Retinoic acid inhibits Th17 polarization and enhances FoxP3 expression through a Stat-3/Stat-5 independent signaling pathway

Immune responses are orchestrated by subsets of CD4⁺ helper and effector T cells. Classically, differentiated T cells have been classified as belonging either to T-helper 1 (Th1) or Th2 lineages.¹  Th1 cells secrete interferon-gamma (IFN-γ) in response to interleukin-12 (IL-12) and require the transcription factors T-box21 (T-bet) and signal transducer and activator of transcription 4 (Stat4)²  and Stat1, whereas Th2 cells secrete IL-4, IL-5, and IL-13 and require the transcription factors GATA-binding protein 3 (GATA-3) and Stat6.³  While this simple paradigm was instrumental in understanding immune responses to model pathogens, it was less clear how such a model explained the pathogenesis of autoimmune disease.

More recently, additional fates for CD4⁺ T cells have emerged that provide more insights into the mechanisms of tolerance and immune-mediated disease. One of these new subsets of T cells, expressing the transcription factor FoxP3, is termed regulatory T cells (Tregs) and serves to inhibit immune responses. Deficiency of FoxP3 in mice and humans is associated with fatal autoimmune disease.^4,5  FoxP3⁺ Tregs normally develop in the thymus, but FoxP3 expression and Treg function also develop when naive CD4⁺ T cells are stimulated in vitro with transforming growth factor-β1 (TGF-β1) and IL-2.⁶  Development and differentiation of Tregs and expression of FoxP3 are dependent upon another transcription factor, Stat5.^7,8 

Another new subset of T cells that may be related to Tregs is a subset that produces the proinflammatory cytokine IL-17 (Th17 cells). Overproduction of IL-17 has been noted in a variety of models of autoimmune disease, including experimental autoimmune encephalomyelitis and collagen-induced arthritis.^9-13  Importantly, increased levels of IL-17 are also seen in human diseases, such as multiple sclerosis, rheumatoid arthritis, and inflammatory bowel disease.^14-16  Th17 cells are generated by the cytokines TGF-β1 and IL-6, the latter of which acts via Stat3 to induce the transcription factor retinoic acid orphan receptor gamma (RORγt).¹⁷  Overexpression of RORγt induces IL-17 production, whereas deficiency of this transcription factor virtually abrogates Th17 differentiation.

RORγt belongs to a superfamily of retinoid nuclear receptors that include the retinoic acid receptors (RARs), retinoid X receptors (RXRs), and other RORs.¹⁸  A connection between retinoids and CD4⁺ T-cell differentiation has long been recognized. Memory T cells from vitamin A–deficient mice have been found to overproduce IFN-γ and underproduce interleukin-4 (IL-4).¹⁹  In contrast, administration of the active metabolite of vitamin A, retinoic acid (RA), has been reported to suppress memory cell IFN-γ production and increase IL-4 secretion.²⁰  Later, it emerged that vitamin A can be metabolized into retinoic acid isomers with different ligand-receptor affinities: 9-cis retinoic acid, with a higher affinity for the RXR family, and all-trans retinoic acid (ATRA), with a higher affinity for the RAR family.²¹  Acting through RXRα, 9-cis retinoic acid skews naive T cells toward the Th2 phenotype.²²  In turn, ATRA actually prevents naive T cells from differentiating into either the Th1 or Th2 phenotypes.²⁰ 

Consistent with the inhibitory effects of ATRA observed in vitro, ATRA analogs have shown benefit in the treatment of a variety of autoimmune diseases.^23-25  For instance, ATRA has long been documented to inhibit EAE in animal models.^24,26  However, a mechanism underlying these findings has not been defined. In light of the recently recognized importance of Th17 and Treg cells in the pathophysiology of EAE, we explored a possible role for ATRA in Th17/Treg polarization.

Wild-type C57BL/6 mice were purchased from Taconic (Hudson, NY). C57BL/6 mice bearing loxP-flanked (fl/fl) alleles of Stat5 and Stat3 and expressing Cre under the control of CD4 have been described previously,^7,27  as have FoxP3 enhanced green fluorescent protein (eGFP) reporter mice.²⁸  Cre-negative littermates or wild-type C57BL/6 mice were used as controls. Il2^−/− mice on a Bl10 Rag2^−/− TcR transgenic (5C.C7) background were obtained from Taconic, along with Bl6 Rag2^−/− TcR transgenic (OTII) animals. All animals were handled and housed in facilities in accordance with National Institutes of Health Animal Care and Use Committee guidelines.

Peripheral T cells were obtained from spleens and lymph nodes of 8- to 12-week-old mice by mechanical disruption through a 70-μm cell strainer. Bulk CD4⁺ T cells were obtained using positive selection with anti-CD4 microbeads using an autoMACS (Miltenyi Biotec, Bergisch Gladbach, Germany). CD4⁺CD62L⁺ cells were obtained by first enriching for CD4⁺ T cells via negative selection using a Mouse CD4⁺ T Cell Isolation Kit (Miltenyi Biotec), followed by positive selection using indirect labeling with anti-CD62L allophycocyanin (APC)-conjugated antibodies (eBioscience, San Diego, CA) and anti-APC microbeads (Miltenyi Biotec). CD4⁺ single-positive thymocytes were obtained by negative selection with anti-CD8 microbeads (Miltenyi Biotec). CD11c⁺ cells were obtained from mouse splenocytes by positive selection with anti-CD11c microbeads (Miltenyi Biotec). Flow cytometry–sorted naive CD4⁺ T cells were obtained by surface staining cells with anti-CD4, anti-CD62L, anti-CD44, and anti-CD25 antibodies (eBioscience), followed by flow cytometric cell sorting using a Mo-Flo cell sorter (Dako, Carpinteria, CA). Cell purity was routinely greater than 99.5%.

All cell cultures were performed in RPMI 1640 (Invitrogen, Carlsbad, CA) supplemented with 10% fetal calf serum, 2 mM glutamine, 100 IU/mL penicillin, 0.1 mg/mL streptomycin, 0.25 μg/mL amphotericin B, and 2 μM β-mercaptoethanol. Cells were activated at a concentration of 10⁶ cells/mL media in 96-well cell culture plates (Nalge Nunc, Rochester NY) either with plate-bound anti-CD3 (5 μg/mL) and anti-CD28 (5 μg/mL) as indicated (BD PharMingen, San Diego, CA) or 2.5 × 10⁵ CD11c⁺ cells/mL pulsed with 1 mM OVA peptide (ISQAVHAAHAEINEAGR; Peptides International, Louisville, KY). Th-neutral conditions (Th0) contained no exogenous cytokines or anticytokines. Th1 cultures were supplemented with 10 ng/mL IL-12 (R&D Systems, Minneapolis, MN) with 10 μg/mL anti–IL-4 (BD PharMingen); Th2 conditions contained 10 ng/mL IL-4 (R&D Systems) with 10 μg/mL anti–IFN-γ (BD PharMingen); and Th17 conditions contained 10 ng/mL IL-6 (Invitrogen, Carpinteria, CA), 5 ng/mL TGF-β1 (R&D Systems), 10 μg/mL anti–IL-4, and 10 μg/mL anti–IFN-γ. Where indicated, IL-2 was added at 100 IU/mL and TGF-β1 alone at 5 ng/mL. ATRA, TTNBP (4-[E-2-(5,6,7,8-tetrahydro-5,5,8,8-tetramethyl-2-naphthalenyl)-1-propenyl]) benzoic acid, and A7980 were obtained from Sigma (St Louis, MO). LE540 and AM580 were obtained from Wako (Osaka, Japan). All retinoids were dissolved in dimethyl sulfoxide (DMSO) at stock concentrations of 0.01 M and stored at −80°C in lightproof containers. Stocks were thrown away after 4 freeze-thaw cycles. Cultures containing retinoids were protected from light throughout the time of culture; unless stated otherwise, ATRA was used at a final concentration of 1 μM.

Various numbers (2500 to 40 000) of natural (CD4⁺ CD25⁺) Tregs or induced (CD4⁺ FoxP3-GFP⁺) Tregs were added to 50 000 naive (CD4⁺ CD25⁻) T cells and 50 000 irradiated Thy-1⁻ cells in a total volume of 200 μL media with soluble anti-CD3 at 1 μg/mL. After 4 days, 0.2 μCi of ³H-thymidine (Amersham Biosciences, Piscataway, NJ) was added, and incorporation was measured after a further 6 hours of incubation using a scintillation counter (PerkinElmer, Waltham MA).

Anti–IFN-γ-APC and fluorescein isothiocyanate (FITC), anti–IL-17-PE, and anti–IL-4-PE and APC antibodies for intracellular cytokine staining were obtained from BD Biosciences. Anti–IL-17 Alexa Fluor 647 and anti–FoxP3-FITC and APC antibodies were obtained from eBioscience. In brief, cells were stimulated for 4 to 5 hours with phorbol myristate acetate, ionomycin, and Brefeldin A. Cells were stained using the Cytofix/Cytoperm Kit (BD Biosciences) with the indicated antibodies. Events were analyzed on a FACS Calibur flow cytometer (BD Biosciences, San Jose, CA). Events were collected and analyzed with FlowJo software version 8.2 (Tree Star, Ashland, OR). Cytokine concentrations of IFN-γ, IL-4, and IL-17 in cell culture supernatants were analyzed by enzyme-linked immunosorbent assay (ELISA) using the appropriate Quantikine kits (R&D Systems).

Total RNA was isolated from cell cultures using the RNeasy kit (Qiagen, Hilden, Germany). RNA was quantitated using a NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE). RNA concentrations were equalized for each experiment before making cDNA using random hexamers and reverse transcriptase (Applied Biosystems, Foster City, CA). Quantitative PCR was performed with an ABI 7500 Fast Real-Time PCR System using Taqman site-specific primers and probes (Applied Biosystems). Results were normalized to β-actin levels. For comparisons, Th-neutral culture conditions were assigned an arbitrary value of 1.

Unless otherwise stated, statistical analysis was performed using the 2-sample, two-tailed Student t test. P values less than .05 were considered significant.

We began our studies by revisiting the effect of ATRA on the polarization of the traditional CD4 T-cell subsets. CD4⁺CD62L⁺ T cells from wild-type C57Bl6 mice were stimulated for 2 days in vitro under neutral, Th1-favoring (IL-12, anti–IL-4), or Th2-favoring (IL-4, anti–IFN-γ) conditions with or without 1 μM ATRA and then expanded for 3 days with polarizing cytokines and IL-2, again with or without ATRA (Figure 1A,B). As expected, the different polarizing conditions led to selective secretion of the signature cytokines. Moreover, consistent with previous reports,^29,30  ATRA reduced IFN-γ and IL-4 production by both Th1 and Th2 cells, respectively.

We next turned our investigation to the effect of retinoids on Th17 differentiation. Naive CD4⁺ T cells from Rag2^−/− TCR-transgenic (OTII) mice were stimulated using CD11c⁺ cells pulsed with cognate peptide for 3 days in vitro under established Th17-favoring conditions (IL-6, TGF-β1, anti–IL-4, and anti–IFN-γ) with or without 1 μM ATRA. On the third day, IL-17 expression was determined by intracellular staining. In the absence of ATRA, the conditions generated substantial proportions of IL-17-producing cells, which were reduced by inclusion of this compound (Figure 2A). We confirmed that this inhibition was dose dependent in polyclonally stimulated naive CD4⁺ T cells stimulated under Th17 conditions by ELISA (Figure 2B). At least in vitro, the differentiation of Th17 cells is related to another subset, inducible Treg cells, as TGF-β1 appears to contribute to the generation of both.²⁸  It was notable, therefore, that in addition to inhibiting Th17-cell generation, ATRA also increased the percentage of FoxP3⁺ cells (Figure 2A) and FoxP3 mRNA (Figure 2C) under otherwise Th17-favoring conditions, the latter in a dose-dependent manner. It should be noted that in contrast with Figure 2A, the differentiation of T cells in Figure 2B,C was induced without antigen-presenting cells, indicating that ATRA must be acting directly upon T cells.

The retinoic acid receptor family is divided into RAR, RXR, and ROR subfamilies, and RAR and RXR family members are known to bind ATRA. We began our investigation by examining which retinoic acid family receptors are expressed by each T-cell subset. Naive CD4⁺ T cells from wild-type C57Bl6 mice were stimulated in vitro for 3 days with plate-bound anti-CD3 and anti-CD28 antibodies under polarizing conditions for Th-neutral, Th1, Th2, or Th17 cells (Figure 3A,B). Receptor expression was analyzed by quantitative RT-PCR. Consistent with the reports of other groups, we found that Th2 cells preferentially express high levels of RXRα.³¹  However, we also found that Th17 cells were distinct in expressing high levels of RARα and RARγ. In addition, Th17 cells also express high levels of the orphan receptor RORγt, which was downregulated by ATRA (Figure 3C).

To test the potential functional significance of RAR in Th17 cells, we compared the effects of (4-[E-2-(5,6,7,8-tetrahydro-5,5,8,8-tetramethyl-2-naphthalenyl)-1-propenyl] benzoic acid (TTNBP), a pan-RAR agonist, and LE540, a pan-RAR antagonist, on FoxP3 and IL-17 expression (Figure 4A). Like ATRA, TTNBP increased the proportion of FoxP3⁺ cells, with a reciprocal decrease in the proportion of IL-17–producing cells. In contrast, the RAR antagonist, LE540, had the opposite effect, actually reducing the ability of TGF-β1 to induce FoxP3, without noticeable effects on IL-17 production. Next, we assessed the relative contributions of RARα and RARγ to FoxP3 expression in naive cells by using the synthetic retinoids AM580 and A7980, selective RARα and RARγ agonists, respectively (Figure 3B). Whereas AM580 was as effective as ATRA at increasing the percentage of FoxP3⁺ cells, A7980 had little demonstrable effect, thus strongly implying a role for RARα as the key player in FoxP3 expression. These data are also notable in that they imply a direct role for retinoids in controlling FoxP3, as the studies were performed using isolated T cells.

Several lines of evidence point to key roles for Stat5 and Stat3 as essential regulators of FoxP3 and IL-17.^27,32-34  Specifically, in the presence of TGF-β1, IL-2 acting through Stat5 is known to increase FoxP3 and inhibit IL-17. In contrast, IL-6 acting through Stat3 increases IL-17 and inhibits FoxP3. The ability of ATRA to both reduce IL-17 and enhance FoxP3 expression in polyclonally stimulated T cells in the presence of IL-6 and TGF-β1 was reminiscent of the effects of altering Stat3 and Stat5 signaling. We therefore examined the relationship between the effect of ATRA and the requirement for these transcription factors. Using T cells from Stat3^fl/flCD4-Cre mice, we first investigated whether ATRA might be inhibiting Th17 cells and promoting FoxP3 indirectly by blocking IL-6 signaling through Stat3 (Figure 5). In the absence of Stat3, the combination of TGF-β1 and IL-6 failed to induce IL-17–producing T cells, as noted previously. In this setting, these cytokines produced a higher proportion of FoxP3⁺ T cells. Nonetheless, the presence of ATRA was able to further increase the expression of FoxP3 in the absence of Stat3, suggesting that ATRA enhances FoxP3 in a Stat3-independent manner.

Next, using thymocytes as a source of naive cells from tissue-specific Stat5^fl/flCD4-Cre knockouts, we evaluated the idea that ATRA might affect Stat5 signaling. As we have shown previously, the addition of IL-2 inhibited IL-17 in a Stat5-dependent manner. However, in the absence of Stat5, ATRA still was able to inhibit IL-17 and combine with TGF-β1 to enhance expression of FoxP3, even in the absence of Stat5 (Figure 6). Thus, we conclude from these data that the negative regulatory effect of ATRA on IL-17 expression and its positive effect of FoxP3 are independent of the Stats typically involved in controlling expression of these key target genes.

Previous work has shown that the presence of IL-2 is critical to induce FoxP3 expression in naive T cells.⁷  The ability of ATRA to induce FoxP3 expression in Stat5-deficient CD4⁺ T cells led us to investigate whether ATRA could replace the action of IL-2 as a cofactor for FoxP3 induction by TGF-β1. Using Rag2^−/− TCR-transgenic (OTII) naive T cells, we examined the effect of the addition of ATRA with or without addition of exogenous IL-2 or its blocking antibody. As shown in Figure 7A, TGF-β1 in the presence of anti–IL-2 was a modest inducer of FoxP3 expression, whereas the addition of human IL-2 greatly increased the effect of TGF-β1. In contrast, the combination of ATRA and TGF-β1 in the presence of anti–IL-2 generated substantial proportions of FoxP3-positive cells (35%). Addition of exogenous IL-2 increased the proportions even further (>90%). In the presence of TGF-β1 and IL-6 and absence of IL-2, substantially greater numbers of IL-17–producing cells are generated, as previously reported.^27,35  The addition of ATRA inhibited IL-17 expression in the absence of IL-2 while enhancing FoxP3 expression (Figure 7A, lower right panel). We confirmed these results using IL-2–deficient Rag2^−/− TCR-transgenic (5C.C7) naive T cells (Figure S1, available on the Blood website; see the Supplemental Materials link at the top of the online article). We next determined whether ATRA-induced FoxP3 cells in the presence of anti–IL-2 were as suppressive as natural (CD4⁺CD25⁺) Tregs (Figure 7B). Using a transgenic FoxP3-GFP mouse, GFP-positive cells induced by 3 days of polyclonal stimulation in the presence of TGF-β1, ATRA, and anti–IL-2 were isolated by flow cytometry and show a similar ability to inhibit polyclonally stimulated CD4⁺ CD25⁺ cells in comparison to natural Tregs.

In this study, we examined the role of the vitamin A metabolite ATRA in Th17 generation and FoxP3 induction. We found that as previously reported, ATRA negatively regulates expression of the effector cytokines IFN-γ and IL-4. In addition, we found that ATRA also directly inhibits T-cell production of the key inflammatory cytokine IL-17. However, given the putative relatedness of Th17 and Treg cells, it was even more striking that, concomitant with reduced IL-17 expression, we noted induction of FoxP3 expression by ATRA. Furthermore, the finding that inhibition of RAR function in isolated T cells inhibited Foxp3 expression and enhanced IL-17 production argues for a role of endogenous retinoids in regulating the balance between Treg and Th17 differentiation. Because of the increasingly prominent roles of these subsets in many autoimmune and autoinflammatory disorders, there is great interest in the relationship between Th17 and Treg cells, both of which can be induced in vitro in the presence of TGF-β1. The present data point to a role of retinoids acting via RARα as the likely candidate for mediating this effect.

Our findings argue that in addition to the well-known regulators of Th17/Treg differentiation, retinoids appear to be another physiologic regulator of CD4⁺ T-cell lineage commitment. The strongest data for this argument are the dramatic effects of inhibiting RARα. In addition, these data provide potential insights into the regulation of CD4⁺ T-cell differentiation in the gut. Mucosal dendritic cells (DCs) are well known to be immunosuppressive.³⁶  These subsets also influence the tropism of T cells by influencing expression of homing receptors.³⁷  Recently, dendritic cells in the gut-associated lymphatic tissue have been shown to synthesize ATRA from dietary beta-carotene.³⁸  The ability of mucosal DCs to produce retinoids may explain why the relative distribution of Th17 cells in the gut is limited, although IL-6 expression is widespread.^17,39  It is possible that DCs use retinoic acid as a T-cell modulator, increasing secretion to suppress inflammation and decreasing production when an inflammatory response is warranted. Thus, retinoid receptors and local production of retinoids may represent novel physiologic regulators of the Th17/Treg axis.

The mechanisms by which retinoids regulate IL-17 and FoxP3 expression are not yet clear. Based on expression and the effect of a selective antagonist, our data would argue that the major mediator of the effect of ATRA is RARα. However, deficiency of RARα results in early postnatal lethality⁴⁰ ; therefore, establishing the precise role of RARα will await tissue-specific deletion strategies. It is notable that ATRA treatment is also associated with downregulation of RORγt, the RAR-related transcription factor that drives Th17 differentiation. It will be important to determine the mechanisms by which RARα influences RORγt expression in CD4⁺ T-cell differentiation. Although Stat3 and Stat5 are key regulators of Th17/Treg differentiation,^{6,7,27,32-34,41}  our data indicate that ATRA appears to exert its effect independently of these factors. As a transcription factor itself, RARα might directly bind the promoters of the Foxp3, Il17, and/or RORγ genes; in the latter case, the expectation would be that it serves as a repressor. Nonetheless, our data suggest a model in which TGF-β1, Stat5, and RARα each provide unique and additive signals to increase FoxP3 expression. Conversely, Stat5 and RARα both contribute to IL-17 suppression, antagonizing the positive effects of TGF-β1 and Stat3.

Beyond potential roles as physiologic regulators of CD4⁺ T-cell differentiation, retinoids and retinoid receptors are attractive pharmacologic agents. The ability to affect Th17/Treg polarization has attracted great clinical interest.^42-44  Altering this ratio in favor of increased numbers of Treg cells would be useful for a wide range of autoimmune and autoinflammatory diseases. Conversely, suppressing Treg formation has potential benefits for developing tumor immunotherapies. In fact, ATRA analogs have shown efficacy in in vivo models of autoimmune disease.^23,25,26  This suggests that retinoid-based pharmaceutical strategies may offer a promising alternative to our current approach of managing immunologic diseases. It will be important to elucidate precisely how retinoids regulate IL-17 and FoxP3 to optimize therapeutic regimens and to generate agents with increased selectivity in these actions.

In summary, ATRA and other ligands for RARα inhibit Th17 polarization and promote FoxP3 expression. This occurs independently of IL-2, Stat5, and Stat3, which have been implicated previously in Th17 and Treg homeostasis. Not only may this be an important physiologic regulator of T-cell lineage commitment but, in addition, ATRA analogs are likely to be useful as pharmacologic agents. Understanding the precise role for retinoic acid in normal Treg development and function remains an interesting subject for future investigation.

K.M.E. is a Howard Hughes Medical Institute–National Institutes of Health Research Scholar.

Contribution: K.M.E. and A.L. designed and performed experiments. T.D., G.S., and Y.K. designed experiments and helped analyze data. E.S. and J.J.O. supervised planning of experiments and assisted in data interpretation.

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

Correspondence: Arian Laurence, Building 10, Room 9N262, 10 Center Drive, MSC-1820, Bethesda, MD 20892-1820; e-mail: laurencea@mail.nih.gov.