Signs of immunosenescence correlate with poor outcome of mRNA COVID-19 vaccination in older adults

Sixty-six individuals (median age 54; range 22–95) with no history of SARS-CoV-2 infection or related symptoms (hereafter ‘unexposed’) were recruited for this study and blood was drawn between 42 and 81 d (median 44 d) post-first vaccination. PCR-confirmed SARS-CoV-2 infected individuals (hereafter ‘exposed’) (n = 49; median age 54 years; range 22–99) were also included (Supplementary Tables 1 and 2)⁶. Although minimal correlates of protection to COVID-19 have not been established so far, vaccine-induced virus-neutralizing antibodies have been implied in protection against infection⁷. Therefore, we first measured SARS-CoV-2 neutralizing antibodies (VN) in our study participants. As previously reported, we found an age-dependent decrease of vaccine-induced neutralizing antibodies (r = −0.579; P < 0.0001) with significantly lower titers in older adults (≥66 years) compared to young (22–40 years; P < 0.0001) and middle-aged (41–65 years; P < 0.01) individuals (Fig. 1a,b). A similar correlation was found with the frequency of Spike-specific IgG memory B cells (MBCs) and total number of IgG MBCs, measured in a subset of individuals (22–40, 14 out of 23; 41–65, 21 out of 25; ≥66, 13 out of 18) (Fig. 1c,d). We could not detect Spike-specific IgA MBCs, perhaps because of their low frequency in peripheral blood below the detection limit of our assay. Overall, we observed reduced induction of SARS-CoV-2-specific antibodies and MBCs in older adults, as the possible consequence of age-related changes affecting B cells⁸. In the absence of protective antibodies, SARS-CoV-2-specific T cells may afford some protection against disease progression and severity, which may be important for older individuals who fail to develop VN antibodies^9,10,11. Nevertheless, like B cells, T cells also undergo age-related alterations. Thus, we investigated whether the magnitude and quality of the SARS-CoV-2-specific T cell responses were also affected by aging. To this end, we measured SARS-CoV-2-specific T cell responses by stimulating peripheral blood mononuclear cells (PBMCs) with pools of overlapping peptides spanning the SARS-CoV-2 S1 and S2 subunits of the Spike protein (homologous to the vaccine strain) using an ex vivo interferon-γ (IFN-γ) enzyme-linked immunospot (ELISpot) assay. In line with the B cell responses, the frequency of Spike-specific T cells declined with increasing age (r = −0.435, P = 0.0003) and Charlson Comorbidity Index (r = −0.417, P = 0.0005) (Fig. 1e and Extended Data Fig. 1a). No association was found between the magnitude of the Spike-specific response and time postvaccination (Extended Data Fig. 1b). Comparison of the response between age groups showed that the overall Spike-specific T cell response in individuals aged over 66 was significantly lower than in the young age group (P < 0.01) and, to a lesser extent (P < 0.05), between middle-aged and older individuals (41–65 versus ≥66 years) (Fig. 1f). Consequently, the proportion of nonresponders in the ≥66 vaccinees was higher than in the other two age groups (27.8 versus 8 versus 4.3%) (Fig. 1g). No significant difference was found between young and middle-aged individuals, indicating that impairment of the vaccine-induced immune response mainly affects older adults.

a, Correlation between age of study subjects and serum virus-neutralizing antibody titer (n = 66). b, Serum virus-neutralizing antibody titers in different age groups (light blue, 22–40 years, n = 23; red, 41–65 years, n = 25; gray, ≥66 years: n = 17). c,d, Frequency of Spike-specific IgG MBCs (c) and number of total IgG MBCs per million of in vitro-expanded PBMCs (d) (light blue, 22–40 years, n = 14; red, 41–65 years, n = 21; gray, ≥66 years, n = 13). e, Correlation between the age of the study participants and frequency of Spike-specific IFN-γ-secreting T cells (n = 66). f, Frequency of IFN-γ SFCs after stimulation with Spike in different age groups (light blue, 22–40 years, n = 23; red, 41–65 years, n = 25; gray, ≥66 years, n = 18). g, Percentage of individuals with Spike-specific response below the cutoff and considered nonresponders. h, Frequency of IFN-γ SFCs in response to nucleocapsid (65 out of 66), membrane (66 out of 66) and Spike (66 out of 66) proteins of SARS-CoV-2 in unexposed vaccinated individuals (n = 66). The numbers below the graph represent the percentage of responders and nonresponders for each tested antigen (orange, Spike; green, nucleocapsid; blue, membrane). Each dot represents a single study participant and the horizontal lines indicate the medians. The cutoff value for a positive response is defined as described in the Methods. The red line represents the linear regression; a two-tailed Spearman test was used to test the significance (r and P values). P values for age groups comparison were determined by two-tailed Kruskal–Wallis test with Dunn’s multiple comparison correction; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; NS, not significant.

Source data

Importantly, the reduced response in individuals over 66 is most likely not due to a general defect of T cell functionality since all individuals, regardless of their age, responded equally well to stimulation with influenza virus antigens and a CD3 antibody (positive control) (Extended Data Fig. 1c,d). Moreover, the Spike-specific response in SARS-CoV-2-exposed individuals was not affected by age (Extended Data Fig. 1e). These findings suggest that age impacts the magnitude of the mRNA vaccine-elicited T cell response but does not affect preexisting memory T cells, like those specific to the influenza virus.

Most (58 out of 66, 87.9%) unexposed vaccinees mounted a T cell response to the Spike peptide pools (S1 and S2), although with a considerable heterogeneity in magnitude (17–510 spot-forming cells (SFCs) per 10⁶ PBMCs). In contrast, a low frequency of nucleocapsid and membrane protein-specific T cells was seen in 21.5% (14 out of 65) and 10.6% (7 out of 66) of the vaccinees, respectively (Fig. 1h). Only one individual with no measurable response to Spike and nucleocapsid peptide pools had a high frequency of membrane-specific T cells (114 SFCs per 10⁶ PBMCs). SARS-CoV-2 nucleocapsid and membrane antigens (not contained in the vaccine) have high sequence homology with those of seasonal human coronaviruses (HCoVs) and cross-reactive T cells to these two antigens have been reported in several studies, although their role in protection against infection is still unresolved^12,13,14,15. Of note, their numbers inversely correlate with age¹⁶. As expected, most of the exposed individuals showed a response to all tested antigens: Spike 89.8% (44 out of 49); membrane, 61.2% (30 out of 49); and nucleocapsid 81.6% (40 out of 49) (Extended Data Fig. 1f).

It has been shown that the mRNA-based COVID-19 vaccine induces Spike-specific CD4⁺ and CD8⁺ T cells¹⁷. Because the ELISpot assay does not allow identification of T cell subsets, we utilized intracellular cytokine staining (ICS) and flow cytometry to further characterize the responding cells. PBMCs were stimulated with the Spike peptide pool and CD3⁺CD4⁺ and CD3⁺CD8⁺ non-naïve T cells were analyzed for the production of IFN-γ, interleukin-2 (IL-2) or tumor necrosis factor-α (TNF-α). The gating strategy is depicted and representative examples of ICS of CD4⁺ and CD8⁺ cells after stimulation with Spike peptides are shown in Extended Data Fig. 2a,b. IFN-γ-producing CD4⁺ and CD8⁺ T cells were detected on mRNA vaccination and, in agreement with the results obtained with the IFN-γ ELISpot assay, although with a higher magnitude when measured by ICS, the Spike-specific response was significantly lower in older adults, compared to young adults (CD4⁺ P < 0.0001; CD8⁺ P < 0.05) and middle-aged individuals (CD4⁺ P < 0.05; CD8⁺ P < 0.01) (Fig. 2a). Younger adults displayed a statistically significant higher frequency of Spike-specific CD4⁺IFN-γ⁺ cells than middle-aged adults (P < 0.05), which was not observed for CD8⁺ T cells. Similarly, age-dependent differences were found for IL-2⁺ (22–40 versus ≥66, P < 0.001; 41–65 versus ≥66, P < 0.05; 22–40 versus 41–65, not significant) and TNF-α⁺ CD4⁺ T cells (22–40 versus ≥66, P < 0.05; 41–65 versus ≥66, and 22–40 versus 41–65 not significant) (Fig. 2a). No differences were found between age groups in relation to CD8⁺ T cells producing IL-2 and TNF-α since IFN-γ dominates the Spike-specific response in vaccinees (Fig. 2a). Noteworthy, comparison of individuals aged 41–65 and ≥66 years exposed to SARS-CoV-2 showed no age-dependent differences in Spike-specific non-naïve CD4⁺ and CD8⁺ T cells producing IFN-γ, IL-2 or TNF-α (Extended Data Fig. 3a,b). Due to the small numbers, we could not include young adults in this comparison. PBMCs from vaccinated older individuals stimulated via CD3 engagement showed similar CD4⁺ T cell responses to those detected in 22–40- and 41–65-year-old individuals (Extended Data Fig. 3c). The frequency of IFN-γ-producing CD8⁺ T cells in response to CD3 increased with age (22–40 versus ≥66, P < 0.01; 41–65 versus ≥66, P < 0.01), but not those producing IL-2 (41–65 versus ≥66, P < 0.05) (Extended Data Fig. 3d).

a, The percentage of CD4⁺ (top) and CD8⁺ (bottom) T cells producing cytokines is shown for each age group (light blue, 22–40 years, n = 22; red, 41–65 years, n = 22; gray, ≥66 years, n = 16). b,c, Percentage of CD4⁺ T cells producing multiple cytokines (black, IFN-γ⁺/IL-2⁺/TNF-α⁺; dark gray, IFN-γ⁺/IL-2⁺; light gray, IFN-γ⁺/TNF-α⁺; white, IL-2⁺/TNF-α⁺), in ELISpot responder vaccinated (light blue, 22–40 years, n = 22; red, 41–65 years, n = 22; gray ≥66 years, n = 16) (b) and SARS-CoV-2-exposed individuals (light blue, 22–40 years, n = 14; red, 41–65 years, n = 19; gray, ≥66 years, n = 15) (c). The Spike-specific response was measured on non-naïve T cells. The gating strategy is depicted in Extended Data Fig. 2. Each dot represents a single donor and the horizontal lines indicate the medians. P values were determined by two-tailed Kruskal–Wallis test with Dunn’s multiple comparison correction; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Source data

In line with these findings, we also observed an age-dependent reduction of CD4⁺ T cells producing two (IL-2⁺IFN-γ⁺, TNF-α⁺IFN-γ⁺ and IL-2⁺TNF-α⁺) and three cytokines (IFN-γ⁺, IL-2⁺, TNF-α⁺) in response to Spike peptide stimulation, with older individuals showing a remarkable decrease of those polyfunctional T cells compared to young adults (67.4%) and the middle-aged group (50.8%) (Fig. 2b). Of interest, a reduction of CD4⁺ polyfunctional T cells could already be observed in middle-aged individuals compared to young adults (33.7%, 41–65 versus 22–40 years) (Fig. 2b). In contrast, SARS-CoV-2 infection induced polyfunctional CD4⁺ T cells in all age groups, although to a slightly lower extent in older adults (Fig. 2c). Furthermore, the frequency of polyfunctional CD4⁺ T cells was similar in vaccinated (0.086%) and SARS-CoV-2-infected (0.089%) younger adults indicating that infected and vaccinated individuals were otherwise comparable (Fig. 2b,c). We investigated the differentiation stage of the non-naïve Spike-specific CD4⁺ T cells in vaccinees and found an age-dependent redistribution of these cells within the central memory T (T_CM) and effector memory T (T_EM) cells (Extended Data Fig. 4). Younger and middle-aged individuals showed a higher percentage of T_EM over T_CM cells, whereas in the group of older adults the frequency of Spike-specific T_CM and T_EM cells was comparable. No differences were found for the effector memory CD45RA⁺ T (T_EMRA) cells.

In summary, these data demonstrated that in older adults, COVID-19 mRNA vaccination elicited a lower frequency of Spike-specific CD4⁺ and CD8⁺ T cells producing cytokines involved in T cell differentiation and proliferation. Furthermore, polyfunctional T cells are involved in protective immunity to virus infections and their low numbers in older adults may contribute to suboptimal protection provided by vaccination in this age group¹⁸.

Because T cell populations are reshaped during aging we sought to investigate whether the differentiation status of circulating T cells correlated with the magnitude of the immune response induced by mRNA vaccination^6,19. To this end, we defined four differentiation subsets of CD4⁺ and CD8⁺ T cells based on the surface expression of the CD45RA and CCR7 molecules (naïve, T_CM, T_EM and T_EMRA) and correlated the proportion of these cells with the frequency of Spike-specific T cells, measured by IFN-γ ELISpot. Overall, we found a decrease of naïve and an accumulation of terminally differentiated T cells, as described previously (Fig. 3a)⁶. This age-dependent reduction of naïve CD4⁺ and CD8⁺ T cells, which was more profound in the CD8⁺ compartment (CD4⁺ r = 0.549, P < 0.0001 upper; CD8⁺ r = 0.743, P < 0.0001 lower) (Fig. 3b), correlated with reduced numbers of Spike-specific IFN-γ SFCs (CD4⁺ r = 0.374, P = 0.0032 and CD8⁺ r = 0.454, P = 0.0002) (Fig. 3c). Comparable results were obtained when the IFN-γ response was measured on non-naïve T cells by ICS (Extended Data Fig. 5a,b). Comparison of the results obtained in the respective age groups showed an age-dependent association between CD4⁺ and CD8⁺ naïve T cells and the Spike-specific T cell response (Fig. 3d). Especially the loss of naïve CD8⁺ T cells in individuals aged ≥66 was inversely correlated with the vaccine-induced T cell response. Of note, no correlation was found with the CD8⁺ T_CM and T_EM subsets and CD4⁺ T_EM and T_EMRA, while a higher frequency CD4⁺ T_CM and CD8⁺ T_EMRA in older individuals correlated inversely with the vaccine-induced T cell response (Extended Data Fig. 5c,d). Remarkably, the proportion of naïve CD4⁺ and CD8⁺ T cells did not correlate with the frequency of T cells directed against the influenza virus, for which immunological memory exists and can be recalled in vitro^20,21 (Extended Data Fig. 5e,f). There was also no correlation with the Spike-specific responses measured in the exposed individuals (Extended Data Fig. 5g,h). Comparison of the three age groups confirmed that there is an age-dependent inverse correlation between naïve CD4⁺ and especially CD8⁺ T cells on the one hand, and the magnitude of the T cell response measured by IFN-γ ELISpot assay on the other. Thus, the lower magnitude of Spike-specific response on COVID-19 mRNA vaccination in older adults correlated with decreased numbers of naïve T cells. This suggests that this subset of T cells may play a role in the COVID-19 vaccination outcome because induction of a primary response to new antigens mainly relies on the activation of naïve T cells^22,23. Of note, a low frequency of naïve T cells has been associated with more severe COVID-19 disease and impaired priming of naïve CD8⁺ T cells in older adults^9,24. Moreover, restricted T cell receptor diversity and altered signaling in naïve T cells may generate a less effective pool of memory cells, which could lead to a suboptimal response, exposing older adults at higher risks of infection and disease severity^19,25,26.

a, CD4⁺ (upper) and CD8⁺ (lower) T cell differentiation subsets (naïve, black; T_CM, gray; T_EM, orange; T_EMRA, light orange) in vaccinated individuals (n = 60). b, Correlation between age and percentage of CD4⁺ (upper, r = −0.549; P < 0.0001) or CD8⁺ (lower, r = −0.743; P < 0.0001) naïve T cells. c, Spike-specific IFN-γ response association with percentage of CD4⁺ (r = 0.374; P = 0.0032) or CD8⁺ (r = 0.454, P = 0.003) naïve T cells. d, Spike-specific response in correlation with the percentage of naïve CD4⁺ (upper) and CD8⁺ (lower) T cells for each age group. Each dot represents a single donor. b–d, A two-tailed Spearman’s test was used to the test the significance (r and P values). b–c, The red line represents the linear regression. The gating strategy is depicted in Extended Data Fig. 2.

Source data

To unveil other potential age-dependent factors that may affect the response to vaccination in older people, we also characterized the magnitude and effector functions of terminally differentiated T cells coexpressing the CD57 and KLRG1 molecules, which have been associated with aging^23,27. Despite a general age-dependent increase of CD4⁺ (Fig. 4a, upper panel) and CD8⁺ (Fig. 4a, lower panel), T cells coexpressing these senescence markers, only the CD8⁺ cells were inversely correlated with the magnitude of the Spike-specific IFN-γ T cell response measured by ELISpot (r = −0.352, P = 0.0058; Fig. 4b) and fluorescence-activated cell sorting analyses (r = −0376, P = 0.003; Fig. 4c), although such correlation does not necessarily imply causality. Of note, some younger individuals showed a high proportion of CD4⁺ and CD8⁺ T cells coexpressing the senescence markers; interestingly, that correlated with the presence of serum antibodies to CMV (Extended Data Fig. 6a,b). This finding supports the notion that chronic infections, like those caused by cytomegalovirus (CMV), may be responsible for the accumulation of CD57⁺ KLRG1⁺ T cells, regardless of age. We found that the CD8⁺ T cells coexpressing such markers were of the T_EM and T_EMRA phenotype and were similarly distributed between age groups (Extended Data Fig. 6c,d). No correlation was found with the percentage of CD4⁺ CD57⁺ KLRG1⁺ T cells.

a–c, Frequency of CD57⁺KLRG1⁺ senescent CD4⁺ (upper) and CD8⁺ (lower) T cells for each age group (light blue, 22–40 years, n = 22; red, 41–65 years, n = 22; gray, ≥66 years, n = 16) (a) and their correlation with Spike-specific IFN-γ response measured by ELISpot (CD4⁺, not significant; CD8⁺, r = −0.352, P = 0.0058) (b) or flow cytometry (CD4⁺, not significant; CD8⁺, r = −0.376, P = 0.003) (c). d, Frequency of circulating T_FH cells in the different age groups (light blue, 22–40 years, n = 22; red, 41–65 years, n = 22; gray, ≥66 years, n = 16), identified based on the expression of CD4⁺ and the homing receptor CXCR5. e, Frequency of Spike-specific T_FH cells producing IFN-γ, IL-2 or TNF-α in vaccinated individuals (light blue, 22–40 years, n = 22; red, 41–65 years, n = 22; gray, ≥66 years. n = 16). f, Proportion of T cells directed to the S1 (filled circle) or S2 (open circle) region of the Spike protein in IFN-γ responder individuals (light blue, 22–40 years, n = 22; red, 41–65 years, n = 23; gray, ≥66 years, n = 13). Spike-specific response of circulating T_FH cells was analyzed on non-naïve CD4⁺ T cells. The gating strategy is depicted in Extended Data Fig. 2. Each dot represents a single study individual and the horizontal lines indicate the medians. The red line represents the linear regression and a two-tailed Spearman test was used to the test the significance (r and P values). P values were determined by a two-tailed Kruskal–Wallis test with Dunn’s multiple comparison correction; *P < 0.05; **P < 0.01; ***P < 0.001.

Source data

Next, since we observed reduced B cell responses (antibodies and memory cells) in older vaccinees, we addressed the question whether alterations of circulating follicular helper T (T_FH) cells related to reduced antibody responses²⁸. Interestingly, the circulating T_FH cells (CD3⁺CD4⁺CXCR5⁺) were numerically similar between the age groups (Fig. 4d). However, when we analyzed the Spike-specific non-naïve circulating T_FH cells, IFN-γ and IL-2 production was lower in the ≥66 group (IFN-γ, P < 0.05; IL-2, P < 0.01) compared to that of young adults (Fig. 4e). Of note, measurement of IFN-γ, IL-2 and TNF-α might have underestimated the frequency of the Spike-specific circulating T_FH cells as other markers (for example, ICOS and IL-21) may be expressed by a larger proportion of circulating T_FH.

No differences were found for the TNF-α-producing circulating T_FH cells (Fig. 4e). Of note, a high frequency of T_H1-like circulating T_FH correlated with strong antibody responses to influenza and other viruses^29,30,31.

We then investigated if, apart from the magnitude of the Spike-specific T cell response, age influenced the specificity of the T cell response. To this end, we tested the response to peptide pools spanning the S1 and S2 regions of the Spike protein, respectively. A similar frequency of S2-specific T cells was observed in all age groups (Extended Data Fig. 5i). However, the frequency of S1-specific T cells was significantly lower in older adults compared to young adults (P < 0.05). Moreover, when we calculated the relative contribution of the S1- and S2-specific response in the individuals who displayed a Spike-specific response, we noticed that older adults had a preferred response to the S2 region (P < 0.01) (Fig. 4f). This is in contrast with middle-aged and young adults, who had similar responses to S1 and S2 (Fig. 4f). In the SARS-CoV-2-exposed individuals, no statistically significant difference was observed (Extended Data Fig. 5j). S2 is more conserved than S1 across coronaviruses; therefore, it may represent a target of preexisting cross-reactive memory B and T cells originally induced by previous infections with ‘common cold’ HCoVs³².

Together with the reduced number of naïve T cells leading to suboptimal responses to a new antigen like the S1 region, this may explain the preferential recognition of the S2 region in older vaccinees³³.

Collectively, our data provide insights into the age-dependent immunological changes that may account for the reduced B and T cells responses observed in older adults on BNT162b2 vaccination; however, interestingly, this is not the case with the SARS-CoV-2 infection. The different antigenic load, anatomical site of antigen encounter (intramuscular versus mucosal), strength and length of T cell receptor engagement (replicating virus versus mRNA-encoded antigen), multiple viral antigens (SARS-CoV-2) versus mRNA-encoded Spike protein, may all account for the differential responses observed in this study. New vaccine approaches or use of substances able to overcome age-related defects would help to provide optimal protection to this vulnerable age group.

Our data suggest a general age-dependent decrease of the adaptive immune responses on BNT162b2 vaccination. However, our study has some limitations. The study was designed as a large cross-sectional epidemiological study and we have used samples from a subset of study individuals to investigate the SARS-CoV-2-specific immune response on vaccination and infection. Therefore, our cohort is heterogeneous and no correction for age and sex was performed since this would have required a larger sample size. Furthermore, the limited number of PBMCs available precluded the possibility to analyze the additional effector functions and biomarkers of the immune cells possibly associated with aging and immunosenescence. The correlation found between aging and redistribution of the T cell differentiation subsets and suboptimal T and B cell immune responses on vaccination do not imply a direct causality since some of the investigated variables (for example, CD8⁺ T cells with senescent phenotype) correlate with age.