Glutaminase inhibition ameliorates cancer-associated fibroblast lipid support of pancreatic cancer cell growth

Abstract

Background

Lipid homeostasis is critical for pancreatic adenocarcinoma (PDAC) cell survival under hypoxic and nutrient-deprived conditions. Hypoxia inhibits unsaturated lipid biosynthesis, compelling cancer cells to depend on exogenous unsaturated lipids to counteract saturated lipid-induced toxicity. Our previous work revealed that cancer-associated fibroblasts (CAFs) secrete unsaturated lipids, primarily lysophosphatidylcholines (LPCs), to alleviate lipotoxic stress in PDAC cells. Here, we conducted a drug screen to identify compounds that bypass the rescue effect of exogenous LPCs on cancer cell survival under stress.

Methods

We employed high-throughput screening of a bioactive chemical library with 3,336 compounds, including FDA-approved drugs and drug-like molecules against defined molecular targets. Two assays were performed: a cytotoxicity assay to exclude indiscriminately toxic compounds at 1 μM and an LPC crosstalk inhibition assay to identify compounds that selectively reduce cancer cell viability in the presence of LPCs under stress conditions.

Results

CB-839, a glutaminase inhibitor, was identified as the most effective compound, selectively inhibiting the LPC-mediated rescue of PDAC cell viability effect without intrinsic cytotoxicity. Mechanistic studies revealed that CB-839 induces cell death by activating the pro-apoptotic ATF4/CHOP pathway, reducing antioxidant production, and increasing reactive oxygen species (ROS). While CB-839 showed limited efficacy against PDAC tumor cells alone in vivo, it modestly inhibited tumor growth in a PDAC-CAF co-implanted subcutaneous mouse model, highlighting its potential to disrupt CAF-mediated nutrient support. Additionally, glutamine antagonists showed more potent tumor-suppressive effects than CB-839.

Conclusion

Our findings emphasize the importance of glutamine metabolism inhibition in suppressing tumor growth and disrupting CAF-mediated crosstalk. We further underscore the potential of glutamine antagonist prodrugs as a strategy to target metabolic vulnerabilities in PDAC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40170-025-00389-z.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is the third leading cause of cancer-associated death in the United States, with a five-year survival rate of 12% [1]. During PDAC progression, the tumor microenvironment (TME) becomes increasingly desmoplastic due to an abnormal accumulation of non-transformed stromal cells and extracellular matrix (ECM) deposition. PDAC TME desmoplasia acts as a physical barrier, impairing vascular function and restricting nutrient and oxygen delivery, leading to severe tissue ischemia [2–4]. PDAC is recognized as one of the most hypoxic human cancers [5, 6], which triggers metabolic reprogramming that supports tumor development and therapeutic resistance.

To adapt and thrive in this challenging and complex TME, PDAC cells rewire their metabolism, exhibiting diverse metabolic phenotypes through the reprogramming of glucose, glutamine (Gln), amino acid, and lipid metabolism [7–11]. Malignant cells frequently elevate glycolytic flux to achieve a higher rate of adenosine triphosphate (ATP) production, known as the Warburg effect [12, 13]. Additionally, PDAC cells enhance nutrient acquisition by increasing scavenging and recycling activities for survival and proliferation. Among these adaptations, alterations in lipid metabolism are particularly prominent. Lipids are a diverse group of biomolecules with varying structures and functions [14], and most originate from fatty acids of different carbon chain lengths and degrees of saturation. Cancer cells enhance lipid biosynthesis and uptake to sustain membrane homeostasis and energy acquisition during rapid proliferation [15]. However, hypoxia impedes the desaturation of de novo synthesized fatty acids, a reaction catalyzed by Δ9 stearoyl-CoA desaturase 1 (SCD1) [16]. Fatty acid desaturation via SCD1 activity requires oxygen as a terminal electron acceptor to convert stearate (C18:0) into oleate (C18:1), which is the most abundant unsaturated fatty acid in the cell and serves as a precursor to other unsaturated lipids (see Fig. 1a). Consequently, hypoxia renders cancer cells dependent on exogenous unsaturated lipids to maintain lipid homeostasis, membrane fluidity, and organellar membrane function [16, 17]. For instance, abnormal saturated lipid accumulation and storage in clear cell renal cell carcinomas (ccRCCs) increase cytoplasmic ceramide and acyl-carnitine levels, which can induce mitochondrial dysfunction and trigger apoptosis [18, 19].

Fig. 1.

Pancreatic cancer cells acquire unsaturated lipids to survive under stress. a Monounsaturated fatty acyl-CoA biosynthesis catalyzed by SCD1. b Chemical structures of LPC (18:0) and LPC (18:1). c Schematic model of accumulating saturated fatty acids due to the blockade of exogenous and endogenous unsaturated fatty acid supply resulting in cancer cell death by activated ER stress pathways (Left). Schematic model of unsaturated LPC in reversing ER stress under lipid imbalance (Right). d Quantification of crystal violet stained live PANC-1 and Su.86.86 under hypoxic and low nutrient stress (SO stress) at 96 h. Cells were treated with 20 μM LPC (18:1), 20 μM LPC (18:0), and 60 μM oleate. e Quantification of crystal violet stained live PANC-1 and Su.86.86 under lipotoxic stress (200 nM SCDi) at 96 h. Cells were treated with 60 μM oleate, 20 μM LPC (18:1), and 20 μM LPC (18:0). Error bar a represents SD of three experimental replicates. All P values were calculated using a two-tailed Student’s t-test or one-way ANOVA for multiple comparisons. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant

Our recent study revealed that cancer-associated fibroblasts (CAFs) play a critical role in secreting unsaturated lipids to help restore lipid homeostasis and reduce lipid saturation stress in hypoxic cancer cells [20]. Additionally, CAFs are significantly less hypoxic than malignant ductal cells, likely due to TME “microdomains” where CAFs are more proximal to blood vessels. Lipidomic analysis revealed that unsaturated lysophosphatidylcholines (LPCs, see Fig. 1b), secreted by CAFs under either hypoxia or normoxia, help alleviate hypoxia-induced endoplasmic reticulum (ER) stress in pancreatic cancer cells. This occurs through their conversion into more fluid phosphatidylcholine (PC) species, which in turn supports membrane stability and integrity [20] . While these findings highlight the importance of CAF-PDAC interactions under hypoxic conditions, further research is needed to explore ways to disrupt this CAF-PDAC crosstalk and mitigate their supportive role for disease progression.

Gln metabolism is also critical for cancer cell proliferation [21, 22], as it supplies fuel for various metabolic pathways, including the Krebs cycle, redox homeostasis, and the synthesis of essential cellular components such as nucleic acids, fatty acids, glutathione (GSH), and other amino acids [23]. Numerous clinical trials have investigated the use of glutaminase (GLS) inhibitors combined with chemotherapy for cancer treatment in conditions such as leukemia, renal cell carcinoma, and non-small cell lung cancer (NCT02071888, NCT02071927, NCT02071862). However, previous studies revealed that these inhibitors failed to significantly suppress the growth of KRAS-driven cancer cells in vivo [24–26]. This resistance is attributed to the capability of pancreatic cancer cells to develop an adaptive metabolic response, such as glutamate (Glu) reaccumulation via Gln-independent, Glu-producing enzymes, underscoring the need for combination therapies to counteract this adaptation and enhance the efficacy of GLS inhibition.

We describe here a high-throughput chemical screen using a library of FDA-approved drugs and pharmacologically active compounds to disrupt the supportive functions of unsaturated exogenous lipids and induce PDAC cell death. This was achieved by supplementing pancreatic cancer cells with key exogenous unsaturated LPCs while simultaneously inhibiting unsaturated fatty acid biosynthesis through SCD inhibitors. Our screening revealed that the GLS inhibitor CB-839 counteracts cell viability rescue effects of unsaturated LPCs by activating the pro-apoptotic ATF4/CHOP signaling pathway and increasing reactive oxygen species (ROS) levels. Although CB-839 did not show direct tumor growth inhibition in PDAC models, we observed a modest reduction in tumor growth in PDAC-CAF co-implanted mouse models. We therefore explored the potential of CB-839 to disrupt PDAC-CAF crosstalk in vivo, highlighting the significance of targeting Gln metabolism for PDAC treatment within the context of PDAC TMEs enriched with CAFs. As stated above, CAFs are essential providers of unsaturated lipids in the nutrient-deprived, hypoxic TME, preventing PDAC cells from ER stress-induced death. However, CB-839 reactivates alternative stress signaling pathways, including ROS elevation and CHOP re-expression, thereby disrupting the supportive role of CAFs within the TME and inducing PDAC cell death in vivo. Gln metabolism has emerged as a promising therapeutic target in cancer therapy, not only through the clinical application of CB-839 but also via the development of Gln antagonist prodrugs with improved tumor exposure and reduced gastrointestinal (GI) toxicity [27, 28]. These approaches are being extensively explored across various cancer types, including glioma, pancreatic cancer, and lung cancer [29–31]. We conclude that glutamine antagonist prodrugs may exert their effects in part by impacting PDAC-CAF crosstalk supporting cancer cell lipid homeostasis.

Methods

Cell culture

Human pancreatic cancer cell lines (PANC-1, Su.86.86) were obtained from ATCC. KPC4662, derived from primary tumors in KPC (KrasLSL-G12D/+; Trp53LSL-R172H/+; Pdx1-Cre) mice, was a gift from Dr. Robert Vonderheide. Murine CAF cell lines (“PSC”) were provided by Dr. Alec Kimmelman. PANC-1, KPC4662, and CAFs were maintained in DMEM with 10% FBS and penicillin/streptomycin, while Su.86.86 was cultured in RPMI-1640 with 10% FBS. All cells were grown in 5% CO₂ at 37°C and routinely tested for Mycoplasma every 3 months. Cells were used for no more than 20-25 passages.

Syngeneic mouse models

1×10⁵ KPC4662 cells were injected subcutaneously, or co-implanted with 5×10⁵ murine CAFs, into each flank of 6–8-week-old female C57BL/6J mice (Jackson ImmunoResearch, Strain #000664). Cells were cultured in DMEM with 10% FBS, resuspended in ice-cold PBS, and mixed 1:1 with Matrigel (BD Biosciences) for a final injection volume of 200 μL. Mice were treated with either vehicle (25% hydroxypropyl-β-cyclodextrin in 10 mmol/L citrate, pH 2.0) or 200 mg/kg Telaglenastat (CB-839, Calithera Biosciences) via oral gavage, twice daily. For DON treatment, animals received 0.25 mg/kg 6-diazo-5-oxo-L-norleucine (DON, Cayman) or vehicle (PBS) via intraperitoneal injection once daily. Tumor volume was measured twice weekly using the formula (width² × length)/2, starting when tumors reached ~100 mm³. At the end of the experiment, mice were euthanized via CO₂ inhalation followed by cervical dislocation, and tumors were harvested for analysis. All procedures were approved by the IACUC of the University of Pennsylvania.

Cell growth assay

PDAC cells (1×10³ per well) were seeded in 384-well plates and subjected to the indicated culture conditions and drug treatments. Cell growth was measured at the specified time points using the CellTiter-Glo 2.0 assay (Promega). For clonogenic assays, PDAC cells (5×10⁴ per well) were seeded in 12-well plates with CAF-conditioned medium, or treated with 20 μM LPC 18:1 (Avanti Polar Lipids, 845875P), 5 mM NAC (Merck, A9165), 5 mM Glutathione ethyl ester (Cayman, 14953), 200 μM α-Ketoglutaric acid sodium salt (Thermo Scientific, 439350050), and 200–400 nM SCD inhibitor (Cayman, CAY10566) cultured under the indicated conditions for 72 hours, and stained with crystal violet. The crystal violet was dissolved in methanol, and cell numbers were quantified by measuring absorbance at OD₅₇₀ using a plate reader.

Annexin V-PI cell viability assay

5×10⁴ cells of each cell line were plated in triplicate in 6-well plates and grown in repleted medium. The next day, cells were washed twice with PBS and replaced with indicated conditions. Cells were cultured in 10% FBS, 0.5% FBS, or 10% delipidated FBS (Gemini Bio-Products, 900–123). For lipid treatment, 60 μM oleic acid (Sigma, O3008), or 20 μM LPC 18:1 was added to cell culture. For CB-839 treatment, 1 μM CB-839 was added to indicated cell culture. Cells and supernatant were collected after 96 hours. Cell viability was detected using FITC-Annexin V PI Kit (BD Bioscience, 556547) according to the manufacturer’s instructions. Flow cytometry was performed using the BD FACSCalibur instrument, with live cells represented as Annexin V and PI double-negative population. Flow data were analyzed with FlowJo software.

RNA isolation and quantitative RT-PCR

Total RNA was extracted by RNeasy Mini Kit (Qiagen, 74104). cDNAs were synthesized with a High-Capacity RNA-to-cDNA kit (Applied Biosystems, 4368814). qRT-PCR was performed on a ViiA7 Real-Time PCR system (Applied Biosystems) with TaqMan master mix (Life Technologies, 4444965). Pre-designed Taqman primers (Life Technologies) were used for detecting XBP1t (HS00231936_M1), XBP1s (HS03929085_G1), DDIT3/CHOP (Hs00358796_G1), 18S (HS03928985_G1). Gene expression levels were normalized to 18S ribosomal RNA.

ROS measurement

5 × 10⁴ Su.86.86 cells were plated in 6-well plates in triplicate under the indicated conditions. Cells were treated with 1 μM CB-839, 5 mM N-acetyl-l-cysteine (Millipore Sigma, A9165) or DMSO for 72 hours. Live cells were washed with PBS, stained with DCFDA (Abcam, 113851) for 1 hour at 37°C, trypsinized, washed twice, resuspended in PBS with 1% BSA, passed through a cell strainer, and analyzed via flow cytometry (BD Accuri C6).

Gas chromatography–mass spectrometry (GC–MS)

Gas chromatography mass spectrometry (GC-MS) analysis was performed as previously described [32]. Briefly, tumors were homogenized, weighed dried and grounded accurately, and then transferred to 1-dram glass vial. Tissue samples were washed with cold PBS, collected into 80% methanol, and cells lysed by three cycles of freeze-thaw. Insoluble materials were pelleted by centrifugation and the metabolite containing supernatant was dried overnight (Speedvac). Dried metabolites were derivatized to methoxime-TBDMS adducts by incubating with methoxyamine hydrochloride (Sigma-Aldrich, 593-56-6) in pyridine (Sigma-Aldrich, 110-86-1) at 70°C for 15 minutes followed by addition of N-tert-butyldimethylsiyl-N-methyltrifluoroacetamide (MTBSTFA, Sigma-Aldrich, 77377-52-7) for 60 minutes. Derivatized samples were injected onto an Agilent Technologies 8890 gas chromatographer with a DB-5MS column (Agilent, G3900) coupled to an Agilent Technologies 5977 GC/MSD. Retention times were verified via external standards. Peak areas were normalized to total ion count (TIC).

High throughput chemical screening

The UPenn Bioactive chemical library was purchased from Selleckchem as DMSO stocks at 10 mM. The library contained 3336 compounds, including 992 FDA approved drugs and 2344 drug-like molecules against defined molecular targets.

For drug cytotoxicity assay, 500 cells were seeded per well in 20 μL of assay medium in Corning 3570 microplate using a MultidropTMCombi Reagent Dispenser (Thermo Scientific). Cells were allowed to attach overnight at 37 °C, 5% CO₂ in a humidified chamber. ~20 hours post-cell plating, 5 uL of assay medium containing 1% DMSO for a final concentration of 0.2% was dispensed to assay plates. Compounds in 50 nL volume were transferred to assay plates using a 384, 50 nL slotted pin tool (V&P Scientific) and a JANUS Automated Workstation (Perkin Elmer) and columns were treated with 50 nM Bortezomib with DMSO. Cells were incubated for 72 hours at 37°C, 5% CO₂ in a humidified chamber. Assay plates were removed from the incubator for 1 hour to equilibrate to room temperature prior to adding 20 μL of Atplite (PerkinElmer). Luminescence was measured on an EnVision Xcite Multilabel Plate Reader (PerkinElmer), using the ultrasensitive luminescence measurement technology. For LPC effect inhibition assay, 5 uL of assay medium containing 1 uM SCDi and 100 uM LPC was dispensed to columns for a final concentration of 0.2 uM SCDi and 20 uM LPC. Cell viability was measured using the same protocol as drug cytotoxicity assays. Raw values from Negative and Positive control wells were aggregated and used to calculate Z’-factors for each assay plate, as a measure of assay performance and data quality, with a Z’-factor >0.5 representing acceptable data. Raw data values of sample wells were normalized to aggregate negative control and Positive plate control wells and expressed as: Percent of Control [POC= (Test well / NegativeControl_avg x 100)]; Z-score [Zscore = (Test well - NegativeControl_avg) / (NegativeControl_sd)].

Statistical analysis

Statistical analyses were performed using GraphPad Prism 9.0.0, with Student’s two-tailed t-test for pairwise comparisons and one-way ANOVA for multiple comparisons. Results are shown as mean ± SEM for at least three independent experiments or mean ± SD for representative experiments at least three technical replicates. Tumor growth curves are presented as mean ± SEM with at least five tumors per group. A p-value < 0.05 was considered significant. Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant.

Results

Pancreatic cancer cells acquire unsaturated lipids to survive under stress

Our previous findings highlighted the crucial role of lipid homeostasis in maintaining cell viability under hypoxic and nutrient-deprived conditions. Under these stress conditions, PDAC cells exhibit reduced proliferation and are more prone to apoptosis due to lipotoxicity, which results from the accumulation of saturated lipids [16, 18, 20]. Inhibiting de novo synthesis of unsaturated fatty acids, either through hypoxia or SCD chemical inhibition, forces pancreatic cancer cells to rely on external unsaturated lipids to prevent saturated lipid accumulation [17, 18]. Additionally, we identified that key lipid metabolites secreted by fibroblasts, specifically unsaturated LPCs (primarily LPC 18:1), help pancreatic cancer cells alleviate cytotoxic stress responses and promote survival (Fig. 1c).

To model the hypoxic and low-nutrient environment in the PDAC TME, we cultured pancreatic cancer cells under 0.5% FBS and 0.5% O₂ (referred to as “SO stress” [16]). As noted, hypoxic PDAC cells depend heavily on external sources of unsaturated lipids to mitigate the buildup of saturated fatty acids. Combining hypoxia with low serum levels further intensifies stress signaling due to unresolved lipid saturation stress. Under SO stress, we observed a consistent reduction in the viability of PANC1 and Su.86.86 cells, which was restored by the addition of unsaturated lipids such as oleic acid (OA) and LPC (18:1) (Fig. 1d, Supplementary Fig. S1a). In contrast, saturated lipids like LPC (18:0) did not improve cell viability (Fig. 1d).

Culturing cells in low serum (0.5% FBS) affects overall growth and limits access to various nutrients. To specifically investigate the role of lipids under hypoxic stress, we cultured PANC-1 and Su.86.86 cells in 10% delipidated (DLP) FBS. Consistent with our findings under SO stress, we observed a significant reduction in cell viability under 0.5% O₂ and 10% DLP FBS conditions (Supplementary Fig. S1b). The addition of OA and LPC (18:1) supported cell viability, highlighting a crucial role of lipid homeostasis, in addition to other nutrients, for cell survival in harsh TMEs.

Because hypoxia influences various aspects of cell metabolism and stress responses, we chose to specifically investigate the role of unsaturated lipids on cell survival during lipotoxic stress by culturing PDAC cells under low serum conditions in normoxia accompanied by pharmacological SCD inhibition. SCD inhibition blocks the rate-limiting step of unsaturated fatty acid biosynthesis, allowing us to study lipid homeostasis without the non-lipid-related metabolic alterations caused by hypoxia. Consistent with SO stress, we observed low viability of PANC1 and Su.86.86 cells under lipotoxic stress. Unsaturated lipids, but not saturated lipids, rescue cell viability (Fig. 1e, Supplementary Fig. S1c).

High throughput screen to target LPC-mediated survival pathways in PDAC

Given our in vitro findings, which elucidated the role of unsaturated LPCs in bolstering PDAC cell viability under hypoxia and nutrient deprivation, our goal was to disrupt LPC utilization by targeting LPC uptake or metabolism in PDAC cells. However, as lipids are essential components of cellular membranes and key energy sources, their regulation is tightly controlled by multiple redundant pathways, making it challenging to effectively target a single protein or regulator. To approach this goal more broadly—targeting LPC-mediated PDAC cell survival both directly and indirectly—we conducted high-throughput chemical screening utilizing a library of ~4,000 small molecules, enriched for FDA-approved drugs (~1500) and compounds with established pharmacological activities (~2500), such as cancer chemotherapeutics, kinase inhibitors, and metabolism inhibitors (Supplementary Table S1). Due to technical limitations preventing the placement of microplate dispensers and automated workstations in a hypoxic chamber, we employed a “lipotoxic” stress culture condition (200 nM SCD inhibitor and 0.5% FBS under normoxia) to mimic lipid saturation stress. This lipotoxic condition effectively simulates stress caused by inhibiting unsaturated fatty acid biosynthesis and restricting external lipid supply [20] .

Consistent with our previous data, PDAC cells displayed increased viability when treated with unsaturated LPCs under lipotoxic stress based on luminescent cell viability assays (Fig. 2a). Su.86.86 cells were selected due to their enhanced dynamic range in live cell numbers between lipotoxic stress and LPC (18:1) treatment. To identify chemicals that specifically inhibit LPC-mediated rescue of PDAC cell viability, we performed two assays with a 5-6-fold window for screening and high Z values for assay plates (Supplementary Fig. S2a-2d). First, we conducted a chemical cytotoxicity assay by culturing Su.86.86 cells with 1 µM of each compound from the bioactive drug library, excluding cytotoxic agents under low serum and normoxic conditions. Second, we performed an LPC inhibition assay under lipotoxic stress with the same concentration of chemicals, supplemented with 20 µM LPC (18:1). After 72 hours of treatment, cell numbers were measured using a luminescence-based assay (Fig. 2b).

Fig. 2.

High throughput screening to target LPC-mediated survival pathways in PDAC. a Luminescence-based measurement of human pancreatic cancer cell numbers under normoxia, 0.5% FBS, treated with 200 nM SCDi and 20 μM LPC (18:1) for 72 h. b Schematic model of chemical screening in the setting of luminescence-based measurements. Cytotoxicity assay was performed on cells cultured at 0.5%FBS and treated with chemicals at 1 μM for 72 h. LPC inhibition assay was performed under SCDi and LPC (18:1) treatment. c Comparison dot plot from the chemical screen. X-axis: Percent of control (POC) for LPC effect inhibition; Y-axis: POC for cytotoxicity. Red box: Hits with POC > 80% in cytotoxicity and POC < 40% in LPC inhibition screens. d Top 19 candidate hits and their corresponding targets. e–g Validation of 1 μM CB-839 (e), 1 μM Geldanamycin (f), 1 μM KW-2478 (g) inhibiting LPC effects on Su.86.86 and PANC-1 cell growth under SCDi for 72 h. Error bar a represents SD of three experimental replicates. All P values were calculated using a two-tailed Student’s t-test or one-way ANOVA for multiple comparisons. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. ns, not significant

Using normalized cancer cell numbers from the screen, relative to a DMSO control, we established a threshold to identify chemicals that allowed high cell viability in the cytotoxicity assay but low cell viability in the LPC inhibition assay. This approach identified 19 candidates that selectively target LPC-mediated pathways in PDAC cells, focusing on cell signaling, kinase inhibition, and metabolic disruption (Fig. 2c and d).

To validate the outcomes of our approach, we conducted further investigation into several chemicals identified among the top candidates. Surprisingly, CB-839 (Telaglenastat), a glutaminase inhibitor, emerged as a leading hit. Glutaminase (GLS) catalyzes the conversion of Gln to Glu, a key step in glutaminolysis, a metabolic pathway frequently upregulated in cancer cells (see Fig. 4a) [33–35]. Our validation experiments confirmed that CB-839 effectively suppressed LPC-mediated PDAC cell survival (both PANC-1 and Su.86.86) under lipotoxic stress (Fig. 2e, Supplementary Fig. S2e), with minimal impact on PANC-1 under normoxic conditions with 0.5% FBS as a control. We also observed that CB-839 does not exacerbate cell death in the absence of LPC addition (Supplementary Fig. S2f). Interestingly, CB-839 also slightly enhanced the viability of Su.86.86 cells in normoxia and low serum conditions.

Fig. 4.

CB-839 elevates ROS levels in cancer cells under lipotoxic stress. a Schematic model of GLS inhibition blocking Glu production, leading to reduced GSH generation and enhanced ROS accumulation. GCLM/GCLC: glutamate-cysteine ligase catalytic heavy-chain subunit (GCLC) and modifier light-chain subunit (GCLM). b Median fluorescence intensity of ROS in Su.86.86 cells under lipotoxic stress treated with LPC (18:1), CB-839, and NAC for 72 h. c Quantification of Su.86.86 cell numbers under lipotoxic stress treated with LPC (18:1), CB-839, and 5 mM NAC for 72 h. Error bar a represents SD of three experimental replicates. All P values were calculated using a two-tailed Student’s t-test or one-way ANOVA for multiple comparisons. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant

Notably, we identified two heat shock protein 90 (HSP90) inhibitors that effectively disrupted LPC-mediated PDAC cell survival under lipotoxic stress. These included geldanamycin, which binds to the unique ADP/ATP-binding pocket of HSP90, and KW-2478 which represents a non-ansamycin HSP90 inhibitor. HSP90 functions as a chaperone that aids in proper protein folding and stability, particularly under thermal stress. Consistent with the screening results, both HSP90 inhibitors effectively blocked LPC-mediated rescue of PDAC cell viability under lipotoxic conditions (Fig. 2f and g). HSP90 inhibition may block LPC-mediated rescue through activation of ER stress by elevating unfolded protein responses. In a previous study, geldanamycin was shown to induce ER stress by interacting with both HSP90 and glucose-regulated protein 94 (GRP94, the ER-resident paralog of HSP90), which blocks cytosolic proteins from maturation and further enhances cellular response to stress in the ER [36, 37]. However, none of the HSP90 inhibitors have been approved by the U.S. Food and Drug Administration (FDA), primarily due to challenges in identifying appropriate biomarkers, limited drug efficacy, and unfavorable toxicity profiles [38]. We utilized HSP90 inhibitors specifically for screening validation purposes, and did not pursue them any further.

Overall, our unbiased chemical screen yielded several promising chemical candidates that specifically reduce PDAC cell survival, even in the presence of unsaturated LPCs.

CB-839 induces additional stress signals and causes cancer cell death in vitro

Our verification experiments confirmed that the GLS inhibitor CB-839 effectively abrogated the protective effects of LPCs in maintaining cell survival under lipotoxic stress. Additionally, CB-839 also exhibited functionality under “SO stress” (Fig. 3a, Supplementary Fig. 3a), suggesting that inhibiting Gln metabolism overrides the LPC pathway in supporting cell viability. Previously, we noted that unsaturated LPCs mitigated ER stress induced by the accumulation of saturated lipids due to an inability to synthesize unsaturated lipids under hypoxia [20] .

Fig. 3.

CB-839 activates ATF4/CHOP stress signals. a Luminescence-based cell number measurement of PANC-1 and Su.86.86 cells under 0.5% O₂, 0.5% FBS, 20 μM LPC (18:1), and 1 μM CB-839 for 72 h. b Schematic model of ER stress pathways. c qRT-PCR analysis of XBP1s/XBP1t, CHOP, and ATF4 mRNAs in PANC-1 cells under hypoxic stress (0.5% O₂, 0.5% FBS), treated with LPC (18:1) and CB-839 for 24 h. d Western blot of spliced XBP1, ATF4, and CHOP protein expression under hypoxic stress (0.5% O₂, 0.5% FBS), treated with LPC (18:1) and CB-839 for 72 h. Error bar a represents SD of three experimental replicates. All P values were calculated using a two-tailed Student’s t-test or one-way ANOVA for multiple comparisons. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant

ER stress is regulated through three major pathways (Fig. 3b): IRE1 (both α and β isoforms) [39], protein kinase RNA-like ER kinase (PERK) [40], and activating transcription factor 6 (ATF6) (both α and β isoforms) [41]. The transmembrane domain of IRE1α is crucial for dimerization or oligomerization in response to saturated fatty acid (SFA)-induced stress, playing a key role in sensing membrane lipid saturation or abnormal lipid compositions [42, 43]. PERK, an upstream kinase, phosphorylates eIF2α (Eukaryotic Initiation Factor 2 alpha), leading to the selective translation of ATF4. ATF4 then regulates CHOP, a pro-apoptotic protein that induces cell death [44]. Additionally, ATF4 is activated under nutrient stress, such as amino acid deprivation [45]. Upon ER stress, ATF6 translocates to the Golgi apparatus, where it undergoes cleavage to release its cytoplasmic domain. This domain functions as a transcription factor, enhancing the expression of genes that facilitate protein folding and ER-associated degradation [46, 47] .

We previously demonstrated that unsaturated LPCs can be utilized for phosphatidylcholine formation, which helps relieve lipid saturation stress on the ER membrane and prevent the activation of IRE1 α /XBP1 pathway in hypoxic PDAC cells [20]. We consistently observed that unsaturated LPC treatment suppressed the IRE1α/XBP1 pathway, one arm of the ER stress response, under both “SO” and lipotoxic conditions (Fig. 3c and d, Supplementary Fig. 3b). However, with CB-839 treatment, despite the presence of unsaturated LPCs, PDAC cells showed increased expression of CHOP and ATF4, indicating that the ATF4/CHOP signaling pathway was reactivated by CB-839, bypassing the IRE1α/XBP1 arm (Fig. 3c and d). GLS inhibition induces Glu deprivation and activates pro-apoptotic receptors, further regulating CHOP signaling to promote cell death [48]. The amino acid availability sensor GCN2 is likely to be activated, promoting ATF4 accumulation due to amino acid deprivation [49]. CB-839 also engaged the ATF4/CHOP pathway under lipotoxic stress (Supplementary Fig. 3b), consistent with its activation under SO stress. Moreover, we found that under normoxia and nutrient-replete conditions, CB-839 fails to enhance ER stress gene expression, as the availability of essential nutrients (e.g., amino acids and unsaturated lipids) prevents engaging these signaling pathways (Supplementary Fig. 3c).

Gln metabolism also plays a crucial role in maintaining cellular redox homeostasis by harnessing enhanced ROS levels (Fig. 4a). Inhibiting Gln metabolism with CB-839 leads to ROS accumulation, even under nutrient-replete conditions and normoxia (Supplementary Fig. 4a). This metabolic pathway contributes significantly to the synthesis of GSH, a pivotal antioxidant responsible for mitigating oxidative stress, including ROS accumulation. Our observations revealed that Su.86.86 cells treated with CB-839 exhibited increased accumulation of ROS, even in the presence of unsaturated LPCs (Fig. 4b). Furthermore, when cells were treated with the ROS scavenger N-acetylcysteine (NAC) or reduced GSH, cell viability was restored (Fig. 4c, Supplementary Fig. 4b). The TCA cycle intermediate α-ketoglutarate (α-KG) can be converted to glutamate through glutamate dehydrogenase activity. However, supplementing Su.86.86 cells with α-KG under SCD inhibition and CB-839 treatment fails to restore cell viability, likely due to the many fates of α-KG beyond glutamate (Supplementary Fig. 4b). These findings suggest that Gln metabolism inhibition by CB-839 treatment circumvents the rescue effect of LPCs in PDAC cells under nutritional stress. This effect is attributed to the reactivation of ER stress and ROS accumulation, thereby highlighting the critical role of Gln metabolism in sustaining cell survival under these circumstances.

CB-839 decreases PDAC-CAF crosstalk in vivo

We previously discovered that CAFs produce unsaturated LPCs, which are utilized by PDAC cells to relieve ER stress and maintain cell viability in harsh tumor microenvironments (Fig. 5a). We validated this CAF-PDAC crosstalk by culturing murine KPC4662 PDAC cells under SO or lipotoxic stress with increasing concentrations of CAF-conditioned media (CM). As expected from our prior study [20] , KPC4662 cells exposed to higher concentrations of CAF CM demonstrated improved survival under stress conditions, indicating the protective role of CAFs in pancreatic cancer cell survival (Fig. 5b).

Fig. 5.

In vivo confirmation of CB-839 effects on PDAC cell survival. a Schematic model of PDAC-CAF crosstalk under hypoxic and nutritional stress in vitro. b Crystal violet staining of KPC4662 cells cultured in murine CAF conditioned media with increasing concentrations. c Schematic model of KPC4662 and KPC4662 co-injected with murine CAFs in C57BL/6 mice, treated with 200 μg/mg CB-839 twice daily. d Tumor volume of KPC4662 syngeneic model treated with vehicle or CB-839 (n = 8 tumors from 4 mice per group). e Tumor volume of KPC4662 + CAF co-implanted syngeneic model treated with vehicle or CB-839 (n = 8 tumors from 4 mice per group). f Endpoint tumor weight in subcutaneous models. g qRT-PCR analysis of CHOP gene expression. h Tumor Gln and Glu quantification by GC-MS. Error bar a represents SD of three experimental replicates. Tumor growth data (mean ± SEM, 8 tumors from 4 mice) analyzed by two-way ANOVA. All other P values were calculated using a two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant

To determine if CB-839 could disrupt this crosstalk in vivo, we utilized a syngeneic subcutaneous mouse model injected with KPC4662 cells alone or with both KPC4662 cells and murine CAFs and treated mice daily with CB-839 (Fig. 5c). At endpoint, no significant tumor growth suppression was observed in the KPC4662-only group with CB-839 treatment (Fig. 5d), consistent with previous findings demonstrating the lack of efficacy of CB-839 in inhibiting tumor growth in vivo [24, 25] . Pancreatic cancer exhibits an adaptive metabolic network to counteract the inhibition of Gln metabolism [24]. As a result, CB-839 fails to exhibit antitumor activity in cell line-derived transplanted mouse models due to Glu accumulation. Asparagine synthetase (ASNS), a Gln aminotransferase, replenishes Glu and contributes to cell survival upon Gln withdrawal. This PDAC metabolic plasticity overcomes the inhibition of Gln metabolism.

While CB-839 consistently failed to demonstrate efficacy in reducing tumor growth in KPC4662 subcutaneous tumors, CB-839 clearly slowed tumor growth in the co-implantation model with KPC4662 and murine CAFs (Fig. 5e and f). Specifically, CB-839 reduced tumor weight in the co-injection group, highlighting that GLS pathway inhibition disrupts the CAF-mediated crosstalk. CB-839 also activated CHOP expression in tumor cells (Fig. 5g) and effectively inhibited GLS activity, as evidenced by increased Gln accumulation in tumors at the expense of Glu (Fig. 5h).

DON potently suppresses tumor growth in vivo

Given challenges encountered with CB-839 in PDAC therapy, an alternative drug, 6-Diazo-5-oxo-l-norleucine (DON), a Gln antagonist, has shown promising efficacy in suppressing cancer cell metabolism and tumor growth [27, 50, 51]. Its prodrugs, DRP-104 and JHU-083, have been developed to mitigate GI toxicity and enhance delivery to target tissues, including tumors and the central nervous system. DON prodrugs are being extensively evaluated in preclinical and clinical trials for various cancers, such as pancreatic cancer, medulloblastoma, and fibrolamellar carcinoma [27, 28, 30, 31, 52].

We validated the role of Gln antagonists as a potent alternative to CB-839 in disrupting the supportive function of unsaturated LPCs in PDAC cell survival under hypoxia and low nutrient stress. As expected, DON efficiently inhibited the rescue of unsaturated LPCs in PDAC cell viability under hypoxic and nutrient-deprived conditions in vitro (Fig. 6a). Furthermore, DON exhibited potent tumor growth suppression in vivo, in both KPC4662 and KPC4662-CAF co-implantation subcutaneous models (Fig. 6b-d) with only negligible effects on mouse body weight (Fig. 6e), indicating high efficacy in blocking both intracellular tumor proliferation and intercellular crosstalk with stromal cells of PDAC. Moreover, the prodrug DRP-104, derived from DON, has been demonstrated to effectively suppress PDAC progression in syngeneic models [31].

Fig. 6.

DON potently suppresses tumor growth in vivo. a Luminescence-based cell number measurement of human pancreatic cancer cells under 0.5% O₂, 0.5% FBS, 20 μM LPC (18:1), and 1 μM DON for 72 h. b Tumor volume of KPC4662 syngeneic model treated with vehicle or DON (10 tumors, 5 mice per group). c Tumor volume of KPC4662 + CAF co-implanted syngeneic model treated with vehicle or DON (10 tumors, 5 mice per group). d Endpoint tumor weight in subcutaneous models. e Mouse body weight with indicated treatments. f Schematic overview of how GLS inhibition disrupts CAF-PDAC crosstalk by reactivating ER stress and increasing ROS in PDAC cells. Error bar a represents SD of three experimental replicates. Tumor growth data (mean ± SEM, 10 tumors from 5 mice) analyzed by two-way ANOVA. All other P values were calculated using a two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. ns, not significant

We previously showed that under hypoxic and nutrient-limited conditions in pancreatic cancer, CAFs serve as key sources of unsaturated lipids to restore lipid homeostasis in cancer cells based on multiple in vitro co-culture assays and extensive lipidomics. CAFs secrete unsaturated LPCs, which are utilized by hypoxic pancreatic cancer cells to alleviate membrane lipid saturation and prevent ER stress, specifically inhibiting IRE1α/XBP1 activation. A high-throughput screen identified key compounds that disrupt LPC-mediated rescue of PDAC cell survival, including the GLS inhibitor CB-839. CB-839 reactivates the ATF4/CHOP pathway by inducing amino acid deficiency, bypassing ER stress reversal by LPCs, to promote cell death. Additionally, CB-839 increases ROS levels by limiting Glu availability, a critical precursor for antioxidant synthesis (Fig. 6f). Although pancreatic cancer exhibits resistance to CB-839 treatment by activating compensatory pathways to replenish Glu, we observed that CB-839 moderately slowed tumor progression in a subcutaneous model with CAF co-implantation. Furthermore, we demonstrated that DON, a potent Gln antagonist, effectively slows PDAC progression in vivo, even when cancer cells are co-implanted with lipid producing CAFs.

Discussion

Lipid homeostasis is essential for cancer cell survival, maintaining membrane integrity and resisting stress signaling activation [17, 18, 53]. However, hypoxia disrupts the synthesis of unsaturated lipids, causing saturated lipid accumulation and elevated ER stress. This compromises membrane fitness, activating sensor proteins within membranes, which can trigger cell death [42, 43]. We previously identified that CAFs secrete unsaturated lipids, particularly LPCs, to support the survival of hypoxic pancreatic cancer cells by alleviating ER membrane saturation stress. Targeting lipid transport between CAFs and PDAC cells seemed like a promising approach to disrupt this protective crosstalk. However, lipid uptake and metabolism are regulated by complex and redundant pathways, posing significant challenges for effective therapeutic targeting. For instance, LPCs (single acyl-chain lipids) might be internalized through various mechanisms as multiple transporters such as the FATP family and CD36 mediate LPC uptake by recognizing acyl-chains. MFSD2A, a reported LPC transporter, is also found on immune cells, complicating selective cancer cell inhibition [54, 55]. Additionally, passive diffusion cannot be excluded as a mode of LPC transfer. Pancreatic cancer cells further adapt to stressful conditions through macropinocytosis [56], enabling them to absorb environmental nutrients, including lipids, to sustain survival. Consequently, inhibiting any single pathway may not fully block LPC uptake or prevent LPC-mediated cell viability maintenance, making lipids quite challenging to target for therapeutic intervention. These limitations motivated us to explore indirect approaches to diminish the role of unsaturated LPCs in supporting cancer cell survival under hypoxia.

We utlized a chemical library including approximately 4,500 compounds to identify drugs capable of inhibiting the pro-survival effect of LPCs on PDAC. The chemical library includes 223 inhibitors targeting various aspects of cell metabolism, such as phospholipases and lipid metabolism regulators like peroxisome proliferator-activated receptors (PPARs) and Liver X receptors (LXRs). The screening strategy involved an initial selection to eliminate cytotoxic compounds, followed by a focused screen for LPC inhibition. However, lipid-related inhibitors did not emerge as standout candidates from this process. Instead, we identified chemicals that indirectly affect the role of LPCs in promoting the survival of pancreatic cancer cells under stress. As such, this drug screening opens alternative avenues for therapeutic intervention.

CB-839, one of the top candidates from our approach, acts as an effective inhibitor of LPC-mediated PDAC cell survival under both SO stress and lipotoxic conditions. As a GLS inhibitor, CB-839 blocks Gln utilization in pancreatic cancer cells, activating amino acid deprivation sensors and inducing stress responses, particularly through ATF4/CHOP. The upregulation of CHOP triggers cellular apoptosis which overcomes the suppression of the IRE1α/XBP1 pathway by unsaturated LPCs. We further investigated the role of Glu, particularly its role as a precursor for GSH synthesis, under lipotoxic stress. In CB-839-treated PDAC cells, we observed elevated ROS levels, corresponding with reduced cell viability. Cell death was prevented by the ROS scavenger NAC, indicating that Gln metabolism is crucial for cancer cell survival under stress, taking precedence over the supportive role of unsaturated LPCs in restoring lipid homeostasis. However, the in vivo efficacy of CB-839 is less pronounced than in cell culture, likely due to the development of compensatory metabolic pathways in pancreatic cancer cells, such as Glu replenishment from Gln-independent sources [24, 25]. In our study, CB-839 treatment showed mild tumor growth inhibition in mice co-implanted with murine PDAC cells and CAFs, suggesting GLS pathway inhibition overcomes PDAC-CAF crosstalk to support tumor growth.

Although DON, a Gln antagonist that broadly inhibits Gln-utilizing reactions [27, 50, 51], was not included in the screening drug library, we also tested it against LPC-mediated rescue of PDAC tumor growth [27, 50, 51]. Unlike CB-839, which specifically targets glutaminase, DON has broader activity and therefore represents a more potent Gln metabolism inhibitor independent of the PDAC TME. DON significantly suppressed tumor growth in both KPC4662 and KPC4662-CAF co-implanted groups. However, prior clinical application of DON was impeded by dose-limiting GI toxicities [27, 31, 57]. A newly designed DON peptide prodrug (DRP-104) effectively decreased Gln flux and induced complete tumor regression with improved GI tolerability [27, 31]. DRP-104 stands as a pioneering prodrug with selective metabolism in target tissues versus toxicity-prone tissues. In recent years, DRP-104 has been undergoing clinical trials in adults with advanced solid tumors under FDA Fast Track designation [52].

In conclusion, our chemical screening has demonstrated that the GLS inhibitor CB-839 effectively suppresses LPC-mediated survival of PDAC cells in vitro. This suppression likely stems from the reactivation of ATF4/CHOP stress pathways or the accumulation of ROS, which counteracts the protective effects of unsaturated LPCs on cancer cell viability under hypoxic and nutrient-deprived conditions. CB-839 emerges as a promising candidate for treating Gln-dependent cancers. Although its application in pancreatic cancer remains limited, we observed a suppressive effect of CB-839 in the context of fibroblast interaction also demonstrated by DON. Additionally, DON prodrugs hold promise for advancing clinical trials.

Authors’ contributions

X.H. and M.C.S. conceived the project and designed the experiments; M.C.S. supervised the overall study. X.H., L.C.K., N.P.L., X.C.,and V.L. performed the experiments. X.H. and M.C.S. wrote the paper. All authors revised and approved the final version of the manuscript.

Funding

This work was supported by R01CA276512 (M.C.S.), R35CA220483 (M.C.S.), National Cancer Institute (NCI) 5T32CA009140 (L.C.K. and N.P.L.), American Cancer Society Postdoctoral Fellowship PF-23-1034739-01-TBE (L.C.K.), and Damon Runyon postdoctoral fellowship DRG2497 (N.P.L.).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.