key: cord-0904936-ppkuk9ow
authors: Emmenegger, M.; Fiedler, S.; Brugger, S. D.; Devenish, S. R. A.; Morgunov, A. S.; Ilsley, A.; Ricci, F.; Malik, A. Y.; Scheier, T.; Batkitar, L.; Madrigal, L.; Rossi, M.; Lynn, A. K.; Saleh, L.; von Eckardstein, A.; Knowles, T. P. J.; Aguzzi, A.
title: Both COVID-19 infection and vaccination induce high-affinity cross-clade responses to SARS-CoV-2 variants
date: 2022-04-10
journal: nan
DOI: 10.1101/2022.04.07.22273545
sha: 7705aed77394d75da41199793529fe14cf9ae589
doc_id: 904936
cord_uid: ppkuk9ow

The B.1.1.529 (omicron) variant has rapidly supplanted most other SARS-CoV-2 variants. Using microfluidics-based antibody affinity profiling (MAAP), we have recently shown that current therapeutic monoclonal antibodies exhibit a drastic loss of affinity against omicron. Here, we have characterized affinity and IgG concentration in the plasma of 39 individuals with multiple trajectories of SARS-CoV-2 infection and/or vaccination as well as in 2 subjects without vaccination or infection. Antibody affinity in patient plasma samples was similar against the wild-type, delta, and omicron variants (KA ranges: 122{+/-}155, 159{+/-}148, 211{+/-}307 M-1, respectively), indicating a surprisingly broad and mature cross-clade immune response. We then determined the antibody iso- and subtypes against multiple SARS-CoV-2 spike domains and nucleoprotein. Postinfectious and vaccinated subjects showed different profiles, with IgG3 (p = 0.002) but not IgG1, IgG2 or IgG4 subtypes against the spike ectodomain being more prominent in the former group. Lastly, we found that the ELISA titers against the wildtype, delta, and omicron RBD variants correlated linearly with measured IgG concentrations (R=0.72) but not with affinity (R=0.29). These findings suggest that the wild-type and delta spike proteins induce a polyclonal immune response capable of binding the omicron spike with similar affinity. Changes in titers were primarily driven by antibody concentration, suggesting that B-cell expansion, rather than affinity maturation, dominated the response after infection or vaccination.

The SARS-CoV-2 B.1.1.529 variant (omicron), considered a WHO variant of concern due to its high 42 transmission rate and its large number of mutations (Han et al., 2022) The antibody response against one or multiple epitopes, elicited upon infection or vaccination is 57 characterized by two properties: affinity and concentration. Those are fundamental, well-defined 58 biophysical parameters; however, until recently it has been challenging to measure them directly in 59 complex heterogeneous mixtures, like serum or plasma. We have recently employed Microfluidic 60 Antibody Affinity Profiling (MAAP) to simultaneously determine the affinity and concentration of 61 antibodies against WT RBD in convalescent sera (Schneider et al., 2022) , to study the antibody-based 62 inhibition of RBD-ACE2 interactions (Fiedler et al., 2021) and to understand memory re-activation and 63 cross-reactivity (Denninger et al., 2021) . We have also characterized the affinity of multiple 64 affinity) and 943.3 µM -1 (KD: 1 nM, i.e. comparatively high affinity). Thus, we observed an 116 approximately 250-fold affinity range within our cohort. The IgG concentrations varied between 3 and 117 49,074 nM (range: 4.2 log10) ( Fig. 2A) . The integrated 2D-density plot revealed a moderate left shift, 118

i.e. overall decreased KA, of antibodies to omicron while none of the variants formed separate clusters 119 Figure 3 . Correlation of affinity and IgG concentrations with clinically relevant parameters does not reveal clear differences between vaccinated and infected subgroups. A. 2D scatter plot with integrated density contours. All quantifiable data points reflecting KA (in M -1 ) and IgG concentration (in M) are plotted. 95% confidence intervals for each point are colored in light red. No distinct clusters were observed among patient groups infected/vaccinated (grey), infected/non-vaccinated (blue), non-infected/vaccinated (yellow), however, the REGN-COV-treated patients (red) clustered separately. B. The same groups as in (A) depicted in a boxplot. Statistical analysis is shown in the graph. The RBD variants are color-coded. C and D. No correlation between age (C) or sex (D) and KA x IgG concentration. E. While Kruskal-Wallis statistical testing indicates that the distributions are significantly different for different disease severities, pair-wise testing with Wilcoxon rank sum test does not result in significance. F. Trend towards increased KA x IgG concentration products in triple vaccinated individuals, without being statistically significant. A: Dotted lines represent the measurements of the same patient sample against different RBD variants. B, D-F: Kruskal-Wallis (KW) with post-hoc Wilcoxon rank sum test (WC) after Holm correction for multiple comparisons was used, with α = 0.01. C: The Pearson correlation coefficient was calculated. C-F: The patient groups are color-coded as in (A), however, the REGN-COV-treated patients were excluded from analyses.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 10, 2022. ; and mostly overlapped. 49% of samples could not be quantified in terms of KA or IgG concentration 120 for omicron (33% for wildtype and 36% for delta) (Fig. 2B) . However, the distributional differences of 121 quantifiable/non-quantifiable samples did not significantly differ (Fisher's exact test, α = 0.01) among 122 any of the antigens, for all samples (n=39), for those with an exclusively infection-and/or vaccine-123 induced antibody response (n=34), or for those treated with the REGN-COV cocktail (casirivimab and 124 imdevimab, n=5), although a trend towards increased evasion of antibody binding for omicron was 125 visible. 126

We then investigated KA (Fig. 2C) , IgG concentrations (Fig. 2D ) and the product of KA x IgG 127 concentration (Fig. 2E) for the three RBD variants. None of the variants displayed a statistically 128 significant deviation (Kruskal-Wallis with post-hoc Wilcoxon rank sum test after Holm correction for 129 multiple comparisons, with α = 0.01). Five patients who received the REGN-COV antibody cocktail 130 displayed the highest IgG concentrations measured, yet their affinities were in the range of the non-131 REGN-COV-treated patients ( Fig. 2A, 2C-E) . In sum, the antibody response following infection and/or 132 vaccination appears less susceptible to a drastic loss in binding against the omicron variant compared 133 to monoclonal antibodies. 134

Correlation of antibody fingerprints with clinically relevant parameters does not reveal 135 clear differences between vaccinated and infected subgroups. 136 We next characterized the affinity/concentration profiles in four patient groups: (1) infected/non-137 vaccinated; (2) non-infected/vaccinated; (3) infected/vaccinated; (4) treated with REGN-COV. We 138 studied the same profile as above, but we color-coded the data points according to the groups of 139 patients (Fig. 3A) . The patients treated with the REGN-COV cocktail clustered separately as expected, 140 whereas the density representations for vaccinated and/or infected patient groups were largely 141 overlapping. Statistical testing showed that the KA x IgG concentration product significantly differed in 142 patients treated with the REGN-COV cocktail versus all other groups (Wilcoxon rank sum test after 143

Holm correction, Fig. 3B) . However, the profiles observed following vaccination and/or infection did 144 not statistically differ among each other. In the following analyses, we have focused solely on those 145 groups with physiological antibody responses and excluded the REGN-COV-treated patients. 146

Correlations of KA x IgG concentration with age (Fig. 3C , the Pearson correlation coefficient R was 147 calculated for the three groups), sex (Fig. 3D ) or with disease severity (Fig. 3E) revealed heterogeneity 148 rather than marked differences. The Kruskal-Wallis test indicated significant distributional differences 149 as a function of disease severity (p-value=0.0084), yet, pair-wise testing with Wilcoxon rank sum test 150 did not result in any significantly changed group after correcting for multiple comparisons. While we 151 observed a trend towards increased KA x IgG concentration with a higher number of vaccinations ( Fig.  152 3F), the distributions did not significantly differ. In conclusion, multiple vaccinations and the 153 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273545 doi: medRxiv preprint Analysis of antibody subtypes, correlation with affinity and global feature profiling.

As infection, vaccination or parameters such as disease severity, age, and number of vaccinations did 158 not clearly correlate with antibody affinity or concentration, we aimed to obtain a more granular view 159 of the antibody compositions. Thus, we used TRABI (Emmenegger et al., 2020 (Emmenegger et al., , 2021 to deeply 160 characterize the antibody iso-and subtypes in our patient collective, to compare it to antibody affinity 161 and concentration, and to seek potential clinical or demographic correlates. 162

We first measured IgG, IgA, IgM, IgG1, IgG2, IgG3, and IgG4 antibodies against the SARS-CoV-2 WT 163 spike ectodomain (ECD), the WT S1 domain, the WT S2 domain, the WT RBD, the delta RBD, and the 164 omicron RBD variants as well as the nucleocapsid (NC) proteins and illustrate them in a heatmap (Fig.  165 4A, purple gradients). The antibody profile mainly revealed that the antibody response, in general, is 166 dominated by IgG, followed by IgA and much less so by IgM and that all IgG subtypes, except IgG2, 167 contributed to the IgG response against the spike-associated domains (Fig. S1A) . The presence of IgG 168 antibodies against the NC, the only protein employed here that is not intrinsically connected to the 169 spike ECD, is indicative of an infection, which was observed in almost all patients with clinically 170 characterized infection with SARS-CoV-2. For NC, the dominant IgG subtype was IgG3 (Fig. S1B) . 171

To validate our results, we repeated the IgG4 measurements against the entire collection of antigens, 172 with the same IgG4-specific secondary antibody (Fig. S1C ) and with the same clone but different 173 storage buffer sold by a different vendor (Fig. S1D) . We observed robust correlations using the same 174 CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; 10 -13 ) as well as the MAAP product (average R over all groups=0.71, p-value=5.4 x 10 -12 ) were well 190

represented by a linear model, for all the three variants. This finding suggested that the titers observed 191 in ELISA primarily reflect antibody concentrations rather than affinities in samples analyzed here. 192

We then reduced the dimensionality across all measured antibodies using principal component 193 analysis (PCA) and projected the linear combinations in two-dimensional space (Fig. 4C, and Fig. S4  194 for a granular view on the variable map). We used three representations using colors and shapes: (1) 195

The infection/vaccination cohorts, where we included patients treated with the REGN-COV antibodies, 196

(2) the number of vaccinations (excluding REGN-COV), (3) disease severity (excluding REGN-COV). 197

Regional clusters of specific antibody iso-and subtypes were annotated in black. Black shapes indicate 198 the mean points of a given group (indicated by color and shape). PCA suggested that while patients 199 with infection (infected/non-vaccinated; infection/vaccinated) clustered towards spike-associated 200 (annotated as S*) IgG3 as well as NC IgG, IgG1, IgG3, and IgA, patients with vaccination (non-201 infected/vaccinated: infected/vaccinated) clustered towards spike-associated IgG4. Spike-associated 202

IgG as well as IgG1 were between the two groups. Moreover, a higher number of vaccinations (two or 203 three) appeared to be linked to spike-associated IgG4 positivity, while fewer vaccinations (none or 204 one) clustered more closely to spike-associated IgG3 as well as to NC IgG, IgG1, IgG3, and IgA. A largely 205 similar pattern was observed with disease severity. No or mild disease clustered more in the region of 206 spike-associated IgG4 while more severe disease courses (hospitalization, oxygen supplementation, 207 ICU) assembled in the region of spike-associated IgG3 and the NC sub-and isotypes referred to above. 208

In sum, this representation evidenced an association between infection, more severe disease, absence 209 of vaccinations, and an IgG3 response against the spike-associated proteins. Conversely, the IgG4 210 response against spike-associated proteins was mainly characterized by vaccination, with higher 211 repeats of vaccinations, and a less severe disease course on average. 212

Based on the patterns identified above, we analyzed potential associations using different methods. 215

We first calculated the Pearson correlation coefficients for all antigens and antibody iso-and subtypes 216 and included additional parameters such as disease severity, immunosuppression, the REGN-COV 217 cocktail, the number of vaccinations, sex, and age, and plotted the significant correlations in a 218 correlogram (Fig. 5A) . Globally, the correlogram indicated a pronounced positive correlation within 219 the iso-or subtypes, which is expected as all domains are contained within the spike ECD, except the 220 NC antigen. IgG1 correlated almost perfectly with IgG, while IgM and IgG2 displayed only spurious 221 correlation with IgG. Disease severity correlated with reactivity against the NC protein, for IgG, IgA, 222 IgG1, and foremostly IgG3 but not for IgG2 or IgG4. A higher number of vaccinations showed negative 223 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; correlations with NC for IgA and IgG1. Sex and age did not display strong correlations in any direction. 224

The same correlogram with all correlations irrespectively of the significance level is shown in Fig. S5 . 225 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273545 doi: medRxiv preprint

We then looked into the antibody reactivity profiles that have shown relevance based on the PCA 226 representation and the correlogram. The p(EC50) values of seven Ig iso-and subtypes for the SARS-227

CoV-2 spike ECD were represented as a ridge plot. We focused only on the spike protein as a well-228 correlated surrogate for other spike domains (see Fig. S2D ), and omitted the NC protein from this 229 analysis. We compared the groups per antibody iso-and subtype looking at the entire distribution by 230 means of Kruskal-Wallis test with post-hoc Wilcoxon rank sum test corrected for multiple comparisons 231

with Holm (α=0.01). Exclusively significant distributional differences were annotated in the in the plot 232 (see Fig. 5B ). Kruskal-Wallis statistics indicated that the IgG responses of the non-infected/vaccinated 233 groups showed a significant change among each other (p-value=0.0024), however, the group-wise 234 testing by means of the Wilcoxon rank sum test did not reach statistical significance due to multiple 235 comparisons that were performed. Likewise, the IgG distributions differed among the degrees of 236 severity (p-value=0.0076, Kruskal-Wallis), but pair-wise testing resulted in non-significant differences 237

While the IgG4, but not overall IgG, IgG1, IgG2, or IgG3 distributions of patients after vaccination 238 showed a trend towards higher p(EC50) values, which were increased after three vaccinations, this 239 observation did not reach statistical significance. However, infection was associated with higher 240 p(EC50) values for IgG (p-value=0.00028, Wilcoxon rank sum test after Holm correction), and IgG3 (p-241 value=0.002, Wilcoxon) but not for IgG1, IgG2 or IgG4. 242 243 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273545 doi: medRxiv preprint

We reported that the monoclonal antibodies deployed against previous variants of SARS-CoV-2 show 245 drastically reduced affinity against the omicron variant (Fiedler et al., 2022) . Here, we found that delta-246 infected and/or vaccinated subjects develop plasma antibodies with similar affinities to all major 247 SARS-CoV-2 clades. Previous SARS-CoV-2 infection (alone or in combination with vaccination), but not 248 vaccination alone, was associated with higher overall anti-SARS-CoV-2 spike IgG, and IgG3 titers. Thus, 249 although antibody profiles following infection differs from that of vaccinated patients who did not 250 encounter the virus, pre-omicron responses showed impressive cross-clade affinities. 251

The high-affinity responses to the omicron spike protein contrast starkly with those of current We next explored the antibody profiles using a feature-based dimensionality reduction and 292 comprehensive correlograms. Both approaches pointed towards a slightly altered antibody profile 293 following vaccination or infection, with a stronger IgG3 response upon infection. However, the clusters 294 were relatively weak and potentially ambiguous in our dataset, although a previous investigation 295 derived similar conclusions regarding differences in the profiles between vaccinated and convalescent 296 individuals (Klingler et al., 2021) . Therefore, we focused on WT spike ECD, by looking at the distribution 297 in a statistical manner. We confirmed that infection alone or in combination with vaccination was 298 associated with higher p(EC50) values for IgG and IgG3 but not for IgG1, IgG2 or IgG4. 299

The limitations of our investigations reside in the number of patients enrolled in the study and the 300 vast number of variables reported, which may constrain the generalizability of results and conclusions. 301 Therefore, all variables underlying this study are available for further studies and for comparison with 302 future cohorts. On the other hand, our findings describing the antibody response of pre-omicron 303 convalescent or post-vaccination sera to the SARS-CoV-2 omicron variant are congruent with those 304 found by others with other methods, including viral neutralization and clinical observations. 305

In conclusion, we have investigated antibody affinity and concentration following infection and/or 306 vaccination in the presence of an antigenic drift. We found that the tolerance to the omicron drift was 307 surprisingly robust, whereas the currently approved therapeutic monoclonal antibodies lost much of 308 their affinity. The most plausible scenario is that antibodies are selected in vivo for immunodominant 309 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; spike domains that are invariant between clades of virus, whereas therapeutic monoclonals were 310 presumably selected in vitro for highest affinity but not for cross-clade protection. Ultimately, our 311 finding, along with others, suggests that the B-cell-mediated immunity, possibly concomitant with a 312 T-cell response, elicited upon infection and/or vaccination might be broad enough to confer a layer of 313 protection in the event of further waves of mutated SARS-CoV-2 variants. 314 315 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. 

For this study, we included residual pre-omicron heparin plasma samples from patients admitted to 318 the University Hospital Zurich, Zurich, Switzerland, whose blood was sent to the Institute of Clinical ELISA.

Serological ELISAs were carried out as previously described (Emmenegger et al., 2020 (Emmenegger et al., , 2021 with 346 minor adjustments. High-binding 1,536-well plates (Perkin-Elmer; SpectraPlate 1536 HB) were coated 347 with 3 µL of 1 µg/mL SARS-CoV-2 spike ECD, WT S1, WT S2, WT RBD, delta RBD, omicron RBD, or NC 348 protein in PBS using Fritz Gyger Certus Flex, incubated at 37 °C for 1 h in a ThermoFisher rotating plate 349 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273545 doi: medRxiv preprint incubator, and washed three times with PBS 0.1% Tween-20 (PBS-T) using Biotek El406. Plates were 350 blocked with 10 µL of 5% milk in PBS-T for 1.5 h using Biotek Multiflo FX peristaltic dispensing 351 technology. Samples inactivated with 1% Triton X-100 and 1% tributyl phosphate were diluted in 352 sample buffer (1% milk in PBS-T), and a serial dilution (range: 0.02 to 1.6 × 10 −4 ) was carried out 353 For quality testing, the same procedure was applied as above using the same clone (HP6025) of the 377 HRP-linked secondary antibody but from a different vendor, including a different storage buffer: 378 mouse anti-human IgG4; Invitrogen; A-10654 at 1:500 dilution. 379

Statistics and data analysis. See Table 2 . 402 

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 10, 2022. ; 

A serological assay to detect SARS-CoV-2 438 seroconversion in humans

Broadly neutralizing antibodies overcome SARS-CoV-2 442

Omicron antigenic shift

Omicron escapes the majority of existing SARS-CoV-2 neutralizing antibodies'

Omicron extensively but incompletely escapes Pfizer 448 BNT162b2 neutralization'

SARS-CoV-2 Omicron-B.1.1.529 leads to widespread 451 escape from neutralizing antibody responses

Reduced neutralisation of SARS-COV-2 Omicron-B.1.1.529 variant by 455 post-immunisation serum

Understanding the role of memory re-activation 458 and cross-reactivity in the defense against SARS-CoV-2', bioRxiv

mRNA-1273 and BNT162b2 mRNA vaccines have reduced 462 neutralizing activity against the SARS-CoV-2 omicron variant

Early peak and rapid decline 466 of SARS-CoV-2 seroprevalence in a Swiss metropolitan region', medRxiv

Anti-prothrombin 470 autoantibodies enriched after infection with SARS-CoV-2 and influenced by strength of antibody 471 response against SARS-CoV-2 proteins

Anti-SARS-CoV-2 antibodies elicited by 475 COVID-19 mRNA vaccine exhibit a unique glycosylation pattern

Serological fingerprints link antiviral activity of 479 therapeutic antibodies to affinity and concentration', bioRxiv

Antibody Affinity Governs the Inhibition 483 of SARS-CoV-2 Spike/ACE2 Binding in Patient Serum

2022) 486 'Receptor binding and complex structures of human ACE2 to spike RBD from omicron and delta 487 SARS-CoV-2

SARS-CoV-2 Antibodies Mediate Complement and Cellular 490 Driven Inflammation

Detection of Antibody Responses Against SARS-CoV-2 in 494 Plasma and Saliva From Vaccinated and Infected Individuals

IgG3 and IgM Identified as Key to SARS-CoV-2 Neutralization in 498 Convalescent Plasma Pools

Kinetic fingerprints differentiate the mechanisms 501 of action of anti-Aβ antibodies

Striking antibody evasion manifested by the Omicron variant of SARS-CoV-505 2'

Establishment of the WHO International 508 Standard and Reference Panel for anti-SARS-CoV-2 antibody', World Health Organization

Considerable escape of SARS-CoV-2 Omicron to 511 antibody neutralization'

Microfluidic characterisation reveals broad 515 range of SARS-CoV-2 antibody affinity in human plasma

Antibodies elicited by SARS-CoV-2 infection or mRNA 519 vaccines have reduced neutralizing activity against Beta and Omicron pseudoviruses

Rapid Generation of 523