MAIT Recognition of a Stimulatory Bacterial Antigen Bound to MR1 Free

Mucosal-associated invariant T (MAIT) cells are an evolutionarily conserved subpopulation of αβ T cells in mammals that are characterized by the expression of a semi-invariant αβ TCR and restriction to the MHC-like protein MR1 (1–3). MAIT cells exhibit an innate-like phenotype and are found predominantly in mucosal tissues, such the gut lamina propria and the lung. MAIT cells are also found at high frequency in the liver, and in blood they can represent up to 10% of the CD4⁻ αβ T cell population (4, 5). MAIT cell development is dependent on MR1 expression, B cells, and an established host commensal flora (2, 6).

Highly conserved across mammalian evolution, MR1 transcripts and protein are detected in most tissues, yet detection on the cell surface is low or undetectable under basal conditions (7, 8). The recent structural elucidation of MR1 (9, 10) revealed an overall backbone structure similar to peptide-presenting, classical MHC molecules but with a putative Ag-binding cavity of smaller dimensions due to the presence of large aromatic and basic residues lining the cavity. The aromatic architecture of this cavity is complementary to the small, ring structure containing ligands identified recently (10) that are derivatives of two B vitamin metabolic pathways: folic acid (vitamin B9) and riboflavin (vitamin B2). 6-Formylpterin (6-FP) was identified as an MR1-bound ligand, potentially covalently attached via Schiff base to lysine 43 found in the ligand-binding cavity of MR1. 6-FP was nonstimulatory to MAIT cells. In contrast, 6,7-dimethyl-8-ribityllumazine (DMRL), the direct precursor to riboflavin, and two DMRL variants were identified as MR1 ligands sufficient to stimulate MAIT cells.

To engage an MR1–Ag complex, human MAIT cells use an αβ TCR composed of a mostly invariant α-chain (Vα7.2/Jα33) that associates with a limited array of β-chains: Vβ2, Vβ13, and Vβ22, among others (11). Diversity in the variable domain of the β-chain arises from Vβ-Dβ-Jβ gene segment rearrangement that generates highly variable CDR3β loops, conferring MAIT TCRs a certain level of diversity despite the Vβ restriction. MAIT cells can be activated upon infection with diverse bacterial or yeast species, including Escherichia coli, Staphylococcus aureus, Mycobacterium tuberculosis, Salmonella typhimurium, Candida albicans, and Saccharomyces cerevisiae, but not viruses (5, 12). Human MAIT cell autoreactivity was also described in MR1-transfected cells in a microbial-independent manner (7, 13, 14), suggesting that endogenous ligands can contribute to MAIT cell reactivity.

Taking advantage of a ligand-independent, cross-species reactivity between human MAIT cells and bovine MR1, our group recently reported the first crystal structure of a complex between a MAIT TCR and MR1, providing a molecular model for how MAIT cells engage their MR1 ligand (9). This model was consistent with mutagenesis of the MAIT TCR/MR1 interaction from other groups (14, 15), and superimposes almost identically with a recently reported complex structure of a human MAIT TCR with human MR1 (16), supporting it as a bona fide model for MAIT TCR engagement of MR1. In this complex, the MAIT TCR bound in a diagonal fashion, reminiscent of classical αβ TCR recognition of MHC–peptide complexes (17) and type II NK T cell recognition of CD1d (18, 19). This docking orientation positioned the CDR3 loops of both the α− and β-chains close to the opening of the ligand-binding cavity. Modeling of the stimulatory ligands identified by Kjer-Nielsen et al. (10) into this complex suggested an important role for Tyr⁹⁵ of the CDR3α loop, because this residue was positioned directly over the opening of this cavity, in hydrogen-bonding proximity to the ribityl chain of the stimulatory compounds DMRL and reduced 6-hydroxymethyl-8-(1-d-ribityl)lumazine (rRL-6-CH₂OH). This complex also identified the three residues on MR1 mediating human/cow cross-reactivity.

In this study, we present our results derived from an engineered, humanized version of bovine MR1 (hbMR1) that was expressed recombinantly in insect cells exposed to E. coli culture supernatant as a source of MAIT cell–reactive ligands. We demonstrate an enhancement of MAIT TCR binding to E. coli–loaded hbMR1 over unloaded hbMR1 and show using mass spectrometry (MS) that the bound MR1 ligand is rRL-6-CH₂OH. Furthermore, our complex of the F7 MAIT TCR with this rRL-6-CH₂OH–loaded MR1 provides critical information about how the MAIT TCR engages MR1 ligands, confirming a pivotal role for Tyr⁹⁵ of the CDR3α loop and establishing a role for the CDR3β loop in ligand binding. This complex, like our previously reported xenoreactive complex, is strikingly similar to the recently reported human MAIT/MR1 complex with a highly similar stimulatory compound (16), reinforcing the evolutionary conservation of this interaction. Last, we present structural evidence that variation in Vβ usage and diversity in the sequence of the CDR3β loops does not affect TCR docking, but it can modulate the affinity of binding to MR1, with and without stimulatory ligand, through variable contacts with MR1 and ligand.

hbMR1 was cloned, expressed, and purified as previously described (9). Briefly, the triple mutant (hbMR1) was generated through overlapping PCR and cloned into the pAcGP67A vector (BD Biosciences) for subsequent expression via baculovirus in Hi5 cells. The recombinant protein was purified with Nickel NTA Agarose (QIAGEN), anion exchange, and size-exclusion chromatography. For preparation of E. coli supernatant, BL21 cells were cultured overnight and then centrifuged at 6000 rpm for 30 min. The supernatant was then passed through a 10-kDa cut-off membrane, and the flow-through was collected. Variable amounts (10–100 ml) of this flow-through were added every 24 h to the Hi5s cells postinfection with the recombinant baculovirus for hbMR1. The expression and purification of the MAIT TCR clones F7, G2, and AE6 were carried out in a similar way to previously described procedures for both Hi5 and E. coli–expression systems (9, 20, 21).

For binding analysis between hbMR1 and the different human MAIT TCR clones, hbMR1 expressed in the presence or absence of the E. coli supernatant was captured to 4 ηm units on a Ni-NTA (NTA) Biosensor in a Blitz System Package (Fortebio), as described previously (9). The binding affinities for the F7, G2, and AE6 MAIT TCR clones were each tested by running increasing concentrations of the TCRs over the immobilized hbMR1 protein in 10 mM HEPES (pH 7.4), 150 mM NaCl. IgG F_C was captured to a similar level and used to subtract nonspecific binding signals. Subtracted responses were then used for calculating K_D with GraphPad Prism by plotting the binding values at equilibrium against the TCR concentrations.

For hbMR1, comparison of hbMR1 protein that was exposed to E. coli supernatant and an untreated control hbMR1 sample were analyzed on a Bruker maXis impact QTOF LC/MS operating in the negative ion mode. Five microliters each of 25 μM hbMR1 was injected onto a Luna NH2 4.6 × 50-mm, 5-μm column in 20 mM ammonium acetate (pH 9) buffer and eluted with an aqueous gradient, as described above. The retention time of rRL-6-CH₂OH was determined by extracted ion chromatograms using the m/z values. Product ions were obtained from target fragmentation at collision-induced voltages of 20eV.

For crystallization purposes, F7, G2, and AE6 recombinant E. coli–derived heterodimeric MAIT TCRs were refolded and purified (9), stoichiometrically mixed with hbMR1 or bovine MR1, and concentrated to 5–10 mg/ml. Crystals of bovine MR1/F7 MAIT TCR complex were used for microseeding fresh sitting drops containing hbMR1/F7 MAIT TCR, bovine MR1/G2 MAIT TCR, and bovine MR1/AE6 MAIT TCR and mother liquor consisting of 100 mM HEPES and various concentrations of ammonium sulfate (0.9–2.0 M). Crystals for all of the complexes appeared at times ranging from 2 to 12 wk.

Prior to data collection, crystals were soaked in mother liquor containing 20% glycerol and then cryo-cooled. All data sets were collected on a MAR300 CCD at beamline 23 ID-D and 23 ID-B at the Advanced Photon Source at Argonne National Laboratory (Lemont, IL) and processed with HKL2000 (22). All datasets were corrected for anisotropic diffraction with the Phenix software suite (23). The structures of the ternary complexes were solved using the published complex structure of bovine MR1/F7 MAIT TCR (PDB 4IIQ) as a search model and removing the TCR CDR loops and all nonprotein coordinates prior to molecular replacement with Phaser (24). The derived solution was refined with the Phenix.refine package (23), including Translation/Libration/Screw vibrational motions (25) at the latest stages of refinements, and combined with manual building in Coot (26) in between the refinement steps. When necessary, PRODRG (27) was used for the generation of ligand libraries and coordinates that were included in the refinement and building process. A random 5% of reflections were taken out of each data set for statistical validation purposes (Rfree) throughout the entire refinement process.

Intermolecular contacts and distances were calculated using the program Contacts from the CCP4 software package (28), interface surface areas were calculated using the PISA server (http://www.ebi.ac.uk/msd-srv/prot_int/pistart.html), and all structural figures were generated using the program Pymol (Schrödinger). Coordinates and structure factors have been submitted to the Protein Data Bank (http://www.ncbi.nlm.nih.gov/pubmed/10592235) under accession codes 4LCC (F7 MAIT TCR/hbMR1), 4L8S (G2 MAIT TCR/bMR1), and 4L9L (AE6 MAIT TCR/bMR1).

Our initial attempts to recombinantly express human versions of MR1 in insect cells failed despite extensive attempts at engineering different constructs. Alternatively, we generated a humanized version of bovine MR1 (9), hbMR1, by mutating the three TCR contact positions that differ between human and bovine MR1 (Ala72Met in MR1 α1 helix and Arg147Gln and Gln151Leu in α2 helix), which we showed modulated MAIT cell xenoreactivity (9). This construct was expressed well and eluted as a monodispersed peak under size-exclusion chromatography. We used the construct in this study to characterize, structurally and biophysically, the recognition of a MAIT cell–reactive ligand bound to hbMR1 by human MAIT TCRs.

Several studies demonstrated MAIT cell reactivity against MR1-transfected APCs infected with different strains of bacteria and/or yeast (5, 10, 12, 14). One of the strains able to trigger MAIT cell response and activation is E. coli. We sought to determine whether a stimulatory ligand could be loaded by expressing, in our insect-expression system, our recombinant hbMR1 in the presence of E. coli supernatant. Protein derived from this method had a noticeable yellow hue, consistent with the presence of a flavonoid substance. Using bio-layer interferometry, we compared the binding kinetics of three MAIT TCRs to the E. coli–loaded hbMR1 compared with that of hbMR1 expressed in the absence of E. coli supernatant (Fig. 1). The binding studies revealed a 20-fold increase in the binding affinity by the F7 MAIT TCR, a >100-fold increase by the G2 MAIT TCR, and a low micromolar affinity interaction for the AE6 MAIT TCR clone, whose interaction with the untreated hbMR1 was not measureable (Fig. 1). These affinities (∼4.5 and 4.9 μM) measured for the MAIT TCR interactions with hbMR1 loaded with a ligand derived from E. coli supernatant are similar to those measured for MAIT TCRs with human MR1 loaded with synthetic rRL-6-CH₂OH (1.65 μM) (16). Differences in MR1-production strategies and ligand loading between these two studies might have resulted in slight differences in affinity. These results suggest that hbMR1 is loaded with a ligand or ligands of bacterial origin that play a direct role in the enhancement of recognition by the MAIT TCR.

These results prompted us to use MS to analyze the content of hbMR1 protein in samples that were exposed or not to E. coli. Normal-phase quadrupole mass analyzer time-of-flight mass analyzer analysis clearly identified a compound with a retention time ∼ 7 min present only in the E. coli–treated sample (Fig. 2A). MS analysis of this compound revealed it to have an m/z ratio of 329.1095, and its fragmentation yielded a pattern nearly identical to that found for rRL-6-CH₂OH (Fig. 2B), a MAIT cell stimulatory ligand characterized from S. typhimurium supernatant (10). This compound is generated as a by-product of the riboflavin synthesis pathway and its structure consists of a lumazine core and a ribityl chain (Fig. 2A, inset). rRL-6-CH₂OH was shown to trigger MAIT cell activation in an MR1-dependent manner, and it displayed the strongest potency in activating MAIT cells among three structurally related compounds (10). Thus, the increased MAIT TCR affinity for the E. coli–treated hbMR1 sample correlates with the presence of this ribityl moiety, which our model suggested (9) participates directly in MAIT TCR recognition of MR1-presented Ag.

To determine the structural basis for this enhanced recognition of rRL-6-CH₂OH–loaded hbMR1 by the MAIT TCR, we used our recombinantly expressed, rRL-6-CH₂OH–loaded MR1 in crystallization screens with the F7 MAIT TCR. Single crystals grown with these components diffracted well and were used to collect a full data set that refined to 3.3 Å (Table I). Overall, the docking mode of the MAIT TCR is highly similar to that observed with bovine MR1 (9) and human MR1 (16), with the MAIT TCR docking in a diagonal orientation with respect to the α1 and α2 helixes of hbMR1 (Fig. 3A). The α-chain CDR loops of the MAIT TCR mostly contact residues of the α2 helix of hbMR1, whereas the β-chain is biased toward the α1 helix of hbMR1 (Fig. 3B, 3C, Table II). These contacts, with the exception of those made with the three bovine-specific amino acids (A72, R147, and Q151), are remarkably similar between this complex and our previously reported MAIT/bMR1 complex (Supplemental Fig. 1). Moreover, the TCR docking of both of these complexes superimpose almost identically on the recently published human MAIT/MR1/RL-6-Me-7-OH complex (Fig. 3B), further validating the evolutionary conservation of MAIT TCR recognition across species. The buried surface area of the MAIT TCR–hbMR1 interface is ?1070 Ų, with 52 and 48% contributed by the α- and β-chains of the MAIT TCR, respectively (Fig. 3C).

Previously, we reported a critical role for the CDR3α loop in the interaction with bMR1; this is confirmed in this complex, with Tyr⁹⁵ of the CDR3α loop positioned directly over the ligand-binding cavity opening. The conformation of the CDR3α loop in this complex is essentially identical to the CDR3α loop conformations noted in the xenoreactive complex (9) and the unliganded human MAIT TCR (15), as well as the human complex (16) (Supplemental Fig. 2), confirming a critical role for Tyr⁹⁵ in MAIT cell recognition of MR1/ligand and the rigidity of the CDR3α loop upon docking. In addition, we see a role for the CDR3β loop in MR1 recognition due to a subtle shift in conformation, resulting in a new contact established with MR1 (G98β with Trp⁶⁹ in the hbMR1 α1 helix). As discussed in more detail below, the CDR3β loop also makes direct contact with the rRL-6-CH₂OH ligand, demonstrating a clear role for the diverse CDR3β loop in ligand discrimination. Despite the flexibility observed in the α2 helix in human MR1 (10), the structure with hbMR1 does not display evident differences in the backbone positioning, suggesting that hbMR1 does not require special conformational rearrangements to be engaged by a MAIT TCR (Supplemental Fig. 3). Slight flexibility was noted in the α2 helix between the human MR1 unliganded (10) and liganded (16) structures; however, this may be due, in part, to the different resolution of the datasets (3.2 Å versus 2.0 and 1.9 Å).

Electron density for the rRL-6-CH₂OH ligand is unambiguous, placing the ligand in a position very similar to that of our model (9) (Fig. 4A) and that observed for 6-FP (9, 10), where the aromatic residues lining the cavity interact with the lumazine moiety, primarily through van der Waals (VDW) and π-stacking interactions (Fig. 4B). Hydrogen bonds are noted between rRL-6-CH₂OH and several MR1 side chains: Ser²⁴, Lys⁴³, Arg⁹⁴, and Tyr¹⁵² (Fig. 4A, Table II); they anchor the ligand stably in the ligand-binding cavity. The ribityl group extends upward toward the opening of the ligand-binding cavity, engaging the MAIT TCR through both the CDR3α and CDR3β loops. Tyr⁹⁵ of the CDR3α loop establishes at least one hydrogen bond with the ribityl group, as we predicted previously (9). However, in this complex, we see a new role for the CDR3β loop in ligand discrimination. A hydrogen bond between the main-chain nitrogen of the CDR3β Glu⁹⁹ is established with the terminal hydroxyl of the ribityl chain of rRL-6-CH₂OH. These contacts provide a clear rationale for the enhancement of MAIT cell binding upon recognition of MR1 loaded with stimulatory ligands, such as rRL-6-CH₂OH. The conformation of the RL-6-Me-7-OH ligand (16) reported recently in human MR1 is flipped in relation to the conformation of rRL-6-CH₂OH reported in this study, resulting in the ribityl chain of RL-6-Me-7-OH being more sequestered in the MR1-binding cavity (Fig. 4C). This is probably not due to differences in the MR1-binding pockets, because the positions of the residues in the two structures superimpose almost perfectly (Fig. 4C). The positioning of the RL-6-Me-7-OH ligand in human MR1 results in only one TCR contact: Tyr⁹⁵ of the CDR3α loop. This likely is the reason for the reduced functional potency of this ligand in relation to rRL-6-CH₂OH (10) and strongly suggests that ligands presented by MR1 adopt unique conformations in the ligand-binding cavity that are dependent on their chemical structure; these conformations can have direct effects on the functional outcome.

To determine whether conformational adjustments play a role in MAIT TCR surveillance of MR1-presented Ags, we compared our crystal structure of the F7 MAIT TCR in complex with hbMR1/rRL-6-CH₂OH with that of F7 in complex with bovine MR1. A small conformational shift of the CDR3β loop was observed, resulting in a hydrogen bond between Gly^98β and the Trp⁶⁹ side chain of hbMR1 (Figs. 3C, 5). These contacts are remote from the three species-specific differences noted in bovine MR1; therefore, it is unlikely that the sequence differences are the cause of this conformational change. Instead, it is more likely that conformational flexibility of the CDR3β loop plays a role in engagement of certain MR1-presented stimulatory ligands. As noted previously, the CDR3β loop does not contact the RL-6-Me-7-OH ligand (16), despite having the same amino acid sequence. The overall Cα backbone structure of hbMR1 compared with that of bovine MR1 is 1.0 root-mean-square deviation, suggesting that hbMR1 does not change substantially upon presentation of a stimulatory ligand or upon engagement of a MAIT TCR (Supplemental Fig. 3).

We demonstrated previously that variation in the β-chain, either through Vβ domain usage or diversity at the CDR3β loop, can modulate the affinity of the MAIT TCR for MR1 (9). The three MAIT TCRs that we examined previously—F7, G2, and AE6 clones—all differ in their CDR3β loop sequences; however, F7 and G2 both use Vβ13.3 (TRBV6-1) and, therefore, share the same CDR1β and CDR2β sequences, whereas AE6 uses Vβ13.2 (TRBV6-2), which differs in both the CDR1β and CDR2β sequences. F7 and G2 have similar binding affinities to bovine MR1 (30–40 μM), but AE6 binding is ∼2-fold weaker (∼70 μM) (9). This also is reflected in the measured affinities to hbMR1/rRL-6-CH₂OH, where F7 and G2 bind with ∼5 μM affinity, and AE6 is ∼2-fold weaker (8.2 μM) (Fig. 1). To understand the molecular basis for our measured affinity difference and the role of Vβ and CDR3β diversity in MR1 binding, we determined the crystal structures of the G2 and AE6 MAIT TCRs in complex with bovine MR1 to 2.9 and 3.4 Å, respectively (Fig. 6, Table I, Supplemental Fig. 4).

Comparison of the three xenoreactive structures shows a conserved docking mode, despite variation in the Vβ domain usage and CDR3β loop sequence (Fig. 6). The contacts of the CDR loops from the α-chains are essentially identical, with minor contact differences attributable to the range of resolution of these complexes. However, variation in contacts of the β-chain were much more extensive. In the F7 complex with bovine MR1, no contact with the CDR1β loop was observed (9), whereas in both G2 and AE6 the CDR1β loop is involved in MR1 recognition (Fig. 6). In the G2 complex, there is only one VDW contact: Asn³⁰ of CDR1β contacts Gln⁷¹ of the MR1 α1 helix. However, the AE6 TCR makes more extensive contacts through the CDR1β loop, with three VDW contacts through Glu³⁰ (with Gly⁶⁸, Gln⁷¹, and Ala⁷² of the MR1 α1 helix) and one through Tyr³¹ with Leu⁶⁵ of the MR1 α1 helix. Glu³⁰ and Tyr³¹ are amino acid residues unique to the Vβ13.2 domain, suggesting that Vβ-encoded sequence variation can play a significant role in MR1 recognition.

Variable positions in the CDR2β loop between Vβ13.3 and Vβ13.2 are also involved in MR1 binding (Fig. 6). In the F7 complex structure, the CDR2β loop has extensive contacts with the α1 helix of MR1 (9); a highly similar set of contacts is also seen in the G2 complex, which shares the same Vβ domain (Vβ13.3), resulting in main-chain loop conformations that are superimposable. AE6, in contrast, establishes fewer contacts through the CDR2β loop, and two of the loop-contact residues (Val⁵⁰ and Ala⁵⁶) differ from the sequence of F7 and G2 (Ala⁵⁰ and Asp⁵⁶). Overall, the AE6 TCR appears to distribute its Vβ contacts over all CDR loops, unlike the CDR2β and CDR3β loop bias observed in F7 and G2.

Finally, all three TCRs differ in the amino acid sequences of their CDR3β loops and adopt different constellations of contacts with MR1. All CDR3β loops bridge the ligand-binding cavity, with contacts observed between both the α1 and α2 helices of MR1 (Fig. 6, lower panel). The CDR3β loop of the F7 TCR is biased toward the α2 helix of MR1, with six VDW contacts and two hydrogen bonds. Only four VDW contacts are made with the α1 helix of MR1 (9). In contrast, the G2 TCR CDR3β loop establishes the majority of contacts (five of six) with the α1 helix; four of these are VDW contacts and one, Asp^97ο with Trp^69Nε1, is a hydrogen-bond (Table III). Of note is a salt bridge established between Asp⁹⁷ of CDR3β and Arg⁶¹ of the α1 helix. Asn⁹⁹ is the only CDR3β residue that contacts the α2 helix, forming a hydrogen bond with Glu^149Nε1. In the AE6 MAIT TCR clone, the contacts are distributed between the α1 and α2 helices of MR1 and are composed of both hydrogen bonds and VDW interactions (Table IV). Pro⁹⁷ and Asp⁹⁸ contact the α1 helix residues Leu⁶⁵ and Trp⁶⁹, whereas Gly⁹⁹ and Gly¹⁰⁰ contact His¹⁴⁸, Glu¹⁴⁹, and Tyr¹⁵² of the α2 helix.

The evolutionary conservation of the MAIT lineage and the molecule to which they respond suggest that this surveillance provides an important function in host protection and/or homeostasis. Yet the modulation of MAIT cell reactivity by MR1-presented ligands has remained unclear. The recent identification of small, ring-based molecules as ligands for MR1 and Ags for MAIT cells (10) expanded our understanding of the signals that are used to engage the MAIT population and opens up a new class of potential MR1 ligands for MAIT cell modulation. In this study, we provide structural, biochemical, and biophysical data revealing the molecular basis of MAIT cell recognition of an MR1-presented stimulatory ligand and how diversity in the MAIT cell population via alternative Vβ gene usage and CDR3β loop diversity can further modulate MAIT cell recognition of MR1 ligand. This work provides a foundation upon which to study the presentation of other MR1-presented Ags and determine how variation in the MAIT population translates into Ag recognition and effector function.

Our crystal structure of the F7 MAIT TCR in complex with hbMR1 loaded with E. coli–derived rRL-6-CH₂OH demonstrates a highly conserved docking orientation that is similar to the unconventional noninvariant NK T TCR–CD1d-sulfatide (18, 19) or classical αβ TCR–MHC–peptide complexes (17), both of which use diversity in their CDR3 loops to probe their variable Ags. The docking of the TCR onto MR1 is nearly identical to that observed in our ligand-independent, xenoreactive complex (9), as well as the recently described human MAIT/MR1 complex (16), with a related, but different, stimulatory ligand, reinforcing the evolutionary conservation of this interaction and strongly suggesting that the TCR-docking footprint does not vary with MR1-presented ligands.

Our structural data also reveal the orientation of the stimulatory rRL-6-CH₂OH in the hbMR1-binding cavity. Very similar to our predicted model and to the placement of the folic acid derivative 6-FP (9, 10), rRL-6-CH₂OH is surrounded by a cluster of aromatic MR1 residues and hydrogen bonds, with Ser²⁴, Lys⁴³, and Arg⁹⁴ side chains in the hbMR1 groove. The ribityl chain emerges from the MR1 cavity, establishing hydrogen bonds with both the CDR3α and CDR3β loops of the MAIT TCR. These additional contacts contributed by Ag to the interaction are consistent with the enhanced binding of MAIT TCRs with rRL-6-CH₂OH–loaded MR1. This orientation is different from that of the RL-6-Me-7-OH ligand resolved recently (16) and provides an excellent explanation for why this ligand is more stimulatory than RL-6-Me-7-OH. Additional ligand contacts with the CDR3β loop likely enhance TCR engagement, resulting in the observed enhanced potency. The diversity in ligand conformations also suggests that the MR1-binding cavity can accommodate a range of structures, suggesting that other small molecules may serve as stimulatory ligands presented by MR1 to MAIT cells.

In our hbMR1–rRL-6-CH₂OH/MAIT TCR complex, the MAIT TCR straddles both MR1 α helices, positioning the semi-invariant CDR3α loop and highly diverse CDR3β loop over the opening of the MR1 ligand-binding cavity. Observed contacts between both CDR3 loops and the rRL-6-CH₂OH ligand confirm their importance in Ag recognition and suggest that MAIT TCRs use both conserved amino acid motifs (Tyr⁹⁵ in the CDR3α loop) and diversity (CDR3β loop residues) to probe MR1-presented Ags. The enhancement of MAIT TCR-binding affinity when MR1 presents rRL-6-CH₂OH (20–100-fold enhancement over unloaded MR1) confirms the importance of these ligand contacts in MAIT TCR recognition of MR1. It is unknown whether the differences in the measured affinities between MAIT TCRs will dictate the effector function of the cell or whether TCR diversity directs specificity to particular MR1-presented ligands. However, our demonstration that MAIT TCRs can recognize MR1-presented Ags with different affinities suggests that MAIT TCR diversity, either through the use of alternate Vβ domains or variation within the CDR3β loop, plays an important role in MAIT cell surveillance. Perhaps TCR variability allows MAIT cells to tune their response to different MR1-presented Ags, or it may imply alternate functions for MAIT cells in host defense or homeostasis that require a different affinity threshold for activation.

The complexes between the G2 and AE6 TCRs and bovine MR1 provide additional new insight into the role of β-chain variation in MR1 recognition. Combined with the F7/bMR1 complex, we have three unique MAIT TCR/MR1 structures for comparison. These structures reveal that variation in the Vβ domain can modify contacts mediated through the CDR1β and CDR2β loops, as seen in the AE6 TCR, which uses Vβ13.2 instead of Vβ13.3. AE6 distributes its Vβ contacts over its CDR1 and CDR2 β loops, whereas F7 and G2 contacts are either exclusive to (F7) or heavily biased toward (G2) the CDR2β loop. Loop-swapping experiments showed the importance of the CDR3β loop (15), but our structures reveal how each TCR establishes a novel constellation of contacts with MR1. Our finding that flexibility in the CDR3β loop plays an important role in MR1 ligand engagement contrasts with the essentially rigid docking of the α-chain CDR loops onto the MR1 ligand surface.

Future research that extends from this work includes determining the identity of the MAIT cell–selecting ligand(s) presented by MR1 during MAIT cell development in the thymus. How is ligand involved in this recognition, or does it merely serve to stabilize MR1 expression on the cell surface? Are there endogenous ligands presented by MR1 that are associated with the reported MAIT cell involvement in autoimmune disorders (29–31)? Finally, given the reported evidence for the infiltration of Vα7.2-Jα33 MAIT cells in kidney and brain tumors (32), are there tumor-derived Ags mediating MAIT cell activity in these diseases? Our structures provide a molecular model by which ligand presentation by MR1 can be studied in each of these areas and, importantly, shed light on the role of MAIT TCR diversity in the engagement of MR1-presented Ags.

We thank the staff of the General Medicine and Cancer Institutes Collaborative Access Team of the Advanced Photon Source (23ID) for the use of and assistance with X-ray beamlines, particularly Ruslan Sanishvili, Steven Corcoran, and Michael Becker for help and advice during data collection. We also sincerely thank Prof. Adelbert Bacher (Institute of Biochemistry and Food Chemistry, Food Chemistry Division, University of Hamburg, Hamburg, Germany) for very helpful discussions and assistance with experimental plans.

The authors have no financial conflicts of interest.