Activation and Functional Specialization of Regulatory T Cells Lead to the Generation of Foxp3 Instability

This information is current as Zhongmei Zhang, Wei Zhang, Jie Guo, Qianchong Gu, of October 5, 2021. Xueping Zhu and Xuyu Zhou J Immunol 2017; 198:2612-2625; Prepublished online 22 February 2017; doi: 10.4049/jimmunol.1601409 http://www.jimmunol.org/content/198/7/2612 Downloaded from

Supplementary http://www.jimmunol.org/content/suppl/2017/02/21/jimmunol.160140 Material 9.DCSupplemental http://www.jimmunol.org/ References This article cites 53 articles, 20 of which you can access for free at: http://www.jimmunol.org/content/198/7/2612.full#ref-list-1

Why The JI? Submit online.

• Rapid Reviews! 30 days* from submission to initial decision

• No Triage! Every submission reviewed by practicing scientists by guest on October 5, 2021

• Fast Publication! 4 weeks from acceptance to publication

*average

Subscription Information about subscribing to The Journal of Immunology is online at: http://jimmunol.org/subscription Permissions Submit copyright permission requests at: http://www.aai.org/About/Publications/JI/copyright.html Email Alerts Receive free email-alerts when new articles cite this article. Sign up at: http://jimmunol.org/alerts

The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD 20852 Copyright © 2017 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

Activation and Functional Specialization of Regulatory T Cells Lead to the Generation of Foxp3 Instability

Zhongmei Zhang,*,† Wei Zhang,* Jie Guo,* Qianchong Gu,*,† Xueping Zhu,*,† and Xuyu Zhou*,†

Accumulating evidence suggests that Foxp3+ cells can downregulate the expression of Foxp3, but whether thymically derived regulatory T cells (tTregs; especially committed tTregs) are capable of downregulating Foxp3 expression and being reprog- rammed into other T effector cells remains controversial. Using a novel tTreg lineage-tracing mouse line, we were able to label epigenetically stable Foxp3+ cells derived from the thymus and demonstrate that mature tTregs are stable under homeostatic conditions. However, TCR engagement and sequential functional specialization of tTregs led to the generation of Foxp3 instability and reprogramming into the Th lineage. We further demonstrated that the signal switch from IL-2 to ICOS during Treg activation induced Treg instability and reprogramming. By using a dual lineage tracing model, we demonstrated that effector Downloaded from Tregs can revert to central Tregs, and this reversion is associated with increasing Foxp3 stability in vivo. The Journal of Immunology, 2017, 198: 2612–2625.

egulatory T cells (Tregs) are a specialized sublineage of the development of tTregs, serving as the alternative insurance to CD4+ T cells that have a crucial role in the control the organism to prevent self-autoreactivity (3). In contrast, in the of immunological self-tolerance and the maintenance of periphery, GALTs exposed to a multitude of commensal bacterial

R http://www.jimmunol.org/ immune homeostasis (1). Foxp3 is the lineage-specifying tran- and food Ags favor the development of pTregs to maintain im- scription factor of Tregs, whose deficiency (such as in scurfy mice mune tolerance at mucosal interfaces (4). To date, there is no and immune dysregulation, polyendocrinopathy, enteropathy, definitive marker to distinguish between these two Treg subsets. X-linked patients) results in a fatal autoimmune syndrome (2). In Thornton et al. (5) demonstrated that Helios is an Ikaros tran- contrast, ablation of Foxp3 from mature Tregs leads to loss of Treg scription factor family member exclusively expressed on tTregs, identity and switching to other types of Th cells. Thus, continued and Ag-specific Foxp3+ T cells induced in vivo by Ag feeding are expression of Foxp3 is indispensable for Treg lineage maintenance devoid of Helios expression. However, others have challenged this and immunosuppressive function (2). conclusion by using different Ag-specific pTreg-inducing systems

Foxp3 can be induced either in Treg precursors in the thymus or (6). Recently, neuropilin 1 (Nrp1) was proposed as a tTreg marker, by guest on October 5, 2021 + in naive CD4 T cells in the periphery by a combination of TCR, but conventional T cells (Tconvs) can transiently upregulate Nrp1 costimulation, and cytokine signals, generating thymically derived upon Ag stimulation (7, 8). Furthermore, slight Nrp1 expression is Tregs (tTregs) and peripherally derived Tregs (pTregs), respec- also detected in newly generated induced Tregs (iTregs) in vitro tively (3). In the thymus, self-agonist ligands preferentially cause and in vivo (7, 8). Recent studies from Rudensky and colleagues (9) demonstrated that the conserved noncoding sequence (CNS)1 *Key Laboratory of Pathogenic Microbiology and Immunology, Institute of Micro- of the Foxp3 gene, which serves as a major TGF-b sensor, is † biology, Chinese Academy of Sciences, Beijing 100101, China; and University of critical for the generation of induced pTregs but dispensable for Chinese Academy of Sciences, Beijing 100049, China tTreg development. Thus, CNS1 dependence could be a more ORCID: 0000-0002-6626-0898 (X. Zhou). reliable criterion for distinguishing tTregs and pTregs, especially Received for publication August 15, 2016. Accepted for publication January 27, 2017. under inflammatory conditions. This work was supported by National Natural Science Foundation of China Grants Although Tregs have shown great therapeutic potential for 31270959 and 31300750 and by National Key Basic Research and Development (973 curing various autoimmune diseases and preventing unwanted Program of China) Grant 2012CB917102. immune responses such as graft-versus-host disease in preclinical The raw RNA sequencing data presented in this article were submitted to the Gene animal models, precautions must be taken to ensure the identity and Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo) under accession number GSE93856. stability of Tregs for translation to the clinic. Our previous fate- Address correspondence and reprint requests to Prof. Xuyu Zhou, Key Laboratory of mapping studies demonstrated that a fraction of Tregs are not Pathogenic Microbiology and Immunology, Institute of Microbiology, Chinese Acad- stable under autoimmune inflammatory conditions. These previ- emy of Sciences, 1st West Beichen, Chaoyang, Beijing 100101, China. E-mail address: ously Foxp3-expressing (exFoxp3) T cells have an activated [email protected] memory T cell phenotype, produce inflammatory cytokines, and The online version of this article contains supplemental material. confer autoimmunity upon adoptive transfer, likely participating in Abbreviations used in this article: BAC, bacterial artificial chromosome; CNS, con- served noncoding sequence; DP, double-positive; exFoxp3, previously Foxp3- a feed-forward loop in promoting the pathogenesis of autoim- expressing; GC, germinal center; iTreg, induced Treg; KI, knock-in; KO, knockout; munity (10, 11). Stable expression of Foxp3 is largely controlled LN, lymph node; MFI, mean fluorescence intensity; Nrp1, neuropilin 1; PD1, pro- by the methylation status of the CNS2 (also known as the Treg- grammed cell death 1; pTreg, peripherally derived Treg; RA, retinoic acid; Rluc, Renilla luciferase; RNA-seq, RNA sequencing; Tconv, conventional T cell; TFH, specific demethylated region) region of the Foxp3 locus (12, 13). follicular helper T; TFR, T follicular regulatory; Tg, transgenic; Treg, regulatory Demethylation of CNS2 in Tregs is mediated by Tet-dependent T cell; tTreg, thymically derived Treg; YFP, yellow fluorescent protein. oxidization and favors the recruitment of multiple transcription Copyright Ó 2017 by The American Association of Immunologists, Inc. 0022-1767/17/$30.00 factors such as Cbfb, Runx1, STAT5, and Foxp3 itself to the www.jimmunol.org/cgi/doi/10.4049/jimmunol.1601409 The Journal of Immunology 2613

CNS2 to further ensure stable Foxp3 expression (14, 15). Re- phenotype to a central Treg phenotype, which was also associ- cently, studies have demonstrated that genetically deleting the ated with stable expression of Foxp3. Together, the signal CNS2 enhancer of Foxp3 results in a destabilized Treg lineage, driving Treg activation and functional specialization also and CNS2-deficient mice develop spontaneous autoimmunity and prompted the generation of Foxp3 instability, which indicates that chronic inflammation (16, 17). an optimal window of signal strength could be crucial for Treg Two different models have been proposed to explain the gen- activation in vivo. eration of exFoxp3 cells. The reprogramming model suggests that Tregs can become unstable due to the remethylation of the CNS2 region of the Foxp3 locus under inflammatory conditions (11). Materials and Methods Thus, Tregs lose their Foxp3 expression and are reprogrammed Mice into Th cells. In contrast, the second model suggests that Tregs are For Foxp3 DCNS1-Cre reporter mice, the hCre-2A-eqFP650-2A-Thy1.1 a stable lineage, but a minor population of uncommitted Foxp3+ cDNA fragment was inserted immediately downstream of the Foxp3 T cells (which either come from transient Foxp3 expression during ATG translational start site by homologous recombination into a 188-kb mouse BAC (from the C57BL/6 genome; clone RP23-143D8) carrying the activation of conventional peripheral T cells or from immature the intact Foxp3 gene. The CNS1 enhancer was deleted by another round tTregs that fail to demethylate CNS2 during thymic Treg devel- of homologous recombination. The modified BAC was purified using a opment) gives rise to exFoxp3 cells (12, 18, 19). These different QIAfilter plasmid maxi kit and microinjected into the pronuclei of a (B6 3 models are not mutually exclusive, and, moreover, accumulating B6D2) F1 mouse fertilized ovum to generate the Foxp3 DCNS1-Cre reporter mice. Three founder mice were generated and screened for evidence suggests that Tregs are a subset of T cells with great Thy1.1 and endogenous Foxp3 expression. For generation of T-bet re- heterogeneity (20). TCR engagement leads to Treg activation and porter mice, a Dre-2A–Thy1.1 fusion DNA fragment was inserted by Downloaded from the generation of effector Tregs, which can use various master homologous recombineering into the ATG translational start site of a transcription factors from other Th lineages for further functional 182-kb BAC clone (RP23-451G13) bearing the entire T-bet locus. The specialization. For example, T-bet–expressing Tregs selectively modified BAC was microinjected, and founders carrying the Dre-Thy1.1 transgene were screened by FACS analysis using peripheral blood cells. For survey the Th1 immune response by upregulating T-bet. Similarly, generation of Rosa26-Rox-Renilla luciferase (Rluc)-stop-Rox-tdTomato the Th2 transcription factor IFN regulatory factor 4 is required for BAC Tg mice, a fusion DNA fragment of CAG promoter-Rox-Rluc-pA- Tregs to control the Th2-mediated autoimmune response, and Rox-tdTomato-WPRE-BGH pA was inserted by homologous recombineer- other transcription factors such as STAT3 and Bcl6 are necessary ing between exons 1 and 2 of BAC clone RP24-85L15 containing the mouse http://www.jimmunol.org/ Rosa26 gene locus. Then, the modified BAC was microinjected. The Tg to shut down Th17-mediated spontaneous colitis and aberrant mice were screened by luciferase activity analysis using peripheral blood germinal center (GC) responses, respectively (20). Campbell and cells. All BAC Tg mice were backcrossed to C57BL/6 mice for at least 12 colleagues (21) divided Tregs into central Tregs and effector Tregs generations. Foxp3-GFP-hCre BAC Tg mice and Rosa26–yellow fluores- based on CD44 and CD62L. Interestingly, the homeostasis control cent protein (YFP) reporter mice have been previously described (10, 25). of effector Tregs is very different from central Tregs; effector The Rosa26-StopFL-P110* mice were purchased from The Jackson Lab- oratory (stock no. 012343). All mice were housed at the specific pathogen- Tregs do not rely on IL-2 but instead depend on continued sig- free animal facility at the Institute of Microbiology, Chinese Academy of naling through the costimulatory receptor ICOS. Whether effector Sciences. All experiments were conducted according to the guidelines of Tregs and functionally specialized Tregs, which frequently have the Institute of Microbiology, Chinese Academy of Sciences Institutional by guest on October 5, 2021 low expression of CD25, can maintain the stability of Foxp3 is Animal Care and Use Committee (permit no. PZIMCAS2011005). still largely unknown. However, several studies suggest that T-bet+ Tregs might lose Foxp3 expression and develop into IFN-g– Abs and flow cytometry producing Th1 effectors (22, 23). Cells were run on a FACSCalibur or FACS Fortessa (BD Biosciences) and To directly address the above issue, we generated two novel fate- analyzed by FlowJo. Fluorochrome-conjugated Abs directed against the tracing models in this study. The first animal model we developed following mouse Ags were used for flow cytometry analysis: CD4 (GK1.5), CD8 (53-6.7), Thy1.1 (OX-7), CD25 (PC61), CD103 (2E7), CD127 is a bacterial artificial chromosome (BAC) transgenic (Tg) mouse (A7R34), GITR (DTA-1), CD69 (H1.2F3), CD62L (MEL14), CD44 (IM7), expressing a Cre-2A–Thy1.1 fusion protein driven from the Helios (22F6), programmed cell death 1 (PD1; 29F.1A12), CD24 (M1/69), CNS1-deleted Foxp3 promoter, which allows the labeling of Qa2 (1-1-2), Foxp3 (FJK-16s), CTLA-4 (UC10-4B9), Ly-6A/E (D7), Itgb1 tTregs by Thy1.1 expression, as well as tracing their progeny. (HMb11), Ki67 (SolA15), Eos (ESB7C2), IFN-g (XMG1.2), CXCR3 (CXCR3-173), CXCR5 (2G8), CCR7 (4B12), and ICOS (C398.4A) were Characterization of these mice demonstrated that the Thy1.1 fu- + + + purchased from BD Pharmigen, BioLegend, or eBioscience. Biotinylated sion protein is mainly expressed by the Nrp1 Helios Foxp3 tTreg Nrp1 polyclonal Ab (BAF566) was purchased from R&D Systems. subset in the periphery and is considerably delayed in the thymus, CXCR5 staining was done with biotinylated anti-CXCR5 for 1 h at room starting only from the late development stage of tTregs. Thus, this temperature, followed by allophycocyanin-labeled or Brilliant Violet 650– new Foxp3 DCNS1-Cre strain provides a unique opportunity to labeled streptavidin (eBioscience and BD Pharmigen). For staining Foxp3 and YFP/tdTomato, cells were prefixed in 1% paraformaldehyde for 2 min genetically trace mature tTregs. We demonstrated that mature at room temperature, then fixed and permeabilized using a Foxp3 fixation/ tTregs are indeed stable under homeostatic conditions, but TCR permeabilization kit (eBioscience) and stained with eFluor 450–labeled activation and sequential functional specialization of tTregs led to Foxp3 (BioLegend). the generation of Foxp3 instability and reprogramming into Th1 and follicular helper T (TFH) cells. The signal switch from IL-2 Ab treatment to ICOS during Treg activation induced tTreg instability and Littermate mice were given 150 mg of the indicated Abs (JES6-1A12, anti– reprogramming. The second animal model we developed is a BAC IL-2; HK5.3, anti-ICOSL; both from Bio X Cell) or PBS by i.p. injection Tg T-bet reporter mouse expressing Dre and Thy1.1 in T-bet+ on days 0, 3, 6, 9, and 12 and sacrificed on day 14. cells. By taking advantage of the dual, noncompeting recombinase In vitro induction of iTregs systems of Cre-lox and Dre-rox, double genetic tracing of Foxp3 2 and T-bet were simultaneously enabled (24). We found that Highly pure naive CD62LhiCD4+GFP T cells from 6- to 8-wk-old Foxp3 sustained T-bet expression preferentially generated Foxp3 in- DCNS1-Cre/GFP knock-in (KI) mice were FACS sorted and cultured with plate-coated anti-CD3 (0.5 mg/ml) and anti-CD28 (1 mg/ml) in the pres- stability, whereas a large fraction of T-bet–experienced cells ence of TGF-b (2 ng/ml), IL-2 (200 U/ml), and retinoic acid (RA; 25 nM) downregulated T-bet expression and decreased the expression of for 4 d before Foxp3-GFP and Thy1.1 expression was analyzed with flow key hallmarks of Treg activation, switching from an effector cytometry. 2614 INSTABILITY OF tTregs IN VIVO

In vitro Treg culture and stability assay cleavage in Tg mice (26). The Cre cassette was inserted imme- Thy1.12YFP+ and Thy1.1+YFP+ Tregs were respectively FACS sorted diately downstream of the Foxp3 ATG translational start site of the from total CD4+ T cells purified from the spleen and lymph nodes (LN) of BAC clone (RP23-143D8) as previously described (27), and the 6- to 8-wk-old Foxp3 DCNS1-Cre-Rosa26-YFP mice. For experiments CNS1 region of Foxp3 was further removed from the BAC by testing the stability of Foxp3 expression after proliferation, cells were another round of homologous recombination (Fig. 1A). Thus, activated for 7–9 d with Dynabeads mouse T activator CD3/CD28 (Life equal expression of the Cre recombinase and Thy1.1 reporter was Technologies) using a bead-to-cell ratio of 2:1 in the presence of 2000 U/ml human IL-2. enabled in tTregs but precluded from pTregs. Founders were generated through microinjection and further screened for Thy1.1 CpG methylation analysis by bisulfite sequencing and Foxp3 expression. Phenotypic analyses of the resulting Foxp3 + Genomic DNA was extracted from different cell subsets and sodium bi- DCNS1-Cre reporter mice revealed that a fraction of CD4 T cells sulfite treated with a Qiagen EpiTect fast DNA bisulfite kit. Then, the expressed Thy1.1, and these Thy1.1+ cells were exclusively modified DNA was amplified by PCR and subcloned into the pGEM-T easy Foxp3+ based on intracellular staining of endogenous Foxp3. vector (Promega) and sequenced (20–40 colonies per subset). The PCR Moreover, the mean fluorescence intensity (MFI) of Foxp3 in primers used in the CpG methylation analysis were as follows: CNS2, + + + forward, 59-AAGGGGGTTTTAATATTTATGAGG-39,reverse,59- CD4 Foxp3 LN cells from Tg mice was comparable to that 2 CCTAAACTTAACCAAATTTTTCTACCA-39 from Tg mice (Fig. 1B). The percentage of Thy1.1+Foxp3+ tTregs among all Foxp3+ cells in the spleen (63.95 6 5.45%) and Adoptive transfer of Tregs LN (68 6 2.8%) was higher than in gut-related Peyer’s patches Total YFP+ cells or Thy1.12YFP+ and Thy1.1+YFP+ Tregs were respec- 6 6 + (55.15 4.45%) and the colon lamina propria (45.95 6.55%) tively FACS sorted from total CD4 T cells isolated from the spleen and (Fig. 1B), which is consistent with the notion that pTregs are LN of 6- to 8-wk-old Foxp3 DCNS1-Cre-Rosa26-YFP mice. Tregs (2 3 Downloaded from 105) were i.v. injected into 2-mo-old TCRb knockout (KO) mice. CD4+ preferentially induced by the gut microenvironment (4). Under TCRb+YFP+ cells were FACS sorted from the recipient mice and analyzed steady-state conditions, both Nrp1 and Helios have previously for Thy1.1 and Foxp3 expression 4 wk after transfer. been described as markers of tTregs (5, 7, 8). Thus, we tested + + 2 2 RNA sequencing and bioinformatics analysis Thy1.1 expression in Nrp1 Helios and Nrp1 Helios Treg subsets and found that 70.45 6 10.96% of Nrp1+Helios+ cells Lymphocytes were isolated from the spleens and LN of Foxp3 DCNS1-Cre- + 6 2 2 + were Thy1.1 . In contrast, only 12.21 3.098% of Nrp1 Helios

Rosa26-YFP mice and enriched for CD4 T cells using magnetic beads. http://www.jimmunol.org/ 2 + + + cells expressed Thy1.1 (Fig. 1C), consistent with the observation Then, Thy1.1 YFP and Thy1.1 YFP Tregs were sorted to a typical 2 2 purity of .97% using Thy1.1 and CD4 Abs. RNA sequencing (RNA-seq) that generation of the Nrp1 Helios subset relies on the CNS1 and bioinformatics analyses were conducted by Novogene. Raw data (raw region of Foxp3 (8). reads) in fastq format were first processed through in-house perl scripts. In To further test whether Thy1.1 expression can be used to trace this step, clean data (clean reads) were obtained by removing reads con- tTregs, we sorted naive conventional GFP2 T cells from Foxp3 taining adapters, reads containing poly-N, and low-quality reads from raw data. At the same time, the Q20, Q30, and GC contents of the clean data DCNS1-Cre/GFP KI double reporter mice, followed by TCR, were calculated. All of the downstream analyses were based on the clean TGF-b, and RA stimulation in vitro (28). In agreement with recent data with high quality. Reference genome and gene model annotation files results from Rudensky and colleagues (9) demonstrating that were directly downloaded from http://ftp.ensembl.org/pub/release-77/gtf/ CNS1 deletion largely limits iTreg induction, strong upregulation mus_musculus/. The index of the reference genome was built using Bowtie by guest on October 5, 2021 v2.2.3, and paired-end clean reads were aligned to the reference genome of Foxp3 was detected after 5 d of stimulation, whereas Thy1.1 using TopHat v2.0.12. We selected TopHat as the mapping tool because it induction was much less efficient (5.4 6 3.1%, Fig. 1C). Further can generate a database of splice junctions based on the gene model an- flow cytometric analyses revealed that the phenotype of Thy1.1+ notation file and thus a better mapping result than other nonsplice mapping Tregs was roughly in line with Thy1.12 Tregs, except for higher tools. HTSeq v0.6.1 was used to count the read numbers mapped to each CD25 and less expression of CTLA-4 and the proliferative marker gene. The expected number of fragments per kilobase of exon per million fragments mapped of each gene was calculated based on the length of the Ki67 (Fig. 1D). Thus, the Tg expression of Thy1.1 and the Cre gene and read counts mapped to this gene. The number of fragments per fusion protein did not disturb endogenous Foxp3 regulation. Taken kilobase of exon per million fragments mapped considers the effect of together, these results demonstrated the high specificity of the sequencing depth and gene length for the read counts at the same time and newly created Foxp3 DCNS1-Cre mice. is currently the most commonly used method to estimate gene expression levels. Prior to differential gene expression analysis, for each sequenced Expression of the Thy1.1-Cre reporter was unexpectedly library, the read counts were adjusted by the edgeR program package through one scaling normalized factor. Differential expression analysis of delayed in the thymus two conditions was performed using the DEGSeq R package (1.20.0). The The CNS1 region of Foxp3 contains multiple binding sites for p values were adjusted using the Benjamini–Hochberg method. Corrected transcription factors (e.g., Smad2/3, NFAT, and RA receptor) and is p = 0.005 and log2 (fold change) = 1 were set as the threshold for sig- nificantly differential expression. Raw RNA-seq data were submitted to the the major sensor of TGF-b in the Foxp3 locus. Aside from pTreg Gene Expression Omnibus database (accession no. GSE93856 at https:// induction, TGF-b signaling is also involved in the early devel- www.ncbi.nlm.nih.gov/geo). opment of tTregs; conditional KO of TGF-bRI using Lck-Cre– mediated deletion postpones the accumulation of neonatal Foxp3+ Statistical analysis thymocytes by a week (29). Therefore, we examined the Thy1.1 Differences between two data sets were analyzed by a paired or unpaired and Foxp3 expression in the thymus of Foxp3 DCNS1-Cre mice. two-tailed Student t test with Prism software (GraphPad Software). Error Interestingly, Thy1.1 was expressed by ∼30% of Foxp3+ cells bars denote SD. within the CD4 single-positive thymocytes (Fig. 2A). Compared to the Thy1.12Foxp3+ subset, these Thy1.1+Foxp3+ cells Results expressed less CD24 and CD69 but expressed high levels of Qa2 Generation of reporter mice for tracing tTregs and CD62L, indicating a more mature status. The percentage of D (Foxp3 CNS1-Cre) Thy1.1-labeled cells in CD24loQa2hiCD69loCD62Lhi mature To trace thymus-derived Foxp3-expressing cells in vivo, we gen- tTregs was ∼60% in the thymus, whereas only a few Thy1.1+ cells erated a novel BAC Tg mouse line that carries genes encoding Cre were detected in the immature counterpart (Fig. 2A). Moreover, recombinase, eqFP650, and the Thy1.1 reporter linked by a self- Thy1.1-labeled Foxp3+ cells also highly expressed two mature cleaving 2A peptide that can efficiently mediate cotranslational tTreg markers: CD25 and Nrp1 (Fig. 2B) (7, 30). Thus, Thy1.1 The Journal of Immunology 2615 Downloaded from http://www.jimmunol.org/ by guest on October 5, 2021

FIGURE 1. Generation of tTreg fate-mapping mice. (A) Construct design of the Foxp3 DCNS1-Cre reporter mice. (B) Thy1.1 and Foxp3 expression in LN cells from wild type (Tg2) and Foxp3 DCNS1-Cre Tg mice (Tg+) is shown (top left); cells were gated on CD4+. The MFIs of Foxp3 in Tregs from Tg2 and Tg+ mice are shown (top right). The percentages of Thy1.1 expression among the Foxp32 and Foxp3+ CD4+ T cells in the indicated tissues are shown. (C) Splenocytes from Foxp3 DCNS1-Cre reporter mice were gated on CD4+Foxp3+ and further gated based on Nrp1 and Helios into Nrp1+Helios+ and Nrp12Helios2 subsets (top left). The histograms show the Thy1.1 expression in the Nrp1+Helios+ (top middle) and Nrp12Helios2 (top right) Tregs. For in vitro iTreg induction, CD4+GFP2CD62Lhi naive cells were sorted from Foxp3 DCNS1-Cre/GFP KI mice, cultured with plate-coated anti-CD3 (0.5 mg/ml) and anti-CD28 (1 mg/ml) in the presence/absence of TGFb (2 ng/ml), IL-2 (200 U/ml), and RA (25 nM) for 4 d, and then (Figure legend continues) 2616 INSTABILITY OF tTregs IN VIVO Downloaded from http://www.jimmunol.org/ by guest on October 5, 2021 FIGURE 2. Thy1.1 labeled the late development of mature Tregs in the thymus. (A) CD4 single-positive thymocytes from Foxp3 DCNS1-Cre reporter mice were separated into Thy1.12Foxp32, Thy1.12Foxp3+, and Thy1.1+Foxp3+ subsets and analyzed for CD24, Qa2, CD69, and CD62L expression by flow cytometry. Representative histograms are shown (top, Thy1.12Foxp32 black, Thy1.12Foxp3+ blue, and Thy1.1+Foxp3+ red). The MFIs of each marker of the indicated population are shown. The percentages of Thy1.1+ cells among immature and mature tTregs are shown. Immature cells were gated on CD4+Foxp3+CD24hiQa2loCD69hiCD62Llo, and mature cells were gated on CD4+Foxp3+CD24loQa2hiCD69loCD62Lhi.(B) Expression of CD25 and Nrp1 by Thy1.1+ and Thy1.12 Foxp3+ cells in CD4 single-positive thymocytes from Foxp3 DCNS1-Cre reporter mice was analyzed by flow cytometry (top). The percentages of each marker among Thy1.1+ and Thy1.12 Foxp3+ cells are shown. (C) Thymocytes from Foxp3 DCNS1-Cre reporter mice were stained with Foxp3 and Thy1.1 and subjected to flow cytometric analysis every 3 d since birth (left); cells were gated on the CD4 single-positive. Mice were analyzed on different days but always compared with an adult mouse. The frequencies of Thy1.1 and Foxp3 expression of neonates relative to the adults are shown (right). Numbers represent the percentages of cells within the indicated gates (A–C). Each dot represents an individual mouse (A and B), and error bars denote SD (A–C). Data are representative of three independent experiments (A and B)orn $ 5(A–C). Mice were 6–8 wk of age at the time of the experiment (A and B). *p , 0.05, **p , 0.01, ***p , 0.001, ****p , 0.0001. SP, single-positive. expression likely begins at the later stage of tTreg development. would affect CNS3 function. We also performed CNS3 To investigate this possibility, we performed a time course study to H3K4me1–chromatin immunoprecipitation assays by comparing evaluate the Foxp3 and Thy1.1 levels in neonates, and a clear Tg+ Thy1.1+ Tregs and Tg2 CD25+ Tregs and found an ∼2-fold delay of Thy1.1 expression (day 6) compared with endogenous increase of H3K4me1 at CNS3 in the Thy1.1+ Tregs, further Foxp3 (day 3) was seen (Fig. 2C). supporting that the function of CNS3 was intact (Supplemental The pioneer element CNS3, which plays an important role in Fig. 1). The inconsistency with the results from the CNS1- tTreg development, is located in close proximity to exon 1 of the deficient mice could be due to the competition of the endoge- Foxp3 locus. The Cre cassette was inserted into the first ATG of nous Foxp3 allele with the Tg allele and/or epigenetic effects Foxp3, and the end of this cassette is still ∼230 bp away from the stemming from the integration site that may amplify the require- CNS3 enhancer. Thus, it is unlikely that the integration of Cre ments for TGF-b signaling (9). Importantly, the novel Thy1.1-Cre

Foxp3 and Thy1.1 expression was analyzed by flow cytometry. (D) Splenocytes from Foxp3 DCNS1-Cre reporter mice were stained with Foxp3, Thy1.1, and a series of markers and analyzed on a FACS Fortessa. The histogram plots are gated on Thy1.12Foxp3+ (blue line) and Thy1.1+Foxp3+ (red line) cells. The percentages and MFIs of the indicated markers are shown (right). Numbers represent the percentages of cells within the indicated gates (B and C). Each symbol represents an individual mouse (B–D); error bars denote SD. Data are representative of three independent experiments (n $ 5). Mice were 6–8 wk of age at the time of the experiment. *p , 0.05, **p , 0.01, ***p , 0.001, ****p , 0.0001. LP, lamina propria; mLN, mesenteric LN; pLN, peripheral LN. The Journal of Immunology 2617 Downloaded from http://www.jimmunol.org/ by guest on October 5, 2021

FIGURE 3. Lineage tracing of mature thymic Tregs. (A) Breeding strategy to generate Foxp3 DCNS1-Cre-YFP mice. (B) Thymocytes and LN cells from Foxp3 DCNS1-Cre 3 Rosa26-YFP mice were analyzed by flow cytometry. Cells were gated on CD4+Thy1.1+Foxp3+ and analyzed for Nrp1 and YFP expression. The percentages of YFP+ cells among Thy1.1+Foxp3+ cells of each organ are shown (right). (C) Highly pure Foxp32, Thy1.12YFP2Foxp3+, and Thy1.1+YFP+Foxp3+ thymocytes and peripheral lymphocytes from Foxp3 DCNS1-Cre 3 Rosa26-YFP mice were sorted and subjected to bisulfite sequencing of the CNS2 region of Foxp3 and intron 1b of eos; the demethylated patterns of each region are shown (top). Amplicons are subdivided by horizontal lines, with each representing an individual CpG motif. The demethylation status of individual CpG motifs is color coded according to the degree of demethylation at that site. The color code ranges from blue (0% demethylation) to yellow (100% demethylation) according to the color scale. Data are representative of three independent experiments. (D) YFP+ cells from the Peyer’s patch were respectively sorted from Foxp3 GFP-Cre 3 Rosa26-YFP and Foxp3 DCNS1-Cre 3 Rosa26-YFP mice and stained with Foxp3 followed by flow cytometric analysis (left). The percentages of exFoxp3 cells among YFP+ cells of the indicated tissues from each mouse were analyzed as in the Peyer’s patch (right). (E and F) FACS-purified YFP+ cells from Foxp3 DCNS1-Cre 3 Rosa26-YFP mice were i.v. injected into 2-mo-old TCRb KO mice, and 4 wk later, splenocytes from recipient mice were analyzed for Thy1.1 and Foxp3 expression. Cells were gated on CD4+TCRb+YFP+, and the percentages of Thy1.12 and exFoxp3 cells are shown (E). Splenic (Figure legend continues) 2618 INSTABILITY OF tTregs IN VIVO reporter mice were able to label late Foxp3+ cells derived from fully commit (10, 33). However, YFP-labeled Tregs are not ter- the thymus, which were bona fide Tregs with higher epigenetic minally differentiated. When highly purified YFP+ cells were stability (31). adoptively transferred into TCR KO mice and analyzed 4 wk later, a dramatic loss of Thy1.1 and Foxp3 expression was observed Lineage tracing of mature thymic Tregs (Fig. 3E). These exTregs were indeed reprogrammed into IFN-g– To evaluate the Cre activity of the new strain, we crossed the Foxp3 producing Th1 cells in the spleen and TFH cells in Peyer’s DCNS1-Cre mouse with the Rosa26-LSL-YFP tracer (25). In the patches. Moreover, the tTreg-derived TFH cells could partially resultant mice, Cre expression in the mature tTregs allows removal rescue the serum IgG1 and IgG2b deficiency in the TCR KO mice of the stop cassette, driving the constitutive expression of YFP (Fig. 3F, 3G). Of note, YFP-labeled exFoxp3 cells were exclu- + (Fig. 3A). Indeed, the large majority (96.71 6 1.03%) of Thy1.1 sively Thy1.12YFP+ (Fig. 3E). Thus, tTregs were a relatively Tregs from secondary lymphoid organs were labeled with YFP, stable subset in steady-state conditions, but they could be indicating high efficiency of Cre activity in the periphery. How- reprogrammed in response to outside signals, such as homeostatic hi ever, the percentage of YFP cells was much lower in the thymus proliferation in an empty host. (53.89 6 9.47%, Fig. 3B), instead of enriching for cells expressing low-to-medium levels of YFP, which represented the recently Development of Foxp3 instability through Treg activation generated Cre-expressing cells due to a delay between Cre ex- A previous study demonstrated that three CpG sites in the CNS1 pression and excision of the stop cassette in the Rosa26-YFP region of Foxp3 also have a similar Treg-specific methylation mice. These YFPlo-med cells also expressed relatively weak pattern as the CNS2 that is demethylated in Tregs but is highly Nrp1, supporting their less mature status (7, 30). Thus, this result methylated in naive T cells and in vitro–induced Tregs. Moreover, validated the high Cre efficiency of the Foxp3 DCNS1-Cre mice, the demethylation of CNS1 depends on Tet enzyme-mediated Downloaded from as well as the notion that the Thy1.1+ cells were indeed generated oxidization of methylcytosine, indicating that CNS1 may also be inside the thymus, not recirculated back from the periphery (32). involved in Treg lineage stability (15). Interestingly, in the Foxp3 Previous studies suggest that Foxp3 expression alone is not DCNS1-Cre 3 Rosa26-YFP strain, ∼10–20% Thy1.12YFP+ cells sufficient to establish the real Treg lineage, which must also include were also found in the LN, spleen, and Peyer’s patches in healthy Treg-specific epigenetic modifications (12). The CNS2 region, also animals, and the frequency increased with age (Supplemental Fig. 2 + known as the Treg-specific demethylated region, is well docu- 2A). Phenotype assessment revealed that Thy1.1 YFP and http://www.jimmunol.org/ mented for Treg lineage stabilization (16, 17). Its methylation Thy1.1+YFP+ cells express similar levels of Foxp3, Nrp1, Helios, status does not depend on Foxp3 expression but is instead con- and CTLA-4 (Supplemental Fig. 2B), indicating that Thy1.12YFP+ trolled by TCR signaling (12, 14). Gradual demethylation of cells maintain the tTreg identity (5, 7, 8). However, in vitro expan- CNS2 during Treg maturation has been described in the thymus sion of these two YFP+ subsets clearly showed that Thy1.12YFP+ (31). Thus, we performed methylation analysis on the sorted pu- cells possess more plasticity (Fig. 4A), which was further supported rified cells using bisulfite sequencing (13). Thy1.1+YFP+ Foxp3+ by separately transferring Thy1.1+YFP+ and Thy1.12YFP+ cells cells clearly possessed a much more demethylated CpG island at into lymphopenic animals (Fig. 4B). Similarly, all of the YFP- the CNS2 region, as well as another “nTreg-Me” region located in labeled exFoxp3 cells did not express Thy1.1 (Fig. 4A, 4B), and eos intron 1b (Fig. 3C) (12). Taken together, our results revealed bisulfite sequencing demonstrated that some of the Thy1.12YFP+ by guest on October 5, 2021 that the unexpectedly delayed expression of the Thy1.1 transgene cells experienced remethylation of the CNS1 region before trans- during Treg thymic maturation provides us a unique opportunity fer (Supplemental Fig. 2C). Thus, our results suggest that loss of to trace the maturation of tTregs in vivo. Thy1.1 precedes the termination of Foxp3, supporting the notion Whether tTregs are capable of downregulating Foxp3 expression that the CNS1 region of Foxp3 also contributes to the stable ex- and being reprogrammed into other T effector cells remains con- pression of Foxp3 (15). Furthermore, Thy1.12YFP+ can serve as a troversial. To date, all fate-mapping systems have identified Tregs useful indicator to track exFoxp3 precursor cells about to terminate by using Foxp3 expression, meaning that the Treg population might Foxp3 expression. be contaminated by uncommitted Tregs that transiently express Interestingly, we observed a higher frequency of Thy1.12 cells Foxp3 and contribute to the exFoxp3 cells (12). To evaluate the among YFP+ cells in nonlymphoid tissues than in lymphoid tis- Foxp3 stability of mature tTregs, we compared previously de- sues (Supplemental Fig. 2D). A previous study suggested that scribed Foxp3 fate-tracing (Foxp3 GFP-Cre 3 Rosa26-YFP) mice attenuated Foxp3 expression in intraislet Tregs might result in with the newly generated Foxp3 DCNS1-Cre 3 Rosa26-YFP Treg instability and functional conversion in diabetic NOD mice strain (10). YFP+ cells from both strains were sorted and fol- (34). However, we observed no difference in the MFI of Foxp3 in lowed by the intracellular staining of endogenous Foxp3 lymphoid and nonlymphoid tissues (Supplemental Fig. 2E). Im- (Fig. 3D). Analysis of spleen, mesenteric LN, peripheral LN, and portantly, tTregs from nonlymphoid tissues had a higher prolif- Peyer’s patches from multiple mice showed that ∼1% of YFP cells eration rate but equivalent apoptotic rate compared with Tregs from the Foxp3 DCNS1-Cre strain lost Foxp3 expression as from lymphoid tissues (Supplemental Fig. 2F, 2G). Thus, Treg compared with 5–10% from the GFP-Cre BAC Tg mice (Fig. 3D) activation was likely linked to Foxp3 instability. (10). This suggests that mature tTregs are quite stable and that the To determine the phenotype of the precursor of exFoxp3 cells, GFP-Cre BAC system also labels cells that transiently express RNA-seq analysis was performed on sorted Thy1.12YFP+ and Foxp3 or that initiate the Treg differentiation program but never Thy1.1+YFP+ cells from the Foxp3 DCNS1-Cre 3 Rosa26-YFP

lymphocytes were stimulated with PMA, ionomycin, and monensin for 3–4 h and examined for Foxp3 and IFN-g expression (F). (G) FACS-purified YFP+ cells from Foxp3 DCNS1-Cre 3 Rosa26-YFP mice or GFP+ cells from Foxp3GFP KI mice were i.v. injected into 2-mo-old TCRb KO mice and, 7 wk later, CXCR5 and PD1 expression among Foxp3+ and exFoxp3 cells in donor cells from Peyer’s patches was determined. Serum collected at different times after transfer was examined for IgG1 and IgG2b by ELISA. Numbers represent the percentages of cells within the indicated gates (B and D–G). Data are representative of three independent experiments with similar results (n $ 4); error bars denote SD (B, F, and G). Mice were 6–8 wk of age at the time of the experiment. *p , 0.05, **p , 0.01, ***p , 0.001, ****p , 0.0001. mLN, mesenteric LN; pLN, peripheral LN; PP, Peyer’s patch; SP, spleen; WT, wild type. The Journal of Immunology 2619 Downloaded from http://www.jimmunol.org/ by guest on October 5, 2021

FIGURE 4. Thy1.12YFP+ cells were the precursors of exFoxp3 cells. (A) Highly purified Thy1.1+YFP+ and Thy1.12YFP+ peripheral lymphocytes were FACS sorted from Foxp3 DCNS1-Cre 3 Rosa26-YFP mice and cultured with Dynabeads mouse T activator CD3/CD28 in the presence of high-dose human IL-2 for 7–9 d. The Thy1.1 and intracellular Foxp3 expression in each population was analyzed by flow cytometry (left). The (Figure legend continues) 2620 INSTABILITY OF tTregs IN VIVO mice. Very similar expression profiles were found in these two more YFP+ cells upregulated CXCR5 and PD1 in Peyer’s patches YFP+ populations. Only 40 genes were differentially upregulated, (Fig. 6A), in agreement with the finding that constitutive STAT5 and 20 genes were downregulated in Thy1.12YFP+ cells signaling in activated CD4+ T cells selectively blocks TFH (Fig. 4C). Thy1.1+YFP+ cells possessed a more bona fide Treg cell differentiation and GC formation, likely by repressing Bcl6 signature, such as upregulated CD25, Eos, Gpr83, Swap70, and expression (39). Clearly, more YFP+ cells lost Thy1.1 expression CCR7 (35). Both CD25 and Eos have previously been reported to followed by termination of Foxp3 expression after IL-2 neutrali- contribute to Foxp3 stability (19, 36), and CCR7 is the key che- zation (Fig. 6A). In contrast, ICOS blockage prevented TFR cell mokine receptor guiding Tregs to IL-2 sources (21). In contrast, generation, resulting in more stable Thy1.1 and Foxp3 expression Thy1.12YFP+ cells upregulated genes that are well documented to (Fig. 6B). Engagement of ICOS-ICOSL leads to recruitment of the be involved in Treg activation, including ICOS, Ki67, Cdc25b, PI3K regulatory subunit and triggers downstream signaling (40). Itgb1, CXCR3, CCR4, CCR2, Gzmb, and Ly6a (Fig. 4D, 4E). To To directly test whether sustained PI3K expression is linked to examine whether Ag-driven activation would affect Thy1.1 sta- Foxp3 instability, we crossed Foxp3 DCNS1-Cre with Rosa26- bility, we generated experimental autoimmune encephalomyelitis StopFL-P110* KI mice (41). We found reprogramming of tTregs, in Foxp3 DCNS1-Cre 3 Rosa26-YFP mice and examined decreased expression of Thy1.1 and Foxp3, and a dramatic gain in lymphocytes from immunized mice with an MHC–peptide tet- the TFH phenotype (Fig. 6C). These findings agree with recent b ramer I-A –MOG38–49, which binds to MOG35–55–peptide- data demonstrating that PTEN plays a crucial role in maintaining specific T cells (37). We found that MOG38–49-specific tTregs the stability of Tregs (42, 43). In conclusion, appropriate activa- included a larger fraction of Thy1.12 cells than polyclonal or tion is necessary for tTreg differentiation in specific immune resting tTregs at the peak stage, supporting the notion that Treg microenvironments, but overactivation raises the risk of Foxp3 activation might be an essential prerequisite in Foxp3 destabi- instability. IL-2 and ICOS controlled the two opposite sides of Downloaded from lization (Supplemental Fig. 3). stability and activation, prompting functional specification. Moreover, we evaluated the Foxp3 stability in two functionally Dual lineage tracing of Foxp3 and T-bet specialized Treg subpopulations, T follicular regulatory (TFR) cells and T-bet+ Tregs, using the Foxp3 DCNS1-Cre 3 Rosa26- To trace the functional specification of Tregs in vivo, we took YFP strain. CXCR5 and PD1 were used to identify TFH and TFR advantage of Cre/lox and Dre/rox, between which no crossover cells from CD4+ T cells within the Peyer’s patches (38). Ac- recombination occurs, to simultaneously examine the dynamics of http://www.jimmunol.org/ companied by gaining expression of CXCR5 and PD1, YFP+ cells Foxp3 and T-bet expression in vivo (24). Another two BAC Tg gradually increased ICOS expression and decreased the expression mouse lines were generated. The first Tg animal we developed of CD25, as previously reported (Fig. 5A). The frequency of was a T-bet BAC Tg mouse that carries genes encoding Dre Thy1.12YFP+ cells dramatically increased following increasing recombinase and the Thy1.1 reporter linked by a 2A peptide CXCR5 and PD1 expression. In the CXCR5hiPD1hi population, (Supplemental Fig. 4A) (26). Phenotypic analyses of the T-bet– + notable loss of Foxp3 (17.08 6 9.04%) was found in the Thy1.12 Dre–Thy1.1 reporter mice showed that a fraction of CD4 T cells subset, whereas the CXCR52PD12 population barely contained expressed Thy1.1. Upregulation of Thy1.1 was detected following + any exFoxp3 cells (3.67 6 2%, Fig. 5B). Thus, once tTregs mi- immunization of this strain (Supplemental Fig. 4B). Thy1.1 cells + by guest on October 5, 2021 grate into the GCs of Peyer’s patches, the microenvironment is were uniformly T-bet based on intracellular staining of endoge- + less favorable for the maintenance of Foxp3 expression and in- nous T-bet, and the large majority of Thy1.1 cells also expressed stead prompts reprogramming into TFH cells. Similar loss of IFN-g (Supplemental Fig. 4B). These results indicated that the Thy1.1, and sequentially Foxp3, was found in CXCR3+T-bet+ surface Thy1.1 expression in T-bet–Dre–Thy1.1 mice indeed Tregs (Fig. 5C, 5D). Taken together, these observations indicate served as a reliable marker for endogenous T-bet. that the well-developed tTregs were not terminally differentiated; The second Tg animal we developed carries the Rosa26 BAC they can further differentiate into different functional Treg sub- transgene with an inserted rox Rluc-stop rox cassette, followed by populations to better adapt to their environments. In contrast, the the gene encoding tdTomato reporter (Supplemental Fig. 4A). In the complicated activating and inflammatory signals in the milieu absence of Dre recombinase, only the Rluc is expressed due to the might terminate tTreg suppression functions and reprogram them presence of a transcriptional stop upstream of tdTomato. The ac- into Th cells. tivity of luciferase from splenocytes and LN cells in the Rosa26-rox tracer (Rosa26-rox-Rluc-stop-rox-tdTomato) mice was significantly IL-2 and ICOS signaling balance the fate of tTregs higher than in the Tg2 control mice (Supplemental Fig. 4C). To Campbell and colleagues (21) divided Tregs into central and ef- evaluate the stability of T-bet expression in vivo, we crossed the fector Tregs and proposed that activated/effector Tregs preferen- T-bet–Dre–Thy1.1 strain with the Rosa26-rox tracer strain. Analysis tially change their IL-2 signaling pathway to ICOS to maintain of splenocytes from the dual reporter mice demonstrated that homeostasis and function in vivo. We hypothesized that this sig- tdTomato expression can be detected in, and is varied within, dif- naling switch actually causes the instability of activated/effector ferent cell types, including CD4+ Tcells,CD8+ T cells, NK cells, tTregs. Indeed, after i.p. injection of anti–IL-2, we found that NKT cells, and B cells (Supplemental Fig. 4D).

percentages of Thy1.12 and exFoxp3 cells among each population are shown (right). (B) Thy1.12YFP+ and Thy1.1+YFP+ cells were i.v. injected into 2-mo- old TCRb KO mice and, 4 wk later, CD4+YFP+ donor cells were sorted from recipients and stained with Thy1.1 and intracellular Foxp3 followed by flow cytometric analysis (left). The percentages of Thy1.12 and exFoxp3 cells among each population are shown (right). (C) Purified Thy1.12YFP+ and Thy1.1+ YFP+ cells were analyzed by RNA-seq. Genes that were upregulated (red circles) and downregulated (blue circles) in Thy1.12YFP+ cells are shown. (D and E) Representative histograms of Eos, CD25, CCR7 (D, top), ICOS, Ly6a, Itgab1, Ki67, and CXCR3 (E, top) staining in splenic Tconvs (black lines), Thy1.12YFP+ (blue lines), and Thy1.1+YFP+ Foxp3 cells (red lines) from Foxp3 DCNS1-Cre-Rosa26-YFP mice are shown. MFIs and percentages (D and E, bottom) of the indicated genes in each population are shown. Numbers in quadrants (A and B) indicate the percentage of cells in each. Each dot represents an individual mouse, and error bars denote SD (E). Data are representative of three independent experiments (n $ 5). Mice were 6–8 wk of age at the time of the experiment. *p , 0.05, **p , 0.01, ***p , 0.001, ****p , 0.0001. The Journal of Immunology 2621

FIGURE 5. tTreg activation led to Foxp3 instability. (A) CD4+YFP+ cells of the Peyer’s patch from Foxp3 DCNS1-Cre 3 Rosa26-YFP mice were divided into CXCR52PD12, CXCR5intPD1int,and CXCR5+PD1+ subsets according to their CXCR5 and PD1 expression (left). ICOS and CD25 expression in three subsets was analyzed and is shown as histograms (mid- dle, black lines represent CXCR52PD12 cells, blue lines represent CXCR5intPD1int cells, and red lines represent CXCR5+PD1+ cells). The MFIs of ICOS and percentages of CD25 of the indicated subsets are shown Downloaded from (right). (B) Thy1.1 and Foxp3 expression in CXCR52PD12, CXCR5intPD1int, and CXCR5+PD1+ subsets are shown (left); cells were gated on CD4+YFP+. The per- centages of Thy1.12 and exFoxp3 cells among the indicated subsets are shown (right). (C) Splenic CD4+YFP+ cells from http://www.jimmunol.org/ Foxp3 DCNS1-Cre 3 Rosa26-YFP mice were separated into CXCR32YFP+ and CXCR3+YFP+ (left) and analyzed for ICOS and CD25 staining. Representative histo- grams are shown (middle, blue lines repre- sent CXCR32YFP2 cells, and red lines represent CXCR3+YFP+ cells). The MFIs of ICOS and CD25 of the indicated subsets are D 2 + shown (right). ( ) Splenic CXCR3 YFP by guest on October 5, 2021 and CXCR3+YFP+ cells were analyzed for Thy1.1 and Foxp3 expression by flow cytometry (left). The proportions of Thy1.12 and exFoxp3 among the indicated subsets are shown (right). Numbers repre- sent percentages of cells within the indi- cated gates (B–D). Data are representative of three independent experiments. Each dot represents an individual mouse, and error bars denote SD (A, B, and D)(n $ 5). Mice were 6–8 wk of age at the time of the ex- periment. *p , 0.05, **p , 0.01, ***p , 0.001, ****p , 0.0001. PP, Peyer’s patch; SP, spleen.

The T-bet–Dre–Thy1.1 3 Rosa26-rox tdTomato strain was data not show). Interestingly, most DP cells did not coexpress further crossed with Foxp3-GFP-Cre 3 Rosa26-lox YFP mice to T-bet and Foxp3 (Fig. 7A). In line with the previous results, T-bet+ generate a tetratransgenic strain (Fig. 7A). In these animals, T-bet+ cells in the DP population did not stably express Foxp3 (Fig. 7A). cells and their progeny were marked by tdTomato expression, and More strikingly, most Foxp3+ cells from the DP subpopulation did Foxp3+ cells and their progeny were marked by YFP expression. not maintain stable expression of T-bet. Moreover, those “exT-bet” Double-positive (DP) (YFP+tdTomato+) cells were clearly de- Foxp3+ (Thy1.12tdTomato+Foxp3+) cells expressed a central Treg tectable in various second lymphoid tissues, including the spleen, phenotype (i.e., downregulation of CXCR3 and upregulation LN, and Peyer’s patch under steady-state conditions (Fig. 7A and of CCR7, CD25, Bcl2, and CD62L; Fig. 7B) (21), supporting a 2622 INSTABILITY OF tTregs IN VIVO Downloaded from http://www.jimmunol.org/

FIGURE 6. IL-2 and ICOS-ICOSL signaling balance the maintenance and functional specialization of tTregs. (A and B) Flow cytometric analyses of by guest on October 5, 2021 CXCR5 and PD1 expression, as well as Thy1.1 and Foxp3 expression, in CD4+YFP+ cells from the Peyer’s patch on day 14 after injection of monoclonal anti–IL-2 (A) or anti-ICOSL (B) into Foxp3 DCNS1-Cre 3 Rosa26-YFP mice on days 0, 3, 6, 9, and 12. (C) Flow cytometric analyses of CXCR5 and PD1 expression (left), as well as Thy1.1 and Foxp3 expression (right), of CD4+YFP+ cells from the spleen of Foxp3 DCNS1-Cre 3 Rosa26-YFP and Foxp3 DCNS1-Cre 3 Rosa26-YFP-PI3K mice. Numbers represent the percentages of cells within the indicated gates (A–C). Data are representative of three independent experiments with similar results, and error bars denote SD (A–C)(n $ 5). Mice were 6–8 wk of age at the time of the experiment. *p , 0.05, **p , 0.01, ***p , 0.001, ****p , 0.0001. PP, Peyer’s patch; SP, spleen. reversion from effector Tregs back to central Tregs. Thus, highly cells. Furthermore, we demonstrated that the signaling switch functionally specialized Tregs have more Foxp3 expression plas- from IL-2 to ICOS facilitates Treg instability. Moreover, by uti- ticity, and the reversion to central Tregs allows them to preserve lizing a T-bet and Foxp3 dual tracing mouse system, we found that lineage stability. T-bet expression by Tregs is largely unstable. These exT-bet Tregs lose their effector phenotype but increase the expression of CD25, Discussion CCR7, and CD62L and revert to central Tregs (21). Our results are Whether tTregs are a stable lineage remains controversial. By using in line with recent observations from a human Treg type 1 diabetes a novel BAC Tg mouse Foxp3 DCNS1-Cre line that enables the mellitus immunotherapy study that demonstrates that expanded 2 genetic tracing of committed tTregs, we found that tTregs highly human Tregs (CD45RA CD45RO+) are long-lived and display express Nrp1, Helios, and CD25, are more highly demethylated in stable Foxp3 expression after adoptive transfer to type 1 diabetes the CNS2 region, and represent a stable population under steady- mellitus patients, and those infused Tregs re-express CD45RA and state conditions. Similar to Tconvs, tTregs that recognize either other markers representing a central Treg phenotype (44). Im- self or foreign Ags in the periphery are activated and differentiated portantly, the reversion from effector to central Tregs is accom- into specialized effector tTregs. However, these processes are two panied by regaining Foxp3 stability. sides of the same coin. Central Tregs expressed high levels of Taken together, our results highlight the different signal re- CCR7 and CD25, eliciting STAT5 signaling to ensure their lineage quirements for activation between Tregs and Tconvs, and we stability. Treg activation led to downregulation of CD25 but up- propose a model according to which the signal strength and/or regulation of ICOS, thus providing different homeostasis signals, duration during Treg activation and specialization determine the mainly through the PI3K/AKT pathway. In this study, we showed different destinies of effector Tregs (i.e., activation versus insta- that when tTregs are activated and develop into a functionally bility). Both Tconvs and tTregs are generated in the thymus and specialized population, such as TFR cells or T-bet+ Th1 Tregs, educated by an elaborate process during which the duration and some of them start to lose their Treg identity and generate Th-like strength of the TCR interaction with the self-peptide-MHC The Journal of Immunology 2623 Downloaded from http://www.jimmunol.org/

FIGURE 7. Dual lineage tracing of Foxp3 and T-bet. (A) Generation of a tetratransgenic strain (far left). Splenic CD4+ cells were divided into YFP and tdTomato DP, YFP or tdTomato single-positive, and YFP and tdTomato double-negative subsets (left). Representative flow cytometric analyses of Thy1.1 (T-bet) and Foxp3 expression in YFP and DP cells are shown (middle). The percentages of exFoxp3 among the single YFP+ and DP subsets (right) and + 2 B Thy1.1 DP and Thy1.1 DP cells (far right) are shown. ( ) Splenocytes from tetratransgenic mice were stained with a series of activity-related markers and by guest on October 5, 2021 analyzed by flow cytometry. Representative FACS plots show the expression of Thy1.1 and indicated markers in CD4+tdTomato+Foxp3+ cells (left); the MFIs of these markers in the indicated cell populations are shown (right). Numbers represent the percentages of cells within the indicated gates (A). Data are representative of three independent experiments with similar results (A and B)(n $ 5). Mice were 6–8 wk of age at the time of the experiment. *p , 0.05, **p , 0.01, ***p , 0.001, ****p , 0.0001. td, tdTomato.

complexes on APCs determine their fate (45). Most thymocytes that ramming might have a unique role in gut IgA production. For bind with high affinity undergo clonal deletion to limit autoim- T cell–deficient CD3ε2/2 mice, GC formation in Peyer’s patches is munity in the periphery. However, some self-agonist ligands with defective, whereas the adoptive transfer of Tregs rather than Tconvs medium-high affinity preferentially develop Tregs (46), which is rescues the defect (50). Most Tregs in the recipients lost their Foxp3 the alternative insurance of the organism to prevent autoreactivity. expression and transformed into IgA-inducing TFH cells in GCs of Both Tconvs and tTregs have the potential to recognize foreign Peyer’s patches. The absence of IgA and disordered IgA selection Ags in the periphery. For Tconvs, high-affinity TCR–MHC en- would affect the balance of gut bacterial communities (51). Treg- gagement may preferentially generate dominant T cell clonal derived TFH cells could support the affinity maturation of IgAs and expansion to combat infection. However, similar TCR–MHC en- maintain the diversity and composition of the gut microbiome. gagement in tTregs may induce strong ICOS/PI3K signaling that Thus, a harmonious intestinal microenvironment is achieved. Ob- is detrimental to Treg suppression activity and further terminate viously, Tregs regulate immune responses in multiple ways. Foxp3 expression. Decreased Foxp3 expression followed by Treg The detailed biology of exT-bet Tregs is largely unknown. activation has been observed in multiple infection models, such as Memory formation is the key characteristic of adaptive immunity, Toxoplasma gondii and Mycobacterium tuberculosis infection (47, and emerging evidence suggests that Tregs have a memory phe- 48). Because high-affinity self-peptide–MHCs are deleted during notype (52). Tregs with a memory phenotype express high levels thymic negative selection, the relatively weak self-peptide–MHCs of effector memory markers, including IL-7R (CD127), CD44, may provide a “just right” activation window for tTregs. This model and CD27 (52, 53). It is unknown whether Tregs, similar to may explain a commonly observed phenomenon in which Tregs Tconvs, can differentiate into effector and central memory cells. It prefer to suppress relatively weak immune responses while being is tempting to speculate that these exT-bet Tregs will give rise to relatively inefficient to control strong immune responses such as to Treg central memory cells. Future studies are needed to test this allo-skin grafts. A recent report demonstrates that Tregs are critical possibility in healthy and various disease models. helpers for the high-avidity CD8 CTL response but are likely In summary, we revisited Treg stability by using two novel lineage downregulated in low-avidity CD8 reactions (49). Treg reprog- tracing systems and propose a model in which the overactivation 2624 INSTABILITY OF tTregs IN VIVO and functional specialization of Tregs leads to the generation of 20. Josefowicz, S. Z., L.-F. Lu, and A. Y. Rudensky. 2012. Regulatory T cells: Foxp3 instability. mechanisms of differentiation and function. Annu. Rev. Immunol. 30: 531–564. 21. Smigiel, K. S., E. Richards, S. Srivastava, K. R. Thomas, J. C. Dudda, K. D. Klonowski, and D. J. Campbell. 2014. CCR7 provides localized access to Acknowledgments IL-2 and defines homeostatically distinct regulatory T cell subsets. J. Exp. Med. 211: 121–136. We thank Prof. Alfred Singer, Dr. Xuguang Tai, and Prof. Lukas T. Jeker for 22. Feng, T., A. T. Cao, C. T. Weaver, C. O. Elson, and Y. Cong. 2011. Interleukin- discussions and reading the manuscript. We thank Dr. Fuping Zhang (Institute 12 converts Foxp3+ regulatory T cells to interferon-g-producing Foxp3+ T cells of Microbiology, Chinese Academy of Sciences, Beijing, China) for providing that inhibit colitis. Gastroenterology 140: 2031–2043. the Foxp3 GFP KI mice. We also gratefully acknowledge Prof. A. Francis 23. Takahashi, R., S. Nishimoto, G. Muto, T. Sekiya, T. Tamiya, A. Kimura, R. Morita, M. Asakawa, T. Chinen, and A. Yoshimura. 2011. SOCS1 is Stewart (Genomics, BioInnovations Zentrum, Technische Universitaet Dresden, essential for regulatory T cell functions by preventing loss of Foxp3 ex- Dresden, Germany) for providing plasmids pCAGGS-Dre-IRES-puro pression as well as IFN-g and IL-17A production. J. Exp. Med. 208: 2055– and ROSA26-rox-neo-rox-lacZ-no-puro, as well as Prof. Hongkui Zeng 2067. (Allen Institute for Brain Science, Seattle, Washington) for providing 24. Anastassiadis, K., J. Fu, C. Patsch, S. Hu, S. Weidlich, K. Duerschke, F. Buchholz, F. Edenhofer, and A. F. Stewart. 2009. Dre recombinase, like Cre, is plasmid CAG-floxed tdTomato. a highly efficient site-specific recombinase in E. coli, mammalian cells and mice. Dis. Model. Mech. 2: 508–515. 25. Srinivas, S., T. Watanabe, C.-S. Lin, C. M. William, Y. Tanabe, T. M. Jessell, and Disclosures F. Costantini. 2001. Cre reporter strains produced by targeted insertion of EYFP The authors have no financial conflicts of interest. and ECFP into the ROSA26 locus. BMC Dev. Biol. 1: 4. 26. Szymczak, A. L., C. J. Workman, Y. Wang, K. M. Vignali, S. Dilioglou, E. F. Vanin, and D. A. A. Vignali. 2004. Correction of multi-gene deficiency References in vivo using a single “self-cleaving” 2A peptide-based retroviral vector. Nat. Biotechnol. 22: 589–594. 1. Sakaguchi, S. 2004. Naturally arising CD4+ regulatory t cells for immunologic self-tolerance and negative control of immune responses. Annu. Rev. Immunol. 27. Zhou, X., L. T. Jeker, B. T. Fife, S. Zhu, M. S. Anderson, M. T. McManus, and Downloaded from 22: 531–562. J. A. Bluestone. 2008. Selective miRNA disruption in T reg cells leads to un- 2. Sawant, D. V., and D. A. A. Vignali. 2014. Once a Treg, always a Treg? controlled autoimmunity. J. Exp. Med. 205: 1983–1991. Immunol. Rev. 259: 173–191. 28. Fontenot, J. D., J. P. Rasmussen, L. M. Williams, J. L. Dooley, A. G. Farr, and 3. Hori, S. 2014. Lineage stability and phenotypic plasticity of Foxp3+ regulatory A. Y. Rudensky. 2005. Regulatory T cell lineage specification by the forkhead T cells. Immunol. Rev. 259: 159–172. transcription factor Foxp3. Immunity 22: 329–341. 29. Liu, Y., P. Zhang, J. Li, A. B. Kulkarni, S. Perruche, and W. Chen. 2008. A 4. Atarashi, K., T. Tanoue, T. Shima, A. Imaoka, T. Kuwahara, Y. Momose, + + G. Cheng, S. Yamasaki, T. Saito, Y. Ohba, et al. 2011. Induction of colonic critical function for TGF-b signaling in the development of natural CD4 CD25 Foxp3+ regulatory T cells. Nat. Immunol. 9: 632–640. regulatory T cells by indigenous Clostridium species. Science 331: 337–341. http://www.jimmunol.org/ 30. Singh, K., M. Hjort, L. Thorvaldson, and S. Sandler. 2015. Concomitant analysis 5. Thornton, A. M., P. E. Korty, D. Q. Tran, E. A. Wohlfert, P. E. Murray, of Helios and Neuropilin-1 as a marker to detect thymic derived regulatory Y. Belkaid, and E. M. Shevach. 2010. Expression of Helios, an Ikaros tran- T cells in naive mice. Sci. Rep. 5: 7767. scription factor family member, differentiates thymic-derived from peripherally 31. Toker, A., D. Engelbert, G. Garg, J. K. Polansky, S. Floess, T. Miyao, U. Baron, induced Foxp3+ T regulatory cells. J. Immunol. 184: 3433–3441. S. Duber,€ R. Geffers, P. Giehr, et al. 2013. Active demethylation of the Foxp3 6. Gottschalk, R. A., E. Corse, and J. P. Allison. 2012. Expression of Helios in locus leads to the generation of stable regulatory T cells within the thymus. peripherally induced Foxp3+ regulatory T cells. J. Immunol. 188: 976–980. J. Immunol. 190: 3180–3188. 7. Yadav, M., C. Louvet, D. Davini, J. M. Gardner, M. Martinez-Llordella, 32. Thiault, N., J. Darrigues, V. Adoue, M. Gros, B. Binet, C. Perals, B. Leobon, S. Bailey-Bucktrout, B. A. Anthony, F. M. Sverdrup, R. Head, D. J. Kuster, et al. N. Fazilleau, O. P. Joffre, E. A. Robey, et al. 2015. Peripheral regulatory 2012. Neuropilin-1 distinguishes natural and inducible regulatory T cells among T lymphocytes recirculating to the thymus suppress the development of their regulatory T cell subsets in vivo. J. Exp. Med. 209: 1713–1722, S1–S19. precursors. Nat. Immunol. 16: 628–634. 8. Weiss, J. M., A. M. Bilate, M. Gobert, Y. Ding, M. A. Curotto de Lafaille, 33. Rubtsov, Y. P., R. E. Niec, S. Josefowicz, L. Li, J. Darce, D. Mathis, C. Benoist, by guest on October 5, 2021 C. N. Parkhurst, H. Xiong, J. Dolpady, A. B. Frey, M. G. Ruocco, et al. 2012. and A. Y. Rudensky. 2010. Stability of the regulatory T cell lineage in vivo. Neuropilin 1 is expressed on thymus-derived natural regulatory T cells, but not + Science 329: 1667–1671. mucosa-generated induced Foxp3 T reg cells. J. Exp. Med. 209: 1723–1742, S1. 34. Wan, Y. Y., and R. A. Flavell. 2007. Regulatory T-cell functions are subverted 9. Josefowicz, S. Z., R. E. Niec, H. Y. Kim, P. Treuting, T. Chinen, Y. Zheng, and converted owing to attenuated Foxp3 expression. Nature 445: 766–770. D. T. Umetsu, and A. Y. Rudensky. 2012. Extrathymically generated regulatory 35. Vandenbon, A., V. H. Dinh, N. Mikami, Y. Kitagawa, S. Teraguchi, N. Ohkura, T cells control mucosal TH2 inflammation. Nature 482: 395–399. and S. Sakaguchi. 2016. Immuno-Navigator, a batch-corrected coexpression 10. Zhou, X., S. L. Bailey-Bucktrout, L. T. Jeker, C. Penaranda, M. Martı´nez- database, reveals cell type-specific gene networks in the immune system. Proc. Llordella, M. Ashby, M. Nakayama, W. Rosenthal, and J. A. Bluestone. 2009. Natl. Acad. Sci. USA 113: E2393–E2402. Instability of the transcription factor Foxp3 leads to the generation of pathogenic 36. Sharma, M. D., L. Huang, J. H. Choi, E. J. Lee, J. M. Wilson, H. Lemos, F. Pan, memory T cells in vivo. Nat. Immunol. 10: 1000–1007. B. R. Blazar, D. M. Pardoll, A. L. Mellor, et al. 2013. An inherently bifunctional 11. Bailey-Bucktrout, S. L., M. Martinez-Llordella, X. Zhou, B. Anthony, subset of Foxp3+ T helper cells is controlled by the transcription factor Eos. W. Rosenthal, H. Luche, H. J. Fehling, and J. A. Bluestone. 2013. Self-antigen- Immunity 38: 998–1012. driven activation induces instability of regulatory T cells during an inflammatory 37. Korn, T., J. Reddy, W. Gao, E. Bettelli, A. Awasthi, T. R. Petersen, autoimmune response. Immunity 39: 949–962. B. T. Ba¨ckstro¨m, R. A. Sobel, K. W. Wucherpfennig, T. B. Strom, et al. 2007. 12. Ohkura, N., M. Hamaguchi, H. Morikawa, K. Sugimura, A. Tanaka, Y. Ito, Myelin-specific regulatory T cells accumulate in the CNS but fail to control M. Osaki, Y. Tanaka, R. Yamashita, N. Nakano, et al. 2012. T cell receptor autoimmune inflammation. Nat. Med. 13: 423–431. stimulation-induced epigenetic changes and Foxp3 expression are independent 38. Crotty, S. 2011. Follicular helper CD4 T cells (TFH). Annu. Rev. Immunol. 29: and complementary events required for Treg cell development. Immunity 37: 621–663. 785–799. 39. Johnston, R. J., Y. S. Choi, J. A. Diamond, J. A. Yang, and S. Crotty. 2012. 13. Floess, S., J. Freyer, C. Siewert, U. Baron, S. Olek, J. Polansky, K. Schlawe, STAT5 is a potent negative regulator of TFH cell differentiation. J. Exp. Med. H.-D. Chang, T. Bopp, E. Schmitt, et al. 2007. Epigenetic control of the 209: 243–250. foxp3 locus in regulatory T cells. PLoS Biol. 5: e38. 40. Gigoux, M., J. Shang, Y. Pak, M. Xu, J. Choe, T. W. Mak, and W. K. Suh. 2009. 14. Zheng, Y., S. Josefowicz, A. Chaudhry, X. P. Peng, K. Forbush, and Inducible costimulator promotes helper T-cell differentiation through phos- A. Y. Rudensky. 2010. Role of conserved non-coding DNA elements in the phoinositide 3-kinase. Proc. Natl. Acad. Sci. USA 106: 20371–20376. Foxp3 gene in regulatory T-cell fate. Nature 463: 808–812. 41. Srinivasan, L., Y. Sasaki, D. P. Calado, B. Zhang, J. H. Paik, R. A. DePinho, 15. Yue, X., S. Trifari, T. A¨ ijo¨, A. Tsagaratou, W. A. Pastor, J. A. Zepeda-Martı´nez, J. L. Kutok, J. F. Kearney, K. L. Otipoby, and K. Rajewsky. 2009. PI3 kinase C.-W. J. Lio, X. Li, Y. Huang, P. Vijayanand, et al. 2016. Control of Foxp3 signals BCR-dependent mature B cell survival. Cell 139: 573–586. stability through modulation of TET activity. J. Exp. Med. 213: 377–397. 42. Huynh, A., M. DuPage, B. Priyadharshini, P. T. Sage, J. Quiros, C. M. Borges, 16. Feng, Y., A. Arvey, T. Chinen, J. van der Veeken, G. Gasteiger, and N. Townamchai, V. A. Gerriets, J. C. Rathmell, A. H. Sharpe, et al. 2015. Control A. Y. Rudensky. 2014. Control of the inheritance of regulatory T cell identity by of PI(3) kinase in Treg cells maintains homeostasis and lineage stability. Nat. a cis element in the Foxp3 locus. Cell 158: 749–763. Immunol. 16: 188–196. 17. Li, X., Y. Liang, M. LeBlanc, C. Benner, and Y. Zheng. 2014. Function of a 43. Shrestha, S., K. Yang, C. Guy, P. Vogel, G. Neale, and H. Chi. 2015. Treg cells Foxp3 cis-element in protecting regulatory T cell identity. Cell 158: 734–748. require the phosphatase PTEN to restrain TH1andTFH cell responses. Nat. 18. Miyao, T., S. Floess, R. Setoguchi, H. Luche, H. J. Fehling, H. Waldmann, Immunol. 16: 178–187. J. Huehn, and S. Hori. 2012. Plasticity of Foxp3+ T cells reflects promiscuous 44. Bluestone, J. A., J. H. Buckner, M. Fitch, S. E. Gitelman, S. Gupta, Foxp3 expression in conventional T cells but not reprogramming of regulatory M. K. Hellerstein, K. C. Herold, A. Lares, M. R. Lee, K. Li, et al. 2015. Type 1 T cells. Immunity 36: 262–275. diabetes immunotherapy using polyclonal regulatory T cells. Sci. Transl. Med. 7: 19. Komatsu, N., M. E. Mariotti-Ferrandiz, Y. Wang, B. Malissen, H. Waldmann, 315ra189. and S. Hori. 2009. Heterogeneity of natural Foxp3+ T cells: a committed regu- 45. Klein, L., B. Kyewski, P. M. Allen, and K. A. Hogquist. 2014. Positive and latory T-cell lineage and an uncommitted minor population retaining plasticity. negative selection of the T cell repertoire: what thymocytes see (and don’t see). Proc. Natl. Acad. Sci. USA 106: 1903–1908. Nat. Rev. Immunol. 14: 377–391. The Journal of Immunology 2625

46. Jordan, M. S., A. Boesteanu, A. J. Reed, A. L. Petrone, A. E. Holenbeck, 50. Tsuji, M., N. Komatsu, S. Kawamoto, K. Suzuki, O. Kanagawa, T. Honjo, M. A. Lerman, A. Naji, and A. J. Caton. 2001. Thymic selection of CD4+CD25+ S. Hori, and S. Fagarasan. 2009. Preferential generation of follicular B helper regulatory T cells induced by an agonist self-peptide. Nat. Immunol. 2: 301–306. T cells from Foxp3+ T cells in gut Peyer’s patches. Science 323: 1488–1492. 47. Oldenhove, G., N. Bouladoux, E. A. Wohlfert, J. A. Hall, D. Chou, L. Dos 51. Kawamoto, S., M. Maruya, L. M. Kato, W. Suda, K. Atarashi, Y. Doi, Y. Tsutsui, Santos, S. O’Brien, R. Blank, E. Lamb, S. Natarajan, et al. 2009. Decrease of H. Qin, K. Honda, T. Okada, et al. 2014. Foxp3+ T cells regulate immuno- Foxp3+ Treg cell number and acquisition of effector cell phenotype during lethal globulin a selection and facilitate diversification of bacterial species responsible infection. Immunity 31: 772–786. for immune homeostasis. Immunity 41: 152–165. 48. Shafiani, S., C. Dinh, J. M. Ertelt, A. O. Moguche, I. Siddiqui, K. S. Smigiel, 52. Rosenblum, M. D., I. K. Gratz, J. S. Paw, K. Lee, A. Marshak-Rothstein, and P. Sharma, D. J. Campbell, S. S. Way, and K. B. Urdahl. 2013. Pathogen-specific A. K. Abbas. 2011. Response to self antigen imprints regulatory memory in Treg cells expand early during Mycobacterium tuberculosis infection but are tissues. Nature 480: 538–542. later eliminated in response to interleukin-12. Immunity 38: 1261–1270. 53. Gratz, I. K., H.-A. Truong, S. H.-Y. Yang, M. M. Maurano, K. Lee, A. K. Abbas, 49. Pace, L., A. Tempez, C. Arnold-Schrauf, F. Lemaitre, P. Bousso, L. Fetler, and M. D. Rosenblum. 2013. Cutting edge: memory regulatory t cells require T. Sparwasser, and S. Amigorena. 2012. Regulatory T cells increase the avidity IL-7 and not IL-2 for their maintenance in peripheral tissues. J. Immunol. 190: of primary CD8+ T cell responses and promote memory. Science 338: 532–536. 4483–4487. Downloaded from http://www.jimmunol.org/ by guest on October 5, 2021