Mitochondria-Targeted Analogues of Metformin Exhibit Enhanced Antiproliferative and Radiosensitizing Effects in Pancreatic Cancer Cells

Metformin (Met) is a synthetic analogue of a naturally occurring biguanides. It is an antidiabetic drug (1, 2) that exerts anticancer effects in diabetic individuals with pancreatic cancer (3, 4). Met exists as a hydrophilic cation (Fig. 1A) at physiologic pH and weakly targets mitochondria (5). A prevailing view is that Met exerts antitumor effects by elevating cellular AMP/ATP ratio and activating the 5′-AMP–activated protein kinase (AMPK)/mTOR pathway (6, 7) and/or by decreasing circulating insulin and blood glucose levels (8). Met also inhibits complex I in the mitochondrial electron transport chain and tumor mitochondrial respiration (9). Met suppresses gluconeogenesis by inhibiting mitochondrial glycerophosphate dehydrogenase (10). The organic cation transporter is responsible for Met uptake into tumor cells, while more lipophilic Met analogues (e.g., phenformin) are taken up into cells via alternate mechanism(s) (11). Phenformin is more potent than Met in inhibiting pancreatic tumor cell proliferation (12). However, phenformin was taken off the market in the United States because of increased incidence of acidosis during antidiabetic therapy (13). Additional clinical research repurposing phenformin as an antitumor drug was recently recommended (14).

Previous reports suggest that mitochondria-targeted cationic agents induce antiproliferative and cytotoxic effects in tumor cells without markedly affecting normal cells (15, 16). For example, conjugating a nitroxide, quinone, a chromanol moiety of α-tocopherol to the triphenylphosphonium (TPP⁺) group via an aliphatic linker increased their antiproliferative effect in tumor cells (15, 16). Selective toxicity to tumor cells as compared with normal cells was attributed to enhanced uptake and retention of TPP⁺-containing compounds in tumor cell mitochondria (16). Met has been used in the clinic for over 50 years and has a very good safety profile (diabetic patients tolerate daily doses of 2–3 g; refs. 1–4). Efforts to improve and enhance efficacy of Met involved modification of structure by attaching alkyl or aromatic groups (e.g., butformin, phenformin; ref. 17; Fig. 1A). We hypothesized that improved mitochondrial targeting of Met by attaching a positively charged lipophilic substituent would result in a new class of mitochondria-targeted drugs with significantly increased antitumor potential. To this end, we synthesized and characterized several Met analogues (e.g., Mito-Met₂, Mito-Met₆, Mito-Met₁₀, Mito-Met₁₂; Fig. 1A) conjugated to an alkyl substituent containing a TPP⁺ moiety (Supplementary Fig. S1). The current results show that Mito-Met₁₀ is nearly 1,000-fold more effective than Met in inhibiting pancreatic ductal adenoma cell (PDAC) proliferation in vitro and more effective than Met in abrogating PDAC tumor growth in vivo. Mito-Met₁₀ inhibited mitochondrial complex I, stimulating superoxide generation and AMPK activation, more potently than Met in MiaPaCa-2 cells. Reports suggest that Met (1 mmol/L) pretreatment followed by radiation resulted in enhanced cancer cell killing (17–19). In this study, we show that Mito-Met₁₀ enhanced radiation sensitivity in PDAC to the same extent as did Met (1 mmol/L) but at a 1,000-fold lower concentration (1 μmol/L). The finding that relatively nontoxic mitochondria-targeted Met analogues alone or in combination with radiotherapy could inhibit pancreatic cancer cell proliferation is highly relevant.

MiaPaCa-2, PANC-1, HPNE, IEC-6, and MCF-10A cell lines were obtained from the ATCC, where they were regularly authenticated. N27 cell line was a gift from Dr. Anumantha Kanthasamy (Iowa State University, Ames, IA). FC-1242 cell line was a gift of Dr. David Tuveson and Dr. Dannielle Engle (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY), and was derived from C57BL/6(B6) KPC transgenic mice that spontaneously developed pancreatic tumors. This cell line was subsequently engineered to express luciferase (FC-1242-luc) to enable monitoring of in vivo tumor growth (20). All cells were obtained over the last five years, stored in liquid nitrogen, and used within 20 passages after thawing.

The mitochondrial function in intact and permeabilized cells was measured using a Seahorse XF96 Extracellular Flux Analyzer (Seahorse Bioscience). Assays in intact cells were performed as described previously (21). Measurement of the activity of mitochondrial respiratory complexes in permeabilized cells was performed according to the manufacturer's instructions. Briefly, intact cells were permeabilized using 1 nmol/L Plasma Membrane Permeabilizer (PMP, Seahorse Bioscience) immediately before oxygen consumption rate (OCR) measurement by XF96. The oxygen consumption derived from mitochondrial complex I or complex II activity was measured by providing different substrates to mitochondria, for example, pyruvate/malate for complex I and succinate for complex II (22, 23). Rotenone, malonate, and antimycin A were used as specific inhibitors of mitochondrial complex I, II, and III, respectively.

Cells were seeded as indicated in 6-well plates and treated with Mito-Met₁₀ or Met for 24 hours. The plates were kept within the incubator and media changed every 3 to 4 days until the control cells formed sufficiently large clones. The cell survival fractions were calculated as before (21).

Cell proliferation was measured using a label-free, noninvasive cellular confluence assay by IncuCyte Live-Cell Imaging Systems (IncuCyte FLR, Essen Bioscience), as described previously (21).

MiaPaCa-2 cells (5 × 10³ cells/well) were seeded in 96-well plates containing Matrigel (Corning). The culture medium (containing appropriate concentration of Met or Mito-Met₁₀) was replaced every two days. At days 3, 7, and 14, the images were acquired using a Nikon Eclipse Ti inverted microscope (Nikon Inc.). Spheroid-forming cells were counted using the Nikon NIS Elements imaging software.

To determine the cytotoxicity of Mito-Met analogues and other TPP⁺-conjugated compounds, we used the Sytox Green–based assay as described previously (16). MiaPaCa-2 cells were treated for 24 hours, and dead cells were monitored in real time in the presence of 200 nmol/L Sytox Green (Invitrogen) under an atmosphere of 5% CO₂:95% air at 37°C. Data are represented as a percentage of dead cells after normalization to total cell number (measured with Sytox Green after a 3-hour treatment with 0.065% Triton X-100) for each group.

Cells were serum-starved for 5 hours (AMPK) or 24 hours (FOXM1). Stimulations with Met or Mito-Met₁₀ were performed for 30 minutes in serum-free medium (AMPK) or for 24 hours in full growth media (FOXM1). After stimulation, cells were washed, lysed using a modified RIPA buffer, and the proteins immunoblotted. Antibodies against FOXM1 (D12D5), cyclin D1 (92G2), phospho-AMPK(40H9), or total AMPK (D5A2) anti-sera were from Cell Signaling Technology. Proteins were detected by chemiluminescence and quantified by densitometric analysis using the FluorChem HD2 software (Cell Biosciences).

PANC1 cells (5 × 10⁵) were seeded into 35-mm dishes and immediately treated with 10 mmol/L Met or 10 μmol/L Mito-Met₁₀ in complete media. At 48 hours after treatment, cells were fixed with 70% ethanol, permeabilized with 0.25% Triton X-100 in PBS, and stained using a propidium iodide/RNase solution (20 μg/mL and 10 μg/mL, respectively). Data were collected on the BD-LSR II Flow Cytometer (BD Biosciences) and analyzed using Flow-Jo (Flow-Jo LLC).

Cells were grown on 10-cm dishes and incubated with the compounds for 24 hours in full media. The protocol for extraction of Mito-Met analogues was the same as described previously, but without butylated hydroxytoluene (16). LC-MS/MS analyses were performed using a Kinetex Phenyl-Hexyl column (50 mm × 2.1 mm, 1.7 μm, Phenomenex) equilibrated with water:acetonitrile mixture (4:1) containing 0.1% formic acid. Compounds were eluted by increasing the content of acetonitrile from 20% to 100% over 4 minutes and detected using the MRM mode.

MiaPaCa-2, HPNE, and MCF-10A cells (5 × 10⁵/dish) were seeded in 6-cm dishes and treated with Mito-Met₁₀ or Met for 24 hours. The control cells were treated with the vehicle (0.1% DMSO), where appropriate. The cells were then exposed to X-radiation (0–6 Gy). After irradiation, cells were suspended and seeded at various densities (100–8,000 cells/well) in 6-well plates for clonogenic assay as described above. The plates were kept within the incubator and media changed every 4–6 days for 2 weeks. Wells with 10 to 50 sufficiently large clones were chosen for calculating the cell survival fractions.

An orthotopic syngeneic engraftment model was used to assess metastatic homing and tumor progression after treatment with Met or Mito-Met₁₀. Six- to eight-week-old C57BL/6 mice were anesthetized with isoflurane and were injected with 1 × 10⁶ luciferase-expressing FC-1242 (FC-1242-luc) cells (24). To generate FC-1242-luc cells, the firefly luciferase gene was cloned into the mammalian expression vector pcDNA3.1/hygro(-) cells transfected using the Lipofectamine 2000 transfection reagent (Life Technologies). A cell line stably expressing firefly luciferase was generated by culturing cells in growth medium supplemented with 0.5 mg/mL hygromycin and limited dilution cloning. A stable FC-1242-luc clone was expanded to generate a frozen stock. Cells were pretreated for 48 hours with Met (1 mmol/L) or Mito-Met₁₀ (0.5 μmol/L) and then orthotopically engrafted to the pancreas as described previously (24–26). Starting the day of implantation, mice were treated daily with 1 mg/kg Met or Mito-Met₁₀ administered via an intraperitoneal injection in a 200-μL volume. Tumor growth and metastasis was monitored using bioluminescence imaging (Lumina IVIS 100, Perkin Elmer) on days 1, 7, and 13 (26, 27). After 13 days, mice were euthanized, and the size of primary tumor determined using calipers. Metastasis to the liver, mesenteric lymph nodes, spleen, and lung was assessed ex vivo using bioluminescence imaging.

All statistical analyses were performed using GraphPad Prism 4. Paired analyses were performed using a Student t test. Multiple comparisons were analyzed using a one-way ANOVA and Tukey post hoc tests to identify pair-wise differences between distinct experimental groups. All values provided represent mean ± SD.

The Mito-Met analogues (Fig. 1A) were synthesized according to the reaction shown in Supplementary Fig. S1E and characterized by nuclear magnetic resonance and mass spectrometry (Supplementary Fig. S1). Mito-Mets (n = 2, 6, 10, and 12) were obtained by reacting the corresponding aminoalkyltriphenylphosphonium substrate with dicyanamide and purified by preparative-HPLC. Experimental details are provided in Supplementary Fig. S1.

To compare the antiproliferative effects of Met and Mito-Met₁₀, human MiaPaCa-2 cells were treated with Met (0.1–2 mmol/L) or Mito-Met₁₀ (0.1–1 μmol/L). Cell proliferation curves, as measured by cell confluence kinetics and representative phase contrast images, are shown in Fig. 1B. Under the conditions used, untreated cells reached 100% confluence in 5 to 6 days. Mito-Met₁₀ treatment (0.2 and 1 μmol/L) inhibited cell growth by 70% and >90%, respectively. In contrast, Met inhibited cell growth by 80% at 1,000-fold higher concentrations (1 mmol/L; Fig. 1B). A similar trend was noticed at 20% and 1% oxygen (Fig. 1B). Mito-Met₁₀ was also significantly more potent than Met in inhibiting proliferation of mouse pancreatic cancer cells, FC-1242 (Supplementary Fig. S2).

We also monitored colony formation in MiaPaCa-2 cells after a 24-hour treatment with different concentrations of Met or Mito-Met₁₀ under similar conditions. Almost no colony formation was detected in cells treated with 3 μmol/L of Mito-Met₁₀ and 3 mmol/L of Met (Fig. 1C, left). The survival fraction analysis (Fig. 1C, right) showed that the IC₅₀ values determined for Met and Mito-Met₁₀ were 1.3 mmol/L and 1.1 μmol/L, respectively. Thus, results from two independent cell growth assays revealed that Mito-Met₁₀ is nearly 1,000-fold more potent than Met in inhibiting MiaPaCa-2 cell proliferation.

To compare the cytotoxic effects of Met, Mito-Met₁₀, and selected other TPP⁺-linked compounds, MiaPaCa-2 cells were treated with Mito-Met₁₀ (100 μmol/L) or Met (30–100 mmol/L) and other TPP⁺-linked compounds for 24 hours and cell death was monitored in real time using the Sytox Green assay (Supplementary Fig. S3). Surprisingly, there was no detectable cell death induced by Mito-Met₁₀ even at 100-fold higher concentrations (100 μmol/L) than needed to completely block cell proliferation. In contrast, other mitochondria-targeted agents (e.g., Mito-CP, Mito-Q, Mito-CP-Ac, Mito-Tempol, and Mito-Chromanol) were considerably more cytotoxic to MiaPaCa-2 cells under these conditions. Intracellular ATP levels measured under these conditions for selected compounds are consistent with the results obtained using the Sytox Green assay.

We also tested the antiproliferative effects of Mito-Met₁₀ and Met in a multicellular tumor spheroid model. As shown in Fig. 2, Mito-Met₁₀ (0.5 μmol/L) more potently inhibited PDAC spheroid growth than Met (10 mmol/L). Thus, both two- and three-dimensional (2D and 3D) cell growth assays indicate that Mito-Met₁₀ is 1,000- to 10,000-fold more effective than Met in inhibiting PDAC proliferation.

We compared the relative antiproliferative potencies of other Mito-Met analogues (Mito-Met₂, Mito-Met₆) with Mito-Met₁₀, phenformin, and Met in normal and cancer cells. Figure 3A compares the antiproliferative effects of Mito-Met₁₀ (left) and Met (right) in MiaPaCa-2 and PANC-1 and in nonmalignant control cells, HPNE, IEC-6, and N27 cells. Mito-Met₁₀ and Met more potently inhibited proliferation of MiaPaCa-2 and PANC-1 cells as compared with normal cells (HPNE, IEC-6, and N27; Fig. 3A). Furthermore, we performed detailed cell viability analyses in Mito-Met₁₀ and Dec-TPP⁺ (same length of the aliphatic linker as Mito-Met₁₀) treated MiaPaCa-2 and PANC-1 cells and in nonmalignant HPNE and IEC-6 control cells. Over a wide range of concentrations and incubation times, Dec-TPP⁺ did not show any selectivity with regard to inhibition of cell viability in PDAC versus nonmalignant cells. However, Met conjugated to TPP⁺ (e.g., Mito-Met₁₀) more potently decreased pancreatic cancer cell viability as compared with normal cell viability, with significantly lower IC₅₀ values in MiaPaCa-2 and PANC-1 cells as compared with HPNE and IEC-6 cells (Supplementary Fig. S4). Figure 3B shows the relative potencies of Mito-Met analogues (Mito-Met₂, Mito-Met₆, and Mito-Met₁₀, phenformin, and Met) in inhibiting MiaPaCa-2 colony formation. Mito-Met₂ and Mito-Met₆ were relatively less potent than Mito-Met₁₀ in inhibiting MiaPaCa-2 cell proliferation (Fig. 3B and Supplementary Fig. S5A and S5B). Next, we investigated the relative uptake of different Mito-Met analogues (Mito-Met₂, Mito-Met₆, and Mito-Met₁₀) and phenformin in MiaPaCa-2 cells by LC-MS (Fig. 3C). Cells were treated with respective Met analogue (1 μmol/L) for 1 hour. As shown in Fig. 3C, there was a dramatic increase in Mito-Met cellular uptake as a function of increasing carbon–carbon side-chain length and Mito-Met₁₀ was taken up nearly 100-fold more than phenformin. Under these treatment conditions, Met uptake was considerably lower than phenformin.

The oxygen consumption rates (OCR) were measured as a readout of mitochondrial function (28). We compared the immediate OCR changes in MiaPaCa-2 cells in response to different concentrations (e.g., at IC₅₀ or higher values determined by clonogenic assay) of Met, Mito-Mets, and phenformin (Fig. 4A). Met dose-dependently decreased the OCR at relatively higher concentrations (≈1–10 mmol/L). In contrast, the more lipophilic phenformin inhibited OCR to the same extent (as did Met) at more than 10-fold lower concentration (Fig. 4A). Mito-Met₁₀ inhibited proliferation most potently at submicromolar concentration, but did not inhibit OCR at 1 μmol/L concentration. However, there was a slight inhibition at a higher concentration of Mito-Met₁₀ (10 μmol/L). Similar results were observed at higher concentrations of Mito-Met₆ and no inhibition was noted with Mito-Met₂ (Fig. 4A).

Next, we monitored the mitochondrial respiration rates in MiaPaCa-2 cells after longer treatment (24 hours) with Met, Mito-Mets, and phenformin followed by a washout and replenishment with fresh assay media. A significant decrease in OCR was observed after a 24-hour treatment with all compounds tested (Fig. 4B). These results demonstrate that the extent of OCR inhibition was dependent on the alkyl chain length with Mito-Met₂, Mito-Met₆, and Mito-Met₁₀ showing similar effects at the concentrations of 25, 16, and 1 μmol/L, respectively (Fig. 4B).

Both PDAC and control nonmalignant cells pretreated for 24 hours with Met (1 mmol/L) or Mito-Met (1 μmol/L) were subjected to X-irradiation, followed by a clonogenic assay. As shown in Fig. 5A, Met pretreatment decreased MiaPaCa-2 cell growth after irradiation, in agreement with previous reports indicating radiosensitizing effects of Met in PDAC (29, 30). Radiation dose–response results show that 1 μmol/L Mito-Met was as effective as 1 mmol/L Met in radiosensitization (Fig. 5A), whereas no radiosensitization was observed in nonmalignant cells (MCF-10A or HPNE) pretreated with Mito-Met₁₀ (Fig. 5A). Enhanced radiosensitivity in MiaPaCa-2 cells could be attributed to an increased uptake and retention of Mito-Met₁₀ in these cells, as compared with normal (MCF-10A) cells (Supplementary Fig. S6). Detailed cell-cycle analysis showed that treatment with Mito-Met₁₀, but not Met, arrested PANC-1 cell growth in the G₁ phase with a concomitant decrease in cyclin D1 levels (Fig. 5B).

Next, we investigated whether treatment with Mito-Met₁₀ or Met would have an effect on the FOXM1 pathway. FOXM1 is a redox-responsive transcription factor that we have previously shown is downregulated in malignant mesothelioma cells upon treatment with mitochondria-targeted nitroxides (31). Treatment of MiaPaCa-2 cells with Mito-Met₁₀, but not Met, resulted in a significant decrease in FOXM1 levels (Fig. 5C, left). Similar results were observed for PANC-1 cells (Supplementary Fig. S7). To better understand potential upstream signaling events, we investigated the AMPK–mTOR energy signaling pathway. The extent of AMPK activation, measured as a ratio of phosphorylated to total protein, showed an approximately 2-fold increase in active AMPK in human (Fig. 5C, middle) and an approximately 4-fold increase in pAMPK in murine FC-1242 PDAC cells treated with Mito-Met₁₀ and Met (Fig. 5C, right). The increase in AMPK activation occurs at a 1,000-fold lower concentration of Mito-Met₁₀ than Met. Treatment with dorsomorphin (Compound C), a potent, reversible AMPK inhibitor, counteracted the antiproliferative effects of Mito-Met₁₀ (Fig. 5D), suggesting a role for AMPK signaling mechanism.

To better understand the role of mitochondrial respiratory complex in Mito-Met and Met-mediated signaling pathways, we used permeabilized and intact cells. The use of permeabilized cells avoids differences in cellular uptake of compounds. Supplementary Figure S8 (left) shows OCR changes in control permeabilized cells and in rotenone- or malonate-treated cells. Rotenone (complex I inhibitor) greatly diminished OCR that was restored by adding succinate. However, in the presence of malonate (complex II inhibitor), addition of succinate did not stimulate OCR. Antimycin A decreased both pyruvate- and succinate-induced OCR. These studies established the use of permeabilized cells in bioenergetics function assay (22, 23). Next, we used this model to probe the effects of Met and Mito-Met analogues. The IC₅₀ values for Met and Mito-Met₁₀ to inhibit complex I-mediated oxygen consumption upon injection to permeabilized cells were determined to be 0.8 mmol/L and 2 μmol/L, respectively (Supplementary Fig. S9). Next, we tested the effects of a 24-hour pretreatment of intact cells with Mito-Met₁₀ or Met on complex I activity. A significant inhibition of complex I-dependent OCR was observed (Fig. 6A and Supplementary Fig. S8). The IC₅₀ values for complex I inhibition in MiaPaCa-2 cells determined for Met and Mito-Met₁₀ were 1.1 mmol/L and 0.4 μmol/L, respectively. This increased potency of Mito-Met₁₀ to inhibit complex I activity is consistent with the >1,000-fold enhanced antiproliferative effect of Mito-Met₁₀ versus Met. In contrast to pancreatic cancer cells, higher concentrations of Mito-Met₁₀ were required to inhibit complex I in HPNE and IEC6 cells (Fig. 6A and Supplementary Fig. S8). These results demonstrate the selectivity of Mito-Met₁₀ in inhibiting mitochondrial respiration in PDACs and suggest that the enhanced antiproliferative effects of Mito-Met₁₀ in PDAC cells are related to their enhanced ability to inhibit complex I activity.

One of the consequences of complex I inhibition is enhanced generation of cellular oxidizing species (32). We used the cell-permeable probe hydroethidine (HE) to detect reactive oxygen species. HPLC traces (Fig. 6B, left) and quantitative analyses of products are shown (Fig. 6B, middle). Results indicate that Mito-Met₁₀ treatment of MiaPaCa-2 cells increased the formation of 2-hydroxyethidium (2-OH-E⁺), a diagnostic marker product of HE/superoxide reaction (Fig. 6B, right). In addition, a marked increase in diethidium (E⁺-E⁺) formation was observed (Fig. 6B, middle), indicating that Mito-Met₁₀ also induced generation of one-electron oxidants. Under these conditions, Mito-Met₁₀ did not stimulate O₂^•− formation in control HPNE cells (Fig. 6B, middle), consistent with its lack of inhibition of mitochondrial complex I at the concentration tested. However, at the concentrations equal or higher than IC₅₀ (Fig. 3A), Mito-Met₁₀ treatment increased the formation of 2-hydroxyethidium (2-OH-E⁺) in HPNE cells (Supplementary Fig. S10).

A proposed model for Mito-Met₁₀–induced antiproliferative effect in PDAC via inhibition of mitochondrial complex I activity, enhanced O₂^•− generation, and AMPK activation is shown in Fig. 6C.

In vivo data show that Mito-Met suppresses tumor growth in a preclinical mouse model (Fig. 6D). Syngeneic KPC cells expressing firefly luciferase (FC-1242-luc) were orthotopically implanted in the pancreas of C57BL/6 mice and tumor growth monitored using bioluminescence imaging (Fig. 6D). Mice orthotopically autografted with FC-1242-luc cells (25) were treated daily with Met (1 mg/kg) or Mito-Met₁₀ (1 mg/kg). Consistent with the cell culture data, Mito-Met₁₀ was considerably more effective than Met in mitigating PDAC growth (Fig. 1B) and the treatment resulted in markedly smaller primary tumors at the completion of the experiment (Fig. 6D). Weekly tracking measurements of total radiance in FC-1242-luc–engrafted mice (Fig. 6D, dotted lines) revealed a significant decrease in percent change of tumor burden throughout the course of the study. Over time, PDAC tumor–bearing mice treated with Mito-Met₁₀ had lower tumor burden when assessed at early, middle, and later time points as visualized by bioluminescent imaging, with smaller tumors at the completion of the study (Fig. 6D). Serum from these animals was collected and hepatic and kidney toxicity tested using standard AST, ALT, AP, and BUN assays, respectively. As predicted from cell culture data (Fig. 3), neither Met nor Mito-Met₁₀ elicited toxicity in vivo (Supplementary Table S1). After administration of Mito-Met₁₀ for two weeks in FC-1242-luc orthotopic mice, we detected an increased accumulation of this compound in liver, kidney, spleen, and tumor tissues (Supplementary Fig. S11). Collectively, results from the in vivo experiments indicate a potent antitumor activity of Mito-Met₁₀, with negligible off-target toxicity.

Human pancreatic ductal adenocarcinoma (PDAC) is the most severe and aggressive form of pancreatic cancer with limited chemo- and radiotherapeutic options to improve survival. Currently available standard-of-care chemotherapy offers limited survival benefit. There is critical unmet need for new therapeutic approaches to mitigate therapeutic resistance mechanisms and maximize multimodal treatment approaches in pancreatic cancer. Here, we have developed new mitochondria-targeted metformin analogues that alone and in combination with radiotherapy markedly inhibited PDAC proliferation. The enhanced potency of Mito-Met₁₀ is attributed to mitochondrial ROS-dependent activation of signal transduction pathway (involving AMPK activation) in PDAC cells. These mitochondria-targeted Met analogues may have significant clinical and translational potential in PDAC treatment.

Delocalized lipophilic cations inhibit tumor cell proliferation through selective accumulation into mitochondria and inhibition of mitochondrial respiration (15, 16, 32). The mitochondrial membrane potential is much higher (more negative inside) in tumor than in normal (nontransformed) cells (33). The cationic compounds tethered to an alkyl chain accumulate preferentially in tumor mitochondria depending upon the alkyl side-chain length (15, 16). TPP⁺-linked agents conjugated to aromatic and heterocyclic groups (Mito-Chromanol, Mito-CP) also exerted selective cytostatic and cytotoxic effects in various tumor cells (15, 16, 34). Met exerts biologic activity through alterations of cellular bioenergetics without undergoing any detectable metabolism itself (1). Selective targeting of cancer cell mitochondrial bioenergetics is an emerging chemotherapeutic strategy (5, 35). We, therefore, surmised that enhancing its mitochondrial uptake would greatly increase Met's biologic activity. Although several lipophilic variants of Met were synthesized and shown to exert increased antitumor potency, none of these modifications included mitochondrial targeting cationic function. In this study, we showed and characterized, for the first time, that fine-tuning of Met structure by attaching a TPP⁺ group tethered to different alkyl chain lengths is synthetically feasible, and that these modified analogues increasingly target tumor mitochondria. Consistent with enhanced intracellular uptake, Mito-Met analogues were more potent than Met in their ability to inhibit PDAC proliferation. The antiproliferative potency of Mito-Met analogues increased with increasing length of the alkyl linker (Mito-Met₁₀ > Mito-Met₆ > Mito-Met₂; Fig. 3B and Supplementary Fig. S5). Elongation of the aliphatic linker to twelve carbon atoms (Mito-Met₁₂) had no further effect on its antiproliferative efficacy, as compared with Mito-Met₁₀ (not shown). A major reason for the selective antiproliferative effect of Mito-Met analogues in tumor cells is due to their preferential accumulation and inhibition of mitochondrial complex I activity in tumor cells as compared with normal, nonmalignant cells.

At present, the mechanism(s) responsible for the enhanced antiproliferative and radiosensitizing effects of Mito-Mets in cancer cells remain unknown. It is likely that Mito-Met₁₀ exerts antiproliferative effects in PDACs via targeting the energy-sensing bioenergetics pathway(s). Mito-Met₁₀ activated AMPK in MiaPaCa-2 cells nearly 1,000-fold more potently than did Met (Fig. 5). AMPK, a master regulator of cellular energy homeostasis, is typically activated by enhanced intracellular AMP (36). Under conditions wherein intracellular ATP levels are decreased along with a concomitant increase in AMP (enhanced AMP/ATP ratio), AMPK is activated via phosphorylation of its threonine-172 residue (37). Previous research has shown that AMPK represses the FOXM1 transcription factor expression through inhibition of the AKT/FOXO3 signaling cascade, leading to regression of cervical cancer cell growth (31). Thus, it is conceivable that Mito-Met₁₀ and related analogues act on the AKT/FOXO3/FOXM1 signaling pathway.

Recently, it was reported that ROS activate two proteins, AMP-activated kinase and hypoxia-inducible factor 1 in C. elegans and that balancing ROS at optimal levels was crucial for their health and longevity (38). Factors responsible for activation of AMPK still need to be ascertained, although mitochondrial ROS was implicated (38). We propose that Mito-Met₁₀–mediated inhibition of complex I and superoxide generation, accompanied by activation of AMPK, play an important role in inhibiting PDAC proliferation (Fig. 6C).

As reported previously (29), pretreatment with Met increased radiosensitivity of MiaPaCa-2 cells. The current data demonstrate that Mito-Met₁₀ is significantly more effective than Met in PDAC radiosensitization. Prevailing views suggest that the antiproliferative effects of Met are mediated by activation of the AMPK pathway and/or improved tumor oxygenation (i.e., decreased hypoxia) due to inhibition of mitochondria, leading to decreased tumor cell respiration in irradiated tumors (29). The enhanced radiosensitivity upon treatment with Mito-Met analogues may be attributed to increased tumor oxygenation (i.e., decreased hypoxia). Tumor hypoxia (pO₂ < 10 mm Hg), an intrinsic property of numerous solid tumors including the pancreas, results from an imbalance between oxygen delivery and oxygen consumption (39). Studies suggest that decreasing oxygen consumption with pharmacologic drugs is an effective route for increasing tumor oxygenation and radiosensitivity (40–42). Met (1–10 mmol/L) reportedly improves tumor oxygenation and enhances tumor radiosensitivity (29). The current results show that Mito-Met₁₀ decreases mitochondrial respiration in MiaPaCa-2 cells after 24 hours. Mito-Met₁₀ inhibited mitochondrial complex I activity and tumor cell respiration at micromolar levels, whereas Met inhibited respiration to a similar extent at millimolar levels. It is likely that Mito-Met₁₀ stimulates tumor oxygenation at concentrations 1,000-fold lower than that of Met. A plausible mechanism by which Mito-Met₁₀ decreased mitochondrial respiration may be due to increased accumulation of Mito-Met₁₀ in mitochondria, leading to enhanced inhibition of complex I in the mitochondrial electron transport chain (43). In the presence of radiation and Mito-Met₁₀, it is possible that two or more mechanisms operate. AMPK-activating drugs increase tumor radiosensitivity (44). Radiation itself activates the AMPK energy sensor pathway (45). However, the degree to which AMPK induces tumor oxygenation and radiosensitivity remains poorly understood. Although Met inhibits growth of glioblastoma cells and mTOR pathway, the effects were found to be independent of AMPK (46). The same study suggests that AMPK could potentially function as a tumor growth supporter. Met-induced mTOR inhibition and suppression of glioma proliferation were attributed to enhanced PRAS40′s association with RAPTOR (46). Clearly, the antiproliferative mechanism of action of Met and Mito-Met may also be related to other mechanisms (i.e., activation of PRAS40/RAPTOR association).

Recent reports suggest that suppression of FOXM1 enhances the radiosensitivity of different human cancer cells (47). Mito-Met₁₀, at concentrations 1,000-fold lower than Met, inhibited mitochondrial respiration, activated AMPK, and significantly decreased FOXM1. Clearly, this is an exciting finding with significant potential to clinical translation and requires additional mechanistic studies. More recently, it was shown that at conventional antidiabetic doses of Met, there was no significant therapeutic effect in patients with advanced pancreatic cancer (48). The investigators suggested that more potent biguanides should be used in cancer treatment because of vastly reduced plasma concentrations typically detected in diabetic cancer patients treated with Met (48). It is conceivable that Mito-Met₁₀ exhibiting a 1,000-fold higher potency than Met would achieve a therapeutically effective plasma concentration in humans.

No potential conflicts of interest were disclosed.

Conception and design: G. Cheng, J. Zielonka, M. Lopez, M. Hardy, M.B. Dwinell, B. Kalyanaraman

Development of methodology: G. Cheng, J. Zielonka, C.S. Barrios, B.D. Johnson, M. Hardy

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Cheng, J. Zielonka, O. Ouari, D.M. McAllister, K.A. Boyle, C.S. Barrios, J. Weber, M. Hardy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Cheng, J. Zielonka, O. Ouari, D.M. McAllister, K.A. Boyle, C.S. Barrios, M. Hardy, M.B. Dwinell, B. Kalyanaraman

Writing, review, and/or revision of the manuscript: G. Cheng, J. Zielonka, K.A. Boyle, C.S. Barrios, B.D. Johnson, M. Hardy, M.B. Dwinell, B. Kalyanaraman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Cheng, M.B. Dwinell, B. Kalyanaraman

Study supervision: G. Cheng, M.B. Dwinell, B. Kalyanaraman

This work was supported by grants from the NIH National Cancer Institute (U01 CA178960 to M.B. Dwinell and B. Kalyanaraman and R01 CA152810 to B. Kalyanaraman), the Medical College of Wisconsin Cancer Center (M.B. Dwinell, B. Kalyanaraman and B. Johnson), and Aix-Marseille Université CNRS (M. Hardy and O. Ouari).

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.