key: cord-326337-s0fp5z1q
authors: Chan, Kui K.; Tan, Timothy J.C.; Narayanan, Krishna K.; Procko, Erik
title: An engineered decoy receptor for SARS-CoV-2 broadly binds protein S sequence variants
date: 2020-10-19
journal: bioRxiv
DOI: 10.1101/2020.10.18.344622
sha: 
doc_id: 326337
cord_uid: s0fp5z1q

The spike S of SARS-CoV-2 recognizes ACE2 on the host cell membrane to initiate entry. Soluble decoy receptors, in which the ACE2 ectodomain is engineered to block S with high affinity, potently neutralize infection and, due to close similarity with the natural receptor, hold out the promise of being broadly active against virus variants without opportunity for escape. Here, we directly test this hypothesis. We find an engineered decoy receptor, sACE22.v2.4, tightly binds S of SARS-associated viruses from humans and bats, despite the ACE2-binding surface being a region of high diversity. Saturation mutagenesis of the receptor-binding domain followed by in vitro selection, with wild type ACE2 and the engineered decoy competing for binding sites, failed to find S mutants that discriminate in favor of the wild type receptor. We conclude that resistance to engineered decoys will be rare and that decoys may be active against future outbreaks of SARS-associated betacoronaviruses.

Zoonotic coronaviruses have crossed over from animal reservoirs multiple times in the past two 19

decades, and it is almost certain that wild animals will continue to be a source of devastating outbreaks. 20

Unlike ubiquitous human coronaviruses responsible for common respiratory illnesses, these zoonotic 21 coronaviruses with pandemic potential cause serious and complex diseases, in part due to their tissue 22 tropisms driven by receptor usage. Severe Acute Respiratory Syndrome Coronaviruses 1 (SARS-CoV-1) 23 and 2 (SARS-CoV-2) engage angiotensin-converting enzyme 2 (ACE2) for cell attachment and entry (1-24 7). ACE2 is a protease responsible for regulating blood volume and pressure that is expressed on the 25 surface of cells in the lung, heart and gastrointestinal tract, among other tissues (8, 9) . The ongoing spread 26 of SARS-CoV-2 and the disease it causes, COVID-19, has had a crippling toll on global healthcare 27 systems and economies, and effective treatments and vaccines are urgently needed. 28

As SARS-CoV-2 becomes endemic in the human population, it has the potential to mutate and 29 undergo genetic drift and recombination. To what extent this will occur as increasing numbers of people 30 are infected and mount counter immune responses is unknown, but already a variant in the viral spike 31 protein S (D614G) has rapidly emerged from multiple independent events and effects S protein stability 32 and dynamics (10, 11). Another S variant (D839Y) became prevalent in Portugal, possibly due to a 33 founder effect (12). Coronaviruses have moderate to high mutation rates. For example, 10 −4 substitutions 34

per year per site occur in HCoV-NL63 (13), an alphacoronavirus that also binds ACE2, albeit via a 35 smaller interface that is only partially shared with the RBDs of SARS-associated betacoronaviruses (14). 36

Additionally, large changes in coronavirus genomes have frequently occurred in nature from 37 recombination events, especially in bats where co-infection levels can be high (15, 16) . Recombination of 38

MERS-CoVs has also been documented in camels (17). This will all have profound implications for the 39 current pandemic's trajectory, the potential for future coronavirus pandemics, and whether drug resistance 40

in SARS-CoV-2 becomes prevalent. 41

The viral spike is a vulnerable target for neutralizing monoclonal antibodies that are progressing 42 through clinical trials, yet in tissue culture escape mutations in the spike rapidly emerge to all antibodies 43 tested (18). Deep mutagenesis of the isolated receptor-binding domain (RBD) by yeast surface display 44 has easily identified mutations in S that retain high expression and ACE2 affinity, yet are no longer bound 45 by monoclonal antibodies and confer resistance (19) . This has motivated the development of cocktails of 46 non-competing monoclonals (18, 20) , inspired by lessons learned from the treatment of HIV-1 and Ebola, 47

to limit the possibilities for the virus to escape. Notably, drug maker Eli Lilly has a monoclonal 48 monotherapy (LY-CoV555) in advanced trials (NCT04427501) where the emergence of resistant virus 49 variants has occurred; the trial has been updated to include an arm with a second monoclonal (LY-50 CoV016). However, even the use of monoclonal cocktails does not address future coronavirus spill overs 51 from wild animals that may be antigenically distinct. Indeed, large screening efforts were required to find 52 antibodies from recovered SARS-CoV-1 patients that cross-react with SARS-CoV-2 (21), indicating 53

antibodies have confined capacity for interacting with variable epitopes on the spike surface, and are 54 unlikely to be broad and pan-specific for all SARS-related viruses. 55

An alternative protein-based antiviral to monoclonal antibodies is to use soluble ACE2 (sACE2) as a 56 decoy to compete for receptor-binding sites on the viral spike (6, (22) (23) (24) (25) of diverse SARS-associated betacoronaviruses that use ACE2 for entry. We further fail to find mutations 78 within the RBD, which directly contacts ACE2 and is where possible escape mutations will most likely 79 reside, that redirect specificity towards the wild type receptor. We conclude that resistance to an 80 engineered decoy receptor will be rare, and sACE2 2 .v2.4 targets common attributes for affinity to S in 81 SARS-associated viruses. 82

The affinities of the decoy receptor sACE2 2 .v2.4 were determined for purified RBDs from the S 85 proteins of five coronaviruses from Rhinolophus bat species (isolates LYRa11, Rs4231, Rs7327, Rs4084 86

and RsSHC014) and two human coronaviruses, SARS-CoV-1 and SARS-CoV-2. These viruses fall 87 within a common clade of betacoronaviruses that use ACE2 as an entry receptor (7). They share close 88 sequence identity within the RBD core while variation is highest within the functional ACE2 binding site 89 (Figures 1 and S1) , possibly due to a co-evolutionary 'arms race' with polymorphic ACE2 sequences in 90 ecologically diverse bat species (28). Affinity was measured by biolayer interferometry (BLI), with 91 sACE2 2 (a.a. S19-G732) fused at the C-terminus with the Fc moiety of human IgG1 immobilized to the 92 sensor surface and monomeric 8his-tagged RBD ( Figure S2 ) used as the soluble analyte. This 93 arrangement excludes avidity effects, which otherwise cause artificially tight (picomolar) apparent 94 affinities whenever dimeric sACE2 2 in solution is bound to immobilized RBD decorating an interaction 95

surface. Wild type sACE2 2 bound all the RBDs with affinities ranging from 16 nM for SARS-CoV-2 to 96 91 nM for LYRa11, with median affinity 60 nM ( SARS-CoV-2 to 3.5 nM for isolate Rs4231, with median affinity less than 2 nM ( Table 1 ). The 100

approximate 35-fold affinity increase of the engineered decoy applies universally to coronaviruses in the 101 test panel and the molecular basis for affinity enhancement must therefore be grounded in common 102

attributes of RBD/ACE2 recognition. 103 

The RBD of SARS-CoV-2 (PDB 6M17) is colored by diversity between 7 SARS-associated CoV strains 105 (blue, conserved; red, variable). 

A deep mutational scan of the RBD in the context of full-length S reveals residues in the ACE2 110 binding site are mutationally tolerant 111

To explore potential sequence diversity in S of SARS-CoV-2 that may act as a 'reservoir' for drug 112 resistance, the mutational tolerance of the RBD was evaluated by deep mutagenesis (32). Saturation 113 mutagenesis was focused to the RBD (a.a. C336-L517) of full-length S tagged at the extracellular N-114

terminus with a c-myc epitope for detection of surface expression. The spike library, encompassing 3,640 115 single amino acid substitutions, was transfected in human Expi293F cells under conditions where cells 116

typically acquire no more than a single sequence variant (33, 34). The culture was incubated with wild 117 type, 8his-tagged, dimeric sACE2 2 at a sub-saturating concentration (2.5 nM). Bound sACE2 2 -8h and 118

surface-expressed S were stained with fluorescent antibodies for flow cytometry analysis ( Figure 2A ).

Compared to cells expressing wild type S, the library was poorly expressed, indicating many mutations 120

are deleterious for folding and expression. A cell population was clearly discernable expressing S 121 variants that bind ACE2 with decreased affinity ( Figure 2B ). After gating for c-myc-positive cells 122 expressing S, cells with high and low levels of bound sACE2 2 were collected by fluorescence-activated 123 cell sorting (FACS), called the ACE2-High and ACE2-Low populations, respectively ( Figure 2C ). Both 124 the expression and sACE2 2 binding signals decreased over minutes to hours during sorting, possibly due 125

to shedding of the S1 subunit. Cells were therefore collected and pooled from three separate FACS 126 experiments for a combined 8 hours sort time. 127 averaging the log 2 enrichment ratios for each of the possible amino acids at a residue position. By adding 143 conservation scores for both the ACE2-High and ACE2-Low sorts we derive a score for surface 144 expression, which shows that the hydrophobic RBD core is tightly conserved for folding and trafficking 145

of the viral spike ( Figure 3A ). By comparison, residues on the exposed RBD surface are mutationally 146

permissive for S surface expression. This matches the mutational tolerance of proteins generally. 147 

For tight ACE2 binding (i.e. S variants in the ACE2-High population), conservation increases for 161 RBD residues at the ACE2 interface, yet mutational tolerance remains high ( Figure 3C ). The sequence 162 diversity observed among natural betacoronaviruses, which display high diversity at the ACE2 binding 163 site, is therefore replicated in the deep mutational scan, which predicts the SARS-CoV-2 spike tolerates 164 substantial genetic diversity at the receptor-binding site for function. From this accessible sequence 165 diversity SARS-CoV-2 might feasibly mutate to acquire resistance to monoclonal antibodies or 166 engineered decoy receptors targeting the ACE2-binding site. binding site (e.g. V362, Y365 and C391) is free to mutate for yeast surface display, but its sequence is 174 constrained in our experiments; this region of the RBD is buried by connecting structural elements to the 175 global fold of an S subunit in the closed-down conformation (this is the dominant conformation for S 176 subunits and is inaccessible to receptor binding) (2, 4, 38, 39). We used targeted mutagenesis to 177

individually test alanine substitutions to all the cysteines in the RBD ( Figure S4 ). We found all cysteine-178

to-alanine mutations severely diminish S surface expression in Expi293F cells, including C391A and 179

C525A on the RBD 'backside' that were neutral in the yeast display scan (36). These differences 180

demonstrate that there are tighter sequence constraints on the RBD in the context of a full spike expressed 181 at a human cell membrane, yet overall we consider the two data sets to closely agree. 182

For binding to dimeric sACE2 2 , we note that interface residues were more tightly conserved in the 183

Starr et al data set (Figure 3D ), possibly a consequence of three differences between the deep mutagenesis 184 experiments. First, our selections for ACE2 binding of S variants at the plasma membrane appears to 185

primarily reflect mutational effects on surface expression, which is almost certainly more stringent in 186 human cells. Yeast permit many poorly folded proteins to leak to the cell surface (40). Second, the yeast 187

selections were conducted at multiple sACE2 concentrations from which apparent K D changes were 188 computed (36); the Starr et al data in this regard is very comprehensive. Due to the long sort times 189 required for our human cell libraries where only a small fraction of cells express spike, we sorted at a 190 single sACE2 2 concentration that cannot accurately capture a range of different binding affinities 191

quantitatively. Third, dimeric sACE2 2 may geometrically complement trimeric S densely packed on a 192

human cell membrane, such that avidity masks the effects of affinity-reducing mutations. Nonetheless, 193 there is overall agreement that ACE2 binding often persists following mutations to the RBD surface, and 194

our data simply suggests mutational tolerance may be even greater than that already observed by Starr et 195 al. 196

Having shown that the ACE2-binding site of SARS-CoV-2 protein S tolerates many mutations, we 198 asked whether mutations might therefore be found that confer resistance to the engineered decoy 199 sACE2 2 .v2.4. Resistance mutations are anticipated to lose affinity for sACE2 2 .v2.4 while maintaining 200

binding to the wild type receptor, and are most likely to reside in the RBD where physical contacts are 201 made. Similar reasoning formed the foundation of a deep mutagenesis-based selection of the isolated 202

RBD by yeast surface display to find escape mutations to monoclonal antibodies, and the results were 203

predictive of escape mutations in pseudovirus growth selections (19). 204

To address whether escape mutations from the engineered decoy might be found in the RBD, we 205

repurposed the S protein library for a specificity selection. Cells expressing the library, encoding all 206 possible substitutions in the RBD, were co-incubated with wild type sACE2 2 fused to the Fc region of 207

IgG1 and 8his-tagged sACE2 2 .v2.4 at concentrations where both proteins bind competitively (25). It was 208

immediately apparent from flow cytometry of the Expi293F culture expressing the S library that there 209

were cells expressing S variants shifted towards preferential binding to sACE2 2 .v2.4, but no significant 210 population with preferential binding to the wild type receptor (Figures 4A and 4B ). Cells expressing S 211 variants that might preferentially bind sACE2 2 (WT)-IgG1 or sACE2 2 .v2.4 were gated and collected by 212 FACS (Figure 4C ), followed by deep sequencing of S transcripts to determine enrichment ratios. There 213 was close agreement between two independent replicate experiments ( Figures 4D-4G ). Most RBD 214 mutations were depleted following sorting, consistent with deleterious effects on S folding and 215 expression. 216 

Soluble ACE2 2 .v2.4 has three mutations from wild type ACE2: T27Y buried within the RBD 233

interface, and L79T and N330Y at the interface periphery ( Figure 5A) . A substantial number of mutations 234

in the RBD of S were selectively enriched for preferential binding to sACE2 2 .v2.4 ( Figure 5B , upper-left 235 quadrant). While sACE2 2 .v2.4-specificity mutations could be found immediately adjacent to the sites of 236 engineered mutations in ACE2 (in particular mutations to S-F486 adjacent to ACE2-L79 and S-T500 237 adjacent to ACE2-N330), major hot spots for sACE2 2 .v2.4-specificity mutations were also mapped to 238 RBD loop 498-506, contacting the region where the ACE2-α1 helix packs against a β-hairpin motif 239

( Figure 5A ). By comparison, there were no hot spots in the RBD for sACE2 2 (WT)-specificity mutations. 240

Indeed, only a small number of mutations were selectively enriched for preferential binding to wild type 241 receptor ( Figure 5B ), and the abundance of these putative wild type-specific mutations barely rose above 242 the expected level of noise in the deep mutagenesis data. In this competition assay, S binding to wild type 243 sACE2 2 is therefore more sensitive to RBD mutations than S binding to engineered sACE2 2 .v2.4. 244 

To determine whether the potential wild type ACE2-specific mutations found by deep mutagenesis 260 are real as opposed to false predictions due to data noise, we tested 24 mutants of S selectively enriched in 261 the wild type-specific gate by targeted mutagenesis (blue data points in Figure 5B ). Only minor shifts 262 towards binding wild type sACE2 2 were observed ( Figure S5 ). Two S mutants were investigated further 263 in sACE2 2 titration experiments, N501W and N501Y, which both retained high receptor binding and 264 displayed small shifts towards wild type sACE2 2 in the competition experiment. N501 of S is located in 265 the 498-506 loop and its substitution to large aromatic side chains might alter the loop conformation to 266 cause steric strain with nearby ACE2 mutation N330Y in sACE2 2 .v2.4. After titrating the concentrations 267 of 8his-tagged sACE2 2 (WT) and sACE2 2 .v2.4 and measuring bound protein to S-expressing cells by flow 268 cytometry, it was found S-N501W and S-N501Y do show enhanced specificity for wild type sACE2 2 , but 269

the effect is small and sACE2 2 .v2.4 remains the stronger binder ( Figure 5C ); these mutations therefore 270

will not confer resistance in the virus to the engineered decoy. By comparison, multiple independent 271 escape mutations are readily found in S of SARS-CoV-2 that diminish the efficacy of monoclonal 272

antibodies by many orders of magnitude (18, 19) . 273

Finally, 8 representative mutations to S predicted from the deep mutational scan to increase 274 specificity towards sACE2 2 .v2.4 (purple data points in Figure 5B ) were cloned and 7 were found to have 275 large shifts towards preferential sACE2 2 .v2.4 binding in the competition assay ( Figure S6 ). These S 276 mutations were Y449K/Q/S, L455G/R/Y and G504K. None of the mutated sites is in direct contact with 277 an engineered residue on sACE2 2 .v2.4 and the molecular bases for specificity changes are therefore 278

ambiguous, but we speculate may involve local conformational perturbations. Validation by targeted 279 mutagenesis therefore confirms that the selection can successfully find mutations in S with altered 280 specificity. The inability to find mutations in the RBD that impart high specificity for the wild type 281

receptor means such mutations are rare or may not even exist, at least within the receptor-binding domain 282

where direct physical contacts with receptors occur. We cannot exclude mutations elsewhere having 283 long-range conformational effects. Engineered, soluble decoy receptors therefore live up to their promise 284 as broad therapeutic candidates against which a virus cannot easily escape. 285

The allure of soluble decoy receptors is that the virus cannot easily mutate to escape neutralization. 287

Mutations that reduce affinity of the soluble decoy will likely also decrease affinity for the wild type 288 receptor on host cells, thereby coming at the cost of diminished infectivity and virulence. However, this 289

hypothesis has not been rigorously tested, and since engineered decoy receptors differ from their wild 290 type counterparts, even if by just a small number of mutations, it is possible a virus may evolve to 291 discriminate between the two. Here, we show that an engineered decoy receptor for SARS-CoV-2 292 broadly binds with low nanomolar K D to the spikes of SARS-associated betacoronaviruses that use ACE2 293

for entry, despite high sequence diversity within the ACE2-binding site. Mutations in S of SARS-CoV-2 294 that confer high specificity for wild type ACE2 were not found in a comprehensive screen of all 295

substitutions within the RBD. The engineered decoy receptor is therefore broad against zoonotic ACE2-296

utilizing coronaviruses that may spill over from animal reservoirs in the future and against variants of 297 SARS-CoV-2 that may arise as the current COVID-19 pandemic rages on. We argue it is unlikely that 298 decoy receptors will need to be combined in cocktail formulations, as is required for monoclonal 299 antibodies or designed miniprotein binders to prevent the rapid emergence of resistance (18, 41), 300

facilitating manufacture and distribution. Our findings give insight into how a potential therapeutic can 301 achieve breath with a low chance of virus resistance for a family of highly infectious and deadly viruses. physiology to exert unacceptable toxicity. For example, the entry receptor for human cytomegalovirus is 312 a growth factor receptor, and growth factor interactions had to be knocked out to make a virus-specific 313 decoy suitable for in vivo administration (43). However, ACE2 in this regard is different and its 314 endogenous activity -the catalytic conversion of vasoconstrictive and inflammatory peptides of the renin-315

angiotensin system -may be of direct benefit for addressing COVID-19 symptoms. During infection, 316

ACE2 activity is dysregulated and the renin-angiotensin system becomes imbalanced, possibly driving 317 aspects of acute-respiratory distress syndrome (ARDS) (44-46). Administration of recombinant sACE2 318 converts angiotensin (Ang) I and II to the protective peptides Ang-(1-9) and Ang-(1-7), respectively, with 319 potential benefits for the pulmonary and cardiovascular systems that include decreased lung elastance, 320

increased blood oxygenation, reduced hypertension and diminished inflammation (44, 45, (47) (48) (49) (50) provide no more than a single coding variant per cell (33, 34). Expi293F cells at 2 × 10 6 / ml were 373 transfected with a mixture of 1 ng coding plasmid (i.e. library DNA) with 1.5 µg pCEP4-ΔCMV carrier 374 plasmid (described in (34) FITC fluorescence for bound sACE2 2 (WT)-8h were collected ( Figure 2C ). Collection tubes were coated 386 overnight with fetal bovine serum prior to sorting and contained Expi293 Expression Medium. Collected 387 cell pellets were frozen at -80°C and were pooled across separate sort experiments prior to extraction of 388 total RNA. 389

The competition selection was performed similarly, with the exception that cells expressing the S library 390

were incubated for 30 minutes in a mixture of 20 nM sACE2 2 .v2.4-8h and 25 nM sACE2 2 (WT)-IgG1. 391

After washing twice, bound proteins were stained for 30 minutes with anti-human IgG-APC (clone 392 HP6017, 1/250 dilution; BioLegend) and anti-HIS-FITC (chicken polyclonal, 1/100 dilution; 393

Immunology Consultants Laboratory). Cells were washed twice and sorted. After gating for the main 394 population of viable cells as described above, the 20 % of cells with the highest FITC-relative-to-APC 395 and highest APC-relative-to-FITC signals were collected ( Figure 4C ). 396

Total RNA was extracted from the collected cells using a GeneJET RNA purification kit (Thermo 397

Scientific). First strand cDNA was synthesized with Accuscript (Agilent) primed with a gene-specific 398

oligonucleotide. The region of S scanned by saturation mutagenesis was PCR amplified as 3 overlapping 399

fragments that together span the full RBD sequence. Following a second round of PCR, primers added 400 adapters for annealing to the Illumina flow cell and sequencing primers, together with barcodes for 401 experiment identification. The PCR products were sequenced on an Illumina NovaSeq 6000 using a 402 2×250 nt paired end protocol. Data were analyzed using Enrich (35), where the frequencies of S variants 403

in the transcripts of the sorted populations were compared to their frequencies in the naive plasmid 404

library. Log 2 enrichment ratios for all the individual mutations were calculated and normalized by 405 subtracting the log 2 enrichment ratio for the wild type sequence across the same PCR-amplified fragment.

A pneumonia outbreak associated with a new coronavirus of probable bat origin. 435 Nature

Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Receptor recognition by novel coronavirus from 439 Wuhan: An analysis based on decade-long structural studies of SARS

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked 444 by a Clinically Proven Protease Inhibitor

Angiotensin-converting enzyme 2 is a functional receptor for the SARS coronavirus. 446 Nature

Functional assessment of cell entry and receptor usage for 448 SARS-CoV-2 and other lineage B betacoronaviruses

Much More Than Just a Receptor for SARS-COV-2. Front

Angiotensin-converting enzyme 2 and angiotensin 1-7: novel therapeutic targets. 452

The D614G mutation in the SARS-CoV-2 spike protein reduces S1 shedding and 454 increases infectivity

Tracking Changes in SARS-CoV-2 Spike: Evidence that D614G Increases 456 Infectivity of the COVID-19 Virus

On the track of the D839Y mutation in the SARS-CoV-2 Spike fusion peptide: 458 emergence and geotemporal spread of a highly prevalent variant in Portugal

Mosaic structure of human coronavirus NL63, one thousand years of evolution

Crystal structure of NL63 respiratory coronavirus receptor-binding 463 domain complexed with its human receptor

Genetic Recombination, and Pathogenesis of Coronaviruses

Evolutionary origins of the SARS-CoV-2 sarbecovirus lineage responsible for 468 the COVID-19 pandemic

Co-circulation of three camel coronavirus species and recombination of 470 MERS-CoVs in Saudi Arabia

Antibody cocktail to SARS-CoV-2 spike protein prevents rapid mutational escape 472 seen with individual antibodies

Complete mapping of mutations to the SARS-CoV-2 spike receptor-binding 474 domain that escape antibody recognition

Ultrapotent human antibodies protect against SARS-CoV-2 challenge via 476 multiple mechanisms

Cross-neutralization of SARS-CoV-2 by a human monoclonal SARS-CoV 478 antibody

Susceptibility to SARS coronavirus S protein-driven infection correlates with 480 expression of angiotensin converting enzyme 2 and infection can be blocked by soluble receptor

Neutralization of SARS-CoV-2 spike pseudotyped virus by recombinant ACE2-Ig

Inhibition of SARS-CoV-2 Infections in Engineered Human Tissues Using 485 Clinical-Grade Soluble Human ACE2

Engineering human ACE2 to optimize binding to the spike protein of SARS 487 coronavirus 2

Engineered ACE2 receptor traps potently neutralize SARS-CoV-2. bioRxiv

High affinity modified ACE2 receptors prevent SARS-CoV-2 infection. bioRxiv

Exceptional diversity and selection pressure on SARS-CoV 493 and SARS-CoV-2 host receptor in bats compared to other mammals

Structural basis of receptor recognition by SARS-CoV-2

Stabilized coronavirus spikes are resistant to conformational changes 497 induced by receptor recognition or proteolysis

Receptor and viral determinants of SARS-coronavirus adaptation to human ACE2

Deep mutational scanning: a new style of protein science

Mapping Interaction Sites on Human Chemokine Receptors by Deep 503 Mutational Scanning

Structural architecture of a dimeric class C GPCR based on co-trafficking of sweet 505 taste receptor subunits

Enrich: software for analysis of protein function 507 by enrichment and depletion of variants

Deep Mutational Scanning of SARS-CoV-2 Receptor Binding Domain Reveals 509 Constraints on Folding and ACE2 Binding

De novo design of ACE2 protein decoys to neutralize SARS-CoV-2. bioRxiv

Distinct conformational states of SARS-CoV-2 spike protein

Molecular architecture of the SARS-CoV-2 virus

Global analysis of protein folding using massively parallel design, synthesis, 516 and testing

De novo design of picomolar SARS-CoV-2 miniprotein inhibitors

THPdb: Database of FDA-approved peptide and protein therapeutics

Engineered receptors for human cytomegalovirus that are orthogonal to normal 522 human biology

Angiotensin-converting enzyme 2 protects from severe acute lung failure

Recombinant angiotensin-converting enzyme 2 improves pulmonary blood flow 526 and oxygenation in lipopolysaccharide-induced lung injury in piglets

The pivotal link between ACE2 deficiency 529 and SARS-CoV-2 infection

Renin-angiotensin-system, a potential pharmacological candidate, in acute 531 respiratory distress syndrome during mechanical ventilation

SARS-CoV-2 and ACE2: The biology and clinical data settling the ARB and 534 ACEI controversy

ACE2 improves right ventricular function 536 in a pressure overload model

Novel ACE2-Fc chimeric fusion provides long-lasting hypertension control and 538 organ protection in mouse models of systemic renin angiotensin system activation

Pharmacokinetics and pharmacodynamics of recombinant human angiotensin-541 converting enzyme 2 in healthy human subjects

A pilot clinical trial of recombinant human angiotensin-converting enzyme 2 in 543 acute respiratory distress syndrome

Novel ACE2-IgG1 fusions with improved activity against SARS-CoV2. 545 bioRxiv

Computational design of a protein-based enzyme inhibitor

Cytometer (BD Biosciences) and data were processed with FCS Express (De Novo Software).

Quantification of myc-S surface expression is detailed in Figure S4 . part supported by NIH award R01AI129719 to E.P. The University of Illinois has filed a provisional 432 patent for engineered decoy receptors and E.P. and K.K.C. are co-founders of Orthogonal Biologics, Inc. 433