The JNK signaling pathway plays a key role in methuosis (non-apoptotic cell death) induced by MOMIPP in glioblastoma

Background Synthetic indolyl- pyridinyl- propenones (IPPs) induce methuosis, a form of non-apoptotic cell death, in glioblastoma and other cancer cell lines. Methuosis is characterized by accumulation of cytoplasmic vacuoles derived from macropinosomes and late endosomes, followed by metabolic failure and rupture of the plasma membrane. However, not all IPPs that cause vacuolization are cytotoxic. The main goals of the present study were to identify key signaling pathways that contribute to methuosis induced by cytotoxic IPPs and to evaluate the anti-tumor potential of a prototype IPP in vivo. Methods We utilized metabolic flux analysis, glucose uptake, immunoblotting, and selective pharmacological inhibitors to compare the effects of closely related cytotoxic and non-cytotoxic IPPs in cultured glioblastoma cells. To determine whether the use of methuosis-inducing IPPs might be feasible in a therapeutic context, we quantified the distribution of our lead IPP compound, MOMIPP, in mouse plasma and brain, and tested its ability to inhibit tumor growth in an intracerebral glioblastoma xenograft model. Results The cytotoxic IPP compound, MOMIPP, causes early disruptions of glucose uptake and glycolytic metabolism. Coincident with these metabolic changes, MOMIPP selectively activates the JNK1/2 stress kinase pathway, resulting in phosphorylation of c-Jun, Bcl-2 and Bcl-xL. At the same concentration, the non-cytotoxic analog, MOPIPP, does not activate these pathways. Pharmacologic inhibition of JNK activity promotes survival, even when cells are extensively vacuolated, but suppression of c-Jun transcriptional activity offers no protection. MOMIPP readily penetrates the blood-brain barrier and is moderately effective in suppressing progression of intracerebral glioblastoma xenografts. Conclusions The results suggest that interference with glucose uptake and induction of JNK-mediated phosphorylation of pro-survival members of the Bcl-2 family represent key events in the methuosis death process. In addition to providing new insights into the underlying molecular mechanism of methuosis, the results indicate that compounds of the cytotoxic IPP class may have potential for further development as therapeutic agents for brain tumors. Electronic supplementary material The online version of this article (10.1186/s12885-019-5288-y) contains supplementary material, which is available to authorized users.


Background
Cancer cells often harbor genetic mutations that diminish the expression or function of proteins that regulate apoptosis (e.g., PTEN, p53, Rb) [1,2]. Consequently, tumors may exhibit reduced sensitivity to therapeutic agents that trigger this classical form of programmed cell death. This limitation has spurred interest in identifying alternative modes of cell death that can be exploited for cancer therapy. Methuosis is one of the most recent additions to the growing array of non-apoptotic cell death mechanisms [3,4]. Cells undergoing methuosis initially exhibit massive vacuolization of macropinosomes and late endosomes. Our recent studies indicate that this is due to defective trafficking of late endosomes to lysosomes, with concomitant homotypic fusion of the affected vesicular compartments [5]. The defect in lysosome-directed trafficking also affects autophagic flux, with resultant accumulation of autophagosomes [5]. Ultimately, the integrity of the cell membrane is compromised and the vacuolated cells rupture in a manner reminiscent of necrosis. These morphological features are distinct from apoptosis and other non-apoptotic cell death processes, and they cannot be prevented by caspase inhibitors or agents that block necroptosis or autophagy [3].
Despite the growing recognition of the morphological hallmarks of methuosis, the specific molecular mechanisms that link vacuolization of endocytic compartments to loss of cell viability remain poorly understood. Our structure-activity studies of MOMIPP and numerous analogs in GBM cells have provided valuable chemical tools to address this question. Specifically, we found that minor structural modifications of the indole ring yielded a functionally distinct sub-group of IPPs that retained the ability to induce robust morphological vacuolization, with greatly reduced cytotoxicity [21,22]. By comparing the effects of MOMIPP with one of the non-lethal analogs (MOPIPP; with propyl substituted for methyl at the 2-position of the indole ring), we noted that cells treated with MOMIPP had more severe inhibition of endolysosomal degradation pathways for EGF and LDL receptors [5]. Coincidentally, MOMIPP shows stronger binding affinity (lower K d ) for PIKfyve than the non-lethal analogs [10], despite the fact that the cells treated with these compounds have similar vacuolated morphologies.
In the present study, the objective was to expand the comparative analysis of cytotoxic versus non-cytotoxic vacuole-inducing IPPs in GBM cells, with the goal of defining pathways essential for triggering cell death. The results indicate that early impairment of glucose uptake and glycolytic metabolism, with attendant activation of JNK signaling and Bcl-2 phosphorylation, are key elements in the methuosis death program.

Methods
Cell culture  14) cell lines were purchased from the American Type Culture Collection (Manassas, VA). Normal human skin fibroblasts were originally derived from a skin biopsy as described previously [23]. All cell lines were maintained in Dulbecco's modified Eagle medium (DMEM; ThermoFisher, Walthham, MA), supplemented with 10% (v/v) fetal bovine serum (FBS) (JR Scientific, Woodland, CA) at 37°C with 5% CO2/95% air. Cell lines were confirmed negative for Mycoplasma contamination by periodic staining with DAPI or use of the PlasmoTest assay (InvivoGen, San Diego, CA).

Immunoblots and antibodies
U251 cells were typically seeded at 1 × 10 6 cells per 10 cm dish and maintained for 24 h in DMEM containing 10% FBS. On the day after plating, test compounds (or an equivalent volume of DMSO) were added in fresh medium and cells were harvested after the indicated times. Cells were lysed in SDS sample buffer (65 mM Tris-HCl, 10% glycerol, 0.5% β-mercaptoethanol, 2% SDS, pH 6.8) and the protein concentration was determined by colorimetric assay (Bio-Rad, Hercules, CA). Equal amounts of protein (80 μg) were subjected to SDS-PAGE, transferred to PVDF membrane. Immunoblot procedures were described previously [24]. Chemiluminescent signals were quantified using an Alpha Innotech FluorChem HD2 imaging system with Alpha View software.
Antibodies against the following proteins were obtained from Cell Signaling Technology (

Phase contrast microscopy
For live cell imaging studies, 1 × 10 5 U251 cells were seeded in 35-mm dishes and incubated for 24 h before commencing treatment with compounds. Cells were examined by phase-contrast microscopy using an Olympus IX70 inverted microscope equipped with a heated stage, a DP80 digital camera, and cellSens™ software (Olympus America, Center Valley, PA).
For immunofluorescence localization of Glut1, cells were grown on coverslips, treated with MOMIPP or DMSO for 4 h, washed with PBS, and then fixed in ice-cold methanol. After pre-blocking with 10% goat serum, cells were incubated with anti-Glut1 polyclonal antibody (Millipore, Cat. No. 07-1401) at 1/50 dilution for 1 h, followed by goat-anti-rabbit IgG, conjugated to AlexaFluor-488 (ThermoFisher) at 1/600 dilution. Fluorescence images were captured with an Olympus IX70 inverted fluorescence microscope, using a DP80 digital camera, and cellSens™ software (Olympus America).

Cell viability
Viability of cells in vitro was assessed by measuring cellular ATP using the CellTiter Glo® luminescence assay according to the manufacturer's protocol (Promega Corp., Madison, WI). U251 cells were seeded in white-walled opaque 96-well plates (2000 cells/well) with four replicate wells for each culture condition. After addition of compounds at the indicated concentrations, cell viability was assayed at a 48 h endpoint. Luminescence was quantified with a Berthold Tech Centro XS3 LB 960 luminometer, using the preinstalled MikroWin software.

Metabolic flux analysis
Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were determined using a Seahorse XF P analyzer, following standard protocols provided by the manufacturer (Agilent Technologies, Santa Clara, CA). Cell plating densities and concentrations of oligomycin and FCCP were optimized according to the recommended protocols. U251 cells were seeded at 15,000 cells per well in DMEM + 10% FBS in Seahorse miniplates. On the day after seeding, cells in parallel wells were changed over to fresh medium containing either 10 μM MOMIPP or 10 μM MOPIPP, with controls containing equivalent amounts of DMSO. Five hours after addition of the compounds, (or DMSO for controls), the cells were washed and switched to the recommended serum-free base medium for each type of metabolic flux assay. After insertion of the pre-hydrated sensor cartridges, the mini-plates were placed in the Seahorse instrument. OCR was measured under basal conditions (+glucose), and after sequential addition of oligomycin and FCCP. In separate assays, ECAR was measure under basal (glucose-free) conditions, and after sequential addition of glucose and oligomycin.

2-deoxy-D-glucose (2-DG) uptake
2-deoxy-D-glucose [1,[2][3] H(N)] (Cat. No. NET549A) was purchased from Perkin Elmer (Waltham, MA). For cells incubated with IPPs for 24 h, cellular uptake of 2-DG was determined essentially as described by Wood et al. [25]. Briefly, U251 cells were plated in 6 cm dishes at 2.5 × 10 5 cells per dish in standard medium. On the next day, fresh medium was added with the indicated compounds and cells were incubated for 24 h. After removal of medium and washing the cells, the cultures were incubated for 5 min at 37 o C in HEPES-buffered saline solution (20 mM HEPES, 140 mM NaCl, 2.5 mM MgSO 4 , 1 mM CaCl 2 , 5 mM KCl, pH 7.4) containing 1 μCi [ 3 H]2-DG and 10 μM unlabeled 2-DG. Parallel dishes were incubated under the same conditions, except that 10 μM cytochalasin B was added to block glucose transporter activity. Reactions were stopped by addition of ice-cold 0.9% NaCl, and after three washes, the cells were harvested in 1 ml lysis buffer: 20 mM Tris, 135 mM NaCl, 1 mM MgCl 2 , 1% Triton-X100, pH 8.0. Tritium in the lysates was quantified in a liquid scintillation counter and values for treated vs. control cultures were compared after subtraction of background from the cytochalasin B-treated cultures. To correct for possible differences in cell density after 24 h drug exposure, a portion of each cell lysate was utilized for quantification of α-tubulin by western blot analysis, and the tritium counts were normalized to the tubulin values before being compared to the controls. For short-term drug exposures (2-4 h), differences in cell density were not a factor, so the net tritium uptake values in control and treated cultures were compared directly without normalization to tubulin.
[ 3 H]2-DG uptake in freshly isolated mouse erythrocytes was assayed essentially as described by Ohmori et al. [26]. Briefly, erythrocytes were pretreated for 10 min at Rubidium uptake 86 Rb (Cat. No. NEZ072001) was purchased from Perkin Elmer. Ouabain-sensitive 86 Rb uptake was measured as described by Galuska et al. [27]. U251 cells were plated in parallel wells of a 12-well dish at 50,000 cells/well. On the next day, fresh medium (DMEM + 10% FBS) was added, with 10 μM MOMIPP or an equivalent volume of DMSO. After 4 h the cells were washed and pre-incubated for 10 min in serum-free DMEM, with or without 1 mM ouabain. Finally, 1 μCi of 86 Rb was added to each culture and incubation was continued at 37°C for an additional 10 min. Reactions were terminated by washing the cells with ice-cold 100 mM MgCl 2 and cell lysates were prepared for liquid scintillation counting. The 86 Rb counts in parallel cultures with ouabain were subtracted from those without ouabain to obtain the net ouabain-sensitive 86 Rb uptake.

C-Jun dominant negative construct
The c-Jun dominant-negative (DN) construct, pMIEG3-Jun DN [28], was provided by Alexander Dent (plasmid # 40350; Addgene, Cambridge, MA). The c-JunDN cDNA, which lacks the sequence encoding the first 122 amino acids of c-Jun, was amplified by PCR, using a forward primer: 5′-AAA AAA GGT ACC ATG ACT AGC CAG AAC ACG CTG CCC AGC GTC-3′ to add a KpnI site and a reverse primer: 5'-AAA AAA-TCT AGA-TCA GCT GGC ATA GTC AGG CAC GTC ATA AGG ATA GCT AAA TGT TTG CAA C-3' to add an XbaI site. The resulting PCR product was ligated into the pTRE-Tight vector (Clontech, Mountain View, CA), to generate pTRE-Tight (c-JunDN). The latter was combined with pTK-Hyg (Clontech) and introduced into the U251-Tet-On cell line, which stably expresses the reverse tetracycline-controlled transactivator (rtTA) [6], using a Nucleofector™ device and kit T reagents from Lonza (Walkersville, MD). The cells were selected in DMEM containing 10% FBS and 200 μg/ml of both G418 and hygromycin. Single colonies were picked, expanded and tested by western blot analysis for expression of HA-tagged c-JunDN in response to addition of 1 μg/ml doxycycline to the medium.

RT-PCR assessment of gene expression
Total RNA was extracted and purified from cultured cells using RNeasy Mini kit, following the manufacturer's protocol (SA Biosciences/Qiagen, Germantown, MD). cDNA was generated by reverse transcription of 100 ng of total RNA with the RT 2 first strand kit (Qiagen). RNA and cDNA were quantified and checked for purity (OD 260/280) using a Nano-Drop-1000 spectrophotometer (Thermo Fisher). RT-PCR reactions were carried out using primers for SERPINA3 and GFAP from Thermo-Fisher and GAPDH primers from Qiagen, using an Applied Biosystems StepOne Plus™ system. After 40 cycles, the ΔC T values for SERPINA3 and GFAP versus the GAPDH standard were calculated. ΔΔC T values indicate the difference between the average ΔC T values for the MOMIPP-treated cells and the controls. The expression fold change was determined as 2 -ΔΔCt .

Turnover and brain uptake of MOMIPP in vivo
All animal studies were performed in compliance with United States Public Health Service Policy on Humane Care and Use of Laboratory Animals, under protocols approved by University of Toledo Institutional Animal Care and Use Committee (Reference Number 107491).
Swiss Webster mice (8-10 weeks, female) were obtained from Charles River Laboratories and housed in ventilated cages on a 12 h light-dark cycle. MOMIPP was dissolved in NSP (10% NMP, 15% Solutol-HS15 in PBS) and administered to mice via intraperitoneal (IP) injection at a dose of 80 mg/kg. For pharmacokinetic studies, 3 mice were used for each time point. Mice were euthanized by CO 2 asphyxiation followed by cervical dislocation at the indicated times after drug injection. Blood was collected by cardiac puncture, held on ice in lithium heparin-coated tubes, and centrifuged at 10,000 x g for 20 min. Plasma samples were stored at − 80°C until analysis. Brains were placed in vials and frozen in liquid N 2 . Prior to quantification of MOMIPP, the frozen brains were weighed and homogenized 1:9 (w/v) in RIPA buffer (150 mM NaCl, 1% NP40, 0.5% sodium deoxycholate, 1% SDS, 50 mM Tris, pH 7.5).
To determine the concentration of MOMIPP, samples of plasma (200 μl) or brain homogenate (500 μl) were extracted with 1 ml ethyl acetate at 37°C for 20 min, followed by centrifugation for 2 min at 16,000×g. 800 μl of the resulting extract was vacuum centrifuged at 30°C for 1 h and the residue was suspended in 100 μl of chromatography elution solution: 30% (v/v) acetonitrile and 0.1% (v/v) formic acid in deionized water. Samples (10 μl) were injected onto a Waters Ascentis Express C18 column (75 × 21 mm, 2.7 μm) with matching guard column on a Waters 2795 HT-Alliance LC Separations Module, and isocratic elution was performed at a flow rate of 0.3 ml/min. MOMIPP was detected via multiple reaction monitoring on a Micromass Quattro Micro Mass Spectrometer (MS) in ESI+ mode with capillary voltage 3.0 kV, source temperature 100°C, desolvation temperature 400°C, desolvation gas flow 650 l/ h, cone gas flow 40 l/hr., and dwell time 0.2 s. MOMIPP was identified as 293.1 > 95.9 at cone voltage 40 V, collision energy 23 V, and a column retention time of 2.1 min. Nitrogen was used as the carrier gas through the mass spectrometer, while argon was used in the collision chamber.
For plasma standards and quality control calculations, solutions of known amounts of MOMIPP were prepared in methanol and the solution was evaporated at 30°C for 1 h in a vacuum centrifuge. 200 μl of mouse plasma (Pel-Freez Biologicals, Rogers, AZ) was added to each prepared tube to give six total plasma standard concentrations ranging from 10 to 100,000 nM, and three quality control samples from 30 to 30,000 nM. Brain standards were prepared in the same way, except that 500 μl of brain homogenate from untreated mice was added, and the measured concentrations for brain tissue were corrected for the 10-fold dilution of brain tissue in RIPA buffer during homogenization. Levels of MOMIPP in plasma were based on volume and expressed as nM concentration. Brain homogenate levels were measured based on tissue mass prior to homogenization, so that these concentrations represent nmol/kg tissue. For ease of comparison, tissue densities were approximated as 1.0 g/ml, so that brain levels could also be expressed as nM concentration.

Anti-tumor efficacy study in an orthotopic xenograft model
Athymic CrTac:NCR-Foxn1<nu> mice (female, 7-8 weeks) were purchased from Taconic Biosciences (Rensselaer, NY) and housed under SPF conditions (three per cage) on a 12 h light-dark cycle in ventilated cages with barrier filters. For initiation of xenografts that could be detected by bioluminescence imaging (BLI), we nucleofected U251 cells with pCMV5neo-Luc, which encodes the firefly luciferase gene excised from pGL3 (Promega, Madison, WI). Clonal selection was performed in medium containing G418. The resulting stable cell line (U251-LUC), was maintained in DMEM supplemented with 10% (v/v) FBS and 200 μg/ml G418 and was periodically tested for uniform luciferase by immunofluorescence analysis using a luciferase antibody (Sigma, Cat. No. L2164).
The intracerebral xenografts were established based on the method described by Ozawa and James [29]. Mice were anesthetized by IP injection with a mixture of ketamine (100 mg/kg) and xylazine (7.5 mg/kg) and placed on a warming pad. A 1 cm incision was made down the midline of the scalp. A small hole was drilled in the skull using a 25 g needle at a point 1 mm anterior and 2 mm lateral to the bregma. 4 × 10 5 cells (suspended in 3 μl of DMEM) were injected into the brain using a 10 μl Hamilton syringe equipped with a 27 g needle, with a cuff placed 3 mm from the tip of the needle to control the depth of injection. The hole in the skull was sealed with sterile bone wax and the skin incision was closed with surgical glue. Post-surgical analgesia was provided with buprenorphine (0.1 mg/kg) administered twice a day for the first two days. Tumor growth was monitored by BLI. All mice had focal BLI in the range of 10 7 -10 8 photons by the fourth day after tumor cell implantation. Mice were then grouped randomly into control and MOMIPP treatment groups (10 mice for the control and 11 mice for the treatment group; one mouse did not survive after surgery). Sample size was determined using Biostat Power and Precision® software, which indicated that 10 mice per group was sufficient to detect a 25% difference in the mean tumor size between control and treated mice (80% confidence), assuming a standard deviation of 20% in each group. MOMIPP (80 mg/kg, in NSP) or vehicle was administered by IP injection every 24 h for 15 consecutive days. Tumor progression was monitored by BLI on the 7th, 11th, and 15th days after commencement of drug treatment. On the 15th day, all mice were euthanized by CO 2 asphyxiation, followed by cervical dislocation. Blood was collected by cardiac puncture and utilized for blood chemistry analysis with a comprehensive diagnostic profile kit and a VetScan VS2 Analyzer (Abaxis, Union City, CA), following protocols recommended by the manufacturer.

Statistics
GraphPad Prism software was used for statistical analyses. Student's unpaired t-test was applied to cell culture studies and Mann-Whitney unpaired test was used for in vivo tumor studies. P values ≤ 0.05 were regarded as significant.

Differential effects of cytotoxic versus non-cytotoxic IPP's on glycolytic metabolism
As shown in Fig. 1a, the morphological effects of the cytotoxic MOMIPP and the non-cytotoxic 2-propyl analog, A B C Fig. 1 MOMIPP selectively impairs glycolytic metabolism compared to the non-cytotoxic analog, MOPIPP. a One day after plating, U251 cells were treated with 10 μM MOMIPP or MOPIPP and examined by phase contrast microscopy after the indicated periods. b & c Five hours after addition of 10 μM MOMIPP or MOPIPP, U251 cells were subjected to metabolic flux analysis to assess oxidative function (OCR) or glycolytic function (ECAR), using a Seahorse® system, as described in the Methods. Three control and three drug-treated cultures were sampled at each time-point. Values are means (± SD), and significance of the differences between the readings at each point was assessed by Student's unpaired t-test. In panel (c), the asterisks indicate that the ECAR values from the MOMIPP-treated cells were significantly different from the control values at all time points (p ≤ 0.05). The ECAR values for the MOPIPP-treated cells were not statistically different from the controls MOPIPP, are indistinguishable at 4 h, with both compounds inducing extreme vacuolization of U251 GBM cells. However, as noted previously [5], by 24 h the cells treated with MOMIPP begin to round up, detach from the surface of the dish, and lyse, whereas the cells treated with MOPIPP remain attached and viable. Loss of cell viability caused by methuosis-inducing IPPs coincides with a decline in cellular ATP [5,8]. Therefore, to extend our comparison of the cellular effects of MOMIPP versus MOPIPP, we asked whether MOMIPP might have selective early effects on cellular metabolic functions in GBM cells. To assess metabolic activity, we began by conducting a Seahorse® metabolic flux analysis starting 5 h after addition of the compounds, measuring oxygen consumption rate (OCR) as an indicator of mitochondrial respiratory capacity (Fig. 1b). The results indicated that basal respiration, proton leak (+oligomycin) and maximal respiration (+FCCP) were not substantially different in cells treated for 5 h with MOMIPP or MOPIPP, compared to controls. However, when glycolytic function was assessed by measuring extracellular acidification rate (ECAR), the results indicated that cells treated with MOMIPP for 5 h experienced a significant reduction in basal and glucosestimulated glycolysis, as well as maximum glycolytic capacity (+oligomycin) (Fig. 1c). In contrast, glycolytic function in the cells treated with MOPIPP was not significantly different from the control (Fig. 1c).
The glycolytic pathway plays a key role in meeting the bioenergetic needs of glioblastoma and other cancer cells [30]. Therefore, the results of the preceding metabolic flux analysis prompted us to look more closely at glucose transport in cells treated with MOMIPP versus MOPIPP. We found that uptake of the non-metabolizable glucose analog, [ 3 H]2-deoxyglucose (2-DG), was reduced by 80% in U251 cells exposed to 10 μM MOMIPP for 24 h (Fig. 2a). By comparison, MOPIPP had a much more modest effect, with an approximate 40% reduction at that time. The effect of MOMIPP on 2-DG uptake was not specific to U251 cells, as variable degrees of suppression were observed in a broad spectrum of human GBM cell lines (Fig. 2b). Since glycolytic function was compromised by MOMIPP within 5 h (Fig. 1c), we repeated the study of 2-DG uptake after only 2 h of drug treatment (Fig. 2c). The results show that suppression of 2-DG is a very early effect of MOMIPP, but not MOPIPP. A dose-response study with MOMIPP indicated that 2-DG uptake was markedly suppressed at ≥2.5 μM (Fig. 2d), which correlates with the GI 50 previously reported for MOMIPP in U251 cells [22]. Given that MOMIPP suppressed 2-DG uptake in a variety of GBM cell lines, we asked whether 2-DG uptake might be affected to the same extent in non-transformed cells. We previously showed that, compared to glioma cells, cultured human skin fibroblast are less sensitive to the cytotoxic effects of MOMIPP [9]. Therefore, in Fig. 2e we assessed the effects of 10 μM MOMIPP on 2-DG uptake in fibroblasts vs. U251 cells after 4 h or 24 h. The results demonstrate that 2-DG uptake in the fibroblasts was not inhibited at 4 h, and was decreased by only 30% at 24 h (compared to 80% inhibition in the U251 cells). Finally, to rule out the possibility that the effect of MOMIPP on 2-DG uptake was due to generalized disruption of cell membrane structure or function, we assayed the activity of Na + ,K + -ATPase by tracking the ouabain-sensitive uptake of 86 Rb + (Fig. 2f). The results indicated that 10 μM MOMIPP had no effect on the sodium pump.
To explore further the mechanism of MOMIPP suppression of 2-DG uptake, we considered the possibility that the compound might directly inhibit glucose transporters or, alternatively, promote sequestration of glucose transporters in the intracellular vacuole compartments. For this purpose, we utilized erythrocyte preparations, which lack endocytic machinery and contain abundant Glut1in their surface membranes. As shown in Fig. 3a, MOMIPP had no direct effect on 2-DG uptake in erythrocytes, while known inhibitors of Glut1 were very effective in reducing 2-DG uptake. On the other hand, immunofluorescence localization of Glut1 in U251 cells revealed prominent sequestration of the transporter in intracellular vacuoles, contrasting with control cells where Glut1 fluorescence was predominantly associated with the cytoplasm and the plasma membrane (Fig. 3b). To confirm that inhibition of 2-DG uptake was related to the formation of endosomal vacuoles, we took advantage of the previously reported ability of the H + -ATPase inhibitor, Bafilomycin A 1 , to prevent the formation of endosomal vacuoles initiated by IPPs and other compounds [8,31]. The results in Fig. 3c&d demonstrate that Bafilomycin A 1 effectively blocked vacuolization induced by MOMIPP, and prevention of vacuole formation abrogated the inhibitory effect of MOMIPP on 2-DG uptake.

Cytotoxic IPPs selectively activate the JNK stress kinase pathway
A number of studies have established that cancer cells are particularly susceptible to oxidative stress and cytotoxicity induced by glucose deprivation or inhibition of glycolysis [32]. Activation of the c-Jun N-terminal kinase (JNK) signaling pathway is a key step in this process [32][33][34]. Thus, we asked whether the cytotoxic methuosisinducing IPPs might have differential effects on JNK, compared to closely related IPPs that induce vacuolization without cytotoxicity. For this purpose, we selected two methuosis inducers, MOMIPP and 2q, and two non-toxic vacuole-inducers, MOPIPP and 2a. The structures and activities of 2q and 2a were reported previously [22]. As shown in Fig. 4a&b, 10 μM MOMIPP and 2q caused substantial vacuolization, cell rounding and detachment, with loss of viability (ATP) by 48 h. In contrast, cells treated with 10 μM MOPIPP or 2a were vacuolated but remained attached, and they showed relatively small decreases in ATP compared to the vehicle-treated control. When JNK activation (phosphorylation) was assessed by western blot analysis after 24 h (Fig. 4c), we found that MOMIPP and 2q induced major increases in JNK1/2 phosphorylation. By comparison, the JNK phosphorylation signals on the blots from cells treated with the non-toxic MOPIPP and 2a were much weaker. Consistent with the activation of JNK, we also observed a parallel increase in phosphorylation of the JNK target, c-Jun, in cells treated with MOMIPP and 2q, but not MOPIPP and 2a (Fig. 4c).
Two MAP kinase kinases, MKK4 and MKK7, mediate phosphorylation of JNK on Tyr and Thr [35]. These kinases can, in turn, be activated by several different upstream kinases (e.g., ASK1, MLK, TAK1). As shown by the western blots in Fig. 4d, MOMIPP and 2q triggered a robust increase in phosphorylation of MKK4. Treatment with MOPIPP and 2a caused a more modest increase. By comparison, neither MOMIPP nor MOPIPP dramatically altered phosphorylation of MKK7 (Fig. 4d). These observations are consistent with the concept that MKK4 responds predominantly to environmental stress, whereas MKK7 is associated with cytokine-mediated pathways [35,36].

Cytotoxic IPPs trigger increased phosphorylation of Bcl-2 and Bcl-xL
In addition to the c-Jun transcription factor, other known substrates for JNK include pro-survival members of the Bcl-2 family, which play key roles in protecting cells from apoptosis and necrosis [37,38] and regulating autophagy. Indeed, phosphorylation of Bcl-2 and Bcl-xL by JNK1 is postulated to interfere with their interactions with proapoptotic members of the Bcl-2 family (e.g., BAX) and the autophagy regulator, Beclin-1, thereby promoting cell death or autophagy [39][40][41]. The results depicted in Fig. 4e indicate that increased phosphorylation of Bcl-2 and Bcl-xL accompanied the activation of JNK in cells treated with the methuosis-inducing compounds, MOMIPP and 2q.

Activation of JNK is an early step in methuosis
To determine if activation of JNK represents an initial step in the evolution of the methuosis phenotype, we examined the phosphorylation of JNK after only 4 h exposure of cells to MOMIPP or MOPIPP. The results show an early increase in phosphorylation of JNK upon treatment with MOMIPP, but not MOPIPP (Fig. 4f). Interestingly, the western blots revealed a time-dependent change in the phosphorylation pattern of the JNK splice variants, with the p54 isoform predominating at 4 h (Fig. 4f ), and the p46 isoform predominating at 24 h (Fig. 4c&f ). Consistent with the activation of JNK, phosphorylation of c-Jun also increased 4 h after addition of MOMIPP (Additional file 1: Figure S1A). In contrast to JNK, there was no detectable change in phosphorylation of the p38 MAP kinase after 4 h of MOMIPP treatment (Additional file 1: Figure S1B). As in the studies conducted at 24 h, we observed substantial increases in phosphorylation of Bcl-2 and Bcl-xL at 4 h in cells treated with MOMIPP (Additional file 1: Figure S1C). However, we did not detect any change in expression or phosphorylation of the pro-apoptotic Bcl-2 family member, BAD (Additional file 1: Figure S1C). An important question raised by the preceding results is whether the activation of JNK is causally related to loss of viability in cells treated with MOMIPP. To address this point we exposed U251 cells to the methuosisinducing IPPs, MOMIPP and 2q, in the presence or absence of a JNK inhibitor, SP600125 (Fig. 5). The JNK inhibitor completely blocked basal and drug-induced phosphorylation of c-Jun (Fig. 5a). Likewise, the inhibitor prevented the increased phosphorylation of Bcl-2 and Bcl-xL triggered by MOMIPP or 2q (Fig. 5b). Most importantly, addition of SP600125 to cells treated with MOMIPP or 2q had a marked protective effect on cell viability. Although the cells remained extensively vacuolated, the JNK inhibitor prevented them from rounding up, detaching and lysing (Fig. 5c). The protective effect of the JNK inhibitor was also reflected in the results of the Cell-TiterGlo® viability assay (Fig. 5d).

Relationship between vacuole formation and JNK activation
The observation that cells treated with MOMIPP or MOPIPP have indistinguishable vacuolated morphologies at early time points, while only the cells exposed to MOMIPP show increased JNK phosphorylation, prompted us to further explore the relationship between JNK activation and endosomal vacuolization. We began by asking whether YM201636, a PIKfyve inhibitor with a chemical structure distinct from MOMIPP (See figure on previous page.) Fig. 4 Methuosis-inducing IPPs specifically trigger activation of the JNK pathway. a U251 cells were treated with the indicated IPPs at a concentration of 10 μM and phase contrast images were obtained after 24 h. b Dose-response studies were carried out with each IPP. Cell TiterGlo® viability assays were performed after incubation for 48 h with compounds at the indicated doses. c-e U251 cells were treated with the indicated IPPs for 24 h, and immunoblot analyses for the indicated proteins were performed on equal amounts of cellular protein as described in the Methods. f U251 cells were treated with the MOMIPP or MOPIPP for either 4 h or 24 h, and equal amounts of cell protein were immunoblotted for phospho-JNK or total JNK as described in the Methods. The [42], could also induce JNK activation and subsequent cell death in U251 GBM cells. Within 4 h, 10 μM YM201636 caused extensive cytoplasmic vacuolization, similar to the phenotype induced by MOMIPP (Fig. 6a). By 48 h, YM201636 and MOMIPP both caused obvious decreases in cell viability at concentrations above 2.5 μM (Fig. 6b). However, comparison of the phosphorylation states of JNK, c-Jun, Bcl-2 and Bcl-xL in cells treated with these compounds demonstrated that only MOMIPP promoted substantial JNK activation and phosphorylation of JNK substrates (Fig. 6c&d). Consistent with the potential link between inhibition of glycolytic activity and JNK activation, inhibition of [ 3 H]2-DG uptake by YM201636 was only 40% after 24 h, compared to 90% by MOMIPP (Fig. 6e).
In light of the divergent effects of the different vacuoleinducing compounds, we asked whether vacuolization of endocytic compartments was required in order for MOMIPP to induce JNK activation. To address this question, we took advantage of the ability of amiloride and its analogs to inhibit macropinocytosis [43]. As shown in Fig. 7a, treatment of U251 cells for 4 h with 5-(N-ethyl-N-isopropyl) amiloride (EIPA) by itself did not alter baseline cell morphology, but when combined with MOMIPP, EIPA caused a marked decrease in the number and size of vacuoles. At the same time, the 3-fold increase in the ratio of phosphorylated to total JNK triggered by MOMIPP was reduced to just 0.5-fold in the presence of EIPA (Fig. 7b). These results indicate that while not all vacuole-inducing compounds trigger JNK activation, in the case of cells treated with MOMIPP the activation of JNK is indeed linked to the accumulation of vacuoles.
Activation of c-Jun-mediated transcription is not required for methuosis c-Jun functions as part of the AP-1 transcription complex, which controls the expression of many genes involved in cell proliferation, survival and death [44]. In view of the marked increase in JNK-mediated phosphorylation of c-Jun triggered by MOMIPP, we carried out a study to determine if transcriptional events regulated by c-Jun are essential for methuosis. We began by generating a stable U251 cell line (U251-c-JunDN) capable of doxycycline (Dox)-inducible expression of a dominant-negative c-Jun construct. The latter binds effectively to DNA but contains a deletion in the transactivation domain, making it a potent inhibitor of endogenous AP-1 activity [28,45]. As shown in Fig. 8a, expression of c-JunDN was tightly controlled by Dox in the stable cell line. To confirm that c-JunDN could in fact suppress c-Jun function in U251 cells, we examined the expression of two genes that are regulated by AP-1 transcriptional complexes in astrocytes and glioma cells; SERPINA3 (α 1 -antichymotrypsin) and GFAP [46]. Transcription of both genes was almost completely blocked when c-JunDN expression was induced by Dox (Fig. 8b).
To determine if c-Jun activation plays a key role in methuosis, U251-c-JunDN cells were treated with MOMIPP, with or without prior induction of c-JunDN. The expression of c-JunDN had no effect on the morphological hallmarks of methuosis (e.g., extreme vacuolization, cell rounding, detachment and lysis) (Fig. 8c), and it did not prevent the loss of cell viability, assessed by the ATP assay (Fig. 8d). Western blot analysis confirmed that expression of c-JunDN did not interfere with stress signaling upstream from c-Jun, since JNK phosphorylation was still increased by MOMIPP (Fig. 8e). Consistent with the preservation of JNK activation in cells expressing c-JunDN, the MOMIPP-induced phosphorylation of Bcl-2 and Bcl-xL was also sustained (Fig. 8e).

Evaluation of MOMIPP pharmacokinetics
The ultimate goal of developing chemical inducers of non-apoptotic cell death in glioblastoma is to employ these compounds as novel therapeutic agents. Therefore, we extended our studies to assess the pharmacokinetic properties of MOMIPP, evaluate its ability to penetrate the blood-brain barrier (BBB), and obtain a preliminary indication of its potential anti-tumor efficacy in a xenograft model.
For initial studies, MOMIPP was prepared in several different solvents and administered via oral (PO) or intraperitoneal (IP) routes at doses ranging from 20 to 80 mg/kg. We learned that the PO route is only effective if mice are fasted, which is impractical for long-term efficacy studies. We also found that MOMIPP is rapidly cleared from the circulation, necessitating high initial doses to achieve sustained plasma levels in the range shown to be effective for inducing methuosis in vitro (≥ 2.5 μM, see Fig. 4b). Based on the results of these preliminary studies, we selected NSP (10% n-methyl-2-pyrrolidone, 15% Solutol HS15, 75% phosphate-buffered saline) as the optimal vehicle and 80 mg/kg as the dose for IP administration of MOMIPP. As shown in Fig. 9a, when mice were given a single injection of the compound, the plasma concentration rapidly reached 40 μM, and remained above 10 μM for 8 h. By 24 h the plasma concentration declined to approximately 1 μM. Of particular interest, the concentrations of MOMIPP in the brain were approximately 50-60% of the concentrations in the plasma up to 8 h, indicating that MOMIPP readily crosses the BBB.

Effect of MOMIPP on tumor progression
Based on the foregoing studies, we elected to initiate a study to evaluate the effects of MOMIPP on the growth of orthotopic GBM xenografts in immunocompromised mice. The U251 human glioblastoma cell line has been widely used to establish aggressive intracerebral tumors with many of the pathological features of human glioblastoma [47]. To permit monitoring of tumor progression by bioluminescence imaging (BLI), we developed a stable U251 cell line engineered to express firefly luciferase (U251-LUC) (Additional file 2: Figure S2). Four days after intracerebral inoculation of U251-LUC cells, mice were sorted randomly into control and treatment groups (Day-0). At this point BLI indicated no significant difference between the groups (control, 1.14 ± 0.67 × 10 8 photons; treatment group 1.11 ± 0.45 × 10 8 photons) (Fig.  9b). Treatment of the mice was then started, with mice in the treatment group receiving daily injections of MOMIPP (80 mg/kg in NSP), while the control group received an equivalent volume of the NSP vehicle. By day-7 the BLI signals in the control group (3.27 ± 0.45 × 10 8 photons) had tripled compared to day-0, indicating substantial A B Fig. 7 Inhibition of macropinocytosis reduces vacuole formation and attenuates the activation of JNK in cells treated with MOMIPP. U251 cells were treated with MOMIPP, EIPA or a combination of both compounds for 4 h. a Cellular vacuolization was assessed by phase-contrast microscopy. b Immunoblot analysis was performed as described in the Methods. A representative blot is shown in the upper panel and the ratios of phosphorylated to total JNK (combined 54 kDa and 46 kDa splice variants) from three separate determinations are depicted in the bar graph tumor growth. By comparison, tumor progression in the MOMIPP treatment group (1.79 ± 1.72 × 10 8 photons) was significantly inhibited (Fig. 9b). Subsequent BLI images acquired on day-11 and day-15 (Fig. 9b) confirmed a continued increase in the size of the tumors in the control group, with significant suppression of tumor progression in the MOMIPP-treated group.
The behavior of the MOMIPP-treated mice was not altered during the course of the study. Their weight remained comparable to the controls until day-15, when the controls began to lose weight, prompting termination of the study (Additional file 3: Figure S3). Blood chemistry profiles obtained for all of the control and treated mice at the end of experiment showed no significant differences in key enzymes and metabolites (Additional file 4: Table S1), suggesting that 15-day treatment with high-dose MOMIPP did not cause systemic toxicity or organ failure.

Discussion
In the present studies, we used MOMIPP and structurallyrelated indolyl chalcones (IPPs) to explore the mechanism of cytotoxicity underlying the form of non-apoptotic cell death termed 'methuosis'. The results indicate that although these compounds initially induce morphological phenotypes that appear identical (i.e., massive vacuolization of macropinosomes and endosomes), the cytotoxic compounds differ from the non-lethal compounds in having distinct effects on glycolytic metabolism and the stress-induced JNK signaling pathway. Within the first several hours after vacuoles appear, the prototype methuosis-inducing compound, MOMIPP, caused a substantial reduction in glucose uptake and a decline in glycolytic function. In contrast, the non-cytotoxic analog, MOPIPP did not have major metabolic effects, despite the fact that in previous studies the number of vacuoles per cell was not statistically different between cells treated with MOMIPP or MOPIPP [5]. One possible explanation for these differential effects emerges from the recent identification of the endosomal phosphoinositide kinase, PIKfyve, as a MOMIPP target [10].
The phosphoinositide products of PIKfyve play key roles in the biogenesis of multivesicular endosomes and the assembly of protein complexes required for vesicular trafficking within the endocytic pathway [11,48]. Indeed, perturbation of the enzyme is known to cause vacuolization of endocytic compartments [49], enlargement and fusion of lysosomes [50], and disruption of autophagic flux [51]. PIKfyve also appears to affect endosome recycling under some circumstances, although this aspect of its function is still not completely understood. For instance, inhibition of PIKfyve disrupts recycling of some junction proteins (claudins) but not others (occludin, E-cadherein) [52]. Of particular relevance for the present study, perturbation of PIKfyve can impair insulin-stimulated glucose uptake and translocation of GLUT4 and GLUT1 from endosomes to the cell surface [53]. It is interesting to note A B Fig. 9 MOMIPP penetrates the blood-brain barrier and inhibits growth of intracerebral U251 xenografts. a MOMIPP (80 mg/kg) was administered by IP injection to female Swiss Webster Mice. At each of the indicated time points after injection, blood and brain tissue were obtained from three mice. Plasma and brain tissue were extracted and MOMIPP concentrations (mean ± SD) were determined by LC/MS as described in the Methods. b Intracerebral xenografts were established in nude mice by injection of 4 × 10 5 U251-LUC cells, as described in the Methods. Daily treatment with MOMIPP (80 mg/kg) or NSP vehicle commenced on Day-0 (four days after cell implantation). The study was terminated on Day-15. BLI was performed on Days 0, 7, 11 and 15, and the differences between the control (n = 10) and MOMIPP-treated (n = 11) groups were evaluated for statistical significance by the Mann-Whitney test for unpaired samples that when Cho et al. [10] compared several IPPs from our series [22], they found that the cytotoxic MOMIPP had a significantly greater affinity for PIKfyve than the non-lethal vacuole inducer, compound, 2a. This raises the possibility that despite similar morphological appearances (i.e., number and size of phase-lucent vacuoles), the cells treated with the more potent PIKfyve inhibitors like MOMIPP may experience a more stringent block in some of the vesicular trafficking functions regulated by this kinase. This concept is supported by our recent study showing that MOMIPP causes a more severe disruption of EGF-receptor and LDL-receptor lysosomal degradation than the non-lethal MOPIPP [5]. Similarly, it is conceivable that compared to MOPIPP and other non-cytotoxic IPPs, MOMIPP may cause greater retention of glucose transporters in non-recycling endosomal vacuoles, ultimately causing a decrease in glucose uptake. While this mechanism merits further investigation, it is not clear whether it alone could account for the differential cytotoxicity of the IPP series. For instance, our finding that the known PIKfyve inhibitor, YM201636, had a much smaller effect on [ 3 H]2-DG uptake than MOMIPP (Fig. 6) leaves open the alternative possibility that MOMIPP causes reduced glycolysis and activation of JNK through interactions with targets other than PIKfyve. In this regard, intriguing possibilities include PFKFB3 and PFKFB4 (6-phosphofructo-2-kinase/fructose-2,6-biphosphatases 3 and 4), which play important roles in regulating glycolytic flux, especially in cancer cells [54,55]. Recently developed antineoplastic inhibitors of PFKFB3 (3PO and PFK15) have chalcone-like structures reminiscent of MOMIPP, with pyridinyl groups as key elements [56,57]. Therefore, in the future it will be interesting to determine if MOMIPP can directly inhibit PFKFB3.
The results from this study demonstrate that the JNK signaling axis plays a critical role in methuosis, a unique form of cell death induced by MOMIPP and other cytotoxic IPPs. Stress-induced activation of the JNK signaling pathway is widely recognized as a stimulus that can lead to apoptosis [35,58]. However, previous studies from our group and others have established that methuosis is distinct from apoptosis, based on absence of classical apoptotic hallmarks (e.g., chromatin condensation, DNA laddering, cell shrinkage) and insensitivity to caspase inhibitors [3,8]. JNK has also has been implicated in regulation of necroptosis, a form of programmed necrosis [58]. But, unlike necroptosis, methuosis is not attenuated by the RIP1 kinase inhibitor, necrostatin-1 (Additional file 5: Figure S4). Thus, methuosis can be added to the growing list of non-apoptotic cell death pathways in which JNK plays a pivotal role [58].
Phosphorylation by JNK increases the activity of c-Jun and other nuclear transcription factors. However, the results of our studies utilizing the c-JunDN construct suggest that MOMIPP-induced cell death does not depend on changes in gene expression mediated by activation of c-Jun. Rather, it appears that post-translational modification of cytoplasmic JNK substrates represents the most likely trigger for cell death. Among the many known JNK targets in the cytoplasm, we focused on Bcl-2 and Bcl-xL, because multi-site phosphorylation of Bcl-2 (T56, S70,T74,S87) and phosphorylation of Bcl-xL (S62) by JNK are thought to disrupt the pro-survival functions of these proteins [39,59,60]. The precise mechanism through which Bcl-2 phosphorylation promotes cell death is still unclear, and it may vary depending on the cell type and the stimulus [61]. Possibilities include: 1) disruption of interactions with pro-apoptotic proteins like Bax, facilitating outer mitochondrial membrane pore formation and release of cytochrome c [62]; 2) alteration of Bcl-2 function in the endoplasmic reticulum, causing discharge of Ca 2+ , with consequential mitochondrial calcium overload [63]; 3) release of Beclin-1 from inhibitory Beclin-1/Bcl-2 complexes, causing increased autophagy [41]; or 4) loss of mitochondrial membrane potential and release of proteins like AIF, which are capable of inducing caspase-independent cell death [64,65]. The latter mechanism merits serious consideration, since methuosis is not blocked by caspase inhibitors [8,15]. Finally, it should be noted that activation of JNK by MOMIPP could have pleiotropic effects beyond phosphorylation of Bcl-2/Bcl-xL. For example, JNK activation is known to amplify production of reactive oxygen species [66], which has the potential to compromise metabolic functions and trigger necrotic cell death [67].
The molecular events linking endosomal vacuolization and changes in glycolytic metabolism to the activation of JNK remain to be defined. Two upstream kinases, MKK4 (SEK1) and MKK7, are known to phosphorylate JNK, with MKK7 mediating signals from proinflammatory cytokines (e.g., TNF) and MKK4 serving as a sensor for environmental stress [35,36]. Previous studies have suggested that glucose deprivation may modulate MKK4 by altering the crosstalk between two JNK-interacting scaffold proteins, JIP1 and JIP3, resulting in activation of ASK1, a kinase that targets MKK4 [68]. Therefore, it is conceivable that the JIP1/3 → ASK1 → MKK4 signaling module may serve as a conduit to activate JNK when endosomal trafficking is disrupted and glucose levels fall during the early stages of methuosis. Our finding that MKK4 was robustly activated by the methuosis-inducing IPPs, MOMIPP and 2q, and less so by the non-cytotoxic vacuole-inducers, MOPIPP and 2a, supports this concept.
Thus far, evaluation of the anticancer activity of methuosis-inducing compounds has been confined mainly to cell culture systems. However, two recent studies have highlighted the potential for exploiting this novel form of cell death for treating cancers in vivo.
In one study Huang et al. [15] identified a unique 4′6'-disubstituted aza-indole that selectively induced methuosis in a broad panel of cancer cell lines in vitro and suppressed the growth of subcutaneous MDA-MB-231 breast cancer xenografts in immunocompromised mice. In a separate study, Ahlstedt et al. [69] found that a quinolonebased methuosis inducer, Vacquinol-1, reduced the size of brain tumors in syngeneic rat models, although no survival advantage was noted. In the present study, we found that MOMIPP readily penetrates the BBB and significantly suppresses the progression of intracerebral GBM xenografts without overt toxicity in nude mice. However, overall growth suppression was modest, and high doses of MOMIPP were required to compensate for the rapid clearance of the compound from the circulation. Even with relatively high daily doses, our pharmacokinetic observations suggest that by the end of each 24 h period the brain levels of MOMIPP would likely fall below the concentrations found to be therapeutically effective when maintained in vitro. For these reasons, we did not carry out survival studies or immunohistochemical analyses of JNK activity in tumor tissues at this stage. The initial findings reported here, coupled with the fact that MOMIPP can kill GBM cells that are resistant to the standard drug, temozolomide [9], suggest that further development of IPPs as possible therapeutic agents for brain tumors is warranted. Improvements in efficacy may be realized by structural modifications that increase potency or prolong drug half-life in vivo. In addition, the incorporation of these compounds into sustained-release formulations or targeted delivery vehicles may prove advantageous. Finally, the new mechanistic insights pointing to decreased glycolytic function and induction of the JNK stress pathway as important precipitating events in methuosis may suggest opportunities for synergistic combinations with other therapeutic agents that alter these pathways [57].