DNA hypomethylation traits define human regulatory T cells in cutaneous tissue and identify their blood recirculating counterparts

Beumer, Niklas; Imbusch, Charles D.; Kaufmann, Tamara; Schmidleithner, Lisa; Gütter, Kathrin; Stüve, Philipp; Marchel, Harriet; Weichenhan, Dieter; Bähr, Marion; Ruhland, Brigitte; Marini, Federico; Sanderink, Lieke; Ritter, Uwe; Simon, Malte; Braband, Kathrin Luise; Voss, Morten Michael; Helbich, Sara Salome; Mihoc, Delia Mihaela; Hotz-Wagenblatt, Agnes; Nassabi, Hadrian; Eigenberger, Andreas; Prantl, Lukas; Gebhard, Claudia; Rehli, Michael; Strieder, Nicholas; Singh, Kartikeya; Schmidl, Christian; Plass, Christoph; Huehn, Jochen; Hehlgans, Thomas; Polansky, Julia K.; Brors, Benedikt; Delacher, Michael; Feuerer, Markus

doi:10.1038/s41590-025-02210-x

Download PDF

Article
Open access
Published: 16 July 2025

DNA hypomethylation traits define human regulatory T cells in cutaneous tissue and identify their blood recirculating counterparts

Nature Immunology volume 26, pages 1315–1328 (2025)Cite this article

7043 Accesses
44 Altmetric
Metrics details

Subjects

Abstract

CD4⁺ regulatory T (T_reg) cells in tissues play crucial immunoregulatory and regenerative roles. Despite their importance, the epigenetics and differentiation of human tissue T_reg cells are incompletely understood. Here, we performed genome-wide DNA methylation analysis of human T_reg cells from skin and blood and integrated these data into a multiomic framework, including chromatin accessibility and gene expression. This analysis identified programs that governed the tissue adaptation of skin T_reg cells. We found that subfamilies of transposable elements represented a major constituent of the hypomethylated landscape in tissue T_reg cells. Based on T cell antigen receptor sequence and DNA hypomethylation homologies, our data indicate that blood CCR8⁺ T_reg cells contain recirculating human skin T_reg cells. Conversely, differences in chromatin accessibility and gene expression suggest a certain reversal of the tissue adaptation program during recirculation. Our findings provide insights into the biology of human tissue T_reg cells, which may help harness these cells for therapeutic purposes.

Identification of human exT_reg cells as CD16⁺CD56⁺ cytotoxic CD4⁺ T cells

Article 10 August 2023

Tissue regulatory T cells: regulatory chameleons

Article 26 March 2021

Harnessing the biology of regulatory T cells to treat disease

Article 16 December 2024

Main

CD4⁺ regulatory T (T_reg) cells are a subset of CD4⁺ T cells that exhibit a specific epigenetic framework¹ and exert immunosuppressive effects². T_reg cells in nonlymphoid tissues are phenotypically and functionally distinct from those in lymphoid tissues^3,4,5,6,7. Nonlymphoid tissue adaptation is accompanied by the ability of tissue T_reg cells to support tissue homeostasis and repair in multiple tissues^{5,8,9,10,11,12,13,14,15}. In mice, tissue T_reg cells were reported to develop in a multistep process from a tissue T_reg progenitor that originates in lymphoid organs^4,6,16 and is dependent on several transcription factors, including BATF^3,4,10,17,18.

Although mouse tissue T_reg cells have been well characterized, human tissue T_reg cells remain incompletely understood, as few studies have focused on healthy human tissue T_reg cells^6,10. T_reg cells in tumor tissues are the target of several clinical trials that aim to deplete CCR8⁺ T_reg cells in various solid tumors (NCT05537740, NCT05635643, NCT05518045, NCT06387628, NCT05007782, NCT05101070, NCT05935098 and NCT06131398). Although CCR8⁺ T_reg cells were thought to be a cancer-specific subpopulation of human T_reg cells^19,20, recent data indicate that CCR8⁺ T_reg cells are also the predominant population of human tissue T_reg cells in nontumor tissues¹⁰. Considering that cancer therapies targeting CCR8⁺ T_reg cells may also deplete a substantial proportion of tissue T_reg cells in healthy tissues, further studies are needed to better understand the identity of tissue T_reg cells. In addition to the CCR8⁺ T_reg cells located within human tissues, a circulating population of CCR8⁺ T_reg cells exists in healthy individuals and individuals with cancer^10,21. This circulating CCR8⁺ T cell subset displays some similarities to tissue T_reg cells, including a notable amount of shared T cell antigen receptor (TCR) sequences^10,21. Epigenetic data, providing long-term information about the differentiation state of cell populations, would allow us to precisely understand the relationship between circulating and tissue CCR8⁺ T_reg cells.

Therefore, we investigated DNA methylation, chromatin accessibility (assay for transposase-accessible chromatin using sequencing (ATAC-seq)) and gene expression data from human blood and skin CD45⁺CD3⁺CD4⁺CD127⁻CD25⁺ T_reg cells at multiple levels, including the transposable element (TE) landscape. TEs are components of the genome that can, or previously could, insert themselves at new genomic locations, a process called transposition^22,23. In humans, they are estimated to account for 40–50% of the genome^24,25,26. TEs can participate in gene regulation by being transcribed into RNAs or being part of promoters and enhancers and are often inactivated by DNA methylation^27,28,29. Our analysis found a substantial number of hypomethylated TE subfamilies in human tissue T_reg cells, which could potentially be reactivated, allowing them to function as enhancers. Further, our methylation data strongly indicated that recirculating tissue T_reg cells are a part of the blood CCR8⁺ T_reg cell population.

Results

Hypomethylation defines skin and blood CCR8⁺ T_reg cells

Tissue T_reg cells are characterized by their expression of CCR8 (Fig. 1a and Extended Data Fig. 1a). To define the relationship between blood CCR8⁺ T_reg cells and tissue T_reg cells, we performed whole-genome bisulfite sequencing to investigate DNA methylation in blood antigen-inexperienced CD3⁺CD4⁺CD8⁻CD25⁺CD127⁻CD45RA⁺CD45RO⁻ T_reg cells, blood antigen-inexperienced CD3⁺CD4⁺CD8⁻CD25⁻CD127⁺CD45RA⁺CD45RO⁻ conventional T (T_conv) cells, skin CD45⁺CD3⁺CD4⁺CD8⁻CD25⁺CD127⁻ T_reg cells, skin CD45⁺CD3⁺CD4⁺CD8⁻CD25⁻CD127⁺ T_conv cells and blood CD3⁺CD4⁺CD8⁻CD25⁺CD127⁻CD45RA⁻CD45RO⁺CCR8⁺ T_reg cells from nine healthy female donors (Tissue Donors 7, 10 and 11 and Blood Donors 3–8; for age, see Methods; Fig. 1b). We analyzed 27,999,538 CpGs with a median coverage of 2–7× per sample (Extended Data Fig. 1b–d). Principal component analysis indicated that blood CD45RA⁺ T_reg and T_conv cells clustered together, away from skin T_reg cells and blood CCR8⁺ T_reg cells, which clustered together, and away from skin T_conv cells (Fig. 1c). Analysis on a chromosomal level showed that, genome-wide, skin T_reg cells and blood CCR8⁺ T_reg cells were hypomethylated compared to blood CD45RA⁺ T_reg cells (Fig. 1d). To extract methylation-level signatures that defined single cell types or groups of cell types, we examined methylation values at individual CpG sites. We specifically looked for CpG sites that separated cell types based on the corresponding methylation values. As such, we looked at the largest numeric gap in methylation between any two cell types. CpGs that met our criteria were considered ‘signature CpGs’, and groups of at least three consecutive CpGs that qualified as signature CpGs were defined as ‘signature regions’ (Fig. 1e). Depending on which cell types were hypomethylated in a signature region, the signature regions were grouped into ‘signature categories’. The signature category that defined skin T_reg cells, skin T_conv cells and blood CCR8⁺ T_reg cells was the largest, encompassing 182,204 signature regions (Fig. 1f, Extended Data Fig. 1d and Supplementary Table 1). In most signature categories, most signature regions were associated with gene elements like introns, except for the ‘skin T_reg, skin T_conv and blood CCR8⁺ T_reg’ signature category, which was dominated by intergenic regions with very low association with CpG islands (Fig. 1g). These data show that blood CCR8⁺ T_reg cells exhibit a global hypomethylation landscape similar to that of tissue T_reg and T_conv cells.

**Fig. 1: DNA methylation separates skin T_reg and blood CCR8⁺ T_reg cells from blood CD45RA⁺ T_reg cells.**

A multiomics comparison identified DMR–peak–gene links

We used publicly available single-cell ATAC-seq (scATAC-seq) data¹⁰ of skin and blood CD4⁺ T cells from four healthy female donors (ATAC Donors 1, 2, 4 and 5; specific age range of skin and blood samples unknown; Methods and Extended Data Fig. 2a,b) and publicly available RNA-sequencing (RNA-seq) data¹⁰ of sorted CD3⁺CD4⁺CD8⁻CD25⁺CD127⁻CD45RO⁻CD45RA⁺ T_reg cells and CD3⁺CD4⁺CD8⁻CD25⁻CD127⁺CD45RO⁻CD45RA⁺ T_conv cells from five healthy female donors of unknown age (Blood Donors 3–5, 9 and 10) to compare DNA methylation changes to chromatin accessibility and gene expression of blood CD45RA⁺ naive T_reg versus T_conv cells. We identified a core naive T_reg cell signature of more than 3,600 differentially methylated regions (DMRs) encompassing regions associated with classical T_reg cell genes, including TNFRSF1B, TNFRSF9, IKZF2, IKZF4 and FOXP3 (Fig. 2a, Extended Data Fig. 2c,e and Supplementary Table 2), and determined DMR-accessible peak–gene links, in which a DMR overlapped with a differentially accessible peak and the associated gene was differentially expressed (hereafter ‘DMR–peak–gene’ links). This approach resulted in 151 DMR–peak–gene links comprising 73 genes (Fig. 2a). We noticed a correlation between hypomethylation changes on one hand and increased accessibility and increased gene expression levels on the other (Fig. 2b and Extended Data Fig. 2d), including DMR–peak–gene links associated with the T_reg cell marker genes TIGIT, IL2RA, FOXP3, CTLA4 and TNFRSF1B as well as new marker genes such as DNAHB or, on the opposite site, LRRN3 and SYTL3 (Fig. 2b,c and Extended Data Fig. 2f).

**Fig. 2: A multiomics comparison defines a core T_reg cell signature.**

To validate the observed methylation differences for some of these links using a different technology, we used amplicon-based bisulfite sequencing in six independent healthy male blood donors (Amplicon Donors 1–6, age unknown). DNA extracted from CD3⁺CD4⁺CD25⁺CD127⁻CD45RA⁺ T_reg and CD3⁺CD4⁺CD25⁻CD127⁺CD45RA⁺ T_conv cells was used to amplify selected regions, including FOXP3, TIGIT, IL2RA, TNFRSF1B, LRRN3 and SYTL3, by PCR followed by bisulfite sequencing. This confirmed the methylation differences for these loci observed by whole-genome bisulfite sequencing (Fig. 2d and Supplementary Table 4). Thus, the DMR–peak-gene link analysis identified new and confirmed known T_reg cell-associated marker genes.

Skin T_reg cell tissue adaptation has a multiomic signature

To define the tissue adaptation program of human skin T_reg cells, we focused on differences between skin T_reg and blood CD45RA⁺ T_reg cells. Therefore, we compared our DNA methylation data to publicly available scATAC-seq data¹⁰ of skin and blood CD4⁺ T cells (four healthy female donors; ATAC Donors 1, 2, 4 and 5; specific age range of blood and skin samples unknown; see Methods) and publicly available RNA-seq data¹⁰ of sorted CD45⁺CD3⁺CD4⁺CD8⁻CD25⁺CD127⁻ skin T_reg cells and CD3⁺CD4⁺CD8⁻CD25⁺CD127⁻CD45RO⁻CD45RA⁺ blood T_reg cells from nine healthy female donors (Tissue Donors 7, 8, 10 and 11 and Blood Donors 3–5, 9 and 10) as described above. This comparison identified a methylation skin T_reg signature comprising 300,199 DMRs (298,457 hypomethylated regions), an accessibility skin T_reg signature comprising 8,914 peaks (3,192 significantly hyperaccessible peaks) and a gene expression skin T_reg signature comprising 6,877 genes (4,049 significantly overexpressed genes; Fig. 3a, Extended Data Fig. 3a and Supplementary Table 3).

**Fig. 3: DMR–peak–gene links define a multiomic signature of skin T_reg cell development.**

Analysis of DMR–peak–gene links identified 1,203 links and showed a notable negative correlation between hypomethylation changes and increased accessibility and gene expression (Fig. 3b and Extended Data Fig. 3b). This ensemble of DMR–peak–gene links contained 813 links (785 DMRs, 793 peaks and 408 genes) that lost methylation and gained accessibility and expression (TNFRSF8, RELB, CCR8, PRDM1 and BATF) and 223 links (210 DMRs, 211 peaks and 132 genes) that gained methylation and lost accessibility and expression (TGFBR2, SELL and IGF1R) in skin T_reg cells (Fig. 3c and Extended Data Fig. 3c). Enrichment of Hallmark gene sets³⁰ among links that lost methylation but gained accessibility and gene expression in skin T_reg cells identified processes including interleukin-2 (IL-2)–STAT5 signaling, TNF signaling and TGFβ signaling (Fig. 3d). We also analyzed genomic regions that experienced strong methylation changes during skin T_reg cell tissue adaptation that were not part of the DMR–peak–gene links and found that most regions (297,577) were hypomethylated (Extended Data Fig. 3d and Supplementary Table 3).

To investigate whether the ‘core naive T_reg’ and ‘skin T_reg’ cell signatures (as defined in Figs. 2a and 3a, respectively) were shared with T_reg cells from other tissues, we generated whole-genome methylation data from CD45⁺CD3⁺CD4⁺CD8⁻CD25⁺CD127⁻ T_reg cells isolated from the subcutaneous fat of three healthy female donors (Tissue Donors 6, 7 and 8). In addition, we used publicly available scATAC-seq¹⁰ data from subcutaneous adipose tissue CD4⁺ T cells (three healthy female donors; ATAC Donors 3–5; age unknown) and bulk RNA-seq data¹⁰ from fat T_reg cells (five healthy female donors; Tissue Donors 5–9). The skin tissue T_reg cell signature defined against blood CD45RA⁺ T_reg cells showed strong similarities between skin and fat T_reg cells (Extended Data Fig. 3e). These results suggest a conserved tissue adaptation program for human T_reg cells.

Methylation of bZIP and bHLH motifs marks skin T_reg cells

To identify transcription factors whose activity may be affected by epigenetic changes in skin T_reg cells, we analyzed the enrichment of transcription factor motifs in DMRs and differentially accessibility peaks between skin T_reg cells and blood CD45RA⁺ T_reg cells (Supplementary Table 5). A range of motifs for basic leucine zipper (bZIP) transcription factors, such as BATF or the composite Jun–AP1, was enriched in skin T_reg cell hypomethylation DMRs and skin T_reg cell hyperaccessibility peaks compared to randomly selected background sets of genomic regions³¹ (Fig. 4a,b and Extended Data Fig. 4a–c), suggesting that regions around bZIP motifs became hypomethylated and were more accessible during skin T_reg cell tissue adaptation. Furthermore, we found an enrichment of several basic helix–loop–helix (bHLH) transcription factor motifs (such as c-Myc, n-Myc and USF1) in skin T_reg cell hypomethylation DMRs, but not in skin T_reg cell hyperaccessibility peaks (Fig. 4c,d and Extended Data Fig. 4d–f), suggesting that the function of bHLH transcription factors was regulated more by changes in DNA methylation than by changes in chromatin accessibility. To determine which of these transcription factors gain activity during skin T_reg cell tissue adaptation, we analyzed the ‘transcriptomic footprints’ of their activity by measuring the enrichment of their target genes among genes that were differentially expressed between skin T_reg cells and blood CD45RA⁺ T_reg cells. Several bZIP and bHLH factors, including Jun, JunB, c-Myc and USF1, had a ‘transcriptomic footprint’ and were expressed in skin T_reg cells (Fig. 4e,f). This gain in activity might indicate an involvement in the tissue adaptation of skin T_reg cells. The genes MLPH and GNA11 contained a chromatin immunoprecipitation with sequencing (ChIP–seq)-confirmed c-Myc or USF1 binding site³² that displayed hypomethylation in skin T_reg cells compared to blood CD45RA⁺ T_reg cells (Fig. 4g and Extended Data Fig. 4g). To assess whether the c-Myc binding sites played a regulatory role for these genes, we targeted their surroundings, as determined by motif analysis and public ChIP–seq data³², using a CRISPR-based transcriptional activation system (CRISPRa) in human primary blood CD3⁺CD4⁺CD25⁺CD127⁻ T_reg cells from six donors (CRISPR Donors 1–6; three donors per guide RNA; see Methods) of unknown sex and age to activate the sites in situ. CRISPRa targeting with specific guide RNAs upregulated MLPH and GNA11 gene expression compared to scrambled control guide RNA (Fig. 4g and Extended Data Fig. 5). These data suggest a regulatory enhancer potential of hypomethylated regions containing c-Myc binding sites.

**Fig. 4: Activity of bZIP and bHLH transcription factors during skin T_reg cell development is associated with binding site hypomethylation.**

Blood CCR8⁺ T_reg cells resemble skin T_reg cells

Next, we investigated the positioning of blood CCR8⁺ T_reg cells relative to blood CD45RA⁺ T_reg cells and skin T_reg cells using our methylation, accessibility and RNA expression skin T_reg cell signatures. Analysis of the 300,199 DMRs indicated that more than 97% (293,604) of the methylation values of the DMRs in blood CCR8⁺ T_reg cells were closer to those in skin T_reg cells than to those in blood CD45RA⁺ T_reg cells (Fig. 5a, Extended Data Fig. 6a,b and Supplementary Table 6). Analysis of about 8,914 accessibility features placed blood CCR8⁺ T_reg cells closer to skin T_reg cells in less than 60% (5,285) of features (Fig. 5a and Extended Data Fig. 6a,b), whereas analysis of gene expression levels showed that less than 45% (2,988) of genes placed blood CCR8⁺ T_reg cells closer to skin T_reg cells than to CD45RA⁺ T_reg cells (Fig. 5a and Extended Data Fig. 6a,b). These differences were seen in each donor analyzed (Fig. 5b). An analysis restricted to DMR–peak–gene links revealed a similar methylation, accessibility and gene expression pattern between blood CCR8⁺ T_reg cells and skin or blood CD45RA⁺ T_reg cells (Extended Data Fig. 6c). These data indicate that the methylation landscape of blood CCR8⁺ T_reg cells was analogous to that of skin T_reg cells, whereas similarities in chromatin accessibility and gene expression were less pronounced. Regions in which skin T_reg cell traits were reflected by methylation, but not by accessibility, in blood CCR8⁺ T_reg cells included BATF, PREX1, a member of the phosphatidylinositol-3,4,5-trisphosphate signaling pathway, and SPRED2, a Ras regulator (Fig. 5c and Extended Data Fig. 6d).

**Fig. 5: DNA methylation patterns reflect similarity of skin T_reg and blood CCR8⁺ T_reg cells.**

To investigate how well the epigenetic properties of blood CCR8⁺ T_reg cells reflected traits linked to transcription factor motifs, we examined DMRs and differentially accessible peaks from the ‘skin T_reg cell hypomethylation’ and ‘skin T_reg cell hyperaccessibility’ signatures as defined above and performed motif enrichment analysis to assess whether transcription factor motifs were enriched in these DMRs and peaks compared to background sets of genomic regions³¹. bZIP motifs were significantly enriched in DMRs that placed blood CCR8⁺ T_reg cells closer to skin T_reg cells than to blood CD45RA⁺ T_reg cells (Fig. 5d,e and Extended Data Fig. 7a), indicating that the epigenetic characteristics surrounding bZIP motifs in skin T_reg cells were well reflected by the methylation level in blood CCR8⁺ T_reg cells. These effects were less pronounced when analyzing chromatin accessibility data around the genomic sites of Jun–AP1(bZIP) and other bZIP motifs (Fig. 5d,f and Extended Data Fig. 7b). We also observed an enrichment of many bHLH motifs among DMRs that placed blood CCR8⁺ T_reg cells closer to skin T_reg cells than to blood CD45RA⁺ T_reg cells (Extended Data Fig. 7c,d), indicating that skin T_reg cell-related methylation traits of bHLH motifs were reflected by blood CCR8⁺ T_reg cells.

To assess whether loss of conservation of skin T_reg cell signature gene expression in blood CCR8⁺ T_reg cells could be detected in identical αβ TCR clones, we analyzed previously published scRNA-seq and scTCR-seq data¹⁰ from skin T_reg cells and blood CCR8⁺ T_reg cells from two healthy female donors (TCR Donors 6 and 7, unknown age; Extended Data Fig. 8a–d). We identified 198 (TCR Donor 6) and 488 (TCR Donor 7) skin T_reg cells that shared TCR sequences with blood CCR8⁺ T_reg cells compared to 565 (TCR Donor 6) and 818 (TCR Donor 7) blood CCR8⁺ T_reg cells that shared TCR sequences with skin T_reg cells (Fig. 5g). In total, 4–10% of all identified blood CCR8⁺ T_reg TCR clonotypes could also be found in skin T_reg cells, whereas only about 0.3% of all identified blood CD45RA⁺ T_reg TCR clonotypes were also found in skin T_reg cells (Extended Data Fig. 8e). Furthermore, more than 38% of TCR clonotypes found in skin T_reg cells were also found in blood CCR8⁺ T_reg cells, compared to less than 2.5% in blood CD45RA⁺ T_reg cells (Extended Data Fig. 8f), indicating considerable clonal TCR overlap between blood CCR8⁺ T_reg cells and skin T_reg cells. To determine the conservation of gene expression patterns in TCR-identical clones between blood CCR8⁺ T_reg cells and skin T_reg cells, we calculated the aggregated expression of genes from the skin T_reg signature (‘skin T_reg cell hyperexpression’ genes and ‘blood CD45RA⁺ T_reg hyperexpression’ genes) into module scores per cell. This analysis indicated that loss of the skin T_reg cell-specific gene expression program and gain of a more naive phenotype occurred once clones left the skin (Fig. 5g).

In mice, CCR8 is a marker for tissue T_reg cells, similar to in humans¹⁰. To experimentally test whether CCR8⁺ T_reg cells represented recirculating tissue T_reg cells, we used Tg(CAG-Kaede)15Kgwa mice, which ubiquitously express the photoconvertible fluorescence protein Kaede (which switches from green to red after exposure to violet light) under the control of a CAG promoter³³. On day 6, after exposing the back skin to 5 min of violet light, we found photoconverted Kaede-red⁺ migratory skin T_reg cells in the skin draining lymph nodes (dLNs; Extended Data Fig. 8g), as has been previously observed³⁴. Approximately 65% of these photoconverted Kaede-red⁺ T_reg cells expressed CCR8 in the dLNs, whereas only 30% of nonconverted Kaede-green⁺ T_reg cells expressed CCR8 (Fig. 5h,i and Extended Data Fig. 8h), indicating that the CCR8⁺ T_reg cell pool is enriched for tissue egressing, recirculating T_reg cells. These results showed a higher conservation of skin T_reg cell methylation traits in blood CCR8⁺ T_reg cells than DNA accessibility and gene expression traits and suggested that blood CCR8⁺ T_reg cells could represent recirculating tissue T_reg cells that lost part of their accessibility and gene expression tissue characteristics.

TE hypomethylation characterizes tissue T_reg cells

TEs contribute to gene regulatory functions, and DNA methylation suppresses their activity^28,35,36. TE elements can be subdivided into classes (including long interspersed nuclear elements (LINEs), long terminal repeats (LTRs), short interspersed nuclear elements (SINEs) and DNA elements) and further into families and subfamilies (Supplementary Table 7). In total, more than 1,100 TE subfamilies can be distinguished with individual integration sites, ranging from a few dozen integration sites to more than 100,000 integrations per subfamily^37,38,39. We asked which TE classes and subfamilies were enriched in epigenetically differential regions of skin T_reg cells versus blood CD45RA⁺ T_reg cells. We found that the TE classes SINE, LTR and DNA as well as 323 TE subfamilies, including LTR18B, HUERS-P2-int, LTR3B, LTR1D, LTR32 and MER125, were significantly enriched in ‘skin T_reg cell hypomethylation’ DMRs compared to the TE distribution expected by chance (Fig. 6a, Extended Data Fig. 9a and Supplementary Table 8). Significant TE enrichment was also detected at the accessibility level in skin T_reg cells (Fig. 6b and Extended Data Fig. 9b). We also directly examined methylation and accessibility around annotated TE insertion sites in skin T_reg and blood CD45RA⁺ T_reg cells. LTR, SINE and DNA TEs were strongly affected by hypomethylation changes in skin T_reg cells (Fig. 6c). In all three TE classes, more TE insertion sites were hypomethylated in skin T_reg cells than in blood CD45RA⁺ T_reg cells, for example, 86% for LTR class TE insertion sites (Fig. 6c and Extended Data Fig. 9c). Similar hypomethylation patterns were also observed in subcutaneous adipose tissue T_reg cells (Extended Data Fig. 9d). Skin T_reg cell signature genes, such as TNFRSF8, PREX1, PRDM1 and BATF, contained hypomethylated TE insertion sites for several TE subfamilies, such as MIR and L2a (Fig. 6d and Extended Data Fig. 9e).

**Fig. 6: TEs are hypomethylated in skin T_reg cells and blood CCR8⁺ T_reg cells.**

Next, we investigated the TE traits of blood CCR8⁺ T_reg cells compared to skin T_reg and blood CD45RA⁺ T_reg cells. DMRs and differential peaks were stratified based on their positioning of blood CCR8⁺ T_reg cells, and we examined whether TEs were enriched among the stratified features. SINE, LINE, LTR and DNA TE classes and 38 individual TE subfamilies, including MIR and L2a, were enriched in DMRs that placed blood CCR8⁺ T_reg cells closer to skin T_reg cells (Fig. 6e). We did not observe any significant TE enrichment at the accessibility level (Fig. 6f), suggesting that TE-related traits in blood CCR8⁺ T_reg cells reflected skin T_reg cell traits at the methylation, but not chromatin accessibility, level. Because TE enrichments may have been biased by different underlying region numbers, we also examined methylation around individual TE insertion sites. The hypomethylation pattern in LTR, SINE, LINE and DNA TEs in CCR8⁺ T_reg cells, but not the hyperaccessibility pattern, was almost indistinguishable from the pattern of skin T_reg cells compared to blood CD45RA⁺ T_reg cells (Fig. 6g and Extended Data Fig. 9f). These findings indicate that TEs represent an integral constituent of the hypomethylation landscape in tissue T_reg cells and that blood CCR8⁺ T_reg cells conserve the TE hypomethylation traits found in tissue T_reg cells.

TE subfamilies gain RNA expression in skin T_reg cells

Because TEs can be transcribed into RNA²³, we investigated how epigenetic methylation changes of TE insertion sites influenced TE expression during skin T_reg cell tissue adaptation. We matched RNA-based sequencing reads to TE consensus sequences and quantified the expression of TEs in the bulk RNA-seq data¹⁰. TEs inserted into larger transcribed regions (genes) can bias this analysis because their expression might be due to transcription of the larger region and not expression of the TE. To avoid such biases, we excluded TE reads that mapped to gene bodies or long noncoding RNAs (lncRNAs). We detected a median number of 1,071 (range: 625–1,561) TE-derived reads across the analyzed samples (Supplementary Table 9). Of 1,103 profiled TEs, 613 displayed detectable expression in at least one cell type (skin T_reg, skin T_conv, blood CD45RA⁺ T_reg, CD45RA⁺ T_conv and CCR8⁺ T_reg cells; Fig. 7a and Supplementary Table 9). Analysis of TEs that changed transcription during skin T_reg cell tissue adaptation identified three TEs that were overexpressed in skin T_reg cells compared to in blood CD45RA⁺ T_reg cells (Fig. 7b), including the TE subfamilies HERVIP10F (labeled ‘HERVIP10F-int’ in the insertion site database) and LTR45B. The insertion sites of HERVIP10F and LTR45B were hypomethylated in skin and subcutaneous fat T_reg cells (Fig. 7c and Extended Data Fig. 10a,b). Blood CCR8⁺ T_reg cells reflected this trend (Fig. 7c). We identified 103 HERVIP10F-int and 72 LRT45B insertion sites that overlapped with ‘skin T_reg cell hypomethylation’ DMRs (Supplementary Table 10). The most strongly hypomethylated insertion sites included a position on chromosome 4, with 74% methylation reduction in skin T_reg cells compared to blood CD45RA⁺ T_reg cells (Table 1). Thus, the TE landscape is an integral part of the hypomethylation traits of tissue T_reg cells, shared by circulating CCR8⁺ T_reg cells.

**Fig. 7: Several TE subfamilies are transcribed in skin T_reg cells.**

Table 1 Locations and methylation of the most strongly hypomethylated HERVIP10F-int and LTR45B insertion sites overlapping with DMRs hypomethylated in skin T_reg cells

Full size table

Discussion

Based on epigenetic DNA analyses, we identified tissue adaptation programs, often involving specific TE subfamilies, in human tissue T_reg cells. We found that CCR8⁺ circulating blood T_reg cells shared a strong epigenetic relationship with tissue T_reg cells, characterized by a similar hypomethylation pattern, while they differed in their chromatin accessibility and RNA expression profile. This indicated that CCR8⁺ T_reg cells in the blood contained recirculating tissue T_reg cells showing signs of a reversal of the tissue adaptation program toward a more naive phenotype.

In mice, T_reg cells are characterized by DNA hypomethylated T_reg cell signature genes, which are essential for T_reg cell functionality, long-term lineage stability¹ and tissue-specific functional adaptations^3,4,7,10. Here, we described about 300,000 DMRs between human skin tissue T_reg cells and naive CD45RA⁺ blood T_reg cells. The sheer number of methylation changes, which we also found in subcutaneous adipose T_reg cells, indicates that T_reg cell tissue adaptation goes along with marked epigenetic reconfiguration. Notably, tissue adaptation programs were dominated by hypomethylation changes. Dominance of hypomethylation was reported for developing lymphocytes and immune cell proliferation⁴⁰, which has been attributed in part to specific late-replicating regions called ‘partially methylated domains’^41,42.

TEs can influence immunological processes by providing an enhancer landscape^39,43. Reactivated TE RNA or DNA can be sensed by the innate immune system, resulting in pathological inflammation^44,45. Research into how TEs affect gene regulation by influencing the enhancer landscape in immune cells is just emerging. It was reported that interferon responses rely on TE co-option⁴⁶, and CD8⁺ T cell activation exploits the TE enhancer architecture⁴⁷. Specific TE families were reported to be associated with tissue adaptation of immune cells³⁸. Here, we found similar patterns in human skin and subcutaneous adipose T_reg cells. We found TE RNA, which was indicative of the reactivation of TE subfamilies, associated with hypomethylation of TE insertion sites in skin tissue T_reg cells.

Our data showed a strong relationship between human circulating CCR8⁺ T_reg cells and tissue T_reg cells, suggesting that blood CCR8⁺ T_reg cells are the recirculating counterpart of tissue T_reg cells. Using cell tracking in mice, we found that dLN CCR8⁺ T_reg cells contained skin recirculating tissue T_reg cells. Therefore, blood CCR8⁺ T_reg cells could become a valuable diagnostic tool, as these cells may have witnessed alterations in a tissue, such as an imminent immune rejection in a transplant situation. The strong epigenetic relationship observed in the methylation data was only partly recapitulated by chromatin accessibility traits, suggesting that methylation changes are a more permanent trait than chromatin accessibility, as previously reported⁴⁸. A recirculating population of human tissue T_reg cells would not be unexpected, as data in mice point in this direction⁴⁹. Mouse parabiosis experiments suggested that depending on the tissue, the mean dwell time of tissue T_reg cells within a certain tissue is 2–8 weeks (ref. ⁴⁹), and work on mouse muscle T_reg cells showed shared muscle T_reg cell TCR sequences in a specific spleen T_reg cell population, compatible with recirculating muscle T_reg cells¹². Our multiomic analysis indicated that recirculating tissue T_reg cells show signs of a regression of the tissue adaptation program toward a more naive state, resembling a precursor-like phenotype, making the distinction between true precursor tissue T_reg cells that never entered the tissue and egressing tissue T_reg cells difficult. The precursor-like phenotype may allow circulating tissue T_reg cells to better readapt to new challenges, while maintaining certain methylation traits that could enforce a more pan-tissue T_reg cell identity. The existence of a pan-tissue T_reg cell identity across various organs and even between human and mouse tissue T_reg cells has been reported^3,4,7,10,49. In conclusion, our data provide insights into the tissue adaptation of human tissue T_reg cells, which may have important implications for therapeutic applications of T_reg cells and disease diagnostics.

Methods

Human donors, sample acquisition, sample sizes and ethics

Human skin and subcutaneous adipose (fat) tissue used for whole-genome bisulfite sequencing and RNA-seq was obtained from seven healthy female donors (named Tissue Donors 5–11) undergoing abdominoplasty procedures after weight loss or epigastric hernia repair after multiple pregnancy. For whole-genome bisulfite sequencing, samples from three donors per cell type were processed. We analyzed skin T_reg cells from Tissue Donors 7, 10 and 11 (median age: 52 years; range: 41–53 years), skin T_conv cells from Tissue Donors 7, 10 and 11 (ages as for skin T_reg cells) and fat T_reg cells from Tissue Donors 6, 7 and 8 (age not available for each donor). For bulk RNA-seq, we profiled samples from four to five donors per cell type. We analyzed skin T_reg cells from Tissue Donors 7, 8, 10 and 11 (age not available for each donor), skin T_conv cells from Tissue Donors 7, 8, 10 and 11 (age not available for each donor) and fat T_reg cells from Tissue Donors 5, 6, 7, 8 and 9 (age not available for each donor). These bulk RNA-seq data have been published previously¹⁰. For flow cytometry analysis, blood and skin samples were obtained from seven healthy female donors undergoing abdominoplasty procedures after weight loss and epigastric hernia repair. For statistical analysis, we used samples from FACS Donors 1–6 (median age: 46 years; range: 31–61 years). For exemplification of our gating strategy, we used samples from FACS Donor 7 (age unknown). Collection of tissue, immune cells and blood samples from these donors was performed in accordance with the Declaration of Helsinki after ethical approval by the local ethics committee at Regensburg University, Germany (reference number 19-1453-101), and signed informed consent.

To obtain human blood samples for whole-genome bisulfite sequencing and RNA-seq, peripheral blood mononuclear cells for CD4⁺ T cell enrichment were isolated from leukocyte reduction chambers from ten healthy female donors of unknown age who donated thrombocytes (termed Blood Donors 3–12). For whole-genome bisulfite sequencing, samples from three donors per cell type were used (blood CD45RA⁺ T_reg cells from Blood Donors 3–5, blood CD45RA⁺ T_conv cells from Blood Donors 3–5 and blood CCR8⁺ T_reg cells from Blood Donors 6–8). For bulk RNA-seq, we profiled three to five samples per cell type (blood CD45RA⁺ T_reg cells from Blood Donors 3–5, 9 and 10, blood CD45RA⁺ T_conv cells from Blood Donors 3–5 and blood CCR8⁺ T_reg cells from Blood Donors 6–8, 11 and 12). These bulk RNA-seq data have been previously published¹⁰. For CRISPR activation, we analyzed T cells from six donors of unknown sex and age. Cells from CRISPR Donors 1–3 were used for the guide RNAs hMLPH_A and hGNA11, and cells from CRISPR Donors 4–6 were used for the guide RNA hMLPH_B. For amplicon-based bisulfite sequencing, we processed samples from six male donors of unknown age. These donors were termed ‘Amplicon Donors 1–6’. Collection of blood samples was performed in compliance with the Declaration of Helsinki after ethical approval by the local ethical committee (reference numbers 19-1414-101 and 19-1614-101) and signed informed consent. Donors provided written informed consent, and their samples were not used for any other analyses in this study.

No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications⁵⁰. Donors were not assigned to different experimental groups, and no randomization was performed. Data collection and analysis were not performed blind to the conditions of the experiments.

In addition, we used public scATAC-seq and scRNA-seq/scTCR-seq data¹⁰. scATAC-seq data contained data on CD4⁺ T cells from blood, skin and subcutaneous fat from five healthy female donors (blood from ATAC Donors 1 and 2, skin from ATAC Donors 4 and 5 and fat from ATAC Donors 3–5) with an average age of 44.6 years (±14; range from 26 to 56 years, ages for individual donors unknown) undergoing abdominoplasty procedures after weight loss and epigastric hernia repair. scRNA-seq/scTCR-seq data contained data on CD4⁺ T cells from blood, skin and subcutaneous fat from two healthy female donors (termed TCR Donors 6 and 7) of unknown age who underwent abdominoplasty procedures after weight loss and epigastric hernia repair. Collection of tissue, immune cells and blood samples from these donors was performed in accordance with the Declaration of Helsinki after ethical approval by the local ethics committee at Regensburg University, Germany (reference number 19-1453-101), and signed informed consent.

All animal experiments were approved by the governmental committee for animal experimentation (Regierungspräsidium Unterfranken) and were conducted in accordance with ethical regulations. Specific details on animal characteristics are provided in a dedicated section below.

Tissue digestion for flow cytometry and FACS of T cells

To isolate T cells from human skin tissue, skin and underlying fat were first mechanically separated. The skin was then cut into small pieces and digested (base medium DMEM (Gibco, 41965), 4 mg ml⁻¹ collagenase type IV (Sigma-Aldrich, C5138), 10 µg ml⁻¹ DNase I (Roche, 11284932001), 2% fetal bovine serum and 10 mM HEPES). Digestion was performed directly in a GentleMACS C tube with the program ‘37_C_Multi_H’ for 90 min, followed by centrifugation and filtration steps. Dead cell removal was performed with a dead cell removal kit (Miltenyi, 130-090-101).

Peripheral blood mononuclear cell isolation and pre-enrichment of blood lymphocytes

To isolate T cells from human blood, leukocyte reduction chambers or buffy coats were used. Leukocytes were first diluted 1:3 with DPBS (Gibco, 14190-094), and the resulting blood and PBS mixture was split into two fractions and underlaid with an equal amount of Pancoll (PAN-Biotech, P04-601000). Samples were centrifuged at 1,000g for 20 min at room temperature, with acceleration set to 4 and brake to 0. The peripheral blood mononuclear cell layer was isolated and washed twice by centrifugation. Antibodies were used, if not indicated otherwise, as recommended by the manufacturer (including amounts and dilutions). Cells were pre-enriched with biotinylated anti-human CD4 (clone OKT4, Biolegend, 317406), biotinylated anti-human CD8 (clone HIT8a, Biolegend, 300904) or PE-labeled or biotinylated anti-human CD25 (clone BC96, Biolegend, 302606 or 302624) at a dilution of 1:20 or 1:100, followed by column-based magnetic separation with anti-biotin or anti-PE ultrapure microbeads (Miltenyi Biotec, 130-105-637 or 130-105-639) following the manufacturer’s protocol. All antibodies were from commercial sources and were validated by the manufacturer.

Preparation of samples for FACS

T cells were isolated and pre-enriched as described in the previous sections. Cells were stained in 1.5-ml Eppendorf tubes or 96-well plates in FACS buffer (2% fetal calf serum in PBS). Surface staining was performed at 4 °C for 20 min in a staining volume of 50–100 µl. Antibodies were used, if not indicated otherwise, as recommended by the manufacturer (including amounts and dilutions). All antibodies were from commercial sources and were validated by the manufacturer. The following antibodies were used for surface staining: anti-human CD3-BV785 (OKT3, Biolegend, 317330 or BD Biosciences, 566782), anti-human CD4-BV711 (OKT4/L200, Biolegend, 317440 or BD Biosciences, 563913), anti-human TCR-β-chain BV785 (IP26, Biolegend, 306742), anti-human CD8-BUV395 (RPA-T8, BD Biosciences, 563795), anti-human CD19-BV605 (HIB19, Biolegend, 302244), anti-human CD25-PE (BC96/2A3, Biolegend, 302606 or BD Biosciences, 57214), anti-human CD45-BUV737 (HI30, BD Biosciences, 748719), anti-human CD45RA-BV605 or CD45RA-BV510 (HI100, Biolegend, 304134/304142 or BD Biosciences, 562886), anti-human CD45RO-BV421 (UCHL1, Biolegend, 304224), anti-human CD127-APC (A019D5, Biolegend, 351342), anti-human CD14 BUV395 (MφP9, BD Biosciences, 563562) and anti-human CCR8-BV421 or BUV395 (433H, BD Biosciences, or 747573) at a dilution of 1:20 or 1:100. Viability staining was performed using eBioscience Fixable Viability Dye eFluor 780 (Thermo Fisher Scientific, 65-0865-14) according to manufacturer’s instructions at a dilution of 1:1,000.

FACS of T cells from blood and skin

T cells were isolated, pre-enriched and stained as described above. BD CS&T beads were used to validate machine functionality. Fluorescence spillover compensation was performed with lymphocytes stained with CD4 (OKT4, Biolegend or BD Biosciences) in the respective colors. Flow cytometry data were analyzed using FACSDiva (version 8.0.2) BD FlowJo (versions 10.6.2 and 10.10.0). Sorting was performed with a BD FACSAriaII or BD FACSFusion cell sorter with an 85-µm nozzle and 45 psi of pressure. Postsort quality controls were performed as applicable. Cells were sorted directly into 500-µl of RLT+ lysis buffer (Qiagen AllPrep DNA/RNA Mini kit, 80204). All procedures were performed in DNA low-bind tubes (Eppendorf, 0030108051). From human skin samples, we sorted 3,500 to 10,000 T_reg cells (CD45⁺CD3⁺TCRβ⁺CD4⁺CD8⁻CD127⁻CD25⁺) and 5,500 to 10,000 T_conv cells (CD45⁺CD3⁺TCRβ⁺CD4⁺CD8⁻CD127⁺CD25⁻). From human peripheral blood samples, we sorted 10,000 CD45RA⁺ T_reg cells (CD45⁺CD3⁺TCRβ⁺CD4⁺CD8⁻CD127⁻CD25⁺CD45RA⁺CD45RO⁻), 10,000 CCR8⁺ T_reg cells (CD45⁺CD3⁺TCRβ⁺CD4⁺CD8⁻CD127⁻CD25⁺CD45RA⁻CD45RO⁺CCR8⁺) and 10,000 CD45RA⁺ T_conv cells (CD45⁺CD3⁺TCRβ⁺CD4⁺CD8⁻CD127⁺CD25⁻CD45RA⁺CD45RO⁻).

Whole-genome bisulfite sequencing

DNA was isolated using a DNA/RNA micro kit (Qiagen, 80284), and DNA quality and concentration were assessed by Qubit (dsDNA HS kit, Q32851) and TapeStation 4200 with Genomic DNA Screentape (Agilent, 5067-5365). Tagmentation-based whole-genome bisulfite sequencing was essentially performed as described previously⁵¹. Two differently barcoded libraries were generated each per sample. Corresponding libraries were pooled in equimolar amounts, and each pool was sequenced on a single lane using an Illumina Novaseq 6000 with paired-end sequencing and 150 cycles (PE150).

scATAC-seq

The scATAC-seq data set was taken from our previous study¹⁰.

Bulk RNA-seq

The RNA-seq data set was taken from our previous study¹⁰. This data set had been generated as follows. Total RNA was isolated using a DNA/RNA micro kit (Qiagen, 80284), and RNA was eluted in 14 µl of RNase-free water. RNA quality was assessed using a TapeStation 4200 and High-Sensitivity RNA Screentape (Agilent, 5067-5579). Eight microliters of RNA was used for generating RNA-seq libraries using a SMART-Seq Stranded kit (Takara, 634444). Indexed libraries were pooled in an equimolar ratio and sequenced on an Illumina NextSeq 550 machine with a NextSeq 500/550 High Output kit v2.5 (75 cycles). DNA was eluted in 30 µl of buffer EB and used for whole-genome bisulfite sequencing as described above.

Generation of amplicon data for validation of methylation differences

Custom bisulfite primers were designed using the bisulfite primer seeker (Zymo Research; for primer sequences, see Supplementary Table 4). The forward primers were extended with barcode sequences for multiplexing (Supplementary Table 4). In addition, forward and reverse primers were extended with partial Illumina sequences as recommended by Genewiz, from Azenta Life Sciences, where sequencing was performed (5′–3′ forward: ACACTCTTTCCCTACACGACGCTCTTCCGATCT; reverse: GACTGGAGTTCAGACGTGTGCTCTTCCGATCT). Primers were synthesized by Sigma-Aldrich. DNA was extracted using a DNeasy Blood and Tissue kit (Qiagen, 69504). Afterward, bisulfite conversion was performed using an EpiTec Bisulfite kit (Qiagen, 59104). For amplicon generation, KAPA HiFi HotStart ReadyMix was used (Roche, 7959079001). Together with bisulfite-converted DNA, barcoded forward and reverse primers and double-distilled water, PCRs were run (initial denaturation at 95 °C for 3 min, 35 cycles start denaturation at 98 °C for 20 s, annealing at primer optimal temperature for 15 s, extension at 72 °C for 15 s, cycle end, final extension at 72 °C for 1 min and hold indefinitely at room temperature). To clean up the PCR products, 1% agarose gels were run, and a MiniElute Gel Extraction kit was used to extract the amplicons (Qiagen, 28604). Afterward, Qubit and TapeStation were run to measure concentration and average base pair size. Taking the barcodes and regions into account, the different amplicons were pooled and sequenced on an Illumina MiSeq machine, yielding 2 × 250 bp paired-end reads. The analysis of the corresponding data is described in ‘Bioinformatics: general’.

CRISPR-based transcriptional activation of MLPH and GNA11

Blood CD45RA⁺ T_reg cells were isolated from peripheral blood mononuclear cells by gradient centrifugation, followed by MACS and FACS isolation as described above. Cells were rested with 100 U ml⁻¹ IL-2 for 24 h before being transduced with dCas9-VP64 and guide RNA using separate lentiviral vectors. Guide RNA sequences were 5′TTAGATGGGAGCTATGTGTG3′ for MLPH_A, 5′TGCTAGTGAGATTTGAGCAC3′ for MLPH_B and 5′AGCCCCTGACCCGAGACTGA3′ for GNA11. A random guide RNA sequence (‘scrambled’; 5′GACCATCTTCTAACCTTACA3′) was used as a negative control. After 3–6 days of incubation, cells transduced with both dCas9-VP64 and guide RNA were sorted by FACS based on expression of co-transduced reporter genes. RNA isolation was performed using an innuPREP RNA Mini kit according to the manufacturer’s specifications. cDNA synthesis was performed according to the manufacturer’s protocol (Reverse Transcriptase II, Life Technologies). Quantitative PCR with reverse transcription was performed with TaqMan probes and master mix in a Viia7 real-time PCR system (all Thermo Fisher Scientific). Gene expression was normalized to housekeeping gene expression (HPRT).

To control for efficient and specific targeting of the guides, classic CRISPR–Cas9 constructs able to cleave DNA containing the same guide RNAs were generated. Guide RNA was generated by mixing transactivating CRISPR RNA (Integrated DNA Technologies, 1072534) and MLPH CRISPR RNAs (sequence: 5′TTAGATGGGAGCTATGTGTG3′), GNA11 CRISPR RNAs (sequence: 5′AGCCCCTGACCCGAGACTGA3′) or scrambled CRISPR RNAs (Integrated DNA Technologies, 1072544) in equimolar concentrations, heating to 95 °C for 5 min and allowing to cool down to room temperature. Guide RNA mixes are combined with Cas9 Enzyme (Integrated DNA Technologies, 1081060) and incubated at room temperature for 30 min to form the RNP complex. Electroporation was performed using a Neon transfection instrument (settings: 1,600 V, 10-ms pulse width, three pulses; Thermo Fisher Scientific) in the presence of an electroporation enhancer (Thermo Fisher Scientific, 1075915) using a Neon transfection system 10 μl kit (Thermo Fisher Scientific, MPK1096). CD4⁺ T cells were cultured with 100 U ml⁻¹ IL-2 and Transact for 3 days before genomic DNA was isolated using a QIAmp DNA Mini kit (Qiagen, 51304). Target DNA was PCR amplified using the following primers: 5′CCACACACACACACGCTATTGGCTCTGC3′ and 5′GCGGACCGTGGCTGTGGCCTGTG3′ for MLPH and 5′CCACACACACACACGCTATTGGCTCTGC3′ and 5′GCGGACCGTGGCTGTGGCCTGTG3′ for GNA11. PCR amplicons were Sanger sequenced, and Tracking of Indels by Decomposition (http://shinyapps.datacurators.nl/tide/) confirmed specific targeting of both guides to the respective genomic position.

Origin, treatment and analyses of Kaede-expressing mice

Tg(CAG-Kaede)15Kgwa mice³³ (genetic background: C57BL/6J) were a gift from R. Linker (Clinic and Polyclinic for Neurology at the University of Regensburg). Animals were housed under specific pathogen-free conditions at the Regensburg University Clinic animal care facility, and the governmental committee for animal experimentation (Regierungspräsidium Unterfranken) approved all experiments involving animals. We used seven adult male and female mice between 30 and 35 weeks of age. Air exchange rates were >8 AC h⁻¹, light/dark cycles were 12 h/12 h, and mice were maintained at ambient temperature (22 ± 2 °C) and a relative humidity of 55 ± 10%. Tg(CAG-Kaede)15Kgwa mice express the photoconvertible fluorescent Kaede protein under the control of a CAG promoter. The excitation maximum of ‘Kaede-green’ was 508 nm. During exposure to violet light, the Kaede-green protein was converted to ‘Kaede-red’, which is characterized by an excitation maximum of 572 nm (ref. ⁵²). For photoconversion of skin, Tg(CAG-Kaede)15Kgwa mice were anesthetized, hair and hair follicles from the back were removed with an electric shaver and depilatory cream, and an area of 8 cm² was exposed to violet light for 5 min (100% intensity at a distance of 10 cm using a CoolLED lamp from the pE-100 series with 440 nm; AHF). Six days after initial exposure, animals were killed, and skin dLNs (inguinal and axillary) were collected. Randomization was not applicable as mice were not assigned to different experimental groups. For the same reason, data collection and analysis were not performed blind to the conditions of the experiments. T cells were isolated from LNs by mechanical dissociation on a 40-µm filter. Cells were pre-enriched for CD25 by staining biotinylated anti-mouse CD25 (clone PC61, Biolegend, 102003), followed by column-based magnetic separation with anti-biotin microbeads (Miltenyi Biotec, 130-105-637) following the manufacturer’s protocol. The following mouse antibodies were used for surface staining: CD4-BUV395 (GK1.5, BD Biosciences, 563790), TCRβ-BV605 (H57-597, Biolegend, 109241), GITR-PE-Cy7 (DTA-1, Biolegend, 126318) and CCR8-BV421 (SA214G2, Biolegend, 150305) or Rat IgG2bk BV421 Isotype Ctrl (RTK4530, Biolegend, 400655) at a dilution of 1:100. Viability staining was performed using eBioscience Fixable Viability Dye eFluor 780 (Thermo Fisher Scientific, 65-0865-14) according to manufacturer’s instructions at a dilution of 1:1,000. Data analysis was performed using FlowJo v10.10 and GraphPad Prism v10.4.1.

Bioinformatics: general

Software versions and corresponding references, if available, are indicated upon first mention of the respective software. All analyses that used the programming language R were performed using R v4.0.0.

Preprocessing of whole-genome bisulfite sequencing data

Alignment and methylation calling was performed based on a previously described workflow⁵³ that uses a customized version of methylCtools⁵⁴. The workflow was modified as follows. Trimmomatic⁵⁵ (v0.30) PE was used to trim adaptor sequences with the options ‘-threads 12 -phred33 ILLUMINACLIP:xxx:2:30:10:8:true SLIDINGWINDOW:4:15 MINLEN:36’. Here, ‘xxx’ refers to a fasta file containing the adaptor sequences. Alignment to the hs37d5 reference genome (based on GRCh37) was performed using BWA-MEM⁵⁶ (v0.7.8) with the ‘-T’ parameter set to 0. Duplicates were marked using sambamba (v0.6.5) with the parameters ‘-t 1 -l 0–hash-table-size = 2000000–overflow-list-size = 1000000–io-buffer-size = 64’ and not removed from the data. We confirmed successful bisulfite conversion in all but one sequencing library by observing low methylation signal at CH sites (Extended Data Fig. 1b). Here, we restricted our analysis to the first 3,000,000 CH sites on chromosome 1. The sequencing library that did not meet this quality criterion was subsequently excluded. For the analysis of CpGs, we combined data on the two opposite cytosines of the same CpG by summing up their read counts. To compute position-wise coverage, we added the counts of methylated and unmethylated reads at each profiled position. One skin T_reg cell sample displayed exceptionally low coverage (Extended Data Fig. 1c) as we had excluded the library with aberrant CH methylation signal. Because all donors were female, we excluded positions on chromosome Y. We then stored the counts of methylated and total reads at each position in a bsseq object (bsseq⁵⁷ package; v1.26.0). We computed smoothed methylation values using the BSmooth function from the bsseq package.

Preprocessing of RNA-seq data

The RNA-seq data set was taken from our previous study¹⁰. It had been preprocessed as follows. For all samples, low-quality bases were removed with Fastq_quality_filter from the FASTX (http://hannonlab.cshl.edu/fastx_toolkit/index.html) Toolkit v0.0.13, with 90% of the reads needing a quality phred score of >20. Homertools³¹ v4.7 was used for poly(A) tail trimming, and reads with a length of <17 were removed. PicardTools v1.78 (https://broadinstitute.github.io/picard/) was used to compute the quality metrics with CollectRNASeqMetrics. With STAR⁵⁸ v2.3, the filtered reads were mapped against human genome 38 using default parameters. Count data and RPKM tables were generated by mapping filtered reads against union transcripts using a custom pipeline. Mapping was performed with bowtie2 (ref. ⁵⁹) v2.2.4 against union human genes: every gene is represented by a union of all its transcripts (exons). The count values (RPKM and raw counts) were calculated by running CoverageBed from Bedtools v2.26.0 of the mapped reads together with the human annotation file (Ensembl 90) in gtf format and parsing the output with custom Perl scripts.

UMAP visualization of scATAC-seq data

We subsetted scATAC-seq data to the relevant cell types and renormalized the data using the term frequency–inverse document frequency (TF-IDF) approach (RunTFIDF function from Signac⁶⁰; v1.2.1). We then identified the most relevant peaks using Signac’s FindTopFeatures function and performed dimensionality reduction with Signac’s RunSVD function. We subsequently corrected for donor-specific effects using the RunHarmony function (group.by.vars = ‘donor’, project.dim = FALSE) from the harmony⁶¹ package (v0.1.0). Finally, we computed a uniform manifold approximation and projection (UMAP) representation using Seurat’s⁶² (v4.0.1) RunUMAP function with the argument ‘dims = 2:50’. This excluded the first harmony-corrected component, which we had previously found to be strongly correlated with sequencing depth.

Principal component analysis

We divided the hs37d5 reference genome into 5,000-bp bins and computed sample-level mean raw methylation across all CpGs in a bin using the getMeth function (regions = bins, type = ‘raw’, what = ‘perRegion’) from bsseq. We then selected bins in which all samples displayed a mean CpG-level coverage of at least 2 and exhibited a nonmissing methylation value. Among these, we identified the 2,000 bins displaying the highest variance across samples and ran a principal component analysis using the prcomp function (scale. = TRUE).

Analysis of global methylation

To visualize methylation across the whole genome, we divided the hs37d5 reference genome into bins of 1,000,000 bases. For each bin and each sample, we then computed the mean raw methylation across all CpGs in that bin using bsseq’s getMeth function (regions = bins, type = ‘raw’, what = ‘perRegion’). We averaged these bin-level values across all samples of a cell type (rowMeans function with ‘na.rm = TRUE’) and computed the deviation of this value from the average value of blood CD45RA⁺ T_conv cells. We removed bins on the mitochondrial chromosome. To plot genome-wide methylation together with chromosome band information, we used the circos.initializeWithIdeogram (species = ‘hg19’, plotType = c(‘ideogram’, ‘labels’), chromosome.index = relevant_chromosomes) function from circlize⁶³ (v0.4.13). We then plotted the methylation values using circlize’s circos.track and circos.yaxis functions. To calculate global methylation values for autosomes and chromosome X, we extracted raw sample-level methylation values for every CpG using bsseq’s getMeth function (type = ‘raw’, what = ‘perBase’). Of these, we selected CpGs displaying a coverage of at least 2 in all samples. We then computed sample-level mean methylation values across CpGs in the respective chromosome categories (all autosomes or chromosome X) and averaged the resulting values across all samples of a cell type.

Extraction of methylation-level cell-type signatures

To extract cell-type signatures at the methylation level, we first generated CpG-level annotations. To this end, we considered all CpGs that displayed a coverage of at least 2 in at least two samples of each cell type. For these CpGs, we extracted sample-level smoothed methylation values using the getMeth function (type = ‘smooth’, what = ‘perBase’) from bsseq. We averaged these values across all samples from a cell type (mean function with ‘na.rm = TRUE’) and ranked the cell types according to these averages. Afterward, we identified the largest and second-to-largest methylation gap between any two adjacent cell types in this ranking. Furthermore, we computed the maximum standard error of the mean (excluding missing values) across the methylation values for all cell types. We annotated a CpG to a signature category if the largest methylation gap was at least 0.15, the ratio between the largest and second-to-largest gap was at least 1.5, and the maximum standard error was not larger than 0.5 times the largest gap. We left all CpGs that did not meet these requirements unannotated. To generate names for signature categories, we divided cell types into two groups based on the location of the largest gap. We named categories according to what cell types were in the group displaying lower methylation (for example, ‘Skin T_reg/Blood CCR8⁺ T_reg’). Based on these CpG-level annotations, we identified signature regions. To this end, we identified groups of consecutive CpGs with the same annotation that were not more than 300 bp apart. We termed these groups of CpGs ‘signature regions’. Afterward, we removed signature regions that contained fewer than three CpGs or displayed a mean coverage (across all CpGs in this region) below 2 in at least one sample. To remove signature regions that were identified due to smoothing artifacts, we performed an additional filtering step based on raw methylation. We fitted a logistic regression model (glm function with the family argument set to ‘binomial’) to sample-level mean raw methylation values (across CpGs; bsseq’s getMeth function with regions = ‘region_intervals’, what = ‘perRegion’, type = ‘raw’), using methylation as an independent variable and group membership (lower or higher methylation) as a dependent variable. We performed leave-one-out cross-validation by ignoring one sample at a time, fitting the logistic model using all remaining samples and using the fitted model to predict group membership (higher or lower methylation) of the left-out sample. If more than one sample was misclassified during this process, we discarded the respective signature region.

Intervals defined with respect to gene bodies and CpG islands

We used locations of gene bodies and exons in the hs37d5 reference genome that were computed based on the Gencode⁶⁴ annotation, version 19. We inferred transcription start sites by the start and end position of gene body intervals together with genomic strand information. For genes on the plus strand, we identified the start of the gene body as the transcription start site. For genes on the minus strand, we identified the end of the gene body as the transcription start site. We ordered exons and introns according to their distance to the transcription start site. We downloaded locations of CpG islands (‘cpgIslandExt’ data set) from https://hgdownload.cse.ucsc.edu/goldenPath/hg19/database/ on 7 March 2022. We defined shores as the 2,000 bp flanking a CpG island on both sides and shelves as the 2,000 bp flanking these shores.

Differential methylation analysis

To identify regions of differential methylation between two cell types, we adapted the strategy for signature extraction (see above) to a two-class comparison. We did not consider the ratio between the largest and second-to-largest methylation gap because the second-to-largest gap does not exist in a two-class comparison. Furthermore, we discarded regions for which at least one sample was misclassified during leave-one-out cross-validation with logistic regression.

Differential peak accessibility analysis in scATAC-seq data

We subsetted scATAC-seq data to the two compared populations and renormalized the data using Signac’s RunTFIDF function. We then analyzed differential accessibility using the FindMarkers (test.use = ‘LR’, latent.vars = ‘peak_region_fragments’) function from Seurat. We considered a peak statistically significant if its Bonferroni-adjusted P value was below 0.05. Sample sizes were 5,687 blood CD45RA⁺ T_reg cells, 6,531 blood CD45RA⁺ T_conv cells and 4,091 skin T_reg cells. For heat map visualization (see Figs. 2a and 3a), we averaged normalized accessibility values across all cells of the respective cell type. We did so separately for each donor who had donated cells from the respective compartment (blood or skin).

Differential expression analysis of RNA-seq data

We considered all genes that had a unique gene symbol and performed differential gene expression analysis based on raw expression counts. For the actual analysis, we used the functions DESeqDataSetFromMatrix (‘design = ~ Donor + Cell_type’ for the comparison between blood CD45RA⁺ T_reg cells and blood CD45RA⁺ T_conv cells and ‘design = ~ Cell_type’ for all other comparisons because of their unpaired sample structure) from the DESeq2 (ref. ⁶⁵) package (v1.30.1) and relevel from the stats package, followed by the functions DESeq and results (α = 0.05) from the DESeq2 package. The inclusion of donor ID as a model covariate was due to the partly paired sample structure in the comparison between blood CD45RA⁺ T_reg cells and blood CD45RA⁺ T_conv cells. For differential TE expression analysis, we deviated from this workflow as described in the respective Methods section. We considered a feature statistically significant if its Benjamini–Hochberg-adjusted P value (‘padj’ column in the results table returned by DESeq2) was below 0.05 and its absolute log₂ (fold change) was above 0.5. Sample sizes were five samples (from five donors) for blood CD45RA⁺ T_reg cells, three samples (from three donors) for blood CD45RA⁺ T_conv cells, four samples (from four donors) for skin T_reg cells and five samples (from five donors) for blood CCR8⁺ T_reg cells.

Computation of TPM values in RNA-seq data

We converted previously computed RPKM values to TPM values by multiplying each RPKM value by 1,000,000 and then dividing by the sum of RPKM values across all genes in the respective sample.

Definition and analysis of DMR–peak–gene links

We defined a DMR–peak–gene link as a combination of a DMR, an overlapping differentially accessible peak and a differentially expressed gene that both the DMR and the peak were associated with. We considered a genomic region (DMR or peak) associated with a gene if it overlapped with the gene body or the 2,000-bp interval upstream of the transcription start site. To identify these associations, we used the gene annotation described in ‘Intervals defined with respect to gene bodies and CpG islands’. To compute differences in methylation, we extracted sample-level mean raw methylation values across all CpGs in a DMR using bsseq’s getMeth function (regions = DMRs, type = ‘raw’, what = ‘perRegion’). We aggregated these to the cell-type level by averaging across all samples of a cell type (rowMeans function with ‘na.rm = TRUE’) and computed the difference between the two cell types. For heat map visualization of accessibility values in DMR–peak–gene links, we averaged normalized accessibility values across all cells of the respective cell type. We did so separately for each donor who had donated cells from the respective compartment (blood or skin).

Generation of track plots

To visualize methylation, chromatin accessibility and other position-wise data in selected genomic regions, we used track plots. We generated methylation tracks by extracting smoothed methylation values for each CpG and sample using bsseq’s getMeth function (regions = regions_of_interest, type = ‘smooth’, what = ‘perBase’). The track plot consists of lines connecting CpG-wise values for each sample. To visualize chromatin accessibility in the same regions, we used BAM files of scATAC-seq data. If necessary, we merged BAM files corresponding to subsets of the same cell type using sambamba merge. We then generated BigWig files containing scATAC-seq signal aggregated into 50-bp genomic bins. For this, we used RPKM normalization using bamCoverage (-of bigwig –-normalizeUsing RPKM) from the deeptools⁶⁶ suite (v3.5.1). We used the import function from the rtracklayer⁶⁷ package (v1.50.0) to load BigWig files into memory. If the plot contained ChIP–seq data, we read the corresponding BigWig files into memory using the same function. To visualize genes in the corresponding regions, we used locations of gene bodies and exons in the hs37d5 reference genome that were computed based on the Gencode annotation, version 19. We inferred transcription start sites as described in ‘Intervals defined with respect to gene bodies and CpG islands’. We aligned all data panels using the plot_grid function (ncol = 1, axis = ‘lr’, align = ‘v’) from cowplot (v1.1.1). To visualize raw methylation values alongside smoothed methylation values, we extracted raw methylation for each CpG site covered in a track plot using bsseq’s getMeth function (regions = region_of_interest, what = ‘perBase’, type = ‘raw’). Each row in the resulting heat maps corresponds to a sample from one donor, and each column corresponds to one CpG site covered in the track plot.

Analysis of amplicon data for the validation of methylation differences

We demultiplexed sequencing reads by looking for exact sequence matches between the starts of R1 reads and the cell-type barcodes used. We removed these barcodes from the reads and subsequently trimmed remaining sequencing adaptors using cutadapt⁶⁸ (v4.8) with the options ‘-a R1_3prime_adaptor -G R2_5prime_adaptor -A R2_3prime_adaptor’. To enable sequencing read alignment, we downloaded the GRCh37 reference genome from http://ftp.ensembl.org/pub/release-75/fasta/homo_sapiens/dna/ on 6 June 2024 and prepared it for usage with the bismark⁶⁹ aligner (v0.20.0) using the bismark_genome_preparation command built into the alignment software. We then performed alignment using the bismark command with the options ‘–-unmapped –-ambiguous –-genome genome_directory’ together with bowtie (v2.3.5.1) and samtools (v1.11.0). To increase alignment yield, we realigned previously unmapped R1 and R2 reads independently using the same alignment command. We merged BAM files from the different alignment steps using the samtools merge command and performed methylation calling on this merged BAM file using the bismark_methylation_extractor command with the options ‘–-comprehensive –-cytosine_report –-genome_folder genome_directory’. To analyze methylation in the targeted genomic regions, we converted the resulting CpG-level methylation file into a structure resembling a Bed file and then extracted all CpGs overlapping with targeted regions using the bedtools (v2.24.0) intersect command with the options ‘-a CpG_methylation_file -b Amplicon_interval_file -u’. We combined data on the two opposite cytosines of the same CpG by summing up their read counts. Coverage was computed as for whole-genome bisulfite sequencing data (see ‘Preprocessing of whole-genome bisulfite sequencing data’). To assess the statistical significance of CpG-wise methylation differences, we used two-tailed Wilcoxon signed-rank tests. For visualization of methylation levels together with their statistical significance, we used the plot_grid (ncol = 1, axis = ‘lr’, align = ‘v’) function from cowplot.

Enrichment analysis of Hallmark gene sets

We downloaded the collection of Hallmark gene sets using the msigdbr function (species = ‘Homo sapiens’, category = ‘H’) from the msigdbr package (v7.4.1) on 16 May 2024. We then identified DMR–peak–gene links displaying hypomethylation, hyperaccessibility and hyperexpression in skin T_reg cells compared to blood CD45RA⁺ T_reg cells. From these links, we extracted all genes and investigated enrichment of Hallmark terms among them using one-tailed Fisher’s exact tests. Our background collection of genes consisted of all genes present in the Gencode annotation, version 19.

Transcription factor analyses

To analyze enrichments of transcription factor motifs in selected genomic regions, we saved the respective region sets in Bed format and used the findMotifsGenome.pl functionality from homer³¹ (v4.11) with the options ‘hg19 -size given’. When we compared two region sets directly instead of characterizing single region sets independently, we specified a custom set of background regions via the -bg option. This was the case in the analysis of blood CCR8⁺ T_reg cell positioning. Within a data type (methylation or chromatin accessibility), we used the ‘closer to skin T_reg’ regions as background for the ‘closer to blood CD45RA⁺ T_reg’ regions and vice versa. To determine whether a motif was enriched or depleted in a region set, we divided the percentage of query regions containing the motif (as computed by homer) by the percentage of background regions containing the motif (as computed by homer). We considered a motif enriched if this quotient was above 1 and depleted otherwise.

To visualize methylation and chromatin accessibility around genomic sites for the relevant motifs, we updated the Seurat object containing scATAC-seq data. To this end, we applied the function UpdateSeuratObject from Seurat, used Signac’s as.ChromatinAssay function to update the assay object with scATAC-seq data, set the genome of the chromatin assay to ‘hg19’ (Signac’s genome function), applied Signac’s CreateFragmentObject (path = fragments_file, cells = colnames(seurat_object)) function and added the resulting fragments object to the Seurat object (Signac’s Fragments function). We then subsetted the Seurat object to the relevant cell types and renormalized data by the TF-IDF approach (Signac’s RunTFIDF function). We used the position frequency matrices (PFMs) that were returned along with homer’s results for relevant motifs. For our analysis, we converted them to a PFMatrixList object using the PFMatrix function (strand = ‘*’, profileMatrix = 1,000 * PFM) and PFMatrixList function from TFBSTools⁷⁰ (v1.28.0). We added motif annotations to the Seurat object using the AddMotifs function (genome = BSgenome.Hsapiens.UCSC.hg19, pfm = PFMs) from Signac. For the visualization of methylation around motif sites, we extracted motif positions from the positions slot of the Motifs object (Signac’s Motifs function). We extended these positions by 200 bases on each side and computed sample-level average raw methylation values across all CpGs therein (bsseq’s getMeth function with arguments regions = ‘extended_regions’, type = ‘raw’, what = ‘perRegion’). We excluded regions that returned missing values in every sample. In the final visualization, we ordered regions by their mean methylation across all visualized samples. For the visualization of chromatin accessibility around motif sites, we used the Footprint function from Signac (motif.name = names_of_relevant_motifs, genome = BSgenome.Hsapiens.UCSC.hg19) together with Signac’s PlotFootprint function.

For the analysis of transcriptomic footprints, we obtained a network of transcription factors and their target genes using the get_dorothea function from decoupleR⁷¹ (v2.3.2). We then analyzed the enrichment of transcription factor target genes among differentially expressed genes based on each gene’s log₂ (fold change) values using the fast gene set enrichment analysis (fgsea) method implemented in decoupleR’s run_fgsea (mat = log2_fold_changes, network = transcriptionfactor_target_network) function.

To identify interesting motif sites for c-Myc and USF1, we extracted motif sites as described above. Our previous analysis had demonstrated enrichment of two motifs for c-Myc. For this c-Myc analysis, we only considered genomic sites of the ‘c-Myc(bHLH)-Unpublished’ motif that overlapped with genomic sites of the ‘c-Myc(bHLH)-GSE11431’ motif. We identified c-Myc/USF1 motif sites that overlapped with a ‘Skin T_reg cell hypomethylation’ DMR, a nondifferential peak and a region of pronounced c-Myc/USF1 ChIP–seq signal. We downloaded ChIP–seq tracks from ENCODE³² via the Gene Expression Omnibus (GEO; accession numbers: GSM822291, GSM822310, GSM822290, GSM803527, GSM803441 and GSM803347) on 9 March 2023. If several replicates were available, we downloaded data for replicate 1.

Analysis of blood CCR8⁺ T_reg cell positionings

We focused on all features that were different between skin T_reg cells and blood CD45RA⁺ T_reg cells. For methylation levels, we extracted sample-level average raw methylation values across all CpGs in a region using bsseq’s getMeth function (regions = DMRs, type = ‘raw’, what = ‘perRegion’). We averaged these across all samples of a cell type (rowMeans function with ‘na.rm = TRUE’). For chromatin accessibility levels, we subsetted the Seurat object with scATAC-seq data to the relevant cell types and renormalized the data by TF-IDF (RunTFIDF function from Signac). We extracted cell-level normalized accessibility data using SeuratObject’s (v4.0.1) GetAssayData function (slot = ‘data’) and aggregated these into cell-type-level means. For the gene expression level, we extracted sample-level TPM values, transformed them by log (100 × TPM + 1) and averaged the resulting values across all samples of a cell type. We then excluded all features for which the extracted (and possibly transformed) values contradicted the differential tendencies found during differential analysis (zero DMRs, zero differential peaks and five differential genes). We scaled cell-type-level values linearly to an interval between 0 and 1. For each feature, we then identified whether its value in blood CCR8⁺ T_reg cells was closer to that in skin T_reg cells or closer to that in blood CD45RA⁺ T_reg cells. For heat map visualization of methylation and gene expression values (see Fig. 5a), we repeated the steps described above without computing the mean across samples so that each sample is displayed independently. For heat map visualization of chromatin accessibility values (see Fig. 5a), we averaged normalized accessibility values across all cells of the respective cell type. We did so separately for each donor who had donated cells from the respective compartment (blood or skin). We ordered features by the difference between blood CCR8⁺ T_reg cells and skin T_reg cells (considering only the methylation level in the case of DMR–peak–gene links). For the analysis of blood CCR8⁺ T_reg cell positionings stratified by blood CCR8⁺ T_reg cell donor, we extracted methylation and gene expression values for each individual blood CCR8⁺ T_reg cell donor. At the methylation level, we only considered DMRs that had nonmissing methylation values in all blood CCR8⁺ T_reg cell donors. To compute donor-level values at the chromatin accessibility level, we considered all cells annotated to the relevant cell types that originated from the respective tissue and then averaged chromatin accessibility values across all relevant cells independently for each donor. For each blood CCR8⁺ T_reg cell donor, we subsequently compared the donor-level methylation, chromatin accessibility and gene expression values to the corresponding cell-type-level values of skin T_reg cells and blood CD45RA⁺ T_reg cells, which had been computed as described above.

Analyses based on TCR clonotypes

We used combined scRNA-seq/scTCR-seq data of human blood and skin T cells from our previous study¹⁰. This data set contains data from two donors (TCR Donors 6 and 7). To identify blood CCR8⁺ T_reg cells and blood CD45RA⁺ T_reg cells, we processed samples 28 (blood T_reg cells, TCR Donor 6), 29 (blood CD45RO⁺ T_reg cells, TCR Donor 7), 24 (blood T_reg cells, TCR Donor 7) and 25 (blood CD45RO⁺ T_reg cells, TCR Donor 7). For each sample, we identified the 3,000 most variable genes using Seurat’s FindVariableFeatures function with the option ‘nfeatures = 3000’. We then scaled expression data using Seurat’s ScaleData function with the options features = ‘rownames(seurat_object)’ and vars.to.regress = ‘nFeature_RNA’. We ran principal component analysis using Seurat’s RunPCA function and generated a UMAP representation using the RunUMAP function (dims = 1:i) from Seurat. The parameter ‘i’ (that is, the number of principal components used) was 10 (samples 24 and 28), 20 (sample 29) and 30 (sample 25). For clustering by the Leiden algorithm, we used Seurat’s FindNeighbors (dims = 1:i) and FindClusters (resolution = y, algorithm = 4) functions. The parameter ‘i’ was the same as above. The parameter ‘y’ (that is, the clustering resolution) was 0.7 (samples 24 and 28), 0.8 (sample 25) or 0.9 (sample 29). To annotate cells, we used several single marker genes next to two gene signatures defining blood CCR8⁺ T_reg cells (Extended Data Fig. 8c–f). Blood CCR8⁺ T_reg cell signature 1 contains genes that are overexpressed in blood CCR8⁺ T_reg cells compared to blood CD45RA⁺ T_reg cells according to our analysis of bulk RNA-seq data (see ‘Differential expression analysis of RNA-seq data’). Blood CCR8⁺ T_reg cell signature 2 contains genes that are overexpressed in blood CCR8⁺ T_reg cells compared to blood CD45RO⁺ (memory) T_reg cells. To generate this signature, we used another scRNA-seq/scTCR-seq sample (sample 30) from TCR Donor 6, which contains sorted blood CCR8⁺ T_reg cells, and compared it to cells in sample 29 using Seurat’s FindMarkers function. We interpreted genes with Bonferroni-adjusted P values below 0.05 as differentially expressed. To generate cell-level scores for the two blood CCR8⁺ T_reg cell signatures, we used the AddModuleScore function from Seurat. For further analysis of blood CD45RA⁺ T_reg cells, we examined blood CD45RA⁺ T_reg cells identified in samples 24 and 28 (Extended Data Fig. 8c,e; note that samples 25 and 29 contain cells sorted for CD45RO expression and thus do not contain blood CD45RA⁺ T_reg cells). For further analysis of blood CCR8⁺ T_reg cells, we investigated blood CCR8⁺ T_reg cells identified in samples 24, 25, 28 and 29 (Extended Data Fig. 8c–f), which we combined with all cells from the CCR8-sorted sample 30. Skin T_reg cells from both donors had previously been identified in this data set (TCR Donor 6: intersection between original cluster 10 and sample R3; TCR Donor 7: intersection between original cluster 8 and sample R7)¹⁰. We considered two cells ‘TCR matched’ if the nucleotide sequence string of their clonotypes was identical and if this sequence string contained at least one TCRα chain sequence and TCRβ chain sequence. To visualize clonotype sharing between cell types, we examined all clonotypes that appeared among skin T_reg cells and contained at least one TCRα chain sequence and TCRβ chain sequence. For each of these clonotypes, we identified TCR-matched cells among blood CCR8⁺ T_reg cells and blood CD45RA⁺ T_reg cells as described above. We then computed the percentage of each clonotype among all cells of the respective population that had nonmissing clonotype information. To compute module scores of our skin T_reg signatures, we used the ‘Skin T_reg cell hyperexpression’ and ‘Blood CD45RA⁺ T_reg hyperexpression’ genes that are shown in Fig. 5a. We restricted single-cell gene expression data to (1) skin T_reg cells that were TCR matched with blood CCR8⁺ T_reg cells, (2) blood CCR8⁺ T_reg cells that were TCR matched with skin T_reg cells and (3) blood CD45RA⁺ T_reg cells from clonotypes that were shared between skin T_reg cells and blood CCR8⁺ T_reg cells. We then applied Seurat’s AddModuleScore function.

TE analyses

To analyze the methylation and chromatin accessibility of TEs, we downloaded RepeatMasker annotations from University of California, Santa Cruz (UCSC, https://hgdownload.soe.ucsc.edu/goldenPath/hg19/database/rmsk.txt.gz), on 14 September 2023. To restrict our analysis to repeat elements that are confidently TE insertion sites, we selected elements that were classified as ‘DNA’, ‘LINE’, ‘LTR’, ‘SINE’, ‘RC’ or ‘Retroposon’ (column 12 of the resulting table). We extracted TE classes from column 12 and TE subfamily names from column 11 of the RepeatMasker table.

To investigate enrichments of TEs in sets of genomic regions, we identified the proportion of regions in the region set overlapping with TE insertion sites. We shuffled regions in the region set across the human genome 10,000 times by saving these regions in Bed format, generating a chromosome length file (lengths were retrieved using seqlengths from GenomeInfoDb (v1.26.7) together with BSgenome.Hsapiens.1000genomes.hs37d5 (v0.99.1) for DMRs and BSgenome.Hsapiens.UCSC.hg19 (v1.4.2) for peaks) and using bedtools shuffle (-noOverlapping) with 10,000 different random seeds and the previously generated chromosome length file. For each of these shufflings, we recomputed the proportion of regions from the (shuffled) region set overlapping with TE insertion sites. We computed an enrichment score as the proportion of regions from the original region set overlapping with TE insertion sites divided by the mean of this proportion across all random shufflings. We computed a one-tailed enrichment P value as the proportion of shufflings for which the overlap between regions from the shuffled region set and TE insertion sites was equal to or larger than the overlap obtained with the original, unshuffled, region set.

To investigate how far blood CCR8⁺ T_reg cells reflect TE-related epigenetic traits of skin T_reg cells, we selected DMRs of the ‘Skin T_reg cell hypomethylation’ class and peaks from the ‘Skin T_reg cell hyperaccessibility’ class. We stratified these by whether they placed blood CCR8⁺ T_reg cells closer to skin T_reg cells or closer to blood CD45RA⁺ T_reg cells. Afterward, we analyzed whether TEs were enriched in one of these strata with respect to the other. To this end, we identified overlaps between TE insertion sites and regions inside the two strata. We then assessed differences in overlap proportions between the two strata using Fisher’s exact test.

We visualized methylation in TE insertion sites by extracting insertion site-level raw methylation values (bsseq’s getMeth function with arguments regions = ‘insertion_sites’, type = ‘raw’, what = ‘perRegion’) and excluding insertion sites for which at least one sample displayed a missing value. We visualized accessibility in peaks overlapping with TE insertion sites by subsetting the scATAC-seq Seurat object to cells from relevant cell types originating from donors that had donated the respective tissue (blood or skin), renormalizing data by TF-IDF (Signac’s RunTFIDF function), extracting normalized accessibility data with SeuratObject’s GetAssayData function (slot = ‘data’) and averaging these across all cells, independently for each donor.

To analyze expression of TEs, we downloaded consensus sequences of TEs after browsing the Giri Repbase⁷² (https://www.girinst.org/repbase/) with the criteria taxon ‘Homo sapiens (Human)’, repeat class ‘Transposable Element’, include elements ‘Autonomous’ and ‘Nonautonomous’ and ‘Simple’ (that is, all) and download ‘Homo sapiens and ancestral (shared) repeats’ on 3 November 2023. We also extracted sequences of full gene bodies in the human genome. To this end, we used the gene body annotations described in ‘Generation of track plots’ and used BSgenome’s (v1.58.0) getSeq function (x = BSgenome.Hsapiens.1000genomes.hs37d5 and as.character = TRUE). Additionally, we downloaded sequences of human lncRNAs from Gencode version 19 (https://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_19/gencode.v19.lncRNA_transcripts.fa.gz) on 19 December 2023. We applied TEspeX⁷³ (v2.0.1) to Fastq files with the options ‘–-TE te_consensus_sequences –-cdna gene_body_sequences –-ncrna lncRNA_sequences –-paired F –-length 76 –-strand ‘no’’ together with zlib (v1.2.12) and Python (v3.8.3). To compute TPM values, we divided raw expression counts by the respective consensus sequence length, divided these quotients by the sum of all quotients within the respective sample and multiplied the resulting value by 1,000,000. For heat map visualization, we clustered TEs by hierarchical clustering using the Euclidean distance and complete linkage. We identified differentially expressed TEs using DEseq2 as described in ‘Differential expression analysis of RNA-seq data’ with the following deviation: we did not run the wrapper function DEseq but executed the steps contained therein independently. The reason for this was that the DESeq function automatically estimates size factors based on the sums of read counts in the expression matrix. However, we reasoned that TE-derived reads account only for a minority of sequencing reads so that the complete read numbers (including non-TE reads) yield a more accurate estimate of the library size. We used DESeq2’s sizeFactors function to manually set size factors to total read counts as returned by TEspeX, multiplied by 1 × 10⁻⁷. Afterward, we used the functions estimateDispersions and nbinomWaldTest from DESeq2.

We computed methylation of HERVIP10F-int and LRT45B insertion sites overlapping with ‘Skin T_reg cell hypomethylation’ DMRs by using bsseq’s getMeth function (regions = overlapping_insertion_sites, type = ‘raw’, what = ‘perRegion’).

Statistics

Unless indicated in the respective figure caption or Methods, the statistical tests performed were two-tailed. We interpreted P values below 0.05 as statistically significant. If possible, test statistics, degrees of freedom, confidence intervals and so on are included in our Supplementary Tables or Source Data. In case of parametric tests, data distribution was assumed to be normal, but this was not formally tested.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Human whole-genome bisulfite sequencing data have been deposited in the European Genome Phenome Archive (EGA; accession number EGAD50000001022) and will be available under controlled access. The following are the file name conventions: labels 1–3: skin T_reg cell samples; labels 4–6: skin T_conv cell samples; labels 13–15: blood CCR8⁺ T_reg cell samples; labels 16–18: blood CD45RA⁺ T_reg cell samples; labels 19–21: blood CD45RA⁺ T_conv cell samples. Access will require application to a data access committee and will require a data sharing agreement. CpG-level methylation data have been deposited in GEO (accession number: GSE286948) and are available without restrictions.

We used data sets that were already used in our previous study¹⁰. Corresponding raw sequencing data have been deposited in the EGA (accession numbers: EGAD00001006779 for scATAC-seq; EGAD00001007665 and EGAD00001007661 for scRNA-seq/scTCR-seq; EGAD00001007663 and EGAD00001007665 for bulk RNA-seq; access requires application to a data access committee and a data sharing agreement). The versions of the corresponding processed data that were used in this study are available on Zenodo (https://doi.org/10.5281/zenodo.14999322 (ref. ⁷⁴) for scATAC-seq; https://doi.org/10.5281/zenodo.14999788 (ref. ⁷⁵) for scRNA-seq/scTCR-seq; https://doi.org/10.5281/zenodo.14999989 (ref. ⁷⁶) for bulk RNA-seq).

In addition, we used several data sets from other sources. The hs37d5 reference genome is publicly available at https://ftp.ncbi.nlm.nih.gov/1000genomes/ftp/technical/reference/phase2_reference_assembly_sequence. The GRCh37 reference genome is available at http://ftp.ensembl.org/pub/release-75/fasta/homo_sapiens/dna. The human genome 38 is available at https://hgdownload.soe.ucsc.edu/goldenPath/hg38/chromosomes. Gencode V19 annotations are available at https://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_19. We downloaded locations of CpG islands (the ‘cpgIslandExt’ data set) from https://hgdownload.cse.ucsc.edu/goldenPath/hg19/database. Hallmark gene sets are available via the msigdbr function (species = ‘Homo sapiens’, category = ‘H’) from the msigdbr package. ChIP–seq data for c-Myc and USF1 are publicly available at GEO (accession numbers: GSM822291, GSM822310, GSM822290, GSM803527, GSM803441 and GSM803347). RepeatMasker annotations are available at https://hgdownload.soe.ucsc.edu/goldenPath/hg19/database/rmsk.txt.gz. We downloaded consensus sequences of TEs from the Giri Repbase⁷² (https://www.girinst.org/repbase/). Sequences of human lncRNAs are available at https://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_19/gencode.v19.lncRNA_transcripts.fa.gz.

Source data for this study are available at Zenodo at https://doi.org/10.5281/zenodo.15607737 (ref. ⁷⁷).

Code availability

We did not develop any new software but used functionality that is publicly available via software packages as described in the Methods. Our wrappercode used for running this publicly available software is published on GitHub, can be accessed without restrictions at https://github.com/DKFZ-ABI/hum_tistreg_methylation and is reusable under a CC BY-NC-SA 4.0 license.

References

Ohkura, N. et al. T cell receptor stimulation-induced epigenetic changes and Foxp3 expression are independent and complementary events required for T_reg cell development. Immunity 37, 785–799 (2012).
Article CAS Google Scholar
Dikiy, S. & Rudensky, A. Y. Principles of regulatory T cell function. Immunity 56, 240–255 (2023).
Article CAS Google Scholar
Delacher, M. et al. Genome-wide DNA-methylation landscape defines specialization of regulatory T cells in tissues. Nat. Immunol. 18, 1160–1172 (2017).
Article CAS Google Scholar
Delacher, M. et al. Precursors for nonlymphoid-tissue T_reg cells reside in secondary lymphoid organs and are programmed by the transcription factor BATF. Immunity 52, 295–312 (2020).
Munoz-Rojas, A. R. & Mathis, D. Tissue regulatory T cells: regulatory chameleons. Nat. Rev. Immunol. 21, 597–611 (2021).
Article CAS Google Scholar
Miragaia, R. J. et al. Single-cell transcriptomics of regulatory T cells reveals trajectories of tissue adaptation. Immunity 50, 493–504 (2019).
Article CAS Google Scholar
DiSpirito, J. R. et al. Molecular diversification of regulatory T cells in nonlymphoid tissues. Sci. Immunol. 3, eaat5861 (2018).
Feuerer, M. et al. Lean, but not obese, fat is enriched for a unique population of regulatory T cells that affect metabolic parameters. Nat. Med. 15, 930–939 (2009).
Article CAS Google Scholar
Cipolletta, D. et al. PPAR-γ is a major driver of the accumulation and phenotype of adipose tissue T_reg cells. Nature 486, 549–553 (2012).
Article CAS Google Scholar
Delacher, M. et al. Single-cell chromatin accessibility landscape identifies tissue repair program in human regulatory T cells. Immunity 54, 702–720 (2021).
Article CAS Google Scholar
Ali, N. et al. Regulatory T cells in skin facilitate epithelial stem cell differentiation. Cell 169, 1119–1129 (2017).
Article CAS Google Scholar
Burzyn, D. et al. A special population of regulatory T cells potentiates muscle repair. Cell 155, 1282–1295 (2013).
Article CAS Google Scholar
Arpaia, N. et al. A distinct function of regulatory T cells in tissue protection. Cell 162, 1078–1089 (2015).
Article CAS Google Scholar
Ito, M. et al. Brain regulatory T cells suppress astrogliosis and potentiate neurological recovery. Nature 565, 246–250 (2019).
Article CAS Google Scholar
Sakai, R. et al. Kidney GATA3⁺ regulatory T cells play roles in the convalescence stage after antibody-mediated renal injury. Cell Mol. Immunol. 18, 1249–1261 (2021).
Article CAS Google Scholar
Li, C. et al. TCR transgenic mice reveal stepwise, multi-site acquisition of the distinctive fat-T_reg phenotype. Cell 174, 285–299 (2018).
Article CAS Google Scholar
Vasanthakumar, A. et al. The transcriptional regulators IRF4, BATF and IL-33 orchestrate development and maintenance of adipose tissue-resident regulatory T cells. Nat. Immunol. 16, 276–285 (2015).
Article CAS Google Scholar
Hayatsu, N. et al. Analyses of a mutant Foxp3 allele reveal BATF as a critical transcription factor in the differentiation and accumulation of tissue regulatory T cells. Immunity 47, 268–283 (2017).
Article CAS Google Scholar
Azizi, E. et al. Single-cell map of diverse immune phenotypes in the breast tumor microenvironment. Cell 174, 1293–1308 (2018).
Article CAS Google Scholar
Plitas, G. et al. Regulatory T cells exhibit distinct features in human breast cancer. Immunity 45, 1122–1134 (2016).
Article CAS Google Scholar
Wang, L. et al. Connecting blood and intratumoral T_reg cell activity in predicting future relapse in breast cancer. Nat. Immunol. 20, 1220–1230 (2019).
Article CAS Google Scholar
Cordaux, R. & Batzer, M. A. The impact of retrotransposons on human genome evolution. Nat. Rev. Genet. 10, 691–703 (2009).
Article CAS Google Scholar
Giassa, I.-C. & Alexiou, P. Bioinformatics and machine learning approaches to understand the regulation of mobile genetic elements. Biology 10, 896 (2021).
Article CAS Google Scholar
Mills, R. E., Bennett, E. A., Iskow, R. C. & Devine, S. E. Which transposable elements are active in the human genome? Trends Genet. 23, 183–191 (2007).
Article CAS Google Scholar
Lander, E. S. et al. Initial sequencing and analysis of the human genome. Nature 409, 860–921 (2001).
Article CAS Google Scholar
Bruno, M., Mahgoub, M. & Macfarlan, T. S. The arms race between KRAB-zinc finger proteins and endogenous retroelements and its impact on mammals. Annu. Rev. Genet. 53, 393–416 (2019).
Article CAS Google Scholar
Ali, A., Han, K. & Liang, P. Role of transposable elements in gene regulation in the human genome. Life 11, 118 (2021).
Fueyo, R., Judd, J., Feschotte, C. & Wysocka, J. Roles of transposable elements in the regulation of mammalian transcription. Nat. Rev. Mol. Cell Biol. 23, 481–497 (2022).
Article CAS Google Scholar
Gebrie, A. Transposable elements as essential elements in the control of gene expression. Mob. DNA 14, 9 (2023).
Article Google Scholar
Liberzon, A. et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst. 1, 417–425 (2015).
Article CAS Google Scholar
Heinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589 (2010).
Article CAS Google Scholar
Encode Project Consortium. An integrated encyclopedia of DNA elements in the human genome. Nature 489, 57–74 (2012).
Article Google Scholar
Tomura, M. et al. Monitoring cellular movement in vivo with photoconvertible fluorescence protein ‘Kaede’ transgenic mice. Proc. Natl Acad. Sci. USA 105, 10871–10876 (2008).
Article CAS Google Scholar
Tomura, M. et al. Activated regulatory T cells are the major T cell type emigrating from the skin during a cutaneous immune response in mice. J. Clin. Invest. 120, 883–893 (2010).
Article CAS Google Scholar
Zhou, W., Liang, G., Molloy, P. L. & Jones, P. A. DNA methylation enables transposable element-driven genome expansion. Proc. Natl Acad. Sci. USA 117, 19359–19366 (2020).
Article CAS Google Scholar
Wells, J. N. & Feschotte, C. A field guide to eukaryotic transposable elements. Annu. Rev. Genet. 54, 539–561 (2020).
Article CAS Google Scholar
Kojima, S. et al. Mobile element variation contributes to population-specific genome diversification, gene regulation and disease risk. Nat. Genet. 55, 939–951 (2023).
Article CAS Google Scholar
Simon, M. et al. Single-cell chromatin accessibility and transposable element landscapes reveal shared features of tissue-residing immune cells. Immunity 57, 1975–1993 (2024).
Article CAS Google Scholar
Schmidleithner, L., Stuve, P. & Feuerer, M. Transposable elements as instructors of the immune system. Nat. Rev. Immunol. https://doi.org/10.1038/s41577-025-01172-3 (2025).
Schuyler, R. P. et al. Distinct trends of DNA methylation patterning in the innate and adaptive immune systems. Cell Rep. 17, 2101–2111 (2016).
Article CAS Google Scholar
Zhou, W. et al. DNA methylation loss in late-replicating domains is linked to mitotic cell division. Nat. Genet. 50, 591–602 (2018).
Article CAS Google Scholar
Endicott, J. L., Nolte, P. A., Shen, H. & Laird, P. W. Cell division drives DNA methylation loss in late-replicating domains in primary human cells. Nat. Commun. 13, 6659 (2022).
Article CAS Google Scholar
Bourque, G. et al. Ten things you should know about transposable elements. Genome Biol. 19, 199 (2018).
Article CAS Google Scholar
De Cecco, M. et al. L1 drives IFN in senescent cells and promotes age-associated inflammation. Nature 566, 73–78 (2019).
Article Google Scholar
Simon, M. et al. LINE1 derepression in aged wild-type and SIRT6-deficient mice drives inflammation. Cell Metab. 29, 871–885 (2019).
Article CAS Google Scholar
Chuong, E. B., Elde, N. C. & Feschotte, C. Regulatory evolution of innate immunity through co-option of endogenous retroviruses. Science 351, 1083–1087 (2016).
Article CAS Google Scholar
Ye, M. et al. Specific subfamilies of transposable elements contribute to different domains of T lymphocyte enhancers. Proc. Natl Acad. Sci. USA 117, 7905–7916 (2020).
Article CAS Google Scholar
Barnett, K. R. et al. ATAC-Me captures prolonged DNA methylation of dynamic chromatin accessibility loci during cell fate transitions. Mol. Cell 77, 1350–1364 (2020).
Article CAS Google Scholar
Burton, O. T. et al. The tissue-resident regulatory T cell pool is shaped by transient multi-tissue migration and a conserved residency program. Immunity 57, 1586–1602 (2024).
Article CAS Google Scholar
Wu, M. J. et al. Mutant IDH1 inhibition induces dsDNA sensing to activate tumor immunity. Science 385, eadl6173 (2024).
Article CAS Google Scholar
Weichenhan, D. et al. Tagmentation-based library preparation for low DNA input whole genome bisulfite sequencing. Methods Mol. Biol. 1708, 105–122 (2018).
Article CAS Google Scholar
Ando, R., Hama, H., Yamamoto-Hino, M., Mizuno, H. & Miyawaki, A. An optical marker based on the UV-induced green-to-red photoconversion of a fluorescent protein. Proc. Natl Acad. Sci. USA 99, 12651–12656 (2002).
Article CAS Google Scholar
Wang, Q. et al. Tagmentation-based whole-genome bisulfite sequencing. Nat. Protoc. 8, 2022–2032 (2013).
Article CAS Google Scholar
Hovestadt, V. et al. Decoding the regulatory landscape of medulloblastoma using DNA methylation sequencing. Nature 510, 537–541 (2014).
Article CAS Google Scholar
Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120 (2014).
Article CAS Google Scholar
Li, H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. Preprint at https://arxiv.org/abs/1303.3997 (2013).
Hansen, K. D., Langmead, B. & Irizarry, R. A. BSmooth: from whole genome bisulfite sequencing reads to differentially methylated regions. Genome Biol. 13, R83 (2012).
Article Google Scholar
Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013).
Article CAS Google Scholar
Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 10, R25 (2009).
Article Google Scholar
Stuart, T., Srivastava, A., Madad, S., Lareau, C. A. & Satija, R. Single-cell chromatin state analysis with Signac. Nat. Methods 18, 1333–1341 (2021).
Article CAS Google Scholar
Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods 16, 1289–1296 (2019).
Article CAS Google Scholar
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587 (2021).
Article CAS Google Scholar
Gu, Z., Gu, L., Eils, R., Schlesner, M. & Brors, B. Circlize implements and enhances circular visualization in R. Bioinformatics 30, 2811–2812 (2014).
Article CAS Google Scholar
Frankish, A. et al. GENCODE reference annotation for the human and mouse genomes. Nucleic Acids Res. 47, D766–D773 (2019).
Article CAS Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Article Google Scholar
Ramirez, F. et al. deepTools2: a next generation web server for deep-sequencing data analysis. Nucleic Acids Res. 44, W160–W165 (2016).
Article CAS Google Scholar
Lawrence, M., Gentleman, R. & Carey, V. rtracklayer: an R package for interfacing with genome browsers. Bioinformatics 25, 1841–1842 (2009).
Article CAS Google Scholar
Martin, M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet J. 17, 10–12 (2011).
Article Google Scholar
Krueger, F. & Andrews, S. R. Bismark: a flexible aligner and methylation caller for bisulfite-seq applications. Bioinformatics 27, 1571–1572 (2011).
Article CAS Google Scholar
Tan, G. & Lenhard, B. TFBSTools: an R/bioconductor package for transcription factor binding site analysis. Bioinformatics 32, 1555–1556 (2016).
Article CAS Google Scholar
Badia, I. M. P. et al. decoupleR: ensemble of computational methods to infer biological activities from omics data. Bioinform. Adv. 2, vbac016 (2022).
Article Google Scholar
Bao, W., Kojima, K. K. & Kohany, O. Repbase update, a database of repetitive elements in eukaryotic genomes. Mob. DNA 6, 11 (2015).
Article Google Scholar
Ansaloni, F., Gualandi, N., Esposito, M., Gustincich, S. & Sanges, R. TEspeX: consensus-specific quantification of transposable element expression preventing biases from exonized fragments. Bioinformatics 38, 4430–4433 (2022).
Article CAS Google Scholar
Beumer, N. Processed single-cell ATAC-seq data of CD4⁺ T cell subsets from skin and blood of healthy human donors. Zenodo https://doi.org/10.5281/zenodo.14999322 (2025).
Beumer, N. Processed single-cell RNA/TCR-seq data of regulatory T cells from blood and skin of healthy human donors. Zenodo https://doi.org/10.5281/zenodo.14999788 (2025).
Beumer, N. Processed RNA-seq data of CD4⁺ T cell subsets from blood and skin of healthy human donors. Zenodo https://doi.org/10.5281/zenodo.14999989 (2025).
Beumer, N. Source data for the publication ‘DNA hypomethylation traits define human regulatory T cells in cutaneous tissue and identify their blood recirculating counterparts’. Zenodo https://doi.org/10.5281/zenodo.15607737 (2025).

Download references

Acknowledgements

We thank the Leibniz Institute for Immunotherapy (LIT) and Research Center for Immunotherapy Mainz (Forschungszentrum für Immuntherapie, FZI) Next-Generation Sequencing core facility, the LIT and FZI flow cytometry core facility and the University Regensburg animal facility for technical support. We thank V. Hofmann, H. Stanewsky, M. Wuttke, J. Raitel, U. Ackermann, R. Eder and I. Fink (all LIT) for technical support. We thank B. Hutter (DKFZ) for help with Gencode annotations. We thank R. Linker (Clinic and Polyclinic for Neurology at the University of Regensburg) for Tg(CAG-Kaede)15Kgwa mice. This work was supported by grants from the European Research Council (ERC-CoG, 648145 REGiREG; to M.F.) and the Deutsche Forschungsgemeinschaft (German Research Foundation) Projektnummer 324392634–TRR221 (to M.F.), Projektnummer 318346496–SFB1292/2 TP19N (to M.D. and F.M.), Projektnummer 490846870–TRR355/1 TPA01 (to J.H. and M.D.) and TPZ02 (to M.D.). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Author information

Malte Simon
Present address: Evotec SE, Hamburg, Germany
These authors contributed equally: Niklas Beumer, Charles D. Imbusch, Tamara Kaufmann.
These authors jointly supervised this work: Benedikt Brors, Michael Delacher, Markus Feuerer.

Authors and Affiliations

Institute of Immunology, University Medical Center Mainz, Mainz, Germany
Niklas Beumer, Charles D. Imbusch, Tamara Kaufmann, Kathrin Luise Braband, Morten Michael Voss, Sara Salome Helbich, Delia Mihaela Mihoc & Michael Delacher
Research Center for Immunotherapy, University Medical Center Mainz, Mainz, Germany
Niklas Beumer, Charles D. Imbusch, Tamara Kaufmann, Federico Marini, Kathrin Luise Braband, Morten Michael Voss, Sara Salome Helbich, Delia Mihaela Mihoc & Michael Delacher
Faculty of Biosciences, Heidelberg University, Heidelberg, Germany
Niklas Beumer & Malte Simon
Division of Applied Bioinformatics, German Cancer Research Center (DKFZ), Heidelberg, Germany
Charles D. Imbusch, Malte Simon & Benedikt Brors
Leibniz Institute for Immunotherapy, Regensburg, Germany
Lisa Schmidleithner, Kathrin Gütter, Philipp Stüve, Harriet Marchel, Brigitte Ruhland, Lieke Sanderink, Uwe Ritter, Malte Simon, Claudia Gebhard, Michael Rehli, Nicholas Strieder, Christian Schmidl, Thomas Hehlgans & Markus Feuerer
Chair for Immunology, University Regensburg, Regensburg, Germany
Lisa Schmidleithner, Kathrin Gütter, Philipp Stüve, Harriet Marchel, Brigitte Ruhland, Lieke Sanderink, Uwe Ritter, Thomas Hehlgans & Markus Feuerer
Division of Cancer Epigenomics, German Cancer Research Center (DKFZ), Heidelberg, Germany
Dieter Weichenhan, Marion Bähr & Christoph Plass
Institute of Medical Biostatistics, Epidemiology and Informatics (IMBEI), University Medical Center Mainz, Mainz, Germany
Federico Marini
Core Facility Omics IT and Data Management (ODCF), German Cancer Research Center (DKFZ), Heidelberg, Germany
Agnes Hotz-Wagenblatt
Department of Dermatology, University Medical Center of the Johannes Gutenberg University, Mainz, Germany
Hadrian Nassabi
Department of Plastic, Hand and Reconstructive Surgery, University Hospital Regensburg, Regensburg, Germany
Andreas Eigenberger & Lukas Prantl
Department of Internal Medicine III, University Hospital Regensburg, Regensburg, Germany
Michael Rehli
Institute of Computational Biology, Helmholtz Munich, Munich, Germany
Kartikeya Singh
Department of Experimental Immunology, Helmholtz Centre for Infection Research, Braunschweig, Germany
Jochen Huehn
RESIST, Cluster of Excellence 2155, Hannover Medical School, Hannover, Germany
Jochen Huehn
Immuno-Epigenetics, Berlin Institute of Health Center for Regenerative Therapies at Charité Universitätsmedizin Berlin, Berlin, Germany
Julia K. Polansky
Immuno-Epigenetics, German Rheumatism Research Centre, Berlin, Germany
Julia K. Polansky
National Center for Tumor Diseases (NCT), Heidelberg, Germany
Benedikt Brors
German Cancer Consortium (DKTK), German Cancer Research Center (DKFZ), Heidelberg, Germany
Benedikt Brors
Medical Faculty Heidelberg and Faculty of Biosciences, Heidelberg University, Heidelberg, Germany
Benedikt Brors

Authors

Niklas Beumer
View author publications
Search author on:PubMed Google Scholar
Charles D. Imbusch
View author publications
Search author on:PubMed Google Scholar
Tamara Kaufmann
View author publications
Search author on:PubMed Google Scholar
Lisa Schmidleithner
View author publications
Search author on:PubMed Google Scholar
Kathrin Gütter
View author publications
Search author on:PubMed Google Scholar
Philipp Stüve
View author publications
Search author on:PubMed Google Scholar
Harriet Marchel
View author publications
Search author on:PubMed Google Scholar
Dieter Weichenhan
View author publications
Search author on:PubMed Google Scholar
Marion Bähr
View author publications
Search author on:PubMed Google Scholar
Brigitte Ruhland
View author publications
Search author on:PubMed Google Scholar
Federico Marini
View author publications
Search author on:PubMed Google Scholar
Lieke Sanderink
View author publications
Search author on:PubMed Google Scholar
Uwe Ritter
View author publications
Search author on:PubMed Google Scholar
Malte Simon
View author publications
Search author on:PubMed Google Scholar
Kathrin Luise Braband
View author publications
Search author on:PubMed Google Scholar
Morten Michael Voss
View author publications
Search author on:PubMed Google Scholar
Sara Salome Helbich
View author publications
Search author on:PubMed Google Scholar
Delia Mihaela Mihoc
View author publications
Search author on:PubMed Google Scholar
Agnes Hotz-Wagenblatt
View author publications
Search author on:PubMed Google Scholar
Hadrian Nassabi
View author publications
Search author on:PubMed Google Scholar
Andreas Eigenberger
View author publications
Search author on:PubMed Google Scholar
Lukas Prantl
View author publications
Search author on:PubMed Google Scholar
Claudia Gebhard
View author publications
Search author on:PubMed Google Scholar
Michael Rehli
View author publications
Search author on:PubMed Google Scholar
Nicholas Strieder
View author publications
Search author on:PubMed Google Scholar
Kartikeya Singh
View author publications
Search author on:PubMed Google Scholar
Christian Schmidl
View author publications
Search author on:PubMed Google Scholar
Christoph Plass
View author publications
Search author on:PubMed Google Scholar
Jochen Huehn
View author publications
Search author on:PubMed Google Scholar
Thomas Hehlgans
View author publications
Search author on:PubMed Google Scholar
Julia K. Polansky
View author publications
Search author on:PubMed Google Scholar
Benedikt Brors
View author publications
Search author on:PubMed Google Scholar
Michael Delacher
View author publications
Search author on:PubMed Google Scholar
Markus Feuerer
View author publications
Search author on:PubMed Google Scholar

Contributions

Study design: M.F., M.D., N.B. and J.H. Analysis of bisulfite sequencing data: N.B. Analysis of scATAC-seq data: N.B. Analysis of RNA-seq data: N.B. Analysis of TEs: N.B. Analyses related to TCR sequences: N.B. and C.D.I. Development and execution of amplicon-based validation experiments: T.K. Contribution to experimental work and/or analyses: L.S., K.G., P.S., H.M., D.W., M.B., B.R., F.M., L.S., U.R., M.S., K.L.B., M.M.V., S.S.H., D.M.M., A.H.-W., H.N., A.E., L.P., C.G., M.R., N.S, K.S., J.K.P., C.S., C.P. and T.H. Original manuscript draft: N.B., M.F., M.D., T.K., C.D.I., L.S., P.S., C.S., M.R., J.H., U.R., K.S. and J.K.P. Supervision of investigation: M.F., M.D., C.D.I., B.B. and C.P. Funding acquisition: M.F., M.D., F.M. and J.H.

Corresponding authors

Correspondence to Michael Delacher or Markus Feuerer.

Ethics declarations

Competing interests

M.D. received personal fees from Odyssey Therapeutics outside the submitted work. The other authors declare no competing interests.

Peer review

Peer review information

Nature Immunology thanks Adrian Liston and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Ioana Staicu, in collaboration with the Nature Immunology team.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Quality control and methylation-level cell type signatures.

a, Frequency of CD45RA⁻CCR8⁺ T_reg cells among CD4⁺CD127⁻CD25⁺ T_reg cells in human blood, fat and skin, 4 donors. b, Distribution of methylation at the first 3,000,000 CH sites of chromosome 1, stratified by sample (3 samples from three donors per cell type) and sequencing library. c, Distribution of CpG-wise coverage values (3 samples from three donors per cell type). d, DNA Methylation in regions belonging to methylation-level cell type signatures with at least 10 regions. Rows correspond to signature regions; columns correspond to samples from individual donors (n = 3 donors per cell type). Numbers on the left indicate the number of regions in each signature category. Parts of this panel are duplicated from Fig. 1f and included for the sake of completeness. Data are representative of three or more independent experiments with three or more individual donors.

Extended Data Fig. 2 Details and examples of the ‘core naïve T_reg ’ signature.

a, Schematic of the study design, describing that blood CD45RA⁺ T_conv cells, blood CD45RA⁺ T_reg cells, blood CCR8⁺ T_reg cells, skin T_conv cells and skin T_reg cells were analyzed with respect to their DNA methylation, chromatin accessibility and gene expression. Schematic generated using BioRender.com. b, UMAP showing scATAC-seq data of the analyzed cell types (encompassing 2 donors per cell type). c, Volcano plots for differential analyses on the chromatin accessibility and gene expression level (corresponding to heat maps in Fig. 2a). Features shown in red are considered statistically significant. Numbers at the top indicate the numbers of features displaying statistical significance. P values for chromatin accessibility, two-tailed likelihood ratio test (see Methods, n = 5,687 blood CD45RA⁺ T_reg cells, 6,531 blood CD45RA⁺ T_conv cells), adjusted using Bonferroni correction; P values for gene expression, two-tailed Wald test based on the raw count matrix as implemented in DESeq2, using donor ID as additional covariate (see Methods; n = 5 T_reg cell donors, 3 T_conv cell donors), adjusted using Benjamini-Hochberg correction. d, Methylation (left), chromatin accessibility (middle) and gene expression (right) of features in DMR-peak-gene links, stratified by quadrant (see Fig. 2b). Labels on the right indicate the genes in each DMR-peak-gene link. Parts of this panel are duplicated from Fig. 2a as the same features are shown. e, Methylation for selected genomic regions. Highlighted regions mark DMRs. Vertical lines at the bottom of the methylation tracks mark CpG sites. f, Smoothed methylation and chromatin accessibility for selected DMR-peak-gene links together with corresponding raw methylation for each shown CpG site (top right heat maps; rows indicate donors; columns indicate CpG sites; gray fields indicate missing values) and expression of the associated gene (bottom right bar charts; mean ± s.d.). Highlighted regions mark DMRs (methylation tracks) and differential peaks (accessibility tracks). Vertical lines at the bottom of the methylation tracks mark CpG sites. Dashed boxes mark the locations of amplicons used for validation of methylation differences. Gene expression P values, two-tailed Wald test based on the raw count matrix as implemented in DESeq2, using donor ID as additional covariate (see Methods; n = 5 T_reg cell donors, 3 T_conv cell donors) with Benjamini-Hochberg correction. Data are representative of two or more independent experiments with two or more individual donors.

Extended Data Fig. 3 Details and examples of the ‘skin T_reg ’ signature.

a, Volcano plots for differential analyses on the chromatin accessibility and gene expression level (corresponding to heat maps in Fig. 3a). Features shown in red are considered statistically significant. Numbers at the top indicate the numbers of features displaying statistical significance. P values for chromatin accessibility, two-tailed likelihood ratio test (see Methods, n = 4,091 skin T_reg cells, 5,687 blood CD45RA⁺ T_reg cells), adjusted using Bonferroni correction; P values for gene expression, two-tailed Wald test based on the raw count matrix as implemented in DESeq2 (see Methods; n = 4 skin T_reg cell donors, 5 blood CD45RA⁺ T_reg cell donors; unpaired sample structure), adjusted using Benjamini-Hochberg correction. b, Methylation (left), chromatin accessibility (middle) and gene expression (right) of features in DMR-peak-gene links, stratified by quadrant (see Fig. 3b). Labels on the right indicate the genes in each DMR-peak-gene link. Parts of this panel are duplicated from Fig. 3a as the same features are shown. c, Smoothed methylation and chromatin accessibility for a selected DMR-peak-gene link together with corresponding raw methylation for each shown CpG site (top right heat map; rows indicate donors; columns indicate CpG sites; gray fields indicate missing values) and expression of the associated gene (bottom right bar chart; mean ± s.d.). Highlighted regions mark DMRs (methylation track) and differential peaks (accessibility track). Vertical lines at the bottom of the methylation track mark CpG sites. Dashed boxes mark the locations of amplicons used for validation of methylation differences. Gene expression P value, two-tailed Wald test (see Methods; n = 4 skin donors, 5 blood donors) with Benjamini-Hochberg correction. Expression bar charts are duplicated from Fig. 5 as they correspond to the same gene. d, Methylation for selected genomic regions. Highlighted regions mark DMRs. Vertical lines at the bottom of the methylation tracks mark CpG sites. e, DMRs (left), differentially accessible peaks (middle) and differentially expressed genes (right; row-scaled) between blood CD45RA⁺ T_reg cells, blood CD45RA⁺ T_conv cells, shown together with corresponding values in skin T_reg cells and fat T_reg cells (‘core naïve T_reg’ signature, upper panel), and between blood CD45RA⁺ T_reg cells, fat T_reg cells and skin T_reg cells (‘skin T_reg ’ signature, lower panel). Rows indicate genomic regions (left, middle) or genes (right). Columns indicate samples from individual donors, with n = 3 donors for methylation data, n = 2 skin T_reg / blood CD45RA⁺ T_reg / blood CD45RA⁺ T_conv donors and 3 fat T_reg donors for accessibility data, or n = 3 blood CD45RA⁺ T_conv donors, 4 skin T_reg donors and 5 blood CD45RA⁺ T_reg / fat T_reg donors for gene expression data. Parts of this panel are duplicated from Figs. 2a and 3a as the same features are shown. Data are representative of two or more independent experiments with two or more individual donors.

Extended Data Fig. 4 Details for transcription factor results with respect to skin adaptation of T_reg cells.

a, Logos for the bZIP motifs highlighted in Fig. 4a (obtained from homer). b, Chromatin accessibility around genomic sites for additional bZIP motifs. c, Methylation (each row corresponds to a motif site ± 200 bases, each column corresponds to one of 3 donors, donors for skin Treg cells and blood CD45RA⁺ T_reg cells are unpaired) around genomic sites for bZIP motifs (n = 29,813 motif sites, 29,546 motif sites, 28,220 motif sites, 28,693 motif sites, 26,628 motif sites, 30,426 motif sites, 27,920 motif sites, 23,868 motif sites, 28,863 motif sites and 18,907 motif sites, respectively, from top to bottom). d, Logos for the bHLH motifs highlighted in Fig. 4c (obtained from homer). e, Chromatin accessibility around genomic sites for additional bHLH motifs. f, Methylation (each row corresponds to a motif site ± 200 bases, each column corresponds to one of 3 donors, donors for skin Treg cells and blood CD45RA⁺ T_reg cells are unpaired) around genomic sites for bHLH motifs. n = 12,050 motif sites, 11,186 motif sites, 9,358 motif sites, 11,415 motif sites and 3,004 motif sites, respectively, from top to bottom). g, Methylation, chromatin accessibility and USF1 ChIP-seq signal from three cell lines around interesting motif sites for USF1. Highlighted regions correspond to selected DMRs (methylation tracks), scATAC-seq peaks (chromatin accessibility tracks) and USF1 motif sites (motif ± 100 bases; ChIP-seq and gene tracks). Vertical lines at the bottom of the methylation tracks mark CpG sites. Parts of this panel are duplicated from Fig. 4g as they display the same genomic regions but focus on a different transcription factor. Data are representative of two or more independent experiments with two or more individual donors.

Extended Data Fig. 5 Targeting of GNA11_A guide and MLPH_A guide.

a, Sanger sequencing of GNA11 locus in human CD4⁺ T cells after CRISPR-Cas9 targeting with guide GNA11_A (top), scrambled control (middle) and not-CRISPRed genomic DNA (bottom) of human T cells. Blue boxes indicate position of GNA11_A guide. Lower part, percentage of aberrant sequences (not matching to genomic control sample). Dashed blue line indicates position of GNA11_A guide. b, Sanger sequencing of MLPH locus in human CD4⁺ T cells after CRISPR-Cas9 targeting with guide MLPH_A (top), scrambled control (middle) and not-CRISPRed genomic DNA (bottom) of human T cells. Blue boxes indicate position of MLPH_A guide. Lower part, percentage of aberrant sequences (not matching to genomic control sample). Dashed blue line indicates position of MLPH_A guide.

Extended Data Fig. 6 The ‘Skin T_reg ’ signature across many cell types and the molecular positioning of blood CCR8⁺ T_reg cells.

a, Feature numbers (that is numbers of DMRs, differentially accessible peaks and differentially expressed genes) underlying blood CCR8⁺ T_reg cell positionings shown in Fig. 5a. b, Methylation, chromatin accessibility and gene expression for the ‘skin T_reg ’ signature in skin T_reg cells, skin T_conv cells, blood CCR8⁺ T_reg cells, blood CD45RA⁺ T_reg cells and blood CD45RA⁺ T_conv cells. Parts of this panel are duplicated from Fig. 3a as the same features are shown. c, Methylation (left), chromatin accessibility (middle) and gene expression (right) of features that belong to DMR-peak-gene links with hypomethylation in skin T_reg cells (top) or blood CD45RA⁺ T_reg cells (bottom). Columns indicate samples from different donors (left, right) or donor-wise mean accessibility across cells (middle). n = 3 donors on the methylation level, 2 donors on the accessibility level, 4 (skin T_reg) and 5 (blood CCR8⁺ T_reg, blood CD45RA⁺ T_reg) donors on the expression level. Annotation bars on the right of each heat map indicate whether blood CCR8⁺ T_reg cells are closer to skin T_reg cells or closer to blood CD45RA⁺ T_reg cells with respect to a feature (based on the mean across donors; see Methods for details). Parts of this panel are duplicated from Fig. 5a as the same features are shown. d, Smoothed methylation and chromatin accessibility in regions where chromatin accessibility of the overlapping peak places blood CCR8⁺ T_reg cells closer to skin T_reg cells (top) and closer to blood CD45RA⁺ T_reg cells (bottom). These regions are shown together with corresponding raw methylation for each displayed CpG site (top right heat maps; rows indicate donors; columns indicate CpG sites; gray fields indicate missing values) and expression of the associated gene (bottom right bar charts; mean ± s.d.). Highlighted regions mark DMRs (methylation tracks) and differential peaks (accessibility tracks). Vertical lines at the bottom of the methylation tracks mark CpG sites. P values, two-tailed Wald test based on the raw count matrix as implemented in DESeq2 (see Methods; n = 4 skin T_reg donors, 5 blood CCR8⁺ T_reg donors, 5 blood CD45RA⁺ T_reg donors; unpaired sample structure) with Benjamini-Hochberg correction. Data are representative of two or more independent experiments with two or more individual donors.

Extended Data Fig. 7 Details of motif analysis regarding the positioning of blood CCR8⁺ T_reg cells.

a, Logos for motifs highlighted in Fig. 5d (obtained from homer). b, Chromatin accessibility around genomic sites for additional bZIP motifs. Parts of these plots are duplicated from Extended Data Fig. 4 as they correspond to the same motifs. c, Enrichment of bHLH motifs in methylation-level region sets defined by different blood CCR8⁺ T_reg cell positionings. P values, one-tailed binomial test implemented in homer assessing enrichment of motif sites among the region sets compared to background sets of random regions (see Methods; n = 6,424 DMRs and 292,033 DMRs for columns 1 and 2, respectively) with Benjamini-Hochberg correction. d, Motifs highlighted in c (obtained from homer). Parts of this figure are duplicated from Extended Data Fig. 5 as they correspond to the same motifs. Data are representative of two or more independent experiments with two or more individual donors.

Extended Data Fig. 8 Details of scRNA-seq/scTCR-seq data analysis and Kaede mice experiments regarding the positioning of blood CCR8⁺ T_reg cells.

a–d, UMAP representations of cells from scRNA-seq/scTCR-seq, showing clusters and cell type annotation next to gene expression and signature values underlying the annotation. Data correspond to blood T_reg cells from TCR Donor 6 (sample 28; a), blood CD45RO⁺ T_reg cells from TCR Donor 6 (sample 29, b), blood T_reg cells from TCR Donor 7 (sample 24, c) and blood CD45RO⁺ T_reg cells from TCR Donor 7 (sample 25, d). Blood CCR8⁺ T_reg signature 1 defines blood CCR8⁺ T_reg cells against blood CD45RA⁺ T_reg cells. Blood CCR8⁺ T_reg signature 2 defines blood CCR8⁺ T_reg cells against blood CD45RO⁺ (memory) T_reg cells. For details regarding the annotation, see the Methods section. e, Stacked bar charts showing TCR clonotype sharing between blood CD45RA⁺ T_reg cells, skin T_reg cells and blood CCR8⁺ T_reg cells. Bar sections indicate the percentage of cells belonging to a particular clonotype. Connections between bar sections indicate that a clonotype was found in two cell types. All clonotypes that appear in skin T_reg cells and contain at least one TCR-α and TCR-β chain are shown (271 clonotypes from TCR Donor 6, 602 clonotypes from TCR Donor 7). f, Percentages of skin T_reg cell clonotypes that were also found among blood CD45RA⁺ T_reg cells and blood CCR8⁺ T_reg cells. g, Kaede-green and Kaede-red expression in skin-draining-lymph-node-derived T_reg cells of Kaede mice 6 days after photoconversion of skin. Cells were pre-enriched for CD25 and gated on size, viability, TCR-β, CD4, CD25 and GITR. Concatenated data of 7 mice. h, CCR8 expression was measured in T_reg and T_conv cells 6 days after light exposure of skin of Kaede mice. Cells were harvested from LN and enriched for CD25 expression. Shown here are the T_conv cells present in CD25-enriched fractions, results for T_reg cells are shown in the main figure. Concatenated data of 7 mice. Data are representative of two or more independent experiments with two or more individual donors/mice.

Extended Data Fig. 9 The contribution of TEs to the epigenetic landscape of skin T_reg cells and blood CCR8⁺ T_reg cells.

a, b, Enrichment of TE classes (left) and subfamilies (right) in regions hypomethylated (a) or hyperaccessible (b) in blood CD45RA⁺ T_reg cells. P values: one-tailed permutation test assessing enrichment of TE insertion sites among DMRs or peaks (see Methods) based on 1,742 DMRs (a) and 5,722 differential peaks (b) displayed in Fig. 3 with Benjamini-Hochberg correction. c, Left, Chromatin accessibility of peaks overlapping with annotated LTR-TE, DNA-TE and SINE insertion sites (n = 2 donors per cell type). White horizontal lines inside violins indicate medians across all peaks. Right, Distribution of differences in mean accessibility (across the two donors) between skin T_reg cells and blood CD45RA⁺ T_reg cells for each peak. Numbers indicate how many insertion sites displayed differences below and above 0. d, Left, Methylation of annotated LTR, DNA-TE and SINE insertion sites as in c with skin T_reg, fat T_reg and blood CD45RA⁺ T_reg (n = 3 donors per cell type). White horizontal lines inside violins indicate medians across all insertion sites. Right, Distribution of differences in mean methylation (across the three donors) between any two cell types for each insertion site. Numbers indicate how many insertion sites displayed differences below and above 0. e, Smoothed methylation and chromatin accessibility around selected TE insertion sites. Highlighted regions mark DMRs (methylation tracks), differential peaks (accessibility tracks) and TE insertion sites overlapping with DMRs (only TE subfamilies enriched among the DMRs; gene tracks). Vertical lines at the bottom of the methylation tracks mark CpG sites (CpG sites overlapping with TEs are shown in red). TE labels correspond to TE subfamilies. Parts of these plots are duplicated from Fig. 3 and Extended Data Fig. 3 as they show the same genomic regions. f, Top, Chromatin accessibility of peaks displaying hyperaccessibility in skin T_reg cells and overlapping with DNA-TE, LINE, LTR-TE and SINE insertion sites (n = 2 donors per cell type). White horizontal lines inside violins indicate medians across all peaks. Bottom, Distribution of differences in mean accessibility (across the two donors) between any two cell types for each peak. Numbers indicate how peaks displayed differences below and above 0. Data are representative of two or more independent experiments with two or more individual donors.

Extended Data Fig. 10 Methylation and accessibility of HERVIP10F-int and LTR45B insertion sites.

a, Left, Methylation of annotated HERVIP10F-int and LTR45B insertion sites in skin T_reg, fat T_reg and blood CD45RA⁺ T_reg cells (n = 3 donors per cell type). White horizontal lines inside violins indicate medians across all insertion sites. Right, Distribution of differences in mean methylation (across the three donors) between any two cell types for each insertion site. Numbers indicate how many insertion sites displayed differences below and above 0. b, Chromatin accessibility of peaks overlapping with annotated HERVIP10F-int and LTR45B insertion sites in skin T_reg, blood CCR8⁺ T_reg and blood CD45RA⁺ T_reg cells. White horizontal lines inside violins indicate medians across all peaks. Data are representative of two or more independent experiments with two or more individual donors.

Supplementary information

Reporting Summary

Supplementary Table 1

All supplementary tables combined into a single workbook.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Beumer, N., Imbusch, C.D., Kaufmann, T. et al. DNA hypomethylation traits define human regulatory T cells in cutaneous tissue and identify their blood recirculating counterparts. Nat Immunol 26, 1315–1328 (2025). https://doi.org/10.1038/s41590-025-02210-x

Download citation

Received: 11 October 2024
Accepted: 09 June 2025
Published: 16 July 2025
Issue date: August 2025
DOI: https://doi.org/10.1038/s41590-025-02210-x