Dual TCR-alpha expression on MAIT cells as a potential confounder of TCR interpretation

Mucosal-associated invariant T (MAIT) cells are innate-like T cells that are highly abundant in human blood and tissues. Most MAIT cells have an invariant T cell receptor (TCR) α chain that uses TRAV1-2 joined to TRAJ33/20/12 and recognize metabolites from bacterial riboflavin synthesis bound to the antigen-presenting molecule, MR1. Recently, our attempts to identify alternative MR1-presented antigens led to the discovery of rare MR1-restricted T cells with non-TRAV1-2 TCRs. Because altered antigen specificity is likely to lead to altered affinity for the most potent known antigen, 5-(2-oxopropylideneamino)-6-D-ribitylaminouracil (5-OP-RU), we performed bulk TCRα and β chain sequencing, and single cell-based paired TCR sequencing, on T cells that bound the MR1-5-OP-RU tetramer, but with differing intensities. Bulk sequencing showed that use of V genes other than TRAV1-2 was enriched among MR1-5-OP-RU tetramerlow cells. Whereas we initially interpreted these as diverse MR1-restricted TCRs, single cell TCR sequencing revealed that cells expressing atypical TCRα chains also co-expressed an invariant MAIT TCRα chain. Transfection of each non-TRAV1-2 TCRα chain with the TCRβ chain from the same cell demonstrated that the non-TRAV1-2 TCR did not bind the MR1-5-OP-RU tetramer. Thus, dual TCRα chain expression in human T cells and competition for the endogenous β chain explains the existence of some MR1-5-OP-RU tetramerlow T cells. The discovery of simultaneous expression of canonical and non-canonical TCRs on the same T cell means that claims of roles for non-TRAV1-2 TCR in MR1 response must be validated by TCR transfer-based confirmation of antigen specificity.


Introduction
Adaptive cellular immunity relies on recombination of the T cell receptor (TCR)-b (TRB), TCR-g (TRG), TCR-a and TCR-d (TRA/TRD) genomic loci during T cell development in the thymus 1 . Remarkable TCR diversity is achieved by combinatorial usage of genome-encoded variable (V), diversity (D), and joining (J) genes, and addition of intervening non-templated (N) nucleotides 2 . Many T cells recognize peptide antigens in the context of highly polymorphic human leukocyte antigen (HLA) molecules 3 . In parallel, some T cells bind non-peptide antigens presented by non-MHC-encoded antigen-presenting molecules, including the MHC-related protein 1 (MR1) and cluster of differentiation (CD)1 proteins (reviewed in 4,5 ). Unlike MHC, CD1 and MR1 proteins are almost monomorphic 6 , and consequently CD1-and MR1-reactive T cells tend to express characteristic TCR motifs, shared by many individuals irrespective of their HLA haplotypes 7 .
These invariant TCR motifs 7 recognize unique antigen classes, including pathogen-derived mycobacterial lipids for CD1b 8 , a-galactosyl ceramides for CD1d 9 and metabolites from active bacterial biosynthetic enzymes for MR1 10 . These invariant TCRs are thought to have co-evolved with cognate nonclassical antigen-presenting molecules in different species 11 .
Due to their potential to elicit generalizable population-level immune responses, donorunrestricted T cells (DURTs), and the antigens they recognize, are attractive targets of vaccination against microbes like Mycobacterium tuberculosis (Mtb) 12 . In particular, mucosal-associated invariant T (MAIT) cells, which recognize antigens presented by MR1, are attractive candidates due to their abundance in the blood 13 , their high reactivity against several bacterial infections 14,15,16,17 , and their documented roles in vaccination 18,19 . MR1 tetramers bind directly to TCRs and allow for unequivocal identification of MAIT cells and more diverse MR1-restricted ab 20 and gd 21 T cells, and provide a unique opportunity to identify novel TCR rearrangements and antigen specificities 22 . Human MAIT TCRa chains display a characteristic complementarity-determining region (CDR3a) formed by a rearrangement between TRAV1-2 and TRAJ33, or sometimes TRAJ12 or TRAJ20, with few non-template encoded (N)-nucleotides 22,23,24 , and a biased preference for some TRB genes 23,25,26 . Diversity in TRB gene usage in MAIT cells is potentially associated with recognition of different microbes 25,27,28,29 or different ligands 30 . These canonical MAIT cells have a preferred specificity for 5-(2-oxopropylideneamino)-6-D-ribitylaminouracil (5-OP-RU) over 6-formylpterin (6-FP) 10,31,32 . Whereas TCR conservation, especially 'canonical' TRAV1-2 usage has been considered a key defining feature of human MAIT cells for decades, a new direction in the field has resulted from identification of 'non-canonical' TRAV1-2-negative (TRAV1-2 -) and gd T cells 21 that recognize MR1 and suggested to have unique antigen specificities 20,33,34,35,36,37 . MAIT cells have broadly reported roles in infection 17 , cancer 38 , and autoimmunity 39 . Hence, defining MAIT TCR motifs can be used to infer pathogenic and protective TCR clonotypes relevant to immunodiagnosis or vaccination.
Several new technologies and algorithms for high-dimensional TCR sequencing analysis have successfully identified clonally expanded populations of antigen-specific T cells, and their TCR motifs among large numbers of blood-and tissue-derived T-cells 40,41,42,43 . These sequencing technologies derive TCR sequences either from single cells, which identify paired TCRa and TCRb 44, 45 , or bulk genomic 46 or transcriptomic sequencing data 41,47 . In this study, we sought to use MR1 tetramers and high throughput TCR sequencing to identify non-canonical TCR patterns. We observed MAIT cell populations with differing binding intensities to the 5-OP-RUloaded MR1 tetramers. We hypothesized that MAIT cells with lower MR1-tetramer binding intensities would reveal unique TCR motifs consistent with lower preferential binding to the 5-OP-RU/MR1 antigen complex. Consistently, we detected an enrichment of TRAV1-2 -TCRs in MR1 tetramer + MAIT cells, especially those with lower MR1-tetramer intensity. However, detailed TCR gene transfer studies revealed that the lower tetramer binding was explained by dual expression of canonical and non-canonical TCRa chains in the same TRAV1-2 + clonally expanded MAIT cells, as opposed to a single non-canonical TCR with lower affinity for MR1-5-OP-RU. Dual TCR expression previously observed in HLA-restricted 48 and CD1d-reactive T cells 49 , but takes on special importance in the MAIT cell system because it can confound the assignment of noncanonical TCRs for MR1 specificity. These data establish the need to validate the antigen specificity of newly-described TCR motifs from large-dimensional sequencing platforms by TCR gene transfer and other alternative techniques 50 .

Human participants
Lima, Peru: We recruited Peruvian participants with active TB disease, or asymptomatic household contacts of TB cases with positive or negative QuantiFERON TB Gold-In-tube test results from Lima, Peru, as described previously 51,52 . Melbourne, Australia: Spleen (SP) lymphoid tissues were collected from deceased donors, whose mortality was caused by conditions other than influenza (DonateLife, Australia), after written

Flow cytometry analysis
The protocol and primary analysis of Peruvian samples by flow cytometry was reported previously 51 . MR1 monomers were obtained from The University of Melbourne, Australia 10,22 , and used to generate tetramers in Boston as previously described 51 . For HEK293T cell validation experiments, we used MR1 tetramers obtained from the National Institutes of Health (NIH) Tetramer Core facility.

Genomic bulk TCR sequencing (Adaptive Biotechnologies, Seattle)
For TCR sequencing from genomic templates, 3900 MR1 tetramer hi and 4500 MR1 tetramer int cells were doubly sorted from PBMC samples from Peruvian donor 58-1 after 14 days of polyclonal T cell expansion. For expansion, 10 6 cells were cultured with 25 × 10 6 irradiated allogeneic PBMC, 5 × 10 6 irradiated allogeneic Epstein Barr Virus transformed B cells, 30 ng/ml anti-CD3 monoclonal antibody (clone OKT3) for 14-16 days, in the presence of 1 ng/ml interleukin-2 (IL-2) 52 . PBMC samples from healthy Boston blood bank donors LP1 and CO2 were not expanded before double cell sorting. Cell numbers obtained from the sorted tetramer hi , tetramer int , and tetramer low populations were 2000, 5800, 3100, respectively, for LP1 and 1100, 4000, and 2300, respectively, for CO2. High-throughput TCR sequencing and assignment of V and J genes was performed for the TCRb locus and the TCRad locus (Adaptive Biotechnologies, Seattle, WA) using a multiplex PCR approach on genomic DNA isolated from sorted T cells using the Qiagen QIAamp DNA Mini Kit, followed by Illumina high-throughput sequencing 46 .

Sorted single cell paired TCR sequencing
Single-cell TCR sequencing was adapted from a previously published protocol 41 . Briefly, single MR1-tetramer-binding cells from Peruvian participant 7-3 and blood bank donors 702A and 703A were sorted into 96-well plate coated with Vapor-Lock (Qiagen) containing Iscript cDNA synthesis mixture (Bio-Rad) and 0.1% triton X-100 for direct cell lysis. Reverse transcription was performed in a thermocycler (25°C for 5', 42°C for 30', 80°C for 5'). Subsequently, cDNA samples were amplified in a nested PCR reaction using Denville Choice Taq Polymerase (Thomas Scientific), using previously described primers 41 . Briefly, the first external reaction contained a mixture of all TCRa and TCRb forward primers, combined at 1 µM each, and reverse TRAC and TRBC primers at 10 µM each: 95°C for 2', 35 cycles of (95°C for 20'', 50°C for 20'', 72°C for 45''), and 72°C for 7'. A second internal PCR reaction used a mix of TCRa forward primers at 1 µM each with a reverse internal TRAC primer at 10 µM, or a mix of TCRb forward primers and reverse TRBC primer, separately at cycling conditions: 95°C for 2', 35 cycles of (95°C for 20'', 56°C for 20'', 72°C for 45''), and 72°C for 7' using previously described primers 41 . Amplicons were analyzed on an agarose gel, and bands were excised using a UV lamp and purified using the QIAquick Gel Extraction Kit (Qiagen) then sent for Sanger sequencing (Genewiz). Sequences were reversecomplemented and analyzed using 4Peaks software and mapped to the reference sequences for the genome-encoded V and J segments for both the TCRa and TCRb genes on the ImMunoGeneTics (IMGT) information system database. The unmapped sequences were considered N-nucleotides, and/or Db segments for TCRb to determine the complementaritydetermining region (CDR)-3. CDR3a and CDR3b amino acid sequences were predicted by in silico translation, showing productive in-frame rearrangements, using the online ExPASy translate tool (https://web.expasy.org/translate/). For Australian samples, single MR1-5-OP-RU-tetramer + TRAV1-2 + PBMCs from healthy donors and spleen tissues were sorted using a FACSAria (BD Biosciences) into 96-well plates. Paired CDR3ab regions were determined using multiplex-nested reverse transcriptase PCR before sequencing of TCRa and TCRb products, as previously described 41,54 , and reported 55 . For paired TCRab analyses, sequences were parsed into the IMGT/HighV-QUEST web-based tool using   Table   1). Frequencies of TRAV1-2 -MAIT cells in blood did not differ by TB status in Peruvian samples (Kruskal-Wallis: p=0.75; Figure 1D). TRAV1-2 -MAIT in these samples ( Figure 1B-C) were similar to frequencies previously reported in other populations 20 representing a minority of T cells (0.6-40%) but they were potentially biologically significant because TCRa diversity diverges from the conventional understanding of MAIT cell function.

During a quantitative study of MAIT cells in a Peruvian
We sought to explain the discrepancy between the low frequencies of TRAV1-2 -MAIT cells as determined by flow cytometry (Figure 1D), and the higher frequencies of TRAV1-2 -TCR a chain sequences identified in sorted MAIT cells, as determined by bulk TCR sequencing ( Figure   1C). Hence, we sorted single cells from populations with different MR1-tetramer binding levels from one Peruvian participant, where we detected three clear MR1-tetramer binding levels (MR1tetramer high , MR1-tetramer int , and MR1-tetramer low ), and applied a previously described nested  (Figure 4, left).
To explain the lack of binding between these TRAV1-2 -TCRs and5-OP-RU-loaded MR1, we took a closer look at the TCRb sequences. Unexpectedly, a single TCRb sequence consisting of TRBV24-1-TRBJ2-5 with a unique CDR3 nucleotide sequence was detected in 10 out of the 15 TRAV16 + single cells (Supplementary Table 3). Interestingly, the same TCRb nucleotide sequence (TRBV24-1-TRBJ2-5) was paired with the canonical MAIT TCRa TRAV1-2-TRAJ33 in 3 wells (Supplementary Table 3). Because the PCR reactions were performed in multiplex format, we hypothesized that this particular T cell clone expressed two different, functional TCRa chains, but that only one of the PCR products dominated the PCR reaction. Hence, to resolve the discrepancy, we re-amplified the templates that initially gave rise to a TRAV16-TRAJ11 PCR products, using only the TRAV1-specific forward primer, which captures the TCRa variable genes TRAV1-1 and TRAV1-2 only, as previously described 41 . Using this approach, 10 out of the 15 templates initially giving rise to TRAV16-TRAJ11 sequences now gave rise to a PCR product that resulted in identical TRAV1-2-TRAJ33 sequences and paired with the same TRBV24-1-TRBJ2-5 TCRb (Supplementary Table 3). Whereas we initially interpreted these results as non-canonical TCRs binding to MR1, the data were more consistent with clonal expansion of a T cell coexpressing one TCR b chain, a TRAV1-2 + invariant MAIT TCR a chain, and an additional, noncanonical TCRa chain. If only the canonical TCRa chain binds MR1, the lower tetramer binding of these TCRs could be caused by competition of two different TCRa chains with the same TCRb chain (TRBV24-1-TRBJ2-5), analogous to what has been described for NKT cells 49 .
Finally, to reproduce our finding of dual TCRa expression on MAIT cells in an independent experiment, we analyzed paired TCR sequences in MR1-tetramer-binding cells from different blood donors 55 . Although in this experiment we sorted all MR1 tetramer-binding T cells, including the MR1-tetramer high ones, we identified cells that co-expressed the canonical invariant TRAV1-2 + TCR a chain with a TRAV1-2a chain in PBMC samples from donors of different ages, as well as healthy spleen tissues of deceased donors ( Figure 5). Collectively, our study suggests that dual-TCRa expression is common among MR1-tetramer-binding MAIT cells in different human populations, tissue types and disease states.

Discussion:
In this study, we hypothesized that TCRs with decreased affinity for MR1-5-OP-RU would reveal new TCR motifs that may prefer MR1 ligands other than 5-OP-RU or correlate with TB disease. Our hypothesis was motivated by the reported expansion of diverse MAIT cell clonotypes following Salmonella challenge of humans in individuals who progress to disease 57 , and the discovery of new antigen classes derived from the related mycobacterium M. smegmatis 28 .
However, our search for new TCR motifs based on differential binding to the 5-OP-RU-loaded MR1 tetramer was confounded by the co-expression of two TCRa chains in the same T cell. The phenomenon of dual TCRa co-expression has been previously described for MHC-restricted 58,59 and CD1d-restricted 49 T cell subsets. Unlike the TCRb locus, the TCRa counterpart is not subject to strict allelic exclusion, so dual TCRa expression is more common 60,61 . TCRa recombination is also known to occur simultaneously on both alleles to maximize productive TCRab recombination and diversity in the TCR repertoire 62 .
The simplest explanation for the lower MR1-tetramer staining, which is also supported by these reports of dual TCRa chains in other systems, is that the canonical MAIT TCR binds to MR1, but the competition of the two TCRa chains to pair with the same pool of available TCRb chains reduces the MR1-tetramer-binding intensity by reducing functional TCR expression on the cell surface. Hence, the hypothesis that these TCRs displayed preferential affinity to different MR1 antigens was not supported by the data. Importantly, our data point to a potentially common artifact in interpreting TCRa sequences, particularly from high-dimensional sequencing data 63 .