key: cord-296657-mymndjvd
authors: Higuchi, Yusuke; Suzuki, Tatsuya; Arimori, Takao; Ikemura, Nariko; Kirita, Yuhei; Ohgitani, Eriko; Mazda, Osam; Motooka, Daisuke; Nakamura, Shota; Matsuura, Yoshiharu; Matoba, Satoaki; Okamoto, Toru; Takagi, Junichi; Hoshino, Atsushi
title: High affinity modified ACE2 receptors prevent SARS-CoV-2 infection
date: 2020-09-16
journal: bioRxiv
DOI: 10.1101/2020.09.16.299891
sha: 
doc_id: 296657
cord_uid: mymndjvd

The SARS-CoV-2 spike protein binds to the human angiotensin-converting enzyme 2 (ACE2) receptor via receptor binding domain (RBD) to enter into the cell. Inhibiting this interaction is a main approach to block SARS-CoV-2 infection and it is required to have high affinity to RBD independently of viral mutation for effective protection. To this end, we engineered ACE2 to enhance the affinity with directed evolution in human cells. Three cycles of random mutation and cell sorting achieved more than 100-fold higher affinity to RBD than wild-type ACE2. The extracellular domain of modified ACE2 fused to the Fc region of the human immunoglobulin IgG1 had stable structure and neutralized SARS-CoV-2 pseudotyped lentivirus and authentic virus with more than 100-fold lower concentration than wild-type. Engineering ACE2 decoy receptors with directed evolution is a promising approach to develop a SARS-CoV-2 neutralizing drug that has affinity comparable to monoclonal antibodies yet displaying resistance to escape mutations of virus.

Coronavirus disease 2019 has spread across the world as a tremendous pandemic and presented an unprecedented challenge to human society. The causative agent of COVID-19, SARS-CoV-2 is a single-stranded positive-strand RNA virus that belongs to lineage B, clade 1 of the betacoronavirus genus 1-3 . The virus binds to host cells through its trimeric spike glycoprotein composed of two subunits; S1 is responsible for receptor binding and S2 for membrane fusion 4 .

Angiotensin-converting enzyme 2 (ACE2) is lineage B clade 1 specific receptor including SARS-CoV-2 3 . The receptor binding domain (RBD) of S1 subunit directly binds ACE2 with high affinity, therefore, it is the most important targeting site to inhibit viral infection. In fact, the RBD is the common binding site of effective neutralizing antibodies identified from convalescent patients [5] [6] [7] .

RNA viruses such as SARS-CoV-2 have high mutation rates 8 , which are correlated with high evolvability including the acquisition of anti-viral drug resistance. Neutralizing antibodies are one of the promising approaches to combat COVID-19. Accumulating evidence demonstrated that monoclonal antibodies isolated from convalescent COVID-19 patients have high potency in neutralizing viruses. However, mutations in the spike gene can lead to the SARS-CoV-2 adaptation to such neutralizing antibodies. In the replicating SARS-CoV-2 pseudovirus culture experiment, escape mutation was observed against monoclonal antibody as early as in the first passage 9 and evasion was seen even against the polyclonal convalescent plasma 10 . Notably, some mutations identified in in vitro replicating culture experiment are present in natural population according to the database 10 Similarly to the anti-RBD antibodies, extracellular domain of ACE2, soluble ACE2 (sACE2), can also be used to neutralize SARS-CoV-2 as a decoy receptor. The therapeutic potency was confirmed using human organoid 10 , and now Apeiron Biologics conducts European phase II clinical trial of recombinant sACE2 against COVID-19. In addition, fusing sACE2 to the Fc region of the human IgG1 has been shown to enhance neutralization capacity 11 as well as to improve the pharmacokinetics to the level of IgG in mice 12 . Most importantly, sACE2 has a great advantage over antibodies due to the resistance to the escape mutation. The virus with escape mutation from sACE2 should have limited binding affinity to cell surface native ACE2 receptors, leading to a diminished or eliminated virulence.

Unfortunately, many reports, including our current study, have revealed that the binding affinity of wild-type sACE2 to the SARS-CoV-2 spike RBD is much weaker (KD ~50 nM) than that of clinical grade antibodies 4, 11, [13] [14] [15] . Thus, the therapeutic potential of the wild-type sACE2 as a neutralizing agent against SARS-CoV-2 is uncertain.

Here we conducted protein engineering with human cell-based directed evolution to improve the binding affinity of ACE2 to the spike RBD. Random mutations were introduced in the protease domain containing the interface to the RBD, then full length ACE2 mutant library was expressed in 293T cells and incubated with fluorescence-labelled RBD. High binding population was sorted and underwent DNA extraction, the bulk of which was further induced with random mutations for the next cycle of selection. Three cycles of screening resulted in an identification of mutant ACE2 clones with more than 100-fold higher binding affinity to the RBD and lower half-maximal inhibitory concentration (IC50) for SARS-CoV-2 pseudotyped lentivirus as well as authentic virus. The present protein engineering system generates a virus-neutralizing drug that has high affinity comparable with antibodies and can resolve the issue of drug resistance caused by escape mutation.

We engineered ACE2 to bind the RBD of the SARS-CoV-2 spike protein with the combination of surface display of mutagenized library and fluorescence-activated cell sorting (FACS) to perform the evolution in 293T human cells. The protease domain (PD) of ACE2 is known to harbor the interface to viral spike protein, located in the top-middle part of ACE2 ectodomain. In this study, ACE2 residues 18-102 and 272-409, referred to as PD1 and PD2, respectively, were mutagenized independently.

Synthetic signal sequence and HA tag were appended and restriction sites were introduced in both sides of PD1 and PD2 by optimizing codon (Fig. 1a) . We used error-prone PCR to mutagenize the protease domain of ACE2 with an average of about one amino acid mutation per 100 bp, then inserted the fragment into the introduced restriction site by homologous recombination. The reaction sample was transformed to competent cell, generating a library of ~10 5 mutants. Mutant plasmid library was packaged into lentivirus, followed by expression in human 293T cells in less than 0.3 MOI (multiplicity of infection) to yield no more than one mutant ACE2 per cell. Cells were incubated with recombinant RBD of SARS-CoV-2 spike protein fused to superfolder GFP (sfGFP; Fig. 1b) . We confirmed the level of bound RBD-sfGFP and surface expression levels of HA-tagged ACE2 with Alexa Fluor 647 in twodimensional display of flow cytometry. Top 0.05 % cells showing higher binding relative to expression level were harvested from ~5 x 10 7 cells by FACS. To exclude the structurally unstable mutants, cells with preserved signal of surface ACE2 were gated. Genomic DNA was extracted from collected cells and mutagenized again to proceed to the next cycle of screening (Fig. 1c) .

Random mutagenesis screening for PD1 was performed 3 times and mutated sequences from top 0.05% population were reconstructed into the backbone plasmid, and expressed in 293T cells individually. One to three hundred clones were validated for the binding capacity to the RBD-sfGFP.

As the selection cycle advances, the two-dimensional distribution of library cells in flowcytometry became broader and higher in RBD-binding signal, and individual clone validation identified several mutants with higher binding capacity (Fig.2a) . To evaluate the neutralization activity in the form of sACE2, we first generated fusion protein of the soluble extracellular domain of mutagenized ACE2 residues 18-614 and sfGFP (sACE2-sfGFP) and used them to compete with the cell-surface WT ACE2 for the RBD binding. To this end, concentration of each mutant sACE2-sfGFP in the cultured medium from transfected cells was quantitatively standardized with sfGFP signal, serially diluted, preincubated with RBD-sfGFP for 30 min, and then transferred to wild-type ACE2 expressing 293T cells. After 30 min, the RBD-bound cells were analyzed in flowcytometry. Higher neutralization activity against the RBD was confirmed for the mutants that have accumulated mutations (Fig. 2b , Table. S1).

Second mutagenesis based on the top hit of first screening, 1-19 mutant, was also performed, but the distribution of the library cells did not expand so much, and we could not isolate clones with significantly higher affinity than the bulk of top 0.05% (Fig. S1 ). We next performed PD2 mutagenesis in both the bulk of top 0.05% and one of the highest mutants of the 3 rd library, the clone 3N39. Again, the binding distribution of the PD2 library cells was similar to the basal cells, suggesting the inability of this strategy to further increase the affinity to RBD. A recent study reported, via deep mutational scanning, that several specific mutations in PD2 were enriched in high RBD-binding clones 13 . When we added these mutations in 3N39, it did not improve the capacity of the RBD neutralization further ( Fig. S2 ).

To identify essential mutation(s) in the high affinity ACE2 mutants, each mutation was altered to wild-type in mutant 3N39, 3J113 and 3j320. Mutant 3N39 contains 7 mutations of A25V, K26E, K31N, E35K, N64I, L79F, and N90H. Among them, individual back-mutation of V25, N31, K35, and F79 to wild-type residues resulted in modest to severe reduction of the RBD-neutralization capacity (Fig. 3a) , while multiple back mutations of E26, I64, and H90 in various combination did not alter the high activity of the original 3N39 ( Fig. 3b) , indicating that A25V, K31N, E35K, and L79F was necessary and sufficient components. In the case of mutant 3J113 that is composed of K31M, E35K, Q60R, S70F, L79F, and N90D, similar back mutation experiment revealed that K31M, E35K, and Q60R was essential (Fig. 3c) . Simultaneous back mutations of two (F70/F79 and F70/D90) but not three (F70/F79/D90) nonessential residues were tolerated ( Fig. 3d) , suggesting that L79F and N90D may exert their positive effect on the activity only when they coexist. The third mutant 3J320 has T20I, A25V, H34A, T78R, T92Q, and Q101H. Single and multiple back mutation experiments showed that H34A, T92Q, and Q101H were essential in securing the high inhibitory activity of 3J320 (Figs. 3e, f).

These top 3 high affinity mutants exhibited higher RBD-neutralization activity when compared with the same sACE2 scaffold carrying the two high affinity mutation sets reported recently 11, 13 (Fig. 3g ).

First mutant library was also sorted in the manner of high and low RBD-sfGFP binding signal (Fig.   S2a ). The affinity value of each mutant was defined as the ratio of high and low read counts. Then, the impact of each amino acid mutation on RBD binding was analyzed as a semi-deep mutational scanning (DMS) (Fig. S2b ). This analysis revealed that some mutations such as K31N, E35K, and Q60R in top 3 high affinity mutants had very mild impact in itself. Simple combination of high value mutations, A25V, K26T, Q42L, L79V and T92K, referred to here as the DMS mutant, and its derivatives showed less neutralization activity than top three high affinity mutants, indicating that each mutation works coordinately in high affinity mutants (Fig. 3h ).

Next, we characterized the binding affinities of the mutant sACE2s for spike RBD using surface plasmon resonance, where IgG1-Fc fused RBD was immobilized as a ligand and the association and dissociation kinetics of his-tagged sACE2 were determined. The KD value of wild type sACE2 was 41.4 nM, whereas those of mutants 3J38 and 3N39 were determined to be 6.5 nM and 0.37 nM (Fig. 4a) .

Analytical size exclusion chromatography (SEC) showed no signs of protein aggregation in the ACE2 mutant samples, confirming that the apparent high affinity was not caused by the avidity effect and the observed KD values represent genuine 1:1 affinity toward RBD (Fig. S4a ). Recombinant soluble ACE2 (rsACE2) was reported to have a fast clearance rate in human blood with a half-life of hours 16, 17 .

Recently, it was demonstrated that a rsACE2 fused with a Fc fragment show high stability in mice 12 as well as higher neutralization activity toward both pseudotyped and authentic SARS-CoV-2 in cultured cells 11 . We formulated our high affinity mutant sACE2s as Fc fusion (sACE2-Fc) and found that the purified proteins were folded well and devoid of aggregation, showing solution behavior indistinguishable from wild type protein (Fig. S4b) . To evaluate their efficacy in neutralizing SARS-CoV-2 infections, affinity-enhanced sACE2-Fc mutants were assayed for viral neutralization against pseudotyped lentivirus and authentic SARS-CoV-2. The IC50 values of wild-type, 3J38, and 3N39 for pseudovirus neutralization in ACE2-expressing 293T cells were 65.2, 8.9, and 0.43 μg/ml, respectively. In the same way, 3N39 neutralized pseudovirus very efficiently with an IC50 value more than 100 times lower than the wild-type in TMPRSS2-expressing VeroE6 cells, (Fig. 4b) . Most importantly, when the neutralization potential against the authentic SARS-CoV-2 in TMPRSS2expressing VeroE6 cells was evaluated, wild-type sACE2-Fc showed no efficiency even at 100μg/ml, whereas 3N39 sACE2-Fc demonstrated significant neutralizing effect in 6.3μg/ml (Fig. 4c) .

SARS-CoV-2 neutralization is one of the preventative or therapeutic approaches against COVID-19. Monoclonal antibodies have become one of the common drug modalities, especially as therapeutics against autoimmune diseases and cancer. As virus-neutralizing antibodies, Palivizumab is clinically used to prevent hospitalization from respiratory syncytial virus infection in high-risk infants 18 , and cocktail of monoclonal antibodies has been shown to reduce mortality from Ebola virus disease 19 .

Engineered recombinant decoy receptor drugs are also developed to neutralize various cytokines including vascular endothelial growth factor, tumor necrosis factor alpha, and CTLA-4 and approved for orbital vascular diseases and rheumatoid arthritis. Recombinant sACE2 or sACE2-Fc fusion protein has potency to neutralize SARS-CoV-2 12,20 , however its modest binding affinity requires higher dose than monoclonal antibody.

We developed the screening system based on the cycle of random mutation and sorting of high affinity population in 293T cells followed by validation of neutralizing activity in a soluble form. In this screening, an additional random mutation was induced in the bulk of sorted mutants, which worked better than mutagenesis in the top mutant. Engineering of decoy receptors with improved affinity was previously reported for cancer-related molecules and ACE2 11, 21, 22 . They used yeast display system to perform directed evolution. Large scale library (~10 7 mutants) was prepared and high affinity mutants were identified by repeating sorting from initial library. Fast growth rate of yeast is suitable for library screening involving repeated sorting and propagation. We on the other hand employed human cells for the display purpose. Since post-translational modification can modulate protein binding affinity, human cell-based screening is better to understand the impact of ACE2 variants on viral affinity and also to proceed biologics development. Repeating mutagenesis after cell sorting without propagation enabled us to conduct screening with relatively small library (~10 5 mutants) and human cells. During the validation, we noticed that high affinity pattern in the flow cytometry assay of full length ACE2 binding RBD-sfGFP did not always correlate with its neutralization activity. Thus, it is evident that experimental validation of each mutation at the level of sACE2 protein was important for efficient identification of high affinity mutants.

Our mutant ACE2s have affinity comparable to typical anti-spike monoclonal antibodies, but they also offer some advantages over antibodies when considered as a drug candidate. Interface of ACE2 to the RBD is larger than that of antibodies, which potentially increases efficacy. Escape mutation to modified ACE2 is likely to result in lower affinity to the native receptor, making such virus much less virulent. SARS-CoV-2 also enters into host cell via endocytosis. SARS-CoV-2 infection is mediated not only by TMPRSS family proteases but by cathepsin L that is catalytically active at pH 3.0-6.5 23 .

Some antibodies are susceptible to impaired affinity at lower pH, leading to lower viral neutralization.

High affinity modified ACE2 fused with Fc is the promising strategy to neutralize SARS-CoV-2. The time frame for running one cycle of mutagenesis and sorting was just one week in our system, and we succeeded in developing optimized mutants in a couple of months without depending on patientsderived cells or tissues. Thus, our system can rapidly generate therapeutic candidates against various viral diseases and may be well suited for fighting against future viral pandemics.

Lenti-X 293T cells were purchased from Clontech and cultured at 37 °C with 5% CO2 in Dulbecco's modified Eagle's medium (DMEM, WAKO) containing 10% fetal bovine serum (Gibco) and penicillin/streptomycin (100 U/ml, Invitrogen). VeroE6/TMPRSS2 cells were a gift from National Institutes of Biomedical Innovation, Health and Nutrition (Japan) and cultured at 37 °C with 5% CO2 in DMEM (WAKO) containing 5% fetal bovine serum (Gibco) and penicillin/streptomycin (100 U/ml, Invitrogen). All the cell lines were routinely tested negative for mycoplasma contamination. For a semi-deep mutational scanning of ACE2 residues 18-102, 30 % of HA positive cells with the highest and 30% with the lowest GFP fluorescence were also collected in 1st mutated library (Fig. S3a) and their genomic DNA was extracted by NucleoSpin Tissue (TAKARA). ACE2 residues 18-102 plasmid sequence was amplified with primers containing adaptor and barcode sequence to perform deep sequencing on the Illumina MiSeq platform using 300nt paired-end protocol. Data were analyzed as follows; high and low gating read count of each mutant was normalized with total counts and log10 ratio of high/low was defined as affinity value. Then, each amino acid mutation-containing mutant affinity values were aggregated.

.

The pcDNA4TO HA-ACE plasmid was transfected into 293T cells (500 ng DNA per ml of culture Kinetic binding measurement using Biacore (SPR)

The binding kinetics of sACE2 (wild-type or mutants) to RBD were analyzed by SPR using a Biacore 

Pseudotyped reporter virus assays were conducted as previously described 25 . A plasmid coding SARS-CoV-2 Spike was obtained from addgene #145032 14 , and deletion mutant CΔ19 (with 19 amino acids deleted from the C terminus) was cloned into pcDNA4TO (Invitrogen) to enhance virus titer 26 .

Spike-pseudovirus with a luciferase reporter gene was prepared by transfecting plasmids (CΔ19, psPAX2, and pLenti firefly) into LentiX-293T cells with Lipofectamine 3000 (Invitrogen). After 48 hours, supernatants were harvested, filtered with a 0.45 μm low protein-binding filter (SFCA), and frozen at -80 °C. ACE2-expressing 293T cells were seeded at 10,000 cells per well in 96-well plate.

Pseudovirus and three-fold dilution series of sACE2-Fc protein were incubated for 1 hour, then this mixture was administered to ACE2-expressing 293T cells. After 1 hour pre-incubation, medium was 

Vero-TMPRSS2 were seeded on 24 well plates (80,000 cells/well) and incubated for overnight. The culture supernatants serially diluted by medium were inoculated and incubated for 2 hours. Culture medium was removed, fresh medium containing 1% methylcellulose (1.5mL) was added, and the culture was further incubated for 3 days. The cells were fixed with 4% Paraformaldehyde Phosphate Buffer Solution (Nacalai Tesque) and plaques were visualized by using a Crystal violet. Table. S1 Amino acid sequence and RBD neutralization activity value of validated mutants.

The value of RBD neutralization activity was calculated as -log2 concentration of 50% RBD-sfGFP bound competing relative to 3N39.

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

A new coronavirus associated with human respiratory disease in China

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

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

A human neutralizing antibody targets the receptor-binding site of SARS-CoV-2

Human neutralizing antibodies elicited by SARS-CoV-2 infection

Potent Neutralizing Antibodies against SARS-CoV-2 Identified by High-Throughput Single-Cell Sequencing of Convalescent Patients' B Cells

Why are RNA virus mutation rates so damn high?

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

Escape from neutralizing antibodies by SARS-CoV-2 spike protein variants

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

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

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

Structural basis of receptor recognition by SARS-CoV-2

Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor

Targeting the degradation of angiotensin II with recombinant angiotensinconverting enzyme 2: prevention of angiotensin II-dependent hypertension

Pharmacokinetics and pharmacodynamics of recombinant human angiotensinconverting enzyme 2 in healthy human subjects

humanized respiratory syncytial virus monoclonal antibody, reduces hospitalization from respiratory syncytial virus infection in high-risk infants. The IMpact-RSV Study Group

Controlled Trial of Ebola Virus Disease Therapeutics

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

An engineered Axl 'decoy receptor' effectively silences the Gas6-Axl signaling axis

Antitumor activity of an engineered decoy receptor targeting CLCF1-CNTFR signaling in lung adenocarcinoma

Genome-wide CRISPR screen reveals host genes that regulate SARS-CoV-2 infection

Human secretory signal peptide description by hidden Markov model and generation of a strong artificial signal peptide for secreted protein expression

Protocol and Reagents for Pseudotyping Lentiviral Particles with SARS-CoV-2 Spike Protein for Neutralization Assays

Retroviral vectors pseudotyped with severe acute respiratory syndrome coronavirus S protein

We would like to thank Sho Hashimoto, Toshiyuki Nishiji, Tomohiro Hino, and Keiko Tamura-