The FDA-Approved Anthelmintic Pyrvinium Pamoate Inhibits Pancreatic Cancer Cells in Nutrient-Depleted Conditions by Targeting the Mitochondria

Pancreatic ductal adenocarcinoma (PDAC) is a lethal cancer with a five-year overall survival rate of 10% (1). Contributing to this high mortality is the resistant nature of PDAC, which has been largely attributed to the PDAC tumor–microenvironment (TME). The PDAC TME is nutrient poor, with decreased vascularization leading to both hypoxic and hypoglycemic conditions, which drive chemoresistance (2–4). Although hypoxic, the levels of oxygenation displayed in PDAC (1.5% oxygen vs. normal 5% oxygen) are still permissive for the effective utilization of the electron transport chain (5, 6). Accordingly, PDAC tumors adapt and become more reliant on oxidative phosphorylation as opposed to glycolysis to increase glucose efficiency (36 vs. 2 ATP per glucose molecule, respectively; refs. 5, 7). Increased reliance on mitochondria in PDAC xenografts has been demonstrated by simultaneously measuring oxygenation, glucose uptake, and glycolysis spatially throughout tumors (6). Accordingly, multiple mitochondrial inhibitors have shown promise in preclinical models for the treatment of PDAC (7–9); thus, there is a strong premise to target mitochondria in the uniquely nutrient-deprived TME of PDAC tumors (4).

To target the nutrient-deprived PDAC TME (3, 5, 6), we utilized the FDA-approved drug pyrvinium pamoate (PP), which has previously been shown to have multiple potential mechanisms of action including targeting mitochondrial pathways (10–16) and to be particularly effective in low glucose settings.

Cell lines (PANC-1, Capan-1, HS766T, MIA-PaCa2, and CFPAC) were purchased from ATCC. HuR knockout clones were generated as previously described (17, 18). PDX line SPC_145 was a gift from Dr. Talia Golan (Sheba Medical Center) (19). All cell lines were tested monthly for mycoplasma and every 6 months STR validated. Cell lines for all experiments were below passage 20.

For IC₅₀ analysis, cells were seeded at 1,000 cells/well in 96-well plates and treated with drugs the next day. Cell viability was assessed after five days utilizing Quant-iT Pico Green (Invitrogen) and a GloMax Explorer Multimode Microplate Reader (Promega). IC₅₀ values were calculated utilizing nonlinear curve fitting in GraphPad Prism 8.0.1. (GraphPad Software).

Colony formation was performed by plating 1,000 cells/well of a 6-well plate, followed by drug dosing the next day. After ten days, colony growth was assessed by staining as previously described (18).

For experiments comparing differing glucose conditions, cells were pretreated for 12–24 hours with indicated glucose concentration prior to drug treatment.

Organoids were cultured as previously described (17). For organoid IC₅₀s, organoids were broken down to single cells and seeded at 1,500 cells/well in a 96-well plate. The next day, cells were treated with drug. After five days, cell viability was determined via CellTiter-Glo (Promega) and a GloMax Explorer Multimode Microplate Reader (Promega). IC₅₀s were calculated utilizing nonlinear curve fitting in GraphPad Prism 8.0.1.

Western blots were performed using the following antibodies α-Tubulin (B-5–1-2, 1:5,000, Thermo Scientific); HuR (3A2, 1:4,000, Santa Cruz Biotechnology), COX-2 (H-62, 1:1,000, Santa Cruz), Total OXPHOS Human WB Antibody Cocktail (ab110411, 1:1,000, Abcam), Anti-TOMM20 antibody (ab56783, 1:1,000, Abcam), and Beta-actin (13E5 1:10,000, Cell Signaling Technology). Membranes were scanned using Odyssey Infrared Imaging System (LI-COR Biosciences) and analyzed with Image Studio Lite Version 5.2.

Cells were plated at 10,000 cells/well in a 24-well plate containing 8-mm coverslips. Cells were treated for 1 hour with PP 0.3 μmol/L and/or 0.1 μmol/L MitoTracker Green FM (Thermo Fisher). PP was imaged using the Texas Red channel. Coverslips were then removed and imaged live within 10 minutes with a Leica DM4 B microscope (Leica Microsystems).

IHC was performed as previously described (20). Patient samples were collected in compliance with IRB 06U.76. Mouse tumors were TUNEL stained using diaminobenzidine (DAB). Five 20× images per group were quantified and compared utilizing a t test in GraphPad Prism 8.0.1.

PANC-1 cells were plated at 6 × 10⁶ cells/10 cm dish. Cells were treated with PP (10 μmol/L), DHTS (10 μmol/L), or DMSO for 4 hours at 37°C prior to collection. Cell pellets were suspended in PBS with suitable protease inhibitor. Aliquots of 5 × 10⁵ cells/100 μL were made in separate tubes. On a thermocycler, cells were exposed to different temperatures (37°C–61°C) for 3 minutes and then kept directly on dry ice. Cells were lysed by freeze–thaw cycle (dry ice to 25°C three times). Cells were centrifuged at 20,000 × g for 20 minutes at 4°C, supernatant was collected and mixed with loading buffer (PBS and SDS loading buffer). Samples were analyzed by immunoblotting.

HS766T, MIA-PaCa2, and PANC-1 cells were seeded at 500,000 cells/10 cm plate and treated with PP at 0.3 μmol/L for 48 hours. Cells were collected and run for phosphoproteomic reverse-phase microarray analysis (previously described; ref. 21). Phosphoproteomic analysis was performed utilizing Genesis version 1.8.1 and Average Linkage WPMGA Pearson Correlation Hierarchical Clustering.

MIA-PaCa2 cells were plated in 6-cm dishes 24 hours prior to treatment with fresh medium containing DMSO or 0.3 μmol/L PP. After treatment for 6 hours, cells were washed with PBS, and metabolites were extracted with 1 mL cold 80% methanol/water on dry ice for subsequent LC-MS/MS analysis (22). For mouse samples, mice were treated with 1 mg/kg PP IP and blood was collected utilizing heperanized tubes and retro-orbital bleeding at the indicated time points. A hydrophilic interaction chromatography method (HILIC) on an Acquity UPLC BEH Amide column (2.1 × 150 mm i.d., 1.7 μm; Waters) was used for compound separation. A Xevo-TQ-S mass spectrometer (Waters) was used to detect all compounds with multiple reaction monitoring (MRM) methods. Peak areas were integrated using Skyline (23) and normalized to cell number. Peak areas were analyzed using MetaboAnalyst 4.0 (http://www.metaboanalyst.ca/MetaboAnalyst/; ref. 24).

Cells were plated at 5,000 cells/well in a 96-well plate. Cells were pretreated for 16–24 hours in media without glucose or FBS. Cells were treated with oligomycin A (10 μmol/L), FCCP (1 μmol/L), gemcitabine (1 μmol/L), or PP (0.3 μmol/L) ± 25 mmol/L glucose for 1 hour, and ATP was quantified utilizing CellTiter-Glo (Promega; cat. #G7570).

OCR and extracellular acidification rate (ECAR) were measured using the Seahorse XFp Analyzer (Seahorse Bioscience). MIA-PaCa2 cells were plated at 22,000 cells/well on XFp cell plates 24 prior to 12 hours treatment with 0.3 μmol/L PP or vehicle. The concentration of FCCP, antimycin A, rotenone, and oligomycin A was 4, 1, 4, and 2 μmol/L. For acute treatment, PP in PBS at 0.3 μmol/L was injected after basal OCR measurement and monitored for 90 minutes, followed by inhibiting the complex IV and then the electron transport chain. For oxygen consumption quantification, total OCR was defined as the basal OCR rates, nonmitochondrial OCR is defined as the OCR subsequent to antimycin A and rotenone treatment, and mitochondrial OCR is defined as the difference between the two.

MIA-PaCa2 cells were plated at 1 × 10⁶ cells/10 cm plate and treated with PP at 0.3 μmol/L for indicated times, at which time cells were collected and DNA extracted (DNeasy blood and tissue kit, Qiagen). Mitochondrial DNA was compared with nuclear DNA using qRT-PCR.

MIA-PaCa2 cells were plated at 1 × 10⁶ cells/10 cm plate and treated with PP at 0.3 μmol/L for 1 hour, at which time cells were collected and RNA extracted (RNeasy Mini Kit, Qiagen). Mitochondrial RNA was quantified using qRT-PCR utilizing 18s as a reference.

MIA-PaCa2 cells were plated at 1 × 10⁶ cells/10 cm plate and treated with PP at 0.3 μmol/L, oligomycin at 4 μmol/L or rotenone at 0.5 μmol/L for 48 hours at which time cells were collected and RNA extracted (RNeasy Mini Kit, Qiagen). Whole-exome sequencing was performed by Novogene utilizing Illumina Platform PE150. Single-sample gene set enrichment analysis was performed using hallmark gene sets from the MSigDB Collections at https://www.gsea-msigdb.org/gsea/msigdb/collections.jsp as previously described (25). Integrative metabolomic and RNA-seq analysis was performed using MetaboAnalyst5.0 (26). Results can be found at GSE176305.

Synergy was determined utilizing BLISS synergy with Combenefit software (27).

We used 6-week-old Athymic female nude mice from Envigo; all procedures were approved by IACUC. Mice were engrafted subcutaneously with 5 × 10⁶ MIA-PaCa2 cells/flank, prepared in 200-μL solution comprised of 1:1 DPBS:Matrigel. For the first PP efficacy, mice were randomized with tumors at 100 mm³ to either vehicle or 1 mg/kg PP 3 times weekly. One PP treated mouse was removed from analysis after utilizing Grubb's outlier test, and one died 11 days into the study. It is unknown if this was drug related. The control arm is the same control arm in a previously published paper as we tested both drugs in vivo at the same time and shared control animals (18).

For the Kaplan–Meier study, mice were randomized when tumor volumes reached 100 mm³ into groups: vehicle and PP IP 1 mg/kg, with 10 mice per arm. Mice were sacrificed when tumor burden reached 2,000 mm³.

For analysis of human-grade oral PP treatment, mice were randomized, and treatment was initiated the day after engraftment. The arms for this experiment were 5 mg/kg, 20 mg/kg, 35 mg/kg, and vehicle. Pyrvin tablets were crushed, resuspended in saline, and dosed orally 5 days/week. There were five mice/arm except for the vehicle arm with ten mice.

For pharmacokinetic IV work, mice were dosed at 1 mg/kg PP (100 μL injection) by tail vein, and blood was collected utilizing heperanized tubes and retro-orbital bleeding at the indicated time points. For IP and oral dosing, mice were dosed either IP at 1 mg/kg or orally at 5, 20, or 35 mg/kg, and serum and tissue samples were collected by Alliance Pharma, with blood collected utilizing cardiac puncture. All samples were analyzed by Alliance Pharma using HPLC (UPLC, Shimadzu), MS/MS (API 4000, AB Sciex).

Similar to the method described (28), mitochondrial DNA and control dsDNA fragments were created by mixing oligonucleotides in annealing buffer (10 mmol/L TrisHCl pH 7.52,50 mmol/L NaCl, and 1 mmol/L EDTA). Annealing sequence using a thermocycler was 2 minutes @ 95°C > cool-down at 0.5°C/minute down to 65°C > 20 minutes @ 65°C > cool-down at 0.5°C/minute down to 55°C > 45 minutes @ 55°C > cool-down at 0.5°C/minute down to 35°C. dsDNA fragments were in buffer (100 mmol/L potassium chloride and 20 mmol/L potassium phosphate) for 1 hour to promote and stabilize G-quadruplex formation. PP was added at a ratio of 1:2 (4 μmol/L pyrvinium and 8 μmol/L DNA) and incubated in the dark for 1 hour at RT and measured using the Bio-Rad ChemiDoc MP (Bio-Rad Laboratories, Inc.) with excitation/emission of 556 nm/573 nm to recognize the spectral and intensity shift previously described by Stockert et al (29). Oligonucleotides used can be found as previously described (28, 30).

Mitochondrial DNA was amplified from PANC-1 DNA using the REPLI-g Mitochondrial DNA Kit (Qiagen). Similar to the method described previously (30), PCR reactions were run with increasing concentrations of PP. Briefly, 25 μL reactions were performed with 0.625 units Ex-Taq DNA Polymerase (Takara), 0.2 mmol/L dNTP, 1 μmol/L of each primer, and 0.375 μL of mtDNA. Reaction mixes were incubated on ice for 15 minutes and then moved to the thermocycler for a sequence of 30 cycles of (10 seconds at 98°C > 30 seconds at 55°C > 15 seconds at 72). Reaction samples were run on a 0.8% agarose gel and imaged on a Bio-Rad ChemiDoc MP (Bio-Rad Laboratories, Inc.).

Primers used are as follows (target fwd/reverse):

• ND1 GGCTACTACAACCCTTCGCT/ACGGGAAGGGTATAACCAACA,

• ND2 ATCATAGCAGGCAGTTGAGG/GGGTGGGTTTTGTATGTTCA,

• Non-G4 TCGAGTCTCCCTTCACCATT/GCACTCGTAAGGGGTGGAT.

For all figures, asterisks are designated for P values (or q-values where appropriate) as follows: *, <0.05; **, <0.01; ***, <0.005; ****, <0.001 unless otherwise designated.

Publicly available data for the activity of PP across the NCI60, and expression levels of genes from microarray data were accessed using CellMinerCDB version 1.2 (31).

To determine PP efficacy in vitro, we tested PP in a panel of PDAC and PDX-derived cell lines where we found PP to be effective at inhibiting cell viability at the nanomolar range (HS766T, MIA-PaCa2, PANC-1, CFPAC, SPC_145 respective IC₅₀ values of 93, 40, 92, 21, 9, 93, 58, and 9 nmol/L). PP effectively inhibited the growth of two PDAC organoid lines KPC and KTC (KTC having an activating KRAS mutation and HuR overexpression and KPC organoids having an activating KRAS mutation and p53 inactivation; ref. 32) with IC₅₀ values of 93 and 58 nmol/L, respectively. These results indicated that PP was effective at inhibiting cell growth in multiple PDAC models.

Several mechanisms of action have been proposed for PP, including the inhibition of the RNA binding protein HuR (15). HuR has been recognized as a vital target in PDAC, and to be active in the nutrient-deprived PDAC TME (3, 32, 33). Utilizing a cellular thermal shift assay we demonstrated that PP did not strongly interact with HuR as compared with the validated HuR inhibitor dihydrotanshinone I (Supplementary Fig. S1A–S1C; ref. 34). To assess if PP was indirectly affecting HuR, we tested PP along with several control therapeutics in HuR KO cells. We found HuR knockout lines were more sensitive to multiple clinically utilized therapeutics tested (Supplementary Table S1); however, they were not resistant to PP (Supplementary Table S1). We confirmed the lack of resistance to PP using a colony formation assay(Supplementary Fig. S1D–S1F). We were unable to supplement these in vitro findings with in vivo experimentation, as the HuR KO cells were xenograft lethal (Supplementary Fig. S1G; ref. 17). These results demonstrate that PP induces PDAC cell death in an HuR-independent manner.

Several other mechanisms of action for PP have been proposed including targeting autophagy, mitochondrial oxidative phosphorylation, WNT response, the unfolded protein response, androgen receptor signaling, and adipogenic differentiation (10–16). To investigate which if any of these mechanisms of action were predominant in PDAC cells in an unbiased manner, we utilized phosphoproteomic reverse-phase microarray (PRPMA). Across HS766T, PANC-1, and MIA-PaCa2 cells, we found a number of significantly affected targets were downregulated, in particular several consistently downregulated genes were regulators of cell-cycle and cellular energetics pathways including S6 ribosomal protein s240/244, RB S780, and mTOR S2448 among others. Of the commonly upregulated genes many were associated with cell scavenging, inflammation, and cell death including cleaved caspase-3, and cleaved lamin A and AKT S473 (Supplementary Fig. S2A-S2C). Based on these findings, we hypothesized that the broad changes observed from our PRPMA data could result from a large shift in cellular metabolism.

To investigate if/how PP was affecting metabolism in PDAC, we performed a comprehensive and unbiased analysis of metabolites after PP treatment. We observed significant modulations in metabolites involved in the citric acid cycle upon PP treatment (Fig. 1A), along with increases in glycolytic metabolites (Supplementary Fig. S2D). This may indicate a compensatory reaction of increased glycolysis in response to PP induced inhibition of the TCA cycle/oxidative phosphorylation. There were several relatively small changes among arginine and proline metabolites (Supplementary Fig. S2E). Interestingly, we observed a large increase in N-carbamoyl aspartate indicating alterations in pyrimidine biosynthesis (Fig. 1B). We also observed an increase in long-chain acylcarnitine levels, but significant decreases in free carnitine along with short and medium acylcarnitine chains (Fig. 1C). In particular, we observed large increases in acylcarnitine C4-OH and acylcarnitine C5-DC, which have been clinically characterized to be markers of mitochondrial disfunction leading to medium/short-chain acyl-CoA dehydrogenase deficiency (M/SCHAD deficiency) and glutaric acidemia type I, respectively (35). These data suggest an alteration of multiple mitochondrial pathways upon PP treatment including an inhibition of the citric acid cycle, alteration of fatty acid metabolism, and inhibition of pyrimidine biosynthesis (Fig. 1D).

To assess the effect of PP on mitochondria, we first confirmed previous reports that PP localizes to mitochondria (36). In PDAC cells, PP localized to the mitochondria for prolonged periods of time (Fig. 2A; Supplementary Fig. S3A and S3B; refs. 12–S14). Chronic treatment (12-hour PP pretreatment) caused a significant decrease of mitochondrial oxygen consumption (i.e., the electron transport chain), while not significantly effecting nonmitochondrial oxygen consumption (Fig. 2B and C). In control cells, acute oligomycin treatment caused an increase in glycolysis (1.44-fold increase P < 0.0001; Fig. 2D), as determined through measuring the extracellular acidification rate (ECAR). This was expected as oligomycin A inhibits mitochondrial ATP generation, and cells frequently upregulate glycolytic flux (glycolysis) to compensate. PP pretreated cells were insensitive to oligomycin A with no increase in glycolysis, consistent with the hypothesis that PP pretreatment inhibited mitochondrial ATP generation (Fig. 2D). Conversely, acute PP treatment caused a loss of mitochondrial oxidative phosphorylation along with a concomitant increase in glycolytic function (1.56-fold increase P < 0.0001; Fig. 2E and F). These results complemented our metabolomics data where PP treatment resulted in altered glycolytic metabolites indicating a potential increase in glycolytic function. To prevent cells from upregulating glycolysis in response to PP treatment, we pretreated PDAC cells with no-glucose media. PP treatment in these glucose-starved PDAC cells resulted in a significant decrease in ATP levels, which did not occur with the negative control gemcitabine, and could be rescued with cotreatment with glucose (Fig. 2G). We demonstrated similar results with other mitochondrial inhibitors and were also able to replicate these results in PANC1 cells (Supplementary Fig. S3C and S3D).

As PP effectively inhibited mitochondrial oxidative phosphorylation, we investigated whether PP was more effective under relevant metabolic stress conditions that occur in the PDAC TME (i.e., hypoxia and hypoglycemia). We demonstrated that PP was similarly effective in normoxic and hypoxic conditions, although it was significantly more effective in hypoglycemic conditions (Supplementary Fig. S3E). We demonstrated that the increased efficacy of PP in low glucose was due to a rapid reduction in ATP levels in glucose-starved PDAC cells, which was followed by a decrease in cellular viability, an effect that was negated with glucose supplementation (Fig. 2H; Supplementary Fig. S3F and S3G).

Overall, we found that acute PP treatment led to mitochondrial dysfunction, which, in a glucose-deplete environment, led to a decrease in ATP levels, followed by cell death. These effects could be partially rescued by glucose (Fig. 2I). Importantly, we demonstrated that these effects on mitochondria were preferential to cancer cells by comparing MIA-Paca2 sensitivity to MRC5 (immortalized fibroblasts) and HPNE (human pancreatic normal epithelial cells). PP caused less of a decrease in ATP production in MRC5 and HPNE cells as compared with MIA-PaCa2 cells (Supplementary Fig. S4A). Accordingly, MRC5 and HPNE cells were 111- and 6.6-fold less sensitive to PP-induced cell kill than MIA-PaCa2 cells (Supplementary Fig. S4B). These data indicate that PDAC cells are more susceptible to PP treatment than normal cells.

To further augment the ability of PP to sensitize PDAC cells to their hypoglycemic TME (4), we cotreated cells with PP and the competitive inhibitor of glucose, hexokinase 2-deoxy-glucose (2DG; ref. 37). Both agents were more effective in low glucose conditions, and they acted synergistically to inhibit ATP production (Supplementary Fig. S4C–S4F). We demonstrated 93.9% decreased viability at clinically achievable levels of 2DG (125 mmol/L; ref. 37), and at low nanomolar levels of PP (25 nmol/L; Supplementary Fig. S4G).

We performed RNA-seq analysis comparing PP with the complex I inhibitor rotenone and the ATP synthase inhibitor oligomycin A. Gene set enrichment analyses indicated that PP significantly affected oxidative phosphorylation, fatty acid metabolism, glycolysis, and gene sets related to cell-cycle progression including E2F targets, MYC targets, and G₂–M checkpoint (Fig. 3A and B). Using MetaboAnalyst 5.0, we integrated metabolomic and RNA-seq data (26) and found that PP affected pathways included the citric acid cycle, oxidative phosphorylation, pyrimidine biosynthesis, glycolysis or gluconeogenesis, central carbon in cancer and cell cycle (Fig. 3C).

Comparing PP with oligomycin and rotenone, we found of the 1,022 genes that PP significantly altered, only 13.7% and 4.5% were similarly altered by rotenone or oligomycin, respectively (Fig. 3D). We also observed divergent gene set enrichment results. Interestingly, for gene sets that encompassed mitochondrial function (i.e., oxidative phosphorylation, fatty acid metabolism, glycolysis), rotenone caused similar changes as compared with PP. Conversely, cell-cycle–related pathway alterations caused by PP (i.e., E2F targets, G₂–M checkpoint, MYC targets) were similarly affected by oligomycin treatment (Supplementary Fig. S4H). These data suggest that PP inhibited the mitochondria in a manner alternative to specific complex I or ATP synthase inhibition.

The mitochondrial genome is small in comparison with the nuclear genome, and the mitochondria encodes only 37 genes, with 13 of these genes encoding enzymes involved in oxidative phosphorylation. We observed that PP caused a significant reduction in all mitochondrial protein coding genes preferentially to changes induced in nuclear genes (Fig. 3E). Furthermore, the downregulation of mitochondrial genes was not seen with either rotenone or oligomycin A. We validated the decrease in mitochondrial RNA in separate qPCR experiments (Supplementary Fig. S5A). We found that PP treatment also caused loss of protein expression of multiple members of the mitochondrial electron transport chain, which was not observed with the mitochondrial inhibitors oligomycin A or rotenone (Fig. 4A and B).

To determine if loss of mitochondrial gene and protein expression could be a precursor or sign of mitophagy, we assessed mitochondrial DNA or mitochondrial mass as markers of mitophagy. Six hours of PP treatment caused a slight but significant decrease in mitochondrial DNA (22.9%; P = 0.0097; Supplementary Fig. S5B). Longer treatments (48 hours) did not cause significant changes in mitochondrial TOMM20 levels, which are indicative of total mitochondrial mass (Supplementary Fig. S5C). These data indicated that PP affected mitochondrial gene and protein expression in a manner that was not caused by specific inhibition of complex I or ATP synthase and was not dependent on mitophagy.

DNA G-quadruplexes are tertiary structures formed in guanine-rich regions of either DNA or RNA. The mitochondrial genome contains significantly more potential G-quadruplex forming regions per kb of DNA than nuclear DNA (38). These structures have roles in normal mitochondrial functions; however, some compounds can bind to and stabilize mitochondrial G-quadruplexes, thereby disrupting mitochondrial transcription (39). Mitochondrial genes, unlike nuclear genes, are transcribed in a polycistronic fashion, and thus, failing of the transcription bubble can inhibit the transcription of multiple genes (39). We hypothesized that PP was binding to and stabilizing mitochondrial G-quadruplexes, thereby disrupting mitochondrial transcription. We found that PP weakly bound negative controls and had an increased binding for a known MYC G-quadruplex and for multiple known and imputed mitochondrial G-quadruplex target sequences (Supplementary Fig. S5D; refs. 28, 30). We confirmed these results using stop-PCR analysis, where PP inhibited two mitochondrial targets with imputed G-quadruplexes (39) at concentrations that did not inhibit targets that lacked G-quadruplexes, similar to the known G-quadruplex inhibitor pyridostatin (Fig. 4C; ref. 40).

We assessed published nuclear G-quadruplexes, and while PP inhibited mitochondrial targets more than rotenone or oligomycin, it did not appear to have similar inhibitory effects on known nuclear G-quadruplex targets (Fig. 4D; ref. 41). Overall, we determined that PP bound to mitochondrial G-quadruplexes and led to a decrease in mitochondrial RNA and protein expression.

In a xenograft model of MIA-PaCa2 tumors, PP treatment three times weekly at 1 mg/kg IP caused a 3.4-fold reduction in tumor growth while not significantly affecting mouse weights (Fig. 5A; P < 0.0001; Supplementary Fig. S5E). Comparing tumors from the vehicle control group with the PP-treated group, we found there was a significant increase in TUNEL staining (24.7-fold; P < 0.0001, Fig. 5B; Supplementary Fig. S5F). Furthermore, PP treatment led to a 1.5-fold increase in median survival (Fig. 5C; P = 0.0036). We treated mice with human-grade PP (Pyrvin 100 mg tablets, Orion Pharma) at 5, 20, and 35 mg/kg. Daily oral treatment of PP led to significant and dose-dependent reductions in tumor growth, although 35 mg/kg daily treatment caused a slight but significant decrease in average mouse weight gain (4.6%, P = 0.0174; Fig. 5D; Supplementary Fig. S5G).

We performed pharmacokinetic profiling of PP to determine its bioavailability. We initially dosed animals IV at 1 mg/kg and determined plasma half-life was 0.54 hours (Supplementary Fig. S6A). We evaluated PP accumulation in tissues by treating mice with single doses of PP at 1 mg/kg IP, or with PO dosing of 5, 20, or 35 mg/kg PP and collected plasma, muscle, fat, and pancreas tissues (Supplementary Table S2A and S2B). We determined across all groups that whereas there was little to no detectable serum accumulation, PP appeared to accumulate at the pancreas and in fat tissue (Supplementary Fig. S6B–S6E). Interestingly, we observed significant accumulation of PP at target organ sites even at concentrations where PP was undetectable in the plasma (Supplementary Fig. S6F and S6G). This may explain the discrepancy where PP has previously been observed to have low oral bioavailability in plasma (42), but has antitumor effects when dosed orally (43, 44).

In vivo plasma metabolomics results from mice treated with PP at 1 mg/kg largely corroborated the hypothesis that PP affected mitochondria systemically in vivo. In particular, we observed an increase in plasma levels of N-carbamoyl aspartate along with a concomitant decrease in orotate, and a trend toward a decrease in all unmodified acylcarnitines tested (Supplementary Fig. S6H and S6I). These data were highly similar to the metabolomic results we observed in vitro, indicating similar systemic alteration in mitochondrial pathways including de novo pyrimidine biosynthesis and fatty acid metabolism.

Beyond PDAC, we explored whether PP was effective in other cancers utilizing publicly available data from the NCI60. We observed that PP appeared to be particularly effective in colon, leukemia, and lung cell lines, with renal cell lines displaying the least sensitivity (Supplementary Fig. S6J). PP efficacy correlated with the expression of mitochondrial related genes MDH2, NDUFB11, HARS2, and inversely correlated with the drug transporter ABCB1 (Supplementary Fig. S6K). These data support the hypothesis that PP works through targeting the mitochondria and may be resisted through the upregulation of drug–efflux pathways.

In this study, we demonstrated that PP is effective for inhibiting PDAC cell growth and proliferation in vitro and in vivo (Fig. 5). We determined through phosphoproteomic, metabolomic, and RNA-seq approaches (Figs. 1 and 3; Supplementary Fig. S2) that PP was acting predominantly through altering cellular metabolism, and validated that PP inhibited mitochondrial oxidative phosphorylation (Fig. 2).

Mitochondrial depleted cells are resistant to PP, confirming PP's action through targeting mitochondria (13); however, how PP inhibits mitochondria is less well understood. Previous work proposed that PP inhibits fumarate reductase, a critical member in an alternative mitochondrial pathway proposed to function in cancer cells in hypoxic conditions whereby mitochondria can generate ATP in an oxygen-independent fashion (12). As PP effectively inhibits cancer cell mitochondria in nonhypoxic settings, specific targeting of this pathway is unlikely. PP has also previously been shown to inhibit complex I in the mitochondria (13). However, it has previously been demonstrated that specifically inhibiting complex I causes an increase in beta oxidation as indicated by slight increases in free carnitine and long-chain acyl carnitines (45). We demonstrated that PP causes a larger increase in long-chain acyl carnitines along with increases in Acylcarnitine C4-OH and Acylcarnitine 5-DC, and decreases in short-chain acylcarnitines and free carnitine (Fig. 1). Taken together, this is indicative of an inhibition of beta oxidation, rather than the increase of beta oxidation that would be expected if PP was solely inhibiting complex I (35, 45). Similarly, the increase in N-carbamoyl aspartate and decrease in orotate observed with PP treatment (Fig. 1; Supplementary Fig. S6) suggest an inhibition of DHODH, which can be caused through the specific inhibition of complex III but not complex I (46). The inhibition of DHODH may further augment PP's efficacy in PDAC as KRAS-driven tumors such as PDAC are uniquely sensitive to DHODH inhibition (47). Furthermore, PP treatment caused transcriptomic changes that were largely distinct from those caused by the complex I inhibitor rotenone (Fig. 3). Overall, our data indicated that PP was not acting through the specific inhibition of complex I, or fumarate reductase, but rather a more global inhibition of mitochondrial function.

We further identified a novel mechanism whereby PP caused ablation of mitochondrial RNA expression (Fig. 3) followed by downregulation of mitochondrial protein expression (Fig. 4). Our results indicate that this ablation of mitochondrial RNA is through PP binding mitochondrial DNA at regions of mitochondrial G-quadruplexes, and thus inhibiting transcription (Fig. 4). This is supported by previous work, which demonstrated that the G-quadruplex stabilizer RHPS4 bound mitochondrial G-quadruplexes and caused a depletion in mitochondrial RNA (39). This novel mechanism underlies the multifaceted mitochondrial effects PP treatment causes through affecting multiple mitochondrial pathways simultaneously.

We demonstrated that PP is more effective at inhibiting ATP levels and cell viability in PDAC cells as compared with noncancerous cells. Beyond this increased sensitivity, the hypoglycemic TME characteristic of PDAC may further sensitize PDAC cancers to PP in comparison with normal tissues (4). Importantly for a predominantly metastatic disease, cancer stem cells which are critical for metastatic spread are critically reliant on oxidative phosphorylation (48). This reliance of primary PDAC tumors and metastases on mitochondrial pathways may be why targeting the mitochondria in PDAC models has shown such promising results (7–9).

Of note, previous work performed in the 1970s indicated that there was poor systemic bioavailability of PP after oral treatment (42). We have demonstrated in this work that PP has therapeutic efficacy for PDAC when dosed orally in mice (Fig. 5D), and this is supported by other studies where oral administration of PP has demonstrated significant systemic efficacy (43, 44). In fact, the doses used in this study span the standard daily dosing of PP in humans (5 mg/kg) to the maximum dose utilized in humans (35 mg/kg; ref. 49). Indeed, 35 mg/kg does not represent a true maximum tolerated dose as higher doses have not been tested, and rhesus monkeys have been treated up to 80 mg/kg daily of PP (50). In preclinical characterization studies of PP, we demonstrated that PP accumulated at target tissues after oral administration (Supplementary Fig. S6) while still being undetectable in plasma. This may explain why PP bioavailability appears to be poor based on drug accumulation in the plasma, yet PP demonstrates in vivo efficacy against tumors. We believe this work supports repurposing PP for the treatment of PDAC. Based on this work, we have received IND approval for the treatment of PDAC patients with PP (IND 144822) and have recently received IRB approval to further explore PP as an anti-PDAC agent in human patients.

M.J. Pishvaian reports personal fees from Merck, Astra Zeneca, Caris, Sirtex, Perthera, RenovoRx, grants from Seattle Genetics, Pfizer, and Halozyme, grants from Genentech, Celldex, Novartis, AbbVie, Tesaro, Bavarian Nordic, Medimmune, BMS, Regeneron, Fibrogen, ARMO, Bayer, Calithera, Curegenix, and Karyopharm outside the submitted work. T. Golan reports grants and personal fees from AstraZeneca and MSD, personal fees from AbbVie, Bayer, and Teva outside the submitted work. E.F. Petricoin reports personal fees and other support from Perthera, Inc., Theralink Technologies, Inc., and Ceres Nanosciences outside the submitted work. J.M. Salvino reports grants from NIH during the conduct of the study; grants from Alliance Discovery, Inc. outside the submitted work. P.K. Singh reports grants from NIH/NCI during the conduct of the study. No disclosures were reported by the other authors.

C.W. Schultz: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. G.A. McCarthy: Conceptualization, formal analysis, investigation, writing–review and editing. T. Nerwal: Formal analysis, investigation, methodology, writing–review and editing. A. Nevler: Conceptualization, validation, investigation, writing–review and editing. J.B. DuHadaway: Investigation. M.D. McCoy: Data curation, software. W. Jiang: Investigation. S.Z. Brown: Writing–review and editing. A. Goetz: Investigation. A. Jain: Investigation, methodology. V.S. Calvert: Investigation. V. Vishwakarma: Data curation, investigation. D. Wang: Data curation, investigation. R. Preet: Investigation. J. Cassel: Investigation. R. Summer: Resources, investigation. H. Shaghaghi: Investigation, methodology. Y. Pommier: Resources. S.A. Baechler: Investigation. M.J. Pishvaian: Visualization. T. Golan: Resources. C.J. Yeo: Resources. E.F. Petricoin: Resources. G.C. Prendergast: Resources. J. Salvino: Resources. P.K. Singh: Resources, writing–review and editing. D.A. Dixon: Resources, writing–review and editing. J.R. Brody: Conceptualization, resources, data curation, supervision, funding acquisition, methodology, project administration, writing–review and editing.

R01CA212600-01 (J.R. Brody), R01CA243445 (D.A. Dixon), R01 CA163649 (P.K. Singh), R01CA216853 (P.K. Singh), P01CA217798 (P.K. Singh), P50CA127297 (P.K. Singh), NCI–CCR Intramural support Z01 006150 (Y. Pommier), T32 training grant NIH/NIGMS T32GM008562 (T. Dhir), the Mary Halinski Pancreatic Cancer Research Fund (J.R. Brody and A. Nevler). Dr. Brody was supported in part by the NCI of the NIH Award Number P30CA056036 SKCC Core Grant (Thomas Jefferson University); OHSU Knight Cancer Institute P30 CA069533 (Druker); in part, by U01CA224012 (NCI, NIH).

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.