Activation and In Vivo Evolution of the MAIT Cell Transcriptome in Mice and Humans Reveals Tissue Repair Functionality

1Mucosal-associated invariant T (MAIT) cells are innate-like T cells that express a “semi-invariant” αβ T cell receptor (TCR) and recognize metabolic derivatives of riboflavin biosynthesis presented on the restriction molecule major histocompatibility complex (MHC)-related protein-1 (MR1). These antigens, which include the potent MAIT cell ligand 5-(2-oxopropylideneamino)-6-D-ribitylaminouracil (5-OP-RU), are produced by a wide variety of bacteria, mycobacteria, and yeasts but are absent from mammals and therefore allow host-pathogen discrimination. MAIT cells have a strong pro-inflammatory phenotype and produce interferon-γ (IFN-γ), tissue necrosis factor (TNF), and IL-17A after phorbol myristate acetate (PMA) and ionomycin stimulation or after infection.

2While baseline frequencies of MAIT cells are low in specific-pathogen-free C57BL/6 mice, we, among others, have previously shown that MAIT cells can be activated and expand in vivo in response to a pulmonary infection with specific intracellular bacteria expressing the riboflavin pathway—Salmonella Typhimurium, Legionella spp, and Francisella tularensis—or in response to synthetic 5-OP-RU accompanied by a Toll-like receptor agonist, providing valuable models to dissect MAIT cell biology.

3To date, the requirements for TCR-dependent activation of MAIT cells in vivo have not been systematically characterized, nor have the consequences of such activation been fully defined. Here we have used MR1 tetramers loaded with 5-OP-RU to specifically identify MAIT cells from human peripheral blood and murine lungs, allowing us to assess the requirements for, and consequences of, MAIT cell activation ex vivo and in vivo. Using a transcriptomic approach on sorted MR1-5-OP-RU tetramer+ cells, we define the transcriptome of an activated MAIT cell in both species and explore how this changes during the resolution of infection and during reinfection.

4Our data reveal strong similarities between MAIT cells in humans and in mice at a transcriptional level; show that MAIT cells displayed the closest similarities to invariant natural killer T (iNKT) cells when activated, but after the resolution of infection were more comparable to γδ T cells; and reveal a previously unknown tissue repair phenotype expressed upon MAIT cell activation in both species.

5First, we aimed to test, systematically, the activation requirements of MAIT cells in vivo in mouse lungs. We have previously shown that pulmonary MAIT cell frequencies in mice can be markedly enhanced by intranasal administration of 5-OP-RU if it is co-administered with S-[2,3-bis(palmitoyloxy)propyl] cysteine (Pam2Cys), CpG ODN 1668, or polyinosinic:polycytidylic acid (poly I:C), which are agonists for TLR2/6, TLR9, and TLR3, respectively. We therefore investigated agonists for each of the murine TLRs, using the maximum doses presented in a literature review of previous studies of these compounds. All animals received the relevant TLR intranasally on day 0. In experimental animals, this was administered in combination with 76 pmol 5-OP-RU on day 0, with repeated inoculae of 76 pmol 5-OP-RU on days 1, 2, and 4. Control mice received the same TLR ligand and 76 pmol of the non-activating MR1 ligand 6-formyl pterin (6-FP) according to the same schedule, or the TLR ligand alone (Table S1). We observed 15- to 180-fold enrichment of pulmonary MAIT cell frequencies (defined as CD3+CD45.2+CD19−MR1-5-OP-RU tetramer+ cells) at day 7, after administration of 5-OP-RU with agonists of TLR3 (high molecular weight poly I:C), TLR4 (lipopolysaccharide from E.coli), TLR2/6 (FSL-1 [Pam2CGDPKHPKSF]), and TLR9 (CpG ODN1826), but not with agonists of TLR1/2 (Pam3CSK4), TLR2 (heat killed Listeria monocytogenes), TLR5 (flagellin from S. Typhimurium), or TLR7 (Imiquimod) (Figure 1), suggesting there is a specific and restricted range of danger signals that are capable of providing the necessary co-stimulus to drive MAIT cell accumulation in response to 5-OP-RU antigens.

6Having defined the minimum requirements for TCR-mediated activation of MAIT cells, we sought to describe in detail the consequences of their activation using a transcriptomic approach to define the basic transcriptome of a MAIT cell in both humans and mice and to determine how this is modulated by activation. Fresh human peripheral blood cells were obtained from three donors. These were cultured for 6 h with (“stimulated”) or without (“unstimulated”) 10 nM 5-OP-RU; magnetically enriched on MR1-tetramer+ cells; and flow-sorted for RNA sequencing of live CD3+TCR Vα7.2+ MR1-5-OP-RU tetramer+ MAIT cells, and of unstimulated naive live CD8+CD45RA+ T cells as a comparator cell type (Table S2). This cell type was selected as a comparator to enhance the contrast with MAIT cells and allow a more comprehensive characterization of the MAIT cell’s full functional phenotype.

7We have previously shown that pulmonary infection of mice with the intracellular pathogen Legionella longbeachae induces strong TCR-mediated MAIT cell activation, and this plays a significant role in host immune protection, thus constituting a physiologically relevant model of in vivo MAIT cell activation. This model infection induces a rapid and sustained expansion of MAIT cells in the lungs (Figure S1), which closely parallels the MAIT cell expansion observed with TCRs and specific TLR ligation in Figure 1. Using this model, we therefore included within the same sequencing experiment live pulmonary CD3+45.2+19−MR1-5-OP-RU tetramer+ MAIT cells, which were magnetically enriched and flow-sorted from the lungs of mice 7 days after infection with 1 × 104 CFU L. longbeachae (“acute”), or at least 12 weeks post infection (“resolution”) or 7 days after a second intranasal infection with 2 × 104 CFU L. longbeachae in mice that had recovered from infection 12 weeks previously (“reinfection”). Because of the very low numbers of MAIT cells in uninfected, specific-pathogen-free mice, it was not possible to obtain robust data on naive MAIT cells from such animals. Instead, live CD3+CD45.2+CD19−CD8+CD44−CD62L+ naive T cells from uninfected mice were used as a comparator cell type, with the objective of providing a comprehensive characterization of MAIT cells’ phenotypic and functional capabilities.

8The number of differentially expressed genes (DEGs) in activated MAIT cells, compared with naive CD8+ T cells, was 4,613 genes in human 5-OP-RU-stimulated MAIT cells and 3,758 genes in acutely infected mice at a false discovery rate (FDR) p value of < 0.05 and minimum log(2) fold change of ±1. (Numbers and full lists of DEGs are shown in Table S3.) These genes constitute the basic transcriptome of an activated MAIT cell in each species. To explore the nature of these gene profiles further, we compared different activation states of MAIT cells. In humans, 3,227 genes were differentially expressed between stimulated and unstimulated MAIT cells and could therefore be considered the direct signature of TCR-mediated MAIT cell activation, while 968 genes were differentially expressed between unstimulated MAIT cells and naive CD8+ T cells and therefore are more related to constitutive differences between T cell lineages (Table S3). In mice, 1,889 genes were differentially expressed between acute infection and the resolution of infection, analogous to the signature of TCR-mediated activation (Table S3).

9Analysis of TCR genes over-represented in MAIT cells confirmed highly significant, selective use of TRAV1-2 with T cell receptor beta variable (TRBV)6-4, TRBV6-1, and TRBV20-1 in humans and Trbv13-3 with Trav1 and Traj33 in mice, in MAIT cells compared to naive CD8 T cells, as expected (Table S4; Figures S2 and S3). We did not observe a specific change in the MAIT cell TRBV repertoire between stages of infection.

10Focused analysis of known cytokine genes confirmed a strong upregulation of several pro-inflammatory type 1 and type 17 cytokines, especially CSF2 (GM-CSF), IL-17A, LIF, TNF, IFN-γ, and IL-17F, which were highly upregulated in both mice and humans in activated MAIT cells (Tables 1 and 2). Expression of selected cytokines was confirmed by flow cytometry (Figures S4 and S5). Likewise, there was significant, but more modest, upregulation of LTA (lymphotoxin A) and CSF1 (M-CSF) in both species. Some features were observed only in one species, notably IL-2 and TNSF14 (LIGHT) produced by activated human MAIT cells and Tnfsf11 (TNF-related activation-induced cytokine E [TRANCE], receptor activator of nuclear factor kappa-B ligand [RANKL]) by murine MAIT cells. In contrast to activation-induced cytokines, expression of the anti-apoptotic cytokine IL-15, implicated in the development and maturation of memory CD8+ T cells, was restricted to MAIT cells in their resting states: human unstimulated MAIT cells or murine MAIT cells at the resolution of infection. In mice, the resolution of infection was also associated specifically with a strong expression of Tnfsf18 (glucocorticoid-induced TNF-related ligand [GITRL]), confirmed by flow cytometry. Expression of cytokine receptors is shown in Table S5, and that of recognized “CD markers” is shown in Table S6.

11Similar analysis of chemokines showed strong activation-induced upregulation of a range of chemokines, including XCL1, CCL3 (MIP1α), CCL4 (MIP1β), and CXCL16, common to both species (Table S7), and of a common array of chemokine receptors CCR6, CXCR6, CCR1, CCR2, and CCR5 (Table S8; Figure S4), underlining a marked evolutionary conservation of MAIT cell function.

12To analyze the transcriptome at the level of pathways, rather than individual genes, we looked for upregulation of pathways using the open-source, manually curated, peer-reviewed Reactome database. The main pathways upregulated in human 5-OP-RU stimulated MAIT cells, compared with naive CD8+CD45RA+ T cells, were related to endoplasmic reticulum stress—the unfolded protein response—and the related pathways IRE-1-α activation of chaperones and XBP1(S) activation of chaperones to chemokine receptor-ligation and to cholesterol biosynthesis (Figure 2A). When stimulated, human MAIT cells were contrasted directly with unstimulated MAIT cells: the activation of chemokine and cytokine signaling pathways—chemokine receptor-ligation, IL-2 signaling, and interleukin receptor Src homology and collagen (SHC) signaling—was more apparent, as was human solute carrier-mediated transmembrane transport (Figure 2B).

13Perhaps reflective of the different context of activation, and consistent with the rapid MAIT cell expansion observed with infection in vivo, the murine MAIT cells activated by the acute L. longbeachae infection showed a very strong activation of cell cycle pathways, as well as signaling by RHO guanosine triphosphatases (GTPases) and chemokine receptor-ligation, with similar dominance of the cell cycle when MAIT cells activated by acute infection were contrasted directly with unstimulated MAIT cells after infection resolution (Figures 2C and 2D).

14MAIT cells are a relatively ancient T cell subset, with both innate and adaptive properties, and they are capable of expressing diverse functions depending on the nature of the pathogenic encounter. Therefore, we sought next to explore the nature of the murine MAIT cell transcriptome by comparing it with the transcriptional profiles for a wide range of other cell types reported within the Immunological Genome Project database (Table S2). Using hierarchical clustering, while the pulmonary naive CD8+CD44−CD62+ T cells clustered with the reference naive CD8+ splenic T cells, activated MAIT cells from acute primary infection or from acute reinfection clustered most closely to iNKT cells (Figure 3). By contrast, after resolution of the infection, MAIT cells clustered most closely with unstimulated splenic γδ T cells. We explored what transcriptional processes might drive this difference through analysis of differentially expressed genes (Table S9) and by using Reactome pathway analysis (Figure S6A). MAIT cells differed from non-thymic precursor iNKT cells by relative upregulation of Plekstrin (Plek), Lyn proto-oncogene (Lyn), SH2 domain containing 1b (Sh2d1b1, a regulator of natural killer cell effector functions), and integrin α X chain protein (Itgax, CD11c). Reactome analysis showed MAIT cells had an upregulation of pathways associated with neutrophil degranulation and pathways associated with cell surface interactions with vasculature. MAIT cells at resolution differed from γδ T cells most by upregulation of cell surface receptors for IL-18 (Il18r1), vitamin D (Vdr), and leukotriene B4 (Ltb4r1) and in pathways associated with RNA transcription.

15Our observation of a distinct cytokine signature after infection resolution suggested that MAIT cells might be capable of performing more diverse functions than a purely pro-inflammatory response to TCR ligation. As observed already, TCR ligation in the absence of a TLR-agonist did not induce proliferation of murine MAIT cells. A wide variety of bacteria, mycobacteria, and yeasts, including many commensal organisms, express the riboflavin biosynthetic pathway and may therefore be a major source of activating MR1 ligands, constitutively, or during breach of a barrier surface. Indeed, MAIT cells require commensal organisms for their expansion. A class of skin-homing Tc17 cells specific to commensal flora, which expresses a “tissue repair” gene signature and can accelerate repair of an epithelial wound, has recently been described. These cells share several features with non-classical T cells, including the type 17 cytokine profile and restriction by another MHC class 1b antigen presentation molecule H2-M3. Therefore, we asked whether this tissue repair phenotype was a shared transcriptional program in MAIT cells. We used gene set enrichment analysis (GSEA) to compare the expression of this set of tissue repair genes (Table S10) with genes differentially expressed in MAIT cells. Indeed, this gene set was markedly enriched in human MAIT cells after 5-OP-RU stimulation (normalized enrichment score [NES] 1.38; familywise error rate [FWER] p < 0.01; Figures 4A and 4B; Table 3). By contrast, there was no significant enrichment of this gene signature when the same analysis was performed on published gene expression data from negatively selected CD3+ T cells stimulated for 24 h with antibody co-stimulation of CD3/CD28 (Figure S6B). Similarly, despite differences in species, time course, and method of MAIT cell activation, the same gene set was even more highly enriched in mice during acute L. longbeachae infection (NES 1.38; FWER p < 0.01; Figures 4C and 4D; Table 3), with enrichment of 10 genes common to both analyses (TNF, CSF2, HIF1A, FURIN, VEGFB, PTGES2, PDGFB, TGFB1, MMP25, and HMGB1).

16Finally, to investigate the extent to which this similarity to H2-M3 restricted T cells was a property specific to MAIT cells, we repeated the previous hierarchical clustering while incorporating published RNA sequencing transcriptomes of three additional cell types: H2-M3 f-MIIINA:H2-M3-tetramer+ CD8+ T cells from mice that had been topically associated with the skin commensal S. epidermidis strain NIHLM087, from which the “tissue repair signature” was first identified and that had been activated in vivo by peptide injection; CCR6+ skin Tc17 cells obtained from mice that were previously associated with this skin commensal; and additional naive CD8+ T cells from secondary lymphoid organs of uninfected specific-pathogen-free mice (Figure 3, gray lozenges; Table S2). As expected, activated H2-M3-specific CD8+ T cells that had been activated by in vivo peptide stimulation clustered immediately adjacent to the MAIT cells from acute infection or reinfection, in which we had observed the tissue repair signature. The skin epithelial cells obtained without recall challenge clustered immediately adjacent to the MAIT cells that were at resolution state, while the additional naive CD8 T cells from secondary lymphoid organs clustered with the other naive CD8+ T cell subsets.

17Here, we have systematically investigated the requirements for TCR-mediated activation of MAIT cells in mice and we delineated both ex vivo in human and in vivo in mice the consequences of this activation at a transcriptomic level. Because of their pro-inflammatory cytokine profile and specificity for a restricted selection of microbially derived small molecules, the most immediately apparent function of MAIT cells has hitherto been the early detection of microbes and initiation of an inflammatory host response. Consistent with previous studies, our data confirm MAIT cells’ capacity for a strong, rapid pro-inflammatory response. However, in contrast to the similarities between activated MAIT cells and iNKT cells, the close similarity at a transcriptional level of resting murine MAIT cells to γδ T cells and the discovery of a clear transcriptional signature for tissue repair suggest that MAIT cells potentially have much broader roles in mucosal immunity.

18The nature of these roles may depend on the context and nature of the cells’ activation. As with iNKT cells, MAIT cells may be activated either via TCR recognition of ligand presented on MR1 or via cytokines alone, in the absence of a TCR signal, as occurs during respiratory viral infection. As we have observed previously, in the absence of inflammatory cytokines, TCR ligation alone is not sufficient to produce MAIT cell proliferation and activation in vivo. Rather, a second signal is required. In vitro in humans, it has been shown that agonists of TLR1, TLR2, and TLR6 can provide this co-stimulus to drive MAIT cell cytokine secretion. Consistent with, and extending previous, observations, we observed here that murine MAIT cells proliferated in response to a different, but similarly restricted, set of TLR agonists: those for TLR3, TLR4, TLR6/2, and TLR9, but not for other TLRs tested. Thus, specific activation of the MAIT TCRs in the context of specific TLR stimulation is sufficient for rapid and robust MAIT cell expansion, which is of similar magnitude and time course to that observed with the Legionella model we used for the transcriptomic experiments.

19While MAIT cells have been considered pro-inflammatory, our data suggest an entirely distinct function for MAIT cells in tissue repair. In both human and murine datasets, activated MAIT cells highly express a shared gene expression signature with murine H2-M3 restricted commensal-specific Tc17 cells recently reported by. Using topical skin colonization with a specific clade of S. epidermidis, this group demonstrated that specific commensal-derived N-formylated peptides presented on H2-M3, another class 1b MHC molecule, could induce tissue-resident Tc17 cells, which provided specific capacity to promote tissue repair and remodeling (MMP25, Furin, PDGFB, TGFB1) and angiogenesis (CSF2, VEGFA, PDGFB). Healing of skin wounds was shown to be accelerated by colonization of H2-M3-sufficient mice with these commensals. A human equivalent of H2-M3 has yet to be identified, but MAIT cells are abundant at barrier sites, allowing close interactions with commensal bacteria possessing an intact riboflavin metabolic pathway. Similar to H2-M3-restricted CD8+ T cells, in this position, MAIT cells are poised to maintain tissue homeostasis in the presence of commensals, thereby limiting inflammation and associated tissue injury. We did not observe enrichment of this tissue repair signature in published gene expression data from TCR-stimulated CD3+ T cells, suggesting this functionality is not a common property of most T cells, but is restricted to only certain T cell subsets. Indeed, when TCR-activated commensal-specific H2-M3-restricted T cells were incorporated into our hierarchical cluster analysis of 100 other lymphocyte subsets, they clustered very closely with the activated MAIT cell subsets, while skin Tc17 cells without activation clustered most closely with MAIT cells at the resolution of infection. These findings provide additional evidence to suggest strong functional similarities between H2-M3-restricted Tc17 cells and the MAIT cell subset specifically. Furthermore, these data are consistent with similar findings of this same tissue repair signature observed by gene set enrichment analysis of MAIT cells when activated by TCR triggering, but not observed in the context of cytokine-mediated activation. These findings might also explain the increased gut permeability observed in Mr1−/− NOD mice compared with Mr1+/− NOD littermates, which suggested a protective role for MAIT cells for maintaining gut homeostasis. It has been speculated this might be mediated by IL-17A and IL-22, which are both important in intestinal homeostasis and, in the case of IL-22, induction of protective mucus-producing goblet cells. In our dataset, both cytokines were strongly upregulated in activated murine MAIT cells. Thus, during mucosal damage, riboflavin-synthesizing pathogens and commensal organisms might provide both the MAIT cell activation to induce the necessary inflammatory response to ensure bacterial clearance and the signals necessary to accelerate healing of the wound. After a successful clearance of infection, or barrier repair, the subsequent reduction in MR1-Ag presentation would ensure this signal declined.

20Commensals might drive the MR1-MAIT cell axis in other ways. MAIT cell expansion requires exposure to a commensal microbiome. Furthermore, commensal microbes have been implicated in enhancing host immunity against pathogens in the respiratory tract. In mice, a Nod2-mediated IL-17A response to upper-respiratory tract commensals enhanced CSF2 (GM-CSF) to promote bacterial killing and clearance by alveolar macrophages. The strong upregulation of CSF2 (GM-CSF) we observed following TCR stimulation in human and murine MAIT cells would be beneficial in the clearance of pathogenic microorganisms that have crossed the mucosal barrier. During tissue homeostasis, commensal-derived MR1 signals might drive lower-level, constitutive expression of GM-CSF needed to maintain alveolar macrophages in a pathogen-responsive state.

21Another novel, prominent feature of MAIT cell activation in both humans and mice was the marked expression of the IL-6 family cytokine leukemia inhibitory factor (LIF). Consistent with our findings of a MAIT cell tissue repair signature, the LIF has been found to protect against epithelial damage in murine models of pneumonia. The LIF is significantly induced during pneumonia and can reduce lung epithelial cell death, promoting the expression of tissue-protective genes essential to lung regeneration and repair and increased mucosal barrier integrity.

22A unique feature of our transcriptomic dataset is that in a single experiment, we were able to analyze MAIT cells from two different species obtained from two different tissues, using different contexts of activation, and yet we observed that the transcriptomic profiles of these MAIT cells were in fact very similar. Thus, the distinctive, common properties of MAIT cells predominate over differences between these cells that might be observed in different contexts. Again, this underlines a strong conservation of functions likely driven by a consistent role in mucosal immunology.

23Nonetheless, we also observed some differences between the human and murine datasets. In particular, the Reactome pathway analysis showed a very prominent gene signature for cell cycle and mitotic pathways not observed in the parallel human pathway analysis. This is unlikely to be due to differences in species, but rather to the different context of cell activation. Several previous studies have shown that to achieve maximal, sustained MAIT cell activation requires cytokine signals in addition to TCR triggering. These signals have been shown to include IL-12, IL-18, IL-15, and type 1 interferons, both in vitro and in vivo. While IL-12 and IL-18 are particularly important for inducing cytokine production from MAIT cells, the induction of MAIT cell proliferation appears to be more dependent on IL-15. In the current study, such synergistic cytokine signals would have been present in the in vivo TLR/5-OP-RU or Legionella infection models, but not in the human in vitro stimulation. In addition, the time courses of these two models were very different: 6 h for the in vitro stimulation, compared with 7 days in vivo, allowing much greater time for cell proliferation. It is therefore more striking that despite these two important differences, the consistent tissue repair signature was apparent in both models.

24Given the wide diversity of conventional and non-classical T cells now recognized, many of which share common transcriptional programs, we applied a comparative approach to analyze the phenotype of MAIT cells, overcoming significant methodological hurdles to compare our RNA sequencing data directly with older microarray expression data in the ImmGen dataset. Activated MAIT cells were most similar to activated invariant iNKT cells, as might be expected from the similarities in surface markers and functional phenotype. This is likely related to shared transcriptional signatures controlled by common transcription factors, not least that that has been described for promyelocytic leukemia zinc finger (PLZF), which defines a distinct surface phenotype and functional capacity in CD161+ NK cells, iNKT cells, and MAIT cells. However, it is interesting that in their resting state, MAIT cells more closely resembled splenic γδ T cells. While MAIT cells differed from γδ T cells by higher expression of specific cell surface receptors—notably the IL-18 receptor, recognized as a classic MAIT cell marker and the vitamin D receptor (Vdr), consistent with their postulated modulation by vitamin D—resting MAIT cells differed more from iNKT cells. These differences were dominated by upregulation of pathways associated with degranulation and pathways associated with cell surface interactions with vasculature, suggesting that after the resolution of infection, MAIT cells may retain more capacity for transmigration and for cytotoxic degranulation than iNKT cells, although this comparison may also have been influenced by the different tissue origins of our MAIT cell data (pulmonary) and the iNKT cell data from ImmGen (spleen and liver). Unlike MAIT and iNKT cells, most γδ T cells are not constrained by a specific MHC restriction; rather, they have different functional profiles associated with the usage of different TCR V gene segments. Depending on the Vγ subset, γδ T cells recognize a diverse range of small microbial metabolites, lipids, self-antigens, and stress-induced proteins and may display a range of functions associated with inflammation, immunoregulation, cytotoxicity, antigen presentation, and promotion of tissue repair. In the absence of the TCR-/TLR-mediated activation, MAIT cells may be fulfilling a different, perhaps homeostatic, function. Indeed, we were able to investigate what this might be by analyzing the transcriptome of MAIT cells in their resting state, outside the context of inflammation. After the resolution of infection, Tnfsf18 (GITRL) is the most strongly upregulated cytokine. The function of GITRL is context dependent, but under resting, non-inflammatory conditions it can negatively regulate NK cells and maintain or expand regulatory T cells’ conditions. Other immunoregulatory cytokines were also upregulated: Tnfsf11 (TRANCE, RANKL) was identified in a commensal-derived immunoregulatory signature, while IL-15 can inhibit T cell apoptosis to maintain memory T cell survival. Together, these data implicate resting MAIT cells in potentially significant immunoregulatory roles.

25In summary, our analysis of TCR-activated MAIT cells demonstrates a pronounced conservation of functions and gene expression profiles between human and murine cells and suggests that beyond type 17/type 1 pro-inflammatory responses to invading microbial pathogens, MAIT cells have the capacity to contribute to immunoregulatory and tissue repair roles likely to be essential for maintaining the integrity of mucosal barrier surfaces in health and disease.

26Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Timothy Hinks (timothy.hinks@ndm.ox.ac.uk).

27C57BL/6JArc mice were bred and housed in the Biological Research Facility of the Peter Doherty Institute (Melbourne, Victoria, Australia). Male mice aged 6–12 weeks, housed under specific pathogen-free conditions were used in experiments, after approval by the University of Melbourne Animal Ethics Committee (1513661 and 1513712). Animals were group housed, and littermates within the same cage were randomly allocated to receive stimuli or control.

28Human peripheral blood mononuclear cells (PBMC) were obtained from healthy volunteer participants (University of Melbourne Human Research Ethics Committee 2002.107). Subjects were adults (Median age 40, range 24-48) of both sexes (3 female, 1 male) who provided written informed consent. Groups of samples of each cell type and experimental condition (Table S2) included 1 male and 2 female samples in all cases. As sample groups were contrasted with each other, rather than within a group, no effect of sex on results will be apparent in the datasets or analyses provided, and hence results are expected to be generalisable to both sexes.

29Cultures of Legionella longbeachae NSW150 were grown at 37°C in buffered yeast extract (BYE) broth supplemented with 30-50 μg/ml streptomycin for 16 hours to log-phase (OD600 0.2-0.6) with shaking at 180 rpm. For the infecting inoculum, bacteria were re-inoculated in pre-warmed medium for a further 2–4 h culture (OD600 0.2–0.6) with the estimation that 1 OD600 = 5x108/ml, sufficient bacteria were washed and diluted in phosphate buffered saline (PBS) with 2% BYE for i.n. delivery to mice. A sample of inoculum was plated onto (BYCE) with streptomycin for verification of bacterial concentration by counting colony-forming units.

305-OP-RU was prepared as described previously. CpG1668 (Sequence: T∗C∗C∗A∗T∗G∗A∗C∗G∗T∗T∗C∗C∗T∗G∗A∗T∗G∗C∗T (∗phosphorothioate linkage) nonmethylated cytosine-guanosine oligonucleotides was purchased from Geneworks (Thebarton, Australia) and Pam2Cys was chemically synthesized and functionally verified in house. Other toll like receptor ligands are detailed in Table S1. Murine and human MR1 and β2-Microglobulin genes were expressed in Escherichia coli inclusion bodies, refolded, and purified as described previously. MR1-5-OP-RU tetramers were generated as described previously.

31Intranasal (i.n.) inoculation with a stimulatory MR1 ligand (76 pmol 5-OP-RU) or a non-activating MR1 ligand (76 pmol 6-FP) and TLR agonist (see Table S1) in a total 50 μL volume was performed on isofluorane-anaesthetised mice on day 0. Additional doses of the relevant MR1 ligand in 50 μL were administered on days 1, 2, and 4. Control mice received TLR agonists alone, in a total volume of 50 μl. For infection experiments mice were inoculated with 1-2 x104 CFU Legionella longbeachae (clinical isolate NSW150; ) in 50 μL PBS.

32Mice were weighed daily and assessed visually for signs of disease, including inactivity, ruffled fur, labored breathing, and huddling behavior. Animals that had lost ≥ 15% of their original body weight and/or displayed evidence of pneumonia were euthanised.

33Mice were killed by CO2 asphyxia, the heart perfused with 10 mL cold Roswell Park Memorial Media-1640 (RPMI, GIBCO) and lungs were taken. To prepare single-cell suspensions lungs were finely chopped with a scalpel blade and treated with 3 mg.ml-1 collagenase III (Worthington, Lakewood, NJ), 5 μg/ml DNase, and 2% fetal calf serum in RPMI for 90 min at 37°C with gentle shaking, and, where relevant, brefeldin A (GolgiPlug, BD Biosciences, San Diego, CA). Lung cells were then filtered (70 μm) and washed with PBS/2% fetal calf serum. Red blood cells were lysed with hypotonic buffer TAC (Tris-based amino chloride) for 5 min at 37°C. Approximately 1.5x106 cells were filtered (40 μm) and used for flow cytometric analysis. To obtain sufficient cells for sorting naive CD8+CD44-CD62L+ cells from infection-naive mice lungs from 2-3 mice per sample were pooled and stained with 0.18 μL anti-CD8-PE, then magnetically enriched using anti-PE beads (Miltenyi) prior to sorting.

34For analysis of systemic MAIT cell distribution lymphocytes were obtained from mesenteric lymph nodes by passing through a 70 μm strainer. Splenocytes were obtained by homogenizing splenic tissue through a 70 μm strainer then preforming red cell lysis prior to staining. Peripheral blood cells were obtained from the inferior vena cava into a heparinised syringe and underwent surface staining prior to red cell lysis with 1 mL of 10% red cell lysis buffer (BD Bioscience) for 5 minutes at room temperature before washing twice with FACS buffer. Hepatic lymphocytes were obtained by perfusing the liver with 8-10 mL PBS, passing through a 40 μm strainer, washing once with PBS the resuspending in 36% Percoll (Sigma) and centrifuging without braking at 800 g for 25 mins at RT over a 70% Percoll underlay. Cells from the interphase were washed with FACS buffer, red cells lysed with 1 mL 10% red cell lysis buffer, cells washed twice and stained for flow cytometry.

35Bacterial infection was determined for L. longbeachae by counting colony-forming units (CFU) obtained from plating homogenized lungs in duplicate from infected mice (x5 per group) on buffered charcoal yeast extract agar (BYCE) containing 30 μg/ml streptomycin and colonies counted after 4 days at 37°C under aerobic conditions.

36Details of flow cytometry antibodies are shown in the Key Resources Table. To block non-specific staining, cells were incubated with MR1-6-FP tetramer and anti-Fc receptor (2.4G2) for 15 min at room temperature and then incubated at room temperature with Ab/tetramer cocktails in PBS/2% fetal calf serum. Dead cells were excluded using 4′,6-diamidino-2-phenylindole (DAPI) for live cell sorting added for 10 mins or by staining for 20 mins in PBS with fixable viability dyes Zombie Yellow (Biolegend, 1:100, 423104) or Live/Dead EF780 (BD Bioscience, 1:1000, 565388).

37For live cell sorting on human peripheral blood mononuclear cells (PBMC) 50ml of heparinised blood were obtained freshly per volunteer, mixed with an equal volume of phosphate buffered saline (PBS) and layered over an equal volume of Ficoll-Paque (GE Healthcare, Chicago, IL) and centrifuged at 800 g for 20 mins at room temperature. Cells were washed twice with PBS, cells counted by trypan blue estimation, then half the cells were resuspended overnight in RPMI with 10% human serum for overnight rest and the other half resuspended in flow cytometry buffer comprising PBS with 2% fetal calf serum and 2 mM EDTA (FACS buffer) for immediate magnetic enrichment and sorting. These cells were stained with surface antibodies (CD3-PE-CF594, CD8-PerCPCy5.5, CD45RA-FITC, TCR Vα7.2-APC and FCγ block for 15 mins at RT, followed by staining with MR1-5-OP-RU tetramer-PE for 20 mins at RT. Tetramer positive cells were positively selected using anti-PE microbeads (10 μL per 10−7 cells, Miltenyi, Cologne, Germany) according to the manufacturer’s instructions. Cells were sorted immediately into ice cold PBS with 10% FCS using an FACSAria III cell sorter (BD Bioscience) selecting live CD3+TCR-Vα7.2+MR1-5OP-RU-tetramer+ MAIT cells from the positive fraction and live CD3+CD8+CD45RA+MR1-Tetramer- cells from the negative fraction (Figure S7A). The following day the remaining cells were stimulated for 6 h with 10 nM 5-OP-RU then magnetically enriched using MR1-5-OP-RU-tetramer-PE and anti-PE microbeads, and live CD3+TCR-Vα7.2+MR1-5-OP-RU-tetramer+ MAIT cells sorted in the same manner. Purity was checked and with an average of 98%. Immediately after sorting cells were centrifuged at 400 g for 5 mins then resuspended in 100 μL of RNA lysis buffer (Agilent Ltd, UK) with 0.7 μL β-mercaptoethanol and stored at −80°C.

38For live cell sorting of murine T cells, CD8 cells were magnetically enriched using anti-CD8-PE and anti-PE microbeads and live CD8+CD44-CD62L+ cells from uninfected mice, or live CD3+CD19-CD45.2+TCRβ+MR1-5-OP-RU-tetramer+ MAIT cells from previously infected mice were sorted as above (Figure S7B).

39For intracellular staining, cells were fixed with 1% paraformaldehyde prior to analysis on LSRII or LSR Fortessa or Canto II (BD Biosciences) flow cytometers. For intracellular cytokine staining Golgi plug (BD Biosciences) was used during all processing steps. Cells stimulated with PMA (phorbol 12-myristate 13-acetate;)/ionomycin (20 ng ml-1, 1 μg ml-1, respectively) for 3 h at 37°C were included as positive controls. Surface staining was performed at 37°C, and cells were stained for intracellular cytokines using the BD Fixation/Permeabilization Kit (BD, Franklin Lakes, NJ) or transcription factors using the transcription buffer staining set (eBioscience) according to the manufacturers’ instructions.

40For validation of key targets identified by RNA sequencing flow cytometry was performed on cryopreserved human PBMC from additional healthy human donors. Samples were defrosted into pre-warmed RPMI with 10% human serum, stained with anti-TCR-Vα7.2-PE or anti-TCR-Vα7.2-PE and magnetically enriched using anti-PE or anti-APC microbeads. 200,000 positively-selected TCR-Vα7.2+ cells or the negative fraction (for naive CD8+45RA+ cells) were co-cultured for 5 hours in the presence of brefeldin A with 100,000 class I reduced (C1R) antigen presenting cells (APCs) which had been previously pulsed for 2 hours with 10 nM 5-OP-RU, or with naive C1R cells (unstimulated control), or with PMA / ionomycin (20 ng ml-1, 1 μg ml-1, respectively), or without any stimulation. Cells were then analyzed by surface and intracellular cytokine staining as above. For validation of murine targets cells were isolated from uninfected mice or mice which had undergone intranasal infection 7 days prior (acute) or 12 weeks prior (resolution) or reinfection 7 days prior, and cytometrically analyzed as described above.

41Cells were lysed in Agilent lysis buffer (Agilent Ltd., UK) containing 100 mM β-mercaptoethanol and passed through a QIAshredder device (QIAGEN, Valencia, US), then RNA extracted using the Absolutely RNA Nanoprep Kit according to the manufacturer’s instructions, including using of DNase I. RNA libraries were prepared at the Melbourne Translational Genomics Platform, Department of Pathology (The University of Melbourne). Briefly, RNA quality and quantity were assessed using the Bioanalyzer 2100 RNA pico kit (Agilent technologies). The input total RNA was normalized to 250 pg per sample and median RIN was 9.7 (range 5.7-10.0). RNA was reverse transcribed and cDNA amplified by in vitro transcription with the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Clontech). First strand cDNA synthesis and tailing by reverse transcription was performed using Clontech’s proprietary SMART (Switching Mechanism at 5′ End of RNA Template) technology. Following first strand synthesis, cDNA was amplified 12 cycles by LD PCR using blocked PCR primers. Amplified cDNA was purified using AMPure XP prior to QC using the bioanalyser 2100 HS DNA kit (Agilent technologies). Library preparation of purified amplified cDNA was performed using Nextera XT library preparation (Illumina, AUS). Following QC, 150 pg of cDNA was tagmented (simultaneously fragmented with adaptors inserted) using Nextera transposons. Molecular barcodes were incorporated during 12 cycles of amplification followed by purification using AMPure XP. The libraries passed a quality checkpoint (Qubit and Bioanalyser HS DNA) prior to normalization and pooling before loading onto the HiSeq 2500 (Illumina, AUS) for paired end sequencing.

42RNA-seq reads were aligned to reference genome sequences using STAR aligner software. Mapped reads were assigned to genomic features using Rsubread R package.

43Genes that were differentially expressed (> 2 fold, p < 0.01, FDR < 0.05) between conditions and their normalized expression values, were generated with EdgeR R package, Partek® Flow®, an online analysis platform for Next Generation Sequencing data (http://www.partek.com/partek-flow/). Pathway enrichment analysis using the Reactome platform was performed using ReactomePA R package. Gene count data transform to log2-counts per million (logCPM) was performed using voom function in limma R package. Gene set enrichment analysis (GSEA) using voom transformed count data was performed using GSEA version 3.0, comparing gene expression data as a whole with the reference gene list obtained from the publication by

44RNA-seq data (logCPM) and ImmGen microarray data were integrated using a common set of Entrez annotated genes; batch effect removal was performed using ComBat algorithm in sva R package. Hierarchical clustering analysis of transcription profiles was conducted in R employing highly variable genes (IQR > 0.75; ) and Euclidian distance.

45Statistical tests were performed using the Prism GraphPad software (version 7.0 La Jolla, CA). Normality of flow-cytometric data was confirmed for the majority of data columns using Shapiro-Wilk tests and comparisons between groups were performed using Student’s t tests on untransformed (MAIT cell frequency) or log-transformed (absolute MAIT cell numbers) data as appropriate. Flow cytometric data analysis was performed with FlowJo10 software (Ashland, OR). Statistical analyses and samples sizes are specified in each figure legend. For standard statistical analyses a two-tailed P value (where relevant, adjusted for multiple comparisons) < 0.05 was considered statistically significant.

46The RNA Sequencing data have been deposited in the Gene Expression Omnibus (GEO) under accession number GSE123805.