Physiological and pathogenic T cell autoreactivity converge in type 1 diabetes

Abstract

Autoimmune diseases result from autoantigen-mediated activation of adaptive immunity; intriguingly, autoantigen-specific T cells are also present in healthy donors. An assessment of dynamic changes of this autoreactive repertoire in both health and disease is thus warranted. Here we investigate the physiological versus pathogenic autoreactive processes in the context of Type 1 diabetes (T1D) and one of its landmark autoantigens, glutamic acid decarboxylase 65 (GAD65). Using single cell gene expression profiling and tandem T cell receptor (TCR) sequencing, we find that GAD65-specific true naïve cells are present in both health and disease, with GAD65-specific effector and memory responses showing similar ratios in healthy donors and patients. Deeper assessment of phenotype and TCR repertoire uncover differential features in GAD65-specific TCRs, including lower clonal sizes of healthy donor-derived clonotypes in patients. We thus propose a model whereby physiological autoimmunity against GAD65 is needed during early life, and that alterations of these physiological autoimmune processes in predisposed individuals trigger overt Type 1 diabetes.

Fig. 1. GAD-specific responses are detected in T1D patients and healthy donors through different orthogonal methodologies.

a, b Scatter plot for IFN-γ and IL-10 ELISPOT stimulation indexes (SI) against GAD in HD (a) and T1D patients (b). Dashed lines represent the thresholds for positivity (SI = 3). c Classification of ELISPOT responses. Shown are percentages of total. Two-sided chi square test (p > 0.05). d Frequency of CD154⁺ CD69⁺ cells from 10 HD and 11 T1D patients after no stimulation (“no antigen”), stimulation with GAD or with SEB. Two-sided Mann Whitney U or unpaired t test. e Phenotype of GAD-specific CD154⁺ CD69⁺ cells from 10 HD and 11 T1D patients (Two-sided Mann Whitney U test. p > 0.05). TN: true naïve. CM: central memory. EM: effector memory. Tscm: stem cell-like memory. Blue: HD. Red: T1D patient. Error bars represent standard deviations.

Fig. 2. GAD-specific cells show qualitative alterations in T1D patients.

a UMAP plot for all GAD-specific cells coloured per cluster. b Details of gene expression per cluster, showing the fraction of cells expressing a gene (dot size) and their expression level (colour). c UMAP plot coloured per type of donor. d CTLA-4 expression levels in cluster #4 (two-sided Mann Whitney U test with Bonferroni multiple comparisons correction. p = 0.043). e IKZF2 expression levels in cluster #10 (two-sided unpaired t test with Welch’s correction. p = 0.0034). f CXCR5 expression levels in cluster #8 (two-sided Mann Whitney U test with Bonferroni multiple comparisons correction. p = 1.97 × 10⁻⁵). g Distribution of HD and T1D cells within cluster #7 subclusters (two-sided Chi-Square test. p = 0.047). h Frequencies of transcription factor-based cell subsets in FOXP3⁺ and FOXP3^neg cells of cluster #5. i CCR5 and Egr2 expression levels in cluster #2 (two-sided Whitney U test with Bonferroni correction. p = 2.7 × 10⁻³ and p = 1.4×10^-6 respectively). j Cell surface phenotype distributions (two-sided Chi-square test. p < 0.0001). c, d, e, f, i blue, HD; red, T1D. g, j white, HD; black, T1D. *: p < 0.05. **: p < 0.01. ***: p < 0.001. ****: p < 0.0001.

Fig. 3. GAD-specific clonotypes are located in multiple gene expression clusters and found in peripheral immune repertoires.

a Circus plot with lines joining cells with identical TCRB and TCRA nucleotide sequences (left) or identical TCRB CDR3 (and unknown or different TCRA) (right). b Circus plots with lines joining cells with similar TCRB CDR3 amino acid sequences (belonging to the same convergence group as per GLIPH algorithm, see Supplementary Table 3), and found in different donors. c Cells whose TCRB CDR3 nucleotide sequence is tracked back in peripheral immune repertoire subsets. d Cells whose TCRB CDR3 amino acid sequence is tracked back in peripheral immune repertoire subsets. c, d blue: HD cells. Red: T1D cells.

Fig. 4. GAD-specific TCRB CDR3 nucleotide sequences can be tracked back into specific peripheral immune cell subsets.

We browsed the GAD-specific TCRB CDR3 nucleotide sequences into 94 peripheral immune cell repertoires: 31 TN (a), 31 CM (b), 16 Treg (c) and 16 Tscm (d). We classified these repertoires into GAD tandem cohort (if both single-cell and deep sequencing took place) and NGS-only cohort (if we only performed deep sequencing). Colour represents the frequency of the given clonotype in peripheral immune cell repertoires.

Fig. 5. GAD-specific TCRB CDR3 nucleotide sequences are public and frequent in specific peripheral immune cell subsets.

a, b Percentage of individuals where GAD- (a) and CMV- (b) specific TCRB CDR3 nucleotide sequences are tracked back. Although each line represents a clonotype, note that many clonotypes show the same behaviour and, as such, they are overlapped in this representation. c Intraindividual frequencies of GAD- and CMV- specific TCRB CDR3 nucleotide sequences (GAD from HD: 1 in TN, 19 in CM, 6 in Treg. GAD from T1D patients: none in TN, 26 in CM, 5 in Treg. CMV: 3 in TN, 32 in CM, 3 in Treg, 6 in Tscm) found in the peripheral immune repertoires (Kruskal Wallis+Dunn). One CMV clonotype exceptionally frequent in the CM subset is not depicted in the figure to ease the visualisation of all other clonotypes (frequency=0.002529). d Ratios of the frequencies of GAD-specific TCRB CDR3 nucleotide sequences in Treg and CM subsets from the same donor (two-sided Mann-Whitney U test). c, d Blue squares: GAD-specific TCRB CDR3 nucleotide sequences found in HD. Red triangles: GAD-specific TCRB CDR3 nucleotide sequences found in T1D patients. White circles: CMV-specific TCRB CDR3 nucleotide sequences. Error bars represent standard deviations. *: p < 0.05.

Fig. 6. GAD-specific amino acid clonotypes are public and convergent in TN and CM peripheral immune cell subsets.

We browsed the GAD-specific TCRB CDR3 amino acid clonotypes into 31 TN (a) and 31 CM (b) peripheral immune cell repertoires. We classified these repertoires into GAD tandem cohort (if both single-cell and deep sequencing took place) and NGS-only cohort (if we only performed deep sequencing). Colour represents the frequency of the tracked back clonotypes in the periphery, while dot size represents the numbers of unique TCRB CDR3 nucleotide sequences coding for each given amino acid clonotype (convergence).

Fig. 7. Some GAD-specific amino acid clonotypes are public in Treg and Tscm peripheral immune cell subsets.

We browsed the GAD-specific TCRB CDR3 amino acid clonotypes into 16 Treg (a) and 16 Tscm (b) peripheral immune cell repertoires. We classified these repertoires into GAD tandem cohort (if both single-cell and deep sequencing took place) and NGS-only cohort (if we only performed deep sequencing). Colour represents frequency of the tracked back clonotypes in the periphery, while dot size represents numbers of unique TCRB CDR3 nucleotide sequences coding for each given amino acid clonotype (convergence).

Fig. 8. GAD-specific TCRB CDR3 amino acid clonotypes from T1D patients show differential publicity and convergence features within peripheral immune repertoires.

a, b: Percentage of peripheral immune repertoires where each GAD-specific TCRB CDR3 amino acid clonotype is tracked into, disaggregated per type of subject (a HD, and b T1D). Orange shade: extremely public clonotypes (present in ≥ 75% of individuals). Pink shade: public clonotypes (present in 25.0%–74.9% of individuals). Green shade: private clonotypes (present in 3.23%–24.9% of individuals) or ultraprivate (only found in the same single-cell donor, 3.23%). For Treg and Tscm, the threshold for “ultraprivate” is 5.88 and 6.25 respectively (faint dotted lines). c Percentage of repertoires where GAD-specific TCRB CDR3 amino acid clonotypes are tracked back into (HD: 32 clonotypes into TN, 51 into CM, 23 into Treg and 10 into Tscm. T1D: 66 clonotypes into TN, 77 into CM, 29 into Treg and 25 into Tscm. Two-sided Mann-Whitney U test. *p = 0.013). Blue: HD. Red: T1D. d Proportions of each publicity-based clonotype category in Treg repertoires (two-sided Chi-square test. * p = 0.0025. Blue: HD. Red: T1D). e–h correlations between number of convergent nucleotide sequences and frequency for each GAD-specific TCRB CDR3 amino acid clonotype tracked back into peripheral immune repertoires (Spearman correlation). e, f correlations for GAD (e) and CMV (f) clonotypes found in TN subsets. g, h correlations for GAD (g) and CMV (h) clonotypes found in CM subsets. e, g Blue: HD-derived GAD clonotypes. Red: T1D-derived GAD clonotypes. Error bars represent standard deviations.

Fig. 9. Frequency, clonal sizes and phenotypes are different between HD- and T1D-derived GAD clonotypes.

a, c Frequency of GAD-specific TCRB CDR3 amino acid clonoytpes in TN (a) and CM (c) peripheral immune repertoires, disaggregated by type of donor (57 clonotypes for HD, 90 for T1D and 64 for CMV). Shown are Tukey boxplots: boxes are interquartile range (25% to 75%), whisker (up) is 75^th percentile plus 1.5 times the interquartile range (IQR). The centre is the 50^th percentile. Individual dots are values that are greater than the whisker. Two-sided Mann Whitney U test with Bonferroni correction (***: p < 0.001). The frequency of each clonotype was calculated as the average frequency for all nucleotide sequences coding for the same amino acid sequence (See “Methods” section). b, d heatmaps showing the means of frequencies, and the corresponding p-values (two sided Mann Whitney U test with Bonferroni correction), of data shown in a, c. e, f we identified GAD clonotypes appearing in peripheral repertoires from HD but not from T1D patients (“HD-only”, in blue), and GAD clonotypes appearing in peripheral repertoires from T1D patients but not from HD (“T1D-only”, in red), and represented them on the UMAP space. GAD clonotypes found in TN repertoires are shown in (e), while those found in CM ones are shown in (f). Arrows in (e) depict expanded clonotypes. g HD-only (blue) and T1D-only (red) GAD clonotypes tracked into CM subsets which show clonal expansions.

Fig. 10. Model of physiological and diabetogenic autoreactivity against GAD.

In HD (left), GAD-specific clonotypes are selected in the thymus, generating GAD-specific TN cells which seem to include “protective” clonotypes. During neonatal remodelling of the pancreas, physiological beta-cell death occurs, releasing GAD. GAD-specific TN cells get activated by antigen in this context of cell death, initiating a repair mechanism, leading to a fully functional pancreas. During this process, GAD-specific effector and memory cells are generated, presenting phenotype features predisposed to tissue remodelling and repair. Equally, peripherally induced GAD-specific Treg (iTreg) cells are generated and, in potential combination with thymic derived Treg (tTreg) cells and other peripheral tolerance mechanisms, keep the memory GAD cells under control. In T1D patients (right), GAD-specific clonotypes are also selected in the thymus- however, alterations during recombination events generate a repertoire of GAD-specific TN cells that is more public. Additionally, “protective” clonotypes are rarer. During pancreas remodelling and GAD release, an altered physiological autoreactive process occurs: the GAD-specific effector and memory cells generated during the process are of altered phenotypes which are not sufficient to drive appropriate remodelling and repair mechanisms. Additionally, effector and memory cell generation is further fuelled by pre-primed TN cells. As such, beta-cell death continues. Less induced Treg cells are generated and, if so, their phenotypic features do not allow them to contain the effector responses. Thymic Tregs and other peripheral tolerance mechanisms, previously shown to be altered in T1D, cannot contain this diabetogenic autoreactivity. This combination of altered physiological and diabetogenic autoreactivity converge to propel beta cell death, leading to overt T1D. GAD-sp: GAD-specific. Some icons have been created in BioRender. Gomez-tourino, I. (2024) BioRender.com/q46s625.