Robust, persistent adaptive immune responses to SARS-CoV-2 in the oropharyngeal lymphoid tissue of children

SARS-CoV-2 infection triggers adaptive immune responses from both T and B cells. However, most studies focus on peripheral blood, which may not fully reflect immune responses in lymphoid tissues at the site of infection. To evaluate both local and systemic adaptive immune responses to SARS-CoV-2, we collected peripheral blood, tonsils, and adenoids from 110 children undergoing tonsillectomy/adenoidectomy during the COVID-19 pandemic and found 24 with evidence of prior SARS-CoV-2 infection, including detectable neutralizing antibodies against multiple viral variants. We identified SARS-CoV-2-specific germinal center (GC) and memory B cells; single cell BCR sequencing showed that these virus-specific B cells were class-switched and somatically hypermutated, with overlapping clones in the adenoids and tonsils. Oropharyngeal tissues from COVID-19-convalescent children showed persistent expansion of GC and anti-viral lymphocyte populations associated with an IFN-γ-type response, with particularly prominent changes in the adenoids, as well as evidence of persistent viral RNA in both tonsil and adenoid tissues of many participants. Our results show robust, tissue-specific adaptive immune responses to SARS-CoV-2 in the upper respiratory tract of children weeks to months after acute infection, providing evidence of persistent localized immunity to this respiratory virus.

in GC and anti-viral memory responses following  In nearly all seropositive participants, we detected S1 + RBD + B cells in PBMCs and both 109 pharyngeal tissues (Fig. 1e), with the exception of two donors (CNMC 91 and 104) who 110 had almost no S1 + RBD + binding B cells in the peripheral blood (Extended Data Fig. 1a). 111 These two donors also had the lowest serum neutralizing antibody titers to WA-1 among 112 our cohort. Surprisingly, one participant (CNMC 32) had high serum neutralization titers 113 but very low percentages of S1 + RBD + B cells, particularly in the oropharyngeal tissues, 114 highlighting heterogeneity in responses to SARS-CoV-2. 115 116 Evaluation of B cell populations by high-dimensional flow cytometry revealed that the 117 majority of S1 + RBD + B cells were CD27 + immunoglobulin (Ig) class-switched memory B 118 cells (IgD -CD38 -CD27 + ) ( Fig. 1f-g, Extended Data Fig. 1b, Supplementary Fig. 1-2), 119 indicating a robust memory B cell response was generated and maintained in the upper 120 respiratory tract as long as 10 months into the convalescent period (Extended Data Fig.  121 1c). These S1 + RBD + memory B cells were primarily IgG + , with lower percentages of IgA + 122 cells compared to total CD27 + memory B cells in the tissue, perhaps reflecting the 123 inflammatory milieu during infection (Extended Data Fig. 1d). Of note, the percentage of 124 S1 + RBD + cells we found among CD27 + switched memory B cells in the oropharyngeal 125 tissue was comparable to that recently reported in lung and lung-draining lymph nodes 126 from convalescent autopsy donors (Fig. 1f, Extended Data Fig. 1e) 13 . 127

128
The predominance of Ig class-switched CD27 + memory B cells among S1 + RBD + B cells 129 suggested that they originated from GC reactions, although the timing of class switching 130 remains controversial 15 . Because the tonsils and adenoids are secondary lymphoid tissues 131 and sites of robust GC formation, we could directly examine the involvement of GCs. Flow 132 cytometric analysis revealed a substantial portion of GC B cells among the S1 + RBD + B 133 cells in both tissues (Fig. 1g). Paired analyses of tonsils and adenoids from the same donor 134 revealed that the adenoids had higher frequencies of S1 + RBD + cells among both total and 135 GC B cells compared to tonsils, perhaps reflecting higher viral exposure due to their 136 location in the nasopharynx (Extended Data Fig. 1f-g). Frequencies of S1 + RBD + B cells in negative (DN, IgD -CD27 -CD38 -CD19 + ) B cells 16,17 . We also saw an expansion of DN B 150 cells among S1 + RBD + B cells in both adenoids and tonsils (Fig. 1g). However, most of 151 these S1 + RBD + DN B cells exhibited characteristics of DN1 (CD21 + CD11c -) cells, which 152 are derived from GCs (Fig. 1i). Only a small portion were DN2 (CD21 -CD11c + ) cells, which 153 originate from extrafollicular B cell activation and were reported to expand in acute severe 154 COVID-19 16 . Our findings, therefore, suggest that robust humoral responses to  CoV-2 associated with intact GC reactions and B cell memory are present in the upper 156 respiratory tract mucosal lymphoid tissue. 157 158

Multimodal single-cell analysis of SARS-CoV-2-specific B cells 159
To investigate B cell responses in greater detail, we sorted S1-binding (S1 + ) and non-160 binding (S1 -) B cells from tonsils, adenoids, and PBMCs from two subjects with a history 161 of COVID-19, as well as one uninfected control (Supplementary Fig. 3a-b). Over 1860 S1 + 162 B cells and 25000 S1 -B cells were captured and characterized by CITE-seq (Cellular 163 Indexing of Transcriptomes and Epitopes by Sequencing), which simultaneously measured cytometric analysis, the majority of S1 + B cells in the tonsils and adenoids were in cluster 172 2, which represented CD27 + memory B cells (Fig. 2c-e). Adenoids and tonsils had a smaller 173 but clear portion of S1 + cells that were in cluster 4, which had a GC B cell gene expression 174 signature and surface protein profile ( Fig. 2b-e, Extended Data Fig. 2a-b). In contrast, S1 + 175 cells in the blood were primarily in cluster 9 ( Fig. 2a-c, e), which was also a CD27 + IgD -176 population (Fig. 2e, lower heatmap) but had different surface marker and gene expression 177 profiles compared to CD27 + IgDmemory B cells in the lymphoid tissues (Fig. 2e, upper  178 heatmap; Extended Data Fig. 2a). S1 + cells from both the peripheral blood and tissues also 179 clustered separately when cells were clustered by transcript expression alone (Extended 180 Data Fig. 3). Furthermore, S1 + memory B cells in cluster 2 had higher expression of CXCR3 181 and HOPX, genes known to be induced by T-bet in T cells 19 , than their S1counterparts, 182 suggesting that they may have developed in a more IFN-g rich environment 183 (Supplementary Table 5  compared to S1 -B cells, indicative of antigen-driven clonal expansion (Fig. 3c). The high 192 mutation frequency in S1 + B cells is consistent with prior work showing that subjects with 193 mild COVID-19 had higher frequencies of hypermutated memory B cells compared to those 194 with severe COVID-19 22 and suggests that these SARS-CoV-2-specific clones underwent 195 somatic hypermutation in GCs. 196 197 Intriguingly, we also observed that a portion of S1 + B cell clones (a total of 83 cells from 29 198 clones: 20 clones from donor 89 and 9 from donor 71) were present in both the tonsils and 199 adenoids (Fig. 3d). The shared S1 + clones were nearly all isotype-switched cells (Extended 200 Data Fig. 2f) and, like the total S1 + B cell population, were comprised primarily of cells from 201 cluster 2 (CD27 + memory B cells) (Fig. 2e). However, a small number of cells from shared 202 clones in the tonsil of one donor were GC B cells (cluster 4) ( Fig. 2e; Supplementary Table  203 6). The distribution of these shared clones across adenoid and tonsil within some clonal 204 lineage trees suggested that B cell clones migrated between these oropharyngeal lymphoid 205 tissues and raised the possibility that class switching can occur before, during, or after 206 SHM (Fig. 3e). Thus, multimodal single cell analysis of the SARS-CoV-2-specific B cells 207 both supports their emergence from GCs and suggests sharing and potential migration of 208 clonally expanded B cells between oropharyngeal lymphoid tissues. 209 210

Expanded GC populations after COVID-19 211
To determine whether prior SARS-CoV-2 infection can broadly alter the immune landscape 212 of mucosal tissues beyond acute infection, we compared the immune cell profiles of tonsils, 213 adenoids, and peripheral blood from individuals with a history of COVID-19 to those 214 without, using both unsupervised analyses and manual gating of high-dimensional flow 215 cytometry data (samples included in each analysis are listed in Supplementary Table 2). 216 To probe cell populations in greater detail, CD19 + B, CD4 + T, and CD8 + T lymphocytes 217 were first gated and then analyzed separately. Adenoids and tonsils were evaluated 218 together, whereas PBMCs were examined on their own, to account for and increase 219 sensitivity for detecting distinct populations in tissues and peripheral blood. 220

221
In the unsupervised analysis of B cell phenotypes, we compared those with prior COVID-222 19 to control subjects while controlling for age and sex. This analysis highlighted 14 223 clusters and revealed more pronounced changes in the adenoids post-COVID-19 ( Fig. 4a-224 b, Extended Data Fig. 4). Clusters 3 and 10 were significantly increased in the adenoids of 225 participants with a history of COVID-19 (Fig. 4b); these clusters represented IgG + and IgM + 226 GC B cells, respectively. In addition, cluster 14, which clustered with naïve B cells, was 227 decreased in both adenoids and tonsils of COVID-19 convalescent subjects ( Fig. 4a-b). In 228 the peripheral blood, a CD127 + IgD + B cell cluster was also decreased following COVID-19 229 Post-COVID-19, we observed that the adenoids had lower percentages of CD4 + T cells 236 (Extended Data Fig. 5a). Unsupervised clustering further underscored differences in 237 COVID-19-convalescent samples ( Fig. 5a-b, Extended Data Fig. 6a-b) that included a 238 reduction in cluster 9, which represents naïve CD4 + T cells (CD45RA + CCR7 + ), in both 239 tonsils and adenoids from COVID-19 convalescent subjects ( Fig. 5a-b). Traditional gating 240 confirmed decreased percentages of naïve CD4 + T cells in lymphoid tissue (Fig. 5c). 241 242 Conversely, cluster 3, which represents a CD57 + PD-1 hi subset, was significantly enriched 243 after COVID-19 in both the adenoids and tonsils ( Fig. 5a-b); manual gating confirmed an 244 expanded CD57 + PD-1 hi CD4 + T cell population in the tissues (Fig. 5d). CD57 has been 245 described as a marker of T cell senescence but is also found on a population of tonsillar 246 GC-Tfh cells 23-25 , a subset of CD4 + T helper cells that provide contact-mediated signals to 247 antigen-stimulated B cells for GC formation and maintenance. Compared to the total CD4 + 248 T cell population in the tissues, the CD57 + PD-1 hi CD4 + T cell population exhibited higher 249 expression of CXCR5 and CD69, indicative of a Tfh phenotype and characteristic of tissue-250 resident memory T (TRM) cells, respectively 6 ( Fig. 5e). Imaging studies revealed that 251 CD57 + PD-1 hi CD4 + T cells were located within the GC (Fig. 5f). Moreover, their frequency   Fig. 5j-k). Similar to the characteristics we found in adenoid CD4 + T cells, 288 regulatory T cells (CD25 + CD127 -) in the adenoids were also more activated after COVID-289 19, with a higher percentage of HLA-DR + CD38 + and CXCR3 + CCR6cells, again 290 suggesting the adenoids may have been primed by a stronger immune response to SARS-291 CoV-2 than the tonsils (Extended Data Fig. 5l-m). Thus, we find an expansion of 292 percentages of Tfh as well as Tfr cells in the tonsils and adenoids that extends into 293 convalescence, providing further evidence for prolonged GC responses to SARS-CoV-2 in 294 the upper respiratory tract of children. 295 296

Enrichment of activated circulating Tfh cells in the blood following COVID-19 297
Because lymphocyte populations in the peripheral blood differ from the tonsil and adenoid, 298 we evaluated PBMCs separately; unsupervised grouping of high-dimensional flow 299 cytometry data revealed two clusters (cluster 5 and cluster 11) that were increased 300 following COVID-19 ( Fig. 6a- Supplementary Fig. 7a-b); both contained circulating Tfh 301 (cTfh)-like cells (CD45RA -CXCR5 + PD-1 + ) that expressed CD38, a marker of recently 302 activated T cells 28 ; cluster 11 was CXCR3 + while cluster 5 was not. Although we did not 303 find increased percentages of total cTfh cells by manual gating, we found that cTfh cells 304 were skewed to a CXCR3 + CCR6phenotype in the COVID-19-experienced group (Fig. 6c); 305 these cells produced IFN-g upon stimulation with PMA and ionomycin (Extended Data Fig.  306 7a). Analogous to prior reports, we also observed an increased frequency of stem cell-like 307 Fig. 7b), 308 perhaps reflecting long-lived memory T cells following recovery from COVID-19 in 309 To identify SARS-CoV-2 antigen-specific CD4 + T cells, we stimulated tonsil, adenoid, and 312 peripheral blood mononuclear cells with spike (S), membrane (M), and nucleocapsid (N) 313 peptide pools and assessed the activation-induced markers (AIM) CD40L, OX40, and 4-314 1BB on T cells. Although we were not able to precisely identify and phenotype the SARS-315 CoV-2-specific T cells in the adenoids and tonsils due to the highly activated status of T 316 cells at baseline without stimulation in these tissues (Extended Data Fig. 7c-

d), SARS-317
CoV-2-reactive CD4 + T cells were identified in the peripheral blood with the greatest 318 responses to the S peptide pool (Fig. 6d-e). By concatenating all the peptide-activated 319 CD4 + T cells, we found that the SARS-CoV-2-responsive CD4 + T cells in the peripheral 320 blood were primarily memory cells that were enriched for CXCR3 + cTfh cells (CD45RA -321

Expanded tissue resident CD8 + cells after COVID-19 327
To further evaluate anti-viral responses, we examined CD8 + T cell in the tonsils and 328 adenoids. With unsupervised clustering, we found that cluster 1, which represented naïve 329 CD8 + T cells, decreased following COVID-19 in the adenoids ( Fig. 7a-b, Extended Data 330 CD8 + T cells (HLA-DR + CD38 + CXCR3 + CCR7 -CD45RA -); cluster 2 expressed higher CD38, 336 while cluster 3 expressed more CD57. Manual gating demonstrated that CD57 + PD-1 + CD8 + 337 T cells were significantly higher in adenoids and tonsils (Fig. 7c), while activated HLA-338 DR + CD38 + CD8 + T cells trended higher in tonsils of the COVID-19-convalescent group 339 (Extended Data Fig. 9d). As in CD4 + T cells, the COVID-19-convalescent adenoids also 340 had significantly more CXCR3 + CCR6 -CD8 + T cells (Tc1 skewed) (Extended Data Fig. 9e). 341 Furthermore, CD8 + T cells in the adenoid produced more IFN-g than those in the tonsils 342 upon PMA/ionomycin stimulation, again indicating the ability of the adenoids to create a 343 more IFN-g rich environment during the anti-viral response (Extended Data Fig. 9f). 344 345 CD8 + T cells expressing the senescence marker CD57 and inhibitory surface protein PD-346 1 are expanded in the peripheral blood of adults with moderate and severe COVID-19; 347 however, the function of these cells and whether they represent a non-functional 348 "exhausted" population is not clear 30,31 . We found that CD57 + PD-1 + CD8 + T cells in the 349 adenoids and tonsils had robust pro-inflammatory cytokine and cytotoxic factor production 350 following PMA and ionomycin stimulation (Extended Data Fig. 9g-h).  Table 7). SARS-CoV-2 was detected in 7 out of 9 FFPE 386 adenoid blocks and 15 out of 22 FFPE tonsil blocks from COVID-19-convalescent 387 individuals, but not in any control tissue samples. In several samples, participants' previous 388 positive PCR from a nasal swab was over 100 days prior to surgery, including one which 389 was 303 days before surgery. Moreover, the copies of viral RNA significantly correlated 390 with the percentage of S1 + RBD + cells among GC B cells in the tonsil (Fig. 8b). Although   CoV-2 test to surgery (n = 10). 678 e. Frequency of S1 + RBD + cells among total CD19 + B cells from PBMC, adenoid, and 679 tonsil from COVID vs. CON (PBMC COVID n = 18, CON n = 33; adenoid COVID n = 16, 680 CON n = 27; and tonsil COVID n = 16, CON n = 30). 681 f. Representative flow cytometry plots demonstrating the percentage of SARS-CoV-2-682 specific (S1 + RBD + ) cells among CD27 + IgDswitched memory B cells in PBMC, adenoid, 683 and tonsil following COVID-19. Gating strategy shown in Supplementary Fig 1-2.   a. Sub-isotype frequencies among S1 + and S1 -B cells from PBMC, adenoid, and tonsil of 718 one COVID-19 convalescent donor (CNMC 89). Labels show the raw number of cells with 719 a given sub-isotype and are only included for sub-isotypes that make up at least 10% of a 720 given category. 721 b. Somatic hypermutation (SHM) frequency among S1 + and S1 -B cells from PBMC,  CD4 + T cells were included in the SPICE analysis (see Supplementary Fig. 6).

Ethics statement 864
This study was approved by the Institutional Review Board (IRB) at Children's National 865 Hospital (IRB protocol number 00009806). Written informed consent was obtained from 866 parent/guardians of all enrolled participants, and assent was obtained from minor 867 participants over 7 years of age. 868

Participant recruitment 869
We recruited 110 children who underwent tonsillectomy and/or adenoidectomy at 870 Children's National Hospital (CNH) in Washington, DC, USA. All children scheduled to 871 undergo tonsillectomy at CNH were eligible. The first 102 participants were recruited from 872 late September 2020 to early February 2021 without screening for prior COVID-19. An 873 additional 2 participants were subsequently recruited with known history of COVID-19, plus 874 6 additional subjects (one of whom turned out to be positive by serology) were recruited in 875 May and June 2021. Because not all tissues or blood were available from each subject, we 876 collected a total of 106 blood samples, 100 adenoids, and 108 tonsils from 110 participants 877 (Supplementary Table 2). No statistical methods were used to predetermine sample size. 878 All participants had negative RT-PCR testing from a nasopharyngeal swab for SARS-CoV-879 2 within 72 hours of the surgery. Demographic information and clinical data were collected 880 through parental questionnaires and chart review and inputted and managed in REDCap, 881 and biologic samples were acquired in the operating room by a separate clinical team at 882

CNH. 883
Eleven participants had previous confirmed SARS-CoV-2 infection with RT-PCR or antigen 884 testing from nasopharyngeal swabs. Another thirteen COVID-19-exposed participants 885 were identified through serum antibody testing and/or identification of B cells that recognize 886 the spike protein of SARS-CoV-2 by flow cytometry (described below). One participant 887 (CNMC 43) had SARS-CoV-2 detected by RT-PCR from the nasopharynx 20 days prior to 888 surgery but had negative serology and no SARS-CoV-2 specific B cells in the tissue or 889 blood. We excluded this subject from our subsequent analysis. 890

Control selection within the cohort 891
Controls for flow cytometric analyses were selected among subjects with no serologic or 892 cellular evidence of prior COVID-19. The primary indication for tonsillectomy in all 24 893 participants with prior COVID-19 was adenotonsillar hypertrophy leading to sleep 894 disordered breathing (SDB) or obstructive sleep apnea (OSA) (Supplementary Table 1 and 895 3) except one participant who had eustachian tube dysfunction. Patients with SDB and 896 OSA both have breathing difficulties during sleep (primarily snoring); however, patients 897 with OSA had polysomnography documenting an apnea-hypopnea index greater than 1, 898 while those with SDB did not undergo polysomnography testing and were diagnosed by 899 clinical history alone. None of the 24 participants with COVID-19 had frequent recurrent 900 tonsillitis (more than 6 episodes in a year) or other medical problems that directly affect the 901 immune system aside from atopic disease, nor did they take immunomodulating 902 medications aside from nasal/inhaled steroid or loratadine within 2 weeks of surgery.  Table 2. 915

Blood and tissue collection 916
Blood samples were obtained just prior to the surgical procedure in the operating room in 917 serum separator tubes (BD) for serum collection and sodium heparin tubes (BD) for 918 peripheral blood mononuclear cells (PBMCs) extraction from an intravenous line placed for 919 anesthesia. Once received in the laboratory on the day of collection, serum separator tubes 920 were spun at 1200g for 10 min, and serum was aliquoted and stored at -80°C. PBMCs 921 were isolated the day after collection by density gradient centrifugation (Lymphocyte 922 Separation Medium, MP Biomedicals) at 1500rpm for 30 min at room temperature with no 923 brake and washed with PBS. If red blood cell contamination was present, cells were lysed 924 with ACK buffer. 925 Tonsil and adenoid tissues were stored in RPMI media with 5% FBS (VWR), gentamicin 926 50mg/mL (Gibco), and 1X antibiotic/antimycotic solution (Gibco) on ice immediately after 927 collection. Tissues were processed the day after collection. A 3-5mm portion of tonsil and 928 adenoid tissue was cut and fixed in 5mL of 10% buffered formalin (Avantik) for 24-48 h. 929 The fixed tissue was then incubated in 70% ethanol until it was paraffin-embedded. The 930 remainder of the tissue was mechanically disrupted and filtered through a 100μm cell 931 strainer to create a single cell suspension, lysed with ACK buffer (Gibco), and washed with 932 PBS three times. Freshly isolated PBMCs and tonsil and adenoid cells were surface 933 stained and analyzed with flow cytometry as described below on the day of processing. 934 The remaining cells were stored in liquid nitrogen in the presence of FBS (VWR) with 10% 935

TMPRSS2 cells (HEK 293T cells that express ACE2 and TMPRSS2 proteins). 956
For the neutralization assay, 50 µL of SARS-CoV-2 S pseudovirions were pre-incubated 957 with an equal volume of medium containing serum at varying dilutions at room temperature 958 (RT) for 1 h, then virus-antibody mixtures were added to 293T-ACE2-TMPRSS2 cells in a 959 96-well plate. The input virus with all SARS-CoV-2 strains used in the current study were 960 the same (2x10 5 Relative light units/50 µL/well). After a 3 h incubation, the inoculum was 961 replaced with fresh medium. Cells were lysed 24 h later, and luciferase activity was 962 measured using luciferin. Controls included cells only, virus without any antibody and 963 positive sera. The cut-off value or the limit of detection for the neutralization assay is 1:10. 964 Data are shown in Supplementary Table 4. 965

SARS-CoV-2 antigen specific B cell detection 967
5 million cells per sample of PBMC, adenoid, or tonsil were resuspended in PBS with 2% 968 FBS and 2 mM EDTA (FACS buffer). Biotinylated S1 and RBD probes (BioLegend) were 969 crosslinked with fluorochrome-conjugated streptavidin in a molar ratio of 4:1. 970 Fluorochrome-conjugated streptavidin was split into 5 aliquots and conjugated to 971 biotinylated S1 and RBD probes by mixing for 20 min/aliquot at 4°C. Cells were first stained 972 with the viability dye, Zombie NIR (1:800 dilution, BioLegend), for 15 min at RT, washed 973 twice and then incubated with True-Stain Monocyte Blocker (BioLegend) for 5 min. An 974 antibody cocktail containing the rest of the surface antibodies, the fluorochrome-975 conjugated S1 and RBD probes, and Brilliant Stain Buffer Plus (BD) were then added 976 directly to the cells and incubated for 30 min at RT in the dark (200uL staining volume). 977 Cells were washed three times and fixed in 1% paraformaldehyde for 20 min at RT before 978 washing again and collecting on a spectral flow cytometer (Aurora, Cytek). Antibodies used 979 in this assay are shown in Supplementary Table 8.