MAIT Cells Promote Tumor Initiation, Growth, and Metastases via Tumor MR1

The tumor microenvironment (TME) contains various immune cells that can promote or suppress tumor growth (1). Conventional αβ⁺ CD8⁺ T cells, which recognize peptide antigens in the context of MHC class I, are key in mediating antitumor responses and thought to be the main cells targeted by current immune-checkpoint inhibitors. However, there is now a growing appreciation that unconventional innate-like T cells, which do not recognize classic peptide antigens, can also be important in regulating tumor immunity, including γδ T cells, natural killer (NK) T cells, and mucosal-associated invariant T (MAIT) cells (reviewed in ref. 2).

MAIT cells are developmentally and functionally dependent on the MHC class I–related protein 1 (MR1) and the host microbiota (3–5). Unlike classic MHC molecules, MR1 presents microbial-derived metabolites to MAIT cells and is evolutionarily conserved across mammals, suggesting the importance of the MAIT TCR–MR1 axis in immunity (6). MR1 presents microbial-derived metabolites that can activate or inhibit MAIT cells (7–9). To date, 5-(2-oxopropylideneamino)-6-d-ribitylaminouracil (5-OP-RU), an adduct of a key riboflavin (vitamin B2) biosynthesis pathway intermediate (5-A-RU) and a glycolysis pathway intermediate (pyruvaldehyde), is the most potent stimulatory MAIT cell antigen (7, 8). In contrast, folate (vitamin B9)-based MR1 ligands such as 6-formyl pterin (6-FP) and its derivative acetyl 6-formylpterin (Ac-6-FP) inhibit MAIT cell activation (9–12). It has also been reported that other aromatic molecules including drugs, drug metabolites, and drug-like molecules (e.g., salicylates, diclofenac and its metabolite) acted as MR1-binding ligands that inhibited or activated MAIT cells (12). In the context of microbial infection, activation of MAIT cells through MR1–TCR engagement results in the rapid secretion of various cytokines including IFNγ, TNF, and IL17A (7, 9, 11, 13–18). Alternatively, MAIT cells can be activated in an MR1-independent, cytokine-dependent manner (18–22).

In mice, MAIT cells are generally low in frequency, although they are enriched in mucosal sites such as the lungs, liver, and intestine (18, 23, 24). In contrast, they are more abundant in humans, representing on average approximately 5% of total blood T cells, 10% of CD8⁺ T cells, and up to 45% of liver T cells (15–17). The majority of mouse MAIT cells are CD4 and CD8 double-negative, although the frequency of MAIT cell subsets varies in a tissue- and strain-specific manner (18, 25). In contrast, most human MAIT cells are CD8 single-positive (17, 25, 26) and potentially may be erroneously identified as conventional CD8⁺ T cells based on an assessment of CD8 positivity alone (e.g., in IHC and standard flow cytometry). Although MAIT cells were previously defined by the coexpression of various cell-surface markers and/or transcription factors, with the development of MR1–Ag tetramers (7, 17), they can now be definitively identified.

MAIT cells have been reported to have either protective or pathogenic roles in bacterial and fungal infection (27); however, their role in antitumor immunity is unknown. Clinically, a number of studies have reported that MAIT cells were present in different human tumors, including colorectal cancer (28–31), kidney and brain cancer (32), liver cancer (33), and multiple myeloma (34, 35). However, these studies identified MAIT cells by staining for TCRVα7.2, in combination with CD3 and CD161, except for the study by Gherardin and colleagues, where a human MR1 tetramer was used (34). In colorectal cancer, the frequency of MAIT cells was reported to be higher in cancer tissues compared with normal adjacent tissues (28–31). In one study, high levels of MAIT cell infiltration in colorectal tumors were associated with poor clinical outcome (29). In contrast, Shaler and colleagues reported that MAIT cell frequencies were lower in colorectal liver metastases compared with healthy liver tissues (36). Similarly, another study reported that MAIT cell frequencies decreased in hepatocellular carcinoma compared with healthy liver tissues and correlated with poor prognosis (33). In the bone marrow of patients with multiple myeloma, MAIT cell frequencies were found to be decreased compared with healthy controls (35). Currently, MAIT cells are proposed to have antitumor activity, because in vitro it was demonstrated that they displayed cytolytic activity against tumor cells when cultured at a high effector:target ratio in the presence of MAIT cell antigen or following PMA/ionomycin stimulation (30, 31, 34). Preclinically, the in vivo functional role of MAIT cells in promoting or suppressing antitumor immunity has not been demonstrated. In this study, we investigated the role of MAIT cells in tumor initiation and control of experimental lung metastases, subcutaneous tumor growth, and their mechanism of action.

Critically, MAIT cell regulation of antitumor immunity has never been investigated in vivo. The role of MAIT cells in experimental tumor metastasis to the lung was investigated by comparing metastasis in C57BL/6 wild-type (WT) and C57BL/6 Mr1^−/− mice, which lack MAIT cells (ref. 5; Fig. 1). Given that MAIT cells respond to microbial metabolites and the microbiota between WT and Mr1^−/− mice was reported to be different (24), we first set up an experiment where we injected B16F10 melanoma cells intravenously into WT or Mr1^−/− mice that were housed separately or cohoused for at least 4 weeks prior to injection. In both settings, the numbers of lung metastases were significantly reduced in Mr1^−/− mice compared with WT mice, suggesting that MAIT cells promoted metastasis (Fig. 1A). This suggested the reduction in metastases was not due to differences in microbiota between WT and Mr1^−/− mice. Nevertheless, to minimize any potential confounding effects of the microbiota on our experiments, cohoused WT and Mr1^−/− mice were used in all the in vivo experiments performed in this study.

To confirm suppression of lung metastases occurred in another tumor model, we injected LWT1, a BRAF^V600E-mutant melanoma cell line, into WT and Mr1^−/− mice (Fig. 1B). Again, the numbers of LWT1 metastases were significantly reduced in Mr1^−/− mice compared with WT mice. Given the critical roles of NK cells and IFNγ in the control of experimental lung metastases (37), we next depleted NK cells or neutralized IFNγ in B16F10 tumor–bearing WT or Mr1^−/− mice (Fig. 1C and D). Although control Ig (cIg)–treated tumor-bearingMr1^−/− mice had fewer lung metastases compared with cIg-treated WT mice, tumor-bearing WT and Mr1^−/− mice depleted of NK cells or neutralized of IFNγ displayed a similar higher level of lung metastases. These results demonstrated that the reduced metastasis observed in Mr1^−/− mice was critically dependent upon NK cells and IFNγ. To further demonstrate that loss of MAIT cells was responsible for the reduction in lung metastases, we generated four-way bone marrow (BM) chimeric mice from WT or Mr1^−/− donors (Fig. 1E). BM reconstitution was confirmed in the recipient mice with engraftment efficiency greater than 95% and it was confirmed that MAIT cells were lacking in mice transferred with Mr1^−/− BM (data not shown). Only mice reconstituted with Mr1^−/− BM displayed reduced tumor metastases (Fig. 1E), suggesting that loss of hematopoietic MR1 (and MAIT cell loss) contributed to lung metastasis suppression in Mr1^−/− mice. In addition to experimental lung metastases, NK cells are also critical for protecting the host from methylcholanthrene (MCA) carcinogen–induced fibrosarcoma (38). Therefore, we injected WT and Mr1^−/− mice with a low (25 μg) or high dose (300 μg) of MCA and monitored their long-term survival (Fig. 1F and G). At a low dose of MCA, Mr1^−/− mice displayed greater resistance to MCA-induced fibrosarcoma than WT mice, with 5 of 21 Mr1^−/− mice and 12 of 21 WT mice developing tumors (Fig. 1F and G). Similarly, this resistance was also observed in Mr1^−/− mice injected with a high dose of MCA compared with WT mice, suggesting that a lack of MAIT cells permitted for better protection against tumor initiation.

To determine how loss of MAIT cells affected tumors growing outside of mucosal sites, we subcutaneously injected SM1WT1 melanoma cells, from which the LWT1 melanoma line was derived, into WT or Mr1^−/− mice. Again, we observed a significant reduction of SM1WT1 tumor growth in Mr1^−/− mice compared with WT mice (Fig. 1H). Furthermore, growth suppression was dependent on NK cells, CD8⁺ T cells, and IFNγ, as SM1WT1 tumor–bearing Mr1^−/− mice depleted or neutralized of this cell type or cytokine, respectively, were unable to suppress tumor growth compared with cIg-treated groups (Fig. 1I–J).

To further demonstrate that MAIT cells had a tumor-promoting function, we asked whether adoptive transfer of MAIT cells into Mr1^−/− mice reversed the reduction in lung metastases (Fig. 2A and B). MAIT cells were identified and sorted using MR1 tetramers (Supplementary Fig. S1A). Using the protocol as indicated in the schematic (Fig. 2A), whole splenocytes (containing MAIT and conventional T cells) from WT mice were cultured in the presence of the activating MAIT cell ligand 5-OP-RU and IL2 for 6 to 7 days (8, 24). Subsequently, 2 × 10⁵ MAIT cells (sorted on MR1–5-OP-RU tetramer) or the remaining Mr1–5-OP-RU tetramer-negative T cells (termed cT) were injected into Mr1^−/− mice 1 day before B16F10 intravenous injection (Fig. 2B). Although Mr1^−/− mice have decreased lung metastases compared with WT mice, this phenotype was strikingly lost in Mr1^−/− mice that received adoptively transferred MAIT cells, and they displayed similar numbers of lung metastases as WT mice (Fig. 2B). In contrast, transfer of cT cells into Mr1^−/− mice did not reverse the observed reduction in lung metastases (Fig. 2B). Upon metastasis enumeration at day 14, we were still able to detect MAIT cells in the lungs of MAIT→Mr1^−/− mice (Supplementary Fig. S1B).

We also expanded MAIT cells from Tcrd^−/− mice, because they have a higher proportion of MAIT cells compared with WT mice (Supplementary Fig. S1C–S1F), and thus logistically it was easier to expand sufficient MAIT cells for subsequent adoptive transfer experiments. To determine if the metastasis-promoting function of MAIT cells was mediated through direct interaction with tumor cells to promote their growth or indirectly through suppression of NK cells, which are the critical effector cells in the control of experimental lung metastases (Fig. 1C; ref. 39), we performed a similar adoptive transfer experiment of MAIT cells or cT cells derived from Tcrd^−/− mice into Rag2cγ^−/− mice, which lack T, B, and NK cells. In contrast to Mr1^−/− mice, adoptive MAIT cell transfer into Rag2cγ^−/− mice did not further increase B16F10 lung metastases in these mice compared with the control media–injected group (Fig. 2C). Similarly, cT cell transfer did not directly promote or suppress lung metastases. Importantly, we also demonstrated that MAIT cells expanded from Tcrd^−/− mice reversed the suppression of B16F10 metastases in Mr1^−/− mice (Supplementary Fig. S1G) similar to WT MAIT cells (Fig. 2B). Overall, our data demonstrated that MAIT cells had metastasis-promoting function.

In both mouse and human, MR1 mRNA transcripts have been reported to be present in different tissues, immune cells (such as antigen-presenting cells), and cell lines (40), although its expression in human cell lines can be upregulated and stabilized following incubation with the MAIT cell ligands 5-OP-RU or Ac-6-FP (8, 41, 42). However, surface MR1 expression on immune cells has been reported to be extremely difficult to detect by flow cytometry (41, 42), which we confirmed, even when two different clones of anti-MR1 (26.5, 8F2.F9) were used (Supplementary Fig. S2A–S2F). As previously reported, low levels of MR1 were detected intracellularly in double-positive mouse thymocytes (42), but not in other immune cells such as NK cells and T cells. In contrast, whether mouse tumor cell lines expressed surface MR1 has not been examined. Therefore, we first determined the cell-surface expression of MR1 on a panel of 11 different mouse tumor cell lines using the 26.5 clone of anti-MR1, which cross-reacts with both mouse and human MR1 (Fig. 2D; Supplementary Fig. S3A and S3B; ref. 42). Basally, the level of MR1 was almost undetectable on the surfaces of all tumor cell lines examined (Fig. 2D; Supplementary Fig. S3A and S3B). When we incubated these cell lines with 100 nmol/L of the activating MAIT cell ligand 5-OP-RU and assessed for MR1 surface expression 4 hours later, interestingly, we observed MR1 expression was substantially upregulated on the cell surface of B16F10 and LWT1 melanoma cells (Fig. 2D). MR1 expression was also upregulated to a varying extent on MC38-parental or MC38-OVA–expressing colorectal adenocarcinoma cells, HcMel3, HcMel12, and SM1WT1 melanoma cells, and MCA1956 fibrosarcoma cells (Supplementary Fig. S3A). However, MR1 was not upregulated on RM1 prostate carcinoma cells, 4T1.2 mammary carcinoma cells, or 3LL lung carcinoma cells (Supplementary Fig. S3B). We selected the 100 nmol/L concentration and 4-hour timepoint to assess MR1 upregulation because we demonstrated in a separate dose titration and time kinetic experiments using B16F10 that these conditions optimally upregulated surface MR1 on these cells (as little as 10 nmol/L ligand upregulated MR1; Supplementary Fig. S3C and S3D). The kinetics of MR1 upregulation on B16F10 cells was similar to what was previously reported for human C1R cells (B-cell lymphoblastoid cell line; ref. 41). Similarly, MR1 was generally negative or very low across a range of different human tumor cell lines, which became upregulated following incubation with 5-OP-RU (Supplementary Fig. S3E–S3G). Next, we assessed how upregulation of MR1 on B16F10 or LWT1 cells prior to intravenous injection to WT or Mr1^−/− mice affected the number of lung metastases. Strikingly, we observed a significantly increased number of lung metastases in WT mice injected with 5-OP-RU–treated compared with DMSO control–treated B16F10 cells and LWT1 cells, and this increase was lost when tumor cells were injected into Mr1^−/− mice (Fig. 2E and F). Interestingly, RM1, which does not upregulate MR1 following incubation with 5-OP-RU, displayed a similar number of metastases when injected into WT or Mr1^−/− mice (Supplementary Fig. S3H). Overall, these results suggested that mouse tumor cell lines that are capable of expressing surface MR1 and presenting activating ligands activate MAIT cells to mediate their suppressive function. Furthermore, our data also suggest that the tumor MR1–MAIT interaction rather than host MR1–MAIT interaction was important in promoting experimental lung metastases.

We next investigated the consequences of MR1 upregulation on B16F10 on NK-cell effector function (Fig. 3). B16F10 cells were incubated with 5-OP-RU or DMSO vehicle control for 4 hours, before cells were washed and intravenously injected into WT mice. On day 5, the lungs were harvested, and NK-cell effector function, as measured by IFNγ production and degranulation (CD107a), was assessed (Fig. 3A). Strikingly, the proportion of NK cells producing IFNγ and their expression levels as measured by mean florescence intensity (MFI) were significantly decreased in the lungs of mice bearing 5-OP-RU–treated compared with DMSO-treated B16F10 cells (Fig. 3B and C). In contrast, NK-cell function was not suppressed in the lungs of Mr1^−/− mice challenged with 5-OP-RU–treated B16F10 cells compared with DMSO-treated B16F10 cells (Fig. 3B and C). Similarly, the proportions of NK cells expressing CD107a and their expression levels were also significantly decreased in the lungs of mice injected with 5-OP-RU–treated B16F10 cells compared with DMSO-treated B16F10 cells (Fig. 3D and E). We observed a similar suppression of NK-cell effector function when the lungs of mice injected with B16F10 cells treated with two different doses of 5-OP-RU were harvested and analyzed at a later time point on day 11 (Supplementary Fig. S4A–S4E). In WT mice that received 5-OP-RU–treated LWT1, we again observed suppression of NK-cell effector function but this did not manifest in Mr1^−/− mice (Fig. 3F–I). Interestingly, we also observed an increased proportion of NK cells producing IFNγ derived from the lungs of Mr1^−/− mice injected with DMSO-treated LWT1 compared with similarly injected WT mice. Overall, these data demonstrated that the interaction of MAIT cells with MR1-expressing B16F10 or LWT1 tumor cells played a critical role in suppressing the antimetastatic activity of NK cells.

Given that MAIT cells can rapidly secrete effector cytokines, we therefore determined whether their activation status and cytokine profile were modulated in the lungs of mice injected with 5-OP-RU– or DMSO-treated B16F10 cells (Fig. 4A). CD69 is a commonly used marker to determine MAIT cell activation status and tissue residency (23), and we observed an increased proportion of MAIT cells that expressed CD69, IL17A, and TNF in the lungs of mice injected with 5-OP-RU–treated compared with DMSO-treated B16F10 tumor cells following restimulation with PMA/ionomycin (Fig. 4B–F). Interestingly, the proportion of MAIT cells producing IFNγ did not change (Fig. 4G). In contrast, in vitro coculture of purified splenic MAIT cells expanded using the same condition as described for Fig. 2 with 5-OP-RU–treated compared with DMSO-treated B16F10 cells produced more IFNγ, IL17A, and TNF (Supplementary Fig. S5A–S5C). This response was specific as no increase in cytokine production was observed in conventional T cells cultured with 5-OP-RU–treated compared with DMSO-treated B16F10 (Supplementary Fig. S5D–S5F). Similar to previous reports that human MAIT cells displayed cytotoxic capability (30, 31, 34), we observed a decrease in the number of 5-OP-RU–treated Cell Trace Violet–labeled B16F10 cells following coculture with MAIT cells (Supplementary Fig. S5G) but not with conventional T cells (Supplementary Fig. S5H). These data suggested that the antitumor phenotype of MAIT cells observed in vitro may not reflect their physiologic role in vivo.

Given the increased proportion of IL17-producing MAIT cells, we asked how the loss of IL17A affected NK-cell effector function. WT or Il17a^−/− mice were injected with 5-OP-RU– or DMSO-treated B16F10 cells, and 5 days later the lungs were harvested for NK-cell analysis. Although the proportions of IFNγ-producing and CD107a-expressing NK cells were significantly reduced in WT mice, as demonstrated earlier, we did not observe this reduction in Il17a^−/− mice (Fig. 4H and I), suggesting that IL17A may be one mechanism by which the antimetastatic activity of NK cells is suppressed. To confirm IL17A was derived from MAIT cells, we adoptively transferred MAIT cells from WT or Il17a^−/− mice into Mr1^−/− mice bearing B16F10 lung metastases (Fig. 4J). MAIT cells that were unable to produce IL17A were significantly less effective at reversing the suppressive phenotype compared with WT MAIT cells, demonstrating that IL17A was partially required for suppression. By inference, there must be additional MAIT cell mechanisms because full reversal of metastasis suppression was not achieved by Il17a^−/− MAIT cells.

Our data suggested the importance of MR1 expression on B16F10 cells to activate the suppressive function of MAIT cells. Using CRIPSR/Cas9, three different single-guide RNAs (sgRNA) targeting the mouse MR1 gene were designed and transfected into B16F10 cells, respectively, whereas transfection of B16F10 cells with an empty vector served as a control (Fig. 5). As shown in Fig. 5A, all three sgRNAs effectively knocked out MR1 in B16F10 cells (B16F10-MR1^KO) compared with vector control–transfected B16F10 tumor cells. Using in vitro assays, we confirmed that loss of MR1 in B16F10 cells did not intrinsically affect their biology, as we observed no changes in their proliferation or migration ability (Supplementary Fig. S6A–S6B). Similarly, coculture of B16F10 cells with 5-OP-RU did not alter the biology of these cells as measured by their proliferative or migratory assays (Supplementary Fig. S6C–S6D). Next, we intravenously injected the three B16F10-MR1^KO or vector control B16F10 tumor cell lines into WT and Mr1^−/− mice and determined their lung metastasis burden 14 days later (Fig. 5B). Strikingly, the number of metastases was dramatically reduced in WT mice injected with the B16F10-MR1^KO cell lines compared with those that received vector control B16F10 (Fig. 5B). Furthermore, for all three B16F10-MR1^KO cell lines, we observed no further decrease in metastases in the Mr1^−/− mice compared with WT mice (Fig. 5B). We demonstrated this decrease in metastases was due to the loss of MR1 and not caused by the transfection process, as the number of metastases was similar in WT mice injected with parental or vector control B16F10 (Supplementary Fig. S6E). In vivo, we also confirmed that the loss of MR1 on B16F10 cells did not intrinsically affect their ability to form lung metastases, as similar numbers were observed between Rag2cγ^−/− mice intravenously injected with B16F10 vector control or B16F10-MR1^KO (sgR3) (Fig. 5C).

When we overexpressed MR1 in B16F10-MR1^KO (sgR3) cells, we showed we could again upregulate MR1 on the cell surface following coculture with 5-OP-RU (Fig. 5D). Interestingly, injection of B16F10 overexpressing MR1 into WT mice increased the number of metastases compared with the GFP vector control–transfected B16F10-MR1^KO (sgR3) cell line even in the absence of pretreatment with 5-OP-RU (Fig. 5E). One possibility is that MR1 overexpression may directly suppress NK-cell function given its similarity to MHC I. However, this is unlikely given that the number of metastases in Mr1^−/− mice injected with B16F10 or B16F10-MR1^KO was similar (Fig. 5B). Another possibility is that the increased availability of MR1 allows more efficient loading of endogenous stimulatory MAIT ligands, which consequently activates more MAIT cells. Some MR1 binding ligands such as 5-OP-RU upregulate and activate MAIT cells, whereas other ligands such as Ac-6-FP upregulate MR1 (Supplementary Fig. S6F) and inhibit MAIT cells (10, 11). To determine whether activation of MAIT cells was required for their suppression of NK cells, Ac-6-FP– or vehicle-treated B16F10 cells were injected into WT or Mr1^−/− mice (Fig. 5F). Strikingly, we observed a decrease in the number of metastases in WT mice injected with Ac-6-FP–treated compared with vehicle-treated B16F10 tumor cells. This level of reduction in metastases was similar to Mr1^−/− mice injected with Ac-6-FP–treated or vehicle-treated B16F10 cells (Fig. 5F). Intriguingly, these data suggested that endogenous MAIT ligands derived from either the microbiota, tumors, and/or the TME activated MAIT cells that can be blocked with an inhibitory MAIT ligand. Overall, our data indicated the potential for tumors to upregulate MR1 and activate MAIT cell immunoregulatory function.

In addition to their use in flow-cytometry analysis, the 26.5 and 8F2.F9 clones of anti-MR1 block MAIT cells from interacting with MR1 (42, 43). Therefore, we asked whether the effect we observed in Mr1^−/− mice was recapitulated with these MR1-blocking antibodies (Fig. 6). MR1 blockade by clone 26.5 on days −1, 0, 3, and 7, relative to B16F10 intravenous injection, effectively suppressed lung metastases to a similar level as observed in Mr1^−/− mice (Fig. 6A). Furthermore, we confirmed the specificity of 26.5, as no further decrease in the number of lung metastases was observed between tumor-bearing Mr1^−/− mice treated with 26.5 or cIg (Fig. 6A). Next, we compared whether giving three or four doses of clone 26.5 or 8F2.F9 (days −1, 0, 3 vs. days −1, 0, 3, and 7) equivalently suppressed metastases (Fig. 6B). Interestingly, suppression of lung metastases was more effective when four doses of 26.5 were given compared with three doses (Fig. 6B). In contrast, administering three or four doses of 8F2.F9 equivalently reduced B16F10 lung metastases, and this suppression appeared to be superior to that observed using the 26.5 clone (Fig. 6B). We also confirmed the specificity of the 8F2.F9 clone for MR1, as no further decrease in the number of lung metastases was observed between tumor-bearing Mr1^−/− mice treated with 8F2.F9 or cIg (Fig. 6B). The lack of any further decrease in lung metastases in Mr1^−/− mice treated with 26.5 or 8F2.F9 also suggested that these antibodies probably do not directly affect B16F10 metastasis (Fig. 6A and B). Furthermore, we again demonstrated the importance of blocking MR1 on tumors, because the numbers of B16F10-MR1^KO (sgR3) lung metastases were not reduced between 8F2.F9- and cIg-treated mice compared with similar groups of treated mice bearing B16F10 parental metastases (Fig. 6C). In addition, we showed that 8F2.F9 treatment significantly reduced the numbers of LWT1 lung metastases compared with cIg treatment (Fig. 6D).

In mice bearing established subcutaneous MCA1956 or SM1WT1 tumors (Fig. 6E and F), treatment with four doses of anti-MR1 significantly suppressed tumor growth compared with cIg-treated groups. Finally, we assessed the therapeutic potential of anti-MR1 therapy in the treatment of established de novo MCA-induced fibrosarcomas (Fig. 6G–I). Although cIg-treated fibrosarcomas grew rapidly and all mice succumbed to their tumors (Fig. 6G), anti-MR1 reduced the growth of most tumors (Fig. 6H and I) and caused the complete rejection of 3 of 21 tumors (Fig. 6G). The therapeutic efficacy of anti-MR1 as a monotherapy was impressive compared with historical single immunotherapies that have been evaluated in this exact same model, including anti–PD-1 (44).

The therapeutic efficacy of anti-MR1 in suppressing the growth of subcutaneous SM1WT1 tumors (Fig. 6F) suggested that MAIT cells may be present in these tumors. One day after the second treatment with anti-MR1 or cIg, MAIT cells were detected in the tumor-infiltrating lymphocytes (TIL) of SM1WT1 tumors, but their proportion was low among all T cells (Fig. 7A and B). Similar proportions of MAIT cells were observed in TILs regardless of cIg or anti-MR1 treatment. Using flow cytometry, we also assessed TILs from end-stage SM1WT1 tumors derived from mice in Fig. 6F. Interestingly, we observed a significant increase in the proportion of total CD45.2⁺ immune cells (Fig. 7C), CD8⁺ T cells (Fig. 7D), and NK cells (Fig. 7E) in the anti-MR1–treated compared with cIg-treated groups. Furthermore, we observed an increased proportion of IFNγ-producing and CD107a-expressing CD8⁺ T cells and NK cells from the TILs of anti-MR1–treated compared with cIg-treated mice following restimulation with PMA/ionomycin (Fig. 7F–I). Overall, these data suggested that the interaction of MR1 on tumors and MAIT cells might be critical in suppressing antitumor immunity in mucosal and nonmucosal sites, and blocking MR1 may represent a new strategy for cancer immunotherapy.

In this study, we demonstrated for the first time that MAIT cells promoted tumor initiation, growth, and metastasis. Using two experimental mouse models of lung metastasis, B16F10 and LWT1, a significant decrease in the number of lung metastases was observed in Mr1^−/− mice compared with WT mice, and this reduction was replicated using two different MR1-blocking antibodies. In addition, Mr1^−/− mice were more resistant to carcinogen-induced fibrosarcoma development compared with WT mice. Furthermore, we demonstrated, using Mr1^−/− mice and WT mice treated with anti-MR1, that loss of MAIT cells or MAIT cell activation via MR1, respectively, affected tumors growing outside of mucosal sites such as subcutaneously inoculated MCA1956 and SM1WT1 tumors and de novo MCA-induced fibrosarcomas. By performing adoptive transfer of purified MAIT cells into B16F10 tumor–bearing Mr1^−/− mice, we reversed the reduction in lung metastases in these mice, whereas conventional T-cell transfer was without effect. We also demonstrated that MAIT cells promoted lung metastases by engaging MR1 on tumor cells resulting in an IL17A-dependent suppression of NK-cell effector function. The importance of MR1 expression on tumor cells was further demonstrated by the increased number of metastases in mice injected with 5-OP-RU–pretreated B16F10 tumor cells. In contrast, deletion of MR1 on B16F10 or pretreatment of B16F10 with Ac-6-FP dramatically reduced the number of lung metastases, whereas reexpression of MR1 reversed this phenotype. These data contrasted with previous dogma that MAIT cells were directly cytotoxic to tumor cells, rather illustrating in vivo that MAIT cells acted to suppress NK-cell effector function and thereby promote tumor metastasis. This is a new concept in that NK cells are known only to be suppressed by regulatory T cells (45–47), or myeloid/granulocyte populations (48–51), or by MHC class I binding on tumors (52). Here, we have described an important MHC class I–like recognition system that can also affect NK-cell effector function.

In our study, we demonstrated that cell-surface expression of MR1 was highly upregulated in some mouse and human tumor cell lines following incubation with 5-OP-RU in a time- and dose-dependent manner, even when MR1 was undetectable at a basal level. Why certain mouse and human tumor cell lines were unable to upregulate MR1 remains to be investigated. Potential explanations include the absence of a functional WT MR1-encoding gene, or the lack of β2M, which is required for MR1 surface expression (40). Interestingly, the lack of MR1 expression on RM1 tumor cells also corresponded with no decrease in lung metastases in Mr1^−/− mice compared with WT mice, suggesting the tumor MR1–MAIT interaction was not operating in this TME. It also suggested that host MR1–MAIT interactions were not critical in the RM1 TME. Indeed, the critical importance of MR1 expression on tumors was clearly illustrated when we deleted MR1 in B16F10 cells to generate three different B16F10-MR1^KO cell lines. Here, we observed a very striking decrease in B16F10-MR1^KO lung metastases compared with similar groups of WT mice injected with control-transfected B16F10 cells. Reexpressing MR1 in B16F10-MR1^KO cells increased the number of lung metastases compared with control-transfected cells. This phenotype was not due to an intrinsic effect mediated by MR1 in B16F10 cells, as in vitro assays demonstrated these cells displayed similar proliferative and migratory ability as vector control–transfected or parental B16F10 cell lines. Furthermore, injecting B16F10-MR1^KO or control-transfected B16F10 cells into Rag2cγ^−/− mice, which lacked T, B, and NK cells, generated a similar number of lung metastases. This demonstrated the equivalent capability of B16F10-MR1^KO– and B16F10-MR1–sufficient cells to form metastases. Finally, we showed that upregulating MR1 on tumor cells increased their metastatic capability. WT mice injected with 5-OP-RU–pulsed B16F10 cells displayed an increased number of lung metastases compared with those that received DMSO-treated B16F10 cells, even when an equal number of both cells were injected. Consequently, NK-cell effector functions including IFNγ production and their degranulation capability were reduced in the lungs of WT mice challenged with 5-OP-RU–treated compared with DMSO-treated B16F10 or LWT1.

Critically, our data also demonstrated that the activation of MAIT cells was required for their suppressive function. Unlike 5-OP-RU, other MAIT cell ligands such as Ac-6-FP and 6-FP, which are derived from vitamin B9, upregulate MR1, but inhibit MAIT cell activation (10, 11). Strikingly, Ac-6-FP–pretreated tumor cells displayed decreased lung metastases compared with vehicle-treated tumor cells, demonstrating that activation of MAIT cells rather than the mere upregulation of MR1 was critical for suppressing antitumor immunity. This also suggested the potential to utilize inhibitory MAIT cell antigen(s) as a therapeutic to suppress MAIT cell activation in tumors. Going forward, a key question is the identification of the source and type of endogenous antigen(s) in mice that activates MAIT cells in the TME. In mucosal sites such as the lungs, the presence of microbial communities containing a complex diversity of bacteria in the lower respiratory tract is now appreciated (53) and can contribute to various pathologies. Using a genetically engineered mouse model of lung cancer, Jin and colleagues demonstrated that the local microbiota associated with tumor growth promoted inflammation and cancer progression via lung-resident γδ T cells (54). We speculate that the presence of tumor cells in the lungs may change the composition of the local microbiome, which may contribute to the release of activating MAIT cell ligands that can upregulate MR1 on tumor cells. Interestingly, a recent study reported that the microbial metabolites in mice controlled the thymic development of MAIT cells (55). Furthermore, the authors demonstrated the ability of 5-OP-RU to rapidly travel from mucosal surfaces to the thymus, where it was captured by MR1 (55). Going forward, it will be important to define the nature of the activating MAIT cell ligands and investigate whether they derived from tumors themselves, the local or gut microbiota, or the TME.

We showed that MAIT cells did not directly promote tumor metastasis, as adoptive transfer of MAIT cells did not increase B16F10 lung metastases in Rag2cγ^−/− mice that lacked T, B, and NK cells. In ex vivo cytokine analysis of MAIT cells derived from the lungs of mice injected with 5-OP-RU–treated B16F10 cells, the proportion of MAIT cells that produced IL17A and TNF, but not IFNγ, increased. We further showed that adoptive transfer of MAIT cells that were unable to produce IL17A into tumor-bearing Mr1^−/− mice partially reversed the suppressive phenotype. This suggested that IL17A derived from MAIT cells was partially involved in suppressing the antimetastatic activity of NK cells, although clearly other molecules are also involved. Adoptive transfer experiments of MAIT cells deficient in different cytokines could be performed to identify which other cytokines in addition to IL17A are involved in suppressing NK-cell antimetastatic function. Overall, our results clearly demonstrated that tumor cells that express and upregulate MR1 with activating ligands do activate MAIT cell suppression of NK-cell antimetastatic activity.

Despite the difficulty in detecting surface MR1 expression on normal cells by flow cytometry, MR1 was previously reported to be expressed on various immune and epithelial cells as inferred from functional assays using MR1-blocking antibodies (41, 42). Thus, it was possible that MAIT cells could be activated by MR1 expressed on both hematopoietic and nonhematopoietic cells. However, the similar number of lung metastases observed in WT→WT and WT→ Mr1^−/− chimeric mice implied that MR1 expression on nonhematopoietic cells was not important for MAIT cell development and activation in a tumor setting. Furthermore, given that the numbers of B16F10-MR1^KO lung metastases were the same between WT and Mr1^−/− mice, this suggests any role of MR1 on host immune and nonimmune cells in activating MAIT cells might be comparatively minor. Our data also suggested that the ability of MAIT cells to suppress NK and/or CD8⁺ T cells may not require MR1 on these cell types, given that they did not appear to express surface MR1 and MAIT cell suppression still occurred in Mr1^−/− mice, where NK cells and CD8⁺ T cells lacked MR1. Potentially, MAIT ligands can be administered intranasally into tumor-bearing mice to assess which immune cells, if any, in addition to tumor cells, upregulate surface MR1. In humans, basal MR1 expression was generally low or undetectable on many of the tumor cell lines we examined, but was upregulated following culture with 5-OP-RU. Going forward, MR1 protein expression on different tumors and tumor-infiltrating immune cells needs to be characterized and correlated with progression-free survival and overall survival in patients with cancer in general and in those receiving immunotherapy. As most human MAIT cells are CD8 single-positive (18, 25), they potentially may be erroneously identified as conventional CD8⁺ T cells, and therefore it may be necessary to perform multiplex IHC to determine the presence of MR1, MAIT cells, and CD8 and NK cells in tumors.

Therapeutically, two different clones of MR1 blocking antibodies (26.5 and 8F2.F9) were as capable of reducing B16F10 and LWT1 lung metastases as loss of MAIT cells in Mr1^−/− mice. Other experiments suggested that their antimetastatic effect was probably mediated through blocking MAIT cell activation rather than having a direct antitumor effect on MR1-expressing tumor cells. Anti-MR1 alone also suppressed subcutaneous tumor growth of two transplanted tumor cell lines, MCA1956 and SM1WT1, and established de novo MCA-induced fibrosarcomas, suggesting that the MR1–MAIT pathway operates in tumors growing in mucosal or nonmucosal sites. SM1WT1 generally does not respond to PD-1 blockade. When we collected end-stage SM1WT1 tumors for flow-cytometry analysis, we observed a significant increase in the proportion of CD8⁺ T and NK cells in the TILs from anti-MR1–treated groups compared with cIg-treated groups. Furthermore, an increased proportion of these CD8⁺ T and NK cells displayed improved effector function as observed by expression of IFNγ and CD107a. Given that MR1 is highly conserved across 150 million years of mammalian evolution, blockade of the tumor MR1–MAIT cell axis in humans may represent a novel approach to relieve immune-mediated suppression of NK and CD8⁺ T cells in the TME. Future experiments and combination approaches can now be used to try to boost NK-cell and CD8⁺ T-cell antitumor activity by blocking tumor MR1 in tumors where MR1 is highly expressed. For example, blocking MR1 on these tumors might allow better retention of NK/CD8⁺ T cells and their effector functions in the context of immune-checkpoint blockade or other approaches that reduce myeloid-mediated immunosuppression. Overall, our data provide evidence that MAIT cells suppress NK-cell antimetastatic function and that blocking MR1 or saturating it with an inhibitory ligand may represent a new therapeutic strategy for cancer immunotherapy.

C57BL/6 WT and gene-targeted mice were bred in-house. C57BL/6 Mr1^−/− mice were kindly provided by James McCluskey (Melbourne University, Melbourne, Australia; ref. 5). C57BL/6 IL17A^−/− mice were provided by Geoffrey R. Hill (QIMR Berghofer Medical Research Institute, Herston, Australia; ref. 56). Tcrd^−/− mice were provided by Ian Frazer (The University of Queensland, Brisbane, Australia). Rag2cγ^−/− mice were generated at QIMR Berghofer Medical Research Institute by crossing Rag2^−/− mice with IL2Rγ^−/− mice (57). WT and Mr1^−/− mice were cohoused in the same cage for at least 4 weeks before experiment initiation unless specifically indicated. Age-matched mice were used in all experiments. All WT and gene-targeted mice used were between the ages of 6 and 14 weeks. All experiments were approved by the QIMR Berghofer Medical Research Institute Animal Ethics Committee.

The mouse melanoma cell lines B16F10, LWT1, and SM1WT1 and the MCA1956 fibrosarcoma cell line were maintained as described previously (37, 44, 58). Anti-MR1 antibody-producing hybridoma cell lines (clone 26.5 IgG2a isotype and 8F2.F9 IgG1 isotype) were maintained in complete RPMI-1640 media containing 10% FBS, 1% l-glutamine, and 1% penicillin/streptomycin (6, 42). All cell lines were routinely tested for Mycoplasma, but cell line authentication was not routinely performed.

Anti-CD8β (clone 53.5.8), anti-IFNγ (clone H22), and control IgG antibody (clone 1-1) were purchased from BioXCell or Leinco. Anti-asialoGM1 (ASGM1) was purchased from Wako Pure Chemicals. Anti-MR1 mAbs were purified from the supernatant of hybridoma cells by protein G affinity resin column (6, 42). The dose and schedule of antibody treatment are indicated in the figure legends. Biotinylated MR1–antigen monomer was conjugated with PE–streptavidin (BioLegend, catalog no. 5544061) to generate MR1 tetramers as described previously (7, 17, 18, 59). MAIT cell antigen 5-OP-RU was generated as described previously (8), whereas Ac-6-FP (60) was bought from Schircks Laboratories. Cell stimulation cocktail (PMA/ionomycin plus protein transport inhibitors) was purchased from Invitrogen Thermo Fisher Scientific (catalog no. 00-4975-93).

The indicated numbers of B16F10 and LWT1 cells were injected intravenously into the tail vein of WT or gene-targeted female or male mice, respectively. In some experiments, B16F10 and LWT1 cells were treated with 5-OP-RU (100 nmol/L, 4 hours) or Ac-6-FP (10 μmol/L, 18 hours) or their respective DMSO or ddH₂O vehicle controls before injection. Cells with viability greater than 90% were used in the experiments. Lungs were harvested on day 14, and surface tumor nodules were counted under a dissection microscope as previously described (37, 39). For therapy experiments, anti-MR1 or cIg (clone 1-1, Leinco) were injected intraperitoneally into mice at the dose and schedule indicated in the figure legends. SM1WT1 or MCA1956 (both 1 × 10⁶) tumor cells were injected subcutaneously into WT or gene-targeted male or female mice, respectively, prior to treatment with cIg or anti-MR1 at time points indicated in the figure legends. For MCA-induced fibrosarcoma, WT and Mr1^−/− male mice were injected subcutaneously in the hind flank with MCA (Sigma-Aldrich) in 100 μL of corn oil as described previously (37, 61) with the doses indicated in the figure legends. For established tumor MCA experiments, mice were treated intraperitoneally with cIg or anti-MR1 from the second palpable tumor measurement twice a week for 6 weeks as indicated in the figure legend. Mice were monitored for fibrosarcoma development over 200 to 250 days. Tumor sizes were determined by caliper square measurements of two perpendicular diameters with data represented as mean ± SEM (mm²) for each group.

BM cells were obtained from the femurs of donor C57BL/6 WT (PTPRCA and CD45.1⁺) mice and Mr1^−/− (C57BL/6 and CD45.2⁺) mice. Two doses of 5.5 Gy of whole-body irradiation were administered to recipient WT and Mr1^−/− mice at 4-hour intervals. Recipient mice were injected intravenously with 5 × 10⁶ BM cells/mouse after irradiation. Mice were provided with water containing neomycin for 4 weeks. Ten weeks after BM transplantation, mice were eye-bled and immune cells were analyzed by flow cytometry using congenic CD45.1 and CD45.2 markers to assess immune cell reconstitution before mice were used experimentally.

Naïve or tumor-bearing lungs or subcutaneous tumor single-cell suspensions were generated as described previously (37) and incubated with anti-CD16/32 (2.4G2) to block Fc receptors on ice prior to surface staining with the antibodies. The following antibodies were used for FACS analysis: anti-CD45.2 (104), anti-TCRβ (H57-597), anti-NK1.1 (PK136), anti-NKp46 (29A1.4), anti-CD45R (B220, RA3-6B2), anti-F4/80 (BM8), anti-CD69 (H1.2F3; all from BioLegend, eBioscience), and MR1 tetramers. For intracellular cytokine staining, cells were surface-stained as described above and then fixed and permeabilized with a Cytofix/Cytoperm Kit (BD Biosciences) followed by staining with anti-IFNγ (XMG1.2), anti-TNF (MP6-XT22), anti-IL17A (TC11-18H10.1), or isotype (eBio299Arm) antibody (all from BioLegend). All data were collected on a Fortessa 4 Flow Cytometer (BD Biosciences) and analyzed with FlowJo v10 Software (TreeStar, Inc.).

Single-cell suspensions from naïve or tumor-bearing lungs or subcutaneous tumors of the indicated groups were incubated in a 96-well U-bottom plate in complete RPMI-1640 media. Cells were incubated in the presence or absence of cell stimulation cocktail (PMA/ionomycin plus protein transport inhibitors; 1,000 times dilution) for 3 to 4 hours as indicated. Cells were then stained for surface markers and intracellular cytokine production. A CD107a staining assay was used to assess the degranulation status of immune cells. Briefly, anti-CD107a (1D4B, BioLegend) antibody was added to single-cell suspensions during the stimulation period before these cells were surface-stained and analyzed by flow cytometry.

MAIT cells were expanded using the protocol described by Varelias and colleagues (24). Briefly, spleens from WT, Tcrd^−/−, or Il17a^−/− mice were mashed through a 40-μm cell strainer and lysed with ACK buffer to remove red blood cells. Splenocytes were cultured in complete RPMI-1640 media containing 50 ng/mL (250 U/mL) mouse IL2 (PeproTech; catalog no. 212-12) and 100 nmol/L 5-OP-RU. On days 6 and 7, cells were harvested for sorting. Anti-CD16/32 (2.4G2) to block Fc receptors were added to the single-cell suspension before staining with biotin-B220 antibody (clone RA3-6B2, Miltenyi Biotec; catalog no. 130-101-928) followed by Streptavidin MicroBeads (Miltenyi Biotec; catalog no. 130-048-101) to deplete B220⁺ cells by MACS. The enriched splenocytes were stained with an antibody cocktail and MAIT cells (B220⁻ F4/80⁻ CD45.2⁺ TCRβ⁺ MR1–5-OP-RU tetramer⁺) and non-MAIT cells (B220⁻ F4/80⁻ CD45.2⁺ TCRβ⁺ MR1–5-OP-RU tetramer⁻) were sorted on an Aria II/Aria III (see gating strategy in Supplementary Fig. S1A). MAIT cells of 85% to 95% purity were used in subsequent experiments.

MR1 was knocked out in B16F10 cells using the CRISPR/Cas9 system. Three mouse Mr1 sgRNAs (Supplementary Table S1) were designed on the CHOPCHOP website (http://chopchop.cbu.uib.no/). Briefly, the three Mr1 sgRNAs were subcloned into PX459 vectors (Addgene; catalog no. 62988), respectively, following the one-step protocol as described by Ran and colleagues (62). PX459 plasmids containing Mr1 sgRNA and pRp GFP-expressing plasmids were cotransfected into B16F10 cells using FuGENE6 Transfection Reagent (Promega; catalog no. E2691). On the following day, cells were cultured in media containing 1 μg/mL puromycin (Sigma-Aldrich) for 2 days. The GFP⁺ cells were sorted using the FACSAria III cell sorter (BD Biosciences) into 96-well flat-bottom plate (1 cell/well) to obtain monoclonal B16F10-MR1^KO cell lines. MR1 knockout in B16F10 cells was verified by determining MR1 surface expression after 100 nmol/L 5-OP-RU stimulation for 4 hours as described above.

B16F10-MR1^KO cells were transfected with MR1-expressing pCMV6-AC-GFP plasmid (OriGene; catalog no. MG205125) or empty control plasmid (OriGene; catalog no. PS100010) using FuGENE6 Transfection Reagent. Two days later, cells were cultured in media containing 500 μg/mL Geneticin Selective Antibiotic (G418 Sulfate; Thermo Fisher Scientific, catalog no. 10131035) for 2 weeks. GFP⁺ cells were sorted to obtain a stable cell line. MR1 expression was verified by determining MR1 surface expression after 100 nmol/L 5-OP-RU stimulation for 4 hours as described above.

GraphPad Prism software was used for statistical analysis. One-way ANOVA with Tukey post hoc test was used for multiple comparisons, Mann–Whitney test for two-group comparison, and log-rank (Mantel–Cox) test for mouse survival. Differences between groups were considered to be statistically significant where the P value was less than 0.05.

M.J.S is funded by research agreements from Bristol-Myers Squibb and Tizona Therapeutics and has an advisory board role for Tizona Therapeutics and Compass Therapeutics. L.L., A.C., and D.P.F are named inventors on a patent application (PCT/AU2013/000742, WO2014005194) and J.Y.W.M., L.L., A.C., Z.C., 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 University of Queensland, Monash University, and University of Melbourne. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Yan, M.J. Smyth, M.W.L. Teng

Development of methodology: J. Yan, S. Allen, M.J. Smyth

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Yan, I. Das, J.Y.W. Mak, B.S. Meehan, A.J. Corbett, A. Varelias, M.J. Smyth

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Yan, I. Das, A. Varelias, M.J. Smyth, M.W.L. Teng

Writing, review, and/or revision of the manuscript: J. Yan, E. McDonald, J.Y.W. Mak, D.P. Fairlie, A.J. Corbett, A. Varelias, M.J. Smyth, M.W.L. Teng

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Yan, S. Allen, E. McDonald, L. Liu, D.P. Fairlie, Z. Chen, M.W.L. Teng

Study supervision: M.W.L. Teng

The authors thank Liam Town, Kate Elder, Brodie Quine, and Gabrielle Kelly for breeding and genotyping of mice throughout this study. We would also like to thank Glen Boyle and Frank Vari for kindly providing human tumor cell lines, and Sebastien Jacquelin and Xian Yang Li for CRISPR/Cas9 technical assistance. We would also like to thank Dillon Corvino, Venkateswar Addala, and John Stagg for their intellectual input and helpful discussions. M.J. Smyth was supported by an NH&MRC Research Fellowship (1078671). M.W.L. Teng was supported by an NH&MRC CDF Fellowship (1159655). J. Yan was supported by a University of Queensland International Scholarship co-funded by QIMR Berghofer. A.J. Corbett is supported by an ARC Future Fellowship (FT160100083). D.P. Fairlie is supported by an NH&MRC Senior Principal Research Fellowship (1117017) and an ARC Centre of Excellence grant (CE140100011).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.