Murine regulatory T cells contain hyper-proliferative and death-prone subsets with differential ICOS expression

Abstract

Regulatory T cells (Treg) are crucial for self-tolerance. It has been an enigma that Treg exhibit an anergic phenotype reflected by hypo-proliferation in vitro following T cell receptor (TCR) stimulation but undergo vigorous proliferation in vivo. We report here that, different from conventional T cells (Tcon), murine Treg are prone to death but hyper-proliferative in vitro and in vivo. During in vitro culture, most Treg die with or without TCR stimulation, correlated with constitutive activation of the intrinsic death pathway. However, a small portion of the Treg population is more sensitive to TCR stimulation, particularly weak stimulation, proliferates more vigorously than CD4⁺ Tcon, and are resistant to activation induced cell death. Treg proliferation is enhanced by IL-2 but less dependent on CD28-mediated costimulation than Tcon. We demonstrate further that the surviving and proliferative Treg are ICOS positive while the death-prone Treg are ICOS negative. Moreover, ICOS⁺ Treg contain much stronger suppressive activity than ICOS⁻ Treg. Our data indicates that massive death contributes to the anergic phenotype of Treg in vitro and suggest modulating Treg survival as a therapeutic strategy for treatment of autoimmune diseases and cancer.

Introduction

Regulatory T cells (Treg) actively suppress autoimmunity and play critical roles for self-tolerance. Natural Treg are generated in the thymus and are governed by the forkhead transcription factor Foxp3. Expression of Foxp3 is critical for Treg generation, maintenance, and function (1–5). Foxp3⁺ Treg are composed of a small portion of T cells with distinct molecular patterns and properties. Different from naïve conventional CD4⁺ Foxp3⁻ T cells (Tcon), Foxp3⁺ Treg express signature molecules such as CD25, GITR, and CTLA4 that are usually expressed by activated Tcon (6–8). However, Treg do not express IL-2 or IFNγ. Instead, they express suppressive cytokines such as IL-10 and TGFβ (9). Moreover, Treg have been marked as `anergic' in vitro and in vivo. In vitro, Treg have been found unable to proliferate following TCR engagement. In vivo, adoptively transferred CD4⁺CD25⁺ Treg that recognize ovalbumin display decreased proliferative responses to antigen challenge in mice (10–13). While these results point to an anergic property of Treg, Treg manifest normal or even stronger homeostatic proliferation after injected into lymphopenic hosts or T cell receptor-induced proliferation in vivo (11, 14). Furthermore, Treg incorporate a higher level of BrdU than Tcon following administration of BrdU into mice, suggesting that Treg are more proliferative than Tcon in normal mice (15). The reasons that lead to these distinct observations and conclusions about Treg properties have been unclear. By tracking Treg proliferation and death simultaneously, we report here that CD4⁺Foxp3⁺ Treg are hyper-proliferative but prone to death in vitro and in vivo as compared to CD4⁺Foxp3⁻ Tcon. During in vitro culture with a TCR stimulating antibody, the majority of Treg die before or during the first division. The remaining surviving subset of Treg is hyper-proliferative to TCR stimulation, particularly when the stimulation is weak. Treg proliferation is enhanced by IL-2 and is less dependent on CD28 costimulation than Tcon. In addition, freshly isolated Treg show enhanced death and proliferation as compared to Tcon. While proliferative Tcon are sensitive to activation induced death, the hyper-proliferative Treg are resistant to activation induced death. Lastly, we demonstrate that the living hyperproliferative Treg are ICOS⁺ and the death prone Treg are ICOS⁻ and that ICOS⁺ Treg contain much stronger suppressive activity than ICOS⁻ Treg. Together, our data indicate that Treg should not be simply defined as anergic. Rather, they contain a predominant ICOS⁻ death-prone and a minor ICOS⁺ hyper-proliferative subsets. Our data provide an explanation for the aforementioned contradictory properties of Treg and suggest that promoting Treg survival could be used as an effective strategy to enhance immune suppression and tolerance.

Methods

Mice

The C57BL6/J, Foxp3GFP (16), CD28^−/−, and B6.PL-Thy1a/CyJ congenic mice were all purchased from the Jackson Laboratory. Thy1.1-Foxp3GFP mice were generated by breeding B6.PL-Thy1a/CyJ mice with Foxp3GFP mice. All mice were housed in an approved pathogen free facility and experiments were performed in accordance with protocols approved by the Duke University Institutional Animal Care and Use Committee.

Flow Cytometry

Single cell suspensions of spleen and LN were stained with antibodies in PBS containing 2% FCS. The antibodies include PE-Cy7-anti-CD4 (RM4–5), APC-Cy7-anti-CD8α (53–6.7), PE-Cy7-anti-ICOS (398.4A), PE- or APC-conjugated anti-CD90.1 (HIS51), PE- or APC-anti-CD90.2 (30-H12) (BD Biosciences). Biotin-conjugated antibodies were developed with PE-, PE-Cy5 or APC-conjugated streptavidin. For intracellular staining, cells were permeabilized using the Foxp3 staining kit (eBioscience) after cell surface staining, followed by PE-anti-Foxp3 (FJK-16s, eBioscience) or unconjugated anti-Ki67 (B56, BD Biosciences). An Alexa Fluor 488-conjugated goat anti-mouse IgG (H+L) (Invitrogen) was used to detect anti-Ki67 antibody. Dying cells were identified using 7AAD or the fixable Violet Live/Dead cells staining kit (Invitrogen). APC-conjugated Annexin V (BD Biosciences) was used to detect apoptotic cell. Stained samples were collected on a Canto II flow cytometer (BD Biosciences). Data were analyzed using FlowJo software (TreeStar, Inc.).

Treg isolation

CD4 T cells were enriched using the APC selection kit (Stem Cell Technologies) or MACS sorting wiht the LD column (Miltenyi Biotec) according to manufacturers' protocol with about 80% purity. The isolated CD4 cells were stained for CD4, CD25, ICOS, 7-AAD, Thy1.1, and Thy1.2 when needed and sorted for live CD4⁺GFP⁻ and CD4⁺GFP⁺ cells. The purities of sorting cells is usually great than 95%.

Proliferation Assay

CFSE proliferation assay was done as previously described (17). Single cell suspension was made from spleens and LNs of mice in IMDM supplemented with 10% FBS, penicillin/streptomycin and glutamine (IMDM-10). Cells were stained with 10 μM CFSE for 9 minutes at room temperature in PBS-0.5% BSA. Labeled cells were left unstimulated or stimulated with an anti-CD3ε antibody (2C-11) at the indicated concentrations at 37°C for different times in a 5% CO₂ incubator. Where indicated, IL-2 (1,000 U/ml, Peprotech) or CTLA4Ig (10 μg/ml, BioXcell) were added at the beginning of the culture. When cultured with IL-2, CD8⁺ cells were depleted using anti-CD8 microbeads and LD Columns according to the manufacturer's protocol (Miltenyi Biotec).

Treg contact inhibition assay

Sorted Thy1.1⁺Thy1.2⁺ Tcon were labeled with CFSE. Splenocytes from TCR β^−/−δ^−/− mice were treated with 25 μg/ml of mitomycin C for 30 min and used as antigen-presenting cells (APCs). In a 96-well U-bottom plate, 1.5 × 10⁵ CFSE-labeled Tcon, the same number of APCs, and different numbers of sorted Thy1.2⁺ ICOS⁺ or ICOS⁻ Treg according to the indicated ratios were added to each well. The cells were left unstimulated or stimulated with an anti-CD3ε (2C11) antibody (1 μg/ml). After 72 h cultures, the cells were stained with PE-Cy7-anti-Thy1.1, APC-anti-Thy1.2, and APC-Cy7-anti-CD4 and analyzed by flow cytometry.

Western blot

CD4⁺GFP⁻ and CD4⁺GFP⁺ cells were sorted into IMDM-10% FBS. Cell pellets of 0.5 to 1 million cells were lysed in 1% Non-diet P-40 buffer (1% NP-40, 50mM Tris, 150mM NaCl) supplemented with protease and phosphatase inhibitors. Protein lysates were boiled for 5 minutes in denaturing sample buffer, separated by SDS-PAGE, and were transferred to a Trans-Blot Nitrocellulose membrane (Bio-Rad Laboratories). After blocking in TBST (10 mM Tris pH 8.0, 150 mM NaCl, 2% Tween 20) supplemented with 3% (w/v) dried skim milk powder, the membrane was probed with indicated primary antibodies in 5% (w/v) BSA-TBST solution followed by HRP-conjugated secondary antibodies in TBST supplemented with 3% (w/v) dried skim milk. Proteins on the membrane were detected by ECL. Antibodies for p-PLC-γ1(#2821), p-RSK1 T359/S363(#9344), p-Foxo1 (#9461S), p-Erk1/2 (#91015), p-p70S6K1 (#9204S), p70S6K1 (#9202),), p-S6(#4838), S6 (#2217), p-4E-BP1 (#2855S), 4E-BP1 (#9644), p-S6(#4838), cleaved Caspase-3 (#9661), cleaved Caspase-9 (#9509), and p-Akt S473 (#9271S) were purchased from Cell Signaling Technology. Anti-GFP (#632376) antibody was purchased from Clontech Inc. Anti-β-actin antibody was from Sigma-Aldrich (A1978).

Real-time PCR

CD4⁺ ICOS⁺Foxp3GFP⁺ and ICOS⁻Foxp3GFP⁺ Treg were isolated by MACS purification and FACS sorting. Total RNA from sorted Treg were prepared with the Trizol reagent. First strand cDNA was synthesized using the Superscript III First-Strand Synthesis System (Invitrogen). Realtime PCR was prepared using the RealMasterMix (Eppendorf) and performed on the Mastercycler® ep realplex² system (Eppendorf). Expressed levels of target mRNAs were normalized with β-actin or snRNA U6, calculated using 2^−ΔΔCT method. Primers used were Bax forward, 5'-tgtttgctgatggcaacttc-3'; Bax reverse, 5'-gatcagctcgggcactttag-3'; Bad forward, 5'-gcacacgccctaggcttgagg-3'; Bad reverse, 5-ggaacatactctgggctggtc-3'; Actin forward, 5'- tgtccaccttccagcagatgt-3', Actin reverse, 5'- tgtccaccttccagcagatgt-3'.

Statistical Analysis

For statistical analysis, two-tail Student t-test was performed. *, p<0.05; **, p<0.01; ***, p<0.001.

Results

Live Foxp3⁺ Treg are hyperproliferative following TCR stimulation in vitro

We compared proliferative responses between CD4⁺Foxp3⁻ Tcon and CD4⁺Foxp3⁺ Treg to different strengths of TCR stimulation in vitro using a CFSE dilution assay. CFSE labeled splenocytes from WT mice were left unstimulated or stimulated with different concentrations of an anti-CD3 antibody for 72 hours. After cell surface staining for CD4 and CD8 as well as a fixable Live/Death staining for dead cells, cultured cells were intracellularly stained for Foxp3. Comparison between total CD4⁺Foxp3⁺ Treg and CD4⁺Foxp3⁻ Tcon revealed much weaker proliferation of Treg than Tcon (Fig 1A), which is consistent with previously published observations that Treg are `anergic' in vitro. However, if gated on live CD4⁺Foxp3⁺ Treg and CD4⁺Foxp3⁻ Tcon (Fig 1B, 1C), we found that Treg had undergone more cell divisions than Tcon, particularly when the stimulation was weak (Fig 1D).

Figure 1. Comparison of proliferation and death between Treg and Tcon in vitro.

CFSE-labeled splenocytes were left unstimulated or stimulated with indicated concentrations of anti-CD3 for 72 hours. Cells were stained with anti-CD4, anti-CD8, and the fixable Live/Death dye followed by intracellular staining for Foxp3. (A). Histogram of CFSE dilution of gated total CD4⁺CD8⁻Foxp3⁺ Treg and CD4⁺CD8⁻Foxp3⁻ Tcon. (B–D). Gating strategy for live cells (B,C) and histograms of CFSE dilution of gated live CD4⁺CD8⁻Foxp3⁺ and CD4⁺CD8Foxp3⁻ cells (D). (E). Simultaneous analysis of proliferation and death of Tcon and Treg following in vitro TCR stimulation. Cell death staining and CFSE dilution of gated total CD4⁺Foxp3⁻ Tcon and CD4⁺Foxp3⁺ Treg in the same experiments as in (A). Data shown represent five experiments.

Differential sensitivity to activation induced death between proliferating Treg and Tcon

The differences of Treg proliferation between total Treg and live Treg prompted us to examine survival and death of Treg during in vitro culture. We analyzed the relationship of cell death vs proliferation of total CD4⁺Foxp3⁺ and CD4⁺Foxp3⁻ cells from the same experiments in figures 1A–1D. As shown in figure 1E, most CD4⁺Foxp3⁻ died after multiple divisions and live CD4⁺Foxp3⁻ T cells had gone through less divisions than dead cells, particularly when the stimulation was strong. In contrast, most of CD4⁺Foxp3⁺ Treg died before the first cell division and the hyperproliferating CD4⁺Foxp3⁺ Treg were mostly live cells. Together, these observations suggest that proliferative CD4⁺Foxp3⁻ Tcon are sensitive to activation induced cell death (AICD) and that most Treg are death prone, but proliferating Treg are resistant to AICD in vitro.

Hyperproliferative Treg are not converted from Tcon

Since some Tcon can convert to Foxp3⁺ following TCR stimulation, this small population of hyper-proliferative Foxp3⁺ cells could be derived from Foxp3⁻ Tcon during in vitro TCR stimulation. To rule out such a possibility, we performed a mixed culture experiment. In this experiment, we utilized the Foxp3GFP mice in which an IRES-GFP cassette is knocked into the 3' untranslated region of the Foxp3 gene so that GFP expression can be used to identify Foxp3⁺ cells. (16) Sorted Thy1.1⁺CD4⁺Foxp3GFP⁻ Tcon from Thy1.1⁺Foxp3GFP mice were mixed with Thy1.2⁺ splenocytes in an amount so that the ratio of Thy1.1⁺ to Thy1.2⁺ Tcon was close to 1:1. The mixed cells were labeled with CFSE and stimulated with anti-CD3. While Thy1.1⁺ and Thy1.2⁺ CD4⁺Foxp3⁻contribute almost equally, more than 95% of live and proliferating CD4⁺Foxp3⁺ cells were Thy1.2⁺ (Fig 2). These observations indicate that while there was a low rate of conversion of CD4⁺Foxp3GFP⁻ to CD4⁺Foxp3⁺ during in vitro TCR stimulation, the contribution of such conversion to the proliferating Treg detected in response to TCR stimulation was minimal and that most of the hyperproliferating CD4⁺Foxp3⁺ cells were natural Treg.

Figure 2. Hyper-proliferative Treg are not derived from Tcon.

Sorted Thy1.1⁺CD4⁺Foxp3GFP⁻ Tcon were mixed with Thy1.2⁺ splenocytes from mice carrying the Foxp3GFP transgene in a ratio that Thy1.1⁺CD4⁺Foxp3GFP⁻ Tcon were almost equal to Thy1.2⁺CD4⁺Foxp3GFP⁻ Tcon in the mixture. The mixed cells were labeled with CFSE and proliferation of Tcon and Treg were similarly examined as in figure 1. (A). Histograms show CFSE dilution of gated live CD4⁺Foxp3⁻ Tcon. Dot plots show Thy1.1 and Thy1.2 staining of gated live CD4⁺Foxp3⁻ Tcon. (B). Histograms show CFSE dilution of live gated CD4⁺Foxp3⁺ Treg. Dot plots show Thy1.1 and Thy1.2 staining of live gated CD4⁺Foxp3⁺ Treg. Data shown represent two experiments.

Increased death and proliferation of freshly isolated Treg

To determine whether the aforementioned in vitro observations reflect properties of Treg in vivo, we examined cell death and proliferative status of freshly isolated T cells. CD4⁺Foxp3GFP⁺ Treg stained positive for annexin V about 3–4 times more frequently than CD4⁺Foxp3GFP⁻ Tcon (Fig 3A, 3B), suggesting more apoptosis of Treg than Tcon in vivo. About 1.5 – 2 fold more Treg than Tcon stained positive for Ki67 (Fig 3C, 3D), a marker of cell proliferation. These data are consistent with the observation that Treg incorporate more BrdU that Tcon in vivo (15). Together, these data are consistent with the aforementioned in vitro data and suggest that Treg are hyperproliferative and prone to death in vivo.

Figure 3. Assessment of proliferative and surviving status of freshly isolated Treg.

(A, B). Increased apoptosis of Treg. Freshly isolated splenocytes and LN cells from Foxp3GFP mice were stained with CD4, CD8, and annexin V. Annexin V staining of CD4⁺Foxp3⁻, CD4⁺Foxp3⁺, and CD8⁺ cells mice was determined by FACS. (A). Representative dot plots. (B). Mean ± SEM presentation of data from multiple mice (n=9). (C,D). Ki67 staining of Treg, Tcon and CD8 T cells. Freshly isolated splenocytes from C57BL6/J mice were stained with CD4 and CD8 followed by intracellular staining for Ki67 and Foxp3. (C). Representative dot plot of Ki67 expression in Treg, Tcon, and CD8 T cells. (D). Mean ± SEM presentation of Ki67 positive Treg, Tcon, and CD8⁺ cells in the spleen and LN from multiple mice (n=5).

Treg proliferation is less dependent on CD28 costimulation than Tcon

T cell activation requires signals from both the TCR and co-stimulatory receptors such as CD28. CD28 deficient mice contain fewer Treg than WT controls. CD28-mediated costimulation has been found to promote human CD4⁺CD25⁺ T cell expansion in vitro (18–20). We examined the role of CD28 costimulation for Treg proliferation using the same experimental system as in figure 1 except that CTLA4-Ig was added to the culture. CTLA4-Ig has high affinities to B7–1 and B7–2 costimulatory ligands and thus blocks CD28-mediated costimulation. As expected, CD4⁺Foxp3⁻ Tcon proliferated less in the presence of CTLA4-Ig than in the absence of this molecule. However, the inhibitory effect of CTLA4-Ig on TCR-induced Treg proliferation was much weaker than on Tcon (Fig 4A).

Figure 4. Differential requirement of CD28-mediated costimulation for TCR-induced Treg and Tcon proliferation.

(A). Treg proliferation is less sensitive to CTLA4-Ig than Tcon. (B). Treg proliferation is less affected by CD28 deficiency than Tcon. Experiments were performed similar to figure 1 except that CTLA4-Ig was added to block costimulation (A) or CD28^−/− T cells were compared to WT T cells (B). Data shown are representative of three experiments.

To further investigate the importance of CD28-mediated co-stimulation on Treg proliferation, we analyzed proliferative responses of CD28^−/− Tcon and Treg to TCR stimulation. As shown in figure 4B, CD28^−/− CD4⁺Foxp3⁻ Tcon proliferated much weaker than WT CD4⁺Foxp3⁻ Tcon. However, CD28^−/− CD4⁺Foxp3⁺ Treg proliferation was only slightly decreased compared to WT CD4⁺Foxp3⁺ Treg. Together, these observations indicate that Treg are less dependent on CD28 costimulatory signal for their proliferation than Tcon.

Altered mTOR signaling and constitutive activation of the intrinsic death pathway in Treg

To investigate the mechanisms that may control Treg proliferation and survival, we compared signaling events between Treg and Tcon. CD4⁺Foxp3⁺ Treg and CD4⁺Foxp3⁻ Tcon were sorted into serum containing medium and lysed immediately after sorting. As shown in figure 5A, Treg contain higher levels of Erk1/2 phosphorylation than Tcon. Erk1/2 usually functions downstream of Ras to promote cell proliferation. The increased Erk1/2 activation correlates with increased proliferation of Treg.

Figure 5. Distinct signaling between Treg and Tcon.

(A–D). Assessment of signaling in serum containing medium. T cells from WT spleens and LNs were isolated by using magnetic beads. CD4⁺FoxP3GFP⁻ Tcon and CD4⁺FoxP3GFP⁺ Treg were sorted into FBS containing medium and were lysed immediately after sorting. Cell lysates were examined for signaling events by immunoblot analysis with the indicated antibodies. (A). Erk1/2 phosphorylation. (B). mTORC1 signaling. (C). mTORC2-Akt signaling. (D). Cleaved caspase 9 and 3. (E–F) Comparison of TCR signaling between Treg and Tcon. Sorted Tcon and Treg as described in (A–D) were washed twice with PBS, rested in PBS at 37°C for 30', and then left unstimulated or stimulated with an anti-CD3 antibody (500A2) for 5 minutes. Cell lysates were subjected to immunoblot analysis with the indicated antibodies. Data shown represent three experiments.

The mammalian target of rapamycin (mTOR) integrates environmental stimuli including growth factors, nutrients, and stress-activated signals to regulate cell metabolism, survival, growth, and proliferation. mTOR signaling is mediated through two signaling complexes: mTOR complex 1 (mTORC1) and mTORC2. mTORC1 is sensitive to rapamycin inhibition, while mTORC2 signaling is insensitive to acute rapamycin treatment. mTORC1 promotes cell growth and proliferation through phosphorylating two major substrates: the 70 kDa ribosomal S6 kinase (S6K1) and the translational repressor 4 elongation factor binding protein 1 (4E-BP1). Activated S6K1 phosphorylates S6 to promote ribosome biogenesis. Phosphorylated 4E-BP1 dissociates from eukaryotic initiation factor 4E (eIF4E) to promote protein translation. mTORC2 phosphorylates Akt at serine 473, leading to an increase of Akt activity (21–23).

Both S6K1 and S6 phosphorylation was lower in CD4⁺Foxp3⁺ Treg than in Tcon. In contrast, 4E-BP1 phosphorylation was higher in Treg than in Tcon (Fig 5B). These observations revealed a dissociation of phosphorylation of S6K1 and 4E-BP1 in Treg. Akt phosphorylation at serine 473 was lower in Treg than in Tcon. In addition, phosphorylation of Foxo1 and GSK3β, events mediated by Akt, was decreased in Treg compared to Tcon. Thus mTORC2 and Akt activities are decreased in Treg (Fig 5C). Consistent with the decreased Akt activity, cleaved active caspase 9 and caspase 3 were both increased in Treg, suggesting enhanced activation of the intrinsic death pathway in Treg (Fig 5D).

TCR signaling in Treg

It has been reported, that Treg display defective signaling transduction following TCR engagement (11). For example, decreased calcium influx and Erk1/2 phosphorylation were reported in TCR-stimulated CD4⁺CD25⁺ Treg (24). To compare TCR signaling between Treg and Tcon, we rested sorted CD4⁺Foxp3⁺ Treg and CD4⁺Foxp3⁻ Tcon in PBS and then stimulated them with anti-CD3. As shown in figure 5E, phosphorylation of PLCγ1, Erk1/2, IκBα, and NFκB was obviously decreased in Treg as compared to Tcon after TCR stimulation, suggesting that proximal TCR signaling and downstream signaling cascades are impaired in Treg. Thus, the aforementioned increase of Erk1/2 phosphorylation in Treg in figure 5A is likely induced by cytokines or other growth factors existing in serum containing medium.

Akt phosphorylation at S473 and Foxo1a phosphorylation were diminished in Treg before and after TCR stimulation (Fig 6B), suggesting that TCR-induced mTORC2 activation is impaired in Treg. While S6K1 phosphorylation was decreased before and after TCR stimulation in Treg, 4E-BP1 phosphorylation in Treg was higher than Tcon before and after TCR stimulation, suggesting again that Treg are distinctive from other cells in mTORC1 signaling.

Figure 6. Effects of IL2 on Treg proliferation and survival.

Experiments were performed and analyzed similarly to those in figure 1 except that IL-2 was added at a concentration of 1000U/ml in the test group. (A). CFSE and Live/Dead staining on gated total Tcon and Treg. (B). Death rates of Tcon and Treg. (C). Overlay of CFSE staining in live Tcon and Treg. Data shown are representative of three experiments.

Promoting Treg proliferation by IL-2

IL-2 plays important roles for Treg homeostasis. Deficiency of IL-2 causes decreased Treg numbers in mice which can lead to autoimmune diseases (25–27). In contrast, injection of IL-2 plus an anti-IL-2 antibody increases Treg numbers in mice that can promote immune suppression (28). We tested how IL-2 may affect the hyperproliferative and death prone properties of Treg. IL-2 can significantly increase proliferation of Tcon (Fig 6A and 6C). IL-2 showed different effects on Tcon survival dependent on the strengths of TCR stimulation. Without TCR stimulation, IL-2 promotes Tcon survival. When TCR stimulation was weak, IL-2 drastically promoted Tcon proliferation as well as death. When TCR stimulation was strong, IL-2 can inhibit Tcon death (Fig 6A and 6B). IL-2 also enhanced Treg proliferation. However, the effect of IL-2 to promote Treg survival was not obvious and could only be observed when stimulated with a high concentration (1 μg/ml) of anti-CD3.

Distinguishing hyperproliferative and death prone Treg subsets by ICOS expression

An important issue that has arisen from our aforementioned data is whether the hyperproliferative Treg and death-prone Treg represent different subsets of Treg. To address this question, it is critical to identify markers that can be utilized to define Treg with these distinct properties. It has been recently reported that human Treg can be divided into ICOS⁺ and ICOS⁻ subsets and that the suppressive activity of the ICOS⁺ Treg appears to be superior to the ICOS⁻ Treg (29–31). ICOS has also been found to be able to promote NKT cell survival (32). We examined the death and proliferative properties of ICOS⁺ and ICOS⁻ Treg. Freshly isolated ICOS⁻ Treg displayed higher death rates than ICOS⁺ Treg (Fig 7A). Moreover, ICOS⁺ Treg stained more intensively for Ki67 than ICOS⁻ Treg, suggesting that more ICOS⁺ Treg and Tcon are in cycling than ICOS⁻ Treg in vivo (Fig 7B). Following in vitro TCR stimulation, the majority of ICOS⁻ Treg died and did not proliferate. However, most ICOS⁺ Treg were alive and proliferated (Fig 7C). ICOS expression on Treg may alleviate the requirement of CD28 for TCR-induced proliferation. ICOS⁺ Treg expressed a higher level of the pro-apoptotic molecule Bad than ICOS⁻ Treg, although they express similar levels of Bax (Fig 7D). Further studies should determine whether differential Bad expression controls survival of these two Treg subsets. Using a contact inhibition assay, we further compared the suppressive activities between these two subsets of Treg. As shown in figure 7E, ICOS⁺ Treg displayed much stronger suppressive activity than ICOS⁻ Treg. The inhibitory effect of ICOS⁺ Treg at 16 to 1 Tcon to Treg ratio was similar to that of ICOS⁻ Treg at 1:1 Tcon to Treg ratio. Together, these data indicate that ICOS expression defines two subsets of Treg with distinct survival and proliferating properties as well as obvious differences of suppressive activities.

Figure 7. Differential expression of ICOS defines proliferative and death-prone Treg subsets.

(A). Assessment of Treg death and survival of freshly isolated Treg based on ICOS expression. Top two panels, CD4 and Foxp3 staining, and ICOS expression on CD4⁺Foxp3⁺ Treg. Bottom two panels, Live/Death staining of ICOS⁺ and ICOS⁻ Treg. (B). Ki67 staining of ICOS⁺ and ICOS⁻ Treg and Tcon. (C). CFSE-labeled splenocytes were stimulated with an anti-CD3 antibody (1μg/ml, 2C11) for 24 or 48 hours. Treg proliferation and survival were examined as in figure 1 except that Treg were gated for ICOS⁺ and ICOS⁻ subsets. Histogram shows overlay of CFSE staining of ICOS⁺ and IOCS⁻ Treg after 48 hours TCR stimulation. (D). Increased Bad expression in ICOS⁻ Treg. Bad and Bax mRNA levels were determined by real-time qPCR with cDNA synthesized from total RNA isolated from sorted ICOS⁺CD4⁺Foxp3GFP⁺ and ICOS⁻CD4⁺Foxp3GFP⁺ Treg as templates. (E). Assessment of suppressive function of ICOS⁺ and ICOS⁻ Treg using the contact inhibition assay. CFSE-labeled Thy1.1⁺Thy1.2⁺ Tcon were mixed with Thy1.2⁺ ICOS⁺ or ICOS⁻ Treg at the indicated ratios as well as mitomycin C treated TCRβ^−/−δ^−/− splenocytes. Cells were left unstimulated or stimulated with an anti-CD3 antibody for 72 hours and then stained for Thy1.1, Thy1.2, and CD4. Top panels show CFSE intensity in gated Thy1.2⁺ Tcon from unstimulated or anti-CD3 stimulated cultures. Bottom panels are overlaid histograms of CFSE intensity in gated Thy1.1⁺Thy1.2⁺ Tcon from anti-CD3 stimulated cultures in the presence of either ICOS⁻ Treg or ICOS⁺ Treg. At the 1:1 Tcon to Treg ratio, only ICOS⁻ Treg were tested.

Discussion

In this report, we simultaneously analyzed Treg proliferation and death using a fixable live/death staining and the CFSE dilution assay. We demonstrated that Treg contain two subsets with distinct death and proliferation properties. The predominant subset is ICOS⁻ and is prone to death, defective in proliferation, and weak in suppression. The rare subset is ICOS⁺ that is characterized by superior survival, hyper-proliferative, and highly suppressive properties. The massive death of ICOS⁻ Treg following in vitro TCR stimulation may be due to the activation of the intrinsic death pathway and contribute to the hypo-proliferative phenotype of Treg during in vitro TCR stimulation. Many studies on Treg proliferation and survival have been performed with CD4⁺CD25⁺ cells. Since most Treg die in vitro, our data suggest that it is critical to use Foxp3 rather than CD25 whenever possible as the marker for Treg isolation when studying Treg, particularly when examining their proliferative responses and death. A small contamination of Tcon in the CD4⁺CD25⁺ population could well outnumber the rare proliferative ICOS⁺ Treg and thus significantly impact data interpretation.

ICOS⁺ and ICOS⁻ Treg have been noted for their difference in several aspects. Human ICOS⁺ Treg use IL-10 to suppress dendritic cells and TGFβ to suppress T cells while ICOS⁻ Treg only employ TGFβ for suppression (30). Human ICOS⁺ Treg have also been found to contain stronger suppressive activity than ICOS⁻ Treg (29, 31, 33). Our data showing that murine ICOS⁺ Treg are much more suppressive than ICOS⁻ Treg are consistent with the data of human Treg. Of note, human ICOS⁺ Treg were found to be more apoptotic than ICOS⁻ Treg (30). At present, the reason for the difference in survival between human and murine ICOS⁺ Treg is unclear. Murine ICOS⁻ Treg become apoptotic within a few hours after in vitro TCR stimulation (data not shown). Since the human Treg were purified based on CD25 expression, it would be of interest to determine whether these CD25⁺ and ICOS⁺ Treg contain activated conventional T cells, which may have a high death rate. ICOS appears not important for Treg generation in the thymus but is involved in Treg homeostasis. Deficiency of either ICOS or its ligand results in decreases of Treg numbers in the peripheral lymphoid organs (31, 34). In addition, expression of ICOS ligand by tumor cells promotes Treg expansion, which may contribute to tumor immune suppression (35). Together, these observations and our data suggest that ICOS may directly promote Treg survival. The importance of ICOS to maintain Treg pool size and survival may help to explain recent finding that inhibition of ICOS signal in mice may reduce Treg number, promote autoimmunity, and exacerbate airway hyperresponsiveness (36–39).

Human CD4⁺ Treg can be divided into resting and activating/memory subsets. It has been reported that the human CD4⁺CD45RO⁺Foxp3⁺CD25^hi Tregs are highly proliferative but susceptible to apoptosis (40). In another report, human CD4⁺ Treg were divided into CD45RA⁺Foxp3^low resting, CD45RA⁻Foxp3^hi activated, and CD45RA⁻Foxp3^low cytokine-secreting subpopulations. Both CD45RA⁺Foxp3^low and CD45RA⁻Foxp3^hi Treg are suppressive while CD45RA⁻Foxp3^low is non-suppressive. In addition, CD45RA⁺Foxp3^low Treg appear to be the precursors of CD45RA⁻Foxp3^hi Treg, which are terminally differentiated and death prone (41). The CD45RA⁺Foxp3^low Treg are more proliferative and survive better than the CD45RA⁻Foxp3^hi Treg following TCR stimulation, which mimics the murine ICOS⁺ and ICOS⁻ Treg to certain extents. However, the differences between these human Treg subsets defined by Foxp3 levels and CD45RA expression are not as drastic as the murine Treg subsets defined by ICOS expression.

Our study has revealed common and different signaling features between freshly isolated Treg in FBS containing medium and in PBS with or without TCR stimulation. Treg in FBS-containing medium (Treg-FBS) display increased phosphorylation of Erk1/2 and Rsk1, substrate for Erk1/2, but decreased phosphorylation of PLCγ1 as compared with Tcon. However, in Treg rested in PBS (Treg-PBS), Erk1/2 and Rsk1 phosphorylation is drastically decreased and is much lower than in Tcon in PBS. The same is true for Treg-PBS stimulated with anti-CD3. These observations suggest that Treg maybe hyper-responsive to growth factors or cytokines in the serum-containing medium such as recently reported leptin (42) and such signaling is independent on PLCγ1 activation. Additionally, distal signaling events that lead to Erk1/2 activation should not be defective in Treg as Erk1/2 phosphorylation occurs at a high level in Treg-FBS. Although Treg are hyper-responsive to yet to be identified stimuli in the serum containing medium, they are defective in TCR signaling. The defects of TCR signaling in Treg is likely proximal to PLCγ1 since its phosphorylation is significantly decreased. Diminished IκBα and NFκB phosphorylation as well as Erk1/2 phosphorylation is like caused by diminished PLCγ1-derived DAG, which is the upstream activator for both the NFκB pathway and the Ras-Erk1/2 pathway (43).

Multiple reports have demonstrated that PI3K and mTOR signaling inhibits Treg differentiation and/or function and Treg display decreased mTOR and Akt activities (42, 44–49). Our studies are consistent with the notion that mTOR signaling is differentially regulated in Treg and Tcon. mTORC2 activation is decreased in both Treg-FBS and Treg-PBS with or without TCR stimulation, indicating the Treg are defective in mTORC2-Akt signaling. At present it is unclear how much the decreased Akt activity in Treg may contribute to Treg death. In both Treg-FBS and Treg-PBS with or without TCR stimulation, S6K1 phosphorylation is decreased but 4E-BP1 phosphorylation is increased as compared to Tcon. The dissociation of S6K1 and 4E-BP1 phosphorylation is quite striking and intriguing since phosphorylation of these two proteins is usually mediated by mTORC1 and occurs concordantly in other cell types (23). Such dissociation could be caused altered composition of mTORC1 leading to change of substrate preference and differential expression of phosphatases for S6K1 and 4E-BP1 in Treg. Of note, S6K1 phosphorylation has been found to be increased in human CD4⁺CD25⁺ T cells in one study. In that study, 4E-BP1 phosphorylation was not examined. In another study, S6K1 as well as 4E-BP1 phosphorylation were both decreased in human CD4⁺CD25⁺ T cells as compared with CD4⁺CD25⁻ cells (46, 48, 49). The reason for the discrepancies between those studies and ours is unclear at present.

In summary, we have demonstrated that Treg display both hyper-proliferative and death-prone properties and that ICOS can define Treg into two subsets based on these properties. The ICOS⁺ Treg are rare, but are superior in survival, are hyper-proliferative in response to TCR stimulation, and contain strong suppressive activity. In contrast, the ICOS⁻ Treg are the predominant subset, are death prone, exhibit defective proliferation capabilities, and contain weak suppressive activity. Our data suggest modulating Treg survival as an important therapeutic strategy for autoimmune diseases and cancer.

1. Abbreviations

Treg

regulatory T cells

Tcon

conventional T cells

TCR

T cell receptor

ICOS

inducible T-cell co-stimulator

mTOR

mammalian target of rapamycin

4E-BP1

translational repressor 4 elongation factor binding protein 1

S6K

70 kDa ribosomal S6 kinase