
==== Front
Sci Adv
Sci Adv
sciadv
advances
Science Advances
2375-2548
American Association for the Advancement of Science

39231218
adn3470
10.1126/sciadv.adn3470
Research Article
Biomedicine and Life Sciences
SciAdv r-articles
Health and Medicine
Immunology
Immunology
Effector memory–type regulatory T cells display phenotypic and functional instability
Safeguarding Treg cell stability
https://orcid.org/0000-0002-1399-787X
Wendering Désirée Jacqueline Conceptualization Data curation Formal analysis Investigation Methodology Validation Visualization Writing - original draft Writing - review & editing 1 2 †
https://orcid.org/0000-0003-4855-5905
Amini Leila Investigation Methodology Resources Writing - original draft Writing - review & editing 3 4 †
https://orcid.org/0000-0003-3142-2890
Schlickeiser Stephan Formal analysis Software Visualization 1 5
https://orcid.org/0000-0002-8170-1276
Farrera-Sal Martí Investigation Methodology 6
https://orcid.org/0000-0003-3719-6652
Schulenberg Sarah Investigation Methodology 6 7
https://orcid.org/0000-0003-1205-7445
Peter Lena Conceptualization Investigation Methodology 6 7
Mai Marco Methodology 6
https://orcid.org/0000-0002-0358-3876
Vollmer Tino Conceptualization Investigation Methodology Writing - review & editing 6
https://orcid.org/0000-0001-5924-8631
Du Weijie Formal analysis Investigation Methodology Resources Validation Visualization Writing - original draft Writing - review & editing 4 8
Stein Maik Investigation 3 8
https://orcid.org/0009-0002-2482-9916
Hamm Frederik Formal analysis Investigation Methodology Visualization 9 10
https://orcid.org/0009-0000-0150-7207
Malard Alisier Investigation Methodology Validation 9
Castro Carla Conceptualization Data curation Formal analysis Funding acquisition Investigation Methodology Project administration Resources Software Supervision Validation Visualization Writing - original draft Writing - review & editing 9
Yang Mingxing Formal analysis 9
https://orcid.org/0009-0009-1284-4095
Ranka Ramon Methodology Resources 10
https://orcid.org/0000-0003-1433-1372
Rückert Timo Investigation Methodology Resources Supervision Writing - review & editing 10
https://orcid.org/0000-0002-6179-0670
Durek Pawel Data curation Formal analysis Software Visualization Writing - review & editing 10
Heinrich Frederik Formal analysis Software 10
https://orcid.org/0000-0002-6423-4637
Gasparoni Gilles Data curation Formal analysis Investigation Project administration 11
https://orcid.org/0000-0002-9649-6922
Salhab Abdulrahman Formal analysis Methodology Software 11 ‡
https://orcid.org/0000-0003-0563-7417
Walter Jörn Funding acquisition Methodology Resources Visualization Writing - original draft 11
https://orcid.org/0000-0002-2189-3579
Wagner Dimitrios Laurin Investigation Methodology Resources Supervision Validation Writing - review & editing 4 8 12 13 §
https://orcid.org/0000-0002-8015-6907
Mashreghi Mir-Farzin Data curation Funding acquisition Investigation Project administration Supervision 10
https://orcid.org/0000-0002-4836-7659
Landwehr-Kenzel Sybille Investigation Methodology Writing - review & editing 2 3 14
https://orcid.org/0000-0003-4727-2540
Polansky Julia K. Conceptualization Funding acquisition Methodology Project administration Supervision Writing - review & editing 9 10
https://orcid.org/0000-0003-0771-4375
Reinke Petra Conceptualization Funding acquisition Methodology Resources Supervision Validation Writing - original draft Writing - review & editing 3 4
https://orcid.org/0000-0002-7743-6668
Volk Hans-Dieter Conceptualization Funding acquisition Methodology Supervision Validation Writing - review & editing 1 4 5 13
https://orcid.org/0000-0001-5964-9179
Schmueck-Henneresse Michael Conceptualization Formal analysis Funding acquisition Methodology Project administration Resources Supervision Validation Visualization Writing - original draft Writing - review & editing 6 *
1 Berlin Institute of Health (BIH) at Charité–Universitätsmedizin Berlin, BIH Center for Regenerative Therapies (BCRT), Development of Biomarkers and Regenerative Therapies, Augustenburger Platz 1, 13353 Berlin, Germany.
2 Hannover Medical School, Institute of Transfusion Medicine and Transplant Engineering, Carl-Neuberg-Str. 1, 30625 Hannover, Germany.
3 Berlin Institute of Health (BIH) at Charité–Universitätsmedizin Berlin, BIH Center for Regenerative Therapies (BCRT), Cell Therapy and Personalized Immunosuppression, Augustenburger Platz 1, 13353 Berlin, Germany.
4 Berlin Center for Advanced Therapies (BeCAT) at Charité–Universitätsmedizin Berlin, Augustenburger Platz 1, 13353 Berlin, Germany.
5 CheckImmune GmbH, Augustenburger Platz 1, 13353 Berlin, Germany.
6 Berlin Institute of Health (BIH) at Charité–Universitätsmedizin Berlin, BIH Center for Regenerative Therapies (BCRT), Experimental Immunotherapy, Augustenburger Platz 1, 13353 Berlin, Germany.
7 Einstein Center for Regenerative Therapies at Charité–Universitätsmedizin Berlin, Augustenburger Platz 1, 13353 Berlin, Germany.
8 Berlin Institute of Health (BIH) at Charité–Universitätsmedizin Berlin, BIH Center for Regenerative Therapies (BCRT), Gene Editing for Cell Therapy, Augustenburger Platz 1, 13353 Berlin, Germany.
9 Berlin Institute of Health (BIH) at Charité–Universitätsmedizin Berlin, BIH Center for Regenerative Therapies (BCRT), Immuno-Epigenetics, Augustenburger Platz 1, 13353 Berlin, Germany.
10 Deutsches Rheuma-Forschungszentrum Berlin, an Institute of the Leibniz Association, Charitéplatz 1, 10117 Berlin, Germany.
11 Saarland University, Institute for Genetics/Epigenetics, Saarbrücken, Germany.
12 Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt–Universität zu Berlin, Institute of Transfusion Medicine, Charitéplatz 1, 10117 Berlin, Germany.
13 Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt–Universität zu Berlin, Institute of Medical Immunology, Augustenburger Platz 1, 13353 Berlin, Germany.
14 Hannover Medical School, Department of Pediatric Pulmonology, Allergy and Neonatology, Carl-Neuberg-Str. 1, 30625 Hannover, Germany.
* Corresponding author. Email: michael.schmueck-henneresse@bih-charite.de
† These authors contributed equally to this work.

‡ Present address: Genomics Data Science Core, Integrated Genomics Services, Sidra Medicine, Doha, Qatar.

§ Present address: Baylor College of Medicine, Center for Cell and Gene Therapy, Houston, TX, USA.

06 9 2024
04 9 2024
10 36 eadn347006 12 2023
30 7 2024
Copyright © 2024 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC).
2024
The Authors
https://creativecommons.org/licenses/by-nc/4.0/ This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

Regulatory T cells (Treg cells) hold promise for sustainable therapy of immune disorders. Recent advancements in chimeric antigen receptor development and genome editing aim to enhance the specificity and function of Treg cells. However, impurities and functional instability pose challenges for the development of safe gene-edited Treg cell products. Here, we examined different Treg cell subsets regarding their fate, epigenomic stability, transcriptomes, T cell receptor repertoires, and function ex vivo and after manufacturing. Each Treg cell subset displayed distinct features, including lineage stability, epigenomics, surface markers, T cell receptor diversity, and transcriptomics. Earlier-differentiated memory Treg cell populations, including a hitherto unidentified naïve-like memory Treg cell subset, outperformed late-differentiated effector memory–like Treg cells in regulatory function, proliferative capacity, and epigenomic stability. High yields of stable, functional Treg cell products could be achieved by depleting the small effector memory–like Treg cell subset before manufacturing. Considering Treg cell subset composition appears critical to maintain lineage stability in the final cell product.

Safeguarding the integrity of regulatory T cell products requires depleting the effector memory–like subset.

http://dx.doi.org/10.13039/100010663 H2020 European Research Council 803992 German Federal Ministry of Education and Research European Union’s Horizon 2020 Research and Innovation Program No 825392 Einstein Center for Regenerative Therapies The state of Berlin and the European Regional Development Fund ERDF 2014–2020, EFRE 1.8/11 Deutsches Rheuma-Forschungszentrum Crossfield project fund of the BIH Research Focus Regenerative Medicine BIH Research Platform Clinical Translational Sciences grant The Leibniz Association Leibniz Competition Collaborative Excellence Grant K59/2017 ‘EpImAge’
==== Body
pmcINTRODUCTION

Regulatory T cells (Treg cells) maintain immune homeostasis, resolve inflammation, and promote self-tolerance (1). They suppress immune responses through various molecular mechanisms and can reshape the immune balance, making them potential therapeutic agents with sustained efficacy in (hyper)inflammatory diseases and unwanted immunity. First clinical trials of Treg cell immunotherapies have shown their feasibility, tolerability, and potential efficacy in establishing and maintaining immune tolerance after solid organ transplantation and hematopoietic stem cell transplantation, as well as autoimmunity and inflammatory bowel diseases (2–6). Treg cells are identified by the lineage marker Forkhead-box-protein P3 (FoxP3) and account for 5% of circulating CD4+ T cells (7, 8). Competition for interleukin-2 (IL-2) via elevated CD25 expression of Treg cells can lead to decreased numbers and function of conventional T cells (TCONV cells), the main drivers of inflammation and autoimmunity (9–14). The IL-7 receptor subunit α (CD127) inversely correlates with FoxP3 expression and suppressive function of Treg cells and is used along with CD25 to purify these cells (CD4+CD25+CD127low) (15). Since FOXP3 is also transiently up-regulated in nonregulatory CD4+CD25− T cells upon activation (16–18), demethylation of the Treg cell–specific demethylated region (TSDR), a CpG-rich element within the FoxP3 locus, is the strongest indicator of a stable Treg cell lineage (19). In extreme inflammation, Treg cells can become unstable, lose their FoxP3 expression and immunosuppressive function, and convert to effector T cells (TEFF cells) (20).

Recent advancements in chimeric antigen receptor (CAR) development and genome editing have enabled the genetic optimization of primary T cell therapies for cancer and infectious disease (21, 22). These technologies are now being applied to improve the specificity and functionality of Treg cells. Although contamination with small amounts of non–Treg cell in first-generation polyclonal Treg cell products used in clinical trials has not raised safety concerns so far [e.g. (2)], their diminishing effect on therapeutic efficacy cannot be ignored. The presence of non-Treg effector cells or the instability of Treg cells in gene-edited Treg cell products with redirected specificities (such as CAR-Treg cells), improved product survival (IL-2 self-support), tissue targeting, or immunosuppressive drug resistance is highly undesirable. If these alterations occur during the manufacturing process and contaminate Treg cell preparations or if specific Treg cell subsets convert to TEFF cells or become instable after infusion, the resulting reactions could compromise the safety and efficacy of Treg cell therapies.

There are limited data on the optimal source of Treg cells for the development of stable and effective suppressive cellular therapies for infusion. Manufacturing pure and stable Treg cell products for cost-effective therapies is challenging because of the small quantity and potential contamination of the product. These issues led to the exploration of stable FoxP3 overexpression through gene transfer or epigenetic modification to improve Treg cell product development (23, 24). However, these methods have limited resistance to excessive inflammation, raising safety concerns and costs. Alternatively, the use of naïve Treg cells offers some resistance to inflammation, but this approach suffers not only from their low frequency in the blood, resulting in low yields (25, 26), but also from the fact that they do not deliver immediate functionality such as memory cells. Within the TCONV cell compartment, central memory T cells (TCM cells) expressing CCR7 and CD45RO and stem cell memory T cells (TSCM cells) expressing CCR7, CD45RA, and CD95 have demonstrated long-lasting effects upon adoptive transfer, in contrast to differentiated effector memory T cells (TEM cells) lacking CCR7 and CD45RA (27–31). This is attributed to their high proliferation potential, ability to self-renew, superior engraftment, and long-term persistence (27, 31–33). Comparable markers for defining functional Treg cell memory subsets have been less clearly defined to date, primarily due to the lack of Treg cell–specific molecules expressed de novo upon activation. CD45RA and CCR7 are differentially expressed on Treg cells, and corresponding subsets analogous to TCM cells (TregCM cells) and TEM cells (TregEM cells) have been described (34–38). Moreover, terminally differentiated, short-lived, and markedly suppressive Treg cells characterized by high expression of FoxP3 have already been identified (7, 26, 39).

Thus, current approaches, such as using naïve Treg cells as starting material for Treg cell production or overexpressing FoxP3 by genetic manipulation (24, 40), are assumed to be insufficient to produce Treg cells in high yields that are effective and stable in an inflammatory environment. We hypothesize that different Treg cell subsets may have distinct phenotypic and functional properties that are optimal for different Treg cell therapy products. Recently, different Treg cell subsets have been described on the basis of surface staining commonly used for TCONV cells (41). To identify the characteristics and subset derivation of Treg cells used in clinical studies, we sorted each Treg cell subset using classical surface markers to define differentiation stages and to study their function, epigenomic stability, propensity to dedifferentiate, transcriptomes, and T cell receptor (TCR) repertoires. Each Treg cell subset exhibited distinct characteristics, with the earlier-differentiated memory Treg cell populations showing superior immunosuppressive function and stability compared to the small subset of late-differentiated TregEM cells. Depletion of TregEM cells before manufacturing resulted in the production of stable Treg cell products. In summary, the composition of Treg cell subsets before manufacturing is critical to maintain lineage stability in the final product.

RESULTS

Composition of subsets defining T cell differentiation stages suggests coherence in memory formation between bulk CD4+ T cells and Treg cells

To investigate the phenotype of TCONV cells and Treg cells ex vivo, we used an extensive flow cytometry panel, which allowed the identification of subsets within both CD4+ T cell lineages. Freshly isolated peripheral blood mononuclear cells (PBMCs) from 53 healthy donors were labeled with specific monoclonal antibodies for flow cytometry analysis (Fig. 1 and Materials and Methods). The gating strategy involved lymphocyte discrimination, doublet exclusion, and the selection of live CD3+CD4+ T cells (Fig. 1A and fig. S1). Within the CD4+ T cell population, Treg cells were identified on the basis of their expression of CD25 and FoxP3 (CD4+FoxP3++CD25++). Conventional memory CD4+ TCM, TEM, and terminally differentiated effector memory T cells (TEMRA) cell subsets were defined according to CD45RA and CCR7 expression (TCM, CCR7+CD45RA−; TEM, CCR7−CD45RA−; TEMRA, CCR7−CD45RA+). After excluding CD45RO+CD62L− memory T cells from the CCR7+CD45RA+ population, the TSCM cell and naïve T cell (TN cell) subsets were defined by means of their differential expression of CD95 and CCR7 (TCM, CCR7+CD45RA−; TEM, CCR7−CD45RA−; TEMRA, CCR7−CD45RA+; TN, CCR7+CD45RA+CD95−; TSCM, CCR7+CD45RA+CD95+). Accordingly, Treg cell memory subsets (TregCM, TregEM, and TregEMRA cells) were defined on the basis of CCR7 and CD45RA expression (TregCM, CCR7+CD45RA−; TregEM, CCR7−CD45RA−; TregEMRA, CCR7−CD45RA+). Similarly, after stringent exclusion of residual memory T cells by excluding CD45RO+CD62L− T cells, memory stem–like CD4+ Treg cells [which we refer to as naïve-like memory (TregNLM) because stem cell properties as in conventional TSCM cells could not be shown so far], and naïve Treg cells (TregN cells) were delineated on the basis of the differential expression of CD95 and CCR7 (TregN, CCR7+CD45RA+CD95−; TregNLM, CCR7+CD45RA+CD95+) (Fig. 1A and fig. S1). All conventional memory and the TN cell subsets could be retrieved from the Treg cell compartment. However, the CM and EM populations within the bulk CD4+ TCONV cells (mean, 32.2 and 7.1%, respectively) were detected only half as frequently as TregCM cells (mean, 52.0%) and TregEM cells (mean, 16.8%). Thus, most Treg cells exhibited a CM phenotype. Notably, naïve cells in the CD4+ T cell bulk (mean, 58.3%) were twice as frequent as naïve cells in the Treg cell population. Given the substantial differences in the frequencies of naïve, CM, and EM cells between bulk CD4+ T cells and Treg cells, it is intriguing that the frequency of TCONVSCM cells in the bulk CD4+ T cell compartment (mean, 2.8%) is very similar to that of TregNLM cells (mean, 2.6%) (Fig. 1B).

Fig. 1. Composition of subsets defining T cell differentiation stages suggests coherence in memory formation between bulk CD4+ T cells and Treg cells.

(A) Gating strategy for defining conventional CD4+ T cells and CD4+CD25+FoxP3+ Treg cells from PBMCs. Illustrated are density plots of flow cytometry data of one representative donor [fluorescence minus one (FMO) controls shown in fig. S1]. SSC-A, Side Scatter Area; FSC-A, Forward Scatter Area; FSC-H, Forward Scatter Height. (B) Summary of ex vivo investigation of defining conventional CD4+ T cells and Treg cell subset distribution of freshly isolated PBMCs based on flow cytometry (A). Statistical analysis for differences between cell subsets by two-way analysis of variance (ANOVA) with Holm-Sidak testing for multiple comparisons is shown. n = 53. Black lines in violin plots show the median. *P ≤ 0.05; ****P ≤ 0.0001. ns, not significant. Scatter plot diagrams show Pearson correlation analysis between (C) bulk conventional CD4+ T cells and their subsets, as well as (D) bulk Treg cells and Treg cell subsets with age. Age (in years) and frequencies of Treg cell subsets are shown on the x and y axes, respectively. (E) Scatter plot diagrams show Pearson correlation analysis between frequencies of Treg cell and conventional CD4+ T cell subsets. Frequencies of conventional CD4+ T cells and Treg cell subsets are shown on the x and y axes, respectively. n = 53. The respective Pearson correlation coefficients (r) are reported in each plot. *P < 0.05 and ****P < 0.00005.

To investigate whether the heterogeneous Treg cell subset composition changes over the course of a lifetime toward a higher abundance of memory cells at the expense of the naïve compartment, analogous to the bulk CD4+ T cell compartment, we analyzed the frequencies of total Treg and Treg cell subsets by Pearson correlation as a function of chronological age (Fig. 1, C to E). The data revealed no significant correlation between bulk Treg cell frequencies and age, suggesting that Treg cell frequencies in peripheral blood remain at a constant level during aging. In contrast, the frequency of EM Treg cells correlated significantly and positively with age, indicating a significant increase in late-differentiated Treg cells with increasing age. TregN and TregNLM cell frequencies correlated significantly and negatively with age, suggesting that the frequencies of both Treg cell subsets decrease with age (Fig. 1C). However, the difference in correlation strength between TregNLM and TregN cells, as determined by Pearson correlation coefficients (r) of −0.3793 and − 0.5519, respectively, suggests a greater decrease in TregN cell frequencies with age compared with TregNLM cells. The frequencies of TregCM cells, which constitute the largest proportion of the Treg cell compartment (Fig. 1B), showed no significant correlation with age, suggesting a stable TregCM cell population as we age (Fig. 1C).

Since the plasticity and differentiation potential within the CD4+ and CD8+ TCONV cell compartments are well elucidated, we investigated possible correlations between the frequencies of Treg cell subsets and their TCONV CD4+ T cell counterpart (Fig. 1E). We found significant positive correlations between all Treg cell subsets and their respective bulk CD4+ T cell populations; hence, as the abundance of either Treg/TCONV cell increases, so does its corresponding counterpart. Notably, the strength of correlation between the TCONVN/TregN cell (r = 0.5287) and TCONVSCM/TregNLM cell (r = 0.5990) subsets is greater than that of TCONVEM/TregEM cell (r = 0.3270) and TCONVCM/TregCM cell (r = 0.5171) subsets, implying that the strongest relationship is between the frequencies of TCONVN/TregN and TCONVSCM/TregNLM cells. Together, these results suggest a coherence in memory formation between Treg cells and bulk CD4+ T cells, as memory-like Treg cell subsets corresponding to their CD4+ TCONV cell counterparts could be detected.

Phenotypical and functional ex vivo characterization of CD4+ TCONV and Treg cell subsets

For a comprehensive phenotypical characterization of Treg cell subsets, we analyzed the mean fluorescence intensity (MFI) of the markers FoxP3 and CD25, which determine the Treg cell lineage. The highest MFIs of CD25 were observed among TregEM cells, which decreased significantly from TregCM to TregNLM to TregN cells (Fig. 2A). The MFI of FoxP3 exhibited significant differences between TregEM and TregN/NLM cells, as well as between TregCM and TregN cells (Fig. 2A). CD31 (platelet endothelial cell adhesion molecule–1), a marker for cells recently migrated from the thymus (42), decreased significantly from TregN to TregNLM to TregCM to TregEM cells. The same pattern of CD31 expression was observed in conventional memory subset counterparts (Fig. 2B). Ki-67, a proliferation marker used to identify dividing cell populations (43), showed an inverse expression profile with the highest expression in TregEM > TregCM > TregNLN > TregN cells (Fig. 2C). Recent TCR engagement was analyzed by CD134 (OX-40) expression (44), revealing the highest expression in TregCM and TregEM cells, followed by TregNLM cells and very low expression in TregN cell subset (Fig. 2D) similar to the pattern observed for TCONV cells.

Fig. 2. Phenotypical and functional ex vivo characterization of CD4+ TCONV cell and Treg cell subsets.

(A) MFI of CD25 and FoxP3 of CD4+CD25+FoxP3+ Treg cells. n = 12. Black lines indicate the mean with SEM. Ex vivo isolated and unstimulated PBMCs were analyzed by flow cytometry on the basis of extracellular, intracellular, and intranuclear proteins labeled with fluorochrome-conjugated monoclonal antibodies. n = 12. Frequencies within Treg cell and conventional CD4+ T cell subsets of CD31 (B), Ki-67 (C), and CD134 (D). n = 6. Results are presented as means ± SEM. Normalized (E) CD45RO, (F) CD45RA, (G) GZMA, (H) ceramide synthase 6 (CERS6), and (I) KLRG mRNA expression in ex vivo fluorescence-activated cell sorted (FACSorted) CD4+CD25+FoxP3+ Treg cell subsets. Depicted are floating bars indicating minimum and maximum values and a line indicating the mean. n = 3. (J) Gating strategy for analyzing ex vivo proinflammatory cytokine production by conventional CD4+ T cells and CD4+CD25+FoxP3+ Treg cells. PBMCs were stimulated with phorbol 12-myristate 13-acetate (PMA)/ionomycin for 6 hours, permeabilized, and intracellularly stained for IFN-γ (K) and IL-2 (L). Illustrated are density plots of flow cytometry data of one representative donor. n = 32 for IFN-γ and n = 24 for IL-2 expression, respectively. Black lines in violin plots show the median. In (A) to (L), statistical significance for differences between cell subsets was determined by two-way ANOVA with Tukey multiple comparison correction *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. (M) Suppression capacity of ex vivo isolated Treg cell subsets. Freshly isolated PBMCs were FACSorted for Treg cell subsets and CD4+CD25− TCONV cells serving as TRESP cells and to control for Treg cell–specific proliferation suppression. Carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled TCONV cells were cocultured with increasing numbers of Treg cells and stimulated with anti-CD3/CD28–coupled microbeads. After 96 hours of culture, proliferation was assessed on the basis of CFSE signal. n = 3. Results are presented as means ± SEM.

Next, we assessed the transcriptional expression levels of CD45RA and CD45RO to distinguish the stages of differentiation. CD45RO mRNA expression, a marker for memory T cells, was the lowest in TregN and TregNLM cell subsets, increased in TregCM cells, and peaked in TregEM cells (Fig. 2E). CD45RA, expressed on naïve CD4+ TCONV cells, exhibited the highest mRNA expression in naïve Treg cells and was only slightly lower in TregNLM cells and almost absent in TregCM and TregEM cells (Fig. 2F). Since thymus-derived activated Treg cells have been reported to express the cytolytic marker granzyme A (GZMA) (45), we also examined its transcriptional expression. Normalized mRNA expression levels of GZMA were generally low, but the EM Treg cells had the highest expression levels, followed by TregCM, TregNLM, and TregN cells, which hardly expressed any GZMA (Fig. 2G). The ceramide synthase 6 (CERS6), which has been shown to play a role in inhibiting T cell activation and differentiation and promoting cellular quiescence by regulating intracellular ceramide production (46, 47), was expressed most abundantly at the transcriptional level in TregN cells, followed by markedly lower levels in TregNLM, TregCM, and TregEM cells (Fig. 2H). While Killer cell lectin-like receptor G1 (KLRG1) has traditionally been associated with a senescent state in TCONV cells (48), we observed increased KLRG1 mRNA expression in early-differentiated TregNLM and TregN cell subsets, which gradually decreased as differentiation progressed (Fig. 2I).

Typically, Treg cells do not produce proinflammatory cytokines such as interferon-γ (IFN-γ) and IL-2 in response to TCR stimuli and can therefore be characterized by absence of proinflammatory cytokine production. Thus, the proinflammatory cytokine profiles were analyzed via intracellular detection of IFN-γ and IL-2 in freshly isolated bulk CD4+ T cells and Treg cell subsets upon polyclonal stimulation with phorbol 12-myristate 13-acetate (PMA)/ionomycin compared to unstimulated cells (Fig. 2J). Treg cell subsets did not produce notable amounts of IFN-γ and IL-2 (each <5%), whereas bulk memory CD4+ T cells produce substantial amounts of both cytokines, thereby serving as a positive control for polyclonal stimulation (Fig. 2, K and L). IFN-γ producers were the most abundant among conventional CD4+ TEM cells, followed by TCM and TSCM cells with comparable frequencies and naïve CD4+ T cells with the lowest frequency (Fig. 2K). IL-2 producers were similarly abundant among conventional CD4+ T cells, including TEM, TCM, and TSCM cells, with naïve CD4+ T cells exhibiting the lowest frequency (Fig. 2L).

Next, we assessed the suppressive capacities of ex vivo isolated Treg cells. To this end, we used an in vitro suppression assay (49) that measures the proliferation of autologous responder T cells (TRESP cells) when cocultured with varying ratios of Treg cells. After 96 hours of coculture, all Treg cell subsets exhibited suppressive capabilities and inhibited proliferation of TRESP cells. TregNLM and TregCM cells were particularly effective in suppressing TRESP cell proliferation, especially at ratios of 1:1 and 2:1 (TRESP:Treg cells), respectively. TregN cells demonstrated the lowest suppression capacities (Fig. 2M). Together, the comprehensive phenotypic and functional characterization of Treg cell subsets ex vivo demonstrated distinct phenotypic and functional properties, including marker expression and suppressive capacity.

Analyses of clonal TCR diversity and FoxP3-locus methylation provide insights into the identity and lineage stability of the distinct Treg cell memory differentiation subsets

Subsequently, we conducted an analysis of the TCR repertoire to study the clonality of T cell lineages during their differentiation process to gain further insight into the pathway of Treg cell differentiation. Among all Treg cell subsets, the most-differentiated Treg cell subset, TregEM cells, showed the highest clonality (0.1027) with the most oligoclonal TCR repertoire (Fig. 3A). Although the clonality of the two donors analyzed differed slightly, TregCM cells had higher diversity in its TCR repertoire (0.0041), making it the population closest to the maximum diversity of TregEM cells. Notably, both TregNLM and TregN cell subsets had remarkably low levels of clonality, suggesting that these two populations have the highest TCR diversity among the subsets analyzed. However, the TregNLM cell subset retained a small but considerably greater degree of clonality (0.0027), suggesting a more oligoclonal TCR repertoire compared with the TregN cell population (0.001) (Fig. 3A). Analyzing the productive frequency of the top 10 clones, it was observed that the TregEM cell subset exhibited higher frequencies in these top clones than the TregCM > TregNLM > TregN cell subsets (Fig. 3B).

Fig. 3. TCR diversity and FoxP3-locus methylation analyses of the distinct Treg cell memory differentiation subsets.

(A) TCRβ repertoire analysis of ex vivo FACSorted Treg cell subsets. n = 2. Mean data are shown. (B) Intuitive view depicting the relative proportions of the top 10 clones of each Treg cell subset of one donor. (C) Ex vivo FACSorted Treg cell subsets were analyzed by quantitative polymerase chain reaction (PCR) to define the percentage of TSDR demethylation. n = 3. Results are presented as means ± SEM. (D) Weighted average DNA methylation in partially methylated domains (PMDs) and highly methylated domains (HMDs) based on reduced representation bisulfite sequencing (RRBS) data. Weighted average methylation in PMD and HMD: RRBS data to determine the weighted average DNA methylation across defined DNA segments, divided into PMD and HMD. A loss of methylation is observed in the order: TregN > TregNLM > TregCM > TregEM. The loss of methylation is more prominent in PMDs. n = 6. Statistical significance for differences between TN, TSCM, TCM, and TEM cells but not bulk conventional CD4+ T cells was determined by t test.

To analyze Treg cell subset–specific lineage stability, we evaluated the degree of selective TSDR demethylation of ex vivo Treg cell subsets. CD4+CD25− TCONV cells, which have been shown to lack demethylated TSDR (19), were also analyzed as an assay-specific reference control. CD4+CD25− TCONV cells displayed a fully methylated TSDR (0.19% demethylation) (Fig. 3C). We observed a mean TSDR demethylation of 90.16% within the bulk Treg cell population. Within the Treg cell subsets, the CM population had the lowest degree of TSDR demethylation (83.89%), followed by the more-differentiated TregEM cell subset (89.32%). Notably, TregN cells exhibited the highest level of TSDR demethylation, reaching 95.03%. This was closely followed by TregNLM cells with a TSDR demethylation rate of 90.75% (Fig. 3C). A progressive loss of DNA methylation in the transcriptionally silent part of the genome (heterochromatin) in TCONV cells (TN > TCM > TEM) that was associated with a proliferative history has been reported (50). To test whether this phenomenon is also found in the different Treg cell subsets, we performed methylation analysis providing data on highly methylated domains (HMDs; representing mainly expressed gene bodies) and partially methylated domains (PMDs; mainly representing the heterochromatin). While the differences between HMDs were less pronounced, we found a strong increase in average methylation levels in PMDs from TregEM to TregCM to TregNLM to TregN cells (Fig. 3D), consistent with data for TCONV cell subsets (50).

In summary, analyses of the TCR repertoire and methylation revealed distinct clonality patterns and lineage stability among the Treg cell subsets, with TregEM cells being the most clonal, TregCM cells being the closest to mean diversity of bulk Treg cells, and TregNLM and TregN cells having the highest TCR diversity. TSDR demethylation identifies the early-differentiated TregN and TregNLM cell subsets as the most stable, while progressive loss of PMD methylation (TregN > TregNLM > TregC > TregEM) indicates an accumulation of proliferation episodes, similar to what has been observed and reported for TCONV cells.

Single-cell transcriptome and surface epitope analysis (CITE-seq) confirms coherence of memory formation between CD4+ T cells and Treg cells

Using canonical surface markers to define Treg cell differentiation stages and subsets, we found phenotypic, functional, clonal, and epigenetic coherence to CD4+ TCONV cells in the orchestration of memory formation (Figs. 1 to 3). However, the definition of Treg cell differentiation stages was based on a limited array of surface proteins, prompting us to validate this strategy with an unbiased approach using single-cell transcriptome analysis. We combined this analysis with cellular indexing of surface epitopes (CITE-seq) to authenticate the expression of surface markers that demarcate T cell differentiation stages, aligning them with those identified in the transcriptomes for unequivocal verification. To this end, we isolated CD4+ T cells and fluorescence-activated cell sorted (FACSorted) Treg cells from four and seven healthy donors respectively, subjecting the cells to CITE-seq and TCR analysis (Figs. 4 and 5). To uncover the transcriptomic interrelationships among individual cells and quantify their similarities, we used Uniform Manifold Approximation and Projection (UMAP) for dimension reduction (Fig. 4) (51). Using oligonucleotide-barcoded antibodies, we implemented manual gating to segregate CD4+ T cell and Treg cell subsets, adhering to the strategy delineated in Fig. 1 and figs. S1 and S2. Subsequently, we compared the transcriptomes of each subset. In addition, we harnessed CITE-seq data to project the manually defined subsets (fig. S2) onto two data-driven methodologies: unbiased clustering (Fig. 4A) and lineage tracing (Fig. 4C), with a particular emphasis on identifying T helper cell (TH cell)–polarized effector subsets (52). This approach unveiled distinct clusters and differentiation patterns along two pseudo-time trajectories for Treg and TCONV cells (Fig. 5A), facilitating cellular delineation. Lineages were confidently annotated by propagating cellular labels from two publicly available reference datasets (53, 54). Figure 5A depicts the distribution of labels along lineage trajectories and their positions within the UMAP, highlighting distinct transcriptomic features unique to each gated subset and external reference phenotype. The expression levels of canonical bona fide Treg cell markers, along with automatically annotated reference labels, underscore the accurate assignment of Treg cell subsets and clusters (Fig. 4, D to F, FOXP3, IL-2RA, and IL-7R, respectively). Unsupervised clustering revealed eight unique cell clusters (fig. S4A). These comprised CD4+ TCONV cells, primarily located in clusters 5 (naïve) and 8 (effector), and Treg cells dispersed across clusters 5 (few TregN cells), 6 (majority TregN cells), and 2 (TregCM cells), as well as clusters 1, 3, and 4 (TregEM cells). Notably, cluster 7 exhibited high expression of cycling genes, indicative of proliferating Treg cells. Figure 5C delineates the differential expression of master transcription factors and functional genes within Treg and TCONV cell lineages along pseudo-time trajectories and furthermore summarizes the surface expression levels of CITE-seq antibodies for both lineages, along with genes that underwent differential regulation along pseudo-time trajectories and thus confirming these established markers for Treg cell subset characterization. The heatmap in Fig. 4G summarizes the most differentially expressed genes between gated subsets, while fig. S3B details those between unbiasedly identified clusters. RNA transcripts associated with effector function (e.g., CD40LG, KLRB1, CCL5, GZMA, and GZMK) were up-regulated in CD4+ TCM and TEM cell subsets (Fig. 4G). Similarly, TN, TSCM, TregN, and TregNLM cell subsets collectively exhibited robust expression of transcripts associated with migration to secondary lymphoid organs, survival, and homeostatic proliferation, features that were less pronounced in memory-type T cell subsets.

Fig. 4. CITE-seq between CD4+ T cells and Treg cells.

(A to F) UMAP representation of FACSorted Treg cells (n = 7) and CD4+ T cells (n = 4) based on top 1000 highly variable genes. (A) Color-encoded cluster membership of unsupervised graph-based clustering using the Louvain community detection method. (B) Distribution of manually gated subsets is shown. Subsets were defined by applying the flow cytometry gating strategy on CITE-seq surface epitope expression levels. (C) Distribution of external cell type identities. Single cells were automatically labeled leveraging two publicly available datasets from bulk RNA-sequenced FACSorted established TH and Treg cell populations. TFH, T follicular helper. (D to F) Gene expression of canonical Treg cell markers. Blue to red rainbow color scale denotes low to high expression. (G) Heatmap representation of average expression of top differentially expressed genes between manually gated subsets.

Fig. 5. Trajectory analysis confirms T cell differentiation into distinct Treg and TH cell lineages.

(A) UMAP representation based on top 1000 highly variable genes (same as Fig. 4) color-encoded for pseudo-time and UMAP-embedded, higher-dimensional differentiation trajectories for Treg cell (left) and TCONV cell (right) lineages. (B) Distribution of cell identity labels and clonal expansion along the lineage trajectories. Pseudo-time values were quantized and binned into uniform cellular distribution for each lineage. Bottom: Average pseudo-time values per bin. Top: Distribution charts for labels assigned by graph-based clustering, manual gating annotation, and reference-based annotation with the same color-coded as in Fig. 4 (A to C). Cells that could not be assigned a label (none) or a clonotype are shown in light-gray. (C) Heatmaps depict the expression of surface epitope markers (top) and expression of differentially regulated genes over pseudo-time and between lineages (bottom). Expression levels were smoothed over pseudo-time trajectories depicted in (A) and (B), respectively. Hallmark differentiation genes, Treg cell functional genes, and master transcription factors are highlighted. ADT, antibody derived tag.

RNA transcripts of genes associated with Treg cell functional attributes [e.g., transforming growth factor–B1 (TGFB1), tumor necrosis factor receptor superfamily member 1B (TNFRSF1B), granulysin (GNLY), aryl hydrocarbon receptor (AHR), GZMA, GZMK, cytotoxic T lymphocyte antigen 4 (CTLA4), IL-2RA, FAS, ectonucleotide triphosphate diphosphohydrolase 1 (ENTPD1), and perforin 1 (PRF1)] exhibited up-regulation in more-differentiated effector-type Treg cells (Fig. 5C). In addition, TregCM and TregEM cell subsets were distinctly characterized by the up-regulation of human leukocyte antigen (HLA) class II–type mRNA (Fig. 4G). Moreover, master transcription factors such as FOXP3 and IKAROS family zinc finger 2 (IKZF2; Helios) were observed to be up-regulated in Treg cells (Fig. 5C). We observed minor induction in transcripts of other TH cell–lineage transcription factors, including GATA3 and RAR-related orphan receptor C (RORC), but not TBX21 (Tbet).

Each CD4+ and Treg cell subset was found at unique coordinates within the UMAP and lineage trajectory, respectively, emphasizing distinct, albeit slightly overlapping, transcriptomic identities of the manually defined subsets (Fig. 4B). While there is some overlap between the subsets, we observed a significant enrichment of Treg cell EM phenotypes and bulk CD4 polarizing differentiation as pseudo-time progresses. This coincided with polarizing TH cell differentiation within CD4+ T cells. TregNLM cells were enriched in early differentiation stages (i.e., naïve and TregCM cells), as well as in cluster 7, marking the end of the trajectory and representing highly proliferating cells (as indicated by mitosis-related genes; Fig. 5C), thus suggesting a recycling of cells between the EM compartments from both ends of the trajectory. This observation is further supported by highly expanded TCR clonotypes along the trajectory, with the highest frequency within the TregEM cell compartment and polarized/differentiated CD4+ TCONV cells (Fig. 5B and fig. S3I). We further explored shared clonalities within Treg cell subsets and other T cell subsets. Although no clonal overlap was detected between Treg and TCONV cell subsets (fig. S3I), the limited number of cells analyzed warrants caution. Thus, while our findings suggest no overlap, this possibility remains inconclusive.

In conclusion, we validated the surface markers defining T cell differentiation stages using single-cell transcriptome analysis, thereby confirming their identities. Through an unbiased single-cell transcriptome analysis combined with CITE-seq, we identify distinct Treg cell subsets and clusters, unveiling differentiation patterns along pseudo-time progression. Expression profiling of canonical Treg cell markers accurately assigns subsets, highlighting functional and transcriptional differences between Treg and TCONV cell lineages. Moreover, the enrichment of Treg cell EM phenotypes and CD4 polarizing differentiation over pseudo-time suggests a cellular recycling process between compartments.

Treg cell memory differentiation subsets exhibit distinct temporal expansion potential in vitro

We set out to investigate the in vitro differentiation pattern of Treg cell subsets following activation and focused on its relevance for adoptive Treg cell therapy. To this end, we studied the expansion potential, phenotype changes, production of proinflammatory cytokines under different conditions, suppression ability, and lineage stability. TregNLM cells share the same protein marker combination as TSCM cells (i.e., CD45RA, CCR7, and CD95), which are known to have enhanced proliferation and self-renewal capacity compared to other TCONV cell subsets. Another objective was therefore to determine whether TregNLM cells exhibit similar characteristics in their proliferation and self-renewal capacity. To this end, we sorted different Treg cell subsets using fluorescence-activated cell sorting (FACS) (Fig. 6, A and B) and expanded them using αCD3/CD28 microbead stimulation every 2 weeks, which simulated antigen-mediated activation (Fig. 6C). In the initial phase of expansion, i.e., the first 3 weeks, TregN and bulk Treg cell subset showed the greatest expansion, followed by TregNLM, TregCM, and TregEM cells (Fig. 6D). From week 3, TregCM cells exhibited the highest proliferation capacity, followed by TregNLM and bulk Treg cells (including all subsets), both of which showed similar levels of expansion (Fig. 6E). Notably, the isolated TregEM cells exhibited the lowest level of expansion (Fig. 6D). Between weeks 4 and 7, all Treg cell subsets reached an increase in their expansion rate, albeit the rates of increase were lower for TregN and TregEM cells (Fig. 6E). To analyze the functional role of TCR stimulation for their expansion potential (fig. S4A), we discontinued bead stimulation at week 2 in one partition of the cultures and concurrently observed that the expansion potential between the subsets was comparable to repetitive αCD3/CD28 stimulation, albeit considerably reduced, indicating the requirement of Treg cell for continuous CD3/CD28 engagement for sustained expansion (fig. S4B).

Fig. 6. FACS strategy for isolating Treg cell subsets—expansion capacity, assessment of Treg cell subset–specific phenotypic, and functional characteristics during expansion.

(A) FACS strategy for isolating Treg cell subsets. The lymphocyte population was gated for singlets, and CD4+ T cells and bulk Treg cells were further defined as CD25highCD127low. From bulk Treg cells, TregCM and TregEM cells were defined as CD45RA−CCR7+ and CD45RA−CCR7−, respectively. CD45RO+ cells were excluded to further identify and isolate CD95+CCR7low TregNLM cells and CD95−CCR7+ TregN cells. All Treg cell populations sorted for expansion are highlighted. FSC-W, Forward Scatter Width; SSC-H, Side Scatter Height; SSC-W, Side Scatter Width. (B) Treg cell population purity was assessed after FACS. (C) FACSorted Treg cell populations were stimulated with anti-CD3/CD28–coated microbeads on day 1 and weeks 1, 2, 4, and 6 (indicated by dots) in the presence of rhIL-2 and rapamycin. Phenotypic and proinflammatory cytokine analysis was performed after 3, 5, and 7 weeks of expansion (indicated by triangles). (D and E) Fold expansion of Treg cell subset–derived cells from a close-up view of weeks 2 to 3 (D) and weeks 2 to 7 (E). n = 6. Results are presented as means ± SEM. (F to I) Examination of the Treg cell subsets phenotypically and functionally during in vitro expansion. (F) Change of Treg cell lineage markers CD25 and FoxP3 of Treg cell subset–derived cells over the entire expansion period of 7 weeks. n = 6. Results are presented as means ± SEM. (G) Changes in the initial naïve/memory phenotype during 7 weeks in culture using their CD45RA and CCR7 expression profile. (H and I) Proinflammatory cytokine production of cultured Treg cell–derived subsets. Treg cell subset–derived cells were stimulated at indicated time points with PMA/ionomycin for 6 hours and stained for IFN-γ (H) and IL-2 (I). For (F) to (I), n = 6. Results are presented as means ± SEM. For (G), statistical significance for differences between cell subsets was determined by two-way ANOVA with Tukey multiple comparison correction. *P ≤ 0.05; **P ≤ 0.01; ****P ≤ 0.0001.

Treg cell subsets exhibited distinct characteristics during in vitro expansion, with TregEM cells displaying loss of CD25/FoxP3, gain of a CM phenotype, increased cytokine production, reduced suppression, and TSDR methylation

We next examined the Treg cell subsets phenotypically and functionally at different time points after in vitro expansion. Throughout a long-term expansion over 7 weeks, all Treg cell subsets, except TregEM cells, stably retained their initial CD25+FoxP3+ expression (Fig. 6F). Approximately 40% of TregEM cells did not express CD25 and FoxP3 at week 7 after expansion despite mammalian target of rapamycin inhibition, which is commonly used to stabilize the Treg cell phenotype (Fig. 6F). Over a period of 7 weeks, we monitored possible changes in the initial subset phenotype based on the expression of CD45RA and CCR7. During the first 3 weeks of expansion, TregNLM (90%) and TregN (80%) cells lost most of their initial naïve CD45RA+CCR7+ phenotype and acquired a CM-like phenotype (CD45RA−CCR7+) that did not change further during weeks 3 to 7 (Fig. 6G). A significant proportion of the TregEM cell subset (about 40%) underwent a phenotype transition to a CM phenotype characterized by CD45RA−CCR7+ that progressed further during weeks 3 to 7 (>60%) (Fig. 6G). TregCM cells largely kept their phenotype and only partially lost it during the first few weeks, with more than 85% retaining a stable CM phenotype at week 7 (Fig. 6G). When we discontinued bead stimulation at week 2 to assess the functional role of TCR stimulation in maintaining the phenotypes of different subsets of Treg cells, we found similar profiles across all subsets (fig. S4, C and D).

It is essential that clinical-grade Treg cell products do only produce marginal amounts of proinflammatory cytokines such as IFN-γ and IL-2 upon TCR-specific stimulation. This absence of cytokine production is considered a central release criterion for ensuring the stability and safety of the Treg cell product in our good manufacturing practice (GMP) facility. Following 3 weeks of expansion, the proportion of IFN-γ and IL-2 producers within Treg cell subsets was minimal (below 2.0% for IFN-γ and 1.21% for IL-2 in TregCM cells), except for TregEM cells, which exhibited 6.19 and 3.10% IFN-γ and IL-2 producers, respectively (Fig. 6, H and I). Thus, the frequency of IFN-γ–and IL-2–producing Treg cells was negligible in all Treg cell subsets until the third week of expansion—the typical expansion time for generation of therapeutic Treg cell products—with the exception of TregEM cells. Although the levels were still relatively low, the IFN-γ production after stimulation was significantly increased in TregEM cells. From weeks 3 to 7, the proportion of IFN-γ producers increased in all memory subsets but remained below 10% for each subset except TregEM cells (Fig. 6H). There was a notable rise in the proportion of IFN-γ producers specifically in the TregEM cell subset, which correlated with the loss of their initial phenotype and decreased expression of FoxP3/CD25 (Fig. 6, F and G). Following the discontinuation of bead stimulation in the second week, we observed consistent trends in the proinflammatory cytokine profiles across different Treg cell subsets, albeit with TregEM cells exhibiting slightly increased cytokine production (fig. S4, E and F).

Considering the plasticity inherent in CD4+ T cells, we expanded our analysis to assess the expression levels of key transcription factors. By examining Tbet, GATA3, and RORγt alongside the canonical marker FoxP3 within expanded Treg cell subsets across various time points during culture, we were able to assess the transcription factors critical for the development, identity, and function of TH1 (Tbet), TH2 (GATA3), and TH17 (RORγt) cell types, and Treg cells (FoxP3) (Fig. 7, A and B). This enabled us to investigate the potential plasticity and propensity of Treg cells to dedifferentiate under the specific culture conditions being studied. Although effector cytokine production is observed in Treg cell subsets during the course of expansion (Fig. 6, H and I), neither subset expressed Tbet or RORγt (Fig. 7, A and B). Notably, by week 3, GATA3 expression becomes evident in both the TregEM and TregCM cell expanded subsets, characterized by the presence of GATA3 single-positive cells distinct from FoxP3 expression (Fig. 7A). GATA3 expression progressively increases over time in the TregEM cell expanded subset (Fig. 7B). In contrast, the TregCM cell subset exhibited a lower level of GATA3 expression, with most cells being double-positive for FoxP3 (Fig. 7A). Minimal GATA3 expression was detected in the TregN and TregNLM cell expanded subsets throughout the entire culture period. However, bulk-expanded Treg cells showed a slight increase in GATA3 expression over time (Fig. 7B). Considering the shift in transcription factors from FoxP3 toward particularly high GATA3 expression observed in the TregEM and TregCM cell subsets, we expanded our analysis to include measurements of other TH cell type–specific cytokines. Specifically, we aimed to elucidate how these changes might lead to effector cytokine production specific to each TH cell type. To achieve this, we measured the production of IL-4, IL-5, IL-13, and granulocyte-macrophage colony-stimulating factor (GM-CSF) (associated with TH2 cells), IL-17 (TH17 cells), and IL-10, IFN-γ, and tumor necrosis factor–α (TNFα) (TH1 cells) in supernatants of Treg cell subsets upon αCD3/28 stimulation (Fig. 7C). We confirmed high IFN-γ and TNFα production primarily in the TregEM cell subset, despite being negative for the transcription factor Tbet (Fig. 7, B and C). IL-17 was detected in both TregEM and TregCM cell cultures, despite both being negative for RORγt (Fig. 7, B and C). Notably, effector TH2 cell type cytokines, including IL-4, IL-5, IL-13, and GM-CSF, were prominently detected in the supernatants of the TregEM cell subset cultures, with their levels continually increasing during the culture period. In addition, we observed high levels of IL-10 production, particularly by the TregEM cell subset (Fig. 7C).

Fig. 7. Dynamics of transcription factor expression and Treg cell subset–specific functional and epigenetic characteristics during expansion.

(A) Expression patterns of transcription factors GATA3, RORγt, and Tbet versus FoxP3 in expanded bulk Treg cells and Treg cell subsets. FACSorted Treg cell populations were cultured according to strategy presented in Fig. 6 (A and B). Cells were stained intracellularly for CD3 and respective transcription factors. Illustrated are density plots of flow cytometry data of one representative donor at day 35 after culture initiation. (B) Bulk Treg cells and Treg cell subsets were analyzed for the transcription factors GATA3, RORγt, Tbet, and FoxP3 after 21, 28, 35, and 42 days of expansion. Quantified data from three donors are presented. Results are presented as connected data points (days) and mean. (C) Cytokine production profiles of expanded Treg cell subsets: IL-4, IL-5, IL-10, IL-13, IL-17, GM-CSF, IFN-γ, and TNFα cytokine production was evaluated in supernatants derived from bulk Treg cell–and Treg cell subset–derived cultures at indicated time points (days 21, 28, and 35) upon 24 hours of αCD3/28 stimulation (Meso Scale Diagnostics assay). n = 3. (D and E) Suppression capacity of expanded Treg cell subset–derived cells. In vitro expanded Treg cell subset–derived populations were cocultured with freshly isolated autologous CFSE+CD3+ TRESP cells and stimulated with αCD3/CD28-coated microbeads for 96 hours. Proliferation of CD4+ and CD8+ TRESP cells was analyzed by CFSE dilution. Percentage suppression of CD4+ and CD8+ TRESP cell proliferation at different TRESP:Treg cell ratios after 3 weeks of expansion. n = 6. Results are presented as means ± SEM. (F) Expanded Treg cell subset–derived cells were analyzed by bisulfite amplicon sequencing of the TSDR to define the percentage of TSDR demethylation (weeks 3, 5, and 7). n = 6. Results are presented as means ± SEM. For (F) to (H), statistical significance for differences between cell subsets was determined by two-way ANOVA with Tukey multiple comparison correction. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.

Next, we assessed the suppressive capacities of the Treg cell subsets after in vitro expansion for 3 weeks. Bulk Treg (including all subsets), TregCM, TregNLM, and TregN cells cocultured with autologous TRESP cells at a 1:1 ratio showed a great ability to suppress 80 to 90% of the proliferation of both CD4+ and CD8+ TRESP cells. TregEM cells, however, showed a lower level of suppression, with only 57.99% suppression of CD4+ TRESP cell proliferation and 48.86% suppression of CD8+ TRESP cell proliferation (Fig. 7D). When the TRESP:Treg cell ratio was increased to 10:1, TregNLM cells demonstrated the highest level of suppression of TRESP cell proliferation among all Treg cell subsets, followed by TregN, TregCM, and bulk Treg cells (Fig. 7E).

We further examined the epigenetic stability of the TSDR after 3, 5 and 7 weeks of expansion (Fig. 7F). At week 3, bulk Treg TregNLM, and TregN cells demonstrated the strongest TSDR demethylation, while TregCM cells showed a 20% lower demethylation in comparison. In contrast, only 13% of the TregEM cell TSDR was demethylated at this time point. At weeks 5 and 7, the differences in TSDR demethylation between the Treg cell subsets remained consistent with the findings from week 3, and all subsets displayed progressive loss of TSDR demethylation. The less-differentiated Treg cell subsets exhibited the highest degree of lineage stability, while TregEM cells demonstrated complete methylation of the TSDR after 7 weeks of culture (Fig. 7F). We found similar profiles across subsets when we discontinued bead stimulation at week 2 to assess the role of TCR stimulation in maintaining TSDR demethylation of different Treg cell subsets (fig. S4G).

Together, we conclude that the TregEM cell subset displays an unstable and plastic phenotype resulting in a pronounced loss of Treg cell functionality toward inflammatory-type cells during expansion. While the underlying molecular reasons for this remain to be determined, we observed an increase in β-galactosidase activity in the TregEM cell subset ex vivo (fig. S6), indicating that TregEM cells display a higher degree of cellular senescence, which has been correlated to an enhanced proinflammatory phenotype (55).

Improved Treg cell products by prior depletion of EM-like Treg cells

We depleted the EM-like Treg cell subset from bulk Treg cells before manufacturing to test the effect of removing this unstable subset on the functionality of the resulting Treg cell product. We used FACS to remove EM cells (CCR7−CD45RA−) from the initial bulk Treg cell population and then compared these cultures with donor-matched bulk Treg cells (Fig. 8A). Over 21 days, we observed a modest increase in the expansion rate of TregEM cell–depleted cell products (Treg − EM) compared to bulk Treg cells that retained the EM population (Fig. 8B). The percentage of FoxP3/CD25-positive cells in the final product at day 21 showed no difference on average (Fig. 8C). However, the bulk Treg cell products were more heterogeneous in terms of their final phenotype, frequencies of FoxP3/CD25, and TSDR demethylation (Fig. 8, C and E). Depletion of the TregEM cell population reduced the overall heterogeneity between Treg cell products from different donors. Further, we aimed to clarify whether depleting the TregEM cell subset from the bulk Treg cell population might result in reduced production of effector cytokines. To this end, we assessed the production of effector cytokines in the supernatants of both bulk Treg cells and TregEM cell compartment–depleted Treg cell–derived cultures (Treg − EM) upon αCD3/28 stimulation (Fig. 8, F to H). We found considerably high cytokine production (IL-4, IL-5, IL-13, GM-CSF, IL-17, IL-10, IFN-γ, and TNFα) primarily in the supernatants of bulk Treg cells and reduced levels in the Treg − EM cultures (Fig. 8, F to H). Thus, depletion of the TregEM cell subset from bulk Treg cells allowed the generation of stable Treg cell products from different donors without major effects on yield.

Fig. 8. Improved Treg cell products by prior depletion of EM-like Treg cells.

(A) Analyses of the effect of depleting the EM-like Treg cell subset before manufacturing the Treg cell product. We used a FACS to deplete EM cells (CD45RA−CCR7−) from the initial Treg cell population (Treg − EM). n = 6. Illustrated are density plots of flow cytometry data derived from the FACS of one representative donor. (B) Fold expansion of bulk Treg cell–and Treg − EM–derived cells of days 7 to 21. n = 6. (C) Treg cell lineage markers CD25 and FoxP3 of bulk Treg cell–and Treg − EM–derived cells at day 21. n = 6. Black lines in violin plots show the median. (D) Memory phenotypes of bulk Treg cell–and Treg − EM–derived cells at day 21. Black lines in violin plots show the median. n = 6. (E) Expanded bulk Treg cell–and Treg − EM–derived cells were analyzed to define the percentage of TSDR demethylation at day 21. Black lines in violin plots show the median. n = 6. (F to H) Impact of TregEM cell subset depletion on effector cytokine production. IL-4, IL-5, IL-13, GM-CSF (F), IL-17, IL-10 (G), IFN-γ, and TNFα (H) cytokine production was evaluated in supernatants derived from bulk Treg cells and Treg − EM at day 21 upon 24 hours of αCD3/28 stimulation. n = 3. (I and J) To evaluate the propensity of Treg cells to dedifferentiate into TH17 cells, a TH17 cell differentiation assay was performed. Treg cells were isolated and either left as bulk Treg cells or depleted of TregEM cells. Treg and Treg − EM cells were expanded for 14 days and then exposed to inflammatory conditions for another 7 days prior. Cells were restimulated with PMA/ionomycin and stained intracellularly for IL-17A. (I) Representative flow cytometry density plots and quantification (J) of IL-17A production of bulk Treg cell (left) and Treg − EM cultures, IL-6Rα knockout (KO), gp130 KO and wild-type (WT) Treg cells, respectively. n = 5. Statistical significance for differences between cells was determined by paired t test, two-tailed. *P ≤ 0.05.

Last, we investigated the propensity of Treg cells to dedifferentiate into TH17 cells in a highly inflammatory environment using a TH17 cell differentiation protocol adapted on the basis of Koenen et al. (56). To test the influence of IL-6 as a proposed key driver of dedifferentiation, we generated a targeted gene knockout to disrupt the IL-6 receptor complex (IL-6Rα or gp130) in bulk Treg cells and TregEM cell compartment–depleted (comprising TregN, TregNLM, and TregCM cells) Treg cell–derived cultures (fig. S7, A to C). Whereas over 4% of wild-type bulk Treg cells produced IL-17 after IL-6–containing inflammatory challenge, IL-6Rα or gp130 knockout bulk Treg cells expressed only about 2% of IL-17 (Fig. 8I). We found that our strategy of depleting the TregEM cell compartment before culture effectively prevented IL-17 production in unstimulated and IL-6–stimulated cultures after in vitro induction of TH17 cells to an extent where no difference between wild-type and edited Treg cells was visible (Fig. 8, I and J). Therefore, depletion of TregEM cells may alleviate the need for genetic manipulation to prevent Treg cell destabilization.

Together, depletion of the EM-like Treg cell subset from bulk Treg cells before manufacturing enhances the stability and homogeneity of resulting Treg cell products, which led to a modest increase in expansion rates without altering the percentage of FoxP3/CD25-positive cells and notably reduced the heterogeneity in phenotype and TSDR demethylation among Treg cell products from different donors. Furthermore, removal of the TregEM cell subset resulted in decreased production of effector cytokines upon stimulation. Our findings indicate that depleting the TregEM cell population may prevent dedifferentiation into inflammatory-type cells, even in highly inflammatory environments, obviating the need for genetic manipulation to maintain stability. Overall, our strategy offers a promising approach for generating stable and functional Treg cell products for therapeutic applications.

DISCUSSION

Therapeutic applications of Treg cells in inflammatory diseases, autoimmunity, transplantation, and gene therapies are promising, and recent advancements in CAR development and genome editing have been used to improve their specificity and functionality. However, several challenges remain before they can be brought into patients, including donor variability, manufacturing failures, and the need for prolonged Treg cell expansion periods. Furthermore, there are still important safety issues, such as obtaining the optimal Treg cell subset and the need to eliminate impurities while preserving Treg cell integrity. The central purpose of our study was to investigate the characteristics of Treg cell subsets with respect to differentiation stages of the in vitro stimulated Treg cell products that will eventually be infused into patients in clinical trials. We examined Treg cell subset derivation, fate, epigenomic stability, plasticity, transcriptome, TCR repertoires, and function for clinical grade manufacturing. Each subset displayed unique functional features, epigenomic lineage stability within the FoxP3 locus, surface marker expression, transcriptome, and TCR diversity. Using single-cell transcriptome analysis and CITE-seq, we validated the coherence of memory formation between CD4+ T cells and Treg cells, confirming surface markers defining T cell differentiation stages, identifying distinct Treg cell subsets and clusters, revealing differentiation patterns along pseudo-time progression, and highlighting functional and transcriptional differences between Treg cell and TCONV cell lineages. Earlier-differentiated memory Treg cell populations suppressed the proliferation of autologous TRESP cells more effectively than EM-like Treg cells. These earlier-differentiated memory Treg cell populations also exhibited the highest proliferative capacity, regardless of the duration and frequency of TCR stimulation. They maintained a stable phenotype after activation and showed high demethylation of the TSDR over time. In contrast, EM-like Treg cell–derived products showed lower TSDR demethylation and diminished immunosuppressive function while producing TH1 and TH17 cell type effector cytokines, despite lacking Tbet and RORγt, respectively. Although maintaining FoxP3 expression, these EM-like Treg cell products displayed elevated levels of GATA3 correlated with augmented production of TH2 cell–like cytokine. By depleting the EM-like Treg cell subset before manufacturing, we successfully produced stable Treg cell products with consistent properties, characterized, for example, by reduced conversion to TH17 cells and thus increased stability against inflammatory triggers.

The success of Treg cell therapy is determined by multiple cell- and patient-dependent factors. The origin of the T cell subset from which the final T cell product is derived may be one of the determining factors for the outcome, as it contributes to cell expansion after infusion, suppressive function, lineage stability, and long-term persistence. However, although the phenotype, such as FoxP3/CD25 positivity of the final Treg cell product, is often described, the Treg cell subsets are rarely studied in terms of their functional and epigenetic properties. However, the functional dissection of the infusion product is a prerequisite for a better understanding and correlation between T cell product and therapeutic outcome. As functional testing remains challenging (49), surrogate release criteria such as epigenetic integrity (TSDR demethylation) or the absence of cytokine production are considered central to ensuring the stability and authenticity of Treg cell products. Furthermore, it must be excluded that the cell product contains potentially contaminating TEFF cells or unstable Treg cell subsets.

Our findings reveal dynamic changes in transcription factor expression profiles, particularly high GATA3 expression in the TregEM cell subset, indicating a shift from a regulatory to an inflammatory phenotype. Despite lacking Tbet and RORγt expression, the TregEM cell subset showed increased production of inflammatory cytokines, including IL-10, IFN-γ, TNFα, and TH2 cell type cytokines, suggesting a loss of Treg cell functionality and acquisition of effector characteristics. Notably, IL-10 production was prominently detected in the TregEM cell subset, suggesting a regulatory-effector hybrid phenotype. These results highlight the plasticity of Treg cell subsets and their potential to transition toward inflammatory-type cells during expansion. The observed plasticity of Treg cell subsets, especially the transition toward an inflammatory phenotype in the EM subset, raises important safety implications. The dynamic changes in transcription factor expression and cytokine production profiles underscore the functional diversity and adaptability of Treg cells in response to inflammatory triggers. Further investigation into the molecular mechanisms governing Treg cell plasticity, especially during in vitro GMP-grade production processes, is imperative. The regulatory-effector hybrid phenotype exhibited by the TregEM cell subset underscores the intricate nature of immune regulation, highlighting the necessity for precision in therapeutic interventions designed to modulate Treg cell plasticity in pathological contexts.

Immunological memory is essential for TEFF cells to respond efficiently to pathogens upon secondary infection. While the benefits of memory are evident in proinflammatory cells, the significance of memory in Treg cells is less clear. Memory Treg cells are thought to reduce tissue damage during heightened proinflammatory responses and support fetal tolerance during pregnancy. Studies confirm the persistence of antigen-specific Treg cells with potent immunosuppressive properties even after antigen elimination (57–59). However, unique markers are lacking, which hinders the study of memory Treg cells. Understanding their function in health and disease requires comprehensive transcriptional profiling and detailed studies of epigenetic and metabolic signatures.

Using a diverse array of methods, including CITE-seq alongside canonical surface marker staining in flow cytometry, we investigated T cell differentiation within the Treg cell lineage. Through this comprehensive approach, we delineated distinct T cell differentiation stages within the Treg cell lineage, facilitating the consistent separation of naïve, effector, and memory Treg cell subsets. Notably, our findings reveal a robust correlation and coherence in memory formation between Treg cell subsets and their CD4+ TCONV cell counterparts, hinting at potential similarities in their differentiation pathways. In addition, the enrichment of Treg cell EM phenotypes and CD4 polarizing differentiation over pseudo-time suggests dynamic cellular interactions and recycling between compartments. Our results support the presence of a TSCM cell–like population within the conventional CD4+ T cell population, similar to what was originally described for CD8+ T cells (27, 32, 60–62). In addition, we found a previously unidentified population among the CD4+ Treg cell subsets, which we termed TregNLM cells. We consider advantageous properties in terms of stability and suppressive capacity to be attractive for immunotherapeutic applications. These TregNLM cells had slightly higher clonality than TregN cells, suggesting their affiliation and identity within the memory T cell pool, whereas the more-differentiated Treg cell subsets (TregCM and TregEM cells) had the least diverse repertoire, possibly due to clonal expansion after antigen encounter, similar to conventional memory T cells. We found a progressive loss of DNA methylation in the heterochromatic parts of the genome with differentiation of TregN cell to memory subsets, as reported for TCONV cells (50), supporting a linear differentiation from TregN to TregNLM to TregCM to TregEM cells. Our findings demonstrated that TregN cells also gradually declined with age with a simultaneous accumulation of TregEM cells (63, 64). In general, our results indicated that memory-like Treg cell subsets (TregCM and TregEM cells) increased with advancing age, similar to conventional memory CD4+ T cells. We further observed increasing expression of FoxP3 and CD25 with increasing differentiation of Treg cell memory subsets, indicating recent activation leading to Treg cell differentiation. However, despite the initial high FoxP3 expression in all subsets, TSDR demethylation decreased in TregEM cells as early as week 3 along with FoxP3 protein expression, indicating a loss of Treg cell stability and suppressive properties during expansion. One possible contributor to the observed functional instability of TregEM cells could be a progressed state of cellular senescence, indicated by increased activity of β-galactosidase (65) and their reduced proliferation capacity. On the basis of the observation of reduced level of DNA methylation in the heterochromatic parts of the genome, accumulation of proliferation episodes might be the driver for senescence acquisition (50, 66). This aligns with the emergence of exTreg, known for their pathogenic role in autoimmune diseases, characterized by IFN-γ production.

We therefore suggest depleting TregEM cells before Treg cells manufacturing to increase the stability and safety of the Treg cell product. Our findings demonstrated that depleting the TregEM cell subset before manufacturing results in a more homogeneous and stable Treg cell product. The expansion of TregEM cell–depleted cells was slightly increased, suggesting that the presence of TregEM cells might hinder the overall expansion potential of Treg cells during in vitro culture. However, the percentage of FoxP3/CD25-positive cells in the final product was not significantly affected, indicating that TregEM cell depletion does not compromise the maintenance of the regulatory phenotype. The enhanced stability of TregEM cell–depleted cell products appears to be particularly important as it ensures consistent and reliable outcomes across different donors. Heterogeneity in Treg cell products from various donors can lead to variable therapeutic responses, which may limit the efficacy of Treg cell–based therapies. Previous studies have shown that under certain conditions, Treg cells can lose their suppressive function and acquire proinflammatory properties, such as TH17 cell–like characteristics. We found that a sorting strategy for CCR7+ Treg cells, which effectively depleted TregEM cells before in vitro TH17 cell induction culture, successfully prevented dedifferentiation of Treg cells. This result indicates that the presence of TregEM cells may contribute to the dedifferentiation of Treg cells under inflammatory conditions and that their removal may preserve the regulatory function of Treg cells even in a proinflammatory environment.

First clinical trials are being launched that exclusively expand TregN cell–containing (and TregNLM cells) CD45RA+ Treg cells for genetic engineering purposes (40, 67), based on their reported stability advantage (68–70). However, a Treg cell subset with the highest expansion capacity, i.e., TregCM cells, is missing in these trials. This subset showed lower TSDR demethylation than TregN and TregNLM cells after expansion (69). However, achieving sufficient numbers of TregN cells through a manufacturing process suitable for clinical applications may not be feasible for all patients, especially the elderly. While targeting TregN/NLM cells could benefit younger patients, broader strategies are essential to ensure efficacy across diverse patient populations, particularly in posttransplantation settings, and among older patients or those with compromised immune systems.

The expansion rate of the TregEM cell subset appeared to be low, although not negligible. Furthermore, it should be cautioned not to extend the expansion rate beyond 5 weeks, as this could lead to functional and phenotypic instability. In vivo expansion of bulk Treg cell products in the patient’s body can only be speculated about, and in vivo follow-up studies are needed to gain further insights.

Overall, our study addresses the challenges in identifying optimal Treg cell subsets for stable and potent suppressive cells, as impurities and phenotypic instability hinder the development of edited Treg cell products and require precautions during manipulations. We examined the characteristics of Treg cells ex vivo and after expansion revealing that earlier-differentiated memory Treg cell populations outperformed late-differentiated EM-like Treg cells in terms of suppressive capacity, higher proliferative capacity, stable phenotypes, and strong demethylation of the TSDR over time. This highlights the importance of considering Treg cell subset composition before expansion in adoptive Treg cell immunotherapy to maintain lineage stability in the final cell product. Next, development of a GMP protocol using the proposed cell separation and clinical studies is needed to further unravel the relevance of our finding for adoptive Treg cell therapy.

MATERIALS AND METHODS

Blood sampling and PBMC isolation

Venous blood samples were collected from 53 healthy donors (30 females and 23 males, 24 to 82 years), who had given their written informed consent. The study was approved by the Charité–Universitätsmedizin Berlin Ethics Committee. PBMCs were isolated using Biocoll (Biochrom) gradient centrifugation.

FACSorting and expansion of Treg cell subsets

Freshly isolated PBMCs were enriched for CD4+ T cells via positive selection by magnetically activated cell separation (CD4 MicroBeads, Miltenyi Biotec) and stained with monoclonal antibodies [CD4 (SK3) and CD25 (2A3), both from BD Biosciences; CD127 (R34.34), from Beckman Coulter; CD45RA (HI100), CCR7 (G043H7), CD45RO (UCHL1), CD62L (DREG-56), and CD95 (DX2), all from BioLegend unless stated] for subsequent Treg cell subset isolation by FACSorting. For specific experiments, the TregEM cell compartment has been depleted from bulk Treg cells by FACSorting to yield CD4+CD25+CD127− Treg cells and CD4+CD25+CD127−CCR7+ Treg − TregEM cell subsets. FACSorted Treg cell subsets were suspended in X-Vivo15 (Lonza) medium supplemented with 10% fetal bovine serum (FBS) (PAA), penicillin (100 U/ml), streptomycin (100 μg/ml) (both Biochrom), recombinant human IL-2 (rhIL-2) (500 U/ml; Proleukin S, Novartis), and 100 μM rapamycin (RAPAMUNE, Pfizer) and placed into humidified incubators at 37°C and 5% CO2. On day 1, Treg cell expansion beads (Miltenyi Biotec), particles loaded with CD3/CD28 antibodies, were added at a bead-to-cell ratio of 4:1. On days 7 and 14, the cells were restimulated at a bead-to-cell ratio of 1:1 and the medium containing all abovementioned supplements. In addition, cells were split and medium changed when on the basis of cell density and/or medium color change.

Phenotype and cytokine profile analysis

Ex vivo isolated PBMC and expanded Treg cell subsets were analyzed for their phenotype and proinflammatory cytokine profile. For detecting intracellular cytokine production, cells were PMA (10 ng/ml) and ionomycin (1 μg/ml) for 4 hours, after which brefeldin A (4 μg/ml; all from Sigma-Aldrich) was added for further 2 hours. To define memory subsets, T cells were extracellularly labeled for LIVE/DEAD Fixable Blue Dead Cell Stain (Invitrogen); stained for surface markers CD25 (2A3, BD Biosciences), CD45RA (HI100), CCR7 (G043H7), CD45RO (UCHL1), CD62L (DREG-56), CD31 (WM59), and CD134 (Ber-ACT35); permeabilized/fixed with Foxp3/Transcription Factor Staining Buffer Set (eBioscience); and stained intracellularly for CD3 (OKT3), CD4 (SK3), CD8 (RPA-T8), CD95 (DX2), FoxP3 (259D/C7, BD Biosciences), Ki-67 (Ki-67), IFN-γ (4S.B3), and IL-2 (MQ1-17H12) (all from BioLegend unless stated). Fluorescence minus one (FMO) controls are shown in fig. S1. FMO controls to set precise gating thresholds by staining cells with all fluorophores except one, enabling accurate discrimination of true-positive signals from background noise and spectral overlap. In specific experiments, bulk Treg cell and expanded Treg cell subsets were analyzed for the transcription factors GATA3 (REA174, Miltenyi Biotec), Tbet (4B10, BioLegend), and RORγt (Q21 559, BD Biosciences). FMO controls are shown in (fig. S5).

GATA3 staining

Ex vivo isolated PBMC-derived bulk CD3+ T cells were extracellularly labeled for LIVE/DEAD Fixable Blue Dead Cell Stain (Invitrogen), stained for the surface marker CRTH2 (CD294, REA598 Miltenyi Biotec), permeabilized/fixed with Foxp3/Transcription Factor Staining Buffer Set (eBioscience), and stained intracellularly for CD3 (OKT3, BioLegend) and GATA3 (REA174, Miltenyi Biotec). Data were acquired and analyzed on a LSR II Fortessa flow cytometer and using FlowJo version 10 software (BD Biosciences, Tree Star) (fig. S5).

Tbet staining

Ex vivo isolated PBMC-derived bulk CD3+ T cells were extracellularly labeled for LIVE/DEAD Fixable Blue Dead Cell Stain (Invitrogen), permeabilized/fixed with Foxp3/Transcription Factor Staining Buffer Set (eBioscience), and stained intracellularly for CD3 (OKT3, BioLegend) and Tbet (4B10, BioLegend). Data were acquired and analyzed on a LSR II Fortessa flow cytometer and using FlowJo version 10 software (BD Biosciences, Tree Star) (fig. S5).

RORγt staining

Isolation and characterization of CD4+ TN cells for TH17 cell polarization in PBMCs. Freshly isolated PBMCs were enriched for CD4+ T cells via positive selection by magnetically activated cell separation (CD4 MicroBeads, Miltenyi Biotec). The following day, the cells were stained with monoclonal antibodies [CD3 (UCHT1) CD4 (OKT4), CCR7 (G043H7), CD25 (M-A251), and CD127 (A019D5), all from BioLegend unless stated, and CD45RA (2H4), from Beckman Coulter] for subsequent CD4+ TN cell FACsorting with on an FACSAria II (BD Biosciences) as Treg cell (CD127−CD25hi)–depleted CD3+CD4+CD45RA+CCR7+.

Cell culture and TH17 cell polarization

Sorted CD4+ TN cells were resuspended in serum-free and phenol red–free TexMACS GMP medium (Miltenyi Biotec) supplemented with penicillin (100 U/ml) and streptomycin (Thermo Fisher Scientific) in a 96-well plate (2 × 105 cells per well in 200 μl). TH17 cell polarization cultures were treated with rhIL-6 (30 ng/ml; R&D Systems), human IL-1b (20 ng/ml; Miltenyi Biotec), rhTGF-β1 (10 ng/ml; R&D Systems), rhIL-23 (30 ng/ml; R&D Systems), anti–IL-4 (2.5 μg/ml; clone 8D4-8, BioLegend), and anti–IFN-γ (1 μg/ml; clone 25718, R&D Systems). Control wells were treated with anti–IL-6 (10 μg/ml; clone 6708, R&D Systems), anti-TGF-β (10 μg/ml; clone 1D11, R&D Systems), anti–IL-4 (2.5 μg/ml), and anti–IFN-γ (1 μg/ml). Cells were activated using anti-CD2/CD3/CD28 activation beads from the T Cell Activation/Expansion Kit (Miltenyi Biotec) at a ratio of 1 bead:2 cells following the manufacturer’s instructions. The cell cultures were incubated at 37°C and 5% CO2. On day 4 following activation, 80 μl of supernatant was removed and replaced with 100 μl of fresh polarization medium. On day 7 following activation, the cell cultures were extracellularly stained with the Zombie UV Fixable Viability Kit (BioLegend) and then stained intracellularly for RORγt using the eBioscience FoxP3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific) following the manufacturer’s instructions. Cells were fixed, washed in permeabilization buffer, and then stained with RORγt (Q21 559, BD Biosciences). Data were acquired and analyzed on a LSR II Fortessa flow cytometer and using FlowJo version 10 software (BD Biosciences, Tree Star) (fig. S5).

Senescence-associated β-galactosidase assay

Freshly isolated PBMCs were used for senescence-associated β-galactosidase activity assessment following the protocol from Debacq-Chainiaux et al. (71). Briefly, lysosomal alkalinization was induced by pretreating subconfluent cells with 100 nM bafilomycin A1 [Sigma-Aldrich, dissolved in dimethyl sulfoxide (Sigma-Aldrich)] for 1 hour in fresh cell culture medium [RPMI 1640 (Gibco) with 10% FBS (Biowest), 50 μM β-mercaptoethanol (Sigma-Aldrich), 25 mM Hepes (Gibco), and 1 mM sodium pyruvate (Gibco)] at 37°C and 5% CO2. Then, 33 μM 5-dodecanoylaminofluorescein di-β-d-galactopyranoside [C12FDG (Invitrogen) dissolved in dimethyl sulfoxide] was added to the cell culture medium, followed by 2 hours of incubation. The cells were harvested and washed three times for ~30 s per wash with 2 ml of phosphate-buffered saline (Gibco) at room temperature. Afterward, cells were stained with monoclonal antibodies [CD3 Alexa Fluor 700 (UCHT1), CD4 APC/Fire (RPA-T4), CD8 BV510 (SK1), CD25 phycoerythrin (M-A251), CD127 APC (A019D5), CD197(CCR7), and BV785 (G04H7), all from BioLegend, and CD45Ra phycoerythrin/Cy7 (2H4) (Beckman Coulter)] for subsequent Treg cell subset characterization by flow cytometry using the Cytek Aurora spectral analyzer.

TRESP cell proliferation suppression assay

For ex vivo FACSorted Treg cell subsets, autologous CD4+CD25− TEFF cells were used both as TRESP cells and negative control to ensure Treg cell integrity and functionality. For expanded Treg cell subsets, freshly isolated and enriched autologous CD3 T cells (Human T cell Enrichment Cocktail, STEMCELL Technologies) were used. Responder cells were stained with 10 μM carboxyfluorescein diacetate succinimidyl ester (CFSE; Sigma-Aldrich) and cocultured at different TRESP:Treg cell ratios. Cells were stimulated with anti-CD3/28–coated microbeads (Treg Suppression Inspector, Miltenyi Biotec) at a cell-to-bead ratio of 1:1 and incubated at 37°C for 96 hours. Subsequently, cells were extracellularly labeled with CD3 (OKT3), CD4 (SK3), and CD8 (RPA-T8) (all from BioLegend) and LIVE/DEAD Fixable Blue Dead Cell Stain (Invitrogen). Data were acquired and analyzed on a LSR II Fortessa flow cytometer and using FlowJo version 10 software (Tree Star). Proliferation was assessed by CFSE dilution, and percentage suppression of proliferation was calculated by relating the percentage of proliferating TRESP cells in the presence and absence of Treg cells, respectively.

TSDR methylation analysis

Genomic DNA from ex vivo FACSorted or expanded T cell subsets was extracted (QIAamp DNA blood mini kit, QIAGEN). Each real-time polymerase chain reaction (PCR) reaction contained a minimum of 60 ng of bisulfite-treated (EpiTect, QIAGEN) genomic DNA or a respective amount of plasmid standard, 10 μl of FastStart universal probe master (Roche Diagnostics), lamda DNA [50 ng/ml; New England Biolabs (NEB)], methylation or nonmethylation-specific probe (5 pmol/ml), and methylation or nonmethylation-specific primers (30 pmol/ml). The samples were analyzed in triplicates on an Applied Biosystems (ABI) 7500 cycler. For some experiments, FOXP3-TSDR methylation analysis was performed by bisulfite amplicon sequencing as previously described (72). Briefly, genomic DNA was isolated using the Quick-DNA Microprep Kit (D3020, Zymo Research, Irvine, USA) according to the manufacturer’s protocol. Up to 200 ng of genomic DNA was bisulfite-converted using EZ-DNA methylation gold kit (D5005, Zymo Research, Irvine, USA). Subsequently PCRs were performed [10 μl of bisulfite-treated DNA, 2× KAPA HiFi Hotstart Uracil+ ReadyMix (KK2802, Kapa Biosystems, USA), 0.25 mM of each deoxynucleotide triphosphate (dNTP), using 0.3 pmol of primers (forward 1, 5′-ACACTCTTTCCCTACACGACGCTCTTCCGATCTTTTGGGGGTAGAGGATTTAGAGGG-3′; reverse 3, 5′-GACTGGAGTTCAGACGTGTGCTCTTCCGATCTCCACATCCACCAACACCCAT -3′)]. Amplicons were purified with QIAquick PCR Purification Kit (28106, QIAGEN, Germany), normalized to 20 ng/μl, and sequenced (2× 300–base pair paired-end). Reads were aligned and evaluated using the bismark package (73).

Quantitative real-time PCR

Total RNA of FACSorted Treg cell subsets was extracted using NucleoSpin RNA II Kit (Macherey-Nagel) following the manufacturer’s instructions. Up to 1 μg of RNA was used for cDNA synthesis according to the QuantiTect Reverse Transcription Kit (QIAGEN) manual. Human hypoxanthine-guanine phosphoribosyltransferase (hHPRT) was used for the normalization of expression levels. Quantitative real-time PCR analysis was performed using TaqMan Universal PCR Master Mix and ready-to-use detections system, containing FAM-TAMRA–labeled probes that were purchased from ABI. All expression levels were analyzed in duplicate using ABI Prism 7500 sequence detection system and associated software (all from ABI). All expression levels were calculated relative to hHPRT expression levels.

TCR next-generation sequencing

CDR3 sequencing was performed on the ImmunoSEQ platform at Adaptive Biotechnologies as previously described by Robins et al. (74) and Sherwood et al. (75). Sequences that did not match CDR3 sequences were removed from the analysis. For further analysis, the standard algorithm as developed by Adaptive Biotechnologies and previously described by Yousfi Monod et al. (76) was used. Genomic DNA was isolated using QIAamp DNA Mini Kit from QIAGEN according to the manufacturer’s instructions. Calculated clonality is an evenness index equal to 1 − normalized Shannon’s entropy. Values for clonality range from 0 to 1, with values near 1 representing samples with one or a few rearrangements (monoclonal or oligoclonal samples) dominating the observed repertoire. Clonality values near 0 represent more polyclonal samples.

CITE-seq analysis

Ex vivo FACSorted Treg cells and conventional CD4+ T cells were labeled with anti-human TotalSeq-C hashtag antibodies (BioLegend) allowing the pooling of samples, followed by labeling with anti-human TotalSeq-C antibodies (BioLegend) targeting a panel of extracellular proteins. Antibody and transcriptome libraries were prepared by using the Chromium Single-Cell 5′ Library & Gel Bead Kit and the Single-Cell 5′ Feature Barcode Library Kit version 1.1 (10x Genomics) following the manufacturer’s instructions. Qubit HS DNA assay kit (Life Technologies) was used for library quantification, and fragment sizes were obtained using the 2100 Bioanalyzer using the High-Sensitivity DNA Kit (Agilent). Sequencing was performed on a NextSeq 500 device (Illumina) using High Output v2 Kits (150 cycles) with the recommended sequencing conditions for 5′ gene expression and feature barcode libraries [read1, 26 nucleotides (nt); read2, 98 nt; index1, 8 nt; 1% PhiX spike-in]. Raw sequence reads were processed using cellranger-5.0.0, including the default detection of intact cells. Mkfastq and count were used in default parameter settings for demultiplexing and quantifying the gene expression. Refdata-cellranger-hg19-1.2.0 was used as reference. Raw unique molecular identifier (UMI) counts were further processed and analyzed using R 4.3.1 according to the osca workflow including normalization, filtering of low-quality cells, Louvain clustering, and UMAP dimensionality reduction. Differentially expressed genes between gated subsets and clusters, respectively, were identified using the findMarkers and multiMarkerStats functions. Manual gating of Treg cell and TCONV cell subsets was applied on all cells previously assigned to either Treg cell (clusters 1 to 7) or TCONV cell (clusters 5 and 8) lineage (fig. S3, A and B). External Treg cell and TH cell subset labels were automatically assigned using SingleR package with two reference datasets (53, 54, 77). Cells that were not annotated with the same T cell subset by both external references were considered ambiguous and thus not assigned. Trajectory analysis was conducted using slingshot and tradeSeq packages (78, 79). To visualize the distribution of cellular labels and clonotypes along the trajectories, pseudo-time of each lineage was quantized into uniform cellular distribution, i.e., bins containing each 100 cells. Heatmaps of surface epitopes and gene expression levels were generated using predictSmooth function on quantized pseudo-time values, while statistical inference was conducted using the negative binomial generalized additive models for each gene along the original pseudo-time values. Heatmaps were generated using ComplexHeatmap package (80).

Reduced representation bisulfite sequencing

Libraries for reduced representation bisulfite sequencing (RRBS) were prepared according to the protocol previously described (81) with minor modifications. Briefly, fresh-frozen cell pellets (30,000 to 130, 000 cells) were lysed by overnight incubation at 55°C after adding 16 μl of lysis buffer (10 mM tris-HCl and 5 mM EDTA) and 4 μl of proteinase K (1 mg/ml; Merck Millipore, Burlington, USA). Proteinase K was then inactivated by incubation with 2.8 μl of 8.14 mM Pefabloc SC (Merck Millipore) for 1 hour at room temperature. Then, samples were cut overnight at 37°C by adding 1 μl of HaeIII (50 U/μl; NEB, Ipswich, USA), 3 μl of 10× CutSmartbuffer (NEB), and 3.2 μl of nuclease-free water. Blunt-end DNA fragments were subjected to A-tailing with Klenow fragment exo- (50 U/μl; NEB) for 30 min at 37°C, followed by inactivation for 20 min at 75°C. Methylated sample–specific sequencing adaptors were ligated using T4 ligase (NEB), followed by bisulfite conversion with the EZ-DNA methylation gold kit (Zymo Research, Irvine, USA). Converted samples were PCR amplified for 15 to 18 cycles and then purified with AMPure XP beads (Beckham Coulter, Brea, USA). Next-generation sequencing libraries were then sequenced for ~30 to 70 million of 100–base pair single reads on a HiSeq 2500 platform (Illumina, San Diego, USA). Raw reads were trimmed with the Trim Galore! wrapper (www.bioinformatics.babraham.ac.uk/projects/trim_galore/) in RRBS mode. Reads were mapped to hg38 genome using the BWA wrapper methylCtools (v0.9.2) (82), and DNA methylation calling was performed as described in Durek et al. (50). Weighted average methylation was calculated and plotted across PMDs and HMDs obtained from TEM cell sample from Durek et al. (50).

Assessment of cytokines in supernatants (the Meso Scale Diagnostics method)

Culture supernatants from the Treg cell subsets expanded for 21, 28, and 35 days were collected, and cytokines were assessed using the human proinflammatory panel 1 kit (for IFN-γ, IL-4, IL-10, IL-13, and TNFα; Meso Scale Diagnostics) and human cytokine panel kit (for IL-5, IL-7, and GM-CSF; Meso Scale Diagnostics), following the manufacturer’s instructions. Measurements were performed at the Immunological Study Lab of the BIH Center for Regenerative Therapies on a Mesoscale Discovery platform. For each sample, the respective optical density values of the analyte concentration, calculated on the basis of a calibration curve, were obtained by subtracting the blank. The mean concentrations and SDs of at least duplicates and two dilutions of each supernatant were calculated. The measurements of the Meso Scale Diagnostics assay were performed and evaluated in accordance with the International Conference on Harmonization Good Clinical Practice Guideline under “Validation of analytical procedures.”

Gene editing of IL-6 signaling pathway

Knockout of IL-6Rα and gp130 was performed by electroporation of CRISPR-Cas9 ribonucleoprotein complexes (RNPs) as previously described (83, 84). Briefly, bulk sorted Treg cell or EM compartment–depleted (ex vivo FACS, CD4+CD25+CD127−CCR7+) Treg cell–derived cultures were expanded until day 7, and harvested and residual Treg cell expansion beads (Miltenyi Biotec) were depleted using a strong magnet stand (see the “FACSorting and expansion of Treg cell subsets” section). Single guide RNAs and recombinant Cas9 protein (IDT) were precomplexed (=RNPs) for 10 to 15 min at room temperature. Single guide RNA: IL6R-ex4-sg1: GGTGCGTCGCCAGTAGTGTC, SpCas9, cleavage site: IL-6R exon 4; gp130 ex4 sg7/sg8: AGAATAATCAACAGTGCATGAGG/GATCTGATGTAACCTTCCCAAGG, SpCas9, cleavage site: gp130 exon 4. The electroporation of Treg cell cultures with the RNPs was performed with the P3 primary cell Kit (Lonza) and nucleofection program “EH-115” on a Lonza 4D-Nucleofector. After electroporation, the cells were rescued in pre-warmed in X-Vivo15 (Lonza) medium supplemented with 10% FBS (PAA), rhIL-2 (500 U/ml; Proleukin S, Novartis), and 100 μM rapamycin (RAPAMUNE, Pfizer) and transferred back into humidified incubators at 37°C and 5% CO2. Treg cells were restimulated with Treg cell expansion beads (Miltenyi Biotec) after an overnight rest.

Analysis of signal transducers and activators of transcription 3 phosphorylation after IL-6 exposure

Treg cells were rested for 24 hours in cytokine-free Treg cell medium [X-Vivo15 (Lonza) supplemented with 10% FBS (PAA)]. A total of 1 × 106 Treg cells were stained with Live/Dead UV Fixable Dye (Thermo Fisher Scientific) according to the manufacturer’s recommendation. Then, Treg cells were resuspended either in prewarmed Treg cell medium containing IL-6 (100 ng/ml; Miltenyi Biotec) with or without soluble IL-6 receptor (10 ng/ml; STEMCELL Technologies) or in no cytokine for 30 min. Immediately at the end of incubation, Treg cells were fixed with prewarmed fixation buffer (BioLegend) at 37°C and incubated for 15 min at 37°C. Then, Treg cells were washed and permeabilization with ice-cold True-Phos permeabilization buffer (BioLegend) at −20°C for 60 min. After permeabilization and fixation, Treg cells were stained intracellularly with monoclonal antibodies for CD3 (UCHT1, Beckman Coulter) and phospho–signal transducers and activators of transcription 3 (13A3-1, BioLegend) in 1× phosphate-buffered saline. Data were acquired and analyzed on a LSR II Fortessa flow cytometer and using FlowJo version 10 software (BD Biosciences, Tree Star).

Treg cell–TH17 cell induction protocol

To evaluate the propensity of Treg cells to dedifferentiate into TH17 cells, a TH17 cell differentiation assay was performed using a protocol adapted from a previous study (56). Briefly, Treg cell cultures (see the “FACSorting and expansion of Treg cell subsets” section), either bulk Treg cell–derived or EM compartment-depleted (ex vivo FACS, CD4+CD25+CD127−CCR7+) Treg cell–derived (each wild-type and knockout of IL-6Rα or gp130), were exposed to irradiated allogeneic PBMCs on day 7 after sorting and activation in the presence of inflammatory cytokines IL-1β (100 ng/ml; Miltenyi Biotec), IL-2 (500 U/ml; Proleukin S, Novartis), and IL-6 (100 ng/ml; Miltenyi Biotec) over 7 days. As controls, Treg cells were expanded with anti-CD3/28 expansion beads (Miltenyi Biotec) and rhIL-2 (500 U/ml; Proleukin S, Novartis) containing X-Vivo15 (Lonza) medium supplemented with 10% FBS (PAA), penicillin (100 U/ml), and streptomycin (100 μg/ml) (both Biochrom). Then, bulk Treg cell–or EM compartment–depleted Treg cell cultures were restimulated with PMA/ionomycin for 6 hours (see the “Phenotype and cytokine profile analysis” section). To define IL-17 production, T cells were permeabilized/fixed with Foxp3/Transcription Factor Staining Buffer Set (eBioscience) and stained intracellularly for CD3 (OKT3, BioLegend) and IL-17 (eBio64CAP17, eBiosciences). Data were acquired and analyzed on a LSR II Fortessa flow cytometer and by using FlowJo version 10 software (BD Biosciences, Tree Star).

Acknowledgments

We would like to express our gratitude to A. Jurisch for contributions and technical assistance (Charité, Berlin, Germany; deceased). We thank A. Schulze and S. Lugo for generation of TSDR amplicon sequencing data and K. Vogt for technical assistance with the PCR-based TSDR analyses. We thank S. Bartosch for proofreading. We would like to acknowledge the technical support of the BIH Cytometry Core Facility. We thank all voluntary blood donors for donations.

Funding: The study was supported in parts by the German Federal Ministry of Education and Research (BIH Center for Regenerative Therapies, 13353, Berlin to H.-D.V., P.R., L.A., and M.S.-H.), a research grants by the Einstein Center for Regenerative Therapies (to L.A. and M.S.-H.), the state of Berlin and the “European Regional Development Fund” (ERDF 2014–2020, EFRE 1.8/11), Deutsches Rheuma-Forschungszentrum (to M.-F.M.), a Crossfield project fund of the BIH Research Focus Regenerative Medicine (to D.J.W., P.R., L.A., H.-D.V., and M.S.-H.), a BIH Research Platform Clinical Translational Sciences grant (to L.A. and M.S.-H.), the Leibniz Association (Leibniz Competition Collaborative Excellence Grant K59/2017 “EpImAge” to J.K.P.), and Deutsche Forschungsgemeinschaft (DFG), CRC1444, P02 to J.K.P. and H.-D.V. This project has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement no. 825392 (ReSHAPE, www.reshape-h2020.eu/) and by the European Research Council (ERC Starting grant “EpiTune,” no. 803992 to J.K.P.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author contributions: Writing—original draft: D.J.W., P.R., J.W., L.A., W.D., and M.S.-H. Conceptualization: D.J.W., P.R., T.V., H.-D.V., L.P., J.K.P., and M.S.-H. Investigation: D.J.W., G.G., T.V., L.A., C.C., A.M., W.D., D.L.W., M.S., M.-F.M., L.P., M.F.-S., F.Ha., S.L.-K., T.R., and S. Schu. Writing—review and editing: D.J.W., P.R., T.V., L.A., H.-D.V., D.L.W., J.K.P., P.D., S.L.-K., T.R., and M.S.-H. Methodology: D.J.W., P.R., R.R., T.V., A.S., C.C., L.A., A.M., H.-D.V., W.D., D.L.W., J.W., L.P., M.S.-H., J.K.P., M.F.-S., M.M., F.Ha., S.L.-K., T.R., and S. Schu. Resources: P.R., R.R., L.A., W.D., D.L.W., J.W., T.R., and M.S.-H. Funding acquisition: P.R., H.-D.V., J.W., M.-F.M., J.K.P., and M.S.-H. Data curation: D.J.W., G.G., M.-F.M., and P.D. Validation: D.J.W., P.R., A.M., H.-D.V., W.D., D.L.W., and M.S.-H. Software: A.S., S.S., P.D. Supervision: P.R., H.-D.V., D.L.W., J.K.P., T.R., and M.S.-H. Formal analysis: G.G., A.S., S.S., W.D., M.S.-H., M.Y., F.H., F.Ha., and P.D. Project administration: G.G., M.-F.M., J.K.P., and M.S.-H. Visualization: D.J.W., S.S., W.D., J.W., F.Ha., P.D., and M.S.-H.

Competing interests: D.J.W., H.-D.V., and M.S.-H. have a patent pending on “Enhancing the lineage stability of therapeutic Treg cell products to inflammatory triggers by selective depletion of the effector memory–like subset.” H.-D.V. is cofounder and CSO of CheckImmune GmbH. D.L.W., P.R., and H.-D.V. are cofounders of TCBalance Biopharmaceuticals GmbH. All other authors declare that they have no competing interests.

Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. There are no restrictions on the use of materials. Single-cell sequencing data are available under DOI https://doi.org/10.6084/m9.figshare.26294932.

Supplementary Materials

This PDF file includes:

Figs. S1 to S7
==== Refs
REFERENCES AND NOTES

1 S. Sakaguchi, T. Yamaguchi, T. Nomura, M. Ono, Regulatory T cells and immune tolerance. Cell 133 , 775–787 (2008).18510923
2 A. Roemhild, N. M. Otto, G. Moll, M. Abou-El-Enein, D. Kaiser, G. Bold, T. Schachtner, M. Choi, R. Oellinger, S. Landwehr-Kenzel, K. Juerchott, B. Sawitzki, C. Giesler, A. Sefrin, C. Beier, D. L. Wagner, S. Schlickeiser, M. Streitz, M. Schmueck-Henneresse, L. Amini, U. Stervbo, N. Babel, H.-D. Volk, P. Reinke, Regulatory T cells for minimising immune suppression in kidney transplantation: Phase I/IIa clinical trial. BMJ 371 , m3734 (2020).33087345
3 S. Landwehr-Kenzel, L. Mueller-Jensen, J. S. Kuehl, M. Abou-El-Enein, H. Hoffmann, S. Muench, D. Kaiser, A. Roemhild, H. von Bernuth, M. Voeller, M. Schmueck-Henneresse, B. Gruhn, U. Stervbo, N. Babel, H. D. Volk, P. Reinke, Adoptive transfer of ex vivo expanded regulatory T-cells improves immune cell engraftment and therapy-refractory chronic GvHD. Mol. Ther. 30 , 2298–2314 (2022).35240319
4 Y. H. Oo, S. Ackrill, R. Cole, L. Jenkins, P. Anderson, H. C. Jeffery, N. Jones, L. E. Jeffery, P. Lutz, R. E. Wawman, A. K. Athwal, J. Thompson, J. Gray, K. Guo, D. Barton, G. M. Hirschfield, T. Wong, P. Guest, D. H. Adams, Liver homing of clinical grade Tregs after therapeutic infusion in patients with autoimmune hepatitis. JHEP Rep. 1 , 286–296 (2019).32039380
5 L. Amini, J. Kaeda, E. Fritsche, A. Roemhild, D. Kaiser, P. Reinke, Clinical adoptive regulatory T cell therapy: State of the art, challenges, and prospective. Front. Cell Dev. Biol. 10 , 1081644 (2023).36794233
6 C. Voskens, D. Stoica, M. Rosenberg, F. Vitali, S. Zundler, M. Ganslmayer, H. Knott, M. Wiesinger, J. Wunder, M. Kummer, B. Siegmund, E. Schnoy, T. Rath, A. Hartmann, H. Hackstein, B. Schuler-Thurner, C. Berking, G. Schuler, R. Atreya, M. F. Neurath, Autologous regulatory T-cell transfer in refractory ulcerative colitis with concomitant primary sclerosing cholangitis. Gut 72 , 49–53 (2023).35428657
7 M. Miyara, Y. Yoshioka, A. Kitoh, T. Shima, K. Wing, A. Niwa, C. Parizot, C. Taflin, T. Heike, D. Valeyre, A. Mathian, T. Nakahata, T. Yamaguchi, T. Nomura, M. Ono, Z. Amoura, G. Gorochov, S. Sakaguchi, Functional delineation and differentiation dynamics of human CD4+ T cells expressing the FoxP3 transcription factor. Immunity 30 , 899–911 (2009).19464196
8 T. Saito, H. Nishikawa, H. Wada, Y. Nagano, D. Sugiyama, K. Atarashi, Y. Maeda, M. Hamaguchi, N. Ohkura, E. Sato, H. Nagase, J. Nishimura, H. Yamamoto, S. Takiguchi, T. Tanoue, W. Suda, H. Morita, M. Hattori, K. Honda, M. Mori, Y. Doki, S. Sakaguchi, Two FOXP3+CD4+ T cell subpopulations distinctly control the prognosis of colorectal cancers. Nat. Med. 22 , 679–684 (2016).27111280
9 C. Baecher-Allan, J. A. Brown, G. J. Freeman, D. A. Hafler, CD4+CD25high regulatory cells in human peripheral blood. J. Immunol. 167 , 1245–1253 (2001).11466340
10 H. Jonuleit, E. Schmitt, M. Stassen, A. Tuettenberg, J. Knop, A. H. Enk, Identification and functional characterization of human CD4+CD25+T cells with regulatory properties isolated from peripheral blood. J. Exp. Med. 193 , 1285–1294 (2001).11390435
11 M. K. Levings, R. Sangregorio, M. G. Roncarolo, Human Cd25+Cd4+ T regulatory cells suppress naive and memory T cell proliferation and can be expanded in vitro without loss of function. J. Exp. Med. 193 , 1295–1302 (2001).11390436
12 W. F. Ng, P. J. Duggan, F. Ponchel, G. Matarese, G. Lombardi, A. D. Edwards, J. D. Isaacs, R. I. Lechler, Human CD4+CD25+ cells: A naturally occurring population of regulatory T cells. Blood 98 , 2736–2744 (2001).11675346
13 L. S. Taams, M. Vukmanovic-Stejic, J. Smith, P. J. Dunne, J. M. Fletcher, F. J. Plunkett, S. B. Ebeling, G. Lombardi, M. H. Rustin, J. W. Bijlsma, F. P. Lafeber, M. Salmon, A. N. Akbar, Antigen-specific T cell suppression by human CD4+CD25+ regulatory T cells. Eur. J. Immunol. 32 , 1621–1630 (2002).12115645
14 D. Dieckmann, H. Plottner, S. Berchtold, T. Berger, G. Schuler, Ex vivo isolation and characterization of CD4+CD25+ T cells with regulatory properties from human blood. J. Exp. Med. 193 , 1303–1310 (2001).11390437
15 N. Seddiki, B. Santner-Nanan, J. Martinson, J. Zaunders, S. Sasson, A. Landay, M. Solomon, W. Selby, S. I. Alexander, R. Nanan, A. Kelleher, B. Fazekas de St, Expression of interleukin (IL)-2 and IL-7 receptors discriminates between human regulatory and activated T cells. J. Exp. Med. 203 , 1693–1700 (2006).16818676
16 M. A. Gavin, T. R. Torgerson, E. Houston, P. DeRoos, W. Y. Ho, A. Stray-Pedersen, E. L. Ocheltree, P. D. Greenberg, H. D. Ochs, A. Y. Rudensky, Single-cell analysis of normal and FOXP3-mutant human T cells: FOXP3 expression without regulatory T cell development. Proc. Natl. Acad. Sci. U.S.A. 103 , 6659–6664 (2006).16617117
17 M. R. Walker, D. J. Kasprowicz, V. H. Gersuk, A. Benard, M. Van Landeghen, J. H. Buckner, S. F. Ziegler, Induction of FoxP3 and acquisition of T regulatory activity by stimulated human CD4+CD25–T cells. J. Clin. Invest. 112 , 1437–1443 (2003).14597769
18 M. E. Morgan, J. H. van Bilsen, A. M. Bakker, B. Heemskerk, M. W. Schilham, F. C. Hartgers, B. G. Elferink, L. van der Zanden, R. R. de Vries, T. W. Huizinga, T. H. Ottenhoff, R. E. Toes, Expression of FOXP3 mRNA is not confined to CD4+CD25+ T regulatory cells in humans. Hum. Immunol. 66 , 13–20 (2005).15620457
19 J. K. Polansky, K. Kretschmer, J. Freyer, S. Floess, A. Garbe, U. Baron, S. Olek, A. Hamann, H. von Boehmer, J. Huehn, DNA methylation controls Foxp3 gene expression. Eur. J. Immunol. 38 , 1654–1663 (2008).18493985
20 X. Zhou, S. L. Bailey-Bucktrout, L. T. Jeker, C. Penaranda, M. Martinez-Llordella, M. Ashby, M. Nakayama, W. Rosenthal, J. A. Bluestone, Instability of the transcription factor Foxp3 leads to the generation of pathogenic memory T cells in vivo. Nat. Immunol. 10 , 1000–1007 (2009).19633673
21 E. A. Walter, P. D. Greenberg, M. J. Gilbert, R. J. Finch, K. S. Watanabe, E. D. Thomas, S. R. Riddell, Reconstitution of cellular immunity against cytomegalovirus in recipients of allogeneic bone marrow by transfer of T-cell clones from the donor. N. Engl. J. Med. 333 , 1038–1044 (1995).7675046
22 M. Inoue, G. E. Plautz, S. Shu, Treatment of intracranial tumors by systemic transfer of superantigen-activated tumor-draining lymph node T cells. Cancer Res. 56 , 4702–4708 (1996).8840987
23 Y. Honaker, N. Hubbard, Y. Xiang, L. Fisher, D. Hagin, K. Sommer, Y. Song, S. J. Yang, C. Lopez, T. Tappen, E. M. Dam, I. Khan, M. Hale, J. H. Buckner, A. M. Scharenberg, T. R. Torgerson, D. J. Rawlings, Gene editing to induce FOXP3 expression in human CD4+ T cells leads to a stable regulatory phenotype and function. Sci. Transl. Med. 12 , eaay6422 (2020).32493794
24 P. Henschel, S. Landwehr-Kenzel, N. Engels, A. Schienke, J. Kremer, T. Riet, N. Redel, K. Iordanidis, V. Saetzler, K. John, M. Heider, M. Hardtke-Wolenski, H. Wedemeyer, E. Jaeckel, F. Noyan, Supraphysiological FOXP3 expression in human CAR-Tregs results in improved stability, efficacy, and safety of CAR-Treg products for clinical application. J. Autoimmun. 138 , 103057 (2023).37224732
25 B. Fritzsching, N. Oberle, E. Pauly, R. Geffers, J. Buer, J. Poschl, P. Krammer, O. Linderkamp, E. Suri-Payer, Naive regulatory T cells: A novel subpopulation defined by resistance toward CD95L-mediated cell death. Blood 108 , 3371–3378 (2006).16868256
26 D. Valmori, A. Merlo, N. E. Souleimanian, C. S. Hesdorffer, M. Ayyoub, A peripheral circulating compartment of natural naive CD4 Tregs. J. Clin. Invest. 115 , 1953–1962 (2005).16007258
27 L. Gattinoni, E. Lugli, Y. Ji, Z. Pos, C. M. Paulos, M. F. Quigley, J. R. Almeida, E. Gostick, Z. Yu, C. Carpenito, E. Wang, D. C. Douek, D. A. Price, C. H. June, F. M. Marincola, M. Roederer, N. P. Restifo, A human memory T cell subset with stem cell-like properties. Nat. Med. 17 , 1290–1297 (2011).21926977
28 P. Graef, V. R. Buchholz, C. Stemberger, M. Flossdorf, L. Henkel, M. Schiemann, I. Drexler, T. Hofer, S. R. Riddell, D. H. Busch, Serial transfer of single-cell-derived immunocompetence reveals stemness of CD8+ central memory T cells. Immunity 41 , 116–126 (2014).25035956
29 F. Sallusto, J. Geginat, A. Lanzavecchia, Central memory and effector memory T cell subsets: Function, generation, and maintenance. Annu. Rev. Immunol. 22 , 745–763 (2004).15032595
30 F. Sallusto, D. Lenig, R. Forster, M. Lipp, A. Lanzavecchia, Two subsets of memory T lymphocytes with distinct homing potentials and effector functions. Nature 401 , 708–712 (1999).10537110
31 C. U. Louis, B. Savoldo, G. Dotti, M. Pule, E. Yvon, G. D. Myers, C. Rossig, H. V. Russell, O. Diouf, E. Liu, H. Liu, M. F. Wu, A. P. Gee, Z. Mei, C. M. Rooney, H. E. Heslop, M. K. Brenner, Antitumor activity and long-term fate of chimeric antigen receptor-positive T cells in patients with neuroblastoma. Blood 118 , 6050–6056 (2011).21984804
32 M. Schmueck-Henneresse, R. Sharaf, K. Vogt, B. J. Weist, S. Landwehr-Kenzel, H. Fuehrer, A. Jurisch, N. Babel, C. M. Rooney, P. Reinke, H. D. Volk, Peripheral blood-derived virus-specific memory stem T cells mature to functional effector memory subsets with self-renewal potency. J. Immunol. 194 , 5559–5567 (2015).25917088
33 L. Amini, T. Vollmer, D. J. Wendering, A. Jurisch, S. Landwehr-Kenzel, N. M. Otto, K. Jürchott, H.-D. Volk, P. Reinke, M. Schmueck-Henneresse, Comprehensive characterization of a next-generation antiviral T-cell product and feasibility for application in immunosuppressed transplant patients. Front. Immunol. 10 , 1148 (2019).31191530
34 N. J. Booth, A. J. McQuaid, T. Sobande, S. Kissane, E. Agius, S. E. Jackson, M. Salmon, F. Falciani, K. Yong, M. H. Rustin, A. N. Akbar, M. Vukmanovic-Stejic, Different proliferative potential and migratory characteristics of human CD4+ regulatory T cells that express either CD45RA or CD45RO. J. Immunol. 184 , 4317–4326 (2010).20231690
35 N. Seddiki, B. Santner-Nanan, S. G. Tangye, S. I. Alexander, M. Solomon, S. Lee, R. Nanan, B. Fazekas de Saint, Persistence of naive CD45RA+ regulatory T cells in adult life. Blood 107 , 2830–2838 (2006).16332974
36 Y. Luo, C. Xu, B. Wang, Q. Niu, X. Su, Y. Bai, S. Zhu, C. Zhao, Y. Sun, J. Wang, M. Liu, X. Sun, G. Song, H. Cui, X. Chen, H. Huang, H. Wang, M. Han, E. Jiang, L. Shi, X. Feng, Single-cell transcriptomic analysis reveals disparate effector differentiation pathways in human Treg compartment. Nat. Commun. 12 , 3913 (2021).34162888
37 K. S. Smigiel, E. Richards, S. Srivastava, K. R. Thomas, J. C. Dudda, K. D. Klonowski, D. J. Campbell, CCR7 provides localized access to IL-2 and defines homeostatically distinct regulatory T cell subsets. J. Exp. Med. 211 , 121–136 (2014).24378538
38 V. Tosello, K. Odunsi, N. E. Souleimanian, S. Lele, P. Shrikant, L. J. Old, D. Valmori, M. Ayyoub, Differential expression of CCR7 defines two distinct subsets of human memory CD4+CD25+ Tregs. Clin. Immunol. 126 , 291–302 (2008).18166500
39 L. Schober, D. Radnai, J. Spratte, A. Kisielewicz, E. Schmitt, K. Mahnke, H. Fluhr, L. Uhlmann, C. Sohn, A. Steinborn, The role of regulatory T cell (Treg) subsets in gestational diabetes mellitus. Clin. Exp. Immunol. 177 , 76–85 (2014).24547967
40 K. Schreeb, E. Culme-Seymour, E. Ridha, C. Dumont, G. Atkinson, B. Hsu, P. Reinke, Study design: Human leukocyte antigen class I molecule A*02-chimeric antigen receptor regulatory T cells in renal transplantation. Kidney Int. Rep. 7 , 1258–1267 (2022).35694562
41 H. Lei, L. Kuchenbecker, M. Streitz, B. Sawitzki, K. Vogt, S. Landwehr-Kenzel, J. Millward, K. Juelke, N. Babel, A. Neumann, P. Reinke, H. D. Volk, Human CD45RA− FoxP3hi memory-type regulatory T cells show distinct TCR repertoires with conventional T cells and play an important role in controlling early immune activation. Am. J. Transplant. 15 , 2625–2635 (2015).25988290
42 S. Tanaskovic, S. Fernandez, P. Price, S. Lee, M. A. French, CD31 (PECAM-1) is a marker of recent thymic emigrants among CD4+ T-cells, but not CD8+ T-cells or γδ T-cells, in HIV patients responding to ART. Immunol. Cell Biol. 88 , 321–327 (2010).20065992
43 J. H. van Dierendonck, J. H. Wijsman, R. Keijzer, C. J. van de Velde, C. J. Cornelisse, Cell-cycle-related staining patterns of anti-proliferating cell nuclear antigen monoclonal antibodies. Comparison with BrdUrd labeling and Ki-67 staining. Am. J. Pathol. 138 , 1165–1172 (1991).1673821
44 I. Gramaglia, A. D. Weinberg, M. Lemon, M. Croft, Ox-40 ligand: A potent costimulatory molecule for sustaining primary CD4 T cell responses. J. Immunol. 161 , 6510–6517 (1998).9862675
45 W. J. Grossman, J. W. Verbsky, W. Barchet, M. Colonna, J. P. Atkinson, T. J. Ley, Human T regulatory cells can use the perforin pathway to cause autologous target cell death. Immunity 21 , 589–601 (2004).15485635
46 B. Ogretmen, Y. A. Hannun, Biologically active sphingolipids in cancer pathogenesis and treatment. Nat. Rev. Cancer 4 , 604–616 (2004).15286740
47 C. Menne, J. P. Lauritsen, J. Dietrich, J. Kastrup, A. K. Wegener, P. S. Andersen, N. Odum, C. Geisler, T-cell receptor downregulation by ceramide-induced caspase activation and cleavage of the ζ chain. Scand. J. Immunol. 53 , 176–183 (2001).11169222
48 D. Voehringer, M. Koschella, H. Pircher, Lack of proliferative capacity of human effector and memory T cells expressing killer cell lectinlike receptor G1 (KLRG1). Blood 100 , 3698–3702 (2002).12393723
49 D. J. Wendering, L. Amini, S. Schlickeiser, P. Reinke, H.-D. Volk, M. Schmueck-Henneresse, The value of a rapid test of human regulatory T cell function needs to be revised. Front. Immunol. 10 , 150 (2019).30804944
50 P. Durek, K. Nordstrom, G. Gasparoni, A. Salhab, C. Kressler, M. de Almeida, K. Bassler, T. Ulas, F. Schmidt, J. Xiong, P. Glazar, F. Klironomos, A. Sinha, S. Kinkley, X. Yang, L. Arrigoni, A. D. Amirabad, F. B. Ardakani, L. Feuerbach, O. Gorka, P. Ebert, F. Muller, N. Li, S. Frischbutter, S. Schlickeiser, C. Cendon, S. Frohler, B. Felder, N. Gasparoni, C. D. Imbusch, B. Hutter, G. Zipprich, Y. Tauchmann, S. Reinke, G. Wassilew, U. Hoffmann, A. S. Richter, L. Sieverling, D. Consortium, H. D. Chang, U. Syrbe, U. Kalus, J. Eils, B. Brors, T. Manke, J. Ruland, T. Lengauer, N. Rajewsky, W. Chen, J. Dong, B. Sawitzki, H. R. Chung, P. Rosenstiel, M. H. Schulz, J. L. Schultze, A. Radbruch, J. Walter, A. Hamann, J. K. Polansky, Epigenomic profiling of human CD4+ T cells supports a linear differentiation model and highlights molecular regulators of memory development. Immunity 45 , 1148–1161 (2016).27851915
51 S. Seth, S. Mallik, T. Bhadra, Z. Zhao, Dimensionality reduction and louvain agglomerative hierarchical clustering for cluster-specified frequent biomarker discovery in single-cell sequencing data. Front. Genet. 13 , 828479 (2022).35198011
52 D. E. Wagner, A. M. Klein, Lineage tracing meets single-cell omics: Opportunities and challenges. Nat. Rev. Genet. 21 , 410–427 (2020).32235876
53 B. J. Schmiedel, D. Singh, A. Madrigal, A. G. Valdovino-Gonzalez, B. M. White, J. Zapardiel-Gonzalo, B. Ha, G. Altay, J. A. Greenbaum, G. McVicker, G. Seumois, A. Rao, M. Kronenberg, B. Peters, P. Vijayanand, Impact of genetic polymorphisms on human immune cell gene expression. Cell 175 , 1701–1715.e16 (2018).30449622
54 G. Monaco, B. Lee, W. Xu, S. Mustafah, Y. Y. Hwang, C. Carre, N. Burdin, L. Visan, M. Ceccarelli, M. Poidinger, A. Zippelius, J. Pedro de Magalhaes, A. Larbi, RNA-seq signatures normalized by mRNA abundance allow absolute deconvolution of human immune cell types. Cell Rep. 26 , 1627–1640.e7 (2019).30726743
55 J. Zhang, T. He, L. Xue, H. Guo, Senescent T cells: A potential biomarker and target for cancer therapy. EBioMedicine 68 , 103409 (2021).34049248
56 H. J. Koenen, R. L. Smeets, P. M. Vink, E. van Rijssen, A. M. Boots, I. Joosten, Human CD25highFoxp3pos regulatory T cells differentiate into IL-17–producing cells. Blood 112 , 2340–2352 (2008).18617638
57 M. D. Rosenblum, I. K. Gratz, J. S. Paw, K. Lee, A. Marshak-Rothstein, A. K. Abbas, Response to self antigen imprints regulatory memory in tissues. Nature 480 , 538–542 (2011).22121024
58 J. H. Rowe, J. M. Ertelt, L. Xin, S. S. Way, Pregnancy imprints regulatory memory that sustains anergy to fetal antigen. Nature 490 , 102–106 (2012).23023128
59 I. K. Gratz, M. D. Rosenblum, M. M. Maurano, J. S. Paw, H. A. Truong, A. Marshak-Rothstein, A. K. Abbas, Cutting edge: Self-antigen controls the balance between effector and regulatory T cells in peripheral tissues. J. Immunol. 192 , 1351–1355 (2014).24442443
60 E. Lugli, M. H. Dominguez, L. Gattinoni, P. K. Chattopadhyay, D. L. Bolton, K. Song, N. R. Klatt, J. M. Brenchley, M. Vaccari, E. Gostick, D. A. Price, T. A. Waldmann, N. P. Restifo, G. Franchini, M. Roederer, Superior T memory stem cell persistence supports long-lived T cell memory. J. Clin. Invest. 123 , 594–599 (2013).23281401
61 M. J. Buzon, H. Sun, C. Li, A. Shaw, K. Seiss, Z. Ouyang, E. Martin-Gayo, J. Leng, T. J. Henrich, J. Z. Li, F. Pereyra, R. Zurakowski, B. D. Walker, E. S. Rosenberg, X. G. Yu, M. Lichterfeld, HIV-1 persistence in CD4+ T cells with stem cell-like properties. Nat. Med. 20 , 139–142 (2014).24412925
62 N. R. Klatt, S. E. Bosinger, M. Peck, L. E. Richert-Spuhler, A. Heigele, J. P. Gile, N. Patel, J. Taaffe, B. Julg, D. Camerini, C. Torti, J. N. Martin, S. G. Deeks, E. Sinclair, F. M. Hecht, M. M. Lederman, M. Paiardini, F. Kirchhoff, J. M. Brenchley, P. W. Hunt, G. Silvestri, Limited HIV infection of central memory and stem cell memory CD4+ T cells is associated with lack of progression in viremic individuals. PLOS Pathog. 10 , e1004345 (2014).25167059
63 M. Beyer, J. L. Schultze, CD4+CD25highFOXP3+ regulatory T cells in peripheral blood are primarily of effector memory phenotype. J. Clin. Oncol. 25 , 2628–2630 (2007).17577047
64 M. Beyer, M. Kochanek, T. Giese, E. Endl, M. R. Weihrauch, P. A. Knolle, S. Classen, J. L. Schultze, In vivo peripheral expansion of naive CD4+CD25highFoxP3+ regulatory T cells in patients with multiple myeloma. Blood 107 , 3940–3949 (2006).16410445
65 Z. Guo, G. Wang, B. Wu, W. C. Chou, L. Cheng, C. Zhou, J. Lou, D. Wu, L. Su, J. Zheng, J. P. Ting, Y. Y. Wan, DCAF1 regulates Treg senescence via the ROS axis during immunological aging. J. Clin. Invest. 130 , 5893–5908 (2020).32730228
66 J. L. Endicott, P. A. Nolte, H. Shen, P. W. Laird, Cell division drives DNA methylation loss in late-replicating domains in primary human cells. Nat. Commun. 13 , 6659 (2022).36347867
67 E. Proics, M. David, M. Mojibian, M. Speck, N. Lounnas-Mourey, A. Govehovitch, W. Baghdadi, J. Desnouveaux, H. Bastian, L. Freschi, G. Privat, C. Pouzet, M. Grossi, P. Heimendinger, T. Abel, D. Fenard, M. K. Levings, F. Meyer, C. Dumont, Preclinical assessment of antigen-specific chimeric antigen receptor regulatory T cells for use in solid organ transplantation. Gene Ther. 30 , 309–322 (2022).35931871
68 R. Arroyo Hornero, G. J. Betts, B. Sawitzki, K. Vogt, P. N. Harden, K. J. Wood, CD45RA distinguishes CD4+CD25+CD127−/low TSDR demethylated regulatory T cell subpopulations with differential stability and susceptibility to tacrolimus-mediated inhibition of suppression. Transplantation 101 , 302–309 (2017).28118317
69 P. Hoffmann, R. Eder, T. J. Boeld, K. Doser, B. Piseshka, R. Andreesen, M. Edinger, Only the CD45RA+ subpopulation of CD4+CD25high T cells gives rise to homogeneous regulatory T-cell lines upon in vitro expansion. Blood 108 , 4260–4267 (2006).16917003
70 J. B. Canavan, C. Scotta, A. Vossenkamper, R. Goldberg, M. J. Elder, I. Shoval, E. Marks, E. Stolarczyk, J. W. Lo, N. Powell, H. Fazekasova, P. M. Irving, J. D. Sanderson, J. K. Howard, S. Yagel, B. Afzali, T. T. MacDonald, M. P. Hernandez-Fuentes, N. Y. Shpigel, G. Lombardi, G. M. Lord, Developing in vitro expanded CD45RA+ regulatory T cells as an adoptive cell therapy for Crohn’s disease. Gut 65 , 584–594 (2016).25715355
71 F. Debacq-Chainiaux, J. D. Erusalimsky, J. Campisi, O. Toussaint, Protocols to detect senescence-associated beta-galactosidase (SA-βgal) activity, a biomarker of senescent cells in culture and in vivo. Nat. Protoc. 4 , 1798–1806 (2009).20010931
72 C. Kressler, G. Gasparoni, K. Nordstrom, D. Hamo, A. Salhab, C. Dimitropoulos, S. Tierling, P. Reinke, H. D. Volk, J. Walter, A. Hamann, J. K. Polansky, Targeted de-methylation of the FOXP3-TSDR is sufficient to induce physiological FOXP3 expression but not a functional Treg phenotype. Front. Immunol. 11 , 609891 (2020).33488615
73 F. Krueger, S. R. Andrews, Bismark: A flexible aligner and methylation caller for bisulfite-seq applications. Bioinformatics 27 , 1571–1572 (2011).21493656
74 H. S. Robins, P. V. Campregher, S. K. Srivastava, A. Wacher, C. J. Turtle, O. Kahsai, S. R. Riddell, E. H. Warren, C. S. Carlson, Comprehensive assessment of T-cell receptor β-chain diversity in αβ T cells. Blood 114 , 4099–4107 (2009).19706884
75 A. M. Sherwood, C. Desmarais, R. J. Livingston, J. Andriesen, M. Haussler, C. S. Carlson, H. Robins, Deep sequencing of the human TCRγ and TCRβ repertoires suggests that TCRβ rearranges after αβ and γδ T cell commitment. Sci. Transl. Med. 3 , 90ra61 (2011).
76 M. Yousfi Monod, V. Giudicelli, D. Chaume, M. P. Lefranc, IMGT/JunctionAnalysis: The first tool for the analysis of the immunoglobulin and T cell receptor complex V-J and V-D-J JUNCTIONs. Bioinformatics 20 , i379-85 (2004).15262823
77 D. Aran, A. P. Looney, L. Liu, E. Wu, V. Fong, A. Hsu, S. Chak, R. P. Naikawadi, P. J. Wolters, A. R. Abate, A. J. Butte, M. Bhattacharya, Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat. Immunol. 20 , 163–172 (2019).30643263
78 K. Van den Berge, H. Roux de Bezieux, K. Street, W. Saelens, R. Cannoodt, Y. Saeys, S. Dudoit, L. Clement, Trajectory-based differential expression analysis for single-cell sequencing data. Nat. Commun. 11 , 1201 (2020).32139671
79 K. Street, D. Risso, R. B. Fletcher, D. Das, J. Ngai, N. Yosef, E. Purdom, S. Dudoit, Slingshot: Cell lineage and pseudotime inference for single-cell transcriptomics. BMC Genomics 19 , 477 (2018).29914354
80 Z. Gu, R. Eils, M. Schlesner, Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics 32 , 2847–2849 (2016).27207943
81 M. Luetke-Eversloh, Q. Hammer, P. Durek, K. Nordstrom, G. Gasparoni, M. Pink, A. Hamann, J. Walter, H. D. Chang, J. Dong, C. Romagnani, Human cytomegalovirus drives epigenetic imprinting of the IFNG locus in NKG2Chi natural killer cells. PLOS Pathog. 10 , e1004441 (2014).25329659
82 V. Hovestadt, D. T. Jones, S. Picelli, W. Wang, M. Kool, P. A. Northcott, M. Sultan, K. Stachurski, M. Ryzhova, H. J. Warnatz, M. Ralser, S. Brun, J. Bunt, N. Jager, K. Kleinheinz, S. Erkek, U. D. Weber, C. C. Bartholomae, C. von Kalle, C. Lawerenz, J. Eils, J. Koster, R. Versteeg, T. Milde, O. Witt, S. Schmidt, S. Wolf, T. Pietsch, S. Rutkowski, W. Scheurlen, M. D. Taylor, B. Brors, J. Felsberg, G. Reifenberger, A. Borkhardt, H. Lehrach, R. J. Wechsler-Reya, R. Eils, M. L. Yaspo, P. Landgraf, A. Korshunov, M. Zapatka, B. Radlwimmer, S. M. Pfister, P. Lichter, Decoding the regulatory landscape of medulloblastoma using DNA methylation sequencing. Nature 510 , 537–541 (2014).24847876
83 M. C. Gundry, L. Brunetti, A. Lin, A. E. Mayle, A. Kitano, D. Wagner, J. I. Hsu, K. A. Hoegenauer, C. M. Rooney, M. A. Goodell, D. Nakada, Highly efficient genome editing of murine and human hematopoietic progenitor cells by CRISPR/Cas9. Cell Rep. 17 , 1453–1461 (2016).27783956
84 L. Amini, D. L. Wagner, U. Rossler, G. Zarrinrad, L. F. Wagner, T. Vollmer, D. J. Wendering, U. Kornak, H. D. Volk, P. Reinke, M. Schmueck-Henneresse, CRISPR-Cas9-edited tacrolimus-resistant antiviral T cells for advanced adoptive immunotherapy in transplant recipients. Mol. Ther. 29 , 32–46 (2020).32956624
