Biased MAIT TCR Usage Poised for Limited Antigen Diversity?

Mucosal-associated invariant T (MAIT) cells are a subset of unconventional T cells that recognize the evolutionarily conserved major histocompatibility complex (MHC) class I-like antigen-presenting molecule known as MHC class I related protein 1 (MR1). Since their rise from obscurity in the early 1990s, the study of MAIT cells has grown substantially, accelerating our fundamental understanding of these cells and their possible roles in immunity. In the context of recent advances, we review here the relationship between MR1, antigen, and TCR usage among MAIT and other MR1-reactive T cells and provide a speculative discussion.


INTRODUCTION
Conventional T cells recognize peptide antigens presented by the major histocompatibility complex (MHC) class I or class II (MHC-I or MHC-II) molecules and elicit a cellular immune response to provide anti-microbial and anti-tumoral immunity for the host. Conventional T cells utilize a cell surface bound T cell receptor (TCR) to recognize peptide-MHC complexes presented on the surface of antigen-presenting cells (APCs). The TCR is a heterodimer comprised of a TCR αand β-chain, each consisting of a variable (V) and constant (C) domain. Three sets of finger-like extensions on the TCRα-and β-chain variable domains, known as complementarity determining loops (CDR1, CDR2, and CDR3), are responsible for recognizing the peptide-MHC-complex and are collectively unique to each TCR. CDR1 and CDR2 loops are germline-encoded by the V gene segments, while the CDR3 loops are formed from somatic recombination of the V gene segments with a joining (J) (V-J) and/or diversity (D) gene segment (V-D-J) during T cell development (1). Together with random nucleotide additions and deletions in the CDR3 loop gene regions, this creates an enormous diversity of TCRs, clonally distributed amongst T cells (2).
Mucosal-associated invariant T (MAIT) cells are a subset of unconventional T cells that are restricted by a monomorphic MHC-I-like molecule, known as MHC class I related protein 1 (MR1) (3)(4)(5)(6). In contrast to conventional T cells, MAIT cells express a "semi-invariant" αβ TCR, meaning they typically express the same TCRα chain paired to a preferred array of TCRβ chains (3). In humans, the "classical MAIT TCR" comprises a TCRα chain encoded by the V gene segment TRAV1-2 juxtaposed to the J gene segment TRAJ33, TRAJ20, or TRAJ12 (5, 7-9) that pairs preferentially with a TCRβ chain encoded by TRBV6 or TRBV20 V gene segment family members (7). In mice, the MAIT TCR is composed of homologous gene segments, a TRAV1-TRAJ33 TCRα chain, that pairs with a TCRβ chain composed of TRBV19 or TRBV13 V gene segments, both of which are murine orthologous segments of human TRBV6 (3,5,10). MAIT cells are abundant in healthy adults, representing on average 3% of blood T cells (11) and can be found throughout peripheral tissues (6,7,10,(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23). The frequency of MAIT cells in laboratory mice is distinctly lower than in humans, although murine MAIT cells are also found in many peripheral organs (24,25). The prototypical antigen presented by MR1 to MAIT cells is the small molecule 5-(2-oxopropylideneamino)-6-D-ribitylaminouracil (5-OP-RU), an adduct of the riboflavin biosynthetic precursor 5-amino-6-Dribitylaminouracil (5-A-RU) and methylglyoxal (26) (Figure 1). See recent reviews for details on the riboflavin biosynthesis and formation of 5-OP-RU from 5-A-RU (31,32). Riboflavin biosynthesis is absent in mammals. Thus, by recognizing 5-OP-RU (25,33,34), and potentially other riboflavin-based ligands presented by MR1 (35), MAIT cells are able to sense a broad range of riboflavin biosynthesis proficient microbes in a highly conserved, innate-like manner, reviewed in (32). Human MAIT cells stimulated with 5-OP-RU rapidly secrete T helper (Th)1 and Th17 type cytokines (11,36,37) as well as cytotoxic granules (38). In mice, lung infection with riboflavin-synthesizing bacteria or co-administration of synthetic 5-OP-RU with adjuvant leads to a significant expansion of MAIT cells with Th1/17 cytokine secreting capacity (25,34,39), enabling MAIT cells to contribute to protection against several pathogens, including Klebsiella pneumoniae (40), Mycobacterium bovis BCG (41), Francisella tularensis (39), E. coli (42), Legionella longbeachae (34), and Clostridium difficile (43). Thus, observations to date suggest MAIT cells are poised, but perhaps not limited to, protecting peripheral tissues from microbial pathogen or commensal breach. In particular, MAIT cells have recently been shown to contribute to tissue repair at barrier sites (44)(45)(46)(47). MAIT cells may also be involved in the tumoral immune response (48)(49)(50)(51)(52), however, elevated MAIT cell numbers at the tumor site in some cancers correlate with a poorer prognosis (49,52). Notably, MAIT cells appear to be subject to a similar fate as conventional T cells during the anti-tumoral immune response, namely: T cell exhaustion, altered functional response, altered frequency, and drug sensitivity (50,(52)(53)(54)(55)(56)(57). A cytokine-modulated (IL-7, IL-12, IL-18) tumor response that occurs independent of, or concurrent with, TCR stimulation should also be considered in the context of tumoral immunity, as MAIT cells are known to respond to inflammatory stimuli in this manner (15,58,59). Furthermore, MAIT cells from healthy donors can efficiently lyse MR1 proficient tumor cells presenting microbial agonists such as 5-OP-RU, suggested as a potential strategy to harness the MAIT cell response therapeutically (56). Perhaps similar in mechanism, disruption of barrier tissues (i.e., colorectal cancers) by tumors may allow invasive growth of commensal bacteria, providing a source of microbial ligand in the context of an inflammatory environment which may trigger anti-tumor MAIT cell responses (48)(49)(50)60). Much is still unknown regarding the response by MAIT cells in the tumoral environment, particularly whether tumor associated, MAIT cell specific MR1 ligands exist and the factors that might drive MAIT cell to become pro-or antitumoral. MAIT cells have, however, attracted some interest as a potential immunotherapeutic target as they possess a number of favorable attributes such as a high precursor frequency, wide tissue distribution, potent cytokine response and cytotoxicity and a donor unrestricted nature (61 (36,62). On the assumption that MR1 would likely adopt a "MHC-I-fold" (63) in the presence of ligand, Kjer-Nielsen et al. folded soluble recombinant MR1 proteins in the presence of bacterial supernatant to capture ligands in the form of stable MR1-ligand-complexes (35). This approach of ligandcapture, combined with mass-spectrometry, and subsequent genetic manipulation of the riboflavin biosynthetic pathway in bacteria, led to the discovery of the pyrimidines; 5-OP-RU and 5-(2-oxoethylideneamino)-6-D-ribitylaminouracil (5-OE-RU), and the substantially less potent, cyclised ribityllumazines; 7hydroxy-6-methyl-8-D-ribityllumazine (RL-6-Me-7-OH); and 7dimethyl-8-D-ribityllumazine (RL-6,7-diMe) as riboflavin-based, MR1-presented, MAIT cell stimulating antigens (26,27,32,35) (Figure 1). While both 5-OP-RU and 5-OE-RU were detected in the supernatant of Escherichia coli and Salmonella enterica serovar Typhimurium, 5-OP-RU is the dominant MAIT cell antigen found in these bacterial supernatants (26). Nevertheless, MR1 tetramers loaded with 5-OE-RU stain a similar proportion of T cells compared to those loaded with 5-OP-RU (26) and synthetic versions of both antigens are similarly potent (64,65). 5-A-RU co-incubated with methylglyoxal is, however, more potent in activating MAIT cells than 5-A-RU co-incubated with glyoxal (26,66), presumably due to differences in the catabolic and metabolic kinetics of 5-OP-RU and 5-OE-RU (64). 5-OP-RU has emerged as the preferred antigen for research, both as synthetic antigen and loaded in MR1 tetramers. Accordingly, MR1-5-OP-RU tetramers have become the gold standard for identifying MAIT cells (67), enabling more accurate MAIT cell detection than using antibodies specific for surrogate markers (TRAV1-2, CD161, and CD8) (11). MR1 presentation of 5-OP-RU and 5-OE-RU involves the formation of a Schiff base between the Lys43 of MR1 and a residual carbonyl group on the ligand. This Schiff base, unique to MR1 ligand presentation amongst MHC molecules, stabilizes, and guides 5-OP-RU and 5-OE-RU into their position on MR1 in the A'-pocket amongst a cradle of aromatic residues (26). Notably, even in the absence of the Schiff base, by mutating Lys43 in MR1 to Ala (68), 5-OP-RU is presented by MR1 in the identical position (26), although the stability of this complex is markedly reduced (28). Unlike the pyrimidine ligands, RL-6-Me-7-OH does not form a Schiff base with MR1, however residues within the A'pocket of MR1 help orient the ligand into a similar position (27). In a physiological context, neutralization of MR1 Lys43 by covalently bound ligand is essential for complete MR1 folding and trafficking from the endoplasmic reticulum (ER) to the cell surface (69).
All classical MAIT TCRs studied to date recognize MR1 presenting riboflavin-based antigens using a surprisingly consistent mode of docking, orthogonally over the cleft of MR1 and forming multiple contacts with aromatic residues that surround the ligand-binding site, notably residues Tyr62 and Tyr152 (26-28, 68, 70). The TCR also directly contacts the antigen, which occupies only 0.6% of the exposed surface in the MR1 cleft (70). Crucial to antigen recognition by classical MAIT TCRs is Tyr95α, a highly conserved residue in the CDR3α loop which directly contacts the 2 ′ -OH group on the antigen ribityl moiety (Figure 2A) (26)(27)(28)68). Structurefunction studies with synthetic derivatives of 5-OP-RU and other riboflavin-based antigens have provided valuable insights on which chemical group modifications affect ligand antigenicity within the MAIT-MR1 axis (64,65). While improving ligand stability, modification of the pyrimidine backbone of 5-OP-RU (C-5 and C-6) (Figure 2A), had only small effects on altering antigen potency (64), perhaps unsurprising as these atoms make no direct contact with the MAIT TCR (26,28,64,65). In contrast, modifications to the ribityl moiety of 5-OP-RU and ribityllumazine ligands had profoundly negative effects on MAIT TCR recognition and antigen potency (65,71). Particularly, removal of the ribityl 2 ′ -and 3 ′ -OH groups or the entire ribityl tail largely abolished MAIT TCR recognition of relevant 5-OP-RU analogs (65,71). In contrast, shortening of the ribityl tail, while preserving the 2 ′ -and 3 ′ -OH groups (ethyl-or propyl-5-OP-RU analogs) did not appreciably affect MAIT TCR recognition (65,71). Notably however, 2 ′ -OH was not required for MAIT TCR recognition of a modified analog of the ribityllumazine RL-6-Me-7-OH (72). Structural modeling suggested the modified ribityl tail of the 2 ′ -deoxy-RL-6-Me-7-OH is flexible enough to be oriented for recognition by the MAIT TCR Tyr95α via a hydrogen bond with the 5 ′ -OH in place of the 2 ′ -OH group (Figure 2A) (72). Although MAIT cells could recognize MR1 presenting the 2-deoxyribityllumazine, this interaction was not sufficient to lead to MAIT cell activation (72), stressing that distinct requirements exist for MAIT cell activation that are not satisfied by TCR engagement alone. Thus, the series of altered metabolite ligands helped to define the boundaries of riboflavin-based antigen diversity tolerated by the MAIT TCR and emphasized the importance of the ribityl tail and Tyr95α in MAIT TCR recognition. In light of this, new microbial ligands have been identified that share the ribityl moiety and some structural features with the riboflavin-based antigens (73). The lumazines 6-(2-carboxyethyl)-7-hydroxy-8-ribityllumazine (photolumazine I; PLI) and 6-(1H-indol-3-yl)-7-hydroxy-8ribityllumazine (photolumazine III; PLIII) and the riboflavin analog 7,8-didemethyl-8-hydroxy-5-deazariboflavin (FO) were isolated from soluble recombinant MR1 expressed in the context of live bacterial infection ( Figure 2B) (73). Although all three ligands had ribityl moieties, only PLI and PLIII, possible adducts of 5-A-RU, were antigenic to MAIT cells, while FO was antagonistic (73). Similarly, three pyrimidines, mercaptopurine, floxuridine, and doxofylline ( Figure 2C) were identified as putative MR1 ligands in an in silico screen of a range of small organic molecules, including drugs, drug metabolites, and drug-like molecules (29). In cellular assays, floxuridine and mercaptopurine were weakly agonistic but did not upregulate MR1 to detectable levels, while doxofylline weakly upregulated MR1 but was not agonistic (29). Further in silico screening of commercial compound libraries recently identified seven weakly agonistic MR1 ligands ( Figure 2D) (30). None of the novel agonist ligands were as potent as 5-OP-RU in cellular assays or were able to upregulate MR1 surface expression, likely due to the inability to form a Schiff base with MR1 Lys43 (30). In contrast, in the same study, two pyrimidines were identified; 3-[(2,6dioxo-1,2,3,6-tetrahydropyrimidin-4-yl)formamido] propanoic acid (DB28) and its derivative methyl 3-[(2,6-dioxo-1,2,3,6tetrahydropyrimidin-4-yl)formamido] propanoate (NV18.1) that significantly modulated MR1 surface expression (30). These ligands bound to MR1 non-covalently (Figure 1), causing MR1 to be retained in the ER and prevented surface expression (30). Importantly, unlike the other ligands, DB28 and NV18.1 were non-stimulatory to PBMC-derived TRAV1-2 + MAIT cells (30).

CHARACTERISTICS OF THE MAIT TCR REPERTOIRE
The classical MAIT TCRα rearrangements (TRAV1-2-TRAJ33/20/12) are highly conserved between genetically unrelated individuals, also referred to as public TCRα chain usage (3,5). Collectively, these TCRα rearrangements account for the majority (∼95%) of MAIT TCR clonotypes in human blood (7-9, 11, 62, 74). Nevertheless, variations of the classical MAIT TCR CDR3α region exist, typically at one or two non-germlineencoded residue positions (CAXXDSSYKLIF, CAVXXXDYKLSF, and CAXXDSNYQLIW for TRAJ12, TRAJ20, and TRAJ33 MAIT CDR3α rearrangements, respectively) (5,11,74). However, MR1reactive clones responding to different microbes (bacteria and yeast) are characterized by more extensive non-germlineencoded differences within the TCRα chain, including what appeared to be microbe-specific MR1-reactive clonotypes in some instances (8). Similarly, systemic infection of healthy volunteers with S. enterica serovar Paratyphi A transiently altered the MAIT TCR repertoire within individuals at the clonotypic level (75). In this case, circulating, overrepresented MAIT clonotypes were reduced during infection but restored to homeostatic proportions at the resolution of infection (75). TCRs from MAIT clonotypes that expanded during infection showed greater stimulatory potential than TCRs from contracted MAIT clonotypes in response to synthetic riboflavin-based antigens or bacterial supernatant (S. enterica Paratyphi A or E. coli) (75), suggesting TCR-dependent but not microbe-specific differences in MAIT clonotypic responses. In mice, stimulation of circulating MAIT cells with the different riboflavin-based antigens induced activation of separate cell clusters based on their activated surface phenotype (66), perhaps involving preferential stimulation of clonal MAIT TCRs. Similarly, in mice, preferential accumulation of MAIT cells with divergent functional phenotypes occurs in response to acute infection by different bacterial species (S. enterica serovar Typhimurium or Legionella longbeachae) (34,46,76), however, divergence in the functional response of MAIT cells has not been correlated with TCR usage in mice. Interestingly, the TCRα chain TRAV1-2-TRAJ33 rearrangement is also found in some CD161 − TRAV1-2 + T cells, and nearly one third of those TRAV1-2-TRAJ33 rearranged TCRs express Tyr95α (74). Whether these cells are also MR1-reactive is not known, although it should be noted that other MHC-reactive T cells and a subset of CD1b-restricted T cells also utilize the TRAV1-2 gene segment (77)(78)(79). A minority of MAIT TCRα TRAV1-2 rearrangements occur with non-classical junctional regions (non-TRAJ12/20/33), some of which do not contain the Tyr95α residue yet still confer MR1-5-OP-RU reactivity (11). These observations indicate that other factors must account for recognition of 5-OP-RU among "MAIT TCR outliers, " as in one case determined by crystallographic analysis (80).
Among MAIT cells, TCRβ chain composition provides the greatest source of variability within the TCR repertoire (7,8,10,11,28,74). Despite this, TRBV6 is a dominant TRBV gene segment used amongst MAIT TCRs, although some variations exist and are dependent on the TCRα junctional segment (7,74). For instance, TRAV1-2-TRAJ12 TCRα chains more commonly pair with TRBV6 (>80%), compared to TRAV1-2-TRAJ33 and TRAV1-2-TRAJ20 rearranged TCRα chains which pair roughly evenly with TCRβ chains encoded by TRBV6 and TRBV20 family members (7,74). Up to one quarter of paired MAIT TCRs express TCRβ chains that are not from TRBV6 or TRBV20 V segment family members (7,74). In line with the less stringent nature of TCRβ chain usage, the CDR3β region among classical MAIT TCRs are non-germline-encoded and quite hypervariable, ranging from 9 to 19 amino acids in length and containing no discernible sequence motifs (9,28,74). A comparison of seven canonical MAIT TCRs (TRAV1-2-TRAJ33 encoded TCRα chain paired with a TRBV6-1 encoded TCRβ chain) with similar CDR3α loops and highly variable CDR3β loop length and composition revealed how CDR3β hypervariability impacts on MAIT TCR recognition of MR1 (28). All of the TCRs bound to MR1-5-OP-RU within a similar range of affinities (K deq ∼2 µM), with the exception of one TCR that bound MR1-5-OP-RU weaker (K deq ∼9.1 µM) (28). Structural analysis of four of the TCRs bound to MR1-5-OP-RU, including the weaker affinity TCR, demonstrated that the overall TCR footprint onto MR1 was conserved (28). Not surprisingly, the greatest variability in MR1 engagement was from contacts by the TCRβ CDR loops and framework regions that differed between TCRs; yet an overall similar buried surface area (BSA) at the interface was achieved (28). CDR3β loop sequence impacted on conformational flexibility of this loop as well as MR1 residues contacted. Importantly, in some structures the CDR3β loop made direct contact with 5-OP-RU, thus fine-tuning MR1 recognition in an antigen-dependent manner (28).
The MAIT TCR repertoire varies throughout the course of life (fetal, neonate, young, and old) and has now been examined in some detail, revealing consistent changes in MAIT cell frequency as we age (7,9,74,80,81). The overall frequency of MAIT cells in blood increases rapidly from low numbers early in life (11,82), typically over the first two-to-three decades, declining slowly thereafter (83,84). In addition, the diversity in clonotypes decreases as certain MAIT cell clones expand over time (84). This is most pronounced in older adults (>65 years), where a small number of private (non-shared) MAIT cell clonotypes dominate the MAIT cell pool (84). It is likely that expansion of naïve MAIT cells in young individuals is a consequence of microbial exposure following thymic egress (36,45,82,85,86). Interestingly, a large proportion of CD161 + T cells found in cord blood express the canonical MAIT TCRα chain (TRAV1-2-TRAJ33) but do not recognize MR1-5-OP-RU tetramers (81). These non-reactive cells are not readily detectable in adults (11,84), suggesting they do not expand akin to MR1-5-OP-RU-reactive MAIT cells. In this regard, some of these cells might feature a TCRβ chain which prohibits reactivity to MR1 presenting riboflavin-based antigens in favor of other antigens furnished by MR1, some of which may be less abundant or less stimulatory than the riboflavin-based antigens for MAIT cells expressing classical MAIT TCRs (28,35,80) or may trigger peripheral tolerance to avoid autoreactive or allergic responses (87). Accordingly, a pairwise comparison of MAIT TCR sequences in cord blood revealed much broader TCRβ diversity compared to those from MAIT cells from adult blood, further highlighting the effect of peripheral exposure on the narrowing of the MAIT TCR repertoire (84).
The same small molecule drug screen that identified three pyrimidines as potential MR1 ligands (described above) also identified 19 non-pyrimidine MR1 ligands, which belonged to one or more of the following diverse classes of molecules: aromatic aldehyde, aromatic carboxylate, phenol alinine, flavone, isoflavone, enone, quinone (29). Most of these non-pyrimidine ligands could not stimulate a MAIT TCR reporter cell line, and in crystal structures analyzed, no TCR contacts with the MR1 ligands were observed. All of the non-stimulatory ligands tested, including the salicylic acid analog 3-formylsalicylic acid (3-F-SA) (Figure 1), competitively inhibited MAIT cell activation to a similar or lesser extent as 6-FP. Some non-pyrimidine ligands were able to stimulate a MAIT TCR reporter cell line to varying degrees, of which diclofenac was most potent and whose activity was assigned to one of its metabolites, 5-OH-diclofenac (29) (Figure 1). Unlike the pyrimidine and ribityllumazine antigens which are contacted in a hydrogen bond by a conserved residue in the MAIT TCRα chain (Tyr95α), the 5-OH group of 5-OHdiclofenac was contacted in a hydrogen bond by Glu99 from the CDR3β loop. This matched the observation that 5-OH-diclofenac only activated one of a panel of MAIT TCR reporter cell lines, indicating that the MAIT TCRβ chain was able to "fine-tune" responsiveness to certain ligands (29), in line with an emerging concept that some MAIT cells are capable of discriminating between different classes of antigens based on differences in the MAIT TCRβ chain (28,73,80,88,89). A weak stimulator, 5formyl-salicylic acid, also only activated one of a panel of MAIT TCR reporter cell lines, suggestive of a similar mechanism (29). Except for diclofenac/5-OH-DCF, all MR1 ligands, of which crystal structure bound to MR1 and complexed with a MAIT TCR were determined as part of this study, featured a Schiff base with Lys43 of MR1 (26,27,29) (Figure 1).
In summary, accumulated data thus far (26)(27)(28)(29)35) indicate that MR1 has the capacity to bind structurally diverse small molecules (150-400 Da). Most of the MR1 ligands (except for RL-6-Me-7-OH and Diclofenac/5-OH-diclofenac) form a Schiff base with Lys43 of MR1. However, regardless of the Schiff base, all ligands are located broadly within a similar location in the A'-pocket of MR1, although some are displaced; or oriented differently, involving mostly rotations with broadly similar depositions of the planes of the aromatic rings (26,27,29,35). One notable exception is diclofenac/5-OH diclofenac, whose ring was not deposed in the same plane, rather, the central ring was essentially perpendicular to that of any of the other MR1 ligands (Figure 1) (29). Both in the context of bacterial infections (73,88), cancer and steady-state (44,90) there is evidence for the existence of additional MR1 ligands, including some that do not appear to be pyrimidines or ribityllumazines (73,88), although the chemical identities of these remain to be defined. Some of these ligands appear to be recognized by T cells that do not necessarily share the phenotype of MR1-5-OP-RU specific MAIT cells, described in detail in the next section.

DIVERSE MR1-REACTIVE T CELLS
A range of MR1-antigen reactivity patterns have emerged for individual T cell clones and the TCRs they express. These reactivity patterns were identified based on (i) binding of T cell clones or TCR reporter cell lines to MR1-antigen tetramer (involving riboflavin-based antigens or other antigens) or mutated MR1 Lys43 to Ala (MR1-K43A) tetramers; and (ii) activation of T cell clones or TCR reporter cell lines in cellular assays with antigen presenting cells expressing physiological levels of MR1 or overexpressing MR1 or MR1-K43A. Notably, this is an active area of research and it is likely that additional reactivity patterns will emerge. Whilst the patterns observed so far are not always absolute, in that one reactivity might be less dominant, for simplicity they can be grouped as outlined below and illustrated in Figure 3 and Table 1. Similarly, a particular T cell phenotype (TCR usage, surface markers, transcription factors) can be exclusively or more frequently associated with one particular antigen reactivity pattern, although in some cases significant overlaps in TCR usage (and other phenotypic characteristics) have been observed for TCRs with different MR1antigen reactivity patterns.
(1) Reactivity to MR1-5-OP-RU/5-OE-RU (and less potent ribityllumazines) but not to other MR1-ligands. This reactivity is mediated by the population of cells referred to as MAIT cells, expressing the canonical MAIT TCR and phenotypic markers associated with the MAIT cell lineage in humans and mice (7,24,26,28,80). The same reactivity can also be mediated by a small fraction of human, TRAV1-2 − T cells with diverse TCR usage (e.g., clone MAV36 that expresses a TRAV36 + TCR), including some cells that feature a MAIT-like phenotype (CD161 + , IL-18Rα + , CD218a + , CD26 + ) (80,89). Notably, a follow-up study revealed that TRAV36 + MR1 reactive T cells are possibly a second public TCR family, alongside MAIT cells, capable of specifically recognizing riboflavin-based antigens in the context of MR1 (80,89). The TRAV36 gene segment of these TCRs was not rearranged with the classical MAIT TRAJ gene segments (TRAJ33, TRAJ12, TRAJ20), and hence lacked a Tyr95α (80,89). Reminiscent of the classical MAIT TCRα chain, they displayed a largely germline-encoded CDR3α loop. In stark contrast to canonical MAIT TCRs, these TCRs featured nearly invariant CDR3β loops of constrained length (14 amino acids) (89). Structural studies on the MAV36 TCR revealed the CDR1a loop was predominant  Table 1 outlines the frequency, TCR usage (where known) and surface characteristics of each subset using this classification. This schematic was inspired by Figure 1 from a recent commentary piece (91).
(4) Reactivity to other MR1-ligands but not MR1-5-OP-RU/5-OE-RU (and less potent ribityllumazines). Such reactivity pattern was first observed for the TRAV1-2 − MAV21 TCR (TRAV21) which specifically recognized pterins but not 5-OP-RU (80). Recently, Crowther et al. identified another TRAV1-2 − T cell clone in blood of a healthy individual (MC.7.G5, TRAV38.2/DV8-TRAJ31, TRBV25.1 TRBJ2.3) which specifically recognized an antigen/s that was expressed or upregulated in cancerous cells but not noncancerous cells (resting, activated, stressed or infected), in an MR1-dependent manner (90). Interestingly, the observed T cell reactivity was donor-unrestricted, with multiple HLAmismatched tumor cells recognized in an MR1-dependent manner, whilst not exerting allo-reactivity, making this T cell clone particularly interesting for immunotherapies (90). Potent activation, including cytotoxicity, was achieved with physiological levels of MR1 and high effector (T cell) to target (APC) ratios. Both Ac-6-FP and bacterial riboflavinbased antigens were not recognized and acted as competitive inhibitors of activation. Whilst a clone with a similar reactivity pattern was also isolated from a second donor, the frequency of similar clones within an individual and their prevalence in the general population and in the context of cancers are unknown. Also, the phenotype of the clones (other than the TCR of one of them) was not described which may provide insights into its developmental origin. Previously, Lepore et al. (44) had identified a population of T cells in healthy individuals that comprised 1 in 2,500-5,000 of circulating T cells. Designated "MR1T" cells, clones of these cells recognized, in the context of physiological levels of MR1 and dependent on MR1, endogenous cellular antigens from various tissues, including cancer tissues, and thus were not cancer specific. One MR1T cell clone was TRAV1-2 + (clone DGA4), but most were TRAV1-2 − , expressing diverse TCRα and β chains (e.g., clone DGB129; TRAV29/DV5-TRAJ23; and TRBV12-4-TRBJ1-1) (44). Like the MC.7.G5 clone, MR1T cell clones (e.g., DGB129) did not recognize 6-FP and their activation was inhibited by 6-FP. However, unlike the MC.7.G5 clone, which did not bind MR1-K43A tetramer or respond to MR1-K43A overexpressing cells (90), MR1T cell clones equally recognized wild type and MR1-K43A overexpressing cells. This suggested to the authors that relevant antigens that were recognized did not depend on Schiff base formation with MR1 for TCR recognition but may also suggest an MR1-centric recognition or MR1 autoreactivity by some MR1T cells for which 5-OP-RU/ribityllumazine are permissive antigens but pterins are not. At least two clones (DGB129 and DGB70) did however, differentially recognize fractions of cell/tumor lysates, demonstrating differential antigen specificity for at least some clones. With the exception for one MR1T clone, all TRAV1-2 − clones were CD161 − and all, except for one, were CD8α + (DGB129, TRAV1-2 − , CD161 − , CD8α + ; DGB70, TRAV1-2 − [TRAV5], CD161 − , CD8 − ). DGA4, the TRAV1-2 + clone, was CD161 + , CD8 − . Phenotypic and functional characterization of MR1T cell clones showed multiple chemokine receptor expression profiles and secretion of diverse effector molecules, suggesting functional heterogeneity within this population that was also distinct from that of MAIT cells. In a T-helper like function, clone DGB129 induced maturation of monocyte derived dendritic cells and another clone suggested the possible contribution of MR1T cells to intestinal epithelial barrier homeostasis.

DISCUSSION
We are beginning to appreciate the diversity of microbial and endogenous (including cancer) antigens presented by MR1 and the heterogenous populations of T cells which recognize them. Independent and repeated identification of these cells has led to the classification "MR1-reactive T cells, " illustrated in Figure 3 and Table 1, inclusive of MR1-restricted MAIT cells, and more phenotypically and functionally heterogenous T cells, some of which are MAIT-like in their phenotype, grouped together as "other MR1-reactive T cells" (44,80,89,91). Whilst TRAV1-2 + T cells are dominant amongst MAIT cells, other MR1-reactive T cells mostly include TRAV1-2 − αβ TCR + T cells (44,73,80,89,90) as well as γδ TCR + T cells (96). Thus, the current stage of research suggests that most MR1-reactive T cells, namely MAIT cells, feature a biased TCR α-chain, and react with a limited repertoire of pyrimidine and ribityllumazine antigens derived from riboflavin biosynthesis. There are also several examples of TCR β-chain usage influencing antigen preference and allowing for antigen crossreactivity by classical MAIT TCRs, and future work might identify additional examples of physiological relevance. The emerging descriptions of other MR1-reactive T cells highlight that the classical MAIT TCR α-chain is not the only solution that warrants MR1-reactivity, however, more work is needed to validate and corroborate the TCR usage and antigen specificities of other, more diverse T cells. In particular, determining the chemical identities and biosynthetic origins of the tumor-associated and endogenous antigens recognized by some of the other MR1-reactive T cells will allow the generation of tetramers which will greatly assist in determining the prevalence of these cells as well as their potentially important role in anti-tumor immunity and other immune-functions. It will also help to better characterize the phenotype (including the TCR usage) and of this diverse set of T cells including public as compared to private existence of individual clones, recently described for TRAV36 + MR1-reactive T cells (89). Another interesting question concerns the lineage development of other MR1-reactive T cells that feature a non-MAIT phenotype. Is it possible that these cells might have originated from another T cell lineage and cross-react across MHC families, including classical MHC molecules (MHC-I and MHC-II) and MHC-Ilike molecules? In TCR α-chain transgenic mice that lack MR1 (Vα19iCα −/− MR1 −/− ) MR1-5-OP-RU tetramer + T cells make up ∼32% of T cells in the spleen as compared to ∼50% in mice proficient in MR1 (Vα19iCα −/− ) (97). The population of these cells mediated both MR1-dependent and MHC-I (H2-K b /H2-D b ) dependent reactivity in vitro (97), although it was not investigated whether these specificities were mediated by single T cells and single TCRs. More recently, MR1-5-OP-RU tetramer + T cells featuring the canonical MAIT TCRα chain (TRAV1-TRAJ33) but not the typical TRBV13 bias were also found in very low percentages in the thymus of non-transgenic mice, Mus musculus castaneus (CAST mice), modified to lack MR1 (86). Thus, MR1-5-OP-RU tetramer + T cells can develop in the absence of MR1, inferring selection may occur on other MHC molecules. Indeed, MR1 is highly conserved with HLA-A2 (39% sequence conservation), suggesting mimicry as a potential mode of crossreactivity by these cells. So far evidence of peripheral T cells that crossreact between classical MHC molecules and MHC-I-like molecules remains elusive.

AUTHOR CONTRIBUTIONS
MS and SE wrote and revised the manuscript, read, and approved the submitted version. All authors contributed to the article and approved the submitted version.