IDO1 Is a Therapeutic Target for Pancreatic Cancer–Associated Depression

Depression is extremely common in patients with pancreatic cancer. Estimates are between 21% and 78% (1–3), with an overall prevalence of 43% across studies (3). The development of depression is highly relevant to the patient with cancer. Epidemiologic studies suggest that patients with cancer and depression have poor treatment compliance and worse overall survival (1, 3, 4). As one may expect, depressed patients with cancer also report worse quality of life, as compared with non-depressed patients (2).

Results from randomized controlled trials testing selective serotonin reuptake inhibitors (SSRI) versus placebo for the treatment of cancer-associated depression have yielded mixed results. Although several randomized trials reveal a reduction in depressive symptoms or even the prevention of depression (5–7), a recent Cochrane review of 10 clinical trials concluded that there was no clear benefit to conventional antidepressant therapy across the spectrum of patients with cancer (8). Data specific to pancreatic cancer are lacking. The potential utility of effective therapies to manage or prevent depression is obvious to providers who treat patients with cancer. This may be particularly important for cancers associated with a high prevalence of depression, such as pancreatic cancer.

The etiology of cancer-associated depression is likely multifactorial and includes diverse factors such as patient awareness of a poor prognosis, physical sequalae of the cancer, chemotherapy-associated toxicities, or pain (9, 10). Intrinsic biologic factors may also be important. New-onset psychiatric symptoms have been repeatedly reported as an early warning sign in patients with pancreatic cancer, similar to unintentional weight loss or new-onset diabetes (3, 10–12). For instance, over 20% of patients diagnosed with pancreatic cancer at our institution exhibited evidence of prodromal psychiatric symptoms on a retrospective chart review (13). Other studies suggest that half of all patients with pancreatic cancer–associated depression present with mood changes or anxiety as a prodromal syndrome (11). The phenomenon was first described nearly a century ago by a neurologist in Philadelphia caring for four different patients who presented with neurocognitive decline (14), and they were eventually diagnosed with advanced pancreatic cancer in the months that followed. Multiple series subsequently reported a similar pattern (2, 3, 11, 13).

Immune mechanisms have been offered as a possible explanation of prodromal psychiatric symptoms. Investigators speculate that cytokine perturbations are associated with a phenotypic acute sickness response. This is supported clinically in patients with melanoma who have been treated with IFN-alpha, and subsequently develop depression (15). Recently, dysregulated tryptophan metabolism has gained traction as a potential driver of cancer-associated psychiatric symptoms (3, 16, 17). Tryptophan is an essential amino acid and substrate for serotonin and melatonin synthesis (18). These metabolites generally improve mood. As a result, medications that increase their concentrations in the brain (SSRIs or melatonin supplements) serve as treatments for depression and insomnia, respectively.

90%–95% of tryptophan is actually shunted through a catabolic pathway that generates kynurenine and other metabolites (Fig. 1A; ref. 19). The first and rate-limiting step in this pathway is the enzymatic synthesis of N-formyl-kynurenine by either tryptophan 2,3-dioxygenase 2 (TDO2), indoleamine 2,3-dioxygenase 1 (IDO1), or indoleamine 2,3-dioxygenase 2 (IDO2; ref. 18). TDO2 is expressed in the brain and liver, and drives hepatic tryptophan metabolism (20). IDO2 is primarily expressed in the liver, kidney, and certain immune cells (21). IDO1 is expressed in skin, lungs, and lymphoid tissue. Importantly, IDO1 is commonly upregulated in cancer (16, 22). IDO1 is believed to drive immune tolerance in tumors, perhaps, through the activation of the aryl hydrocarbon receptor by kynurenine metabolites (23).

Germane to this work, kynurenine metabolites counter the mood-related benefits of serotonin and melatonin, and actually promote depression and anxiety (16, 17, 24). In fact, multiple studies have established a role for these metabolites in the onset of depression in mice (24, 25). In addition, investigators have shown abnormal kynurenine metabolites in patients with cancer, as compared with control patients (26). Greater pancreatic tumor burden has also been associated with increased kynurenine levels (17). To our knowledge, no studies have examined this pathway and its relationship to depression in well-controlled preclinical cancer models.

Small-molecule inhibitors of IDO1 were developed to activate the antitumor immune response (16, 27). Unfortunately, a recent trial tested epacadostat (a selective IDO1 inhibitor; ref. 28) and pembrolizumab [a programmed cell death protein (PD)-1 inhibitor] in patients with advanced melanoma, and failed to demonstrate any additive survival benefit of epacadostat to the PD-1 inhibitor (29). As a result, enthusiasm has diminished for this class of drugs as a cancer therapy. Herein, we propose an alternative use for IDO1 inhibitors not to directly treat cancer, but instead to improve pancreatic cancer–associated depression through disruption of tryptophan–kynurenine metabolism.

Human pancreatic cancer cell lines (MIA PaCa-2 (RRID:CVCL_0428), PANC-1 (RRID:CVCL_0480), BxPC-3 (RRID:CVCL_0186) were obtained from the ATCC and passaged at least two times before experimentation. Murine pancreatic cancer cells (KPC K8484: KrasG12D; Trp53R172H/⁺; Pdx1-Cre) were provided by the laboratory of Dr. Darren Carpizo (30, 31). Sanger sequencing was performed by the genomics core at Case Western Reserve University. Mycoplasma screening was performed monthly using a MycoAlert detection kit (ATCC, 30–1012K; all cell lines were negative for Mycoplasma as of 12/2021, before original article submission). Cell lines were maintained at 37°C and 5% CO₂. All cell lines were grown in McCoy's 5A media (Thermo Fisher Scientific, 2192439) supplemented with 10% FBS, 1% penicillin/streptomycin, and prophylactic doses of Plasmocin (Life Technologies, MPP-01–03). McCoy's media contain ascorbic acid (2.84 μmol/L), a cofactor required for serotonin production, and have been previously used for in vitro assays of tryptophan metabolism (32). IDO1 expression was induced in vitro with IFNγ (18, 33). For assays involving IFNγ treatments (human: Invitrogen, PHC4033; murine: PeproTech, 315–05), a dose of 12.5 ng/mL was used. Murine TNFα (PeproTech, 315–01A) was also used for select in vitro studies at a dose of 5 ng/mL (34).

RNA was extracted from cultured cells after 24 hours of pharmacologic treatment with a given drug using a PureLink RNA isolation kit (Life Technologies, 12183025) and treated with DNase (Life Technologies, AM2222). cDNA was synthesized using 1 μg of total RNA (Thermo Fisher Scientific, 01127021). Primers for IDO1 (human: Hs00984148_m1; murine: Mm00492586_m1) were used and normalized to 18S (human: Hh99999901_s1; murine: Mm03928990_g1; all Thermo Fisher Scientific). All PCR reactions were performed in triplicate using iTaq Universal Probes Supermix (Bio-Rad, 1725131). RT-qPCR acquisition was performed using a Bio-Rad CFX96.

For in vitro Western blots, total protein was extracted using RIPA buffer (Thermo Fisher Scientific, 89901) from cells cultured under study conditions for 48 hours. Western blots performed on resected mouse tumors were homogenized in cell lysis buffer (Cell Signaling Technology, 9803S). Cell lysates were supplemented with a protease–phosphatase inhibitor (Thermo Fisher Scientific, WC320075). Protein was quantified using the BCA Protein Assay (Thermo Fisher Scientific, 23225). Lysates were separated by electrophoresis using 4%–12% Bis-Tris Plus gels (Thermo Fisher Scientific, NM04120) and protein bands were transferred to PVDF membranes (Thermo Fisher Scientific, 2PM22091–01). Membranes were blocked for 1 hour in a 5% milk solution at room temperature and then probed with antibodies against IDO1 (murine: Cell Signaling Technology Cat #68572, RRID:AB_2799750; human: Cell Signaling Technology Cat #86630, RRID:AB_2636818) overnight at 4°C. Alpha-tubulin (Proteintech Cat# 11224–1-AP, RRID:AB_2210206), GAPDH (Santa Cruz Biotechnology Cat# sc-47724, RRID:AB_627678), and ß-actin (Santa Cruz Biotechnology Cat# sc-47778, RRID:AB_626632) were used as loading controls. Blots were probed with secondary antibodies designed for an Odyssey imaging system (Li-Cor, D10831–15 or D10901–15).

Cell viability was assessed by DNA quantitation (PicoGreen dsDNA assay, Life Technologies, P7589). Gemcitabine chemotherapy was used as a positive control when assessing for cell death. As an independent assay of cell viability, a clonogenic assay was performed. Approximately 1,000 cells were seeded in a 6-well plate. After 24 hours, 100 nmol/L epacadostat (or vehicle) and IFNγ (to induce IDO1 expression in cell culture) were added to the culture media. Upon completion of the experiments, colonies were fixed in a crystal violet solution (BD Biosciences, 0296874) and photographs were obtained to visually depict cell viability. All cell viability assays were performed after four to five days of treatment.

All experiments involving mice were approved by the Case Western Reserve University Institutional Animal Care Regulations and Use Committee (IACUC, protocol 2018–0063). Mice were maintained under pathogen-free conditions in a dedicated animal facility. Eight-to-10-week-old, female, C57BL/6J mice (RRID:IMSR_JAX:000664) were purchased from The Jackson Laboratory. Eight-week-old, female, NOD scid gamma (NSG; RRID:IMSR_JAX:005557) mice were also purchased from The Jackson Laboratory. Mice were anesthetized using inhaled isoflurane gas and prepped and draped in a sterile fashion. After ensuring an appropriate depth of anesthesia, a 0.5-cm left subcostal incision was made and the peritoneal cavity was entered. The spleen and tail of the pancreas were externalized and 10,000 KPC cells suspended in 30 μL of a 40:60 mixture of Matrigel (Corning, 354234) and PBS were carefully injected into the pancreas. For non–tumor-bearing controls, an acellular mixture of Matrigel and PBS was similarly injected into the tail of the pancreas to control for the physical stress of the operation (i.e., a sham surgery). The pancreatic tail was left undisturbed for 60 seconds to limit leakage from manipulation. The pancreas was then returned to the peritoneal cavity, the incision was closed, and the mouse was allowed to recover under close observation. On postoperative day 7, the presence of viable pancreatic tumors was confirmed by bioluminescence imaging after intraperitoneal injection of 100 μL D-Luciferin (Perkin-Elmer, 2898979, 50 mg/mL). Mice with confirmed tumors were randomized to the indicated treatment groups.

Approximately 100,000 KPC cells were suspended in 150 μL of a 40:60 mixture of Matrigel and PBS and injected into the left flank of C57BL/6J mice. Once tumors reached approximately 150 mm³, mice were randomized to the indicated treatment groups. Tumor volumes were measured twice per week using calipers. Tumor volumes were measured using the formula: (length × width²)/2. Experiments were terminated once tumors became ulcerated (approximately 500–600 mm³), in the interest of animal welfare.

Behavioral experiments were designed in consultation with members of the Mouse Behavioral Phenotyping Core at Case Western Reserve University. C57BL/6J mice underwent behavioral testing at baseline (i.e., before tumor implantation) and between 5–6 weeks after tumor implantation (referred to as the “late” time point throughout the results section). The late time point was a period in time where mice had palpable tumors, but did not yet exhibit systemic signs of illness or toxicity related to tumor progression. NSG mice underwent testing after 3 weeks of tumor growth, due to rapid disease progression. For direct comparison purposes, C57BL/6J also underwent testing after 3 weeks of tumor growth.

For indicated experiments, mice were administered an IDO1 inhibitor (epacadostat, MedChemExpress, HY-15689, 50 mg/kg), SSRI (escitalopram, Selleck Chemicals, S4064, 10 mg/kg), or vehicle (10% PEG-400, 4% Tween-80, and 86% saline). Epacadostat is known to have near complete selectivity to IDO1, as compared with TDO2 or IDO2 (28). Epacadostat also has activity against both human and murine IDO1, as the IDO1 gene is conserved across species (35). Drugs were administered once daily via oral gavage at the indicated doses, as described previously (28, 36–38).

Mice were placed into the open field arena for a total of 15 minutes. Video recordings with automated analyses were performed. The total distance traveled and percentage immobile were used as a surrogate marker for overall physical mobility (39). These data were interpreted as a general gauge of mouse well-being. For instance, comparable mobility in tumor-bearing and non–tumor-bearing mice was interpreted to reflect a modest tumor burden imposing only a minimal amount of systemic toxicity. The total times within pre-specified standardized zones (inner, middle, and outer) were used as a measure of anxiety-like behavior. Greater time in the outer zone (i.e., close to the arena walls), also known as thigmotaxis, is a validated marker of anxiety-like behavior (39). Investigators were not blinded for this analysis, because measurements were automatically calculated by the video recording software.

Mice were gently placed into an inescapable 2,000 mL beaker filled with room temperature water. Testing lasted for a total of 6 minutes. The percentage immobile was recorded for each mouse, with greater immobility acting as a marker of depressive-like behavior (40). Following testing, mice were rapidly dried and placed into a warm cage. Mice were closely monitored during their recovery. Investigators were blinded during this analysis.

Mice were carefully affixed by their tail using laboratory tape to a horizontal shelf and dangled approximately 18 inches above the tabletop. Testing lasted for a total of 6 minutes. The percentage immobile was recorded for each mouse, with greater immobility acting as a marker of depressive-like behavior (40). Mice treated with vehicle or epacadostat were video recorded at the late time point. The primary investigator (J.J. Hue) was not blinded during this measurement. At a later time, a separate investigator (E.S. Katayama) reviewed all recordings and independently measured the time immobile.

Mice were placed at the center of the maze where they faced a closed arm at the beginning of the test. Testing lasted a total of 5 minutes. Video recordings with automated analyses were performed. The percentage of time spent in open and enclosed arms was recorded for each mouse. Greater time in the enclosed arm was used as a marker of anxiety-like behavior (41). In addition, the number of line crossings was counted. Investigators were not blinded for this analysis, as measurements were automatically obtained by the video recording software.

At indicated time points, blood was harvested from mice via submandibular venipuncture and serum was extracted. Kynurenine (Biovision, E4629) and serotonin (Biovision, E4294) analyses were performed according to the manufacturer's instructions.

For murine metabolomics analyses, whole blood was extracted via retro-orbital venipuncture under anesthesia. Mice were subsequently euthanized. Blood was extracted after 4 weeks of tumor growth for comparisons between tumor-bearing and non–tumor-bearing C57BL/6J mice. For tumor-bearing mice undergoing treatment, blood was collected after completion of behavioral studies (i.e., 6–7 weeks after tumor implantation). Metabolic profiles in tumor-bearing mice treated with vehicle, escitalopram, and epacadostat were then compared.

For human analyses, preoperative blood samples were obtained from patients undergoing pancreatectomy at our institution after obtaining informed consent (Institutional Review Board protocol 2018–0966). Patients who received neoadjuvant therapy were excluded. Metabolic profiles of patients with pancreatic ductal adenocarcinoma (PDAC) were compared with patients with various benign diagnoses (chronic pancreatitis, benign neuroendocrine tumor, mucinous cystic neoplasm, pseudopapillary neoplasm, among others; Supplementary Fig. S1E).

For both human and murine samples, we extracted polar metabolites from 100 to 150 μL of serum using 80% high-performance liquid chromatography (HPLC)-grade methanol. Samples were shipped on dry ice and assays were performed by the Mass Spectrometry Core at Beth Israel Deaconess Medical Center. Metabolites were quantified by LC-MS analysis using 5500 QTRAP hybrid quadrupole mass spectrometer (AB/SCIEX) coupled to a Prominence HPLC system (Shimadzu) using amide HILIC chromatography (Waters) at pH 9.2 (42). Specifically, metabolites in the tryptophan pathway (tryptophan, kynurenine, kynurenic acid, anthranilic acid, xanthurenic acid, and quinolinic acid) were quantified. Abundances for each metabolite were normalized to levels in mice receiving vehicle or patients with PDAC. Because of variability in tryptophan concentrations in human samples, all metabolomics data are presented as ratios (e.g., kynurenine/tryptophan, kynurenic acid/tryptophan, etc.), as has been previously used (43, 44). Raw abundances (i.e., not normalized to tryptophan) are also provided in Supplementary Figures.

KPC orthotopic tumors and normal pancreatic tissue were sectioned and preserved in 10% formalin. Samples were embedded in paraffin and probed for IDO1 (1:200; Santa Cruz Biotechnology, mIDO-48: sc-53978) and tryptophan hydroxylase (Tph1; 1:200; Millipore Sigma, AB15570–1) by the Tissue Resources Core Facility of the Case Comprehensive Cancer Center. Photographs of three to five representative tumors from each treatment group were obtained using a Zeiss Axio Scope A1 microscope at both ×2.5 and ×20 magnifications.

For qPCR and cell viability assays, data are presented as mean ± standard error of means. Quantification of IDO1 in Western blot analyses were normalized to the specified loading control. All in vitro experiments were performed in duplicate at a minimum for each cell line. ELISA and LC-MS data are depicted using box plots, with all data points displayed. The horizontal line indicates the median value, and the box includes the interquartile range (25%–75%). For behavioral analyses, the Wilcoxon rank-sum test was used for statistical comparisons between treatment groups due to the non-normal distribution of data points. Box plots were also used to graphically depict behavioral data. Median survival was assessed using the Kaplan–Meier method and compared using the log-rank test. Statistical comparisons were made using StataSE v16.1 (Statacorp LLC, RRID:SCR_012763) and graphical displays of in vitro and behavioral studies were made using GraphPad Prism 9 (RRID:SCR_002798). A P value was considered statistically significant when <0.05 (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

The data generated in this study are available upon request from the corresponding author.

The Cancer Genome Atlas database demonstrated increased IDO1 mRNA expression in human pancreatic cancer specimens, relative to normal pancreatic tissue (Fig. 1B), consistent with past reports (16). Comparisons of other tryptophan-related enzymes revealed an elevation of TDO2 in tumors, but no increase was observed in Tph1 or IDO2 (Supplementary Fig. S1A–S1C). In a cohort of patients undergoing pancreatic resection at a single institution (University Hospitals), serum metabolites were compared in patients with pancreatic cancer and benign pancreatic disease (Fig. 1C; Supplementary Fig. S1D and S1E). Patients with pancreatic cancer had significantly higher levels of kynurenine, as well as downstream metabolites (kynurenic acid, anthranilic acid, xanthurenic acid, and quinolinic acid) as compared with patients with benign pancreatic pathologies.

IDO1 was also assayed at mRNA and protein levels across murine and human cell lines. The enzyme was minimally detected or undetectable at baseline (Supplementary Fig. S2A and S2B). As previously shown (18, 33), IDO1 expression was strongly induced by treatment with an inflammatory cytokine present in the tumor microenvironment (IFNγ). The induction of IDO1 protein levels was further increased with the addition of TNFα (Supplementary Fig. S2C), validating past reports (34). Together, these data suggest that conditions in the tumor microenvironment could stimulate tryptophan metabolism toward the kynurenine pathway via IDO1 to generate neurotoxic metabolites and promote depressive-like behavior. The hypothesis provided a rationale to consider a direct link between the kynurenine pathway and neuro-cognitive behavioral symptoms in mouse models of pancreatic cancer.

Behavioral studies were performed on immunocompetent C57BL/6J mice with orthotopically transplanted syngeneic pancreatic tumors, and compared with mice who underwent sham surgery without tumor implantation. The open field test was initially used to profile the general health of mice at the tested time point, and guard against the possibility that behavioral changes were attributable to systemic illness from advanced cancer. Both tumor-bearing and non–tumor-bearing mice experienced a decrease in the total distance traveled during the open field test over time, reflecting a similar degree of habituation (Fig. 2A). Perhaps unexpectedly, tumor-bearing mice were actually more mobile to a small extent at the late time point. This finding demonstrated that tumor burden did not adversely affect overall mobility at the time of behavioral analyses and provides indirect evidence that the extent of disease had yet to overwhelm overall fitness in the animals. Similarly, non–tumor-bearing mice actually had a higher percentage immobile at the later time point in the open field test (as compared with baseline), unlike tumor-bearing mice (Fig. 2B). The two groups had comparable increases in time spent in the outer zone of the open field arena at the late time point (as compared with baseline). There was a corresponding decrease in the amount of time spent in the middle and inner time zones in both groups (Fig. 2C–E). In the aggregate, these findings suggested that tumor burden at the tested timepoints had not appreciably affected the general wellness of the mice.

Non–tumor-bearing C57BL/6J mice performed similarly at baseline and the late time point in the forced swim test (median percentage immobile: 8.8% vs. 8.2%, P = 0.419), whereas tumor-bearing mice exhibited a significant increase in the time immobile at the late time point (Fig. 2F, 12.0% vs. 16.0%, P = 0.001; P < 0.001 as compared with non–tumor-bearing mice at the late time point). Tumor-bearing mice were also more immobile during the tail suspension test, as compared with the non–tumor-bearing mice at the late time point (Fig. 2G, 15.4% vs. 24.1%, P = 0.003). Notably, the time immobile decreased at the late time point (relative to baseline) in the non–tumor-bearing mice, but not in the tumor-bearing mice. These data collectively point to an increase in depressive-like behavior in immunocompetent tumor-bearing mice.

These experiments were repeated in immunocompromised NSG mice. After just 3 weeks of tumor growth, tumor-bearing NSG mice exhibited a substantial decrease in total distance traveled during the open field test, which was not seen with C57BL/6J mice (Fig. 3A). This was likely due to rapid tumor growth in NSG mice as demonstrated by a significant increase in body weight (Fig. 3B) and evidence of significant disease burden on bioluminescence imaging (Fig. 3E). Despite this reduction in physical mobility, tumor-bearing NSG mice performed similarly during the forced swim (Fig. 3C) and tail suspension (Fig. 3D) tests when compared with non–tumor-bearing NSG mice. Conversely, tumor-bearing C57BL/6J mice exhibited depressive-like behavior relative to their controls on both behavioral tests, despite equivalent physical mobility, similar to the aforementioned results from the later timepoint (Fig. 2). These data demonstrate the importance of the tumor microenvironment (e.g., IFNγ) as a driver of IDO1-dependent metabolism and depressive symptoms.

IDO1 was absent in normal pancreatic tissue by IHC, but expression was pronounced in syngeneic murine tumors harvested from immunocompetent mice (Supplementary Fig. S3A and S3B). In contrast, when the same murine KPC cells were orthotopically injected into immunocompromised mice, IDO1 expression was virtually undetectable (Supplementary Fig. S3C). Whole-tumor lysates were analyzed for IDO1 protein expression. This revealed increased IDO1 levels in tumors, as compared with pancreatic tissue from the non–tumor-bearing mice (Fig. 2J). These findings further underscore the presence of IDO1 expression in mouse pancreatic tumors and highlight a requirement of activated immune cells in the tumor microenvironment for IDO1 induction. These data are consistent with the in vitro requirements for IFNγ stimulation of IDO1 expression described above (Supplementary Fig. S2). Serum kynurenine measurements reinforced these observations. Levels were similar in non–tumor-bearing mice over the study period, but increased substantially in tumor-bearing mice over the same interval (Fig. 2H). The increase in serum kynurenine in tumor-bearing mice was confirmed via independent measurements by LC-MS (Fig. 2I). In contrast, serum tryptophan levels were not different based on tumor-bearing status (Fig. 2I).

We also assessed for changes in the non-kynurenine arm of tryptophan metabolism. Somewhat surprisingly, serum serotonin levels actually increased over the study period in tumor-bearing mice (Supplementary Fig. S4A). Tph1 protein levels were generally similar in normal pancreatic tissue and KPC orthotopic tumors (Supplementary Fig. S4B and S4C), similar to human data (Supplementary Fig. S1A). These findings likely discount serotonin-related changes as a relevant driver of depressive-like behavior in the current model.

Our data in aggregate point to a possible therapeutic role for an IDO1 inhibitor (e.g., epacadostat) as a treatment for pancreatic cancer–associated depression, because this drug should modulate levels of kynurenine and downstream metabolites. For therapeutic experiments in cell culture and in vivo, a commonly used SSRI, escitalopram, was also tested as a comparator. Neither of these drugs had any appreciable effect on pancreatic cancer cell viability in vitro, even at high doses (PicoGreen assay, Supplementary Fig. S5A and S5D). Inconsequential effects on cell viability with epacadostat (100 nmol/L) were confirmed in a clonogenic colony formation assay (Supplementary Fig. S5E and S5H). This same dose of epacadostat, however, did have an appreciable biologic effect on the tryptophan–kynurenine axis, as evidenced by a compensatory increase in IDO1 protein expression (i.e., negative feedback; Supplementary Fig. S2D).

Similar to in vitro studies, neither epacadostat, nor escitalopram, improved survival in mice with orthotopic pancreatic tumors (P = 0.892, Supplementary Fig. S6A). All treatments were well tolerated (Supplementary Fig. S6B). The lack of therapeutic efficacy was validated using a subcutaneous allograft model, where neither epacadostat, nor escitalopram, slowed tumor growth relative to vehicle (Supplementary Fig. S6C).

In the open field test, as before, there was a decrease in the total distance traveled when comparing the baseline with late time points for tumor-bearing mice treated with vehicle or epacadostat (Fig. 4A), possibly reflecting habituation. Epacadostat-treated mice traveled slightly less distance relative to mice receiving vehicle, but there was no difference in the percentage immobile (Fig. 4B). There were no differences in time spent in the outer, middle, or inner zones among mice receiving vehicle or epacadostat (Supplementary Fig. S7A–S7C).

Tumor-bearing mice receiving vehicle experienced an increase in immobility during the forced swim test (Fig. 4C, median percent immobile: 3.0% vs. 6.6%, P = 0.002), replicating the prior experiment presented in Fig. 2F. In contrast, tumor-bearing mice treated with epacadostat did not experience such an increase in immobility (2.9% vs. 3.2%, P = 0.560). The percentage immobile during the tail suspension test was similar for tumor-bearing mice receiving vehicle over the study period (32.7% vs. 39.3%, P = 0.071), but tumor-bearing mice receiving epacadostat experienced less immobility (Fig. 4D, 31.6% vs. 25.0%, P = 0.007).

Similar trends in physical mobility were assessed and observed in the open field test for mice receiving escitalopram (Fig. 4E and F). There was some habituation during the experiment and no difference in the percent immobile. Tumor-bearing mice receiving escitalopram spent more time in the outer zone (i.e., greater anxiety-like behavior), and less time in the middle and inner zones relative to mice who received vehicle (Supplementary Fig. S7G and S7I). In contrast with epacadostat, escitalopram was ineffective at mitigating depressive-like behavior in either the forced swim or tail suspension tests (Fig. 4G and H). These data suggest that escitalopram is rather ineffective at treating pancreatic cancer–associated psychiatric symptoms in this mouse model.

Neither epacadostat, nor escitalopram, modified mouse behavior during the elevated plus maze test; all mice spent a greater proportion of time in the closed arm and had a reduction in line crossings by the end of the study period (Supplementary Fig. S7D–S7F and Supplementary Fig. S7J–S7L), which suggests neither medication was effective as an anxiolytic.

Serum samples from representative mice receiving vehicle or epacadostat were analyzed using LC-MS (Fig. 4I, Supplementary Fig. S7M). Relative tryptophan levels were comparable, whereas a 95% reduction in kynurenine levels was observed in mice receiving epacadostat. The drug consistently decreased other downstream kynurenine metabolites as a general pattern, but these did not reach statistical significance. By comparison, serum tryptophan and kynurenine levels were similar between mice receiving vehicle or escitalopram (Fig. 4J; Supplementary Fig. S7N). Serum serotonin levels increased over the study period, but were similar at both time points between vehicle and epacadostat treatment groups (Supplementary Fig. S8A).

As observed above in cell culture experiments, whole-tumor lysates of mice receiving epacadostat demonstrated a compensatory increase in IDO1 levels, indirectly validating on-target activity (Fig. 4K). By IHC analysis, mice treated with epacadostat experienced a drastic increase in IDO1 expression, as compared with vehicle or escitalopram-treated mice (Fig. 5A and C). Tph1 protein levels were similar in the three treatment groups (Supplementary Fig. S8B and S8D).

The preclinical evaluation of cancer-associated psychiatric conditions is limited, partly because of the challenge in measuring depression in small animals. One previous study used subcutaneous human pancreatic cancer xenografts in immunocompromised mice and demonstrated that fluoxetine (an SSRI) moderately improved exploration and grooming during open field testing relative to vehicle (45). Immunocompetent mice bearing subcutaneous murine colon cancer allografts have been shown to exhibit depressive-like behavior, with increased time immobile during the forced swim test and reduced sucrose preference relative to non–tumor-bearing mice (46). Other groups have induced depression in tumor-bearing mice using the chronic unpredictable mild stress model to study the effectiveness of antidepressants (47). Importantly, these past studies have all examined the efficacy of SSRIs (or analgesics) on cancer-associated depression.

An early report from 1998 linking the tryptophan–kynurenine axis to the immune system demonstrated that IDO1 played a key role in modulating maternal immune tolerance to a developing fetus (48). Subsequent studies have examined IDO1 as a therapeutic target, with the hope that IDO1 blockade would augment the effectiveness of the antitumor immune response. Toward this end, epacadostat and other IDO1 inhibitors have been paired with validated immunotherapies (e.g., checkpoint inhibitors) in preclinical models (49) and in patients (27, 29). Unfortunately, a recent placebo-controlled trial failed to show any therapeutic survival benefit in patients with advanced melanoma (29). As a result, the promise of IDO1 inhibitors has been called into question.

Our work, along with other prior works, suggests that IDO1 inhibitors still have a promising clinical application, beyond immunotherapy. Numerous studies have consistently shown that epacadostat dramatically reduces circulating kynurenine levels in non–tumor-bearing humans and mice (28, 38, 50). These metabolic alterations may have clinical ramifications for mood, because administration of systemic kynurenine to mice causes behavioral changes and an anxiety-like phenotype (24). In addition, a different IDO inhibitor, DWG-1036, reduced depressive-like behavior in a murine model of Alzheimer's disease (51). This line of investigation is particularly compelling for tumors like pancreatic cancer, where IDO1 expression is substantially elevated. A study of 17 patients with resected pancreatic cancer was evaluated for tryptophan pathway metabolite levels and mood scores. Statistically significant correlations were identified between increased tumor burden and higher kynurenine levels, and between poor mood scores and lower kynurenic acid to tryptophan ratio (17). Despite these compelling studies, to our knowledge, there have not been any focused examinations of the tryptophan–kynurenine axis in animal models of pancreatic cancer or the impact of IDO1 inhibitors on depressive-like behavior in preclinical cancer models. Moreover, the prior study of kynurenine pathway metabolites in pancreatic cancer patients did not test a non-cancer control group. Thus, the present work provides the strongest foundation to target this pathway as a novel treatment for cancer-associated depression in patients with pancreatic cancer.

In this report, we confirm that IDO1 expression is upregulated by immune factors in the pancreatic tumor microenvironment (52). This notion is underscored by high IDO1 expression in orthotopic tumors in immunocompetent mice, but not immunodeficient mice. As a consequence, immunocompetent mice bearing pancreatic cancer exhibit signs of depression that were appreciable despite minimal changes in overall wellness and physical mobility. This was not true among immunocompromised mice, suggesting that the immune system plays an integral role in cancer-associated depression. Pharmacologic inhibition of IDO1 with epacadostat effectively reduced circulating kynurenine levels in pancreatic cancer–bearing mice with a compensatory increase in IDO1 expression, and an associated improvement in depressive-like symptoms. The behavioral results were confirmed with two independent tests (forced swim and tail suspension). Although low-dose epacadostat (50 mg/kg, daily; ref. 28) was used in our experiments, doses up to 2,000 mg/kg are well-tolerated by mice (28) and future studies are needed to determine whether higher doses offer even greater protection against depressive-like behavior. Along these lines, genetic ablation of IDO1 in mice could offer additional support of the role of IDO1 in pancreatic cancer–associated depression.

We focused on IDO1 over the known homologous and analogous enzymes, because it is upregulated in multiple cancer types, and an available, highly selective small-molecule inhibitor has been safely administered to patients in multiple clinical trials (27, 29). Therefore, our data can be rapidly translated to patients. Inhibition of TDO2 may have a similar effect on cancer-associated depression as IDO1 blockade, because TDO2 plays a similar metabolic role and also appears to be upregulated in pancreatic tumors (Supplementary Fig. S1B). Early investigations are underway trialing combination IDO1–TDO2 inhibitors in humans and future studies should examine the effect of TDO2 inhibition alone and in combination with IDO1 inhibition.

This work provides further evidence that kynurenine and related circulating metabolites produced by the tumor potentially drive an underlying paraneoplastic syndrome characterized by unwanted neuroactive side effects. Notably, changes in serotonin levels do not appear to be a major factor driving pancreatic cancer–associated depression in the models used in these experiments. Although the present data confirm recent human studies that epacadostat lacks antitumor capabilities without any effect on pancreatic cancer growth or mouse survival, an important clinical benefit was nevertheless detectable. As of December 2021, none of the 59 active clinical trials testing IDO1 inhibitors in patients included measures of mood or depression as important outcome metrics. Ultimately, this research and future studies could be foundational for clinical trials testing epacadostat or other IDO1 inhibitors as targeted antidepressants in patients with pancreatic cancer.

Mice with pancreatic cancer exhibit depressive-like behavior relative to control mice, with an associated increase in serum kynurenine levels. These findings mirror results in patients with pancreatic cancer. Treatment of pancreatic cancer–bearing mice with epacadostat improved depressive-like behaviors and dramatically reduced serum kynurenine levels. These data provide compelling evidence for future studies in mice and patients to examine epacadostat as a bona fide treatment for pancreatic cancer–associated depression.

No disclosures were reported.

J.J. Hue: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. H.J. Graor: Conceptualization, data curation, investigation, methodology, writing–review and editing. M. Zarei: Data curation, formal analysis, investigation, methodology, writing–review and editing. E.S. Katayama: Data curation, validation, investigation, methodology, writing–review and editing. K. Ji: Investigation, writing–review and editing. O. Hajihassani: Writing-review and editing. A.W. Loftus: Writing-review and editing. A. Vaziri-Gohar: Writing-review and editing. J.M. Winter: Conceptualization, resources, supervision, funding acquisition, writing–original draft, writing–review and editing.

Grant support for this research comes from American Cancer Society MRSG-14–019–01-CDD and 134170-MBG-19–174–01-MBG, NCI R37CA227865–01A1, and the University Hospitals research start-up package (to J.M. Winter). The authors would like to acknowledge the Tissue Resources Core Facility of the Comprehensive Cancer Center of Case Western Reserve University and University Hospitals of Cleveland (P30 CA43703), namely Jenifer Mikulan and Adam Kresak, for their expertise in tissue processing and IHC. In addition, the authors acknowledge Ian Adams and Dr. D. Katz in the Mouse Behavioral Phenotyping Core for their expertise in design of behavioral studies. The authors would like to acknowledge the Mass Spectrometry Core at Beth Israel Deaconess Medical Center for their assistance with metabolomic assays. Finally, we are grateful for additional support from numerous donors to the University Hospitals pancreatic cancer research program, including the John and Peggy Garson Family Research Fund, the Jerome A. and Joy Weinberger Family research fund, Robin Holmes-Novak in memory of Eugene, Rosi and Saby Behar, and Fred and Brittan DiSanto.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC Section 1734.