Prior SARS-CoV-2 Infection Enhances Initial mRNA Vaccine Response with a Lower Impact on Long-Term Immunity

Owing to SARS-CoV-2 infection, as well as the administration of different platforms of COVID-19 vaccines, ∼50% of the world’s population has acquired immunity against SARS-CoV-2 (1–3). Following successive waves of distinct viral variants and the increase in breakthrough cases, new booster regimens have been introduced, all resulting in a complex virological and immunological landscape. Currently, the population is composed of heterogeneous groups, from rare immunologically naive individuals to those who have experienced multiple SARS-CoV-2 Ag exposures (through natural infection and/or vaccination). Both infection and vaccination can induce immunological protection from SARS-CoV-2, which has been successful in controlling the morbidity and mortality associated with COVID-19 (3, 4). However, questions about the trajectory of the adaptive immune responses and long-term immunity with respect to priming and repeated Ag exposure remain poorly understood.

It is well recognized that mRNA-based vaccination can induce robust humoral and cellular immunity and prevent severe disease caused by SARS-CoV-2 (5–13). High levels of anti-Spike Abs are induced by mRNA vaccination, which wane from vaccine peak levels and gradually plateau to a set point (4, 14–16). Despite this decline, subjects with undetectable or impaired humoral responses show better outcomes from severe disease and hospitalization, indicating a critical role of cellular immunity in disease protection (17, 18). mRNA vaccines elicit functional T cell immunity, with CD4⁺ and CD8⁺ T cell subsets contributing to protective immune responses and long-term immunological memory (5, 7, 9). Spike-specific memory T cells can be reactivated as soon as 24 h after Ag encounter, resulting in activation and expansion (10, 12, 19–21). Specifically, CD4⁺ T follicular helper (Tfh) cells have a key role in developing memory B cells, plasma cells, and Abs (10, 11, 19, 22). In addition, CD4⁺ Th1 cells support the quality of antiviral effector memory (EM) functions (1, 11, 23, 24), and central memory (CM) and EM T cell subsets help with leukocyte recirculation, tissue access, and overall magnitude and durability of immunity upon Ag re-exposure (11, 12, 19, 25).

Even with the success of mRNA in inducing a coordinated humoral and cellular immune response, less is known about the long-term effect of recurrent mRNA boosters. Estimating the magnitude and durability of mRNA vaccine responses in the population has become a challenge because of the many individuals with hybrid immunity (immunity developed through a combination of SARS-CoV-2 infection and vaccination) varying rates and timings of past infection and vaccination. Previous studies have examined the COVID-19 mRNA vaccine-induced response in individuals previously infected with SARS-CoV-2, and an enhanced response shortly after priming vaccination was observed (10, 12, 19, 26). However, longitudinal data examining the kinetics of multiple immune compartments simultaneously, from the initial response to vaccination to the late memory phase, remain poorly explored.

In this study, we characterized the quality, magnitude, and durability of immune memory over a three-dose monovalent Pfizer BNT162b2 mRNA vaccination in immunologically naive and previously infected individuals to better understand potential features of lasting vaccine effectiveness. We analyzed the Spike-specific responses in samples collected at six time points ranging from prevaccination up to 1 y after initial immunization. We had a unique set of individuals who received only the mRNA vaccination and were not infected before or during the vaccination schedule, allowing us to evaluate the long-term vaccine-generated response. A side-by-side comparison group was composed of individuals primed by natural infection before the vaccination but without breakthrough infections, allowing us to understand how pre-existing immunity induced by SARS-CoV-2 infection shapes the long-term memory response. Our data demonstrated that a third immunization expressing the same Spike protein improves the quantity and quality of long-term memory IgG but has a minimal improvement on T cell frequency and longevity. Additionally, the pre-existing immunity modulates the initial mRNA vaccination response by increasing Ab levels and inducing specialized CD4⁺ T cell subsets associated with cytotoxic functions and maturation of Ab-producing B cells. However, the little or absent benefit of priming infection was demonstrated in long-term immunity as both SARS-CoV-2–naive and preimmune individuals showed similar memory quality and quantity at the end of the three-dose vaccination regimen.

In 2020, the SPARTA (SARS SeroPrevalence and Respiratory Tract Assessment) program, funded by the U.S. National Institutes of Health, was initiated to understand the immune responses elicited following infection with SARS-CoV-2 and/or vaccination against COVID-19. Participants between 22 and 65 y of age were enrolled starting in October 2020 with written informed consent in Athens and Augusta, GA, Memphis, TN, or Los Angeles, CA (27, 28). The study procedures, informed consent, and data collection documents were reviewed and approved by the WIRB Copernicus Group Institutional Review Board (approval no. 202029060) (27, 29). Forty individuals were identified as “SARS-CoV-2–naive” (no SARS-CoV-2 infection before vaccination) or “prior COVID” (previous SARS-CoV-2 infection before vaccination) based on Ag test (PCR), self-reporting, and/or anti–receptor-binding domain (RBD) Ab levels. Individuals who had COVID before the vaccination (prior COVID) were infected between July 2020 and January 2021, covering the Wuhan wave, and samples were collected ∼180 (±5) d after onset of symptoms. Participants received the three vaccinations of Pfizer BNT162b2 monovalent mRNA vaccines and were followed longitudinally. There was no reported history of chronic health conditions or breakthrough infection during the study observation. Cohort and demographic information are listed in Table I.

PBMC samples used in this study were collected at six time points: prevaccine baseline (time point 0), ∼30, 90, and 180 (±5) d after the second vaccination (time points 1, 2, and 3, respectively), and ∼30 and ∼180 (±5) d after the third vaccination (time points 4 and 5) (Fig. 1). BD Vacutainer serum separation tubes (BD Biosciences, Franklin Lakes, NJ) containing whole blood were centrifuged at 1000 × g for 10 min. After centrifugation, the supernatant containing the serum was isolated from the gel layer and RBC pellet. Next, the sera were heat-inactivated in a 56°C water bath for 45 min to disable any infectious SARS-CoV-2 virus and stored at −80°C. For PBMC preparation, whole blood was collected in BD Vacutainer K2 EDTA tubes, homogenized, and transferred to BD Vacutainer cell preparation tubes. Next, cell preparation tubes were centrifuged at 1800 × g for 20 min. The supernatant containing plasma was removed and PBMCs were washed in PBS and submitted to lysis of RBCs. Cells were washed one more time, resuspended in freezing media (90% bovine serum with 10% DMSO), and stored in liquid nitrogen.

ELISA was performed as previously published (29). SARS-CoV-2 RBD protein (USA/WA1/2020 strain) was expressed following the protocol described in Ecker et al. (30) using the vector pCAGGS containing the SARS-related coronavirus 2 Wuhan-Hu-1 Spike glycoprotein RBD. This plasmid was provided from The Biodefense and Emerging Infections Research Resources Repository (BEI Resources) under catalog no. NR-52309. Briefly, Immulon 4HBX (Thermo Fisher Scientific, Waltham, MA) plates were coated with 100 ng/well recombinant SARS-CoV-2 RBD protein, incubated with heat-inactivated serum samples. IgG Abs were detected using HRP-conjugated goat anti-human IgG detection Ab (SouthernBiotech, Birmingham, AL) at a 1:4000 dilution, and colorimetric development was accomplished using 100 µl of 0.1% ABTS (bioWORLD, Dublin, OH) solution with 0.05% H₂O₂ for 18 min at 37°C. The reaction was terminated with 50 µl of 1% (w/v) SDS (VWR International, Radnor, PA). Colorimetric absorbance was measured at 414 nm using a PowerWave XS plate reader (BioTek, Winooski, VT). All samples and controls were run in duplicate, and the mean of the two blank-adjusted OD values was used in downstream analyses. IgG equivalent concentrations were calculated based on a 7-point standard curve generated by a human IgG reference protein from plasma (Athens Research and Technology, Athens, GA) and verified on each plate using human sera of known concentrations.

The neutralization activity of plasma samples against the first SARS-CoV-2 virus detected in Wuhan (SARS-CoV-312 2/INMI1-Isolate/2020/Italy: MT066156) was evaluated using a cytopathic effect–based assay as previously described (31). The experiments were performed in biosafety level 3 laboratories at Toscana Life Sciences (Siena, Italy). Biosafety level 3 laboratories were authorized by a certified biosafety professional and annually examined by local authorities. In brief, sera samples were incubated with a SARS-CoV-2 viral solution containing 100 median tissue culture infective dose of the virus. After 1 h of incubation at 37°C with 5% CO₂, sera/virus mixtures were added to a 96-well plate containing a subconfluent Vero E6 cell monolayer. Then, plates were incubated for 3–4 d at 37°C with 5% CO_2, and the virus-induced cytopathic effect was analyzed using an inverted optical microscope by two independent operators. All samples were tested in duplicates starting at a 1:10 dilution and then 2-fold serially diluted. In each plate, a positive control (human plasma with high neutralizing titers against SARS-CoV-2) and a negative control (human plasma negative for SARS-CoV-2 neutralization) were used as previously described (31).

Enumeration of SARS-CoV-2 Spike and RBD-specific Ab-secreting cells (ASCs) from polyclonally stimulated PBMCs was determined using a FluoroSpot assay platform. In brief, 70% (v/v) ethanol preconditioned high-binding polyvinylidene difluoride filter plates were coated with an anti-human Igκ/λ mixture for total Ig plates. For Ag-specific plates, a 6× His-tagged full-length SARS-CoV-2 Spike protein (27, 30) at 10 µg/ml in PBS was applied to ethanol preconditioned wells precoated overnight at 4°C with 10 µg/ml anti-His tag Ab (BioLegend, San Diego, CA) and incubated overnight at 4°C to improve Ag absorption, as previously described (32). Following one wash with 150 μl of PBS, plates were blocked with 150 μl of B cell media (RPMI 1640 media supplemented with 10% FBS, 2 mM l-glutamine, 1% penicillin/streptomycin) for 1 h at room temperature. PBMCs were previously thawed and cultured at 2 × 10⁶ cells/ml for 5 d under B-Poly-S reagent stimulation (TLR7/8 agonist R848 + recombinant human IL-2) (Cellular Technology Limited, Cleveland, OH). Polyclonally stimulated PBMCs were washed, recounted, and added at 3-fold serial dilutions starting from 1 to 3 × 10⁵ cells/well. Plates were then incubated for 16 h at 37°C, 5% CO₂, and spot-forming units were visualized using the human IgA/IgG/IgM three-color ImmunoSpot kit from Cellular Technology Limited according to the manufacturer’s instructions. Plates were scanned using an S6 Ultimate M2 ImmunoSpot reader and counted using the ImmunoSpot 7.0.28.5 Analyzer software (Cellular Technology Limited).

We employed peptides provided by BEI Resources to characterize SARS-CoV-2–specific T cell response on flow cytometry assays. We pooled 181 peptides spanning the whole Spike glycoprotein of SARS-CoV-2 (USA-WA1/2020, GenPept: QHO60594). The peptides were 17- or 13-mers, with 10-aa overlaps (BEI Resources, catalog no. NR-52402). The pool was made by resuspending individual peptides in 100% acetonitrile, then mixing and lyophilizing. Next the pool was resuspended in 100% DMSO to a final concentration of 1 mg/ml/peptide.

An activation-induced marker (AIM) assay was performed as previously described with the following modifications (10, 23, 33–38). PBMCs were thawed by warming frozen cryovials in a 37°C water bath and resuspending cells in 10 ml of RPMI 1640 media supplemented with 10% FBS, 2 mM l-glutamine, and 1% penicillin/streptomycin in the presence of Benzonase (25 U/ml). Cells were centrifuged at 500 × g for 7 min, supernatants removed, and cells washed with media without Benzonase. Cells were resuspended in culture media (10 ml of RPMI 1640 media supplemented with 10% FBS, 2 mM l-glutamine, 1% penicillin/streptomycin, 2-ME, sodium pyruvate, and nonessential amino acids) and density was adjusted to 1 × 10⁶ cells/200 µl/well. Cells were added in round-bottom 96-well plates and rested overnight in a humidified incubator at 37°C, 5% CO₂. After 16 h, plates were centrifuged, the supernatant was discarded, and cells were blocked at 37°C for 15 min by adding 100 µl of anti-CD40 mAb (Miltenyi Biotec) at a final concentration of 0.5 µg/ml. After blocking, 100 µl of Spike-specific peptide pool (1 µg/ml/peptide final concentration) was added (200 µl final volume). Stimulation with an equimolar amount of DMSO was performed as a negative control. Staphylococcal enterotoxin B (1 µg/ml) and CMV (1 µg/ml, PepTivator CMV pp65, Miltenyi Biotec) were included as positive controls. After 24 h, cells were washed (500 × g, 10 min, 4°C) in PBS and incubated for 20 min at room temperature with LIVE/DEAD (near-infrared red, no. L10119, Thermo Scientific) and Fc receptor blocking solution (human TruStain FcX, BioLegend) (50 µl diluted in PBS). Cells were spun again, carefully resuspended (vortex), and 50 µl of surface Abs diluted in FACS buffer (PBS supplemented with 0.5% BSA and 2 mM EDTA) was added and incubated for 30 min at room temperature. Cells were washed twice with FACS buffer and resuspended in 1% paraformaldehyde in PBS for fixation. Cells were rewashed and resuspended in the FACS buffer before data acquisition. Ag-specific CD4⁺ and CD8⁺ T cells were measured as a percentage of AIM+ cells (OX40⁺CD137⁺ and CD69⁺CD137⁺, respectively). Ag-specific memory (CD45RA, CCR7) and circulating Tfh (cTfh) (CXCR5⁺) cells were defined as a percentage of CD4⁺ T AIM+ (OX40⁺CD40L⁺) cells. Full gating strategy can be found in Supplemental Fig. 1A. All samples were acquired on a NovoCyte Quanteon flow cytometer system with four lasers (Agilent Technologies, Santa Clara, CA). A list of Abs used in this panel can be found in Supplemental Table I. Ag-specific AIM+ CD4⁺ and CD8⁺ T cells were determined by the stimulation index (SI = stimulated/paired unstimulated). A response >2 was considered positive. The Ag-specific AIM+ subsets were measured as background (DMSO) subtracted data, with a minimal DMSO level set to 0.005%. The limit of quantification was calculated using the mean of all negative controls plus 2-fold SD.

Intracellular cytokine staining was performed as previously described with the following modifications (10, 23, 37, 38): PBMCs were rested, blocked with anti-CD40 mAb, and cultured in the presence of a SARS-CoV-2 Spike-specific peptide pool (1 µg/ml) and control stimulus as described above. GolgiPlug containing brefeldin A and GolgiStop containing monensin (BD Biosciences, San Diego, CA) were added after 2 h into the culture and incubated for 16 h more (18 h total incubation). After incubation, cells were washed and stained for surface Abs, as described above. Cells were washed twice with FACS buffer, permeabilized, and fixed for 20 min at room temperature (Fix/Perm, no. 554714, BD Biosciences). Afterward, cells were washed twice with permeabilization buffer, and intracellular cytokines Abs were added and incubated for 30 min at room temperature. Cells were washed once with permeabilization buffer, once with FACS buffer, and finally resuspended in FACS buffer for data acquisition. All samples were acquired on a NovoCyte Quanteon flow cytometer system with four lasers (Agilent Technologies, Santa Clara, CA). A list of Abs used in this panel can be found in Supplemental Table I. The gates were applied to determine CD4⁺ CD40L⁺IFN-γ⁺, CD40L⁺IL-2⁺, or CD40L⁺TNF-α⁺ and CD8⁺ IFN-γ⁺, IL-2⁺, or TNF-α⁺ T cells. Full gating strategy can be found in Supplemental Fig. 1B. A Boolean analysis was performed to determine the multifunctional profiles. Ag-specific CD4⁺ and CD8⁺ T cells were measured as background (DMSO) subtracted data. The limit of quantification for Ag-specific CD4⁺ and CD8⁺ T cell responses was calculated using the mean 2-fold SD of all negative controls.

Flow cytometry data were analyzed using FlowJo 10.7.1. Statistical analyses were performed in GraphPad Prism 9.4 unless otherwise stated. The statistical details of the experiments are provided in the respective figure legends. Data are plotted as the median and interquartile range (IQR). Mann–Whitney U or Wilcoxon tests were applied for unpaired or paired comparisons, respectively. Differences among three or more groups were evaluated using Kruskal–Wallis and a Dunn posttest for multiple comparisons. The positivity rate comparison was determined by χ². Details pertaining to significance are also noted in the respective legends.

Forty participants from the SPARTA cohort were selected for detailed analysis of their collected serum and blood cell samples (Fig. 1). Nineteen of the participants had previously been infected with SARS-CoV-2 Wuhan strain between July 2020 and January 2021, and received mRNA vaccination ∼180 (±5) d after prior infection. The remaining participants were immunologically naive and had no detectable anti-RBD Abs prior to vaccination. Of the 40 participants, 29 were women (15 preimmune and 14 immunologically naive), and 37 (93%) were white. Samples were collected from participants at time of enrollment (baseline, T0) (Table I). Following the initial two vaccination regimen of Pfizer BNT162b2 mRNA vaccination, samples were collected at 30 (T1), 90 (T2), and 180 d (T3) postvaccination. Approximately 180 d after the second vaccination, participants were vaccinated a third time and samples were collected at 30 (T4) and 180 d (T5) postvaccination (Fig. 1).

Anti-RBD binding Abs were measured in collected serum samples. Vaccination induced robust circulating serum Ab responses to the RBD of the Spike protein in all subjects following the second and third vaccinations, regardless of immune status (i.e., naive or preimmune) (Fig. 2A, 2B). These Abs were maintained at detectable levels at the end of the three-dose vaccination regimen in all participants (Fig. 2B). Enrolled participants that were previously infected with SARS-CoV-2 had detectable RBD-specific IgG levels prior to vaccination. Following the second immunization (T1), preimmune participants had significantly higher Ab concentrations compared with immunologically naive participants (p = 0.005) (Fig. 2B). Six months after the second vaccination (T3), there was a 4-fold decline in anti-RBD IgG levels in the sera collected from participants who were previously infected. An enhanced decline (T1–T3, 10-fold decline) was observed in immunologically naive participants with Ab concentrations significantly lower than in preimmune participants (p < 0.0001) at T3. There was a significant increase in the Ab levels in both preimmune and naive participants following the third vaccination (T3 versus T4, p < 0.0001, Fig. 2A), but there was no significant difference in the peak Ab levels between naive and preimmune participants (T4, Fig. 3B). At 180 d following the third vaccination (T5), there was a slight decline in the anti-RBD IgG levels (∼1.5- to 2.5-fold) in both participants groups (Fig. 2B). Taken together, there was a difference in anti-RBD Ab kinetics and a difference in magnitude of vaccine and memory response between naive and preimmune groups following the second vaccination. In contrast, the third vaccination equally boosted naive and preimmune individuals and induced a stable and sustained response 6 mo later.

To determine whether the elicited Abs had neutralization activity against SARS-CoV-2, each serum sample was tested in vitro for the ability to prevent viral infection of susceptible cells. Sera from ∼50% of participants who were infected with SARS-CoV-2 had neutralizing Abs against the ancestral Wuhan strain prior to vaccination (T0) (Fig. 2C, 2D). Following the second immunization (T1), all preimmune participants had neutralizing Abs that were maintained throughout the vaccination schedule (T1–T5). One-hundred percent of immunologically naive participants had neutralizing titers after the second vaccination (T1). However, it was followed by 8-fold decay 6 mo later (T1–T3) and a significantly lower positivity rate compared with preimmune individuals (T3 = 42%, p < 0.001) (Fig. 2C, 2D). Interestingly, the magnitude of the neutralizing Ab titers and the positivity rate of two vaccinations in naive participants (T3 median = 10.42%) were comparable to the response from natural infection before vaccination (preimmune, T0 median = 15.54%) (Fig. 2D). Preimmune participants had significantly higher neutralizing Ab levels following the third vaccination (T4) than did naive participants (Fig. 2D). Six months after the third vaccination (T5), preimmune and naive participants had similar kinetics (∼2-fold decay) and Ab titers (p = 0.11) (Fig. 2C, 2D). Additionally, participants previously infected with SARS-CoV-2 tended to respond better than naive participants at the end of study observation (T5 = 100 versus 93% positivity rate, median = 400 versus 140 Ab titers, respectively). Overall, both RBD binding and neutralizing assay results showed that preimmune participants had a significantly higher peak Ab after initial immunization (T0–T3), but it was not as high as after the third dose (T4–T5).

Specific SARS-CoV-2 Spike and RBD ASCs were measured following the second (T1) and third mRNA vaccinations (T4) (Fig. 3A). One-hundred percent of individuals previously infected with SARS-CoV-2 had IgG Spike-specific ASCs prior to vaccination (T0) and showed a significantly higher frequency compared with naive participants (T0, p = 0.03) (Fig. 3B). The second dose of mRNA vaccine did not significantly increase these Spike-specific ASCs, whereas the third vaccination significantly boosted the number of IgG ASCs in both groups (T1 versus T4; naive p = 0.005 and preimmune p = 0.031) (Fig. 3B). Naive individuals showed a higher response compared with preimmune following the same number of Ag exposures (i.e., T4 from naive versus T1 from preimmune). The IgG ASC RBD-specific responses followed a similar pattern at the prevaccination (T0) and postboost time points (T1, T4) but with a lower frequency compared with Spike (Fig. 3C). IgM preimmune and naive participants had similar frequencies of Spike-specific ASCs after the second vaccination, with no significant changes from baseline (Fig. 3D). The same was observed in RBD measurements, but with a higher response in both groups (Fig. 3E). Following the third vaccination (T4), ∼70% of participants had a rise in the number of IgM Spike- and RBD-specific ASCs (Fig. 3D, 3E). The number of IgA Spike- and RBD-specific ASCs were low at baseline in preimmune individuals and did not significantly increase following the second or third vaccinations. In contrast, IgA-specific ASCs were significantly boosted in naive participants (Fig. 3F, 3G). In addition, there was a significantly higher number of IgA RBD-specific ASCs isolated from naive participants compared with preimmune participants following the third vaccination (T4) (Fig. 3G).

In the comparison of the relative composition of Ab isotypes per group and time point, IgM was the predominant isotype at the prevaccination time point in naive individuals, ranging from 56 to 75% of Spike- and RBD-specific ASCs, respectively. Following the second vaccination (T1), participants had predominantly IgG-secreting Spike- and RBD-specific ASCs (78–95% of isolated ASCs), which were maintained following the third vaccination (88–93% of isolated ASCs). In preimmune participants, most isolated ASCs secreted IgG prior to vaccination (∼50–60%). The number of secreting IgG ASCs increased after consecutive vaccinations. The number of IgA-secreting ASCs represented a minority during the course of vaccinations. The peak in IgA-secreting ASC values occurred following the first encounter with the Spike Ag (i.e., T0 for preimmune and T1 for naive individuals) (Fig. 3H). Overall, preimmune participants retained Spike- and RBD-specific IgG-specific memory B cells generated during the previous SARS-CoV-2 infection. However, there were no significant differences in these responses compared with naive participants at any time point.

SARS-CoV-2 Spike-specific CD4⁺ T cell responses induced by vaccination were first measured utilizing a flow cytometric AIM assay following stimulation of PBMCs with a peptide pool spanning the entire Spike protein sequence (Fig. 4A). AIM+ CD4⁺ T cells were defined by the upregulation of CD137 (4-1BB) and CD134 (OX40). An SI (the ratio of AIM+ T cells in stimulated over unstimulated samples) >2 was used to determine T cell activation. Most participants (74%) who were previously infected with SARS-CoV-2 had detectable Spike-specific CD4⁺ T cell populations prior to vaccination that were significantly higher than those of naive participants (T0, p = 0.01) (Fig. 4B, 4C). However, there was no significant increase in AIM+ CD4⁺ T cells after the second mRNA vaccination (T0 × T1, p = 0.733). SARS-CoV-2 Spike-specific CD4⁺ T cells were detectable in all naive participants following the second mRNA vaccination (T0 × T1, p = 0.001) and had a comparable SI median (∼3.5) compared with preimmune participants (Fig. 4B, 4C). Both groups had a similar and slight drop (∼1.5-fold change) of AIM⁺ CD4⁺ T cell responses 6 mo later (T3) (Fig. 4B). The SARS-CoV-2 Spike-specific CD4⁺ T cells did not significantly increase following the third vaccination (T3 × T4) in either naive or preimmune participants; however, the number of participants that had detectable CD4⁺ T cells increased to 89–100% (Fig. 4C). Six months following the third vaccination (T5) AIM+ CD4⁺ T cells contracted again and dropped ∼2-fold (T4 × T5) in both groups of participants (Fig. 4B). One year after initial vaccination (T5), preimmune and naive individuals had a similar frequency of Spike-specific CD4⁺ T cells and number of responders. No significant difference between naive and preimmune participants was observed between the same number of exposures (three times) at vaccine response time points (p = 0.350, T1 from preimmune and T4 from naive). However, preimmune individuals had a significantly higher response than did naive individuals at memory time points (p = 0.002, T3 from preimmune and T5 from naive). Overall, preimmune participants had fewer significant changes and a higher positivity rate compared with naive participants during the course of observation.

SARS-CoV-2–specific CD8⁺ T cells were defined by the upregulation of CD137 (4-1BB) and CD69 utilizing the AIM assay (Fig. 5A). In contrast to Spike-specific CD4⁺ T cells, only a few preimmune participants (33%, SI > 2) had detectable CD8⁺ T cell responses prior to vaccination. Following the second mRNA vaccination, the number of CD8⁺ T cells from preimmune participants increased almost 2-fold, but returned to the baseline levels 6 mo later (T3). The third vaccination induced a similar fold increase and then decline as the second vaccination, with ∼25% of the participants maintaining Ag-specific CD8⁺ memory T cells (Fig. 5B, 5C). Forty-four percent of naive participants had detectable Ag-specific CD8⁺ T cell responses following the second vaccination (T1), but overall were at approximately baseline levels for 6 mo (T0–T3). Twenty-five percent of naive participants continued to maintain Ag-specific CD8⁺ T cells 6 mo after the second vaccination (T3). The third vaccination induced the highest change (1.5-fold increase), peak of vaccine response, and positivity rate (50%). Following 6 mo after the third vaccination, 13% of naive participants had AIM⁺ CD8⁺ T cells (Fig. 5B, 5C). At 12 mo from initial vaccination, naive and preimmune participants had similar CD8⁺ T cell dynamics, both with peak responses following each vaccination and returned to baseline levels. No significant difference between naive and preimmune subjects was observed between the same number of exposure (three times) at vaccine response time points (T1 from preimmune and T4 from naive) and memory time points (T3 from preimmune and T5 from naive). Overall, vaccination recalled pre-existing CD8⁺ T cell immunity from previous infection and generated a low-magnitude booster response.

Spike-specific CD4⁺ (Fig. 6A) and CD8⁺ T (Supplemental Fig. 2A) cells were characterized by differentiation status using CCR7 and CD45RA markers to define naive, CM, EM, and terminally differentiated EM T cell (TEMRA) populations. The distribution of memory subsets of SARS-CoV-2–specific CD4⁺ T cells (AIM+ cells) in the peripheral blood was predominately CM at all time points (Fig. 6B, 6C). Preimmune individuals had a slightly higher fraction (28%) of CD4⁺ EM cells compared with naive (12%) individuals prior to vaccination (T0). Naive individuals after receiving the second vaccination (T1) demonstrated a shift from the naive to EM subset, which was similar to the infection-generated response in preimmune individuals (T0 = 6 mo postinfection). During the course of the vaccination schedule, the memory T cell subset distribution of these Spike-specific CD4⁺ T cells did not change substantially. Six months after last vaccination (T5), CD4⁺ EM cells in preimmune individuals returned to the baseline values, but maintained a larger proportion than the vaccine-generated response observed in the naive group (10%) (Fig. 6B). The distribution of CD8⁺ T cells was mainly composed of EM and TEMRA subsets (Supplemental Fig. 2B, 2C). Prior to vaccination, 46% of Spike-specific CD8⁺ T cells in the preimmune group were composed mainly of the TEMRA subset. In contrast, naive individuals presented a large fraction of naive subsets (43%) at baseline (T0). After the second vaccination (T1), naive individuals reached the same distribution observed at the infection-generated response (T0) in preimmune individuals. These proportions stayed relatively consistent throughout the course of vaccination in preimmune subjects, and there were no statistically significant changes in naive individuals. One year after the initial immunization (T5), the naive fraction of CD8⁺ Spike-specific cells in naive individuals returned to baseline values, whereas the preimmune participants maintained a predominance of the TEMRA subset. These data indicate that prior infection durably shapes the differentiation state of the memory immune compartment (Supplemental Fig. 2B).

Spike-specific cTfh cells within the AIM+ subset were assessed (Fig. 7A). The number of Spike-specific cTfh cells were ∼2-fold higher (p = 0.023) in participants previously infected with SARS-CoV-2 than in naive participants prior to vaccination (T0, Fig. 7B). Preimmune subjects were significantly boosted after the second vaccination (T0 × T1, p = 0.009) (Fig. 7C) and maintained a significantly higher frequency of cTfh cells compared with naive group (T1, p = 0.014) (Fig. 7B). There was no significant difference between naive and preimmune participants following the third vaccination (T4). At T5, cTfh cell frequency was significantly higher (p = 0.034) in preimmune than in naive participants (Fig. 7B) and significantly correlated with binding and neutralizing Abs (p = 0.043 and 0.039, respectively).

Cytokine-producing Spike-specific CD4⁺ T cells were determined within activated CD40L⁺ cells (Fig. 8). Participants previously infected with SARS-CoV-2 had a significantly higher frequency of IFN-γ–, IL-2–, and TNF-α–secreted CD4⁺ T cells than did naive participants prior to vaccination (T0) (Fig. 8C). The second dose (T1) induced an increase of Spike-specific pre-existing memory (IFN-γ = 2.1-, IL-2 = 1.6-, and TNF-α = 1.9-fold increase) (Fig. 8B) to significantly higher levels than for naive donors (T1) (Fig. 8C). These levels were maintained for up to 6 mo until receiving the third vaccination. Following the third vaccination (T4), preimmune participants had a slight increase in the frequency of functional CD4 cells, which significantly declined during the next 6 mo (T5) (Fig. 8B). There were increased T cells secreting Th1 cytokines (IFN-γ = 5.7-, IL-2 = 5.7-, and TNF-α = 4.4-fold increase) following the second vaccination (T1) in naive participants (Fig. 8A). The response from naive participants followed similar dynamics as preimmune individuals, all with a significant decay at the end of the three-dose vaccination schedule (T5) but remaining above detectable levels (Fig. 8A).

Analysis of T cell polyfunctionality in preimmune participants showed that 42% of the CD40L⁺ CD4⁺ T cells produced at least two cytokines and 15% produced all three cytokines at prevaccination (T0) (Supplemental Fig. 3A). Naive participants reached similar levels after the second vaccination (T1). Both group participants maintained similar profiles after the second vaccination (T1) until receiving the third vaccination (T4). Six months after the third vaccination (T5), naive and preimmune individuals lost this multifunctional profile, with 75% of activated CD40L⁺ CD4⁺ T cells producing predominantly one cytokine out of IFN-γ, TNF-α, and IL-2. Overall, SARS-CoV-2 infection modulated the vaccine response by inducing more robust functional Th1 memory in recovered individuals, but after initial priming (first two vaccinations), vaccine recipients established an equivalent memory response.

A low number of cytokine-producing Spike-specific CD8⁺ T cells (IFN-γ, TNF-α, and IL-2) was observed in both naive and preimmune participants following the second or third vaccinations (Fig. 9A, 9B). There were no significant differences in the frequency of cytokine-producing Spike-specific CD8⁺ T cells between naive and preimmune participants during the course of vaccination (Fig. 9C). In the multifunctional profile, ∼25% of CD8⁺ T cells secreted two cytokines and only 2% secreted three cytokines during the course of vaccination (Supplemental Fig. 3B).

Correlation analysis showed a relationship between AIM CD4⁺ T cells from previous exposure (regardless of vaccination or infection) to neutralizing Ab titers at late time points (Fig. 10). AIM+ CD4⁺ memory T cells elicited following SARS-CoV-2 infection (T0) significantly correlated with neutralizing Abs titers detected following the third vaccination (T5). Following the second vaccination (T1), CD4⁺ T cells positively correlated with neutralizing Ab responses at subsequent time points (T2–T5). At the end of the study observation (T5) in preimmune participants, Spike-specific CD4⁺ T cells and neutralizing Abs significantly had a coordinated response. A similar pattern was observed in naive participants after the second exposure (T3 AIM⁺ CD4⁺ positively correlated with T5 Wu neutralizing Ab). No relationship was observed in naive participants at 6 mo after the third vaccination (AIM+ CD4⁺ and Wu neutralizing Ab at T5). These data indicate that early induction of Spike-specific CD4⁺ T cells, either by infection or vaccination, is associated with increased humoral responses.

Because mRNA vaccines have been administered globally, it is crucial to understand their immunological features in different immunological settings. The repeated Ag exposures through the vaccination on the development of SARS-CoV-2 memory T cells and Ab responses are essential for determining susceptibility to subsequent infections and informing booster vaccination strategies. Few longitudinal data exist that simultaneously examine multiple arms of the immune response from the same immunologically naive people prior to vaccination with no breakthrough infections. Also, there are still gaps in the literature on the magnitude and duration of the immune response conferred by previous infection(s) among individuals who have been vaccinated (i.e., the effectiveness of hybrid immunity). Our goal was to characterize the longitudinal Spike-specific immune responses (Ab, memory B and T cell) induced by three vaccinations of the BNT162b2 vaccine and determine whether people previously infected with the Wuhan strain would elicit differential long-term immune memory compared with immunologically naive participants. This study supports two overall conclusions: 1) repeated Spike exposure increases the quantity and quality of Ab memory response, but has a reduced ability to improve T cell responses, and 2) natural infection enhances the vaccine responses shortly after initial immunization with a minimum impact in long-term immunity.

The magnitude and positivity rate of binding and neutralizing IgG titers progressively increased following the vaccination and declined in subsequent months. Notably, the third dose significantly enhanced the humoral response by increasing the IgG magnitude compared with the second dose, right after the immunization as well as 6 mo later. It supports the important role of additional boosters in the improvement of the quality of humoral responses and longevity. Preimmune individuals also benefited from the third immunization and demonstrated a more stable response with fewer changes during the course of the vaccination schedule compared with naive individuals. A higher magnitude of Ab titers has been shown in hybrid immunity (i.e., people with natural immunity who are subsequently vaccinated) compared with vaccination only (39–42). Consistent with previous data (14, 43–45), preimmune individuals presented higher early immune response patterns than did naive participants. However, although there were higher IgG Ab levels in preimmune participants shortly after vaccination, these enhanced effects gradually diminished and were not observed in the long term.

Although we observed a benefit from previous infection in serum Ab response after a second dose of mRNA vaccination, the same was not observed for Spike and RBD IgG-specific memory B cells. We detected a higher number of Ag-specific IgG ASCs in preimmune participants prior to vaccination, indicating a rapid recall of pre-existing immunity generated by previous infection(s) and corroborating the serological observations. However, upon the second and third vaccinations, both subject groups recalled a similar frequency of B memory cells. Of note, previous studies have shown that Ag-specific B memory cells from previously infected individuals declined following vaccination, whereas they increased in immunologically naive people (12, 14). Even though additional analysis is needed, the Spike-based vaccine may result in a high-affinity selection and more prolonged activity on germinal centers (46, 47) leading to a superior recall responses than the infection.

In contrast to Ab responses, Spike-specific T cell frequencies were boosted by the third vaccination, but the levels did not significantly increase in the peak and memory response compared with the second dose. Repeated Spike vaccinations recalled activated T cells that sustained the response over time. Less than a 2-fold difference in Spike-specific CD4⁺ T cell frequencies was observed between the immune peak and 6 mo following the booster vaccination, indicative of durable vaccine T cell memory. In contrast, CD8⁺ T cell responses were slightly boosted and unchanged over time, corroborating previous observations (1, 10, 19). The type of vaccine, route, interval dosing, and individual immune status can affect the T cell response longitudinally (4, 48–51). We have observed moderate differences in circulating Spike-specific memory CD4⁺ T cells between vaccination only and hybrid immunity. T cells are retained postinfection and recalled after mRNA vaccination. However, unlike IgG Ab responses, the enhanced response is not sustained following the initial vaccination peak in preimmune individuals. Immune responses were enhanced by the presence of pre-existing cross-reactive CD4⁺ T cell memory over time, and previously infected participants had a consistently higher magnitude of CD4⁺ AIM+ cells and a larger percentage of Spike-specific responders at all time points assessed. Additionally, hybrid immunity induced a better long-term CD4⁺ response at the late memory stages (6 mo postvaccination) when comparing the same number of exposures.

The immune response from SARS-CoV-2 infection is expected to surpass that observed in participants who were only administered the vaccine (52). Infection generates a greater magnitude and breadth of B and T cell repertoires targeting SARS-CoV-2 structural, nonstructural, and accessory proteins (48, 50). Furthermore, despite the induction of robust circulating humoral and cellular immunity, current COVID-19 mRNA vaccines likely do not provoke sufficient levels of mucosal immunity in the human lower respiratory. Respiratory mucosal Abs and memory T and B cells are likely among the early responders and rapidly recalled during Ag re-encounter (53). SARS-CoV-2 infection induces peripheral virus-specific T cells with phenotypic characteristics of tissue homing (26). This suggests that the low Spike-specific CD4⁺ T cell responses after the second vaccination in preimmune individuals may have resulted from these cells preferentially leaving the blood compartment and migrating either to draining lymph nodes to participate in germinal center reactions or to lungs during inflammation. Additionally, previous data have shown that breakthrough infections robustly activate cellular responses and induce more polyfunctional CD4⁺ cells compared with vaccination only (Ref. 48 and M.M. Painter, T.S. Johnston, K.A. Lundgreen, J.J.S. Santos, J.S. Qin, R.R. Goel, S.A. Apostolidis, D. Mathew, B. Fulmer, J.C. Williams, et al., manuscript posted on medRxiv, DOI: 10.1101/2023.02.05.527215). The detection of robust and functionally efficient virus-specific T cell (not only Spike) responses is an important component of the defense mechanisms of the host against developing COVID-19 (48, 50, 51). However, it is still unclear whether exposure history can be rewritten and expanded through heterologous vaccination. Previously published data support the hypothesis that other non-Spike viral proteins could broaden cellular immune response potency and consequently mediate lasting vaccine effectiveness against severe COVID-19 disease and hospitalization.

To further understand the effect of preimmunity on vaccine-induced responses, we expanded our observations to specific T cell subsets. Memory subsets, Th1 cytokine-producing cells, and cTfh cells have implications for long-term protective immunity and can induce response as early as neutralizing Abs (13, 14, 19, 54, 55). In this study, previously infected individuals had changes in CD4⁺ T cell subsets, suggesting that the pre-existing Spike memory CD4⁺ T cells are recalled upon vaccination and impact the CD4⁺ T cell repertoire. SARS-CoV-2 infection resulted in long-lived Spike-specific CM and EM cells, concomitant with previous data (11, 12, 19). CM cells remained predominant after consecutive vaccinations, suggesting an increased potential for long-term persistence, as they have longer in vivo half-lives than EM and TEMRA cells. Vaccine-generated CD4⁺ T cells showed a similar response, indicating similar quality and quantity memory, but hybrid immunity had improved long-term responses with a larger percentage of effector subsets at the memory phase. The recirculation of CM and EM cells and access to secondary lymphoid tissues have implications on Ag encounters and rapid effector responses. The vaccination also exhibited an enhanced antiviral functional profile in preimmune individuals, including substantial Tfh cells and IFN-γ expression. Although higher levels of IFN-γ IL-2, and TNF-α were recalled after the second immunization, cTfh cells also had higher frequencies at a late stage of observation and correlated with neutralizing Ab levels. This indicates that infection-generated cTfh cells may both accelerate B cell priming and Ab responses and also increase the robustness of long-term humoral immunity, as evidenced by the higher neutralizing Ab titers. In contrast, Th1 cells are not involved in B cell maturation and instead support cellular and innate immunity against pathogens (22).

This study also demonstrated that humoral immune responses were significantly impacted by the presence of pre-existing SARS-CoV-2–specific CD4⁺ T cell memory. Infection-induced CD4⁺ T cells had a long-lasting benefit for the Spike-specific IgG Ab response. Additionally, primed CD4⁺ T cell responses were correlated with boosted neutralizing Abs and, therefore, an early indicator of the induction of long-term humoral immunity. Immune responses elicited by viral infection contributed to a coordinated long-term B and T cell immunity following mRNA vaccinations (1, 10, 11, 47). This may be due to the rapid mobilization of infection-induced memory CD4⁺ T cells, which augmented humoral immunity postvaccination.

There are a few caveats to note in this study. Although PBMCs provide insight into peripheral immune signatures, these cells do not reflect distinct features at the site of infection. Moreover, Spike-specific memory cells can be efficiently activated and undergo proliferative expansion, but expanded cells can exit the bloodstream and accumulate in tissues at the sites of infection. Also, although peptide stimulation assays can provide an estimate of the total magnitude of the CD8⁺ T cell response, they can underestimate the frequency of epitope-specific T cells, in part due to the high background. Lastly, evidence supports (56) an important role of the early activation of virus-specific T cells in the rapid reduction of viral replication, and thus testing full-length Spike could have obscured the recall of the additional SARS-CoV-2 repertoire induced during the infection.

Overall, SARS-CoV-2 preimmunity induced by infection impacted the immune responses to the mRNA COVID-19 vaccination in this cohort. The findings presented demonstrate that immunity generated by infection guides multiple arms of the adaptive immune responses following mRNA vaccination. Infection-induced immunity enhances the humoral and T cell responses following initial mRNA vaccination in the form of neutralizing activity, CD4⁺ memory phenotype, cTfh cells, and effector functions (Th1 cytokines), but it has a slight benefit on long-term immunity. The findings of the development of SARS-CoV-2 memory T cell and Ab responses on the two immunologically heterogeneous groups based on prime infection are relevant for determining susceptibility to subsequent infections and informing booster vaccination strategies. The early expansion of functional T cell subsets coupled with Ab production presented in our study may have implications on rapidly controlling SARS-CoV-2 replication and attenuation of disease severity. Thus, additional immunizations that resemble the immune response to SARS-CoV-2 infection, preferentially enhancing key hallmarks of long-term antiviral immunity (Th1-secreting cells, cTfh cells, and EM cell subsets) will likely confer lasting protection against SARS-CoV-2 infection.

The authors have no financial conflicts of interest.

We thank Cellular Technology Limited LLC, and in particular Greg A. Kirchenbaum, for technical support with the FluoroSpot experiments; the Center for Tropical & Emerging Global Diseases Cytometry Core at the University of Georgia, in particular Julie Nelson for assistance on flow cytometry assays; and the Biodefense and Emerging Infections Research Resources Repository (BEI Resources) for provide the peptide array (NR-52402) and the vector pCAGGS (NR-52309).