key: cord-0270669-b4h1jnko
authors: Pongracz, T.; Nouta, J.; Wang, W.; van Meijgaarden, K. E.; Linty, F.; Vidarsson, G.; Joosten, S. A.; Ottenhoff, T. H. M.; Hokke, C. H.; de Vries, J. J. C.; Arbous, S. M.; Roukens, A. H. E.; Wuhrer, M.
title: Immunoglobulin G1 Fc glycosylation as an early hallmark of severe COVID-19
date: 2021-11-20
journal: nan
DOI: 10.1101/2021.11.18.21266442
sha: 2cb5497c065e3c1ef04d87ee67a4fdcacd547c47
doc_id: 270669
cord_uid: b4h1jnko

Background Immunoglobulin G1 (IgG1) effector functions are impacted by the structure of fragment crystallizable (Fc) tail-linked N-glycans. Low fucosylation levels on severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike protein specific (anti-S) IgG1 has been described as a hallmark of severe coronavirus disease 2019 (COVID-19) and may lead to activation of macrophages via immune complexes thereby promoting inflammatory responses, altogether suggesting involvement of IgG1 Fc glycosylation modulated immune mechanisms in COVID-19. Methods In this prospective, observational single center cohort study, IgG1 Fc glycosylation was analyzed by liquid chromatography - mass spectrometry following affinity capturing from serial plasma samples of 159 SARS-CoV-2 infected patients. Findings At baseline close to disease onset, anti-S IgG1 glycosylation was highly skewed when compared to total plasma IgG1. A rapid, general reduction in glycosylation skewing was observed during the disease course. Low anti-S IgG1 galactosylation and sialylation as well as high bisection were early hallmarks of disease severity, whilst high galactosylation and sialylation and low bisection were found in patients with low disease severity. In line with these observations, anti-S IgG1 glycosylation correlated with various inflammatory markers. Interpretation Association of low galactosylation, sialylation as well as high bisection with disease severity suggests that Fc-glycan modulated interactions contribute to disease mechanism. Further studies are needed to understand how anti-S IgG1 glycosylation may contributes to disease mechanism and to evaluate its biomarker potential. Funding This project received funding from the European Commission's Horizon2020 research and innovation program for H2020-MSCA-ITN IMforFUTURE, under grant agreement number 721815.

Background 49

Immunoglobulin G1 (IgG1) effector functions are impacted by the structure of fragment crystallizable 50 (Fc) tail-linked N-glycans. Low fucosylation levels on severe acute respiratory syndrome coronavirus 51 2 (SARS-CoV-2) spike protein specific (anti-S) IgG1 has been described as a hallmark of severe 52 coronavirus disease 2019 (COVID-19) and may lead to activation of macrophages via immune 53 complexes thereby promoting inflammatory responses, altogether suggesting involvement of IgG1 Fc 54 glycosylation modulated immune mechanisms in Methods 56

In this prospective, observational single center cohort study, IgG1 Fc glycosylation was analyzed by 57 liquid chromatographymass spectrometry following affinity capturing from serial plasma samples 58 of 159 SARS-CoV-2 infected patients. 59

Findings 60

At baseline close to disease onset, anti-S IgG1 glycosylation was highly skewed when compared to 61 total plasma IgG1. A rapid, general reduction in glycosylation skewing was observed during the disease 62 course. Low anti-S IgG1 galactosylation and sialylation as well as high bisection were early hallmarks 63 of disease severity, whilst high galactosylation and sialylation and low bisection were found in patients 64 with low disease severity. In line with these observations, anti-S IgG1 glycosylation correlated with 65 various inflammatory markers. 66

Interpretation 67

Association of low galactosylation, sialylation as well as high bisection with disease severity suggests 68 that Fc-glycan modulated interactions contribute to disease mechanism. Further studies are needed to 69 understand how anti-S IgG1 glycosylation may contributes to disease mechanism and to evaluate its 70 biomarker potential. 71 3 Antibody glycosylation against the spike (S) protein of patients infected with severe acute respiratory 78 syndrome SARS-CoV-2 has been reported as a potentially important determinant of COVID-19 79 disease severity. Studies have hitherto focused on afucosylation, a modification on immunoglobulin 80 G1 (IgG) Fc-tail-linked N-glycans that enhances effector functions. Most of these studies featured 81 limited sample numbers or were imperfectly matched with respect to demographic and other important 82 confounding factors. Our lab has contributed to some of these studies, and we additionally searched 83

for research articles on PubMed and Google Scholar from January 2020 to October 2021. To date, only 84 two groups studied anti-S IgG1 glycosylation, which resulted in overall three publications found. 85 However, none of these groups found a severity marker between hospitalized non-ICU and ICU 86 patients or studied dynamic changes. Instead, exclusively fucosylation at the first available timepoint 87 has been associated with disease severity between severely ill inpatients and mild outpatients. 88

In this prospective, observational single center cohort study, we investigated the severity marker 90 potential of anti-S IgG1 glycosylation in severe and mild hospitalized COVID-19 cases, and correlated 91 these findings with numerous inflammation and clinical markers. Our study reveals low galactosylation 92 and sialylation as well as high bisection on anti-S IgG1 as early hallmarks of severe COVID-19, after 93 correction for age and sex effects. In line with these observations, anti-S IgG1 glycosylation correlated 94 with many inflammatory markers. As days since onset is one of the major confounders of anti-S IgG1 95 glycosylation due to its highly dynamic nature, we additionally confirmed our findings in time-matched 96 patient subgroups. We believe anti-S IgG1 glycosylation may be applicable for patient stratification 97 upon hospitalization. 98

1

Introduction 103

The current global coronavirus disease 19 (COVID-19) pandemic caused by the novel coronavirus 104 severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been leading to extensive 105 hospitalizations worldwide. 1 To date, more than 253 million infections and more than 5 million deaths 106 have been reported. 2 SARS-CoV-2 is an enveloped virus and its uptake by target cells in the respiratory 107 tract is mediated by the spike glycoprotein. 1 Recent reports have indicated that the high inter-individual variability in COVID-19 disease severity 3 120 may partly be explained by low Fc fucosylation of anti-SARS-CoV-2 spike protein-specific (anti-S) 121

IgG1. The lack of core fucose on these specific antibodies early on during disease points to their 122 potential proinflammatory role in severe illness. 6,13,14 Literature suggests, that in particular membrane-123  embedded foreign antigens, such as the SARS-CoV-2 spike protein, induce low fucosylated IgG1  124  responses, which in combination with high titers may lead to excessive macrophage activation and  125 drive COVID-19 associated pathology including acute respiratory distress syndrome. 6,13 126

Here, we study the dynamics of IgG1 Fc glycosylation and its association with clinical parameters in 127 a longitudinal cohort of 159 hospitalized COVID-19 patients, analyzing a total of 1300 longitudinal 128 patient samples. We report on the association of early anti-S IgG1 glycosylation signatures with disease 129 severity and various inflammatory markers, indicating its biomarker potential. 130 

Anti-S IgG was captured using a setup that resembles a conventional ELISA: IgGs were affinity-156 captured from plasma using recombinant trimerized spike-protein-coated Maxisorp NUNC-Immuno 157 plate (Thermo Fisher Scientific, Roskilde, Denmark), whereas total IgG was affinity-captured using 158 protein G Sepharose Fast Flow 4 beads, as described previously. 13,16 Antibodies were eluted using 100 159 mM formic acid and the samples were dried by vacuum centrifugation. Samples were reconstituted in 160 25 mM ammonium bicarbonate and subjected to tryptic cleavage, as described elsewhere. 16 Samples 161 5 belonging to a single patient were prepared and measured consecutively on the same plate, except for 162 follow-up samples after hospitalization period. On each plate, at least 3 Visucon-F plasma standards 163

(dating pre-COVID-19) and 3 blanks were included. 164

Glycopeptides were separated and detected using an Ultimate 3000 high-performance liquid 166 chromatography (HPLC) system (Thermo Fisher Scientific, Waltham, MA) hyphenated to an Impact 167 quadrupole time-of-flight mass spectrometer (Bruker Daltonics, Billerica, MA), as described. 16 168 2.5 Liquid chromatographmass spectrometry data processing 169

MzXML files were generated from raw liquid chromatographmass spectrometry (LC-MS) spectra. 170

An in-house developed software, LaCyTools was used for the alignment and targeted extraction of raw 171 data. 17 Alignment was performed based on the average retention time of minimum three abundant IgG1 172

glycoforms. The targeted extraction list included analytes of the 2 + and 3 + charge states and was based 173 on manual annotation of the mass spectra as well as on literature. 18,19 A pre-COVID-19 plasma pool 174 (Visucon-F) was measured in triplicate in each plate to assess method robustness and was as well used 175 as negative control. All spectra below the average intensity plus three times the standard deviation of 176 negative controls was excluded from further analysis. Signals were integrated by covering a minimum 177 of 95% of the area of the isotopic envelope of glycopeptide peaks. Inclusion of an analyte for the final 178 data analysis was based on quality criteria such as signal-to-noise (> 9), isotopic pattern quality (< 25% 179 deviation from the theoretical isotopic pattern), and mass error (within a ± 20 ppm range). Furthermore, 180

analytes that were present in at least 1 out of 4 anti-S IgG1 spectra (25%) were included in the final 181 analysis. 182

Circulating cytokine and chemokine levels were determined in serum using commercially available 184 bead based multiplex assays using the BioPLex 100 system for acquisition as previously described. 20 185

Standard curves were included in the kits and, in addition, a pooled serum sample of 4 hospital admitted 186 COVID-19 patients was included as internal reference in all assays. Quantitative detection of SARS-CoV-2 anti-S1/S2 IgG antibodies was performed using the DiaSorin 199 LIAISON platform. The CLIA assay consists of paramagnetic microparticles coated with distally 200 biotinylated S1 and S2 fragments of the viral surface spike protein. RLUs proportional to the sample's 201 anti-S1/S2 IgG levels are converted to AU/mL based on a standardized master curve. 202 All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint 6 Semi-quantitative detection of SARS-CoV-2 anti-RBD IgM antibodies was performed using the 203

Wantai IgM-ELISA (CE-IVD) kit (Sanbio). 24 Briefly, the IgM u-chain capture method was used to 204 detect IgM antibodies using a double-antigen sandwich immunoassay using mammalian cell-expressed 205 recombinant antigens containing the RBD of the spike protein of SARS-CoV-2 and the immobilized 206 and horseradish peroxidase-conjugated antigen. Sample/Cut-off index OD values of 1 and higher were 207 considered positive per the manufacturer's instructions. 208

Semi-quantitative detection of SARS-CoV-2 anti-S1 IgA antibodies was performed using the 209

Euroimmun IgA 2-step ELISA. 25 Ratio values of 1·1 and higher were considered positive per the 210 manufacturer's instructions. 211

The severity score is based on the 4C mortality score. 26 The 4C mortality score is a prediction score 213 calculated at admission, and the severity score calculated in our cohort represents the daily clinical 214 disease severity, and thus is dependent on parameters that can change over time. Therefore, the fixed 215 parameters of the 4C score were removed (i.e. age, sex at birth, number of comorbidities. Daily oxygen 216 flow for non-ICU patients (L/min) and p/f ratio (kPa) and FiO2 (%) for ICU patients were added to our 217 severity score (Table S2) . 218

Relative intensity of each glycopeptide species in the final analyte list was calculated by normalizing 220

to the sum of their total areas (Table S3) . Structurally similar glycopeptide species were used for the 221 calculation of derived traits fucosylation, bisection, galactosylation and sialylation (Table S4) . Anti-S 222 and total IgG1 glycosylation traits were compared using a Wilcoxon signed-rank test ( Figure 1 , Table  223 S5), while a Wilcoxon rank-sum test was used to compare non-ICU and ICU patients as well as severity 224 score groups ( in each statistical question (Table S5-7) . Spearman's correlation was used to explore associations 227 between glycosylation traits and age ( Figure S2 ), as well as between glycosylation traits and 228 inflammatory markers and titers ( Figure 5 , Table S8 ). To assess method repeatability, the inter-plate 229 variation for the 14 analytes included in the final analysis was calculated for the standards, which was 230 2·4%. All statistical analyses and visualizations were performed in R, version 4.1.0 (R Foundation for 231

Statistical Computing, Vienna, Austria) and RStudio, version 1.4.1717 (RStudio, Boston, MA). 232

The funders had no role in study design, data collection, data analysis, data interpretation, or writing 234 of the report. 235

Results 236

Both anti-S and total IgG1 glycosylation signature of 159 COVID-19 patients (39 female and 119 237 male) and corresponding timepoints were analyzed during their entire hospitalization period. The 238 patient demographics and the comprehensive cohort characteristics are presented in Table 1 and Table  239 S1, respectively. Follow-up samples after hospital discharge were available for 19 patients (Table S1 , 240 Figure S5 ). LC-MS was employed to analyze Fc glycosylation on the glycopeptide level after tryptic 241 digestion, which allowed the identification of 14 glycoforms. The found glycoforms were consistent 242 with previous reports on anti-S IgG1 glycosylation 13,14 , from which fucosylation, bisection, 243 All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint 7 galactosylation and sialylation were calculated (Table S3-4) . Overall, a total of 650 total IgG1 and 650 244

anti-S IgG1 glycosylation profiles were determined. 245 

The Fc glycosylation signatures of anti-S and total IgG1 were compared pairwise at hospitalization 249 with regard to fucosylation, bisection, galactosylation and sialylation ( Figure 1 , Table S5 ). 250

Fucosylation of anti-S was significantly lower than total (fold change (FC): 0·93; p-value: 3·4×10 -24 ) 251

( Figure 1A , Table S5 ). Notably, a prominently low anti-S fucosylation (<85%) was found for 56 252 patients, with a few patients showing levels as low as 66% ( Figure 1A) . 

Next, we explored the changes of glycosylation over time. Anti-S glycosylation was found to be highly 263 dynamic, but also total IgG1 glycosylation showed changes in the course of the disease ( Figure S6 ). 264

Both anti-S and total IgG1 galactosylation was found to be confounded by age and sex ( Figure S2 ) in 265 line with literature on IgG Fc glycosylation. 27 Therefore, delta (Δ) values were calculated by 266 subtracting total from anti-S IgG1 levels to eliminate the confounding effect, and used hereafter 267 ( Figure S2-3) . 268 All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. severity score (grey = NA). The shape displays whether a patient passed away (square) or was discharged alive (circle).

The black dashed line with a grey 95% confidence interval band is a cubic polynomial fit over the shown datapoints to 274 illustrate overall dynamics. Late timepoints and two outliers are shown in the Supplementary Material due to spatial 275 constraints ( Figure S4-5) , as well as anti-S and total IgG1 glycosylation dynamics ( Figure S6 ).

The longitudinal samples allowed us to establish the time course of ΔIgG1 glycosylation during 277 hospitalization, normalized for day of onset of symptoms (Figure 2 , Table S1 ). Interestingly, all 278 glycosylation traits showed a transient pattern for most patients, and were characterized by profound 279 dynamics, as illustrated by the timelines of individual patients (as indicated by differential line 280 coloring) and by the fit cubic polynomial line (Figure 2) . Fucosylation (Figure 2A ) and bisection 281 ( Figure 2C ) showed a rapid increase within days and weeks after onset of the disease, followed by a 282 plateau and approximation of the glycosylation patterns of total IgG1 ( Figure S6 ). In contrast, 283 galactosylation ( Figure 2B ) and sialylation ( Figure 2D ) quickly declined in the first days and weeks, 284

with the decrease continuing for a long period albeit at lower pace. At the moment of hospital discharge 285

anti-S galactosylation and sialylation were still slightly higher than total IgG1. Since 19 patients 286 returned for follow-up sampling after hospital discharge, we noted that for most, fucosylation and 287 bisection largely remained constant or slightly increased, whilst galactosylation and sialylation 288 continued to decrease since the last available timepoint ( Figure S5) . 289

To investigate whether Fc glycosylation was associated with intensive care unit (ICU) admission, 291 patients were stratified based on treatment need. This resulted in two groups: 1) patients who at some 292 point during hospitalization were admitted to the ICU, and 2) patients who were not enrolled to ICU 293 treatment at all (non-ICU) during hospitalization. ΔIgG1 glycosylation derived traits fucosylation, 294 bisection, galactosylation and sialylation of the above groups were compared both at time of 295 hospitalization and at the time point of their highest disease severity (Figure 3 , Table S6 ). 296 All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (Table S6) . *, ****: p-value < 0·05, 0·0001, respectively. Glycosylation 303 dynamics of ICU and non-ICU patients between day 10 and 25 are shown in Figure S8 . Figure 3D) . This difference was maintained or even more 308 pronounced at the time of highest disease severity (FC: 0·61, 0·26, 0·34; p-value: 1·9×10 -10 , 4·1×10 -309 12 , 3·4×10 -9 , for Δbisection, Δgalactosylation and Δsialylation, respectively) (Table S6) . Fucosylation 310 levels of the ICU group were higher at the time of highest disease severity (FC 0·62; p-value: 0·012), 311

but remained similar at the time of hospital admission ( Figure 3A) . To confirm that the observed 312 effects were not confounded by vast glycosylation dynamics, a subset of non-ICU and ICU patients 313

were created and compared, which resulted in comparable observations with regards to Δbisection, 314

Δgalactosylation and Δsialylation as shown above (Figure S7-8) . 315

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint Patients were stratified into three groups based on their severity score: 1) severity score between 0-5 317 (low severity), 2) 6-11 (intermediate severity) and 3) 12-17 (high severity). Similarly as before, ΔIgG1 318 glycosylation traits were compared both at time of hospitalization and at time of highest disease 319 severity (Figure 4, Table S7 ). 320 321 

Color indicates ICU (red) and non-ICU (blue) patients. A Wilcoxon rank-sum test was used to compare the different 326 severity score groups (Table S7) . *, **, ****: p-value < 0·05, 0·01, 0·0001, respectively.

ΔBisection was found to be increased in groups with increased disease severity (Figure 4C) , whereas 328

Δgalactosylation ( Figure 4B ) and Δsialylation ( Figure 4D ) patterns were found to be decreased with 329

increased disease severity at the time of hospitalization (Table S7) . These observations were largely 330 maintained at highest disease severity (Figure 4, Table S7 ). Higher fucosylation marked the time of 331 highest disease severity, but remained rather stable at the time of hospital admission between all groups 332 ( Figure 4A, Table S7 ). To confirm that the observed effects were not confounded due to profound 333 glycosylation dynamics, subsets of patients matched for the time since disease onset were compared, 334 which resulted in similar observations with regards to Δgalactosylation and Δsialylation as shown 335 12 above, whereas we could not exclude a potential confounding effect for the bisection signature, maybe 336 caused by swift glycosylation dynamics, low sample size, or the combination thereof ( Figure S9 ). 337

Apart from ICU admission and severity score, we tested acute respiratory syndrome, ventilation and 338 survival, and found Δbisection being higher for patients at baseline who passed away later ( Figure  339  S10) . 340

Multiple inflammatory mediators (in serum) and clinical parameters were measured for patients 342 enrolled during the first wave of the pandemic. These include members of the CXC, CCL and CX3C 343 chemokine families, cytokines and corresponding soluble receptors, acute phase proteins and other 344 mediators involved in the immune response as well as severity scores and anti-viral antibody titers. In 345 general, negative associations were found between Δgalactosylation and Δsialylation and positive 346 associations for Δbisection and Δfucosylation with inflammatory markers at baseline. One notable 347 exception was a strong negative correlation between anti-RBD IgM levels and Δbisection and 348

Δfucosylation at baseline and at highest severity, respectively. ΔSialylation associated negatively with 349 various chemokines, such as CCL24 (r = -0·45), CX3CL1 (r = -0·43), CCL25 (r = -0·34), certain 350 cytokines, such as IL-8 (r = -0·29), IFN-γ (r = -0·3) and several other variables ( Figure 5 , Table S8 ). 351 All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint 13 352 All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint 

Comparable, and largely overlapping negative associations were found for Δgalactosylation as for 356

Δsialylation: CCL24 (r = -0·55), CX3CL1 (r = -0·56), CCL25 (r = -0·41), IL-8 (r = -0·44), INF-γ (r = 357 -0·4) and TNF-β (r = -0·33). Conversely, Δbisection associated positively with IL-8 (r = 0·56), CCL25 358 (r = 0·52) and CX3CL1 (r = 0·56). Additionally, severity score negatively correlated with 359

Δgalactosylation (r = -0·55) and Δsialylation (r = -0·41) and positively with Δbisection (r = 0·46). 360

Positive associations were found between Δfucosylation and inflammatory markers, including CCL17 361 (r = 0·41) and IL-8 (r = 0·34). The above described baseline correlations were comparable to those at 362 the time of highest disease severity, but a vast body of associations were temporary ( Figure 5 , Table  363 S8). 364

Discussion 365

In this study, we analyzed total and anti-S IgG1 Fc glycosylation of 159 COVID-19 patients at different 366 timepoints during their clinical illness. Although several studies reported on the importance of (anti-S) 367

IgG1 Fc glycosylation and its association with disease severity in COVID-19 6,13,14,28,29 , this study 368 involves a large, single center cohort that confirms specific anti-S IgG1 glycosylation features as an 369 early hallmark of severe COVID-19 in an age-and sex-corrected, time-matched dataset at baseline, 370

and in the longitudinal dimension. 371 Afucosylated IgG1 B cell responses have recently been described to characterize immune reactions 372

against membrane-embedded antigens in general, and in particular against viral infections caused by 373 enveloped viruses such as COVID-19. 13 Foregoing studies showed that severe, hospitalized patients 374 exhibit a decreased anti-S IgG1 fucosylation as compared to mild, non-hospitalized patients. 6,13,14 375

Accordingly, we likewise observed proinflammatory, low-fucosylation signatures on anti-S as 376 compared to total IgG1, but found no difference in fucosylation comparing hospitalized ICU patients 377 versus hospitalized non-ICU patients, which is in line with a previous report on anti-SARS-CoV-2 378 receptor binding domain (anti-RBD) IgG1 fucosylation. 14 Therefore, based on the early existence of 379 these proinflammatory signatures in some of the patients, we hypothesize that low fucosylation -380 potentially even lower before measurable seroconversion, as hypothesized before 13on anti-S IgG1 381 may act as an early inflammatory signal that promotes the development of a more severe disease in 382 COVID-19 patients, resulting in hospital admission. However, disease severity between hospitalized 383 patients could not be further distinguished based on anti-S IgG1 fucosylation. Furthermore, hardly any 384 negative associations were found between anti-S IgG1 fucosylation and inflammatory markers in this 385 study, unlike in previous reports, where in vitro experiments demonstrated that the stimulation of 386 isolated macrophages with recombinant, glycoengineered anti-S or patient sera-derived low-fucose 387

IgG1 antibodies trigger higher proinflammatory cytokine release than those with normal fucose 388 levels. 6,13,14 However, high proinflammatory cytokine levels are not necessarily present in all severe 389 patients 30 , and this contrasting observation suggests a different regulation and/or the temporal 390 resolution of fucosylation and cytokine production dynamics in vivo. Additionally, beyond or in 391 combination with low anti-S IgG1 fucosylation a pre-existing risk factor may plays a role in COVID-392 19 disease severity, which hitherto remained unclear. 29 Of note, the anti-S and anti-RBD IgG1 Fc 393 glycosylation data were all determined from the circulation, and it is unclear to which extent this would 394 reflect the inflammatory pattern and glycosylation profile of anti-S antibodies in the lung. Our results 395 demonstrate that the proinflammatory fucosylation signature that is observed at the early time points 396 in the disease tends to fade with the course of the disease, which one may interpret as a shift towards a 397 more anti-inflammatory Fc glycosylation profile that is maintained over time. The absence of core 398 15 fucose is known to enhance a proinflammatory immune response by activating FcγRIII receptors on 399 monocytes, macrophages and NK cells. 10 Decreased fucosylation on specific IgG1 has been described 400

in HIV 13,31 and dengue fever 32 , as well as in alloimmune diseases. 33-37 However, whilst afucosylation 401 of specific IgG1 plays a protective role in HIV, it clearly marks high disease severity in dengue, 402 alloimmune diseases or COVID-19. 6,13,14 Furthermore, low total IgG1 fucosylation has been associated 403 with outcome of pediatric meningococcal sepsis indicating a systemic inflammation due to the potential 404 accumulation of airway infections during early childhood. 38 Even though the origin of low fucose IgG 405 responses is seemingly linked to antigen context and affect mostly specific antibodies 13 , the 406 mechanisms underlying the dynamics of antibody glycosylation remain elusive. 407

Besides afucosylation, a transient, decreased bisection was found on anti-S IgG1. Recent reports 408

suggest that severe COVID-19 patients present low levels of bisection both on total IgG (Fc and Fab 409 combined) 29 and anti-S IgG1 13 relative to mild cases. In contrast, no difference was found in anti-RBD 410

IgG1 bisection between ICU and non-ICU patients in age-and sex-matched patients 14 , albeit these 411 disease groups were largely comparable to the ones in our study. While bisection associated positively 412

with ICU admission, disease severity and survival in our study, it lacks functional relevance based on 413 our current understanding and has no effect on FcγRIII or C1q binding. 10,39 414

Elevated galactosylation and sialylation of anti-S IgG1 were associated with a less severe disease 415 course upon hospitalization, and no ICU admission. Similar observations were made in a previous 416 report, where severe COVID-19 was characterized by lower anti-S IgG1 galactosylation and sialylation 417 than mild COVID-19. 13 Interestingly, both anti-S and total IgG1 galactosylation and sialylation 418 decrease by advancing age. As Larsen et al. compared anti-S IgG1 galactosylation and sialylation of 419 imperfectly age matched patient groups without age correction, the disease and age effects remained 420

indiscernible. 13 We describe decreased anti-S IgG1 galactosylation in ICU patients as compared to 421 non-ICU patients, and analogously, markedly lower specific IgG1 galactosylation has been shown to 422 characterize the more severe, active phase of tuberculosis as compared to its latent counterpart. 40 Even 423 though more and more reports support that elevated levels of galactosylated IgG are associated with 424 the activation of the classical complement pathway 10,12,41 , galactosylation was associated with 425

increased disease severity in this study, possibly due to the fact that complement can contribute to the 426 increased inflammation both directly, and through inducing a chemotactic response through C5a, 427 thereby increasing cellular infiltration to inflamed tissues such as the lung. 42 Elevated sialylation levels 428 on anti-S IgG1 were associated with increased disease severity in the current report. Sialylation has 429 been broadly described as critical in mediating anti-inflammatory activity 43-45 , yet it remains to be 430 elucidated whether sialylated IgG exerts an anti-inflammatory effect in COVID-19. 431

Conclusions 432

This study established anti-S IgG1 bisection, galactosylation and sialylation as a unique combination 433 of features that associate with ICU admission and disease severity in hospitalized COVID-19 patients. 434

These features were additionally associated with markers of inflammation. Hence, we believe anti-S 435 IgG1 glycosylation may be applicable for patient stratification upon hospitalization. The glycosylation 436

profiles are highly dynamic, the drivers of which remain elusive and to be investigated in future studies. 437 6

Contributors 438 T. P.: Data (pre)processing, data curation, formal analysis, validation, investigation, visualization, 439 statistical analysis, data interpretation, conceptualization, writingoriginal draft preparation. J. N.: 440 sample preparation, data acquisition (IgG Fc glycosylation), W. W.: sample preparation (IgG Fc 441 All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ;

Clerc All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint sialylation levels (all in grey), as normalized to total IgG levels by subtracting total from anti-S IgG1 glycosylation levels. Baseline timepoints are shown. Shown in the inset are the Spearmen correlation coefficients (R) and pvalues, respectively. IgG1 bisection is known to increase, whereas galactosylation and sialylation are known to decrease upon aging. 1 Correction for the age confounding effect was performed by normalizing to total IgG levels, as illustrated by the weak and non-significant Spearman correlations and p- values, respectively (B, D, F, H Δsialylation. Correction for the age and sex confounding effect was performed as described above ( Figure S2 ). All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. , (B) galactosylation, (C) bisection and (D) sialylation, as shown during hospitalization (n=111). Line colours correspond to a single COVID-19 patient, whilst the colour gradient in the circles/squares indicates the corresponding severity score (grey = NA). The shape displays whether a patient passed away (square) or was discharged alive (circle). The black dashed line with a grey 95% confidence interval band is a cubic polynomial fit over the shown datapoints to illustrate overall dynamics. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. sialylation, as shown during the hospitalization period and follow-up (n=19). Line colours correspond to a single COVID-19 patient, whilst the colour gradient in the circles/squares indicates the corresponding severity score (grey = NA). The circle and shape display whether the timepoint corresponds to a follow-up sample (square) or to a sample taken during hospitalization (circle). The black dashed line with a grey 95% confidence interval band is a cubic polynomial fit over the shown datapoints to illustrate overall dynamics. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint Supplementary Figure 6 . Anti-S and total IgG1 glycosylation dynamics during the entire hospitalization period. The time-course of glycosylation derived traits (A) fucosylation, (B) bisection, (C) galactosylation and (D) sialylation as shown during the hospitalization period (n=111). Anti-S IgG1 dynamics are shown in the left facets, whereas total IgG1 dynamics in the right facets in each panel. Line colors correspond to a single COVID-19 patient, whilst the colour gradient in the circles/squares indicates the corresponding severity score (grey = NA). The circle and shape display whether the patient passed away (square) or was discharged alive (circle) from the hospital. The black dashed line with a grey 95% confidence interval band is a cubic polynomial fit over the shown datapoints to illustrate overall dynamics. Note that the confounding effect of age largely influences the observed bisection ( Figure S2C) , galactosylation ( Figure S2E) and (Figure S2G ) sialylation pattern, thereby has been corrected for age effects ( Figure S2 ). All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint the Spearmen correlation coefficients (R) and p-values, respectively. The red (ICU) and blue (non-ICU) lines are linear regression lines and the corresponding band indicates the 95% confidence interval. The dashed vertical lines indicate group means. Comparison of corresponding ΔIgG1 (B) fucosylation, (D) bisection, (F) galactosylation and (H) sialylation levels between ICU and non-ICU patients. All datapoints correspond to baseline samples (time of hospitalization). All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. D, H, L) . Bisection negatively associated with, survival, and sialylation negatively associated with ventilation. No other associations were found. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint

A pneumonia outbreak associated with a new coronavirus 464 of probable bat origin

An interactive web-based dashboard to track COVID-19 in real 466 time

Prevalence of Asymptomatic SARS-CoV-2 Infection : A Narrative 468 Review

Antibody responses to SARS-CoV-2 in patients with 470 COVID-19

Characteristics of SARS-CoV-2 and COVID-19

High titers and low fucosylation of early human anti-474

SARS-CoV-2 IgG promote inflammation by alveolar macrophages

Circulating multimeric immune complexes drive 476 immunopathology in COVID-19

Mouse and human FcR effector functions

Mechanisms of disease: The human N-glycome

Decoding the Human Immunoglobulin G-Glycan 482

Unique carbohydrate-carbohydrate interactions are required 485 for high affinity binding between FcgammaRIII and antibodies lacking core fucose

Fc Galactosylation Promotes Hexamerization of 488

Leading to Enhanced Classical Complement Activation

Afucosylated IgG characterizes enveloped viral 491 responses and correlates with COVID-19 severity

Proinflammatory IgG Fc structures in patients 493 with severe COVID-19

Potent neutralizing antibodies from 495 COVID-19 patients define multiple targets of vulnerability

A Targeted Liquid Chromatography-Mass

Spectrometry Data Processing Package for Relative Quantitation of Glycopeptides

High throughput isolation and glycosylation analysis of 502

IgG-variability and heritability of the IgG glycome in three isolated human populations

Immunoglobulin G glycosylation in aging and diseases

The datasets generated for this study are available on request from the corresponding author. 462

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint Supplementary Figure 9 . Patients in varying severity score groups 0-5 (red), 6-11 (green) and 12-17 (dark blue) and corresponding ΔIgG glycosylation derived traits in a "days since onset of symptoms" subset of patients to confirm that the observed differences (Figure 4) All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this this version posted November 20, 2021. ; https://doi.org/10.1101/2021.11.18.21266442 doi: medRxiv preprint