Na+ current properties in islet α- and β-cells reflect cell-specific Scn3a and Scn9a expression

Mouse pancreatic β- and α-cells are equipped with voltage-gated Na+ currents that inactivate over widely different membrane potentials (half-maximal inactivation (V0.5) at −100 mV and −50 mV in β- and α-cells, respectively). Single-cell PCR analyses show that both α- and β-cells have Nav1.3 (Scn3) and Nav1.7 (Scn9a) α subunits, but their relative proportions differ: β-cells principally express Nav1.7 and α-cells Nav1.3. In α-cells, genetically ablating Scn3a reduces the Na+ current by 80%. In β-cells, knockout of Scn9a lowers the Na+ current by >85%, unveiling a small Scn3a-dependent component. Glucagon and insulin secretion are inhibited in Scn3a−/− islets but unaffected in Scn9a-deficient islets. Thus, Nav1.3 is the functionally important Na+ channel α subunit in both α- and β-cells because Nav1.7 is largely inactive at physiological membrane potentials due to its unusually negative voltage dependence of inactivation. Interestingly, the Nav1.7 sequence in brain and islets is identical and yet the V0.5 for inactivation is >30 mV more negative in β-cells. This may indicate the presence of an intracellular factor that modulates the voltage dependence of inactivation.

r αand β-cells express both Na v 1.3 and Na v 1.7 Na + channels but in different relative amounts. r The differential expression explains the different properties of Na + currents in αand β-cells. r Na v 1.3 is the functionally important Na + channel α subunit in both αand β-cells. r Islet Na v 1.7 channels are locked in an inactive state due to an islet cell-specific factor.
Mouse pancreatic βand α-cells are equipped with voltage-gated Na + currents that inactivate over widely different membrane potentials (half-maximal inactivation (V 0.5 ) at −100 mV and −50 mV in βand α-cells, respectively). Single-cell PCR analyses show that both αand β-cells have Na v 1.3 (Scn3) and Na v 1.7 (Scn9a) α subunits, but their relative proportions differ: β-cells principally express Na v 1.7 and α-cells Na v 1.3. In α-cells, genetically ablating Scn3a reduces the Na + current by 80%. In β-cells, knockout of Scn9a lowers the Na + current by >85%, unveiling a small Scn3a-dependent component. Glucagon and insulin secretion are inhibited in Scn3a −/− islets but unaffected in Scn9a-deficient islets. Thus, Na v 1.3 is the functionally important Na + channel α subunit in both αand β-cells because Na v 1.7 is largely inactive at physiological membrane potentials due to its unusually negative voltage dependence of inactivation. Interestingly, the Na v 1.7 sequence in brain and islets is identical and yet the V 0.5 for inactivation is >30 mV more negative in β-cells. This may indicate the presence of an intracellular factor that modulates the voltage dependence of inactivation.

Introduction
Insulin and glucagon are the principal hormones involved in glucose homeostasis. They are secreted from the βand α-cells, respectively, of the pancreatic islets in response to changes in plasma glucose levels, with insulin secretion being stimulated and glucagon secretion being suppressed by glucose elevation . The regulation of their release is impaired in diabetes mellitus: insulin secretion is insufficient and glucagon levels are inappropriately elevated (Dunning et al. 2005;Ashcroft & Rorsman, 2012). In both cell types, electrical activity is crucial for hormone release . This electrical activity is orchestrated by the concerted activity of several different types of ion channel (Rorsman et al. 2011).
Voltage-gated Na + channels play a central role in cellular excitability (Hille, 2001). They are dually influenced by the membrane potential. A large rapid depolarization results in Na + channel opening and an inward Na + current. However, if depolarization is sustained, the channel enters a non-conducting 'inactivated' state, resulting in a decline in current flow. Thus, the magnitude of the Na + current depends on the potential at which the membrane normally sits.
Mammalian Na + channels consist of a pore-forming α subunit and auxiliary β subunits that modify channel gating (Catterall, 2000). There are nine different genes (Scn1a-5a and Scn8a-11a) that encode Na + channel α subunits (Catterall et al. 2005). The channels resulting from these genes are named Na v 1.1-Na v 1.9 (Goldin et al. 2000). Confusingly, the numbers of the genes and the protein do not always match. Thus, Na v 1.6 and Na v 1.7 are encoded by the genes Scn8a and Scn9a, respectively (Goldin et al. 2000). Currently, four isoforms of the β subunits are known: referred to as Scn1b, Scn2b, Scn3b and Scn4b (Koopmann et al. 2006). With few exceptions (Na v 1.5, Na v 1.8 and Na v 1.9; Catterall et al. 2005), voltage-gated Na + channels are sensitive to the neurotoxin TTX.
Pancreatic islet cells are equipped with TTX-sensitive voltage-gated Na + channels (Gopel et al. 2000a,b;Vignali et al. 2006). Na + channels are critical for glucose-induced electrical activity and insulin secretion in human β-cells (Pressel & Misler, 1990;Braun et al. 2008). However, their role in rodent β-cells is puzzling because they exhibit an arcane voltage dependence of inactivation. Surprisingly, when holding at membrane potentials found under physiological conditions, depolarisation evokes almost no Na + current in β-cells (Plant, 1988;Gopel et al. 1999;Lou et al. 2003). By contrast, large Na + currents can be evoked in α-cells held at physiological membrane potentials and blocking these channels with TTX slows the upstroke of the action potential and inhibits glucagon secretion (Zhang et al. 2013).
It has been proposed that the different inactivation properties of Na + currents in αand β-cells are attributable to different α or β subunit composition (Lou et al. 2003). Here, we tested this hypothesis using a combination of biophysical measurements of islet cell Na + currents and single-cell PCR analyses. The functional significance of the different Na + channel α subunits on electrical activity and hormone secretion was dissected using mice in which specific α subunits were genetically deleted.
Our data confirm the important role of Na + channels in glucagon secretion, and provide evidence for a previously unrecognised role for Na + channels in insulin secretion. They also raise the possibility that β-cells contain a factor that shifts the voltage dependence of inactivation of Na v 1.7 Na + channels to hyperpolarised levels, rendering the channel functionally inactive at physiological membrane potentials.

Ethics
All animal experiments were conducted in accordance with the UK Animals Scientific Procedures Act (1986) and University of Oxford ethical guidelines. Human pancreatic islets were isolated with ethical approval and clinical consent at the Diabetes Research and Wellness Foundation Human islet Isolation Facility (Oxford).
One series of experiments (see Fig. 5A) was performed on tdRFP-negative cells in islets from mice expressing the red fluorescent protein (tdRFP) under the control of the somatostatin promoter (see supplementary material in Chera et al. 2014). It was ascertained that tdRFP-expressing cells contain somatostatin.

Preparation of pancreatic islets
Mice were killed by cervical dislocation, the pancreases quickly removed and islets isolated by collagenase digestion. Islets were used for acute experiments and were not maintained in tissue culture for more than 16 h.

Whole-cell patch-clamp recordings
Whole-cell Na + currents were recorded in αand β-cells within intact islets using the standard whole-cell configuration as previously described (Gopel et al. 1999). The measurements were performed using EPC-9 or EPC-10 patch-clamp amplifiers and Pulse software (HEKA Electronics, Lambrecht/Pfalz, Germany). Currents were compensated for capacitive transients and linear leak using a −P/4 protocol. The currents were filtered at 2.9 kHz and digitised at >10 kHz.

Hormone measurements
Batches of 10-20 size-matched islets were preincubated in 0.3 ml of a modified KRB containing 1 mM glucose and 2 mg ml −1 BSA for 1 h at 37°C, followed by a 1 h test incubation in 0.3 ml of the same medium supplemented with glucose, receptor antagonists, metabolic inhibitors and ion channel activators/blockers as indicated. Insulin and glucagon were assayed using Millipore RI-3K (Billerica, MA, USA) and Eurodiagnostica RB 306 (Malmö, Sweden) radioimmunoassays, respectively. The somatostatin receptor (SSTR) antagonist BIM23056 was obtained from Tocris (Bristol, UK).

Pancreas perfusion
Dynamic measurements of insulin secretion were performed using in situ pancreatic perfusion. Briefly, the aorta was cannulated by ligating above the coeliac artery and below the superior mesenteric artery, and the pancreas was perfused with KRB solution at a rate of ß0.45 ml min −1 using an Ismatec (Glattbrugg, Switzerland) Reglo Digital MS2/12 peristaltic pump. The perfusate was maintained at 37°C with a Warner Instruments temperature control unit TC-32 4B in conjunction with a tube heater (Warner Instruments P/N 64-0102, Hamden, CT, USA) and a Harvard Apparatus (Holliston, MA, USA) heated rodent operating table. The effluent was collected, using a Teledyne (Thousand Oaks, CA, USA) ISCO Foxy R1 fraction collector, by cannulating the portal vein. The pancreas was first perfused for 20 min with 1 mM glucose before commencing the experiment to establish the basal rate of secretion.

[Ca 2+ ] i imaging
Confocal [Ca 2+ ] i imaging experiments were conducted essentially as previously reported (Girard et al. 2009). Islets were superfused with the KRB described above plus glucose and TTX or tolbutamide as indicated, at a flow rate of 1 ml min −1 and 37°C.

Identification of αand β-cells in intact islets
In all voltage-clamp experiments, the identity of the α-cells was established by immunocytochemistry following injection of the cell with biocytin (0.5 mg ml −1 ) via the recording electrode (Zhang et al. 2007). In the membrane potential recordings, α-cells were identified based on their spontaneous action potential firing with 1 mM glucose; β-cells were identified as cells that were electrically silent and hyperpolarized with 1 mM glucose and had characteristic oscillatory electrical activity when exposed to 8 mM glucose (Gopel et al. 2000a). Cells identified by these criteria had Na + channel inactivation properties characteristic of αand β-cells, as indicated by voltage-clamp measurements in cells identified by immunocytochemistry. For [Ca 2+ ] i measurements, β-cells were identified by the induction of [Ca 2+ ] i oscillations at 6 mM glucose that convert into a (more) sustained [Ca 2+ ] i elevation at 20 mM glucose.

RNA purification and quantitative RT-PCR
RNA purification and quantitative RT-PCR in mouse islet cells were carried out as described elsewhere (Braun et al. 2009. Details of primer sequences used will be provided on request. Some experiments were performed using mRNA collected from αand β-cell fractions purified by fluorescence-activated cell sorting, as described previously (De Marinis et al. 2010) Preamplification and single-cell RT-PCR Individual cells were collected using 2 μl 0.5% NP-40 lysis buffer. cDNA was synthesized as above with a reaction volume of 10 μl (Bengtsson et al. 2005;Stahlberg & Bengtsson, 2010). Single-cell cDNA was preamplified for 10 cycles with the TaqMan PreAmp Master Mix Kit (PN4364130, Applied Biosystems, Foster City, CA, USA) and TaqMan assays according to the manufacturer's instructions. In the subsequent quantitative PCR (qPCR), 2.5 μl preamplification product diluted 1:20 was used. The presence of transcript was defined as a positive signal from the TaqMan probe after 40 PCR cycles. Amplicon size was confirmed by agarose gel electrophoresis.

Identification of Scn9a splice variants in mouse islets
Total RNA purified from mouse islets and brain was reverse-transcribed using a High Capacity RNA-to-cDNA Kit (Applied Biosystems). PCR was performed with Scn9a gene-specific primers and the resulting PCR products were cloned using a Zero Blunt TOPO PCR cloning kit (Invitrogen, Carlsbad, CA, USA) and sequenced.

Data analysis
All data are given as mean values ± SEM of the indicated number of experiments (n).
All electrophysiological experiments were performed on the indicated number of cells obtained (unless otherwise stated) from at least three different animals (usually more). Current amplitudes have been normalised to cell capacitance and are expressed as pA pF -1 . Inactivation was analysed by applying 200 ms conditioning pulses (from −180 to −5 mV) followed by a test pulse to 0 mV (applied at a frequency of 1 Hz). Peak current amplitude during the test pulse (I) was normalized to that observed when the conditioning pulse went to −180 mV (I max ). Data are presented as h Ý (= I/I max ). The Na + current activation parameters were determined as described previously (Gonoi & Hille, 1987): the Na + conductance (G) was calculated from the measured Na + current amplitudes (I) by the relation G = I/(V − V r ) where V is the membrane potential and V r is the estimated reversal potential (Nernst potential), calculated from the extra-and intracellular Na + concentrations in media used. Values of G at a given V were normalised to the maximal G (G max ). Inactivation and activation curves were fit to single or double Boltzmann functions as appropriate. Peak current amplitudes, the activation (τ m ) and inactivation (τ h ) time constants, and inactivation properties were estimated using the electrophysiology analysis software package Pulsefit (Heka Electronics. Lamprecht/Pfalz, Germany).
Action potential peak voltages were measured as reported elsewhere (Zhang et al. 2013); the number of action potentials for each experimental condition/cell averaged >40.
Statistical significances were evaluated using Student's t test or ANOVA (for multiple comparisons), as appropriate.

Molecular characterization of Na + channel subunits in mouse and human pancreatic islets
In mouse pancreatic islets, Scn9a was the dominant α subunit, being expressed at levels 6-fold higher than Scn3a and Scn8a. Of the β subunits, only Scn1b and Scn3b were detected (Fig. 1A). The high expression of Scn9a and Scn1b in islets is in agreement with a previous report (Ernst et al. 2009).
We performed single-cell PCR to determine which subunits are expressed in αand β-cells, respectively (Fig. 1B). All cells used for this analysis possessed voltage-gated Na + currents, as confirmed by whole-cell patch-clamp recordings prior to mRNA collection. Yet most cells (40 out of a total of 52 β-cells and 24 out of a total of 37 α-cells, from three mice) lacked detectable mRNA for any of the Na + channel α subunits, perhaps due to low mRNA copy numbers and/or non-continuous gene transcription (Raj et al. 2006). In those α-cells in which transcripts were detected, Scn3a and Scn9a were found equally often. Importantly, 2 of the 13 α-cells contained mRNA for both Scn3a and Scn9a. In β-cells (Fig. 1B, lower), Scn9a was the most abundant transcript (4.5-fold more frequent than Scn3a). No β-cells and only one α-cell expressed Scn8a. Of the β subunits, Scn3b was found 2.7-fold more often than Scn1b in α-cells, whereas Scn1b predominated in β-cells (detected 4.5-fold more often than Scn3b).
We also measured Na + channel transcripts in pure αand β-cell fractions. In α-cells (Fig. 1C, top), Scn9a was the most abundant transcript but relatively high levels of Scn3a and Scn8a were also found. Among the β subunits, Scn3b was predominantly expressed (>4-fold higher than Scn1b). In β-cells (Fig. 1C, bottom), Scn9a was expressed at levels 7-to 20-fold higher than Scn8a and Scn3a, respectively, and Scn1b was expressed at 6-fold higher levels than Scn3b. In agreement with the whole-islet data, β subunits were expressed at much lower levels than Scn9a in both αand β-cell fractions. Thus, the data obtained from purified αand β-cell populations are in good agreement with those obtained from single αand β-cells.

Figure 1. Expression of Na + channel subunits in pancreatic islet cells
A, qPCR on mouse islets (n = 3 preparations). Data have been normalized to expression of Scn9a, which in turn was normalised against mean expression of ribosomal protein S29 (Rps29) and hypoxanthine guanine phosphoribosyl transferase 1 (Hprt). B, single-cell PCR analyses of α-(top) and β-cells (lower). Data are presented as the percentage of α-cells (or β-cells) containing the indicated Na + channel transcript and come from 13 α-cells (identified by the presence of glucagon mRNA), and 12 β-cells (containing insulin mRNA). Cells come from three mice. C, as in A but using pure α-(top) and β-cell fractions (lower).

Properties of Na + currents in mouse αand β-cells
As our PCR analyses indicated that αand β-cells may contain Na + channels of different molecular composition, we next investigated whether this might give rise to biophysically distinct Na + currents. All electrophysiological data reported here were obtained from identified α or β-cells in intact acutely isolated pancreatic islets. In β-cells, two types of responses were observed. In 70% of β-cells (7/10 cells), no Na + current was seen when the holding potential was −70 mV but large Na + currents were evoked when the cells were subsequently hyperpolarised to −180 mV. In the remaining β-cells (n = 3), voltage-gated Na + currents were elicited from −70 mV that increased only ß2-fold when the cells were hyperpolarised to −180 mV. Figure 2A compares the mean Na + I-V relationship evoked from holding potentials of −70 or −180 mV.
We analysed the voltage dependence of Na + current inactivation using a standard two-pulse protocol (Fig. 2B). In most β-cells (n = 7), inactivation was well described by a single Boltzmann function with half-maximal inactivation (V 0.5 ) at −97 ± 1 mV. However, in the subset of cells in which Na + currents could be evoked from −70 mV, inactivation was best described as a sum of two Boltzmann functions with V 0.5 at −97 ± 6 and −50 ± 1 mV, comprising 49 ± 10 and 51 ± 10% of the maximum current, respectively (n = 3). When the data from all 10 cells were pooled, inactivation was biphasic with V 0.5 at −95 ± 2 and −50 ± 1 mV, comprising 85 ± 8 and 15 ± 8% of the maximum current, respectively (Fig. 2B).
In α-cells, large Na + currents were invariably observed when holding at −70 mV and hyperpolarization to −180 mV only marginally increased (+20%) the peak current amplitude (Fig. 2C). Figure 2C compares the I-V relationships recorded when cells were held at −70 or −180 mV. A biphasic inactivation curve was observed in all 6 cells analysed (Fig. 2D). The more negative component had a V 0.5 of −84 ± 5 mV and accounted for only 26 ± 3% of the total Na + current. The principal component (comprising 74 ± 3%) had a V 0.5 of −47 ± 1 mV.
Also shown in Fig. 2B and D are the Na + current activation curves. In α-cells, activation was monotonic and half-maximal at −23 ± 2 mV (n = 6). In the subset of β-cells where part of the Na + current inactivated at more positive membrane potentials, activation was also monotonic and half-maximal at −23 ± 1 mV (n = 3). The corresponding value in β-cells where all Na + current inactivated at more negative voltages (V 0.5 ß−100 mV) was −24 ± 3 mV (n = 7). The activation curve shown in Fig. 2B is based on the data from all 10 β-cells.
The finding that all α-cells (and some β-cells) exhibit biphasic voltage-dependent Na + current inactivation suggests that Na + channel subunits with widely different inactivation properties are coexpressed in individual J Physiol 592.21 cells. This echoes the single-cell PCR measurements that provided direct evidence for the expression of both Scn3a and Scn9a within the same α-cell.
Families of voltage-clamp currents recorded from βand α-cells during membrane depolarisation from −180 to −20, −10 and 0 mV are shown in Fig. 3A and B. The voltage dependence of the time constants of activation (τ m ) and inactivation (τ h ) of the Na + currents in αand β-cells are compared in Fig. 3C and D. In both αand β-cells, inactivation became progressively more rapid (seen as a decrease in τ h ) as the cells were depolarised. The differences between αand β-cells were subtle but τ h was higher in β-cells at voltages above −10 mV.

Role of Na + channels in islet cell electrical activity: effects of TTX
The inactivation properties of the β-cell Na + currents suggest that Na + channels will be almost fully inactivated at physiological membrane potentials in β-cells and thus contribute little (if at all) to β-cell electrical activity and hormone secretion. By contrast, Na + channels are likely to be functionally more important in glucagon-secreting α-cells. We examined this idea by performing membrane potential recordings from αand β-cells within intact pancreatic islets, using the perforated patch configuration (in which cell metabolism remains intact).
Unlike β-cells, α-cells are electrically active at 1 mM glucose (Zhang et al. 2013). Figure 4A shows α-cell action potentials displayed on a slow (left) and expanded (right) time base under control conditions, during application of the Na + channel blocker TTX (0.1 μg ml −1 ) and following washout of the toxin. TTX consistently reduced the peak voltage of the action potentials: from +2 ± 5 mV (n = 6) to −12 ± 1 mV (P < 0.01). Thus, Na + channels are important for the upstroke of the action potential in α-cells.
In β-cells, TTX had more variable effects. In 3 out of 6 cells, TTX barely affected action potential firing elicited by 8 mM glucose (Fig. 4B). However, in the remaining three

. Properties of voltage-gated Na + currents in αand β-cells in intact islets
A, I-V relationships recorded from holding potentials of −70 mV (•) or −180 mV (•) in β-cells (n = 10). Cells were depolarised to voltages between −60 and +50 mV in 10 mV increments. Note that the more hyperpolarized holding potential leads to the generation of large Na + currents. B, voltage dependence of Na + current inactivation (h Ý ) and activation (G/G max ) in β-cells. The curves represent a double Boltzmann fit to the data (n = 10). C and D, as in A and B but experiments were conducted on α-cells (n = 6). * P < 0.05, * * P < 0.01, * * * P < 0.001 for differences between holding at -70 and -180 mV. β-cells, TTX markedly reduced action potential amplitude (Fig. 4C): peak voltage decreased from −8 ± 1 mV in the absence to −20 ± 3 mV in the presence of TTX (n = 3; P < 0.03, paired t test). When all six cells were averaged, peak voltage decreased from −18 ± 5 to −24 ± 2 mV, but this difference did not quite reach statistical significance (P < 0.07, paired t test).
In the presence of 20 mM glucose (Fig. 4D), when β-cells were permanently depolarised to a membrane potential of −47 ± 1 mV, TTX reduced the peak action potential voltage from −25 ± 7 to −29 ± 6 mV (n = 4; P < 0.05, paired t test). Note that whereas slow membrane potential oscillations sometimes persist in β-cells exposed to 20 mM glucose (as exemplified by the cell shown in Fig. 4D), the response to the higher glucose concentration more often consists of nearly uninterrupted action potential firing without lengthy repolarised electrically silent intervals (not shown but see Kanno et al. 2002)

Effects of TTX on glucose-induced [Ca 2+ ] i increases in β-cells and insulin secretion
In β-cells, action potential firing culminates in the opening of voltage-gated Ca 2+ channels. We tested the effects of glucose and TTX on β-cell [Ca 2+ ] i by confocal imaging in fluo4-loaded mouse islets (Fig. 5A). In most cells (104 out of 110 cells) in which [Ca 2+ ] i was low and stable at 1 mM glucose, increasing glucose to 6 mM did not induce any [Ca 2+ ] i oscillations whereas a subsequent increase to 20 mM elicited a sustained elevation of [Ca 2+ ] i (Fig. 5A, i).
However, in a small subset of cells (n = 6), [Ca 2+ ] i oscillations were evoked by 6 mM glucose. In these cells, addition of TTX reversibly reduced the frequency of the [Ca 2+ ] i oscillations from 2.0 ± 0.7 to 0.5 ± 0.2 min −1 (P < 0.05) and decreased the area under the curve (above baseline) by 72 ± 14% (P < 0.05). Increasing glucose further, from 6 to 20 mM, resulted in a more sustained elevation of [Ca 2+ ] i (Fig. 5A, ii). The frequency of the rapid [Ca 2+ ] i oscillations observed at threshold glucose concentrations, and their conversion to a non-oscillatory sustained increase at higher glucose levels, echos the effects of glucose on β-cell electrical activity (Henquin & Meissner, 1984). In intact islets, β-cells are electrically coupled by gap junctions and this synchronises electrical activity and [Ca 2+ ] i oscillations in a K ATP channel-dependent fashion (Rocheleau et al. 2006). Unlike what is seen at higher glucose concentrations , the [Ca 2+ ] i oscillations observed at 6 mM glucose were not well synchronised across the islet (Fig. 5B). It is possible that this is a consequence of the K ATP channel activity being slightly greater in the non-active cells and that the lower input resistance prevents synchronisation of electrical activity mediated by current spread via the gap junctions.
We acknowledge that we cannot exclude the possibility that some of the cells responding to 6 mM glucose were somatostatin-secreting δ-cells, which have a lower threshold for secretion than β-cells. To test this possibility, we performed experiments on islets from mice expressing red fluorescent protein (tdRFP) under the control of the somatostatin promoter. Some tdRFP-negative cells (likely to be β-cells) responded to 6 mM glucose with TTX-sensitive [Ca 2+ ] i oscillations that converted to a sustained elevation when glucose was subsequently increased to 20 mM (Fig. 5A, iii). The oscillatory pattern in tdRFP-positive cells (i.e. δ-cells) was rather different and consisted of larger and more regular [Ca 2+ ] i oscillations that were easily distinguished from those seen in β-cells. Interestingly, ß30% of tdRFP-positive δ-cells exhibited spontaneous [Ca 2+ ] i oscillations at 1 mM glucose (Fig. 5A,  iv). Collectively, the data suggest that a small number of β-cells are indeed active at 6 mM glucose.
We next examined the effects of TTX on insulin secretion using the perfused mouse pancreas preparation (Fig. 5C, D). Under control conditions, increasing glucose from 1 to 6 mM evoked a transient ß7-fold stimulation of insulin secretion (Fig. 5C, D). It is likely that this is released from the small subset of β-cells that show  [Ca 2+ ] oscillations in 6 mM glucose (Fig. 5A, ii and  iii). Subsequent stimulation with 20 mM glucose evoked an insulin secretory response that was >20-fold higher than that elicited by 6 mM glucose (Fig. 5C, D). This correlated with an 18-fold increase in the number of cells showing [Ca 2+ ] i oscillations. Thus, it seems that much of the increase in glucose-dependent insulin secretion reflects recruitment of additional β-cells. At both 6 and 20 mM glucose, insulin secretion was strongly inhibited by the Na + channel blocker TTX: the inhibitory effect averaged 74 ± 8 and 68 ± 7%, respectively (Fig. 5D).
The strong suppression of glucose-induced insulin secretion by TTX seen in the perfused pancreas experiments (Fig. 5D) contrasts with the weaker effect of TTX on glucose-induced insulin secretion found in static incubations of isolated islets (Fig. 5E, left). Somatostatin-secreting δ-cells possess large TTX-sensitive voltage-gated Na + currents (Gopel et al. 2000a). As expected, δ-cells identified by expression of tdRFP generated large (100 mV) over-shooting action potentials when exposed to 20 mM glucose, and these were reversibly suppressed by TTX (Fig. 5F). We reasoned that TTX-induced suppression of somatostatin secretion might result in relief from paracrine inhibition of insulin secretion in static incubations (Renstrom et al. 1996;Hauge-Evans et al. 2009). To test this hypothesis, we compared insulin secretion in the absence and presence of the SSTR subtype 5 (SSTR5) antagonist BIM23056 (SSTR5 is the predominant SSTR subtype in mouse β-cells; Strowski et al. 2003). The SSTR5 antagonist stimulated insulin secretion at 8 mM glucose by ࣙ50% (Fig. 5E, right). While TTX had little effect in the absence of BIM23056, it exerted a strong inhibitory effect (ࣙ60%) in its presence. By contrast, in the perfused mouse pancreas, BIM23056 did not increase glucose-induced insulin secretion (data not shown).
Collectively, the data of Figs 4 and 5 suggest that Na + channels play a previously unrecognised role in the initiation of β-cell electrical activity, [Ca 2+ ] i oscillations and insulin secretion. We next attempted to establish the molecular identity of these Na + channels. Na + currents in Scn3a +/and Scn3a -/αand β-cells Expression analysis (Fig. 1) suggests that Na v 1.3 (Scn3a) might contribute to the Na + current in both βand α-cells. We examined this possibility by comparing αand β-cell Na + currents in cells from heterozygous Scn3a +/-(that contain one functional Scn3a allele) and Scn3a −/− knockout mice (that contain no functional alleles). Figure 6A shows Na + current I-V relationships for Scn3a +/and Scn3a −/− β-cells. The magnitude of the Na + current in Scn3a +/β-cells was slightly smaller than in wild-type β-cells (compare with Fig. 2A) but this difference was not statistically significant. However, that in Scn3a −/− β-cells was reduced by ß35%. Figure 6B compares the voltage dependence of Na + current inactivation in β-cells from Scn3a +/and Scn3a −/− mice. Biphasic inactivation was observed in 4 of the 7 Scn3a +/β-cells analysed: one component, accounting for 69 ± 20% of the current, inactivated with a V 0.5 of −94 ± 5 mV and a second component, comprising 31 ± 20%, inactivated with a V 0.5 of −49 ± 7 mV. In the remaining 3 cells, inactivation was monophasic with a V 0.5 of −94 ± 3 mV. Mean data from all seven Scn3a +/β-cells are shown in Fig. 6B. These values are not significantly different from those of wild-type β-cells (Fig. 2B). In Scn3a −/− β-cells, but where 20 mM glucose elicited a sustained increase in [Ca 2+ ] i . This trace is representative of 104 of 110 cells in 9 islets from 4 different mice. ii, example of a cell that responded with TTX-sensitive [Ca 2+ ] i oscillations when exposed to 6 mM glucose (6 of 110 cells). Subsequent elevation of glucose to 20 mM resulted in a more sustained increase in [Ca 2+ ] i . iii, recording from a β-cell (identified by the absence of tdRFP fluorescence) in an islet from an Sst-tdRFP mouse that generated [Ca 2+ ] i oscillations at 6 mM glucose and a sustained [Ca 2+ ] i elevation when glucose was increased to 20 mM. iv, recording from a δ-cell (identified by the presence of tdRFP fluorescence); note the large, regular spontaneous [Ca 2+ ] i oscillations at 1 mM glucose which are little affected by increasing the glucose concentration (unlike what is seen in β-cells). B, poor synchronization of [Ca 2+ ] i oscillations in β-cells exposed to 6 mM glucose. The traces were taken from two β-cells in the same islet. The dotted vertical lines indicate the peaks of the [Ca 2+ ] i oscillations in cell 1. C, stimulation of insulin secretion by 6 mM (left) and 20 mM (right) glucose in the absence (•; n = 8) or presence (•; n = 9) of TTX (0.1 μg ml −1 ). Note the different ordinate scales. The grey areas highlight the same data points displayed using the different ordinate scales. For clarity, statistical significances are not shown. D, area under the curve (AUC) measured in response to 6 or 20 mM glucose in the absence and presence of TTX (0.1 μg ml −1 ; as indicated). Note the different ordinate scales for 6 and 20 mM glucose. * P < 0.05 (n = 8-9). E, insulin secretion measured at 1 or 8 mM glucose (Glc) with or without TTX in the absence (left) or presence (right) of 100 nM BIM23056 (n = 6-8 replicates using islets from eight mice). * * * P < 0.001 vs. 1 mM glucose alone; † P < 0.01 vs. 1 mM glucose and TTX; ‡ ‡ P < 0.01 vs. 8 mM glucose; § § P < 0.01 vs. 8 mM glucose and BIM23056. F, reversible suppression of action potential firing by TTX in a tdRFP-expressing δ-cell. Note stimulation of action potentials when glucose was increased from 1 to 20 mM (1G and 20G). TTX was applied at a concentration of 0.1 μg ml −1 as indicated by the bold horizontal line.
inactivation was invariably monophasic and occurred with a V 0.5 of −102 ± 2 mV (n = 18). The amplitude of the Na + current in Scn3a +/α-cells was slightly smaller than that observed in control animals (compare Figs 2A and 6C) but this difference was not statistically significant. Figure 6C shows I-V relationships recorded from Scn3a +/and Scn3a −/− α-cells using a holding potential of −180 mV. Total ablation of Scn3a reduced the maximum current by ß85% compared to that observed in Scn3a +/α-cells.
As in wild-type α-cells (Fig. 2D), Na + current inactivation was biphasic in all Scn3a +/α-cells (n = 11, Fig. 6D). One component, comprising 24 ± 10% of the total current, inactivated at negative voltages (V 0.5 = −91 ± 2 mV) and another component, accounting for 76 ± 10% of the total current, inactivated at more positive voltages (V 0.5 = −38 ± 2 mV). These values are comparable to those observed in wild-type cells (Fig. 2D). In Scn3a −/− α-cells, the Na + current component inactivating at voltages above −50 mV was abolished and the relationship was well described by a single Boltzmann function with a V 0.5 of −94 ± 7 mV (n = 5). However, there was a hint of a small component of inactivation between −90 and −50 mV that may represent Scn8a, expression of which was detected in α-cells (see Fig. 1).

Influence of ablating Scn3a on Na + current kinetics
We examined the consequence of Scn3a ablation on the time constants of activation (τ m ) and inactivation (τ h ) in αand β-cells. Activation and inactivation kinetics in Scn3a +/α-cells were similar to those observed in wild-type α-cells (compare Figs 7A and B and 3C and D). Ablation of Scn3a slowed activation (increased τ m ; Fig. 7A) and accelerated inactivation (reduced τ h ; Fig. 7B). In β-cells, ablation of Scn3a did not measurably affect Na + current kinetics, presumably because Scn3a/Na v 1.3 only accounts for a fraction (<20%) of the β-cell Na + current.

Na + currents in α-Scn9a −/− α-cells
We next measured Na + currents after α-cell-specific knockout of Scn9a. Figure 8C illustrates the I-V relationships for Scn9a +/and Scn9a −/− α-cells. Ablation of Scn9a reduced the Na + current amplitude by <10%; the residual current inactivated with a V 0.5 of ß−35 mV (data not shown), in reasonable agreement with that obtained for the principal Na + current component observed in wild-type α-cells (Fig. 2D). Na + current activation and inactivation kinetics in α-Scn9a +/and α-Scn9a −/− α-cells were identical (data not shown), presumably because Na + current kinetics in α-cells are principally determined by non-Na v 1.7 channels (probably Na v 1.3). We conclude that Na v 1.7 Na + channels contribute only marginally to the α-cell Na + current.

Influence of ablating Scn9a on Na + current kinetics
Na + current activation and inactivation kinetics in β-Scn9a +/and β-Scn9a −/− β-cells are compared in Fig. 9. In β-Scn9a +/β-cells, Na + current kinetics were comparable to those seen in wild-type β-cells ( Fig. 3C  and D). Whereas the effects on inactivation (τ h ) were not statistically significant, activation tended to be slower in β-Scn9a -/β-cells and τ m measured at −10 mV increased from ß0.3 to ß0.7 ms. No effects on Na + current kinetics were seen in α-Scn9a -/α-cells (data not shown), presumably reflecting the small contribution of Na v 1.7 Na + channels to the α-cell Na + current.
In Scn3a +/islets, increasing glucose from 1 to 8 mM inhibited glucagon secretion by >60% (Fig. 10A, left), similar to what is observed in wild-type islets (not shown here but see Macdonald et al. 2007). In Scn3a -/islets (Fig. 10A, right), glucagon secretion at 1 mM glucose was reduced by two-thirds compared to that seen in Scn3a +/islets and although elevating glucose to 8 mM resulted in further suppression, the magnitude of this effect was reduced by 70%. The lower rate of glucagon secretion in Scn3a-deficient islets was not due to a decreased glucagon content, which averaged 939 ± 99 and 804 ± 23 pg per islet in Scn3a +/and Scn3a −/− islets, respectively (not statistically different).

Pancreatic islets do not express an islet-specific Scn9a isoform
The inactivation of Na v 1.7 Na + channels (encoded by Scn9a) we observe in mouse β-cells occurs at voltages 30 mV more negative than reported for the same channels in sensory neurones (Herzog et al. 2003). One possible explanation for this discrepancy is that β-cells and neurones express distinct Na v 1.7 isoforms. However, although two splice variants were detected in both mouse islets and brain tissue, named isoforms A and B (Fig. 11), their deduced amino acid sequences were 100% identical.

Na v 1.7 inactivation is controlled by an intracellular islet cell factor
Na v 1.7 heterologously expressed in HEK 293 cells inactivates with a V 0.5 of −70 mV (Cox et al. 2006), which J Physiol 592.21 is 30-40 mV more positive than we observe in β-cells. The finding that Na v 1.7 is identical in islets and brain therefore suggests that islet cells contain a factor that shifts Na v 1.7 inactivation toward unphysiologically negative membrane potentials. We tested the possible involvement of a diffusible (cytosolic) factor by comparing Na + current inactivation in β-cells at various times after establishment of the whole-cell configuration. We measured V 0.5 as −98 ± 1 and −104 ± 1 mV (n = 3; P < 0.02) at 2 and 15 min after starting the recordings, respectively (Fig. 12A). It is evident that the shift is small and in the opposite direction, and thus cannot explain why Na + channel inactivation occurs at such negative voltages in β-cells.

Na + channel inactivation in human β-cells
Unlike rodent β-cells, Na + currents in human β-cells are reported to inactivate at physiological membrane potentials (Misler et al. 1992;Braun et al. 2008). Recent RNA sequencing data indicate that SCN9A represents a small fraction (25%) of Na + channel transcripts on human β-cells (Nica et al. 2013). We therefore reassessed the voltage dependence of inactivation in human β-cells. Two components of inactivation were observed in 4 out of 9 human β-cells: one component accounting for 23 ± 3% of the current inactivated with a V 0.5 of −105 ± 2 mV and a second component comprising 77 ± 3% with a V 0.5 of −41 ± 1 mV. In the remaining 5 cells, inactivation was monotonic with a V 0.5 of −46 ± 1 mV (Fig. 12B). The contribution of the component inactivating at negative voltage averaged 10 ± 4% of the total Na + current when data from all nine cells were combined. Its smallness probably explains why it escaped detection in the earlier study (Braun et al. 2008).

Discussion
Our data demonstrate that αand β-cells express the same types of voltage-dependent sodium channel α subunits: Na v 1.3 (Scn3a) and Na v 1.7 (Scn9a). Although Scn8a (Na v 1.6) is also expressed in both αand β-cells, electrophysiological data indicate that Na v 1.6 channels do not contribute substantially to the Na + current in either cell type.
Na + channel density in the plasma membrane (measured as the maximum Na + current that can be evoked from a hyperpolarized membrane potential) is approximately equal for both cell types but Na + currents evoked from physiological resting potentials are far larger in α-cells (100 pA pF -1 ) than β-cells (20 pA pF -1 ). The reason for this difference is that the relative expression of Na v 1.3 and Na v 1.7 differs, with Na v 1.3 being expressed at high levels in α-cells and Na v 1.7 being the dominant subtype in β-cells. Furthermore, Na v 1.7 channels in both αand β-cells are locked in a functionally inactive state, so that they contribute little to the physiologically activatable Na + current density. Thus, in β-cells, where Na v 1.7 dominates, little Na + current can be evoked from physiological resting potentials. As a consequence, Na v 1.7 plays a relatively minor functional role in β-cells, despite the fact that it accounts for 80-90% of the transcripts. In α-cells, by contrast, where Na v 1.3 dominates, large Na + currents can be activated at physiological membrane potentials (−60 mV) and blocking Na + channels with TTX abolishes the stimulation of glucagon secretion evoked by low glucose (Zhang et al. 2013).

Mouse αand β-cells express different Na + channel subunits: impact on Na + channel inactivation
Although Scn9a/Na v 1.7 is expressed in both αand β-cells (Vignali et al. 2006), we observed that Scn9a is quantitatively far more important in β-cells than α-cells (>80 vs. 45% of all Na + channel transcripts, respectively). In agreement with this finding, 80-90% of the β-cell Na + current flows through Na v 1.7 channels, but only 10-20% of the α-cell Na + current. Conversely, Na v 1.3 is more important in α-cells, in terms of both transcripts (30-50%) and Na + current density (70-90%). Thus, there is a reasonable agreement between the electrophysiological and molecular biological data.

Role of Na + channels in β-cells
We found that the Na + channel blocker TTX inhibited electrical activity, [Ca 2+ ] i and insulin secretion in the perfused pancreas. These data are in apparent disagreement with our previous finding that insulin secretion is little affected by TTX in isolated mouse islets (Macdonald et al. 2007). We attribute this discrepancy to accumulation of secreted somatostatin during static incubations of isolated islets, which leads to suppression of insulin secretion. By blocking somatostatin secretion, we hypothesise that TTX relieves the paracrine inhibition of insulin secretion and that this effect masks the suppression of insulin secretion produced by inhibition of Na + channels in β-cells. Indeed, when somatostatin signalling in β-cells was pharmacologically prevented, insulin secretion in isolated islets became TTX-sensitive (Fig. 5E).
At first sight, a major role for voltage-gated Na + channels in β-cell electrical activity and insulin secretion seems unlikely given that the Na + current is almost fully (>90%) inactivated at −70 mV, which is close to the most negative membrane potential attained in the absence of glucose. However, 30% of β-cells possess a component of Na + current that inactivates only at more positive potentials. Inactivation of this current is half-maximal at ß−50 mV, which is comparable to the plateau potential from which the action potential originates at glucose concentrations ࣙ20 mM. Although Figure 10. Effects of ablating Scn3a and Scn9a on insulin and glucagon secretion A, glucagon secretion measured during 1 h static incubations at the indicated concentrations of glucose and TTX from Scn3a +/-(left) and Scn3a −/− (right) islets (n = 6 replicates using islet from three Scn3a +/and five Scn3a −/− mice). * P < 0.05, * * P < 0.01 vs. 1 mM glucose (same mouse strain) and † † P < 0.01 for comparison between Scn3a +/and Scn3a −/− islets. B, as in A but comparing α-Scn9a +/-(n = 11 replicates using islet from three mice) and α-Scn9a −/− islets (n = 6 replicates using islets from five mice). * P < 0.05; * * P < 0.01 vs. 1 mM glucose. C, as in A but insulin was measured. * P < 0.05 vs. 1 mM glucose (same strain); † P < 0.05 vs. wild-type. D, as in C but measurements were made in β-Scn9a +/and β-Scn9a −/− islets (n = 5-6 replicates using islets from five Scn9a +/and seven Scn9a −/− mice). * P < 0.05 vs. 1 mM glucose (same strain). J Physiol 592.21 Figure 11. Predicted amino acid sequences of Na v 1.7 α subunits from mouse brain and islets Amino acid sequences of splice variants of Scn9a from mouse brain and pancreatic islets are shown as a Clustal W alignment. Alternative mouse splice forms found in GenBank (NP_061340, 1975 amino acids;and AAI72147, 1984 amino acids) are shown as references. Identical residues are marked with asterisks. Putative transmembrane this current is small compared to that which can be evoked from hyperpolarised membrane potentials, its magnitude is comparable to that of the voltage-gated Ca 2+ current. Thus, it is likely to contribute to action potential firing. Indeed, we found that that action potential height is reduced by TTX. Insulin exocytosis is steeply voltage-dependent (Gopel et al. 2004) and it is possible that a 4-6 mV reduction of action potential height accounts for much of the suppression of glucose-induced insulin secretion produced by TTX at high (20 mM) glucose concentrations.
Electrophysiological characterization of Scn3a and Scn9a knockout mice suggests that Na v 1.3 represents the functionally important Na + channel subtype in β-cells. This conclusion is underlined by the strong reduction of glucose-induced insulin secretion in Scn3a −/− islets. Na + currents at physiological resting potentials are observed in only 30% of β-cells, but it is possible that very small Na v 1.3 Na + currents (too small to be detected) may be present in the remaining 70% of cells and contribute to the initiation and upstroke of the action potentials. Because of the high input resistance of the β-cell in the presence of glucose (Ashcroft & Rorsman, 2004), even such small Na + currents may have a dramatic depolarising effect on the β-cell membrane potential. At high glucose concentrations when K ATP channel activity is low in all β-cells, initiation of Na v 1.3 Na + channel-dependent electrical activity in a subset of cells may, via current spread through the gap junctions, trigger Ca 2+ -dependent electrical activity in neighbouring β-cells in which Na v 1.7 Na + channels are completely inactivated. These considerations suggest that Na v 1.3 channels, despite their low copy number, may play a previously unrecognised modulatory role in glucose-induced electrical activity and insulin secretion in mouse β-cells.
In human β-cells, Na + channels play a prominent role in action potential generation and insulin secretion is TTX sensitive (Misler et al. 1992;Braun et al. 2008). Our data suggest that Na v 1.7 is indeed present in human β-cells and inactivates at hyperpolarised voltages (ß−105 mV). However, consistent with the relatively low expression of SCN9A in β-cells (25% of all Na + channel transcripts; Nica et al. 2013), it comprises only a small component of the total current. The fact that non-Na v 1.7 currents segments (S1-S6) are highlighted with continuous lines on top. Sequences of both Na v 1.7 α subunit isoforms from islets were identical to those expressed in brain and the GenBank sequence AAI72147, except the region highlighted. Isoform A (isoA) contains the sequence encoded by extended exon 12 (extension highlighted), while isoform B (isoB) contains the sequence encoded by conventional exon 12. Note that only the sequences of domain I and the cytoplasmic loop between domains I and II are shown because the rest of the protein is 100% identical in all isoforms. dominate in human β-cells can explain why their Na + currents inactivate at physiological membrane potentials.
Ablation of Na v 1.7 (Scn9a) in mice had little effect on glucose-induced insulin secretion, consistent with the fact that this current is completely inactivated at resting membrane potentials. Thus, the physiological role of Na v 1.7, which accounts for 80-90% of the total Na + current, in β-cells is obscure.

Role of Na + channels in α-cells
The effects of the Na + channel blocker TTX on glucagon secretion and α-cell electrical activity confirm that voltage-gated Na + channels are important for the upstroke of the α-cell action potential. Application of TTX resulted in a ß15 mV reduction of action potential height (from +2 to −12 mV). The inhibitory effect of TTX on glucagon secretion in wild-type islets is quantitatively similar to that  Fig. 2B but using β-cells that only contained a Na + current component inactivating at negative membrane potentials. The inactivation curves were recorded immediately after establishment of the whole-cell configuration (•) and 15 min later (•). Data are mean ± SEM recorded in the same cells at two different times (n = 3). B, voltage dependence of Na + current inactivation in human β-cells that contained one (•, n = 5) or two (•, n = 4) components of inactivation when cells were held at −180 mV. The curves represent single (black) or double (grey) Boltzmann fits to the data (all cells from one organ donor). produced by genetically ablating Scn3a −/− . This argues that Na v 1.3 represents the functionally important Na + channel in α-cells. Because of the steep voltage dependence of exocytosis in α-cells, the reduction in action potential height produced by TTX can be expected to exert a large inhibitory effect (75%) on exocytosis (Zhang et al. 2013), sufficient to account for the suppression of glucagon secretion.
As in β-cells, Na v 1.7 is unlikely to contribute to action potential firing or glucagon secretion because of its inactivation properties. This idea is supported by the finding that ablation of Na v 1.7 has little effect on glucagon secretion.
Human α-cells possess a similar complement of Na + channel subunits, and other ion channels, to human β-cells Nica et al. 2013). The question therefore arises as to why glucose stimulates insulin secretion, but inhibits glucagon secretion. We have argued elsewhere that this is principally because the magnitude of the K ATP conductance is much smaller in α-cells Zhang et al. 2013). As a consequence, α-cells are electrically active at low glucose concentrations and membrane potential-dependent inactivation of Na + channels mediated by glucose will have a strong inhibitory effect on electrical activity and glucagon secretion. In β-cells Na + channel inactivation also occurs, but the stimulatory effect on glucose on electrical activity (and thus secretion) dominates.
In β-cells, inactivation of Na v 1.7 proceeds at such negative voltages that very few (if any) of these channels remain 'activatable' at the normal resting potential (−70 mV). Why this is the case is an enigma. Mouse Na v 1.7 currents in dorsal root ganglion cells inactivate with a V 0.5 of ß−70 mV (Herzog et al. 2003), ß30 mV more positive than the corresponding value in β-cells. This puzzling difference does not reflect expression of a splice variant because the β-cell Na v 1.7 sequence is identical to that found in neurones. This argues that there must be something in β-cells that shifts inactivation towards negative membrane potentials. The identity of this factor remains unknown. However, the fact that Na v 1.7 inactivation does not shift towards more depolarised potentials following extended intracellular dialysis suggests the involvement of a non-diffusible factor.
We speculate that the atypical inactivation of Na v 1.7 in islet cells results from the interaction of Na v 1.7 with a protein (or lipid) present in β-cells but not in neurones. We can exclude a role of Na + channel β subunits J Physiol 592.21 because inactivation of the Na v 1.3-independent current component (representing Na v 1.7) was identical (V 0.5 at −100 mV) in αand β-cells despite these two cell types expressing β 3 and β 1 subunits, respectively. It is notable that Na v 1.3 and Na v 1.6 (as suggested by the Na + current measurements in human β-cells; Fig. 12) channels undergo voltage-dependent inactivation at physiological membrane potentials. Thus, the impact of this modulation appears specific for Na v 1.7 channels. The nature of this modulator remains to be determined. Na v 1.7 channels play a key role in pain sensation (Dib-Hajj et al. 2008), and thus identification of the putative modulator might provide a new target for therapeutic drugs directed at alleviating pain. J Physiol 592.21

Funding
Financial support was received from the Wellcome Trust (grant nos. 884655, 089795, 095531) and the European Union (322620). L.G. and N.J.R. hold Wellcome Trust OXION PhD studentships