An Interaction between Kynurenine and the Aryl Hydrocarbon Receptor Can Generate Regulatory T Cells

It has long been recognized that the immune system is in a fine balance between immunity and self-tolerance. The concept of suppressor T cells playing a role in this balance was first proposed in the 1970s (1). Efforts to identify these cells were generally unsuccessful, and their very existence was brought into question in the early 1980s by molecular biologists who failed to locate an elusive suppressor gene in the mouse MHC class II locus (2). The suppressor T cell concept was dropped and remained out of vogue until it re-emerged as the CD4⁺CD25⁺ regulatory T cell (Treg), first described in detail by Sakaguchi in 1995 (3). Since that time, numerous studies have characterized these cells and the role they play in autoimmunity, control of infection, and transplant rejection. Identification of FoxP3, a transcription factor for Treg development, has led to further characterization of the importance of regulation (4, 5). More recently, a new Th cell lineage, termed Th17, was described (6, 7). These IL-17–secreting cells are thought to play a major role as effectors in autoimmunity and transplant rejection. Interestingly, these new data have led investigators to question previously held beliefs about terminal cell differentiation and stability of Tregs, as the ability of Tregs to redifferentiate into Th17 cells in the appropriate inflammatory milieu has now been described (8, 9).

Recent publications implicate the aryl hydrocarbon receptor (AHR) as a central player in T cell differentiation. The AHR is best known as the receptor for 2,3,7,8-tetrachlorodibenzo-p-dioxin (dioxin, or TCDD). Activation of the AHR by this environmental pollutant can lead to a range of toxic endpoints, including hepatocellular damage, epithelial changes, cancer, birth defects, thymic involution, and immunosuppression (10, 11). Although the AHR is well known for its role in toxicology, this receptor has also been shown to play a role in vascular and hematopoietic development (12, 13). Moreover, several potential endogenous ligands have been shown to bind to the AHR with variable affinity and potency (14). The fact that this receptor has been conserved in evolution (15) and that invertebrate orthologs of the AHR are well preserved and yet do not bind to TCDD (16) is consistent with the idea that a physiologically relevant endogenous ligand for this receptor exists (17).

In evaluating the mechanisms for immunosuppression seen postexposure to TCDD, it was discovered that activation of the AHR with TCDD leads to the generation of Tregs in vitro or in vivo (18), and, alternatively, activation with a different endogenous ligand, 6-formylindolo[3,2-b]carbazole (FICZ), leads to Th17 cell formation (19). Although this unusual pharmacology in which one AHR agonist diverts T cells toward regulators and another agonist generates effectors is surprising, multiple studies have confirmed the importance of the AHR in the generation of Th17 cells both in vitro and in vivo (20, 21). Regarding Treg generation, the direct relationship of regulatory cells to the AHR has been less clear (22). This has led investigators to question whether the AHR truly has a direct effect on the generation of these cells (23, 24).

Our laboratory has focused on the role of indolylic products as potential endogenous ligands of the AHR (25, 26). Therefore, we began to think about the potential for an interaction between the AHR and the IDO pathway. The IDO enzyme catalyzes the rate-limiting step of tryptophan degradation along the kynurenine pathway (27). IDO is present and activated in subsets of dendritic cells (DCs; particularly plasmacytoid DCs, or pDCs) and thought to be central to Treg generation from T cell precursors by DC–T cell interactions (28, 29). The exact mechanistic pathway by which IDO leads to Tregs has been debated, and both tryptophan starvation and direct effects of tryptophan metabolites (including kynurenine) have been proposed (30–32). In addition to a connection via indole metabolism, the IDO–AHR interaction was particularly interesting in light of the observation that IDO may be upregulated by the AHR (33, 34) and that kynurenine and related metabolites may be AHR agonists (35–37).

In this report, we demonstrate an important role for kynurenine, the first tryptophan metabolite of the IDO pathway, in Treg generation. We provide evidence that kynurenine activates the AHR at a dose clinically relevant in humans and leads to Tregs in vitro. The role for the AHR in this process is supported by two observations. First, kynurenine does not influence Treg generation in AHR-null T cells. Second, kynurenine can be shown to activate the AHR using classical response genes, such as Cyp1a1 and Cyp1b1. In our model, the AHR in T cells is required for the generation of Tregs by kynurenine. We further define the importance of the AHR for optimal generation of Tregs by TGF-β and characterize potential mechanisms for this.

C57BL/6J wild-type (WT) and BALB/c mice were obtained from The Jackson Laboratory (Bar Harbor, ME). AHR-null (AHR-deficient B6) mice on a C57BL/6J background (13) were bred and maintained under specific pathogen-free conditions. All animal experiments were carried out according to institutional guidelines approved by the University of Wisconsin School of Medicine and Public Health Animal Care and Use Committee (Madison, WI).

The method of murine bone marrow-derived DCs (BMDCs) was performed as previously described (38). Briefly, bone marrow was obtained from mice femurs. After RBC lysis, the cells were plated in six-well plates with a density of 1 × 10⁶/ml in complete RPMI 1640 media supplemented with 30 ng/ml GM-CSF. On day 3, nonadherent cells and 75% of culture media were exchanged for fresh media. On day 6, the cells were either harvested as immature DCs or cultured an additional day to maturity by again exchanging 75% of the media, with the addition of 50 ng/ml LPS. A total of 80% of the cell population stained positive for CD11c by flow cytometry. For analysis of mRNA expression, cells were treated with or without TCDD (10 nM). As mentioned above, 50 ng/ml LPS was used for maturation of BMDCs. There is a previous publication that LPS alone can lead to IDO (39). It should be clarified that this only occurred when higher doses of LPS (5 μg/ml) were used. This response is dose dependent, as seen by other investigators (33, 40, 41). As an additional control, an LPS titration was performed with BMDCs as further confirmation, the results of which are in the Supplemental Materials (Supplemental Fig. 1). No increase in IDO mRNA was seen until at least 100 ng/ml LPS was used in these assays.

A mouse hepatoma cell line H1L6.1c3, stably carrying a dioxin-responsive element (DRE)-driven firefly luciferase reporter gene [a gift from Dr. Denison, University of California, Davis, CA (42)] was maintained with 0.3 mg/ml G418 in completed DMEM media. Briefly, 0.6 × 10⁶ cells were seeded in each well of a six-well plate overnight and were then treated with different concentrations of TCDD, l-tryptophan, l-kynurenine, hydroxykynurenine, hydroxyanthranilic acid, anthranilic acid, nicotinamide, and quinolinic acid for the time specified. All of the kynurenines¹ (including kynurenine) were purchased from Sigma-Aldrich (St. Louis, MO), and purity was listed at ≥98% confirmed by HPLC. They were placed in solution in 0.5 M HCL, as recommended by the manufacturer for maximum solubility. Cells were lysed by lysis buffer (Promega, Madison, WI), and the luciferase assay was performed by using a BD monolight 3010 luminometer (BD Biosciences, San Jose, CA). The relative light unit is the indicator of luciferase expression level. All experiments were repeated three times, and each sample was tested in triplicate each time.

Total RNA was extracted using the reagents: RNeasy Mini Kit and RNase-Free DNase Set (Qiagen, Valencia, CA). A total of 500 ng total RNA in each group was used for RT reaction (iScript cDNA Synthesis Kit, Bio-Rad, Hercules, CA; or High-capacity cDNA Reverse Transcription Kits, Applied Biosystems, Foster City, CA). The relative quantitation PCR for IDO1 (Mm00492586-m1), GAPDH (4352339-E0806018), and IFN-γ (Mm99999071-m1) were performed in the Applied Biosystems 7900HT Fast Real-Time PCR System (Applied Biosystems), and TaqMan Universal PCR Master Mix (Applied Biosystems) was used as a reaction reagent. The relative quantitation PCR for Foxp3, Cyp1a1, Cyp1b1, and GAPDH were processed by the Bio-Rad iCycler (Bio-Rad) and iQ SYBR Green Supermix (Applied Biosystems) used as the reaction reagent.

Naive CD4 T cells were isolated from spleens of C57BL/6J and AHR-null mice using the CD4 CD62L Isolation Kit (Miltenyi Biotec, Auburn, CA) and an autoMACS. This kit includes a depletion mixture, including the addition of a CD25 and an anti-TCRγ/δ⁺ Ab. CD62L is expressed on naive T cells and downregulated upon activation. A small subset of central memory T cells also express CD62L and could be included in this separation. These represent a very small proportion of the final separation, and we will refer to separated cells as naive T cells. Cells were tested for purity postsorting and consistently showed >90% purity for CD4⁺CD62⁺CD25⁻ cells. An example of analysis of the separations is included in the Supplemental Material (Supplemental Fig. 2A). Viability at the beginning of culture was typically >98% as seen by trypan blue staining. For quantitative PCR (qPCR) analysis, 2–5 × 10⁵ cells were cultured in each well of a 96- well round-bottom plate coated with 0.5 μg/ml anti-CD3 and anti-CD28 overnight and then washed with PBS twice before seeding the cells. The naive T cells were maintained in F10 media supplemented with 10% heat-inactivated FBS, 100 μg/ml streptomycin, 100 U/ml penicillin, 50 μm 2-ME, 25 mM HEPES, and 2 mM l-glutamine and were treated with 10 nM TCDD, 100 nM FICZ, 2–10 ng/ml TGF-β (as specified), 50 μM kynurenine, or 25 μM each kynurenine (hydroxykynurenine, hydroxyanthranilic acid, anthranilic acid, nicotinamide, and quinolinic acid). After 5 d, the cultured cells were harvested for RNA assay. Prior to this, cells were checked for viability using live-dead staining with flow cytometry. The majority of cells were viable, but dead cells were gated out in flow cytometry analysis. For flow cytometric analysis, purified naive T cells were stimulated with the CD3/CD28 T cell Activation/Expander Kit (Miltenyi Biotec) for 5 d. As indicated, cultures were supplemented with recombinant cytokines and reagents: human TGF- β1 (R&D Systems, Minneapolis, MN), mouse IL-6 (20 ng/ml; R&D Systems), kynurenine, FICZ (100 nM), and AHR antagonist CH-223191 (Calbiochem, San Diego, CA).

To stain for Foxp3, T cells were first surface stained with anti-CD4 and anti-CD25 and then fixed and permeabilized with the Fixation/Permeabilization buffer (eBioscience, San Diego, CA) for 30 min at 4°C. Following this, cells were stained with Pacific Blue-conjugated anti-Foxp3. For intracellular IL-17 staining, T cells were first stimulated with 50 ng/ml PMA (Sigma-Aldrich) and 800 ng/ml ionomycin (Sigma-Aldrich) for 4 h in the presence of GolgiStop (BD Pharmingen, San Diego, CA) for the final 2 h. Cells were then fixed and permeabilized with the Fixation/Permeabilization buffer (eBioscience) and then stained with PE-conjugated anti–IL-17. All Abs were from eBioscience. Flow cytometric analysis was performed using an LSR-II (BD Biosciences).

Naive CD4⁺CD25⁻ T cells were isolated from WT and AHR-null mice and cocultured with BALB/c pDCs isolated using the Miltenyi Mouse pDC Isolation Kit (Miltenyi Biotec) at a ratio of 20:1 or 10:1 (example of pDC separation in Supplemental Fig. 2B). CpG, FICZ, and kynurenine were added at the start of culture. On day 5, cells were harvested and subjected to flow cytometric analysis.

We first examined the role that the AHR in DCs might play in directing T cell differentiation. Based on previously published data (29, 33, 34), we considered the possibility that IDO may be a mediator of the cross talk between these cells. We initially set out to confirm that IDO could be induced by activation of the AHR. Fig. 1 shows that cultured BMDCs, when exposed to the AHR agonist TCDD, led to Cyp1a1 induction, confirming activation of the AHR in these cells. Exposure to TCDD also increased IDO mRNA levels. DCs produced Cyp1a1 and IDO in both immature and mature states, indicating they did not have to be activated to have this response. AHR-null DCs did not exhibit IDO production postexposure to TCDD. In Supplemental Fig. 3, mRNA for IDO was analyzed at earlier time points than 48 h used above to control for any indirect affect that may have led to IDO expression. As early as 7 h postculture, IDO mRNA was generated.

Given that IDO can be generated by AHR activation in DCs and that tryptophan breakdown products have been known to generate AHR ligands, we examined all of the tryptophan breakdown products of the IDO pathway for their ability to activate the AHR. We employed a mouse cell line of hepatoma cells, termed Hepa1. These cells have been transfected with a luciferase reporter gene fused to the DRE (43). We tested each of the commercially available substrates in the kynurenine pathway, and compared their ability to activate the DRE with TCDD. Interestingly, kynurenine itself showed the strongest activity (Fig. 2A). Peak activity was at 5 h, with a dose (50 μm) that is comparable to levels encountered clinically in humans in areas of inflammation postactivation of IDO (44) (Fig. 2A). All other breakdown products showed less DRE activity, with decreasing peaks of activity the further down the kynurenine pathway that products were examined. As seen in Fig. 2B, we further tested this cell line using real-time PCR and found that postexposure to kynurenine in vitro, a substantial increase in Cyp1a1 and Cyp1b1 mRNA was seen, confirming activation of the AHR by this ligand. FICZ was also tested to compare its efficacy as an activator of the DRE in comparison with kynurenine. It is a potent ligand, with a peak at a later time point (10 h), seen in Fig. 2C.

The observation that IDO can be upregulated in DCs in an AHR-dependent manner and that kynurenine activates the AHR led us to consider that kynurenine could generate Tregs through the AHR. Using the well-documented technique for Treg generation with TGF-β and Ab stimulation (45), we initially exposed naive CD4⁺ T cells from WT and AHR-null animals to 2 ng/ml TGF-β and analyzed collected mRNA for FoxP3 expression. As seen in Fig. 3A, WT cells generated FoxP3, >40 times the response in AHR nulls. To further support this finding, we performed a similar experiment exposing WT or AHR-null cells to Ab stimulation and a higher dose of TGF-β and measured Treg generation by flow cytometry. This is represented in Fig. 3B, in which optimal Treg populations were generated in WT cells (29.1% in this representative assay), with a muted response from null cells (11.6% in this same assay). By titrating doses of TGF-β, we were able to yield increasing numbers of CD25⁺FoxP3⁺ cells seen by flow cytometry, represented graphically in Fig. 3C. When naive CD4⁺ T cells were separated from AHR-null animals, increasing doses of TGF-β had little effect on the generation of Tregs (Fig. 3B, 3C). To further test the importance of AHR-ligand binding in Treg generation, we repeated Ab stimulation of naive WT T cells with titrating doses of TGF-β, and this time included an AHR antagonist (CH-223191) known to competitively bind to the receptor. As seen in Fig. 3D, the addition of antagonist blocked the increase of Tregs seen by flow cytometry postexposure to TGF-β in vitro.

The above experiments all demonstrate the importance of the AHR in Ab-stimulated Treg generation via TGF-β. Given our suspicion that cell–cell contact is important in AHR-dependent Treg generation, we employed an in vitro system separating pDCs and exposing them to allogeneic naive CD4⁺ T cells (pDCs were derived from BALB/c mice, and naive T cells from C57BL/6J mice). This system was previously shown to be dependent on IDO for successful generation of Tregs (29). We were able to repeat the findings that pDCs exposed to CPG led to significant generation of FoxP3⁺ Tregs in WT allogeneic naive T cells (Fig. 3E). When naive T cells were isolated from AHR-null mice, a low percentage of Tregs were identified prior to manipulation. Addition of CpG did increase Treg generation, but the expression was dramatically less robust than in the WT cells (Fig. 3E).

As it is well known that IDO leads to the generation of Tregs, and it appears that kynurenine activates the AHR, the next step was to assess whether kynurenine can directly lead to FoxP3 expression. We first cultured mouse naive CD4⁺ T cells with Ab stimulation for 5 d in the presence of 10 ng/ml of TGF-β, kynurenine, TCDD, or FICZ. RNA was then harvested and tested for the presence of FoxP3. As seen in Fig. 4A, top panel, only TGF-β and kynurenine led to significant induction of FoxP3 RNA. We analyzed FoxP3 induction in triplicate in 11 separate biological experiments and achieved a fold change of 3.2, which was significantly increased from control with a p value of 0.017. When AHR-null cells were used (Fig. 4A, bottom panel), only TGF-β (10 ng/ml) yielded significant induction of FoxP3 RNA. To further show that kynurenine is leading to FoxP3 mRNA via an interaction with the AHR and not in some indirect way dependent on TGF-β, we cultured mouse naive CD4⁺ T cells as above with Ab stimulation either with or without kynurenine. RNA was then harvested and tested for Cyp1a1, Cyp1b1, and TGF-β. As seen in Fig. 4B, kynurenine exposure led to significant amounts of Cyp1a1 compared with control, but an increase of TGF-β mRNA over control was not seen, making it unlikely that kynurenine is acting indirectly by generating this cytokine. We then performed a similar experiment, exposing naive T cells to Ab stimulation with and without kynurenine and after 5 d of culture used flow cytometry to assess for FoxP3. As seen in Fig. 4C, cells that were not exposed to kynurenine showed minimal FoxP3 expression, but those exposed to kynurenine had a significant shift, with 24.4% more FoxP3 expression in analyzed cells in one representative assay. We performed this experiment six times and observed significant FoxP3 protein expression in four out of six assays. This was also repeated with titrating doses of kynurenine with or without the presence of an AHR antagonist, and these data further confirm that kynurenine at a dose of 50 μM leads to FoxP3⁺ T cells, and the AHR antagonist decreased the percentage of FoxP3⁺ cells both at baseline and postexposure to kynurenine.

We next used the in vitro system separating pDCs and exposed them to allogeneic naive CD4⁺ T cells. As seen in Fig. 4D, in WT T cells, the addition of CpG exhibited the greatest expression of Tregs, but kynurenine also yielded elevated Treg generation. We also tested FICZ, and this ligand not only did not lead to an increased Treg population, but also decreased the percentage of CD4⁺CD25⁺FoxP3⁺ cells compared with untreated control. When AHR-null mice were used as the source for naive T cells, very few Tregs were generated when exposed to pDCs. The addition of kynurenine did not cause any enhancement of Treg formation, presumably due to the lack of the AHR on the T cells. As further confirmation, we repeated the DC/T cell coculture with a 1:20 ratio of allogeneic BALB/c DCs to naive C57BL/6J T cells multiple times and present the summarized data graphically. As shown in Fig. 4E, the 1:20 ratio yielded generation of Tregs by flow, and kynurenine led to significant Treg generation as compared with untreated cells. When repeated with AHR-null T cells, CpG was able to yield Tregs, but kynurenine did not, supporting the dependence of the function of kynurenine on the expression of AHR by the T cell.

To assess the importance of the AHR on DCs in this model, we performed this experiment using AHR-null cells as the source for pDCs (on a C57BL/6J background), and naive T cells were taken from WT mice (BALB/c). As seen in Supplemental Fig. 4, under these circumstances, similar generation of Tregs was seen using null pDCs as WT pDCs, indicating, at least in this model, that the presence of the AHR is necessary on the T cell, and not the pDC, for optimal Treg generation.

The data in this section highlight that the AHR on T cells is activated by kynurenine and leads to Treg induction.

To better understand the role of the AHR in TGF-β–dependent Treg generation, we next extracted total RNA from naive T cells, either fresh or after 20 h or 3 d of culture, and conducted real-time PCR for AHR expression. Culture conditions included Ab stimulation with FICZ, kynurenine, or TGF-β. There is AHR expression at baseline (Fig. 5A), which increases >4-fold at 20 h with exposure to TGF-β and remains >3-fold elevated at 3 d. We additionally looked at Cyp1b1 expression at 20 h and 3 d, and, as seen in Fig. 5A, FICZ and kynurenine led to 20 and 50 times mRNA production over baseline at 20 h, respectively, with Cyp1b1 levels remaining 20 times elevated at 3 d after kynurenine exposure. Culturing in the presence of TGF-β did lead to a small increase in Cyp1b1 (∼4 times over baseline at 20 h), but much less than seen with FICZ or kynurenine. To assess whether AHR upregulation secondary to TGF-β would potentiate binding of ligands to the AHR, we compared the expression of Cyp1a1 and Cyp1b1 after kynurenine exposure with and without TGF-β, which is represented in Fig. 5B. The response is strongly enhanced after TGF-β exposure, shifting the curve up significantly, implying that TGF-β does potentiate the binding of kynurenine to the AHR when this ligand is present in the culture.

Given that the AHR has also been implicated in the generation of Th17 cells when bound to certain ligands (FICZ), we wondered whether IDO pathway products could also favor Th17 morphology when present in a milieu favoring Th17 generation. We used the Th17-generating conditions described previously, based on exposure of naive CD4⁺ T cells to IL-6 and TGF-β (21). We first repeated the finding that FICZ leads to enhancement of IL-17⁺ cells (23) (Fig. 6). As mentioned previously, FICZ is thought to act primarily through the AHR, confirming that the AHR can promote T cell differentiation to both Treg and Th17 differentiation depending on the milieu. We then tested kynurenine and found no effect on the generation of Th17 cells (Fig. 6). This indicates that activation of the AHR with different ligands can lead to entirely different outcomes depending on the surrounding milieu.

Collectively, the data represent a novel way that IDO, via the kynurenine pathway, leads to Treg generation. It is generally believed that T cell differentiation depends on interactions between DCs and T cells (46–48), with IDO playing a role in this. The above findings establish a direct relationship between the AHR and the fate of T cells in vitro. IDO is generated by pDCs. This leads to tryptophan metabolism and kynurenine formation. Kynurenine binds to the AHR in T cells, leading to differentiation to CD25⁺FoxP3⁺ T cells. Absence of the AHR in T cells prevents this effect. Kynurenine was previously identified in a review of AHR ligands (35), which corresponds to the knowledge that tryptophan metabolites can lead to ligands of the AHR; the finding that this ligand–receptor interaction leads to Treg generation is novel. Whereas Fig. 1 does indicate that IDO induction by DCs can be stimulated via the AHR, the physiologic significance of this needs to be further defined. IFN-γ does stimulate IDO in AHR-null DCs (data not shown), and Supplemental Fig. 4 would suggest that Tregs can be induced by AHR-null pDCs in coculture assays, but the AHR needs to be present on T cells for optimal Treg generation.

One issue that needs to be addressed is the reliance of TGF-β on the AHR. Fig. 3 indicates that optimal generation of Tregs by TGF-β is dependent on the presence of the AHR, similar to a report published previously (21). This decrease in Treg production is demonstrated both in T cells obtained from null mice and also with the use of the AHR antagonist (Fig. 3D). It is unlikely that this cytokine binds directly to the AHR, given its structure (we have tested this in a DRE luciferase assay with no response to TGF-β alone). More likely, it is secondary to the effect demonstrated in Fig. 5, which shows that TGF-β with Ab stimulation leads to an upregulation of the AHR in culture. This effect is seen in the first 24 h in culture and seems to persist at least 3 d, according to our data and the literature (21, 49). It is important to note that CD4⁺CD25⁻ T cells do express the AHR prior to its upregulation, which is demonstrated by both Western blot (21, 49) and DNA microarray (21) and in our own data (Fig. 5). This is further demonstrated by the fact that exposure of naive T cells to kynurenine or FICZ leads to mRNA transcription of Cyp1a1 and Cyp1b1, which would only occur in the presence of the AHR. It is possible that AHR upregulation alone leads to Treg generation after TGF-β exposure, which would correlate with a previously published experiment in which cotransfection of a construct coding for mouse AHR into a bacterial artificial chromosome with FoxP3 tagged with a Renilla luciferase reporter led to upregulation of Renilla activity (18). Another possibility is that the upregulation of the AHR allows endogenous ligands present in the system (either in media or secreted from cells during inflammation) to bind to the increased receptor with enhanced effect. This is supported by the data in Fig. 4C, which indicate that the AHR antagonist reduces the amount of FoxP3⁺ cells seen at baseline, as well as the fact that we do see some Cyp1a1 and Cyp1b1 induction in T cells postexposure to TGF-β (Fig. 5A, 5B). Perhaps there is ongoing binding between the AHR and endogenous ligands (which may include kynurenine in an in vivo system), and blocking the receptor blocks this Treg-generating effect. Although endogenous ligands may play a role in this differentiation, the effects of kynurenine in our assays far outweigh ligands that may already be present in the media. We tested the role of the AHR with TGF-β in Fig. 5B, in which kynurenine was titrated in culture with CD4⁺CD25⁻ T cells either in the presence of or without TGF-β. When this cytokine was present, the response of Cyp1a1 and Cyp1b1 was dramatically elevated when exposed to increasing doses of kynurenine, much higher than was seen with TGF-β alone. This would further support that kynurenine is a ligand of the AHR and that TGF-β potentiates the effect of kynurenine binding by increasing the amount of receptor, far beyond what was seen with any ligand already present in the media.

A second important question is whether kynurenine undergoes catabolism, and is it actually a metabolite or breakdown product that is binding the AHR. It is possible that kynurenine in solution, like other small molecules, undergoes breakdown through various mechanisms. We used a 0.5 M solution of HCL to get kynurenine into solution, as per the recommendations of the manufacturer regarding maximum solubility. In an effort to test for breakdown of kynurenine, we have conducted HPLC with diode array detection to analyze kynurenine in this solution, as well as in buffered solution at a physiologic pH (50–53). In addition, the manufacturer has tested kynurenine in HCL by HPLC and found it to be >98% pure, which is similar to our findings (Supplemental Fig. 5). As can be seen in the figure, kynurenine dissolved in HCL at 1 d shows minimal decomposition, whereas kynurenine dissolved in buffer shows decreased purity. It was more difficult to dissolve kynurenine in buffered solution, which took up to 8 h at 37°C, as opposed to kynurenine in HCL, which went rapidly into solution and stayed in solution throughout the assays. We did find that kynurenine in bicarbonate buffer did not strongly activate the DRE in the luciferase assay when made fresh, but after a few days in culture displayed strong activity, which may correlate with solubility issues. When we examined kynurenine in HCL at 3 wk and 7 mo by HPLC, it continued to show minimal breakdown, indicating there is stability when placed in this solution. It is well known in the literature that the isoforms of kynurenine can be modified depending on the surrounding milieu in vivo (50–53), and it is difficult to rule out that this may occur to some degree in our assays. Nevertheless, if it is a metabolite of kynurenine binding to the AHR, it is still formed early in the kynurenine pathway, generated by effects of IDO on tryptophan. This does not diminish the importance of these findings, still linking IDO and the kynurenine pathway to the AHR.

It is fascinating that some ligands (like FICZ) activate the AHR, as seen by the luciferase assays and mRNA analysis, but do not lead to FoxP3 expression. Other ligands (kynurenine) activate this same receptor and do lead to FoxP3 on T cells. We have considered how different ligands might activate the same receptor and lead to disparate outcomes in protein generation. One hypothesis that we are investigating is that kynurenine itself may be enzymatically modified by the cytochrome P450 enzymes that are induced by the AHR, whereas FICZ may not. The product of this modification may directly lead to FoxP3 induction as opposed to a direct effect of kynurenine itself. This would explain this differential effect of these AHR ligands. This hypothesis is further supported by the fact that TGF-β can generate FoxP3 in our assays as early as 3 d, whereas the kynurenine-induced generation is typically not seen prior to 5 d (data not shown), despite the fact that the AHR is activated by kynurenine within a few hours. We will test this theory further by using known inhibitors of Cyp1a1 and Cyp1b1 enzymes [trans-stilbenes (54)] and experimenting with our recently generated DRE cluster null mice (55), which have dysfunctional Cyp1a1 and Cyp1a2 enzymes.

Regarding the concentration of kynurenine used in these experiments, the dose chosen was physiologic (44, 56), comparable to levels encountered in humans in areas of inflammation. It is also consistent with observations on the amounts of kynurenine and other tryptophan breakdown products generated in vitro by DCs (57) (5–50 μM range). In areas without an ongoing immune response, concentrations of the kynurenines are significantly lower, in the nanomolar range, but increase 1000-fold in microenvironments of inflammation. The fact that kynurenine concentrations in vivo only reach the doses examined in our assays in areas of inflammation could serve as a way to localize the IDO-AHR–dependent Treg generation. Conceivably, this would allow an immune response to commence, leading to IFN-γ and other inflammatory cytokines, which would then induce DCs to generate IDO. Ultimately, kynurenine levels would build up sufficiently to interact with the AHR and generate Tregs (and AHR expression would be enhanced by TGF-β expression, further potentiating the effects of kynurenine), hence muting the immune response. In areas without inflammation, kynurenine doses would be inadequate to generate significant amounts of AHR-dependent Tregs.

These data shed light on the direct mechanism of IDO, which has been controversial. Two theories on the function of IDO in Treg generation have been proposed. The first is that IDO leads to tryptophan depletion, and this relative starvation leads to cell cycle arrest in some populations, favoring generation of Tregs (32). The second theory is that the tryptophan catabolites themselves have a more direct role in the generation of Tregs, supported by some observational studies but without any clearly understood mechanism (30, 31). The data in this paper strongly support the second theory, and experiments to elucidate the exact molecular responses elicited after binding of kynurenine or FICZ to the AHR are underway. It is still plausible that tryptophan depletion continues to play a role in the IDO-dependent generation of Tregs, and this paper does not exclude this. In fact, as mentioned in our 1Materials and Methods section, we did perform the assays with naive T cells alone exposed to kynurenine in a low tryptophan media (F10), based on previously published data (32) and our own experience that the yield in this media is superior to when RPMI media (with higher tryptophan content) is used. We have contemplated that the reason low tryptophan solutions lead to more regulatory cells is that tryptophan itself is known to form other potential ligands of the AHR, including FICZ. In our in vitro assay, we use a low tryptophan solution with kynurenine as the primary AHR ligand. Perhaps this shifts the balance toward generation of regulatory cells, whereas when tryptophan is present to start with, additional ligands can be generated that shift the balance toward an effector response. It is possible that both mechanisms play a role and in fact may explain why DCs are so potent at generating Tregs in coculture assays (as they generate IDO and deplete the tryptophan present in the assays). Of course, other mechanisms, including cell–cell contact and cytokine release, may also be involved.

The data in this manuscript further our understanding of the emerging role of the AHR as a key player in the differentiation of T cells to Tregs, as well as the ultimate balance of regulatory and effector responses in immunity. As this pathway is further characterized, improved understanding will enhance our knowledge of T cell differentiation and yield new strategies of modulating the balance of regulation and effector response using ligands of the AHR.

Disclosures C.A.B. has served as a scientific consultant to Dow Chemical on issues related to dioxin toxicity.