Data from participants of three longitudinal SARS-CoV-2 vaccination studies within one framework with similar study design were combined. Two were observational COVID-19 vaccination cohort studies: one focusing on adults 18 to 60 years of age (NL76440.041.21, EudraCT: 2021-001357-31), and one on the ageing population over 60 years of age (NL74843.041.21, EudraCT: 2021-001976-40). In these studies, vaccines were rolled-out per age group from old to young according to the national vaccination campaign. For the primary vaccination series, vaccinees received BNT162b2 (Pfizer/BioNTech), mRNA-1273 (Moderna), ChAdOx1 nCoV-19 (AstraZeneca), or Ad26.COV2.S (Janssen) vaccines; and for booster vaccination either BNT162b2 or mRNA-1273 vaccines were given, as per national policy ( Figure 1A ). Due to limited number of participants vaccinated with ChAdOx1 nCoV-19 and Ad26.COV2.S, the majority of comparisons for statistical purposes have been performed on the BNT162b2 and mRNA-1273 vaccine groups.

The third study concerns an intervention study (Vital) investigating seasonal influenza and pneumococcal conjugate vaccine responses in older adults (≥65 years) compared with younger and middle-aged adults (25-49; 50-64 years). With a protocol amendment, vaccine responses to a primary series of mRNA-1273 followed by a booster dose with BNT162b2 have been added to the study (NL69701.041.19, EudraCT: 2019-000836-24) ( Figure 1A ). Ethical approval was obtained through the Medical Research Ethics Committee Utrecht in all three studies. All participants provided written informed consent. All trial-related activities were conducted according to Good Clinical Practice, including the provisions of the Declaration of Helsinki.

Samples from participants who experienced a SARS-CoV-2 infection prior or within 2 weeks after providing a post-vaccination sample were excluded from analysis in this study. Additionally, participants who experienced an infection in between the two doses of the BNT162b2 or mRNA-1273 primary vaccination series were also excluded. SARS-CoV-2 infection of the participants was determined by self-reporting a positive COVID-19 test (PCR or antigen) and/or anti-Nucleoprotein IgG antibody titers above 14.3 BAU/ml. Additionally, individuals with anti-Spike IgG antibody titers above 10.1 BAU/ml prior to primary vaccination were also defined as infected, as previously reported (20).

When studying effect of age for some analysis the study participants were divided in three different age groups (18-59, 60-69, and 70+ years), which were based on the age-related risk groups defined by the Dutch government that guided the COVID-19 booster vaccination strategy in the Netherlands.

Heparinized blood samples were collected from participants prior to vaccination (P0), 28 days post completion of the primary series (P28), prior to booster vaccination (B0) and 28 days post booster vaccination (B28) ( Figure 1A ). From all blood samples, peripheral blood mononuclear cells (PBMC) were isolated using Lymphoprep (Progen) and cryopreserved at −135°C.

Synthetic overlapping peptide pools covering the complete Spike protein from the different SARS-CoV-2 variants were all purchased from JPT; Ancestral (PM-WCPV-S-1), B.1.1.7 (Alpha; PM-SARS2-SMUT01-1), B.1.351 (Beta; PM-SARS2-SMUT02-1), P.1 (Gamma; PM-SARS2-SMUT03-1), B.1.617 (Delta; PM-SARS2-SMUT06-1), B.1.1.529 BA.1 (Omicron BA.1; PM-SARS2-SMUT08-1). All peptide pools contained 315 15-mers with 11 amino acid overlap. Peptide pools were reconstituted in 50µl DMSO and stored at -20°C until use according to the manufacturers’ instructions.

IFNγ T cell ELISpot was performed as described before (21). PBMCs were thawed and plated in AIM-V medium (Gibco BRL, 12055-083) with 2% heat-inactivated human serum (Sigma, H6914) into each well, at 2x10⁵ cells/well (or 5x10⁴ cells/well for the positive control), of 96-well plates pre-coated with anti-human IFNγ antibodies (5 µg/ml, clone 1-D1K, Mabtech, 3420-3-1000). Subsequently, cells were stimulated in triplicate with DMSO (negative control), PHA (positive control; 1 µg/ml), or Spike peptide pools (0.65 µg/ml per peptide) for 22-24 hours (37°C, 5% CO₂). Next, plates were washed with PBS+0.05% Tween 20 and incubated with anti-human IFNγ biotinylated antibody (1 μg/ml, 7-B6-1, MabTech, 3420-6-1000) in PBS+0.5% FBS for 1 hour. After washing with PBS+0.05% Tween20, the plates were incubated with Extravidin Alkaline Phosphatase (1 μg/ml; Sigma, E2636) in PBS+0.5% FBS for 1 hour. Subsequently, after washing with PBS+0.05% Tween20, the plates were washed with PBS and developed with TMB or 5-bromo-4-chloro-3-indolyl phosphate (BCIP)/nitro blue tetrazolium (NBT) (SigmaFast; Sigma, B5655). The reaction was stopped after approximately 7 minutes using H₂0 and the plates were dried for 15 minutes. Spots were analyzed with CTL reader and software, and the number of spots from DMSO controls was subtracted from total spot numbers induced by antigen-specific stimulation. Samples that showed a low PHA response (<33 Spot Forming Units (SFU)/5x10⁴ PBMC) compared to other samples from the same donor were excluded from the analysis due to technical issues. Samples showing no response were calculated as 0.15 SFU/2x10⁵ PBMCs (representing half of the minimum of 1 spot per triplicate). Finally, results are reported as IFNγ SFU per 1x10⁶ PBMCs.

An in-house developed multiplex immunoassay was used to quantify anti-CMV IgG antibodies (22). Thresholds to determine CMV-seropositivity were adapted from a previous study (23). Individuals were categorized as seronegative if IgG concentration was <4 relative units (RU)/ml; as borderline if ≥4 RU/ml and < 7.5 RU/ml; and as seropositive if ≥7.5 RU/ml. Donors classified as borderline were excluded from analysis when CMV status was taken into consideration (see Table 1 ).

Values below the detection limit of the ELISpot assay (0.75 SFU/10⁶ PBMCs) were recorded at 0.75 and missing data were excluded list-wise. Log₂ of fold change (log₂FC) was used to describe the changes in Spike-specific T cell ELISpot values between different timepoints. Data were visualized by means of boxplots, violin plots and linear trends with dots representing individual datapoints. In boxplots, median and interquartile range are provided.

For univariate hypothesis testing, figure legends specify the utilized statistical test. On the other hand, we employed linear models to establish multivariate associations between the longitudinal vaccine response, baseline characteristics (age, sex, CMV status), and the different vaccines received. Owing to the different vaccination regimes received at different timepoints, BNT162b2 or mRNA-1273 at P28 and B0; and BNT162b2/BNT162b2, BNT162b2/mRNA-1273, and mRNA-1273/BNT162b2 at B28, two separate models were required. To estimate the IFNγ T cell response to primary vaccination at timepoints P28 and B0, we established a mixed-effects regression model. As fixed effects, sex, an interaction term between age and CMV status, and an interaction term between timepoint and the type of primary vaccine were considered. As random effects, a random intercept per donor and a random slope per timepoint were defined ( Table 2 ). On the other hand, we established a fixed-effects model to estimate the IFNγ T cell response to the different primary-booster vaccination regimes at B28, replacing the primary vaccine factor by the primary-booster vaccination regime ( Table 3 ). Lastly, to evaluate the induction of the T cell response by the different primary-booster vaccine combinations and the effect of time in between the two vaccines, we established a fixed-effects model to estimate the log₂FC between the IFNγ T cell response at B28 and at P28. To tackle the multi-collinearity issue between age and time to booster vaccination, these two variables were centralized around their mean. The interaction in between the resulting variables, sex, and the log₂FC between the response at B0 and at P28 were defined as main effects ( Table 4 ).

A p-value of <0.05 was regarded as statistically significant and data analysis and visualization were performed in RStudio software (R version 4.3.0, R Core Team).

We investigated COVID-19 vaccine-induced T cell responses in different age groups (range 18-99 years) of generally healthy adults in the Dutch population without history of SARS-CoV-2 infection(s) after their primary vaccination series and the first booster vaccination. The vaccination and sampling schemes are shown in Figure 1A and the characteristics of the different vaccine groups are given in Table 1 . In total 125 participants received the BNT162b2 (Pfizer/BioNTech; two doses) vaccine for their primary series, 71 the mRNA-1273 (Moderna; two doses), 3 the ChAdOx1 nCoV-19 (AstraZeneca; two doses), and 9 the Ad26.COV2.S (Janssen; 1 dose) vaccine. Booster vaccination was followed up in a total of 81 BNT162b2- and 40 mRNA-1273-primed participants.

Age and sex composition was slightly different between the BNT162b2 and mRNA-1273 primary vaccinated groups, with the first composed of more younger female adults than the latter, yet not significantly ( Table 1 ). Regarding CMV status, CMV seropositivity was significantly different across age groups in this dataset (p-value 0.0076; Kruskal-Wallis), but randomly occurring across sex and primary vaccination groups. The distribution of CMV-seropositive donors across age groups in this cohort was approximately 45% up to 69 years old and 70% among ≥70 years old donors, which reproduces the previously described CMV incidence in the general Dutch population (18, 19)).

First, we evaluated the T cell responses induced after primary series with different COVID-19 vaccines (pre (P0) versus post (P28) vaccination). For this, the number of IFNγ-producing T cells was assessed by stimulating PBMCs with an overlapping peptide pool spanning the entire sequence of the Spike protein from the ancestral strain ( Figure 1B ). In the vast majority of the participants in all vaccine groups, a robust Spike-specific T cell response was induced ( Figure 1C ). Notably, higher T cell responses were induced by mRNA-1273 compared to BNT162b2 vaccination ( Figures 1C, D ).

When interrogating the effect of age in the vaccine-induced Spike-specific T cell response, we observed an inverse relation between the age of the vaccinees and the Spike-specific T cell frequency induced by primary vaccination in both BNT162b2 and mRNA1273 vaccine groups ( Figure 1D ). Stratifying participants according to CMV seropositivity and age groups (18-59, 60-69, and 70+ years), revealed that only in seropositive individuals ≥70 years old, this inverse relation was significant ( Figure 1E ). This is in line with the significant lower T cell frequencies in the 70+ versus the 18-59 age group for both BNT162b2- or mRNA1273-vaccinated CMV-seropositive individuals, although a trend in reduced levels in the 70+ age group is also found for CMV-seronegative participants in the mRNA1273-vaccinated group ( Figure 1F ). Of note, the higher responses induced by primary mRNA-1273 versus BNT162b2 vaccination was only observed in the CMV-seropositive 18-59 age group ( Figure 1E ). No significant effect of the sex of the vaccinees on the T cell response was found, even across age groups ( Supplementary Figures 1A, B ). Likewise, the inverse relation of Spike-specific T cell responses and age for CMV-seropositive donors was significant for both sexes, yet with a similar trend for CMV-seronegative males ( Supplementary Figure 1C ).

Together, these results indicate that all primary COVID-19 vaccine regimens induced a significant Spike-specific T cell response, with the mRNA-1273 vaccine inducing higher responses than the BNT162b2 vaccine mainly in the youngest adult age group (18-59). Age seems to negatively impact the height of the induced T cell response upon primary COVID-19 vaccination, which is most pronounced among CMV-seropositive older adults (70+).

Next, we investigated possible differences in the decline of the T cell response after primary vaccination across age groups and CMV status. For this purpose, we compared T cell responses at 28 days post primary series (P28) to pre-booster vaccination (B0). As expected, T cell frequencies were lower at the B0 compared to the P28 timepoint ( Figure 2A ), yet remained significantly higher than pre-vaccination (P0) levels in all age groups (data not shown). A similar tendency was seen when further stratifying the data by CMV status and primary-booster vaccination regimes ( Supplementary Figure 2A ). The interval between completion of the primary series and booster vaccination ranged between 4-9 months, which could affect the detected Spike-specific T cell frequencies at B0. We observed a weak inverse relation between the B0 T cell levels and the length of the interval between primary and booster vaccination for both BNT162b2 and mRNA-1273 vaccinees, albeit not significant for the latter ( Figure 2B ). Additionally, waning of the T cell response between BNT162b2- versus mRNA-1273-vaccinated participants and across the different age groups was similar, as the log₂ of fold change (Log₂FC) of the B0/P28 response was comparable. However, taking CMV status into account, the 18-59 age group showed more waning (i.e. lower Log₂FC) in the mRNA-1273 versus BNT162b2 vaccine group only among CMV-seropositive participants ( Figure 2C ). Moreover, no correlation was apparent between the interval between primary and booster vaccination and the decline in the Spike-specific T cell response in both CMV-seropositive and -seronegative vaccinees ( Figure 2D ).

Together, this indicates that in all age groups the induced Spike-specific T cell response waned between 4 and 9 months after the primary series, but T cells were still detectable in the peripheral blood. This suggests maintenance of a Spike-specific memory T cell population that could be recalled when encountering the Spike antigen.

Subsequently, we investigated to what extent a booster vaccine dose, given 4-9 months after completion of the primary series, increased the Spike-specific T cell frequencies. This revealed that for the 18-59 and 60-69 age groups, the Spike-specific T cell response was significantly elevated at 28 days upon booster vaccination (B28 versus B0), with a similar trend for the 70+ age group ( Figures 3A, B ). Additionally, the ultimate number of boosted IFNγ-producing T cells of the 70+ age group were significantly lower than those in the younger adult age groups ( Figure 3A ).

Differences were observed when interrogating the different primary-booster vaccination regimens. Both groups with BNT162b2 as a primary vaccine (i.e. BNT162b2/BNT162b2 and BNT162b2/mRNA-1273) showed a significant increase in the Spike-specific T cell response upon booster vaccination (B28 versus B0) ( Figure 3C first and second panels), reaching similar levels as shown after the primary series at P28 ( Figure 3D ). In contrast, the mRNA-1273/BNT162b2 primary-booster combination resulted in a limited boosting effect as reflected by a negative log₂FC, and strikingly, T cell frequencies even tended to be lower after booster vaccination (B28) than after the primary vaccination series (P28) (P28 versus B28, p-value 0.051; Figures 3C, D ). This effect was not directly correlated with age ( Figure 3E ) nor with the time interval between primary and booster vaccination ( Figure 3F ). Notably, the ultimate boosted T cell frequencies at B28 were comparable between all primary-booster vaccine combination groups ( Figure 3C ). Moreover, CMV status did not affect these observations across any of the age groups ( Supplementary Figures 2A-E ).

Together, we gathered evidence that recovering the post-primary Spike-specific T cell response upon booster vaccination is affected by the given primary-booster vaccination regimen, with the mRNA-1273/BNT162b2 mounting limiting boosting effects.

We additionally applied a multivariate analysis to investigate the effect of age, sex and CMV status or their interactions on the vaccination induced Spike-specific T cell response. Importantly, these models confirmed the significant decrease of the Spike-specific T cell response with aging among CMV-seropositive donors ( Supplementary Figure 2F , filled lines; Tables 2 , 3 ); the higher primary T cell response induced by mRNA-1273 compared to BNT162b2 ( Supplementary Figure 2F , P28; Table 2 ); and the limited boosting capacity of the heterologous mRNA-1273/BNT162b2 vaccine combination independently of the time interval between vaccines ( Supplementary Figure 2G , Table 4 ).

Conversely, these models also challenged previous observations made from non-parametric univariate analysis. For example, upon adjusting for age, sex, CMV status and the type of primary vaccine received, the estimated T cell response does not support the significant decrease with age independently of CMV status ( Table 2 ). In addition, while no significant differences were seen when comparing the IFNγ T cell response between sexes, the results of the multivariate model depicted in Table 3 estimated higher vaccine-induced T cell responses for females compared to males at B28. Likewise, despite no differences were found when comparing the T cell response at B28 in the two different BNT162b2-primed vaccine groups, the model estimates a higher response for donors receiving the BNT162b2/mRNA-1273 versus the BNT162b2/BNT162b2 primary-booster vaccination regimen. Owing to the poor fitness of both models and potential confounding, evidenced by the low R² values depicted in Tables 2 – 4 , it should be noted that there might be other more informative variables not considered in the present study that could improve the longitudinal estimation of the T cell response to COVID-19 vaccination.

During the COVID-19 pandemic, multiple VOCs emerged. Several studies indicate that vaccination-induced T cell responses are highly cross-reactive to the Spike proteins of SARS-CoV-2 variants (3, 5–8). To assess cross-reactivity, we performed a paired analysis just for those participants who showed an IFNγ T cell response against the Spike protein from the ancestral strain above the detection limit of the assay (>0.75 SFU/10⁶ PBMCs). Across the whole cohort, we found a slight reduction in cross-reactivity to the Spike protein of Alpha (B.1.1.7), Delta (B.1.617), and Omicron BA.1 (B.1.1.529 BA.1) versus the ancestral variant after primary vaccination. This effect was similar upon booster vaccination, except for the Delta variant ( Figure 4A ). Taking age into account, the reduction in reactivity to Alpha, Delta, as well as Beta was seen in the 18-59 age group after primary vaccination (P28), while for Omicron BA.1, it was observed in both the 18-59 and 60-69 age groups ( Figure 4B ). After booster vaccination (B28), the reduction in the T cell response was only observed for Omicron BA.1 (B.1.1.529 BA.1) in the 60-69 age group. No significant reduction to all tested VOCs was found for the 70+ age group ( Figure 4B ), which might be a consequence of the higher variation in the T cell response observed in this group. Additionally, the observed reduction in cross-reactivity was diminished upon booster vaccination, indicating the capacity of a third vaccination to boost vaccine-induced cross-reactive T cells against VOCs. CMV seropositivity of the vaccinees did not affect cross-reactivity of vaccine-induced T cells within any of the age groups ( Supplementary Figure 3 ), and similar patterns were observed across different primary-booster vaccine combinations (data not shown).

Overall, this indicates that there is only a small reduction in the cross-reactivity of ancestral vaccine-induced Spike-specific T cells against Alpha, Beta, Delta, and Omicron BA.1 variants. Booster vaccination seems to be required to generate a similar Spike-specific T cell response towards the vaccine ancestral strain and the emerging VOCs.