High-content screening identifies a role for Na+ channels in insulin production

Insulin production is the central feature of functionally mature and differentiated pancreatic β-cells. Reduced insulin transcription and dedifferentiation have been implicated in type 2 diabetes, making drugs that could reverse these processes potentially useful. We have previously established ratiometric live-cell imaging tools to identify factors that increase insulin promoter activity and promote β-cell differentiation. Here, we present a single vector imaging tool with eGFP and mRFP, driven by the Pdx1 and Ins1 promoters, respectively, targeted to the nucleus to enhance identification of individual cells in a high-throughput manner. Using this new approach, we screened 1120 off-patent drugs for factors that regulate Ins1 and Pdx1 promoter activity in MIN6 β-cells. We identified a number of compounds that positively modulate Ins1 promoter activity, including several drugs known to modulate ion channels. Carbamazepine was selected for extended follow-up, as our previous screen also identified this use-dependent sodium channel inhibitor as a positive modulator of β-cell survival. Indeed, carbamazepine increased Ins1 and Ins2 mRNA in primary mouse islets at lower doses than were required to protect β-cells. We validated the role of sodium channels in insulin production by examining Nav1.7 (Scn9a) knockout mice and remarkably islets from these animals had dramatically elevated insulin content relative to wild-type controls. Collectively, our experiments provide a starting point for additional studies aimed to identify drugs and molecular pathways that control insulin production and β-cell differentiation status. In particular, our unbiased screen identified a novel role for a β-cell sodium channel gene in insulin production.

Insulin production is the central feature of functionally mature and differentiated pancreatic β-cells. Reduced insulin transcription and dedifferentiation have been implicated in type 2 diabetes, making drugs that could reverse these processes potentially useful. We have previously established ratiometric live-cell imaging tools to identify factors that increase insulin promoter activity and promote β-cell differentiation. Here, we present a single vector imaging tool with eGFP and mRFP, driven by the Pdx1 and Ins1 promoters, respectively, targeted to the nucleus to enhance identification of individual cells in a high-throughput manner. Using this new approach, we screened 1120 off-patent drugs for factors that regulate Ins1 and Pdx1 promoter activity in MIN6 β-cells. We identified a number of compounds that positively modulate Ins1 promoter activity, including several drugs known to modulate ion channels. Carbamazepine was selected for extended follow-up, as our previous screen also identified this use-dependent sodium channel inhibitor as a positive modulator of β-cell survival. Indeed, carbamazepine increased Ins1 and Ins2 mRNA in primary mouse islets at lower doses than were required to protect β-cells. We validated the role of sodium channels in insulin production by examining Nav1.7 (Scn9a) knockout mice and remarkably islets from these animals had dramatically elevated insulin content relative to wild-type controls. Collectively, our experiments 2015 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. 1

. Introduction
Insulin production and secretion from pancreatic β-cells is carefully regulated to sustain normoglycaemia. However, type 1 or type 2 diabetes develops when circulating insulin levels drop below the threshold that marks the body's minimum requirements for glucose homeostasis. In type 1 diabetes, this occurs when β-cells are almost entirely destroyed via autoimmune attack [1]. In type 2 diabetes, β-cell exhaustion is thought to occur in the context of high-demand coincident with systemic insulin resistance [1,2]. We have also shown that insulin production, via effects on basal circulating insulin, is a key determinant of diet-induced obesity [3]. Thus, changes in insulin production have critical pathophysiological consequences.
A defining feature of a fully differentiated β-cell is robust insulin production. This truism has guided efforts to produce β-cell replacements from human embryonic stem cells [4]. Several groups, including ours, have recently explored the concept of β-cell dedifferentiation [5][6][7][8][9][10][11][12][13][14], and elegant in vivo work has established β-cell dedifferentiation as a new mechanism for β-cell failure in uncontrolled type 2 diabetes [15][16][17]. We have shown previously that the differentiation state can be measured in real-time at the single-cell level using a fluorescent reporter of Ins1 promoter activity in rodent and human β-cells [5,7,9,18]. Specifically, our strategy employed a single 'ratiometric' dual-reporter lentivirus wherein eGFP was driven by a fragment of the Ins1 promoter and mRFP was driven by a fragment of the Pdx1 promoter [7]. However, this original dual-reporter construct employed untargeted fluorescent proteins that distributed across the whole cell cytoplasm and nucleus, limiting its use for applications requiring automated cell counting. Thus, in this study, we sought to generate an improved imaging tool for assessing insulin production and differentiation state in living β-cells.
Here, we report the development of an improved Ins1/Pdx1 dual-reporter lentivirus and validate it against our previously published tool. We employ this new reporter construct in a multi-parameter, high-content screen of off-patent drugs and identify several small molecules capable of regulating β-cell maturity and/or insulin gene expression, without significant effects on cell viability. Notably, we identified carbamazepine, an ion channel-modulating drug we have previously shown to have favourable effects on β-cell survival, as a small molecule capable of driving Ins1 promoter activity in MIN6 cells and increasing insulin mRNA expression in primary islets. Collectively, our observations form the basis for additional studies on numerous factors, including carbamazepine.

Construct design and cloning
The original pTigerPdx1mRFP/Ins1eGFP construct was cloned as previously described [7]. To generate the new nuclear-targeted pTigerPdx1mRFP-NLS/Ins1eGFP-NLS vector, the nuclear localization signal (NLS) from simian virus large T-antigen (sequence: 'GATCCAAAAAAGAAGAAAGGTA) was synthesized in tandem using custom oligos (Integrated DNA Technologies, IA, USA) flanked by compatible ends for downstream ligations. The single-stranded oligos were annealed and incorporated at the 3 ends of each of the mRFP and eGFP sequences preceding the stop codons in the original vector.
Fluorescence-activated cell sorting (FACS) was conducted on an Influx sorter (Cytopeia) at the Life Sciences Institute core facility. Cells were excited with a 488 nm laser (530/40 emission) and a 561 nm laser (610/20 emission).
The Prestwick Library of 1120 drugs includes a diverse array of chemicals, including off-patent drugs (www.prestwickchemical.com). Compounds were pinned from 96-well round bottom plates (Corning) at approximately 15 μM using a 0.7 mm diameter 96-pin tool equipped onto a Biorobotics Biogrid II robot. The final compound concentration was approximately 8.5 μM, and we recognize that this would not have been an ideal dose for all drugs in the library.
Insulin content was measured after acid ethanol extraction in the same number of islets of similar size from approximately 1 year old female Nav1.7 (Scn9a) knockout mice, which have previously been characterized [24].

Electrophysiology
HEK 293 cells were cultured in Quantum 286 complete medium with L-glutamine and penicillin/ streptomycin. One day before transfection, they were seeded on polylysine-coated glass coverslips. The sodium channel Nav1.7 (NM_002977.3) was transiently transfected with a pcDNA3 expression vector (total cDNA 0.225 μg/coverslip) by using Lipofectamine TM 2000 according to the manufacturer's guide (Life Technologies, Thermo-Fisher). After transfection, the cells were grown at 37 • C, 5% CO 2 for 40-48 h before beginning electrophysiological experiments. Islets were isolated according to previously published protocols [25].
Currents were recorded at room temperature from HEK 293 or single islet cells using the standard whole-cell technique. The measurements were performed using an EPC-9 patch-clamp amplifier (HEKA Electronics, Lambrecht/Pflaz, Germany) and PULSE (v. 8.54) software. Currents were compensated for capacitive transients with the built-in compensation. No leak compensation was done. Patch pipettes were pulled from borosilicate glass with a resistance of 2.8-4.0 MΩ when filled with the pipette solution. The pipette solution contained (mM): 120 CsCl, 2 MgCl 2 , 10 EGTA, 10 TEA-C, 10 HEPES, pH 7.2. The standard extracellular medium consisted of (mM): 150 NaCl, 2 KCl, 2 CaCl 2 , 1 MgCl 2 , 10 HEPES, 10 glucose, pH 7. 4. Carbamazepine was dissolved as stock solution in DMSO and diluted to the final concentration in extracellular solution on each experimental day. Currents were measured by standard 100 ms depolarizations from a holding potential of −80 mV (HEK 293 cells) or −120 mV (islet cells). The steady-state inactivation properties of sodium currents were studied using a conventional two-pulse protocol in which a 30 ms test depolarization to −10 or 0 mV was preceded by a 100 ms conditioning pulse (V pre ) to voltages between −140 and 0 mV (HEK 293 cells) or between −180 and −40 mV (islet cells). Data for steady-state inactivation were fitted by a Boltzmann equation: where I/I max is the relationship between the current amplitudes (I) and the maximal current amplitude at the most negative V pre (I max ), A is the saturated level of I/I max , V 0.5 is the potential at which I/I max is half-maximal and k represents the slope factor. Dose-inhibition curves were calculated by a Hill equation. Electrophysiological data were analysed with ORIGIN v. 6.1 (OriginalLab Corp., Northampton, MA, USA).

Statistical analysis
Data of replicated experiments are expressed as mean ± s.e.m., unless otherwise indicated. Results were considered statistically significant when p < 0.05 using Student's t-test. 3. Results

Development of a nuclear-targeted Ins1/Pdx1 dual-reporter lentivirus
Fluorescent proteins driven by known promoter sequences provide windows into gene expression at the single-cell level [8]. The inclusion of green fluorescent and red fluorescent promoter reporters on the same lentiviral backbone in our first construct ensured that each cell had the potential to express both fluorescent proteins and permitted ratiometric analyses of signals with a known stoichiometry [5][6][7]9,18]. To improve upon our original dual reporter, especially for automated cell counting applications, we targeted eGFP and mRFP to the nucleus by way of a standard NLS (figure 1a). Consistent with the behaviour of the original dual-reporter lentivirus, FACS purification of cells infected with the nuclear-targeted dual-reporter displaying only Pdx1 promoter activity (i.e. mRFP-positive without significant eGFP signal) had significantly less Ins1 and Ins2 mRNA when compared with cells expressing robust Ins1 promoter activity (figure 1b,c). As expected, the nuclear-targeted dual-reporter construct was advantageous for delineating single cells, even in highly clumped cultures and at low magnification (figure 1d). Based on these studies, we have confidence that the nuclear-targeted Ins1/Pdx1 dual promoter reporter behaves in a similar fashion to our original imaging tool, with the notable improvement of nuclear targeting.

High-content screening for drugs that alter Ins1 and/or Pdx1 promoter activity
We assessed the utility of this new tool by conducting a three-parameter high-content screen to identify drugs that might increase insulin expression. We chose MIN6 cells for this high-content screen because they express lower levels of insulin relative to primary cells in our hands. Others have employed subdifferentiated cell models in similar screens [26], as it is easier to see increases in Ins1 promoter activity under such conditions. We chose the Prestwick library of off-patent drugs because mechanisms of action are known for many of the compounds [23]. Prior mechanistic knowledge has the potential to provide insights into the basic biology of insulin expression in β-cells.
MIN6 cells stably expressing the nuclear-targeted Ins1/Pdx1 dual promoter reporter were continuously cultured in a low concentration of Hoechst 33342 during drug treatment (figure 2a), which we have previously shown does not have significant effects on cell survival [27]. These cells were treated with 1120 drugs as well as an array of controls. These controls included: DMSO controls,    drug 'hits' that increase Pdx1 + /Ins1 + mature cells Figure 2. Self-organizing response groupings of a three-parameter high-content screen to assay Ins1 promoter activity, Pdx1 promoter activity and cell proliferation/survival. (a) Experimental design for a three-parameter, high-throughput imaging screen using MIN6 βcells. (b) Self-organizing heat maps of nine result groups generated with ACUITYXPRESS software. In these heat maps, red represents high and green represents low. Screen hits were chosen based on their clustering with positive controls. For example, follistatin is a positive control for a factor that increases Ins1 promoter activity and β-cell differentiation. The blow-up of group 4 (bottom panel) contains the hits that were followed up in this study.
ACUITYXPRESS software generated self-organizing maps that took into account the overall properties of each treatment, including cell number, percentage of Pdx1-positive but Ins1-low cells ('immature cells' [7]), and percentage of Pdx1-positive and Ins1-high cells ('mature cells' [7]), and then clustered similar treatments together (figure 2b). This approach resulted in nine response groupings. For example, response group 1 contained treatments that increased cell number, but were mostly low or neutral with regards to Ins1 or Pdx1 promoter activity. Response groups 2, 3, 5, 6 and 8 contained drugs with largely neutral effects on the parameters tested. Response group 7 contained treatments that increased Ins1 promoter activity, but that also had deleterious effects on cell viability. We focused on response group 4 which contained 117 treatments that nominally increased Ins1 promoter activity independent of changes in cell number. We specifically focused on the top approximately one-third of response group 4 where changes in Ins1 promoter activity were particularly robust (figure 2b, bottom panel). This approach identified dapsone, allantoin, minaprine dihydrochloride, nifedipine, amitryptiline hydrochloride, metformin, carbamazepine as drugs of interest that cluster with positive controls such as 10% FBS and follistatin, which we have previously shown can increase Ins1 promoter activity in β-cells [5].
We next manually examined the Ins1 promoter activity and Pdx1 promoter activity data as a function of total cell number ( figure 3). The data were normalized around the DMSO controls. We looked for treatments with Ins1 promoter activity outside of two median absolute deviations (2MAD; figure 3a), but with relatively normal Pdx1 promoter activity (figure 3b), indicative of a high Ins1/Pdx1 promoter activity ratio. Dapsone, minaprine and todralazine increased Ins1 promoter activity while having modest positive effects on viability. Carbamazepine, isoconazole, allantoin, amitryptiline, nifedipine, metformin and ascorbic acid all increased Ins1 promoter activity by these criteria, while having neutral effects on β-cell number (i.e. viability). Thus, there was excellent overall agreement between the two approaches. Electronic supplementary material, table S1 provides a full list of screen results.
As we have previously shown [23], a substantial portion of drugs in the Prestwick library of compounds is toxic to pancreatic β-cells (figure 3c). Interestingly, several factors, including cycloheximide, the established inhibitor of protein synthesis, increased Ins1 promoter activity despite causing robust β-cell death. A few of the drugs (e.g. merbromin) are fluorescent, which interfered with the assay, and were ignored.

Follow-up validation of drugs that increase insulin expression
In order to determine whether increases in Ins1 promoter activity could be reflected in real changes in insulin mRNA, we conducted follow-up studies using validated qPCR primers [5,7,9,18]. For these experiments, we employed primary mouse islet cells, which represent a more physiologically relevant cell model than the MIN6 cell line. We examined the mRNA expression of both rodent insulin genes, Ins1 and Ins2, as well as Pdx1, a transcription factor that supports insulin transcription, and is critical for adult β-cell function and survival [28,29]. Of the drugs chosen for follow-up, nine of 10 increased insulin mRNA expression in primary mouse islets ( figure 4). Allantoin, isoconazole, minaprine and carbamazepine showed statistically significant increases in Ins1 and/or Ins2 mRNA expression above the DMSO control (figure 4). Carbamazepine increased Ins1, Ins2 and Pdx1 mRNA expression in these experiments ( figure 4). Together, these results suggest that our nuclear-targeted Ins1/Pdx1 dual-reporter system can identify novel regulators of insulin production.
The observation that carbamazepine increased insulin expression was interesting, because we have recently implicated the same drug as protective in the context of a toxic cytokine cocktail [23]. Thus, we performed a more detailed dose-response analysis of the effects of carbamazepine on Ins1, Ins2 and Pdx1 mRNA expression in primary mouse islets. These experiments showed that carbamazepine increased Ins1 mRNA in primary islets even at low doses (figure 5), orders of magnitude below the doses (100 μM) at which this drug has consistent, statistically significant effects on primary β-cell survival [23]. Thus, in our studies, carbamazepine may be a more potent regulator of insulin expression than of β-cell survival.

Carbamazepine directly inhibits the Nav1.7 sodium channel isoform
We and others have recently characterized the expression and function of sodium channels in mouse pancreatic β-cells [23,24]. Gene expression analysis in purified β-cells revealed that the gene encoding Nav1.7 (Scn9a) was the most highly expressed of these channels [23,24]. Here, we determined that carbamazepine inhibits Nav1.7 Na + currents ( figure 6). Specifically, Nav1.7 channels expressed in HEK293 cells were dose-dependently and voltage-dependently inhibited by carbamazepine. The steadystate inactivation curve was shifted by more than 30 mV to hyperpolarized potentials in the presence of 300 μM carbamazepine ( figure 6a,b). The half-maximal inactivation of Nav1.7 current in HEK 293 cells was IC 50 = 130 μM. Similar observations under identical conditions were made in normal mouse islet cells. This effect was also voltage-dependent as shown by shifting the holding potential in islets from −120 to 80 mV ( figure 6c,d).    in (a,b), which show 2xMAD (median absolute deviation), were then assessed as possible hits. Hits were chosen for their effects on several outputs such as inducing β-cell maturation by increasing the relative number of mature Pdx1 + /Ins1 + cells (top) and concomitantly reducing the relative number of immature Pdx1 + /Ins1 low cells (middle). Positive controls are marked in orange. 3. 5

. Insulin production in sodium channel knockout mice
We reasoned that if inhibition of β-cell sodium channels stimulates insulin gene expression in vivo, then mice lacking the most abundant sodium channel isoform would have accumulated increased levels of insulin protein within their islets. Indeed, and quite remarkably, isolated islets from Nav1. 7   mice had more than a threefold higher level of insulin content ( figure 7). The striking elevation of insulin content contrasts with a relatively minor inhibition of glucose-stimulated insulin release from Nav1.7 knockout islets [24]. This observation strongly suggests that sodium channel activity specifically and negatively modulates insulin production in vivo over the long term.

Discussion
The goals of this study were to develop and validate an improved high-content screening-compatible Ins1/Pdx1 ratiometric lentiviral reporter, and to use this fluorescent probe to identify drugs capable of year old, wild-type or Nav1.7 (Scn9a) knockout mice (n = 5, 6). * p < 0.05.
modulating pancreatic β-cell differentiation and maturation. Indeed, our nuclear-targeted fluorescent reporters improved the ability to identify and count single cells, while retaining the fidelity of the original construct [5,7,9]. We expect that this improved dual-reporter lentivirus has the potential to be also useful for many studies on the fundamental biology of β-cell state and heterogeneity. Our high-content screening results demonstrate the use of this live-cell imaging tool for the identification of factors that increase insulin expression, and point to previously unexpected roles for ion channels in β-cell function. Insulin production is a cardinal function of pancreatic β-cells and it is under tight control at transcriptional, translational and post-translational levels [30][31][32]. Insulin production is modulated by nutrients including glucose, and perhaps by insulin itself [29][30][31][33][34][35][36]. Here, we present evidence that several pharmacological compounds can affect insulin production. In many cases, it is the first evidence that a specific drug class that can affect β-cells. We observed statistically significant stimulation of Ins1 mRNA expression with allantoin, isoconazole, minaprine and carbamazepine. Allantoin is also referred to as 5-ureidohydantoin or glyoxyldiureide, and is best known as a component of cosmetics. Isoconazole is an antifungal drug. Minaprine is a discontinued anti-depressant drug thought to work via reversible inhibition of monoamine oxidase A [37]. It is not likely that these three drugs have a path to clinical application for in vivo diabetes therapy, although the pathways they target might eventually be exploited in vitro to increase insulin yields for cell therapy applications.
Carbamazepine, on the other hand, may have therapeutic potential. Carbamazepine is a usedependent inhibitor of sodium channels [23], and we have confirmed that it has this function in pancreatic β-cells. Mouse pancreatic β-cells have a relatively unique complement of sodium channel genes, with Nav1.7 (Scn9a) being the most prominently expressed mRNA [23,24]. Importantly, it appears that it may be possible to target these channels pharmacologically with only relatively modest effects on glucose-stimulated insulin secretion [23,24,[38][39][40][41][42]. Indeed, this hypothesis was further supported by our observation that islets from lifelong Nav1.7 (Scn9a) knockout mice had increased insulin content, but relatively normal insulin secretion [24]. We have previously shown that carbamazepine, at concentrations much higher than those that modulate insulin production in this study, does not exhibit a robust, acute effect on insulin secretion in primary mouse islets using the gold-standard perifusion assay [23]. Thus, while chronically reduced insulin production can indirectly reduce insulin secretion [3,34], the converse is not true in the cases of carbamazepine and Nav1.7 knockout that increase insulin production, but do not increase insulin secretion [23,24]. Notwithstanding, a recent luciferase-based high-throughput screen identified Na + channel modulators as potential regulators of insulin secretion in insulinoma cell lines [43], indicating that this area requires further study. It should be noted that a series of studies have shown that carbamazepine can have direct effects on K ATP channels [44,45]. Given the importance of K ATP channels in β-cell physiology [40], it is likely that this drug acts through multiple mechanisms to affect this cell type.
Collectively, our studies suggest that carbamazepine or derivatives, regardless of the exact molecular mechanism or mechanisms, may hold promise for increasing insulin production, promoting β-cell differentiation, and/or protecting β-cells from stress. Additional future work in this area, including in vivo experiments, is warranted.