Activated β-catenin in Foxp3+ regulatory T cells links inflammatory environments to autoimmunity.

Foxp3+ regulatory T cells (Treg cells) are the central component of peripheral immune tolerance. Whereas a dysregulated Treg cytokine signature has been observed in autoimmune diseases, the regulatory mechanisms underlying pro- and anti-inflammatory cytokine production are elusive. Here, we identify an imbalance between the cytokines IFN-γ and IL-10 as a shared Treg signature present in patients with multiple sclerosis and under high-salt conditions. RNA-sequencing analysis on human Treg subpopulations revealed β-catenin as a key regulator of IFN-γ and IL-10 expression. The activated β-catenin signature was enriched in human IFN-γ+ Treg cells, as confirmed in vivo with Treg-specific β-catenin-stabilized mice exhibiting lethal autoimmunity with a dysfunctional Treg phenotype. Moreover, we identified prostaglandin E receptor 2 (PTGER2) as a regulator of IFN-γ and IL-10 production under a high-salt environment, with skewed activation of the β-catenin-SGK1-Foxo axis. Our findings reveal a novel PTGER2-β-catenin loop in Treg cells linking environmental high-salt conditions to autoimmunity.

Introduction

The homeostatic maintenance of T cells is finely tuned by Treg cells. Treg cells play a distinct role from the other CD4+ T cells in dampening prolonged inflammation and preventing aberrant autoimmunity[1]. Although Treg cells are potent suppressors of immune function, the number of Treg cells is often normal in a variety of autoimmune diseases, including multiple sclerosis (MS)[2, 3]. These observations suggest that not only a quantitative, but also a functional dysregulation of Treg cells contributes to the development of autoimmunity.

Treg cells display their suppressive capacity through both contact-dependent and cytokine-mediated mechanisms[4]. Treg cells demonstrate substantial heterogeneity and the balance between pro- and anti-inflammatory populations is finely regulated to maintain immunologic homeostasis[4]. IFN-γ marks dysfunctional Treg cells in patients with autoimmunity (MS[5] and T1D[6]) and cancer (glioblastoma[7]). Additionally, Treg cells producing the anti-inflammatory cytokine IL-10 play prominent roles in suppressing the immune response at environmental interfaces and development of mature memory CD8+ T cells to prevent autoimmunity and chronic infection in mice[8, 9]. These studies suggest that the balance between IFN-γ and IL-10 production in Treg cells is central in the maintenance of immune homeostasis; however, the molecular mechanisms underlying this regulatory balance are not known.

Human autoimmune disease results from an interplay between genetic factors and environmental triggers. In this regard, MS is an autoimmune disease that results from the complex interaction of predominantly common genetic variants and environmental factors[10], with 233 common risk haplotypes identified to date[11,12]. Several environmental factors are associated with an increased risk of MS including vitamin D insufficiency, smoking, obesity, and a high salt diet (HSD)[13]. Previous studies showed that a HSD exacerbated neuroinflammation in the experimental autoimmune encephalomyelitis (EAE) model of MS, and that higher salt concentration within the physiological range skewed naive CD4+ T cells into pro-inflammatory TH17 cells and impaired Treg suppressive function through induction of IFN-γ expression[14, 15, 16]. Studies using murine models of autoimmune disease are accumulating to support this theory[17, 18] and recent magnetic resonance imaging studies revealed higher sodium intensity in acute MS lesions compared to chronic lesions, suggesting more sodium accumulation within the pathogenic microenvironment in MS brain[19]. However, it remains unknown whether a high salt diet has a direct impact on MS clinical activity[20].

β-catenin is an essential component of the canonical Wnt signaling pathway and involved in a variety of biological processes including carcinogenesis, stem cell maintenance, organogenesis, and aging[21, 22]. Although β-catenin and canonical Wnt signaling have been studied in immune system, the specific mechanisms by which β-catenin affects Treg function and their role in modulating cytokine production by Treg cells, in particular in the context of human autoimmune disease, is poorly understood.

Here, we show that the imbalance between IFN-γ and IL-10 is a shared Treg signature observed in the patients with MS and high salt environment. By performing unbiased RNA-seq analysis on human Treg subpopulations, we identify β-catenin as central in maintaining Treg function and regulating both IFN-γ and IL-10 cytokine production. Moreover, we clarify a role for β-catenin in mediating the high salt-induced pro-inflammatory signature by creating a feed forward loop with PTGER2, which is uniquely upregulated under high salt conditions. Our findings suggest that the β-catenin-PTGER2 axis serves as a bridge between environmental factors and autoimmune disease by modulating Treg function and this axis may be involved in the pathogenesis of autoimmune disease.

Results

Treg cytokine imbalance in multiple sclerosis and high salt environment

Previous studies have identified a pro-inflammatory Treg population characterized by the secretion of IFN-γ. This population is dysfunctional both in vitro and in vivo, and a high frequency of this population is associated with autoimmune disease[5, 6, 7]. However, the balance between pro- and anti-inflammatory Treg populations has not been defined. To address this question, we evaluated the production of pro-inflammatory (IFN-γ) and anti-inflammatory (IL-10) cytokines by circulating human Treg cells from healthy subjects and patients with MS by flow cytometry. Based on our observation that CD25hiCD127lo-negCD45RO+ Treg cells (memory Treg cells; mTreg cells) are the major source for effector cytokine expression in human Treg cells (Supplementary Fig. 1), we focused on mTreg cells, so as to avoid the potential bias caused by the variable ratio of naive Treg cells and mTreg cells between subjects. We found that mTreg cells isolated from MS patients (MS-Treg) produced more IFN-γ and less IL-10 compared to healthy donors, and the ratio of IFN-γ to IL-10 producing Treg cells further highlights this imbalance (Fig. 1a, b). Furthermore, we examined the mRNA expression of IFNG and IL10 genes in mTreg cells without PMA plus iomomycin stimulation, better reflecting the situation in vivo, and identified a trend similar to that seen in protein expression (Fig. 1c).

Figure 1.

IFN-γ:IL-10 balance of human Treg cells in MS and high salt treatment.

(a) Representative flow cytometric analysis of ex vivo human Treg cells (CD4+CD25hiCD127lo-negCD45RO+) isolated from healthy donor and MS patient. FACS isolated Treg cells were stimulated with PMA plus iomomycin for 4 h followed by intracellular staining for IFN-γ and IL-10. Data are representative of twelve experiments. (b) IFN-γ and IL-10 cytokine profiles of ex vivo human Treg cells. Percentage of IFN-γ and/or IL-10 producing Treg cells was shown (left). Ratio between IFN-γ positive and IL-10 positive Treg cells was plotted (right). (HC; n=12, MS; n=14 subjects) *P<0.05, **P<0.01 (two-way ANOVA with Sidak’s multiple comparisons test (left), two-sided Student’s t-test (right)). (c) IFNG and IL10 gene expression was determined on ex vivo Treg cells by qPCR. Ratio between IFNG and IL10 expression was shown at the right (HC; n=36, MS; n=27 subjects). P values were calculated by two-sided Student’s t-test. (d) Representative flow cytometric analysis of IFN-γ and IL-10 production in human Treg cells stimulated with anti-CD3 and anti-CD28 in the normal media (Control) or media supplemented with additional 40 mM NaCl (NaCl) for 96 h. Data are representative of eight experiments. Percentage of IFN-γ and/or IL-10 producing Treg cells was shown at the bottom (n=21 subjects). P values were calculated by two-sided Student’s t-test. (e) IFNG and IL10 mRNA expression was assessed at 96 h following stimulation as in (d) (n=41 subjects). Ratio between IFNG and IL10 expression was plotted (right). Each symbol represents an individual subject. P values were calculated by two-sided Student’s t-test. (f) mRNA expression kinetics of IFNG and IL10 from nine different time points were plotted (n=6 subjects). *P<0.05, **P<0.01, ***P<0.001 (two-way ANOVA with Sidak’s multiple comparisons test). Data were represented as mean (b-e), or mean +/− SEM (e).

We recently demonstrated that Treg cells exposed to high salt concentrations exhibited a dysfunctional phenotype with a pro-inflammatory cytokine signature skewed towards IFN-γ[16]. We sought to determine whether high salt could also impair the IFN-γ:IL-10 balance and found that the high salt environment caused an increase in IFN-γ and decrease in IL-10 production in human mTreg cells after 96 h in culture (Fig. 1d, e). Gene expression kinetics of IFNG and IL10 by using qPCR identified early (8 h) and late (96 h) waves of gene expression. High salt stimulation suppressed the early wave of IFNG and IL10, and enhanced the late wave of IFNG but not IL10 (Fig. 1f). These findings suggest that the imbalance of IFN-γ:IL-10 induced by continuous exposure to high salt conditions, which is not observed at the phase of acute response to high salt stress, might capture the dysfunctional Treg properties in the setting of autoimmunity.

β-catenin regulates relative production of IFN-γ and IL-10 in human Treg cells

The molecular mechanisms underpinning the balance between IFN-γ and IL-10 in Treg cells were largely unknown. To address this question, we performed RNA-sequencing (RNA-seq)-based genome-wide transcriptome analysis on human Treg subsets defined by IFN-γ and IL-10 production. mTreg cells isolated from peripheral mononuclear cells of healthy subjects were stimulated with PMA plus iomomycin for 4 h ex vivo. After applying cytokine capture kits for IFN-γ and IL-10, we sorted four different subpopulations (IFN-γ single positive (IFN-γSP), IL-10 single positive (IL10SP), IFN-γ and IL-10 double positive (DP), and double negative (DN)), and we performed RNA-seq on each subpopulation (Fig 2a). We identified 672 differentially expressed genes between IFN-γSP and IL10SP and the four populations could be distinguished by their gene expression profiles (Fig. 2b). Of note, the IFN-γ-producing populations were highly distinct from IFN-γ-negative populations, suggesting that IFN-γ-secreting Treg cells represent a more dominant signature than IL-10-secreting Treg cells. We also identified ten clusters of co-expressed genes (C1-C10) across the populations. IFN-γ and IL-10-associated genes are enriched in C9 and C10 (e.g. CXCR3, CD226, and NKG7) and C1 and C2 (e.g. MAF, SOCS3, and NOTCH2), respectively.

Figure 2.

Transcriptional profiling of IFN-γ and IL-10 producing human Treg subsets.

(a) Experimental workflow for RNA-seq with IFN-γ and/or IL10 producing Treg subpopulations. (b) Heatmap of 672 differentially expressed genes between IFN-γSP and IL10SP. 10 clusters are identified and representative genes for each cluster are shown. (c) GSEA enrichment plots of KEGG Wnt signaling pathway between INF-γSP vs. DN, INF-γSP vs. IL10SP, IL10SP vs. DN, and DP vs. DN. Results were calculated from three subjects performed in the same batch. Normalized enrichment score (NES) and false discovery rate (FDR) were represented at the bottom of each plot.

To predict the key transcriptional regulators that account for IFN-γ and IL-10 production, we performed an upstream regulator analysis in Ingenuity Pathway Analysis (IPA), using differentially expressed genes from each population (Supplementary Table 1). We identified β-catenin (CTNNB1) as one of the top upstream regulators in the Treg populations producing IFN-γ and/or IL-10 compared to DN. Intriguingly, β-catenin was ranked as the top-ranked upstream regulator in the comparison between IFN-γSP and IL10SP. These results suggest that β-catenin plays a critical role in driving the production of both IFN-γ and IL-10 in Treg cells, especially for IFN-γ. We also identified several upstream regulators that have been demonstrated to have critical roles in maintaining Treg function, including MYB, SATB1, NFATC2, and KLF2, suggesting that our upstream regulator analysis provides a reliable readout.

In agreement with these findings, gene-set-enrichment analysis (GSEA) identified significant enrichment of the Wnt-β-catenin signaling pathway in IFN-γ-producing Treg subsets (Fig. 2c). IFN-γSP exhibited the highest enrichment score for the Wnt-β-catenin signaling pathway. Further GSEA analysis with different gene sets also provided similar results (Supplementary Fig. 2a). Taken together, these findings indicate that Wnt-β-catenin signaling is more activated in IFN-γ-secreting Treg cells than in other human Treg subpopulations.

β-catenin is stabilized in the IFN-γ secreting Treg population

We first confirmed that β-catenin is stabilized and transcriptionally active in IFN-γSP compared to DN in ex vivo Treg cells by examining the level of Active β-catenin, the dephosphorylated form of β-catenin with established active transcriptional activity[23] (Fig. 3a). Notably, the DP and IL10SP also exhibited increased Active β-catenin expression compared to DN ex vivo, suggesting that β-catenin signaling is important not only for IFN-γ but also for IL-10 production in Treg cells, consistent with our upstream regulator and enrichment analyses. To exclude the possibility that PMA plus iomomycin stimulation affected β-catenin stability, we measured Active β-catenin levels in CXCR3+ TH1-like Treg cells, which contain most of the IFN-γ-producing Treg cells[24] without PMA plus iomomycin stimulation; these analyses confirmed that Active β-catenin expression was significantly increased in the CXCR3+TH1-like Treg population (Fig. 3b). In agreement with these data, the downstream β-catenin target genes, AXIN2 and TCF7, and the protein TCF-1 (encoded by TCF7) were upregulated in IFN-γSP compared to DN ex vivo (Fig. 3c, Supplementary Fig. 2b). This was consistent with previously published microarray data for IFN-γ-positive and IFN-γ-negative Treg cells[25] (Supplementary Fig. 2c).

Figure 3.

β-catenin is stabilized in the IFN-γ secreting Treg subpopulation.

(a) Relative expression level of Active β-catenin on ex vivo Treg subpopulations analyzed by flow cytometry (n=11 subjects). Fold change in gMFI over DN were depicted. *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). gMFI, geometric mean fluorescence intensity. (b) Expression level of Active β-catenin between CXCR3- and CXCR3+ ex vivo Treg cells from healthy controls. Representative histogram (left) and summary of results (n=7 subjects) (right). P values were calculated by two-sided Student’s t-test. (c) Gene expression of Wnt−β-catenin signaling target genes (AXIN2 and TCF7) assessed by RNA-seq (n=8 subjects). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (d) Relative expression level of Active β-catenin on Treg cells stimulated with anti-CD3 and anti-CD28 for 4 days, followed by 4 h PMA plus iomomycin stimulation and intracellular cytokine staining for IFN-γ and IL-10 (n=12 subjects). Fold change in gMFI over DN were depicted. *P<0.05, **P<0.01, ***P<0.001 (one-way ANOVA with Tukey’s multiple comparisons test). (e) Expression level of β-catenin on Treg cells stimulated with anti-CD3 and anti-CD28 in the presence (TH1) or absence (TH0) of IL-12 for 4 days. β-catenin level on IFN-γ positive/negative Treg populations was determined after 4 h PMA plus iomomycin stimulation (left) (n=4 subjects). Representative histogram for β-catenin expression was shown (right). P values were calculated by two-sided Student’s t-test. (f, g) Frequency of IFN-γ and IL-10 positive cell number (f) and gene expression of IFNG and IL10 by qPCR (g). Treg cells were stimulated with anti-CD3 and anti-CD28 in the presence of Wnt/β-catenin signaling inhibitor PKF115–584 (PKF), IL-12 (TH1), or IL-12 and PKF115–584 (TH1+PKF) (n=4 subjects) *P<0.05, **P<0.01, ***P<0.001 (one-way ANOVA with Tukey’s multiple comparisons test). (h, i) Relative frequency of IFN-γ and IL-10 positive cell number (fold of scramble shRNA/control condition) (h) and gene expression of IFNG and IL10 by qPCR (i). Treg cells were transduced with a scramble shRNA or a CTNNB1 shRNA and cultured in TH0 or TH1 condition for 5 days (h; n=7 subjects, i; n=5 subjects). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). Data are representative of two experiments (e, f) or are from more than three experiments. Data were represented as mean +/− SD.

To examine whether the in vitro model can recapitulate the ex vivo results, we assayed Active β-catenin levels on each of the Treg subsets after four days of culture with anti-CD3 and anti-CD28 stimulation. IFN-γ-producing Treg populations (IFN-γSP and DP) showed higher Active β-catenin expression compared to IL10SP and DN (Fig. 3d), indicating that stabilization of β-catenin is more enhanced in IFN-γSP compared to IL10SP under T cells receptor (TCR) stimulation. IL-12 is an essential cytokine to induce IFN-γ-producing pathogenic Treg cells under TCR stimulation[5]. We found that upregulation of β-catenin was also observed in IL-12-induced TH1-like Treg cells, especially in the IFN-γ-producing population (Fig. 3e). To determine if Wnt-β-catenin signaling was necessary for IFN-γ production in TH1-like Treg cells, we blocked β-catenin signaling with the β-catenin signaling inhibitor, PKF115–584 (PKF). Treg cells treated with PKF exhibited a significantly reduced production of IFN-γ (Fig. 3f, g). IL-10 production was also suppressed by PKF treatment, albeit less dramatically than IFN-γ. Knocking down the CTNNB1 gene in Treg cells using short hairpin RNA (shRNA) (Supplementary Fig. 2d) ameliorated IL-12-induced IFN-γ and IL-10 production (Fig. 3h, i). These data suggest that β-catenin plays a critical role in IFN-γ and IL-10 induction in human Treg cells, but more profoundly in IFN-γ production under TCR stimulation.

Constitutive activation of β-catenin in Treg cells induces Scurfy-like autoimmunity

To ascertain the physiological relevance of β-catenin signaling in Treg cells, we generated Treg-specific β-catenin stabilized mice by crossing Foxp3-IRES-Cre mice (Foxp3Cre)[26] with Ctnnb1ΔEx3 mice[27] (Supplementary Fig. 3a), where the active form of β-catenin was specifically induced in Treg cells. In these Ctnnb1ΔEx3/Foxp3Cre mice, β-catenin was highly stabilized in Foxp3+ Treg cells, but not on Foxp3— non-Treg cells (Fig. 4a, Supplementary Fig. 3b). Ctnnb1ΔEx3/Foxp3Cre mice spontaneously developed a hunched posture, crusting of the ears, eyelids and tail and showed thymic atrophy, splenomegaly and lymphadenopathy (Fig. 4b). Histologic analysis demonstrated lymphocyte infiltration into several tissues, such as lung, pancreas, liver, and intestine, representing systemic inflammation in Ctnnb1ΔEx3/Foxp3Cre mice (Fig. 4c). This Scurfy-like fulminant autoimmunity led to premature death within 40 days of birth with 100% penetrance (Fig. 4d).

Figure 4.

Treg specific activation of β-catenin induces Scurfy-like autoimmunity.

(a) Flow cytometric analysis of β-catenin on peripheral lymph node Treg cells and Foxp3– CD4+ T cells (CD4T) from Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre mice. Data are representative of four experiments. (b) Images of 4-week-old Foxp3Cre mouse and Ctnnb1ΔEx3/Foxp3Cre mouse (left). Representative pictures of thymus, peripheral lymph nodes, and spleens isolated from 4 week-old Foxp3 or Ctnnb1ΔEx3/Foxp3Cre mice. (c) Hematoxylin and eosin staining of thymus, spleen, liver, intestine, pancreas, and lung sections from 4 week-old Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre mice. Scale bars, 300 μm in the lower magnification and 150μm in the higher magnification. Results in (b, c) are representative of six experiments. (d) Survival of Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre mice. (e) The percentage of Treg cells within CD4+ T cells and the cell numbers of Treg cells in spleen (top) and thymus (bottom) from Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre mice at 3 weeks (3wks) and 5 week old (5wks) of age (n=2–4 mice). (f) Flow cytometric analysis of CD4+ and CD8+ T cells in peripheral lymph nodes and spleens from Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre mice at the age of 3 weeks. Cell count and percentages of CD4+ and CD8+ T cells among CD3+ T cells from the spleens were shown at the bottom (n=3–4 mice). **P<0.01 ***P<0.001, ****P<0.0001 (two-way ANOVA with Sidak’s multiple comparisons test). Data were represented as mean +/− SD. Data are representative of three experiments.

The balance between Treg cells and effector T cells is critical to maintain T cell homeostasis both in central and peripheral lymphoid tissue. The percentage of Treg cells within thymic CD4+ T cells of Ctnnb1ΔEx3/Foxp3Cre mice remained at the same level as Foxp3Cre mice by the age of 3 weeks and even increased at the age of 5 weeks; however, the number of Treg cells in thymus began to decline at the age of 3 weeks in Ctnnb1ΔEx3/Foxp3Cre mice (Fig. 4e). In contrast, Ctnnb1ΔEx3/Foxp3Cre mice displayed an increased number of CD4+ and CD8+ conventional T cells in secondary lymphoid organs (Fig. 4f) and higher expression of effector cytokines such as Ifng, Il4, and Il10, but not Il17a (Supplementary Fig. 3c). Downregulation of Rorc in both Treg cells and conventional CD4+ T cells is the opposite for Ctnnb1ΔEx3/CD4Cre mice[28], highlighting the difference between Treg cells from Ctnnb1ΔEx3/CD4Cre mice and Ctnnb1ΔEx3/Foxp3Cre mice. We examined Helios expression to characterize the functional stability of Treg cells[29], and found that Ctnnb1ΔEx3/Foxp3Cre Treg cells lost Helios expression, supporting the unstable and dysfunctional feature of Ctnnb1ΔEx3/Foxp3Cre Treg cells (Supplementary Fig. 3d). These results suggest that forced expression of a stabilized form of β-catenin in Treg cells influences their functional stability in the periphery more than in the central compartment.

In vitro suppression assay revealed that Ctnnb1ΔEx3/Foxp3Cre Treg cells showed less suppressive activity compared to Foxp3Cre Treg cells (Fig. 5a). Given that the direct interaction of β-catenin with Foxo1 has been reported[30, 31], we noted that the morphological and pathophysiological phenotype of Ctnnb1ΔEx3/Foxp3Cre mice was similar to that of Treg-specific Foxo1 depletion mice, which exhibited fulminant autoimmunity and disrupted Treg function with aberrant IFN-γ expression[32]. To identify transcriptional changes in β-catenin stabilized Treg cells, we measured the gene expression signature of Ctnnb1ΔEx3/Foxp3Cre Treg cells by genome-wide DNA microarrays (Fig. 5b). Further assessment with GSEA revealed similar transcriptional profiles between Ctnnb1ΔEx3/Foxp3Cre Treg cells and Foxo1-depleted Treg cells (Supplementary Fig. 3e). In agreement with this observation, phosphorylated Foxo1 and Foxo3a were increased in Ctnnb1ΔEx3/Foxp3Cre Treg cells compared to Foxp3Cre Treg cells (Fig. 5c). To determine whether β-catenin and Foxo1 are directly interacting with each other, we performed an in situ proximity ligation assay (PLA) on human Treg cells and detected the PLA signal in human Treg cells (Fig. 5d). Collectively, our results indicate that β-catenin regulates the pro-inflammatory TH1-skewing program in Treg in concert with the Foxo pathway.

Figure 5.

β-catenin stabilized Treg cells represent dysfunctional phenotype with phosphorylation of Foxo1.

(a) Representative histogram of CFSE dilution for Treg suppression assay. Yellow; Foxp3 Teff only, Blue; Foxp3Cre Treg cells and Foxp3Cre Teff at 1:1 ratio, Red; Ctnnb1ΔEx3/Foxp3Cre Treg cells and Foxp3Cre Teff at 1:1 ratio (left). Bar graph shows percentage of suppression (right) (n=3). P values were calculated by two-sided Student’s t-test. (b) Gene expression profile of Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre Treg cells by microarray analysis. (c) Flow cytometric analysis on peripheral lymph node Treg cells from Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre mice. Quantification of gMFI for indicated molecules was shown (Foxp3Cre; n=4 mice, Ctnnb1ΔEx3/Foxp3Cre; n=4 or 5 mice). P values were calculated by two-sided Student’s t-test. Data were represented as mean and mean +/− SD. (d) Representative immunofluorescence images of human Treg cells with PLA signal for β-catenin-Foxo1 interaction (red) and Foxp3 staining (green). Nuclei were stained with DAPI (blue). Scale bar; 5μM. Data were representative of three experiments.

High salt environment activates the β-catenin-SGK1-Foxo axis and produces IFN-γ/IL-10 imbalance

It has been shown previously that the PI3K-AKT1-Foxo axis played a pivotal role in inducing IFN-γ-producing dysfunctional Treg[25]. Furthermore, we observed that p-Foxo1/3a and SGK1 were upregulated in Ctnnb1ΔEx3/Foxp3Cre Treg cells. To assess if the SGK1-Foxo axis is activated in human Treg subpopulations, we examined p-SGK1 and p-Foxo1 levels by flow cytometry and found that both were highly expressed in the IFN-γ-producing Treg population ex vivo (Supplementary Fig. 4a). Additionally, we demonstrated the direct interaction between β-catenin and Foxo1 in IFN-γ-producing human Treg cells by using PLA (Supplementary Fig. 4b). These findings prompted us to hypothesize that activation of β-catenin is involved in high salt-induced IFN-γ production as an upstream regulator of the SGK1-Foxo axis. Higher expression of Active β-catenin, p-SGK1, and p-Foxo1 was observed specifically in the IFN-γ-producing human Treg subset under high salt conditions (Fig. 6a), but not in the IL-10-producing subset (Supplementary Fig. 4c). β-catenin target genes (AXIN2 and TCF7) and TCF-1 protein were also upregulated in human Treg cells treated with increased salt concentration (Fig. 6b, Supplementary Fig. 4d). Additional salt treatment skewed the IL-12 induced, TH1-like Treg to produce more IFN-γ and less IL-10, suggesting that the high salt environment might exacerbates the IFN-γ-skewing pathogenic Treg signature that resembles the MS-Treg phenotype (Fig. 6c, Supplementary Fig. 4e). To determine if β-catenin activation was necessary for IFN-γ induction under high salt conditions, we pharmacologically blocked β-catenin signaling with two different Wnt-β-catenin signaling inhibitors, PKF and IWR-1. PKF or IWR-1 significantly downregulated high salt-induced IFN-γ expression in human Treg cells (Fig. 6d, Supplementary Fig. 4f). These results were also observed upon genetic knock down of CTNNB1 by shRNA (Fig. 6e-g). Given that SGK1 is a target of β-catenin signaling[33], we then tested the impact of inhibiting β-catenin signaling on SGK1 activation. PKF or IWR treatment prevented SGK1 phosphorylation under high salt stimulation in human Treg cells (Supplementary Fig. 4g). Next, we measured the production of IFN-γ and IL-10 and Foxo1 phosphorylation in the presence or absence of SGK1 inhibitor (GSK650394; SGK1-i). SGK1-i ameliorated high salt-induced IFN-γ production and Foxo1 phosphorylation, but had no impact on IL-10 production (Supplementary Fig. 4h, i). These results suggest that β-catenin positively regulates salt-induced IFN-γ expression through activation of the SGK1-Foxo axis. We extended the analysis to human effector T cells (Teff, CD4+ CD25) and Jurkat T cells. Both of these displayed active β-catenin signaling under high salt condition (Supplementary Fig. 5a). In Teff cells there was an imbalance of IFN-γ and IL-10 production in agreement with our Treg data (Supplementary Fig. 5b). In addition, we generated β-catenin-depleted Jurkat T cells by using CRISPR/Cas9 technology, and found that high salt-induced SGK1 and Foxo1 phosphorylation were attenuated in β-catenin knockout Jurkat T cells (Supplementary Fig. 5c). These data, along with the evidence from non-immune cells[34], supports our hypothesis that the β-catenin-SGK1-Foxo1 axis is activated by high salt stimulation.

Figure 6.

High salt environment induces β-catenin signal activation and IFN-γ:IL-10 cytokine imbalance.

(a) Flow cytometric analysis of Active β-catenin, p-SGK1 (Thr256), and p-Foxo1 (Ser256) expression in human IFN-γ-producing Treg cells. Treg cells were stimulated with anti-CD3 and anti-CD28 in the presence (NaCl) or absence (Control) of additional 40 mM NaCl for 96 h followed by 4 h PMA plus iomomycin stimulation (Active β-catenin; n=18 subjects, p-SGK1; n=13 subjects, p-Foxo1; n=10 subjects). P values were calculated by two-sided Student’s t-test. (b) mRNA expression kinetics for Wnt-β-catenin target genes (AXIN2 and TCF7) from nine time points were plotted and each dots represent the average of four different experiments. *P<0.05, **P<0.01, ***P<0.001 (two-way ANOVA with Sidak’s multiple comparisons test). Data were represented as mean +/− SEM. (c) IFNG mRNA expression in human Treg cells cultured in TH0 or TH1 condition in the presence (NaCl) or absence (Control) of additional 40 mM NaCl for 96 h (n=19 subjects). *P<0.05, **P<0.01, ***P<0.001 (one-way ANOVA with Tukey’s multiple comparisons test). (d) IFNG mRNA expression in human Treg cells stimulated in the presence (NaCl) or absence (Control) of additional 40 mM NaCl with and without Wnt/inhibitor PKF115–584 (PKF) or IWR-1 (IWR) for 96 h (n=7–10 subjects). *P<0.05 (one-way ANOVA with Tukey’s multiple comparisons test). (e) Representative flow cytometric analysis of IFN-γ and IL-10 production in human Treg cells transduced with a non-targeted shRNA or a CTNNB1 shRNA and cultured in the normal media (Control) or media supplemented with additional 40 mM NaCl (NaCl) for 96 h. Data are representative of three experiments. (f) IFNG and IL10 mRNA expression on Treg cells, and (g) frequency of IFN-γ and IL-1 0 producing Treg cells relative to control/scramble shRNA condition were shown. Treg cells were treated as in (e) (f; n=9 subjects, g; n=8 subjects). *P<0.05, **P<0.01, ***P<0.001 (one-way ANOVA with Tukey’s multiple comparisons test). Data were represented as mean +/− SD.

We next explored the molecular mechanisms underlying high salt-induced β-catenin activation. First, we examined whether Wnt ligands play a role in this aberrant activation of β-catenin signaling. We used the fragment crystallizable region fused to the cysteine-rich domain of Frizzled-8 protein (Fzd8-FC), which is known to act as a scavenger of Wnt ligands, to inhibit the effect of Wnt ligands on Treg cells. Activation of β-catenin assessed by Active β-catenin level or production of IFN-γ and IL-10 were not affected by Fzd8-FC treatment in control or high salt conditions, suggesting the dispensable role for Wnt ligands in high salt-induced activation of β-catenin (Supplementary Fig. 6a, b). Although a salinity stress sensor has not been fully described in mammalian cells, a number of pathways contributing to the salt stress response have been identified[34, 35]. Within these pathways, we focused on AKT kinase because it is well known to regulate β-catenin signaling via direct phosphorylation of β-catenin[36] or indirectly through GSK3β, which is a negative regulator of β-catenin. Indeed, the PI3K-AKT pathway was highly enriched in the IFN-γ-producing Treg subset and AKT phosphorylation at Ser473 was increased in IFN-γ-producing Treg (Supplementary Fig. 6c, d). We then investigated whether β-catenin could be directly activated by AKT by examining AKT-specific phosphorylation of β-catenin (Ser522), which stabilizes β-catenin[36]. Phosphorylation of β-catenin (Ser522) was increased in a high salt environment and this effect was reversed by the AKT inhibitor MK2206, indicating that activation of AKT is responsible for stabilizing β-catenin during high salt stimulation[34] (Supplementary Fig. 6e). Furthermore, we demonstrate that phosphorylation of GSK3β at Ser9, which is an important site of phosphorylation by AKT, was increased by high salt stimulation and that amounts of both p-AKT and p-GSK3β were not affected by silencing β-catenin (Supplemental Fig. 6f), suggesting both of them act upstream of β-catenin. These data indicate that AKT regulates β-catenin activation by both direct and indirect mechanisms under high salt conditions.

A high salt-induced PTGER2-β-catenin loop leads to imbalance between IFN-γ and IL-10

Both IFN-γ and IL-10 are upregulated in IL-12-induced TH1-like Treg cells in a β-catenin-dependent manner (Fig. 3f-i). However, IL-10 expression was significantly suppressed by high salt treatment, in contrast to IFN-γ. In fact, the β-catenin-SGK1-Foxo axis was not activated in IL-10SP after high salt treatment (Supplementary Fig. 4c). Additionally, the effect of high salt on IL-10 production could not be explained by activated β-catenin signaling (Fig. 6e, f, Supplementary Fig. 4f). Thus, we hypothesized that there might exist a factor that can be uniquely induced in the high salt environment but not in IL-12-driven TH1 conditions, resulting in IL-10 inhibition. We compared the gene expression profiles of Treg cells between control media and IL-12 supplemented media (TH1), and also between control media and NaCl supplemented media (NaCl). Among the group of differentially expressed genes in each comparison, we identified six genes that were upregulated in high salt conditions but downregulated in TH1 conditions, and four genes that were regulated in the opposite direction, which are potentially able to account for the high salt-induced IFN-γ:IL10 imbalance (Fig. 7a).

Figure 7.

PTGER2 is a unique factor regulating IFN-γ and IL-10 in conjunction with β-catenin under high salt condition.

(a) Venn diagrams showing the overlapped genes between the genes upregulated in NaCl treatment (NaCl Up) and downregulated in TH1 condition (TH1 Down) (left), and between the genes downregulated in NaCl treatment (NaCl Down) and upregulated in TH1 condition (TH1 Up) (right). (b) PTGER2 mRNA expression in human Treg cells (left) and TH17 cells (right). Treg cells were stimulated with anti-CD3 and anti-CD28 in the normal media (Control) or media supplemented with additional 40 mM NaCl (NaCl) for 96 h (n=14 subjects). Naive CD4+ T cells were cultured in the normal TH17 condition (Control) or TH17 condition supplemented with additional 40 mM NaCl (NaCl) for 72 h (n=6 subjects). P values were calculated by two-sided Student’s t-test. (c) Representative flow cytometric analysis of IFN-γ and IL-10 production in human Treg cells transduced with a scramble shRNA or a PTGER2 shRNA and cultured in the normal media (Control) or media supplemented with additional 40 mM NaCl (NaCl) for 96 h. Data are representative of three experiments. (d) Relative frequency of IFN-γ and IL-10 producing Treg cells, and (e) relative expression level of Active β-catenin in Treg cells were shown. Treg cells were treated as in (c) (d, e; n=8 subjects). *P<0.05, **P<0.01, ***P<0.001 (one-way ANOVA with Tukey’s multiple comparisons test). Data were represented as mean +/− SD.

PTGER2 is known to regulate the production of cytokines in a context-dependent manner[37, 38, 39, 40, 41], and the action of PTGER2 on cytokine production, especially IFN-γ production in T cells, was affected by the strength of PI3K-AKT signaling[42]. Since we have observed a role for PTGER2 in promoting the pathogenic phenotype by modulating IFN-γ:IL10 balance in TH17 cells[39] and high salt treatment induces a pathogenic TH17 cell signature, we hypothesized that PTGER2 regulates the IFN-γ:IL-10 balance in salt stimulated Treg cells. Indeed, PTGER2 was upregulated after high salt treatment in human Treg cells and TH17 cells (Fig. 7b) and was highly expressed in IFN-γSP compared to IL10SP (Supplementary Fig. 7a).

Given the evidence of a positive relationship between β-catenin signaling and PTGER2[43, 44], we investigated whether β-catenin and PTGER2 build an autoregulatory loop during chronic high salt exposure. We used Jurkat T cells, and demonstrated that high salt-induced PTGER2 was suppressed by genetic deletion of CTNNB1 (Supplementary Fig. 7b) and that PTGER2 knockdown could partially ameliorate high salt-induced β-catenin activation (Supplementary Fig. 7c). These results suggest the presence of a β-catenin-PTGER2 feed-forward loop under high salt conditions. PTGER2 silencing abolished the high salt-induced IFN-γ:IL-10 imbalance on human Treg cells, and eliminated the high salt-induced activation of β-catenin in IFN-γSP, while it did not affect Active β-catenin level in IL10SP (Fig. 7c-e), suggesting that high salt-induced IFN-γ depends on the PTGER2-β-catenin loop, but IL-10 suppression by high salt is dependent on PTGER2 per se but not on β-catenin.

We further investigated if the high salt-induced PTGER2-β-catenin loop could be amplified in cells where AKT is activated, such as in IFN-γ producing Treg, but not in the cells with lower AKT activity, such as IL-10 producing Treg. We then tested the impact of modulating AKT signaling on IFN-γ and IL-10 production in Treg cells under high salt conditions via increasing CD28 co-stimulation. High salt-induced IFN-γ production was boosted by strengthening AKT signaling with higher CD28 co-stimulation (Supplementary Fig. 7d)[42]. By contrast, high salt-induced IL-10 inhibition was not altered. Together, these data indicated that high salt induces a positive feedback loop between β-catenin and PTGER2 in conjunction with activated AKT status, resulting in amplification of IFN-γ production in Treg　cells.

Stabilized β-catenin is observed in Treg cells from mice fed a high salt diet and MS patients

To examine if β-catenin is stabilized under high salt conditions in vivo, we fed wild type mice with either a normal salt diet (NSD), containing 0.4% of NaCl, or a HSD, containing 4% NaCl, and assayed β-catenin expression on Treg cells. We found that β-catenin and phosphorylated Foxo1/3a/4, assayed by a monoclonal antibody for phosphorylation sites on Foxp1 (Thr24), Foxo3a (Thr32), and Foxo4 (Thr28), were increased in Treg cells from HSD mice (Fig. 8a). Next, we determined if β-catenin is more stabilized in MS-Treg as compared to Treg cells from healthy subjects. The level of Active β-catenin in IFN-γ producing Treg was increased in MS patients compared to healthy subjects (Fig. 8b). We also found a positive correlation between IFN-γ production and the level of Active β-catenin in MS-Treg but not in healthy subjects (Fig. 8c). Furthermore, to investigate the link between PTGER2 expression and Active β-catenin level or IFNG expression in MS-Treg, we assessed the expression of these factors in Treg cells from healthy subjects and MS patients. Notably, higher expression of PTGER2 and Active β-catenin level are correlated with IFNG expression in MS-Treg but not in Treg cells from healthy subjects (Fig. 8d, e). These findings provide in vivo evidence to support our hypothesis that the PTGER2-β-catenin loop plays an important role in the salt-induced malfunction of Treg cells and links this salt signature to the pathogenic profile of MS-Treg (Supplementary Fig. 8).

Figure 8.

Stabilized β-catenin associated with IFN-γ and PTGER2 expression in Treg cells from MS patients.

(a) Flow cytometric analysis of Treg cells from the mesenteric lymph nodes of wild type mice fed a normal diet (ND) or a high-salt diet (HSD) for 3 weeks. Quantification of gMFI for β-catenin and p-Foxo1/3a/4 were shown (ND; n=4, HSD; n=4). P values were calculated by two-sided Student’s t-test. (b) Flow cytometric analysis of ABC level in ex vivo Tregs of healthy controls and MS patients (HC; n=14 subjects, MS; n=11 subjects). P value was calculated by two-sided Student’s t-test. Data were represented as mean +/− SD. (c-e) Correlation plots (c); between the percentage of IFN-γ−producing Treg cells and gMFI of Active β-catenin, (d); between IFNG and PTGER2 mRNA expression, (e); between Active β-catenin level and PTGER2 mRNA expression level in healthy subjects and MS patients. Linear regression is shown with 95% confidence interval (dotted lines). Correlation statistics by two-sided Spearman rank correlation test.

Discussion

Loss of Foxp3+ regulatory T cell function is associated with a number of autoimmune diseases, and previous works have linked environment factors to autoimmunity through affecting Treg cell homeostasis. Here we extend this concept yet also describe a novel molecular mechanism using both murine and human system. Our results demonstrate a novel role of β-catenin as a regulatory molecule for Treg functional plasticity and also provide molecular mechanisms that link high salt environment to autoimmune disease.

Our transcriptional profiling of human Treg subsets based on IFN-γ and IL-10 production provides new insights into Treg heterogeneity. We identify β-catenin as a key regulator and demonstrate its role in skewing Treg into a dysfunctional state in human Treg cells and in murine models. Although several studies have demonstrated the role of β-catenin in Treg cells, it is still unknown how β-catenin contributes to Treg function and the effector cytokine signature. One study demonstrated β-catenin as an anti-inflammatory factor in the context of generating long-lived suppressive Treg cells via anti-apoptotic gene induction[45], and two previous studies showed that activation of β-catenin provokes Treg dysfunction, leading to exaggerated colitis in a murine model[28, 46]. We show that Treg-specific stabilization of β-catenin resulted in loss of suppressive properties of Treg cells and induced a lethal Scurfy-like phenotype in mice, which is consistent with the latter aspect. We believe our Foxp3-Cre-based Treg-specific intervention provides direct evidence for the role of β-catenin in Treg cells.

The incidence of autoimmune diseases has been increasing in the last half century, which cannot be explained by genetic adaptation. Thus, there is great interest in studying the interplay between genetic risks and environmental triggers[47]. Among several environmental triggers, a high salt diet might increases the incidence of autoimmune diseases, though this as yet requires further epidemiologic investigations[48]. Previous studies have shown that higher salt concentration affects on TH17 cell development and Treg stability, manifest by aberrant IFN-γ production[16]. Of note, our findings that IFN-γ:IL-10 imbalance is observed not only in MS-Treg cells but also under high salt condition suggested that the salt-induced signature may overlap with the MS pathogenic profile. The importance of IFN-γ:IL-10 in the context of salt-induced immune alteration was supported by a previous study showing that increased sodium content in colon tissue of HSD mice resulted in excessive inflammation in IBD models[18]. Interestingly, β-catenin signaling was activated in both Treg and TH17 cells under stimulation with high salt (data not shown). Furthermore, here we demonstrate that PTGER2 accounts for high salt-induced IFN-γ:IL-10 imbalance in Treg cells by creating a positive feed forward loop with β-catenin. Given that Treg cells can produce PGE2[49] and that PGE2 is enriched in EAE lesions[50], PGE2-PTGER2 signaling could be amplified in Treg cells under high salt conditions and also in the MS lesion. However, the role of PGE2 in EAE and MS remains unclear and further investigation is needed.

In summary, we provide genome-wide transcriptomic profiles of human ex vivo Treg subpopulations, which unveil the heterogeneity of Treg cells in terms of IFN-γ and IL-10 production. Aberrant β-catenin activation modulates Treg cytokine plasticity and integrity in both human and murine Treg cells. Under the high salt environment, this effect occurs in conjunction with upregulation of PTGER2, resulted in establishing the feed-forward loop between PTGER2 and β-catenin. Of note, Treg cells from MS patients display positive correlations among IFN-γ production, PTGER2 expression, and Active β-catenin level, which are not observed in Treg cells from healthy subjects. Together, our study in humans with autoimmune disease and confirmed in mouse models indicates that the PTGER2-β-catenin axis serves as a bridge between environmental factors and autoimmune disease by modulating Treg properties.

Methods

Study subjects

Peripheral blood was drawn from healthy individuals and patients with MS after informed consent and approval by the Institutional Review Board at Yale University. The patients were diagnosed with either Clinically Isolated Syndrome (CIS) or Relapsing-Remitting MS (RRMS) by 2010 MacDonald Criteria and were not treated with any immunomodulatory therapy at the time of the blood draw. All experiments conformed to the principles set out in the WMA Declaration of Helsinki and the Department of Health and Human Services Belmont Report. Clinical characteristics of evaluated MS patients are listed in Supplementary Table 2.

Mice

C57/BL6/J mice were purchased from the Jackson Laboratory or CLEA Japan. FIC mice26, and Ctnnb1ΔEx3 mice[27] have been described, and mice backcrossed into the C57/BL6/J strain were used. Ctnnb1ΔEx3/Foxp3Cre mice were studied at 3–5 weeks of age. For high salt diet (HSD) experiments, six-week-old male wild type mice were fed with normal chow (control group) or sodium-rich chow containing 4% NaCl (Research Diets; HSD group) with normal tap water for 3 weeks. Cells were isolated from the spleen and/or mesenteric lymph nodes and Treg cells were analyzed by flow cytometry. All experiments were approved by the University of Tokyo Ethics Committee for Animal Experiments and strictly adhered to the guidelines for animal experiments of the University of Tokyo.

Human T cell isolation and culture

Peripheral blood mononuclear cells (PBMCs) were isolated from donors by Ficoll-Paque PLUS (GE Healthcare) or Lymphoprep (Stemcell) gradient centrifugation. Total CD4+ T cells were isolated by negative magnetic selection using a CD4 T cell isolation kit (Stemcell) and CD4+CD25hiCD127lo-negCD45RO+ Treg cells were sorted on a FACS Aria (BD Biosciences). Treg cells were cultured in RPMI 1640 medium supplemented with 5% Human serum, 2 nM L-glutamine, 5 mM HEPES, and 100 U/ml penicillin, 100 μg/ml streptomycin, 0.5 mM sodium pyruvate, 0.05 mM nonessential amino acids, and 5% human AB serum (Gemini Bio-Products). 96-well round bottom plates (Corning) were pre-coated with anti-human CD3 (UCHT1) (1–2 μg/ml) and used for Treg in vitro culture with soluble anti-human CD28 (28.2) (1–5 μg/ml) (BD Bioscience) and human IL-2 (50 U/ml). Human IL-2 was obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH). TH1-Treg cells were induced with human recombinant IL-12 (20 ng/ml) (R&D). The Wnt/β-catenin inhibitor PKF115–584 (Tocris) was used at 200 nM and IWR-1 (Tocris) was used at 20 μM. SGK1 inhibitor GSK650394 (Tocris) was used at 10 μM. AKT inhibitor MK2206 (Tocris) was used at 5 μM. Frizzled 8 FC chimera protein (R&D) was used at 500 ng/ml.

Suppression assay

CD4+CD25+ Treg cells and CD4+CD25- T responder cells were isolated from spleen and lymph node from Foxp3Cre mice or Ctnnb1ΔEx3/Foxp3Cre mice by using CD4+CD25+ Regulatory T Cell Isolation Kit (Miltenyi Biotec) and were sorted on a FACS Aria (BD Biosciences). T responder cells were labeled with CFSE and then co-cultured with Treg cells (5 × 104) at 1:1 ratio in RPMI 1640 medium supplemented with 10% FBS (HyClone), 50 μM 2-Mercaptoethanol (Sigma-Aldrich), 1x GlutaMAX, 50 U/ml penicillin, and 100 μg/ml streptomycin with Dynabeads Mouse T-Activator CD3/CD28 at 2:1 bead-to-cell ratio. The proliferation of T responder cells was determined at day 4 by FACS on a Verse instrument (BD Bioscience).

Quantitative PCR

Total RNA was extracted using RNeasy Micro Kit (QIAGEN), or ZR-96 Quick-RNA kit (Zymo Research), according to the manufacturer’s instructions. RNA was treated with DNase and reverse transcribed using TaqMan Reverse Transcription Reagents (Applied Biosystems) or SuperScript IV VILO Master Mix (Invitrogen). cDNAs were amplified with Taqman probes (Taqman Gene Expression Arrays) and TaqMan Fast Advanced Master Mix on a StepOne Real-Time PCR System (Applied Biosystems) according to the manufacturer’s instructions. mRNA expression was measured relative to B2M expression.

Flow cytometry analysis

Single-cell suspensions were prepared from PBMCs or mouse tissues and stained with fixable viability dye for 10 min at RT, followed by staining with surface antibodies for 30 min at 4°C. For intracellular staining, cells were fixed and permeabilized with the Foxp3 Fix/Perm buffer set (eBioscience) for 1 h at 4°C, followed by staining with intracellular antibodies. For cytokine staining, cells were stimulated with PMA (50 ng/ml) and ionomycin (for ex vivo Treg cells; 1000 ng/ml, for in vitro cultured Treg cells; 250 ng/ml) in the presence of GolgiPlug (BD Bioscience) for 4 h at 37°C. Antibodies and reagents used for flow cytometric analysis are listed as follows: for human samples; anti-CD4 (RPA-T4), anti-CD25 (MA251), anti-CD127 (HIL-7R-M21), anti-CD45RO (UCHL1), anti-IFN-γ (B27), anti-β-catenin (14/Beta-Catenin), anti-TCF1 (S33–966), PE-Cy™7 Streptavidin from BD Bioscience, anti-IL-10 (JES3–9D7), anti-Tbet (4B10) from BioLegend, anti-phospho Foxo1 (S256) polyclonal, anti-phospho SGK1 (T256) polyclonal (from Bioss), anti-phospho β-catenin (Ser522) (D8E11), anti-phospho AKT (Ser473) (D9E), anti-phospho GSK3β (Ser9) (D85E12), anti-Foxo1 (C29H4) from CST, anti-IFN-γ (4S.B3), anti-Foxp3 (PCH101) from eBioscience, anti-active β-catenin (8E7) from Millipore, IFN-γ secretion assay (APC), IL-10 secretion assay (PE) from Miltenyi, for mice samples; anti-Foxp3 (FJK-16s), anti-CD3 (145–2C11), anti-CD4 (RM4–5), anti-CD8 (53–6.7), anti-GATA3 (TWAJ), anti-RORgt (B2D) from eBioscience, anti-Helios (22F6) from BioLegend, anti-phospho Foxo1 (S256) (E1F7T), anti-phospho Foxo3a (Ser253) (D18H8), anti-phospho Foxo1(T24)/3a(T32)/4(T28) (4G6) from CST, anti-SGK1 (Y238) (from Abcam), for both human and mice samples; Zombie Aqua™ Fixable Viability dye from BioLegend. Stained samples were analyzed with a BD FACS Verse or an LSR Fortessa flow cytometer (BD Bioscience). Data were analyzed with FlowJo software (Treestar).

RNA-seq library preparation and data analysis

Preparation of cells for RNA-seq:

For the ex vivo Treg subpopulations, CD4+CD25hiCD127lo-negCD45RO+ memory Treg cells from healthy donors were sorted and immediately stimulated with PMA (50 ng/ml) and iomomycin (1000 ng/ml) for 4 h. By combining IFN-γ secretion assay (APC) and IL-10 secretion assay (PE) (Miltenyi), Treg cells were labeled based on the expression of IFN-γ and IL-10. To avoid RNA degradation, cells were kept in CellCover (Anacyte) before a second round of sorting. For in vitro cultured Treg cells in TH1 or high salt conditions, mTreg cells were cultured in each condition for four days as described. Cells were harvested and immediately processed for cDNA preparation. Samples were collected from four healthy subjects for identification of the TH1 signature and five healthy subjects for identification of the high salt signature.

cDNA and library preparation and sequencing:

cDNAs were generated directly from resorted and harvested cells using the SMART-Seq v4 Ultra Low Input RNA Kit for sequencing (Takara/Clontech). Barcoded libraries were generated by the Nextera XT DNA Library Preparation kit (Illumina) and sequenced with a 2 × 100 bp paired-end protocol on the HiSeq 2000 Sequencing System (Illumina).

RNA-seq data analysis:

RNA-seq analysis was performed using Partek flow (v6.6). First, RNA-seq reads were trimmed and mapped to the hg19 genome reference using STAR (2.5.0e). Aligned reads were quantified to the gene level using Partek’s E/M algorithm and gene annotation from Ensembl release 75. Gene-level quantitations were normalized by dividing the gene counts by the total number of reads, following by addition of a small offset (0.001). The offset was added to enable log2 transformation and the value of the offset was determined by exploring the data distribution. Differential expression was assessed by fitting Partek’s log-normal model with shrinkage (comparable in performance to limma-trend). Genes having geometric mean below 1.0 were removed from the analysis.

For ex vivo Treg subpopulation data, differentially expressed genes (Fold change > 1.5, P value < 0.05) were used for functional analysis using IPA and upstream regulator analysis (https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis/). Gene set enrichment analysis was performed on normalized gene expression counts of RNA-seq data or microarray data as described previously.

For in vitro TH1-Treg and high salt Treg data, differentially expressed genes (Fold change > 2, P value < 0.05) were used.

Microarray

For the Oligo DNA microarray analysis, total RNA samples were extracted from sorted CD4+CD25hi Treg cells of Foxp3Cre mice or Ctnnb1ΔEx3/Foxp3Cre mice. Microarray analysis was performed with a 3D-Gene Mouse Oligo chip 24k (Toray Industries Inc., Tokyo, Japan). Total RNA was labeled with Cy5 by using the Amino Allyl MessageAMP II aRNA Amplification Kit (Applied Biosystems). The Cy5-labeled aRNA pools were mixed with hybridization buffer, and hybridized for 16 h. The hybridization signals were obtained by using a 3D-Gene Scanner and processed by 3D-Gene Extraction software (Toray Industries Inc.). Detected signals for each gene were normalized with the global normalization method (the median of the detected signal intensity was adjusted to 25).

Histology

Mouse tissues were fixed in Ufix (Sakura Finetek Japan) and embedded in paraffin. 6-μm tissue sections were stained with haematoxylin and eosin.

Lentiviral transduction for shRNA gene silencing and CRISPR/Cas9-mediated gene deletion

Lentiviral plasmids encoding shRNAs were obtained from Sigma-Aldrich and all-in-one vectors carrying CTNNB1 sgRNA/Cas9 with GFP reporter were obtained from Applied Biological Materials. Each plasmid was transformed into One Shot® Stbl3™ chemically competent cells (Invitrogen) and purified by ZymoPURE plasmid Maxiprep kit (Zymo research). Lentiviral pseudoparticles were obtained after plasmid transfection of 293FT cells using Lipofectamine 2000 (Invitrogen). The lentivirus-containing media was harvested 48 or 72 h after transfection and concentrated 40 – 50 times using Lenti-X concentrator (Takara Clontech). Sorted Treg cells were stimulated with plate-bound anti-CD3 (1 μg/ml) and soluble anti-CD28 (1 μg/ml) for 24 h and transduced with lentiviral particles by spinfection (1000 x g for 90 min at 32°C) in the presence of Polybrene (5 μg/ml) on the plates coated with Retronectin (50 μg/ml) (Takara/Clontech) and anti-CD3 (1–2 μg/ml). Human Jurkat T cells were directly transduced with lentiviral particles by spinfection. Five days after transduction, cells were sorted on the basis of expression of GFP. GFP expressing human Jurkat T cells were further purified by FACS at least three times before using for experiments.

Proximity ligation assay (PLA)

PLA was performed with Duolink In Situ Detection Reagents Orange (Sigma) according to the manufacturer’s recommendation with minor modifications. Treg cells were cultured for four days and harvested, and cells were fixed with 2% paraformaldehyde for 10 min at RT. Fixed cells were incubated in Foxp3 Fix/Perm buffer set for 30 min at 4°C, followed by staining with mouse anti-β-catenin (14/Beta-Catenin) and rabbit anti-Foxo1 (C29H4) for 1 h at RT in Foxp3 staining buffer. Cells were washed and stained in Foxp3 staining buffer with the secondary mouse PLUS and rabbit MINUS antibodies for 30 min at RT. Cells were washed in TBS (0.01 M Tris, 0.15 M NaCl) with 0.5% BSA and the ligation reaction was performed at 37°C for 30 min, followed by the amplification reaction at 37°C for 100 min. Cells were washed in TBS (0.2 M Tris, 0.1 M NaCl) with 0.5% BSA and stained with anti-Foxp3 (PCH101) or anti-IFN-γ (B27) antibody for 30 min at 4°C. Cells were analyzed with a 60x or 100x objective on a Leica DM6000 CS confocal microscope.

Statistical analysis

All statistical analyses were performed using GraphPad Prism 6 (GraphPad Software). Detailed information about statistical analysis, including tests and values used, is provided in the figure legends. Values of P<0.05 or less were considered significant.

Memory Treg cells are the main source of effector cytokines IFN-γ and IL-10.

(a) Sorting strategy for memory and naive Treg cells from circulating human CD4+ T cells. (b) mRNA expression of IFNG and IL10 gene on memory and naive Treg cells (memory Treg cells; n=35 subjects, naive Treg cells; n=16 subjects). P values were calculated by two-sided Student’s t-test. Data were represented as mean +/− SD.

β-catenin signaling is activated in IFN-γ producing human Treg subset.

(a) GSEA enrichment plots of four different Wnt signaling pathway gene sets between IFN-γSP vs. DN. (b) Flow cytometric analysis of TCF-1 expression on ex vivo Treg subpopulations relative to DN (n=8 subjects). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (c) AXIN2 and TCF7 mRNA expression in IFN-γ+ and IFN-γ– human Treg populations assessed by DNA microarray (n=7 subjects). P values were calculated by two-sided Student’s t-test. (d) CTNNB1 gene expression on Treg cells transduced with a non-targeted shRNA or a CTNNB1 shRNA and cultured for 5 days (n=10 subjects). P values were calculated by two-sided Student’s t-test. (d) Frequency of IFN-γ and IL-10 positive cell number. Treg cells were stimulated with anti-CD3 and anti-CD28 in the presence of SGK1 inhibitor GSK650394 (SGK1-i), IL-12 (TH1), or IL-12 and GSK650394 (TH1+ SGK1-i) (n=6 subjects) ***P<0.001 (one-way ANOVA with Tukey’s multiple comparisons test). Data were represented as mean +/− SD.

Treg specific activation of β-catenin induces IFN-γ secreting dysfunctional phenotype with Scurfy-like autoimmunity.

(a) Schematic of the wild-type and targeted CTNNB1 allele. (b) Flow cytometric analysis of β-catenin and Foxp3 in peripheral lymph nodes, spleen, and thymus CD4+ T cells from Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre mice. (c) Expression for classical helper cytokines and transcription factors in both Treg cells (Treg) and T effector cells (CD4+ CD25neg; non-Treg) assessed by qPCR (n=4 mice). *P<0.05, **P<0.01 (two-sided Student’s t-test). (d) Flow cytometric analysis of Foxp3 and Helios expression on CD4+ T cells in peripheral lymph nodes and spleen from Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre mice at the age of 3 weeks. Percentages of Foxp3+ and/or Helios+ CD4+ T cells isolated from lymph nodes are shown at the bottom. (n=5–6 mice) ***P<0.001 (two-way ANOVA with Sidak’s multiple comparisons test). (e) GSEA enrichment plot between Foxp3Cre and Ctnnb1ΔEx3/Foxp3Cre Treg cells using the gene set that is positively regulated by Foxo1 (left) and negatively regulated by Foxo1 (right) identified from the comparison between wild-type vs. Foxo1 KO Treg cells (GSE40655). Normalized enrichment score (NES) and false discovery rate (FDR) are indicated. Data were represented as mean +/− SD.

High salt activates the β-catenin/SGK1/Foxo axis in IFN-γ-producing human Treg cells.

(a) Flow cytometric analysis of p-SGK1 (left) and p-Foxo1 (right) level in human Treg subsets. Treg cells were stimulated with anti-CD3 and anti-CD28 for 96 h followed by 4 h PMA plus iomomycin stimulation, and the expression of p-SGK1 and p-Foxo1 were determined by intracellular staining in each subset. (n=12 subjects; p-SGK1, n=10 subjects; p-Foxo1) (b) Representative immunofluorescence images of human Treg cells with PLA signal for β-catenin-Foxo1 interaction (red) and IFN-γ staining (green). Nuclei were stained with DAPI (blue). PLA signal (arrowheads) was observed in an IFN-γ positive cell (arrow). Data were representative of three experiments. (c) Flow cytometric analysis of Active β-catenin, p-SGK1 (Thr256), and p-Foxo1 (Ser256) expression in human IL-10 producing Treg cells. Treg cells were stimulated with anti-CD3 and anti-CD28 in the presence (NaCl) or absence (Control) of additional 40 mM NaCl for 96 h followed by PMA plus iomomycin stimulation for 4 h (Active β-catenin; n=15 subjects, p-SGK1; n=13 subjects, p-Foxo1; n=10 subjects). P values were calculated by two-sided Student’s t-test. (d) Flow cytometric analysis of TCF1 expression in human IFN-γ producing Treg cells. Treg cells were cultured as in (c) (n=3 subjects). P values were calculated by two-sided Student’s t-test. (e) Relative frequency of IFN-γ and IL-10 positive cell number (fold of control condition) in human Treg cells cultured as in Fig. 6c (n=11 subjects). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (f) Relative frequency of IFN-γ and IL-10 positive cell number (fold of control condition) in human Treg cells cultured as in Fig. 6d (n=9 subjects). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (g) Flow cytometric analysis of p-SGK1 expression in human IFN-γ producing Tregs. Treg cells were cultured as in Fig. 6d (n=6 subjects). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (h) Frequency of IFN-γ and IL-10 positive cell number in human Treg cells stimulated in the presence (NaCl) or absence (Control) of additional 40 mM NaCl with and without Wnt/inhibitor GSK650394 (SGK1-i) for 96 h (n=11 subjects). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (i) Flow cytometric analysis of p-Foxo1 expression in human IFN-γ producing Treg cells. Treg cells were cultured as in (h) (n=8 subjects). *P<0.05 (one-way ANOVA with Tukey’s multiple comparisons test). Data were represented as mean +/− SD.

The β-catenin-SGK1-Foxo axis is also activated in Teff and human Jurkat T cells under high salt conditions.

(a) Flow cytometric analysis of Active β-catenin expression in human T effector cells (Teffs) and human Jurkat T cells. Human Teff were stimulated as well as Treg cells for 96 h (n=8 subjects). Human Jurkat T cells were cultured without TCR stimulation for 120 h (n=12). Both were cultured in the presence (NaCl) or absence (Control) of additional 40 mM NaCl. P values were calculated by two-sided Student’s t-test. (b) Relative frequency of IFN-γ and IL-10 positive cell number (fold of control condition) in human Teffs cultured as in (a) (n=8 subjects). P values were calculated by two-sided Student’s t-test. (c, d) Flow cytometric analysis of p-SGK1 (Thr256) (c) and p-Foxo1 (Ser256) (d) in human Jurkat T cells. Human Jurkat T cells were transduced with scramble gRNA (CRISPR/Scramble) or CTNNB1 targeted gRNA (CRISPR/CTNNB1) with Cas9. Both cell lines were cultured in the presence (NaCl) or absence (Control) of additional 40 mM NaCl without TCR stimulation for 120 h (n=4). **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). Data were represented as mean +/− SD.

High salt induced β-catenin activation via AKT is independent of Wnt ligands.

(a) Relative frequency of IFN-γ and IL-10 positive cell number (fold of control condition) in human Treg cells. Treg cells were stimulated with anti-CD3 and anti-CD28 in the presence of Fzd8-FC (Fzd), additional 40 mM NaCl (NaCl), or Fzd8-FC and NaCl (NaCl + Fzd) (n=7 subjects). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (b) Relative expression level of ABC in human Treg cells cultured as in (a). (n=7 subjects) *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (c) GSEA enrichment plots of PI3K/AKT pathway gene sets between IFN-γSP vs. IL10SP. Normalized enrichment score (NES) and false discovery rate (FDR) are indicated at the bottom of each plot. (d) Relative expression level of p-AKT on Treg subsets. Treg cells were stimulated with anti-CD3 and anti-CD28 for 4 days, followed by 4 h PMA plus iomomycin stimulation and intracellular cytokine staining for IFN-γ and IL-10 (n=5 subjects). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (e) Flow cytometric analysis of β-catenin phosphorylation at s522 in human Jurkat T cells. Human Jurkat T cells were stimulated in the presence of AKT inhibitor MK2206 (AKT-i), additional 40 mM NaCl (NaCl), or MK2206 and NaCl (NaCl + AKT-i) (n=4). (f) Flow cytometric analysis of GSK3β phosphorylation at s9 and AKT phosphorylation at s473 in human Jurkat T cells. Human Jurkat T cells were prepared as in Supplementary Fig. 6c. (n=4). *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). Data were represented as mean +/− SD.

PTGER2-β-catenin loop is activated by high salt stimulation.

(a) PTGER2 expression assessed by RNA-seq on ex vivo Treg subpopulations (n=8 subjects). (b) Flow cytometric analysis of PTGER2 in human Jurkat T cells. Human Jurkat T cells were prepared as in Supplementary Fig. 6c. (n=4). **P<0.01, ***P<0.001 (one-way ANOVA with Tukey’s multiple comparisons test). (c) Flow cytometric analysis of ABC in human Jurkat T cells. Human Jurkat T cells were transduced with a scramble shRNA or a PTGER2 shRNA and cultured in normal media (Control) or media supplemented with additional 40 mM NaCl (NaCl) for 120 h. (n=4) *P<0.05, **P<0.01 (one-way ANOVA with Tukey’s multiple comparisons test). (d) Representative flow cytometric analysis of IFN-γ and IL-10 production in human Treg cells cultured in the normal media (Control) or media supplemented with additional 40 mM NaCl (NaCl) with anti CD3 (2μg/ml) and different concentration of anti CD28 (1, 2, 5 μg/ml) for 96 h. Relative frequency of IFN-γ and IL-10 producing Treg cells are shown at the bottom (n=4 subjects). *P<0.05, **P<0.01 (two-way ANOVA with Sidak’s multiple comparisons test). Data were represented as mean +/− SD.

Schematic model of the role of PTGER2 and the AKT-β-catenin-SGK1-Foxo axis for the production of IFN-γ and IL-10 in Treg cells.

AKT-β-catenin signaling balances between IFN-γ and IL-10 production in Treg cells. Under high salt conditions, PTGER2 was increased and established the positive feed forward loop with β-catenin, resulted in amplified activation of the β-catenin-SGK1-Foxo axis in IFN-γ-producing Treg cells.

Upstream regulator analysis in IPA between each Treg subpopulations

Lists of the top-ranking genes identified by IPA analysis as upstream regulators between each Treg subpopulations. Tables show statistically significant (overlap P value <0.05) upstream regulators in each comparison (Genes that could not be calculated for fold change were blank). CTNNB1 gene, which codes β-catenin protein, was highlighted in red.

Clinical characteristics of evaluated MS patients