Developmental Progression and Interrelationship of Central and Effector Regulatory T Cell Subsets

This article is featured in In This Issue, p.3497

Regulatory T cells that express Foxp3 are essential for establishing and maintaining self-tolerance and for limiting immune responses (1, 2). To mediate such activity, Foxp3⁺ Tregs exhibit substantial heterogeneity that is phenotypically distinguishable and appears to be analogous to that described for activated T effector (Teff) cells (3, 4). Recent evidence supports a model whereby effector Tregs (eTregs) with an activated phenotype are derived from a more quiescent pool of naive, central, or memory Tregs (5, 6). Central Tregs (cTregs) are characterized by expression of CCR7 and high levels of CD62L and lower expression of CD44 (7, 8). Ly-6C marks at least some cTregs, and these cTregs exhibit less suppressive activity and express lower affinity TCRs when compared with Ly-6C⁻ Tregs (9, 10). cTregs are primarily localized in secondary lymphoid tissues, and their homeostasis depends in part on signaling through the IL-2R (8). eTregs are characterized by expression of CD69, CD103, and Klrg1; high expression of CD44, GITR, and ICOS; and chemokine receptors that direct tissue localization, for example, CXCR3, CCR6, and CCR9 (11). Correspondingly, eTregs dominate tissue sites such as the gut mucosa, lung, fat, and acutely injured skeletal muscle (12). The development of eTregs from cTregs depends on TCR signaling and activity associated with the transcriptional regulator IFN regulatory factor 4 (IRF4) (6, 7, 13). Genome-wide expression profiling indicates that eTregs are distinguished from cTregs by increased levels of mRNAs for suppressive molecules such as IL-10, tissue-related chemokine receptors, and proapoptotic molecules, but lower levels of prosurvival mRNAs (14, 15). Overall, the activation of eTregs leads to their expansion into highly suppressive, but short-lived cells well suited to regulate inflammatory responses in nonlymphoid tissues.

Much has also been learned concerning the heterogeneity of Tregs based on examining their TCR repertoire, primarily using TCR transgenic models to limit their specificities. This work demonstrated that the Treg TCR repertoire is highly diverse, with only a partial and somewhat limited overlap with the repertoire found on conventional T cells (16–18). These studies indicate that most Tregs in the periphery are thymus derived, with peripherally developed Tregs more obvious in the gut mucosa. Tregs show heterogeneity in their TCR repertoire based on the anatomical location of draining secondary lymphoid tissues (19) or within distinct nonlymphoid tissue sites (20). This finding is consistent with localized responses by Tregs to tissue-specific self-Ags.

The above findings and additional studies indicate that cTregs give rise to eTregs (4, 14) and more robust clonal expansion is associated with eTregs (21). However, the relationship between phenotypically distinct subsets within cTregs and eTregs has not been extensively studied. Indeed, it is not known whether there is additional heterogeneity within cTregs and whether their development into eTregs represents a default pathway for cTregs that undergo clonal expansion. We also do not know whether clonal expansion of eTregs is broadly associated with these cells or rather is a property of an eTreg subset.

In the current study, these issues were addressed in part by examining mRNA expression of immune-related genes using the Nanostring platform, and the TCR repertoire by next generation sequencing for subsets of cTregs and eTregs. Adoptive transfer studies of purified subsets of Tregs were performed to assess their capacity to develop into eTregs and suppress autoimmunity. Our study is consistent with a model in which cTregs are not solely precursor cells for eTregs and that extensive clonal expansion in response to self-Ag is restricted to subsets of eTregs.

Foxp3/RFP reporter (22), Foxp3/GFP reporter (23), IL-2Rβ^−/− (24), IL-2Rβ^Y3 (referred to as Y3) (25), CD45.1-congenic, and TCR-β transgenic Tg(Tcrb)93Vbo/J (26) mice, all on the C57BL/6 genetic background, were bred within the specific pathogen-free animal facility at the University of Miami. Purified Tregs (1.5 × 10⁵) or Treg subsets (1 × 10⁵) from Foxp3-reporter mice or CD45.1 spleen cells (10 × 10⁶) were adoptively transferred i.v. All other studies used adult age-matched mice typically 8–16 wk of age. All animal studies were reviewed and approved by the Institutional Animal Care and Use Committee at the University of Miami.

The following mAbs to the indicated molecules, the fluorescent labels, and their sources were used in this study: CD4 (GK1.5, FITC; RM4-5, PE, BioLegend; PE-Cy7, BD Pharmingen); CD62L (MEL-14, allophycocyanin-Cy7, BioLegend; MEL-14, eVolve 605, eBioscience); CD69 (H1.2F3, PE-Cy7, BioLegend); CD103 (2E7, PE, eBioscience); Klrg1 (2F1/KLRG1, allophycocyanin, BioLegend, Biotin, eBioscience); Foxp3 (FKJ-16s, Percp-Cy5.5, eFluor450, eBioscience); TCR Vβ 8.1, 8.2 (KJ16-133.18, PE, BioLegend); IL-2, (Jes6-5H4, allophycocyanin, BioLegend); IL-17 (ebio17b7, PE-Cy7, eBioscience); IFN-γ (XMG1.2, PE, BioLegend); CD25 (PC-61, PE, BD Pharmingen; PE-Cy7, BioLegend); CD103 (2E7, allophycocyanin, eBioscience); CD39 (24DMS1, PE, eBioscience); Ki67 (B56, Alexa700, BD Pharmingen); Bcl-2 (BCL/10C4, PE, BioLegend); CD44 (PGP-1, FITC, Alexa647, laboratory made); CD5 (53-7.3, PerCP-Cy5.5, BioLegend); ICOS (15F9, PerCP-eFluor710, eBioscience); CTLA4 (UC10-4F10-11, PE, BD Pharmingen); CD73 (eBioTY/11.8, PE, PE-Cy7, eBioscience); Thy-1.1 (HIS51, PerCP-Cy5.5, FITC, eBioscience); and Ly-6C (AL-21, allophycocyanin-Cy7, Biotin, BD Pharmingen). Streptavidin conjugates used are as follows: PE-Strep (BioLegend) and PE-Cy7 Strep, PerCP-eFluor710-Strep, and allophycocyanin-Strep (eBioscience). We conjugated anti-CD4 (GK1.5) and anti-CD8 (53.6.7) using Alexa Fluor 700 NHS Ester from Molecular Probes, according to the manufacturer’s instructions. Samples were analyzed on a LSR-Fortessa-HTS. More than 100,000 events were collected per sample. To purify Tregs and Treg subsets, total CD4⁺ T cells were obtained using anti-CD4 magnetic MicroBeads (Miltenyi Biotec). The cells were then stained and sorted based on the expression of the Foxp3 reporter dye and cell surface markers using FACS Aria-II flow cytometry. Cells were usually >95% pure.

RNA was isolated from purified T cell populations, typically from 1–6 × 10⁵ cells, using the TRIzol reagent (Life Technologies), followed by ethanol precipitation in the presence of 40 μg glycogen (Affymetrix USB, ultrapure, MB grade) as a carrier. RNAs were dissolved in 12 μl RNase-free water, and 10 μl of this RNA was reverse transcribed in a 20 μl reaction. First, the RNA, poly(T)-oligo(T)16mer (5 μM) (Life Technologies) and dNTPs (500 μM) (Promega) were preheated at 65°C for 5 min. After cooling, recombinant RNasin RNase inhibitor (20 U) (Promega), DTT (5 mM), SuperScript III Reverse Transcriptase (10 U) (Life Technologies), and its buffer were added, and the cDNA was synthesized at 50°C for 1 h. The primers and probes for TaqMan PCR are shown in Supplemental Fig. 1A. The AY02 clonotype and Cα probes were synthesized and labeled 5′ with 6-carboxyfluorescein and 3′ with tetramethylrhodamine by Applied Biosystems. TaqMan reactions were performed in a volume of 20 μl containing cDNA (6 μl for the AY02 and 1 μl for the Cα reactions), the appropriate forward and reverse primers (500 nM) and gene-specific probe (250 nM), MgCl₂ (2.5 mM), and 2.5× 5′ MasterMix containing Taq DNA polymerase, dNTPs, and buffer components (Fisher Scientific). TaqMan PCR conditions were a single cycle at 50°C for 2 min and 95°C for 10 min, followed by 40 cycles at 95°C for 15 s and 60°C for 60 s using StepOnePlus Real-Time PCR Systems (Applied Biosystems). A standard ramp speed was used. For each reaction, amplification plots of threshold cycles were determined. Every experiment contained a positive (plasmid with the AY02 clonotype) and negative control (plasmid with an irrelevant Vα2 clonotype). The AY02 plasmid was used for the standard curves by dilution (10²–10⁸ copies) of this plasmid.

Mouse IL-2 (eBioscience, Recombinant Protein Carrier-Free) and anti-murine IL-2 Ab (JES6-1A12; Bio X Cell) were mixed at a 2:1 molar ratio and incubated for 30 min at room temperature. Each mouse received 1 μg IL-2/5 μg JES6-1A12 complex in 200 μl PBS i.p. for consecutive 3 d.

The indicated Treg subsets were lysed in RLT buffer (Qiagen). The lysates from 1 × 10⁴ cell equivalents were used for RNA profiling using the nCounter Mouse Immunology Gene Expression Code Set (NanoString) at the Oncogenomics Core at the University of Miami. All analysis of expression data, including normalization, log₂ transformation, determining fold change ratios between groups, Student’s two-tailed t testing, and calculation of false discovery rates was carried out using NSolver version 2.5, a statistical analysis program provided by Nanostring Technologies, according to the manufacturer’s recommendations.

DNA was isolated from the indicated Treg subsets using the QIAamp DNA Micro Kit (Qiagen). Amplification and sequencing of TCR-β CDR3 regions were performed using the immunoSEQ platform (Adaptive Biotechnologies, Seattle, WA). ImmunoSEQ combines multiplex PCR with high-throughput sequencing and a sophisticated bioinformatics pipeline for TCR-β CDR3 region analysis (27, 28). Sufficient sequence information was obtained to assign V subgroups and J regions associated with each CDR3.

Cells (1–2 × 10⁶) were cultured in 24-well flat-bottom plates with complete RPMI 1640 medium (29) containing PMA (50 ng/ml), ionomycin (1 μM), and brefeldin A (BioLegend) for 4 h. Cells were harvested, washed, and stained for cell surface markers and then fixed, permeabilized, and stained to detect IL-2, IL-17, and IFN-γ by flow cytometry.

Data were analyzed by one-way ANOVA applying Tukey’s multiple comparison test or unpaired two-tailed Student t test using GraphPad Prism 6.0 software, as indicated in the figure legends. All data are reported as the mean ± SD. In all figures, significant differences are indicated as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Ly-6C⁺ Tregs express lower affinity TCRs, are less suppressive, and are cTregs based on expression of high levels of CD62L and low levels of CD44 (9, 10). Nevertheless, it remains unclear whether all cTregs express Ly-6C. FACS analysis of peripheral splenic Tregs showed that only a subset of CD62L^hi cTregs expresses Ly-6C (Fig. 1A). Representative FACS profiles (Supplemental Fig. 2) and analysis of the resulting data (Fig. 1B) revealed that Ly-6C⁺ and Ly-6C⁻ cTregs and eTregs expressed similar levels of Foxp3. Ly-6C⁺ cTregs expressed the lowest levels of the activation molecules CD103, Klrg1, CD44, and ICOS; the proliferative marker Ki67; the Treg functional molecules CD39, CD73, and CTLA4; and CD5, whose levels directly correlate with the affinity of the TCR (30), but expressed the highest levels of Bcl-2. This expression pattern by the Ly-6C⁺ cTreg subset is consistent with long-lived slowly replicating cells that express TCRs with a relatively low affinity for self-Ag and low suppressive activity. In most cases, Ly-6C⁻ cTregs exhibited an intermediate phenotype when compared with Ly-6C⁺ cTregs and eTregs. Two exceptions were CD5 and CD25, whose levels were the highest for Ly-6C⁻ Tregs; this phenotype is consistent with a Treg subset that expresses higher affinity TCRs and preferentially responds to IL-2.

To further explore the relationship between these cTreg subsets and eTregs, Nanostring gene expression profiling was performed using an array that measures 561 immunologically relevant genes. Results from all mRNAs are shown in Supplemental Table 1A. Hierarchical clustering of the 150 differentially expressed mRNAs revealed that Ly-6C⁺ and Ly-6C⁻ cTreg subsets were much more related to each other than to eTregs (Fig. 1C). Many of these genes were similarly expressed by both cTreg subsets (Fig. 1D, regions 2 and 4; Fig. 1E), but for others expression by Ly-6C⁻ cTregs tended to move toward that of eTregs (Fig. 1D, regions 1, 3, and 5). Examples of such genes shown in Fig. 1E, which are associated with region 3, include Nt5e, Entpd1, Pdcd1, Cxcr3, Ccr6, Cxcr5, Ccr2, Irf4, Ahr, Casp1, and Casp3. Overall, these data show that cTregs express the lowest levels of molecules associated with Treg suppressive function, activation, tissue-seeking chemokine receptors, transcription factors related to TCR signaling, and regulation of suppressive pathways, but have the highest expression of molecules associated with T cell survival and regulation of Jak/Stat signaling. These properties were almost always the most polarized for Ly-6C⁺ cTregs, whereas the Ly-6C⁻ cTregs showed a tendency for higher activation, progressing toward eTregs.

Deep sequencing was performed for TCR-β for splenic Ly-6C⁺ and Ly-6C⁻ CD62L^hi CD69⁻ CD103⁻ cTregs and CD62L^lo CD69⁺ eTregs to further explore the relationship between these subsets. For these three Treg subsets from two replicate experiments, DNA was isolated from ∼7 × 10⁵ cells (range 4–11 × 10⁵). Productive sequence reads ranged from 3.7 to 9.5 × 10⁵, yielding ∼23,000–54,000 unique sequences with high repertoire diversity, as clonality of the repertoire measured between ∼0.02–0.05, where 1 equals monoclonal. Sequence coverage (mean 17.6; range 9.9–30.5) was similar for all samples. No skewing of V genes and J regions was noted between these subsets.

When assessing the entire Vβ repertoire of the deduced productive protein V region sequences, including the V gene and CDR3s, repertoires for these three Treg subsets overlapped by ∼10–12% (Fig. 2A). This result is consistent with a modest interrelationship among the specificities expressed by these subsets. The number of individual protein V sequences that overlapped between these subsets was also similar (∼3500–4000). The abundance of these shared sequences was relatively low, especially between the Ly-6C⁺ versus Ly-6C⁻ cTregs. The clonality of the shared sequences, that is, same nucleotide and amino acid V region sequences, was also low (∼32–42%) (Fig. 2B), consistent with convergent recombination characterizing some of the sequences within the overlapping TCR-β repertoires of cTregs and eTregs.

The 10 most abundant clonotypes were identified for Ly-6C⁺ and Ly-6C⁻ cTregs and eTregs. After normalization, eTregs showed much larger clone sizes than cTregs (Fig. 2C). This result is consistent with greater clonal expansion by eTregs. The distribution of the top 20 clonotypes of each subset between the cTreg and eTreg subsets was determined (Fig. 2D). The top sequences for Ly-6C⁺ cTregs were least frequently detected in eTregs, whereas the most frequent sequences in eTregs were almost entirely excluded from cTregs. Ly-6C⁻ cTregs showed an intermediate pattern where they were more readily found in Ly-6C⁺ cTregs and eTregs. Overall, these data support the notion that some Ly-6C⁺ and Ly-6C⁻ cTregs develop toward eTregs, which occurs more frequently in the latter, but also raise the possibility that some cTregs, especially Ly-6C⁺ Tregs, persist mainly as cTregs.

Past genome-wide profiling showed that CD69⁺, Klrg1⁺, and CD103⁺ Tregs appear to be more activated than total Tregs, cTregs, or CD62L^lo Tregs (14, 15). We refined this analysis to better quantify immunological properties of eTregs after fractionation into four subsets (Fig. 3A); all were CD62L^lo, but they varied in their expression of CD69 and CD103, with CD103 found on a minority of peripheral eTregs. FACS analysis showed that CD69⁻ CD103⁻ eTregs showed the least activated phenotype based on lower expression of Klrg1, CD44, and ICOS, but higher expression of Ly-6C and Bcl-2 (Fig. 3B). Highest proliferation was associated with CD69⁻ Tregs based on Ki67 expression. These cells showed lower CD25 and Foxp3 levels, consistent with the notion that this response was largely IL-2 independent (21). Treg functional molecules were distinctively expressed such that all subsets expressed equivalent levels of CD39, but the highest levels of CTLA4 and CD73 were associated with CD103⁺ and CD69⁺ eTregs, respectively.

Nanostring profiling was also performed for total cTreg and eTreg subsets. The expression of all mRNAs is shown in Supplemental Table 1B. Hierarchical clustering of the 193 differentially expressed mRNAs revealed that CD103⁻ eTreg subsets are more related to cTregs, with CD62L^lo CD69⁻ CD103⁻ Tregs the closest subset to cTregs (Fig. 3C). CD62L^lo CD69⁺ CD103⁺ Tregs were more distantly related to cTregs. Inspection of the clustering of individual mRNAs resulted in three major patterns of mRNA expression. One pattern is generally increasing expression of mRNAs from CD62L^lo CD69⁻ CD103⁻→CD62L^lo CD69⁺ CD103⁺ eTregs (Fig. 3D, regions 1 and 4). Representative mRNAs include Ctla4, Nt5e, Icos, Tigit, Ccr2, Ccr6, Cxcr3, Ahr, Prdm1, Maf, and Rorc (Fig. 3E). A second pattern of mRNAs is relatively high expression by CD103⁺ eTreg subsets (Fig. 3D, region 2), which include Entpd1 and Ccr3 (Fig. 3E). The third pattern is a somewhat favored expression of mRNAs related to the CD69⁺ eTreg subset, unrelated to whether they express CD103 (Fig. 3D, region 5). Some genes in this group include Irf4, Ebi3, Pdcd1, Tbx21, Ccr8, Il10, Cxcr5, and Sell (Fig. 3E). Very few mRNAs were found distinctly expressed only by the most highly active eTregs, CD62L^lo CD69⁺ CD103⁺ (Fig. 3D, region 3). A notable mRNA in this group is Pparg, which is important for tissue-residing Tregs (31). Collectively, these data support the existence of three main eTreg subsets: a transitional less activated CD62L^lo CD69⁻ CD103⁻ subset, and more highly activated CD62L^lo CD69⁺ and CD62L^lo CD103⁺ subsets.

Deep sequencing was performed for TCR-β for Treg subsets isolated based on distinctive expression of CD62L, CD69, and CD103 from wild-type (WT) mice. DNA was isolated from ∼4 × 10⁵ cells (range 2–6 × 10⁵). The total number of sequence reads (∼275,000–350,000), the number of unique sequences (∼16,000–32,000), and the repertoire diversity (∼0.04–0.09), where 1 equals monoclonal, and sequence coverage (mean 10.8; range 7.0–15.4) was similar for all samples. No skewing of V genes and J regions was noted between these subsets.

Comparing the deduced TCR-β protein V region sequences, including the V gene and CDR3s, in the entire repertoire between all samples showed that the four eTreg subsets were more related to each other than cTregs (Fig. 4A). In accordance with the mRNA expression analysis (Fig. 3D), the highest relationship of TCR-β repertoires was noted between CD62L^lo CD69⁺ CD103⁻ and CD62L^lo CD69⁺ CD103⁺ subsets and then between CD62L^lo CD69⁻ CD103⁺ and CD62L^lo CD69⁺ CD103⁺ subsets (Fig. 4A). The clonality of these shared sequences, that is, clonotypes encoded by a single DNA sequence, followed a similar hierarchy with the lowest clonality for cTregs and the highest for CD62L^lo CD69⁺ CD103⁻, CD62L^lo CD69⁻ CD103⁺, and CD62L^lo CD69⁺ CD103⁺ eTregs (Fig. 4B). Thus, these data show that these eTreg subsets are related to each other, with relatedness and clonality increasing for Tregs that express a more activated phenotype.

To further examine the relationship among Treg subsets, the 20 most abundant clonotypes within each subset were identified and evaluated for the extent to which they appeared in other Treg subsets. Heat maps, in which yellow represents the absence of a specificity, revealed that the dominant clonotypes associated with cTregs were largely absent in the other four Treg subsets, whereas the dominant specificities for the more activated Tregs were largely absent from the CD62L^hi Tregs (Fig. 4C). Substantial overlap in individual clonotypes was found between the four eTreg subsets. As the most dominant sequences most likely represent TCR specificities associated with clonal expansion, this finding provides further evidence supporting a unique composition of some of the Treg specificities within the CD62L^hi cTregs. The relatively high relatedness and clonality of TCR-β clonotypes for the CD62L^lo CD69⁺ CD103⁻, CD62L^lo CD69⁻ CD103⁺, and CD62L^lo CD69⁺ CD103⁺ and their high presence within all CD62L^lo subsets (Fig. 4A–C) are also consistent with highly interrelated cell populations that have undergone clonal expansion.

To estimate which Treg subsets underwent the greatest clonal expansion, we plotted the numbers of sequence reads for the 20 top clonotypes for each Treg subset (Fig. 4D). The relative clone sizes were the largest for the CD62L^lo CD69⁺ CD103⁻, CD62L^lo CD69⁻ CD103⁺, and CD62L^lo CD69⁺ CD103⁺ Treg subsets, whereas lower clonal expansion is noted for CD62L^lo CD69⁻ CD103⁻ Tregs. Thus, the highly activated phenotype of these cells correlates with a potential for high clonal expansion in response to self-Ags.

The distinctive repertoire of cTreg versus eTreg subsets raised the question of the overall capacity of cTregs to generate eTregs. To test this point, highly purified Ly-6C⁺ and Ly-6C⁻ cTregs and CD62L^lo CD69⁺ eTregs were transferred to IL-2Rβ^Y3 (Y3) mice (Fig. 5A), and their development into eTregs was assessed 2 wk posttransfer. Y3 mice express a mutant IL-2Rβ in T lineage cells, where three critical tyrosine residues in the cytoplasmic tail were mutated to phenylalanine. The low IL-2R signaling by the mutant IL-2Rβ is sufficient to support largely normal Treg development and homeostasis (25). Engraftment of donor WT Tregs in lympho-replete Y3 mice is much more extensive when compared with WT recipients, probably due to favorable competition for IL-2 and permits following the development of small numbers (1 × 10⁵) of donor Treg subsets.

Two weeks posttransfer, donor Tregs were noted in the spleen (Fig. 5B) and mesenteric lymph node (MLN) of Y3 mice. This analysis revealed that the large majority (>95%) of the donor cells expressed Foxp3 (data not shown), indicating that donor Tregs were stable. Quantitative assessment indicated greater engraftment by Ly-6C⁻ cTregs when compared with Ly-6C⁺ cTregs. Phenotypic analysis (Fig. 5C) is consistent with essentially all donor cTregs developing into eTregs. Notably, donor Tregs showed uniformly high expression of CD44, ICOS, and CTLA4, a characteristic phenotype of eTregs (Fig. 5D).

Most donor Tregs were actively proliferating based on Ki67 expression, and this most likely accounts for the increase in Klrg1, which marks Tregs that have undergone extensive proliferation and terminal differentiation. The heterogeneous expression of CD62L, CD69, and CD103 by the donor cells, regardless of whether the input Tregs were derived from a cTreg or an eTreg, suggests that the expression of these molecules is under the control of environmental cues. Somewhat surprisingly, Ly-6C⁺ and Ly-6C⁻ donor Tregs yielded increased development into CD103⁺ Tregs when compared with recipients that received CD62L^lo CD69⁺ eTregs. This finding suggests that the programming of cTregs and eTregs is distinctive with respect to the subtypes of eTregs that are supported. Overall, these data indicate that within the environment of IL-2Rβ^Y3 recipients, at least some cTregs develop into eTregs and that eTregs appear not to revert into cTregs.

Public TCR-α clonotypes were identified for peripheral Tregs from a TCR-β transgenic mouse line and after their transfer into IL-2Rβ^−/− mice (32). For one such clonotype, designated AY02, a TaqMan RT-PCR assay was developed to quantify this clonotype using a previously described approach (33) and the primers and probes shown in Supplemental Fig. 1A and 1B. This TaqMan assay is specific for AY02 as a PCR product was not detected when six other clonotypes were assayed, some with CDR3s closely related to the AY02 clonotype (Supplemental Fig. 1C). This TaqMan assay accurately measured the frequency of the AY02 clonotype in the context of irrelevant Vα2 TCRs except in a large excess of an irrelevant V region that also contained Jα23, which is also used by AY02 (Supplemental Fig. 1D), which is most likely due to primer competition. This condition is not expected when assessing the AY02 frequency in a polyclonal Vα2 TCR repertoire because the Jα23 containing TCRs will not be dominant.

The frequency of the AY02 clonotype was determined for unfractionated Tregs and cTreg and eTreg subsets, using standard curves established with plasmid DNA containing the AY02 clonotype (Supplemental Fig. 1E). The AY02 clonotype was detected in Tregs from the thymus, spleen, and MLN (Fig. 6A) at a frequency usually between ∼0.2 and 2% of the total Tregs. Although we typically used Tg⁺ TCR-α^+/+ mice, the potential for biallelic expression of AY02 did not obviously affect our determination of AY02 because this clonotype was detected at an equivalent frequency for splenic Tregs from Tg⁺ TCR-α^+/− mice (Fig. 6A). The detection of AY02⁺ Tregs in the thymus indicates that this clonotype marks Tregs of thymic origin. AY02 was not detected in thymic (n = 5) or peripheral (n = 8) CD4⁺ Foxp3⁻ T cells (data not shown), indicating that this specificity is unique to Tregs. After fractionation into Treg subsets, the AY02 clonotype was most evident in CD62L^hi CD69⁻ cTregs, but was also detected at a significantly lower amount in CD62L^lo CD69⁻ eTregs; AY02 was not detected in CD62L^lo CD69⁺ eTregs (Fig. 6B). When cTregs were fractionated into Ly-6C⁺ and Ly-6C⁻ cTregs and CD103⁺ and CD103⁻ eTregs, AY02 was highly associated only with Ly-6C⁺ cTregs (Fig. 6C). Thus, in the steady state, Tregs expressing the AY02 clonotype are largely restricted to Ly-6C⁺ cTregs while being largely absent from CD69⁺ and/or CD103⁺ eTregs.

The fortuitous detection of the AY02 specificity in Ly-6C⁺ cTregs provided an opportunity to test whether conditions that support extensive Treg proliferation and activation drove the development of AY02-expressing cTregs into eTregs. One approach was to treat mice with IL-2/anti–IL-2 agonist complexes (IL2-IC) to enhance the proliferation of Tregs and increase eTregs, as assessed by expression of CD103 and Klrg1 (Fig. 6D). When the distribution of the AY02 clonotype in IL2-IC–treated mice was assessed in Treg subsets based on expression of CD62L and CD69, the AY02 clonotype remained highly preferentially associated with the CD62L^hi CD69⁻ subpopulation (Fig. 6E).

The detection of AY02⁺ Tregs in the MLN (Fig. 6A) raised the possibility that eTregs with this clonotype might be readily found in the intestinal lamina propria. The expansion of Tregs by IL2-IC facilitated the isolation of sufficient numbers of Tregs to test this point. However, the intestinal lamina propria contained very few AY02⁺ Tregs (Fig. 6E), providing additional data that this Treg specificity is not substantially associated with tissue-associated activated eTregs.

Another approach was to adoptively transfer TCR-β transgenic Tregs into neonatal IL-2Rβ^−/− mice, which promotes narrowing and reshaping of the donor Treg TCR repertoire (32) and more effectively promoted the development into eTregs (Fig. 6D). The AY02 clonotype was detected in donor TCR-β transgenic Tregs from 9 of 10 individual IL-2Rβ^−/− recipients (Fig. 6F), despite repertoire reshaping. The typically lower frequency of AY02 when compared with input Tregs might reflect some tendency to select against this specificity or might reflect that Treg activation and proliferation favor CD62L^lo CD103⁺ eTregs and that AY02⁺ Tregs were excluded from this Treg subset. To assess this possibility, we examined the distribution of the AY02 clonotype among subsets of engrafted donor Tregs based on expression of CD62L and CD69 (Fig. 6G). Again, the AY02 clonotype was largely restricted to the CD62L^hi CD69⁻ subset with a frequency similar to the input Tregs. Collectively, these data indicate that CD62L^hi Tregs do not simply represent a pool of cells that ultimately give rise to eTregs, but may also contain a population of Tregs whose specificities remain largely restricted to cTregs, perhaps with optimal activity in lymphoid tissues.

To test the suppressive activity of cTregs and eTregs, we compared the capacity of purified CD62L^hi CD69⁻ CD103⁻ Klrg1⁻ cTregs to CD62L^lo CD103⁺ eTregs to prevent autoimmunity after transfer into neonatal IL-2Rβ^−/− mice. Controls included untreated IL-2Rβ^−/− mice and IL-2Rβ^−/− mice that received either total purified Tregs or unfractionated spleen cells as a source of total Tregs. Donor Tregs readily engrafted and persisted within the peripheral lymph node (PLN), MLN, as well as lamina propria of the small intestine (SI-LP) and colon lamina propria of IL-2Rβ^−/− recipients at largely equivalent levels (Supplemental Fig. 3A). For the PLN and MLN, an increased frequency of Tregs was noted for recipients that received cTregs, consistent with some preference for lymphoid tissues. However, the transferred cTregs readily developed into CD62L^lo CD103⁺ eTregs that were present at only slightly lower levels than occurred for recipients of CD103⁺ eTregs (Supplemental Fig. 3B). Thus, these data further confirm that some cTregs are precursors to eTregs and that the resulting eTregs adopted a similar phenotype in comparison with mice that received purified eTregs.

Untreated IL-2Rβ^−/− mice develop rapid systemic autoimmunity that leads to a wasting syndrome and inflammation in multiple tissues, particularly the lung, liver, and intestines, that typically results in death by 8–12 wk of age. Neonatal IL-2Rβ^−/− mice that were adoptively transferred with CD62L^hi cTregs or CD103⁺ eTregs were outwardly normal without evidence of wasting when analyzed 10–12 wk posttransfer. Therefore, we focused on early changes associated with autoimmunity in IL-2Rβ^−/− mice, that is, increased lymph node cellularity and increased proportion of activated T cells. With respect to suppression of these autoimmune symptoms, lymphoproliferation in the PLN and a high frequency of CD4⁺ CD44^hi CD62L^lo Teff cells in the PLN and MLN were not seen for recipients that received cTregs or CD103⁺ eTregs (Fig. 7A, 7B). However, increased cellularity in the MLN was noted for IL-2Rβ^−/− recipients that received cTregs (Fig. 7A). This increase was much greater than what was seen for untreated IL-2Rβ^−/− mice. This result most likely reflects the age difference at the time of assay, 4–6 wk for the untreated IL-2Rβ^−/− mice and 10–12 wk for the Treg transferred mice, as the former, due to severe autoimmunity, must be tested by this age.

The increased MLN cellularity within cTreg-treated IL-2Rβ^−/− mice raised the possibility that this Treg subset is less effective in controlling tolerance within the gut mucosa. Indeed, the MLN from these mice showed slightly increased numbers of CD4⁺ T cells (Fig. 7C), including IL-2– and IL-17–secreting Teff cells (Fig. 7D). Furthermore, when comparing the cTreg versus CD103⁺ eTreg transferred IL-2Rβ^−/− recipients, the SI-LP, but not colon lamina propria, of the cTreg-treated mice showed a significant increase in the total number of CD4⁺ T cells (Fig. 7C) and IL-2– and IL-17–producing Teff cells (Fig. 7D). Thus, eTregs derived from cTregs are less effective in maintaining immune homeostasis in the SI-LP.

Tregs from secondary lymphoid tissues, which are dominated by thymic-derived Tregs, are highly heterogeneous cells that are classified into two major subsets, cTregs and eTregs, based on their distinctive phenotype, homeostasis, anatomical preference, and transcriptional regulation (5–8, 12, 13). This study refines this view by showing that there are distinctive subpopulations of cTregs and eTregs and by directly considering the interrelationships between these subsets. These data support the view that there are two main populations of cTregs that are distinguished by expression of Ly-6C and three subpopulations of eTregs that are distinguished by their distinctive expression of CD62L, CD103, and CD69. These three eTreg subsets are distinct from a fourth terminal short-lived eTreg subpopulation, marked by expression of Klrg1 (14).

Extensive immune profiling and TCR repertoire analysis demonstrate that Ly-6C⁺ and Ly-6C⁻ cTregs are related to each other, but represent distinctive cTreg subsets. Many immunological properties are similarly exhibited by these two subsets. However, a set of immune-related mRNAs was identified that show expression by Ly-6C⁻ cTregs moving toward eTregs. On average Ly-6C⁻ Tregs express TCRs with a higher affinity for self-Ags, based on higher expression of CD5, which reflects the intrinsic affinity of the TCR for positively selected self-Ags (30). This greater responsiveness to self-Ags provides one condition to promote a somewhat more activated immune profile by Ly-6C⁻ cTregs. Another difference between Ly-6C⁺ and Ly-6C⁻ cTreg subsets is that at the steady state the Ly-6C⁻ Tregs showed greater proliferation as measured by Ki67 expression. This increase correlated with heightened expression of CD25 that would promote responsiveness to IL-2. Indeed, the homeostasis of cTregs has been reported to be largely dependent on IL-2 (21). Our findings suggest that IL-2–dependent homeostasis of cTregs primarily reflects the action of IL-2 on Ly-6C⁻ cTregs.

One function of Ly-6C⁺ and Ly-6C⁻ cTregs is to provide a pool of naive Tregs that become activated into CD62L^lo eTregs. This conclusion is based on following the fate of cTregs after transfer into Y3 or IL-2Rβ^−/− mice and shared TCR repertoires between cTreg subsets and eTregs. This latter result is in line with other studies that have found the same TCR clonotype in resting and activated Tregs (34–37). However, our study shows that Ly-6C⁺ cTregs are less efficient in developing into eTregs when directly transferred in vivo. Furthermore, the dominant TCR Vβ clonotypes of Ly-6C⁺ cTregs were detected at a low frequency in eTregs, and reciprocally the prevalent TCR Vβ clonotypes of eTregs were also at a low frequency in Ly-6C⁺ cTregs. In comparison with eTregs, the dominant TCR Vβs of Ly-6C⁺ cTregs were more often found in Ly-6C⁻ cTregs. Thus, expression of Ly-6C marks a quiescent and relatively unactivated population of Tregs with lower affinity TCRs that have a greater tendency to develop into Ly-6C⁻ cTregs rather than into eTregs. Thus, one simple interpretation of our data is that cTregs serve primarily as precursors of eTregs and the latter mediate suppressive function required for self-tolerance.

Several recent studies have shown that eTregs are more suppressive than cTregs, including Ly-6C⁺ Tregs (7, 9, 14). Consistent with this view, the transfer of CD103⁺ eTregs more fully protected IL-2Rβ^−/− mice from autoimmunity than recipients of cTregs. This result was somewhat surprising because in this model cTregs readily gave rise to eTregs, and full protection was expected. IL-2Rβ^−/− mice that received cTregs were effectively protected from systemic autoimmunity and colitis, but they showed symptoms consistent with inflammation in the lamina propria of the small intestine. Thus, greater suppressive activity of the CD103⁺ eTregs may not fully explain their high efficacy. The effective control of autoimmunity associated with IL-2Rβ^−/− mice that received donor eTregs may also be the result of distinctive TCR repertoires of cTregs and eTregs, whereby cTregs lack TCR specificities for some tissue-associated Ags, whereas the TCR repertoires of eTregs were already selected to be more responsive to such Ags. Nevertheless, we cannot rule out that the initial distinctive immune properties of the donor cTregs may also contribute to inflammation in the small intestine of IL-2Rβ^−/− recipients, even though they developed into eTregs. In particular, the inefficient development of eTregs from Ly-6C⁺ Tregs within the donor cTregs may have effectively lowered the number of Tregs available to suppress disease relative to those mice that received the CD103⁺ eTregs.

Besides their role of precursors to eTregs, some of our findings also raise the possibility of a more active role of cTregs, including Ly-6C⁺ cTregs, in maintaining self-tolerance. In particular, the public AY02 clonotype, which was identified as largely restricted to Ly-6C⁺ cTregs, did not substantially shift toward more highly activated eTregs, even when Tregs expressing this clonotype were placed under conditions that support substantial proliferation and activation, that is, stimulation with IL2-IC or transfer into IL-2Rβ–deficient mice. The detection of the AY02 in the latter situation suggests that AY02 may be a useful Treg clonotype, as this specificity persisted under conditions in which substantial reshaping of the Treg TCR repertoire occurs to suppress autoreactive T cells (32), that is, adoptive transfer of Tregs into IL-2Rβ^−/− recipients. Given the lower affinity of Ly-6C⁺ Tregs and the inability to drive AY02 into highly activated Tregs, AY02⁺ Tregs were not expected to be maintained at a relatively high frequency by cTregs, if these Tregs did not promote immune tolerance. In addition, the public AY02 clonotype was readily found in Tregs within the spleen and especially in the MLN, but was poorly detected in the lamina propria of the small intestine and colon. This finding markedly differs from past work in which Treg specificities associated with a tissue site were predominately found in the lymph nodes that drain that tissue (19) and suggests that some Tregs prefer to persist within lymphoid tissue.

We propose that CD62L^hi cTregs may also function to directly suppress autoreactive T cells within secondary lymphoid tissues. The distinctive immune profiles and TCR repertoire of cTregs raise the possibility that these Tregs use a distinctive mechanism to maintain tolerance, perhaps based on the action of TGF-β and CTLA4, whose mRNAs are abundant in these cells. The action of these cells may be particularly important under homeostatic nonautoimmune conditions when the clonal frequency of many autoreactive T cells is expected to be low and may be effectively regulated by individual cTregs at low frequencies. Once the immune system is perturbed in a manner that causes particular autoreactive T cells to become more abundant, some cTregs develop into eTregs to try to contain their activity in lymphoid tissues as well as the target tissues that express the relevant autoantigens.

Our immune profiling and TCR repertoire analysis identified three main interrelated subsets of eTregs, that is, CD62L^lo CD69⁻ eTregs, with an immune profile more closely related to cTregs, and more activated CD62L^lo CD69⁺ and CD62L^lo CD103⁺ eTregs. Each subset showed a preference for expression of particular mRNAs. However, these expression patterns do not support the existence of mutually exclusive subsets, but a preference for somewhat distinctive gene expression most likely based on environmental or niche signaling received by the eTregs. For example, CD62L^lo CD69⁻ eTregs generally showed low expression of many mRNAs related to T cell activation, suppressive function, and tissue homing. Nevertheless, CD62L^lo CD69⁻ CD103⁺ eTregs expressed some features of a more activated Treg. Levels of other mRNAs showed a gradient of expression from low to high in comparison with cTregs across all the eTreg subsets. Collectively, these data imply two concurrent processes are ongoing to shape the eTreg compartment. One is moving from a lower (cTreg) to a progressively higher (eTreg) state of activation that may be dependent on a common set of external stimuli. TCR signaling and transcriptional regulation by IRF4 represent required signals for production of eTregs (6, 7, 13). These and other niche-related processes then may act in concert to shape the distinctive immune profiles of eTreg subsets. For example, recent TCR signaling and type-1 IFNs favor expression of CD69⁺ (38), and TGF-β upregulates CD103 (39). A clear interrelationship among these eTreg subsets is illustrated by the high degree of overlap of their TCR Vβ repertoires and the dominant clonotypes. This finding is analogous to that reported for Teff subsets (40).

Our data support a model whereby at least some eTregs are derived from cTregs. One key step is the development of CD62L^lo CD69⁻ CD103⁻ eTregs from either Ly-6C⁺ or Ly-6C⁻ cTregs. CD62L^lo CD69⁻ CD103⁻ eTregs might be considered a key transitional step, in which, depending upon the strength of TCR signaling and other niche signals, they develop into CD69⁺ and/or CD103⁺ eTregs. Highly activated eTregs are more likely to be derived from Ly-6C⁻ cTregs due to their higher affinity TCRs. Due to the lower affinity of the TCRs expressed by Ly-6C⁺ Tregs, they are less apt to develop into highly activated Tregs, and instead their development may sometimes be aborted as Ly-6C⁻ cTregs or transitional CD62L^lo CD69⁻ CD103⁻ eTregs, as seen for Tregs expressing the AY02 clonotype. Furthermore, when Ly-6C⁺ or Ly-6C⁻ wild-type cTregs were transferred into Y3 mice, these Tregs uniformly developed into eTregs, but some still expressed high levels of CD62L. Thus, some eTregs, perhaps at the transitional eTreg stage, may support the re-expression of CD62L to promote traffic of some eTregs into secondary lymphoid tissue.

We thank the flow cytometry cores of the Sylvester Comprehensive Cancer Center and the Diabetes Research Institute at the University of Miami for help with FACS analysis and sorting.

The authors have no financial conflicts of interest.