Dense core secretory vesicles revealed as a dynamic Ca²⁺ store in neuroendocrine cells with a vesicle-associated membrane protein aequorin chimaera

In most mammalian cells, the ER (Streb et al., 1984; Rizzuto et al., 1993; Brini et al., 1993) and Golgi complex (Pinton et al., 1998) are believed to represent the major mobilizable intracellular Ca²+ stores (Rutter et al., 1998). Uptake of Ca²+ into these stores is mediated largely by sarco(endo)plasmic reticulum Ca²+–ATPases (SERCAs)^* (Moller et al., 1996) and helps to maintain resting cytosolic Ca²+ concentrations ([Ca²+]_c) at levels (∼10⁻⁷ M) some four orders of magnitude lower than in the extracellular space (∼10⁻³ M). Release of Ca²+ from the ER and Golgi complex is provoked by hormones and other agonists which generate inositol 1,4,5 trisphosphate (IP₃) (Berridge, 1993) to open intracellular receptors (Mikoshiba, 1997). In certain cell types, such as skeletal muscle, these stores are also accessed by “Ca²+-induced Ca²+ release”, via receptors for the insecticide, ryanodine (Lai et al., 1988; Takeshima et al., 1989).

Secretory vesicles of both the dense core (Hellman et al., 1976; Hutton et al., 1983) and small synaptic types (Andrews and Reese, 1986) possess a substantial proportion of the total Ca²+ in cells specialized for peptide secretion. However, despite many studies, the role of secretory vesicles as Ca²+ buffers or stores remains controversial (Pozzan et al., 1994; Rutter et al., 1998; Yoo, 2000).

Mobilization of dense core vesicle Ca²+ by IP₃ was originally suggested as a possibility in chromaffin cells (Yoo and Albanesi, 1990). Later data suggested that IP₃ receptors may be present on the vesicle membrane in insulin-secreting cells (Blondel et al., 1994), though subsequent experiments showed the antibody used cross-reacted with insulin (Ravazzola et al., 1996). IP₃, as well as the receptor (RyR) agonist cyclic ADP ribose (cADPr) (Galione, 1994) have been reported to release Ca²+ from individual acinar cell zymogen granules (Gerasimenko et al., 1996), although contamination of these preparations (e.g., with ER or Golgi apparatus–derived vesicles) is a potential problem (Yule et al., 1997). Finally, evidence for the direct participation of secretory granules in the control of cytoplasmic Ca²+ concentration was recently provided in intact goblet cells (Nguyen et al., 1998).

At present, the molecular mechanisms responsible for Ca²+ uptake into secretory vesicles are also a matter of controversy. At high Ca²+ concentrations (>50 μM), uptake into dense core vesicles in permeabilized hypophysial cells (Troadec et al., 1998; Thirion et al., 1999) and into isolated chromaffin cell granules (Krieger-Brauer and Gratzl, 1982) can occur via Na⁺/Ca²+ exchange, while synaptic vesicles also accumulate Ca²+ via a Ca²+-H⁺ antiport system (Goncalves et al., 1998). However, the pump/exchange mechanisms active at physiological [Ca²+]_c are uncertain.

To examine the pathways by which Ca²+ crosses the limiting membrane of the dense core secretory vesicle of living islet β-cells, we have developed a new approach to monitor the free Ca²+ concentration within the secretory vesicle matrix ([Ca²+]_SV) using recombinant targeted aequorin (Rizzuto et al., 1995). Cloned from the jellyfish Aequorea victoria (Inouye et al., 1985), aequorin is a calcium-sensitive bioluminescent protein (Cobbold and Rink, 1987), previously used to measure free Ca²+ concentrations in a variety of subcellular organelles (Rutter et al., 1998). Importantly, aequorin activity is less severely inhibited at low pH values (<6.5; Blinks, 1989) than Ca²+ probes based on green fluorescent protein (GFP) (Miyawaki et al., 1997; Baird et al., 1999; Emmanouilidou et al., 1999). If appropriately targeted, this probe should allow Ca²+ concentrations to be measured in the acidic environment of the secretory granule interior (Orci et al., 1985).

Vesicle-associated membrane protein (VAMP)2/synaptobrevin (Sudhof et al., 1989) is a vesicle-specific SNARE with a single transmembrane-spanning region. Expression of chimaeric cDNA encoding a fusion protein between VAMP2 and aequorin (VAMP.Aq) has therefore allowed the intravesicular-free Ca²+ concentration to be monitored dynamically in live MIN6 β-cells. With this approach, we show that Ca²+ is actively pumped into dense core vesicles when [Ca²+]_c increases, and may be released via RyR, but not IP₃, receptors. This release may be important at sites of high intracellular Ca²+, including sites of exocytosis at the plasma membrane.

Chimeric cDNA encoding hemagglutinin (HA)1-tagged aequorin, fused to VAMP2 (Sudhof et al., 1989), was generated as shown in Fig. 1 A.

Immunocytochemical analysis of MIN6 cells transfected with VAMP.Aq cDNA revealed close overlap with insulin staining (Fig. 1 B). Explored at a higher resolution by immunoelectron microscopy (Fig. 1 C), VAMP.Aq immunoreactivity was highly enriched in 61 of 148 (41.2%; n = 11 cells) vesicles colabeled for insulin (Fig. 1 C). Analyzed by single labeling for VAMP.Aq, staining of the ER, Golgi apparatus, and small synaptic-like microvesicles (Reetz et al., 1991) was very low, while reactivity was also present on the plasma membrane and in endosomes (see the legend to Fig. 1 and Discussion).

Given the high total Ca²+ content of secretory vesicles (Hutton et al., 1983), we used the approach adopted previously to measure Ca²+ in the ER lumen (Montero et al., 1995). Apoaequorin was reconstituted at a low free Ca²+ concentration (Montero et al., 1995), achieved by depleting cells of Ca²+ (Materials and methods). Depletion of vesicle Ca²+ had no marked effect on glucose or K⁺-stimulated insulin secretion, or on vesicle motility (Pouli et al., 1998b; Tsuboi et al., 2000; unpublished data).

To determine the response of the expressed aequorins to Ca²+ in situ, permeabilized cells were incubated at buffered Ca²+ concentrations in the presence of ionomycin and monensin (Fig. 2, C and E). The sensitivity to Ca²+ (at pH 7.0) of the VAMP.Aq chimaera was similar to that reported previously for mutant (D¹¹⁹A) aequorin (Montero et al., 1995). Intravesicular pH in intact cells was determined using a fusion construct between VAMP2 and a mutated, pH-sensitive GFP (pH.fluorin(e); Miesenbock et al., 1998), and gave a pH value of 6.3 ± 0.02 (n = 85 cells; Fig. 2 A). Confirming that this low intravesicular pH was unlikely to significantly affect VAMP.Aq, near identical calibration data were obtained at pH 5.7 (unpublished data). Therefore, we used the constants obtained in vitro (Montero et al., 1995) to calculate [Ca²+] from the fractional rate of aequorin consumption (F) according to: [Ca²+] = 1.44√10^(LOG[F]−^3.4) (Rutter et al., 1993).

To monitor uptake of Ca²+ into vesicles in living cells, we provoked rapid Ca²+ influx into cells through store-operated channels (Parekh and Penner, 1997). During reintroduction of CaCl₂ to cells previously perifused in EGTA, [Ca²+]_c approached 500 nM, then fell back to 150–200 nM (Fig. 2 B). By contrast, steady state free [Ca²+] in the secretory vesicle matrix was 30–90 μM (51; 7.5 μM; n = 5 separate cultures; Fig. 2 D), whereas free [Ca²+] in the ER lumen was fivefold higher (249; 12.9 μM; n = 3 preparations; Fig. 2 F).

Both the rate and extent of the [Ca²+] increases in the secretory vesicles and the ER were strongly dependent on the presence of added ATP in permeabilized cells (Fig. 3, A and D). The increase in [Ca²+]_SV upon reintroduction of Ca²+ ions was completely inhibited by the P-type Ca²+ pump inhibitor, orthovanadatate, at >100 μM (Fig. 3 B), but was insensitive to 10 μM orthovanadatate, a concentration which strongly inhibits the plasma membrane Ca²+-ATPase (Carafoli, 1991). Preincubation with the specific SERCA inhibitor, thapsigargin (Thastrup, 1990), and perifusion with cyclopiazonic acid (CPA) (Mason et al., 1991) had no effect on the changes in [Ca²+]_SV (Fig. 3 C and Table I), but markedly (>85%) inhibited ER [Ca²+] increases (Fig. 3 E).

Arguing against vesicular Ca²+ uptake via Ca²+/H⁺ exchange, monensin, an Na⁺/H⁺ exchanger, had no effect on [Ca²+]_SV increases in permeabilized cells (Fig. 4 A). Similarly, ionomycin (a Ca²+/H⁺ exchanger) decreased [Ca²+]_SV marginally (∼20%; Fig. 4 B), which is consistent with the inability of this ionophore to bind to and transport Ca²+ against a prevailing H⁺ gradient, but caused a rapid and complete decline in [Ca²+]_SV in the additional presence of monensin (Fig. 4 B). Neither the vacuolar H⁺-ATPase inhibitor, bafilomycin (Fig. 4, C and D), nor the protonophore carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), had any effect on vesicular Ca²+ uptake (Fig. 4 E).

[Ca²+]_SV increases were also completely unaffected by raising [Na⁺] from 0, 2, or 140 mM (Fig. 4 F), and by simultaneous blockade of Na⁺/Ca²+ exchange and Ca²+/H⁺ exchange (zero Na⁺, 300 nM bafilomycin; unpublished data).

To achieve large changes in intracellular IP₃ levels, we coexpressed the metabotropic glutamate receptor, mGluR5, with VAMP.Aq or ER.Aq. Stimulation of mGluR5-transfected cells with the specific agonist, (S)-3,5-dihydroxyphenylglycine (DHPG) (Thomas et al., 2000) caused a significant decrease in [Ca²+]_ER (Fig. 5 E), but an increase in [Ca²+]_SV (Fig. 5 A). DHPG had no significant effect on the distribution of vesicles (imaged after expression of VAMP.GFP/pHfluorin (r); Fig. 5 D), indicating that enhanced exocytosis and exposure of the VAMP.Aq to the extracellular medium was unlikely to contribute to the observed increases in [Ca²+]_SV (Fig. 5 A, and see Discussion).

Similarly, in permeabilized cells, IP₃ had no effect on [Ca²+]_SV (Fig. 5, B and C), while causing a large decrease in [Ca²+]_ER (Fig. 5, F and G).

The RyR agonists caffeine or 4-chloro-3-ethylphenol (4-CEP) (Zorzato et al., 1993) provoked rapid decreases in [Ca²+]_SV and [Ca²+]_ER (Fig. 6, A and C) in intact cells. Similarly, cADPr (Galione, 1994) caused clear decreases in both parameters in permeabilized cells (Fig. 6, B and D), and the effects of cADPr were strongly potentiated by palmitoyl CoA (Fig. 6, B and D), a coagonist of RyRs (Chini and Dousa, 1996).

The above results suggested that activation of RyR on vesicles should cause an increase in cytoplasmic [Ca²+]_c, even after the depletion of the ER Ca²+ pool. Added to fura-2–loaded MIN6 cells, 4-CEP caused a substantial increase in [Ca²+]_c in the absence of external Ca²+ (Fig. 7 A). This [Ca²+]_c increase was partly retained after depletion of SERCA-dependent stores, giving a peak [Ca²+]_c increase of 30–40% of that in control cells (Fig. 7 B versus A). The [Ca²+]_c increase elicited by 4-CEP was also partially retained after treatment of cells with ionomycin (Fig. 7 C), but completely abolished after treatment with ionomycin plus monensin, to deplete acidic Ca²+ stores (Fig. 7 D). Demonstrating that the effects of 4-CEP on [Ca²+]_SV were likely mediated by RyR, immunoreactivity to these channels was revealed on vesicle membranes by direct immunoelectron microscopy (Fig. 7 E).

Exposure of islet β- or MIN6 cells to glucose or other nutrients causes an increase in intracellular-free [ATP] (Kennedy et al., 1999) closure of ATP-sensitive K⁺ (K_ATP) channels (Bryan and AguilarBryan, 1997) and Ca²+ entry via voltage-sensitive (L-type) Ca²+ channels (Safayhi et al., 1997).

High (20 mM) glucose, or a combination of nutrient secretagogues (Ashcroft and Ashcroft, 1992), caused a large increase in steady state [Ca²+] both in the ER and in vesicles (Fig. 8, A and C), with no significant change in vesicular distribution (Fig. 8 B), which is consistent with the increases in [Ca²+]_c seen under these conditions (Grapengiesser et al., 1988).

We show here that aequorin can be targeted to the lumen of dense core secretory vesicles in living cells, with no incorporation into the ER, Golgi, or trans-Golgi network using the cell's own protein-sorting machinery (Fig. 1). However, a proportion of VAMP.Aq was also present on the plasma membrane and endosomes (Fig. 1 C). Since our recordings were made at either high (1.5 mM, intact cells) or low (400 nM, permeabilized cells) external Ca²+ concentrations, the plasma membrane–located photoprotein should be either rapidly exhausted, or inactive, respectively. In any case, it would appear that plasma membrane–targeted aequorin is not reconstituted, since we would have expected to see a “spike” of luminescence upon readdition of high Ca²+ concentrations to intact cells (e.g., Fig. 3 C). None was seen, suggesting that the mistargeted protein may be inactivated by unknown mechanisms during recycling between the plasma membrane and endosomes.

Similarly, endosomal aequorin is likely to be inactive under our conditions, since (a) the total endosomal Ca²+ content is low in neuroendocrine cells (Pezzati et al., 1997) and (b) free [Ca²+] is particularly low (∼3.0 μM) in acidic endosomes (Gerasimenko et al., 1998); active VAMP.Aq was located exclusively in an acidic compartment (Fig. 4 B). Simulation of the contribution of endosomal VAMP.Aq reveals that our measurements may slightly underestimate [Ca²+]_SV (by <10%; unpublished data).

Mutated aequorins respond well to Ca²+ concentrations over the range (30–90 μM) which pertains within the secretory vesicle, as well as the rest of the secretory pathway (200–300 μM) (Montero et al., 1997; Pinton et al., 1998), without significant interference from the low pH environment. This feature of aequorin contrasts with GFP-based Ca²+ probes (Miyawaki et al., 1997; Baird et al., 1999; Emmanouilidou et al., 1999), whose fluorescence is strongly reduced at acidic pHs (Miesenbock et al., 1998; Fig. 2 E). Thus, aequorin may represent the most suitable molecularly targetable Ca²+ reporter presently available for the secretory vesicle interior.

Despite possessing a larger total Ca²+ content (Andersson et al., 1982; Hutton et al., 1983; Nicaise et al., 1992), secretory vesicles displayed a significantly lower free [Ca²+] than the ER or Golgi apparatus. Our measured values for [Ca²+]_SV (∼50 μM) correspond fairly well to measurements using other techniques in isolated chromaffin granules (Krieger-Brauer and Gratzl, 1982; 24 μM; null point titration), respiratory tract goblet cells (24 μM; calcium orange 5 N fluorescence; Nguyen et al., 1998) and platelet α-granules (12 μM; null point titration; Grinstein et al., 1983). Thus, free Ca²+ represents <0.05% of the total vesicular calcium content of β-cell secretory vesicles (assuming a total Ca²+ concentration of 50–100 mM; Hutton et al., 1983). Secretory vesicles appear to have the highest Ca²+-buffering capacity of all subcellular organelles so far examined, with the percentage of free Ca²+ being much higher both in the cytosol (∼2%) and in the ER (∼10%; Pozzan et al., 1994). Importantly, resting [Ca²+]_SV was well below the K_M for Ca²+ of proinsulin-processing enzymes (Davidson et al., 1988).

The identity of the Ca²+ binding proteins (or other molecules) responsible for chelating free Ca²+ in these vesicles is unknown. Chromogranins (Yoo and Albanesi, 1990), or the mammalian homologue of Tetrahymena thermophila granule lattice protein 1 (Grlp1) (Chilcoat et al., 1996), are each strong possibilities. In addition, Ca²+ chelation by small molecules, such as ATP (Hutton et al., 1983), may also be involved. Finally, it is likely that in islet β-cells the insulin crystal itself also participates in chelating vesicular Ca²+ (Palmieri et al., 1988; see below).

The dense core secretory vesicle pool of islet β-cells has previously been considered relatively inert (Howell et al., 1975; Prentki et al., 1984). We provide evidence here that net uptake of Ca²+ into the dense core secretory vesicle population occurs during activated Ca²+ influx (Figs. 2 and 3), and in response to a receptor agonist (Fig. 5) or to nutrients (Fig. 8). However, and as discussed below (see also Fig. 9), these measurements do not exclude the possibility that discrete vesicle pools may experience different [Ca²+]_SV changes.

We considered the possibility that the increases in [Ca²+]_SV observed upon challenge of cells with agonists (Fig. 5) or nutrients (Fig. 8) may be due in part to the activation of exocytosis, and thus the exposure of aequorin within the vesicle matrix to the extracellular Ca²+ concentrations (1.5 mM). However, this phenomenon is likely to contribute negligibly to the observed signals, since only a tiny fraction of the total vesicle population (the “primed” pool) in β-cells (5–20/13,000 per min; Rorsman, 1997) undergoes fusion, even during the maximal stimulation of exocytosis. Consistent with this, we could detect no significant net movement of VAMP.GFP fluorescence to the plasma membrane after stimulation with DHPG (Fig. 5 D) or nutrients (Fig. 8 B).

Changes in intravesicular Ca²+ concentration detected with vesicle-targeted aequorin seem likely to result largely from the flux of Ca²+ ions across the vesicle membrane. However, two caveats are important: (a) Ca²+ release from/association with binding sites (or a “polyanionic matrix”) (Nguyen et al., 1998) within the vesicle may also occur, especially if the intravesicular concentrations of other ions (e.g., K⁺) change; (b) the number of Ca²+ ions that are transported is unknown, since this depends on the binding to intraorganellar sites. As discussed above, the present studies demonstrate that the ratio of free to bound Ca²+ within the secretory vesicles is much lower than that in the ER or Golgi lumen. This would appear at first to suggest that changes in intravesicular Ca²+ concentration must involve the flux of a very large number of Ca²+ ions. However, an alternative possibility is that there are two (or more) pools of bound Ca²+ in the secretory vesicle, one (the larger) comprising Ca²+ which is tightly bound (perhaps to a structural component such as the insulin crystal) and the other a more loosely bound and readily exchangeable pool. Depletion of this latter, “labile” pool, may thus lead to a substantial decrease in the free vesicular [Ca²+], while involving the release into the cytosol of relatively few Ca²+ ions.

The present data suggest that Ca²+ accumulation into vesicles is catalyzed at physiological [Ca²+]_c chiefly by a P-type Ca²+-ATPase. Previous studies have indicated that transport of Ca²+ into the Golgi apparatus is mediated partly by a SERCA pump, and partly by another, unidentified thapsigargin-insensitive system (Pinton et al., 1998). This second ATP-dependent Ca²+ uptake system may be closely related to ATP2C1 (Hu et al., 2000), the mammalian homologue of the yeast Golgi Ca²+ transport ATPase, PMR1 (Sorin et al., 1997). Supporting this view, mRNA-encoding ATP2C1 is abundant in MIN6 cells, and polyclonal antibodies raised to ATP2C1 reveal a punctate pattern of intracellular staining (unpublished data). Intriguingly, although patients defective in the ATP2C1 gene develop skin lesions (“Hailey-Hailey” disease; Hu et al., 2000), it is unclear whether these individuals also tend to suffer from neuroendocrine or other disorders (e.g., diabetes mellitus).

The present results suggest that IP₃ is unlikely to stimulate the release of Ca²+ from dense core secretory vesicles directly (Fig. 5).

By contrast, we now show (Figs. 6 and 7) that functional RyR channels are present on the dense core vesicle membrane of MIN6 cells. Type II RyR mRNA is present in ob/ob mouse islets, mouse β-TC3 cells (Islam et al., 1998), and in rat islets (Holz et al., 1999). RyR II protein has been detected in INS1 β-cells (Gamberucci et al., 1999), and ryanodine binding sites revealed in human islets (Holz et al., 1999), mouse islets and MIN6 cells (Varadi and Rutter, 2001). Evidence for functional RyRs on the ER INS-1 β-cells (Maechler et al., 1999) has also recently been provided. The present data indicate that dense core vesicles may represent a large fraction (∼30%) of the total RyR–accessible Ca²+ pool in MIN6 β-cells (Fig. 7 B). However, the proportion of the total vesicular Ca²+ pool which is mobilizable via RyR is small. Thus, the [Ca²+]_c increases observed when the total acidic Ca²+ pool was emptied with CPA, ionomycin, and monensin (Fig. 7 D) were much larger than those apparent when cells were treated with 4-CEP after CPA and ionomycin (Fig. 7 C).

In the present studies, cADPr caused an apparent release of Ca²+ from both the ER and from the secretory vesicles in permeabilized MIN6 β-cells. Moreover, we have recently observed (Varadi and Rutter, 2001) that photolysis of caged cADPr increases [Ca²+]_c in MIN6 cells. These findings were surprising in light of earlier results using ob/ob mouse islets (Islam et al., 1993) or INS-1 β-cells (Rutter et al., 1994), where cADPr was ineffective in permeabilized cells. However, it should be noted that the present studies may provide greater sensitivity to small Ca²+ fluxes by measuring changes in [Ca²+] inside the secretory vesicle or ER.

Could changes in intracellular cADPr concentration play a role in the regulation of insulin secretion as proposed by Takasawa et al. (1993)? Supporting this view, mice homozygous for inactivation of the CD38 (ADPribose cyclase) gene display impaired glucose tolerance (Kato et al., 1999). However, no change in the islet cADPr content was observed by others in response to acutely elevated glucose concentrations in vitro (Scruel et al., 1998), suggesting further studies are necessary.

Dense core secretory vesicles rapidly take up Ca²+ ions via an ATP-dependent non-SERCA Ca²+ pump (see Scheme, Fig. 9). Ca²+ accumulated into vesicles may later be released via RyRs once these vesicles reach a high local Ca²+ concentration beneath the plasma membrane (Neher, 1998) or elsewhere in the cell. This release may further contribute to the high local [Ca²+]_c in the vicinity of the submembrane vesicles (Emmanouilidou et al., 1999), and may be important to trigger exocytosis (Scheenen et al., 1998).

Cell culture reagents were obtained from GIBCO BRL or Sigma-Aldrich and molecular biologicals from Roche Diagnostics. Fura-2 and cADPr were from Sigma-Aldrich and IP₃ was from Molecular Probes.

VAMP2 cDNA was amplified by PCR using specific primers (forward: 5′-AAG CTT ACC ATG TCG GCT ACC GCT GCC ACC-3′; reverse: 5′-AAG CTT AGT GCT GAA GTA AAC GAT GAT-3′, HindIII sites are underlined). This fragment was then inserted upstream of the HindIII-EcoRI fragment encoding HA1-tagged D¹¹⁹-A aequorin (Kendall et al., 1992) in plasmid pBS+. Chimeric VAMP.Aq cDNA was shuttled as a NotI-XhoI fragment into pcDNAI (Invitrogen; Fig. 1 A).

ER.Aq (Montero et al., 1995) and VAMP.Aq cDNAs were transferred into plasmid pShuttleCMV (He et al., 1998) as Kpn1-Xho1 and Xba1-HindIII fragments, respectively. The resultant constructs were recombined with vector pAdEasy-1, and viral particles were generated as described (Ainscow and Rutter, 2001).

MIN6 β-cells (Miyazaki et al., 1990; passages #20 and #30) were cultured in DME, with additions as given previously (Ainscow and Rutter, 2001). Cells cultured on 13 mm poly-l-lysine–coated coverslips were transfected with VAMP.Aq or ER.Aq plasmid DNA (1 μg) using Lipofectamine (Promega) in Optimem I serum-free medium (GIBCO BRL), or infected with adenoviruses as described (Ainscow and Rutter, 2001). The total quantity of reconstituted aequorin was higher in cells transfected with ER.Aq than VAMP.Aq cDNAs due to the inclusion in the former of intronic regulatory elements. Thus, typical integrated counts/culture (Fig. 2) after the expression of VAMP.Aq were 3 × 10⁵ and 1.5 × 10⁶ for conventional and adenoviral expression, respectively, and 2 × 10⁶ and 2 × 10⁷ for ER.Aq.

Cells were Ca²+ depleted by incubation with ionomycin (10 μM), the SERCA inhibitor CPA (10 μM; Mason et al., 1991; Pinton et al., 1998), and monensin (10 μM) in modified Krebs-Ringer bicarbonate buffer (KRB; 140 mM NaCl, 3.5 mM KCl, 0.5 mM NaH₂PO₄, 0.5 mM MgSO₄, 3 mM glucose, 10 mM Hepes, 2 mM NaHCO₃, pH 7.4) supplemented with 1 mM EGTA for 5 min at 4°C. Aequorin was reconstituted in 0.1 mM EGTA, 5 μM coelenterazine (cytosolic aequorin) or coelenterazine n (Shimomura et al., 1993; VAMP.Aq, ER.Aq) for 1–2 h at 4°C.

Intact cells were perifused (2 mL/min⁻¹) with KRB plus additions as stated (37°C) close to a photomultiplier tube (ThornEMI), and emitted light was collected (one data point) with a photon-counting board (ThornEMI; Cobbold and Rink, 1987). Where indicated, cells were permeabilized with digitonin (20 μM, 1 min at 22°C) and subsequently perifused in intracellular buffer (IB) (140 mM KCl, 10 mM NaCl, 1 mM KH₂PO₄, 5.5 mM glucose, 2 mM MgSO₄, 1 mM ATP, 2 mM Na⁺ succinate, 20 mM Hepes, pH 7.05) at 37°C.

Cells were loaded with 5 μM fura-2-AM (Sigma-Aldrich) for 40 min (37°C) in KRB plus 0.05% Pluronic F-127 (BASF). [Ca²+]_c was monitored in single cells at 37°C as described (Ainscow et al., 2000), using a Leica DM-IRBE inverted microscope (40× objective), and a Hamamatsu C4742-995 charge-coupled device camera driven by OpenLab (Improvision) software.

Permeabilized cells were loaded with 3 μM acridine orange (Tsuboi et al., 2000) before imaging (37°C) on a Leica DM-IRBE^TM inverted optics epiflurosecence microscope, at 493 ± 10 excitation, 530 nm emission (filters were from Chroma Techology).

48 h after transfection or infection, cells were fixed in 3.7% (vol/vol) paraformaldehyde and probed with antibodies essentially as described (Pouli et al., 1998a).

Gelatin-embedded, aldehyde-fixed cells were frozen in liquid nitrogen and ultrathin cryosections were obtained with a Reichert-Jung Ultracut E. Immunogold localization was performed as described (Confalonieri et al., 2000). Sections were immunostained either with rabbit anti-HA antibody (Sigma-Aldrich), or with a sheep anti-RyR antibody (Molecular Probes) followed by 10-nm protein A-gold. For double labeling, anti-HA–labeled sections were incubated with 1% glutaraldehyde in 0.1 M Na₂PO₄, pH 7.0, followed by incubation with a guinea pig antiinsulin antibody (Dako) and 15-nm protein A-gold. Sections were examined with a ZEISS EM 902 electron microscope. No labeling was detected in control sections (unpublished data).

Cells were imaged at 37°C on a Leica SP2 confocal spectrophotometer, (488 nm excitation) using a 100× oil immersion objective.

Free [Ca²+] and [Mg²⁺] were calculated using “METLIG” (Rutter and Denton, 1988). Data are presented as the mean ± SEM for the number of observations given, and statistical significance calculated using Student's t test.

We thank Dr. Mark Jepson and Alan Leard (Bristol MRC Cell Imaging Facility) for their assistance with cell imaging, Dr. Nadia Todesco for help with the construction of VAMP.Aq cDNA, Maria Cristina Gagliani for help in electron microscopy, and Dr. Elek Molnar for the provision of mGluR5 cDNA and DHPG.

Supported by grants to G.A. Rutter from the Medical Research Council (UK), Human Frontiers Science Program, the Wellcome Trust, the Biotechnology and Biological Research Council, DiabetesUK, and the European Union; and to R. Rizzuto from Italian “Telethon” (Projects 850 and E0942), the Italian University and Health Ministries, the Italian Space Agency, The Italian National Research Council, The Armenise Harvard Foundation, and the Italian Association for Cancer Research.