An autoimmune transcriptional circuit drives FOXP3+ regulatory T cell dysfunction

Abstract

Autoimmune diseases, among the most common disorders of young adults, are mediated by genetic and environmental factors. Although CD4⁺FOXP3⁺ regulatory T cells (T_regs) play a central role in preventing autoimmunity, the molecular mechanism underlying their dysfunction is unknown. Here, we performed comprehensive transcriptomic and epigenomic profiling of T_regs in the autoimmune disease multiple sclerosis (MS) to identify critical transcriptional programs regulating human autoimmunity. We found that up-regulation of a primate-specific short isoform of PR domain zinc finger protein 1 (PRDM1-S) induces expression of serum and glucocorticoid-regulated kinase 1 (SGK1) independent from the evolutionarily conserved long PRDM1, which led to destabilization of forkhead box P3 (FOXP3) and T_reg dysfunction. This aberrant PRDM1-S/SGK1 axis is shared among other autoimmune diseases. Furthermore, the chromatin landscape profiling in T_regs from individuals with MS revealed enriched activating protein-1 (AP-1)/interferon regulatory factor (IRF) transcription factor binding as candidate upstream regulators of PRDM1-S expression and T_reg dysfunction. Our study uncovers a mechanistic model where the evolutionary emergence of PRDM1-S and epigenetic priming of AP-1/IRF may be key drivers of dysfunctional T_regs in autoimmune diseases.

Fig. 1.. Deep transcriptomic analysis of mT_regs and T_convs highlights PRDM1 up-regulation in MS.

(A) Schematic of study design. mT_convs or mT_regs were sorted by FACS from peripheral blood CD4⁺ T cells from individuals with MS and HCs. Bulk RNA-seq was performed on the discovery cohort, and DEGs were identified (HC, n = 20; MS, n = 26). The selected DEGs were validated by an independent validation cohort (HC, n = 23; MS, n = 16). (B) Volcano plots showing DEGs for mT_regs and mT_convs between individuals with MS and HCs. (C) Overlapped DEGs between mT_regs and mT_convs. (D) qPCR validation for PRDM1 expression in both discovery and validation cohorts. (E) Protein validation for BLIMP1 expression using flow cytometry (HC, n = 12; MS, n = 9). APC, allophycocyanin; gMFI, geometric mean fluorescence intensity. (F) Heatmap depicting expression patterns of the selected six genes in mT_reg/Fr2 eT_reg across 12 autoimmune diseases. Fr2 eT_reg and Fr1 nT_reg in the study by Ota et al. correspond to our mT_regs and nT_regs, respectively. Other autoimmune diseases included SLE, IIM, systemic sclerosis (SSc), mixed connective tissue disease (MCTD), Sjögren’s syndrome (SjS), rheumatoid arthritis (RA), Behçet’s disease (BD), adult-onset Still’s disease (AOSD), AAV, and Takayasu arteritis (TAK). Data were extracted from Ota et al. (30) and Gao et al. (31). *P < 0.05, **P < 0.01. Statistical significance was computed by unpaired t test in (D) and (E).

Fig. 2.. Single-cell dual-omics analysis reveals elevated PRDM1 in effector T_regs in MS.

CD4⁺ T cells isolated from 10 individuals (five HCs and five with MS) were analyzed and shown. (A) Surface protein–guided CD4⁺ T cell subtype annotation. Four CD4⁺ T cell subtypes were distinguished by CD25, CD127, CD45RO, and CD45RA protein expression. (B) tSNE based on gene expression for CD4⁺ T cells. (C) FOXP3, IKZF2, PRDM1, and BACH2 expression is shown by tSNE visualization (all cells that passed quality control are plotted). (D) Combined differential analysis for bulk- and scRNA-seq in mT_regs. Representative DEGs with pseudobulk analysis with scRNA-seq are shown. Gene expression changes in MS relative to control samples for the indicated genes were computed at the single-cell level (gray and red lines represent the average of HC or MS, respectively). The effect sizes of bulk (t-statistics) and scRNA-seq DEG analysis were compared, reporting Spearman’s rank correlation values between the two types. Correlation analyses were conducted separately in mT_regs and mT_convs. The correlations across 8272 (mT_reg) and 8262 (mT_conv) genes that were marginally significant in the bulk-level analysis were assessed. The size of dots is scaled proportionally to the average number of mRNA reads quantified within each batch and condition; the y axis (disease effect) shows average gene expression after adjusted by confounding factors by matching; the error bars capture 1 SD for the disease effects in Bayesian inference. (E) Combined differential analysis for bulk- and scRNA-seq in mT_convs. Data are shown as in (D). (F) Subcell type annotation of mT_regs based on CITE-seq. (G) Heatmaps showing the marker genes (italic font) and proteins (regular font) to define subcell types for mT_regs. Standardized logarithm-transformed gene expression (log1p) values were calculated and summarized at each subcell type level. (H) The changes in PRDM1 and FOXP3 expression are shown at subcell type level for mT_regs from individuals with MS and controls. Data are shown as in (D).

Fig. 3.. The alternative short PRDM1 isoform is elevated in MS mT_regs.

(A) Schematic of PRDM1 short and long isoforms. PR; PR domain, Pro/Ser; proline/serine-rich region, ZnF; five C2H2 zinc fingers. (B) ATAC-seq [mT_regs and nT_regs from HC (n = 8)], DHS (ENCODE primary T_regs, Roadmap primary T cells, ENCODE monocytes, Roadmap primary B cells), and HiDRA peaks at PRDM1 locus. Accessible chromatin regions of human mT_regs are highlighted in orange. (C) PRDM1-S and PRDM1-L isoform expression across nine different immune cell types in peripheral blood was assessed by bulk RNA-seq (n = 6 HCs). (D) Western blot analysis of BLIMP1 expression from eight different immune cell types in peripheral blood of HCs. Conventional BLIMP1 and alternative BLIMP1-S are distinguished by different sizes. (E) PRDM1-S and PRDM1-L gene expression was assessed by bulk RNA-seq in mT_regs between MS and HC samples. **P < 0.01; statistical significance was computed by one-way analysis of variance (ANOVA) with Dunn’s multiple comparisons tests. (F) Heatmaps depicting Spearman’s correlation for curated immune-related genes with PRDM1-S or PRDM1-L in peripheral blood samples from HCs or individuals with MS. Different patterns of correlation for PRDM1-S and PRDM1-L with HC or MS samples are highlighted in three boxes (boxes 1 to 3).

Fig. 4.. PRDM1-S induces SGK1 expression and T_reg dysfunction.

(A) Volcano plot showing statistical significance and FC for genes differentially expressed by PRDM1-S OE primary mT_regs. (B) SGK1 expression was assessed by qPCR for OE of PRDM1-S and PRDM1-L with primary human mT_regs (top) and Jurkat T cells (bottom). ***P < 0.001, ****P < 0.0001; Statistical significance was computed by one-way ANOVA with Dunn’s multiple comparisons tests. (C) The top track illustrates BLIMP1-S binding sites (positive signal, purple) over control IgG binding (negative signal, gray), as determined by the MACS2 bdgcomp function. Significant peaks in BLIMP1-S binding over IgG control were identified using SEACR (blue bars). BLIMP1 ChIP-seq data from H549 cells (light blue track) and ATAC-seq data from human primary T_regs (green track) were obtained from ENCODE and included in the figure. In addition, as a positive control, H3K4me3 CUT&RUN was performed concurrently with BLIMP1-S and IgG control CUT&RUN and is depicted at the bottom (magenta track). BLIMP1-S binding regions are highlighted in red boxes. (D) In vitro T_reg suppression assay with human primary T_regs. T effector cell proliferation was assessed after 5 days of coculture with T_regs transduced with GFP control vector or PRDM1-S OE vector. **P < 0.01; statistical significance was computed by two-way repeated measures ANOVA. (E) Flow cytometry analysis for FOXP3 in primary T_regs overexpressing PRDM1-S compared with GFP control (n = 10). **P < 0.01; statistical significance was computed by paired t test.

Fig. 5.. AP-1 and IRF TF binding are enriched in MS mT_regs.

(A) Schematic of ATAC-seq experiments on mT_regs from individuals with MS (n = 26) and HCs (n = 21). (B) TF motif enrichment analysis in mT_regs between MS and HC by HINT. IRF and AP-1 motifs are highlighted in red and blue, respectively. (C) TF footprint enrichment analysis in mT_reg isolated from individuals with MS and HCs by HINT and TOBIAS. TF footprints that are significantly enriched in mT_regs from individuals with MS are highlighted in red.

Fig. 6.. Identification of active enhancer element for PRDM1-S with AP-1 and IRF binding.

(A) Schematic experimental overview. (B) CRISPRa validation for top 20 PRDM1-associated regulatory elements. Top: Top 20 accessible chromatin elements that are associated with PRDM1 expression are highlighted in orange. Potential interactions of regulatory elements with the PRDM1 gene are analyzed with the GeneHancer database and shown on the top. Middle and bottom: CRISPRa-induced expression of PRDM1-S (middle) and PRDM1-L (bottom) was assessed by qPCR. Detailed information for all 20 regions is shown in table S4. (C) Validation CRISPRa and CRISPRi experiments for the #2 peak using peak #2–targeted sgRNA-1 (n = 4). (D) Top: H3K27ac, H3K4me1, and H3K4me3 MINT-ChIP signal on the #2 peak region in mT_regs from HC and MS samples. Four replicates of HC mT_regs and two replicates of MS mT_regs are merged into one representative track, respectively. Middle: Footprint analysis on #2 peak region with TOBIAS footprint score. Bottom: AP-1 and IRF ChIP-seq signals identified on #2 peak region in ENCODE data are shown. AP-1/IRF composite motif identified in #2 peak region is highlighted. *P < 0.05, **P < 0.01, and ****P < 0.0001; statistical significance was computed by one-way ANOVA with Dunn’s multiple comparisons tests.