Machine Learning Guided Design of High-Affinity ACE2 Decoys for SARS-CoV-2 Neutralization.

A potential therapeutic strategy for neutralizing SARS-CoV-2 infection is engineering high-affinity soluble ACE2 decoy proteins to compete for binding to the viral spike (S) protein. Previously, a deep mutational scan of ACE2 was performed and has led to the identification of a triple mutant variant, named sACE22.v.2.4, that exhibits subnanomolar affinity to the receptor-binding domain (RBD) of S. Using a recently developed transfer learning algorithm, TLmutation, we sought to identify other ACE2 variants that may exhibit similar binding affinity with decreased mutational load. Upon training a TLmutation model on the effects of single mutations, we identified multiple ACE2 double mutants that bind SARS-CoV-2 S with tighter affinity as compared to the wild type, most notably L79V;N90D that binds RBD similarly to ACE22.v.2.4. The experimental validation of the double mutants successfully demonstrates the use of machine learning approaches for engineering protein-protein interactions and identifying high-affinity ACE2 peptides for targeting SARS-CoV-2.

An ACE2 decoy can be administered by inhalation and potently targets omicron variants of SARS-CoV-2.

Monoclonal antibodies targeting the SARS-CoV-2 spike (S) neutralize infection and are efficacious for the treatment of COVID-19. However, SARS-CoV-2 variants, notably sublineages of B.1.1.529/omicron, have emerged that escape antibodies in clinical use. As an alternative, soluble decoy receptors based on the host entry receptor ACE2 broadly bind and block S from SARS-CoV-2 variants and related betacoronaviruses. The high-affinity and catalytically active decoy sACE22 .v2.4-IgG1 was previously shown to be effective against SARS-CoV-2 variants when administered intravenously. Here, inhalation of aerosolized sACE22 .v2.4-IgG1 increased survival and ameliorated lung injury in K18-hACE2 mice inoculated with P.1/gamma virus. Loss of catalytic activity reduced the decoy's therapeutic efficacy, which was further confirmed by intravenous administration, supporting dual mechanisms of action: direct blocking of S and turnover of ACE2 substrates associated with lung injury and inflammation. Furthermore, sACE22 .v2.4-IgG1 tightly binds and neutralizes BA.1, BA.2, and BA.4/BA.5 omicron and protects K18-hACE2 mice inoculated with a high dose of BA.1 omicron virus. Overall, the therapeutic potential of sACE22 .v2.4-IgG1 is demonstrated by the inhalation route and broad neutralization potency persists against highly divergent SARS-CoV-2 variants.

An engineered ACE2 decoy receptor can be administered by inhalation and potently targets the BA.1 and BA.2 omicron variants of SARS-CoV-2.

Monoclonal antibodies targeting the SARS-CoV-2 spike (S) glycoprotein neutralize infection and are efficacious for the treatment of mild-to-moderate COVID-19. However, SARS-CoV-2 variants have emerged that partially or fully escape monoclonal antibodies in clinical use. Notably, the BA.2 sublineage of B.1.1.529/omicron escapes nearly all monoclonal antibodies currently authorized for therapeutic treatment of COVID-19. Decoy receptors, which are based on soluble forms of the host entry receptor ACE2, are an alternative strategy that broadly bind and block S from SARS-CoV-2 variants and related betacoronaviruses. The high-affinity and catalytically active decoy sACE2 2 .v2.4-IgG1 was previously shown to be effective in vivo against SARS-CoV-2 variants when administered intravenously. Here, the inhalation of sACE2 2 .v2.4-IgG1 is found to increase survival and ameliorate lung injury in K18-hACE2 transgenic mice inoculated with a lethal dose of the virulent P.1/gamma virus. Loss of catalytic activity reduced the decoy’s therapeutic efficacy supporting dual mechanisms of action: direct blocking of viral S and turnover of ACE2 substrates associated with lung injury and inflammation. Binding of sACE2 2 .v2.4-IgG1 remained tight to S of BA.1 omicron, despite BA.1 omicron having extensive mutations, and binding exceeded that of four monoclonal antibodies approved for clinical use. BA.1 pseudovirus and authentic virus were neutralized at picomolar concentrations. Finally, tight binding was maintained against S from the BA.2 omicron sublineage, which differs from S of BA.1 by 26 mutations. Overall, the therapeutic potential of sACE2 2 .v2.4-IgG1 is further confirmed by inhalation route and broad neutralization potency persists against increasingly divergent SARS-CoV-2 variants.

Engineered ACE2 decoy mitigates lung injury and death induced by SARS-CoV-2 variants.

Vaccine hesitancy and emergence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants of concern (VOCs) escaping vaccine-induced immune responses highlight the urgency for new COVID-19 therapeutics. Engineered angiotensin-converting enzyme 2 (ACE2) proteins with augmented binding affinities for SARS-CoV-2 spike (S) protein may prove to be especially efficacious against multiple variants. Using molecular dynamics simulations and functional assays, we show that three amino acid substitutions in an engineered soluble ACE2 protein markedly augmented the affinity for the S protein of the SARS-CoV-2 WA-1/2020 isolate and multiple VOCs: B.1.1.7 (Alpha), B.1.351 (Beta), P.1 (Gamma) and B.1.617.2 (Delta). In humanized K18-hACE2 mice infected with the SARS-CoV-2 WA-1/2020 or P.1 variant, prophylactic and therapeutic injections of soluble ACE22.v2.4-IgG1 prevented lung vascular injury and edema formation, essential features of CoV-2-induced SARS, and above all improved survival. These studies demonstrate broad efficacy in vivo of an engineered ACE2 decoy against SARS-CoV-2 variants in mice and point to its therapeutic potential.

Computationally Designed ACE2 Decoy Receptor Binds SARS-CoV-2 Spike (S) Protein with Tight Nanomolar Affinity.

Even with the availability of vaccines, therapeutic options for COVID-19 still remain highly desirable, especially in hospitalized patients with moderate or severe disease. Soluble ACE2 (sACE2) is a promising therapeutic candidate that neutralizes SARS CoV-2 infection by acting as a decoy. Using computational mutagenesis, we designed a number of sACE2 derivatives carrying three to four mutations. The top-predicted sACE2 decoy based on the in silico mutagenesis scan was subjected to molecular dynamics and free-energy calculations for further validation. After illuminating the mechanism of increased binding for our designed sACE2 derivative, the design was verified experimentally by flow cytometry and BLI-binding experiments. The computationally designed sACE2 decoy (ACE2-FFWF) bound the receptor-binding domain of SARS-CoV-2 tightly with low nanomolar affinity and ninefold affinity enhancement over the wild type. Furthermore, cell surface expression was slightly greater than wild-type ACE2, suggesting that the design is well-folded and stable. Having an arsenal of high-affinity sACE2 derivatives will help to buffer against the emergence of SARS CoV-2 variants. Here, we show that computational methods have become sufficiently accurate for the design of therapeutics for current and future viral pandemics.

An engineered decoy receptor for SARS-CoV-2 broadly binds protein S sequence variants.

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, because of 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 that an engineered decoy receptor, sACE22v2.4, 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.

Engineered High-Affinity ACE2 Peptide Mitigates ARDS and Death Induced by Multiple SARS-CoV-2 Variants.

Vaccine hesitancy and continuing emergence of SARS-CoV-2 variants of concern that may escape vaccine-induced immune responses highlight the urgent need for effective COVID-19 therapeutics. Monoclonal antibodies used in the clinic have varying efficacies against distinct SARS-CoV-2 variants; thus, there is considerable interest in engineered ACE2 peptides with augmented binding affinities for SARS-CoV-2 Spike protein. These could have therapeutic benefit against multiple viral variants. Using molecular dynamics simulations, we show how three amino acid substitutions in an engineered soluble ACE2 peptide (sACE2 2 .v2.4-IgG1) markedly increase affinity for the SARS-CoV-2 Spike (S) protein. We demonstrate high binding affinity to S protein of the early SARS-CoV-2 WA-1/2020 isolate and also to multiple variants of concern: B.1.1.7 (Alpha), B.1.351 (Beta), P.1 (Gamma), and B.1.617.2 (Delta) SARS-CoV-2 variants. In humanized K18-hACE2 mice, prophylactic and therapeutic administration of sACE2 2 .v2.4-IgG1 peptide prevented acute lung vascular endothelial injury and lung edema (essential features of ARDS) and significantly improved survival after infection by SARS-CoV-2 WA-1/2020 as well as P.1 variant of concern. These studies demonstrate for the first time broad efficacy in vivo of an ACE2 decoy peptide against multiple SARS-CoV-2 variants and point to its therapeutic potential.

Machine learning guided design of high affinity ACE2 decoys for SARS-CoV-2 neutralization.

A potential therapeutic candidate for neutralizing SARS-CoV-2 infection is engineering high-affinity soluble ACE2 decoy proteins to compete for binding of the viral spike (S) protein. Previously, a deep mutational scan of ACE2 was performed and has led to the identification of a triple mutant ACE2 variant, named ACE2 2 .v.2.4, that exhibits nanomolar affinity binding to the RBD domain of S. Using a recently developed transfer learning algorithm, TLmutation, we sought to identified other ACE2 variants, namely double mutants, that may exhibit similar binding affinity with decreased mutational load. Upon training a TLmutation model on the effects of single mutations, we identified several ACE2 double mutants that bind to RBD with tighter affinity as compared to the wild type, most notably, L79V;N90D that binds RBD with similar affinity to ACE2 2 .v.2.4. The successful experimental validation of the double mutants demonstrated the use transfer and supervised learning approaches for engineering protein-protein interactions and identifying high affinity ACE2 peptides for targeting SARS-CoV-2.

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

The spike (S) protein of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) binds angiotensin-converting enzyme 2 (ACE2) on host cells to initiate entry, and soluble ACE2 is a therapeutic candidate that neutralizes infection by acting as a decoy. By using deep mutagenesis, mutations in ACE2 that increase S binding are found across the interaction surface, in the asparagine 90-glycosylation motif and at buried sites. The mutational landscape provides a blueprint for understanding the specificity of the interaction between ACE2 and S and for engineering high-affinity decoy receptors. Combining mutations gives ACE2 variants with affinities that rival those of monoclonal antibodies. A stable dimeric variant shows potent SARS-CoV-2 and -1 neutralization in vitro. The engineered receptor is catalytically active, and its close similarity with the native receptor may limit the potential for viral escape.

An engineered decoy receptor for SARS-CoV-2 broadly binds protein S sequence variants.

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, sACE2 2 .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 (RBD) 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. Variant N501Y in the RBD, which has emerged in a rapidly spreading lineage (B.1.1.7) in England, enhances affinity for wild type ACE2 20-fold but remains tightly bound to engineered sACE22.v2.4. We conclude that resistance to engineered decoys will be rare and that decoys may be active against future outbreaks of SARS-associated betacoronaviruses.

Introduction of a polar core into the de novo designed protein Top7.

Design of polar interactions is a current challenge for protein design. The de novo designed protein Top7, like almost all designed proteins, has an entirely nonpolar core. Here we describe the replacing of a sizable fraction (5 residues) of this core with a designed polar hydrogen bond network. The polar core design is expressed at high levels in E. coli, has a folding free energy of 10 kcal/mol, and retains the multiphasic folding kinetics of the original Top7. The NMR structure of the design shows that conformations of three of the five residues, and the designed hydrogen bonds between them, are very close to those in the design model. The remaining two residues, which are more solvent exposed, sample a wide range of conformations in the NMR ensemble. These results show that hydrogen bond networks can be designed in protein cores, but also highlight challenges that need to be overcome when there is competition with solvent.

An examination of the relationship between active site loop size and thermodynamic activation parameters for orotidine 5'-monophosphate decarboxylase from mesophilic and thermophilic organisms.

Closure of the active site phosphate gripper loop of orotidine 5'-monophosphate decarboxylase from Saccharomyces cerevisiae (ScOMPDC) over the bound substrate orotidine 5'-monophosphate (OMP) activates the bound substrate for decarboxylation by at least 10(4)-fold [Amyes, T. L., Richard, J. P., and Tait, J. J. (2005) J. Am. Chem. Soc. 127, 15708-15709]. The 19-residue phosphate gripper loop of the mesophilic ScOMPDC is much larger than the nine-residue loop at the ortholog from the thermophile Methanothermobacter thermautotrophicus (MtOMPDC). This difference in loop size results in a small decrease in the total intrinsic phosphate binding energy of the phosphodianion group of OMP from 11.9 to 11.6 kcal/mol, along with a modest decrease in the extent of activation by phosphite dianion of decarboxylation of the truncated substrate 1-(beta-D-erythrofuranosyl)orotic acid. The activation parameters DeltaH(double dagger) and DeltaS(double dagger) for k(cat) for decarboxylation of OMP are 3.6 kcal/mol and 10 cal K(-1) mol(-1) more positive, respectively, for MtOMPDC than for ScOMPDC. We suggest that these differences are related to the difference in the size of the active site loops at the mesophilic ScOMPDC and the thermophilic MtOMPDC. The greater enthalpic transition state stabilization available from the more extensive loop-substrate interactions for the ScOMPDC-catalyzed reaction is largely balanced by a larger entropic requirement for immobilization of the larger loop at this enzyme.

Mechanism of the orotidine 5'-monophosphate decarboxylase-catalyzed reaction: effect of solvent viscosity on kinetic constants.

Orotidine 5'-monophosphate decarboxylase (OMPDC) is an exceptionally proficient catalyst: the rate acceleration (k(cat)/k(non)) is 7.1 x 10(16), and the proficiency [(k(cat)/K(M))/k(non)] is 4.8 x 10(22) M(-1). The structural basis for the large rate acceleration and proficiency is unknown, although the mechanism has been established to involve a stabilized carbanion intermediate. To provide reaction coordinate context for interpretation of the values of k(cat), k(cat)/K(M), and kinetic isotope effects, we investigated the effect of solvent viscosity on k(cat) and k(cat)/K(M) for the OMPDCs from Methanothermobacter thermautotrophicus (MtOMPDC) and Saccharomyces cerevisiae (ScOMPDC). For MtOMPDC, we used not only the natural OMP substrate but also a catalytically impaired mutant (D70N) and a more reactive substrate (FOMP); for ScOMPDC, we used OMP and FOMP. With MtOMPDC and OMP, k(cat) is independent of solvent viscosity, indicating that decarboxylation is fully rate-determining; k(cat)/K(M) displays a fractional dependence of solvent viscosity, suggesting that both substrate binding and decarboxylation determine this kinetic constant. For ScOMPDC with OMP, we observed that both k(cat) and k(cat)/K(M) are fractionally dependent on solvent viscosity, suggesting that the rates of substrate binding, decarboxylation, and product dissociation are similar. Consistent with these interpretations, for both enzymes with FOMP, the increases in the values of k(cat) and k(cat)/K(M) are much less than expected based on the ability of the 5-fluoro substituent to stabilize the anionic intermediate; i.e., substrate binding and product dissociation mask the kinetic effects of stabilization of the intermediate by the substituent.

Mechanism of the orotidine 5'-monophosphate decarboxylase-catalyzed reaction: evidence for substrate destabilization.

The reaction catalyzed by orotidine 5'-monophosphate decarboxylase (OMPDC) involves a stabilized anionic intermediate, although the structural basis for the rate acceleration (k(cat)/k(non), 7.1 x 10(16)) and proficiency [(k(cat)/K(M))/k(non), 4.8 x 10(22) M(-1)] is uncertain. That the OMPDCs from Methanothermobacter thermautotrophicus (MtOMPDC) and Saccharomyces cerevisiae (ScOMPDC) catalyze the exchange of H6 of the UMP product with solvent deuterium allows an estimate of a lower limit on the rate acceleration associated with stabilization of the intermediate and its flanking transition states (>or=10(10)). The origin of the "missing" contribution, or=10(10)), is of interest. Based on structures of liganded complexes, unfavorable electrostatic interactions between the substrate carboxylate group and a proximal Asp (Asp 70 in MtOMPDC and Asp 91 in ScOMPDC) have been proposed to contribute to the catalytic efficiency [Wu, N., Mo, Y., Gao, J., and Pai, E. F. (2000) Proc. Natl. Acad. Sci. U.S.A. 97, 2017-2022]. We investigated that hypothesis by structural and functional characterization of the D70N and D70G mutants of MtOMPDC and the D91N mutant of ScOMPDC. The substitutions for Asp 70 in MtOMPDC significantly decrease the value of k(cat) for decarboxylation of FOMP (a more reactive substrate analogue) but have little effect on the value of k(ex) for exchange of H6 of FUMP with solvent deuterium; the structures of wild-type MtOMPDC and its mutants are superimposable when complexed with 6-azaUMP. In contrast, the D91N mutant of ScOMPDC does not catalyze exchange of H6 of FUMP; the structures of wild-type ScOMPDC and its D91N mutant are not superimposable when complexed with 6-azaUMP, with differences in both the conformation of the active site loop and the orientation of the ligand vis a vis the active site residues. We propose that the differential effects of substitutions for Asp 70 of MtOMPDC on decarboxylation and exchange provide additional evidence for a carbanionic intermediate as well as the involvement of Asp 70 in substrate destabilization.

Structural basis for substrate specificity in phosphate binding (beta/alpha)8-barrels: D-allulose 6-phosphate 3-epimerase from Escherichia coli K-12.

Enzymes that share the (beta/alpha) 8-barrel fold catalyze a diverse range of reactions. Many utilize phosphorylated substrates and share a conserved C-terminal (beta/alpha) 2-quarter barrel subdomain that provides a binding motif for the dianionic phosphate group. We recently reported functional and structural studies of d-ribulose 5-phosphate 3-epimerase (RPE) from Streptococcus pyogenes that catalyzes the equilibration of the pentulose 5-phosphates d-ribulose 5-phosphate and d-xylulose 5-phosphate in the pentose phosphate pathway [J. Akana, A. A. Fedorov, E. Fedorov, W. R. P. Novack, P. C. Babbitt, S. C. Almo, and J. A. Gerlt (2006) Biochemistry 45, 2493-2503]. We now report functional and structural studies of d-allulose 6-phosphate 3-epimerase (ALSE) from Escherichia coli K-12 that catalyzes the equilibration of the hexulose 6-phosphates d-allulose 6-phosphate and d-fructose 6-phosphate in a catabolic pathway for d-allose. ALSE and RPE prefer their physiological substrates but are promiscuous for each other's substrate. The active sites (RPE complexed with d-xylitol 5-phosphate and ALSE complexed with d-glucitol 6-phosphate) are superimposable (as expected from their 39% sequence identity), with the exception of the phosphate binding motif. The loop following the eighth beta-strand in ALSE is one residue longer than the homologous loop in RPE, so the binding site for the hexulose 6-phosphate substrate/product in ALSE is elongated relative to that for the pentulose 5-phosphate substrate/product in RPE. We constructed three single-residue deletion mutants of the loop in ALSE, DeltaT196, DeltaS197 and DeltaG198, to investigate the structural bases for the differing substrate specificities; for each, the promiscuity is altered so that d-ribulose 5-phosphate is the preferred substrate. The changes in k cat/ K m are dominated by changes in k cat, suggesting that substrate discrimination results from differential transition state stabilization. In both ALSE and RPE, the phosphate group hydrogen bonds not only with the conserved motif but also with an active site loop following the sixth beta-strand, providing a potential structural mechanism for coupling substrate binding with catalysis.