Blocking Tryptophan Catabolism Reduces Triple-Negative Breast Cancer Invasive Capacity

Triple-negative breast cancer (TNBC) has a high rate of recurrence as a metastatic disease within the first few years of diagnosis (1–3). Anchorage-independent survival (also known as anoikis resistance) is known to facilitate metastasis (4–7). We previously demonstrated that upregulation of the kynurenine (KYN) pathway for tryptophan (TRP) catabolism through NF-κB occurs in anoikis-resistant/anchorage-independent TNBC cells. Inflammatory cytokines, IL1β, and TNFα that activate NF-κB also support anchorage independence (8). Expression of tryptophan 2,3-dioxygenase 2 (TDO2), a rate-limiting enzyme in the KYN pathway, is increased in anoikis-resistant TNBC (9–11) and ovarian cancer cells (12) grown under anchorage-independent conditions compared with two-dimensional culture, and furthermore, anchorage-independent survival is dependent on TDO2 activity (13).

TDO2 and the nonhomologous enzymes indoleamine 2,3-dioxygenase 1 and 2 (IDO1 and IDO2) evolved separately to catabolize TRP to formyl-KYN and KYN. IDO1 is expressed in most tissues, as is IDO2, albeit at much lower levels (14). TDO2 is normally expressed in the brain and liver (14). However, TDO2 and IDO1 can be expressed in various cancers. KYN binds to the aryl hydrocarbon receptor (AhR) and promotes cancer cell survival and motility in human glioblastoma cells (15) and breast and ovarian cancer cells (8, 13). In addition to binding AhR in cancer cells, secreted KYN can bind to AhR in immune cells, resulting in the functional attenuation of cytotoxic T cells and expansion of regulatory T cells, to facilitate an immunosuppressive microenvironment (14–16). In publicly available primary breast cancer databases, high tumor TDO2 correlates with poor overall survival (16). Although TDO2 mRNA and proteins are higher than those of IDO1 in TNBC, both enzymes can be expressed (8, 16), and we find this to be the case in ovarian cancer as well (12, 13).

Current small-molecule inhibitors of TDO2 or IDO1 are tryptophan mimetics (14). In clinical trials, IDO1 inhibitors such as epacadostat were combined with immune checkpoint blockade, such as anti-programmed death-ligand 1 (anti–PD-L1) or anti–programmed cell death protein 1 (anti-PD-1; refs. 17, 18); however, no significant improvement in tumor response was observed (19). Dual IDO1/TDO2 inhibitors are still under preclinical investigation and are currently in phase I clinical trials (20, 21). Although the mechanisms of resistance to selective TDO2 or IDO1 inhibitors are largely unexplored, resistance is hypothesized to arise from the compensatory potential of these enzymes that catabolize TRP (17). Thus, dual inhibition of TDO2 and IDO1 may be necessary to block TRP catabolism effectively.

Here, we examined TNBC response to inflammatory cytokines that increase endogenous TDO2 and measured flux through the TRP catabolism pathway by tracing stable isotope-labeled TRP in anchorage-independent versus standard two-dimensional culture, with or without inflammatory cytokines and with or without a new dual IDO1/TDO2 inhibitor. We find that knockdown (KD) of TDO2 can lead to a compensatory increase in IDO1, indicating the necessity for dual inhibition of both enzymes. We find that AT-0174, a new small-molecule dual TDO2/IDO1 inhibitor, significantly reduced TRP catabolites produced by TNBC cells and reduced anchorage-independent survival and invasion through a TDO2/KYN/AhR/zinc finger E-box binding homeobox 1 (ZEB1) axis. Collectively, our findings demonstrate that dual TDO2/IDO1 inhibition is a promising therapeutic avenue for the blockade of TRP catabolism to reduce TNBC progression to metastatic disease.

BT549 (No. HTB-122; RRID: CVCL_1092) was purchased from ATCC and cultured in RPMI 1640 medium supplemented with 10% FBS, 1% non-essential amino acids (NEAA), and 10 μg/mL insulin. MDA-MB-453 (No. HTB-131; RRID: CVCL_0418) was obtained from ATCC and cultured in DMEM with 10% FBS. SUM159PT (RRID: CVCL_5423) was obtained from the University of Colorado Cancer Center (UCCC) Cell Technology Shared Resource and grown in Ham’s F12 media with 5% FBS, 1% 4-(2-Hydroxyethyl)piperazine-1-ethane-sulfonic acid (HEPES), 10 µg/mL insulin, and 0.3% hydrocortisone. MDA-MB-231 (No. HTB-26; RRID: CVCL_0062) was purchased from ATCC and cultured in 5% FBS, 1% HEPES, 1% NEAA, 1% glutamine, 1% penicillin–streptomycin, and 10 µg/mL insulin. All cell lines were authenticated by short tandem repeat analysis and confirmed Mycoplasma negative in the UCCC Cell Technology Shared Resource before the study. Cells were last tested on May 2, 2024. For experiments in anchorage-independent culture (forced suspension condition), the culture plates were coated with 12 mg/mL of poly-2-hydroxyethyl methacrylate (poly-HEMA; Cat. # P3932, Sigma-Aldrich) and dried overnight before cell seeding.

A total of 5 × 10⁵ MDA-MB-453 cells were seeded in T25 flasks. After 24 hours, the cells were treated with either vehicle (water) or 10 ng/mL of IL1β + TNFα for 24 hours. A total of 3,000 cells were collected for single-cell RNA sequencing (scRNA-seq) with a read depth of 75,000 reads/cell by using the 10× Genomics platform (RRID: SCR_019326) and NovSeq 6000 system (Illumina) at the UCCC Genomics Shared Resource Facility (RRID: SCR_021984) according to 10× sample preparation guideline. For data analysis, Cell Ranger (v3.1.0; ref. 22) was used to process the FASTQ files to generate cell and gene count tables using unique molecule identifiers (UMI) with the GRCh38 genome (compiled by 10× Genomics, refdata-gex-GRCh38-2020-A). The Seurat (v4.0.4; ref. 23) pipeline was used for downstream quality control and analysis. Cell Ranger–filtered data were read into Seurat. Host genes were removed if identified in fewer than 10 cells. Cells were filtered out if they contained ≥50,000 UMI, ≤3,000 genes, ≥5,000 genes, or ≥15% of UMIs coming from mitochondria. The two samples were integrated for visualization only using Seurat with normalization by the sctransform with the difference between S and G₂–M scores regressed out. The top 30 principal components were used to perform Uniform Manifold Approximation and Projection. Prior to differential expression and plotting gene expression, data were normalized by dividing gene counts by total counts per cell and multiplied by 10,000 followed by natural-log transformation. Raw and processed data are deposited in Gene Expression Omnibus (GEO; GSE237918). Individual analysis of the control cells was previously published (24).

AT-0174 treatment set: 3 × 10⁵ BT549 cells were seeded in a six-well plate coated with 12 mg/mL poly-HEMA for forced suspension culture and followed by 10 µmol/L AT-0174 or vehicle (DMSO) treatment. BT549 shTDO2 set: 3 × 10⁵ BT549 TDO2-KD shRNA constructs and shScramble (shSCR) control cells were seeded in a six-well plate. MDA-MB-453 set: 5 × 10⁵ MDA-MB-453 cells were seeded in a six-well plate coated with/without 12 mg/mL poly-HEMA for forced suspension/regular cultures. All the conditions were conducted in biological triplicate. After 48 hours, the total RNAs were extracted using RNeasy Mini Kit (Cat. # 74104, QIAGEN). For library preparation and sequencing, RNA quality was verified using a High Sensitivity ScreenTape Assay on the TapeStation 2200 system (Agilent Technologies) and measured using a NanoDrop 1000 (Thermo Fisher Scientific). Library construction was performed using the Universal Plus mRNA Library Kit (NuGEN Technologies), and sequencing was performed on the NovaSeq 6000 instrument (Illumina) using paired-end sequencing (2 x 150 bp) by the UCCC Genomics Shared Resource (RRID: SCR_021984). For data analysis, RNA sequencing (RNA-seq) data were processed using the nf-core/rnaseq pipeline (version 3.12.0; ref. 25). Briefly, Illumina adapters were removed using Cutadapt (version 3.4; ref. 26) as part of the Trim Galore (0.6.7) package (https://doi.org/10.5281/zenodo.7598955). Reads were aligned using STAR (version 2.7.9a; ref. 27) to the human transcriptome (GRCh38, gene annotation from Ensembl release 104) and quantified using Salmon (version 1.10.1; ref. 28). Raw data with counts by gene were generated using tximport (https://doi.org/10.12688/f1000research.7563.1) on Salmon-quantified data. Normalized data were generated as counts per million (29). Differential expression was calculated between groups using the limma R package (30). Gene set enrichment analysis (GSEA) was performed using fold change and the fgsea R package with hallmark gene sets from the Molecular Signatures Database (31), which were downloaded using the msigdbr R package (https://CRAN.R-project.org/package=msigdbr). Normalized enrichment score was plotted based on GSEA and Benjamini–Hochberg adjusted P value. The subset of downregulation gene expression heatmap on the hallmark epithelial–mesenchymal transition (EMT) pathway was displayed. Raw and processed RNA-seq data were deposited in GEO (GSE253239).

The TDO2 ORF overexpression (OE) construct (Cat. # OHS6085-213574308, Horizon Discovery) and empty vector control PLX-304 (RRID: Addgene_25890) were obtained from the CCSB-Broad Lentiviral Expression Library (Horizon Discovery) at the University of Colorado Functional Genomics Facility. TDO2-KD shRNA constructs and shSCR control were obtained from MISSION(R) TRC Lentiviral shRNA Collection (MilliporeSigma), and sequences are listed in Supplementary Table S1. The lentiviral particles were generated by the UCCC Functional Genomics Shared Resource (RRID: SCR_021987). For virus transduction, SUM159PT or BT549 cells were seeded in a six-well plate and transduced with 500 µL to 1 mL of viral particles and then selected with 10 µg/mL blasticidin (Cat. # ant-bl, InvivoGen) or 10 μg/mL puromycin for 7 to 10 days. The TDO2 expression level after OE/KD was confirmed by immunoblotting or qPCR.

¹³C₁₁ L-TRP (CLM-4290-H-0.1) was purchased from Cambridge Isotope Laboratories, Inc. TDO2 inhibitor 680C91 (Cat. # SML-0287) and IDO1 inhibitor epacadostat (Cat. # S7910) were purchased from Sigma-Aldrich and Selleckchem, respectively. TDO2/IDO1 dual inhibitor AT-0174 was obtained from Antido Therapeutics, and the molecular structure has been published in Wu and colleagues (32). For isotope tracing, MDA-MB-453 cells were seeded in media supplemented with 16 mg/L labeled TRP following the indicated treatment. A total of 5 × 10⁵ MDA-MB-453 cells or 3 × 10⁵ BT549 cells were seeded in six-well plates and treated with 1 or 10 µmol/L AT-0174 under attached or suspension culture conditions. The cell pellets and corresponding conditioned media were harvested at 24 or 48 hours of treatment, cells counted, and samples snap-frozen. Metabolomic analyses were performed at the University of Colorado School of Medicine Mass Spectrometry Metabolomics Shared Resource Facility (RRID: SCR_021988). Medium samples were thawed on ice and diluted 1:25, and cell pellets were extracted at 2 × 10⁶ cells/mL in ice-cold lysis solution (5:3:2 MeOH:ACN:H₂O) as previously described (33, 34). Samples were vortexed vigorously at 4°C for 30 minutes and then centrifuged at 4°C, 10,000 g for 10 minutes to isolate solids (35). The supernatants were transferred to autosampler vials and analyzed on a Vanquish UHPLC system coupled to a Q Exactive mass spectrometer (Thermo Fisher Scientific). Extracts were resolved over a Kinetex C18 column (150 × 2.1 mm × 1.7 µm; Phenomenex) at a flow rate of 450 µL/minute using 5-minute gradients in positive and negative ion polarity modes as previously described (35, 36). Samples were randomized and introduced to the mass spectrometry (MS) via electrospray ionization with the MS scanning in full MS mode (2 µscans) over the range of 65 to 950 m/z (10 µL injection for cells; 20 µL injection for media). Technical replicates were injected every 6 to 12 samples to ensure instrument stability (36). Metabolites were annotated and integrated using MAVEN (Princeton University; RRID: SCR_022491) in conjunction with the Kyoto Encyclopedia of Genes and Genomes database (RRID: SCR_012773). Peak quality was determined using blanks, technical mixes, and ¹³C₁₁ natural abundance along with an in-house standard library (37). All studies were conducted in biological triplicate or above and unitless peak areas presented. The metabolomics data is publicly available (see data availability section).

BT549 shTDO2-82 or BT549 shAhR-85 cells were transfected with 1.5 µg pcDNA3.0-ZEB1 ORF using Lipofectamine 2000 (Cat. # 11668019, Invitrogen) according to the manufacturer’s protocol. After 24 hours, cells were harvested for use in invasion assays.

For the assay, 24-well format 8.0-μm culture inserts (Cat. #353097, Falcon) were precoated with 200 µg/mL Cultrex UltiMatrix Reduced Growth Factor Basement Membrane Extract (Cat. # BME001-05, Bio-Techne). For BT549 shTDO2 or MDA-MB-453 shTDO2, cells were cultured under attached conditions. For BT549 or SUM159PT, cells were cultured under attached or suspension conditions and pretreated with 10 µmol/L AT-0174, 10 µmol/L AhR inhibitor StemRegenin 1 (Cat. # S2858, Selleckchem), or their combination for 48 hours. The next day, 5 × 10⁴ cells were seeded on the top of the inserts, and the lower compartment was supplemented with the complete media. After 24 hours, the inserts were fixed with 10% neutral buffered formalin and stained with 0.1% crystal violet, and the upper cells were removed by cotton swabs. For quantification, crystal violet–stained invaded cells were dissolved in 10% acetic acid/water and the absorbance measured at 570 nm.

BT549 or MDA-MB-453 shTDO2/shSCR cells (3,000 cells/well) were seeded in 96-well ultralow attachment plates followed by AT-0174 treatment, and the cell growth was measured using CellTiter-Glo (Cat. # G7571, Promega), according to the manufacturer’s instructions at day 7.

Immunoblotting: Cells were lysed using T-PER Tissue Protein Extraction Reagent (Cat. # 78510, Thermo Fisher Scientific) containing 1× Halt Protease Inhibitor Cocktail (Cat. # 87786, Thermo Fisher Scientific), and the protein concentration was quantified using the Pierce BCA Protein Assay Kit (Cat. # 23225, Thermo Fisher Scientific). Generally, 10 to 40 µg of total protein lysate was mixed with SDS loading buffer and resolved in SDS-PAGE and then transferred on Immobilon-FL membranes (Cat. # IPFL00010, MilliporeSigma). The membranes were then blocked with 3% Bovine Serum Albumin (BSA)/Tris-buffered saline with 0.1% Tween 20 (TBST) for 1 hour and hybridized with primary antibodies overnight at 4°C. After washing with TBST, the membranes were incubated with the secondary antibodies for 1 hour, and the fluorescent signal was detected using a LI-COR Odyssey Infrared Imaging System (RRID: SCR_013715). Primary antibodies used in this study were TDO2 (1:1,000 dilution, Cat. # ab259359, Abcam), IDO1 (1:1,000 dilution, Cat. # 86630S, Cell Signaling Technology), total NF-κB p65 (1:1,000 dilution, Cat. # 8242S, Cell Signaling Technology), phospho–NF-κB p65 (1:1,000 dilution, Cat. # 3033S, Cell Signaling Technology), ZEB1 (1:1,000 dilution, Cat. # 70512, Cell Signaling Technology), and α-tubulin (1:5,000 dilution, Cat. # T5168, Sigma-Aldrich; RRID: AB_477579). The secondary antibodies included Goat anti-Rabbit IgG Alexa Flour 680 (1:5,000 dilution, Cat. #21109, Thermo Fisher Scientific; RRID: AB_2535758) or IRDye 680RD Goat anti-Mouse IgG (1:5,000 dilution, Cat. # 926-68070, LI-COR Biosciences; RRID: AB_10956588). The protein densitometry was quantified using ImageJ software (RRID: SCR_003070). For IHC, cell pellets were paraffin embedded and subjected to IHC for TDO2 (1:50 dilution, Cat. # MABN1537, MilliporeSigma) as previously described (8).

Total RNA was isolated using RNeasy Mini Kit (Cat. # 74104, QIAGEN). A measure of 1 µg of RNA was transcribed to cDNA using qScript cDNA SuperMix (Cat. # 95048, Quantabio) according to the manufacturer’s instructions. For qPCR, 10 µL of qPCR mixture contained 1.25 µL of cDNA (three times dilution from the reverse transcription cDNA product with nuclease-free water), 0.8 µL each of 10 µmol/L forward and reverse primers, 5 µL of PowerUp SYBR Green Master Mix (Cat. # 25742, Applied Biosystems/Thermo Fisher Scientific), and 2.15 µL nuclease-free water, and the SYBR amplification signal was detected using the 7500 Fast Real-Time PCR System (Applied Biosystems) and software (RRID: SCR_014596). GAPDH served as an internal control. Fold change was calculated based on 2^−ΔΔCT method using the formula ΔΔCt = [Ct (target) group − Ct (GAPDH) group] − [Ct (target) control − Ct (GAPDH) control], in which Ct represents threshold response and control signifies nontreatment or vehicle. The primers (5′–3′) were synthesized by Integrated DNA Technologies, and the sequences are listed in Supplementary Table S2.

The breast invasive carcinoma (The Cancer Genome Atlas Pan-Cancer Atlas; n = 1,084) dataset was analyzed using the cBioPortal platform (https://www.cbioportal.org/) to inquire the correlation of AhR and ZEB1 mRNA expression. The Gene expression-based Outcome for Breast cancer Online platform (https://co.bmc.lu.se/gobo/gobo.pl) was utilized to analyze the expression of AhR and ZEB1 in 51 breast cancer cell lines and the expression of TDO2, AhR, and ZEB1 in 1,881 clinical breast tumors of different subtypes. Expression of multiple genes is averaged with median-centered analysis. The public dataset for AhR chromatin immunoprecipitation analysis (GSE127649, doi: 10.17989/ENCSR412ZDC) was used with inquiry of ZEB1 and its upstream regulatory region.

All the experiments in this study were performed in biological triplicate at three independent times and analyzed using GraphPad Prism v.9.4.1 (RRID: SCR_002798). A P value < 0.05 was set for statistical significance. All the statistical methods are described in the corresponding figure legends and presented with mean ± SD or SEM.

scRNA-seq data for MDA-MB-453 cells treated with vehicle or IL1β + TNFα were deposited in GEO as GSE237918. Bulk RNA-seq data for BT549 cells in suspension culture treated with AT-0174 or TDO2 KD (shTDO2) was deposited in GEO as GSE253239. The metabolomics data presented in this study is available at the NIH Common Fund's National Metabolomics Data Repository (NMDR) website, the Metabolomics Workbench (https://www.metabolomicsworkbench.org), where it has been assigned the following Study IDs: ST003484, ST003485, ST003486, ST003487, ST003488, ST003489, ST003490, ST003491.

We previously reported increased NF-κB activation in TNBC cells under anchorage-independent culture conditions compared with 2D culture conditions (5). TNBC cells surviving under anchorage-independent conditions increase TRP catabolism by specifically upregulating TDO2. Inflammatory cytokines IL1β and TNFα that activate NF-κB also increase TDO2 (38). As we had previously observed that even in TNBC cell lines, not all cells upregulate TDO2 in suspension culture with or without NF-κB–activating cytokines, we sought to identify the characteristics of cells that express and upregulate this enzyme. We utilized the MDA-MB-453 cell line, which has high endogenous TDO2 in approximately 20% of cells. The cells were treated with 10 ng/mL of IL1β and TNFα for 24 hours, and IHC for TDO2 was performed followed by scRNA-seq. Genes coexpressed in cells with TDO2-high expression at baseline without cytokine treatment are listed in Supplementary Table S10. An increase in TDO2 protein with inflammatory cytokines was confirmed by IHC and immunoblotting (Fig. 1A and B). Interestingly, the cytokine treatment upregulated TDO2 mRNA and protein but did not upregulate IDO1 (Fig. 1B), and these results were confirmed by Uniform Manifold Approximation and Projection analysis (Fig. 1C). Differential gene expression between IL1β + TNFα–treated and vehicle-treated cells demonstrated significant upregulation of genes associated with inflammatory cytokines, including S100 calcium-binding protein A8/A9 and chemokine ligand 2 along with the TRP catabolism enzymes TDO2, kynureninase (KYNU), and kynurenine 3-monooxygenase (Fig. 1D). The complete list of differentially expressed genes (with and without the addition of inflammatory cytokines) is provided in Supplementary Table S3. GSEA pathway analysis of scRNA-seq data in response to cytokine treatment and bulk RNA-seq data of MDA-MB-453 cells surviving in forced suspension culture both showed positive enrichment of TNFα signaling via NF-κB activation and cellular amino acid catabolic process (Fig. 1E; Supplementary Fig. S1). scRNA-seq also demonstrated increased TDO2, KYNU, and kynurenine 3-monooxygenase expression upon exposure to inflammatory cytokines (Fig. 1F). The significant upregulation of TDO2 with NF-κB–activating cytokine treatment was also observed in the TNBC lines SUM159PT and BT549 (Fig. 1G).

To measure the flux of TRP through the KYN pathway over time, we performed tracing of labeled ¹³C₁₁-TRP via MS in MDA-MB-453 cells cultured for 24 and 48 hours in TRP-depleted media supplemented with ¹³C₁₁-TRP treated with or without IL1β + TNFα (Fig. 2A). Incorporation of heavy (¹³C₁₁-labeled) TRP versus light (unlabeled) isotopologues of intracellular TRP (to test uptake) and downstream metabolites formyl-KYN and KYN (to test TRP catabolism) demonstrated a significant and rapid decrease in the level of TRP (Fig. 2B), formyl-KYN (Fig. 2C), and KYN (Fig. 2D) in response to IL1β + TNFα treatment as compared with vehicle-treated cells. The catabolites 3-hydroxykynurenine and kynurenic acid significantly increased with treatment (Fig. 2E and F). The percentage (%) isotope ¹³C₁₁ labeled is shown in Supplementary Fig. S2. Tracing of labeled TRP showed that IL1β + TNFα increased flux through the TRP catabolism pathway, leading to a decrease in the levels of metabolites upstream of KYN and an accumulation of KYN and downstream metabolites. In addition, we performed ¹³C₁₁-TRP tracing in MDA-MB-453 cells cultured under forced suspension conditions as compared with 2D culture (attached). Notably, intracellular TRP, formyl-KYN, and KYN increased in cells surviving in suspension as compared with the attached condition (Supplementary Fig. S3). Thus, both inflammatory cytokine treatment and anchorage independence increased the flux of TRP through the KYN pathway in TNBC, as a result of the increased TDO2.

To examine targeting TDO2 in invasive TNBC, BT549 cells were transduced with lentivirus carrying shRNA specific for TDO2. Immunoblotting confirmed the reduction in TDO2 at the protein level (Fig. 3A). Formyl-KYN and KYN were significantly reduced in the cells generated with one of the shRNA constructs (TDO2-82) but not reduced in cells created with another construct (shTDO2-98; Fig. 3B). To investigate this, we interrogated IDO1 protein levels and discovered that IDO1 was 56-fold upregulated in the shTDO2-98 cells (Fig. 3A). The shTDO2-98 BT549 cells were thereafter denoted as having IDO1 compensation (or IDO1 comp). KD of TDO2 also caused a significant reduction in the AhR target gene cytochrome P450 1A1/1B1 (CYP1A1 and CYP1B1) in shTDO2-082 cells without an increase in IDO1, but these genes were not reduced in the shTDO2-98 cells (IDO1 comp) in which IDO1 showed a compensatory increase (Fig. 3C).

As we observed that IDO1 compensation is possible following TDO2 KD, we tested the dual TDO2/IDO1 inhibitor AT-0174 that acts as a competitive TRP mimetic that can bind both enzymes (32). MDA-MB-453 and BT549 TNBC cells cultured in suspension (in which TDO2 is increased) were treated with or without two doses of AT-0174 (1 and 10 µmol/L; Fig. 4A). Although no significant TRP changes were observed in MDA-MB-453 (Fig. 4B), TRP significantly accumulated in the cells and media in BT549 (Fig. 4E) upon treatment with AT-0174, indicating that TRP was not being catabolized. We measured both intracellular levels of TRP and its catabolites, as well as the depletion of TRP or secretion of catabolites into the media. We found that secreted formyl-KYN significantly decreased with AT-0174 treatment at 1 and 10 µmol/L in MDA-MB-453 (Fig. 4C), whereas 10 µmol/L AT-0174 decreased both intracellular and secreted formyl-KYN in BT549 cells (Fig. 4F). AT-0174 at 10 µmol/L treatment significantly decreased intracellular and secreted KYN in both cell lines (Fig. 4D and G). We also stably OE TDO2 in SUM159PT cells (Supplementary Fig. S4A) and found significantly higher secreted formyl-KYN and KYN compared with those secreted by control empty vector containing cells (Supplementary Fig. S4B). Furthermore, AT-0174 treatment significantly reduced KYN levels secreted into the media of TDO2 OE cells (Supplementary Fig. S4C). To test whether the inhibitors against TDO2/IDO1 reduced TRP catabolites in the BT549 shTDO2-98 cells (IDO1 comp), we utilized 680C91 (a selective TDO2 inhibitor), epacadostat (a selective IDO1 inhibitor), and the dual inhibitor of both IDO1 and TDO2 AT-0174 and conducted metabolomics to measure formyl-KYN and KYN. Epacadostat and AT-0174 significantly decreased secreted formyl-KYN and KYN in the media from shTDO2-98 (IDO1 comp) cells, but 680C91 (the TDO2 inhibitor) did not (Fig. 4H), suggesting the upregulation of TRP catabolism because IDO1 compensation can be reduced using the inhibitors that target IDO1. We also used MDA-MB-231, which expresses high endogenous IDO1, and found AT-0174 or epacadostat reduces KYN (Fig. 4I). To further test the effect of AT-0174 on flux through the TRP catabolism pathway, we performed ¹³C₁₁-TRP tracing in MDA-MB-453 under attached versus suspension culture conditions with or without 1 or 10 µmol/L AT-0174 treatment (Fig. 5A; Supplementary Fig. S5A). A measure of 10 µmol/L AT-0174 significantly increased intracellular TRP (Fig. 5B), but no significant changes in secreted TRP (Supplementary Fig. S5B). In contrast, AT-0174 decreased intracellular and secreted formyl-KYN and KYN levels (Fig. 5C and D; Supplementary Fig. S5C and S5D). No significance of kynurenic acid (Fig. 5G; Supplementary Fig. S5F) was observed in response to the treatment, suggesting that KYN was mainly processed to the direct downstream pathway. Notably, AT-0174 significantly decreased the production of 3-hydroxykynurenine (Fig. 5E; Supplementary Fig. S5E) and nicotinamide adenine dinucleotide (Fig. 5F), the terminal metabolite in the pathway, showing that dual inhibition of TDO2/IDO1 effectively reduced the flux of TRP through the KYN pathway.

To investigate whether dual inhibition of TDO2/IDO1 affects proliferation in anchorage-independent growth and invasion of TNBC, invasion assays were performed using transwell chambers coated with Cultrex, and invasive capacity was measured after 24 hours. TDO2 KD (Supplementary Fig. S6A) or AT-0174 significantly reduced anchorage-independent growth and invasion in BT549 and MDA-MB-453 cell lines (Fig. 6A; Supplementary Fig. S6B and S6C). In contrast, the BT549 shTDO2-98 (IDO1 comp) cells showed no significant difference in invasive capacity compared with shSCR control (Fig. 6A). Notably, treatment of BT549 shTDO2-98 (IDO1 comp) cells with AT-0174 or the AhR inhibitor StemRegenin significantly reduced invasion, and the combination of these two inhibitors further reduced invasion (Fig. 6B). AT-0174 treatment also decreased invasion in BT549 and SUM159PT TDO2 OE cells compared with vehicle control (Supplementary Fig. S7). To test invasive capability following anchorage-independent culture, when endogenous TDO2 increases, we seeded cells under forced suspension conditions for 24 hours and then treated with or without AT-0174, StemRegenin, or both. We found that AT-0174 or StemRegenin significantly reduced invasion (Fig. 6C; Supplementary Fig. S8A) and that treatment with both drugs further decreased invasion. Moreover, exogenous KYN significantly increased invasion, but treatment with AT-0174 did not affect invasion in the presence of KYN, as expected, because KYN binding to AhR is downstream of the action of AT-0174 on TDO2/IDO1 (Fig. 6D; Supplementary Fig. S8B). Conversely, the AhR inhibitor StemRegenin significantly reduced invasion in the context of KYN treatment as it inhibits AhR activity. To further investigate the effects of combinational treatment with the dual TDO2/IDO1 inhibitor and AhR inhibitor, we performed invasion assays with BT549 cells treated with three doses of AT-0174 and StemRegenin, alone or in combination, and found that the two drugs together resulted in synergistic effects with a synergy score of 11.11 (Fig. 6E). To elucidate if IDO1 also mediates invasion, we used MDA-MB-231 cells, which express high endogenous IDO1 (Fig. 6F), and treated them with AT-0174, epacadostat, StemRegenin, or the combination. We found that AT-0174 and epacadostat significantly also reduced invasion (Fig. 6G). Collectively, these data show that the inhibition of TDO2/IDO1 decreased invasion through KYN-mediated activation of AhR.

To elucidate the effect of TDO2 on TNBC invasive capacity, we performed RNA-seq in BT549 cells under anchorage-independent culture conditions with or without AT-0174 treatment and in BT549 cells with TDO2 shRNA or control shRNA. Comprehensive differential gene expression and GSEA results are listed in Supplementary Tables S4–S9. GSEA showed a decrease in TNFα-mediated NF-κB signal gene signatures with AT-0174 treatment, suggesting that the TDO2/IDO1 inhibitor may reduce NF-κB activation during anchorage independence. In the shTDO2-82 cells (with no IDO1 compensation), TGFβ and KRAS signatures are decreased compared with the control shRNA cells. Both shTDO2-82 and pharmacologic inhibition with AT-0174 positively enriched the mTOR signature, a known response to TRP accumulation (39), indicating that TRP catabolism was inhibited. Importantly, the hallmark of the EMT pathway, particularly mesenchymal genes, was decreased with AT-0174 and shTDO2-82 (Fig. 7A and B). Conversely, BT549 shTDO2-98 (IDO1 comp) cells did not show a decrease in the mesenchymal gene signature or mTOR signal (Fig. 7C), suggesting that IDO1 activity may regulate EMT. Although somewhat controversial, the mesenchymal phenotype is generally associated with cancer invasive capacity (40–42). To further examine epithelial and mesenchymal gene sets, we examined the mesenchyme-associated genes N-cadherin, vimentin, and ZEB1. These genes were significantly decreased in response to AT-0174 treatment (Fig. 7D) as in shTDO2-82 cells (Fig. 7E). However, in shTDO2-98 (IDO1 comp), N-cadherin and vimentin did not decrease (Fig. 7F and G). RNA-seq also showed that AT-0174 treatment or TDO2 KD decreased matrix metalloproteases MMP1 and MMP2 (Fig. 7D and E), and these proteases are known to be involved in invasion through extracellular matrix remodeling. Taken together, our RNA-seq data suggests that inhibition of TDO2 reduces mesenchymal gene sets, but mesenchymal genes were not reduced in cells that exhibit a compensatory increase in IDO1.

Next, we explored the potential mechanism that inhibition of TDO2 decreases invasion. We focused on ZEB1 as we observed the decrease in ZEB1 expression in response to TDO2 inhibition by RNA-seq. ZEB1 drives the EMT by transcriptional suppression of the genes encoding the cell–cell adhesion molecule E-cadherin (43, 44), and it promotes invasion in many carcinomas (45–47). Importantly, the ZEB1 promoter region contains predicted AhR-binding sites, and an AhR chromatin immunoprecipitation sequencing database (48) showed enrichment of AhR binding in the ZEB1 transcript (Supplementary Fig. S9). Indeed, in the shTDO2-82 BT549 cells, we found that both ZEB1 mRNA and protein were significantly decreased but were not decreased in the shTDO2-98 (IDO1 comp) cells (Fig. 8A; Supplementary Fig. S10A). IDO1 compensation restored ZEB1 levels, possibly because of the increase in KYN (shown in Fig. 3B) and KYN-mediated AhR transcription of ZEB1. To elucidate if ZEB1 is driven by AhR, we knocked down AhR and found decreased ZEB1 (Fig. 8B). Also, AT-0174 and the AhR inhibitor StemRegenin 1, each significantly reduced ZEB1 mRNA and protein and the combination caused a further decrease in protein expression (Fig. 8C; Supplementary Fig. S10B). To address whether KYN-mediated AhR transcription drives ZEB1 expression, we used exogenous KYN and found that treatment with KYN increased ZEB1 in BT549 cells (Fig. 8D; Supplementary Fig. S10C), suggesting that KYN drives AhR transcription/translation to increase ZEB1 expression. To test if the TDO2/AhR signaling axis drives TNBC invasion through ZEB1, we transfected the ZEB1 overexpression vector in the TDO2-KD (shTDO2) or AhR-KD (shAhR) cells. ZEB1 OE restored the invasive capability under both conditions (Fig. 8E and F). These data demonstrate that TDO2/KYN/AhR drives TNBC invasion through ZEB1.

The Cancer Genome Atlas dataset (Pan-Cancer Atlas) from patients with breast cancer showed a significant positive correlation between AhR and ZEB1 (Fig. 8G). Moreover, the Gene expression-based Outcome for Breast cancer Online dataset (49) showed that TNBC cell lines expressed a higher average of AhR and ZEB1 expression compared with that in hormone receptor–positive HER2⁺ breast cancer cell lines (Fig. 8H). Furthermore, the average of the expression of TDO2, AhR, and ZEB1 genes is higher in basal-like breast cancer (TNBC) than that in luminal A/B [estrogen receptor–positive (ER⁺)] breast cancer and when generally comparing ER⁻ versus ER⁺ tumors (Fig. 8I). The individual gene expression in the dataset above is shown in Supplementary Fig. S11A–S11C. Thus, the TDO2/AhR/ZEB1 axis may serve as a therapeutic target in TNBC. Although primary breast cancer tumors were examined in these clinical datasets for expression levels of TDO2, AhR, and ZEB1, these genes may be expressed at even higher levels in circulating tumor cells (anchorage-independent condition).

In our prior study (8), we found that anchorage-independent survival led to an increase in L-type amino acid transporter 1 and an increase in TDO2 (8). Here, we show that both anchorage independence and stimulation with inflammatory cytokines activate the NF-κB signaling cascade, leading to upregulation of TDO2 but not IDO1-mediated TRP catabolism. The IL6/STAT3 axis also increases TDO2, and consequent KYN production leads to the activation of AhR in response to inflammation in various types of cancer (11, 50). IHC and scRNA-seq solidified our previous observation that TNFα and IL1β increase TDO2 mRNA and protein levels in a heterogeneous fashion, even within TNBC cell lines (8, 13).

Here, we find that IDO1 can be induced in the context of TDO2 suppression. In lung cancer cells, IDO1 KD led to an increase in TDO2 (32), suggesting that both enzymes can increase in a compensatory manner. Although the exact feedback mechanism underlying the ability of these two enzymes to sense and compensate for each other remains to be determined, these results emphasize the need for a dual inhibitor that targets both enzymes. Here, we tested one such inhibitor, AT-0174 (32).

Structurally, TDO2 and IDO1 are distinct but share similarities in the catalytic sites that process TRP (14, 51). Prior inhibitors tested in the clinic specifically targeted IDO1, to determine if inhibiting the production of KYN would enhance the activity of immune checkpoint blockade therapies such as anti–PD-L1/PD-1 (18, 52). The rationale for the combination therapy was due to secreted IFNγ from cytotoxic T cells driving expression of PD-L1/PD-1 and TDO2/IDO1 in cancer cells in an immunosuppressive feedback loop (53) because KYN binding to AhR inhibits the viability and function of cytotoxic T cells but expands regulatory T cells (54). However, phase III clinical trials (NCT02752074/ECHO-301/KEYNOTE-252) combining epacadostat (that specifically targets IDO) with pembrolizumab (anti–PD-1) in advanced melanoma resulted in no significant improvement in patient outcomes (55, 56). It is possible that the tumors in these epacadostat trials were more dependent on TDO2 than IDO1 or that TDO2 activity/dependency increased in the face of targeting IDO1, leading to an inefficient reduction of KYN production. Our recent study in ovarian cancer demonstrated that TDO2/IDO1 targeting alone attenuates PD-L1 expression (13). The TDO2 inhibitor 680C91 has been used in preclinical studies (14, 57) but is not practical in vivo given its poor pharmacokinetic properties. Here, the dual TDO2/IDO1 inhibitor AT-0174 blocks TRP catabolism in TNBC cells dependent on TDO2, in cells that acquire compensatory IDO1, and in the MDA-MB-231 cells that express more IDO1.

We previously reported that TDO2 is targeted by microRNA-200c, which is lost in TNBC, permitting TDO2 expression (38). Significantly, the loss or silencing of miR-200c in TNBC facilitates EMT because members of the miRNA-200 family directly target and suppress ZEB1 and many other mesenchymal and neuronal (non-epithelial) genes (58, 59). In this study, we find that both dual TDO2/IDO1 inhibition and inhibition of AhR significantly reduce ZEB1. Importantly, TDO2 KD or the dual TDO2/IDO1 inhibitor reduced TNBC invasive capacity by reducing levels of ZEB1 transcript and protein. In cells in which IDO1 increases in a compensatory manner (upon TDO2 KD), invasive capacity was restored, indicating that both enzymes mediate TNBC invasion, likely through KYN-mediated activation of AhR, as AhR inhibition also decreased invasion and exogenous KYN increased invasion.

In summary, here, we demonstrate that both inflammatory cytokines and anchorage-independent survival upregulate TDO2 expression and activity in TNBC cells and that TDO2 inhibition can lead to IDO1 compensation. Thus, dual inhibition of both enzymes is necessary to block TRP conversion to KYN in TNBC effectively. The TDO2/IDO1 dual inhibitor AT-0174, which can be orally administered and is in a phase I clinical trial (ACTRN12623000956606), blocked TRP catabolism and inhibited KYN/AhR-mediated TNBC invasive capacity. We have also tested this inhibitor in ovarian cancer preclinical models and observed that AT-0174 enhanced the antitumor effects of chemotherapy (13). As the present study was limited to in depth in vitro analyses of TRP catabolism via the KYN pathway in TNBC, we are currently designing in vivo studies to test the effects of dual TDO2/IDO1 inhibition on both primary TNBC and metastasis and ways to analyze direct effects versus effects on the immune system in immune-competent models. In vivo modeling will also determine if dual TDO2/IDO1 inhibition influences primary tumor growth or whether its utility as an adjuvant therapy will reduce the risk of recurrence as a metastatic disease.

M.M. Williams reports grants from NIH/NCI during the conduct of the study. B.G. Bitler reports grants from NIH/NCI, Department of Defense, and American Cancer Society outside the submitted work. J.K. Richer reports grants from NIH/NCI and the Department of Defense outside of the submitted work. No disclosures were reported by the other authors.

L.-W. Kuo: Resources, data curation, software, formal analysis, validation, investigation, methodology, writing–original draft, editing, and project planning. L.S. Crump: Resources, data curation, software, validation, investigation, writing–review and editing. K. O’Neill: Conceptualization, data curation, software, formal analysis, investigation, writing–review and editing. M.M. Williams: Data curation, software, investigation, methodology, writing–review and editing. J.L. Christenson: Data curation, software, formal analysis, investigation, writing–review and editing. N.S. Spoelstra: Data curation, formal analysis, writing–review and editing. M.K. Roy: Data curation, software, formal analysis, methodology. A. Argabright: Data curation, formal analysis. J.A. Reisz: Supervision, writing–review and editing. A. D’Alessandro: Supervision, writing–review and editing. M.P. Boorgula: Data curation, software, formal analysis, writing–review and editing. A. Goodspeed: Data curation, software, supervision, writing–review and editing. M. Bickerdike: Resources, writing–review and editing. B.G. Bitler: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing. J.K. Richer: Conceptualization, funding acquisition, project administration, writing–review and editing.

The study was supported by the Department of Defense (W81XWH-20-1-0617-BC191031 to J.K. Richer), NIH/NCI (R37CA261987 and R01CA285446 to B.G. Bitler), and Training Program in Cancer Biology through the University of Colorado Cancer Center (L.-W. Kuo). Through the use of the Genomics Shared Resource Core Facility, Bioinformatics and Biostatistics Shared Resource, Mass Spectrometry Shared Resource (Metabolomics), and Pathology Shared Resource Research Histology Core Facility at the University of Colorado Cancer Center, this study was partly supported by the NIH P30CA046934. This work was supported by the Alpine HPC system, jointly funded by the University of Colorado Anschutz Medical Campus, UC Boulder, Colorado State University, and the National Science Foundation (award 2201538).