Report

Microbial metabolites control the thymic development of mucosal-associated invariant T cells

See allHide authors and affiliations

Science  25 Oct 2019:
Vol. 366, Issue 6464, pp. 494-499
DOI: 10.1126/science.aaw2719

Commensals rule the MAITrix

Mucosal-associated invariant T (MAIT) cells play an important role in mucosal homeostasis. MAIT cells recognize microbial small molecules presented by the major histocompatibility complex class Ib molecule MR1. MAIT cells are absent in germ-free mice, and the mechanisms by which microbiota control MAIT cell development are unknown (see the Perspective by Oh and Unutmaz). Legoux et al. show that, in mice, development of MAIT cells within the thymus is governed by the bacterial product 5-(2-oxopropylideneamino)-6-d-ribitylaminouracil, which rapidly traffics from the mucosa to the thymus, where it is captured by MR1 and presented to developing MAIT cells. Constantinides et al. report that MAIT cell induction only occurs during a limited, early-life window and requires exposure to defined microbes that produce riboflavin derivatives. Continual interactions between MAIT cells and commensals in the skin modulates tissue repair functions. Together, these papers highlight how the microbiota can direct immune cell development and subsequent function at mucosal sites by secreting compounds that act like self-antigens.

Science, this issue p. 494, p. eaax6624; see also p. 419

Abstract

How the microbiota modulate immune functions remains poorly understood. Mucosal-associated invariant T (MAIT) cells are implicated in mucosal homeostasis and absent in germ-free mice. Here, we show that commensal bacteria govern murine MAIT intrathymic development, as MAIT cells did not recirculate to the thymus. MAIT development required RibD expression in bacteria, indicating that production of the MAIT antigen 5-(2-oxopropylideneamino)-6-d-ribitylaminouracil (5-OP-RU) was necessary. 5-OP-RU rapidly traveled from mucosal surfaces to the thymus, where it was captured by the major histocompatibility complex class Ib molecule MR1. This led to increased numbers of the earliest MAIT precursors and the expansion of more mature receptor-related, orphan receptor γt–positive MAIT cells. Thus, a microbiota-derived metabolite controls the development of mucosally targeted T cells in a process blurring the distinction between exogenous antigens and self-antigens.

Mucosal-associated invariant T (MAIT) cells are evolutionarily conserved T cells that recognize vitamin B2 precursor derivatives [5-(2-oxopropylideneamino)-6-d-ribitylaminouracil (5-OP-RU)] presented by the major histocompatibility complex class Ib molecule MR1 (1). These derivatives are produced by most bacteria and yeasts but not by animal cells (2). MAIT cells have been implicated in human pathologies associated with microbiotal dysbiosis (36). In mice, mature, receptor-related orphan receptor γt–positive (RORγt+) MAIT cells play beneficial roles in maintaining intestinal homeostasis (5, 7). Germ-free (GF) mice lack MAIT cells (8, 9), but the underlying mechanisms are unknown. Here, we report that vitamin B2 metabolites are directly transferred into the thymus and presented to thymocytes, driving the intrathymic expansion of RORγt+ MAIT cells. Thus, commensal bacteria influence barrier homeostasis by controlling the thymic production of T cells targeted to mucosae.

To investigate the role of the microbiota in MAIT development, we compared thymic MAIT frequencies of mice under specific pathogen–free (SPF) and GF conditions (10). As recently described (8), the number of MR1:5-OP-RU tetramer+ (MAIT) cells was reduced in the thymus (Fig. 1A), as well as in the spleen (Fig. 1B), lungs (Fig. 1C), and colon (fig. S1A), of GF mice. By contrast, invariant natural killer T (iNKT) cells, a T cell subset with different specificity but parallel development (11, 12), were not reduced in GF mice (fig. S1B) (13). We then characterized the developmental defect of MAIT cells as described previously (8). HSA+CD44 (Heat-stable antigen–positive, CD44-negative) immature MAIT cells (8) were reduced in GF mice (Fig. 1, D and E). A few cells reached the HSACD44+ most mature stage (Fig. 1, D and E) and expressed the lineage-defining transcription factor PLZF (promyelocytic leukemia zinc finger) at normal levels (fig. S1C). These MAIT cells may be selected by empty MR1 or by MR1 loaded with unknown self-ligands. We compared the sublineage choice of mature MAIT cells in SPF and GF mice using CD122 and CD138 to identify the T-bet+ MAIT1 and RORγt+ MAIT17 subsets, respectively (fig. S1D). MAIT17 cells were decreased in GF mice in both the thymus (Fig. 1, F and G, and fig. S1F) and spleen (fig. S1E). Accordingly, MAIT cells in GF mice were skewed toward interferon (IFN)-γ production with reduced interleukin (IL)-17 production (Fig. 1H).

Fig. 1 Rapid development of MAIT17 cells upon colonization with commensal microbes.

(A to C) Top: Representative flow cytometry of thymocytes (A), splenocytes (B), and lung cells (C) from SPF and GF B6 mice. Data are shown as mean ± SEM; N = 3 independent experiments; n ≥ 1 biological replicates per experiment. exGF, GF mice co-housed with SPF mice. Bottom: MAIT cell numbers in thymus (A) or MAIT cell frequencies in spleen (B) and lungs (C) of the indicated mice. Dotted lines represent the limit of detection as defined by the mean frequency of 6-FP:MR1 tetramer+ events. (D) Representative HSA and CD44 expression in MAIT thymocytes. Data are shown as mean ± SEM; N = 4 experiments, n ≥ 1 replicates. (E) Numbers of HSA+CD44 and HSACD44+ MAIT thymocytes in the indicated mice. (F) Representative CD122 and CD138 expression in HSACD44+ MAIT thymocytes. Data are shown as mean ± SEM; N = 3 experiments, n ≥ 1 replicates. (G) Numbers of MAIT1 and MAIT17 thymocytes in the indicated mice. (H) IL-17 and IFN-γ expression in HSACD44+ MAIT thymocytes after phorbol 12-myristate 13-acetate and ionomycin stimulation. A total of eight to nine thymi were pooled together; N = 2 experiments, n = 1 replicate. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 by Mann–Whitney U tests.

To directly assess the role of microbiota on MAIT development, adult GF mice were cohoused with SPF mice. Upon microbial colonization, MAIT frequencies increased first in the thymus, then in the spleen and lungs (Fig. 1, A to C), without reaching in the periphery the levels of mice colonized at birth. In the thymus, numbers of both immature HSA+ and mature CD44+ MAIT cells increased and reached SPF levels 2 weeks after microbial colonization (Fig. 1D). The increase in mature thymic MAIT numbers was accounted for by the expansion of MAIT17 cells, whereas MAIT1 numbers remained unchanged (Fig. 1E). Thus, commensal microbes are necessary for the complete thymic maturation of MAIT17 cells.

MAIT cells may expand in the periphery and circulate back to the thymus, thereby explaining the reduced frequency of mature MAIT cells in GF thymi. We used parabiosis of B6-MAITCast mice, which exhibit increased frequency of MAIT cells compared with C57BL/6 mice (fig. S2A) (14), to address this possibility (fig. S2B). Only 2.5 ± 1% (SEM) of mature thymic MAIT cells originated from the other parabiont, indicating minimal migration from the periphery (Fig. 2A). Few MAIT cells also migrated into Mr1−/− parabionts (fig. S2C). Accordingly, MAIT17 cells were highly proliferative (Ki67+) within the thymus but not in the spleen (Fig. 2B). Finally, MAIT frequency was unaffected in the thymus, spleen, and lungs of B6-MAITCast Jh−/− mice (fig. S2, D and E), which lack B cells. Thus, B cells are not required for MAIT development.

Fig. 2 MAIT17 cells receive TCR signals and proliferate in the thymus.

(A) Representative flow cytometry of MAIT cells in the blood (left) and thymus (center) of CD45.1/2 and CD45.2/2 congenic B6-MAITCast parabiotic mice. Data are shown as mean ± SEM; N = 3 independent experiments, n = 2 biological replicates per experiment. Right: Chimerism of MAIT cells in the indicated tissues. (B) Top: Representative flow cytometry of HSACD44+ MAIT cells from thymus and spleen. Data are shown as mean ± SEM; N = 4 experiments, n ≥ 1 replicates. Bottom: Proportion of Ki67+ cells in the indicated MAIT subset. (C) Top: Representative flow cytometry of HSACD44+ MAIT cells from thymus and spleen; N = 3 experiments, n ≥ 1 replicates. Bottom: Mean fluorescence intensity (MFI) of Nur77 staining in the indicated MAIT subset relative to CD4CD8+ cells from the same sample. (D) Top: Representative flow cytometry of HSACD44+ MAIT cells from thymus of the indicated mice. Data are shown as mean ± SEM; N = 3 experiments, n ≥ 1 replicates. Bottom: Number of MAIT1 and MAIT17 cells in the thymus or spleen of the indicated mice. (E to G) Left: Representative flow cytometry of the indicated MAIT subset from the thymus of SPF or GF mice; N ≥ 2 experiments, n ≥ 2 replicates. Right: Proportion of Ki67+ MAIT17 cells (E), MFI of Nur77 staining relative to CD4CD8+ cells (F), and proportion of TCR Vβ8+ MAIT cells (G) in the thymus of SPF or GF mice. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 by Student’s t tests.

Given the similarities in the thymic differentiation process of iNKT and MAIT cells (12, 15, 16), their discordant dependency on microbiota suggested a role for T cell receptor (TCR) specificity and antigen (Ag) availability in the thymus. Therefore, we measured expression levels of Nur77, a transcription factor correlated with TCR signaling strength (17). Nur77 was expressed in MAIT17 cells in the thymus but not in the spleen or in the MAIT1 cells of either organ (Fig. 2C). Thus, MAIT17 cells receive strong TCR signals in the thymus. We then investigated whether the phosphatidylinositol 3-kinase (PI3K) pathway, which controls proliferation in lymphocytes (18), was involved in MAIT development. Deletion of phosphatase and tensin homolog (PTEN), a negative regulator of PI3K activity, had no effect on the number of MAIT1 cells, yet caused a 53-fold increase in MAIT17 numbers in the thymus but not the spleen (Fig. 2D). This suggested that the PI3K pathway was active in thymic MAIT17 cells. The discrepancy between MAIT1 and MAIT17 behaviors may result from the expression of several inhibitory receptors, such as the TCR inhibitor Ptpn22, by MAIT1 cells (fig. S2F) (12). The ability of both MAIT subsets to signal through the TCR was evaluated in vitro. Although the majority of MAIT17 cells expressed Nur77 in response to 5-OP-RU, only a fraction of MAIT1 cells did (fig. S2G), indicating that TCR signaling was inhibited in this subset. Thus, only MAIT17 cells receive strong TCR signals and proliferate in a PI3K-dependent manner in the thymus.

In GF mice, MAIT17 cells expressed lower levels of Ki67 (Fig. 2E), as well as lower levels of Nur77 (Fig. 2F), indicative of reduced TCR signaling. Nur77 expression was lower in immature HSA+ cells from GF mice, which indicates that the microbiota contributed to TCR stimulation in these cells as well. We recently reported increasing TCR Vβ8 expression in MAIT cells during thymic maturation (19). In GF mice, Vβ8 expression did not increase with MAIT maturation (Fig. 2G), consistent with a role for microbiota-derived Ags in shaping the thymic development of MAIT cells.

The ability of a bacterial strain to trigger MAIT TCRs strictly correlates with the ability to produce an intermediary compound of the vitamin B2 biosynthesis pathway, 5-amino-ribityl uracil (5-A-RU), which reacts with methylglyoxal to generate 5-OP-RU (2022). Because microbial metabolites may circulate in the body (23, 24), and colonization with commensal microbes results in the expansion of Nur77hi MAIT cells in the thymus, we investigated whether 5-OP-RU could be captured and presented by thymic cells. In vitro incubation with synthetic 5-OP-RU, but not 5-A-RU, resulted in the up-regulation of MR1 at the cell surface of thymocytes, indicating efficient capture by intracellular MR1 (fig. S3A). Using an in vitro assay, we compared the ability of cells from different tissues to capture and present synthetic 5-OP-RU. Thymic cells were 50 times more efficient than splenocytes at activating 5-OP-RU:MR1–specific TCR-Tg reporter cells in vitro (Fig. 3A and fig. S3B). Thymic cells from GF mice also captured and presented synthetic 5-OP-RU to reporter cells in vitro (fig. S3C). By contrast, thymocytes were no better than splenocytes at capturing and presenting the iNKT ligand α-galactosylceramide (αGalCer) to αGalCer:CD1d–specific TCR Tg cells (fig. S3D).

Fig. 3 Vitamin B2 metabolite 5-OP-RU can cross mucosal barriers and reach the thymus for presentation to MAIT cells.

(A) Relative activation of reporter splenocytes incubated with total cells from the indicated tissues pulsed with 5-OP-RU. Data are shown as mean ± SEM; N = 3 independent experiments, n = 1 biological replicate per experiment. (B) Activation of reporter splenocytes incubated with total cells from the indicated tissues harvested 3 hours after intraperitoneal injection of 5-OP-RU. Data are shown as mean ± SEM; N = 2 experiments, n = 1 replicate. (C) Left: Representative flow cytometry of MAIT17 thymocytes 24 hours after intraperitoneal injection of the indicated molecule (5 nmol); N = 2 experiments, n ≥ 2 replicates. Right: Proportion of Ki67+ MAIT17 cells in the indicated condition. PBS, phosphate-buffered saline. (D) Activation of reporter splenocytes incubated with thymic APCs isolated by fluorescence-activated cell sorting and pulsed with 5-OP-RU. Data are shown as mean ± SEM; N = 2 experiments. DC, dendritic cell; mTEC, medullary thymic epithelial cell; cTEC, cortical thymic epithelial cell. (E) Left: Representative flow cytometry of reporter splenocytes incubated with thymic APCs from Mr1−/− or Mr1+ mice previously injected with 5-OP-RU (5 nmol). Data are shown as mean ± SEM; N = 3 experiments, n ≥ 1 replicates. Right: Activation of reporter splenocytes incubated with the indicated APCs. (F) Representative flow cytometry of reporter splenocytes incubated with total thymocytes from mice to which αGalCer (5 nmol) or 5-OP-RU (5 nmol) was painted on intact skin 3 hours before tissue harvest. As a positive control, thymocytes were incubated with 5-OP-RU (1 μM). Data are shown as mean ± SEM; N = 3 experiments, n ≥ 1 replicates. (G) Summary of the data obtained as in (F). (H) Activation of reporter splenocytes incubated with thymocytes or splenocytes from mice having received 5-OP-RU (5 nmol) either on the skin (left panel) or through oral gavage (right panel) for the indicated times before tissue harvest; N = 2 experiments, n ≥ 1 replicates. **p < 0.01, ***p < 0.001, ****p < 0.0001 by Student’s t tests.

To study the in vivo dynamics of 5-OP-RU, we injected decreasing doses of 5-OP-RU (or αGalCer control) intraperitoneally into wild-type or Mr1−/− mice. In mice administered with αGalCer, cells from the spleen and liver, but not the thymus, activated αGalCer:CD1d–specific reporter cells in vitro (fig. S3D). Thus, thymocytes did not capture circulating αGalCer, confirming the current view of the thymus being hermetic to circulatory Ags. By contrast, in mice receiving 5-OP-RU, thymic cells were more efficient at activating 5-OP-RU:MR1–specific reporter cells in vitro than were cells isolated from the periphery (Fig. 3B). Thus, systemic 5-OP-RU was efficiently captured and presented by thymic cells. The presentation of 5-OP-RU was MR1 dependent as thymic cells from Mr1−/− mice did not activate reporter cells. Neither 5-A-RU nor the nonstimulatory MR1 ligand acetyl-6-formylpterin (Ac6FP) activated reporter cells after injection (fig. S3E). Accordingly, injection of 5-OP-RU, but not 5-A-RU, triggered Ki67 expression in endogenous MAIT17 cells (Fig. 3C). Single-positive thymocytes, double-positive (DP) thymocytes, dendritic cells, and medullary and cortical thymic epithelial cells isolated as outlined in fig. S3F were all able to activate reporter cells upon in vitro incubation with 5-OP-RU. However, DP thymocytes were the most potent antigen-presenting cells (APCs) (Fig. 3D). After 5-OP-RU injection, DP thymocytes, dendritic cells, and epithelial cells all captured and presented 5-OP-RU in an MR1-dependent fashion (Fig. 3E).

We then investigated whether 5-A-RU or 5-OP-RU could reach the thymus from outside of the body. When painted onto the intact skin of mice, no αGalCer was detected in any tissue (fig. S3G). By contrast, only thymic cells from mice painted with 5-OP-RU, but not 5-A-RU, activated reporter cells (Fig. 3, F and G). Thus, some 5-OP-RU from the skin was captured in the thymus but not by other sampled tissues. The cutaneous application of 5-OP-RU, but not 5-A-RU, also activated endogenous thymic MAIT cells, as well as splenic MAIT cells to a lesser extent (fig. S3H). 5-OP-RU was detected in the thymus 1 hour after cutaneous application or after oral gavage (Fig. 3H). Thus, exogenous 5-OP-RU rapidly reaches the thymus and is presented to MAIT thymocytes by various APCs.

MAIT cell development may be directly dependent upon the production of 5-OP-RU by commensal bacteria. To control for microbial factors not associated with 5-OP-RU production, we used Escherichia coli strains with genetic deletion of vitamin B2 enzymes either upstream (ΔRibD) or downstream (ΔRibE) of 5-A-RU production (22) (Fig. 4A).

Fig. 4 5-OP-RU produced by commensal bacteria controls MAIT cell development in the thymus.

(A) Left: Representative flow cytometry of reporter splenocytes incubated with thymocytes previously cocultured with 107 colony-forming units/ml of the indicated strain of E. coli. Data are shown as mean ± SEM; N = 2 independent experiments, n = 2 biological replicates per experiment. Right: Summary of the data. (B) Left: Representative flow cytometry of cells from the thymus (top) and lungs (bottom) of the indicated mice; N ≥ 2 experiments, n ≥ 2 replicates. Right: MAIT cell numbers in the thymus (top) and MAIT frequencies in the lungs (bottom) of the indicated mice. (C) Left: Representative HSA and CD44 expression in MAIT thymocytes. Data are shown as mean ± SEM; N ≥ 2 experiments, n ≥ 2 replicates. Right: Numbers of HSA+CD44 and HSACD44+ MAIT thymocytes in the indicated mice. (D) Left: Representative CD122 and CD138 expression in HSACD44+ MAIT thymocytes in the indicated mice. Data are shown as mean ± SEM; N ≥ 2 experiments, n ≥ 2 replicates. Right: Numbers of MAIT1 and MAIT17 cells in the indicated mice. (E) Left: Representative flow cytometry of MAIT thymocytes in mice injected with PBS or with 1 nmol of 5-OP-RU for 2 weeks. Data are shown as mean ± SEM; N = 2 experiments, n ≥ 1 replicates. Right: Numbers of HSACD44+ MAIT thymocytes in the indicated conditions. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 by Student’s t tests (A) or by Mann–Whitney U tests (B to E).

GF mice were monocolonized with either ΔRibD or ΔRibE E. coli. MAIT frequency remained low in thymus and lungs of mice colonized with ΔRibD bacteria. By contrast, colonization with ΔRibE bacteria induced MAIT development and migration into lungs (Fig. 4B). 5-OP-RU production by bacteria was required for thymic expansion of MAIT17 cells (Fig. 4, C and D). 5-OP-RU production also controlled increased numbers of immature HSA+ MAIT cells (Fig. 4C). This may reflect either positive selection of new MAIT cells or the expansion of preselected HSA+ MAIT cells. These results were confirmed with the wild-type commensal bacterial species Enterococcus hirae (vitamin B2 proficient) and Enterococcus faecalis (vitamin B2 deficient) (25) (fig. S4, A to E). Thus, 5-OP-RU production by commensal bacteria is required for the complete maturation of thymic MAIT cells.

To test whether 5-A-RU-derived Ags are sufficient to induce MAIT development, SPF and GF mice were injected with various doses of 5-OP-RU. These injections induced a dose-dependent depletion of thymic MAIT cells in GF mice (Fig. 4E), likely a result of negative selection. MAIT cells were also depleted from the thymus of SPF mice but required higher doses of 5-OP-RU, suggesting that additional factors induced by the microbiota promoted MAIT survival. MAIT development was not impaired in Myd88−/− or Tlr3−/− mice (fig. S4F), ruling out a role for Toll-like receptor–dependent microbial sensing. Because myeloid differentiation factor 88 is necessary for signaling through IL-1 receptor family members, IL-1, IL-18, and IL-33 are therefore dispensable for MAIT development. Aryl hydrocarbon receptor, IFN-γ, and IFN-γ receptor were also dispensable for MAIT development (fig. S4F), suggesting the involvement of other microbiota-induced mediators in MAIT expansion and maintenance.

To test whether 5-OP-RU, together with an unknown bacterially induced factor(s), controlled MAIT maturation, low doses of 5-OP-RU (10 pmol) were injected into GF mice (Fig. 4E) and GF mice monocolonized with ΔRibD bacteria (Fig. 4, B to D). Injections of 5-OP-RU resulted in increased MAIT numbers in the thymus (Fig. 4B), in association with increased HSA+ and CD44+ MAIT numbers (Fig. 4C). The injection of Ac6FP or 5-A-RU had no effect on MAIT cells (Fig. 4B), indicating that methylglyoxal from the mouse cannot react in vivo with 5-A-RU to form agonist ligands. Thus, bacterial metabolites such as 5-OP-RU drive the thymic expansion of MAIT17 cells.

The results of this study help to explain how the microbiota can influence the host at distant sites through the production of metabolites.

Supplementary Materials

science.sciencemag.org/content/366/6464/494/suppl/DC1

Materials and Methods

Figs. S1 to S4

Table S1

References (2632)

References and Notes

Acknowledgments: We thank V. Dangles-Marie, M. Garcia, I. Grandjean, the mouse facility technicians, and the flow cytometry core at Institut Curie. We also acknowledge G. Eberl, S. Latour, and A. Lehuen for mice; P. Serror for bacteria; and N. Manel and S. Amigorena for discussions and for reviewing the manuscript. We thank the NIH tetramer core facility (Emory University) for providing CD1d and MR1 tetramers. The MR1:5-OP-RU tetramer technology was developed jointly by J. McCluskey, J. Rossjohn, and D. Fairlie, and the material was produced by the NIH Tetramer Core Facility as permitted to be distributed by the University of Melbourne. Funding: F.L. was supported by a Marie-Skłodowska Curie individual fellowship (706353) from the European Commission (H2020). This work was supported by the Institut National de la Santé et de la Recherche Médicale, Institut Curie, and Agence Nationale de la Recherche (ANR) [Blanc (neoMAIT, MAIT, and diabMAIT) and Labex DCBIOL]. O.L.'s group is supported by the Equipe Labellisée de la Ligue Contre le Cancer. Author contributions: F.L. designed the project and performed experiments, analyzed data, and wrote the paper. C.D. designed and performed experiments. Y.E.M. and A.D. performed experiments and analyzed data. D.B., E.P., A.F., and A.B. performed experiments. M.Sal., J.G., and S.R. provided technical support and advice. A.E.M., K.N., M.Sar., and F.S. developed and provided reagents. B.R. provided mice. O.L. supervised the project, designed experiments, and wrote the paper. Competing interests: The authors declare no competing interests. Data and materials availability: All experimental data are available in the main text or the supplementary materials.

Stay Connected to Science

Navigate This Article