Evidence of Diminished Glucose Stimulation and Endoplasmic Reticulum Function in Nonoscillatory Pancreatic Islets

Abstract

Pulsatility is a fundamental feature of pancreatic islets and a hallmark of hormone secretion. Isolated pancreatic islets endogenously generate rhythms in secretion, metabolic activity, and intracellular calcium ([Ca²⁺]_i) that are important to normal physiological function. Few studies have directly compared oscillatory and nonoscillatory islets to identify possible differences in function. We investigated the hypothesis that the loss of these oscillations is a leading indicator of islet dysfunction by comparing oscillatory and nonoscillatory mouse islets for multiple parameters of function. Nonoscillatory islets displayed elevated basal [Ca²⁺]_i and diminished [Ca²⁺]_i response and insulin secretory response to 3–28 mm glucose stimulation compared with oscillatory islets, suggesting diminished glucose sensitivity. We investigated several possible mechanisms to explain these differences. No differences were observed in mitochondrial membrane potential, estimated ATP-sensitive potassium channel and L-type calcium channel activity, or cell death rates. Nonoscillatory islets, however, showed a reduced response to the sarco(endo)plasmic reticulum calcium ATPase inhibitor thapsigargin, suggesting a disruption in calcium homeostasis in the endoplasmic reticulum (ER) compared with oscillatory islets. The diminished ER calcium homeostasis among nonoscillatory islets was also consistent with the higher cytosolic calcium levels observed in 3 mm glucose. Inducing mild damage with low-dose proinflammatory cytokines reduced islet oscillatory capacity and produced similar effects on glucose-stimulated [Ca²⁺]_i, basal [Ca²⁺]_i, and thapsigargin response observed among untreated nonoscillatory islets. Our data suggest the loss of oscillatory capacity may be an early indicator of diminished islet glucose sensitivity and ER dysfunction, suggesting targets to improve islet assessment.

The loss of islet calcium oscillations correlates with impaired calcium regulation and reduced insulin secretion, suggesting oscillatory capacity is important to islet health and function.

As in many endocrine systems, pancreatic islets secrete hormones such as insulin and glucagon in pulses. Glucose is the primary trigger of insulin secretion and oscillatory activity in the pancreatic β-cell. As glucose rises, glucose is transported into β-cells through the GLUT2 transporter (1,2) and metabolized by a series of biochemical processes that involves glycolysis, increased endoplasmic reticulum (ER) activity, and increased mitochondrial ATP production (for review see Ref. 3). The resulting increase in the ATP/ADP closes ATP-sensitive potassium channels (K_ATP channels) (4,5,6), leading to calcium influx through L-type calcium channels, which triggers exocytosis of insulin-containing secretory granules and ensuing pulses of insulin release. Once blood glucose is returned to its basal level through insulin action, ATP/ADP levels drop in the β-cell, leading to the reopening of K_ATP-channels that, in turn, shuts off the glucose-induced electrical activity (7,8,9).

Insulin pulses result from complex interactions involving mitochondria, ER, plasma membrane ion channels, and other cellular processes that produce metabolic and electrical oscillations in the β-cell. At the level of the islet, the interplay among β-cells, counterregulatory glucagon-secreting α-cells, and other types of islet cells produces additional layers of complexity in rhythmic activity (10,11). These rhythms, ranging in period from seconds to about 5 min (12,13), occur in vivo and in vitro when islets are maintained in stimulatory glucose (∼8–15 mm). The slower form of oscillation (∼5 min) has been reported in ion channel activity (14,15), intracellular calcium ([Ca²⁺]_i) (16), metabolism (17,18,19), and insulin secretion (20). [Ca²⁺]_i is a key component along the insulin secretion pathway, and oscillations in islet [Ca²⁺]_i have been directly and causally linked with pulses of insulin secretion in vitro (21,22,23). Pulsatile insulin patterns measured from conscious mice in vivo have also been associated with similar rhythms in [Ca²⁺]_i from the isolated islets of these same mice (24). Insulin pulsatility thus originates from intrinsic islet oscillations in [Ca²⁺]_i and other processes (20).

Whereas pulsatility appears to be a natural function of islets both in vivo and in vitro, it has been hypothesized that disruptions in rhythmic function may be an early biomarker of islet dysfunction (25,26). In healthy humans, insulin pulses have been detected as frequently as one pulse per 4–10 min, which is consistent with measurements in other mammals (20). A reduction in the amplitude and regularity of insulin pulses has been linked with patients with type 2 diabetes (T2D) (27,28,29,30). The close relatives of diabetic patients also demonstrate considerable degradation of pulsatile insulin secretion despite reporting clinically normal glucose tolerance and normal insulin sensitivity (31,32). Aberrant insulin pulsatility has also been reported in specific forms of diabetes and in other metabolic disorders, including mature onset diabetes of the young type 2 (33), Tarui’s disease or glycogen storage disease type-VII (29), and maternally inherited diabetes and deafness (34) as well as other conditions such as hypertension (35) and obesity (36,37). Although these studies suggest a correlation between the loss of pulsatile insulin secretion and metabolic dysfunction, they are limited in determining the nature and extent of this association.

As a more systematic approach to determine whether a link exists between islet oscillatory capacity and overall islet function, we addressed the matter at the level of the isolated islet. Although it is well accepted that damage (38,39) or disease (40,41) can eliminate islet oscillatory capacity among other deleterious effects, to our knowledge few, if any, studies have directly examined the loss of oscillatory capacity for its effects on other aspects of islet function. Because eliminating oscillations is difficult without specifically blocking processes that are crucial to normal stimulus-secretion coupling (such as blocking glycolysis or K_ATP channels), we addressed this issue by comparing oscillatory and nonoscillatory islets that were isolated from healthy adult mice and cultured under identical control conditions. Among these seemingly healthy islets, 60–70% demonstrated oscillatory capacity.

We tested the hypothesis that the loss of the capacity to generate oscillations in [Ca²⁺]_i is a leading indicator of islet dysfunction by directly comparing oscillatory and nonoscillatory islets at several points along the stimulus-secretion pathway including glucose-stimulated changes in [Ca²⁺]_i, mitochondrial membrane potential, ion channel activity, and ER calcium handling. Our findings indicate that nonoscillatory islets have diminished glucose stimulation and ER calcium storage compared with oscillatory islets. Similar results were found by mildly damaging islets with proinflammatory cytokines, suggesting that the loss of oscillations may serve as an early biomarker for islet dysfunction.

Materials and Methods

Mice and islet isolation

Male CD-1 mice weighing 25–35 g (Charles River Laboratories, Wilmington, MA) were housed in a pathogen-free facility in the Center for Comparative Medicine at the University of Virginia for use in all studies. Mice were euthanized according to Institutional Animal Care and Use Committee-approved protocol, and their pancreatic islets were isolated by collagenase digestion and Histopaque centrifugation using a modified version of a previously published protocol (42). After clamping the common bile duct at the duodenum, the pancreas was perfused through the common bile duct with 5 ml of 1.4 mg/ml collagenase P (Roche Diagnostics, Indianapolis, IN) in fresh Hanks’ balanced salt solution (HBSS; Invitrogen, Carlsbad, CA) with 1% BSA and 4.2 mm sodium bicarbonate. The perfused pancreas was removed and incubated at 37 C for 8–11 min in 1 ml HBSS solution. After incubation, samples were shaken vigorously by hand. Samples were then washed with HBSS solution and pellet was resuspended. Samples were strained through mesh with 35 linear openings per inch. Samples were centrifuged and resuspended in room temperature Histopaque 1077 (Sigma-Aldrich, St. Louis, MO). Equal volumes of HBSS solution were gently added onto Histopaque layer resulting in a discontinuous gradient. Samples were centrifuged for 20 min at 1200 rpm in Centrifuge 5810R (Eppendorf, Netheler, Germany) at room temperature. Supernatant was washed, and islets were transferred to a petri dish containing RPMI 1640 supplemented with 11 mm glucose (Invitrogen). All islets were incubated overnight to allow sufficient recovery time from collagenase digestion before any experiments were performed.

Cytokines and drug treatments

The cytokine combination chosen for this study is used widely as a means of inducing inflammatory responses in islets (43,44,45,46). Although cytokines induce cell death at higher doses, we also observed clear effects on islet function at low enough doses to avoid overt cell death as measured by propidium iodide (PI) staining. Mouse forms of cytokines (B&D Scientific, Franklin Lakes, NJ) were used at the following concentrations: 20 pg/ml for TNF-α, 200 pg/ml for interferon-γ, and 10 pg/ml for IL-1β in PBS. Islets were treated overnight with cytokines after a full day of recovery from the isolation procedure. All experimental tests were performed 2 d after isolation.

Glucose-stimulated [Ca²⁺]_i (GSCa)

[Ca²⁺]_i was measured using the ratiometric [Ca²⁺]_i indicator fura-2/AM using methods modified from those previously described (24). Briefly, islets were dye-loaded and recorded in solution containing the following in millimoles: 11 glucose, 130.5 NaCl, 3 CaCl₂, 5 KCl, 2 MgCl₂, and 10 HEPES, pH 7.3. All islets were maintained in 11 mm glucose at all steps before studies of oscillations to avoid transient shifts in oscillatory patterns during the experiment (47,48). Islets were loaded with 3 μm fura-2 AM (30–40 min), washed, and then transferred to a small volume chamber (Warner Instruments, Hamden, CT) mounted on the stage of a BX51WI fluorescence microscope (Olympus, Tokyo, Japan). Islets were perifused with a peristaltic pump (Gilson, Middleton, WI) at about 35 C by an in-line heater (Warner Instruments). Images were taken sequentially from 340 nm and then 380 nm excitation to produce each [Ca²⁺]_i ratio from emitted light at 510 nm using a Hamamatsu ORCA-ER camera (Hamamatsu Photonics, Hamamatsu, Japan). Excitation light from a xenon burner was supplied to the preparation via a light pipe and filter wheel (Sutter Instrument Co., Novato, CA).

Paired images were recorded every 5 sec for 15 min. After recording [Ca²⁺]_i in 11 mm glucose for 15 min, islets were incubated in 3 mm glucose for 15 min and then recorded to determine islet response to 28 mm glucose stimulation. The glucose-stimulated calcium was defined as the difference in [Ca²⁺]_i levels between 28 vs. 3 mm glucose as measured by fura-2 AM ratio (340/380 nm fluorescence). Data were analyzed with IP Lab software version 4.0 (Scanalytics, Rockville, MD).

Glucose-stimulated insulin secretion (GSIS)

Islets were first assessed for oscillatory capacity as described above. Oscillatory islets were then separated from nonoscillatory islets and transferred by pipette individually from the recording chamber to a 48-well plate for static GSIS measurements. Islets were tested for insulin secretion as described previously (49,50) but using only three islets per replicate. Briefly, islets were preincubated at 37 C and 5% CO₂ for 1 h in a standard Krebs-Ringer bicarbonate (KRB) solution and then washed and incubated in KRB supplemented with 3 mm glucose for 1 h followed by a 1-h treatment with KRB containing either 11 or 28 mm glucose. The supernatant was collected after each treatment and insulin concentration in the supernatant was measured by an EIA method (Mercodia Inc., Uppsala, Sweden) with a mouse insulin standard. The intraassay variation was less than 4% and inter-assay variation was less than 10%.

Mitochondrial membrane potential

Rhodamine 123 (rh123) was used to measure changes in mitochondrial membrane potential as reported previously (51,52). Briefly, islets were loaded with 5 μm rh123 for 15–20 min and then imaged as described above for [Ca²⁺]_i but with single wavelength excitation of 488 nm and 510 nm emission.

Cell death measurements

After each [Ca²⁺]_i recording, islets were treated with 20 μg/ml PI and incubated for 10 min. Islets were imaged once under bright-field illumination to determine the islet borders and imaged again to measure PI fluorescence using 535 nm excitation and 617 nm emission.

Data analysis and statistics

[Ca²⁺]_i recordings were assessed for a variety of oscillatory measurements. Oscillatory capacity was calculated as the percentage of islets with oscillatory activity among all islets recorded. The period of [Ca²⁺]_i oscillations was directly measured from the start of one cycle to the start of the next, and the amplitude of oscillations was measured peak to nadir. [Ca²⁺]_i patterns were also analyzed by the pulse detection algorithm CLUSTER8 (53), as used previously (54) using the following parameters: 10-sec minimum peak and minimum nadir size (2 points), 2.0 for t-score to detect peaks and nadirs, and no minimum value for peak amplitude. False positives were kept less than 5% by using the sd of [Ca²⁺]_i during 3 mm glucose (a nonoscillatory glucose concentration). To improve the detection of possible false negatives, the data were reanalyzed using a t-score of 1, producing the same results. In a comparison of a subset of 14 recordings, oscillations were detected in 11 of 14 records by CLUSTER 8 analysis with mean period of 378 ± 13 sec and oscillatory amplitude 0.33 ± 0.03 (340/380 nm ratio) as compared with a period of 370 ± 10 sec (P = 0.62) and amplitude of 0.30 ± 0.03 (340/380 nm ratio, P = 0.60) by direct measurement. Based on the similar finding, direct measurement was used for the remaining calculations. A two-tailed t test was used for two-group comparisons, with P < 0.05 used as an indication of statistical significance. For the insulin studies, because each trial contained equal numbers of matched oscillatory and nonoscillatory islets, a two-tailed paired t test was used to compare each trial of oscillatory vs. nonoscillatory islets. Statistical analysis was performed using Prism version 4 software (GraphPad Inc., La Jolla, CA).

Results

Oscillatory islets display more robust responses to glucose stimulation

A hallmark of pancreatic islet function is the capacity to generate oscillations. We tested the hypothesis that this activity is important to not only the normal function of the islet, but also that the loss of activity may be a harbinger of islet dysfunction. In this study, all islets were cultured, pretreated, and then recorded in solutions containing 11 mm glucose to avoid disruptions in steady-state oscillatory patterns. As shown in Fig. 1A, two of three islets displayed oscillations, which is representative of the 123 of 198 islets that displayed oscillations, approximately 62% overall among all experiments conducted. All islets, whether oscillatory or not, had relatively smooth surfaces with few shedding cells visible, suggesting little or no physical damage among any of the islets used in these studies. We also compared cell death rates using PI for a subset of islets but found no difference in PI values between untreated oscillatory islets [425 ± 19 arbitrary units (a.u.), n = 14] and nonoscillatory islets (467 ± 40 a.u., n = 16, P = 0.37).

Figure 1.

Oscillatory islets and GSCa. A, Representative examples of calcium patterns during continuous 11 mm glucose exposure. B, GSCa (3–28 mm glucose stimulation) for the same islets as in A. C, Comparison of oscillatory (n = 34) and nonoscillatory (n = 15) islets for mean values of [Ca²⁺]_i in 3 and 28 mm glucose (G). D, Comparison GSCa for oscillatory and nonoscillatory islets. *, P < 0.05. #, Islet number.

To investigate possible differences in glucose sensitivity between oscillatory and nonoscillatory islets, we monitored oscillatory activity in 11 mm glucose for 15 min and then exposed islets to 3 mm glucose for 15 min and recorded islet responses to stimulation with 28 mm glucose. During the 11 mm glucose phase, oscillatory islets maintained lower mean [Ca²⁺]_i (1.23 ± 0.03 ratio 340/380 nm) compared with nonoscillatory islets (1.32 ± 0.02 ratio 340/380 nm, P < 0.05). This difference may be attributed to the nadirs among oscillatory islets reducing the mean [Ca²⁺]_i or perhaps to a left-shifted glucose sensitivity among nonoscillatory islets. As shown in Fig. 1B, the most robust GSCa responses were observed among islets that displayed large amplitude oscillations, whereas the representative nonoscillatory islet (no. 5) displayed the most attenuated GSCa of the three islets shown. Figure 1C shows that oscillatory islets maintained lower mean basal [Ca²⁺]_i in 3 mm glucose compared with nonoscillatory islets (P < 0.05). Mean peak [Ca²⁺]_i also trended lower among nonoscillatory islets in 28 mm glucose, but this was not significant (P = 0.08). As shown in Fig. 1D, however, glucose stimulated calcium response (GSCa, measured by the difference in calcium between 28 and 3 mm glucose) was reduced among nonoscillatory islets stimulated to 28 mm glucose. These data indicate that the overall response to glucose stimulation among oscillatory islets is enhanced or better maintained in comparison with nonoscillatory islets.

We next investigated whether these differences in GSCa translated into differences in insulin secretion. Islets were assessed for oscillatory capacity and then transferred to different wells in groups of three based on whether the islets were oscillatory or not. No difference in islet size was observed among treatment groups as measured by islet area in pixels squared from [Ca²⁺]_i recordings (oscillatory, 17370 ± 1480 pixels²; nonoscillatory 18520 ± 2450 pixels²) GSIS was then measured by exposing each group of three islets to 3 mm glucose solution for 1 h and then 28 mm glucose for 1 h (n = 7). As shown in Fig. 2A, no differences were observed when islets were stimulated to 11 mm glucose (Fig. 2A). Oscillatory islets stimulated to 28 mm glucose appeared to secrete more insulin than the nonoscillatory islets (Fig. 2B), although this was not significant (P = 0.25). We also examined the increase in insulin between basal and stimulated conditions. As shown in Fig. 2C, nonoscillatory islets showed a significant reduction in GSIS in the 28 mm glucose condition (P < 0.05), but no difference was observed in the 11 mm glucose condition (P = 0.60). These data are consistent with our GSCa findings and suggest that nonoscillatory islets are less able to maintain stimulus-secretion coupling, particularly during high-glucose stimulation.

Figure 2.

GSIS. A and B, Insulin secretion during 1 h of treatment in 3 mm glucose (3G) followed by 1 h in 11 mm (11G; A, n = 10 sets of islets) or 28 mm glucose (2B/G, n = 7 sets of islets). C, Stimulation of insulin secretion measured as the difference in insulin secretion between stimulated and basal glucose levels. *, P < 0.05. Osc, Oscillatory; nonosc, nonoscillatory.

Oscillatory and nonoscillatory islets show similar ion channel activity

To address possible mechanisms that could produce the observed differences in GSCa, we first looked at calcium flux through plasma membrane ion channels. Because ion channels are integral to the regulation of insulin secretion and to generating oscillations, we compared ion channel activity of oscillatory vs. nonoscillatory islets. Islets were first recorded under 11 mm glucose to measure oscillatory capacity and then stimulated with 28 mm glucose between recordings to completely saturate the glucose response, thus closing a large majority of K_ATP channels and opening L-type calcium channels, the two most dominant ion channels involved in insulin secretion. Islets were then exposed to 250 μm diazoxide in 28 mm glucose to open K_ATP channels, washed with 28 mm glucose, treated with nifedipine to block L-type calcium channels, and washed again. As shown by representative examples, oscillatory islets (Fig. 3A) and nonoscillatory islets (Fig. 3B) had similar responses to the ion channel modulators (as summarized in Fig. 3C). We also tested islets under basal glucose (3 mm) conditions by adding 30 mm potassium chloride (KCl) as a general depolarizing agent and observed only a trend toward slightly greater stimulation (P = 0.11) among oscillatory vs. nonoscillatory islets as summarized in Fig. 3C. These data suggest that ion channels are not primarily responsible for the differences in GSCa between oscillatory and nonoscillatory islets. The only difference between oscillatory and nonoscillatory islets in these studies was in stimulation from 11 to 28 mm glucose, which is consistent with the observed GSCa data.

Figure 3.

Oscillatory and nonoscillatory islets show similar ion channel activity. A and B, Representative examples of [Ca²⁺]_i records from oscillatory islets (A) and nonoscillatory islets (B) that were raised from 11 to 28 mm glucose (G) to saturate glucose stimulation and then treated with 250 μm diazoxide (Diaz) to open K_ATP channels, washed, and treated with 50 μm nifedipine (Nifed) to block L-type calcium channels. Parallel lines (A and B) indicate a pause in the recording to allow [Ca²⁺]_i to stabilize. C, Mean changes in ratiometric [Ca²⁺]_i during diazoxide and nifedipine treatment (n = 14 oscillatory and n = 27 nonoscillatory islets) and during 30 mm KCl stimulation in 3 mm glucose (3G; n = 19 oscillatory and n = 36 nonoscillatory islets). *, P < 0.01.

Oscillatory and nonoscillatory show similar mitochondrial responses

Pyruvate, a product of glucose metabolism, is metabolized by mitochondria in the tricarboxylic acid cycle to produce the ATP that stimulates insulin secretion. To determine whether mitochondrial activity might explain GSCa differences, we compared oscillatory and nonoscillatory islets for mitochondrial membrane potential changes in response to glucose stimulation and sodium azide, a mitochondrial inhibitor. We used rh123, a small, positively charged fluorescent probe that targets to the highly negatively charged mitochondria (51,52,55). As the mitochondria increase their membrane potential gradient in response to stimuli like glucose, the rh123 concentrates and self-quenches to reduce fluorescence.

Islets were loaded with fura-2 AM and imaged to first identify oscillatory or nonoscillatory [Ca²⁺]_i patterns. A representative example of [Ca²⁺]_i patterns in 11 mm glucose is shown for oscillatory islets in Fig. 4A and nonoscillatory islets in Fig. 4B (top panels). Islets were next exposed to 3 mm glucose for 30 min and then imaged to determine changes in mitochondrial membrane potential in response to glucose and sodium azide exposure as shown in Fig. 4, A and B (bottom panels). As shown in Fig. 4C, mean percent change in rh123 fluorescence levels associated with mitochondrial membrane potential did not differ between oscillatory (n = 55) and nonoscillatory (n = 17) islets in terms of glucose stimulation from basal levels or gradient capacity as determined by response to sodium azide response (P > 0.30). These findings thus suggest that glucose metabolism does not differ between oscillatory and nonoscillatory islets at the level of mitochondrial energy production.

Figure 4.

Oscillatory and nonoscillatory show similar mitochondrial membrane potential changes. A, A representative example of an islet displaying [Ca²⁺]_i oscillations in 11 mm glucose (top) and mitochondrial membrane potential changes after glucose stimulation (3 to 28 mm) and challenge with sodium azide (NaN3) (bottom). B, A representative example of a nonoscillatory islet as described in A. C, Comparison of mean percent changes in rh123 fluorescence between all oscillatory (Osc; n = 55) and nonoscillatory (nonosc; n = 17) islets. No differences were observed. 3G, 3 mM Glucose; 28G, 28 mM glucose.

Nonoscillatory islets demonstrate diminished ER calcium handling

Because the ER is a substantial calcium-storing organelle that is also important to protein synthesis and normal cellular function, we examined the ER as a possible mediator of some of the observed differences in calcium handling. To determine calcium storage in the ER, we used thapsigargin to block sarco(endo)plasmic reticulum calcium ATPase (SERCA) pumps, preventing additional sequestration of calcium in the ER (56). The thapsigargin-induced increase in cytosolic calcium is thus indirectly representative of stored calcium in the ER. As shown in Fig. 5, oscillatory islets showed substantial calcium release in the presence of thapsigargin (Fig. 5A), whereas nonoscillatory islets showed significantly reduced calcium release from the ER (Fig. 5B). Mean differences in thapsigargin-induced calcium responses are shown in Fig. 5C, suggesting that nonoscillatory islets may have disrupted ER calcium storage.

Figure 5.

Comparison of thapsigargin (thaps)-induced calcium release from oscillatory (A) and nonoscillatory islets (B). C, Mean changes [Ca²⁺]_i during each treatment [n = 20 oscillatory (osc) and n = 16 nonoscillatory (nonosc) islets]. 3G, 3 mM Glucose; 11G, 11 mM glucose. *, P < 0.05.

Cytokine treatment reproduces the reduced function of nonoscillatory islets

Proinflammatory cytokines play a key role in the dysfunction and destruction of pancreatic islets. Because cytokines have potent and pleiotropic effects, we used a combination of cytokines at very low concentrations [10–100 times lower than typical doses (43,44,45,46)] that do not induce frank cell death. IL-1β, interferon-γ, and TNF-α are the most abundant inflammatory cytokines from the infiltrating immune cells related to type 1 diabetes (T1D) and inflict direct inhibitory and cytotoxic effects on pancreatic β-cells (43,44,45,46). At this cytokine dose, no significant effect was observed on cell death as measured by mean PI fluorescence intensity [untreated islets: 324 ± 6 a.u. (n = 43 islets), cytokine treated 345 ± 11 a.u. (n = 30 islets), P = 0.13]; however, cell death was significantly increased at 5 times this dose or greater (data not shown).

As shown in Fig. 6, cytokine-treated islets showed a significant reduction in GSCa. Two representative examples of untreated (Fig. 6A) and cytokine-treated islets (Fig. 6B) indicated a significant increase in basal [Ca²⁺]_i (P < 0.001), a decrease in peak calcium after 28 mm glucose stimulation (P < 0.05) as summarized in Fig. 6C, and a decrease in GSCa (untreated, 0.21 ± 0.02; cytokine-treated, 0.075 ± 0.01 ratio 340/380 nm, P < 0.001). Note that the two examples of untreated islets in Fig. 6A showed oscillations, whereas the cytokine-treated islets in Fig. 6B did not (data not shown). Untreated islets also showed substantial thapsigargin-induced calcium release (Fig. 6D), whereas cytokine-treated islets did not (Fig. 6E), indicating cytokines greatly reduce ER calcium storage as demonstrated in Fig. 6F and also by others (46). These cytokine effects were qualitatively very similar to the observations made among nonoscillatory islets (compare Figs. 1C and 6C), suggesting that the loss of oscillatory capacity may be indicative of ER stress and/or other early signs of islet dysfunction.

Figure 6.

Cytokine-induced islet effects are consistent with nonoscillatory islets. A and B, Representative traces of GSCa for untreated (A) and low-dose cytokine-treated islets (B). C, Mean basal [Ca²⁺]_i and glucose-stimulated [Ca²⁺]_i for untreated (n = 49) and cytokine-treated (n = 45) islets. D and E, Thapsigargin-induced ER calcium release from untreated (D) and cytokine-treated islets (E). F, Bar graph of mean calcium change after thapsigargin-induced [Ca²⁺]_i release from untreated (n = 7) and cytokine-treated (n = 10) islets. 3G, 3 mM Glucose; 28G, 28 mM glucose. *, P < 0.05; ***, P < 0.001.

Discussion

[Ca²⁺]_i plays a prominent and well-described role in the coupling of glucose stimulation and insulin secretion in pancreatic islets (57,58). We used [Ca²⁺]_i as a sensitive end point to examine differences between oscillatory and nonoscillatory islets in response to glucose stimulation and found that nonoscillatory islets had significantly elevated basal [Ca²⁺]_i and reduced [Ca²⁺]_i and insulin responses to glucose stimulation. To attempt to identify possible causes of these differences, three key points along the stimulus-secretion pathway were investigated: 1) calcium flux through ion channels, 2) mitochondrial membrane potential, and 3) ER calcium handling. We did not observe significant differences in mitochondrial membrane potential or ion channel activities; however, our data suggest deficiencies in ER function among nonoscillatory islets. Nonoscillatory islets displayed reduced [Ca²⁺]_i response to thapsigargin, suggesting reduced SERCA pump activity and ER calcium storage. We also showed that mildly damaging islets with very low doses of proinflammatory cytokines produced similar effects on oscillatory capacity and ER calcium handling, suggesting that these effects may represent the initial stages of stress-mediated islet dysfunction.

The ER serves several important functions that are crucial to cell survival, including posttranslational modification, folding, and assembly of newly synthesized proteins. Several conditions can perturb ER homeostasis including hyperglycemia and calcium depletion of the ER lumen, which can lead to ER stress, a well-established marker for islet dysfunction related to both T1D and T2D (59,60). Proinflammatory cytokines, for example, have been shown to mediate immune responses and β-cell loss in T1D by inducing severe ER stress through nitric oxide-mediated depletion of ER calcium and inhibition of ER chaperones (59). In the db/db mouse model of type 2 diabetes, islets displayed minimal SERCA activity, diminished glucose-stimulated calcium responses, loss of oscillations, and reduced insulin secretion compared with wild-type controls. Furthermore, islets from wild-type mice that were cultured for several days in the presence of thapsigargin demonstrated dramatically reduced glucose-stimulated insulin secretion and behaved qualitatively similar to islets from db/db mice (40). The Goto-Kakizaki and neonatal streptozotocin models of T2D also showed deficiencies in glucose stimulation, oscillatory capacity, and ER calcium handling (41). These studies thus demonstrate a clear link between ER stress and reduced islet function.

The ER is also a high-affinity and high-capacity organelle for calcium. The ER sequesters calcium when cytoplasmic calcium levels are high and releases calcium when cytoplasmic calcium levels are low (61,62). Calcium is pumped into the ER through two forms of the SERCA pumps (SERCA2b and SERCA3) that are expressed in islets (63,64,65,66). Knockout studies produced embryonic lethal SERCA2(−/−) mice, suggesting a critical housekeeping role for SERCA2b in cell survival (67). In contrast, SERCA3(−/−) mice are viable, but their islets do not display the elevated basal [Ca²⁺]_i observed in our study (65). Dysfunction in SERCA2b but not SERCA3 is thus consistent with our observed rise in basal [Ca²⁺]_i and differences in thapsigargin response among nonoscillatory and cytokine-treated islets. Of interest, proinflammatory cytokines have been shown to reduce SERCA2b expression, raise basal [Ca²⁺]_i, and reduce thapsigargin-induced calcium release in islets in a manner consistent with our data (46). Recent mathematical analysis, however, suggests that SERCA2b inhibition may have only transient effects on cytosolic calcium without additional activation of store-operated currents (68). Thapsigargin can also indirectly activate store-operated channels such as calcium release-activated current and K slow by depleting ER calcium (69,70). A thapsigargin-sensitive plasma membrane sodium-channel has been identified as well (71). Determining the mechanisms underlying these deficiencies in ER calcium handling and the contributions of store-operated channels will require further study.

Although our findings thus suggest a link between ER dysfunction and the loss of oscillatory capacity, determining the causal nature of this link is difficult to directly investigate. Does the loss of periodic activity and quiescence lead to ER dysfunction or do disruptions in normal ER activity eliminate the capacity to generate oscillations? Experimental (11,62,72,73) and mathematical modeling studies (47,68,74) suggest that the ER participates in the processes that generate islet oscillations, and thus, it is possible that ER stress may lead to the observed loss of oscillatory capacity from islets. Rather than eliminate oscillations, however, acute thapsigargin treatment augments the amplitude of oscillations and increases their frequency (62,66,73). Although this suggests that ER calcium flux is not requisite to generating oscillations, these studies were conducted with control islets that normally oscillate, whereas our study specifically separated nonoscillatory from oscillatory islets. As an alternative interpretation, the loss of oscillatory capacity could mean that islets do not have periods of rest between secretory volleys, which may lead to β-cell dysfunction from overstimulation (75,76). In this interpretation, the loss of oscillations is the cause of ER stress and islet dysfunction.

Whereas much of the focus of this discussion has been on the ER, we note that other aspects of islet function were investigated but did not seem to be linked with the loss of oscillations. First, no differences were observed in islet mitochondrial membrane potential, suggesting similar mitochondrial ATP production and use between oscillatory and nonoscillatory islets. Reduced mitochondrial function has been reported using rh123 in comparisons of isolated β-cells with intact islets (52) and among islets from diabetic rodents (77), thus demonstrating the validity of the technique to detect differences in function. It should be noted that rh123 does not directly measure ATP but is rather a reflection of the mitochondrial gradient that drives ATP production (51,55). Direct measurement of ATP/ADP has been used to predict islet transplant success (78), and it is possible that shifts in ADP/ATP could alter intracellular calcium and potentially SERCA function. The lack of any effect on mitochondrial membrane potential in our study thus supports, but cannot confirm, the hypothesis that the mitochondrial function of oscillatory and nonoscillatory islets is similar. Second, we reported a slight trend toward decreased ion channel response to KCl depolarization (P = 0.11) among nonoscillatory islets. This could indicate early signs of islet fatigue, which is consistent with the reduced insulin secretion among nonoscillatory islets in response to 28 mm glucose stimulation. Third, we note that altered communication within the islet due to α-cell dysfunction (21,79,80) or disruption in gap junction communication (81,82,83,84) could also contribute to the loss of oscillations but was not directly investigated. Our intent in this study was to examine the islet as a single microorgan and examine processes that were dysfunctional within the islet as a whole. On balance, although other effects cannot be ruled out, ER calcium handling appears to be the most pronounced and acute effect and thus a prime target for further study and intervention.

In conclusion, our findings suggest that the loss of oscillatory capacity is a highly sensitive biomarker for islet dysfunction, which may be useful for further elucidating early markers in the pathophysiology of diabetes and other metabolic disorders and identifying functional vs. dysfunctional islets for use in human islet transplantation. With regard to islet transplantation, our findings suggest that reducing the metabolic activity of cultured islets may preserve their viability and prolong their function because ER stress is directly related to cellular metabolic demands. This hypothesis is consistent with findings of improved human islet viability under conditions of reduced temperature or glucose concentration in culture (85), which both reduce cellular energy demands. We conclude that elucidating the mechanisms that alter or eliminate islet oscillations may help to identify the early causes of islet dysfunction and destruction and aid in determining the efficacy of treatments designed to improve islet viability.