Gemcitabine (Sandoz S.p.a), Irinotecan (Fresenius Kabi Italia S.r.l.), and Oxaliplatin (Accord Healthcare S.L.U.) were stored at 4 °C, and 5-fluorouracil (Teva Italia S.r.l.) was at room temperature (RT). Sodium succinate hexahydrate was was dissolved in phosphate-buffered saline (PBS; Thermo Scientific).

The pancreatic adenocarcinoma cell lines Panc1, MiaPaCa2, HS776T, Suit2, and PaCa3 were grown in RPMI-1640 supplemented with 10% fetal bovine serum (FBS) and 50 µg/mL gentamicin sulfate (all from Gibco/Life Technologies, USA) and were maintained at 37 °C with 5% CO₂. CSCs were obtained as previously described^[4]. Briefly, adherent cells were washed twice in 1× PBS (Gibco/Life Technologies, USA) and then cultured in stem-specific medium, i.e., DMEM/F-12 without glucose (US Biological Life Sciences, USA) supplemented with 1 g/L glucose, B27, 1 µg/mL fungizone, 1% penicillin/streptomycin (all from Gibco/Life Technologies, USA), 5 µg/mL heparin (Sigma/Merck), 20 ng/mL epidermal growth factor (EGF), and 20 ng/mL fibroblast growth factor (FGF) (both from PeproTech, United Kingdom). Panc1 cells were cultured in flasks with a hydrophobic surface specifically designated for the growth of suspension cells and were maintained at 37 °C with 5% CO₂ in the stem-specific medium 8 weeks (8 w), refreshed twice a week with new medium. Before each experiment, cells were passed through a cell strainer (40 µm) to separate and maintain only the cell aggregates/spheres, which were trypsinized to obtain single-cell suspension. Bright-field images were acquired with an inverted microscope (Axio Vert. A1, Zeiss, Germany). In all the assays, both parental cells and CSCs were cultured in proliferating conditions and did not undergo contact inhibition–induced growth arrest.

The human pancreatic cancer cell line Panc1, along with the other cell lines used in this study (MiaPaCa2, HS776T, Suit2, and PaCa3) were previously purchased from American Type Culture Collection (ATCC) or kindly provideded by colleagues at University of Verona, Italy. Ethical approval was not required for this study as all cell lines were commercially obtained or provided by collaborators.

TRAP1-knockout Panc1 cell line was generated as previously reported by Cannino et al.^[12]. In brief, guide RNAs specific for human TRAP1 were annealed and cloned into the lentiCRISPRv2 vector (Addgene plasmid #52961). The construct was co-transfected into HEK293T cells together with the packaging plasmids pMDLg/pRRE, pRSV-Rev, and pMD2.G to produce lentiviral particles. These recombinant lentiviruses were subsequently used to transduce Panc1 cells, and stable knockout cells were selected using 1 µg/mL puromycin.

Total RNA was extracted from about 1 × 10⁶ cells using TRIzol Reagent (Life Technologies, USA) and 1 µg of RNA was using a first-strand cDNA synthesis kit. Real-time quantification was performed in duplicate samples by SYBR-Green detection chemistry with GoTaq quantitative polymerase chain reaction (qPCR) Master Mix (Promega, USA) on a QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific, USA). The primers used were listed in Table 1.

The cycling conditions used were: 95 °C for 2 min, 40 cycles at 95 °C for 3 s, 60 °C for 30 s, 95 °C for 15 s, and 60 °C for 1 min. The average of cycle threshold of each triplicate was analyzed according to the 2^-∆∆Ct method using SDHA as an endogenous control.

CDH1 levels were also analyzed following treatment with succinate. Cells were seeded in 60-mm Ø plates (3.5 × 10⁵ cells/plate) and, the day after, were treated for 48 h with succinate at different concentrations: 0.5, 2, and 20 mM.

Panc1 parental cells (P), TRAP1-KO cells, and CSCs at 2, 4, and 8 weeks of culture were plated in 24-well cell culture plates (3 × 10⁴ cells/well) in their respective medium and were incubated at 37 °C with 5% CO₂. Viable cells were counted by Trypan Blue dye exclusion after 2, 4, 7, and 10 days of culture. The doubling time was calculated using the formula: T = (T₂ - T₁) × log2/log(Q₂/Q₁), where: T₁, day 0; T₂, day 10; Q₁, cell number at day 0; and Q₂, cell number at day 10.

To test the cell viability after treatment, cells were seeded in 96-well plates (5 × 10³ cells/well) and, on the following day, were incubated with compounds at the following concentrations: 50 µM Gemcitabine, 100 µM Oxaliplatin, 200 µM Irinotecan, and 1 mM 5-fluorouracil. At the end of the treatment (48 h for all compounds), cell growth was measured by Resazurin assay (Immunological Sciences, Italy) according to the manufacturer’s protocol and fluorescence was measured by Tecan Infinite 200 PRO microplate reader (Ex: 535 nm, Em: 590 nm).

To prepare the samples, frozen cell pellets were resuspended in lysis buffer {i.e., 1 mM Na₃VO₄, 1 mM NaF, 2 mM ethylenediaminetetraacetic acid (EDTA), 0.2 mM phenylmethylsulfonyl fluoride (PMSF), 150 mM NaCl, 1× complete protease inhibitor cocktail, and radioimmunoprecipitation assay (RIPA) buffer pH 8.0 [150 mM NaCl, 50 mM Tris-HCl, 1% Igepal, 0.5% sodium deoxycholate, and 0.1% sodium dodecyl sulfate (SDS)]}. The lysate was centrifuged at 12,000 × g for 10 min at 4 °C and the supernatant was used for immunoblotting. Protein concentration was measured with the Bradford Reagent (SERVA electrophoresis, Heidelberg, Germany) using bovine serum albumin (BSA) as a standard. Thirty µg of protein suspended in SDS loading buffer were run on 12% SDS polyacrylamide gels and electrotransferred to polyvinylidene difluoride (PVDF) membranes (Merck Millipore, Burlington, MA, USA). Membranes were then incubated for 1 h at RT with blocking solution, i.e., 5% nonfatdry milk in tris-buffered saline with Tween 20 (TBST) pH 7.5 (100 mM Tris-HCl, 0.1% Tween-20, and 0.9% NaCl). Then, the membranes were incubated with primary antibodies at appropriate dilutions in blocking solution overnight at 4 °C. Primary antibodies included the following: TRAP1 (1:1,000, sc-73604 Santa Cruz Biotechnology), E-cadherin (1:1,000, #A20798 ABclonal), phospho-DRP1 (p-DRP1; 1:1,000, #4494), dynamin-related protein 1 (DRP1; 1:1,000, #8570), mitochondrial fission factor (MFF; 1:1,000, #E5W4M), optic atrophy 1 (OPA-1; 1:1,000, #80471), mitofusin 1 (MFN1; 1:1,000, #14739), and mitofusin 2 (MFN2; 1:1,000, #11925). All antibodies were obtained from Cell Signaling Technology unless otherwise specified. Horseradish-peroxidase-conjugated anti-mouse (1:8,000; KPL #074-1806) or anti-rabbit (1:10,000; KPL #074-1516) were used as secondary antibodies. The immunocomplexes were visualized by chemiluminescence using the ChemiDoc MP Imaging System (Bio-Rad Laboratories, Hercules, CA, USA), and the intensity of the chemiluminescence response was measured by processing the image using NIH ImageJ software (http://rsb.info.nih.gov/nih-image/). Amido black staining was used to confirm loading in different lanes.

Panc1 P, TRAP1-KO cells, CSCs 2w, and CSCs 2w TRAP1-KO were processed for the ultrastructural morphological and morphometric evaluation of mitochondria by transmission electron microscopy (TEM). Cell samples were fixed with 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, at 4 °C for 2 h. Cell samples were then rinsed in PBS, postfixed with 1% OsO₄ and 1.5% potassium ferrocyanide for 1 h at 4 °C, dehydrated with acetone, and embedded in Epon 812 resin. Ultrathin sections (80-nm thick) were stained with lead citrate for 1 min and observed with a Philips Morgagni transmission electron microscope operating at 80 kV and equipped with a Megaview III camera for digital image acquisition. Morphometric evaluation of the mitochondrial area, aspect ratio, cristae extension, cristae width, and cristae junction diameter were performed. Micrographs (×18,000) of 50 randomly selected mitochondria from 20 different cells were taken and the area and the lengths of the major and the minor mitochondrial axes were measured. The aspect ratio parameter (index of mitochondrial elongation) was calculated as the ratio of the two lengths. The length of the outer and the inner mitochondrial membrane was measured (micrographs at ×36,000) in 30 mitochondria per sample, and the inner/outer membrane ratio was calculated as an assessment of the cristae extension independent of mitochondrial size. Cristae width and junction diameter were measured in 15 mitochondria for each sample (micrographs at ×44,000).

The measurements were performed using ImageJ software (NIH Image, Bethesda, MD, USA). The means ± standard error (SE) were calculated for all of the mitochondrial parameters and the statistical comparison was performed with the Kruskal–Wallis non-parametric test followed by the Mann–Whitney test for pairwise comparison. Statistical significance was set at P ≤ 0.05.

Extracellular oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured with the Seahorse XF24 Extracellular Flux Analyzer (Seahorse Bioscience, USA).

Panc1 P cells, Panc1 TRAP1-KO P cells (3 × 10⁴ cells/well), and CSCs 2w TRAP1-KO (6 × 10⁴ cells/well), were plated in the V7 XFe-24-well cell culture microplate precoated with 3 mg/mL collagen I (Gibco/Life Technologies, USA) and cultured at 37 °C with 5% CO₂ overnight to reach the appropriate confluence. On the day of the assay, cells were incubated in Seahorse XF DMEM Medium (Seahorse Bioscience, cat. No. 103575-100) supplemented with 10 mM glucose, 1 mM pyruvate, 2 mM glutamine, pH 7.4, for 1 h in a non-CO₂ incubator.

For real-time adenosine triphosphate (ATP) rate assay, OCR, and ECAR were recorded at the baseline and after sequentially adding oligomycin (ATP synthase inhibitor; 1 µM) through port A, and a combination of rotenone (complex I inhibitor; 1 µM), and antimycin A (complex III inhibitor; 1 µM) through port B.

For Mito stress test OCR and ECAR were recorded at the baseline and after sequentially adding oligomycin A (1 µM; port A), Carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP, 1.5 μM; port B), and rotenone/antimycin A (1 µM / 1 µM; port C).

Raw OCR and ECAR were normalized to the DNA content per well which was quantified with the CyQUANT Cell proliferation assay kit (Thermo Fisher Scientific, Cat. No. C35007), according to the manufacturer’s instructions, on a Victor X4 multilabel microplate reader (Perkin Elmer). Real-time ATP rate assay results were expressed as the calculated percentage contribution of glycolysis and OXPHOS to total ATP production, as well as the ratio between mitochondrial and glycolytic ATP production rates. Mito Stress Test results were expressed as fold change relative to the parental (P) cell line. Data from at least three independent biological experiments were reported and statistically analyzed.

Results are presented as mean ± SE of at least three biological replicates. Statistical significance was determined using two-tailed Student’s t-test or one-way analysis of variance (ANOVA) for multiple comparisons, as appropriate. When ANOVA was performed, Tukey’s post hoc test was used for pairwise comparisons. Data analyses were carried out using GraphPad Prism software (version 7.0), and statistical significance was defined as P < 0.05.

TRAP1 is emerging as a crucial target for cancer therapy, due to its expression in diverse neoplastic tissues, together with its expression in different cancer cell lines, including PDAC cell models [Supplementary Figure 1A]. We evaluated TRAP1 protein expression by Western blotting in five different PDAC cell lines (MiaPaca2, Panc1, HS776T, Suit2, and PaCa3; Supplementary Figure 1B). While all cell lines displayed TRAP1 expression, Suit2 cells exhibited the highest level. However, as Suit2 cells failed to generate well-dedifferentiated CSCs in our models, we focused our analysis on Panc1 cells, the most thoroughly characterized CSC model in our laboratory. To investigate the role of TRAP1 in Panc1-derived CSCs, we analyzed its protein expression during progressive de-differentiation at three time points: 2 weeks (2 w), 4 weeks (4 w), and 8 weeks (8 w) of culture in the specific stem culture medium. We have previously shown that these time points display a progressive stem phenotype and different metabolic properties^[4,16]. After 2 weeks in culture, CSCs exhibit increased glycolytic activity^[4] and increased TRAP1 expression levels compared to both parental (P) cells [Figure 1A and B]. Based on these findings, we stably knocked out TRAP1 in Panc1 P cells and subsequently derived the relative CSCs [Figure 1A]. Both P cell types (with or without TRAP1) exhibited a similar morphology. However, TRAP1-KO CSCs generated smaller and fewer spheroids, characterized by less defined borders than their wild-type counterparts [Figure 1C]. These differences were observed up to 8 weeks in culture, indicating a long-lasting effect of TRAP1 on the CSC phenotype. Given that TRAP1 expression is maximally induced after 2 weeks in culture, we focused on this time point its role in modulating stemness.

Analysis of cell proliferation [Figure 1D and Supplementary Table 1], stem marker expression (ZEB1, SOX2, NANOG, and OCT3/4; Figure 1E), and cell viability after exposure to four different chemotherapy agents [Figure 1F] showed no differences between P cells and CSCs, regardless of TRAP1 expression. Nonetheless, TRAP1-KO CSCs had a slower proliferative capacity, an increase in stem marker expression, and increased chemotherapy resistance compared to parental cells.

As TRAP1-KO CSCs were unable to form spheroids, we analyzed the messenger RNA (mRNA) levels of various genes involved in cell adhesion, such as CD44, CD29, MUC-4, MUC-16, CDH1, CDH3, and E-selectin. Among these, CD44 expression remained unchanged. In contrast, CD29, MUC-4, and MUC-16 were regulated similarly in both wild-type and TRAP1-KO CSCs compared to P cells; specifically, CD29 expression decreased while MUC-4 and MUC-16 increased. This suggests that their modulation depends primarily on the acquisition of stem properties [Figure 2A]. Notably, TRAP1 ablation increased the expression of CDH1, CDH3, and E-selectin in P cells [Figure 2A].

The CDH1 gene, which encodes E-cadherin, deserves further consideration. Cells lacking TRAP1 expression upregulated both CDH1 mRNA [Figure 2A] and E-cadherin protein levels [Supplementary Figure 2A]. In addition, CDH1 mRNA expression was low in Suit2 cells [Supplementary Figure 2B], supporting an inverse correlation between TRAP1 levels and CDH1 expression.

Previous studies have reported that TRAP1 downregulates the enzymatic activity of succinate dehydrogenase, leading to the accumulation of the oncometabolite succinate^[17]. Therefore, we analyzed whether CDH1 expression is regulated in a succinate-dependent manner. We treated both P and TRAP1-KO cells with varying concentrations of succinate and analyzed CDH1 expression levels. As shown in Figure 2B, the expression levels of CDH1 decreased significantly in Panc1 TRAP1-KO cells treated with succinate, while no changes in CDH1 expression were observed in TRAP1-expressing P cells. These data indicate that TRAP1 decreases CDH1 expression via a succinate-dependent mechanism.

Since TRAP1 is a molecular chaperone involved in maintaining mitochondrial homeostasis^[9], we investigated whether mitochondrial morphology, which is closely linked to functional activity^[18], undergoes TRAP1-dependent changes. Transmission electron microscope (TEM) analysis revealed that mitochondria undergo extensive structure remodeling in P cells in the absence of TRAP1 [Figure 3A]. Specifically, TRAP1-KO mitochondria exhibit a significant increase in area, aspect ratio (major-to-minor axis length ratio), and cristae width [Figure 3B].

In dedifferentiated cells, the mitochondria of TRAP1-KO CSCs showed greater cristae extension and reduced crista junction diameter compared with those of P cells. Notably, a comparison between P cells and CSCs, both lacking TRAP1, further confirms these differences, alongside a significantly lower aspect ratio and cristae width [Figure 3B].

From a molecular perspective, mitochondrial dynamics is tightly controlled by specific proteins, such as DRP1 and MFF that regulate mitochondrial fission, and OPA1, MFN1, and MFN2 that elicit mitochondrial fusion^[19]. As reported in Figure 4 and consistent with our previous work^[16], the expression of the active form of DRP1 (p-DRP1) is decreased in TRAP1-expressing CSCs compared to their parental counterpart. This is in line with the increased levels of OPA1 observed in these cells (Figure 4B, green histograms). In contrast, when TRAP1 is deleted, the expression of these proteins remains unchanged, whereas MFF is strongly down-regulated in TRAP1-KO parental cells. Conversely, MFF expression is strongly increased in TRAP1-KO CSCs after 2 weeks in culture (Figure 4B, orange histograms), corroborating the TEM observations. Regarding the other proteins involved in mitochondrial fusion, Figure 4 shows that MFN1 is regulated only in TRAP1-KO cells, whereas MFN2 levels remain consistent among all cell types. Altogether, these data show that ablation of TRAP1 in parental cells is associated with more elongated mitochondria that present larger cristae, whereas the absence of TRAP1 in CSCs prompts mitochondria fragmentation and extension of tightly connected cristae, consistent with MFF upregulation.

To evaluate the effect of TRAP1 on cell bioenergetics, we analyzed the ATP production rates from distinct metabolic pathways. While total ATP production rate remained unchanged across P cells, TRAP1-KO cells, and CSCs 2w TRAP1-KO, the energetic source shifted significantly towards OXPHOS utilization when TRAP1 was knocked out [Figure 5A, Supplementary Figure 3A and B]. Indeed, TRAP1 ablation in P cells significantly decreased the glycolytic ATP production rate while enhancing OXPHOS-derived ATP; this effect was further enhanced in de-differentiated TRAP1-KO CSCs [Figure 5A]. These data suggest that TRAP1 loss increases mitochondrial respiration, while limiting the glycolytic flux. To validate this metabolic rewiring, we assessed mitochondrial OXPHOS using a standard Mitostress Test [Figure 5B and Supplementary Figure 3C]. Despite comparable basal respiration, the mitochondrial ATP-linked respiration was significantly higher in both P and CSC TRAP1-KO cells. Moreover, both TRAP1-KO cell lines exhibited a stronger FCCP-stimulated increase in maximal respiration, indicating an enhanced ability to utilize mitochondrial oxidative capacity under high energetic demand. Notably, this effect was significantly more pronounced in TRAP1-KO CSCs compared to TRAP1-KO P cells, correlating with the thicker cristae observed by TEM in the former cells [Figure 3B]. Finally, the reduced proton leak and higher coupling efficiency in TRAP1-deficient cells [Figure 5B] further support a more tightly coupled mitochondrial state and a lower propensity for mitochondrial membrane potential dissipation.