Mtb HLA-E-tetramer-sorted CD8+ T cells have a diverse TCR repertoire

Summary HLA-E molecules can present self- and pathogen-derived peptides to both natural killer (NK) cells and T cells. T cells that recognize HLA-E peptides via their T cell receptor (TCR) are termed donor-unrestricted T cells due to restricted allelic variation of HLA-E. The composition and repertoire of HLA-E TCRs is not known so far. We performed TCR sequencing on CD8+ T cells from 21 individuals recognizing HLA-E tetramers (TMs) folded with two Mtb-HLA-E-restricted peptides. We sorted HLA-E Mtb TM+ and TM− CD8+ T cells directly ex vivo and performed bulk RNA-sequencing and single-cell TCR sequencing. The identified TCR repertoire was diverse and showed no conservation between and within individuals. TCRs selected from our single-cell TCR sequencing data could be activated upon HLA-E/peptide stimulation, although not robust, reflecting potentially weak interactions between HLA-E peptide complexes and TCRs. Thus, HLA-E-Mtb-specific T cells have a highly diverse TCR repertoire.


ll OPEN ACCESS
HLA-E is further conserved among vertebrate species with orthologues found in mice (Qa-1 b ), rhesus macaques (RMs) (Mamu-E), and cynomolgus macaques (Mafa-E), facilitating rapid translation of pre-clinal research to humans. 191][22][23] Vaccination with an SIV/rhesus CMV strain 68-1 vector expressing the SIVgag protein resulted in broad and effective Mamu-E CD8 + T cell responses in nearly all vaccinated RMs, and Qa-1 b T cell responses in Mtb-infected mice were essential to control and protect against Mtb.Altogether, these studies suggest that HLA-E is a promising vaccination target for Mtb and other pathogens to induce protection across a genetically diverse population.
Recently, the improvement of predictive HLA-E peptide-binding algorithms enabled the identification of high-affinity Mtb peptides. 17ext to peptide identification, we investigated the HLA-E-Mtb-restricted T cell repertoire in more detail to deepen our current understanding on the diversity of HLA-E T cells in general and in TB specifically.T cells that recognize monomorphic molecules are called donor unrestricted T cells (DURTs) and generally express invariant TCRs that are possibly conserved across individuals. 24,25Previous studies on cytolytic T cell clones recognizing conserved HLA-E-restricted UL40-derived epitopes from CMV in four donors revealed a nearly monoclonal expansion in the TRBV region, which was different between donors. 15,26However, more recently, TCR profiling on 13 HLA-E-restricted T cell clones from one donor recognizing the HLA-E-restricted HIV-derived peptide RL9HIV and on 13 HLA-E-restricted T cell clones from two donors specific for Sars-CoV-2-derived HLA-E peptide binders revealed a more diverse TCR repertoire with variations between clones in the usage and pairing of variable (V) and joining (J) fragments in the TCRa and b chains. 27,28Thus, the question remains if HLA-E-restricted TCRs can be considered invariant, like other DURTs, or are diverse.
In this study, we performed TCR repertoire as well as RNA-sequencing analysis on CD8 + T cells sorted with HLA-E TMs folded with two Mtb-derived HLA-E-specific peptides in 21 South African individuals. 16We show that the TCR repertoire of Mtb HLA-E TM + T cells is diverse across and within individuals, with no indications for conservation of TCRs between individuals.We further show that HLA-E Mtb TM + TCRs selected using GLIPH2 could be activated after stimulation with the HLA-E Mtb peptides in both Jurkat cells and primary T cells using Zap70 phosphorylation as a readout, although with relatively low signals compared with classical HLA-Irestricted T cells.

RESULTS
HLA-E Mtb TM + CD8 + -sorted TCRs contain heterogeneous variable and joining a and b fragments To unravel the TCR repertoire of HLA-E-restricted Mtb-specific CD8 + T cells, we stained PBMCs from 21 South African individuals with HLA-E*01:03 tetramers (TMs) folded with two Mtb-derived peptides that are established binders to HLA-E*01:03, called p34 and p55.The gating strategies for sorting 1,000-2,000 HLA-E TM + CD8 + T cells are shown in Figure S1A.We validated correct folding of the HLA-E*01:03 monomers with HPLC and tetramerization and integrity of the monomers using LILRB1 staining. 18,29In our staining and sorting approach, we always combined two unrelated HLA-E TMs (e.g., p34 and p55 or p34 and another published Mtb-specific HLA-E-restricted peptide 16 ) with different fluorochromes in the same staining to control for possible a-specific staining of the HLA-E TMs.Double-positive cells were considered a-specific and only single-positive cells were sorted (Figures S1B and S1C; Table S1 for TM/peptide combinations).HLA-E TM + cells were either sorted in bulk or as single cells and were used for downstream RNA-seq (bulk) and TCR-seq (single cells) analysis.We initially wanted to compare the sequencing data of p34 and p55 HLA-E TM + T cells with the well described and characterized HLA-E-restricted UL40-derived epitope from CMV; however, only two individuals in our cohort showed detectable HLA-E UL40 TM staining, which was the reason we compared the sorted p34 and p55 HLA-E TM + T cells with the total CD8 + T cell pool (i.e., CD8 + T cells not sorted based on HLA-E TM staining for these two Mtb epitopes, called CD8 + TM À T cells).
TCRab fragments were extracted from bulk RNA-seq data and from single-cell sequencing data with the algorithm MiXCR.MiXCR identifies single Va, Vb, Ja, Jb, and CDR3ab fragments from sequence data via sequence alignments.Bulk RNA-seq on CD8 + TM À T cells from 17 donors combined revealed a heterogeneous TCR repertoire, as expected (Figure 1A, left; Table S2).Both Va, Vb and Ja, Jb fragments were diverse in these CD8 + TM À T cells as reflected by the low frequencies of single Va and Vb fragments (data not shown).Bulk RNA-seq on p55 HLA-E TM + -sorted T cells from 16 donors combined revealed a similarly diverse Va and Vb repertoire (Figure 1A, left).Likewise, bulk RNA-seq analysis per donor on p34 and p55 HLA-E TM + -sorted T cells revealed a diverse TCR repertoire as well albeit with a certain degree of preference per donor for certain individual Va or Vb fragments, as shown for the two representative donors in Figure 1A, right panel.Although only two donors had sufficient UL40 HLA-E TM + -sorted T cells, the Va and Vb fragments were more homogeneous/conserved, particularly in the Vb fragments (Figure 1A, right).In addition, there was a different Vb fragment preferentially used between the two donors: TRBV20-1 in donor 4 and TRBV5-1 and -14 in donor 21.
In addition to Vab fragment identification from the bulk RNA-seq data, we also assessed the conservation of the CDR3a and b region via clonotype identification using MiXCR on the bulk RNA-seq data (Figure 1B).The majority of CDR3a and b sequences in the CD8 + TM À T cell population (17 donors), including the p34 (2 donors) and p55 (16 donors) HLA-E TM + T cell populations, were present with a copy number of 1-5, suggesting that the clonality of the CDR3 sequences was low, indicating diversity in these populations (Figure 1B).In addition, the frequency of possibly clonally expanded (copy number of 10 or more) CDR3a and b sequences was low as well in these populations for each donor (Figure 1C).However, compared with p34 and p55 HLA-E TM + -sorted TCRs, UL40 HLA-E TM + -sorted TCRs had a higher frequency of CDR3ab sequences with a copy number of 10 or more, suggesting a more clonally expanded T cell population (Figure 1C).As such, the heterogeneity in the CDR3ab and Va, b regions of Mtb HLA-E TM + TCRs, but not in UL40 HLA-E TM + TCRs, suggests that HLA-E Mtb TM + TCRs are more diverse as determined from the RNA-seq data.Besides the identification of TCRab V and J fragments from the bulk RNA-seq data, we also looked at the conservation of paired V and J fragments using the sequencing data on single-cell-sorted TM + T cells for p34 and p55 and CD8 + TM À T cells.Analysis at the single-cell level does not only permit analysis of the V-J fragment usage but can also access a and b pairing and will permit selection of TCRs for downstream analysis.Varying per donor, between 4 and 64 V-J pairs were identified in the single-cell dataset, which indicates that multiple different V-J pairs can bind to the same HLA-E/peptide complex.The V-J pairing was diverse for CD8 + TM À T cells as well.Examples of two individual donors are shown in Figure 2A.The median CDR3 length in all HLA-E p55 and p34 TM + -sorted T cells was 41 nucleotides (1.0 SD) in the TCRa chain, whereas the CDR3b was slightly longer with a median length of 44 nucleotides (1,4 SD) (Figure 2B).The V(D)J-insert in the TCRb chain in the single-cell-sorted TM + T cells was also longer (median 12 G 1.5 SD) than in the TCRa chain (median 4 G 0.6 SD).Paired V-J analysis on the sorted single HLA-E p55 and p34 TM + T cells per donor also revealed that the number of unique V-J pairs for both chains increased proportionally, indicating equal variation in the a and b chains (Figure 2C).HLA-E allele expression (i.e., *01:01 homozygous or *0101:*0103 heterozygous) did not influence aor b-chain variability, indicating no effect of genotype.Duplicate V-J pairings showed slight variations for some donors, which suggests that some V-J combinations are more preferred than others (Figure 2C).The number of CDR3 clonotypes with multiple copies was calculated per donor (Figure 2D) and most TCRa-and b-CDR3s were present as a single copy, suggesting diversity of V(D)J pairs in HLA-E p34 and p55 TM + TCRs within and between donors, including for CD8 + TM À TCRs.Similar clonotype analysis and comparisons could not be performed on single-cell-sorted HLA-E UL40 TM + T cells because of the limited number of donors with a detectable TM + population.Large variation in the number and usage of Vab and Jab fragments in single-cell-sorted HLA-E Mtb TM + TCRs Figure 3A shows the copy number of single Va, Vb and Ja, Jb fragments in 17 donors extracted from the single-cell data.Stacked bars show the diversity as well as inter-donor variations in the number and usage of V and J fragments.To further substantiate TCR diversity quantitatively instead of numerically, we calculated the Gini-Simpson index as well as the inverse Simpson index on the single-cell data and RNA-seq data combined.The Gini-Simpson index indicates the probability that two randomly selected TCRs from the same population have a different clonotype.The index value can be between 0 and 1, where 1 represents a TCR repertoire of infinite diversity (i.e., polyclonal repertoire) and 0 indicates a monoclonal repertoire.CD8 + TM À , p34 TM + , and p55 TM + TCRs had a Gini-Simpson index close to 1, indicating a more polyclonal TCR repertoire, whereas UL40 TM + TCRs had a lower index (0.73), suggesting a relatively more monoclonal population (Figure 3B).The inverse Simpson index represents the inverse fraction of a particular clonal type relative to the total number of identified clonal types.An increasingly higher number represents a polyclonal repertoire, as shown for the total CD8 + TM À T cells (Figure 3C).Compared with the total CD8 + TM À T cells, p34 and p55 TM + T cells had a lower index, suggesting a less polyclonal population compared with CD8 + TM À T cells but still far from monoclonal.However, UL40 TM + TCRs had a low index, close to 0, indicating an almost monoclonal population, which confirms the low number of TCR clonotypes detected in the bulk RNA-seq data for UL40 TM + TCRs in the two donors (Figures 1A-1C).

HLA-E Mtb TM + CD8 + T cells have a different gene expression profile compared with CD8 + TM À T cells
Next to extraction of TCR sequences from the bulk RNA-seq data, we also analyzed and compared the gene expression profile of HLA-E p55 TM + CD8 + T cells with CD8 + TM À T cells.We did not include HLA-E p34 TM + CD8 + T cells because we could only sort p34 TM + CD8 + T cells from two donors.This comparison showed that 276 genes were differentially expressed (DE), of which only one gene, i.e., PLXNB1, was significantly downregulated on p55 TM + CD8 + T cells (Figures S2A and S2B).Plexin B1 is a receptor for Semaphorin 4. Signaling via Plexin B1 activates cell migration of neuronal, epithelial, and tumor cells, including angiogenesis.The function of PLXNB1 on T cells has not been unraveled yet.Principal-component analysis on all genes further substantiated the differences between total CD8 + TM À and p55 HLA-E TM + T cells and confirmed that the two T cell populations can be discriminated based on the gene expression profile (Figure S4A).Gene Ontology (GO) enrichment analysis revealed that the functional phenotype of p55 HLA-E TM + T cells was also different compared to bulk CD8 + TM À T cells.The top hits from the GO enrichment analysis were neutrophil activation, positive regulation of cytokine production + peptidyl-tyrosine phosphorylation, and NF-kB activation (Figure S3).Apart from neutrophil activation, these enriched functions point toward a metabolically activated phenotype of the p55 HLA-E TM + T cells, although we cannot deduce their differentiation phenotype.
The top 25 genes with the highest Log2FoldChange differential expression between total CD8 + TM À T cells and p55 HLA-E TM + CD8 + T cells were analyzed further to identify if the proteins were specific for Mtb HLA-E TM + CD8 + T cells (Figure S2C).These top 25 genes are involved in processes related to the Golgi apparatus and endoplasmic reticulum (ER), signaling transduction pathways (e.g., transforming growth factor b [TGF-b] and Wnt signaling), metabolic processes (e.g., ubiquitin ligases), and cell growth.RNA-seq read counts of these 25 genes were increased in the p55 HLA-E TM + -sorted population relative to the total CD8 + TM À T cell pool (Figure S2C).We next validated if the proteins were also upregulated on p55 HLA-E TM + T cells using cell surface staining in combination with TM staining.We selected genes for this validation if the gene was expressed in five donors or more, was an established cell surface protein, and had an average read count above 50.Eight of twenty-five genes fulfilled the criteria for protein-level validation but due to unavailability of flow cytometry antibodies for STAB1 and AQP9, only CD300b, TLR4, CCR2, THBD, FZD1, and WLS could be validated on 11 independent donors from the same TB endemic cohort (which were not included in the sequencing data) (Figure S4B).CD300b, TLR4, CCR2, THBD, and FZD1 were, in agreement with the RNA-seq data, significantly more expressed on the cell surface of p55 HLA-E TM + -sorted T cells, whereas only WLS, a carrier protein involved in Wnt signaling, was not (Figure S4C).

Identification of TCRs with high peptide specificity using GLIPH2 clustering
We next wanted to functionally validate HLA-E TM + TCRs that were common in the single-cell TCR-seq data.To identify these common HLA-E TM + TCR sequences, we applied the GLIPH2 cluster 30,31 algorithm on TCR sequences identified from the single-cell TCR-seq data and bulk RNA-seq data combined.GLIPH2 clusters TCRb sequences based on short similarity regions (enriched motifs) in the CDR3b region that are determined from the total TCRb dataset via sequence alignment.The total input dataset included 3,211 TCRb sequences.These sequences included TCRbs that could bind to HLA-E p34 and p55 TMs, and CD8 + TM À TCRbs.Initially, 527 GLIPH2 clusters containing 1,489 TCRb sequences were identified, which were filtered to exclude clusters containing sequences from less than three donors, resulting in 179 GLIPH2 clusters (Figure 4A).From these clusters, we selected clusters that contained TCRb sequences with a higher p34/p55 TM specificity compared to bulk CD8 + TM À TCRs.This selection resulted in 41 GLIPH2 clusters (Figure 4B).These clusters were then further separated for specificity to p34 or p55 TMs, which revealed that these clusters could be grouped into 11 categories (Figure 4B).We ultimately selected TCRb sequences from these 11 categories using the following approach: (1) We grouped all CDR3b sequences per category from the bulk and single-cell TCR-seq data that have the GLIPH2 motifs shown in Table 1.(2) We selected CDR3b sequences with more than 1 GLIPH2 motif.
(3) We selected CDR3b sequences that were derived from the single-cell TCR-seq dataset to find the pairing TCRa.(4) We selected CDR3b sequence with restricted VJ gene variation.
(5) If possible, we preferred CDR3b sequences that were found in multiple donors or that had GLIPH2 motifs from >1 categories (i.e., the CDR3b sequence was found in multiple categories).( 6) We only selected CDR3b sequences from categories that had the highest percentage specificity for either p34 or p55 found in our analysis to prevent the inclusion of possibly peptide a-specific TCRs.This approach ultimately led to the selection of 16 TCRb and paired TCRa sequences from categories 1, 2, 3, 4, 8, 9, and 11, with a balanced distribution between p34 and p55 specificity (Tables 2A and 2B).This table also shows which GLIPH2 motifs are present in the selected TCRbs, indicated by the category number in the first column.For instance, TCR 13 contains all motifs present in categories 1 and 2.

Functional validation of GLIPH2-selected TCRab sequences
The 16 TCRab sequences shown in Tables 2A and 2B were transduced into Jurkat cells that did not express endogenous TCR sequences or into isolated primary CD3 + T cells derived from three healthy buffy coats. 32HLA-E-restricted T cells have an unorthodox cytokine profile, with the capacity to secrete both Th1 and Th2 cytokines.Functional readout methods based on cytokine secretion, such as ELISpot or ICS that are commonly used for classical HLA-I T cells, are difficult, because of the overall low cytokine levels and the variation in the specific cytokines secreted by specific HLA-E T cells.We therefore used Zap70 phosphorylation as a readout for early TCR activation, which is a cytokine-independent method.Based on the transduction efficiencies, we selected 4 p34 HLA-E TM + -sorted TCRs and 4 p55 HLA-E TM +sorted TCRs transduced in Jurkats and 3 p34 HLA-E TM + -sorted TCRs and 1 p55 HLA-E TM + -sorted TCR transduced in primary CD8 + T cells (after gating on the CD8 + T cell population within the transduced CD3 + T cell population) to validate their peptide specificity via measuring Zap70 activation after stimulation with peptide-loaded K562 cells expressing HLA-E*01:03 or with HLA-E*01:03 TMs loaded with Mtb peptides (for Jurkat cells only).Zap70 is activated early after TCR signaling via phosphorylation of tyrosine (Y) at positions 292 and 319, among other sites, referred to as Y292 and Y319.In one of our previous studies, Zap70 phosphorylation was a good indicator for activation of HLA-E-specific T cell clones. 12We determined the level of phosphorylation at Y292 and Y319 and calculated the percentage of peptide-specific phosphorylation at Y292 and Y319 relative to the mean fluorescent intensity (MFI) of the control peptide (p44) and, in the case of K562 stimulation, unloaded condition.The left panels in Figures 5A-5C show representative examples of the MFI of phosphorylated Y292 and Y319 for the peptide-specific condition and the control conditions in TCR-transduced Jurkat cells and TCR-transduced primary CD8 + T cells after stimulation with peptide-loaded K562 cells or HLA-E TMs (Jurkat cells only).Representative gating strategies are shown in Figure S5 for both primary CD8 + T cells and Jurkat cells.We selected the highest phosphorylation site per TCR and per independent experiment/donor after peptide stimulation and summarized these results in Figures 5A-5C right panels for both transduced Jurkat cells and primary CD8 + T cells per stimulus.The Mtb peptide that was used for sorting T cells is shown below the heatmaps.These results show that most TCRs could be activated in an HLA-E/peptide-specific manner with one of the stimuli in both Jurkat cells and primary CD8 + T cells.The shift in Zap70 phosphorylation was not very strong, likely reflecting weak interaction between the TCR and the HLA-E peptide complex.There was inter-experimental variation as well as variation between TCRs in the maximal signal of Zap70 (A) p34 TCR sequences.(B) p55 TCR sequences.TCR 8 specific for p55 was selected based on the high clonal number and not on the GLIPH2 results.
activation after peptide specific stimulation but overall TCRs showed repeatedly a positive signal above background.Most Mtb HLA-E TM + -sorted TCRs, however, had a comparable signaling capacity as the published HLA-E-restricted TCR KK50.4,which recognizes the UL40-derived peptide with high affinity, at least for primary CD8 + T cells (Figure 5C, right). 33lthough the transduction efficiency of the TCRs in primary CD8 + T cells was similar between donors, there was variation between donors in the percentage phosphorylation that could be achieved upon peptide stimulation for the same TCR suggesting a donor-dependent effect (e.g., TCR 8 and 18).Likewise, the percentage phosphorylation upon peptide stimulation in TCR-transduced Jurkat cells showed variation between experiments and between stimuli (Figures 5A and 5B).In general, HLA-E TMs induced a higher level of phosphorylation compared with peptide-loaded K562 cells expressing HLA-E.Nevertheless, the overall percentage of phosphorylation upon peptide stimulation was higher in primary CD8 + T cells than in Jurkat cells.

DISCUSSION
We describe, what is to our knowledge, the first elaborate in-depth TCR repertoire analysis of CD8 + T cells binding to two Mtb peptides presented in HLA-E*01:03.Our TCR sequencing results reveal that TCRs sorted based on binding to HLA-E TMs loaded with p55 and p34 have a heterogeneous repertoire within and between individuals.Several V and J gene segments in both TCRa and b chains were identified on sorted HLA-E TM + p55 or p34 TCRs.TCRs showed limited clonal expansions, as confirmed quantitively with a close-to-1 Gini-Simpson index, revealing a TCR Vab profile that was almost comparable with that of bulk CD8 + TM À T cells.However, the number of Mtb-peptide-specific (functional) TCRs is potentially overestimated as TM staining was used to identify Mtb HLA-E-specific TCRs.This was also found previously and might be an inherent problem of using TMs to identify peptide-specific functional TCRs. 34Despite this limitation, our TCR sequencing results suggest that HLA-E TCRs are less conserved compared to previously described other DURTs. 35e also showed that HLA-E Mtb TM + CD8 + -sorted T cells have a distinctive gene expression profile compared to bulk CD8 + TM À T cells, showing upregulation of all but one of the 275 differentially expressed genes in HLA-E Mtb TM + CD8 + T cells, suggesting a more (transcriptionally) activated state of HLA-E Mtb TM + CD8 + T cells compared to bulk CD8 + TM À T cells.Differences between bulk CD8 + TM À T cells and Mtb HLA-E TM + -sorted CD8 + T cells were also found previously in cellular assays, showing that TM + CD8 + T cells had a higher production of interferon gamma (IFN-g), interleukin-10 (IL-10), IL-4, and granulysin, suggesting that Mtb HLA-E TM + T cells are transcriptionally and functionally distinct compared to bulk CD8 + TM À T cells. 13The activated state of Mtb HLA-E TM + T cells compared to bulk CD8 + TM À T cells could have been induced upon TM binding.However, the read count of several genes known to be upregulated upon T cell activation (e.g., granulysin, IFN-g, IL-2) were not increased in the sorted HLA-E TM + population compared to the sorted HLA-E TM À population.
The large variation of TCRs binding HLA-E/Mtb TM complexes is remarkable, suggesting that HLA-E-restricted T cells are more similar to conventional MHC-restricted T cells compared to DURTs in terms of their TCR profile. 24The diversity of classical TCRs recognizing Mtb was illustrated before in two recent studies that performed TCR repertoire analysis on sorted activated (i.e., CD154-and CD69-positive) CD4 + T cells obtained from PBMCs of Mtb-infected individuals briefly stimulated with Mtb lysate. 30,36Similar to our results on Mtb TM-sorted CD8 + HLA-E T cells, GLIPH2 analysis in both studies identified various unique TCRb sequences that could be clustered into >100 unique TCR specificity groups based on shared CDR3b motifs identified by GLIPH2.Unlike our approach, these studies used GLIPH2 to determine the cognate peptide and HLA-II allele restriction of the identified TCR clusters to curate their approach as Mtb lysate contains many epitopes.The TCR repertoire of these conventional CD4 + T cells recognizing various HLA-II alleles seems to be comparable with the diversity we have identified for CD8 + HLA-E Mtb TM + T cells.In contrast, a recent study revealed a restricted TCR repertoire and invariant TCRa usage in MAIT cells, which also recognize a monomorphic antigen presentation molecule and are well known as DURTs. 35These MAIT cells were sorted from PBMCs stimulated with Mtb lysate on the basis of CD161 expression. 35These studies, suggest that Mtb can be recognized by a potentially Figure 5. Zap70 phosphorylation of TCR-transduced Jurkat cells and primary CD8 + T cells sorted with TMs folded with p34 or p55 (A) Left: Representative MFI data after stimulating TCR 20 transduced Jurkat cells with HLA-E TMs.Graph shows the control conditions, including the heterologous peptide, in grey and peptide specific condition in black.Right: Percentage peptide specific Zap70 phosphorylation of 8 TCRs in Jurkat cells; 4 sorted with p34 TMs and 4 sorted with p55 TMs, after stimulation with HLA-E TMs loaded with control and specific peptides.Left: Representative MFI data after stimulating TCR 20 transduced Jurkat cells with HLA-E TMs.Graph shows the control conditions, including the heterologous peptide, in grey and peptide specific condition in black.Right: Percentage peptide specific Zap70 phosphorylation of 8 TCRs in Jurkat cells; 4 sorted with p34 TMs and 4 sorted with p55 TMs, after stimulation with HLA-E TMs loaded with control and specific peptides.(B) Left: Representative MFI data after stimulating TCR 13 transduced Jurkat cells with peptide-loaded K562 HLA-E cells.Graph shows the control conditions, including the heterologous peptide, in grey and peptide specific condition in black.Right: Percentage peptide specific Zap70 phosphorylation of 8 TCRs in Jurkat cells; 4 sorted with p34 TMs and 4 sorted with p55 TMs, after stimulation with K562 HLA-E cells loaded with control and specific peptides.(C) Left: Representative MFI data after stimulating TCR 18 transduced primary CD8 + T cells with peptide-loaded K562 HLA-E cells.Graph shows the control conditions in grey and peptide specific condition in black.Peptide specific phosphorylation in primary CD8 + T cells relative to the control peptide was determined via gating on CD8 + T cells within the CD3 + T cell population.Right: Percentage peptide specific Zap70 phosphorylation in transduced primary CD3 + T cells derived from 3 donors after stimulation with K562 HLA-E cells loaded with control and specific peptides.Each box in A-C represents an independent experiment for each TCR as indicated on the X-axis.The percentage representing each box was the highest peptide-specific phosphorylation site for that experiment and TCR.The independent experiment numbers on the Y-axis do not correlate and should be considered as individual experiments.A cross implies that there was no data for this TCR due to unsuccessful transduction into primary CD3 + T cells or because the TCR Jurkat line was not tested for that repetition (n).
diverse TCR repertoire in the context of HLA-E and further suggest that HLA-E T cells are more similar to conventional T cells than other DURTs.
Similar to our findings for Mtb, a diverse TCR profile was identified before for 13 HLA-E TCRs recognizing the HIV-derived RL9HIV peptide presented in HLA-E, which were obtained from one donor and 13 HLA-E TCRs in two donors recognizing Sars-CoV-2 peptides. 27,28These TCRs revealed diverse V-J pairs in both a and b chains, including heterogeneity in the CDR3 region.In contrast, we observed a more clonally expanded TCR population in two donors, particularly in the TRBV region, within the CD8 + T cell population recognizing the UL40-derived HLA-E-specific peptide from CMV.In one donor, the TRBV20-1 fragment was dominant, whereas in the other donor TRBV5-1 was dominant.Similar monoclonal expansions and donor dependency were described before in HLA-E T cells from four donors recognizing various epitopes derived from the UL40 protein.Cytolytic T cell clones derived from these donors generated using a limited dilution culturing system expressed either TRBV16, 9, 22, and 5-1 with a percentage ranging from 60% to 80%. 15,26However, this seems rather unique for UL40 epitopes presented in HLA-E, as in that specific case also the genotype of HLA class Ia alleles restricts the induction of HLA-E-specific responses.Even though a more restricted TCR repertoire could indicate memory formation, it is known that DURTs (although not evaluated for HLA-E) due to their well-described effector and innate-like character generally do not differentiate to a memory phenotype, as was shown before in the context of BCG vaccination. 37he lack of genetic diversity of HLA-E TCRs recognizing UL40-derived epitopes compared with TCRs recognizing Mtb or other pathogenderived peptides could be explained by the sequence similarities between UL40 epitopes and HLA-Ia leader-sequence-derived peptides, which CMV exploits to avoid lysis of the infected cell.It is conceivable that most of the higher affinity TCRs recognizing UL40 peptides identical to self-peptides undergo negative selection in the thymus to avoid the development of self-reactive T cells.The limited number of UL40 TCRs that escape negative selection possibly share the same or similar sequences, leading to a more conserved population.For other pathogens, as we have shown previously for Mtb, various epitopes with sequences divergent from self-peptides can be presented by HLA-E; thymic selection likely permits the generation of a diverse TCR repertoire. 17The recognition of HLA-E/Mtb peptide complexes by multiple TCRab combinations is possibly also beneficial for vaccination with HLA-E peptides as the HLA-E/peptide complex can be recognized by a larger pool of TCRs.In that line, it is interesting to know if a diverse TCR repertoire remains after vaccination with HLA-E peptides as a primary response.
We selected HLA-E Mtb TM + TCRs for functional validation based on the GLIPH2 analysis performed on the bulk RNA-seq and single-cell TCR sequence data.The activation of these TCRs using Zap70 phosphorylation of Y292 and Y319 as a readout after K562 and TM stimulation showed variable results between experiments and TCRs sorted for the same Mtb peptide.Some TCRs displayed relatively clear Zap70 signals; however, several TCRs had limited Zap70 activation after peptide stimulation.Possibly, this is the result of low affinity interactions that are difficult to distinguish from a-specific interactions, therefore not ruling out a degree of a-specificity.A recent study on HLA-E TCRs from T cell clones recognizing SARS-CoV-2-derived peptides also found weak functional T cell responses upon peptide stimulation and dim TM staining that could be explained by the low TCR affinity for HLA-E. 28Besides this, the requirements for HLA-E TCR activation (i.e., co-stimulation signals and cytokines) are not known, and possibly, experiments to induce TCR activation were performed under suboptimal conditions.
Altogether, this study is another step forward in the understanding of the diverse and complex HLA-E-Mtb-restricted immunity, which is crucial in the advancement of TB vaccine design.

Limitations of the study
There are some aspects in our studies that could be improved in future work using latest state-of-the-art methods currently available.First, the validity of our sorting and RNA-seq on sorted CD8 + HLA-E TM + T cells could have been improved with the use of dual TM labeling methods followed directly by single-cell RNA-seq.However, to prevent sorting of a-specific HLA-E TM + cells, we stained with two unrelated HLA-E TMs attached to different fluorochromes and sorted only the single positive HLA-E TM + population.With this setup also cells with a-specific HLA-E TM staining would not be included in the analysis.Second, our statements on p34 and p55 HLA-E TM + -specific gene signatures and TCR profiles based on the transcriptomic data could have been strengthened by comparing and contrasting to other ([non-]classical) T cell subsets, such as HLA-E-restricted CMV-specific T cell responses and HLA-A2-restricted responses, as well as by including several of our recently predicted Mtb-derived HLA-E-specific peptides.Although we intended to have a CMV-specific control population, the individuals in the studied cohort did not have detectable responses toward the HLA-E peptide from CMV that we included as a control for our sequencing analysis.Third, we included individuals from one geographical location, namely South-Africa, but the frequency of circulating HLA-E T cells and the capacity to recognize HLA-E-specific Mtb peptides could depend on the geographical location and TB burden.In addition to this point, we did not perform sex-related analyses.However, the proportion of male to female in our cohort was almost 1:1 (10 males and 11 females), which suggests that the data are representative for both sexes.Lastly, we only included two Mtb-specific HLA-E-restricted epitopes in this study.Including more HLA-E-specific Mtb epitopes might provide a broader overview of the HLA-E Mtb T cell repertoire and would increase the chance of detecting HLA-E/Mtb T cells in the circulation.Due to the limited PBMC material available for each individual, analyses had to be prioritized and were necessarily limited.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

METHOD DETAILS Flow cytometric sorting and analysis
PBMCs from 21 participants (Table S1) were thawed, spun down and resuspended in 200 mL Iscove's modified Dulbecco's medium (IMDM, Gibco Life technologies) and treated for 10 min at RT with 80.000 u/mL DNAse I (Calbiochem).After washing, the PBMCs were counted (CASY cell counter, Roche), resuspended in PBS +0.1% BSA and divided into 2-5x10 6 cells per staining.The cells were then incubated for 10 min at RT with 10 mg/mL CD94 mAb (Clone HP-3D9, BD Bioscience) to block HLA-E tetramer (TM) binding to the CD94/NKG2A receptor complex. 13LA-E TMs were produced as described previously and monomer folding was confirmed with mass spectrometry and staining to LILRB1 expressing cell lines. 17,18,29HLA-E TM staining was performed for 15 min at 37 C; containing an equal mix of HLA-E*01:01 and HLA-E*01:03 TMs, loaded with Mtb peptide #p55 (VMATRRNVL, 0.4 mg TM/10 6 cells) conjugated with phycoerythrin (PE), Mtb peptide #p34 (VMTTVLATL, 0.8 mg TM/10 6 cells) conjugated to allophycocyanin (APC) or loaded with CMV-UL40 glycoprotein derived peptide (VLAPRTLLL, 0.2 mg/10 6 cells) conjugated with APC.PBMCs were stained with 2 different HLA-E TMs labeled with different fluorochromes.PBMCs were subsequently stained for surface expression of: CD3 (BV510, clone UCHT1, BD Biosciences), CD4 (BV650, clone OKT4, Biolegend), CD8 (FITC, clone HIT8a, BD Biosciences).Samples were stained in the dark for 30 min at 4 C, washed in PBS/0.1% BSA.A live/dead stain (Vivid fixable violet dye, Invitrogen) was performed on all samples according to the manufacturer's protocol.Cells were then washed twice in PBS/0.1% BSA and resuspended to 5-10 x 10 6 PBMCs/mL, strained through a 50 mM filter (CellTrics filters, Sysmex) and then acquired on a FACSAria-III cell sorter (BD Biosciences) using FACSDiva software (v8.0.2 BD Biosciences).We sorted only HLA-E TM + cells if a single HLA-E TM + population was visible.A minimum of 800 sorted cells were immediately lysed in in 250 mL QIAzol Lysis Reagent (Qiagen) and stored at À80 C for RNA sequencing.In addition, 100 single cells were sorted and collected in 96-well plates for identification of the single cell TCR sequences.The sorting populations of peptide specific TM-subsets were identified using the gating strategy as shown in Figure S1 and comprised the subsequent gating on the lymphocyte population according to the FCS vs. SSC plot, single cells, live cells, CD3 + cells, CD4 À CD8 + cells and TM + cells.The peptide specific populations sorted for TCR sequencing per donor are shown in Table S1.

HLA-E genotyping
HLA-E*01:01 and/or *01:03 genotypes were determined with the R107G SNP rs1264457 Taqman SNP genotyping assay (Applied Biosystems).At least 10,000 PBMCs were frozen in MQ/1%BSA (Sigma-Aldrich) at À20 C. For analysis samples were thawed and incubated 5 min at 80 C, 10 min at 56 C with 1 mL Proteinase K (Qiagen) and 2 min at 80 C. Subsequently 25-mL qPCR reaction mixes were prepared containing 5-10 mL cell lysate, 12.5 mL Taqman Universal PCR 203 Mastermix (Thermo Fisher Scientific) and 0.15 mL Taqman probe mix rs1264457 (Applied Biosystems).Fluorescence was measured using a Quantstudio6 Flex (Qiagen) during qPCR conditions of 95 C for 5 min, 40 cycles of 95 C for 20 s, 52 C for 20 s, and 72 C for 45 s, followed by 1 cycle of 72 C for 7 min.

RNA sequencing
Following flow cytometric sorting, RNA was isolated from the QIAzol samples according to the manufacturer's protocol using an miRNeasy Mini Kit (Qiagen) and eluted into DNA Lobind Eppendorf Tubes (VWR).RNA integrity and yield was tested on the Agilent 2100 Bioanalyzer.Quality control steps were included to determine RNA and RNA library quality and quantity.Samples that did not pass QC were excluded.RNA from samples of CD8 + TMp55 + cell population paired with the donors' CD8 + population from 16 donors were successfully acquired and suitable for RNA-seq.For 2 donors, we obtained RNA from CD8 + TM CMV + samples.For 1 donor we solely sequenced the CD8 + population as we could not acquire any TM + cells.RNA was amplified to cDNA using the SMART-Seq version 4 Ultra Low Input RNA Kit for Sequencing (Takara Bio).The resulting cDNA was used to prepare a sequencing library.Libraries were sequenced on an Illumina HiSeq 2500 platform to acquire single-end reads of 50 base pairs in length (0.5 Gb/sample), resulting in approximately 17 million reads per sample of which 80% could be aligned against the human genome.

RNA sequencing analysis
Reads were aligned against the human reference genome: Homo sapiens GRCh38.91.Analysis of reads was performed in R (version 3.5.0,R Foundation, software: RStudio v1.2) using the DESeq2 package for the normalization and detection of differentially expressed genes.DESeq2 parameters used were: log2 foldchange (LFC) threshold = 1, read count >4 per gene, adjusted p value cut off = 0.05 and the DESeq2 multifactor design formula = '$donor + cell population'.The cell populations compared were CD8 + and CD8 + TMp55 + .No samples were excluded.The DESeq2 model and analysis processes are described in Love et al. 2014. 42Briefly, the differential expression analysis in DESeq2 uses a generalized linear model using a negative binomial distribution with fitted mean and a gene-specific dispersion parameter.The raw RNA-seq data files are deposited at GEO (GEO accession number: GSE236154).
Single cell cDNA synthesis, multiplex nested PCR, and sanger sequencing TM-specific sorted single cells were collected in PCR plates.cDNA was synthesized from these single cells with SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific) in 2.5-mL reaction mixes, each containing 0.25 mL of SuperScript III reverse transcriptase, 0.5 mL of 53 VILO reaction mix and 0.1% Triton X-100 (Sigma-Aldrich).Mixes were incubated at 25 C for 10 min, 42 C for 120 min, and 85 C for 5 min TCR transcripts from each cell were amplified by multiplex nested PCR in 25-mL reaction mixes containing 2.5 mL of cDNA.The external PCR was performed with 0.75 U of Taq DNA polymerase (Qiagen), 2.5 mL of 103 PCR Buffer (Qiagen) (containing KCl, (NH 4 ) 2 SO 4 and 15 mM MgCl 2 ), 0.5 mL of 10 mM deoxynucleotide triphosphate (Invitrogen) and 2.5 pmol each of the external TRAV, TRBV, TRAC and TRBC primers.PCR conditions were 95 C for 5 min, followed by 35 cycles of 95 C for 20 s, 52 C for 20 s, and 72 C for 45 s, followed by 1 cycle of 72 C for 7 min.The internal PCR used CoralLoad PCR Buffer (Qiagen) containing two marker dyes for gel loading.2.5 mL External PCR product served as template for two separate PCRs that incorporated either (a) internal sense TRAV primers and antisense TRAC primer or (b) internal sense TRBV primers and antisense TRBC primer.TCR segment-specific primers as published by Wang et al. 2012 were synthesized (Sigma-Aldrich), reconstituted and multiplexed to a concentration of 5 mM each for 40 external/internal pairs of sense TRAV, 28 sense TRBV, and 1 of each antisense TRAC and TRBC primers. 38Presence of internal PCR products was confirmed on a 2% agarose gel and aliquots (5 mL) of relevant samples were purified with 2 mL ExoSAP-IT Express reagent (Applied Biosystems) during 4 min at 37 C and 2 min at 80 C. If necessary, products with multiple bands on gel were run over a 1% low-melt agarose gel, single bands were selected and treated with the Wizard SV Gel and PCR Clean-Up System (Promega) according to manufacturer's protocol.Transcripts were Sanger sequenced with 10 pmol of relevant internal antisense primer, either TRAC (TGTTGCTCTTGAAGTCCATAG) or TRBC (TTCTGATGGCTCAAACACAG), by BaseClear (Leiden, the Netherlands).
(paired) TCRa/b analysis TCRab gene fragments from single cell Sanger sequencing, as well as TCRab gene fragments from bulk RNA-seq data, were identified -and paired for single cell data-using MiXCR software version 2.1.9.TCR gene segments were designated according to ImMunoGeneTics nomenclature (IMGT, http://www.imgt.org). 43CDR3a/b pairs displaying the same amino acid sequences and V/J gene usage were defined as clonotypes.Pairwise overlap circos plots were created and BasicStats calculated using VDJtools version 1.2.1 (VDJtools) and Gini-Simpson and inverse Simpson indices were calculated using the tcR package (tcR) in R version 3.5.1 (R). 44,45Graphs and statistics were performed using GraphPad software version 8.0.1 (Prism, La Jolla, USA).Obtained CDR3a/b amino acid sequences from bulk RNA-seq as well as from single cell TCR sequencing were used to construct clone network analysis with the GLIPH2 algorithm. 30,31CR sequences are available at the VDJdb database: (https://github.com/antigenomics/vdjdb-db/issues/351).TCR sequences from sorted single cells with a high clonal number and high significance score as determined by GLIPH2 were selected for transductions and are shown in Tables 2A and 2B.

MP71flex vector design
The MP71flex retroviral vector was used to clone VDJ TCR segments of HLA-E TM specific TCRab chains.This vector contains several unique cut-sites which allows the exchange of variable TCRab domains into the vector and contains codon-optimized murine TCRab constant domains to increase the expression and pairing of the introduced TCRab chains. 32,39Furthermore, the TCRa and b chains are connected with a porcine-teschovirus P2A sequence.TCRab and CDR3ab sequences of the HLA-E restricted TCRs recognizing the human cytomegalovirus UL40-derived epitope VMAPRTLIL were derived from the published HLA-E T cell clone KK50.4.Mtb or other CMV HLA-E restricted TCRab and CDR3ab sequences were derived from HLA-E TM sorted T-cells and selected using the GLIPH2 cluster algorithm. 30TCR vector constructs were designed in silico with the software program Geneious Prime (version 2021.0).Synthesis and cloning of TCR constructs into the MP71flex retroviral vector were performed at BaseClear (Leiden, The Netherlands).

Transfections of Phoenix GALV packing cells
Human 293-based Phoenix GALV packaging cells were transfected with 2.6 mg MP71flex retroviral vector containing TCRab and CDR3ab sequences recognizing previously identified Mtbor CMV-derived peptides. 46Retroviral vectors were packaged using Fugene HD transfection reagent (Promega) in serum-free Opti-MEM I medium (Invitrogen).Virus supernatant was harvested 24 or 48 h post transfection.For all T cell transductions, fresh or thawed virus supernatant was spun down for 20 min 2000G at 4 C on 30 mg/mL RetroNectin (TaKaRa Bio) coated and 2% Human Serum Albumin (HSA) blocked non-tissue culture treated plates (Greiner bio-one CellStar).Virus supernatant was removed from the wells before transduction.

Figure 1 .
Figure 1.Sorted Mtb HLA-E TM + CD8 + T cells have a diverse TCR repertoire TCR encoding RNA transcripts were extracted from the total RNA sequencing dataset by in silico alignment using MiXCR.(A) Pie charts show the frequency of TRAV and TRBV variable fragment usage in both pooled donors (left) and per donor (right).Fragments present in R5% of all TCRa/b in pooled donors (CD8 + TM À n = 17, p55 + n = 16) or R5% of all TCRa/b per donor (p34 + n = 2, UL40 + n = 2, p55 + n = 2) are named, and the remaining fragments are summarized as ''other''.(B) Frequency of CDR3ab sequences present as a single copy or with 2-5, 6-9, 10-100, or >100 copies.Each dot represents the frequency of CDR3 sequences in a donor with that clonal number.Data are depicted from n = 17 total CD8 + TM À samples, n = 16 TMp55 + -sorted samples, n = 2 TMp34 + , and n = 2 TMUL40 + -sorted samples.(C) The percentage of CDR3ab sequences present with R10 copies per donor per population as stated on the y axis.

Figure 2 .
Figure 2. VDJ pairs in sorted p34 and p55 HLA-E TM + T cells have a diverse character (A) Donor-specific variation in V-J segment pairing in CDR3 junctions for TCRa and b chain in Mtb TM + CD8 + T cell populations.A single TCRa and b chain VDJ recombination was performed on single-cell-sorted HLA-E p34 and p55 TM + T cells.Representative plots for p55 + (left) and p34 + (right) populations in respectively donor 1 (*01:01-*01:03) and donor 16 (*01:01) are shown.The chord diagram connects segment pairs by corresponding V-J pair frequency.The total number of identified CDR3s is respectively 32 and 30.(B) CDR3 and insert length in number of nucleotides.Variation in length and corresponding median length of the CDR3 nucleotide sequence (left) and the median number of inserted random nucleotides in the CDR3 sequence (right) were calculated using the BasicStats function of VDJ tools R plug (n = 17).Box shows the median lengths/inserts min to max.(C) Variation in TCRa and b chain concerning V-J segment pairing.Number of unique V-J pairings (left) or duplicate pairings (right) was scored for TM + populations per donor (n = 17).The total number of CDR3s was dependent on sequence sample size.Heterozygous HLA-E*01:01-*01:03 donors are indicated by filled dots; homozygous HLA-E*01:01 donors are open (no HLA-E*01:03 donors identified).(D) Distribution of repetitive a (left) and b (right) clonotypes over different donors.Frequency of CDR3 clonotype replicates on amino acid sequence divided per number of copies.Dots represent pooled data from 20 donors (n = 15 p55 + and n = 5 p34 + ), and data are retrieved from the single-cell TCR dataset.

Figure 3 .
Figure 3. V and J fragment usage and frequency in p55 and p34 TM + T cell populations (A) Absolute number of copies per TRAV, TRAJ, TRBV, and TRBJ fragment (top to bottom) in each of the 17 donors for either p55 + or p34 + TM + TCRs.Data was obtained from the single-cell TCR sequencing dataset.Bar sizes correlate to the absolute number of copies per fragment.Data are shown per donor and per population.(B) CDR3 sequence diversity represented with the Gini-Simpson index.Gini-Simpson score of 1 indicates infinite clonal variation.Open dots represent the singlecell TCR dataset (n = 15 p55 + , n = 5 p34 + ); closed dots represent the RNA-sequencing TCR dataset (n = 17 CD8 + TM À , n = 16 p55 + , n = 2 p34 + , n = 2 UL40 + ).(C) CDR3 sequence diversity represented with the inverse Simpson index.An inverse Simpson index of 1 defines a homogeneous TCR population with only 1 clonotype, and this index increases in parallel to the diversity to a limit related to the total number of reads.

Figure 4 .
Figure4.GLIPH2 analysis to identify TCRb sequences enriched for either the p34 or p55 peptide (A) One hundred seventy-nine GLIPH2 clusters were identified in the bulk and single-cell sequence data of TCRs sorted with p34 or p55 TMs after filtering to exclude GLIPH2 clusters containing sequences from less than three donors.Plot shows clusters that are enriched for p34/p55 specificity in red or for bulk CD8 + TM À in blue (scale on right side).Size of cluster represents the number of TCR sequences in that cluster (increasingly larger dot = increasingly higher number of TCR sequences).(B) Forty-one GLIPH2 clusters from the 179 clusters shown in (A) remain after selecting for clusters with enrichment for p34 (yellow) or p55 (green) over bulk CD8 + TM À (scale on right side).These clusters were grouped into 11 categories containing overlapping GLIPH2 motifs.Size of cluster represents the number of TCR sequences in that cluster (increasingly larger dot = increasingly higher number of TCR sequences).

iScience Article Table 1 .
GLIPH2 identified motifs in the 11 categories

Table 2 .
TCRab sequences selected with GLIPH2 analysis for transduction in Jurkat cells and primary CD3 + T cells