Tuning of human MAIT cell activation by commensal bacteria species and MR1 dependent T cell presentation

Development of a functional assay for screening MAIT-TCR stimulation by bacteria

MAIT cell activation is mostly restricted to metabolites from bacterial riboflavin pathway, expressed by many bacterial species (both pathogenic and commensal). Thus, it remains unclear whether MAIT cells can discriminate between the composition of thousands of different commensal bacterial species in the human microbiota. However, screening hundreds or thousands of bacteria for their capacity to stimulate human MAIT cells is challenging due to their relatively small and highly variable frequency in human blood. To overcome this, we aimed to develop a robust functional assay that was highly sensitive, high-throughput, and could recapitulate activation of primary MAIT cells by diverse commensals or pathogenic bacterial species found in the human microbiota (Fig. 1A). Accordingly, we utilized a human EBV- transformed and transporter associated with antigen processing (TAP)-deficient B cell line called T2 cells⁴¹ as antigen presenting cells (APC), which express MR1 on cell surface (Fig. 1B). To determine the role of MR1 in this assay, we either further overexpressed MR1 on T2 cells, or using CRISPR/Cas9 gene editing deleted MR1 or B2M (β₂-microglobulin, β2m) gene (Fig. 1B), as cell-surface presentation of MR1 molecules requires expression of β2m⁴ Wild type and MR1^hi and MR1^- T2 cells were pre-incubated with E. coli supernatant or tryptic soy broth (TSB) medium used for culturing E. coli, as control, washed and co-cultured with peripheral blood mononuclear cells (PBMC) for 24 hrs. Subsequently, activation of Vα7.2⁺CD161⁺ CD8⁺ MAIT cells in PBMC (Fig. 1C) was assessed by upregulation of CD25 and CD69 surface molecules (Fig. 1D). Both wild type T2 (T2-wt), which expresses MR1, and MRl-overexpressing T2 cells (T2-MR1^hi) upregulated CD25 and CD69 expression specifically on CD8⁺ MAIT cells in PBMCs, although T2-MR1^hi cells were more efficient, especially as E. coli supernatant was further diluted (Fig. 1D). In contrast, MAIT cells were not stimulated with either MRl-knockout T2 cells (T2-MR1^−/−) pre-incubated with E. coli supernatant (Fig. 1D, bottom panel) or MR1- expressing T2 cell lines with TSB controls (Fig. 1D), showing that specificity of activation is dependent on E. coli secreted metabolites and MR1 expression. We proceeded with subsequent experiments using T2-MR1^hi cells as they were the most sensitive, likely due to increased capacity of MR1 to capture lower levels of metabolites in the bacterial supernatants.

Using PBMCs to screen hundreds or even thousands of bacterial species for specific MAIT stimulatory activity is highly challenging, due to large cell number requirements and donor-to- donor variation in MAIT-cell frequency and responses^{1, 2,4} To overcome this, we engineered a MAIT cell specific TCR (eMAIT-TCR) with mouse TCRβ constant domain (mTCRβ), which avoids endogenous TCR mispairing, as previously described^{42, 43} Accordingly, TCR-invariant Vα7.2 sequence with Jα33 gene segment was paired with TCR-Vβ2 domain conjugated with the Jβ2.1 (TRBJ2–1) segment (Fig. 2A) for its preferential usage with the dominant MAIT-TCR- Vβ2 ^44–46 We then ectopically expressed this eMAIT-TCR on human conventional primary CD8+ T cells or a Jurkat cell line for comparison^{8, 11}. The synthesized eMAIT-TCR Vα-Vβ gene construct was subcloned into a lentiviral vector, which contained GFP as a marker, and used to transduce the Jurkat cell line or activated purified human CD8+ T cells as previously described⁴³. The Jurkat cell line or the expanded CD8+ T cells transduced with eMAIT-TCR- Vβ2 encoding vector co-expressed a GFP marker and as mentioned above, the mTCRβ, which can be stained with a specific antibody (Fig. 2B). Cells gated on mTCRβ+GFP+ subset expressed both Vα7.2 and Vβ2 parts of the eMAIT-TCR both in Jurkat or CD8+ T cells (Fig.2B) and only mTCRβ+ cells were stained with MR1–5-OP-RU tetramer (Fig. 2C), which specifically binds to MAIT-TCRs as described¹¹, and thus confirming this as MAIT-TCR.

Upon stimulation of eMAIT-TCR-Vβ2⁺ cells with T2-MR1^hi cells pre-incubated with E. coli supernatant, Jurkat cells only induced CD69, whereas, primary CD8+ T cells expressing the eMAIT-TCR greatly upregulated both CD25 and CD69 expression (Fig. 3A). The eMAIT-TCR expressing T cells were also activated in a dose dependent manner by MAIT cell stimulating riboflavin metabolite (5-ARU) (Fig. S1). T2 cells lacking MR1 expression (either through MR1 or β2Μ, which is required for MR1 expression expression^{47, 48,}_I gene-targeted CRISPR/Cas9 mediated deletion) did not activate eMAIT-TCR expressing cells either by E. coli supernatant or 5-ARU (Fig. 3B). Together, these findings conclusively show that the activation of eMAIT- TCR+ cells is MRl-dependent.

To demonstrate that eMAIT-TCR stimulation is also dependent on metabolites produced from riboflavin biosynthesis pathway of E. coli, we repressed the promoter element of ribA gene of the riboflavin pathway, by CRISPRi targeting in E. coli since ribA is required for production of MAIT-TCR stimulatory metabolites^{49, 50}. Repression of ribA mRNA transcription was confirmed by qPCR and resulted in a reduced capacity of E. coli to stimulate eMAIT-TCR- expressing T cells (Fig. 3C). Further, E. coli growth time appeared to have little impact on MAIT stimulatory metabolite production since even 1 hour growth culture was sufficient for optimal stimulation of eMAIT-TCR+ T cells (Fig. S2).

As functional outputs for TCR-mediated activation, we also measured cytokine secretion and cytotoxicity from primary CD8+ T and Jurkat cell lines. Stimulation of eMAIT-TCR+ CD8+ primary T cells with E. coli supernatant induced high level of IFNγ secretion (Fig. 3D). The primary T cells expressing eMAIT-TCR also displayed high cytotoxicity against presenting T2- MR1^hi cells pre-cultured with either E. coli supernatant (Fig. 3E and 3F) or with 5-ARU metabolite (Fig. 3F). In contrast, these effector functions were not observed in the stimulated eMAIT-TCR-expressing Jurkat cells (Fig. 3D and 3F). Thus, these findings further validate that the primary T cells are both more sensitive for activation and display physiologically relevant effector functions upon eMAIT-TCR stimulation.

Screening of MAIT-TCR stimulation with diverse bacterial species

After establishing a proof-of-principle with να7.2+νβ2+ eMAIT-TCR, to expand the breadth of MAIT-TCR representation we engineered additional TCRs that contained the constant Vα7.2- Jα33 and Jβ2.1 but was paired with 20 different Vβ segments. These synthetic TCRs were individually expressed on CD8+ T cells and Jurkat cells, which were confirmed using Vβ- specific antibodies (Fig. S3). The TCR-engineered CD8+ primary T cells and Jurkat cells were then stained with MR1 tetramers pre-loaded either with 5-OP-RU or 6-FP as a negative control. Only 6 of the Vβ segments (Vβ 2, 12, 13.2, 13.5, 13.6, 20) bound MR1–5-OP-RU, albeit at differing levels (Fig. 4, A and B), possibly due to changes in overall TCR affinities. We then tested all TCRs for their capacity to signal in T cells using functional assay described above. We found that only 7 of the engineered-TCRs (all those stained with MR1–5-OP-RU and Vβ 13.1) were also stimulated by E. coli supernatant (Fig. 4C), and that staining intensity of these with MR1–5-OP-RU correlated with the proportion of cells activated (Compare Fig. 4B with 4C). In fact, the differences between strength of activation of these 7 eTCRs were much more evident when they were activated with varying concentrations of 5-ARU and activation was measured as IFNγ secretion (Fig. 4D), which showed remarkably similar hierarchy of Vβ MAIT-TCR activation both with E. coli supernatant and in MR1-tetramer staining intensity (Fig. 4D, middle and right panels). Henceforth, we refer to these 7 functional engineered-TCRs as eMAIT-TCRs, which were subsequently used in the functional bacterial screens as described below.

We next screened 47 human-associated bacterial species in the eMAIT-TCR assay for their stimulatory capacity. These were comprised primarily of commensals and belonging to different phyla present in different mucosal tissues as commensals, in addition to a few environmental strains potentially present in different human tissues (Fig. 5A, tissue localization column) based on 16S rRNA and shotgun metagenomic studies^{51, 52}. The eMAIT-TCR with Vβ2 was the most sensitive for detecting stimulation with supernatants from the bacterial species, with more variable activation of other engineered TCRs, assessed by CD25 and CD69 (Fig. 5A, bars under each eMAIT-TCR). This was again consistent with their binding patterns to MR1-tetramers (Fig. 4A and 4B). Species that lacked riboflavin pathway also completely lacked stimulatory capacity for any of the eMAIT-TCRs (Fig. 5A, gray strip). To more broadly quantify the capacity of different bacteria for their eMAIT-TCR stimulatory capacity, we added the percent activation by each eMAIT-TCR for each species to create a cumulative stimulation index (Fig. 5B). This analysis showed remarkable clustering of bacterial species based primarily on phyla (Fig. 5B and 5C) despite significant phylogenetic distance between the species themselves. Indeed, species belonging to Bacteroidetes and Proteobacteria had statistically higher MAIT-TCR stimulation capacity compared to species Firmicutes or Actinobacteria (Fig. 5C). Importantly, measuring activation, by representative bacterial supernatants from each phylum, of primary CD8+ MAIT cells in healthy adult PBMCs, from different donors, recapitulated their respective stimulatory capacity (Fig. 5D) and correlated with cumulative stimulation of eMAIT-TCR+ CD8+ T cells (Fig. 5E). These findings suggest that activation of eMAIT-TCRs by diverse bacterial supernatants reflects their stimulatory capacity for primary MAIT cells.

To rule out the possibility that the eMAIT-TCR or primary MAIT cell activation could be affected by different bacterial density or concentration, given different growth kinetics for each species, we counted each species at their stationary phase as described in methods. While there was nearly 3-log difference in concentration among different species, there was no significant correlation between the bacterial counts and eMAIT-TCR stimulation (Fig. 5F), which is also consistent with the time course experiment using E. coli cultures where metabolite concentrations quickly reached high levels sufficient for MAIT-TCR activation before stationary phase (Fig. S2).

These findings suggested that bacteria could possess the riboflavin pathway yet not synthesize sufficient levels of riboflavin or secrete the metabolites that are required for MAIT cell activation. To confirm this, we measured riboflavin secretion from a select group of high and low MAIT-stimulating bacteria that possessed the riboflavin pathway using mass spec analysis. Those lacking the pathway were used as controls. For this experiment, the bacteria were grown overnight and then washed to remove the media, counted and resuspended in PBS at similar numbers and incubated for an additional 2 hours. The mass spec analysis of riboflavin in bacterial supernatants were then performed as described in detail in methods section. Indeed, the high-stimulator bacteria species secreted greater amount of riboflavin compared to low- stimulator bacteria (Fig. 6A).

However, it is also possible that MR1-binding metabolites produced by some bacteria may not be as stable. To address this, we added bacteria directly on to primary human macrophages in the presence of human serum. In this assay, bacteria would be taken up by the macrophages and their metabolites then bind to MR1 in macrophages and be presented to eMAIT-TCR cells. We then measured secretion of IFNγ from the eMAIT-TCR cells after co-culturing with macrophages and bacteria. Remarkably, the activation of T cells as measured by IFNγ secretion again was greater with high- vs low-stimulator bacteria or those that lacked the riboflavin pathway (Fig. 6B). In a similar experimental approach, we also added the bacteria directly to PBMCs to test activation of primary MAIT cells. We found a similar activation profile, as assessed by specific upregulation of CD25 (Fig. 6C), only on primary MAIT cells with high- vs low-stimulator bacteria (Fig. 6D) but not in any non-MAIT CD8+ T cells which did not show any activation in the presence of the bacteria. To further confirm that this was MR1-dependent T cell activation, we used a blocking MR1 antibody which completely abolished the activation of primary MAIT cells compared to isotype control antibody (Fig. 6E). Together these findings suggest that, at least for the bacteria tested in these experiments, the level of riboflavin production is appeared to be the main driver of their capacity to stimulate MAIT cells.

Stimulation of MAIT-TCR by bacterial antigens in the absence of antigen presenting cells

During course of eMAIT-TCR stimulation assays, we observed that supernatants from high stimulatory bacteria (E. coli or C. normanense) could activate eMAIT-TCR expressing cells alone, that is in the absence of T2 cells (Fig. 7A and 7B). This was a surprising result and suggested that primary human T cells may express sufficient level of MR1 for T-T cell presentation of bacterial metabolites. Although expression of MR1 in primary T cells was very low, deletion of either MR1 or β2Μ gene (as mentioned above, required for MR1 cell surface expression) using CRISPR/Cas9 gene editing in primary T cells resulted reduction in intensity of MR1 antibody staining (Fig. 7C). Furthermore, MR1 expression was increased upon incubation of the cells with 5-ARU (Figure S4), using a previously described protocol⁵³ suggesting that primary T cells can express MR1 on cell surface. Importantly, deletion of β₂m also allowed us to negatively sort the β₂m-deleted cells (Fig. 7D), to generate primary T cells completely devoid of MR1 cell surface expression. Indeed, we found that β₂m^−/− sorted CD4+ or CD8+ T cell subsets, pre-incubated with the bacterial supernatants as antigen-presenting cells (Fig. 7E), did not stimulate eMAIT-TCR+ CD8+ T cells (Fig. 7F) or primary MAIT cells (Fig. 7G). This demonstrates that T cells express sufficient level of MR1 to present bacterial metabolites to trigger MAIT-TCRs.

We next compared three different bacterial species (C. normanense, E. coli and B. subtilis) with different MAIT-cell stimulatory capacity in the absence or presence of T2-MR1^hi cells. While both primary MAIT cells within CD8+ T cells or eMAIT-TCR+ T cells were stimulated in the presence or absence of T2-MR1^hi cells with C. normanense or E. coli supernatants, the activation level based on CD25 expression were higher in the presence of T2-MR1^hi cells. Activation of primary MAIT cells alone in the presence of B. subtilis was much lower or undetectable (Fig. 7H). These experiments using bacteria supernatants revealed a gradient of MAIT-TCR stimulation strengths from resting CD8+ T cells (primary MAIT) and effector eMAIT-TCR+ cells to T2 cells overexpressing MR1, that was also dependent on stimulation capacity of the bacterial species.

To better assess and compare the presenter capacity of primary T cells, we next repeated similar experiments using 5-ARU metabolite as the antigen. In these experiments, we further sorted CD4+ T cells into naive (CD45RO-CCR7+; TN) and memory (CD45RO+; TM) subsets, as previously described⁵⁴. In addition, a portion of the CD4+ T cells were also activated using anti- CD3+anti-CD28 beads (Invitrogen) and expanded in IL-2 culture for two weeks to prepare effector T cells (TE). CD8+ T cells and B cells were also purified from the same donors using Dynabead (Invitrogen) positive selection. CD4+ T cell subsets and B cells were then incubated with different concentrations of 5-ARU for 4 hours, washed, and then added on to autologous PBMCs. In addition, we also added 5-ARU directly onto PBMCs or purified CD8+ T cells, which contain majority of primary MAIT cells. After 2 days, activation of primary MAIT cells within PBMC was determined through staining with MAIT cell markers and expression of CD25 and CD69 (Fig. 8A). Similar to bacterial supernatant stimulations, we found that T cell subsets pre-cultured with 5-ARU stimulated primary MAIT cells in a concentration-dependent manner (Fig. 8B). However, naive T cells required high concentration of 5-ARU and had the lowest capacity of activating MAIT cells, compared to memory or effector T cells (Fig. 8B). On the other hand, 5-ARU directly applied to PBMCs (that contain APCs such as dendritic cells and macrophages) or presented by purified primary B cells was relatively stronger in MAIT cell activation compared with T cell subsets used as APCs (Fig. 8B). As expected T2 cells had the highest stimulatory capacity (Fig. 8B).

We next determined whether stimulation of MAIT cells through T cell presentation of metabolites would also induce effector functions and determined how this compared to B cells or direct addition to PBMC. Accordingly, in the same experiments described above, after co-culture of 5-ARU presenting cells with PBMC, MAIT cells were stained after 22 hours, with surface markers and for intracellular expression of TNF, Granzyme B (GzmB) and IFNγ. As shown in representative flow cytometer data (Fig. 8C), sizeable portion of MAIT cells expressed both GzmB and TNF, when 5-ARU was added directly to PBMC. Concordant with results from the cell surface activation (Fig. 8B), memory or effector T cells induced lower levels of GzmB, or TNF or IFNy compared to B cell presentation or direct activation of PBMC (Fig. 8C and 8D), and that GzmB or cytokine expression from naive T cells was either very low or undetectable (Fig. 8C and 8D). Taken together, we conclude that primary T cells, particularly memory and effector T cells have the capacity to stimulate MAIT cells in antigen-specific and MRl-restricted manner.

PBMC and T cell purification

Healthy adult blood was obtained from New York Blood Center, New York, NY). PBMCs were isolated using Ficoll-paque plus (GE Health care). CD4⁺ T, CD8⁺ T, and CD19⁺ B cells were purified using Dynal CD4 Positive, CD8 Positive and CD19 positive Isolation Kits (Invitrogen) respectively. CD4⁺, CD8⁺ T and CD19+ B cells were >99% pure as assessed by flow cytometry staining with respective antibodies. Purified CD4⁺ T cells were sorted by flow cytometry (FACSAria; BD Biosciences) based on CD45RO expression into Naive T cells (CD4 TN) (CD45RO^-) and Memory T cells (CD4 TM) (CD45RO⁺), sorted subsets were >98% pure. In some experiments, CD4⁺ T cells were activated using anti-CD3/anti-CD28 coated beads (Invitrogen) and expanded for 2 weeks (TE). Cells were cultured at 37°C and 5% CO₂ in complete RPMI 1640 medium (RPMI supplemented with 10% Fetal Bovine Serum (FBS, Atlanta Biologicals), and 1% penicillin/streptomycin (Corning Cellgro).

Cell surface and intracellular staining and flow cytometry analysis.

Cells were resuspended in staining buffer (PBS + 2% FBS) and incubated with fluorochrome- conjugated antibodies for 30 min at 4^oC. Stimulation of eMAIT-TCR⁺ T cells was determined by staining with following antibodies: mouse constant TCRβ-PE (eBioscience), CD3-APC/Cy7, CD69-PerCP/Cy5.5, CD25-APC. Primary MAIT cell (CD3⁺CD8⁺Va7.2⁺CD161⁺) stimulation was also determined by staining with antibodies CD25, CD69 and CD4-BV605, CD8-A700, Va7.2-PE and CD161-BV421 (all from Biolegend). In cytotoxicity experiments, T2 cells were labeled with Cell Trace Violet (CTV) (Invitrogen) and pre-incubated with bacteria supernatant for 4 hours. After incubation, cells were washed and added to eMAIT cells. After 2 days, cells were stained with eFluor 506-conjugated Fixable viability dye (eBioscience) and eMAIT surface markers and then analyzed by flow cytometry. MR1 expression on T2 cells was assessed with anti-MR1-PE and β2m expression with antiβ₂m-APC (all from Biolegend). Expression of eMAIT-TCRs was determined by expression of GFP, Va7.2 and respective Vβ antibody staining (kit from Beckman Coulter).

For intracellular cytokine staining, PBMCs were stimulated with the riboflavin metabolite 5- amino-4-D-ribitylaminouracil dihydrochloride (5-ARU, Toronto Research Chemicals, Canada) for 16 hours, then cultured additional 6 hours in the presence of GolgiStop (BD Biosciences). Cells were then washed and stained for surface markers using anti-CD3-Alexa Fluor 532 (eBioscience), anti-CD4-Pacific Blue, anti-CD8-Brilliant Violet 570, anti-Vα7.2-PE, anti- CD161- Brilliant Violet 421 and Fixable Viability Dye-eFluor 506 (Biolegend) and permeabilized, using a fixation/permeabilization buffer (eBioscience). Fixed cells were then stained intracellularly with anti-GzmB-Alexa Fluor 647, anti-TNFα-PE/Dazzle 594 and anti- IFNγ-APC.Cy7 (Biolegend). For MR1 blocking experiments, cells were pre-incubated with MR1 blocking antibody or mouse IgG2a isotype control (both from Biolegend) for 2 hours before stimulation with 5-ARU or bacteria.

Identification of MAIT cells with MR1 tetramers was described in previous study¹¹. APC- conjugated human MR1 tetramers which are pre-loaded with either 5-OP-RU or 6-FP (negative control) were provided by NIH Tetramer Core Facility. Engineered MAIT-TCR expressing CD8+ T or Jurkat cells were stained with hMR1 tetramers for 40 min at room temp followed by mTCRβ surface staining for 30min. All flow cytometry analysis was performed using LSR Fortessa X-20 flow cytometer (BD Biosciences) or SP6800 (Sony) flow cytometer and analyzed using FlowJo software (Tree Star).

Preparation of bacteria supernatants

Bacterial strains were grown to stationary phase in the appropriate medium (Table S1) and centrifuged at 4000rpm to pellet the cells. The supernatants were then filtered through a 0.22- micron PES filter (EMD Millipore) to remove any remaining cells, and aliquoted for testing in the MAIT cell assay. Bacterial strains were obtained from the American Type Culture Collection (ATCC), DSMZ-German Collection of microorganisms and cell culture- and BEI Resources. Anaerobic bacterial species (marked * in Table S1) were grown in the anaerobic chamber (BACTRON, Shel Lab) with the conditions including 90% N₂, 5% H₂, 5% CO₂. K. pneumoniae strain was isolated from a patient by Bo Shopsin at New York University School of Medicine. Staphylococcus epidermidis Tu3298 was provided by Michael Otto at NIH.

Preparation of human macrophages and culture with bacteria

Monocytes were purified from PBMC of blood as previously described⁸⁶ and cultured with RPMI media with 5% human serum and GM-CSF for one day at 30 thousand cells/well in 96- well-plate. Selected bacteria were counted, centrifuged and resuspended in the RPMI+5% human serum media and added to macrophage cultures at varying numbers. After 4 hours of culture, eMAIT-TCR cells were added to the same cultures. After 2 days, IFNγ, was measured in the supernatants by a bead assay (Qbeads, Intellicyt) according to manufacturer’s instructions and analyzed with iQue Screener plus (Intellicyt) flow cytometer.

CRISPR/Cas9 gRNAs and vectors.

LentiCRISPR_V2 (Addgene plasmid #52961)⁸⁷, was a gift from Feng Zhang. To generate gene- specific 20-nucleotide target sequences with minimum off-target scores, the online CRISPR design tool was used ( which was developed by the Zhang Lab). The sequences were selected to precede 5’-NGG protospacer-adjacent motif (PAM) sequences. 20-nucleotide sequences 5’- GAACCTCGCGCCTGATCACT-3’ and 5’-GAGTAGCGCGAGCACAGCTA-3’ were designed to target the MR1 gene and the B2M (β₂m) gene, respectively. The oligonucleotides were purchased from Eurofins Genomics. Next, the oligonucleotides were annealed and cloned into the single guide RNA scaffold through Bsmbl-BsmbI sites in the LentiCRISPR_V2 vector. The cloned constructs were then transformed into NEB Stable Competent E. coli (New England Biolabs). The transformants were cultured overnight before plasmid DNA was isolated using QIAprep Spin Miniprep Kit and Qiacube (Qiagen). Successful cloning with the annealed oligo was confirmed by determining 2kb difference relative to the conventional LentiCRISPR_V2 vector that lacks the oligo after digesting the vectors with both Notl and BamHI restriction enzymes and running on an agarose gel. Human MR1 transcript variant 1 (NM_001531) cDNA clone was purchased from Origene. The clone was then sub cloned into lentiviral vector encoding GFP marker⁴³.

Synthetic engineering and cloning of MAIT-TCRs.

Engineered TCRs were constructed as described in Figure 2A. Peptide sequences of the gene segments were downloaded from Ensembl Genome Browser. Transcript ID for each gene segment and sequences for p2A and Jβ2–1 linkages were listed in Table S2. The segments were connected in the following order: Vα7.2-Jα33-mTCRα-p2A-Vβ-Jβ2--mTCRβ, converted into a single open reading frame (ORF) and synthesized by Genscript. The synthesized eMAIT-TCR ORFs with each of the 21 different Vβ gene segments were sub-cloned into a lentiviral vector encoding GFP marker⁴³.

Lentiviral production and transduction.

The confirmed lentiviral constructs including MR1 or eMAIT-TCRs encoding vectors and the CRISPR constructs were co-transfected with the packaging plasmids VSVG, pLP1 and pLP2 into HEK293T cells using Lipofectamine™ 3000 (Invitrogen) according to the manufacturer’s protocol. Lentiviral supernatants were collected, filtered and stored as previously described ⁴⁴ Cell lines or primary T cells were transduced with lentiviruses line was infected with MOI of 1–5. For primary T cells, lentiviruses were added to purified CD8⁺ and CD4⁺ T cell subsets, which were stimulated using anti-CD3/anti-CD28 coated beads (Invitrogen) at 1:2 (bead: cell). After three days, the cells were selected either with 0.5–1 μg/ml puromycin antibiotic for 3–5 days until more than 95% of non-transduced cells were dead or sorted based on marker gene.

Bacteria counting

The Bacteria Counting Kit, which enumerates bacteria by flow cytometry, was purchased from Invitrogen (MP07277). The kit includes a nucleic acid stain that penetrates both gram-negative and gram-positive bacteria, producing green-fluorescent signal. The kit also includes a calibrated suspension of microspheres. Signals from both the stained bacteria and the beads were detected in the green fluorescence channel and were distinguished on a plot of forward scatter versus fluorescence using a flow cytometer. The density of the bacteria in the sample was determined from the ratio of bacterial signals to microsphere signals in the FACS plot as the commercial protocol suggests.

E. coli riboflavin gene knockout via CRISPR/Cas9

CRISPR approach was used to achieve gene knockdown. Respective E. coli rib A gene sequence was extracted from KEGG⁸⁸ including 100-nucleotide upstream of the start codon. Custom-made scripts were used to design spacers targeting the start sequence of the genes (non-template strand). The spacer oligos using primers (RibA_spacer_F, 5’- AAACGTTCTTCAAATCCCACCATCG and RibA_spacer_R, 5’-AAAACGATGGTGGGATTTGAAGAAC) were cloned into pdCas9 ⁵⁰. pdCas9 was a gift from Luciano Marraffini (Addgene plasmid # 46569). Spacer cloning into pdCas9 was carried out with 5-alpha competent E. coli (NEB) as described in Addgene spacer cloning manual and the manufacturers manual. The transformed cells were plated on tryptic soy plates with chloramphenicol. Colonies were confirmed for the presence of the spacer in pCas9 plasmid using the primer pdCas9_f (5’-GTCTGAGCAAGAAATAGGCA) binding Cas9 and the respective reverse spacer pdCas9_r (5’-CGGAATGGACGATCACACTACTC) through PCR as well as a primer binding downstream of the integrated spacer pdCas9_r and the respective forward spacer. Positive colonies were grown overnight. For the eMAIT-TCR stimulation assay supernatants were obtained after centrifugation of the overnight culture for 5 min at 4000rpm followed by sterile filtration (0.2μm Fisherbrand Syringe Filters). Cell pellets were used for RNA extraction.

RNA isolation and rtPCR

E. coli overnight cultures were used to extract the RNA using the Qiagen RNAeasy kit. On- column DNase digestion step was carried out according to the manufacturer’s manual. To improve purity of the RNA, a second round of DNase digestion was performed by adding 1μl of DNase to each 10μg of RNA eluted using the TURBO DNA-free Kit (ThermoFisher Scientific). After 30 min, another 1μl of DNase was added to the RNA and the reaction was quenched after 30 min according to the manufacturer’s manual. Integrity of the RNA was checked on an agarose gel.

Reverse transcription was performed according to the manufacturer’s instructions using the High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific). Per sample a reaction lacking the reverse transcriptase was included to account for DNA contamination. 0.4μg of RNA was used in each reaction. QRT-PCR primers (RibA-QPCR_F, 5’- CAGCTTAAACGTGTGGCAGA and RibA-QPCR_R, 5’-CGTCACCGGTCAGACATTC) were designed using the RealTime qPCR Assay tool from IDT and ordered from IDT. Off-target binding was checked by aligning the primer and spacer against the E. coli genome. Each sample was tested in replicates using the following protocol: 5μl 2x PowerUp SYBR Master Mix (Applied Biosystems), 0.25μΜ Primer mix (5μΜ), 2.25μl H2O and 2.5μl of RNA. RT-PCR was carried out using the ViiA-7 Real-Time PCR System (ThermoFisher Scientific). The cysG gene was used as the control housekeeping gene (CysG-QPCR-F, 5’-GGCGAAGAGCTGGAAACA and CysG-QPCR-R, 5’-CGTGAGTGGAATACCCGAATAG) and ⁸⁹. The Deltadelta CT- method implemented in the Quant-Studio Real-Time PCR Software (v1.1) was used to analyze the RT-PCR data.

Riboflavin pathway analysis

Riboflavin biosynthesis genes in bacteria whose genome sequences have been annotated by Kyoto Encyclopedia of Genes and Genomes (KEGG) were identified according to the annotation in KEGG pathway map00740. For unannotated bacterial genomes, riboflavin biosynthesis genes were identified in the following steps. First, all KEGG-annotated bacteria riboflavin biosynthesis genes (KEGG orthologues K01497, K01498, K02858, K00793 and K00794) were downloaded using the KEGG REST R package (version 1.14.0), and grouped according to KEGG orthology. Second, for each unannotated microbial genome, genes were predicted using PRODIGAL with default parameters ⁹⁰. Third, potential riboflavin biosynthesis genes were identified by aligning the KEGG-annotated riboflavin biosynthesis genes to the predicted genes using UBLAST⁹¹, with parameters evalue=1e-9 and accel=0.6. Next, top hits from the blast search were validated using SMARTBLAST based on consistency in domain predictions. Finally, bacterial species that contain KEGG orthologues K01497, K01498, K02858, K00793 and K00794 are classified as species that have the riboflavin biosynthesis pathway.

Mass spectrometry quantitation of riboflavin in bacterial supernatants

Bacteria supernatant in PBS was acidified with 20% formic acid in water. Stable isotopic riboflavin-¹³C,¹⁵N₂ (VB2; Toronto Research Chemicals, Canada) was spiked as the internal standard. The spiked supernatants were desalted using an Oasis HLB 96-well μElution plate (Waters, Waltham, MA), following manufacturer’s recommended procedures. Each sample was eluted with 50 μl of the elution solution, concentrated, and reconstituted to 20 μl with 0.2% formic acid in water. Riboflavin in each sample was quantified used the method of liquid chromatography-stable isotope dilution multiple reaction monitoring mass spectrometry (LC-SID-MRM MS)⁹² with transitions of 377.15/172.10, 375.15/198.10, and 377.15/243.10 for positive ions of riboflavin and 380.10/173.11, 380.10/200.06, 380.10/246.07 for VB2. The column (Atlantis dC18, 5 μm, 0.3 mm x 150 mm; Waters, Waltham, MA) temperature was 40 °C. Solvent A was water/acetonitrile/formic acid = 98.8:1.0:0.2 (v/v/v), solvent B was water/acetonitrile/formic acid = 1.0:98.8:0.2 (v/v/v), and the flow rate was 9 μl/min. Each sample was analyzed in triplicate. Chromatograms were processed using the Skyline software and ratios of the peak area for riboflavin to that for VB2 were obtained. Original concentrations of riboflavin in bacterial supernatants were calculated using these ratios and the absolution concentration of VB2 in the analysis samples. Limit of quantitation for riboflavin was determined using the method of “blank and low-concentration samples”⁹³. PBS treated with the same desalting procedure was used as the blank.

Statistical analysis

Correlation with the Spearman’s rank test and two-tailed Mann-Whitney U tests were performed using GraphPad Prism software. No outliers were excluded in any of the statistical tests. Scatter dot plots report the median center lines with 25^th and 75^th quartile (interquartile range, IQR), and each data point represents an independent measurement. Correlation analysis was assessed with Spearman’s rank test. Bar plots report the mean and standard deviation or the standard error of the mean. Threshold of significance for all tests was set at <0.05.