Figures and data
![](https://prod--epp.elifesciences.org/iiif/2/97665%2Fv1%2Fcontent%2F594762v1_fig1.tif/full/max/0/default.jpg)
The TFH-to-TR1 conversion is associated with massive closure of previously open chromatin regions and gene silencing. BDC2.5mi/I-Ag7-NP-induced Tet+ (TET); KLH-DNP-induced TFH (TFH), and Tconv cells (Tconv) from female NOD mice (n=4, 8 and 1 mice, respectively) were sorted and processed for 10X GEX+ATAC multiome. (A) Left: multidimensional analysis of scRNAseq and ATACseq data using weighted nearest neighbor (WNN). Cell type prediction is based on K-means clusterization and predicted cell subtype classification of scRNAseq data. Middle and right: scRNAseq (middle) and scATACseq (right) UMAP plots with K-means clusterization and predicted cell subtype classification based on scRNAseq data. The legends’ colors correspond to different T cell types and the legends’ shape correspond to sample type. (B) Hierarchical clustering of multiome data based on Euclidean distance of all the clusters found in all sample types. (C) Bar plot with the number of differentially open chromatin regions in TFH.1 (including Tet+ TFH.1 and KLH-DNP-induced TFH.1 cells), TR1/TR1-like cells, as compared to Tconv cells (adjusted P <0.05). Color depicts the proportion of OCRs that are shared with KLH-DNP-induced TFH.1 (vs Tconv cells) (yellow) or that appear de novo in TR1/TR1-like cells (grey). (D) Volcano plot of Wilcox’s differential analysis of scRNAseq data from the multiome dataset comparing TFH.1 and TR1/TR1-like cells. Not overlapping dots were labelled. Genes mentioned in the text are boxed. (E-F) Jitter plots depicting log2FC in gene expression between TR1/TR1-like and TFH.1 cells as measured by scRNAseq (adjusted P<0.05) for genes associated with OCRs found in TFH.1 cells (left; closed in TR1/TR1-like cells) or TR1/TR1-like (right; closed in TFH.1 cells) (E), or in both TFH.1 and TR1/TR1-like cells (F) as measured by scATACseq (adjusted P<0.05). Dot (gene) colors define the -log10 (adjusted P) for RNA expression of Wilcox test; dot sizes are proportional to the number of differential OCRs associated with each gene. All genes are labelled except when overlapping. Chromatin closure by TR1 cells was significantly associated with gene downregulation as determined via Chi-square test.
![](https://prod--epp.elifesciences.org/iiif/2/97665%2Fv1%2Fcontent%2F594762v1_fig2.tif/full/max/0/default.jpg)
Chromatin accessibility in BDC2.5mi/I-Ag7-NP-induced Tet+ and KLH-DNP-induced TFH cells versus gene expression. (A) Euler’s plot comparing the number and sharing of differential OCRs between BDC2.5mi/I-Ag7-NP-induced Tet+ and KLH-induced TFH cells relative to their Tconv counterparts (n=4, 3 and 3, respectively) (adjusted P<0.01). (B) Associations between the number and status of differentially open or closed chromatin sites and gene expression as measured via bulk ATACseq (adjusted P<0.01) and RNAseq (adjusted P<0.01 and |log2FC|>2), respectively. Data correspond, from left to right, to BDC2.5mi/I-Ag7-NP-induced Tet+ vs Tconv and KLH-induced TFH vs Tconv. Each dot represents a gene, and its size is proportional to the number of associated OCRs. Dot color represents RNAseq differential analysis’ -log10 of False Discovery Rate (FDR). Only genes with the highest log2FC value in each condition were labelled. (C) Associations between the status of chromatin accessibility in BDC2.5mi/I-Ag7-NP-induced Tet+ vs. KLH-induced TFH cells and gene expression as measured via bulk ATACseq (adjusted P<0.01) and RNAseq (adjusted P<0.01 and |log2FC|>2), respectively. Each dot represents a gene, and its size is proportional to the number of associated OCRs. Dot color represents RNAseq differential analysis’ -log10 of False Discovery Rate (FDR). Only genes with the highest log2FC value in each condition were labelled. P values in (B) and (C) were calculated using Chi-square.
![](https://prod--epp.elifesciences.org/iiif/2/97665%2Fv1%2Fcontent%2F594762v1_fig3.tif/full/max/0/default.jpg)
Genome-wide distribution of H3K4me3, H3K27me3, and H3K27ac marks and active enhancers in BDC2.5mi/I-Ag7-NP-induced Tet+ cells and KLH-DNP-induced TFH cells versus differential gene expression. (A) Clustering heatmaps of all the regions enriched for H3K4me3 (top), H3K27me3 (middle) and H3K27Ac (bottom) deposition in BDC2.5mi/I-Ag7-NP-induced Tet+ cells, KLH-DNP-induced TFH cells and Tconv cells. The intensity of the red color is proportional to the magnitude of enrichment vs. the corresponding background. (B) Bar plot comparing the relative percentages (X axis) and absolute numbers (number annotations in each bar) of differentially marked regions for each histone modification (FDR<0.01). Top, middle and bottom rows correspond to KLH-DNP-induced TFH vs Tconv, BDC2.5mi/I-Ag7-NP-induced Tet+ vs Tconv and BDC2.5mi/I-Ag7-NP-induced Tet+ vs KLH-DNP-induced TFH comparisons, respectively.
![](https://prod--epp.elifesciences.org/iiif/2/97665%2Fv1%2Fcontent%2F594762v1_fig4.tif/full/max/0/default.jpg)
Gene methylation status in BDC2.5mi/I-Ag7-NP-induced Tet+, KLH-DNP-induced TFH and Tconv cells. (A) Venn diagram of differentially methylated regions (DMRs) in BDC2.5mi/I-Ag7-NP-induced Tet+ and KLH-DNP-induced TFH vs. Tconv cells. (B) Venn diagrams of differentially hypo-methylated (left) or hyper-methylated (right) regions shared by BDC2.5mi/I-Ag7-NP-induced Tet+ cells and KLH-DNP-induced TFH cells as compared to their Tconv counterparts. Gene names found in the 106 TFH/TR1/Treg gene list in Supplementary Table 1, with shared methylation status, are indicated. (C) Bar plot showing the numbers of differentially hypo- or hyper-methylated regions (separated by gene region: promoter, gene body or distal intergenic) in BDC2.5mi/I-Ag7-NP-induced Tet+ and KLH-DNP-induced TFH vs Tconv cells, respectively. DMRs are classified, from left to right, into those only found in BDC2.5mi/I-Ag7-NP-induced Tet+ cells (Only Tet), KLH-DNP-induced TFH cells (only TFH), or found in both subsets (Shared by tet+ and TFH). (D) Pie-donut chart showing the distribution of DMRs (hyper and hypo-methylated status) in genes associated with OCRs (n=328) shared by TFH.1 and TR1/TR1-like cells (identified via scMultiome). (E) Jitter plot comparing differential gene expression between TR1/TR1-like and TFH.1 cells (as determined by scMultiome) and differential methylation status, focusing on the genes that remain open at the TR1/TR1-like cell stage, as determined by scMultiome. DMR-associated genes are classified based on their cell pool specificity (i.e. only found in BDC2.5mi/I-Ag7-NP-induced Tet+ cells (Only Tet+), KLH-DNP-induced TFH cells (only TFH), or both (Shared by Tet+ and TFH)). Color depicts the methylation status (hypo- or hyper-methylated) of the regions associated with these genes. No significant correlation between methylation status and differential gene expression was found.
![](https://prod--epp.elifesciences.org/iiif/2/97665%2Fv1%2Fcontent%2F594762v1_fig5.tif/full/max/0/default.jpg)
Relative contribution of different epigenetic marks to changes in the expression of differentially expressed TF-coding genes during the TFH to TR1 cell conversion.
Heatmap depicting the presence of different epigenetic marks (from left to right: differential methylation, differential OCRs, differential H3K27ac deposition, differential H3K4me3 deposition and differential H3K27me3 deposition) in BDC2.5mi/I-Ag7-NP-induced Tet+ vs KLH-DNP-induced TFH cells. Data correspond to differentially expressed TF-coding genes between TR1 and TFH.1 cells as determined by the scMultiome analyses. Differential epigenetic data is scaled for each technique and when multiple genomic regions are associated to a gene, the average is provided. Genes are arranged from most to least upregulated, followed by least to most downregulated (Fold Change). No differentially enriched sites for H3K27Ac histone deposition were associated with differential expression of these genes.
![](https://prod--epp.elifesciences.org/iiif/2/97665%2Fv1%2Fcontent%2F594762v1_fig6.tif/full/max/0/default.jpg)
TR1 cells inherit hypomethylated active enhancers from their TFH precursors. (A) Euler’s plot showing total active enhancer sharing by BDC2.5mi/I-Ag7-NP-induced Tet+ cells, KLH-DNP-induced TFH cells, and Tconv cells. Overlapping of active enhancers in BDC2.5mi/I-Ag7-NP-induced Tet+ cells and KLH-DNP-induced TFH cells was significantly higher than overlapping between BDC2.5mi/I-Ag7-NP-induced Tet+ and Tconv cells (Pearson’s Chi-squared test with Yates’ continuity correction P<2.2e-16). (B) Jitter plot comparing differential gene expression between TR1/TR1-like and TFH.1 cells (as determined by scMultiome), focusing on the genes that remain open at the TR1/TR1-like cell stage, and both distribution and number of active enhancers per gene as a function of their cell pool specificity (i.e., only found in BDC2.5mi/I-Ag7-NP-induced Tet+ cells (Only Tet+), KLH-DNP-induced TFH cells (only TFH), or both (Shared by Tet+ and TFH)). Color and size depict the region type (gene body or intergenic) and the number of active enhancers per gene, respectively. Gene names are displayed for all the genes except when more than 20 dots are in the same region of the plot. No significant correlation between active enhancer distribution and differential gene expression (Wilcox test for differential analysis: adjusted P < 0.05) was found using Pearson’s Chi-square test, P = 0.76. (C) Figure displays genome tracks for the various readouts examined herein in Tconv CD4+ T-cells (left), KLH-DNP-induced TFH cells (middle), and BDC2.5mi/I-Ag7-NP-induced Tet+ cells (right). Tracks correspond, from top to bottom, to reads for RNAseq (n=4), ATACseq (n=3), ChIPseq (n=1) for H3K4me3, H3K27me3 and H3K4me3 immunoprecipitation, respectively; ChiPseq (n=1) for Tox-2 (TFH cells), Irf4 (Th17 cells), and cMaf (Th17 cells) deposition, respectively (see main text for references), active enhancers and differential methylation, respectively. Visualization reads were normalized to total sequencing depth per sample using BPM (Bins per Million) and, where replicates were available, height mean per bin was also computed. ChIPseq data was also normalized for input (non-immunoprecipitated) sequenced reads. Height (y-axis) is equivalent to the normalized number of mapped reads in each region. Active enhancers (ACT ENH) were predicted as overlapping regions of OCR (ATACseq) and H3K27Ac deposited peaks (ChIPseq H3K27Ac) and are depicted as absent or present in each region. DIFF METH shows differentially methylated regions (n=3) obtained comparing BDC2.5mi/I-Ag7-NP-induced Tet+ cells and KLH-DNP-induced TFH to Tconv cells. Height corresponds to the relative mean methylation value for each region.
![](https://prod--epp.elifesciences.org/iiif/2/97665%2Fv1%2Fcontent%2F594762v1_fig7.tif/full/max/0/default.jpg)
Extensive sharing of differentially methylated active enhancers carrying Tox2, IRF4 and/or c-MAF-binding sites in BDC2.5mi/I-Ag7-NP-induced Tet+ and KLH-DNP-induced TFH cells vs. Tconv cells. (A) Euler’s plots with differentially methylated active enhancers for each comparison: BDC2.5mi/I-Ag7-NP-induced Tet+ cells vs Tconv (Tet+; red), KLH-DNP-induced TFH vs Tconv (TFH; orange) and Tconv vs BDC2.5mi/I-Ag7-NP-induced Tet+ cells (Tconv; green). Pie charts on the left show the methylation status (hypomethylated or hypermethylated) of each subset of active enhancers for the same subset-to-subset comparisons. The BDC2.5mi/I-Ag7-NP-induced Tet+ and KLH-DNP-induced TFH cell subsets share significantly more differentially methylated active enhancers amongst each other than with Tconv cells (Pearson’s Chi-squared test with Yates’ continuity correction; P < 2.2e-16). (B) Jitter plot of differentially methylated active enhancers in BDC2.5mi/I-Ag7-NP-induced Tet+ cells (Tet+) and KLH-DNP-induced TFH (TFH) vs Tconv cells vs. differential gene expression along the TFH.1-TR1 axis as determined by scMultiome. Plot corresponds to genes that remain accessible as TFH.1 cells transdifferentiate into TR1/TR1-like cells. The differentially methylated enhancers linked to these genes were classified as being specific for BDC2.5mi/I-Ag7-NP-induced Tet+ cells (Only Tet+), exclusive for KLH-DNP-induced TFH cells (Only TFH) or shared by both Tet+ and TFH cells. Gene bubble color depicts -log10 of adjusted P value from scRNAseq differential analysis, and bubble size depicts the number of regions associated per gene. Gene names are displayed for all the genes except when more than 20 dots are in the same region of the plot. No statistically significant associations between the presence of differentially hypo- or hypermethylated active enhancers and gene expression differences were found (Chi-square Test: P = 0.32). (C) Overlap of active enhancers with Tox2, IRF4 and c-MAF binding sites. Histogram plot compares the relative proportion of the active enhancers linked to genes that are accessible in both TR1/TR1-like and TFH.1 cells (as defined via scMultiome) and that are upregulated in TR1/TR1-like vs TFH.1 cells (as determined via scMultiome), which overlap with binding sites for Tox-2 in TFH cells, and IRF4 or c-MAF in Th17 cells. Bars correspond from left to right, to: 1) binding sites shared by both BDC2.5mi/I-Ag7-NP-induced Tet+ and KLH-DNP-induced TFH cells (Shared Tet+ & TFH); 2) exclusively found in the BDC2.5mi/I-Ag7-NP-induced Tet+ pool (Only Tet+); 3) shared by BDC2.5mi/I-Ag7-NP-induced Tet+, KLH-DNP-induced TFH and Tconv cells (Shared Tet+ & TFH & Tconv); and 4) shared by BDC2.5mi/I-Ag7-NP-induced Tet+ and Tconv cells (Shared Tet+ & Tconv).