Insulin Secretion and Ca2+ Dynamics in β-Cells Are Regulated by PERK (EIF2AK3) in Concert with Calcineurin*

Background: A genetic deficiency in protein kinase R (PKR)-like endoplasmic reticulum kinase (PERK) in human and mice results in insulin-dependent diabetes. Results: Acute inhibition of PERK impairs insulin secretion, secretagogue-stimulated Ca2+ influx, and sarcoplasmic endoplasmic reticulum Ca2+-ATPase activity through a calcineurin-dependent pathway. Conclusion: PERK and calcineurin regulates Ca2+ dynamics underlying insulin secretion. Significance: Our findings provide insights into the intracellular mechanisms underlying stimulated insulin secretion. Protein kinase R (PKR)-like endoplasmic reticulum kinase (PERK) (EIF2AK3) is essential for normal development and function of the insulin-secreting β-cell. Although genetic ablation of PERK in β-cells results in permanent neonatal diabetes in humans and mice, the underlying mechanisms remain unclear. Here, we used a newly developed and highly specific inhibitor of PERK to determine the immediate effects of acute ablation of PERK activity. We found that inhibition of PERK in human and rodent β-cells causes a rapid inhibition of secretagogue-stimulated subcellular Ca2+ signaling and insulin secretion. These dysfunctions stem from alterations in store-operated Ca2+ entry and sarcoplasmic endoplasmic reticulum Ca2+-ATPase activity. We also found that PERK regulates calcineurin, and pharmacological inhibition of calcineurin results in similar defects on stimulus-secretion coupling. Our findings suggest that interplay between calcineurin and PERK regulates β-cell Ca2+ signaling and insulin secretion, and that loss of this interaction may have profound implications in insulin secretion defects associated with diabetes.

Insulin secretion from the endocrine pancreatic ␤-cells is driven by a rapid influx of Ca 2ϩ from the extracellular space or from internal stores into the cytoplasm, which stimulates exocytosis of the insulin granules (1,2). The uptake of Ca 2ϩ is initially stimulated by nutrient secretagogues such as glucose or non-nutrient secretagogues such as acetylcholine. Glucosestimulated insulin secretion (GSIS) 2 occurs through a well characterized pathway whereby glucose is taken up and metab-olized to generate ATP, which in turn inhibits the ATP-sensitive potassium channel (K ATP ) resulting in depolarization of the plasma membrane and activation of the voltage-dependent Ca 2ϩ channel (VDCC). In addition to VDCC-mediated Ca 2ϩ influx, one or more of the transient receptor potential (TRP) channels may contribute to the rapid rise in Ca 2ϩ (3). The endoplasmic reticulum (ER) has a large Ca 2ϩ storage capacity and acts to both buffer the cytoplasm Ca 2ϩ and to release Ca 2ϩ in response to non-nutrient secretagogues. Sarcoplasmic endoplasmic reticulum Ca 2ϩ -ATPase (SERCA) plays an important role in regulating ER Ca 2ϩ by pumping Ca 2ϩ into the ER and is required to maintain high ER Ca 2ϩ levels. In the case of glucosestimulated insulin secretion the ER acts as an important buffer for the massive influx of Ca 2ϩ into the cytoplasm (4) and can extend stimulated secretion by releasing Ca 2ϩ into the cytoplasm by a Ca 2ϩ -induced Ca 2ϩ release mechanism. Depletion of ER Ca 2ϩ induces the store-operated Ca 2ϩ channel (SOCC) at the plasma membrane, which not only results in ER Ca 2ϩ refilling but also increases cytoplasmic Ca 2ϩ stimulating insulin secretion. Although the role of ER Ca 2ϩ stores in regulating insulin secretion is still incompletely understood, its importance is underscored by the observation that dysfunctions in or inhibition of SERCA residing in the ER membrane result in ablation of stimulated insulin secretion (5)(6)(7)(8).
Besides glucose, non-fuel secretagogues such as acetylcholine or carbachol can also drive insulin exocytosis (9). Acetylcholine, released by intrapancreatic nerve endings during the preabsorptive and absorptive phases of feeding, causes the release of internal Ca 2ϩ stores largely located in the ER by activating phospholipase C. Activation of phospholipase C results in the generation of inositol triphosphate (IP 3 ), which causes the release of ER Ca 2ϩ into the cytosol via the IP 3 receptor channels (9,10). ER Ca 2ϩ release further triggers store-operated Ca 2ϩ entry (SOCE) to maintain a sustained elevation of cytosolic Ca 2ϩ level and insulin secretion (10,11). PERK (EIF2AK3), an eIF2␣ kinase in the ER membrane, is essential for normal development and function of the insulinsecreting ␤-cell (12)(13)(14)(15). Perk loss of function mutations in humans and mice result in insulin-dependent permanent neo-natal diabetes due to insufficient insulin secretion from the pancreas (12,14). PERK has also been shown to play a key role in regulating the ER stress and the unfolded protein response in cultured cells that are subjected to severe stress conditions (16,17). However, the relevance of the ER stress response pathway to the normal developmental and physiological functions of PERK in ␤-cells has been questioned and remains controversial (18,19). Previous attempts to identify the primary functions of PERK were confounded by the myriad dysfunctions within ␤-cells including ablated insulin synthesis and secretion, delayed development and proliferation of the ␤-cells, and a massive accumulation of proinsulin in the ER (14,19,20) as well as dysfunctions in other organs and tissues (13,14,21). Recently a highly selective PERK inhibitor (denoted throughout as PERKi in text and PI in figure legends) was developed by GlaxoSmithKline, Inc. (22). When applied to animal models, it recapitulated the major pancreatic defects seen in Perk-deficient mice and humans including ␤-cell dysfunction and atrophy of the exocrine pancreas (22,23). The GlaxoSmithKline (GSK) PERK inhibitor provided us the means to acutely inhibit PERK activity and assess the immediate impact on insulin secretion and intracellular Ca 2ϩ dynamics in the pancreatic ␤-cells prior to the onset of severe cellular dysmorphies.
Calcineurin (CN) (14), a Ca 2ϩ -dependent phosphatase, plays similar roles to PERK in regulating insulin secretion, ␤-cell proliferation, and glucose homeostasis (24,25) suggesting that PERK and CN may be acting through related pathways. Supporting this hypothesis, Bollo and co-workers (26) discovered that CN and PERK interact and modulate each others activity as a function of the concentration of cytoplasmic Ca 2ϩ . Moreover, CN has been shown to dephosphorylate calnexin and thereby relieve repression of SERCA activity and ER Ca 2ϩ uptake.
We show herein that acute inhibition of PERK or CN in ␤-cells rapidly suppresses glucose-stimulated insulin secretion. Unexpectedly we discovered that ablation of their activities also strongly abrogates glucose-stimulated Ca 2ϩ uptake into the cytoplasm and restoration of Ca 2ϩ to the ER following stimulated release. We speculate that the major function of PERK in the pancreatic ␤-cell is to coordinate Ca 2ϩ dynamics between the ER and the cytoplasm during stimulus-coupled secretion of insulin, and that this regulation is mediated through CN.

EXPERIMENTAL PROCEDURES
Reagents-GSK2606414 PERK inhibitor was a kind gift from Jeffrey Axten and Rakish Kumar, GlaxoSmithKline, Collegeville, PA. PERKi was prepared as a 10 mM stock solution in dimethyl sulfoxide and diluted immediately before use. Chlorogenic acid (Sigma), cypermethrin (Sigma), and ionomycin (Calbiochem) were dissolved in a ϫ10,000 dimethyl sulfoxide working concentration and diluted immediately before use.
Cell Culture-INS1 832/13 (obtained from Dr. Christopher Newgard, Duke University) and MIN6 cells (provided by Dr. Jun-Ichi Miyazaki, Osaka University, Japan) were cultured as previously described (27). INS1 832/13 cells containing a shorthairpin RNA directed against the rat Perk mRNA (shPerk) were obtained from Dr. Fumihiko Urano (University of Massachusetts). The shPerk is stably integrated into the genome of INS1 832/13 ␤-cell lines and under the inducible regulation of doxy-cycline. The INS1 832/13 shPerk cells were cultured in a tetracycline-free environment to avoid leaky expression of shPerk. Full details of treatment to various cell lines were described in figure legends.
Islet Isolation and Primary ␤-Cell Culture-Human pancreatic islets were obtained through the Integrated Islet Distribution Program and first allowed overnight recovery in fresh RPMI1640 medium with 10% fetal bovine serum, 1% antibiotic antimycotic solution (Sigma), and 5.5 mM glucose at 5% CO 2 , 95% air. Rat islets were isolated from 2-3-month-old Sprague-Dawley rats (purchased from Charles River) using a modified Histopaque-1077 separation method (28) and cultured in the same way as human islets.
For primary ␤ cell culture, islets were disassociated by trypsin (0.125% in PBS; 4 min at 37°C) to release single cells and evenly plated on coverslips for 48 h before experiments. For Ca 2ϩ imaging experiments utilizing cells isolated from islets we did not specifically identify ␤ cells, which comprise ϳ75% of the islet. However, for all cytosolic Ca 2ϩ measurements we selected cells that exhibited a positive response to glucose, which selected against ␣-glucagon cells that comprise ϳ20% of the islet and do not respond positively to high glucose stimulation (29).
Insulin Secretion-Insulin concentrations were determined by immunoassay (Meso Scale Discovery) and were normalized to total protein concentration. For studies of insulin secretion, isolated islets or cultured ␤-cell line were first cultured overnight at 37°C (5% CO 2 ) in RPMI1640 medium containing 10% fetal bovine serum and 5.5 mM glucose. Samples were then incubated at 37°C in KRB-HEPES buffer (pH 7.4) with 1% bovine serum albumin for assigned pretreatment and insulin stimulation as described in figure legends. At the end of the 30-min stimulation, the supernatant was assayed for secreted insulin (by Meso Scale Discovery), and cells/islets were sonicated in acid ethanol and assayed for total insulin (by Meso Scale Discovery) and total protein (by Bio-Rad Protein Assay).
Immunocytochemistry-Cultured ␤-cells were subjected to fixation and permeabilization with 4% formaldehyde and 0.1% Triton X-100. The cells were then denatured with 1 N HCl for 20 min, followed by 5% horse serum (Invitrogen) in PBS for 1 h. The following primary antibodies were applied overnight at 4°C: insulin (1:500, Abcam), proinsulin (1:200, Beta Cell Biology Consortium and Hytest), and NFATc1 (1:200, Thermo Scientific). Appropriate secondary antibodies conjugated with Alexa Fluor 350, 488, or 555 dye (Molecular Probes) were used (1:400 dilution) to visualize the labeled cells. Anti-fade reagent with DAPI (Invitrogen) was used to mount slides and label nucleic region. Fluorescence images were captured with a Nikon Eclipse E1000 and Image-Pro Plus (Phase 3 Imaging Systems, GE Healthcare, Inc.).
For NFATc1 translocation measurement, all procedures for immunocytochemistry and image collection were done at the same time under identical conditions to allow direct comparison between treatments. For data analysis, NIH Image J software was used. The nucleic and cytosolic areas were traced based on DAPI and the insulin signal, and after background subtraction for each area, NFATc1 translocation was quanti-fied for each cell by calculating ratio of pixel density of NFATc1 in the nucleic area to the density in cytoplasm.
Voltage-dependent Ca 2ϩ Current Density Measured by Whole Cell Patch Clamp-Whole cell patch clamp recordings were performed using the Multiclamp 700A patch clamp amplifier (Molecular Devices, Palo Alto, CA). Tetrodotoxin (1 M) and tetraethylammonium (15 mM) were added during recording to block voltage-dependent sodium channels and potassium channels. For Ca 2ϩ current recording experiments, the membrane potential was held at Ϫ70 mV baseline. A series of depolarizing voltage steps with 10-mV increments were delivered at 5-s intervals, to elicit voltage-dependent Ca 2ϩ responses. Data were collected using pClamp 9 software (Molecular Devices, Palo Alto, CA), sampled at 10 kHz and filtered at 1 kHz. Off-line data analyses of Ca 2ϩ currents amplitude were performed using pClamp 9 software. All experiments were performed at room temperature.
PERK autophosphorylation was measured using anti-PERK blot. Phosphorylated PERK band (PERK(P)) and total PERK band (PERK) of each sample were traced and the pixel density was measured for each sample with background subtraction.
Cytosolic Ca 2ϩ Measurement by Fura2 Ca 2ϩ Imaging-The cytosolic Ca 2ϩ level was measured using the ratiometric Ca 2ϩ indicator Fura2-AM following the procedure of Roe and coworkers (30). After dye loading, coverslips (12 mm) were transferred to a perfusion chamber (Warner Instruments Series 20 open bath chamber) mounted on a Nikon TE-2000-S inverted microscope with a ϫ20 objective and a high 340/380 nm transmittance filter for Ca 2ϩ ratio imaging (Chroma Technology). Cells were perfused in KRB-HEPES with a constant flow rate of 1-2 ml/min at 37°C. Details of treatment were described in figure legends. Multiple cells were randomly picked per operation. Ratios of the fluorescent emission signals under excitation at 340 over 380 nm (R) were collected by Simple PCI imaging software (C-Imaging) and further normalized to the average ratios before stimulation (R 0 ).

ER Ca 2ϩ Measurement by Fluorescence Resonance Energy Transfer (FRET)-based Ca 2ϩ
Imaging with ER-targeted Cameleon-ER Ca 2ϩ level was measured using FRET-based D1ER cameleon probe following the protocol of Michael Roe and co-workers (30). Cells were transduced with adenovirus encoding the ER-targeted cameleon D1ER (provided by Dr. Michael Roe, SUNY Upstate Medical University). The medium was replaced with fresh RPMI1640 medium after 2 h of transduction and cultured for an additional 48 h before experiments. Coverslips (25 mm) were placed into a glass coverslip dish (mode MSC-TD, Warner instruments) fit on PDMI-2 microincubator (Harvard Apparatus). Cells were perfused with KRB-HEPES buffer at a constant rate of 3 ml/min at 37°C. Full details of treatments were provided in the figure legends. Images were obtained at 5-s intervals using a Nikon Eclipse E600FN microscope with a ϫ60 1.0 numerical aperture water immersion objective (Nikon) controlled by Metavue software (Universal Imaging Corp.). Filters used for the dual emission ratio (CFP excitation 430/25, CFP emission 470/30; YFP emission 560/80) were placed in filter wheels and combined with the dual dichroic mirror at 505 nm (505dcxr, Chroma). Data were represented as ratios of YFP over CFP signal intensity (R) normalized to the average ratios prior to stimulation (R 0 ). The samples used for the ER Ca 2ϩ experiments on primary human and rat islet cells were from the same batch of islets as those used the for cytosolic Ca 2ϩ experiments. The ␤-cell fraction was estimated based on cytosolic Ca 2ϩ measurement under glucose stimulation.
Statistical Analysis-All numeric data were represented as mean Ϯ S.E. For Ca 2ϩ imaging data, results are presented as averages from Ͼ3 separate experiments. The area under the curve was measured for each biological individual and used to estimate the percentage difference between treatments. Statistical significance was determined using Student's t testing.

Inhibition of PERK Activity Recapitulates ␤-Cell Dysfunctions
Seen in Genetic Ablation of Perk-Previously we showed that loss of function mutations of Perk in mice (PKO) led to an impacted ER phenotype in a substantial fraction of ␤-cells (30 -40%) characterized by accumulation of proinsulin and other client proteins in the ER and failure of anterograde trafficking to the Golgi (19,20,27). This phenotype can be readily detected using immunohistochemical labeling of insulin and proinsulin in mouse islets of Langerhans (Fig. 1A, top). To test whether inhibition of the enzymatic activity of PERK results in the impacted-ER phenotype, we employed the use of newly developed PERK inhibitor GSK2606414, which is a high affinity ligand of the catalytic site that competes with ATP (22,23). Approximately 20% of the INS1 832/13 cells treated 24 h with 1 M PERKi exhibited the same impacted ER phenotype seen in PKO mice (Fig. 1A, lower view). We next determined whether PERKi reduced PERK activity. Depletion of ER Ca 2ϩ stores causes activation of PERK and phosphorylation of its substrate eIF2␣ (31,32). Exposing INS1 832/13 cells 30 min to cyclopiazonic acid (CPA), an inhibitor of SERCA, led to PERK activation and phosphorylation of eIF2␣ (Fig. 1B, lanes 4, 8, and 12). Pre-treatment with PERKi for 20 min abolished CPA-induced PERK activation and eIF2␣ phosphorylation (Fig. 1B, lanes 3, 7,  and 11). Therefore the PERKi can be used as an effective tool for investigating the immediate effects of PERK inhibition on the scale of minutes, which is ϳ20 h before severe cellular dysfunctions are first seen.
Acute Inhibition of PERK Activity Impairs Glucose-dependent Insulin Secretion-Previously we showed that glucose-stimulated insulin secretion was ablated in islets isolated from neonatal PKO mice (19). In the present study, this result was confirmed by genetic knockdown of Perk in INS1 832/13 ␤-cells bearing a tetracycline-operated shPerk transgene (denoted as INS1 832/13 shPerk cells). After 24 h administration of 2 g/ml of doxycycline, the Perk mRNA level was reduced to 39.7 Ϯ 3.9% of WT cells (n ϭ 6, p Ͻ 0.001) and GSIS was reduced by 57.6 Ϯ 2.2% (p Ͻ 0.001, Fig. 2A). To determine whether acute inhibition of PERK by PERKi impacts insulin secretion, PERKi was employed for 20 min before 20 mM glucose stimulation. PERK inhibition led to reduction of GSIS in INS1 832/13 cells (Fig. 2B, left panel) and in islets isolated from rats (Fig. 2B, middle panel) and humans (Fig. 2B, right panel) by 34.6 Ϯ 3.8% (p Ͻ 0.01), 27.1 Ϯ 9.0% (p ϭ 0.058), and 35.6 Ϯ 5.4% (p Ͻ 0.01), respectively. In addition, we also measured insulin secretion of INS1 832/13 cells in response to 8 mM glucose, which is a more physiological concentration observed postprandially. PERK inhibition also led to a reduction of GSIS by 22.0 Ϯ 3.6% (p Ͻ 0.001, Fig. 2C). Because an increase in [Ca 2ϩ ] c is a key regulator of GSIS, we measured the effect of PERKi on glucose-stimulated changes in [Ca 2ϩ ] c in rat and human pancreatic islets of Langerhans. Pretreatment with 1 M PERKi for 20 min significantly lowered the glucose-induced rise in [Ca 2ϩ ] c by 61.9 Ϯ 13.3% (p Ͻ 0.01) in rat islets and 55.9 Ϯ 8.4% (p ϭ 0.053) in human islets (Fig. 2D). These results indicate that PERK activity is required for normal glucose-stimulated insulin secretion and suggests this modulation is mediated by mechanisms underlying Ca 2ϩ signaling.
PERK Inhibition Represses Stimulated Ca 2ϩ Influx-Glucose-stimulated Ca 2ϩ signaling is regulated by Ca 2ϩ influx through the plasma membrane and release of Ca 2ϩ from ER Ca 2ϩ stores (33,34 Fig. 3A) and a 30.8 Ϯ 5.0% reduction in insulin secretion (p Ͻ 0.01, Fig. 3B). In addition, PERKi (1 M, 20 min) reduced KCl-stimulated [Ca 2ϩ ] c by 44.1 Ϯ 3.1% in rat (p Ͻ 0.001) and 40.9 Ϯ 9.3% in human (p Ͻ 0.05) primary ␤-cells (Fig. 3, C and D). These findings suggested that PERK-dependent Ca 2ϩ influx contributes to insulin exocytosis. To determine how rapidly PERKi reduced KCl-stimulated Ca 2ϩ uptake, INS1 832/13 cells were exposed to PERKi only 100 s prior to a series of 25 mM KCl pulses. The inhibitory effect of PERKi was first detected only 10 min (p Ͻ 0.001, Fig. 3E) after addition of the inhibitor. To determine whether PERK activity was modulated by changes in [Ca 2ϩ ] c , the autophosphorylation level of PERK was measured in ␤-cells treated with 50 mM KCl. PERK autophosphorylation was induced by 50 mM KCl in the Ca 2ϩ -contained solution but not in the Ca 2ϩ -free solution (Fig. 4A), suggesting that an increase in [Ca 2ϩ ] c induced PERK activation. We also found a significant elevation of PERK activity by exposing cells to 50 mM CaCl 2 (Fig. 4B). Taken together, these findings suggest that PERK activity is both activated by increases in [Ca 2ϩ ] c and regulates Ca 2ϩ influx through the plasma membrane.
PERK Inhibition Does Not Directly Affect VDCC Current-Ca 2ϩ influx through VDCCs plays a dominant role in contributing to the rise in [Ca 2ϩ ] c after stimulation of ␤-cells with high glucose or KCl. To determine whether PERKi negatively regulates the VDCC, VDCC-dependent Ca 2ϩ current density was measured by patch clamp electrophysiology. Unexpectedly, in cells treated with PERKi, VDCC current density was not significantly different from control cells (Fig. 5A) suggesting that the PERK-dependent regulation of Ca 2ϩ uptake involves other Ca 2ϩ or ion channels.
Store-operated Ca 2ϩ Entry Is Impaired by Acute PERK Inhibition-Pancreatic ␤-cells express other plasma membrane Ca 2ϩ channels including store-operated Ca 2ϩ channels and transient receptor potential channels that either conduct Ca 2ϩ or modulate the membrane potential (33). 2-Aminoethoxydiphenyl borate (2-APB) has been shown to inhibit SOCCs and some of the TRP channels (3,35). We found that 2-APB and PERKi exhibit similar effects on KCl-stimulated Ca 2ϩ influx, and the combination of 2-APB and PERKi showed no further reduction in Ca 2ϩ influx (Fig. 5B). Moreover, 2-APB treatment reduced KCl-stimulated insulin secretion by 55.8 Ϯ 2.2% in INS1 832/13 cells (p Ͻ 0.001, Fig. 5C). Taken together, these findings suggested that PERKi and 2-APB inhibit the same Ca 2ϩ signaling mechanism, which plays a significant role in regulating insulin secretion.
To determine whether SOCC activity is affected by acute ablation of PERK, store-operated Ca 2ϩ entry was measured by using 250 M carbachol following the methods of Liu and Gylfe (11). PERK inhibition had no effect on the initial peak in cytosolic Ca 2ϩ (Fig. 5, D and E), which is driven by ER Ca 2ϩ release. However, the second phase, characterized by a gradual decrease and plateau, was reduced by 31.0 Ϯ 7.0% in PERKinhibited cells (p Ͻ 0.05, Fig. 5D). Extracellular Ca 2ϩ influx through SOCC is known to be the source of the second phase (11), which was confirmed by the absence of a second phase when extracellular Ca 2ϩ was absent (Fig. 5E). To determine whether this negative impact of PERK inhibition on SOCC may adversely impact carbachol-stimulated insulin secretion, INS1 832/13 cells were pretreated with PERKi and then stimulated with carbachol. PERK inhibition reduced carbachol-stimulated insulin secretion by 57.4 Ϯ 7.1% (p Ͻ 0.05, Fig. 5F).
An alternate method to measure SOCC activity (36) was utilized to determine the effect of PERKi on SOCE. In these studies, cells were bathed in Ca 2ϩ -free solutions and then ER Ca 2ϩ stores were depleted using the SERCA inhibitor CPA. [Ca 2ϩ ] c was measured after reconstituting extracellular Ca 2ϩ in the presence of a VDCC inhibitor nifedipine. The resultant Ca 2ϩ influx reflected SOCE through SOCCs. As expected, administration of extracellular Ca 2ϩ lead to a rapid increase of [Ca 2ϩ ] c that was attenuated by 21.7 Ϯ 1.9% in INS-1 832/13 cells (p Ͻ 0.05, Fig. 5G) and 43.9 Ϯ 5.0% in primary rat ␤-cells (p ϭ 0.18, Fig. 5H) pretreated with PERKi.
PERK Regulates ER Ca 2ϩ Reuptake through Modulation of SERCA Pump Activity-To determine whether acute loss of PERK activity impacts ER Ca 2ϩ dynamics, ER Ca 2ϩ levels in response to carbachol were measured using the FRET-based probe D1ER cameleon (37). Application of 250 M carbachol in INS1 832/13 cells led to an immediate loss of ER Ca 2ϩ followed by a rapid reuptake (Fig. 6A). Consistent with the absence of an effect of PERK inhibition on the transient increase in [Ca 2ϩ ] c following carbachol administration (Fig. 5, D and E), PERKi did not affect the rate or extent of Ca 2ϩ extrusion from the ER (Fig.  6A). However, ER Ca 2ϩ reuptake was significantly reduced by 41.5 Ϯ 10.2% in PERK-inhibited cells (p Ͻ 0.05, Fig. 6A). More- over, this effect of PERKi was not mediated through 2-APBsensitive channels (Fig. 6B), suggesting that PERK may regulate SERCA-mediated ER Ca 2ϩ reuptake independently of its effects on SOCE. To explore this possibility, we estimated SERCA pump activity using an alternative method (31), which utilizes the reversible SERCA inhibitor CPA. Control cells treated with vehicle showed rapid refilling of ER after washout of CPA, whereas in PERK-inhibited ER Ca 2ϩ restoration was reduced by 76.1 Ϯ 5.2% (p Ͻ 0.01, Fig. 6C, left panel). PERKi also attenuated the ER Ca 2ϩ refilling by 56.2 Ϯ 6.8% in primary rat ␤-cells (p Ͻ 0.05, Fig. 6C, middle panel) and 39.6 Ϯ 6.7% in primary human ␤-cells (p Ͻ 0.01, Fig. 6C,  right panel). Taken together, our data suggest that PERK positively regulates SERCA particularly when ER Ca 2ϩ stores are being replenished.
Interaction of SERCA and Calnexin Is Negatively Regulated by PERK-SERCA pump activity is negatively regulated through interaction with calnexin, an ER chaperone protein   (26,38). To determine whether PERK affects the interaction between SERCA and calnexin, a co-immunoprecipitation experiment with SERCA antibody was performed. Cells treated with PERKi showed a 2.4-fold increase in calnexin coimmunoprecipitated with SERCA (Fig. 6, D and E), suggesting that PERK negatively regulates the interaction of SERCA and calnexin in ␤-cells. This result is consistent with our finding that PERK positively regulates SERCA pump activity.
These findings raised the possibility that the interplay between PERK and CN is an important regulatory component of subcellular Ca 2ϩ signaling dynamics in pancreatic ␤-cells. To directly test whether CN regulates Ca 2ϩ dynamics, INS1 832/13 cells were treated with the CN inhibitor CPM. CN-inhibited cells showed a 59.7 Ϯ 5.7% reduction in cytosolic Ca 2ϩ influx (p Ͻ 0.001, Fig. 7B) and a 65.9 Ϯ 4.7% reduction in insulin secretion (p Ͻ 0.01, Fig. 7C) in response to 25 mM KCl, which was similar to the impact of PERK inhibition. We further determined whether SOCE was affected by CN activity. Cells treated with CPM exhibited a 38.0 Ϯ 4.7% (p Ͻ 0.001) decrease in SOCE (Fig. 7, D and E), whereas activation of CN by CGA restored SOCE in the PERK-inhibited cells to control levels (p Ͻ 0.05, Fig. 7, D and E). As expected, SOCE in CPM-treated samples could not be recovered by CN activator CGA (Fig. 7F). Furthermore, cells treated with CPM showed a 78.0 Ϯ 3.9% FIGURE 6. PERK regulates ER Ca 2؉ reuptake through modulation of SERCA pump activity. A, [Ca 2ϩ ] ER of INS1 832/13 cells in response to carbachol was measured using FRET-based probe D1ER cameleon and expressed as a YFP/CFP ratio (R) normalized to that before stimulation (R 0 ) (R/R 0 ). Cells were exposed to PI or vehicle 20 min before the experiment. Shown are means of biological individuals (n Ͼ 5, area under the curve (AUC) of ER Ca 2ϩ uptake phase: control ϭ 100 Ϯ 7.4%, PI ϭ 58.5 Ϯ 10.2%, p Ͻ 0.05). B, [Ca 2ϩ ] ER of INS1 832/13 cells in response to carbachol was measured by FRET-based D1ER probe. Cells were exposed to 2-APB with or without PI for 20 min before experiment. Data are represented as means of all biological individuals (n Ͼ 5, AUC of ER Ca 2ϩ uptake phase: control ϭ 100 Ϯ 11.9%, PI ϭ 60.2 Ϯ 18.4%, p ϭ 0.051). C, SERCA-mediated ER Ca 2ϩ reuptake of INS1 832/13 cells (left, n Ͼ 5), primary rat islet cells (middle, n Ͼ 4), and primary human islet cells (right, n Ͼ 5). Based on estimation of the ␤-cell fraction from glucose-stimulated cytosolic Ca 2ϩ measurement, 75% of the cells measured were ␤-cells. Samples were pre-treated with 20 mM glucose in KRB for 1 h with CPA plus vehicle/PI added in the last 20 min. Shown are means of biological individuals (quantification of AUC showed: left, control ϭ 100 Ϯ 12.3%, PI ϭ 23.9 Ϯ 5.2%, p Ͻ 0.01; middle, control ϭ 100 Ϯ 16.4%, PI ϭ 43.8 Ϯ 6.8%, p Ͻ 0.05; right, control ϭ 100 Ϯ 11.9%, PI ϭ 60.4 Ϯ 6.7%, p Ͻ 0.01). D, representative Western blots (WB) from three independent experiments showing calnexin (CNX) protein immunoprecipitated (IP) with SERCA antibody. INS1 832/13 cells were pretreated with 20 mM glucose in KRB for 1 h, with CPA plus PI/vehicle added in the last 20 min. E, quantitative analysis of IP Western by measuring pixel density of blots. Data represent as intensity of CNX blot relative to WT control and shown as mean Ϯ S.E. (n ϭ 12, *, p Ͻ 0.05).
(p Ͻ 0.05) decrease in ER Ca 2ϩ reuptake, whereas CGA brought ER Ca 2ϩ reuptake in the PERK-inhibited cells back to control levels (p Ͻ 0.05). Taken together, the results were consistent with the hypothesis that PERK regulates Ca 2ϩ signaling in a CN-dependent pathway.

DISCUSSION
Perk is among a small number of genes that are so important for maintaining glucose homeostasis that their loss results in permanent neonatal diabetes (43,44), the most severe form of diabetes. Previously we showed that the diabetic phenotype of Perk genetic deficiency was due to the absence of PERK in pancreatic ␤-cells (19). However, the normal function of PERK in the ␤-cell has remained elusive. Initially it was proposed that the main function of PERK was to provide an adaptive response to presumed fluctuations in normal ER stress by temporarily repressing protein synthesis so as to ease client protein load in the ER (14,16,17). Subsequent studies on Perk-deficient mice and cultured ␤-cells, however, have generally not supported this hypothesis. Rather, we discovered dysfunctions in the basic  Fig. 2A (n ϭ 7, *, p Ͻ 0.05; **, p Ͻ 0.01). D, SOCC-mediated cytosol Ca 2ϩ influx in INS1 832/13 cells was measured using Fura2. Cells were exposed to the suggested chemicals 20 min before experiments. E, quantitative analysis of Fig. 6D based on AUC calculation. Data represent the percentage changes relative to WT control. Shown are mean Ϯ S.E. (n Ͼ 16, *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). F, SOCC-mediated cytosol Ca 2ϩ influx in INS1 832/13 cells was measured using Fura2 following the procedure as described in the legend to Fig. 5G. Cells were exposed to the suggested chemicals 20 min before experiments (n Ͼ 15, AUC did not show significance between two treatments). G, SERCA-mediated ER Ca 2ϩ refilling in INS1 832/13 cells was measured using the FRET-based D1ER probe following the same procedure as described in the legend to Fig. 5B. Cells were exposed to the indicated chemicals 20 min before experiments. H, quantitative analysis of Fig. 6F based on AUC measurement. Data represents as percentage changes relative to WT control. Shown are mean Ϯ S.E. (n Ͼ 5, *, p Ͻ 0.05; **, p Ͻ 0.01).
processes of the pancreatic ␤-cell including ablated glucosestimulated insulin secretion (19,27) and proinsulin trafficking (20). Because Ca 2ϩ is a key driver of insulin granule exocytosis and PERK is activated by ER Ca 2ϩ depletion, we postulated that these defects seen in GSIS could be due to misregulation of intracellular Ca 2ϩ . The use of genetic tools to probe PERK function is confounded by compensatory adaptive responses that can obfuscate interpretations of underlying mechanisms. The development of a highly selective PERK inhibitor provided by GlaxoSmithKline (22) provided us a powerful tool to determine the immediate consequences of ablating PERK function. We found that acute inhibition of PERK enzymatic activity rapidly leads to a reduction in both glucose-and KCl-stimulated insulin secretion, thus recapitulating the defect in stimulated insulin secretion when Perk is genetically ablated. Moreover, PERK inhibition resulted in pronounced suppression of glucose-and KCl-stimulated Ca 2ϩ influx in both rat and human primary ␤-cells. To provide an independent method to confirm that PERK regulates stimulated Ca 2ϩ uptake in ␤-cells, we employed a genetic method to acutely knockdown PERK expression through the use of shRNA directed against the Perk mRNA. We found that KCl-stimulated Ca 2ϩ influx was strongly suppressed in ␤-cells that were knocked down for PERK expression.
Glucose or KCl stimulation of ␤-cells induces a major influx of Ca 2ϩ into the cytosol that occurs largely through the voltagedependent Ca 2ϩ channels, and therefore we anticipated that the negative effect of the PERKi on Ca 2ϩ uptake was due to inhibition of the VDCCs. However, direct measurement of the VDCC-dependent Ca 2ϩ current showed no impact of PERK inhibition. Other plasma membrane ion channels such as storeoperated Ca 2ϩ channels and transient receptor potential channels, are known to either conduct Ca 2ϩ or to synergize VDCC by further enhancing depolarization (3,11,(45)(46)(47). The SOCC and some of the key TRP channels (e.g. TRPM2) are inhibited by 2-APB (48,49). We found that like PERKi, 2-APB significantly suppresses insulin secretion as well as KCl-induced cytosol Ca 2ϩ entry. Moreover, co-treatment of 2-APB and PI does not further inhibit KCl-stimulated cytosol Ca 2ϩ entry consistent with the hypothesis that they act through the same pathway. Therefore we postulate that PERK regulates Ca 2ϩ entry via one or more 2-APB sensitive ion channels.
Inasmuch as SOCE is purported to be dependent upon ER Ca 2ϩ release (50), we examined the impact of PERK inhibition on ER Ca 2ϩ release and reuptake. We found that PERKi does not induce ER Ca 2ϩ depletion nor does it influence Ca 2ϩ release stimulated by inhibition of SERCA or activation of IP 3 receptor. However, restoration of ER Ca 2ϩ , in the wake of ER Ca 2ϩ depletion/release, is strongly impaired by PERK inhibition. Because restoring ER Ca 2ϩ is largely dependent upon SERCA, we propose that PERK positively regulates SERCA activity under conditions of diminished ER Ca 2ϩ when SERCA is maximally activated. This proposal is supported by previous studies that PERK is activated under conditions of low ER Ca 2ϩ and repressed when ER Ca 2ϩ is at its high steady-state level (31). Thus PERK may function to induce SERCA to a highly activated state when rapid restoration of ER Ca 2ϩ is required during acetylcholine-stimulated insulin secretion or when the ER is acting to buffer Ca 2ϩ influx into the cytoplasm during glucose-stimulated insulin secretion (4). PERK and SERCA are unique to multicellular eukaryotes and neither is present in yeast. We postulate that emergence of PERK during evolution was in response to the need to modulate SERCA activity vis à vis intracellular Ca 2ϩ dynamics.
How PERK regulates glucose-and KCl-stimulated Ca 2ϩ entry and SERCA activity is an intriguing question in light of the well known activity for PERK of modulating gene expression via the eIF2␣ pathway. Inasmuch as Perk deficiency and an eIF2␣ knock-in mutation of the regulatory phosphorylation site in mice both result in proinsulin trafficking defect and diabetes (19,52), it is clear that at least some of the functions of PERK in ␤-cells are mediated by eIF2␣ phosphorylation. However, the speed at which acute PERK inhibition can negatively affect Ca 2ϩ entry into the cytoplasm and restoration of ER Ca 2ϩ argues against the involvement of the eIF2␣-mediated pathway and its subsequent regulation of protein synthesis. Hence we turned toward examining the possibility that PERK may mediate Ca 2ϩ regulation through interaction with calcineurin (14), a Ca 2ϩ /calmodulin-dependent protein phosphatase. In addition to a multiplicity of functions in modulating Ca 2ϩ /calmodulindependent processes in the cell, CN was more recently shown to bind to PERK and increase its activation (53). The CN-dependent activation of PERK was also shown to be enhanced by increased cytoplasmic Ca 2ϩ levels. In pancreatic ␤-cells, we confirmed that KCl-induced Ca 2ϩ uptake increased PERK activity. We discovered that the effects of inhibiting CN on intracellular Ca 2ϩ signaling in ␤-cells recapitulated the effects seen in PERK-inhibited cells including blunting glucose and KCl-stimulated Ca 2ϩ uptake, restoration of ER Ca 2ϩ , and inhibition of SOCE. Moreover, hyperactivation of CN was able to significantly reverse the negative effects of PERKi on Ca 2ϩ regulation, supporting the hypothesis that CN lies downstream of PERK in modulating intracellular Ca 2ϩ . Because immunosuppression therapy using CN inhibitors often leads to diabetes, CN has been suspected of playing an important role in ␤-cell development and function (24). Recently a targeted mutation of the calcineurin b1 (Cnb1) mouse gene has revealed that CN is required for normal insulin secretion, ␤-cell proliferation, and glucose homeostasis (25). Indeed, the postnatal diabetic progression in Cn1b KO mice closely mimics that of Perk KO mice.
We propose that PERK mediates its regulation of SERCA via a sequence of steps including PERK activation of CN, inactivation of calnexin, and release of calnexin inhibition of SERCA (Fig. 8). The restoration of ER Ca 2ϩ as synergized by this pathway may be particularly important to non-nutrient secretagogues such as acetylcholine that act through the phospholipase C-IP 3 pathway. Stimulated release of ER Ca 2ϩ stores via the IP 3 receptor not only helps to drive insulin secretion but also the resultant ER Ca 2ϩ depletion further induces Ca 2ϩ uptake through plasma membrane SOCC. Thus PERK-and CN-dependent regulation of SERCA and SOCC may also orchestrate non-nutrient secretagogue stimulation of insulin secretion.
The identities of the plasma membrane channels or receptors that mediate PERK-and CN-dependent regulation of Ca 2ϩ uptake stimulated by glucose and KCl are as yet uncertain.
Treatment of ␤-cells with glucose or KCl results in elevation rather than depletion of ER Ca 2ϩ , and therefore participation of SOCE in the initial Ca 2ϩ spike would seem unlikely because SOC is argued to be induced by ER Ca 2ϩ depletion (50). However, we found that 2-APB, a potent inhibitor of SOCC (35), strongly blunts glucose-and KCl-stimulated Ca 2ϩ uptake. We speculate that either SOCC is activated by the rapid increase in externally derived Ca 2ϩ or that 2-APB blocks a channel or receptor that is activated by the rise in cytosolic Ca 2ϩ conducted by the VDCC. The dependence of SOCC activation on ER Ca 2ϩ depletion has been challenged, in part, by the finding that under physiological conditions the ER does not experience substantial Ca 2ϩ depletion (54). In addition, SOCC has been shown to be sensitive to small changes in cytoplasmic Ca 2ϩ . Further studies will be required to identify the factors regulated by PERK and CN, which modulate the rapid rise in cytoplasmic Ca 2ϩ via the VDCC in response to glucose stimulation of ␤-cells.
Why PERK has evolved to regulate Ca 2ϩ dynamics is likely tied to how PERK activity is regulated. Although others have proposed that PERK is primarily activated by the accumulation of unfolded proteins in the ER under stress conditions (55), we noted previously that PERK is normally activated in highly secretory tissues such as the pancreas and therefore likely to be activated by normal physiological processes (14,18). In this study we discovered that elevation of cytosolic Ca 2ϩ , as induced by KCl, stimulated PERK activation. This is in contrast to activation of PERK associated with ER Ca 2ϩ depletion that occurs when SERCA is inhibited. However, it should be noted that ER Ca 2ϩ depletion results in a transient increase in cytosolic Ca 2ϩ , which may be the cause of PERK activation. Bollo and co-workers (53) have suggested that CN may act as a Ca 2ϩ sensor between the cytoplasm and ER. We further speculate that PERK and CN may act together to monitor and coordinate the balance between cytoplasmic and ER Ca 2ϩ levels, which is critically important in the regulation of stimulated insulin secretion.
The discovery that PERK regulates cellular Ca 2ϩ dynamics may have profound implications to PERK function in other organs and tissues such as the nervous system where Ca 2ϩ plays a dominant role in synaptic transmission (56). Interestingly, PERK and CN have been shown to be required for flexibility in hippocampal-dependent memory extinction in spatial memory tasks (51,57), and we suggest that this function may be dependent upon PERK and CN modulation of intracellular Ca 2ϩ dynamics. PERK activity may be directly regulated by cytoplasmic Ca 2ϩ or indirectly through interaction with calcineurin, a known cytoplasmic Ca 2ϩ sensor. In contrast Ca 2ϩ depletion in the ER can result in PERK activation. PERK-dependent regulation of Ca 2ϩ uptake in the ER is likely to be mediated through calcineurin dephosphorylation of calnexin and disassociation with SERCA. Release of CNX from SERCA then results in activation of SERCA and restoration of ER Ca 2ϩ . SERCA plays an important role in buffering Ca 2ϩ entering of the cytoplasmic during glucosestimulated secretion and restoring ER Ca 2ϩ following acetylcholine-stimulated insulin secretion. PERK and CN also act together to regulate plasma membrane channels and receptors such as TRP and SOCC to regulate glucose-stimulated Ca 2ϩ entry into the cytoplasm.