NHC-gold compounds mediate immune suppression through induction of AHR-TGFβ1 signalling in vitro and in scurfy mice

Gold compounds have a long history of use as immunosuppressants, but their precise mechanism of action is not completely understood. Using our recently developed liver-on-a-chip platform we now show that gold compounds containing planar N-heterocyclic carbene (NHC) ligands are potent ligands for the aryl hydrocarbon receptor (AHR). Further studies showed that the lead compound (MC3) activates TGFβ1 signaling and suppresses CD4+ T-cell activation in vitro, in human and mouse T cells. Conversely, genetic knockdown or chemical inhibition of AHR activity or of TGFβ1-SMAD-mediated signaling offsets the MC3-mediated immunosuppression. In scurfy mice, a mouse model of human immunodysregulation polyendocrinopathy enteropathy X-linked syndrome, MC3 treatment reduced autoimmune phenotypes and extended lifespan from 24 to 58 days. Our findings suggest that the immunosuppressive activity of gold compounds can be improved by introducing planar NHC ligands to activate the AHR-associated immunosuppressive pathway, thus expanding their potential clinical application for autoimmune diseases.

T he aryl hydrocarbon receptor (AHR), a ligand-dependent transcription factor of the basic helix-loop-helix/Per-Arnt-Sim family, often acts as a sensor in response to environmental alternation 1 . The AHR gene is ubiquitously expressed in vertebrates and has an important role in development 2 . Its liganddependent activation is tightly controlled by a range of physiological responses 3 . The AHR was initially considered a mediator of the xenobiotic metabolism to detoxify dioxins and other toxic chemicals 3 . Because xenobiotic metabolism is part of the human defense system, the AHR is widely expressed in the immune system, including lymphocytes, macrophages and natural killer cells 4,5 . Compelling evidence has shown that ligand-mediated AHR activation triggers transforming growth factor beta (TGFβ) production, thereby, for instance, decreasing the expression of interleukin-2 (IL-2) and regulating the differentiation of CD4 + T cells to repress immune activity [6][7][8][9] . Velhoen et al. also reported that ligand-activated AHR contributed to the formation of Th17 T cells, a key mediator in the pathology of autoimmune diseases, implicating that the influence of AHR on the self-immune tolerance is largely related to the binding ligand 10 . So far, a large body of chemicals has been identified as endogenous or exogenous AHR ligands 2 .
Autoimmune diseases, such as rheumatoid arthritis, inflammatory bowel disease, multiple sclerosis and type 1 diabetes mellitus, are characterized by over activity of the immune system, leading to tissue damage 11 . Immunodysregulation polyendocrinopathy enteropathy X-linked (IPEX) syndrome is a rare, heritable autoimmune lymphoproliferative disorder associated with severe diarrhea, diabetes, eczema, erythroderma, and liver diseases 12 . It is caused by the mutation of the FOXP3 gene and occurs in males in the early months of life. Immunosuppressive medications, like methotrexate (MTX), is used for the treatment 13 . However, at present bone marrow transplantation is the sole possible cure with many risks of complications; thus research is ongoing to identify and develop safer therapies for patients with IPEX 12 . In scurfy mice, a mouse model of human IPEX syndrome, treatment with TGFβ-differentiated FOXP3 + T cells from wt mice or autologous transplantation of hematopoietic stem cell containing a lentiviral vector for FOXP3 overexpression has been shown to sufficiently prevent skin inflammation 14,15 . Although the efficacy of gold compounds against rheumatoid arthritis was tested in the 20 th century by Jacques Forestier 16 , little is known about their effect on IPEX syndrome.
Recently, we established a liver-on-a-chip system consisting of a microfluidic chip equipped with two interconnected chambers to investigate cellular responses to liver-mediated drug metabolites 17,18 . Applying this system, we screened our own small molecules collection including some organometallics [19][20][21] . Interestingly, we found that N-heterocyclic carbene (NHC) gold complexes including our lead structure [di-(1,3-diethylbenzylimidazol-2-ylidene)]gold(I) iodide (MC3) ( Supplementary  Fig. 1A) and other gold organometallics are potent AHR ligands. In line with this effect, MC3induced TGFβ1 and inhibited immune response in vitro and in vivo; however chemical and genetic inhibition of the AHR or TGFβ1 signaling pathways antagonized these effects. Thus, our data identify planar NHC gold organometallic compounds as activators of the AHR-TGFβ1 axis and show for the first time the potential of activating this signaling cascade as a complementary therapeutic target to enhance the immunosuppressive effect of gold compounds.

Results
Gold compounds induce CYP1A1 expression and are AHR ligands. We recently developed a semi-automated, remotely controllable microfluidic chip system composed of two interconnected microfluidic chambers (Chamber 1 and 2, Supplementary Fig. 1B). The basic technical specifications of this system have been described previously 17 and are outlined in Supplementary Fig. 1B. We used hepatocellular carcinoma cell line HepG2 as a metabolite generator, which expresses the majority of human cytochromes P450 (CYPs) and is widely used in various microfluidic systems 17,18,22,23 . Thus, the compound of interest is metabolized in the chamber 1 (HepG2 cells) and cellular responses to metabolic products can be recorded in the chamber 2. In this study cellular viability was determined by the number of propidium iodide positive cells over total cells (Hoechst staining) and toxicity was calculated as the percentage of cleaved caspase 3. We screened our own small molecules collection (~200), including several organometallics we recently synthesized 21,24,-26 . The treatment of a gold complex bearing an NHC ligand MC3 ( Supplementary Fig. 1A) led to significant differences in cell viability between chambers (Fig. 1b) and toxicity ( Supplementary Fig. 1C). In addition, we studied the stability of MC3 and found that MC3 shows exceptional chemical stability in aqueous and polar environments at physiological temperatures over several days. Some decomposition occurred after extended periods (>24 h) in highly apolar media (Supplementary Fig. 1D), suggesting that this detoxification was associated with hepatocyte-mediated metabolism.
A number of exogenous and endogenous AHR ligands share planar structures with maximal dimensions of 14 Å × 12 Å × 5 Å, including the prototype 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) 3 . Considering that MC3 carries a planar NHC ligand ( Supplementary Fig. 1A), we hypothesized that MC3 can act as an AHR ligand and be metabolized by CYP1s. Indeed, global gene expression profiles obtained from DNA microarray analysis revealed that CYP1A1 was the top gene upregulated by MC3 among the~20,000 genes tested, whereas other CYPs remained nearly unaffected (Fig.  1a). In addition, the KEAP1-NRF2 signalling cascade, which has been reported as a downstream pathway of AHR 27 , was barely affected by MC3 treatment (Fig. 1a). MC3-mediated CYP1A1 expression was confirmed by RT-PCR, which showed~100-fold induction of CYP1A1 after MC3 treatment (Fig. 1b). This induction was evident as early as 30 min after MC3 treatment and increased over time in a concentration-dependent manner (Supplementary Fig. 1E and 1F). Of note, 1 h MC3 treatment induced higher CYP1A1 expression than the prototypic AHR agonist TCDD ( Supplementary Fig. 1G). To characterize the structural features responsible for the upregulation of CYP1A1 we tested two additional benzylimidazole-based, MC2 and MC4 19 , as well as two imidazole-based NHC gold compounds, IO1 and IO2 ( Supplementary Fig. 1H) 28,29 . We found that all NHC-gold compounds elevated CYP1A1 levels ( Supplementary Fig. 1F), whereas hydrophilic metal-free salt, 1,3-diethylbenzimidazoliumiodide (MC1), did not. For comparison, we tested the effect of auranofin ( Supplementary Fig. 1H), an FDA-approved gold complex for treatment of rheumatoid arthritis with covalent bonds to phosphine and thiol ligands in a linear arrangement. Results showed that auranofin is not an AHR ligand, because expression of CYP1s was not affected by its presence (Supplementary Fig. 1I). Taken together, our results indicate that MC2, MC3, and MC4 are activators of the CYP1 family most likely owing to the planar NHC structure.
Results obtained with immunostaining and immunoblotting showed that increased CYP1A1 protein expression after MC3 treatment is accompanied by the nuclear translocation of AHR (Supplementary Fig. 2A-2C, and Supplementary Fig. 6), an indicator of active AHR 30 . Co-incubation with resveratrol, a putative AHR antagonist 9 , led to a significant reduction of both TCDD-and MC3-mediated CYP1A1 gene expression (Fig. 1c).
We monitored the enzymatic activity of CYP1s in living cells using a commercially available dye Vividye BOMCC as described previously 18 . The result confirmed that resveratrol neutralized MC3-induced enzymatic activity of CYP1s ( Supplementary  Fig. 2D) and toxicity ( Supplementary Fig. 2E).
With the help of CRISPR/CAS9 genome editing, we generated HepG2 AHR knockdown (KD, Supplementary Fig. 2F) cells and observed significant reductions of TCDD-and MC3-induced CYP1A1 expression at the transcription and translation levels (Figs. 1d and e and Supplementary Fig. 6). Together, our results demonstrate that MC3 and other NHC-containing gold compounds are potent AHR ligands.
Immunosuppressive effect of MC3 in human primary T cells. Previous findings have shown that hepatocytes are part of the defense system and express a large number of immune proteins to regulate T-cell activity 31 . In line with this, MC3 treatment in HepG2 cells influenced a significant number of genes related to immune response ( Fig. 2a and Supplementary Fig. 3A), suggesting that MC3 modulates immune response through its interaction with the AHR.
To characterize direct immunomodulatory effects of MC3, we investigated different parameters of T-cell activation in freshly isolated human primary CD4 + T cells treated with MC3. Experiments were conducted using a previously published protocol, in which resting CD4 + T cells are incubated overnight with the drug and then activated by stimulating CD3-CD28 receptors 32,33 . We compared the effect of MC3 to those of auranofin and TCDD. MTT assays first confirmed that at the optimized concentration MC3 (0.5 µM), auranofin (0.5 µM), and TCDD (10 nM) did not have a significant effect on the viability of resting CD4 + T cells ( Supplementary Fig. 3B). To assess the effect of each drug treatment on T-cell antigen receptor (TCR) proximal signaling events, we analyzed the ability to spread on a TCR-stimulatory surface and form circumferential F-ACTIN rich rings (Fig. 2b). This system mirrors key events of TCR engagement by antigen presenting cells and the observed actin polymerization is essential for downstream TCR signaling 34 .
Although T-cell spreading and F-ACTIN ring formation were observed upon treatment with TCDD or auranofin, MC3 prevented cell spreading and significantly reduced ACTIN polymerization in response to TCR engagement (Fig. 2b). This defect in proximal TCR signaling translated in reduced downstream signaling because MC3 treatment led to significantly decreased cell surface levels of the T-cell activation marker CD25 ( Fig. 2c and Supplementary Fig. 3C), and, importantly, intracellular production of the key cytokine IL-2 ( Fig. 2d and Supplementary Fig. 3D) were significantly decreased upon treatment with MC3. TCDD did not show significant effects on T-cell activation, and, in line with previous results 35 , treatment with auranofin reduced the levels of CD25, IL-2 and CD38 expression (Fig. 2c-4e and Supplementary Fig. 3C-3E). None of the drugs significantly modulated the expression levels of the CD3 co-receptor, suggesting that the decreased activation after MC3 treatment is not owing to altered availability of binding targets for the activating beads ( Supplementary Fig. 3F and Supplementary Fig. 3G). Furthermore, impaired activation in MC3-treated lymphocytes was associated with decreased viability (Supplementary Fig. 3H), which is in line with previous reports 32, 36 . Our results demonstrate that MC3 interferes with CD4 + T-cell activations and unlike auranofin, also affects TCR early signaling events, suggesting a non-overlapping effect of these drugs in different steps of T-cell activation.

MC3-mediated immune suppression is TGFβ1-dependent.
Some potential mechanisms of action for the immunosuppressive effect of gold compounds such as thioredoxin reductase inhibition (TrxR) have been postulated; however, the precise molecular targets are still unknown 16,20 . As shown in Supplementary  Fig. 4A, the result from the in vitro enzymatic activity assay revealed that MC3 is a much weaker TrxR inhibitor (<30% inhibition at 0.5 µM) than auranofin, which has an EC50 of 0.0005 µM 20 . Most likely, TrxR is not a primary target of MC3. Early studies on primary T cells have highlighted the immunosuppressive function of TGFβ1 37,38 . The AHR signaling pathway has been reported to regulate TGFβ1 production and modulate immune responses 5,6 . In good agreement, our global gene expression profile revealed that MC3-regulated signaling pathways were involved in TGFβ1 production, epithelial-mesenchymal transition and wound healing ( Fig. 2a and Supplementary Fig. 3A). Thus, we investigated the role of TGFβ1 in MC3-mediated immune suppression in T cells. We screened a series of immortalized human T lymphocyte-related cells and found that TGFβ1 induction is accompanied by increased CYP1A1 expression after MC3 treatment, except for Jurkat cells, which express a low level of AHR ( Fig. 3a and Supplementary  Fig. 4B) 39 . We compared the MC3-mediated gene induction of CYP1A1, TGFβ1, and IL-2 in SupT1 cells lacking AHR using either CRISPR/CAS9 (AHR KD) or siRNA (AHR siRNA). AHR deficiency clearly compensated the effect of MC3 on all four genes ( Supplementary Fig. 4C).
Moreover, data obtained with an enzyme-linked immunosorbent assay confirmed the MC3-mediated increase of TGFβ1 only in SupT1 cells expressing AHR (Fig. 3b). Phosphorylation of SMAD3 is required for mediating TGFβ1 signaling 40 , which was elevated by MC3 in SupT1 cells, but remained unaffected in AHR KD cells and in presence of the TGFβ1 inhibitor SB4 ( Fig. 3c and Supplementary Fig. 4D and Supplementary Fig. 6). SMAD4 is required to transfer the cytosolic mediators of TGFβ1 signaling into nuclei and to trigger downstream expression of their target genes 40 . The treatment of either AHR siRNA or SMAD4 siRNA led to the inhibition of MC3-mediated SMAD3 phosphorylation, suggesting that AHR is an upstream of TGFβ/SMAD in this context ( Fig. 3d and Supplementary Fig. 6). Furthermore, the cellular level of IL-2 expression was significantly reduced by MC3 in AHR-deficient SupT1 cells (Fig. 3e).
In addition to repress IL-2 expression, TGFβ1 can also induce FOXP3 + Treg cells to suppress immune activity 41 . Thus, we selected FOXP3 as the second indicator of TGFβ-related immunosuppression induced by MC3. Adding TGFβ1 to production or e CD38 expression. For b, n = 3 donors, for c-e, n = 5 donors. Data were analyzed by one-way ANOVA or Friedman test. An appropriate transformation (Logit) was employed to restore normality when appropriate. One-way ANOVA t test was performed. *p < 0.05, **p < 0.01, ***p < 0.001; lower and upper ends of the bars indicate the minimum and maximum values, respectively, and the centre represents the median. Error bars ± SD. The source data for 2c-e are provided as Supplementary Data 1.
SupT1 cells promoted FOXP3 gene expression (Fig. 3f) and increased the percentage of CD4 + FOXP3 + cells (Figs. 3g and Supplementary Fig. 4E). As expected, MC3 treatment induced FOXP3 mRNA expression levels and increased the proportion of CD4 + FOXP3 + cells ( Fig. 3f and g). Either the chemical inhibition of AHR activity using resveratrol or the genetic knockdown of AHR compensated the MC3-induced FOXP3 expression. Moreover, the chemical inhibitor of the TGFβ1 receptor (SB4) or genetic knockdown SMAD4 (siS4) repressed the differentiation to FOXP3 + cells, whereas overexpression of SMAD3 (OES3) further increased FOXP3 expression. Taken together, our in vitro results demonstrate that MC3 activates TGFβ1 to suppress immune activity at least by the reduction of IL-2 expression and/or activation of FOXP3 + Treg.
Enrichment of FOXP3 + Treg in MC3-treated murine lymphocytes. To confirm the involvement of TGFβ in murine primary cells, we used a validated model of murine lymphocytes wherein TGFβ induction results in enrichment in Treg cells. In freshly isolated mouse CD4 + T cells, MC3 increased the proportion of Treg cells (defined as CD4 + CD25 + FOXP3 + ) from 4 to 14% at a concentration of 0.1 µM (Fig. 4a and Supplementary  Fig. 5A-C). In addition, the expression of AHR-related genes such as Cyp1a1 and Cyp1a2 (but not Cyp3a4), as well as Tgfβ1/2/ 3, were elevated ( Supplementary Fig. 5D). Moreover, resveratrol and SB4 neutralized the effect of MC3 (Fig. 4b). Thus, these results demonstrate that MC3 also repressed immune response in mouse CD4 + T cells in an AHR-TGFβ-dependent manner.  (Figs. 4c and d). In addition, we observed a remarkable reduction of autoimmune skin manifistations, such as scales, crust, and erosions on ears and tails. In some cases, severe inflammation on the ears was completely abolished even at 60 days (Fig. 4e). MC3-treated mice exhibited a delayed onset of skin manifestations and clearly lower clinical histopathology scores of the affected skin during the lifespan of N,N− dimethyl formamide (DMF)-treated scurfy mice g Comparison of body weight between mock treatment and MC3 treatment in WT and SC mouse. Because of spontaneous death in DMF-treated scurfy mice, no control group is available after 29 days. h Biodistribution of gold in various tissues isolated from MC3-treated WT mice at the indicated point times. Statistical analysis was performed by different tests: b One-way ANOVA t test was performed to compare more than two groups, for statistical analysis of survival ratios in C the Log-rank (Mantel-Cox) test was used. d Two-tailed unpaired t test with Welch's correction and in f+g two-way ANOVA was performed *p < 0.05, **p < 0.01, ***p < 0.001, ****p < < 0.0001. Lower and upper ends of the bars indicate the minimum and maximum values, respectively, and the centre represents the median. Error bars ± SD. The source data for 4a-d, 4f-h are provided as Supplementary Data 1.
( Fig. 4f) and significantly greater body weight than DMF-treated mice (Fig. 4g). We also observed the abundant gold in spleen, kindey, ear skin, and tail skin detected by total reflection X-ray fluorescence spectrometer, which might explain the immunosuppressive effect of MC3 in scurfy mice (Fig. 4h). Ectopic expression of activation markers, CD25 and CD69, was found in scurfy CD4 + T cells 43 . Analysis of immunosuppressive markers showed that MC3 reduced these CD4 + T-cell activation markers (CD25 + in lymph node (LN) and CD69 + in spleen) ( Fig. 5a and b); however, the proliferation of MC3-treated CD4 + T cells in scurfy mice was not affected (Supplementary Fig. 5E and S5F). In line with our previous results in vitro, MC3 induced the expression of Tgfβ1in the liver, thymus, and spleen as detected by RT-qPCR (Fig. 5c) and in situ hybridization (Fig. 5d). MTX is an immunosuppressant and has been used for the treatment of IPEX patients 13 . We isolated LN supernatant from scurfy mice treated with MTX or MC3. The quantified result using Legendplex clearly showed a significantly increased level of total TGFβ1 in the presence of MC3 (Fig. 5e). Furthermore, immunohistochemisty showed that MC3 significantly increased the expression of cytosolic and nucleic AHR in scurfy mice (Fig. 5f). This evidence supports the idea that MC3 is active in vivo as an immune suppressant and that, noteworthy, it may act through a Treg independent mechanism despite acting through TGFβ.

Discussion
Our results highlight a novel mechanism of action for MC3-like (i.e., carriers of a planar NHC ligand) gold compounds exerted by potent AHR activation, leading to downstream TGFβ1 production and immune repression in vitro and in vivo (Fig. 5g). The therapeutic potential of this signaling cascade was demonstrated by the effects of MC3, which led to reduced systemic inflammation and organ damage in scurfy mice. Recently, Masiuk et al. reported that the injection of autologous hematopoietic stem cells with lentivirus-based overexpression of functional FOPX3 could rescue autoimmune phenotypes in scurfy mice; however, lifespan was not improved probably owing to the low efficacy of transplantation 15 . By contrast, our study demonstrated that MC3 treatment clearly prolongs the lifespan of scurfy mice. In scurfy mice, the effective concentration of MC3 is 0.16 mg/kg/3 days and we did not find any toxic effect, suggesting a good clinical tolerance for MC3.
The MC3 gold(I) complex was initially developed as an anticancer agent 19,20 . Our recently developed liver-on-a-chip microfluidic system enabled us to reveal that MC3 and other NHC-based organometallics are potent AHR agonists and effectively induce CYP1A1 expression. AHR is a ligandactivated transcription factor, which is of importance in xenobiotic metabolism and immune responses 1 . CYP1s are the major family of enzymes involved in AHR-mediated xenobiotic metabolism 44,45 . Computational scanning of human, mouse, and rat genes revealed the presence of dioxin response elements in the promoter regions of several thousand genes, including TGFβ1/2/3 and various ILs 46 . The interaction between the AHR and TGFβ1 signaling pathway has been reported previously 7,47 . Our data corroborate this interaction by showing that MC3-induced TGFβ1 expression does not occur in AHR-deficient cells or in the presence of an AHR inhibitor (Fig. 5g). Likewise, the immunosuppressive effects of MC3 disappeared in both cases. Moreover, several reports have indicated that structurally-distinct AHR ligands differently modulate AHR activity 9,10 , which may account for the inability of TCDD to induce TGFβ1 and related immune repression in our experiments.
TGFβ1 is widely expressed in immune cells 48 and its deficiency causes multiple organ inflammation that leads to lethality in early mouse embryonic development 8 , indicating the essential role of TGFβ1 in adaptive immune response. Early experiments showed that TGFβ1 blocks IL-2 transcriptional expression to inhibit Tcell proliferation 37,38 . In line with this, we observed decreased IL-2 production in SupT1 CD4 + cells and primary CD4 + T cells after MC3 treatment. In the presence of MC3, primary CD4 + T cells were markedly less sensitive to exogenous stimulation as shown by the reduction of T-cell activation markers CD38 and CD25. In addition, the formation of immunological synapses was impaired as evidenced by the disruption of F-ACTIN rearrangements, which is one primary early steps of CD4 + T-cell activation 49 . Notably, both MC3 and auranofin, impaired CD4 + Tcell activation, but their mechanisms of action only partially overlapped: the latter did not exert an effect on F-ACTINmediated polarization. Moreover, in line with the finding that TGFβ1 promotes CD4 + T cells to Tregs by activating FOXP3 expression 9,50 , we detected an increased level of CD4 + FOXP3 + in human and mouse CD4 + T cells treated with MC3. Because scurfy mice lack functional Tregs, this suggests that the observed effect of MC3 was associated with other TGFβ1-related immunosuppressive signaling pathways (Fig. 5g). In general, functional TGFβ1 binds to receptors and phosphorylates regulatory-SMAD2/3, which in turn forms a complex with common SMAD4 and activates its downstream gene expression 40 . Reducing SMAD4 expression or using a chemical inhibitor of the TGFβ1 receptor compensated the immunosuppressive effect of MC3, whereas SMAD3 overexpression enhanced the function of MC3, indicating that MC3-mediated immunosuppression depends on TGFβ1 activity (Fig. 5g).
We observed that all three tested NHC-gold(I) complexes, but not auranofin, are AHR agonists. It is likely that organometallics containing planar NHC ligand are AHR agonists and can be metabolized by the CYP1 family. However, this hypothesis needs to be confirmed by adding various NHC-containing organometallics. Our in vivo data showed FOXP3-independent immunosuppression of MC3. This effect, as well as its FOXP3-dependent effect, needs to be elucidated in other animal models. It would also be interesting to examine how the AHR signaling pathway compensates loss-of-FOXP3 induced autoimmune response. Taken together, our results reveal that gold compounds containing planar NHC ligands utilize an alternative signaling pathway, which could explain their immunosuppressive effect and may have potential therapeutic use in immunology and related clinical applications. The mechanism of action showed herein overlaps only partially with that of a clinically approved gold compound, namely auranofin, As the clinical testing of auranofin to reduce immune activation in multiple diseases is expanding 51 , leaving open the possibility to combine different classes of gold compounds to enhance their immunosuppressive effect in vivo.
Cell culture. HepG2 and HeLa cells were cultured in Dulbecco's modified eagle medium, and A301 and SupT1 in Roswell Park Memorial Institute (RPMI) 1640 containing 10% fetal bovine serum (FBS) and 1% Pen/Strp under 5% CO 2 at 37°C in a humidified atmosphere and treated with compounds as indicated in the text. The percentage of CD4 + CD69 + cells in the LN and spleen. c Relative expression of Tgfβ1 in the liver, thymus, and spleen isolated from WT and SC mice treated with MC3. The data were normalized to respective mock treatment. d In situ hybridization of Tgfβ1 expression in the liver and thymus isolated from WT and scurfy mice. Scale bar: 100 µM. e The quantification of secreted total TGFβ1 from PMA/IONO restimulated lymph node cells isolated from scurfy mice treated either with immunosuppressant methotrexate (MTX; n = 5) or MC3 (n = 6) or the mock (DMF, n = 2) control using Legendplex. f Immunohistochemistry analysis on the expression of AHR in non-treated and MC3-treated scurfy mice. Scale bar: 60 µM. g Summary of MC3-mediated immunosuppression in an AHR-TGFβ-dependent manner. The number of mice used for LD analysis n WT mock = 11, n WT MC3 = 11, n SC mock = 3, and n SC MC3 = 5; for spleen analysis n WT mock = 10, n WT MC3 = 13, n SC mock = 3 and n SC MC3 = 5. For Legendplex, n SC mock = 2, n SC MC3 =6, and n SC MTX = 5. One-way ANOVA t test was performed. *p < 0.05, **p < 0.01, *** p < 0.001. The source data for 5a, 5b, 5c, and 5e are provided as Supplementary Data 1.
Primary CD4+ T cells were isolated from total blood obtained from healthy donors in agreement with the regulations of the Institute for Clinical Transfusion Medicine and Cell Therapy (IKTZ, in Heidelberg) The isolation of primary human CD4 + T cells was performed by negative selection using the RosetteSep Human CD4+ T-Cell Enrichment Cocktail according to manufacturer's instructions. Murine CD4+ T cells were isolated from inguinal lymph nodes of WT mice via MACS separation using the negative selection isolation kit EasySep mouse CD4+ T Cell (Stemcell, Cologne, Germany) according to manufacturer's instructions. Isolated CD4 + T-cells were seeded at 10 6 cells/mL in RPMI1640 containing 10% FBS and left untreated or treated with either 0.1-0.5 µM MC3, 0.5 µM auranofin, or 10 nM TCDD. Following overnight incubation, cells were either assayed by MTT and flow cytometry (resting cells) or activated by stimulating the CD3-CD28 receptors with Dynabeads Human T-Activator or Dynabeads Mouse T-Activator, respectively (Thermo Fisher Scientific, Darmstadt, Germany) according to manufacturer's instructions and left for 24-72 h in an incubator at 5% CO 2 and 37°C. Activating beads were then magnetically removed and cells were used for flow cytometric analysis or for evaluating TCR-mediated actin ring formation.
To monitor proliferation ability after treatment with MC3, autanofin or TCDD, isolated murine CD4+ T cells were stained with CFSE (Thermo Scientific) for 20 min at 37°C, followed by incubation with RPMI medium containing 10% FCS to stop the staining reaction. siRNAs were purchased from (Thermo Scientific) and knockdown experiments were followed the previous work 25 . SMAD3 overexpression plasmid was purchased from Addgene (#14052) and performed as described before 53 .
Mycoplasma contamination was determined routinely.
Microfluidic device. Cells were cultured in a chip consisting of two interconnected chambers. Each chamber comprises three inlet/outlet ports for cell seeding and culture medium supply, and a cell culture area surrounded by a guidance barrier. Prior to seeding, chips where coated with Collagen R 0.2% solution (Serva, Germany) for 45 min at 37°C and thoroughly washed with PBS. Cell type specificity in each designated chamber was achieved by the chip design and the seeding procedure, blocking the three ports of the opposing chamber. After 4 h medium was carefully exchanged and cells were left overnight to fully adhere. For microfluidic studies a low pressure syringe pump (nemesis, CETONI, Germany) was used in combination with a 1.0 ml syringe H 1/4"-28 Tubing Connector (ILS, Germany). Cells were treated with an intermitted flow profile of 5 min flow at 100 µL/h and 55 mins of 0 µL/h for a period of 48 h.
Cell apoptosis determination on chips. Cells were treated with increasing concentrations of MC3 for 24 h and fixed with 4% PFA at room temperature for 15 min and blocked with blocking buffer (5% goat serum, 1% BSA and 0.3% Triton X-100 in PBS) for 30 min. After aspirating the blocking buffer, cells were incubated with cleaved Caspase3 antibody (1:200, Cell Signaling, NEB, Germany) for 1 h. The secondary antibody (Goat anti-rabbit Alexa Flor 594; Dianova, Germany) was added and incubated for 30 min. Hoechst 33342 (1 µg/mL in PBS) was used to visualize nuclei. Images were taken on BIOREVO fluorescence microscope (BZ9000, KEYENCE). Image J software (the particle count function) was used to calculate the percentage of viability, as we described previously 17 .
CRISPR/Cas9 knockdown AHR. The CRISPR/Cas9 AHR plasmid (SC-400297-NIC-NIC) was purchased from Santa Cruz (Germany). The sequences of gRNA are CGGTCTCTATGCCGCTTGGA for strand A and TAATACAGAGTTGGACCG TT for strand B. Lipofectamine 3000 (Life Technologies, Germany) was used for transfection. According to previous work 53 , the transfection was carried out in a 24-well plate with seeding density of 400,000 cells/well for HepG2 and 50,000 cells/ well for SupT1. A mixture of 0.5 µg/well DNA and 1 µL Lipofectamine 3000 was added and incubated for 24 h. For the selection, 1 µg/mL puromycin was used. The established HepG2 AHR KD and SupT1 AHR KD cell lines were cultivated in respective medium containing 10% FCS, 1% PS, and 1 µg/mL puromycin for HepG2 and 3 µg/mL for SupT1 cells. The medium was changed two times per week.
Sulforhodamine B assay (SRB assay) and (3-(4,5-Dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide assay (MTT assay). Effects on cell growth were determined in cell lines using the SRB assay and MTT assay at an initial cell density of 5000 cells/well in a 96-well plate as previously reported 54  After subtraction of the blank, absorbance values were expressed as percentage of controls.
qRT-PCR. Quantitative real-time reverse transcription-PCR was performed according to manufacturer's instructions (Lightcycle 96, Roch, Germany) 19 . In brief, total RNA was isolated from cells using RNeasy kit from Qiagen. cDNA was generated by reverse-transcription of equivalent quantities of RNA using Proto-Script First Strand cDNA synthesis kit from NEB. qRT-PCR was performed using SYBR Green PCR master mix (qPCRBIO SyGreen Mix Lo-Rox, Nippon Genetics, Germany). Primer sequences are listed in Supplementary Table 1. Actin was used as an endogenous control.
Immunoblotting. Cell extracts were homogenized in urea-lysis buffer (1 mM EDTA, 0.5% Triton X-100, 5 mM NaF, 6 M Urea, 1 mM Na 3 VO 4 , 10 µg/mL Pepstatin, 100 µM PMSF and 3 µg/mL Aprotinin in PBS) as previously described 55 . The immunoblot was detected by using ECL solution. In all, 40 µg of total protein was resolved on 10% SDS-PAGE gels and immunoblotted with specific antibodies. Primary antibodies were incubated at a 1:1000 dilution in blocking buffer with gentle agitation overnight at 4°C. The results shown are representative of at least three independent experiments 55 .
Immunocytochemistry. HepG2 cells were seeded in a Ø-12 mm cover slip coated with Geltrex (Life Technologies, Germany) at a density of 200,000 cells/well. After 24 h cells, were incubated with MC3 as described in the text, fixed with 4% PFA at RT for 15 min, and blocked with blocking buffer (5% goat serum, 1% BSA and 0.3% Triton X-100 in PBS) for 30 min. The blocking solution was aspirated and incubated with AhR and CYP1A1 antibodies, respectively. The secondary antibody (Goat anti-rabbit Alexa Flor 594; Dianova) was added and incubated for 30 min.
Flow cytometry. Staining of surface markers (CD3, CD25, and CD38) was performed as previously described 57 60 . Prepossessing involved default image processing (with median-based local background), default Illumina removal for outlying observations, mean summarization of bead-level observations into probe-level data, quantile normalization of probe-level data, and log2 transformation. The resulting probes were matched to genes using annotations from the package illuminaHu-manv4.db 1.26.0. After filtering probes that poorly matched the annotated genes (quality status of no-match and bad) as well as low-expression probes (detection score > 0.05), 13,010 probes were available for clustering analysis with STEM, a tool for clustering short time series data that can differentiate between real and random temporal expression patterns. For this purpose, we selected one probe per gene (the one with highest variance, if multiple probes matched to single gene) and converted the data into fold changes with respect to the not-treated HepG2. We analyzed the data as short time series of length three (corresponding to time point 0, 1, and 24 h after treatment). We applied STEM in default mode, except for the following settings: at least one measurement with absolute fold change of 0.65; significance of model profiles corrected by false discovery rate method; cellular component terms, terms with evidence code IEA and NAS were excluded from the Gene Ontology 61 enrichment analysis.
Analysis of TVR-induced actin polymerization. Primary human CD4 + T-cells were isolated from total blood in healthy donors (in agreement with the regulations of the Institute for Clinical Transfusion Medicine and Cell Therapy, IKTZ, in Heidelberg). The detailed protocol can be found in the supplementary information. TCR signaling induced circumferential actin polymerization was quantified as previously described 62 . In brief, stimulatory coverslips were prepared by coating microscope glass coverslips with a 0.01% poly-L-lysine solution for 10 min at RT. Subsequently poly-L-lysine was aspirated and the coverslips were coated with 10 µg/ ml anti-CD3ε (clone HIT3a αCD3ε; BD Biosciences) in PBS and placed in a humidified chamber following 3 h incubation at 37°C. The stimulatory coverslips were washed with PBS and 100 µl of drug treated T cells (1 × 10 6 /100 µl) were loaded and allowed to adhere for 5 min before fixation with 3% PFA for 20 min at RT. The fixed cells were permeabilized for 2.5 min with 0.1% Triton X-100 and blocked with 3% BSA in PBS. Incubation with Phalloidin-Alexa-488 (1:1000 in PBS; Life Technologies) for 1 h was used to visualize F-actin. Samples were mounted on glass slides with Mowiol and analyzed by epifluorescence (Olympus IX81 S1F-3, cellM software). At least 100 cells per experiment and condition were analyzed for their ability to form a circumferential F-actin ring.
In vivo experiments. Scurfy or WT neonates were treated 3 days after birth with either 0.16 mg/kg MC3 or DMF as solvent control. 100 µl of MC3 or DMF diluted in medium was injected subcutaneously every 3 days. Body weight was controlled every 3 days. After 24 days, mice were killed and inguinal lns and spleen were used to generate single cell suspension, which was further analyzed by flow cytometry. The single cell suspension was stained with fixable dye Zombi Aqua (Biolegend), followed by staining with fluorescence-labeled antibodies against surface markers CD4 (RM4-4), CD25 (PC61), CD69 (H1.2F3), CD62L (MEL-14) in FACS buffer (PBS containing 1% FCS) for 30 min at 4°C. For nuclear staining, the cells were fixed with fix-perm buffer (eBioscience) for 45 min at 4°C. For characterization of regulatory T cells, the transcription factor foxp3 (FJK-16s) was stained for 30 min at 4°C in perm buffer. Analysis was performed with flow cytometer (Gallios, Beckman Coulter, Sinsheim, Germany). FACS data analysis was performed using the Kaluza software.
In situ hybridization. Paraffin blocks were prepared from different organs (spleen, thymus, kidney, and liver) and different treatments (in the same block for blind study) and cut into 4 μm slices, mounted on polyl-lysine-coated slides, air-dried O/N at 37°C and stored at 4°C (48 h).
Definition of clinical score. The different grades of skin changes were scored as follows: no sign of skin inflammation is displayed by score 0, score 1 describes mild flaking of ear and tail skin, if flaking, erythema, and thickening gets stronger and affects more areas a clinical score 2 is reached. Mice with strong skin manifestations, including open wounds, erosions, and necrotic areas of ear and tail are defined by clinical score 3.
Statistics and reproducibility. Data were analyzed with the GraphPad Prism software (v7.01, La Jolla, CA, USA). One-way analysis of variance (ANOVA) followed by Tukey's multiple comparisons test was employed when normal distribution could be assumed. Otherwise normal distribution was restored using the Logit transformation, followed by one-way ANOVA, or data were analyzed with an appropriate non parametric test, i.e., Friedman test followed by Dunn's multiple comparisons test.
All experiments have been independently repeated at least for three times, from which similar results were obtained.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
DNA microarray can be obtained from GEO (Accession code: GSE125146). Supplementary Data 1 contains the source data underlying the graphs and charts presented in the article, including Figs. 1B, 1C, 1D, 2B, 2C, 2D, 2E, 3A, 3B, 3E, 3F, 3G, 4A, 4B, 4C, 4D, 4F, 4G, 4H, 5A, 5B, 5C, and 5E, as well as supplementary figs. 1G, 1I, 2E,3B, 3G, 3H, 4A, 4B, 4C, 4D, 5A, 5C, 5D, 5E, and 5F. Received: 2 April 2019; Accepted: 28 November 2019; Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.