key: cord-0885642-52r4mmbp
authors: Kim, Sang Il; Noh, Jinsung; Kim, Sujeong; Choi, Younggeun; Yoo, Duck Kyun; Lee, Yonghee; Lee, Hyunho; Jung, Jongtak; Kang, Chang Kyung; Song, Kyoung-Ho; Choe, Pyoeng Gyun; Kim, Hong Bin; Kim, Eu Suk; Kim, Nam-Joong; Seong, Moon-Woo; Park, Wan Beom; Oh, Myoung-don; Kwon, Sunghoon; Chung, Junho
title: Stereotypic Neutralizing VH Clonotypes Against SARS-CoV-2 RBD in COVID-19 Patients and the Healthy Population
date: 2020-07-02
journal: bioRxiv
DOI: 10.1101/2020.06.26.174557
sha: 8900c51157715cdcc78277666b01a94d71bf5bbb
doc_id: 885642
cord_uid: 52r4mmbp

In six of seven severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) patients, VH clonotypes, encoded by either immunoglobin heavy variable (IGHV)3-53 or IGHV3-66 and immunoglobin heavy joining (IGHJ)6, were identified in IgG1, IgA1, and IgA2 subtypes, with minimal mutations, and could be paired with diverse light chains, resulting in binding to the SARS-CoV-2 receptor-binding domain (RBD). Because most human antibodies against the RBD neutralized the virus by inhibiting host cell entry, we selected one of these clonotypes and demonstrated that it could potently inhibit viral replication. Interestingly, these VH clonotypes pre-existed in six of 10 healthy individuals, predominantly as IgM isotypes, which could explain the expeditious and stereotypic development of these clonotypes among SARS-CoV-2 patients.

yet been identified. Here, we report stereotypic nAbs for SARS-CoV-2, which were identified by mapping nAbs onto deep immunoglobulin repertoires that were profiled from infected patients. One of these stereotypic nAbs was perfectly naïve and was encoded by immunoglobin heavy variable (IGHV)3-53/IGHV3-66 and immunoglobin heavy joining (IGHJ)6. Furthermore, we also found that these exact VH clonotypes pre-exist in the majority of the healthy population, predominantly as an IgM isotype, which immediately provoked the hypothesis that individuals with this VH clonotype may be able to rapidly evolve potent nAbs and experience favorable clinical features, similar to the human immunodeficiency virus (HIV)-1 response observed among individuals who acquire a unique VH clonotype, featuring a very long heavy chain complementarity determining region (HCDR)3, following exposure to syphilis infection 14 .

To obtain monoclonal nAbs against SARS-CoV-2, we collected blood samples from seven SARS-CoV-2-infected patients (patients A-G) and used them to generate human antibody libraries. Similar to SARS-CoV, SARS-CoV-2 also uses a spike (S) protein for receptor binding and membrane fusion 15 . This protein interacts with the cellular receptor angiotensinconverting enzyme II (ACE2) to gain entry into the host cell 16, 17 . A previous report suggested that a human monoclonal antibody, which reacted with the receptor-binding domain (RBD), within the S1 region of the S protein, could hinder the initial interaction between the virus and the cell, effectively neutralizing SARS-CoV-2 13 . We confirmed the reactivity of the sera derived from patients against recombinant SARS-CoV-2 S and RBD proteins. Patients A and E, who presented with extensive pneumonic infiltrates, also showed high plasma IgG levels against all recombinant SARS-CoV-2 nucleocapsid (NP), S, S1, S2, and RBD proteins, which could be detected 11, 17, and 45 days after symptom onset in Patient A and 23, 44, and 99 days after symptom onset in Patient E (Supplementary Table 1 and Supplementary Fig. 1) . Notably, the sera samples from Middle East respiratory syndrome coronavirus (MERS-CoV) patients cross-reacted with the SARS-CoV-2 S protein, showing a higher titer against the S2 domain, and vice versa (Supplementary Fig. 1 and 2) , suggesting the potential risk for ADE. We generated four human antibody libraries, utilizing a phage display system, based on the blood samples from Patient A, which were collected on days 17 and 45 (A_d17 and A_d45), and Patient E, which were collected on days 23 and 44 (E_d23 and E_d44). After biopanning, we successfully isolated 38 single-chain variable fragment (scFv) clones that were reactive against recombinant SARS-CoV-2 RBD in an enzyme immunoassay (Supplementary Fig. 3 and Supplementary Table 2 ). The half-maximal binding of these scFv-human kappa light chain fragment (hCκ) fusion proteins with the coated antigens occurred at concentrations ranging from 0.32 to 364 nM, which was compatible with the findings of previous reports that have described human monoclonal antibodies against SARS-CoV-2 RBD 10,13 . Then, we tested whether these antibody clones could inhibit the binding between recombinant SARS-CoV-2 S protein and Vero E6 cells expressing the We also performed deep profiling of the immunoglobulin (IG) repertoire in three chronological blood samples each from patients A and E and two chronological samples from each of the other five patients. Then, we searched for nAb clonotypes that possessed identical VJ combinations and perfectly matched HCDR3 sequences, at the amino acid level among the immunoglobulin heavy chain (IGH) repertoires of Patients A and E. One and five nAb clonotypes were successfully identified in Patients A and E, respectively (Fig. 1a) . Notably, three nAbs (A-2F1, E-3A12, and E-3B1) were encoded by IGHV3-53/IGHV3-66 and IGHJ6 (Fig. 1a) . These two VH genes, IGHV3-53*01 and IGHV3-66*01, are identical at the amino acid level, except for the H12 residue (isoleucine in IGHV3-53 and valine in IGHV3-66), and only five nucleotide differences exist between their sequences. Furthermore, four clonotypes were IgG1, and two clonotypes class-switched to IgA1 and IgA2 when examined 44 days after symptom onset (Fig. 1a) . These clonotypes had a very low frequency of somatic mutations (1.03% 0.51%), which was compatible with findings regarding other nAbs in previous reports 9, 10 . Then, we collected all VH sequences from the seven patients and searched the clonotypes of 11 nAbs that were encoded by the same VH and JH genes and showed 66.6% or higher identity in the HCDR3 sequence, at the amino acid level (Supplementary Fig. 6 ).

Interestingly, clonotypes that were highly homologous to the E-3B1 nAb were found among six of seven patients, with a total of 55 sequences among the isotypes IgG1 (Patient A, B, D, E, F, and G), IgA1 (Patient E and G), and IgA2 (Patient E) (Supplementary Table 3 ). These clonotypes shared nearly identical VH sequences (92.78% 1.40% identity at the amino acid level), with E-3B1 displaying an extremely low frequency of somatic mutations (0.77% 0.93%). Among these 55 clonotypes, 22 unique HCDR3s were identified, at the amino acid level, and eight unique HCDR3s existed in more than one patient. To test the reactivity of clonotypes homologous to E-3B1 against the SARS-CoV-2 S protein, we arbitrarily sampled 12 IGH clonotypes (Fig. 1b) , containing five different HCDR3s, from the IGH repertoires of six patients. The genes encoding these IGH clonotypes were chemically synthesized and used to construct scFv genes, using the Vλ gene from the E-3B1 clone. Then, the reactivities of these scFv clones were tested in an enzyme immunoassay. Three clones (E-12, A-32, and B-33) reacted against the recombinant S and RBD proteins (Fig. 1b) . Then, scFv libraries were constructed, using the A-11, A-31, E-34, A,B,G-42, G-44, D-51, F-53, E-52, and A-54 genes, and the Vκ/Vλ genes were amplified from Patients A, E, and G. Consequently, we confirmed that all 12 IGH clonotypes were reactive against both recombinant S and RBD proteins when paired with eight different Vκ and Vλ genes (Fig. 1b,c) . Moreover, all seven patients possessed these Vκ/Vλ clonotypes with identical VJ gene usage and perfectly matched LCDR3 amino acid sequences ( Supplementary Fig. 7 ). In particular, IGLV2-14/IGLJ3, IGLV3-19/IGLJ2, and IGLV3-21/IGLJ2 were frequently used across all seven patients (Supplementary Fig. 8 and 9) . Because E-3B1 effectively inhibited the replication of SARS-CoV-2 (Fig. 1d) , these 55 clonotypes are likely to neutralize the virus when paired with an optimal light chain.

Among these IGH clonotypes, A,B,G-42 was quite unique, presenting no somatic mutations and containing an HCDR3 (DLYYYGMDV) formed by the simple joining of IGHV3-53 and IGHJ6. This naïve VH sequence existed in the IGH repertoire of three patients (A, B, and G), as IgG1, IgG1, or IgG1 and IgA1 subtypes, respectively (Table 1) . More interestingly, the IGH clonotypes encoded by IGHV3-53/IGHV3-66 and IGHJ6 that possessed an HCDR3 (DLYYYGMDV) with zero to one somatic mutation residues could be identified within the IGH repertoire of six of 10 healthy individuals, predominantly as an IgM isotype (Table 1) , based on publicly available IGH repertoires 18 . The A,B,G-42 clonotype showed light chain plasticity and paired with five Vκ /Vλ genes to achieve RBD binding. In particular, the Vκ gene (2J6H) accumulated only five somatic mutations (1.4% divergence). None of the 12 clones, including A,B,G-42, reacted against the recombinant RBD proteins from either SARS-CoV or MERS-CoV ( Supplementary Fig. 10 ). In our prior experiment, none of the 37 identified MERS-RBD-binding human monoclonal antibodies, from two patients, were encoded by IGHV3-53/IGHV3-66 and IGHJ6 (Supplementary Table 4) 19 . Therefore, the presence of these stereotypic-naïve IGH clonotypes in the healthy population, and their light chain plasticity to achieve SARS-CoV-2 RBD binding, may be unique to SARS-CoV-2, which might provide a rapid and effective humoral response to the virus among patients who express these clonotypes. These findings provide the majority of the population possess germline-precursor B cells, encoded by IGHV3-53/IGHV3-66 and IGHJ6, which can actively initiate virus neutralization upon SARS-CoV-2 infection.

To further elucidate the preferential use of IGHV3-53/IGHV3-66 and IGHJ6 genes during the generation of SARS-CoV-2 RBD-binding antibodies, we extracted 252 predicted RBDbinding clones from our biopanning data (See Methods). We previously showed that antibody clones with binding properties can be predicted by employing next-generation sequencing (NGS) technology and analyzing the enrichment patterns of biopanned clones 20,21 . Although the IGHJ4 gene was more prominent in the IGH repertoires of the seven patients, similar to healthy human samples 18,22 , the predicted RBD-binding clones primarily used the IGHJ6 gene (Fig. 1e) . Furthermore, the predicted RBD-binding clones showed the dominant usage of IGHV3-53/IGHJ6 and IGHV3-66/IGHJ6 pairs, which was not observed in the whole IGH repertoires of patients (Fig. 1f) .

Naïve B cells typically undergo somatic hypermutations, clonal selection, and class-switching following antigen exposure. We examined the chronological events that occurred in all IGH clonotypes identified in patients and those that were reactive against the SARS-CoV-2 RBD.

We categorized RBD-reactive clones into three groups: neutralizing antibodies (neutralize), binding-confirmed antibodies (bind), and binding-predicted antibodies (predicted). In all three groups, these IGH clonotypes appeared and disappeared along the disease course and showed a low frequency of somatic mutations (Fig. 2c ,d) and rapid class-switching, especially to IgG1, IgA1, and IgA2. In the entire IGH repertoire of the patients, naïve-derived IGH clonotypes with minimal somatic mutations (< 2.695%  0.700%) showed increased IgG3 and IgG1 subtypes, and the proportion of IgG1 subtype was dramatically increased for a period (Fig. 2a,b and Supplementary Fig. 11 ). Furthermore, these naïve-derived IGH clonotypes were detected as IgA1 and IgG2 subtypes in patients A and E, as minor populations (Fig. 2a,b) , and as the IgA2 subtype in Patient E (Fig. 2b) . To summarize, RBDreactive IGH clonotypes rapidly emerged and underwent class-switching, to IgG1, IgA1, and IgA2, without experiencing many somatic mutations. However, this dramatic temporal surge of naïve IGH clonotypes, with rapid class-switching, occurred across the entire IGH repertoire of patients and was not confined to those reactive to the SARS-CoV-2 RBD.

Because several mutations within the RBD have been identified along the course of the SARS-CoV-2 pandemic, worldwide 23 , we examined the probability of emerging escape mutants from the IGH repertoire induced by the wild-type virus infection. Our E-3B1, A-1H4, A-2F1, A-2H4, and E-3G9 nAbs successfully bound to recombinant mutant RBD  proteins (V341I, F342L, N354D/D364Y, V367F, A435S, W436R, G476S, and V483A) in a   dose-dependent manner, with compatible reactivity against recombinant wild-type RBD protein ( Supplementary Fig. 12) . Therefore, the human IGH immune repertoire may provide effective protection against most current SARS-CoV-2 mutants.

In response to SARS-CoV-2 infection, most human IGH repertoires efficiently generated clonotypes encoded by IGHV3-53/IGHV3-66 and IGHJ6, which could be paired with diverse light chains, for both RBD binding and virus neutralization, with few to no somatic mutations. These clonotypes underwent swift class-switching to IgG1, IgA1, and even IgA2 subtypes. The expeditious development of these IGH clonotypes would be possible because the naïve-stereotypic IGHV3-53/IGHV3-66 and IGHJ6 clonotypes pre-exist in the majority of the healthy population, predominantly as an IgM isotype. In line with our findings, several groups have previously reported potent human nAbs, composed of either IGHV3-53 or IGHV3-66 and IGHJ6 genes, using single B cell sequencing technology 9-13 . Furthermore, the crystal structures of two IHGV3-53 neutralizing antibodies were determined which showed that two key motifs within HCDR1 and HCDR2 encoded in the IGHV3-53 germline are making contact with RBD 24 . Therefore, the preferential use of IGHV3-53/IGHV3-66 and IGHJ6 in the development of nAbs to SARS-CoV-2 appeared prominent. From these observations, we hypothesize that the existence of this unique, naïve IGH clonotype would provide near-immediate protection to some people exposed to SARS-CoV-2, and a very favorable clinical course, unlike SARS-CoV or MERS-CoV. In addition, the chronological follow-up of IGH clonotypes, encoded by the IGHV3-53/IGHV3-66 and IGHJ6 genes, along with their class-switching events, would be valuable for the development of a safe and effective vaccine.

clonotypes that are highly homologous to E-3B1, and the predicted RBD-binding clones that were enriched through biopanning. Stereotypic nAb VH clonotypes against the SARS-CoV-2 RBD, encoded by IGHV3-53/3-66 and IGHJ6, were found in six of seven patients. a, Characteristics of nAbs discovered in patients A and E. b, IGH clonotypes that are highly homologous to E-3B1 and reactive against recombinant SARS-CoV-2 S and RBD proteins. The right column shows the results of the phage ELISA. All experiments were performed in quadruplicate, and the data are presented as the mean  SD. c, List of diverse IGL clonotypes that can be paired with the IGH clonotypes from b to achieve reactivity. d, Measurement of viral RNA in the culture supernatant of Vero cells after SARS-CoV-2 infection e, J and f, VJ gene usage in the IGH repertoire of patients (upper) and the binding-predicted IGH clones (bottom). For the VJ gene usage heatmap, the frequency values for the IGH repertoire of all seven patients were averaged and are displayed (upper) along with those of the predicted RBDbinding IGH clones (bottom). N/A: not applicable RBD-reactive IGH clonotypes rapidly undergo class-switching events to IgG1, IgA1, and IgA2, with few somatic mutations. (a,b) IGH repertoires of a, Patient A and b, Patient E were analyzed 11, 17, and 45 (A_d11, A_d17, A_d45) days and 23, 44, and 99 (E_d23, E_d44, E_d99) days after symptom onset, respectively. IGH repertoires were examined according to divergence from the germline and the isotype composition of the sequences. Values for divergence from the germline were calculated separately for each isotype and are presented as violin plots, ordered by the classswitch event. The bar graphs on the top of the violin plots represent the proportion of each isotype in the repertoire. (c,d) Mapping of three types of RBD-binding IGH sequences (neutralize, bind, and predicted), derived from either c, Patient A or d, Patient E, against the corresponding IGH repertoire. The positions of the RBD-binding IGH sequences in the divergence value were annotated as dot plots, on the same scale used for a and b. Bar graphs on the top of the dot plots indicate the isotype compositions of the sequences in the repertoire.

The healthy samples based on publicly available IGH repertoires or patient identification can be found in the sample column. Clonotypes were mapped according to identical VJ gene usage of IGHV3-53/IGHV3-66 and IGHJ6 and perfectly matched HCDR3 at the amino acid level. The read counts of the mapped sequences in the repertories of each samples were annotated in the occurrence column. For the clonotypes with multiple occurrences, mean and standard deviation of divergence were represented. The proportion of each isotypes were indicated for the all samples. 

Pre-processing of the NGS data for the IG repertoire. The raw NGS forward (R1) and reverse (R2) reads were merged by PEAR, v0.9.10, in default setting 27 . The merged reads were q-filtered using the condition q20p95, which results in 95% of the base-pairs in a read having Phread scores higher than 20. The location of the primers was recognized from the qfiltered reads while allowing one substitution or deletion (Supplementary 28, 29 . From the aligned reads, the frequency of each nucleotide was calculated, and a consensus sequence of each sub-cluster was defined using the frequency information.

Then, the read count of the consensus sequence was re-defined as the number of UMI subclusters that belong to the consensus sequences.

Sequence annotation, functionality filtering, and throughput adjustment. Sequence annotation consisted of two parts, isotype annotation and VDJ annotation. For annotation, the consensus sequence was divided into two sections, a VDJ region and a constant region, in a location-based manner. For isotype annotation, the extracted constant region was aligned with the IMGT (international immunogenetics information system) constant gene database 30 .

Based on the alignment results, the isotypes of the consensus sequences were annotated.

Then, the VDJ regions of the consensus sequences were annotated, using IgBLAST, v1.8.0 31 . Among the annotation results, V/D/J genes (V/J genes for VL), CDR1/2/3 sequences, and the number of mutations from the corresponding V genes were extracted, for further analysis.

Divergence values were defined as the number of mutations identified in the aligned V gene, divided by the aligned length. Then, the non-functional consensus reads were defined using the following criteria and filtered-out: 1. sequence length shorter than 250 bp; 2. existence of stop-codon or frame-shift in the full amino acid sequence; 3. annotation failure in one or more of the CDR1/2/3 regions; and 4. isotype annotation failure. Then, the functional consensus reads were random-sampled, to adjust the throughput of the VH data (Supplementary Table 5 ). Throughput adjustment was not conducted for VL data (Supplementary Table 6 ).

Pre-processing of the biopanning NGS data. Pre-processing of the biopanning NGS data was performed as previously reported, except for the application of the q-filtering condition q20p95 instead of q20p100 32 .

Overlapping IGH repertoire construction. To investigate the shared IGH sequences among the patients, we defined the overlapping IGH repertoire of the patients. First, histograms for the nearest-neighbor distances of the HCDR3 amino acid sequences were calculated for the repertoire data. A hierarchical, distance-based analysis, which was reported previously 33 , was applied to the HCDR3 amino acid sequences, to cluster the IGH sequences at a functional level. The IGH sequences for all repertoire data could be approximated into a bimodal distribution, allowing the functionally similar IGH sequences to be extracted by capturing the first peak of the distribution ( Supplementary Fig. 13 ). Threshold values for each data set were defined as the nearest-neighbor distance value of those points with a minimum frequency between the two peaks of the distribution. Then, the minimum value among all threshold values, 0.113871, was used to construct the overlapping IGH repertoire, which means that 11.3871% of mismatches in the CDR3 amino acid sequence were allowed in the overlapping IGH repertoire construction. To construct the overlapping IGH repertoire, the repertoire data sets of all patients were merged into one data set. The IGH sequences in the merged data set were then clustered, using the following conditions: 1. the same V and J gene usage; and 2. mismatch smaller than 11.3871% among the CDR3 amino acid sequences.

Subsequently, clusters containing IGH sequences from more than one patient were included in the overlapping IGH repertoire data set.

Extraction of binding-predicted clones. From each round of biopanning (rounds 0, 2, 3, and 4), the VH genes were amplified and subjected to NGS analysis, using the MiSeq platform, as described previously 21 . Binding-predicted clones from biopanning were defined by employing frequency the values of the NGS data from four libraries, A_d17, A_d45, E_d23, and E_d44, at each round of biopanning. The enrichment of clones primarily occurred during the second round of biopanning, based on the input/output virus titer values for each round of biopanning and the frequencies of the clones in the NGS data ( Supplementary Fig. 14) . Then, the frequency information in the NGS data sets for biopanning rounds 0, 2, 3, and 4 was subject to principal component analysis (PCA), for dimension reduction. Accordingly, principal component (PC)1 and PC2, which represented clone enrichment and clone depletion, respectively, were extracted. In the biopanning data, PC1 was primarily composed of the frequencies in rounds 2, 3, and 4, whereas PC2 was primarily composed of the frequency in round 0 ( Supplementary Fig. 15 ). Thus, we defined PC1-major clones as the predicted clones, by setting constant threshold values on the PC1 value and the ratio between PC1 and PC2 (Supplementary Table 7 ). Subsequently, 94.74% of the RBD-binding clones were successfully mapped to the predicted clones ( Supplementary Fig. 15 ).

For the VH gene, the cDNA prepared for the NGS analysis was used. For the VΚ and Vλ genes, total RNA was used to synthesize cDNA, using the Superscript IV First-Strand Synthesis System (Invitrogen), with oligo(dT) primers, according to the manufacturer's instructions. Then, the genes encoding VΚ/Vλ and VH were amplified, from the oligo(dT)-synthesized cDNA and the cDNA prepared for NGS analysis, respectively, using the primers listed in Supplementary Data and materials availability: Raw sequencing data will be submitted shortly. Computer codes and processed data will be deposited on Github. All other data that supporting the findings of this study are available from the corresponding author on reasonable request. Nat Rev Immunol, doi:10.1038/s41577-020-0321-6 (2020). Plasma samples from seven SARS-CoV-2 patients were diluted (1:100) and added to plates coated with recombinant SARS-CoV-2 spike, S1, S2, or N proteins, fused to HIS tag, or RBD protein, fused to human Cκ domain. The amount of bound IgG was determined using antihuman IgG (Fc-specific) antibody. ABTS was used as the substrate. All experiments were performed in duplicate, and the data are presented as the mean  SD.

Recombinant SARS-CoV-2 RBD-coated microtiter plates were incubated with varying concentrations of scFv-hCκ fusion proteins. HRP-conjugated anti-human Ig kappa light chain antibody was used as the probe, and TMB was used as the substrate. All experiments were performed in duplicate, and the data are presented as the mean  SD.

The recombinant scFv-hFc fusion proteins (200 nM or 600 nM) were mixed and incubated with recombinant SARS-CoV-2 S glycoprotein (200 nM) fused with a HIS tag at the Cterminus. After incubation with Vero E6 (ACE2 + ) cells, the relative amount of bound, recombinant SARS-CoV-2 S glycoprotein was measured using a FITC-conjugated anti-HIS antibody. For each sample, 10,000 cells were monitored.

The recombinant scFv-hCκ fusion proteins were mixed with 2,500 TCID50 of SARS-CoV-2 (BetaCoV/Korea/SNU01/2020, accession number MT039890), and the mixture was added to the Vero cells. After 0, 24, 48, and 72 h of infection, the culture supernatant was collected to measure the viral titers. Supplementary Fig 6. Mapping of the 11 nAbs to the overlapping IGH repertoire.

a, The number class-switched IGH sequences in the overlapping repertoire, mapped to nAbs. The allowed number of CDR3 amino acid sequence substitutions during the mapping process is represented on the x-axis of the plot, after normalizing against the sequence length. The number of mapped sequences was normalized against the total number of IGH sequences in each patient, and their sum is represented in the y-axis of the plot. b, The number of patients expressing the overlapping class-switched IGH sequences, which were mapped to the nAbs. The x-axis is the same as described for a and the y-axis indicates the number of patients. Supplementary Fig 7. Existence of VL that can be paired with the stereotypic VH.

VL was mapped according to identical VJ gene usage and perfectly matched LCDR3 sequences at the amino acid level, which were identified in the IGL repertoires of all seven patients (A-G). The frequency values of the mapped sequences in the repertoires of all sampling points were summed. Patient identification can be found above each bar graph.

The frequency values of all sampling points were averaged and represented for each patient. Patient identification can be found at the top left corner of each heatmap.

The frequency values of all sampling points were averaged and are represented for each patient. Patient identification can be found at the left top corner of each heatmap.

Recombinant SARS-CoV-2, SARS-CoV, or MERS-CoV RBD protein-coated microtiter plates were incubated with phage clones. HRP-conjugated anti-M13 antibody was used as the probe, and ABTS was used as the substrate. All experiments were performed in quadruplicate, and the data are presented as the mean  SD. Values of divergence from the germline were calculated separately, for each isotype, and are presented as violin plots, class-switching event order. The bar graphs above the violin plots represent the proportions of each isotype.

Recombinant wild-type or mutant (V341I, F342L, N354D, V367F, R408I, A435S, G476S, and V483A) SARS-CoV-2 RBD protein-coated microtiter plates were incubated with varying concentrations of scFv-hFc fusion proteins. HRP-conjugated anti-human IgG antibody was used as the probe, and ABTS was used as the substrate. All experiments were performed in triplicate, and the data are presented as the mean  SD.

The frequency values of the histograms were approximated by the binned kernel estimation method, in the Gaussian kernel setting (black line). The threshold value for each patient was set as the x value of the points with a minimum frequency value between two peaks of the bimodal distribution (red vertical line). The x and y values of the threshold-setting point are indicated in the upper right corner of each histogram.

The x-and y-axes represent the frequency values for the NGS data in each biopanning round. The line on the scatter plots indicates the identity line (y = x). Input and output virus titer values are also presented, above the matched scatter plots.

Information regarding the PC weight vectors and the cumulative proportion of variance explained by the PCs are listed on the left side of the plots. PCA plots for PC1 and PC2 on shown on the right side of the plots. The binding-predicted clones were defined based on the value of PC1 and the ratio between PC1 and PC2, by setting a constant threshold value for each. The population of clones defined as predicted clones is marked in pink. The clones known bind to SARS-CoV-2 RBD are marked in red. P-246  SSSYYWG SIYYSGSTYYNPSLKS  RKWLRGAFDI  IGHV4-39  IGHJ3 0  E  G1  P-251  YYWIG  IIYPGDSDTRYSPSFQG  RSTTVGWLDY  IGHV5-51  IGHJ4 0.008734  E  G1  P-252  SYWMS  NIKQDGSEKYYVDSVKG  RVYYYGWLDV  IGHV3-7  IGHJ6 0.001456  A  G3|G1  P-261  SYGIH  LISYDGSDKYYADPVKG  SSWLRGAFDY  IGHV3-30  IGHJ4 0.038961  A  G1  P-268  SYYMH  IINPSGGSTSYAQKFQG  SSWYKLGFDP  IGHV1-46  IGHJ5 0  E  G1  P-269  SSSYYWG SIYYSGSTYYNPSLKS  TPWLRGAFDY  IGHV4-39  IGHJ4 0.001082  E  G3|G1|A1  P-275  SYEMN  YISSSGSTIYYADSVKG  TQWLRGAFDI  IGHV3-48  IGHJ3 0  A  G1  P-315  VNYMT  LIYSGGSTYYADSVKG  VLPYGDYADF  IGHV3- 

Polyclonal and convergent antibody response to Ebola virus vaccine rVSV-ZEBOV

Multi-Donor Longitudinal Antibody Repertoire Sequencing Reveals the Existence of Public Antibody Clonotypes in HIV-1 Infection