Decreasing Cx36 Gap Junction Coupling Compensates for Overactive K_ATP Channels to Restore Insulin Secretion and Prevent Hyperglycemia in a Mouse Model of Neonatal Diabetes

Glucose-stimulated insulin secretion from β-cells in the islet is regulated via a series of metabolic and electrical events. The ATP-sensitive K⁺ channel (K_ATP channel) provides a central role in coupling increases in the ATP/ADP ratio after the metabolism of glucose, to membrane depolarization, elevated intracellular free calcium activity ([Ca²⁺]_i), and insulin granule exocytosis (1). The K_ATP channel is made up of inward-rectifying K⁺ channel Kir6.2 and sulfonylurea receptor 1 (Sur1). In humans, mutations in the genes encoding Kir6.2 (KCNJ11) and Sur1 (ABCC8), which reduce the sensitivity of ATP inhibition and lead to overactive K_ATP channels (gain-of-function mutations), are the most common cause of neonatal diabetes mellitus (NDM) (2). Transgenic mice that express ATP-insensitive K_ATP channels in β-cells can recapitulate the human disease (3–5). These mice exhibit marked hyperglycemia and reduced plasma insulin, and islets from these mice show a suppression of glucose-stimulated [Ca²⁺]_i and insulin release (4,6). This demonstrates that a reduction of glucose-dependent excitability underlies the development of diabetes caused by K_ATP channel mutations. Inhibition of K_ATP channels with sulfonylureas can recover elevations in [Ca²⁺]_i and insulin secretion in islets of these mouse models (3,4,6). As such, sulfonylureas can be applied to treat patients with NDM associated with K_ATP channel mutations (7,8). Nevertheless, some mutations render the K_ATP channel insensitive to sulfonylureas (7,9,10).

Cellular interactions within the islet have long been known to be important for enhancing the regulation of insulin release (11) and are also important for regulating electrical activity (12). Connexin-36 (Cx36) gap junction channels regulate islet electrical activity by coupling K_ATP-regulated membrane depolarization between β-cells of the islet. This synchronizes oscillations in membrane depolarization and [Ca²⁺]_i at elevated glucose (13,14), leading to a coordination of first-phase insulin release and pulsatile second-phase insulin release (15). Under normal conditions, Cx36 gap junction channels also enhance a suppression of spontaneous elevations in [Ca²⁺]_i at basal glucose (16,17); upon a knockout of Cx36, elevated [Ca²⁺]_i is observed at lower glucose levels (14,18). This suppression likely occurs as a result of the long-established presence of β-cell heterogeneity (19), where inexcitable cells in the islet suppress membrane depolarization and [Ca²⁺]_i in more excitable cells via gap junction coupling (16,18).

Given that Cx36 gap junction channels coordinate K_ATP-regulated membrane potential, we hypothesized that their action in suppressing [Ca²⁺]_i and insulin occurs more generally under conditions of K_ATP channel opening. Therefore, similar to their normal action at lower glucose levels, we hypothesized that Cx36 gap junctions would inappropriately enhance the suppression of [Ca²⁺]_i and insulin secretion at elevated glucose in the presence of overactive ATP-insensitive K_ATP channels in NDM. Therefore, we hypothesized that an absence of Cx36 gap junction coupling upon expression of ATP-insensitive K_ATP channels would reduce this suppression and lead to spontaneous elevations in [Ca²⁺]_i at elevated glucose. We anticipated that this [Ca²⁺]_i elevation would stimulate sufficient insulin release to prevent the severe hyperglycemia that emerges in these animals. However, we anticipated that suppression of [Ca²⁺]_i and insulin release would be maintained at low glucose levels; in essence, a decrease in Cx36 will left shift the dose response to compensate for a right shift due to overactive K_ATP channels. Here, we tested this by introducing a knockout of Cx36 gap junction channels into a mouse model of NDM that expresses ATP-insensitive K_ATP channels in the β-cell under control of an inducible Cre^ER-recombinase (3). We tested whether mice with reduced or absent Cx36 gap junction coupling showed an improvement in glucose homeostasis and whether islets from these mice showed a recovery of glucose-stimulated [Ca²⁺]_i and insulin secretion.

Experiments were performed in compliance with the relevant laws and institutional guidelines and were approved by the University of Colorado Institutional Animal Care and Use Committee. Mice expressing Rosa26-Kir6.2^{[ΔN30,K185Q]} (Kir6.2^{[ΔN30,K185Q]} gain-of-function mutant), Pdx-Cre^ER (pancreas-specific inducible Cre), and Cx36^−/− (global Cx36 knockout) were generated as described previously (3,20,21) and supplied by collaborating laboratories. β-Cell conditional expression of Kir6.2^{[K185Q,ΔN30]} is achieved through crossing Rosa26-Kir6.2^{[ΔN30,K185Q]} and Pdx-Cre^ER mice to excise a loxP-flanked stop codon to drive Kir6.2^{[K185Q,ΔN30]} expression. Green fluorescent protein (GFP) is coexpressed via an internal ribosome entry site. Rosa26-Kir6.2^{[ΔN30,K185Q]} and Pdx-Cre^ER mice either had normal gap junction coupling (to generate Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} mice) or were first separately crossed with Cx36^−/− mice to achieve a homozygous deletion and then bred together (to generate Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice). Cx36^+/+ or Cx36^−/− littermate mice lacking Rosa26-Kir6.2^{[ΔN30,K185Q]} and/or Pdx-Cre^ER were used as controls. Mice were studied at generation F3–F5. To prevent genetic drift, new breeders were generated by crossing Cx36^+/+- and Cx36^−/−-expressing breeders every two to three generations.

Each experimental group (age-matched Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} and Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} with respective littermate Cx36^+/+ and Cx36^−/− controls) received five daily doses of tamoxifen (50 mg/g body weight) at experimental day 1–5. Blood glucose was measured daily as previously performed (3) with a glucometer (Ascensia Contour; Bayer). Plasma insulin was measured at day 29 from blood samples centrifuged at 14 krev/min for 10 min and assayed using mouse ultrasensitive insulin ELISA (Alpco). Glucose tolerance tests were performed at day 30–32. Littermate or age-matched mice were fasted overnight for 16 h and received intraperitoneal injection of 2 g/kg body weight of glucose, and blood glucose was measured on tail vein blood samples preinjection (0 min) and 15, 30, 60, 90, and 120 min post–glucose delivery. Insulin tolerance tests (ITTs) were performed at day 30–32. Littermate mice were fasted for 6 h and received intraperitoneal injection of 0.75 units/kg body weight of human recombinant insulin (Novolin; Novo Nordisk), and blood glucose was measured on tail vein blood samples preinjection (0 min) and 15, 30, 45, 60, and 90 min postinjection.

At day 30–36, islets were isolated from pancreata of each experimental mouse and maintained in islet medium (RPMI medium, 10% fetal bovine serum, 11 mmol/L glucose, 100 units/mL penicillin, 100 μg/mL streptomycin) at 37°C under humidified 5% CO₂ for 24 h. For static insulin secretion measurements, islets (five per column, in duplicate) were preincubated for 60 min at 37°C in Krebs-Ringer buffer (128.8 mmol/L NaCl, 5 mmol/L NaHCO₃, 5.8 mmol/L KCl, 1.2 mmol/L KH₂PO₄, 2.5 CaCl₂, 1.2 mmol/L MgSO₄, 10 mmol/L HEPES, 0.1% BSA, pH 7.4) plus 2 mmol/L glucose and then incubated for 60 min at 37°C in Krebs-Ringer buffer plus different glucose concentrations and/or reagents as indicated. After the incubation period, the medium was removed and insulin concentration assayed using mouse ultrasensitive insulin ELISA. To estimate insulin content, islets were lysed in 1% Triton X-100 and frozen at −20°C overnight.

All isolated islets were imaged using established methods (13), in polydimethylsiloxane microfluidic devices (16), maintained at 37°C, with imaging medium (125 mmol/L NaCl, 5.7 mmol/L KCl, 2.5 mmol/L CaCl₂, 1.2 mmol/L MgCl₂, 10 mmol/L HEPES, 2 mmol/L glucose, 0.1% BSA, pH 7.4).

To measure [Ca²⁺]_i response and dynamics, islets were loaded with 4 μmol/L Fura Red-AM (Invitrogen) for 2 h at room temperature and imaged on a Marianas spinning-disk microscope (3I) with a 40× 1.3 NA Plan-Neofluar oil-immersion objective (Zeiss). Images were acquired 10 min after glucose stimulation. Fura Red was imaged using a 488-nm diode laser for excitation and a 580- to 655-nm band-pass emission filter (Semrock). GFP was imaged using a 488-nm diode laser for excitation and a 500- to 550-nm band-pass emission filter. There was no detectable cross-talk between Fura Red and GFP.

To measure [Ca²⁺]_i concentration, islets were loaded with 2 μmol/L Fura 2-AM for 30 min at room temperature and imaged on an Eclipse-Ti widefield microscope (Nikon) with a 20× 0.75 NA Plan Apo objective. Images were acquired 10 min after glucose stimulation or immediately after KCl stimulation. Fura 2 was imaged sequentially using 340- and 380-nm (±5 nm) Arc-lamp excitation and a 470- to 550-nm band-pass emission filter (Chroma).

To measure NAD(P)H, islets were imaged on an LSM510 microscope (Zeiss), with a 40× 1.2 NA C-Apochromatic water-immersion objective. NAD(P)H autofluorescence was imaged with two-photon excitation using a 710-nm mode-locked Ti:sapphire laser (Coherent) and a 400- to 500-nm band-pass emission filter (Chroma) and nondescanned detector. GFP fluorescence was imaged using a 488-nm Ar⁺ laser line for excitation and a 500- to 550-nm band-pass emission filter. Z-stacks of six images were acquired at 2-μm spacing. No GFP fluorescence was detected in the NAD(P)H channel.

To measure mitochondrial membrane potential changes, islets were loaded with 50 nmol/L Rhodamine 123 for 20 min at 37°C and imaged on an LSM510 microscope with a 40× 1.2 NA C-Apochromatic water-immersion objective. Images were acquired 10 min after glucose stimulation. Rhodamine 123 was imaged using a 488-nm Ar⁺ laser line for excitation and a 500- to 550-nm band-pass emission filter. Z-stacks of six images were acquired at 2-μm spacing.

Images collected on the different microscope systems (3I and Zeiss; Nikon) were analyzed offline in Matlab (Mathworks) using established methods (6,18,22). For Fura 2, 340- and 380-nm intensities were averaged across each islet, and the “background” intensity averaged over an unstained islet subtracted. Time-averaged [Ca²⁺]_i concentration was calibrated from the background-subtracted time-averaged 340-nm/380-nm intensity ratio using the Fura 2 calibration kit (Invitrogen).

NAD(P)H and Rhodamine 123 fluorescence were averaged across each islet, with Z positions corrected for focal drift. The mean NAD(P)H autofluorescence for each experimental group was normalized to the average autofluorescence in Cx36^+/+ islets at 2 mmol/L. The mean Rhodamine 123 fluorescence of each islet was normalized to the fluorescence at 2 mmol/L glucose.

To estimate the proportion of the islet showing elevations in [Ca²⁺]_i, Fura Red images were first smoothed using a 3 × 3 average filter. The variance (intensity fluctuation) was calculated over the time course of each pixel. A threshold variance was calculated for a silent cell, identified as quiescent at 2 mmol/L glucose, which represents image noise. An active area showing [Ca²⁺]_i fluctuations was defined as having a variance significantly greater (95% confidence) than the variance of the silent cell.

To determine high-GFP (GFP⁺) cells, GFP images were first smoothed using a 3 × 3 average filter. A threshold intensity was calculated from the autofluorescence in the GFP channel from control (GFP⁻) Cx36^+/+ or Cx36^−/− islets. A GFP⁺ region was identified as having GFP fluorescence greater than the threshold intensity.

Inducible β-cell expression of mutant ATP-insensitive K_ATP channel subunit (Kir6.2^{[ΔN30,K185Q]}) forms K_ATP channels with mixed mutant and wild-type Kir6.2 (3), which reduces the sensitivity of ATP inhibition. To test whether reduced Cx36 gap junction coupling could recover sufficient elevation in [Ca²⁺]_i and insulin secretion in the presence of ATP-insensitive K_ATP channels, age-matched Rosa-Kir6.2^{[ΔN30,K185Q]};Pdx-Cre^ER–expressing mice with normal Cx36 gap junction coupling (Cx36^+/+;Kir6.2^{[ΔN30,K185Q]}) and Rosa-Kir6.2^{[ΔN30,K185Q]};Pdx-Cre^ER–expressing mice with a lack of Cx36 gap junction coupling (Cx36^−/−;Kir6.2^{[ΔN30,K185Q]}) were selected along with littermate control mice (Cx36^+/+ and Cx36^−/−, respectively). Each group of these four mice underwent tamoxifen injections to induce Kir6.2^{[ΔN30,K185Q]} expression and were studied in parallel.

In Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} mice, after the induction of Kir6.2^{[ΔN30, K185Q]} expression, ad libitum–fed blood glucose levels rapidly rose to >500 mg/dL compared with littermate Cx36^+/+ control animals (Fig. 1A). Plasma insulin levels were substantially and significantly diminished in Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} mice compared with littermate Cx36^+/+ controls (Fig. 1B). In Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice, which lack β-cell gap junction coupling, after the induction of Kir6.2^{[ΔN30,K185Q]} expression, ad libitum–fed blood glucose levels remained unchanged compared with Cx36^−/− littermate control animals (Fig. 1C). Insulin levels were not significantly reduced in Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice compared with littermate Cx36^−/− controls (Fig. 1D). At day 30 postinduction, Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} mice showed substantially greater blood glucose compared with Cx36^+/+ controls, whereas Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} showed no significant increase compared with littermate Cx36^−/− controls, and similar to Cx36^+/+ controls (Fig. 1E). Over the 30 days postinduction, Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} mice showed significantly less weight gain compared with Cx36^+/+ littermate controls, whereas Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} showed no significant difference compared with Cx36^−/− littermate controls (Fig. 1F).

In Cx36^+/−;Kir6.2^{[ΔN30,K185Q]} mice with ∼50% gap junction coupling, after the induction of Kir6.2^{[ΔN30,K185Q]} expression, ad libitum–fed blood glucose levels also rose compared with littermate control animals (Supplementary Fig. 1). This initial elevation was slightly reduced compared with that in Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} mice studied in parallel. However, mice ultimately progressed to similar high blood glucose levels >500 mg/dL. Plasma insulin levels were also significantly diminished in Cx36^+/−;Kir6.2^{[ΔN30,K185Q]} mice compared with control mice.

Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} mice also showed substantially greater fasting blood glucose compared with Cx36^+/+ controls and poor glucose tolerance during an intraperitoneal glucose tolerance test (IPGTT), as expected given the marked elevation in fed glucose levels (Fig. 2A). Whereas Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice showed similar fasting blood glucose compared with Cx36^−/− controls, they showed significantly elevated glucose levels during an IPGTT, indicating a reduced glucose tolerance (Fig. 2B). During an ITT, Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice showed only a small elevation in blood glucose levels compared with Cx36^−/− controls, indicating a small reduction in insulin sensitivity (Fig. 2C). Therefore, in the presence of Cx36 gap junction coupling, expression of ATP-insensitive K_ATP channels results in severe diabetes, whereas in the absence of Cx36 gap junction coupling, only glucose intolerance occurs.

To discover whether the improved glycemic control originated from improvements in stimulus secretion coupling within the islet, we isolated islets from Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} and Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice with their respective littermate controls at 30 days postinduction. Islets isolated from Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} and Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice showed similar levels of GFP coexpression above the level of autofluorescence (Fig. 3A), with on average 50% (each ranging from 30 to 65%) of the islet GFP⁺ (Fig. 3B). This indicates similar levels of high mutant Kir6.2^{[ΔN30,K185Q]} expression in each case.

Significantly elevated insulin secretion was observed at elevated glucose levels (20 mmol/L) in islets from Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice compared with low glucose levels (2 mmol/L), whereas no significant elevation in insulin secretion was observed in islets from Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} mice (Fig. 3C). At elevated glucose levels (20 mmol/L), insulin secretion from Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets was significantly lower compared with Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} islets, as well as compared with Cx36^+/+ and Cx36^−/− control islets. Whereas mean glucose-stimulated insulin secretion from Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} islets was significantly greater than Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets, it was ∼50% less than Cx36^+/+ and Cx36^−/−control islets. Under elevated KCl stimulation, similar levels of insulin secretion were observed in all cases (Fig. 3C). Insulin content was also slightly reduced in Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets compared with Cx36^+/+ controls, whereas similar insulin content was observed in both Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} and Cx36^−/− islets (Fig. 3D). Therefore, in the absence of Cx36 gap junction coupling, ATP-insensitive K_ATP channels have a reduced effect in suppressing glucose-stimulated insulin secretion.

We next examined potential alterations in [Ca²⁺]_i in islets from these mice. At elevated (11 mmol/L) glucose levels, transient elevations (oscillations) in [Ca²⁺]_i were observed in Cx36^+/+ control islets (Fig. 4A, top). These oscillations were rarely observed in Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets at elevated glucose levels (Fig. 4A, middle) but observed frequently in many cells of Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} islets at elevated (20 mmol/L) glucose (Fig. 4A, bottom). The proportion of cells showing transient elevations in [Ca²⁺]_i was significantly greater in Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} islets compared with Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets at elevated glucose (Fig. 4B). [Ca²⁺]_i oscillations were observed at a greater extent in low-GFP⁻ cells in Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} islets compared with high-GFP⁺ cells (Fig. 4B and Supplementary Fig. 2), and to a similar extent as Cx36^+/+ control islets. Interestingly, a large proportion of high-GFP⁺ cells still showed [Ca²⁺]_i oscillations in Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} islets, albeit with a plateau fraction of <10%, compared with between 30 and 90% in low-GFP⁻ cells (Fig. 4A). Upon elevated glucose levels, the time-averaged free-calcium concentration was also significantly reduced in Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets compared with Cx36^+/+ islets (Fig. 4C). At high glucose levels, Cx36^−/− islets normally show a reduced time-average free-calcium concentration compared with wild-type islets (18), and no significant difference was observed between Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} and Cx36^−/− islets. Again, similar elevations in free-calcium concentration were observed in all groups upon elevated KCl. Therefore, in the absence of Cx36 gap junction coupling, ATP-insensitive K_ATP channels have a reduced effect in suppressing glucose-stimulated [Ca²⁺]_i, consistent with the observed recovery of glucose-stimulated insulin secretion.

To further test whether modulating gap junction coupling can enhance [Ca²⁺]_i and insulin secretion upon K_ATP channel opening, this time in the presence of normal β-cell heterogeneity, we applied varying concentrations of the K_ATP activator diazoxide. At 11 mmol/L glucose in Cx36^+/+ islets, 100 μmol/L diazoxide suppressed [Ca²⁺]_i elevations across most of the islet (Fig. 5A, top), whereas in Cx36^−/− islets, many cells still showed [Ca²⁺]_i elevations (Fig. 5A, bottom). [Ca²⁺]_i elevations were significantly greater in Cx36^−/− islets upon 100 μmol/L diazoxide but not upon 250 μmol/L diazoxide (Fig. 5B). Similarly, upon 100 μmol/L diazoxide at 11 mmol/L glucose, insulin secretion from Cx36^−/− islets was significantly greater than Cx36^+/+ islets. Therefore, in a more general case of K_ATP channel opening, a reduction in gap junction coupling elevates [Ca²⁺]_i and insulin secretion.

A secondary consequence of the chronic hyperglycemia and hypoinsulinemia that occurs in Kir6.2^{[ΔN30,K185Q]}-expressing mice is marked mitochondrial dysfunction (6), characterized by reduced mitochondrial NAD(P)H accumulation and reduced mitochondrial membrane depolarization at elevated glucose. We quantified these parameters in mice with reduced gap junction coupling that are protected from diabetes. Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets showed elevated NAD(P)H at low levels of glucose compared with control Cx36^+/+ islets, but similar low levels of NAD(P)H were observed in Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} islets and control Cx36^−/− islets (Fig. 6A and B). Similar levels of NAD(P)H were observed at elevated glucose in all sets of islets (Fig. 6A). As a result, Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets showed significantly reduced glucose-stimulated accumulation of NAD(P)H compared with other experimental groups (Fig. 6C). Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets also showed significantly reduced mitochondrial membrane depolarization compared with Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} islets at elevated glucose, as indicated by Rhodamine 123 fluorescence (Fig. 6D–F). Therefore, reducing gap junction coupling in Kir6.2^{[ΔN30,K185Q]}-expressing mice leads to an absence of secondary mitochondrial dysfunction.

In this study, we tested whether a reduction in islet Cx36 gap junction coupling could compensate for overactive ATP-insensitive K_ATP channels in a model of NDM, and therefore prevent the emergence of diabetes. Upon a knockout of Cx36, the elevation in glucose-stimulated [Ca²⁺]_i and insulin secretion that we measured explains the normalization in blood glucose levels. Based on results presented here and prior studies, we propose the following mechanisms of action, schematically represented in Fig. 7A. A reduction in Cx36 gap junctions effectively left shifts the dose response of [Ca²⁺]_i (although insulin remains suppressed at low glucose due to [Ca²⁺]_i-independent mechanisms of suppression [18]), whereas expression of overactive K_ATP channels effectively right shifts the dose response of [Ca²⁺]_i and insulin secretion (with a largely suppressed response over the physiological glucose range in the presence of Cx36). Therefore, in the presence of overactive K_ATP channels, a reduction in Cx36 partially normalized the dose response, albeit with disrupted insulin secretion dynamics associated with a loss of Cx36. This is further detailed in Fig. 7B, where in the normal presence of Cx36, inexcitable cells that are present due to heterogeneity prevent membrane depolarization and [Ca²⁺]_i elevations in normally excitable cells at elevated glucose, via a gap junction–mediated current. This suppresses [Ca²⁺]_i elevations and insulin release across the islet. In the absence of Cx36 gap junction coupling, this current is absent (17) and therefore normally excitable cells are free to depolarize, elevate [Ca²⁺]_i, and release insulin at elevated glucose levels.

Previous work has shown that membrane depolarization and [Ca²⁺]_i elevations in excitable cells of an islet can be suppressed by gap junction coupling to inexcitable cells. This was demonstrated in islets expressing a K_ATP loss-of-function mutation (closes in the absence of ATP) at low (∼2 mmol/L) glucose (16), where a reduction of gap junction coupling elevated [Ca²⁺]_i. Similar observations have been made in normal islets at basal (∼5 mmol/L) glucose (17,18). Here, in the presence overactive ATP-insensitive K_ATP channels (Fig. 4), or diazoxide-induced K_ATP channel opening (Fig. 5), a reduction of gap junction coupling also elevated [Ca²⁺]_i. This shows that a general principle exists whereby gap junction channel closure/deletion can (at least partially) counteract the effect of K_ATP channel opening to elevate [Ca²⁺]_i.

Importantly, in the presence of a gain-of-function mutation to the K_ATP channels, there was no elevation in [Ca²⁺] and insulin release at lower glucose levels. As described above, likely a reduction in Cx36 imparts a left shift to the glucose-stimulated [Ca²⁺]_i response rather than a constitutive elevation, counteracting the right shift after ATP-insensitive mutant K_ATP channel expression. Thus, a reduction in Cx36 cannot counteract the very strong K_ATP opening that results from low glucose and ATP-insensitive K_ATP channels. This suggests that a Cx36 gap junction reduction would be unlikely to counteract very strong mutations to the K_ATP channel, as is also suggested by very high diazoxide treatment (Fig. 5).

Interestingly, in islets from Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice, even those high-GFP⁺ cells that were inexcitable in the presence of gap junctions showed some [Ca²⁺]_i elevations. These were less frequent than in the more excitable low-GFP⁻ cells but significantly more than in islets of Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} mice (Fig. 4). Electrically uncoupled cells will likely show some stochastic “channel noise,” as suggested by modeling studies (23,24), which may be sufficient to transiently depolarize and elevate [Ca²⁺]_i. This suggests that electrically uncoupled cells may inherently behave different to β-cells in normal islets, warranting further study.

Importantly, the elevated [Ca²⁺]_i at elevated glucose levels that follows a reduction of gap junction coupling also led to elevated insulin secretion (Figs. 3 and 5). This contrasts with the behavior at lower glucose levels in normal islets (18), where this effect is normally masked due to other mechanisms of cell-cell communication that suppress insulin release independent of [Ca²⁺]_i. However, it was suggested that this other communication mechanism(s) that suppresses insulin is inactive at elevated glucose (18,25), as we have demonstrated.

Of further importance, the gap junction–mediated suppression of islet [Ca²⁺]_i and insulin release is physiologically important. By eliminating gap junction coupling in mice expressing ATP-insensitive K_ATP channels, we eliminated hyperglycemia (Fig. 1). Gap junction coupling in the islet has been shown to play an important physiological role: to coordinate and enhance the first phase of insulin secretion and coordinate the pulsatile second phase of insulin secretion (15). In its absence, glucose intolerance occurs, and we observed glucose intolerance in Cx36^−/−;Kir6.2^{[ΔN30,K185Q]} mice, albeit slightly less than previously observed (15), likely due to the younger age of the mice. This physiological role gives a new fundamental understanding of the multicellular properties of the islet, specifically how the electrical coupling of β-cells via gap junction channels can be critically important to coordinate K_ATP channel–regulated electrical activity to regulate insulin secretion and glucose homeostasis.

The majority of cases of NDM (∼60%) occur as a result of gain-of-function mutations to the Kir6.2 or Sur1 subunits of the K_ATP channel (26). In the normal presence of Cx36, we observed similar results to a prior mouse model that expresses a Kir6.2 mutation associated with NDM (4), including marked hyperglycemia, reduced weight gain, reduced islet insulin content, and suppression of glucose-stimulated [Ca²⁺]_i and insulin. Despite the reduced insulin content, normal depolarization-induced [Ca²⁺]_i and insulin release was observed, also as previously observed (4). Given that a reduction of Cx36 gap junction coupling so dramatically normalized glycemic control, we anticipate an improvement in islet function would occur in humans with NDM. However, some subtle differences must be discussed.

The distribution of Kir6.2^{[ΔN30,K185Q]} expression is likely more heterogeneous in islets from this mouse model compared with islets of humans with the disease, due to the stochastic and variable nature of Pdx-Cre^ER–mediated recombination (21). Therefore, the heterogeneity in β-cell function, from which the gap junction–mediated suppression of [Ca²⁺]_i and insulin partly occurs, will not be exactly the same. However, it is well established that human and mouse β-cells lacking cellular proximity are very heterogeneous in their glucose response (27,28). This suggests that a Cx36 gap junction reduction will elevate glucose-stimulated [Ca²⁺]_i and insulin even in the presence of uniform Kir6.2^{[ΔN30,K185Q]} expression. Results from diazoxide-treated islets (Fig. 5) support this, as diazoxide treatment likely provides uniform K_ATP opening across the islet. In this case, the endogenous β-cell heterogeneity would be closer to that of human islets with NDM mutations. Further, gap junction coupling suppresses excitability in other conditions of K_ATP channel opening at lower glucose levels. In Cx36^−/− islets, elevated [Ca²⁺]_i is observed in ∼50% of β-cells at 5 mmol/L glucose (18), and in ∼40% of β-cells upon 100 μmol/L diazoxide (Fig. 5), each similar to that after expression of Kir6.2^{[ΔN30,K185Q]} (Fig. 4). The precise quantitative balance between β-cell heterogeneity and gap junction coupling in determining islet function remains to be determined. Nevertheless, the presence of any heterogeneity will lead to gap junction coupling suppressing [Ca²⁺]_i over a certain range of glucose levels. Likely, the broader the heterogeneity, the greater the difference in the presence and absence of gap junction coupling (18) and the more dramatic the effect in modulating gap junction coupling.

The discovery that NDM can be caused by gain-of-function mutations to the Kir6.2 and Sur1 subunits of the K_ATP channel has meant that patients can switch from insulin therapy to oral sulfonylurea treatment with improved glycemic control (7). Although successful for treating NDM, sulfonylurea treatment has been reported to be associated with hypoglycemic episodes (29), which may occur due to constitutive glucose-independent K_ATP channel closure, which reduces the low glucose regulation of insulin secretion. Chronic sulfonylurea therapy can also cause glucose intolerance (30), possibly resulting from overstimulation of [Ca²⁺]_i. As discussed above, a modulation of gap junction coupling retains a suppression of [Ca²⁺]_i and insulin secretion at low glucose (Figs. 3, 4, and 7) and thus would be anticipated to lessen potential hypoglycemia. More importantly, some Kir6.2 and Sur1 mutations reduce the sensitivity of sulfonylurea inhibition (7,9,10), reducing sulfonylurea effectiveness. We speculate that modulating Cx36 gap junction coupling may provide an alternative route to elevate glucose-stimulated [Ca²⁺]_i and insulin secretion, particularly in the presence of sulfonylurea-insensitive mutations.

Here, a near-complete reduction in gap junction coupling (>95%, Cx36^−/−) normalized blood glucose levels, but a partial reduction (∼50%, Cx36^+/−) had a minimal effect (Supplementary Fig. 1). This Cx36 dose response is similar to how reducing Cx36 gap junction coupling disrupts glucose tolerance via first-phase and second-phase insulin dynamics (15). As a potential therapeutic target, we therefore anticipate that a gap junction inhibition of >50% would be required. Current inhibitors are weak with nonspecific effects, although a recent study developed a novel screen for gap junction modulators (31), which may yield more potent and specific inhibitors. A partial disruption to gap junction coupling can also result from hyperglycemia (32). However, in Cx36^+/+;Kir6.2^{[ΔN30,K185Q]} islets, significant coupling is still present, as shown by the coordinated residual [Ca²⁺]_i. Therefore, any Cx36 decrease resulting from hyperglycemia would likely be insufficient to significantly alter insulin release in this model.

In principle, we anticipate that reducing gap junction coupling could similarly correct defects in insulin release caused by mutations proximal to [Ca²⁺]_i influx, including glucokinase (33,34) and genes that regulate mitochondrial function (35,36). Defects in proximal steps could affect the “amplifying” pathways that regulate insulin release (37), and reduce the effect of gap junction modulation. However, altering the K_ATP regulation of membrane potential can compensate for an absence of glucokinase (38). Therefore, modulating Cx36 gap junction coupling may be more broadly applicable to normalize islet function.

To summarize, we have shown that by reducing gap junction coupling between β-cells in a model of NDM caused by expression of ATP-insensitive K_ATP channels, we can eliminate severe hyperglycemia and islet dysfunction. This is achieved through a novel pathway where a reduction in Cx36 can partially compensate for overactive K_ATP channels and prevent suppression of electrical activity across the islet. This restores glucose-stimulated [Ca²⁺]_i, insulin secretion, and glucose homeostasis. This yields a better understanding of how the islet functions as a coupled unit of β-cells and may ultimately provide a potential therapeutic target for treating NDM and other monogenic forms of diabetes.

Acknowledgments. The authors thank Colin G. Nichols (Washington University, St. Louis, MO) and Maria S. Remedi (Washington University, St. Louis, MO) for helpful comments and suggestions in performing this study and for reviewing this manuscript.

Funding. This study was primarily supported by National Science Foundation grant DGE0742434 (T.H.H.), National Institutes of Health grant R00-DK-085145 (to R.K.P.B.), and University of Colorado internal funds. Experiments on the 3I Marianas spinning disk microscope and the Zeiss LSM510 two-photon microscope were performed through the use of the University of Colorado Anschutz Medical campus advanced light microscopy core (P30-NS-048154 and UL1-RR-025780).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. L.M.N., M.P., and T.H.H. researched data. R.K.P.B. designed experiments, researched data, and wrote the manuscript. R.K.P.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Part of this study was presented in abstract form at the 73rd Scientific Sessions of the American Diabetes Association, Chicago, IL, 21–25 June 2013.