Regulation of Glucokinase by Intracellular Calcium Levels in Pancreatic β Cells*

Glucokinase (GCK) controls the rate of glucose metabolism in pancreatic β cells, and its activity is rate-limiting for insulin secretion. Posttranslational GCK activation can be stimulated through either G protein-coupled receptors or receptor tyrosine kinase signaling pathways, suggesting a common mechanism. Here we show that inhibiting Ca2+ release from the endoplasmic reticulum (ER) decouples GCK activation from receptor stimulation. Furthermore, pharmacological release of ER Ca2+ stimulates activation of a GCK optical biosensor and potentiates glucose metabolism, implicating rises in cytoplasmic Ca2+ as a critical regulatory mechanism. To explore the potential for glucose-stimulated GCK activation, the GCK biosensor was optimized using circularly permuted mCerulean3 proteins. This new sensor sensitively reports activation in response to insulin, glucagon-like peptide 1, and agents that raise cAMP levels. Transient, glucose-stimulated GCK activation was observed in βTC3 and MIN6 cells. An ER-localized channelrhodopsin was used to manipulate the cytoplasmic Ca2+ concentration in cells expressing the optimized FRET-GCK sensor. This permitted quantification of the relationship between cytoplasmic Ca2+ concentrations and GCK activation. Half-maximal activation of the FRET-GCK sensor was estimated to occur at ∼400 nm Ca2+. When expressed in islets, fluctuations in GCK activation were observed in response to glucose, and we estimated that posttranslational activation of GCK enhances glucose metabolism by ∼35%. These results suggest a mechanism for integrative control over GCK activation and, therefore, glucose metabolism and insulin secretion through regulation of cytoplasmic Ca2+ levels.

Vector Preparation-Circularly permuted mCerulean3 (mCer3) (19) variants were created using the four-primer PCR method, with the GGSGG linker sequences encoded by sense (s) primer 5Ј-T GGC AGC GGT GGC ATG GTG AGC AAG G-3Ј and antisense (as) primer 5Ј-CC ACC GCT GCC A CC CTT GTA CAG CTC G-3Ј. These were used in conjunction with the following end primers to generate fragments of the appropriate length: cp49, 5Ј-TTT A CCG GTC CGC ACC ATG ACC ACC GGC AAG-3Ј (s) and 5Ј-CCC AGA TCT GAG TCC GGA GCA GAT GAA CTT CAG GG-3Ј (as); cp157, 5Ј-TTTA CCG GTC CGC ACC ATG CAG AAG AAC GGC-3Ј (s) and 5Ј-CCC AGA TCT GAG TCC GGA CTT GTC GGC GGT-3 (as); cp173, 5Ј-TTT ACC GGT CGC CAC CAT GGA CGG CAG CGT G-3Ј (s) and 5Ј-CCC AGA TCT GAG TCC GGA CTC GAT GTT GCA GTT GC-3Ј (as); cp195, 5Ј-TTT A CCG GTC CGC ACC ATG CTG CCC GAC AA-3Ј (s) and 5Ј-CCC AGA TCT GAG TCC GGA CAG CAC GGG GC-3Ј (as). N-and C-terminal fragments were linked and amplified by PCR prior to insertion into derivative pQE9 and C1 vectors using AgeI and BglII restriction sites (20). Subsequent cloning into FRET-GCK (8) from C1 vectors was accomplished using NheI and BglII restriction sites. The sequences were confirmed by DNA sequencing (Genewiz). Generation of recombinant proteins and spectroscopic characterization were performed as described previously (19). The ER-localized channelrhodopsin (ChR)-mCherry-ER construct was created by fusing the ER retention sequence from pER-ECFP (Clontech) to the C terminus of hChR2(H134R-mCherry (21) using BsrGI and XbaI restriction sites. Adenoviral vectors were constructed by subcloning the FRET-GCK biosensor coding sequence into an existing pAdTrack-CMV vector. Viruses were then generated by the Mid-Atlantic Nutrition Obesity Research Center core facility using the AdEasy system (22).
Fluorescence Microscopy-FRET imaging and NADH/ NADPH (collectively referred to as NAD(P)H) autofluorescence assays were performed using confocal/two-photon microscopy (5) or wide-field microscopy (6) methods as indicated and as described previously. Imaging devices, conditions, and the imaging buffer have been described in detail elsewhere (5,6). Specimens for Fluo-4 imaging were prepared by labeling cells with acetoxymethyl ester conjugate (Life Technologies, 5 M, 30 min, 37°C) according to the recommendations of the manufacturer. Data were collected in the wide field under 470-nm LED illumination and a high-efficiency enhanced GFP filter cube (Zeiss) for collection.
Quantitative fura-2/FRET imaging was performed essentially as described previously (23) but incorporated 360-and 380-nm LED illumination for collecting fura-2 fluorescence. Ca 2ϩ concentrations were obtained using the fura-2 Ca 2ϩ imaging calibration kit (Invitrogen) according to the kit instructions. For experiments utilizing ChR-mCherry-ER, FRET biosensor illumination was changed to 400 nm, whereas 455-nm pulses were used to activate ChR. ChR-mCherry-ER fluorescence was captured using 530-nm LED illumination with the Zeiss dsRed (no. 43) filter set prior to experimentation. ChR was activated by manual triggering of 455-nm LED fluorescence for ϳ1-s intervals as indicated, except for collection of quantitative data for curve fitting. For those experiments, the Zeiss AxioVison Smart Experiments function was used to precisely activate ChR using three trains of 15 1-s pulses. FRET ratios from maximum cytoplasmic [Ca 2ϩ ] obtained before and after pulsing were normalized to FRET ratios obtained from mutant GCK proteins resistant to S-nitrosylation (5,8). Crosstalk fluorescence was subtracted from FRET signals as described previously (23). The minimum FRET-GCK ratio was approximated from using a sensor containing the C371S mutation (FRET ratio, 0.27 Ϯ 0.009 (S.E.), n ϭ 10 independent replicates). The maximum FRET ratio was estimated by treatment of cells expressing the FRET-GCK with 100 M diethylamine NONOate (Calbiochem) (FRET ratio, 0.42 Ϯ 0.015, n ϭ 15 independent replicates). Curve fitting and statistical analysis was performed using GraphPad Prism software. Image analysis was performed using Zeiss Axiovision software or ImageJ. The non-ratio images in Figs. 3F, 3G, and 4B were pseudocolored cyan and yellow using Adobe Photoshop. Ratio images were generated from the original images by using ImageJ software for smoothing, image ratioing, masking the background, and applying the pseudo-color look-up table. ImageJ was also used to generate the kymograph in Fig. 4B.

ER Ca 2ϩ Couples Receptor
Signaling to GCK Activation-To study dynamic GCK regulation, we utilized two methods we established previously as useful assays for studying GCK activation by NO in living cells. First, we developed a FRET-GCK biosensor on the basis of the GCK structure that can sensitively report activation by S-nitrosylation (8,24). Second, we can quantify glucose-stimulated changes in NAD(P)H autofluorescence (5-7). Because GCK is rate-limiting for glucose metabolism (25), changes in GCK activity are reflected by glucose-dependent rises in NAD(P)H (26). Therefore, stimulation of GCK S-nitrosylation results in corresponding potentiation of glucose-stimulated rises in NAD(P)H autofluorescence (5). These previously established assays were used to probe the relationship between changes in cytoplasmic Ca 2ϩ and GCK activation.
To test whether insulin-stimulated FRET-GCK biosensor activation requires Ca 2ϩ release from the ER, we used inhibitors of the two primary ER Ca 2ϩ channels: 2-aminoethoxydiphenyl borate (2-APB) (27) to block inositol 1,4,5-trisphosphate receptors (IP 3 Rs) and inhibitory concentrations of ryanodine (28,29) to block Ca 2ϩ release from ryanodine receptors (Fig. 1A). Both of these agents inhibited activation of FRET-GCK by insulin. Furthermore, stimulating concentrations of ryanodine (2.5 nM) also activated the FRET-GCK sensor, indicating that inducing ER-Ca 2ϩ release is sufficient for activating GCK. Importantly, these inhibitors did not signifi-cantly affect the basal FRET ratio (Fig. 1B). Similarly, potentiation of glucose metabolism by insulin, as measured by increased NAD(P)H autofluorescence, was inhibited by application of ER Ca 2ϩ channel blockers (Fig. 1C), supporting a role for ER Ca 2ϩ in activating GCK. Consistent with our FRET-GCK experiment, the activating dosage of ryanodine (2.5 nM) also potentiated glucose metabolism (Fig. 1C). The effects of the treatments on cytoplasmic Ca 2ϩ were confirmed in cells labeled with Fluo-4 (Fig. 1D). Co-stimulation with insulin and ER Ca 2ϩ channel blockers significantly decreased cytoplasmic Ca 2ϩ levels through mechanisms that are not entirely clear but may be related to direct association of phosphorylated IRS proteins with sarcoendoplasmic reticulum calcium transport ATPase Ca 2ϩ pumps on the ER (30).
We next tested whether ER Ca 2ϩ release is a general mechanism for GCK activation by testing whether GLP-1, a G proteincoupled receptor agonist, also requires ER Ca 2ϩ to activate GCK. Inhibition of ER Ca 2ϩ release blocked activation of the FRET-GCK biosensor ( Fig. 2A), potentiation of glucose metabolism (Fig. 2B), and GLP-1-stimulated increases in cytoplasmic Ca 2ϩ (Fig. 2C). These findings are consistent with the required mobilization of the ER Ca 2ϩ pool for GCK activation. Therefore, release of ER Ca 2ϩ appears to be an important regulator of GCK activity and glucose metabolism in ␤ cells.
Glucose Stimulates GCK Activation-Our observations raise the important question of whether glucose itself can regulate GCK through the S-nitrosylation mechanism. Although receptor signaling is known to enhance Ca 2ϩ -induced Ca 2ϩ release from intracellular stores (31)(32)(33)(34)(35), Ca 2ϩ -induced Ca 2ϩ release can be observed in the absence of receptor-mediated signaling (36,37), albeit to a lesser extent. Although we did not find conclusive evidence supporting glucose-specific regulation of GCK in our initial studies (8), our earliest-generation FRET-GCK biosensors were hampered by incorporating fluorescent proteins with low brightness and unstable fluorescence (19) and may have missed smaller glucose effects because of the poor signal-to-noise ratio of the biosensor.
An optimized FRET-GCK sensor with improved brightness and sensitivity was developed by incorporating circularly permuted cyan fluorescent proteins (CFPs) derived from mCer3 (19). These new CFPs were created by linking the natural N and C termini of mCer3 by a GGSGG linker. New N termini were created at amino acid positions 49, 157, and 195 (Fig. 3A). Each new mCer3 variant was substituted into the FRET donor position in the FRET-GCK sensor and expressed in ␤TC3 cells. Cells were stimulated with GLP-1 to activate the sensor, and the increase in FRET ratio, or dynamic range, was normalized to the standard deviation of FRET ratios in unstimulated cells (Fig.  3B). The optimized FRET-GCK sensor containing cp173-mCer3 provided contrast over 10 S.D. and was the FRET-GCK sensor used for the rest of this study. The variant incorporating cp157-mCer3 was also bright and outperformed the previousgeneration sensor, leading us to further characterize the fluorescence properties of this protein in addition to cp173-mCer3. The excitation and emission spectra of cp157-mCer3 (⑀, 25000 M Ϫ1 cm Ϫ1 ; quantum yield, 0.77) and cp173-mCer3 (⑀, 25000 M Ϫ1 cm Ϫ1 ; quantum yield, 0.86) were comparable with the parent mCer3 protein (Fig. 3, C-E). Furthermore, we examined the subcellular distribution of the optimized FRET-GCK sensor containing cp173-mCer3. Stimulation with GLP-1 (Fig. 3F) or 2.5 nM ryanodine (Fig. 3G) resulted in similar activation patterns as those observed using earlier versions of the FRET-GCK sensor (8,24) and is consistent with activation on secretory granules and translocation to the cytoplasm (2,7).
To further explore the conditions underlying the molecular regulation of the FRET-GCK sensor, we tested the effect of glucose fasting conditions (8) 2؉ . A, ␤TC3 cells expressing the FRET-GCK sensor were glucose-starved for 3 h prior to treatment as indicated. Cells were observed by fluorescence microscopy, and FRET ratios were normalized to baseline conditions as described previously (6). Treatment was with 100 nM insulin (Ins, 5 min) or a stimulatory dose of ryanodine (Ry, 2.5 nM). Inhibitors of IP 3 Rs (100 M 2-APB) and ryandonine receptors (100 M ryanodine) were applied prior to stimulation. Error bars indicate mean Ϯ S.E. (n Ն 5 independent replicates). **, p Ͻ 0.01; *, p Ͻ 0.05; ns, not significant as determined by ANOVA. B, FRET ratios from ␤TC3 cells expressing the FRET-GCK sensors and left untreated or treated with 100 M 2-APB or 100 M ryanodine. The ratios were normalized to the untreated group. ANOVA was used to determine statistical significance. ns, p Ͼ 0.05 compared with the untreated group, Tukey's multiple comparison test, n ϭ 5 independent replicates; error bars indicate mean Ϯ S.E. C, -fold increase in NAD(P)H autofluorescence was measured by two-photon microscopy of cultured ␤TC3 cells as described previously (5). Treatments were as in A, with the inclusion of 5 mM glucose (Glc) with the stimulatory dose. Results were normalized to pretreatment NAD(P)H fluorescence (n Ն 15 independent replicates). ***, p Ͻ 0.001 by ANOVA; ns, not significant as determined by ANOVA. D, effects of the selected treatments on intracellular Ca 2ϩ in cells labeled with Fluo-4 (n ϭ 5 independent replicates). *, p Ͻ 0.05; **, p Ͻ 0.01 by ANOVA compared with the insulin-treated group; error bars indicate mean Ϯ S.E.). Change in fluorescence was normalized to baseline (⌬F/F 0 ).
GLP-1 and insulin (Fig. 3H). Our previous studies have noted that a brief glucose starvation period does not adversely affect ␤TC3 cells (6), but it is unclear whether stimulation of FRET-GCK could occur in cells exposed to low amounts of glucose that do not elicit secretion. Changing the glucose fasting conditions to 2 mM did not significantly affect the ability of GLP-1 or insulin to stimulate GCK activation (Fig. 3H). Further analysis by two-factor ANOVA (p Ͼ 0.05) did not show significant interaction between basal glucose conditions and stimulation. Prestimulated FRET ratios for low-glucose and glucose-free conditions were not significantly different (p Ͼ 0.05, ANOVA, Tukey multiple comparison test). We also found that increasing cAMP levels with forskolin or isobutylmethylxanthine activated the FRET-GCK sensor (Fig. 3I), consistent with previous work tying cAMP levels to GLP-1 activation of GCK (38). The ability of these treatments to affect cAMP levels in ␤TC3 cells was confirmed in cells expressing the ICUE3 cAMP FRET indicator (39) (Fig. 3J). Fluctuations in FRET-GCK activation patterns (Fig. 4, A and  B) were observed in the presence of 5 mM glucose. Representative traces are shown in Fig. 4A (red and blue), and the FRET ratio increases over time compared with cells left unstimulated (black). Coincident changes to FRET-GCK activation and Ca 2ϩ were observed in response to glucose (Fig. 4C) in cells labeled with fura-2 to track Ca 2ϩ changes (23). Even so, the FRET-GCK ratio decreased rapidly, likely from diffusion out of the region of interest. This is supported by our previous work on the compartmentalized nature of GCK regulation and by the low magnitude of the FRET ratio change. By 3 min, glucose-activated FRET decreases to ϳ50% of the levels observed for stimulation through cell surface receptors or manipulators of cAMP (Fig. 3,  H and I).
The relatively weak stimulation of GCK by glucose may be related to the ␤TC3 cell model, which shows key differences from islets with respect to their glucose-stimulated Ca 2ϩ behaviors. Notably, they lack the ability to oscillate and are known for their highly variable response to glucose (33). MIN6 cells, however, are known to exhibit oscillatory Ca 2ϩ behaviors (17, 40 -42) similar to those reported in islets. To investigate potential regulation of GCK by Ca 2ϩ oscillations, MIN6 cells expressing the optimized FRET-GCK sensor were labeled with fura-2. Under low-glucose conditions (2 mM, Fig. 5), Ca 2ϩ remained low, and no changes in GCK biosensor FRET were observed. Increasing glucose levels to 20 mM in the presence of tetraethylammonium (41,43) induced Ca 2ϩ oscillations (Fig.  5A). Strong activation of FRET-GCK was observed only in response to the initial Ca 2ϩ influx. The response persisted for several minutes with a slow deactivation. The response is likely maximal because the sensor did not respond to later Ca 2ϩ oscillations. Physiologic glucose levels (7.5 mM) also produced similar results, although with dissimilar patterns of Ca 2ϩ oscillations (Fig. 5, B and C).
Optical Release of ER Ca 2ϩ Can Activate GCK-To quantify the relationship between cytoplasmic Ca 2ϩ concentration and GCK activation more systematically, we developed an approach to optically stimulate release of Ca 2ϩ from the ER. However, resting Ca 2ϩ levels at 2 mM glucose in ␤TC3 cells and MIN6 cells (Fig. 6A) proved to be highly variable and poorly suited to such an approach. A third cell line, INS1E cells, displayed consistently low levels of intracellular Ca 2ϩ , making it most suitable for quantifying the relationship between intracellular Ca 2ϩ levels and GCK activation using an optically gated cation channel. An mCherry-tagged channelrhodopsin was targeted to the ER (ChR-mCherry-ER) (Fig. 6B). ChRs are non-selective cation channels that can be activated optically (44,45). Insertion of an ER retention sequence into the C terminus of the mCherry epitope tag constrained its localization to the ER, as evidenced by colocalization with a luminal ER Ca 2ϩ sensor (Fig. 6B). Using optical pulses, we were able to recreate similar changes in cytoplasmic Ca 2ϩ in cultured ␤ cells, as observed naturally (Fig. 6C). Furthermore, when co-expressed with an ER-localized FRET Ca 2ϩ sensor (D1) (46), optical stimulation of ChR-mCherry-ER resulted in depletion of Ca 2ϩ from the ER, as indicated by a change in the FRET ratio (Fig. 6D). Prolonged elevation of the cytoplasmic Ca 2ϩ concentration was achieved by manipulation of pulsing conditions (Fig. 6E). This permitted quantification of the relationship between the cytoplasmic Ca 2ϩ concentration and GCK biosensor activation (Fig. 6F). Half maximal activation of the biosensor was observed at ϳ400 nM Ca 2ϩ , with a steep Hill coefficient in the 4 -6 range (95% confidence intervals). These values   agree well with the measured half-maximal activity of NOS in vitro (ϳ300 nM) (47) and the proposed stoichiometry of Ca 2ϩ -calmodulin-mediated activation of dimeric NOS (12,48).
To explore whether glucose can activate GCK in a more physiologic context, the optimized FRET-GCK biosensor was expressed in primary mouse islets. Fluctuations in sensor activation were observed under continuous glucose stimulation (Fig. 7A) or upon transition from 2 to 7.5 mM glucose stimulation (Fig. 7B). Fluctuations in GCK activation were observed in parallel with fluctuations in NAD(P)H fluorescence (Fig. 7B). Furthermore, glucose failed to activate two mutant sensors that are not S-nitrosylated (Fig. 7C, white). Glucose-stimulated changes in NAD(P)H were also reduced in the cells expressing theses mutant sensors (Fig. 7C, white) even though these mutations do not substantially alter glucose phosphorylation kinetics in vitro (2,5). These results suggest that the glucose-stimulated GCK S-nitrosylation potentiates metabolism by an additional ϳ35%.

Discussion
The Optimized FRET-GCK Sensor and ChR-ER Are Enabling Technical Advances-Here we developed two novel reagents that enable a more sensitive study of the regulation mecha-    FEBRUARY 5, 2016 • VOLUME 291 • NUMBER 6

Intracellular Ca 2؉ Controls GCK Activation
nisms that control ␤ cell glucose metabolism. Creation of the circularly permuted mCer3 variants provided an efficient method for sensor optimization. Simple swapping of mCerulean for the brighter mCer3 did improve the brightness of the probe but not its performance in this case, as defined by improvement of the ratio of the dynamic range to measurement precision. This is consistent with our experience with upgrading the first generation ECFP to Cerulean, where the dynamic range between activated and deactivated FRET conformations was decreased slightly (24). Notably, circular permutation of mCer3 was well tolerated, and the optimized FRET-GCK sensor containing cp173-mCer3 displays greatly improved contrast between S-nitrosylated and denitrosylated conformations while retaining the advances in measurement precision associated with mCer3 development (19). To our knowledge, this is the first use of an optically gated, non-selective cation channel to stimulate release of Ca 2ϩ from the ER. ChR-mCherry-ER was useful for systematic exploration of the relationship between GCK activation and ER Ca 2ϩ release. Unlike permeabilized cell approaches, the integrity of the cytoplasmic environment remains intact. Here we show that periodic Ca 2ϩ oscillations can be recreated in culture. Therefore, we expect that this approach will be useful for exploring the function of Ca 2ϩ oscillations not only for ␤ cell physiology but for other systems where oscillatory activity is known to occur (49). Even so, our approach has special utility for studying ␤ cells and should permit close examination of the spatial aspects of GCK regulation (50,51) and enable a more direct exploration of the connection between metabolic and Ca 2ϩ oscillations in living ␤ cells. Finally, the optical approach for modulating ER Ca 2ϩ offers advantages over pharmacological agents, like 2-APB, that have off-target effects and complex pharmacology (52).
Cytoplasmic Ca 2ϩ and GCK Activation-Our data suggest a tight connection between a rise in the cytoplasmic Ca 2ϩ concentration and GCK activation. The initial Ca 2ϩ influx following exposure to stimulatory concentrations of glucose seems to be most important and, in islets and MIN6 cells, leads to a prolonged activation of GCK spanning several minutes. Even in ␤TC3 cells, low levels of activated GCK tended to rise over time, fitting well with what is known about the mechanism of GCK activation. Association with NOS on secretory granules is essential for GCK activation (7) and results in dissociation of the complex and release of GCK into the cytoplasm. The mechanism behind denitrosylation of GCK is unknown but does not appear to be rapid enough to allow fast cycling of GCK activation. Rather, the model we favor is a burst of GCK activation following exposure to post-prandial conditions (i.e. high glucose and hormones such as GLP-1), perhaps contributing to the first phase of insulin secretion or the highly dynamic first The asterisks indicate pulses of 455-nm LED light (1 s). D, INS1E cells co-expressing ChR-mCherry-ER and the D1 FRET-based Ca 2ϩ sensor were exposed to activated light pulses (indicated by bars). Depletion of the ER-Ca 2ϩ is indicated by a decrease in the FRET ratio. Data were smoothed for presentation clarity (GraphPad Prism, second-order smoothing). E, closely spaced pulses were used to raise the cytoplasmic Ca 2ϩ concentration of an INS1E cell expressing ChR-mCherry-ER. F, fura-2 was used to quantify the cytoplasmic Ca 2ϩ concentration in cells expressing ChR-mCherry-ER and the FRET-GCK biosensor. Optical pulses were used to raise cytoplasmic Ca 2ϩ concentration, and FRET-GCK activation was normalized as described under "Experimental Procedures." Data were fit using a log(agonist) versus normalized response model. responder type of secretion observed in vivo (53), although that remains to be proven.
Also unclear is the relationship between glucose and hormonal activation of GCK. Strong stimuli, such as those used to induce GCK oscillations in MIN6 cells, seem to saturate GCK activation, whereas stimulation of islets with physiologic levels of glucose does seem to allow the possibility of low-frequency GCK oscillations. The actual physiology underlying hormone stimulation of islet function is not well understood because it is difficult to distinguish between systemic and direct actions of hormones such as insulin (54), GLP-1 (55), and kisspeptin (56,57). Synergistic GCK activation between glucose and hormonal signaling can occur at multiple levels, especially those influencing cAMP dynamics (58), as suggested by our finding that increasing cAMP levels with forskolin or isobutylmethylxanthine can activate GCK. A detailed investigation will be required to understand the nuances of GCK regulation during co-stimulation.
With regards to the dependence of GCK on ER Ca 2ϩ versus Ca 2ϩ influx, our pharmacological studies strongly support a role for ER Ca 2ϩ , but, as noted above, interpretation of these studies can be difficult because such agents tend to be less specific than desired. Even so, it is questionable whether extracellular Ca 2ϩ influx during an action potential would raise cytoplasmic Ca 2ϩ levels high enough to activate GCK. Previous measurements have found that extracellular Ca 2ϩ is buffered quickly and raises Ca 2ϩ levels only within a short distance from the plasma membrane (59). Furthermore, we have shown that extracellular Ca 2ϩ influx is not essential for activating GCK because it can be observed in the presence of voltage-gated Ca 2ϩ channel blockers (8). Taken together with our finding that activation of our GCK sensor by ER-ChR requires raising cytoplasmic Ca 2ϩ to ϳ300 -400 nM, GCK activation almost certainly requires mobilization of the ER Ca 2ϩ pool. The proposed ER-Ca 2ϩ mechanism also makes sense from the vantage point of spatial regulation of GCK activation, given that release of ER Ca 2ϩ is more likely to reach NOS-GCK complexes on secretory granules than extracellular Ca 2ϩ because the vast majority of insulin granules are located in the cell interior away from the plasma membrane.
Dynamic Ca 2ϩ Oscillations and GCK Activation-Both cytoplasmic Ca 2ϩ and metabolism (60) are known to exhibit oscillatory behaviors, and connections between the two have been observed (61). Suggested mechanisms tying metabolic oscillations to Ca 2ϩ oscillations most frequently invoke PFK2, although this is not without controversy (62). The dual oscillator theory was derived from computational muscle models and emphasizes PFK2 activity. Importantly, these models assume a permissive role for GCK and static activity (63). Although this assumption should hold true for tissues, such as skeletal muscle, that phosphorylate glucose using hexokinase, ␤ cell metabolism has a much different relationship with glucose because of its dependence on GCK. As Fridlyland and Phillipson (64) point out, ␤ cells are notable for the "near-dominant" control of glycolysis by GCK activity . Indeed, even in the dual oscillator model, a precise amount of GCK activity is needed for the simulations to work because oscillations disappear with too little or too much GCK activity (63). Future work is required to fully unravel the influence of dynamic GCK activity on metabolic and Ca 2ϩ oscillations.
Mechanism of GCK Control by ER Ca 2ϩ Channels-Our results also show sensitivity of GCK regulation to both IP 3 R and ryanodine receptor regulation. This was not unexpected because both receptor systems couple strongly to the generation of IP 3 (15,16). Even so, it is unclear how generalizable our findings are with regard to the specific receptors involved in ER Ca 2ϩ release. Some studies clearly favor a role for IP 3 Rs over ryanodine receptors in receptor-mediated potentiation of [glucose] mM FIGURE 7. Glucose-stimulated regulation of GCK in islets. A, the optimized FRET-GCK sensor was expressed in an isolated mouse islet using adenoviral vectors. Dynamic changes in the FRET ratio were observed by fluorescence microscopy in an islet stimulated with 7.5 mM glucose. B, FRET-GCK activation was tracked along with changes in NAD(P)H fluorescence in an islet cell stimulated with 7.5 mM glucose (total). The change from 2 to 7.5 mM glucose is indicated by the dashed line. Normalization was to resting (2 mM glucose) values. C, islet cells expressing WT sensors or those with mutations that block GCK S-nitrosylation (V367M and C371S). FRET ratios and NAD(P)H autofluorescence were measured by fluorescence microscopy before (2 mM glucose) and after addition of glucose (7.5 mM total). Error bars indicate mean Ϯ S.E. n Ն 9 independent replicates; **, p Ͻ 0.01; ***, p Ͻ 0.001; ANOVA. secretion (65), although expression of ryanodine receptors in islets is indeed prevalent, and their high conductance for Ca 2ϩ suggests a prominent role in ␤ cell Ca 2ϩ -induced Ca 2ϩ release (66). Additional mechanisms for ER Ca 2ϩ release have also been described, such as through nicotinic acid adenine dinucleotide phosphate (67). Even so, it is likely that individual mechanisms will prove to be synergistic, given the relationship between cytoplasmic Ca 2ϩ concentration and GCK activation.
We favor an integrative model where islets respond to a variety of factors in the extracellular milieu, including glucose, gapjunctional communication, and hormonal and, likely, neuronal inputs to produce a coordinated and integral secretory output. Naturally, further investigation will be required to quantitatively assess the control strength of the various possible regulatory mechanisms over GCK activation. Our optimized FRET-GCK sensor is perhaps a useful tool for such an investigation.