PPARδ Orchestrates a Prometastatic Metabolic Response to Microenvironmental Cues in Pancreatic Cancer

Abstract

The pronounced desmoplastic response in pancreatic ductal adenocarcinoma (PDAC) contributes to the development of a microenvironment depleted of oxygen and nutrients. To survive in this hostile environment, PDAC cells use various adaptive mechanisms that may represent therapeutic targets. In this study, we showed that nutrient starvation and microenvironmental signals commonly present in PDAC tumors activate PPARδ to rewire cellular metabolism and promote invasive and metastatic properties both in vitro and in vivo. Mild mitochondrial inhibition induced by low-dose etomoxir or signals from tumor-associated macrophages altered the lipidome and triggered the downstream transcriptional program of PPARδ. Specifically, PPARδ reduced mitochondrial oxygen consumption and boosted the glycolytic capacity by altering the ratio of MYC and PGC1A expression, two key regulators of pancreatic cancer metabolism. Notably, genetic or pharmacologic inhibition of PPARδ prevented this metabolic rewiring and suppressed both invasiveness in vitro and metastasis in vivo. These findings establish PPARδ as a central driver of metabolic reprogramming in response to starvation and tumor microenvironmental cues that promotes a prometastatic phenotype in PDAC, suggesting that PPARδ inhibition could serve as a therapeutic strategy to combat PDAC progression.

Significance: Nutrient starvation and microenvironmental signals activate PPARδ in pancreatic cancer to support survival and metastasis by promoting metabolic plasticity and invasiveness, providing a strong rationale for developing PPARδ-targeted therapies for pancreatic cancer.

Graphical abstract

Figure 1.

Induction of EMT-like phenotype and metabolic switch in PDAC cells upon starvation. A, Representative images illustrating morphologic changes for PDAC-354 cells in response to treatment for 72 hours with the complex I inhibitor metformin (3 mmol/L), the β-oxidation inhibitor etomoxir (20 μmol/L), the complex II inhibitor malonate (5 mmol/L), the pyruvate carrier inhibitor UK5099 (100 μmol/L), or tumor-like conditions [low pH (HCl 50 μmol/L) + low glucose (1 mmol/L) + 3%O₂]. Scale bar, 50 μm. B, Expression of EMT-associated genes (ZEB1, SNAI1, SNAI2, LOXL2, and VIM) was determined by RT-qPCR after cells were treated for 48 hours as indicated in A or with MCM. Pooled data for PDAC-185, A6L, 215, 253, and 354 (n ≥ 4 for each cell type). Data are normalized to HPRT1. Eto, etomoxier; Mal, malonate; Met, metformin. C, PDAC-215, 253, and 354 cells were treated with MCM or 20 μmol/L etomoxir for 48 hours and seeded in modified Boyden invasion chambers containing 20% FBS in the lower compartment. The number of invasive cells was analyzed after 16 hours (n = 7–11). D, GFP⁺-Luciferase⁺-PDAC-354 cells were treated with control, MCM, or 20 μmol/L etomoxir for 48 hours and then injected intrasplenically to assess their metastatic capacity (n = 9 mice/group). Representative photographs of liver metastasis (top) and subsequent hematoxylin and eosin staining (bottom). Scale bar, 200 μm. E, Representative OCR profile for PDAC-253 cells treated for 48 hours as indicated (mitochondrial stress test). O, ATP synthase inhibitor oligomycin; F, mitochondrial OXPHOS uncoupler FCCP [carbonyl cyanide-4 (trifluoromethoxy) phenylhydrazone]; A+R, complex III inhibitor antimycin A + electron transport change inhibitor rotenone. F, Maximal and ATP-linked respiration (resp) in non-CSC vs. CSC cells. Bars represent pooled data from PDAC-215, 253, and 354, showing individual data points corresponding to each PDX (n = 5–7). G, Representative extracellular acidification rate (ECAR) profile for PDAC-253 cells treated for 48 hours as indicated (glycolysis test). G, glucose; O, ATP synthase inhibitor oligomycin; 2DG, glycolysis inhibitor 2-deoxy-glucose. H, Glycolysis, glycolytic capacity, and glycolytic reserve in adherent (non-CSC) vs. sphere-derived cells (CSC). Bars represent pooled data from PDAC-215, 253, and 354, showing individual data points corresponding to each PDX (n = 4–5). All data are represented as the mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001. See also Supplementary Figs. S1–S3.

Figure 2.

Transcriptomic and scRNA-seq analyses identify metabolic switch and EMT induction upon pseudostarvation. Cells were treated for 48 hours with vehicle (Ctrl), MCM, and 20 μmol/L etomoxir (Eto). A, Gene expression profile as assessed by a carbohydrate metabolism PCR array in PDAC-354 cells. Heatmap showing only genes whose expression was significantly altered (n = 3). B, Left, PDAC-003 cells were treated as indicated and were then subjected to scRNA-seq (10X Genomics Chromium platform). Unsupervised clustering of viable PDAC cells exposed to Ctrl, MCM, or etomoxir, represented as Uniform Manifold Approximation and Projection (UMAP) plots. Different clusters are color-coded. Right, boxplots illustrating gene set enrichment results for the EMT and glycolysis (Hallmark data set) for different clusters in Ctrl vs. MCM and etomoxir treatment, respectively. Differences in enrichment scores between treatments were assessed using the Mann–Whitney U test. C, Expression of EMT hallmark signature and PPARD in single cancer cells (PDAC-002 and 021) displayed as unsupervised clusters and color-coded for allocated treatment. See also Supplementary Figs. S4 and S5.

Figure 3.

PPARD expression is linked to metabolic switch and EMT in patients with PDAC. A, Expression levels of PPARD in PDAC tumors (T) vs. surrounding normal tissue (N) included in the TCGA and GTEx projects. B, Patients were dichotomized for PPARD expression [higher (n = 58) and lower (n = 81) expression compared with the mean; RNA-seq V2 RSEM values]. Kaplan–Meier curves for disease-free survival are shown. Dotted lines denote the confidence intervals. Survival, 19.48 vs. 13.53 months. C and D, A TMA with 108 cases was stained by IHQ using an anti–PPARδ antibody. Representative images depicting different signal intensities (weak vs. strong) and localizations (nuclear vs. cytoplasmatic) are shown in C. Magnification, ×400. Scale bar, 50 μm. Arrows, stained cancer cells. D, Signal intensity was scored by Allred immunoreactive score, and patients were dichotomized into low (n = 34; scores 0–6) vs. high (n = 59; scores 7–8) PPARδ expression groups. Kaplan–Meier curves for overall survival are shown. Dotted lines denote the confidence intervals. Survival, 17.5 vs. 11 months. E, Correlation between PPARD tumor expression levels family members and an EMT-associated signature composed of SNAI1, SNAI2, and ZEB1. F, Gene sets enriched in the transcriptional profile of tumors belonging to the top PPARD high-expression group compared with the bottom low-expression group in the TCGA data series (PAAD, pancreatic adenocarcinoma). Shown are the normalized enrichment score (NES) values for each pathway using the Hallmark gene sets, meeting the significance criteria: nominal P value of < 0.05, FDR < 25%. G, Enrichment plot for TNFα, EMT, glycolysis, and hypoxia hallmarks in PPARD-high vs. -low samples shown in F, indicating values of normalized enrichment score and FDR q values. H, Representative images of costaining of PPARδ (nuclear and cytoplasmatic, red) and HIF1α (nuclear brown) by IHQ in several patients shown in C and D. Magnification, ×400. Scale bar, 50 μm. Arrows, costained cancer cells. See also Supplementary Fig. S6.

Figure 4.

Activation of PPARδ initiates invasiveness and metastasis. A,PPARD mRNA expression upon 24 to 48 hours of treatment with MCM, etomoxir (Eto), and 5 μmol/L of the PPARδ agonist GW0742. Pooled data of PDAC-215, 253, and 354 cells (n = 4–7). B, Representative Western blot after 48 hours of treatment in PDAC-354 cells. C, PPARδ activity, measured as binding to the PPAR response element, following stimulation with MCM, etomoxir, and the PPARδ agonist GW0742 for 24 hours (n = 5). D, CUT&Tag analysis of PPARδ protein binding at the UCP1, PGC1A, and SNAI2 loci. WashU Epigenome browser tracks showing CUT&Tag signals at the mentioned loci with the indicated transcription start site (TSS). Blue signals represent PPARδ binding in control (Ctrl) conditions, and red signals represent PPARδ binding upon 24 hours of etomoxir treatment in PDAC-002 cells. E, Lipidomics analyses for PDAC-215 and PDAC-253 cells treated for 24 hours with MCM and etomoxir. OPLS-DA analysis showing the most represented lipids common for PDAC-215 and 253 for each experimental conditions vs. the control condition (n = 3). F, Venn diagram indicating the number of lipid species for each experimental group. The four common upregulated lipids for all four conditions are indicated in the square. G, Invasive capacity of cells treated for 48 hours with the PPARδ agonists L-165 and GW0742 (5 μmol/L). Cells were placed in modified Boyden invasion chambers containing 20% FBS in the lower compartment, and the number of invasive cells was assessed after 16 hours (n = 4–8). H, Experimental metastasis assay of PDAC-354-GFP-Luc cells pretreated with GW0742 for 48 hours. After intrasplenic injection, mice received three more daily doses of GW0742 (0.3 mg/kg i.v.). IVIS imaging (left) and quantification of the total CK-19 area in the livers 9 weeks after implantation (right). All data are represented as the mean ± SEM. *, P < 0.05; **, P < 0.01. See also Supplementary Figs. S7–S9.

Figure 5.

PPARδ controls the balance between OXPHOS and glycolysis, linked to EMT and metastasis. A,In vitro invasion in PDAC-215, 253, and 354 cells stably transduced with inducible lentiviral vectors expressing either a nontargeting short hairpin RNA (NT shRNA) or three different shRNAs against PPARD (sh#1, sh#2, and sh#3). Transduced cells were pretreated with doxycycline for 24 hours, then incubated with MCM, etomoxir (Eto), or L-165 for 48 hours, and finally plated in modified Boyden chambers for 16 hours (n = 7). B, Top, ZsGreen expression by RT-qPCR in liver homogenates from an in vivo metastasis assay of PDAC-354 cells stably expressing either the NT or the sh#1 against PPARD. Cells were pretreated with doxycycline and/or 20 μmol/L etomoxir for 48 hours. After intrasplenic implantation, mice were treated with oral doxycycline (2 mg/mL; drinking water) and etomoxir (15 mg/kg, i.p. daily) for 7 days, when splenectomies were performed. Bottom, numbers indicate the percentage and total number of micrometastases in each experimental group. C, PDAC-215, 253, and 354 transduced cells as in A were pretreated with doxycycline for 24 hours, then incubated with MCM, etomoxir, or L-165, and then tested for ATP-linked respiration (top) and glycolytic capacity (bottom) after additional 24 hours (n = 8). D, Mitochondrial stress test (top row) and glycolysis test (bottom row) following treatment with control (Ctrl) or the PPARδ agonists L-165 or GW0742. Left column, representative OCR and extracellular acidification rate (ECAR) profiles for PDAC-253. Right column, pooled data for PDAC-215, 253, and 354 cells (n = 6–9). Glyco, glycolytic; Max res, maximum respiration; O, ATP synthase inhibitor oligomycin; F, mitochondrial OXPHOS uncoupler FCCP [carbonyl cyanide-4 (trifluoromethoxy) phenylhydrazone]; A+R, complex III inhibitor antimycin A + electron transport change inhibitor rotenone. G, glucose; 2DG, glycolysis inhibitor 2-deoxy-glucose. E, ATP-linked respiration (left) and maximal respiration (right) for control vs. GW0742-treated cells following treatment with or without palmitate-BSA (FAO assay). PDAC-354 cells were treated with 10 μmol/L GW0742 for 48 hours prior to the assay (n = 5). In A, C, and D, the bars represent pooled data from PDAC-215, 253, and 354, showing individual data points corresponding to each PDX. All data are represented as the mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001 vs. control, ###, P < 0.001 vs. palmitate.

Figure 6.

PPARδ rewires cellular metabolism regulating MYC/PGC1A balance. A, Expression of MYC, PGC1A, and MYC/PGC1A ratio in PDX-354 after mitochondrial energy deprivation during 48 to 72 hours (n = 4–7). B,MYC and PGC1A reporter assay. Promoter activity was estimated as luciferase bioluminescence at the indicated times following treatment with PPARδ agonist GW0742 or PPARD overexpression (PPARD OE; n = 3–5). C, PDAC-354 cells were transduced with inducible lentiviral vectors expressing either a nontargeting short hairpin RNA (NT shRNA) or two different shRNAs against MYC (sh#1 and sh#2) or the complete cDNA of PGC1A. Effect of MYC knockdown (shMYC, pooled data for sh#1 and sh#2) or PGC1A overexpression (PGC1A OE) on invasiveness in response to treatment with 5 μmol/L PPARδ agonist L-165 for 48 hours (n = 6–8). D, CUT&Tag analysis of MYC protein binding at the CCNE, PGC1A, SNAI2, and VIM loci. WashU Epigenome Browser tracks showing CUT&TAG signals at the mentioned loci with the indicated transcription start site (TSS). Blue signals represent MYC binding in control (Ctrl) conditions, and red signals represent MYC binding upon 24 hours of etomoxir (Eto) treatment in PDAC-002 cells. E–H, PDAC-215 and 354 cells were transduced as in A, pretreated with doxycycline for 48 hours, and then incubated with MCM or etomoxir. E, OCR changes for maximal respiration (Max resp; left) and ATP-linked respiration (n = 4; right). F, Glycolytic (Glyco) capacity (left) and reserve (n = 4; right). G,ZEB1 gene expression. H, Invasive capacity (n = 10). I, PDAC-354 cells were treated with MCM or 20 μmol/L etomoxir for 48 hours in the presence or absence of the MYC/Max interaction inhibitor Mycro3 (25 μmol/L). Cells were then seeded in modified Boyden invasion chambers containing 20% FBS in the lower compartment. The number of invasive cells was assessed after 16 hours (n = 5). In E, F, and H, the bars represent pooled data from PDAC-215 and 354, showing individual data points corresponding to each PDX. All data are represented as the mean ± SEM. #, P < 0.05; ##, P < 0.01; ###, P < 0.001 vs. unstimulated control. *, P < 0.05; **, P < 0.01; ***, P < 0.001 vs. NT. See also Supplemenatry Figs. S10–S13.

Figure 7.

Therapeutic targeting of PPARδ impairs invasion in vitro and metastasis in vivo. A, PDAC-215, 354, and 265 cells were pretreated with PPARδ antagonists GSK0660 (10 μmol/L) and GSK3787 (10 μmol/L) and inverse agonist DG172 (2.5 μmol/L) for 1 hour and then treated (PDAC-215 and 354) or not (PDAC-265) with MCM or etomoxir (Eto) for 48 hours. Invasion over 16 hours was assessed in modified Boyden invasion chambers (n = 6). B–F, Spontaneous metastasis upon orthotopic injection of 10⁵ metastatic PDAC-265-GFP-Luc cells (n = 8 mice/group). Following implantation, mice were treated daily with either vehicle, the PPARδ agonist GW0742 (0.3 mg/kg i.p.), or the PPARδ antagonist GSK3787 (3 mg/kg i.p.) until termination of the experiment at week 9, when mice became moribund. Tumor growth was assessed by weekly IVIS. B, Metastasis onset evaluated as hGAPDH absolute copy number in livers. C, Expression of PPARD and downstream targets in pancreatic tumors measured by RT-qPCR. D, Quantification of PPARδ protein expression relative to β-actin that was used as loading control, measured by Western blot. E, Expression levels of PPARδ by IHQ (representative images). F, Expression levels of CK-19 in liver sections (top, representative images) or c-MYC (brown) and VIM (purple) in pancreatic tumors was measured by IHQ (bottom, representative images). MYC and VIM stainings were quantified using the Allred score, and median scores per group are shown as text inserts. All data are represented as the mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001 vs. control cells; #, P < 0.05; ##, P < 0.01; ###, P < 0.001 vs. control or single treatment. See also Supplementary Fig. S14.