Human CD25^highFoxp3^pos regulatory T cells differentiate into IL-17–producing cells

In mice, IL-17A (IL-17)–producing T cells have been established as an important T-helper (Th) effector lineage, clearly distinct from the Th1 or Th2 lineage.^1,2  Infectious disease mouse models indicate that IL-17–producing cells (Th17) mediate protection against extracellular pathogens.^3-5  However, on the downside, Th17 cells also appear to be the driving force in the pathogenesis of several autoimmune diseases.^2,6 

Furthermore, mouse studies showed that differentiation of Th17 cells from naive CD4 T cells requires the concomitant activity of IL-6 and transforming growth factor-β (TGFβ).^7-9  The key transcription factor driving Th17 cell differentiation is the orphan nuclear receptor RORγt, which is needed for constitutive expression of IL-17.¹⁰  In in vivo studies in mice, IL-17– and RORγt-expressing cells were shown to be present in the lung and digestive mucosal compartments,¹¹  and especially throughout the intestinal lamina propria.¹⁰ 

In humans, IL-17 is associated with many inflammatory disorders such as rheumatoid arthritis, asthma, multiple sclerosis, lupus, Crohn disease, and allograft rejection.^3,12-15  Recently, human RORγt-positive IL-17–producing T cells were identified under physiologic conditions in peripheral blood and tonsils,^15-17  being contained within the CCR6⁺ (CCR4⁺)CD4⁺ memory T-cell population.^15,16  Most recent data show that human Th17 differentiation, distinct from mouse Th17 development, is under control of IL-1β, IL-6, and IL-23.^18,19 

In mice, the development of Th17 cells was described to be linked to that of regulatory T cells (Tregs) in a reciprocal fashion, whereby under the influence of TGFβ, IL-6 levels determine the outcome.^8,9  In vitro, activated Tregs promoted Th17 cell differentiation from CD4 T cells,^9,20  likely through their production of TGFβ. In addition, in vivo the transfer of Treg enhanced IL-17 production in a mouse model of systemic autoimmune disease.²¹  Next, to this reciprocal relationship, it was recently demonstrated that IL-17–producing cells directly develop from mouse Tregs.²⁰  This is a remarkable finding, because Tregs are typically associated with T-cell tolerance and immune homeostasis.²²  Different Treg subsets have been described, and even within the naturally occurring CD4^posCD25^high Foxp3^pos population heterogeneity is evident. This heterogeneity reflects differences in differentiation or developmental stage, trafficking properties, and suppressor capacity.^23-25  Although the lineage differentiation of helper T cells is very well accepted, that of Tregs is only scarcely recognized. In humans, naive and memory-like CD4^posCD25^posFoxp3^posTreg have been defined in peripheral blood, discriminated by the expression of CD45 isoforms CD45RA and CD45RO.^25-28  Also in vitro differentiation of human Foxp3⁺ Tregs has been demonstrated based on CD45 isoform switching.^25,29  Recently, we described 2 human Treg subsets originating from the naturally occurring CD4⁺Foxp3⁺ Treg pool,²⁴  made apparent by differential expression of CD27, homing receptors such as CD62L, and suppressor function. Gradually, the notion emerges that the regulatory arm of the immune response may develop along similar lines as the effector arm. This would imply that Treg development might exhibit a degree of plasticity that meets local requirements and thereby transgresses lineage barriers. Here, we fuel this notion by showing that isolated highly purified CD4^posCD25^highCD27^posCD45RA^negFoxp3^pos Tregs can differentiate into IL-17–producing cells, given that antigen-presenting cells (APCs), in particular monocytes, and the cytokines IL-2 or IL-15 are present. Moreover, we reveal an important role for the IL-1β/IL-1Ra system in regulating Treg differentiation into IL-17–producing cells. In addition, IL-23 and IL-21 were found to enhance this process. This IL-17 induction was found to be dependent on epigenetic modification as evidenced by use of an HDAC inhibitor.

Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation (Lymphoprep; Nycomed-Pharma AS, Oslo, Norway) of buffy coats obtained from healthy donors upon written informed consent according to the Declaration of Helsinki with regard to scientific use. To obtain high-purity T-cell populations, first CD4^pos T cells were purified from PBMCs by negative selection using monoclonal antibodies (mAbs) directed against CD8 (RPA-T8), CD14 (M5E2), CD16 (3G8), CD19 (4G7), CD33 (P67.6), and CD56 (B159) (BD Biosciences, Erembodegem, Belgium) combined with sheep anti–mouse Ig–coated magnetic beads (Invitrogen, Breda, The Netherlands). This resulted in a CD4^pos T-cell enrichment of more than 95% and the absence of CD8^pos T cells. Purified CD4^pos T cells were labeled with anti–CD27-FITC (M-T271; DAKO, Glostrup, Denmark), CD25PE-(MA251; BD Biosciences), or CD45RA-ECD (2H4; Beckman-Coulter, Mijdrecht, The Netherlands); thereafter CD25^highCD27^posCD45RA^neg cells (Tregs) were isolated by high-purity flow cytometric cell sorting using an Altra cell sorter (Beckman-Coulter). A rerun was performed to analyze the cell purity of the sorted cells; sorted cells were always more than 98% pure. Where indicated, sorted CD4^posCD25^negCD27^pos T cells (Teffs) were used for comparison. Monocytes or B cells were positively isolated from PBMCs with magnetic bead isolation using CD14 microbeads (Miltenyi-Biotec, Utrecht, The Netherlands) and CD19 Dynalbeads (Invitrogen, Breda, The Netherlands), respectively, according to the manufacturers' instruction.

Cells were cultured in culture medium (RPMI-1640 with glutamax supplemented with pyruvate [0.02 mM[, 100 U/mL penicillin, 100 μg/mL streptomycin [all from Gibco, Paisley, United Kingdom], and 10% human pooled serum [HPS]) at 37°C, 95% humidity, and 5% CO₂, in 96-well round-bottom plates (Greiner, Frickenhausen, Germany).

Recombinant human cytokines IL-2 (rIL-2; 12.5 U/mL, proleukine; Cetus, Amsterdam, The Netherlands) and IL-15 (rIL-15;10 ng/mL), IL-1β (rIL-1β;50 ng/mL), IL-21 (rIL21;12.5 ng/mL) (Biosource, Etten-Leur, The Netherlands), IL-1 receptor antagonist (rIL-1Ra;125 ng/mL), IL-6 (rIL-6;50 ng/mL), IL-23 (rIL-23;50 ng/mL), TGFβ (rTGFβ; 3 ng/mL) (R&D Systems, Abingdon, United Kingdom), granulocyte colony-stimulating factor (G-CSF) (rGCSF; 50 ng/mL, Neutropen; Amgen, Breda, The Netherlands), and trichostatin A (TSA, 50 ng/mL; Sigma-Aldrich, Zwijndrecht, The Netherlands) were used. All agents were added at the start of the cultures.

Cells were phenotypically analyzed by 4- or 5-color flow cytometry as described previously.³⁰  The following conjugated mAbs were used: CD3-(UCHT1), CD4-(MT310), CD8-(DK25), CD27-(M-T271), CD45RA-(4KB5), CD45RO-(UCHL1) FITC- or PE-labeled (DAKO), CD25-(M-A251) PE, CD127-(M21) PE, CCR4-(1G1) PeCy7, CCR6-(11A9; BD Biosciences) PE- or biotin-labeled, CXCR3-(1C6/CXCR3) PeCy5 (BD Biosciences), CXCR4-(12G5) PeCy5 (eBioscience, Uithoorn, The Netherlands), CCR7-(150503) FITC or PE (R&D Systems), CD4-(T4) ECD, CD4-(T4) PeCy5, and CD62L-(DREG54) ECD (Beckman-Coulter). The detection of biotinylated antibodies was performed with streptavidin conjugated to PeCy5, PeCy7 (Beckman-Coulter), or Quantum dots (Qdot605; Invitrogen, Breda, The Netherlands). Isotype-matched antibodies were used to define marker settings. Intracellular analysis of Foxp3-(PCH101) FITC or PE and IL-17-(6CAP17) PE (eBioscience) was performed after fixation and permeabilization, using Fix and Perm reagent (eBioscience). Before intracellular cytokine measurements, the cells were stimulated for 4 hours with PMA (12.5 ng/mL) plus ionomycin (500 ng/mL) in the presence of Brefeldin A (5 μg/mL; Sigma-Aldrich).

To study cell division by flow cytometry, 10⁶ cells of interest were labeled with 0.5 μM CFSE (Molecular Probes, Leiden, The Netherlands) as described previously.³⁰ 

Cell samples were measured on a FC500 flow cytometer (Beckman-Coulter), and flow cytometry data were analyzed using CXP software (Beckman-Coulter).

Sorted CD25^highCD27^pos Treg or CD25^negCD27^pos Teff (2.5 × 10⁴) populations were stimulated with 0.5 to 10⁵ HLA-mismatched γ-irradiated (30 Gy) stimulator PBMCs (HLA typing was conducted as described previously³⁰ ), or anti-CD3 plus anti-CD28–coated beads (Invitrogen, Breda, The Netherlands) in 200 μL culture medium in the absence or presence of recombinant human cytokines. To analyze cytokine production, cells were stimulated with PMA (12.5 ng/mL) plus ionomycin (50 ng/mL; Sigma-Aldrich) for the indicated culture time.

The suppressor capacity of regulatory cells was studied in a coculture suppression assay. CD4^posCD25^neg T cells (2.5-5 × 10⁴) were stimulated with anti-CD3 plus anti-CD28–coated beads (Invitrogen, Breda, The Netherlands) in a cell:bead ratio of 4:1, in the absence or presence of increasing numbers of Tregs. T-cell proliferation was analyzed by ³H-thymidine incorporation using a Gas Scintillation Counter (Matrix-96 Beta-counter; Canberra-Packard, Meriden, CT). To this end, 0.037 MBq (1 μCi) ³H-thymidine (ICN-Pharmaceuticals, Irvine, CA) was added to each well, cells were harvested after 8 hours of culture, and ³H-thymidine incorporation was measured. The ³H-incorporation is expressed as mean plus or minus SD counts per 5 minutes of at least triplicate measurements

Human IL-1, IL-2, IL-4, IL-5, IL-6, IL-10, IL-13, IL-17, IFNγ, TNFα, and G-CSF were determined in the supernatant of the MLC cultures and primary T-cell cultures, respectively, using human cytokine multiplex kits (Bioplex-system; Bio-Rad, Veenendaal, The Netherlands) according to the manufacturer's instructions. The sensitivity of the cytokine assay was less than 5 pg/mL for all cytokines measured.

Total RNA was extracted with the RNeasy Micro Kit according to the manufacturer's instructions (Qiagen, Hilden, Germany). cDNA was synthesized using the SuperScript III First-Strand Synthesis System and Oligo(dT)₂₀ primers (Invitrogen, Carlsbad, CA). Transcripts were quantified by real-time quantitative polymerase chain reaction (PCR) on an ABI-PRISM 7700 Sequence Detector using predesigned TaqMan GeneExpression Assays and reagents according to the manufacturer's instructions (Applied Biosystems, Foster City, CA). Probes with the following Applied Biosystems assay identification numbers were used: TBX21, Hs00203436_m1; GATA3, Hs00231122_m1; FOXP3, Hs00203958_m1; IL17A, Hs00174383_m1; TGFB1, Hs00171257_m1; RORC variant 1, Hs00172858_m1; and RORC variant 1 + 2, Hs01076112_m1. Data were normalized using the human HPRT1 Endogenous Control (4333768T; Applied Biosystems) and expressed as relative arbitrary units. Because there is no probe available for RORC variant 2, and the RORC variant 1 was not detected in freshly isolated cells (data not shown), we used a probe that detects both variants and took the levels obtained to be indicative for RORC variant 2 (RORγt).¹⁸ 

A standard 2-tailed t test or Wilcoxon matched pairs test was used for statistical analysis; P values of .05 or less were considered significant and are indicated with an asterisk (*). Nonsignificant data are indicated as NS.

In contrast to the latest studies, describing Th17 differentiation from the naive human Foxp3⁺CD4 T-cell pool, recent findings on Th17 differentiation in murine models have prompted us to evaluate the differentiation of IL-17–producing cells from human CD4⁺ Tregs. We set out to study a highly pure and well-defined human CD4^posCD25^highCD27^posCD45RA^neg Treg population, freshly isolated from peripheral blood by flow cytometric cell sorting. A phenotypic analysis before sorting established the broader phenotype of our target CD4^posCD25^high CD27^posCD45RA^neg Treg population to be as follows: cells consistently expressed CD45RO (not shown), Foxp3, and CD62L. The majority expressed CCR4 and CXCR4, approximately half of the population expressed CCR6 and HLA-DR, few cells expressed CCR7, CXCR3, and IL-23 receptor, whereas CD127 expression was lacking (Figure 1A).

For comparison, the CD4^posCD25^negCD27^posCD45RA^neg T-cell population was positive for CD127 and CD62L, the majority expressed CXCR4, approximately half of the population expressed CCR7, less than 20% expressed the IL-23 receptor, whereas expression of Foxp3, CCR4, CCR6, and HLA-DR was virtually lacking. Accordingly, after sorting, the resultant CD4^posCD25^highCD27^posCD45RA^neg Tregs (from here on referred to as CD25^highCD27^pos Tregs) all ex-pressed Foxp3 and lacked CD127, whereas in contrast, sorted CD4^posCD25^negCD27^posCD45RA^neg T cells (CD25^negCD27^pos Teffs) all lacked Foxp3 and were all positive for CD127 (Figure 1B). Subsequently, we analyzed the expression of mRNA encoding transcription factors that are specific for Tregs (Foxp3), Th1 cells (Tbet), Th2 cells (GATA3), and Th17 cells (RORγt; Figure 1C). The CD25^highCD27^pos Treg population, in contrast to the CD25^negCD27^pos Teff population, showed high expression of Foxp3. Then, the suppressor function of the freshly isolated and sorted CD25^highCD27^pos Tregs was measured, which as anticipated, showed a dose-dependent suppression of CD4^pos T-cell proliferation, in contrast to the CD25^negCD27^pos Teffs (Figure 1D).

The differential production of cytokines by T cells after stimulation is a hallmark of T-cell differentiation.⁶  Therefore, we determined the cytokine-producing potential of sorted CD25^highCD27^pos Tregs upon stimulation with PMA plus ionomycin. In contrast to CD25^negCD27^pos Teffs, CD25^highCD27^pos Tregs hardly produced IL-2, IFNγ, or TNFα, but were able to produce IL-5, IL-13, and IL-10 (Figure 2A). Stimulation did not enhance mRNA levels of TGFβ (data not shown). Interestingly, the Treg population was found to produce higher levels of IL-17 than the CD25^negCD27^pos Teffs. Next, we analyzed IL-17 production of sorted CD25^highCD27^pos Tregs after stimulation with allogeneic PBMCs in the presence of rIL-2 plus rIL-15 (Figure 2B). This stimulation of Tregs also resulted in IL-17 production, which always occurred within a defined timeframe of 6 up to 9 days of culture. In the Treg cultures, little IFNγ production was found (Figure 2B). Both CD25^highCD27^pos Tregs and CD25^negCD27^pos memory Teffs showed proliferation; the latter demonstrated more vigorous cell growth²⁴  (data not shown). The production of IL-17 by Tregs upon stimulation with PBMCs and rIL-2 plus rIL-15 on day 8 of culture was confirmed by intracellular cytokine staining after PMA plus ionomycin stimulation (Figure 2C). The percentage of IL-17–producing cells was significantly higher in the CD25^highCD27^pos Treg population compared with CD25^negCD27^pos memory Teffs (summarized in Figure 2D). The majority of IL-17–producing Tregs (> 63%) were single IL-17 producers; 37% produced both IL-17 and IFNγ (data not shown).

Intrigued by the observation that activated CD25^highCD27^pos Tregs produced IL-17, we decided to study this phenomenon in more detail. To confirm IL-17 production by Foxp3⁺ Tregs, we analyzed IL-17 production by sorted CD25^highCD27^pos Tregs after stimulation with allogeneic PBMCs and rIL-2 plus rIL-15 by flow cytometry. Early after stimulation, CD25^highCD27^pos Tregs coexpressed Foxp3 and IL-17 (Figure 3A). Importantly, dividing Tregs expressed IL-17 and gradually lost Foxp3 expression (Figure 3A). This would suggest that single-positive IL-17 cells pass through a Foxp3/IL-17 double-positive stage (Figure 3A,B). At the end of the culture, an inverse relationship between Foxp3 expression and IL-17 production was observed (Figure 3B). Next, we addressed the question as to whether the production of IL-17 after allogeneic PBMC stimulation paralleled the previously observed differentiation of CD25^highCD27^pos Treg into CD27^pos and CD27^neg T-cell subsets.²⁴  To examine this, high-purity sorted CD25^highCD27^pos Tregs were labeled with CFSE and activated with allogeneic stimulator PBMCs in the presence of exogenously added rIL-2 plus rIL-15, and analyzed by flow cytometry at day 8 of the cultures. As shown previously, cell division–related differentiation of the CD25^highCD27^pos Tregs took place as indicated by the emergence of both CD27^pos and CD27^neg T-cell populations with low CFSE content (Figure 3C). IL-17–producing cells were predominantly found after multiple divisions of the starting CD25^highCD27^pos population (Figure 3D). Moreover, IL-17 appeared to be exclusively produced by the resultant CD27^neg cells (Figure 3E). To confirm that the IL-17–producing cells resided within the CD27^neg population, which was generated as a consequence of CD25^highCD27^pos Treg differentiation, again peripheral CD25^highCD27^pos Tregs were sorted, labeled with CFSE, and stimulated with allogeneic stimulator PBMCs and rIL-2 plus rIL-15. This typically results in the differentiation of CD25^highCD27^pos Tregs into CD27^neg and CD27^pos cells, such as shown in Figure 3C. After 10 days of culture, the emerged CD27^neg and CD27^pos cells were sorted based on CD27 expression and low CFSE content. These sorted, differentiated CD27^pos and CD27^neg cells were restimulated with PMA plus ionomycin and analyzed for the production of IL-17 in the culture supernatants. Indeed, IL-17 appeared to be produced exclusively by the CD27^neg population that had differentiated from the CD25^highCD27^pos Treg population, but not from the CD25^negCD27^pos population (Figure 3F). Thus, after cell division–related differentiation of CD25^highCD27^pos Tregs, the resultant CD27^neg cells are the main producers of IL-17.

T-cell differentiation is strongly determined by the cytokine microenvironment. Therefore, we here analyzed which cytokines and stimulatory conditions contribute to the differentiation of CD25^highCD27^pos Tregs into IL-17–producing cells. First, to analyze the effect of exogenous rIL-2 plus rIL-15, growth factors that are routinely added to our cultures to facilitate Treg proliferation, high-purity sorted CD25^highCD27^pos Tregs were activated with allogeneic stimulator PBMCs in the absence or presence of rIL-2, rIL-15, or the combination thereof. At day 8 of the cultures, the IL-17–producing capacity was analyzed by flow cytometry (Figure 4). In the absence of rIL-2 and rIL-15 supplementation, the CD25^highCD27^pos Tregs were unable to produce IL-17 (Figure 4A; Figure S1, available on the Blood website; see the Supplemental Materials link at the top of the online article). Overall, the presence of rIL-2, rIL-15, or the combination of both resulted in comparable numbers of IL-17–producing cells. No additive effect of rIL-2 and rIL-15 was observed (data summarized in Figure 4B).

Second, we examined the need for antigen-presenting cells (APCs) in the differentiation of CD25^highCD27^pos Tregs into IL-17–producing cells. The number of IL-17–producing cells was higher upon stimulation of CD25^highCD27^pos Tregs with allogeneic PBMCs compared with stimulation with anti-CD3 plus anti-CD28 mAb–coated beads, both in the presence of IL-2/IL-15 (Figure 5A). Notably, both stimulation methods induced Treg proliferation²⁴  (data not shown). Importantly, also the secretion level of IL-17 (mean fluorescence intensity [MFI] values) was higher after PBMC stimulation. Within the stimulating PBMC fraction, the CD14^pos monocytes were endowed with the most salient capacity to stimulate differentiation of CD25^highCD27^pos Tregs into IL-17–producing cells (Figure 5B), stressing the importance of APCs in this process.

Given that stimulation with APCs, compared with anti-CD3 plus anti-CD28 mAb–coated bead stimulation, was superior in stimulating the differentiation of Tregs into IL-17–producing cells, we next analyzed which APC-related cytokines were produced in case of APC stimulation versus anti-CD3 plus anti-CD28 stimulation. Compared with anti-CD3 plus anti-CD28 bead stimulation, in case of allogeneic APC stimulation of sorted CD25^highCD27^pos Tregs, we typically found increased IL-1β, IL-6, and G-CSF levels (Figure 5C).

These findings prompted us to analyze whether the addition of exogenously added rIL-1β, rIL-6, or rG-CSF to cultures of APC-stimulated CD25^highCD27^pos Tregs in the presence of rIL-2/rIL-15 enhanced the generation of IL-17–producing cells. Again, the presence of rIL-2/rIL-15 appeared a prerequisite for the emergence of IL-17–producing cells (Figure 4), as in their absence neither rIL-1β, rIL-6 alone, nor both in combination had any effect (Figure 6A; Figure S1). This is probably due to the lack of proper blast formation and subsequent proliferation (Figure S1). In addition, addition of G-CSF had no effect (not shown). In sharp contrast, the addition of rIL-1β together with rIL-2/rIL-15 resulted in a firm increase in the number of IL-17–producing cells. This effect was not the result of increased proliferation, as rIL-1β did not further promote Treg proliferation, such as observed upon stimulation with PBMCs and rIL-2/rIL-15 (data not shown). rIL-6 alone was not effective, but in individual cases IL-6 did enhance the effect of IL-1β. The observed effect of IL-1β was far more pronounced for CD25^highCD27^pos Tregs than for CD25^negCD27^pos Teffs. To determine the direct contribution of IL-1β on the differentiation of Tregs into IL-17–producing cells, rIL-1 receptor antagonist (IL-1Ra) was added to the cultures. The addition of rIL-1Ra did lead to a significant reduction in the number of IL-17–producing cells (Figure 6B, and summarized in 6C). In addition, based on results obtained in the mouse, we analyzed the effect of TGFβ supplementation. We observed no enhancing effect of the addition of TGFβ to the cultures (Figure S2).

Thus, Treg signaling via IL-2 or IL-15 appears a prerequisite to induce differentiation of regulatory T cells into IL-17–producing cells. Once IL-2/IL-15 are present, the local cytokine microenvironments may then fine-tune the effector fate of the Treg population, as is clearly demonstrated by the opposite effect of IL-1 and IL-1Ra.

Next to IL-1β, IL-23 has been reported to induce IL-17 production in human conventional T cells.^18,19  To address the role of IL-23 in our experimental system, we used neutralizing anti–IL-23 mAb. The addition of anti–IL-23 mAb to Tregs stimulated with allogeneic PBMCs in the presence of IL-2/15 or IL-2/IL-15/IL-1β did not hamper the differentiation of Tregs into IL-17–producing cells (Figure 6D). The neutralizing capacity of the anti–IL-23 mAb was demonstrated by preventing rIL-23–enhanced IL-17 induction (Figure 6D). Second, we analyzed whether rIL-23 or rIL-21, the latter was effective in generation of Th17 cells in mice,^31,32  had the potential to enhance Treg differentiation into IL-17–producing cells. Upon Treg stimulation with PBMCs and IL-2 plus IL-15, the coaddition of rIL-23 or rIL-21 (IL-21 was effective in 50% of our experiments) led to an increased production of IL-17–producing cells (Figure 6E). In the absence of IL-2 plus IL-15, rIL-23 and rIL-21 were not effective (data not shown).

Histone/protein deacetylases (HDACs) regulate chromatin remodeling, gene expression, and the functions of many transcription factors and nonhistone proteins. Optimal Treg function required acetylation of the forkhead domain of Foxp3.³³  The HDAC inhibitor (HDACi) trichostatin A (TSA) increased Foxp3 gene expression, as well as the production and suppressive function of regulatory T cells.³³  We thus hypothesized that inhibition of HDAC activity might lead to reduced differentiation of Tregs into IL-17–producing cells, thus favoring maintenance of Foxp3-expressing cells. To examine this, sorted Tregs were stimulated with allogeneic PBMCs in the presence of IL-2 plus IL-15 and IL-1β to optimally stimulate differentiation of Tregs into IL-17–producing cells, with and without HDACi. HDAC inhibition almost completely prevented the differentiation of Tregs into IL-17–producing cells (Figure 6F), while sustaining numbers of Foxp3-expressing cells (Figure 6F). This occurred irrespective of the effect of the HDACi on proliferation (data not shown). Thus, the data imply that the mechanism of Treg differentiation into IL-17–producing cells in our system is dependent on epigenetic modification.

As shown in Figure 3, after the differentiation of CD25^highCD27^pos Tregs into IL-17–producing cells, these cells were found to reside within the emerging CD27^neg population. Here, we sought to determine which cell surface marker positively identifies this IL-17–producing population. First, we measured expression of the homing/chemokine receptors CCR6, CCR4, CCR7, CXCR3, CXCR4, and CD62L upon allogeneic stimulation of CD25^highCD27^pos Tregs and, for comparison, of CD25^negCD27^pos Teffs. The CD27^neg T cells that differentiated from CD25^highCD27^pos Tregs clearly showed a high number of CCR6-expressing cells (> 50%; Figure 7A). This was distinct from both CD27^pos cells that differentiated from CD25^highCD27^pos Tregs, and CD27^pos and CD27^neg cells differentiating from the CD25^negCD27^pos Teffs. The majority of CCR6-expressing CD27^neg cells lacked expression of CCR4 (Figure 7B). Indeed, this CD27^negCCR4^negCCR6^pos cell fraction, obtained after allogeneic stimulation of CD25^highCD27^pos, produced IL-17, especially after addition of IL-1β (Figure 7C), which was confirmed after sorting of the CCR6^pos cells (Figure 7D). These CCR6^pos cells showed high expression of the Th17-associated transcription factor RORγt mRNA that was paralleled by high IL-17mRNA levels (Figure 7E). This was in contrast to the CCR6^neg population that revealed low amounts of RORγt and an absence of IL-17 mRNA (Figure 7E).

Given the fact that previously CCR6 has been associated with proinflammatory Th17 cells, we set out to exclude that our finding was not so much a result of outgrowth of preexisting Th17 CCR6^pos cells, but indeed a result of differentiating Tregs. Thus, we performed CCR6 subsorting of freshly isolated Tregs. We provided evidence that CCR6^pos, and more importantly, that CCR6^neg cells differentiated into IL-17–producing cells (Figure 7F), thereby further substantiating our findings on Treg plasticity.

Like Th1 and Th2, it is now generally assumed that Th17 cells evolved to provide adaptive immunity to encounter specific classes of pathogens.⁶  In mice, differentiation of naive CD4^pos T cells into Th17 cells is controlled by the presence of the cytokines TGFβ and IL-6 in the microenvironment.^7-9  Of note, Th17 development from naive CD4 T cells is clearly different in the human system^18,19,34  where IL-1, IL-6, and IL-23 were found to be key in this process. In contrast to the latest studies, describing Th17 differentiation from the naive T-cell pool, we have focused our studies on the plasticity of the T regulatory arm of the immune system, and analyzed differentiation of human CD4⁺Foxp3⁺ Tregs into IL-17–producing cells.

Our main finding, as summarized in Figure S4, is that stimulation of highly purified human CD4^posCD25^highCD27^posCD45RA^negFoxp3^pos Tregs with allogeneic PBMCs in the presence of exogenous r-IL2/rIL-15 led to a subset of IL-17–producing cells. IL-2 and/or IL-15 appeared obligatory for this differentiation process, which was most salient in the presence of APCs such as monocytes. Induction of IL-17 production by CD4^posCD25^highCD27^posCD45RA^negFoxp3^pos Tregs required cell division that was accompanied by the loss of CD27 and Foxp3 protein expression. Single-cell analysis of differentiated Tregs indicated that the IL-17–producing cells expressed CCR6 and the IL-17–associated transcription factor RORγt mRNA. Cells still demonstrated suppressive capacity, albeit to a lesser extent than freshly isolated Tregs. Coexpression of IL-17 and IL-21, IL-22, and IFNγ (∼ 10%, ∼ 47%, and ∼ 37% of IL-17–producing cells, respectively) was observed (H.J.P.M.K. and I.J., unpublished observations, February 2008). Finally, given that IL-2 or IL-15 is present, the IL-1/IL-1R antagonist system was found to control this Treg differentiation program. We demonstrated that the differentiation of Tregs into IL-17–producing cells depends on histone/protein deacetylase (HDAC) activity. Inhibition of HDAC activity prevented differentiation into IL-17–producing cells and resulted in sustained Foxp3 expression. Support for this latter observation came from a recent study demonstrating that optimal Treg function requires acetylation of the forkhead domain of Foxp3, leading to increased Foxp3 gene expression and the production of regulatory cells with suppressive function.³³  Our findings indicate that human CD45RA⁻ Tregs (and CD45RA⁺ Tregs; data not shown) differentiate into IL-17–producing T cells, thereby revealing a high plasticity of Tregs, which is controlled by cytokines in the local microenvironment and as our data suggest is regulated by epigenetic remodeling.

We have thoroughly examined the possibility that the IL-17–producing cells in our cultures arose from contaminating CD4^posCD25^pos effector, memory cells or CD4^posCD25^neg T cells. First of all, the acquisition of the IL-17 phenotype that we describe in Tregs was conducted using a well-defined population of highly purified CD4^posCD25^highCD27^posCD45RA^neg peripheral Tregs. These cells expressed the transcription factor Foxp3 and lacked expression of the IL-7 receptor (CD127), both of which are considered as hallmarks of true Tregs.^35,36  Moreover, given that effector/memory cells express CD127,^37,38  this clearly differentiates these cells from Tregs. Second, both Tregs and activated (effector/memory) T cells express CD25. However, they do differ in expression level; Tregs express high levels of CD25, whereas activated CD4^pos T cells express intermediate levels of CD25.^35,36,39  In our culture system, we readily found IL-17 production in the sorted CD25^highCD27^pos Treg population, but it was virtually absent in the CD25^intermediateCD27^pos T cells (data not shown). Third, for comparison, memory CD4^posCD25^negCD27^posCD45RA^neg T cells that were Foxp3^neg and CD127^pos were included throughout our experiments. These cells, compared with CD4^posCD25^high Tregs, always responded in a completely different fashion with respect to cytokine production profile (almost no IL-17 production), cell surface marker expression, and the presence of transcription factors (Figures 1,2). Importantly, IL-17–producing cells were not simply the result of outgrowth of already differentiated CCR6^pos Th17 cells, since they emerged from both purified CCR6^pos and CCR6^neg Treg subsets (Figure 7F).

To further substantiate our finding that bona fide Tregs can acquire an IL-17 phenotype, we showed that a highly dedicated Treg subset (CD25^highFoxp3^highCD27^posCD62L^posCTLA4^high; revealing 50% inhibition at Treg/Teff ratios of at least 1:100)²⁴  was still capable of differentiating toward the IL-17 phenotype (data not shown). Together this implies that it is highly unlikely that in our experimental system the IL-17–producing cells were a result of contaminating outgrowth.

The expression of chemokine receptors has been instrumental in the characterization of T-cell subsets.⁴⁰  We found that the IL-17–producing cells that differentiated from Tregs were contained within the CD27^neg cell population, expressed CCR6, and showed expression of RORγt mRNA. Strikingly, others have recently reported that under physiologic conditions and at inflamed sites, human IL-17–producing cells were contained within the CCR6^posCD4^pos memory T-cell population,^15,16  which gives rise to questions on the origin of these tissue-infiltrating cells, which are so far unresolved. The expression of CCR6 enables migration induced by the ligands CCL20 and β-defensin, which are associated mainly with inflamed sites,^41,42  as well as in peripheral tissues such as the skin.⁴³  The latter adds to the tissue-infiltrating and injurious properties assigned to Th17 cells.

In the current study, we show that Treg differentiation into IL-17–producing cells is cell division–dependent. In accordance, we demonstrate that IL-2 or IL-15, which share common β and γ chains, are obligatory in this differentiation process. Indeed, Tregs strongly depend on growth factors (in particular IL-2) that are crucial for their development, survival, and proliferation.^44-46  Of note, Th17 differentiation from naive T cells in the mouse was antagonized by IL-2 signaling.^47,48  Thus, the acquisition of IL-17 phenotype within the T-regulatory arm of the immune system may require additional control mechanisms distinct from the factors regulating the naive T-cell developmental pathways.

We here report that the proinflammatory cytokine IL-1β, in the presence of IL-2 and/or IL-15, enhanced IL-17 production in human Tregs, a process that was counteracted by its natural antagonist IL-1Ra. Although endogenous IL-6 was produced in our experimental setup, IL-6 mAb-neutralizing experiments did not prevent the emergence of IL-17–producing cells (data not shown), and exogenously added rIL-6 did not enhance IL-17. Addition of rIL-6 to exogenously added rIL-1β provided inconsistent results in terms of IL-17 enhancement (variation between individuals). Neutralizing IL-23 mAbs did not prevent Treg differentiation into IL-17 cells, suggesting a limited role for IL-23 in our experimental setup of PBMC stimulation in the presence of IL-2 + IL-15. The addition of TGFβ, associated with the induction of Th17 cells in mice, had no enhancing effect. Only very recently, it was shown that the differentiation of human naive CD4^pos T cells into Th17 effector cells is governed by IL-1β, IL-6, and IL-23, but not TGFβ.^18,19  Interestingly, in our hands, IL-1β, but also IL-23, had major effects on the induction of IL-17 in Tregs, whereas this effect on the similarly treated Teff population was much less outspoken. The biologic implication of the IL-1/IL-1Ra control system has been made explicit in both mice and humans. Abnormal T-cell activation caused by the imbalance of the IL-1/IL-1Ra system was responsible for the development of experimental autoimmune encephalomyelitis,⁴⁹  and in IL-1Ra knockout mice IL-17 production was increased and associated with increased arthritis severity.⁵⁰  In line with this, the blocking of IL-17 suppressed inflammation at the level of the joint in experimental arthritis.⁵¹ 

In human autoimmune disease, in MS patients, IL-1β was present in lesions, and patients with a high IL-1β to IL-1Ra production ratio were at a greater risk of having a family member with relapse-onset MS than those with a low ratio.⁵²  Indeed, IL-1 has been reported to break tolerance by promoting both proliferation and cytokine production of CD4^pos effector/memory T cells, and by attenuating Treg suppressor function.⁵³  Our data suggest that the latter may be explained by induction of IL-17–producing cells from Tregs. Whether these Treg-derived IL-17–producing cells indeed have tissue-injurious properties remains to be established.

To date, information on human Treg development and differentiation is limited. Importantly, we now demonstrate that human Foxp3⁺ Treg also possess an effector differentiation program resulting in IL-17 production. Our data suggest that epigenetic modification underlies this phenomenon. This flexibility in Treg differentiation programming may have evolved to anticipate local microenvironmental needs and serves to control the wide variety of immune responses that take place at distinct anatomic sites. In addition, because thymic egress of naive cells is rapidly lost with age, this type of plasticity, whereby predefined lineage barriers are crossed, may in fact serve a more general phenomenon enabling homeostatic maintenance and optimal immune surveillance.

This feature of Treg development can be exploited to facilitate vaccination strategies and tumor immunotherapy. However, on the downside, it may have implications for the clinical application of ex vivo–expanded Tregs as part of tolerance induction strategies.

We thank R. Woestenenk and J. van Velzen, Central Hematology Laboratory (CHL), Radboud University Nijmegen Medical Centre (RUNMC) for expert cell sorting, and H. Tijssen and E. Fasse, Department of Blood Transfusion and Transplantation Immunology (ABTI), RUNMC for RT-PCR analysis and technical assistance.

Contribution: H.J.P.M.K., R.L.S., P.M.V., A.M.H.B., and I.J. designed research; H.J.P.M.K., R.L.S., P.M.V., and E.R. performed research; H.J.P.M.K., R.L.S., P.M.V., E.R., A.M.H.B., and I.J. analyzed data; and H.J.P.M.K., A.M.H.B., and I.J. wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: H. Koenen, Radboud University Nijmegen Medical Centre, Department of Blood Transfusion and Transplantation Immunology (Route 469), Postbox 9101, 6500 HB Nijmegen, The Netherlands; e-mail: h.koenen@abti.umcn.nl.