Dynamic transition of Tregs to cytotoxic phenotype amid systemic inflammation in Graves' ophthalmopathy

Abstract

Graves' disease (GD) is an autoimmune condition that can progress to Graves' ophthalmopathy (GO), leading to irreversible damage to orbital tissues and potential blindness. The pathogenic mechanism is not fully understood. In this study, we conducted single-cell multi-omics analyses on healthy individuals, patients with GD without GO, newly diagnosed patients with GO, and treated patients with GO. Our findings revealed gradual systemic inflammation during GO progression, marked by overactivation of cytotoxic effector T cell subsets, and expansion of specific T cell receptor clones. Importantly, we observed a decline in the immunosuppressive function of activated Treg (aTreg) accompanied by a cytotoxic phenotypic transition. In vitro experiments revealed that dysfunction and transition of GO-autoreactive Treg were regulated by the yin yang 1 (YY1) upon secondary stimulation of thyroid stimulating hormone receptor (TSHR) under inflammatory conditions. Furthermore, adoptive transfer experiments of the GO mouse model confirmed infiltration of these cytotoxic Treg into the orbital lesion tissues. Notably, these cells were found to upregulate inflammation and promote pathogenic fibrosis of orbital fibroblasts (OFs). Our results reveal the dynamic changes in immune landscape during GO progression and provide direct insights into the instability and phenotypic transition of Treg, offering potential targets for therapeutic intervention and prevention of autoimmune diseases.

Keywords: Autoimmune diseases; Autoimmunity; Bioinformatics; Genetics.

Figure 1. Single-cell atlas of expression and chromatin accessibility revealed enhanced inflammation in the systemic immune environment during GO progression.

(A) Schematic representation of single-cell multi-omics sequencing (scRNA-Seq, scATAC-Seq, and scTCR-Seq) experimental design. PBMCs were isolated from healthy controls (Healthy, n = 10), patients with Graves’ disease without ophthalmopathy (GH, n = 10), patients with Graves’ ophthalmopathy (GO, n = 10), and corticosteroid-treated patients (Treated, n = 11), followed by processing using the 10X Genomics platform. (B) UMAP plots depicted major immune cell types in peripheral blood based on scRNA-Seq and scATAC-Seq datasets. (C and D) UMAP plots illustrated the expression levels (C) and gene activity scores (D) of typical marker genes for major immune cell types. (E) Violin plots showed the expression and activity scores of Graves’ disease, inflammatory, and IFNG signal gene sets among Healthy, GH, and GO groups. Data are represented as the median IQR. ****P_adj < 0.00001 by Mann-Whitney U test. (F) Scatter plots depicted the correlation of Graves’ disease gene and inflammatory gene scores across all samples in Healthy, GH, and GO groups. (G–I) Flow density plots and dot plots displayed the expression levels of IL-6 (G), TNF-α (H), and IFN-γ (I) in peripheral blood from Healthy, GH, and GO donors (Healthy, n = 9; GH, n = 9; GO, n = 9). Data are represented as the median IQR. *P_adj < 0.05, **P_adj < 0.001, ***P_adj < 0.0001 by Mann-Whitney U test.

Figure 2. The dynamic transcriptional characteristics of T cell subpopulations during the progression of GO.

(A) Major T cell subtypes in the scRNA-Seq dataset. (B and C) All subtypes of CD4 T (B) and CD8 T cells (C) in both scRNA-Seq and scATAC-Seq datasets. (D and E) The relative abundance of clonotypes and distribution of the GO top 10 unique clonotype among CD4 T cell subtype (D) and CD8 T cell subtype (E). (F) The expression levels for the cytotoxicity functional genes in CD4 CTL. Data are represented as the median IQR. ****P_adj < 0.00001 by Mann-Whitney U test. (G) Chromatin accessibility peak of GZMB and FGFBP2 in CD4 CTL. (H) qPCR showed the gene expression levels of GZMB, FGFBP2, CTSW, and PRF1 in CD4 CTL (Healthy, n = 6; GO, n = 6). Data are represented as the median IQR. **P < 0.001 by Mann-Whitney U test. (I and J) The scores of exhaustion score (I) and aTreg functional gene score (J) in aTreg; heatmap displayed the expression levels of representative functional genes. Data are represented as the median IQR. ****P_adj < 0.00001 by Mann-Whitney U test. (K) The expression levels of the top 10 GO-specific TRAV and TRBV genes in aTreg. (L and M) The expression of TGF-β1 (L) and CTLA4 (M) from 3 groups; dot plots indicate the mean expression intensity and target cell proportions (Healthy, n = 9; GH, n = 9; GO, n = 9). Data are represented as the median IQR. ***P_adj < 0.0001 by Mann-Whitney U test. *P_adj < 0.05, **P_adj < 0.001. (N–P) The proliferation (N), inhibition (O) and IFN-γ (P) production of CD8⁺ T cells co-cultured with Treg (Healthy, n = 5; GH, n = 5; GO, n = 5). Data are represented as the median IQR. *P_adj < 0.05 by Mann-Whitney U test. The experiments were repeated 3 times.

Figure 3. aTreg acquired CTL transcriptional signature under the high-inflammatory conditions of GO.

(A) Pseudotime analysis of CD4 T cell subpopulations. (B) Dendrogram of CD4 CTL hdWGCNA and representative genes for each module. (C) UMAP plot showed the expression of different module genes in CD4 T cell subpopulations. (D) The expression of module 3 genes in CD4 T cell subpopulations. (E) The expression differences of module 3 genes in Tfh and aTreg between GH and GO groups. Data are represented as the median IQR. ***P < 0.0001 by Mann-Whitney U test. (F) Pseudotime analysis of aTreg, rTreg, and CD4 CTL. Left: Pie chart displayed the proportions of cells in different states over time for Healthy, GH, and GO groups. Right: Distribution of cells in different states along the branching trajectory over time. (G) The scores of CD4 CTL transcriptional feature gene set in aTreg from Healthy, GH, and GO groups; heatmap displayed the expression levels of representative CD4 CTL functional genes in aTreg from 3 groups. Data are represented as the median IQR. ****P_adj < 0.00001 by Mann-Whitney U test. (H) Chromatin accessibility plots of KLRC1 and KLRC4 genes in aTreg from 3 groups. (I) Upset plot of aTreg and CD4 effector T cell clonotype sharing. On left, total number of expanded clones for each subpopulation is displayed. On top, the total number of shared clones is displayed. Incidences of clone sharing between aTreg and CD4 CTL are highlighted. (J and K) The expression of KLRC1 (J) and FGFBP2 (K) in peripheral blood CD4⁺CD25⁺FOXP3⁺ Treg from 3 groups; dot plots indicate the target cell proportions (Healthy, n = 9; GH, n = 9; GO, n = 9). Data are represented as the median IQR. *P_adj < 0.05, **P_adj < 0.001, ***P_adj < 0.0001 by Mann-Whitney U test.

Figure 4. YY1 promoted the cytotoxic transition of autoreactive Treg upon secondary antigen stimulation in an inflammation-dependent manner.

(A) Venn diagram illustrated the number of consistently upregulated DEGs in aTreg of GO group compared with the GH group and healthy group. (B) Venn diagram displayed the overlap between the consistently upregulated DEGs in aTreg of GO group and the CTL signature gene set. (C) The transcription factor regulatory network of aTreg consistently upregulated CTL signature genes predicted based on JASPAR. (D) TF motif enrichment analysis of consistently overrepresented sequences in GO. (E) GO analysis of consistently upregulated genes regulated by YY1. (F) Experimental model diagram of the toxic shift of Treg in vitro: Treg were isolated from healthy and GO donors, expanded in culture for 7 days, followed by LV-YY1 shRNA knockdown or LV-YY1 overexpression. After 2 days, these cells were treated with TSHR recombinant protein and/or proinflammatory cytokines (TNF-α, IFN-γ) for 48 days. (G) qPCR histogram showed the expression levels of 4 specific genes about the results of Treg cytotoxic transition under different treatment conditions. All intergroup P values can be found in Supplemental Table 5. Each group consists of 3 independent samples (each n = 3). Data are represented as the median IQR. All experiments were repeated 3 times.

Figure 5. Cytotoxic phenotype of Treg was alleviated as inflammation decreased after corticosteroid therapy.

(A) Bar chart showed the GO functional enrichment of downregulated DEGs in the corticosteroid-treated group compared with the GO group. (B) The cell proportions of CD4 T cell subpopulations within the GO and Treated groups. Data are represented as the median IQR. **P < 0.001 by Mann-Whitney U test. (C) The expression and activity scores of Graves’ disease, inflammatory, and IFNG signal gene sets between the GO and Treated groups. Data are represented as the median IQR. ****P < 0.00001 by Mann-Whitney U test. (D) Flow density plots and dot plots showed the expression intensity of IL-6, TNF-α, and IFN-γ in peripheral blood PBMCs from the GO and Treated groups (GO, n = 9; Treated, n = 9). Data are represented as the median IQR. ***P < 0.0001 by Mann-Whitney U test. **P_adj < 0.001. (E) Violin plots show the CTL signature score of aTreg between the GO and Treated groups. Data are represented as the median IQR. ****P < 0.00001 by Mann-Whitney U test. (F) Chromatin accessibility plots for KLRC1 and KLRC4 genes in aTreg between the GO and Treated groups. (G) Flow cytometry plots displayed the expression of KLRC1 and FGFBP2 in peripheral blood CD4⁺CD25⁺FOXP3⁺ Treg in Treated groups; dot plots indicating the target cell proportions (GO, n = 9; Treated, n = 9). **P_adj < 0.001. Data are represented as the median IQR. ***P < 0.00001 by Mann-Whitney U test.

Figure 6. Pathogenic role of Treg cytotoxic transition to the orbital lesions of GO.

(A) Schematic diagram of the EGFP labeled autoreactive Treg adoptive transfer experiment. (B) The proportion of EGFP⁺KLRC1⁺ cells within CD4⁺ cells in the peripheral blood of Healthy and GO mice after adoptive transfer experiment, representing the cytotoxic transition of Treg in vivo (Healthy, n = 3; GO, n = 9). Data are represented as the median IQR. *P < 0.05 by Mann-Whitney U test. (C) Immunofluorescence staining demonstrated the presence of KLRC1⁺EGFP⁺ cells in the orbital tissues of GO model mice. (D and E) The major subtypes of orbital tissues (D) and integrated CD4 T cells from PBMCs and tissues (E) based on scRNA-Seq data of Healthy donors and patients with GO; the upper right UMAP plot illustrates the batch effects. (F) The ligand-receptor interactions involving CD4 CTLs significantly altered in GO among 3 LPF subgroups, MYF, and COF. lipofibroblast, LPF; myofibroblast, MYF; conventional orbital fibroblast, COF. Red represents inflammation-regulating ligand-receptor interactions, while green indicates ligand-receptor interactions related to extracellular matrix remodeling. (G) Experimental model of the pathogenic effect of Treg cytotoxic transformation on localized lesions in the orbit. (H) Immunofluorescence showed the expression of specific proteins of OF cells in the coculture system of healthy control and GO model mouse groups; the histograms represent the quantitative statistics of these proteins. Data are represented as the median IQR. **P < 0.001 by Mann-Whitney U test. (I) qPCR histogram showed the expression of specific genes in cells of the coculture system. Each group consists of 3 independent samples (each n = 6). Data are represented as the median IQR. **P < 0.001 by Mann-Whitney U test. All experiments were repeated 3 times.