Ion Channel Drugs Suppress Cancer Phenotype in NG108-15 and U87 Cells: Toward Novel Electroceuticals for Glioblastoma

Glioblastoma is a rapidly progressing brain cancer that is very difficult to treat. Given that many aspects of cell and tissue behavior are controlled by electric signaling, we sought to test whether drugs that target ion channel proteins might be effective at controlling the spread and functionality of glioblastoma cells in culture. Testing aspects of cell growth and physiology, we show that several novel combinations of ion channel drugs, which are already approved in human patients for other purposes, are highly effective against two types of glioblastoma cells. This facilitates the development of new strategies to address cancer by repurposing the large class of ion channel drugs against cancer.

Glioblastoma is a lethal brain cancer that commonly recurs after tumor resection and chemotherapy treatment. Depolarized resting membrane potentials and an acidic intertumoral extracellular pH have been associated with a proliferative state and drug resistance, suggesting that forced hyperpolarization and disruption of proton pumps in the plasma membrane could be a successful strategy for targeting glioblastoma overgrowth. We screened 47 compounds and compound combinations, most of which were ion-modulating, at different concentrations in the NG108-15 rodent neuroblastoma/glioma cell line. A subset of these were tested in the U87 human glioblastoma cell line. A FUCCI cell cycle reporter was stably integrated into both cell lines to monitor proliferation and cell cycle response. Immunocytochemistry, electrophysiology, and a panel of physiological dyes reporting voltage, calcium, and pH were used to characterize responses. The most effective treatments on proliferation in U87 cells were combinations of NS1643 and pantoprazole; retigabine and pantoprazole; and pantoprazole or NS1643 with temozolomide. Marker analysis and physiological dye signatures suggest that exposure to bioelectric drugs significantly reduces proliferation, makes the cells senescent, and promotes differentiation. These results, along with the observed low toxicity in human neurons, show the high efficacy of electroceuticals utilizing combinations of repurposed FDA approved drugs.

Glioblastoma (GBM) is a cancer of the brain that has a high mortality rate and a short median survivability time of around 14 to 15 months [1]. Prognosis is particularly poor for glioblastoma patients due to the failure of standard interventions in preventing tumor re-growth. Currently, the standard treatment for GBM is resection of the main tumor followed by radiation and temozolomide (TMZ) administration [2]. However, reoccurrence of GBM is common (90%) and usually arises from the area immediately next to the resected area [3,4]. Thus, the invasiveness and hypermigratory nature of these residual cancer cells make resection alone ineffective; additional interventions that prevent the proliferation of these cells are needed to complement current methodologies.

Targeting specific cell types in glioblastoma has proven difficult due to their heterogeneous nature. Although the WHO classifies glioblastoma as a grade IV astrocytoma, GBM likely arises from a combination of both neural stem cells (NSCs) in the subventricular zone (SVZ) and oligodendrocyte progenitor cells (OPCs) that differentiate from those NSCs [5,6,7,8]. Further phenotypic blurring occurs when oncogenic gene expressions lead to an astrocytic or oligodendroglial appearance [9]. In Vivo, OPC-like cells are enriched at the periphery of GBM tumors. These cells and microglia serve as a source of growth factors that give GBM cells more stemness, facilitate growth and proliferation, and create a supportive tumor microenvironment [10]. In addition, GBM cells in the border niche have upregulated levels of transcripts for drug efflux transporters and radioresistance [10], which may underlie resistance to TMZ treatment during radiation [10]. Thus, the proliferative capability of these border areas and resistance to drug therapy and radiation likely explain the high reoccurrence of GBM after tumor resection and chemotherapy. Treatment strategies that target the proliferative capability of GBM and bypass typical drug efflux transporters could potentially increase the survivability of this fatal disease.

Differentiation as a treatment strategy for GBM has recently been explored by several labs. Instead of killing the cancer cells, this method attempts to remediate the cancer cells so that they no longer proliferate or invade [11,12,13]. Studies have successfully used all-trans-retinoic acid (ATRA), molecules that upregulate cAMP production, bone morphogenic protein 4 (BMP4), and most recently Pam3CSK4, a synthetic Toll-like receptor 2 (TLR2) ligand to differentiate GBM cells [14,15,16,17,18,19]. However, retinoid treatment can be prone to resistance effects, and some of the other methods suffer from off-target effects that can require invasive targeted treatment strategies and limit in vivo use [20].

Alongside canonical biochemical factors and biomechanical forces, cell and tissue-level order are now known to be regulated by bioelectrical signaling [12,21]. The transmembrane voltage (Vmem) of cells is regulated by ion channels, which are not only important targets in embryonic channelopathies [22,23,24] but also increasingly seen as cancer targets [22,25,26,27,28,29,30,31,32,33,34,35,36]. Overall, adult non-proliferative cells tend to have hyperpolarized membrane potentials (around −40 mV to −90 mV), whereas stem cells, embryonic cells, and other highly proliferative cells are much more depolarized (around −5 mV to −40 mV) [22,37]. Cancer cell Vmem tends to have much more depolarized membrane potentials than their non-cancerous equivalents [24,37]. Indeed, transformed cells can be detected in vivo in animal models using voltage-sensitive fluorescent dyes based on their abnormal bioelectric signature [38]. Importantly, the resting membrane potential is not only a marker but is functionally instructive for cell behavior. Classic work by Cone showed that Vmem is an important regulator of proliferation in terminally differentiated cells or cancer [39,40,41], and this has been confirmed in recent work linking depolarization with a plastic, undifferentiated, highly proliferative state. For example, treatment of mesenchymal stem cells with depolarizing drugs inhibited their differentiation into adipocytes or osteoblasts, suggesting that hyperpolarization is required for differentiation [42,43,44,45,46]. Importantly, in vivo experiments in Xenopus tadpoles showed that co-expression of hyperpolarizing ion channels, or optogenetic activation of channels with human oncogenes, can prevent the tumors induced by oncogene expression alone [34,38,47,48].

In mammalian models, several recent studies have shown that ion channelopathies that affect resting membrane potential are present in many cancers and play a key role in cell proliferation, progression through the cell cycle, and metastasis [25,29,31,36,49,50,51,52,53,54,55,56].

The location of many ion channels in the plasma membrane also makes them attractive therapeutics targets. Agonists or antagonists used to modulate Vmem can be chosen for binding sites on the outer side of the membrane, bypassing the challenges posed by increased drug efflux transporters in glioblastoma stem cells (GSCs). Furthermore, ion channel transcripts have been shown to be upregulated in GSCs, including SCN8A, which encodes a sodium channel, KCNB1, which encodes a voltage-gated potassium channel, and GRIA3, which encodes an ionotropic glutamate receptor that is non-selective for monovalent cations [53]. Other studies have found multiple changes in a variety of potassium channels in GBM, including an increase in BK channel expression, a decrease in hERG1, and a decrease in Kir4.1 [57,58]. In addition, intracellular alkalization induced by dysfunction in proton transportation has been shown to increase drug resistance in GBM and contributes to extracellular acidification, which provides GSCs with an optimized niche and facilitates the release of large oncosomes, a type of extracellular vesicle that can transform neighboring cells [59,60,61]. Thus, due to the location and importance of these channels in cancer cell proliferation, migration, and metastasis, ion channels provide an optimal target for the development of new GBM treatments.

A large catalogue of FDA approved pharmaceuticals that modulate ion channels exist, creating an opportunity to directly assess the role of Vmem in cancer progression. To discover effective interventions for glioblastoma using known small molecule modulators of the bioelectric state, we first used the cell line, NG108-15, to screen for Vmem-modifying compounds that could potentially promote differentiation in glioblastoma cells [62]. NG108-15 cells are a hybrid formed from mouse N18TG2 neuroblastoma cells with rat C6-BU-1 glioma cells and is a popular model system in neuronal differentiation studies and shows cancer stem cell characteristics. This line was chosen due to its extensive use in determining the factors necessary for neuronal differentiation, its close resemblance to NSCs and GSCs, and its well-characterized membrane properties and electrophysiology [63,64,65]. Our goal was to find compounds that would safely abrogate the proliferative potential in NG108-15 cells under high serum conditions, normally prohibitive for differentiation and representative of the high plasma protein content found in vivo, revealing robust candidates for subsequent assessment in the U87 (ATCC) human glioblastoma line, and for future testing in patient-derived GSCs [66,67,68,69].

To screen candidate drugs for anti-cancer activity, we first created new cell lines as resources for future studies in this field: an NG108-15 and U87 stable line expressing a Fluorescence Ubiquitin Cell Cycle Indicator (FUCCI) [70] and a cell membrane mTurquoise tag for revealing whole-cell morphology. Automated electrophysiology was used to determine the overall plasma membrane potential in NG108-15 cells in the presence of the drug compound treatment. For U87 cells, we used dyes to determine cytoplasmic calcium levels, internal pH (pHi), transmembrane potential and lysosomal pH to gain comprehensive profiling of the drugs’ effects on cell physiology. Finally, a beta-galactosidase stain and immunocytochemistry were used to analyze senescence, cell cycle inhibitor levels, and differentiation status of treated NG108-15 and U87 cells. In addition, we used human neurons to perform a Live/Dead assay and a senescence assay to identify treatments that were toxic to non-cancerous cells. We identified several combinations of compounds that strongly and specifically reduce proliferation and promote induction of differentiation, demonstrating that this class of electroceuticals, most of which are already FDA approved for human use in other conditions, are good candidates for cancer remediation in GBM.

Stocks of Rapamycin (R8781, Sigma, St. Louis, MO, USA), Retigabine (SML0325, Sigma), Minoxidil (M4145, Sigma), NS1643 (sc-2041353, Santa Cruz Biotechnology, Dallas, TX, USA), Lamotrigine (L3791, Sigma), Zolmitriptan (SML0248, Sigma), Cariporide (SML1360, Sigma), Topiramate (T0575, Sigma), Pantoprazole Sodium Hydrate (P0021, Sigma), Fenofibrate (F6020, Sigma), Acetazolamide (A6011, Sigma), Quercetin (Q4951, Sigma), Temozolomide (2706, Tocris, Bristol, UK), Dexamethasone (D4902, Sigma), ONO-RS-082 (O0766, Sigma), Topotecan hydrochloride (4562, Tocris), CKD 602 (5125, Tocris), (Z)-4-hydroxytamoxifen (3412, Tocris), Lansoprazole (2582, Tocris), Chlorzoxazone (C4397, Sigma), and Sodium Butyrate (B5887, Sigma), were made at 1000× concentration in DMSO. Stocks of Gabapentin (G154, Sigma) and Cisplatin (232120, EMD Millipore, Burlington, MA, USA) were made in water at 1000×. Dibutyryl cAMP sodium salt (D0627, Sigma) was the only compound dissolved directly in cell culture media at 1 mM concentration.

NG108-15 cells (ATCC), passage 7–10 were cultured in growth media containing DMEM medium with high glucose and no phenol red or sodium pyruvate (31053-028, ThermoFisher, Waltham, MA, USA), which was supplemented with 2 mM Glutamax, 10% FBS, HAT supplement, and 100 U/mL penicillin/streptomycin. Cells were disassociated with Accutase and passaged when 70% confluent and maintained at 37 °C with 5% CO2. The media were changed every two days. For live cell assays, plates were coated with 1/50 dilution of no phenol red growth factor reduced (GFR) Matrigel (356231, Corning, NY, USA) in imaging media made of Fluorobrite DMEM (A1896701, ThermoFisher) with no added supplements. Cell culture medium was switched to Fluorobrite DMEM with 2 mM Glutamax, 10% FBS, HAT supplement, and 100 U/mL penicillin/streptomycin for all live cell imaging experiments, except for a cAMP with rapamycin treatment which was conducted in the exact same media except with 1% FBS. All drug screens, including controls, were performed using 0.02% DMSO in the high serum Fluorobrite DMEM.

U87 (ATCC) cells were cultured under the same conditions as the NG108-15 cells minus the HAT supplement and used at passage 7–10 for experiments. All cells plated for antibody staining and senescence assays were cultured on polyethylenimine (181978, Sigma) coated plates at 25 µg/mL in 150 mM NaCl solution for 1 h at room temperature followed by a rinse of PBS and an additional coating of 1/50 Matrigel as described above.

Human neuronal cells were differentiated from human induced neural stem cells (hiNSCs) (passage 7–10), a generous gift from David Kaplan, as described previously [71]. Briefly, hiNSCs were grown on mouse embryonic fibroblast (MEF) feeder cells until ready to differentiate into neurons. Cells were then seeded on Poly-D-Lysine (A3890401, ThermoFisher) and laminin (L2020, Sigma) coated 96-well plates at a density of 128,000 cells/mL. Cells were differentiated in Neurobasal (12348017, ThermoFisher) media supplemented with 2% B27 (17504044, ThermoFisher), 1% Glutamax, and 1% antibiotic-antimycotic for 7 days with media changes every 2 days. At the end of the 7 days the treatments were added in the differentiation media for 3 days.

The ES-FUCCI construct containing a hygromycin resistance cassette was subcloned using XmnI and SalI from the plasmid ES-FUCCI, which was a gift from Pierre Neveu (Addgene (Watertown, MA, USA) plasmid #62451; http://n2t.net/addgene:62451 (9 March 2022); RRID:Addgene_62451) [70]. The CAG pPalmitoyl-mTurquoise2 construct was subcloned from the plasmid pPalmitoyl-mTurquoise2 using BamHI and NotI, which was a gift from Dorus Gadella (Addgene plasmid #36209; http://n2t.net/addgene:36209 (9 March 2022); RRID:Addgene_36209) [72]. All subcloned fragments were cloned into a pENTR1A plasmid with a CAG promoter and multiple cloning site (MCS) followed by an SV40 poly(A) using the same sites as were used in excising the fragment from the parent plasmid. In the case of ES-FUCCI, the CAG promoter, MCS, and poly(A) were removed from the pENTR1A plasmid with SpeI, which was blunted, and SalI prior to ligation with the fragment. The resulting pENTR1A ES-FUCCI was then Gateway LR clonased (11791020, ThermoFisher) into the hyperactive piggyBac transposase-based, helper-independent, and self-inactivating delivery system, pmhyGENIE-3, a gift from Stefan Moisyadi [73,74]. The pENTR1A CAG pPalmitoyl-mTurquoise2 was cloned into a pmhyGENIE-3 containing a neomycin resistance gene in the backbone. The resulting plasmids, HypG3 Hygro ES-FUCCI and HypG3 NeoBB CAG pPalmitoyl-mTurquoise2 were used for subsequent transfections.

All transgenic cell lines were made by transfecting cells at 30% confluence with 500 ng of appropriate HypG3 plasmid via 1 µL of lipofectamine 3000 (L3000008, ThermoFisher) per well of 24-well plate containing 500 µL of culture medium. The reagent was removed after 24 h, and fresh culture medium was added. Cells were allowed to recover for 24 h prior to selection with 1000 µg/mL G418 or 200 µg/mL hygromycin. After selection, cells were serially diluted into 96-well plates, and single colony clones were expanded. Clones showing robust growth and strong expression were chosen for subsequent experiments.

Both cell lines were plated at 5000 cells per mL into black-walled flat-bottom 96-well plates coated with 1/50 dilution of Matrigel. Images of cells were taken on Day 0 and every subsequent day for 10 days using a Zeiss Axio1 fitted with an on-stage incubator and kept at 37 °C with 5% CO2. A 5× objective was used along with filters for YFP, RFP, and CFP, and careful calibration of each plate was undertaken to ensure that the same fields of view were imaged each day. Drugs were added after the initial images were taken and then changed every two days until day 6. For recovery experiments, after imaging on day 6, all drugs were removed, and cells were put into imaging media without any added drugs. This media was changed every two days until day 10. Analysis of growth and snap shots of FUCCI reporters over the 6 days or 10 days was conducted using a CellProfiler [75] pipeline designed by the Broad. The fractional difference in cell number was calculated by dividing the total number of cells each day by the total number of cells from day 0. The ratio of cells in each cell cycle stage every day was determined by counting all the nuclei of one color and dividing it by the total number of nuclei. Blind quality control was performed on all the images to make sure that any fibers or image artifacts were not incorrectly counted by the program.

NG108-15 cells were plated on PEI and Matrigel-coated plates at 15,000 or 30,000 cells/mL, U87 cells at 10,000 cells/mL and grown in imaging media for 6 days with drug treatment. Fresh media was added every 2 days. On day 6, cells were fixed with 4% formaldehyde in PBS for 30 min for all antibody staining, except for BrdU which was fixed for 15 min, and for the senescence assay, which was fixed for 20 min using the Senescence Beta-Galactosidase Staining Kit (9860, CellSignaling, Danvers, MA, USA) and thereafter stained according to manufacturer’s protocol. After fixing, cells were washed twice with PBS and were permeabilized in PBS with 0.3% Triton-X for 15 min, then washed again twice with PBS. The cells for the BrdU assay were additionally treated with 1N HCl for 10 min on ice, followed by treatment with 2N HCl for 50 min at room temp. Cells were then blocked with standard blocking buffer comprised of TBS containing 10% goat serum, 0.1% BSA, and with 0.05% Tween 20 or without detergent for antibodies against phosphorylated proteins or monoclonal antibodies except for Anti-YAP and Anti-BrdU staining, which were followed according to manufacturer’s protocol. Primary antibodies were added at the following concentrations: 1:250 anti-S100 beta (GTX129573, Genetex, Irvine, CA, USA), 1:500 anti-GFAP (AB5804, EMD-Millipore, Burlington, MA, USA), 1:300 anti-SOX10 (ab155279, Abcam, Cambridge, UK), 1:100 MAP2 (4542, CellSignaling), 1:400 anti-cleaved caspase 3 (9661, CellSignaling), 1:300 anti-Cx43 (STJ2411, St.

John’s Laboratory, London, UK), 1:800 anti-p27Kip1 (3698, CellSignaling), 1:150 anti-BrdU (5292, CellSignaling), 1:125 anti-LC3B (2775, CellSignaling), 8 µg/mL anti-O4 (MAB1326, R&D Systems, Minneapolis, MN, USA), 1:1000 anti-TH (RPCA-TH, EnCor, Gainesville, FL, USA), 1:4000 anti-NSE (RPCA-NSE, EnCor), 1:4000 anti-NFM (RPCA-NF-M, EnCor), 0.5 µg/mL anti-TujI (801202, BioLegend, San Diego, CA, USA), 1:400 anti-NF-KB p65 (8242, CellSignaling), 1:125 anti-phospho CREB (05-807, EMD-Millipore), 1:250 anti-vimentin (ab92547, Abcam), 1:125 anti-YAP (14074, CellSignaling). Antibodies were diluted in their respective blocking buffer and incubated overnight at 4 °C. The next day, cells were washed with TBS-T or TBS for 3× for 5 min each, and secondary antibody was added as follows: 1:1000 donkey anti-mouse 647 (A-31571, ThermoFisher) or 1:1000 donkey anti-rabbit 647 (A-31573, ThermoFisher). Each was diluted in blocking buffer along with 2.5 µg/mL Hoechst 33342 and incubated for 1 h at room temperature. Cells were then washed again with TBS-T or TBS for 3× for 5 min each and covered with Gelvatol. Cells were imaged with an EVOS M7000 system at 20X, with at least 50% of each well in a 96-well plate imaged and analyzed using a CellProfiler pipeline for measuring integrated intensity, mean intensity, or nuclear to cytoplasmic ratio of mean intensities [76]. LC3B stained cells were marked as positive for autophagy only if very bright puncta were detected.

For the Live/Dead assay, 1 µM Calcein Green AM and 0.5 µM ethidium homodimer-1 (L3224, ThermoFisher) were added to PBS along with 10 µg/mL of Hoechst to make the staining solution. Half the medium was removed from the human neuronal cells that had been incubated with treatments for 3 days and replaced with staining solution. This was conducted four times to make sure that cells did not detach. Cells were then incubated for 15 min at 37 °C and imaged. For resting membrane potential staining, DiBAC4(3) was used. U87 cells were seeded at 10,000 cells per mL on Matrigel-coated plates as described above. Cells were treated with drugs for 6 days with changes in media every other day. On day 6 the media was removed and washed once with dye buffer consisting of Hank’s Balanced Salt Solution (HBSS) with 20 mM HEPES at pH 7.4, and then the staining solution, consisting of dye buffer with 2 µM DiBAC4(3), was added to the cells and allowed to incubate for 30 min at 37 °C. The staining solution was then removed, and fresh staining solution containing the treatments were added to the cells. For cytoplasmic calcium staining, Fluo-4 AM was used. The staining solution consisted of a dye buffer with 4 µM Fluo-4 AM and a 1:1 ratio of Pluronic F-127 (20% in DMSO) and allowed to incubate for 30 min at room temperature. The staining solution was then removed, and fresh dye buffer was added and was allowed to incubate for 20 min at 37 °C. Then the dye buffer containing the treatments were added to the cells. For cytoplasmic pH staining, pHrodo Green was used. The staining solution consisted of a dye buffer with a 1:1000 dilution of the pHrodo Green stock and a 1:100 dilution of the PowerLoad concentrate (P35373, ThermoFisher). The cells were allowed to incubate for 30 min at 37 °C.

The staining solution was removed and washed once with dye buffer, then fresh dye buffer containing the treatments were added and allowed to incubate for 5 min at 37 °C. A calibration curve was made by instead adding the components of the intracellular pH calibration kit (P35379, ThermoFisher). For lysosomal pH staining, LysoSensor Green was used. The staining solution consisted of a dye buffer with a 1 µM dilution of LysoSensor Green DND-187 (L7535, ThermoFisher). The cells were incubated as above, and the staining solution was removed and replaced with a dye buffer containing 2.5 µg/mL Hoechst 33342 for 10 min. Hoechst was removed, and cells were washed with dye buffer one time before adding dye buffer containing the treatments. Cells were imaged using an EVOS M7000 system outfitted with a GFP filter cube for stains and a DAPI filter for Hoechst. Images were then analyzed using a CellProfiler pipeline.

The SyncroPatch 384PE platform (Nanion Technologies®, Munich, Germany) was used to perform the automated high-resolution whole-cell patch-clamp recordings in a 384-well plate format. The experiments were performed within one hour after the harvesting process. The assays were carried out in medium resistance single-hole chips (4–5 MΩ).

Cells were harvested 48–72 h after seeding 1–2 × 106 cells in T175 culture flasks (Falcon), then cells were rinsed with PBS (5 mL) and treated with 3 mL Accutase (STEMCELL Technologies, Vancouver, BC, Canada) for 5 min at 37 °C, re-suspended in 10 mL of serum-free media and pelleted at 160 g for 5 min at RT. The supernatant was discarded, and cells were re-suspended in serum-free DMEM medium with high glucose and no phenol red or sodium pyruvate (31053-028, ThermoFisher), and physiological extracellular solution (pECS) 50% (v:v). The cells were kept until the moment of the experiment in a temperature-controlled dedicated reservoir at 10 °C and shaken at 200 rpm as described [77].

The chip was primed with the following solutions (in mM), physiological extracellular solution (pECS) 10 HEPES, 140 NaCl, 5 Glucose, 4 KCl, 2 CaCl2, 1 MgCl2, 295–305 mOsm pH 7.4 (NaOH). Internal recording solution (in mM) 20 EGTA, 50 KCl, 10 NaCl, 60 KF, 10 HEPES at pH 7.2, and 285 mOsm. 15 µL of the cell suspension (50% v/v pECS/Medium no serum) was added to each well to a final density of 50–80 K cells/mL. Cell capture was promoted by holding negative pressure of −100 mbar for 20 s. After the capture, successive hyperpolarization steps from −30 mV to −100 mV were applied to foster the electrical seal formation. The seal was enhanced by transient addition of a high Ca2+ extracellular solution (80 mM NaCl, 3 mM KCl, 10 mM CaCl2, 10 mM MgCl2, 10 mM HEPES at pH 7.4 and 298 mOsm. High Ca2+ solution was washed out by successive external exchanges replacing half of the volume of the well each time with the external recording solution. All recording solutions were prepared with ultrapure MilliQ water (18 MΩ-cm).

After the formation of the Giga-seal, a −250 mbar pressure was applied to break the membrane patch. Once in whole-cell configuration, the resting membrane potential was monitored in current-clamp configuration “I = 0”, the values shown correspond to the average of seven 1 s length sweeps, with an inter-sweep time of 10 s. Cells were also recorded in voltage-clamp mode to characterize the different voltage-dependent conductance present in NG108-15 cells.

Compounds in DMSO stocks were dissolved 1–2 h before the experiment in pECS to a final concentration of 2× the working concentration. 40 µL of the 2× compound solution was applied to each well to a final volume of 80 µL. Before the compound addition, a 5 min recording in 0.2% DMSO was used as a baseline to normalize the changes in the Vmem.

NG108-15 cells containing the FUCCI cell cycle reporter and palmitoyl-mTurquoise2 fluorescent membrane tag were incubated with compounds by themselves and in combination, in media containing high serum (known to be prohibitive for differentiation in this line), to test for their ability to suppress proliferation (Supplementary Figures S1 and S2). The compounds chosen for testing included compounds known to alter the membrane potential of the cell by decreasing proton efflux [78], increasing potassium efflux [79], or decreasing sodium influx [80], and others chosen for their potential combinatorial effects with the ion modulating drugs, including cell cycle-specific disruptors and autophagy-inducing compounds (Supplementary Table S1). In addition, we also tested the most clinically relevant compound currently used in the standard treatment of glioblastoma, temozolomide (TMZ). A subset of the compounds exhibiting the highest efficacy in the initial screens are shown in Table 1.

Many of the tested compounds showed a significant difference in proliferation from the controls after 6 days of treatment, but only a few of these were explored in more detail due to their novelty (Table 1). Figure 1 shows the best of the individual and combination treatments chosen when compared to control at day 6. Note we did not include chlorzoxazone combinations in the NG108-15 analysis shown, due to its poor performance in the U87 cell line. The stacked bar graphs show the ratio of cells in different parts of the cell cycle as indicated by the FUCCI cell cycle reporter.

The proliferation of NG108-15 cells treated with NS1643, a human Ether-a-go-go (hERG) activator and KCNQ2, KCNQ4, KCNQ2/3 potassium channel potentiator [85] at 20 µM and 50 µM alone were able to lower proliferation significantly about 1.6- and 2.4- fold decrease compared to control respectively (Figure 1A). In fact, NS1643 at 20 µM was very effective at lowing cell proliferation when combined with pantoprazole, a proton pump inhibitor known to inhibit the expression of the vacuolar-ATPases (V-ATPases) in human gastric adenocarcinoma [86]; this combination worked significantly better than pantoprazole or NS1643 alone, with a fold decrease of 7.2. NS1643 at 50 µM also significantly lowered proliferation when combined with rapamycin (autophagy inducer), with a 4.6-fold decrease and worked better than rapamycin or NS1643 alone. Retigabine, which opens the voltage-activated potassium channels KCNQ2-5 [87,88], significantly lowered cell proliferation by about 1.6-fold decrease compared to control, but its combination with rapamycin or pantoprazole worked better than any of those compounds alone with a 2.9 and 6-fold decrease as compared to control, respectively. Pantoprazole at 100 µM, however, was the most effective compound alone (5.1-fold decrease) or in combination with lamotrigine which blocks voltage-gated sodium channels [89], or NS1643, or rapamycin, with fold decreases compared to control of 6.8, 7.2, and 9.3 respectively. These three combinations worked better than one of our positive controls, which consisted of a treatment of 1 mM cAMP with rapamycin at 200 nM in full serum media (5.1-fold decrease). This same positive control treatment is known to terminally differentiate these cells when in low serum [90], which we confirmed. However, the use of cAMP is problematic clinically due to its many off-target effects [91].

The cell cycle data in Figure 1B reveal that pantoprazole increases the proportion of cells in early S and that its combinations can also increase the proportion of cells in G1. Rapamycin treatment alone increased the proportion of cells in G1, while NS1643 treatment did not seem to affect the cell cycle proportion. It is worthwhile to note that these compounds were effective on NG108-15 cells, while temozolomide (TMZ), the standard glioblastoma treatment, was not, a situation found in many GBM cases [92].