Potent NTD-Targeting Neutralizing Antibodies against SARS-CoV-2 Selected from a Synthetic Immune System

The majority of neutralizing antibodies (NAbs) against SARS-CoV-2 recognize the receptor-binding domain (RBD) of the spike (S) protein. As an escaping strategy, the RBD of the virus is highly variable, evolving mutations to thwart a natural immune response or vaccination. Targeting non-RBD regions of the S protein thus provides a viable alternative to generating potential, robust NAbs. Using a pre-pandemic combinatorial antibody library of 1011, through an alternate negative and positive screening strategy, 11 non-RBD-targeting antibodies are identified. Amongst one NAb that binds specifically to the N-terminal domain of the S protein, SA3, shows mutually non-exclusive binding of the angiotensin-converting enzyme 2 receptor with the S protein. SA3 appears to be insensitive to the conformational change and to interact with both the “open” and “closed” configurations of the trimeric S protein. SA3 shows compatible neutralization as S-E6, an RBD-targeting NAb, against the wild type and variant of concern (VOC) B.1.351 (Beta) of the SARS-CoV-2 pseudo virus. More importantly, the combination of SA3 with S-E6 is synergistic and recovers from the 10-fold loss in neutralization efficacy against the VOC B.1.351 pseudo virus.


Introduction
SARS-CoV-2 has infected over 700 million people and caused over 6 million deaths as of March 2023 (https://covid19.who.int/ (accessed on 8 March 2023)), which poses a big challenge to the global public health system [1]. In the global fight against coronaviruses, neutralizing monoclonal antibodies provide not only potential treatment but also valuable epitopes for vaccine development. Most antibody studies focused on the coronavirus spike (S) protein, a key structural viral protein for its recognition and membrane fusion to host cells [2]. This is critical to the coronavirus in determining host infectivity and transmission capacity. Generally, the S protein is divided into two subunits: the S1 subunit responsible for receptor binding and the S2 subunit responsible for cell membrane fusion [3]. The S1 subunit consists of a receptor-binding domain (RBD) that binds to the receptor protein of host cells, angiotensin-converting enzyme 2 (ACE2) [4], and a N-terminal domain (NTD) [2,5]. The gene fragment encoding RBD of the S protein is the most variable portion of the coronavirus genome, generating a series of variants of concern (VOC), i.e., Beta, Delta, and Omicron [6,7]. Monoclonal antibodies targeting the S protein with strong neutralizing extracellular domain of the SARS-CoV-2 spike sequence was modified to get the prefusion conformation as follows: the furin cleavage site (residues 682-685) was mutated to GSAS, KV (residues 986-987) was mutated to PP, and. a HRV 3C cleavage site was inserted before the 6-histidine tag or hFc tag [21]. Combinatorial antibodies in the scFv-Fc format or full-length human IgG1 construct in this same pFuse expression vector were expressed in HEK293F cells for 4 days and then purified by Mabselect chromatography according to the manufacturer's instructions. Purified recombinant antigens and antibodies were buffer exchanged into the PBS buffer (pH 7.4) using centrifugal concentrators (Amicon Ultra-4, Merck Millipore, #UFC803096, Tullagreen, Carrigtwohill, Ireland).

Phage Panning
SARS-CoV-2 spike non-RBD region-specific scFv antibodies were selected from a combinatorial human scFv antibody library (10 11 members) displayed on phages after three rounds of affinity enrichment. In each round, the phage particles were first incubated with biotinylated S-RBD-hFc immobilized on the Streptavidin (SA)-coated magnetic beads (#11206D, Invitrogen, Thermo Fisher Scientific Baltics UAB, Vilnius, Norway) for the negative selection. Phages that did not bind to S-RBD-hFc were collected and further incubated with biotinylated S-trimer-His immobilized on the SA-coated magnetic beads for positive selection. Antigen-bound phages were eluted with Glycine-HCl (pH 2.2) after each round of screening. XL1-Blue cells were used to amplify the output phages for the next round of panning. After three iterations, colonies were picked and analyzed by phage ELISA. All positive clones were sequenced using Sanger sequencing.

ELISA
ELISA procedures were described previously [19]. Phage ELISA was used to select positive clones after panning. Briefly, antigens at a concentration of 2 ng µL −1 were coated on 384-well-plates (20 µL per well) at 4 • C overnight. After blocking and PBST washing, the antigen coated plates were incubated with 50 µL XL1-Blue culture supernatants containing phages in each well at room temperature (RT) for 1 h. Anti-M13 bacteriophage antibody conjugated with HRP (20 µL per well, dilution factor 1:5000; #11973-MM05T-H, Sino Biological, Beijing, China) and the ABTS solution (20 µL per well, Roche, #11684302001) were used as the secondary antibody and the detection substrate, respectively, following PBST washes. Absorbance at 405 nm was measured (Enspire, PerkinElmer, Singapore).
To determine the binding properties of antibodies to antigens, briefly, the antibodies or antigens were coated on 96-well plates (2 ng µL −1 in 100 µL of PBS buffer per well). Biotinylated antigen or antibody solutions with 1:5 serial dilution in PBS buffer from 100 to 0.0064 nM were incubated in the coated wells (100 µL per well) at RT for 1 h. Streptavidin conjugated with HRP (ThermoFisher Scientific, #21130) was used as the secondary detector. The coloration of the substrate and absorbance detection are the same as described above.
For selected antibodies to the S protein and the S-RBD, the recombinant antibodies were coated. Biotinylated S-trimer-His and S-RBD-His solutions were added to the precoated wells.
For the competition assay, SA3-hFc or S-E6-IgG4 were coated. 2 nM biotinylated S-trimer-His was pre-incubated with of various concentrations ACE2-mFc or SA3-hFc at RT for 1 h. Then the mixture (100 µL) was added to the pre-coated wells and incubated for another 1 h at RT.
For comparison of the "open" and "closed" configurations of the S protein, SA3-hFc was coated. Biotinylated S-trimer-His with or without 50 µM Linoleic Acid (LA) pre-incubated were incubated in the pre-coated wells.

Affinity Determination by Biolayer Interferometry (BLI)
Binding affinities were performed by BLI on the Octet RED96 (Molecular Devices LLC, San Jose, CA, USA). Biotinylated SA3-hFc was loaded onto a SA biosensor (#18-5019, Sartorius, ForteBio, PALL Life Sciences, Shanghai, China) at 5 µg mL −1 in PBST-B buffer (PBS containing 0.02% Tween-20 and 0.05% BSA). The SA-bio-SA3 sensor was dipped into PBST-B for 60 sec to establish a baseline, then incubated in antigen solutions of various concentrations (1:2 serial dilution from 200 to 6.25 nM) to record the progression curves of association, followed by dissociation progression in a PBST-B buffer. The S-trimer-His antigens with and without 50 µM LA pre-incubation was compared in the conformation sensitive detection.
Sensor regeneration and equilibration, R max , k on , k off and K D fitting were carried out as previously described.

Size-Exclusion-High-Performance Liquid Chromatography
20 µL antibody solution in a PBS buffer (pH 7.4, 0.5 µg µL −1 ) was applied to an Agilent Bio SEC-5, 500A HPLC system using PBS buffer (pH 7.4) as the mobile phase at a flow rate of 0.5 mL min −1 . The percentage of aggregation and degradant compositions of a testing antibody was monitored by absorbance change at wavelength 280 nm at the corresponding retention times.

Pseudovirus-Based Neutralization Assay
Pseudo virus (PSV)-based neutralization assay was described previously [19]. Briefly, PSV of SARS-CoV-2 wild-type or mutant spike ∆19 (19 amino acids truncated at the C-terminus) with mCherry and Luciferase were first produced in HEK293T cells according to the reference. Then, HEK293T cells expressing hACE2 were seeded into a 96-well, white-opaque plate at a density of 1 × 10 4 per well. Testing antibodies serially diluted in DMEM with 10% FBS (dilution factor: 3.16, from 200 nm to 6.3 fM) were incubated with an equal volume of PSV at 37 • C for 30 min. No PSV or PSV in the absence of antibodies were set as controls for normalization. 100 µL of the mixtures were transferred to the HEK293T/hACE2 cells. Fresh media was changed 16 h after treatment for an additional culture at 48 h. PSV transduction was evaluated by luciferase activity using the Bright-Lumi Firefly Luciferase Reporter Gene Assay Kit (Beyotime, #RG015M) according to the manufacturer's instructions. Data fitting was carried out in GraphPad Prism 8.3. The combination index (CI) in the PSV neutralization assay was calculated by the CompuSyn program to evaluate the synergism according to the program instruction. CI > 1 indicates antagonism, CI = 1 indicates additive effect, and CI < 1 indicates synergism.

Chemical Crosslinking and Mass Spectrometry
S-trimer-His and SA3-hFc were crosslinked using collision-induced dissociation (CID)cleavable cross-linker, disuccinimidyl sulfoxide (DSSO) following the described procedure [22]. Briefly, S-trimer-His and SA3-hFc were mixed in PBS and incubated for 30 min at RT. Crosslinking was performed for 30 min at RT after adding DSSO (ThermoFisher Scientific, #A33545) to the mixture with 100-fold molar excess and quenching it with excess Tris (1 M, pH 8.0) for 10 min at RT. Then the crosslinked products were digested with chymotrypsin. The LC-MS n data of digested peptides were collected on the Orbitrap Fusion Tribrid (ThermoFisher Scientific) with an on-line NanoLC system and analyzed using the CID-MS 2 -MS 3 strategy as previously described [23].

Quantification and Statistical Analysis
All values in the text and figures are presented as the mean ± SEM of independent experiments with given sizes (n). Graphs were compiled, and statistical analyses were performed with Prism software (GraphPad). Statistical significance was evaluated with the two-tailed unpaired t-test when comparing two groups and with the one-way analysis of variance (ANOVA) when comparing more than two samples. Other statistical details are indicated in the figures and legends.

Selection of Antibodies Targeting the Non-RBD Regions of SARS-CoV-2 Spike Protein
The biotinylated recombinant SARS-CoV-2 spike RBD fused with human Fc (S-RBD-hFc) and the full-length spike extracellular domain with T4F motif and C-terminal Histag (S-trimer-His) immobilized on streptavidin (SA)-coated magnetic beads were panned against a pre-pandemic combinatorial scFv antibody phage library containing 10 11 members generated from peripheral blood mononuclear cells of 50 healthy donors in 1999 [20]. In order to select non-RBD-binding antibodies of interest, an alternate negative/positive panning strategy was developed for a total of three rounds of panning (Figure 1a). The binding signal ratios of S-trimer-His to S-RBD-hFc phagemids dramatically rose to 3-and 16-folds in the second and third rounds of the process (Figure 1b). Positive phages from round 3 were examined using phage ELISA against S-RBD-hFc and S-trimer-His. Eleven unique clones were identified by Sanger sequencing and re-confirmed to bind S-trimer-His but not S-RBD-hFc (Figure 1c,d). The 11 enriched positive clonal sequences were shown to harbor multiple somatic hypermutations (SHMs) and derive from multiple different germlines (Table S1). Three clones, SA3, SA4, and SB4, derived from the same germlines (IGHV1-18 and IGKV1-39) showed 11 identical SHMs in the heavy chain and various SHMs (0, 4, 6, respectively) in the light chain. Purified SA3, SA4, and SB4 fused with mFc were characterized and validated to bind S-Trimer-His and spike proteins expressed on the cell surface (Supplementary Material Figure S1).

Kinetic Characterization and Mode of Interaction for Non-RBD Targeting Antibody SA3
SA3 was first converted into the formats of scFv-linked human Fc-tag (SA3-hFc) and full-length IgG1 (SA3-IgG1). Both formats had good thermostability at 4 • C or room temperature (RT) over a period of 7 days' incubation ( Figure S2). Similar to that of SA3-mFc, the hFc and IgG1 forms of SA3 were shown to bind to the S-trimer-His with an apparent EC 50 value of 0.15 ± 0.02 ( Figure 2a) and 4.4 ± 0.5 nM (Figure 2b), respectively, but not the RBD domain. SA3-hFc compared to the full length SA3-IgG1 showed at least one order of magnitude more potent binding and was used in the following experiments.
SA3-hFc appeared to display a "fast-on, slow-off" kinetics to S-trimer-His with an apparent dissociation constant (K D ) value of 0.15 ± 0.01 nM determined by biolayer interferometry (BLI) using a 1:1 fitting model (Figure 2c). Similarly, BLI failed to detect any interaction between SA3-hFc and S-RBD-His.
The binding of SA3-hFc to S-trimer-His in the presence or absence of Linoleic Acid (LA) which was previously shown to bind the RBD of SARS-CoV-2 S protein tightly (K D = 41.1 ± 16 nM) and lock the S protein in the "closed" conformation to reduce interaction with ACE2 in vitro [24] was carried out by BLI and showed no measurable difference in binding affinities (Figure 2c). This implies strongly a distinct binding feature of SA3 that is independent of the interaction between ACE2 and S protein. Consistent with this notion, in a competitive binding experiment of SA3-hFc, ACE2-mFc, and S-trimer-His, SA3-hFc and ACE2-mFc showed mutually non-exclusive binding to S-trimer-His (Figure 2d).
We ectopically expressed spike proteins of all seven coronaviruses that infect humans [5,25] including SARS-CoV-2, SARS-CoV, HCoV-229E, HCoV-OC43, HCoV-NL63, HCoV-HKU1, and MERS-CoV on the surface of HEK293T cells and tested the crossreactivity of SA3 to them. FACS analyses showed that SA3 specifically bound to the S protein of SARS-CoV-2 but not the other six coronaviruses (Figure 2e). The numbers besides the clone names in the right panel represent the branch lengths, indicating the evolutionary distances between two consecutive nodes in the phylogenetic tree.

Kinetic Characterization and Mode of Interaction for Non-RBD Targeting Antibody SA3
SA3 was first converted into the formats of scFv-linked human Fc-tag (SA3-hFc) and full-length IgG1 (SA3-IgG1). Both formats had good thermostability at 4 °C or room temperature (RT) over a period of 7 days' incubation ( Figure S2). Similar to that of SA3-mFc, the hFc and IgG1 forms of SA3 were shown to bind to the S-trimer-His with an apparent EC50 value of 0.15 ± 0.02 (Figure 2a) and 4.4 ± 0.5 nM (Figure 2b), respectively, but not the  tion, in a competitive binding experiment of SA3-hFc, ACE2-mFc, and S-trimer-His, SA3-hFc and ACE2-mFc showed mutually non-exclusive binding to S-trimer-His (Figure 2d).

Epitope Mapping of SA3
To dissect the mode of interaction between the SARS-CoV-2 S protein and SA3, we sought to determine the binding epitopes of SA3 on the S protein using high-resolution mass spectrometry (MS) and mutagenesis. The S-trimer-His and SA3-hFc were crosslinked and digested, followed by "shot-gun" tandem MS analyses. Integrative analyses of CID-induced cleavage of interlinked peptides in MS/MS 2 and MS 3 of single peptide chain fragment ions revealed high crosslinking scores (Figure 3a) (Figure 3b and Figure S3-S6). These peptides locate in the RBD and NTD of the S protein, where the light chain of SA3 appears to interact with both the NTD and RBD, and the heavy chain interacts with the RBD (Figure 3a-c). To further validate the NTD-targeting of SA3, the binding of SA3-hFc to S-NTD-His was examined by ELISA, which showed an apparent EC 50 value of 16 ± 2 nM (Figure 3d).  To identify the critical epitope residues involved in SA3 binding, we performed FACSbased alanine substitutions at the peptide regions of Spike 185-189 , Spike 204-208 , Spike 415-419 , and Spike 456-460 . Comparing with the NTD-targeting antibody 4A8 [16], Spike 185-189 on the N4 loop of the S protein was found to be the main interacting site for SA3-hFc (Figure 3e). Each individual residue on Spike 185-189 was next mutated to alanine to identify the critical residues. As shown in Figure 3f, point mutation in F186A abolished the binding of SA3-hFc to S proteins, while K187A and L189A showed partially disrupted binding of SA3-hFc. The N4 loop at the NTD of the S protein, therefore, most likely consists of the epitope residues for SA3.

Mutually Non-Exclusive Binding and Synergistic Neutralization of SA3 and S-E6 against SARS-CoV-2
The mutually non-exclusive binding of SA3-hFc vs. ACE2-mFc to the S protein ( Figure 2d) observed above suggests SA3 is most likely orthogonal to an RBD-binding antibody with the S protein. S-E6 was previously identified as an RBD-targeting NAb through competitive blockage of the ACE2 attachment and binding of viral spikes [19]. The neutralization activities of the NTD-targeting antibody SA3-hFc against wild-type (WT) and variant SARS-CoV-2 spikes were evaluated in a PSV infection assay and compared to S-E6-IgG4, which showed compatible potent neutralization against SARS-CoV-2 WT and variants in a dose-dependent manner. The apparent NT 50 values of SA3-hFc against WT, Beta, and Delta were determined to be 0.063 ± 0.011, 0.23 ± 0.05, and 1.5 ± 0.2 nM, respectively (Figure 4a). For Omicron (BA.1), a significant decrease in SA3 neutralization potency occurred with the dramatic remodeling of the NTD surface of the Omicron [26]. However, it still maintained over 90% inhibition of efficacy at high concentrations (1 µM). To validate whether the binding of S-E6 and SA3 to the S protein is orthogonal, competitive binding to S-trimer-His was tested, which showed mutually non-exclusive binding (Figure 4b). PSV assay was next applied in the neutralization of Beta variant of SARS-CoV-2. As expected, combination of S-E6 and SA3 showed synergistic neutralization activity with paired antibodies displaying more potent neutralization than each alone (Figure 4c). The combination index (CI) value of the 1:1 orthogonal antibody mixture at NT50 was determined to be 0.53 indicating a nearly 50% enhancement in viral neutralization.

Discussion
Epitope mapping studies on SARS-CoV-2 NAbs revealed several sites on the S protein that appeared with high frequency in host immune responses [9,15,27,28]. The majority of the highly immunogenic sites are located in the RBD of the S protein since the RBD is highly exposed on the viral membrane and attachment site of the virus to the host cell To validate whether the binding of S-E6 and SA3 to the S protein is orthogonal, competitive binding to S-trimer-His was tested, which showed mutually non-exclusive binding (Figure 4b). PSV assay was next applied in the neutralization of Beta variant of SARS-CoV-2. As expected, combination of S-E6 and SA3 showed synergistic neutralization activity with paired antibodies displaying more potent neutralization than each alone (Figure 4c). The combination index (CI) value of the 1:1 orthogonal antibody mixture at NT 50 was determined to be 0.53 indicating a nearly 50% enhancement in viral neutralization.

Discussion
Epitope mapping studies on SARS-CoV-2 NAbs revealed several sites on the S protein that appeared with high frequency in host immune responses [9,15,27,28]. The majority of the highly immunogenic sites are located in the RBD of the S protein since the RBD is highly exposed on the viral membrane and attachment site of the virus to the host cell [9][10][11]15]. However, as a defense strategy of viruses, RBD is highly susceptible to mutations through genetic drift and host immune adaptation. Here we present an approach beyond human immune responses in vivo to select antibodies targeting the non-RBD sites on the S protein using the CAL. The natural CAL is a synthetic immune system that consists of not only an infinite diversity of antibody sequences but also "fossil records" of the immune responses of each individual donor within the library [17][18][19]22,29]. In our previous study of SARS-CoV-2, multiple potent RBD-targeting antibodies with strong neutralizing activities were discovered from the CAL constructed before the COVID-19 pandemic [19]. Further in the current study, we demonstrated that epitopes of weak immunogenicity in the natural immune system could be sensitized and enhanced to elicit strong immune responses in a synthetic immune system.
The negative screening step helped to eliminate interference from RBD-binding antibodies. A total of 11 non-RBD targeting antibodies were identified, representing four different germlines for heavy chains (IGHV5-51, 3-30, 1-18, and 1-69). Amongst SA3, specific binding to the N4 loop in NTD outside the supersite was observed [15,16,28] with high affinity binding to spike (K D = 0.15 ± 0.01 nM) and strong neutralizing activity. Six convalescent non-RBD targeting antibodies from the CoV-AbDab database that share the same germline of heavy and light chains (IGHV1-18/IGKV1-39) with SA3 showed no neutralization efficacy against SARS-CoV-2 [30][31][32][33] (Figure S7). Detailed sequence alignment indicates that SA3 contains 11 SHMs, compared to much lower levels of SHMs in those from convalescent patients. In acute infections (during the first week of COVID-19 infection), most antibodies undergo minimal SHM with limited clonal expansion [34]. The previous observation of patterns of convergence in potent NAb lineages across different donors [35] and the lack of neutralization activity for the convalescent non-RBD targeting antibodies strongly implicate the necessity of repeated immune stimulation for weak immunogenic epitopes. This illustrates well the value of CAL, with its "fossil record" of providing antibodies to weak immunogens.
Comparing to the published NTD-targeting neutralizing antibodies isolated from convalescent patients with typical IC 50 values inhibiting SARS-CoV-2 infection between 2 and 700 ng/mL (about 10 nM to 5 µM) [12,15,30,36,37], SA3 screened from the in vitro synthetic immune system showed more potent neutralizing activity with an IC 50 value of 0.063 ± 0.011 nM against SARS-CoV-2 pseudo virus. SA3 and the RBD-targeting antibody S-E6 were also shown to bind to the S protein in a mutually non-exclusive manner, which makes them an ideal orthogonal pair. Indeed, SA3 and S-E6 showed a remarkable additive effect in the pseudoviral neutralization assay of the SARS-CoV-2 VOC B.1.351 variant. The selection strategy of orthogonal antibodies with weak immunogenicity presented here facilitates the development of new therapeutic antibodies, detection methods, and preventative vaccines to meet the challenge of present and future pandemics. For instance, in view of the reductions in response to SARS-CoV-2 vaccines in patients with autoimmune diseases such as systemic sclerosis and immunosuppressant treatment, more vaccine doses or vaccination prior to planned immunosuppression is recommended [38]. Therefore, additional therapeutic antibody treatment, especially for those with less immunogenic epitopes, might be beneficial to the populations to compensate for the deficiency.