Abstract

Introduction

   Higher frequencies of mucosal-associated invariant T (MAIT) cells were
   associated with an increased adaptive response to mRNA BNT162b2
   SARS-CoV-2 vaccine, however, the mechanistic insights into this
   relationship are unknown. In the present study, we hypothesized that
   the TNF response of MAIT cells supports B cell activation following
   SARS-CoV-2 immunization.

Methods

   To investigate the effects of repeated SARS-CoV-2 vaccinations on the
   peripheral blood mononuclear cells (PBMCs), we performed a longitudinal
   single cell (sc)RNA-seq and scTCR-seq analysis of SARS-CoV-2 vaccinated
   healthy adults with two doses of the Pfizer-BioNTech BNT162b2 mRNA
   vaccine. Collection of PBMCs was performed 1 day before, 3 and 17 days
   after prime vaccination, and 3 days and 3 months following vaccine
   boost. Based on scRNA/TCR-seq data related to regulatory signals
   induced by the vaccine, we used computational approaches for the
   functional pathway enrichment analysis (Reactome), dynamics of the
   effector cell-polarization (RNA Velocity and CellRank), and cell-cell
   communication (NicheNet).

Results

   We identified MAIT cells as an important source of TNF across
   circulating lymphocytes in response to repeated SARS-CoV-2 BNT162b2
   vaccination. The TNF^high signature of MAIT cells was induced by the
   second administration of the vaccine. Notably, the increased TNF
   expression was associated with MAIT cell proliferation and efficient
   anti-SARS-CoV-2 antibody production. Finally, by decoding the
   ligand-receptor interactions and incorporating intracellular signaling,
   we predicted TNF^high MAIT cell interplay with different B cell
   subsets. In specific, predicted TNF-mediated activation was selectively
   directed to conventional switched memory B cells, which are deputed to
   high-affinity long-term memory.

Discussion

   Overall, our results indicate that SARS-CoV-2 BNT162b2 vaccination
   influences MAIT cell frequencies and their transcriptional effector
   profile with the potential to promote B cell activation. This research
   also provides a blueprint for the promising use of MAIT cells as
   cellular adjuvants in mRNA-based vaccines.

   Keywords: SARS-CoV-2, mRNA vaccine, MAIT cells, immune response, TNF, B
   cells, single-cell RNA/TCR-sequencing

1. Introduction

   Mucosal-associated invariant T (MAIT) cells are unconventional T cells
   defined by their semi-invariant αβ T cell receptor (TCR) composed of
   TCR α-chain Vα7.2 (TRAV1-2) joined to Jα33 (TRAJ33) and paired in
   humans with TCR β-chain V[β]2 (TRBV20) and V[β]13 (TRBV6)
   ([65]1–[66]3). The TCR of MAIT cells is restricted to the
   non-polymorphic major histocompatibility complex (MHC) class I-like
   protein MR1 that recognizes bacterial- and yeast-derived riboflavin
   metabolites ([67]4, [68]5). MAIT cells are abundant in peripheral
   tissues including the liver and the gut ([69]6, [70]7), and represent
   up to 10% of all peripheral blood (PB) circulating T lymphocytes
   ([71]8). MAIT cells provide a rapid, innate-like effector response to
   several microbial infections, which leads to migration, proliferative
   expansion, and cytotoxic activity with the release of effector
   molecules such as granzymes, perforin, and cytokine secretion such as
   tumor necrosis factor (TNF), interferon (IFN)-γ, and interleukin
   (IL)-17 ([72]2, [73]9–[74]11). In addition, MAIT cells can be activated
   in an MR1-independent manner via toll-like receptor (TLR) ligands (DNA,
   RNA, and LPS), IFN-α/β and IL-12, IL-18, and IL-15 ([75]10,
   [76]12–[77]14). Cytokine-dependent activation of MAIT cells has been
   shown in a range of viral infections including Dengue fever (DENV),
   influenza, hepatitis C virus (HCV), human immunodeficiency virus (HIV),
   Hantavirus, and, recently, SARS-CoV-2 ([78]15–[79]20).

   MAIT cells have been shown to enhance B cell functionality. Indeed,
   experimental approaches in vitro provided evidence that MAIT cells are
   capable of promoting the differentiation of memory B cells into
   plasmablasts and favouring antibodies (Ab) production ([80]21, [81]22).
   Moreover, both in vivo infection (Vibrio cholerae, Simian
   immunodeficiency virus (SIV)) and vaccination (Shigella dysenteriae,
   SIV) increase MAIT cell frequencies and their cytokine secretion that
   positively correlate with B cell activation, specific Ab production,
   class switching, and memory establishment ([82]23–[83]25). On the other
   hand, the activation status of MAIT cells correlates with auto-Abs
   production in different autoimmune diseases. Indeed, an auto-Ab
   reduction was observed upon MAIT cell inhibition in MR1-deficient mice
   or upon MR1-inhibitory ligand binding ([84]22, [85]26, [86]27).

   Recently, higher frequencies of MAIT cells were associated with an
   increased adaptive response to mRNA BNT162b2 SARS-CoV-2 vaccine
   ([87]28). However, the molecular patterns and dynamics of this response
   are unknown. In the present study, we conducted a comprehensive,
   longitudinal single-cell RNA-sequencing (scRNA-seq) and scTCR-seq
   analysis of MAIT cell activation tracked in the PB of healthy adults
   who received two doses of BNT162b2 vaccination. We observed that, among
   PB lymphocytes, TRAV1-2^+ KLRB1^+ MAIT cells represent an important
   source of TNF after the booster vaccine. This increased expression of
   TNF in MAIT cells correlates with their activation and polyclonal
   proliferation. Finally, we provide a map of predicted interactions
   between MAIT and B cells established upon vaccination, indicating a
   possible TNF-dependent downstream signaling in conventional switched
   memory B cells.

2. Materials and methods

2.1. Study design

   This study was designed to assess MAIT cell immune responses after
   immunization with two doses of Pfizer-BioNTech BNT162b2 mRNA vaccine in
   individuals without previous SARS-CoV-2 infection. We used scRNA-seq
   paired with scTCR-seq techniques to uncover the dynamics of MAIT cells
   upon repeated BNT162b2 vaccination. In specific, we performed a
   longitudinal study on six volunteers, collecting peripheral blood
   mononuclear cell (PBMCs) samples (n=28) 1 day before (P0), 3 and 17
   days after prime vaccination (P1 and P2, respectively), and 3 and 99
   days following vaccine-boost (P3 and P4, respectively).

2.2. PBMCs isolation

   Freshly PBMCs were isolated through Lympholyte^®-H Cell Separation
   density gradient solution (Cedarlane Laboratories, Burlington, North
   Carolina, USA) according to the manufacturer’s instructions and frozen
   in Fetal Bovine Serum (FBS, EuroClone) supplemented with 10% of the
   cryoprotective dimethyl sulfoxide (DMSO, PanReac AppliChem) as
   previously published ([88]29).

2.3. Anti-SARS-CoV-2 IgG Ab titration

   Anti-SARS-CoV-2 IgG Ab titration was performed at the Humanitas
   Research Hospital by a ready-to-use ELISA (enzyme-linked immunosorbent
   assay) kit (Ref. no. COV19G.CE) for diagnostic use (DIA.PRO; Diagnostic
   Bioprobes Srl, Italy) and following the manufacturer’s procedures. The
   IgG Ab levels were measured by the optical density (OD) of 450/620-630
   nm. The cutoff OD was evaluated by the formula: cutoff OD= negative
   control (NC) + 0.250. The specific concentration of IgG was evaluated
   by the kit-provided standards.

2.4. Single-cell RNA and TCR library preparation

   Libraries for scRNA-seq were prepared using the Chromium Single Cell
   Platform with a Single Cell 5′ Library and Gel Bead Kit (10X Genomics,
   1000006). Thawed PBMC samples were evaluated for viability prior to
   scRNA-seq analysis, and all were ≥95% viable. Cells were resuspended in
   a volume equivalent to 10,000 target cells for each sample and were
   individually loaded onto a Chromium single-cell controller (10X
   Genomics) to generate single-cell gel beads-in-emulsion (GEMs).
   Captured cells were then lysed and the released RNA was barcoded
   through reverse transcription in individual GEMs. Complementary DNAs
   (cDNA) were generated and split to generate additional libraries of αβ
   TCR amplicons. Complementary DNAs were amplified, and the quality was
   assessed using an Agilent 4200 Tapestation.

   The scRNA and scTCR libraries were sequenced using an Illumina
   Novaseq6000 sequencer with a paired-end 150-bp (PE150) reading strategy
   (performed by CapitalBio Technology).

2.5. Processing scRNA-seq data

   The scRNA-seq reads were aligned to the GRCh38 (version
   refdata-gex-GRCh38-2020-A, 10X Genomics) human genome reference, and
   UMIs were quantified using Cellranger 5.0.0 (10X Genomics). Subsequent
   analyses were performed using the Python package Scanpy v1.8.1
   ([89]30), under Python v3.8 if not stated otherwise. Raw data matrices
   of all samples were merged and cells with fewer than 600 expressed
   genes, or 1,200 UMIs, greater than 8% mitochondrial genes were removed.
   Data were then log-normalized with a scale factor of 10,000. Highly
   variable genes (HVGs, 2,000) were identified using the Seurat
   dispersion-based methods. Data from each sample (n=28) were integrated
   by the Scanpy package. In brief, principal component analysis (PCA) was
   performed using HVGs via ARPACK implementation of singular value
   decomposition (SVD). Principal components (PCs) were used to integrate
   data from different samples by the Harmony algorithm using the
   Harmonypy package v0.0.6 ([90]31). Neighbors were identified using the
   top 50 components of Harmony-corrected PCA embeddings, and the
   clustering analysis was performed using the Leiden algorithm ([91]32)
   with an initial resolution of 0.1. The clusters were then embedded in
   two dimensions by using Uniform Manifold Approximation and Projction
   (UMAP) algorithm. ([92]33).

2.6. Cluster marker identification and cell-type annotation

   Differential expression gene (DEG) analysis between clusters was
   carried out to find markers for each of the identified clusters by
   using the Wilcoxon rank-sum test (Scanpy). Genes with an adjusted
   p-value < 0.05 and expressed by at least 10% of cells in the cluster at
   a minimum Log[2]-FoldChange of |0.25| were considered significant. A
   total of 147,160 CD3^+ cells were identified based on the expression of
   canonical cell-type markers and subjected to a next round of
   dimensionality reduction, and unsupervised clustering as described
   above. From CD3^+ T cells, a total of 6,290 MAIT cells embedded in
   cluster 9 were profiled and annotated based on the expression of
   canonical cell-type markers, and further re-clusterized. For B cells, a
   total of 21,013 cells embedded in clusters 7, 11, 33, and 32 were
   profiled and annotated based on the expression of canonical B cell
   markers, and further re-clusterized.

2.7. Reactome pathway analysis

   Pathway enrichment analysis was performed using the Reactome web tool
   ([93]https://reactome.org/) ([94]34). The Reactome pathways of cell
   type were enriched using DEGs with adjusted p-value < 0.05 and
   Log[2]-FoldChange > |0.25|. Only enriched terms with FDR < 0.05 were
   selected as significant and visualized by R package ggplot2 under R
   v4.2.1. Dots are colored by FDR values and sized by the number of DEGs
   enriched in each pathway.

2.8. RNA velocity analysis

   RNA velocity analysis ([95]35) was performed using the Python package
   scVelo v0.2.4 to infer the future states of a cell from unspliced and
   spliced mRNA counts. The matrices of spliced and unspliced counts were
   quantified using the Python library velocyto v0.17 from the aligned bam
   file generated by CellRanger. Counts matrices in loom file format were
   loaded to scVelo and merged into the anndata object. Genes detected in
   less than 20 unspliced and spliced counts were filtered out, and the
   counts were then normalized. The filter_genes_dispersion() function was
   used to identify the top 2,000 HVGs. The scv.pp.moments() function was
   used to compute the first- and second-order moments among nearest
   neighbors in the PCA space. The velocities and velocity graphs were
   estimated using the generalized dynamical model. Velocities were
   visualized on UMAP embedding. Initial and terminal states were computed
   by using the Python package CellRank v1.5.0 ([96]36)
   (cr.tl.initial_states, cr.tl.terminal_states). The latent time was
   computed using the scVelo function scv.tl.recover_latent_time() by
   using the CellRank initial and terminal states.

2.9. scTCR-seq analysis

   The scTCR-seq reads were aligned to the GRCh38 (version
   refdata-gex-GRCh38-2020-A, 10X Genomics) human genome reference, and
   consensus αβ TCR annotations were performed using Cellranger VDJ tools
   3.1.0 (10X Genomics). Subsequent analyses were performed using the
   Python package Scirpy v0.10.1 ([97]8) under Python v3.8 if not stated
   otherwise. TCR annotations were merged with transcriptomics data using
   the scirpy.pp.merge_with_ir() function. TCR quality control was
   computed by using the ir.tl.chain_qc() function; only cells with
   TRAV1-2 TCR α chain paired with β chains and cells with TCR α chain
   TRAV1-2 paired with β chains plus an additional α and/or β chain were
   incorporated into the analysis. Paired invariant TRAV1-2 MAIT cells
   were detected in 54% of 6,290 MAIT cells. α and β chain combinations
   were visualized via chord diagrams using the circlize package v0.4.8.
   Clonotypes were defined according to the amino acid sequence identity
   of α and β CDR3 regions. Clonal cells were defined as clonotypes that
   appeared in at least two cells. The alpha diversity of clonotypes was
   calculated by computing the Normalized Shannon Entropy (Shannon index)
   ([98]1) within time points. α and β CDR3 sequence lengths were computed
   by using the ir.tl.spectratype() function.

2.10. NicheNet analysis

   NicheNet analysis was performed using the R package nichenetr v1.1.0
   ([99]37) to predict ligand-target links between interacting cells. We
   applied NicheNet analysis to investigate the cell-cell interactions
   (CCIs) between MAIT cells, which act as the sender population, and B
   cells, which act as the receiver population. Both for the sender
   population and the receiver one, we identified cells at P3. Only genes
   expressed in at least 10% of the sender or receiver population were
   considered. The gene sets of interest were identified as the DEGs in B
   cells between time points P3 and P0 with adjusted p-value < 0.05 and
   Log[2]-FoldChange > |0.25|. Ligand activities were predicted using the
   predict_ligand_activities() function using the Pearson correlation
   coefficient and ranked according to the Pearson correlation
   coefficient. Target genes were predicted by looking at the regulatory
   potential scores between ligand and target genes of interest, while the
   ligand-receptor network was predicted by looking at the prior
   interaction potential. The target genes and the receptor of the
   top-ranked ligands were visualized via chord diagrams using the
   circlize package v0.4.8.

2.11. DNA proliferation score

   The proliferation scores were computed on both total MAIT cells (c9,
   among all the CD3^+ lymphocytes) and on subcluster 1 (sc1) of the
   re-clustering of the MAIT cells. The genes implicated in the DNA
   replication pathway found in Reactome Pathway Analysis (H3F3B, H2AFX,
   DBF4, UBB, UBE2S, UBC, FOSB, and CCNL1), with the addition of MKI67,
   were used to calculate the score by the Scanpy-implemented function
   scanpy.tl.score_genes(). Null gene expression cells were then excluded
   from the analysis.

2.12. Statistics

   Statistical analyses of scRNA-seq data were performed using GraphPad
   PRISM software version 9.5.0 (La Jolla, California, USA). Statistical
   differences were calculated with ANOVA tests as specified in the
   legend. Statistically significant P values were represented with
   GraphPad style and summarized with the following number of asterisks
   (*): *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001.

2.13. Study approval

   The enrollable healthy individuals were those who planned to be
   vaccinated according to the national Italian COVID-19 vaccination
   program at the Humanitas Research Hospital from November 2020 to April
   2021. Recruitment of volunteers was performed according to the
   Declaration of Helsinki, and all the individuals signed written
   informed consent. The collection of healthy SARS-CoV-2 vaccinated
   subjects’ PB samples for research purposes was ethically approved by
   the Institutional Review Board (IRB) of Humanitas Research Hospital
   (HRH) (approval 738/20).

3. Results

3.1. MAIT cells exhibit substantially altered transcriptomes following
SARS-CoV-2 vaccination

   To investigate the effects of repeated SARS-CoV-2 vaccinations on the
   transcriptomic profile of the PB circulating MAIT cells, we performed a
   longitudinal scRNA-seq study on six volunteers vaccinated with two
   doses of the SARS-CoV-2 Pfizer-BioNTech BNT162b2 mRNA vaccine.
   Collection of PBMCs was performed 1 day before (P0), 3 and 17 days
   after prime vaccination (P1 and P2, respectively), and 3 and 99 days
   following vaccine boost (P3 and P4, respectively) ([100]Figure 1A).
   Among lymphocytes, MAIT cells were identified as those expressing CD3
   (CD3E), invariant T cell receptor (TCR) α chain TRAV1-2 region-encoding
   segment, and canonical markers CD8 (CD8A), CD161 (KLRB1), SLC4A10
   (encoding solute carrier family 4 member 10), and CD127 (IL7R)
   ([101]Figure 1B; [102]Supplemental Figure 1A). In total, we refined
   6,290 MAIT cells that UMAP analysis projected to cluster 9 (c9).

Figure 1.

   [103]Figure 1
   [104]Open in a new tab

   SARS-CoV-2 vaccination shapes the transcriptional pattern of MAIT
   cells. (A) Schematic overview of the experimental design. Collections
   of PBMCs were performed 1 day before (P0), 3 and 17 days after prime
   vaccination (P1 and P2, respectively), and 3 and 99 days following
   vaccine boost (P3 and P4, respectively). (B) UMAP clustering (left
   panel) projection of the integrated PB CD3^+ T cells from all subjects
   (s01-s06; a total of 147,160 CD3^+ cells). The dot plot (middle panel)
   and the feature plots (right panel) show the expression of canonical
   markers used for the annotation of MAIT cells. (C) The bar plot shows
   the number of identified DEGs across the different time points compared
   to P0. For all the figures, DEGs were defined as follows: (i) absolute
   value of average |log[2] FC| ≥ 0.25 for upregulated and downregulated
   genes; (ii) adjusted P values (adj. p) ≤ 0.05, and (iii) detected in at
   least 10% of cells (min. pct ≥ 10%). (D) Transcriptional pattern of
   MAIT cells at different time points in terms of DEGs. The heatmap shows
   the gene expression across different time points of the top 50 DEGs
   (rows) for time point P3 compared to P0. (E) The dot plot shows a
   selection of significantly enriched pathways with FDR values < 0.05,
   identified among DEGs at each time point for the MAIT cell cluster
   using the Reactome pathway browser. Dots are colored by FDR values and
   sized by the number of DEGs enriched in each pathway. (F) The scatter
   plot shows the median of the proliferation-modular score of MAIT cells
   calculated as AddModuleScore of nine genes associated with the
   proliferation signature (MKI67, H3F3B, H2AFX, DBF4, UBB, UBE2S, UBC,
   FOSB, and CCNL1). The unpaired parametric ANOVA test was used for the
   statistical analysis, statistically significant P value was represented
   as (*); ****P < 0.0001. (G) Kernel density for the TNF embedded on UMAP
   projection. (H) The ridge plot shows the TNF expression levels (x-axis,
   log-UMI) and the related cell frequency of cells (y-axis) across the
   different time points. Null gene expression cells were excluded from
   the analysis. The dotted line highlights the changes in TNF gene
   expression level across the different time points. (I) The circular bar
   plot (left panel) shows the frequency of TNF-expressing cells across
   all the CD3^+ clusters, where each bar is colored for the scaled TNF
   mean expression. The dot plot (right panel) shows the expression of TNF
   in both MAIT cell (c9) and γδ T cell clusters (c11) across all the time
   points.

   Examining the transcriptional pattern of MAIT cells at different time
   points in terms of differentially expressed genes (DEGs) using P0 as
   the baseline, we observed a higher number of modulated genes shortly
   after the vaccine boost ([105]Figure 1C). Indeed, MAIT cells at P3
   increased expression of early activating marker CD69 and several
   members of the transcription factor (TF) families including genes of
   AP-1 (i.e., JUN, JUNB, JUND, FOSB, and FOSL2), and NF-kB (i.e., NFKBIA,
   NFKBIZ, NFKB1, MAP3K8, and RELB) ([106]Figure 1D; [107]Supplemental
   Table). To further provide evidence of MAIT cell responsiveness to
   vaccines, we used pathway-based analysis instead of the
   ‘one-gene-at-a-time’ approach by using Reactome
   ([108]https://www.reactome.org) ([109]34). The vaccine boost, compared
   to the first immunization, provided a much wider response related to
   the modulation of pathways, including TNF and IFN-γ (encoded by IFNG)
   cytokines, IL-12 signaling, TCR-signaling, and pathways related to
   SARS-CoV-2 host interactions and activation of the cell cycle when
   compared to the first immunization ([110]Figure 1E). To further provide
   evidence of cycling MAIT cells, we analyzed the presence of gene
   proliferation signature score (MKI67, H3F3B, H2AFX, DBF4, UBB, UBE2S,
   UBC, FOSB, and CCNL1) showing a significant increase after the second
   dose of vaccine ([111]Figure 1F). We then assessed whether the modified
   cytokine-related pathways were associated with their up- or
   down-modulation. Longitudinal analysis of the cytokine-specific gene
   expression showed that, in immunized individuals, MAIT cells were
   distinguished for high expression levels of TNF ([112]Figure 1G).
   Importantly, this TNF^high signature at P3 was a consequence of a
   progressive increase in TNF expression after vaccination
   ([113]Figure 1H). These data implicate MAIT cells as an important
   source of TNF across circulating T lymphocytes after the booster
   vaccine ([114]Figure 1I). Expression of TNF in MAIT cells exceeded
   those of other cells such as gamma delta (γδ) T cells, in which, the
   high TNF expression was unrelated to the second dose of the vaccine,
   thus indicating that the vaccine-induced activation of MAIT cells
   differs from other innate-like lymphocytes, although it may not be
   exclusive. On the contrary, in MAIT cells we found very low expression
   levels of IFNG and IL-12 (IL12A) ([115]Supplemental Figure 1B). In
   addition, despite the known propensity of MAIT cells to produce IL-17
   (IL17A) ([116]11), no expression of this gene was detected.

3.2. Polyclonal MAIT cell proliferation following SARS-CoV-2 vaccination
converges on the effector TNF polarization

   Considering the heterogeneity of MAIT cells ([117]38, [118]39), we
   performed additional analysis. Notably, all MAIT cells detected in the
   c9 were subjected to further re-clustering analysis, which generated
   six sub-clusters (sc0-5) ([119]Figure 2A). We centered our analysis on
   TRAV1-2^+ cells embedded in sc0-4 that differ from each other in their
   transcriptomic profiles ([120]Supplemental Figure 2A). Regarding the
   kinetics and amplitude of DEGs, we found a heterogeneous response
   across sub-clusters with major changes detected in sc1
   ([121]Figure 2B). Indeed, refined Reactome pathway enrichment analysis
   performed for each MAIT sub-cluster confirmed sc1 as the most
   responsive sub-cluster after the second dose of the vaccine
   ([122]Supplemental Figure 2B). According to our previous observation,
   cells projected to sc1 showed a higher expression level of TNF, which
   increased after the second dose of the vaccine ([123]Figure 2C).
   Additionally, cells in sc1 showed an enhanced proliferation score over
   the pre-immunization level, which correlated with an increase in the
   sc1 relative frequency distribution across MAIT cells ([124]Figure 2D;
   [125]Supplemental Figure 2C). Consequently, we found significant
   DNA-replication-related pathways enriched in sc1 ([126]Supplemental
   Figure 2B).

Figure 2.

   [127]Figure 2
   [128]Open in a new tab

   Single-cell tracking of responsive MAIT cells following vaccination.
   (A) UMAP clustering projection (left panel) of the integrated PB MAIT
   cells from all subjects (s01-s06; a total of 6,290 MAIT cells). The
   violin plot (right panel) shows the expression of canonical markers
   used for the annotation of MAIT cells. (B) The bar plot shows, for each
   cluster, the number of identified DEGs across the different time points
   compared to P0. For all figures, DEGs were defined as follows: (i)
   absolute value of average |log[2] FC| ≥ 0.25 for upregulated and
   downregulated genes; (ii) adjusted P values (adj. p) ≤ 0.05, and (iii)
   detected in at least 10% of cells (min. pct ≥ 10%). (C) Kernel density
   (left panel) of the TNF embedded on UMAP projection. The box plot
   (right panel) shows the TNF mean expression across the different time
   points in sub-cluster 1 (sc1). The unpaired parametric ANOVA test was
   used for the statistical analysis, statistically significant P value
   was represented as (*); **P < 0.01. (D) The bar plot shows the
   frequency (%) of cluster distribution across all the time points. Cell
   numbers were normalized to the total loaded cell number per time point.

   We then wondered whether the vaccine-induced proliferation of TRAV1-2^+
   MAIT cells modifies their TCR repertoire. We linked our scRNA-seq data
   to the individual αβ TCR repertoire generated at single-cell
   resolution. Regardless of vaccination, TCR-encoding segment TRAV1-2 of
   analyzed MAIT cells paired with TRAJ33 and, at much lower frequencies,
   with TRAJ12 or TRAJ20. In addition, analyzed subjects showed
   oligoclonal TCRβ chain usage, mainly TRBV20 and TRBV6 ([129]Figure 3A;
   [130]Supplemental Figures 3A, B). However, the proportion of the TCRβ
   chain usage was subjected to individual changes upon vaccination
   ([131]Supplemental Figure 3C), indicating the subject-specific
   alteration of the MAIT TCR repertoire. On the other hand, constant
   CDR3α and CDR3β sequence lengths were maintained ([132]Supplemental
   Figure 3D). The PB TRAV1-2^+ MAIT cells presented a highly polyclonal
   TCR repertoire as shown by the Shannon index, ranging from 0
   (monoclonal) to 1 (polyclonal), which was maintained after the vaccine
   boost, indicating a polyclonal proliferation of the MAIT cells
   ([133]Figure 3B). Consequently, we found a prevalence of numerically
   low cell-size clones (1-3 cells) at all time points ([134]Figure 3C).
   Importantly, this polyclonal proliferation of MAIT cells at P3 was not
   the consequence of the first vaccination but was rather due to the
   independent activation upon the boost. Indeed, in tracking the fate of
   individual MAIT cell clonotypes with regard to shared clones between
   P0, P1, and P3, we did not find any specific clonotype expansion after
   the booster vaccine ([135]Figure 3D; [136]Supplemental Figure 3E).

Figure 3.

   [137]Figure 3
   [138]Open in a new tab

   Changes in MAIT cell TCR repertoire following vaccination. (A) The
   chain pairing of different AVs/AJs pairs with BVs chains at all the
   time points and subjects together are displayed as chord diagrams,
   where ribbons connecting chains are proportional to the number of
   paired AVs/AJs with BVs chains. (B) MAIT-TCR repertoire diversities
   were estimated by the normalized Shannon Index ranging from 0
   (completely monoclonal) to 1 (completely polyclonal). The bar plot
   shows the Shannon Index of time points P0, P1, and P3. (C) The pie
   charts show the frequency (%) distribution of the expanded MAIT-TCR
   clones grouped by cell sizes across time points P0, P1, and P3. (D) The
   Venn diagram (left panel) shows the overlapping MAIT-TCR clonotypes
   among time points P0, P1, and P3. The alluvion plot (right panel) shows
   the longitudinal tracking of the overlapping MAIT-TCR clonotypes among
   time points P0, P1, and P3. Each stratum represents a unique MAIT-TCR
   clonotype highlighted by a different color. The colored bands between
   columns represent shared clones among different time points.

   To further infer the activation relatedness of MAIT sub-clusters
   defined above, we estimated the RNA velocity by using the scVelo
   toolkit combined with the CellRank algorithm projected on the
   pre-computed UMAP ([139]35) ([140]Figure 4A). This approach, based on
   the ratio of spliced to unspliced pre-RNA transcripts and
   transcriptional similarities, is used to identify the directional
   transitions between cell clusters and hence construct the effector
   trajectories that account for the speed and direction of motion
   ([141]35, [142]40). This analysis visualized the inferred connectivity
   between MAIT sub-clusters and identified a trajectory with initial and
   transient states going to the final state, which is characterized by
   high values of latent time ([143]Figure 4B; [144]Supplemental Figure
   4). From velocity streamlines, we observed that MAIT cells with higher
   latent time overlapped with highly vaccine-responsive cells in sc1. By
   correlating gene expression with cell latent time, we visualized gene
   expression cascades specific to the MAIT cellular trajectory upon
   vaccination ([145]Figure 4C). The top genes driving the pronounced
   dynamics behavior and ranked from largest to smallest, showed important
   overlap with the vaccine-related activation profile of MAIT cells
   ([146]Figure 1D) (i.e., TNFAIP3, FOS, FOSB, FOSL2, JUN, JUNB, CXCR4,
   CD69, NFKBIA, NFKBIZ, KLF6, and ZFP36) to finally converge to their
   high TNF profile.

Figure 4.

   [147]Figure 4
   [148]Open in a new tab

   TNF^high MAIT cells converge towards a responsive profile following
   vaccination. (A) RNA velocity projection on UMAP plot. Each dot
   represents a cell. Cells are colored according to corresponding
   clusters. The arrows represent the RNA velocity field. (B) Latent time
   projection on UMAP plot. Each dot represents a cell colored according
   to its latent time value. (C) Heatmap of the top 50 latent
   time-dependent lineage driver genes identified by scVelo. The X-axis
   represents cells ordered by latent time (from left to right) and
   different colors correspond to each gene’s scaled (Z-scored) expression
   in each cell.

   Taken together, repeated immunizations induced transcriptional changes
   of MAIT cells linked to their proliferation and to the reinforcement of
   their effector TNF polarization.

3.3. Transcriptional TNF^high state of MAIT cells is linked to B cell
response following immunization

   Previous findings showed that the frequency of the MAIT cell
   compartment positively correlates with the anti-SARS-CoV-2 IgG levels
   in BNT162b2 mRNA vaccinated subjects ([149]28). Since some SARS-CoV-2
   infected or vaccinated individuals may elicit a cellular sensitization
   without evidence of the virus-specific Ab production ([150]41,
   [151]42), we performed a comparative analysis between the
   concentrations of anti-SARS-CoV-2 IgG and the level of the
   MAIT-associated TNF in our dataset. Importantly, we found an
   overlapping trend between these two parameters that showed the highest
   magnitude early after the vaccine boost in relation to their
   pre-vaccine status ([152]Figure 5A).

Figure 5.

   [153]Figure 5
   [154]Open in a new tab

   Transcriptional prediction of MAIT and B cells cross-talk following
   immunization. (A) Comparative analysis between the mean TNF expression
   level (red line) and the circulating anti-SARS-CoV-2 IgG concentrations
   (AU/mL, blue line) of the subject enrolled in the study across the
   different days post-immunization. The paired parametric ANOVA test was
   used for the statistical analysis, statistically significant P values
   were represented as (*): *P < 0.05, **P < 0.01 (B) UMAP clustering
   projection (left panel) of the integrated PB B cells and ASCs from all
   subjects (s01-s06; a total of 20,722 B cells and 291 ASCs). The dot
   plot (right panel) shows the expression of markers used for the
   annotation of the different maturation stages of B cells and ASCs,
   where clusters have been ordered following the maturation path (IGHD,
   IGHM, TCL1A, VPREB3, PLD4, and CD38 for transitional B cells, enriched
   in cluster 4; IGHD, IGHM, TCL1A, IL4R, and FCER2 for naïve B cells,
   enriched in clusters 0, 5, and 6; FCRL3, FCRL5, RHOB, and FGR for
   double negative (DN) B cells, cluster 3; CD27, PARP15, TCF4, AP3B1, and
   CD1C for unswitched memory (USM) B cells, enriched in cluster 1; CD27,
   CRIP1, CRIP2, TCF7, COCH, and AIM2 for switched memory (SM) B cells,
   enriched in cluster 2; CD27, MZB1, XBP1, and PRDM1 for ASCs, enriched
   in cluster 7). (C) The outcome of NicheNet’s ligand-target pairs
   regulating genes differentially expressed by the SM B cells between
   time points P3 and P0 (target genes). The top 20 prioritized ligands
   expressed by MAIT cells at time point P3, ranking according to the
   Pearson correlation coefficient, are ordered from higher (top) to lower
   values (bottom). (D) On the left, the ligand-target matrix between the
   top 20 prioritized MAIT ligands expressed at time point P3 and the
   predicted target genes expressed by the SM B cells, where the matrix is
   colored according to the regulatory potential values. On the right, the
   chord diagram shows the interaction between MAIT-derived TNF-ligand at
   time point P3 with its predicted target genes on SM B cells at P3,
   where the ribbons connecting chains are proportional to the regulatory
   potential scores. (E) The chord diagram shows the interaction between
   MAIT-derived ligands at time point P3 with receptors expressed by SM B
   cells at time point P3, where the ribbons connecting chains are
   proportional to the prior interaction potential. TNF-receptor pairs are
   highlighted in red. (F) The heatmap shows the mean gene expression of
   maturation, migration, and activation–related genes in the SM B cells
   across the different time points identified from the NicheNet analysis
   among the predicted TNF target genes (underlined) together with genes
   found in the literature with similar functions.

   We then turned to the question of whether MAIT cells can provide help
   to B cell response. First, we examined B cell populations in our
   scRNA-seq dataset. Among total PBMCs, B cells and Ab-secreting cells
   (ASCs) were identified based on the expression of canonical markers
   including CD19, MS4A1, XBP1, MZB1, and PRDM1 ([155]43). B cells were
   annotated in clusters 7, 11, and 33, while ASCs were identified in
   cluster 32 ([156]Supplemental Figure 1A). To further characterize these
   populations, B cells and ASCs were re-clustered and eight (0–7)
   clusters were obtained ([157]Figure 5B). Based on previously published
   B cell subset transcriptomic signatures ([158]43–[159]46), we annotated
   cluster 4 as transitional B cells, clusters 0, 5, and 6 as naïve B
   cells, cluster 1 as unswitched memory (USM) B cells, cluster 3 as
   double negative (DN) B cells, cluster 2 as switched memory (SM) B
   cells, and cluster 7 as ASCs.

   To investigate the MAIT cell-B cell interaction we used NicheNet, an
   algorithm that infers active ligand-target links between interacting
   cells ([160]37). Based on the gene expression changes detected upon
   vaccination at P3, we predicted networks of the ligand-receptor and
   ligand-target gene links between MAIT cells (defined as the senders)
   and specific B cell subsets (defined as receivers). Clusters 5-7 of B
   cells with no significant DEGs upon vaccination were excluded from the
   NicheNet analysis. In order to predict which MAIT cell ligands most
   likely affected the gene expression in specific B cell subsets, for
   each couple of sender-receiver cells we ranked the ligand activity,
   obtaining a list of the top 20 prioritized MAIT ligands
   ([161]Figure 5C; [162]Supplemental Figure 5). According to the
   fundamental role played by CD40L-CD40 in T-B cell interactions, CD40LG
   had a very high activity in the interaction of MAIT cells with all B
   cell sub-clusters, as assessed by its high-ranking position among the
   prioritized MAIT ligands. Other MAIT ligands involved in the activation
   of all B cell subsets included HMGB1, NAMPT, and CCL4, all genes known
   to be relevant for B cell activation ([163]44, [164]47, [165]48).

   Focusing on the role of TNF in MAIT-B cell interactions, TNF appeared
   in the list of the top 20 prioritized MAIT ligands only when
   considering the interactions between MAIT cells and cluster 2, which
   identifies SM B cells. Importantly, molecular ligand-target gene
   interactions related specifically to TNF-communication were closely
   linked to the B cell activation program that includes different
   transcription factors (TFs) (i.e., members of TF families AP-1: JUNB,
   JUN, and FOS; NF-κB: NFKBIA, NFKB1, and RELB; and IRF1, IER2, ETS1, and
   NR4A2) and molecules such as CXCR4, CD83, and BCL3 ([166]Figure 5D).
   Looking for specific ligand-receptor interaction in SM B cells, we
   found that TNF prioritized the B cell membrane TNF-binding receptor
   TNFRSF1B (TNFR2) ([167]Figure 5E), whose expression is a characteristic
   of memory B cells ([168]49).

   Furthermore, we stratified the gene expression of the activation,
   maturation, and migration molecules identified above, together with
   known SM B cell-related functional molecules (i.e., activation markers
   TNFRSF13C and ICOSLG, and the lymphoid follicles homing marker GPR183)
   ([169]50–[170]52), along the different time points. Notably, the
   expression of all these markers in the SM B cell cluster was maximally
   increased at time point P3 ([171]Figure 5F), thus suggesting that the
   BNT162b2 mRNA vaccine boost can promote MAIT cell activity and favor B
   cell activation and migration towards lymphoid follicles where the GC
   reaction occurs.

   Taken together, our data reflect a possible TNF-dependent mechanism of
   MAIT cell action to influence B cell response following mRNA-based
   vaccination. Moreover, the levels of TNF expression in circulating MAIT
   cells in the PB could serve as a useful, non-invasive predictor
   cellular biomarker of memory B cell response.

4. Discussion

   As a component of the innate immune system, MAIT cells respond quickly
   to infectious agents but have also exhibited activation and increased
   frequencies in response to vaccination in humans ([172]24, [173]53),
   mice ([174]54), and macaques ([175]25, [176]55). In the present
   longitudinal study, we evaluated the effects of the mRNA-based vaccine
   composed of a priming and boosting regimen. At single-cell resolution,
   we found that the second vaccine-induced polyclonal MAIT cell
   activation resulted in their increased frequencies and boosted the
   transcriptional profile of high TNF expression.

   In this study, many aspects of MAIT cell response were not understood,
   including the earliest events leading to their activation. Unlike
   bacteria, the vaccine is unable to generate riboflavin-derived ligands
   to activate MAIT cells ([177]4, [178]5), hence we deduced that the
   ability of MAIT cells to provide immune responses in mRNA vaccine
   immunogenicity was cytokine-licensed. Indeed, in the boosting phase,
   several cytokines such as IL-15, IL-12, IL-18, and type-I and II IFNs
   are released and are important for shaping both humoral and cellular
   protective immunity ([179]56, [180]57). Moreover, cytokines, such as
   IL-15, early after the boost correlated with spike antibody responses
   ([181]56). Previous observations showed that cytokine-related
   activation of MAIT cells was associated with increased production of
   IFN-γ ([182]14, [183]58). Here, we highlighted the critical role of TNF
   as an immune effector released by MAIT cells in response to the
   mRNA-based vaccine, suggesting the specific cytokine pattern of MAIT
   cell activation in the mRNA SARS-CoV-2 vaccine.

   Several new data have shed light on the role of cytokine-driven MAIT
   cell activation in vaccines. For instance, cytokine-driven activation
   of MAIT cells was observed in adenoviral-vector platforms ([184]19,
   [185]25, [186]59). The specific mechanism for activation in the
   adenovirus vaccine has not yet been defined but likely involves lymph
   nodes as the site of priming ([187]19). This raises another question on
   the relationship and circulation of MAIT cells between blood and
   different tissue compartments after mRNA SARS-CoV-2 immunization. The
   trafficking pattern and relationship of MAIT cells in non-lymphoid
   tissues, and lymph/lymph nodes, are still under investigation. Some
   studies provided evidence that MAIT cells located in mucosal barrier
   tissues are functionally distinct from their blood counterparts in
   regard to phenotype and functional properties ([188]60, [189]61). On
   the other hand, V. Voillet et al. provided evidence that MAIT cells are
   able to egress tissues and enter the lymph during steady-state
   conditions ([190]62). Consequently, a highly overlapping clonotype
   usage of MAIT-TCR repertoires was observed in blood and lymph.
   Therefore, further analysis should be performed to understand whether
   in our setting the TNF-boosted MAIT cells found in the PB derived from
   tissue sites.

   MAIT cells are capable to support Ab production and provide help to B
   cells in a vaccine setting. In SIV vaccination, blood and
   bronchoalveolar lavage (BAL) MAIT cells showed a greater capacity to
   secrete cytokines/chemokines (i.e., IL-6, IL-8, IL-10, IL-21, TNF,
   IFN-γ, CCL3, and CCL4) associated with help for B cell activation,
   migration, and regulation ([191]25). Although the specific role of TNF
   in this study was not assessed, the culture of MAIT cell
   derived-supernatants with B cells led to greater tissue-like memory B
   cell frequencies. TNF is a potent pleiotropic cytokine critical for
   cell trafficking, inflammation, and host defense against various
   pathogens. In mammals, TNF has been proven as an important
   co-stimulator of B cells for their polyclonal expansion on primary
   immune responses ([192]63). Moreover, TNF is required for the proper
   development and maintenance of B cell follicles and GC development
   ([193]64–[194]67).

   It is important to notice that TNF inhibitor treatment causes a lower
   serologic response and impaired memory B cell differentiation due to
   the repeated mRNA SARS-CoV-2 vaccination. In fact, patients undergoing
   anti-TNF treatment for chronic inflammatory disease showed a
   significantly lower humoral response after the SARS-CoV-2 mRNA vaccine
   ([195]68). After two doses of the mRNA-based BNT162b2 vaccine, these
   patients showed defects in the formation of antibody (Ab)-secreting B
   cells, affinity-matured memory B cells, and a dramatic reduction in Ab
   longevity ([196]68–[197]73). Therefore, we can hypothesize that TNF
   plays a crucial role in response to the mRNA-based BNT162b2 vaccine.

   Several recent studies also suggested the adjuvant role of MAIT cells
   in the context of SARS-CoV-2 vaccination. For instance, a study in vivo
   provided evidence of MAIT cells promoting CD40L-dependent activation of
   lung-associated dendritic cells (DCs), expansion of follicular T helper
   (Tfh) cells, and production of antigen-specific mucosal IgA ([198]74).
   Notably, this mouse model combined with protein antigens from
   SARS-CoV-2 promoted neutralizing Ab production. Moreover, the flow
   cytometry approach in BNT162b2 vaccinated subjects showed that the PB
   MAIT cell compartment was associated with the magnitude of both the
   adaptive CD4 T cell response and the humoral response to SARS-CoV-2
   spike after vaccination ([199]28). These findings indicated an
   unexpected association between the PB MAIT cells and the immune
   response to the BNT162b2 vaccine. Whereas in the present study, we
   provided molecular bases that link MAIT and B cell activation upon
   BNT162b2 vaccination. In fact, the advantage of having performed a
   single-cell transcriptional landscape of MAIT cells early (72 hours)
   after the vaccine boost allowed the detection of a TNF^high activated
   profile of PB MAIT cells associated with B cell activation and IgG
   production. Notably, with this approach, we observed that activation of
   SM B cells, which are the B cell subset responsible for high-affinity
   long-term memory humoral immune responses, could rely on the effector
   TNF profile of MAIT cells.

   We acknowledge some limitations associated with this study, such as the
   relatively small number of analyzed individuals, which did not allow us
   to perform a direct correlation analysis between the TNF response of
   MAIT cells and the level of anti-SARS-CoV-2 Ab production. In addition,
   the computational analysis highlighted a possible TNF-mediated
   mechanism of action of MAIT cells to promote B-cell response that needs
   to be further tested in an experimental model. Therefore, further
   studies will be needed not only to confirm the utility of MAIT cells as
   biomarkers for vaccine response, but also to demonstrate the link
   between TNF, MAIT, and B cell function in response to the SARS-CoV-2
   mRNA-based vaccine.

   Overall, our results indicate that SARS-CoV-2 BNT162b2 vaccination
   shapes MAIT cell transcriptional effector profile with the potential to
   promote B cell activation and serve as cellular adjuvants in mRNA-based
   vaccine platforms.

Data availability statement

   The data presented in the study are deposited in Zenodo, at
   [200]https://zenodo.org/record/8060366,
   [201]https://doi.org/10.5281/zenodo.8060366, with accession number
   8060366 [202]https://zenodo.org/record/8060366,
   [203]https://doi.org/10.5281/zenodo.8060366.

Ethics statement

   The studies involving human participants were reviewed and approved by
   Institutional Review Board (IRB) of Humanitas Research Hospital (HRH)
   (approval 738/20). The patients/participants provided their written
   informed consent to participate in this study.

Author contributions

   PM and ST analyzed scRNA/scTCR-seq data, performed statistics, and
   wrote the manuscript; SB and SDB supervised B cell analysis and wrote
   the manuscript; RP provided assistance in scRNA/scTCR-seq analysis and
   in the manuscript’s preparation; IS, SR, LT, and IP designed,
   performed, and provided assistance in scTCR-seq analysis; VC, AsC, MC,
   AnC, SF, AF, and AD processed biological specimens and performed
   experiments; AV, FC, and CV recruited individuals subjected to
   vaccination, managed sample storage, and assisted in experimental
   settings; JM and DM equally directed, designed, and supervised the
   study, analyzed data, and wrote the manuscript. All authors contributed
   to the article and approved the submitted version.

Acknowledgments