Multiparameter Screening Reveals a Role for Na⁺ Channels in Cytokine-Induced β-Cell Death

Abstract

Pancreatic β-cell death plays a role in both type 1 and type 2 diabetes, but clinical treatments that specifically target β-cell survival have not yet been developed. We have recently developed live-cell imaging-based, high-throughput screening methods capable of identifying factors that modulate pancreatic β-cell death, with the hope of finding drugs that can intervene in this process. In the present study, we used a high-content screen and the Prestwick Chemical Library of small molecules to identify drugs that block cell death resulting from exposure to a cocktail of cytotoxic cytokines (25 ng/mL TNF-α, 10 ng/mL IL-1β, and 10 ng/mL IFN-γ). Data analysis with self-organizing maps revealed that 19 drugs had profiles similar to that of the no cytokine condition, indicating protection. Carbamazepine, an antiepileptic Na⁺ channel inhibitor, was particularly interesting because Na⁺ channels are not generally considered targets for antiapoptotic therapy in diabetes and because the function of these channels in β-cells has not been well studied. We analyzed the expression and characteristics of Na⁺ currents in mature β-cells from MIP-GFP mice. We confirmed the dose-dependent protective effects of carbamazepine and another use-dependent Na⁺ channel blocker in cytokine-treated mouse islet cells. Carbamazepine down-regulated the proapoptotic and endoplasmic reticulum stress signaling induced by cytokines. Together, these studies point to Na⁺ channels as a novel therapeutic target in diabetes.

Pancreatic β-cell death is pathologically elevated in type 1 and type 2 diabetes, as well as in failing islet transplants. In type 1 diabetes, infiltrating autoreactive T cells secrete proinflammatory cytokines, such as TNF-α, IL-1β, and IFN-γ, which effectively coerce the vast majority of pancreatic β-cells into programmed cell death (1–4). Preventing β-cell death could reduce the burden on at-risk families, and protection of the few remaining β-cells in type 1 diabetic patients has the potential to delay disease progression (3). Furthermore, massive β-cell death before, during, and after clinical islet transplantation reduces the success of this promising therapy for type 1 diabetes (5–7). Type 2 diabetes results from the eventual loss of functional β-cell mass, which is in part due to increased β-cell death (2, 8). Thus, there is an urgent clinical need to develop or repurpose drugs that can enhance β-cell survival in both type 1 and type 2 diabetes.

Ion channels are a critical class of drug targets, and significant effort has been devoted to understanding the ionic basis of glucose-stimulated insulin secretion (9). Particular attention has been focused on K⁺ and Ca²⁺ channels, whereas Na⁺ channels are much less well understood despite their presence on β-cells (9). Moreover, the role of plasma membrane ion channels in cell fate decisions, such as apoptosis, remains to be fully elucidated. Unbiased screens are critical for the identification of new pathways involved in programmed β-cell death. We have recently developed rich multiparameter screening platforms with which multiple aspects of programmed cell death can be assessed simultaneously in a high-throughput manner (10).

Here, we report the results of a multiparameter, image-based high-throughput screen to identify drugs that prevent β-cell death in the context of cytotoxic cytokines, designed to mimic conditions that precipitate type 1 diabetes (11). We identified several novel antiapoptotic drugs, including carbamazepine, a use-dependent Na⁺ channel inhibitor. The confirmation of these effects with another use-dependent Na⁺ channel inhibitor strongly suggests a previously unappreciated role for Na⁺ channels in pancreatic β-cells. These data will support therapeutic efforts to inhibit β-cell death in diabetes.

Materials and Methods

Primary islet isolation, cell culture, and perifusion

Pancreatic islets were isolated from 12- to 20-week-old male or female C57BL/6J mice or MIP-GFP mice (The Jackson Laboratory) using collagenase and the filtration technique described previously (12, 13). Mice were housed in accordance with the University of British Columbia Animal Care Committee guidelines. The islets were further hand picked using a brightfield microscope. Islets were cultured overnight (37°C and 5% CO₂) in RPMI 1640 medium (Invitrogen) with 9.9 mM glucose (Sigma-Aldrich), 100 U/mL penicillin, 100 μg/mL streptomycin (Invitrogen), and 10% vol/vol fetal bovine serum (FBS) (Invitrogen). Dynamic insulin secretion was measured by perifusion and RIA (14). MIN6 cells were cultured in DMEM (Invitrogen) containing 22.2 mM glucose, 100 U/mL penicillin, 100 μg/mL streptomycin, and 10% vol/vol FBS. Chemicals were from Sigma-Aldrich unless specified otherwise.

Multiparameter cell death screening platform

MIN6 cells expressing the eBFP2-DEVD-eGFP fluorescence resonance energy transfer (FRET) sensor for detection of caspase-3/7 activation were seeded into 96-well plates and cultured as described previously (13, 15). Cells were stained with 50 ng/mL Hoechst 33342 (Invitrogen), 0.5 μg/mL propidium iodide (PI) (Sigma-Aldrich), and a 3:500 dilution of annexin V–conjugated to Alexa Fluor 647 (Invitrogen). Cells were treated with test drugs in the presence of 22.2 mM glucose containing DMEM supplemented with 10% vol/vol FBS and a cytokine cocktail (25 ng/mL TNF-α, 10 ng/mL IL-1β, and 10 ng/mL IFN-γ; R&D Systems). After 30 hours of treatments, cells were imaged with ImageXpress Micro (Molecular Devices) at 37°C and 5% CO₂. The fluorescent emission of Hoechst 33342 and enhanced blue fluorescent protein 2 (eBFP2), excited at 386/23 nm, was captured with a 445/20 nm filter. PI excitation and emission wavelengths passed through 562/40 and 624/40 nm filters, respectively. Alexa Fluor 647 excitation and emission wavelengths passed through 628/40 and 692/40 nm filters, respectively. Enhanced green fluorescent protein (eGFP) excitation and emission wavelengths passed through 472/30 and 520/35 nm filters, respectively. FRET between eBFP2 and eGFP decreases upon activation of caspase-3/7 and was measured using a 520/35 nm filter and normalized to eBFP2 emission intensity. Time-lapse images for the secondary screens were captured every 2 hours and analyzed using MetaXpress Software (Molecular Devices).

Chemical library

We used the Prestwick Chemical Library of 1120 drugs, which includes a diverse array of chemicals, including off-patent drugs (www.prestwickchemical.com). Compounds were pinned into the 96-well round bottom plates (Corning) at ∼15 μM using a 0.7-mm diameter 96-pin tool equipped onto a Biorobotics Biogrid II robot. Final compound concentrations of ∼8.5 μM were attained when treatments were added to the cells.

Electrophysiology

Ionic currents were recorded with standard whole-cell patch clamp methods. Patch pipettes were pulled from 1.5-mm OD thin-walled borosilicate glass (World Precision Instruments) to have tip resistances of 1 to 2 MΩ in standard recording solutions. The internal (pipette) solution contained 135 mM CsCl, 10 mM HEPES, and 10 mM EGTA, with pH adjusted to 7.2 using CsOH. The external solution contained 140 mM NaCl, 2 mM MgCl₂, and 10 mM HEPES, with pH adjusted to 7.4 using NaOH. Currents were recorded with an Axopatch 200B patch clamp amplifier and Digidata 1322A analog-to-digital converter (Molecular Devices), filtered at 10 kHz, and digitized at 50 kHz. For collection of current-voltage relationships, P/4 leak subtraction was used, with a holding potential of −120 mV. For measurement of use-dependent inhibition, it was not possible to perform P/N leak subtraction, so only data from cells with small amounts of leak current were used. In these cases, we used offline linear leak subtraction, although this had very small effects on the appearance of the data.

RT-PCR and immunoblotting

Total RNA was isolated from fluorescence-activated cell sorting (FACS)–purified mouse primary islet cells using an RNeasy Mini kit (Qiagen). cDNA was generated using qScript cDNA SuperMix (Quanta Biosciences) and quantitative RT-PCR was conducted using TaqMan probes from Applied Biosystems and PerfeCTa qPCR SuperMix (Quanta) on a StepOnePlus device (Applied Biosystems). Relative gene expression changes were analyzed by the 2^−ΔCt method, with Hprt1 or Ppia as reference genes. Immunoblotting was performed as described previously (14).

Calcium imaging

Mouse islet cells were transfected with the D3cpv, D1ER, or 4mtD3cpv FRET probe (16–18) for the detection of cytosolic, endoplasmic reticulum (ER), or mitochondrial Ca²⁺. In brief, dispersed mouse islet cells were transfected using the Neon system (Invitrogen) and imaged 48 hours later with ImageXpress Micro (Molecular Devices) at 20-minute intervals at 37°C and 5% CO₂. Cells were imaged in RPMI 1640 medium with 20 mM glucose (14).

Statistical analysis

Data are expressed as means ± SEM unless otherwise indicated. Results were considered statistically significant at a value of P < .05 using the Student t test or ANOVA, where appropriate, with Prism software from GraphPad Software Inc.

Results

High-throughput screen for discovering drugs that protect β-cells

Image-based, high-throughput screening requires a high degree of reproducibility. For our screens, we chose the MIN6 β-cell line, which we and others have shown undergoes stereotypical apoptotic programmed cell death in response to proinflammatory cytokines (10). Apoptosis is the mode of cell death that is most commonly assessed in the fields of islet biology and diabetes research (1, 2), and guidelines on the measurement of apoptosis consistently state that multiple parameters should be assessed to distinguish it from other forms of cell death (19, 20). In the current study, we used a 4-parameter live-cell imaging approach to assess the effects of drugs on cytokine-induced β-cell apoptosis. To assess cell number, MIN6 cells were continuously cultured in a low concentration of Hoechst 33342, which we have previously shown does not have significant effects on cell survival (10). To assess phosphatidylserine translocation to the plasma membrane outer surface, an early event in the classic apoptotic cascade (21, 22), cells were cultured in a low concentration of Alexa Fluor 647–conjugated annexin V, which we have also shown to be nontoxic (10). Caspase-3 activation was assessed using a FRET sensor we recently modified to include eBFP2 and eGFP (10). The detection of late-phase cell death was monitored with incorporation of PI, which enters the cell only after the irreversible compromise of the plasma membrane. The combination of these fluorescent probes provided us with 4 distinct parameters for the analysis of β-cell death (Figure 1).

Figure 1.

Multiparameter, high-content screening for compounds that promote β-cell survival. MIN6 cells stably expressing the caspase-3 eBFP-DEVD-eGFP FRET sensor under the Ins1 promoter were seeded into 96-well plates and stained with Hoechst 33342, PI, and Alexa Fluor 647–conjugated annexin V. Cell death was induced with a cytokine cocktail of 25 ng/mL TNF-α, 10 ng/mL IL-1β, and 10 ng/mL IFN-γ, and cells were treated with an 8.5 μM concentration of each compound in the Prestwick Chemical Library, which includes US Food and Drug Administration–approved drugs. Cells were imaged after 30 hours of treatment with ImageXpress Micro, and images were analyzed using MetaXpress software. A–D, Z score values calculated from the numbers of total cells, PI-positive cells, annexin V–positive cells, and activated caspase-3–positive cells were determined and displayed as a heatmap where red and green represent high and low numbers, respectively. The top hits that promoted β-cell survival were selected from automated clustering into 20 different self-organizing maps (A). Scatter plots of PI (B), annexin V (C), and caspase-3 (D) vs total cell were generated. CBZ, carbamazepine.

Multiple antiapoptotic drugs identified with the high-throughput, high-content assay

We screened the Prestwick Chemical Library of 1120 drugs in an effort to identify chemicals and also molecular mechanisms that modulate apoptotic MIN6 cell death induced by a well-established toxic cocktail of cytokines. The results were automatically analyzed into self-organizing maps using the Acuity Express software program (Figure 1A). This approach groups the treatments according to similarity, taking into account the 4 measured cell death–associated parameters. Using this method, 19 distinct drugs showed a pattern that was similar to the “no cytokine” condition, meaning that these were the closest to completely blocking the apoptosis-inducing effects of the cytokines (Figure 1A). A variety of chemicals, from several classes, populated this small “hit” list (Table 1). Notably, we identified 5 vitamins, 5 antibiotics or antifungals, 6 drugs that modulate plasma membrane ion channels and/or membrane receptors, and 3 miscellaneous drugs that had significant protective effects on β-cell survival. The identification of direct antiapoptotic effects in a β-cell model was novel for most of the drugs. For 9 of these chemicals, we were unable to find evidence that they had been linked to apoptosis in any cell type. A literature search revealed that 4 of these chemicals had previously been shown to be antiapoptotic in another cell type, whereas 3 had been shown to be proapoptotic in other systems (Table 1). Two of these drugs, loperamide and carbamazepine, are known to either increase or decrease apoptosis, depending on the context. The possibility that these drugs might have specific effects on β-cells was intriguing.

Table 1.

Summary of “Hits” Identified as Similar to the “No Cytokine” Condition Using Self-Organizing Maps

Hit Drug	Description	Known Effects on Apoptosis?	References
Pantothenic acid	Vitamin B5	Antiapoptotic in some systems	60
Calciferol	VitaminD (D2, D3)	Proapoptotic in some systems	61
Cholecalciferol	VitaminD (D3)	Unknowa
Tocopherol (R,S)	Vitamin E	Antiapoptotic in some systems	62, 63
Trolox	Vitamin E	Antiapoptotic in some systems	64
Amphotericin B	Pore-forming antibiotic, antifungal	Proapoptotic in some systems	65–69
Ceftazidime	Antibiotic	Proapoptotic in some systems	70
Cephalosporanic acid	Antibiotic	Unknowna
Butoconazole	Antifungal	Unknowna
Adamantanamine	Antiviral	Unknowna
Alprenolol	Nonselective β-blocker, 5-HT1A antagonist	Unknowna
Loperamide	μ-Opioid agonist, calcium channel inhibitor	Antiapoptotic or proapoptotic	71–73
Buflomedil	Vasodilator	Unknowna
Tolazoline	Adrenergic blocking agent, peripheral vasodilator	Unknowna
Carbamazepine	Na+ channel inhibitor, anticonvulsant, mood stabilizer	Antiapoptotic or Proapoptotic	74–76
Carcinine	β-Alanylhistamine, imidazole dipeptide	Antiapoptotic in photoreceptors	77
Meticrane	Diuretic	Unknowna
Bisacodyl	Laxative	Unknowna
Iopamidol	Contrast agent	Unknowna

^aListed as unknown as a search for the drug name and “apoptosis” yielded no results in PubMed.

The use of our multiparameter screening approach was validated when we identified mitoxantrone dihydrochloride as a drug that could completely prevent PI incorporation in the presence of toxic cytokines (Figure 1B). However, this was a false hit. In fact, there was a very high percentage of dying cells in the presence of this compound as shown by the dramatic cell loss, high annexin V fluorescence, and high caspase-3 activity (Figure 1, B–D), but this topoisomerase inhibitor had apparently prevented the PI from binding to the DNA. Had we not used a rich multiparameter approach in our initial screen, valuable time and resources could have been wasted following up this false-positive effect.

Carbamazepine blocks Na⁺ currents in primary mouse pancreatic β-cells

We chose a single drug for secondary validation. Of the 19 unique hits, we chose to focus on carbamazepine, a Na⁺ channel inhibitor, for 4 reasons. First, we were interested by the fact that carbamazepine has been found to have both proapoptotic and antiapoptotic effects in other systems. Second, ion channels are a highly druggable class of membrane proteins. Third, little was known about the role of voltage-gated Na⁺ channels in programmed cell death, in β-cells or in other cell types. Fourth, Na⁺ channels are not thought to play a major role in normal β-cell physiology, raising the possibility that targeting these channels might be specific to a state of β-cell stress and/or β-cell death.

Voltage-gated Na⁺ channels have not been extensively studied in β-cells, relative to other plasma membrane channels, so it was critical to evaluate the Na⁺ currents in our primary islet cell cultures and determine whether they were inhibited by carbamazepine. We used voltage clamp electrophysiology in the whole-cell configuration to measure Na⁺ in unambiguously identified β-cells from transgenic mice expressing green fluorescent protein (GFP) under the control of the Ins1 promoter (ie, MIP-GFP mice) (23). Indeed, many MIP-GFP β-cells in our cultures expressed Na⁺ currents with both a rapidly inactivating transient phase and an unusual sustained phase that was not sensitive to tetrodotoxin (Figure 2A and data not shown). Voltage-gated Na⁺ currents were most frequently measured from GFP-positive β-cells compared with β-cells identified simply by their larger morphology, suggesting that these currents may be more prominent in “mature” β-cells marked by activation of the mouse Ins1 promoter (24–26) than in immature β-cells or in other islet cell types.

Figure 2.

Use-dependent inhibition of voltage-gated Na⁺ currents in β-cells from MIP-GFP dispersed islets. A, GFP-positive cells were identified by epifluorescence, and whole-cell patch clamp recording was performed. Cells were held at −100 mV and pulsed in 10-mV steps between −90 mV and +50 mV. Rapidly decaying inward Na⁺ currents were apparent, along with a sustained inward current. The rapidly decaying current component was well inhibited by carbamazepine (CBZ). B, Peak current-voltage relationships for Na⁺ currents in the presence and absence of carbamazepine. C, Current traces depict pulses 1 and 25 in a repetitive train of 20-ms depolarizations from −100 mV to +10 mV (10-Hz frequency). In the presence of carbamazepine, there is a significant tonic block of the current starting at pulse 1 and further current reduction over the course of the pulse train. The sustained current component was not sensitive to use-dependent inhibition. D, Dispersed MIP-GFP islet cells were FACS sorted, and the expression of voltage-gated Na⁺ channels (Na_v) and β-subunits (Na_vβ) was detected by quantitative RT-PCR (n = 3, mean ± SEM).

Importantly, we confirmed the molecular mechanism of carbamazepine. As expected, carbamazepine treatment resulted in the use-dependent inhibition of β-cell Na⁺ channels (Figure 2, A–C). The expression of voltage-gated Na⁺ channels and associated β-subunits was detected in FACS-purified β-cells (Figure 2D). Our results were consistent with those of other studies indicating the expression of Na_v1.3 and Na_v1.7 in mouse islet cells (27) and further indicated that Na_v1.7 was predominantly expressed in β-cells (Figure 2D). These data collectively demonstrated that many mature β-cells express voltage-gated Na⁺ currents and validate the fact that carbamazepine can block these channels, but they do not necessarily show that the antiapoptotic effects of carbamazepine result from the use-dependent inhibition of Na⁺ channels.

Use-dependent Na⁺ channel blockers protect primary mouse β-cells from apoptosis

The use-dependent inhibition of Na⁺ currents by carbamazepine was reminiscent of the effects of lidocaine and other classic local anesthetics/anticonvulsants. To test the hypothesis that use-dependent inhibition of Na⁺ channels protects β-cells from cytokine-induced apoptosis, we chose to compare the effects of carbamazepine with those of lidocaine, which is also a use-dependent Na⁺ channel blocker. We compared these effects with those of tetrodotoxin, a Na⁺ channel blocker that is not use-dependent. Remarkably, both carbamazepine and lidocaine prevented cytokine-induced death of primary islet cells in a dose-dependent manner (Figure 3, A and B). On the other hand, tetrodotoxin was not protective (Figure 3C). We further tested whether these drugs might suppress programmed cell death induced by ER stress, which has been shown to share some mechanistic similarities with cytokine-induced apoptosis (28). Interestingly, all 3 drugs partially protected primary islet cells from death associated with the sarcoendoplasmic reticulum calcium ATPase inhibitor, thapsigargin (Figure 3, D–F). Studies with MIP-GFP islet cells further determined that the protection against cytokine-induced islet cell death was β-cell specific (Figure 4, A–C). The non–β-cell population, indicated by the lack of GFP expression, did not display any increase in cell death after cytokine treatment (Figure 4C). These data establish previously unappreciated roles for Na⁺ channels in β-cell death induced by cytokines and ER stress.

Figure 3.

Na⁺ channel inhibitors can reduce cytokine-induced primary islet cell death. Dispersed islet cells seeded into 96-well plate were stained with Hoechst 33342 and PI. Cells were treated with a cytokine cocktail (25 ng/mL TNF-α, 10 ng/mL IL-1β, and 100 ng/mL IFN-γ) (A–C) or 1 μM thapsigargin (D–F), in combination with 0.01 to 100 μM concentrations of Na⁺ channel inhibitors carbamazepine, lidocaine, and tetrodotoxin. Cells were imaged with ImageXpress Micro and the percentage of PI-positive (PI+) cells was calculated. Insets, area under the curve for the last 10 hours was calculated (n = 3–6, mean ± SEM; ✻, P < .05 compared with cytokine-treated cells).

Figure 4.

Na⁺ channel inhibitors can reduce cytokine-induced primary β-cell death. A–C, Dispersed MIP-GFP islet cells seeded into 96-well plates were stained with Hoechst 33342 and PI. Cells were treated with a cytokine cocktail (25 ng/mL TNF-α, 10 ng/mL IL-1β, and 100 ng/mL IFN-γ), in combination with 0.01 to 100 μM carbamazepine (CBZ). Cells were imaged with ImageXpress Micro, and the percentage of PI-positive cells in the islet cell (A), β-cell (B), and non–β-cell (C) populations was calculated. Insets, area under the curve for the last 10 hours was calculated (n = 4–6, mean ± SEM; ✻, P < .05 compared with cytokine-treated cells). D, Insulin level in medium collected 70 hours after treatment was measured (n = 5).

Carbamazepine does not affect insulin secretion

It was not clear what effects carbamazepine might have on insulin secretion. This is important because any β-cell–protecting drug should not have deleterious effects on insulin secretion. Moreover, it was important to test whether carbamazepine protected β-cells by modulating insulin secretion, which can have autocrine prosurvival effects (29, 30). Mouse islet cells treated with carbamazepine in the presence of cytokines did not display significant differences in insulin levels in the media collected 70 hours after treatment compared with that from islets treated with cytokines alone (Figure 4D). Islet perifusion studies further showed that carbamazepine does not acutely modulate insulin secretion under either basal or glucose-stimulated conditions (Figure 5). Thus, carbamazepine can protect β-cells from toxicity without significant effects on insulin secretion. These data also imply that the prosurvival effects were independent of autocrine insulin signaling.

Figure 5.

Insulin secretion is not modulated by acute carbamazepine (CBZ) treatment. Mouse islets were perifused with Krebs-Ringer bicarbonate buffer containing 3 or 15 mM glucose in combination with 100 μM carbamazepine (n = 3, mean ± SEM). DMSO, dimethyl sulfoxide.

Carbamazepine modulates cytosolic, ER, and mitochondrial Ca²⁺

We next assessed the molecular mechanisms associated with the protective effects of carbamazepine. Na⁺ channel modulation could influence intracellular Ca²⁺ levels by altering the electrical activity of β-cells (31, 32) and disruption of Ca²⁺ homeostasis can induce cell death (33). Long-term cytosolic Ca²⁺ imaging with genetically encoded Ca²⁺ biosensors revealed the elevation of intracellular Ca²⁺ in dying cells after treatment with a cocktail of proinflammatory cytokines and showed that this was reduced by carbamazepine (Figure 6A, right). The only cells that survived when treated with cytokines alone maintained a lower level of cytosolic Ca²⁺ (Figure 6A, left). These observations are consistent with the concept that carbamazepine protects β-cells by suppressing cytokine-induced Ca²⁺ excitotoxity via the use-dependent inhibition of Na⁺ channels.

Figure 6.

Carbamazepine (CBZ) modulates cytosolic, ER, and mitochondrial Ca²⁺ signaling. Dispersed mouse islet cells were transfected with D3cpv (A), D1ER (B), or 4mtD3cpv (C) genetically encoded Ca²⁺ biosensors to image cytosolic, ER, or mitochondrial Ca²⁺, respectively. Cells were incubated in 20 mM glucose, serum-free RPMI 1640 medium before the treatments with cytokines (red), cytokines and 100 μM carbamazepine (black), or no cytokines (green) at the indicated time (arrow). Left, Ca²⁺ levels of cells that remained alive during the 38 hours of imaging. Inset, area under the curve from the time of treatment to the end of the time course (n = 6–28 cells; ✻, P < .05 compared with cytokine treated). Right, Ca²⁺ levels of cells that died during the 38 hours of imaging. The FRET/cyan fluorescent protein ratios were normalized to the average reading from 5 to 6 hours before PI incorporation (0 hours). Inset, area under the curve from the time of Ca²⁺ influx or depletion to the time of PI incorporation (n = 11–26 cells; *, P < .05 compared with cytokine treated).

We also assessed Ca²⁺ levels under the same conditions within 2 key organelles, the ER and mitochondria. Interestingly, we also found that carbamazepine treatment could prevent the changes in the level of ER Ca²⁺ induced by cytokines and delay cell death after induction of Ca²⁺ depletion (Figure 6B, right). The prolonged depletion of ER Ca²⁺ was not associated with changes in mitochondrial Ca²⁺, suggesting that carbamazepine treatment may not alter the Ca²⁺-mediated crosstalk between ER and mitochondria (Figure 6C, right). Overall, these experiments demonstrate that inhibition of voltage-gated Na⁺ channel activity can modulate changes in Ca²⁺ disruptions induced by cytokines.

Carbamazepine decreases proapoptotic and ER stress signaling

Proinflammatory cytokines can induce apoptotic and ER stress signaling in β-cells (34, 35). Previous studies have also shown that proinflammatory cytokines can induce p38 MAPK activation and that TNF-α can enhance Na⁺ channel activity in mouse dorsal root ganglion neurons through a p38 MAPK-dependent mechanism (36). Thus, we considered the possibility that carbamazepine acts in part by blocking the activation of Na⁺ channels induced by p38 MAPK. Indeed, in islet cells treated with cytokines, we observed increases in cleaved caspase-3 and C/EBP homology protein (CHOP) levels and increases in activation of the p38 MAPK and c-Jun NH₂-terminal kinase (JNK) pathways (Figure 7, A–D). Use-dependent blockage of voltage-gated Na⁺ channels by carbamazepine and lidocaine could decrease the elevation in cleaved caspase-3, CHOP, and phosphorylated JNK induced by cytokines (Figure 7, A, B, and D). Surprisingly, carbamazepine and lidocaine displayed differential effects on the phosphorylation of p38 MAPK (Figure 7C). The prosurvival effects of carbamazepine were associated with the down-regulation of proapoptotic and ER stress signaling but not with changes in BAD phosphorylation at serine-112 (Figure 7, E and F). Together, these experiments describe mechanisms involved in the protection of β-cells from cytokine- and ER stress–induced apoptosis by use-dependent Na⁺ channel inhibitors.

Figure 7.

Carbamazepine down-regulates cytokine-induced proapoptotic and ER stress signaling. A–E, Mouse islet cells were treated with 100 μM carbamazepine (CBZ), lidocaine (LIDO), or tetrodotoxin (TTX) under 20 mM glucose serum-free conditions for 24 hours. Immunoblotting of cleaved caspase-3, CHOP, phosphorylated and total p38 MAPK, JNK, and BAD is shown (n = 4–6; ✻, P < .05 compared with cytokine treated). F, Model of effects of carbamazepine on cell death signaling in β-cells.

Discussion

The goal of the present study was to identify small molecule drugs that protect pancreatic β-cells from programmed cell death induced by proinflammatory cytokines and to uncover the mechanisms of action. Our high-throughput screen identified a number of drugs as powerful prosurvival agents, including the use-dependent Na⁺ channel blocker carbamazepine. This observation prompted us to investigate the roles of Na⁺ channels in β-cells and pointed to an unexpected role of Na⁺ channels in β-cell death.

The physiological and pathophysiological roles of Na⁺ channels in β-cells remain poorly understood. Voltage-gated Na⁺ channels have been identified on β-cells from mice, rats, dogs, and humans (9, 37–42). It has been shown that Na⁺ channels are more active in canine and human β-cells than in rodent cells (40). Heterogeneity of Na⁺ currents and Na⁺ channel expression has been observed between subpopulations of β-cells (27). In our hands, the use of MIP-GFP cells permitted the recording of Na⁺ channels from many more cells, perhaps suggesting an enrichment of Na⁺ channels in mature β-cells. We have previously shown that β-cells with Ins1 promoter expression (ie, MIP-GFP cells) represent a subgroup of β-cells in a mature state (24–26). Na⁺ currents have been reported previously in MIP-GFP β-cells (23).

Insulin secretion from human β-cells is modulated by blockers of voltage-dependent Na⁺ channels (43). However, in general, studies with chemical inhibitors have suggested that Na⁺ channels have relatively modest effects on glucose-stimulated insulin secretion (9, 37, 39, 42). Our data suggest that use-dependent blockage of Na⁺ channels does not influence insulin secretion. Glucose induces cytoplasmic Na⁺ oscillations in pancreatic β-cells (44). Na⁺ channel activation modulates the frequency of Ca²⁺ oscillations in β-cells (32). In β-cells, increased ATP can shift the current-voltage and the voltage-dependent inactivation curves to the right (45). On the other hand, lifelong, complete knockout of the major regulatory Na⁺ channel β1-subunit (Scn1b, a component of Na_v1.7) reduces insulin secretion (46). Tetrodotoxin blocks insulin secretion induced by carbachol (47). Na⁺ channels have also been shown to play a role in glucagon release from α-cells (48). Beyond hormone secretion, virtually nothing has been reported on roles for Na⁺ channels in β-cell fate decisions. To the best of our knowledge, our study is the first to report that Na⁺ channels participate in programmed cell death in β-cells or any endocrine cell type.

The role of Na⁺ channels in neuronal apoptosis has been the focus of several studies. Overall, these studies paint a complex picture. The potent Na⁺ channel blocker tetrodotoxin has been shown to protect neurons from hypoxia (49). Similarly, 4-[4-fluorophenyl]-2-methyl-6-[5-piperidinopntyloxy] pyrimidine hydrochloride, an inhibitor that blocks both Ca²⁺ and Na⁺ channels, protects neurons after ischemia (50). Moreover, Na⁺ channel mutations correlate with increased survival and disease severity in glioblastoma multiforme (51). On the other hand, mice lacking the Scn2a subunit gene, a component of the Na_v1.2 channel, exhibit significant neuronal apoptosis (52). Neurons in culture and in vivo show activity-dependent survival, whereby inhibition of Na⁺ channels leads to apoptotic cell death (53–55) and selective activation of Na⁺ channels promotes survival (56). Other mutations in Na⁺ channels can have proapoptotic effects in the heart (57, 58). Together, these studies suggest that Na⁺ currents may play a deleterious role in neurons subjected to hypoxic or ischemic stresses, whereas a certain level of Na⁺ channel activity in basal conditions is essential for activity-dependent survival. This role is similar to the well established dual roles for Ca²⁺ channels in cellular survival. Collectively, these data support a model whereby plasma membrane excitability must be kept within a tight range to maintain optimal survival.

Before our study, very little was known about the effects of carbamazepine or other use-dependent Na⁺ channel inhibitors on pancreatic β-cells. Interestingly, it was recently shown that carbamazepine can rescue the trafficking defects of mutant K_ATP channels involved in congenital hyperinsulinism (59). Whether these effects of carbamazepine are involved in its ability to protect β-cells from apoptosis remains to be elucidated, but with the lack of changes in insulin secretion upon carbamazepine exposure, we do not favor a major role for this pathway in the effects observed in our study. In our view, it is more likely that carbamazepine and lidocaine protect β-cells via direct actions on Na⁺ channels and membrane excitability, similar to what may occur in neurons. Our studies suggest that inhibition of Na⁺ channels could reduce excessive membrane depolarization, thereby preventing the cytokine-stimulated influx of Ca²⁺ through voltage-gated Ca²⁺ channels. In addition, carbamazepine treatment extended the ER Ca²⁺ depletion time before cell death and improved the ER Ca²⁺ depletion threshold, which could have contributed to the observed down-regulation of ER stress signaling. Additional mechanistic studies into the prosurvival effects of carbamazepine are warranted.

In conclusion, we report the results of an unbiased high-throughput, high-content multiparameter screen that identified 19 drugs capable of protecting β-cells from cytokine-induced apoptosis. The unexpected observation that carbamazepine and lidocaine protected primary β-cells from apoptosis afforded new insight into the role of Na⁺ currents in apoptosis. Although, the current study focused on the validation of use-dependent Na⁺ channel inhibitors as antiapoptotic agents, additional insights are likely to come from the pursuit of other hits from our screen. The therapeutic impact for diabetes is intriguing especially because carbamazepine and many of the hits are already approved for the treatment of patients. Undoubtedly, these studies will continue to paint a complex picture of programmed cell death in the pancreatic β-cell.

Note Added in Proof

While our manuscript was under review, Cnop et al reported that carbamazepine protected the INS-1 beta-cell line from palmitate-induced death (Diabetes Epub Dec 30 2013).

Funding Statement

This work was supported by grants from the JDRF and the Stem Cell Network (to J.D.J.). Y.H.C.Y. was supported by a scholarship from the Natural Sciences and Engineering Research Council of Canada.