Research ArticleCORONAVIRUS

Antibody-like proteins that capture and neutralize SARS-CoV-2

See allHide authors and affiliations

Science Advances  14 Oct 2020:
Vol. 6, no. 42, eabd3916
DOI: 10.1126/sciadv.abd3916

Abstract

To combat severe acute respiratory syndrome–related coronavirus 2 (SARS-CoV-2) and any unknown emerging pathogens in the future, the development of a rapid and effective method to generate high-affinity antibodies or antibody-like proteins is of critical importance. We here report high-speed in vitro selection of multiple high-affinity antibody-like proteins against various targets including the SARS-CoV-2 spike protein. The sequences of monobodies against the SARS-CoV-2 spike protein were successfully procured within only 4 days. Furthermore, the obtained monobody efficiently captured SARS-CoV-2 particles from the nasal swab samples of patients and exhibited a high neutralizing activity against SARS-CoV-2 infection (half-maximal inhibitory concentration, 0.5 nanomolar). High-speed in vitro selection of antibody-like proteins is a promising method for rapid development of a detection method for, and of a neutralizing protein against, a virus responsible for an ongoing, and possibly a future, pandemic.

INTRODUCTION

The coronavirus disease 2019 (COVID-19) pandemic, caused by the severe acute respiratory syndrome–related coronavirus 2 (SARS-CoV-2), has negatively and deeply affected our lives and societies. To control the COVID-19 pandemic in addition to any emerging, heretofore unknown pathogens in the future, it is of paramount importance to rapidly generate multiple high-affinity antibodies or antibody-like proteins (ALPs) against virus proteins, thus developing both a detection method for the virus (1) and a neutralizing protein against it (2).

Identification of a cross-reactive antibody among those developed against SARS-CoV is the fastest way to obtain a therapeutically useful antibody (37). In principle, however, this approach would yield lower-affinity antibodies with cross-reactivity to the original virus. The isolation of monoclonal antibodies from virus-infected patients is another possible approach (8), but the collection of B cell samples and the identification and large-scale production of effective antibodies represent a long-term research and development project.

In vitro selection is advantageous in terms of the isolation speed of ALPs because time-consuming animal immunization is not necessary to generate ALPs. This method also allows the use of immunoglobulin protein scaffolds, including single-chain variable fragments (9, 10), single-domain antibodies (11), and non-immunoglobulin proteins. Non-immunoglobulin proteins such as a lipocalin (12), a domain of fibronectin (13), the Z domain of protein A (14), the ankyrin repeat motif (15), and the SH3 domain of Fyn (16) allow a greater chance to obtain high-affinity ALPs.

Phage display is the method that is used most commonly for in vitro selection (17). ALP libraries were expressed in Escherichia coli and displayed on the surface protein of phage. The E. coli transformation step using phage DNA, the efficiency of which limited the library size practically from 109 to 1011, is the main disadvantage of phage display. To generate highly diverse protein libraries, typically 1012 to 1013, a cell-free translation system has been used for mRNA display (1820). However, this display method requires multistep manipulations to synthesize protein/mRNA/puromycin linker (PuL) complexes, the transcription of the mRNA, the attachment of the PuL to the mRNA, and the translation of library proteins. These multistep manipulations render mRNA display a time-consuming method, with more than weeks typically being necessary to complete six to seven rounds of the procedure (2 to 3 days per round).

We previously modified the mRNA display method and developed a high-speed in vitro selection method, i.e., transcription-translation coupled with association of PuL (TRAP) display, for the identification of macrocyclic peptides (fig. S1A) (21). In the TRAP display, macrocyclic peptide/mRNA/PuL complexes are automatically synthesized via the simple addition of the peptide template DNA, thus skipping two time-consuming steps: transcription of mRNA and ligation of PuL. We reported the completion of six rounds of selection in approximately 14 hours using the TRAP display and obtained macrocyclic peptides with nanomolar affinity to their target protein. We also performed selection against the vascular endothelial growth factor receptor 2 (VEGFR2) and obtained a macrocyclic peptide with VEGF-induced angiogenesis inhibitory activity (22).

Here, we report the development of an improved TRAP display to facilitate the selection of ALPs (Fig. 1A and fig. S1A). Subsequently, we performed selection trials against the human EGFR1 (HER1) and HER2 using a nanobody (the camelid single-domain antibody) (11, 23, 24) and a monobody (the 10th type III domain of human fibronectin) (13) as backbone proteins. We further used this method to rapidly obtain monobodies against the S1 subunit of the SARS-CoV-2 spike protein. The selected high-affinity monobodies were used to capture SARS-CoV-2 particles from the nasal swab samples of patients. We also tried to increase the affinity of a monobody by dimerization. The monomeric and dimeric monobodies were used as a neutralizing protein against SARS-CoV-2 infection.

Fig. 1 Selection of monobodies against the SARS-CoV-2 spike protein using the TRAP display.

(A) Schematic representation of the TRAP display. Monobody/mRNA complexes were synthesized simply by adding the template DNA to the TRAP system. After reverse transcription (RT), selection, and PCR, the amplified DNAs were added to the TRAP system to reproduce a monobody library for the subsequent round of selection. (B) Structure of the monobody backbone (10th type III domain of fibronectin, 1TTG). The BC and FG loops (labeled in blue) were randomized with the indicated number of residues. (C) Progress of the TRAP display selection. After each round of selection, the recovered cDNA was quantified by real-time PCR. The recovery of cDNA was calculated by dividing the amount of recovered cDNA by the theoretical amount of mRNA/PuL (1 μM). At the fourth round, the selection pressure was increased by decreasing the target concentration from 20 to 2 nM. After the sixth-round selection, the recovered DNA (*) was sequenced. The sequences of the monobodies are shown in Table 1. Stv, streptavidin; Bio, biotin; NTD, N-terminal domain; Pu, puromycin.

RESULTS

Development of the improved TRAP display for ALP selection

We started our study by measuring the display efficiencies of the nanobody and monobody on the mRNA because they determine the diversity of ALP pools and the selection efficiency against a target protein. The monobody DNA template and the DNA-PuL (dPuL; fig. S1B) were added to the TRAP system, followed by determination of the display efficiency of the monobody on the mRNA. In contrast with our expectations, less than 3% of the dPuL formed a monobody/mRNA/dPuL complex; moreover, most of the dPuL (87%) did not form an mRNA/dPuL complex (fig. S1C). As an mRNA sufficient to capture the dPuL (1 μM) was produced within 10 min in the reaction mixture (fig. S2A), an unexpected reaction occurred that prohibited the formation of the mRNA/dPuL complex.

We found that the DNA portion of the dPuL complementary to the 21 mers of the 3′-end mRNA sequence (fig. S1B) was transcribed to RNA in a promoter-independent manner (25) and formed an unwanted DNA/RNA duplex, thus inhibiting formation of the mRNA/dPuL complex (fig. S2B). To prevent promoter-independent transcription, we added a 2′–methoxy (OMe) modification to the ribose because of its reported prevention of T7 promoter–dependent transcription (26). The 2′-OMe modification prevented promoter-independent transcription even after an incubation of 30 min (fig. S2B).

We then used the 2′-OMe–modified PuL (mPuL) in the TRAP display. As expected, formation efficiency of the mRNA/mPuL complex increased from 11% (dPuL) to 60% (mPuL), and display efficiency increased from 3% (dPuL) to 22% (mPuL) (fig. S1C). We also analyzed the display efficiency of the anti-triclocarban nanobody (23, 24) on the mRNA. Formation efficiency of the mRNA/mPuL complex was increased, from 9% (dPuL) to 45% (mPuL); display efficiency of the nanobody on the mRNA also increased, from 2% (dPuL) to 15% (mPuL) (fig. S1C).

The TRAP display was very useful for simplifying the selection procedure, but it was not ideal for first-round selection because the diversity of the library was limited by the amount of the DNA template added in the translation mixture (see fig. S2C for the optimization of DNA concentration). To maximize the diversity of libraries, we usually stock an mRNA pool and add it to the translation reaction after the formation of the mRNA/PuL complex in the first-round selection. Therefore, we also analyzed the display efficiency of the nanobody/monobody on the mRNA in the first-round format. We found that the display efficiencies were similar to that of the TRAP display format for both the monobody (19%) and nanobody (15%) (fig. S1D).

Next, we conducted a selection trial using the monobody and nanobody libraries against EGFR1 and HER2 (fig. S3). High-affinity binders (KD = 1.3 and 3.5 nM for EGFR1 and KD = 4.4 and 13 nM for HER2) were obtained from the monobody library, whereas mid-affinity ones were obtained from the nanobody library (KD = 47 and 32 nM for EGFR1 and KD = 11 and 9.9 nM for HER2).

High-speed selection of monobodies against the SARS-CoV-2 spike protein S1 subunit

According to the results presented above, we focused on the monobody for the subsequent library construction. We introduced 8 or 10 random residues at the BC loop (Pro25-Val29) and 10 or 12 random residues at the FG loop (Gly77-Lys86) of the monobody (Fig. 1B), according to previous reports (13, 20, 27, 28). We used a Tyr-, Ser-, Gly-, and Trp-rich codon mix to construct the random residues (27, 28). The preparation of the mRNA library took 6 months after we ordered the oligonucleotides due to the oligonucleotide synthesis using trinucleotide phosphoramidites (29) and the large-scale polymerase chain reaction (PCR) and transcription procedures that were applied to maintain the diversity of the monobody library. However, once the pool of mRNA was prepared, it could be used for multiple in vitro selection studies against various targets due to the large amount of mRNA produced.

With the pool mRNA in our hands, we received the two biotinylated targets, the SARS-CoV-2 spike protein S1 subunit (S1-biotin) and the receptor binding domain (RBD-HRP-biotin), and started the TRAP display selection (Fig. 1A). In the first round of the selection, an mRNA template, rather than a DNA one, was used to maximize the diversity of the monobody library (1 μM mRNA in a 500-μl scale; monobody display efficiency on the mRNA of about 10%; 3 × 1013 calculated molecules; fig. S4). We used mixed targets in the first round to optimize the effort of the large-scale preparation of the cell-free translation system. From the second round, we used the TRAP display to boost the speed of selection and simultaneously conducted selection procedures against each target. We observed an enrichment of binder monobodies against both targets in the third round of selection (Fig. 1C). To obtain higher-affinity monobodies, we performed an additional three rounds of selection using a 10 times lower target concentration (2 nM). After the six rounds of selection, the recovered complementary DNAs (cDNAs) of monobodies were sequenced. Notably, because of the high-speed in vitro selection feature of the TRAP display (Fig. 1A), we successfully completed the selection within 3 days and obtained the monobody sequences on the 4th day.

RBD-specific monobodies with low- to sub-nanomolar KD values

As the same clones were enriched through the selection against both the S1 subunit and the RBD, we selected the nine most enriched clones in the pool (Table 1) for further study. To confirm the binding activity, we immobilized the purified monobodies (table S1 and fig. S5) on a microwell plate and added either the S1 subunit (S1-HRP) or the (RBD-HRP) conjugated with streptavidin–horseradish peroxidase (Stv-HRP). The results of the enzyme-linked immunosorbent assay (ELISA) showed that the all monobodies were able to bind to both targets (Fig. 2A). The wild-type (WT) monobody (the control protein with the original loop sequences) did not bind to either of the targets. These results clearly indicated that the selected loop sequences were responsible for the binding to the RBD of the S1 subunit.

Table 1 Monobodies obtained by the TRAP display selection against the SARS-CoV-2 spike protein S1 subunit.

The table contains the following information: sequences of the BC and FG loops of the monobodies, the kinetic parameters determined by BLI, the inhibitory activity of the S1 subunit and ACE2 interaction, SARS-CoV-2 S1 subunit specificity, and clones with binding competition. Mutations were identified in the body sequence of clone 6 (S55W), clone 11 (L62M), and clone 12 (D67H). A Ser-to-Thr mutation in the BC loop of clone 11 was introduced during cloning. These mutations were reversed in the clones 6b, 11b, and 12b. Clones TD4, TD6b, and TD12b are tandem dimers of the corresponding clones. N.D., not detected.

View this table:
Fig. 2 Binding activity of each clone against the SARS-CoV-2 spike protein S1 subunit and the RBD.

(A) Determination of the binding domain. A schematic representation of the ELISA is provided on the left. S1-HRP or RBD-HRP (10 nM) was added to the monobody-immobilized microplate. Error bars indicate SD of each experiment (in triplicate). (B) Determination of kinetic parameters by BLI. A schematic representation of the BLI is provided at the top. S1-biotin was immobilized on a streptavidin-sensor chip, and monobodies (2.5 to 40 nM) were used in the kinetic analysis. Some mutations were identified in clones 6, 11, and 12. These mutations were reversed in the clones 6b, 11b, and 12b. The data are depicted in blue, and the fit data to a 1:1 binding model are shown in black. The determined kinetic parameters of the monobodies are provided in Table 1. WT, 10th type III domain of fibronectin; RLU, relative luminescence units; Nus, Nus-Tag fused at the C terminus of the monobody.

To study the stability of monobodies, C-terminally biotinylated monobodies prepared by sortase A reaction (monobody-biotin; fig. S5) were incubated overnight at various temperatures, and the remaining binding activity was measured by ELISA using an S1 subunit–immobilized microplate and Stv-HRP. We found that all monobodies were stable at 37°C and that five of them retained half of the binding activity after incubation at 50°C overnight (fig. S6).

We further assessed the kinetic parameters of all nine monobodies using biolayer interferometry (BLI). S1-biotin was immobilized on the streptavidin-immobilized sensor, and various concentrations of monobodies were added to determine the kinetic parameters based on global fitting (Fig. 2B and Table 1). The seven clones exhibited a sub-nanomolar to nanomolar-level affinity against the S1 subunit (clone 1, KD = 2.47 nM; clone 4, KD = 1.94 nM; clone 6, KD = 0.76 nM; clone 10, KD = 2.02 nM; clone 11, KD = 3.23 nM; clone 12, KD = 1.37 nM; clone 18, KD = 0.65 nM). Since mutations were found in clone 6 (S55W), clone 11 (L62M and S to T in the BC loop), and clone 12 (D67H), we also determined the kinetic parameters of the back mutants (6b, 11b, and 12b). The affinities were similar to those observed for the parental ones, suggesting that these mutations do not contribute to the binding affinity. It was interesting to obtain high-affinity monobodies because the in vitro selection of ALPs sometimes required additional affinity maturation after the initial selection to obtain high-affinity monobodies. We believe that this finding is attributable to either the high antigenicity of the S1 subunit or the high diversity/quality of the library prepared using trinucleotide phosphoramidite synthesis (2729).

Monobodies bound to the SARS-CoV-2 spike protein S1 subunit but not to that of SARS-CoV

Next, we studied the specificity of the monobodies for the SARS-CoV-2 spike protein S1 subunit because this parameter is important for the development of a future diagnostic assay for this virus. Since the SARS-CoV-2 spike protein sequence exhibits greater similarity with that of SARS-CoV than it does with other human coronaviruses (30), we decided to test the specificity of the monobodies against the S1 subunit of SARS-CoV-2 and SARS-CoV. The pull-down experiment using monobody-immobilized beads showed four monobodies (clones 4, 6, 9, and 10) bound to the SARS-CoV-2 S1 subunit but not to the SARS-CoV S1 subunit (Fig. 3A), suggesting that these monobodies are specific for the SARS-CoV-2 S1 subunit. In turn, four monobodies (clones 1, 11, 12, and 18) bound to both targets, and clone 16 did not bind to either of the S1 subunits due to a low binding affinity (Table 1). Clone 16 pulled down a spike protein trimer (Fig. 3A), probably due to a multivalent effect of the target.

Fig. 3 Characterization of monobodies and the application for the sandwich ELISA.

(A) A schematic representation of the pull-down assay is provided at the top. The SARS-CoV S1 subunit (100 nM), SARS-CoV-2 S1 subunit (100 nM), or SARS-CoV-2 S protein trimer (100 nM) were pulled down by monobody-immobilized beads. The supernatant and the heat elution from the beads were loaded onto SDS–polyacrylamide gel electrophoresis (PAGE), followed by staining with SYPRO Ruby. (B) A schematic representation of the ELISA is provided on the left. ACE2-HRP (1 nM) was mixed with each monobody (100 nM) and was added to an S1-immobilized microplate. (C) ACE2-HRP (1 nM) was mixed with each monobody (10 nM clones 4, 6, 9, or 10) and was added to an S1-immobilized microplate (before). Alternatively, a monobody was added to an ACE2-HRP–incubated S1-immobilized microplate (after). (D) A schematic representation of the sandwich ELISA is provided on the left. The S1 subunit (10 nM) was added to the monobody-immobilized microplate. The S1 subunit bound with the capturing monobody was identified by detecting monobody-HRP (10 nM). All possible combinations were tested. (E) Titration curves of the SARS-CoV-2 S1 subunit, SARS-CoV-2 spike protein trimer, and SARS-CoV S1 subunit. After incubation of the analyte with the detecting monobody-HRP (clone 12, 1 nM), the solution was added to the capturing monobody (clone 10)–immobilized microplate. Error bars indicate SD of each experiment (in triplicate). S1, SARS-CoV or SARS-CoV-2 spike protein S1 subunit or SARS-CoV-2 spike protein trimer; N.D., not determined; MK, protein maker (kDa).

Monobodies inhibited the interaction between angiotensin-converting enzyme 2 and the S1 subunit

To determine whether the binding site of the monobodies overlapped with that of the angiotensin-converting enzyme 2 (ACE2), we also tested the inhibition of the S1 subunit/ACE2 interaction (31) afforded by the monobodies. The S1 subunit was immobilized on a microplate, and the premixed solution of ACE2-HRP and each monobody was added. The results of ELISA showed that monobody clones 4, 6, 9, and 10 at a concentration of 100 nM inhibited the binding of the S1 subunit to ACE2 (Fig. 3B); moreover, the inhibition was retained for three of the monobodies (clones 4, 6, and 10) at the low concentration of 10 nM (Fig. 3C). These monobodies also exhibited specificity toward the SARS-CoV-2 S1 subunit (Fig. 3A), probably because the ACE2-binding subdomain of the RBD contains more substitutions than other regions of the spike protein (5, 6).

Application of monobodies in sandwich ELISA

The nine monobodies studied above exhibited binding activity toward the S1 subunit. This observation led us to develop a sandwich ELISA for the detection of the S1 subunit because it requires a pair of antibodies that recognize distinct epitopes of the S1 subunit. To explore for these monobody pairs, we tested all combinations of the cloned monobodies in a sandwich ELISA format and found that the monobodies recognized three different epitopes (Fig. 3D and Table 1). Monobody clones 1, 11, 12, and 18 shared an epitope that was not located at the ACE2-binding surface (Fig. 3, B and D). The FG loop sequences of these monobodies were similar (xLWGYxTxWD; with x representing a nonconserved residue), suggesting that they are the main antigen-binding loop. Monobody clones 6 and 10 shared another epitope that was likely located in the ACE2-binding subdomain (Fig. 3, B and D). These clones exhibited a conserved sequence in the FG loop (xYxGPxxYGxxEx), which will be investigated in a future study. Monobody clones 4 and 9 shared another epitope that was also likely located in the ACE2-binding surface (Fig. 3, B and D). No conserved sequence was detected in the BC loop or the FG loop of these clones. Therefore, we identified 16 pairs that can be used in further sandwich ELISA. Among them, we tested clone 10 as a capturing monobody and clone 12 as a detecting monobody because the pair gave one of the highest signal-to-noise ratios in the sandwich ELISA experiment (Fig. 3D). The results of sandwich ELISA showed that a combination of monobodies could specifically detect the SARS-CoV-2 spike protein trimer and the S1 subunit (Fig. 3E). Although the sensitivity of this system was not sufficient to detect the target in actual samples (103 particles of SARS-CoV-2 in a 100-μl solution, corresponding to 1 fM spike protein), its combination with other detection methods, such as digital ELISA (32) or loop-mediated isothermal amplification (33), may help to overcome the problem of the detection limit in a future study.

Capture of SARS-CoV-2 virions using monobodies from a cultivated sample

Next, we studied the binding of monobodies to SARS-CoV-2. Each monobody-biotin was incubated with SARS-CoV-2 particles in phosphate-buffered saline (PBS), and the viral particles bound to the monobodies were collected on the magnetic beads. A reverse transcription quantitative PCR (RT-qPCR) assay of the SARS-CoV-2 RNA extract in the supernatants and the bead eluents showed that a large proportion of the SARS-CoV-2 particles were detected in the bead eluents for all monobodies, except for the WT monobody (Fig. 4A), proving the binding activity of all selected monobodies toward SARS-CoV-2 particles.

Fig. 4 Monobody binding to SARS-CoV-2 from culture and patient samples.

(A) Monobody binding activities toward SARS-CoV-2 particles. SARS-CoV-2 particles bound with monobody-biotin were pulled down by M280-streptavidin beads. The RNA genomes in the supernatant and on the beads were quantified by RT-qPCR after RNA extraction. (B) The detection limits of the traditional RT-qPCR method and the pull-down RT-qPCR method. Various concentrations of SARS-CoV-2 (0.1 to 10,000 particles/μl) were assayed. SARS-CoV-2 particles were collected from 700 μl of solution by pull down using monobody clone 4, and viral RNA was quantified by RT-qPCR after RNA extraction. Alternatively, the original solution (14 μl) was subjected to RT-qPCR. (C) Pull down of SARS-CoV-2 particles from nasal swab samples (35 μl) of patients. The number of RNA genomes in the supernatant and on the beads was quantified by RT-qPCR after RNA extraction. (D) Pull-down direct RT-qPCR using nasal swab samples of patients. SARS-CoV-2 particles were collected from 400 μl of nasal swab samples by pull down and were directly added to an RT-qPCR reaction mixture. Alternatively, an original nasal swab solution (2.5 μl) was added to an RT–droplet digital PCR (ddPCR) reaction mixture. Error bars indicate SD of RT-qPCR results (in triplicate). N.D., not detected.

On the basis of the experiment described above, we realized that this pull-down method would also be useful for enhancing the detection limit of SARS-CoV-2. To test this idea, we suspended SARS-CoV-2 particles in a PBS solution at various concentrations (0.1 to 10,000 particles/μl) and analyzed the number of particles by pull-down RT-qPCR using monobody clone 4 or by a traditional RT-qPCR. The result showed a wide linear dynamic range of this method (Fig. 4B). The detection limit was increased from 1 to 0.1 particles/μl.

Monobody capture of SARS-CoV-2 virions from nasal swab samples of patients

We then performed a pull down of SARS-CoV-2 particles from nasal swab samples of patients (Fig. 4C). Notably, the actual patient samples (sometimes in the form of highly viscous liquids) were very different from cultured samples, primarily because they contain many unknown components. We tested 16 samples and found that the monobody captured SARS-CoV-2 even in these nasal swab samples. To our knowledge, this is the first example of an antibody or an ALP to capture SARS-CoV-2 from patient nasal swab samples. For samples 8, 15, and 16, the capture rates were notably lower than the remaining cases. We reasoned that this might be because the monobody binding site was hidden by neutralizing antibodies developed in the patients.

Last, we performed a pull-down direct RT-qPCR for nasal swab samples (400 μl) of patients (Fig. 4D). We tested four samples this way and found that the monobody captured SARS-CoV-2 even in higher-volume nasal swab samples. The number of RNA genomes was increased 25 to 77 times after the pull-down technique was used. Moreover, RNA genome in the nasal swab sample 4 was detected only by the pull-down direct RT-qPCR but not by traditional direct RT-qPCR, suggesting the increased sensitivity of the pull-down direct RT-qPCR assay.

Binding activity of monobody tandem dimers

A dimeric form of a protein binder usually showed higher affinity than the monomeric one (7, 16); therefore, we next synthesized the tandem dimers of a monobody. We chose clones 4, 6b, and 12b to synthesize the corresponding tandem dimers (TD4, TD6b, and TD12b) because these clones showed the highest affinity in each group defined by the binding competition assay (Fig. 3D). As expected, the affinities of these tandem dimers were markedly improved (Fig. 5A and Table 1): The KD value displayed less than 1 pM due to the immeasurable koff value (10−7/s).

Fig. 5 The affinity of monobody tandem dimers and the neutralization of SARS-CoV-2 infection.

(A) Determination of kinetic parameters of tandem dimers by BLI. S1-biotin was immobilized on a streptavidin-sensor chip, and the tandem dimers (2.5 to 40 nM) were used in the kinetic analysis. (B) The biotinylated tandem dimers of monobody were immobilized on a streptavidin-sensor chip (*), and the S1 subunit (2.5 to 40 nM) was used in the kinetic analysis. TD4, TD6b, and TD12b are tandem dimers of the corresponding clones 4, 6b, and 12b connected with a (GGGGS)3 linker. The data are depicted in blue, and the fit data to a 1:1 binding model are shown in black. The determined kinetic parameters of the monobodies are provided in Table 1. (C) Monobody-mediated neutralization of SARS-CoV-2 infection in VeroE6/TMPRSS2 cells. The x-axis value indicates the final concentration of the indicated monobody (6b, TD6b, or WT) for each assay well. The experiment was performed with triplicate samples. The copy number of each RNA in the supernatant is plotted.

To understand why the dimerization of monobodies markedly improved the affinity, we first immobilized the decreased amounts of S1 protein on the sensor chip and analyzed monobody dissociation. The results showed that the monobody dimers still dissociated very slowly (fig. S7A). Next, when the monobodies were immobilized on the sensor chip and the S1 protein was used as an analyte, we observed a similar KD value for clone 6b (Fig. 2 versus fig. S7B and Table 1) and no improvement on the KD values for the tandem dimers (Fig. 2 versus Fig. 5B and Table 1). These results suggest that the tandem connection of monobodies does not improve the affinity but bringing up the bridging effect, i.e., an avidity effect.

Monobodies exhibited a high neutralizing activity against SARS-CoV-2 infection

The monobody clone 6b showed high affinity against the S1 subunit of the SARS-CoV-2 spike protein and inhibited the interaction between ACE2 and the S1 subunit (Fig. 3C). These findings led us to analyze the neutralizing activity of monobody clones 6b and TD6b against SARS-CoV-2 infection. We added fourfold serially diluted monobody solutions to VeroE6 cells expressing the transmembrane serine protease (TMPRSS2), VeroE6/TMPRSS2 cells, (34) in 96-well culture plates, followed by infection with SARS-CoV-2 [105 TCID50 (median tissue culture infectious dose)/ml] for 1 hour at 37°C. After 36 hours of incubation with fresh medium, SARS-CoV-2 mRNA amounts in the supernatant were measured by one-step RT-qPCR. The monobody clone 6b showed a very low half-maximal inhibitory concentration (IC50, 0.5 nM; Fig. 5C), even in the monomeric form. Its neutralizing activity was increased in the dimeric form (IC50, 0.4 nM), although the data obtained for 390 pM TD6b fluctuated, probably due to the high sensitivity of VeroE6/TMPRSS2 cells. These IC50 values were comparable to those of the high-affinity human neutralizing antibodies reported recently (3537), indicating the usefulness of the TRAP display for the selection of highly active ALPs.

DISCUSSION

Here, we developed a high-speed in vitro selection method to obtain ALPs against various targets. We demonstrated a constant development of high-affinity binders (KD values in the low- or sub-nanomolar range) against EGFR1, HER2, and the SARS-CoV-2 spike protein. Furthermore, our selected monobodies recognized a recombinant spike protein and SARS-CoV-2 particles, demonstrating the applicability for a sandwich ELISA for a rapid antigen test. Using the monobody before performing pull-down RT-qPCR also increased the detection limit of SARS-CoV-2. The monobody captured SARS-CoV-2 in nasal swab samples of patients with COVID-19. Actual patient samples are very different from cultured samples because they consist of inhomogeneous liquid and contain many unknown components, including various proteases, mucins, and neutralization antibodies. We believe that the monobodies procured in our study will soon be useful to develop effective diagnostic tools and that these tools will contribute to the worldwide effort to overcome the COVID-19 pandemic. The monobody also had a very low IC50 value against SARS-CoV-2 infection, which was comparable to those of the antibodies that were reported recently as being highly active for neutralizing SARS-CoV-2 (3537). Because the backbone of the monobody was the 10th type III domain of human fibronectin, it might be useful as a neutralizing ALP against SARS-CoV-2 infection.

We received the SARS-CoV-2 spike protein S1 subunit on 7 April 2020 as a target of the TRAP display selection and, using the methods described herein, obtained high-affinity monobody sequences on 10 April 2020. The high-speed feature of the improved TRAP display is, therefore, useful for rapid response to the subspecies of SARS-CoV-2 (38, 39) in addition to potential new viruses causing future pandemics.

MATERIALS AND METHODS

Materials

The oligonucleotides were purchased from either FASMAC Co. Ltd. (Japan) or Nippon Bio Service (Japan). The primer/probe set for RT-qPCR and positive control RNA for N2 were purchased from Applied Biosystems Inc. (NY, USA). The synthetic DNAs were acquired from GenScript (NJ, USA). The sequences of the primers and synthetic DNAs are listed in data file S1. SARS-CoV-2 protein active trimer (SPN-C52H8), human ACE2 protein (AC2-H5257), biotinylated 2019-nCoV S1 subunit–Avitag (S1N-C82E8), biotinylated 2019-nCoV S protein RBD-Avitag (SPD-C82E9), SARS-CoV-2 (COVID-19) S1 subunit (S1N-C52H4), and SARS S1 subunit (S1N-S52H5) were all purchased from ACROBiosystems (DE, USA). EGFR/Fc chimera and ErbB2/Fc chimera were obtained from R&D Systems (MN, USA). Creatine kinase, creatine phosphate, and E. coli transfer RNAs were purchased from Roche Diagnostics (Japan). The restriction enzymes were obtained from New England Biolabs (MA, USA).

Preparation of monobody mRNA libraries for selection against SARS-CoV-2 targets

To prepare an A-fragment DNA of the monobody library, FN3F0.F83 (1 μM), FN3F1-2.F29(P) (1 μM), and the FN3FF1coR8.F73 (0.5 μM) or FN3FF1coR10.F79 (0.5 μM) were ligated by T4 DNA ligase (75 μl in total, 75 pmol for each oligonucleotide) with an assistance of Fn3an1.R20(3NH2) (2 μM) and Fn3an2-1.R20(3NH2) (2 μM). As codons for the randomized residues, we used a codon mix with the following ratios: 20% Tyr, 10% Ser, 15% Gly, 10% Trp, and 3% each of all the other amino acids except for Cys, which is similar to the original cocktail (30% Tyr, 15% Ser, 10% Gly, 5% Phe, 5% Trp, and 2.5% each of all the other amino acids except for Cys) (27, 28). After the ligation, the entire mixture was added to the reaction mixture [10 mM tris-HCl (pH 8.4), 100 mM KCl, 0.1% (v/v) Triton X-100, 2% (v/v) dimethyl sulfoxide, 2 mM MgSO4, 0.2 mM each deoxynucleotide triphosphate (dNTP), 0.375 μM T7SD8M2.F44, 0.375 μM FN3BsaI.R40, and 2 nM Pfu-S DNA polymerase] and the amplified by PCR (15 ml in total, seven cycles of PCR). B-fragment DNA was prepared by the same procedure using FN3FF2co.F72(p), FN3F3coR10.F70(p), FN3F3coR12.F76(p), and Fn3an3.R20(3NH2) for ligation and FN3BsaI.F33 and FN3Pri2.R44 for amplification.

The amplified A-fragment DNA and B-fragment DNA were purified by phenol/chloroform extraction and isopropanol precipitation. One end of each DNA product was digested with Bsa I (New England Biolabs, MA, USA), as per the manufacturer’s protocol, and the DNA products were then purified by phenol/chloroform extraction and isopropanol precipitation. The products were ligated to each other (1 μM, 200 μl) to synthesize full-length DNA products, and they were amplified using T7SD8M2.F44, G5S-4Gan21-3.R42, and Pfu-S DNA polymerase (60 ml in total, four cycles of PCR). The products were purified through phenol/chloroform extraction and isopropanol precipitation. The DNA template was transcribed by in vitro run-off transcription, and the mRNA was purified by isopropanol precipitation, followed by polyacrylamide gel electrophoresis (PAGE) purification. The mRNA/hexachloro-fluorescein (HEX)–mPuL was prepared by a similar procedure described above. The resulting complex was directly used in the first round of selection.

In vitro selection of monobodies against SARS-CoV-2 spike protein S1 subunit and RBD of the S1 subunit by the improved TRAP display

For first-round selection, 1 μM mRNA/puromycin-OMe linker was added to a reconstituted translation system, and the reaction mixture (500 μl) was incubated at 37°C for 30 min. After the reaction, 41.7 μl of 200 mM EDTA (pH 8.0) was added to the translation mixture. A reverse transcription buffer [41.1 μl of 0.78 M tris-HCl (pH 8.4), 1.16 M KCl, 0.37 M MgCl2, and 0.08 M dithiothreitol (DTT)], 5 mM dNTPs (66.7 μl), 100 μM FN3S.R29 (10 μl), and 28.7 μM in-house moloney murine leukemia virus reverse transcriptase (HMLV) (27.5 μl) were added to the translation mixture, and the resulting solution was incubated at 42°C for 15 min. The buffer was exchanged to HBST buffer [50 mM Hepes-KOH (pH 7.5), 300 mM NaCl, 0.05% (v/v) Tween 20] using Zeba Spin Desalting Columns. To remove the bead binders, the resulting solution was mixed with Dynabeads M-280/M-270 streptavidin (1:1) (Thermo Fisher Scientific) at 25°C for 10 min. The supernatant was mixed with 4.8 μl of 7.3 μM biotinylated 2019-nCoV S1 subunit–Avitag and 2.1 μl of 16.5 μM biotinylated 2019-nCoV S protein RBD-Avitag and then incubated at 25°C for 5 min. After recovering the target proteins with Dynabeads M-270 streptavidin, the resulting beads were washed with HBST buffer thrice, and PCR premix (1 ml) was added. The beads were heated at 95°C for 5 min, and the amount of the eluted cDNAs was quantified by SYBR green–based qPCR using T7SD8M2.F44 and FN3Lip.R20 as primers. The eluted cDNAs were PCR-amplified using T7SD8M2.F44, G5S-4Gan21-3.R42, and Pfu-S DNA polymerase and was purified by phenol/chloroform extraction and isopropanol precipitation. The posttranslation and post–reverse transcription reaction mixtures were analyzed using the same procedure as described above.

For the second-round selection, the resulting DNA (about 3 to 10 nM in final concentration) was added to the TRAP system, and the reaction mixture (40 μl) was incubated at 37°C for 30 min. After the reaction, 8 μl of 100 mM EDTA (pH 8.0) was added to the translation mixture. Reverse transcription mixture [24 μl of 150 mM tris-HCl (pH 8.4), 225 mM KCl, 75 mM MgCl2, 16 mM DTT, 1.5 mM dNTPs, 7.5 μM primer, and 3.4 μM HMLV] were added to the translation mixture, and the resulting solution was incubated at 42°C for 15 min. The buffer was exchanged to HBST buffer using Zeba Spin Desalting Columns. To remove the bead binders, the resulting solution was mixed thrice with M270/M280 magnetic beads (1:1) at 25°C for 20 min. Each target (20 nM final concentration) was added to the half portion of the supernatant. After removing the supernatant, the beads were washed with HBST buffer thrice, and the PCR premix was added to the beads. Quantitation of cDNA and amplification and purification of DNA were performed by the same procedure as described for the first-round selection.

For the third- to sixth-round selection, the procedure was performed similarly as that of the second round, except for the volume of the reaction mixtures (5 μl), the target concentration (2 nM in the fourth to sixth rounds), and bead washing conditions (an extensive wash in the sixth-round selection against S1 subunit). After the sixth-round selection, the sequences of the recovered DNAs were analyzed using an Ion Torrent instrument (Thermo Fisher Scientific).

Affinity measurement of selected nanobodies and monobodies

The affinity measurement was performed against the EGFR/Fc chimera or the ErbB2/Fc chimera immobilized on an anti-human immunoglobulin G Fc capture biosensor (ForteBio), biotinylated 2019-nCoV S1 subunit–Avitag, or biotinylated monobodies immobilized on a streptavidin biosensor (ForteBio), using Octet system (ForteBio, CA, USA), as described in the manufacturer’s instructions. The binding assay was performed at 30°C in the buffer D. Each step in the binding assay was as follows: equilibration for 150 s, association for 900 s, and dissociation for 900 s (EGF and HER2); equilibration for 150 s, association for 600 s, and dissociation for 600 s (S1 subunit). Buffer D′ [50 mM Hepes-KOH (pH 7.5), 300 mM NaCl, 0.1% (v/v) Tween 20, and 1% (w/v) PEG-6000] was used when the S1 subunit was an analyte.

Binding assay of monobodies against SARS-CoV-2 S1 or S1 RBD proteins

The ELISA microplate with 96 wells was coated with 100 μl of 100 nM monobody in HBS2 buffer [25 mM Hepes-K (pH 7.5) and 150 mM NaCl] overnight. The microplate was washed once with HBST2 buffer [HBS2 with 0.05% (v/v) Tween 20], blocked with 1% (w/v) bovine serum albumin (BSA) in HBS2 buffer at room temperature for 1 hour, and then washed once with HBST2 buffer. S1-HRP or RBD-HRP was prepared by mixing S1-biotin or RBD-biotin with the same amount of Stv-HRP in HBST2BP buffer [HBST2 with 0.1% (w/v) BSA and 0.1% (w/v) PEG-6000] and was added to the microplate with a final concentration of 10 nM. The microplate was incubated at room temperature for 1 hour. The microplate was washed five times with HBST2, and 200 μl of SuperSignal ELISA Pico was added to each well. The chemiluminescence was measured immediately using the SpectraMax M5 plate reader (Molecular Devices).

Pull down of S proteins with monobody-immobilized beads

Biotinylated monobody (1 μM, 3.3 μl) in HBST2BP buffer was mixed with Dynabeads M-280 streptavidin (10 mg/ml, 6.6 μl, prewashed by HBST buffer) at room temperature for 5 min. After removing the supernatant, the beads were washed twice with HBST buffer, and 2.5 μl of 100 nM each analyte (SARS-CoV-2 spike protein S1 subunit, SARS-CoV spike protein S1 subunit, and SARS-CoV spike protein trimer) in buffer D was added to a one-third fraction of the beads. After mixing at room temperature for 30 min, the supernatant was collected. The beads were washed twice with HBST buffer. The supernatant and heat elution from the beads were loaded on SDS-PAGE gel and was stained using SYPRO Ruby.

Inhibition of ACE2/CoV-2 S1 subunit interaction by monobodies

ELISA was performed using a microplate with 96 wells, coated with 100 μl of 2 nM SARS-CoV-2 spike protein S1 subunit in HBS2 buffer overnight. The microplate was washed once with HBST2 buffer, blocked with 1% (w/v) BSA in HBS2 buffer at room temperature for 1 hour, and then washed once with HBST2 buffer. Biotinylated ACE2 complexed with same amount of Stv-HRP in HBST2BP buffer was added to a final concentration of 1 nM together with 100 or 10 nM monobody. Alternatively, a monobody (10 nM in total) was added to the ACE2-HRP–preincubated S1-immobilized microplate. The microplate was incubated for 1 hour at 37°C. The plate was then washed five times with HBST2 buffer. After the addition of SuperSignal ELISA Pico (200 μl) to each well, the chemiluminescence was recorded immediately using a SpectraMax M5 plate reader.

Sandwich ELISA using monobodies

The sandwich ELISA was performed using a microplate with 384 wells, which was coated with 25 μl of 100 nM of the capturing monobodies (nine clones and WT) in HBS2 buffer overnight. The microplate was washed once with HBST2 buffer, blocked with 1% (w/v) BSA in HBS2 buffer at room temperature for 1 hour, and then washed once with HBST2 buffer. SARS-CoV-2 S1 subunit (10 nM) in HBST2BP buffer was added. The microplate was incubated at room temperature for 1 hour and was subsequently washed once with HBST2 buffer. Biotinylated detecting monobody (nine clones and WT) complexed with same amount of Stv-HRP in HBST2BP buffer was added to a final concentration of 10 nM, and the plate was incubated at room temperature for 30 min. The microplate was washed five times with HBST2 buffer, and 50 μl of SuperSignal ELISA femto (Thermo Fisher Scientific) was added to each well. The chemiluminescence was immediately recorded using the SpectraMax M5 plate reader.

For the other sandwich ELISA, a microplate with 96 wells was coated with 100 μl of 100 nM monobody clone 10 in HBS2 buffer overnight. The microplate was washed once with HBST2, blocked with 1% (w/v) BSA in HBS2 buffer at room temperature for 1 hour, and washed once with HBST2 buffer. One of the analytes, the SARS-CoV-2 S1 subunit, the spike protein trimer, or the SARS-CoV S1 subunit, was incubated with the detecting monobody-HRP (clone 12, 1 nM) on ice for 1 hour. The solution was then added to the capturing monobody–immobilized microplate and was incubated at room temperature for 1 hour. The microplate was washed five times with HBST2 buffer, and 200 μl of SuperSignal ELISA femto was added to each well. The chemiluminescence was immediately recorded using the SpectraMax M5 plate reader.

Binding assay of monobodies toward SARS-CoV-2 particles

The SARS-CoV-2 virus was isolated from VeroE6 cells (American Type Culture Collection) from a nasal swab sample of a patient in the Nagoya Medical Center, Japan. Virus-infected cells were grown in Dulbecco’s modified Eagle’s medium (Sigma-Aldrich) supplemented with 10% fetal bovine serum and penicillin (100 U/ml) and streptomycin (100 μg/ml) (Thermo Fisher Scientific). Four days after infection, the supernatant was harvested. The supernatant was clarified by centrifugation and filtration through a 0.45-μm pore-size syringe filter (Merck Millipore) and used for the virion-binding assay. The virus particle amounts were measured using droplet digital PCR (ddPCR) using the One-Step RT-ddPCR Advanced Kit for Probes (Bio-Rad) with a primer/probe set for N2 (data file S1) (40).

SARS-CoV-2 particles (5 × 104 particles/140 μl) and 10 nM monobody in HBST buffer were incubated at 25°C for 10 min. This was followed by the addition of Dynabeads M-280 streptavidin (10 mg/ml, 10 μl, prewashed by PBS buffer), which were incubated with the solution for 10 min. After the incubation, the supernatant was collected for further analysis. The beads were washed three times with HBST and were suspended with 140 μl of HBST buffer. Viral RNA was extracted from both the supernatant and the bead suspension using the QIAamp Viral RNA Mini Kit (QIAGEN), as described in the manufacturer’s protocol. The 5 μl of the 60 μl of RNA-containing solution was added to an RT-qPCR reaction mixture (One-Step PrimeScript III RT-qPCR Mix, Takara Bio) using a primer/probe set for N2 (40). The target copy numbers were determined using the Thermal Cycler Dice Real Time System III (Takara Bio), which constantly showed about three times higher number of RNA genomes than that of above RT-ddPCR.

Pull-down RT-qPCR for study of the detection limit of SARS-CoV-2 particles

SARS-CoV-2 virus particles (0.1 to 10,000 particles/μl based on viral RNA coy number, 700 μl) were incubated with 2 nM monobody clone 4 in PBS [10 mM phosphate (pH 7.4), 2.7 mM KCl, and 137 mM NaCl] containing 0.05% (v/v) Tween 20 at room temperature for 10 min. The solution was mixed with Dynabeads M-280 streptavidin (10 mg/ml, 8 μl, prewashed by PBS buffer) at 25°C for 10 min. The beads were washed with HBST three times and were suspended in HBST. Viral RNA from the bead suspension or the original solution (14 μl) was extracted using the QIAamp Viral RNA Mini Kit. For RT-qPCR, the One-Step PrimeScript III RT-qPCR Mix with the N2 primer/probe set was used. The target copy numbers were determined using the Thermal Cycler Dice Real Time System III (Takara Bio), as aforementioned.

Pull down of SARS-CoV-2 in nasal swab specimens from patients

For the pull-down assay using nasal swab specimens from patients, we used residual nasal samples after diagnostic tests. The study was approved by the ethical committee at the Nagoya Medical Center (registration no. 2019-087) and conducted according to the tenets of the Declaration of Helsinki. Written informed consent for the use of the residual samples was obtained from all participants. Thirty-five microliters of each nasal sample was mixed with 2 nM monobody clone 4 to a total volume of 100 μl in HBST buffer and incubated at 25°C for 10 min. The mixture was then incubated for another 10 min with Dynabeads M-280 streptavidin (2.5 mg/ml, 40 μl, prewashed by HBST buffer). The RNA copies in both the supernatant and the bead fraction were measured using RT-qPCR, as described above.

Pull-down direct RT-qPCR assay of SARS-CoV-2 in nasal swab specimens from patients

Each residual sample of nasal swab from four patients was used for this assay. Briefly, 0.5% Tween 20 (45 μl) and monobody clone 4 (5 μl of 200 nM) were added into 400 μl of the swab sample in PBS. The mixture was incubated at 25°C for 10 min, followed by addition of Dynabeads M-280 streptavidin (2.5 mg/ml, 10 μl, prewashed by HBST buffer). After 10 min of incubation, the beads were washed with 1 ml of HBST three times and resuspended with 2.5 μl of HBST containing carrier RNA. The bead suspension or the original swab solution (2.5 μl) was mixed with an RT-qPCR reaction mixture (47.5 μl; PrimeDirect Probe RT-qPCR Mix, Takara Bio) with a primer/probe set (data file S1). The RNA copy numbers were determined using the Thermal Cycler Dice Real Time System III (Takara Bio), as aforementioned. The detection level of positive control RNA using the PrimeDirect Probe RT-qPCR Mix was ≳25 copies per well.

Virus neutralization assay

SARS-CoV-2 neutralization assay was performed using VeroE6/TMPRSS2 cells (34) that were obtained from the JCRB Cell Bank, Ibaraki, Japan. The cells (5 × 103 cells per well, 50 μl) were seeded in 96-well culture plates and incubated at 37°C for 18 hours before infection. Monobodies were fourfold serially diluted (from 400 nM to 24.4 pM), and 20 μl of the diluted monobodies were added into each culture well. Cells were infected with 10 μl of SARS-CoV-2 (105 TCID50/ml) for 1 hour at 37°C. The supernatant was removed, and 80 μl of Dulbecco’s modified Eagle’s medium (Sigma-Aldrich) was supplemented with 10% fetal bovine serum and penicillin (100 U/ml) and streptomycin (100 μg/ml) (Thermo Fisher Scientific). After incubation at 37°C supplied with 5% CO2 for 36 hours, the culture supernatants were harvested. The SARS-CoV-2 RNA amounts in the supernatants were measured by RT-qPCR using the SARS-CoV-2 Direct Detection RT-qPCR Kit (Takara Bio). The IC50 values were determined using the GraphPad Prism version 6.0.

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/sciadv.abd3916/DC1

https://creativecommons.org/licenses/by-nc/4.0/

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

REFERENCES AND NOTES

Acknowledgments: We thank Y. Baba and Y. Hasegawa for initial arrangement of the collaboration between the Murakami group and the Iwatani group for this research. Funding: This work was supported by the Grant-in-Aid for Scientific Research (A) (General) (grant number 15H02006 to H.M.), Grant-in-Aid for Scientific Research (B) (General) (grant number 19H03482 to Y.I.), Grant-in-Aid for Scientific Research on Innovative Areas (grant number 20H04704 to G.H.), and Grant-in-Aid for Early-Career Scientists (grant number 18K14332 to T.F.) from the Japan Society for the Promotion of Science and a donation from H.M. Author contributions: T.K., T.F., S.U., K.I., S.K., and T.S. carried out the nanobody/monobody selection experiment. Y.I. designed the virus detection study, and K.M. conducted the experiment. Y.Y. collected the virus samples. H.M. wrote the manuscript with support from T.K., T.F., Y.I., and G.H. H.M. supervised the project. Competing interests: T.K., Y.I., T.F., S.U., G.H., and H.M. are inventors on the provisional patent application (U.S. application no. 63/026,802, filed 19 May 2020) submitted by the Tokai National Higher Education and Research System and National Hospital Organization Nagoya Medical Center that covers monobody sequences against the SARS-CoV-2 spike protein. All other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. The monobody genes can be provided by H.M., as well as pending scientific review and a completed material transfer agreement. Requests for the genes should be submitted to H.M. Additional data related to this paper may be requested from the authors.
View Abstract

Stay Connected to Science Advances

Navigate This Article