A Highly Ca²⁺-sensitive Pool of Granules Is Regulated by Glucose and Protein Kinases in Insulin-secreting INS-1 Cells

Glucose triggers insulin release from pancreatic β cells by closing ATP/ADP-gated K⁺ channels, which leads to membrane depolarization, Ca²⁺ influx through voltage-gated Ca²⁺ channels, and [Ca²⁺]_i-triggered exocytosis of insulin-containing granules. Glucose also promotes exocytosis from β cells independently from effects on K⁺(ATP) channels (Eliasson et al., 1997; Takahashi et al., 1999; Rosengren et al., 2002). For example, elevated ATP levels may increase the number of vesicles in a readily releasable state. Activation of protein kinases such as cAMP-dependent protein kinase (PKA; Jones et al., 1986; Ammala et al., 1993a; Gillis and Misler, 1993; Rosengren et al., 2002) and PKC (Jones et al., 1985; Ammala et al., 1994) modulate Ca²⁺-triggered exocytosis from β cells and thus provides a mechanism for hormonal modulation of insulin secretion through second messengers (for review see Jones and Persaud, 1998).

A combination of patch-clamp techniques with photorelease of caged Ca²⁺ allows the Ca²⁺ sensitivity of exocytosis to be directly measured. With these techniques, individual steps in the stimulus–secretion cascade can be isolated so that the detailed mechanisms whereby second messengers modulate exocytosis can be determined. Previous studies using caged Ca²⁺ in β cells show that the Ca²⁺ sensitivity of exocytosis is relatively low such that [Ca²⁺]_i > 25 μM is necessary to elicit rates of exocytosis similar to those measured during voltage-clamp depolarization (Barg et al., 2001). Thus, insulin granules are postulated to be located very close to Ca²⁺ channels in order to sense high [Ca²⁺]_i levels during Ca²⁺ influx (also see Wiser et al., 1999).

We have recently reported heterogeneity in the Ca²⁺ sensitivity of exocytosis in adrenal chromaffin cells. In particular, we found a highly Ca²⁺-sensitive pool (HCSP) of granules that can be released at [Ca²⁺]_i levels in the low μM range (Yang et al., 2002). Here we investigate the Ca²⁺ sensitivity of exocytosis in the INS-1 insulin-secreting cell line. INS-1 cells, like primary β cells, undergo Ca²⁺-triggered exocytosis of insulin that is modulated by hormones, glucose, and protein kinases (Asfari et al., 1992; Rosengren et al., 2002). We find that INS-1 cells have an HCSP in addition to the less Ca²⁺-sensitive readily releasable pool (RRP) that has previously been characterized. We find that the size of the HCSP in INS-1 cells, but not the Ca²⁺-sensitive rate constant of release, is increased by glucose and agents that activate PKA and PKC. The HCSP may be responsible for release of insulin that is not tightly coupled to Ca²⁺ influx through Ca²⁺ channels.

Rat insulinoma INS-1 cells (provided by C. Wollheim, University of Geneva, Geneva, Switzerland) were maintained in culture media consisting of RPMI 1640 medium supplemented with 50 μM 2-mercaptoethanol, 2 mM l-glutamine, 10 mM HEPES, 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mM pyruvate, and 10% FBS. Cells were kept in a humidified 37°C incubator with 5% CO₂ and were passaged by trypsinization and subcultured once a week. Cell suspensions were plated onto glass coverslips (25 mm OD) coated with 0.3 mg/ml poly d-lysine. Experiments were performed 1 or 2 d after cells were plated out.

The standard bath solution consisted of (in mM) 140 NaCl, 5.5 KCl, 1 MgCl₂, 5 CaCl₂, 10 HEPES titrated to pH 7.2 with NaOH. The bath solution contained 2.6 mM glucose, whereas higher glucose concentrations were achieved by adding an appropriate volume of a 1 M glucose stock solution. Standard pipette solution consisted of (in mM) 110 l-glutamic acid, 110 N-methylglucamine, 8 NaCl, 1 MgCl₂, 2 Na₂-ATP, and 40 HEPES titrated to pH 7.2 with CsOH. The Ca²⁺ indicator dyes fura-2FF (K⁺ salt; Teflabs) and bisfura-2 (K⁺ salt; Molecular Probes) were included in the pipette solution at an equimolar ratio (0.1 mM) to allow measurement of [Ca²⁺]_i over a wide dynamic range (Voets, 2000). The Ca²⁺ cage Nitrophenyl-EGTA (0.5–1 mM; Ellis-Davies and Kaplan, 1994), 80% loaded with CaCl₂, was added to the pipette solution. We adjusted the amount of added CaCl₂ and Nitrophenyl-EGTA to achieve baseline [Ca²⁺]_i values typically in the range of 0.5–0.7 μM. Elevated basal [Ca²⁺]_i is commonly used in caged Ca²⁺ experiments in order to increase the size of the RRP (e.g., Voets, 2000) and we observed that the HCSP was often quite small or absent when basal [Ca²⁺]_i was <∼200 nM (not depicted). The rate of C_m increase after achieving the whole-cell configuration was typically 1 fF/s or less and amperometric events were rarely observed (not depicted), thus the “basal” rate of exocytosis appeared to be low before stimulation.

All chemicals were from Sigma-Aldrich except otherwise indicated. PKA inhibitory peptide (PKI) was from Alexis. Bisindoylymaleimide I was from Calbiochem.

The Ca²⁺ indicator combination was calibrated in cells using whole-cell recording in a manner similar to that used by Voets (2000). Eight solutions with free [Ca²⁺]_i of 0, 0.31, 1.2, 3.6, 7.2, 28, 127 μM and 10 mM were prepared using Ca²⁺ buffers EGTA (K_D = 150 nM at pH 7.2, total concentration 10 mM), N-hydroxyethylethylenediaminetriacetic acid (K_D = 3.6 μM, concentration 10 mM), or 2-ol-N, N′-tetraacetic acid (K_D = 81 μM, concentration 30 mM). Dissociation constants for the buffers are from Martell and Smith (1974)(–1989). Whole-cell recordings were made using each of the calibration solutions in the pipette, and the ratio of fluorescence emission (535 ± 25 nm) for 340 and 365 nm excitation were recorded after allowing 2 min for the pipette solution to diffuse into the cell. Three to five recordings were made for each calibration solution. The following equation was fit to the eight measured ratio values (R) to find the five unknown parameters (Voets, 2000):

Inversion of the equation was used to calculate [Ca²⁺]_i from the ratio R measured in caged Ca²⁺ experiments.

Most experiments were performed at room temperature (∼22°C) with the exception of the experiments depicted by the squares in Fig. 2. In these experiments, cells were warmed to 30–32°C by perfusing warmed bath solution. The temperature of the cells was confirmed by placing a small thermistor in the middle of the recording chamber.

Whole-cell patch-clamp measurements were performed using an EPC-9 patch-clamp amplifier and PULSE acquisition software (HEKA). Pipettes (2–4 MΩ) were pulled from Kimax glass capillaries, coated with dental wax at their tips, and fire polished. The pipette potential was held at a dc value of −70 mV except during membrane depolarization. Capacitance measurements were performed using the “sine + dc” software lock-in amplifier method implemented in PULSE software (Pusch and Neher, 1988; Gillis, 2000). The assumed reversal potential was 0 mV and the sinusoid had an amplitude of 25 mV and a frequency of 1.5 kHz.

A monochromator (Polychrome 4, TILL/ASI) coupled to the epifluorescence port of an Olympus IX-50 microscope with a fiber-optic cable excited the Ca²⁺ indicators at 340 and 365 nm. A two-port condensor (TILL/ASI) combined the monochromator excitation path with that of a flash lamp (TILL/ASI). A 40X 1.15 NA water immersion lens (U-APO; Olympus) focused the excitation light and collected fluorescent light. The fluorescent light (535 ± 25 nm) was measured using a photodiode mounted in a viewfinder (TILL/ASI).

Amperometric measurements were performed with carbon-fiber electrodes purchased from ALA, Inc. The tip of the carbon-fiber electrode was positioned to just touch the surface of a cell using a micromanipulator (PCS-5000; Burleigh). The amperometric current (I_AMP) generated by oxidation of 5-HT was measured using a patch-clamp amplifier (EPC-9; HEKA) while holding the potential of the carbon-fiber electrode at +700 mV. INS-1 cells were incubated in media containing 0.6 mM 5-HT and 0.6 mM 5-hydroxytryptophan for 5–10 h before initiating amperometric measurements (Smith et al., 1995; Zhou and Misler, 1996). The amperometric signal was filtered at 200 Hz and sampled at 1 kHz.

Data analysis and curve fitting were performed using Igor software (Wavemetrics). Analysis of amperometric spikes was performed using software described in Segura et al. (2000). Results and histograms are expressed as mean ± SEM. Statistical comparisons were performed using Student's t test. * denotes P < 0.05, whereas ** denotes P < 0.01.

We introduced caged Ca²⁺ (0.5–1 mM nitrophenyl-EGTA loaded ∼80% with CaCl₂) into INS-1 cells through the patch pipette during whole-cell recording. A combination of a high affinity Ca²⁺ indicator (bisfura-2) and a lower affinity indicator (fura-2FF) were also included in the pipette solution to allow ratiometric measurement of [Ca²⁺]_i over a wide concentration range (Voets, 2000). Fig. 1 A depicts a typical experiment where flash photolysis of the cage elevates [Ca²⁺]_i from a baseline value of 450 nM to 3.3 μM and leads to a rise in membrane capacitance (ΔC_m) of ∼150 fF over several seconds that presumably reflects an increase in membrane surface area due to exocytosis. Note that the rate of exocytosis (dC/dt) declines despite the maintained elevation of [Ca²⁺]_i. The most common explanation given for transient exocytosis with a maintained Ca²⁺ stimulus is depletion of one or more discrete pools of secretory vesicles. Closer inspection of Fig. 1 A reveals that there are at least two kinetic phases of exocytosis: a small, rapid burst followed by a larger, but slower phase. Fitting a sum of two exponentials to the ΔC_m trace reveals that the rapid phase has an amplitude of 18 fF and a time constant of 33 ms, whereas the slower component has an amplitude of 181 fF and is released with a time constant of 1.75 s. We do not know with certainty the unitary increase in capacitance that results from the fusion of an individual granule in INS-1 cells. However, if we assume that granules in INS-1 cells contribute about the same amount of membrane as those in mouse β cells (1.7 fF/granule; Ammala et al., 1993b) then the fast kinetic phase corresponds to exocytosis of ∼10 granules, whereas ∼100 granules are released during the slower phase.

The slower kinetic phase has an amplitude and kinetic rate similar to that previously reported for exocytosis from the RRP in β cells (Takahashi et al., 1997; Barg et al., 2001). The faster component has not previously been identified as a distinct phase of exocytosis in β cells, presumably because its small amplitude makes it easy to be overlooked. We have reported a similar kinetic phase of exocytosis from adrenal chromaffin cells that we attribute to release of an HCSP of granules.

Release from the HCSP is most prominent at low μM [Ca²⁺]_i, however, release from the larger RRP is often too slow at these [Ca²⁺]_i values to be reliably measured. Fig. 1 B depicts a “hybrid stimulus” protocol we developed to estimate the size of the HCSP and RRP in a single sweep. We first use flash photolysis of caged Ca²⁺ to release the HCSP followed by 10 depolarizing pulses (30 ms in duration to +20 mV) to release most or all of the RRP. In the example of Fig. 1 B, the HCSP is ∼19 fF, whereas the increase in C_m after 10 depolarizing pulses (ΔC₁₀) is ∼142 fF, similar in size to the RRP defined from caged Ca²⁺ experiments (Fig. 1 A and Fig. 2). In the following experiments, we will use ΔC₁₀ as a crude indicator of the size of the RRP, but it is not necessarily a direct measure because clear exhaustion of a releasable pool (i.e., smaller jumps in C_m for late pulses in the pulse train) is not always evident, perhaps because the RRP is being refilled during the train.

Fig. 2 presents a plot of the amplitude and rate constant (inverse time constant) of exponential fits to the ΔC_m time course as a function of [Ca²⁺]_i measured after the flash. Note that the rates, but not the amplitudes, of the kinetic phases of exocytosis depend on [Ca²⁺]_i after the flash. This observation is consistent with the hypothesis that the two kinetic phases of exocytosis correspond to depletion of two discrete pools of granules. We define the small pool underlying the faster kinetic phase as the HCSP (Fig. 2, circles and squares), whereas the slower phase is due to release from the previously described RRP (Fig. 2, triangles). The size and [Ca²⁺]_i-dependent rate constant of release of the RRP are similar to values previously reported in β cells. For example, rate constants for a [Ca²⁺]_i value of 10 μM are ∼0.64 s⁻¹ in Barg et al. (2001), ∼1.0 s⁻¹ in Takahashi et al. (1997), and ∼10 s⁻¹ in Wan et al. (2004). Note in Fig. 2 that the rate of release from the HCSP is an approximately linear function of [Ca²⁺]_i, whereas release from the RRP is more steeply dependent on [Ca²⁺]_i. The rate constant of exocytosis from the RRP has previously been shown to depend on [Ca²⁺]_i raised to the 3rd–5th power in β cells (Takahashi et al., 1999; Barg et al., 2001), chromaffin cells (Heinemann et al., 1994; Voets, 2000), melanotrophs (Thomas et al., 1993), hair cells (Beutner et al., 2001), and neurons (Heidelberger et al., 1994; Bollmann et al., 2000; Schneggenburger and Neher, 2000). A shallow dependence of the rate of release from the HCSP on [Ca²⁺]_i is also observed in chromaffin cells (Yang et al., 2002) and in rat β cells (Wan et al., 2004). Nevertheless, the release rate from the HCSP may drop steeply for [Ca²⁺]_i < 1 μM, because the baseline rate of exocytosis preceding photoelevation of [Ca²⁺]_i is very low (unpublished data).

We performed most experiments at room temperature (∼22°C), which is sufficient to support exocytosis from most neuroendocrine cells (e.g., Thomas et al., 1993; Dinkelacker et al., 2000), whereas temperatures >∼28°C are usually thought to be necessary to support exocytosis from primary β cells (e.g., Atwater et al., 1984; Gillis and Misler, 1992; Renstrom et al., 1996). We performed a set of experiments to see if the size or time course of release of the HCSP is affected by raising the temperature to 30–32°C. The squares in Fig. 2 represent exponential fits to data obtained at this higher temperature range, whereas the circles are from experiments performed at room temperature. These data demonstrate that neither the number of granules in the HCSP nor the Ca²⁺-dependent rate of release are steeply dependent on temperature in INS-1 cells. We also find that depolarization-evoked ΔC_m responses in INS-1 cells are not highly sensitive to temperature over this range (unpublished data).

The first vesicles released during membrane depolarization are generally thought to be located nearby Ca²⁺ channels and have been termed the immediately releasable pool (IRP; Horrigan and Bookman, 1994; Barg et al., 2001). We designed an experiment to test if the HCSP is released in response to brief membrane depolarization due to its high sensitivity to Ca²⁺. Fig. 3 presents a sample experiment, representative of responses from eight cells, where photoelevation of [Ca²⁺]_i to 1.8 μM is still able to elicit release from the HCSP following three brief (10 ms) depolarizing pulses to deplete the IRP. Note that the ΔC_m responses to the second and third pulses are substantially smaller than the response to the first pulse, demonstrating that the first pulse exhausted the IRP and that it takes substantially longer than 200 ms for the IRP to refill. Since an HCSP is still present after exhaustion of the IRP, the HCSP must be a distinct population of vesicles from the IRP, and a substantial fraction of the vesicles in the HCSP do not appear to be located particularly close to Ca²⁺ channels.

Glucose stimulates insulin release through membrane depolarization triggered by closure of K⁺(ATP) channels, but also has one or more roles in “priming” insulin granules for exocytosis. Fig. 4 A presents sample experiments following the protocol of Fig. 1 B to investigate modulation of the HCSP size with preincubation of cells in various concentrations of glucose. The HCSP is very small or absent in cells incubated in 2.6 mM glucose but is maximally stimulated at 10 mM glucose. The ΔC₁₀ response also was sensitive to glucose and roughly doubled in size between 2.6 and 10 mM glucose, whereas the Ca²⁺ current and total Ca²⁺ influx (Q_Ca) are unaffected by the sugar. Results summarized in Fig. 4 B suggest that the size of the HCSP may be more sensitive to glucose than the RRP.

The second messenger cAMP stimulates exocytosis from β cells by increasing the amount of release for a given amount of Ca²⁺ influx (Jones et al., 1986; Ammala et al., 1993a; Gillis and Misler, 1993; Renstrom et al., 1997; Eliasson et al., 2003). Fig. 5 A depicts sample experiments demonstrating that including cAMP (0.2–1 mM) in the pipette solution increases the size of the HCSP and the ΔC₁₀ response. The enhancement of exocytosis by cAMP can be completely blocked by including PKI (0.5 mM), a peptide inhibitor of PKA, in the pipette. Including a membrane permeant analogue of cAMP (CPT-cAMP, 100 μM) in the bath solution also increases the size of the HCSP and the ΔC₁₀ response, and this enhancement is completely blocked by the PKA inhibitor Rp-cAMP (0.5 mM, Fig. 5 B). The results summarized in Fig. 5 B suggest that the HCSP and RRP are both increased approximately twofold by cAMP and that this effect appears to be completely attributable to activation of PKA in INS-1 cells.

The rate constant and amplitude of release from the HCSP with cAMP in the pipette are depicted with filled circles in Fig. 6. The control values from Fig. 2 (open circles and triangles) are also included in Fig. 6 for comparison. These data demonstrate that cAMP does not affect the kinetics or intrinsic Ca²⁺ sensitivity of release from the HCSP.

Activation of PKC by phorbol esters also potently increases the amount of exocytosis for a given influx of Ca²⁺ in β cells (Ammala et al., 1994). In chromaffin cells, activation of PKC increases the size of the HCSP approximately sixfold while increasing the RRP by two to threefold (Gillis et al., 1996; Yang et al., 2002). Fig. 7 demonstrates that activation of PKC with bath application of PMA (100 nM) increases the HCSP and the ΔC₁₀ response by two to threefold in INS-1 cells. This enhancement can be completely blocked by bisindoylymaleimide I (1 μM), a selective inhibitor of PKC.

The rate constant and amplitude of the HCSP obtained from cells incubated in PMA are plotted as asterisks in Fig. 6. Thus PMA, like cAMP, does not affect the time course and Ca²⁺ sensitivity of release from the HCSP. A similar affect of PMA on the amplitude, but not the time course, of release from the HCSP has previously been reported in chromaffin cells (Yang et al., 2002).

We tested combinations of glucose, cAMP, and PMA in paired experiments to determine if the enhancement of exocytosis by these agents involves common mechanisms. Fig. 8 (A and B) demonstrates that preincubation of cells in 10 mM glucose results in larger granule pool sizes than preincubation in 2 mM glucose even in the presence of high concentrations of cAMP and PMA, thus it is likely that glucose enhances exocytosis by mechanisms independent of the PKA and PKC stimulatory pathways. In contrast, Fig. 8 C demonstrates that the actions of PKA and PKC on enhancing pool sizes are not additive and therefore may involve a common mechanism.

Carbon-fiber amperometry has proven to be a powerful tool to measure the contents and release kinetics of individual catecholamine-containing vesicles. The technique can also be used to measure release of 5-HT from β cells preincubated with 5-HT and a precursor (Smith et al., 1995; Zhou and Misler, 1996). 5-HT is accumulated in insulin-containing granules, and release of this electro-oxidizable indolamine has been shown to be an accurate and specific marker of insulin granule exocytosis in β cells (Aspinwall et al., 1999).

Fig. 9 A depicts a sample trace where flash photolysis elevates [Ca²⁺]_i to 3.7 μM, leading to release of the HCSP. The flash is followed 1 s later by a train of depolarizing pulses to release granules from the RRP. Spikes of oxidative current are recorded during both phases of release with each spike reporting release of 5-HT from an individual granule. Significant cell-to-cell variability of responses occurs because only those granule fusion events that occur in the ∼10–15% of the membrane that is directly adjacent to the carbon-fiber microelectrode are captured. Fig. 9 B presents the averaged response from 22 sweeps taken from 17 cells. The amperometric current is integrated to give an indicator of cumulative release (Q_amp) and is scaled to match the maximal ΔC_m signal. Note that the ratio of Q_amp to ΔC_m is similar for release of the HCSP and the RRP, suggesting that the granules in these two pools have a similar diameter (see discussion). The good correspondence between the time course of Q_amp and ΔC_m also demonstrates that endocytosis does not greatly affect the kinetics of ΔC_m under our recording conditions.

Fig. 9 C is a cumulative histogram comparing the charge of individual spikes released from the HCSP (dashed line, n = 78) and the RRP (solid line, n = 153) from 17 cells. Q_amp is 0.11 ± 0.02 pC for HCSP spikes and 0.16 ± 0.07 pC for RRP spikes. Fig. 9 D presents a cumulative histogram of t_1/2, defined as the time interval between the rising and falling phase of a spike where I_amp assumes the half-maximal value. The value of t_1/2 is 4.16 ± 0.61 ms for HCSP spikes versus 3.84 ± 0.42 ms for RRP spikes. Thus the quantal content and the time course of release of 5-HT from granules is similar for the RRP and the HCSP, and it is therefore unlikely that vesicles in the two pools differ in gross morphology.

We describe a small (∼14 granule), but relatively rapid, burst of exocytosis in INS-1 cells that is prominent when [Ca²⁺]_i is rapidly elevated into the low μM range. We attribute this burst of exocytosis to release of an HCSP of vesicles with characteristics that are very similar to what we previously described in adrenal chromaffin cells. A companion article in this issue (Wan et al., 2004) reports an HCSP in rat β cells with a size nearly identical to what we observe but with somewhat higher sensitivity to [Ca²⁺]_i. Exocytosis with a similar sensitivity to Ca²⁺ has also recently been reported in rod photoreceptors (Thoreson et al., 2004) and in pituitary gonadotropes (Zhu et al., 2002).

The size, but not the rate constant of release, of the HCSP is sensitive to cAMP and the phorbol ester PMA (Fig. 6; also see Wan et al., 2004). The enhancement of pool size by PMA can be completely blocked with bisindolylmaleimide I (Fig. 7) and other specific PKC blockers (Wan et al., 2004), supporting the hypothesis that phorbol esters enhance exocytosis from neuroendocrine cells by activating PKC. In contrast, the enhancement of release from hippocampal nerve terminals by phorbol esters is reported to be mediated by munc-13 (Rhee et al., 2002). We find that PKI and Rp-cAMP, specific inhibitors of PKA, can completely block the ability of cAMP to increase the size of the HCSP and RRP. Similarly, Wan et al. (2004) report a complete block of the forskolin-induced enhancement of HCSP and RRP pool sizes by Rp-cAMP. These findings are surprising because previous reports describe only a partial block of the cAMP-induced enhancement of depolarization-evoked exocytosis by PKA inhibitors in mouse β cells and INS-1 cells (Renstrom et al., 1997; Rosengren et al., 2002). This has led to the hypothesis that cAMP enhances exocytosis through a PKA-independent pathway such as cAMP-GEFII (Eliasson et al., 2003). It should be noted that we use a relatively high concentration of PKI in the pipette (0.5–1 mM) because, e.g., high concentrations of PKI are necessary to rapidly and completely block PKA-dependent activation of the CFTR Cl⁻ channel during whole-cell recordings in cardiac myocytes (Hwang et al., 1993). The effects of saturating doses of cAMP and PMA on enhancing granule pool sizes do not appear to be additive (Fig. 8 C), therefore PKA and PKC may converge on a common mechanism of action.

Activation of PKA or PKC increases the size of both the HCSP and the RRP by approximately twofold (Figs. 5 and 7), thus PKA and PKC do not appear to change the fraction of releasable granules that are in a highly Ca²⁺-sensitive state in INS-1 cells. In contrast, PKC increases the fraction of granules in the HCSP in chromaffin cells (Yang et al., 2002), and both PKC and PKA increase the fraction of granules in the HCSP in rat β cells (Wan et al., 2004). Our data suggest that the fraction of granules in the highly Ca²⁺-sensitive state may be higher when glucose is elevated in INS-1 cells (Fig. 4), but further study is necessary to confirm this trend. If so, then this would provide a mechanism for glucose to increase the sensitivity of insulin granule exocytosis to modest increases in [Ca²⁺]_i. Further studies of the modulation of HCSP and RRP pool sizes could benefit from using the perforated patch configuration to maintain a more physiological intracellular milieu (Horn and Marty, 1988; Gillis et al., 1991).

Our data show that the size and release kinetics of the HCSP (Fig. 2) and RRP (not depicted) are not highly temperature sensitive in INS-1 cells over a range (22–32°C) where the size of the depolarization-evoked ΔC_m responses change many fold in primary β cells (Gillis and Misler, 1992; Renstrom et al., 1996). This finding likely indicates a difference in the temperature sensitivity of insulin secretion between INS-1 cells and primary β cells (but see Sheu et al., 2003). Another intriguing possibility is that the high basal (preflash) [Ca²⁺]_i in our experiments (0.5–0.7 μM) overcomes a temperature-sensitive step in vesicle priming.

Preincubation of INS-1 cells in glucose increases the size of the HCSP and the RRP (Fig. 4; also see Rosengren et al., 2002). Glucose is often thought to influence insulin secretion by increasing the ATP concentration or the ATP/ADP ratio in the cell. However, during whole-cell recording, the ATP concentration is clamped to the value in the pipette solution (2 mM) following a brief interval of dialysis. Thus, the enhancement of vesicle pool size must either be a long-lived effect of previous elevation of ATP or some other mechanism that is resistant to whole-cell dialysis. Preincubation of cells in 10 mM glucose results in larger granule pool sizes than incubation in 2 mM glucose even when combined with a high concentration of cAMP (1 mM in the pipette) or PMA (100 nM in the bath). Thus glucose does not work exclusively by activating PKA or PKC to increase granule pool sizes. Further experiments are needed to determine if the glucose effect we observe is mediated by membrane depolarization and Ca²⁺ influx that occurs before the cells are voltage clamped for our recording.

β cells contain small GABA-containing synaptic-like vesicles in addition to dense-core insulin-containing granules (Thomas-Reetz and De Camilli, 1994), so it is natural to question whether distinct kinetic phases of exocytosis measured as membrane capacitance changes correspond to release of morphologically distinct types of vesicles (Takahashi et al., 1997). However, it has recently been estimated that release of GABA-containing synaptic-like vesicles only contributes ∼1% of the capacitance signal in β cells (Braun et al., 2004). Our measurements of quantal 5-HT release using carbon-fiber microelectrodes supports the hypothesis that the HCSP and the “conventional” releasable pool (the RRP) consist of the same type of vesicle. Simultaneous measurement of insulin and 5-HT release with modified carbon fiber electrodes demonstrates that 5-HT is released exclusively from insulin-containing granules (Aspinwall et al., 1999). We find that the amount of 5-HT released from a single vesicle and oxidized on the surface of the carbon fiber (Q_amp) is similar for the two kinetic phases of exocytosis (Fig. 9 C). The time course of release of 5-HT from individual vesicles in the two pools also does not appreciably differ (Fig. 9 D). Since the single-granule Q_amp and the ratio of the cumulative ΔC_m to Q_amp are the same for the two pools (Fig. 9 B), the ΔC_m due to fusion of a single vesicle is likely to be the same for the HCSP and the RRP. The single-vesicle ΔC_m is proportional to the diameter squared, so vesicles in the HCSP have a similar size as vesicles in the RRP. Thus the HCSP does not appear to be a morphologically distinct type of vesicle, but rather a distinct granule “state.” It seems likely that vesicles in the highly Ca²⁺-sensitive state posses either a different Ca²⁺ sensor or a different conformation of Ca²⁺-sensing proteins.

The rate constant of exocytosis from the HCSP has a shallower dependence on [Ca²⁺]_i than the rate constant of release of the RRP (Fig. 2). Thus the rate of exocytosis from the HCSP dominates for [Ca²⁺]_i <∼10 μM, but will actually be slower than the rate of release from the RRP at higher [Ca²⁺]_i levels. This suggests that the highly Ca²⁺-sensitive triggering mechanism may have a higher affinity for Ca²⁺, but a lower efficacy in triggering fast exocytosis. Thus, it would appear that the HCSP release mode is specialized for mediating slow exocytosis in response to modest elevations of [Ca²⁺]_i. Our finding that the bulk of the HCSP is not released in response to brief depolarization (Fig. 3) suggests that many of the vesicles in the HCSP do not reside close to Ca²⁺ channels. Thus, the HCSP is likely to respond to global elevations of [Ca²⁺]_i that result from action potential trains or release of Ca²⁺ from internal stores by hormone agonists. The modulation of the HCSP by PKA and PKC may explain how a phorbol ester can stimulate insulin secretion at a subthreshold glucose concentration (Bozem et al., 1987) and how PKA and PKC can elicit exocytosis from β cells at substimulatory [Ca²⁺]_i values (Jones et al., 1985, 1986). In contrast, the RRP release mechanism is specialized for fast, synchronous release of granules that experience high [Ca²⁺]_i during action potentials due to their close proximity to Ca²⁺ channels.

We thank Prof. Claes Wollheim for providing the INS-1 cell line and Vanessa Melton for technical assistance. We also thank Miguel A. Brioso and Ricardo Borges (Universidad de La Laguna, Tenerife, Spain) for providing software support for analysis of amperometric spikes.

This work was supported by National Institutes of Health R01 NS40453 to K.D. Gillis.

Olaf S. Andersen served as editor.