Memory B Cells and Memory T Cells Induced by SARS-CoV-2 Booster Vaccination or Infection Show Different Dynamics and Responsiveness to the Omicron Variant

In December 2019, an outbreak of unknown etiology with apparently viral pneumonia emerged in Wuhan, China. On January 9, 2020, the World Health Organization announced the discovery of a novel coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which is the pathogen responsible for coronavirus disease 2019 (COVID-19). The COVID-19 pandemic has spread globally, infecting >500 million people and causing >6.5 million deaths. However, since the development and widespread administration of mRNA vaccines BNT162b2 (Pfizer) and mRNA-1273 (Moderna) encoding the SARS-CoV-2 spike protein as well as the emergence of the omicron strain, the fatality rate, which was estimated to be ∼3% in early 2020, has decreased to <0.3% currently, and severe cases have been greatly reduced (1, 2).

Vaccination against SARS-CoV-2 has shown a high preventive effect worldwide, especially for reducing severe illness and death. This vaccination induces a robust Ab response, particularly the emergence of neutralizing Abs (3–5). However, serum Ab levels drop markedly several months after SARS-CoV-2 infection and vaccination (6). Conversely, persistence and the significance of immunological memory have not been fully investigated.

Mutations frequently occur in the spike glycoprotein in SARS-CoV-2 variants, which can alter viral transmission and immune recognition (7, 8). Omicron variants have amino acid mutations that are concentrated in the receptor-binding domain (RBD) of the spike protein, and there is a significant concern that the initial vaccines based on the Wuhan strain may be less effective in preventing infection. However, vaccination could establish immunological memory in B and T cells (5, 9, 10), and it is expected that the third or fourth booster vaccination will strongly reduce COVID-19 severity. However, immune responses after SARS-CoV-2 vaccination vary widely among individuals, and the interaction between humoral and T cell immunity in defense against COVID-19 has not been fully elucidated.

The humoral response by B cells, specifically Ab production and class switching, is closely regulated by follicular Th (Tfh) cells. Tfh cells play a central role in facilitating germinal center reactions and promoting differentiation of cognate B cells for Ab secretion (9). CXCR5⁺ circulating Tfh (cTfh) cells are present in human blood and subdivided into cTfh1, cTfh2, cTfh17, and cTfh1/17 subsets (11, 12). Although cTfh subsets have been reported to be associated with neutralizing Ab titers in patients with COVID-19 (13, 14), the relationship between cTfh and humoral responses in vaccinated subjects has not been reported.

In the current study, we comprehensively examined T and B cell immunological memory before and after the third BNT162b2 vaccine booster injection in healthy Japanese volunteers. B cell memory was assessed using FACS for surface expression levels of anti-RBD IgG Abs, and the ELISPOT assay was used to measure the number of B cell that produce anti-RBD Abs. T cell memory was also examined using FACS analysis of surface activation markers and ELISPOT assay for IFN-γ after culturing PBMCs with spike protein peptide pools. We investigated memory B and T cell interactions using several approaches. We also evaluated the effects of vaccination from various angles, including cross-reactivity of memory cells to omicron variants.

This study was approved by the Ethics Committee at Keio University School of Medicine (20200063) and conducted in compliance with the tenets of the Declaration of Helsinki. Informed consent was obtained from all participating individuals.

Forty-three healthy individuals from Keio University and Hospital, Tokyo, Japan, were enrolled, and their demographic information is shown in Table I. All healthy individuals had no known history of any significant systemic diseases, including autoimmune disease, diabetes, kidney or liver disease, or malignancy. They were vaccinated three times with a 3-wk interval between the first and second vaccinations and an 8-mo interval between the second and third vaccinations. Eighty-eight patients who had COVID-19, which was diagnosed using approved RT-PCR tests for SARS-CoV-2 with nose swabs or saliva, and who were hospitalized at Keio University Hospital between April and December 2020 were enrolled (15). All patients were infected with the Wuhan-hu-1 strain or the same strain with D614G mutation (15). Patient information is listed in Table II and Supplemental Table I. Blood samples from patients were collected at outpatient visits over 6 mo to 1 y.

Blood was diluted with PBS and then gently loaded onto the Lymphoprep (Serumwerk Bernburg AG, Bernburg, German) layer with a density of 1.007 ± 0.001 g/ml (20°C) followed by density gradient centrifugation (820 × g, 25°C, 30 min). Plasma samples were aliquoted and stored at −20°C after density gradient centrifugation. Cells were washed with PBS containing 0.5% BSA and 2 mM EDTA and then cryopreserved in Cellbanker 1 Plus (Takara Bio, Kusatsu, Japan) at −80°C until use.

Whole spike protein and the spike protein RBD were prepared as described previously (16, 17). Briefly, the RBD was inserted into pcDNA3.4 with a streptavidin binding peptide (SBP) tag at the C terminus and produced using the Expi293 Expression System (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer’s instructions. The supernatant containing the RBD was collected and purified using Streptavidin Sepharose High Performance beads (Cytiva, Tokyo, Japan). The protein purity was determined by SDS-PAGE, and the concentration was determined using a bicinchoninic acid Protein Assay Kit (Thermo Fisher Scientific). Biotinylated SARS-CoV-2 spike RBD, His, and Avi tags of both Wuhan and omicron types were purchased from ACROBiosystems (Newark, DE).

The RBD that we made was diluted to 5 μg/ml in PBS and coated onto flat 96-well plates (442404; Thermo Fisher Scientific) overnight. When using commercially sourced RBD, 10 μg/ml of streptavidin (S4762; Sigma-Aldrich, St. Louis, MO) was coated overnight, and RBD at the concentration of 0.5 μg/ml was applied on streptavidin-coated plates after washing with PBS. The plates were blocked with blocking buffer (Bethyl Laboratories, Montgomery, TX) for 30 min at room temperature. After washing with PBS containing 0.05% Tween-20, plates were incubated with plasma diluted 1:1000–5000 using diluent buffer (blocking buffer with 0.05% Tween-20) for 2 h at room temperature. After washing, plates were incubated with Peroxidase-AffiniPure F(ab′)₂ Fragment Goat Anti-Human IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) at a 1:5000 dilution for 1 h. After the final washing, plates were incubated with TMB Substrate Set (BioLegend, San Diego, CA) for 10 min. Reactions were stopped with H₂SO₄, and the OD at 450 nm was measured. Recombinant human SARS-CoV-2 spike S1 IgG lyophilized Ab (clone AM009105; BioLegend) was used as a standard.

The 15-mer peptide pools for SARS-CoV-2, PepTivator SARS-CoV-2 Prot_S (130-126-700), S1 (130-127-041), and S+ (130-127-311) were obtained from Miltenyi Biotec (Bergisch Gladbach, Germany) and mixed equally to use for stimulation.

As peptide pools derived from the omicron variant, the PepTivator SARS-CoV-2 Prot_S B.1.1.529 Mutation Pool (130-129-928; Miltenyi Biotec) and SARS-CoV-2 Spike Glycoprotein B.1.1.529-Omicron (RP30121; GenScript, Piscataway, NJ) were used with the simultaneous control of the wild-type, PepTivator SARS-CoV-2 Prot_S B.1.1.529 WT Reference Pool (130-129-927; Miltenyi Biotec) and SARS-CoV-2 Spike Glycoprotein-crude (RP30020; GenScript). A 10-aa overlapping peptide mixture pool was prepared, the details of which will be published elsewhere.

A human IFN-γ ELISpot Plus kit (Mabtech AB, Nacka Strand, Sweden) was used in accordance with the manufacturer’s instructions. PBMCs were thawed and washed with culture medium (RPMI 1640 containing 10% FBS, 2 mM l-glutamine, 1 mM sodium pyruvate, nonessential amino acids, 10 mM HEPES, penicillin-streptomycin, and 2-ME). After incubation with culture medium for 30 min, 3 × 10⁵ cells was added to each well in the ELISPOT plates, which were blocked with culture medium, in the presence of 0.1 μg/ml (∼60 nM) peptide pool. In some experiments, PBMCs were cultured with oligopeptide mixtures for 1–2 d, and then cells were applied into ELISPOT plate wells. The spots were tested using detection Ab and streptavidin–alkaline phosphatase. Spots were read using a CTL-ImmunoSpot S5 Series Analyzer (Cellular Technology Limited, Shaker Heights, OH). The spots >0.0051 mm² (vaccinated volunteers) or 0.0033 mm² (infected patients) were counted as positive, and the numbers of spots with peptide stimulation were subtracted by those without peptide stimulation.

We followed standard ELISPOT assay for B cell memory using the Human IgG Single-Color ELISPOT Assay (Cellular Technology Limited) (18). For polyclonal activation, 2 × 10⁵ PBMCs were cultured for 4 d with 1 μg/ml R-848 (resiquimod; MedChemExpress, Monmouth Junction, NJ), an agonist of TLR7/8, in the presence of 5 ng/ml recombinant human IL-2 (PeproTech, Cranbury, NJ) and applied onto the ELISPOT plate, which was coated with anti-human Igκ and anti-human Igλ capture Abs and then blocked as described in the manufacturer’s instructions. After overnight culture, cells were washed out, and 0.5 μg/ml SBP-tagged RBD or biotinylated RBD was added. To detect total IgG-secreting cells, anti-human IgG detection Ab was used. Anti-RBD Ab among the total IgG was detected using streptavidin–alkaline phosphatase and substrates in the kit. Spots >0.0018 mm² detected by SBP-tagged RBD and >0.0044 mm² detected by biotinylated RBD or anti-human IgG were counted as positive.

The activation-induced marker (AIM) assay was performed as previously described (19). PBMCs were cultured in the presence of 60 nM spike peptide pool or 1 μg/ml recombinant spike protein for 1–2 d. After the incubation with human FcR Blocking Reagent (Miltenyi Biotec) and fixable viability dye (FVD)-eFlour 780 (BioLegend), cells were stained with CD4 Alexa Fluor-700 (RPA-T4; BioLegend), CD8a-PerCP/Cy5.5 (HIT8a; BioLegend), CD137-allophycocyanin (4B4-1; BD Biosciences, Franklin Lakes, NJ), and OX40-PE/Cy7 (Ber-ACT35; BioLegend) Abs. For memory and cTfh analysis, CXCR5-BB515 (RF8B2; BD Biosciences), CXCR3-Brilliant Violet (BV) 421 (G025H7; BioLegend), and CCR6-PE (11A9; BD Biosciences) were stained as previously described (19). cTfh cells were divided into cTfh1 (CXCR5⁺CXCR3⁺CCR6⁻), cTfh2 (CXCR5⁺CXCR3⁻CCR6⁻), and cTfh17 (CXCR5⁺CXCR3⁻CCR6⁺) subsets.

To detect anti-RBD IgG Ab-expressing B cells, the thawed PBMCs were incubated with 10 μg/ml SBP-tagged RBD or 1 μg/ml biotinylated RBD plus human FcR Blocking Reagent in PBS containing 1% FBS, 2 mM EDTA, and 0.04% NaN₃ for 30 min at 37°C. After washing, the cells were stained with a mixture of streptavidin-BV410 (BioLegend) and streptavidin-PE/Cy7 (Invitrogen), mAbs against CD20-allophycocyanin (2H7; BioLegend), CD3-allophycocyanin/Cy7 (HIT3; BioLegend), CD11b-eFluor-780 (M1/70; Invitrogen), IgG-BV510 (G18-145; BD Biosciences), and CD19-PE (HIB19; BioLegend), and FVD-eFluor 780 (Thermo Fisher Scientific). Stained samples were applied on an FACSCanto II (BD Biosciences) or CytoFLEX S (Beckman Coulter, Brea, CA) and analyzed using FlowJo10.5.3 (Tree Star, Sheffield, U.K.).

PBMCs (3 × 10⁵ cells) were cultured with 60 nM of a 15-aa oligopeptide mixture of the spike protein (S-peptide) or 1 μg/ml recombinant whole spike protein (S-protein) for 4 d in the presence of 5 ng/ml recombinant human IL-2 in 96-well round-bottom plates and then washed with culture medium twice. Culture was continued for another 3 d without S-peptide or S-protein. Culture supernatant was harvested, and the anti-RBD Ab concentration was measured using ELISA.

PBMCs were cultured for 24 h as described above, and total RNA was purified using the ReliaPrep RNA Miniprep System (Promega, Madison, WI). cDNA was obtained using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystem Waltham, MA). Real-time PCR was performed by the CFX Connect Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA) with SsoAdvanced Universal SYBR Green Supermix (Bio-Rad Laboratories). The following primers were used: human (h)HPRT forward, 5′-TGAGGATTTGGAAAGGGTGT-3′; hHPRT reverse, 5′-CCTCCCATCTCCTTCATCAC-3′; hIL21 forward, 5′-GTGAATGACTTGGTCCCTGAA-3′; hIL21 reverse, 5′-AAGCAGGAAAAAGCTGACCA-3′; hIL4 forward, 5′-TCAAAACTTTGAACAGCCTCA-3′; hIL4 reverse, 5′-CTTGGAGGCAGCAAAGATGT-3′; hIFNG forward, 5′-TGCCAGGACCCATATGTAAA-3′; hIFNG reverse, 5′-TCCATTATCCGCTACATCTGAA-3′; hIL10 forward, 5′-CTGGGGGAGAACCTGAAGA-3′; and hIL10 reverse, 5′-GGCCTTGCTCTTGTTTTCAC-3′.

The numerical data were statistically analyzed and visualized using Prism 9 software (GraphPad Software, San Diego, CA). Multiple data were analyzed with Kruskal–Wallis test and followed by Dunn multiple-comparisons test. Unpaired data were analyzed using the Mann–Whitney U test. Differences with p values <0.05 were considered significant: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 (NS for p ≥ 0.05). The relationship between variables was studied using a simple linear regression model in the software R (https://www.R-project.org/).

Serum Ab levels and B cell memory were evaluated from 43 volunteers listed in Table I at the following three time points: 3 wk after the second vaccination (V2_3w); 8 mo after the second vaccination (V2_8m); and 3 wk after the third vaccination (V3_3w). As reported (6, 20), plasma anti-RBD Ab levels drastically increased at V2_3w compared with those of prevaccination and then decreased to almost undetectable levels at V2_8m. Approximately 3 wk after the third vaccination (V3_3w), Ab levels recovered to similar or even higher levels compared with those at V2_3w (Fig. 1A, Supplemental Fig. 1A).

To measure RBD-specific memory B cells (RBD-MBCs), we established the following two methods: FACS analysis and ELISPOT assay using purified RBD containing an SBP tag (17). As shown in (Fig. 1B and Supplemental Fig. 1B, RBD-specific IgG-expressing B cells were detected with two different RBD fluorochromes after CD19⁺ and IgG⁺ gating, as reported (21). RBD-specific IgG-expressing B cells could be detected at V2_3w (0.263 ± 0.197% in CD19⁺IgG⁺ cells) and increased up to 8 mo after the second vaccination (0.535 ± 0.281%) (Fig. 1B, 1C, Supplemental Fig. 1C). Because most of the RBD-specific IgG-expressing B cells exhibited CD19⁺IgG⁺CD20^high memory phenotypes (Fig. 1D), we confirmed them as RBD-MBCs. RBD-MBCs further increased after the third vaccination (V3_3w) (1.759 ± 1.055%).

Next, we evaluated anti-RBD Ab-secreting B cells by the ELISPOT assay. To facilitate differentiation of memory B cells to Ab-secreting plasma cells, PBMCs were cultured with R848, a potent agonist of TLR7/8 (22, 23). Without R848 stimulation, no or few spots appeared, indicating that most spots were derived from the Ab from plasma cells that were converted from memory B cells in the blood (Fig. 1E). The number of RBD-specific Ab-secreting B cells (RBD-ASCs) detected using an RBD-SBP tag was counted and normalized by dividing by the total number of IgG-secreting B cells detected using anti-IgG Ab (Fig. 1E, 1F). Consistent with RBD-MBCs detected by FACS, RBD-ASCs were detected at V2_3w (0.137 ± 1.139% per IgG-secreting cells), progressively increased at V2_8m (0.510 ± 0.408%), and profoundly enhanced by the third vaccination (1.489 ± 1.114%) (Fig. 1F, Supplemental Fig. 1D). The frequencies of RBD-MBCs and RBD-ASCs were well correlated (Pearson correlation r = 0.817) (Fig. 1G). Consistent with other reports (9, 10), these data indicate that although the Ab levels against SARS-CoV-2 had severely decreased 8 mo after the second vaccination, memory B cells were circulating in the blood, and they were progressively increased by the booster vaccination.

Next, to evaluate T cell response, IFN-γ ELISPOT assay and AIM⁺ FACS assays were performed after 1–2 d of PBMC stimulation with 15-mer or 10-mer peptide pools of the SARS-CoV-2 spike protein (24). (Fig. 2A shows representative IFN-γ spots that were detected after 1 or 2 d of culture that appeared at V2_3w, decreased at V2_8m to about half the level, and then were restored after the third vaccination (Fig. 2B, Supplemental Fig. 1E for 1-d culture, and Supplemental Fig. 1F for 2-d culture). The activation markers OX40 and CD137 were also examined using the AIM assay after culture with the 15-mer S-peptide. Activation markers were upregulated in CD4⁺ T cells, but there was minimal upregulation in CD8⁺ T cells (Fig. 2C, 2D), suggesting that S-peptide stimulated mostly CD4⁺ T cells. In contrast, 10-mer peptide pools induced IFN-γ expression from CD8⁺ T cells but not from CD4⁺ T cells after 2 d of culture (Supplemental Fig. 1G), indicating that 10-mer peptides are more suitable than S-peptide for detection of Ag-specific CD8⁺ T cells. After stimulation with the 10-mer peptide pool, IFN-γ spots were also highly maintained regardless of the third booster vaccination (Fig. 2E, 2F). Thus, we confirmed that both CD4⁺ and CD8⁺ T cell memories were maintained after 8 mo with a relatively low decline, and they recovered to the initial levels by the booster vaccination.

cTfh cell subsets were reported to be associated with Ab response in COVID-19 convalescent patients (14, 21, 25, 26). We observed CXCR5⁺ cTfh subsets in OX40⁺CD137⁺ activated CD4⁺ cells after S-peptide stimulation for 2 d at V2_3w, an early time point in the induction of immune response. As shown in (Fig. 2G, cTfh1 (CXCR3⁺CCR6⁻), cTfh17 (CXCR3⁻CCR6⁺), cTfh2 (CXCR3⁻CCR6⁻), and cTfh1/17 (CXCR3⁺CCR6⁺) populations were determined, and the frequency of each subset is shown in (Fig. 2H. Correlation between these subsets and RBD-ASCs was examined at the V2_3w and V3_3w time points. A positive correlation between the cTfh1 subset and RBD-ASCs at V3_3w (Fig. 2I) was observed, but not at V2_3w (Supplemental Fig. 1H). Conversely, cTfh17 was negatively correlated with RBD-ASCs at V3_3w but not at V2_3w. No correlation between cTfh2 and RBD-ASCs was found. Thus, a higher cTfh1 response may result in more memory B cells. These data are consistent with those of a previous report showing a correlation between cTfh1 and B cell memory in case of infection (25).

Next, we attempted to make B cells produce the Abs in vitro. We expected memory B cell activation and differentiation into Ab-producing plasma cells by in vitro culture in the presence with memory CD4⁺T cells and T cell Ags. PBMCs were cultured with the S-peptide or S-protein for 4 d in the presence of IL-2, and the culture was continued for another 3 d without S-peptide or S-protein. The culture supernatant was removed, and the anti-RBD Ab concentration was measured using an ELISA. As shown in (Fig. 3A, anti-RBD Ab was detected in several, but not all, PBMCs from vaccinated volunteers. Ab was more frequently detected in samples that were stimulated with S-peptide than by the S-protein, and only a few PBMC samples produced anti-RBD Ab in response to both stimulations. The most frequent Ab production was observed in PBMCs 3 wk after the third vaccination (V3_3w), and lowest Ab production was observed at 8 mo after the second vaccination (V2_8m).

To clarify the relationship between the in vitro Ab-producing ability and memory T and B cells, we separated the samples at V3_3w into two groups, a high Ab-producing group (E003, E005, E014, E015, E019, E021, E024, E025, E027, E029, E030, and E031: no stimulation, 3.446 ± 8.819 ng/ml; S-peptide stimulation, 28.910 ± 23.554 ng/ml; and S-protein stimulation, 7.008 ± 12.213 ng/ml) and a low Ab-producing group (E001, E004, E007, E010, E011, E012, E016, E017, E018, E020, E023, and E026: no stimulation, 1.517 ± 4.176 ng/ml; S-peptide stimulation, 0.655 ± 0.822 ng/ml; and S-protein stimulation, 0.360 ± 0.433 ng/ml), and Ab production and T cell and B cell responses were analyzed. Abs were highly produced when stimulated with the S-peptide than with the S-protein stimulation in the Ab-producing group (Fig. 3B), suggesting that TCR stimulation of CD4⁺ Th cells plays a critical role in this culture system. When the numbers of memory B cells (RBD-ASCs) are compared between the two groups, the high Ab group contained slightly more RBD-ASCs than the low Ab group (Fig. 3C). The high Ab-producing group exhibited a higher T cell response ability (Fig. 3D: without S-peptide; 5.10 ± 4.37 and 9.00 ± 5.36 spots in the low and high group, respectively; with S-peptide, 137.67 ± 62.90 and 266.11 ± 117.10 spots). These data suggest that, in addition to a high B cell memory, T cell responses play important roles in Ab production in vitro.

We also evaluated memory B and T cells in 88 convalescent COVID-19 patients who had asymptomatic, mild, moderate, severe, or critical symptoms, which are listed in Table II. Information about each patient is presented in Supplemental Table I. Although asymptomatic participants provided only one blood sample, most other participants provided blood samples two or three times at various intervals. All severe and critical patients and 4 of 39 moderate patients received systemic steroid treatment. We measured RBD-specific B cells with FACS (Fig. 4A) and memory T cells with IFN-γ ELISPOT assay (Fig. 4B) using samples from 30 to 80 d postinfection. Memory B cells were barely detectable in infection and/or vaccination-free individuals and asymptomatic patients (Supplemental Fig. 2A). Although memory B cells were detected at the convalescent stage in all mild, moderate, severe, and critical patients, the frequency of RBD-specific B cells was lower than that in vaccinated samples (Figs. 1C, 4A). RBD-specific T cell memory was detected at the convalescent stage in moderate, severe, and critical patients (Fig. 4B, Supplemental Fig. 2B). Although the frequencies of IFN-γ–positive memory T cells vary among patients, some of these patients have levels that are comparable to those in vaccinated patients (Fig. 2B). These data suggest that SARS-CoV-2 infection resulted in less efficient MBC generation and memory T cell induction at certain levels compared with vaccination.

Longitudinal analysis of RBD-MBC and T cell frequencies was performed as a function of days after PCR-confirmed infection (Fig. 4C, 4D for overall patients, Supplemental Fig. 2C, 2D for patients categorized by symptom severity). Similar to vaccination, RBD-MBCs increased progressively, whereas memory T cells decreased slowly. Thus, there is less difference in the RBD-MBC and T cell frequencies among the symptom categories in the samples >80 d postinfection (Supplemental Fig. 2E, 2F). These data are consistent with previous reports showing the development and persistence kinetics of memory B and T cells elicited by natural infection with SARS-CoV-2 (24, 25).

Vaccination with the parental Wuhan-type SARS-CoV-2 has been shown to develop Abs that can cross-react with other variants, such as β and Δ variants (27–30). As expected, the serum Ab titer at V3_3w was significantly lower against the o BA.1 variant RBD, at ∼30% of the levels shown in response to the parental Wuhan strain RBD (Fig. 5A). To evaluate whether memory cells are cross-reactive against the omicron variant, we examined RBD-specific memory T cells using the IFN-γ ELISPOT assay (Fig. 5B) and RBD-MBCs using the ELISPOT assay (Fig. 5C) and FACS (Fig. 5D). As reported (31, 32), T cell memory responded equally to both the Wuhan and omicron-type spike peptide pools (Fig. 5B). We used two commercially available peptide pools that contain only mutated peptides or whole spike peptide pools, but both showed similar results.

Unexpectedly, the number and frequency of RBD-MBCs detected using the ELISPOT and FACS assays were not severely reduced against the omicron RBD (69.4% [V2_8m] and 77.9% [V3_3w], (Fig. 5C; and 88.3% [V2_8m] and 56.0% [V3_3w], (Fig. 5D). Representative ELISPOT data showed that WT and omicron RBDs did not react with PBMCs collected before vaccination (Supplemental Fig. 3A). These results are consistent with others’ reports (33, 34). As shown by a representative FACS profile (Supplemental Fig. 3B), the mean fluorescence intensity was lower for omicron RBD than for the Wuhan-type RBD (38.7% [V2_8m] and 40.1% [V3_3w], (Fig. 5E). These data indicate that although memory B cells elicited by the BNT162b2 vaccine are reactive to the omicron RBD, their avidity to the omicron RBD may be lower than that to the Wuhan-type RBD, and therefore, the Ab titer against the omicron RBD is low (Fig. 5A). Because the T cell response was normal against omicron, infection with the omicron variant or vaccination with the omicron-type spike protein may induce affinity maturation that is high enough to protect against an omicron infection.

Although it is known that memory B and T cells emerge as a result of BNT162b2 vaccination, the interaction between B and T cells is not well understood. We found that existing Abs gradually decrease after vaccination and infection, but memory B cells producing anti-RBD Abs progressively increase in the plasma of vaccinated and infected individuals, which is similar to other reports (9, 29, 33, 34). Memory B cells persisted up to 8 mo after the second injection, and they further increased after the third booster. The decrease in serum Ab levels is probably due to a short lifespan of the Ab-producing plasma cells. However, it is not clear why memory B cells continue to increase in the absence of infection after the second vaccination. One possibility is that the amount of spike protein produced by the mRNA vaccine is so high that B cells are still differentiating into plasma cells 3 wk after the second vaccination; therefore, memory B cells require longer time to accumulate. As shown recently, memory T and B cells are present in the bone marrow, spleen, lung, and multiple lymph nodes for up to 6 mo postinfection (35, 36). It is possible that the production of B cell memory is ongoing and that it takes time for them to emerge into the blood. To verify such possibilities, it would be necessary to measure both the IgM-type memory B cells and the IgG type and to measure memory B cell levels in the lymph nodes. However, memory B cells persist for a longer time than expected after vaccination; they expand with booster vaccination, and they probably also expand with natural omicron infection. In contrast, memory T cells that suppress severe disease can respond to both Wuhan and omicron to the same degree, and they live at least >8 mo. Thus, fourth and fifth booster vaccination is questionable at least for young healthy people who acquired sufficient memory B and T cells by the third vaccination.

It has been very difficult to control the time of the blood sample collection from infected individuals at hospital outpatient, and we just estimated the period from infection date to blood sample collection date. Even with these limitations, we estimated a persistence of memory cells in patients, which suggests a similar tendency, increase in memory B cells, and decay in memory T cells, to the memory cells derived from vaccination.

Recent studies indicate that omicron infection to unvaccinated individuals shows the limited neutralization Ab of only omicron itself; however, infection to vaccinated individuals induces higher neutralization Ab titers against all SARS-CoV-2 variants (37, 38). This observation is reinforced by our finding that memory B cells induced by Wuhan-type vaccine still can bind to omicron RBD, but just with low avidity. It is likely that more common memory B cells that can neutralize both Wuhan and omicron strains may be selected or undergo affinity maturation after the third vaccination (29). The number of MBCs likely increases after vaccination because they are being generated in germinal centers that can last for months (36). The affinity maturation supporting this has also been measured (29).

To detect RBD-specific B cells in this study, we performed FACS analysis and ELISPOT assay. FACS analysis detects the expression of surface Ig bound to RBD, whereas ELISPOT assay can detect cells that actually secrete anti-RBD Abs. Although this system may detect Ab-secreting cells other than IgG, there is a positive correlation between the frequency of RBD-MBCs detected by these two methods. The ELISPOT method has a significant advantage in the number of cells required for the measurement (10⁵ cells) compared with FACS analysis, which requires a large number of PBMCs. This is an important issue when there are limited samples of human origin.

Although much of the focus on vaccine efficacy and protective effects has focused on the role of neutralizing Abs, T cell responses play an important role in the resolution of infection. Memory T cells persist but gradually decline after vaccination or infection, as described by other investigators (9, 39). Even after the third booster, IFN-γ–producing memory T cells recovered to the level of the second vaccination, but they did not exceed these levels. Although T cell memory generally persists even a decade after vaccination (40), it is possible that this number of memory T cells is sufficient to suppress severe injury by virus infection. Additionally, memory Tcells maintain reactivity to a relatively wide range of viral Ags such as omicron variants. This is consistent with the fact that BNT162b2 vaccine administration maintains a low protective effect against omicron but not against severe disease (41, 42). Again, the effect of additional booster vaccination should be carefully evaluated.

An important role of Th cells is to provide helper signals to Ab-producing B cells. In this study, we found that CXCR3⁺CCR6⁻ cTfh1 correlated positively with B cell memory, whereas CXCR3⁻CCR6⁺ cTFh17 showed a negative correlation. This is consistent with the importance of IFN-γ from cTfh1 for the class switch to IgG. cTfh, especially cTfh1, was similarly reported to correlate positively with neutralizing Abs in COVID-19 convalescent patients (13, 14). Additionally, in vaccinated patients, an early response of cTfh may influence the subsequent Ab response (26). In addition to these observations, we have now established a system to induce B cell differentiation by stimulating memory CD4⁺ T cells in vitro. Because in vitro Ab production was dependent on stimulation with S-peptide, it is highly likely that this response is T cell–dependent. To show T cell dependence more directly, experiments using purified T and B cells would be necessary. We examined IL-21, IFN-γ, and IL-4, which are representative Tfh cytokines in this in vitro system; however, we could not find statistically significant differences between Ab high and low groups (Supplemental Fig. 1I). Further study is necessary to define what factors are important for the differentiation of ASCs from memory B cells. We noticed that in vitro Ab production of samples 8 mo after the second vaccination (V2_8m) was much lower than the V2_3w samples, even though memory B cell levels are higher in V2_8m samples than in V2_3w samples. Three weeks after vaccination may still stimulate innate immunity sufficient for memory cTfh cell activation, which is necessary for memory B cell activation and plasma cell differentiation. Further detailed analysis will reveal the function of memory T cells in activating and amplifying B cell memory.

In this study, we also evaluated memory B and T cells induced by infection and compared them to those induced by vaccination. Because three variants of concern strains, α, β, and γ, were not detected in airport quarantine until December 2020, most patients whom we examined were infected with the Wuhan-hu-1 strain or the variant having only D614G mutation (15). This is ideal for comparing the immune memory of patients infected with the Wuhan strain with that of individuals inoculated with BNT162b2 designed for the Wuhan strain. Future investigation will be done to compare memory responses of vaccinated and infected individuals to the omicron variant.

At the beginning of the CIVID-19 pandemic in 2020, it was speculated that Asian races might have a kind of infection resistance factor, because the number of infected cases was low in Asian countries, including Japan. However, from July to August 2022, Japan had the highest number of infections in the world, >200,000/d, and it is highly likely that there is no Asian-specific infection-resistant factor, at least against the omicron strain. In contrast, the response and memory-forming ability of Japanese to mRNA vaccines were not different from those in Europe and the United States, and no unique characteristics of Japanese patients were found in terms of vaccine response.

Although immunological memories are considered to be induced similarly by infection and vaccination (11), there seems to be a difference when we carefully compared the data; memory T cell induction was not as low with infection, but B cell memory was much lower than after the third vaccination. This reason is not clear at present; however, presence of cross-reactivity of memory T cells to circulating “common cold” coronavirus and SARS-CoV-2 (19, 43). Unlike memory B cells, memory T cells appear to slowly decrease, and they do not show a drastic increase by booster vaccination. Although other studies have shown that patients with severe disease have the strongest memory response (35), our present study showed that patients with moderate disease showed the strongest memory response. All of the patients with severe disease in our study received systemic administration of steroids, which may have inhibited the development of memory cells. However, a robust recall response in both B cell and T cell memory is expected by natural infection or further vaccine boosting.

Finally, we examined cross-reactivity with omicron mutants. Vaccine-induced T cell memory was as responsive to the Wuhan-type spike protein peptide as it was to the omicron-type peptide with similar efficiency. Although the BCRs of memory B cells bound omicron-type RBDs, their affinity was probably severely reduced, as indicated by the decrease in fluorescence intensity of the RBD signals in the FACS analysis. However, it is possible that infection with omicron or vaccination with omicron-type recombinant spike protein may induce affinity maturation by helping memory Tfh cells to produce more effective Abs.

We thank Yukiko Tokifuji, Noriko Yumoto, and Yasuko Hirata (Keio University) for providing technical support.

M.M. reports equipment, drugs, or supplies provided by MBL Corporation, Sysmex Corporation, and Qiagen. Y.U., M.W., and M.M. have patents pending to Keio University. M.T. and H.S. receive royalties from MBL Corporation for SARS-CoV-2 ELISA kits. The other authors have no financial conflicts of interest.