Three groups of donors were enrolled in this study. The first one was composed of nine COVID-19 recovered patients (hereafter called REC; mean age of 35.1 ± 11.1 years), with a mean of 131.1 days (range 64–165 days) from last infection during follow-up visits at the Infectious Diseases Clinics of the Azienda Ospedaliero-Universitaria Policlinico di Modena. All REC had symptoms consistent with COVID-19 and positive PCR-based testing for SARS-CoV-2 over the period of March 2020–August 2020. Within this group, four patients were classified as severe (35.3 ± 5.68 years) while five patients were moderate (35.0 ± 7.4) according to World Health Organization guidelines (30). Given that there were no differences between moderate and severe recovered individuals and their low number, they were considered as a unique group for the statistical analysis. Twenty-three vaccinated donors were enrolled in this study, and they were divided into two groups: one was composed of 11 donors with a mean of 31.1 days (range 30–35 days) after the third dose of SARS-CoV-2 vaccine (hereafter defined MIX; 27.0 ± 4.5 years); these subjects were vaccinated with three different vaccines (first dose: ChAdOx1; second dose: BNT162b2; third dose: mRNA-1273). The second group was composed of 12 donors with a mean of 33.9 days (range 26–44 days, hereafter defined RNA; 35.3 ± 11.3 years) after being vaccinated with two different RNA vaccines (first and second doses: BNT162b2; third dose: mRNA-1273). Each participant, including healthy donors, provided informed consent according to the Helsinki Declaration, and all uses of human material have been approved by the local Ethics Committee (Comitato Etico dell’Area Vasta Emilia Nord, protocol number 177/2020, 11 March 2020) and by the University Hospital Committee (Direzione Sanitaria dell’Azienda Ospedaliero Universitaria di Modena, protocol number 7531, 11 March 2020). The clinical characteristics of all participants are reported in Table 1 and in the Methods section.

To investigate the percentage of Ag-specific T cells, we used T-cell receptor (TCR)-dependent activation-induced marker (AIM) assays to identify and quantify SARS-CoV-2-specific CD4⁺ T cells (31–34). We stimulated peripheral blood mononuclear cells (PBMCs) from nine REC patients and 11 MIX and 12 RNA donors overnight with 15-mer peptides with 11-amino acid overlap, covering the complete sequence of Wuhan SARS-CoV-2 spike glycoprotein (see Methods for details).

The phenotype of Ag-specific T cells (i.e., those CD137⁺CD69⁺) within CD4⁺ T cells, hereafter termed Ag⁺CD4⁺ T cells, was first analyzed by manual gating and compared with the non-Ag-specific CD4⁺ T cell counterparts (CD137⁻CD69⁻, hereafter called Ag⁻CD4⁺). Ag⁺CD4⁺ T cells showed different cell subset distributions (in terms of the expression of differentiation markers such as CD45RA, CCR7, CD28, and CD95) and Th cell polarization (evaluated by the expression of CCR6 and CXCR3). Ag⁺CD4⁺ T displayed a low percentage of naive (N, CD45RA⁺CCR7⁺CD28⁺CD95⁻) and higher frequencies of memory compartment such as central memory (CM; CD45RA⁻CCR7⁺CD28⁺CD95⁺), transitional memory (TM; CD45RA⁻CCR7⁻CD28⁺CD95⁺), effector memory (EM; CD45RA⁻CCR7⁻CD28⁻CD95⁺), and T_SCM (CD45RA⁺CCR7⁺CD28⁺CD95⁺) and a similar percentage of terminally differentiated effector memory (EMRA; CD45RA⁺CCR7^-CD28^-CD95⁺) ( Supplementary Figures S1A, B ). Considering T-cell polarization, in comparison with Ag⁻CD4⁺ T cells, those Ag-specific displayed a higher percentage of Th1 (CXCR3⁺CCR6⁻), Th17 (CXCR3⁻CCR6⁺), and Th1/Th17 (CXCR3+CCR6+) and a lower percentage of Th0/Th2 (CXCR3⁻CCR6⁻) ( Supplementary Figure S1C ).

To gain a more detailed overview on the differentiation status and Th-polarization, we took advantage of unsupervised FlowSOM clustering. This analysis revealed a total of 19 clusters, and within these, six clusters represented SARS-CoV-2-reactive CD4⁺ T cells expressing CD69 and CD137 ( Figures 1A, B ; Supplementary Figure S2 ).

The frequencies of the different clusters of T cells within Ag⁻CD4⁺ T cells were similar in the three groups of individuals as shown in Supplementary Figure S3 . We focused our attention on Ag⁺CD4⁺ T cells that were selected and reclustered. We obtained 10 clusters, representing different subpopulations of Ag⁺ T cells. We found naive T cells that were defined as CD45RA⁺CCR7⁺CD27⁺ CD95⁻, T_SCM as CD45RA⁺CCR7⁺CD27⁺CD95⁺, CM Th1 as CCR7⁺CD45RA⁻CCR6⁻CXCR3⁺ (CM Th1), CM Th0/Th2 as CCR7⁺CD45RA⁻CCR6⁻CXCR3⁻, CM Th17 as CCR7⁺CD45RA⁻CCR6⁺CXCR3⁻, CM CXCR5⁺ as CCR7⁺CD45RA⁻CXCR5⁺PD-1⁻, circulating T follicular helper as CCR7⁺CD45RA⁻CXCR5⁺PD-1⁺ (cTfh), TM Th1 as CCR7⁻CD45RA⁻CD28⁺CXCR3⁺ (TM Th1), TM Th0/Th2 as CCR7⁻CD45RA⁻CD28⁺CCR6⁻CXCR3⁻, and TM Th17 as CCR7⁻CD45RA⁻CD28⁺CCR6⁺CXCR3⁻( Figure 1C ).

The percentage of total CD4⁺ and Ag⁺ CD4⁺ T cells was similar among the three groups ( Figures 1D, E ). Despite that, within the latter, we observed a different distribution of the populations among REC and vaccinated groups (both MIX and RNA). RNA displayed higher percentages of CM Th1, CM Th0/Th2, and TM Th1 if compared to those in REC subjects. Moreover, both MIX and RNA showed a lower percentage of cTfh cells ( Figure 1F ). No differences were found between Ag⁺CD4⁺ T-cell clusters of MIX and RNA. Similar percentages of all other clusters were present in REC, MIX, and RNA subjects ( Supplementary Figure S4 ).

The AIM assay was used for CD8⁺ T-cell analysis to identify and quantify SARS-CoV-2-specific (see Methods). We first manually gated different subpopulations of T cells on the basis of differentiation markers and cytotoxic-polarization markers (Tc-polarization). We observed that Ag⁺CD8⁺ T cells, if compared to Ag⁻CD8⁺ T lymphocytes, displayed lower percentages of N and higher percentages of T_SCM, CM, and TM; similar percentages of both EM and EMRA were found ( Supplementary Figures S5A, B ). In terms of Tc-polarization, similar percentages of Tc1 cells were found within Ag⁺ and Ag⁻CD8⁺ T cells. However, Ag⁺CD8⁺ T cells were characterized by higher percentages of both Tc17 and Tc1/Tc17 and lower percentages of Tc0/Tc2 ( Supplementary Figure S5C ).

As for CD4⁺ T-cell analysis, we applied unsupervised analysis and found 21 clusters, of which six were SARS-CoV-2-reactive CD8⁺ T cells ( Figures 2A, B ; Supplementary Figure S6A ). Considering Ag⁻CD8⁺ T cells, both MIX and RNA showed increased levels of TM Tc17 CD69⁺, TM Tc0/Tc2, and TM Th0/Th2 PD1⁺ CXCR5⁺ if compared to those of REC subjects. Furthermore, RNA showed a higher percentage of TM Tc1 PD1⁺ CXCR5⁺ compared to those of REC and MIX ( Supplementary Figure S7 ). Ag⁺CD8⁺ T cells were selected and after reclustering, 11 clusters were identified. Besides naive T cells and T_SCM, defined as CD45RA⁺CCR7⁺CD27⁺CD28⁺CD95⁻ and CD45RA⁺CCR7⁺CD27⁺CD28⁺CD95⁺, respectively, we found two clusters of CM T cells defined as follows: CM Tc1 PD-1⁻ that were CD45RA⁻CCR7⁺CD28⁺CXCR3⁺PD-1⁻ and CM Tc1 PD-1⁺ that were CD45RA⁻CCR7⁺CD28⁺CXCR3⁺PD-1⁺.

Among effector Ag⁺CD8⁺ T cells, we found five clusters defined as TM Tc1 PD-1⁻ (CCR7⁻CD45RA⁻CD28⁺CXCR3⁺PD-1⁻), TM Tc1 PD-1⁺ (CCR7⁻CD45RA⁻CD28⁺CXCR3⁺PD-1⁺), TM Tc0/Tc2 (CCR7⁻CD45RA⁻CD28⁺CCR6⁻CXCR3⁻), TM Tc17 (CCR7⁻CD45RA⁻CD28⁺CCR6⁺CXCR3⁻), and EM Tc1 CD57⁺ PD-1⁺ (CCR7⁻CD45RA⁻CD28⁻CD57⁺PD-1⁺). Moreover, three populations of effector memory cells reexpressing CD45RA (EMRA) were detected, i.e., EMRA Tc1 CD57⁻ PD-1⁺ (CCR7⁻CD45RA⁺CXCR3⁺CD57⁻PD-1⁺), EMRA Tc1 CD57⁺ PD-1⁺ (CCR7⁻CD45RA⁺CXCR3⁺CD57⁺PD-1⁺), and EMRA Tc1 CD57⁺ PD-1⁻ (CCR7⁻CD45RA⁺CXCR3⁺CD57⁺PD-1⁻) ( Figure 2C ).

Similar percentages of total CD8⁺ and Ag⁺CD8⁺ T cells were found among the three groups ( Figures 2D, E ). However, within the Ag⁺ population, we observed increased percentages of EM Tc1 CD57⁺ PD-1⁺ in both vaccinated groups if compared to that in the recovered ones ( Figure 2F ). Furthermore, MIX and RNA showed increased levels of EMRA Tc1 CD57⁺ PD-1⁺ terminal effector CD8⁺ T cells compared to that in REC. Finally, we observed that the percentage of EMRA Tc1 CD57⁺ PD-1⁻ terminal effector CD8⁺ T cells was higher in MIX compared to those of both REC and RNA ( Figure 2E ). Similar percentages of all other subpopulations were found among REC, MIX, and RNA ( Supplementary Figure S8 ).

Besides Th-polarization, the functional properties of Ag⁺-specific T cells were investigated by measuring the percentages of cells producing IFN-γ, tumor necrosis factor (TNF), interleukin (IL)-2, IL-17, and granzyme B (GZMB), along with the expression of the degranulation marker CD107a. The percentages of cells producing cytokines were assessed following 16 h of in vitro stimulation with SARS-CoV-2 peptide pool covering the complete sequence of Wuhan SARS-CoV-2 spike glycoprotein. The gating strategy is reported in Supplementary Figure S9 .

REC displayed a higher percentage of CD4⁺ T cells producing TNF, IL-2, and IL-17 than that in the MIX group, but not with respect to that of the RNA group. Furthermore, a higher percentage of CD4⁺ T cells producing IL-2 and TNF was observed in RNA compared to that in MIX subjects. Similar percentages of IFN-γ, CD107a, and GZMB were found among the three groups ( Figure 3A, B ).

Polyfunctional properties were investigated in CD4⁺ and CD8⁺ T cells by analyzing the simultaneous production of TNF, CD107a, IFN-γ, IL-2, and IL-17 using the bioinformatic Simplified Presentation of Incredibly Complex Evaluation (SPICE) tool. Among CD4⁺ T cells, REC exhibited a different polyfunctionality profile from those who had been vaccinated ( Figure 3C ). In particular, REC displayed a higher percentage of CD4⁺ T cells simultaneously producing IL-2 and TNF compared to those in MIX and RNA. The percentage of CD4⁺ T cells producing TNF or IL-17 was higher in REC compared to those in both vaccinated groups. Moreover, RNA exhibited higher percentages of CD4⁺ T cells simultaneously producing IL-2 and TNF or IL-2 alone compared to those in MIX. Furthermore, we found that both vaccinated groups displayed higher percentages of cells defined as “highly polyfunctional” as simultaneously producing CD107a, IFN-γ, IL-2, and TNF compared to those in REC ( Figure 3D ). The functional properties of CD8⁺ T were similar between the three groups ( Supplementary Figure S10 ).

SARS-CoV-2 antibodies decline already as early as 21 days after infection or vaccination (6). However, long-lived MBCs constitute a durable long-term memory and provide a rapid recall response differentiating into high-affinity matured plasma cells (35). For this reason, we measured the frequencies of circulating SARS-CoV-2 spike-specific B cells (Ag⁺ B cells) (see Methods).

Similar percentages of total B cells were found among the three groups ( Figure 4A ). However, both MIX and RNA showed a higher percentage of Ag⁺ B cells (defined as CD45⁺CD19⁺decoy⁻Spike-BUV661⁺Spike-BV650⁺) when compared to that in REC ( Figure 4B ).

By applying manual gating, we observed that Ag⁺ B cells compared to its Ag⁻ counterpart displayed a lower percentage of naive B cells and an increased percentage of memory switched, memory unswitched and of CD27⁻IgD⁻ B cells ( Supplementary Figure S11A ). Moreover, after vaccination or SARS-CoV-2 infection, ~42%–96% of Ag⁺ B cells were IgG⁺. This percentage decreased to ~5%–22% in the Ag⁻ B cells, where ~69%–90% of cells were IgD⁺IgM⁺ ( Supplementary Figure S11B ). Furthermore, the percentage of IgA⁺ B cells was higher in the Ag⁻ compartment ( Supplementary Figure S11B ).

To deeply characterize both Ag⁻ and Ag⁺ B cells, we took advantage of unsupervised clustering. The analysis revealed 15 clusters, spanning from naive to atypical B cells (atBCs; CD21⁻CD27⁻CD38⁻) (36) ( Figures 4C, D ; Supplementary Figure S12A ).

Besides naive and transitional B cells (TrB), respectively defined as naive: CD20⁺CD21⁺CD24⁺CD38⁻IgD⁺IgM⁺ and TrB: CD20⁺CD21⁺CD24⁺CD38⁺IgD⁺IgM⁺, we found five clusters of MBCs defined as follows: MBC IgD⁺ IgM⁺ (CD20⁺CD21⁺CD24⁺CD27⁺IgD⁺IgM⁺), MBC IgA⁺ (CD20⁺CD21⁺CD24⁺CD27⁺IgA⁺), MBC IgG⁺ (CD20⁺CD21⁺CD24⁺CD27⁺IgG⁺), MBC IgA⁺ CD71⁺ (CD20⁺CD21⁺CD24⁺CD27⁺IgA⁺CD71⁺), and MBC IgG⁺ CD71⁺ (CD20⁺CD21⁺CD24⁺CD27⁺IGA⁺CD71⁺). Among plasmablasts (PBs), we found the following three clusters: PB IgA⁺ as CD27⁺CD71⁺CD38⁺⁺IgA⁺, PB IgM⁺ as CD27⁺CD71⁺CD38⁺⁺IgM⁺, and PB IgG⁺ as CD27⁺CD71⁺CD38⁺⁺IgG⁺. Together with naive, TrBs, MBCs, and PBs, we identified five clusters of atBCs, i.e., atBC1 as CD21⁻CD27⁻CD20⁺IgG⁺, atBC2 as CD21⁻CD27⁻CD24⁺CD20⁺IgG⁺, atBC3 as CD21⁻CD27⁻CD20⁺IgD⁺, atBC4 as CD21⁻CD27⁻CD20⁺IgD⁺IgM⁺, and atBC5 as CD21⁻CD27⁻CD20⁺CD24⁺.

Within Ag⁻ B cells, MIX and RNA showed higher levels of MBC IgD⁺IgM⁺ and lower levels of atBC5 compared to those in REC ( Supplementary Figure S13 ). Within Ag⁺ B cells, MIX and RNA displayed lower percentages of naive, MBC IgA⁺, and atBC4 B cells if compared to those in REC, while the percentages of MBC IgG⁺ CD71⁺ and atBC2 were significantly higher ( Figure 4E ). Moreover, REC displayed a higher percentage of atBC4 cells if compared to those in MIX and RNA ( Figure 4E ). Similar percentages of all other subpopulations were found among REC, MIX, and RNA subjects ( Supplementary Figure S14 ).

In addition, we measured IgG antibodies able to bind the spike and the RBD of the S1 subunit of the spike protein (the latter known as neutralizing antibodies). We observed that both vaccinated groups had higher levels of anti-spike and anti-RBD-binding IgG compared to those in REC subjects ( Figure 4F ).

The principal component analysis (PCA) computed using the complete phenotype of Ag⁺ B and T cells, CD4⁺ T cell polyfunctionality, plasmatic anti-spike, and anti-RBD antibodies showed that the group of REC clusters in a different position of the two-dimensional PCA space if compared to MIX and RNA, which are almost entirely overlapping ( Figure 5A , left). Immune features related to the amount of MBC IgA, CD107a⁻IFN-γ⁻IL2⁻TNF⁺IL17⁻, CD107a⁻IFN-γ⁻IL2⁺TNF⁺IL17⁻, CD107a⁻IFN-γ⁻IL2⁻TNF⁻IL17⁺, and naive B cells (more abundant in REC subjects) were the main drivers of the clusterization of samples in two different areas ( Figure 5A , right). Moreover, the picture of PCA contribution also reveals that both vaccinated groups were characterized by increased levels of MBC IgG CD71⁺, anti-spike, and anti-RBD IgG antibodies ( Figure 5A , right).

By using the same parameters used to perform the PCA, we assessed the existence of immunological correlations between the variables within the REC, MIX, and RNA groups. It is to note that in REC, but not in the MIX and RNA groups, a strong positive correlation was present among the percentages of MBC IgA CD71⁺ and all polyfunctional CD4⁺ T-cell subsets ( Figure 5B , Supplementary Figure S15 ). The percentages of MBC IgD⁺ IgM⁺, transitional, and naive B cells inversely correlate with all polyfunctional CD4⁺ T-cell subsets ( Figure 5B , Supplementary Figure S15 ).

Up to 30 ml of blood was collected from each patient in vacuettes containing ethylenediaminetetraacetic acid (EDTA). Blood was immediately processed. Isolation of PBMCs was performed using Ficoll-Hypaque according to standard procedures (48). PBMCs were stored in liquid nitrogen in fetal bovine serum (FBS) supplemented with 10% dimethyl sulfoxide (DMSO). Plasma was stored at -80°C until use.

Isolated PBMCs were thawed and rested for 6 h. After resting, CD40-blocking antibody (0.5 µg/ml final concentration) (Miltenyi Biotec, Bergisch Gladbach, Germany) was added to the cultures 15 min before stimulation. PBMCs were cultured in a 96-well plate in the presence of 15-mer peptides with 11-amino acid overlap, covering the complete sequence of Wuhan SARS-CoV-2 spike glycoprotein (PepTivator SARS-CoV-2 Prot_S complete, Miltenyi Biotec, Bergisch Gladbach, Germany) together with 1 μg/ml of anti-CD28 (Miltenyi Biotec, Germany). PBMCs were stimulated for 18 h at 37°C in a 5% CO₂ atmosphere in complete culture medium (RPMI 1640 supplemented with 10% FBS and 1% each of L-glutamine, sodium pyruvate, nonessential amino acids, antibiotics, 0.1 M 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), and 55 μM β-mercaptoethanol) (33, 34). For each stimulated sample, an unstimulated one was prepared as a negative control. After stimulation, cells were washed with Phosphate buffer saline (PBS) and stained with PromoFluor IR-840 (Promokine, PromoCell, Heidelberg, Germany) for 20 min at room temperature (RT). Next, cells were washed with FACS buffer (PBS added with 2% FBS) and stained with the following fluorochrome-labeled monoclonal antibodies (mAbs) for 30 min at 37°C: CXCR5-BUV661, CCR6-BUV496, and CXCR3-BV785. Finally, cells were washed with FACS buffer and stained for 20 min at RT with Duraclone IM T-cell panel (Beckman Coulter, Brea, CA, USA) containing CD45-Krome Orange, CD3-APC-A750, CD4-APC, CD8-AF700, CD27-PC7, CD57-Pacific Blue, CD279 (PD1)-PC5.5, CD28-ECD, CCR7-PE, and CD45RA-FITC and added with three other fluorescent mAbs, i.e., CD69-BV650, CD137-BUV395, and CD95-BV605. Samples were acquired on a CytoFLEX LX flow cytometer (Beckman Coulter). All reagents used for T-cell phenotype are reported in Supplementary Table S1 . All mAbs added to DuraClone IM T cells were previously titrated on human PBMCs and used at the concentration giving the best signal-to-noise ratio. The gating strategy used to identify CD4⁺ and CD8⁺ T cells is reported in Supplementary Figure S16 .

Isolated PBMCs were thawed and rested for 6 h. PBMCs were then cultured in the presence of 15-mer peptides with 11-amino acid overlap, covering the complete sequence of Wuhan SARS-CoV-2 spike glycoprotein (PepTivator SARS-CoV-2 Prot_S complete, Miltenyi Biotec, Bergisch Gladbach, Germany) together with 1 μg/ml of anti-CD28 (Miltenyi Biotec, Germany). PBMCs were stimulated for 16 h at 37°C in a 5% CO₂ atmosphere in complete culture medium (RPMI 1640 supplemented with 10% FBS and 1% each of L-glutamine, sodium pyruvate, nonessential amino acids, antibiotics, 0.1 M HEPES, and 55 μM β-mercaptoethanol). For each stimulated sample, an unstimulated one was prepared as a negative control. All samples were incubated with protein transport inhibitor containing brefeldin A (Golgi Plug, Becton Dickinson) and monensin (Golgi Stop, Becton Dickinson) and previously titrated concentration of CD107a-PE (BioLegend, San Diego, CA, USA). After stimulation, cells were washed with PBS and stained with LIVE/DEAD fixable Aqua (ThermoFisher Scientific, USA) for 20 min at RT. Next, cells were washed with FACS buffer and stained with surface mAbs recognizing CD3-PE.Cy5, CD4-AF700, and CD8-APC.Cy7 (BioLegend, San Diego, CA, USA). Cells were washed with FACS buffer and fixed and permeabilized with the Cytofix/Cytoperm buffer set (Becton Dickinson Bioscience, San Jose, CA, USA) for cytokine detection. Then, cells were stained with previously titrated mAbs recognizing IL-17-PE-Cy7, TNF-BV605, IFN-γ-FITC, IL-2-APC, and GZMB BV421 (all mAbs from BioLegend). Samples were acquired on an Attune NxT acoustic cytometer (ThermoFisher Scientific, USA). Supplementary Table S1 reports mAb titers, clones, catalog numbers, and type of fluorochrome used in the panel. Gating strategy used to identify and analyze the intracellular cytokine production of CD4⁺ and CD8⁺ T lymphocytes is reported in Supplementary Figure S9 .

Thawed PBMCs were washed twice with RPMI 1640 supplemented with 10% FBS and 1% each of L-glutamine, sodium pyruvate, nonessential amino acids, antibiotics, 0.1 M HEPES, 55 μM β-mercaptoethanol, and 0.02 mg/ml DNAse. PBMCs were washed with PBS and stained using viability marker PromoFluor IR-840 (Promokine, PromoCell, Heidelberg, Germany) for 20 min at RT in PBS. Next, cells were washed with PBS and stained for 15 min at RT with streptavidin-AF700 (decoy channel; ThermoFisher Scientific, USA) to remove false-positive SARS-CoV-2-specific B cells. After washing with FACS buffer, cells were stained with biotinylated full-length SARS-CoV-2 spike protein (R&D Systems, Minneapolis, USA) labeled with different streptavidin-fluorophore conjugates. Full-length biotinylated spike protein was mixed and incubated with streptavidin-BUV661 (Becton Dickinson) or streptavidin-BV650 (BioLegend) at a 6:1 mass ratio for 15 min at RT. All samples were stained with both fluorescent biotinylated biotinylated spike protein for 1 h at 4°C. Then, cells were washed with FACS buffer and stained for 20 min at RT with DuraClone IM B cells (Beckman Coulter, Brea, CA, USA) containing the following lyophilized directly conjugated mAbs: anti-IgD-FITC, CD21-PE, CD19-ECD, CD27-PC7, CD24-APC, CD38-AF750, anti-IgM-PB, and CD45-KrO to which the following drop-in antibodies were added: CD71-BUV395, CD20-BV785, anti-IgG-BUV496, and anti-IgA-PerCP-Vio700. Samples were acquired on a CytoFLEX LX flow cytometer (Beckman Coulter). A minimum of 1,000,000 cells per sample were acquired. All reagents used for B-cell phenotype are reported in Supplementary Table S1 . All mAbs added to DuraClone IM B cells were previously titrated on human PBMCs and used at the concentration giving the best signal-to-noise ratio. The gating strategy used to identify Ag⁻ and Ag⁺ B cells is reported in Supplementary Figure S17 .