Switched and unswitched memory B cells detected during SARS-CoV-2 convalescence correlate with limited symptom duration

Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), the causative agent of the pandemic human respiratory illness COVID-19, is a global health emergency. While severe acute disease has been linked to an expansion of antibody-secreting plasmablasts, we sought to identify B cell responses that correlated with positive clinical outcomes in convalescent patients. We characterized the peripheral blood B cell immunophenotype and plasma antibody responses in 40 recovered non-hospitalized COVID-19 subjects that were enrolled as donors in a convalescent plasma treatment study. We observed a significant negative correlation between the frequency of peripheral blood memory B cells and the duration of symptoms for convalescent subjects. Memory B cell subsets in convalescent subjects were composed of classical CD24+ class-switched memory B cells, but also activated CD24-negative and natural unswitched CD27+ IgD+ IgM+ subsets. Memory B cell frequency was significantly correlated with both IgG1 and IgM responses to the SARS-CoV-2 spike protein receptor binding domain (RBD). IgM+ memory, but not switched memory, directly correlated with virus-specific antibody responses, and remained stable over time. Our findings suggest that the frequency of memory B cells is a critical indicator of disease resolution, and that IgM+ memory B cells play an important role in SARS-CoV-2 immunity.


Introduction 29
There have now been over 24 million reported cases of SARS-CoV-2, including at least 830,000 30 deaths worldwide (1). As the work to develop effective vaccines and therapies to control the 31 pandemic progresses, it is important to develop reliable approaches for assessing durable 32 immunological memory. Identification of a correlate of immunity to SARS-CoV-2 has been 33 challenging, as clinical presentation and serological profiles vary between patients. Rare SARS-34 CoV-2-specific antibodies with potent neutralizing capacity have been isolated from recovered 35 COVID-19 patients (2). Additionally, acute COVID-19 patients have been observed to have 36 perturbations of immune profiles, and have been grouped into three or more immunotype 37 clusters (3). On the basis of these findings, we focused on correlates of clinical outcomes in 38 convalescent plasma donors that could be inferred through cell-based assays. Recent studies 39 have correlated B cell responses in some individuals with immunity and protection (4). These analyses reveal substantial heterogeneity not just in the B cell immunophenotype 108 between subjects, but also within the B cell memory compartment itself. 109 All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Shorter symptom duration was correlated with increased switched and IgM + memory B 110 cell frequencies in convalescent subjects 111
While we observed diverse B cell subsets in convalescent subjects, it was still unclear which, if 112 any, of these subsets were correlated with clinical outcomes following symptomatic  infection. An association between frequencies of memory B cells and enhanced recovery 114 following COVID-19 pneumonia was reported (8). We therefore analyzed our B cell 115 immunophenotyping dataset for its correlation with self-reported symptoms in our convalescent 116 cohort. Our analysis revealed a significant negative correlation between the duration of COVID-117 19 symptoms and the frequency of memory B cells within the B cell compartment, as well as for 118 the IgM + memory B cell subset (Figure 4a-b). A similar trend was observed for switched 119 memory B cell frequency (Figure 4c). 120 To determine whether these observations were due solely to the time of sampling, we analyzed 121 the frequency of B cell subsets in convalescent subjects and the number of days between last 122 symptom and sample collection, as well as the number of days between symptom onset and 123 sampling. IgM + and switched memory B cell frequencies were stable or enhanced over time. We 124 failed to observe a correlation between IgM + or switched memory B cell frequency with the time 125 since the last reported symptom (Figure 4d-e), or the time since symptom onset (Figure 4g-h). 126 As was expected, longer symptom duration correlated with increased frequency of naïve and 127 transitional B cells, due to contraction of the plasmablast response (Figure 4g-k). Neither age 128 nor gender were observed to have a statistically significant influence on any of the clinical or 129 immunophenotypic parameters examined in this dataset (data not shown). 130 8 CD24-negative CD19 + plasmablasts (Figure 1g). Despite this trend, this subset was not 134 significantly correlated with the duration of COVID-19 symptoms at the time point at which we 135 sampled (Figure 4l). Plasmablast frequency among convalescent B cells did appear to wane 136 over the time since last symptom, and symptom onset, confirming the contraction of the acute 137 response (Figure 4f, i). Collectively, these data suggest that the presence of B cell memory is a 138 durable clinical correlate of shorter duration of COVID-19 disease. 139

Memory B cell frequency was correlated with anti-RBD antibody production 140
We next addressed whether the frequency of B cell memory in convalescent subjects correlated 141 with the generation of anti-spike receptor-binding domain (RBD) antibodies. The spike RBD is 142 thought to be required for SARS-CoV-2 binding and entry via the ACE2 receptor, and both 143 inhibitory and neutralizing anti-RBD antibodies have been identified in infected and recovered 144 subjects (4, 11). In seropositive convalescent subjects, IgG1 and IgM anti-spike RBD were 145 significantly correlated with CD24 + CD38-negative memory B cell frequency (Figure 5a-b, f). (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

baseline. 158
We next addressed whether the memory B cells were a stable population, or waned, as did the 159 plasmablast response. We re-sampled a subset of the convalescent cohort at least 3 months 160 after the initial visit (Figure 1, Supplemental Figure 1). Results from this longitudinal analysis 161 showed a contraction of the plasmablast response (Figure 6a). We also observed that memory 162 B cell frequencies and subset distribution were maintained or increased in most subjects 163 activation-associated B cell subsets, the frequency of T-bet + , CD11c + , DN, and activated B cells 167 did not change significantly over time for the cohort as a whole (Figure 6g-j). These trends were 168 not obviously impacted by the stage of convalescence when the first sample was obtained. 169 None of the B cell subsets analyzed exhibited statistically significant changes over the 3-month 170 period wherein we assessed our convalescent cohort, although the plasmablast compartment 171 contraction suggests a return to B cell homeostasis. These findings support a role for switched 172 and unswitched memory B cells in the maintenance of stable, durable SARS-CoV-2 immunity. 173 All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. . https://doi.org/10.1101/2020.09.04.20187724 doi: medRxiv preprint COVID-19 symptom duration. Moreover, IgM + memory correlated strongly with the anti-RBD 176 IgG1 antibody response. These data suggest that a protective memory response occurs in at 177 least some COVID-19 patients, preceded or accompanied by the generation of IgM + memory B 178 cells. We envision three possible explanations for these findings. First, it is possible that 179 memory B cells identified in some individuals were generated in response to a previous 180 coronavirus infection. Coronaviruses as a group likely generate cross-reactive B and T cells 181 responses (12). The observation that the anti-RBD IgG1 response was correlated with IgM + 182 memory cell frequency is paradoxical, however, given that IgM + memory cells don't produce 183 switched immunoglobulin. We propose that IgM + memory cells are generated in abundance 184 during coronavirus infections, and that some of these enter germinal centers and undergo class 185 switching following a related coronavirus infection, thereby contributing to enhanced IgG1 186 production. The capacity of IgM + memory cells to preferentially enter germinal centers upon 187 activation has been well-documented in mouse and human studies (13,14,15). This 188 characteristic versatility of IgM + memory cells could be advantageous for immunity to 189 pathogens, such as the coronaviruses, where infections with closely related strains often occur. 190 The lack of correlation between the frequency of resting memory B cells and CD11c + and/or T-191 bet + B cells in convalescent subjects was unexpected, given the pivotal role these molecules 192 play in type-1 B cell immunity. We hypothesize that these factors may play a key role during the 193 acute phase and during chronic viral infections, but are not essential during the convalescent 194 phase of SARS-CoV-2 infection. Additional prospective studies and kinetic analyses of 195 previously-exposed and naïve individuals will help to resolve this question. 196 All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. . https://doi.org/10.1101/2020.09.04.20187724 doi: medRxiv preprint cell expansion, is that naïve subjects whose B cells received more efficient T cell help during 198 primary infection generated a larger pool of memory B cells. This explanation is consistent with 199 the close relationship between memory and pathogen-specific antibody production we 200 observed. This explanation would suggest that T cells contributed to a better germinal center 201 response in some individuals. In contrast, subjects whose B cells received insufficient or has been suggested that B cell responses to severe acute COVID-19 disease have similar 220 characteristics to those observed in SLE (5). The variability in the penetrance of this phenotype 221 All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. . https://doi.org/10.1101/2020.09.04.20187724 doi: medRxiv preprint may be explained by the influence of gender and pre-existing autoimmunity. Future work using 222 animal models of COVID-19 may help to resolve these questions. 223 Our studies support other work that has reported that germinal center-derived memory B cells 224 are likely generated following SARS-CoV-2 infection, and that these memory B cells are a 225 durable correlate of an effective primary response (4, 8). These studies challenge early claims 226 of waning immunity shortly after SARS-CoV-2 infection. Our data also show that some 227 individuals can be identified as having better natural immunity. We also propose that both IgM + 228 and switched memory B cells may provide a good indication of vaccine efficacy, and that 229 individuals with large numbers of IgM + memory B cells may be better protected from future re-230 infection with homotypic or heterotypic infection. Our findings suggest that IgM + memory B cells 231 are central to the COVID-19 adaptive immune response, and highlight the need for prospective 232 SARS-CoV-2 studies to determine whether large memory B cell populations are a pre-existing 233 correlate of protection, or a durable measure of the antiviral immune response. 234 All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. Unit starting from March 2020 and is ongoing. Participants meeting eligibility criteria were adults 238 aged 18 or older who have tested positive for SARS-CoV-2 and are at least 14 days past their 239 first symptom. Exclusion criteria included the inability to give informed consent and/or an 240 inability to donate plasma or blood transfusion in the past. Study participants were interviewed 241 by study staff, and presented to the SUNY Upstate Clinical Research Unit for peripheral blood 242 collection. Information regarding symptoms, including dates of first and last symptom, was self-243 reported. Donors were questioned about acute symptoms such as fever, shortness of breath, 244 sore throat, cough that impacted activity, and fatigue that impacted activity. These indications 245 were used to calculate dates of symptoms retrospectively. Lingering symptoms such as loss of 246 taste and smell, mild cough or tickle in the throat, or lingering fatigue that did not impact their 247 daily activity were not considered part of the acute illness and therefore not included in the 248 length of illness. For donors reporting no symptoms, the date of positive RT-PCR test was used 249 for the start and stop date of symptoms. These subjects were not included in correlative 250 analysis of symptom duration. Healthy control subjects were adults aged 18 or older who denied 251 infection with or known exposure to SARS-CoV-2. Healthy controls were screened by anti-RBD 252 plasma ELISA to confirm negative exposure status. Sample size was determined based on 253 subject availability. All samples were de-identified following collection, and researchers 254 conducting assays were blinded to clinical data until final comparative analysis. 255 Blood sample processing and storage 256 PBMCs were obtained following gradient centrifugal separation of peripheral blood using Cell 257 Preparation Tubes (CPT) (BD) for 30 minutes at 1700 x g. Plasma was separated, aliquoted, 258 All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. . https://doi.org/10.1101/2020.09.04.20187724 doi: medRxiv preprint counting on a Coulter particle counter (Becton-Dickinson). PBMCs were either directly stained 260 for flow cytometry (initial visit samples), or frozen slowly to -80°C in FBS and DMSO for short-261 term storage (3-month visit samples). Flow cytometry panel was validated using a sample of 262 fresh vs. frozen PBMCs to ensure comparability in target detection. Plasma samples were heat-263 inactivated (56°C for 30 minutes) prior to use in assays. 264

Flow Cytometry 265
The following antibodies used for flow cytometry were obtained from BioLegend: CD21 (Bu32) Plasma samples were first heat-inactivated at 56°C for 30 mins before use in assays. 281 Recombinant Twin-Strep-tagged RBD protein was purified from 293T cells transfected with 282 All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. BioTek Synergy LX multi-mode plate reader. Area under the curve (AUC) analysis was 296 performed using peak identification set to a 10% minimum change from baseline. Plasma 297 ELISAs for total human immunoglobulin isotype quantitation were performed using the Human 298 Immunoglobulin Isotyping LEGENDplex 6-plex kit (BioLegend) according to the manufacturer's 299 instructions. Data were collected using a BD LSR II flow cytometer and analyzed using 300 LEGENDplex Data Analysis Software. 301

Statistical Analysis 302
Statistical analyses were performed using GraphPad Prism software (v8.4.3). Analysis of 303 correlation between flow cytometry, total serum immunoglobulin ELISA data, and continuous 304 clinical data was performed using Pearson correlation coefficients for data sets equal to or 305 larger than 35 values, or nonparametric Spearman's Rank correlation for data sets with fewer 306 than 35 values. p-values are two-tailed and 95% confidence intervals shown where noted in 307 All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. . https://doi.org/10.1101/2020.09.04.20187724 doi: medRxiv preprint figure legends. Statistical analysis of cell subset frequency between healthy and the total 308 convalescent donor cohort from flow cytometry assays was performed using unpaired 309 nonparametric Mann-Whitney test with two-tailed p-values and 95% confidence intervals. 310 Multiple comparison analysis between each convalescent subject subgroup was done with 311 Kruskal-Wallis test with Dunn's correction. Adjusted p value was used to determine family-wise 312 significance at alpha = 0.05. Statistical analysis of anti-RBD plasma ELISA data was performed 313 using area under the curve analysis with peak identification set to a 10% minimum change from 314 baseline. NS indicates a p value > 0.05, *p, < 0.05, **p, < 0.01, ***p, < 0.001, and ****p, 315 (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 preprint this version posted September 5, 2020. The authors have declared that no conflict of interest exists. 342

Data Availability 343
All data generated or analyzed during this study are included in this article and supplementary 344 data, or available from the corresponding author upon reasonable request. 345 All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. (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 preprint this version posted September 5, 2020. (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 preprint this version posted September 5, 2020. (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 preprint this version posted September 5, 2020. All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020.  IV.1.72

Late Convalescent
All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020.  (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 preprint this version posted September 5, 2020.   All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. . https://doi.org/10.1101/2020  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
(g) (f) All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020.  All rights reserved. No reuse allowed without permission.
(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 preprint this version posted September 5, 2020. All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
(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 preprint this version posted September 5, 2020.