GOT1 Inhibition Induces Extracellular Matrix Remodeling in Pancreatic Cancer

TL;DR

Inhibiting GOT1 in pancreatic cancer changes the tumor environment by altering the extracellular matrix and reducing cancer cell survival.

Contribution

The study reveals how GOT1 inhibition affects the tumor microenvironment through matrix remodeling and metabolic crosstalk.

Findings

GOT1 inhibition alters matrix organization by upregulating matrix-related proteins.

Stromal cells upregulate metabolic programs like glutamine metabolism and oxidative phosphorylation.

Cell death occurs under tissue-like conditions but not with cytotoxic drugs alone.

Abstract

Pancreatic cancer cells rely on glutamine to sustain their survival in the stiff and poorly vascularized tumor microenvironment (TME). Inhibiting glutamic‐oxaloacetic transaminase 1 (GOT1) is a strategy to target glutamine metabolism and impair cancer cell functions. However, it remains unclear how cellular and extracellular elements of the TME respond to GOT1 inhibition. We engineered a pancreatic TME model ‘on a dish’ and recreated the metabolic interactions. Stromal cells remodeled the extracellular matrix and upregulated metabolic programs, including glutamine metabolism, oxidative phosphorylation, and central carbon metabolism. Cell responses to GOT1 inhibition were modulated by TME elements, with reductions in cell viability and proliferation occurring only under tissue‐like conditions. GOT1 inhibition altered matrix organization by upregulating different matrix‐related proteins, while it did not enhance cell responses to cytotoxic drugs. Our findings uncover the metabolic crosstalk within the TME and show that metabolism‐targeting treatments directly impact stromal elements of pancreatic cancer.

Using tumor tissue engineering, we recreated pancreatic cancer and found that inhibiting glutamic‐oxaloacetic transaminase 1 (GOT1) induces extracellular matrix remodeling and secretome rewiring, as well as promotes cell death.

Figures

Proteomic profiling of multicellular 3D cultures. (a) Schematic of our pancreatic TME model, containing human PDAC cells, cancer‐associated fibroblasts and peripheral blood mononuclear cells embedded within GelMA‐HAMA hydrogels. (b) Cell metabolic activity and (c) DNA content of multicellular 3D cultures using different PDAC cells. n = 3, = p ≤ 0.01, = p ≤ 0.0001. (d) Immunostaining of multicellular 3D cultures depicting the presence of cancer cells (KRT‐19), cancer‐associated fibroblasts (αSMA), and macrophages (CD68). (e) Gene ontology enrichment analysis, (f) annotated matrisome, and (g) volcano plot showing significantly up/downregulated matrisome proteins in multi‐ compared to monocellular 3D cultures. Blue = upregulated, red = downregulated. FDR cut‐off = 0.05, log2(fold‐change) ≥ 2, p‐value ≤ 0.05. (h) Immunostaining of patient‐derived PDAC tissues (scale bars = 200 µm) and multicellular 3D cultures (scale bars = 30 µm) illustrating the expression of types I and VI collagen. TME, tumor microenvironment. PDAC, pancreatic ductal adenocarcinoma. GelMA, gelatin methacrylate. HAMA, hyaluronic acid methacrylate. KRT‐19, cytokeratin‐19. αSMA, alpha‐smooth muscle actin.

Metabolic profiling of multicellular 3D cultures. (a) Metabolic pathway analysis and (b) heatmap of amino acid metabolism microarray of multi‐ compared to monocellular 3D cultures grown in GelMA‐HAMA hydrogels. FDR cut‐off = 0.05, p‐value ≤ 0.05. (c) Principal component analysis of the metabolome of mono‐ and multicellular 3D cultures. (d) FEELA pathway analysis indicating up/downregulated metabolic pathways, and (e) relative quantification of metabolites associated with central carbon metabolism in multi‐ versus monocellular 3D cultures. GelMA, gelatin methacrylate. HAMA, hyaluronic acid methacrylate.

Effects of GOT1 inhibition in the pancreatic TME. (a) Schematic of glutamine metabolism. (b) Determination of half‐maximal inhibitory concentration (IC50) of the GOT1 inhibitor using mono‐ and multicellular 3D cultures. (c) Cell viability of mono‐ and multicellular 3D cultures. n = 3, = p ≤ 0.0001, (d) live/dead staining (scale bar = 100 µm), and (e) relative metabolite quantification following GOT1 inhibition. (f) Gene ontology enrichment analysis, and (g) volcano plot showing the significantly up/downregulated matrisome proteins of treated versus non‐treated multicellular 3D cultures. Blue = upregulated, red = downregulated. FDR cut‐off = 0.05, log2(fold‐change) ≥ 2. (h) Immunoblot of multicellular 3D cultures upon GOT1 inhibition. (i) Cell metabolic activity for PDAC cells cultured in type III collagen. n = 6, * = p ≤ 0.05, = p ≤ 0.01, * = p ≤ 0.001. (j) Cytokine profiling of multicellular 3D cultures following GOT1 inhibition. GOT1, glutamic‐oxaloacetic transaminase 1. TME, tumor microenvironment. PDAC, pancreatic ductal adenocarcinoma.

Analyses of cell responses following combined GOT1 inhibition and treatment with cytotoxic drugs. (a) Schematic of experimental rationale. (b) Cell viability of mono‐ and multicellular 3D cultures grown in GelMA‐HAMA hydrogels following treatment with GOT1 inhibitor ± gemcitabine and nab‐paclitaxel. n = 3, * = p ≤ 0.05, * = p ≤ 0.001. (c) Gene ontology enrichment analysis, (d) signaling pathway analysis, and (e) volcano plot showing significantly up/downregulated matrisome proteins of treated multicellular 3D cultures compared to non‐treated cells grown in GelMA‐HAMA hydrogels. Blue = upregulated, red = downregulated. (f) Relative metabolite quantification in multicellular 3D cultures following treatment with GOT1 inhibitor ± gemcitabine and nab‐paclitaxel. FDR cut‐off = 0.05, log2(fold‐change) ≥ 2. GOT1, glutamic‐oxaloacetic transaminase 1. GelMA, gelatin methacrylate. HAMA, hyaluronic acid methacrylate.

Multicellular 3D cultures including patient‐derived PDAC cells and their response toward GOT1 inhibition. (a) Micrograph of multicellular 3D cultures grown in GelMA‐HAMA hydrogels (scale bar = 100 µm). (b) Cell metabolic activity and (c) DNA content of mono‐ and multicellular 3D cultures including patient‐derived PDAC cells in GelMA‐HAMA hydrogels. (d) Cell viability, and (e) DNA content of mono‐ and multicellular TKCC‐22‐LO cultures following treatment with GOT1 inhibitor ± gemcitabine and nab‐paclitaxel. n = 3, * = p ≤ 0.05, = p ≤ 0.01, * = p ≤ 0.001, = p ≤ 0.0001. (f) Live/dead staining (scale bar = 100 µm), and (g) heatmap of amino acid metabolism microarray of multicellular 3D cultures treated with the GOT1 inhibitor compared to non‐treated cells. PDAC, pancreatic ductal adenocarcinoma. GOT1, glutamic‐oxaloacetic transaminase 1. GelMA, gelatin methacrylate. HAMA, hyaluronic acid methacrylate.

Full text

Introduction

1

Glutamine is a central nutrient for pancreatic ductal adenocarcinoma (PDAC) and supports bioenergetic and biosynthetic programs required for tumor growth and metastasis [1]. This amino acid is converted into glutamate and other metabolites, serving as a key source of carbon and nitrogen to fuel the tricarboxylic acid (TCA) cycle. This process generates glutathione and reduced nicotinamide adenine dinucleotide phosphate (NADPH) to maintain the intracellular redox balance [2, 3]. The increased glutamine uptake and anaplerosis (generation of intermediates to replenish the TCA cycle) by PDAC cells are driven by KRAS mutations, an oncogenic activation present in the majority of patients [4]. KRAS‐associated rewiring of glutamine metabolism includes the upregulation of glutamic‐oxaloacetic transaminase 1 (GOT1) [5]. GOT1 catalyzes the conversion of aspartate into oxaloacetate, which leads to increased NADPH and ultimately reduces the levels of reactive oxygen species [6]. GOT1 inhibition induces PDAC cell death in vitro and reduces tumor growth in vivo, without impacting normal pancreatic cells, such as stellate and exocrine cells [5, 6]. In addition, GOT1 inhibition has no impact on other cancer cells (e.g. colorectal cancer), underscoring its PDAC‐specific action [7]. Loss of its mitochondrial counterpart, GOT2, also impairs PDAC cell proliferation [8]. Despite the promising findings on modulating glutamine metabolism in PDAC and the development of inhibitory compounds, blocking GOT1 and GOT2 has not yet been translated into improved clinical treatments [9]. Instead, recent studies have explored alternative strategies to target glutamine metabolism by using glutamine antagonists to impair glutaminolysis and prevent cancer progression [10, 11].

One of the reasons for the slow progress in targeting glutamine metabolism and other metabolic programs is the lack of a full understanding of the metabolic crosstalk within the tumor microenvironment (TME) [12]. In pancreatic tumors, cancer cells are surrounded by a fibrotic extracellular matrix (ECM) and multiple cell types, including cancer‐associated fibroblasts (CAFs) and immune cells [13]. The complex interactions between the cellular and extracellular elements of the TME shape many aspects of tumor biology, including metabolic rewiring [14]. For example, the abundant concentration of crosslinked ECM molecules, mainly collagen isotypes and hyaluronic acid, creates a stiff landscape which alters mitochondrial activity and energy generation in PDAC cells [15, 16]. Likewise, the ratio of cleaved to intact type I collagen fibers in the TME modulates macropinocytosis (endocytosis of large extracellular volumes) and mitochondrial biogenesis in PDAC cells, whereas hyaluronic acid may be a substrate to fuel their growth [17, 18]. Regarding the cellular elements, glutamine deprivation induces macropinocytosis in CAFs to sustain their cell fitness and support the survival of PDAC cells [19]. Induced by the protein Netrin G1, CAFs also release glutamine and glutamate in the TME, which prevents PDAC cell death under nutrient‐deficient conditions [20]. PDAC cell‐derived lactate drives the pro‐tumorigenic polarization of macrophages, whereas the glycolytic reprogramming of neutrophils mediates the proliferation of cancer cells and helps suppress T cells [21, 22]. Recreating the pancreatic TME and its metabolic crosstalk advances our knowledge of the role of cell‐cell and cell‐matrix interactions in shaping cancer metabolism, identifying targetable metabolic vulnerabilities, and cell responses to metabolism‐based treatments [23].

Here, we developed a tissue‐engineered 3D platform and combined proteomics and metabolomics to investigate the linkage between TME elements and glutamine metabolism in PDAC. Our pancreatic TME model closely resembled key cellular and extracellular features of human PDAC tissues, revealing that CAFs and immune cells remodel the ECM, while upregulating glutamine metabolism, oxidative phosphorylation (OXPHOS), and central carbon metabolism. We observed that treatment with a GOT1 inhibitor reduced cell viability and proliferation of multicellular 3D cultures by 50% and altered the ECM by upregulating collagen isotypes and other matrix‐related proteins. GOT1 inhibition reduced cell viability to similar levels as those with treatment using gemcitabine and nab‐paclitaxel, without enhancing cell responses when combined to these drugs. Screening of patient‐derived cells validated the effect of GOT1 inhibition on PDAC cells and glutamine metabolism. Our work highlights the importance of considering both cellular and extracellular elements of the TME for the development and adequate testing of metabolism‐targeting therapies.

Materials and Methods

2

PDAC Cells

2.1

Human PDAC cell lines BxPC3, MIA PaCa‐2, and PANC‐1 were obtained from the European Collection of Authenticated Cell Cultures. Patient‐derived cells were established and characterized at The Kinghorn Cancer Centre (TKCC) as described elsewhere (Table S1) [24], and provided by the Australian Pancreatic Cancer Genome Initiative. PDAC cells were grown as adherent monolayers using Dulbecco's Modified Eagle Medium (DMEM, Gibco 10569010) supplemented with 10% fetal bovine serum (FBS, Gibco 26140079) and 1% penicillin‐streptomycin (P/S, Gibco 15140122). TKCC‐19‐LO, ‐22‐LO, and ‐25‐LO cells were grown using Iscove's Modified Dulbecco's Medium (IMDM, Gibco 12440053) supplemented with 20 ng/mL human recombinant epithelial growth factor (hEGF, Invitrogen PHG0313), 12.5 µg/mL human apo‐transferrin (Sigma‐Aldrich T1147), 0.2 IU/mL Actrapid Penfill human insulin (Novo Nordisk), 0.5x MEM vitamin solution (Gibco 11120052), 20% FBS, and 1% P/S. TKCC‐06 cells were grown using DMEM/F12 (Gibco 11320033) supplemented with 15 mM HEPES (Gibco 15630080), 10 ng/mL hEGF, 0.1 IU/mL insulin, 40 ng/mL hydrocortisone (Sigma‐Aldrich H0888), 0.12% glucose (Gibco A2494001), and 7.5% FBS. All cells were routinely tested for mycoplasma.

Isolation of Patient‐Derived CAFs

2.2

CAFs were isolated from two different patient‐derived PDAC tissues obtained from the Monash Medical Centre with informed patient consent (Monash University Human Research Ethics Committee, ID 35707). Tissue samples were sectioned using a sterile scalpel and incubated in 5 mL of Roswell Park Memorial Institute (RPMI) 1640 Medium (Gibco 22400089) supplemented with 5 mg/mL collagenase V (Sigma Aldrich C9263) at 37°C, 5% CO_2_ for 45 min. Samples were centrifuged at 1,500 rpm for 5 min, and cell pellets were resuspended in DMEM/F12 (Gibco 10565042) supplemented with 10% FBS and 1% P/S. Cells were grown until reaching 80% confluence and passaged using TrypLE Express Enzyme (Gibco 12605010). Immunocytochemistry was performed as described below to assess the expression of alpha‐smooth muscle actin (αSMA) and type I collagen (COL1), two characteristic markers of myofibroblastic cancer‐associated fibroblasts (myCAFs) (Figure S1a). Following the first passage, CAFs were grown using DMEM supplemented with 10% FBS and 1% P/S.

3D cell Cultures

2.3

GelMA hydrogels were synthesized as previously described [25]. GelMA precursor solution was lyophilized for 7 days and stored at ‐20°C until usage. Hydrogel solutions were prepared by dissolving GelMA powder at 5% (wt/vol) and HAMA (PhotoHA Stiff, Cellink 5212) at 0.1% in PBS containing 0.15% lithium phenyl‐2,4,6‐trimethylbenzoylphosphinate (LAP, Sigma‐Aldrich 900889) for 1 h at 37°C. Matrigel (Corning 356231) was thawed on ice and kept at 4°C until usage. Human peripheral blood mononuclear cells (PBMCs, Stemcell 700251) were thawed one day prior to the experiment using RPMI medium supplemented with 10% FBS and 1% P/S. Expression of characteristic immune cell markers (e.g. macrophages—CD80, CD163; T cells – CD3) was assessed by qPCR as described below (Figure S1b). PDAC cells (2.1 × 10^5^ cells/mL) were resuspended in hydrogel solutions as mono‐ or multicellular 3D cultures including CAFs (4.2 × 10^5^ cells/mL) and PBMCs (4.2 × 10^5^ cells/mL). Cell‐containing hydrogel solutions were dispensed as droplets of 37.5 µL on a Sigmacote (Sigma SL2)‐treated glass slide and crosslinked at 405 nm for 4 min. Hydrogel constructs were transferred to non‐treated 48‐well plates containing 500 µL of PDAC or TKKC cell medium. For glucose studies, cell‐laden hydrogels were cultured using 500 µL of glucose‐free DMEM medium (Gibco 11966025). For indirect multicellular 3D cultures, PDAC and stromal cells (CAFs and PBMCs) were seeded in separate hydrogels, and one cell‐laden construct of each 3D culture condition was transferred to non‐treated 24‐well plates containing PDAC cell medium. For Matrigel, 37.5 µL of cell‐containing matrix was dispensed directly into non‐treated 48‐well plates and incubated at 37°C, 5% CO_2_ for 15 min for polymerization. Mono‐ and multicellular 3D cultures in either matrix were grown for 14 days, and media was renewed on days 7 and 11. For 3D cultures in type III collagen hydrogels, 50 µL of collagen solution (Cellink 5019) at 0.5 and 1 mg/mL were polymerised in non‐treated 96‐well plates for 4 h. Then, 5 × 10^3^ PANC‐1 cells were seeded on top of the collagen matrix and cultured using 200 µL of PDAC cell medium for 7 days. 3D cell cultures were observed by brightfield microscopy using an EVOS 5000 (Thermo Fisher Scientific).

Cell Metabolic Activity and Proliferation Assays

2.4

Cell metabolic activity and proliferation were quantified using Prestoblue (Invitrogen A13261) and CyQUANT (Invitrogen C7026) assays, respectively, as described elsewhere [26]. Prestoblue assays rely on the intracellular reduction of resazurin to resorufin by metabolically active cells, whereas CyQUANT assays measure total DNA content as an indicator of cell proliferation [26]. Following 1 and 14 days of 3D culture, cell‐laden hydrogel or Matrigel samples were transferred to a 48‐well plate containing 200 µL of 10% Prestoblue solution in phenol‐red free DMEM (Gibco 21063029) and incubated at 37°C, 5% CO_2_ for 45 min. For collagen hydrogels, cell metabolic activity was measured following 1, 4, and 7 days of 3D culture. Cell‐free samples were used as negative controls. A volume of 90 µL was pipetted in duplicates in a black, clear bottom 96‐well plate (PerkinElmer 6005182), and fluorescence signals were measured using a VICTOR Nivo multimode microplate reader (PerkinElmer). Cell‐laden samples were washed with PBS and frozen at −20°C for at least 24 h prior to CyQUANT assays. Samples were incubated with 0.5 mg/mL proteinase K (Gibco 25530015) in phosphate‐buffered EDTA at 65°C for 16 h, and treated with 40 µg/mL RNase A (Invitrogen 12091021) at room temperature for 1 h. Samples were subjected to CyQUANT reagent dye for 5 min protected from light, and fluorescence signals measured using a microplate reader.

Drug Treatment

2.5

Mono‐ and multicellular 3D cultures were established as described above and grown for 7 days and then treated with 25 µM GOT1 inhibitor (MedchemExpress HY‐122723) alone or combined with 100 nM gemcitabine (Sigma‐Aldrich G6423) and 100 nM nab‐paclitaxel (Bristol‐Myers Squibb), which were dissolved directly into PDAC or TKCC cell media. Treatment was renewed every 48 h over 7 days of 3D culture. For combination treatments, cells were first treated with GOT1 inhibitor on days 7 and 9, followed by gemcitabine and nab‐paclitaxel on days 11 and 13. Cell viability was quantified by Prestoblue assay and normalized to the vehicle control group.

Western Blot

2.6

Cells were recovered from hydrogel constructs or 6‐well microplates and lysed in RIPA buffer (Thermo Fisher Scientific 89900) supplemented with protease inhibitor cocktail (Roche 5892970001). Protein concentration of samples was quantified using the BCA Protein Assay Kit (Thermo Fisher Scientific 23227), and 15 µg of protein lysate per sample was resolved on SDS‐PAGE gels and transferred to a PVDF membrane (Millipore, IPVH00010). Membranes were blocked in 5% nonfat dry milk dissolved in Tris‐buffered saline (TBS) buffer and incubated with primary antibodies (Table S2). Membranes were washed with TBS‐Tween, incubated with secondary antibodies (Table S2), and protein bands were visualized using an Alliance Q9 mini system (Uvitec).

Immunocytochemistry

2.7

Cell‐laden samples and/or cell monolayers were fixed in 4% paraformaldehyde overnight, washed with 0.1 M glycine (Sigma‐Aldrich 50046) in PBS, and permeabilized with 0.3% Triton‐X 100 (Sigma‐Aldrich T9284). Samples were blocked with 10% FBS in PBS for 2 h, and incubated with primary antibodies (Table S2) diluted in 5% FBS/PBS at 4°C overnight. After washing with PBS, samples were incubated with secondary antibodies (Table S2) diluted in 5% FBS/PBS followed by 0.8 U/mL rhodamine‐conjugated phalloidin (Invitrogen R415) diluted in PBS. Samples were counterstained with 1 µg/mL DAPI (Invitrogen D1306) diluted in PBS at room temperature for 30 min, protected from light. Immunofluorescent images were acquired using a FV3000 confocal laser scanning microscope (Olympus).

PCR Array and qPCR

2.8

Mono‐ and multicellular 3D cultures were treated with 1 mg/mL collagenase I solution (Sigma‐Aldrich C9407), and cells were washed with PBS. RNA was extracted using a RNeasy Micro Kit (QIAGEN 74104) as per the manufacturer's instructions. RNA samples were quantified by spectrophotometry (DeNovix DS‐11 FX), and 400 ng of each sample was converted into first‐strand cDNA using the RT2 First Strand Kit (QIAGEN 330401). Then, cDNA samples were mixed with RT2 SYBR Green mastermix (QIAGEN 330501) and subjected to a 384‐well RT2 Profiler PCR array for human amino acids (QIAGEN 330231 PAHS‐130Z). Real‐time PCR was performed using a Quantstudio Pro 6 system (Applied Biosystems). The cycle threshold (Ct) of each gene was determined and normalized to the average expression level of the housekeeping genes (ACTB, B2M, GAPDH, HPRT1, RPLP0) and compared to the control (ΔΔCt). Relative gene expression between groups was calculated using the 2^−ΔΔCt^ method. For qPCR, PBMCs and PANC‐1 cells were pelleted and resuspended in 1 mL of TRI reagent (Sigma T9424), and RNA was isolated as per the manufacturer's instructions. Then, 2 µg of RNA was used to synthesize cDNA using the High‐capacity cDNA RT kit (Applied Biosystems 4368814), diluted in a 1:25 ratio for quantitative PCR (qPCR) analyses using a CFX96 Touch Real‐Time PCR Detection System, with the target forward and reverse primers of target genes (Table S3).

Cytokine and Chemokine Profiling

2.9

Multicellular 3D cultures were treated with the GOT1 inhibitor as described above, and cell culture medium was collected on day 14. Relative changes in cytokine and chemokine secretion to non‐treated cells were measured using the Human Cytokine Array C5 (RayBiotech, AAH‐CYT‐5‐4) as per the manufacturer's instructions. Briefly, membranes were blocked for 30 min at room temperature and incubated with 1 mL of cell culture medium overnight. Membranes were washed with buffers I and II and incubated with the biotinylated antibody cocktail and HRP‐conjugated streptavidin. Signals were visualized using an Alliance Q9 mini system (Uvitec) and quantified by densitometry using ImageJ.

Statistical Analysis

2.10

Differences were assessed using one‐way ANOVA with Brown‐Forsythe and Welch's corrections for ungrouped datasets, and two‐way ANOVA with Tukey's, Bonferroni's, or Šídák's multiple comparison post hoc tests for grouped data, using GraphPad Prism. Results with p‐values below 0.05 were considered statistically significant (* = p ≤ 0.05; ** = p ≤ 0.01; *** = p ≤ 0.001; **** = p ≤ 0.0001). Only statistically significant p‐values are shown.

Further information on methodology is available in the supplementary file.

Results

3

Stromal Cells Shape the Extracellular Matrix and Metabolic Processes

3.1

To recreate the pancreatic TME, we cultured human PDAC cells together with patient‐derived CAFs and PBMCs encapsulated in gelatin methacrylate‐hyaluronic acid methacrylate (GelMA‐HAMA) hydrogels (Figure 1a). These cell populations represent the most abundant cellular components of PDAC tissues, whereas GelMA‐HAMA models collagen and hyaluronic acid, two key extracellular molecules in the pancreatic TME [16, 27]. Multicellular 3D cultures of MIA PaCa‐2 cells showed the highest levels of metabolic activity and DNA content, indicative of cell viability and proliferation, followed by PANC‐1 and BxPC3 cells (Figure 1b,c). In the absence of CAFs and PBMCs, monocellular BxPC3 cultures had a lower metabolic activity and DNA content, whereas PANC‐1 and MIA PaCa‐2 cells had the highest metabolic activity and proliferation, respectively (Figure S2a,b). Given that MIA PaCa‐2 cells have a mesenchymal phenotype, and BxPC3 cells had the lowest cell growth and lack a KRAS mutation [28], we selected PANC‐1 cells for subsequent studies. PANC‐1 cells formed spheroids and sustained the expression of the cell proliferative marker Ki67 in 3D culture (Figure S2c,d). To mechanically characterize our pancreatic TME model, we measured the local and bulk stiffness using atomic force microscopy and indentation testing. GelMA‐HAMA hydrogels had a local stiffness of 3.42 ± 0.05 kPa and bulk stiffness of 16.7 ± 0.67 kPa, representing the biomechanical properties of human PDAC tissues [29]. The addition of HAMA was essential to elevate the stiffness, given that GelMA‐only hydrogels had a local and bulk stiffness of 0.52 ± 0.09 kPa and 6.1 ± 0.69 kPa, respectively. GelMA‐based matrices strongly contrast to Matrigel, a matrix commonly used for 3D cultures, which had a local stiffness of 0.04 kPa and a non‐measurable bulk stiffness due to its liquid‐like behavior (Figure S2e,f). We characterized the different cell populations in our pancreatic TME model by immunostaining of cytokeratin‐19, α‐smooth muscle actin and CD68 (Figure 1d). PDAC cells grew as spheroids surrounded by CAFs and immune cells, resembling the cell organization observed in patient‐derived tissue samples where pancreatic ducts are embedded within stromal cells (Figure S2g). Spheroids had a diameter of 100.55 ± 34.41 µm when co‐cultured with stromal cells and a diameter of 121.09 ± 42.11 µm when cultured alone. PDAC cells cultured in GelMA‐only hydrogels showed similar morphologies to our pancreatic TME model, growing as spheroids of 100 ± 25.74 µm in monocellular 3D cultures and 77.52 ± 25.42 µm in multicellular 3D cultures. Cells embedded in Matrigel formed irregular cell aggregates without a defined morphology (Figure S2h,i). Overall, our results demonstrate the suitability of our tumor tissue‐engineered platform to resemble the typical elements of the pancreatic TME.

Proteomic profiling of multicellular 3D cultures. (a) Schematic of our pancreatic TME model, containing human PDAC cells, cancer‐associated fibroblasts and peripheral blood mononuclear cells embedded within GelMA‐HAMA hydrogels. (b) Cell metabolic activity and (c) DNA content of multicellular 3D cultures using different PDAC cells. n = 3, ** = p ≤ 0.01, **** = p ≤ 0.0001. (d) Immunostaining of multicellular 3D cultures depicting the presence of cancer cells (KRT‐19), cancer‐associated fibroblasts (αSMA), and macrophages (CD68). (e) Gene ontology enrichment analysis, (f) annotated matrisome, and (g) volcano plot showing significantly up/downregulated matrisome proteins in multi‐ compared to monocellular 3D cultures. Blue = upregulated, red = downregulated. FDR cut‐off = 0.05, log2(fold‐change) ≥ 2, p‐value ≤ 0.05. (h) Immunostaining of patient‐derived PDAC tissues (scale bars = 200 µm) and multicellular 3D cultures (scale bars = 30 µm) illustrating the expression of types I and VI collagen. TME, tumor microenvironment. PDAC, pancreatic ductal adenocarcinoma. GelMA, gelatin methacrylate. HAMA, hyaluronic acid methacrylate. KRT‐19, cytokeratin‐19. αSMA, alpha‐smooth muscle actin.

To profile our pancreatic TME model, we compared the proteome of multicellular 3D cultures to their monocellular counterparts (Figure S3a, Table S4). The proteome (5,086 identified proteins) was significantly different, having 320 and 305 proteins uniquely expressed by mono‐ and multicellular 3D cultures, respectively, while 4,461 proteins were present in both (Figure S3b). Differential expression analyses revealed that multicellular 3D cultures upregulated 72 proteins (e.g. PTGS2, NNMT, TGM2, CEMIP, MGST1; Table S5). PTGS2 is a pro‐inflammatory enzyme involved in establishing the immunosuppressive TME [30]. CEMIP promotes the degradation of hyaluronic acid and induces pro‐inflammatory processes in solid tumors [31]. In contrast, 29 proteins (e.g. SLC25A22, CDK13, RAB5B, GBP4, PDE5A, SYNM, TRAF1; Table S5) were downregulated. SLC25A22 is a mitochondrial glutamate carrier and is linked to the immunosuppressive TME of KRAS‐mutant cancers [32]. Our results suggest that our pancreatic TME model recapitulates features of the immune landscape in PDAC.

Gene ontology (GO) enrichment analysis revealed that stromal cells upregulated cellular components and molecular functions of processes associated with the ECM, including structural constituents and tensile strength. Stromal cells upregulated proteins associated with prostaglandin, prostanoid, olefinic compound, and unsaturated fatty acid metabolism (Figure 1e; Figure S3c, and Table S6). These upregulated ECM programs attest the key role of stromal cells in remodeling the architecture of the TME, which leads to the stiff and fibrotic appearance of PDAC tissues [16]. The upregulated metabolic processes are linked to sustaining the immunosuppressive features of the pancreatic TME, as previously reported [30, 33, 34]. Further analysis of ECM‐related proteins (referred to as matrisome [35]) revealed an increased expression of factors classified as ECM‐affiliated proteins (e.g. ANXA6), regulators (e.g. LOX, PLOD2, TGM2, TIMP3), glycoproteins (e.g. IGFBP7, PXDN), secreted factors (e.g. GDF15) and collagens (e.g. COL1A1, COL1A2, COL6A1, COL6A2, COL6A3, COL12A1). In contrast, LAMC2 was downregulated, aligning our findings with the proteome of PDAC and stromal cells [36] (Figure 1f,g, Table S5). Indeed, the upregulation of types I and VI collagen in our platform matched the profile seen in patient‐derived PDAC tissues as validated by immunostaining (Figure 1h; Figure S2j).

In parallel, we grew multicellular 3D cultures in GelMA hydrogels and Matrigel and compared their proteome profiles with that of our pancreatic TME model. A total of 40 proteins were commonly upregulated in GelMA‐based cultures, whereas 33 proteins (e.g. MGAT1, CENIH4; Table S5) were uniquely upregulated, 26 proteins (e.g. DGKZ, MAP2K1, RAD1; Table S5) were uniquely downregulated, and only four proteins (GBP4, PDE5A, SYNM, TRAF1; Table S5) were commonly downregulated (Figure S4a,b). To the best of our knowledge, these proteins are not central to PDAC biology. Multicellular 3D cultures in Matrigel had 19 proteins (e.g. CEMIP2, MRPL4, MIPEP; Table S5) uniquely upregulated, and six proteins (e.g. ADO, CDK9, PTPRA; Table S5) uniquely downregulated, with none overlapping with our pancreatic TME model (Figure S4a,b, Table S5). Cells grown in GelMA hydrogels also upregulated biological processes, cellular components and molecular functions associated with the ECM (mainly collagen isotypes), but none related to cell metabolism. Cells grown in Matrigel upregulated only two mitochondrial‐related processes and showed differential expression of matrisome‐related HTRA1 and LAMC1 (Figure S5a–d, Tables S5 and S6). This shows that the hydrogel matrix significantly influences the cell proteome and metabolism. Overall, our results indicate that stromal cells are central in shaping the extracellular and metabolic processes in the pancreatic TME.

Stromal Cells Rewire Glutamine Metabolism in the Pancreatic TME

3.2

We further explored the role of stromal cells in driving the metabolic signature in PDAC by comparing multi‐ to monocellular 3D cultures. Our pancreatic TME model showed an increased expression of 6% of proteins associated with alanine, asparagine, and glutamine metabolism, and a decreased expression of 3% of proteins linked to this metabolic pathway. These amino acids mediate the metabolic crosstalk between PDAC cells and CAFs, promoting tumor growth and metastasis [10, 11]. In addition, the expression of 6% of proteins associated with glycolysis/gluconeogenesis, and 11% of proteins related to the pentose phosphate pathway were decreased, whereas 6% of proteins of OXPHOS were increased (Figure 2a). These findings suggest that, in our pancreatic TME model, stromal cells favor energy production using amino acids as substrates rather than glucose. To validate the modulation of amino acid metabolism, we performed a PCR microarray, which confirmed the up/downregulation of genes related to glutamine metabolism (e.g. GLS, CPS1, ASS1, CAD, ASL, NIT2) in multicellular 3D cultures compared to their monocellular counterparts (Figure 2b). The upregulation of GLS suggests an increase in glutaminase, an enzyme that catalyzes the hydrolysis of glutamine to glutamate, fueling this metabolic pathway [11].

Metabolic profiling of multicellular 3D cultures. (a) Metabolic pathway analysis and (b) heatmap of amino acid metabolism microarray of multi‐ compared to monocellular 3D cultures grown in GelMA‐HAMA hydrogels. FDR cut‐off = 0.05, p‐value ≤ 0.05. (c) Principal component analysis of the metabolome of mono‐ and multicellular 3D cultures. (d) FEELA pathway analysis indicating up/downregulated metabolic pathways, and (e) relative quantification of metabolites associated with central carbon metabolism in multi‐ versus monocellular 3D cultures. GelMA, gelatin methacrylate. HAMA, hyaluronic acid methacrylate.

Next, we performed untargeted metabolomics, revealing a significant difference in the metabolomic profiles of multi‐ and monocellular 3D cultures (Figure 2c, Tables S7 and S8). FEELA pathway enrichment analysis based on the quantified metabolites showed that stromal cells upregulate key metabolic pathways in our pancreatic TME model, including central carbon metabolism, mTOR signaling and pentose phosphate pathways. Both purine, pyrimidine and nucleotide metabolism, and valine, leucine and isoleucine biosynthesis were decreased (Figure 2d, Table S9). Upon relative quantification of metabolites associated with central carbon metabolism in cancer, we found that glutamine, glutamic acid, malic acid, citric acid, pyruvic acid, and succinic acid were more abundant in multi‐ compared to monocellular 3D cultures, whereas aspartic acid and phosphoenolpyruvic acid were decreased (Figure 2e). The metabolome of multicellular 3D cultures grown in GelMA hydrogels and Matrigel was also significantly different to their monocellular counterparts (Figure S6a–d; Table S8), in particular central carbon metabolism, and relative metabolite levels of glutamine, glutamic acid and glucose (Figure S6e). Taken together, our findings demonstrate that stromal cells drive the rewiring of glutamine metabolism and other pathways in the pancreatic TME, and that our platform captures this PDAC‐specific profile.

Modulating Glutamine Metabolism Remodels the Extracellular Matrix and Secretome

3.3

Given the upregulation of glutamine metabolism, we sought to understand how stromal cells and ECM molecules influence cell responses to the modulation of this pathway via GOT1 inhibition. GOT1 is a key enzyme in glutamine metabolism and has emerged as a targetable metabolic vulnerability in PDAC (Figure 3a) [5]. We treated 3D cultures with a GOT1 inhibitor and found its half‐maximal inhibitory concentration (IC_50_) of 20 ± 1.3 µM for monocellular 3D cultures and 25 ± 2.5 µM for its multicellular counterpart (Figure 3b). Given our focus on the multicellular TME, we adopted the IC_50_ of 25 µM for subsequent experiments. Then, we determined the effect of GOT1 inhibition on other PDAC cells and observed that the cell viability of multicellular BxPC3 cultures was reduced to 62.92 ± 8.37%. Monocellular BxPC3 cultures had a cell viability of 6.6 ± 0.63%, the lowest across all cells tested, suggesting an nonspecific effect of GOT1 inhibitor on KRAS‐wildtype cells. Multicellular MIA PaCa‐2 cultures were less responsive to GOT1 inhibition and had a reduced cell viability of 80.16 ± 7.03% (Figure 3c). These differential cell responses suggest that the concentration of GOT1 inhibitor should be adjusted on a cell‐by‐cell basis. Live/dead staining of mono‐ and multicellular PANC‐1 cultures revealed that GOT1 inhibition induced cell death (Figure 3d; Figure S7a). Notably, the matrix composition influenced cell responses to GOT1 inhibition. While GelMA‐based mono‐ and multicellular PANC‐1 cultures sustained a cell viability of 66.00 ± 3.04% and 73.24 ± 8.68%, respectively, upon GOT1 inhibition, cells grown in Matrigel and stromal cell cultures showed no effect (Figure S7b,c).

Effects of GOT1 inhibition in the pancreatic TME. (a) Schematic of glutamine metabolism. (b) Determination of half‐maximal inhibitory concentration (IC50) of the GOT1 inhibitor using mono‐ and multicellular 3D cultures. (c) Cell viability of mono‐ and multicellular 3D cultures. n = 3, **** = p ≤ 0.0001, (d) live/dead staining (scale bar = 100 µm), and (e) relative metabolite quantification following GOT1 inhibition. (f) Gene ontology enrichment analysis, and (g) volcano plot showing the significantly up/downregulated matrisome proteins of treated versus non‐treated multicellular 3D cultures. Blue = upregulated, red = downregulated. FDR cut‐off = 0.05, log2(fold‐change) ≥ 2. (h) Immunoblot of multicellular 3D cultures upon GOT1 inhibition. (i) Cell metabolic activity for PDAC cells cultured in type III collagen. n = 6, * = p ≤ 0.05, ** = p ≤ 0.01, *** = p ≤ 0.001. (j) Cytokine profiling of multicellular 3D cultures following GOT1 inhibition. GOT1, glutamic‐oxaloacetic transaminase 1. TME, tumor microenvironment. PDAC, pancreatic ductal adenocarcinoma.

Next, we profiled the metabolome of mono‐ and multicellular 3D cultures upon GOT1 inhibition. Relative metabolite quantification showed that GOT1 inhibition significantly decreased metabolites associated with glutamine metabolism and OXPHOS in monocellular 3D cultures. Among these metabolites, glutathione, NADH and NADPH are essential to maintain the intracellular redox state, and lead to cell death when unbalanced [10]. Conversely, multicellular 3D cultures sustained higher levels of these metabolites, along with increased aspartic acid, glutamine and NADP (Figure 3e). To delineate these differences, we profiled PDAC cells indirectly co‐cultured with stromal cells in 3D and observed increased levels of all glutamine‐ and OXPHOS‐related metabolites (Figure S7d). Our results suggest that stromal cells modulate cell responses to GOT1 inhibition and supply metabolites to rescue PDAC cells. While cells grown in GelMA hydrogels had a decrease in NAD, NADH and NADPH, cells grown in Matrigel did not show any significant changes, and rather promoted viability of treated cells (Figure S7e).

Given the ability of cancer cells to rewire their metabolism and explore alternative substrates, we assessed cell responses in our pancreatic TME model to GOT1 inhibition in the absence of glucose. Glucose is a central metabolite for cancer cells and provides intermediates to restore inhibited metabolic programs, including OXPHOS [1]. In the absence of glucose, GOT1 inhibition reduced cell viability to 21.77 ± 1.91%, with monocellular 3D cultures reaching a cell viability as low as 11.88 ± 2.66% (Figure S8a). To validate the effects of GOT1 inhibition, we knocked down GOT1 in PANC‐1 cells using siRNA and quantified cell viability 7 days post‐transfection in 3D culture. We found significantly reduced GOT1 levels, and cell viability declined by 25% (Figure S8b,c). Collectively, our results show that cell responses to GOT1 inhibition are mediated by stromal elements of the pancreatic TME.

To decipher the effects of GOT1 inhibition within the pancreatic TME, we analyzed the proteome of our model following GOT1 inhibition. GO enrichment analysis of multicellular 3D cultures post‐treatment revealed the upregulation of multiple biological processes, cellular components, and molecular functions associated with the matrisome (Figure 3f, Table S6). We observed a significant upregulation of collagens (e.g. COL1A1, COL3A1), ECM glycoproteins (e.g. FBN1, FN1, IGFBP7, PCOLCE, SPARC, THBS2), ECM regulators (e.g. CST3, LOX, SULF1, TIMP3) and a secreted factor (GDF15), and downregulated VW5A5 (Figure 3g, Table S5). These proteins are linked to the structure and organization of the ECM, indicating its remodeling in our platform. In contrast, monocellular 3D cultures showed the upregulation of a different collagen isotype (COL18A1), ECM glycoproteins (e.g. LTBP1, LTBP2) and downregulated only LAMC2 (Figure S9a; Table S5). These results suggest that GOT1 inhibition has a differential impact on the matrisome of PDAC and stromal cells, and alters the composition and organization of the ECM by mainly modulating stromal cells. We validated the significant upregulation of COL3A1 in our pancreatic TME model (Figure 3h; Figure S9b), and assessed the effects of COL3A1 on PDAC cells. PANC‐1 cells cultured in type III collagen gels (0.5 and 1 mg/mL) exhibited reduced cell metabolic activity, suggesting that this ECM protein may inhibit cell growth (Figure 3i). In addition, our proteomic data showed that GOT1 inhibition did not alter the expression of markers linked to invasion and migration (e.g. CDH2, CTNNB1, VIM; Table S5), indicating that GOT1 inhibition did not shift PDAC cells toward a more mesenchymal phenotype at the protein level. Last, we determined the effect of GOT1 inhibition on the secretome by assessing the levels of important cytokines and chemokines in our pancreatic TME model. We found that IL‐6 and CCL2 were increased, while CCL5 and CXCL1 were decreased (Figure 3j). IL‐6 and CCL2 are immunosuppressive factors that help recruit and differentiate monocytes and impair the function of T cells [37, 38]. CCL5 has been linked to macrophage‐derived drug resistance and reduced T cell infiltration, whereas CXCL1 is implicated in neutrophil recruitment [39, 40]. Further analyses of IL‐6 levels confirmed a 2.3‐fold increase in this cytokine following treatment with the GOT1 inhibitor (Figure S9c). These results suggest that GOT1 inhibition may reshape the immune compartment of the pancreatic TME and alter the function of different immune cell types.

Importantly, the results observed in our pancreatic TME model strongly contrast the effect of GOT1 inhibition on multicellular 3D cultures grown in GelMA hydrogels and Matrigel. While ECM glycoproteins (e.g. FBLN1, PCOLCE, THBS2), ECM regulators (e.g. MMP2, TIMP3), collagens (e.g. COL1A1, COL1A2, COL5A2, COL6A1), and proteoglycans (e.g. BGN, VCAN) were upregulated in cells grown in GelMA hydrogels, the expression of matrisome‐related proteins was not affected in cells grown in Matrigel upon GOT1 inhibition (Figure S9d,f, Table S5). Monocellular 3D cultures in GelMA and Matrigel also showed a contrasting matrisome profile compared to their multicellular counterparts (Figure S9e,g). Our findings indicate that cell responses to GOT1 inhibition are a function of TME elements.

GOT1 Inhibition Did Not Enhance the Effects of Cytotoxic Drugs

3.4

Based on the cell responses in our pancreatic TME model to GOT1 inhibition, we sought to explore the effects of combining GOT1 inhibition with the first‐line chemotherapeutics gemcitabine and nab‐paclitaxel (Gem/PTX; 100 nM; Figure 4a). The viability of multicellular 3D cultures was reduced to 30.12 ± 9.95% following exposure to cytotoxic drugs alone, while in combination with GOT1 inhibition, cell viability decreased to 15.07 ± 1.61%. Monocellular 3D cultures responded similarly to both treatment regimens, without significant differences between the treatment with cytotoxic drugs alone and the combination with GOT1 inhibition (Figure 4b). In contrast, multicellular 3D cultures grown in GelMA hydrogels and Matrigel were more resistant to the combined treatment, maintaining cell viabilities of 35.56 ± 3.81% and 73.78 ± 23.31%, respectively (Figure S10a,b).

Analyses of cell responses following combined GOT1 inhibition and treatment with cytotoxic drugs. (a) Schematic of experimental rationale. (b) Cell viability of mono‐ and multicellular 3D cultures grown in GelMA‐HAMA hydrogels following treatment with GOT1 inhibitor ± gemcitabine and nab‐paclitaxel. n = 3, * = p ≤ 0.05, *** = p ≤ 0.001. (c) Gene ontology enrichment analysis, (d) signaling pathway analysis, and (e) volcano plot showing significantly up/downregulated matrisome proteins of treated multicellular 3D cultures compared to non‐treated cells grown in GelMA‐HAMA hydrogels. Blue = upregulated, red = downregulated. (f) Relative metabolite quantification in multicellular 3D cultures following treatment with GOT1 inhibitor ± gemcitabine and nab‐paclitaxel. FDR cut‐off = 0.05, log2(fold‐change) ≥ 2. GOT1, glutamic‐oxaloacetic transaminase 1. GelMA, gelatin methacrylate. HAMA, hyaluronic acid methacrylate.

To further characterize the effects of combined treatments in our pancreatic TME model, we performed proteome and GO enrichment analyses. Among the significantly upregulated cellular programs, we observed changes in both extra‐ and intracellular components, protein binding and folding, and metabolic processes (Figure 4c, Table S6). We found increased expression of proteins associated with DNA replication, nucleotide excision repair, and cell cycle, and those related to PPAR signaling, while ECM‐receptor interactions were downregulated (Figure 4d). In contrast to GOT1 inhibition alone, its combination with gemcitabine and nab‐paclitaxel did not significantly alter ECM‐associated programs. Despite this, analysis of matrisome‐related proteins showed that the combined treatment led to increased COL3A1, COL1A1, COL12A1, FN1, THBS2, and TIMP3 expression, among others (Figure 4e, Table S5). GO enrichment analysis of multicellular 3D cultures grown in GelMA hydrogels and Matrigel revealed upregulated cellular programs similar to those of cells grown in GelMA‐HAMA hydrogels following combined treatment (Figure S10c,d), with increased matrisome‐related proteins (e.g. COL5A2, COLA1, COL1A2) in GelMA hydrogels and (e.g. COL6A3, COL6A1, COL4A1, COL4A2, COL18A1) Matrigel (Figure S10e,f, Table S5). Combined treatments reduced all glutamine‐ and OXPHOS‐related metabolites in multicellular 3D cultures (Figure 4f, Table S8). Overall, our results indicate that despite the absence of a synergistic effect of GOT1 inhibition in combination with cytotoxic drugs, the combination treatment differentially impacts elements in the pancreatic TME.

Patient‐derived PDAC Cells Have Differential Responses Toward GOT1 Inhibition

3.5

As a proof‐of‐concept we cultured four different patient‐derived PDAC cells together with stromal cells in GelMA‐HAMA hydrogels. PDAC cells formed organoid‐like structures within 1–7 days of 3D culture (Figure 5a). Among these patient‐derived cells, three (TKCC‐06, undifferentiated; ‐22‐LO, poorly differentiated and ‐25‐LO, undefined) were obtained from primary PDAC tissues and one (TKCC‐19‐LO, moderately differentiated) from liver metastatic tissue, and all harbored a KRAS mutation [24]. In addition, TKCC‐06, ‐19‐LO and ‐22‐LO were TP53 wild‐type, whereas ‐25‐LO carried a TP53 mutation (Table S1). Multicellular TKCC‐22‐LO cultures had the highest metabolic activity and DNA content upon 14 days of 3D culture, followed by TKCC‐25‐LO and ‐19‐LO cells. Conversely, multicellular TKCC‐06 cultures had a lower metabolic activity and DNA content due to the shorter 3D culture period (7 days) and different medium composition (see methods; Figure 5b,c).

Multicellular 3D cultures including patient‐derived PDAC cells and their response toward GOT1 inhibition. (a) Micrograph of multicellular 3D cultures grown in GelMA‐HAMA hydrogels (scale bar = 100 µm). (b) Cell metabolic activity and (c) DNA content of mono‐ and multicellular 3D cultures including patient‐derived PDAC cells in GelMA‐HAMA hydrogels. (d) Cell viability, and (e) DNA content of mono‐ and multicellular TKCC‐22‐LO cultures following treatment with GOT1 inhibitor ± gemcitabine and nab‐paclitaxel. n = 3, * = p ≤ 0.05, ** = p ≤ 0.01, *** = p ≤ 0.001, **** = p ≤ 0.0001. (f) Live/dead staining (scale bar = 100 µm), and (g) heatmap of amino acid metabolism microarray of multicellular 3D cultures treated with the GOT1 inhibitor compared to non‐treated cells. PDAC, pancreatic ductal adenocarcinoma. GOT1, glutamic‐oxaloacetic transaminase 1. GelMA, gelatin methacrylate. HAMA, hyaluronic acid methacrylate.

GOT1 inhibition significantly reduced cell viability of monocellular 3D cultures but not multicellular 3D cultures using patient‐derived PDAC cells. Cell viability of TKCC‐22‐LO monocultures was reduced to 25.57 ± 9.84%, whereas cell viability of their multicellular counterpart was not affected (Figure 5d,e). Live/dead staining revealed viable and smaller spheroids following GOT1 inhibition compared to non‐treated spheroids, in both mono‐ and multicellular cultures (Figure 5f; Figure S11a). When treated with gemcitabine and nab‐paclitaxel (Gem/PTX), cell viability of both mono‐and multicellular TKCC‐22‐LO cultures decreased to 34.7 ± 2.12% and 29.52 ± 1.50%, respectively. GOT1 inhibition combined with cytotoxic drugs reduced cell viability of TKCC‐22‐LO monocultures to 23.78 ± 6.62% and multicellular 3D cultures to 26.34 ± 4.23% (Figure 5d). The DNA content correlated well with cell viability, other than for the reduced cell proliferation of multicellular TKCC‐22‐LO cultures of 62.66 ± 5.55% upon GOT1 inhibition, with negligible impact on cell viability (Figure 5e). Mono‐ and multicellular TKCC‐19‐LO and ‐25‐LO cultures showed a similar response to the individual and combined treatments. Notably, cell viabilities of TKCC‐19‐LO and ‐25‐LO monocultures treated with cytotoxic drugs were higher than their respective multicellular counterparts. In addition, the cell viability and DNA content of multicellular TKCC‐25‐LO and ‐19‐LO cultures were not affected by GOT1 inhibition (Figure S11b–e). Mono‐ and multicellular TKCC‐06 cultures treated with the GOT1 inhibitor had a reduced cell viability to 32.05 ± 0.58% and 57.06 ± 3.52%, respectively, while the cytotoxic drugs reduced their cell viability to 79.80 ± 4.22% and 43.14 ± 2.65%, respectively. However, these results were not reflected in a reduced DNA content, indicating that both treatment regimens impact cell metabolic activity rather than cell proliferation (Figure S11f,g).

Lastly, we sought to uncover the effects of GOT1 inhibition on the amino acid metabolism of patient‐derived cells. PCR microarray profiling of multicellular TKCC‐22‐LO cultures showed that GOT1 inhibition downregulated genes associated with glutamine metabolism (e.g. ALDH4A1, ASL, ASS1, CAD, CPS1, GFPT1, GLS, GLUD1, GOT2, NIT2, PPAT). Likewise, genes associated with arginine (e.g. ACY1, NOS2, SAT1) and proline (e.g. DAO, PRODH, PYCR1) metabolism were downregulated. Both amino acids are used by PDAC and stromal cells to sustain their proliferation within the pancreatic TME [15, 41]. Conversely, the upregulation of LDHA and MAOB suggests the activation of compensatory bioenergetic pathways, such as glycolysis, to sustain cell survival and proliferation following a disrupted glutamine metabolism [42] (Figure 5g). Overall, our results show that patient‐derived PDAC cells have differential responses to GOT1 inhibition, which are modulated by the presence of stromal cells and affect several amino acid‐related metabolic pathways.

Discussion

4

Using tumor tissue engineering, we show that stromal cells shape the ECM and metabolic processes, and GOT1 inhibition remodels the pancreatic TME. To date, most studies still rely on cell monolayers and animal models [16, 18, 19, 21, 22], neglecting the importance of human TME elements and differences between species. In particular, cell monolayers are grown in the absence of stromal cells and do not capture cell‐cell interactions, for example, between cancer cells and CAFs and their role in shaping cysteine and glutamine metabolism, and glycolysis in PDAC [20, 43, 44]. Miniaturized tissue replica, or organoids, have been used to study the TME and metabolic crosstalk between cancer cells and their surroundings, bridging the gap between cell monolayers and animal models [23]. However, these models often depended on Matrigel, a soft, undefined laminin‐enriched hydrogel which is no substitute for the ECM of solid tumors [27]. Specifically in PDAC, cancer cells undergo delamination in the early stages of the disease and reside within a collagen‐rich microenvironment in direct contact with CAFs and immune cells, underscoring the limitation of Matrigel‐based cultures in truly replicating the human pancreatic TME [36, 45, 46]. By incorporating tumor tissue engineering, advanced 3D models are constructed using bioengineered hydrogels [15, 23] or CAF‐derived ECMs [17, 20] that closely mimic the composition and mechanical properties of PDAC tissues and allow decoding the complex metabolic interactions within the TME [27].

Through proteomic and metabolomic analyses, we found that stromal cells not only remodeled the ECM but also altered the metabolic landscape within the pancreatic TME. Consistent with the literature, the incorporation of stromal cells upregulated ECM‐related processes, thereby promoting the secretion of collagen isotypes, regulatory molecules and glycoproteins. The upregulation of types I and VI collagen, fibronectin and lysyl oxidase mirrored the matrisome of PDAC tissues, and is linked to poor prognosis and drug resistance [36]. The upregulation of prostaglandin and prostanoid metabolic processes suggests the establishment of a pro‐inflammatory TME, which drives the differentiation of monocytes into immunosuppressive macrophages [47]. In our pancreatic TME model, we found that stromal cells regulated glutamine metabolism, OXPHOS and central carbon metabolism. While CAF‐derived glutamine is integral to PDAC metabolism and cell survival, the extent of this metabolic crosstalk, and the involvement of immune cells are not fully understood [5]. In breast cancer, glutamine is a precursor for CAF‐derived type I collagen, and is explored as a substrate for energy production and in facilitating CAF migration [41, 48]. Likewise, glutamine metabolism has been linked to type I collagen and fibronectin deposition by CAFs in lung cancer [49]. Given the upregulation of different collagens in our platform, we observe that a similar correlation between glutamine metabolism and ECM remodeling also applies to PDAC. OXPHOS is explored by pancreatic CAFs to oxidize lipids and support tumor growth, whereas central carbon metabolism sustains the bioenergetics of PDAC and stromal cells [50]. The increasing levels of both glutamine and OXPHOS in multicellular 3D cultures suggest that our platform resembles essential metabolic PDAC features. Future studies may include the metabolomic profiling of patient‐derived PDAC tissues, however, the handling and processing of tissue samples (e.g. fresh versus snap‐frozen) as well as patient‐to‐patient variability must be considered. The distinct proteomic and metabolomic profiles of multicellular 3D cultures grown in softer, undefined biomaterials (GelMA and Matrigel, respectively) underscore the importance of recreating the mechanical and biochemical properties of the ECM of tumor tissues in vitro.

To modulate glutamine metabolism, we subjected our pancreatic TME model to GOT1 inhibition. Targeting this cytoplasmic enzyme in PDAC significantly impairs the proliferation of cell monolayers and tumor growth in murine models by disturbing the intracellular redox balance and triggering cancer cell death by ferroptosis [5, 6]. While the inhibition of its mitochondrial counterpart ‐ GOT2 – is regulated by the pancreatic TME, how the stromal elements influence cell responses toward GOT1 inhibition remains unclear [8, 51]. We show that cell responses toward GOT1 inhibition are regulated by the cellular and extracellular elements of the pancreatic TME. Multicellular 3D cultures had a slight reduction in cell viability and proliferation compared to 3D monocultures, despite decreasing glutathione, NADH and NADPH in both 3D settings. Analysis of indirect multicellular 3D co‐cultures revealed that stromal cells are not affected by GOT1 inhibition and provide metabolites to sustain glutamine metabolism in PDAC cells. It remains unclear whether this metabolic crosstalk might influence other pathways in PDAC or in other stromal cells. CAF‐derived cysteine confers resistance to ferroptosis in PDAC cells, preventing oxidative damage by enhancing glutathione synthesis [43]. Real‐time metabolic flux measurements and targeted metabolomic analyses would help elucidate the metabolic interactions occurring between PDAC and stromal cells in this context.

We found that GOT1 inhibition upregulated many structural and regulatory proteins associated with the ECM, and altered the secretion of cytokines linked to the immune microenvironment in our pancreatic TME model. The overexpression of types I and III collagen and fibronectin in multicellular 3D cultures but not in 3D monocultures indicate that changes in the composition of the ECM are mainly derived from stromal cells. To date, the correlation between ECM remodeling following GOT1 inhibition in PDAC, and to our knowledge, in other solid tumors, has not yet been reported. Targeting glutamine metabolism by blocking glutaminase in PDAC cells co‐cultured with CAFs decreased the synthesis of types I, IV and V collagen [52]. In lung cancer, glutamine deprivation in cancer cells increases the production of type I collagen and fibronectin by CAFs through a YAP‐mediated mechanism [49]. The role of type III collagen in PDAC also remains unclear, while this protein sustains the dormancy of head and neck carcinoma cells [53]. We observed a reduction in the metabolic activity of PDAC cells grown within type III collagen matrices, suggesting that PDAC cells may become quiescent upon alterations in the ECM due to GOT1 inhibition. Last, we found that GOT1 inhibition modulated the levels of different cytokines, suggesting alterations in the immune landscape of our platform. Among those, the upregulation of IL‐6 and CCL2 (both associated with monocyte recruitment and T cells suppression [37, 38]) contrasts the increased infiltration of T cells and pro‐inflammatory macrophages in the pancreatic TME following glutaminase inhibition [52]. Together, our findings underscore the need to better understand the distinct effects of glutamine‐targeting strategies in PDAC and how specific metabolic pathways might be interrupted for improved therapeutic outcomes.

We conducted our 3D studies using standard culture medium supplemented with glucose (4500 mg/L) and glutamine (584.0 mg/L) at supraphysiological levels. Under these conditions, alternative metabolic pathways support cell survival and proliferation following GOT1 inhibition. These mechanisms may include the upregulation of GOT2, which can partially compensate for aspartate and α‐ketoglutarate metabolism, as well as support increased OXPHOS, lipid oxidation, and glutamine‐dependent TCA cycling [54]. In addition, elevated glucose availability may enable alternative routes for NADPH production, particularly through the pentose phosphate pathway [5]. Using glutamine‐free conditions, CAF‐derived nucleotides are taken up by PDAC cells to sustain their glucose levels, survival and proliferation [55]. We found that glucose deprivation had a comparable effect in reducing cell viability in our pancreatic TME upon GOT1 inhibition.

Given that GOT1 inhibition did not impair stromal cells and that PDAC cell responses were influenced by TME elements, we explored the effect of combining the GOT1 inhibitor with gemcitabine and nab‐paclitaxel. Previous work found that GOT1 inhibition improved responses of Matrigel‐based PDAC organoids toward treatments with gemcitabine via SIRT5 activation, but at different dosages [2]. This suggests that dosing parameters of both inhibitor and drugs may need to be optimized to assess any potential enhancement or synergy [56] in our pancreatic TME model. In this initial investigation, biological processes and signaling pathways affected by treatments with gemcitabine and paclitaxel aligned with those found in treated patient‐derived xenografts [57, 58], highlighting the dominant cytotoxic effect in our 3D studies. Our matrisome analysis revealed that GOT1 inhibition combined with cytotoxic treatment upregulated ECM‐related proteins as compared to GOT1 inhibition alone. Gemcitabine induces the expression of collagen isotypes and ECM regulators, and stiffening of the ECM [59, 60], thus, we hypothesize that the combined treatment using GOT1 inhibition and cytotoxic drugs may enhance these processes. Therapeutic strategies using stromal‐targeting compounds, such as LOX inhibitors [59], would be suitable to block ECM remodeling due to combined GOT1 inhibition and cytotoxic drugs, and may circumvent drug resistance upon matrix stiffening [61].

Finally, we validated our preclinical platform by integrating patient‐derived PDAC cells and treatments with the GOT1 inhibitor alone or in combination with gemcitabine and nab‐paclitaxel. Due to patient heterogeneity, there are no ‘one size fits all’ treatments, and ongoing clinical trials (e.g. NCT05296421) aim to target the metabolic vulnerabilities in PDAC. We found that GOT1 inhibition efficiently blocked the metabolic activity of KRAS‐mutant monocellular 3D cultures, independent of their differentiation status, and had a limited impact on multicellular 3D cultures. Our 3D patient‐derived cultures were grown in medium supplemented with growth factors (e.g. insulin, epithelial growth factor) as previously reported [24], which may have induced the proliferation of CAFs and activated compensatory metabolic pathways [8]. This underscores again the importance of therapeutic combination strategies using cytotoxic drugs and stromal‐targeting compounds (e.g. NCT04634539), and the need for advanced 3D models that integrate stromal cell populations. In addition, the incorporation of cancer and stromal cells from the same donor may allow for a patient‐specific representation of the tumor immune microenvironment. Our findings demonstrate the effect of inhibiting GOT1 on patient‐derived cells, paving the way for the adoption of our platform to screen other metabolism‐targeting strategies.

Although our model replicated key molecular features of the pancreatic TME, it remains unclear whether the ECM organization and architecture, including fibers alignment and porosity, are similar to those of desmoplastic PDAC tissues [17]. Characterizing these structural elements combined with single‐cell and/or spatially resolved analyses, may reveal how distinct cell populations and their cross‐talk shape the ECM [62], particularly following GOT1 inhibition. This includes the interactions between PDAC and cancer‐supporting cells (e.g. different CAF subtypes [63] and PBMCs), as well as with normal fibroblasts and the contribution of these cell types in remodeling the TME. Despite not fully mirroring human physiology, animal models remain the gold standard for validating in vitro findings and evaluating the effects of metabolism‐targeting strategies [23]. For example, GOT1 inhibition impaired PDAC cell monolayers and reduced tumor growth, whereas inhibition of its mitochondrial counterpart (GOT2) did not translate in vivo [5, 8]. These observations underscore the need for further advance tissue‐engineered 3D models to capture human‐specific features, and for integrating these platforms with other models to decipher the metabolic interactions in PDAC and guide the development of more effective therapies.

Conclusions

5

We report that targeting glutamine metabolism by GOT1 inhibition remodels the pancreatic TME. Our tumor tissue‐engineered platform shows that stromal cells actively shape the ECM while enhancing key metabolic programs, including glutamine metabolism. GOT1 inhibition altered the ECM composition and decreased cell viability and proliferation in multicellular 3D cultures, only under tissue‐like conditions. Although no synergistic effect was observed when combining GOT1 inhibition with cytotoxic drugs, our findings highlight the anti‐cancer potential of metabolism‐targeting compounds. This approach may also be explored to target other KRAS‐mutant tumors, such as lung cancer. We highlight the growing need for defined 3D cancer models to advance the screening and development of metabolic inhibitors, as well as supporting metabolism‐focused studies.

Author Contributions

R.C. and D.L. conceived the study and designed the experiments. R.C. performed the majority of the experiments. S.H. performed the immunohistochemistry. M.S. performed the bioinformatic analysis. C.K.B., J.R.S., and R.B.S. performed the LC‐MS experiments. E.S. maintained TKCC cells. D.C. provided the patient‐derived tissue samples for CAF isolation. K.S. and A.T. performed the mechanical characterization. D.L. supervised the project. R.C. and D.L. drafted the manuscript. All authors read and revised the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: advs73505‐sup‐0001‐SuppMat.docx.

Supporting File 2: advs73505‐sup‐0002‐Supplementary Tables.zip.