Critical Role of Gap Junction Coupled KATP Channel Activity for Regulated Insulin Secretion

Pancreatic β-cells secrete insulin in response to closure of ATP-sensitive K+ (KATP) channels, which causes membrane depolarization and a concomitant rise in intracellular Ca2+ (Cai). In intact islets, β-cells are coupled by gap junctions, which are proposed to synchronize electrical activity and Cai oscillations after exposure to stimulatory glucose (>7 mM). To determine the significance of this coupling in regulating insulin secretion, we examined islets and β-cells from transgenic mice that express zero functional KATP channels in approximately 70% of their β-cells, but normal KATP channel density in the remainder. We found that KATP channel activity from approximately 30% of the β-cells is sufficient to maintain strong glucose dependence of metabolism, Cai, membrane potential, and insulin secretion from intact islets, but that glucose dependence is lost in isolated transgenic cells. Further, inhibition of gap junctions caused loss of glucose sensitivity specifically in transgenic islets. These data demonstrate a critical role of gap junctional coupling of KATP channel activity in control of membrane potential across the islet. Control via coupling lessens the effects of cell–cell variation and provides resistance to defects in excitability that would otherwise lead to a profound diabetic state, such as occurs in persistent neonatal diabetes mellitus.

Introduction b-cells within the intact islet of Langerhans exhibit synchronous glucose-dependent bursts of electrical activity. Dissociated b-cells also show increased electrical activity at elevated glucose concentrations, but this activity is variable from cell to cell, with some cells being electrically ''silent'' and others continuously ''bursting'' at any given glucose concentration [1][2][3][4]. Synchronization of electrical activity via gap junctions has long been argued as essential for normal glucose-dependent insulin secretion in the intact islet [5][6][7][8].
In particular, islets devoid of gap junction-forming Connexin 36 (Cx36) exhibit limited synchronization of glucose-stimulated intracellular Ca 2þ (Ca i ) [8]. To test the extent and role of this coupling in the islet electrical response, it would be desirable to examine within the intact islet the behavior of cells that would be silent or bursting as isolated cells. To date, it has not been possible to know which cell is which in the setting of the intact islet. However, we have generated transgenic mice containing two electrically distinct types of b-cells that uniquely permit us to track the behavior of specific individual cells within the intact islet [9]. As previously shown, these mice express a b-cell-specific, dominant-negative Kir6. 2[AAA]-GFP transgene, in which the pore-forming subunit of the ATP-sensitive K þ (K ATP ) channel is rendered nonfunctional. Mosaic expression of the Kir6. 2[AAA]-GFP transgene results in a high-level expression in approximately 70% of the b-cells, and these expressing cells are distributed randomly throughout the islet [9,10]. Thus, K ATP channels are functionally ''knocked out'' of 70% of cells, yet the intact islets still show glucose-dependent insulin secretion, with an even steeper concentration dependence of insulin secretion than wild-type (WT) islets [9].

Results/Discussion
To examine the behavior of individual cells within the intact islet, we developed a microfluidic device to hold pancreatic islets stationary while under continuous fluid flow (see Figure S1) [11]. We compared the glucose-dependent Ca i responses of WT and Kir6.2[AAA] transgenic islets ( Figure 1). Near the periphery of the islet, fluorescence from a Ca 2þ sensor (Fura Red) was relatively uniform in all cells ( Figure  1A), and spectrally separable from green fluorescent protein (GFP) fluorescence ( Figure 1B). Both the GFP-positive andnegative populations of b-cells (K ATP -absent and K ATPpresent cells, respectively) within a Kir6.2[AAA] transgenic islet showed identical Ca i -responses ( Figure 1C). WT islets were ''Ca i inactive'' across the tissue below 8 mM glucose (i.e., Ca i remained low and no oscillations were observed; Figure  1D), consistent with previous studies [9,11]. Kir6.2[AAA] islets were also completely Ca i inactive across the tissue at low glucose concentrations (2 and 4 mM), but they exhibited Ca i activity at slightly lower glucose levels than WT (6 mM). This leftward shift in glucose dependence of Ca i in Kir6.2 [AAA] islets is consistent with the leftward shift in whole-islet glucose-dependent insulin secretion ( Figure 1E) [9]. Kir6.2 À/À and SUR1 À/À mice, which completely lack K ATP channels, show essentially no glucose dependence of electrical activity, Ca i , or insulin secretion [12][13][14]. Thus, the near-normal glucose response of Ca i ( Figure 1D) and insulin secretion ( Figure 1E) by Kir6.2[AAA] islets are initially somewhat surprising, since K ATP is functionally knocked out in approximately 70% of the b-cells. Paracrine signaling could influence the electrical synchronization between islet cells, for instance by d-cell secretion of somatostatin [15]. However, these data suggest that K ATP channel activity in the residual 30% of cells is coupled through gap junction conductance that is strong enough to hyperpolarize the knockout cells. This suggestion, in turn, predicts that glucose dependence will be lost in dispersed Kir6.2[AAA]-GFP cells. Accordingly, we measured the glucose-stimulated Ca i response and insulin secretion from dispersed cell preparations ( Figure 2). Image fields of isolated Kir6.2[AAA] b-cells show uniform loading of the Ca 2þ indicator dye (Fura-2; Figure 2A), and no overlap (right) transgenic islets. Islets were exposed to the indicated glucose concentrations for greater than 10 min by changing the reagent well solution. Images were then collected at 1.5-s intervals over a period of 125 s. In contrast to the traces found in (C), each trace in each panel (solid, dot, and dashed) represents a whole single islet. These islet traces are done to show the dose -response of individual islets rather than synchrony across the tissue. WT islets showed no oscillatory behavior until treated to 8 mM glucose or greater (responses of three different islets are shown). Kir6.2[AAA] islets were also quiescent at low glucose concentrations; however, these islets showed Ca i oscillations at 6 mM glucose or above (n ¼ 5 islets). (E) Glucose-stimulated insulin secretion from WT and Kir6.2[AAA] islets (mean 6 SEM, n ¼ 17 and 15, respectively). Islets were also exposed to 16.7 mM glucoseþ glibenclamide (Glib) to achieve maximum possible insulin secretion. The difference in means 6 95% confidence interval is shown in Figure S2A. DOI: 10.1371/journal.pbio.0040026.g001 with GFP fluorescence ( Figure 2B). Both the Ca i response and insulin secretion were essentially glucose independent in dispersed Kir6.2[AAA] cells ( Figure 2C and 2D). This is in striking contrast to the Ca i response and insulin secretion from dispersed WT cells which are strongly glucose dependent, but is consistent with previous results from SUR knockout mice [16].
It was previously shown that dispersed GFP-positive Kir6.2[AAA] b-cells are continuously depolarized in nonstimulatory glucose concentrations [9]. To examine whether cell-cell coupling reconstitutes glucose-dependent membrane potentials in intact islets, we imaged intact islets using a membrane potential-sensitive dye ( Figure 3A). At 2 mM glucose, a few cells on the islet periphery were noticeably more depolarized than others. These bright cells were never incubations. This incubation time shows K ATP channel-dependent insulin secretion, but does not encompass K ATP channel-independent insulin secretion shown to occur in knockout islets beyond 3 h of incubation [13]. The difference in means 6 95% confidence interval for (C) and (D) are shown in Figure S2B and S2C, respectively). DOI: 10.1371/journal.pbio.0040026.g002 seen to co-localize with GFP fluorescence (not shown), and were likely a-cells. The rest of the islet cells showed relatively dim uniform fluorescence, consistent with uniformly polarized membranes, and all of these cells brightened (depolarized) upon glucose stimulation. GFP-expressing and non-expressing cells had indistinguishable membrane polarization ( Figure 3A). Thus, b-cells across the Kir6.2[AAA] islet have similar resting membrane polarization, and all depolarize with increased glucose concentrations.
In contrast to electrical activity, we have not observed metabolic coupling of b-cells [11,17], although there may be feedback between electrical activity and metabolic response. The Kir6.2[AAA] islets provide a unique opportunity to examine this potential feedback. We used intrinsic NAD(P)H fluorescence as a measure of cellular metabolic status [11,18]. In comparison to WT, Kir6.2[AAA] islets showed no difference in baseline NAD(P)H intensity, but showed a slight (;1 mM) leftward shift in the (glucose) response ( Figure 3B). This shift is comparable to that of glucose-stimulated Ca i response and insulin secretion (see Figure 1D and 1E), suggesting a feedback modulation of metabolism during electrical activity. Such feedback could result from Ca i influx [19] or insulin signaling [20,21] since both likely modulate bcell metabolism and occur when the tissue is electrically active. We further compared the baseline and stimulated responses of dispersed Kir6.2[AAA] cells ( Figure 3C). Baseline NAD(P)H intensity was significantly elevated in GFPpositive cells (1.3-fold vs. GFP negative at 2 mM glucose), but was similar upon glucose stimulation. The elevated basal metabolism in GFP-positive cells with no difference in glucose-stimulated metabolism is again consistent with a modulating effect of Ca i or insulin. Thus, even though metabolism is not directly coupled between cells in the islet [11,17], entrained electrical activity and shifted Ca i responses can influence metabolic response.
The above results suggest a model whereby the Kir6.2[AAA] islet response is due to coupling of K ATP channel activity, presumably through gap junctions. This model predicts that inhibition of gap junctions would recapitulate the uncoupled Ca i activity found in dispersed Kir6.2[AAA] cells. Thus, we examined glucose-stimulated Ca i responses in the presence of a gap junction inhibitor, 18a-glycyrrhetinic acid (aGA) [22]. Ca i measurements of WT islets ( Figure 4A) showed no b-cell oscillations at sub-threshold glucose levels in the presence or absence of either 10 or 50 lM aGA. In contrast, spontaneous Ca i oscillations were observed in the presence of 50 lM aGA in four of five Kir6.2[AAA] islets at 2 mM glucose, and in six of six Kir6.2[AAA] islets at 4 mM glucose ( Figure 4B, Videos S1-S3). Unlike the synchronous oscillations across the entire islet that are seen in the absence of aGA ( Figure 4A, bottom trace, and Figure 1C), these oscillations occurred asynchronously in small groupings of cells ( Figure 4A, middle trace). The resulting asynchronous Ca i activity in portions of Kir6.2[AAA] islets is consistent with partial decoupling of the K ATP channel activity by the inhibition of gap junction conductance.
The data presented here indicate that glucose control of membrane potential, Ca i , and insulin secretion all depend critically on the gap junctional coupling of K ATP channel activity across neighboring cells within the islet ( Figure 5). It has previously been shown that knockout of Cx36 in islets results in reduction of glucose-stimulated electrical synchrony [8]. Our results provide additional experimental evidence for gap junctional coupling between islet b-cells. Furthermore, our results demonstrate the mechanism by which the coupling of K ATP channel activity via gap junctions results in near-normal glucose-dependent electrical and Ca i responses in the intact islet, even when a majority of b-cells are essentially glucose insensitive. More specifically, this coupling allows b-cells in low glucose with normal K ATP channel activity to clamp the membrane potential of neighboring cells. The electrical properties of isolated b-cells are very variable [2][3][4]. This variability along with our results suggests that the rise in basal insulin secretion from Cx36 knockout islets is due to an inability of quiescent cells to clamp the membrane potential of other b-cells with low stimulation threshold. This may also explain an evolutionary drive to organize these secretory cells into a unique electrical syncytium. Such a mechanism provides a protective effect against hypoglycemia if individual cell properties change under pathological or physiological stimuli. Factors that reduce K ATP channel density, such as channel subunit trafficking mutants [23,24], could do so to the point where individual cells become glucose insensitive. However, the remaining glucose-sensitive cells in the islet with sufficient residual K ATP channel activity will maintain a glucose response of the whole islet. Such cell coupling would help resist the development of persistent hypoglycemic hyperinsulinemia (PHHI), unless compounded by additional defects, as may be the case in human PHHI [25].
Our results also have implications for the understanding of diabetic phenotypes resulting from K ATP channel gain of function or from other causes of hypoexcitability. It is now clear that overactive K ATP channels cause neonatal diabetes due to inexcitability of b-cells and failure to secrete insulin [26,27]. Our data demonstrate that cell-cell coupling can ensure maintained hyperpolarization in all cells if K ATP channel activity in just a few reaches a threshold of inexcitability. However, this also suggests that the domination of K ATP channel strength on neighboring cell membrane potentials, may permit elevated K ATP channel activity in only a few b-cells to clamp the remainder at a hyperpolarized potential, thereby stopping secretion. Such a mechanism may underlie the profound diabetic state in persistent neonatal diabetes mellitus [27]. Equivalent gain-of-function model systems [26] and the experimental approach that we describe here should allow a critical test of the syncytial involvement in these types of neonatal diabetes.

Materials and Methods
Islet isolation and secretion measurements. Islets were isolated as previously described [28,29] and maintained in complete RPMI medium 1640 containing 10% fetal bovine serum and 11 mM glucose at 37 8C under 5% humidified CO 2 for 24-72 h. Islets were dispersed to single cells as described previously [9]. Insulin release was measured from pancreatic islets (ten per well) or dispersed cells (equivalent to 15 islets per well) in static 1-h incubations as described previously [9].
Confocal imaging of microfluidic device trapped islets. Devices were fabricated using the elastomer polydimethylsiloxane (PDMS) as described elsewhere [30] (see Figure S1 for details of fabrication and use). Islets were labeled with 4 lM of either Fluo-4, acetoxymethyl (AM) or Fura Red, AM at room temperature for 1 to 2 h in imaging buffer (125mM NaCl, 5.7 mM KCl, 2.5 mM CaCl 2 , 1.2 mM MgCl 2 , 10 mM HEPES, 2 mM glucose, and 0.1% bovine serum albumin [pH 7.4]). One-and two-photon microscopy was performed on a LSM 510 microscope with a 20 3 0.75 NA Fluar objective lens (Carl Zeiss, Thornwood, NewYork, United States). The device was held on the microscope in a humidified temperature controlled stage (Carl Zeiss) for imaging at 37 8C. Fluo-4 (Molecular Probes, Eugene, Oregon, United States) was imaged using the 488-nm laser line and the long- pass 505-nm emission filter. NAD(P)H intensity was imaged with a 710-nm mode-locked Ti:Saph laser (;3.5 mW at the sample) and fluorescence collection through a non-descanned detector with a custom 380-to 550-nm filter (Chroma, Rockingham, Vermont, United States) [18]. We collected four focal planes separated by 3-lm intervals. Fura Red (Molecular Probes) was imaged using the 488-nm laser line and a band-pass filter (620-680 nm). The GFP label of Kir6.2[AAA] islets was imaged using the 488-nm laser and band-pass (540/20) filter, with no noticeable bleed-through of the Fura Red signal. Membrane potential was measured with bis-(1,3-diethylthiobarbituric acid)trimethine oxonol (DiSBAC 2 (3); Molecular Probes). This dye accumulates in the cells in a Nernst-dependent manner [31]. We flowed the indicated solutions with 500 nM of DiSBAC 2 (3). These solutions were allowed to equilibrate for approximately 30 min to allow dye penetration before imaging. DiSBAC 2 (3) images were collected by 543-nm excitation and a long-pass 560-nm filter. GFP fluorescence in these same islets was imaged separately using 488-nm excitation and a 500-530-nm band-pass filter.
Single cell Ca i measurements. Prior to imaging, cells were labeled with 2 lM Fura-2, AM (Molecular Probes) for 30 min at room temperature and washed with imaging buffer containing 2 mM glucose. Imaging was done on a TE300 Eclipse microscope (Nikon, Melville, New York, United States) using the 40 3 1.3 NA DIC PlanFluor lens, side-mounted CoolSnap HQ camera (Roper Scientific, Tucson, Arizona, United States), and Fura-2 filter set (Chroma). The cells were placed on a stage enclosed in Plexiglas and kept in humidified air at 37 8C. The system was controlled with Metamorph 6.0 (Universal Imaging, Downington, Pennsylvania, United States). Image pairs were collected at 2-s intervals for 2 min after incubation for at least 10 min in 2, 6, and 10 mM glucose.
Image analysis. Images were analyzed with MetaMorph 5.0 (Universal Imaging). For Fluo-4 analysis, islets/regions of islets were outlined and their intensities versus time were plotted. From these plots, the relative intensity change (F/F 0 ) and frequencies were calculated. For Fura-2 analysis, individual cells were outlined for calculation of the background corrected 340:380-nm intensity ratio. Cells were categorized as ''calcium active'' if they had a ratio that was either more than two standard deviations larger than that observed for WT cells in 2 mM glucose or demonstrated a change of more than one standard deviation change during the 2-min observation period. Figure S1. A Microfluidic Device, Designed to Hold a Pancreatic Islet Stationary in a Fluid Stream, was Fabricated Using the Elastomer Polydimethylsiloxane (PDMS) The fabrication of the device is described in [30]. (A) A dye-loaded microfluidic device. The image is labeled to show the Reagent well, Islet In/Out port, Wall trap area, and Waste port. This device relies on gravity flow from either the Reagent well or Islet In/Out port to the Waste port. Islets were brought into the device through the In/Out port. The islet travels through the main channel of the device until it comes in contact with Wall area. (B) A differential interference contrast image of a pancreatic islet at the Wall trap area of the microfluidic device. This islet is in the main channel (height ;100 lm and width ;600 lm) touching the wall trap (bottom of image), with fluid flowing by gravity from the top to bottom of the image. Islets trapped in a microfluidic device channel with approximately 100 lm height are pressed against the coverslip surface for optimal microscopic imaging. The islet is held stationary by the coverslip, ceiling and wall trap. All the islets studied maintain their shape throughout the experiments, which suggests limited sheer pressure. We have previously shown that WT islet Ca i and NAD(P)H responses are unperturbed in similar devices [11]. (C) Multiple image planes of sulforhodamine B (0.2 lM) dye as it is flown past a device-trapped islet. The z-distance from the coverslip is shown at the bottom-left corner of each image. Note that the dye solution is not observed in the wall trap area above 16 lm of depth. In this region of the device, the height of the channel drops from approximately 100 to 15 lm. This channel height allows fluid to flow while blocking islet movement. Once islets are trapped in the device, we plugged the In/Out port and started gravity flow from the Reagent well. (D) Changing the Reagent well solution (100 ll) to contain sulforhodamine B (0.2 lM) resulted in complete change of solution at .1 min. The half-maximum of this change was 27 s, which corresponds to a flow rate of approximately 1.2 ll min À1 , which is below the calculated maximum velocity (;3 ll min À1 ), but slightly faster than the calculated flow rate (;0.7 ll min À1 ). At this flow rate, we observed no noticeable warping of the islet, which is consistent with low sheer. Furthermore, the islets remained stationary after long periods of flow time (minutes to hours) and during reagent solution changes. Keeping the islet in such a stationary position facilitates time-lapse imaging and the observation of similar regions after many different treatments. DOI: 10.1371/journal.pbio.0040026.sg001 (4 MB TIF). Video S1. This Video Shows a Kir6.2[AAA] Islet Treated to 2 mM Glucose (2 mM Glc) The elapsed time covers a period of 190 s at 1.57 frames/s. Some individual cells are observed to asynchronously increase in intensity. These responses are similar to those observed in WT islets at nonstimulatory glucose concentrations [11], and are consistent with an a-cell response. DOI: 10.1371/journal.pbio.0040026.sv001 (407 KB MOV).

Supporting Information
Video S2. This Video Shows the Same Kir6.2[AAA] Islet Shown in Video S1, but Treated to 2 mM Glucose with 10 lM aGA (2 mM Glc þ 10 uM GA) The elapsed time covers a period of 190s at 1.57 frames/s. Similar to the response found with 2 mM glucose (Video S1), some individual cells are observed to asynchronously increase in intensity. These responses are similar to those observed in WT islets at nonstimulatory glucose concentrations [11], and are consistent with an a-cell response. Video S3. This Video Shows the Same Kir6.2[AAA] Islet Shown in Video S1, but Treated to 2 mM Glucose with 50 lM aGA (2 mM Glc þ 50 uM GA) The elapsed time covers a period of 383 s at 1.57 frames/s. White arrows are shown to highlight synchronous Ca i responses among groups of cells within the islet, consistent with a b-cell response. However, these responses occur in two distinct regions of the islet at different time points indicating asynchrony between cell groupings. DOI: 10.1371/journal.pbio.0040026.sv003 (899 KB MOV).