Alanine-scan Mutagenesis of the 6C2 MAIT TCR
The 6C2 MAIT TCRα and β chains were expressed in a TCR-negative hybridoma. The TCR-expressing hybridoma produced large amounts of IL-2 (10 to 100-fold over background) when co-cultured with LM1.8 fibroblasts transduced with a mouse MR1-encoding construct 
and the response could be blocked by the addition of anti-MR1 but not isotype control mAbs (data not shown), thereby reproducing the reactivity of the original 6C2 TCR 
Individual residues in the CDRs of the 6C2 TCR were substituted with alanine and each mutant chain was expressed together with the appropriate wild-type partner. Each hybridoma was sorted for similar level of TCR surface expression and the sorted cells of each mutant were demonstrated to have equivalent responses to anti-CD3, anti-CD28-coated plate stimulation (data not shown). Stimulation of these hybridomas using the LM1.8 fibroblast overexpressing MR1 (, A and B) provided us with the pattern of reactivity of these mutants to the presumably self-antigen(s) expressed in fibroblast cells and presented by MR1 molecules or, alternatively, to ligand(s) provided by the culture media. Several interesting observations can be drawn from these results. First, the requirement for residues encoded within the invariant TCRα chain, especially the CDR1α loop, is much more pronounced than it is for residues within the TCRβ chain. Residues within the CDR1α (T26α, G28α, F29α, N30α, G31α), CDR2α (Y48α, V50α, L51α) and CDR3α (D92α, S93α, Y95α, I98α) loops were all necessary for the recognition of the self-antigen-MR1 complex.
Mutational analysis of the mouse 6C2 MAIT TCR in response to MR1-overexpressing antigen presenting cells.
Based on the mouse TRAV1 (used by MAIT cells), TRAV11 (used by iNKT cells) and TRAV19 (used by conventional T cells) sequences, we searched databases for orthologue genes in sixteen different species of placental mammals and performed a phylogenic analysis rooted to the TRAV1S1 sequence from Oncorhynchus mykiss
(Rainbow Trout) as a reference. Based on this analysis, the different mammalian TRAV1 genes appear more closely related to each other than the TRAV11 and TRAV19 genes are, suggesting that TRAV1 genes might have been more conserved in the course of evolution than those of TRAV11 or TRAV19 (Fig. S1A
). Sequence alignment showed that all the MAIT TCRα residues involved in the recognition of antigen-MR1 are conserved, thereby providing a potential explanation for the unique use of the TRAV1 and TRAJ33 gene segments by the MAIT TCR (Fig. S1B
In addition, several residues within the 6C2 Vβ chain affected the recognition of mouse MR1. Mutations of histidine 27 in the CDR1β loop and of Y46β in the CDR2β loop abolished reactivity to MR1 on fibroblasts. Interestingly, tyrosine residues at position 46 and 48 of the CDR2β loop of Vβ8 family gene segments have been proposed as important evolutionarily conserved residues for the generic recognition of MHC molecules 
. Thus our results potentially extend the involvement of the Y46β residue to the recognition of MR1.
Two residues, G95β and E96β, within the variable CDR3β loop appeared important to the recognition of a putative self antigen(s) presented by MR1 transfected fibroblasts. Although the glycine residue is unlikely to represent a direct contact residue, its flexibility might allow for the CDR3β loop to adopt the “right” configuration necessary for recognition, suggesting the possibility that different MAIT TCRs might use such flexibility to fine-tune their antigen specificities. Further mutational analysis of other autoreactive mouse MAIT TCRs will be necessary to demonstrate whether these results are restricted to the 6C2 TCR or can be extended the MAIT population in general.
Overall, the majority of residues important for recognition of self-MR1 in the MAIT TCR are concentrated in the TCRα chain, however, the TCRβ chain contributes a significant portion of these residues as well. This supports the idea that the degree of conservation of the MAIT TCRα chain is important for reactivity to MR1 and the usage of several different TCRβ chains might allow for more flexibility. These results are in agreement with a recent study that identified the energetically important residues for the recognition of MR1 by the human MAIT TCR with the exception that no residue within the human Vβ chain was found essential in mediating MR1-restricted activation of the human MAIT TCR 
MAIT TCR Mutations Reveal a Similar Pattern of Reactivity to MR1-transfected B cells
The ubiquitous expression of MR1 transcripts suggests that several cell types might be able to present antigens to MAIT cells. However, it remains unclear whether MR1 might be presenting antigen(s) common to all APCs or whether different APCs might each express a unique set of antigen(s). In support of the latter possibility, B cells are uniquely required for the expansion of MAIT cells in the periphery 
We compared the reactivity of each of our MAIT TCR mutants with the LM1.8 fibroblast and the CH27 B cells, which overexpress MR1 on their surface. The pattern of responses of the hybridoma collection to B cells was similar to that obtained with fibroblasts (, C and D). Minor differences between the two APC types were noticed for residues V49β, D51β and E96β but they probably reflect the stimulatory effectiveness of each APC than actual differences in MR1 recognition. Indeed, rat MR1-transfected CH27 cells, which provide a stronger stimulatory signal to the 6C2 hybridoma 
, did not perturb the pattern of recognition of the different TCR mutants (data not shown).
Thus, the self antigen(s) presented by MR1 expressed on cell lines of two different origins (fibroblast and B cells) appear to be similar in their ability to engage a MAIT TCR.
MAIT TCR Mutants Reveal a Different Pattern of Reactivity to Cells Co-cultured with E. coli
Various strains of bacteria and yeast activate MAIT cells in an MR1-dependent manner 
. A still unidentified antigen(s) common to these microbes is thought to be directly presented by MR1 to MAIT cells. The 6C2 hybridoma responded to MR1 transfectants but was unresponsive to untransfected APCs. Addition of E. coli
to the cultures induced a response by the hybridoma when added to the untransfected APCs while it only minimally increased the autoreactive response (Fig. S2
). MR1 mRNA and protein are ubiquitously expressed, including in the untransfected APC cell line used 
, but the cell surface expression of endogenous MR1 is extremely low and is not sufficient to readily trigger a response 
. Addition of E. coli
to the hybridoma in absence of APCs also does not trigger any IL-2 release from the hybridoma suggesting that the observed response is not due to autopresentation. Yet, the hybridoma response can be blocked by the addition of anti-MR1 mAbs when E. coli
is added to the untransfected line. These results suggest that addition of E. coli
might induce endogenous MR1 to accumulate on the surface of APCs at levels that are sufficient to activate MAIT cells. However, these expression levels remain too low to be detected serologically (
and data not shown).
Bacteria could induce the response either by directly providing a product or by indirectly activating the production of a self-ligand that, when bound to MR1, engages the TCR and activates MAIT cells. The former idea is likely correct since previous studies have indicated that MAIT cell activation occurs independently of Toll-like receptors or other innate pathways and that MAIT cells can be activated by fixed APCs cultured in the presence of bacteria 
Several mutations in both the TCRα and TCRβ chains had different effects depending whether the response was directed against self or the bacteria-induced response (). In the TCRα chain, alanine substitution of T26α in the CDR1α and Y48α, V50α, S93α and I98α in the CDR2α and CDR3α loops affected the autoreactivity but not the bacteria-induced response. In contrast, mutation of R91α in CDR3α had the opposite affect since it did not influence the autoreactive response but decreased the bacteria-induced response. For the TCRβ chain, H27β, Y46β, and G95β were differentially required between the autoreactive and bacteria-induced responses since alanine substitution of these residues resulted in loss of autoreactivity, but had no effect on the E. coli-induced response.
Mutational analysis of the mouse 6C2 MAIT TCR in response to antigen presenting cells infected with E. coli.
The differential sensitivity to TCR sequence changes of the two responses could be detected using either fibroblasts or B cells as APCs, although overall the response of the mutants to the B cells was lower (). Altogether, these results suggest that an E. coli-derived antigen(s) is directly recognized by the MAIT TCR in a manner that is similar but not identical to the recognition of the self-antigen(s). It is tempting to speculate that different antigen(s) might be recognized in the two situations or that the bacteria-derived antigen induces MR1 to adopt a slightly different conformation than the self-antigen(s).
Effect of MR1 Mutations on the Reactivity of the 6C2 MAIT TCR
Next, we analyzed how 10 different mutations localized in the predicted helical residues of the MR1 molecule 
affected the response of the 6C2 hybridoma to MR1-overexpressing APCs or its response to the E coli
-derived ligand(s). Two mutations located on the α1 helix (L65 and G68) and two mutations on the α2 helix of MR1 (A163 and Y152) affected the autoreactivity of the 6C2 hybridoma (). Only the residues L65 and Y152 remained essential to the reactivity of the 6C2 TCR towards the E. coli
-derived ligand. These centrally located residues (), demonstrate an extremely focused energetic hot spot on the MR1 molecule necessary for the recognition by the 6C2 TCR. Interestingly, this hot spot is located next to the Q151 residue. Such position is encoded by a leucine in human MR1 and was shown to be the sole reason that human MR1 fails to activate mouse MAIT cells 
The MAIT hybridoma response to MR1 mutants.
The E. coli-derived MAIT Antigen is Resistant to Proteinase K Digestion and Lipid Extraction
The nature of the antigen(s) recognized by MAIT cells remains to be determined. Previous experiments have suggested that it was sensitive to protease digestion 
. Glycolipids have also been proposed as MAIT cell antigens 
A <10 kD fraction isolated from E. coli contains the activity that stimulates the 6C2 hybridoma but not the OVA-specific DO11.10 hybridoma or the α-Galcer-reactive DN32.D3 hybridoma (data not shown). We therefore independently spiked the fraction with either ovalbumin or α-Galcer.
The fraction was further treated, or not, with proteinase K and used to individually stimulate the DO11.10 or the MAIT 6C2 hybridoma. Proteinase K treatment of the OVA-spiked fraction completely abrogated the response of the DO11.10 hybridoma (), while the stimulation of the 6C2 hybridoma was unaffected ().
The MAIT ligand is resistant to digestion with proteinase K and is not part of the lipid component of <10 kD E. coli culture.
Alternatively, the fraction was extracted with the Folch method to separate polar from non-polar compounds 
. The organic, interphase and aqueous phases were isolated, dried and resuspended in complete media. Each phase was then used to stimulate the 6C2 or DN32.D3 hybridomas. As expected, αGC activity was found within the organic and interphases but was absent from the aqueous phase (). In sharp contrast, only the aqueous phase contained the stimulatory activity for the MAIT 6C2 hybridoma (). Altogether, these data show that the MAIT ligand is rather hydrophilic but is resistant to proteinase K digestion. These characteristics contrast with the nature of most of the currently known antigens presented to T cells (ie. peptides by conventional MHC class I and II, formylated-peptides by H2-M3, phospholipids, glycolipids and lipopeptides by members of the CD1 family), and argue that MAIT cells might be directed at an unusual class of antigen(s).
While our manuscript was under review, Kjer-Nielsen and colleagues 
reported the first crystal structure of human MR1 in complex with 6-formyl pterin, a vitamin B9 metabolite. Further, they showed that other metabolites originating from the riboflavin (vitamin B2) metabolic pathway could stimulate human MAIT cells in the presence of an Epstein-Barr virus (EBV)-transformed B cell lymphoblast transfected with a human-MR1-encoding construct 
. These results are in good agreement with the findings described above and suggest that vitamin B metabolites might represent a new class of antigens that are presented by MR1 for MAIT-cell immunosurveillance of microbial infections.