Alternative splicing of MR1 regulates antigen presentation to MAIT cells

Mucosal Associated Invariant T (MAIT) cells can sense intracellular infection by a broad array of pathogens. These cells are activated upon encountering microbial antigen(s) displayed by MR1 on the surface of an infected cell. Human MR1 undergoes alternative splicing. The full-length isoform, MR1A, can activate MAIT cells, while the function of the isoforms, MR1B and MR1C, are incompletely understood. In this report, we sought to characterize the expression and function of these splice variants. Using a transcriptomic analysis in conjunction with qPCR, we find that that MR1A and MR1B transcripts are widely expressed. However only MR1A can present mycobacterial antigen to MAIT cells. Coexpression of MR1B with MR1A decreases MAIT cell activation following bacterial infection. Additionally, expression of MR1B prior to MR1A lowers total MR1A abundance, suggesting competition between MR1A and MR1B for either ligands or chaperones required for folding and/or trafficking. Finally, we evaluated CD4/CD8 double positive thymocytes expressing surface MR1. Here, we find that relative expression of MR1A/MR1B transcript is associated with the prevalence of MR1 + CD4/CD8 cells in the thymus. Our results suggest alternative splicing of MR1 represents a means of regulating MAIT activation in response to microbial ligand(s).

While MHC Class I molecules traditionally present peptides to stimulate CD8 + T cell responses, MR1 binds and presents microbial small molecule metabolites to MAIT cells 10,15 . These antigens were first described as intermediates in the riboflavin synthesis pathway, but recent reports have highlighted the increasing diversity of the MAIT ligand repertoire [15][16][17] . For example, MR1 also presents antigen(s) from S. pyogenes, a bacteria that cannot synthesize riboflavin 4 . In silico screens showed that MR1 can bind a range of synthetic compounds, including commonly prescribed pharmaceuticals 16 , We have recently shown through metabolomics analysis that MAIT stimulatory antigens include ligands distinct from those generated in the riboflavin synthesis pathway 5 .
The underlying gene organization of MR1 is distinct from canonical MHC molecules. MR1 is non polymorphic, highly conserved across species and individuals, with the transcript ubiquitously expressed [18][19][20] . MR1 pre-mRNA undergoes alternative splicing to produce multiple isoforms, which have been demonstrated at the transcript level to be expressed in human tissues and cell lines 21 . The structure of MR1 is similar to that of MHC Class I molecules, with α1 and α2 domains that bind ligand, an α3 domain that interacts with β2-microglobulin, and a transmembrane domain for surface expression 21,22 . The full length isoform, MR1A, contains all encoded exons and can stimulate MAIT cells. The shorter isoform MR1B lacks the α3 domain but does encode the ligand binding and transmembrane domains. The function of MR1B is remains incompletely understood. Overexpression of MR1B in a fibrosarcoma model suggested a functional role for MR1B in stimulating MAIT cells following infection with E. coli 23 . MR1C is a putative soluble isoform, lacking both the α3 and transmembrane domains, however, there are no reports on its expression or function.
Here, we sought to determine the role of MR1 isoforms in the presentation of microbial ligand to MAIT cells. We show MR1A and MR1B transcripts are detectable across human tissues, with considerable variation in isoform expression among donors and tissues. We developed a lung epithelial cell line deficient in MR1 and utilized this system to show that MR1B can antagonize MR1A in the presentation and/or processing of mycobacterial antigen(s). While MR1A is observed in the ER and vesicular compartments, MR1B appears to reside primarily in intracellular vesicles. Finally, we show that surface expression of MR1A on CD4 + CD8 + MR1 expressing thymocytes is associated with relative abundance of MAIT cells in the thymus. Taken together, our results suggest that the splice variant MR1B can regulate the response of MAIT cells to intracellular infection.

MR1A and MR1B are ubiquitously expressed.
To study expression of the MR1 splice variants, we used Snaptron, a tool for exploring exon-exon junction expression across thousands of publicly available RNA sequencing (RNA-seq) samples 24,25 . We analyzed exon-exon junctions corresponding to inclusion or exclusion of exon 4 to distinguish MR1A from MR1B, using nearly 10,000 RNA-seq samples from 31 non-diseased tissue sites in over 500 deceased individuals from the Genotype Tissue Expression (GTEx) project (see Materials and Methods) 26 . Here, we show that both MR1A and MR1B are expressed across human tissues, as seen in prior, nonquantitative studies ( Fig. 1) 18,21,27 . The ratio of MR1A to MR1B varied across tissues, with higher relative MR1A transcript observed in blood, bone marrow, liver, and lung, and lower relative MR1A observed in uterine cervix, breast, small intestine, and colon. Interestingly, we observed that in all the tissues queried, the MR1A/MR1B ratio was consistently less than 0.5, suggesting that all the tissues express higher relative MR1B than MR1A. As these data are from the total mRNA for a given tissue, this analysis does not take into account diversity of the individual cell types that comprise each tissue.
Quantitative RT-PCR analysis reveals differing levels of MR1 splice variants in antigen presenting cells. To explore the relative MR1 splice variant expression specifically in primary cells, we focused on macrophages, monocyte derived DC as well as epithelial cells, all of which can present Mtb-derived antigens to MAIT cells 12 . Using quantitative real-time PCR (qRT-PCR), we generated a standard curve for each splice variant by using plasmids designed to be specific to each isoform based on unique exon-exon junctions ( Fig. 2A). We found that these primers were specific in that we did not observe amplification in a Bronchial epithelial cell line (Beas2B) lacking MR1, and that transfection with plasmids encoding each isoform were specifically amplified, with no amplification detected for the other splice variants. ( Fig. 2A and Supplemental Fig. 1). In this manner, we were able to quantify absolute as well as relative amounts of the MR1 splice variants. Macrophages and DC were prepared from PBMC 28 . We observed that in both dendritic cells (DCs) and macrophages, both MR1A and MR1B transcripts were readily detected (Fig. 2B). To determine the expression of MR1 splice variants in primary human epithelial cells, we generated single cell suspensions from lung parenchyma and small intestine and subsequently sorted epithelial cells by EpCam (CD326) expression (Fig. 2C, top-Lung Epithelial Cells, bottom-Small Intestine Epithelial Cells). Using qRT-PCR, we found that the relative expression of MR1A was again similar to MR1B. Interestingly, both lung and small intestine epithelial cells expressed nearly 2-log lower levels of MR1 splice variants than PBMC derived myeloid cells. We observed a discrepancy between the quantities of the MR1 isoforms in isolated epithelial cells from tissue versus bulk tissue from the RNA-seq analysis in Fig. 1. Therefore, we performed the same analysis on single cell suspensions derived from unfractionated small intestine tissue and observed ratios of MR1A/MR1B similar to what was observed in Fig. 1. (Fig. 2D). This indicates that the relative expression of MR1A and MR1B vary greatly between tissue types, as well as cell types within each tissue.
As primary large airway epithelial cells (LAEC) are capable of presenting mycobacterial antigen to stimulate MAIT cells, we prepared LAEC from human airway tissue, and confirmed that these cells expressed for EpCam ( Fig. 2E)  MR1B inhibits T cell activation by MR1A. We next sought to assess the function of the MR1 isoforms in the absence of endogenous MR1. As previously described in A549 cells, CRISPR/Cas9 was used to generate a bronchial epithelial cell line (Beas2B) that lacked the gene for MR1 (Beas2B:MR1_KO) 29 . To identify MR1 deficient cells, we performed limiting dilution cloning, and then screened for the absence of 6FP induced MR1 surface stabilization (Fig. 3A) 30 . Functionally, we found that these clones could not stimulate an MR1-restricted clone in response to M. smegmatis infected APC (Fig. 3B). In contrast, stimulation of both HLA-B45 and HLA-E restricted clones was not decreased, and in the case of HLA-B45, was slightly, but not significantly, increased, indicating that the knockout was specific to MR1 and not to MHC Class I in general (Fig. 3B). We next generated expression vectors containing the genes for MR1A, MR1B, or MR1C under the control of the CMV promoter. To distinguish MR1A from MR1B or MR1C following transfection, we tagged MR1A with GFP and MR1B and MR1C with RFP (pCI:MR1AGFP, pCI:MR1BRFP, pCI:MR1CRFP). We then transfected these plasmids individually into Beas2B:MR1_KO and A549:MR1_KO cells and infected the cells with M. smegmatis overnight in order to stimulate MAIT cells (Fig. 3C). Flow cytometry was used to confirm protein expression both in single transfectants and co-transfected cells (Fig. 3D). While MR1C protein was detectable by flow cytometry, it was expressed at markedly lower levels than MR1B both in single and cotransfected samples (Fig. 3D). We verified that transcript of each isoform was expressed by performing qRT-PCR on cells transfected with pCI:MR1AGFP, pCI:MR1BRFP, pCI:MR1CRFP (Supplemental Fig. 1).
As we found that neither MR1B nor MR1C could present antigen to MAIT cells, we postulated that these isoforms could work in concert with MR1A to modulate MAIT activation. To test this hypothesis, we coexpressed pCI:MR1AGFP with pCI:MR1BRFP or pCI:MR1CRFP in Beas2B:MR1_KO and A549:MR1_KO cells, and subsequently infected the cells with M. smegmatis. Coexpression of MR1A with MR1B resulted in a 40% decrease in MAIT cell activation in both cell lines with M. smegmatis infection (Fig. 3E, left, middle). However, coexpression of MR1C with MR1A did not lead to inhibition of MR1A-mediated presentation of M.smegmatis-derived ligand. While we observed inhibition of TCR activation when there was a 1:1 ratio of MR1B:MR1A, we observed a more marked inhibition of MR1A-dependent antigen presentation when we transfected a 2:1 ratio of MR1B:MR1A expressing plasmid, keeping the quantity of MR1A fixed (Supplemental Fig. 2). We selected this higher dose of MR1B for all subsequent experiments, as this quantity of plasmid was most similar to RNA data (Fig. 1). As antigens derived from M. smegmatis can be presented exogenously, we sought to test a mycobacteria for which recognition depends on intracellular infection 14 . As shown in Fig. 3E   In brief: RNA was isolated from 5e04 cells and qRT-PCR was performed to quantify the absolute amounts of MR1A or MR1B mRNA as measured by a standard curve designed to be specific for each isoform.
(B) qRT-PCR performed on monocyte-derived macrophages and monocyte-derived dendritic cells generated from human PBMC (n = triplicate wells from one experiment, error bars represent mean and standard deviation from the mean). (C) Human lung parenchyma (top) and small intestine lamina propria (bottom) were stained with antibodies against CD45, CD20, CD14 and CD326 (EpCAM). CD45-CD20-CD14-CD326 + cells were sorted for mRNA isolation and cDNA synthesis. Briefly, cells were gated sequentially on cell shape, singlet cells, live, CD3-, CD45-/CD14-/CD20-, CD326 + (Epithelial Cells). qRT-PCR was performed as described twice in triplicate (n = 6, error bars represent mean and standard deviation from the mean). (D) qRT-PCR was performed on unfractionated small intestine from 4 donors. Experiment was performed twice with triplicate wells. (n = 6, error bars represent mean and standard error of the mean of 6 replicates) (E) Human Large Airway Epithelial Cells (LAEC) were isolated and cultured from the upper airway and stained with an antibody to either isotype (IgG2b) or CD326 (EpCAM) to verify surface expression of EpCAM. mRNA was isolated and qRT-PCR performed as described. (n = 3 replicates from 1 experiment, error bars represent mean and standard deviation from the mean).  www.nature.com/scientificreports/ also observed following infection with BCG. Co-expression of MR1B or MR1C with MR1A resulted in a slight increase in constitutive MR1A surface protein abundance as measured by flow cytometry (Fig. 3F). Additionally, coexpression of MR1B with MR1A did not impede antigen presentation of peptide antigen by Beas2B:MR1_KO cells to HLA-B45 restricted T cells, indicating that the inhibition was specific to MR1 (Supplemental Fig. 3). Taken together, these results indicate that MR1B is functioning selectively to inhibit T cell activation by MR1A.

MR1B resides in the same intracellular compartments as MR1A.
In the absence of ligand, MR1A has been found to reside in the endoplasmic reticulum as well as intracellular vesicles 11,13,14 . The subcellular localization of MR1B is not fully understood, however, transient overexpression of MR1B in a fibrosarcoma cell line showed that MR1B is capable of forming stable homodimers at the cell surface 23 . However, these studies were performed in cell lines expressing endogenous MR1 isoforms. To elucidate the subcellular localization of MR1A and MR1B in knockout cell lines, we individually expressed pCI:MR1AGFP or pCI:MR1BRFP in Beas2B:MR1_KO cells and performed live cell imaging (Fig. 4A). MR1A (left) was detectable in both the endoplasmic reticulum as well as intracellular vesicles, as has been previously described 11,13,14 . As shown in Fig. 4A (right), MR1B had a similar intracellular distribution to MR1A, particularly localizing to intracellular vesicles. Co-expression of pCI:MR1AGFP and pCI:MR1BRFP in Beas2B:MR1_KO cells demonstrated association of MR1A and MR1B in similar compartments (Fig. 4B). This suggested that MR1B could depend on its close proximity to intracellular MR1A for the inhibition we observed. It has been previously established that translocation of MR1A to the cell surface is dependent on the availability of ligands such as 6FP 13,30 . As MR1A and MR1B reside in the same intracellular compartments, we postulated that MR1B could interfere with either the translocation of MR1A to the cell surface or the engagement with the MAIT T cell receptor on the cell surface. Beas2B:MR1_KO cells transfected with either pCI:MR1AGFP or pCI:MR1BRFP were exposed to 6FP overnight and imaging was performed to detect cell surface translocation. While we observed cell surface expression of MR1A, MR1B remained in intracellular vesicles (Fig. 4C). Therefore, these results suggest that MR1B-mediated antagonism is mediated intracellularly, prior to MR1A cell-surface translocation, rather than to directly block the MAIT TCR.
MR1B expression limits abundance of MR1A protein.
In the experiments described above, both MR1A and MR1B were expressed under the control of the CMV promoter. However, as described above, both the abundance and relative expression of each isoform was highly variable. To control the timing and expression of MR1A, we developed a MR1A construct under the control of a doxycycline-dependent promotor. To accomplish this, we stably transduced Beas2B:MR1_KO cells with a lentivirus encoding a tetracycline-inducible MR1AGFP (Beas2B:MR1KO_Tet_MR1AGFP) and sorted the transduced cells based on GFP expression following overnight administration of doxycycline (Fig. 5A) 14 . We confirmed that these cells were capable of presenting  www.nature.com/scientificreports/ M. smegmatis-derived antigen to MAIT cells upon overnight induction of MR1A expression with doxycycline. Using this cell line, we could control the expression of MR1A expression by administering doxycycline either before or after transfecting the cells with pCI:MR1BRFP ( Fig. 5B-left). In all conditions we performed infections 48 h following MR1B transient transfection to ensure optimal expression of MR1B and consistency with prior experiments. In our prior work, we had used doxycycline inducible MR1AGFP to define the optimal timing for expression of MR1AGFP 14 . We chose to induce MR1AGFP expression with doxycycline for 18 h either prior to transfection of MR1BRFP, or prior to flow cytometry or ELISpot, as this would ensure maximal expression of MR1AGFP, based on prior work 14 . To ensure that MR1AGFP expression was not artificially increased due to prolonged exposure to doxycycline, we showed that roughly 85% of these cells expressed MR1AGFP both before and after transfection with empty vector (Fig. 5B, right). Interestingly, transfection of pCI:MR1BRFP into Beas2B:MR1KO_Tet_MR1AGFP cells prior to the induction of MR1AGFP expression resulted in a 20% increase in MR1AGFP-negative cells, as compared to cells with doxycycline-induced MR1A expression prior to the transfection of pCI:MR1BRFP (Fig. 5C). This suggested that the expression of MR1B prior to the induction of MR1A could modulate total MR1A protein abundance. In contrast, the expression of pCI: MR1CRFP either pre or post induction of MR1A had no effect on MR1A protein production (Fig. 5D).
To assess the impact of MR1BRFP expression prior to induction of MR1AGFP expression on the recognition of microbial ligands, we repeated the experiment as described in Fig. 5B, and infected the cells with M. smegmatis (Fig. 5D). We observed that, compared to the no-doxycycline control, Beas2B_MR1KO:Tet_MR1AGFP transfected with an empty vector were capable of stimulating MAIT cell activation upon the administration of doxycycline (Fig. 5D). Upon transfection of MR1B in MR1A expressing cells, we observed that when MR1A and MR1B were both expressed, there was a 30% decrease in the activation of MAIT cells, as seen in prior experiments with MR1A expressed using the CMV promoter. However, we observed an approximately 50% inhibition of MAIT activation when MR1B was expressed prior to induction of MR1A protein, which might reflect the lower MR1A protein abundance under these conditions (Fig. 5D, left). Expression of pCI:MR1CRFP pre or post induction of MR1A expression did not impact MR1A-mediated expression or antigen presentation, indicating this was not a general consequence of transfection (Fig. 5E, 5F). Thus, prior expression of MR1B may inhibit the abundance of MR1A protein, either through competing for chaperones necessary for MR1A folding, or by competing for factors required for protein synthesis of MR1.
Expression of MR1 splice variants in CD4/CD8 MR1-expressing thymocytes. Amongst PBMC, surface expression of MR1 has been difficult to observe in the absence of either exogenous or bacterially-derived antigen(s) 31 . We have previously shown that CD4 + CD8 + double positive (DP) thymocytes constitutively express high levels of cell-surface MR1 32 . It has been previously reported that the percentage of MR1-expressing thymocytes varies from 0.5 to 15% of DP thymocytes 32 . The role of MR1 expressing DP thymocytes in the selection of MAIT thymocytes is not well understood 33 . To determine whether MR1 + DP thymocytes are capable of presenting mycobacterial antigen to activate MAIT cells, we enriched for MR1-expressing DP thymocytes from human thymus using positive selection following surface staining for MR1A (Table 1). To determine if these cells could stimulate MAITs, the MR1 + enriched fraction as well as the MR1 depleted fraction were incubated for 1 h with M. smegmatis (MOI = 3) and tested for their ability to activate MAITs. As shown in Fig. 6A, the MR1-expressing DP thymocytes could activate MAIT cell production of IFNγ as compared to thymocytes not expressing cell surface MR1.
To assess whether relative expression of the MR1 isoforms is associated with prevalence of MR1A expressing thymocytes, we sorted MR1 expressing DP thymocytes from nine donors (Fig. 6B). We then measured expression of MR1A and MR1B mRNA by qRT-PCR, which we then compared to surface MR1A expression in these donors (Fig. 6C). We observe that there was not a strong association between MR1A mRNA and surface expression of MR1 (R 2 = 0.03354, p = 0.637), but there was a significant association between MR1B transcript and the frequency of MR1 + DP thymocytes (R 2 = 0.4486, p = 0.0484). Intriguingly, we observed a strong positive association (R 2 = 0.8196, p = 0.0008) between relative MR1A/MR1B transcript and the frequency of MR1 + DP thymocytes.
To explore the relationship between DP MR1 expressing thymocytes and MAIT abundance in the thymus, we used the tetramer of MR1 bound to 5-OP-RU (MR1/5-OP-RU) to identify MAITs (Fig. 6D) 34 . Among CD3 thymocytes, we observed frequencies of MAIT cells, with variation across donors (0.01-0.45% of live CD3 + cells, Table 1). Intriguingly, the frequency of MR1 expressing DP thymocytes was associated with the frequency of MR1/5-OPRU + thymocytes (Fig. 6E) (R 2 = 0.4709, p = 0.0412). However, the absolute amounts of MR1A and MR1B mRNA did not appear to be significantly correlated with frequency of MR1/5-OP-RU + thymocytes (Fig. 6F). These results suggest a potential role for alternative splicing of MR1 in contributing to selection and development of MAIT cells in the thymus.

Discussion
In this report, we show that MR1B, an alternatively spliced form of MR1, inhibits mycobacterial antigen presentation to MAIT cells by the full-length variant, MR1A. Both MR1A and MR1B are ubiquitously expressed, but with considerable variation between both discrete cell and tissue types as well as individual donors. These results suggest that MR1B can serve to modulate the recognition of MR1 ligands by MR1A. We postulate that in addition to microbial recognition, MR1B could play a role in the prevention of unwanted MAIT cell activation, and in the development of MAIT cells.
These results raise a number of questions about the mechanism(s) by which MR1B mediates this inhibitory function. As we found that MR1B did not traffic to the cell surface following exposure to exogenous ligand, we believe that MR1B is not directly blocking TCR engagement with MR1A. Additionally, as MR1A and MR1B reside in the same intracellular compartments, we postulate that the inhibitory effect of MR1B could depend   www.nature.com/scientificreports/ upon its proximity to MR1A. For example, MR1B may be competing for ligand, limiting the ability of MR1A to bind and present antigen to MAIT cells, or reducing MR1A folding and stability. Alternatively, MR1B could compete for binding to a molecular chaperone which is necessary for the loading and/or stabilization of the MR1A molecule following exposure to bacterial ligands, or for subsequent translocation of the loaded MR1Aligand complex to the cell surface. While coexpression of MR1A and MR1B resulted in decreased MAIT activation, we did not observe a decrease in MR1A protein abundance or surface expression. However, the co-transfection approach cannot replicate the variations in MR1A and MR1B expression and relative expression that we observed in our analysis of gene expression across cells and tissues. To address this, we have developed an MR1A expression system dependent on doxycyline, that in turn allows to vary the timing and abundance of MR1A relative to MR1B. Here, we found that expression of MR1B prior to MR1A resulted in a decrease in total MR1A protein, further inhibiting the presentation of ligands to MAIT cells. Further studies involving analysis of MR1B protein structure, ligand binding, or possibly its associated chaperones could would be logical next steps to elucidate the mechanism by which MR1B is acting to inhibit MR1A.
We found that both the abundance and relative expression of MR1A and MR1B varies considerably across tissues and donors. At present, the significance of this variation is not clear with regard to MAIT cell prevalence in tissues, their state of activation, and their ability to respond to MR1 antigens. Preliminarily, we note that tissues such as blood and spleen expressed relatively higher levels of MR1A/MR1B, while uterus, breast, and colon expressed lower MR1A/MR1B transcript. As these data were derived from mRNA from the whole tissue rather than specific cell types, the full significance of these data will require further investigation. Importantly, we observe that in unfractionated single cells prepared from the small intestine, we observed relative expression of MR1A/MR1B transcript that was similar to what we observed in our analysis of large datasets. This would suggest that different cell types within a given tissue can express markedly different levels of MR1 isoforms. However, the wide variation would lead us to speculate that the relative abundance of MR1B could be associated with diminished activation of tissue resident MAITs. This could have significance with regard for the ability to recognize microbial infection or be associated with a propensity for autoimmunity.
Alternative splicing is often regulated by genetic modifications such as changes in DNA methylation or single nucleotide polymorphisms (SNPs) 35,36 . Recently, an intronic SNP in MR1 was shown to be associated with susceptibility to tuberculosis 37 . These results raise the possibility that regulation of MR1A and MR1B presentation is controlled at the level of gene expression. Alternately, we postulate microbial infection could alter expression of MR1 splice variants.
Alternative splicing is common and represents an important means of controlling gene expression. For instance, dysregulated splicing has been associated with both autoimmunity and inflammation [38][39][40] . However, the role of alternative splicing in the regulation of MHC and subsequent acquisition of adaptive immunity is incompletely explored. An α3 domain deletion mutant of HLA-A11 can promote evasion of NK cell mediated  www.nature.com/scientificreports/ cytotoxicity of the HIV-infected cell 41 . Alternative splicing of the non-classical molecule HLA-G has been reported and is hypothesized to be associated with protection of the developing fetus 42,43 . These reports suggest that a putative role for alternative splicing is to control immune cell activation and therefore target immune responses to microbial infection. A recent study suggested that viral infection downregulates surface expression of MR1 and thereby presentation of MR1-bound antigen to MAIT cells, through the HSV-1 protein US3 44 . The mechanism of this is unknown but suggests that impeding MR1 function can further regulate MAIT cell activity. Here, the relationship of MR1 splice variants in the context of both infection and autoimmunity remains to be explored. Our results demonstrate an association between the relative MR1A/MR1B expression and prevalence of MR1A-expressing CD4 + CD8 + thymocytes. CD1d expressing DP thymocytes have been shown to contribute to the selection of iNKT cells, and MR1 + DP thymocytes have been suggested to play a similar role in the selection of MAIT cells in the thymus 45,46 . Furthermore, we find that the prevalence of MAIT cells in the thymus was associated with MR1A-expressing DP thymocytes. While the frequency of MR1A expressing DP thymocytes varied, cell surface of MR1 protein was relatively constant between donors. Therefore, it is clear that the cell surface expression of MR1 is not strictly dependent on the relative abundance of the MR1A and MR1B splice variants. Recent work by Legoux and colleagues demonstrated that intrathymic development of murine MAIT cells could be governed, in part, by metabolites produced by commensal bacteria presented by MR1 47 . Therefore, mechanisms regulating MR1 expression and capacity to bind and present antigen, could potentially also contribute to thymic development of MAIT cells. We observed that relative MR1A/MR1B expression on DP thymocytes was associated with the prevalence of MR1 + DP thymocytes. Additionally, the prevalence of MR1 + DP thymocytes was associated with that of MAIT precursors in the thymus. How DP MR1 expressing thymocytes function in MAIT development and the role of alternative splicing in this development will require detailed in vitro modeling. One limitation of our study is that we did not have matched PBMC from our thymus donors, and therefore, we were not able to establish the relationship of intrathymic MR1 expressing cells and the prevalence of MAITs in the blood. Taken together, our results suggest a potential role of alternative splicing in modulating presentation of microbial ligand to MAIT cells.

Materials and methods
Statement. All experiments and methods were performed in accordance with relevant guidelines and regulations.
Reagents and antibodies. Doxycycline (Sigma-Aldrich. St. Louis, Missouri) was resuspended to 2 mg/ mL in sterile water and was used at 2 µg/mL. 6-formylpterin (Schirck's Laboratories, Bauma, Switzerland) was resuspended to 3 mg/mL in 0.01 M NaOH. Control vehicle used was 0.01 M NaOH. Antibodies utilized are listed in Table 2 Microorganisms and preparation of Apcs. M Expansion of T-cell clones. T-cell clones were cultured in the presence of X-rayed (3,000 cGray using X-RAD320, Precision X-Ray Inc.) allogeneic PBMCs, X-rayed allogeneic LCL (6,000 cGray) and anti-CD3 monoclonal antibody (20 ng ml −1 ; Orthoclone OKT3, eBioscience, San Diego, USA) in RPMI 1,640 media with 10% human serum in a T-25 upright flask in a total volume of 30 ml. The cultures were supplemented with IL-2 on  ). Then, the plate was washed three times with sterile PBS and blocked for 1 h at room temperature with RPMI 1,640 media containing 10% heat-inactivated HS pool. Then, the APCs and T cells were prepared as described above and co-incubated overnight. Antigen presenting cells and antigen used are noted in the relevant figure. T-cell clones were added at 5 × 10 3 per well. The plate was incubated overnight at 37 °C and then washed six times with PBS containing 0.05% Tween. The plate was then developed as previously described and analyzed using an AID ELISPOT reader (Strassberg, Germany) 5 .
Human subjects. All  cell lines. BEAS-2B and A549 cell lines was obtained from ATCC. A549:MR1_KO were generated as described previously 29 . All cell lines were cultured in DMEM (Gibco via ThermoFisher Scientific, Waltham, MA) supplemented with L-glutamine and 10% heat inactivated FBS. All cell lines were confirmed to be mycoplasma free.  www.nature.com/scientificreports/ Generation of lentivirus particles. Lentivirus production was performed as described previously 48 .
Briefly, lentiviruses were produced by co-transfection of HEK 293T cells with the lentiviral vectors that contained the gene of interest (pCDH-sgRNA or pCW-Cas9) and the packaging plasmids (psPAX2 and pMD2.G). Transfection was performed with Lipofectamine 3,000 (Thermo Fisher, Waltham, USA) according to manufacturer's instructions. Cells were cultured in DMEM supplemented with 10% FBS. 48 h after transfection, lentivirus-containing supernatants were harvested, centrifuged for 5 min at 1,250 rpm and filtered through a 0.45 µm filter. The Beas2B_MR1KO cell line was transduced with a lentivirus expressing pDMLV_Tet_MR1AGFP by spinoculation. Transduced cells were exposed to doxycycline overnight to promote expression of MR1AGFP and subsequently sorted on an InFlux cell sorter based on GFP expression. Cells were subsequently cultured, validated for MR1AGFP expression by flow cytometry, and utilized for further experiments.
Generation of a Beas2B MR1 knockout cell line. CRISPR/Cas9-mediated genome-editing for MR1 was performed as described previously 48 . Wildtype Beas2B cells were seeded in a 6 well plate and transduced by spinoculation with a lentivirus encoding the CRISPR/Cas9 gene with guide RNA specifically targeted to MR1 that had been previously used to generate an A549 MR1 knockout cell line 29 . Following this, limiting dilution was performed on transduced cells. Cells were grown in DMEM media containing 10% fetal bovine serum and 1% gentamicin antibiotic. Three clones were validated as knocked out for MR1 protein expression, as well as functionally unable to stimulate MAIT clones, and cryopreserved in 90% fetal bovine serum/10% DMSO. One clone was selected and utilized for all future experiments (Beas2B:MR1_KO).
Human tissue sources of antigen presenting cells. PBMCs were isolated from the peripheral blood of healthy donors using Ficoll-Paque gradients. Monocyte-derived DCs and macrophages were generated by the method of Romani 28 . PBMCs obtained by apheresis were resuspended in 2% human serum in RPMI and were allowed to adhere to a T-75 flask at 37 °C for 1 h. After gentle washing twice with PBS, nonadherent cells were removed and 10% human serum in RPMI containing 30 ng ml −1 of IL-4 (Immunex, Seattle, USA) and 30 ng ml −1 of granulocyte-macrophage colonystimulating factor (Immunex, Seattle, USA) was added to the adherent cells to generate dendritic cells. For generation of macrophages cells, adherent cells were incubated in 10% human serum in RPMI for 1 h. The cells were X-rayed with 3,000 cGray using X-RAD320 (Precision X-Ray Inc, North Branford, USA) to prevent cell division. After 5 days, cells were harvested with cell-dissociation medium (Sigma-Aldrich, Gillingham, UK) and used as APCs in assays.
Large airway epithelial cells were generated as described previously from the upper airway of human donors and cryopreserved in heat-inactivated fetal bovine serum with 10% DMSO 2 .
Thymocytes: Thymus tissue was ground in a GentleMACS dissociator with of DMEM plus 10% FBS to form a single cell suspension. The suspension was cryopreserved at 2-3 × 10 8 cells/ml in a 90% FBS/10% DMSO freezing solution with a post-thaw viability of approximately 50% RnA isolation, cDnA synthesis and qpcR analysis. Total RNA was isolated using TRIZOL (Life Technologies, Carlsbad, USA) and phenol chloroform extraction followed by an RNAeasy Kit (Qiagen, Venlo, Netherlands). cDNA was synthesized using a High Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific, Waltham, USA). qPCR was performed using SYBRGreen Power Master Mix (ThermoFisher Scientific, Waltham, USA) on a Step One Plus Real-Time PCR System (Applied Biosystems, Foster City, USA). Primers were designed to be specific for each gene assay and generated by the DNA services core and IDT Technolo- www.nature.com/scientificreports/ gies. The absolute quantification method was used using by generating a standard curve with a plasmid specific for each gene assayed. Primer sequences are noted in and plated into 1.5 mm glass bottom chamber slides (Nunc, Roskilde, Denmark; ThermoFisher Scientific, Waltham, USA) and incubated at 37° C and 5% CO 2 . After 48 h, live images were acquired on a high-resolution wide field CoreDV system (Applied Precision, Pittsburgh, Pennsylvania) with a Nikon Coolsnap ES2 HQ. For 6FP surface stabilization assay, cells were transfected as above, and administered 50 uM 6FP or an equivalent volume of 0.01 M NaOH for 16 h. Cells were fixed for 20 min in 4% PFA and stained with DAPI (ThermoFisher Scientific, Waltham, USA) to visualize nucleus before acquisition on a wide field CoreDV system. Each image was acquired as Z-stacks in a 1,024 × 1,024 format with a 60 × objective (NA 1.42).

Dynamics of MR1A and MR1B expression. pCI_MR1BRFP or pCI_MR1CRFP was transfected into
Beas2B_MR1KO cells overexpressing a Tet_MR1AGFP lentivirus either 24 h prior to or 16 h post induction of MR1A expression. Flow cytometry was performed as described earlier to measure MR1AGFP expression or MR1BRFP expression using a BD Symphony and data were analyzed using FlowJo software. For functional analyses, cells were harvested on ice, infected 1 h with a titrating dose of M. smegmatis, and utilized as antigen presenting cells (5 × 10 3 /well) in an ELISPOT assay following incubation with an MAIT clone (5 × 10 3 /well). The ELISPOT was set up and developed as described above.
flow cytometry staining and fAcS sorting. Lung and small intestine cell suspensions generated as described earlier were thawed, washed with PBS twice and blocked in FACS buffer. 2e06 cells were stained with antibodies against surface CD45, CD326 (EPCam), CD14, and CD20, for 30 min on ice in the dark, as well as Live/Dead Fix Aqua for viability. Cells were harvested on ice, washed in PBS, resuspended in FACS buffer, and sorted on a BD InFlux sorter for Live, CD45-CD14-, and CD20-CD326 + cells in order to isolate CD326 + epithelial cells. Cells were sorted into TRIZOL for immediate RNA isolation and subsequent quantitative real time PCR analysis. Sorting of MR1 + DP thymocytes: Thymocytes were thawed and blocked with FACS buffer, and 2e06 cells were stained with antibodies for the surface markers CD3, CD4, CD8, and MR1-biotin (clone 26.5), or the isotype control (IgG2a-biotin) for 30 min on ice in the dark. Cells were washed twice with cold PBS and stained with streptavidin-PE (1:2000 dilution) for 15 min in the dark. Cells were washed twice and resuspended in FACS buffer and sorted on a BD InFlux Sorter for Live, CD3 + CD4 + CD8 + MR1 expressing thymocytes, which was confirmed based on an isotype control for MR1. Cells were sorted into TRIZOL for immediate RNA isolation and subsequent quantitative real time PCR analysis. For quantifying the frequency of MAIT thymocytes, thymocytes were thawed, blocked with FACS buffer, and 2e06 cells were incubated with the MR1/5-OP-RU tetramer or the MR1/6FP tetramer (1:500 dilution, NIH Tetramer Core facility, Emory, USA), for 45 min at RT in the dark. Following this, antibodies to the surface markers CD3, CD4, and CD8 and Live/Dead Fix Aqua viability stain were added at manufacturer recommended concentrations. Cells were incubated at RT in the dark for 15 min, washed twice with PBS, and fixed with 1% paraformaldehyde. Acquisition was performed on a BD Symphony flow cytometer and data were analyzed using FlowJo software (TreeStar, Ashland, USA). All antibodies utilized are listed in Table 4 www.nature.com/scientificreports/ Gtex analysis. Figure 1 depicts the approximate ratio of exon 4 inclusions to skips in MR1 across normal tissue RNA-seq samples comprising the GTEx project 26 . To generate this figure, we first used Snaptron to query GTEx RNA-seq samples for numbers of reads spanning each of three exon-exon junctions, whose hg38 coordinates are (a) chr1:181,050,287-181,052,234, (b) chr1:181,052,511-181,053,572, and (c) chr1:181,050,287-181,053,572. Junctions (a) and (b) point to inclusion of exon 4, while junction (c) points to exclusion of exon 4. For a given sample, we averaged the numbers of reads spanning junctions (a) and (b) and divided the result by the number of reads spanning junction (c) to estimate the ratio of exon 4 inclusion to exclusion 25 . We subsequently used Mathematica 10.4 to plot results across samples. Scripts to reproduce Fig. 1 are available at https :// githu b.com/nello re/mr1. Data analysis. Data were analyzed and plotted using Prism 7 GraphPad Software (La Jolla, California).
Statistical significance was determined using unpaired Student's two-tailed t-test, unless otherwise indicated. For comparison of DP MR1 + thymocyte gene expression with frequency of DP thymocytes or of MAIT thymocytes, Pearson's correlation calculation was performed. For comparisons of functionality between transfected antigen presenting cells, linear regression analysis was performed and significant differences between slopes and y-intercepts were measured and reported. Error bars in the figures indicate the standard deviation, standard error of the mean, or the data set range as indicated in each figure legend. P values < 0.05 were considered significant (*P < 0.05; ** P < 0.01; *** P < 0.001).