key: cord-0857768-ahevy38l
authors: Jeffery-Smith, Anna; Burton, Alice R; Lens, Sabela; Rees-Spear, Chloe; Patel, Monika; Gopal, Robin; Muir, Luke; Aiano, Felicity; Doores, Katie J; Chow, J. Yimmy; Ladhani, Shamez N; Zambon, Maria; McCoy, Laura E; Maini, Mala K
title: SARS-CoV-2-specific memory B cells can persist in the elderly despite loss of neutralising antibodies
date: 2021-05-31
journal: bioRxiv
DOI: 10.1101/2021.05.30.446322
sha: 850cca658cabf88f9b793455ae5060bae2e4be5e
doc_id: 857768
cord_uid: ahevy38l

Memory B cells (MBC) can provide a recall response able to supplement waning antibodies with an affinity-matured response better able to neutralise variant viruses. We studied a cohort of vulnerable elderly care home residents and younger staff, a high proportion of whom had lost neutralising antibodies (nAb), to investigate their reserve immunity from SARS-CoV-2-specific MBC. Class-switched spike and RBD-tetramer-binding MBC with a classical phenotype persisted five months post-mild/asymptomatic SARS-CoV-2 infection, irrespective of age. Spike/RBD-specific MBC remained detectable in the majority who had lost nAb, although at lower frequencies and with a reduced IgG/IgA isotype ratio. Functional spike/S1/RBD-specific recall was also detectable by ELISpot in some who had lost nAb, but was significantly impaired in the elderly, particularly to RBD. Our findings demonstrate persistence of SARS-CoV-2-specific MBC beyond loss of nAb, but highlight the need for careful monitoring of functional defects in RBD-specific B cell immunity in the elderly. One sentence summary Circulating class-switched spike and RBD-specific memory B cells can outlast detectable neutralising antibodies but are functionally constrained in the elderly.

The human coronavirus SARS-CoV-2 has had a particularly devastating impact on the elderly, 45 who are at much greater risk of morbidity and mortality. 1,2 Understanding the nature of a 46 successful immune response in those who have avoided these outcomes and cleared SARS-47

CoV-2 after a mild infection, despite advanced age, is key to protecting this vulnerable group 48 in the future. Whether older survivors of SARS-CoV-2 infection are able to mount robust and 49 durable responses with the potential to provide long-term protection from reinfection, and 50 from emerging viral variants, remains to be understood. Insights into the strengths and 51 limitations of the immune response in those who have had a successful outcome of natural 52 infection can inform the future optimisation of vaccines. It is also crucial to understand the 53 nature of the immune protection afforded to previously infected individuals whilst they await 54 vaccination, especially with the ongoing delays in rollout and the lag in provision to low and 55 middle-income countries. 56 57 Antibodies, in particular the neutralising fraction, provide a vital frontline defence to achieve 58 protective immunity against viruses. An initial waning of antibody titres is typically seen after 59 resolution of an acute viral infection. 3, 4 In the case of some viruses, long-lived plasma cells are 60 then able to maintain antibodies for decades. 5-7 By contrast, in the months following infection 61 with other viruses, including human coronaviruses like SARS-CoV-2, neutralising antibodies 62 continue to wane and can drop below the threshold of detection in a proportion of 63 individuals. 3, [8] [9] [10] [11] [12] [13] Even if antibodies are maintained, they may fail to provide sufficient 64 functional flexibility to cross-recognise viral variants. 14-16 However, antibody responses of 65 inadequate titre or unable to cross-recognise variants can be compensated by a second line 66 of defence provided by antigen-specific memory B cells (MBC), that are poised to react rapidly 67 upon pathogen re-encounter. [17] [18] [19] Not only can MBC provide a faster response on re-exposure 68 to the virus, they are also able to diversify in the face of a mutating virus, resulting in more 69 potent, affinity-matured antibody response and enhanced resistance to viral mutations. 9,20 70 71 In this study, we therefore analysed whether MBC develop in elderly subjects following the 72 resolution of SARS-CoV-2 infection and whether they can maintain functionality once 73 neutralising antibodies (nAb) have waned. To address these questions, we studied elderly 74 residents that had recovered from SARS-CoV-2 infection following outbreaks in three care 75 homes in the UK, who had mild or asymptomatic infection, a substantial proportion of whom 76 lost detectable nAb by five months after outbreak resolution. MBC were compared between 77 the elderly care home residents and younger staff to assess the impact of ageing. We 78 identified MBC specific for SARS-CoV-2 spike and RBD that persisted when serum nAb had 79 completely waned. Their frequency, phenotype, isotype and function were analysed 80 according to age and/or nAb loss, to inform the assessment and boosting of durable immunity 81 in the elderly. 82 83

antibodies 86

To study the role of MBC, we obtained PBMC from a subset (n=32) of a large cohort who 87 survived COVID-19 with mild/asymptomatic infection after outbreaks in three care homes in 88

April 2020 (Methods and Table S1 ). 21, 22 The care home cohort subset was selected to have a 89 wide range of nAb titres detectable against live virus at the first sampling timepoint (T1, May 90 2020, Fig.1 ). By end September 2020 (T2, five months), 32% of all participants sampled had 91 stable or increasing nAb to live virus. In contrast, 22% had declining titres, and 38% had lost 92 detectable nAb (Fig.1a,b) . 93 94 To compare MBC frequencies in those who had maintained or lost nAb, we stained PBMC 95 with SARS-CoV-2 spike trimer tetramers, made by pre-incubating recombinant biotinylated 96 trimeric spike protein with fluorescently-conjugated streptavadin. 15 Dual staining with spike 97 tetramers with two distinct fluorochromes was used to enhance the discrimination of true 98 antigen-specific MBC (Fig.1c) , as described previously. 23-25 Frequencies of antigen-specific 99 responses were calculated within the memory fraction of B cells (CD19 + CD20 + excluding IgD + , 100 CD38 hi and CD21 + CD27naïve fractions, gating strategy in Fig.S1a , as previously described 26 ). 101

A threshold for background non-specific staining was set at mean+ 2SD of staining seen in an 102 uninfected control cohort derived from the same care homes (seronegative at both time 103 points, Table S1 ) and from pre-pandemic healthy donor samples (Fig.S1b) . 104 105 Spike-specific MBC were detectable in 28 of the 32 tested 5 months post-infection (Fig.1d) . 106

The frequency of spike-specific MBC was reduced in those who had lost nAb compared to 107 those in whom they were still detectable (Fig.1e) . Of note, however, most of those (85%) who 108 had lost detectable nAb still had some persistent spike-specific MBC, a comparable 109 proportion to that in the group maintaining nAb (Fig.1f) . The frequency of spike-specific MBC 110 correlated significantly with the strength of the nAb response (nAb titre to live virus) at 5 111 months (Fig.1g) ; however, there was partial discordance due to detection of spike-specific 112 MBC in most individuals with no nAb (dotted box, Fig.1g) . 113 114 Next, we analysed the MBC response specifically directed against RBD since this is the region 115 within spike which many SARS-CoV-2-specific nAb target. 15,27-29 RBD-specific MBC were 116 identified by gating on dual spike tetramer-staining populations that also stained with a 117 tetramer formed from recombinant biotinylated RBD protein pre-incubated with 118 fluorescently-conjugated streptavidin (Fig.1h) . RBD-specific responses were detectable in 26 119 of the 28 with sufficient magnitude spike-specific MBC responses (>20 dual-spike+ cells) to 120 allow analysis of the RBD-co-staining cells (Fig.1i) . The frequency of RBD-specific MBC was 121 significantly reduced in the group who had lost nAb compared to those with stable (or waning 122 but still detectable) nAb (Fig.1i) . However, as noted with spike-specific MBC, some RBD-123 specific MBC remained detectable in most of the cohort, irrespective of whether or not they 124 had lost nAb (Fig.1j) . Overall, the magnitude of RBD-specific MBC correlated with nAb titres, 125 although again there was some discordance due to RBD MBC in those who had lost nAb 126 (dotted box in Fig.1k ). Importantly, both the RBD positive and RBD negative components of 127 the spike-specific B cell response significantly correlated with nAb titres, though slightly more 128 robustly for the RBD positive subset (Fig.1k, Fig.S1c ). This highlights the importance of the 129 RBD as the major target for neutralising antibodies, whilst also underscoring the contribution 130 of antibodies targeting regions outside of the RBD (for example the NTD of the spike 131 protein 15,29-31 ) to the neutralising antibody response, at the 5 month timepoint in this cohort. 132 133 These data therefore revealed the persistence of detectable, albeit reduced, MBC specific for 134 both spike and RBD in most people whose nAb titres against live virus had fallen below the 135 threshold of detection. Thus, loss of detectable nAb 5 months after asymptomatic/mild 136 infection is frequently compensated by the presence of a memory response primed to 137 respond upon re-exposure. 138

Comparable persistence of spike and RBD-specific MBC in elderly care home residents and 141 younger staff 142

The care home cohort was constructed to sample two comparator groups: elderly residents 143 (median age 86yrs, range 66-96) and a control group of younger staff (median age 56yrs, 144 range 41-65). Five months after asymptomatic/mild infection, similar proportions of staff and 145 residents had lost detectable nAb (Fig.2a) , and those who maintained them had similar titres 146 ( Fig.2b) . We postulated that there may, nevertheless, be a defect in the maintenance of 147 spike/RBD-specific MBC in the elderly compared to younger age group. However, spike-148 specific MBC were maintained at similar frequencies and in comparable proportions of the 149 elderly residents and younger staff (Fig.2c,d) . There were no clear trends for spike-specific 150 MBC to decrease with increasing age, even in residents in their nineties (Fig.2e) . 151 152 Similarly, RBD-specific MBC were equally well-maintained in the residents and staff (Fig.2f,g) , 153

with no decline in their frequencies (as a fraction of total MBC) with increasing age (Fig.2h) . 154 RBD-specific MBC comprised a variable proportion of the total spike-specific MBC response 155 (4.6 to 41.0%; median 24.0%), the remainder representing B cells targeting non-RBD regions 156 of spike. The proportions of RBD and non-RBD-binding spike-specific MBC again showed no 157 changes with age ( Fig.2i) . 158 159 160

Having identified and quantified antigen-specific B cells with tetramer staining, we were able 162 to apply high-dimensional multiparameter flow cytometry to phenotype these low frequency 163 populations without any in vitro manipulation. We investigated the immunoglobulin isotype, 164 memory phenotype, homing markers and transcription factor usage of spike and RBD-specific 165 B cells, and global B cells (Fig.3) . 166

The vast majority of SARS-CoV-2 MBC expressed IgG, with a similar isotype distribution 168 observed between spike and RBD-specific MBC (Fig.3a,b,c) . However, individuals with 169 persistent nAb had a higher frequency of IgG isotype expressing spike-and RBD-specific MBC 170 than their counterparts who had lost nAb (Fig.3b,c) , indicating the establishment of a robust, 171 class-switched memory response in these individuals. In contrast, individuals whose nAb had 172 waned below detectable limits had lost more IgG, and had a relative preservation of IgA class-173 switched spike-and RBD-specific MBC (Fig.3b,c) . Elderly residents similarly showed a trend 174 towards less IgG on their spike-specific MBC but, overall, no significant skewing of their 175 immunoglobulin class-switching compared to younger staff (Fig.3b Having found that antigen-specific MBC could persist following complete waning of 208 circulating nAb, we wanted to confirm their potential for functional recall upon re-209 encountering SARS-CoV-2. We therefore used cultured B cell ELISpots to examine the capacity 210 of persistent SARS-CoV-2-specific MBC to differentiate into plasmablasts capable of secreting 211

IgG capable of binding recombinant trimeric spike, S1 or RBD proteins. 212 213 ELISpots were performed using PBMC from 24 seropositive care home residents and staff, 214 with the threshold for detection set at the highest observed value in an uninfected controls 215 group (five seronegative care home residents and five pre-pandemic controls). Only 216 individuals with responses detectable in a control total IgG well were included in analysis. 217

Where responses were too numerous to count (TNTC), the highest number of spot-forming 218 cells (SFCs) observed in the maximal response to the respective protein was used (Fig.S3a) . 219 220 Functional recall responses to SARS-CoV-2 trimeric spike protein were observed in 21 of the 221 24 seropositive individuals tested, with ELISpots tending to be positive in more of those who 222 had maintained nAb (Fig.4a) . However, the majority of those who had lost detectable nAb 223 still had a spike-specific response by ELISpot, with no significant difference in their magnitude 224 compared to the nAb group (Fig.4a ). ELISpots showed similar results for IgG binding S1 and 225 RBD, with a trend to a lower proportion of positive results in those who had lost nAb but no 226 significant difference in the magnitude of B cell recall responses in those maintaining serum 227 nAb or not (Fig.4b,c) . 228

The magnitude of RBD recall response assessed by ELISpot showed a significant correlation 230 with both spike and RBD MBC detection by tetramer staining (Fig.4d ,e). However, there was 231 some discordance due to individuals who had tetramer-binding spike or RBD B cells that did 232 not produce detectable IgG by ELISpot (dotted boxes, Fig.4d ,e), mainly in those who had lost 233 nAb. Importantly, these data revealed that circulating antigen-specific B cells can be detected 234 in the absence of functional recall. 235 236 Next, we compared functional responses to all three proteins for each individual, ranked 237 according to nAb status and age. Individuals with strong recall to spike (as measured by 238

ELISpot) tended to also have strong responses to S1 and RBD, whereas others had weak 239 responses to all three antigens (Fig.4f ). Functional MBC recall responses decreased with 240 increasing age in both the groups, regardless of maintenance of serum nAb (Fig.4f) . Thus, 241 elderly residents had significantly lower ELISpot MBC responses against spike, S1, and 242 particularly RBD, than the younger staff group (Fig.4g ,h,i). Focusing on elderly residents who 243 had lost nAb, we found that none of these individuals sustained MBC capable of functional 244 recall to RBD (Fig.4j) . One strategy to combat antibodies that are waning or unable to cross-recognise emerging 294 variants is the use of booster vaccines. Our finding that the elderly have impaired 295 differentiation of their persistent spike/RBD-specific MBC into antibody producing cells 296 detected by ELISpot assays provides biological rationale for a potential need for more 297 frequent booster vaccination in this high-risk group. The frequency, phenotype and class-298 switching of antigen-specific B cells did not reveal obvious changes in the elderly group to 299 account for this functional defect, other than an increase in the CD27 + CD21subset. The 300 activated CD27 + CD21subset of MBC has recently been noted to remain expanded in some 301 resolved COVID-19 patients, 50 consistent with emerging literature supporting the possibility 302 of prolonged antigen persistence, exemplified by a recent study detecting SARS-CoV-2 in the 303 small bowel four months after asymptomatic infection. 9 Our finding of more antigen-specific 304 CD27 + CD21 -MBC in the older age group raises the possibility there is more prolonged antigen 305 persistence and resultant B cell activation following SARS-CoV-2 infection in the elderly. 306

However, the ageing immune system is characterised by a tendency to low-level chronic 307 inflammation, 51,52 which could also contribute to prolonged activation of SARS-CoV-2 MBC. 308

Analogous to our findings in elderly care home residents, both older subjects and those with 309 HIV have been found to have persistent circulating MBC but defective plasmablast formation, 310 resulting in reduced influenza vaccine-induced antibodies. 53,54 Such age-related defects in B 311 cell responses to vaccination have been attributed to a combination of B cell intrinsic 312 senescence and defective T follicular helper cells (Tfh) in germinal centres. 55-57 313 314 A caveat to our study is that we were only able to study circulating B cells, whilst additional 315 recall responses may be compartmentalised within the mucosa. A recent study suggested 316 mild infection can stimulate mucosal SARS-CoV-2-specific IgA secretion in the absence of 317 circulating antibodies. 58 The bias towards the retention of IgA+ spike/RBD-specific MBC in 318 those who had lost all detectable serum nAb to live virus could therefore be reflective of a 319 stronger mucosal response in these individuals. An increase in mucosal-homing IgA responses 320 has been described as a feature of the ageing immune response, 59 consistent with the older 321 composition of our cohort. Alternatively, the relative preservation of IgA rather than IgG 322 spike/RBD-specific MBC in those with the fastest waning nAb may simply reflect the recent 323 observations that IgA dominates the early nAb response to SARS-CoV-2 infection, 56 and may 324 not decline as fast as the IgG response. 9,60 Since our ELISpot assays did not measure the 325 function of IgA isotype B cells, we may have under-estimated the full extent of residual SARS-326

CoV-2-specific responses, particularly in those with a more IgA-skewed response. In addition, 327 several studies have shown that the magnitude of the MBC response to SARS-CoV-2 continues 328 to increase beyond six months, 9, relative preservation of IgA antigen-specific MBC in those with waned serum nAb raises the 343 possibility that mucosal sequestered immunity may outlast that detectable in the circulation. 344

Increased expansion of activated MBC in the elderly highlights the need to investigate 345 whether they are more prone to prolonged stimulation from persistent reservoirs of SARS-346

CoV-2 antigen. A finding of concern was the lack of detectable functional recall to RBD in 347 elderly donors who had lost nAb; given that RBD is the dominant site for nAb this supports 348 the need for additional monitoring and/or booster vaccines to maintain sufficient antibodies 349 to neutralise emerging variants in this highly vulnerable group. Clinical and laboratory data including age, gender, symptom status at T0 and SARS-CoV-2 RT-387 PCR status at T0 were available for all participants. (Table S1) Recombinant S1 protein constructs spanning SARS-CoV-2 residues 1-530 for ELISpot were 400 produced as previously described. 28,30 Briefly, codon-optimised DNA fragments were cloned 401 into mammalian expression vector pQ-3C-2xStrep to create plasmids, which were then 402 transfected into Expi293F cells growing at 37 in 5% CO2 atmosphere using ExpiFectamine 403 reagent (Thermofisher Scientific). Proteins were purified by strep-tag affinity and 404 subsequently size exclusion chromatography. 405 406

High dimensional multiparameter flow cytometry was used for ex vivo identification of spike 408 and RBD-specific B cells. Two panels (surface and intranuclear) of monoclonal antibodies 409 (mAbs) were used to phenotype global and antigen specific subsets (Table S2) Continuous data that did not follow a normal distribution were described as medians with 463 interquartile ranges and differences compared using the 

Do antibody positive healthcare workers have lower SARS-CoV-2 608 infection rates than antibody negative healthcare workers? Large multi-centre 609 prospective cohort study (the SIREN study)

Humoral and circulating follicular helper T cell 613 responses in recovered patients with COVID-19

Durable SARS-CoV-2 B cell immunity after mild 616 or severe disease

Prior SARS-CoV-2 infection rescues B and T 619 cell responses to variants after first vaccine dose

Maturation and persistence of the anti-622

SARS-CoV-2 memory B cell response

Why does COVID-19 disproportionately affect 625 older people? Aging (Albany NY)

Inflamm-aging. An evolutionary perspective 627 on immunosenescence

The generation of memory B cells is 631 maintained, but the antibody response is not, in the elderly after repeated influenza 632 immunizations

Antibody but not memory B-cell responses are 634 tuned-down in vertically HIV-1 infected children and young individuals being 635 vaccinated yearly against influenza

Age-related factors that affect 638

B cell responses to vaccination in mice and humans

How T follicular helper cells and the germinal centre response change 641 with age

A booster dose enhances 643 immunogenicity of the COVID-19 vaccine candidate ChAdOx1 nCoV-19 in aged mice

Systemic and mucosal antibody responses 646 specific to SARS-CoV-2 during mild versus severe COVID-19

Immunosenescence and mucosal 649 immunity: Significant effects of old age on secretory IgA concentrations and 650 intraepithelial lymphocyte counts

IgA dominates the early neutralizing antibody 653 response to SARS-CoV-2

Persistence of SARS-CoV-2 specific B-and T-cell 656 responses in convalescent COVID-19 patients 6-8 months after the infection

ADNKA overcomes SARS-CoV2-mediated NK cell inhibition through non-spike 662 antibodies. bioRxiv

PD-1 blockade partially recovers 665 dysfunctional virus-specific B cells in chronic hepatitis B infection

Cryo-EM Structure of the 2019-NCoV Spike in the 668

Prefusion Conformation

Optimization of a human IgG B-cell ELISpot assay for the analysis of vaccine-induced 671

B-cell responses

Tracking human antigen-specific memory B 674 cells: A sensitive and generalized ELISPOT system

Sample and clinical data acquisition: AJS, FA

Writing -original draft: AJS, MKM; Writing -review & 702 editing: All authors

with SARS-CoV-2 spike tetramers on MBC (CD3-CD14-CD19+CD20+CD38

CD21+CD27-)) for previously infected (left) and uninfected (right) individuals

Frequency of dual spike-specific MBC (D) in infected (n=32) and uninfected (n=13) and (E) in 761 infected individuals with (nAb, n=19) and without (no nAb, n=13) detectable nAb at T2

Dashed lines indicate threshold for spike-specific responses determined by uninfected 763 controls (Supplementary figure 1b). (F) Proportion of infected individuals with detectable 764 spike-MBC above the threshold stratified by presence (nAb, n=19) and absence (no nAb

Correlation between frequency of spike MBC and live 766 virus nAb titres in infected individuals (n=32). (H) Representative FACS plots of dual staining 767 with SARS-CoV-2 spike tetramers on MBC (top panel) and RBD tetramer on dual spike 768 specific cells (lower panel) of an infected individual. Minimum number of cells in spike-769 specific gate required for RBD probe analysis = 20 (I) Frequency of RBD-specific MBC in 770

Proportion of infected 772 individuals with detectable RBD MBC stratified by presence (nAb, n=16) and absence (no 773 nAb, n=10) of detectable nAb at T2. (K) Correlation between frequency of RBD MBC and live 774 virus nAb titres in all infected individuals (n=29). (A) Wilcoxon matched pairs

E, I) Bars indicate median and interquartile range

J) Fisher's exact test ; (F) p= 0.6285, (J) p= 0.6664. (G, K) Dotted 777 box indicates individuals with discordant MBC and nAb response. Spearman's rank 778 correlation. nAb

MBC, memory B cell, RBD, receptor binding domain

Analysis of RBD specific MBC only in those with ≥20 cells in spike-specific gate

n=6) and resident (blue, n=8) status. (D) Representative FACS plots of CD21 and CD27 gating 868 on spike-specific (top panel), RBD-specific (middle panel), and global

CD19+CD20+CD38lo/neg IgD-MBC from an infected individual. (E) Frequency of

-MBC subsets of spike-specific MBC stratified by 871 presence (nAb, n=16) and absence (no nAb, n=9) of detectable nAb at T2 ordered by 872 increasing age. (F) Frequency of CD21-CD27+

RBD-specific MBC stratified by presence (nAb, n=13) and absence (no nAb, n=4) of 874 detectable nAb at T2 ordered by increasing age

MBC stratified by presence (nAb, n=13) and absence (no nAb, n=4) of detectable nAb at T2, 876 and by staff (grey, n=7) and resident (blue, n=10) status

B) IgG p=0.0382; ns, IgA p=0.0045; ns, IgM ns

Analysis of individuals ≥50 cells in the relevant parent gate for all phenotypic analysis