Endogenous antigens shape the transcriptome and TCR repertoire in an autoimmune arthritis model

Abstract

The development of pathogenic autoreactive CD4+ T cells, particularly in the context of impaired signaling, remains poorly understood. Unraveling how defective signaling pathways contribute to their activation and persistence is crucial for identifying new therapeutic targets. We performed bulk and single-cell RNA-Seq (scRNA-Seq) and single-cell T cell receptor sequencing (scTCR-Seq) to profile a highly arthritogenic subset of naive CD4+ T cells from BALB/c-Zap70*W163C (SKG) mice, which develop CD4+ T cell-mediated autoimmune arthritis driven by a hypomorphic mutation in Zap70 - a key TCR signaling kinase. Despite impaired signaling, these cells exhibited heightened expression of T cell activation and cytokine signaling genes but diminished expression of a subset of tolerogenic markers (Izumo1r, Tnfrsf9, Cd5, S100a11) compared with WT cells. The arthritogenic cells showed an enrichment for TCR variable β (Vβ) chains targeting superantigens (Sags) from the endogenous mouse mammary tumor virus (MMTV) but exhibited diminished induction of tolerogenic markers following peripheral antigen encounter, contrasting with the robust induction of the negative regulators seen in WT cells. In arthritic joints, cells expressing Sag-reactive Vβs expanded alongside detectable MMTV proviruses. Antiretroviral treatment and Sag-reactive T cell depletion curtailed SKG arthritis, suggesting that endogenous retroviruses disrupted peripheral tolerance and promoted the activation and differentiation of autoreactive CD4+ T cells into pathogenic effector cells.

Keywords: Autoimmune diseases; Autoimmunity; Immunology; T cell receptor; Tolerance.

Figure 1. Prearthritic naive SKG T cells demonstrate enhanced T cell activation.

(A) Experimental schematic of the bulk RNA-Seq analysis. neg, negative. (B) PCA based on transcriptomics data from bulk RNA-Seq reveals the distribution of SKGNur and WTNur GFP^lo and GFP^hi CD4 naive T cell subsets as shown in A (n = 3 per subgroup). (C) Heatmap showing the expression of 991 significantly DEGs [|log₂(fold change [FC])| >1, adjusted (adj.) P < 0.05] from pairwise comparisons for all samples grouped by subgroup. Hierarchical clustering was used to group DEGs into 6 modules (indicated by dendrogram and row annotation color bar on left). (D) Enrichment plots of TCR signaling and cytokine pathways from gene set enrichment analysis (GSEA) analysis of all Gene Ontology Biological Process (GOBP) pathways for ranked genes from SKGNur GFP^hi and WTNur GFP^hi differential expression analysis. NES, normalized enrichment score.

Figure 2. scRNA-Seq reveals heterogeneity in naive CD4⁺ T cells, highlighting a subset defined by TCR signaling genes.

(A) Experimental design of paired scRNA- and TCR-Seq of sorted GFP^hi and GFP^lo naive CD4⁺ T cells. V(D)J, variable, diversity, and joining sequences. (B) Uniform manifold approximation and projection (UMAP) of 99,074 naive T cells from 8 samples in A, colored according to merged clusters. (C) Violin plots of log-normalized expression of marker genes for each cluster. Black box highlights the T.4N_Nr4a1 cluster; red box highlights genes uniquely expressed in the T.4N_Nr4a1 cluster. (D) Volcano plot of DEGs in the T.4N_Nr4a1 cluster versus all other cells. Dots are colored according to significant overexpression (|log₂(FC)| >1, adj. P < 0.05) in the T.4N_Nr4a1 cluster (red) or in other cells (dark gray), or to no significant difference (light gray). Labeled genes are colored according to their role in TCR signaling regulation: positive (red) or negative (blue). Heatmap shows average expression of the labeled genes by subgroup normalized by the standard scale for each gene. (E) Volcano plot of DEGs from SKGNur versus WTNur GFP^hi cells in the T.4N_Nr4a1 cluster. Dots are colored according to significant overexpression (|log₂(FC)| >0.2, adj. P < 0.05) in WTNur (orange) or SKGNur (blue) GFP^hi cells or no significant difference in expression between groups (gray). Labeled genes involved in TCR signaling are colored as in D. (F) Enrichment plots of TCR activation and signaling pathways from GSEA analysis of GOBP pathways for ranked DEGs of SKGNur versus WTNur GFP^hi cells from the T.4N_Nr4a1 cluster. (G) Violin plots show expression of “natural anergy” genes in WTNur, SKGNur GFP^lo, and GFP^hi CD4⁺ naive cells from the T.4N_Nr4a1 cluster. Heatmap displays average gene expression by subgroup. Both are normalized by the standard scale. Asterisks in the heatmap labels in E and G indicate significant differential gene expression between SKGNur and WTNur GFP^hi cells.

Figure 3. T.N4_Nr4a1 cells segregate into 2 distinct TCR signaling modules that segregate acute from chronically antigen-activated T cells.

(A) Correlation matrix shows hierarchical clustering of Spearman’s correlation of the top 25 HVGs that positively and negatively correlated with Nr4a1 expression across all cells (SKG and WT). Diagonal gray boxes represent a correlation of 1; dark gray boxes mark distinct gene modules from genes that positively correlated with Nr4a1 expression. (B) UMAP plots show expression levels in all cells of the indicated marker genes positively correlated with Nr4a1, as identified in A. The scale represents the log-transformed normalized gene counts. (C) Volcano plot shows DEGs for SKG and WT cells in the T.4N_Nr4a1 cluster that expressed (log-normalized expression >1) Egr2 or Tnfrsf9, with dots colored according to significant overexpression (|log₂(FC)| >0.5, adj. P < 0.05) in Egr2-expressing (brown) or Tnfrsf9 -expressing (teal) cells. (D) Enrichment plots from GSEA of GSE17974 data on pathways of time-course in vitro activation of CD4⁺ T cells with anti-CD3 plus anti-CD28 for ranked genes from DEG analysis of cells in the T.4N_Nr4a1 cluster that express Egr2 versus Tnfrsf9. (E) Heatmap of the average expression of peripheral tolerance defect signature genes from WTNur and SKGNur GFP^hi cells expressing Egr2 or Tnfrsf9 in the T.4N_Nr4a1 cluster, normalized by the standard scale for each gene. Teal asterisks next to genes in the heatmap mark significant differential expression between SKGNur GFP^hi and WTNur GFP^hi cells in the Tnfrsf9 subcluster.

Figure 4. Trajectory analysis of T.4N_Nr4a1 cells orthogonally uncovers acute versus chronically antigen-activated T cell states with a distinct distribution in the SKGNur GFP^hi subset.

(A) UMAP of cells from the T.4N_Nr4a1 cluster colored according to latent time. (B and C) Smoothed gene expression analysis of cells in the T.4N_Nr4a1 cluster of selected genes with the highest expression earlier (B) or later (C) along the latent time axis. (D) Probability densities of latent time distribution of cells from the T.4N_Nr4a1 cluster assigned to 4 distinct clusters (labeled stages 1–4) by a Gaussian mixture model. (E) Predicted transitions from the partition-based graph abstraction (PAGA) algorithm between cells from the stages indicated in D. (F) Heatmap of single-cell, standard scale–normalized expression of genes ordered top to bottom by peak expression at earlier to later latent times. Chosen genes are the top 300 highest-confidence genes used in the modeling of latent time. Column annotation bar indicates stage assignment of the cell in each column. (G and H) Probability densities of latent time distribution for GFP^hi (G) and GFP^lo (H) cells from WTNur and SKGNur mice, with P values determined by Kolmogorov-Smirnov test.

Figure 5. SKG CD4⁺ T cells harbor a biased TCR Vβ gene repertoire.

(A and B) Scatterplot of the mean frequency of cells expressing each TRBV (A) or TRAV (B) gene for the SKGNur GFP^hi samples versus the WTNur GFP^hi samples. Dots for each TRBV and TRAV genes are sized according to the FDR using a 1-sided, 1-tailed paired t test with B-H correction comparing frequencies in SKGNur GFP^hi versus SKGNur GFP^lo cells. Dots are colored as either significantly enriched (FDR <0.1) in SKGNur GFP^hi cells (dark blue), significantly enriched in SKGNur GFP^lo cells (light blue), or not significantly enriched in either cell subgroup (black). Dots for significant TRBV genes are labeled with the TRBV gene name. Labels for TRBV genes that were significantly enriched in SKGNur GFP^hi cells and were also more highly expressed in SKGNur GFP^hi samples versus WTNur GFP^hi samples are bolded. (C) Bar plot of the mean value of cells expressing each TRBV gene as a percentage of all cells in each sample with an assigned TRBV. Bars are colored according to subgroup and ordered with the TRBV genes enriched in SKGNur GFP^hi cells from A, followed by the other TRBV genes ordered by increasing overall frequency. (D) Bar plots of the frequency of cells for each of the 2 replicate mice in each subgroup expressing the indicated TRBV genes which were significantly enriched in SKGNur GFP^hi. (E and F) Representative FACS plots (E) of naive peripheral CD4⁺ T cells with the indicated TCR Vβ protein usage determined by flow cytometry in GFP^lo and GFP^hi T cells from LNs of WTNur and SKGNur mice prior to arthritis induction and quantification (F), where bar graphs depict the mean frequency (± SEM). n = 3–4 mice per group. The experiment was repeated at least 3 times. *P < 0.05 and **P < 0.01, for FDR (2-tailed paired Student’s t test with B-H correction) or P value (exact permutation test).

Figure 6. Arthritogenic CD4⁺ T cells are enriched for TCR Vβs that are likely driven by endogenous Sags.

(A and B) Representative FACS plots (A) of peripheral naive or memory CD4⁺ T cells or joint CD4⁺ T cells, with the indicated TCR Vβ protein usage determined by flow cytometry in CD4⁺ T cells from dLNs or joints of SKGNur mice 2.5 weeks after arthritis induction with zymosan or mice treated with PBS vehicle (as seen in Supplemental Figure 8B) and quantification (B), where bar graphs depict the mean frequency (± SEM). (C) Bar graphs showing the GFP MFI (± SEM) of CD4⁺ T cells bearing the indicated Vβs from arthritic joints of SKG mice. n = 7 mice pooled from 2 experiments (also reported in Supplemental Figure 8D). (D and E) Representative FACS plots of (D) peripheral naive or memory CD4⁺ T cells or joint CD4⁺ T cells with the indicated TCR Vβ protein usage determined by flow cytometry in GFP^lo (light blue) and GFP^hi (dark blue) T cells from LNs or joints of SKGNur mice 2.5 weeks after arthritis induction with zymosan and quantification (E), in which bar graphs depict the mean frequency (± SEM). n = 7 mice per group pooled from 2 experiments. *P < 0.05, **P < 0.01, and ***P < 0.001, for FDR by 2-tailed paired t test with B-H correction; P value by exact permutation test (B and E); or FDR by linear mixed-effects model with B-H correction (C).

Figure 7. Sag-reactive SKG CD4⁺ T cells show impaired tolerance, defective signaling, and a Th17 differentiation bias.

(A and B) Volcano plots of DEGs within WT (A) or SKG (B) GFP^hi subsets for comparison of Sag-reactive TRBV_enriched cells with non-Sag-reactive TRBV_nonenriched cells with clored dots indicating significant overexpression (|log₂(FC)| >0.4, adj. P < 0.05) in TRBV_enriched (purple) or TRBV_nonenriched (green). (C) Volcano plots of DEGs within the TRBV_enriched cells in the T.4N_Nr4a1 cluster for comparison to WT or SKG GFP^hi mice with dots colored by significant overexpression (|log₂(FC)| >0.4, adj. P <0.05) in WT GFP^hi (orange) or SKG GFP^hi (blue) cells. (D) Left panel: Representative FACS plots of the gating strategy for peripheral CD4⁺CD25^– naive Sag-reactive (Vβ3⁺Vβ5⁺Vβ1⁺Vβ12⁺) and Sag^– (Vβ3^–Vβ5^–Vβ11^–Vβ12^–) T cells from LNs of WT or SKG mice. Right panel: Histograms showing surface marker expression in unstimulated cell subsets (quantified in Supplemental Figure 9B). Data represent 6 mice per genotype from 4 independent experiments. (E) Histograms display phosphorylated ERK (pERK) levels in Sag-reactive and non-Sag-reactive CD4⁺CD25⁻ T cells gated on naive markers (CD62L^hiCD44^lo). Cells were from WT and SKG mice after TCR crosslinking for 2 minutes with anti-CD3ε (αCD3) or stimulation with PMA plus ionomycin (PMA+I). Data represent at least 4 mice per group from 3 independent experiments. (F) Left panel: FACS plots show IL-17⁺ cell frequencies in Sag-reactive and non-Sag-reactive CD4⁺ T cells after restimulation with PMA plus ionomycin or vehicle control. Naive CD4⁺CD25^– T cells from SKG mouse LNs were cultured for 4 days under pathogenic or nonpathogenic (non-path) Th17 conditions. Right panel: Quantification of mean IL-17⁺ cell frequencies in Sag-reactive (Sag Vβ⁺) and Sag-negative (Sag Vβ^–) CD4⁺ T cell subsets. Results represent the mean ± SEM. n = 8 independent biological replicates per condition (each dot represents data pooled from 2 mice). The experiment was repeated 3 times. *P < 0.05 and **P < 0.01, by 2-tailed paired t test. Unstim, unstimulated.

Figure 8. Arthritis pathogenicity partially localizes to Sag-reactive SKG T cells.

(A) Experimental set-up: SKG mice were treated with Truvada (n = 12) or vehicle control (n = 8) on day –16 prior to arthritis induction, with i.p. administration of zymosan on day 0. (B and C) Arthritis score for SKG mice after zymosan injection (B), with results plotted as the AUC (C). (D) Arthritis-free survival plotted as a Kaplan-Meier curve for the results from A and B, which are representative of 2 independent experiments. (E) Sorted SKG CD4⁺CD25^– T cells of the indicated Vβ T cell populations were adoptively transferred into SCID mice that were monitored for arthritis development. (F–H) Arthritis score for SCID mice after adoptive transfer (F) and plotted as the AUC (P = 0.08) (G) and probability of arthritis-free survival (H). n = 6–8 mice per group. Results are representative of 2 independent experiments. *P < 0.05 and **P < 0.01, by 2-tailed Welch’s t test (C and G) or log-rank Mantel-Cox test (D and H).

Figure 9. Sag-reactive SKG T cells evade central and peripheral tolerance and contribute to SKG arthritis.

Impaired TCR signaling in SKG mice leads to a more self-reactive repertoire and escape of autoreactive, along with MMTV Sag-reactive, CD4⁺ T cells into the periphery. Chronic encounter with peripheral antigens and innate immune stimuli activates these T cells (identified as GFP^hi cells) via their TCR. As a result of impaired TCR signal transduction, SKG mice show reduced induction of TCR negative regulators and fail to have a fully established protective anergy state upon antigen encounter. Consequently, in the setting of certain environmental cues (e.g., IL-6 signaling), SKG T cells encountering endogenous antigens differentiate into pathogenic IL-17–producing effector T cells that cause erosive arthritis.