Research ResourcesT CELLS

A divergent transcriptional landscape underpins the development and functional branching of MAIT cells

See allHide authors and affiliations

Science Immunology  22 Nov 2019:
Vol. 4, Issue 41, eaay6039
DOI: 10.1126/sciimmunol.aay6039

Charting MAIT cell development

Mucosal-associated invariant T (MAIT) cells are a distinct subset of T cells that recognize vitamin B metabolites presented by MHC class I–related protein (MR1). Here, using bulk and single-cell RNA- seq and flow cytometric analysis, Koay et al. have characterized the development of murine and human MAIT cells. In addition to providing us a clearer picture of MAIT cell development in the thymus, their studies have uncovered several molecules that play key roles in regulating this process. Using knockout mouse strains, they have confirmed the functions some of these molecules, including SAP (SLAM adaptor protein), SATB1, CXCR6, and CCR7 in regulating MAIT cell development.


MR1-restricted mucosal-associated invariant T (MAIT) cells play a unique role in the immune system. These cells develop intrathymically through a three-stage process, but the events that regulate this are largely unknown. Here, using bulk and single-cell RNA sequencing–based transcriptomic analysis in mice and humans, we studied the changing transcriptional landscape that accompanies transition through each stage. Many transcripts were sharply modulated during MAIT cell development, including SLAM (signaling lymphocytic activation molecule) family members, chemokine receptors, and transcription factors. We also demonstrate that stage 3 “mature” MAIT cells comprise distinct subpopulations including newly arrived transitional stage 3 cells, interferon-γ–producing MAIT1 cells and interleukin-17–producing MAIT17 cells. Moreover, the validity and importance of several transcripts detected in this study are directly demonstrated using specific mutant mice. For example, MAIT cell intrathymic maturation was found to be halted in SLAM-associated protein (SAP)–deficient and CXCR6-deficient mouse models, providing clear evidence for their role in modulating MAIT cell development. These data underpin a model that maps the changing transcriptional landscape and identifies key factors that regulate the process of MAIT cell differentiation, with many parallels between mice and humans.


Mucosal-associated invariant T (MAIT) cells are a population of unconventional T cells that detect vitamin B2 (riboflavin) metabolites presented by the monomorphic antigen-presenting molecule MR1 (1, 2). Most MAIT cells express a semi-invariant T cell receptor (TCR) comprising a TRAV1-2/TRAJ33 TCR-α chain, paired with a limited array of TCR-β chains (enriched for TRBV19 and TRBV13 in mice and TRBV6 and TRBV20 in humans). MAIT cells are highly abundant in humans, representing up to 10% of circulating T cells and up to 45% of T cells in some tissues such as liver (1). However, for reasons that are unclear, they are also highly variable between individuals. Although the consequences of low or high MAIT cells are uncertain in humans, in mice, many studies have demonstrated that MAIT cell deficiencies can be important in immunity to bacterial and viral infections (1, 3, 4) and in autoimmune and inflammatory diseases (1, 5). It is therefore critical that we understand how the development and homeostasis of MAIT cells are regulated.

MAIT cells are known to develop in the thymus through a three-stage process in mice and humans (6). In mice, the first stage is defined as CD24+CD44, the second stage is CD24CD44, and the third stage is CD24CD44+. In humans, the corresponding stages are defined as CD27CD161, CD27+CD161, and CD27+CD161+ cells (6). The progression through each stage is thought to depend on the interaction of developing MAIT cells with MR1 expressed on cortical thymocytes (7). MR1 is important not only for their initial selection toward the MAIT cell lineage but also for their subsequent maturation in the thymus (6). The progression from stage 1 to stage 2 also depends on the microRNA processing enzyme Drosha (6), microRNA miR-181a/b-1 (8), and the presence of undefined microbial factors because this process is impaired in germ-free mice (6). The progression from stage 2 to stage 3 is dependent on the transcription factor promyelocytic leukemia zinc finger (PLZF) and the cytokine interleukin-18 (IL-18) (6). Stage 3 MAIT cells are closest in phenotype to mature MAIT cells in the periphery, although further extrathymic maturation of these cells is required for them to reach a fully mature phenotype (6, 9). Thus, the transition from stage 2 to stage 3 represents key events in MAIT cell maturation where MAIT cells not only enter their final maturation stage but also diversify into functionally distinct subsets. A major subset of stage 3 cells in the thymus (MAIT17) produces IL-17 and expresses the transcription factor RORγt (retinoic acid receptor–related orphan receptor γt), whereas another subset (MAIT1) produces interferon-γ (IFN-γ) and expresses the transcription factor T-bet (6, 10). Thus, it seems that either within or just before stage 3 of MAIT cell development, there are some important differentiation events that contribute to the divergent MAIT cell populations, and it is likely that these cells do not all group into one transcriptionally homogeneous population.

In this study, we have carried out bulk and single-cell RNA-sequencing (RNA-seq) transcriptomic analysis of MAIT cells from both mice and humans at each stage of their development. These results have revealed many genes that are sharply up- or down-regulated as MAIT cells develop, and there are substantial commonalities between the two species. On the basis of this information, we have used a series of gene-deleted mice to identify factors that control the development of this unique arm of the immune system. Furthermore, we have also dissected stage 3 MAIT cells into transcriptionally distinct subpopulations, which has led us to construct a more detailed developmental model that reflects the functional bifurcation events that see MAIT cells branch into two distinct lineages. Collectively, our findings provide a major advance in our understanding of how MAIT cells develop and differentiate and the factors that control this process.


Differentially expressed genes associated with intrathymic mouse MAIT cell development

We previously described a three-stage intrathymic development pathway for MAIT cells in mice and humans. In mice, the pathway begins with CD24+CD44 stage 1 MAIT cells, which develop into CD24CD44 stage 2 MAIT cells and then into CD24CD44+ stage 3 MAIT cells (6) (Fig. 1A). To investigate the molecular mechanisms that underpin this process, we compared the transcriptomes of thymic immature stage 1 and mature stage 3 MAIT cells from mouse thymus using bulk (100 cells) RNA-seq. We identified 314 differentially expressed genes (DEGs) between stages 1 and 3, of which 196 were significantly up-regulated and 118 significantly down-regulated between stages 1 and 3 (Fig. 1B). A list of the top 50 DEGs is shown in (Fig. 1C), and a full list of these genes can be found in table S1. As a strong validation of the RNA-seq–based approach, DEGs included several markers that we had previously identified for each population (6); thus, stage 1 cells expressed higher Cd24a, Cd4 and Cd8b, Cd69, and Sell (CD62L), whereas stage 3 cells expressed higher Cd44, Icos, Il18r (CD218), Itgae (CD103), Il7r (CD127), Zbtb16 (PLZF), Klrb1c (NK1.1), and Rorc (RORγt). These data revealed many additional genes that were up-regulated [such as Cxcr6, Ccr6, Maf, Sdc1 (Syndecan 1, CD138), GzmB (Granzyme B), and Ly6e] or down-regulated (such as Slamf1, Slamf6, Tcf7, Satb1, Tox, Ccr7, Ccr9, and Lef1) as MAIT cells mature in the thymus (Fig. 1, B and C). To determine whether these transcriptional changes were associated with transitions between stage 1 and stage 2 or stage 2 and stage 3, we used single-cell RNA-seq to separately probe these populations because isolating sufficient numbers of stage 2 cells for bulk RNA-seq is quite rare and difficult. These data demonstrated that the transition from stage 1 to stage 2 is accompanied by 173 transcripts that increased and 91 genes that decreased (Fig. 1D). Most of the transcripts (70 of 184) that increased encode ribosomal proteins (Rps and Rpl), suggesting that this transition is associated with increased cellular activity and protein production, consistent with a recent study showing that stage 2 is linked to increased proliferation (8). Gene set testing performed on the DEGs using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases also highlighted the heavy enrichment of ribosomal activity pathways between stage 2 and stage 1 (fig. S2). In addition, the transition from stage 1 to stage 2 was associated with changes in some important immunological genes, including increased Zbtb16 (PLZF), Il18r (CD218), Il23r, and Ccr2 and decreased SlamF1, Satb1, Cd28, and Cd24a (already validated as the surface marker that separates stage 1 from stage 2 cells). A comparison of stage 2 with stage 3 shows that most of the changes identified in association with the bulk RNA-seq comparison of stage 1 and stage 3 (Fig. 1B) were associated with this transition.

Fig. 1 Transcriptomic analysis of mouse MAIT cell development.

(A) Flow cytometric analysis of three stages of mouse thymic MAIT cells after MR1-5-OP-RU tetramer enrichment. MAIT cell stages are defined with CD24 and CD44; stage 1 (S1; CD24+CD44) in blue, stage 2 (S2; CD24CD44) in green, and stage 3 (S3; CD24CD44+) in red. (B) MD plot showing gene expression comparison of bulk cell–purified stage 3 versus stage 1 MAIT cells. Colored dots indicate genes that significantly up-regulated in stage 3 (red) and stage 1 (blue). Colored numbers represent the total number of DEGs. (C) Table lists the top 50 most DEGs within each subset. (B and C) Data are from three pooled biological replicates each generated from a pool of thymi from five mice. (D) Gene expression comparison of single cell–purified stage 2 versus stage 1 (left) and stage 3 versus stage 2 (right) MAIT cells. (E) Phenotypic analysis of thymic MAIT cell stages for expression of transcription factors LEF1 (encoded by gene Lef1), SATB1 (Satb1), TCF1 (Tcf7), and Bach2 (Bach2); SLAM molecules CD150 (Slamf1), Ly108 (Slamf6), and CD319 (Slamf7); costimulation receptors CD27 (Cd27) and CD28 (Cd28); glycoprotein CD138 (Sdc1); and chemokine receptors CCR2 (Ccr2), CCR6 (Ccr6), CCR7 (Ccr7), and CXCR6 (Cxcr6). Histograms depict stage 1 MAIT cells in blue, stage 2 MAIT cells in green, stage 3 MAIT cells in red, and CD4+CD8+ double-positive (DP) thymocytes in black. Colored numbers indicate mean fluorescence intensity (MFI) values of each marker for the respective cell population. (A and E) Data are representative of at least two independent experiments with a total of two separate samples (pools of three thymi).

To further validate these transcriptional data, we performed flow cytometry analysis with a selection of antibodies and confirmed that transcription factors LEF1, SATB1, and TCF1 (Tcf7) down-regulated as MAIT cells develop, whereas BACH2 transiently increased between stage 1 and stage 2 and decreased at stage 3 (Fig. 1E). Surface receptors CD138, CCR2, CCR6, and CXCR6 were up-regulated, whereas CCR7, CD27, CD28, CD150 (Slamf1), and Ly108 (Slamf6) down-regulated at the protein level between stages 1, 2, and 3 (Fig. 1E). Given that Slamf1 and Slamf6 were clearly down-regulated as MAIT cells mature, we also examined Slamf7 (CD319), another SLAM family member that showed the opposite trend in our transcriptome analysis. This protein was clearly detected on a small subset of stage 3 thymic MAIT cells, reflecting heterogeneity in stage 3 (Fig. 1E). Accordingly, MAIT cells undergo major changes in gene and protein expression as they mature from stage 1 to stage 2, and even more so between stages 2 and 3. Furthermore, our data identify a number of candidate genes that might be important for the development of MAIT cells within the thymus.

Separation of mature stage 3 MAIT cells into functionally distinct subsets

As we have previously reported (6, 10), thymic stage 3 mouse MAIT cells include two functionally distinct subsets, defined as IFN-γ–producing T-bet+ MAIT1 and IL-17–producing RORγt+ MAIT17 cells (Fig. 2A). To compare these cell types at the transcriptomic level, we avoided cell fixation and permeabilization and/or activation by identifying cell surface markers to separate them. We investigated several possible candidate markers that are differentially expressed by stage 3 MAIT cells (Fig. 1C and table S1), including some that are expressed by related NKT1 and NKT17 cells (11, 12). These included the following: CD138 (Syndecan 1), CD43 [high glycosylation (HG) variant], CD122 (IL-2Rβ), CD127 (IL-7Rα), and CD319 (SLAMf7) (Fig. 1E). We tested these markers against RORγt and T-bet intracellular staining (Fig. 2B). RORγt+ cells stained higher for CD138, CD43HG, and CD127, whereas T-bet+ cells stained higher for CD319 and CD122. The combination of CD319 versus CD138 provided a clear and mutually exclusive separation, allowing us to identify MAIT1 and MAIT17 cells, respectively, within the stage 3 MAIT cell population without fixation and permeabilization (Fig. 2C). We also observed that some stage 3 MAIT cells did not express CD319 or CD138, which we surmised may encompass uncommitted stage 3 cells.

Fig. 2 Bifurcation and transcriptomic comparison of mature mouse MAIT cells.

(A) Flow cytometric analysis of mature mouse thymic MAIT cells after HSA complement–mediated depletion showing RORγt and T-bet expression in CD44+ stage 3 MAIT cells. (B) Panel of markers showing expression of CD138 (Sdc1), CD319 (Slamf7), CD43HG (Spn), CD122 (Il2rb), and CD127 (Il7r) relative to RORγt and T-bet in stage 3 MAIT cells. (C) Flow cytometry profile of stage 3 MAIT cells costained with CD138 and CD319. (D) Gene expression comparison of purified CD319+ versus CD138+ stage 3 MAIT cells. Colored dots indicate genes significantly up-regulated in CD319+ cells (red) and CD138+ cells (blue). Colored numbers represent the total number of DEGs. (E) Tables list genes enriched corresponding to CD319+ (red, top) and CD138+ (blue, bottom) MAIT cell subsets. (F) Flow cytometry analysis for expression of PLZF, CD44, ICOS, Ly6c, and NK1.1 relative to RORγt+ and T-bet+ mature MAIT cells. (G) LEF1, SATB1, CD28, and CD103 expression on CD138+- and CD319+-stained mature MAIT cells. Colored numbers indicate MFI of each marker for the respective CD138+ or CD319+ MAIT cells. (A to C, F, and G) Data are representative of at least two independent experiments.

We isolated CD319+ (MAIT1) and CD138+ (MAIT17) single cells and performed single-cell RNA-seq comparison of these two mature thymic MAIT cell populations (Fig. 2D). More than 1000 genes were differentially expressed between the mature CD138+ and CD319+ stage 3 MAIT cell subsets, 783 genes that were more highly expressed by the MAIT17 cells and 275 genes that were more highly expressed by the MAIT1 cells (Fig. 2, D and E). As expected from our cytometry-based analysis in Fig. 2 (B and C), CD319+ MAIT1 cells had higher gene expression of genes encoding CD319 (Slamf7), CD122 (Il2rb), and IFN-γ (Ifng), whereas CD138+ MAIT17 cells expressed higher levels of the genes encoding CD138 (Sdc1), CD127 (Il7ra), RORγt (Rorc), and IL-17A (Il17a) (Fig. 2, D and E). In addition to these genes, MAIT1 cells were distinguished by higher expression of genes encoding natural killer (NK) family members including Klr family members and many other genes including Lef1, Nkg7, Klrb1c, Ly6c2, GzmB, Bcl2, Stat4, Ms4a4b, and Cd28 (Fig. 2, D and E). In contrast, MAIT17 cells had higher expression of genes encoding IL-17–related molecules such as Il17a, Il17f, Il22, Il23r, Il17rb, and Il17re and also higher expression of Blk, Maf, Icos, Zbtb16, Tnfrs25, and Cd44 (Fig. 2D). Flow cytometry–based analysis of MAIT1 and MAIT17 subsets was used to validate some of these results (Fig. 2F). For example, RORγt+ MAIT17 cells expressed moderately higher PLZF and CD44 and much higher levels of inducible T cell costimulator (ICOS) compared with T-bet+ MAIT1 cells. Conversely, Ly6c and NK1.1 (Kkrb1c) were positive on many (but not all) MAIT1 cells, whereas most MAIT17 cells did not express these markers. Using CD138 and CD319 as surrogate markers, we confirmed LEF1 expression and CD28 as being lower on MAIT17 cells, whereas SATB1, which was expressed slightly higher in MAIT1 cells at the transcript level (Fig. 2D), showed no clear difference at the protein level (Fig. 2G). In addition, MAIT17 cells expressed much higher levels of CD103 (Itgae) compared with MAIT1 (Fig. 2G). Together, these data demonstrate that MAIT1 and MAIT17 cells are transcriptionally and functionally distinct populations that emerge as MAIT cells develop in the thymus.

Lineage diversification within stage 3 of MAIT cell development

Given that MAIT1 and MAIT17 cells exhibit a highly divergent transcriptional program, we sought to determine whether this happens before or after they reach stage 3 of MAIT cell development in the thymus. Using the combination of cell surface CD138 and CD319 expression, it was clear that stage 1 and stage 2 MAIT cells (Fig. 1D), and some stage 3 cells (Fig. 3A), lacked expression of these markers. This raised the possibility that MAIT1 and MAIT17 cells emerge after MAIT cells arrive at stage 3 and moreover that some newly arrived stage 3 cells may lack MAIT1 and MAIT17 transcriptomic signatures and will more closely resemble less mature stage 2 cells. To investigate these subsets of stage 3 cells in greater detail, we compared stage 3 (CD24CD44+) cells, sorted into three subpopulations (CD138CD319+, CD138+CD319, and CD138CD319), alongside immature stage 1 (CD24+CD44) and stage 2 (CD24CD44) cells, using single-cell RNA-seq analysis. Principal component analysis and Monocle analysis of these populations provided a clear demonstration of the marked differences between thymus-derived CD319+ (MAIT1) and CD138+ (MAIT17) stage 3 cells (Fig. 3B). Moreover, it demonstrated that the CD138 CD319 stage 3 cells were heterogeneous, with some sitting between the CD138+ and CD319+ populations based on principal component analysis dimension 2, whereas others appeared to merge with the CD138+ subset (Fig. 3B). There was less overlap between CD138CD319 stage 3 cells and the CD319+ subset, although some CD319 cells merged with the CD138CD319 subpopulation in the middle (Fig. 3B). This principal component analysis also showed that immature stage 1 and stage 2 cells were clearly distinct from most stage 3 cells including CD138+, CD319+, and CD138CD319 cells, although some CD138CD319 cells fell between immature stage 1 and stage 2 cells and CD138+ and CD319+ cells on dimension 1 (Fig. 3B). In addition, Pseudotime analysis carried out using Monocle on DDRtree clustering of the cell subsets also showed two trajectories branching from an intermediate CD138CD319 subpopulation toward terminally differentiated (CD138+ and CD319+) stage 3 subsets (Fig. 3B and fig. S3A). Thus, CD138CD319 cells may include candidates for newly arrived stage 3 MAIT cells that share similarities with stage 2 precursors.

Fig. 3 Lineage diversification within stage 3 of MAIT cell development.

(A) Flow cytometry profile of mouse MAIT cell subsets present within the mouse thymus single cell–sorted for RNA-seq analysis. Graph depicts percentages of each stage 3 subpopulation across eight samples. (B) Principal components analysis on the respective sorted subsets in (A) (left), and Monocole DDRtree showing the clustering of respective MAIT cell subsets (right). (C) Heatmap showing the DEGs across MAIT cell subsets ordered by decreasing Cd4 (top) and Lef1 (bottom) expression, grouped across developmental subsets (n = 363 in total, n = 75 stage 1, n = 34 stage 2, n = 81 CD138 CD319 double-negative stage 3, n = 95 CD138+ stage 3, and n = 78 CD319 stage 3). (D) Flow cytometry analysis of mature CD44+ MAIT cells for CD138 and CD319 expression (left). Analysis of CD4+ subset (yellow) from mature CD44+ CD319 MAIT cells (middle). Histogram overlay analysis of LEF1 expression between CD319 and CD138CD319 CD4+ stage 3 MAIT cells (right). (E) Scatter plot graph shows percentage of CD4+ subset within stage 2 and CD319 stage 3 MAIT cells. Graph shows percentage of LEF1+ cells within CD4+ S2 and CD4+ S3 MAIT cell populations.

Heatmap analysis of the top genes that separate each of the populations is shown in Fig. 3C, and the full heatmap is provided in fig. S1. The columns are grouped into each of the different MAIT cell subsets as described above and further organized for each subset on the basis of CD4 expression level because CD4 is highly expressed by stage 1 and 2 cells but is absent in most mature stage 3 cells (6). This heatmap analysis again highlights the many similarities between stage 1 and stage 2 MAIT cells (Fig. 1D) and how a major transcriptomic change is apparent when comparing these to most of the stage 3 cells. Similar to Fig. 2 (D and E), these data also reflect the extensive transcriptomic differences between CD319+ MAIT1 and CD138+ MAIT17 cells (Fig. 3C). It was also apparent that many of the stage 3 CD138CD319 cells more closely resembled CD138+ MAIT17 stage 3 cells, but consistent with the principal component analysis (Fig. 3B), a subset of the CD138CD319 cells, outlined by the black rectangle (Fig. 3C), appeared to be more similar to stage 2 cells in their overall transcript signature. For example, these cells express Cd4 at mid to high levels, in contrast to most other stage 3 cells that do not express Cd4. Because Lef1 was found to be the top gene most positively associated with Cd4, a second heatmap ordered via Lef1 expression illustrated a similar subset of cells within this black rectangle (Fig. 3C). These Cd4- and Lef1-expressing stage 3 cells also expressed higher levels of many other genes that are more in line with the stage 2 signature than other stage 3 cells, including Tcf7, Itm2a, and Satb1, and they showed little expression of MAIT1- or MAIT17-related genes (fig. S1). CD4- and LEF1-expressing CD138CD319 stage 3 cells were almost entirely encompassed within the cells in the center of the principal component and Monocle analysis (Fig. 3B), and Cd4 and Lef1 also showed a similar expression trajectory, when Pseudotime is overlaid onto Monocle analysis (fig. S3B). To exclude the possibility that the CD4+LEF1+ cells were contaminating stage 2 cells during the single-cell RNA-seq sorting, we used flow cytometry to examine stage 3 MAIT cells and confirmed that a clear subset (~10%) of CD4+ cells was present within the CD138CD319 population (Fig. 3D). Moreover, intracellular transcription factor staining revealed that these also expressed LEF1, in contrast to most other stage 3 cells (Fig. 3, D and E). Thus, these data suggest that CD4+LEF1+CD319CD138 stage 3 cells are the next stage beyond stage 2 and represent newly arrived stage 3 cells that have not yet committed to either the MAIT1 or MAIT17 lineages. To directly investigate this developmental sequence, we reanalyzed experiments from our earlier study (6), examining stage 3 cells from very young mice (2-week-old mice). We report here that a much higher percentage of stage 3 cells were CD4+ compared with in older mice (fig. S4A).

Aside from the CD4+LEF1+CD138CD319 stage 3 cells, which also seem to be a less mature subset of stage 3 cells based on their DEG profile (fig. S4B), the remaining population of CD4LEF1CD138CD319 stage 3 cells contained a mix of cells whose characteristics were more in common with CD138+ MAIT17 cells than with CD319+ MAIT1 cells. This is clearly seen using mean-difference (MD) plot analysis where relatively few genes were differentially expressed between LEF CD138CD319 stage 3 cells and CD138+ MAIT17 cells (fig. S4C), whereas many genes were differentially expressed between LEF1+ CD138CD319 stage 3 cells and CD138+ MAIT17 cells (fig. S4D). Of the many genes that distinguished CD4LEF1CD138CD319 cells from CD319+ cells, some of the most noteworthy included the following: Slamf7, GzmA, Ccl5, and Bcl2 (all increased); Ccr6, Ccr7, and Icos (decreased); and some of the IL-17–associated genes such as Rorc, Il17ra, and Il17re. The enrichment of these lineage-specific signatures was recapitulated in KEGG pathway analysis performed between these subsets (fig. S2B). Of the few genes that distinguished CD4LEF1CD138CD319 and CD138+ cells, notable genes included increased Klf4 and proto-oncogene transcription factors Fos, FosB, and Jun (fig. S4, C and D).

Together, these data highlight a high degree of diversity within stage 3 of MAIT cell development. These data suggest that there are distinct subpopulations of stage 3 MAIT cells, the earliest arrivals being the CD4+LEF1+ stage 3 cells and the most functionally mature being the CD319+ MAIT1 or CD138+ MAIT17 cells, as well as a population of CD4LEF1CD138CD319 cells that comprises a mixture of cells, including some resembling CD138+ MAIT17 cells, which we propose represent putative precursors of mature CD138+ MAIT17 cells. Conversely, we were unable to detect a clear population of CD4CD319CD138 cells that might reflect MAIT1 precursors, suggesting that this transition happens quite rapidly and/or that CD319 is an early marker associated with the differentiation of this functionally distinct sublineage. The detection of some CD319+ cells in the middle range of the principal component analysis plot (Fig. 3B) supports this possibility.

Factors that regulate MAIT cell differentiation

Because many of the genes that were sharply modulated as MAIT cells mature are known to be important in the development or function of at least some types of T cells, we selected a panel of these genes to examine their importance in MAIT cell maturation. Because SLAM family members were differentially expressed between immature stage 1 MAIT cells and mature stage 3 MAIT cells (Fig. 1), we investigated MAIT cell development in mice lacking SLAM-associated protein (SAP) that transmits SLAM-based intracellular signals (13). This analysis revealed a key role for SLAM family members in regulating MAIT cell development beyond stage 1, such that stage 2 MAIT cells were numerically diminished and stage 3 MAIT cells were essentially absent, with the few residual stage 3 cells enriched for CD4+ (fig. S4A), supporting the hypothesis that these could be early stage 3 cells. Furthermore, we were unable to detect MAIT cells in spleens of SAP-deficient [SAP knockout (KO)] mice (Fig. 4A). Conventional TCRβ+ MR1-tetramer cells were present in moderately higher numbers in these SAP-deficient samples (fig. S5a), whereas as expected, these mice also had a reduction in CD1d-restricted NKT cells and only a slight decrease in the percentage, but not number, of γδ T cells (fig. S5B).

Fig. 4 Identification of SAP, SATB1, Zap70, CXCR6, and CCR7 as key factors that regulate MAIT cell development with gene-deleted mice.

(A to E) Three panels depicting analysis of MAIT cells WT control mice and SAP KO (A), Satb1m1Anu/m1Anu (B), SKG (C), CXCR6 KO (D), and CCR7 KO (E) mice. Left panel of flow cytometry analysis shows detection of MAIT cells from whole and MR1-5-OP-RU tetramer–enriched thymus of respective mice, analysis of MR1-tetramer–enriched MAIT cells for CD24 and CD44 expression. Scatter graphs depict absolute numbers and percentage of MAIT cells of T cells in thymus and spleen (A to E) and lymph nodes (E) of the respective mouse strains. Bar graphs in (E) show the proportion of RORγt and T-bet subset of stage 3 MAIT cells in WT and CCR7 KO thymus, spleen, and lymph nodes. Data are representative of two independent experiments with a total of six to nine mice per group (A to E; mean ± SEM). *P < 0.1 **P < 0.01 ***P < 0.001, ****P < 0.0001 (Mann-Whitney U rank sum test; A to E).

SATB1 is a chromatin-organizing protein with crucial functions in T cells, including in Foxp3+ T regulatory (Treg) cell development and function (14) and NKT cell development (15). This gene was differentially expressed as MAIT cells develop, with highest expression in stage 1 MAIT cells (Fig. 1, B to D). We examined a previously unpublished Satb1m1Anu/m1Anu mouse model that is unlike Satb1-deficient mice, which are runted and typically die by 3 weeks after birth (16). Mice homozygous for the Satb1m1Anu allele are healthy and have a normal life span but exhibit abnormalities in T cell development (fig. S6). Satb1m1Anu/m1Anu mice show a major defect in MAIT cell development beyond stage 2 with a marked reduction of stage 3 MAIT cells (Fig. 4B). MAIT cells were also markedly diminished in the spleens of Satb1m1Anu/m1Anu mice (Fig. 4B). As anticipated, Satb1m1Anu/m1Anu mice also had a major reduction in CD1d-restricted NKT cells (15) but no significant change in γδ T cells, aside from a moderately reduced number of these in the spleen (fig. S5B). Conventional TCRβ+ MR1-tetramer cell numbers were not significantly different in Satb1m1Anu/m1Anu samples, apart from a verified increase in the CD4+ Foxp3+ Treg compartment (figs. S5A and S6).

Next, we examined Zap70-deficient (SKG) mice (17) because we have previously shown that MAIT cell development requires ongoing MR1-dependent signaling, suggesting a role for TCR signaling, and Zap70-deficient mice were previously reported to lack NKT cells (18). We found that MAIT cells are diminished in the thymus and spleen of the Zap70-deficient mice, although all subsets were still detectable (Fig. 4C), supporting a role for TCR signaling not only in the initial emergence of Stage 1 MAIT cells but also in MAIT cell maturation to stage 3. We did not detect a change in thymus NKT cells, and furthermore, these cells were more frequent in spleens of Zap70-deficient mice. TCRβ+ MR1 tetramer T cells in the thymus were present in normal numbers in Zap70-deficient mice (fig. S5).

As several chemokine receptors were strongly modulated as MAIT cells mature, we examined whether signaling through some representative chemokine receptors is important for MAIT cell maturation. CXCR6-deficient mice exhibited a near-complete block in MAIT cell development between stages 2 and 3, and no MAIT cells could be detected in the spleen or lymph nodes of CXCR6-deficient mice (Fig. 4D). CXCR6-deficient mice also had a moderate reduction in the number of CD1d-restricted NKT cells in the thymus and a major reduction of these cells in the spleen (Fig. S5B). In contrast, CCR7-deficient mice had a partial defect in MAIT cell development with a large reduction in the percentage and number of MAIT cells in the thymus and spleen but not the lymph nodes. CCR7 KO mice retained a clear population of MAIT cells at stage 3, that on further analysis showed to be heavily biased toward T-bet+ MAIT1 cells compared with MAIT cells from wild-type (WT) mice (Fig. 4E), demonstrating that CCR7 is selectively required for the differentiation of Rorγt+ MAIT17 subset. This effect was even more pronounced for MAIT cells in spleens of CCR7-deficient mice, but no difference was observed for these cells in lymph nodes (Fig. 4E). In addition, these results also suggest that CD138+ MAIT cells might preferentially home to, or expand in, lymph nodes and that CCR7 is not required for their presence in this site. CCR7 KO mice also had a major reduction in CD1d-restricted NKT cells and, to a lesser extent, γδ T cells (fig. S5B). Conventional TCRβ+ MR1-tetramer cells were present in normal numbers in CXCR6- and CCR7-deficient samples (fig. S5A). Mice deficient in Bach2, a repressor of effector differentiation regulated by PLZF (19), exhibited no change in MAIT cell percentage or number despite its elevated expression in early-stage MAIT cells (fig. S7).

DEGs associated with human MAIT cell development

In humans, there are important parallels with mouse MAIT cell development, including the existence of a three-stage development pathway where stage 1 MAIT cells are CD27CD161, stage 2 MAIT cells are CD27+CD161, whereas stage 3 MAIT cells are CD27+/−CD161+ (Fig. 5A) (6).

Fig. 5 Transcriptomic analysis of human MAIT cell development.

(A) Flow cytometric analysis of three stages of human Vα7.2+ thymic MAIT cells of CD3+ cells before and after TRAV1-2 enrichment. MAIT cell stages are defined with CD27 and CD161; stage 1 (S1, CD27CD161) in blue, stage 2 (S2, CD27+CD161) in green, and stage 3 (S3, CD27+/-CD161+) in red. (B) MD plot showing gene expression comparison of bulk cell–purified stage 3 versus stage 1 human MAIT cells. Colored dots indicate genes that significantly up-regulated in stage 3 (red) and stage 1 (blue). Colored numbers represent the total number of DEGs. (C) Table lists the top 50 most DEGs within each subset. (D) Gene expression comparison of purified stage 2 versus stage 1 human MAIT cells and stage 3 versus stage 2 human MAIT cells. (B to D) Data are representative from two separate sorts, with a total of eight donor samples. (E) Phenotypic analysis of human thymic MAIT cell stages for expression of transcription factors SATB1, LEF1, and TCF1; CD1 molecules CD1a, CD1b, and CD1d; HLA-A, HLA-B, HLA-C; and chemokine receptors CCR5 and CCR6. Histograms depict stage 1 MAIT cells in blue, stage 2 MAIT cells in green, stage 3 MAIT cells in red, CD4+CD8+ DP thymocytes in black. (F) Flow cytometry analysis of CD4+ (yellow) and CD4 (pink) CD161+ thymic MAIT cells for LEF1 expression. Scatter plots show percentage CD4+ cells within CD161+ thymic MAIT cells and percentage LEF1+ cells within this CD4+ population. (E and F) Data indicate mean ± SEM and are representative of at least three experiments using three to seven separate donor samples.

To explore the changing transcriptomic landscape during human MAIT cell development, we carried out RNA-seq analysis of different stages of thymic MAIT cells. A comparison of stage 1 to stage 3 revealed 625 DEGs, 390 of which were increased and 235 were decreased as MAIT cells mature (Fig. 5B). The 50 most significant of each category are listed in Fig. 5C, the full list of DEGs is provided in table S2, and a heatmap is provided in fig. S8. Our previous study identified higher protein expression of CD218 (IL18R), CD161 (KLRB1), CD27, and PLZF (ZBTB16) in stage 3 MAIT cells compared with stage 1 MAIT cells (6), and this was also reflected by an increase in gene expression of these molecules (Fig. 5, B and C). In addition, we detected a significant increase in SLAMF7, GZMA (granzyme A), GZMK, CXCR6, CCR5, CCR6, IL2RB (CD122), and IL7RA (CD127), ABCB1 encoding the adenosine triphosphate–binding cassette multidrug efflux protein 1 (MDR1) that is prominently expressed in peripheral blood MAIT cells (20, 21), and CEBPD encoding for the bZIP transcription factor C/EBPδ recently shown to be important for MAIT cell trafficking (22). Conversely, there were many genes that were down-regulated in stage 3 human MAIT cells, including: RAG1, SATB1, Lef1, Tcf7, CCR9, BACH2, CD8β, and CD4 (Fig. 5, B and C). Note that many of these were similarly modulated between stages 1 and 3 in mouse MAIT cells, including ZBTB16, CXCR6, CCR6, CCR9, SLAMF7, IL2RB, IL7R, IL18R, RAG1, SATB1, LEF1, TCF7, BACH2, CD8β, and CD4 (Fig. 1, B to D, and Fig. 5, B and C).

We next investigated where most of these transcriptional changes occurred by comparing stage 1 with stage 2 and stage 2 with stage 3 (Fig. 5D). This revealed that only 84 genes were significantly modulated (51 up and 33 down) between stage 1 and stage 2, whereas a much greater change took place between stage 2 and stage 3 (205 up and 148 down). As expected, one of the genes that increased between stage 1 and stage 2 was CD27, which is consistent with its use as a marker to separate these stages. HLA-A, HLA-B, and HLA-C genes and CCL5 increased between stages 1 and 2, and RAG1, RAG2, and CD1e decreased, which is consistent with the relative maturity of stage 2 cells. As expected, most of the changes between stages 2 and 3 were the same as those detected between stages 1 and 3. Accordingly, these data reveal that most changes in gene expression occur between stage 2 and stage 3 in humans, which is similar to that in mice (Fig. 1), further validating the three-stage thymic development pathway that we originally proposed (6).

Using flow cytometry, we were able to validate results from these RNA-seq studies, showing a decrease in expression of SATB1, LEF1 and TCF1 transcription factors, and CD1 proteins (CD1a, CD1b, and CD1d), as human MAIT cells mature from stages 1 to 3 (Fig. 5E). Conversely, human MAIT cells up-regulate expression of human leukocyte antigen (HLA) class I, CCR5 and CCR6 molecules as they mature intrathymically (Fig. 5, E and F). These findings, in conjunction with previously validated factors (CD27, CD161, CD218, and PLZF) (6), provide assurance that the modulated transcripts detected in our RNA-seq–based assay are reflective of actual changes in protein expression as MAIT cells mature. Considering that we detected a CD4+ LEF1+ population that appears to be early stage 3 cells in mice and that these proteins were similarly expressed in human MAIT cells, we looked for the presence of a comparable population in humans. We found that about 15% of human stage 3 cells expressed CD4, and these were all LEF1+ (Fig. 5G). This suggests that a similar population within stage 3 cells expresses analogous factors in both mice and human thymus.

Further transcriptional modulation associated with extrathymic MAIT cell differentiation

We have previously documented that human MAIT cells appear to undergo further, extrathymic maturation after leaving the thymus, increasing their capacity to produce cytokines and switching from CD8αβ to CD8αα (6). Hence, we carried out RNA-seq analysis on sorted stage 3 MAIT cells from human thymus and blood samples from the same donors. These data revealed considerable diversity between mature thymic versus extrathymic MAIT cells, even despite the blood coming from very young thymus blood–matched donors (Fig. 6, A and B). Of those that increased, some of the immunologically relevant genes were TGFB1, NFKB1, IRF1, CD83, CXCR4, and SOCS3, whereas those that decreased included LEF1, CCR7, CXCR6, EOMES, CD2, ID3, IL7R, SOX4, TOX2, and SATB1. We have recently profiled SATB1 expression in human T cells where thymic T cell progenitors have the highest levels of SATB1 that is subsequently down-regulated in mature peripheral CD4+ and CD8+ T cells (23). We validated via flow cytometry that CCR7 and LEF1 expression decreased between the thymus and young blood stage 3 MAIT cells (Fig. 6C). It has previously been reported that CD8αα MAIT cells arise extrathymically because most stage 3 cells in the thymus are CD8αβ (6, 9). We found that CD8β was reduced on average in peripheral MAIT cells, but the difference did not quite reach significance in our analysis. Together, these data reveal that human MAIT cells continue to mature and differentiate after leaving the thymus, which may be a result of extrathymic exposure to antigen, cytokines, or other factors. Previous studies have examined the transcriptional profile of human MAIT cells, defined as CD161-associated genes in one study (24) or Vα7.2+ CD161+ T cells in another (25). We found considerable overlap of these reported genes with the DEGs revealed in our RNA-seq–based study on purified MAIT cells and could further segregate these gene regulation events in a temporal manner. Thus, we grouped the genes that stood out in these signatures into two stages: those that arise intrathymically between stages 1 and 3 and those that arise after MAIT cells leave the thymus (Fig. 6D).

Fig. 6 Transcriptomic analysis of extrathymic MAIT cell development.

(A) Flow cytometry profile with CD27 and CD161 expression of Vα7.2+ MAIT cells of CD3+ cells from young (thymus donor) and adult peripheral blood sample. (B) MD plot showing gene expression comparison of bulk cell–purified stage 3 MAIT cells from young peripheral blood versus thymus sample and associated list of the most DEGs. Data were derived from five young blood and three thymus donor samples. (C) Flow cytometry analysis of CCR7 and LEF1 expression on stage 3 thymus MAIT cells, stage 3 young blood MAIT cells, and young blood conventional T cells. Data are representative of at least three experiments using three separate donor samples. (D) Classification of known DEGs between MAIT and conventional T cells and known core transcriptional signature of CD161-associated up-regulated genes into intrathymic (between thymic stage 3 and stage 1) and extrathymic (between young blood and thymus) regulation events.

We also carried out RNA-seq–based comparison of MAIT cells from mouse spleen (essentially all stage 3) to stage 3 MAIT cells from mouse thymus. This analysis revealed unexpectedly few genes (19) that were differentially expressed between these populations, 5 that were increased and 14 that were decreased. Immunologically relevant genes among those that were increased include Dtx1 (Deltex1) and Ccl5, whereas among those that were decreased include Ly6a, Klra3, and Klra9. The main message to arise through these data seems to be that there is remarkable consistency between the transcriptome of stage 3 MAIT cells in mouse thymus and MAIT cells in spleen (fig. S9).

In summary, these data have provided a major advance to our current understanding of MAIT cell development, identifying a new population of CD4+ LEF1+ stage 3 cells that appears to be present in both mice and humans and many genes that are modulated through the three-stage progression that are similar between mice and humans. This suggests that these conserved transcriptional changes play a key role in MAIT cell development in both species. We have integrated the information derived from our findings in this paper into a new model of MAIT cell development, as illustrated in Fig. 7.

Fig. 7 A new model of MAIT cell development.

Dashed lines depict likely pathways based on transcriptome analysis. Phenotypic signature of each stage is shown in blue text in boxes. Factors that regulate each differentiation step are depicted above the arrow for each relevant step: Green text represents newly defined factors that support the relevant step, and black text represents previously defined factors from this study. This model is based on mouse MAIT cell development.


We have previously presented a model of MAIT cell development that progressed through three distinct stages in mice and humans. Here, using these stages as a starting point, we have mapped the transcriptional changes that occur as MAIT cells progress through each stage, in both species. Our results have highlighted a greater level of complexity in the stage 3 phase of MAIT cell development, allowing many key advancements to the existing model developed in our previous publication (6). This has resulted in a new model (Fig. 7) that integrates some of the key findings from this paper.

Our findings have confirmed that stages 1 and 2 are clearly separate populations with numerous transcriptional changes, and there are many more transcriptional changes that separate stages 2 and 3. This is true for both mice and humans, and indeed, many of the changes align between both species, which is quite reassuring that mouse MAIT cell development represents a viable model for understanding this process in humans. We also identified an unexpected degree of heterogeneity in mouse stage 3 cells, and with the power of single-cell RNA-seq, we were able to separate these into at least four distinct subpopulations. Of these, we have identified a new stage 3 subpopulation, defined as CD4+ LEF1+ stage 3 cells, that seems to bridge the gap between stage 2 and the bulk of stage 3 cells This is illustrated by principal component analysis and is experimentally supported by its enrichment in stage 3 cells from very young mice and SAP-null mice. At the other end of the stage 3 spectrum, we have populations of functionally mature MAIT1 and MAIT17 cells that are capable of IFN-γ or IL-17A production, respectively. Although we had previously identified these two populations (6, 10), here, we demonstrate that they are remarkably distinct at the transcriptomic level and furthermore, that they can be readily separated using CD138 and CD319 cell surface markers. A comparison of the genes expressed by MAIT1 and MAIT17 cells with those recently published for NKT1 and NKT17 cells highlights considerable overlap between these functionally similar subsets (12). Moreover, a recent study focused on tissue-resident MAIT cells showed that RORγt+ MAIT17 cells and RORγt MAIT cells (enriched for MAIT1 cells), identified from the lung, liver, and spleen, shared similarity to NKT1 and NKT17 cells isolated from the same tissue, although clear differences were observed between tissues for each cell type (26). In our study, we focused on MAIT cell subsets in the thymus to understand where this transcriptional bifurcation is likely to occur during intrathymic MAIT cell development. We propose that MAIT1 and MAIT17 cells arise in the thymus, within stage 3, from a subset that was not labeled with anti-CD138 or CD319. These cells are themselves diverse, including the CD4+LEF1+ population that more closely resembles stage 2 cells and CD4LEF1CD138CD319 cells, which most closely resemble MAIT17 cells. Thus, we have proposed a model pathway that integrates these stage 3 subpopulations into the earlier three-stage pathway of MAIT cell development (Fig. 7) (6). This pathway postulates that stage 2 cells (which express CD4 and LEF1) first arrive as CD4+LEF1+ stage 3 cells. These CD4+ stage 3 cells then transition to either CD319+ MAIT1 or CD138+ MAIT17 cells. The CD138CD319CD4LEF1 stage 3 population may contain intermediates for both MAIT1 and MAIT17 cells, although as a population they more closely resembled the MAIT17 lineage and may reflect a continuum with MAIT17 cells.

We were able to validate many of the transcriptional changes observed by antibody-mediated detection of the relevant proteins, for example, SLAM family members (Slamf1, Slamf6, and Slamf7) varied widely at the level of transcripts, and this was verified by antibody staining of the relevant MAIT cell subsets. We were also able to directly test the biological significance of some of the key factors detected in this study, demonstrating a critical role that several of these gene products play in the process of MAIT cell development. An initial report suggested that the SLAM-SAP-Fyn pathway was not critical for the development of MAIT cells because SAP-deficient patients contained normal frequencies of MAIT cells (27). However, the expression of SLAMF1 and SLAMF6 in early MAIT cell precursors, and of SLAMF7 (CD319) by MAIT1 cells, prompted us to reexamine the role of SLAM family members using SAP-deficient mice, revealing that MAIT cell numbers are drastically reduced in these mice. The dissimilarities in MAIT cell frequency observed between species are likely due to the large expansion of MAIT cells that occurs in humans but not mice (6, 2831). Furthermore, while this paper was under review, another study has been recently published that also demonstrated down-regulation of expression of SLAMF6 as MAIT cells developed and showed impaired MAIT cell development in SAP-deficient mice (32). CCR7 was recently shown to be expressed by RAG2+ early MAIT cell precursors (33), and we validated this finding here by using transcriptome and flow cytometry analysis. Furthermore, we demonstrate that a deficiency in CCR7 results in a reduction of thymic and splenic MAIT cells and a bias toward MAIT1 cells in these organs. CCR7 might be required for thymocyte migration because previous studies showed that NKT cell development is impaired in CCR7 KO mice, likely because of diminished thymocyte migration into the thymic medulla (34). We also showed that mice that lack Zap70 have a drastic reduction in thymic MAIT cells suggesting a role for TCR signaling in MAIT cell development, although these mice have reduced thymocyte survival and this might be caused by diminished Bcl-XL, RORγ, and TCF1/LEF1 in these mice (35, 36). Notwithstanding the lack of formal validation for the role of LEF1 with LEF1-deficient mouse models, we recently demonstrated that Tcf7, encoding TCF1, is essential for MAIT cell development with virtually no detectable MAIT cells in Tcf7 KO mice (37), indicating a Tcf7-dependent step at the earliest stage of MAIT cell development (Fig. 7). Given that TCF1 and LEF1 are similarly regulated as MAIT cells mature and that they have a similar role in conventional T cell development (38) and both play coordinated roles in NKT cell development (39, 40), we suggest that, similar to TCF1, LEF1 will also be a key factor in regulating MAIT cell development.

Our data identify many other genes that are differentially expressed as MAIT cell develop in the thymus, and indeed the data from this manuscript may serve as a basis for many new experiments that explore the development, function, and differentiation of these cells. For example, c-maf is up-regulated in stage 3 thymic MAIT cells, specifically MAIT17 cells, and it was recently reported to regulate the expression of known IL-17 effector program genes (Blk, Rorc, and Il17a) in IL-17–producing mouse γδ T cells (41). We demonstrate here that these same genes are preferentially expressed by MAIT17 cells compared with MAIT1 cells and less mature MAIT cells. Thus, we have markedly expanded the list of known factors that control MAIT cell development before stage 1 (SKG), between stages 2 and 3 (SLAM family members via SAP, CXCR6, and SATB1), and the functional branchpoint within stage 3 (CCR7).

It was encouraging that there were many homologous/orthologous genes that are similarly regulated in both human and mouse MAIT cell maturation through stages 1 to 3, in terms of transcription factor and surface protein expression. As some key examples: PLZF, IL-7 receptor (CD127), IL-18 receptor (CD218), CD161 (NK1.1/KLRB1), SATB1, LEF1, and CD4 undergo similar regulation between stage 1, 2, and 3 cells, in both mice and humans. Considering the large evolutionary gap between humans and mice and that many genes in humans are simply not expressed or differentially regulated in mice and vice versa, we feel that the two models from each species are perhaps unexpectedly well aligned, and information gained from mouse models is valuable for understanding intrathymic MAIT cell development in both species. We note that these observations build on previous findings that human thymic stage 3 MAIT cells do not show the same clear bifurcation in the thymus as mouse stage 3 MAIT cells, but the bifurfication event in human MAIT cell development appears to occur postthymically and seems to be influenced by different tissue microenvironments (4244).

In summary, this combined transcriptomic-based study has mapped in fine detail the molecular changes that occur in MAIT cells as they develop in the thymus of mice and humans. Many of these have been validated by flow cytometry, and we have established the functional significance of several of these as regulators of MAIT cell development using gene KO mice. The results provide a new and far more detailed model of how MAIT cells develop in mice and humans and provide a framework upon which future studies of these cells, and the factors that control their generation, will be founded.



C57BL/6 WT, BALB/c WT, CXCR6–green fluorescent protein (C57BL/6 background) (45), Satb1m1Anu/m1Anu (C57BL/6 background), Bach2CD4-cre (C57BL/6 background) (46), and SKG (BALB/c background) (17) mice were bred in-house at the Department of Microbiology and Immunology Animal House (University of Melbourne). Satb1m1Anu/m1Anu mice were generated from the Satb1m1Anu mutation. The Satb1m1Anu allele (NM_001163630.1:c.1500T > A), which encodes leucine instead of phenylalanine at amino acid 393 of SATB1, was identified in a screen for immunological regulators on the basis of the chemical mutagen N-ethyl-N-nitrosourea (47). The Satb1m1Anu mutation was identified by exome sequencing (48) in three littermate mice that exhibited decreased CD44 expression on CD4+ and CD8+ T cells. Similar phenotypic abnormalities in mice with T cell–specific ablation of SATB1 expression were observed in Satb1m1Anu/m1Anu mice (fig. S6) (14). SAP KO (Sh2d1a−/−) mice (49), and age- and sex-matched controls were bred and housed in specific pathogen-free conditions at the Garvan Institute Biological Testing Facility or Australian BioResources. CCR7 KO mice were sourced from JAX (006621) and were bred and maintained at the University of Adelaide Animal Facility behind a specific pathogen–free barrier in individually ventilated caging. CCR7 WT controls were age- and sex-matched WT C57BL/6J that were bred and maintained in the same room and on the same rack. All mice used were aged between 6 and 10 weeks at time of sacrifice. All procedures using mice were approved by the University of Melbourne Animal Ethics Committee, the Garvan Institute–St. Vincent’s Animal Ethics Committee, University of Adelaide Animal Ethics Committee, the Australian National University Animal Experimentation Ethics Committee, or the Walter and Eliza Hall Institute Animal Ethics Committee. Single-cell suspensions from mouse thymus, spleen, and lymph nodes were prepared as previously described (10).

Human blood and tissue

Adult peripheral human blood samples were obtained from the Australian Red Cross Blood Service under agreement number 13-04VIC-07. Young human peripheral blood samples and matching thymus were obtained from the Royal Children’s Hospital (RCH; Victoria, Australia). These thymic and matched peripheral blood data were derived from eight donors between the ages of 2 months and 4 years old.

Experiments were conducted in accordance with the University of Melbourne Human Research and Ethics committee guidelines (reference numbers 1035100 and 1443540) and RCH Human Research Ethics Committee Approval (reference number 24131 G). Blood mononuclear cells were isolated by Ficoll-Paque Plus density gradient centrifugation (GE Healthcare). Donor thymi were cut into small pieces and passed through a 70-μm cell strainer into ice-cold RPMI 1640 medium containing 2 mM EDTA before being washed into phosphate-buffered saline + 2% fetal calf serum [fluorescence-activated cell sorting (FACS) buffer].

Magnetic bead enrichment of thymic MAIT cells

Using 5-OP-RU formed in situ from synthetic 5-A-RU and methylglyoxal, mouse and human MR1 tetramers were generated and biotinylated as previously described (5052). Biotinylated MR1-5-OP-RU monomers were tetramerized with streptavidin conjugated to either phycoerythrin (PE) (BD Pharmingen) or Brilliant Violet 421 (BioLegend). Single-cell suspensions of mouse thymus were prepared and stained with PE-mouse MR1-5-OP-RU tetramers before magnetic bead enrichment using anti-PE microbeads as per manufacturer’s instructions (Miltenyi Biotec). One independent enriched sample represents two to three pooled thymi unless otherwise specified. Single-cell suspensions of human thymus were enriched for TRAV1-2+ cells by staining for TRAV1-2-PE antibody, followed by magnetic bead enrichment using anti-PE microbeads (Miltenyi Biotec).

Flow cytometry, sorting, and intracellular transcription factor staining

Mouse and human cells were stained with viability dye 7-aminoactinomycin D (Sigma-Aldrich) and cell surface antibodies concurrently with MR1-tetramers, at room temperature (RT) for 30 min. A list of antibodies used is detailed in table S4. Transcription factors were assessed by staining with antibodies after the cells were surface-stained and permeabilized with the eBioscience Foxp3 Fixation/Permeabilization kit, according to the manufacturer’s instructions. Apart from intracellular analysis, mouse cells were analyzed unfixed after staining, whereas human cells were fixed with 1% paraformaldehyde. Analysis was carried out using a BD LSRFortessa equipped with a 561-nm yellow-green laser, and data were processed using FlowJo software (TreeStar). Mouse cells are gated on B220 lymphocytes and human cells on CD14CD19 lymphocytes after dead cell and doublet exclusion. Flow cytometric sorting for live, unfixed bulk (100 cells) or single cells was performed using a BD FACSAria III cell sorter. Flow cytometry gating strategy is shown in fig. S10.

Cell sorting for RNA-seq

Bulk cell sorting involved sorting three separate samples for 100 cells as biological and technical repeats. Bulk and single MAIT cells were flow-sorted into a chilled 384-well polymerase chain reaction (PCR) plate (Greiner Bio-One, no. 785290) containing 1.2 μl of primer-lysis mix {20 nM indexed polydT primer (custom made, IDT), 1:6,000,000 dilution of ERCC RNA spike-in mix (Ambion no. 4456740), 1 mM deoxynucleotide triphosphates [dNTPs; New England BioLabs (NEB), no. N0446S)], 1.2 U of SUPERase IN Rnase Inhibitor (Thermo Fisher Scientific, no. AM2696), 0.2% Triton X-100 solution (Sigma-Aldrich, no. 93443), diethyl pyrocarbonate (DEPC) water (Thermo Fisher Scientific, no. AM9920)} using a BD FACSAria III flow cytometer (BD Biosciences, San Jose, CA, USA). Sorted plates were sealed, centrifuged for 1 min at 3000 rpm and immediately frozen upside down at −80°C until further processing using an adapted CelSeq2 protocol (53).

Complementary DNA generation

Sorted plates were thawed on ice and briefly centrifuged. To lyse the cells and anneal the mRNA capture primer, we incubated the plate at 65°C for 5 min and immediately chilled it on ice for at least 2 min before adding 0.8 μl of reverse transcription reaction mix [in 2-μl reverse transcription reaction: 1× First-Strand Buffer (Invitrogen, no. 18064-014), 20 mM dithiothreitol (DTT; Invitrogen, no. 18064-014), 4 U of RNaseOUT (Invitrogen, no. 10777-019), and 10 U of SuperScript II (Invitrogen, no. 18064-014)]. The plate was incubated at 42°C for 1 hour and 70°C for 10 min and chilled to 4°C to generate first-strand complementary DNA (cDNA). For second-strand cDNA synthesis, 6 μl of second strand reaction mix were added [1x NEBNext Second Strand Synthesis buffer (NEB, no. E6111S); NEBNext Second Strand Synthesis Enzyme Mix: 2.4 units DNA Polymerase I (Escherichia coli), 2 units RNase H, 10 units E. coli DNA Ligase (NEB, no. E6111S), DEPC water (Thermo Fisher Scientific, no. AM9920)]. The plate was incubated at 16°C for 2 hours to generate double-stranded cDNA.

Pooling and cleanup

Cleanup was performed with 1.2× NucleoMag NGS Cleanup and Size Select magnetic beads (MACHEREY-NAGEL, no.7449970.5) according to the manufacturer’s instruction. To reduce the amount of beads for each 100 μl of pooled sample, we added 20 μl of beads and 100 μl of bead binding buffer [20% polyethylene glycol, molecular weight 8000 and 2.5 M NaCl (pH 5.5)]. The cDNA was eluted in 6.4 μl of DEPC water and kept with beads for the following in vitro transcription (IVT) reaction.

In vitro transcription

IVT reaction mix (9.6 μl) [1.6 μl of each of the following: adenosine triphosphate, guanosine triphosphate, cytidine triphosphate, and uridine triphosphate solution; 10× T7 buffer; and T7 enzyme (MEGAscript T7 Transcription Kit—Ambion, no. AM1334)] was added and incubated at 37°C for 13 hours and then chilled and kept at 4°C. To remove leftover primers, we added 6 μl of ExoSAP-IT For PCR Product Cleanup (Affymetrix, no. 78200), and the sample was incubated at 37°C for 15 min and then chilled and kept at 4°C.

RNA fragmentation and cleanup

Chemical heat fragmentation was performed by adding 5.5 μl of 10× fragmentation buffer (RNA Fragmentation Reagents, no. AM8740) to the sample and incubating in preheated thermal cycler at 94°C for 2.5 min followed by immediately chilling on ice and adding 2.75 μl of Fragmentation Stop buffer (RNA Fragmentation Reagents, no. AM8740). The fragmented amplified RNA was purified using 1.8× RNAClean XP beads (Beckman Coulter, no. A63987) according to the manufacturer’s instruction and eluted in 6 μl of DEPC water of which 5 μl (no beads) was transferred to a fresh tube for library preparation.

5′-Tagged random hexamer reverse transcription

The fragmented RNA was transcribed into cDNA using 5′-tagged random hexamer primers 9 (GCCTTGGCACCCGAGAATTCCANNNNNN), introducing a partial Illumina adapter as described in CEL-Seq2 as well (53). To remove RNA secondary structure and anneal the mRNA capture primer, we added 1 μl of tagged random hexamer (100 μM) and 0.5 μl of 10 mM dNTPs (dNTP solution set, NEB, no. N0446S) to the sample and incubated it at 65°C for 5 min and immediately chilled on ice for at least 2 min before adding 4 μl of reverse transcription reaction mix [in 10-μl reverse transcription reaction: 1 First-Strand Buffer (Invitrogen, no. 18064-014), 20 mM DTT (Invitrogen, no. 18064-014), 4 U of RNaseOUT (Invitrogen, no. 10777-019), and 10 U of SuperScript II (Invitrogen, no. 18064-014)].

Library amplification and cleanup

The PCR primers introduce the full-length adaptor sequence required for Illumina sequencing (for details, see Illumina small RNA PCR primers). PCR was performed in 12.5 μl using half of the 5′-tagged random hexamer reverse transcription sample as a template [1× KAPA HiFi HotStart ReadyMix (Kapa Biosystems, no. KK2602), and 400 nM each primer]. The final PCR-amplified library was submitted to two consecutive 1× NucleoMag NGS Cleanup and Size select magnetic beads (MACHEREY-NAGEL, no. 7449970.5) according to the manufacturer’s instructions. The final library was eluted in 20 μl of 10 mM Trizma hydrochloride solution (Sigma-Aldrich, no. T2319).

RNA-seq data analysis

CelSeq2 RNA-seq reads were mapped to either the GRCm38 mouse genome or the GRCh38 human genome using the Subread aligner (54) and assigned to genes using scPipe (55) with ENSEMBL v86 annotation. Gene counts were exported as a matrix by scPipe with unique molecule identifier–aware counting and imported into R (56). For 100-cell bulk samples, cells were removed from further analysis if they failed to achieve 5000 total counts. In single-cell samples, cells were removed using the detect_outlier function from the scPipe. In 100-cell bulk samples, genes were filtered out if they failed to achieve five counts per million in three samples, and in single cells, genes were required to be expressed in at least five samples and an average of two or more counts in those samples. Heatmaps were generated on normalized expression values using heatmap2 from the gplots package with row normalization. Dimensionality reduction was performed on normalized log2-count-per-million expression values with size factors from computeSumFactor in scran (57). MD plots were generated using limma (58) and Glimma (59). Differential expression analyses used negative binomial GLMs and likelihood ratio tests from edgeR (60, 61). Genes with a false discovery rate (62) of <0.0 5 were deemed differentially expressed. We used monocle2 (v2.12.0) (63) to estimate single-cell trajectories and create subsequent visualizations. Starting from the scran-normalized data, the 1000 most highly variable genes were selected and used as input to construct a Discriminative Dimensionality Reduction with Trees (DDRTree). The DDRTree was then used to estimate a pseudotime trajectory by the orderCells() function in monocle2, rooting the tree using the stage 1 cells. Gene set testing was performed on the DEGs using goana (64) and kegga from the limma package for GO (65) and KEGG (66) pathway database, respectively.


Fig. S1. Full heatmap data for Fig. 3C.

Fig. S2. GO and KEGG pathway enrichment analysis in MAIT cell development.

Fig. S3. Trajectory analysis on sorted single cells.

Fig. S4. MD profiles of DEGs of Lef1+ stage 3 MAIT cells compared with other subsets.

Fig. S5. Conventional TCRβ+, NKT, and γδT cell compartments in SAP, Satb1m1Anu/m1Anu, SKG, Cxcr6 KO, and Ccr7 KO mice.

Fig. S6. Abnormal Foxp3+ Treg phenotype of mice homozygous for the Satb1m1Anu allele.

Fig. S7. MAIT cell development appears to be normal in Bach2-deficient mice.

Fig. S8. Full heatmap data for Fig. 5.

Fig. S9. DEGs between bulk splenic and thymic mature MAIT cells.

Fig. S10. Flow cytometry gating strategies and representative sorting purities.

Table S1. Full list of DEGs from bulk (100 cells) purified thymic stage 3 versus stage 1 mouse MAIT cells.

Table S2. Full list of top DEGs between stage 3 and stage 1 human MAIT cells.

Table S3. List of antibodies used.


Acknowledgments: We thank the staff in the Doherty Institute, University of Queensland, University of Adelaide, the Australian National University, and the Walter and Eliza Hall Institute Animal Facilities and Biological Resource Facilities (BRF) for animal husbandry assistance. We thank T. Luke and the staff at the Doherty Institute Flow Cytometry Facility, S. Monard and the staff at the WEHI Flow Cytometry Facility, and V. Jameson and the staff at the Melbourne Brain Centre Flow Cytometry Facility for flow cytometry support. We thank Pam Schwartzberg at NIH for generating and providing SAP KO mice. We also thank S. Turner and A. Uldrich for helpful discussions and J. Bier for technical assistance. Funding: This work was supported by the National Health and Medical Research Council of Australia (NHMRC): 1083942, 1113293, and 1140126 to D.I.G.; NHMRC 1122277 and 1054925 to G.T.B.; NHMRC 1071916 to K.K.; and NHMRC 1016953, 1113904, 585490, and 1081858 and the National Institutes of Health (U19-AI100627) to C.C.G. H.-F.K. is supported by an NHMRC ECF Fellowship (1160333); D.I.G., H.-F.K., D.P.F., and D.G.P. are also supported by the Australian Research Council CE140100011. D.G.P. is supported by a CSL Centenary Fellowship; K.K. is supported by an NHMRC Senior Research Fellowship (1102792); A.K. is supported by an NHMRC Fellowship (1139607); and D.P.F., D.I.G., and G.T.B. are supported by NHMRC Senior Principal Research Fellowships (1117017, 1117766, and 1135898, respectively). This study was made possible through the Victorian State Government Operational Infrastructure Support and Australian Government NHMRC Independent Research Institute Infrastructure Support scheme. Author contributions: H.-F.K., D.G.P., and D.I.G., conceptualized and formulated the study, designed experiments, interpreted results, and wrote the manuscript. H.-F.K., D.G.P., D.A.-Z., T.B., S.S., and P.F.H. performed and interpreted experiments and bioinformatics analysis. S.R.D., I.C., L.M., C.E.W., I.E.K., Y.d’U., S.P.B., J.Y.W.M., Y.S., C.M.R., T.S., A.K., Z.C., S.N., K.K., L.K.M., S.R.M., E.K.D., D.P.F., J.M., C.C.G., M.E.R., G.T.B., and S.H.N. provided mice, human tissue samples, key reagents, and intellectual input and revised the manuscript. Competing interests: J.M. and D.P.F. are named inventors on a patent application (PCT/AU2013/000742, WO2014005194), and J.Y.W.M., Z.C., J.M., and D.P.F. are named inventors on another patent application (PCT/AU2015/050148, WO2015149130) involving MR1 ligands for MR1-restricted MAIT cells owned by the University of Queensland, Monash University, and University of Melbourne. The remaining authors declare that they have no competing financial interests. Data and materials availability: The RNA-seq profiling data that support the findings of this study have been deposited in the Gene Expression Omnibus repository with the accession number GSE137350. All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Immunology

Navigate This Article