Helios Controls a Limited Subset of Regulatory T Cell Functions Free

A functional immune system is dependent on the maintenance of gene expression, and transcription factors play critical roles in regulating gene expression at specific stages of development. The Ikaros transcription factor family is one such family whose expression is indispensable for immune system development and function. A targeted mutation of the DNA binding domain of Ikaros (Ikzf1) acts in a dominant-negative fashion to functionally delete the entire Ikaros family and results in the lack of all lymphoid lineages, whereas a targeted mutation of the dimerization domain of Ikaros results in the lack of B cells, NK cells, and fetal T cells, as well as reduced numbers of thymic dendritic cells (1, 2). Mice with a similar targeted mutation of the conserved dimerization domain of Aiolos (Ikzf3) exhibit an increase in germinal centers and activated B cells and an elevation in serum IgG and IgE levels; these mice eventually develop B cell lymphomas (3). Aiolos also was shown to inhibit IL-2 expression in Th17 cells (4). Finally, a targeted mutation of Helios (Ikzf2) is neonatal lethal (5). This finding was surprising in light of the fact that Helios expression was originally thought to be confined to a subset of T cells (6, 7) and was later shown to be restricted specifically to regulatory T cells (Tregs). However, mice with a T cell–specific deletion of Helios (crossed to CD4^Cre) appeared to have normal Treg function (8) (A.M. Thornton and E.M. Shevach, unpublished observations).

Tregs that are characterized by the transcription factor Foxp3 suppress immune activation in a dominant manner and play a critical role in the maintenance of self-tolerance. Distinct phenotypic subpopulations of murine Tregs were described recently, broadly dividing Tregs into naive (resting or central, CD44^loCD62L^hiCCR7⁺) and effector (memory, CD44^hiCD62L^loCCR7⁻) subpopulations (9–11). Effector/memory populations of Tregs can be characterized by their ability to express Th lineage–specific transcription factors, in addition to Foxp3, and they can control their corresponding T effector responses. Thus, T-bet, which is required for Th1 differentiation, is also induced in Tregs during a Th1 inflammatory response and is required for Treg control of Th1 responses (12). Similarly, Treg expression of IRF4 is important for suppression of Th2 responses (13), and Treg-specific deletion of STAT3 results in a failure to control Th17 responses (14). This model may be more complex than originally proposed because Treg-specific deletion of GATA3 or T-bet alone had no effect, but the deletion of both resulted in severe autoimmunity characterized by increased cytokines of all subsets (IL-4, IFN-γ, and IL-17) (15, 16). Although the transcription factor Bcl-6 controls T follicular helper (T_FH) cell development and is expressed by Foxp3⁺ T follicular regulatory (T_FR) cells, mice with a Treg-specific deletion of Bcl-6 do not exhibit spontaneous inflammatory disease; instead, they develop enhanced Th2-mediated airway inflammation following immunization (17–20). Bcl-6–deficient Tregs express higher levels of GATA3 compared with wild-type (WT) Tregs, and it appears that Bcl-6 controls the Th2 inflammatory activity of Tregs by repressing GATA3 function (20, 21).

Our previous studies demonstrated that Helios was expressed by 70–80% of mouse and human Foxp3⁺ peripheral Tregs and suggested that expression of Helios allowed the differentiation of thymus-derived Tregs (tTregs) from peripherally induced Tregs (pTregs) (8). Helios was not expressed in Ag-specific Foxp3⁺ T cells induced by Ag feeding or in Tregs induced in germ-free mice following exposure to bacteria (22). However, this finding has been called into question (23–26); the function of Helios and the role it may play in Tregs are still unknown. Although expression of Helios appeared to be Treg specific, our subsequent studies demonstrated that Helios expression is induced in activated T cells following immunization in vivo (A.M. Thornton, J. Galant, and E.M. Shevach, unpublished observations) and that Helios can be expressed by both Th2 cells and T_FH cells in vivo (27). These latter studies raised the possibility that our previous failure to observe a phenotype in mice with a T cell–specific deletion of Helios (Helios^fl/fl × CD4^Cre) may have resulted from the deletion of Helios in both Tregs and conventional CD4 T (Tconv) cells. To definitively examine the function of Helios in Tregs, we generated mice with a Treg-specific deletion of Helios by crossing Helios^fl/fl mice to Foxp3^YFP-Cre mice. These mice developed normally, but exhibited splenomegaly, lymphadenopathy, and lymphocytic infiltrates in nonlymphoid tissues, particularly the salivary glands and liver, with increasing age. Most notably, Tconv cells in Helios^fl/fl × Foxp3^YFP-Cre mice displayed an activated, Th1 phenotype and had lymphoid follicular hyperplasia, increased numbers of germinal centers, and increased serum Ig levels secondary to the failure of T_FR cell function. In vitro Helios-deficient Treg suppressor function was normal, as was their capacity to inhibit the induction of inflammatory bowel disease (IBD) in vivo. However, Helios-deficient Tregs failed to control the expansion of pathogenic T cells derived from scurfy mice in an adoptive-transfer model. In competitive settings, Helios-deficient Tregs also displayed a marked survival disadvantage. Taken together, these studies demonstrate that Helios plays a critical role in controlling certain aspects of Treg-suppressive function, differentiation, and survival.

Mice bearing loxP-flanked (^fl/fl) alleles of Ikzk2 (Helios) on a C57BL/6 background were generated by Ozgene (Bentley Dc, Australia) (8). Ikzf2^fl/fl mice were bred to mice expressing Cre recombinase under control of the Foxp3 promoter (Foxp3-YFP Cre) (Jackson Laboratory, Bar Harbor, ME) to generate Treg-specific Helios-deficient mice. B6.SJL and B6.SJL RAG^−/− mice (expressing the CD45.1 congenic marker) were obtained by the National Institute of Allergy and Infectious Diseases (NIAID) and were maintained by Taconic Farms (Germantown, NY) under contract to NIAID. All animal protocols used in this study were approved by the NIAID Animal Care and Use Committee.

The following staining reagents were used: allophycocyanin anti-CD95 (15A7), PE anti-CD19 (eBio1D3), PE anti–PD-1 (J43), CXCR5 biotin (SPRCL5), allophycocyanin eFluor 780 anti-CD4 (RM4-5), eFluor 450 anti-CD19 (eBio1D3), Alexa Fluor 700 anti-CD44 (IM7), FITC anti-Helios (22F6), PE anti-CD25 (PC61.5), PE anti-CD69 (H1.2F3), PE anti-CD62L (MEL-14), PE anti–IFN-γ (XM61.2), and eFluor 450 anti-CD4 (RM4-5) (all from eBioscience, San Diego, CA). FITC anti-GL7 (GL7), Pacific Blue anti-B220 (RA3-6B2), PE-Cy7 streptavidin, FITC anti-CD4 (RM4-5), PE anti-CXCR3 (CD183) (CXCR3-173), and PE anti-OX40 (OX-86) were purchased from BioLegend (San Diego, CA). FITC anti-CD45Rb (16A), PE anti-CD103 (M290), PE anti–Ki-67, PE anti-CD8a (53-6.7), PE anti-CD25 (7D4), PE anti–Bcl-2, and CD16/32 (24G2) were purchased from BD Biosciences (San Jose, CA). Alexa Fluor 488 anti-GFP was purchased from Life Technologies (Grand Island, NY).

Thymus, spleen, Peyer’s patches, and lymph nodes (LNs) were harvested from mice at the indicated ages. Unless noted, staining was performed using the Foxp3 Staining Buffer Set (eBioscience), according to the manufacturer’s protocol. For cytokine staining, cells were stimulated for 4 h with Cell Stimulation Cocktail (eBioscience) and stained for surface molecules, followed by intracellular staining with BD Cytofix/Cytoperm (BD Biosciences), according to the manufacturer’s protocol. Flow cytometry was performed on an LSR II (BD Biosciences) and analyzed using FlowJo software (TreeStar, Ashland, OR). Staining for YFP was done during the intracellular staining using an anti-GFP Ab (Life Technologies).

Male and female Helios^fl/fl and Helios^fl/fl × Foxp3^YFP-Cre mice were sent to the National Institutes of Health Division of Veterinary Resources to be assessed. Gross necropsies and blood chemistries were performed by a Division of Veterinary Resources pathologist.

Spleen, salivary glands, and liver from Helios^fl/fl and Helios^fl/fl × Foxp3^YFP-Cre mice were sent to American Histo Labs (Gaithersburg, MD) for sectioning and H&E staining. Images were taken in the Biological Imaging Section, NIAID, National Institutes of Health. For histology scores, the degree of infiltrate in liver and lung was determined by two independent scorers and was scored blind.

The Mouse Anti-Nuclear Antigens (ANA) Total Ig kit was purchased from Alpha Diagnostic International (San Antonio, TX). The Mouse anti-dsDNA ELISA kit was purchased from BioVendor Research and Diagnostic Products (Asheville, NC). Mouse IgG1 ELISA Ready-SET-Go!, Mouse IgG2a ELISA Ready-SET-Go!, Mouse IgG2b ELISA Ready-SET-Go!, Mouse IgG3 ELISA Ready-SET-Go!, Mouse IgM ELISA Ready-SET-Go!, Mouse IgA ELISA Ready-SET-Go!, and Mouse IgE ELISA Ready-SET-Go! sets were purchased from eBioscience.

CD4⁺CD45Rb^hi (4 × 10⁵) FACS-sorted cells were transferred alone or with CD4⁺CD25⁺ (2 × 10⁵) FACS-sorted cells from Helios^fl/fl or Helios^fl/fl × Foxp3^YFP-Cre mice to B6.SJL RAG^−/− mice. Mice were monitored for weight loss for the indicated time.

Splenocytes (2 × 10⁶) from 21-d-old male scurfy mice were transferred i.v. to B6 RAG^−/− mice alone or with CD4⁺CD25⁺ (5 × 10⁵) FACS-sorted cells from Helios^fl/fl or Helios^fl/fl × Foxp3^YFP-Cre mice. At 6 wk, splenocytes were analyzed for CD4 and CD8 expression.

For the generation of mixed bone marrow chimeras, recipient female B6.SJL RAG1^−/− mice were sublethally irradiated (550 rad) and reconstituted i.v. with a 1:1 mixture of bone marrow cells from B6.SJL (CD45.1) mice and bone marrow cells from either C57BL/6 (CD45.2) mice or Helios-deficient (CD45.2) mice (for a total of 2 × 10⁶ cells). Mice were analyzed 8 wk after reconstitution.

Spleens were placed in Tissue-Tek optimum cutting temperature compound (Sakura Finetek), snap-frozen in liquid N₂, and stored at −80°C until sectioned by HistoServ (Gaithersburg, MD). Affixed cryostat sections (6 μm thickness) were dried at 25°C, fixed in ice-cold acetone for 10 min, and dried again at 25°C. Slides were rehydrated in 1× TBS (pH 7.6) and placed in a humidifier chamber. Sections were blocked for 2 h at 25°C with IHC/ICC Blocking High Protein Buffer (eBioscience). Sections were stained overnight at 4°C in blocking buffer with anti-IgD (11-26c) PerCP eFluor 710 (eBioscience), anti-CD4 (GK1.5) Alexa Fluor 647 (BioLegend), and anti-GL7 Alexa Fluor 488 (BioLegend). Slides were washed in 1× TBS for 5 min with gentle agitation and mounted with Fluoromount-G with DAPI (eBioscience). Images were visualized and collected using a Leica SP8 inverted five-channel confocal microscope (Leica) and analyzed with Imaris 8.1 (Bitplane, Oxford Instruments).

For primary responses, mice were immunized i.p. with 2 × 10⁸ SRBCs purchased from Lampire Biological Laboratories (Pipersville, PA). Spleens were harvested and analyzed by flow cytometry after 7 d.

The CD4⁺CD25⁻ and CD4⁺CD25⁺ cell populations were sorted using a FACSAria cell sorter. The genomic DNA was isolated using a DNeasy Kit (QIAGEN, Valencia, CA), bisulfite conversion was performed using a EpiTect Bisulfite Kit (QIAGEN), and the samples were prepared for pyrosequencing using PyroMark Gold Q96 Reagents and run on the PyroMark Q96 ID (both from QIAGEN).

CD4⁺CD25⁻ and CD4⁺CD25⁺ cells from LN and spleen were labeled with Pacific Blue anti-CD4 and PE anti-CD25 and sorted using a FACSAria cell sorter. Alternatively, CD4⁺YFP⁻ cells from Foxp3^Cre mice and CD4⁺YFP⁺ Tregs from Foxp3^Cre mice or heterozygous Helios^fl/fl × Foxp3^Cre/+ mice were sorted using a FACSAria cell sorter. Suppression assays were performed as previously described (8).

All data are mean ± SD. Comparisons between groups were analyzed using unpaired Student t tests (Prism GraphPad). Statistical significance was established at p < 0.05.

To elucidate the role of Helios, we generated a mouse with a Treg-specific deletion of Helios by breeding Helios^fl/fl mice that contain loxP sites flanking exon 8 (8) to Foxp3^YFP-Cre mice (28). Although the percentage of Foxp3⁺ T cells in the offspring was normal, expression of Helios could not be detected in Foxp3⁺ T cells by flow cytometry (Fig. 1A). It should also be noted that the small number of Foxp3⁻Helios⁺ T cells present in C57BL/6 mice (8) and in Helios^fl/fl mice (Fig. 1A) was unchanged in Helios^fl/fl × Foxp3^Cre mice. It was recently reported that some conditional alleles are subject to promiscuous expression of the Foxp3^YFP-Cre allele with deletion of the allele in Foxp3⁻CD4⁺ T cells (29). However, CD4⁺Foxp3⁻ cells in these Helios^fl/fl × Foxp3^Cre mice do not express YFP, thus confirming the specificity and fidelity of the Treg-specific Helios deletion.

At 6 mo of age, discernable abnormalities were observed in Helios^fl/fl × Foxp3^Cre mice. A marked lymphadenopathy, specifically in the inguinal, brachial, and axillary LNs, was noted (Fig. 1B), along with splenomegaly (Fig. 1C). The absolute total cell numbers were increased in all lymphoid organs, whereas CD4⁺ T cells and Tregs were increased in most lymphoid organs (Fig. 1D). Interestingly, Helios^fl/fl × Foxp3^Cre mice exhibited a greatly reduced amount of s.c. and visceral adipose tissue, which was accompanied by significant weight loss (Fig. 1B, Supplemental Fig. 1A, 1B). Notably, these mice also displayed signs of hepatic lipidosis and marked enlargement of the liver, with increased levels of alkaline phosphatase and lactate dehydrogenase present in the serum (Supplemental Fig. 1A, 1C). Perivascular lymphoid aggregates were observed in several organs, including the submandibular salivary gland and liver, of Helios-deficient mice (Fig. 2A, 2B). Older mice (>9 mo) frequently presented with distended abdomens due to enlargement of the liver. Serum Ig levels were also measured, and elevations in IgE, IgA, IgM, and IgG2a were observed in Helios-deficient mice, whereas a decrease in IgG3 was noted (Fig. 2C). Finally, anti-nuclear Abs could also be detected in the sera of Helios-deficient mice; although there was a trend toward higher levels of anti-dsDNA Abs, it was not statistically significant (Fig. 2D).

To further investigate the phenotype of the T cells from Helios^fl/fl × Foxp3^Cre mice, we examined a variety of activation markers. Tconv cells from 2-mo-old Helios^fl/fl × Foxp3^Cre mice showed slightly elevated expression of activation markers compared with either Foxp3^Cre or Helios^fl/fl mice (data not shown). However, striking differences between Tconv cells from Helios^fl/fl × Foxp3^Cre mice and both Foxp3^Cre and Helios^fl/fl mice were observed at 6 mo. Tconv cells from Helios^fl/fl × Foxp3^Cre mice exhibited a markedly activated phenotype, with significantly elevated percentages of CD44^hiCD62L^lo, PD-1, Nrp-1, ICOS, and CD69 cells (Fig. 3A). Moreover, Tconv cells from Helios^fl/fl × Foxp3^Cre mice displayed a Th1 phenotype, with elevated expression of CXCR3 and an increased capacity to secrete IFN-γ compared with cells from control mice, concurrent with increased expression of T-bet (Fig. 3B, 3C). Excess production of other cytokines (IL-4 and IL-17) was not observed (data not shown). In parallel, CD8⁺ T cells from Helios^fl/fl × Foxp3^Cre mice displayed modest changes at 2 mo (data not shown) but displayed a significant increase in total cell number (Supplemental Fig. 2A). CD8⁺ T cells from Helios^fl/fl × Foxp3^Cre mice also displayed increased activation, as measured by CD44^hiCD62L^lo expression, and a significant increase in IFN-γ production at 6 mo (Supplemental Fig. 2B).

To explain this Th1 dysregulation, we examined Tregs from 6-mo-old Helios^fl/fl × Foxp3^Cre mice. These cells exhibited a slightly activated phenotype, with higher percentages of CD44^hiCD62L^lo cells and elevated expression of Nrp-1, ICOS, and CD69; however, CD25 expression was unchanged (Fig. 4A). Tregs from Helios^fl/fl × Foxp3^Cre mice also had elevated expression of CXCR3 and, more importantly, had an increased capacity to secrete IFN-γ compared with cells from control mice (Fig. 4B), suggesting that Th1-like Tregs were less functional and unable to control Th1 responses. Significantly, Tregs that secreted IFN-γ also expressed T-bet (Fig. 4C). Taken together, a Treg-specific deletion of Helios results in the slow development of a complex systemic autoimmune disease with a Th1 phenotype, which is accompanied by hepatosplenomegaly, hepatic lipidosis, and a loss of s.c. and visceral fat.

Although the absolute number of Tregs in 6-mo-old Helios^fl/fl × Foxp3^Cre mice was elevated, it remained possible that the composition of the peripheral T cell pool was altered in the absence of Helios, with a disordered ratio of tTregs/pTregs. Selective demethylation of an evolutionarily conserved element within the Foxp3 locus, the TSDR, was shown to be critical for the regulation of Treg stability, and pTregs may display less stable Foxp3 expression (30). To examine Treg stability in Helios^fl/fl × Foxp3^Cre mice, we sorted CD4⁺CD25⁺ and CD4⁺CD25⁻ populations from Helios^fl/fl and Helios^fl/fl × Foxp3^Cre mice and examined methylation of the TSDR. In both control and Helios-deficient Treg populations, the TSDR was equally demethylated at both 2 and 6 mo of age (Fig. 5A, data not shown). Thus, Helios does not directly control the stability of Foxp3 expression.

We next tested Tregs from Helios^fl/fl × Foxp3^Cre mice in an in vitro suppression assay. First, Helios-deficient Tregs were nonresponsive to stimulation with anti-CD3 but proliferated in a manner similar to control Tregs to TCR stimulation in the presence of IL-2. In the in vitro suppression assay, control and Helios-deficient CD4⁺CD25⁺ Tregs equally suppressed the proliferative responses of CD4⁺CD25⁻ responder cells from both control and Helios^fl/fl × Foxp3^Cre mice (data not shown). Because Tregs in Helios^fl/fl × Foxp3^Cre mice have a modestly activated phenotype, and activated Tregs are more suppressive, we wished to examine Helios-deficient Tregs that developed under more physiologic conditions. We crossed female Helios^fl/fl × Foxp3^Cre/Cre mice to male Helios^fl/fl × Foxp3^wt/wt mice to produce female heterozygous Helios^fl/fl × Foxp3^Cre/+ mice. Because Foxp3 is X-linked, and genes on the X chromosome are subject to random inactivation of one allele, female mice that are heterozygous for YFP-Cre should have 50% of Tregs that use the Foxp3^wt allele and express Helios and 50% of Tregs that use the Foxp3^Cre allele and, thus, do not express Helios and are marked by YFP. Because these heterozygous female mice possess a mixture of WT Tregs and Helios-deficient Tregs, they do not develop the pathology described in Helios-deficient mice. When assayed, YFP⁺ Tregs from Helios^fl/fl × Foxp3^Cre/+ mice were nonresponsive to stimulation with anti-CD3, but they proliferated to TCR stimulation in the presence of IL-2 in a manner similar to control Tregs (Fig. 5B). In the in vitro suppression assay, YFP⁺ Tregs from Foxp3^Cre and Helios^fl/fl × Foxp3^Cre/+ mice equally suppressed the proliferative responses of CD4⁺YFP⁻ responder cells (Fig. 5C).

We also compared the suppressive capacity of control and Helios-deficient Tregs to inhibit T cell activation in vivo. We transferred CD4⁺CD45Rb^hi naive cells alone into RAG^−/− recipients or cotransferred control or Helios-deficient Tregs and examined their capacity to inhibit the development of IBD (Fig. 5D). Control and Helios-deficient Tregs equally inhibited disease induction. To evaluate the suppressive capacity of Helios-deficient Tregs under more inflammatory conditions, we transferred splenocytes from a premoribund scurfy mouse (containing highly activated T cells) to RAG^−/− mice and cotransferred Helios-deficient or control Tregs (Fig. 5E). At 6 wk posttransfer, control Tregs, but not Helios-deficient Tregs, markedly suppressed the expansion of scurfy CD4⁺ T cells, and both control and Helios-deficient Tregs suppressed the expansion of scurfy CD8⁺ T cells. Transfer of scurfy splenocytes results in severe lymphocytic infiltration in the lung and liver (Fig. 5F). Cotransfer of control Tregs was protective in both the lung and liver, but Helios-deficient Tregs were only partially protective. These results suggest that Helios regulates some, but not all, of the Treg-suppressive mechanisms.

We next generated mixed bone marrow chimeras to assess the fitness of Helios-deficient Tregs in a competitive environment. After 8 wk of reconstitution, the thymus and spleen were harvested. In WT/WT mice, reconstitution in the thymus and spleen was roughly equal, as expected (data not shown). In the thymus of WT/Helios-deficient mice, reconstitution of CD4⁺Foxp3⁻ and CD4⁺Foxp3⁺ cells was roughly equal, demonstrating that thymic development of Helios-deficient Tregs is not impaired by the absence of Helios (Fig. 6A). However, in the spleen there was a striking decrease in the ratio of WT/Helios-deficient Tregs, suggesting that, although Helios-deficient Tregs were generated in the thymus, once in the periphery, their survival was greatly impaired compared with control Tregs (Fig. 6A, 6B).

Because studies demonstrated that Tregs can be divided into two distinct subpopulations based on their differential expression of CD44 and CD62L (9–11, 31, 32), we examined CD44 and CD62L expression in the mixed chimeric mice. We found that Helios-deficient Tregs were deficient in cells of the activated/effector phenotype and were preferentially of the naive/central phenotype (CD62L^hiCD44^lo) (Fig. 6C). The phenotype of Helios-deficient Tregs in this chimeric environment must be contrasted with their phenotype in the parental Helios^fl/fl × Foxp3^Cre mouse, in which the majority of Tregs exhibit an activated effector phenotype. These data suggest that, in a competitive setting, Helios-deficient Tregs fail to differentiate to effector Tregs or, alternatively, die during the transition from naive to an activated state or are unstable in the activated state.

To further examine the phenotype of Helios-deficient Tregs under more physiologic conditions, we further analyzed heterozygous female Helios^fl/fl × Foxp3^Cre/+ mice, because they do not develop the pathology described in Helios^fl/fl × Foxp3^Cre mice; in addition, unlike the bone marrow chimeric model that involves radiation of the RAG2^−/− recipient, Tregs develop in a normal environment. Using YFP expression as a marker of Helios-deficient Tregs, we confirmed that the thymus contains a similar percentage of YFP⁺ Helios-deficient Tregs and YFP⁻ WT Tregs; thus, the thymic development of YFP⁺ Helios-deficient Tregs does not appear to be impaired (Fig. 7A). However, in secondary lymphoid tissue, there is a reduction in YFP⁺ Helios-deficient Tregs to ∼35% of total Tregs as early as 2 mo of age (data not shown). These differences were more pronounced at 5 mo, with YFP⁺ Helios-deficient Tregs further reduced to only 23% of the total Treg population (Fig. 7A, 7B). A comparison of the phenotype of YFP⁺ Helios-deficient and YFP⁻ WT Tregs revealed that YFP⁺ Helios-deficient Tregs preferentially expressed a more naive phenotype, with reduced levels of activation markers (CD44/CD62L, OX40, CD69) (Fig. 7C), in agreement with the data from bone marrow chimeras. Furthermore, CD25 expression on YFP⁺ Helios-deficient Tregs in heterozygous mice was increased, and the proliferation rate, as measured by the cell cycle marker Ki67, was decreased, again consistent with a naive Treg phenotype (Fig. 7C). Finally, we examined the expression of IFN-γ and T-bet. In the absence of an inflammatory environment in female heterozygous Helios^fl/fl × Foxp3^Cre/+ mice, YFP⁺ Helios-deficient Tregs and YFP⁻ Tregs had comparable percentages of T-bet⁺IFNγ⁺ cells within the same mouse (Supplemental Fig. 3). Thus, the instability of Helios-deficient Tregs is a consequence of the activated environment. Together, these data suggest that, in a competitive setting, Helios-deficient Tregs are unable to differentiate or survive, particularly in an activated, effector state.

We did not find any significant difference in the levels of expression of a broad array of pro- and antiapoptotic genes between thymic or peripheral YFP⁺ and YFP⁻ Tregs in heterozygous female mice (data not shown). However, unstimulated Tregs from Helios^fl/fl × Foxp3^Cre mice expressed lower levels of Bcl-2 compared with Helios^fl/fl mice (Fig. 7D), but no differences were seen with regard to the mRNA expression of Bim, Bcl-x_L, Bad, Bid, Bax, or Mcl-1 (data not shown). Upon activation, Bcl-2 expression was further decreased in Tregs from Helios-deficient mice (Fig. 7D). The decreased expression of Bcl-2 in Helios-deficient Tregs was paradoxical, because the total Treg population is enhanced in Helios^fl/fl × Foxp3^Cre mice (Fig. 1D). When we correlated Ki-67 staining with CD44 expression in control and Helios-deficient Tregs, we noted that CD44^hi Helios-deficient Tregs had decreased Ki-67 compared with control CD44^hi Tregs, whereas CD44^lo Helios-deficient Tregs had a higher level of Ki-67 than Helios^fl/fl CD44^lo Tregs (Fig. 7E). Thus, it appears that, in both competitive and noncompetitive conditions, Helios-deficient CD44^hi Tregs are at a competitive disadvantage. In Helios^fl/fl × Foxp3^Cre mice, this deficiency is compensated for by enhanced expansion of the CD44^lo population.

Given the increased spleen size and the pan-hypergammaglobulinemia in Helios^fl/fl × Foxp3^Cre mice (Figs. 1C, 2C), we stained spleen sections to assess the presence of germinal centers. Helios^fl/fl × Foxp3^Cre mice possess more follicles and strikingly increased germinal centers (Fig. 8A). This suggests that Helios might play a role in the differentiation of T_FR cells, specialized Tregs that control T_FH cells and germinal center responses (17–19). We first compared the presence of T_FH cells and T_FR cells in unimmunized control and Helios^fl/fl × Foxp3^Cre mice (Fig. 8B). Unimmunized control mice have few follicular T cells, based on CXCR5⁺ and PD-1^hi expression, but ∼45% are Foxp3⁺ T_FR cells. In marked contrast, unimmunized Helios^fl/fl × Foxp3^Cre mice have an enhanced percentage of PD-1⁺CXCR5⁺ follicular T cells, but very few are Foxp3⁺ T_FR cells. In addition, Helios^fl/fl × Foxp3^Cre mice have a higher percentage of germinal center B cells (Fig. 8C). Upon immunization with SRBCs, we observed an increase in T follicular cells in control mice but no change in the already high percentage of T follicular cells and germinal center B cells in Helios^fl/fl × Foxp3^Cre mice (Fig. 8D, 8E). Because the percentage of T_FR cells was drastically reduced in Helios^fl/fl × Foxp3^Cre mice, this suggests that, in the absence of Helios, Tregs are unable to differentiate into T_FR cells. However, as a result of the increased cellularity of the spleen and the increased percentage of T follicular cells, calculation of the absolute number of T_FR cells showed that, in fact, they are present and increased in Helios^fl/fl × Foxp3^Cre mice (Fig. 8F). Calculation of the absolute number of T_FH cells showed a 75-fold increase in Helios^fl/fl × Foxp3^Cre mice.

We further examined Tregs for the presence of T_FR cells in heterozygous Helios^fl/fl × Foxp3^Cre/+ female mice. Gating on CD4⁺ Foxp3⁺ cells from 2-mo-old heterozygous mice, the YFP⁺ Helios-deficient T_FR population in heterozygous mice was totally absent (Fig. 8G). Similar results were observed in 5-mo-old heterozygous mice (data not shown). Upon immunization with SRBCs, we observed an increase in T follicular cells in heterozygous Helios^fl/fl × Foxp3^Cre/+ female mice, but no T_FR cells were derived from the YFP⁺ Helios-deficient population (Fig. 8H).

We found that selective deletion of Helios in Foxp3⁺ Tregs results in the development of a systemic autoimmune disease in mice beginning at 2 mo of age. At 6 mo of age, Helios^fl/fl × Foxp3^Cre mice exhibit generalized splenomegaly, lymphadenopathy, lymphocyte activation, expansion of Th1 T effector cells, hypergammaglobulinemia, and increased size/numbers of lymphoid follicles and germinal centers. Our results suggest that Helios plays a critical role in controlling Treg-mediated regulation of Th1 and T_FH effector functions. In addition, Helios^fl/fl × Foxp3^Cre mice have a marked abnormality of lipid metabolism characterized by a complete absence of s.c. and visceral fat and markedly enlarged livers with lipidosis. None of these abnormalities was observed in aged mice with a deletion of Helios in mature T cells (CD4-Cre) or in T cell precursors (Vav-Cre). Therefore, it is likely that Helios plays an important functional role in Tconv cells and that loss of Helios expression in Tconv masks the effect of loss of Helios in Tregs. Studies are underway to elucidate the role that Helios may play in CD4⁺Foxp3⁻ cells.

The phenotype of the Tregs in Helios^fl/fl × Foxp3^Cre mice differed from the Treg phenotype observed in both the mixed bone marrow chimeras and heterozygous Helios^fl/fl × Foxp3^Cre/+ mice. The former expressed an activated phenotype, whereas Helios-deficient Tregs in the competitive environments had a predominantly naive phenotype. It is likely that, in the noncompetitive setting, Helios-deficient Tregs are continuously activated and are attempting to control the Th1 T effector and T_FH responses. In competitive settings, mice are phenotypically normal, and Helios-sufficient Tregs effectively control T effector cell function. Therefore, it is unlikely that Helios controls the differentiation of naive Tregs to T effector/memory Tregs, because the Tregs in Helios^fl/fl × Foxp3^Cre mice express an activated phenotype, whereas the failure to observe activated Helios-deficient Tregs in the competitive environments is consistent with a role for Helios in controlling Treg fitness. Although the absolute numbers of Tregs in Helios^fl/fl × Foxp3^Cre mice were elevated, they expressed reduced levels of Bcl-2 and decreased percentages of Ki-67⁺CD44^hi cells. This is also consistent with the decrease in the survival of activated/effector Helios-deficient Tregs in heterozygous females and bone marrow chimeras. Although Bcl-2–deficient Tregs do not appear to have any defects in survival in mixed bone marrow chimeras generated from mice with a global deletion of Bcl-2 (33), survival of Tregs in mice with a Treg-specific deletion of Bcl-2 has not been reported, and the role of Bcl-2 in Helios-deficient Tregs needs to be examined further. It is likely that the enhanced numbers of Tregs in Helios^fl/fl × Foxp3^Cre mice are secondary to enhanced proliferation of the Helios-deficient CD44^lo population or increased thymic output.

One other problem in interpreting our results, particularly in the competitive environment models, is that Foxp3-Cre is expressed in all Foxp3⁺ cells, including the 30% of peripheral Tregs that do not express Helios. We postulated that Foxp3⁺Helios⁻ Tregs are pTregs, but several studies (34) raised questions about this finding. Nevertheless, it is likely that a majority (∼60%) of Foxp3⁺Helios⁻ Tregs are pTregs. Tregs in the Helios-deficient mouse may still be a heterogeneous population that consists of the typical 30% Helios⁻ cells and 70% Helios positive cells that no longer express Helios. Alternatively, Tregs that normally express Helios may not survive in a competitive environment once Helios has been deleted, and the remaining Treg population may consist exclusively of Helios⁻ cells. At this time, it is difficult to rule out either possibility.

Helios appears to regulate two distinct suppressor-effector functions of the Foxp3⁺ T cell population. Helios-deficient Tregs were as efficient as control Tregs in suppressing the activation of naive CD4⁺ T cells in vitro and were as competent as control Tregs in preventing the induction of IBD induced by transfer of CD45Rb^hi cells to immunodeficient recipients. Although Helios^fl/fl × Foxp3^Cre mice have an expanded pool of Tregs that express CXCR3, IFN-γ, and T-bet, Th1 responses were not controlled. Surprisingly, the marked increase in IFN-γ–producing T effector cells (both CD4⁺ and CD8⁺) was not accompanied by pathologic evidence of organ-specific autoimmunity. The failure to control Th1 T effector function was also clearly documented when we transferred T effector cells (primarily Th1 cells) from moribund scurfy mice to immunodeficient mice. Control, but not Helios-deficient, Tregs were able to control expansion of the CD4⁺ T effector cells. Helios-deficient Tregs controlled the expansion of CD8⁺ scurfy T effector cells, suggesting that Tregs may use different mechanisms to suppress CD4⁺ and CD8⁺ T effector cells.

Helios-deficient Tregs were also incapable of controlling T_FH cell function, because the most prominent feature of the phenotype of Helios^fl/fl × Foxp3^Cre mice was a marked increase in the size and number of lymphoid follicles and germinal centers, as well as hypergammaglobulinemia. Abnormal lymphoid follicle formation was also observed in the salivary glands. Although the ratio of T_FR/T_FH cells was markedly reduced in Helios^fl/fl × Foxp3^Cre mice, the absolute number of T_FR cells in the spleen was normal or increased secondary to the splenomegaly. Furthermore, the levels of Bcl-6 in Helios-deficient Tregs are comparable to those of control Tregs (data not shown). Thus, it appears that T_FR cells, which express Bcl-6, can be generated in the absence of Helios, but their suppressor functions are compromised. Because T_FR cells are thought to arise from tTregs, not pTregs (17, 19), this result is consistent with the preferential expression of Helios in tTregs. It also suggests that the normal 30% of Tregs that are CD4⁺Foxp3⁺Helios⁻ are unable to differentiate into fully functional T_FR cells. The defect in T_FR cell function in Helios^fl/fl × Foxp3^Cre mice resembles, in part, the defect in T_FR cell function seen in mice with a Treg-specific deletion of TRAF3 (35). TRAF3^−/− mice exhibited milder signs of generalized dysregulated Treg dysfunction (e.g., elevated Th1 cells in peripheral tissues). The changes in T follicular cell responses (elevated T_FH and decreased T_FR cells) were primarily seen after immunization. The defect in T_FR cell function in these mice appeared to be related to a decreased expression of ICOS. In marked contrast to TRAF3^−/− Tregs, Helios-deficient Tregs isolated from the spleens of 6-mo-old mice expressed very high levels of ICOS (Fig. 4A). Further studies of the T_FR cell population in Helios^fl/fl × Foxp3^Cre mice should be helpful in determining what specific suppressor mechanisms are used by T_FR cells.

Obesity, insulin resistance, diabetes, and nonalcoholic fatty liver disease (NAFLD) are associated with inflammation in adipose tissue that is accompanied by proinflammatory M1 macrophages and TNF-α (36). Serum levels of TNF-α were not elevated in Helios^fl/fl × Foxp3^Cre mice with NAFLD (M. Sebastian and A.M. Thornton, unpublished observations). More recently, another subpopulation of effector Tregs in adipose tissue was described (37, 38). Adipose tissue–specific Tregs are characterized by expression of the transcription factor PPARγ and control adipose tissue inflammation via IL-10. At present, it is difficult to link the development of the severe NAFLD seen in Helios^fl/fl × Foxp3^Cre mice to the Tregs that are normally resident in fat, because Helios^fl/fl × Foxp3^Cre mice completely lack adipose tissue and are not obese. NAFLD in Helios^fl/fl × Foxp3^Cre mice is only seen in mice 6 mo of age and older, and its severity is of variable penetrance. It remains possible that this phenotype is not a direct effect of loss of Helios in Tregs but an indirect phenomenon, perhaps secondary to the development of autoantibodies to a receptor involved in lipid metabolism. Further studies of the potential role that Helios plays in adipose tissue–specific Tregs are under investigation. It will also be of interest to determine whether the NAFLD phenotype can be transferred by T cells or by Ab.

Taken together, our studies indicate that Helios plays a complex role in Treg function and probably in the function of Tconv cells. The major effect of the loss of Helios expression in Tregs appears to be disruption of T_FR cell function, with the slow development of a systemic autoimmune disease in the absence of immunization or pathogen challenge. It also remains possible that the activated Th1 phenotype is secondary to the failure of the Helios-deficient Tregs to control T_FH cell function and that the activated Th1 phenotype is secondary to abnormal differentiation of the uncontrolled T_FH cell population. It was claimed that Helios promotes binding of Foxp3 to the IL-2 promoter and, thereby, controls IL-2 expression by Tregs (39). Short hairpin RNA–mediated downregulation of Helios resulted in IL-2 production by Tregs and loss of Treg suppressive capacity in vitro and in vivo in the IBD model. It is difficult to reconcile our findings that specific deletion of Helios in Tregs had no effect on Treg suppressive function in vitro or in vivo in the IBD model. It remains possible that the differences in the studies are secondary to redundant mechanisms mediated by other members of the Ikaros gene family, but it is also possible that the short hairpin RNA approach induced expression of dominant-negative proteins that interfered with other members of the Ikaros family. Ikaros itself binds to the IL-2 promoter and represses IL-2 gene expression. Although Baine et al. (39) claimed that Helios bound to the IL-2 promoter, using chromatin immunoprecipitation assays, it is difficult to interpret their results because the polyclonal anti-Helios Ab used in their studies was reactive with the C-terminal domain of Helios, which is conserved among Ikaros family members. Ikaros, but not Helios, was identified as 1 of the 361 proteins associated with Foxp3 in mass spectrometric analyses of Foxp3 complexes (40). Helios also was not identified as one of the critical transcription factors involved in the regulation of the Treg gene signature (41). Thus, Helios appears to play a niche role in the spectrum of Treg suppressive functions (42, 43), and further analysis of its specific target requires a focus on T_FR cell suppressive functions.

During the review of this article, Kim et al. (44) reported that Helios plays a critical role in Treg function and is required for the stable inhibitory activity of Tregs due to decreased activation of the STAT5 pathway. Most of their experiments used mice that have a global deficiency in Helios, and it is difficult to compare their results with our studies in mice with a conditional deletion of Helios in Tregs. Kim et al. (44) observed autoimmunity in aged Helios^−/− mice and in bone marrow chimeras in which CD4 or CD8 cells were deficient for Helios. However, we never observed any signs of Treg dysfunction, T cell activation, or autoimmune disease in aged Helios^fl/fl × CD4^Cre mice or Helios^fl/fl × Vav^Cre mice. Similar to our results, Kim et al. (44) observed significant T cell activation in 6-mo-old Helios^fl/fl × Foxp3^Cre mice, but they also reported manifestations of autoimmune pathology in liver, lung, and pancreas that we did not observe. Although lymphoid infiltrates were present in the salivary glands and liver in our mice, we did not detect any signs of autoimmune-mediated organ pathology. They also reported a defect in the ability of Tregs from Helios^fl/fl × Foxp3^Cre mice to protect in the IBD model that we did not find, but this may be secondary to differences in the microbiome in the colonies.

One of the most interesting findings reported by Kim et al. (44) was that STAT5b was a target for Helios binding and that reduced STAT5 activation was responsible for reduced expression of Foxp3 in Helios^fl/fl × Foxp3^Cre mice. They conclude that this is consistent with the contribution of STAT5 for Foxp3 stability. We also noted reduced Foxp3 expression in Helios^fl/fl × Foxp3^Cre mice and in the Helios-deficient cells in female Helios^fl/fl × Foxp3^Cre/+ mice. However, we did not observe any differences in STAT5 activation (data not shown). It remains possible that the reduction in Foxp3 expression that they observed represents a hypomorphic feature of the Foxp3^YFP-Cre allele, as previously reported (29).

Lastly, one of the major differences between the two studies is that severe autoimmune disease was primarily observed when Kim et al. (44) generated bone marrow chimeras from mice with either global or Treg-specific deletions of Helios. We also observed severe autoimmune disease in bone marrow chimeras generated from Helios^fl/fl × Foxp3^Cre mice (data not shown). It is likely that the thymic generation of Tregs in bone marrow chimeras (generated in irradiated RAG^−/− adult mice) differs markedly from the normal neonatal selection of Tregs (45). Helios is expressed very early in the thymic development of Foxp3⁺ Tregs, and a deficiency of Helios may contribute to the alterations in Treg function observed in bone marrow chimeras.

Abnormalities in Treg function have been noted in almost all human autoimmune diseases, based largely on decreased suppressor function in the standard Treg-suppression assay in vitro (46). However, deletion of the expression of several transcription factors (including Helios) that are critical for Treg function in the mouse, although precipitating autoimmune disease, may not result in abnormalities in Treg suppressor function in vitro (35). Because the major manifestation in Helios^fl/fl × Foxp3^Cre mice is defective T_FR cell function, resulting in a disease resembling systemic lupus erythematosus, it would be highly desirable to develop in vitro assays of T_FR cell function for humans that might be useful both diagnostically and to monitor therapy.

We thank Deborah D. Glass for technical help, Hideyuki Ujiie for assistance with the scurfy transfer model, and Sundar Ganesan for assistance with immunohistochemistry and imaging.

The authors have no financial conflicts of interest.