Nanobodies against SARS-CoV-2 RBD from a Two-step Phage Screening of Universal and Focused Synthetic Libraries

Coronavirus disease 2019 (COVID-19) is an evolving global pandemic, and nanobody (Nb) is recognized as a potential diagnostic and therapeutic tool for infectious disease. Here, we designed and synthesized a humanized and highly diverse phage Nbs library hsNb-U (Humanized synthetic Nbs Library - Universal). We expressed and puri�ed the SARS-CoV-2 receptor-binding domain (RBD), and screened this univeral library against the RBD protein target. Then, the CDR1 and CDR2 sequences of �ve leads obtained from the hsNb-U phage panning were combined with randomly mutated CDR3 to construct a targeted (focused) phage display library, hsNb-RBD, for subsequent phage panning and screening. From the obtained sequences, we expressed 45 unique anti-RBD candidate Nbs. Among the selected Nbs, eight were found to be highly expressed, and �ve of these show high-anity to RBD (EC 50 less than 100nM). Finally, we found that Nb39 can compete with angiotensin converting enzyme 2 (ACE2) for binding to RBD. Overall, this two-step strategy of synthetic phage display libraries enables rapid selection of SARS-CoV-2 RBD nanobodies with potential therapeutic activity, and this two-step strategy can potentially be used for rapid discovery of Nbs against other targets.


Background
Since December 2019, a novel, highly transmissible severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2, COVID-19) [1,2] has erupted on a large scale worldwide and spread rapidly.As of May 2023, more than 750 million people have been infected and about 7 million lives have been claimed.These numbers are still rising.The global COVID-19 pandemic poses serious challenges to patients, health care systems, and economic and social activity.
Although the SARS-CoV-2 vaccine is widely used around the world, the vaccine's protective effect is greatly reduced in people with weakened immunity system, such as elderly or people with immune-compromised conditions.Vaccine alone is not enough to end the pandemic.The development of neutralizing antibodies or related passive immunization molecules to prevent and treat SARS-CoV-2 will always be unmet needs [3].
Monoclonal antibodies (mAb) have had tremendous success in treating a variety of disease, and several mAb have been approved for the treatment of COVID-19 [4][5][6][7][8][9].However, virus like SARS-CoV-2 continuously evolve through mutations during genome replication.Since the outbreak, multiple variants of SARS-CoV-2 have been identi ed [10], named in Greek letters by the World Health Organization (WHO).Omicron variant is more transmissible than other variants and is insensitive to some of the anti-SARS-CoV-2 mAbs that have been developed for treatment or prophylaxis [11,12].
The high production costs, large doses needed, and low-temperature requirements for transportation and storage associated with traditional mAbs make it challenging to be cost-effective for large scale applications.Single-domain antibodies or variable domain of the heavy chain of HACbs (VHH) from camelids, having a smaller molecular weight (15 kD) than traditional mAbs (150 kD), are commonly termed nanobodies (Nbs) [13,14].Nbs have shown great potential in biomedical applications, including cancer, infection, in ammation, and other diseases [15][16][17].The rst Nbs -based medicine was approved to treat acquired thrombotic thrombocytopenic purpura (aTTP) in 2018 [18].
Nbs have several important advantages over traditional mAbs, such as higher thermal stability, higher solubility, and easier penetration into tissues [19].Due to their minimal size, they are particularly suited to reach hidden epitopes such as crevices of target proteins [20].Nbs can be expressed in prokaryotic systems with lower production costs [21,22].Although for therapeutic applications, it will be necessary to produce under GMP, which may increase the production cost to a level comparable to mAbs production.Nbs can be easily bioengineered into novel bivalent / multivalent / multispeci c and high-a nity molecules [23,24].Especially, Nbs are stable and small, and they can be aerosolized for direct delivery to the lungs.Nbs provide possible opportunities for rapid production of antiviral drugs.
At present, SARS-CoV-2 Nbs are mainly obtained through the "in vivo" method [25][26][27][28][29][30][31].The recombinant spike (S) protein or receptor-binding domain (RBD) protein is used to immunize camelid animals such as camels or alpacas.However, "in vivo" screening methods require a long development period (usually > 3 months) from antigen to nal speci c Nbs. So, it is di cult to develop and generate antibodies against new virus mutant strains or new viruses quickly and at low cost.Therefore, rapid and e cient "in vitro" screening becomes an important approach by combining phage display technology with naïve [32] or synthetic [33][34][35] Nbs libraries.
Here, we designed and synthesized a highly diverse phage library of humanized Nbs, and screened it to obtain Nbs binding to RBD.The CDR1 and CDR2 of the obtained lead sequences were then assembled with randomly mutated CDR3 to construct a focused (targeted) library, hsNb-RBD, speci cally targeting RBD.By this two-stage method, Nbs can be selected within ten working days, which is considerably faster than "in vivo" method, which requires the repetitive immunization prior to binder selection by phage display.Finally, we screened and obtained several Nbs that bind to RBD with high a nity.Overall, we have established an e cient two-stage method for rapid development of humanized Nbs targeting SARS-CoV-2 RBD.This two-stage method has the potential to strengthen our ability to respond to the current COVID-19 pandemic, possible new variants and even other diseases.

Results
1. Design and construction of a synthetic Nb library hsNb-U.
We used a Nbs framework (Fig. 1-2) based on a soluble human germline immunoglobulin heavy-chain variable region (IGHV3-66*01), which has been shown to be an ideal alternative to camel Nbs [36].For highly variable complementary determine regions (CDRs), the sequence library (Fig. 1-2) was designed based on the work of McMahon et al [34].
A xed size of 7 amino acids (AAs) was chosen for the CDR1 and 12 AAs was chosen for the CDR2.It is observed that Nbs CDR3 varies greatly in length, and contributes most to antigen-binding a nity and speci city.Therefore, three sizes (10 AAs, 14 AAs and 18 AAs) were chosen for CDR3.For variable positions not fully randomized (such as the rst amino acid in CDR1 region), we designed degenerate codons using a web-based tool SwiftLib [37].The designed DNA sequence is further optimized for E. coli expression based on codon usage.The sub-library was constructed using NNC and NNK respectively for fully randomized amino acids, and then combined in the phage screening.The full-length Nb sequences were assembled by overlap-extension PCR from DNA fragments with degenerated codons.The full-length Nb sequences were cloned into the pComb3Xss vector, and then transformed into E. coli TG1 strain by electroporation in multiple batches.Finally, > 1 10 9 clones were obtained.Quality control was carried out using Sanger sequencing of 50 randomly picked clones, and no redundant clone was found.In this way, we generated a large and highly diverse library of humanized Nbs library hsNb-U (Humanized synthetic Nbs Library-Universal) (as shown in Fig. 1-2).

Initial screening of anti-RBD Nbs
SARS-CoV-2 expresses a surface Spike (S) glycoprotein composed of two subunits, S1 and S2, and forms the homotrimeric S protein [38] which can interact with host cells.The interaction between the SARS-CoV-2 and the host cell is mediated by the RBD of the S1 subunit, which binds to the peptidase domain (PD) of Angiotensin-converting enzyme 2 (ACE 2) [38].After that, the S2 subunit undergoes a drastic conformational change and triggers membrane fusion [38].Therefore, the S protein RBD has become one of the most important targets for developing SARS-CoV-2 antibodies.
We obtained the recombinant SARS-CoV-2 S protein RBD (as shown in Fig. 3A-B) through the Bac-to-Bac baculovirus expression system and puri ed by Ni-NTA a nity column followed by Superdex-75 gel ltration column.Two rounds of phage panning were performed against the recombinant RBD from the hsNb-U library.Enrichment of Nbs-displaying phages against the RBD were monitored by enzyme-linked immunosorbent assay (ELISA).Candidate phages enriched more than 5-fold over the bovine serum albumin (BSA) control protein were selected as initial leads.We identi ed 5 high-a nity clones: H1, F3, E5, A6, H6 (Fig. 3D).

Anti-RBD Nbs screening from the focused library
From the hsNb-RBD library, two rounds of panning were performed against the recombinant RBD.Dozens of candidate Nbs-displaying phage clones with more than 30-fold enhanced ELISA signals over BSA were selected (Fig. 4).On average, the Nbs phage clones screened from the targeted hsNb-RBD library had stronger a nity to the RBD than those from the universal hsNb-U library.These results showed that hsNb-RBD library, with CDR1 and CDR2 optimized according to the lead sequences, had improved neutralizing capacity against RBD compared with the initial library hsNb-U.

Binding a nity of the top Nb candidates
For E. coli expression of phage-screened Nbs, we selected 45 highest-a nity clones with diverse CDR sequences to encompass a variety of biophysical, structural and potentially different antiviral properties.We found that 13 of them had no expression (28.8%), 23 had low expression (51%), and only 8 candidates had high expression (17.7%) (Fig. 5A).We puri ed these 8 Nbs and tested their RBD binding by ELISA, from which we identi ed ve high a nity RBD-speci c Nbs (Fig. 5B-E).Amino acid sequences of the ve highest a nity RBD Nb candidates were shown as Fig. 5E.Among them, Nb39 and Nb42 had the highest a nity to RBD, however, the Nb42 protein was less stable and precipitated after freezing and thawing.Nb39 has good stability, high a nity to RBD, and the half-inhibitory concentration reaches about 10 nM (Fig. 5B-E).

Candidate Nbs compete with ACE2 for RBD binding
To test whether the Nb candidates compete with angiotensin converting enzyme 2 (ACE2) for RBD binding.We obtained the extracellular domain of ACE2 protein through the Bac-to-Bac baculovirus expression system, and puri ed it by Ni-NTA a nity column followed by Superdex-75 gel ltration column (Fig. 6A).
To investigate whether the binding of candidate Nbs to RBD is competitive with ACE2, we developed a competition assay to determine the binding of Nb39 to RBD with and without ACE2 by microscale thermophoresis (MST).The binding a nity of Nb39 to RBD is similar as the one (Kd = 14.0 ± 11.6 nM) determined by the ELISA method without ACE2 (Fig. 6B).However, the observed a nity of Nb39 to RBD drops 10-fold (Kd = 140.3± 91.8 nM) in the presence of 5nM ACE2 (Fig. 6B).The Kd values were analyzed using the MO.A nity Analysis software (NanoTemper Technologies).This indicates that Nb39 can bind the same site on RBD as ACE2 does.Also, the results indicate Nb39 and ACE2 have a similar binding a nity to RBD, and Nb39 binds RBD of the SARS-CoV-2 spike protein and inhibits the interaction between RBD and ACE2.

Discussion
Nbs have several important advantages over traditional antibodies, including low cost, high thermal stability, small molecular weight, and nebulization for direct delivery to the lungs.Nbs can be nebulized, inhaled, and administered directly to the site of infection, with rapid onset of action, high local drug concentration/high bioavailability, and high patient compliance (needle-free) [40,41], making them very attractive agents against respiratory viruses.In recent years, research on their application against respiratory pathogens has also accelerated.For example, Nbs against MERS-CoV [42], H1N1 [43], H5N1 [44], in uenza [45] and so on has been documented.ALX-0171, a trivalent Nb that neutralizes Respiratory syncytial virus (RSV), directly prevents or treats RSV infection in the lungs of cotton rats [41].
High-quality Nbs are promising candidates for the treatment of COVID-19 pneumonia [46][47][48].The development of highly effective anti-SARS-CoV-2 Nbs may provide an important means for multifunctional, cost-effective prevention, treatment and immediate diagnosis.Xiang et al. immunized camels with recombinant RBD and identi ed several high-e ciency SARS-CoV-2 neutralizing Nbs using proteomics methods [28].These heat-stable Nbs can be massproduced rapidly by microorganisms, and are resistant to freezing, drying and aerosolization [28].They further developed the most e cient tri-valent Tri-Nb21 into PiN-21 aerosol, which can effectively prevent and treat Syrian hamsters infected by SARS-CoV-2 at an ultra-low dose, greatly reducing viral load and preventing lung damage and viral pneumonia [27,28].
The synthetic Nbs library uses gene synthesis technology to introduce random DNA sequences at speci c sites, which is highly controllable and allows fast and e cient screening against target proteins.However, due to combinatorial explosion, the possible Nbs sequences space is extremely vast and much larger than the capacity of phage library.In this study, we rst screened universal hsNb-U library against the target protein, and obtained ve high-a nity lead sequences.We further assembled the CDR1 and CDR2 sequences from these ve initial leads with randomly mutated CDR3 to construct a second library hsNb-RBD, which is more focused on the target protein (RBD).Multiple Nb candidates were obtained from the hsNb-RBD library.Among them, Nb39 has good stability, high binding a nity to RBD with the EC50 reaching 3nM, and Nb39 can compete with ACE2 for binding to RBD.
The probability of screening high-a nity RBD-binding Nbs from the focused hsNb-RBD library is higher than from the universal hsNb library, and the identi ed Nbs have higher a nities than the lead sequences.These results validate the feasibility and effectiveness of this two-stage screening strategy, where a universal and diverse library is initially screened to obtain lead sequences for construction of an antigen-speci c library for a second-stage screening (Fig. 7).This strategy allows screening of high a nity SARS-CoV-2 Nbs within ten working days, which is relatively quickly compared to immunized animals.Moreover, compared to the traditional one-step synthetic library screening (universal library in the rst step), the second step focused library is more e cient in screening.This strategy can be extended to the screening of other targets.
In addition, the current rapid development of computational technology and arti cial intelligence (AI) has facilitated the development of protein structure prediction and computer-aided drug design.At present, large-scale co-evolution analysis is the commonly used algorithms for predicting the 3D structure of proteins based on gene sequences and performs quite well [49].This algorithm is used by Google's AlphaFold, which can accurately predict protein structure from protein sequence within minutes [50,51].With the development of theoretical chemistry and computational biophysics, our understanding of the physical nature of protein folding and interactions has improved, and various software has been developed to model and simulate proteins and other biomolecules.David Baker et al. used their Rosetta software to design a variety of proteins with unnatural structures or proteins with high a nity to speci c targets [52].The Rosetta software has also been extended to perform high-precision modeling of antibodies [53].Computer-aided nanobody development may become a very important tool in the future.
In summary, here we report a synthetic Nb platform for rapid screening of anti-RBD Nbs (Fig. 7), and this pipeline can be extended to screening of other targets.These Nbs may be promising candidates for COVID-19 prevention, treatment, or as reagents to facilitate SARS-CoV-2 vaccine development.This twostep strategy can be used to rapidly developed new Nbs against mutant virus strains, and address the need for continues virus mutation in a pandemic.We believed that ongoing research will surely pave the way to a safer world.

1.Construction of humanized Nbs library
The full-length Nb DNA sequences were obtained by a series of overlap-extension PCR (OE-PCR) using oligonucleotides (with degenerated codons) purchased from external primer synthesis services.PCR was performed using high delity DNA polymerase (Phusion Green DNA polymerase, Thermo sher) for 20~32 cycles (with annealing temperature chosen according to Thermo sher's online T m calculator).Finally, the full-length Nb DNA fragments were digested with S I (New England Biolabs) and cloned into phagemid pComb3Xss (NBbiolab, China).The recombinant vector was electro-transformed (Bio-Rad MicroPulser electroporator) into TG1 bacteria at 2.5kV (0.2cm cuvette) and ~5.2ms time constant, Pre-warmed SOC medium was added and incubated at 37℃ with shaking at 250 rpm for 1 h. 10 μl of the culture was 10-fold serially diluted and plated on 2×TY agar plates containing 1.5% glucose ( nal concentration, the same below) and 75mg/mL carbenicillin.The plates were incubated overnight at 37℃, and the diversity of the library was calculated next day by counting the colonies.

2.Nbs Screening from phage library
96-well plates (Corning, high binding surface) were coated with 100 μl of 100 μg/ml puri ed protein (RBD or BSA) for 2 hours at room temperature (RT), and blocked with PBS buffer containing 2% milk powder (w/v) for 1 hour at RT. Phages library were incubated with immobilized antigen for 1 hour and then washed with PBST (PBS buffer supplemented with 0.5% Tween 20).Bound phages were eluted with 100μl of 20μg/ml trypsin, and were used to infect TG1 bacteria culture (OD = 0.2 ~ 0.8) at 37°C for 45 min.The eluted phage library was ampli ed according to the protocol described in above section.The antigenspeci c-binding of phages library after each round of panning was assessed by polyclonal phage ELISA.Single-clone phage ELISA were also carried out using colonies on phage titration plates.

3.Enzyme-linked Immunosorbent Assay (ELISA)
The entire ELISA procedure was carried out at room temperature.96-well plates (Corning #3690) were coated with 100 μl of 100 μg/ml puri ed protein (RBD or BSA) for 2 hours, and blocked with PBS buffer containing 2% milk powder (w/v) for 1 hour.For polyclonal phage ELISA, phages from each round of panning were incubated with immobilized antigen and bound phages were detected with anti-M13-horseradish peroxidase (HRP) polyclonal antibody (Thermo sher, MA5-29950).For the puri ed antibody binding assay, serially diluted Nbs (with HA-tag) solutions were added and incubated for 1.5 h, and bound Nbs s were detected with monoclonal anti-HA-HRP antibody.The enzyme activity was measured with the subsequent addition of substrate EL-TMB and signal reading was carried out at 450 nm using a Microplate Spectrophotometer.

4.Protein expression and puri cation
The gene sequences of the Nbs were ampli ed with PCR and subcloned into a pET-21(a+) expression vector, which contains a C-terminal 6xHis+HA tag.The expression construct was transformed into a BL21(DE3) chemially competent E. coli for protein expression.
The overnight culture with the selected colony was inoculated in one Liter LB media with correct antibiotics.The temperature was decreased to 18℃when OD600 of culture reached 0.6, the recombinant Nbs protein expressing was induced overnight with 0.5mM IPTG.Bacterial was harvested and resuspended in lysis buffer (50mM PBS, 2mM PMSF, pH 7.4).Protein was puri ed with Ni column (HiTrap Excel, GE Healthcare) and gel ltration (Superdex S75 column, GE Healthcare).RBD (R319-F541) and human ACE2 ectodomain(S19-D615) protein were expressed with Bac-to-Bac Baculovirus Expression System (Invitrogen).The corresponding gene of two proteins were subcloned into a modi ed pFastBac1 vector (Invitrogen), which contains a N-termial GP67 secreting signal peptide sequence and a C-terminal 6xHis puri cation tag.The expressing construct was transformed into bacterial DH10bac competent cell, the recombinant bacmid was extracted and transfected into sf9 insect cell with Cellfectin II reagent (Invitrogen).After two-rounds ampli cation, the recombinant baculovirus with hightiter were harvested and mixed with Hi5 insect cell (2x10 6 cells per mL).After 60 hours infection, the cell culture containing the secreted proteins was harvested.Protein puri ed with Ni-column (HiTrap Excel, GE Healthcare) and gel ltration column (Superdex 200, GE Healthcare); PBS buffer was used for all puri cation steps.

5.Micro Scale Thermophoresis (MST)
The binding a nity between Nbs proteins and RBD was measured with the MST NT.115 device (NanoTemper Technologies).RBD was labeld with MonolithTM RED-NHS labeling kit with the manufacturer's protocol.The labeled RBD protein was diluted with binding buffer (PBS-T: 20mM PBS, 0.05% Tween-20, pH=7.4) before it was used in the experiment.A 20nM nal concentration of the labeled protein was mixed (1:1) with the sequentially diluted nanbobodies.For the competition assay, the same procedure was followed except the PBST-buffer was containing 5nM ACE2 ectodomain.All measurements were triplicated with 50% LED medium MST power.The Mo.A nity Analysis software (NanoTemper Technologies) was used for the data analysis.6.Nb39-RBD complex structure prediction with Snugdock [39] We performed nanobody-antigen docking with the SnugDock module in Rosetta 3. Before running SnugDock, we prepared models for Nb39 and RBD, respectively, as below.
For Nb39, an initial model was rst obtained from Alphafold2 [54] prediction (monomer mode with Nb39 sequence as input and default settings, see below) by selecting the model with the highest average pLDDT score.The PDB residues were renumbered according to the Chothia antibody-numbering scheme.Then the initial model's CDR3 conformation was further optimized by the Antibody_H3 module in Rosetta 3 with default settings.1000 conformations were generated in total and the 5 top-scored conformations were selected for docking.
For RBD, to account for possible conformational exibility, we prepared multiple models from three different sources: (i) the RBD structure extracted from the complex of RBD with the Fab fragments of two neutralizing antibodies (PDB 6xdg); (ii) a model predicted by Alphafold2 (monomer mode with RBD sequence as input and default settings, see below); (iii) a model extracted from the Nb39-RBD complex model predicted by Alphafold-multimer [55] (with both RBD and Nb39 sequences as input and default settings) by selecting the model with the highest average pLDDT score.
For nanobody-antigen docking, the Snugdock method [39] in Rosetta 3 was employed with ensemble docking to mimic conformer selection and induced t by performing simultaneous optimization of the antibody (nanobody)-antigen rigid-body positions and the CDR loops.The best complex conformation was selected according to the ranking score. Figures

Figure 3 Initial
Figure 3 Initial screening of anti-RBD Nbs.(A-B) Expression and puri cation of the recombinant RBD through baculovirus expression system and puri ed by Ni-NTA followed by gel ltration.C. The sequences of the CDR1 and CDR2 of the ve high-a nity RBD-binding Nbs.D. The results of phage ELISA of candidate phages.The error bars represent S.D. from three independent experiments.The statistical difference was measured by paired two-sided Student's t test.

Figure 4 The
Figure 4 The results of phage ELISA of candidate phages.The error bars represent S.D. from three independent experiments.The statistical difference was measured by paired two-sided Student's t test.

Figure 6 Candidate
Figure 6 Candidate Nb39 compete with ACE2 for RBD binding.(A) Puri ed ACE2 recombinant protein with >90% protein purity.Protein purity was quanti ed by Gel-Pro analyzer.(B) The competition of candidate Nbs with ACE2 for RBD binding by MST.(C) model of Nb39-RBD complex by SnugDock.The CDR1, CDR2, and CDR3 of Nb39 are shown in blue, green, and magenta, respectively, whereas the rest of Nb39 are in gray, and RBD in light orange.(D) The hydrogen bond between His32 (H32) of CDR1 and Glu166 (E166) of RBD.(E) The binding interface between CDR3 and RBD where key residues were shown in sticks.