Abstract
Pancreatic ductal adenocarcinoma (PDAC) is a highly metastatic disease with few effective treatments. Here we show that the mitochondrial calcium uniporter (MCU) promotes PDAC cell migration, invasion, metastasis, and metabolic stress resistance by activating the Keap1-Nrf2 antioxidant program. The cystine transporter SLC7A11 was identified as a druggable target downstream of the MCU-Nrf2 axis. Paradoxically, despite the increased ability to uptake cystine, MCU-overexpressing PDAC demonstrated characteristics typical of cystine-deprived cells and were hypersensitive to cystine deprivation-induced ferroptosis. Pharmacologic inhibitors of SLC7A11 effectively induced tumor regression and abrogated MCU-driven metastasis in PDAC. In patient-derived organoid models in vitro and patient-derived xenograft models in vivo, MCU-high PDAC demonstrated increased sensitivity to SLC7A11 inhibition compared with MCU-low tumors. These data suggest that MCU is able to promote resistance to metabolic stress and to drive PDAC metastasis in a cystine-dependent manner. MCU-mediated cystine addiction could be exploited as a therapeutic vulnerability to inhibit PDAC tumor growth and to prevent metastasis.
Elevated mitochondrial calcium uptake in PDAC promotes metastasis but exposes cystine addiction and ferroptosis sensitivity that could be targeted to improve pancreatic cancer treatment.
Introduction
Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal tumor malignancies. Even among patients with resectable early-stage PDAC, 75% will develop liver metastasis within 1–2 years after complete removal of primary tumor (1). It is urgent to understand molecular mechanisms underlying PDAC metastasis and to develop novel therapeutic strategies for metastatic PDAC.
Mitochondria are crucial in pancreatic cancer tumorigenesis and progression (2–4). In addition to their roles in cell metabolism, mitochondria are essential for Ca2+ signaling and redox balance (5). The mitochondrial calcium uniporter (MCU) is the pore-forming subunit of the MCU complex, which consists of MCU, the scaffold protein EMRE, and the calcium-sensitive inhibitory regulatory subunits MICU1 and MICU2 (6, 7). MCUb is a paralog of MCU that serves as a dominant negative regulator of MCU (8). The activities of MCU can be regulated by change in the ratios between MCU and MCUb and the binding of MICU1/2 to Ca2+. In addition to MCU, the mitochondrial Ca2+ levels are regulated by NCLX, a Na+-Ca2+ exchanger responsible for mCa2+ efflux (9). Ca2+ uptake by MCU is important for buffering cytosolic Ca2+ increase and regulating mitochondrial oxidative phosphorylation (OXPHOS; refs. 6, 7, 10, 11). On the other hand, the MCU-mediated mitochondrial Ca2+ overload might lead to cell death due to opening of the mitochondrial permeability transition pores and excessive production of reactive oxygen species (ROS; refs. 7, 11–13). Paradoxically, there are reports of MCU upregulation or NCLX downregulation in metastatic cancer, implicating a prometastasis role for mitochondrial Ca2+ signaling during tumor progression (14–17).
Cancer cells are characterized by simultaneous increase in oxidative stress and antioxidant response (18). The Keap1-Nrf2 pathway is the predominant antioxidant response and xenobiotic detoxification mechanism in mammalian cells (19–21). Keap1 is a cytosolic oxidative stress sensor that sequesters Nrf2 for ubiquitination and proteasomal degradation (19). Under oxidative stress, the oxidation of multiple Cys residues of Keap1 blocks its binding to Nrf2 (20), which allows the stabilization of Nrf2 protein and the transcription of antioxidant response element (ARE)-harboring genes (22–24). This pathway is upregulated in PDAC (25, 26) due to oncogenic KRAS-promoted transcription of Nrf2 (26). The activation of Keap1-Nrf2 promotes cell migration, invasion, and oxidative stress resistance in metastatic cancer cells (21, 27–29).
Ferroptosis is a type of iron and lipid peroxidation–dependent cell death (30–34). Cancer cells rely on cystine uptake for glutathione (GSH) synthesis, which is required for GSH peroxidase 4 (GPX4) to reduce lipid peroxidation (33, 34). Upon cystine deprivation or GPX4 inhibition, iron rapidly catalyzes the accumulation of lipid peroxide through Fenton reaction, which eventually compromises membrane integrity and leads to cell death (30–32, 34). Mitochondrial membrane hyperpolarization driven by OXPHOS has recently been implicated in promoting ferroptosis induced by cystine deprivation (35). However, the role of mitochondrial Ca2+ signaling in ferroptosis remained unclear.
In this study, we investigated the role of MCU in pancreatic cancer metastasis and progression. Our data supported that MCU promotes PDAC progression through activating the Keap1-Nrf2 circuit and the cystine addiction in MCU-overexpressing PDAC is a therapeutic vulnerability that could be exploited to prevent metastatic recurrence.
Materials and Methods
Patients and human tissue specimens
Tumor samples that included lymph node samples were obtained from 132 patients with PDAC treated at the Tianjin Cancer Hospital from 2014 and 2018. All 132 patients underwent radical pancreaticoduodenectomy with R0 margins confirmed by two pathologists. None of the patients had received chemotherapy or radiotherapy at the time of sample collection. All patients were treated with systemic gemcitabine-based chemotherapy after operation for six cycles after surgery. No radiotherapy was given before or after surgery. Postoperative follow-up of patients were conducted every 3 months initially. Overall survival (OS) was defined as the time interval from the date of surgery to that of death due to any cause or that of the last follow-up. Relapse-free survival (RFS) was calculated from the date of surgery to that of local recurrence or metastasis. Local recurrence or metastasis was diagnosed by radiological examination (contrast-enhanced CT or MRI scanning). All patients provided written informed consent. The study protocols were approved by the Ethics Committee of the Tianjin Cancer Institute and Hospital.
IHC analysis
Consecutive sections of formalin-fixed, paraffin-embedded tumors were subjected to IHC analysis for MCU and NRF2 using a DAB (3,3′-diaminobenzidine) substrate kit (Maixin) according to the manufacturer's instructions. The slides were deparaffinized in xylene and rehydrated through graded ethanol to water before staining. All sections were treated with EDTA (pH 8.0) for antigen retrieval and with 3% H2O2 for the inactivation of endogenous peroxidase. Sections were incubated with anti-MCU (Sigma, PA5-109304, at 1:500 dilution) antibodies or anti-NRF2 (Abcam, ab-62352, at 1:350 dilution) overnight at 4°C. After wash, the sections were stained with a secondary antibody for 30 minutes at room temperature. PBS was substituted for each primary antibody as a negative control. Five random fields were examined under a light microscope. Immunoreactivity was semiquantitatively scored according to the estimated percentage of positive tumor cells as described previously. IHC-staining results were blindly and independently performed by two pathologists who were blinded to the clinical data. Staining intensity was scored 0 (negative), 1 (low), 2 (medium), and 3 (high). Staining extent was scored 0 (0% stained), 1 (1%–25% stained), 2 (26%–50% stained), and 3 (51%–100% stained). The final score was determined by multiplying the intensity scores with staining extent and ranged from 0 to 9. Final scores (intensity score ×percentage score) less than or equal 4 were considered as low staining, more than 4 were high staining. Antibodies used in IHC staining can be found in Supplementary Table S5.
ROS measurement
Mitochondrial ROS (mROS; superoxide anion) were detected using 5 μmol/L MitoSox Red staining (Invitrogen) for 30 minutes at 37°C according to the manufacturer's instructions and analyzed using a FACS Calibur flow cytometer (BD Biosciences). Forward and side scatter data were collected (10, 000 events per sample).
Animal experiments
All animal experiments were performed according to protocols approved by Institutional Animal Care and Use Committee at the Tianjin Cancer Institute and Hospital or the Penn State College of Medicine.
For the orthotopic tumor models, 1 × 106 luciferase-labeled PDAC cells (Panc-1, SW1990, or PDX677) were resuspended in 40 μL 0.25 mg/mL Matrigel (Corning) with PBS buffer, then orthotopically injected into the carefully exposed pancreas of 5-week-old female nude mice (36, 37). The pancreas was then returned to the peritoneal cavity, the abdominal wall and the skin was closed with skin clips. For drug treatments, mice were treated with either saline vehicle (Veh.), sulfasalazine (SAS, 100 mg/kg), or imidazole ketone erastin (IKE; 20 mg/kg) everyday via intraperitoneally starting from day 7. Noninvasive bioluminescent imaging (BLI) and analysis were performed as described previously using Xenogen IVIS 200 (38) starting from 1 week after surgery. Tumor growth was analyzed by BLI. Six to 7 weeks after surgery, mice were euthanized and the BLI signals in primary tumor (pancreas) or distant metastatic sites (liver, peritoneal cavity) were determined through ex vivo BLI. The visible metastatic lesions in the gut, mesentery, and liver of each mouse were counted. The primary and metastatic pancreatic tumor were excised, fixed in formalin, and embedded in paraffin. Half of the tissues were subjected to hematoxylin and eosin (H&E) and IHC staining as described for human PDAC tumors. Histologic staining on paraffin sections (H&E) were carried out using standard protocols. Following digital capture on Olympus BX51, histology images were processed using Affinity Photo software, using three filters/adjustments applied evenly across the entire image: unsharp mask, white balance, and levels adjustment.
For patient-derived xenograft (PDX) experiments, four PDX lines with MCU-high or MCU-low expression levels were used. The MCU expression levels were determined according to the IHC staining of the patient tumor tissues and PDX tissues. PDX tumors were cut into small pieces (about 8 mm3) and implanted into 5-week-old female NSG mice subcutaneously (10 mice for each PDX line). When the tumor volume reached 50–70 mm3, mice were randomized into vehicle control group or IKE group. A total of 100 μL IKE (20 mg/kg) or saline (vehicle control) were administrated daily via intraperitoneal. Tumor growth was monitored every 5 days using a caliper and tumor volumes were calculated by the following formula: Volume = 1/2 ×L1× (L2)2, where L1 is the long axis and L2 is the short axis. Mice were sacrificed after 8 weeks and tumors were harvested. Tumor tissues were immediately fixed in buffered formalin and embedded in paraffin. Tissue slides (5 μm) were prepared and H&E staining were performed for histopathologic analysis according to instructions. IHC of 4-HNE (abcam, ab48560) staining were performed to evaluate lipid peroxidation status of tumor tissues.
Biotin labeling of oxidized KEAP1
Cells were lysed for 15 minutes on ice in biotin labeling lysis buffer (BLLB: 50 mmol/L Tris-HCl pH 7.0, 5 mmol/L EDTA, 120 mmol/L NaCl, 0.5% Igepal-630) containing protease inhibitors and 100 mmol/L maleimide (Sigma-129585). Insoluble material was then removed by centrifugation at 20,000 × g for 10 minutes at 4°C, the cleared supernatant was transferred to a fresh Eppendorf tube and protein concentration was determined by the Bradford assay. Protein concentration was adjusted to 1 μg/μL with BLLB, SDS was added from a 10% stock to a final concentration of 1% and the cell lysates were incubated at room temperature for 2 hours rotating. To remove unreacted maleimide, proteins were subsequently precipitated by adding 5 volumes of acetone preequilibrated at −20°C and incubated for 20 minutes at −20°C. The preparations were centrifuged at 20,000 × g for 10 minutes at 4°C, supernatants removed and discarded and precipitated protein pellet was air dried. The pellet was then resuspended in 200 μL BLLB containing 1% SDS, 10 mmol/L DTT, and 0.1 mmol/L biotin-maleimide (Sigma-B1267, stock dissolved in dimethylformamide) to reduce the remaining, previously oxidized, sulfhydryl groups and allow their reaction with biotin-maleimide. Proteins were again precipitated with 5 volumes of methanol (−20°C) as above, the dried pellet was resuspended in 500 μL of BLLB, incubated with 10 μL of a 50% slurry of Pierce Streptavidin Agarose (Thermo Fisher Scientific, 20353) rotating at 4°C for 2 hours. The beads were then washed four times with BLLB and resuspended in SDS-PAGE loading buffer for SDS-PAGE analysis and Western blotting with the KEAP1 antibodies.
RNA sequencing
Three biological replicates each of control or MCE overexpress (OE) Panc-1 cells were cultured in glucose limited medium (1 mmol/L Glc) for overnight. The mRNA was extracted with RNeasy kit (Qiagen) and used for the preparation of RNA-sequencing (RNA-seq) library. Directional (stranded) libraries for the paired-end sequencing of PANC-1 cells were generated with an Illumina platform. After filtering low expression genes, raw read count data were used to calculate FDR and P value by applying edgeR method.
Ferroptosis treatment
Erastin (Selleck, S7242), Z-VAD-FMK (Selleck, S7023), necrostatin-1 (Nec-1; Selleck, S8037), the mitoTEMPO (mT; Millipore, C906P50), ferrostatin-1 (Fer-1; Cayman, 17729), CGP37157 (Sigma, 75450-34-9) and cyclosporine A (CsA; Selleck, S2286) were each dissolved in DMSO (Sigma). Deferoxamine (DFO; Sigma, 138-14-7) was dissolved in deionized water. For ferroptosis assay cells were seeded in 6-well plates and treated with cystine deprivation (cystine-free medium), glutamate (50 mmol/L), or erastin (4 μmol/L) in the presence or absence of Z-VAD-FMK (ZVAD, 50 μmol/L), Nec-1 (10 μmol/L), CsA (10 μmol/L), DFO (50 μmol/L) or Fer-1 (1 μmol/L). The effect of mT (2 μmol/L), NCLX inhibitor CGP37157 (2 μmol/L), DFO (50 μmol/L), Fer-1 (1 μmol/L) treatment on cystine deprivation or erastin-induced ferroptosis were determine in some experiments. In most experiments, PDAC cells were treated between 16–24 hours except for AsPC-1 cells, which were treated for 48 hours.
Measurement of lipid peroxidation
Lipid peroxide was analyzed by flow cytometry. Cell lines were plated in quadruplicate at the cell numbers indicated previously for the 12-well plate format. Cells were seed overnight and were subjected to various compound treatments for indicated times. Cells were then incubated for 30 minutes in live-cell imaging solution containing the pertinent ROS dye at the following concentrations: C-11 BODIPY (Invitrogen; 2 μmol/L). Cells were then washed with PBS, trypsinized with 0.25% trypsin, and neutralized with 10% FBS in PBS. A minimum of 10,000 cells were analyzed per condition. For C-11 BODIPY, signal was analyzed in the FITC channel. Software analysis and histogram generation was carried out using FlowJo v10. For Lipid peroxidation confocal imaging, cells were seeded at a density of 2.5 × 105 per well on coverslips placed in a 6-well dish and grown overnight in DMEM. Cells were treated with cystine-free medium for 16 hours, the coverslips were then mounted to a Teflon chamber. C11-BODIPY (Invitrogen; 5 μmol/L) were added to each well 15 minutes before measurements. Reduced and oxidized C11-BODIPY were detected at emission 590 and 510 nm, respectively.
Organoid culture and experiment
Human pancreatic tissues were obtained from patients undergoing surgery for pancreatic cancer who consented to donate their samples for research. The collection of samples was approved by the Ethics Committee of the Tianjin Cancer Institute and Hospital. The study is compliant with all relevant ethical regulations regarding research involving human participants. Human PDAC organoid cultures were established and cultured as described previously (39). To ensure purity of the sorted cells, the sorters were set to “single-cell” mode. CK19-positive single cells were sorted into cooled, Matrigel-covered wells of 96-well plates (Costar) in at least technical triplicate. After sorting, 10 μL/well undiluted Matrigel was allowed to solidify in an incubator before organoid media was added. For human organoids, human complete feeding medium (hCPLT) contained human wash medium with BSA (advanced DMEM/F-12, 10 mmol/L HEPES, 1 × GlutaMAX supplement, 100 μg/mL primocin, and 0.1% BSA), 10 mmol/L nicotinamide, 1 × Wnt3a-conditioned medium, 1 X R-spondin1-conditioned medium, 100 ng/mL mNoggin, 1 × B27 supplement, 100 μg/mL primocin, 100 ng/mL hFGF, 1.25 mmol/L N-acetylcysteine, 1 μmol/L PGE2, 50 ng/mL hEGF, 10 μmol/L Y27632, 10 nmol/L hGastrin and 500 nmol/L A 83–01. PDAC organoid (3 × 103 cells/well) were seeded in 96-well plates. After 7 days in culture, the medium was replaced with hCPLT, erastin (10 μmol/L), CGP37157 (5 μmol/L) or Erastin+CGP37157, followed by incubation for 3 days. Cell growth was analyzed by microscopy and by measuring ATP concentrations with CellTiter-Glo 3D Viability Assay kit (Promega Corporation).
Glutamine uptake and glutamate export
A total of 1.2 × 105 cells were plated onto 12-well plate and allowed to attach for 12 hours. The cells were fed with fresh medium. After 6-hour incubation, the conditioned medium were collected and the concentration of glutamine and glutamate in the conditioned medium and fresh medium were measured using YSI 2700 Biochemistry Analyzer (Marshal Scientific). The glutamine uptake and glutamate export by PDAC cells were calculated on the basis of the decreases in glutamine and increases in glutamate concentrations in conditioned medium relative to fresh medium. For H2O2 treatment, 100 μmol/L H2O2 was added.
FITC-cystine uptake assay
A total of 3 × 105 cells were plated onto 6-well plate for overnight. The cells were then incubated with fresh medium containing 5 μmol/L FITC-cystine (SCT047, Sigma). Incubate the cells in a 37°C CO2 incubator for 30 minutes. The cells were resuspended in ice-cold PBS containing 1% FBS after trypsinization and used for flow cytometry analysis.
Statistical analysis
Statistical analyses were performed with SPSS 18.0 software and GraphPad Prism version 5.0.3. Student t test or ANOVA for unpaired data were used to compare mean values. The Student t test for paired data was used to compare mean values for in vivo data. A Spearman rank correlation coefficient test was carried out for testing the association between ordinal variables. Kaplan–Meier survival curves were made according to these cut-off points; the log-rank test was used to obtain a P value for the significance of Kaplan–Meier curves’ divergence. All P values less than 0.05 were considered statistically significant. *, P < 0.05.
Data availability
The RNA-seq data generated in this study are publicly available in Gene Expression Omnibus at GSE199692.
Results
MCU overexpression in PDAC correlates with poor prognosis and metastatic progression
Calcium signaling is frequently dysregulated during tumorigenesis and cancer progression (40, 41). To understand the role of mitochondrial Ca2+ signaling in PDAC progression, we determined the expression levels of MCU complex subunits in paired PDAC tissue (T) and para-tumor normal pancreatic tissue (N) from 6 PDAC patients (Fig. 1A and B). MCU protein (Fig. 1A and B) levels were elevated in tumor tissues by approximately 4-fold when compared with paired normal tissues, while the levels of MCUb and the inhibitory subunit MICU2 were decreased. The levels of MICU1 were decreased by approximately 2-fold while the levels of EMRE did not change significantly (Fig. 1A and B). The upregulation of MCU in PDAC was confirmed in another cohort of 10 pairs of patients with PDAC (Supplementary Fig. S1A and S1B).
The immunofluorescence staining of MCU colocalized with NDUFB8 in PDAC tissues, which is consistent with the function of MCU as a mitochondrial inner membrane channel (Supplementary Fig. S1C). In IHC staining (Fig. 1C and D), MCU expression was very low in the normal pancreas but elevated in pancreatic intraepithelial neoplasia (PanIN) and carcinoma tissues. MCU levels were increased in carcinoma when compared PanIN in the same patient; PanIN-3 had stronger MCU staining than PanIN-1 (Fig. 1C). MCU overexpression significantly correlated with differentiation, lymph node metastasis and tumor–node–metastasis (TNM) stage (Fig. 1D; Supplementary Table S1). MCU overexpressing group had significantly poorer OS and RFS in our cohort of 132 patients with PDAC (Fig. 1E). The correlation between MCU expression and patient survival was also confirmed in in the The Cancer Genome Atlas (TCGA) RNA-seq database (Supplementary Fig. S1D). Cox proportional hazards analysis indicates that MCU expression levels strongly correlates with OS and RFS in univariate analysis and is an independent predictor of OS in multivariate analysis (Supplementary Table S2). Taken together, our data indicate that MCU expression levels increased during PDAC initiation and progression.
MCU promotes metabolic stress resistance and PDAC metastasis
To understand the role of MCU in PDAC progression and metastasis, we took either OE or knockout (KO) MCU in PDAC cell lines (Supplementary Fig. S2A). MCU KO abrogated mitochondrial Ca2+ uptake without affecting thapsigargin-induced endoplasmic reticulum Ca2+ release or store-operated calcium entry (Fig. 2A; Supplementary Fig. S2B), while MCU OE was sufficient to promote mitochondrial Ca2+ uptake (Fig. 2B). Mitochondrial calcium uptake activities strongly correlated with MCU protein levels in cell lines and PDX lines (Fig. 2C and D; Supplementary Fig. S2C). MCU KO in PDAC cells remarkably inhibited cell migration, invasion, and soft agar colony formation (Supplementary Fig. S2D–S2E). The decrease in migration and invasion in MCU KO cells could be rescued by the ectopic MCU (Supplementary Fig. S2F and S2G). Conversely, MCU OE promoted cell migration and invasion (Supplementary Fig. S2H).
The glucose concentration in the tumor microenvironment could be as much as 10 times lower than in the blood circulation. Although MCU KO had no effect on cell survival when cultured under physiologi concentration of glucose, it increased cell death when PDAC cells were exposed to glucose limitation (1 mmol/L Glc; Supplementary Fig. S2I). The metabolic stress sensitivities in MCU KO cells were more pronounced in glucose-free conditions (Fig. 2E), which could be rescued by ectopic MCU (Supplementary Fig. S2G). Conversely, MCU OE increased resistance to glucose deprivation (Fig. 2F).
Cell migration and invasion are critical for metastatic dissemination, while metabolic stress resistance promotes metastatic colonization (42). Therefore, we evaluated the role of MCU in PDAC metastasis in an orthotopic xenograft model. We used short hairpin RNA (shRNA) to knockdown (KD) MCU and used MCU-EGFP to rescue the KD in Panc-1 cells (Supplementary Fig. S2J). After orthotopic implantation of luciferase-labeled Panc-1 cells, visible development of multiple luminescence foci were detected on day 49 in the control and rescue groups (Fig. 2G, red arrow; Supplementary Fig. S2K), which was indicative of metastasis. Ex vivo examination confirmed the development of multiple metastatic lesions in the liver and the peritoneal cavity in control and MCU rescue groups (Fig. 2H; Supplementary Fig. S2K–S2N). Although MCU KO only modestly decreased the tumor weight of orthotopic Panc-1 tumor (Fig. 2I; P = 0.083), the development of metastases was completely abrogated (Fig. 2J). Ectopic expression of MCU in KD cells was able to restore metastasis to liver, peritoneal cavity and promote orthotopic tumor growth (Fig. 2G and J; Supplementary Fig. S2J–S2N). To further evaluate the role of MCU in PDX metastasis, we knocked out MCU in the MCU-high PDX 677 line. As shown in Fig. 2K and L, S2O and S2P, the orthotopically injected PDX677 was able to established primary tumor in the pancreas and metastasis to the liver and peritoneal cavity. MCU KO reduced the growth of primary tumor (Fig. 2K) and the average number of metastatic lesions (Fig. 2L) and the number of mice that developed liver metastasis (Supplementary Fig. S2O–S2Q). The lack of liver metastasis in the KO group was confirmed by H&E staining (Supplementary Fig. S2Q). These data support that MCU is required for PDAC metastasis.
MCU activates Keap1-Nrf2 antioxidant response in PDAC by increasing mROS
It has been previously reported that MCU promotes breast cancer cell migration and invasion by activation of HIF1α (16). However, we found that MCU KO or OE had no effects of HIF1α protein levels in PDAC (Supplementary Fig. S3A). To understand the mechanisms underlying MCU-mediated PDAC metastasis, we interrogated TCGA RNA-seq database to examine genes that correlated with MCU in patients with PDAC. Interestingly, MCU expression highly correlated with multiple Nrf2 target genes containing ARE (Fig. 3A). The overexpression of MCU in PDAC cells was sufficient to increase Nrf2 levels (Fig. 3B), which was further enhanced by Glc-limitation. Conversely, MCU KO remarkably decreased the levels of Nrf2 protein under Glc limitation (Fig. 3C). MCU overexpression also increased the mRNA levels of classical Nrf2 target genes (Fig. 3D) and promoted the binding of Nrf2 to ARE elements (Supplementary Fig. S3B). ARE luciferase reporter assay indicated that MCU KO inhibited, while MCU OE enhanced, Nrf2 transcriptional activities (Supplementary Fig. S3C). Taken together, our data indicate robust activation of Nrf2 by MCU in PDAC cells.
We found no difference in Nrf2 mRNA levels after MCU OE or KO (Supplementary Fig. S3D). However, there was significant promotion of Keap1 oxidation in MCU OE PDAC cells (Fig. 3E). Conversely, MCU KO remarkably inhibited Keap1 oxidation (Fig. 3F). It has been previously reported that MCU-mediated mitochondrial Ca2+ uptake increased mROS (16). We confirmed elevated mROS levels in MCU OE PDAC cells (Supplementary Fig. S3E–S3F) and depressed mROS levels in MCU KO cells (Supplementary Fig. S3G–S3H). MCU levels in PDX lines correlated with basal mROS levels, (Supplementary Fig. S3I). Treatment of Panc-1 and AsPC-1 cells with mT abrogated MCU-mediated Keap1 oxidation and Nrf2 accumulation (Fig. 3G). Taken together, our data suggest that mROS is required for MCU-mediated Nrf2 activation and Keap1 oxidation.
MCU expression correlates with Nrf2 levels in patients with PDAC
To determine whether MCU might indeed regulate Nrf2 among patients with PDAC, we used IHC to evaluate the correlation between MCU and Nrf2 staining in patients with PDAC (Fig. 3H). Strong Nrf2 staining was remarkably associated with poor OS, RFS, differentiation status, and pTNM stages among patients with PDAC (Fig. 3I; Supplementary Table S3). Importantly, MCU-high patients also had stronger Nrf2 staining (Fig. 3H and J). Our data support that MCU promotes PDAC progression through Nrf2 upregulation and activation in patients with PDAC.
The Keap1-Nrf2 circuit is required for MCU-mediated metastasis
To determine whether Nrf2 is responsible for the prometastasis effects of MCU, we used two independent shRNA to KD Nrf2 expression in control and MCU OE cells (Supplementary Fig. S4A). Nrf2 KD reduced the migration speed of individual cells without affecting cell viability under glucose replete conditions (Fig.S4B–S4D; Supplementary Video S1–S4). Nrf2 KD abrogated MCU-mediated increase in cell migration, invasion (Fig. 4A–C; S4E–S4G), and resistance to glucose deprivation (Fig.4D). Conversely, we activated Nrf2 in MCU KO AsPC1 and MiaPaCa2 cells by knocking down Keap1 (Fig. 4E). Keap1 KD was able to increase Nrf2 and rescue Nrf2 levels in MCU KO cells (Fig. 4E). Nrf2 activation with Keap1 depletion rescued cell migration, invasion, and restore resistance to glucose deprivation in MCU KO cells (Fig. 4F–I). Our previous data showed that mROS was required for MCU to activate the Keap1-Nrf2 signaling. Therefore, we further investigated the effect of mT treatment in PDAC cells. mT treatment inhibited cell migration, invasion, and wound closure in control and MCU OE cells (Supplementary Fig. S4H–S4K).
To further investigate the role of Nrf2 in MCU-mediated metastasis, we implant control or MCU OE SW1990 cells with or without Nrf2 KD into the pancreas of nude mice. Eight weeks after implantation, mice were euthanized, the tumor weight, number of metastatic lesions in the peritoneal cavity, and the liver were evaluated (Fig. 4J–M; Supplementary Fig. S4L–S4N). MCU OE increased the average number of metastatic lesion by approximately 2-fold without affecting primary tumor size (Fig. 4J–M). The shRNA depletion of Nrf2 abrogated MCU-mediated metastasis (Fig. 4L and M; Supplementary Fig. S4L and S4M) and reduced tumor growth (Fig. 4J and K). The lack of liver metastasis in Nrf2 KD groups was confirmed with H&E staining (Supplementary Fig. S4N). These data support that MCU promotes pancreatic cancer cell migration, invasion, and metabolic stress resistance through activation of Nrf2.
MCU-Nrf2 signaling promotes SLC7A11 expression and cystine uptake in PDAC
To identify targetable effectors responsible for MCU-mediated PDAC metastasis and progression, we evaluated MCU-mediated transcriptomic changes in pancreatic cancer cells. A total of 182 genes were upregulated (and 180 downregulated) in MCU OE Panc-1 cells under Glc-limited conditions. Upregulated genes include multiple ARE genes (Supplementary Fig. S5A; Supplementary Table S4), which is consistent with the notion that MCU activates Nrf2-mediated antioxidant response. 8 of the 182 upregulated genes are also positively correlated with MCU expression in TCGA PDAC cohort (Spearman rank r > 0.4; Fig. 5A). The top two of those eight genes are Nrf2 target genes (SLC7A11 and IDH1; Fig. 5B). Of particular interest is SLC7A11, a functional subunit of the cystine/glutamate antiporter (system xc−) responsible for the uptake of cystine and export of glutamate (Fig. 5B; refs. 30, 43). Although cyst(e)ine is a nonessential amino acid, cancer cells depend on SLC7A11-mediated cystine uptake for GSH synthesis and antioxidant response. Interestingly, SLC3A2, the regulatory unit of system xc−, and SLC38A2 (SNAT2, responsible for neutral amino acid such as glutamine transport) and GCLC/GCLM (responsible for the rate-limiting step of GSH synthesis) were among the upregulated genes in MCU OE cells (Supplementary Fig. S4A and S4B; Supplementary Table S4). MCU-high PDX lines (PDX 677 and 977) had higher levels of SLC7A11 than MCU-low PDX samples (PDX 202 and 081; Fig. 5C). MCU OE increased, while MCU KO decreased the levels of SLC7A11 (Fig. 5D). MCU-mediated upregulation of SLC7A11 was abrogated by Nrf2 KD (Fig. 5E and F). MCU OE also increased the binding of Nrf2 to SLC7A11 promoter (Fig. 5G).
System xc− exports glutamate while importing cystine. Glutamine is the main source of glutamate used by system xc− for cystine uptake (44). MCU OE increased glutamine uptake and glutamate secretion in pancreatic cancer cells, especially when cells were subjected to oxidative stress induced by H2O2 (Fig. 5H and I). Conversely MCU KO inhibited glutamine uptake and glutamate secretion (Fig. 5H and I). Overexpression of MCU significantly increased the uptake of FITC-cystine in Panc-1 cells (Fig. 5J and K), which was blocked by erastin. Conversely, MCU KO inhibited FITC-cystine uptake (Fig. 5J and K). Taken together, our data indicate that the MCU upregulates SLC7A11 and cystine uptake through Nrf2.
Activation of mitochondrial Ca2+ signaling promotes cystine addiction and sensitizes PDAC cells to ferroptosis
Cystine deprivation induced the transcription of “Ferroptosis Signature” genes characterized by activation of the Gcn2-eIF2α-ATF4 branch of integrated stress response (43, 45). Our RNA-seq data revealed robust upregulations of Ferroptosis Signature genes in MCU-overexpressing cells (Fig. 6A and B). We hypothesized that mROS production in MCU-upregulated PDAC might lead to increased consumption of cysteine required for antioxidant response. The high demand for cystine in MCU OE PDAC cells could lead to intracellular cysteine deficiency despite SLC7A11 upregulation. MCU OE PDAC cells indeed had lower intracellular levels of cysteine/cystine, which could be rescued by supplementation of additional cystine (1 mmol/L; Supplementary Fig. S6A). LC/MS assay coupled with NEM (n-ethyl maleimide) derivatization showed that, MCU OE also reduced the levels of intracellular cysteine, which could be rescued by cystine supplementation (Fig. 6C).
We observed marked upregulation of ATF4 protein levels and the expression Ferroptosis Signature transcripts in MCU OE cells (Fig.6D and E). Importantly, cystine supplementation in the medium reduced the levels of ATF4 and Ferroptosis Signature genes in MCU OE cells (Fig. 6D and E).
We hypothesized that cystine addiction in MCU-overexpressing PDAC might constitute a therapeutic vulnerability. To investigate this possibility, we used three different approaches (cystine-free medium, glutamate, or erastin) to suppress cystine uptake. MCU-overexpressing PDAC cells were hypersensitive to cystine deprivation–induced cell death (Fig. 6F; Supplementary Fig. S6B). In contrast, MCU slightly reduced sensitivities to ferroptosis induced by GPX4 inhibitor RSL3 (Supplementary Fig. S6C). Conversely, MCU KO inhibited ferroptosis induced by cystine deprivation (Supplementary Fig. S6D). Cystine deprivation–induced cell death in MCU OE cells could be rescued by ferroptosis inhibitor Fer-1, but not by ZVAD, Nec-1, or CsA (Fig. 6F), which suggested that MCU promotes ferroptosis induced by cystine deprivation (but not by GPX4 inhibition).
It was recently reported that mROS plays a critical role in promoting ferroptosis (35, 46). We observed that MCU promoted lipid peroxidation in PDAC cells 3 to 6 hours after cystine deprivation and some of the lipid peroxidation colocalized with mitochondria (Fig.6G; Supplementary Fig. S6E). Flow cytometry analysis confirmed MCU OE elevated lipid peroxidation, while MCU KO decreased lipid peroxidation (Fig. 6H and I; Supplementary Fig. S6F and S6G) in cystine-deprived PDAC cells. The MCU-mediated lipid peroxidation could be rescued by mROS scavenging with mT (Fig. 6H and I). mT also abrogated the ferroptosis-promoting effects of MCU (Fig. 6J). Taken together, these data support that MCU-mediated mROS is responsible for cystine addiction and promoting ferroptosis in PDAC.
It has been previously reported that inhibition of NCLX was able to increase mCa2+ and mROS levels (17, 47). As shown in Supplementary Fig. S6H, inhibition of NCLX with the pharmacologic inhibitor CGP37157 did not affect PDAC cell viability in the presence of cystine; however, when combined with cystine deprivation CGP37157 robustly increased ferroptosis (Supplementary Fig. S6H) and lipid peroxidation (Supplementary Fig. S6I and S6J). The effect of CGP37157-mediated cell death and lipid peroxidation could be abrogated by DFO or Fer-1 (Supplementary Fig. S6H and S6J). Our data suggested that elevation of mCa2+ is sufficient to sensitize PDAC cells to ferroptosis.
MCU-driven cystine addiction is a therapeutic vulnerability that can be exploited to prevent PDAC metastasis
Next, we examined whether MCU-driven cystine addiction could be exploited as a therapeutic vulnerability using SLC7A11 inhibitors SAS and IKE (Fig. 7A and B; Supplementary Fig. S7A). BLI data showed that SAS and IKE have much stronger inhibitory effects on tumor growth in MCU-overexpressing groups (Fig. 7A and B). After euthanasia, the primary tumor growth and distant metastases in these mice were analyzed (Fig. 7C-F). MCU significantly promoted metastasis to both the liver and the peritoneal cavity (Fig. 7C–E) and moderately increased tumor weight (Fig. 7F). The medium BLI signals from metastatic lesion in the MCU OE (vehicle) group were more than 10 times higher than that in the control group (vehicle; Fig. 7E). These data support the notion that MCU promotes pancreatic cancer metastasis.
The IKE treatment in the MCU OE group resulted in complete tumor regression in 50% of mice (Fig. 7A and B; Supplementary Fig. S7B) and abolished the development of liver and peritoneal metastasis (Fig. 7C–E). SAS treatment led to strong inhibition of both liver and peritoneal metastasis in the MCU OE group (Fig. 7C–E). The efficacies of IKE was much stronger than SAS, probably due to the fact that IKE (IC50: 34 nmol/L) was a much more potent inducer of ferroptosis than SAS (IC50: 160 μmol/L). H&E staining of harvested orthotopic tumor revealed formation of lipid droplet-like structures (Fig. 7G, arrow) in SAS- and IKE-treated mice, which is consistent with a previous report (45). SAS- and IKE-induced larger and more numerous lipid droplets, more intense 4-HNE staining and large areas of necrosis in MCU OE groups (Fig. 7G and H; Supplementary Fig. S7C), which is suggestive of increased ferroptosis in these tumors. Taken together, these data support that MCU-overexpressing PDAC are hypersensitive to cystine deprivation and SLC7A11 inhibitors could be used to prevent metastatic recurrence among these patients with PDAC.
MCU-high PDAC is more sensitive to cystine deprivation
We prepare patient-derived organoids (PDO) from 3 patients with MCU-high and 3 patients with MCU-low PDAC (Supplementary Fig. S8A) and tested their sensitivities to erastin. As shown in Fig. 8A and B, erastin treatment inhibited the growth of the MCU-low and MCU-high PDO lines by 52.6% ± 3.1% and 76.0% ± 9.1%, respectively, which suggested patients with MCU-high PDAC might be more sensitive to cystine-deprivation treatment. To further examine this notion, we implanted PDX tumors derived from 2 of the MCU-high patients (patient 677 and 977) and 2 of the MCU-low patients (patient 022 and 081) into NSG mice (Supplementary Fig. S7B). After the tumors reached 50–70 mm3, mice were treated with vehicle control or IKE (20 mg/kg) for 40 days. IKE treatment inhibited the growth of the two MCU-high PDX tumor lines by 73.4% ± 4.5% (PDX677) to 88.4% ± 2.7% (PDX977; Fig. 8C). In contrast, the inhibitory effects of IKE on the growth of MCU-low PDX tumors were only 14.7% ± 12.8% (PDX202) to 31.4% ± 8.0% (PDX081; Fig. 8C). IKE treatment induced significantly more lipid droplets and stronger 4-HNE staining in MCU-high PDX tumors (Fig. 8D and E; Supplementary Fig. S7C and S7D). To determine whether MCU plays a causative role in cystine addiction, we used CRISPR/Cas9 to KO MCU in cell lines derived from the MCU-high PDX677 (Supplementary Fig. S8E). MCU KO abrogated the antitumor effects of IKE treatment in this MCU-high PDX line (Fig. 8F and G; Supplementary Fig. S8F), which suggested that MCU is required for sensitivities to IKE. Taken together, MCU-high PDAC are more sensitive to SLC7A11 inhibition than MCU-low tumors. The cystine addiction in patients with MCU-high PDAC could be exploited as a therapeutic vulnerability to inhibit tumor growth and prevent metastasis.
Discussion
Pancreatic cancer is a highly metastatic tumor malignancy with poor prognosis and limited treatment options. Here we showed that MCU is upregulated during pancreatic cancer initiation and progression. MCU promotes cell migration, invasion, and metabolic/oxidative stress resistance in cell culture and xenograft models. The mitochondria are the major source of ROS in the cell. The elevated mCa2+ increases mROS by promoting the tricarboxylic acid cycle and oxidative phosphorylation (11). ROS play complicate and context-dependent role in cancer initiation and progression. Although it is known that excessive ROS increase could lead to cell death, it is also well established that modest ROS production could promote cell migration, invasion, and metastasis by activating HIF1α signaling and inhibiting protein phosphatases such as PTEN (18). Our data show that by activating the Keap1-Nrf2 antioxidant response, MCU-mediated mROS signal contribute to promoting metabolic stress resistance and metastatic progression.
Cancer cells depend on SLC7A11-mediated cystine uptake for GSH synthesis antioxidant response (44). GSH is required for GPX4 to reduce lipid peroxidation and prevent ferroptosis (48). Although mitochondria might not be essential for ferroptosis, mitochondria-derived ROS promotes ferroptosis induced by cystine deprivation, but not by GPX4 inhibitors (35). By increasing mROS, MCU remarkably increased lipid peroxidation and ferroptosis under cystine deprivation. Interestingly, MCU modestly promote resistance to ferroptosis induced by GPX4 inhibitor RSL3. Among cysteine metabolites, Coenzyme A has been shown to have anti-ferroptosis activities (45).
Our data supported that the MCU-mediated antioxidant response and metastatic progression has high demand for cystine uptake from the tumor microenvironment. When cystine is available, MCU-mediated mROS oxidizes Keap1 and upregulates Nrf2 antioxidant response to promote metabolic stress resistance and pancreatic cancer metastasis. However, upon SLC7A11 inhibition or cystine deprivation, MCU-mediated mROS promotes lipid peroxidation and exacerbates ferroptosis. Therefore, the cystine addiction in MCU-overexpressing PDAC could be potentially exploited as a metabolic vulnerability. There are FDA-approved drugs (e.g., sorafenib, SAS) that could be repurposed to inhibit SLC7A11 (43). More specific inhibitors with higher potencies and favorable pharmacokinetics (e.g., IKE) have also been developed (49). These inhibitors could be employed to prevent metastatic recurrence in MCU-overexpressing PDAC.
Authors' Disclosures
S. Yang reports grants from NCI during the conduct of the study. No disclosures were reported by the other authors.
Authors' Contributions
X. Wang: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. Y. Li: Data curation, investigation. Z. Li: Data curation, investigation. S. Lin: Data curation. H. Wang: Data curation, investigation. J. Sun: Data curation, investigation. C. Lan: Data curation, investigation. L. Wu: Data curation, investigation. D. Sun: Data curation. C. Huang: Resources, methodology. P.K. Singh: Writing–review and editing. N. Hempel: Methodology, writing–review and editing. M. Trebak: Methodology, writing–review and editing. G.M. DeNicola: Resources, writing–review and editing. J. Hao: Conceptualization, supervision, funding acquisition, writing–review and editing. S. Yang: Conceptualization, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.
Acknowledgments
The authors thank Dr. Katherine Aird for assistance with the use of YSI 2700 Biochemical Analyzer, Ryan Yoast for assistance with mitochondrial calcium uptake assay, and Yuka Imamura for help with RNA-seq data analysis. S. Yang is supported by grants from the NIH (R01CA233844, R01CA256911). J. Hao is supported by National Natural Science Foundation of China (grants 81720108028, 81672431).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.