Autocrine Signaling Underlies Fast Repetitive Plasma Membrane Translocation of Conventional and Novel Protein Kinase C Isoforms in β Cells*

PKC signaling has been implicated in the regulation of many cell functions, including metabolism, cell death, proliferation, and secretion. Activation of conventional and novel PKC isoforms is associated with their Ca2+- and/or diacylglycerol (DAG)-dependent translocation to the plasma membrane. In β cells, exocytosis of insulin granules evokes brief (<10 s) local DAG elevations (“spiking”) at the plasma membrane because of autocrine activation of P2Y1 purinoceptors by ATP co-released with insulin. Using total internal reflection microscopy, fluorescent protein-tagged PKCs, and signaling biosensors, we investigated whether DAG spiking causes membrane recruitment of PKCs and whether different classes of PKCs show characteristic responses. Glucose stimulation of MIN6 cells triggered DAG spiking with concomitant repetitive translocation of the novel isoforms PKCδ, PKCϵ, and PKCη. The conventional PKCα, PKCβI, and PKCβII isoforms showed a more complex pattern with both rapid and slow translocation. K+ depolarization-induced PKCϵ translocation entirely mirrored DAG spiking, whereas PKCβI translocation showed a sustained component, reflecting the subplasma membrane Ca2+ concentration ([Ca2+]pm), with additional effect during DAG spikes. Interference with DAG spiking by purinoceptor inhibition prevented intermittent translocation of PKCs and reduced insulin secretion but did not affect [Ca2+]pm elevation or sustained PKCβI translocation. The muscarinic agonist carbachol induced pronounced transient PKCβI translocation and sustained recruitment of PKCϵ. When rise of [Ca2+]pm was prevented, the carbachol-induced DAG and PKCϵ responses were somewhat reduced, but PKCβI translocation was completely abolished. We conclude that exocytosis-induced DAG spikes efficiently recruit both conventional and novel PKCs to the β cell plasma membrane. PKC signaling is thus implicated in autocrine regulation of β cell function.

PKCs are involved in various aspects of ␤ cell function, like proliferation, differentiation, and death, as well as insulin secretion (11), but the precise role of various isoforms is difficult to define because of a lack of selective pharmacological tools and potential problems with compensatory mechanisms and functional redundancy in genetic ablation studies. PKC-activating phorbol esters were found early to stimulate insulin secretion (12,13), an effect mediated by sensitization of the secretory machinery to Ca 2ϩ (14,15). Although the involvement of PKC in the regulation of insulin release by G protein-coupled receptor stimuli is well established (11), its role in glucose-stimulated secretion is controversial. Down-regulation of PKC activity has little effect on the secretory response to glucose (16,17), and experiments with PKC inhibitors have yielded conflicting results (18 -23). Studies with genetic ablation of different PKC isozymes indicate that PKC␦ and / are important for insulin secretion either by direct effects on the exocytosis machinery (24) or by controlling the expression of genes important for ␤ cell differentiation (4). However, adenoviral overexpression of wild-type and kinase-dead forms of PKC␣ and ␦ failed to affect glucose-stimulated insulin secretion in another study (18). Also, the functional importance of PKC⑀ is unclear. Expression of a dominant negative mutant suppressed exocytosis in isolated ␤ cells (25), and a specific PKC translocation inhibitor reduced insulin secretion from rat islets (21). However, functional ablation of the protein did not affect glucose-induced insulin secretion and even amplified that from islets treated with fatty acids (26).
Activation of conventional and nPKCs typically involves their translocation to the plasma membrane. For nPKCs, this process is mostly DAG-driven, and for cPKCs, it depends on a combination of Ca 2ϩ and DAG dynamics (27,28). We recently discovered that glucose induces rapid DAG elevations in restricted regions of the ␤ cell plasma membrane with durations of Ͻ10 s. They reflect the exocytotic release of adenine nucleotides with autocrine feedback activation of P2Y 1 purinoceptors (29). Moreover, these DAG microdomains induced PKC activation, measured as translocation of fluorescencelabeled myristoylated alanine-rich C-kinase substrate (MARCKS). However, it remains unclear whether the DAG microdomains trigger PKC translocation or whether DAG spiking merely activates PKCs already present at the plasma membrane. It also remains to be establish which subclasses of PKCs are activated and whether the different isoforms respond differentially to the glucose-induced DAG signaling pattern. Therefore, we analyzed the translocation dynamics of various fluorescence-tagged PKC isoforms and DAG dynamics as well as the subplasma membrane Ca 2ϩ concentration ([Ca 2ϩ ] pm ) in insulin-secreting ␤ cells using total internal reflection fluorescence (TIRF) microscopy.

Results
Depolarization-induced DAG Spiking Triggers nPKC-dependent MARCKS Phosphorylation-TIRF imaging of MIN6 insulinoma cells co-expressing an mCherry-tagged DAG sensor and the PKC activity detector MARCKS-GFP showed that K ϩ -mediated membrane depolarization induced repetitive, brief (Ͻ10 s) and pronounced DAG elevations. Each DAG spike was associated with less rapid dissociation of MARCKS-GFP from the plasma membrane (Fig. 1, A and B). Gö 6976, an inhibitor of cPKCs, had little effect on these responses, whereas Gö 6983, targeting both conventional and nPKCs, immediately suppressed MARCKS-GFP translocation without affecting DAG spiking (Fig. 1, A and B). These data indicate that DAG spiking primarily activates nPKCs in ␤ cells.
Glucose-induced Plasma Membrane Translocation of nPKCs Reflects DAG Spiking-MIN6-cells were next co-transfected with the DAG biosensor and different GFP-tagged PKC isoforms. All nPKCs tested (␦, ⑀, and ) showed rapid, transient, and repetitive glucose-induced translocation between the cytoplasm and the plasma membrane in response to glucose, whereas the muscarinic agonist carbachol induced sustained membrane association, almost perfectly mirroring simultaneously measured DAG patterns (Fig. 2, A-C). The glucose-induced DAG spikes were often spatially confined, and the nPKCs translocated specifically to the membrane regions with elevated DAG, as illustrated for PKC⑀ in Fig. 2D. PKC⑀ was further investigated to evaluate the DAG dependence of the PKC translocation. Like glucose stimulation, membrane depolarization with a high K ϩ concentration resulted in parallel DAG spiking and PKC⑀-GFP translocation (Fig. 3A). Secretagogue-induced DAG spiking in ␤ cells is due to exocytotic release of ATP with autocrine feedback activation of P2Y 1 purinoceptors, which in turn activates phospholipase C (29). Consistent with DAG spiking underlying the brief plasma membrane binding of PKC⑀, both events were prevented by the P2Y 1 receptor inhibitor MRS 2179 (Fig. 3A). Similarly, introduction of MRS 2179 before K ϩ depolarization prevented the appearance of DAG spikes as well as the concomitant transient PKC⑀ translocation (Fig. 3B) without affecting the sustained depolarization-induced increase of [Ca 2ϩ ] pm (Fig. 3C). Conversely, brief DAG increases caused by repeated 5-s applications of 0.1 M P2Y 1 receptor agonist MRS 2365 caused transient translocations of PKC⑀ to and from the plasma membrane, whereas continuous application of the drug induced sustained responses (Fig. 3D).
The stable acetylcholine analogue carbachol activates phospholipase C, and the resulting increases in DAG and cytoplasmic Ca 2ϩ concentrations induce PKC activation. Two 5-min periods of carbachol stimulation 15 min apart resulted in comparable plasma membrane DAG increases and PKC translocation dynamics (Fig. 4A). PKC⑀-GFP translocation to the plasma membrane was sustained and typically very similar to the DAG dynamics (Fig. 4A). Omission of Ca 2ϩ from the extracellular medium together with addition of EGTA and the Ca 2ϩ -ATPase inhibitor cyclopiazonic acid to prevent elevations of the cyto-  plasmic Ca 2ϩ concentration reduced carbachol-induced DAG production, probably as a result of elimination of positive feedback from Ca 2ϩ on phospholipase C (31), and PKC⑀-GFP translocation was consequently also slightly reduced (Fig. 4, B and C). These findings show that the glucose-induced DAG microdomains are associated with translocation of nPKCs to the plasma membrane and reinforce the notion that DAG is sufficient for driving translocation of nPKCs.

Glucose Induces Complex Oscillatory cPKC Translocation
Reflecting both Ca 2ϩ and DAG Dynamics-Experiments were also performed with MIN6 cells co-expressing the DAG biosensor and either of the GFP-tagged cPKC isoforms PKC␣, ␤I, or ␤II. Elevation of the glucose concentration to 11 mM induced prominent, repetitive translocation of all three cPKC isoforms (Fig. 5, A-C). The translocation pattern of PKC␤II differed slightly from that of PKC␣ and PKC␤I. Of 10 cells expressing PKC␤II-GFP, four showed DAG increases and PKC translocation even at a substimulatory glucose concentration (data not shown). In the remaining cells, 11 mM glucose induced prominent, repetitive translocation of PKC␤II to the plasma membrane with kinetics strikingly similar to that of DAG, and few glucose-induced PKC␤II recruitments to the membrane occurred without a concomitant DAG spike (Fig. 5B). Carba-chol triggered pronounced and sustained recruitment of PKC␤II to the plasma membrane, mirroring the DAG dynamics (Fig. 5B).
The translocation pattern of PKC␤I consisted of a small, sustained increase of fluorescence with superimposed, very pronounced (Ͼ3-fold increases in fluorescence) repetitive translocation peaks that only partially reflected parallel DAG spiking (Fig. 5, C and D). Although carbachol transformed the DAG pattern to a sustained increase, PKC␤I continued to show fast, repetitive translocation from an elevated level that sometimes fluctuated slowly (Fig. 5, C and D). The fast PKC␤I translocations that did not coincide with DAG spikes are most likely driven by rapid glucose-induced changes in the cytoplasmic Ca 2ϩ concentration. Fig. 5, E and F, shows an example of a glucose-stimulated cell with slow oscillations of PKC␤I translocation (duration Ͼ1 min) and superimposed spiking almost perfectly mirroring slow and fast oscillations of [Ca 2ϩ ] pm . With PKC␤I, it was more difficult to observe spatially restricted translocation coinciding with DAG microdomains than it was with the nPKCs. This was because PKC␤I translocation spikes were so frequent and clearly dependent on factors in addition to DAG. Fig. 5G shows one of the rather infrequent examples of an isolated PKC␤I translocation event paralleled by local DAG generation.
Membrane depolarization with a high K ϩ concentration resulted in sustained plasma membrane translocation of PKC␤I-GFP with superimposed spiking (Fig. 6A). This PKC␤I-GFP spiking was inhibited when DAG spiking was prevented by P2Y 1 receptor inhibition with MRS 2179 (Fig. 6, A and B). However, the sustained translocation of PKC␤I-GFP was unaffected by the drug (Fig. 6A), showing dynamics strikingly similar to those of [Ca 2ϩ ] pm , with a slight time-dependent decline (Fig. 6, C and D). Likewise, introduction of MRS 2179 before K ϩ depolarization did not affect the subsequent [Ca 2ϩ ] pm elevation and stable PKC␤I translocation but prevented the appearance of rapid DAG and [Ca 2ϩ ] pm spikes as well as the concomitant transient PKC␤I translocations (Fig. 6, E and F).
In the presence of 3 mM glucose, carbachol triggered an initial distinct peak of PKC␤I-GFP translocation to the plasma membrane, followed within a few seconds by dissociation and stabilization at a sustained plateau corresponding to 15% Ϯ 3% of the peak translocation (n ϭ 19; Fig. 7, A and B). Similar kinetics were not always observed for DAG production, and the plateau phase corresponded to 83% Ϯ 5% of the peak amplitude (n ϭ 19, Fig. 7A). PKC␤I translocation dynamics were reminiscent of those of [Ca 2ϩ ] pm , which were dominated by a pronounced initial peak caused by inositol-1,4,5-trisphosphate-mediated mobilization of Ca 2ϩ from the endoplasmic reticulum, followed by a modest stable elevation, mainly reflecting store-operated Ca 2ϩ entry (Fig. 7B) (32). The sustained PKC␤I-GFP translocation probably reflects the Ca 2ϩ dependence of DAG binding (33,34). Omission of Ca 2ϩ from the extracellular medium together with addition of EGTA and cyclopiazonic acid prevented the carbachol-induced rise of [Ca 2ϩ ] pm (Fig. 7B) and effectively prevented plasma membrane translocation of PKC␤I-GFP (Fig. 7, C and D). Our findings reinforce the notion that the translocation of cPKCs is strictly   (5)(6)(7)(8)(9)(10). Although 100 M carbachol or rise of glucose elicited pronounced DAG signaling, there was no consistent PKC-GFP translocation to the plasma membrane (Fig. 8A). A weak tendency of carbachol to increase PKC-GFP membrane fluorescence was probably unspecific and reproduced in control experiments with plasma membrane-anchored GFP (data not shown). Because PKC-GFP might be responsive to phosphatidylinositol 3,4,5-trisphosphate, we investigated the effect of insulin. At 100 -300 nM, insulin induced a phosphatidylinositol 3,4,5-trisphosphate increase of similar magnitude as glucose but without recruitment of PKC-GFP to the plasma membrane (Fig. 8B).

Inhibition of P2Y 1 Receptors Suppresses Glucose-induced
Insulin Secretion-To clarify whether DAG spiking and PKC translocation contributed to glucose-induced insulin release, we measured insulin secretion dynamics from MIN6 pseudoislets. Increase of glucose from 3 to 20 mM induced pulsatile insulin release that deteriorated after introduction of 10 M MRS 2179 (Fig. 9A). After averaging data from five experiments, pulsatility was no longer evident, but there was significant suppression of insulin release during MRS 2179 exposure (time-average secretion, 64% Ϯ 9% of the preceding 10-min control period; p Ͻ 0.01; Fig. 9B).

Discussion
This study demonstrates that exocytosis-induced microdomains of DAG recruit both conventional and nPKCs to the plasma membrane in ␤ cells. Autocrine signaling is thus involved in controlling the spatiotemporal dynamics of PKC activity in glucose-stimulated ␤ cells. In contrast to the view that DAG changes slowly and that Ca 2ϩ accounts for the rapid translocation of PKC (28,35), this study highlights that fast, oscillatory PKC translocation can be determined by DAG spiking, whereas more sustained translocation is Ca 2ϩ -dependent.
PKC translocation dynamics has previously been investigated in insulin-secreting ␤ cells. However, glucose-induced plasma membrane recruitment of PKC␤I-GFP has not been observed before, probably because of insufficient time resolution and difficulties to specifically detect changes in plasma membrane fluorescence (25). PKC␣ and PKC␤II have been reported to undergo oscillatory membrane translocation in glucose-stimulated primary ␤ cells (18,23), probably reflecting slow oscillations of the free cytosolic Ca 2ϩ concentration (23,36). Endogenous PKC␣ has been found to strongly associate with the plasma membrane during the first phase of insulin secretion, followed by a dissociation and subsequent reassociation during second-phase insulin release (10). We now discovered more complex and rapid translocation dynamics that were strikingly parallel with DAG spiking. We propose that the previously described translocation dynamics merely reflect the Ca 2ϩ -driven component of cPKC activation and that improved temporal resolution and selective imaging of the plasma membrane now reveal a superimposed component of transient, rapid, and local DAG-mediated plasma membrane binding of cPKCs. Subtle variations in translocation patterns among closely related isoforms indicate that the PKCs are fine-tuned to respond differently to distinctive DAG and Ca 2ϩ signals.
Although DAG-dependent activation of nPKCs is well established, it is notable that all tested isoforms were capable to respond to glucose-induced DAG spiking with strikingly rapid translocation to the plasma membrane. Several previous studies failed to detect glucose-induced plasma membrane association of PKC␦ (10,36,37), and in HEK293 cells, this isoform translocates to the endoplasmic reticulum after activation of purinergic receptors (38). Another study stressed a difference between the ␦ and ⑀ isoforms, with PKC⑀ translocating to the cell periphery and PKC␦ to perinuclear sites in response to glucose (21). The presently observed similarity in plasma membrane translocation dynamics of the two isoforms was therefore somewhat surprising, especially because PKC␦ and ⑀ often have different or even opposing, effects (39). However, a similar spatiotemporal pattern at the plasma membrane does not exclude distinct activation at other subcellular sites. Endogenous PKC⑀ was reported to associate with insulin staining near the nucleus, and glucose was found to induce changes in staining intensity, reflecting the biphasic insulin response of the perfused rat pancreas (10). Although the spatial resolution in the latter study did not permit a definite localization of the staining to the secretory granules, confocal imaging of PKC⑀-GFP supported a glucoseinduced association of PKC⑀ to insulin granules in INS1E insulinoma cells, with the most prominent effect near the plasma membrane (25). No granular pattern was evident in the presently observed PKC⑀ distribution, indicating that the protein also can interact directly with the plasma membrane.
PKC differed from the other nPKCs in that transient membrane translocation sometimes occurred without a simultaneously detected DAG elevation. The reason may be a higher DAG affinity of the C1 domain of PKC compared with that of the PKC␥-derived DAG sensor (27). PKC has been found previously to locate to the cytoplasm in rat islets (5) and to membranes in RINm5F cells (8), but, unlike in this study, carbachol did not affect the distribution pattern in either case.
The aPKCs lack a Ca 2ϩ -binding C2 domain, and their C1 domain variant is unable to bind DAG (1,2). The carbacholinduced membrane translocation is therefore likely mediated by protein-protein interactions via the PB1 domain, which is specific for aPKCs (1,2). Carbachol-induced translocation of PKC to the plasma membrane has been detected previously by Western blotting and suggested to mediate carbachol-stimulated insulin release in RINm5F cells (8). Another study using various PKC inhibitors concluded that glucose-induced insulin secretion is, at least in part, dependent on activation of an aPKC isoform (20). Using immunohistochemistry, Warwar et al. (10) demonstrated that glucose induces transient translocation of PKC to the plasma membrane, corresponding to first-phase insulin secretion, and that prolonged stimulation led to accumulation of PKC in the nucleus. The present findings do not support the view that glucose or carbachol cause rapid association of PKC with the plasma membrane but do not allow conclusions about its localization or activity in other subcellular compartments. Our findings strengthen the idea that DAG spiking underlies the secretagogue-induced, repetitive, brief plasma membrane associations of novel and cPKCs. They also emphasize the requirement of DAG production for nPKC translocation and suggest that cPKCs can associate with the plasma membrane without prominent increases in DAG concentration. The modest, stable DAG elevation caused by membrane depolarization in the presence of MRS 2179 (Figs. 3B and 6D) probably results from Ca 2ϩ -driven phospholipase C activity (29, 40 -42). Membrane binding of cPKCs is also mediated via interaction of the Ca 2ϩ -binding C2 domain with other lipids, primarily phosphatidylserine, phosphatidylinositol-4,5-bisphosphate, and phosphatidylinositol 3,4,5-trisphosphate (43,44). The C2 domain of nPKCs is not required for membrane binding. Instead, their C1 domain has an increased affinity for DAG (33). Moreover, the C1 domain binds selectively and stereospecifically to phosphatidylserine-containing membranes in a DAG-dependent manner (33,34), which is consistent with our finding that PKC⑀ did not translocate when DAG spiking was prevented.
The very short periods of PKC activation may explain why it has been difficult to detect increased PKC-mediated phosphorylation in glucose-stimulated ␤ cells (18). Our data clearly show that a brief DAG spike with concomitant PKC translocation is sufficient for inducing protein phosphorylation. In ␤ cells, the MARCKS protein seems to be phosphorylated primarily by an nPKC. The transient nature of these PKC signaling events should be suitable for regulation of rapid processes such as exo-cytosis. Indeed, several components of the exocytosis machinery, including SNAP25 and Munc18, are targets for PKC (45,46). Studies of the effects of PKC inhibitors on insulin secretion have nevertheless yielded contradictory results (18 -23). The inconsistencies may in part be explained by different effects of PKC on the initial and late phases of secretion (23). Another possibility is that PKC is important for the periodic pattern of insulin release. Accordingly, inhibition of the purinergic feedback mechanism that underlies the DAG spiking and transient PKC translocation markedly perturbed pulsatile insulin secretion (Fig. 9). Similar findings have been reported previously from the perifused rat pancreas using a conventional immunoassay (47) and from isolated mouse insulinoma cells using a single-cell optical assay (29). The disrupted pulsatility was associated with either inhibition (Ref. 29 and this study) or stimulation (47) of insulin secretion, a discrepancy that is probably due to differences in experimental preparations. The effects of adenine nucleotides on ␤ cells are pleiotropic and involve the action of several purinergic receptors (reviewed in Refs. 48,49). Recent work in human ␤ cells confirmed that P2Y 1 receptors mediate autocrine stimulation of insulin secretion (50). Further investigations are required to define the functional importance of the various PKC isoforms in ␤ cells. In conclusion, insulin secretagogues induce transient DAG microdomains that rapidly recruit both conventional and novel PKCs to the ␤ cell plasma membrane. The findings implicate PKC signaling in the autocrine regulation of ␤ cell function and emphasize a role of DAG for rapid kinetic control of PKC-dependent cellular processes. for the relative amplitudes of the peak and following plateau of simultaneously recorded DAG production and PKC␤I translocation in response to two consecutive carbachol applications in the presence or absence of Ca 2ϩ (n ϭ 18 and 10 cells from three and two experiments for control and 0 Ca 2ϩ , respectively). *, p Ͻ 0.05; **, p Ͻ 0.0007; ***, p Ͻ 3 ϫ 10 Ϫ5 for the difference from the control (ctrl).

Experimental Procedures
Reagents and DNA Constructs-MRS 2179 tetrasodium salt, MRS 2365, Gö 6976, and Gö 6983 were purchased from Tocris Bioscience (Bristol, UK). Bovine serum albumin was from Roche Diagnostics and physiological salts for the experimental buffer from Merck. Cyclopiazonic acid, EGTA, and HEPES were purchased from Sigma-Aldrich (St. Louis, MO). The plasmids containing GFP-labeled PKCs and MARCKS-GFP were provided by Prof. Tobias Meyer (Stanford University) and Prof. Naoaki Saito (Kobe University), respectively. A DAG biosensor based on the C1 domain tandem repeat in PKC␥ (␥C1aC1b-mCherry) was created as described earlier (29), and the genetically encoded Ca 2ϩ sensor R-GECO (51) was used for measurements of the cytoplasmic Ca 2ϩ concentration. mCherry targeted to the plasma membrane by a CAAX motif was used as plasma membrane marker.
Cell Culture and Transfection-If not otherwise stated, all cell culture reagents were from Life Technologies. Insulin-secreting MIN6 insulinoma cells (30) of passages 17-31 were cultured in DMEM containing 25 mM glucose and supplemented with 2 mM glutamine, 70 M 2-mercaptoethanol, 100 units/ml penicillin, 100 g/ml streptomycin, and 15% fetal calf serum and kept at 37°C in a humidified atmosphere with 5% CO 2 . Cells were transfected while being seeded onto 25-mm coverslips (Menzel-Gläser, Thermo Fisher Scientific, Waltham, MA) coated with polylysine (0.01 mg/ml). For each coverslip, ϳ0.2 million cells were suspended in 100 l of Opti-MEM medium containing 0.5 l of Lipofectamine TM 2000 with up to 0.3 g of plasmid DNA and plated onto the glass. After 3 h, when the cells were attached, the transfection was interrupted by addition of 3 ml of complete culture medium. Experiments were conducted after 13-36 h of further culture. For insulin secretion experiments, 1.5 million MIN6 cells were allowed to form pseudoislets by culture in a 60-mm polystyrol Petri dish (Sarstedt, Nümbrecht, Germany) for 4 days.
TIRF  with the cells was mounted in an open 50-l chamber and superfused with buffer at a rate of 0.3 ml/min. All experiments were performed at 37°C.
The plasma membrane localization of GFP-tagged PKCs and the DAG translocation biosensor as well as the intensity of the Ca 2ϩ sensor fluorescence were measured with a TIRF microscopy setup consisting of an Eclipse Ti microscope (Nikon) with a TIRF illuminator and a ϫ60, 1.45 numerical aperture objective. The 488-nm line of a multiline argon laser (ALC 60X, Creative Laser Production, Munich, Germany) and the 561-nm line from a diode-pumped solid-state laser (Jive, Cobolt AB, Solna, Sweden) were selected by narrow bandpass filters (Semrock, Rochester, NY) in a filter wheel (Sutter Instruments, Novato, CA) and used for excitation of GFP (488 nm), mCherry, and R-GECO (561 nm). Fluorescence was detected with a backilluminated DU-897 electron multiplying charge-coupled device camera (Andor Technology, Belfast, Northern Ireland) controlled by MetaFluor software (Molecular Devices Corp., Downington, PA). Emission wavelengths were selected with a filter wheel (Sutter Instruments) and the following filters: 527/ 27-nm half-bandwidth for GFP and 584-nm long pass for mCherry and R-GECO (Semrock Rochester, NY). For timelapse recordings, images or image pairs were acquired every 1 s. The laser beam was blocked by a shutter (Sutter Instruments) between image captures to minimize exposure to the potentially harmful light.
Measurements of Insulin Secretion-Groups of 20 -24 pseudoislets were placed in a 20-l chamber superfused at 130 l/min with a similar experimental buffer as described above but with 2.6 mM CaCl 2 . After 35 min of superfusion with buffer containing 3 mM glucose, the perifusate was collected in 30-or 60-s fractions during elevation of the glucose concentration to 20 mM and addition of MRS 2179. The collected medium was immediately put on ice, and after appropriate dilution, insulin was measured using a human insulin AlphaLISA immunodetection kit (PerkinElmer Life Sciences) according to the protocol of the manufacturer.
Data Analysis-Fluorescence intensities were logged from individual cells as a function of time using MetaFluor (Molecular Devices) and expressed relative to the initial fluorescence after subtraction of background (F/F 0 ). The response amplitudes and area under the curve were evaluated using Igor Pro software (Wavemetrics Inc.) and are presented as mean Ϯ S.E. Statistical analysis was performed with paired or two-sample equal variance Student's t test as appropriate.
Author Contributions-A. W. and A. T. designed the study and wrote the paper. A. W. and Q. Y. performed the experiments. All authors analyzed the data and approved the final version of the manuscript.