A Subset of 50 Secretory Granules in Close Contact With l-Type Ca²⁺ Channels Accounts for First-Phase Insulin Secretion in Mouse β-Cells

It is well established that glucose-stimulated insulin secretion follows a characteristic biphasic time course (Fig.1B) (10). Shortly (1–2 min) after elevation of the glucose concentration, a transient component of insulin secretion is observed. In mouse islets, this component has a duration of <10 min and reaches a peak secretory rate of 50–100 pg of insulin per minute and islet (first phase). The secretory rate then declines and plateaus at a lower secretory rate of 10–80 pg per minute and islet (depending on the experimental conditions), which continues as long as glucose remains elevated (second phase). Only fuel secretagogues are capable of eliciting the second phase, and when insulin release is initiated by nonmetabolizable stimuli, such as high extracellular K⁺ and tolbutamide, only first-phase release is observed (11).

The secretory rates quoted above can be converted to number of granules undergoing release, as a single granule contains 2 femtogram insulin (equivalent to ∼250,000 molecules of insulin) and assuming that a typical islet contains 1,000 β-cells. As discussed elsewhere (5,12), these calculations suggest that first-phase insulin secretion corresponds to the total release of 50–100 granules per β-cell and that second-phase release amounts to 5–40 granules per cell and minute.

Before considering our experimental data, a few words about the method we used to measure exocytosis (capacitance measurements) may be appropriate. A confounding feature of capacitance measurements is that they almost invariably indicate rates of exocytosis much higher than expected from reported values of insulin secretion. Accordingly, the 5 granules per minute we derived above as the lower value of glucose-induced second-phase insulin secretion corresponds to 300 granules (2.5% of total granule number) per hour. However, the maximum rate of capacitance increase observed in β-cells is equivalent to 600 granules (5% of total granule number) per second. Thus, there appears to be a discrepancy of almost four orders of magnitude! However, it is important to remember that this high rate of exocytosis detected by capacitance measurements only proceeds for a brief period (<0.5 s) and that exocytosis in the steady state is considerably slower (<15 granules per second). It also deserves pointing out that capacitance measurements only report changes in the cell surface area (equal to membrane fusion). Obviously, there might be significant delays between the fusion of the granules with the plasma membrane, the establishment of the (narrow, ∼1 nm) fusion pore and cargo release. Recent observations in our laboratory suggest that there is indeed a significant delay (1–10 s) between membrane fusion and release of insulin (S.B., C.S. Olofsson, J. Schriever-Abeln, A. Wendt, S. Gebre-Medhin, E.R., P.R., submitted). Interestingly, there is evidence that the opening of the fusion pore in eosinophils is modulated by Ca²⁺ and protein kinase C (13). If this also occurs in β-cells, then regulation of insulin secretion at a very distal level (i.e., subsequent to elevation of cytoplasmic Ca²⁺ and membrane fusion) can be envisaged.

The pancreatic β-cell is an interesting example of Ca²⁺-induced hormone release because it is capable of a high-speed exocytosis, even though it is equipped with <500 Ca²⁺ channels (estimated by nonstationary fluctuation analysis) (14). The β-cell has a cell capacitance of 5 pF. Using a conversion factor of 10 femtofarad (fF)/μm², we can estimate that the cell surface area is ∼500 μm², indicating a Ca²⁺-channel density of <1 Ca²⁺ channel/μm². This should be compared with the Ca²⁺-channel density in adrenal chromaffin cells (the archetypal endocrine cell), which has been estimated to range between 9 and 20 Ca²⁺ channels (15,16). Although the Ca²⁺-channel density of the β-cell is thus only 5–10% of that in the chromaffin cells, the secretory capacity is not correspondingly reduced. The maximum rate of exocytosis measured in mouse β-cells during a voltage-clamp depolarization is 1.2 pF/s, 40% of the 3 pF/s observed in chromaffin cells (17). This also holds true if we compare the Ca²⁺-channel density and peak exocytotic rate in β-cells with the corresponding values in glucagon-producing α-cells. The Ca²⁺-current density in α-cells measured during depolarizations from −70 mV to 0 mV at 2.6 mmol/l extracellular Ca²⁺ is 39 ± 4 pA/pF (n = 9), threefold higher than the 13 ± 2 pA/pF (n = 8; P < 0.001) observed in β-cells of the same preparation and using the same experimental protocol (standard whole-cell recordings at +32°C and using pipettes containing [mmol/l] 125 Cs-glutamate, 10 CsCl, 10 NaCl, 1 MgCl₂, 5 HEPES, 0.05 EGTA, 3 Mg-ATP, and 0.1 cAMP). Despite this, the maximum rate of exocytosis is only 35% higher in the α-cell (1.7 pF/s) (18) than in the β-cell (1.2 pF/s). Collectively, these findings argue that there exists a tight assembly of Ca²⁺ channels and secretory granules in the β-cell and that the coupling between Ca²⁺ entry and exocytosis is more efficient than in other secretory cells such as glucagon-releasing β-cells and adrenaline-producing chromaffin cells.

What is the mechanism behind the economical use of Ca²⁺ in the β-cell? We have previously proposed that l-type Ca²⁺ channels and secretory granules colocalize in the β-cell (19). More recently, we demonstrated that a peptide of the α_1C-subunit of the l-type channel (the only l-type channel in β-cells isolated from Naval Medical Research Institute mice) (14), corresponding to the synprint site (Lc_753-893; “synprint”), binds to immobilized syntaxin, SNAP-25, and p65 (20). It has been shown that injecting the corresponding region of N-type Ca²⁺ channels abolishes fast neurotransmission (21). In β-cells, exogenous synprint suppressed exocytosis evoked by single (500 ms) voltage-clamp depolarizations. The latter effect could not be attributed to inhibition of Ca²⁺ entry or interference of exocytosis per se as witnessed by the lack of effect on the peak Ca²⁺ current and exocytosis triggered by a global increase in cytoplasmic Ca²⁺-concentration ([Ca²⁺]_i) through flash photolysis of “caged” calcium (20). Thus, exocytosis in pancreatic β-cells is more neuron-like than what was perhaps anticipated.

Taken together with biochemical data, it seems likely that the synprint region of the l-type Ca²⁺ channels tethers it to the secretory granules, thus ensuring that the exocytotic machinery is rapidly exposed to very high Ca²⁺ concentrations upon Ca²⁺-channel opening (Fig.2A). Such a concept would account for both the ability of the β-cell to perform high-speed exocytosis despite the low Ca²⁺-channel density and the fact that exocytosis in the β-cell normally proceeds only during the depolarization (22). Adding a high concentration of exogenous synprint peptide competes for the binding site on the granule with the endogenous polypeptide loop that is part of the l-type Ca²⁺ channels (Fig. 2B). This leads to dissociation of the Ca²⁺ channel–SNARE/synaptotagmin complex, thereby increasing the average distance between channels and docked granules. This interpretation is supported by the finding that, although exocytosis elicited by single depolarizations is abolished by the synprint peptide, it can be rescued by a train of depolarizations applied at a high frequency (Fig. 2C). This is presumably a consequence of the interval between two successive depolarizations (500 ms) being too short for [Ca²⁺]_i to return to basal, leading to a progressive accumulation of Ca²⁺ inside the cell. Eventually, the global Ca²⁺ concentration has risen sufficiently to trigger exocytosis of granules situated too far away from the Ca²⁺ channels to be released by a single stimulation.

Analysis of the kinetics of depolarization-evoked exocytosis has revealed a rapid component of capacitance increase (presumably the β-cell equivalent of IRP) with an amplitude of ∼100 fF that is released with a time constant of 75 ms and that plateaus within 100–200 ms (Fig. 3A). This component is highly sensitive to inclusion of the synprint peptide, suggesting that it reflects the release of granules that have complexed with the Ca²⁺ channels (14). Interestingly, the size of this component is comparable to a rapid component of capacitance increase observed in response to a step elevation of [Ca²⁺]_i by photorelease of caged Ca²⁺, which likely corresponds to RRP (Fig. 3C) (5). Thus, it appears as if all RRP granules in the β-cells are situated in close proximity of the Ca²⁺ channels and thus identical to the IRP. If there were many more RRP granules than IRP granules, then a rapid global increase in Ca²⁺ would be expected to produce a much larger exocytotic response than that elicited by brief membrane depolarizations. However, estimating pool sizes from experiments using caged Ca²⁺ is complicated by concomitant and rapid endocytosis and the capacitance changes observed represent the net changes, which might result in a too low value of RRP. Indeed, when only including experiments lacking endocytosis, RRP was found to be >200 fF (C. Olofsson, S. Göpel, S.B., J. Galvanovskis, X. Ma, A. Salehi, P.R., L.E., submitted). We also estimated IRP using the two-pulse protocol first described by Gillis et al. (23). The size of IRP was estimated by two 100-ms pulses (i.e., the time where depolarization-evoked exocytosis starts to plateau; cf., Fig. 3A) applied with an interval of 200 ms (Fig.3 B). We thus derived a size of IRP of 102 ± 18 fF (n = 5), equivalent to 50 granules using a conversion factor of 2 fF/granule (22), similar to the synprint-sensitive component. It therefore seems justifiable to conclude that in mouse β-cells, up to 50% of the release-competent RRP granules are situated in the immediate vicinity of a Ca²⁺ channel (14).

If the IRP granules in the β-cell are all tethered to the Ca²⁺ channels via the synprint region of the Ca²⁺-channel protein, then the length of the peptide can be used to estimate the maximum distance between the mouth of the Ca²⁺ channel and the Ca²⁺ sensor. The synprint region contains 140 residues. Given that each residue contributes 1.5 Å (as expected, assuming α-helical organization) to the peptide chain, then the total length of the synprint peptide is ∼20 nm. From this, we infer that an IRP granule is situated ≤10 nm from a Ca²⁺ channel. This close to the Ca²⁺ channels, the Ca²⁺ concentration can be expected to rise to very high levels (24). Using caged Ca²⁺ to produce uniform and rapid increases in [Ca²⁺]_i, we have estimated the Ca²⁺ dependence of exocytosis in the β-cell. The speed of exocytosis was half maximal at ∼20 μmol/l [Ca²⁺]_i, and concentrations >30 μmol/l are required to attain the maximum speed of capacitance increase observed during voltage-clamp depolarizations (1.2 pF/s) (14). This Ca²⁺ dependence of exocytosis is close to that observed in preparations with a much higher Ca²⁺-channel density, such as the adrenal chromaffin cell (25) and central synapses (26,27). However, in a cell with as few Ca²⁺ channels as the β-cell, such high Ca²⁺ concentrations cannot arise, except in the immediate vicinity of the Ca²⁺ channels. For insulin secretion to occur, it is therefore essential that the release-competent secretory granules are situated close to the points of Ca²⁺ entry.

What is the functional advantage of tethering Ca²⁺ channels to the secretory granules? It seems unlikely that speed is the prime objective because insulin acts systemically following its release into the blood stream. Rather, the arrangement is likely to be a consequence of the low Ca²⁺-channel density. With so few Ca²⁺ channels, it is naturally important that the Ca²⁺ entering the cell is made use of in the most efficient way. This is achieved if the Ca²⁺ channels and secretory granules assemble into a functional complex. By this arrangement, the β-cell can achieve large exocytotic responses over long periods, while minimizing the Ca²⁺ influx and the subsequent expenditure of metabolic energy to restore the Ca²⁺ concentration to the resting level. Removal of Ca²⁺ from the cytoplasm by ATP-consuming Ca²⁺-ATPases can be expected to lower the cytoplasmic ATP/ADP ratio. Indeed, such an effect has been observed following depolarization-induced Ca²⁺ entry (28). This is obviously undesirable in a cell that utilizes ATP-regulated K⁺ channels to control the membrane potential. Excessive Ca²⁺ entry can then be expected to lower the cytoplasmic ATP/ADP ratio sufficiently to activate the K_ATP channels and thus prematurely repolarize the β-cell. Apparently, there is less need for the α-cell to economize with Ca²⁺ entry. The Ca²⁺-channel density is threefold higher in the β-cells than in the β-cell, but the maximum speed of exocytosis is only 35% greater. Although we have not yet tested the effects of the synprint peptide in α-cells, it is tempting to speculate that its effects are less pronounced than in the β-cell. Because of the high Ca²⁺-channel density in the α-cell, many RRP granules may be situated close enough to the Ca²⁺ channels to undergo rapid release following the opening of the channels (i.e., they functionally behave as IRP) and they need not be physically attached to the channel protein. A similar argument has been invoked in chromaffin cells to account for an ultrafast component of depolarization-induced exocytosis in these cells (17). Indeed, less efficient coupling between Ca²⁺ entry and secretion in α-cells and chromaffin cells is not unexpected, given that both adrenaline and glucagon need only be released for brief periods, but at a high rate under certain physiological conditions, such as hypoglycemia and intense labor. By contrast, insulin must be provided, more or less continuously, throughout life.

Electron microscopy on various neuronal and endocrine cells (29–31) has revealed a subset of secretory granules that are docked beneath the plasma membrane, but the precise relationship between the docked pool and RRP is not entirely clear. A similar situation appears to exist in the pancreatic β-cell, where the number of ultrastructurally docked granules is ∼10-fold larger than the functionally determined size of IRP/RRP (Olofsson C, Göpel S., S.B., J. Galvanouskis, X. Ma, A. Salehi, P.R., L.E., submitted). Therefore, refilling of RRP following its emptying does not necessarily involve the physical translocation of secretory vesicles. Recent experiments using a combination of capacitance measurements and confocal imaging of EGFP-tagged secretory granules (S.B., C.S. Olofsson, J. Schriever-Abeln, A. Wendt, S. Gebre-Medhin, E.R., P.R., submitted) indicate that replenishment of RRP principally reflects priming of granules already docked below the membrane. Given that replenishment of RRP does not correlate with any granule movements and the data suggesting that all RRP granules associate with the Ca²⁺ channels, it appears necessary to postulate a mechanism for disassembly of the Ca²⁺ channel/granule complexes following exocytosis, so that the Ca²⁺ channels are liberated to associate with new granules. If this is the case, then the Ca²⁺ channels must possess the ability to move laterally within the plasma membrane at sufficient speed to account for the recovery of RRP. It should therefore be interesting to study the mobility of fluorescently tagged Ca²⁺ channels in the plasma membrane and how agents that accelerate RRP recovery affect their mobility.

As remarked above, the bout of exocytosis that can be released immediately upon elevation of [Ca²⁺]_i corresponds to only ∼0.5% of the total number of granules. The ability of the cell to maintain the exocytotic capacity in the long term depends on the continuous replenishment of RRP by mobilization of granules from a “reserve pool.” The latter process involves ATP-, Ca²⁺-, and temperature-dependent “priming” reactions (3–6).

Clearly, several sequential steps are necessary to render the granule release-competent. Here we will only consider the possibility that the cell’s metabolic state influences granule priming, a feature that may be particularly important in the β-cell that controls fuel homeostasis of the entire body. Figure 4A (left) shows the increases in cell capacitance evoked by a train of 10 depolarizations. It can be seen that the total increase amounts to ∼500 fF. In a series of nine experiments, the total increase in cell capacitance evoked by the train averaged 363 ± 61 fF, 3.5-fold larger than the size of RRP estimated by the two-pulse protocol (see above). This suggests that RRP must have turned over several times during the pulse train and that granule priming operates at sufficient speed to allow RRP recovery within a few seconds. We subsequently repeated the experiment in the presence of 5 mmol/l Mg-ADP but maintained the concentrations of ATP and cAMP at 3 mmol/l and 0.1 mmol/l, respectively (Fig.4 A), middle). Exocytosis under the latter conditions is much reduced and the total increase in cell capacitance elicited by the train was only ∼100 fF (100 ± 16 fF in a series of four experiments). However, close inspection revealed some interesting features. Thus, the capacitance increase evoked by the first depolarization was hardly affected and averaged 130 ± 35 fF and 90 ± 11 fF under control conditions and in the presence of ADP, respectively. The latter value is close to the size of IRP/RRP estimated as described above, suggesting that exocytosis of granules that have already proceeded into RRP is unaffected by ADP. The inhibitory action of ADP is instead exerted at the level of granule priming and replenishment of RRP, manifested as the loss of the progressive capacitance increase normally observed in response to depolarization numbers 2–10.

We also tested the effects of rapid application of ADP (3.5 mmol/l) by release from a caged precursor. Figure 4 B shows an experiment in which exocytosis was elicited by intracellular dialysis with a medium containing 3 mmol/l Mg-ATP, 0.1 mmol/l cAMP, and 1.5 μmol/l [Ca²⁺]_i (mixture of 9 mmol/l Ca²⁺ and 10 mmol/l EGTA, pH 7.15). The rate of capacitance increase under control conditions averaged 30 ± 5 fF/s. Application of ADP reduced the rate of exocytosis to 3 ± 4 fF/s (n = 6; P < 0.01 vs. control). However, the inhibitory effect of ADP was not instantaneous, but occurred with a latency of 4 ± 1 s. During this period, the cell capacitance increased by 98 ± 10 fF. The latter value is again close to that which remained releasable during the train of depolarizations in the presence of ADP and the size of RRP estimate by the two-pulse protocol.

How does the metabolic state, as evidenced by the effects of changing the ATP/ADP ratio, regulate exocytosis? Experiments using Cl⁻ channel blockers, such as 4,4-diisothiocyanostilbene–2,2-disulfonate (DIDS), as well as omission of intracellular Cl⁻, implicate granular Cl⁻ fluxes in the regulation of exocytosis (32). Indeed, two antibodies against the Cl⁻ channel ClC-3 labeled β-cell granules. The antibody anti–ClC-3-C, which has been demonstrated to block ClC-3 channel function, abolished exocytosis in β-cells when included in the intracellular solution (32). This was neither due to unspecific interaction nor steric hindrance of the exocytotic machinery, because the antibody anti–ClC-3-N, which lacks effect on ClC-3 channel activity, was without inhibitory action on exocytosis. It is therefore not unexpected that intracellular application of DIDS (0.1 mmol/l) abolishes exocytosis elicited by a train except for the response to the first depolarization (Fig. 4A, right). The effects of the Cl⁻ channel blocker clearly resemble those of ADP (cf. Figure 4 A, middle), suggesting that granular Cl⁻ uptake is somehow required in the preparation of granules for release.

What is the significance of granular Cl⁻ uptake? We can discard the possibility that net accumulation of electrolytes and associated water uptake promotes exocytosis via increased intragranular hydrostatic pressure (33–35) because varying the osmolarity of the cytosol over a wide range (300–1,000 mOsm) did not influence exocytosis (32). Alternatively, the electrogenic H⁺ pumping, involved in the acidification of the granule interior, might require a shunt conductance to prevent the development of a large electrical potential across the granular membrane (36). Indeed, both inhibition of the granular v-type H⁺-ATPase with bafilomycin and disruption of the granular pH gradient using the protonophore CCCP (carbonyl cyanide m-chlorophenylhydrazone) resulted in strong suppression of Ca²⁺-induced exocytosis (32). This implicates granular acidification as an important step in the priming of the secretory granules. We verified this concept by correlating exocytosis to changes in intragranular pH. In these experiments, intragranular pH was measured with the fluorescent probe LysoSensor green, which accumulates in acidic compartments (mostly granules in the β-cell) and that is virtually nonfluorescent at neutral pH. Figure 5 A shows changes in cell capacitance evoked by intracellular dialysis with a solution containing 1.5 μmol/l Ca²⁺, 3 mmol/l Mg-ATP, and 0.1 mmol/l cAMP. It can be seen that cell capacitance increases by >2 pF over 60 s, and the average rate of capacitance increase was 32 ± 5 fF/s (n = 26). Including 5 mmol/l ADP in the intracellular solution resulted in strong inhibition of exocytosis, and the rate of capacitance increase averaged 8 ± 3 fF/s (n = 12; P < 0.01 vs. control [no ADP]). The effects of ADP were virtually identical to those of DIDS, and the mean rate of capacitance increase in the presence of the Cl⁻-channel blocker was 8 ± 4 fF/s (n = 8; P < 0.01). Figure 5 B correlates the exocytotic responses to changes of intragranular pH in cells that had been preloaded with LysoSensor. It can be seen that the granules acidify promptly following establishment of the whole-cell configuration and wash-in of the high ATP-solution. This acidification could be prevented by inclusion of ADP. Again, the effects of ADP on granule pH were mimicked by DIDS.

Figure 5 C attempts to summarize the chain of events that are involved in granule priming. We propose that granular acidification is a decisive event and depends on the simultaneous activity of the v-type H⁺-ATPase and the ClC-3 Cl⁻ channel. Granular Cl⁻ uptake counteracts the development of an insurmountable electric field across the membrane (positive inside), which would prevent further H⁺ pumping and intragranular acidificaiton. In this scenario, Ca²⁺ and ATP enhance granular Cl⁻ uptake and thus facilitate H⁺ pumping and granule priming, whereas ADP, by inhibiting ClC-3 (compare with the effects of DIDS), prevents granular acidification and priming. There are similarities between this concept and the recently postulated role of granular glutamate uptake in insulin secretion from rat β-cells (37). In both cases, the uptake of the anion by allowing granule acidification increases granule releasability.

This mechanism may not be limited to pancreatic β-cells, as Cl⁻ is required for hormone release in a variety of other cells, including pancreatic acinar cells (38) and pituitary melanotrophs (39). Very recent data also suggest that ClC-3 channels fulfill an important function in the acidification of synaptic vesicles (40). What is the role of granular acidification in granule priming? Possibly, low granular pH promotes exocytosis by inducing conformational changes in SNARE proteins, rendering them more fusogenic (41).

Based on the findings reviewed above, we propose the following model to account for biphasic insulin secretion (Fig. 6). We believe that first-phase insulin secretion is attributable to the release of RRP granules. A host of experimental data suggests that this pool is limited to 50 granules (giving rise to a capacitance increase of 100 fF upon fusion) in mouse pancreatic β-cells (Fig. 6 A). A remarkable feature of the β-cell is that nearly all RRP granules are situated in close proximity to the few (<500 per cell) Ca²⁺ channels and undergo exocytosis within 100–200 ms (Fig. 6 B). The size of RRP granules compares favorably with the total number of granules that are released during first-phase insulin secretion (Fig.6 C). We further propose that refilling of IRP/RRP by priming of granules accounts for second-phase insulin secretion. Capacitance measurements using intracellular dialysis with high [Ca²⁺]_i buffers (1.5 μmol/l) to trigger exocytosis have revealed that secretion continues for several minutes at a steady rate of 20–30 fF/s (10–15 granules/s). The total increase in cell capacitance under these conditions amounts to ≤5 pF, which is equivalent to the release of ∼2,500 secretory granules. Obviously, the 50 granules originally residing in RRP can only account for a few percent of this response, and we therefore conclude that most of the capacitance increase observed under these experimental conditions is due to release of granules mobilized from the reserve pool. The rate of granule mobilization suggested by the capacitance measurements described above (10–15 granules/s) is >10-fold higher than that required to account for second-phase insulin secretion (0.1–0.7 granules/s). Thus, factors other than the supply rate of granules set an upper limit to second-phase insulin secretion. The same conclusion was also reached by Daniel et al. (42) using different methods.

As discussed above, the secretory granules gain release competence subsequent to their intragranular acidification. The observations that ADP prevents granular acidification, granule priming, and refilling of RRP may explain why stimulation with tolbutamide or high extracellular K⁺ only produces a transient stimulation of insulin release when they are applied in the absence of glucose. This is because only granules that have proceeded into RRP are capable of undergoing exocytosis under these experimental conditions, and the replenishment of this pool is largely prevented by the high cytoplasmic ADP concentration.

Type 2 diabetes is associated with metabolic disturbances interfering with the generation of ATP (8,9). The finding that ADP prevents the refilling of RRP may thus also account for the low insulin secretory capacity associated with the disease. It is therefore of interest that sulfonylureas such as tolbutamide, in addition to closing K_ATP channels, also counteract the ability of ADP to inhibit exocytosis (32), an effect that must involve processes other than K_ATP channel closure.

Financial support was obtained from the Swedish Medical Research Council (Grants 8647, 9890, 12334, 12708, and 13147), the Swedish Diabetes Association, the Royal Physiographic Society, the Juvenile Diabetes Foundation International, the Knut and Alice Wallenbergs Stiftelse, the Swedish Society for Medical Research, the Crafoord Foundation, the European Community, and the Novo Nordisk Foundation.

We thank our colleagues in Lund for sharing unpublished data.