Omicron BA.2 breakthrough infection enhances cross-neutralization of BA.2.12.1 and BA.4/BA.5

BNT162b2-vaccinated individuals after Omicron BA.1 breakthrough infection have strong serum neutralizing activity against Omicron BA.1, BA.2, and previous SARS-CoV-2 variants of concern (VOCs), yet less against the highly contagious Omicron sublineages BA.4 and BA.5 that have displaced previous variants. As the latter sublineages are derived from Omicron BA.2, we characterized serum neutralizing activity of COVID-19 mRNA vaccine triple-immunized individuals who experienced BA.2 breakthrough infection. We demonstrate that sera of these individuals have broadly neutralizing activity against previous VOCs as well as all tested Omicron sublineages, including BA.2 derived variants BA.2.12.1, BA.4/BA.5. Furthermore, applying antibody depletion we showed that neutralization of BA.2 and BA.4/BA.5 sublineages by BA.2 convalescent sera is driven to a significant extent by antibodies targeting the N-terminal domain (NTD) of the spike glycoprotein. However, neutralization by Omicron BA.1 convalescent sera depends exclusively on antibodies targeting the receptor binding domain (RBD). These findings suggest that exposure to Omicron BA.2, in contrast to BA.1 spike glycoprotein, triggers significant NTD specific recall responses in vaccinated individuals and thereby enhances the neutralization of BA.4/BA.5 sublineages. Given the current epidemiology with a predominance of BA.2 derived sublineages like BA.4/BA.5 and rapidly ongoing evolution, these findings helped to inform development of our Omicron BA.4/BA.5-adapted vaccine.

Antigenically, BA.2.12.1 exhibits high similarity to BA.2 but not BA.1, whereas BA.4 and BA.5 differ considerably from BA.2 and even more so from BA.1, in line with their genealogy (15,16). Whereas some amino acid changes in the RBD are shared between all Omicron sublineages, the alteration L452Q is only found in BA.2.12.1 and is the only residue that distinguishes its RBD from that of the BA.2 variant. The L452R and F486V alterations are BA.4/BA.5 specific, whereas S371F, T376A, D405N, and R408S are shared by BA.2 and its descendants BA.2.12.1 and BA.4/BA.5 but not BA.1 (fig. S1). These amino acid exchanges are associated with further escape from vaccine-induced neutralizing antibodies and therapeutic antibody drugs targeting the wildtype S glycoprotein (6,15,(17)(18)(19)(20). The NTDs of BA.2 and its descendants are antigenically closer to the wild-type strain and lack several amino acid changes, insertions, and deletions that occurred in BA.1 ( fig. S1). For instance, Δ143-145, L212I, or ins214EPE, which rendered the BA.1 variant resistant to a panel of NTD-directed monoclonal antibodies raised against the wild-type S glycoprotein, is not found in BA.2 and descendants (21,22).
Omicron BA.1 breakthrough infection of BNT162b2-vaccinated individuals augments broadly neutralizing activity against Omicron BA.1, BA.2, and previous VOCs at levels similar to those observed against SARS-CoV-2 wild type (10,23). BA.1 breakthrough infection of triple BNT162b2-vaccinated individuals induced a robust recall response, primarily expanding memory B cells against epitopes shared broadly among variants rather than inducing B cells specific to BA.1 only. Neutralization of the latest Omicron sublineages BA.4 and BA.5 was not enhanced, and geometric mean titers (GMTs) were rather comparable to those against the phylogenetically more distant SARS-CoV-1.
Given that Omicron BA.2 is more closely related to BA.4/BA.5 than to BA.1, we asked whether BA.2 breakthrough infection would shift cross-neutralization activity more toward these most recent Omicron sublineages. We compared the neutralization of different Omicron sublineages by serum samples from three different cohorts of individuals triple vaccinated with mRNA COVID-19 vaccines, namely, from individuals with no history of SARS-CoV-2 infection and individuals who experienced breakthrough infection with either BA.1 or BA.2. In addition, we characterized the contribution of serum antibodies targeting the S glycoprotein RBD versus the NTD to Omicron sublineage neutralization. Our data will increase current understanding of Omicron immune escape mechanisms and the effects of immunization on variant cross-neutralization and thus help guide further vaccine development.

Cohorts and sampling
This study investigated serum samples from three cohorts: from triple BNT162b2-vaccinated individuals who were SARS-CoV-2 naïve at the time of sampling (BNT162b2 3 ; n = 18), from individuals vaccinated with three doses of mRNA COVID-19 vaccine (BNT162b2/mRNA-1273 homologous or heterologous regimens) who subsequently had a breakthrough infection with Omicron at a time of BA.1 dominance (mRNA-Vax 3 + BA.1; n = 14), or from triple mRNA-vaccinated individuals with a breakthrough infection at a time of BA.2 dominance (mRNA-Vax 3 + BA.2; n = 19). For convalescent cohorts, relevant intervals between key events such as the most recent vaccination and infection are provided in Fig. 1 and tables S1 to S3. Sera were derived from the biosample collections of BNT162b2 vaccine trials and from a noninterventional study researching vaccinated patients who had experienced Omicron breakthrough infection. A subset of the samples included in this study had also been used in our previous investigation of the effect of BA.1 breakthrough infection on serum-neutralizing activity and memory B cell repertoire (10).
Because BA.4 and BA.5 share an identical S glycoprotein sequence, we refer to them as BA.4/5 in the context of the pVNT. In addition, we assayed SARS-CoV (herein referred to as SARS-CoV-1) to detect potential pan-Sarbecovirus-neutralizing activity (26). As an orthogonal test system, we used a live SARS-CoV-2 neutralization test (VNT) that analyzes neutralization during multicycle replication of authentic virus (SARS-CoV-2 wild-type strain and VOCs including BA.4, except Omicron BA.2.12.1) with the antibodies present during the entire test period.
To compare the cohorts with regard to neutralization breadth irrespective of the magnitude of antibody titers, we normalized the VOC pVN 50 GMTs against the wild-type strain. The ratios showed that BA.4/5 cross-neutralization was significantly (P < 0.05) stronger in mRNA-Vax 3 + BA.2 (GMT ratio of 0.37) as compared with mRNA-Vax 3 + BA.1 and BNT162b2 3 sera (GMT ratios of 0.18 and 0.17) (Fig. 2B). Cross-neutralization of Omicron BA.2.12.1 by mRNA-Vax 3 + BA.2 sera (GMT ratio of 0.53) was slightly stronger than by mRNA-Vax 3 + BA.1 sera (GMT ratio of 0.43) and even significantly (P < 0.01) stronger than by BNT162b2 3 sera (GMT ratio of 0.26). A separate analysis including only the BNT162b2-vaccinated individuals within those . For convalescent cohorts, relevant intervals between key events such as the most recent vaccination, SARS-CoV-2 infection, and serum isolation are indicated. All values specified as median-range. The age/sex composition of cohorts is further detailed in tables S1 to S3. Data for the reference cohorts BNT162b2 3 and mRNA-Vax 3 + BA.1 were previously published (10), except for newly generated BA.2.12.1 neutralization data. N/A, not applicable. Schematic was created with BioRender.com three cohorts confirmed that BA.2 breakthrough infection was associated with considerable BA.4/5 cross-neutralization (BA.4/5 to wild-type GMT ratio of 0.36), whereas after BA.1 breakthrough infection, pVN 50 GMTs against BA.4/5 were~6-fold lower than those against wild-type (i.e., GMT ratio of 0.17) ( fig. S2, A to C). Crossneutralization of BA.2 and BA.2.12.1 by sera of the BA.1 or BA.2 convalescents was stronger than that of BNT162b2 triple-vaccinated SARS-CoV-2 naïve individuals.
The authentic live SARS-CoV-2 virus neutralization assay provided VOC-neutralizing titers that strongly correlated with those from the pVNT assay ( fig. S3) and largely confirmed the major findings in Fig. 2. Again, in this assay, 50% virus neutralization (VN 50 ) GMT against Omicron BA.2 in BNT162b2 3 sera was strongly reduced compared with that against wild type (P < 0.0001), whereas sera from both convalescent groups exhibited strong neutralizing activity, with VN 50 GMTs comparable to those against the wild-type strain (Fig. 3A). Reduction of neutralizing activity against Omicron BA.4 was less pronounced in the BA.2 convalescent cohort as compared with BNT162b2 3 and mRNA-Vax 3 + BA.1 cohorts (VN 50 GMTs~2.5-fold as compared with~15-fold and 5fold lower than against the wild-type strain, respectively). In line with the pVNT data, magnitude-independent analyses via the calculated ratios of VOC VN 50 GMTs against the wild-type strain showed that BA.4 cross-neutralization was stronger in the mRNA-Vax 3 + BA. In aggregate, these data demonstrated that Omicron BA.2 breakthrough infection of vaccinated individuals is associated with broad neutralizing activity against all tested Omicron sublineages and previous SARS-CoV-2 VOCs. In particular, our data indicate that breakthrough infection with BA.2 is more effective (~2-fold higher cross neutralization) than that with BA.1 at refocusing neutralizing antibody responses toward the BA.4/BA.5 S glycoprotein.

Neutralization of Omicron BA.2 and BA.4/5 by sera of triple mRNA-vaccinated BA.2 convalescent individuals is mediated to a large extent by NTD-targeting antibodies
To dissect the role of serum antibodies binding either to the RBD or to the NTD of the S glycoprotein for neutralization of SARS-CoV-2 SARS-CoV-2 VOC pVN 50 GMTs normalized against the wild-type strain pVN 50 GMT (ratio VOC to wild type). Group geometric mean ratios with 95% confidence intervals are shown. The nonparametric Kruskal-Wallis test with Dunn's multiple comparisons correction was used to compare the VOC GMT ratios between cohorts. **P < 0.01 and *P < 0.05. Serum was tested in duplicate.
The depletion experiments removed >97% of all RBD binding antibodies and >74% of all NTD binding antibodies ( fig. S4B). Depleted sera were subsequently tested in pVNT assays. RBD-antibody depletion strongly diminished neutralizing activity against the wildtype strain in sera from all cohorts, whereas neutralizing activity was mostly retained (>80% remaining activity) upon depletion of NTD binding antibodies ( Fig. 4A and table S11). Neutralization of Omicron BA.1 was completely abrogated upon depletion of RBD binding antibodies and largely unaffected by NTD binding antibody depletion. For neutralization of BA.2, RBD antibody depletion almost completely abolished neutralizing activity of mRNA-Vax 3 + BA.1 sera (about 2% residual neutralization activity). The reduction of neutralizing titers for BNT162b2 3 and particularly mRNA-Vax 3 + BA.2 sera was less severe with~12 and~24% remaining neutralizing activity, respectively. In contrast, depletion of NTD binding antibodies did not considerably affect the neutralizing activity of BNT162b2 3 and mRNA-Vax 3 + BA.1 sera (~91 and~99% of undepleted control, respectively), whereas neutralizing activity of mRNA-Vax 3 + BA.2 sera was reduced to~50%. A similar pattern was seen after RBD antibody depletion for neutralization of BA.4/5, with strongly reduced neutralizing activity of mRNA-Vax 3 + BA.1 sera (~3% residual activity) versus less severe reductions for BNT162b2 3 and mRNA-Vax 3 + BA.2 sera (~20 and~26% remaining activity, respectively). Depletion of NTD binding antibodies had a larger impact for BA.4/5 neutralization compared with BA.2, with remaining neutralizing activity of BNT162b2 3 and mRNA-Vax 3 + BA.1 sera of~70 and~90% respectively, again with the strongest effect (~48% of undepleted control) on mRNA-Vax 3 + BA.2 sera. Serum concentrations of NTD-or RBD-targeting antibodies were not significantly different between the cohorts ( fig. S5), suggesting that differential neutralization activities may be attributable to differences in antibody potencies rather than serum antibody levels.
As an orthogonal approach, we assessed the neutralizing activity of sera from those three cohorts of vaccinated individuals against a pseudovirus harboring an engineered hybrid S glycoprotein consisting of the Omicron BA.1 N terminus including the NTD (amino acids 1 to 338) and the BA.4/5 C terminus including the RBD. The pVN 50 GMT against the Omicron BA.1-BA.4/5 hybrid pseudovirus in sera from BNT162b2 3 was moderately yet significantly GMTs normalized against the wild-type strain VN 50 GMT (ratio VOC to wild type). Group geometric mean ratios with 95% confidence intervals are shown. The nonparametric Kruskal-Wallis test with Dunn's multiple comparisons correction was used to compare the VOC GMT ratios between cohorts. ****P < 0.0001, **P < 0.01, and *P < 0.05. Serum was tested in duplicate.
In aggregate, the data obtained in both experiments indicate that across all these VOCs RBD binding antibodies make a major contribution to neutralization. Another key finding is that exposure to BA.1 (that differs substantially from previous VOCs in its NTD; fig. S1) boosts recall responses of vaccine-induced neutralizing antibodies that primarily bind the RBD, whereas exposure to BA.2 S glycoprotein (with an NTD more closely related to previous VOCs) can build on existing memory and elicits a considerable recall of NTDtargeting antibodies that in turn contributes substantially to the neutralization of BA.2 and BA.4/5.

DISCUSSION
Omicron BA.1 breakthrough infection in individuals vaccinated with mRNA vaccines BNT162b2 or mRNA-1273 or an inactivated virus vaccine boosts serum-neutralizing titers against VOCs, including BA.2 (10,15,23), but not against BA.2.12.1 or BA.4/ BA.5. The immune escape has been attributed to boosting of preexisting neutralizing antibody responses that recognize epitopes shared between the SARS-CoV-2 wild-type strain and Omicron BA.1 but are in part absent in BA.2.12.1, BA.4, and BA.5 due to alterations at key residues including L452Q/L452R and F486V (15). Here, we reported that BA.2 breakthrough infection was associated with broadly neutralizing activity including BA.2 and its descendants BA.2.12.1, BA.4, and BA.5. These findings are in agreement with recent publications (19,28) and suggest that the higher sequence similarity of BA.2 to BA.2.12.1 and BA.4/5 in the S glycoprotein RBD and the NTD drives more efficient cross-neutralization as compared with breakthrough infections with the antigenically more distant BA.1 variant. In particular, BA.1 breakthrough infection may not elicit a strong recall of NTD-specific memory B cells owing to the substantial alterations within the BA.1 NTD, given that breakthrough infection with heterologous SARS-CoV-2 strains primarily expands a memory B cell repertoire against conserved S glycoprotein epitopes (10,23). Our data obtained in antibody depletion and hybrid pseudovirus experiments showed that NTD binding antibodies had a substantial contribution to neutralizing activity against Omicron BA.4/5 in triple-vaccinated BA.2 convalescent sera, whereas neutralizing activity in BA.1 convalescent sera largely relied on RBD binding antibodies. This finding is consistent with the observation that NTD binding antibodies isolated from BA.2 breakthrough infected individuals do not neutralize BA.1 (29). Together, these findings extend our knowledge on how vaccinations and boosters with the current wild-type strain-based vaccines together with breakthrough infections with the various VOCs shape the immunity patterns within the population and are material to inform further vaccine development and adaptation in response to current and emerging VOCs.
Our findings are based on retrospective analyses of samples derived from different studies, using relatively small sample sizes and cohorts that are not fully aligned in terms of intervals between vaccine doses, intervals between the most recent vaccine dose and infection, and demographic characteristics such as age and sex of individuals. Although the SARS-CoV-2 naïve cohort was triple vaccinated with BNT162b2, and the Omicron breakthrough cohorts were triple vaccinated with BNT162b2 or mRNA-1273 or a heterologous regimen of the mRNA COVID-19 vaccines, key findings held true when only looking at a BNT162b2-vaccinated subset. Studies investigating long-lived plasma cell, memory B cell, and T cell immunity in cohorts with additional subjects could provide further insights into the mechanisms underlying the broad neutralizing activity associated with Omicron BA.2 breakthrough infection and corroborate our findings.
Notwithstanding the importance of vaccination with currently approved wild-type strain-based vaccines such as BNT162b2 that offer effective protection from severe disease by current VOCs including Omicron BA.1 and BA.2 (30,31), our findings highlight that consideration of rapidly evolving epidemiological landscapes and newly emerging SARS-CoV-2 variants is of high importance for guiding vaccine adaptation programs. Our data suggest that a vaccine adapted to the sequence of BA.2 or its descendants may enhance neutralization breadth against newly emerging VOCs, including the currently predominating BA.4/BA.5 sublineages.

VSV-SARS-CoV-2 S variant pseudovirus generation
In brief, human embryonic kidney 293T/17 monolayers [American Type Culture Collection (ATCC) CRL-11268] cultured in Dulbecco's modified Eagle's medium with GlutaMAX (Gibco) supplemented with 10% heat-inactivated fetal bovine serum (fetal bovine serum, Sigma-Aldrich) (referred to as medium) were transfected with Sanger sequencing-verified SARS-CoV-1 or variantspecific SARS-CoV-2 S expression plasmid with Lipofectamine LTX (Life Technologies) following the manufacturer's instructions. At 24 hours after transfection, the cells were infected at a multiplicity of infection of three with VSV-G-complemented VSVΔG vector. After incubation for 2 hours at 37°C with 7.5% CO 2 , cells were washed twice with phosphate-buffered saline (PBS) before medium supplemented with anti-VSV-G antibody (clone 8G5F11, Kerafast Inc.) was added to neutralize residual VSV-Gcomplemented input virus. VSV-SARS-CoV-2-S pseudotype-containing medium was harvested 20 hours after inoculation, passed through a 0.2-μm filter (Nalgene), and stored at −80°C. The pseudovirus batches were titrated on Vero 76 cells (ATCC, CRL-1587) cultured in medium. The relative luciferase units induced by a defined volume of a SARS-CoV-2 wild-type strain S glycoprotein pseudovirus reference batch previously described in Muik et al. (25), which corresponds to an infectious titer of 200 transducing units (TU) per ml, was used as a comparator. Input volumes for the SARS-CoV-2 variant pseudovirus batches were calculated to normalize the infectious titer based on the relative luciferase units relative to the reference.

Pseudovirus neutralization assay
Vero 76 cells were seeded in 96-well white, flat-bottom plates (Thermo Fisher Scientific) at 40,000 cells per well in medium 4 hours before the assay and cultured at 37°C with 7.5% CO 2 . Each individual serum was serially diluted twofold in medium, with the first dilution being 1:5 (SARS-CoV-2 naïve triple BNT162b2 vaccinated; dilution range of 1:5 to 1:5120) or 1:30 (triple vaccinated after subsequent Omicron BA.1 or BA.2 breakthrough infection; dilution range of 1:30 to 1:30,720). In the case of the SARS-CoV-1 pseudovirus assay, the serum of each individual was initially diluted 1:5 (dilution range of 1:5 to 1:5120). VSV-SARS-CoV-2-S/ VSV-SARS-CoV-1-S particles were diluted in medium to obtain 200 TU in the assay. Serum dilutions were mixed 1:1 with pseudovirus (n = 2 technical replicates per serum per pseudovirus) for 30 min at room temperature before being added to Vero 76 cell monolayers and incubated at 37°C with 7.5% CO 2 for 24 hours. Supernatants were removed, and the cells were lysed with luciferase reagent (Promega). Luminescence was recorded on a CLARIOstar Plus microplate reader (BMG Labtech), and neutralization titers were calculated as the reciprocal of the highest serum dilution that still resulted in a 50% reduction in luminescence. For depletion studies, resolution with regard to neutralization titers was increased to discriminate smaller than twofold differences on an individual serum level. Neutralization titers were determined by generating a four-parameter logistical fit of the percent neutralization at each serial serum dilution. The pVN 50 titer was reported as the interpolated reciprocal of the dilution yielding a 50% reduction in luminescence. Results for all pseudovirus neutralization experiments were expressed as GMTs of duplicates. If no neutralization was observed, then an arbitrary titer value of half of the limit of detection (LOD) was reported. Tables of the neutralization titers are provided (tables S4 to S6 and S11). SARS-CoV-2 wild-type strain, and Alpha, Beta, Delta, BA.1, BA.4/5 VOC, and SARS-CoV-1 pseudovirus-neutralizing GMTs for the SARS-CoV-2 naïve BNT162b2 triple-vaccinated cohort and the triple-vaccinated BA.1 convalescent cohort were previously reported in Quandt et al. (10). Only the BA.2.12.1 neutralization data were newly generated from serum samples for this study.
Depletion of RBD or NTD binding antibodies from human sera SARS-CoV-2 wild-type strain S glycoprotein RBD-and NTDcoupled magnetic beads (Acro Biosystems, catalog nos. MBS-K002 and MBS-K019; 40 μg of RBD/mg beads and 38 μg of NTD/mg beads, respectively) were prepared according to the manufacturer's instructions. Beads were resuspended in ultrapure water at 1 mg of beads per ml, and a magnet was used to collect and wash the beads with PBS. Beads were resuspended in serum to obtain 20 μg of RBD-or NTD-bait per 100 μl of serum. A mock depletion (undepleted control) was performed for each serum by adding 0.5 mg of biotin-saturated MyOne streptavidin T1 Dynabeads (Thermo Fisher Scientific, catalog no. 65601) per 100 μl of serum. Beads were incubated with human sera for 1 hour with gentle rotation. A magnet was used to separate bead-bound antibodies from the depleted supernatant. Depleted and undepleted sera were analyzed for cross-neutralization capacity using pseudovirus neutralization assays. Depletion efficacy for both RBD and NTD binding antibodies was determined by a multiplexed electrochemiluminescence immunoassay (Meso Scale Discovery, V-Plex SARS-CoV-2 panel 1 kit, catalog no. K15359U-2). A table of the neutralization titers is provided (table S11).

Analysis of SARS-CoV-2-specific serum IgG
Serum immunoglobulin G (IgG) levels directed against NTD domain, RBD domain, S1 spike, and S2 spike were determined using a V-Plex SARS-CoV-2 panel 1 kit (Meso Scale Diagnostics LLC) according to the manufacturer's protocol. Each individual serum was diluted 1:50,000 in diluent 100. The plate was analyzed on a MESO QuickPlex SQ 120 imager (Meso Scale Diagnostics).

Statistical analysis
The statistical method of aggregation used for the analysis of antibody titers is the geometric mean and for the ratio of SARS-CoV-2 VOC titer and wild-type strain titer is the geometric mean and the corresponding 95% confidence interval. The use of the geometric mean accounts for the nonnormal distribution of antibody titers, which spans several orders of magnitude. The Friedman test with Dunn's correction for multiple comparisons was used to conduct pairwise signed-rank tests of group geometric mean neutralizing antibody titers with a common control group. The Kruskal-Wallis test with Dunn's correction for multiple comparisons was used to conduct unpaired signed-rank tests of group GMT ratios with a common control group. The Wilcoxon matched-pairs signedrank test was used for pairwise comparison of the Omicron BA.4/ 5 neutralizing pVN 50 titers with titers against the Omicron BA.1-BA.4/5 hybrid pseudovirus. Spearman correlation was used to evaluate the monotonic relationship between nonnormally distributed datasets. For all conducted tests, P values of <0.05 were considered statistically significant. All statistical analyses were performed using GraphPad Prism software version 9.

Supplementary Materials
This PDF file includes: Figs. S1 to S6 Tables S1 to S11 Other Supplementary Material for this manuscript includes the following: MDAR Reproducibility Checklist View/request a protocol for this paper from Bio-protocol.