Baby hamster kidney (BHK)-21, a BHK cell line (ATCC CCL-10), cells were cultured in Roswell Park Memorial Institute 1640 medium containing 5% fetal bovine serum (FBS) (Gibco). The BHK-21 cells were stably transduced with the lentiviral vector harboring angiotensin-converting enzyme (ACE2) and selected with puromycin (Sigma) at 10 μg/mL to obtain BHK-ACE2 cells, as described previously (22). Human embryonic kidney 293 (HEK293), an HEK cell line (ATCC CRL-1573), cells were grown in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% FBS. SARS-CoV-2 The SARS-CoV-2 WT strain (hCoV-19/Taiwan/4/2020) and B.1.1.529.1 (hCoV-19/Taiwan/16804/2021, BA.1 variant), and B.1.1.529.5 (hCoV-19/Taiwan/689423/2022, BA.5 variant) were provided by the Taiwan Centers for Disease Control, Ministry of Health and Welfare, and propagated using Vero E6 cells supplemented with 2% FBS. Passage-3 virus was used for all the studies described here. Viral stocks were contamination-free, and viral titers were determined by plaque assay, followed by storage of aliquots at −80°C until further use in experiments.

Fifty adult participants (aged 27–63 years) with good physical health (mild-to-moderate well-controlled comorbidities were permitted) were enrolled between March 2022 and December 2022 in the Tri-Service General Hospital in Taiwan. According to the primed COVID-19 vaccines they had received, the participants were subdivided into three groups: AAM (n = 18), AMM (n = 16), and MMM (n=16). Participants in AAM had received two standard prime doses of AZ (0.5 mL/dose containing 5 × 10¹⁰ viral particles via intramuscular injection) followed by a single dose of Mod (100 μg administered at 0.5 mL via intramuscular injection). The median durations between the two doses of AZ and the single booster of Mod were 35 and 89 days, respectively. Participants in AMM had received one standard prime dose of AZ (0.5 mL/dose containing 5 × 10¹⁰ viral particles via intramuscular injection) followed by two doses of Mod (100 μg administered at 0.5 mL via intramuscular injection). The median durations between the one dose of AZ and two doses of Mod were 80 and 92 days, respectively. Participants in MMM had received one standard prime dose of Mod (100 μg administered at 0.5 mL via intramuscular injection) followed by two doses of Mod (100 μg administered at 0.5 mL via intramuscular injection). The median durations between the one dose of Mod and two doses of Mod were 80 and 92 days, respectively. Blood samples were collected 180-190 days after receiving the three vaccine doses.

Serum SARS-CoV-2 anti-receptor-binding domain (RBD) and anti-nucleocapsid immunoglobulin G (IgG) antibodies were assessed by enzyme-linked immunosorbent assay (ELISA) as previously described (23). In brief, ninety-six-well plates were coated with SARS-CoV-2 RBD and Nucleocapsid protein (5 μg/mL), incubated at 4°C overnight: (1) wild-type (WT): SARS-CoV-2 (COVID-19) Spike RBD protein, His tag (active) (GTX136090-pro, GeneTex), (2) BA.1: SARS-CoV-2 (COVID-19) Spike RBD Protein, Omicron/BA.1 variant, His tag (GTX136716-pro, GeneTex), (3) BA.2: SARS-CoV-2 (COVID-19) Spike RBD Protein, Omicron/BA.2 variant, His tag (GTX136905-pro, GeneTex), (4) BA.4/5: SARS-CoV-2 (COVID-19) Spike RBD Protein, Omicron/BA.4/5 variant, His tag (GTX137098-pro, GeneTex), (5) BQ.1: SARS-CoV-2 (COVID-19) Spike RBD Protein, Omicron/BQ.1 variant, His tag (GTX137879-pro, GeneTex), (6) XBB.1: SARS-CoV-2 (COVID-19) Spike RBD Protein, Omicron/XBB.1 variant, His tag (GTX138115-pro, GeneTex), and (7) Nucleocapsid: SARS-CoV-2 (COVID-19) Nucleocapsid Protein, His tag (GTX135592-pro, GeneTex). After incubation, the plates were washed once with wash buffer and blocked with 10% bovine serum albumin (BSA) (Cyrusbioscience) per well for 2 h at room temperature. Heat-inactivated serum was serially diluted fourfold with 1% BSA at 1:50 and added to the wells, and the plates were incubated for 2 h at room temperature. After washing, anti-human IgG-horseradish peroxidase (1:100,000) (GTX26759, GeneTex) was added to the plates and incubated for 1 h at room temperature. After washing, 3,3’,5,5’-tetramethylbenzidine (TMB) substrate (Invitrogen) was added to each well and incubated for 5 min, and the reaction was stopped with 1 M sulphuric acid. Optical density (OD) was measured at 450 nm using an ELISA reader (BioTek). Endpoint titers were established using a nonlinear sigmoidal four-parameter logistic (4PL) model curve fit to calculate the reciprocal serum dilution that reached the OD450 values of the pre-pandemic sera + 3 SD.

A pseudovirus carrying the variant of concern (VOC) spike protein of SARS-CoV-2 was constructed, as previously described. In brief, 60-μL Lipofectamine 3000^® transfection reagents (ThermoFisher) were mixed with 500 μL serum-free DMEM, sat at room temperature for 5 min, and subsequently mixed with the following three DNA plasmids that were diluted in 500 μL serum-free DMEM for another 20 min: pLAS3w-FLuc-Ppuro (9.5 μg), pCMV-Δ8.91 (Gag-Pol provider, 6.5 μg), and the indicated spike plasmids (4.5 μg): pcDNA3.1_spike_del19 (Addgene #155297), SARS-CoV-2 Omicron Strain S gene Human codon_pcDNA3.1(+) (B.1.1.529/BA.1) (GenScript # MC_0101274), pcDNA3.3 SARS2 Omicron BA.2 (Addgene #183700), pCAGGS SARS-CoV-2 BA.4/5 Spike (Addgene #186031), pCAGGS SARS-CoV-2 BQ.1.1 Spike (Addgene #193710), and pCAGGS SARS-CoV-2 XBB Spike (Addgene #195287). This DNA–lipofectamine mixture was co-transfected to HEK-293T cells (4 × 106 cells per 10-cm dish) and incubated at 37°C in a 5% carbon dioxide incubator. After overnight incubation for 16 h, the transfected cells were replenished with fresh medium for subculture. At 48 h post-transfection, the pseudovirus containing culture medium was collected by centrifugation at 1,000 × g for 10 min to removes unwanted cells or large debris, followed by passing the clarified medium through a 0.45-μm filter (Millipore Corporation. Billerica, MA, USA). Virus can be stored at 4°C for immediate use or frozen at –80°C. Pseudovirus titers were determined using a p24 ELISA kit (Takara Bio).

BHK-21/ACE2 cells were seeded at a density of 4 × 10⁴ cells/well in 24-well plates 16 h before the experiment. For neutralization assay, 40 μL of heat-inactivated sera was started with a 1:16 dilution in complete medium containing 2% FBS, followed by two-fold serial dilutions in duplicate samples, and then incubation with 40 μL of pseudovirus (1 ng p24) for 1 h at 37°C. On the day of infection, the cells were washed twice with PBS, and 100 μL of serum–virus mixture was added to the cells and incubated for 48 h. The cells were quenched by adding 100 μL of Bright-Glo luciferase substrate (Promega) directly to each well, and the luciferase activity was measured using a Synergy H4 luminometer (BioTek). Background values, monitored from uninfected cells, were consistently below 400 relative luminescence units, and pre-pandemic sera were used as the negative control for the neutralization assay. Sera diluted at 1:16 provided results in the range of the background relative to light unit levels. A pVNT50 > 1:16 serum dilution was regarded as positive.

Serum samples were heated at 56°C for 30 minutes to inactivate complement; two-fold serial dilutions, starting at a concentration of 1:5, were mixed with an equal volume of viral solution containing 100 PFU of SARS-CoV-2. The serum-virus mixture was incubated for 1 hour at 37°C in a humidified atmosphere with 5% CO2. After incubation, the mixture at each dilution was added to Vero E6 cells and incubated at 37°C for 1 hour. Cells were subsequently cultured with DMEM containing 2% FBS and 1.4% methylcellulose for 72 hours. After culturing, plaques were stained and counted. Neutralizing antibody titers were defined as the reciprocal of the maximum dilution of serum that reduced the virus titer by 50% compared to the negative control sera, and PRNT50 below 1:5 serum dilution was considered negative.

Peripheral blood mononuclear cells (PBMCs) derived from the participants were isolated from anticoagulant-treated whole blood using Ficoll-Paque™ PLUS density gradient medium (Cytiva #17144003) as previously described (23). For isolate PBMCs, blood diluted with PBS was gently layered over an equal volume of Ficoll in a Falcon tube and centrifuged for 30 min at 400 × g without braking. The PBMCs were gently removed using a Pasteur pipette and added to a warm medium or PBS to remove any remaining platelets. The pelleted cells were counted, and the percentage viability was estimated using trypan blue staining. The isolated PBMCs were stored in liquid nitrogen until use in assays.

Cryopreserved Peripheral blood mononuclear cells (PBMCs) were thawed and washed with RPMI 1640 medium containing human AB serum (Sigma-Aldrich), L-glutamine (Gibco), and HEPES buffer (Gibco). The cells were then plated in U-bottom 96-well plates at 1 × 10⁶ cells per well and incubated overnight. PBMCs were stimulated with peptide pools of SARS-CoV-2 spike WT (PepTivator^® SARS-CoV-2 Prot_S, Miltenyi Biotec), BA.1 variant (PepTivator^® SARS-CoV-2 Prot_S B.1.1.529/BA.1 Mutation Pool, Miltenyi Biotec), or BA.5 variant (PepTivator^® SARS-CoV-2 Prot_S B.1.1.529/BA.5 Mutation Pool, Miltenyi Biotec), along with co-stimulation antibodies. Positive control cells were stimulated with phytohemagglutinin (ThermoFisher), and negative control cells were treated with DMSO in PBS.

After 24 hours of stimulation, PBMCs were washed and stained with Zombie Red Fixable Viability Dye (BioLegend), followed by staining with antibodies against specific markers. The markers included anti-hCD3 BV510, anti-hCD4 BV605, anti-hCD8 FITC, anti-hCD134 (OX40) BV421, anti-hCD137 APC, anti-hCD69 PE-Cy7, anti-hCD197 (CCR7) PerCP-Cy5.5, and anti-hCD45RA APC-Cy7 (BioLegend). Flow cytometry analysis was conducted using the Attune™ NxT Flow Cytometer, and the gating strategy can be found in Supplementary Figure 3 .

The amount of antigen-specific interferon (IFN)-γ- or interleukin (IL)-2-secreting T cells was evaluated by enzyme-linked immunosorbent spot (ELISpot) assays as previously described (22). The mean SFC value of duplicate peptide pool-stimulated PBMCs was calculated and normalized by subtracting the mean of the negative control replicates (control medium), and the cut-off value for background T-cell responses was defined as the mean SFC value of seronegative PBMCs derived from healthy unvaccinated donors + 3 SD (8.5 SFC/106 PBMCs). The results are expressed as SFC per million PBMCs.

This study was approved by the Institutional Review Board of Tri-Service General Hospital (TSGHIRB No. B202105097). Informed consent was obtained from all the enrolled participants. Work with infectious SARS-CoV-2 has been approved by the Institutional Biosafety Committee (IBC) and was performed in the high biocontainment Biosafety Level 3 (BSL-3) facilities of the Institute of Preventive Medicine (IPM), National Defense Medical Center (NDMC), which are approved for such use by the Taiwan Centers for Disease Control, under license D1-109-0030#1123 and D1-111-0017#2028 to institutional guidelines.

Statistical analyses were performed using GraphPad Prism version 5. Anti-RBD IgG titers and pVNT₅₀ values are presented as medians and IQRs. A nonlinear sigmoidal 4PL model was used to determine the endpoint titers of anti-RBD IgG and pVNT₅₀ in each serum sample. The statistical significance of the endpoint titers and pVNT₅₀ was measured among experiments using a one-way analysis of variance (ANOVA) with Tukey’s multiple comparison test. One-way ANOVA performed the statistical significance of the T cell AIM assay with a Tukey’s post-hoc test for multiple pairwise comparisons.

Blood samples were collected from 50 vaccinated individuals between March 2022 and December 2022. In group AAM, 10 (55.6%) were male and 8 (44.4%) were female, with a median age of 43.5 years. In group AMM, 8 (50%) were male and 8 (50%) were female, with a median age of 44.5 years. In group MMM, 9 (56.3%) were male and 7 (43.7%) were female, with a median age of 43.5. Only 38 participants above were involved in the T cell AIM assay. All 50 participants were monitored to be uninfected by using anti-nucleocapsid ELISA. Detailed information can be found in Table 1 .

Inducing antibodies targeting the receptor-binding domain (RBD) of the SARS-CoV-2 spike protein is crucial for blocking virus infection. Thus, we determined the RBD-specific IgG titers of individuals with different vaccination schedules against different SARS-CoV-2 variants using ELISA ( Figure 1 ). In the AAM group, the mean endpoint titers against WT, BA.1, BA.2, BA.4/5, BQ.1, and XBB.1 were 16,353, 3,630, 2,004, 2,011, 1,051, and 729.8 respectively ( Figure 1A ). Similarly, in the AMM group, the mean endpoint titers against WT, BA.1, BA.2, BA.4/5, BQ.1, and XBB.1 were 15,773, 3,304, 1,123, 804.8, 427.7, and 304.5 respectively ( Figure 1B ). The MMM group exhibited the highest antibody titers against the WT spike, with mean endpoint titers of 78,352, 6,983, 6,359, 1,490, 781.2, and 545.1 against WT, BA.1, BA.2, BA.4/5, BQ.1, and XBB.1, respectively ( Figure 1C ). Overall, across all groups, we observed a decline in RBD-specific IgG titers against the BA.1, BA.2, BA.4/5, BQ.1, and XBB.1 variants compared to the WT (Tukey’s multiple comparisons test p.adj <0.0001). When comparing the vaccine regimens, the MMM group had the highest antibody titer against the WT, BA.1, and BA.2 variants compared to the AAM or AMM group. However, when confronted with the BQ.1 and XBB.1 variants, the MMM group did not display a comparable antibody titer to the AAM or AMM group, suggesting that the antibodies induced by the WT vaccine may lose their ability to effectively recognize the Omicron BQ.1 and XBB.1 spike proteins, which harbors a significant number of mutations in the RBD.

Next, we measured neutralizing titers against the emerging variants by using the pseudovirus and infectious virus neutralization assay ( Figure 2 ). The AAM group showed the median of 50% pseudovirus neutralization titer (pVNT50) values of 2,104 (WT), 424.6 (BA.1), 141.7 (BA.2), 109.2 (BA.4/5), 66.18 (BQ.1.1), 66.12 (XBB.1). Similarly, the AMM group exhibited mean pVNT50 values of 1,867 (WT), 448.9 (BA.1), 154.5 (BA.2), 111.8 (BA.4/5), 62.53 (BQ.1.1), and 52.18 (XBB.1), and the MMM group had mean pVNT50 values of 8,330 (WT), 1,064 (BA.1), 897.4 (BA.2), 531.0 (BA.4/5), 114.1 (BQ.1.1), and 80.60 (XBB.1) ( Figure 2A ). Significant differences in neutralizing activity were observed when comparing the WT-induced polyvalent sera against the tested variants (WT, BA.1, BA.2, BA.4/5, BQ.1.1 and XBB.1) across different vaccination combinations (AAM, AMM, and MMM group), with the most substantial reduction in neutralizing antibody titers against the XBB.1 variant in the AAM, AMM, and MMM group, with a mean fold decrease of 31.8, 35.8 and 103, respectively ( Figure 2A ). Besides, the MMM group had the best neutralizing effect against the WT, BA.1, BA.2 and BA.4/5 compared to the AAM or AMM group ( Supplementary Figures 1A-D ). However, when faced with the BQ.1.1 and XBB.1 variants, which has accumulated a vast number of mutations in the RBD region, the neutralization titers are significantly reduced, regardless of the vaccination schedules ( Supplementary Figures 1E, F ).

In addition, we further evaluated the neutralizing activity by using the infectious virus neutralization assay ( Figure 2B ). The mean 50% plaque reduction neutralization test (PRNT50) values for the AAM group were 3,057 for the WT strain, 247.2 for the BA.1, and 93.6 for the BA.5 variants. Similarly, for the AMM group, the mean PRNT50 values were 3,041 for WT, 259.5 for BA.1 and 100.7 for BA.5. The MMM group exhibited mean PRNT50 values of 9,086 for WT, 847.5 for BA.1 and 310.5 for BA.5, showing the best neutralizing effect against the WT, BA.1 and BA.5 compared to the AAM or AMM group ( Figure 2B ). A significant reduction in the mean PRNT50 values for the BA.1 and BA.5 variant compared to WT was observed across different vaccine combinations. The AAM, AMM, and MMM groups showed approximately 12-fold, 11-fold, and 10.5-fold reductions for BA.1, and 32-fold, 30-fold, and 29.3-fold reductions for BA.5, respectively, in neutralizing activity compared to WT ( Figure 2B ). The detailed pVNT50 and PRNT50 values were summarized in the Supplementary Table 1 . Overall, our findings suggest that the Omicron variants may possess specific mutations or epitope changes that affect the recognition and binding of neutralizing antibodies induced by the evaluated vaccination regimens.

Cellular immune responses, including CD4+ and CD8+ T cell responses, play a crucial role in protecting against different SARS-CoV-2 variants infection. Thus, we next assessed spike-specific cellular immune responses using the AIM (activation-induced markers) assay ( Figure 3 ; Supplementary Figures 3 , 4 ) and analyzed the memory phenotype within SARS-CoV-2-reactive AIM+ CD4 and CD8 T cells ( Figure 4 ; Supplementary Figure 5 ). The PBMCs from 38 vaccinated individuals (16 from AAM, 12 from AMM, and 10 from MMM) within the participants from previous results were involved in the following experiments, allowing for simultaneous tracing of both humoral and cellular responses in the vaccinated individuals.

Flow cytometry was used to detect AIM expression on CD4+ (OX40+CD137+) and CD8+ (CD69+CD137+) T cells ( Figure 3A ; Supplementary Figure 3A ). The mean percentages of AIM+ CD4+ T cells stimulated with spike WT, BA.1, and BA.5 peptide pools were 0.42%, 0.18%, and 0.21%, respectively, while the mean percentages of AIM+ CD8+ T cells under peptide stimulation were 0.21%, 0.18%, and 0.15% ( Figures 3B, C ). These results indicated that compared to WT, CD4 T cell reactivity against BA.1 and BA.5 showed significant reductions, with a 2.3-fold and 1.9-fold decrease, respectively. Besides, we observed a similar reduction in the AAM group when comparing the Omicron variants to the WT, not only in AIM expression on CD4+ T cells but also in the secretion of IFN-γ and IL-2 detected by ELISPOT assays ( Supplementary Figures 2 , 3 ). Unlike the results from the AIM+ CD4+ assay, there was no significant difference detected in the mean percentages of AIM+ CD8+ T cells among the WT, BA.1, and BA.5 spike treatment groups ( Figure 3C ), suggesting that CD8+ T cells showed considerable cross-reactivity to Omicron even 8 months after vaccination. Moreover, we found that the population of AIM+ CD8+ T cells remained consistently higher in the MMM group than in the AAM group, whereas similar results could not be found in AIM+ CD4+ T cells ( Supplementary Figure 4 ). The detailed T cell reactivities were summarized in the Supplementary Table 2 .

We further explored memory phenotypes within AIM+ CD4+ and CD8+ T cell subsets among vaccinated individuals with diverse regimens, including naïve T cells (T_Naïve), central memory T cells (T_CM), effector memory T cells (T_EM), and CD45RA+ effector memory T cells (T_EMRA) in response to WT, BA.1 or BA.5 variants ( Figure 4 and Supplementary Figure 5 ). 8 months after receiving three vaccine doses, the CD4+ spike-specific memory T cells were primarily T_CM and T_EM, while CD8+ subsets were mainly T_EM and T_EMRA, excluding naive cells. In CD4+ T cells, we noted a decline in the naïve population and a preferential enrichment in the T_CM and T_EM populations compared to the bulk counterpart ( Figures 4A, B ). Conversely, for CD8+ T cells, the difference was less pronounced. However, WT spike-specific AIM+ CD8+ T cells exhibited significantly higher proportions of T_EMRA and lower T_EM percentages than the bulk population ( Figures 4A, C ). When focusing on distinctions between the variants, we noted a rise in the population of T_EM cells in response to Omicron variants compared to WT, coupled with a decrease in T_Naïve cells. The detailed memory phenotypes of T cell subsets were summarized in the Supplementary Table 3 .

Across varying vaccine combinations, memory subsets showed consistent trends: diminished naïve T cells and increased effector memory in CD4+ T cells ( Supplementary Figure 5 , left panels), with fewer differences observed in CD8+ T cells between original WT and variants ( Supplementary Figure 5 , right panels). In summary, these findings offer additional insights into T cell memory phenotypes in vaccinated individuals primed by the ancestral WT spike protein, revealing unique subsets that may contribute to long-term cross-reactive immunity against Omicron variants.