Mutual Influence of ROS, pH, and CLIC1 Membrane Protein in the Regulation of G₁–S Phase Progression in Human Glioblastoma Stem Cells

Gliomas are the major group of primary malignant brain tumors. Among them, glioblastoma (GB, WHO grade IV) is the most aggressive and lethal (1, 2). To date, no effective therapies for GB have been developed. Surgical removal of GB is often ineffective due to its fast spread and high rate of infiltration in brain parenchyma. Chemotherapy is also problematic due to the presence of the blood–brain barrier and the high expression of antiapoptotic proteins and drug efflux transporters on cancer cells. Current pharmacologic treatments for GB target highly proliferating but nontumorigenic cancer bulk cells but spare the tumorigenic cancer stem cells (CSC), which, being refractory to therapy, determine the poor prognosis of GB patients. Although usually quiescent, CSCs are the reservoir for a potential tumor recurrence (3, 4). Isolated CSCs, which contain all the necessary genetic information to faithfully reproduce the original tumor, can then be used as a model to look for new diagnostic markers and screen potential drugs. In addition, therapies directed specifically against CSCs could offer better clinical outcomes (4). The greatest challenge of GB treatment is likely connected to the need to preferentially target CSCs.

We have previously demonstrated the pivotal role of CLIC1 protein in the tumorigenic potential of CSCs isolated from human GBs, and showed that CLIC1 functional expression in cell membranes is essential for the proliferation and self-renewal of GB CSCs (5, 6). CLIC1 was cloned from a human monocytic cell line (7). First identified on nuclear membranes, it was then also found on plasma membranes (7, 8). This protein is highly conserved among vertebrates, ubiquitously expressed in different tissues, and overexpressed in tumors (9, 10). CLIC1 has been characterized as a metamorphic protein (11), coexisting as both a soluble cytoplasmic form and a membrane-associated element. The transition between its hydrophilic and hydrophobic localizations is promoted by several factors. In vitro studies on an artificial lipid bilayer and isolated cells have demonstrated that changes in both intracellular pH (pHi) and oxidative state are responsible for conveying CLIC1 protein to the membrane (9, 12–14). Once in the membrane, the protein is associated with increased membrane permeability, mainly to chloride ions. It is still controversial whether CLIC1 is the protein that actually forms the ionic pathways or whether it modulates an existing population of ion channels. Supporting the first hypothesis, previous experiments on different cell types using ionic channel blockers (15), siRNA (5, 6), and point mutation (16) have demonstrated a clear link between CLIC1 presence in the lipid bilayer and an increased transmembrane chloride current. However, in both cases, the activated protein functions to increase ionic membrane flux (17). CLIC1 protein downregulation and direct blockade of CLIC1-activated membrane current (5, 6) result in CSC accumulation in the G₁ phase and an elongation of CSC cell cycling time (5, 6).

Over the last two decades, researchers have defined an important role for ion channels in tumor development and growth, and they are now novel targets for cancer therapy. Channel dysfunctions may have a strong effect on cell physiology and signaling, and it has become clear that abnormal triggering of ion channel activity could support the high proliferation rate, mobility, and invasiveness of tumor cells. Ion channels expression is often altered in tumor cells. During the change from a physiological toward a neoplastic state, like several other proteins, most ionic permeabilities result overexpressed or altered in their function (9, 18, 19). The main impediment to using ionic channels as therapeutic targets is their diffuse presence in the cell membrane, both in physiologic and pathologic conditions. This is not the case for CLIC1-associated ionic current (10). The chloride membrane current in GB CSCs, as well as in several tumor cell lines, can be downregulated by CLIC1 siRNA and blocked by IAA94, a specific inhibitor of CLIC1-induced chloride current (5, 6). Conversely, there is no measurable current in differentiated GB CSCs or umbilical cord mesenchymal stem cells (MSC), although CLIC1 protein is abundant in MSC cytoplasm (5). Furthermore, the abundance of CLIC1 protein and its associated membrane current density is directly proportional to the tumor's aggressiveness (6). CLIC1 protein's cytoplasmic role, if any, is still unclear and requires more investigation. As previously reported, CLIC1 translocation from the cytosolic aqueous phase to the membrane lipid bilayer is modulated by pH changes and fluctuating oxidative levels within the internal face compartment (9, 14). In general, cellular metabolic activity rises as internal pH increases. Oscillations in pHi contribute to the control of cell-cycle progression and cell division in both prokaryotic and eukaryotic cells. A rapid alkalinization of pHi may be important for G₁–S phase progression (18). High pHi, associated with a more acidic extracellular space, is involved in the processes of neoplastic transformation and tumor development (19, 20). Oxidative-level oscillations in the intracellular compartment also contribute to the regulation of cell-cycle progression (21, 22). Menon and colleagues have shown that a transient increase in pro-oxidant levels early in G₁ is required for the cells to enter S phase (23). This periodicity of intracellular redox state requires a delicate balance between ROS production and its subsequent removal by antioxidants (24, 25). Therefore, ROS overproduction and/or the inhibition of antioxidant efficiency could perturb the redox cycle, which in turn could lead to aberrant proliferation.

Alterations in cells' basal levels of oxidation typically contribute to many tumorigenic processes (22, 26, 27). Thus, it is not surprising that CLIC1 channel activity, induced by oxidation, is higher in hyperproliferating tumor cells. As we previously demonstrated in activated microglia, cancer cells could take advantage of a feed-forward mechanism between CLIC1 channel activity and ROS production by NADPH oxidase (12).

In this study, we highlight the crucial role of CLIC1 membrane function in GBM CSCs during cell-cycle progression. Our results demonstrate that CLIC1 protein in the plasma membrane of GB CSCs is instrumental in stabilizing, in a time-dependent manner, the hyperactivated metabolic state defined as “allostasis” (28). Therefore, inhibiting CLIC1 function should help reestablish the homeostatic condition typical of quiescent cells. The uniqueness of CLIC1 functional expression in CSC membranes makes it a promising potential pharmacologic target for GB treatment.

Indanyloxyacetic acid 94 (IAA94), diphenyleneiodonium chloride (DPI), 3-amino-6-chloro-N-(diaminomethylene)-5-(ethyl(isopropyl)amino)pyrazine-2-carboxamide (EIPA), and N-Formyl-L-methionyl-L-leucyl-L-phenylalanine (FMLP) were purchased from Sigma Aldrich.

Antibodies used were mouse monoclonal anti-human CLIC1 (Santa Cruz Biotechnology, sc 81873), rabbit monoclonal anti-cyclin D1 (Cell Signaling Technology, 92G2, 1:1,000), rabbit monoclonal anti-p27 Kip1 (Cell Signaling Technology, D37H1, 1:1,000), anti-mouse Alexa Fluor 488 (Life Technologies), and secondary antibody-HRP conjugated (Sigma-Aldrich). In some experiments, we used a custom-made antibody against the last 30 amino acids of the amino terminus (AB-NH₂) of CLIC1 (6, 8, 15).

We obtained a CSC culture from a 48-year-old GB patient (coded as GB3). We also used two other CSC preparations from different patients (GB23 and GB19) to reproduce key experiments. CSCs, isolated as described in previous work (29, 30), were grown in stem cell–permissive medium enriched with 10 ng/mL human bFGF and 20 ng/mL human EGF (31, 32). We validated the CSC properties as described (32), in particular, tumorigenicity was confirmed by orthotropic xenograft in NOD-SCID mice after Policlinico San Martino-IST Institutional Animal Care and Use Committee approval. CSCs, grown as a monolayer on Matrigel, were synchronized in G₁ phase of the cell cycle through 60 hours of starvation in growth factor–free medium. Cells were released from the G₁ blockade by changing to complete stem cell–permissive medium. Human umbilical cord MSCs were obtained and characterized following the minimal criteria proposed by the International Society for Cell Therapy, as previously described (6). For CSC and MSC isolation, we collected samples after the Institutional Ethical Committee approved the patients' written-informed consent.

GB CSCs or MSCs were plated in a 24-well plate (2 × 10⁴/well) and treated for 72 to 96 hours with different inhibitors. At the end of the incubation time, we collected and counted the cells using the Trypan Blue Dye exclusion test (Thermo Fisher Scientific) and the Countess II FL Automated Cell Counter (Thermo Fisher Scientific).

shRNA sequence targeting human CLIC1 (target sequence: 5′-GATGATGAGGAGATCGAGCTC-3′) was cloned using EcoRI and AgeI sites in pLKO.1-TCR cloning vector (Addgene). As control SCRAMBLE, shRNA cloned in pLKO.1 vectors was used (Addgene). Lentiviral vectors were produced by cotransfection of pLKO and packaging plasmids psPAX2 (Addgene) and pMD2.G (Addgene) into HEK293T. Virus was harvested at 60 hours after transfection, and infections were carried out to stably express constructs in GB CSCs.

Patch-clamp electrophysiology was performed in a perforated-patch whole-cell configuration as previously reported (5). The voltage protocol consisted of 800 ms pulses from −40 mV to +60 mV (20 mV voltage steps). The holding potential was set according to the resting potential of the single cell (between 40 and −80 mV); we applied a 15 ms prepulse of −40 mV before starting the voltage steps.

CLIC1-mediated chloride currents were isolated from the cells' other ionic currents by perfusing the specific inhibitor IAA94 (100 μmol/L). Patch-clamp solutions (22) included the following: bath (mmol/L): 125 NaCl, 5.5 KCl, 24 HEPES, 1 MgCl₂, 0.5 CaCl2, 5 D-glucose, 10 NaOH, pH 7.4; pipette (mmol/L): 135 KCl, 10 HEPES, 10 NaCl, 1 MgCl₂, pH 7.2. In the experiments shown in Fig. 7, the bath solution pH was changed at the value reported in the figure legend.

Cells (2 × 10⁵ cells each) were collected by directly scraping with a lysis buffer (LB; in water: 0.25 mol/L Tris-HCl pH 6.8, 4% SDS, 20% glycerol). Complete lysis of all cells was obtained by 10 minutes of incubation in LB at 95°C. Concentration of protein lysates was assessed by BCA assay (Pierce). We added LDS Sample Buffer 4X (Thermo Fisher Scientific) to samples consisting of equal amounts of protein and loaded each sample onto a 12% SDS-PAGE under reducing conditions. Resolved proteins were electroblotted onto a nitrocellulose membrane (Bio-Rad Laboratories) and probed with specific antibodies. Western blot data were normalized on housekeeping protein (tubulin) and plotted.

CSC monolayers of 5 × 10³ cells, seeded on 12-mm diameter cover glass, were fixed with 2% paraformaldehyde and incubated with specific antibodies. To visualize the nuclei, we treated the cells with DAPI and mounted them on microscope slides using a glycerol-based mounting agent. Samples were observed under a confocal microscope (Leica TCS SP2) with a Leica HCX PL APO 63X/1.4NA oil immersion objective for Fig. 2. Images were analyzed using ImageJ software.

We measured pHi using the probe carboxy SNARF-1 AM (carboxy SemiNaphthoRhodaFluor acetoxymethyl ester, Life Technologies), calibrated with high-potassium buffers and nigericin as described in Chow and Hedley's protocol (33). Briefly, CSCs were harvested, centrifuged for 10 minutes at 600 rpm, and resuspended in 1X PBS (1 × 10⁶ cells/mL PBS). Carboxy SNARF-1 AM was added at the final concentration of 5 μmol/L, and cells were incubated for 30 minutes at 37°C in the dark. Samples were centrifuged at room temperature, resuspended in 500 μL of the desired solution, incubated for 20 minutes at room temperature in the dark, and then analyzed by flow cytometry with a FACS Aria (BD Biosciences) to determine the 640/580 fluorescence ratio. The solutions used included: samples (in mmol/L) 125 NaCl, 5.5 KCl, 24 HEPES, 1 MgCl₂, 0.5 CaCl₂, 5 glucose, 10 NaOH, pH 7.4 ± desired treatment; standard high-potassium buffers (pH 6.8–7.0–7.2–7.4–7.6–7.8) + nigericin (final concentration 2 μg/mL).

To measure the redox conditions of the cells, we performed experiments using 2′,7′- dichlorofluorescin diacetate (DCFH-DA, Sigma). CSCs were plated as a monolayer and incubated with DCFHDA, which passively diffuses into cells as a nonfluorescent probe. DCFH-DA is de-esterified by intracellular esterase and becomes a highly fluorescent adduct that is trapped inside the cell upon oxidation, which allows for rapid quantification of oxygen-reactive species in response to oxidative metabolism.

After 24 hours of treatment, cells were incubated with Hank's Buffered Salt Solution (HBSS) with added DCFH-DA (final concentration 20 μmol/L) for 30 minutes at 37° in the dark. For short-term treatments, the compounds were added right before measurements in fresh HBSS (0.137 mol/L NaCl, 5.4 mmol/L KCl, 0.25 mmol/L Na₂HPO₄, 0.1 g/L glucose, 0.44 mmol/L KH₂PO₄, 1.3 mmol/L CaCl₂, 1.0 mmol/L MgSO₄, and 4.2 mmol/L NaHCO₃).

Cells were imaged in vitro using a Nikon Ti-E inverted fluorescence microscope (Nikon) with a CFI Plan Apo VC 20X objective. Excitation light was produced by a Prior Lumen 200 PRO fluorescent lamp (Prior Scientific) at 470/40 nm. Images were collected with a Hamamatsu Dual CCD Camera ORCA-D2 (Hamamatsu Photonics), and emissions were gathered with a 505–530 nm bandpass filter. Exposure time was 100 ms with no binning for the excitation wavelength, and images were acquired every minute for 31 minutes. Filters and a dichroic mirror were purchased from Chroma (Chroma Technology). The NISElement (Nikon) was used as platform to control the microscope, illuminator, camera, and postacquisition analyses.

It is important to note that the data reported in this article are not an absolute value of the cells cytoplasmic oxidation state. They are rather a measurement of the ROS production/ROS elimination slope detected at any given time and consequently they give an idea of the general oxidative level of the cells.

Cells were washed in PBS, fixed with ice-cold 95% ethanol at 4°C for 1 hour, and resuspended in staining solution (PBS, 20 mg/mL RNAse A, 50 μg/mL propidium iodide, 0.5% Triton X-100, all from Sigma-Aldrich). DNA content was quantified by FACScalibur (BD Bioscience) as reported (34). Cell-cycle profile was determined by analysis of Listmode data using ModFit LT software (Verity Software House). At least 10,000 events were collected gating single nuclei and excluding aggregates.

For proliferation experiments, we plated CSCs in a 6-well plate (3 × 10⁵ cells/well) as monolayer. Time lapses were performed using a ScanR system (Olympus) based on either an IX81 inverted microscope equipped with a Hamamatsu ORCA-Flash4 camera and driven by CellSens software (Olympus), or an Eclipse TE200-E microscope (Nikon) equipped with a Cascade II-512 camera (Photometrics) driven by Metamorph software.

During the entire observation process, cells were kept in a microscope stage incubator (Okolab) for environmental control. We analyzed the cells using a 20x LUCPlanFLN (NA 0.45) or a 10X PlanFluor (NA 0.30) objective. Images were acquired every 15 minutes for 36 hours and were analyzed using Fiji software (35).

We used TIRF microscopy in different experiments to visualize the spatial-temporal dynamics of CLIC1 near the cell surface. Using a Leica AM TIRF microscope and an Andor iXon DU-885 camera, we obtained time-lapse video with images taken every 15 minutes for 3 or 7 hours with a HCX PL APO 63X/1.47NA objective. CSCs were transfected with a plasmid carrying the CLIC1-GFP sequence, seeded in adhesion on circular cover glasses (Ø = 25 mm), and then synchronized in G₁ phase. After 36 hours, cells were released in complete medium or in external solution at pH = 8 (see Bath Solution). Emitted fluorescence was recorded with a Filtercube GFP-T ET TIRF MC filter for laser line 488 (ex-dic-em: 475/40-500-530/50).

In ROS and TIRF experiments, we determined fluorescence intensities for every image over regions of interest corresponding to the cells analyzed by subtracting the specific background and normalizing the values to the first one detected for each cell. All analyses were performed using Fiji software (35).

We repeated all experiments at least twice; quantitative data were collected from experiments performed in triplicate or quadruplicate and expressed as mean ± SEM, except for time-lapse experiments (mean ± SD). CLIC1-mediated current density/voltage relationships were analyzed as following: for each condition, we plotted every experiment and extrapolated the linear regression for the curve (all significant, with R² ≥ 0.9). For analyses of two conditions, we used a two-sample t test; otherwise, we performed one-way ANOVA analysis on the slopes of every experimental group.

In the box chart plots, the solid lines within the boxes represent the median values. The square within the boxes is the average, and the boxes show the 25th and 75th percentile range of the measured elements. The maximum and minimum values are depicted as horizontal bars, and the symbols to the right of the box are the numbers of trials.

After 60 hours of starvation, human GB CSCs were synchronized and accumulated (more than 90%) in G₁ phase of the cell cycle. Synchronized cells released in complete medium progressed to the S phase (5), and the small percentage of cells (∼10%) already in late S and G₂ phase underwent their first division after 8 to 12 hours. CLIC1 contribution to cell-cycle progression is already visible for these cells. GB CSCs incubated with IAA94, a CLIC1 ionic conductance inhibitor, displayed a longer transition time before the G₂–M phase. Figure 1A shows the results of a time-lapse observation of CSC cell-cycle dynamics confirming the effect of CLIC1-membrane protein inhibitor. The same results were obtained by incubating CSCs with a custom-made antibody directed against the extracellular portion of the protein, which is known to inhibit CLIC1 function at the membrane level (6, 8, 15).

According with cell-cycle progression measurements, CLIC1 inhibition has a most pronounced effect on G₁–S transition time. Flow cytometry analysis of DNA content distribution shows that CLIC1 inhibition by IAA94 treatment delays the transition of GB CSCs from G₁ phase (Fig. 1B). In fact, cells treated with IAA94 required about 8 hours more to reach similar percentage of cells in S phase as shown in Fig. 1B and Supplementary Fig. S1.

Also analysis of GB CSC growth curve reflected the elongation of the G₁–S phase in condition of CLIC1 inhibition. In IAA94 as well as in CLIC1-silenced cells, CSCs' proliferation is strongly downregulated (Fig. 1C). In particular, although shRNA transfection does not completely eliminate the protein (Supplementary Fig. S2), cell growth is drastically compromised.

Western blot evaluation of cyclin D1 and p27 expression every 2 hours after G₁-synchronization release established the starting time of G₁–S transition in control cells (Fig. 2A and B). IAA94 treatment of released cells shifted the time of cyclin D1 increase between 2 and 4 hours (Fig. 2C and D). The plot in Fig. 2D supports the results obtained by FACS analysis.

Immunolocalization of CLIC1 with a commercial antibody confirms its metamorphic nature. At the end of starvation period, CLIC1 protein appeared diffuse in the cytoplasm. After 8 hours in complete medium, when the cell cycle had progressed to S phase, the protein localized close to the plasma membrane and concentrated around the nucleus in several cells (Fig. 3A). We quantified CLIC1 presence in the membrane in three ways: (i) cytofluorimetric analysis, (ii) perforated patch-clamp configuration experiments, and (iii) TIRF measurements. For the first method, we collected CSCs at 0 and 8 hours after synchronization release and incubated live cells with a custom-made antibody (AB-NH₂) directed against the external portion of CLIC1 transmembrane protein (8). The higher fluorescence signal detected at 8 hours was proportional to the translocation of the protein from the cytoplasm to the membrane. We confirmed the specificity of the detected signal by competition with the peptide used for immunization (Fig. 3B). Furthermore, whole-cell electrophysiological measurement of IAA94-sensitive membrane current not only supported the idea of a protein-rich membrane, but also confirmed the increase of CLIC1-associated chloride current during G₁ phase progression (Fig. 3C). The presence of other IAA94-sensitive member of the CLIC family (namely CLIC5A, ref. 36) in GB CSCs was excluded by qRT-PCR and thus confirms the specificity of above results (Supplementary Fig. S3). The slope of the current/voltage relationships shown in Fig. 3C represents membrane conductance. At 8 hours after synchronization release (138 pS, triangles), it is almost 5 times higher compared with cells measured just after release (32 pS, circles) or 12 hours later (35.3 pS, diamonds), and to randomly cycling CSCs (23.5 pS, squares).

Finally, we monitored CLIC1 membrane-localization dynamics using the TIRF procedure. We observed CSCs transiently transfected with CLIC1-GFP plasmid, looking for surface fluorescence in two time windows: 0–7 hours and 7–12 hours after synchronization release (Fig. 3D). Membrane fluorescence suddenly increased and reached the maximum value 4 hours after cell release. The fluorescence level held steady up to 8 hours, when it began a slow decline, returning to basal signal after 12 hours.

It has been reported that metabolic conditions are chronically altered in GB cancer cells compared with normal stem cells. The cytoplasmic redox ratio and pH_i levels are not exceptions. In particular, because a transient increase in reactive oxygen species (ROS) is essential for cell-cycle progression from G₁ to S phase (21), we would expect GB CSCs to have a greater ability to generate ROS. Similar considerations can be applied to pH levels: previous measurements showed higher pH_i values in GB cell lines compared with nontumor cells (20). ROS ratio between production and antioxidant action monitored for 30 minutes in randomly cycling CSCs is completely abolished in the presence of IAA94 (Fig. 4A). Similar results have been obtained with AB-NH₂. Both treatments reduced ROS content in GB CSCs to the level of MSCs. pHi is also perturbed by CLIC1 inhibition. Randomly cycling GB CSCs display an average pH_i of 7.46. Impairing CLIC1-associated membrane current by treatment with IAA94 or AB-NH₂ for 24 hours caused pronounced cytoplasmic acidification, reaching values proximal to MSC pH_i (Fig. 4B). The link between CLIC1-associated membrane current and NADPH oxidase (12) appears to be shared also by the NHE1 proton pump, one of the membrane transporters responsible for the control of pHi (20). Both elements are upregulated in GB cell membranes (20, 37, 38) and are specifically inhibited by DPI (38) or EIPA (39), respectively. CSC membrane current recording shows that CLIC1 current density is inhibited at 8 hours after synchronization release by AB-NH₂, DPI, and EIPA (Fig. 4C), confirming the functional link between the three membrane proteins (Supplementary Figs. S4 and S5).

The results showing a tight interconnection between NADPH oxidase, NHE1 proton pump, and CLIC1-associated transmembrane chloride current suggest that CSC proliferation could be reduced by downregulating one of these three membrane proteins. Incubating CSCs for 72 hours with IAA94, AB-NH₂, DPI, or EIPA resulted in a drastic reduction of proliferation (Fig. 5A). However, in MSCs the reduction of cell viability induced by IAA94 and AB-NH₂ was negligible compared with the marked effect of DPI and EIPA (Fig. 5B). The inserts in Fig. 5A and B depict single examples of complete time course evaluating cell growth in GB CSCs and MSCs evidencing that only in CSCs AB-NH₂ was able to affect proliferation. The reduction in the number of both CSCs and MSCs treated with membrane oxidase (DPI) and proton pump (EIPA) inhibitors was mainly due to cell death, whereas IAA and AB-NH2 reduce cell number by reducing the proliferation rate of GB CSC (Fig. 5C and D).

As we previously demonstrated, modifying CLIC1 membrane function affects ROS and pH_i dynamics. NADPH oxidase and NHE1 are also mutually dependent. Figure 6 depicts the changes that occur in ROS ratio (i) and pH modification (ii) when one of the two events is compromised. As a comparison, the plots also show the effects of acute and prolonged CLIC1 current inhibition on ROS content and pH. All the average data reported were expected, except for the early effect on pH caused by IAA94. Short-term CLIC1 inhibition not only reduces ROS unbalance but also exacerbates cytoplasmic alkalinization. This result suggests that the link between CLIC1 and NHE1 is not direct but is likely mediated by NADPH oxidase function. Timing disjunction is evident in a parallel study of the kinetic parameters after CSC release from starvation. Figure 7 shows the ROS dynamics (i), CLIC1 current density (ii), and cytoplasmic pH evaluation (iii) from cell release through the G₁–S transition (see Fig. 1). ROS accumulation and membrane ionic mechanism were previously characterized in microglia cells (12). However, although NADPH oxidase activity shows a peak around 3 to 4 hours, CLIC1 current reaches a maximum after 8 hours, synchronous with the maximum level of cytoplasmic alkalinization. As previously reported (21), a temporary oxidation increase can be interpreted as a permissive signal to cell-cycle progression. Instead, the transient alkalinization of the cytoplasm occurs toward the end of the G₁–S transition, which suggests several potential functional interpretations. One hypothesis is that pH change works as an off signal, downregulating the G₁–S transition machinery. CLIC1-associated current drops synchronously with the maximum cytoplasmic pH value. The ability to mediate the interaction between CLIC1 and the biological membrane by pH change is well-documented in previous publications (14, 37, 40, 41). In perforated whole-cell experiments, external pH was systematically dropped in the external solution from 7.4 to 7.0 and to 6.5 in randomly cycling CSCs. Electrophysiologic experiments measuring CLIC1 current demonstrated that sudden acidification of extracellular pH inhibits CLIC1 current (Fig. 8A). Figure 8B shows the average current inhibition relative to IAA94 action for external pH 7 and 6.5. Delivering the specific inorganic blocker before pH acidification prevents any further decrease in membrane current, which confirms that pH changes and IAA94 act on the same membrane target. Measurements of current inhibition lag time after perfusing the test solution (Fig. 8C) suggest that acidification does not work as an ion channel blocker. Removing CLIC1 protein from the membrane could be an alternative way to reduce membrane current. As already shown in Fig. 3 and confirmed in Fig. 8D (empty circles), TIRF experiments demonstrate that due to the presence of CLIC1-GFP at the membrane level of GB CSCs, the fluorescent signal progressively fades 8 hours after release from G₁ accumulation. This occurs synchronously with the intracellular alkalinization peak (Fig. 7C) and with CLIC1-associated current decrease (Fig. 7B). TIRF measurements obtained while maintaining the external solution at pH 8 (open square) show no fluorescent signal dimming (Fig. 8D).

Human GB CSC-enriched cultures show a steady-state condition in which both ROS balance and pHi are chronically altered. Randomly cycling GB CSCs have a 5-fold increase in their ability to produce ROS and a pH_i almost 0.3 units more alkaline than human MSCs (Fig. 4A and B). Many different solid tumors demonstrate similar, consistent overexpression and upregulation of several metabolic cell components (42–45). Physiologically, this cellular hyperactivation is a transient condition. The process of achieving a new state of stability through physiologic changes is known as “allostasis” (28). This phenomenon is essential to maintain cell viability during fluctuating basal conditions. However, when stress is persistent and cells are unable to restore a basal homeostatic state, allostasis can become chronic. Most of the cells are not able to cope with these new conditions prolonged in time and they die. Maintaining allostasis is energetically very costly and requires cells with altered basal metabolism. As an example, a chronic condition of oxidative stress results in an overproduction of antioxidant component to balance the system. This is the case for GB CSCs, in which the most evident allostatic outcome is a high rate of cell division (5, 6). Several proteins present in the allostatic GB cell plasma membrane are deregulated in either expression or function (20, 46–48). Altering any of these proteins can disrupt the allostatic equilibrium, causing drastic functional changes and, in some cases, even cell death. Theoretically, many of these proteins can be considered as valuable pharmacologic targets. However, most of them also play important roles in physiologic homeostatic conditions. Consequently, modifying their functions would affect not only cancer cells but also healthy cell populations. On the contrary, CLIC1 protein only becomes continuously active as a membrane charge carrier during periods of chronic stress, whereas during homeostatic conditions, it is essentially irrelevant. This is the most important result of our investigation. Figure 5 shows that the inhibition of NADPH oxidase or the NHE1 proton pump, both overexpressed in CSCs, causes cell death in CSCs and healthy MSCs. On the contrary, impairing CLIC1 membrane protein, using either the chloride channel blocker IAA94 or the specific antibody against the external portion of the protein, prolongs CSC cell-cycle duration and leaves MSC functions unaltered. Confirming this result, GB CSC growth curves in the presence of IAA94 show that the early apoptosis rate is less than 10% (Supplementary Fig. S6) even though the proliferation rate drops drastically (Fig. 4A). The present work suggests that CLIC1 protein and its associated chloride permeability are principal elements involved in the stabilization of GB CSC allostasis. Consistent with previous results, CLIC1 membrane current and NADPH oxidase are part of a feed-forward mechanism that sustains ROS production, even in GB CSCs (12). This result suggests that GB CSCs are in a constant state of oxidative stress. ROS accumulation is considered to be a signal to promote cell-cycle progression from G₁ to S phase (23). Our working hypothesis is that CLIC1 acts as cell-cycle accelerator in GB CSCs.

As a metamorphic protein, CLIC1 membrane localization is stimulated by increased ROS generation. Simultaneously, by promoting transmembrane chloride flux, CLIC1 compensates for negative charges released by NADPH oxidase, creating a self-sustaining metabolic pathway (12). Our experiments suggest that to trigger G₁–S transition, ROS levels must reach a defined threshold. We can hypothesize that this increased ROS level can be due either to an increase in the production or a decrease in the scavenging or even both the phenomena at the same time. The feed-forward mechanism generated by the cooperative activity of the membrane oxidase and CLIC1 sets the limiting rate of cell-cycle progression kinetics based on the active number of the two elements: the more proteins involved, the faster the transition. Cytoplasm alkalinization is another characteristic of GB CSCs. Our experimental setting highlights the mutual modulation of the three elements: ROS, pH_i and chloride flux. In randomly cycling CSCs, pHi is chronically elevated by almost 0.3 units and is functionally linked with both NADPH oxidase and CLIC1 activities (Fig. 5). Blocking NADPH oxidase or inhibiting CLIC1 membrane function for prolonged periods results in cytoplasmic acidification (Fig. 5B). However, within minutes of suppressing CLIC1 current, we observed an unexpected transient intracellular alkalinization (Fig. 5B). This suggests the proton pump still runs after chloride inflow stops, lowering internal [H⁺]. Thus, no direct link appears to exist between internal pH and CLIC1-associated chloride current. However, after 24 hours of CLIC1 inhibition, pH_i acidifies to MSC levels. The link is represented by NADPH oxidase: CLIC1 inhibition impairs ROS production, causing downregulation of NHE1 function.

The role of the oxidative stress peak is well-established (21). Our results support ROS accumulation as a trigger for G₁–S progression. However, it is not clear whether cytoplasmic pH modulation has a role during the cell cycle. Time-lapse experiments on synchronized cells show ROS production growing early after G₁ release in parallel with increasing membrane current (Fig. 6A and B). pHi depicts a cytosolic alkalinization peak delayed in the G₁–S time progression (Fig. 6C). All our experiments, from immunolocalization, FACS analysis, and electrophysiology (Fig. 2) to TIRF (Figs. 2 and 7), show a gradual reduction of both CLIC1 membrane localization and its associated current synchronous with transient intracellular alkalinization (Fig. 6B and C). We have previously demonstrated that CLIC1 chloride current sustains ROS production by a feed-forward mechanism (12). Based on our electrophysiology and dynamic localization data in the presence of pH changes (Fig. 7), we propose that transient alkalinization can reduce CLIC1 activity and, as a consequence, ROS production, by determining de facto the beginning of a new cell-cycle phase.

In conclusion, GB CSCs survive and proliferate in a “stable” condition in which all cell systems governing cell-cycle progression (including NADPH oxidase, NHE1 proton pump, CLIC1, and others; refs. 20, 46, 48, 49) are upregulated. CLIC1 ability to balance CSC allostatic state means that CLIC1 membrane permeability can be considered an allostatic state stabilizer. Its function is negligible in homeostasis but fundamental during chronic stress. Among the three elements we considered in these experiments, CLIC1 is the only target that can be proposed for potential pharmacologic therapy with minimal side effects. Inhibiting CLIC1-associated membrane current would not only compromise proliferation, but would also return CSCs from chronic allostasis to steady homeostasis. Furthermore, in light of drug repositioning approaches for cancer research (50), CLIC1 direct inhibition is one of the proposed mechanisms for the antitumoral effects of the antidiabetic drug metformin (5), which paves the way for the development of a novel class of inhibitors.

No potential conflicts of interest were disclosed.

Conception and design: T. Florio, M. Mazzanti

Development of methodology: M. Peretti, F.M. Raciti, V. Carlini, S. Barozzi, A. Pattarozzi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Peretti, F.M. Raciti, V. Carlini, S. Sertic, S. Barozzi, M. Garré, A. Pattarozzi, A. Daga, A. Costa

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Peretti, F.M. Raciti, V. Carlini, I Verduci, A. Pattarozzi, F. Barbieri, T. Florio, M. Mazzanti

Writing, review, and/or revision of the manuscript: M. Peretti, F.M. Raciti, V. Carlini, I Verduci, A. Daga, A. Costa, T. Florio, M. Mazzanti

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I Verduci

Study supervision: T. Florio, M. Mazzanti

We are thankful to Francesca Cianci and Xueting Bai for assistance in several molecular biology experiments, Francesca Casagrande and Ilaria Costa for confocal imaging and time-lapse microscopy, and Erika Pizzi for critical reading. This work was supported by grant no.16713, IG 2015 to M. Mazzanti from the Italian Association for Cancer Research (AIRC), Italy.

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.