrin domain. Models of the trACE2-Fc (Figures 1C, D) and EflACE2-Fc (Figures 1A, B) decoy proteins were generated using Alphafold 2 (73, 74) and EflACE2-Fc was docked to SARS-CoV-2 spike RBD (Figure 1E). A comparison of the RBD docked to the Alphafold prediction of the EflACE2 and native ACE2 structures aligned similarly with the observed structure (PDB ID: 6M0J) with an RMSD of 2.733 Å and 2.940 Å respectively, and overlayed the RBD-EflACE2-Fc docked structure with an RMSD of 0.573 Å. This indicates that the affinity enhancing mutations do not impact the docking position of the SARS-CoV-2 spike RBD using this modelling approach. To evaluate the interaction with trimeric spike and assess the relative distance between ACE2 catalytic domains and adjacent RBD, the EflACE2-Fc-SARS-CoV-2 spike RBD or the trACE2-Fc-SARS-CoV-2 spike RBD model was overlayed on the RBD of chain A of the observed Spike-ACE2 complex structure (PDB ID: 7VXM) (Figures 1F, G). This showed that the ACE2 dimer in the EflACE2-Fc construct is not able to bind adjacent RBD on a single spike trimer due to distance restraints. Though the two ACE2 catalytic domains in the trACE2-Fc construct are likely not dimeric, through the lack of a collectrin domain (36), restraints imposed by disulphide bonding at the N-terminus of the Fc similarly act to restrict the distance between the ACE2 domains and likely also prevent binding to adjacent RBD for this construct (Figure 1G).

The key rationale for the development of a ACE2 decoy antiviral protein as a biosecurity agent against a future pandemic is the presumption it will similarly inhibit SARS-CoV-2 variants and ACE2 tropic coronaviruses generally. Indeed, using an ELISA flACE2-Fc bound near equally to both the ancestral RBD and RBD from the beta, gamma and delta VOCs (Figure 1H).

The activity of ACE2-Fc against SARS-CoV-2 variants was further addressed using a bead array. The inhibition of binding of biotinylated trACE2-Fc to an established array of 24 SARS-CoV-2 spike RBD variants (71) by unlabeled ACE2-Fc decoys was examined. Inhibition of binding to RBD-WT followed the hierarchy trACE2-Fc-WT < flACE2-Fc-WT < EflACE2-Fc-WT (IC₅₀ = 114, 80, 10 nM respectively) with effective inhibition of binding to all the individual RBD variants reached, with IC₅₀ values within two-fold of that observed with the ancestral RBD (Figure 2). Thus, across the array of RBD variants the average IC₅₀ values (110 ± 4; 86 ± 4 nM; 9.5 ± 0.9 nM) simply replicated this hierarchy of increasing neutralization potency over trACE2-Fc-WT as variants with increased affinity for ACE2 have equivalent increased susceptibility to inhibition by ACE2, including the N439K, S477N and E484K RBDs and other variants associated with escape from neutralizing antibodies (11, 54, 77). This contrasted sharply with the neutralizing mAb S35 where binding to the L455F and A475V RBD variants was abrogated. Furthermore, the decoy proteins were also effective inhibitors of binding to the spike proteins of the SARS and NL63 beta-coronaviruses (Figure 2). This illustrates the intrinsic resistance of ACE2 based antiviral decoys to escape by spike mutation and their applicability to other viruses that also use ACE2 for entry.

In addition to fusion to wild-type (WT) IgG1-Fc, these ACE2 formats were also fused to a Fc carrying novel substitutions of histidine 429 (Eu numbering) with phenylalanine (H429F) or tyrosine (H429Y), or in the adjacent residue, a known IgG hexamerising mutation E430G (78, 79). A glycan-modified form of trACE2-Fc was also produced in the presence of the mannosidase inhibitor kifunensine (trACE2-Fc-kif). The recombinant ACE2-Fc fusion proteins were purified first by anion exchange followed by size exclusion chromatography (SEC) and comprised largely a single species by SEC (Figure 3A) except the H429Y mutant Fc proteins which in all formats were resolved by SEC as oligomeric and monomeric species (Figure 3B).

Fc modification did not affect the intrinsic affinity for the SARS-CoV-2 spike RBD (e.g. trACE2-Fc-WT, K_D = 28.6nM; flACE2-Fc-WT, 25.2 nM; flACE2-Fc-H429F, K_D = 23.2 nM, Figures 3C-E) which was comparable with that of the reported affinity 22nM for the flACE2 (43). As expected, the EflACE2-Fc WT protein with the enhanced RBD-binding mutant ACE2 domain showed a ~30-fold increase in affinity to K_D = 0.7 nM (Figure 3F) (43) compared to the flACE2-Fc.

Native PAGE (N-PAGE) analysis showed that ACE2-Fc WT fusion proteins migrate as a single species, at ~ 260 kDa for trACE2-Fc and at > 260 kDa for the flACE2-Fc and EflACE2-Fc fusion proteins, reflecting the additional presence of the collectrin domain (Figure 3G). Notably the ACE2-Fc-H429Y variants (e.g. trACE2-Fc-H429Y Fc, Figure 3G, 5^th trACE2-Fc lane “Y”) migrated in N-PAGE as several distinct higher molecular weight oligomer species, that were not apparent in denaturing SDS-PAGE, i.e. these comprise non-covalent oligomers. N-PAGE shift analysis showed that the normal and enhanced ACE2 (e.g. trACE2-Fc and EflACE2-Fc) proteins, and the Fc mutants, had high-specific binding activity for SARS-CoV-2 spike RBD, visualized by their shift to high molecular weight complexes following interaction with SARS-CoV-2 spike RBD-Ig (RBD-Ig “+” lanes, Figure 3G). When quantified by ELISA the ACE2-Fc proteins bound the bivalent ligand RBD-Ig with subnanomolar avidity and were unaffected by mutation of the Fc, excepting the oligomer forming H429Y Fc mutants which exhibited weaker binding (Figure 3H).

The antiviral activities of the ACE2-Fc fusion proteins were determined in a microneutralization assay using SARS-CoV-2 infection of Vero cells (64) where the EC₅₀ endpoint corresponds to neutralization of ~99% of the inoculum virions (66). The SARS-CoV-2 neutralization endpoint (EC₅₀ 2.70 µM) of the unfused truncated ectodomain (trACE2 alone) was improved ~10-fold by its fusion with the unmodified wildtype Fc region of IgG1 (trACE2-Fc-WT, EC₅₀ 283 nM), consistent with its improved binding avidity (Figure 4A). In accord with its increased intrinsic affinity for the RBD (Figure 3F), the EflACE2-Fc-WT (EC₅₀ 11 nM) was a further ~11 and 25-fold more inhibitory than the unmodified flACE2-Fc-WT (EC₅₀ 124 nM) and trACE2-Fc-WT respectively (Figures 4A–C). Thus overall, EflACE2-Fc-WT (Figure 4C) was ~240-fold more active in virus neutralization than trACE2 alone (Figure 4A).

Of the five Fc modifications, the oligomeric (og) form of the H429Y Fc mutants fused with any ACE2 format, consistently displayed superior neutralization activity within its ACE2 format class. Thus, the oligomeric trACE2-Fc-H429Y_og, isolated by SEC, had a neutralization activity (EC₅₀ 21.9 nM) that was 13-fold improved over the monomeric trACE2-Fc-WT (EC₅₀ 283nM, Figure 4A). Similarly, flACE2-Fc-H429Y_og (EC₅₀ 10.0 nM) showed greater potency than flACE2-Fc-WT (EC₅₀ 124 nM) (Figure 4B). Indeed, it was equivalent in neutralization activity to the EflACE2-Fc WT neutralization (EC₅₀ 10.6 nM). Finally, the most potent inhibitor, EflACE2-Fc-H429Y_og (EC₅₀ 4.23 nM) (Figure 4C), was ~ 600-fold more active than the monovalent trACE2 (Figure 4A). This improved neutralization by the H429Y decoy contrasted with the H429F and the E430G modifications which did not significantly alter SARS-CoV-2 neutralization activity in any ACE2-Fc format (Figures 4A-C). As a comparator the laboratory equivalent of the therapeutic mAb REGN 10933 (casirivimab) had an EC₅₀ of 3.6 nM, (n = 2).

The Fc receptors of leukocytes and serum complement provide the two major effector systems harnessed normally by the Fc portion of antibodies. FcγR functions, which may include ADCC, phagocytosis and clearance of opsonized viruses are important antiviral effector mechanisms and are increasingly found to play a protective role during SARS-CoV-2 infection (57, 58, 60, 61, 80, 81). The interaction of FcγRIIIa with the ACE2-Fc fusion proteins was evaluated by flow cytometry using Ramos cells expressing SARS-CoV-2 spike protein (Ramos-S cells) opsonized with the different formats of ACE2-Fc. Dimeric recombinant soluble FcγRIIIa (68) bound the Fc-WT and H429F and E430G mutant fusion proteins within each ACE2 format class near equivalently. However, the H429Y mutation of the Fc largely ablated FcγR binding in all the ACE2-Fc formats (Supplementary Figure S1A). The loss of FcγRIIIa binding was not due to lack of opsonization of the Ramos-S cells by the H429Y variants as all ACE2-Fc proteins showed similar binding of the SARS-CoV-2 spike protein on the cell surface (Table 2).

Next, FcγRIIIa activation by ACE2-Fc fusion proteins was evaluated as a validated surrogate of ADCC (67). FcγRIIIa was activated by Ramos-S cells opsonized with a Fc-WT fusion of any ACE2 format (Figure 5A). The flACE2-Fc-WT induced FcγRIIIa-mediated activation of the reporter cell at 2.7-fold lower concentration than the trACE2-Fc-WT (EC₅₀ 1.7 nM) (Figure 4A, p < 0.0001, Supplementary Figures S1B, C), indicating that inclusion of the ACE2 collectrin domain, improved FcγRIIIa activation. Notably, the increased affinity of the EflACE2-Fc WT for SARS-CoV-2 spike protein, did not increase FcγRIIIa activation above that of flACE2-Fc WT (Figure 5A).

However, the most potent FcγRIIIa activation was achieved following glycan-modification by kifunensine (82) during the production of the trACE2-Fc. Thus, despite the lower activity of the trACE2-Fc format, FcγRIIIa activation by trACE2-Fc-kif exceeded that of the flACE2-Fc and EflACE2-Fc and approached that of the therapeutic anti-CD20 mAb rituximab used as a comparator on the CD20⁺ Ramos-S cells (Figure 5A; Supplementary Figure S1B). Thus, the hierarchy of FcγRIII activation by the proteins was trACE2-Fc-kif > EflACE2-Fc WT ~ flACE2-Fc- WT > trACE2-Fc-WT.

In accord with the FcγRIIIa binding data (Supplementary Figure S1A), modification of ACE2-Fc decoys by the H429F or E430G mutation had only modest effects on FcγRIIIa activation (Figure 5A; Supplementary Figures S1B, C). In contrast, the H429Y mutation in all ACE2-Fc formats ablated FcγRIIIa activation of cells which is consistent with their abrogated binding to FcγRIIIa (Supplementary Figure S1A). Thus, while enhancing virus neutralization, the H429Y modified Fc in trACE2-Fc, flACE2-Fc and EflACE2-Fc formats were largely inactive in FcγR binding and consequently unable to activate cells via FcγRIIIa (Figure 5A).

The second major Fc-dependent effector system is the classical complement pathway. The activation of complement by the ACE2-Fc proteins was tested initially by ELISA for the capacity to fix complement components C5b-9 and then to mediate complement-dependent killing of cells expressing SARS-CoV-2 spike protein. The fixing of C5b-9 (Supplementary Figure S1D) which forms the membrane attack complex, was achieved by all Fc fusions but was enhanced by both the H429F Fc mutation and the hexamerising E430G mutation of trACE2-Fc compared to unmodified Fc-WT. Despite the H429Y mutated Fc preforming oligomers, which might be anticipated to confer superior complement fixation, this was not apparent and the trACE2-Fc-H429Y oligomer form, showed similar C5b-9 fixation as the trACE2-Fc-WT and glycomodifed trACE2-Fc-kif, (Supplementary Figure S1D).

Despite the ELISA showing that the trACE2-Fc-WT fusions with an unmodified Fc fix C5b-9, analysis of cell killing showed that the unmodified Fc-WT fusion proteins of any ACE2 format failed to mediate significant CDC (Figure 5B). In stark contrast to this CDC inactivity, both the H429F and the hexamerising E430G Fc mutants of trACE2-Fc fusion proteins were remarkably active in mediating complement lysis of Ramos-S cells (Figure 5B). The monomeric form of trACE2-Fc-H429Y_mn was active, although substantially less potent than the H429F mutant. H429 in the Fc is thus a site for modification that remarkably potentiates the Fc’s capacity for stimulating complement-mediated target lysis.

In this study we examined different antiviral functions of ACE2-Fc virus decoy proteins. As in IgG antibodies, the Fc-region drives the effector responses mediated by the IgG-Fc fusion proteins. In vivo SARS-CoV-2 challenge models (57–61) have found Fc immune functions of antibodies decreased virus load, spread from nasal tissue to major organ systems, cytokine storm and inflammation, and mortality. In contrast, ablating Fc function resulted in increased disease severity or mortality (58–60, 83). Recently a non-neutralizing human mAb with Fc-enhanced ADCC activity conferred partial protection in a SARS-CoV-2 infection model and contributed to complete protection in combination with a neutralizing mAb (80). Indeed, ADCC potency is an indicator of humoral responses that protect against severe disease in humans (81). While complement activation features in the pathophysiology of severe COVID-19 it is likely to be initially protective (84) and is an identified function of anti-SARS-CoV-2 therapeutic antibodies (85). The Fc portion is thus an important element to optimize for the development of ACE2-Fc as an anti-SARS-CoV-2 antiviral molecule and for viral entry receptors fused to Fc more generally.

Hence, we have manipulated three major antiviral activities of ACE2-Fc by modifying its Fc portion to enhance the existing decoy (neutralization) action of the ACE2 component, complement mediated killing and activation of FcγR. Firstly, the H429Y mutation, in the Fc CH3 domain outside the Fcγ receptor or complement contact sites of the CH2 domain, resulted in the formation of oligomers of the decoy protein which resulted in improved neutralization potency. The improved neutralization activity and oligomeric nature of the H429Y Fc mutant decoys mimic the polymeric antibody classes, IgA (86) and IgM (87) where avidity contributes to the efficacy of SARS-CoV-2 neutralization. Fc : Fc interactions are a recognized property of IgG antibodies (88) and their stabilization by mutation can lead to the formation of in solution oligomers (89). In contrast, the E430G modification of IgG is known to promote “on-target” oligomerization (hexamerization) of IgG (79), but did not significantly alter SARS-CoV-2 neutralization activity in any ACE2-Fc format. In contrast, the H429Y Fc mutation enhanced neutralization potency of all formats of the ACE2-Fc decoy proteins. H429Y Fc mutation in combination with the inclusion of the collectrin domain and the triple ACE2 mutations enhancing affinity for S (43), (i.e. EflACE2-Fc-H429Y) resulted in an overall 600-fold increased SARS-CoV-2 neutralization potency over that of the monomeric truncated ACE2 domain. The neutralization potency of EflACE2-Fc-H429Y (4.2 nM) was comparable to that of the laboratory equivalent of the therapeutic mAb REGN 10933 (casirivimab, 3.6 nM). A feature of the H429Y mutation was the loss of binding by FcγRIIIa. Mutations at the CH2/CH3 interface can affect low affinity FcγR binding to the Fc (90), suggesting these sites, though distant, can affect each other (91).

Secondly, the phenylalanine substitution of histidine 429 (H429F) of the ACE2-Fc proteins did not enhance neutralization but did transform CDC against S expressing targets. This improved CDC activity was like that of E430G mutated ACE2-Fc, a known “on-target” Fc-hexamerising mutation, a format optimal for C1 binding and activation (78, 79). Lastly, FcγR potency of trACE2-Fc was improved by modifying the Fc glycan (82) to enhance FcγRIII binding (92). It is likely that similar treatment of flACE2-Fc WT and EflACE2-Fc WT, or alternatively, amino acid substitution to increase affinity for FcγRIIIa (93), would similarly further improve their FcγRIII activating potency. Notably, FcγRIII activation was a little reduced for the decoy lacking the collection domain, indicating the formatting of Fc-fusion proteins can impact Fc-mediated activity.

We have demonstrated ACE2-Fc to be a potent agent against SARS-CoV-2, not only for neutralization but also for the harnessing of Fc-mediated effector functions. The exemplar Fc modifications demonstrated herein illustrate the potential for the tuning of Fc function to optimize virus neutralization, FcγR interaction and complement activation. This selection of desired functional profiles could aid the deployment of broadly effective ACE2-Fc, mAbs and other Fc therapeutics. There has been a rapid progression of multiple different SARS-CoV-2 mAbs to clinical use that is likely to herald increased deployment of mAbs clinically for infectious diseases. The optimization of Fc functions will make a significant difference to their clinical success. Furthermore, the world remains susceptible to new pandemics and vaccine escape variants. Thus, an antiviral decoy comprising optimized Fc fusion to a viral entry receptor such as ACE2-Fc, is an important option for deploying a rapid first line of defense to contain new zoonotic viral threats while vaccines, mAbs and antiviral drugs are being developed.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. PDB coordinate files are available from the authors on request.

BW and PMH conceived and planned the experiments. BW, LK, HT, SE, EL, KC, L-JC, FM, WL, NG, SP, GH, PP, and JC performed the experiments and analyzed the data. JB, DG, LB, MvZ, AW, AC, W-HT, KS, SK, and PMH provided supervision and analyzed the data. BW and PMH wrote the manuscript with input from all other authors. All authors contributed to the article and approved the submitted version.

The Medical Research Future Fund (MRFF 2002073) and Victorian State Government COVID research funding supported the research of AW, DG, W-HT, SK, PH, and SP, LB (1175865) with contribution from the Victorian Operational Infrastructure Support Program and Australian Government NHMRC Independent Research Institutes Infrastructure Support Scheme. NHMRC project grants supported PH, BW. (1145303). Investigator Grants are held by JB (1173046), KS (1177174) and DG (2008913) and program grants by SK (1149990), LB (1055214). DG (1117766), W-HT, SK, MZ (1117687) and AW receive National Health and Medical Research Council (NHMRC) fellowships. DG is supported by the Australian Research Council (ARC; CE140100011). NG was supported by an ARC DECRA Fellowship (DE210100705). SP and LB funded by a Heart Foundation Vanguard Grant (105798). The Melbourne WHO Collaborating Centre for Reference and Research on Influenza is supported by the Australian Government Department of Health. W-HT is a Howard Hughes Medical Institute–Wellcome Trust International Research Scholar (208693/Z/17/Z).

Authors PMH and BW are inventors on a provisional patent filing by the Burnet Institute.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.