Main

Degradation of Trp by indoleamine-2,3-dioxygenases 1 and 2 (IDO1/2) in tumours and tumour-draining lymph nodes inhibits antitumour immune responses1,2,3,4,5 and is associated with a poor prognosis in various malignancies6. Inhibition of IDO1/2 suppresses tumour formation in animal models1,3 and is currently tested in phase I/II clinical trials in patients with cancer7. The relevance of Trp catabolism to human tumour formation and progression is, however, as yet unknown.

TDO degrades Trp to Kyn in human brain tumours

A screen of human cancer cell lines revealed constitutive degradation of Trp and release of high micromolar amounts of Kyn in brain tumour cells, namely glioma cell lines and glioma-initiating cells (GICs), but not human astrocytes (Fig. 1a). IDO1 and IDO2 did not account for the constitutive Trp catabolism in brain tumours (Supplementary Fig. 1a–e). Instead, tryptophan-2,3-dioxygenase (TDO), which is predominantly expressed in the liver (Supplementary Fig. 3a, b) and is believed to regulate systemic Trp concentrations6, was strongly expressed in human glioma cells (Supplementary Fig. 1b) and correlated with Kyn release (Fig. 1b). Pharmacological inhibition or knockdown of TDO blocked Kyn release by glioma cells, whereas knockdown of IDO1 and IDO2 had no effect (Fig. 1c, d and Supplementary Fig. 2a), thus confirming that TDO is the central Trp-degrading enzyme in human glioma cells. In human brain tumour specimens, TDO protein levels increased with malignancy and correlated with the proliferation index (Fig. 1e–g and Supplementary Fig. 2b–d). As described previously8, healthy human brain showed only weak TDO staining in neurons (Fig. 1e). TDO expression was not confined to gliomas but was also detected in other cancer types (Supplementary Fig. 3b, c). Lower Trp concentrations were measured in the serum of patients with glioma (Fig. 1h). These may not have translated into increased Kyn levels (Fig. 1h) because Kyn is taken up by other cells and metabolized to quinolinic acid. Indeed, accumulation of quinolinic acid was detected in TDO-expressing glioblastoma tissue (Fig. 1i and Supplementary Fig. 3d).

Figure 1: TDO degrades Trp to Kyn in human brain tumours.
figure 1

a, Trp (left) and Kyn (right) content in the supernatants of human astrocytes (hAs), glioma cell lines and GICs (T323) cultured for 72 h and measured by high-performance liquid chromatography (HPLC) (n = 4). b, Correlation between TDO mRNA and Kyn release by human glioma cells measured by quantitative RT–PCR and HPLC (n = 4). c, Kyn concentrations in the supernatants of U87 glioma cells cultured for 48 h in the presence of the TDO inhibitor 680C91 (black bars) or its solvent (white bars; n = 4, P = 0.005, 0.002 and 0.0009 for 1, 5 and 10 μM 680C91, respectively). d, Kyn release by glioma cells after knockdown of TDO (black bars; P = 0.000007, 0.0007 and 0.00006, respectively), IDO1 (dark grey bars) or IDO2 (light grey bars) by siRNA (n = 3). e, Upper panel: weak neuronal TDO expression in healthy brain tissue. Lower panel: TDO expression in glioblastoma (WHO grade IV). TDO staining is in red. Asterisk indicates necrosis; arrowheads indicate the border to infiltrated brain tissue. Insets: single tumour cells (arrows) infiltrating the adjacent brain tissue. Magnifications: ×40 (main panels), ×400 (upper inset) and ×100 (lower inset). f, Plot of TDO expression (H-score; see Supplementary Methods) in brain tumours of increasing malignancy (WHO grades II–IV; grade II, n = 18; grade III, n = 15; grade IV, n = 35). g, Correlation of the Ki-67 proliferative index with the TDO H-score in gliomas of different WHO grades (n = 42). h, Trp (left) and Kyn (right) concentrations in the sera of 24 patients with glioblastoma and 24 age-matched and sex-matched healthy controls, measured by HPLC. i, Quantification of staining with quinolinic acid in healthy human brain tissue (white bar; n = 5) and glioblastoma tissue (black bar; n = 5). The data distribution in f and g is presented as box plots, showing the 25th and 75th centiles together with the median; whiskers represent the 10th and 90th centiles, respectively. Error bars indicate s.e.m.

PowerPoint slide

Effects of TDO-mediated Kyn release on immune cells

Kyn suppresses allogeneic T-cell proliferation9. Allogeneic T-cell proliferation was inversely correlated with Kyn formation by glioma-derived TDO (Fig. 2a and Supplementary Fig. 4a, b). Knockdown of TDO in glioma cells (Supplementary Fig. 4c, d) restored allogeneic T-cell proliferation, and the addition of Kyn to the TDO knockdown cells prevented the restoration of T-cell proliferation (Fig. 2b). The proliferation of CD4+ and CD8+ T cells stimulated by the T-cell receptor was inhibited by Kyn in a concentration-dependent manner (Supplementary Fig. 4e). In addition, knockdown of TDO resulted in enhanced lysis of glioma cells by alloreactive peripheral blood mononuclear cells (Supplementary Fig. 4f). Finally, decreased infiltration with leukocyte common antigen (LCA)-positive and CD8+ immune cells was observed in sections of human glioma with high TDO expression in comparison with those with low TDO expression (Fig. 2c), indicating that Kyn formation by TDO may suppress antitumour immune responses. In vivo experiments in immunocompetent mice demonstrated that tumours expressing TDO grew faster and had a higher proliferation index than TDO-deficient control tumours (Fig. 2d and Supplementary Fig. 4g–i). TDO activity in tumours suppressed antitumour immune responses in vivo, as demonstrated by a decreased release of interferon-γ by tumour-specific T cells and tumour cell lysis by spleen cells of mice bearing TDO-expressing tumours in comparison with mice bearing TDO-deficient tumours (Fig. 2e, f).

Figure 2: Paracrine and autocrine effects of TDO-mediated Kyn release by glioma cells.
figure 2

a, Correlation of the proliferation of peripheral blood mononuclear cells (PBMCs) cultured with allogeneic glioma cell lines with the Kyn release of the glioma cells (n = 3). b, Proliferation of PBMCs cultured with allogeneic TDO-expressing control U87 glioma cells (sh-c) in comparison with U87 glioma cells with a stable short hairpin RNA-mediated knockdown of TDO (sh-TDO), with or without 100 μM Kyn (black bars), in comparison with PBMCs alone with or without 100 μM Kyn (white bars; n = 3). c, Quantification of LCA+ cells (left) and CD8+ cells (right) stained in human glioma sections with low TDO expression (H-score < 150, white bar; n = 12 for LCA, n = 10 for CD8) and in human glioma sections with high TDO expression (H-score ≥ 150, black bar; n = 17 for LCA, n = 10 for CD8). d, Growth of Tdo-deficient GL261 murine glioma cells stably transfected with Tdo (filled circles) or empty vector (open circles) injected subcutaneously into the flank of C57BL/6N mice was monitored using metric callipers (n = 6). Tumour weight (g) was calculated as 0.5 × length (cm) × width2 (cm2). e, Interferon-γ release by T cells of mice bearing subcutaneous Tdo-expressing tumours (black bar) in comparison with T cells of mice bearing Tdo-deficient tumours (white bar) after re-stimulation with glioma lysates measured by ELISpot (n = 3). f, Lysis of GL261 murine glioma cells by spleen cells of mice with Tdo-expressing GL261 tumours in comparison with those with subcutaneous Tdo-deficient GL261 tumours, measured by chromium release (n = 4). g, Quantification of the migrated distances of sh-c (open squares) and sh-TDO (filled circles) cells into a collagen matrix (n = 3, P = 0.004, 0.0005 and 0.01 for 24, 48 and 72 h, respectively). h, Clonogenic survival of sh-c (white bar) and sh-TDO (black bar) U87 cells (n = 3). i, Matrigel Boyden chamber assay of U87 glioma cells in the absence or presence of 70 μM Trp without or with 30 or 60 μM Kyn (n = 3). j, Clonogenic survival of LN-18 glioma cells in the absence or presence of 70 μM Trp without or with 30 or 60 μM Kyn (n = 3). k, Representative cranial MRIs, haematoxylin/eosin staining (H&E) and nestin staining of CD1 nu/nu mice implanted with sh-c (upper panels) or sh-TDO (lower panels) U87 glioma cells. The images are representative of two independent experiments (n = 6). l, Tumour weight of sh-c (white bars) and sh-TDO (black bars) U87 glioma cells injected subcutaneously in the flank of CD1 nu/nu mice that were treated either with anti-asialo GM1 antibody (asialo) for NK-cell depletion or control IgG (IgG) (n = 8). Error bars indicate s.e.m.

PowerPoint slide

Effects of TDO-mediated Kyn release on glioma cells

We next assessed the autocrine effects of Kyn on glioma cells. Although no differences in cell cycle progression were detected between controls and glioma cells with TDO knockdown (Supplementary Fig. 5a), knockdown of TDO decreased motility and clonogenic survival (Fig. 2g, h and Supplementary Fig. 5b–d). This was mediated by Kyn because exogenous addition of Kyn restored motility and clonogenic survival in the absence of Trp (Fig. 2i, j and Supplementary Fig. 5e, f), suggesting that Kyn increases the motility of malignant glioma cells. In GICs, sphere formation was enhanced in response to Kyn (Supplementary Fig. 5g). Finally, tumour formation was impaired when TDO knockdown tumours were orthotopically implanted in the brains of nude mice, which are devoid of functional T cells (Fig. 2k and Supplementary Fig. 5h, i). To analyse whether TDO-mediated inhibition of antitumour natural killer (NK)-cell responses, which are functional in nude mice, might account for the impaired formation of TDO knockdown tumours, we compared subcutaneous tumour growth in the presence or absence of NK cells. NK-cell depletion (Supplementary Fig. 5j) enhanced the growth of both control and TDO knockout tumours but did not restore the growth of TDO knockout tumours to that in controls (Fig. 2l and Supplementary Fig. 5k), suggesting that Kyn generated by constitutive TDO activity enhances the malignant phenotype of human gliomas in an autocrine manner in the absence of functional antitumour T-cell and NK-cell responses.

Kyn activates the aryl hydrocarbon receptor

For a better understanding of the molecular mechanisms underlying the autocrine effects of Kyn on glioma cells, we performed microarray analyses of Kyn-treated glioma cells revealing broad induction of aryl hydrocarbon receptor (AHR) response genes by Kyn (Fig. 3a and Supplementary Figs 6a, b and 7). Pathway analyses showed that the 25 genes that were most strongly induced by Kyn treatment in U87 cells at 8 h and at 24 h were all directly or indirectly regulated by the AHR (Fig. 3a and Supplementary Fig. 6b). The AHR is a transcription factor of the basic helix–loop–helix (bHLH) Per–Arnt–Sim (PAS) family, which is activated by xenobiotics such as benzo[a]pyrene and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)10. Malignant glioma cell lines and GICs express the AHR constitutively (Supplementary Fig. 6c)11, and upregulation of AHR target genes by Kyn was confirmed in two different glioma cell lines (Supplementary Fig. 6d, e). Kyn led to translocation of the AHR into the nucleus after 1 h, thus showing an immediate effect of Kyn on the AHR (Fig. 3b, c and Supplementary Fig. 8a). In accordance with this, western blot analyses of Kyn-activated tumour cells showed decreased cytoplasmic localization paralleled by increased nuclear accumulation of the AHR comparable to that induced by TCDD (Fig. 3d). In the nucleus the AHR forms a heterodimer with the AHR nuclear translocator (ARNT) that interacts with the core-binding motif of the dioxin-responsive elements (DRE) located in regulatory regions of AHR target genes12. Kyn induced DRE–luciferase activity in glioma cells, with a concentration giving half-maximal response (EC50) of 36.6 μM (Fig. 3e, Supplementary Fig. 8b). AHR activation was unique to Kyn in a panel of Trp catabolites (Supplementary Table 1). An ethoxyresorufin-O-deethylase (EROD) assay confirmed the induction of the functional AHR target gene CYP1A1, encoding cytochrome P450, family 1, subfamily A, polypeptide 1, with an EC50 of 12.3 μM for Kyn (Supplementary Fig. 8c). Radioligand binding assays with mouse liver cytosol from Ahr-proficient and Ahr-deficient mice showed that Kyn binds to the AHR with an apparent Kd of roughly 4 μM (Fig. 3f).

Figure 3: Kyn activates the AHR.
figure 3

a, Connection of the 25 genes that were most strongly induced by Kyn treatment in U87 cells after 8 h to AHR signalling (red, upregulation; green, downregulation). b, Translocation of green fluorescent protein (GFP)-tagged AHR into the nucleus of mouse hepatoma cells, which do not degrade Trp, after 3 h treatment with 50 μM Kyn, 50 μM Trp or 1 nM TCDD (negative control: medium). c, Ratios of nuclear to cytoplasmic fluorescent intensity in cells with GFP-tagged AHR after 3 h of indicated treatment (negative control: medium; positive control: 1 nM TCDD, 50 μM Kyn). The data distribution is represented by box plots, showing the 25th and 75th centiles together with the median; whiskers represent the 10th and 90th centiles, respectively (P < 0.001, one-way analysis of variance on ranks, followed by Dunn’s method). d, AHR western blots of two different nuclear (N) and cytoplasmic (C) fractions each of control (lanes 1 and 2), Kyn-treated (lanes 3 and 4) and TCDD-treated (lanes 5 and 6) human LN-229 glioma cells. e, Dioxin-responsive element (DRE) chemical activated luciferase gene expression in U87 glioma cells treated with the indicated Kyn concentrations (n = 2). f, Radioligand binding assay with indicated concentrations of l-[3H]Kyn using mouse liver cytosol from Ahr-proficient and Ahr-deficient mice. Specific binding was calculated by subtracting the radioactivity measured in Ahr-deficient cytosol from that of Ahr-proficient cytosol (n = 4). g, CYP1A1 mRNA expression in sh-AHR LN-308 glioma cells (black bars) in comparison with controls (sh-c; white bars) treated with 100 μM Kyn, 1 nM TCDD or controls (n = 4). h, mRNA expression of AHR target genes in sh-TDO (black bars) in comparison with sh-c U87 glioma cells (white bars; n = 4) (AHRR, P = 0.02; CYP1A1, P = 0.0007; IL1Α, P = 0.001; IL1B, P = 0.0000006; IL6, P = 0.0047; IL8, P = 0.01; PAI-2, P = 0.0005; TIPARP, P = 0.06). Error bars indicate s.e.m.

PowerPoint slide

Activation of the AHR and upregulation of AHR-regulated gene expression in response to Kyn were inhibited by the AHR antagonist dimethoxyflavone or knockdown of AHR (Fig. 3g and Supplementary Fig. 8g–k), indicating that Kyn is a specific agonist of the AHR. The involvement of the same or similar AHR residues in the binding to Kyn, TCDD and 3-methylcholanthrene was confirmed by the fact that dimethoxyflavone inhibited the activation of the AHR by all three ligands (Supplementary Fig. 8g–i).

The effects of TDO-derived Kyn are mediated by the AHR

The endogenous production of Kyn in glioma cells was sufficient to activate the AHR, because knockdown of TDO decreased the expression of AHR-regulated genes (Fig. 3h and Supplementary Fig. 8l–o). Because mean Kyn concentrations of 37.01 ± 13.4 μM were measured in U87 xenografts (n = 6), sufficient Kyn concentrations to activate the AHR were also reached in vivo.

In accordance with activation of the AHR by TDO-derived Kyn, expression of the AHR target gene TIPARP in LCA+ immune cells was observed only in human glioma sections expressing TDO (Fig. 4a). To determine whether TDO influences antitumour immune responses through the AHR we analysed the infiltration of immune cells in human glioma sections in relation to their AHR expression. Indeed, infiltration by LCA+ and CD8+ immune cells was decreased in sections of human gliomas with high AHR expression compared with those with low AHR expression (Fig. 4b). To analyse the contribution of host AHR expression to tumour growth, we compared the growth of murine tumours with and without Tdo expression in Ahr-deficient and Ahr-proficient mice. The growth of Tdo-expressing tumours was attenuated in Ahr-deficient mice when compared with Ahr-proficient mice (Fig. 4c) indicating that AHR-mediated host effects enhance tumour growth. Staining of LCA+ immune cells in the tumours revealed that expression of TDO decreased the infiltration with LCA+ immune cells in Ahr-proficient mice but not in Ahr-deficient mice (Fig. 4d and Supplementary Fig. 9a), suggesting that TDO-mediated suppression of antitumour immune responses through the AHR contributes to the host effects enhancing the growth of Tdo-expressing tumours. In addition, while in Ahr-proficient mice the expression of Tdo strongly enhanced tumour growth in comparison with tumours not expressing Tdo, the same effect was observed in Ahr-deficient mice, although to a much smaller extent (Fig. 4c). Because murine glioma cells express functional AHR (Supplementary Fig. 9b–e), these results suggest that the increase in tumour growth mediated by TDO in Ahr-deficient mice is due to autocrine effects of TDO on the tumour cells themselves.

Figure 4: The autocrine and paracrine effects of TDO-derived Kyn are mediated through the AHR.
figure 4

a, Immunofluorescence staining of LCA and TIPARP in human glioma sections with low or high TDO expression. Magnification ×400. b, Quantification of LCA+ cells (left) and CD8+ cells (right) stained in human glioma sections with low AHR expression (Histoscore < 150, white bar; n = 10 for LCA, n = 8 for CD8) and in human glioma sections with high AHR expression (H-score ≥ 150, black bar; n = 12 for LCA, n = 12 for CD8). c, Tumour weight measured 15 days after subcutaneous injection of murine GL261 glioma cells with and without Tdo expression in the flanks of Ahr-proficient mice (white bars) or Ahr-deficient mice (black bars; n = 6). d, Quantification of LCA+ immune cells stained in the subcutaneous Tdo-proficient and Tdo-deficient GL261 tumours in Ahr-proficient and Ahr-deficient mice presented as box plots, showing the 25th and 75th centiles and the median (n = 4). WT, wild type. e, Migration of sh-c LN-308 glioma cells (white bars) and LN-308 glioma cells with knockdown of the AHR by two different shRNAs (grey bars, sh-AHR1; black bars, sh-AHR2) in the presence or absence of 100 μM Kyn (n = 4). f, Clonogenicity of sh-c (white bars) and sh-AHR (black bars) LN-308 glioma cells with or without 100 μM Kyn (n = 3). g, Growth of AHR-proficient (filled circles) and AHR-deficient (open circles) human LN-308 glioma cells injected subcutaneously into the flank of CD1nu/nu mice was monitored with metric callipers (n = 7). Tumour weight was calculated as in Fig. 2d. P values: day 5, 0.147; day 9, 0.546; day 12, 0.027; day 16, 0.008; day 19, 7.18 × 10−4; day 22, 1.77 × 10−5; day 26, 1.57 × 10−5; day 30, 8.49 × 10−4; day 34, 8.26 × 10−4; day 41, 0.022; day 45, 0.0477. Error bars indicate s.e.m.

PowerPoint slide

This notion is supported by the fact that Kyn failed to induce motility of human glioma cells after AHR knockdown (Fig. 4e). In addition, the increase in clonogenic survival in response to Kyn was abolished in glioma cells with a knockdown of AHR (Fig. 4f). Finally, in vivo experiments demonstrated that induced knockdown of AHR in human glioma cells inhibited tumour growth in immunocompromised mice (Fig. 4g and Supplementary Fig. 9f), underscoring the importance of AHR signalling for the autocrine effects of Trp degradation.

TDO-derived Kyn activates the AHR in human cancer

Next we aimed at investigating whether TDO-derived Kyn activates the AHR in human brain tumour tissue. Indeed, TDO expression correlated with the expression of the AHR and AHR target genes in human glioma tissue (Fig. 5a–c and Supplementary Fig. 9g), indicating that constitutive TDO expression in glioma cells produced sufficient Kyn levels to activate the AHR. To address whether the TDO–Kyn–AHR signalling pathway is also activated in cancers other than glioma, we analysed microarray data of diverse human tumour entities. Interestingly, TDO expression correlated with the expression of the AHR target gene CYP1B1 not only in glioma (Fig. 5c) but also in B-cell lymphoma, Ewing sarcoma, bladder carcinoma, cervix carcinoma, colorectal carcinoma, lung carcinoma and ovarian carcinoma (Fig. 5d, Supplementary Fig. 10a and Supplementary Table 2). This finding indicates that the TDO–Kyn–AHR pathway is not confined to brain tumours but seems to be a common trait of cancers. Analysis of the Rembrandt database revealed that the overall survival of patients with glioma (WHO grades II–IV) with high expression of TDO, the AHR or the AHR target gene CYP1B1 was reduced in comparison with patients with intermediate or low expression of these genes (Fig. 5e and Supplementary Fig. 10b)13. Finally, in patients with glioblastoma (WHO grade IV)14, the expression of the AHR targets CYP1B1, IL1B, IL6 and IL8, which are regulated by TDO-derived Kyn in glioma cells (Fig. 3h and Supplementary Fig. 5d, e), were found to predict survival independently of WHO grade (Fig. 5f and Supplementary Fig. 10c), thus further underscoring the importance of AHR activation for the malignant phenotype of gliomas. In summary, these data suggest that endogenous tumour-derived Kyn activates the AHR in an autocrine/paracrine fashion to promote tumour progression (Fig. 5g).

Figure 5: TDO-derived Kyn activates the AHR in diverse human cancers, and AHR activation predicts survival in patients with glioma.
figure 5

a, Correlation of TDO expression (red) and AHR expression (brown) in consecutive sections of human glioblastoma tissue. Arrows indicate vessels for orientation. Magnification ×40; insets ×200. b, Correlation between TDO and AHR expression in human glioma tissue based on H-scores of TDO and AHR, calculated by Spearman rank correlation (n = 26). c, Correlation between TDO and CYP1B1 expression in microarray data of human glioblastoma (n = 396) analysed by Spearman rank correlation. d, Correlation between TDO and CYP1B1 expression in microarray data of human bladder cancer (left; n = 58), human lung cancer (centre; n = 122) and human ovarian carcinoma (right; n = 91) analysed by Spearman rank correlation. e, Survival probabilities of patients with glioma (WHO grades II–IV) with high expression (red) of TDO (left) or AHR (right) compared with those of patients with intermediate (blue) or low (green) expression of these genes derived from Rembrandt13. f, Survival probabilities of patients with glioblastoma with high expression (red) of the AHR target gene CYP1B1 compared with those of patients with low (green) expression of CYP1B1 derived from the glioblastoma data set of the Cancer Genome Atlas network14 (n = 362). g, Synoptic figure highlighting the autocrine and paracrine effects of TDO-derived Kyn on cancer cells and immune cells through the AHR.

PowerPoint slide

Discussion

Cancer-associated immunosuppression by Trp degradation has until now been attributed solely to the enzymatic activity of IDO in cancer cells and tumour-draining lymph nodes. IDO inhibition is therefore currently being evaluated as a therapeutic strategy in the treatment of cancer in clinical trials7 despite some off-target effects on human cancer cells15. We show that TDO is strongly expressed in cancer and is equally capable of producing immunosuppressive Kyn. In IDO-negative glioma cells, TDO seems to be the sole determinant of constitutive Trp degradation, indicating that TDO represents a novel therapeutic target in glioma therapy. In fact, an orally available TDO inhibitor has recently been developed16. Inhibition of TDO may not only restore antitumour immune responses but also act on the tumour cell intrinsic malignant phenotype, because we delineated the importance of constitutive Trp degradation to sustain the malignant phenotype of cancer by acting on the tumour cells themselves.

Emerging evidence indicates a tumour-promoting role of the AHR. AHR activation promotes clonogenicity and invasiveness of cancer cells11,17. Transgenic mice with a constitutively active AHR spontaneously develop tumours18,19, and the repressor of the AHR (AHRR) represents a tumour suppressor in multiple human cancers20. The aberrant phenotype of Ahr-deficient mice points to the existence of endogenous AHR ligands21. Although different endogenously produced metabolites such as arachidonic acid metabolites, bilirubin, cyclic AMP, tryptamine and 6-formylindolo[3,2-b]carbazole (FICZ) have been shown to be agonists of the AHR22, their functionality has not been convincingly demonstrated in a pathophysiological context such as cancer or immune activation. The search for endogenous ligands of the AHR is therefore continuing.

We now link these two important pathways contributing to cancer progression by showing that Trp catabolism leads to AHR activation, and we provide evidence of a pathophysiological human condition that is associated with the production of sufficient amounts of a functionally relevant endogenous AHR ligand. Our results reveal a differential response of primary immune cells and transformed cancer cells to AHR-mediated signals, which is in line with various toxicological studies using the classical exogenous AHR ligands TCDD and 3-methylcholanthrene11,17,23. Exposure to these xenobiotics leads to profound suppression of cellular and humoral immune responses23, while also promoting carcinogenesis and inducing tumour growth11,17. These cell-specific differences in AHR effects are likely to depend on the expression of factors differentially regulating AHR signal transduction such as the AHRR20 as well as cell-specific transcription factor crosstalk shaping the response to AHR activation24.

It is likely that Kyn-mediated activation of the AHR is not only relevant in the setting of cancer. For instance, activation of the mouse and human AHR by agonistic ligands induces regulatory T cells25,26,27,28. Mice with a poor-affinity AHR suffer from exacerbated autoimmune encephalomyelitis in the absence of an exogenous ligand26, and Trp catabolites suppress autoimmune neuroinflammation29,30, suggesting that activation of Trp catabolism represents an endogenous feedback loop to restrict inflammation through the AHR. In fact, exogenous Kyn is involved in the regulation of immune cells in mice through the AHR31,32. Kyn concentrations sufficient to activate the AHR are also generated by IDO in response to inflammatory stimuli (Supplementary Fig. 11a–c). In a broader context, a significant number of malignancies arise from areas of mostly chronic infection and inflammation33, where Trp catabolism in the tumour microenvironment is activated and sustains local immune suppression34. Activation of the AHR by Kyn generated in response to inflammatory stimuli may thus constitute a previously unrecognized pathway connecting inflammation and carcinogenesis.

Methods Summary

TDO expression was analysed by immunohistochemistry in human tumours. Its relevance for Trp degradation was determined by genetic knockdown or overexpression of TDO. Trp and Kyn were measured in cell culture supernatants, human sera and xenograft tissue by high-performance liquid chromatography. Mixed leukocyte reactions, chromium release, ELISpot and staining of immune cells in tumour tissues were used to assess the immune effects of TDO activity. Cell cycle analysis, Matrigel and spheroid invasion assays, scratch assays, sphere formation assays and clonogenicity assays were employed to analyse the autocrine effects of TDO activity. All animal procedures followed the institutional laboratory animal research guidelines and were approved by the governmental authorities. Orthotopic implantation of human glioma cells with and without stable knockdown of TDO into CD1nu/nu mice, subcutaneous injection of these cells into NK-depleted or wild-type CD1nu/nu mice and subcutaneous injection of murine Tdo-proficient and Tdo-deficient GL261 cells into syngeneic C57BL/6N mice were performed to analyse the autocrine and paracrine effects of TDO activity in vivo. Microarray analysis of Kyn-treated human glioma cells was performed to identify signalling pathways activated by Kyn. Analysis of AHR translocation, DRE–luciferase assays and radioligand binding assays confirmed activation of the AHR by Kyn. Pharmacological inhibition and stable knockdown of the AHR (in vitro and in vivo) proved that the effects of Kyn are AHR dependent. Injection of Tdo-proficient and Tdo-deficient tumour cells into Ahr+/+ and Ahr−/− mice was used to address the contribution of host effects to TDO-mediated cancer promotion. Finally, stainings, microarray data and clinical data on human tumour tissues were used to analyse whether TDO activates the AHR in human cancers and how this affects survival.