Longitudinal Isolation of Potent Near-Germline SARS-CoV-2-Neutralizing Antibodies from COVID-19 Patients

Summary

The SARS-CoV-2 pandemic has unprecedented implications for public health, social life, and the world economy. Because approved drugs and vaccines are limited or not available, new options for COVID-19 treatment and prevention are in high demand. To identify SARS-CoV-2-neutralizing antibodies, we analyzed the antibody response of 12 COVID-19 patients from 8 to 69 days after diagnosis. By screening 4,313 SARS-CoV-2-reactive B cells, we isolated 255 antibodies from different time points as early as 8 days after diagnosis. Of these, 28 potently neutralized authentic SARS-CoV-2 with IC₁₀₀ as low as 0.04 μg/mL, showing a broad spectrum of variable (V) genes and low levels of somatic mutations. Interestingly, potential precursor sequences were identified in naive B cell repertoires from 48 healthy individuals who were sampled before the COVID-19 pandemic. Our results demonstrate that SARS-CoV-2-neutralizing antibodies are readily generated from a diverse pool of precursors, fostering hope for rapid induction of a protective immune response upon vaccination.

Graphical Abstract

Highlights

• •Isolation of highly potent SARS-CoV-2-neutralizing antibodies
• •Longitudinal sampling reveals early class-switched neutralizing response
• •SARS-CoV-2 S-protein-reactive antibodies show little somatic mutation over time
• •Potential antibody precursor sequences identified in SARS-CoV-2-naive individuals

In a longitudinal analysis of SARS-CoV-2-infected people, Kreer et al. find highly potent neutralizing antibodies that use a broad spectrum of variable (V) genes and show low levels of somatic mutations. They also identify potential precursor sequences of these SARS-CoV-2-neutralizing antibodies from virus-naive individuals, sampled before the COVID-19 pandemic. This could indicate that neutralizing antibodies can be readily generated from existing germline antibody sequences found in the general population.

Introduction

By June 2020, over 8.4 million severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections and over 450,000 casualties of the associated coronavirus disease 2019 (COVID-19) were reported (Dong et al., 2020; Huang et al., 2020; Zhou et al., 2020; Zhu et al., 2020). The exponential spread of the virus has caused countries to shut down public life with unprecedented social and economic consequences. Therefore, decoding SARS-CoV-2 immunity to promote development of vaccines as well as potent antiviral drugs is an urgent health need (Sanders et al., 2020).

Monoclonal antibodies (mAbs) targeting viral surface proteins have been demonstrated to effectively neutralize viruses such as Ebola virus (EBOV) (Ehrhardt et al., 2019; Flyak et al., 2016; Saphire et al., 2018), respiratory syncytial virus (RSV) (Kwakkenbos et al., 2010), influenza virus (Corti et al., 2011; Joyce et al., 2016; Kallewaard et al., 2016), or human immunodeficiency virus 1 (HIV-1) (Walker et al., 2009; Huang et al., 2016a, 2016b; Scheid et al., 2011; Schommers et al., 2020; Wu et al., 2010). The most prominent target for an antibody-mediated response on the surface of SARS-CoV-2 virions is the homotrimeric spike (S) protein. The S protein promotes cell entry through interaction of its receptor-binding domain (RBD) with angiotensin-converting enzyme 2 (ACE2) (Hoffmann et al., 2020; Walls et al., 2020). Antibodies that target the S protein are therefore of particular interest to combat the current pandemic (Burton and Walker, 2020; Sempowski et al., 2020).

SARS-CoV-2 infection induces a humoral immune response of varying magnitude (Duan et al., 2020; Ni et al., 2020), and antibody levels depend on several factors, including disease severity (Long et al., 2020a; Wang et al., 2020b). SARS-CoV-2-reactive as well as SARS-CoV-2-neutralizing antibodies have now been isolated from COVID-19 survivors (Brouwer et al., 2020; Cao et al., 2020; Hansen et al., 2020; Ju et al., 2020; Liu et al., 2020a; Robbiani et al., 2020; Seydoux et al., 2020; Shi et al., 2020; Wu et al., 2020; Zost et al., 2020), immunized animals (Hansen et al., 2020; Wang et al., 2020a; Wrapp et al., 2020a), and phage display libraries (Li et al., 2020; Liu et al., 2020b; Yuan, 2020; Zeng et al., 2020). Such antibodies are of great value to elucidate neutralization mechanisms, inform vaccination strategies, and potentially treat and prevent SARS-CoV-2 infection, as demonstrated in animal models (Cao et al., 2020; Rogers et al., 2020; Zost et al., 2020). The antibodies described target various sites on the S protein, including the RBD (Brouwer et al., 2020; Liu et al., 2020a; Seydoux et al., 2020). However, the affinities and neutralization activities of the reported antibodies vary strongly, and the potential for SARS-CoV-2 escape mutations highlights the need to carefully develop antibody-mediated strategies (Baum et al., 2020). Moreover, little is known about the likelihood of generating such neutralizing antibodies and how they evolve over time, which will be critical for the development of a broadly active SARS-CoV-2 vaccine.

Here, we isolated and sequenced 4,313 S-protein-reactive memory B cells from 12 SARS-CoV-2-infected individuals as early as 8 days after diagnosis (16 days after onset of symptoms). Five patients were followed over a period of 8–69 days after diagnosis to investigate the dynamics of antibody development against SARS-CoV-2. Besides presenting 28 neutralizing antibodies that are currently being evaluated for clinical application, we provide evidence that antibodies develop early after SARS-CoV-2 infection with limited ongoing somatic hypermutation. Finally, we identified potential precursor sequences of potent SARS-CoV-2-neutralizing antibodies in naive B cell repertoires from healthy individuals who were sampled before the SARS-CoV-2 pandemic.

Results

SARS-CoV-2-infected individuals develop a polyclonal memory B cell response against the S protein

To investigate the antibody response against SARS-CoV-2, we collected blood samples from seven COVID-19 patients (38–59 years of age) between 8 and 36 days after diagnosis (Figure 1A; Table S1). Five patients presented with mild symptoms, including dry cough, fever, and dyspnea, whereas two patients were asymptomatic (Table S1). Purified plasma immunoglobulin G (IgG) of all seven individuals showed binding to the full trimeric S ectodomain (Wrapp et al., 2020b) by ELISA, and half-maximal effective concentrations (EC₅₀) ranged from 3.1–96.1 μg/mL (Figure 1B; Table S2). Moreover, neutralizing IgG activity was determined against authentic SARS-CoV-2, showing 100% inhibitory concentrations (IC₁₀₀) between 78.8 and 1,500 μg/mL in five of seven patients (Figure 1B; Table S2). To decipher SARS-CoV-2-specific B cells and antibody response on a molecular level, we performed single-B-cell sorting and sequence analysis of all individuals. Using flow cytometry, we detected that between 0.04% (±0.06) and 1.02% (±0.11) IgG⁺ B cells reacted with the S ectodomain (Figures 1C and S1). From these, we isolated a total of 1,751 single B cells and amplified IgG heavy and light chains using optimized PCR protocols (Figure 1C; Table S3) (Kreer et al., 2020a; Schommers et al., 2020). Sequence analysis revealed a polyclonal antibody response with 22%–45% clonally related sequences per individual and 2–29 members per identified B cell clone (Figure 1D; Table S3). We conclude that a polyclonal B cell response against the SARS-CoV-2 S protein was initiated in all studied COVID-19 patients.

Figure 1.

SARS-CoV-2 infection induces a polyclonal B cell and antibody response

(A) Scheme of cross-sectional sample collection (see also Table S1).

(B) Binding to the trimeric SARS-CoV-2 S ectodomain (ELISA, EC₅₀) and authentic SARS-CoV-2 neutralization activity (complete inhibition of VeroE6 cell infection, IC₁₀₀) of cross-sectional plasma-purified IgG samples. Bar plots show arithmetic or geometric means ± SD of duplicates or quadruplicates for EC₅₀ and IC₁₀₀, respectively. Abbreviation is as follows: n.n., no neutralization as defined by IC₁₀₀ > 1,500 g/mL IgG.

(C) Dot plots of IgG⁺ B cell analysis. Depicted numbers (percent ± SD) indicate average frequencies of S-reactive B cells (see also Tables S2, S3, and Figure S1).

(D) Clonal relationship of S ectodomain-reactive B cells. Individual clones are colored in shades of blue and green. Numbers of productive heavy-chain sequences are depicted in the center of the pie charts. Clone sizes are proportional to the total number of productive heavy chains per clone.

Figure S1.

Gating strategy for single-cell sorting, related to Figures 1 and 2

CD19⁺ B cells isolated by MACS were used and cell aggregates were excluded by FSC. Living CD20⁺ IgG⁺ cells were gated and cells with a positive SARS-CoV-2 S ectodomain staining were selected for single cell sort.

Longitudinal analysis of the SARS-CoV-2 antibody response

To delineate the dynamics of the SARS-CoV-2 antibody response, we obtained longitudinal blood samples from an additional five infected individuals at three time points spanning 8–69 days after diagnosis (Figure 2A; Table S1). Across the different individuals, EC₅₀ (S ectodomain binding) and IC₁₀₀ (SARS-CoV-2 neutralization) values of plasma IgG ranged from 1.54–129 μg/mL and 78.8–1,500 μg/mL, respectively (Figure 2B; Table S2). For each individual, however, this response remained almost unchanged over the studied period (Figures 2A and 2B).

Figure 2.

SARS-CoV-2-specific IgG⁺ B cells readily develop after infection with recurring B cell clones and a preference for the V_H gene segment 3-30

(A) Scheme of longitudinal sample collection. The viral RNA load from nasopharyngeal swabs is indicated in red (copies [cp] per milliliter, right y axis). ^∗The viral load for IDFnC1 is given as positive or negative result (see also Table S1).

(B) Binding to the trimeric SARS-CoV-2 S ectodomain (ELISA, EC₅₀) and authentic SARS-CoV-2 neutralization activity (complete inhibition of VeroE6 cell infection, IC₁₀₀) of longitudinal purified plasma IgG samples. n.n., no neutralization as defined by IC₁₀₀ > 1,500 µg/mL IgG. Bar plots show arithmetic or geometric means ± SD of duplicates or quadruplicates for EC₅₀ and IC₁₀₀, respectively.

(C) Percentage of SARS-CoV-2 S ectodomain-reactive IgG⁺ B cells over time (mean ± SD; see also Tables S2, S3, and Figure S1).

(D) Clonal relationship over time. Individual clones are colored in shades of blue and green. Numbers of productive heavy-chain sequences per time point are given in the center of pie charts.

(E) Frequencies of V_H gene segments (top), CDRH3 length and CDRH3 hydrophobicity (bottom left), as well as V_H gene germline identity and IgG isotype of clonal and non-clonal sequences (bottom right) from all 12 subjects and time points. NGS reference data from 48 healthy individuals (collected before the outbreak of SARS-CoV-2) are depicted in red (see also Tables S1 and S2). Bar and line plots show mean ± SD.

(F) Ratio of κ and λ light chains in non-clonal (top, gray) and clonal (bottom, blue) sequences (see also Figure S2).

To investigate B cell clonality and antibody characteristics on a single-cell level, we proceeded to sort S ectodomain-reactive IgG⁺ B cells from all five subjects at the different time points (t1, t2, and t3). We found up to 0.65% SARS-CoV-2 S-reactive B cells, with higher frequencies at later time points and with the number of days after diagnosis being predictive of the fraction of SARS-CoV-2 S-reactive B cells (p = 0.0016) (Figure 2C). From a total of 2,562 B cells, we detected 254 B cell clones (Table S3). Fifty-one percent of these clones (129) were detected recurrently, suggesting persistence of SARS-CoV-2 S-reactive B cells over the investigation period of 2.5 months. When separated by individual time points, the fraction of clonally related sequences ranged from 18%–67% across patients and remained constant or showed only moderate decreases over time (Figure 2D).

Next, we analyzed the single-cell Ig sequences (6,587 productive heavy and light chains) from all 12 patients (Figures 2E, 2F, and S2). Here, clonally related and non-clonal sequences similarly presented a broad spectrum of heavy-chain variable (V_H) gene segments, normally distributed heavy-chain complementarity-determining region 3 (CDRH3) lengths, symmetrical CDRH3 hydrophobicity distributions, and a predominance of the IgG1 isotype (Figure 2E). However, in comparison with repertoire data from healthy individuals, IgV_H 3-30 was overrepresented, and clonal sequences more often facilitated κ over λ light chains (76% in clonal versus 62.5% in non-clonal sequences, p = 0.0029) (Figures 2F and S2). Finally, V_H genes of S-reactive B cells were, on average, less mutated than V_H genes from healthy IgG⁺ repertoires (median identity of 98.3 versus 94.3, p < 0.0001) (Figures 2E and S2). We concluded that a SARS-CoV-2 S-reactive IgG⁺ B cell response readily develops after infection with the same B cell clones detectable over time and a preference for facilitating V_H gene segment 3-30.

Figure S2.

Light-chain characteristics of sorted single cells, related to Figure 2

Left and middle graphics: frequencies of V_L gene segments of clonal and non-clonal sequences are shown (κ left, λ middle). Shown on the right are ratios of κ and λ within the single sample sets in clonal and non-clonal sequences. A two-tailed Wilcoxon matched-pairs signed rank test was performed on κ / λ ratios to test for significance.

Isolation of highly potent near-germline SARS-CoV-2-neutralizing antibodies from COVID-19 patients

To determine antibody characteristics and isolate potent neutralizing antibodies, we cloned a total of 312 matched heavy- and light-chain pairs from all 12 patients. These antibodies were primarily selected on the basis of clonality with the aim to include at least one clonal member of several (but at least three) clones per individual (Table S3). However, the polyclonal response (median clone size of 2) (Table S3) and the lack of differences between clonal and non-clonal sequences suggested the presence of weakly expanded clones in the non-clonal fraction. We therefore also included about one-third randomly selected non-clonal sequences (83 antibodies) for production. From 255 successfully produced IgG1 antibodies, 79 (31%) bound to the full trimeric S ectodomain (Wrapp et al., 2020b) by ELISA with EC₅₀ values ranging between 0.02 μg/mL and 5.20 μg/mL (Figure 3A). Of these, 30 antibodies showed SARS-CoV-2 reactivity via a commercial diagnostic system (Euroimmun IgG detection kit) (Figures 3A and 3B; Table S4). Surface plasmon resonance (SPR) analyses using the RBD as an analyte for 13 SARS-CoV-2-interacting antibodies gave dissociation constant (K_D) values as low as 0.02 nM (Table S4). By determining the neutralization activity against authentic SARS-CoV-2, we found a total of 28 neutralizing antibodies among 9 of 12 individuals, and IC₁₀₀ values ranged between 100 μg/mL (assay limit) and 0.04 μg/mL (Figures 3C and 3D). Of note, neutralizing activity was mainly detected among high-affinity antibodies (Figure 3B; Table S4), and a positive correlation between neutralization and binding could be detected (Spearman's correlation coefficient [r_s] = 0.429, p = 0.023) (Figure 3E).

Figure 3.

Infected individuals can develop potent near-germline SARS-CoV-2-neutralizing antibodies that preferentially bind to the S-protein RBD

(A) Interaction of isolated antibodies with the SARS-CoV-2 S ectodomain by ELISA. Binding antibodies (blue) were defined by an EC₅₀ of less than 30 μg/mL and an optical density 415–695 nm (OD_415–695) of 0.25 or more (data not shown).

(B) EC₅₀ values (mean of duplicates) of SARS-CoV-2 S ectodomain-interacting antibodies per individual. Neutralizing antibodies are labeled in shades of red (see also Figure S5 and Table S4).

(C) Authentic SARS-CoV-2 neutralization activity (complete inhibition of VeroE6 cell infection, IC₁₀₀, in quadruplicates) of S-ectodomain-specific antibodies (red).

(D) Geometric mean potencies (IC₁₀₀) of all neutralizing antibodies.

(E) Correlation between S ectodomain binding (EC₅₀) and neutralization potency (IC₁₀₀). The correlation coefficient r_S and approximate p value were calculated by Spearman’s rank-order correlation (see also Figure S3).

(F) Epitope mapping of SARS-CoV-2 S ectodomain-specific antibodies against the RBD, truncated N-terminal the S1 subunit (aa 14–529), and a monomeric S ectodomain construct by ELISA. S2 binding was defined by interaction with monomeric S but not RBD or S1. Antibodies interacting with none of the subdomains were specified as conformational epitopes or not defined.

(G) Top: frequencies of V_H gene segments for non-neutralizing and neutralizing antibodies. Clonal sequence groups were collapsed and treated as one sample for calculation of the frequencies. Shown on the bottom are the CDRH3 length (left) and V_H gene germline identity (right) of non-neutralizing and neutralizing antibodies (see also Figure S4).

To better characterize the interaction between the SARS-CoV-2 S protein and reactive antibodies, we determined binding to a truncated N-terminal S1 subunit (including the RBD), the isolated RBD, and a monomeric S ectodomain. We found 27 of 28 neutralizing antibodies binding to the RBD but only 29% of the non-neutralizing antibodies, suggesting that the RBD is a major site of vulnerability on the S protein. Epitopes for non-neutralizing antibodies included the N-terminal S1 domain and conformational epitopes (Figure 3F; Table S4). Notably, neutralizing and non-neutralizing antibodies were characterized by a broad distribution of V_H as well as light-chain variable (V_L) gene segments and a preference for κ light chains (Figures 3G and S4). Moreover, 31 of 79 binding and 11 of 28 neutralizing antibodies demonstrated germline identities of 99%–100%, and no correlation was detected between neutralizing activity and the level of somatic mutation (Figure 3G; Table S4; Figure S3).

Figure S4.

V_L gene distribution in non-neutralizing and neutralizing antibodies, related to Figure 3

(A) Frequencies of V_L gene segments for non-neutralizing (left, gray) and neutralizing antibodies (right, red). Clonal sequence groups were collapsed and treated as one sample for calculation of the frequencies.

(B) Ratio of λ and κ light chains for neutralizing (left) and non-neutralizing S-ectodomain-specific antibodies (bottom, blue).

Figure S3.

Correlation of binding and neutralization with V_H gene characteristics, related to Figure 3

Correlation plots of EC₅₀ values of binding or neutralizing antibodies or IC₁₀₀ values of neutralizing antibodies with CDRH3 lengths or V_H germline identities. Spearman correlation coefficient r_S and approximate p values are given.

Finally, we performed a HEp-2 cell autoreactivity assay. Of 28 neutralizing antibodies, 4 showed low to moderate levels of autoreactivity (Figure S5; Table S4), and 2 of them also reacted with envelope proteins of other viruses (i.e., Ebola glycoprotein and HIV-1 gp140) (Table S4). In summary, these data show that SARS-CoV-2-neutralizing antibodies develop from a broad set of different V genes and are characterized by a low degree of somatic mutations. Moreover, we were able to isolate highly potent neutralizing antibodies that are promising candidates for antibody-mediated prevention and therapy of SARS-CoV-2 infection.

Figure S5.

Autoreactivity of selected SARS-CoV-2-binding and -neutralizing antibodies, related to Figure 3

HEp-2 cells were incubated with SARS-CoV-2 S-ectodomain antibodies at concentrations of 100 μg/mL and analyzed by indirect immunofluorescence. Representative pictures of the scoring system are shown.

Investigating ongoing somatic hypermutation in SARS-CoV-2 binding and neutralizing antibodies

To investigate the development of somatic mutations over time, we longitudinally analyzed 129 recurring B cell clones that comprised 17 binding and 6 neutralizing antibodies. To this end, we phylogenetically matched all members of a B cell clone at a given time point with the most closely related member at the consecutive time point (331 pairings in total). Mean mutation frequencies in either direction (i.e., toward higher or lower V gene germline identities) were 0.51% ± 0.61%, 0.08% ± 0.51%, and 0.01% ± 0.19% per week for all, binding, and neutralizing clonal members, respectively (Figure 4A, top). When averaging the V_H gene germline identity of concurrent clonal members, we found a moderate increase in somatic mutations over time (Figure 4A, bottom). Changes were similar for binding and neutralizing subsets, with one exception among the neutralizing antibodies that accumulated about 5% nucleotide mutations over the investigated period (Figure 4A, bottom). In line with this finding, neutralizing antibodies isolated on days 8–17 and days 34–42 after diagnosis showed V_H gene germline identities of 97.5% and 97.0%, respectively (Figure 4B). We concluded that SARS-CoV-2-neutralizing antibodies carry similar levels of somatic hypermutation independent of the time of isolation.

Figure 4.

Dynamics of somatic mutations for SARS-CoV-2-specific antibodies

(A) Distribution of mutation rates per week for clonal members (top) and median change in V_H germline identity normalized by the first measurement for each longitudinal clone (bottom).

(B) V_H gene germline identity of neutralizing antibodies from different time points. Shown on top is the mean ± SD for groups of antibodies from early or late time points (two-tailed Mann-Whitney U test). Shown on the bottom are the V_H germline identities of all isolated neutralizing antibodies depending on the time between diagnosis and blood sample collection (see also Table S4).

Potential precursor sequences of SARS-CoV-2-neutralizing antibodies can be identified among healthy individuals

The low rate of somatic mutations in the majority of binding and neutralizing antibodies emphasizes the requirement for the presence of distinct germline recombinations in the naive human B cell repertoire. To estimate the frequency of potential precursor B cells, we performed unbiased heavy- and light-chain next-generation sequencing (NGS) of the naive B cell receptor repertoires from 48 healthy donors (Table S5). All samples were collected before the SARS-CoV-2 outbreak and comprised a total of 1.7 million collapsed reads with 455,423 unique heavy, 170,781 κ, and 91,505 λ chain clonotypes (defined as identical V and joining (J) gene segment pairing and the same CDR3 amino acid sequence). Within this dataset, we searched for heavy and light chains that resemble the 79 SARS-CoV-2-binding antibodies (Figure 5A). For 14 of 79 tested antibodies, we found 61 heavy-chain clonotypes with identical V/J pairs and similar (±1 amino acid [aa] in length and up to 3 aa differences) CDRH3s in 28 healthy individuals (Figures 5B and 5C), including one exact CDRH3 match (MnC2t1p1_C12). For light chains, we identified 1,357 κ chain precursors with exact CDR3 matches that cover 41 of 62 antibodies and 109 λ chain precursors that represent 7 of 17 antibodies (Figures 5B and 5C). All 48 naive repertoires included at least one κ and one λ chain precursor. When combining heavy- and light-chain data, we found both precursor sequences of 9 antibodies in 14 healthy individuals (Figure 5C). Importantly, among these potential sequence precursors, we found close similarities to three potent neutralizing antibodies (CnC2t1p1_B4, HbnC3t1p1_G4, and HbnC3t1p2_B10). Although the NGS repertoire data did not include pairing information of heavy- and light-chain combinations, we found matched heavy- and light-chain sequences despite small sample sizes of, on average, 9,500 heavy and 2,000–3,500 light chain clonotypes per individual. We thus conclude that potential SARS-CoV-2-binding and -neutralizing antibody precursor sequences are likely to be abundant in naive B cell repertoires.

Figure 5.

Sequence precursor frequencies of SARS-CoV-2-specific antibodies in naive repertoires of healthy individuals

(A) Strategy for sequence precursor identification from healthy naive B cell receptor (BCR) repertoires. HC, heavy chain; KC, κ chain; LC, λ chain; V_H and V_L, heavy- and light-chain V gene; CDRH3 and CDRL3, heavy- and light-chain CDR3.

(B) Number of clonotypes in healthy naive B cell receptor repertoires (n = 48) with matched V/J genes from SARS-CoV-2-binding antibodies (n = 79), plotted against the CDR3 difference. Bars of included potential sequence precursors are highlighted in shades of blue. For heavy chains, CDR3s were allowed to differ 1 aa in length and contain up to 3 aa mutations. For light chains, only identical CDR3s were counted.

(C) Number of different antibody heavy and light chains for which precursors were identified and number of different individuals from which precursor sequences were isolated. Numbers in overlapping circles indicate that both heavy and light chains were detected.

See also Table S5.

Discussion

Neutralizing antibodies can effectively target pathogens, and their induction is a key objective of vaccination strategies (Fauci and Marston, 2015; Mascola and Montefiori, 2010; Walker and Burton, 2018; Zolla-Pazner et al., 2019). A detailed understanding of the human antibody response to SARS-CoV-2 is therefore critical for the development of effective immune-mediated approaches against the COVID-19 pandemic (Burton and Walker, 2020; Koff et al., 2013; Kreer et al., 2020b; Sempowski et al., 2020).

Using B cell microcultures, single-cell cloning, or high-throughput single-cell sequencing, SARS-CoV-2-neutralizing antibodies have been identified from a limited number of SARS-CoV-2-infected individuals and survivors of infection with the related betacoronavirus SARS-CoV-1. Most of these antibodies target the RBD of the S protein and interfere with viral binding to its host cell receptor angiotensin-converting enzyme 2 (ACE2) (Andreano et al., 2020; Brouwer et al., 2020; Cao et al., 2020; Hansen et al., 2020; Ju et al., 2020; Pinto et al., 2020; Robbiani et al., 2020; Seydoux et al., 2020; Shi et al., 2020; Wu et al., 2020; Zost et al., 2020). Demonstrating their potential for clinical application, SARS-CoV-2-neutralizing antibodies can dampen or prevent infection in animal models (Cao et al., 2020; Rogers et al., 2020; Zost et al., 2020), although antibody combinations might be needed to restrict the potential for viral escape and resistance (Baum et al., 2020). Through single-cell analysis of more than 4,000 SARS-CoV-2 S-protein-reactive B cells from 12 infected individuals, we identified highly potent human monoclonal SARS-CoV-2-neutralizing antibodies that preferentially target the RBD. These antibodies fully block authentic viral infection at concentrations as low as 0.04 μg/mL and therefore provide a potential option for prevention and treatment of SARS-CoV-2 infection.

Affinity maturation of antibodies through somatic hypermutation and clonal B cell selection is a hallmark of the adaptive immune response to pathogens. Chronic infections, such as HIV-1 infection, can result in extensive somatic mutation and amino acid substitutions in more than 30% of the V-gene-encoded region compared with their germline sequences (Klein et al., 2013; Scheid et al., 2011; Schommers et al., 2020; Wu et al., 2010). In contrast, SARS-CoV-2-neutralizing antibodies isolated from different individuals show a limited degree of somatic mutation (Brouwer et al., 2020; Ju et al., 2020; Robbiani et al., 2020; Rogers et al., 2020; Seydoux et al., 2020). However, previously described SARS-CoV-2 S-reactive antibodies were obtained from single time points, preventing conclusions regarding the molecular dynamics of the B cell response to SARS-CoV-2. By longitudinally analyzing the memory B cell response in five individuals for up to 2.5 months after SARS-CoV-2 transmission, we reveal that there is little additional somatic hypermutation or clonal B cell expansion over time. One explanation for this observation is the potentially limited antigenic B cell stimulation because of rapid viral clearance, to which the cellular response and additional immune factors might contribute (Grifoni et al., 2020; Ni et al., 2020). In contrast, somatic mutation over several months despite only a brief period of systemically circulating antigen can be observed after infection with EBOV (Davis et al., 2019) or the yellow fever vaccine strain 17D (Wec et al., 2020a). However, given the high binding affinity of near-germline IgG antibodies against SARS-CoV-2 in the range of nano- to picomolar K_D values, these antibodies might limit antigen access to the germinal center, the site of affinity maturation (Zhang et al., 2013). Although recent data suggest that neutralizing titers correlate with severity of infection (Long et al., 2020b; Wang et al., 2020b), it remains elusive whether this effect is caused by ongoing somatic hypermutation that results in higher antibody potency or ongoing production of highly potent antibodies that were initially generated.

SARS-CoV-2 S-reactive antibodies could be isolated from three SARS-CoV-2-naive individuals, although these antibodies were only weakly binding (Wec et al., 2020b). Whether the observation of such antibodies is a consequence of prior exposure to other human coronaviruses, as suggested for SARS-CoV-2-reactive T cells from healthy donors (Braun et al., 2020; Grifoni et al., 2020), and whether such antibodies and cells can provide background immunity remains to be elucidated. Notably, our deep sequencing analysis of the naive B cell receptor repertoires of 48 individuals sampled before the pandemic identified potential heavy- and/or light-chain precursor sequences of potent SARS-CoV-2-neutralizing antibodies in every single individual. In addition to the limited mutation rate and broad use of antibody gene segments across potent SARS-CoV-2-neutralizing antibodies, these results suggest that protective antibodies can be widely and readily induced by vaccination.

STAR★Methods

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Human IgG-APC (Clone G18-145)	BD Biosciences	Cat#550931; RRID: AB_398478
Anti-Human CD20-Alexa Fluor 700 (Clone 2H7)	BD Biosciences	Cat#560631; RRID: AB_1727447
Anti-Human IgD-Pe-Cy7 (Clone IA6-2)	BD Biosciences	Cat#561314; RRID: AB_10642457
Anti-Human IgM-FITC (Clone G20-127)	BD Biosciences	Cat#555782; RRID: AB_396117
Anti-Human CD27-PE (Clone M-T271)	BD Biosciences	Cat#560985; RRID: AB_10563213
Peroxidase AffiniPure Goat Anti-Human IgG, Fcγ fragment specific	Jackson ImmunoResearch	Cat#109-035-098; RRID: AB_2337586
AffiniPure Goat Anti-Human IgG, Fcγ fragment specific	Jackson ImmunoResearch	Cat#109-005-098; RRID: AB_2337541
Anti-6X His tag antibody	Abcam	Cat#ab9108; RRID: AB_307016
Bacterial and Virus Strains
E. coli DH5α	Thermo Fisher Scientific	Cat#18263012
BavPat1/2020	European Virus Archive global	Cat#026V-03883
Biological Samples
PBMCs, Plasma, and IgGs of donor IDFnC1	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDFnC2	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDCnC2	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDMnC1	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDMnC2	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDMnC4	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDMnC5	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDHbnC1	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDHbnC2	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDHbnC3	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDHbnC4	This paper	N/A
PBMCs, Plasma, and IgGs of donor IDHbnC5	This paper	N/A
PBMCs of 48 healthy blood donors	This paper	N/A
Chemicals, Peptides, and Recombinant Proteins
DMSO	Sigma-Aldrich	Cat#D2650; CAS: 67-68-5
DAPI	Thermo Fisher	Cat#D1306; CAS: 581-88-4
DTT	Promega	Cat#P1171
RNasin	Promega	Cat#N2515
RNaseOUT	Thermo Fisher	Cat#10777019
SuperScript IV Reverse Transcriptase	Thermo Fisher	Cat#18090050
Platinum Taq DNA Polymerase	Thermo Fisher	Cat#10966034
Platinum Taq Green Hot Start	Thermo Fisher	Cat#11966034
Q5 Hot Start High Fidelity DNA Polymerase	NEB	Cat#M0493L
T4 DNA Polymerase	New England Biolabs	Cat#M0203L
NP-40	Thermo Fisher	Cat#85124
dNTP Mix	Thermo Fisher	Cat#R1122
DTT	Sigma Aldrich	Cat#GE17-1318-01
SMARTScribe Reverse Transcriptase	Takara Bio	Cat#639537
Branched Polyethylenimine, 25 kDa	Sigma-Aldrich	Cat#408727; CAS: 9002-98-6
FreeStyle Expression Medium	Thermo Fisher	Cat#12338001
Protein G Sepharose 4 Fast Flow	GE Life Sciences	Cat#17061805
HiTrap MabSelect Protein A column	GE Life Sciences	Cat#8408255
ABTS solution	Thermo Fisher	Cat#002024
Dulbecco’s Modified Eagle Medium (DMEM)	GIBCO	Cat#11960-044
Fetal bovine serum (FBS)	Sigma-Aldrich	Cat#F9665
Sodium Pyruvate	Thermo Fisher	Cat#11360-070
L-Glutamine	Thermo Fisher	Cat#25030024
HEPES	Thermo Fisher	Cat#15630-080
GlutaMAX	Thermo Fisher	Cat#35050-061
MEM NEAA	Thermo Fisher	Cat#11140-050
FBS, Qualified	GIBCO	Cat#10270-106
0.05% Trypsin-EDTA (1x)	GIBCO	Cat#25300-096
Pen Strep	GIBCO	Cat#15070-063
L-Glutamine 200 mM (100x)	GIBCO	Cat#250030-123
Ni-NTA Agarose	Macherey-Nagel	Cat#745400.25
Strep-Tactin®XT Superflow® 50% suspension	IBA lifesciences	Cat#2-4010-010
10x Buffer BXT; Strep-Tactin®XT Elution Buffer	IBA	Cat#2-1042-025
10x Buffer W; Strep-Tactin®/Strep-Tactin®XT Wash Buffer	IBA	Cat#2-1003-100
Critical Commercial Assays
NOVA Lite Hep-2 ANA Kit	Inova Diagnostics / Werfen	Cat#066708100
CD19-Microbeads	Miltenyi Biotec	Cat#130-050-301
Microscale Antibody Kit (DyLight 488)	Thermo Fisher Scientific	Cat#53025
Molecular Probes Alexa Fluor 488 NHS Ester (Succinimidyl Ester)	Thermo Fisher Scientific	Cat#10266262
SARS-CoV-2 detection ELISA	Euroimmun	Cat#EI 2606-9601 G
Deposited Data
Cloned and tested SARS-CoV-2-neutralizing antibodies	This paper	GenBank: MT658806 - MT658861
Experimental Models: Cell Lines
293-6E cells	NRC	NRC file 11565
HEK293T cells	ATCC	Cat#CRL-11268
Vero C1008	ATCC	Cat#CRL-1586, RRID: CVCL_0574
Oligonucleotides
Single cell PCR Primer	Kreer et al., 2020a	N/A
Random Hexamer Primer	Thermo Fisher	Cat#SO142
SLIC heavy chain reverse primer (GGGTGCCAGGGGGAAGACC GATGGGCCCTTGGTCGAGGC)	This paper	N/A
SLIC kappa chain reverse primer (CTCATCAGATGGCGGGAAGA TGAAGACAGATGGTGCAGCCACCGTACG)	This paper	N/A
SLIC lambda chain reverse primer (GAAGCTCCTCACTCGAGGG YGGGAACAGAGTG)	This paper	N/A
Recombinant DNA
Human antibody expression vectors (IgG1, Igλ, Igκ)	Tiller et al., 2008	N/A
Plasmid encoding SARS-CoV-2 S ectodomain (amino acids 1−1208 of SARS-CoV-2 S; GenBank: MN908947)	Wrapp et al., 2020b	N/A
pCAGGS Plasmid encoding RBD of the SARS-CoV-2 spike protein (GenBank: MN908947; aa:319-541)	Stadlbauer et al., 2020	N/A
pCAGGS encoding SARS-CoV-2 S ectodomain “monomer” without trimerization domain (GenBank: MN908947; aa:1-1207)	This paper	N/A
pCAGGS encoding SARS-CoV-2 S1 subunit (GenBank: MN908947; aa:14-529)	This paper	N/A
pCAGGS-EBOV GPΔTM-GCN4-HIS-Avi, encoding the EBOV Makona glycoprotein (GP) ectodomain (GenBank: KJ660347; aa:1-651)	Ehrhardt et al., 2019	N/A
pCAGGS-YU-2 gp140-GCN4-HIS-Avi, encoding the HIV-1YU2 gp140 ectodomain (GenBank: M93258; aa:1- 683)	Ehrhardt et al., 2019	N/A
Software and Algorithms
Geneious R10 and Geneious Prime	Geneious	RRID: SCR_010519
Prism	GraphPad	RRID: SCR_002798
Python 3.6.8	Python Software Foundation; https://www.python.org/	RRID: SCR_008394
SciPy	SciPy developers	RRID: SCR_008058
IgBLAST 1.13.0	Ye et al., 2013	RRID: SCR_002873
Clustal Omega 1.2.3	Sievers et al., 2011	RRID: SCR_001591
pRESTO 0.5.11	Vander Heiden et al., 2014	RRID: SCR_001782
FlowJo 10.5.3	FlowJo, LLC	https://www.flowjo.com
Prism 7	GraphPad	https://www.graphpad.com/
MacVector 16.0.9	MacVector	https://www.macvector.com
Adobe Illustrator CC 2018	Adobe	https://www.adobe.com
Other
Amicon MWCO 30 kDa/10 kDa	Merck Millipore	Cat#Z677108
Leica DMI-microscope	Leica Biosystem	N/A
BD FACSAria Fusion Cell Sorter	BD Bioscience	N/A

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Florian Klein (florian.klein@uk-koeln.de).

Materials Availability

Reasonable amounts of antibodies will be made available by the Lead Contact upon request under a Material Transfer Agreement (MTA) for non-commercial usage.

Data and Code Availability

Nucleotide sequences of all SARS-CoV-2-neutralizing antibodies were deposited at GenBank (accession numbers MT658806 - MT658861). Further antibody sequences and NGS data of healthy individuals will be shared by the Lead Contact upon request.

Experimental Models and Subject Details

SARS-CoV-2-infected individuals and sample collection

Samples were obtained under a study protocol approved by the Institutional Review Board of the University of Cologne and respective local IRBs (study protocol 16-054). All 12 participants (six females, six males; Table S1) provided written informed consent and were recruited at hospitals or as outpatients. Sites of recruitment were Munich Clinic Schwabing for IDMnC1, IDMnC2, IDMnC4, and IDMnC5; the University Hospital of Frankfurt for patients IDFnC1 and IDFnC2; and the Institute of Virology at the University Hospital Cologne for patient IDCnC2. Patients IDHbnC1-5 were recruited as outpatients in the county Heinsberg. Sample size was estimated based on previous studies (Corti et al., 2011; Ehrhardt et al., 2019; Schommers et al., 2020) to sufficiently inform on B cell receptor repertoires and yield neutralizing antibodies. Participants were enrolled and allocated to single blood draws or longitudinal follow-up based on the epidemiology of the infection and participants’ availability.

Method Details

Isolation of peripheral blood mononuclear cells (PBMCs), plasma and total IgG from whole blood

Blood draw collection was performed using EDTA tubes and/or syringes pre-filled with heparin. PBMC isolation was performed using Leucosep centrifuge tubes (Greiner Bio-one) prefilled with density gradient separation medium (Histopaque; Sigma-Aldrich) according to the manufacturer’s instructions. Plasma was collected and stored separately. For IgG isolation, 1 mL of the collected plasma was heat-inactivated (56°C for 40 min) and incubated with Protein G Sepharose (GE Life Sciences) overnight at 4°C. The suspension was transferred to chromatography columns and washed with PBS. IgGs were eluted from Protein G using 0.1 M glycine (pH = 3.0) and buffered in 0.1 M Tris (pH = 8.0). For buffer exchange to PBS, 30 kDa Amicon spin membranes (Millipore) were used. Purified IgG concentration was measured using a Nanodrop (A280) and samples were stored at 4°C.

SARS-CoV-2 S protein expression and purification

The construct encoding the prefusion stabilized SARS-CoV-2 S ectodomain (amino acids 1−1208 of SARS-CoV-2 S; GenBank: MN908947) was kindly provided by Jason McLellan (Texas, USA) and described previously (Wrapp et al., 2020b). In detail, two proline substitutions at residues 986 and 987 were introduced for prefusion state stabilization, a “GSAS” substitution at residues 682–685 to eliminate the furin cleavage site, and a C-terminal T4 fibritin trimerization motif. For purification, the protein is C-terminally fused to a TwinStrepTag and 8XHisTag. Protein production was done in HEK293-6E cells by transient transfection with polyethylenimine (PEI, Sigma-Aldrich) and 1 μg DNA per 1 mL cell culture medium at a cell density of 0.8 10⁶ cells/mL in FreeStyle 293 medium (Thermo Fisher Scientific). After 7 days of culture at 37°C and 5% CO₂, culture supernatant was harvested and filtered using a 0.45 μm polyethersulfone (PES) filter (Thermo Fisher Scientific). Recombinant protein was purified by Strep-Tactin affinity chromatography (IBA lifescience, Göttingen Germany) according to the Strep-Tactin XT manual. Briefly, filtered medium was adjusted to pH 8 by adding 100 mL 10x Buffer W (1 M Tris/HCl, pH 8.0, 1.5 M NaCl, 10 mM EDTA, IBA lifescience) and loaded with a low pressure pump at 1 mL/min on 5 mL bedvolume Strep-Tactin resin. The column was washed with 15 column volumes (CV) 1x Buffer W (IBA lifescience) and eluted with 6 × 2.5 mL 1x Buffer BXT (IBA lifescience). Elution fractions were pooled and buffer was exchanged to PBS pH 7.4 (Thermo Fisher Scientific) by filtrating four times over 100 kDa cut-off cellulose centrifugal filter (Merck).

Cloning and expression of different SARS-CoV-2 S protein subunits and Ebola surface glycoprotein

The RBD of the SARS-CoV-2 spike protein (GenBank: MN908947; aa: 319-541), fused to a hexahistidine tag, was expressed from a plasmid kindly provided by Florian Krammer (Stadlbauer et al., 2020) through transient calcium phosphate transfection of HEK293T cells. After 7 days of culture at 37°C and 5% CO₂, culture supernatant was harvested and filtered using a 0.45 μm PES filter (Thermo Fisher Scientific). RBD protein was purified via the hexahistidine tag using Ni-NTA agarose (Macherey-Nagel). To this end, filtered culture supernatant was mixed with an equal volume of 2x NPI10 buffer (1x NPI10 buffer: 50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, pH 8). Ni-NTA beads were equilibrated with NPI10 and added to the cell culture supernatant/NPI10 mix (1 mL bed volume per 1000 mL original cell culture supernatant), followed by incubation at 4°C over night with constant rotation. Beads were harvested by centrifugation at 500 g for 5 min at 4°C and washed twice in NPI10 (100 mL per 1000 mL original cell culture supernatant), followed by centrifugation at 500 g for 5 min at 4°C. Beads were additionally washed three times in NPI20 (50 mM NaH₂PO₄, 300 mM NaCl, 20 mM imidazole, pH 8). Beads were then transferred in NPI20 to Polyprep chromatography columns (BioRad, 2 columns per 1000 mL original cell culture supernatant) and washed with 10 mL NPI20 per column. Protein was eluted with 5 mL NPI250 (50 mM NaH₂PO₄, 300 mM NaCl, 250 mM imidazole, pH 8). Buffer was exchanged to PBS using 10 kDa Amicon spin columns (Millipore).

SARS-CoV-2 S ectodomain “monomer” without trimerization domain (GenBank: MN908947; aa:1-1207) and S1 subunit (GenBank: MN908947; aa:14-529) regions of the spike DNA were amplified from a synthetic gene plasmid (furin site mutated) (Wrapp et al., 2020b) by PCR. PCR products were cloned into a modified sleeping beauty transposon expression vector containing a C-terminal thrombin cleavage and a double Strep II purification tag. For the S1 subunit, the tag was added at the 5′ end and a BM40 signal peptide was included. For recombinant protein production, stable HEK293 EBNA cell lines were generated employing the sleeping beauty transposon system (Kowarz et al., 2015). Briefly, expression constructs were transfected into the HEK293 EBNA cells using FuGENE HD transfection reagent (Promega). After selection with puromycin, cells were induced with doxycycline. Supernatants were filtered and the recombinant proteins purified via Strep-Tactin®XT (IBA Lifescience) resin. Proteins were then eluted by biotin-containing TBS-buffer (IBA Lifescience), and dialyzed against TBS-buffer.

Ebola surface glycoprotein (EBOV Makona, GenBank: KJ660347; GP aa:1-651) and HIV-1 gp140 (strain YU2, GenBank: M93258; Env aa:1-683), both lacking the transmembrane domain and containing a GCN4 trimerization domain (Ehrhardt et al., 2019), were expressed by transient transfection of HEK293-6E cells using PEI, following the same protocol and culture conditions as for the prefusion stabilized SARS-CoV-2 S ectodomain described above. After 7 days, supernatants were filtered using a 0.45 μm polyethersulfone (PES) filter (Thermo Fisher Scientific) and proteins were purified by affinity chromatography through their hexahistidine tag with Ni-NTA agarose (Macherey-Nagel) following the same protocol as for the RBD purification described above.

Isolation of SARS-CoV S ectodomain-specific IgG⁺ B cells

B cells were isolated from PBMCs using CD19-microbeads (Miltenyi Biotec) according to the manufacturer’s instruction. Isolated B cells were stained for 20 min on ice with a fluorescence staining-mix containing 4’,6-Diamidin-2-phenylindol (DAPI; Thermo Fisher Scientific), anti-human CD20-Alexa Fluor 700 (BD), anti-human IgG-APC (BD), anti-human CD27-PE (BD) and DyLight488-labeled SARS-CoV-2 spike protein (10μg/mL). Dapi⁻, CD20⁺, IgG⁺, SARC-CoV-2 spike protein positive cells were sorted using a FACSAria Fusion (Becton Dickinson) in a single cell manner into 96-well plates. All wells contained 4 μL buffer, consisting of 0.5x PBS, 0.5 U/μL RNAsin (Promega), 0.5 U/μL RNaseOUT (Thermo Fisher Scientific), and 10 mM DTT (Thermo Fisher Scientific). After sorting, plates were immediately stored at −80°C until further processing.

Antibody heavy/light chain amplification and sequence analysis

Single cell amplification of antibody heavy and light chains was mainly performed as previously described (Kreer et al., 2020a; Schommers et al., 2020). Briefly, reverse transcription was performed with Random Hexamers (Invitrogen), and Superscript IV (Thermo Fisher Scientific) in the presence of RNaseOUT (Thermo Fisher Sicentific) and RNasin (Promega). cDNA was used to amplify heavy and light chains using PlatinumTaq HotStart polymerase (Thermo Fisher Scientific) with 6% KB extender and optimized V gene-specific primer mixes (Kreer et al., 2020a) in a sequential semi-nested approach with minor modifications to increase throughput (unpublished data). PCR products were analyzed by gel electrophoresis for correct sizes and subjected to Sanger sequencing. For sequence analysis, chromatograms were filtered for a mean Phred score of 28 and a minimal length of 240 nucleotides (nt). Sequences were annotated with IgBLAST (Ye et al., 2013) and trimmed to extract only the variable region from FWR1 to the end of the J gene. Base calls within the variable region with a Phred score below 16 were masked and sequences with more than 15 masked nucleotides, stop codons, or frameshifts were excluded from further analyses. Clonal analysis was performed separately for each patient. All productive heavy chain sequences were grouped by identical V_H/J_H gene pairs and the pairwise Levenshtein distance for their CDRH3s was determined. Starting from a random sequence, clone groups were assigned to sequences with a minimal CDRH3 amino acid identity of at least 75% (with respect to the shortest CDRH3). 100 rounds of input sequence randomization and clonal assignment were performed and the result with the lowest number of remaining unassigned (non-clonal) sequences was selected for downstream analyses. All clones were cross-validated by the investigators taking shared mutations and light chain information into account.

Next generation sequencing and evaluation of healthy control IgG⁺ and naive B cell repertoires

B cell receptor repertoire sequence data were generated by an unbiased template-switch-based approach as previously described (Ehrhardt et al., 2019; Schommers et al., 2020). In brief, PBMCs from 48 healthy individuals (samples collected before the SARS-CoV-2 outbreak from the Institute of Transfusion Medicine of the University Hospital of Cologne under protocol 16-054 approved by the Institutional Review Board of the University of Cologne) were enriched for CD19⁺ cells with CD19-microbeads (Miltenyi Biotec). For each individual, 100,000 CD20⁺IgG⁺ and 100,000 CD20⁺IgD⁺IgM⁺CD27⁻IgG⁻ B cells were sorted into FBS (Sigma-Aldrich) using a BD FACSAria Fusion. RNA was isolated with the RNeasy Micro Kit (QIAGEN) on a QiaCube (QIAGEN) instrument. cDNA was generated by template-switch reverse transcription according to the SMARTer RACE 5′/3′ manual using the SMARTScribe Reverse Transcriptase (Takara) with a template-switch oligo including an 18-nucleotide unique molecular identifier (UMI). Heavy and light chain variable regions were amplified in a constant region-specific nested PCR and amplicons were used for library preparation and Illumina MiSeq 2 × 300 bp sequencing. Raw NGS reads were pre-processed and assembled with an in-house pipeline based on custom Python scripts, IgBLAST (Ye et al., 2013), Clustal Omega (Sievers et al., 2011), and the pRESTO toolkit (Vander Heiden et al., 2014) as previously described (Ehrhardt et al., 2019). In brief, raw reads were initially filtered by Phred quality score (mean of 25 or higher) and read length (250 bp or longer). Unique molecular identifiers (UMIs) were extracted and paired reads were pre-annotated with IgBLAST. Based on IgBLAST pre-annotation, an additional molecular identifier (MID) was extracted by taking consecutive 18 nucleotides (nt) starting 12 nt downstream of the end of framework region (FWR) 3. For error correction, reads with the same UMIs were grouped. Reads that did not match the most abundant V gene call or had more than 1 nt difference to any other read were removed from their assigned UMI group. Assuming that ungrouped and removed reads may result from RT, PCR, or sequencing errors within the UMI, the remaining single as well as the removed reads were re-grouped by their MID. MID groups with a unique V gene call and no more than 1 nt difference between included UMIs were re-defined as a novel UMI group. Clustal omega was used to align all reads within each corrected UMI group and aligned sequences were collapsed to build consensus reads. For consensus building, base calls were weighted by their quality (1 - error probability) and bases with the highest quality-weighted frequencies were taken as the consensus. Paired consensus reads were assembled with the pRESTO AssemblePairs module and a minimal overlap set to 6 nt. Assembled sequences were annotated with IgBLAST and productive sequences were kept for analyses. To minimize the influence of sequencing and PCR errors, NGS-derived sequences were only evaluated for UMI groups with at least three reads. For the identification of overlapping clonotypes in healthy individuals a maximum of one amino acid length difference and three or less differences in absolute amino acid composition of CDR3s were considered as similar.

Cloning and production of monoclonal antibodies

Antibody cloning from 1^st PCR products was performed as previously described (Schommers et al., 2020; Tiller et al., 2008) by sequence and ligation-independent cloning (SLIC; von Boehmer et al., 2016) with a minor modification. In brief, 1 μL of the 1^st PCR product was used for amplification with Q5 Hot Start High Fidelity DNA Polymerase (New England Biolabs) as well as specific forward- and reverse primers including overhangs for subsequent SLIC into expression vectors (IgG1, Igλ, Igκ; Tiller et al., 2008). In contrast to the published protocol, PCR amplification for SLIC assembly was performed with extended forward primers based on 2^nd PCR primers (Kreer et al., 2020a) covering the complete endogenous leader sequence of all heavy and light chain V genes (unpublished data) and reverse primers specific for the 5′ end of heavy and light chain constant regions. PCR conditions were 98°C for 30 s; 35 cycles of 98°C for 10 s, 72°C for 45 s; and 72°C for 2 min. PCR products were purified (NucleoSpin 96 PCR Clean-up, Macherey Nagel), cloned into expression vectors by SLIC using T4 DNA polymerase (NEB) and chemical competent Escherichia coli DH5α. After verification of positive colonies by colony PCR and Sanger sequencing, plasmids were amplified and purified from midi cultures (Macherey Nagel).

Monoclonal antibodies were produced by transient co-transfection of 293-6E cells with human heavy chain (IgG1 isotype) and light chain antibody expression plasmids using polyethylenimine (PEI) (Sigma-Aldrich). Following transfection, cells were maintained in FreeStyle 293 Expression Medium (Thermo Fisher) and 0.2% penicillin/streptomycin (Thermo Fisher) at 37°C and 6% CO₂ and kept under constant shaking at 90-120 rpm. Seven days after transfection, supernatants were harvested and clarified by centrifugation and subsequent filtration using PES filters. For antibody purification, Protein G-coupled Sepharose beads (GE Life Sciences) were incubated with culture supernatants. After washing, antibodies were eluted from the Protein G-coupled beads in chromatography columns using 0.1 M glycine (pH = 3) and buffered using 1 M Tris (pH = 8). Finally, buffer exchange to PBS was performed using Amicon spin membranes (Millipore). Antibody concentrations were determined using UV spectrophotometry (Nanodrop, Thermo Fisher) and antibodies were stored at 4°C until further use.

ELISA analysis to determine antibody binding activity to SARS-CoV-2 S and subunit binding

ELISA plates (Corning 3369) were coated with 2 μg/mL of protein in PBS (SARS-CoV-2 spike ectodomain, RBD, or N-terminal truncated S1) or in 2 M Urea (SARS-CoV-2 spike ectodomain “monomer” lacking the trimerization domain) at 4°C overnight. For SARS-CoV-2 spike ectodomain ELISA, plates were blocked with 5% BSA in PBS for 60 min at RT, incubated with primary antibody in 1% BSA in PBS for 90 min, followed by anti-human IgG-HRP (Southern Biotech 2040-05) diluted 1:2500 in 1% BSA in PBS for 60 min at RT. SARS-CoV-2 spike subunit ELISAs were done following a published protocol (Stadlbauer et al., 2020). ELISAs were developed with ABTS solution (Thermo Fisher 002024) and absorbance was measured at 415 nm and 695 nm. Positive binding was defined by an OD ≥ 0.25 and an EC₅₀ < 30 μg/mL. The commercial anti-SARS-CoV-2 ELISA kit for immunoglobulin class G was provided by Euroimmun (Euroimmun Diagnostik, Lübeck, Germany). Antibody detection was done according to manufacturer’s instructions and a concentration of 50 μg/mL of antibodies and 2 mg/mL of plasma IgG was used. The samples were tested using the automated platform Euroimmun Analyzer 1.

Virus neutralization test

SARS-CoV-2 neutralizing activity of poly-IgG samples or human monoclonal antibodies was investigated based on a previously published protocol for MERS-CoV Koch and Dahlke (2020). Briefly, samples were serially diluted in 96-well plates starting from a concentration of 1,500 μg/mL for plasma-isolated IgG and 100 μg/mL for monoclonal antibodies. Samples were incubated for 1 h at 37°C together with 100 50% tissue culture infectious doses (TCID₅₀) SARS-CoV-2 (BavPat1/2020 isolate, European Virus Archive Global # 026V-03883). Cytopathic effect (CPE) on VeroE6 cells (ATCC CRL-1586) was analyzed 4 days after infection. Neutralization was defined as absence of CPE compared to virus controls. For each test, a positive control (neutralizing COVID-19 patient plasma) was used in duplicates as an inter-assay neutralization standard.

Surface Plasmon Resonance (SPR) measurements

For SPR measurement, the RBD was additionally purified by size exclusion chromatography (SEC) purification with a Superdex200 10/300 column (GE Healthcare). Binding of the RBD to the various mAbs was measured using single-cycle kinetics experiments with a Biacore T200 instrument (GE Healthcare). Purified mAbs were first immobilized at coupling densities of 800-1200 response units (RU) on a series S sensor chip protein A (GE Healthcare) in PBS and 0.02% sodium azide buffer. One of the four flow cells on the sensor chip was empty to serve as a blank. Soluble RBD was then injected at a series of concentrations (i.e., 0.8, 4, 20, 100, and 500 nM) in PBS at a flow rate of 60 μL/min. The sensor chip was regenerated using 10 mM Glycine-HCl pH 1.5 buffer. A 1:1 binding model was used to describe the experimental data and to derive kinetic parameters. For some mAbs, a 1:1 binding model did not provide an adequate description for binding. In these cases, we fitted a two-state binding model that assumes two binding constants due to conformational change. In these cases, we report the first binding constants (K_D¹).

HEp-2 Cell Assay

Monoclonal antibodies were tested at a concentration of 100 μg/mL in PBS using the NOVA Lite HEp-2 ANA Kit (Inova Diagnostics) according to the manufacturer’s instructions, including positive and negative kit controls on each substrate slide. HIV-1-reactive antibodies with known reactivity profiles were included as additional controls. Images were acquired using a DMI3000 B microscope (Leica) and an exposure time of 3.5 s, intensity of 100%, and a gain of 10.

Quantification and Statistical Analysis

Flow cytometry analysis and quantifications were done with FlowJo10. Statistical analyses were performed using GraphPad Prism (v7), Microsoft Excel for Mac (v14.7.3), Python (v3.6.8), and R (v4.0.0). To test for a significant increase in S-reactive B cells over time (Figure 2C and results section ‘Longitudinal analysis of the SARS-CoV-2 antibody response’) Linear Mixed Effects Models (R-function nlme::lme(), Pinheiro et al., 2020) were applied to the composite of all longitudinal data points, with individuals having their own intercept. V gene usage, CDRH3 length and V gene germline identity distributions for clonal sequences (Figure 2E) were determined for all input sequences without further collapsing. CDRH3 hydrophobicity (Figure 2E) was calculated based on the Eisenberg-scale (Eisenberg et al., 1984). To test for a significant difference of κ / λ ratios between clonal versus non-clonal sequences, D’Agostino-Pearson normality test (Prism, GraphPad) was used to test for normality and a Wilcoxon matched-pairs signed rank test was performed (Prism, GraphPad) on κ / λ ratios of each subset (Figure S2). V gene statistics for neutralizer and non-neutralizer (Figure 3G) were calculated from collapsed clonal sequences. For correlation analyses (Figures 3E and S3), spearman’s rank correlation coefficients were calculated in Prism (GraphPad). For longitudinal analyses on mutation frequencies of recurring clones (Figure 4A), a multiple sequence alignment for the B cell sequences was calculated with Clustal Omega (version 1.2.3) (Sievers et al., 2011) using standard parameters. From this, a phylogenetic tree of the sequences was estimated with RAxML through the raxmlGUI (version 2.0.0-beta.11) (Edler et al., 2019) using the GTRGAMMA substitution model (RAxML version 8.2.12) (Stamatakis, 2014). Based on the phylogenetic tree distances, all variants of a clone at a given time point were matched to variants at the consecutive time point and the slope between the pairs was computed. Hamming distances between the pairs were determined and normalized for sequence length and time difference to calculate the mean mutation frequency per day. Given the median slope per clone, a one-sided Wilcoxon Signed Rank Test was applied to test whether the slopes are equal to zero, with the alternative hypothesis that the slopes are smaller than zero. For visualizing the change of V_H gene germline identity over time (Figure 4A, bottom), the germline identity for each clone was normalized by its median value at the first-time measurement and the median slope was plotted. To test for a difference in V_H gene germline identity between neutralizing antibodies that were isolated at early or late time points, antibodies were separated into two groups (8–17 and 34–42 days, respectively). D’Agostino-Pearson normality test (Prism, GraphPad) was used to test for normality and a two-tailed unpaired Mann-Whitney U test was performed (Prism, GraphPad).

Published: July 7, 2020