Mitochondrial Ca²⁺ controls pancreatic cancer growth and metastasis by regulating epithelial cell plasticity

SUMMARY

Endoplasmic reticulum to mitochondria Ca²⁺ transfer is important for cancer cell survival, but the role of mitochondrial Ca²⁺ uptake through the mitochondrial Ca²⁺ uniporter (MCU) in pancreatic ductal adenocarcinoma (PDAC) is poorly understood. Here, we show that increased MCU expression is associated with malignancy and poorer outcomes in patients with PDAC. In isogenic murine PDAC models, Mcu deletion (Mcu^KO) ablated mitochondrial Ca²⁺ uptake, which reduced proliferation and inhibited self-renewal. Orthotopic implantation of MCU-null tumor cells reduced primary tumor growth and metastasis. Mcu deletion reduced the cellular plasticity of tumor cells by inhibiting epithelial-to-mesenchymal transition (EMT), which contributes to metastatic competency in PDAC. Mechanistically, the loss of mitochondrial Ca²⁺ uptake reduced the expression of the key EMT transcription factor Snail and secretion of the EMT-inducing ligand TGF-β. Snail re-expression and TGF-β treatment rescued deficits in Mcu^KO cells and restored their metastatic ability. Thus, MCU may present a therapeutic target in PDAC to limit cancer-cell-induced EMT and metastasis.

Graphical Abstract

In brief

Weissenrieder et al. show that mitochondrial Ca²⁺ uniporter (MCU) deletion reduces growth and metastasis of murine KPCY cells by reducing cell-autologous induction of epithelial-to-mesenchymal transition (EMT). Induction of EMT with Snail expression or TGF-β treatment rescues growth and metastatic deficits. These findings have implications for therapeutic development in PDAC.

INTRODUCTION

Pancreatic cancer is one of the most lethal cancers in the United States, with the most common form, pancreatic ductal adenocarcinoma (PDAC), having a 5-year survival rate of only ~13%.^1–3 Patient treatment is hampered by late diagnosis, early metastasis, and poor treatment responses. The vast majority of patients with PDAC present with metastatic disease, which is often the cause of death.⁴ PDAC is generally heterogeneous, refractory to most treatments, and driven by currently un-targetable driver mutations, although some progress has been made with specific KRAS mutations.⁵ Thus, while targeted therapies have vastly improved survival in many other malignancies, such as breast and prostate cancer, the standard of treatment for PDAC remains resection and cytotoxic chemotherapy.^1,5

Genetic regulators of the metastatic cascade remain largely undefined, suggesting that non-genetic cellular plasticity contributes to the underlying biological processes driving metastasis. Such plasticity changes tumor cell biology to alter metabolic requirements, responses to chemotherapeutics, growth rates, and responses to the immune system to promote cell survival, growth, and metastasis. Given the early dissemination of tumor cells in PDAC development, cellular plasticity is of particular interest. Epithelial-to-mesenchymal transition (EMT) has emerged as a means for tumor cells to gain pro-metastatic features through a progressive loss of epithelial markers, such as E-cadherin (ECAD), and an increase in mesenchymal markers, such as N-cadherin.^6–10 Classically, this transition is mediated by multiple transcription factors, including Snail, Slug, and Twist^6,7,9,10 downstream of signaling pathways including TGF-β^8,11. These EMT transcription factors actively repress the epithelial program and promote a pro-invasive mesenchymal phenotype that facilitates metastasis. Recently, lineage-labeled genetically engineered mouse models (GEMMs) of PDAC have revealed that partial or hybrid EMT states, where tumor cells co-express both epithelial and mesenchymal genes, are prevalent in pancreatic cancer.¹² Distinct from classical or complete EMT, these partial EMT states are regulated by protein localization, metabolism, and second messenger signaling.^12–16 Notably, EMT has been linked to metabolic alterations and enhanced Ca²⁺ signaling.^6,14,17–20 Despite observations that cytoplasmic Ca²⁺ signaling promotes EMT, little is known regarding the mechanisms that link intracellular Ca²⁺ homeostasis to EMT.

While Ca²⁺ signaling is known to strongly affect many cancer-related phenotypes, the role of mitochondrial Ca²⁺ signaling in PDAC is poorly understood.¹⁴ Ca²⁺ signaling in other cancers has been implicated in pro-tumor phenotypes, including therapeutic resistance and modulation of cellular identity through plasticity events such as EMT.^14,21–23 Many cancer cells appear to be “addicted” to Ca²⁺ flux from the endoplasmic reticulum (ER) to mitochondria, which may represent a therapeutic vulnerability.²⁴ Canonically, this signaling occurs at mitochondria-associated membranes (MAMs), where Ca²⁺ released by ER-localized inositol 1,4,5-trisphosphate receptors (IP₃Rs) is taken up by the mitochondrial Ca²⁺ uniporter (MCU) channel complex. Critically, when MCU is lost, this uptake does not occur.²⁵ The Ca²⁺ released by IP₃Rs is rapidly taken up by MCU in a quasi-synaptic manner at MAMs due to the large electrochemical gradient of the mitochondrial inner membrane (membrane potential ~ −150 to −180 mV) and the close apposition of ER and mitochondrial membranes at these sites (10–25 nm).^2,26 This close apposition allows for ER Ca²⁺ release by constitutive low-level openings of IP₃Rs and their activation in response activation of phospholipase C-coupled receptors to result in changes of mitochondrial [Ca²⁺] that regulate mitochondrial function. Previous reports have suggested that PDAC cells may depend on this flux to resist metabolic stress, since loss of MCU creates a dependency on cystine in human PDAC cells through an antioxidant-related pathway.^24,27 Here, we provide evidence that targeting mitochondrial Ca²⁺ uptake has therapeutic value in PDAC. We observe profound effects of MCU expression on PDAC tumor cell plasticity, survival, growth, and metastasis in vivo and in vitro, and elucidate a novel relationship between MCU and EMT.

RESULTS

MCU is upregulated in human and murine pancreatic cancers

We examined tissue and publicly available datasets to identify links between MCU expression, tumorigenesis, and patient outcomes. Consistent with previously reported oncogenic functions for MCU in other cancers,^{24,25,27–32} MCU protein expression is highly upregulated in PDAC tumor cells compared with normal tissue (Figure 1A), and higher MCU gene expression is associated with poorer survival outcomes in the TCGA-PAAD (The Cancer Genome Atlas-Pancreatic Adenocarcinoma) cohort (Figure 1B). Similarly, MCU complex components Micu1 and Smdt1 (EMRE) mRNA levels are increased in KPC organoids (Figures S1A and S1B). In the TCGA cohort, survival rates were generally not affected by other MCU complex members (data not shown). Analysis of publicly available single-cell RNA sequencing (GEO: GSE229413) from patients with PDAC^33,34 indicated that MCU complex members were more highly expressed in PDAC-like cells than in normal-type ductal or acinar-like populations (Figures S1C and S1D). MCU itself was most highly expressed in PDAC cells compared with cells from normal-type ductal or acinar populations (Figures S1E and S1F). Higher MCU expression in pancreatic tissue is correlated with KRAS mutations, the most common driver mutations in PDAC (Figure 1C). Human PDAC cell lines show faster rates of mitochondrial Ca²⁺ uptake when exposed to 2.5μM Ca²⁺ in a permeabilized cell assay compared with normal human pancreatic ductal epithelial control cells (Figures 1D and 1E), indicating increased MCU activity in PDAC. Permeabilized cell assays directly measure mitochondrial Ca²⁺ uptake activity by reducing confounding effects of varied expression or function of upstream proteins (Gq-coupled receptors, IP₃R, etc.) and Ca²⁺ buffering which complicate the analysis of whole-cell, agonist-based stimulation protocols. These findings are consistent with previous reports suggesting that cancer cells may be addicted to ER-to-mitochondrial Ca²⁺ uptake²⁴ and that they may be more tolerant of higher mitochondrial [Ca²⁺], with implications for apoptosis resistance.¹⁴ Together, these support the notion that MCU is a putative oncogenic driver that may facilitate tumorigenesis in patients with PDAC.

Figure 1. MCU expression and function are associated with malignancy in human and murine PDAC.

(A) MCU is highly expressed in human PDAC tissue but not in normal pancreas, as seen by immunohistochemistry. Scale bars, 100 μm.

(B) High MCU mRNA expression is associated with poor survival outcomes in the TCGA-PAAD cohort. Cohorts were split at the 50th percentile, and log rank p = 0.0477 via Kaplan-Meier survival analysis.

(C) High MCU mRNA expression is associated with Kras mutations in the TCGA-PAAD cohort, obtained from cBioPortal (n = 87 per group).

(D) Mitochondrial Ca²⁺ uptake is more rapid and complete in human cancer cell lines Panc-1 and MiaPaCa-2 compared with “normal” human pancreatic ductal epithelial (HPDE) control cells. Assays carried out in biological triplicate.

(E) Quantification of mitochondrial Ca²⁺ uptake rates in (D).

(F) Immunofluorescence imaging of tissues from normal-type CY mice, pancreatic intra-epithelial neoplasia lesion-developing KCY mice, and PDAC tumor-bearing KPCY mice show high MCU expression in KCY and KPCY tissues, particularly in tumor lesions of KPCY. Scale bars, 100 μm. n = 3 per group.

(G) Mcu mRNA expression is increased in KC and KPC organoids over wild-type ductal organoids in a publicly available dataset from Tuveson and co-workers (GEO: GSE63348, n = 6–7 per group).³⁷ Statistical analysis of survival is by Kaplan-Meier analysis, while two-group analyses were carried out with Student’s t test. Data with three groups were analyzed with one-way ANOVA with Dunnett’s post hoc. Bars represent standard error of mean (SEM). *p ≤ 0.05, **p ≤ 0.01, ***p ≤0.001.

To gain a deeper understanding of when Mcu expression is turned on during PDAC progression, we stained for MCU in a mutant Kras- and gain-of-function Tp53-driven PDAC GEMM (Kras^LSL-G12D/+; Trp53^LSL-R172H/+; Pdx1-Cre; R26^{LSL-Yfp/LSL-Yfp}, “KPCY” mice). This mouse model is driven by mutations in Kras and Trp53, the two most common driver mutations in human PDAC that are mutated in 90% and 75% of patients, respectively.¹ The Cre-inducible Rosa26-LSL-Yfp allele labels tumor cells of the Pdx1 lineage, enabling identification of PDAC tumor cells of epithelial origin.¹⁵ Normal (exocrine acinar and endocrine islet) cells from Pdx1Cre; R26^LSL-Yfp (or “CY”) mice express appreciable levels of MCU (Figure 1F), consistent with previous reports.^25,35,36 MCU expression is upregulated in pancreatic intra-epithelial neoplasia lesions from Kras^G12D/+; Pdx1-Cre; R26^{LSL-Yfp/LSL-Yfp} (KCY) mice and in YFP⁺ PDAC tumor cells from KPCY mice (Figure 1F). Notably, YFP-negative stromal cells from KPCY mice express less MCU (Figure 1F) than YFP-positive tumor cells (Figure 1F), demonstrating that tumor cells upregulate MCU during tumorigenesis. Consistently, Mcu mRNA expression was significantly elevated with increased malignancy in a previously published RNA sequencing dataset (GEO: GSE63348)³⁷ comparing organoids developed from the pancreas of WT, KC, and KPC mice (Figure 1G). Taken together, these data demonstrate that MCU expression and mitochondrial Ca²⁺ uptake are upregulated in tumor cells from human patients with PDAC, a phenomenon that is recapitulated in the KPCY murine model of PDAC.

MCU promotes malignant properties of PDAC cells in vitro

Since MCU expression is associated with malignant phenotypes in human and murine models, we employed isogenic murine models of Mcu^KO to assess the role of mitochondrial Ca²⁺ signaling in pancreatic cancer development, growth, and metastasis. Knockout of this single gene prevents the function of the MCU complex, ablating Ca²⁺ uptake into the mitochondria in response to increases in Ca²⁺ near the channel.²⁵ Cell lines were generated from Kras^G12D/+; Tp53^R172H/+; Pdx1Cre; R26^{LSL-Yfp/LSL-Yfp}; Mcu^loxP/loxP (KPCY-Mcu^Cre-KO) mice and MCU was re-expressed at physiologically relevant levels (i.e., KPCY-Mcu^rescue; Figure 2A). The fidelity of this knockout and re-expression system was verified by western blot analysis of clonal cell lines from each genotype (Figure 2B, and long exposure, S1G). KPCY-Mcu^rescue cells express V5- and His-tagged MCU at similar or reduced levels to endogenous MCU from a previously generated KPCY murine tumor cell line, 2838.c3.³⁸ As expected, mitochondria in KPCY-Mcu^Cre-KO cells failed to take up Ca²⁺ from the bath, in contrast to those in KPCY-Mcu^rescue cells (Figure 2C), indicating that re-expressed MCU is functional. KPCY-Mcu^Cre-KO cells had reduced proliferation rates compared with KPCY-Mcu^rescue cells (Figure 2D, ~50% reduction), as well as reduced wound healing (Figure 2E), spheroid formation (Figure 2F), transwell migration (Figure 2G), and transwell invasion (Figure 2H). Strikingly, KPCY-Mcu^Cre-KO cells were nearly incapable of forming spheroids in anchorage-independent growth conditions, suggesting a lack of self-renewal capacity. Notably, even low levels of MCU expression were associated with rescued phenotypes (Mcu^rescue clone 9, e.g.,), highlighting the importance of this uptake activity in cancer cells.

Figure 2. Genetic deletion of Mcu inhibits mitochondrial Ca²⁺ uptake and reduces growth and motility phenotypes in vitro.

(A) Schematic of the development of isogenic Mcu^Cre-KO and Mcu^rescue cell lines from KPCY-Mcu^Cre-KO murine ductal cells.

(B) Mcu^Cre-KO clones express no MCU that is restored by stable re-expression of MCU-V5-His; 2838.c3, a KPCY-Mcu^WT cell line, included as positive control. Pair of bands seen in Mcu^rescue cell lines due to partial degradation of the His tag.

(C) Mitochondrial Ca²⁺ uptake is ablated in Mcu^Cre-KO cells and restored by stable expression of Mcu.

(D) Mcu^Cre-KO cells proliferate more slowly than paired Mcu^rescue isogenic cell lines.

(E) Eighteen-hour wound-healing assay of Mcu^Cre-KO and Mcu^rescue cells (20×).

(F) Anchorage-independent spheroid formation. Mcu^Cre-KO spheroids, when present, are small and misshapen, in contrast to large, smooth-textured Mcu^rescue spheroids. Scale bars, 50 μm.

(G) Migration activity in a 24-h transwell assay with FBS-containing medium as chemoattractant.

(H) ECM invasion capacity in 24-h transwell assay using FBS as a chemoattractant.

(I) Schematic of the development of isogenic Mcu^CRISPR-KO and Mcu^WT cell lines from KPCY murine ductal cells.

(J) Mcu^CRISPR-KO clones express no MCU, in contrast to Mcu^WT isogenic controls.

(K) Mitochondrial Ca²⁺ uptake in Mcu^WT cells is ablated in Mcu^CRISPR-KO cells.

(L) Mcu^CRISPR-KO cells proliferate more slowly than Mcu^WT isogenic cell lines.

(M) Twenty-four-hour wound-healing assay of Mcu^CRISPR-KO and Mcu^WT cells.

(N) Anchorage-independent spheroid formation. Mcu^Cre-KO spheroids, when present, are small, fragmented, and irregular, compared with large, smooth-textured Mcu^WT spheroids. Scale bars, 100 μm.

(O) Transwell migration activity.

(P) ECM invasion activity. N > 3 per experiment. Cell count experiments analyzed with two-way ANOVA with Sidak’s post hoc. One-way ANOVA with Sidak’s post hoc was employed for all other experiments. Bars represent SEM. Box and whisker plots in (F) and (N) are min, 1st quartile, median, 3rd quartile and max. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

To ensure that these observed phenotypes were not due to compensatory mechanisms in response to the in vivo knockout of MCU, we also developed an isogenic CRISPR-KO model of a KPCY-Mcu^WT cell line, 2838.c3 (Figure 2I). To reduce potential off-target effects of the expression of Cas9, the vector was only transiently expressed before single-cell clones were isolated. KPCY-Mcu^CRISPR-KO cells expressed no MCU protein (Figure 2J) and lacked mitochondrial Ca²⁺ uptake (Figure 2K). Consistent with the Cre-mediated Mcu knockout and rescue models, the deletion of Mcu with CRISPR strikingly reduced proliferation (Figure 2L), wound healing (Figure 2M), spheroid formation (Figure 2N), migration (Figure 2O), and invasion (Figure 2P). Observed deficits in cell growth (Figure S1H) and motility (Figure S1I) in Mcu^CRISPR-KO cells were largely reversed by transient transfection with MCU, suggesting that these effects are MCU specific and not due to off-target effects. Therefore, we conclude that inhibition of mitochondrial Ca²⁺ uptake by MCU knockout strongly reduces in vitro phenotypes associated with malignancy, metastasis, and invasion in PDAC.

MCU promotes tumor growth and metastasis in murine xenografts

To further examine the function of mitochondrial Ca²⁺ signaling in PDAC development, the behaviors of these isogenic cell lines were interrogated in vivo in orthotopic implantation models of PDAC. Despite being proliferative in vitro, KPCY-Mcu^Cre-KO cells failed to form primary tumors after orthotopic implantation into the pancreas of C57BL/6 mice, in striking contrast to KPCY-Mcu^rescue cells (Figures 3A–3C). Notably, whereas YFP⁺ liver metastases were observed in 80% of animals implanted with KPCY-Mcu^rescue cell lines, none were observed in mice implanted with KPCY-Mcu^Cre-KO cells (Figures 3A and 3D–3F). No differences were observed in overall body weight between the two cohorts (Figure S2A). KPCY-Mcu^Cre-KO cell-injected animals lacked pancreatic or metastatic lesions, in contrast to KPCY-Mcu^rescue cell-injected mice (Figures 3D–3F and S2B). To evaluate the metastatic ability of MCU-null or -expressing PDAC tumor cells, KPCY-Mcu^Cre-KO or KPCY-Mcu^rescue cells were injected into the tail veins of C57BL/6 mice. Similar to the results in the orthotopic implantation assay, KPCY-Mcu^Cre-KO tail vein-injected animals failed to form metastatic colonies, whereas KPCY-Mcu^rescue cells efficiently colonized the lung (Figures 3G–3I). No differences in body weight were observed (Figure S2C). Thus, the lack of metastases seen in the orthotopic model was not solely due to the inability to form a primary lesion.

Figure 3. MCU ablation reduces tumor growth and metastasis in in vivo xenograft models. Mice were injected orthotopically in the pancreas with 100,000 cells (Mcu^Cre-KO or Mcu^rescue) and aged for 21 days, until tumors were palpable and mice began to show symptoms. n = 3–5 mice per group.

(A) Representative bright-field and YFP images of liver and pancreas of C57bl/6J mice injected orthotopically with Mcu^Cre-KO or Mcu^rescue cells.

(B) Quantification of number of mice with tumors (n = 3–5 mice per group).

(C) Total mass of pancreas (n = 3–5 mice per group).

(D) Percent mice with liver metastases (n = 3–5 per group).

(E) Liver mass of mice injected with Mcu^Cre-KO and Mcu^rescue cells (n = 3–5 per group).

(F) Representative images of liver tissue stained with H&E. Scale bar, 100 μm.

(G) Representative bright-field and YFP images of the lungs of C57/bl6J mice injected in the tail vein with 100,000 Mcu^Cre-KO or Mcu^rescue cells and aged for 14 days (n = 5 mice per group).

(H) Quantification of lung colonization in the tail-vein injection model (n = 5 mice per group).

(I) Lung mass from tail-vein injection model.

(J) Representative bright-field and YFP images of pancreas, liver, and lung from C57/bl6 mice orthotopically implanted with 100,000 Mcu^CRISPR-KO or Mcu^WT cells in the pancreas and aged for 13 (n = 3 per group) or 27 days (n = 4 per group).

(K) Quantification of pancreatic mass (n = 3–4).

(L) Percent of mice with liver metastases (n = 3–4).

(M) Representative H&E staining of liver tissue. Scale bars, 100 μm. Number of mice with tumors or metastases were compared with Fisher’s exact test. Tissue masses were compared with Student’s t test. Scale bar, 1 mm. Bars represent SEM. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

As KPCY-Mcu^Cre-KO cells failed to form tumors upon orthotopic implantation, we performed additional experiments in an orthologous model by injecting isogenic clonal cell lines generated after in vitro knockout of MCU (KPCY-Mcu^CRISPR-KO) or control parental lines (KPCY-Mcu^WT). MCU deletion reduced primary tumor burden at 13- and 27-day post-implantation (Figures 3J and 3K). This parental cell line has been characterized previously as having a micrometastatic phenotype, with evidence of YFP⁺ tumor cells in the liver and lung at 27 days post-implantation (Figures 3J, 3L, 3M, and S2D). Metastatic burden was reduced in animals injected with KPCY-Mcu^CRISPR-KO cells, with little to no evidence of YFP⁺ micrometastases by fluorescent dissection microscopy or pathological analysis of H&E images (Figures 3J, 3L, 3M, and S2D). In addition, KPCY-Mcu^CRISPR-KO tumor-bearing mice had fewer ascites and spleen metastases compared with Mcu^WT mice (Figure S2F). Lung lesions were not observed in Mcu^CRISPR-KO mice, whereas occasional small lung metastases (which did not affect total lung weight) were observed in Mcu^WT mice (Figures 3J, S2E, and S2F). Overall body weight was not affected by tumor-specific MCU deletion (Figure S2G). Both isogenic cell lines formed poorly differentiated tumors, with no differences in relative proportions of differentiated area (Figure S2I). The similar results in multiple in vivo models of tumor cell-specific deletion of Mcu suggests that mitochondrial Ca²⁺ uptake significantly supports the growth and metastasis of PDAC tumors.

In the GEMM KPCY-Mcu^Cre-KO model (Figure S3A), we did not observe improvements in survival (Figure S3B), percent of mice with metastases (Figure S3C), or pancreatic mass (Figure S3D) in Mcu^Cre-KO mice compared with KPCY-Mcu^WT mice. Mcu^Cre-KO mice had modestly reduced liver mass (Figure S3E) and no difference in lung mass (Figure S3F). There were trends in KPCY-Mcu^Cre-KO animals toward reduced tumor cell proliferation (Figures S3G–S3I), increase in the apoptosis marker cleaved caspase-3 (Figures S3J–S3L) and decreased number of metastatic tumor cells (Figures S3M–S3O). Collectively, these data indicate that there was a trend toward a reduction in primary and metastatic tumor growth upon the loss of tumor cell Mcu in KPCY animals, but the discrepancy between phenotypes observed in GEMMs versus orthotopic injection models suggests that there exist compensatory mechanisms upon germline in vivo knockout of Mcu.

MCU loss reduces EMT

We observed distinct morphological differences between MCU-KO- and MCU-expressing isogenic cell lines in vitro. MCU-KO cells were more epithelial (semicuboidal, grew in plaque-like formations, and were less motile), whereas MCU-expressing isogenic lines had a more mesenchymal identity (spindle morphology and more motile) (Figure 4A). We posited that these differences could be explained by mitochondrial Ca²⁺-dependent induction of an EMT program, a key pathway implicated in tumor cell plasticity and metastasis.^{6,7,10,14,15,19,20,39} To confirm this shift in cellular identity in vivo, we stained MCU-KO- and MCU-expressing orthotopic tumor tissues for ECAD, a marker of the epithelial state whose loss indicates EMT induction. ECAD expression was markedly reduced in MCU-expressing tumor cells (KPCY-Mcu^WT and KPCY-Mcu^rescue) compared with MCU-KO cells (KPCY-Mcu^CRISPR-KO and KPCY-Mcu^Cre-KO) (Figures 4B and S4A), indicating that mitochondrial Ca²⁺ signaling facilitates EMT in PDAC tumors. Of note, KPCY-Mcu^WT cells expressed basal levels of the key EMT transcription factor, Snai1 (Snail), whereas it was undetectable in Mcu^CRISPR-KO cells (Figure 4C). Snail is known to repress ECAD expression and potently induce EMT downstream of TGF-β signaling, a canonical EMT-inducing signal in cancer.^40,41 To gain further insights into the role of mitochondrial Ca²⁺ signaling in EMT, we interrogated transcriptional differences between MCU-expressing and MCU-KO cells by RNA sequencing (GEO: GSE287625). Unsupervised hierarchical clustering demonstrated that the different cell lines clustered by Mcu expression rather than by parental cell-line source, suggesting a strong effect of MCU expression on overall transcriptional programs (Figure 4D; Table S1). Consistent with the observed shift in epithelial cell identity by cell morphology and ECAD expression in MCU-deficient cells, Gene Ontology and gene set enrichment analysis (GSEA) indicated that EMT was one of the top significantly altered gene sets between Mcu-expressing and Mcu-KO cells, with KO cells having reduced enrichment for EMT genes (Figures 4E and 4F; Tables S2–S5). EMT-related genes clustered strongly by Mcu expression (Figure 4G), and several EMT transcription factors were identified by CHeA3 analysis (Figure S4B). CHeA3 employs geneset libraries from RNA sequencing, ChIP-seq, and co-occurrence datasets to allow for identification of probable transcription-factor activation (as measured by a composite rank) from a gene list of significantly altered genes, more efficiently than other comparable methods.⁴² Similar results were seen in the genetic model, where tumor tissue from KPCY-Mcu^KO mice contained more YFP⁺/ECAD^LOW cells, which marks tumor cells undergoing EMT that have lost their epithelial cell identity and have begun to disseminate (Figures S4C–S4E). Thus, MCU expression strongly promotes an EMT transcriptional program in PDAC. Of possible biological relevance of the EMT phenotype for human patients, approximately 7.8% of PDAC cells in a single-cell RNA sequencing dataset co-express the epithelial marker CDH1 (which codes for ECAD) and the mesenchymal marker, CDH2 (which codes for NCAD) (Figure S4F).^33,34 MCU was detected in the majority of cells regardless of CDH1 and CDH2 status (Figures S4G and S4H), including dual positive CDH1⁺CDH2⁺ cells that likely capture cells undergoing EMT or in a partial EMT state. The high number of apparently MCU-null cells in this dataset is likely due to the low numbers of transcripts generated per cell.

Figure 4. Mcu expression is associated with EMT.

(A) Representative 20× bright-field images of Mcu^CRISPR-KO, Mcu^Cre-KO, and their isogenic control cells. Scale bars, 100 μm.

(B) Representative immunofluorescence images of tissue from the primary lesions of 27-day Mcu^CRISPR-KO and Mcu^WT orthotopic injections from Figures 3J–3M. Scale bars, 100 μm. n = 3 per group. Mcu^WT cells express little ECAD, suggesting extensive EMT has occurred.

(C) Western blot of Snail in Mcu^CRISPR-KO and Mcu^WT cells.

(D) Heatmap of RNA sequencing genes (as Z score). When unsupervised hierarchical clustering is applied to Mcu-knockout and Mcu-expressing isogenic cell lines, they independently group into Mcu^KO and Mcu^rescue groups, suggesting that Mcu expression strongly influences transcriptional regulation. Deposited at GEO: GSE287625.

(E) GSEA enrichment pathways by normalized enrichment score (NES), colored by false discovery rate (FDR).

(F) GSEA enrichment analysis plot for EMT gene set, indicating upregulation in Mcu^WT cells compared with Mcu^CRISPR-KO cells.

(G) Heatmap of top 30 leading-edge genes from GSEA of EMT genes in Mcu^WT and Mcu^CRISPR-KO clones, showing clear differences between groups (results shown as Z score).

TGF-β and Snail rescue MCU^KO phenotypes in vitro

To confirm the induction of an EMT program in MCU-expressing tumor cells, we examined protein expression levels of key EMT pathway components. As noted, under basal, untreated conditions, Mcu^WT cells expressed the EMT transcription factor Snail at higher levels than in Mcu^CRISPR-KO cells (Figure 4C). Higher Snail expression in Mcu^WT cells was associated with increased levels of secreted TGF-β, a known inducer of EMT, in the cell culture medium (Figure 5A), despite similar expression of Tgfb1 transcripts (Figures S5A–S5C). The reduced secretion of TGF-β in Mcu^CRISPR-KO was accompanied by a paradoxical increase in intracellular TGF-β levels (Figure 5B), suggesting a potential dysregulation of secretion machinery upon loss of Mcu. A cytokine array assay of the media of Mcu^WT and Mcu^CRISPR-KO cells indicated that the levels of many secreted factors was lower in the absence of MCU (Figure 5C). We speculated that the promotion of TGF-β secretion by MCU-mediated mitochondrial Ca²⁺ uptake could mechanistically link MCU expression with increased Snail expression and possibly EMT. To test the hypothesis that Mcu^CRISPR-KO cells expressed lower levels of EMT markers as a consequence of lower secretion of EMT-induction factors, we used TGF-β treatment and stable Snai1 overexpression (Snail^OE), both well-characterized orthogonal methods of EMT induction, in the isogenic Mcu^WT and Mcu^CRISPR-KO cell lines. Treatment with TGF-β (10 ng/mL for 72 h) reduced the epithelial cell marker ECAD and increased the expression of mesenchymal markers N-cadherin, vimentin, and Snail, independent of Mcu status (Figure 5D). Of note, Mcu^CRISPR-KO cells expressed higher levels of the TGF-β receptor (Tgfbr1) transcripts and similar levels of Tgfbr2 (Figure S5D) compared with Mcu^WT cells. Furthermore, similar EMT phenotypes were induced in both cell lines by stable Snai1 overexpression (Figures 5E and 5F). Thus, functional EMT machinery remains intact in Mcu^CRISPR-KO cells, despite their lower levels of basal Snail expression and absence of EMT phenotypes.

Figure 5. Exogenously induced EMT ameliorates many deficits seen in Mcu^CRISPR-KO cells.

(A) Mcu^WT cells secrete more TGF-β into the medium, as measured by ELISA of the medium at 48 h.

(B) Intracellular TGF-β is increased in Mcu^CRISPR-KO cells compared with Mcu^WT after 10,000 cells were incubated for 48 h in complete media, measured by ELISA.

Values normalized to cell count.

(C) After 48 h incubation, secreted factors were decreased in media supernatants from Mcu^CRISPR-KO cells (Eve Technologies MD32 panel). Data shown as Z score.

(D) Exogenous treatment with 10 ng/mL TGF-β for 72 h increases N-cadherin and vimentin protein levels and reduces E-cadherin levels independent of MCU expression.

(E and F) Stable Snai1 expression increases Snail and N-cadherin expression and reduces E-cadherin levels in Mcu^WT cells (E) and Mcu^CRISPR-KO cells (F).

(G) Flow cytometry plot for ECAD surface expression of untreated Mcu^WT and Mcu^CRISPR-KO cells indicates that Mcu^CRISPR-KO cells express more ECAD.

(H) Flow cytometry plots for Mcu^CRISPR-KO and Mcu^WT cells treated with 10 ng/mL TGF-β or stable Snail expression.

(I) Quantification of % Ecad_low cells stably expressing Snail (from H).

(J) Quantification of %Ecad_low cells treated with 10 ng/mL TGF-β (from H).

(K) Snail expression increases cell growth of Mcu^CRISPR-KO cells to levels comparable with those of Mcu^WT cells.

(L) Snail overexpression increases 24-h transwell migration of Mcu^CRISPR-KO cells to a level intermediate between Mcu^WT cells with or without Snail overexpression.

(M) Seventy-two-hour cell counts indicate that TGF-β increases Mcu^CRISPR-KO cell growth to levels comparable with Mcu^WT cells treated with TGF-β. Treatment with a TGF-β neutralizing antibody (anti-TGF-β) significantly reduced proliferation in Mcu^WT, but not Mcu^CRISPR-KO cells. Cell count data are analyzed with two-way ANOVA with Sidak’s post hoc. Two group data are analyzed with Student’s t test. Bars represent SEM. Box and whiskers plot in (A) are min, 1st quartile, median, 3rd quartile, and max. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

Western blot analysis of whole-cell lysates, which measures total ECAD protein expression throughout the cell, gives an incomplete picture of epithelial versus mesenchymal identity. Rather, membranous ECAD is important for maintaining epithelial cell identity and loss of ECAD from the membrane is associated with the EMT phenotype.^6,14,19,20 Flow cytometric quantification of surface ECAD in non-permeabilized cells revealed that Mcu^CRISPR-KO cells have higher baseline surface ECAD expression than Mcu^WT cells (Figure 5G), consistent with their more epithelial nature. Induction of EMT by overexpression of Snai1 or TGF-β treatment more strongly reduced surface ECAD levels in Mcu^WT cells relative to Mcu^CRISPR-KO cells (Figures 5H–J). Collectively, these results indicate that MCU-expressing cells are poised to undergo EMT during tumorigenesis through enhanced expression of EMT transcription factors, and that inhibition of mitochondrial Ca²⁺ uptake abrogates the ability of pancreatic tumor cells to lose their epithelial cell identity, with implications for their metastatic ability.

Remarkably, Snai1 overexpression rescued Mcu^KO-associated deficits in tumor cell clonogenicity (Figure S5E), proliferation (Figure 5K), wound healing (Figure S5F), and transwell migration (Figure 5L) to levels comparable with those of Mcu^WT cells. TGF-β treatment also increased cell proliferation in both groups (Figure 5M). TGF-β neutralizing antibody reduced unstimulated growth of Mcu^WT cells but had no effect on basal proliferation of Mcu^CRISPR-KO cells (Figure 5M), consistent with higher TGF-β secretion in Mcu^WT cells (Figure 5A) and indicative of divergent TGF-β signaling upon loss of Mcu. Similarly, wound healing was also increased by TGF-β treatment in both Mcu^CRISPR-KO and Mcu^WT cells, and treatment with the TGF-β neutralizing antibody reduced this phenotype only in the Mcu^WT cells (Figure S5G). In the Mcu^Cre-KO and Mcu^rescue models, TGF-β stimulation increased the proliferation of only the MCU-expressing cells (Figure S5H), while transient transfection with Snail increased growth in all contexts (Figure S5I). However, both TGF-β treatment and Snail expression were able to increase wound healing (Figures S5J and S5K) and transwell migration (Figures S5L and S5M) of Mcu^Cre-KO cells, albeit at lower levels. This reduced response rate may be due to the transient nature of the treatments, or it could suggest that the germline knockout of Mcu may have induced intrinsic alterations in the ability of the cells to respond to EMT-inducing signals. MCU expression may alter reactive oxygen species (ROS) production and stress responses through transcription factors such as HIF1α⁴³ and NRF2.²⁷ MCU knockout enhanced basal ROS production (Figure S5N), whereas HIF1α and NRF2 protein levels were similar in MCU-expressing and KO cells (Figure S5O). Similarly, these transcription factors were not identified as activated by CheA3 analysis (Figure S4B). Overall, these results suggest that inhibition of mitochondrial Ca²⁺ uptake reduces malignant phenotypes, in part through reducing cell-autologous secretion of pro-tumorigenic signals, but loss of MCU does not prevent the responses to these signals. Furthermore, EMT induced by stable Snail expression or TGF-β can induce key malignant phenotypes in tumor cells lacking MCU-mediated mitochondrial Ca²⁺ signaling, highlighting molecular redundancy in the EMT pathway.

Snail expression rescues Mcu^CRISPR-KO phenotypes in vivo

To determine if the phenotypic changes observed in vitro upon Snail overexpression were maintained in vivo, we implanted Mcu^CRISPR-KO cells stably expressing Snail or empty vector (EV) into the pancreas of C57BL/6 syngeneic mice. This model was chosen because, in contrast to the Mcu^Cre-KO cells, the Mcu^CRISPR-KO cells successfully form tumors, the MCU-KO phenotypes were directly related to MCU expression in these cells as evidenced by addback experiments, and the approach more closely models inhibition of MCU after a tumor has formed as a possible therapeutic strategy. Snail expression in the Mcu^CRISPR-KO cells increased the primary tumor burden (Figures 6A and 6B) and enhanced metastatic ability (Figure 6C). Increased tumor and metastatic burden were associated with decreased surface ECAD expression in YFP⁺ tumor cells (Figure 6D), indicative of tumor cells undergoing EMT and entering into the metastatic cascade. Continued expression of ECAD in adjacent YFP-negative wild-type ductal cells in the Snail^OE tumors (Figure 6D) indicates that cell-intrinsic mechanisms within the tumor cells mediate EMT induction. The loss of ECAD expression and appearance of more mesenchymal morphology in histological sections strongly indicates that overexpression of Snail robustly induces EMT in Mcu^CRISPR-KO cells in vivo. Therefore, inhibition of mitochondrial Ca²⁺ signaling appears to reduce the propensity of cells to cell-autologously induce EMT, but not the ability of cells to respond to EMT induction from exogenous sources. Snail expression was able to induce resistance of Mcu^WT cells, but not Mcu^CRISPR-KO cells, to commonly used chemotherapeutics gemcitabine (Figures 6E and 6F) and 5-fluorouracil (Figures 6G and 6H). Indeed, Mcu knockout appears to sensitize cells to these agents in a manner that cannot be reversed by Snail expression, highlighting a potential for targeting MCU therapeutically.

Figure 6. Snail is sufficient to drive tumor growth and metastasis in MCU-deleted tumors.

(A) Representative bright-field (BF) and YFP images of pancreas and liver from C57/bl6J mice orthotopically implanted with 100,000 Mcu^CRISPR-KO+EV or Mcu^CRISPR-KO+Snail^OE cells for 21 days (n = 3 per group). Scale bar, 1 mm

(B) Pancreatic mass from orthotopic injection model (n = 3 per group).

(C) Percent of mice with metastases (n = 3 per group). (D) Representative immunofluorescence images of Mcu^CRISPR-KO+EV or Mcu^CRISPR-KO+Snail^OE cells stained for YFP (lineage tracer), DAPI (nuclear marker), and ECAD (epithelial marker). Mcu^CRISPR-KO+EV cells show robust staining of ECAD in YFP-expressing tumor cells, but YFP⁺ Mcu^CRISPR-KO+Snail^OE cells only poorly co-express ECAD (n = 3 per group). A resident, normal-type duct is shown to demonstrate that epithelial cells in the host robustly express ECAD. Scale bars, 100 μm.

(E) After 24-h treatment, 300 nM gemcitabine (GemC) reduces cell proliferation more effectively in Mcu^KO than in Mcu^WT cells, but Snail has a protective effect only in Mcu^WT. Data presented as percent of vehicle control.

(F) Western blot for caspase-3, cleaved caspase-3, and tubulin (loading control). Cells treated with 300 nM or 1 μM gemcitabine for 24 h showed evidence of apoptosis, as measured by caspase-3 cleavage.

(G) After 48-h treatment, 10 μM 5-fluorouracil gemcitabine reduces cell proliferation more effectively in Mcu^KO than in Mcu^WT cells, but Snail has a protective effect only in Mcu^WT cells. Data presented as percent of vehicle control.

(H) Western blot for caspase-3, cleaved caspase-3, and tubulin (loading control). Cells treated with 1 μM 5-fluorouracil (5-FU) for 48 h showed evidence of apoptosis, as measured by caspase-3 cleavage. Two-group data analyzed with Student’s t test, and proportion data analyzed by chi-squared test. Cell proliferation data analyzed with two-way ANOVA with Sidak’s post hoc.Bars represent SEM. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

DISCUSSION

Here, we identify mechanisms by which the mitochondrial Ca²⁺ influx channel MCU supports oncogenic and pro-metastatic functions of pancreatic tumor cells in murine KPCY models of PDAC (Figure 7). MCU deletion reduces PDAC cell motility, clonogenicity, and proliferation in vitro and tumor growth and metastasis in vivo. Mechanistically, inhibition of mitochondrial Ca²⁺ uptake restricts tumor cell plasticity by reducing EMT and entry into the metastatic cascade, ultimately blocking their ability to colonize distant niches.

Figure 7.

Schematic of effects of MCU deletion in the presence and absence of exogenous EMT induction with stable Snail overexpression or TGF-β treatment. Graphic generated in BioRender

Mitochondrial Ca²⁺ homeostasis plays dual roles in cells. Excessive mitochondrial uptake can result in Ca²⁺ overload and cell death by apoptotic and necrotic mechanisms. Conversely, mitochondrial Ca²⁺ is a critical control mechanism for the regulation of basal bioenergetics and for enhanced energy production during periods of increased metabolic demands that are likely encountered during multiple steps of the metastatic cascade. Cancer cells may have a “Goldilocks zone” for Ca²⁺ signaling through MCU, wherein sufficient influx is necessary to support proliferation and metabolism, but excessive signaling contributes to toxicity. This set-point may differ significantly based on cell type, environmental conditions, driver mutations, or expression levels of other proteins within the signaling pathway. Such differences could contribute to observed variability in the effects of MCU expression on survival and tumor growth in different cancers. For example, according to TCGA datasets, high MCU expression levels in melanoma and kidney tumor tissues are associated with enhanced patient survival, in contrast to the opposite associations observed in PDAC, liver, and breast cancer patients.⁴⁴ We previously showed that inhibition of ER-to-mitochondrial Ca²⁺ flux is selectively toxic for cancer cell lines, which may suggest that its inhibition could be tolerated by patients while maintaining anticancer efficacy. However, only a few, non-selective agents that directly modulate MCU activity have been identified.^45,46 In the absence of reliable means to pharmacologically inhibit MCU, we have demonstrated through proof-of-principle genetic deletion experiments that MCU drives PDAC disease aggressiveness. Whole-body knockout of MCU in outbred mice has few effects.^24,28,47 These observations highlight the potential of targeting MCU as a therapeutic vulnerability in cancer.

In contrast to the strong phenotypes observed in our in vitro and in vivo models, the lack of an overt phenotype in our knockout KPCY-Mcu^Cre-KO animals may be due to compensatory mechanisms in response to the chronic loss of MCU. Whereas whole-body Mcu knockout (Mcu^−/−) is lethal in the C57BL/6 genetic background, mice are viable with no overt phenotypes when Mcu is knocked out in a mixed genetic background.⁴⁷ Tamoxifen-inducible Cre-mediated Mcu knockout mice have overt phenotypes in the heart,^47,48 whereas germline Mcu deletion mice do not.^49,50 Thus, there exist unknown compensatory mechanisms in response to the chronic loss of MCU that enable mice to survive with no obvious phenotypes.⁵¹ Similarly, we suggest that compensatory mechanisms contribute to tumor development, cellular survival, and proliferation in the chronic, germ-line absence of MCU. The compensatory mechanisms are currently under investigation. In future experiments, an acute loss of MCU by inducible Cre-mediated deletion in dual recombinase Flp-O-based “KPF” animals⁵² may show more obvious phenotypes.

We found that lack of MCU expression in PDAC restricts EMT. However, when exogenous EMT-inducing pressures were applied through stable expression of Snail or application of exogenous TGF-β, Mcu^KO cells remained competent to undergo EMT. Such plasticity suggests that extrinsic induction of EMT, for example by secretion of factors such as TGF-β by cancer-associated fibroblasts, may reduce the effectiveness of MCU inhibition as an anticancer target. Such findings may at least also partially explain the lack of a phenotype in our genetic model of MCU deletion in the KPCY background, in contrast to the striking phenotypes in the xenograft models. While these observations may suggest that targeting MCU as a mono-therapeutic approach might not be fruitful, it is possible that a combinatorial approach, for example with inhibitors of TGF-β signaling^53–55 could be efficacious. Such approaches emphasize the need to develop specific pharmacology for MCU-mediated mitochondrial Ca²⁺ uptake that is currently lacking.

While the reduction of malignant phenotypes and basal tendency toward EMT in the context of MCU deletion is at least partially due to inhibition of cell-autologous secretion of protumorigenic factors such as TGF-β, other processes likely contribute as well. These may include alterations in metabolism, activation states of other signaling pathways, or proclivity for senescence. Notably, EMT phenotypes have been linked to metabolic alterations, and mitochondrial Ca²⁺ is known to be an important regulator of ATP production and the synthesis of biochemical intermediates produced by flux through the TCA cycle.^6,14,17,18 Future studies to examine the role of MCU in regulating these pathways in PDAC are warranted, as they may also present as potential targets for co-targeting.

Limitations of the study

This study has a number of limitations. Firstly, this study primarily uses mouse models of pancreatic cancer and thus studies murine Mcu. While MCU is highly homologous between human and murine models, there remains a possibility that human and murine models respond differently to MCU deletion or expression. Future functional studies in human patient-derived PDAC organoids will be necessary to determine if our results translate to human systems. Similarly, this study uses conventional Cre-loxP alleles to delete Mcu in mouse models of pancreatic cancer simultaneously with the induction of mutant Kras and gain-of-function p53 to drive tumorigenesis. This limits the scope of our research because Mcu was knocked out before the tumor was formed. However, in human patients, we may expect to target MCU after detectable cancer, or a precancerous lesion, has formed. To clarify the effects of MCU in a more treatment-relative paradigm, future studies should attempt to delete Mcu in already-formed tumors using dual-recombinase mouse models with inducible Mcu deletion that would uncouple Mcu loss from tumorigenesis and allow for the study of the acute effects of Mcu loss on PDAC progression and metastasis. This could also be achieved through the use of inducible shRNA-mediated knockdown in already-formed orthotopic tumors. Similarly, our cell culture models are limited by their stable nature, which may induce some level of compensatory mechanisms within knockout cells that may fundamentally alter their reliance on mitochondrial Ca²⁺ handling through MCU. While our results demonstrate that the loss of Mcu alters the induction of EMT in tumor cells, compensation has yet to be explored and may reveal a more complete understanding of the effect of mitochondrial Ca²⁺ on tumor cell states. Finally, our work primarily explores the role of mitochondrial Ca²⁺ within tumor cells, yet tumor cells are vastly outnumbered by host cells in PDAC, thus giving an incomplete picture of the role of MCU in pancreatic cancer. As such, future work should determine the role of stromal MCU in regulating tumor initiation, growth, and progression.

RESOURCE AVAILABILITY

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, J. Kevin Foskett (foskett@pennmedicine.upenn.edu).

Materials availability

All unique/stable reagents generated in this study, such as mice and cell lines, are available from the lead contact with a completed materials transfer agreement.

STAR★METHODS

Detailed methods are provided in the online version of this paper and include the following:

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Human cell lines were sourced from ATCC and Kerafast. Panc-1 was initially isolated from a pancreatic neoplasm of a 56-year-old male, and MiaPaCa2 are from a pancreatic neoplasm of 65-year-old male. The control cell line, HPDE, was developed from normal-type pancreatic ductal epithelial cells of a 63-year-old female.

For the KPCY GEMM model, 3 male and 8 female KPCY-Mcu^WT mice and 10 male and 7 female KPCY-Mcu^Cre-KO mice were generated. Data were pooled for male and female animals, as there was no difference in trend or significance between male and female mice (data not shown).

All murine parental cell lines were developed from experimental mice as previously described,³⁸ and xenograft hosts were sex-matched to their xenograft hosts. The KPCY-Mcu^Cre-KO cell line was generated from a male mouse. Paired isogenic cell lines were implanted into n = 3–5 male mice. For tail vein assays, n = 5 was used. KPCY-Mcu^WT cells (previously published as 2838. c3) were sourced from a female mouse and implanted into female mice. At day 13, 3 mice per group were harvested, and 4 mice per group were harvested on day 27. Finally, for the KPCY-Mcu^CRISPR-KO ± Snail expression experiments in Figure 6, n = 3 mice were used.

For imaging studies, such as in Figure 1F, three mice were used per condition, with a mixture of males and females. No phenotypic differences were noted between sexes.

METHOD DETAILS

Reagents and cell lines

Panc-1 (CRL-1469, from pancreatic neoplasm of a 56-year-old male) and MiaPaCa2 (CRL-1420, from pancreatic neoplasm of 65-year-old male) cells were obtained from ATCC. All murine parental cell lines were developed from experimental mice as previously described.³⁸ HPDE (Kerafast H6c7, from 63-year-old female) were grown in Keratinocyte SFM + EGF + bovine pituitary extract (Invitrogen 17005042) supplemented with 1x antibiotic-antimycotic (A/A, Gibco 15240–062). All other cell lines were maintained in DMEM (Corning 10–013CM) + 10% fetal bovine serum (FBS, Hyclone SH30071.03) + 1x A/A. All cell lines were maintained in a humidified incubator at 37°C and 5% CO₂ with media changes or passaging every 2–3 days and used within 20 passages.

For all stable clones, single-cell clones were developed by transfection with Lipofectamine 3000 (Thermo Fisher Scientific, L3000001) for 48 h, then selection of a polyclonal cell line with a given antibiotic and subsequent isolation of single-cell clones via limiting dilution. When possible, control lines expressing empty vectors were used as negative controls. Clones were verified for MCU expression by western blot (WB) and mitochondrial Ca²⁺ uptake assay. KPCY-Mcu^rescue lines were generated by stable expression of pCMV-Mcu-V5-His-puro, selected with 8 μg/mL puromycin then maintained under 2 μg/mL puromycin. For KPCY-Mcu^CRISPR-KO, pLenti-CRISPR-V2-sgMCU-mCherry (a generous gift from Mohamed Trebak) transfected cells were selected by limiting dilution. For Snail expression, KPCY-Mcu^CRISPR-KO and isogenic KPCY-Mcu^WT cells were transfected with pCDH-Snailpuro and selected with 8 μg/mL puromycin, then maintained at 2 μg/mL thereafter. When indicated, cells were treated with 10 ng/mL TGFβ (Millipore Sigma SRP3171) in culture; this was replenished every 2 days.

Cell proliferation assays

20,000 cells/well were plated in 24-well tissue culture treated plates in 2 mL of media unless otherwise noted. At given time points, media was aspirated, cells were rinsed with 1 mL 1x DPBS, and wells were trypsinized with 250 μL 0.25% trypsin for ~3 min until detachment. Trypsinized cells were mixed with 250 μL complete, FBS-containing media and counted manually by hemocytometer, with the average of two technical replicates taken as the value. Three separate wells were counted at each time point per condition, and three independent cell count experiments were carried out.

Pancreatic cancer patient samples

Human PDAC or normal pancreatic tissues obtained from the UMass Center of Clinical and Translational Sciences Biorepository and derived retrospectively from patients undergoing surgery at UMass Memorial Hospital were consented under the IRB approved protocol no. H-4721. De-identified FFPE tumor specimens were cut into 5-μm sections and IHC staining was performed as described above. MCU primary antibody was stained at 1:200 (Sigma #HPA016480).

Tissue staining and imaging

Tissues were isolated from mice, placed in cassettes in zinc formalin fixative, and stored at 4°C overnight. Then, tissue cassettes were transferred to 70% ethanol in distilled water at 4°C until further processing. All tissues were paraffin embedded, sectioned, and stained with hematoxylin and eosin by the Molecular Pathology and Imaging Core (MPIC) at the University of Pennsylvania (Center for Molecular Studies in Digestive and Liver Diseases - P30DK050306, RRID: SCR_022420).

For immunofluorescent staining, tissue sections were deparaffinized in Xylene, rehydrated, and antigen retrieval was performed with R-Buffer A (Electron Microscopy Sciences 62706–10). Slides were blocked and permeabilized for 1 h at room temperature with 5% donkey serum in 0.3% PBS-Triton X and then slides were left in primary antibody in 5% donkey serum in 0.3% PBS-Triton X overnight at 4°C. Primary antibodies used were MCU (HPA016480, Sigma-Aldrich), ECADHERIN (Clone M108, Takara), Ki-67 (ab16667, Abcam), and green fluorescent protein (ab6673, Abcam), which recognizes YFP. Slides were mounted in Fluoromount-G Mounting Medium with DAPI (Invitrogen 00–4959-52). Imaging was completed on a Leica Thunder Tissue Imager and analyzed on QuPath.⁵⁶

Mitochondrial Ca²⁺ uptake assays

Mitochondrial Ca²⁺ uptake was observed via high-speed spectrofluorimetry, as previously described.²⁵ Trypsinized cultured cells were resuspended in complete media to 8×10⁶ cells in 10 mL and incubated for 20 min at 37°C and 5% CO₂, and then centrifuged for 3 min at 1000 rpm. The media was then aspirated, and the pellet resuspended in Ca²⁺-free extracellular-like buffer (ECM: 20 mM HEPES-NaOH, 120 mM NaCl, 5 mM KCl, 1 mM KH₂PO₄, 0.2 mM MgCl₂, 0.1 mM EGTA, pH 7.4) made using dH₂O treated with BT Chelex 100 resin (Bio-Rad, 143–2832), spun down at 1000 rpm for 3 min, then resuspended in 1.5 mL Ca²⁺-free intracellular-like buffer (ICM: 20 mM HEPES-NaOH, 10 mM NaCl, 120 mM KCl, 1 mM KH₂PO₄, 5 mM succinate, pH 7.5) made using dH₂O treated with BT Chelex 100 resin and transferred to cuvettes. Fluorescence was monitored in a fluorimeter with multiwavelength excitation and emission (Delta RAM, PTI) at a constant temperature of 37°C. Fura2-FF (AAT Bioquest, 21028, Kd = 5.5 μM) excited at 340 nm and 380 nm was monitored at 535 nm emission. TMRE (Molecular Probes, T669) was excited at 560 nm and emission monitored at 595 nm. Reagents were added as follows: T = 0, ICM-cell suspension; T = 25 s, 1 μM Fura-FF and 10 nM TMRE; T = 50 s, 0.004% digitonin; T = 100 s, 2 μM thapsigargin (Sigma, T9033); T = 300 s, 10 μM CGP37157 Tocris, 1114); T = 500 s, 2.5–10 μM CaCl₂; T = 700 s, 2 μM CCCP; T = 900 s, 1 mM EGTA; and T = 950 s, 1 mM CaCl₂ (Sigma-Aldrich, 21115). Bolus concentrations of Ca²⁺ were selected as the highest concentration at which provided cells were able to take up all of the exogenously supplied Ca²⁺ without loss of mitochondrial membrane potential, as measured by TMRE fluorescence. To determine extramitochondrial Ca²⁺ concentration (${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\mathrm{bath}}$) based on the ratiometric calibration of Fura-FF, we used the following equation:

$${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\mathit{bath}}=\left(\frac{\mathrm{R}-R_{min}}{R_{max}-\mathrm{R}}\right)\mathrm{*}\left(\frac{S{f}_2}{S{b}_2}\right)\mathrm{*}Kd$$

where $R$ is the ratio of Fura-FF fluorescence at 340/380 nm excitation; $R_{\mathit{min}}$ is the $R$ measured with [Ca²⁺] = 0; ${\mathrm{R}}_{\mathit{max}}$ is $R$ measured at saturating [Ca²⁺]; Sf2 is fluorescence of Fura2-FF excited at 380 nm and [Ca²⁺] = 0. Sb2 is fluorescence of Fura2-FF excited at 380 nm at saturating [Ca²⁺]; and $Kd$ is the dissociation constant of Fura2-FF (5.5 μM). To quantify rates of Ca²⁺ uptake, a single exponential was fitted from the initial peak after Ca²⁺ addition (T = 500 s) until a steady state was reached (T = 650 s). TMRE fluorescence was used to verify that mitochondria had not undergone permeability pore transition during the course of uptake experiments.

Invasion and migration assays

Cultured cells were plated at a density of 20,000 cells/well in the top of transwell invasion (Millipore Sigma ECM550, following manufacturer’s instructions) or migration (Corning, 3464) plates in serum-free DMEM. Complete DMEM +10% FBS was used as an attractant in the bottom of the plate. After 24 h, the tops of the wells were cleaned and cells on the bottom of the membrane were fixed in 4% paraformaldehyde (Electron Microscopy Sciences, 15713) and stained with 10 μg/mL DAPI in 1% Triton X-100 (Sigma T9284). Three different fields of view were imaged at 10x and analyzed by ImageJ. For transient transfection models, cells were plated at 250,000 cells/well in 6-well plates and transfected with 1 μg of plasmid per well for 48 h before plating for transwell assays as previously described above. For TGF-β treatment experiments, 100,000 cells/well in 6-well plates were treated with 10 ng/mL TGF-β in complete growth media for 48 h before plating as above.

Wound healing assays

Cultured cells were plated in complete media at 500,000 cells/well of 12-well tissue culture treated plates and incubated overnight. The next day, monolayers were scratched by hand with a 200 μL pipet tip to create a wound. Media was changed, and plates were imaged at 10x. For transient transfection models, cells were plated at 250,000 cells/well and transfected with Lipofectamine 3000 and 0.5 μg/well of plasmid for 48h before scratching. For TGF-β treated cells, cells were plated at 250,000 cells/well and treated for 48 h before scratching. Plates were incubated for 18–24 h as noted then imaged again in the same conditions, except in the case of transiently transfected cells. Relative migration area was calculated by:$\frac{Area\left(0h\right)-Area\left(24h\right)}{Area\left(0h\right)}*100\%$. Experiments were repeated 2–3 times, with biological replicates in triplicate per experiment.

Clonogenic assays

Cells were plated at 10,000 cells/well in 6-well tissue culture treated plates and incubated for 7 d. Cells were then rinsed with 1x DPBS, incubated at room temperature in crystal violet fix/stain solution (1% methanol, 1% paraformaldehyde, 0.5% crystal violet in 1x DPBS) for 1 h, then gently rinsed with water until the wells run clear. After drying overnight, the plates were imaged with a GeneSys GBox and quantified for clonogenic area with ImageJ using the ColonyArea plug-in.

Spheroid formation assays

Tumorsphere formation assays were carried out as previously described with a few changes.⁵⁷ Briefly, 200 cells were plated in 200 μL spheroid media [DMEM/F12 (1:1) + 20 ng/mL epidermal growth factor (Sigma Aldrich, St. Louis, MO), 5 μg/mL insulin (Sigma Aldrich), 10 ng/mL basic fibroblast growth factor (Sigma Aldrich) and 0.4% bovine serum albumin (Research Products International, Mount Prospect, IL). For each experiment, media was freshly mixed with growth factors and 1x B-27 (Gibco) was added to spheroid media immediately before use. in low-adhesion plates coated with Aggrewell Antiadherence solution (Stem Cell Tech #07010) by adding 200 μL of solution per well, then centrifuging at 1300x g for 5 min before aspiration and rinsing with 1xDPBS. Outer wells were filled with 1xDPBS to reduce evaporation. Spheroids were counted manually on a microscope after 7 d of incubation.

Western blotting

Media was aspirated from culture plates and cells were rinsed with 1x DPBS before thorough aspiration. Cells were then harvested in RIPA buffer +200 μM PMSF + 1x cOmplete Mini protease inhibitor cocktail (EDTA free, Roche 11836170001) via scraping on ice, transferred to labeled tubes, and rotated for 1.5–2 h at 4°C. Samples were then centrifuged at 13000+ rpm and 4°C for 10 min and transferred to new tubes and quantified via Pierce BCA assay according to manufacturer’s instructions, with samples diluted 1:5. Samples were mixed to load 10–20 μg protein/well with 4x Laemmli buffer (Biorad 1610747) according to manufacturer specifications with the exception of chemotherapeutic-treated cells, which were loaded at 4 μg/well due to low cell density in test conditions. Precision Plus Dual Color Standard (Biorad 161–0374) served as protein Standard. Samples were run on NuPAGE 4–12% bis/tris mini protein gels (NP0321–0323) at 100 V in MES running buffer (NP0002) for ~1 h then transferred for 1 h at 100 V onto Immobilon-P PVDF membrane on ice. After blocking for 1.5 h at room temperature on a rocker in 5% w/v dry milk in TBST (1x tris-buffered saline +0.1% Tween 20), blots were incubated overnight on a shaker at 4°C. After rinsing, blots were incubated in secondary antibody (Cell Signaling Technologies: anti-mouse-HRP, 7076, and/or anti-rabbit-HRP, 7074). Blots were then rinsed with TBST and visualized with a GeneSys GBox and SuperSignal West PICO Plus ECL reagent (ThermoFisher Scientific, 34577). Results were quantified as appropriate with ImageStudio Lite.

Murine studies

Orthotopic implantation of tumor cells was performed as previously described.^s58,59 Briefly, mice were anesthetized with isoflurane and a sterile field around the abdomen was prepared. An incision was made in the upper left quadrant of the abdomen and the body of the pancreas was exposed. Then 1.0×10⁵ cells in 100 μL sterile DMEM were injected into the tail of the pancreas with an insulin syringe. KPCY-derived tumor cells from C57BL/6J mice were injected into 8–10-week old female C57BL/6J mice (000664, The Jackson Laboratory). The formation of a liquid bleb at the injection site verified a successful injection. After injection, a sterile pad was held to the injection site to prevent tumor cells from leaking into the abdominal cavity. Finally, the pancreas was placed back into the abdomen and then the peritoneum and skin were sutured closed with 4–0 coated sutures.

For tail vein injections, an insulin syringe was loaded with 1.0×10⁵ KPCY cells in 100 μL sterile DMEM and injected into the tail vein of 8–10-week old female C57BL/6 animals (000664, The Jackson Laboratory). Lungs were harvested at the indicated time points, imaged for YFP and bright field, and then formalin fixed for downstream analysis. All vertebrate animals were maintained, and experiments were conducted, in compliance with all regulations and NIH guidelines for animal research and were approved by the University of Pennsylvania and University of Massachusetts Chan Medical School Institutional Animal Care and Use Committees.

Flow cytometry

Surface ECAD levels were measured by a previously described method.¹⁴ Briefly, cells were removed from culture plates and dissociated into single cells using Hank’s Enzyme Free Cell Dissociation Solution (S-004-C, EMD Millipore). Cells were stained using anti-ECAD (147308, BioLegend) or isotype control (400418, BioLegend) in FACS buffer for 15 min on ice in the dark. Cells were washed in FACS buffer and then stained with DAPI and filtered through a 70-μm strainer to create a single cell suspension. Flow cytometry was run on an LSR II at the University of Pennsylvania Flow Cytometry Core. These experiments measure surface ECAD levels only, as cells were not permeabilized.

Secretion quantification

For the quantification of secreted factors in the media, 100,000 cells per replicate of Mcu^CRISPR-KO and Mcu^WT cells were plated in 6-well tissue culture treated plates in 1.5 mL complete growth media in triplicate. Cells were incubated at 37°C and 5% CO₂ for 48 h before harvest. Cell culture media/supernatant was harvested and stored in sterile Eppendorf tubes at −80°C before shipment. Undiluted media was assayed by Eve Technologies for secretion of chemokines and cytokines represented in their Mouse Discovery 32-plex assay (MD32). Data is shown as a heatmap of the Z score of each secreted factor which was quantified in all lines tested.

TGF-β ELISA

To quantify TGF-β secretion into the media, 50,000 cells/well of Mcu^CRISPR-KO and Mcu^WT cells were plated in 48-well plates for in 1 mL complete media. For intracellular TGF-β quantification, 100,000 cells/well were plated in 6-well plates in 1.5 mL complete media for 48 h, then trypsinized with 500 μL 0.25% trypsin, mixed with 500 μL complete media, and then counted manually with a hemocytometer. Cells were then spun down at 1500 rpm for 3 min, rinsed with 1xDPBS, and then spun down again and harvested in 150 μL RIPA buffer +200 μM PMSF + 1x cOmplete Mini protease inhibitor cocktail (EDTA free, Roche 11836170001). In either case, samples were quantified with the Abcam mouse TGF-β 1 ELISA kit (ab119557) according to manufacturer’s instructions. Briefly, samples were diluted 1:10 in assay buffer, treated with 1:11 dilution of 1N HCl for 1 h, then neutralized with 1:12 1N NaOH. Values were compared to a standard curve of TGF-β from 31.25 to 2000 pg/mL. Samples were loaded into the rehydrated and washed ELISA plate and incubated for 2 h on a shaker at room temperature before rinsing and incubating with 100 μL biotinylated antibody per well for 1 h on a shaker at room temperature. After rinsing, 100 μL streptavidin-HRP was added and incubated for 30 min on a shaker at room temperature. Then, wells were rinsed and 100 μL TMB substrate was added to each well and incubated for 30 min in the dark. 100 μL stop solution was added, and the plate was immediately read on a plate reader at A450-A660. Data was corrected for dilution factor and normalized to cell number.

RNA sequencing

RNA was isolated with a RNeasy Mini kit (Qiagen 74104) from 10 to 15-cm dishes and sequenced by Novogene with a NovoSeq PE150 at ~20 M paired-end reads. Raw reads were processed with Salmon and DESeq2 before analysis with GSEA and GO. At least 3 biological replicates were used for each experimental condition, and principal components were used to verify the reproducibility of replicates.

Single Cell RNA sequencing data analysis

Single cell RNA sequencing data from tumor tissue of human pancreatic cancer patients was obtained from GSE229413 and converted to a Loupe browser object. Within Loupe Browser, cell populations consistent with PDAC, PanIN, Ductal type 1, and acinar-like populations were identified. Heat maps and quantifications of gene expression within these populations were generated with this software. Data are presented as heat maps as a split UMAP plot and as violin plots, both of log2 normalized values.

QUANTIFICATION AND STATISTICAL ANALYSIS

Unless otherwise noted, all experiments were carried out as 3 separate, independent experiments with at least 3 biological replicates per experiment. Data were analyzed with GraphPad Prism (versions 8–10) or R, unless otherwise noted. For all normally-distributed two-group data, Student’s T test was used. For multigroup, one-independent variable data, one-way ANOVA with Sidak’s posthoc was used. When two independent variables were present (i.e., in ±MCU, ±Snail experiments and time courses), two-way ANOVA with Sidak’s posthoc was employed. All data were assessed for normality with Kolmogorov-Smirnov tests and for outliers with ROUT with Q = 5% before analysis.

Supplementary Material

SUPPLEMENTAL INFORMATION

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.115627.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
beta Tubulin Monoclonal Antibody	Invitrogen	32-2600; RRID: AB_2533072
MCU (D2Z3B) Rabbit mAb	Cell Signaling	14997; RRID: AB_2721812
Vimentin (D21H3) XP® Rabbit mAb	Cell Signaling	5741; RRID: AB_10695459
N-Cadherin (D4R1H) XP® Rabbit mAb	Cell Signaling	13116; RRID: AB_2687616
E-Cadherin (24E10) Rabbit mAb	Cell Signaling	3195; RRID: AB_2291471
Snail (C15D3) Rabbit mAb	Cell Signaling	3879; RRID: AB_2255011
Anti-rabbit IgG, HRP-linked Antibody	Cell Signaling	7074; RRID: AB_2099233
Anti-mouse IgG, HRP-linked Antibody	Cell Signaling	7076; RRID: AB_330924
HIF-1α (D1S7W) XP® Rabbit mAb	Cell Signaling	36169; RRID: AB_2799095
Nrf2 (D1Z9C) XP® Rabbit mAb	Cell Signaling	12721: RRID: AB_2715528
Anti-MCU antibody produced in rabbit	Sigma Aldrich	HPA016480; RRID: AB_2071893
Monoclonal Anti-mouse E-cadherin (Clone ECCD-2)	Takara	M108; RRID: AB_2895157
Rabbit anti-Ki67 antibody SP6	Abcam	ab16667; RRID: AB_302459
Caspase-3 (D3R6Y) Rabbit mAb	Cell Signaling	14220: RRID: AB_2798429
Cleaved Caspase 3 (Asp175) (5A1E) Rabbit mAb	Cell Signaling	9664: RRID: AB_2070042
Anti-Green Fluorescent Protein Antibody (Goat Polyclonal)	Abcam	ab6673; RRID: AB_305643
Tgf beta-1,2,3 Monoclonal Antibody (1D11), Unconjugated, Species Reactivity: Human, Host: Mouse/IgG1	Thermo Fisher Scientific	MA5-23795; RRID: AB_2609812
Chemicals, peptides, and recombinant proteins
Triton X-100	Sigma	T9284
TGF beta	Millipore Sigma	SRP3171
Basic Fibroblast Growth Factor	Sigma Aldrich	F0291
B-27 supplement	Life Technologies	17504-044
Epidermal Growth Factor	Sigma Aldrich	E5036
Insulin	Life Technologies	A11429IJ
Bovine Serum Albumin	Sigma Aldrich	A9576
Aggrewell Antiadherence solution	Stem Cell Technologies	07010
cOmplete Mini protease inhibitor cocktail, EDTA free	Roche	11836170001
NuPAGE 4-12% bis/tris mini protein gels	Invitrogen	NP0321-0323
Immobilon P PVDF membrane	Millipore	IPVH00010
Precision Plus Dual Color Standard	Biorad	161-0374
NuPAGE™ MES SDS Running Buffer (20X)	Invitrogen	NP0002
Puromycin dihydrochloride, 10 mg/mL	Life Technologies	A11138-03
Geneticin	Invitrogen	10131035
Lipofectamine 3000	Thermo Fisher Scientific	L3000001
Fluoromount-G™ Mounting Medium with DAPI	Invitrogen	00-4959-52
Critical commercial assays
Mouse Cytokine/Chemokine 32-Plex Discovery Assay® Array	Eve Technologies	MD32
Mouse TGF beta 1 ELISA kit	Abcam	ab119557
RNeasy Mini kit	Qiagen	74104
ECMatrix Cell Invasion Assay 8 uM	Millipore Sigma	ECM550
Corning® 6.5 mm Transwell® with 8.0 μm Pore Polyester Membrane Insert, Sterile	Corning	3464
Deposited data
Mitochondrial Ca2+ controls pancreatic cancer growth and metastasis by regulating epithelial cell plasticity	Weissenrieder JS, Foskett JK	GEO: GSE287625
Expression Analysis of Normal and Neoplastic Mouse Pancreatic Ductal Organoids	Tuveson DA, Baker LA	GEO: GSE63348
Analysis of donor pancreata defines the transcriptomic signature and microenvironment of early neoplastic lesions (reanalyzed from dbGap phs002071	Pasca di Magliano M, Carpenter ES, Elhossiny AM	GEO: GSE229413
Experimental models: Cell lines
Panc-1	ATCC	CRL-1469
HPDE	Kerafast	H6c7
MiaPaCa-2	ATCC	CRL-1420
2838.c3 murine KPCY line	Stanger lab	Stanger lab
1151 murine KPCY-McuCre-KO	This paper	This paper
2838.c3 murine KPCY-McuWT pLentiCRISPR-EV clones 15&22	This paper	This paper
2838.c3 murine KPCY-McuCRISPR-KO clones 17 and 42	This paper	This paper
1151 murine KPCY-McuCre-KO empty vector clones 1 & 3	This paper	This paper
1151 murine KPCY-Mcurescue clones 2,3,4,5,6, & 9	This paper	This paper
2838.c3 KPCY-McuWT EV cl 15 + pCDH-EV	This paper	This paper
2838.c3 KPCY-McuWT EV cl 15 + pCDH-Snail	This paper	This paper
Experimental models: Organisms/strains
C57BL/6J Mice	Jackson Laboratories	000664
Pdx1-Cre;KrasG12D/+;Tp53R172H/+;R26Lsl-Yfp/Lsl-Yfp;Mcufl/fl mice in C57BL/6J background	This paper	This paper
Recombinant DNA
pCMV6-A-MCU-V5-His-puro	Generated in lab	DOI: pnas.2005976117
pCMV6-A-EV-puro	Origene	PS100025
pCDH-Snail-blasti		N/A
pCDH-EV-blasti
pLentiCRISPR V2-sgMCU	Mohamed Trebak lab	N/A
pLentiCRISPR V2-EV	Addgene	52961
Software and algorithms
Prism	GraphPad	N/A
R	N/A	N/A
VisiView	Visitron	N/A
FIJI	ImageJ	N/A
QuPath		PMID: 29203879
Loupe Browser	10x Genomics	N/A

Highlights.

• Mcu knockout reduces pancreatic cancer growth and metastasis

• MCU knockout reduces EMT in vitro and in vivo

• EMT induction by Snail expression or exogenous TGF-β rescues growth and metastatic defects

Data and code availability

RNA sequencing data have been deposited at Gene Expression Omnibus (GEO) as GSE287625. and are publicly available as of the date of publication. This paper analyzes existing, publicly available data, accessible at GSE63348, TCGA-PAAD, and GSE229413. This paper does not report original code.