Engineering a “muco‐trapping” ACE2‐immunoglobulin hybrid with picomolar affinity as an inhaled, pan‐variant immunotherapy for COVID‐19

Abstract Soluble angiotensin‐converting enzyme 2 (ACE2) can act as a decoy molecule that neutralizes severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) by blocking spike (S) proteins on virions from binding ACE2 on host cells. Based on structural insights of ACE2 and S proteins, we designed a “muco‐trapping” ACE2‐Fc conjugate, termed ACE2‐(G4S)6‐Fc, comprised of the extracellular segment of ACE2 (lacking the C‐terminal collectrin domain) that is linked to mucin‐binding IgG1‐Fc via an extended glycine‐serine flexible linker. ACE2‐(G4S)6‐Fc exhibits substantially greater binding affinity and neutralization potency than conventional full length ACE2‐Fc decoys or similar truncated ACE2‐Fc decoys without flexible linkers, possessing picomolar binding affinity and strong neutralization potency against pseudovirus and live virus. ACE2‐(G4S)6‐Fc effectively trapped fluorescent SARS‐CoV‐2 virus like particles in fresh human airway mucus and was stably nebulized using a commercial vibrating mesh nebulizer. Intranasal dosing of ACE2‐(G4S)6‐Fc in hamsters as late as 2 days postinfection provided a 10‐fold reduction in viral load in the nasal turbinate tissues by Day 4. These results strongly support further development of ACE2‐(G4S)6‐Fc as an inhaled immunotherapy for COVID‐19, as well as other emerging viruses that bind ACE2 for cellular entry.


| INTRODUCTION
Despite the ready availability of numerous vaccines against COVID-19, there continues to be major demand for safe and effective antivirals that are easily administered, particularly among high-risk individuals, and do not suffer risk for drug-drug interactions. 1Furthermore, in light of the ever-shifting landscape of immune evasive variants of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), there is a clear need to develop pan-variant antiviral therapies that may be broadly effective against a spectrum of coronaviruses or other respiratory viruses.SARS-CoV-2 gains entry into cells when the receptor binding domain (RBD) of its spike protein (S) binds to angiotensinconverting enzyme 2 (ACE2) on the target cell's surface, 2 similar to SARS-CoV-1 3 and NL63-CoV. 43][14][15] Unfortunately, since S-proteins only bind ACE2 with modest affinity, the neutralizing potencies of such ACE2 decoys are also rather modest. 5,6To overcome the limited affinity, several groups have engineered higher affinity ACE2 variants by random mutagenesis and selection using yeast surface display. 5,6,8Nevertheless, the directed evolution strategy leaves open the possibility of an escape SARS-CoV-2 variant that binds wildtype ACE2 receptor but not the mutated ACE2 molecule.A mutated ACE2 may also have lower affinity to other viruses that bind ACE2.Alternatively, it is possible to increase the overall avidity of the molecule to the virus by increasing the valency of ACE2 binding domains per molecule 12,[16][17][18] ; however, this would likely result in much larger constructs that are in turn more difficult to manufacturer, stabilize, and formulate.We thus sought a strategy to improve the binding of wildtype ACE2 to the virus without changing the sequence of ACE2 from the native human protein, while maintaining the same binding domain valency as conventional IgGs.
Cryo-electron microscopy of SARS-CoV-1 showed that $50-100 S proteins are present on the virus surface, with an average spacing of $15 nm. 18The trimeric form of the S-protein spike also results in a large distance between any two of the three S protein monomers within an individual spike trimer.Such distances limit the ability for a single antibody or ACE2-conjugate to simultaneously bind to two distinct S-proteins.These structural insights motivated us to improve the binding of our ACE2-decoy by tuning the presentation of the two ACE2 domains.Specifically, we replaced the VH-CH1 domain of the standard IgG1 Fab with the extracellular domain of ACE2 only (residues 18-614; no self-dimerizing collectrin domain), and further introduced an extended 30-amino acid flexible linker between the ACE2 fragment and CH2 in order to increase the reach of each of the two ACE2 arms, which should increase the probability of bivalent binding across spike trimers (Figure 1c).SARS-CoV-2, 19 just like SARS-CoV-1 20 and NL63 and HKU1 coronaviruses, 4,[21][22][23] infects strictly via the apical side of airway epithelium (i.e., airway lumen), and predominantly sheds progeny virions back into airway mucus (AM) overlaying the epithelium.5][26] Building off of our previous discovery that the Fc domain of IgG1 possess a weak affinity to mucins that can enable potent trapping of viruses in mucus, 24,27,28 we engineered here a "muco-trapping" ACE2-immunoglobulin construct (ACE2-(G 4 S) 6

| Designing ACE2-(G 4 S) 6 -Fc
To establish how ACE2 may associate with a trimeric Spike protein, we first generated the reported spike protein structures in the "three up" RBDs conformation. 29We then determined whether an ACE2-Fc molecule could bind to two RBDs on the same S protein trimer (Figure S1a).When one of the two ACE2 domains on ACE2-Fc engages any one of the three RBD in the spike trimer, the remaining ACE2 domain becomes oriented upwards, away from the S protein, due to the lack of flexibility and length in the hinge of IgG1 connecting ACE2 to Fc.Thus, it is not likely that the ACE2-Fc can bind bivalently to two RBDs on the same Spike protein (or to two RBDs on two different Spike proteins on the virus).The same limitation holds for the conventional ACE2-Fc encompassing the collectrin domain, since the dimerization of the collectrin domain limits the reach of the adjacent ACE2 fragments.
The estimated distances between RBDs within a spike trimer ranged from 60 to 100 Å when the three RBDs were in the "three-up" conformation.To bridge the distance and add flexibility to the molecule, we introduced a (GGGGS) 6 flexible linker with a length of $10 nm between the extracellular ACE2 fragment and IgG1-Fc to form bivalent ACE2-(G 4 S) 6 -Fc (Figure 1c, Figure S1b.Since the flexible linker is present on each of the two heavy chains, the two ACE2 fragments on ACE2-(G 4 S) 6 -Fc can theoretically span distances nearly twice that length, that is, up to $20 nm.

| Biophysical characterization of ACE2-Fc and ACE2-(G 4 S) 6 -Fc
We next proceeded to produce and purify both ACE2-Fc and ACE2-(G 4 S) 6 -Fc.We first analyzed both molecules in Native-PAGE; the single bands for both confirmed that they existed as monomers, although they ran at an apparent molecular weight $350 kDa (Figure S1c) instead of the expected theoretical 194 kDa.We next analyzed their molecular weight using size exclusion chromatography/ multi-angle light scattering (SEC/MALS).We found ACE2-Fc and ACE2-(G 4 S) 6 -Fc possessed MW of $208 kDa and $212 kDa, respectively, in good agreement with their theoretical size (Figure S1e).Both molecules were predominantly found in the monomeric form: ACE2-Fc and ACE2-(G 4 S) 6 -Fc were $85% and $91% monomer after simple Protein A purification, respectively, with the remaining fraction corresponding to oligomers of the proteins and aggregates.We   2 was used to generate all the protein models, and UCSF Chimera X 1.1 3 was used to render the model for publication.ACE2-Fc and ACE2-(G 4 S) 6 -Fc were constructed using models 6M17 for ACE2, 4 1HZH for human IgG, 5 and 1EIB for GGGGS linker. 6CE2-Fc and ACE2-(G 4 S) 6 -Fc bound to S protein were generated by matching the ACE2 of ACE2-Fc and ACE2-(G 4 S) 6 -Fc with the RBD-bound ACE2 in the "all-up" S protein model 7A98.7 The predicted 3D model of ACE2-Fc with collectrin domain was modified from Reference 1. ACE2, angiotensin-converting enzyme 2; RBD, receptor binding domain.
consistently obtained greater yields with ACE2-(G 4 S) 6 -Fc production, with an average yield of $86 mg per 500 mL of transiently transfected culture, which is more than double our typical yield with ACE2-Fc under identical conditions ($36 mg of protein per 500 mL of culture; Figure S1d).Finally, we assessed the stability of the molecules using differential scanning calorimetry.The melting temperature T M for ACE2-(G 4 S) 6 -Fc was $52 ± 0.6 C (Figure S3).This result is comparable to previously reported values for the T M of ACE2 decoys: $52 C for truncated ACE2-Fc (residues 18-640) and $55 C for full length ACE2-Fc with the collectrin domain (residues 18-740). 8
The key advantage of ACE2 decoys over conventional monoclonal antibodies (mAbs) is the ability to bind any SARS-CoV-2 variant.To confirm ACE2-(G 4 S) 6 -Fc can indeed bind different SARS-CoV-2 variants, we also measured initially its binding affinity to B.1.1.7 (UK) and B.1.351(SA) spike proteins using ELISA (Figure 2c), and, later, against more recent variants including the Omicron family (Figure 2d).The binding affinity of ACE2-(G 4 S) 6 -Fc to all tested variants was highly comparable (Figure 2d).The ACE2-(G 4 S) 6 -Fc evaluated in Figure 2d for breadth of binding across variants of interest was produced using stable CHO pools that have since been advanced into cGMP development.As a control, we also evaluated the binding of RGN10933, a mAb developed by Regeneron that is part of the REGN-COV2 mAb cocktail that originally received EUA from the FDA prior to the rise of evasive variants.While RGN10933 bound WT and alpha variant S proteins with comparable binding affinity in our ELISA assay ($0.3 ± 0.04 ng/mL and $0.3 ± 0.08 ng/mL for WT and alpha variants, respectively), we were unable to detect binding against the beta variant, in line with prior reports. 16These results underscore the utility of ACE2-(G 4 S) 6 -Fc across all SARS-CoV-2 variants.

| Neutralization potencies of ACE2 decoys in pseudovirus infection assays and against live virus
We next sought to determine whether the increased apparent binding affinity of ACE2-(G 4 S) 6 -Fc correlates with greater neutralizing activity.
We measured the neutralization potencies of ACE2-(G 4 S) 6   RdRp stands for the RNA-dependent RNA polymerase, which was the gene target of the RT-PCR used in this experiment.IC 50 for Alpha and Gamma variants of 0.96 μg/mL and 0.57 μg/mL, respectively.ACE2, angiotensin-converting enzyme 2; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2. N.S., *, and *** represent "non-significant", p < 0.05, and p < 0.001, respectively.SARS-CoV-2 VLPs exhibited rapid diffusion in AM, as reflected by the high mean-squared displacement (MSD) values and the slope of the MSD versus time scale plot (Figure 4a).In contrast, ACE2-(G 4 S) 6 -Fc effectively trapped SARS-2 VLP in AM, as quantified by the markedly lower MSD values and lower slope value.This corresponds to a $10-fold reduction in the average hD eff i of the SARS-CoV-2 VLPs compared to naïve AM (Figure 4b), and a nearly 14-fold reduction in the fast moving viral populations, defined as the fraction of viruses possessing sufficient diffusivity to diffuse across $50 μm layer in $1 h) even at just 1 μg/mL conc. in AM (Figure 4c).In contrast, neither ACE2-Fc nor CR3022, a high-affinity mAb against S protein, were unable to reduce viral mobility to the same extent as ACE2-(G 4 S) 6 -Fc.
We then proceeded to assess whether ACE2-(G 4 S) 6 -Fc could facilitate muco-trapping at lower concentrations.We observed a slight dosedependent trapping of SARS-CoV-2 VLPs, with slightly more effective trapping at higher ACE2-(G 4 S) 6 -Fc concentrations, although the difference was not statistically significant (Figure 4d,e).The fraction of fast-moving SARS-CoV-2 VLPs was reduced at all ACE2-(G 4 S) 6 -Fc concentrations tested (Figure 4f).
We were interested in investigating whether ACE2-(G 4 S) 6 -Fc could similarly trap other ACE2-targeted virions in AM.Since the causative agent of SARS, SARS-CoV-1, also binds to human ACE2 for cellular entry, we next assessed whether ACE2-(G 4 S) 6 -Fc could trap SARS-CoV-1 VLPs in human AM, and again found ACE2-(G 4 S) 6 -Fc mediated potent trapping of SARS-CoV-1 VLPs in human AM (Figure S2).This further substantiates the broad value of this decoy approach for any pathogen that may target human ACE2, not just variants of SARS-CoV-2.
The most direct method to achieve therapeutic concentrations of mAb in the respiratory tract, particularly the lung airways, is to directly deliver the mAb via inhalation. 30Vibrating mesh nebulizers (VMNs) can nebulize protein therapeutics without generating local heating and shearing that can degrade proteins.We thus tested whether ACE2-(G 4 S) 6 -Fc could be stably nebulized using a VMN.Using a Philip's InnoSpire Go VMN, we nebulized ACE2-(G 4 S) 6 -Fc, collected the resulting aerosols using a two-chamber twin glass impinger setup designed to capture aerosols >6.4 μm (upper chamber) and <6.4 μm (lower chamber) following European Pharmacopeia 5.0, 31  (Figure 5).Native PAGE also confirmed there was no separation of the heavy chains or detectible aggregation (Figure S3) and differential scanning fluorimetry showed a peak on first derivative ratio around 54 C.
5][36] Here, we first evaluated initiating daily intranasal dosing of ACE2-(G 4 S) 6 -Fc starting at 4 and 24 h postinfection.While we did not observe significant reduction of the viral titers in the deep lung, we saw highly impressive reduction of the viral titers in the nasal turbinates (Figure 6).The majority of the animals receiving treatment starting at 4 and 24 h postinfection had viral titers reduced to below the detection threshold, and the mean reduction ranged from $40-fold to $20-fold ( p < 0.05), respectively (Figure 6a).The treated animals also exhibited substantially less weight loss compared to saline control.
Building off these results, we further attempted initiating treatment at 48 h postinfection.Surprisingly, despite this very late initiation of therapy, ACE2-(G 4 S) 6 -Fc treatment still provided a $10-fold reduction in viral load in the nasal turbinate tissues at the 96 h postinfection time point (Figure 6b).The reduction in viral titers again translated to substantial protection against weight loss over the 2-day period ( p = 0.03).were consistently $25-30 ng/mL against S-proteins from various SARS-CoV-2 variants (Figure 2d).These binding affinities are comparable to a number of mAbs that have demonstrated clinical utility against sensitive SARS-CoV-2 strains.

| DISCUSSION
As part of the unprecedented global effort to develop effective antiviral therapies against SARS-CoV-2, there were multiple neutralizing mAbs that were quickly advanced into clinical development.Indeed, mAbs were the first class of therapeutics approved as antivirals for COVID-19, and the fact that every one of the mAbs advanced into Phase 2/3 studies consistently saw strong efficacy underscores the overall effectiveness of antiviral mAbs.Indeed, when administered early in the course of infection, the antiviral mAbs were generally highly effective ($70%-85%) at reducing risk of hospitalization and death. 37,380][41][42] This underscores the limited effectiveness of a mAb cocktail approach in preventing viral escape against SARS-CoV-2.
Given the concerns about escape mutants, it would be highly advantageous to develop a single mAb-based product that is not at risk of viral escape.Furthermore, given that there are already three human coronaviruses that target ACE2 as the primary host entry receptor (SARS-CoV-2, SARS-CoV-1, and HCoV-NL63), we consider it quite plausible that another ACE2-targeting respiratory virus will emerge with pandemic potential.Soluble ACE2 decoys can block infections by both SARS-CoV-1 43 and SARS-CoV-2. 9,44For these reasons, we, along with a number of other groups, became interested in developing ACE2 decoys, with the goal of enabling a universal immunotherapy against all ACE2-targeted viruses.Contrary to the strategy of engineering ACE2 mutants for improved affinity to S protein, we made a concerted approach to utilize only the WT ACE2 fragment, thus eliminating the potential risk of viral escape mutants that bind WT ACE2 but cannot be captured by the ACE2 mutant.We also believe maintaining Fc effector function to be critical to the therapeutic efficacy of antiviral mAbs, and thus retained the wildtype IgG1-Fc.These considerations led us to (i) eliminate the collectrin domain on ACE2, (ii) optimize the linkage between the extracellular fragment of ACE2 with the IgG1-Fc, and (iii) incorporate a "muco-trapping" Fc.We found that ACE2-(G 4 S) 6 -Fc, engineered with these features, not only was able to achieve exceptionally consistent binding to all variants of SARS-CoV-2, but also could be stably nebulized and trap SARS-CoV-2 VLPs in human AM.These features likely contributed to the effective suppression of viral load in nasal turbinates and clinical signs in hamsters in this study, even when dosed as late as 48 h postinfection.These results strongly substantiate the promise for advancing ACE2-(G 4 S) 6 -Fc into clinical development; we are currently undertaking the necessary preclinical activities and GMP manufacturing to do so.

Early work with ACE2 decoys against SARS-CoV-2 infection
showed that simply linking ACE2 to an Fc domain inhibited infection in vitro with a modest EC 50 of $24 μg/mL, while retaining effector function of the Fc. 13 This finding motivated investigators to pursue strategies to further improve ACE2 decoys.One common approach is to conduct directed evolution to introduce targeted mutations to recombinant ACE2 using structural knowledge and screening of mutant libraries. 5,6,8These strategies resulted in engineered ACE2 decoys with EC 50 s to S protein closer to 150 ng/mL. 45One study tested an inhaled engineered ACE2 decoy in a K18-hACE2 transgenic mouse model of COVID-19, which prolonged survival and reduced viral load consistent with our findings in hamsters in this work. 46In contrast with using targeted mutations to increase affinity of ACE2 decoy molecules, other groups have sought to improve affinity by increasing ACE2 valency, ranging from dimeric ACE2-Fc (fusion of ACE2 to IgG1-or IgG4-Fc), trimeric ACE2, 15,47 tetrameric ACE2, 12 to hexameric constructs via the oligomerization motif from an IgM Fc domain to create hexavalent ACE2-based molecules, 48 or via incorporating an IgM tailpiece (PTLYNVSLVMSDTAGTCY) that enable pentamerization or hexamerization of IgGs. 17Indeed, trimerization of Fc-null ACE2 decoys by fusing ACE2 to T4 foldon or helix bundle through flexible linkers increased the affinity of the molecule to SARS-CoV-2 S protein by 5-to 10-fold compared to the monomer, and shifted the IC 50 from >7500 ng/mL to hundreds of ng/mL.One report of an IgM-like inhalable ACE2 fusion protein was shown to neutralize SARS-CoV-2 variants with an IC 50 of $13-131 ng/mL. 17In another study, hexavalent ACE2-based molecules were demonstrated an IC 50 of 0.18 ± 0.04 nM. 48The potencies achieved herein with ACE2-(G 4 S) 6 -Fc are comparable to those previously observed with the trimeric ACE2 and the "IgM" versions of ACE2 presentation on a mass concentration basis.While our strategy is more akin to the multimerization approach, it differs in that we believe two ACE2 fragments per molecule would confer sufficient binding avidity for the molecule to be highly active at concentrations we can readily achieve physiologically, if we could extend the reach of the two ACE2 fragments.Indeed, by utilizing a longer extended linker, ACE2-(G 4 S) 6 -Fc achieved binding affinity and neutralization potencies comparable to many of the affinity-matured or highly multimerized ACE2 decoys, and conferred effective protection of the upper respiratory tract in vivo.Finally, we are also unique in our focus to engineering molecules with a "muco-trapping" Fc.
Intranasal delivery of ACE2-(G 4 S) 6 -Fc conferred highly effective suppression of SARS-CoV-2 in the nasal turbinates of hamsters, the key site for virus replication and dissemination, even when treatment was withheld until 48 h after infection.Notably, prior studies in the same hamster model have shown that intravenously administered mAbs were unable to effectively reduce viral load in the nasal turbinates of hamsters, despite some protection in the lower airways. 33We believe the lower respiratory tracts in rodents, particularly the deep lung, have greater blood permeability than the upper airways; this enhances the relative basal-to-apical mAb distribution, which in turn enabled modest protection.In contrast, following intranasal delivery, we were likely achieving and sustaining highly inhibitory concentrations of ACE2-(G 4 S) 6 -Fc in the nasal turbinates, which reduces the viral load locally.These results directly demonstrate that topical treatment of a mAb-based antiviral could afford meaningful antiviral activity.
Due to stark anatomical differences in the respiratory tract compared to humans, hamster is typically not used for inhalation studies; instead, models like dogs are frequently utilized.Unfortunately, we were limited to the fact that hamster was the only available animal model for SARS-CoV-2 infection at the time, and to this day represents the gold standard utilized by nearly all candidate therapies for COVID-19.As a nebulization setup was not available in the BSL-3 facility, we could not actually evaluate direct nebulized delivery of our molecule into the lungs of the animals.This limited us to administering our treatments by intranasal dosing.While adequate for delivery to the nasal turbinates, intranasal administration notoriously results in spotty distribution within the lungs of rodents.Thus, we consider the limited anti-viral effects observed in the lungs of hamsters in this work is likely an artifact from non-uniform pulmonary distribution following intranasal instillation of drug, a notorious challenge with intranasal delivery to rodents.Such non-uniform delivery will likely be resolved by the use of VMNs in human studies, which offers control of delivery throughout the entire respiratory tract.
ACE2 preferentially traffics to the apical membrane of airway epithelium; this is consistent with the apical infection and apical shedding observed with SARS-CoV-1 20 and SARS-CoV-2. 19This pathophysiology also means that viruses are concentrated either within the luminal airway epithelium or in the AM directly overlaying the cells.In turn, they are difficult to reach by systemically-dosed antivirals, particularly biologics.Indeed, antiviral mAbs take days to reach Cmax locally in the respiratory tract, and typically far less than 1% of the injected dose actually does so.For instance, a Phase 2 study of the antiinfluenza mAb CR6261 showed that Cmax in the airways only occurred on Days 2-3 post-dosing, with Cmax in the nasal fluid well below the IC 50 of the mAb. 49In contrast, direct delivery into the respiratory tract by inhalation mitigates these biodistribution limitations.Inhalation achieves highly inhibitory concentrations of the drug virtually instantaneously, which is expected to maximize efficacy and increase the time window for initiating treatment, while also minimizing the dose needed.While jet nebulizers and ultrasonic nebulizers generate sufficient heat and shearing to denature proteins, the latest generations of VMN can support stable nebulization of various biologics as long as the formulation is suitably optimized.Indeed, we recently demonstrated stable nebulization of a potent muco-trapping mAb (IN-006) against SARS-CoV-2 using VMN that met regulatory requirements for initiating clinical studies, 50 and completed a Phase 1 trial demonstrating the safety, tolerability, and surprisingly prolonged pulmonary PK in healthy adults. 51Given that ACE2-(G 4 S) 6 -Fc can be stably nebulized as IN-006, we anticipate a similar once per day inhaled dosing regimen would similarly sustain highly inhibitory levels of ACE2-(G 4 S) 6 -Fc in both the upper and lower respiratory tracts in humans, much greater than what could be achieved via systemic delivery.
Finally, mucus is often overlooked in the context of immune protection against foreign pathogens.This is not surprising as IgG has long been thought to mediate protection systemically by interacting with the immune cells and complement proteins.Nonetheless, there are abundant quantities of IgG naturally present in human AM at levels that are comparable to sIgA, implicating a potential role for IgG in mucosal protection of the respiratory tract.The notion that IgG can work in tandem with mucins to immobilize pathogens in mucus was only demonstrated less than a decade ago by us, first with engineering mAbs that could immobilize Herpes Simplex Virus in cervicovaginal mucus leading to sterilizing immunity against vaginal Herpes transmission, 27,52 and more recently in engineering antibodies that can effectively trap SARS-CoV-2 53 and Ebola VLPs 24 in fresh human AM.Thus, IgG's potential muco-trapping effector function is generally overlooked in the development of antiviral mAb interventions.Nonetheless, muco-trapping offers a number of highly desirable features, including directly intervening in the viral life cycle by immediately halting the spread of infection, unlike small molecule antivirals that fail to inhibit already shed progeny virus.Furthermore, by harnessing natural mucociliary and/or cough-driven AM clearance, muco-trapping mAbs can directly remove potentially inflammatory viral antigens from the respiratory tract, [26][27][28]52,[54][55][56] thus attenuating the pneumonia and inflammation that results in severe COVID. Interestngly, higher affinity between IgG-Fc and mucins does not enhance muco-trapping potencies; instead, there is an optimal weak affinity supported by specific Fc glycosylation that enables IgG to quickly diffuse through mucus and accumulate on the surface of the virus, while enabling effective trapping of mAb/virion complexes.10,11 Here, we produced our ACE2 immunodecoy using specific mammalian cell lines that we previously found to generate Fc N-glycosylation supporting the muco-trapping effector function.Such N-glycosylations would likely be absent from molecules generated from bacterial and fungal systems.Finally, combined with the outstanding safety of biologics, we anticipate inhaled "muco-trapping" mAbs and immunodecoys will eventually become a broadly adopted approach as early therapy for both mild and moderate COVID-19 and other acute respiratory viral infections.$45%. Cell ulture supernatants were concentrated 10-fold by tangential flow (Sartorius Vivaflow 50 crossflow cassette system with 100,000 MWCO cassette with polyethersulfone membrane).The concentrated supernatant was incubated with 1 mL of Ni-Nta agarose resin (Qiagen) overnight before being recovered with a gravity-flow column (Bio-Rad).The resin was then washed with several column volumes of PBS with 20 mM imidazole, followed by elution with PBS with 500 mM imidazole.The proteins were then buffer exchanged into PBS or 20 mM Tris with 120 mM sucrose and 20 mM sodium chloride pH 7 using Spin-X ® UF 50k MWCO PES spin columns.Following buffer exchange, the proteins in Tris-sucrose buffer were flash frozen in liquid nitrogen before been stored at À80 C.After 24 h at 4 C, the VLPs were then aliquoted and stored at À80 C.

| 3D models
UCSF Chimera 1.14 60 was used to generate all the protein models, and UCSF Chimera X 1.1 61 was used to render the model for publication.ACE2-Fc and ACE2-(G 4 S) 6 -Fc were constructed using models 6M17 for ACE2, 62 1HZH for human IgG, 63 and 1EIB for GGGGS linker. 64ACE2-Fc and ACE2-(G 4 S) 6 -Fc bound to S protein were generated by matching the ACE2 of ACE2-Fc and ACE2-(G 4 S) 6 -Fc with the RBD-bound ACE2 in the "all-up" S protein model 7A98. 29The predicted 3D model of ACE2-Fc with collectrin domain was modified from reference 65.

| SEC-MALS measurements of purified proteins and native PAGE
Solutions containing 1.0 mg/mL ACE2-(G 4 S) 6 -Fc or Ace2-Fc were prepared in standard buffer.One hundred microliters of these solutions were then loaded into a Superdex 200 Increase 10/300 GL (Cytivia) mounted on a NGC Quest 10 FPLC (BioRad).The column was pre-equilibrated with PBS, and the whole run was performed at 0.5 mL/min.The molecular weight of proteins eluting from the column was determined using a Mini Dawn multi-angle light scattering detector and its companion software Astra8 (Wyatt), assuming an extension coefficient of 1.92.Molecular weight was calculated for two independent batches of ACE2-(G 4 S) 6 -Fc and two batches of Ace2-Fc.For native PAGE, 5 μg of protein were loaded onto 3%-12% Bis-Tris gels (Invitrogen), and the gels were ran as described by the manufacturer's protocol.

| Scanning differential fluorimetry
The melting temperature of ACE2-LFC was determined by nanoDSF

| Neutralization assays
A series of 10 four-fold serial dilutions were made of ACE2-Linker-Fc or ACE2-Fc or IgG starting at 20 μg/mL in OptiMEM.Ten microliters of each dilution was added to wells of a 96-well plate in triplicate.To each of these dilutions 0.5 μL of SARS-CoV-2 pseudotyped lentivirus (diluted in OptiMEM to 10 μL, MOI 1, titer estimated using infection of HEK293-ACE2 cells, a female sex cell line) was added and -Fc) that binds different variants of SARS-CoV-2 S proteins, potently neutralizes different SARS-CoV-2 pseudovirus and live virus, effectively traps SARS-CoV-2 virus like particles in human AM, can be stably nebulized, and effectively reduces SARS-CoV-2 infections in hamsters.This molecule is now being actively advanced into clinical development as an inhaled treatment for ACE2-binding pathogens, including SARS-CoV-2.

F I G U R E 1
Computational prediction of hypothetical structures of different ACE2 decoys.(a) ACE2-Fc fusion comprised of the entire ACE2 molecule including the collectrin domain linked to IgG1-Fc (ACE2(740)-Fc). 1 (b) ACE2-Fc fusion containing the extracellular domain of ACE2 without collectrin domain, linked to Fc with no flexible linker.(c) ACE2 fragment without collectrin domain linked to human IgG1-Fc via a 30 amino acid glycine-serine flexible linker (ACE2-(G 4 S) 6 -Fc).The flexible linker, shown in gray, allows for much longer reach of the two ACE2 domains.UCSF Chimera 1.

2. 5 |
ACE2-(G 4 S) 6 -Fc effectively traps SARS-CoV-2 virus-like particles (VLPs) in human AM and can be stably nebulized To evaluate whether ACE2-(G 4 S) 6 -Fc can trap SARS-CoV-2 in human AM, we prepared fluorescent SARS-2 VLP by co-expressing S protein with GAG-mCherry fusion construct and visualized the mobility of hundreds to thousands of individual virions in different samples of fresh human AM isolated from extubated endotracheal tubes.These studies were completed before EUA as granted for any COVID vaccines, and thus these specimens likely contained little, if any, endogenous antibodies that could bind and trap SARS-CoV-2 VLPs, unlike our previous work with influenza. 26Not surprisingly, in the absence of exogenously added antibodies or ACE2-decoy molecules, F I G U R E 3 Neutralization potency of different ACE2 decoys against pseudovirus and live SARS-CoV-2 virus.(a) Representative infectivity curves of pseudotyped SARS-CoV-2 virus across different concentrations of ACE2 decoys.(b) IC 50 values estimated from the binding curves.Each data point represents independent experiments.Student's t-test was used for comparisons across molecules.(c) Neutralization of live SARS-CoV-2 virus.Vero E6 cells were infected 48 h prior at MOI of 0.01 before treatment with ACE2-(G 4 S) 6 -Fc.
and measured the binding affinity of the recovered nebulized ACE2-(G 4 S) 6 -Fc via S-protein ELISA.We did not observe any appreciable loss in binding affinity of ACE2-(G 4 S) 6 -Fc recovered from either the upper or lower chamber, compared to ACE2-(G 4 S) 6 -Fc that was not nebulized F I G U R E 4 ACE2-(G 4 S) 6 -Fc more effectively traps SARS-2 VLP in human AM than ACE2-Fc or CR3022, a control anti-SARS-CoV-2 mAb, as measured by high-resolution multiple particle tracking.(a) Mean squared displacements versus time scale, (b) ensemble averaged effective diffusivities (hD eff i), and (c) % fast-moving SARS-CoV-2 VLPs in naïve AM treated with different molecules.(d) Mean squared displacements versus time scale for different concentrations of ACE2-(G 4 S) 6 -Fc, (e) ensemble averaged effective diffusivities (D eff ), and (f) % fast-moving SARS-CoV-2 VLPs in naïve AM treated with different concentrations of ACE2-(G 4 S) 6 -Fc.MSD and D eff were compared across drug constructs or across drug concentrations using one-sided t tests, and unadjusted p values are shown as *p < 0.05, **p < 0.01, and ***p < 0.001.ACE2, angiotensin-converting enzyme 2; AM, airway mucus; MSD, mean squared displacement; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; VLP, virus-like particle.

2. 6 |
Intranasal delivery of ACE2-(G 4 S) 6 -Fc reduces viral load in the nasal turbinates of hamsters infected with SARS-CoV-2 As an in vivo proof-of-concept, we conducted a pilot study assessing the efficacy of intranasal delivery of ACE2-(G 4 S) 6 -Fc in golden Syrian Hamsters infected with live SARS-CoV-2.SARS-CoV-2-infected hamsters develop clinical signs of weight loss, lethargy, hunched back posture, ruffled fur, and rapid breathing, and histopathological changes with high viral loads in the upper and lower respiratory tract, making them a suitable model for testing mAb-based approaches despite differences in anatomy of the respiratory tract.
CHO cells represent the gold standard of mAb manufacturing.To prepare for IND-enabling preclinical studies, we generated and selected stable CHO cell lines that produced ACE2-(G 4 S) 6 -Fc.Encouragingly, we are able to produce ACE2-(G 4 S) 6 -Fc with >2 g/L productivity, generated research cell bank clones, and selected a clone to advance toward master cell banking for GMP production.The measured EC 50

F I G U R E 6
Efficacy of topical ACE2-(G 4 S) 6 -Fc treatment in a hamster model of COVID-19.(a) Viral load in the lung and nasal turbinate tissues of SARS-CoV-2-infected hamsters collected at 4 days post infection, groups treated either 4 or 24 h after infection.(b) Viral load in the nasal turbinates in animals that were withheld treatment until 48 h post infection.(c) Change in bodyweight over the course of the study in animals treated at 4 h or 24 h post infection, compared to control.(d) Change in bodyweight over the course of the study in animals treated at 48 h post infection versus control.PFU/mL or bodyweight were compared across treatment groups using one-sided t tests, and unadjusted p values are shown as *p < 0.05, **p < 0.01.

Fluorescent
VLPs were made by cotransfection of pGAG-mcherry plasmid (kind gift from Gummuluru lab) and CoV-2 S protein plasmid in a 1:1 ratio.Nonreplicating lentivirus pseudotyped with SARS-CoV-2 alpha (B.1.1.7)spike protein were created using the following plasmids, in a 1:1:1:2 ratio: pMDLg/pRRE, pRSV-REV, SARS-CoV-2 alpha spike, and pLL7 GFP.Nonreplicating lentivirus pseudotyped with SARS-CoV-2 beta spike (B.1.351)were created using the same plasmids/ratio above with SARS Cov2 alpha spike replaced with SARS Cov2 beta spike.All plasmids were purified using NucleoBond Xtra Midi Plus EF kit.The plasmids were transfected into LVMaxx using the LVMaxx Transfection kit.Each VLP was made in 60 mL cultures, and harvested after 48 h.The VLPs were purified using 25% sucrose (in 25 mM HEPES/130 mM NaCl) cushion spin protocol.Three milliliters of 25% sucrose solution was add to each Beckman Coulter ultracentrifuge tube, which then had 7 mL of cell culture supernatant gently layered on top.The tubes were then spun at 36,000 rpm for 2.5 h at 4 C.The sucrose/supernatant was then aspirated off, and 20 μL of 10% sucrose solution was placed on top of the VLP pellet.
using a Promethius NT.48 (Nanotemper Technologies).Samples were heated up from 25 C to 95 C at a rate of 1 C/min.Samples were measured in triplicate.Reported data is the average of three independent repeats.

4. 10 |
ELISAELISA binding assays were performed using 96-well half-area plates (Fisher Scientific, Costar 3690) coated with 0.5 μg/mL of S protein and incubated overnight at 4 C. ELISA plates were blocked the following day with 5% (w/v) milk (LabScientific MSPP-M0841) with Tween 20 (Fisher Scientific BP337-100) at a 1:2000 dilution at room temperature for 1 h.Samples were diluted in 1% (w/v) milk with Tween 20 at a 1:10,000 dilution and plated once the blocking had commenced and the 5% milk had been discarded.Samples were incubated at room temperature for 1 h, and the solution was discarded after the 1 h incubation.Plates were then washed with PBS containing Tween 20 at a 1:2000 dilution four times.A peroxidaseconjugated goat anti-human IgG Fc antibody (Rockland 709-1317) was diluted in 1% milk with Tween 20 at a 1:5000 dilution, plated, and incubated at room temperature for 1 h.The solution was then discarded and washed with PBS with Tween 20 two times, followed by washing with just PBS two more times.Plates were developed with TMB solution (ThermoFisher 34029), and development was stopped by adding 2 N HCl (Sigma-Aldrich 320331).The absorbance at 450 nm and 595 nm was then measured with a microplate photodetector (Fisher Scientific, accuSkan FC).