Vertical RAS pathway inhibition in pancreatic cancer drives therapeutically exploitable mitochondrial alterations

Abstract

Oncogenic KRAS mutations drive metabolic reprogramming in pancreatic ductal adenocarcinoma (PDAC). Src-homology 2 domain-containing phosphatase 2 (SHP2) is essential for full KRAS activity, and promising dual SHP2/mitogen-activated protein kinase (MAPK) inhibition is currently being tested in clinical trials. Exploitable metabolic adaptations may contribute to invariably evolving resistance. To understand the metabolic changes induced by dual inhibition, we comprehensively tested human and murine PDAC cell lines, endogenous tumor models, and patient-derived organoids, which are representative of the full spectrum of PDAC molecular subtypes. We found that dual SHP2/mitogen-activated protein kinase kinase (MEK1/2) inhibition induces major alterations in mitochondrial mass and function, impacts reactive oxygen species (ROS) homeostasis and triggers lipid peroxidase dependency. Anabolic pathways, autophagy and glycolysis were also profoundly altered. However, most strikingly, mitochondrial remodeling was evident, persisting into a therapy-resistant state. The resulting vulnerability to the induction of ferroptotic cell death via the combination of vertical SHP2/MEK1/2 with glutathione peroxidase (GPX4) inhibition was largely independent of the PDAC molecular subtype and was confirmed with direct targeting of RAS. The triple combination of SHP2/MEK1/2 inhibition and the ferroptosis-inducing natural compound withaferin A suppressed tumor progression in an endogenous PDAC tumor model in vivo. Our study offers a metabolic leverage point to reinforce RAS pathway interference for targeted PDAC treatment.

Fig. 1

Pharmacological SHP2 and/or MEK inhibition reprograms PDAC cell metabolism. a Transcriptomic classification of human and murine PDAC cell lines into a spectrum of basal-like to classical subtypes. b Proliferation of PDAC cell lines treated with DMSO, SHP099 (15 μM), trametinib (10 nM), or both. The data represent the SDs from 12 wells in one experiment. c Left: Metabolic inhibitors combined with MAPK pathway inhibition. Right: Example of a proliferation assay showing color-coded MAPK treatment comparisons at ~90% confluency, with or without metabolic inhibitors. d–g Relative cell proliferation under MAPK pathway inhibition with low (L), medium (M), and high (H) metabolic inhibitor concentrations in PDAC cells. The heatmap represents the mean of eight wells across two independent experiments; the metabolic inhibitor concentrations are depicted in Supplementary Fig. 2. For the crossed-out areas, no data could be obtained. h Mitochondrial mass in PDAC cells was measured via flow cytometry, and three to four independent experiments were performed. i Intracellular ROS by flow cytometry, with tert-butyl hydroperoxide (TBHP) as the ROS-inducing control; mean of two to four independent experiments. j Spare respiratory capacity and glycolytic reserve in PDAC cells, as assessed via the ECAR and OCR in three experiments with six technical replicates each. Statistical significance was determined via one-way ANOVA in (b, h, i, j), with comparisons made against corresponding DMSO controls. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Fig. 2

Pharmacological SHP2 and/or MEK inhibition reshapes both the intra- and extracellular metabolite profiles. a PCA of extracellular metabolites in PDAC cell lines treated for 72 h with DMSO, SHP099 (15 μM), trametinib (10 nM), or the combination (four samples per treatment; 62 metabolites for MIA PaCa II, 54 for PANC-1 and YAPC). b Summarized PCA contribution factors that indicate how much each variable influences the principal components. PCA contribution of human PDAC cell lines at 12, 48, and 72 h of treatment, showing variable loadings on the first two principal components, which capture the most variance. The values for the various metabolic pathways (e.g., amino acids) represent the means of the individual metabolites shown in Supplementary Fig. 5. c Heatmap of intra- and extracellular metabolite levels in cells treated with SHP099, trametinib, or both vs. DMSO controls (set to 1). The color gradient shows the fold changes, with dark red indicating values > 3 for clarity. For the crossed-out areas, no data could be obtained. d KEGG-based gene set enrichment analysis (GSEA) was performed on human and murine PDAC cells treated with DMSO or the combination of SHP099 and trametinib for 48 h. Each colored square represents a significant upregulation or downregulation of at least one gene set within a metabolic category relative to DMSO, with the individual gene sets listed in Supplementary Fig. 7b. “Not definitively assignable” indicates unclear metabolic classification. Gene sets with an adj. P < 0.25 are shown. Statistical significance was determined via one-way ANOVA in (c), with comparisons made against corresponding DMSO controls. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

Fig. 3

Dual SHP2 and MEK inhibition causes mitochondrial adaptations in vivo. The treatment schedule for the KPC mice is shown at the top left. The mice were treated with vehicle (control), SHP099 (75 mg/kg), trametinib (1 mg/kg), or their combination every other day. The mouse icon was sourced from BioRender.com. a Dynamics of KPC tumors in mice, from which tumor interstitial fluid was collected for analysis via LC‒MS/MS. b Metabolite abundance under targeted therapy relative to that in vehicle-treated tumors, organized by metabolic pathways. Supplementary Fig. 11 details individual metabolites. c Tumor dynamics in KPC mice, with tumors processed for electron microscopy. d Left: Mitochondrial diameter in PDAC tumor cells from KPC mice, shown for all therapy arms. Each dot represents a single mitochondrion. Right: Same data separated by short (≤2 weeks) and long (>2 weeks) therapy durations. e Representative mitochondrial morphology in KPC-PDAC tumor cells under targeted therapy. f Tumor dynamics in KPC mice, with tumors processed for whole-tissue RNA sequencing. g PCA of KPC tumor transcriptomes following therapy. h In silico deconvolution of RNA-sequenced KPC tumors. i KEGG-based GSEA of KPC tumor transcriptomes. Each colored square represents a significant up- or down regulation of at least one gene set within a metabolic category relative to the vehicle reference, with the individual gene sets listed in Supplementary Fig. 7b. Gene sets with an adj. P < 0.25 are shown. Statistical significance in (b, d) was determined by one-way ANOVA against vehicle controls: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. G3P glyceraldehyde-3-phosphate, MMA methylmalonic acid, 5-HIAA 5-hydroxyindoleacetic acid, 2-AAA 2-aminoadipic acid, GSH glutathione, MSO methionine sulfoximine, SAM S-adenosylmethionine, SAH S-adenosylhomocysteine, CysGly cysteinylglycine, THF tetrahydrofolate

Fig. 4

Mouse PDAC tissue single-cell transcriptomics confirm the involvement of mitochondria in the adaptive response to dual SHP2 and MEK inhibition. a Left: Batch-corrected UMAP plot of cells from PDAC tissue from individual KPC mice treated with vehicle (acting as a control), trametinib (1 mg/kg), or the combination of SHP099 (75 mg/kg) and trametinib. KPC mice treated for ≤2 weeks (short) or >5 weeks (long) were included. Right: UMAP plot highlighting the subclustering of the tumor cell/epithelial clusters. b Dot plot displaying the expression levels of selected genes across various cell types. c PDAC subtype signature enrichment and its contributions to the tumor cell/epithelial subclusters. Gene sets according to Collisson et al., Moffitt et al., and Chan-Seng-Yue (CSY) et al. were applied. d Percent distribution of the indicated treatments across different tumor cell/epithelial subclusters. e Cluster-specific gene set enrichment analysis within the tumor cell/epithelial compartment, including the top 500 up- and downregulated differentially expressed genes. f Treatment-specific GSEA within the tumor cell/epithelial compartment, based on the top 500 up- and downregulated DEGs. In (e, f), each colored square represents a significant up- or downregulation of at least one gene set within a metabolic category, with the individual gene sets listed in Supplementary Fig. 7b. Gene sets with an adj. P < 0.25 are shown. DCs dendritic cells, MDSCs myeloid-derived suppressor cells

Fig. 5

RAS pathway inhibition triggers lipid peroxidase dependency. a Left: Representative images of MDA staining. Right: Quantification of MDA levels in KPC tumors treated with vehicle, SHP099 (75 mg/kg), trametinib (1 mg/kg), or their combination. Each dot represents the histoscore of a single mouse, calculated as the mean of five to ten tumor regions analyzed per animal. b Heatmap showing relative lipid peroxidation (C11-BODIPY fluorescence) across multiple PDAC cell lines. The values are normalized to those of the DMSO control. TBHP served as a positive control, and ferrostatin-1 served as an antioxidant rescue control. The quantification results are shown in Supplementary Fig. 12b. Crossed-out areas indicate missing data. c Detection of lipid peroxidation via C11-BODIPY fluorescence in the context of RAS inhibition. Left: Effect of the addition of RMC-7977 (10 nM). Experiments were conducted in seven independent PDAC cell lines, each measured in two technical replicates per condition. Right: Lipid peroxidation in cells resistant to RMC-7977 (in purple, 10 nM) or MRTX1133 (in magenta, 10 µM). Experiments were conducted in three independent PDAC cell lines, each measured in two technical replicates per condition. ML210 concentrations were adapted to individual cell line sensitivity: low/medium/high concentrations - MIA PaCa II: 0.25/0.5/1 µM; KPC495: 5/10/20 µM; KPC382: 50/75/100 µM; PANC-1: 0.1/0.25/1 µM; YAPC: 10/15/20 µM; KPC323: 50/75/100 µM; KPC330: 10/20/40 µM. All values are shown relative to the corresponding treatment without ML210. d Determination of ML210 IC₅₀ values across different treatment conditions in multiple PDAC cell lines. e ΔIC₅₀ values for human and murine PDAC cell lines treated with DMSO or RMC-4550 + trametinib are shown, calculated as the IC₅₀ in the presence of MitoTEMPO relative to the IC₅₀ without MitoTEMPO. The values are plotted on a logarithmic scale. f Transcriptomic analysis classifying human PDAC organoids into basal-like or classical subtypes via ssGSEA. The color code represents the normalized enrichment score (NES) of each PDAC subtype signature. g Row-scaled intensity (z-score) of glycolytic and lipogenic genes across PDAC organoid samples. Rows and columns were hierarchically clustered via the complete linkage method with Euclidean distance as the similarity metric. h Relative proliferation of organoids treated with SHP099 (15 μM) + trametinib (10 nM) or RMC-7977 (10 nM) with or without ML210 (15 µM). The values are normalized to those of untreated controls. The data are based on two to four biological replicates with five technical replicates each. i Increased subtype-specific proliferation-inhibitory effects of ML210 in combination with the respective MAPK-pathway-targeted therapies. In (b) (left panel), statistical significance was assessed via one-way ANOVA, with black asterisks indicating differences between the untreated (DMSO) and treatment groups and blue asterisks denoting differences between the groups treated with or without ML210. Statistical comparisons with the DMSO-treated controls shown in (c, d) were also performed via one-way ANOVA. As shown in (i), one-way ANOVA was conducted to assess differences between subtypes in general. As shown in (h), differences between groups with and without ML210 were evaluated via unpaired t-tests. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Fig. 6

Combined SHP2/MEK/GPX4 inhibition delays tumor progression in vivo. a Schematic representation of the treatment trial with triple SHP099 (75 mg/kg) + trametinib (1 mg/kg) ± withaferin A (4 mg/kg) and the respective controls. Treatments were administered every other day. The mouse icon was sourced from BioRender.com. b Initial pancreatic volume relative to body weight before the start of treatment. c Individual volume changes over time relative to the baseline volume. d Left: Tumor dynamics over time are shown as the tumor-to-body weight ratio (%). The error bars indicate the standard error of the mean (SEM). Right: Pancreatic volumes at individual timepoints relative to the corresponding baseline volume. Each dot represents the mean tumor volume of multiple mice per treatment group and time point, normalized to the therapy-specific baseline volume at therapy initiation. e Endpoint pancreatic weights relative to body weight across treatment groups. f Kaplan‒Meier survival curves comparing treatment groups analyzed via the Mantel‒Cox log-rank test. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated via the Mantel‒Haenszel method. For statistical analysis, unpaired two-tailed t-tests were performed in (b, e), whereas a paired two-tailed t-test was used in (d)