Breakthrough infection elicits hypermutated IGHV3-53/3-66 public antibodies with broad and potent neutralizing activity against SARS-CoV-2 variants including the emerging EG.5 lineages

The rapid emergence of SARS-CoV-2 variants of concern (VOCs) calls for efforts to study broadly neutralizing antibodies elicited by infection or vaccination so as to inform the development of vaccines and antibody therapeutics with broad protection. Here, we identified two convalescents of breakthrough infection with relatively high neutralizing titers against all tested viruses. Among 50 spike-specific monoclonal antibodies (mAbs) cloned from their B cells, the top 6 neutralizing mAbs (KXD01-06) belong to previously defined IGHV3-53/3-66 public antibodies. Although most antibodies in this class are dramatically escaped by VOCs, KXD01-06 all exhibit broad neutralizing capacity, particularly KXD01-03, which neutralize SARS-CoV-2 from prototype to the emerging EG.5.1 and FL.1.5.1. Deep mutational scanning reveals that KXD01-06 can be escaped by current and prospective variants with mutations on D420, Y421, L455, F456, N460, A475 and N487. Genetic and functional analysis further indicates that the extent of somatic hypermutation is critical for the breadth of KXD01-06 and other IGHV3-53/3-66 public antibodies. Overall, the prevalence of broadly neutralizing IGHV3-53/3-66 public antibodies in these two convalescents provides rationale for novel vaccines based on this class of antibodies. Meanwhile, KXD01-06 can be developed as candidates of therapeutics against SARS-CoV-2 through further affinity maturation.

Spike (S) protein on virus surface mediates viral entry into host cells and is therefore the main target of vaccines and antibody therapeutics.Structurally, S protein is trimeric and each monomer consists of two functional subunits: S1 for engaging the receptor angiotensin converting enzyme 2 (ACE2) and S2 for driving fusion of viral and cellular membranes.S1 contains an amino-terminal (N-terminal) domain (NTD), a receptor-binding domain (RBD) and two carboxy-terminal (C-terminal) domains (CTD1 and CTD2).RBD further consists of a core structure and a receptor binding motif (RBM) that contacts with ACE2.S2 includes the N-terminal fusion peptide and its proximity region, heptad repeat 1 (HR1), central helix, stem helix, HR2, transmembrane region, and cytoplasmic tail.Although all the extracellular domains are susceptible to antibody binding, the majority of neutralizing antibodies target RBD with the rest recognizing NTD, S2, CTD or other epitopes [11,12].
Understanding the characteristics of broadly neutralizing antibodies (bnAbs) against SARS-CoV-2 will guide the rational design of vaccines and antibody therapeutics with broad protection.In this study, we identified two convalescents of breakthrough infection with relatively high neutralizing titers against all tested viruses including BQ and XBB lineages.Among 50 spike-specific monoclonal antibodies (mAbs) cloned from their B cells, the top 6 neutralizing mAbs (KXD01-06) belong to IGHV3-53/3-66 public antibodies and exhibit broad neutralizing capacity.Deep mutational scanning reveals that they target relatively conserved sites on RBD.Genetic and functional analysis indicates that the extent of somatic hypermutation is critical for their breadth.These findings provide rationale for the development of vaccines and antibody therapeutics based on IGHV3-53/3-66 public antibodies with broad and potent neutralizing activity.

Identification of individuals with broad neutralizing activity
To characterize bnAbs elicited by infection or vaccination, we collected peripheral blood samples from 11 donors (Table 1).Except Donor 11 who was not vaccinated with any SARS-CoV-2 vaccine, all the other donors were vaccinated with either two doses of mRNA vaccine (Donor 1 and 2), or two doses of inactivated vaccine (Donor 3-6), or three doses of inactivated vaccine (Donor 7-10).During the wave of infection in December 2022, all donors except Donor 10 were infected with SARS-CoV-2, probably BA.5 or BF.7 variants as they were main circulating strains at that time [24].
We measured plasma binding antibody titers against prototype (wild-type, WT), BA.4/5 and XBB.1.5spike by ELISA (Fig 1A and 1B).The antibody titers vary in a wide range with highest titers from Donor 2 and lowest titers from Donor 10 and 11.Moreover, antibody titers against WT, BA.4/5 and XBB.1.5spike gradually decrease, which correlate with the number of mutations on these spikes.We further measured plasma neutralizing antibody titers against 19 pseudoviruses including WT, Delta, BA.

Isolation of mAbs from two convalescents
As Donor 1 and 2 have higher neutralizing titers against XBB and XBB.1 than other donors, we chose their blood cell samples to isolate mAbs.With WT spike as bait, we sorted single antigen-specific memory B cells and plasmablasts (S1A Fig  .These data suggest that a large portion of neutralizing antibodies from these two convalescents belong to previously defined IGHV3-53/3-66 public antibodies.
We also tested the neutralizing activity of these mAbs against the same panel of pseudovirues used for plasma samples (Figs 3 and S2B).Surprisingly, KXD01-06 exhibit broad neutralizing capacity against SARS-CoV-2 viruses.Specifically, KXD01-03 are able to neutralize variants ranging from WT to emerging variants including EG.5, EG.5.1, FL.1.5 and FL.1.5.1, although they show reduced potency against later variants than earlier variants.The neutralizing breadth of KXD04-06 ranges from WT to XBB.1.16,which is consistent with their binding breadth.KXD09 is another bnAb with reduced neutralizing potency against multiple variants including Delta, BF.7, BQ.1, XBB lineages and the emerging variants.In contrast to other KXD antibodies, KXD07 and KXD08 have limited neutralizing breath and potency.Among the published mAbs, BD55-1205 and BD56-1854 show broad and potent neutralizing antibodies against all SARS-CoV-2 viruses, while the other antibodies have limited neutralizing breadth.Taken together, we identified 6 IGHV3-53/3-66 public antibodies with broad neutralizing capacity against SARS-CoV-2.

Mapping escape mutations of KXD01-06
To explore the molecular basis for broad neutralization and neutralization escape, we performed deep mutational scanning (DMS) to map escape mutations of KXD01-06 (Figs 4A and  S3).We first filtered out RBD mutants losing binding to ACE2 from two independently-constructed mutant libraries based on BA.4/5 RBD.Sequence analysis of yeasts after this round of sorting reveals that the distribution of mutations is rather random (S3A and S3B Fig) .Then we sorted RBD mutants with reduced binding to antibodies.In previously reported method [25,26], mutants post second sorting are directly processed to sequencing.Here we performed

PLOS PATHOGENS
Broadly neutralizing antibodies against SARS-CoV-2 an additional round of sorting to further remove mutants maintaining binding to antibodies or losing binding to ACE2 (Figs 4A and S3C).Sequence analysis of mutants post third sorting shows that escape mutations of KXD01-06 are limited to residues including D420, Y421, L455, F456, N460, A475 and N487.In addition, the escape mutations of each antibody display a different pattern regarding to frequency, implying that there are minor differences in their interactions with RBD (Fig 4B ).
We analyzed the variations on the 7 residues among sequences of WT SARS-CoV-2, representative SARS-CoV-2 variants and SARS-CoV-1 (Fig 4C ).No mutations are found on D420, Y421 and N487 and only SARS-CoV-1 shows variation on A475.F456L and N460K are common mutations shared by SARS-CoV-2 variants and SARS-CoV-1.Moreover, mutations on L455 are found in SARS-CoV-1 and the emerging HK.3.To track the variations on these residues, we collected spike sequences from 2019 December to 2023 October (S4 Fig) .So far, the mutation frequency on D420, Y421, A475 and N487 are still nearly 0. In contrast, N460K appeared in the second half of 2022 and reached 100% quickly.In the past several months, F456L has become a common mutation among the circulating strains along with the emergence of L455F.The accumulation of mutations on spike is consistent with the gradually reduced neutralizing potency of KXD01-06 against emerging variants.
To confirm the effects of mutations on those residues to neutralization evasion, we measured the neutralizing activity of KXD01-06 against BA.4/5 pseudoviruses with escape mutations (Fig 4D).Overall, KXD01-06 are largely or completely escaped by viruses carrying mutations identified by DMS.It is noteworthy that KXD01 maintains relatively potent neutralizing activity against virus with N460F, consistent with above neutralization profile of KXD01 against variants with N460K such as BA.2.75, BQ.1, XBB lineages.Taken together, these results demonstrate that KXD01-06 can be escaped by current and future variants with mutations on residues including D420, Y421, L455, F456, N460, A475 and N487.
To investigate the function of those mutations, we reverted the IGHV region of KXD01-06 to IGHV3-53/3-66 germline sequences.Then we compared the binding activity of mature antibodies and germline antibodies to WT and variant RBDs (Figs 5D and S5).Although each pair of antibodies bind WT RBD with similar affinities, the germline versions partially or almost completely lose binding to BA.2.75, BA.4/5, BQ.1, XBB.1.5 and EG.5.1 RBDs.Oppositely, we introduced common mutations including T28I, S31R and Y58F to IGHV region of P2C-1F11 as it lacks these mutations.Consistent with above results, these mutations dramatically increase the binding activity of P2C-1F11 to variant RBDs while not to WT RBD (Figs 5E and S5).Previous studies reported that Y58F can increase the binding of IGHV3-53/ 3-66 antibodies to WT RBD by 10-1000 fold [27,28].To figure out this inconsistency, we introduced T28I, S31R and Y58F to the germline version of P2C-1F11 (P2C-1F11-GL) individually or in combination (Figs 5F and S5).Consistent with previous findings, these mutations substantially increase the binding of P2C-1F11-GL to WT RBD.However, these mutations have limited or little effects on the binding of P2C-1F11-GL to BA.4/5, BA.2.75, BQ.1, XBB.1.5 and EG.5.1 RBDs, suggesting that more mutations beyond common mutations such as T28I, S31R and Y58F are required to confer P2C-1F11-GL high affinity to variant RBDs.We also analyzed the neutralizing breadth of P2C-1F11 with T28I/Y58F and T28I/S31R/Y58F.Consistent with a prior study [28], these mutations substantially improve the neutralizing activity of P2C-1F11 against Omicron variants, although the IC50s against BQ.1 and XBB lineages are still rather high.Taken together, these results elucidate that the extent of somatic hypermutation is critical for the neutralization breadth of IGHV3-53/3-66 public antibodies.
The development of bnAbs against highly variable pathogens is a key question for immunology and vaccinology.Currently, the development of HIV-1 bnAbs is most studied and a co-evolution model is proposed based on longitudinal analysis of the race between viruses and B cell lineages [29].According to this model, envelop glycoprotein (Env) from transmitted/ founder (T/F) virus leads the rare B cell precursors of bnAbs to undergo clonal expansion and somatic hypermutation.Subsequently, B cell lineages with desired mutations are selected by Envs from variants of T/F virus for further diversification.After iterative selections, B cell lineages gradually accumulate high levels of somatic hypermutations and eventually mature to produce bnAbs.Overall, the development of HIV-1 bnAbs is a rare event due to the following roadblocks.First, the precursor B cells of HIV-1 bnAbs are extremely rare in human B cell repertoire [30].Second, those precursor B cells generally show poor binding activity with most Envs [31,32].Third, HIV-1 bnAbs have 10-30% mutations and some mutations are intrinsically improbable [33,34].Compared with HIV-1 bnAbs, there are no such roadblocks in the development of broadly neutralizing IGHV3-53/3-66 antibodies against SARS-CoV-2.As the dominant antibodies targeting RBD, IGHV3-53/3-66 antibodies are prevalent in human B cell repertoire [35,36].Moreover, IGHV3-53/3-66 germline antibodies already have moderate to high affinities to SARS-CoV-2 [37,38].In addition, our study shows that the mutation level of broadly neutralizing IGHV3-53/3-66 antibodies is much lower than HIV-1 bnAbs and those mutations are commonly found in IGHV3-53/3-66 antibodies.These findings suggest that IGHV3-53/3-66 antibodies are promising targets for vaccines aiming to elicit bnAbs against SARS-CoV-2.
Regarding to mutations accumulated during antibody affinity maturation, we show that the mutations on IGHV3-53/3-66 substantially enhance the binding activity of KXD01-06 to variant RBDs instead of WT RBD, and thus we speculate that those mutations are selected by variant RBDs during breakthrough infection, which is consistent with co-evolution model of HIV-1 bnAbs.However, IGHV3-53/3-66 antibodies with broad neutralizing activity are also identified in WT vaccinees and WT convalescents as mentioned above, suggesting that variants are not essential for the development of neutralizing breadth across VOCs.Therefore, more studies are required to understand the role of antigen variation in the development of IGHV3-53/ 3-66 antibodies with broad neutralizing activity.
According to the classification of RBD-specific antibodies [11,39,40], IGHV3-53/3-66 antibodies mainly fall into two groups: RBD class 1 (RBS-A; RBD-2a) and RBD class 2 (RBS-B; RBD-2b).IGHV3-53/3-66 antibodies in RBD class 1 have short CDRH3 and bind to RBM using germline-encoded NY and SGGS motifs in CDRH1 and CDRH2.They only bind to RBD in the up conformation and neutralize SARS-CoV-2 by ACE2 blocking.In contrast, IGHV3-53/3-66 antibodies in RBD class 2 have long CDRH3 and can bind to RBD in both up and down conformation.They mainly contact with the RBD ridge and its nearby regions, and also neutralize SARS-CoV-2 by ACE2 blocking.According to this classification, KXD03, which has a long CDRH3 with 21 amino acids, is supposed to target different epitopes with the other 5 mAbs.Indeed, KXD01-06 share similar escape map with P2C-1F11, which is a representative of RBD class 1 antibody with escape mutations distributed on D420, Y421, L455, F456, N460, P463, Y473, A475 and N487 [26].On the other hand, although IGHV3-53/3-66 antibodies with broadly neutralizing activity have been widely reported and characterized [8,[19][20][21][22][23]28,[41][42][43], they generally use short CDRH3.Here we show that KXD03, which has a long CDRH3, is able to neutralize all tested SARS-CoV-2 viruses and is even more potent to variants with L455F than KXD01 and KXD02.Taken together, these observations suggest that KXD03 represents a unique type of IGHV3-53/3-66 antibodies with long CDRH3, which will be characterized structurally in future.
Regarding to the surveillance of SARS-CoV-2, emerging variants like EG.5, EG.5.1, FL.1.5 and FL.1.5.1 have F456L mutation, which falls into the escape maps of KXD01-06.Although these variants show similar resistance to plasma neutralization as XBB and XBB.1, they partially escape the neutralization of KXD01-03 and completely escape the neutralization of KXD04-06.Moreover, addition of L455F can further increases the neutralization resistance of EG.5.1 to KXD01-03.Consistent with studies published recently [9,10], these results demonstrate the evolution potential of SARS-CoV-2 and highlight the need to monitor current variants with those mutations.
We acknowledge that there are several potential limitations of this study.First, we only identified 50 spike-specific mAbs, which may be not enough to represent the antibody response in the two individuals.Second, we did not collect blood samples before the breakthrough infection.So we are not able to compare IGHV3-53/3-66 antibodies generated before and after the breakthrough infection, which is helpful to understand the development of broadly neutralizing IGHV3-53/3-66 antibodies.Finally, the profiles of neutralization and escape mutations suggests that there are some minor differences in the interactions of KXD01-06 with spike.In future studies, structure analysis can be performed to further elucidate these differences.
The rapid emergence of SARS-CoV-2 VOCs highlights the urgent need to develop vaccines and antibody therapeutics with broad protection against current and future SARS-CoV-2 variants.This study demonstrates that IGHV3-53/3-66 public antibodies have enormous potential to develop broad and potent neutralizing activity through antibody affinity maturation, which provides rationale for the development of novel vaccines and antibody therapeutics based on this class of antibodies.

Ethics statement
This study was approved by the Ethics Committee of Hefei Institutes of Physical Science, Chinese Academy of Sciences (Approval Number: YXLL-2023-47).All donors provided written informed consent for collection of information, analysis of plasma and PBMCs, and publication of data generated from their samples.

Human samples
Peripheral blood samples were collected from 11 donors (Table 1).Plasma and peripheral blood mononuclear cells (PBMCs) were separated from blood by Ficoll density gradient centrifugation.

PLOS PATHOGENS
Broadly neutralizing antibodies against SARS-CoV-2 to the wells and incubated at 37˚C for 1 hr.HRP-conjugated goat anti-human antibodies (Zen-bio, 550004; 1:5000 dilution) were added to the wells and incubated at 37˚C for 1 hr.TMB substrate (Sangon Biotech, E661007-0100) was added to the wells and incubated at room temperature for 5 mins.The reaction was stopped by TMB Stop Solution (Sangon Biotech, E661006-0500) and absorbance at 450 nm was measured.

Competition ELISA
WT RBD was coated onto 96-well ELISA plates (100 ng/well) and incubated at 4˚C overnight.After blocking with PBS containing 10% FBS, biotinylated ACE2, COV2-2196, COV2-2130 and P2C-1F11 were mixed with competing mAbs at 1:50 molar ratio.The mixtures were added to the wells and incubated at 37˚C for 1 hr.Streptavidin-HRP (GenScript, M00091; 1:5000 dilution) was added to the wells and incubated at 37˚C for 1 hr.TMB substrate (Sangon Biotech, E661007-0100) was added to the wells and incubated at room temperature for 5-10 mins.The reaction was stopped by TMB Stop Solution (Sangon Biotech, E661006-0500) and absorbance at 450 nm were measured.The percentages of signal decrease caused by competing mAbs were calculated.

Deep mutational scanning
Yeast libraries displaying BA.4/5 RBD mutants were kindly provided by Prof. Yunlong Cao at Peking University.Three rounds of FACS were performed to enrich RBD mutants losing binding to mAb but maintaining binding to ACE2.In the first round, yeasts were stained with ACE2 and ACE2-positive yeasts were sorted and expanded.In the second round, yeasts were stained with mAbs and mAb-negative yeasts were sorted and expanded.In the third round, yeasts were stained with ACE2 and mAbs simultaneously.ACE2-postive but mAb-negative yeasts were sorted.With the yeasts post the third sorting as template, PCR was performed to amplify RBD fragment from the plasmid.The PCR products were cloned to T vector and sequenced by Sanger-sequencing.The sequences are aligned with reference (BA.4/5) and mutations are identified.Then the frequency of mutations on each amino acid is analyzed.

Data analysis
Sequences of antibody heavy and light chain variable regions from single B cells were analyzed with IgBlast (https://www.ncbi.nlm.nih.gov/igblast/).Sequence alignment was performed either by MEGA or by the Muscle v5 algorithm.Sequence logos displaying mutation profiles were created with the Pandas, Bio, Matplotlib, Seaborn, and Logomaker packages in Python 3.8.16environment.FACS data were analyzed with Flowjo.ELISA and neutralizing data were analyzed and plotted with Graphpad Prism 8.

Fig 1 .
Fig 1. Characterization of plasma from SARS-CoV-2 convalescents or vaccinees.(A) Measurement of antibody titers against WT, BA.4/5 and XBB.1.5spikes by ELISA.Data are represented as the mean ± SD. (B) Summary of antibody titers.Statistical analysis was performed by two-tailed Wilcoxon matched-pairs signed rank test.*P < 0.05, **P < 0.01.(C) Measurement of neutralizing antibody titers against a panel of pseudovirues.Data are represented as the mean ± SD. (D) Summary of neutralizing antibody titers.For plasma with no neutralizing activity at 20-fold dilution, a number between 10-20 is given as titer to separate the curves.The numbers on top are geometric mean titers against the viruses.The titers against WT are compared with titers against other viruses.Statistical analysis was performed by two-tailed Wilcoxon matched-pairs signed rank test.*P < 0.05, **P < 0.01.All results are representatives of two independent experiments, in which duplicates are performed.https://doi.org/10.1371/journal.ppat.1011856.g001

Fig 2 .
Fig 2. Isolation and characterization of mAbs from Donor 1 and 2. (A) Summary of mAb screening by ELISA.Undiluted supernatant from 293T cells transfected with antibody-expressing vectors was tested against WT spike and RBD.OD450 was measured and positive clones were identified with 0.2 as cut-off.(B) Summary of 50 spike positive mAbs.The mAbs are ordered based on the neutralizing activity of 293T supernatant.OD450s and neutralization percentages are color-coded, with darker red indicates higher values.(C) Frequency of light chain V gene usage among 1208 IGHV3-53/3-66 public antibodies.The red arrows indicate V genes used by KXD01-06.https://doi.org/10.1371/journal.ppat.1011856.g002

Fig 3 .
Fig 3. Comparison of epitope, binding activity and neutralizing activity among KXD01-09 and reported antibodies.Epitopes were mapped by competition ELISA.ACE2, COV2-2196, COV2-2130 and P2C-1F11 were labeled with biotin.The percentages of their binding to RBD competed by KXD01-09 and reported antibodies were measured by ELISA.The values are color-coded with darker red indicates more competition.The binding activity against WT and variant RBDs was measured by ELISA.The neutralizing activity was measured by pseudoviruses.The highest antibody concentration used to determine EC50 and IC50 is 10 μg/ml.EC50s and IC50s were calculated by least squares fit.The values of EC50s and IC50s are color-coded with dark blue indicates lower values and dark red indicates higher values.All results are representatives of two independent experiments, in which duplicates are performed.https://doi.org/10.1371/journal.ppat.1011856.g003

Fig 4 .
Fig 4. Mapping of escape mutation.(A) FACS plots illustrating the process of library screening.In the first sorting, ACE-2-positive mutants were sorted.In the second sorting, mAb-negative mutants were sorted.In the third sorting, ACE2-positive but mAb-negative mutants were sorted.(B) Sequence analysis of mutants post the third sorting.The sequences were aligned with BA.4/5 RBD and then the mutations were identified.The left two columns represent mutation frequency across RBD (331-531).The right column represents mutation profiles on 420, 421, 455, 456, 460, 475 and 487.(C) Mutation profile of representative spike sequences.(D) Neutralizing activity of KXD01-06 against BA.4/5 with escape mutations.Data are represented as non-linear fit curves calculated by least squares fit.For (A-B), two independently constructed libraries are screened as two independent experiments.For (D), the results are representatives of two independent experiments, in which duplicates are performed.https://doi.org/10.1371/journal.ppat.1011856.g004