Introduction

On the surface of all nucleated cells, class I major histocompatibility complex (MHC-I) molecules bind and present a repertoire of peptides derived from the endogenous proteome as a means of immune surveillance by CD8 T cells (1). In an analogous process, the non-classical MHC-I-related protein 1 (MR1) displays small molecule ligands derived from the endogenous and exogenous metabolome to Mucosal-associated invariant T (MAIT) cells and MR1-restricted T cells (2). The MR1 processing and presentation pathway begins with the assembly of a nascent MR1 heavy chain with the light chain beta-2 microglobulin (β₂m) in the endoplasmic reticulum (ER) (3). The assembly process is facilitated by dedicated molecular chaperones: tapasin, which is restricted within the ER-resident peptide-loading complex (PLC), and the homologous, PLC-independent TAP-binding protein related (TAPBPR) (4, 5). However, unlike classical MHC-I, MR1 molecules are only transiently expressed at the cell surface and are rapidly internalized (3, 6). Under steady state conditions, nascent MR1 molecules are localized to the ER, vesicular pools, and endocytic compartments (7-9). These empty MR1 molecules are highly unstable and prone to aggregation. Therefore, MR1 folding and metabolite binding are cooperative processes. The availability of robust ligands triggers MR1 to egress out of the ER, and traffic to the cell surface (4, 10). Alternatively, partially folded MR1 can capture ligands on the cell surface or within vesicular pools, endocytic compartments, and phagosomes to orchestrate host responses (11). Finally, MR1 has a cytoplasmic tyrosine-based motif that is recognized by the endocytic adaptor protein 2 (AP2) complex. This interaction controls the kinetics of MR1 internalization from the cell surface, and minimizes recycling to define the duration of metabolite presentation (12). The importance of MR1 for human health is highlighted by the recent discovery that the human herpesviruses HSV-1 and CMV can disrupt MR1 expression through direct inhibition of the physiological ligand loading process (13).

Canonical MR1 ligands are derived from bacterial metabolite synthesis pathways (14, 15). The two prominent classes are lumazine/pyrimidine derivatives of riboflavin/vitamin B2 (16, 17) and pterins derived from folate/vitamin B9 (18, 19). While RL-6,7-DiMe, Ac-6-FP, and 5-OP-RU all upregulate MR1 surface expression, RL-6,7-DiMe and 5-OP-RU are agonists while 6-FP is antagonistic to T cells (14, 20). Other microbial MR1 ligands, such as 7,8-didemethyl-8-hydroxy-5-deazariboflavin (FO), photolumazines, hesperidin and riboflavin, exhibit pleiotropic properties (15). In contrast to the six polymorphic pockets in the peptide binding groove of MHC-I (A-F), the MR1 groove comprises only A’ and F’ pockets (18, 21). All MR1 ligands characterized to date bind exclusively to the A’ pocket rather than the much shallower MR1 F’ pocket (20). It has been proposed that the formation of a Schiff base between the conserved K43 buried in the A’ pocket and metabolites promotes stable MR1 folding (3, 18). However, several potent metabolites have been identified which associate with MR1 in a non-covalent manner but still upregulate surface expression levels (14, 22, 23). Consistent with this observation, MR1 K43A mutants can still bind ligands, albeit with decreased thermal stabilities (19, 24), suggesting that K43 Schiff base formation may not be the only determinant for MR1 egress out of the ER, and that both covalent and non-covalent ligands can form stable complexes with MR1. These knowledge gaps preclude a detailed understanding of what defines a physiologically relevant repertoire of ligands from the cellular metabolome (15).

Previous studies aiming to stabilize the MR1 structure for recombinant protein production and ligand screening applications have used either a chimeric approach where human α₁/α₂ platform is coupled to bovine α₃/β₂m domains (hpMR1)(25), or, more recently an “open” MR1 format, with an engineered disulfide bond at the heavy and light chain interface (26). We have recently developed SMART MHC proteins, which consist of a single-chain α₁/α₂ binding groove with the α₃ and β₂m domains replaced by a helical stabilizing domain (27). Here, we leverage the SMART platform to create a stable minimalistic system for studying MR1 in vitro. We demonstrate that SMART-MR1 can capture different classes of known ligands and maintains binding to cognate T cell receptors (TCRs), while also offering new avenues for studying MR1 interactions by solution NMR. We also solve the three-dimensional structure of SMART-MR1 in complex with A-F7, illustrating that SMART-MR1 presents ligands and binds to TCRs similarly to native MR1. Our results demonstrate that SMART-MR1 is a miniaturized system with native-like properties, expanding the potential for studying MR1 using standard biochemical and structural techniques.

Results

Design and purification of SMART-MR1

To determine whether MR1 is a suitable candidate for adapting it to the SMART system, we compared the interface formed by β₂m and the α₁/α₂ “platform” between the structures of human metabolite-bound MR1 (PDB 6PUC) and peptide-bound HLA-A*02:01 (A02, PDB 9SKO). Despite sharing limited (43.6%) overall sequence homology, the α₁/α₂ domains of MR1 and A02 bear considerable structural similarity (Fig. 1A). In both structures, the interface is formed by polar and hydrophobic contacts between key β₂m and α₁/α₂ residues that are conserved between A02 and MR1. Specifically, W60 from β₂m is nestled within a hydrophobic pocket formed by the conserved residue pair M98/95 and F8 under the α₁/α₂ β-sheet, and this interaction plays a central role in both structures. The W60 ring forms a hydrogen bond with D122/118 in A02/MR1, further stabilizing the β₂m/heavy chain interface. Indeed, a key tryptophan at the analogous position in the interface was shown to be favorable when screening for different SMART MHC stabilizing domains (27). Additional stabilizing polar contacts are formed between β₂m residues R3 and D53 with the analogous residues R48/46 and G120/116 in A02/MR1. This striking degree of structural similarity of the β₂m/heavy chain interface in MR1 and A02 suggests that the SMART MHC stabilizing domain can also form the necessary contacts required to stabilize the α₁/α₂ MR1 platform. To this end, we generated a SMART-MR1 construct by joining the N-terminus of the MR1 α₁/α₂ domain to the C-terminus of the SMART MHC stabilizing domain via a rigid linker. Using AlphaFold3, we generated a structural model of SMART-MR1 and found that it closely recapitulated the features of the SMART MHC X-ray structure. In this model, the stabilizing domain formed a network of favorable contacts with the MR1 α₁/α₂ platform. Notably W17 of the SMART-MR1 occupied the hydrophobic pocket under the α₁/α₂ β-sheet, playing an equivalent role to W60 from β₂m in the native MR1 structure (Fig. 1B).

Figure 1.

We next sought to express and purify SMART-MR1 from E. coli lysates. Our previous work has shown that, while SMART MHC can be expressed in soluble, peptide-receptive form, refolding of the same protein in the presence of excess peptide results in enhanced native structural features and improved affinity for cognate T cell receptors (TCRs) (27). These results suggest that refolded SMART proteins confer superior functionality. We therefore chose to refold SMART-MR1 from E. coli inclusion bodies using standard procedures (25, 28). Purification of SMART-MR1 refolded in the presence of the covalent ligand 6-FP by size-exclusion chromatography (SEC) resulted in a monodisperse peak at the expected molecular weight (Fig.1C and SI Appendix, Fig. S1A). To evaluate the quality of the purified SMART-MR1/6-FP complex we compared its thermal stability to full-length hpMR1/6-FP using differential scanning fluorimetry (DSF). The thermal stability of MR1 complexes is highly dependent on the occupancy a ligand within the A’ pocket (5), thus MR1 unfolding is coupled with dissociation of the ligand from the protein. The melting temperature (T_m, °C) of SMART-MR1/6-FP was nearly identical to hpMR1/6-FP (56.0 vs 55.9 °C), indicating that SMART-MR1 can bind ligands and present them in a native-like manner (Fig. 1D and SI Appendix, Fig. S1B). Together, these results demonstrate that the SMART stabilizing domain can bolster the ligand-binding α₁/α₂ platform of MR1 effectively serving as a functional surrogate of the α₃ and β₂m domains.

Engineered SMART-MR1 can accommodate several classes of known ligands

MR1 is capable of binding to multiple chemically distinct classes of small molecule metabolites. To explore whether SMART-MR1 can capture these diverse ligands, we refolded SMART-MR1 in the absence (empty) or presence of vitamin B9 metabolites (Ac-6-FP, 6-FP), vitamin B2 metabolites (5-OP-RU), and drug or drug-like molecules (3FSA, 5FSA) (Fig. 2A). For all ligands tested, refolding yields improved compared to refolding in the absence of ligand (Fig. 2B and SI Appendix, Fig. S2 A and C). Moreover, T_m values could only be determined when SMART-MR1 was refolded in the presence of a ligand (Fig. 2C and SI Appendix, Fig. S2B and C), likely due to the binding of the hydrophobic dye to the groove of empty SMART-MR1. T_m values also correlated with refolding yields (R² = 0.503, Fig. 2D). These results indicate that SMART-MR1 binds to different classes of known MR1 ligands, and that, like native MR1, its protein stability is highly dependent upon the presence of bound ligand.

Figure 2.

SMART-MR1 is properly conformed and binds the A-F7 TCR with high affinity

To further characterize the structural and biochemical properties of SMART-MR1, and to explore potential applications of our system, we applied solution-based biophysical techniques. A major advantage of SMART-MR1 is that it drastically reduces the size of the complex compared to native MR1 (43 vs 29 kDa). This reduction in size mitigates relaxation losses during NMR experiments, allowing for facile backbone assignments and characterization of interactions with small molecules, immunoreceptors and molecular chaperones, whereas the study of WT MR1 necessitates more elaborate sample preparation with site-selective isotopic labelling (5). To test the feasibility of amide-based NMR experiments, we refolded isotopically labeled SMART-MR1 with ²H, ¹³C, and ¹⁵N in complex with 6-FP, and performed 2D TROSY experiments (Fig. 3A). The 2D TROSY spectrum was well dispersed, as expected for a properly conformed globular protein of 29 kDa, indicating that residue-level information on dynamics and interactions with different binding partners can be obtained using standard solution NMR methods.

Figure 3.

We next sought to determine whether SMART-MR1 can be used as a tool to screen for ligand binding. We refolded SMART-MR1 with the non-covalent ligand DCF and used a fluorescent 5-OP-RU analog (JYM20) (4) to perform competition experiments in the presence of a second ligand using fluorescence polarization (FP)(Fig. 3B). Upon incubation of SMART-MR1/DCF with JYM20 for 72 hours at room temperature, we detected an increase in the FP of JYM20 as it transitions into a high MW protein complex. Addition of a competitive ligand (Ac-6-FP) to the sample caused a significant reduction of FP in a dose dependent manner.

Next, we tested binding of SMART-MR1 to an established cognate TCR using isothermal titration calorimetry (ITC) (Fig. 3C). A-F7 is a MAIT cell-derived TCR that specifically binds to MR1 molecules presenting 5-OP-RU. A-F7 binding to SMART-MR1/5-OP-RU was exothermic, driven by a high enthalpy change of -13.5 kcal/mol, and showed a dissociation constant (K_D) of 356 nM, as measured by ITC. This result could suggest an affinity enhancement for SMART-MR1 relative to values reported previously for A-F7 binding to native MR1 (K_D of 1.1 uM), albeit these experiments were performed using Surface Plasmon Resonance (29). These results demonstrate that SMART-MR1 provides a system for high-resolution solution biophysics by NMR, high-throughput ligand screening, and TCR binding studies.

Cryo-EM structure of the SMART-MR1/A-F7 TCR complex reveals a native antigen recognition mechanism

To determine whether SMART-MR1 presents ligands and binds to TCRs in a native-like manner, SEC-purified SMART-MR1/5-OP-RU/A-F7 TCR complex was used to prepare grids for cryogenic electron microscopy (cryo-EM). To overcome the potential preferred orientation issue of the complex during vitrification, 0.01% fluorinated octyl-maltoside was added to the sample prior to grid preparation. Micrographs from cryo-EM data collection revealed an even distribution of particles and 2D classes showed different orientations of the complex. The data processing resulted in a nearly complete cryo-EM map (3.08 Å) of the complex with well-defined densities that allowed unambiguous register of the protein sequence and 5-OP-RU ligand (SI Appendix, Fig. S3 and Table S1).

The cryo-EM structure of SMART-MR1 displayed remarkable agreement with the predicted AlphaFold3 model and the native MR1 crystal structure (Fig. 4A). W17 of SMART-MR1 adopts a similar conformation to W60 of β₂m, supporting our prediction that a Trp at this position is key for stabilizing the underside of the MR1 α₁/α₂ domain. We observed a continuous density connecting K116 of SMART-MR1 (analogous to K43 in native MR1) to 5-OP-RU, confirming that SMART-MR1 covalently captures ligands through the formation of a Schiff base (SI Appendix, Fig. S4A). The conformation of 5-OP-RU and its interactions within the binding groove remain conserved across the SMART-MR1 and native MR1 structures (Fig. 4B and C). The uracil ring of 5-OP-RU is burrowed within a hydrophobic pocket formed by Y80/W142 and is further stabilized by hydrogen bonds to R82 and S97. The hydroxyl groups of the ribityl chain form a network of polar interactions with R82, R167, Y225, Q226, and W229.

Figure 4.

We also observed that the mechanism of 5-OP-RU recognition by A-F7 is nearly identical across the SMART and native MR1 structures. Y95 of the A-F7 complementarity determining region (CDR) 3α reaches into the MR1 binding groove and forms polar contacts with Y225 and W229 in addition to the ribityl group of 5-OP-RU (Fig. 4D). Furthermore, all six CDR loops of A-F7 dock onto the SMART-MR1 binding groove similarly to native MR1 (Fig. 4E). CDR loops 1α and 2α form contacts with the α₁helix of SMART-MR1, whereas CDR loops 1β and 2β interact with the α₂ helix. CDR loops 3α and 3β are positioned centrally above the MR1 binding groove, while only CDR3α is in a favorable position to extend into the A’ pocket to make direct contact with the ligand. Together, these structural results demonstrate that SMART-MR1 presents ligands and binds to TCRs in a manner nearly indistinguishable from native MR1.

Discussion

Biochemical characterization of native MR1 presents challenges due to the fact that its stability is coupled to the presence of a ligand within the binding groove, in addition to association with β₂m. Here, we demonstrate that the α₁/α₂ metabolite-presenting platform of MR1 can be stabilized through interactions with a synthetic protein domain, mitigating any concerns caused by β₂m dissociation while maintaining binding to known ligands. Our biophysical data confirm that SMART-MR1 is stable and properly conformed, and that SMART-MR1 binds to cognate TCRs with high affinity. Our structural results demonstrate SMART-MR1 presents ligands in a native-like manner, and that TCRs bind to SMART and native MR1 through identical interactions.

Identifying new ligands and TCRs is critical for understanding MR1 biology in the context of cancer and infectious disease and for developing immunotherapies (22, 29-35). SMART-MR1 provides a platform for rapid exploration of ligand and TCR interactions by various orthogonal techniques. Screening ligands is possible by using SMART-MR1 for high-throughput competitive FP assays. The reduced size and high expression yield of SMART-MR1 allows for mapping of interactions with TCRs by NMR, as illustrated by our recent work using SMART MHC (36). As shown by our current structural determination of the complex with A-F7, SMART MR1 provides an attractive system for high-resolution structural studies of metabolite recognition by MAIT cell receptors using cryo-electron microscopy. Finally, while our work focuses on the MR1*01 allele which has a frequency 71% in humans (37), expanding SMART-MR1 to include less frequent alleles can help to determine the effect of polymorphisms on ligand binding and TCR recognition. In summary, SMART-MR1 offers a streamlined and adaptable system that maintains the functional features of MR1, while enabling a wide range of biochemical, structural, and screening approaches for characterizing emerging MR1-restricted antigens relevant to MAIT cell biology, and human disease.

Materials and methods

MR1 ligands

MR1 ligands were purchased from the following suppliers: Ac-6-FP (Cayman Chemical no. 23303), 6-FP (Cayman Chemical no. 14247), 3FSA (3-formylsalicylic acid) (Acros Organics no. 377070250), 5FSA (5-formylsalicylic acid) (Millipore Sigma no. F17601), DCF (Millipore Sigma no. D6899). 5-OP-RU was synthesized in-house from 5-N-RU, which was generously provided by E. Adams (University of Chicago). Chemical structures were prepared using ChemSketch (ACD/Labs).

Protein expression, refolding, and purification

Codon optimized DNA encoding SMART-MR1, hpMR1, or bovine β2m was transformed into BL21(DE3) E. coli (New England Biolabs). Proteins were expressed in autoinducing media at 37 °C overnight (38). Inclusion bodies were isolated and solubilized using guanidine hydrochloride (39). For in vitro refolding of SMART-MR1, 60 mg of solubilized inclusion body was slowly diluted into 300 mL of refolding buffer (0.4 M arginine–HCl, 2 mM EDTA, 4.9 mM reduced L-glutathione, 0.57 mM oxidized L-glutathione, 100 mM Tris pH 8.0) at 4 °C while stirring in the absence (empty) or presence of 2 mg of ligand. For hpMR1 refolding, a 50 mg mixture of hpMR1:bβ₂m at a 1:1 molar ratio was slowly diluted into 1 L of refolding buffer containing 2 mg ligand. Refoldings proceeded at 4 °C for four days without stirring. Solutions were then dialyzed into buffer containing 150 mM NaCl, 25 mM Tris pH 8.0 overnight. The refolding mixtures were concentrated using Pellicon Single-Pass Tangential Flow Filtration (Millipore Sigma) and Amicon ultra centrifugal filters (Millipore Sigma). Purification of refolded MR1 proteins was performed by size-exclusion chromatography using a HiLoad 16/600 Superdex 200 pg column at 1 mL/min in running buffer (150 mM NaCl, 25 mM Tris pH 8.0).

Differential scanning fluorimetry

DSF was used to assess the thermal stabilities of refolded MR1 molecules. 10 μM protein was mixed with 10× SYPRO Orange dye (Invitrogen) in a buffer of 150 mM NaCl and 20 mM sodium phosphate pH 7.4. 20 μL samples were loaded into MicroAmp Optical 384-well plate and the experiment was performed on a QuantStudio 5 real-time PCR machine with excitation and emission wavelengths set to 470 nm and 569 nm. The temperature was incrementally increased at a rate of 1 °C per minute between 25 and 95 °C. Data analysis and fitting were performed in GraphPad Prism v9.

Nuclear magnetic resonance

To generate isotopically-labeled SMART-MR1, proteins were expressed in M9 media containing ²H₂O, 3 g/L [²H, ¹³C] glucose, and 1 g/L of ¹⁵NH₄Cl. Protein expression was induced with 1 mM IPTG at OD₆₀₀=0.6 followed by shaking at 37 °C for 5 hours. Inclusion bodies and refolding mixtures were prepared as described above. SEC purification of triple-labeled protein was performed in NMR buffer (100 mM NaCl, 20 mM sodium phosphate, pH 7.2). NMR samples were prepared with 650 μM protein and 5% ²H₂O. ¹H-¹⁵N TROSY HSQC spectra were collected at 800 MHz ¹H magnetic field. The data were processed in NMRPipe (40) and analyzed in POKY (41).

Fluorescence polarization

The association of a TAMRA-labeled, fluorescent 5-OP-RU analog (JYM20) (4) to SMART-MR1/DCF was monitored using fluorescence polarization (FP). 10 nM JYM20 was incubated with 10 μM SMART-MR1/DCF in the absence or presence of increasing amounts of Ac-6-FP for 72 hours at room temperature in the dark. To determine the optimal incubation conditions for ligand binding, several association assays were performed at room temperature and in 4° C across various incubation times. We found that 48 hours was not sufficiently long enough to facilitate complete association and that at 96 hours a degree of protein degradation had occurred. Additionally, experiments performed at 4° C did not reach full equilibrium, even after 72 hours. Reactions performed at room temperature for 72 hours were able to both reach equilibrium and avoid degradation, a finding consistent across the literature (42). Excitation and emission values used to detect the fluorescence of JYM20 were 531 and 595 nm on a SpectraMax iD5 plate reader. Raw parallel (I_II) and perpendicular emission intensities (I_⊥) were collected and converted to polarization (mP) values using the equation 1000*[(I_II-(G*I_⊥))/(I_II+(G*I_⊥))] with a G-factor of 0.33 for JYM20. The data was analyzed using GraphPad Prism v9.

Isothermal titration calorimetry

ITC was performed using a MicroCal VP-ITC system (Malvern Panalytical, Westborough, MA). All proteins were dialyzed into buffer containing 150 mM NaCl, 25 mM Tris pH 8.0. Syringe containing 200 μM A-F7 was titrated into a calorimetry cell containing 20 μM SMART-MR1/5-OP-RU at 25°C. Following an initial 5 μL injection, injection volumes were 10 μL for a duration of 20 s and spaced 240 s apart to allow for a complete return to baseline. Data were processed and analyzed with Origin software. Isotherms were fit using a one-site ITC binding model with the first data point excluded from analysis. ITC was performed with one technical replicate.

Cryo-EM sample preparation

The A-F7 TCR/SMART MR1-5-OP-RU complex was purified using SEC and best peak fraction (0.3 mg/mL) was used for grid preparation. Quantifoil R 1/2 300 Mesh, Cu grids were glow discharged at 15 mA for 60s. Prior to the grid preparation, 0.01% fluorinated octyl maltoside was added to the sample to mitigate any preferred-orientation problems. A sample volume of 3 μL was applied to the grids and subjected to vitrification using a Vitrobot Mark IV system (ThermoFisher) at 4°C with 100 % humidity. The sample was blotted for 3.0 s with a blot force of 2 using Standard Vitrobot 595 filter paper (Ted Pella, Inc) and the grids were plunge-frozen into liquid-nitrogen-cooled liquid ethane.

Cryo-EM data collection, processing, and model building

Cryo grids of the complexes were imaged at 165,000× nominal magnification using a Falcon 4 detector (ThermoFisher) on a Glacios 2 (ThermoFisher) microscope operating at 200 kV with a calculated pixel size of 0.6975 Å. Automated image collection was performed using EPU with a nominal defocus range of –0.8 to –2.0 μm.

For A-F7 TCR/SMART MR1-5-OP-RU complex, 5,448 micrographs were collected using EPU. Data processing was carried out in CryoSPARC v4.7.1. Micrographs were aligned using Patch-motion correction and the Contrast Transfer Function (CTF)-corrected using Patch CTF Estimation in CryoSPARC. Micrographs with CTF fit worse than 10 Å were discarded. Blob picker was used to pick 2,299,717 particles from entire dataset with a minimum particle diameter of 90 Å to a maximum of 180 Å. Particles picked by blob picker were inspected and extracted with a box size of 384 pixels to yield 1,673,666 particles. A round of 2D classification was performed to select best classes (408,724 particles) that look like the desired complex. Ab-initio reconstruction was performed using these particles to get two volumes. One volume containing 248,649 particles was refined to 3.16 Å resolution map. However, this map looked distorted upon rotating by 90° and the density was fragmented in multiple regions. The other volume containing 287,776 particles yielded a nearly complete and less distorted map with 3.08 Å overall resolution, thus used for model building.

For the densities corresponding TCR and MR1 heavy chain, previously solved structure of human MAIT A-F7 TCR in complex with human MR1-5-OP-RU (PDB ID: 6PUC) without α3 domain and β₂m was used for docking. For the density corresponding to the stabilization domain, bundle consisting of three N-terminal helices of miniaturized HLA-A*02 (PDB ID: 9NDS) was used for docking. Both models were docked in the cryo-EM density map using UCSF Chimera. The model was manually adjusted in Coot and refined using real-space refinement in PHENIX package.

Data availability

Atomic coordinates for the SMART-MR1/A-F7 TCR complex have been deposited in the PDB under the accession code 10VM.

Supporting information

Figure S1.

Figure S2.

Figure S3.

Figure S4.

Supplementary Table S1.

Acknowledgements

This work was delivered as part of the MATCHMAKERS Team supported by the Cancer Grand Challenges partnership funded by Cancer Research UK (CGCATF-2023/100004) and the National Cancer Institute (OT2CA297575) and The Mark Foundation for Cancer Research. We acknowledge support by grant NIGMS (7R35GM125034) to N.G.S.

Additional information

Funding

Cancer Research UK (CRUK) (CGCATF-2023/100004)

• Nikolaos Sgourakis

HHS | NIH | National Cancer Institute (NCI) (OT2CA297575)

• Nikolaos Sgourakis

HHS | NIH | National Institute of General Medical Sciences (NIGMS) (7R35GM125034)

• Nikolaos Sgourakis