Regulation of two insulin granule populations within the reserve pool by distinct calcium sources

Pancreatic β cells play an essential role in glucose homeostasis by secreting insulin in response to a variety of stimuli, especially a rise in blood glucose levels. Insulin is stored in large dense-core secretory granules. Although an individual β cell in a mouse islet contains over 10,000 granules (Dean, 1973), only a small fraction (<5%) exist in a functionally defined readily releasable pool (RRP) (Barg et al., 2002). These granules are docked at the plasma membrane (PM) and are released upon stimulation of secretion. By contrast, the remaining granules exist in a nonreleasable state that is commonly referred to as the reserve pool. During glucose-stimulated insulin secretion, a small fraction of granules from the reserve pool must be mobilized to refill the RRP and sustain insulin secretion (Rorsman et al., 2000). However, the mechanisms underlying granule mobilization in these cells are poorly understood.

One important question is whether insulin release and granule trafficking are regulated by the same mechanisms. While granule mobilization and refilling of the RRP are required for sustained insulin release, previous studies have reported that stimulating insulin secretion by Ca²⁺ influx does not stimulate granule movement (Hisatomi et al., 1996; Niki et al., 2003). Insulin release is initiated by glucose metabolism that increases the ATP/ADP ratio, leading to inhibition of the K_ATP channel and subsequent depolarization of the cell membrane. Membrane depolarization by glucose or KCl then results in influx of Ca²⁺ through voltage-gated calcium channels (Berggren and Larsson, 1994). Ca²⁺ influx can also stimulate Ca²⁺ efflux from the endoplasmic reticulum (ER) through a process called Ca²⁺-induced Ca²⁺ release (Lemmens et al., 2001). Despite its importance in glucose-stimulated insulin secretion, the role of Ca²⁺ in granule mobilization remains to be elucidated.

Two types of secretory granule movement have been described in β cells (Ivarsson et al., 2004; Pouli et al., 1998; Varadi et al., 2003), PC-12 cells (Burke et al., 1997; Lang et al., 2000; Pouli et al., 1998) and chromaffin cells (Steyer and Almers, 1999; Tsuboi et al., 2001). Whereas the majority of insulin granules display a random but restricted movement with short displacement, there is a small population of fast-moving granules that rapidly move much longer distances, at velocities of several μm/second. Previous studies have not specifically focused on the regulation of this fast-moving granule population, since their small sample pool showed only a few fast-moving granules. Because the fast-moving granules constitute such a small minority of the reserve pool, a large data set is needed for a quantitative assessment.

To examine the regulation of granule movement in the reserve pool, especially that of the small fast-moving population, we developed a fully automated approach for measurement of granule mobility. Tracking individual granules across an image series is currently done in a semi-automated manner. This approach is usually poor in tracking granules over long distances, limits the analysis to a small fraction of the granules, and is susceptible to bias from the observer-defined granules selected for analysis. We addressed these issues by enabling observer-independent tracking of a large number of fluorescently labeled granules in each cell. We developed custom image analysis software utilizing two different tracking algorithms designed to specifically track each type of granule movement (Li et al., 2004). Our analysis pool of over 20,000 granules allowed us to quantify population dynamics and examine potential mechanisms.

In this study, we consider how granule mobilization from the reserve pool may contribute to the refilling of the RRP. Our analysis quantifies two distinct populations of insulin granules within the reserve pool identifiable by their relative mobility. We aimed to explore two questions: (1) are these two populations differentially regulated?; (2) what is the significance of the fast-moving granules? Our results show that the two populations respond differently to stimulation by insulin secretagogues and are regulated by separate Ca²⁺ sources. Extracellular Ca²⁺ influx activates a small population of fast-moving granules; intracellular Ca²⁺ efflux from the ER initiates mobilization of the larger population of slow-moving granules. This differential regulation of the two granule populations appears to utilize localized distribution of spatially distinct Ca²⁺ pools. We also show that both granule populations require actin rearrangement and microtubule integrity for movement. Finally, our data are consistent with the model that the fast-moving granules act as an intermediate pool to enable refilling of the RRP.

Fluorescent protein expression vectors were obtained from BD Biosciences (Palo Alto, CA). DNA isolation reagents were from Qiagen (Valencia, CA). Restriction enzymes were obtained from New England Biolabs (Beverly, MA). PCR primers were from Integrated DNA Technologies (Coralville, IA). FluoZin-3, tubulin antibody, jasplakinolide, Alexa Fluor 488 phalloidin and fluorescently labeled secondary antibodies were obtained from Molecular Probes (Eugene, OR). Guinea pig anti-insulin was obtained from Linco Research, Inc. (St Charles, MO). The phogrin construct was a generous gift from John Hutton (University of Colorado, Denver, CO). S-(–)-1,4-dihydro-2,6-dimethyl-5-nitro-4-[2-(trifluoromethyl)phenyl]-3-pyridinecarboxylic acid methyl ester (Bay K8644), 1,4-dihydro-2,6-dimethyl-4-(3-nitrophenyl)-3,5-pyridinedicarboxylic acid 2-methoxyethyl 1-methylethyl ester (nimodipine), and all other chemicals were from Sigma-Aldrich. Cell culture reagents were made by the Media and Reagents Core of the Diabetes Research and Training Center at Vanderbilt University.

βTC3 cells secrete insulin in a regulated manner very similar, but not identical, to that of intact β cells (D'Ambra et al., 1990). βTC3 cells were maintained in sodium bicarbonate-buffered Dulbecco's modified Eagle's medium (DMEM) containing 15% horse serum and 2.5% FBS, 5 mM glucose, 100 IU/ml penicillin, and 100 μg/ml streptomycin (Life Technologies, Inc.) in an atmosphere of 5% CO₂. MIN6 cells were cultured in similar DMEM supplemented with β-mercaptoethanol and 15% heat inactivated FBS. Generation of phogrin-EGFP was as previously described (Emmanouilidou et al., 1999). Proinsulin-EGFP was made from proinsulin-ECFP (Rizzo et al., 2002) by substitution of the EGFP coding sequence of pEGFP-N3 for ECFP using BamHI and BsrGI restriction sites. The proinsulin-EYFP-DsRed construct was also made from proinsulin-ECFP by substitution of the EYFP coding sequence from pEYFP-N3 for ECFP using the BamHI and BsrGI restriction sites. The cDNA sequence encoding DsRed (pDsRed-C1) was then amplified by PCR (sense primer: 5′-TAGGTACCATGGTGCGCTCC-3′; antisense primer 5′-ATGGGCCCCTGAGCAGGAAC-3′) and inserted C-terminal to EYFP using KpnI and ApaI restriction sites. ECFP-ER was from BD Biosciences. Plasmid DNAs were introduced into βTC3 and MIN6 cells suspended in Dulbecco's PBS in a 2-mm gapped cuvette by ten 50-μsecond square-wave pulses of 300 V at 500-msecond intervals with a BTX ECM830 electroporator (Holliston, MA). Transfected cells were plated on glass-bottomed coverslip dishes (MatTek Corp., Ashland, MA) in regular growth medium and switched to growth medium containing 2 mM glucose for 48 hours. Cells stably transfected with phogrin-EGFP were generated using G418 (Mediatech, Inc., Herndon, VA). Cells were equilibrated in BMHH buffer (125 mM NaCl, 5.7 mM KCl, 2.5 mM CaCl₂, 1.2 mM MgCl₂, and 10 mM Hepes, and 0.1% BSA, pH 7.4) for 4 hours (basal, unstimulated condition) prior to microscopy.

Confocal fluorescence microscopy was performed using an Axiovert 100M inverted microscope equipped with an LSM 510 laser scanning unit and a 63× 1.4 NA plan Apochromat objective (Carl Zeiss, Inc.). A 40× 1.3 NA plan Apochromat objective was used for FluoZin-3 experiments. 457, 488 and 514 nm argon and 543 nm helium-neon laser lines were used to excite ECFP, EGFP/Alexa Fluor 488/FluoZin-3, EYFP, and DsRed, respectively. Emitted light was passed through bandpass filters for collection of ECFP (470-510 nm), EGFP/Alexa Fluor 488/FluoZin-3 (505-530 nm), EYFP (530-550 nm) and a long-pass filter (560 nm) was used for DsRed.

The secretion assay was performed with a Nikon TE300 inverted wide-field microscope equipped with a 40× 1.3 NA plan Apochromat objective using a standard GFP filter combination. Images were acquired with MetaMorph imaging software (Universal Imaging Corp., Downingtown, PA).

Cells were first fixed for 30 minutes with 4% paraformaldehyde in PBS at 4°C, permeabilized with 0.1% Triton X-100, and blocked with 5% goat serum. For insulin detection, they were then stained with guinea pig anti-insulin and Alexa Fluor 546-conjugated goat anti-guinea pig antibodies. Mouse α-tubulin and Alexa Fluor 488 goat anti-mouse antibodies were used to detect microtubule distribution. Alexa Fluor 488 phalloidin was used to identify actin microfilaments.

Sequential images of βTC3 cells expressing phogrin-EGFP were taken for 2-6 minutes at 1-second intervals. A minimum laser power that would give us a useful florescence signal was used and the laser intensity was kept the same for all cells. Cellular phototoxicity caused by laser illumination was tested by mimicking acquisition conditions used in real experiments but using the buffer alone as the stimulus. Cells were first imaged for 2 minutes to establish the unstimulated profile. A 20× concentrated stock of stimulus was then added to the cells on the microscope stage. A final concentration of 20 mM glucose, 30 mM KCl, 1 μM Bay K8644, 5 μM nimodipine, 2 μM thapsigargin, 10 mM caffeine, 10 μg/ml cytochalasin D, 1 μM jasplakinolide and 10 mM nocodazole was used. Cells were allowed 2 minutes to react to glucose before the stimulated profile was obtained. In the case of nocodazole, cytochalasin D, jasplakinolide and thapsigargin, cells were treated for 20 minutes and a profile taken before stimuli were added. All manipulations, as well as fluorescence microscopy, were done on the temperature-controlled microscope stage maintained at 37°C using the Zeiss stage incubator.

Summation projection of all background-corrected confocal slices was produced using the MetaMorph imaging software (Universal Imaging Corp., Downingtown, PA). The degree of bleed-through and colocalization between the channels was determined as described previously (Mallet and Maxfield, 1999; Mukherjee et al., 1999). To obtain parameters for granule movement, images were first background corrected (Hao and Maxfield, 2000) using MetaMorph before running through imaging analysis software developed by Li et al. (Li et al., 2004). The individual granules are identified automatically by filtering the images to reduce the noise with wiener and median filters, thresholding the images with a combination of optimum single and double threshold algorithms, and further eliminating spurious points with morphological operations. Two different tracking algorithms were used to track each type of granule movement. Each granule was first subjected to a simple tracking algorithm that detected the overlapping regions between successive frames. If such a region was found, the algorithm would record the granule's position in the current frame. This algorithm worked well for the granules that moved short distances. If no overlapping region was found, the second tracking algorithm was activated. Briefly, the program calculates parameters about each detected vesicle in the first frame, including size, brightness, and position. Next, a simple linear model is used to predict the position of the vesicle in the next frame. Then, the image is segmented locally to find possible granules within a box placed around the predicted location. When several potential granules are detected within the box the most likely candidate is chosen based on its size, intensity, and position. When a granule is lost, the tracking program performs analysis only for the frames in which it showed a path. Before the program was applied to all the granules, the output from our tracking software was first manually inspected and then compared with that from Metamorph to ensure accuracy. Our tracking program out performs others by the fact that it not only traces granules with high accuracy, it can track many granules simultaneously and automatically.

βTC3 cells were plated at a low density and cultured for 3 days after transfection. A cluster of cells was used to produce a detectable amount of secreted insulin and Zn²⁺. Two identical groups of cells were used for this experiment. One group was used to monitor insulin release using sequential images taken in the presence of 2 μM FluoZin-3 before and after adding glucose. The cells were then washed three times and incubated for 1 hour in glucose-free buffer, which was then replaced with fresh glucose-free buffer containing FluoZin-3. A time series was taken after glucose washout. As a control for autofluorescence and focal plane drift, fluorescence in the extracellular medium was monitored in the absence of FluoZin-3 while glucose was added as before. Fluorescence was quantified in three small regions in the extracellular medium as determined by the differential interference contrast image. The other group of cells, transfected with phogrin-EGFP, was used to monitor the granule movement. Time series of granule movement were first taken before and after adding glucose, and again after glucose wash-out. All time series were acquired for the same field of cells.

Cells were transfected with proinsulin-EYFP-DsRed. One image was taken of EYFP and DsRed for the pH measurement of granules. The two channels were scanned alternately in a line-by-line fashion, having only one laser line and one detector channel on at each time. Focusing was done using the DsRed channel, which would result in ∼0.1% photobleaching of DsRed (Baird et al., 2000). A time series was then acquired with only the DsRed channel to measure granule movement. To verify pH dependency of proinsulin-EYFP-DsRed, cells expressing the construct were fixed with paraformaldehyde and permeabilized with Triton X-100. In situ images were taken of cells incubated with PBS and adjusted to pHs from 7.5 to 5.5 using predetermined amounts of HCl. The fluorescence ratio of EYFP/DsRed at each pH was normalized to that at 7.5 in order to pool the results from different cells. Furthermore, raising the pH with NH₄Cl caused an increase in the fluorescence ratio of EYFP/DsRed in live cells. Images of the same field of cells were taken using the same settings before and after 30 mM NH₄Cl was added. Fluorescence intensity was normalized to that of the images before NH₄Cl was added.

Visualization of β cell granules has been greatly enhanced by the utilization of fluorescent proteins targeted to secretory granules. Phogrin (phosphatase on granules of insulinoma cells) is well established for the study of granule dynamics in pancreatic β cells because of its specialized localization to the insulinoma dense-core granule membranes (Pouli et al., 1998; Wasmeier and Hutton, 1996). We fused phogrin with enhanced green fluorescent protein (EGFP). To examine whether phogrin-EGFP, in a stably transfected βTC3 cell line, was targeted to insulin secretory granules, the localization of phogrin-EGFP was compared with that of immunostained insulin. Fig. 1A-C shows that most phogrin-EGFP-labeled granules contained insulin. In order to avoid artifacts created by the immunofluorescence procedure, we also looked at live cells containing both phogrin-EGFP and an enhanced cyan fluorescent protein (ECFP) construct that was targeted to the interior of secretory granules by the connecting peptide segment of the murine proinsulin II (proinsulin-ECFP) (Watkins et al., 2002; Rizzo et al., 2002). As shown in Fig. 1D-F, phogrin-EGFP colocalized very well with the co-transfected proinsulin-ECFP, with >95% of the phogrin-EGFP-labeled granules also containing proinsulin-ECFP. These results indicate that most of the phogrin-EGFP expressed in βTC3 cells was efficiently targeted to the insulin secretory granules.

Insulin is released from β cells in response to various extracellular stimuli. Of the secretagogues used in this study, KCl causes the largest stimulation of insulin secretion in βTC3 cells and also the most dramatic change in intracellular granule movement. Sequential images of cells expressing phogrin-EGFP were taken to examine granule dynamics in pancreatic β cells. Fig. 2 shows three frames from a representative set of sequential images taken of the same cell. Several granules were selected from each panel (Fig. 2A-C) and their positions were tracked through 30 frames (the corresponding lower panels in Fig. 2D-F). The accompanying time-lapse movie can be seen in the supplementary material. Most granules display small, confined movements under basal, unstimulated conditions (Fig. 2D). Adding KCl caused a slowing down in the overall granule dynamics (Fig. 2E). However, a small population (∼8%) of granules underwent much longer excursions (Fig. 2F). Similar results were obtained using another β cell line, MIN6 (Table 1). Table 1 lists the changes in granule mobility under all the experimental conditions used in this study.

Our granule tracking analysis shows that the two subsets of granules respond differently to stimulation. Two distinct types of motion were observed with KCl (Fig. 2E,F). In order to take a closer look at each population, we used histograms to segment the granules according to their movement (Fig. 3). The peaks in the histograms represent the slow-moving population. The percentage of fast-moving granules is shown as bar graphs in the inset of each panel. For this study, the fast-moving population was defined as granules with speeds >0.4 μm/second, which was at least two standard deviations above the average speed (118.3±10.6 to 286±27.5 nm/second) under all conditions. In unstimulated cells, the fast-moving population accounted for only 2.1±0.3% of the total granules. This number increased to 5.8±0.6% and 8.5±0.7% when glucose and KCl, respectively, was added (Fig. 3A,B, Table 1). To see if glucose had any further effect on the two granule populations segregated by KCl, we included glucose before and after KCl was added (Fig. 3C,D). Although an increase in the speed of the slow-moving granules was seen when glucose was present, the fraction of fast-moving granules did not increase (Fig. 3C inset, red vs blue bars; compare Fig. 3D and B inset red bars, Table 1). These results indicate that conditions other than ATP production are required to activate a large fraction of granules to a more mobile state under KCl stimulation.

In addition to the average velocity, we characterized granule motion using net displacement and mean-square displacement (MSD). First, we wanted to make sure that the granules with higher velocities (fast-moving granules) were indeed the ones that showed greater net displacements. As shown in Fig. 3E, there was correlation of speed and net displacement for granules undergoing diffusion. More importantly, this panel shows that fast-moving granules were more likely to experience directed motion. To confirm this, we selected 20 granules from each of the three speed intervals: <0.2 μm/second, 0.2-0.4 μm/second and >0.4 μm/second, and characterized their motion using MSD. Granule movement could be classified into three types (Ivarsson et al., 2004): caged motion, in which the MSD values rapidly reached a plateau for longer time intervals (represented by curve i in Fig. 3F), random diffusion, in which the MSD values were fitted to a linear function (Fig. 3F, curve ii), and directed motion, in which the MSD values were fitted to a second degree equation (Fig. 3F, curve iii). Table 2 shows that the slowest granules largely exhibited caged motion and the fast granules mostly displayed directed movement. These additional analyses show that the granule movement could be quantified using the average velocity.

It has been shown that insulin granule exocytosis occurs by complete fusion and that direct recycling of granules occurs only rarely (Ma et al., 2004). Using evanescence microscopy, it was found that there was a significant decrease in phogrin-EGFP fluorescence when phogrin-EGFP labeled granules interacted with the PM (Tsuboi et al., 2000). These data suggest that the majority of the granules in our analysis have not undergone exocytosis. To rule out the possibility that granule `kiss and run' occurrence (Tsuboi and Rutter, 2003) could affect our characterization of the granule movements, we used proinsulin-EGFP to image the insulin granules. Unlike the membrane-bound phogrin-EGFP, proinsulin-EGFP was released when insulin granules interacted with the PM, leaving the recycled granules non-fluorescent. Using this construct, which excluded granules having already undergone exocytosis, we were able to obtain data on glucose- and KCl-stimulated granule movement very similar to that obtained with phogrin-EGFP (data not shown).

Most of the granules imaged in this study were situated near the PM adherent to the coverslip. To confirm that the granule movement in this plane was representative of the entire cell, we looked at the granules in the cell center (Table 1). The overall average speed was higher (17.8±2.0%) under unstimulated condition compared with granules at the PM. Unlike granules near the PM, long-distance moving granules at the cell center were only weakly activated by KCl (1.3±1.2% vs 6.4±1.0% increase, after KCl treatment, of the fast-moving population in the center and at the PM, respectively). This result suggests that granule mobilization upon KCl stimulation does not occur uniformly in the entire cell and that there is preferential activation at the cell periphery.

Although two modes of granule movement have been described previously, the mechanisms regulating these motions are unclear. While glucose had a small stimulatory effect on both populations of granules, KCl significantly activated the fast-moving granules (Fig. 3A,B). Data in Fig. 3C,D also show that adding glucose with KCl further stimulated only the slow-moving granules. These results would be expected if the fast-moving granules were more critically dependent on Ca²⁺ than slow-moving granules. This is because high concentrations of KCl are known to induce a much larger and more rapid Ca²⁺ influx than glucose (Graves and Hinkle, 2003), and although it increases insulin secretion, raising glucose levels in the presence of high KCl does not further elevate the intracellular Ca²⁺ concentration (Henquin et al., 2002). We, therefore, hypothesize that intracellular Ca²⁺ may differentially regulate the trafficking of the two granule populations.

To test this idea, we first looked at changes in granule motion when cytosolic Ca²⁺ levels were altered. Bay K8644 is an established L-type calcium-channel activator that increases the mean open time and opening probability of the channels in a variety of cells (Schramm et al., 1983), including pancreatic β-cells (Larsson-Nyren and Sehlin, 1996; Roe et al., 1996; Smith et al., 1989). Nimodipine, which potently inhibits L-type calcium channels, has been used to prevent the depolarization-induced Ca²⁺ rise in β cells (Garcia-Barrado et al., 1996). When cells were pretreated with these reagents that interacted with the voltage-dependent calcium channels, the largest changes in granule movement were seen in the fast-moving granules (Fig. 4A,B insets). This population of granules significantly increased with Bay K8644 and decreased with nimodipine. Much smaller changes were observed for the slow-moving population (Fig. 4A,B). These results indicate that the two types of granule motion respond differently to changes in cytosolic Ca²⁺ and that the fast-moving granules are closely regulated by Ca²⁺ influx from the L-type calcium channels.

We next examined the other major intracellular Ca²⁺ source. Upon stimulation, cytosolic Ca²⁺ increases as a result of Ca²⁺ influx from extracellular media and Ca²⁺ efflux from intracellular stores (Rojas et al., 1994; Theler et al., 1992). We tested the role of ER Ca²⁺ stores in regulating insulin granule movement. Thapsigargin, an inhibitor of ER Ca²⁺-ATPases (Islam and Berggren, 1993), is often used to block intracellular Ca²⁺ pumps. Under conditions that depleted the ER Ca²⁺ stores, thapsigargin treatment caused a significant decrease in granule movement, even in the presence of glucose (Fig. 4C; Table 1). However, pretreating the cells with thapsigargin had little effect on the stimulation of fast-moving granules by KCl (compare Fig. 4C and Fig. 3D inset red bars, Table 1, P>0.1). To elicit the opposite effect of thapsigargin, caffeine was used to stimulate Ca²⁺ release from the ER by activating ryanodine receptors located on the ER membranes (Islam et al., 1998). Caffeine alone produced a larger stimulatory effect than glucose on the slow-moving population, but failed to activate the fast-moving population (Fig. 4D, Table 1, P>0.1). Similar to thapsigargin, caffeine treatment did not affect the fast-moving population when KCl was added (Fig. 4D vs Fig. 3B inset red bars, Table 1, P>0.1). These results indicate that Ca²⁺ efflux from the ER affects the majority of insulin granules (slow-moving population), whereas the small percentage of granules that move long distances (fast-moving population) are regulated independently of the ER Ca²⁺ efflux.

To explore how the two granule populations are regulated by separate Ca²⁺ pools, we investigated the granule movements with respect to their proximity to the PM or the ER, since it has been shown that upon KCl stimulation, granules close to the plasma membrane experience a greater increase in Ca²⁺ at their surface than the ones deeper inside the cell (Emmanouilidou et al., 1999). Several attempts were made to measure the distances between the granules and the PM. However, it became apparent to us that it was technically difficult to determine the precise location of the PM because of its complex three dimensional structure. However, we could obtain reliable and reproducible measurements of the position of each granule in relation to the ER by the fluorescence intensity of the co-transfected ECFP-ER at that location. ECFP-ER contains both the targeting and retrieval sequences for the ER (Roderick et al., 1997). Fig. 4E shows the distance between a granule and the ER (indicated by ECFP-ER fluorescence) versus its speed of movement upon glucose stimulation. Relative to the other granules in the reserve pool, the fast-moving granules, shown in red, are farthest from the ER and thus least sensitive to Ca²⁺ efflux from the ER. This result agrees with our data showing that the fast-moving granules are less regulated by ER Ca²⁺ stores than the slow-moving population (Fig. 4C,D insets). Also consistent with this hypothesis is the observation that KCl has a smaller stimulatory effect on the fast-moving population of granules located in the cell center, away from the site of extracellular Ca²⁺ influx (Table 1). When the same type of experiment shown in Fig. 4E was performed using ECFP-Golgi instead of ECFP-ER, no correlation was found between the granule speed and the distance to the Golgi (Fig. 4F).

Since actin rearrangement is a Ca²⁺-mediated process (Gilman and Mattson, 2002) that has been shown to increase insulin secretion (Thurmond et al., 2003; Wilson et al., 2001), we wanted to assess the role of the cytoskeleton network in insulin granule mobility. We first verified, in βTC3 cells, the minimal dosage of the two drugs used that was needed to disrupt the actin network. Fig. 5 shows F-actin fluorescently labeled with phalloidin and treated with either cytochalasin D, which disrupts actin filaments and inhibits actin polymerization (Fig. 5C,D), or jasplakinolide, which binds to F-actin and prevents depolymerization (Fig. 5E,F) (Wilson et al., 2001). Cytochalasin D decreased basal movement of the granules, but significantly enhanced granule movement upon KCl addition (Fig. 5G, Table 1). Jasplakinolide, by contrast, reduced granule movement under both basal and KCl-stimulated conditions, as well as completely blocked the long-distance migration of the granules (Fig. 5H, Table 1). These data show that actin rearrangement is required for the trafficking of both granule populations upon stimulation.

To examine whether microtubule formation had a role in the differential regulation of the two granule populations, we disrupted the microtubules with nocodazole in βTC3 cells prior to looking at granule movement (Fig. 6A-D). Nocodazole had an inhibitory effect on the granule movement under basal conditions, causing it to drop 20.8±2.6%. The more striking effect was the inability of KCl to activate the long-distance moving granules in nocodazole-treated cells (Fig. 6E, Table 1). This result indicates that both populations of granules are actively transported on microtubules, consistent with the idea that the structural integrity of microtubules is important for sustained insulin secretion (Varadi et al., 2002a).

Granule mobilization has been implicated as a major event leading to insulin secretion. To see if the fluorescently labeled granules undergo exocytosis when exposed to glucose or KCl, we monitored the disappearance of proinsulin-EGFP. Proinsulin-EGFP is released when insulin granules fuse with the PM, thus the fluorescence of proinsulin is lost. KCl and glucose caused a 2.2±0.2% and 3.1±0.3% decrease in fluorescence, respectively, 5 minutes after they were added (Fig. 7A, blue and red lines). Pretreating the cells with nocodazole severely impeded granule release after the initial exocytosis (Fig. 7A, black line), consistent with our observation that granule movement was significantly reduced in nocodazole-treated cells (Fig. 6E). Our observation also supports the notion that granule mobilization is required for sustained insulin secretion.

We hypothesized that the fast-moving granules act as an intermediate pool between the RRP and the traditionally defined reserve pool, and may contribute to refilling the exocytosed granules from the RRP. To see whether there is a correlation between the fast-moving granules and insulin release, we measured the size of the fast-moving population under basal and stimulated conditions. Zn²⁺ indicators (such as FluoZin-3) have been successfully used to study insulin release in pancreatic β cells because insulin and Zn²⁺ are co-stored in secretory vesicles and co-released by exocytosis (Gee et al., 2002). The FluoZin-3 signal does not accumulate, consistent with diffusional dilution of released Zn²⁺ (Qian et al., 2000). Granule movement and Zn²⁺ secretion were analyzed under three conditions, i.e. basal unstimulated, stimulated with glucose, and back to basal condition after glucose washout. Fig. 7B shows a synchronized change in the fast-moving population (blue bars) and insulin release (indicated by FluoZin-3 in red), suggesting that the fast-moving granules are activated upon insulin secretion when there is a need to replenish the granules depleted from the RRP, and are reduced when insulin secretion decreases.

To further test if the fast-moving granules could be responsible for refilling the RRP, we examined the pH values of the insulin granules using a pH sensitive construct. We took advantage of the fact that the brightness of enhanced yellow fluorescent protein (EYFP) is highly dependent on pH (Llopis et al., 1998), while the red fluorescent protein (DsRed) fluorescence is relatively resistant to pH (Baird et al., 2000). The dependence of the fluorescence ratio of proinsulin-EYFP-DsRed is demonstrated in Fig. 7C. pH titration of the construct shows the fluorescence ratio of EYFP to DsRed decreases as the pH becomes acidic. Adding a permeant base, NH₄Cl, caused a pH rise, and resulted in an increase in the fluorescence ratio. Fig. 7D shows that there is an overall inverse relationship between the pH of a granule and its speed of movement. The fast-moving granules, shown in red, are the most acidic in the reserve pool. Since granule acidification has been shown as an important step during the preparation of granules for exocytosis (Barg, 2003; Hutton, 1989), our data show that the fast-moving population is well suited to becoming release-competent upon stimulation.

In this manuscript, we have reported a mechanism whereby secretory granules from the reserve pool can replenish the RRP using the same Ca²⁺ influx that causes insulin release. One of the strengths of this study lies in our ability to automatically segment and track a large number of granules (over 20,000). No prior work in the literature documents such a large scale automated individual granule tracking. Our custom-developed image analysis software (Li et al., 2004) enabled us to assay quantitatively the small – yet important – fast-moving population within the reserve pool. This subset of granules has been described previously with limited quantification (Ivarsson et al., 2004; Lacy et al., 1975; Pouli et al., 1998; Somers et al., 1979; Tsuboi et al., 2000; Varadi et al., 2003). Owing to limited sample size, it was concluded that the intracellular movement of insulin granules was regulated separately from insulin exocytosis (Hisatomi et al., 1996; Niki et al., 2003). We found this to be true for the slow-moving population, which accounts for over 90% of the granules. However, Ca²⁺ influx also plays a role in insulin granule trafficking through its involvement in the fast-moving granules. Our study, therefore, provides evidence that both insulin secretion and granule trafficking respond to a central Ca²⁺-dependent regulatory machinery in pancreatic β cells. Furthermore, our data indicate that fast-moving granules may contribute to the refilling of the RRP (Fig. 7), suggesting that refilling of the RRP could be regulated by the same mechanism as insulin secretion.

Granule populations within the β cell are not yet clearly defined but they have often been described in terms of the RRP and the reserve pool. The RRP remains very small even under maximal stimulatory conditions (Eliasson et al., 1997). Interference with the refilling of the RRP has been suggested to contribute to the secretory defect of type II diabetes (Rorsman et al., 2000). The sustained phase of insulin secretion involves the recruitment of granules from an intracellular site (Varadi et al., 2002b). We propose that the fast-moving population serves as an intermediate pool that provides a critical means for replenishing secreted granules in the second phase of insulin secretion. We show that fast-moving granules are activated at time points after initial insulin secretion (Fig. 7B). Furthermore, the relatively acidic pH values of the fast-moving granules are closely associated with those of the granules in the RRP (Fig. 7D). Our data strongly suggest an essential role of the fast-moving population in the translocation of insulin granules from the reserve pool to the PM. Nevertheless, our study does not rule out the possibility that the slow-moving granules may also play a role in refilling the RRP. This population of granules could contribute to diffusional granule mobility and facilitate redirection and switching between different microtubules before embarking on a directed movement (Ivarsson et al., 2004). However, our data, and that in the literature, do not support the notion that insulin granules are able to reach the PM by mere diffusion, especially over long distances. As pointed out by Ivarsson et al. (Ivarsson et al., 2004), granule diffusion is restricted within functional `cages' of ∼0.9 μm diameter. Another study also reported that there is a lack of vesicle movement in space not occupied by microtubules, and suggested that free diffusion plays a minimal role in long-distance transport (Varadi et al., 2003).

A major question in cell biology is how one signaling molecule such as Ca²⁺ can activate different mechanisms to control many diverse processes. Here, we provide an example of two granule populations regulated preferentially by separate cytosolic Ca²⁺ pools (Fig. 4). Our observations indicate that while the fast-moving population is activated primarily by Ca²⁺ influx through voltage-gated calcium channels on the PM, the slow-moving population is regulated by Ca²⁺ efflux from the ER through Ca²⁺-induced Ca²⁺ release. We also show that the intracellular localization of insulin granules in relation to the PM and the ER is probably the underlying mechanism of such differential regulation by Ca²⁺ (Fig. 4C). It has been shown that Ca²⁺ distribution resulting from extracellular influx and from intracellular efflux is different (Martin et al., 1997; Theler et al., 1992), and that there exist steep spatial gradients of Ca²⁺ within the β cell (Ammala et al., 1993). This heterogeneous distribution of cytosolic Ca²⁺ from the two different sources has been speculated to exert distinct and co-operative influences on the β cell secretory machinery (Niki, 1999). The spatially separate Ca²⁺ pools, such as the glucose-induced microgradients of Ca²⁺ localized just beneath the PM of the β cell (Martin et al., 1997), could provide high Ca²⁺ concentrations locally for stimulation of subsets of granules. Indeed, Ca²⁺ measurements at the surface of β cell granules showed that a small population of granules located close to the PM displayed a greater Ca²⁺ concentration at the granule surface compared with granules located farther away from the PM (Emmanouilidou et al., 1999). Differential Ca²⁺ signaling caused by extracellular influx and intracellular efflux has been shown to contribute to the release of IL-1 β and IL-1 α from macrophages (Brough et al., 2003), and replenishment of two synaptic vesicle pools at the neuromuscular junction are also separately mediated by Ca²⁺ influx and efflux (Kuromi and Kidokoro, 2003).

Current understanding is limited as to how Ca²⁺ may regulate insulin granule traffic at the molecular level. Granule movement is believed to involve protein phosphorylation by Ca²⁺-dependent protein kinases (Ashcroft, 1994). It is speculated that through the activity of Ca²⁺/calmodulin-dependent protein kinases (Gromada et al., 1999), most likely myosin light chain kinase in β cells (Iida et al., 1997), that energy is generated for granule traffic (Niki, 1999). The microtubule-associated protein MAP-2 (Krueger et al., 1997) and the actin-binding protein synapsin I (Krueger et al., 1999) are both substrates for Ca²⁺/calmodulin-dependent protein kinase II in β cells. In addition, activation of granule mobilization is impeded when Ca²⁺/calmodulin-dependent protein kinase II is inhibited (Gromada et al., 1999).

The cytoskeleton network plays a key role in insulin transport. It has been shown that actin remodeling is necessary for glucose-stimulated insulin secretion (Li et al., 1994; Wilson et al., 2001). However, whether actin rearrangement also acts to facilitate the directed movement of the fast-moving granules has not been tested. We show that transient reduction of actin filaments activates both types of granule movement and in turn promotes refilling of the RRP, leading to greater insulin secretion. F-actin is a target of Ca²⁺-dependent signaling cascades (Staiger and Franklin-Tong, 2003) and increases in cytosolic Ca²⁺ stimulate actin depolymerization (Gilman and Mattson, 2002; Staiger and Franklin-Tong, 2003). Our data, along with these reports, suggest that glucose and other secretory stimuli exert their effects on granule trafficking partly through Ca²⁺-mediated actin rearrangement. Microtubules have been shown to be involved in the recruitment of secretory vesicles to the PM (Varadi et al., 2002b). We show that both types of granule movement are affected by the nocodazole treatment, indicating that both populations of granules are attached to the microtubules. Their intracellular movement reflects either microtubule remodeling or active transport on the microtubules.

In summary, the use of live cell imaging and advanced image analysis reveals a small population of insulin granules within the reserve pool that is highly mobilized upon stimulation of secretion. This fast-moving population, which is probably responsible for refilling the RRP, accounts for less than 10% of granules. A detailed regulatory mechanism can be deduced only when a vast pool of granules is examined. Unlike conclusions from other studies, we have now shown a potential mechanism by which the refilling of the RRP is controlled by the same processes that are central to glucose-stimulated insulin secretion.

We would like to thank J. C. Hutton for the phogrin construct, L. Ballester and J. Veale for their contribution to this work, and S. C. Gunawardana and L. Sethaphong for helpful discussions. This work was supported by NIH Grants DK53434 and GM72048 (D.W.P.), US NIH Research Service Awards DK60275 (M.A.R.) and DK59737 (J.V.R.). Software development was supported by the Vanderbilt Advanced Computer Center and NIH grant LM07613.