CD8+ T cells specific for the islet autoantigen IGRP are restricted in their T cell receptor chain usage

CD8+ T cells directed against beta cell autoantigens are considered relevant for the pathogenesis of type 1 diabetes. Using single cell T cell receptor sequencing of CD8+ T cells specific for the IGRP265-273 epitope, we examined whether there was expansion of clonotypes and sharing of T cell receptor chains in autoreactive CD8+ T cell repertoires. HLA-A*0201 positive type 1 diabetes patients (n = 19) and controls (n = 18) were analysed. TCR α- and β-chain sequences of 418 patient-derived IGRP265-273-multimer+ CD8+ T cells representing 48 clonotypes were obtained. Expanded populations of IGRP265-273-specific CD8+ T cells with dominant clonotypes that had TCR α-chains shared across patients were observed. The SGGSNYKLTF motif corresponding to TRAJ53 was contained in 384 (91.9%) cells, and in 20 (41.7%) patient-derived clonotypes. TRAJ53 together with TRAV29/DV5 was found in 15 (31.3%) clonotypes. Using next generation TCR α-chain sequencing, we found enrichment of one of these TCR α-chains in the memory CD8+ T cells of patients as compared to healthy controls. CD8+ T cell clones bearing the enriched motifs mediated antigen-specific target cell lysis. We provide the first evidence for restriction of T cell receptor motifs in the alpha chain of human CD8+ T cells with specificity to a beta cell antigen.

Autoreactive CD8 + T cells are in all likelihood key mediators of the pancreatic beta cell destruction leading to type 1 diabetes 1-6 . T cell receptor (TCR)-mediated recognition of (auto-) antigenic peptides presented on MHC class I molecules is a prerequisite for CD8 + T cell mediated target cell destruction. Several islet autoantigen epitopes presented on MHC class I molecules 5,[7][8][9][10][11][12][13] and assays to measure and quantify CD8 + T cell responses against these epitopes have been described [14][15][16][17] . However, information on the TCR repertoire of autoantigen specific CD8 + T cells in type 1 diabetes is so far limited to the TCR sequencing of propagated CD8 + T cell clones 18 , TCR sequence information of single TCR chains of isolated bulk autoantigen specific CD8 + T cells 19 , or CDR3 spectrotype data on bulk autoantigen specific CD8 + T cells 20 . These studies do not provide clonotype information and have not been able to show restricted TCR usage by the autoreactive CD8 + T cells.
TCRs are heterodimers consisting of TCR α -and β -chains and TCR diversity results from combinatorial rearrangements of variable (V), joining (J), and, for TCR β , also the diversity (D) gene segments. V-(D)-J sequences of both chains constitute the hypervariable complementary determining region 3 (CDR3) which provides the major contact point with the antigenic peptide and, therefore, determines antigen specificity of the T cell. The unique combination defines a clonotype. Although TCR clonotypes can be promiscuous in their binding to MHC-peptide complexes 21 , TCRs that recognize epitopes of viral and tumour antigens often have preferred CDR3 motifs or gene usage [22][23][24][25][26] , indicating that some structural restriction of the MHC-peptide binding region of the TCR plays an important role in the selection and expansion of clones.
In this study, we interrogated the TCR repertoire of CD8 + T cells directed against an epitope of an islet autoantigen using single cell TCR sequencing in order to determine whether there is TCR selection in islet autoantigen-specific CD8 + T cells. We chose the islet-specific glucose-6-phosphatase catalytic subunit related protein (IGRP) antigen as a model islet autoantigen, since an HLA A*0201 restricted peptide, IGRP 265-273 , has been identified and IGRP 265-273 directed CD8 + T cells have been detected in the pancreatic islets of organ donors with type 1 diabetes 27 . Additionally, the occurrence and quantification of CD8 + T cells directed against the islet autoantigen IGRP has been demonstrated to have prognostic value on autoimmune diabetes development in NOD mice 4,28 . Our findings suggest that, as described for virus-specific CD8 + T cells, there is selection and expansion of a restricted TCR repertoire in islet-antigen specific CD8 + T cells.

Results
T cell receptor sequencing reveals dominant clonotypes and common alpha chains for IGRP-specific CD8 + T cells. We initially tested our TCR sequencing approach using CD8 + T cells that stained positive with MHC class I multimers loaded with a bona fide Influenza peptide epitope (Flu MP 58-66 ; Supplementary Fig. S1a). From the analysed cells, we identified new as well as previously described 22,23,29,30 Flu MP 58-66 -specific T cell receptor chains (see Supplementary Table S1). We noted inter-individual sharing of TCR α -chains among the analysed Flu-specific cells (Supplementary Fig. S1b) and, in accordance with previous reports 22,23,29,31 , we observed preferential usage of TRBV19 (72.2%), TRBJ2-7 (31.1%) and TRAJ42 (24.4%) genes in the analysed Flu MP 58-66 -specific CD8 + T clonotypes ( Supplementary Fig. S1c). These findings endorse the approach taken to obtain antigen-specific TCR information.
We proceeded to analyse CD8 + T cells with specificity against the type 1 diabetes autoantigen epitope, IGRP 265-273 . Consistent with a previous report 19 , we observed no difference in the frequency of CD8 + T cells that stained positive with multimers loaded with the HLA-A*0201 restricted peptide epitope IGRP 265-273 between healthy controls (median frequency, 0.02% of CD8 + T cells) and children with recent onset of type 1 diabetes (0.01%; P = 0.11) or long-standing type 1 diabetes (0.01%; P = 0.06; Fig. 1a; Supplementary Table S2). However, three patients showed a prominent and distinct population of IGRP 265-273 -specific CD8 + T cells (T1D-1, T1D-2, T1D-3; Fig. 1b). The IGRP 265-273 -specific CD8 + T cells in at least two of these patients had an antigen-experienced memory phenotype, which was in contrast to a predominantly naive phenotype of IGRP 265-273 -specific CD8 + T cells in other patients tested and in healthy controls ( Supplementary Fig. S2).
Paired TCR α -and β -chain sequences of 411 IGRP 265-273 -specific CD8 + T cells were obtained from patients T1D-1, T1D-2, and T1D-3. Dominant TCR α -and β -chain combinations or clonotypes, defined as a unique combination of TCR α -and β -chains, were observed in each of the three patients ( Fig. 1c and Supplementary Table S1). Moreover, the dominant IGRP 265-273 -specific CD8 + T cell clonotypes persisted in samples taken more than 10 months apart from patient T1D-1 ( Supplementary Fig. S3). Of note, the major TCR α -chains were shared among the three patients, paired with a number of different TCR β -chains ( Fig. 1c and Supplementary Table S1). Interestingly, as revealed by combined index sorting of IGRP 265-273 multimer-positive CD8 + T cells and TCR sequencing for T1D-1, those cells that expressed the dominant TCR α -chain detected in T1D-1 and T1D-3 (IGRP α 1; depicted as red piece of pie charts in Fig. 1c) had the highest median fluorescence intensity in the IGRP 265-273 -specific multimer staining (Fig. 1d) as compared to the IGRP α 2-containing cells (depicted as yellow piece of pie charts in Fig. 1c; P < 0.05) and cells with other TCR α -chains (P < 0.001).
Taken together, by analysing the TCR repertoires of IGRP 265-273 directed CD8 + T cells from patients with type 1 diabetes, we identified dominant IGRP 265-273 -specific clonotypes that were persistent in sequential samples and common usage of IGRP 265-273 -specific CD8 + T cell-derived TCR α -chains between individuals.
A part of the SGGSNYKLTF motif (GSNA/YKLT) was previously found in T lymphocytes isolated from pancreatic islets of NOD mice [32][33][34] , as well as in 2 of 53 CD4 + T cell clones isolated from islets of a HLA-A*0201 positive human patient with type 1 diabetes 35 . We, therefore, searched for the SGGSNYKLTF motif in our antigen-specific T cell database. Of the 51 IGRP 265-273 -specific CD8 + T cell clonotypes identified in both patients and controls, 19 (37.3%) had the SGGSNYKLTF motif. In comparison, 2 out of 90 (2.2%; P < 0.0001) of the TCR α -chains from Flu MP58-66-specific CD8 + T cell clonotypes, 0 out of 18 (0%; P = 0.0015) of the TCR α -chains from CMVpp65-specific CD8 + T cell clonotypes 24 , and 29 out of 1492 (1.9%, P < 0.0001) TCR α -chains from GAD65-or Tetanus toxoid-specific CD4 + T cell clonotypes analysed in previous studies 36 had the SGGSNYKLTF motif arguing for an enrichment of this motif in IGRP-specific cells and against a general bias for this motif in antigen-specific T cells.
Dominant IGRP 265-273 -specific CD8 + T cell TCR α-chain and SGGSNYKLTF motif frequencies in CD8 + T cell repertoires. TCRs raised against specific antigens are proposed as potential disease biomarkers 37 . We were, therefore, interested in determining whether any of the TCR α -chains identified in IGRP 265-273 directed cells were associated with type 1 diabetes. We first applied next generation sequencing of  ). Using index sorting, multimer binding cells were isolated as single cells for TCR sequencing. TCR α -chain sequencing information was subsequently attributed to the individual sorted cells and visualized using index sorting fluorescence intensity information (middle). Cells for which no TCR α -chain information was obtained are depicted in gray. The bottom plot compares PE-multimer fluorescence intensities (y axis) of cells expressing the dominant TCR α -chains IGRP α 1 or IGRP α 2, and other TCR α -chains. Lines indicate median values and significant differences between groups according to one-way ANOVA using Dunn's multiple comparison test are marked (*P = 0.01-0.05; ***P < 0.001).
the TCR α -chain repertoire to determine whether the approach could detect the dominant IGRP 265-273 -specific chains IGRP α 1 and IGRP α 2 in bulk CD8 + T cells from the original patients T1D-1, T1D-2, and T1D-3 ( Supplementary Fig. S4). Consistent with the single cell TCR sequencing results (Fig. 1), we retrieved message for IGRP α 1 exclusively within the libraries of donors T1D-1 and T1D-3. Message for the dominant IGRP α 2 was detected within both the libraries of T1D-1 and T1D-2.
We, therefore, performed TCR α -chain next generation sequencing on naïve and memory CD8 + T cells from a new set of HLA A*0201 positive donors comprising control children (n = 14), multiple islet autoantibody (AAb+ ; n = 13) positive children, and recent onset patients with type 1 diabetes (n = 8). Seven of the 45 TCR α -chains identified in IGRP 265-273 multimer-sorted single cells were detected in either the naïve (7 TCR α -chains) or memory (3 TCR α -chains) CD8 + T cells from these individuals (Fig. 3). Within memory CD8 + T cells, IGRP α 2, the dominant TCR α -chain shared by patients T1D-1 and T1D-2, was detected in none of 11 control children as compared to 2 (20%) of 10 autoantibody positive children and 3 (43%) of 7 recent onset type 1 diabetes patients (P = 0.042 controls vs patients). Moreover, IGRP α 2 read frequencies were higher in the memory CD8 + T cell repertoires of patients as compared to controls (P = 0.026). No other differences were observed.
We next examined whether the SGGSNYKLTF motif is per se associated with type 1 diabetes ( Supplementary Fig. S5). TCR α -chains that contained the SGGSNYKLTF motif were detected in all donors and their abundance in naïve or memory CD8 + T cells was not significantly increased in AAb+ (median, naive 0.77% and memory 0.89%) or recent onset type 1 diabetes patients (median, 1.06% and 0.63%) as compared to control children (median, 1.07% and 0.73%), arguing against a general association of TCRs containing this motif with disease.

Discussion
We applied single cell TCR α -and β -chain sequencing to assess the TCR repertoire of CD8 + T cells directed against IGRP 265-273 as a model target epitope of type 1 diabetes. We identified dominant IGRP-specific clonotypes in patients that were persistent in sequential samples, detected sharing of dominant TCR α -chains among patients, and identified a motif that is overrepresented in IGRP 265-273 -specific CD8 + T cells as compared to T cells with other antigen specificities.
The low frequency of islet-antigen-specific CD8 + T cells in the peripheral blood of type 1 diabetes patients has thus far hampered a detailed analysis of their TCR repertoire. The overall low frequency of IGRP 265-273 -specific cells in samples from both type 1 diabetes patients and controls observed in this study is in agreement with previous reports 17,19 that looked at frequencies and phenotypes of CD8 + T cells specific for IGRP and additional autoantigen-specific epitopes, although one of these reports suggested increased frequencies of IGRP-specific CD8 + T cells in patients as compared to controls 17 . The identification of three patients with expanded populations of memory CD8 + T cells specific to IGRP 265-273 allowed us to isolate and analyse a large number of IGRP 265-273specific cells. Although it is possible that our findings on TCR from these patients are not representative of all subjects, it is remarkable that we found sharing of TCR α -chains between the patients and that sharing was marked by a TRAJ53 encoded motif. These findings, as well as the impressive correlation between TCR α -chain and multimer staining intensity, the persistency of expanded clonotypes over time, and the ability of clones derived from multimer-positive CD8 + T cells to antigen-specifically lyse target cells speak for bona fide antigen-specific cells. Further supporting the methodological approach, we find similar TCR gene usage for Flu MP 58-66 -specific CD8 + T cells to that reported in previous reports 22,23,29,30 and identical TCR chain CDR3 sequences and public motifs contained in them 38 .
Selection of antigen-specific T cells in our study was based upon multimer staining. As we demonstrated, different TCRs varied in their multimer binding intensities, suggesting high and low affinity TCRs. Studies in the NOD mouse have demonstrated that TCR avidity against IGRP is relevant to disease stage, with higher avidity clones marking a feature of disease progression 39 . Studies in the NOD mouse argue against IGRP as a primary target in the autoimmune disease process 40 and describe a predominant presence of IGRP directed CD8 + T cells in later stages of disease 28 . It is, therefore, potentially interesting to find expansion of high affinity IGRP 265-273specific TCRs in the patient with long-standing type 1 diabetes. We did not, however, study sufficient numbers  of patients to conclude that the expanded population with a memory phenotype is related to disease duration or accumulation over time.
The repertoire of IGRP-specific cells was highly enriched in TCR α -chains comprising the motif SGGSNYKLTF that is encoded by the TRAJ53 gene, especially when combined with the TRAV29/DV5 gene. Interestingly, this motif, and a highly similar motif encoded by Traj42 (SGGSNAKLTF), has been identified previously in autoreactive and pancreatic islet derived T cells in NOD mice 32,34 , including CD8 + T cells reactive to Igrp peptide 33 and CD4 + T cells reactive to the Insulin peptide B:9-23 41 . Moreover, TCRs targeting the insulin B:9-23 peptide presented by the I-A g7 MHC class II molecule frequently use the Vα gene segment Trav5D-4 rearranged to the Jα gene segments Traj53 and Traj42 42,43 . Thus, islet antigen reactive T cells in NOD mice and man appear to have a recurring preference for certain TCR α -chain genes. Such selection, if true, could be explained by structural similarities across multiple epitope/MHC targets.
Finally, in an attempt to identify molecular markers of type 1 diabetes, we identified one TCR α -chain that, in memory CD8 + T cells, was restricted to patients with type 1 diabetes or pre-type 1 diabetes. The association was not striking, but suggests that it may be possible to identify panels of clonotypes that mark type 1 diabetes-relevant CD8 + T cell activity. A limitation is that we did not perform next generation sequencing for the TCR β -chain. Nevertheless, our findings support the use of approaches similar to those used in this study to extend the knowledge of beta cell antigen-specific TCR repertoires. Moreover, it is possible that both the expansion of autoreactive T cells we observed for IGRP and an apparent preference for certain TCR α -chain genes in responses to different autoantigens reflects cross-reactivity to unknown non-self antigens 44 , which may be favoured by selection of T cells that bear TCRs with a moderate ability to cross-react with multiple peptides 45 .

Methods
Subjects. Samples were obtained from 75 individuals who had the HLA-A*0201 allele (Supplementary Table S2).
These included 27 patients with type 1 diabetes (16 female, 11 male; median age at blood sampling: 22.9 years; range, 5.3-45.9 years), 14 of whom had a diabetes duration of less than one year (median disease duration: 0.04 years; range, 0.01-0.94 years) and 13 patients with a type 1 diabetes duration of more than one year (median disease duration: 14.33 years; range, 1.1-41.5 years); 13 non-diabetic autoantibody positive relatives of patients with type 1 diabetes (3 female, 10 male; median age: 11.6 years; range, 3.8-18.7 years) who had autoantibodies against at least two of four islet antigens (Insulin, GAD65 Peptides. Peptides used in this study were purchased from Mimotopes, Australia at > 95% purity, as confirmed using HPLC and mass spectrometry. (VLFGLGFAI) or a control peptide of the HLA-A2 protein (HLA-A2 140-149 ; YAYDGKDYIA) were purchased from Immudex (Copenhagen, Denmark). Cells were acquired on a Becton Dickinson ARIA II or ARIA III flow sorter with FACS Diva software (for index sorting experiments, FACS Diva 8 software was used) and analysed using FlowJo software (FlowJo LLC Ashland, OR, USA). Unless stated otherwise, cells were stained in PBS supplemented with 1% pooled human AB serum for 30 min on ice followed by at least two washing steps using the same buffer and staining with dead cell marker before cell analysis. Multimer staining was carried out on thawed or fresh PBMC samples after overnight resting at 37 °C/5% CO 2 in X-VIVO 15 supplemented with 5% pooled human AB serum. Cells were resuspended in PBS supplemented with 5% FBS and 50 nM dasatinib and incubated for 30 min at 37 °C. 2 × 10 6 cells were stained with multimers for 10 min at room temperature in the same buffer followed by an additional 20 min incubation at 4 °C in the presence of the respective antibodies. Cells were washed twice with PBS/5% FBS and incubated with 7AAD or Sytox Blue for 10 min immediately before flow cytometry. Gating strategies for the analysis and cell sorting of multimer stained samples are exemplarily shown in Supplementary Fig. S7. Gates for antigen-specific CD8 + T cell frequency analysis were based on staining using the control multimers or, when available, according to staining data of PBMC spiked with the respective antigen-specific CD8 + T cell clone. TCR sequencing. RT-PCR amplification and sequencing of expressed TCR α -and β -chain genes from single CD8 + T cells, as well as subsequent cloning of PCR fragments, were performed as previously described 46 . Analysis of TCR α -and TCR β -chain sequences and junction peptide amino acid sequence extraction was conducted with reference to the International ImMunoGeneTics Information System database 47 . Obtained junction peptides were then analyzed using KNIME 2.11.2 48 .
The TCR α -chain repertoire of 2 × 10 5 flow-sorted total, naïve (CD45RA + CCR7 + ) or memory (CD45RA +/− CCR7 − and CD45RA − CCR7 + ) peripheral blood derived CD8 + T cells from patients and controls was analysed using next generation sequencing as previously described 36  were excluded. TCR CDR3 region sequence extraction and PCR error correction was performed with MiTCR 49 . All additional sequence analysis was conducted in KNIME 2.11.2.
Cloning of antigen specific CD8 + T cells. Flu MP 58-66 -and IGRP 265-273 -specific CD8 + T cell clones were propagated from multimer positive single cells according to a previously described protocol 5 . In brief, purified CD8 + T cells were stained with the respective multimers and viable multimer positive CD8 + T cells were flow-sorted and seeded at 1 cell/well in 96-well plates containing 10 5 irradiated allogeneic PBMCs per well in X-VIVO15 supplemented with 5% Cellkine (Zeptometrix Corp., NY, USA), IL-7 (10 ng/ml), IL-15 (0.1 ng/ml) and PHA-M (5 μ g/ml). CD8 + T cell clones were then stimulated every 14 days at a 1:5 clone to feeder ratio (or smaller in the initial expansion phase, i.e. until a 1:5 ratio per well of a 96 well plate was achieved) with irradiated allogeneic PBMCs that alternately were used either in combination with PHA-M as described above or were preloaded with peptide (FluMP 58-66 or IGRP 265-273 ; 10 μ g/ml).
Cytotoxicity assay. Cytotoxicity of propagated clones was analysed by radioactive 51 Cr release assays.
HLA-A*0201 expressing K562 cells (K562/A*0201 50 ; kindly provided by Prof. Thomas Wölfel; Johannes Gutenberg Universität, Mainz, Germany) were used as target cells. Briefly, 5 × 10 3 51 Cr labelled target cells were seeded into V-shaped 96 well plates followed by effector cells (CD8 + T cell clones) in varying ratios. Assays were conducted in a final volume of 200 μ l /well in X-VIVO15 medium supplemented with 5% AB serum, IL-7 (10 ng/ml) and IL-15 (0.1 ng/ml). Assays were conducted in triplicate and cultures were incubated for 20 h at 37 °C/5%CO 2 in the presence of FluMP 58-66 or IGRP 265-273 peptides (10 μ g/ml) or solvent (DMSO) control. After 4 or 20 hours incubation, 25 μ l of supernatant from each well was transferred into 96 well sample plates containing 150 μ l scintillation liquid per well. After 5 minutes shaking at room temperature 51 Cr release was measured with a scintillation counter (1450 MicroBeta TriLux, Perkin Elmer). Maximum and spontaneous 51 Cr release was determined from cells treated with lysis buffer or assay medium, respectively. Specific cytotoxicity was calculated using the formula: % specific release = (experimental-spontaneous release) × 100/(maximum-spontaneous release).
Statistical analysis. Frequencies of multimer directed cells between groups were compared using the Mann Whitney U test. Fluorescence intensity values of index-sorted cells with different TCR α -chains were compared using one-way ANOVA with Dunn's multiple comparison test. TCR α CDR3 motif frequencies between antigen-specific T cell clonotypes were compared using Fisher's exact test. The presence or absence of TCR α -chain sequences in next generation sequencing information was compared between groups using Fisher's exact test. Read frequencies were compared using Mann-Whitney U test. Statistical analysis was performed using GraphPad prism (version 5.04) or R (version 3.2.5).