Abstract

Introduction

   Pre-existent pools of coronavirus-specific or cross-reactive T cells
   were shown to shape the development of cellular and humoral immune
   responses after primary mRNA vaccination against SARS-CoV-2. However,
   the cellular determinants of responses to booster vaccination remain
   incompletely understood. Therefore, we phenotypically and functionally
   characterized spike antigen-specific T helper (Th) cells in healthy,
   immunocompetent individuals and correlated the results with cellular
   and humoral immune responses to BNT162b2 booster vaccination over a
   six-month period.

Methods

   Blood of 30 healthy healthcare workers was collected before, 1, 3, and
   6 months after their 3rd BNT162b2 vaccination. Whole blood was
   stimulated with spike peptides and analyzed using flow cytometry, a
   13-plex cytokine assay, and nCounter-based transcriptomics.

Results

   Spike-specific IgG levels at 1 month after booster vaccination
   correlated with pre-existing CD154+CD69+IFN-γ+CD4+ effector memory
   cells as well as spike-induced IL-2 and IL-17A secretion. Early
   post-booster (1-month) spike IgG levels (r=0.49), spike-induced IL‑2
   (r=0.58), and spike-induced IFN‑γ release (r=0.43) correlated
   moderately with their respective long-term (6-month) responses.
   Sustained robust IgG responses were significantly associated with
   S-specific (CD69+±CD154+±IFN-γ+) Th-cell frequencies before booster
   vaccination (p=0.038), especially double/triple-positive type-1 Th
   cells. Furthermore, spike IgG levels, spike-induced IL‑2 release, and
   spike-induced IFN‑γ release after 6 months were significantly
   associated with increased IL‑2 & IL‑4, IP‑10 & MCP1, and IFN‑γ & IP‑10
   levels at 1 month post-booster, respectively. On the transcriptional
   level, induction of pathways associated with both T-cell proliferation
   and antigen presentation was indicative of sustained spike-induced
   cytokine release and spike-specific IgG production 6 months
   post-booster. Using support vector machine models, pre-booster
   spike-specific T-cell frequencies and early post-booster cytokine
   responses predicted sustained (6-month) responses with F1 scores of
   0.80-1.00.

Discussion

   In summary, spike-specific Th cells and T-cellular cytokine signatures
   present before BNT162b2 booster vaccination shape sustained adaptive
   cellular and humoral responses post-booster. Functional T-cell assays
   might facilitate early identification of potential non-responders.

   Keywords: COVID-19, immunity, memory cells, immune monitoring,
   antibodies, cytokines, transcriptomics

1. Introduction

   Since the beginning of the Coronavirus disease 2019 (COVID-19)
   pandemic, we witnessed a uniquely fast and successful development of
   vaccines against severe acute respiratory syndrome coronavirus-2
   (SARS-CoV-2). Extensive evidence suggests that booster vaccination,
   i.e., additional vaccine doses following primary immunization, provides
   enhanced protection against death or severe COVID-19 ([45]1). However,
   as the virus is continuously evolving, breakthrough infections (BTIs)
   following vaccination have become more frequent, given the emergence of
   immune escape variants of SARS-CoV-2 ([46]2–[47]4).

   Many conditions associated with suppression of the adaptive immune
   response can considerably hamper the effect of COVID-19 vaccination
   ([48]5, [49]6). However, the relative contributions of the different
   components of the adaptive immune system to vaccine-induced immune
   protection, including the determinants of long-term immunity after
   booster vaccination, remain incompletely understood. In particular, the
   role of pre-existing antigen-specific T-helper (Th) cells remains to be
   fully defined in the context of booster vaccination.

   Even though pre-existing, seasonal coronavirus-specific T cells can
   confer some protection against SARS-CoV-2 infection ([50]7), these
   cells do not seem to be a major contributor to the development of
   immunity against SARS-CoV-2 after infection or vaccination ([51]8).
   Instead, pre-existing memory T cells reacting to the SARS-CoV-2 spike
   protein (S) might even impair the development of a varied T-cell and
   neutralizing antibody response after the primary BNT162b2 vaccination
   series ([52]8). This effect has been shown to be age-dependent, as
   elderly vaccinees have developed a diverse coronavirus-specific T-cell
   repertoire over their lifetime. These cells cross-react with SARS-CoV-2
   epitopes, albeit with low avidity ([53]9, [54]10). Thus, this response
   does not mature or diversify into higher-avidity antigen receptor
   variants and cannot initiate high-quality responses ([55]8). On the
   other hand, pre-existing S-reactive T cell populations in younger
   vaccinees exhibit naïve phenotypes (CCR7^+CD45RA^+), have high
   plasticity, and can initiate diverse and high-quality responses, i.e.,
   consistent production of neutralizing antibodies and cytokines after
   primary immunization ([56]8). Cytotoxic (CD8^+) T-cell responses were
   shown to vary considerably between individuals and most vaccinees
   develop only low frequencies of SARS-CoV-2 specific CD8^+ T cells
   following the first three vaccinations ([57]11).

   Significant variation in the robustness and longevity of antibody and
   T-cell responses to booster vaccination against SARS-CoV-2 has been
   extensively documented ([58]12–[59]17). However, the determinants of
   sustained adaptive immune responses to booster vaccination against
   SARS-CoV-2, including the role of T cells present pre-booster, remain
   scarcely studied. Therefore, we phenotypically and functionally
   characterized S-specific Th cells in healthy, immunocompetent
   individuals and correlated the results with cellular and humoral immune
   responses to BNT162b2 booster vaccination over a six-month period.
   Thereby, we identified S-specific Th cell populations integral to
   short-term T-cell expansion, along with cytokine profiles in response
   to S peptides that correlate with longevity of key antiviral effector
   functions. These insights might help to identify potential
   low-responders to SARS-CoV-2 booster vaccination.

2. Methods

2.1. Subjects and blood collection

   As part of an ongoing registry study (Corona Registry Study number
   20–426, approved by the ethics committee of the Ludwig Maximilians
   University Munich), thirty healthy volunteers were recruited among
   employees of the University Hospital of Augsburg. All subjects had
   received primary vaccination with two doses of BNT162b2 (Pfizer Inc.,
   New York, NY, USA and BioNTech SE, Mainz, Germany) and were enrolled
   before their third dose of BNT162b2 (first booster). Exclusion criteria
   were active infections, receipt of immunomodulatory therapy, pregnancy,
   positive nucleocapsid-specific (N) IgG titers, and underlying
   conditions associated with significant impairment of anti-COVID-19
   immunity (e.g., diabetes mellitus, cancer, prior stem cell
   transplantation, or autoimmune diseases). Lithium
   heparin-anticoagulated whole blood and serum samples were collected
   immediately before the third vaccination (T0), as well as 1 (T1), 3
   (T3), and 6 months (T6) after booster vaccination ([60] Supplementary
   Figure S1 ). Demographic data of study participants are summarized in
   [61]Supplementary Table S1 for each individual timepoint. No additional
   booster doses were given during the study period.

2.2. Antibody measurement

   Serum was tested for N- and S-specific IgG using a COBAS e801
   immunoassay analyzer with Elecsys Anti-SARS-CoV-2 and Anti-SARS-CoV-2 S
   (Hitachi Ltd., Tokyo, Japan and Roche Diagnostics GmbH, Basel,
   Switzerland) kits. Samples exceeding the upper limit of detection were
   diluted and re-measured. Serum remainders were cryopreserved for
   analysis of neutralizing antibody titers (NAT). Subjects developing
   BTIs were excluded from the time of first positive
   nucleocapsid-specific IgG titer (>1.0 cut-off index). Likewise,
   subjects who missed a follow-up measurement were excluded from all
   subsequent timepoints due to unknown BTI status. NAT were determined as
   previously described ([62]18).

2.3. Preparation of stimulation tubes for whole blood-based immunoassays

   As published previously ([63]19, [64]20), 2.7-mL blood collection tubes
   without anticoagulant were pre-loaded with co-stimulatory antibodies
   α-CD28 and α-CD49d plus the following stimuli: (i) negative control (no
   antigen), (ii) SARS-CoV-2 S peptides, (iii) or positive control
   (mixture of CPI and CEF peptides). Details regarding antigens,
   reagents, manufacturers, and concentrations are provided in
   [65]Supplementary Table S2 .

2.4. Flow cytometric characterization of spike-reactive T cells

   Stimulation tubes prepared as described above were inoculated with 500
   µL whole blood. After 4 h of incubation at 37°C, brefeldin A was added
   at a final concentration of 10 µg/mL, followed by incubation for
   another 16-18 h at 37°C. After erythrocyte lysis using Buffer EL
   (Qiagen, Hilden, Germany), extracellular staining was performed using
   CD197(CCR7)-BV650 (BioLegend, San Diego, USA), CD3-VioBright R720,
   CD4-VioBlue, CD8-VioGreen, CD45RO-APC-Vio770, CD137-APC,
   CD279(PD1)-VioBright 515, and 7-AAD (all from Miltenyi Biotec, Bergisch
   Gladbach, Germany) according to the manufacturer’s instructions.
   Intracellular staining with CD69-PE-Vio615, CD154-PE-Vio770, and
   IFN-γ-PE (all from Miltenyi Biotec) was performed using the Inside
   Stain Kit (Miltenyi Biotec) according to the manufacturer’s
   instructions. Stained cells were analyzed using a Cytoflex S flow
   cytometer (Beckman Coulter, Brea, USA) and Cytexpert software (version
   2.4.0.28, Beckman Coulter, Brea, USA). Downstream analysis was
   performed with Kaluza (version 2.1, Beckman Coulter).

2.5. Functional characterization of spike-reactive T cells by cytokine and
transcriptome analysis

   For cytokine secretion assays and transcriptome analysis, stimulation
   tubes ([66] Supplementary Table S2 ) were incubated for 20 – 24 h at
   37°C without addition of brefeldin A. Concentrations of secreted
   cytokines were measured in supernatants using the 13-plex LegendPlex
   Essential Immune Response Kit (BioLegend).

   The residual cell pellet underwent erythrocyte lysis and was then
   cryopreserved in 1 mL RNAprotect (Qiagen). Total RNA was isolated using
   the RNeasy Plus Mini Kit (Qiagen). Transcripts for 785 immune-related
   genes were quantified using the nCounter sprinter platform and Human
   Immune Exhaustion panel (Nanostring, Seattle, USA). Data was analyzed
   using the nSolver Analysis Software, with elimination of flagged
   datasets, background thresholding to the median of negative controls,
   and normalization to the panel’s 12 housekeeping genes (geometric
   mean). Thereafter, the ratio of normalized mRNA counts in S-stimulated
   samples versus unstimulated background controls was determined for each
   subject. After classification of subjects by serological and cytokine
   responses, as detailed in the Results section, median
   S-stimulated/background ratios were determined for each cohort.
   Median-to-median ratios (MMRs) between high- and low-responders were
   calculated and imported into Ingenuity Pathway Analysis (Qiagen). Core
   analysis was performed to determine canonical pathway enrichment.
   Enrichment of canonical pathways was considered significant at a
   Benjamini-Hochberg adjusted p-value <0.05. For inclusion in figures and
   to determine biological significance of pathway activation/inhibition,
   absolute z-score thresholds of ≥1.5 and ≥2 were applied, respectively.
   The data discussed in this publication have been deposited in NCBI’s
   Gene Expression Omnibus ([67]21) and are accessible through GEO Series
   accession number [68]GSE273994
   ([69]https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE273994).

2.6. Statistical analysis

   To determine S-specific T cell frequencies and cytokine release,
   results of unstimulated controls were subtracted from S-stimulated
   responses. Unless indicated otherwise, individual results and medians
   are shown. Spearman rank coefficients were used for correlation
   analysis. Depending on the data format, Mann-Whitney-U-test, Wilcoxon’s
   rank sum test, Kruskal-Wallis test, or Friedman test were used for
   significance testing, as specified in the figure legends. Dunn’s
   post-test was used for multi-group comparisons. For analyses with 5 or
   more readouts, the Benjamini-Hochberg test was used to identify
   significant results with a false discovery rate of <0.2. Data analysis
   and visualization was performed using Excel 365 (Microsoft Corporation,
   Redmond, USA) and Prism v9 (GraphPad Software, Boston, USA).

   Support Vector Machine (SVM) analyses were performed using an online
   calculator ([70]22). Top and bottom responder classification of
   subjects at T6 (based on S IgG, S-induced IL-2, IFN-γ, or a combination
   of all three readouts) were predicted by training linear kernel models
   with independent variables pre-selected using the significance tests
   described above. Models were generated in multiple iterations by
   successively adding independent variables to those yielding the highest
   F1 scores in the previous iteration. This process was repeated until
   either all additional variables worsened F1 scores, no further
   improvement in F1 score was achieved in two successive iterations, a
   maximum of four markers had been selected, or an optimal result was
   achieved (F1 = 1.00). Among combinations of independent variables
   yielding the same F1 score, those with the fewest independent variables
   were selected.

3. Results

3.1. Multifaceted humoral and cellular anti-spike response peak at 1 month
post-booster

   We enrolled a total of 30 subjects at T0. Five of them were male and 25
   were female. Their median age was 51 years (range 26-65 years). Only 7
   subjects had a known previous SARS-CoV-2 infection. One subject at T1,
   6 at T3, and 14 at T6 were excluded due to positive N antibodies or
   missed follow-up ([71] Supplementary Figure S1 ).

   Compared to baseline (pre-booster), we observed strong and significant
   increases in S IgG and NAT in all subjects 1 month after booster
   vaccination (median NAT of 160 at T1 vs. 5 at T0, p<0.001,
   [72]Figure 1A ), followed by a steady decrease during the 6-month
   follow-up period (median NAT of 80 and 40 at T3 and T6, respectively).
   These trends were corroborated by matched analysis of subjects with
   complete follow-up data (T0 vs. T1 in [73]Supplementary Figure S2A , T0
   vs. T1 vs. T6 in [74]Supplementary Figure S3A ).

Figure 1.

   [75]Figure 1
   [76]Open in a new tab

   Longitudinal evolution of spike-specific immune responses. Humoral and
   cellular spike (S)-specific immune responses were monitored before
   booster vaccination (T0) as well as 1 (T1), 3 (T3), and 6 months (T6)
   post-booster. (A) S IgG and neutralizing antibody titers. (B)
   S-specific Th cell frequencies among viable CD3^+CD4^+ lymphocytes
   based on the expression of CD154, IFN-γ, and/or CD69. (C)
   Characterization of the (background-adjusted) S-specific Th cell
   population expressing one or more of the activation markers CD154
   (orange), IFN-γ (blue) or CD69 (green). (D) Phenotypic characterization
   of unstimulated and S-reactive Th cells based on CD45RO and CCR7
   expression levels: T[E]/T[EMRA] (blue), CD45RO^-CCR7^-; T[EM] (red),
   CD45RO^+CCR7^-; T[CM] (grey), CD45RO^+CCR7^+; T[N]/T[SCM] (yellow),
   CD45RO^-CCR7^+. (E) Background-adjusted S-induced secretion of IFN-γ,
   IL-2, IL-4, IL-10, IL-17A, and CXCL10. (A, B, E) Individual values,
   median, interquartile range (boxes), and spread (whiskers) are shown.
   Paired/matched analyses are shown in [77]Supplementary Figure S2 and
   [78]Supplementary Figure S3 . (A–C, E) Kruskal-Wallis test with Dunn’s
   post-test. * p<0.05, ** p<0.01, *** p<0.001. Significance indicators in
   (C) refer to comparisons at T1/T3/T6 versus T0. S, spike; T[E]/T[EMRA],
   effector T cells and effector memory T cells re-expressing CD45RA;
   T[EM], effector memory T cells; T[CM], central memory T cells;
   T[N]/T[SCM], naïve T cells and stem cell memory T cells.

   Likewise, S-specific Th-cell reactivity, as measured by intracellular
   expression of CD154 (median, 0.086% at T1 vs. 0.062% at T0), CD69
   (0.597% vs. 0.248%), and IFN-γ (0.054% vs. 0.030%), or any combination
   of these markers (0.627% vs. 0.282%), moderately increased after
   booster vaccination ([79] Figure 1B ; [80]Supplementary Figures S2B ,
   [81]S3B ). However, due to considerable inter-individual baseline
   variation, statistical significance for most markers was only seen upon
   paired T0/T1 analysis (CD154, p=0.014; IFN-γ, p<0.001; any marker,
   p=0.048, [82]Supplementary Figure S2B ).

   At all timepoints, >20% of S-specific Th cells expressed more than one
   activation marker, with a minor and non-significant increase at T1
   (25.8% double/triple-positive, [83]Figure 1C ). With double- and
   triple-positivity, the predominant phenotype of S-specific Th cells
   shifted from a central memory phenotype (T[CM]) to an effector memory
   phenotype (T[EM]) from T1 onwards ([84] Figure 1D ). Likewise, the
   proportion of S-specific T[EM] Th cells increased while the T[CM]
   phenotype decreased after booster vaccination compared to baseline,
   especially at T6 (p<0.01 for 7/7 [T[EM]] and 6/7 [T[CM]] combinations
   of activation markers, respectively, [85]Figure 1D ).

   Notably, S-induced release of several T-cellular cytokines (IFN-γ,
   IL-2, IL-4, IL-10, IL-17) and chemokines associated with IFN-γ release
   (CXCL10) in S-stimulated whole blood was markedly stronger at T1 vs. T0
   ([86] Figure 1E ), with statistical significance reached for all
   markers in at least one of the two paired/matched analyses ([87]
   Supplementary Figures S2C , [88]S3C ). S-induced release of all
   cytokines reverted to pre-booster levels by T6 ([89] Figure 1E ,
   [90]Supplementary Figures S2C , [91]S3C ). Taken together, these
   findings indicate broad yet mostly short-lived expansion and activation
   of S-specific Th cells, paralleled by a more persistent shift in
   effector/memory phenotypes of Th cells.

3.2. IFN-γ^+ S-specific Th cells present pre-booster shape cytokine and
antibody responses

   For all post-booster assessments, we observed high inter-individual
   variability of serological and T-cell responses ([92] Figure 1 ). Most
   readouts showed no association with age. However, increasing age
   correlated with higher S IgG at T1 (ρ=0.52, p=0.004) but less robust
   sustained S-specific T-cell responses at T6 (significant for %
   S-specific Th cells, ρ=-0.60, p=0.016) ([93] Supplementary Figure S4 ).

   We hypothesized that inter-individual variation may be caused by
   differences in memory and effector Th subsets present at the time of
   booster vaccination (i.e., at T0). Consistent with a previous study
   ([94]23), we found no significant correlation between NAT and
   S-specific T cells in our cohort. Therefore, we focused on correlations
   of frequencies of S-specific T cells, their phenotypes, and double- or
   triple-activation marker positivity at T0 with post-booster S IgG as
   well as S-specific IFN-γ and IL-2 secretion, which are widely
   considered surrogates of protective adaptive immunity ([95]24–[96]26).
   Interestingly, we found a strong and significant positive correlation
   of post-booster (T1) cytokine responses with total pre-booster double-
   and triple-positive IFN-γ producing S-specific Th cells (ρ=0.45-0.56),
   especially the naive/stem cell memory (T[N]/T[SCM]) sub-population
   (ρ=0.26-0.48, [97]Figure 2A ). Additionally, we found a strong
   association of S-specific IFN-γ or IL-2 release at T1 with S-specific
   IL-2 at T0 (ρ=0.62-0.77, p<0.001), a cytokine known to be abundantly
   produced by T[SCM] cells ([98]27, [99]28) ([100] Supplementary Figure
   S5A ).

Figure 2.

   [101]Figure 2
   [102]Open in a new tab

   Impact of antigen-specific Th cells at T0 on key determinants of
   anti-spike immunity at T1. (A) Spearman correlation of pre-existing
   (T0) spike (S)-specific T helper (Th) cells (expressing CD154, IFN-γ,
   and/or CD69) and their respective memory subsets (T[E]/T[EMRA],
   CD45RO^-CCR7^-; T[EM], CD45RO^+CCR7^-; T[CM], CD45RO^+CCR7^+;
   T[N]/T[SCM], CD45RO^-CCR7^+) with T1 S IgG, S-induced IL-2 release and
   S-induced IFN-γ release. (B) Determination of bottom-10 (B10) and
   top-10 (T10) responders based on T1 serum S IgG levels (red/green),
   S-induced IL-2 secretion (purple/gold), and S-induced IFN-γ secretion
   (blue/orange). Median values are indicated by black bars. (C) T0
   frequencies of double- and triple-positive IFN-γ^+ S-specific Th cells
   among Th lymphocytes depending on response classification at T1. (D)
   Phenotypic characterization of IFN-γ^+ double- and triple-positive
   S-specific Th cells at T0 depending on response classification at T1:
   T[E]/T[EMRA] (blue), T[EM] (red), T[CM] (grey), T[N]/T[SCM] (yellow).
   (E) Frequencies of pre-existing IFN-γ^+ S-specific double- and
   triple-positive T[E]/T[EMRA] among Th cells in B10 and T10 S IgG
   producers at T1. (F) Frequencies of pre-existing IFN-γ^+ S-specific
   double- and triple-positive T[N]/T[SCM] among Th cells in B10 and T10
   IL-2 responders to S stimulation at T1. (C, E, F) Individual and median
   values (columns) are shown. Mann-Whitney-U test. • p<0.05, false
   discovery rate (FDR)>0.2; * p<0.05, FDR<0.2; ** p<0.01, FDR<0.2; ***
   p<0.001, FDR<0.2. B10, T10, bottom 10, top 10 responder; FDR, false
   discovery rate; S, spike; T0, T1, T6, 0, 1, or 6 months post booster;
   Th, T helper cells; T[E]/T[EMRA], effector T cells and effector memory
   T cells re-expressing CD45RA; T[EM], effector memory T cells; T[CM],
   central memory T cells; T[N]/T[SCM], naïve T cells and stem cell memory
   T cells.

   Although S IgG serum levels at T1 did not correlate with absolute
   pre-booster frequencies of S-specific T cells, we found a modest
   positive correlation of S IgG levels at T1 with differentiated
   (T[E]/T[EMRA]) IFN-γ^+ S-specific T cells and a higher proportion of
   T[E]/T[EMRA] or T[EM] IFN-γ^+ S-specific T cells at T0 ([103] Figure 2A
   ). Moreover, S IgG levels at T1 showed a significant positive
   correlation (ρ=0.39-0.54) with S-induced release of various Th1, Th17,
   and associated innate effector cytokines at T0 ([104] Supplementary
   Figure S5A ). Collectively, these results indicate significant
   contributions of antigen-specific Th cells present pre-booster,
   especially IFN-γ^+ (Th1) cells, to antibody and cellular responses
   after booster vaccination. Moreover, these findings suggest that
   distinct baseline phenotypes of pre-existing antigen-specific Th cells
   are associated with antibody (T[E]/T[EMRA]) versus T-cellular cytokine
   responses (T[N]/T[SCM]).

   To corroborate these results, we used a previously described ([105]8)
   “top/bottom-10” approach to classify antibody and T cellular responses
   at T1 ([106] Figure 2B ). Based on this classification, pre-existing
   S-specific IFN-γ^+ double- and triple-positive Th frequencies were
   significantly associated with S-induced IL-2 release at T1
   (p=0.018-0.043). Likewise, triple-positive (IFN-γ^+CD69^+CD154^+)
   S-specific T cell frequencies at T0 were significantly associated with
   S-induced IFN-γ release at T1 (p=0.035, [107]Figure 2C ). Although
   top-10 and bottom-10 S IgG responders showed similar frequencies of
   total double- and triple-positive IFN-γ producing Th cells ([108]
   Figure 2C ), top-10 S IgG responders had increased proportions of
   pre-existing S-specific double- and triple-positive IFN-γ producing
   T[EM] and T[E]/T[EMRA] cells at T0 ([109] Figure 2D ), consistent with
   the correlation analysis shown above. In fact, only one of the
   bottom-10 S IgG responders at T1 had measurable quantities of these
   cells at T0, whereas most top-10 responders had S-specific IFN-γ
   producing T[EM] and T[E]/T[EMRA] cells co-expressing at least one of
   the activation markers CD69 and/or CD154 (p=0.049, [110]Figure 2E ).
   Vice versa, compared to bottom-10 responders, top IL-2 responders at T1
   had more IFN-γ^+ S-specific double- and triple-positive T[N]/T[SCM]
   cells at T0, especially IFN-γ^+ CD69^+ cells (p=0.033, [111]Figure 2F
   ). Additionally, the top/bottom-10 classification also corroborated
   significantly greater S-induced production of IL-2 (p=0.019) and IL17-A
   (p=0.014) at T0 in top-10 T1 spike IgG producers, with similar yet
   non-significant trends for IFN-γ (p=0.061), MCP1 (p=0.073), and TNF-α
   (p=0.052) ([112] Supplementary Figure S5B ).

   Altogether, these analyses suggest that S-reactive double- and
   triple-positive IFN-γ^+ Th cells present before booster vaccination are
   strongly associated with short-term serological and T-cellular
   post-booster responses, with disparate baseline phenotypes favoring
   stronger T-cell versus serological responses. Specifically, avid
   T-cellular cytokine responses seem to rely on pre-existing double- and
   triple-positive IFN-γ T cells, including less-differentiated
   populations (T[N]/T[SCM]), while S-specific T[EM] and T[E]/T[EMRA]
   cells present pre-booster contributed to robust S-specific IgG levels.

3.3. Cytokine production of S-specific Th cells at 1 month post-booster
determines sustained reactivity of the cellular and humoral immune response
at 6 months post-booster

   Given the marked inter-individual variation in long-term serological
   and T cell responses ([113] Figures 1A, B ), we next sought to
   elucidate whether antigen-specific Th cells present pre-booster (T0)
   and/or the early booster-induced response determine S-specific adaptive
   immunity at T6. S IgG levels and S-induced release of IL-2 and IFN-γ at
   T1 were found to moderately correlate with their respective levels at
   T6 (ρ=0.43-0.58, [114]Figure 3A ), suggesting partially overlapping
   immunological determinants of short- and long-term post-booster
   responses.

Figure 3.

   [115]Figure 3
   [116]Open in a new tab

   Pre- and post-booster predictors of sustained antibody and cytokine
   responses at T6. (A) Spearman correlation of T1 (peak) and T6 (long
   term) spike (S)-specific IgG levels, S-induced IL-2 release, and
   S-induced IFN-γ release. Bottom-half (B) and top-half (T) long-term
   responses are color-coded in red/green for IgG levels, purple/gold for
   S-induced IL-2, and blue/orange for S-induced IFN-γ based on T6 values.
   (B) Spearman correlation of S-specific Th cells (expressing CD154,
   IFN-γ, and/or CD69) and their respective effective/memory subsets
   (T[E]/T[EMRA], CD45RO^-CCR7^-; T[EM], CD45RO^+CCR7^-; T[CM],
   CD45RO^+CCR7^+; T[N]/T[SCM], CD45RO^-CCR7^+) at T1 with measurements of
   S IgG, S-induced IL-2, and S-induced IFN-γ at T6. (C) Frequencies of
   pre-existing (T0) IFN-γ^+ S-specific Th cells of B (red) and T (green)
   S IgG producers at T6 and characterization based on the expression of
   one or more activation markers (CD154, IFN-γ, and/or CD69). (D)
   Spearman correlation of S-induced cytokine secretion levels at T0
   (upper panels) and T1 (lower panel), respectively, with S IgG,
   S-induced IL-2, and S-induced IFN-γ at T6. (E) S-induced IL-2 and IL-4
   secretion at T1 of bottom- (red) and top-half (green) long term S IgG
   producers at T6. (F) S-induced IP-10 and MCP1 secretion at T1 of
   bottom- (purple) and top-half (gold) long term IL-2 responders to S
   stimulation at T6. (G) S-induced IFN-γ and IP-10 secretion at T1 of
   bottom- (blue) and top-half (orange) long-term IFN-γ responders to S
   stimulation at T6. (C, E-G) Individual and median values (columns) are
   shown. Mann-Whitney-U test. • p<0.05, false discovery rate (FDR)>0.2; *
   p<0.05, FDR<0.2; ** p<0.01, FDR<0.2; *** p<0.001, FDR<0.2. B, T,
   bottom, top half long term responder; FDR, false discovery rate; S,
   spike; T0, T1, T6, 0, 1, or 6 months post booster; Th, T helper cells;
   T[E]/T[EMRA], effector T cells and effector memory T cells
   re-expressing CD45RA; T[EM], effector memory T cells; T[CM], central
   memory T cells; T[N]/T[SCM], naïve T cells and stem cell memory T
   cells.

   Double- and triple-positive IFN-γ^+ S-specific Th cells at T0 strongly
   correlated with both S IgG levels and S-induced release of IL-2 and
   IFN-γ at T6 (ρ=0.54-0.64), whereas no consistent association with
   specific effector/memory Th phenotypes was seen ([117] Figure 3B ).
   Accordingly, the total pre-booster frequencies of S-specific T cells,
   the proportion of IFN-γ^+ S-specific T cells, the proportion of double-
   and triple-positive S-specific T cells, and frequencies of IFN-γ^+
   S-specific T cells co-expressing either CD69 or CD154 were
   significantly higher in the top half of serological responders at T6
   than in bottom-half responders (p=0.010-0.037, [118]Figure 3C ).

   In sharp contrast, not S-specific T-cell frequencies ([119]
   Supplementary Figure S6 ) but Th cytokine release and associated innate
   effector cytokines at T1 ([120] Figure 3D ) were significantly
   associated with S IgG levels and S-induced release of IL-2 and IFN-γ at
   T6. Specifically, S-induced release of IL-2, the Th1 signature cytokine
   IFN-γ, the IFN-γ-induced effectors IP-10 and MCP1, and the Th2
   signature cytokine IL-4 at T1 were significantly associated with at
   least one of the three studied surrogates of long-term (T6) responses
   ([121] Figures 3D-G ).

   Altogether, these data support the following model: Short-term antibody
   responses are mainly determined by pre-existing IFN-γ^+ S-specific Th
   cells. Early post-booster T cell activation, Th cytokine release, and
   associated cytokine responses are determined by total double- and
   triple-positive S-specific Th-cell frequencies, including less
   differentiated phenotypes (T[N]/T[SCM]). Early post-booster T-cell
   activation and cytokine release, in turn, promotes long-term S-specific
   antibody and T cellular immunity.

3.4. Transcriptomic analysis reveals upregulation of both adaptive and innate
immunity pathways in top responders

   To obtain a more detailed picture of the early post-booster immune
   landscape associated with long-term serological and T cellular
   responses, we performed n-Counter-based transcriptomics and pathway
   enrichment analysis on S-stimulated whole blood at T1 ([122] Figure 4A
   ) and compared individuals classified as T1 and T6 triple (S IgG, IL-2,
   and IFN-γ) top versus bottom responders ([123] Figure 4B ). To
   functionally validate transcriptional analyses from stimulated whole
   blood, we first compared T1 pathway enrichment based on T1 response
   classification. Indeed, we found strong and highly significant positive
   correlations between T1 cytokine responses and expression of a broad
   array of adaptive and innate effector pathways, whereas exhaustion
   pathways (CTLA-4 and PD-1) were suppressed in triple top responders
   ([124] Supplementary Figure S7 ).

Figure 4.

   [125]Figure 4
   [126]Open in a new tab

   Transcriptional networks associated with long-term post-booster
   responses. (A) Screening algorithm used to identify triple top- and
   bottom-responders at T6 based on spike (S)-specific IgG, S-induced IL-2
   release, and S-induced IFN-γ release. (B) Pathways with significantly
   stronger expression at T1 in T6 triple-top responders. Pathways with
   absolute z-scores ≥1.5 and Benjamini-Hochberg-adjusted (BH-adj.)
   p-values <0.05, i.e., -log[10](BH-adj. p) ≥1.3 are displayed. No
   pathways suppressed in triple-top responders reached the z ≤-1.5
   threshold. Activated pathways with absolute z-scores ≥2 are considered
   biologically significant (horizontal line). (C) (Simplified) gene
   expression network and predicted interactions, as determined by
   Ingenuity Pathway Analysis (core analysis, graphical summary). Orange
   interaction nodes indicate stronger induction/activation at T1 in T6
   triple-top responders. BH-adj., Benjamini-Hochberg-adjusted; S, spike;
   T1, T6, 1 or 6 months post booster; Th, T helper cells.

   When subsequently comparing transcriptional pathway enrichment at T1
   based on their response classification at T6, individuals with high
   levels of S IgG, S-induced IL-2, and IFN-γ at T6 displayed enrichment
   of several pathways associated with strong adaptive immune activation
   (e.g., IL-2 signaling, OX40 signaling) at T1 ([127] Figure 4C ).
   Interestingly, pathway analysis at T1 also revealed that several
   pathways and genes involved in pattern recognition receptor signaling,
   innate immune cell activation, antigen presentation, and inter-cellular
   crosstalk were significantly stronger activated in T6 IgG, IL-2, and
   IFN-γ top responders than in bottom responders ([128] Figure 4C ). In
   summary, these findings support a mechanistic link between short- and
   long-term immune protection and suggest that feedback loops between
   pre-existing and/or newly formed S-specific T cells and
   antigen-presenting cells (APCs) might contribute to both short- and
   long-term adaptive immune responses.

3.5. S-specific Th-cell present at the time of booster vaccination and
surrogates of early post-booster responses predict sustained production of
spike-specific IgG, IL-2, and IFN-γ

   Given our prior observations that both antigen-specific Th-cell
   frequencies pre-booster and early post-booster responses were
   associated with surrogates of adaptive long-term immune responses, we
   next analyzed the predictive performance of such markers for
   stratification of individual long-term (T6) responses. To that end, we
   tested various combinations of markers that significantly correlated
   with S IgG and S-specific secretion of IL-2 and IFN-γ at T6 ([129]
   Figures 3E, F ) and performed SVM analyses.

   Here, a combination of pre-existing S-specific Th cells or,
   specifically, IFN-γ^+ double- and triple-positive Th cells at T0 and S
   IgG, S-induced release of IFN-γ, IL-2, IL-4, IP-10, and MCP-1 at T1
   predicted top- versus bottom-half classification of S IgG and
   S-specific secretion of IL-2 and IFN-γ with high precision and recall
   performance (≥0.80 and 1.00, respectively), leading to F1 scores of
   ≥0.89. Further, the above combination of pre-booster and early
   post-booster markers showed 100% precision, recall, and accuracy for
   identification of triple (i.e., S IgG and S-induced IFN-γ and IL-2) top
   and bottom responders at T6 ([130] Table 1A ). Remarkably, inclusion of
   T1 S IgG as an immune variable consistently led to inferior predictive
   F1 scores ([131] Supplementary Figure S8 ). Altogether, these findings
   underscore the importance of functional T-cell analyses in the
   monitoring of adaptive immunity after SARS-CoV-2 booster vaccinations
   and corroborate a model whereby S-specific Th cells detectable
   pre-booster and the magnitude and quality of early post-booster Th-cell
   jointly determine long-term adaptive immune protection.

Table 1.

   Support vector machine analysis of T0 and T1 immune markers to predict
   sustained spike-specific (S) IgG and S-induced cytokine response at T6.
   T6 outcome Combination yielding highest F1 score F1 score Precision
   Recall Accuracy
   (A) Including flow cytometric readouts
   S IgG T1 IL-2 0.89 0.80 1.00 0.88
   T1 IP-10
   S-specific IL-2 T0 S-specific Th cells 0.94 0.89 1.00 0.94
   T1 IL-4
   T1 IP-10
   T1 MCP-1
   S-specific IFN-γ T0 IFN-γ^+ MF Th cells 1.00 1.00 1.00 1.00
   T1 IFN-γ
   T1 IL-2
   T1 IP-10
   Triple classification T0/1 8 possible combinations of 2 markers each
   [132]^a 1.00 1.00 1.00 1.00
   (B) Cytokine release only
   S IgG T1 IL-2 0.89 0.80 1.00 0.88
   T1 IP-10
   S-specific IL-2 T1 MCP-1 0.80 0.86 0.75 0.81
   S-specific IFN-γ T1 IFN-γ 0.94 0.89 1.00 0.94
   T1 IP-10
   Triple classification T1 3 possible combinations of 2 markers each
   [133]^b 1.00 1.00 1.00 1.00
   [134]Open in a new tab
   ^a

   (i) T0 IFN-γ^+ MF Th cells + T1 IFN-γ, (ii) T0 IFN-γ^+ MF Th cells + T1
   IL-2, (iii) T0 IFN-γ^+ MF Th cells + IL-4, (iv) T0 IFN-γ^+ MF Th cells
   + T1 MCP-1, (v) T0 S-specific Th cells + T1 IP-10, (vi) T1 IFN-γ + T1
   IP-10, (vii) T1 IFN-γ + T1 MCP-1, or (viii) T1 IP-10 + T1 MCP-1.
   ^b

   (i) IFN-γ + IP-10; (ii) IFN-γ + MCP-1, and (iii) IP-10 + MCP-1.

   MF, multifunctional (i.e., double-/triple-positive); S, spike; T0, T1,
   T6, timepoints 0, 1, or 6 months post booster-vaccination; Th, T helper
   cells.

   However, due to the limited feasibility of flow cytometric analysis for
   clinical routine use ([135]29–[136]33), and to avoid reliance on repeat
   sampling at T0 and T1, we compiled additional SVM models relying
   exclusively on S-induced cytokine release at T1 ([137] Supplementary
   Figure S9 , [138]Table 1B ). Although some of the resulting F1 scores
   were slightly lower than in the initial model including flow cytometry
   at T0, classification of long-term responses based on S-induced
   cytokine release at T1 was feasible with high F1 scores of 0.80-1.00
   ([139] Supplementary Figure S9 , [140]Table 1B ). In particular,
   cytokine release assays at T1 facilitated correct prediction of all
   triple top and bottom responders at T6.

4. Discussion

   Several groups have characterized the cellular adaptive immune response
   to primary SARS-CoV-2 vaccination, whereas the cellular determinants of
   sustained immune protection after booster vaccinations are incompletely
   understood. Herein, we identified several components of
   antigen-specific Th-cell responses pre- and post-booster that are
   associated with strong and sustained serological and cytokine responses
   after booster vaccination. Specifically, peak S-specific IgG correlated
   with pre-existing IFN-γ^+ T[E]/T[EMRA] cells while double- and
   triple-positive IFN-γ^+ Th cells present before booster vaccination,
   including T[N]/T[SCM] subsets, were associated with stronger cytokine
   secretions in response to booster vaccination (especially IL-2 and
   IFN-γ). These cytokine responses in turn determined the long-term
   production of S IgG over the 6-month observation period. On the
   transcriptional level, IL-2, OX40, and other genes associated with
   T-cell proliferation, were indicative of sustained S-reactive cytokine
   release and stable S-specific IgG levels.

   This underlines the cellular immune response’s essential role as a
   facilitator and predictor of antibody production and long-term adaptive
   immune protection after vaccinations. Specifically, interaction of
   T-cellular CD154 with its receptor CD40 on B cells is essential for
   germinal center formation, antibody class switching, and affinity
   maturation ([141]34). Although we found that IFN-γ-producing effector
   and effector memory Th cells correlate with development of IgG
   responses, we cannot precisely pinpoint the causative T-cell type. The
   most obvious candidates would be T-follicular helper cells, which
   promote development of high-affinity IgG in B-cell follicle germinal
   centers ([142]35). Previous studies showed concomitant development and
   comparable peak frequencies of circulating T-follicular helper and
   IFN-γ-producing Th cells following SARS-CoV-2 vaccination ([143]36).
   Surprisingly, these T-cell populations both correlate with development
   of specific IgG responses ([144]36). While it is counter-intuitive that
   IFN-γ-producing T-cells, in particular Th1 cells, might support IgG
   development, previous data indicate that type-1 T-follicular helper
   cells, which transiently produce IFN-γ, induce T-bet expression in
   B-cells which is required for development of antiviral IgG isotypes
   ([145]37, [146]38). Alternatively, the association between IFN-γ and
   IgG production might reflect the concomitant development of
   T-follicular helper cells and IFN-γ-producing Th cells. Further studies
   using more defined Th-cell markers are needed to fully elucidate which
   cell types support the development of SARS-CoV-2-specific IgG, and the
   underlying mechanisms.

   Furthermore, it was previously shown that S-specific CD4^+ T[EM] cells
   one month after the 2^nd dose of primary SARS-CoV-2 mRNA vaccination
   determined sustained responsiveness of S-specific T cell responses 6
   months post-booster ([147]39). Although not considered the driver of
   responses to primary SARS-CoV-2 vaccination, there has been evidence of
   pleiotropic effects of pre-existing cellular immunity on response to
   both primary and booster vaccinations. For instance, pre-existing
   memory B cells were associated with increased S-specific and
   neutralizing antibody levels ([148]39). As discussed in the
   Introduction, pre-existing SARS-CoV-2-crossreactive T cells and their
   age-dependent phenotypes influenced the quality, avidity, and diversity
   of the developing adaptive responses to SARS-CoV-2 after the primary
   vaccination series ([149]8). Herein, we found that responses to
   subsequent booster vaccinations tend to benefit from the S-specific
   (effector) memory cell pool developed in response to the initial
   vaccination.

   Interestingly, our findings further suggest that feedback loops of
   antigen-specific T cells and APCs are critical for enhanced adaptive
   immune responses after booster vaccination. This is consistent with
   prior evidence suggesting that upregulated IFN-α and IFN-γ receptor
   genes, natural killer cell signaling pathways, and production of
   reactive oxygen species and nitric oxide in macrophages at day 3 after
   flu vaccination predicted increased inhibitory antibody levels against
   influenza hemagglutinin at day 28 ([150]40).

   Our findings might have translational implications. Given the natural
   inter-individual heterogeneity and variability of responses to mRNA
   vaccination, even in healthy and seemingly immunocompetent individuals
   ([151]13, [152]15, [153]16), simple functional immunoassays might allow
   for individualized prediction of the quality or longevity of immune
   responses and guide recommendations regarding additional booster doses.
   By monitoring pre-existing spike-specific Th cells and early
   post-booster Th responses with simple whole blood-based assays, we
   obtained a robust prediction of the vaccinees’ anti-spike immune
   response at 6-months post-booster. Notably, in the SVM model presented
   here, evaluation based on T cellular and cytokine secretion data
   exceeded the predictive performance of early serological responses
   (i.e., spike IgG) alone.

   This study has several limitations. Due to its small scale, we could
   not assess protection against severe or symptomatic SARS-CoV-2 BTIs.
   Nine of our 30 subjects developed anti-nucleocapsid antibodies
   indicative of BTIs but our study was not sufficiently powered to
   determine associations of the identified immune surrogates and BTI
   status. However, to ensure that we specifically analyze responses to
   the first COVID-19 booster vaccination after prior primary (2-dose)
   BNT162b2 vaccination, we excluded subjects developing anti-nucleocapsid
   antibodies indicative of a SARS-CoV-2 infection from subsequent
   immunoassay timepoints. These included subjects missing one or more
   appointments, as we could not rule out an intermittent infection due to
   the short half-life of nucleocapsid antibodies. This finally resulted
   in only 16 subjects available for the 6-month assessment, limiting
   statistical power. Additionally, we recruited our subjects from a
   limited pool of healthy and mostly middle-aged healthcare workers.
   Although some associations with the subjects’ age were seen, this study
   was not designed or powered to investigate nuanced age-dependent
   differences in T-cell biology and transcriptional profiles, as
   previously suggested ([154]8, [155]41). Furthermore, we did not
   evaluate T-cell reactivity to mutated spike peptides of any variant of
   concern, where reduced vaccine efficacy might be expected ([156]39,
   [157]42–[158]44). Lastly, given possible underreporting of asymptomatic
   or mild infections and unreliability of N antibodies to trace
   infections in the distant past, we could not assess the role of prior
   COVID-19 infection on immune status and response to booster
   vaccination.

   Despite these limitations, we identified T-cellular activation and
   cytokine markers predictive of short- and long-term response to mRNA
   booster immunization against SARS-CoV-2. We found that pre-existing
   S-specific Th cell populations and their respective cytokine profiles
   determine the quality and longevity of adaptive immune responses to
   booster vaccinations, underscoring the influence of pre-existing
   antigen-specific CD4^+ T cells on the robustness of adaptive humoral
   responses. Given the rapid evolution of vaccine technologies,
   implementation of simple and facile whole blood-based T-cellular
   cytokine release assays in early stages of future vaccine trials could
   help to preemptively identify potential low responders. As our
   immunoassay platform could be adapted to any protein antigen, expansion
   of this technology to other vaccine-preventable diseases or even
   post-exposure prophylaxis would be easily feasible.

Acknowledgments