key: cord-0927160-ex1d0ohw
authors: Ehling, Roy A.; Weber, Cédric R.; Mason, Derek M.; Friedensohn, Simon; Wagner, Bastian; Bieberich, Florian; Kapetanovic, Edo; Vazquez-Lombardi, Rodrigo; Di Roberto, Raphaël B.; Hong, Kai-Lin; Wagner, Camille; Pataia, Michele; Overath, Max; Sheward, Daniel J.; Murrell, Ben; Yermanos, Alexander; Cuny, Andreas P.; Savic, Miodrag; Rudolf, Fabian; Reddy, Sai T.
title: SARS-CoV-2 reactive and neutralizing antibodies discovered by single cell sequencing of plasma cells and mammalian display
date: 2021-12-22
journal: Cell Rep
DOI: 10.1016/j.celrep.2021.110242
sha: 1a6c52bd218d0b4eabdc6171352e8a611d6753f6
doc_id: 927160
cord_uid: ex1d0ohw

Characterization of COVID-19 antibodies has largely focused on memory B cells, however it is the antibody-secreting plasma cells that are directly responsible for the production of serum antibodies, which play a critical role in resolving SARS-CoV-2 infection. However, little is known about the specificity of plasma cells, largely because plasma cells lack surface antibody expression, thereby complicating their screening. Here, we describe a technology pipeline that integrates single-cell antibody repertoire sequencing and mammalian display to interrogate the specificity of plasma cells from 16 convalescent patients. Single-cell sequencing allows us to profile antibody repertoire features and identify expanded clonal lineages. Mammalian display screening is employed to reveal that 43 antibodies (of 132 candidates) derived from expanded plasma cell lineages are specific to SARS-CoV-2 antigens, including antibodies with high affinity to the SARS-CoV-2 receptor binding domain (RBD) that exhibit potent neutralization and broad binding to the RBD of SARS-CoV-2 variants (of concern/interest).

Serum antibody responses against SARS-CoV-2 play a critical role in resolving viral infection (Hung et representing a relevant target for antibodies. The S protein contains the S1 subunit, including the 43 receptor-binding domain (RBD) and the S2 subunit. S1 is highly immunogenic and a target for 

Earlier this year, the emergence of variants such as beta and gamma that showed resistance to several 

To interrogate the specificity of PCs in convalescent patients recovered from COVID-19, we describe a 92 pipeline that integrates single-cell antibody repertoire sequencing, genome editing and high-throughput 93 mammalian display screening (Fig. 1) . We first perform single-cell antibody sequencing of PCs isolated 109 PBMCs were isolated from a set of 16 convalescent patients, which was part of a larger cohort biobank 110 (Kaltenbach et al., 2020) . Among the selected group of 16, all patients were confirmed to be SARS-CoV- antibody responses to SARS-CoV-2 S1 and nucleocapsid protein (NCP) antigens (data not shown).

Clinical characteristics of the patients are summarized in Fig. 2 ; all patients displayed mild symptoms 114 (no hospitalization required) and with the exception of patients 2, 14 and 15, none of the patients 115 required any assistance ( Fig. 2A, left) . PBMCs were collected within 9-24 days after onset of symptoms 116 (average: 13 days) or 0-17 days after resolution of the disease (average ~5.1 days). A follow-up ELISA 117 on patient serum (Fig. 2B ) failed to detect S1 specific IgA+IgG titers in two patients (F5921513, 118 F5931620), and IgG titers in four patients (F5921407, F5921795, F5921393, F5921915), thus highlighting 119 a discordance to POCTs. Sensitivity and specificity of the particular POCT used was >92% and 99%, 120 respectively (Rudolf et al., 2020) . The two patients that lacked both S1-specific IgA and IgG titers 121 (F5921513, F5931620) showed IgM specificity to SARS-CoV-2 NCP (Fig. 2B) .

Single-cell antibody repertoire sequencing and analysis of PCs from convalescent patients

To interrogate the antibody repertoire of PCs, we performed single-cell sequencing. First, PCs were 125 isolated from PBMCs based on CD138 expression using magnetic beads. Next, single-cell sequencing 126 libraries were prepared using the 10X Genomics Chromium system and the V(D)J protocol, which 127 combines gel encapsulation and DNA-barcoding to tag antibody transcripts of both HC and LC 128 originating from the same cell. Processing the raw reads resulted in approximately 8.3 million reads per 

was also observed (Fig 3A, Fig S2) . Next, we measured the percentage identity to germline V-gene, 140 which served as a proxy for somatic hypermutation (SHM); this showed as expected that IGHG 141 (94.4±0.1%) and IGHA (93.4±0.1%) had lower V-gene identity than IGHM (98.1±0.05%), however on a 142 repertoire level this indicated that class-switched antibodies had only a minor degree of SHM relative 143 to repertoires from other viral infections (e.g., lymphocytic choriomeningitis virus) (Neumeier et al., 144 2021a) (Fig 3B) . We then examined the clonal expansion profiles per patient, identifying various degrees 145 of clonal expansion, with several patients showing highly expanded IGHG and IGHA clones (Fig 3C) . The infection. We therefore compared the germline profiles from our single-cell repertoire data to the bulk 156 repertoire data from these two studies and observed very similar trends in IGHV gene usage, including 157 an increased usage of IGHV3-30, IGHV3-30-3, IGHV3-33 and IGHV5-51 relative to a healthy control 158 repertoire data set (Ghraichy et al., 2020) (Fig 3D) . While an increased IGHJ6 usage was observed for 159 the dataset from Nielsen et al., the overall usage distribution is similar for all datasets (Fig 3E) . Analysis 160 of the CDRH3 length distribution also showed no major difference neither between COVID-19 repertoire 161 studies and the healthy control nor across isotype (Fig S2-S6 ).

We then compared the pairing ratios of IGHV with IGKV or IGLV from our patient cohort with that of and healthy pairing ratios were generally very similar, notably IGKV3-20 and IGKV1-39 were both shown   165 to pair with a wide range of IGHV, while others such as IGLV1-47 were only observed in combination 166 with a limited number of IGHV (i.e., IGHV1-2) (Fig. S1) . However, this can at least partly be explained 167 by the increased relative frequencies at which LCs such as IGKV3-20 are observed in overall repertoires.

Additional analysis of paired CDRL3 and CDRH3 length distributions did not reveal any trends or 169 preferences in COVID-19 antibody repertoires (Fig. S2A-G) . Table S2B ). This statistical method aids in determining true binders from 246 background (false negatives of binders) and accounts for an uneven distribution of sequences in the 247 parental Ab+ populations. Highly enriched antibody sequences, as well as sequences differentially 248 enriched for one of the S1 or S2 subunits were considered binders to the respective SARS-CoV-2 antigen 249 and selected for downstream characterization. In addition, neutral or weakly depleted sequences were 250 also chosen to help determine the enrichment cutoffs for binding vs. non-binding antibodies.

Based on enrichment, a subset of 43 sequences was selected and Cas9 HDR was performed on nPnP 252 cells to generate mAb cell lines. Flow cytometry on the 43 monoclonal PnP cell lines was performed to 253 verify specificity to either SARS-CoV-2 S1, RBD or S2 antigen ( Fig. 5B-D) . Out of 43 tested candidates, 254 37 displayed strong or sufficient Stokes shifts in the fluorescence channel (561 nm) to confirm their 255 specificity. For some candidates, strongly trailing peaks indicated an incomplete staining possibly due 256 to differential antibody expression and stability. Nevertheless, these antibodies can still be considered 257 specific to SARS-CoV-2 antigen due to a fluorescence signal in both 561 and 633 nm channels ( Fig.   258   S4A ). Additionally, an unrelated mFc-tagged antigen (here: anti-hCD69-mFc) was used to verify that the 259 observed signal was not a result of mFc interactions (selected samples remained negative for the 260 unrelated mFc-tagged antigen, while positive for S1 and S2 antigens) (Fig. S4B ). S1 binders ( Fig. 5B) 

were also labeled with RBD antigen, revealing that at least three clones appeared to bind to the RBD 262 ( Fig. 5C ) and could potentially be neutralizing antibodies. Flow cytometry confirmed that most of the 263 highly enriched clones, but surprisingly also some differentially depleted clones, showed specificity to CoVs: 229E, HKU1, OC43 and NL63, in addition to SARS-CoV-2 antigens S1, S2 and NCP (Fig. S5) . As a 272 reference, the CR3022 supernatant showed strong binding to SARS-CoV-1 and SARS-CoV-2 spike, while

showing no cross-reactivity with any other CoV antigen. The antibodies from Pool A showed more 274 binding to all CoV antigens, while the Pool B antibodies appeared to be more focused on S1 and NCP 275 antigens, which was correlated with what was observed by FACS.

From the previously missing sequences from Pool B, sAb HDR templates were subsequently repeated 277 for Cas9 HDR integration into nPnP cells (in bulk), followed by FACS to determine their antigen 278 specificity. While no candidates were checked individually, Ab+ sorting revealed that all but one of the 279 13 mAbs were found to express productive antibody on their surface; antigen-sorting and deep 280 sequencing analysis on RBD, S1 and S2 positive events led to identification of two additional RBD 281 binders (mAb-56, mAb-76) and four additional S2 binders (mAb-78, -80, -111, -115) ( Fig. S3B-C) .

Patient and SARS-CoV-2-specific antibody sequence analysis 283 43 of the 132 (32.5%) antibody sequences selected from the highly expanded PC clonal lineages could 284 be confirmed to be specific for SARS-CoV-2 S1, RBD or S2 antigen (Table S2C) . By examining these 285 SARS-CoV-2-specific sequences and which patients they originated from, we observed patient-related 286 trends (Table S3) . For instance, ~67% (6/9) of antibody sequences selected from patient F5921407 were 287 SARS-CoV-2 antigen-specific, while for another patient only 10% (1/10) of tested antibody sequences 288 demonstrated antigen binding. As the number of selected sequences varied per patient, the mean hit-289 rate per patient was slightly higher at 37%. From our cohort, serum from two of the patients (F5921513, 290 F5921620) showed a complete lack of S protein reactive antibodies by ELISA (Fig. 2B) , which is in stark (mAb-1, mAb-9, mAb-104 and mAb-135) from highly expanded PCs of these two patients were shown 294 to bind to S1 or S2 antigens. The 37 initially observed SARS-CoV-2-specific antibody sequences were 295 distinct from each other and did not strongly correlate with trends observed in the overall repertoire. (Fig. 5E, Fig. S6A ). The most frequently observed V H -gene (IGHV3-30, seen in 8 binders) 298 was also present at an increased frequency in both the overall repertoires as well as the pool of 132 299 selected sequences. The CDRH3 lengths ranged from 6 to 26 (CDRL3: 8 and 15) ( Fig. S6B ) and the set 300 of binders exhibited a broad range of SHM with V-gene identities between 85%-100% (Fig. S6C ).

Notably, 11 of the antibodies possessed no SHM, with a germline identity of 100% in both their V H and 302 V L sequences. Comparing our results to previously published SARS-CoV-2-specific antibody sequences, of 94%, which was the same as the average for the whole repertoire. Next, we measured the binding 325 affinities of the RBD-specific antibodies by using a biotinylated prefusion stabilized SARS-CoV-2 S1 326 antigen and biolayer interferometry (BLI). All three antibodies exhibited similar binding kinetics, with 327 apparent equilibrium dissociation constants (K D ) values of 1.61 nM (mAb-50), 1.96 nM (mAb-64) and 328 1.78 nM (mAb-82) (R 2 = 0.99) (Fig. 6A-C) . Antibodies derived from B mem have thus far been reported to 329 routinely display affinities in the nanomolar (K D ≈ 1 x 10 -9 ) to subnanomolar (K D < 1 x 10 -9 ), and even 330 in some cases the low picomolar range (K D ≈ 1 x 10 -12 ) ( 

we performed epitope mapping of our PC-derived antibodies by using an in-tandem competition assay: 337 SARS-CoV-2 S1 was immobilized to Streptavidin biosensors, followed by incubation with an antibody 1 338 and a secondary, competing antibody 2 at a lower concentration. As controls we used the CR3022 339 antibody or the extracellular domain of the ACE-2 receptor (Fig. 6D) . As expected, competition and self- epitope. The epitope targeted by mAb-50 did not overlap with either mAb-64 or mAb-82 (Fig. 6E) .

Finally, to determine if RBD-specific antibodies were indeed able to neutralize SARS-CoV-2, we repertoires with sequences from the CoV-AbDab, we were able to identify 45 antibodies with an amino 375 acid sequence similarity above 80% in CDRH3 (Table S4 ). In addition, we also found that mAb-64 shared sequence space, to be tested for specificity to SARS-CoV-2 antigens.

In order to enable the rapid screening and discovery of SARS-CoV-2 reactive candidates within highly 383 expanded PC lineages, we adapted a mammalian display platform based on the previously reported further characterize specificity to SARS-CoV-2 antigens. Thereby, we could show that in most of the 390 patients in our cohort (11/16) there were highly expanded PCs producing antibodies with specificity to 391 SARS-CoV-2. While we were unable to discover antibodies from all patients with specificity to S1 or S2, 392 this could be explained by the fact that some patients had low or no detectable serum antibody titers 393 against S protein (Fig. 2B ) or in some patients there was low single-cell sequencing depth that may 394 have compromised clonal expansion analysis (Fig. 3C) . In both Pool A and Pool B, antibodies showed 395 cross-reactivity to other human CoV antigens, including HCoV-229E, OC43, HKU1 and NL63, but did not 396 show any specificity to MERS-CoV-Spike or SARS-CoV-1 Spike, which is in contrast to the cross-specific 

Briefly, genomic DNA was extracted from 1x10 6 cells using a PureLink™ Genomic DNA mini Kit (Thermo, 738 kinetics buffer at 20 nM (mAb-50) and 60 nM (mAb-64, -82) for 60s, followed by (4) a dissociation step 739 in 1X kinetics buffer for 300s. Fitting was performed using the ForteBio Octet HTX software package as 

Cell supernatant containing secreted antibodies was analyzed for the presence of coronavirus specific 698 responses. First, supernatant was collected and filtered with 0.2 µm filters (Sartorius, 16534-K). 96-well 699 microtiter plates were coated with recombinant coronavirus antigens purchased from SinoBiological: 700 SARS2-Spike S1+S2 (40589-V08B1)

1 hour at RT, cell supernatants were diluted 704 1:2 in 2% milk PBS (PBSM) and added to the microtiter plate for 1 hour at RT. Antigen-specific responses 705 were detected using Strep-Tactin®-HRP (iba lifesciences

#34028) as a substrate. The absorbance of each sample was measured at 450 nm as well as 630 nm. All 708 samples reported here were interrogated in the same plate

ForteBio) at 25˚C, shaking 712 at 1000 rpm. Kinetics assays were performed with Streptavidin (SA) Biosensors (ForteBio, Cat-No. 18-713 5019) with the following steps: (0) Hydration of SA Biosensors in 1X kinetics buffer for 30 mins

18-1105). (1) Baseline equilibration in 1X kinetics buffer for 60s

Baseline for 300s (4) Loading of antigen (SARS-CoV-2 S1 Avi-tag, Cat-No

C82E8) at 100 nM for 300 s (5) Quenching/Blocking of sensors in 50 µg/mL hIgG in 1X KB for 30s

Antibody binding: sensors immersed into purified antibody at 1.9 -500 nM in 1X KB for 350 s

Dissociation in 1X KB for 700 s (8) Regeneration of sensors. Curve fitting was performed using the 719

ForteBio Octet HTX data analysis software using a 1:1 model, and a baseline correction using a reference 720 sensor

Kinetic analysis of antibody pairs was performed using Streptavidin (SA) Biosensors 722 (ForteBio, Cat-No. 18-5019) using the "binning assay" definition with the following steps

SA Biosensors in 1X kinetics buffer for 30 mins (ForteBio, Cat-No. 18-1105). (1) Baseline equilibration 724 in 1X kinetics buffer for 60s

Loading of antigen (SARS-CoV-2 S1 Avi-tag, Cat-No. S1N-C82E8) at 41.6 nM for 45 s

SinoBiological, Cat-No. 10108-H08H) or Buffer (1X KB) at 166 nM for 800 s, followed by (7) short 728 dissociation step for 10 s and then (8) the association of Ab2 at 78 nM for 400 s. (9) Regeneration of 729 sensors. Curve fitting was performed using the ForteBio Octet HTX data analysis software using a 1:1 730 model

Convalescent plasma in the management of moderate covid-19 in adults in India: open label 805 phase II multicentre randomised controlled trial (PLACID Trial)

MiXCR: software for comprehensive adaptive immunity profiling

Techniques to Study Antigen-Specific B Cell 810 Responses

Potent neutralizing 813 antibodies from COVID-19 patients define multiple targets of vulnerability

Potent Neutralizing Antibodies against SARS-CoV-2 Identified by High-Throughput

Single-Cell Sequencing of Convalescent Patients' B Cells

The 819

SARS-CoV-2 Neutralizing Antibody LY-CoV555 in 823 Outpatients with Covid-19

Influenza Infection in Humans Induces Broadly Cross-826 Reactive and Protective Neuraminidase-Reactive Antibodies

Germline 828 immunoglobulin genes: Disease susceptibility genes hidden in plain sight?

The igraph software package for complex network research. 831 InterJournal Complex Syst

Large-scale sequence and structural 834 comparisons of human naive and antigen-experienced antibody repertoires

Convergent selection in antibody repertoires is revealed 838 by deep learning

Deep Sequencing of B Cell 841 Receptor Repertoires From COVID-19 Patients Reveals Strong Convergent Immune Signatures. 842 Front

Maturation of the 845 Human Immunoglobulin Heavy Chain Repertoire With Age

Integrated pipeline for the accelerated discovery 848 of antiviral antibody therapeutics

Learning the High-Dimensional Immunogenomic Features That Predict Public

Systems Analysis Reveals High Genetic and Antigen-Driven 854 Predetermination of Antibody Repertoires throughout B Cell Development

Complex heatmaps reveal patterns and correlations 857 in multidimensional genomic data

An alpaca 860 nanobody neutralizes SARS-CoV-2 by blocking receptor interaction

Studies in humanized mice and convalescent humans yield a 863 SARS-CoV-2 antibody cocktail

Both Neutralizing and Non-Neutralizing Human H7N9 866 Influenza Vaccine-Induced Monoclonal Antibodies Confer Protection

SARS-CoV-2 Cell Entry Depends on 870 ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

SARS-CoV-2 shedding and seroconversion among 874 passengers quarantined after disembarking a cruise ship: a case series

The neutralizing 877 antibody, LY-CoV555, protects against SARS-CoV-2 infection in nonhuman primates

Human neutralizing antibodies elicited by SARS-CoV-2 infection

Initial characterisation of ELISA assays and the 883 immune response of the clinically correlated SARS-CoV-2 biobank SERO-BL-COVID-19 884 collected during the pandemic onset in Switzerland

Simultaneous Treatment of Human Bronchial Epithelial Cells with Serine and Cysteine Protease 888

Inhibitors Prevents Severe Acute Respiratory Syndrome Coronavirus Entry

Stereotypic neutralizing VH antibodies against SARS-CoV-2 spike protein 892 receptor binding domain in COVID-19 patients and healthy individuals

Structure-guided multivalent 896 nanobodies block SARS-CoV-2 infection and suppress mutational escape

An improved single-chain Fab platform for 898 efficient display and recombinant expression

Longitudinal Isolation of Potent Near-901

Germline SARS-CoV-2-Neutralizing Antibodies from COVID-19 Patients

Potent neutralizing antibodies against multiple epitopes on SARS-904 CoV-2 spike

The stringdist Package for Approximate String Matching

Fast and easy isolation of CD27-positive human memory 909 B cells using EasySep Releasable RapidSpheres

High-throughput antibody engineering in mammalian cells by CRISPR/Cas9-mediated 912 homology-directed mutagenesis

Optimization of therapeutic antibodies by predicting 915 antigen specificity from antibody sequence via deep learning

Antibody Combination against SARS Coronavirus: Synergy and Coverage of Escape Mutants

Large-scale network analysis reveals 921 the sequence space architecture of antibody repertoires

B cells engineered to express pathogen-specific antibodies protect against infection

Single-cell sequencing reveals clonally 927 expanded plasma cells during chronic viral infection produce virus-specific and cross-reactive 928 antibodies

Phenotypic determinism 931 and stochasticity in antibody repertoires of clonally expanded plasma cells

RColorBrewer: ColorBrewer Palettes

Convergent Antibody Responses to SARS-CoV-2

The generation of 938 antibody-secreting plasma cells

Antibody discovery and engineering by enhanced CRISPR-Cas9 integration 941 of variable gene cassette libraries in mammalian cells

A functional BCR in human IgA and IgM plasma cells

Cross-neutralization of SARS-CoV-2 by a human 946 monoclonal SARS-CoV antibody

Immunogenomic 948 engineering of a plug-and-(dis)play hybridoma platform

Broad and potent activity against 951 SARS-like viruses by an engineered human monoclonal antibody

CoV-AbDab: the 953 coronavirus antibody database

Monoclonal antibodies isolated 956 without screening by analyzing the variable-gene repertoire of plasma cells

The discovery of plasma cells: An historical note

Convergent antibody responses to SARS-961

CoV-2 in convalescent individuals

Isolation of potent SARS-CoV-2 neutralizing antibodies and protection 964 from disease in a small animal model

Clinical Characterisation of Eleven Lateral Flow Assays for Detection of 967 COVID-19 Antibodies in a population

Viral epitope profiling of COVID-19 patients reveals cross-970 reactivity and correlates of severity

A Randomized Trial of 973 Convalescent Plasma in Covid-19 Severe Pneumonia

Potent binding of 2019 novel coronavirus spike protein by a SARS coronavirus-specific 976 human monoclonal antibody

pRESTO: a toolkit for processing high-throughput 979 sequencing raw reads of lymphocyte receptor repertoires

Transient expression of human antibodies in mammalian cells

Broad neutralization of SARS-related viruses 984 by human monoclonal antibodies

Immune cell profiling of COVID-19 patients in the recovery stage by single-cell 987 sequencing

Potent and broad HIV-neutralizing antibodies in 993 memory B cells and plasma

Rapid cloning of high-affinity human monoclonal antibodies 996 against influenza virus

A high-throughput platform for population reformatting 999 and mammalian expression of phage display libraries to enable functional screening as full-1000 length IgG

A highly 1002 conserved cryptic epitope in the receptor binding domains of SARS-CoV-2 and SARS-CoV

Structural basis of a shared antibody response to SARS-CoV-2

Cross-1006 reactive neutralization of SARS-CoV-2 by serum antibodies from recovered SARS patients and 1007 immunized animals

SARS-CoV-2 specific and neutralising antibodies from plasma cells of patients who recovered from COVID-19 using an integrated pipeline of single-cell sequencing and mammalian display. The described method herein enables the functional profiling of expanded plasma cells