University of Birmingham Optical control of insulin release using a photoswitchable sulfonylurea

Sulfonylureas are widely prescribed for the treatment of type 2 diabetes mellitus (T2DM). Through their actions on ATP-sensitive potassium (K ATP ) channels, sulfonylureas boost insulin release from the pancreatic beta cell mass to restore glucose homeostasis. A limitation of these compounds is the elevated risk of developing hypoglycemia and cardiovascular disease, both potentially fatal complications. Here, we describe the design and development of a photoswitchable sulfonylurea, JB253, which reversibly and repeatedly blocks K ATP channel activity following exposure to violet-blue light. Using in situ imaging and hormone assays, we further show that JB253 bestows light sensitivity upon rodent and human pancreatic beta cell function. Thus, JB253 enables the optical control of insulin release and may offer a valuable research tool for the interrogation of K ATP channel function in health and T2DM.

T ype 2 diabetes mellitus (T2DM) is a global health-care epidemic associated with life-changing sequelae ranging from blindness to cancer 1,2 . This endocrine disease, which currently affects 1 in 12 of the adult population worldwide, involves a disturbance of normal glucose homeostasis due to failure of the pancreatic beta cell mass to adequately compensate for increased peripheral insulin resistance 3 . As such, the rescue of insulin release through the coaxing of beta cell activity remains a therapeutically desirable approach for the long-term restoration of normal glucose levels.
While sulfonylureas are widely prescribed because of their effectiveness and relative inexpensiveness, they have a range of off-target effects that limits their therapeutic use. For example, sulfonylureas can provoke prolonged episodes of low blood glucose due to hyperinsulinemia 17 , elevate cardiovascular disease risk 18 and induce weight gain 19 . Conversely, there is a lack of tools for the precise functional dissection of K ATP channels located not only in the pancreas, but also in the brain 20,21 , heart 22 and vascular smooth muscle 23 . With this in mind, we set out to combine the glucose-lowering attributes of sulfonylureas with the exquisite spatiotemporal control conferred by possession of photoresponsive elements 24,25 .
Here, we showcase JB253, a 'fourth-generation' sulfonylurea based on glimepiride that bears an azobenzene photoswitch, endowing K ATP channels with remarkable photocontrollable properties (Fig. 1a). We demonstrate that JB253 offers sensitive, reversible and repeated manipulation of K ATP channel state and beta cell activity with visible light, yielding optical control over insulin release. Thus, JB253 may allow the selective targeting of K ATP channels in the pancreas and elsewhere.

Results
Design and synthesis of JB253. Distinct substitution patterns are found within different classes of arylsulfonylurea drugs: while there may be a variety of moieties on the aryl-ring, ranging from a simple methyl group in tolbutamide to more complex structures such as a linked pyrrolidinone in glimepiride (Fig. 1b), the terminal nitrogen in sulfonylureas is usually substituted with an aliphatic group. We reasoned that, to generate a photoswitchable analogue, the aromatic core of the sulfonylurea drugs could be extended to an azobenzene. Furthermore, we aimed for a cyclohexyl substituent on the urea moiety that mimics the corresponding substituent on glimepiride. Using a simple threestep procedure commencing with sulfanilamide, N,N-diethylaniline and cyclohexyl isocyanate, JB253 could be synthesized rapidly and inexpensively in large quantities via the sulfonamide-azobenzene (E)-4-((4-(diethylamino)phenyl)-diazenyl)benzenesulfonamide ( Fig. 1c; Supplementary Figs 2 -4). Initial photochromic characteristics of JB253 were measured using a ultraviolet/visible (Vis) spectrophotometer equipped with a monochromator, affording a single broad band as expected for a push-pull-azobenzene-system (l max ¼ 472 nm) (Fig. 1d).
Azobenzenes are known to be photoconverted between their cis-and trans-state by excitation with different wavelengths of light, or alternatively by illumination and dark-relaxation. Indeed, JB253 was readily converted to its cis-state by applying wavelengths ranging from l ¼ 400 to 500 nm (peak l ¼ 472 nm), while thermal relaxation to its trans-state occurred rapidly in the dark. X-ray diffractometry revealed a high degree of structural similarity between trans-JB253 and glimepiride crystals 26 (Fig. 1e) (see Supplementary Table 1). Whereas glimepiride in solution rotates freely around its ethylene carbon chain to adopt various possible binding conformations, JB253 is rigid unless illuminated and so can only adopt two conformations depending on isomeric state (that is, trans-or cis-).
Attaching lipophilic azobenzene units normally renders molecules poorly soluble in water and aqueous buffers, an obvious drawback for their use in biological systems. JB253, however, demonstrates excellent water solubility (Z0.1 mM) when diluted from a 50 mM stock solution in dimethyl sulfoxide (DMSO), presumably due to its acidity (pK a (trans-JB253) ¼ 4.76; see Supplementary Fig. 5). These features were a promising entry point for our subsequent studies using mammalian tissue.
JB253-binding studies. To determine the binding affinity of JB253 to SUR1 relative to a known sulfonylurea (that is, glimepiride), [3H]-glibenclamide displacement assays were performed. JB253 bound SUR1 with a 1,000-fold lower affinity compared with glimepiride, and this was unaffected by illumination (halfmaximal inhibitory concentration (IC 50 ) ¼ 8.3 nM versus 17.6 mM versus 14.8 mM for glimepiride versus trans-JB253 versus cis-JB253, respectively) (Fig. 2a). However, owing to the potential for rapid thermal dark-relaxation during the wash cycles (see below), we were unable to exclude a role for trans-to cis-isomerization in strengthening JB253 binding affinity. Therefore, to compare the activity profiles of trans-JB253, cis-JB253 and glimepiride using a functionally relevant readout, concentration-response experiments were conducted in mouse islets. The effector concentration for half-maximum response (EC 50 ) of cis-JB253 for cytosolic Ca 2 þ rises was found to be 675 nM, similar to that obtained for glimepiride in the same system (EC 50 glimepiride ¼ 399 nM) (Fig. 2b). The concentration-response curve for glimepiride was right-shifted in the presence of a saturating concentration of trans-JB253, demonstrating the presence of competitive agonism even under dark conditions (Fig. 2b).
Since most sulfonlyureas have been reported to bind and activate Exchange Protein directly Activated by cAMP 2A (Epac2A) 27 , an important mediator of insulin secretion 28,29 , the presence of interactions with JB253 was assessed using a Förster resonance energy transfer (FRET)-based approach. To enable this, a full-length Epac2-camps biosensor containing the sulfonylureabinding site was encoded in HEK293t cells 29 . Confirming the existence of sufonylurea-Epac2A interactions, application of either glimepiride (Fig. 2c) or cis-JB253 (Fig. 2d) decreased FRET to a similar extent (DR/R o ¼ 0.052 versus 0.064 a.u., glimepiride versus JB253, respectively; NS, not significant, Student's t-test).
JB253 allows photoswitching of K ATP channels. We sought first to investigate whether JB253 could yield optical control over K ATP channel activity using a system free from confounding effects of glucose metabolism. To enable this, K ATP channels were heterologously expressed in HEK293t cells by transfection with plasmids encoding the Kir6.2 and SUR1 subunits along with green fluorescent protein. Tolbutamide and diazoxide-sensitive inward-rectifying K þ currents could be recorded in transfected cells, confirming the functional assembly of K ATP channels. In the dark state, JB253 partly reduced K þ current amplitude within a few seconds. This was, however, a fraction of that observed during 500 mM tolbutamide application ( Supplementary  Fig. 6a,b). Subsequent illumination of JB253 with wavelengths between 400 and 500 nm further closed the channel (Fig. 3a), with B45-72% block being achieved relative to that recorded using 500 mM tolbutamide (Supplementary Table 2). The reversal potential was close to the expected equilibrium potential for K þ , and this was unaffected by molecule orientation ( À 90.0 ± 1.8 versus À 87.8±1.6 mV, dark versus illuminated; not significant) ( Supplementary Fig. 6c,d). As such, JB253 possesses the advantageous property of becoming a high-affinity K ATP channel blocker upon illumination.
Using a wavelength of 400 nm, heterologously expressed K ATP channels could be repeatedly opened and closed in JB253-treated   Tables 3 and 4). Upon exposure to 400 nm, rapid block was observed (t on ¼ 0.4 s) (Fig. 3c). When the light source was shut off, thermal relaxation was fast, returning the K ATP channel to baseline levels within a couple of seconds (t off ¼ 1.5 s) ( Fig. 3c; Supplementary Tables 3 and 4). While maximal photoblock of K ATP channels was observed at 460 nm, significant effects were also obtained with violet light (that is, 405 nm), which was more compatible with fluorescence imaging of pancreatic beta cell function (see below). Confirming that JB253 could photoswitch endogenous K ATP channels, hyperpolarizing currents were reversibly blocked in MIN6 beta cells following illumination ( Supplementary Fig. 7).

Functional interrogation of beta cells within mouse islets.
Stimulus-secretion coupling in beta cells relies on the closure of K ATP channels, Ca 2 þ influx through VDCC and release of insulin granules 12 . We therefore attempted to manipulate beta cell activity by optically controlling K ATP channels with JB253. Using functional multicellular Ca 2 þ imaging to monitor cell activity directly in situ within intact islets 30,31 , increases in cytosolic free Ca 2 þ , assumed largely to emanate from beta cells under the conditions used here 32 , could be evoked following global illumination using a 405-nm laser ( Fig. 4a) (n ¼ 10 recordings). Just over half (54%) of the Fluo-2-loaded population responded to illumination with synchronous Ca 2 þ rises ( Fig. 4a,b). Demonstrating the utility of JB253 for the fine control of beta cell function, discrete Ca 2 þ oscillations, thought to underlie generation of insulin pulses 33,34 , could be imposed using repeat exposure to 405 nm (Fig. 4c). As anticipated, high doses of tolbutamide and diazoxide were able to augment and suppress, respectively, the effects of JB253 (Fig. 4d,e) (n ¼ 4-6 recordings).
The wavelength required to excite Fluo-2 (l ¼ 491 nm), a commonly used Ca 2 þ indicator, could potentially lead to K ATP channel closure in JB253-treated islets due to cis-isomer formation. We therefore decided to repeat the above experiments using the red (l ¼ 561 nm)-excited Ca 2 þ indicator X-Rhod1. Identical results were obtained for studies with Fluo-2 ( Fig. 5a,b) (63% responsive X-Rhod1-loaded cells) (n ¼ 3 recordings), confirming that the photostationary state of JB253 during brief (263 ms) pulses of 491-nm light was alone insufficient to close K ATP channels in the islet preparation. Likewise, global Ca 2 þ oscillations could be induced in JB253-treated islets by more prolonged illumination with 440-nm and 491-nm laser lines and, as expected from the electrophysiological recordings, both wavelengths appeared to activate a slightly larger cell population (Fig. 5c,d). These observations were unlikely due to K ATP channel closure at 561 nm, since JB253 was unable to photoswitch K þ currents at wavelengths 4560 nm ( Supplementary Fig. 8).
Demonstrating the spatial precision of JB253, a single islet from a doublet could be activated using a targeting laser without significantly stimulating its neighbour (B200 mm from center to center) (Fig. 5e), in part aided by the high molecular extinction coefficient (38,670 mol À 1 cm À 1 at 485 nm; see Supplementary  Fig. 9). Lastly, JB253 did not appear to be cytotoxic to islets, as necrosis indices showed no significant differences in cell death versus DMSO alone (Fig. 5f). Manipulation of human islet function using JB253. To underline the translational potential of JB253 for use in man, Ca 2 þ -imaging experiments were repeated using isolated human islets of Langerhans. As observed for their mouse counterparts, beta cells within JB253-treated human islets responded to 440 and 491 nm with large intracellular Ca 2 þ rises, and oscillations could be coaxed simply by turning the laser on and off (Fig. 6a-c). JB253 effects appeared to be due to K ATP blockade, as they could be mimicked and reversed using tolbutamide and diazoxide, respectively (Fig. 6c,d).
Optical stimulation of insulin release using JB253. To cement the link between photocontrol of K ATP channels, [Ca 2 þ ] i and insulin secretion, islets (from n ¼ 9 mice) were incubated in the presence of JB253 while exposing to either dark (no illumination), 405 nm or 485 nm. Insulin release was similar in control experiments (5 mM glucose, shown to sensitize beta cells to sulfonylurea 35 ) and JB253-treated islets in the dark, suggesting that any K ATP channel block and VDCC activity detected under these conditions was subthreshold for triggering Ca 2 þ -activated exocytosis (Fig. 7). By contrast, JB253-treated islets secreted almost four-to eightfold more insulin following illumination, and this could be partially reversed using diazoxide (Fig. 7). When exposed to 485 nm light, JB253 was equipotent to glimepiride at stimulating insulin secretion (Fig. 7).

Discussion
In the present manuscript, we have described the development and testing of JB253, a chemical chimera of glimepiride and an azobenzene, which allows light-induced closure of K ATP channels. In the primary tissue employed here, viz islets of Langerhans, this translates to activated Ca 2 þ flux and insulin release.
The principles of photopharmacology, that is, the control of biological function with small-molecule photoswitches, are now well established 25,36,37 . In particular, azobenzene photoswitches have been employed as photochromic neurotransmitters and neuromodulators 38,39 , ion channel blockers 24,40 , covalently bound ion channel gates 41 and enzyme inhibitors 42,43 . With respect to ion channels, however, they have mostly been used to optically control excitable cells in the mammalian nervous system, and none have directly targeted K ATP channels. These are ubiquitously expressed channels that contribute to membrane potential in a number of cell types including hypothalamic and hippocampal neurons, cardiac myocytes, vascular smooth muscle and neuroendocrine cells 11,[20][21][22][23]44 . Importantly, K ATP channels translate metabolic state to transmembrane potential and, in pancreatic beta cells, are central to glucose-stimulated insulin secretion [10][11][12] . Since electrical status is generally correlated to biological output in excitable tissues, JB253 may provide a useful tool for investigating K ATP channel function under a range of normal and pathological states.
The prevailing view of sulfonylurea action is one of SUR1 binding, K ATP channel closure and alterations to beta cell membrane potential [7][8][9] . However, recent studies have also invoked a K ATP channel-independent signalling pathway whereby sulfonylurea may alter insulin release via Epac2A interactions [27][28][29] . Using radioactive displacement assays in combination with FRET experiments, JB253 was found to interact with both SUR1 and Epac2A. While SUR1 affinity for JB253 was much lower than glimepiride, we were unable to properly assess the active cis-state due to rapid thermal backrelaxation. As such, a role for illumination in strengthening any interaction cannot be excluded, for example, by altering binding conformation due to isomerization. Nonetheless, JB253 and glimepiride possess similar EC 50 values for intracellular Ca 2 þ rises and, when applied at the same concentration, both compounds stimulated almost identical levels of insulin secretion. Thus, JB253 possesses a similar activity profile to glimepiride, most likely due to signalling via pathways generally acknowledged to underlie sulfonylurea action.
In addition to photopharmacology, optogenetic and artificial light-sensitive K þ channels are equally applicable to the remote control of electrically responsive cells, including beta cells 45,46 . However, therapeutic potential in humans is limited by the requirement for genetic manipulation, high activation irradiances and the hyperpolarizing effects of recombinantly expressed Kir6.2. Alternatively, an implantable synthetic optogenetic transcription device has recently been shown to improve bloodglucose homeostasis in a mouse model of T2DM via the expression and secretion of incretin 47 . However, debate still exists as to whether incretin-based therapies are associated with increased risk of pancreatitis and pancreatic adenocarcinoma 48,49 . Nonetheless, similar concepts have recently been extended to designer fusion molecules and may in the future be adopted for insulin release 50 . By contrast, due to its favourable profile as an exogenously applied sulfonylurea that is sensitive to light, JB253 has advantages both as a research tool and as an anti-diabetic agent. We note, however, that confirmation of glucose-lowering effects in rodents is required before studies using JB253 can be extended to man.
In the context of photodynamic therapy, light penetration in human tissues has been studied in detail and is now well understood 51 . Although, the current activation wavelength of JB253 (400-500 nm) limits deep tissue penetration, for example, through the skin, we have recently begun to develop variants that can be switched at longer wavelengths. In addition, stimulated by the brisk development of optogenetics 52 , devices that can deliver light to target tissues with minimal invasiveness and high spatial precision have emerged 53,54 , although their application to the pancreas is untested.
We therefore speculate that JB253, or related photoswitchable molecules, which regulate K ATP channels, may have an impact on human medicine and research. A long-standing challenge in endocrinology has been the inability to properly recreate the dynamics that underlie pulsatile hormone release, a prerequisite  for proper downstream organ function 55 . Furthermore, disparate biological systems can use similar or identical molecular components. For example, K ATP channels are also expressed in the heart and brain, including in neuronal populations tasked with the central regulation of glucose homeostasis and counterregulatory responses 20,56 . Photopharmacology has the ability to target drug activity to the primary site of dysfunction with high spatial and temporal resolution 37 . JB253 holds particular promise in this regard. It is non-cytotoxic and can be used to repeatedly modulate rodent and human beta cell activity, the basis for recreating the oscillatory activity known to orchestrate hormone pulses 33,34 . Its light dependency means  In summary, we have designed and synthesized a lightsensitive sulfonylurea, JB253, which has a broad spectrum of application due to conferment of photoswitching on K ATP activity.

Methods
Chemical synthesis. (E)-4-((4-(Diethylamino)phenyl)diazenyl)benzenesulfonamide (1). Sulfanilamide (2.00 g, 11.61 mmol, 1.0 eq.) was dissolved in 2.4 M HCl and cooled to 0°C. Under vigorous stirring, a solution of NaNO 2 (0.96 g, 13.91 mmol, 1.2 eq.) in 6 ml water was added dropwise until the solution turned pale yellow. The formed diazonium salt was stirred under ice-cooling for an additional 10 min before it was transferred into a solution of N,N-diethylaniline (1.73 g, 11.61 mmol, 1.84 ml, 1.0 eq.) in a 1/1 mixture of 1 M NaOAc/MeOH. The solution turned to dark red and was allowed to warm to room temperature under stirring. The crude product was extracted with EtOAc (3x), and the combined organic layers were washed with brine and dried over MgSO 4 . Flash column chromatography (25% EtOAc/i-hexanes) yielded 1.45 g (4.37 mmol) of the desired product as a red powder in 38% yield. General chemistry. Flash column chromatography was carried out on silica gel 60 (0.040-0.063 mm) purchased from Merck. Reverse phase flash column chromatography was carried out on Waters C18 silica gel (0.055-0105 mm, 125 Å). Reactions and chromatography fractions were monitored by thin-layer chromatography on Merck silica gel 60 F254 glass plates. The spots were visualized either under ultraviolet light at 254 nm or with appropriate staining method (iodine, p-anisaldehyde, KMnO 4 ) followed by heating.
A Varian MAT CH7A mass spectrometer was used to obtain low-and highresolution electron impact mass spectra. Low-and high-resolution ESI mass spectra were obtained on a Varian MAT 711 MS instrument operating in either positive or negative ionization modes.
Solvents for column chromatography and reactions were purchased in HPLC grade or distilled over an appropriate drying reagent before use. If necessary, solvents were degassed either by freeze-pump-thaw or by bubbling N 2 through the vigorously stirred solution for several minutes. Unless otherwise stated, all other reagents were used without further purification from commercial sources.
Ultraviolet/Vis spectra were recorded on a Varian Cary 50 Bio UV-Visible Spectrophotometer using Helma Suprasil precision cuvettes (10 mm light path).
LC-MS was performed on an Agilent 1260 Infinity HPLC System, MS-Agilent 1100 Series, Type: 1946D, Model: SL, equipped with a Agilent Zorbax Eclipse Plus C18 (100 Â 4.6 mm, particle size 3.5 micron) RP column.  Infrared spectra were recorded on a PerkinElmer Spectrum BX-59343 instrument as neat materials. For detection, a Smiths Detection DuraSam-plIR II Diamond ATR sensor was used. The measured wavenumbers are reported in cm À 1 .
Melting points were measured on an EZ-Melt apparatus (Stanford Research Systems) and are uncorrected.
Extinction coefficients were measured on a BMG Labtech Omega Series FLUOstar microplate reader with clear flat-bottom white 96-well plates by full spectra acquirement in low-K þ external bath buffer (containing in mM: 3 KCl, 118 NaCl, 25 NaHCO 3 , 2 CaCl 2 , 1 MgCl 2 , 10 HEPES; NaOH to pH 7.4). All JB253containing solutions (dilution series, n ¼ 4: 10 nM; 100 nM; 1 mM; 10 mM; 25 mM; 50 mM) were background substracted and fitted with a linear slope. Volumes were 100 ml each, which resulted in a path length of l ¼ 2.94 mm. pK a measurements and data processing were performed using the same instrument and protocol as in the study by Martinez and Dardonville 57 . Full absorbance spectra (280-800 nm) were acquired and background subtracted before spectral differences were calculated. The total change of maximal positive and maximal negative difference was calculated and plotted against pH. Sigmoidal fit of the obtained plot gave access to the pK a .
Crystallography. X-Ray data collection was performed on a Bruker D8Venture at 173 K using MoKa-radiation (l ¼ 0.71073 Å). The APEX2 (v2012.4-3, Bruker AXS Inc.) software and embedded programs were applied for the integration, scaling and multi-scan absorption correction of the data. The structures were solved by direct methods with SIR97 58 and refined by least-squares methods against F2 with SHELXL-97 59 . All non-hydrogen atoms were refined anisotropically. The C-bound hydrogen atoms were placed in ideal geometry riding on their parent atoms, N-bound hydrogen atoms were refined freely. The crystallographic data for JB253 is available free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif (accession ref. CCDC 1014606).
Mouse islet isolation. Male and female CD1 and C57BL6 mice (8-20 weeks) were maintained in a specific pathogen-free facility under a 12 h light-dark cycle with ad libitum access to water and food. Animals were euthanized using a schedule-1 method and pancreatic islets isolated by collagenase digestion. All procedures were regulated by the Home Office according to the Animals (Scientific Procedures) Act 1986 of the United Kingdom (PPL 70/7349), and study approval granted by the Animal Welfare and Ethical Review Body of Imperial College. No randomization was used for animal experimentation, since mice were only used as tissue donors.
Human islet isolation. Human islets were isolated from deceased heart-beating donors (n ¼ 3) at transplantation facilities in Pisa and Edmonton with the relevant national and local ethical permissions, including consent from next of kin where required, and cultured in Roswell Park Memorial Institute medium supplemented with 5.5 mM D-glucose, 10% foetal calf serum, 100 U/ml penicillin, 100 mg/ml streptomycin and 0.25 mg/ml fungizone (37°C, 5% CO 2 ). All studies involving human tissue were approved by the National Research Ethics Committee London (Fulham), REC #07/H0711/114. Calcium imaging. Islets were loaded for 30-45 min in Fluo2-AM (10 mM) or for 2-5 min with X-Rhod1 (5 mM) diluted with a mixture of DMSO (0.01%, wt/vol) and pluronic acid (0.001%, wt/vol; all Invitrogen) in a bicarbonate buffer containing in mM: 120 NaCl, 4.8 KCl, 1.25 NaH 2 PO 4 , 24 NaHCO 3 , 2.5 CaCl 2 , 1.2 MgCl 2 and 5 D-glucose. Functional multicellular Ca 2 þ imaging was performed using a Zeiss Axiovert M200 fitted with a Nipkow spinning-disk head (Yokogawa CSU-10) and a Â 10/0.3 numerical apperture objective adjusted for chromatic aberration (EC Plan-Neofluar, Zeiss). Pulsed excitation (frequency ¼ 0.5 Hz; exposure ¼ 263 ms) was delivered at 491 nm and emitted signals recorded at 500-550 nm with a back-illuminated 16-bit EM-CCD camera (ImageEM 9100-13; Hamamatsu). During recording, islets were maintained at 35-36°C in the presence of 50 mM JB253 using a custom-manufactured perfusion and heating system (Digital Pixel). Drugs were introduced through the perfusion system at the indicated time points and concentrations. Violet light was delivered by a 405±5-nm laser coupled to the side port of the microscope and configured to fill the back of the objective with light using an Optospot (Cairn Research). Blue light was delivered using 440 ± 5nm and 491±5-nm diode lasers controlled by a laser merge module (Spectral Applied Research) to allow simultaneous exposure and acquisition. For single islet targeting, a 473±5 nm laser was coupled to a custom-manufactured dichroic array (Cairn Research), allowing user-directed steering of a collimated laser spot across the field of view. Signals were normalized using F/F min where F is fluorescence at a given time point and F min is minimum fluorescence.
Cytotoxicity assay. Islets were incubated with either DMSO or JB253 for 1 h before staining with 3 mM of calcein-AM (live) and 2.5 mM of propidium iodide (dead). Absorbance/emission was detected at 491/525 and 561/620 nm for calcein and PI, respectively. The area of dead:live cells was calculated as a unitary ratio and the observer blinded to treatment identity.
Epac2 imaging. For Epac2 imaging, HEK293t were transfected with the full-length construct for Epac2-camps containing the cAMP-and sulfonylurea-binding domains 29 (a kind gift from Prof. Jin Zhang, Johns Hopkins University) before imaging 60 using a HEPES-bicarbonate buffer containing in mM: 120 NaCl, 4.8 KCl, 24 NaHCO 3 , 0.5 Na 2 HPO 4 , 5 HEPES, 2.5 CaCl 2 , 1.2 MgCl 2 and 5 D-glucose. Excitation was delivered at 440 nm and emitted signals captured using cerulean (530 nm) and citrine (470 nm) filters. FRET was calculated as the ratio of Cerulean (CFP):Venus (YFP) fluorescence. Signals were normalized using R/R o, where R is the ratio at a given time point and R o is the minimum ratio.
Statistical analysis. Data distribution was determined using the D'Agostino omnibus test. Non-multifactorial pairwise comparisons were made using the Student's t-test. Interactions between multiple treatments were assessed using one-way analysis of variance followed by pairwise comparisons using Bonferroni's post hoc test. Nonlinear regression was used to calculate the EC 50 of normalized and log-transformed concentration-response curves. For [3H]-glibenclamide displacement assays, data points were fitted to the Hill equation before calculation of the halfmax value. In all cases, analysis was performed using Graphpad Prism (Graphpad Software) and IgorPro, and experimental numbers reported as independent biological replicates. No animals or data were excluded from the analysis and results were considered significant at Po0.05. Effect sizes in islets/cells are usually sufficiently large that multiple animals/independent replication is a more important determinant of power in studies requiring statistical comparison.