The molecular basis for Mucosal-Associated Invariant T cell recognition of MR1 proteins

Significance

Mucosal-associated invariant T (MAIT) cells are a highly conserved lineage of αβ T cells found in most mammals. These cells express a T-cell receptor of low diversity that recognizes vitamin metabolites presented by the MHC-related protein, MR1. Despite the evolutionary divergence of MR1 from other MHC proteins, we have found that MAIT T-cell receptors recognize MR1 using similar molecular strategies as that of the highly diverse, conventional αβ T cells, which recognize classical MHC molecules presenting peptide fragments. Our results also shed light onto how MR1-presented antigens can modulate the MAIT–T-cell receptor affinity and MAIT cell stimulation.

Abstract

Mucosal-associated invariant T (MAIT) cells are an evolutionarily conserved αβ T-cell lineage that express a semi-invariant T-cell receptor (TCR) restricted to the MHC related-1 (MR1) protein. MAIT cells are dependent upon MR1 expression and exposure to microbes for their development and stimulation, yet these cells can exhibit microbial-independent stimulation when responding to MR1 from different species. We have used this microbial-independent, cross-species reactivity of MAIT cells to define the molecular basis of MAIT-TCR/MR1 engagement and present here a 2.85 Å complex structure of a human MAIT-TCR bound to bovine MR1. The MR1 binding groove is similar in backbone structure to classical peptide-presenting MHC class I molecules (MHCp), yet is partially occluded by large aromatic residues that form cavities suitable for small ligand presentation. The docking of the MAIT-TCR on MR1 is perpendicular to the MR1 surface and straddles the MR1 α1 and α2 helices, similar to classical αβ TCR engagement of MHCp. However, the MAIT-TCR contacts are dominated by the α-chain, focused on the MR1 α2 helix. TCR β-chain contacts are mostly through the variable CDR3β loop that is positioned proximal to the CDR3α loop directly over the MR1 open groove. The elucidation of the MAIT TCR/MR1 complex structure explains how the semi-invariant MAIT-TCR engages the nonpolymorphic MR1 protein, and sheds light onto ligand discrimination by this cell type. Importantly, this structure also provides a critical link in our understanding of the evolution of αβ T-cell recognition of MHC and MHC-like ligands.

Mucosal-associated invariant T (MAIT) cells are a highly conserved T-cell subset found in most mammalian species (1–4). In humans, they can constitute up to 10% of circulating double-negative T cells, although they are much less frequent in mice (1, 5, 6). Most MAIT cells lack expression of the CD4 or CD8 coreceptors, although many MAIT cells express the αα form of the CD8 coreceptor (1). In humans, these cells are found at moderate frequency in the intestine and represent up to ∼50% of T cells in the liver (7). The cells exhibit an effector-memory phenotype and express the CD161 receptor (6). Their presence as mature effector cells in the periphery is dependent on B cells and the gut commensal flora (6, 8). Stimulated human MAIT cells can express both proinflammatory cytokines (IFN-γ, TNF-α, and IL-17) and cytolytic effectors (granzyme B) (7, 9, 10). MAIT cells are known best for their reactivity against various microorganisms from both bacterial and fungal origin (9, 10). These microorganisms include several important human pathogens, such as Mycobacterium tuberculosis, Salmonella typhimurium, and Staphylococcus aureus. Indeed, a significant proportion of the nonclassically restricted responding T cells in M. tuberculosis-infected individuals were determined to be of the MAIT lineage (9). MAIT cells have also demonstrated autoreactivity and have been associated with various autoimmune disorders (11, 12); they have also been found in both kidney and brain tumors (13).

MAIT cells are characterized by expression of a semi-invariant T-cell receptor that, in humans, is composed of Vα7.2 (TRAV1-2)/Jα33 α-chain paired with either Vβ13 (TRBV6) or Vβ2 (TRBV20) β-chains. MAIT T-cell receptors (TCRs), particularly the Vα domain, are highly evolutionarily conserved, exhibiting high amino acid sequence identity between human and cow (75%) and human and mouse (72%) (1). This sequence identity reaches 85% and 90%, respectively, when only the CDR loop-region sequences are compared. The CDR3α loop, encoded by the region of Vα-Jα rearrangement, is strictly conserved in length and highly conserved in amino acid sequence in MAIT TCRs, both within and between species. Junctional variability has been noted at only two of the CDR3α loop positions (underlined XXs): CAXXDSNYQL. In contrast, the CDR3β loop (encoded by the Vβ-Dβ-Jβ rearrangement) is highly diverse in MAIT TCRs. This evolutionary and sequence conservation is reminiscent of the semi-invariant TCRs of type I invariant natural killer T (iNKT) cells that recognize lipids presented by the MHC-like protein CD1d.

MAIT cells are restricted to the MHC class I-like molecule MR1, which is similarly conserved throughout mammalian evolution. The MR1 gene is encoded outside of the MHC locus, found ∼23 Mb away from the CD1 genes on Chromosome 1 in humans. Like CD1 molecules, MR1 exhibits low intraspecies variation and high-sequence conservation between mammalian species (3, 14). MR1 is ubiquitously expressed (14); however, cell-surface expression is low or undetectable in endogenous cells but is robust in MR1-transduced or -transfected cells (2). Recently, the precursor to riboflavin, 6,7-dimethyl-8-(1-d-ribityl)lumazine (DMRL), and chemical derivatives of it were shown to be presented by MR1 and could stimulate MAIT cells (15), shedding light on the mechanism of MAIT cell immunosurveillance of microbial species.

In addition to microbial reactivity, microbial-independent autoreactivity of human MAIT cells toward MR1-transfected cells has been noted (3, 16, 17), suggesting MAIT reactivity is not completely dependent on the presence of an exogenous, microbial antigen. Furthermore, microbial-independent cross-species reactivity between mouse and human MAIT cells and their MR1 orthologs (3) has been previously observed. Position 151 on the MR1 α2 helix was implicated in cross-species reactivity (16), and was further shown to enhance human MAIT cell reactivity when mutated to an alanine (18). Taken together, these observations suggest that MAIT cell reactivity to MR1 appears to be a balance between MR1 availability on the cell-surface, which may be modulated by endogenous, small-molecule ligands that can support MR1 expression and stability, and other features that can enhance TCR binding, such as species-specific differences in the α-helices, in particular at position 151, or the presence of stimulatory small-molecule ligands from exogenous sources.

MAIT cells have similarities to iNKT cells in their effector-phenotypes, their recognition of highly conserved MHC-like molecules (MR1 and CD1d, respectively) and their use of semi-invariant TCRs. Like MAIT cell TCRs, iNKT TCRs similarly have a conserved α-chain. The α-chain in iNKT TCRs is completely invariant within the iNKT cell population, but iNKT β-chains have preferred Vβ domains and diverse CDR3β loops. Structures of mouse and human iNKT TCR complexes with CD1d presenting various lipids have established that TCR recognition of CD1d is unlike classical αβ TCR recognition of MHC-peptide and follows a mostly conserved docking orientation mediated predominantly via the CDR3α loop via residues contributed by the Jα18 gene segment (19). Similarly, extensive mutagenesis of MAIT TCRs has established a bias of α-chain contribution to MR1 binding and mutagenesis of MR1 has provided a tentative footprint on the MR1 α1 and α2 helices (18). However, lacking definitive structural evidence of this interaction, the docking footprint of the MAIT TCR on MR1 remains unclear, as does the role of MR1-presented antigen in MAIT TCR engagement. We have taken advantage of the microbial-independent cross-species reactivity of MAIT cells to study the molecular basis for MAIT cell binding to MR1. Here we present the 2.85 Å structure of a human MAIT TCR in complex with bovine MR1. Our structure reveals a docking orientation reminiscent of diverse αβ TCR recognition, such as that observed in classical αβ TCR recognition of MHC/peptide (20) or noninvariant NKT recognition of CD1d/sulfatide (21, 22), and is distinctly different from that of the iNKT TCR bound to CD1d (19). There is a clear bias toward contacts contributed by the conserved MAIT α-chain CDR loops; however, the CDR2β and diverse CDR3β also contribute to binding. Our structure establishes the molecular basis for MAIT cell xeno-reactivity, providing an excellent model by which to understand the bona fide MAIT cell recognition of MR1 and how stimulatory ligand presentation from MR1 can enhance MAIT TCR binding and lead to MAIT cell activation.

Results

MR1 Has a Small Ligand-Binding Cavity Lined with Aromatic and Basic Residues.

Gene sequences for human MAIT TCRs F7, G2, and AE6 were derived from the published sequences of Tilloy et al. (1) and were synthesized via overlapping PCR. These MAIT TCRs were expressed in Escherichia coli and refolded to produce full-length heterodimeric TCRs, whereas a single-chain version of bovine β₂microglobulin (β₂m)-MR1 was expressed in insect cells for our biophysical and structural studies. Crystals of the MAIT TCR F7/MR1 complex were optimized and used to collect a complete dataset, which was finally refined to 2.85 Å resolution (Table 1). There was one complex of MAIT TCR/MR1 in the asymmetric unit.

Table 1.

Data collection and refinement statistics (molecular replacement)

Data collection and refinement	MAIT TCR F7 – MR1
Data collection
Space group	P 21 21 21
Cell dimensions
a, b, c (Å)	82.045, 87.41, 156.372
α, β, γ (°)	90.00, 90.00, 90.00
Resolution (Å)	50–2.85 (2.9–2.85)
Rsym	0.079 (0.681)
I/σI	14.63 (2.72)
Completeness (%)	99.55 (97.57)
Redundancy	4.1 (4.1)
Refinement
Resolution (Å)	2.86
Total No. reflections	108,886
No. unique reflections	26,556
Rwork/Rfree	0.2129/0.2695
No. atoms
Protein	6,285
Ligand/ion	28
Water	5
B-factors
Protein	65.20
Ligand/waters	73.50
R.m.s. deviations
Bond lengths (Å)	0.004
Bond angles (°)	0.88

Values in parentheses are for highest-resolution shell.

In the complex, bovine MR1 adopted an overall binary structure of a class I MHC fold associated with the β₂m subunit (Fig. 1A), highly similar to that of HLA-A2 and to the recent published human MR1 structure (Fig. 1B) (15). Of note was a slight shift in the α2 helix of our structure in relation to unliganded human MR1 (Fig. 1B); the flexibility of this region was further reflected in the two molecules of human MR1 in the asymmetric unit (shown as light and dark magenta), where a dramatic shift (up to 11 Å) was evident in this region. The structural differences between these molecules suggest the α2 helix is highly flexible and can adapt structurally in a context-dependent manner. MR1 has a cavity smaller (∼760 Å) than that of classical class I MHC molecules, similar to that noted previously (15). With the exception of a small opening or “portal,” this cavity is mostly closed on the left-most side, where it is lined with aromatic (tryptophan, tyrosine, and phenylalanine) and basic (arginine and lysine) residues, giving this region an overall basic charge (Fig. 1C), suggesting it can accommodate small molecule species that are polar and potentially negatively charged. Due to its similarity in location to the A pocket of classical class I MHC and A′ cavity of CD1 molecules, we refer to this region as the A′ cavity, similar to the designation used in ref. 15.

Fig. 1.

MR1 has a small, basic, aromatic-lined cavity optimal for binding small ligands. (A) Side view of a ribbon diagram of bovine MR1. The heavy chain, composed of the α1, α2, and α3 domains, is shown in cyan, β₂m in teal. (B) Comparison of bovine MR1 (bMR1) α1 and α2 platform domain (shown in cyan) with that of HLA-A2 (PDB ID code: 1DUZ, shown in purple); rmsd of the Cα backbones of the two platform domains is 0.581. At the bottom is a comparison of bovine MR1 with that of the two human MR1 structures identified in ref. 15 (PDB ID code: 4GUP). Bovine MR1 is shown in cyan and the two MR1 human structures (molecule “a” and molecule “b”) are shown in lavender and magenta, respectively. The 11 Å shift between the α2 helices of human MR1 molecules “a” and “b” is indicated by a red line. (C) Electrostatic surface representation of the MR1 cavity shown on a ribbon diagram of the MR1 CA backbone. The basic residues contributing to the overall basic charge of the cavity are shown (Inset), highlighted in yellow.

Clear electron density exists for a ligand in our MR1 structure (Fig. 2), enclosed by the aromatic (Y7, Y62, W156), basic (R9, K43, R94), and polar (S24 and T34) residues lining the A′ cavity. The electron density is similar in nature to the 6-FP previously identified as bound by human MR1 refolded in the presence of mammalian cell-culture media (15). Because MR1 used for this study was expressed in insect cells, there may be alternative compounds that can be presented that are either synthesized by insect cells or present in the insect cell media. The continuous electron density of this ligand with lysine 43 suggested that some, if not all, of this ligand was covalently attached via a Schiff base, as noted previously (15).

Fig. 2.

Ligand presentation by MR1. An omit electron density map of the residues and ligand found in the MR1 cavity. Side-chain electron density (1.5σ) is shown in green. 6-FP, identified in ref. 15, has been modeled into the ligand electron density (1σ) shown in yellow.

MAIT TCR Straddles MR1 α1 and α2 Domains in a Classical Docking Orientation.

The F7 MAIT TCR docks on MR1 in a perpendicular fashion, reminiscent of conventional αβ TCR recognition of MHC-peptide, with an orientation similar to that of the type II, noninvariant NKT TCR binding to CD1d presenting sulfatide (Fig. 3) (21, 22). The MAIT TCR/MR1 complex interface buries ∼1180 Ų of surface area, within the range of buried surface areas observed in αβ TCR/MHCp (23) and type II NKT TCR/CD1d structures (21, 22). The α-chain buries the majority of the surface area (∼55%) with its contacts biased toward the α2 helix of MR1. The β-chain buries ∼45%, with contacts biased toward the MR1 α1 helix (Fig. 4 and Table S1). This binding orientation is unlike that of the invariant NKT TCR recognition of CD1d, where the docking orientation is rotated more than 90° with most of the CD1d contacts established with the α chain, the CDR3α loop in particular (Fig. 3B, Lower Right). The MAIT TCR is oriented such that the CDR3α loop is situated over the A′ portal, perfectly positioned for contact with a suitable ligand should it be presented in this cavity.

Fig. 3.

The MAIT TCR docking orientation on MR1. (A) Ribbon diagrams of the complex structures of the human MAIT TCR with bovine MR1; conventional αβ TCR with MHC-peptide; type II NKT TCR with CD1d-sulfatide; and type I iNKT TCR with CD1d-α-Galactosylceramide (PDB ID codes: 2CKB, 4EI5, and 3HUJ, respectively). The TCR α- and β-chains are shown, respectively, in pink and marine (MAIT), yellow-orange and orange (αβ TCR), light green and green (type II NKT), and skyblue and darkblue (type I NKT). MR1 is shown in cyan, classical MHC in yellow, CD1d in green and slate for the type II and type I complexes, respectively. (B) Positioning of the CDR loops onto the respective MHC or MHC-like surfaces. MR1, classical MHC, and CD1d are all shown in white in a semitransparent surface representation. The CDR1, CDR2 and CDR3 loops from the respective TCRs are shown as they are positioned in each of the complexes. The dashed black lines represent the axis of binding, derived from a line extending between the conserved intrachain disulfides of the Vα and Vβ Ig domains. The loop coloring is the same as that defined in A.

Fig. 4.

Contacts mediated by the MAIT CDR loops on the MR1 surface. Shown in each panel are the side-chains of the CDR1α, CDR2α, CDR3α, CDR2β, and CDR3β loops and the residues they contact on the MR1 surface. No contacts were noted for the CDR1β loop. (Right, Upper) The distribution of VDW contacts (Upper) versus hydrogen-bond and electrostatic interactions (Lower). Contacts from the α-chain are shown in pink; β-chain in marine and residues contacted by both α- and β-chains are colored purple. Hydrogen-bond contact distances are ≤ 3.3 Å and are shown as dashed yellow lines.

Comparison of the human MAIT TCR structure in our complex with bovine MR1 with an unliganded human MAIT TCR reported by Reantragoon et al. (18) reveals only minor structural differences in the Vα domain conformation (Fig. S1), with an rmsd of 0.6 as determined by the DALI server (http://ekhidna.biocenter.helsinki.fi/dali_lite/start_). The CDR loop conformations of the Vα domains are highly similar between the liganded and unliganded TCRs (Fig. S1), suggesting that little conformational adjustments of the Vα CDR loops are required for recognition of MR1. Of note is the striking structural similarity of the CDR3α loop side-chains between these TCRs; even though these loops vary at two amino acid positions (90 and 91), the remaining side-chain conformations are remarkably conserved, including Y95. However, the positioning of the Vβ CDR loops are different, mainly because of the use of different Vβ domains in these two MAIT TCRs. This finding suggests that the Vβ CDR loops may be more adaptable during engagement of MR1.

All three CDR loops of the MAIT TCR α-chain contribute to MR1 recognition (Fig. 4 and Table S1). The germ-line–encoded CDR1α and CDR2α loops exclusively establish contacts with MR1 residues of the α2 helix. The majority of contacts are van der Waals (VDW); however, several key hydrogen-bonding contacts are established with these loops, in particular the main-chain atoms of F29 from the CDR1α loop hydrogen-bond with side-chains of N155 and E160 of MR1, and E55 of the CDR2α loop forms a salt-bridge and hydrogen-bond with R147 of MR1 (Table S1). One α-chain contact lies outside of the canonical CDR loops, R66 from β-strand-D of the Vα domain establishes VDW and weak electrostatic interactions with E159 of the α2 helix. The CDR3α loop establishes both VDW and hydrogen-bond contacts with both the α1 and α2 helices, with the majority of contacts coming from Y95. This residue hydrogen-bonds with R61 of the α1 helix and W156 of the α2 helix and forms VDW contacts with R61, L65, W69 of the α1 helix, and Y152 and W156 of the α2 helix. Y95 is also positioned directly over the portal containing the ligand density, suggesting this residue may establish direct contact with a stimulatory ligand. The main-chain nitrogen of Q96 of the CDR3α loop also hydrogen-bonds with the R61 side-chain of MR1. Of note, all of the residues of the CDR3α loop that contact MR1 derive from the Jα33 (TRAJ33) segment used in MAIT TCRs. The MAIT TCR α-chain has been characterized in several different species and is highly conserved. Examination of the residues of the TCR α-chain observed to contact MR1 in this complex structure reveals that only 2 of the 12 α-chain residues used in contacting MR1 are variable across human, mouse, and cow (Fig. S2). Human E55 is a lysine in the mouse Vα homolog, whereas human S93 is a glycine in the cow MAIT Vα domain.

Contacts of the MAIT TCR β-chain to MR1 are restricted to the germ-line–encoded CDR2 loop and the diverse CDR3 loop. With the exception of T54 and T55, all contacts from the CDR2β (Y48, A50, S51, and D56) are VDW; T54 and T55 establish hydrogen-bonds with the side-chain of Q64 (Fig. 4 and Table S1). Six CDR3β loop residues (W96, T97, G98, E99, G100, and S101) establish contacts with the α1 and α2 helices, the majority of them from VDW interactions. S101 from the CDR3β loop makes hydrogen-bond contacts with main-chain and side-chain atoms of E149 of the MR1 α2 helix. In contrast to the conserved α-chain contacts, those of the β-chain are less conserved both within and between species. Only T55, which establishes VDW and hydrogen-bond contacts with Q64 of MR1, is conserved in the human Vβ domains used in MAIT TCRs; this position is variable in its mouse counterparts (Fig. S2).

Mutagenesis analyses of autoreactive MAIT TCR/MR1 interactions (17) and those in the presence of microbial infection (17, 18) are both consistent with the overall footprint of the human MAIT TCR on bovine MR1 described here. Both the mouse and human studies describe an α-chain bias in MR1 recognition (18), consistent with the major contribution to binding coming from the MAIT TCR α-chain observed in this complex structure. Most of the residues in either human or mouse MAIT TCR α-chains that were determined to be important for MR1 recognition (17, 18) establish contacts with MR1 in our structure, specifically: G28α, F29α, N30α, Y48α, V50α, L51α, S93α, N94α, and Y95α. It is important to note that these residues are strictly conserved between mouse and human MAIT TCRs. Of particular importance is residue Y95α, which establishes eight contacts with MR1 and is perfectly positioned to establish contacts with a ligand presented in the A′ cavity. Mutagenesis of this residue abrogated MAIT cell autoreactivity and microbial-dependent MR1 stimulation. As discussed later, modeling of stimulatory compounds identified by Kjer-Nielsen et al. (15) into the A′ cavity of MR1 in our complex structure supports the importance of this residue in ligand recognition. The presence of a stimulatory ligand likely supplies additional contacts that overcome the binding threshold required for MAIT cell activation.

Positions 72, 147, and 151 Mediate the Xeno-Reactive Interactions Between Human MAIT TCRs and Bovine MR1.

To identify positions contributing to MAIT cell cross-reactivity, the amino acid differences between human and bovine MR1 were mapped onto our MR1 structure (Fig. 5). Three differences are at contact sites with the human MAIT TCR: A72 of the α1 helix and R147 and Q151 of the α2 helix, which are M72, Q147 and L151 in human MR1, respectively (Fig. S3). Although position 72 is contacted only by one residue of the CDR3β loop (W96), Q151 and R147 are surrounded by residues V50, L51, and E55 of the CDR2α loop (Figs. 4 and 5). Position 151 has been noted to be a position that modulates MAIT cell reactivity during investigation of xeno-reactivity of MAIT cells to MR1 (3). In addition, mutagenesis of this position in human MR1 to alanine stimulated microbial-independent human MAIT cell reactivity (18), suggesting it plays an important role in MAIT TCR binding. This effect of position 151 could be a direct result of a steric interaction of the 151 side-chain (a leucine in humans) interfering with the CDR2α loop binding, or this residue may affect the positioning of the neighboring residue at 147. R147 forms a salt-bridge with E55 of the CDR2α loop, although the equivalent position in human, glutamine, would also be capable of forming hydrogen-bond interactions with E55.

Fig. 5.

Sites of variability between human and bovine MR1. Variable sites are indicated in pink on a semitransparent surface representation of bovine MR1. The contacting CDR loops of the MAIT TCR are shown, with the side-chains where contacts with variable residues are established. One residue in the α1 helix differs between human and cow, position 72, which is an alanine in cow and a methionine in human. Two positions vary in the α2 helix, 147 and 151, which are arginine and glutamine in cow, respectively, and glutamine and leucine in human, respectively. Of note is position 151, noted previously to mediate species cross-reactivity (3) and, upon mutation to alanine, induce potent autoreactivity by MAIT cells (18).

To evaluate the effect of these positions on MAIT TCR binding to MR1, we first performed surface plasmon resonance (SPR) on the F7, G2, and AE6 TCRs (Fig. 6, Upper) with wild-type bovine MR1 to assess the affinity of this xeno-reactivity. F7 and G2 MAIT TCRs bound wild-type bovine MR1 with low but similar affinity, 31 μM and 39 μM, respectively. The AE6 TCR, which uses a different Vβ domain (TRVB6_2 in lieu of TRVB6_1 in F7 and G2), and thus differs from F7 and G2 in the CDR2β and CDR3β loops (Fig. S4), bound wild-type bovine MR1 with nearly a twofold weaker affinity, 74 μM. We then measured, using bio-layer interferometry (Blitz), the affinity of the F7 MAIT TCR for single bovine MR1 mutants expressing the human residues at positions 72 (A to M), 147 (R to Q), and 151 (Q to L) and a “humanized” version containing all three mutations. The A72M mutation enhanced binding ∼190% (Fig. 6, Lower) (∼19.5 μM) (Fig. S5), whereas the R147Q mutation decreased binding to ∼40% of wild-type levels (∼91 μM). Mutation of Q151 to leucine decreased binding to an undetectable level. However, the F7 TCR bound the triple “humanized” mutant with ∼55% the affinity of wild-type (∼65 μM). Although position 151 played the most significant role in MAIT TCR recognition as revealed through our mutagenesis, both positions 72 and 147 modulated binding, consistent with our observed footprint of the MAIT TCR on the MR1 surface.

Fig. 6.

Affinity of human MAIT TCRs for bMR1 and humanized bMR1. (Upper) SPR curves are shown between three human MAIT TCRs [F7, G2, and AE6 (1)] and bovine MR1. Dissociation constants (K_D) were derived from equilibrium analysis, the fits are shown for their respective SPR curves. (Lower) Residues that differ between human and bovine MR1 (72, 147, and 151) were converted to the human sequence and measured for changes in binding by the human F7 MAIT TCR by bio-layer interferometry (Blitz). Values shown are percent binding of wild-type (K_Dmut/K_Dwt).

Finally, to probe the influence of the CDR3β loop on MAIT TCR recognition of MR1, we have included a mutation of position E149 of MR1 to alanine in our binding studies. E149 of MR1 is the only residue to which hydrogen-bonds are formed by the CDR3β (S101); however, it also establishes VDW contacts with E99, G100, and S101 (Fig. 3 and Table S1). Mutation of this position to alanine had little effect upon MAIT TCR binding, suggesting that specific, side-chain–mediated contacts of the CDR3β may not play a central role in MR1 recognition.

These results establish that: (i) MAIT TCRs recognize MR1 with a range of affinities with contributions from the Vβ domain (TRVB6_1 TCRs in this study bind more strongly than TRVB6_2); (ii) positions 72, 147, and 151 of MR1 act in concert to modulate MAIT cell xeno-reactivity by enhancing the MAIT TCR binding affinity of MR1 approximately twofold; and (iii) the variable CDR3β loop of this MAIT TCR (F7), although establishing numerous VDW contacts with MR1, may not play a critical role in MR1 recognition. This finding does not, however, rule out the possibility that the CDR3β loop may play an important role in modulating MAIT TCR binding and that variability at this loop between MAIT TCRs may result in different thresholds of MR1-dependent stimulation, similar to what was observed in iNKT cell modulation by CDR3β loop variability (24).

Optimized Ligand Docking in MR1 Provides Additional Contacts to the MAIT TCR Through Y95 in the CDR3α.

To explore the effect of MR1 presented ligand on MAIT TCR activation, we used Autodock Vina to model how the stimulatory compounds identified in ref. 15 could be presented by MR1 and enhance binding of the MAIT TCR. The riboflavin precursor, DMRL and its chemical derivative, reduced 6-hydroxymethyl-8-d-ribityllumazine (rRL-6-CH₂OH), each underwent 10 independent Vina docking runs. Side-chains within the MR1 binding pocket, as well as Y95α of the MAIT TCR, were allowed to be flexible. Each round generated nine ligand conformations, ranked by their calculated binding energies, yielding the same top-scoring structure for each ligand in 8 of the 10 runs. The best poses reported energies of −9.2 and −8.4 kcal/mole for DMRL and rRL-6-CH₂OH, respectively. Shown in Fig. 7 are the orientations of these two ligands in the MR1 binding cavity. In each case, the lumazine group is stacked between the aromatic residues Y7 and Y62 and enclosed by the polar residues R9, S24, K43, and R94. W156 also contributes to VDW contacts with both ligands. The orientation of the lumazine group, although in the same plane, is flipped between the two ligands. DMRL has its two methyl groups at the 6 and 7 position oriented toward the more nonpolar side of the binding cavity, whereas the two carbonyl groups are oriented toward the polar residues. rRL-6-CH2OH, in contrast, has its hydroxymethyl group oriented toward the polar side-chains and the carbonyl groups oriented toward the hydroxyls of Y7 and Y62. Despite the flipped orientations of the lumazine group, the ribityl chain is essentially positioned in the same spot. In both cases, the first hydroxyl of the ribityl chain is positioned within hydrogen-bonding distance to Y95 of the CDR3α loop, providing an important ligand-mediated contact that could enhance the binding affinity of the MAIT TCR and initiate T-cell activation. We also performed an identical docking analysis of the nonstimulatory, noncovalently bound compound, 6-FP (Fig. S6), and found similar conformations of this ligand in the MR1 groove, suggesting these are the preferred orientations of the aromatic rings in the binding pocket.

Fig. 7.

MAIT TCR engagement of stimulatory antigens revealed through ligand modeling. Optimized positions of the DMRL (and its chemical derivative, reduced rRL-6-CH₂OH) are shown in the MR1 ligand binding cavity. DMRL is shown in yellow, rRL-6-CH₂OH in green. The side-chains lining the cavity, as well as residue 95 of the CDR3α loop, were allowed to move; the alternate conformations of these side-chains from that of our original MR1 structure are shown in white. These positions represent the most energetically favorable of eight alternate conformations, with energies of −9.2 and −8.4 kcal/mole for DMRL and rRL-6-CH₂OH, respectively. The conformations of both ligands are such that the first hydroxyl group of the ribityl chain engages the MAIT TCR through residue Y95 of the CDR3α loop (shown in pink), via a hydrogen bond (≥3.3 Å, shown as yellow dashed line). The chemical structures of DMRL and rRL-6-CH₂OH are shown beneath their modeled positions.

Discussion

MAIT cells constitute a considerable percentage of circulating CD8⁻/CD4⁻ and CD8αα⁺ T cells and have been shown to respond to important human microbial pathogens such as M. tuberculosis, S. typhimurium, and S. aureus through their recognition of MR1. The broad conservation of MAIT cells in mammals, restriction to an equally highly conserved MHC molecule, and their prevalence in barrier, mucosal tissues suggests that MAIT cell microbial detection has been an adaptation widely used in mammalian host defense. In addition to microbial detection, MAIT cells have been shown to demonstrate microbial-independent autoreactivity when MR1 is expressed at high levels on antigen-presenting cells, which may provide a link to MAIT cell involvement in autoimmune disorders and tumor immunosurveillance (11, 12, 25). This microbial-independent reactivity of MAIT cells is further enhanced by modification of a single amino acid residue, position 151 on the α2 helix. This residue was found to be central to the cross-species reactivity used in our study. We have capitalized on the close sequence identity between bovine and human MR1 and the microbial-independent reactivity between bovine MR1 and human MAIT cells to provide insight into how MAIT cells, through their semi-invariant TCR, recognize MR1. We also extend a model for the role of the riboflavin precurosor, DMRL and its derivative, rRL-6-CH2OH (15), in MAIT TCR engagement during microbial detection.

Despite the similarity of the semi-invariant MAIT TCR to that of iNKT cells, the MAIT TCR engages MR1 with a footprint and orientation unlike that seen in iNKT/CD1d complex structures. Instead, the orientation of the MAIT TCR on MR1 is highly similar to that of diverse type II NKT TCRs bound to CD1d/sulfatide (21, 22), and classical αβ TCR recognition of MHC-peptide (20). Of particular note is the focus of the germ-line–encoded CDR loops (CDR1 and CDR2) onto the MR1 helices; this is highly reminiscent of classical αβ TCR recognition, whereby the CDR1 and CDR2 loops of the α-chain establish contacts with the α2 helix and those of the β-chain contact the α1 helix (20). In lieu of a peptide, the MAIT TCR CDR3α and -β loops establish contacts with both helices and, in the case of the CDR3α loop, acts as a sensor for MR1-presented antigens. Several lines of evidence support the dominance of the Vα domain in MR1 recognition. This domain is essentially invariant in the MAIT cell population (only two residues in the CDR3α loop vary) and in our structure there is a bias of the docking footprint on MR1 toward the Vα CDR loops. Previous mutagenesis of MAIT TCR CDR loops (17, 18) are consistent with the Vα dominance in our structure, and finally, there is no significant conformational change of the CDRα loops, or their side-chains, between liganded and unliganded MAIT TCRs. Finally, mutation of an MR1 contact residue with the CDR3β loop, E149, showed little effect on MAIT TCR binding. Taken together, these data suggest that MAIT TCRs engage MR1 predominantly through the Vα CDR loops, whereas the contacts mediated by the Vβ CDR loops provide ancillary contacts that can vary across the different Vβ domains and the diverse CDR3β loops present in MAIT TCRs.

The MAIT α-chain has only two variable positions located within the CDR3 loop, where the V-J junction occurs during rearrangement. All of the CDR3α loop contacts noted in our structure, however, derive from the amino acids encoded by the Jα33 gene segment used during MAIT Vα rearrangement, a key similarity to that of invariant NKT cells, where the contribution of the Jα18 gene segment is pivotal for iNKT TCR recognition. Y95 is a particularly important residue in the MAIT CDR3α loop as it not only establishes contacts with many MR1 residues, but is also positioned directly over the opening to the MR1 binding cavity where 6-FP and related compounds are bound and where our modeling has favored binding of the stimulatory riboflavin derivatives DMRL and rRL-6-CH2OH. It is likely that contacts established by this residue “sense” presentation of stimulatory ligands by MR1, contributing to enhanced T-cell engagement and subsequent MAIT cell activation. The minor contacts that the MAIT TCR establishes with our modeled stimulatory compounds is in direct contrast to what is observed in conventional αβ TCR recognition of MHC-peptide, where TCR contacts with the peptide comprise 20–30% of the contact surface (23), and iNKT TCR contact with CD1d/aGalCer, where ∼36% of TCR contacts are with the lipid headgroup (19). Indeed, differences in the headgroup of the lipid antigen presented by CD1d that are recognized by iNKT TCRs can have profound effects on iNKT effector outcome (26). It is conceivable that other, yet to be identified, stimulatory compounds might establish more contacts with the TCR and could modulate MAIT cell activity; however, the data to date suggests a minor role for ligand in TCR engagement.

The F7 and G2 MAIT TCRs examined in our study retain the ability to bind our “humanized” MR1 (bovine MR1 with the three species-specific differences at positions 72, 147, and 151 mutated to their respective human amino acids: M, Q, and L) with a K_d of ∼60 μM, suggesting that human MAIT TCRs can bind human MR1 in the absence of exogenous stimulatory ligands. This finding is consistent with the previously observed microbial-independent autoreactivity and may be the basis for MAIT cell reactivity in autoimmunity and tumor surveillance. Presentation of certain ligands that can provide additional contacts to the TCR, such as DMRL or rRL-6-CH₂OH, would enhance activation. However, these contacts with the proposed stimulatory compounds that we have noted in our complex structure and through modeling are not substantial in relation to the other contacts between the TCR and MR1, suggesting that T-cell stimulation through TCR engagement is a very finely tuned process. Although mutagenesis of Y95 of the CDR3α chain was shown to abrogate microbial-dependent MAIT cell activation (18), this residue establishes numerous contacts with MR1 independent of bound ligand and only participates in one hydrogen bond with our modeled stimulatory compounds. Thus, the trigger for MAIT cell activation likely comes from a combination of two factors, the first being enhanced stability of MR1 on the cell surface; we suggest this can be through binding a range of ligands that are suitable for docking in the cavity of MR1, and second, through presentation of compounds that can provide additional contacts to the MAIT TCR, such as our modeled conformations of DMRL and rRL-6-CH₂OH.

Although DMRL and rRL-6-CH₂OH have been shown to induce MAIT cell stimulation, we propose that there are additional classes of endogenous and exogenous antigens that may stabilize MR1 expression on the cell surface and may also provide additional contacts to the MAIT TCR through position Y95α. These compounds may be acidic in nature (electrostatically compatible with the basic MR1 groove) and aromatic, such as what is observed with DMRL, rRL-6-CH2OH, and other derivatives also shown to stimulate MAIT cells.

Self/nonself recognition by T cells is a key feature of immunity; this includes self/foreign peptide recognition (via classical MHC molecules) by classical αβ T cells and lipid recognition by CD1-restricted T cells. However, the closest analogy to MAIT cell recognition of small molecule metabolites presented by MR1 may be detection of metabolite phosphoantigens by human Vγ9Vδ2 T cells, whereby both foreign and self metabolites can provide potent stimulatory signals to this T-cell lineage. Although the molecular mechanisms behind this activation process have only begun to be elucidated (27, 28), it is testament to the importance of metabolic processes to immune assessment of the health of the cell. It may very well be that MAIT surveillance is based on similar endogenous and exogenous metabolite profiling.

Materials and Methods

Recombinant MR1 Expression and Purification.

The cDNAs corresponding to the ectodomain region of the bovine β₂m and bovine MR1, linked through a glycine-serine flexible linker, were cloned in-frame with a C-terminal 6- or 12xHis tag into the pAcGP67A vector (BD Biosciences). This single-chain bovine β₂m-Gly-Ser-MR1 construct (scMR1) was expressed in Hi5 cells via baculovirus transduction. The recombinant protein was first captured following addition of Nickel NTA Agarose (Qiagen) and then subjected to anion exchange chromatography in a MonoQ column (GE Healthcare). For crystallographic purposes, the protein was then treated with Carboxypeptidase A (Sigma) as previously described (29), followed by a final step of size-exclusion chromatography.

Recombinant MAIT TCR Expression and Purification.

The cDNA samples for the MR1-reactive MAIT TCR clones F7, G2, and AE6 were synthesized via overlapping PCR from sequences reported by Tilloy et al. (1). These cDNAs were modified to bear T48C and S57C mutations at the α- and β-constant chain domains, respectively, and were separately cloned into different versions of the pAcGP67A. Each chain contains a C-terminal 3C protease site followed by either acidic or basic zippers and a 6xHis tag. For each clone, both α- and β-chains were coexpressed in Hi5 cells via baculovirus transduction. The heterodimeric TCRs were captured with Nickel NTA Agarose and further purified by anion exchange and size-exclusion chromatography. For SPR studies, the purified TCRs were treated with 3C protease and the digested sample loaded onto a Nickel NTA agarose column. The His-tag–free fraction was collected from the flow-through and used as analyte for SPR. For crystallization purposes, the MAIT TCR chains were separately expressed in E. coli as inclusion bodies and then each clone refolded, treated with Carboxypeptidase A and purified as previously described (29).

Generation of MR1 and MAIT TCR Mutants.

The scMR1 mutants: A72M, R147Q, E149A, Q151L, and the triple mutant, A72M/R147Q/Q151L were generated through overlapping PCR with specific primers containing the desired mutations. This mutant was expressed in insect cells as described for the wild-type counterpart.

MR1 and MAIT TCR Biophysical Interaction Analysis.

The 12xHis tag scMR1 or the Q151L mutant were captured to a stable level of 1,500 RUs on the surface of a NiCl₂-treated NTA sensor chip (GE Healthcare). Insect cell-derived recombinant FC was captured in the flow channel 1 to subtract nonspecific binding events. Increasing concentrations (0.625, 1.25, 2.5, 5, 10, 20, and 40 μM) of each TCR clone were injected at a flow rate of 30 μL/min using 10 mM Hepes pH 7.4, 150 mM NaCl, and 0.005% Tween-20 as running buffer. The traces of the reference-subtracted binding signals were plotted and GraphPad Prism was used to determine the affinity constants for the interactions with the immobilized MR1.

All interaction analyses of the MR1 mutants were carried out in real time by bio-layer interferometry in a Blitz System Package (Fortebio). The 12xHis scMR1 or mutants of it were captured to a stable level of 4 ηm units on a Ni-NTA (NTA) Biosensor. A series of MAIT TCR concentrations ranging from 0.37 to 90 μM were ran over the immobilized MR1 protein in 10 mM Hepes pH 7.4, 150 mM NaCl, and 0.005% Tween-20 and the association and dissociation binding traces were recorded. FC was used as a “nonbinding” control partner. Responses from measurements with FC were subtracted from those with MR1 and interaction affinity K_ds were calculated with GraphPad Prism by plotting the binding values at equilibrium against the TCR concentrations.

Ternary Complex Formation and Crystallization.

Equimolar amounts of refolded F7 MAIT TCR and scMR1 were mixed and concentrated to 8.5 mg/m. Initial crystals were found in 0.1 M [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid] pH 7.0 and 1.5 M ammonium sulfate. These crystals were crushed and used for microseeding fresh drops. This procedure yielded single crystals that were used for data collection.

Crystallographic Data Collection, Structure Determination, and Refinement.

Crystals were cryo-cooled in mother liquor supplemented with 20% (vol/vol) glycerol before data collection. All datasets were collected on a MAR300 CCD at beamline 23 ID-D at the Advanced Photon Source at the Argonne National Laboratory and processed with HKL2000 (30). The structure of the ternary complex was solved by molecular replacement with the program Phaser (31) and using as search model entities the coordinates of the ligandless HLA-B*4103 (PDB ID code 3LN4) for β₂m-MR1 and the previously solved MAIT TCR coordinates (PDB ID code 4DZB) with the CDR loops omitted for the F7 MAIT TCR. Refinement was accomplished with Phenix software suite (32) and Refmac5 (33) by initially dividing the molecules into rigid bodies and subsequent jelly-body and restrained refinement. Next, extensive cycles of manual building in Coot (34) and restrained refinement were carried out and missing regions were built in accordance to the electron density maps. PRODRG (35) was used for the generation of the ligand-modified lysine in MR1 and included in subsequent refinement and manual building steps. All of the refinement procedure was performed taking a random 5% of reflections and excluding them for statistical validation purposes (R_free).

Structure Analysis.

Intermolecular contacts and distances were calculated using the program Contacts from the CCP4 software package (36), interface surface areas were calculated using the PISA server (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 for the MAIT TCR–scMR1 complex have been deposited in the Protein Data Bank under the ID code 4IIQ.

Docking of Stimulatory Ligands.

Autodock Vina v1.1.2 (37) was used to perform docking calculations on the ligands DMRL, rRL-6-CH2OH, and 6-FP to the complex structure. Because of the stochastic nature of Autodock Vina, we ran each ligand calculation 10 times, yielding the same top scoring structure for each ligand in 8 of the 10 runs. Vina was run using an x, y, z box size of 24, 18, 28 centered at x, y, z coordinates −17.1, 37.0, 38.6 with flexible residues. The flexible residues were TYR7, ARG9, SER24, LYS43, TYR62, ARG94, and TRP156 on MR1 (chain C) and Y95 on the MAIT TCR α-chain (chain A). All other Vina parameters were set to the default.