Multiple Isomers of Photolumazine V Bind MR1 and Differentially Activate MAIT Cells

Key Points

• A new bacterially produced MR1 ligand has been identified in four isomeric conformations.

• Isomers of a newly identified MR1 ligand differentially activate MAIT cells.

Abstract

In response to microbial infection, the nonclassical Ag-presenting molecule MHC class I–related protein 1 (MR1) presents secondary microbial metabolites to mucosal-associated invariant T (MAIT) cells. In this study, we further characterize the repertoire of ligands captured by MR1 produced in Hi5 (Trichoplusia ni) cells from Mycobacterium smegmatis via mass spectrometry. We describe the (to our knowledge) novel MR1 ligand photolumazine (PL)V, a hydroxyindolyl-ribityllumazine with four isomers differing in the positioning of a hydroxyl group. We show that all four isomers are produced by M. smegmatis in culture and that at least three can induce MR1 surface translocation. Furthermore, human MAIT cell clones expressing distinct TCR β-chains differentially responded to the PLV isomers, demonstrating that the subtle positioning of a single hydroxyl group modulates TCR recognition. This study emphasizes structural microheterogeneity within the MR1 Ag repertoire and the remarkable selectivity of MAIT cell TCRs.

Introduction

As a cornerstone of adaptive immunity, MHC molecules sample a variety of intracellular ligands for inspection by T lymphocytes at the plasma membrane. Whereas conventional MHC class I peptide ligands have been studied for decades, the discovery of unconventional, monomorphic MHC-related molecules has recently revealed that T cells can also detect immune and cellular dysregulation through the presentation of nonproteinaceous biomolecules (1–4). The unconventional MHC class I–related protein 1 (MR1) presents ligands derived from microbial riboflavin synthesis and related small molecules to MR1-restricted T cells, of which mucosal-associated invariant T (MAIT) cells are the most abundant and best characterized subset (4–7).

MAIT cells are an abundant population of innate-like T cells that have primarily been characterized in their protective qualities in microbial infection. In the setting of infectious disease, predominantly tissue-resident MAIT cells recognize MR1/metabolite complexes and activate in a cytotoxic or Th1-like cytokine response (8–10), which contributes to mitigating infections. These data suggest that MAIT cells inhibit initial bacterial colonization, lending credence to the concept that MAIT cells bridge the innate and adaptive compartments to help combat a broad variety of bacterial infections. Indeed, in Gram-negative and mycobacterial infection mouse models, activation of MAIT cells is positively correlated with survival and convalescence (5, 11–13). In more recent years, MR1/ligand complexes have been implicated in additional, nonmicrobial contexts. For example, MR1/ligand complexes have also been found to activate protective T cells during cancerous transformation, where MAIT cells were found to directly kill tumor cell lines in vitro and leukemia in vivo (14). Furthermore, MAIT cells have been implicated as contributors to autoimmune reactions, although the importance of MR1-presented small-molecule ligands to autoimmunity has yet to be elucidated (15, 16). There also seems to be evidence for a MAIT cell TCR-dependent pathway of tissue repair and coordination of the immune response (17–19). Thus, the MR1/MAIT cell axis is increasingly considered as a complex sensor of metabolic dysregulation rather than a sensor of only microbial infections.

Although early work clearly showed that MR1 presents microbial Ags, the chemical nature of those Ags was not discovered until 2012 (4). Because MR1 ligands are small-molecule metabolites, their characterization was initially hampered by technical limitations such as stability in solution and the requirement of complex of chromatographic and mass spectrometry (MS) techniques. Nevertheless, concerted efforts from several laboratories have since led to the characterization of several MAIT cell Ags, primarily by integrating MHC immunobiology, synthetic and natural products chemistry, new approaches in MS, and in silico screens (7, 20–22). Initial explorations into the nature of the MR1 ligand repertoire first identified the ligand 6-formyl pterin (6-FP) (4), which induces strong surface expression of MR1 yet is not known to activate MAIT cells. However, the ribitylpyrimidine and ribityllumazine classes of MR1 ligands, which include molecules such as 6-hydroxymethyl-8-d-ribityllumazine (rRL-6-CH₂OH) and pyrimidine-5-(2-oxopropylideneamino)-6-d-ribitylaminouracil (5-OP-RU) (23, 24), respectively, were soon discovered. These Ag classes, which are thought to arise from the riboflavin biosynthetic pathway, are both capable of stimulating MAIT cells, owing in part to their ribityl moiety. Indeed, in structural studies, this chemical feature has been shown to protrude from the MR1 ligand binding groove and directly contact the CDR3α of the MAIT cell TCR (23–25). After these hallmark papers, the identification of microbial MR1 ligands had slowed due to the intrinsic challenges of discovering novel molecules. As indicated by in silico methods, the binding groove of MR1 is hypothetically capable of binding many small molecules (20, 21); thus, microbial pathogens might provide a series of undiscovered yet related metabolites that could additionally be presented by MR1. We have seen this with the discovery of photolumazines (PLs) (26) as MR1 ligands, a series of ribityllumazine small molecules produced by several bacterial species that activate MAIT cells (22).

To address the challenge of novel ligand discovery, we have recently used a bait-and-hook system to capture microbial Ags within recombinant MR1. In our previous work using this method (22), we identified and characterized two ribityllumazine compound MR1 ligands, PLI and PLIII, for which the names derive from the convention used in the discovery of similar molecules (26, 27). In this study, we expand on this approach to describe PLV, a hydroxyindolyl-ribityllumazine, the presence of each of its isomers in mycobacterial supernatants, and the differential reactivity of these isomers both in the context of presentation in MR1 and recognition by MAIT cells. Our findings confirm that microorganisms can produce a diversity of antigenic structures based on a common core (ribityllumazine) and demonstrate that microheterogeneity, that is, small structural shifts in similar isomers, can result in profound effects on antigenic presentation and T cell recognition. These findings imply that differences in pathogen metabolism, such as those that occur in response to treatment, may in turn influence the immune system’s control of infection.

Materials and Methods

Cell culture and supernatant purification

Hi-5 cells were grown in Insect-XPRESS serum-free media (Lonza) supplemented with 2 mM l-glutamine (Gemini Bio). Mycobacterium smegmatis isolate mc²155 (American Type Culture Collection) or the RibA overexpresser (provided by Dr. Ruchi Jain Dey) were grown in 7H9 medium supplemented with 0.5% glycerol, 0.05% Tween 80, and 10% ADC (albumin-dextrose-catalase) supplement. For Florisil enrichment of bacterial supernatants, the RibA-overexpressing line was grown to stationary phase, then allowed to continue shaking for 24 h. The supernatant was cleared of cells, then filtered using a 3-kDa molecular mass cutoff Amicon centrifugal concentrator. The concentrator flowthrough was applied to a gravity-packed Florisil column pre-equilibrated in HPLC-grade water, then the column was washed with two column volumes of HPLC-grade water. The column was developed with 10% acetone in HPLC-grade water, dried under vacuum using a CentriVap concentrator (Labconco, Kansas City, MO), and resuspended in water for injection onto an AB Sciex 5600 mass spectrometer.

Recombinant human platform MR1 production and purification

Methods describing the production of recombinant human platform MR1 (hpMR1) loaded with heterogeneous ligands from bacteria are described in detail in Harriff et al. (22) and Ladd et al. (28). Briefly, we expressed a recombinant, secreted, single-chain bovine/human MR1 chimera (hpMR1) composed of the bovine α3 and β₂-microglobulin domains and the human α1 and α2 domains. At the time of baculoviral induction of the Hi-5 cells (induced at 2 × 10⁶ cells/ml), M. smegmatis was added to the Hi-5 culture at a density of 4 × 10⁴ cells/ml, as determined using the assumption that OD₆₀₀ of 1.0 is equivalent to 1.5 × 10⁸ M. smegmatis cells/ml. After 72 h, the supernatant was cleared by centrifugation, filtered, and buffer exchanged into HEPES-buffered saline, after which the protein was purified by affinity chromatography with Ni-NTA. The 12× His tag was cleaved by recombinant (in-house) 3C protease, and then the protein was purified by size exclusion chromatography and concentrated to 2 μg/μl.

Nano liquid chromatography–tandem MS

Samples containing hpMR1 were injected with no fractionation or separation using an Eksigent NanoLC (nanoscale liquid chromatography) 415 reversed-phase HPLC. A trap-elution method was used with a 5-mm-long, 350-μm-internal diameter ChromXP C18 trap column by Optimize Technologies with 3-μm particles and 120-Å pores. After desalting, samples were then loaded onto a 150-mm-long, 75-μm-internal diameter ChromXP Eksigent C18 reversed-phase separation column with 3-μm particles and 120-Å pores. An acetonitrile gradient was run at pH 3.0 using two solvents. Solvent A was 0.1% formic acid in HPLC-grade water, and solvent B was 0.1% formic acid in 95% acetonitrile in HPLC-grade water. After column equilibration with 2% solvent B, samples were loaded at 5 μl over 10 min onto the trap column and injected through the separation column at 300 nl/min with a gradient of 10–40% B for 70 min, followed by a clearance gradient of 40–80% B for 7 min.

Sample passing through the separation column was ionized using the PicoView ion source by New Objective, through an 11-cm, 20-μm-inner diameter pulled glass emitter that tapered to 10 μm at the tip. An ionized sample was injected into an AB Sciex TripleTOF 5600 quadrupole time of flight (TOF) mass spectrometer with source voltage set at −2600 V in the negative ion mode. Data-dependent acquisition was the method used for data collection.

For PLV stability area under the curve (AUC) acquisition, PLV-7 was resuspended at 0.0225 mM and split into three aliquots, with each aliquot placed in one of three temperatures. Then, 25 μl was removed from each aliquot and direct injected into an AB Sciex TripleTOF 5600 quadrupole TOF mass spectrometer using the PicoView ion source with a voltage set at −2600 V in the negative ion mode. Data were acquired in 3-min acquisitions with a mass range of 30–800 Da.

Multiple reaction monitoring high-resolution MS

Five microliters of the Florisil-enriched RibA overexpresser supernatant, 5 µl of 1.0 × 10⁻⁷ mol of individual PLV synthetics, and a 1:1:1:1 ratio of the same molarity for the combination of synthetics were subjected to high-resolution multiple reaction monitoring (MRM-HR) MS. Samples were injected using the previously detailed nanoLC and trap column. In addition to the ChromXP C18 trap column by Optimize Technologies, a Waters ACQUITY M-class HSS T3 75-µm × 150-mm column with 1.8-µm particles was used for nonproteinaceous samples. A TOF-MS method of 24 min was used with the same parameters as the nanoLC–tandem MS (MS/MS) injection method, save that samples were loaded onto the trap column over 4 min, and the gradient of 10–40% solvent B lasted from 2 min until 16 min, followed by the clearance gradient of 40–80% solvent B until 20 min. The following precursor ions were examined with a total TOF-MS cycle time of 1.8005 s, a total accumulation time of 0.249966 s, and a minimum and maximum TOF mass range of 200–500 Da: 241.07, 275.10, 325.12, 327.09, 329.11, 362.10, 375.13, 385.10, 405.11, 428.12, 444.12, 460.11, 453.12, 488.10, and 485.12. The product ion scan accumulation times were 0.100037 s each, with a TOF mass range of 30–500 Da.

Data analysis

For MRM-MS, raw .wiff files were imported into Skyline (29) and examined searching for between 4 and 10 transitions for each target precursor ion, found in Supplemental Fig. 2. Ions were then manually curated by making sure transitions corresponded to precursor elution times.

For nanoLC-MS/MS, extracted ion chromatograms (XICs) were drawn using PeakView (v1.2) on the TOF-MS experiments, with a center using the precursor ion mass previously identified, and the width set at 0.05. XICs were exported and plotted in R using ggplot2 (30) after smoothing using the DIAlignR (31) package (Gaussian smoothing, sampling time of 0.5 s for the intact protein samples and 0.05 s for the MRM). AUCs were determined from the point of peak start until peak tailing, regardless of acquisition time.

An AUC was calculated for direct injections by selecting a 1-min window from 1.5 to 2.5 min and calculating the sum intensity of the 444.12 parent ion.

Statistical analysis of ion abundance using MarkerView

Raw .wiff files were converted to .mzxml files using msconvert from the ProteoWizard suite, with 64 bit compression. A minimum retention time of 5.00 min and maximum retention time of 45.00 min was used to initially pare down ions, and then a minimum of 10.00 min and maximum of 30.00 min were used when most ions fell within this range. No subtraction offset was used, and a minimum spectral peak width of 0.05 Da was used, as well as a minimum retention time peak width of three scans and a noise threshold of 1. Charge states were assigned, and the retention time tolerance was 6.00 min with a mass tolerance of 0.05 Da due to the nature of these intact protein injections. A maximum number of peaks was set to 5000 with no isotope filtering. Three biological replicates were examined with three technical replicates within each biological replicate, and output data were examined in GraphPad Prism version 9.3.1.

Global Natural Products Social Molecular Networking

The .mzxml files used for MarkerView were also uploaded to the Global Natural Products Social Molecular Networking server using an FTP client, then .mzxml files were subjected to the METABOLOMICS-SNETS (version 1.2.5) workflow. Files were loaded into their respective groups, with precursor ion mass tolerance set to 0.02 Da, and fragment ion mass tolerance set to 0.02 Da. The Min Pairs Cos setting was set to 0.7, the Network TopK setting was set to 10, maximum connected component size to 100, minimum matched fragment ions to 3, and minimum cluster size to 2, with MSCluster selected to run. Library search options included setting the library search minimum matched peaks to 3, with analog searching enabled with a score threshold set to 0.6 and maximum analog search mass difference at 50. All remaining settings in the workflow were set at default. The networks were visualized using Cytoscape 3.9.0 (32).

Synthesis of PLV isomers

Briefly, methyl 2-(4-hydroxy-1H-indol-3-yl)-2-oxoacetate, methyl 2-(5-hydroxy-1H-indol-3-yl)-2-oxoacetate, methyl 2-(6-hydroxy-1H-indol-3-yl)-2-oxoacetate, or methyl 2-(7-hydroxy-1H-indol-3-yl)-2-oxoacetate (39 mg, 0.18 mmol) were dissolved in 1 ml of DMF. 5-A-RU HCl (14 mg, 0.045 mmol) was dissolved in 1 ml of H₂O. The solutions were combined (Supplemental Fig. 2A), after which the reaction was heated for 20 min at 120°C with microwave irradiation. The reaction was filtered, then purified by reversed-phase HPLC (Waters Atlantis T3 30- × 150-mm column, water/acetonitrile, 20 ml/min flow rate, 0.9 ml injection volume, 20 min 10–40% B, 5 min 40% B). The fractions containing product were combined and lyophilized. All products were observed by MS at the expected m/z of 444 (M-1).

Flow cytometry

Using an MR1 overexpressing cell line, BEAS-2B:MR1-GFP, cells (33) were grown in a six-well tissue culture plate to ∼70% confluency before being incubated by either the vehicle control, 6-FP, or the PLV synthetic isomers at specified concentrations for 16 h. Cells were collected on ice and stained with the anti-MR1 26.5 Ab conjugated to allophycocyanin (BioLegend) for 40 min in 2% human serum, 2% goat serum, and 0.5% FBS. T cell clones (1e5) were stained with anti-CD3, anti-CD4, anti-CD8, or anti-TCRαβ Abs for 30 min in 2% FBS in PBS buffer. Cells were washed, fixed, and analyzed with a BD FACSymphony flow cytometer and FACSDiva software (BD Biosciences). All analyses were completed with FlowJo (Tree Star) and GraphPad Prism version 9.3.1.

ELISPOT assays

Dendritic cells (DCs) were plated in equivalent numbers indicated in the figure legends and used as APCs in an ELISPOT assay with IFN-γ production by noted MAIT cell clones or magnetic bead–purified (Miltenyi Biotec) ex vivo CD8⁺ T cells as the readout as previously described (12). Synthetic PHA, M. smegmatis supernatant, M. smegmatis, and deazalumazine (DZ) (34) PLI or PLV isomers were added to the cells at indicated concentrations and incubated for 1 h prior to the addition of specified MAIT cell clones or CD8⁺ T cells. ELISPOT plates were then incubated for 18 h prior to development. Blocking was performed using the unconjugated anti-MR1 26.5 Ab at 5 µg/ml and an IgG2a isotype control, added at 5 µg/ml for 1 h prior to the addition of synthetic ligand.

Results

MS analysis of ligands eluted from hpMR1 uncovers novel ligands

To capture ligands produced from M. smegmatis during coculture with our eukaryotic expression system, we expressed a soluble, single-chain, bovine–human chimeric MR1 construct (hpMR1) using a baculoviral expression system as described previously (22). The purified, M. smegmatis–loaded MR1 produced in this system (hpMR1^{+M. smegmatis}) from three biological replicates potently activated MAIT cells when compared with MR1 expressed in sterile conditions (hpMR1^−Bact) (22). This production method ensures that MAIT cell–activating ligands generated by M. smegmatis are bound within the purified MR1 prior to mass spectrometric examination of the ligands themselves.

To determine the ligands bound by hpMR1^{+M. smegmatis}, we compared the ligands associated with hpMR1^{+M. smegmatis} to those associated with hpMR1^−Bact by MS. Following negative ion mode injection of the MR1/ligand complexes onto a nanoLC electrospray ionization TOF mass spectrometer, individual ion intensities from hpMR1^{+M. smegmatis} were compared with ions observed in hpMR1^−Bact using a t test. To eliminate low abundance species, only the 5000 highest intensity ions were compared. Ions found to have a −log(p) value of >1.3 and a log₂ fold change >1 were examined manually. Manual curation involved determining whether 1) the ion had clear, distinct precursor ion peaks via an XIC, 2) was truly found in hpMR1^{+M. smegmatis} samples and not in the hpMR1^−Bact, 3) was present in at least two biological replicates and all technical replicates of a respective sample, and 4) had clear collision-induced disassociation (CID) fragment spectra. Using these parameters, six ions were found to be significantly enriched in hpMR1^{+M. smegmatis} samples (Fig. 1A): m/z of 385.116 and 428.121, previously identified as PLI and PLIII, respectively (22), and four others, one of which had an m/z of 444.116. Each of these ions exhibited a neutral loss of 134.05 Da as a common feature, consistent with the loss of a ribityl group. The prospect of these putative ligands containing a ribityl chain, and thus having activating potential, caused us to investigate these data with additional methods.

FIGURE 1.

MS analysis of intact hpMR1 cocultured with M. smegmatis reveals a diverse array of ions. (A) Volcano plot showing significance and log-fold change between hpMR1^{+M. smegmatis} and the control (hpMR1^−Bact). Dots above dashed line indicate ions with a p value <0.05. After manual curation, eliminated dots are shown in black and the prioritized ions are highlighted in red. Previously characterized ions, m/z 385.1 and 428.1, are annotated with the chemical structure (22). (B) The network of ions containing that of riboflavin (375.1), with each node labeled with the m/z of the ion. The number of scans of ions found in hpMR1^{+M. smegmatis} samples is represented by the color of the node on a graded scale, where low-intensity ions are light red and the high-intensity ions are dark red. White indicates the ion was found in the hpMR1^−Bact negative control.

In a workflow orthogonal to the t test comparative analysis, we examined the samples using the Global Natural Products Social Molecular Networking server (35). We compared precursor ion and CID fragment spectra to a number of different public spectral libraries to determine either the identity of putative ligands, or, if not previously identified, to determine the relatedness of these compounds to other known small molecules. Given that many of the known ligands for MR1 are derivatives of riboflavin biosynthesis, we initially focused our attention on the riboflavin ([M-H]⁻ m/z of 375.131) molecular network. Following elimination of ions found in the negative controls and focusing on ions scanned three or more times, 29 ions from the hpMR1^{+M. smegmatis} samples were indicated in the riboflavin network as enriched (Fig. 1B). These ions were color-coded based on the number of scans, with light red indicating low intensity and dark red indicating a higher intensity. After manual examination of these data, three ions remained enriched to the hpMR1^{+M. smegmatis} samples: m/z 428.121, previously identified as PLIII (22), m/z 453.116, and m/z 444.116. Of note, the ion m/z 444.116 was also identified in the volcano plot as being unique to hpMR1^{+M. smegmatis} samples. Given that this ion was found in both workflows and seemed to be related to PLIII, we named this ligand PLV and prioritized it for structural elucidation and characterization.

PLV exists as multiple isomers

To determine the structure of this putative ligand, we examined the MS data for PLV in detail, which showed a clear m/z 444.116 XIC peak with a retention time of ∼28 min in hpMR1^{+M. smegmatis} samples with no peak present in hpMR1^−Bact controls (Fig. 2A). Unexpectedly, the m/z 444.116 LC peak “tailed” asymmetrically, with several smaller peaks appearing between 28 and 34 min. However, it was difficult to assess the meaning of this tailing with an XIC analysis. From the accurate mass and isotopic distribution, we calculated the chemical formula of putative PLV as C₁₉H₁₉N₅O₈, which matched the [M-H]⁻ precursor ion peak within 1.8 ppm. Examination of the CID fragment spectrum revealed a distinctive neutral loss of 134.05 Da from fragment ion 444.128 to fragment ion 310.072, indicating the potential neutral loss of a ribityl chain (Fig. 2B). Further examination of the fragment spectrum allowed for several possible PLV structures to be proposed (Fig. 2C). Given the proximity of m/z 444.116 to PLIII in the molecular network, we found it most likely that PLV is also a 6-indolyl-7-oxo-ribitiyllumazine. Differing from PLIII by a single oxygen, PLV is likely a hydroxylated derivative of PLIII, with the additional oxygen likely installed on the benzene ring of the indole moiety. However, because the CID fragment spectrum could not resolve the exact position of the hydroxyl group, we chemically synthesized each of the four PLV isomers (Supplemental Fig. 2) to determine which species was captured in our coculture system with M. smegmatis.

FIGURE 2.

PLV is a hpMR1 ligand. (A) XIC for m/z 444.1 drawn for the injections of intact hpMR1 produced in the absence (black) or presence (red) of M. smegmatis. (B) MS/MS fragment spectrum for m/z 444.1 from the injections of intact hpMR1^{+M. smegmatis}. High-intensity fragments are annotated with the structure they represent (red, bold), projected onto the putative structure of PLV (gray). (C) Structure of PLV. The four positions at which the additional hydroxyl of PLV could be attached are labeled as such (4, 5, 6, and 7) and colored individually.

Multiple isomers of PLV bind hpMR1 and are produced by M. smegmatis

To validate that the fragmentation pattern of our synthetic reference isomers matched that of PLV eluted from hpMR1^{+M. smegmatis}, we subjected each to mass spectrometric analysis. All four isomers exhibited an identical CID fragment spectrum pattern as the PLV eluted from hpMR1^{+M. smegmatis} (Fig. 3A). Because this was insufficient to clarify the positions of the hydroxyl on the indole moiety, we then determined the relationship among the retention times of the synthetic standards using nanoLC and the MRM-HR method, such that we could compare this pattern to that seen for PLV eluted from hpMR1 (Fig. 3B, Supplemental Fig. 2). PLV-4 eluted the latest of the four isomers, PLV-7 eluted earliest, and in between eluted PLV-5 and PLV-6 with similar retention times. Indeed, the elution pattern of the four combined synthetics showed three distinct retention times, with a broad peak that encompassed PLV-5 and PLV-6 (Fig. 3C). This demonstrated that a nanoLC MRM-HR method can discriminate among three of the four closely related PLV isomers.

FIGURE 3.

Multiple isomers of PLV bind hpMR1 and are produced by M. smegmatis. (A) Mirror plot comparing the MS/MS fragmentation spectra for m/z 444.1 between synthetic PLV-7 and hpMR1^{+M. smegmatis} (red). All four synthetic isomers had identical fragment spectra (data not shown). (B) XICs for m/z 444.1 from each of the synthetic PLV isomers generated using the MRM method. The color of the line corresponds with the color given in Fig. 2C. (C and D) XICs of m/z 444.1 for (C) all four synthetic PLV isomers combined and (D) Florisil-enriched supernatant of the M. smegmatis RibA2 overexpresser.

We then used this MRM-HR method to determine which isomers are biosynthesized by bacteria; specifically, we assessed the supernatant of a strain of M. smegmatis overexpressing RibA2 (GTP cyclohydrolase II) (36) for the presence of the four PLV isomers. The supernatant was first fractionated by a Florisil column to purify riboflavin-related metabolites (37). The MRM-HR chromatogram of the Florisil eluate demonstrates that M. smegmatis produces at least three of the PLV isomers (Fig. 3D), matching that which we observed for the four combined synthetics (Fig. 3C). Because this established that M. smegmatis produces multiple isomers of PLV, we then examined XICs of m/z 444.116 from hpMR1^{+M. smegmatis} to determine which isoforms hpMR1 captures. The three peaks in the “tailing” XIC of m/z 444.116 (Fig. 2A) are thus distinguished, matching the pattern of both the combined synthetic isomers of PLV and the enriched M. smegmatis supernatant (Fig. 3C, 3D). Moreover, using a modified version of MR1 ligand stability testing protocols in an aqueous solution (24, 38), a PLV-7 time-course at 4, 21, and 37°C indicates that PLV is stable as compared with 5-OP-RU (Supplemental Fig. 3A) and that PLV remains soluble over these time courses (data not shown). Taken together, these data indicate that multiple isomers of PLV are produced by M. smegmatis and loaded into the groove of hpMR1 in a coculture system.

Functional testing of PLV reveals an MR1 ligand with differential MAIT cell activating potential

Having demonstrated that M. smegmatis produces multiple isomers of PLV that are then loaded into MR1, we next tested the capacity of each of these ligands to stabilize MR1 at the cell surface. In the absence of ligand, MR1 is predominantly sequestered in the endoplasmic reticulum, but upon ligand introduction, it is rapidly trafficked to the cell surface. Therefore, beyond the bait-and-hook method of production used to collect our initial data, the presence of cell surface–localized MR1 represents an additional indicator of the ability of a ligand to bind MR1 in vivo. Using a flow cytometry assay to assess MR1 conformational integrity with each PLV synthetic, all four isomers stabilized MR1 on the cell surface of MR1 overexpressing BEAS-2B:MR1-GFP cells (33), with the effect induced by PLV-4, the weakest, and PLV-7, the strongest (Fig. 4A, 4B). All MAIT cell clones showed similar MFI when tested for surface TCR levels, indicating that they express similar levels of TCR (Supplemental Fig. 3B, 3C).

FIGURE 4.

Functional testing of PLV reveals an MR1 ligand with differential MAIT cell activating potential. (A) Surface stabilization of endogenous human MR1 using 100 µM of known PLI, 6-FP, or the four synthetic PLV isomers and controls. The gray outlined histogram represents the NaOH vehicle, and the black outline represents the experimental ligand, each color coded to match Fig. 2C. (B) MFI of each ligand. Average of three independent experiments. (C and D) Activation of TRAV1-2⁺ MAIT cell clones D481-C7 (C) and D426-G11 (D) by PLI or PLV isomers as indicated by the number of IFN-γ spot-forming units in an ELISPOT assay. Twenty thousand DCs and 10,000 MAIT cells were used per well. (E) Activation of clone D481-C7 using an anti-MR1 Ab for MR1 blocking, as measured by an IFN-γ ELISPOT assay, using 50 μM of each ligand. Twenty thousand DCs and 10,000 MAIT cells were used per well.

We were interested in which of these isomers could activate MAIT cells when presented by endogenous human MR1. To this end, we incubated DCs with the four synthetic PLV ligands and used them as APCs in an IFN-γ ELISPOT assay with a panel of MAIT cell clones expressing diverse TCRs (Supplemental Fig. 1D), as well as DCs infected with M. smegmatis (Supplemental Fig. 3D). MAIT clone D481-C7 was previously shown to recognize PLI and PLIII whereas clone D426-G11 does not respond to these Ags (22). Remarkably, both D481-C7 and D426-G11 displayed differential activation by the PLV isomers that was dependent on the location of the hydroxyl group: PLV-4 did not activate, PLV-5 showed weak activation, PLV-6 showed moderate activation, and PLV-7 provided moderate to strong activation of D481-C7 (Fig. 4C). All four isomers of PLV exhibited a similar activation profile with MAIT cell clone D426-G11 (Fig. 4D). Intriguingly, while these two MAIT cell clones shared the same TCR β-chain, three other MAIT cell clones expressing diverse TCR β-chains did not produce IFN-γ in response to any of the PLV isoforms (Supplemental Fig. 1A–C). Additionally, blocking this interaction with the anti-MR1 26.5 Ab eliminates the IFN-γ response, indicating that this effect is MR1-dependent (Fig. 4E). While this activation was replicable for D481-C7 and D426-G11, a two-donor panel of CD8⁺ cells from donors with a high percentage of MAIT cells showed no activation against PLV-7, although they also showed no activation against the D481-C7 clone agonist PLI (22) and very little activation with regards to the synthetic MAIT clone agonist deazalumazine (34) (Supplemental Fig. 3E, 3F). In composite, these results demonstrate that the isomers of M. smegmatis–derived PLV are differentially recognized by T cells in the context of MR1, and that this differential activation is due only to the positioning of a single hydroxyl group.

Discussion

Our exploration of T cell–activating MR1 ligands from M. smegmatis in this study led to the identification of structural microheterogeneity among a number of PLV isomers. A focus on riboflavin secondary metabolites such as PLV is rooted in initial studies of MR1 Ag presentation that identified the potent ribityllumazines (4) and later 5-A-RU derivatives (24), suggesting riboflavin biosynthesis as an origin for bacterial MR1 ligands. While there is evidence for a broader diversity of ligands, these findings, and others (39, 40), provided a starting point in understanding how MR1 ligands launch cellular immunity. The examination of MR1 ligands in the Mycobacteria genus in our previous study revealed that these samples, which do not contain appreciable quantities of 6-hydroxymethyl-8-d-ribityllumazine or 5-OP-RU, potently activate MAIT cells (22). Using refinements of past techniques, we characterized a putative ligand that repeatedly appeared in statistical analyses, m/z 444.116. Through structural elucidation, it became apparent that four isomers of this putative ligand may exist, and upon synthesizing all four, we determined that at least three, and likely all four, were produced by M. smegmatis and loaded into the MR1 groove. We next discovered that each isomer stabilized MR1 on the cell surface of human cells with endogenous MR1, although to varying extents. Furthermore, all four of these isomers exhibited differential activation potential when tested against a panel of MAIT cell clones. The two clones that did respond, D481-C7, previously shown to respond to PLI and PLIII (22), and D426-G11, which had not, did so in a consistent manner wherein the response weakened as the position of the hydroxyl was shifted from the 7-position to the 4-position of the indole. Because PLV-4 was shown to have only weakly stabilized MR1 on the cell surface, this effect links, at least to some degree, MR1 surface stabilization by the PLs and MAIT cell activation. Based on these findings, it is clear that the shifting of a hydroxyl from carbon to carbon on an indole moiety in PLV changes the activation potential of the ligand. Thus, microbe-generated microheterogeneity within MR1 ligands modulates T cell reactivity. Determining the biosynthetic origin of these isomers is an ongoing effort.

It is a surprise that, as the methods for MR1 ligand identification were refined and broadened, additional small molecules derived from the riboflavin pathway continue to emerge as MR1 ligands. This may suggest that most MR1 ligands derive from this pathway or, perhaps, that the discovery of ligands derived from this pathway is attributed to the current technological approaches to ligand discovery, and that broadening our search beyond our current techniques may yield other ligands unrelated to riboflavin. Note that the methods used in the current study have limitations: bacterial cells grown in eukaryotic cell medium are exposed for a limited time to the ligand-loading mechanism, which may alter the profile of ligands presented, and the expression of a secreted protein that may not sample the endosomal compartment could prevent sampling of important ligands introduced through additional ligand-loading mechanisms, although to date we have seen no evidence that this is the case. Despite these limitations, tetramers of MR1 produced using the methods in the current study robustly stain MAIT cells (22). As these methods continue to evolve, MR1 ligands originating from other metabolic pathways may well emerge.

From a T cell perspective, although the repertoire of αβ TCR rearrangements observed for MR1T cells continues to expand, implying the role of additional cognate ligands, emerging data also suggest that even minor differences among conventional MAIT cell TCRs allow for differential recognition of metabolites (33, 41). Indeed, our data show an interesting correlation between the TRBV usage and PLV recognition, furthering the hypothesis that the TCRβ “tunes” the semiconserved mechanism of TRAV1-2⁺ MAIT cell TCR engagement and may facilitate discrimination among structurally similar derivatives of the riboflavin pathway (6). The exact mechanisms underlying the interplay between ribityllumazine/ribitylpyrimidine microheterogeneity and TCR sequence, however, is yet to be explained. In silico, we have observed that ligands varying at positions obscured from the TCR promote differential dynamics of the ribityl moiety when bound within MR1, which we have postulated may promote their differential recognition by MAIT clones (23, 34, 42). In the current study, it is remarkable that the subtle positioning of the hydroxyl group around the indole moiety influences TCR/ligand interactions; ongoing structural studies of MR1/TCR ternary complexes will hopefully answer these questions.

In summary, we report the identification and characterization of a (to our knowledge) new MR1 ligand, PLV, that is represented by four distinct isomers. Our data show that M. smegmatis produces these four PLV isomers that act as MR1 ligands for varied recognition by T cells. The discovery of this microheterogeneity within a single ligand opens the door to a wide variety of unrealized secondary metabolites as MR1 ligands. The identification and characterization of additional MR1 ligands that arise during microbial growth, both those identified in this study and others yet to come, will add to the growing repertoire of MR1 ligands of immunologic significance.

Disclosures

The authors have no financial conflicts of interest.