BNT162b2 vaccination induces durable SARS-CoV-2–specific T cells with a stem cell memory phenotype

Description SARS-CoV-2 mRNA vaccination induces stem cell memory T cells specific for the virus that persist up to 6 months. SARS-CoV-2 mRNA vaccination induces long-term T cell responses Vaccinations for SARS-CoV-2 are being administered globally to halt the spread of the infection. Despite the prevalence of these vaccines, the long-term immune responses they induce are still poorly characterized. Here, Guerrera et al. took longitudinal blood samples from 71 BNT162b2 mRNA–vaccinated individuals and looked at the T cell responses to vaccination. They found that the vaccine induced polyfunctional CD4+ and CD8+ T cell memory responses that peaked 2 weeks after the second dose of the vaccine and contracted at 6 months after vaccination. Despite this waning, the mRNA vaccine induced T stem cell memory cells after the first dose that remained stable out to 6 months. Thus, BNT162b2 provides robust, long-term T cell memory responses against SARS-CoV-2. Vaccination against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is effective in preventing hospitalization from severe COVID-19. However, multiple reports of breakthrough infections and of waning antibody titers have raised concerns on the durability of the vaccine, and current vaccination strategies now propose administration of a third dose. Here, we monitored T cell responses to the Spike protein of SARS-CoV-2 in 71 healthy donors vaccinated with two doses of the Pfizer-BioNTech mRNA vaccine (BNT162b2) for up to 6 months after vaccination. We found that vaccination induced the development of a sustained anti-viral CD4+ and CD8+ T cell response. These cells appeared before the development of high antibody titers, displayed markers of immunological maturity and stem cell memory, survived the physiological contraction of the immune response, and persisted for at least 6 months. Collectively, these data show that vaccination with BNT162b2 elicits an immunologically competent and long-lived SARS-CoV-2–specific T cell population.

Mass vaccination against COVID-19 is effective and protects against severe COVID-19 in real-world settings (34,35). Decreases of antibody levels and increases of breakthrough infections point to waning immunity of vaccination over time (36)(37)(38)(39). However, despite the decline in antibody levels, protection from severe disease and hospitalization remains high (40,41), suggesting that persisting cellular immunity drives the immune response and prevents viral dissemination when antibodies disappear. Studies on immune memory to other coronaviruses have shown that cellular immunity can be detected for up to 17 years after initial infection in the absence of antibodies (9,(40)(41)(42), with cellular immunity persisting for up to 15 months after SARS-CoV-2 infection (43). The study of the cellular immune components induced by vaccination and the assessment of their persistence is fundamental; in this respect, a recent study has shown that robust adaptive immune responses can be detected for up to 6 months after mRNA vaccination (15).
Here, we performed a longitudinal study looking at the T cell responses and anti-receptor binding domain (RBD) antibodies in 71 health care workers and scientists vaccinated with the BNT162b2 vaccine following the European Medicines Agency (EMA)-approved two-dose vaccination schedule, up to 6 months after the first dose. We investigated expression of phenotypic markers associated with the differentiation, polarization, and cytotoxic functions of Spikespecific T cells and monitored the evolution of the adaptive immune response induced by mRNA vaccination. Vaccination induced the generation of immunologically competent Spike-specific T cells, including potentially long-lived memory stem cells, pilasters of durable immunity. This information provides further information on the durability of the vaccine, which may inform vaccination strategies going forward.

Induction and persistence of anti-RBD antibodies
The antibody response to vaccination with BNT162b2 was measured in serum samples obtained at baseline (T0), on the day of the boost (T1), 14 days later (T2), and 6 months after the first dose (T3). All individuals in our cohort were devoid of anti-RBD antibodies at baseline, and significant levels appeared in 100% of individuals only after the second dose (fig. S1A) (T1: median 28; T2: median 1786; T3: median 517). Age and sex are key variables in immunity induced by vaccination, whose effectiveness decreases with age and is usually lower in males (44). In our cohort, antibody levels induced by BNT162b2 were comparable between males and females (fig. S1B), similar to other studies (45). Also, we found that antibody titers negatively correlated with age at all time points, confirming previous results (fig. S1C) (45)(46)(47). Thus, the data show that BNT162b2 induces the production of anti-RBD antibodies, which decrease over time but are maintained at high levels for at least 6 months.

Induction and durability of the Spike-specific T cell response
To investigate the cellular immune response induced by vaccination, we exposed freshly obtained peripheral blood mononuclear cells (PBMCs) to peptide pools spanning the entire sequence of the Spike (S) protein. Blood collection was performed at the same time points as outlined above. To fully capture the antigen-specific T cell response and to maximize sensitivity, two separate assays for the detection of the expression of surface activation-induced markers (AIM) and for intracellular cytokine staining (ICS) were set up (see fig. S2 for gating strategy). All assays were performed on freshly isolated PBMCs to obtain accurate absolute cell counts.
AIM + CD4 + T cells were defined by upregulation of CD40L and CD69 (48), whereas CD137 and CD69 expression identified the AIM + CD8 + subset (5) (Fig. 1A). After paired background subtraction from parallel unstimulated cultures, and having set a threshold for positivity at 0.008% for CD4 + T cells and 0.079% for CD8 T cells (see Materials and Methods for further details), we found that 97% of donors had detectable numbers of AIM + CD4 + cells at baseline (T0; Fig. 1, B and C, and fig. S3A). Twenty-one days after the first dose of vaccine, these cells were expanded sixfold compared with baseline (T1) and further still at 14 days after the boost (T2). Six months after the first dose (T3), AIM + CD4 + cell numbers were still fivefold higher compared with baseline and were detected in all donors. AIM + CD8 + T lymphocytes were present at baseline in only 18% of the donors (Fig. 1, B and C, and fig. S3A). After vaccination, 87% showed Spike-specific CD8 + T cells, which had expanded 11-fold from baseline; after further expansion after the second dose, these cells remained detectable after 6 months in 88% of donors ( Fig. 1, B and C, and fig. S3A). The total magnitude of the T cell response (including both CD4 + and CD8 + cells) increased significantly after priming (T1) and rechallenge (T2) but decreased 6 months after the first dose (T3) (Fig. 1D and fig. S3B). The stimulation index (S.I.; the ratio of AIM + T cells in stimulated over unstimulated samples) was used to determine T cell activation (49), and the cutoff for positive responses was arbitrarily set at 3. The S.I. of AIM + CD4 + cells increased markedly after the first dose, remained at high levels after boosting, and decreased after 6 months, remaining more than fivefold higher compared with baseline (Fig. 1E). The S.I. of AIM + CD8 + T cells peaked 14 days after the boost, and declined by one-third at the latest time point (Fig. 1E). More than 70% of individuals in our cohort showed AIM + CD4 + cells with S.I. > 3 already at baseline, and this proportion increased to 100% 6 months after receiving the first dose (Fig. 1F). On the other hand, only 7% of individuals showed AIM + CD8 + with S.I. > 3 cells at baseline, with this fraction increasing to 60 and 75% after priming and boosting, and then decreasing to 63% at the latest time points (Fig. 1F). The frequency of S-specific CD4 + (and not CD8 + ) T cells induced by vaccination inversely correlated with age only at T1, although a tendency toward reduced numbers of activated T cells with increasing age was observed at all time points ( fig. S3C), again with no significant differences between males and females ( fig. S3D).
In univariate regression, the number of AIM + cells correlated with antibody levels only after priming and not at subsequent time points ( fig. S3E), suggesting that humoral and cellular immune responses follow different kinetics. Multivariate regression modeling indicated that the number of AIM + CD4 + cells 21 days after priming was the best predictor of anti-RBD levels after 6 months, as expected from a T cell-dependent B cell response (Fig. 1G).
Two donors in the cohort presented exceptionally high frequencies of CD4 + and CD8 + AIM + cells at baseline ( fig. S4A). Vaccination boosted the frequency of AIM + cells, but antibody responses remained within the 25th and 75th percentile of the distribution of the whole cohort at all time points ( fig. S4B). Thus, vaccination induces detectable and robust antigen-specific T cells that develop before high antibody titers, with most T cell expansion occurring after the first dose and persisting for up to 6 months.

Cytokine production by S-specific T cells
Cytokine production was assessed by ICS after stimulation overnight with the peptide pools in the presence of monensin and brefeldin ( Fig. 2A). Measurement of cytokine production showed that, at baseline, 45 and 46% of the tested donors had IFN- + CD4 + and CD8 + cells, respectively (Fig. 2, B and C). Three weeks after the first dose, 83% of individuals had IFN- + CD4 + T cells, whereas CD8 + IFN- + were found in 71% of individuals. Two weeks after the boost, 81 and 68% of individuals had IFN- + CD4 + and CD8 + T cells, respectively; these cells were maintained for 6 months. Fifty-three percent of donors at baseline showed IL-2 + CD4 + T cells; this fraction increased to 84% after the first dose then reached 94% after the boost, and remained detectable after 6 months. Polyfunctional IFN- + IL-2 + CD4 + T cells were induced in 99% of the individuals only after the booster dose and persisted for at least 6 months after the first dose (Fig. 2D). CD8 + T cells coexpressing IFN- + and lysosomal associated membrane glycoprotein (CD107a) increased significantly after the boost (Fig. 2D). Induction of cytokine-secreting S-specific T cells was equivalent in both sexes at all time points (fig. S5A). The number of tumor necrosis factor- (TNF)-secreting CD4 + and CD8 + cells was highest at the peak of the secondary response, 14 days after the second dose (fig. S5B); at the same time point, polyfunctional TNF + IFN- + cells decreased significantly, and, after 6 months, they were at prevaccination levels. Further analysis of CD107a, TNF, and granzyme B coexpression among CD4 + and CD8 + T cells producing IFN- + revealed changing patterns between time points (P < 0.001 for all comparisons except for IFN- + CD4 + T cells between T1 and T2; Fig. 2E), with the fraction of polyfunctional CD4 + cells increasing at each time point, and a predominance of CD8 + cells with two functions at the latest time point, indicating dynamic changes of functional programs in antigen-specific T cells after boost and contraction.
At the peak of the response (T2), cytokine secretion was also measured with a MACSPlex assay in the supernatants of cultures stimulated with the peptide pools for the AIM assays. Production of high levels of interferon- (IFN-) and interleukin-1 (IL-2) was found, whereas IL-17 and IL-4 were barely detectable (<5 pg/ml), confirming the T helper 1 (T H 1) differentiation profile of S-specific cells ( fig. S5C). Thus, vaccination induces the emergence of a robust CD4 + and CD8 + cytokine response by T cells after priming, whereas (G) A linear regression model was fitted to test for T1 significant predictors of T3 anti-RBD antibody serum levels. Variables were log 10 -scaled when their distribution was not normal. Model's R 2 was 0.18. **P < 0.01. The number of AIM + CD4 + cells per milliliter was deemed as significant **P < 0.01. Continuous and dotted lines represent linear regression and 95% confidence intervals, respectively. AU, Arbitrary Units. full effector functions marked by polyfunctionality are acquired after the boost, and then maintained for at least 6 months.

Differentiation features of Spike-specific T cells
We then characterized the antigen-specific T cells induced by vaccination through the definition of their differentiation status using CCR7 and CD45RA markers to define naïve (N), central memory (CM), effector memory (EM), and terminally differentiated effector (EMRA) populations (Fig. 3, A and B, and fig. S2). The study of the frequencies of these subsets within AIM + cells in the peripheral blood showed that, at all time points, EM cells dominate the CD4 + T cell subset, whereas CD8 + T cells also present a large proportion of terminally differentiated cells (Fig. 3, A and B). Six months after the first dose, CD4 + S-specific cells that had survived the physiological contraction of the immune response were mostly CM and EM cells, whereas, in the CD8 + subset, these cells also included a significant fraction of terminally differentiated effectors (Fig. 3, A and B).
Among the desirable outcomes of vaccination lies the generation of a pool of memory stem cells, which can rapidly and efficiently differentiate in highly effective and polyfunctional lymphocytes in case of re-encounter with their nominal antigen (50). Stemness includes longterm persistence, a key aspect in this age of pandemics and uncertainties on the durability of the novel vaccines. Thus, we searched for these cells within the antigen-responsive CD4 + and CD8 + subsets. After the first dose, CD4 + AIM + CD45RA + CD27 + CCR7 high CD95 + cells, representing CD4 + T SCM (T stem cell memory) cells (51), were detectable in 88% of individuals (see fig. S6 for gating strategy); 2 weeks after the second dose, this fraction was still 88%; after 6 months, 91% of individuals showed these cells (Fig. 3C). CD8 + T SCM followed similar kinetics, with 94, 87, and 96% of individuals showing these cells at T1, T2, and T3, respectively.
To investigate the possible impact of T SCM on future immune responses, we used a general linear model with stepwise selection aiming to optimize Akaike information criterion with leave-one-out A partial permutation test (10,000 iterations, Monte Carlo simulation) was applied on distributions into combinatorial gates. ***P < 0.001; ****P < 0.0001.

of 12
cross-validation. We tested if the number of early (i.e., T1) CD4 + T SCM induced by vaccination significantly predicted immunological outcomes at future time points. We found that the number of T SCM induced after priming was a significant predictor of the number of both CD4 + -and CD8 + -activated T cells at the latest time point (Fig. 3D). To exclude the possibility that outliers were present in the CD4 + T SCM distribution, we performed multiple Grubbs and Rosner tests and no outliers were found (tables S1 and S2). These findings show that vaccination with BNT162b2 induces a T cell population with features of longevity, which remain numerically stable in the peripheral blood for at least 6 months and predict future T cell responses.

Effector features of Spike-specific CD4 + and CD8 + cells
We next explored the phenotypic changes occurring in AIM + T cells after boosting and 6 months thereafter by unbiased computational analysis. We used FlowSOM clustering to identify four CD4 + T cell clusters segregated by the expression pattern of CD137, CD39, Inducible costimulator (ICOS), programmed cell death protein 1 (PD-1), Human The relevance of T1 CD4 + T SCM cells per milliliter in predicting T3 AIM + CD4 + (top) and CD8 + (bottom) cells per milliliter was tested with two generalized linear models with stepwise selection aiming to optimize Akaike information criterion. Model's R 2 was 0.16 and 0.34 for AIM + CD4 + and CD8 + cells, respectively. The number of CD4 + T SCM in T1 cells per milliliter was deemed as significant in predicting both CD4 + and CD8 + AIM + cell numbers in T3 (*P < 0.05). Continuous and dotted lines represent linear regression and 95% confidence intervals, respectively. leukocyte antigen -DR (HLA-DR), CD25, CXCR5, CD95, CCR7, CD45RA, CD38, and CD127 (table S3) and superimposed those clusters in a Uniform Manifold Approximation and Projection (UMAP) plot generated by embedding the same set of markers (Fig. 4A). AIM + CD4 + T cells were distributed differently in the UMAP plot at the three different time points (Fig. 4B), highlighting an overall phenotypic shift. Among these cell clusters, clusters 3 and 4 were unchanged in frequency, whereas clusters 1 and 2 showed significant changes over time (Fig. 4C). Cluster 2 increased in frequency at each time point and was characterized by high expression of activation markers (CD25, CD38, CD39, HLA-DR, and CD137) and of the typical markers of T follicular helper (T fh ) cells (i.e., expression of ICOS, PD-1, and CXCR5; Fig. 4, A and C). Cluster 1 was composed of cells that showed the lowest expression of CD25, CD38, CD39, HLA-DR, and CD1379, corresponding to a nonactivated profile, and decreased with time (Fig. 4, A and C).
Data analysis through manual gating and measurement of singlemarker expression on the antigen-specific T cells revealed the emergence and persistence in time, albeit at lower levels after 6 months, of CD4 + subsets displaying PD-1, ICOS, and CXCR5 ( fig. S7), typical T fh markers. At the 6-month time point, expression of activation markers such as CD25, CD39, CD38, and HLA-DR was greatly reduced. Linear regression modeling indicated that the fraction of PD-1 + ICOS + within CD4 + AIM + T cells was a significant predictor of antibody levels at the 6-month time point (fig. S8).
S-specific CD8 + lymphocytes were remodeled by vaccine doses, as shown by the distinct distribution with time along the UMAP axes (Fig. 4E). Two clusters, comprising highly (cluster 4) and slightly less (cluster 3) activated AIM + CD8 + T cells, showed transient expansion after the booster dose (Fig. 4, D and F). This was in contrast to cluster 1 (naïve-like cells), which remained stable after a decrease in frequency at T2, and to cluster 2 (CCR7 + CD127 + CD45RA − CM CD8 + T cells), which was not significantly reduced at T2 but increased in size at 6 months. Manual analysis of AIM + CD8 + T cells confirmed the transient increase at T2 of CD38 + , HLA-DR + , and CD25 + cell frequencies, whereas CD39 + and PD-1 + cells steadily decreased or increased, respectively, at T2 and T3 compared with T1 ( fig. S7). Thus, antigen-specific T cells acquire phenotypic features of activation and functional capability after the booster, and most of these attributes are less evident and partially replaced, in the long run, by characteristics distinctive of more quiescent memory cells.

DISCUSSION
Multiple different vaccines are being administered globally to prevent the infection and spread of SARS-CoV-2. Here, we studied the adaptive immune response induced by vaccination with BNT162b2 to characterize Spike-specific T cell responses and to understand whether vaccine-induced T cells displayed features of longevity. We found Spike-specific CD4 + and CD8 + memory T cells that peaked 2 weeks after the boost and remained detectable in the peripheral blood for up to 6 months. These findings are in line with those of Goel et al. (15), who have recently described the establishment and durability of memory T cells after mRNA vaccination, and with previous studies investigating cellular immunity after SARS-CoV2 infection (2,26,28). Also, we found antigen-specific population of T SCM , which persisted overtime.
In our cohort, nearly all individuals harbored Spike-specific T cells at baseline, likely due to the presence of a pool of memory clones cross-reactive with other coronaviruses. Preexisting T cell crossreactivity to endemic coronaviruses, such as common cold coronaviruses, has been described in various proportions of SARS-CoV-2-naïve individuals (5,9,23,24,49,(52)(53)(54). Bacher et al. (54), using freshly isolated cells, detected Spike-specific T cells in all tested donors, similar to our findings. The presence of SARS-CoV-2 cross-reactive T cells associates with an enhanced cellular and humoral response to vaccination (49). The individuals in our cohort with exceptionally high numbers of AIM + CD4 + cells at baseline did not respond to vaccination with an equally exceptional antibody response. This could be due to the fact that cross-reactive T cells mainly target the S-II region of the Spike protein (49), which does not contain the RBD-binding region; thus, measurement of only anti-RBD antibodies may not inform on the whole spectrum of the humoral response induced by vaccination.
An optimal antibody response is the consequence of a competent underlying T cell response, which relies on the presence of T fh cells that interact with B lymphocytes in germinal centers to promote high-affinity antibody production and the generation of longlived memory B cells (55). High antibody titers induced by influenza vaccination positively correlate with the frequency of T cells expressing follicular helper molecules, including CXCR5 and ICOS (56,57). Induction of T fh cells also occurs in response to the Spike protein (58,59), and our finding of Spike-specific cells expressing CXCR5, PD-1, and ICOS suggests that mRNA vaccination is effective in generating CD4 + cells able to interact with B cells, in line with other studies (13,15,17,60). Accordingly, the simultaneous measurement of serum levels of anti-RBD antibodies showed the effective induction of the humoral arm of the adaptive response in all individuals in our cohort. Although antibody titers did decline with time, they were maintained at high levels for the entire period of our observation. Our data also converge with a recent study showing that the magnitude of antigen-specific CD4 responses at early time points predicted antibody levels after 6 months, confirming the central role of CD4 + T cells in instructing the humoral response (15).
CD4 + T cells also sustain immune responses through the production of cytokines, and we find that mRNA vaccination elicits a vigorous T H 1-skewed response, with production of IFN-, IL-2, and TNF by Spike-specific cells, and undetectable levels of IL-4 and IL-17. After 6 months, T cell responses were maintained, with the persistence of a numerically consistent pool of polyfunctional memory antigenspecific T cells (61). A correlation between both magnitude and polyfunctionality of T cell responses and resistance to infection or favorable disease evolution has been described in successful vaccine settings (62)(63)(64)(65)(66)(67)(68), and polyfunctional T cells correlate to protection against influenza (69). Polyfunctional cells were also detected in the CD8 + Spike-specific T cell subset, which produced IFN- and TNF and expressed granzyme B and the surrogate marker of degranulation CD107a, in agreement with previous studies (12,15,17,70,71). CD8 + T cells, however, were less expanded compared with the CD4 + subset and showed lower S.I., probably due to the fact that the peptide pool used in this study, consisting of 15-mers, is not optimal for presentation through major histocompatibility complex I and thus for CD8 + T cell recognition (72), so these responses might have been underestimated.
We found that mRNA vaccination also induced CD4 + and CD8 + T SCM , which were stable throughout the 6-month period of this study. T SCM cells provide a memory reservoir with multipotent capacity and have been shown to persist for decades (73-76). The assessed by Friedman test followed by post hoc two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli. *P < 0.05; **P < 0.01; ***P < 0.001. (G) T cell marker measurements from both CD4 + and CD8 + AIM + cells were normalized across the three time points, and the relative positivity for each marker is displayed on the radar plots. Measurements from CD4 + T cells are shown in red, those from CD8 + cells are blue, and cumulative measurements from both subsets are purple. The resulting plots illustrate the main features of T cells responding to a pool of peptides derived from the Spike protein of SARS-CoV-2. Dashed circles indicate neutralizing anti-Spike antibody levels. Histograms represent CD4 + (red) and CD8 + (blue) cell counts along the timeline. Syringes indicate the time point of vaccine administration, and tubes correspond to the day of blood sampling. Cyto + : aggregation of absolute numbers of CD4 + and CD8 + cell producing at least one cytokine among IFN-, IL-2, and TNF; IFN- MFI: mean fluorescence intensity of the IFN- signal in IFN- + CD8 + (blue) or CD4 + (red) cells.
detection of TSCM in vaccinated individuals is thus suggestive of the establishment of long-lived immunity. The number of T SCM induced after priming predicted future T cell responses, and the stability of this memory population may point in the direction of durable immunity against SARS-CoV2.
Last, immune responses to vaccines have been shown to be higher in females (44,77). Moreover, the male sex is associated with severe COVID-19 and death (78). Thus, we searched for a sex bias in the immunological parameters investigated. We did not find differences in vaccine-induced immune responses between males and females.
This study has the inevitable limits of human studies, consisting mainly of the wide genetic and immunological variability of the study population, and with blood being the only available source of material. Thus, we only studied circulating lymphocytes, which may be different from those who reside at mucosal barriers and which confer immediate protection against infection. Also, we did not investigate Spike-specific B lymphocytes, and dosage of antibody titers was the only readout of successful induction of humoral immunity. T fh and T SCM cells were identified only by the phenotypic characterization of surface markers, and functional studies, such as measurement of IL-21 production to identify T fh cells, were not carried out to match the phenotype. Further time points will be necessary to measure effective durability of anti-SARS-CoV-2 immune responses. Moreover, although donors were equally distributed for age and sex, our sample was limited in size.
On the whole, the results of this study can be visualized as the dynamic and integrated emergence of a Spike-specific adaptive immune response, characterized by a T cell response that precedes the development of high levels of anti-RBD antibodies (Fig. 4G). T cells induced after the first encounter with the Spike protein are mostly effector cells, which secrete intermediate levels of cytokines, express the highest levels of granzyme B, are not polyfunctional, and only some have T fh characteristics; cells with features of stemness and longevity also appear. After the boost, the peak of the response shows a fully activated, cytotoxically empowered, multifunctional T cell population, including cells of the T fh phenotype; at this time point, the highest levels of anti-RBD antibodies are detected in the serum. What remains after 6 months is a population of T cells with features of polyfunctionality and markers of T fh cells; at this time point, Spike-specific T cells that have survived the immunological contraction are highly specific and theoretically prone to give rise to effective and rapid antiviral responses, both by possibly sustaining the production of antibodies and by exerting cytotoxicity toward already infected cells; concomitantly, the pool of T SCM is stably maintained. This is in line with clinical real-world data showing that vaccine effectiveness in preventing severe COVID-19 is maintained above 90% at least for 6 months (79)(80)(81). Protection from infection, however, does significantly decrease with time and likely correlates to the weaning antibody titers (21,38), which provide an immediate shield against infection.
Cellular responses to mRNA vaccination thus include richly heterogeneous and dynamic memory T cell populations showing features associated with protective immunity. Previous studies performed at early time points after vaccination have described the induction of memory T H 1 and T fh CD4 + T cells along with cytotoxic CD8 + T cells (13,16,17,82,83). Observations at more distant time points (6 months) have shown stability of SARS-CoV-2 memory T cells, with a faster decline of the CD8 + subset (15). Another study found equivalent levels of CD4 + and CD8 + cells 6 months after vaccination, as observed after natural infection (13). Here, we have shown that a sizable population of T SCM is promptly induced by vaccination, but whether these memory progenitors will provide added protection upon viral exposure remains to be established.

Study design
This work started to define cellular immune responses against the SARS-CoV-2 Spike-protein induced by mRNA vaccine administration. To this aim, we enrolled 71 individuals from health care workers and scientists operating at the Santa Lucia Foundation scheduled for vaccination with Pfizer-BioNTech BNT162b2 between 12 January and 2 February 2021 (table S4). All donors signed informed consent forms approved by the Ethical Committee of the Santa Lucia Foundation. Venous blood was obtained immediately before the first dose (T0), 21 days thereafter, at the time of boosting (T1), two weeks after the second dose (T2), and 6 months after the first dose (T3). The use of freshly obtained blood cells permitted the detection of fragile markers, avoided the bias introduced by freezing/thawing procedures, and provided the possibility to precisely calculate absolute cell counts, a measure routinely used to guide clinical decisions in infectious diseases, such as HIV infection (84).

Evaluation of anti-SARS-CoV-2 antibodies
The measurement of anti-RBD antibodies was performed by electrochemiluminescence sandwich immunoassay through Roche Elecsys Anti-SARS-CoV-2 S (Roche Diagnostics, Switzerland). Antibody levels were measured on a Cobas 601 modular analyzer (Roche diagnostics, Switzerland), using a cutoff of 0.8 U/ml. Elecsys Anti-SARS-CoV-2 S U/ml measurements are equivalent to World Health Organization International Standard Binding Arbitrary Units per milliliter (BAU/ml).

AIM assay
In vitro stimulation of freshly obtained PBMCs was performed as previously described (85). Briefly, 200 l of cell suspension (10 × 10 6 cells/ml) was seeded in U-bottom 96-well plates and stimulated for 18 hours with PepTivator SARS-CoV-2 protein S, S1, and S+ peptide pools (1 mg/ml each; Miltenyi Biotec) or with an equal volume of water. Purified CD40 (0.5 g/l; Miltenyi Biotec) was added at culture start. Supernatants and cells from these culture wells were collected for cytokine measurement and flow cytometry, respectively. The threshold for positivity for background-subtracted values was set using the median 75th percentile of values obtained in negative control cultures (86).

Intracellular cytokine staining
PBMCs were incubated with peptide pools or water for 1 hour and then BV421-conjugated antiCD107a (1:200 dilution; BD Biosciences), monensin, and brefeldin A (5 M and 10 g/ml, respectively; both from Sigma-Aldrich) were added to the cultures. After an additional 17 hours, cells were harvested and directly stained for flow cytometry.

Flow cytometry staining and acquisition
Postculture cells were pelleted in V-bottom 96-well plates and resuspended in 30 l of antibodies at preoptimized concentrations and diluted in Brilliant Stain Buffer (BD Biosciences), then incubated for 15 min at room temperature (RT). Cell pellets were fixed either in FoxP3 fixation/permeabilization buffer (Thermo Fisher Scientific) for 20 min at RT (AIM assay) or in 4% formaldehyde in phosphate-buffered saline (ICS assay). ICS was performed by incubating cells in 30 l of antibodies against cytokines and granzyme B diluted in a saponin (Sigma-Aldrich) 0.5% w/v solution. The complete list of antibodies used for surface and intracellular staining is shown in tables S5 and S6, respectively. Samples were acquired on a fully equipped CytoFLEX LX within 6 hours from staining and fixation. Quality control beads (Beckman Coulter) were used daily to check and standardize instrument performances. Absolute counts were obtained by a volumetric, lyse-no-wash, flow cytometry approach. Briefly, whole blood was stained (table S7). After 15 min of incubation at RT, fluorescenceactivated cell sorting lysing solution (BD Biosciences) was added. After complete red blood cell lysis, samples run on a CytoFLEX with a volume-calibrated peristaltic pump. The number of cells per milliliter automatically derived by the instrument was then multiplied by 10,000 to determine the absolute CD4 and CD8 T cells on a milliliter basis. Absolute counts of AIM + , cytokine + , and related subsets were then performed multiplying their proportions by the absolute count of parent lineages. Data were analyzed with FlowJo v.10.7. UMAP embedding and FlowSOM clustering of AIM + cells were performed with FlowJo (UMAP ver3.1 and FlowSOM ver2.6 plugins). A total of 2500 CD4 and 990 CD8 AIM + T cells, equally sampled from 10 donors per three time points, were used to run each algorithm.
Cytokine measurements IL-4, IL-17, TNF, and IFN- were measured in frozen culture supernatants from the AIM assays described above by a bead-based sandwich enzyme-linked immunosorbent assay (MACSPlex Cytokine Kit, Miltenyi Biotec) according to the manufacturer's instructions.

Statistical analysis
Statistical analyses were performed with GraphPad Prism 9, Spice 6.1, and R programming language using the following libraries: caret 6.0, FactoMineR 2.4, and factoextra 1.0.7. Details of the statistical tests applied for each experiment are illustrated in the respective figure. Time points were compared by Kruskall-Wallis or Friedman test followed by post hoc Dunn's or two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli. Statistical significance was inferred if P < 0.05. In all figures, P values are indicated by *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Outlier detection was performed with multiple Grubbs' test and Rosner's test using the outliers and EnvStats packages in R, respectively.

SUPPLEMENTARY MATERIALS
www.science.org/doi/10.1126/sciimmunol.abl5344 Figs. S1 to S8 Tables S1 to S7 Raw Data View/request a protocol for this paper from Bio-protocol.