Inhibition of Tryptophan-Dioxygenase Activity Increases the Antitumor Efficacy of Immune Checkpoint Inhibitors

Cancer immunotherapy induces sustained clinical responses in a significant fraction of patients with cancer; however, a majority of patients fail to respond (1, 2). Several resistance mechanisms could be involved, including the presence of immunosuppressive pathways acting in the tumor microenvironment. Some of these immunosuppressive pathways may be amenable to pharmacologic inhibition, which could increase the efficacy of immunotherapy (3, 4). One of these mechanisms relies on the catabolism of l-tryptophan (Trp), which is converted into kynurenine (Kyn) by two distinct enzymes, indoleamine 2,3-dioxygenase (IDO1) and tryptophan 2,3-dioxygenase (TDO). Local tryptophan depletion combined with kynurenine production impairs cellular immunity, by reducing T-cell proliferation, inducing differentiation of regulatory T cells and driving dendritic cells towards a tolerogenic phenotype (5–7). In humans, IDO1 expression is restricted in normal tissues (8), but is strongly induced by IFNγ (9, 10). IDO1, which is highly abundant in inflammatory sites, appears to contribute to suppressing immune responses to prevent immunopathology (11, 12). In addition to this immunoregulatory effect, IDO1 activity can also promote pathogenic inflammation favoring tumor angiogenesis (13).

The exact mechanisms involved in the immunosuppressive effects of Trp catabolism by IDO1 remain unclear. In particular, the respective role of Trp depletion versus Kyn production in the microenvironment has been debated. Trp depletion was initially suggested to impair T-cell proliferation through the activation of an integrated stress response involving GCN2 stress kinase (14), but this was not confirmed in other studies (15, 16). Trp depletion can also lead to deactivation of the mTOR pathway (17). Kyn and/or its derivatives also contribute to immunosuppression through the induction of T-cell apoptosis, the induction of regulatory T-cell differentiation, possibly via activation of the aryl hydrocarbon receptor (AhR), or the upregulation of PD-1 on CD8⁺ T cells, also via AhR activation (18–22). Kyn degradation by in vivo administration of kynureninase increases the efficacy of checkpoint inhibitors or cancer vaccines (23).

IDO1 is largely described as an immunosuppressive protein, playing a major role in feto-maternal tolerance (24), but also in cancer, as around 50% of human tumors express IDO1 (8, 25). IDO1 expression in human tumors can either be constitutive, driven by COX-2 (26), or induced by IFNγ released by tumor-infiltrating lymphocytes (8, 11, 27). Tumoral expression of IDO1 is often associated with poor prognosis and a more aggressive tumor phenotype (8, 28). IDO1 inhibitors (IDOi) have been developed and tested in numerous preclinical studies, showing synergistic antitumor activity when combined with immunotherapies such as vaccination or immune checkpoint blockade (25, 29–31). These results have prompted the clinical development of IDOi (32, 33). However, the first IDOi that completed phase III testing failed to improve the clinical activity of anti-PD1 in patients with melanoma (34, 35). The reasons for this clinical failure remain unclear, but could be related to an insufficient inhibition of IDOi activity and/or the involvement of other Trp-degrading enzymes (36, 37).

The other enzyme that converts Trp into Kyn is TDO, which is encoded by the gene TDO2. As opposed to IDO1, TDO is constitutively and highly expressed in normal liver tissue, where it mainly degrades excess dietary Trp to maintain systemic levels of approximately 80 μmol/L (11, 12, 38). This function is illustrated by the 9-fold higher levels of Trp in the serum of the TDO-KO mice compared with wild-type (WT) mice (39). TDO is expressed in human tumors, particularly in hepatocarcinomas, melanomas, bladder carcinomas, and glioblastomas, which can produce Kyn, activate AhR, enhance tumor cell survival, and reduce infiltration by immune cells (40, 41). TDO-transfected P815 mouse tumors resist immune rejection by mice immunized against P1A, the dominant MAGE-type antigen expressed in these tumors (40). Tumor rejection is restored upon treatment of the mice with a TDO inhibitor (TDOi; ref. 40). These results provide proof of concept for the clinical development of TDOi, which is ongoing. However, very few other studies have characterized the immunosuppressive role of TDO in cancers. In particular, the immune effects of the modulation of systemic Trp by TDO inactivation have not been addressed. In this report, we further described the role of TDO in modulating antitumor immune responses and the antitumor efficacy of checkpoint inhibitors using a novel TDOi or TDO-KO mice. In the accompanying manuscript, we characterize the expression of TDO in human tumors and normal tissues and define the TDO-expressing cell types and their impact on T-cell infiltration (42). Together, these studies will help to better define the role of TDOi in cancer immunotherapy.

The mouse colon carcinoma line MC38 was obtained in 2013 from Dr P. Berraondo (University of Navarra, Navarra, Spain). The mouse colon carcinoma line CT26 was obtained from ATCC. CT26 cells were transfected with empty vector (pCDneo, #15ACV2VC from GeneArt) or with a vector coding for mTdo2 (#15ADFX6C from GeneArt) using Lipofectamine 3000. CT26ev and CT26mTDO cells were selected with neomycin. Expression of TDO in CT26mTDO cells was validated by RT-PCR, western blot, and enzymatic activity assay (Supplementary Fig. S1). CT26mTDO cells were cultured in the presence of TDOi LM10 (20 μmol/L; refs. 40, 43). Derivatives of the P815 mouse mastocytoma cell line were obtained in our lab (25, 40). A172 (human glioblastoma) cells were purchased from Antisense Pharma and were authenticated in 2018 by short tandem repeat profiling (Promega Powerplex hs 16). Human HEK-293-EBNA (293E) cells were purchased from InvivoGen and transfected with hTDO2 or hIDO1 as described previously (25, 40). They were not authenticated in the past year. We also transfected murine cells P815B with hTDO2 using the pEF6/V5-His vector described in ref. 40 to obtain P815B-hTDO clone 19. We also transfected human 293E and murine P815B cells with expression vectors pEF6/V5-His encoding the human INDOL1 (referred hereafter as IDO2) full-length (FL) or short-length (SL) open reading frame [starting at methionine 14, reportedly more active than the full-length (ref. 44; ref seq: NM_194294)], to obtain 293E-hIDO2(FL) clone 5, P815B-hIDO2(FL), and P815B-hIDO2(SL). The hIDO2(FL) was obtained from lipopolysaccharide-treated dendritic cells RNA using the following primers: Forward (Fwd)-AGG-CCA-CCA-CAA-GAA-TGT-TGC and Reverse (Rev)-GCA-GCC-TCC-TAA-CCA-CGT-G. The hIDO2(SL) ORF was subcloned from pEF6 hIDO2(FL) using the following primers: Fwd-CTT-CAA-ACA-AAA-TAA-TGG-AGC-CCC-AC and Rev-GCA-GCC-TCC-TAA-CCA-CGT-G and inserted into vector pEF6/V5-His. The two plasmids were sequenced and we excluded the presence of SNP rs10109853 (R248W polymorphism; ref. 45). Expression of hIDO2(FL) and (SL) in both cell lines was validated by RT-PCR (Supplementary Fig. S2). All cell lines were tested for Mycoplasma in 2019.

To determine the IC₅₀ of PF06845102/EOS200809 (the TDOi used in this study) on TDO-expressing cells, we used human glioblastoma cells A172, which constitutively express TDO (40, 41); hTDO2-transfected human cells 293E-hTDO (clone 119; ref. 40); mTdo2-transfected murine cells P815B-mTDO (clone 12; ref. 40); murine cells CT26mTDO described above; hIDO1-transfected human cells 293E-hIDO1 (clone 17); mIdo1 transfected murine cells P815B-mIDO1 (clone 6; refs. 25, 40), and hIDO2-transfected 293E and P815B cells as described in the previous section. Cells were incubated (2 × 10⁵ cells/well) in a final volume of 200 μL of Iscove's modified Dulbecco medium (GIBCO-Thermo Fisher Scientific; which contains 80 μmol/L Trp) supplemented with 2% FCS in presence of the TDO inhibitor at different concentrations (1–10,000 nmol/L) for 24 hours. Cells were centrifuged and 55 μL of supernatant was mixed with 55 μL of 12% (wt/vol) trichloroacetic acid. After centrifugation, supernatants were used to quantify Kyn concentration by high-performance liquid chromatography (HPLC) based on the retention time and the UV absorption (360 nm for Kyn; ref. 40).

C57BL/6J Ola Hsd mice (Envigo), C57BL/6 IDO1-KO mice (Jackson Laboratory), and C57BL/6 TDO-KO mice (provided by Dr Hiroshi Funakoshi and Dr Toshikazu Nakamura from Osaka University, Osaka, Japan; ref. 39) were bred at the animal facility of the Ludwig Institute for Cancer Research (Brussels, Belgium). Mice were used at 8–12 weeks of age. C57BL/6 and C57BL/6 TDO-KO mice were bred to obtain heterozygous offspring, which were intercrossed to obtain Tdo2^+/−, Tdo2^+/+ (WT), and Tdo2^−/− (TDO-KO). Homozygous TDO-KO and WT mice littermates were selected by PCR on tail genomic DNA using primers: mTdo2: Fwd-GTA-TCT-ATG-GAG-GAC-AAT-GAA-G, Rev-GAT-GAA-TAG-GTG-CTC-GTC-ATG; neomycin-resistance gene: Fwd-GTT-CTT-TTT-GTC-AAG-ACC-GA, Rev-TTT-CCA-CCA-TGA-TAT-TCG-GC (Supplementary Fig. S3A). TDO-KO and IDO1-KO mice were mated to obtain double IDO1/TDO-KO mice, which were selected with the above Tdo2 primers, and the following primers: Ido1: Fwd-TGG-AGC-TGC-CCG-ACG-C, Rev-TAC-CTT-CCG-AGC-CCA-GAC-AC; neomycin-resistance gene: Fwd-CTT-GGG-TGG-AGA-GGC-TAT-TC, Rev-AGG-TGA-GAT-GAC-AGG-AGA-TC (Supplementary Fig. S3C). Homozygous IDO1/TDO-KO mice were then bred independently. BALB/c mice (Charles River) were maintained at the animal facility of the Institute for Medical Immunology (Gosselies, Belgium). Animal studies were conducted in accordance with national and institutional guidelines for animal care and with the approval of the Comité d'Ethique pour l'Expérimentation Animale from the Secteur des Sciences de la Santé, UCLouvain [2015/UCL/MD/14 and 2012-03 (BALB/c)].

Mice received a single oral gavage (200 μL) of the indicated dose of TDOi PF06845102/EOS200809 suspended in methocel vehicle after 10-minute sonication.

Blocking monoclonal anti-CTLA4 (9H10), anti-PD1 (RMP1-14), and control isotype IgG2a (2A3) were purchased from BioXCell and were injected intraperitoneally at the doses indicated in the figure legends. The standard and Trp-low diets were prepared by Special Diet Services Ltd, as described by Maeta and colleagues (46). Besides Trp, the two diets were similar in composition, except for a small excess of the other amino acids in the Trp-low diet. Amino acid compositions (%) were as follows in the standard/Trp-low diets, respectively: Arg 0.91/1.11; Lys 0.66/1.01; Met 0.22/0.55; Cys 0.24/0.36; Trp 0.18/0.06; His 0.35/0.55; Thr 0.49/0.72; Ile 0.54/0.74; Leu 0.98/1.10; Phe 0.66/0.73; Val 0.69/0.73; Tyr 0.49/0.37; Gly 1.11/1.84; Glu 3.17/2.76; Pro 1.20/0.46; Ser 0.56/0.46; Ala 0.16/0.46. Respective diets were given to mice starting 1 day before tumor injection. TDOi PF06845102/EOS200809 is described in the published patent application US 20150225367 (47). Epacadostat (Synnovator; ref. 48) and TDOi PF06845102/EOS200809 were put in suspension after 10-minute sonication in methocel (vehicle) and administered twice a day by oral gavage at 100 mg/kg in a volume of 200 μL. Tumors were obtained by subcutaneous injection of 1 × 10⁵ (CT26mTDO) or 5 × 10⁵ cells (MC38). Tumor size was measured twice a week with a caliper, and the volume reported with the formula (w² × L/2, where w is width and L is length). Liver metastases were established by injecting 2 × 10⁵ MC38 cells in one half of the spleen of anesthetized mice (49). Three minutes after injection, hemisplenectomy was performed to remove the injected half and reduce the outgrowth of tumors in the spleen. Three weeks later, mice were sacrificed and liver metastases were counted by visual inspection.

Sera were collected after 5-minute centrifugation (8,000 × g) of blood collected in the retro-orbital sinus. Tumor interstitial fluids (ISF) were obtained by performing a low centrifugation (30 minutes, 420 × g, 10°C) on a 20 μm nylon filter for each total tumor sample (50–300 mg) as described previously with no modification (50). Tumor homogenates were harvested after mixing pieces of 50 to 300 mg tumor mass in PBS with Ultra-Turrax (IKA-T10, Sigma Aldrich). Trp and Kyn concentration were quantified by LC/MS-MS or HPLC-UV analysis after treatment with acetonitrile or trichloroacetic acid (TCA; ref. 40).

Fifteen-micron–thick frozen tumor sections were used for RNA extraction and RT-qPCR with the following primers and probes (annealing 1 minute at 60°C): β-Actin (gene Actb): forward (Fwd)-CTC-TGG-CTC-CTA-GCA-CCA-TGA-AG; reverse (Rev)-GCT-GGA-AGG-TGG-ACA-GTG-AG, probe-ATC-GGT-GGC-TCC-ATC-CTG-GC; Ido1: Fwd-GTA-CAT-CAC-CAT-GGC-GTA-TG, Rev-CGA-GGA-AGA-AGC-CCT-TGT-C, probe-CTG-CCC-CGC-AAT-ATT-GCT-GTT-CCC-TAC; Tdo2: Fwd-GTA-TCT-ATG-GAG-GAC-AAT-GAA-G, Rev-GAT-GAA-TAG-GTG-CTC-GTC-ATG, probe-CCT-CCT-TTG-CTG-GCT-CTG-TTT-ACA-CC; CD8β (gene Cd8b): Fwd-GTG-GTT-GAT-GTC-CTT-CCT-ACA-A, Rev-TCC-GCA-CAC-AGT-AAA-AGT-AGA-C, probe-AAT-GCC-AGC-AGA-AGC-AGG-ATG-CAG-ACT-A; IFNγ (gene Ifng): Fwd-TCA-AGT-GGC-ATA-GAT-GTG-GAA-GAA, Rev-TGG-CTC-TGC-AGG-ATT-TTC-ATG, probe-TCA-CCA-TCC-TTT-TGC-CAG-TTC-CTC-CAG; Samples were assayed in duplicates and mRNA copies were calculated with internal standards and reported for 1,000 cells, using β-actin for normalization (supposing a mean of 2,000 β-actin mRNA copies per cell).

MC38 tumors were frozen in Tissue-Tek O.C.T Compound and seven-micron–thick sections were obtained using a cryostat (CryoStar NX70, Thermo Fisher Scientific). Thawed sections were fixed for 5 minutes in 4% formaldehyde and antigen retrieval was performed by heating the sections in microwave 10 mmol/L citrate buffer pH 6 0.05% Tween20. The detailed immunostaining protocol was previously described and used without modification (26). The antibodies used were the following: rabbit anti-mouse CD8α (clone D4W2Z, #98941T, Cell Signaling Technology, dilution 1/1,000), rabbit anti-human CD3 (clone SP7, #16669, Abcam, dilution 1/200), rabbit anti-mouse CD4 (clone EPR19514, #183685, Abcam, dilution 1/200), and labeled polymer-HRP goat anti-rabbit (Dako) as the secondary antibody. After washing, sections were finally stained with a 1/25 dilution of Highdef DAB chromatogen (Enzo) in Highdef DAB substrate buffer (Enzo) for 20 minutes and counterstained with hematoxylin for 5 minutes. Spleen was used as a positive control. Negative controls included liver and omission of primary antibody on duplicate tumor sections. Stained slides were mounted and digitalized with a Pannoramic P250 Flash III slide scanner (3DHISTECH) and were analyzed with CaseViewer (3DHISTECH). CD8⁺, CD4⁺, and CD3⁺ cells were quantified with the Halo Image Analysis software. Invasive margins were defined by the software as 200 μm–thick edges around tumor nests.

Spleen cells and peripheral blood mononuclear cells (PBMC) from TDO-KO and wild-type mice were stained with anti-CD45-AlexaFluor 700, anti-CD3-PE, anti-CD4-FITC, and anti-CD8a-PerCP (all from BioLegend) and analyzed on a LSRFortessa (BD Biosciences). Data were analyzed using FlowJo.

Statistical analyses were performed using Prism 6 (GraphPad Software). Comparisons between two groups were performed using a two-tailed Mann-Whitney unpaired t test, while two-way ANOVA with Bonferroni post hoc test was used to compare the effects of different treatments on tumor growth. Survival curves were generated by Kaplan–Meier method and compared by using a log-rank test (Mantel–Cox).

The linear mixed model was obtained with the statistical software JMP, after log transformation of tumor volumes, where volumes smaller than 20 mm³ where fixed at log(20). Fixed effects included time (third degree polynomial function), treatment, and experiment with all their interactions. Random effects included a random intercept and slope per mouse. An average treatment difference in tumor growth over all experiments was tested by a joint hypothesis test for the interaction terms of time and treatment with effect coding of all categorical variables. Normalized tumor growth curves per experiment were constructed with the residual error term from a model as described previously (51) without adding the treatment term. This way, only the variation caused by treatment remains, allowing to graphically assess a common treatment effect over all experiments.

We developed a new selective TDOi, starting from a high-throughput screening with recombinant hTDO, followed by hit-to-lead and lead-optimization medicinal chemistry steps (47). The final compound PF06845102/EOS200809, whose structure is shown in Table 1, was synthesized as described in (47) and found to inhibit the enzymatic activity of human TDO with an IC₅₀ of 230 nmol/L and 285 nmol/L in a cellular assay using human glioma cells A172 (which constitutively express high levels of TDO) or hTDO2-transfected 293E cells, respectively (Table 1). It did not inhibit human IDO1 activity in transfected 293E cells (Table 1). Like previously described TDO inhibitors 680C91 and LM10, EOS200809 is based on a 6-fluoro indole scaffold (40, 43, 52). When tested side by side in a standardized cellular assay, PF06845102/EOS200809 was two times more potent than 680C91, more than 100-fold more potent than LM10, and more selective with regard to several important heme-containing cytochromes (Supplementary Table S1). The inhibitor was also tested on cells expressing the full-length human IDO2 or the short, reportedly active form of hIDO2 (44), but these cells showed no tryptophan-degrading activity (Table 1), indicating that hIDO2 is not active in physiologic conditions, in line with the reported high K_m of the enzyme (6.8 mmol/L; ref. 53). The compound also inhibited murine TDO activity (IC₅₀ 104 nmol/L) while murine IDO1 activity was unaffected (Table 1). The pharmacokinetic properties of PF06845102/EOS200809 in mice showed high exposures (AUC of 45,000 ng × h/mL at 100 mg/kg twice a day) and half-lives of 2.5 hours (i.v.) and 4 hours (orally; Table 1). These properties are predicted to achieve effective TDO inhibition in vivo with a twice-a-day oral administration. Preliminary toxicology studies performed in rats implied that the compound was well tolerated at doses of 30 and 100 mg/kg and did not reveal any safety signals.

To confirm activity in vivo, we treated naïve mice with increasing doses of TDOi by oral gavage. We collected liver and plasma and measured concentrations of Kyn and Trp. We observed a dose-dependent decrease of Kyn in liver (Fig. 1A), and a concomitant increase of Trp in plasma (Fig. 1B), confirming a sustained in vivo inhibition of Trp degradation by TDO in the liver. To confirm the activity of the TDOi in the tumor microenvironment, we established tumors by injecting mice with colon carcinoma CT26 cells transfected with an empty vector (CT26ev) or with a plasmid encoding mTdo2 (CT26mTDO). This cell line was confirmed to express TDO mRNA, protein, and enzymatic activity (Supplementary Fig. S1A–S1C), while neither cell line expressed mIdo1 or mIdo2 (Supplementary Fig. S1A). In addition, TDO protein was detected by western blot analysis in explants from CT26mTDO, but not from CT26ev tumors (Supplementary Fig. S1B). We treated mice with the TDOi (100 mg/kg) by oral gavage and collected the tumoral ISF 2 hours later. As expected, ISF from CT26mTDO tumors contained reduced Trp and increased Kyn as compared with control CT26 tumors (Fig. 1C). Upon treatment with PF06845102/EOS200809, Trp levels in ISF of CT26mTDO tumors went back to normal values indicating efficient inhibition of TDO (Fig. 1C). Of note, Kyn levels remained high in this setting, likely because of IDO1 expression induced in these tumors by the IFNγ produced by antitumor T cells (29) and because of increased availability of tryptophan as an IDO1 substrate. In the serum of CT26mTDO tumor–bearing mice treated with the inhibitor, Trp levels were increased while Kyn levels were unaffected (Supplementary Fig. S4), similar to what happened in nontumor-bearing mice (Fig. 1B).

To evaluate the antitumor effect of TDO inhibition in combination with the administration of checkpoint inhibitors, we used the CT26mTDO model described above. This was needed because we could not identify a mouse tumor line expressing TDO naturally (Supplementary Table S2), in contrast with human tumor cell lines, several of which constitutively express TDO at a high level (40). We inoculated BALB/c mice subcutaneously with CT26mTDO cells or CT26ev cells. Nine days later, we injected mice with anti-CTLA4 and/or treated them with PF06845102/EOS200809 by oral gavage (Fig. 2). Monotherapy with anti-CTLA4 or TDOi did not significantly affect progression of either tumor type. However, in mice with CT26mTDO tumors that received the combination of anti-CTLA4 and TDOi we observed improved antitumor efficacy, with 6 of 10 mice showing complete regression and no palpable tumor after 100 days, as compared with 2 of 10 mice treated with anti-CTLA4 single agent (Fig. 2A). The combined treatment was not effective in mice with control tumors CT26ev, which did not express TDO (Fig. 2B). In total, the effect of the combination was compared with anti-CTLA4 monotherapy in three independent experiments with CT26mTDO tumors. Although a high degree of individual variability was present in all experiments, we observed, in each experiment, more mice with regressing tumors in the combination group as compared with the monotherapy group. This difference became obvious when the tumor growth curves of all three experiments were combined and normalized to be plotted on the same scale (Fig. 2C). Overall, the improved antitumor effect of the combination group compared with anti-CTLA4 was highly significant (P = 0.0004). These results indicate a synergistic antitumoral effect of TDO inhibition and anti-CTLA4 therapy in TDO-expressing CT26 tumors. These data further document the immunosuppressive role of TDO in the tumor microenvironment and indicate that TDOi could be used in combination with checkpoint inhibitors in patients bearing TDO-expressing tumors.

So far, available evidence supporting an immunosuppressive role of TDO in cancer was obtained with murine cell lines transfected with Tdo2. Several human tumor cell lines do express TDO naturally, and a number of human tumor samples express TDO2 mRNA, indicating natural expression of TDO in the tumor microenvironment (11, 40). To better mimic the human situation, we searched for murine tumors that express TDO naturally. We collected a series of primary tumors and metastases obtained by implanting common tumor lines into syngeneic mice or by injecting tamoxifen into genetically engineered TiRP melanoma mice (refs. 54, 55; Supplementary Table S2). We collected tumors at different time points and measured Tdo2 expression by RT-qPCR. Surprisingly, none of these 15 tumor models expressed Tdo2 transcripts. These data suggest that, unlike human cancers, mouse tumors do not contain TDO-expressing cells.

We injected a series of mouse tumor lines into TDO-KO mice, to investigate the potential role of host TDO expression. Because of the lack of hepatic TDO expression, these mice have a defect in Trp homeostasis and have higher levels of systemic Trp (39). In line with this, we measured an increased Trp concentration in the serum of the TDO-KO mice (430 μmol/L) as compared with their WT or heterozygous (70 to 100 μmol/L) counterpart (Fig. 3A; Supplementary Fig. S3B). The link between TDO deficiency and high tryptophanemia is supported by the observation of similar metabolic changes in an independent strain of TDO-KO mice (56) and in a human with inherited TDO deficiency, who presented chronic hypertryptophanemia (57). In contrast, IDO1-KO mice had normal Trp levels in their serum (Fig. 3A). Kyn, the main product of both TDO and IDO1, was increased in TDO-KO mice but was totally absent in IDO1-KO mice (Fig. 3B). This observation indicated that circulating Kyn was produced by IDO1 but not by TDO. This is likely because the Kyn produced by TDO is further degraded in the liver, which expresses all the enzymes of the Kyn pathway, while the inflammatory and immune cells that express IDO1 in the periphery do not express these enzymes. The higher Kyn levels observed in TDO-KO mice would result from higher amounts of Trp being degraded by IDO1 in peripheral tissues. This is entirely related to the increased levels of circulating Trp, the IDO1 substrate, in these mice, as we found no increased Ido1 or Ido2 expression in tissues (liver, colon, and lymph nodes) from TDO-KO mice as compared with WT mice (Supplementary Fig. S5). In line with the role of IDO1 in producing systemic Kyn, we measured almost no Kyn in the serum of double KO mice (IDO1^−/− TDO^−/−), despite very high Trp levels (Fig. 3A and B). These results indicated that liver TDO activity, by degrading dietary Trp, maintained Trp homeostasis by reducing blood levels of Trp, while blood Kyn was produced exclusively by IDO1 in the periphery. The notion that circulating Kyn is produced by IDO1 and not by TDO also explains why we saw no decreased Kyn in the serum of mice treated with the TDOi (Fig. 1B; Supplementary Fig. S4).

Even though we observed no TDO expression in murine tumors, we were still interested to see whether the drastic changes in systemic tryptophan metabolism observed in TDO-KO mice would affect anti-tumor immune responses, given the high sensitivity of T-cell mediated responses to Trp catabolism (14, 17, 58–60). We used the colon adenocarcinoma line MC38, described as being immunogenic (61–63) and able to generate liver metastases (64–66). We counted the number of liver metastases three weeks after intrasplenic injection followed by hemisplenectomy. We observed less metastases in TDO-KO mice as compared with WT littermate mice (Supplementary Fig. S6). These results suggested that TDO-KO mice were more resistant to metastasis development, potentially because of a better antitumor immune response.

To further explore this point, we evaluated the primary growth of subcutaneous MC38 tumors. We observed similar growth rates in TDO-KO and WT littermate mice. However, upon treatment with immune checkpoint inhibitor anti-CTLA4 (Fig. 4A) or anti-PD1 (Fig. 4B), we observed a significant decrease of tumor growth in the TDO-KO mice but not in the WT littermates. To confirm the involvement of T cells in this difference, we collected tumors and analyzed the expression of CD8β by RT-qPCR, as a measure of CD8 T-cell infiltration. CD8β mRNA expression was significantly increased in tumors from TDO-KO mice treated with anti-PD1, while they were not in WT mice treated with anti-PD1 (Fig. 4C). We also evaluated T-cell infiltration by IHC in MC38 tumors treated with anti-PD1, and quantified the cells expressing CD8, CD4 or CD3 in the invasive margin and the central tumor region (Fig. 4D–F; Supplementary Fig. S7). We confirmed a higher CD8⁺ cell infiltration in tumors from TDO-KO as compared with WT mice, while there was no clear difference for CD4⁺ and CD3⁺ cells. Of note, there was no difference in the proportion of CD4⁺ and CD8⁺ T cells in the spleen and PBMC of TDO-KO and WT naïve nontumor-bearing mice, as measured by flow cytometry (Supplementary Fig. S8). Altogether, these results suggest that immune checkpoint inhibitors induce stronger antitumor CD8⁺ T-cell response in TDO-KO mice as compared with WT mice, resulting in better tumor control.

To determine whether this better efficacy of checkpoint inhibitors was related to the high systemic Trp and Kyn concentrations, we tried to normalize blood Trp and Kyn levels by changing the diet of TDO-KO mice. After 24 hours of feeding mice on a diet with a Trp content reduced from 0.18% to 0.06%, the blood of TDO-KO mice contained normal levels of Trp and Kyn (Supplementary Fig. S9), while their weight and behavior were similar to those of mice fed on standard diet. We then repeated the anti-PD1 experiment with MC38 tumors in TDO-KO mice fed on either standard or low-Trp diet. Interestingly, while anti-PD1 again showed antitumor efficacy in TDO-KO mice fed on standard diet, there was no antitumor efficacy in TDO-KO mice fed on the Trp-low diet (Fig. 5A and B). At the end of the experiment, we collected serum and tumors and analyzed Trp and Kyn concentrations by HPLC. The 6-fold higher Trp levels observed in the serum of TDO-KO mice were reduced when the mice were fed on Trp-low diet, down to levels equal or below those of WT mice fed on standard diet (Fig. 5C). Levels of Trp in the tumor homogenate and ISF reflected those in the blood, with high levels (600 μmol/L) in TDO-KO mice fed on standard diet and lower levels in TDO-KO fed on Trp-low diet, which were similar to those of WT mice (Fig. 5D and E). Interestingly, the Kyn concentrations in the tumor homogenate and in the ISF from the TDO-KO mice were not or only barely affected by the diet change. These results indicated that the high systemic Trp concentration observed in TDO-KO mice was required for the antitumor effect of the anti-PD1.

To explain the improved antitumor efficacy of anti-PD1 in mice with high systemic Trp, we hypothesized that MC38 tumors might express IDO1 and thereby suppress proliferation and/or activity of antitumor T lymphocytes in WT mice by degrading Trp locally. In TDO-KO mice, the very high Trp concentration at the tumor site might overload IDO1 activity and prevent it from reducing Trp down to the levels required for immunosuppression. To test this hypothesis, we measured the expression of Ido1, IFNγ, and Tdo2 by RT-qPCR in the tumors from TDO-KO and WT mice treated with anti-PD1 or IgG2a, collected between 6 and 15 days after MC38 challenge (Fig. 6A). The results confirmed expression of Ido1 in these tumors, which was slightly increased after anti-PD1 treatment. A similar pattern was observed for IFNγ, indicating an ongoing immune response (Fig. 6A). As expected, no Tdo2 expression was detected in any of the MC38 tumor samples (Supplementary Table S2). These tumors contained high Kyn concentrations, which were increased 4- to 7-fold in both the tumor homogenate and the ISF as compared with the levels measured in the serum (Fig. 5C–E). Because TDO is not expressed in these tumors, this high Kyn concentration was likely due to IDO1 activity.

To confirm the role of IDO1, we inhibited IDO1 activity in vivo using epacadostat, considering that if IDO1 was indeed responsible for the lack of effect of anti-PD1 in WT mice, then its inhibition should increase anti-PD1 efficacy in WT mice. This would not be the case in TDO-KO mice because IDO1 activity would already be functionally “inactivated” by the high Trp levels. We administered epacadostat twice a day by oral gavage in TDO-KO or WT mice bearing MC38 tumors and treated with anti-PD1. As shown in Fig. 6B and C, IDO1 inhibition significantly improved the antitumor effect of the anti-PD1 in WT mice, but had no effect in TDO-KO mice. At the end of the experiment, we collected the serum and the tumor homogenates to measure Trp and Kyn (Supplementary Fig. S10). Epacadostat administration did not increase Trp levels, which remained high in TDO-KO and low in WT mice. However, Kyn levels were reduced in samples from mice receiving epacadostat, both in sera and tumor homogenates, in line with IDO1 inhibition.

Altogether, these results suggested that the increased systemic levels of Trp resulting from TDO inactivation could overcome local immunosuppression induced by IDO1 expression at the tumor site. Therefore, TDOi might be used in patients with cancer to increase the efficacy of immunotherapy, even if there is absence of TDO expression at the tumor site.

Our results confirm the benefit of blocking TDO activity to increase the efficacy of immunotherapy. This is based on experiments using the novel TDOi PF06845102/EOS200809, which was optimized for in vivo application and further clinical development. When EOS200809 was tested side by side with other TDOi 680C91 and LM10, it was two times more potent than 680C91, and more than 100-fold more potent than LM10. EOS200809 was also more selective than 680C91, particularly in regard to several important heme-containing cytochromes, which were significantly inhibited by 680C91 but not by EOS200809. Compound 680C91 has poor stability in human and mouse microsomes, which was not the case with EOS200809, whose microsomal intrinsic clearance was divided by three on both human and mouse microsomes. The metabolic stability of EOS200809 was confirmed in mouse pharmacokinetic studies, which showed the compound as moderately cleared (30 mL/minute/kg) and with excellent oral bioavailability (80% at 100 mg/kg), giving exposures (AUC 45,000 ng × h/mL at 100 mg/kg, C_average 5,200 nmol/L) sufficient to achieve good TDO target coverage in vivo, considering that the IC₅₀ is between 200 and 300 nmol/L. These features are drastically improved as compared with 680C91, which is known to be poorly bioavailable through the oral route (40). They are in line with our observation that EOS200809 increased blood Trp levels more than 2-fold upon oral administration. A comprehensive review compares the structure and features of TDOi described in the patent literature, including EOS200809 (67). Although a few compounds show in vitro activity similar to EOS200809, no pharmacokinetic/pharmacodynamic data are available for these compounds. EOS200809 is therefore the first TDO-selective inhibitor reported with a favorable pharmacokinetic/pharmacodynamic profile after oral administration, allowing its preclinical evaluation in vivo in mouse tumor models and its potential subsequent clinical development.

EOS200809 synergized with the checkpoint inhibitor anti-CTLA4 in CT26 tumors engineered to express TDO. It was not possible to test this approach with mouse tumors expressing naturally TDO, because no such tumors could be identified among the 15 mouse tumors that we screened. However, a number of human tumors naturally express TDO (40, 41). Therefore, the results obtained with CT26-mTDO tumors are relevant to human TDO-expressing tumors and confirm previous observations indicating immune escape of TDO-expressing tumors (41), including our findings that rejection of TDO-transfected P815B tumors by mice immunized again tumor antigen P1A can be restored upon treatment with a TDOi (40). The expression of TDO in human tumors has been extensively characterized using novel antibodies in the accompanying report (42). Some tumors, such as hepatocarcinoma, constitutively express TDO at a high level in the tumor cells themselves and represent prime indications for TDO inhibitors (42). Other tumors, including glioblastomas, bladder, pancreatic, and colon carcinomas, express TDO in stromal cells belonging to the tumor vasculature (42). The testing of TDOi in clinical trials is therefore warranted, particularly in view of the failure of the first clinical trial testing IDO1 inhibitor epacadostat in combination with anti-PD1 (34). Although there are many potential reasons for this outcome, it is possible that IDO1 inhibitors need to be combined with TDOi to reveal their full potential (36, 37).

Studies with IDO1-KO mice have revealed additional roles of IDO1-mediated Trp catabolism in shaping protumorigenic inflammation that favors angiogenesis and recruitment of myeloid-derived suppressor cells (13, 68–70). IDO1 inhibitors also synergize with immunogenic chemotherapy in autochthonous mouse tumor models (68, 71–75). Although TDO expression in nonhepatic tissues is limited, it will be interesting to determine whether TDO plays a similar inflammatory modulator role and whether TDOi could also synergize with chemotherapy.

We observed that modulating TDO activity also regulated immune responses against tumors that do not express TDO, as checkpoint inhibitors displayed better efficacy in TDO-KO mice bearing MC38 tumors, as compared with wild-type mice. Because these tumors do not express TDO, this effect depended on the increased circulating Trp that was observed in TDO-KO mice and resulted from the missing liver TDO to degrade excess dietary Trp and regulate tryptophan homeostasis. The antitumor effect of anti-PD1 was lost when TDO-KO were fed on a Trp-low diet that normalized circulating Trp levels. We inferred that high Trp levels in the blood and tumors of TDO-KO mice overcame the capacity of locally expressed IDO1 to induce a Trp depletion that was sufficient to induce effective immunosuppression. In line with this explanation, IDO1 was expressed in these MC38 tumors (Fig. 6A), and the treatment of wild-type mice with the IDO1 inhibitor epacadostat restored a similar efficacy of anti-PD1 as that observed in TDO-KO mice (Fig. 6B and C). Altogether, these results suggest that TDOi might be clinically beneficial also for the treatment of human tumors expressing little or no TDO, provided that they can increase systemic Trp to levels similar to those present in TDO-KO mice. At least two published reports support this notion that inhibition of hepatic TDO may help antitumor immunity. The first is our previous study of TDO-expressing P815B tumors, in which TDO inhibitor LM10 promoted better rejection of control P815B tumors that did not express TDO (40). Another article showed a decrease in the number of lung metastases in mice inoculated intravenously with Lewis lung carcinoma (LLC) cells and subsequently treated with the TDOi 680C91 (76).

The respective role of IDO1 and TDO in Trp homeostasis is unclear. Here, mice that were knockout for both IDO1 and TDO were viable and fertile. We confirmed the previous reports describing high Trp and Kyn in the blood of TDO-KO mice (39). High Trp is explained by the loss of liver TDO activity required to degrade dietary Trp. High Kyn is more difficult to explain, and presumably results from Trp degradation by peripheral IDO1-expressing cells, namely dendritic cells in lymphoid organs, inflammatory cells, and some scattered cells in the lung, the female genital epithelia, and, in the mouse, the epididymis (8). High Trp likely induced high Kyn production by IDO1, even in the absence of Ido1 or Ido2 overexpression. Our study fully confirms this prediction, by showing that both IDO1-KO and double IDO1/TDO-KO mice had barely detectable Kyn in their blood (Fig. 3). We thus conclude that the circulating level of Trp is controlled by hepatic TDO, while circulating Kyn is produced by IDO1 in the periphery.

Our results shed some light on the respective role of Trp depletion and Kyn production in the immunosuppression triggered by Trp catabolism. We observed that the 7-fold higher Trp concentration present in the blood of TDO-KO mice was solely responsible for the increased antitumor effect of anti-PD1. In fact, these mice had elevated blood Kyn. When TDO-KO mice were fed on a Trp-low diet able to normalize blood Trp, the antitumor effect of anti-PD1 was lost. In this setting, blood Kyn was also reduced. Hence, the increased efficacy of anti-PD1 in TDO-KO mice (high Kyn) and its loss upon dietary Trp restriction (low Kyn) does not fit with a major immunosuppressive role of Kyn in this setting. On the other hand, a major role for Trp depletion would imply that, in WT mice, some extracellular regions of the tumor microenvironment exhibit low Trp concentrations and locally impact tumor-infiltrating T cells. This was not clearly observed in our measurements of Trp concentration in tumor homogenates or interstitial fluid, in which Trp concentration remained at a physiologic level (∼ 80 to 120 μmol/L). Two reasons may explain this result. First, it is likely that intracellular Trp was much higher than extracellular Trp, particularly in tumors expressing IDO1 or TDO which overexpress solute carrier transporters (SLC) involved in Trp uptake, such as SLC7A5 and SLC1A5 (77–80). This is why we analyzed ISF in addition to tumor homogenates. Even though the protocol we used to collect ISF was conceived to limit contamination by intracellular fluid (50), we cannot exclude some cell lysis during the centrifugation step, resulting in overestimation of Trp levels in the extracellular environment. To avoid this bias, it might be useful to use a better method to collect extracellular fluid, such as microdialysis (81). Second, it is possible that extracellular Trp was reduced only in certain areas of the tumor, as suggested by quantitative mass spectrometry imaging of the spatial distribution of Kyn and Trp in IDO1-expressing tumors (82). Although Trp concentrations appear to play an important role, our results did not rule out an immunosuppressive role of Kyn. Trp levels in the blood and in the tumor homogenate were barely affected by treatment of both TDO-KO and WT mice with the IDO1 inhibitor epacadostat, while Kyn levels were reduced in both mouse strains. This was consistent with inhibition of IDO1 activity, which increased the efficacy of anti-PD1 in WT mice. As discussed above, our experimental approach did not capture real concentrations of extracellular metabolites in subareas of the tumor microenvironment. However, these results suggest that both high Trp (TDO-KO mice) and low Kyn (epacadostat) may contribute to relieving immunosuppression orchestrated by the IDO1/TDO pathway.

Our results described a new TDOi, EOS200809, and demonstrated its potential to improve the efficacy of checkpoint inhibitors against TDO-expressing tumors. EOS200809 could be useful for the therapy of human TDO-expressing tumors, such as hepatocarcinomas. Our observations in TDO-KO mice also suggest that TDO blockade could favor immunotherapy of TDO-negative tumors, by increasing Trp to overcome IDO1-mediated immunosuppression.

S. Crosignani is Vice President, Preclinical R&D at and has ownership interest (including patents) in iTeos Therapeutics. G. Driessens is a full time employee at iTeos Therapeutics. B.J. Van den Eynde is a SAB member and has ownership interest (including patents) in iTeos Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: F. Schramme, G. Driessens, B.J. Van den Eynde

Development of methodology: F. Schramme, L. Pilotte, V. Stroobant, G. Driessens, B.J. Van den Eynde

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Frederix, D. Hoffmann, L. Pilotte, V. Stroobant, J. Preillon, G. Driessens

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Schramme, D. Hoffmann, L. Pilotte, J. Preillon, G. Driessens, B.J. Van den Eynde

Writing, review, and/or revision of the manuscript: F. Schramme, L. Pilotte, J. Preillon, G. Driessens, B.J. Van den Eynde

Study supervision: B.J. Van den Eynde

Other [design and drug metabolism and pharmacokinetic (DMPK) characterization of the TDO2 inhibitor]: S. Crosignani

The authors thank Dr. Hiroshi Funakoshi, Dr. Toshikazu Nakamura, and Prof. Michael Platten for providing TDO-KO mice; Christophe Lurquin, Pedro J. Gomez Pinilla, Guy Warnier, and the Laboratory Animal Facility team for the production and the genotyping of TDO-KO, IDO1-KO, and IDO1/TDO-KO mice; Sofie Denies for statistical advice; Dr. Pascal Brouillard, senior manager of the Genomics Platform of UCLouvain, for genotyping of the cell lines; and Auriane Sibille for editorial assistance. This work was supported by grants from the Ludwig Cancer Research (to L. Pilotte, V. Stroobant, and B. J. Van den Eynde), the Fonds Scientifique pour la Recherche – FNRS, de Duve Institute and UCLouvain (Belgium), UCLouvain (to F. Schramme), iTeos Therapeutics (to S. Crosignani, K. Frederix, J. Preillon, and G. Driessens), and FNRS-FRIA (grant no.: 1.E082.14; to D. Hoffmann).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.