IMPACT OF THE BIVALENT COVID-19 BOOSTER ON SPIKE-SPECIFIC T CELL RESPONSES

Abstract

The emergence of SARS-CoV-2 variants during the COVID-19 pandemic prompted updates to the original vaccines to improve immunogenicity against novel SARS-CoV-2 variants. Since 2022, vaccination guidelines introduced updated mRNA boosters, starting with a bivalent formulation that included the original and Omicron BA.4/BA.5 spike sequences, which emerged in late 2021. In this study, we characterized the spike-specific T cell response following the bivalent booster vaccination in humans. We performed an in-depth profiling of T cells, assessing activation markers, cytokine production and the impact of previous COVID-19 infection in the response to different spike variants. Overall, our results are consistent with the bivalent booster having a limited influence on spike-T cell responses, with similar magnitude and functionality observed toward the Omicron BA.4/BA.5 variant and the ancestral spike. Nonetheless, the booster proved to be beneficial for immunocompetent individuals with poor or declining T cell responses, increasing the frequency of specific T cells in blood.

INTRODUCTION

The health crisis during the COVID-19 pandemic accelerated the development and approval of multiple vaccines in record time, representing an unprecedented biomedical achievement [1–3]. As a result, numerous subsequent studies have defined the immunological responses of the approved vaccines in parallel [4–12].

At the end of 2021, the SARS-CoV-2 variant of concern B.1.1.529, named “Omicron”, displayed an unprecedented number of mutations, increased transmissibility, and substantial escape from existing neutralizing antibodies [13]. Subsequently, several Omicron subvariants emerged, including BA.2, BA.4, BA.5, and XBB, each exhibiting further mutations that contributed to increased transmissibility and partial immune evasion. In response, mRNA vaccines were updated in September 2022 to a bivalent formulation, including a copy of the spike protein from the Omicron BA.4/BA.5 variants alongside the original sequence. One year after, in September 2023, recommendations were revised to endorse the use of a new monovalent vaccine formulation targeting the Omicron XBB.1.5 variant. Currently (2025), the approved vaccines are monovalent formulations targeting the SARS-CoV-2 JN.1 lineage (https://www.fda.gov/vaccines-blood-biologics/industry-biologics/covid-19-vaccines-2025-2026-formula-use-united-states-beginning-fall-2025).

The bivalent booster vaccine showed superior relative efficacy against both infection and severe disease during the circulation of BA.4, BA.5, and subsequent Omicron subvariants, compared to equivalent doses of the monovalent vaccine [14, 15]. However, other authors have questioned its superiority compared to the monovalent formulation based on affinity and levels of neutralizing antibodies [16–18]. The first reports addressing T cell immunity to the booster showed that the bivalent booster did not substantially augment T cell responses, and elicited comparable magnitude of antigen-specific responses against the ancestral strain, Omicron [17] and the subsequent variant XBB.1 [19]. Follow-up studies have similarly described negligible differences in T cell responses between monovalent and bivalent boosters [20]. Those observations are consistent with the notion that T cells are highly cross-reactive and less susceptible to immune evasion at the population level than antibodies [21]. While breakthrough infection [22] and bivalent vaccination [23] can boost T cell responses and induce new epitopes recognizing variant-specific mutations, they primarily expand an existing pool of memory T cells that are cross-reactive to both the ancestral and BA.4/5 spike proteins.

Our study seeks to evaluate not only the magnitude but also the comprehensive functional profile of T cell responses induced by the bivalent BA.4/5 booster. To achieve this, we employed a multiparametric flow cytometry approach integrating activation markers, phenotypic profiling, and cytokine production in a longitudinal cohort of fully vaccinated donors who were sequentially boosted with both monovalent and bivalent formulations. Using the same methods, we also compared T cell responses to the spike proteins from the ancestral strain, Omicron BA.4/5, and XBB.1 variant, and evaluated the effect of prior SARS-CoV-2 infection on the vaccine-induced immunity.

MATERIALS AND METHODS

Human sample donors

Experiments were approved by the institutional review board of IRB #: 20-1007 Protocol Title: SARS-CoV-2 Serological Network (SeroNet). Donor recruitment and blood processing were performed by Feinstein Institutes for Medical Research (Manhasset, NY).

Blood samples were collected between 12/12/2021 and 12/22/2022 from 35 healthy adults who received the standard two dose immunization with two doses of the original mRNA COVID-19 vaccine, either BNT162b2 COVID-19 Vaccine (2023-2024 formula) (developed by Pfizer/BioNTech), or mRNA-1273 COVID-19 Vaccine (2023-2024 formula) (developed by Moderna). The second vaccination was between 20-40 days after the first. In all cases, this was followed by third (“booster”) immunization of the same vaccine within 180-319 days following second dose. Blood from the pre-monovalent booster timepoint was collected on the same day as the vaccine administration, except for four participants (1 collected 1 day prior, 2 collected 3 days prior, and 1 collected 12 days prior). Samples post-monovalent booster were collected 23 to 36 days after the vaccination (median value= 28 days).

Subjects underwent a second booster immunization with a bivalent WT/BA.4/5-adapted vaccine preparation. Either BNT162b2-WT/BA.4/5 or mRNA-1273.214 were administered between 245-395 days after the last vaccination with the original booster vaccines. Blood from pre-booster timepoint was collected the same day of vaccine administration. Samples post-bivalent booster were collected 17 to 48 days after the shot, median value= 31 days.

Peripheral blood mononuclear cells (PBMCs) isolation

Peripheral blood was collected in CPTTM tubes (BD Biosciences). After centrifugation cells were aspirated and washed in cold RPMI and 40% FCS. Cells were then frozen in RPMI 1640 Complete media, 20%FBS, 7.5% DMSO using a controlled rate freezer (Thermo Scientific Model 7450 CryoMed) in aliquots of 10⁶ cells per ml.

Peptide synthesis and Megapool preparation

Peptide megapools (MPs) were designed and synthesized to allow simultaneous testing of a large number of epitopes. Briefly, individual peptides were synthesized as crude material (TC Peptide Lab, San Diego, CA) and solubilized in dimethyl sulfoxide (DMSO) at a concentration of 10-20 mg/mL. Peptides were pooled into their corresponding MPs, sequentially lyophilized and resuspended in DMSO at a final concentration of 1 mg/mL [24].

Spike MPs comprised of overlapping 15-mers peptides by 10 amino acids spanning the entire sequence of the ancestral Wuhan sequence, and a selection of the SARS-CoV-2 Omicron subvariants BA.4/5 and XBB.1, allowing for simultaneous detection of CD4⁺ and CD8⁺ T cells responses [21, 22]. The list of mutations compared to ancestral were previously reported in [25, 26].

CD4⁺ and CD8⁺ cytomegalovirus (CMV) MPs were used as a control and are based on experimentally defined epitopes available in the Immune Epitope Database (IEDB) [24]. The CD4RE MP, a mixture of experimental defined epitopes from non-spike region of SARS-CoV-2 was used to identify previous COVID-19 infection. In samples with limited number of cells, we prioritized stimulation with ancestral > BA.4/5 > CD4RE > XBB.1 MPs.

Activation-induced markers (AIM) and intracellular cytokine staining assay

Cryopreserved cells were thawed by diluting them in 10 mL pre-warmed complete RPMI 1640 with 5% human AB serum (Gemini Bioproducts) and 50 U/mL benzonase (Sigma-Aldrich, St. Louis, MO) and centrifuged at 1200 rpm for 7 min. Cells were counted and plated in 96-well U bottom plates with 1x10⁶ cells per well.

The complete list of antibodies and reagents used in this panel can be found in Table 2. Fifteen minutes prior to the addition of peptides stimuli, cells were treated with 5 μg/mL anti-human CD40 antibody and stained with anti-human CXCR5 Brilliant Violet 421 and CCR7 Brilliant Violet 711 antibody (Table 2), as previously described [5, 11, 27]. PBMCs were then stimulated for 24 hours with either ancestral, BA.4/5 or XBB.1 spike MPs or CMV MP at 1 μg/mL. Phytohaemagglutinin (PHA) at 1 μg/mL was used as positive control. Equimolar amount of DMSO was used as negative control for each sample. A control donor with known response to CMV was used to ensure reproducibility between experiments.

Table 2.

List of reagents

Reagent	Clone	Source	Catalog #	Dilution / Concentration
Golgi-Plug		BD Biosciences	555029	1:1000
Golgi-Stop		BD Biosciences	554724	1:1000
LIVE/DEAD Fixable Blue Dead Cell Stain		Thermo Fisher Scientific	L23105	1:50
Human BD Fc Block		BD Biosciences	564220	1:20
Brilliant Stain Buffer Plus		BD Biosciences	566385	1:10
CD40 anti-human	HB14	Miltenyi	130-094-133	0.5 μg/mL
Brilliant Violet 421 anti-human CD185 (CXCR5)	J252D4	Biolegend	356920	1:400
Brilliant Violet 711 anti-human CD197 (CCR7)	G043H7	Biolegend	353228	1:160
BV605 anti-human CD69	FN50	BD Biosciences	562989	1:250
PE-Cy5 anti-human CD137	4B4-1	Biolegend	309808	1:500
BUV805 anti-human CD8	SK1	BD Biosciences	612889	1:200
Brilliant Violet 510 anti-human CD16	3G8	Biolegend	302048	1:200
Brilliant Violet 510 anti-human CD14	63D3	Biolegend	367124	1:200
Brilliant Violet 510 anti-human CD20	2H7	Biolegend	302340	1:200
PE-Cy7 anti-human CD134 (OX40)	ACT35	BD Biosciences	563663	1:160
BUV395 anti-human CD3	UCHT1	BD Biosciences	563546	1:100
Brilliant Violet 570 anti-human CD45RA	HI100	Biolegend	304132	1:100
PE anti-human CD279 (PD-1)	eBioJ105	Thermo Fisher Scientific	12-2799-42	1:80
cFluor B548 anti-human CD4	SK3	Cytek Biosciences	R7-20043	1:50
Alexa Fluor 700 anti-human Granzyme B	GB11	BD Biosciences	560213	1:1000
APC anti-human TNF alpha	Mab11	Thermo Fisher Scientific	17-7349-82	1:625
FITC anti-human IFN gamma	4S.B3	Thermo Fisher Scientific	11-7319-82	1:500
PE-Dazzle 594 anti-human IL-10	JES3-19F1	Biolegend	506812	1:200
BUV737 anti-human IL-2	MQ1-17H12	BD Biosciences	612836	1:80
APC-efluor-780 anti-human CD154 (CD40 Ligand)	24-31	Thermo Fisher Scientific	47-1548-42	1:80

During the last 4 hours of stimulation, cells were treated with Golgi-Plug and Golgi-Stop, along with anti-human CD69 BV605 and CD137 PE-Cy5 abs (Table 2). PBMCs staining was performed as previously reported [5, 11, 27, 28]. Cells were washed, incubated with human Fc-block, and stained with LIVE/DEAD viability marker for 15 min according to the manufacturer’s instruction (Thermo Fisher). Subsequently, cells were washed and stained with anti-human antibodies (CD3 BUV395, CD4 cFluor B548, CD8 BUV805, CD14 Brilliant Violet 510, CD16 Brilliant Violet 510, CD20 Brilliant Violet 510, OX40 PE-Cy7, CD45RA Brilliant Violet 570, PD-1-PE and CCR7 Brilliant Violet 711 (Table 2) in PBS with 3% Fetal Bovine Serum and 10% Brilliant Staining Buffer Plus for 30 min at 4 °C. Finally, cells were fixed in 4% paraformaldehyde (Sigma-Aldrich) for 10 min at 4 °C, followed by washing with permeabilization buffer (PBS with 5 μg/mL saponin, 10% bovine serum albumin and 1% azide) at room temperature (RT).

Intracellular staining was performed as previously described [5, 11, 27]. Briefly, anti-human antibodies (Granzyme B Alexa Fluor 700, TNF alpha APC, IFN gamma FITC, IL-10 PE-Dazzle 594, IL-2 BUV737 and CD40 Ligand PerCP-eFluor 710) were resuspended in permeabilization buffer with 10% human AB serum and Brilliant Staining Buffer Plus for 30 min at RT. Samples were acquired on a Cytek Aurora flow cytometer (Cytek Biosciences, Fremont, CA).

Quantification and statistical analysis

Samples were analyzed using FlowJo 10.9 software. A representative gating strategy is shown in Supplementary Figure 1A–E. AIM positive-specific T cells are defined for co-expressing the markers OX40⁺CD137⁺ for CD4⁺ T cells and CD69⁺CD137⁺ for CD8⁺ T cells as previously described in [5, 28–30]. Functional spike-specific T cells are defined by their production of either IFN-γ, TNF-α, IL-2, IL-10 or granzyme B, along with the expression CD40L for CD4⁺ T cells or CD69 for CD8⁺ T cells as previously utilized in [5, 21] .

Data was normalized with a minimal level of sensitivity of 0.005 [5, 28]. The percentage of virus-specific T cell responses is shown as background (DMSO) subtracted data. A limit of detection (LOD) and limit of sensitivity (LOS) were calculated for each population. LOD was calculated as the upper 95% Confidence Interval of the of negative control (DMSO) values. LOS was calculated as median + 2 x Standard Deviation (SD) of negative control (DMSO) values. Stimulation Index (SI) was also calculated for each sample and population, as % of response divided by % of response in DMSO control. Responses > than LOS and SI > 2 for CD4⁺ T cells or SI > 3 for CD8⁺ T cells were considered positive [5, 28]. Responses with SI < 2 for CD4⁺ T cells or SI < 3 for CD8⁺ T cells were normalized to LOD. To define multifunctional profile of T cells producing IFN-γ, TNF-α, IL-2, IL-10 and/or granzyme B, the minimal level of response was set to 0.001%. LOD and LOS values were calculated based on the IFN-γ, population. Background (DMSO) subtracted data was calculated, with positive values defined as > 0.005% and SI > 2.

All samples showed high activation response to PHA stimulation (Supplementary Figure 1F). Percentage of positive AIM (OX40⁺CD137⁺) CD4⁺ T cells to CMV incubation in control donor differed less than 20% (coefficient of variation) between experiments (Supplementary Figure 1G).

Statistical analysis and data representation were performed using GraphPad Prism 9.5. The statistical details and plot descriptions are provided in the respective figure legends. Overall, non-parametric Wilcoxon paired tests were used for paired samples when normality assumptions were not met (e.g., same donor over time or under different stimuli), and Mann-Whitney U tests were applied for unpaired comparisons (e.g., infected vs. non-infected donors).

RESULTS

Bivalent COVID-19 booster vaccine cohort

To assess the virus-specific T cell response elicited by the booster vaccination, we recruited 35 adult donors that received three doses of the original mRNA COVID-19 vaccine (either BNT162b2 or mRNA-1273), and one additional boost of the bivalent WT/BA.4/5-adapted mRNA vaccine (either BNT162b2-WT/BA.4/5 or mRNA-1273.214) (Figure 1A–B). Our cohort was limited in size and skewed toward older (range 27-80, median age 59, IQR 16), and predominantly female participants (71% female), which may limit the generalizability of our findings (Table 1).

Figure 1. Effect of monovalent and bivalent COVID-19 boosters on spike-specific circulating T cells.

(A) Schematic representation of monovalent and bivalent COVID-19 mRNA vaccines and (B) experimental design. See methods for detailed information.

(C) Representative gating strategy of spike-specific T cell expressing activation markers, detected after bivalent booster.

(D, E) Percentage of ancestral spike-specific T cell responses measured by activation induced markers OX40⁺CD137⁺ (CD4⁺ cells) (D) or CD69⁺CD137⁺ (CD8⁺ cells) (E). Values are shown as percentages over total CD4⁺ or CD8⁺ T cells in logarithmic scale. Values above the dotted line (LOS) are considered positives or responders. Bold line represents the geometric mean in each time point. Thin grey lines connect paired samples. Statistical differences between paired timepoints were calculated using two tailed Wilcoxon signed rank test, with p-values depicted on top. Fold change mean was calculated from paired samples, as the ratio between post-booster and pre-booster responses.

(F) Correlation analyses of monovalent and bivalent Pre- and Post- booster T cell responses (upper row), measured as percentages of AIM⁺ T cells, and Pre-booster T cell responses and fold change levels after boosters (lower row) (calculated as paired % of response Post-booster / % of response Pre-booster). Dots indicate individual values; black thin line depicts nonlinear regression line. Non parametric Spearman correlation was calculated and plotted on the graph, with r = correlation coefficient and p = statistical significance.

(G, H) Functionality of ancestral spike-specific CD4⁺ (G) and CD8⁺ (H) T cell responses induced by bivalent COVID-19 booster, measured by expression of CD40L (CD4⁺) or CD69 (CD8⁺) and production of IFN-γ, TNF-α, IL-2, IL-10 or granzyme B (secreted-effector⁺ T cells). Values are shown as percentages over total CD4⁺ or CD8⁺ T cells in logarithmic scale. Values above the dotted line (LOS) are considered positives or responders. Green bold lines represent the geometric mean in each time point. Thin grey lines connect paired samples. Statistical differences between paired timepoints were calculated using two tailed Wilcoxon signed rank test, with p-values depicted on top. Fold change mean was calculated from paired samples, as the ratio between post-booster and pre-booster responses.

Table 1.

Characteristics of donor cohort.

Characteristic	Total	BNT162b2 / BNT162b2-WT/BA.4/5	mRNA-1273 / mRNA-1273.214	BNT162b2 / mRNA-1273.214	mRNA-1273 / BNT162b2-WT/BA.4/5
Donors, n (%)	35 (100%)	27 (77%)	6 (17%)	1 (3%)	1 (3%)
Gender, n (%)
Female	25 (71%)	19 (70%)	5 (83%)	1 (100%)	0
Male	10 (29%)	8 (30%)	1 (17%)	0	1 (100%)
Age in years, median (IQR)	59 (16)	59 (16)	58 (20)	27 (0)	67 (0)
Race or ethnicity, n (%)
White	24 (69%)	19 (70%)	3 (50%)	1 (100%)	1 (100%)
Asian	5 (14%)	5 (19%)	0	0	0
Black or African American	3 (9%)	2 (7%)	1 (17%)	0	0
Black or African American, Asian	1 (3%)	1 (4%)	0	0	0
White, Multirace	1 (3%)	0	1 (17%)	0	0
Other	1 (3%)	0	1 (17%)	0	0
Medical history
Without autoimmunity or cancer	26 (74%)	22 (81%)	3 (50%)	1 (100%)	0
With autoimmunity	3 (9%)	1 (4%)	2 (33%)	0	0
Diagnosed with cancer	4 (11%)	3 (11%)	0	0	1 (100%)
With autoimmunity and cancer	2 (6%)	1 (4%)	1 (17%)	0	0

All donors received a complete mRNA vaccination series (two doses of the original vaccine from the same manufacturer), followed by a monovalent booster from the same manufacturer 250 ± 70 days later, and a subsequent bivalent booster 320 ± 75 days after the first booster. In all except two cases, the original monovalent vaccine series and the bivalent booster were from the same manufacturer. 27 donors were immunized with the BNT162b2 + BNT162b2-WT/BA.4/5 combination, while 6 donors were immunized with mRNA-1273 + mRNA-1273.214 vaccines. Of the 2 remaining subjects, one received initial BNT162b2 immunizations followed by mRNA-1273.214 bivalent booster, and one received the original mRNA-1273 vaccines, followed by the BNT162b2-WT/BA.4/5 bivalent booster. Peripheral blood mononuclear cells (PBMCs) were collected on the day of the bivalent booster vaccination, reflecting the T cell response before the booster. A sample post booster vaccination was collected 17-48 days after.

Limited effect of boosters on spike-specific T cell magnitude

Our first goal was to measure ancestral spike-specific CD4⁺ and CD8⁺ T cell responses. For that, we measured the expression of activation-induced markers (AIM) by flow cytometry (Figure 1C), consistent with our prior work characterizing spike-specific T cell responses in [5, 28–30].

Previous to each boost, an existing population of spike-specific CD4⁺ T cells was detected in most donors, as expected (75% responders pre-monovalent, 84% responders pre-bivalent) (Figure 1D). After the administration of each booster, there was a slight, non-significant increase in spike-specific activated T cells, with a mean fold-change of 2.1x for monovalent and bivalent was observed and an increase in the percentage of responders (90% post-monovalent, 91% post-bivalent). However, differences were not statistically significant between pre- and post-boosters (p=0.09 for both comparisons using two-tailed Wilcoxon signed-rank test) (Figure 1D).

Comparable levels of pre-existing spike-specific CD8⁺ T cells were noted in 8% and 10% of donors subsequently receiving, monovalent or bivalent boosters, respectively (Figure 1E). The levels of spike-specific CD8⁺ T cell responders increased to 24% and 31% following booster vaccination. Similar to CD4⁺ cells, we also observed a slight increase in the frequency of activated CD8⁺ T cells, with a mean fold change of 2.3x (monovalent) and 2.1x (bivalent). Interestingly, the increase of spike-specific CD8 cells following the bivalent (fourth dose) but not the monovalent booster (third dose) was statistically significant (p=0.049 and p=0.2, respectively) (Figure 1E). CD69 can be also upregulated on bystander-activated cells, resulting in a higher background than CD4⁺ AIM⁺ (OX40⁺CD137⁺) cells. The subsequent normalization may raise the CD8⁺ AIM⁺ threshold, potentially underestimating the number of donors with positive CD8⁺ responses. Nonetheless, CD69⁺CD137⁺ has been widely used to identify antigen-specific CD8⁺ T cells and was retained here for consistency across studies [5, 28–30].

Overall, both the monovalent and bivalent shots lead to a moderate, though not significant, increase of the percentage of both spike-specific CD4⁺ cells when stimulated with the ancestral spike peptide pool. However, statistically significant higher levels spike-specific CD8⁺ T cells were observed after the bivalent shot. This could be influenced by the pre-existing levels of spike-specific T cells before boosting, as shown by da Silva Antunes et al. (2025) [28].

Memory phenotype of spike-specific T cells remains unchanged after boosters

Next, we characterized the memory phenotype of ancestral spike-specific T cells after bi-or monovalent vaccination. Cells were classified based on their surface expression of CCR7 and CD45RA into Naive, Central Memory, Effector Memory or T cells re-expressing CD45RA (TEMRA) as previously described [11] (Supplementary Figure 2A).

As previously reported, most spike-specific CD4⁺ T cells expressed a central or effector memory profile [11], while the most dominant subset in the spike-specific CD8⁺ T cells was TEMRA cells, followed by effector memory cells [31]. Overall, no significant differences in the frequencies of antigen-specific memory subsets were observed after the bivalent or monovalent booster shot.

Similarly, no changes in fraction of follicular T helper cells (CD4⁺ T cells expressing the markers OX40⁺CD40L⁺CXCR5⁺PD-1⁺) were observed comparing both, the monovalent (p=0.23 using two-tailed Wilcoxon signed-rank test) or bivalent booster (p=0.50) (Supplementary Figure 2B).

Lower levels of pre-existing spike-specific CD4⁺ cells are associated with higher expansion post-booster

To investigate whether the differences in fold change elicited by either of the booster shots were related to the initial number of antigen-specific T cells we compared the percentages of specific CD4⁺ T cells pre- and post-booster vaccination, paired by donor (Figure 1F, upper panel). We found a positive correlation between the percentages of spike-specific T cells before and after boosting for both monovalent (Spearman correlation, r=0.48, p=0.02) and bivalent vaccines (r=0.43, p=0.01). Specifically, donors with the highest percentages of spike-specific CD4⁺ T cells pre-booster also had the highest responses after the shots while donors with low frequencies of spike-specific CD4⁺ cells continued demonstrated the lowest percentages after boosting.

Next, we compared the frequencies of spike-specific CD4⁺ T cells pre-booster to their paired fold-change levels after vaccination (Figure 1F, lower panel). In our cohort, 13 out of 21 participants with measurable fold changes exhibited increased T cell responses (fold change >1) following the monovalent vaccination, and 18 out of 31 following the bivalent vaccination. The analysis revealed a strong negative correlation, particularly after the bivalent booster (Spearman r = −0.67, p < 0.0001), indicating that donors with the lowest pre-existing responses exhibited proportionally greater increases in the frequency of spike-specific T cells following the booster. Conversely, donors with the highest pre-booster responses were generally associated with fold changes below 1. Although similar trends were observed with the monovalent booster, the correlation only significant in the bivalent booster group. These observations applied only to the CD4⁺ subset, as spike-specific CD8⁺ AIM⁺ responses declined more rapidly in blood, being detectable in just 8% of participants before the third booster and 10% before the bivalent booster.

When comparing longitudinal samples from the same individuals, we evaluated whether fold-change levels following the monovalent booster could predict T cell responses to the bivalent booster (Supplementary Figure 2C). However, no significant correlation was observed, suggesting that individual responses to the monovalent booster does not predict responsiveness to the subsequent bivalent booster. No correlations in T cell responses were observed with participant age, sex, or other reported characteristics.

In summary, the bivalent vaccination elicited a greater effect on those donors with lower pre-existing frequencies of spike-specific CD4⁺ T cells prior to the booster.

The bivalent booster enhances the frequency of cytokine-producing T cells

Next, we aimed to characterize the functional profile after mono- and bivalent vaccination. We measured the ability of secreting IFN-γ, TNF-α, IL-2, IL-10 and Granzyme B (GzB) by flow cytometry, in co-expression with the activation markers CD40L (CD4⁺ cells) or CD69 (CD8⁺ cells) (Supplementary Figure 2D), as previously described in our studies characterizing spike-specific T cell functionality [5, 21].

As observed previously, most donors exhibited high levels of cytokine-producing CD4⁺ T cells pre-boosters (Figure 1G), whereas only a subset of donors showed cytokine production by CD8⁺ T cells (Figure 1H). The analysis of individual molecules revealed that the bivalent booster (administered as a fourth shot), but not the monovalent (administered as the third shot), nearly doubled the percentage of spike-specific IFN-γ⁺ CD4⁺ T cells, calculated as either frequencies (from 0.016% to 0.027% geometric mean) (p=0.003, two-tailed Wilcoxon matched-pairs signed rank test) or fold change (2.5x). Similarly, spike-specific TNF-α⁺ CD4⁺ T cells significantly increased post-bivalent boost (0.028% to 0.048% geometric mean; p=0.002; 2.3× fold change), along with a modest rise in spike-specific IL-2⁺ CD4⁺ T cells (0.021% to 0.030% geometric mean; p=0.003). While these changes were observed only with the bivalent booster, they may be attributable to dosing sequence (third vs fourth dose) rather than inherent effects of the monovalent versus bivalent vaccine. No significant changes were detected in spike-specific IL-10⁺ or Granzyme B⁺ CD4⁺ T cells in any of the cohorts.

When analyzing the functional profile of the spike-specific CD8⁺ T cells, statistically significant changes following the bivalent booster were observed for two cytokines following the bivalent booster. The frequency of spike-specific IFN-γ⁺ CD8⁺ T cells doubled after bivalent vaccination (0.014% to 0.030% geometric mean; p=0.03), with the proportion of responders increasing from 16% to 37% (21% absolute increase). Spike-specific TNF-α⁺ CD8⁺ T cells increased significantly after both the monovalent and bivalent boosters (p=0.02 and p=0.03 respectively). In contrast, no differences were observed in IL-2⁺ or IL-10⁺ spike-specific CD8⁺ T cells. We also observed a noticeable, but not statistically significant (p=0.06), increase in Granzyme B⁺ spike-specific CD8⁺ T cells exclusively after the bivalent booster, limited to a subset of donors (8 out of 35; 3% responders pre-booster; 23% post-booster).

In summary, the bivalent booster, administered as a fourth COVID-19 vaccine dose, but not the prior monovalent booster, significantly enhanced spike-specific CD4⁺ T cells producing IFN-γ, TNF-α, and IL-2, as well as IFN-γ⁺ spike-specific CD8⁺ T cells.

Bivalent booster does not enhance T cell polyfunctionality

Next, we performed a combinatorial analysis to characterize the ability of spike-specific T cells to produce one or multiple cytokines (monofunctional vs polyfunctional). Polyfunctional T cells are typically associated with greater functional maturity and improved outcomes against various infections, including COVID-19 [32, 33].

At this resolution, we detected some statistically significant, although modest, changes between pre- and post-booster populations. The bivalent booster increased frequencies of monofunctional IFN-γ⁺ (p=0.046, Wilcoxon matched-pairs signed rank test), TNF-α⁺ (p=0.033) or IL-10⁺ (p=0.004) spike-specific CD4⁺ T cells, as well as a marginally significant increase in polyfunctional CD4⁺ T cells co-expressing IFN-γ, TNF-α, and IL-10 (p=0.049) (Figure 2A). The average proportions of monofunctional or polyfunctional spike-specific CD4⁺ T cells were not substantially altered by either the monovalent or bivalent booster, with the majority of responding cells producing only a single cytokine (58% pre-monovalent, 67% post-monovalent, 57% pre-bivalent, 63% post-bivalent) (Figure 2A, right pie charts).

Figure 2. Effect of monovalent and bivalent COVID-19 boosters on the multifunctional profile of spike-specific T cells.

Multifunctional profiles of spike-specific CD4⁺ (A) and CD8⁺ (B) T cells expressing CD40L (CD4⁺) or CD69 (CD8⁺) and cytokines (secreted-effector⁺ T cells) pre and post monovalent and bivalent booster. Boolean analysis was performed to identify combinations of IFN-γ, TNF-α, IL-2, IL-10 or granzyme B (GzB) expression. Each secreted-effector⁺ profile combination was considered positive with >0.005% and SI >2. Values are shown as percentages over total CD4⁺ or CD8⁺ T cells in logarithmic scale. Values above the dotted line (LOS) are considered positives or responders. Horizontal black lines indicate the geometric mean of each population. Statistical differences between pre and post vaccination paired samples were calculated using Wilcoxon matched-pairs signed rank test, with p-values > 0.05 depicted. Pie charts indicate the average percentage of the sum of cells with 1, 2, and 3 or more functions for pre and post booster samples.

Within the CD8⁺ T cell population, an increase in monofunctional Granzyme B⁺ spike-specific CD8⁺ T cells was observed in some donors (Wilcoxon matched-pairs signed rank test, p=0.031) (Figure 2B). Additionally, we noted an increase in double-positive CD8⁺ T cells co-expressing IFN-γ and Granzyme B (p=0.02), and cells producing TNF-α and Granzyme B (p<0.001) after the bivalent booster.

The bivalent booster did not shift the overall T cell profile toward increased polyfunctionality (Figure 2B, right pie charts). The majority of CD4⁺ T responses were monofunctional, whereas some CD8⁺ T cell subsets expanded after boosting, particularly those producing Granzyme B alone or co-expressing TNF-α and Granzyme B. In conclusion, although the bivalent booster increased certain monofunctional cytokine responses, it did not enhance the overall polyfunctionality of spike-specific T cells.

Comparable T Cell Responses to ancestral, Omicron, and XBB.1 spikes

Our next aim was to compare the T cell responses against the spike proteins of three different SARS-CoV-2 variants of concern (VOC): Ancestral, Omicron and XBB.1. We sought to elucidate whether the bivalent booster could enhance the overall T responses not only against the encoded BA.4/BA.5 variants, but also to XBB.1, a subsequent variant containing additional spike mutations. To assess this, PBMCs were stimulated with spike-derived megapools (MPs) specific to each variant, followed by AIM and cytokine analysis as previously described.

We assessed T cell responses against three VOC before and after the monovalent and bivalent boosters. For both CD4⁺ and CD8⁺ T cells, we compared the responses in terms of magnitude (AIM⁺ cells, Figure 3A), functionality (percentage of cytokine-producing T cells, Figure 3B), and the distribution of polyfunctional profiles (Supplementary Figure 3). As previously demonstrated, no statistically significant differences were found in any of the analyses performed. T cell frequencies and functional responses to the three VOC were comparable, with p-adjusted values ≥ 0.05 (Wilcoxon matched-pairs signed rank test). Similarly, no meaningful differences among variants were found in either fold changes or the proportion of individuals mounting a detectable response.

Figure 3. COVID-19 bivalent booster does not modify the overall spike-specific T cell response among variants.

(A) Percentage of specific T cells recognizing ancestral, Omicron BA.4/5 or XBB.1 SARS-CoV-2 spike, measured by activation induced markers OX40⁺CD137⁺ (CD4⁺ cells) or CD69⁺CD137⁺ (CD8⁺ cells) following bivalent booster.

(B, C) Percentage of specific secreted-effector⁺ CD4⁺ (B) or CD8⁺ T cells (C) recognizing ancestral, Omicron BA.4/5 or XBB.1 SARS-CoV-2 spike following bivalent booster. Values are shown as percentages over total CD4⁺ or CD8⁺ T cells in logarithmic scale. Values above the dotted line (LOS) are considered positives or responders. Horizontal black lines indicate the geometric mean. Statistical differences were calculated using Wilcoxon matched-pairs signed rank test corrected for multiple comparison using the Holm and Šidak method, with adjusted p-values specified for each comparison.

Our results are consistent with previous observations that CD4⁺ and CD8⁺ T cells similarly recognize and are stimulated by spike antigens from SARS-CoV-2 variants, both prior and following administration of the bivalent booster. While the bivalent vaccine was designed to enhance immunity against Omicron BA.4 and BA.5, the results show no overall differences in T cell responses to the variants tested. The bivalent booster did not change the responses to BA.4/5 or XBB.1 spikes compared to the ancestral version, suggesting limited improvement in T cell reactivity toward the new VOC.

Classification of samples based on SARS-CoV-2 exposure before the bivalent booster

Given the collection dates, it is expected for a big fraction of the donors to have had prior exposure to SARS-CoV-2. We aimed to assess whether previous natural infection influences the T cell response to the bivalent booster. We focused on the bivalent booster (31 samples pre-bivalent booster and 35 samples post-bivalent booster) due to insufficient samples or unknown infection status from the monovalent booster timepoints. We used a T cell-based assay previously described in [34], which allows to distinguish T cell responses of subjects based on their SARS-CoV-2 infection and vaccination history with more than 88% of accuracy. Briefly, cells were cultured with either the spike MP, or the CD4-RE MP, composed by experimentally defined non-spike SARS-CoV-2 epitopes. A positive CD4⁺ AIM response toward the CD4-RE MP would reflect activation of antigen-specific CD4⁺ T cells from the memory compartment that developed after a natural SARS-CoV-2 infection.

Accordingly, a total of 27 samples out of 66 (41%) tested positive to previous SARS-CoV-2 infection (Figure 4A). 34 samples (52%) tested only positive to the spike MP. Five samples could not be evaluated using the CD4RE assay because of their limited number of cells.

Figure 4. Previous exposure to SARS-CoV-2 has no effect on the T cell responses following bivalent booster.

(A) COVID-19 clinical classification of pre- and post-bivalent booster samples. CD4⁺ T cell responses to ancestral spike and CD4RE MPs were measured as percentage of AIM⁺ (OX40⁺CD137⁺) and plotted in two dimension comparing Stimulation Index for CD4RE versus ancestral spike MPs. Samples (n=66) were divided in three groups: Vaccinated and unexposed (I-V+) (SI bellow CD4RE threshold 3.3) (n=34), vaccinated and infected (I+V+) (SI > 3.3) (n=27) and missing values (MV) (n=5), which lacked CD4RE values.

(B) COVID-19 clinical classification of the donors. Donors that did not respond to CD4RE MP (SI bellow threshold) before and after the bivalent booster were considered Non-infected (n=12). Donors responding to CD4RE MP pre-bivalent boost (SI above threshold) were considered Infected (n=12). Donors that did not fall into one of those categories were excluded.

(C- F) Comparison of T cell specific responses to ancestral spike based on COVID-19 infection status of the donors: Non-infected (Non), with no detectable virus exposure, or Infected (In), with previous exposure to COVID-19. Percentages of AIM⁺ cells (C,E) and cytokine-secreting cells of interest (D,F) are plotted in logarithmic scale from total CD4⁺ or CD8⁺ cells. Values above the dotted line (LOS) are considered positives or responders. Horizontal black lines indicate the geometric mean. Statistical differences were calculated using two-tailed Mann Whitney U test, with p-values specified for each comparison.

(G) Comparison of specific T cells recognizing ancestral, Omicron BA.4/5 or XBB.1 SARS-CoV-2 spike, separated by COVID-19 infection status of the donors: Non-infected (Non), with no detectable virus exposure, or Infected (In), with previous exposure to COVID-19. Percentages of AIM⁺ cells and cytokine-secreting CD4⁺ or CD8⁺ cells of interest are plotted in logarithmic scale from total CD4⁺ or CD8⁺ cells. Values above the dotted line (LOS) are considered positives or responders. Horizontal black lines indicate the geometric mean. Statistical differences were calculated using Wilcoxon matched-pairs signed rank test corrected for multiple comparison using the Holm and Šidak method, with adjusted p-values specified for each comparison.

A total of 12 (out of 35) donors tested positive prior to the bivalent vaccine administration and were classified as “Infected” (Figure 4B). A total of 12 donors (out of 35) tested always negative to SARS-CoV-2 infection and were classified as “Non-infected”. Donors without complete CD4-RE information and donors that tested positive only at the post-booster timepoint were excluded from the ensuing analysis.

Effect of previous SARS-CoV-2 infection on bivalent booster T cell responses

Ancestral-spike responses were compared between previously infected and non-infected donors. Prior to the booster, infected donors did not exhibit higher CD4⁺ T cell responses to spike; in fact, their responses were slightly lower and more heterogeneous compared to non-infected individuals (p = 0.06, using two-tailed Mann-Whitney U test) (Figure 4C). Additionally, the percentage of responders was lower in the infected group (75%) compared to the non-infected group (92%). However, following administration of the bivalent booster, spike-specific CD4⁺ T cell responses were comparable between the two groups, with similar frequencies of activated cells (p = 0.37) and responder rates. No significant differences were observed in the magnitude of spike-specific CD4⁺ T cells secreting IFN-γ, TNF-α (Figure 4D), IL-2, IL-10, Granzyme B (Supplementary Figure 4A), or in the distribution of monofunctional and polyfunctional CD4⁺ T cell subsets (Supplementary Figure 4C, upper panel).

Similarly, with no discernible differences in the overall magnitude (Figure 4E) or functionality (Figure 4F, Supplementary Figure 4B, Supplementary Figure 4C, lower panel) for spike-specific CD8⁺ T cells, driven by a previous SARS-CoV-2 infection were detected.

Lastly, we assessed whether prior SARS-CoV-2 infection influenced CD4⁺ and CD8⁺ T cell responses to VOC ancestral, BA.4/5, and XBB.1. Donors were stratified based on their infection history, and T cell responses to each variant were compared both before and after administration of the bivalent booster. We observed broadly homogeneous CD4⁺ and CD8⁺ T cell responses across previously infected and non-infected individuals (p-adjusted values ≥ 0.05 using Wilcoxon matched-pairs signed rank test) (Figure 4G). Across all populations of cytokine-secreting spike-specific T cells, only one significant difference was observed: the number of IFN-γ⁺ CD4⁺ T cells responding to Omicron was lower than that responding to the ancestral strain in the non-infected group following the bivalent booster (adjusted p = 0.02, Wilcoxon matched-pairs signed-rank test) (Supplementary Figure 5). However, no differences were observed in the comparison between the ancestral strain and the subsequent variant XBB.1. While the observed difference was statistically significant, its overall impact may be limited.

DISCUSSION

The present study analyzed the human T cell responses elicited by the COVID-19 bivalent booster vaccination. The bivalent vaccine is an updated version of the original mRNA COVID-19 vaccine, containing both the genomic spike sequence of the Wuhan isolate and the Omicron subvariants BA.4 and BA.5. Introduced in mid-2022, the bivalent booster aimed to counteract immune evasion by circulating SARS-CoV-2 Omicron variants. Indeed, clinical data indicate that the bivalent booster offers improved protection against severe outcomes from Omicron infection compared to the monovalent booster [14, 15]. Previous work demonstrated that the bivalent booster increases spike binding-antibody levels [18, 35, 36], neutralizing antibody levels against the Omicron spike remain low [16, 17], with limited affinity for ensuing variants such as XBB.1 [18, 37]. Here, we investigated potential differences induced by the monovalent and bivalent boosters through a detailed phenotypical and functional characterization of spike-specific T cells.

Previous work by us and other support a pivotal role of T cells in limiting COVID-19 severity. While T cell immunity may not prevent infection, it reduces virus replication and protects from severe disease (reviewed in [38]). Recent work following multiple monovalent booster vaccinations has shown that T cell responses plateau after the first two doses and remain relatively stable thereafter [28]. Consistent with these findings, in our cohort we observed high levels of pre-existing CD4⁺ T cells prior to the third and fourth shots (corresponding to the first and second boosters).

In our cohort, the bivalent booster (administered as a fourth COVID-19 dose approximately one year after the previous one) had a limited effect on the frequency of spike-specific T cells in peripheral blood towards the studied variants, consistent with the expected plateau and with the previous reports on BA.1- [39] and BA.4/BA.5-adapted boosters [17, 19]. This work failed to detect differences in T cell responses to SARS-CoV-2 variants across any subset or functional profile. We and others have previously shown that T cells exhibit broad responsiveness to multiple variants of SARS-CoV-2 [21, 40] after monovalent vaccination. Moreover, only a small fraction of the Omicron mutations (studied in variant B.1.1.529) were predicted to impact the T cell epitopes [21]. More recently, the work by Sop et al. (2024) [23] on the bivalent booster epitope mapping demonstrated that the vaccine mostly elicits CD4⁺ T cell clonotypes cross-reactive with both ancestral and BA.4/5 spike epitopes. Even with hybrid immunity, the expansion of novel variant-specific epitopes (only 9%-15% of the total epitope repertoire) occurs alongside a predominant boost of T cells recognizing cross-reactive epitopes shared among variants [22]. Our results confirm that the bivalent booster preserves the overall magnitude of spike-specific T cell responses, regardless of the variant.

Despite the reduced boosting effect (with most fold changes close to 1) observed after multiple vaccinations, as previously reported [28], our data importantly indicate that the bivalent booster has a greater impact in individuals with low levels of circulating spike-specific T cells prior to booster vaccination. In our cohort, these lower pre-boost T cell responses could not be could they be predicted based on the response to the previous booster, or attributed to age or other reported participant characteristics, which may be attributable to the limited size and homogeneity of the cohort. Given that measuring antigen-specific T cells is challenging and not routinely performed in clinical practice, it will remain difficult to predict who will benefit most from booster vaccination based on T cell levels. However, in conjunction with the documented decline in antibody levels over time [41], our findings support the continued benefit of booster vaccination for at least a subset of the population and may justify its broader use as a precautionary measure.

The percentages and levels of observed T cell responses pre- and post-boost were comparable with previous studies of the initial COVID-19 monovalent vaccination (two doses) [5], except for a lower percentage of responders exhibiting positive CD8⁺ AIM⁺ cell responses. Zhang et al. (2022) reported CD8⁺ T cell responses in approximately 50% of individuals six months after full vaccination [5], whereas in our cohort, only 31% of participants exhibited responses following the bivalent booster. This difference may be attributable to inherent methodological variability between studies or intercohort variation, given that our study involved a small and predominantly older female-biased group. Nonetheless, we consistently observe across studies that the fraction of CD8⁺ AIM⁺ responders occurs at a lower frequency than that of CD4⁺ AIM⁺ responders [5].

While we did not observe substantial quantitative differences in spike-specific T cells, we discovered some qualitative differences comparing the monovalent (third dose) and bivalent booster (fourth dose). We investigated potential differences induced by the monovalent and bivalent boosters through a detailed characterization of spike-specific T cells. A primary limitation of this work is that the monovalent and bivalent boosters were administered sequentially (third and fourth doses, respectively), making direct comparisons between their immune responses challenging. Consequently, any differences observed should be interpreted cautiously. On the other hand, this approach also offers valuable insight into how individuals respond to both boosters over time, revealing that the response to the monovalent booster does not necessarily predict the response to the bivalent one.

We observed that the BA.4/BA.5-adapted booster, but not the monovalent, particularly increased the percentage of spike-specific monofunctional CD4⁺ T cells secreting IFN-γ and TNF-α. Again, this effect may reflect the sequential administration rather than the vaccine formulation itself, with no differences observed between responses to the ancestral, Omicron BA.4/5, or XBB.1 spikes. Effective control of SARS-CoV-2 has been strongly associated with Th1-driven responses, induced either by natural infection or vaccination [11, 29, 42, 43]. Our findings indicate that the bivalent booster administered as a fourth dose helps sustain this Th1-skewed immune profile. We also observed an increase in monofunctional IL-10-producing CD4⁺ T cells. While IL-10 has been implicated in limiting immunopathology [44, 45], the increase in these T cells may represent a transient response to the booster.

In contrast to the findings reported by Trieu et al. (2025) [39], we did not detect an increase in CD4⁺ follicular helper T cells following the bivalent booster. Nonetheless, we acknowledge that intracellular detection of CD40L may artifactually increase apparent Tfh frequencies and thereby contribute to differences observed across studies. Our results also differ from Urschel et al. (2024), who reported a predominance of polyfunctional T cells both before and after the bivalent booster administration [51]. These discrepancies could be partly attributable to methodological differences, including the stimulation period (6h vs 24h) or the activation markers selected for co-expression with cytokines (CD69 vs CD40L for CD4⁺ T cells). Instead, our findings align with previous studies of the initial mRNA vaccine series, which likewise observed a predominance of spike-specific monofunctional CD4⁺ T cells [5], using a similar experimental approach and set of markers as in this study.

To our knowledge, this is the first description of the COVID-19 bivalent booster inducing CD8⁺ T cells expressing Granzyme B in a subset of individuals, either alone or in combination with additional effector molecules such as TNF-α. Their contribution in COVID-19 defense has been broadly confirmed [27, 46–49], also in the context of Omicron (variant B.1.1.529) infection in macaques [50].

Our final aim was to evaluate the impact of prior natural SARS-CoV-2 infection on the T cell response to the bivalent booster. Individuals with a history of infection had likely encountered a broader range of viral epitopes and experienced additional antigen exposure. Given the rapid waning of antibody levels, our T cell-based assay has demonstrated greater accuracy than nucleocapsid serology in identifying past infections [34]. We observed largely comparable CD4⁺ and CD8⁺ T cell responses to the bivalent booster against the three VOC, irrespective of infection history, consistent with findings by Urschel et al. (2024), who categorized donors based on self-reported infection or serological data [51]. However, our approach to assessing infection history has limitations. We could not determine the number or timing of infections per individual, nor assess the influence of recent versus earlier infections or their timing relative to the full vaccination schedule. Additionally, the infecting SARS-CoV-2 variant(s) remain unknown. Each of these factors could potentially influence the T cell responses.

As SARS-CoV-2 evolves, with JN.1 prevailing globally and vaccines reverting to a monovalent JN.1 formulation, our study remains important for guiding vaccine development. Spike divergence between BA.4/BA.5 and JN.1 (~15-20 mutations) is smaller than that between the ancestral strain and BA.4/BA.5 (~32-34). Together with prior studies, our findings suggest that the high degree of T cell cross-reactivity across variants makes it unlikely that current vaccine updates will alter cellular immunity, and individuals with robust spike-specific T cell responses are expected to retain effective protection against JN.1. It should be noted that this study did not assess the effects of the bivalent WT/BA.4/5 booster versus a hypothetical monovalent BA.4/5 booster, both administered as fourth doses in a parallel design, and thus, no inferences should be made regarding monovalent versus bivalent vaccine strategies.

The present work demonstrates that, despite minimal changes in the overall T cell responses and no significant differences in the recognition of spike variants, the bivalent booster appears to expand certain circulating Th1 CD4⁺ and cytotoxic CD8⁺ T cell populations. This functional shift may facilitate a faster and more effective response upon re-exposure to the virus. Our results underscore the importance of booster vaccination, especially for individuals with poor or waning T cell immunity, and highlight the need for ongoing immunological surveillance.