ER stress increases store-operated Ca2+ entry (SOCE) and augments basal insulin secretion in pancreatic beta cells

Type 2 diabetes mellitus (T2DM) is characterized by impaired glucose-stimulated insulin secretion and increased peripheral insulin resistance. Unremitting endoplasmic reticulum (ER) stress can lead to beta-cell apoptosis and has been linked to type 2 diabetes. Although many studies have attempted to link ER stress and T2DM, the specific effects of ER stress on beta-cell function remain incompletely understood. To determine the interrelationship between ER stress and beta-cell function, here we treated insulin-secreting INS-1(832/13) cells or isolated mouse islets with the ER stress–inducer tunicamycin (TM). TM induced ER stress as expected, as evidenced by activation of the unfolded protein response. Beta cells treated with TM also exhibited concomitant alterations in their electrical activity and cytosolic free Ca2+ oscillations. As ER stress is known to reduce ER Ca2+ levels, we tested the hypothesis that the observed increase in Ca2+ oscillations occurred because of reduced ER Ca2+ levels and, in turn, increased store-operated Ca2+ entry. TM-induced cytosolic Ca2+ and membrane electrical oscillations were acutely inhibited by YM58483, which blocks store-operated Ca2+ channels. Significantly, TM-treated cells secreted increased insulin under conditions normally associated with only minimal release, e.g. 5 mm glucose, and YM58483 blocked this secretion. Taken together, these results support a critical role for ER Ca2+ depletion–activated Ca2+ current in mediating Ca2+-induced insulin secretion in response to ER stress.

Type 2 diabetes mellitus is characterized by impaired glucose-stimulated insulin secretion in the setting of insulin resistance (1)(2)(3). Insulin secretion from pancreatic beta cells is triggered by glucose-induced Ca 2ϩ entry triggered by the closure of K ATP channels (4 -6). In many preparations, Ca 2ϩ entry is manifested by regular oscillations in cytosolic Ca 2ϩ , where each oscillation in turn provokes the release of insulin granules (4,(7)(8)(9)(10). Maintaining intracellular Ca 2ϩ homeostasis is critical for proper insulin secretion and for retaining beta-cell fitness.
In mammalian cells, such as the pancreatic beta cell, the ER 2 is the intracellular organelle where proteins of the secretory pathway are synthesized and initially packaged for export (11). In addition, the ER maintains protein quality control (12) and serves as a Ca 2ϩ reservoir that sequesters but also can release free Ca 2ϩ into the cytosol to generate a physiological signal (13)(14)(15). Ca 2ϩ is pumped into the ER lumen via sarco/endoplasmic reticulum Ca 2ϩ -ATPases (SERCA pumps) and released to the cytosol through the triggered activation of inositol trisphosphate and/or ryanodine receptors in the ER membrane (16 -21).
Beta cells undergo apoptosis after sustained exposure to the ER stress inducers tunicamycin (TM), thapsigargin, dithiothreitol (DTT), and high glucose or saturated fatty acid (22)(23)(24)(25)(26)(27). These conditions activate the unfolded protein response (UPR) through various mechanisms to restore normal proteostasis and preserve beta-cell function and viability (22,23). For instance, TM inhibits GlcNAc phosphotransferase, the key enzyme involved in the N-glycosylation of proteins, which in turn leads to the misfolding of glycoproteins in the ER (28). The resulting ER stress causes UPR activation, which in turn may restore proper protein folding and trafficking, increase the protein-folding capacity of the cell, and cause the degradation of misfolded proteins. In addition, further activation of the UPR inhibits new protein synthesis to reduce the protein load of the ER during times of increased stress.
Disrupted ER homeostasis has been proposed to be a potential cause of T2DM (14), and increasing evidence has emerged suggesting that the ER stress cascade is activated in islets from T2DM patients and from animal models of diabetes (23,29,30). We have recently discussed the potential links between disrupted ER homeostasis and altered beta-cell function in a review article (31). Other groups have also proposed the relevance of UPR signaling to beta-cell loss and the pathology of diabetes (32). We wished to advance the study of ER stress in disrupting specific beta-cell function, such as ER Ca 2ϩ -handling, cytosolic Ca 2ϩ oscillations, and insulin secretion, and we also wanted to determine how these changes in turn affected long-term beta-cell survival.
In our study, we used TM to experimentally-induce ER stress in insulin-secreting INS-1(832/13) cells or isolated mouse islets. ER stress responses in the form of UPR end points, ER and cytosolic Ca 2ϩ levels, insulin secretion, and beta-cell death were measured at various time points after exposing islets or cells to TM to determine the timeline of these events. TM treatment increased cytosolic Ca 2ϩ and insulin secretion, even in 5 mM glucose, a level that is below the normal glucose threshold of insulin secretion and the triggering of cytosolic Ca 2ϩ oscillations. We further found that this abnormal Ca 2ϩ signaling resulted from the activation of store-operated Ca 2ϩ entry (SOCE), most likely due to a stress-induced reduction of ER Ca 2ϩ concentration. The possible significance of this novel mechanism for augmenting insulin secretion for patients with T2DM is discussed.

Tunicamycin induced the ER stress response and apoptosis
Tunicamycin, a commonly-used pharmacological inducer of ER stress in beta cells, inhibits protein glycosylation (22,(33)(34)(35). To investigate the relationship between ER stress, ER Ca 2ϩ , and cytosolic Ca 2ϩ , we systematically measured the concentration of Ca 2ϩ in the cytosol and ER in parallel with UPR markers to establish their respective time courses following TM treatment. Changes in the three canonical ER stress-response markers, spliced XBP1, CHOP, and BiP, were determined at the mRNA or protein level. Mouse islets or insulin-secreting INS-1(832/13) cells were treated with vehicle (DMSO) as a control or TM for 6, 12, or 16 h in 11 mM glucose-containing medium prior to extracting total-cell mRNA and making whole-cell protein lysates.
As shown in Fig. 1, A and B, XBP1 splicing increased after 6 h of TM treatment in both INS-1(832/13) cells and mouse islets, whereas total XBP1 levels were unchanged. Similarly, as shown in Fig. 1C, CHOP increased after 6 h of TM treatment in INS-1(832/13) cells. In contrast, as shown in Fig. 1, D-G, levels of BiP protein only increased after 12 h of exposure to TM. XBP1 splicing is known to be an early event in the UPR, whereas the up-regulation of BiP expression has been reported to be more delayed (22,23,36,37).
Apoptosis occurs in a variety of cell types as a consequence of prolonged ER stress (23), and previous studies have shown that TM induces cell death in INS-1(832/13) cells and other cell lines (35,(38)(39)(40)(41)(42). To determine the presence of apoptosis, we assayed the level of cleaved PARP protein, an established marker of apoptosis (43,44), by Western blotting. As shown in Fig. 2, A and B, a band corresponding to cleaved PARP was visible at 89 kDa in lysates obtained from INS-1(832/13) cells exposed to TM for 12 h or more. Cleaved PARP was barely detected in any of our cell samples under control conditions or if TM exposure was for 6 h or less. It thus appeared that TM only triggered significant apoptosis after 12 h. The percentage of cleaved versus total PARP was monitored and is shown in Fig.  2B to rule out the effect of uneven protein loading. In addition, the percentage of INS-1(832/13) cells that take up propidium iodide (PI), a dye that is indicative of cell death, only increased after 16 h of TM treatment, compared with DMSO-treated controls (Fig. 2C). Cell death as assessed using this marker was not observed at any of the earlier time points studied. Fig. 2D shows there was a 4-fold increase in the number of cells in the sub-G 1 phase following 24 h of exposure to TM, indicating they were late-stage apoptotic cells compared with DMSO-treated controls.
Taken together, TM triggered a classic ER stress response in INS-1(832/13) cells after 6 h, whereas apoptosis was only seen after 12 h. Quantitative beta-cell death, in turn, was evident much later, after about 16 h of TM treatment, as evidenced by increased propidium iodide uptake.

Tunicamycin led to ER Ca 2؉ loss
As mentioned, TM has been used to induce ER stress in several studies of beta cells (22,23,33,35). The ER plays an important role in beta-cell function because it is the site where pro-teins of the secretory pathway are folded and processed in preparation for transport to the Golgi apparatus (15,45), and it is the location where proteostasis occurs (45,46). In terms of cellular Ca 2ϩ homeostasis, the ER also has a central role in this process due to its ability to sequester and buffer cytosolic Ca 2ϩ and serves as a releasable Ca 2ϩ source in response to surface membrane G-protein-coupled receptor signaling, and it supplies Ca 2ϩ to Ca 2ϩ -binding ER-resident protein chaperones that act to ensure proper protein folding (47,48).
To test whether TM altered ER Ca 2ϩ levels in our system, the ER Ca 2ϩ probe D4ER was transiently expressed in islet beta cells using an adenovirus delivering the d4er gene placed behind the rat insulin promoter 2. Islets were then treated with vehicle control (DMSO) or TM (10 g/ml) for 6, 12, or 16 h, and then ER Ca 2ϩ was measured. Fig. 3A shows ER Ca 2ϩ normalized to the initial FRET ratio (F0), expressed in relative units, as a function of time, and the effect of thapsigargin (TG, 1 M) is shown for both control and TM-treated islets. TG is a SERCA blocker that is well-known to deplete ER Ca 2ϩ by blocking Ca 2ϩ uptake into the ER (49). Only beta cells that responded to TG are shown in Fig. 3A; these constituted ϳ50% of the beta cells

ER stress increases SOCE and beta-cell insulin secretion
tested. As shown in Fig. 3B, TM caused a decline of steady-state ER Ca 2ϩ in islets compared with DMSO after 6, 12, and 16 h of treatment.

Tunicamycin increased cytosolic free Ca 2؉ under sub-threshold glucose conditions
Mouse islets do not typically show oscillations in cytosolic Ca 2ϩ or electrical activity when acutely exposed to glucose concentrations Ͻ7 mM (2,(50)(51)(52). To determine the relationship between ER stress and cytosolic free Ca 2ϩ in our experimental system, mouse islets were exposed to TM or vehicle control (DMSO) in standard RPMI 1640 medium for 6, 12, or 16 h. Following this treatment, cytosolic free Ca 2ϩ and islet electrical activity were recorded in parallel studies using an extracellular recording solution containing 5 mM glucose.
As shown in Fig. 4A, cytosolic free Ca 2ϩ in control islets did not display oscillatory activity in 5 mM glucose solution, as expected (2,(50)(51)(52). In contrast, islets treated with TM exhibited islet Ca 2ϩ oscillations or Ca 2ϩ transients when exposed to the TM for 6 h or more. 40% of islets treated with TM for 6 h displayed free Ca 2ϩ oscillations compared with those treated with DMSO (Fig. 4B). Treatment with TM for 12 or 16 h resulted in a greater percentage of oscillating islets.
The plateau fraction, frequency, and amplitude of oscillating islets as well as their baseline Ca 2ϩ levels were analyzed and plotted in Fig. 5. Plateau fraction, oscillation frequency, and amplitude were not plotted for control islets as they did not exhibit oscillations. Statistically significant increases in baseline Ca 2ϩ levels were observed after 6 and 12 h of TM exposure (Fig. 5B).

Changes in electrical activity occurred in parallel with changes in Ca 2؉ oscillations
Our observation that islets treated with TM exhibited cytosolic Ca 2ϩ oscillations ( Fig. 6) was next confirmed by separate measurements of islet electrical activity, obtained using perforated patch-clamp recordings. TM-treated beta cells thus exhibited oscillations in islet membrane potential in 5 mM glucose, which was rarely observed in control islets exposed to the same glucose concentration, as was found for Ca 2ϩ . However, the occurrence of oscillations was related to the duration of TM treatment. As shown in Fig. 6, islets subjected to TM for 6 h showed occasional oscillations in 5 mM glucose, whereas islets treated for 12 or 16 h showed regular oscillations having an average period of 5-8 min. Importantly, the oscillations we observed in TM-treated islets in 5 mM glucose strongly resembled those of normal islets exposed to glucose concentrations Ͼ7-8 mM (53).

Tunicamycin increased insulin secretion under sub-threshold glucose conditions
When beta cells are depolarized, Ca 2ϩ influx through voltage-gated Ca 2ϩ channels leads to a rise in cytosolic Ca 2ϩ concentration that triggers the release of insulin granules from the cell (4, 10, 51, 54). To test whether the changes we observed in islet electrical and cytosolic Ca 2ϩ activity in response to TM

ER stress increases SOCE and beta-cell insulin secretion
treatment were sufficient to elicit insulin secretion even under normal subthreshold conditions, islets were pretreated with DMSO or TM in standard RPMI 1640 medium (including 11 mM glucose) for 6, 12, or 16 h. After treatment, islets were thoroughly washed, and a static incubation protocol was used to measure insulin secretion in 5 mM glucose. As shown in Fig. 7A, insulin secreted into the medium was significantly increased after 12 h or more of TM exposure, whereas islet insulin content was unchanged. Expressed another way, TM exposure for 12 h or more resulted in greater insulin secretion as a percent of insulin content, compared with controls ( Fig. 7B). The time course of increased secretion closely paralleled the increase in cytosolic Ca 2ϩ or electrical activity depicted in Figs. 4 -6. They support the hypothesis that the activation of cytosolic Ca 2ϩ activity by ER stress in 5 mM glucose was triggered by increased islet electrical activity and was sufficient to release more insulin from the beta cell. Secreted insulin and percent insulin content were both higher after 6 h of TM compared with DMSO exposure, but the differences were not statistically significant. The unique aspect of the 6-h time point will be addressed under the "Discussion." We also point out that the magnitude of the secretion response of TM-treated islets in 5 mM glucose is still much lower than that seen in response to 11 mM or more glucose.

Tunicamycin increased cytosolic Ca 2؉ , membrane potential oscillations, and insulin secretion through store-operated Ca 2؉ entry
SOCE links reduced ER Ca 2ϩ concentration to the activation of voltage-independent, plasma membrane Ca 2ϩ channels that can replenish the depleted ER, with Ca 2ϩ entering the cell from the extracellular space (55)(56)(57). To determine whether SOCE played a role in mediating the oscillations we observed following chronic ER stress and ER Ca 2ϩ lowering, we tested whether YM58483, a selective blocker of membrane SOCE channels, interfered with our physiological end points (58 -60). As shown in Fig. 8, A and B, both cytosolic Ca 2ϩ oscillations and the electrical activity observed in TM-treated islets in 5 mM glucose were abruptly abolished by YM58483 treatment. These results show that SOCE, which normally plays little or no role in the genesis of glucose-induced islet electrical oscillations (61), was here facilitated by TM-induced ER stress in beta cells, presumably because TM reduced ER Ca 2ϩ . Importantly, YM58483 also blocked TM-induced insulin secretion in islets bathed in 5 mM glucose (Fig. 8, C and D). In contrast, the addition of YM58483 had no effect on insulin secretion, islet electrical activity, or intracellular Ca 2ϩ in control islets (Fig. 8, A-D).
At the molecular level, the main components of SOCE are stromal interaction molecule-1 (STIM1) and Ca 2ϩ releaseactivated Ca 2ϩ channel protein 1 (ORAI1). STIM1 is an ER Ca 2ϩ sensor, whereas ORAI1, which is found on the plasma membrane, is the pore-forming subunit of functional SOCE. When STIM1 senses ER Ca 2ϩ depletion, STIM1 molecules aggregate and interact with ORAI1 at ER-plasma membrane junctions, and this complex mediates Ca 2ϩ influx through SOCE (62,63). To confirm the results we obtained at the molecular level, siRNA was used to knock down STIM1 in INS-1(832/ 13) cells. Transfecting INS-1(832/13) cells with siRNA-STIM1 (siSTIM1) reduced STIM1 mRNA by ϳ80 -90% (Fig. 9A) and

ER stress increases SOCE and beta-cell insulin secretion
STIM1 protein by ϳ70 -75% (Fig. 9, B and C) compared with treatment with control siRNA (siCon). STIM1 reduction did not result in significant up-regulation of ORAI1, suggesting the cells did not compensate for the loss of STIM1 (Fig. 9, B and C). The percentage of cells showing cytosolic Ca 2ϩ transients in 5 mM glucose was decreased in siSTIM-transfected cells (ϳ40%) compared with siCon-transfected control cells (ϳ10%) following 16 h of TM exposure (Fig. 9, D and E). Even after 16 h of DMSO exposure, controls showed no change in their Ca 2ϩ activities.
As an alternative to blocking SOCE channels with YM, we tested whether removing extracellular Ca 2ϩ was similarly able to abolish the cytosolic Ca 2ϩ oscillations we observed in TMtreated islets in 5 mM glucose. Removing extracellular Ca 2ϩ confirmed the results we obtained with YM, supporting the hypothesis that the oscillations seen after TM treatment indeed require increased influx of extracellular Ca 2ϩ (Fig. S1A). However, applying other SOCE channel blockers, 2-aminoethoxydiphenyl borate (2APB) or SKF96365 (SKF), acutely at the end of a Ca 2ϩ -imaging experiment surprisingly increased cytosolic Ca 2ϩ levels in both control and experimental groups (Fig. S1, B and C) (64,65). 2APB and SKF are nonselective SOCE inhibitors as they also inhibit other channels over a similar concentration range (66).
Although ER Ca 2ϩ decreased in TM-treated islets compared with controls, blocking SOCE with YM58483 had little or no measurable effect on the ER Ca 2ϩ levels of either control or TM-treated beta cells (Fig. 10A). This finding was unexpected, but will be addressed further under the "Discussion." Blocking SOCE with YM58483 also did not affect any of the TM-induced UPR end points we measured (Fig. 10B).
Previous reports have shown that elevated cytosolic Ca 2ϩ is detrimental to beta cells (67). Thus, preventing excessive cytosolic Ca 2ϩ elevation due to overactive SOCE might have at least partly protected beta cells from cell death induced by prolonged exposure to TM. However, as shown in Fig. 10C, 24 h of treatment with TM increased cell death in INS-1(832/13) cells, but we found no protection afforded by the inclusion of YM58483.

Tunicamycin did not affect cytosolic free Ca 2؉ under above-threshold glucose conditions
To maintain glucose homeostasis, beta cells secrete insulin when blood glucose concentration rises. Islets exhibit oscillations in cytosolic free Ca 2ϩ when exposed to 7 mM or more of glucose (52). After isolated mouse islets were exposed to TM or vehicle control (DMSO), free Ca 2ϩ and insulin secretion were measured in parallel in 11 mM glucose. As shown in Fig. 11A, both control and experimental groups showed Ca 2ϩ oscillations in 11 mM glucose. The percentages of oscillating islets we observed were very similar between the two groups (70 -80%), whereas the remaining islets tended to go to a plateau (Fig.  11B). The frequency of the oscillations observed in TM-treated islets was higher than for controls, whereas no significant change was observed in plateau fraction, baseline Ca 2ϩ , or oscillation amplitude (Fig. 11, C-F). In addition, we found no significant change in insulin secretion between experimental and control groups after they were stimulated with 11 mM glucose for 30 min (Fig. 11, G and H).

Other ER stress inducers also increased cytosolic free Ca 2؉ under sub-threshold glucose conditions
Besides tunicamycin, ER stress can be induced by treating islets with thapsigargin or high glucose (22,23). As shown in Fig. 12A, mouse islets exposed to thapsigargin (200 nM) for 16 h exhibited oscillatory cytosolic Ca 2ϩ levels despite being in 5 mM glucose. Similarly, mouse islets cultured in medium containing 25 mM glucose to induce stress also exhibited cytosolic Ca 2ϩ oscillations (Fig. 12B). These oscillations were also abruptly abolished by YM58483 treatment. DMSO-treated or 11 mM glucose-cultured islets did not exhibit Ca 2ϩ oscillations in 5 mM glucose solution, however, as expected.

Effect of tunicamycin on gene expression
As SOCE activated in response to TM treatment in our study, we also assayed the level of STIM1 and ORAI1 expression under these same experimental conditions. As shown in Fig.  13A, we observed a protein band corresponding to STIM1, as expected, and an additional smaller molecular-weight band in lysates from TM-treated INS-1(832/13) cells after 16 h of treat-

ER stress increases SOCE and beta-cell insulin secretion
ment. The intensity of the upper band for STIM1 was not significantly altered in response to TM compared with controls (Fig. 13B). ORAI1 protein was also unchanged by TM treatment (Fig. 13, A and B), as reported in another recent study (68). GLUT2 protein was also measured in INS-1(832/13) cells after 6 h of TM treatment compared with control, and no change in protein expression was found (Fig. S2, A and B).

Discussion
In this study, we sought to delineate the temporal relationship between the induction of ER stress, altered beta-cell function, and altered beta-cell viability, focusing on the role of ER and cytosolic Ca 2ϩ in these processes. Studies were carried out by exposing mouse islets or INS-1(832/13) cells to the glycosylation inhibitor tunicamycin for up to 24 h. We found that UPR activation appeared to be linked to a reduction in ER Ca 2ϩ and a phase of increased extracellular Ca 2ϩ influx linked to ER Ca 2ϩ unloading. The Ca 2ϩ oscillations that were triggered by storeoperated Ca 2ϩ influx were sufficient to trigger the release of insulin, even in normally sub-threshold glucose. Cell death was found to occur much later, e.g. after 16 h post-treatment and appeared to be independent of the early phase of SOCE-mediated Ca 2ϩ influx and concomitant insulin secretion.
Previous research carried out using many types of cells has shown that thapsigargin, a SERCA blocker, which prevents ATP-dependent Ca 2ϩ sequestration by the ER, unloads the ER Ca 2ϩ store, triggering SOCE (16,60). Activated SOCE results in increased cytosolic Ca 2ϩ , which serves to replenish the ER Ca 2ϩ pool (69). The recent findings reported by Yamamoto et al. (70) indicate that tunicamycin decreases ER Ca 2ϩ by increasing ryanodine receptor 2 activity, which in turn elicits spontaneous cytosolic Ca 2ϩ transients that are seen after raising extracellular Ca 2ϩ concentration. We agree with Yamamoto et al. (70) that ryanodine receptors (RyRs) are likely involved in ER stressinduced ER Ca 2ϩ lowering, as we observed inhibitory effects of the RyR blocker ryanodine (data not shown). However, we propose a very different interpretation in this paper. Our data show that ER stress conditions activate a Ca 2ϩ current mediated by SOCE channels under low-glucose conditions, which likely occurs secondary to ER Ca 2ϩ depletion by tunicamycin.
The normal glucose threshold for islet oscillations in our hands is near 7 mM (2, 50), which means that TM-induced ER stress in a sense increased the sensitivity of islets to glucose concentration. In our view, glucose-induced islet Ca 2ϩ oscillations are induced despite the low level of glucose by the activation of SOCE-mediated Ca 2ϩ current, which depolarizes the  Insulin secretion was measured by acutely exposing 10 islets to 5 mM glucose for 30 min for each experimental condition. A, both secreted insulin and insulin remaining in the extracted islets were quantified in triplicate by ELISA and normalized to total protein concentration (BCA protein assay). B, percent insulin that was secreted was obtained by dividing the secreted insulin by total insulin (the sum of secreted insulin and insulin in the lysate). Values shown are means Ϯ S.E. #, p Ͻ 0.05 compared with control conditions; n ϭ 3 mice.

ER stress increases SOCE and beta-cell insulin secretion
beta-cell membrane to threshold despite incomplete closure of beta-cell K ATP channels. The evidence for this interpretation is as follows. 1) The Ca 2ϩ oscillations we observed strongly resemble those of control islets exposed to glucose Ͼ7 mM, suggesting a common origin. 2) The Ca 2ϩ oscillations of stressed islets were completely blocked by the selective SOCE blocker YM58483 (58, 59); notably, this drug had no effect on untreated control islets. 3) Patch-clamp electrophysiology confirmed the electrical nature of the ER stress-induced oscillations, and as for the Ca 2ϩ oscillations, the electrical bursting we observed in 5 mM glucose was similarly abolished by YM58483. 4) The percentage of Ca 2ϩ oscillations was decreased in TMtreated siSTIM1-knockdown INS-1(832/13) cells compared with controls. Taken together, these data are in strong support of a plasma membrane-delimited mechanism, and they rule out intracellular store Ca 2ϩ release as the proximal cause of the Ca 2ϩ oscillations we observed in TM-treated islets, although we believe ER Ca 2ϩ depletion by ER stress indirectly caused the oscillations by triggering SOCE.
Physiologically, when blood glucose rises, K ATP channel closure mediates plasma membrane depolarization, which in turn increases cytosolic Ca 2ϩ , which then drives insulin secretion (2,5,8). Membrane potential changes in mouse beta cells have been shown to precede changes in cytosolic Ca 2ϩ under physiological conditions (5). The cytosolic Ca 2ϩ oscillations shown in Fig. 4A occurred in parallel with membrane potential oscillations in 5 mM glucose saline in response to TM treatment, as shown in Fig. 6. In simultaneous measurements of cytosolic Ca 2ϩ and insulin secretion, each oscillation in islet Ca 2ϩ has been shown to be well-synchronized with a pulse of insulin secretion (4,5,10).
Although cytosolic Ca 2ϩ was increased after 6 h of TM treatment, the changes in insulin secretion and percent insulin (Fig.  7) we measured at this time point were not statistically significant compared with controls, although the means we obtained were greater than controls. This may be explained by our observation that less than 40% of islets displayed elevated cytosolic Ca 2ϩ within 6 h of TM treatment (Fig. 4B). Our results at the  6-h time point may thus underestimate the amount of insulin secretion seen in response to TM because it included both responding and nonresponding islets.

ER stress increases SOCE and beta-cell insulin secretion
YM58483 did not affect the extent of ER Ca 2ϩ depletion that followed TM treatment (Fig. 10A), which was surprising. This could be due to several possible factors. 1) The influx of Ca 2ϩ due to SOCE may have been too small to cause a detectable change in ER Ca 2ϩ due to limits in the Ca 2ϩ sensitivity of the D4ER Ca 2ϩ probe. 2) SERCA expression and/or function might also be reduced by TM treatment, such that under these pathophysiological conditions SOCE is capable of mediating an electrical current and Ca 2ϩ oscillations but not significant ER store refilling. ER stress has in fact been reported to cause reduced SERCA2b expression in beta cells, which supports this idea (18,71,72). 3) The ER may become so leaky to Ca 2ϩ after TM treatment that a modest activation of SOCE was unable to do enough to measurably refill the ER, like turning on a small hose to refill a very leaky barrel.
Our results support the hypothesis that TM-triggered betacell death occurs as a consequence of ER Ca 2ϩ depletion and that SOCE activation is a separate action that is unrelated to the ultimate fate of the cell, as shown in Fig. 14. Similar observations and conclusions were made in studies of thapsigargintreated LNCaP, PC3, and MCF7 cells (49). Thapsigargin caused the unloading of ER Ca 2ϩ and resulted in cell death despite genetic knockdown of the SOCE components STIM1 and/or ORAI1 in this case. Therefore, ER Ca 2ϩ depletion due to ER stress appeared to be an important contributor to thapsigargininduced cell death, instead of SOCE activation and increased cytosolic Ca 2ϩ (49).
As shown in Fig. 13, A-D, we found two bands corresponding to STIM1 protein. The upper band of STIM1 expression at 77 kDa and ORAI1 expression remained unchanged. Both STIM1 and ORAI1 are known to be N-linked glycosylated proteins (62, 63, 73, 74). Other investigators also observed no change in ORAI1 in response to induced ER stress, whereas

ER stress increases SOCE and beta-cell insulin secretion
STIM1 responded to TM treatment. The slightly smaller molecular weight STIM1 species, representing nonglycosylated STIM1, were reported. Blocking STIM1 glycosylation led to diminished SOCE (73,74). Evans-Molina and co-workers (68) have recently reported that STIM1 was down-regulated in a diabetes model, whereas overexpressing STIM1 restored SOCE under high-glucose conditions. However, Evans-Molina and co-workers (68) propose that SOCE is an essential driver of glucose-induced Ca 2ϩ oscillations (15 mM glucose) under normal conditions and that SOCE is impaired in response to proinflammatory cytokines or palmitate-mediated stress conditions. In contrast, we propose that SOCE is not involved in the triggering or modulation of glucose-induced Ca 2ϩ oscillations in untreated control islets, but it is activated by ER stress, resulting in the appearance of Ca 2ϩ oscillations under subthreshold glucose conditions (5 mM glucose) by virtue of this abnormal triggering mechanism, which in essence shifts the glucose sensitivity of the islet to the left where islet Ca 2ϩ activity could then contribute to the production of high basal insulin release. Different glucose conditions may account for the different interpretations.
The justification for our use of insulin-secreting INS cells in addition to mouse islets in this paper relates to the small amount of tissue available for biochemical and molecular studies if just islets were used. For example, analyzing propidium iodide levels with flow cytometry in order to quantify cell death is extremely challenging if primary beta cells are used. INS-1(832/13) cells are one of the most commonly used insulinsecreting cell lines that display many important characteristics of primary beta cells. Importantly, INS-1(832/13) cells are very responsive to glucose (75). According to Fig. 1, A and B, INS-1(832/13) cells had identical UPR responses as isolated islets. Thus, we believe that the molecular studies done in INS-1(832/ 13) cells, while not perfectly reflecting what we might expect if islets or primary beta cells were used in their place, are reasonable surrogates for the primary cells with regard to UPR activation and cell death. This is not likely to be true regarding physiology where our methods are well-attuned to studying primary islets and their oscillatory and secretory characteristics.
In summary, as shown in Fig. 14, we propose that TM-induced beta-cell death occurs through ER Ca 2ϩ depletion, whereas SOCE and concomitantly increased cytosolic Ca 2ϩ were required for our finding increased insulin secretion under stress conditions. During prediabetes, which is associated with insulin resistance, the pancreatic beta cell is thought to compensate for rising levels of glucose by increasing insulin secretion and, if that fails, increasing beta-cell mass, provided the cells are capable of doing so (76). However, long-term hyperinsulinemia, and the increased metabolic workload it represents, can potentially exhaust the beta cell and promote beta-cell death (77). In our results, TM-induced ER stress resulted in increased beta-cell electrical activity, cytosolic Ca 2ϩ oscillations, and insulin secretion by activating SOCE. Blocking SOCE by applying YM58483 to stressed but not control islets abolished ER stress-triggered increases in electrical activity, cytosolic Ca 2ϩ oscillations, and insulin secretion (Fig. 8, A-D). Therefore, the increased insulin secretion data not only confirmed that TM increased cytosolic Ca 2ϩ oscillations, but it also established SOCE as the key mechanism.

ER stress increases SOCE and beta-cell insulin secretion
This report shows that SOCE is a key player in ER stressinduced cytosolic Ca 2ϩ oscillations and insulin secretion, and we suggest that this pathway must work in parallel with the UPR and cell-death pathways. The cytosolic Ca 2ϩ oscillations we observed clearly resulted from electrical oscillations and not Ca 2ϩ -induced Ca 2ϩ release from the ER. Thus, these results suggest the possibility that in T2DM or under prediabetic conditions increased secretion due to SOCE activation may contribute to the increased basal insulin secretion that is a hallmark of type 2 diabetes. Combining our findings with more detailed mechanistic and pharmacological studies on SOCE activity in prediabetes may disclose additional valuable information and perhaps novel treatment strategies.

Experimental procedures
Materials TM, YM, TG, 2APB, and SKF were all obtained from Cayman Chemical. Small interfering RNAs (siRNAs) were purchased from Thermo Fisher Scientific. Table S1, A and B, contains a complete list of PCR primers and antibodies, respectively. ECL reagents was obtained from Bio-Rad.

Isolation of pancreatic islets and islet pretreatments
Pancreatic islets were isolated from male Swiss-Webster mice (3 months of age; 25-35 g) according to the regulations of the University of Michigan Committee on the Use and Care of Animals (UCUCA), using previously described methods (78) and with an approved protocol. Isolated islets from a given mouse were divided into control and experimental groups, and both were cultured in standard RPMI 1640 medium containing 11 mM glucose, 10% fetal bovine serum (FBS), 10 mM HEPES, 1% penicillin/streptomycin, and 1% sodium pyruvate. Control islets were incubated with DMSO, whereas test islets were pretreated with 10 g/ml tunicamycin.

Cell culture and transfection
INS-1(832/13) cells were grown in RPMI 1640 medium containing 11 mM glucose, 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, 10 mM HEPES, and 1% sodium pyruvate. INS-1(832/13) cells were grown in 10-cm culture dishes, 6-well plates, or T25 flasks at 37°C in a 5% CO 2 -humidified atmosphere. Cells obtained ϳ70% confluence prior to the initiation of experimentation. INS-1(832/13) cells were transfected with Figure 11. Tunicamycin did not affect cytosolic free Ca 2؉ under above-threshold glucose conditions. Isolated pancreatic mouse islets were treated with a vehicle control (DMSO) or TM (10 g/ml) for 16 h in 11 mM glucose. A, responses of cytosolic free Ca 2ϩ to solution containing 11 mM glucose under the indicated conditions. B, percentage of oscillating islets. Summary findings for the data are shown in C-F. C, plateau fraction. D, baseline values. E, oscillation frequency. F, oscillation amplitude. G and H, insulin secretion was measured by acutely exposing 10 islets to 11 mM glucose for 30 min for each experimental condition. Both secreted insulin and insulin content (G) and the percent insulin (H) were quantified as described in Fig. 7. All values shown are means Ϯ S.E. ##, p Ͻ 0.01; n ϭ at least three mice.

ER stress increases SOCE and beta-cell insulin secretion
STIM1-specific siRNA or negative control siRNA using Lipofectamine RNAiMAX reagent as described in the manufacturer's protocol (Invitrogen). The treated cells were assessed by real-time PCR and Western blotting.

Real-time PCR
Total RNA was extracted from INS-1(832/13) cells or islets using the RNeasy mini kit (Qiagen) according to the manufa-cturer's instructions. One g of total RNA from INS-1(832/13) cells or 0.4 g of total islet RNA was reverse-transcribed using Superscript RT II (Invitrogen) according to the manufacturer's instructions. Real-time experiments were carried out using a SYBR Green PCR master mix (Applied Biosystems) with the primers shown in Table S1A. Raw threshold-cycle (C T ) values were obtained using Step One software, and mean C T values were calculated from triplicate PCRs for each sample. Data

ER stress increases SOCE and beta-cell insulin secretion
were presented as RQ values (2 Ϫ⌬⌬CT ) with expression presented relative to an endogenous control, HPRT1.

Western blotting
Total protein was obtained by treating INS-1(832/13) cells or mouse islets with KHEN lysis buffer (50 mM KCl, 50 mM HEPES, 10 mM EGTA, 1.92 mM MgCl 2 ; pH 7.2) and then separating proteins using 4 -12% SDS-PAGE and transferring them to nitrocellulose membranes. Membranes were blocked in 5% w/v nonfat dry milk or 5% BSA in 1ϫ TBST containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 0.1% Tween 20. Blots were incubated overnight with primary antibodies diluted in 5% nonfat dry milk in 1ϫ TBST at 4°C as described in Table S1B. Blots were incubated with horseradish peroxidase-conjugated mouse anti-rabbit antibodies or goal anti-mouse antibodies, and these were visualized using ECL reagents.

Fura-2/AM imaging
Islets were loaded with Fura-2/AM (2.5 M) for 45 min in medium containing 5 mM glucose prior to imaging. Islets were then transferred to a 1-ml perfusion chamber containing 5 mM glucose imaging buffer for 6 min, followed by 10 -30 min of perfusion with this solution at ϳ1 ml/min. Imaging buffer contained the following (in mM): 140 NaCl, 3CaCl 2 , 5 KCl, 2 MgCl 2 , 10 HEPES, and 5 glucose. Ratiometric fura-2 imaging was carried out using 340/380 nm excitation and collecting 502 nm emission, as described previously (78). The fluorescence data were acquired using Metafluor.

FRET measurements
To measure ER Ca 2ϩ , we utilized a previously described ERlocalized FRET biosensor, D4ER (79). The sensor was selectively expressed in the beta cells of intact mouse islets using an adenovirus and under the control of the rat insulin promoter, as described previously (79). The same system described above for Fura-2/AM imaging was employed here. D4ER imaging was carried out using 430 nm excitation and 470/535 nm ratiomet-ric emission. The imaging solution used contained (in mM): 140 NaCl, 3 CaCl 2 , 5 KCl, 2 MgCl 2 , 10 HEPES, 5 glucose, and 0.2 diazoxide (Dz). Dz was included to keep the K ATP channel in its open state to prevent oscillatory Ca 2ϩ activity and improve the signal/noise ratio and stability of the ER Ca 2ϩ recordings. FRET ratios were acquired using Metafluor, and mean values were calculated using Prism.

Analysis of cytosolic Ca 2؉ recordings
Traces containing cytosolic Ca 2ϩ oscillations were analyzed using MATLAB (Mathworks) to obtain the plateau fraction (PF), periods, baseline ratios, and relative amplitudes of Ca 2ϩ oscillations, as described (50). PF was calculated as the activephase duration divided by the period of each oscillation (50). Only islets displaying oscillations were assigned a PF, and those exhibiting a persistent plateau phase were assigned a PF value of 1.0.

Electrophysiology
Islet membrane potential was measured using a perforatedpatch whole-cell current clamp as described (53). Electrophysiological recordings were made from single beta cells in intact islets treated with TM for 6, 12, and 16 h, respectively. Islets treated with vehicle medium were used to make control recordings. Only one beta cell in each intact islet was patched. Membrane potential of each beta cell in an intact islet was recorded in current-clamp mode after perforated-patch configuration was established. The external recording solution contained the following (in mM): 140 NaCl, 3 CaCl 2 , 5 KCl, 2 MgCl 2 , 10 HEPES, and 5 glucose.

Assays of cell death
INS-1(832/13) cells were dislodged from T25 flasks with 0.05% trypsin and after gentle shaking, and PI was applied to label dead cells, as described in the manufacturer's protocol (Sigma). The percentage of PI-positive cells was determined using a flow cytometer provided by the Flow Cytometry Core of the University of Michigan.

Assays of apoptosis
INS-1(832/13) cells were harvested as described above under "Assays of cell death" and fixed in cold 70% ethanol and stored in 4°C. Before measurement, PI was added as described in the manufacturer's protocol (Sigma), and the percentage of apoptotic cells was determined by calculating the percentage of sub-G 1 cells in the DNA content histogram using a flow cytometer provided by the Flow Cytometry Core of the University of Michigan.

Glucose-stimulated insulin secretion assay
Islets were washed with glucose-free KRB buffer (115 mM NaCl, 4.7 mM KCl, 1.2 mM MgSO 4 ⅐7H 2 O, 1.2 mM KH 2 PO 4 , 20 mM NaHCO 3 , 16 mM HEPES, 2.56 mM CaCl 2 ⅐2H 2 O, 0.2% BSA) for 30 min at 37°C and then incubated with KRB buffer containing 5 mM or more glucose for an additional 30 min. The supernatant and islets were then collected separately to determine insulin content using a mouse insulin ELISA kit according to the manufacturer's instructions (Crystal Chem).