Polarized expression of G protein-coupled receptors and an all-or-none discharge of Ca2+ pools at initiation sites of [Ca2+]i waves in polarized exocrine cells.

In the present work we examined localization and behavior of G protein-coupled receptors (GPCR) in polarized exocrine cells to address the questions of how luminal to basal Ca(2+) waves can be generated in a receptor-specific manner and whether quantal Ca(2+) release reflects partial release from a continuous pool or an all-or-none release from a compartmentalized pool. Immunolocalization revealed that expression of GPCRs in polarized cells is not uniform, with high levels of GPCR expression at or near the tight junctions. Measurement of phospholipase Cbeta activity and receptor-dependent recruitment and trapping of the box domain of RGS4 in GPCRs complexes indicated autonomous functioning of G(q)-coupled receptors in acinar cells. These findings explain the generation of receptor-specific Ca(2+) waves and why the waves are always initiated at the apical pole. The initiation site of Ca(2+) wave at the apical pole and the pattern of wave propagation were independent of inositol 1,4,5-trisphosphate concentration. Furthermore, a second Ca(2+) wave with the same initiation site and pattern was launched by inhibition of sarco/endoplasmic reticulum Ca(2+)-ATPase pumps of cells continuously stimulated with sub-maximal agonist concentration. By contrast, rapid sequential application of sub-maximal and maximal agonist concentrations to the same cell triggered Ca(2+) waves with different initiation sites. These findings indicate that signaling specificity in pancreatic acinar cells is aided by polarized expression and autonomous functioning of GPCRs and that quantal Ca(2+) release is not due to a partial Ca(2+) release from a continuous pool, but rather, it is due to an all-or-none Ca(2+) release from a compartmentalized Ca(2+) pool.

In the present work we examined localization and behavior of G protein-coupled receptors (GPCR) in polarized exocrine cells to address the questions of how luminal to basal Ca 2؉ waves can be generated in a receptor-specific manner and whether quantal Ca 2؉ release reflects partial release from a continuous pool or an all-or-none release from a compartmentalized pool.

Immunolocalization revealed that expression of GPCRs in polarized cells is not uniform, with high levels of GPCR expression at or near the tight junctions. Measurement of phospholipase C␤ activity and receptor-dependent recruitment and trapping of the box domain of RGS4 in GPCRs complexes indicated autonomous functioning of G q -coupled receptors in acinar cells. These findings explain the generation of receptor-specific
Ca 2؉ waves and why the waves are always initiated at the apical pole. The initiation site of Ca 2؉ wave at the apical pole and the pattern of wave propagation were independent of inositol 1,4,5-trisphosphate concentration. Furthermore, a second Ca 2؉ wave with the same initiation site and pattern was launched by inhibition of sarco/endoplasmic reticulum Ca 2؉ -ATPase pumps of cells continuously stimulated with sub-maximal agonist concentration. By contrast, rapid sequential application of sub-maximal and maximal agonist concentrations to the same cell triggered Ca 2؉ waves with different initiation sites. These findings indicate that signaling specificity in pancreatic acinar cells is aided by polarized expression and autonomous functioning of GPCRs and that quantal Ca 2؉ release is not due to a partial Ca 2؉ release from a continuous pool, but rather, it is due to an all-or-none Ca 2؉ release from a compartmentalized Ca 2؉ pool.
Ca 2ϩ signaling by G protein-coupled receptors (GPCR) 1 in-volves the generation of inositol 1,4,5-trisphosphate (IP 3 ) in the cytosol and Ca 2ϩ release from the endoplasmic reticulum (1). In polarized exocrine cells, Ca 2ϩ release is not uniform but occurs in the form of Ca 2ϩ waves. Rooney et al. (2) reported a unique initiation site and propagation pattern of GPCR-evoked Ca 2ϩ waves in hepatocytes. Kasai et al. (3) first described the unique feature of initiation of Ca 2ϩ waves at the apical pole and their propagation to the basal pole of pancreatic acini. This phenomenon was later confirmed in pancreatic acinar cells (3)(4)(5)(6) and was extended to other exocrine cells (7,8). Subsequent studies showed that expression of high levels of all IP 3 receptor (IP 3 Rs) isoforms at the apical pole accounts for the initiation of Ca 2ϩ waves at this site (7,9). Furthermore, the apical pole showed higher sensitivity to Ca 2ϩ release by IP 3 than other regions of the cell, including the basal pole (3,10,11). However, since the discovery of the polarized Ca 2ϩ waves in exocrine cells, it remained a mystery how IP 3 can be generated in the apical pole to initiate the waves. Early functional (12) and radioligand (13) localization of receptors in tissue slices indicated that GPCRs are expressed in the basal membrane. Therefore, it was assumed that during GPCR stimulation IP 3 generated in the basal pole diffuses to the apical pole to initiate Ca 2ϩ release and waves (14). This assumption has several difficulties. For example, at maximal stimulus intensity Ca 2ϩ release starts within few milliseconds of cell stimulation. The diameter of exocrine acinar cells is about 20 m. This requires exceptionally high rate of diffusion of IP 3 in the cytosol of this cells. In pancreatic acini, Ca 2ϩ release events can remain confined to the apical pole (3,4). This requires continuous traffic of IP 3 through the cytosol without causing Ca 2ϩ release. Another alternative is generation of IP 3 in or close proximity to the apical pole. This requires localization of GPCR at the apical pole. With the development of suitable antibodies for immunolocalization of GPCR and the use of isolated cell clusters that increase the resolution of immunolocalization, it became possible to re-examine GPCR localization in relation to initiation of Ca 2ϩ waves.
Another aspect of Ca 2ϩ release in terms of initiation and propagation of Ca 2ϩ waves is the architecture of the Ca 2ϩ pool and the dynamics of Ca 2ϩ release. Propagation of Ca 2ϩ waves requires either sequential Ca 2ϩ release from a compartmentalized pool or release from different sections of a continuous pool along the path of the Ca 2ϩ wave. A unique property of Ca 2ϩ release from internal stores, the quantal feature of Ca 2ϩ release (15), can reflect the spatial organization of the Ca 2ϩ pool that is needed for propagation of Ca 2ϩ waves. Ca 2ϩ release evoked by either stimulation of GPCR (15), activation of the IP 3 receptors (15)(16)(17), or activation of the ryanodine receptors (18) has quantal properties, that is at submaximal stimulus intensity only part of the Ca 2ϩ pool is released and at increased stimulus an increased fraction of the pool is released. Two main models were proposed to explain quantal release. The first is an all-or-none release of Ca 2ϩ from a compartmentalized pool that has a continuum of sensitivity to IP 3 (15,19,20). The second model proposes phasic Ca 2ϩ release from a homogeneous pool where the phase of release at a given IP 3 concentration is determined by gating of IP 3 Rs activity by Ca 2ϩ content remaining in the stores (21). The first model fits well with the experimentally observed variable sensitivity of different cellular regions of pancreatic acini to IP 3 -mediated Ca 2ϩ release (3,11). On the other hand, recent work (22) in Xenopus oocytes reports that the quantal behavior of Ca 2ϩ release stems from rapid adaptation of the IP 3 R channels of a continuous Ca 2ϩ pool and claimed to refute the compartmentalization model.
The constancy of initiation of Ca 2ϩ waves in the apical pole by all GPCR of pancreatic acini and the variable sensitivity of different cellular regions to IP 3 -mediated Ca 2ϩ release prompted us to determine localization of GPCR in pancreatic acini and to examine their autonomous behavior and what mechanism they use to evoke quantal release. We report that expression of GPCR is highly enriched in the apical pole, at or just underneath the tight junctions. This provides an explanation of how Ca 2ϩ waves are initiated at the apical pole. The initiation site and propagation pattern of Ca 2ϩ waves are receptor-specific in the same cells. This appears to be the result of autonomous coupling of receptors to G proteins and thus operation of GPCR signaling complexes. Finally and most notably, by using agonist concentration jump protocols and inhibition of the sarco/endoplasmic reticulum Ca 2ϩ -ATPase (SERCA) pump of partially stimulated cells, we provide evidence that the quantal properties of Ca 2ϩ release are due to an all-or-none Ca 2ϩ release from a compartmentalized Ca 2ϩ pool.

EXPERIMENTAL PROCEDURES
Materials-Thapsigargin was from Alexis. The anti-muscarinic monoclonal antibody (mAb) M35 was from Argene. The hybridoma cells producing the 5H9 mAb against muscarinic type 3 receptor (M3R) and the specificity of the Abs were described elsewhere (23). Two pAbs that recognize different epitopes of the cholecystokinin (CCK) receptor were prepared using peptide antigens, and their specificity was extensively characterized and verified by blocking with peptides that were used to raise the anti-CCKR antibodies (24). A recombinant M2R was a generous gift from Dr. Eliott Ross (University of Texas Southwestern Medical Center, Dallas). Anti-IP 3 R2 pAb was a generous gift from Dr. Akihiko Tanimura (University of Hokkaido, Ishikari-Tobetsu, Japan). The box domain of RGS4 (4Box) was prepared as described (25). Anti-IP 3 R3 mAb and anti ZO1 pAbs were purchased from ABR and Zymed Laboratories Inc., respectively. Anti-ZO1 mAb was obtained from the Hybridoma Bank at the University of Iowa.
Preparation of Pancreatic Acini and Single Acinar Cells-Acini were prepared from the pancreas of 100 -150-g rats by limited collagenase digestion as described previously (6). After isolation, the acini were resuspended in a standard solution A containing (in mM) 140 NaCl, 5 KCl, 1 MgCl 2 , 1 CaCl 2 , 10 Hepes (pH 7.4 with NaOH), 10 glucose, and 0.1% bovine serum albumin and kept on ice until use. Doublet or triplet acinar cell clusters were obtained by incubation of a minced pancreas in a 0.025% trypsin, 0.02% EDTA solution for 5 min at 37°C. After washing with solution A supplemented with 0.02% soybean trypsin inhibitor, doublets and triplets were liberated by a 7-min incubation at 37°C in the same solution that also contained 160 units/ml pure collagenase. The cells were washed with solution A and kept on ice until use.
[Ca 2ϩ ] i Imaging-Pancreatic acinar cells were loaded with Fura 2, and [Ca 2ϩ ] i was imaged as detailed before (6). Fura 2 fluorescence was measured at a single excitation wavelength of 380 nm, by averaging eight consecutive images for each time point. Under these conditions and using a frame size of 256 ϫ 240 pixels, recording was at a resolution of 90 ms/averaged image. During perfusion with the control solution and just before the first stimulation, the image of resting cells was acquired and was taken as the fluorescence signal at time 0 (F 0 ). Pixel values of all subsequent images were divided by this image, and the traces and images are the calculated F t /F 0 , where F t is the fluorescence at time t.
To calculate the distance between initiation sites from which Ca 2ϩ waves were triggered by different agonists or sequential application of the same agonist in the same cell, the initiation sites of all waves were marked on the first image, and the image was transferred to Adobe Photoshop. A right triangle was drawn using the ray between the two initiation sites as the hypotenuse. The x and y axes originated from the left and top plane, respectively. The distance between the initiation sites was then calculated by means of the Pythagorean theorem. The distances from multiple experiments were averaged and are given as means Ϯ S.E. Although this procedure allowed quantitative analysis of the distance between initiation sites, such analysis has limitations inherent to the imaging procedure. The most critical limitation is the spatial resolution of cell imaging. The initiation sites and the waves are three-dimensional, whereas imaging at a speed needed to capture Ca 2ϩ waves can be done only in two dimensions. Another limitation of the measurement is the temporal resolution of our recording system. This resulted in initiation sites that occupy 1-2 m 2 . Nevertheless, the differences in Ca 2ϩ wave initiation sites and propagation pattern observed when cells were stimulated with two different agonists were sufficiently large to allow quantitative analysis of our measurements.
Electrophysiology-The whole cell configuration of the patch clamp technique was used for measurement of Ca 2ϩ -activated Cl Ϫ current, which correlates with changes in [Ca 2ϩ ] i near the plasma membrane (26). The experiments were performed with single acinar cells perfused with solution A. The standard pipette solution contained the following (in mM): 140 KCl, 1 MgCl 2 , 0.5 EGTA, 5 ATP, 10 Hepes (pH 7.3 with KOH) with or without 100 nM 4Box, as described in previous studies (27). The 4Box was dialyzed against an ATP-free pipette solution and concentrated to about 5 M with a centricone system. Seals of 6 -10 gigohms were produced on the cell membrane, and the whole cell configuration was obtained by gentle suction or voltage pulses of 0.5 V for 0.3-1 ms. The patch clamp output (Axopatch-1B, Axon Instruments) was filtered at 20 Hz. Recording was performed with patch clamp 6 and a Digi-Data 1200 interface (Axon Instruments). All traces shown were recorded at a holding potential of Ϫ40 mV.
Measurement of IP 3 -IP 3 levels were measured by a radioligand assay as described elsewhere (28). Acini in solution A were stimulated with the indicated combination of agonists for 10 s at 37°C. The reactions were stopped by addition of 20 l of ice-cold 20% perchloric acid to 200 l of samples, vigorous mixing, and incubation on ice for at least 10 min to allow precipitation of proteins. The supernatants were collected and transferred to clean tubes. Standards of IP 3 were prepared in the same manner. The perchloric acid was removed and IP 3 extracted by the addition of 0.15 ml of Freon and 0.15 ml of tri-n-octylamine. IP 3 content in the aqueous phase was measured by displacement of [ 3 H]IP 3 using microsomes prepared from bovine brain cerebella (28).
Immunocytochemistry-The immunostaining procedure was described previously (29). In brief, cells attached to glass coverslips were fixed and permeabilized with cold methanol. After removal of methanol, nonspecific sites were blocked by a 1-h incubation in blocking medium prior to incubation with 50 l of blocking medium containing control serum (controls) or the following antibodies: 1:500 dilution of the M35 mAb that recognizes all muscarinic receptors; the same mAbs that were reabsorbed with recombinant M2R; 1:250 dilution of the 5H9 mAb recognizing the M3R; 1:200 dilution of pAb specific for the M3R; 1:50 dilution of pAb1 and a 1:100 dilution of pAb2 that recognize the CCK receptor; 1:200 dilution of mAb against IP 3 R3; 1:100 dilution of pAb against IP 3 R2; 1:20 dilution of mAb against ZO1; and 1:100 dilution of pAb recognizing ZO1. After incubation with the various primary antibodies overnight at 4°C, the cells were washed three times with the incubation buffer. For the experiment with the 5H9 mAbs in Fig. 4A, two different protocols were used. The first was a standard protocol using permeabilized cells. Under these conditions, all pools of M3R were detected. Because this antibody recognizes an extracellular epitope of the M3R diagnostic of Sjögren syndrome (23), it was used to label the receptors in the plasma membrane by incubating a 0.5-ml suspension of intact cells with 50 l of 5H9 mAb for 30 min at 37°C. The cells were then washed 3 times with phosphate-buffered saline to remove excess antibodies before fixation with 0.5 ml of cold methanol. All steps postfixation were the same as with permeabilized cells. The primary Abs were detected with goat anti-rabbit or anti-mouse IgG tagged with fluorescein or rhodamine. Images were collected with a Bio-Rad MRC 1024 confocal microscope.

Receptor-specific Ca 2ϩ
Waves-Repetitive stimulation of pancreatic acinar cells with the same high concentration of agonist evokes Ca 2ϩ waves with the same initiation site and propagation pattern (6). The quantal properties of agonistmediated Ca 2ϩ release (15) raised the question of whether partial and maximal discharge of the pool generates the same or different Ca 2ϩ waves. In particular, it was of interest to determine whether the wave initiates from the same site at low and high IP 3 concentrations generated by weak and intense agonist stimulation. The properties of Ca 2ϩ waves evoked by repetitive, brief stimulation with increasing concentration of agonist (5, 50, and 500 M carbachol) are depicted in Fig. 1. Similar results were obtained in seven different cell preparations. All experiments were performed with clusters of doublet or triplet cells to ensure polarized expression of signaling protein was maintained. The arrowheads in the first image of each series show that the initiation site is constant and is independent of the IP 3 concentration used to initiate the wave. The calculated distance between initiation sites (see "Experimental Procedures") averaged 0.19 Ϯ 0.05 m (n ϭ 7). The diameter of an acinar cell was about 20 m. In addition, the pattern of wave propagation remained constant at all agonist concentrations. Increased stimulation intensity only increased the rate of Ca 2ϩ wave propagation.
Constant initiation site and propagation patterns of Ca 2ϩ waves were also observed with repetitive stimulation of another GPCR, the CCK receptor (see below). However, in previous work (6) we reported that stimulations of multiple GPCRs in the same rat pancreatic acinar cell evoked receptor-specific Ca 2ϩ waves in terms of initiation sites and propagation patterns. Cancela et al. (30) used mouse pancreatic acinar cells to suggest that Ca 2ϩ waves are stochastic and are similar for all GPCRs in a given cell. Therefore, we examined Ca 2ϩ waves evoked by carbachol and CCK stimulation in the same cell of small acinar clusters. Fig. 2 shows three examples of five experiments with similar results. The cells were stimulated with submaximal agonist concentrations to increase the temporal resolution of the waves. The white (carbachol) and magenta (CCK) arrowheads in the first image of each experiment show that, without exception, the initiation site of Ca 2ϩ waves was different between the two agonists. The calculated distance between the carbachol and CCK initiation averaged 4.5 Ϯ 1.5 m (range 1.2-9.1 m, n ϭ 5). Although the waves initiated by stimulation of the two receptors propagated along the cell periphery, differences in wave patterns were noticeable in most experiments. Close examination of the results in Fig. 5A of Cancela et al. (30) also show that in a single isolated mouse pancreatic acinar cell activation of the muscarinic and CCK receptors resulted in different initiation sites of Ca 2ϩ waves. This is particularly evident when the third image of the acetylcholine series is compared with the second image of the CCK series. The Ca 2ϩ waves evoked by the two GPCRs had different spatial propagation pattern.
Autonomous Behavior of Signaling Complexes-The constancy of Ca 2ϩ waves evoked by repetitive stimulation of the same GPCR, their independence of stimulus intensity, and the different Ca 2ϩ wave evoked by activation of different GPCRs in the same cell suggest autonomous functioning of GPCRs. We used two experimental protocols to test this prediction. In the first set of experiments we examined IP 3 production by maximal agonist concentrations added individually or in combination. Although maximal stimulation of each of the GPCRs can mobilize the entire Ca 2ϩ pool (see below), Fig. 3A shows that all combinations of agonists produced a nearly additive increase in IP 3 concentration. Hence, each GPCR can activate different pools of cellular PLC␤ or distinct portions of a single pool that is in excess to the total number of GPCRs.
In the second protocol we used the RGS domain of RGS4 (4Box) to demonstrate autonomous functioning of the GPCRs in the same cell. The 4Box lacks receptor recognition domain and is a poor inhibitor of signaling in resting cells (31). The 4Box binds to activated ␣-subunits of G proteins (25,32) and is nearly as effective as full-length RGS4 in stimulating GTPase activity of G␣ q (31). Consequently, upon cell stimulation the 4Box is recruited to signaling complexes to inhibit their activity and remains trapped in the complexes for as long as the cells are stimulated (see Ref. 27 for further details on the mode of action of the 4Box). We took advantage of the mode of interaction of the 4Box with GPCR signaling complexes to examine whether maximal activation of one GPCR type is sufficient to recruit 4Box to all GPCRs complexes, or stimulation of each GPCR complex is needed to inhibit signaling of the stimulated complex. Fig. 3, B and C, shows that stimulation of cells with 1 mM carbachol resulted in complete mobilization of the Ca 2ϩ pool, as evident from the failure of 10 nM CCK8 and 100 nM bombesin (BS) to increase [Ca 2ϩ ] i in cells continuously stimulated with carbachol. When the cells were infused with 100 nM 4Box, stimulation with 1 mM carbachol resulted in a maximal initial signal that terminated rapidly and returned to base line at continuous stimulation with carbachol, due to receptor-dependent recruitment and trapping of 4Box (27) . Fig. 3, D and E, shows that at continuous stimulation with carbachol the cells responded with a maximal signal to CCK8 (n ϭ 15) and BS stimulation (n ϭ 4). The response to CCK8 and BS had the same kinetics as that to carbachol, with a rapid onset and inhibition, and it remained inhibited for the duration of cell stimulation. This behavior is best explained if the various receptors are coupled to separate pools of cellular G␣ q and PLC␤.

Polarized Expression of Ca 2ϩ Signaling Proteins and Receptors-Polarized and restricted expression of Ca 2ϩ channels and
Ca 2ϩ pumps in the luminal pole of epithelial cells appears to be important for initiation and propagation of polarized Ca 2ϩ waves (7,33). Co-immunoprecipitation of various Ca 2ϩ signaling proteins indicates organization of the proteins into complexes in cellular microdomains (29). GPCRs are believed to be expressed mostly in the basal membrane. However, a recent report (34) suggested the possible expression of selective GPCRs next the luminal pole in rat goblet cells. With the development of selective antibodies that recognize several GPCRs, it was of interest to re-examine localization of GPCRs in pancreatic acini. Fig. 4 demonstrates the co-localization of the type 3 muscarinic (M3Rs) and CCK receptors (CCKRs) with Ca 2ϩ -signaling proteins. Both receptors are expressed at high levels at the apical pole of pancreatic acini. Several antibodies used to localize the M3Rs revealed two cellular pools of these receptors, a plasma membrane and a Golgi-like pools. Double staining with pAb recognizing M3Rs and ZO1, a specific tightjunction protein (Fig. 4A, top panels), or IP 3 R2 (not shown) showed that M3Rs localized in a non-uniform fashion with high levels at the lateral border, in close proximity or at the tight junctions, and low levels at the basal region (n ϭ 10). However, the M35 and the 5H9 mAbs showed different patterns of localization. In permeabilized cells, the two mAbs showed strong labeling of an intracellular compartment and lower staining of the plasma membrane (n ϭ 9 for M35, n ϭ 6 for 5H9). This is particularly evident with the M35 mAb (Fig. 4, lower middle  panel). The staining with this mAb was eliminated by adsorbing the mAb with recombinant M2Rs (Fig. 4, lower left panel), demonstrating specificity of the Abs. Because the 5H9 mAbs recognize an extracellular epitope evident of Sjögren syndrome (23), the plasma membrane staining was isolated by incubating intact cells with the Ab prior to cell fixation and permeabilization. Fig. 4A, lower right panel, shows that the M3Rs receptors are highly localized to the lateral membrane next to tight junctions (n ϭ 4).
Selective localization was also observed with the CCKRs using two different pAbs recognizing different domains of the receptor, a pAb raised against amino acids 30 -42 and a pAb raised against amino acids 411-429 of the type A CCKR (24). The two Abs specifically recognized recombinant and native CCKRs. Localization of the CCKRs at the apical pole of the lateral membrane (Fig. 4B) was similar to that of the M3Rs at the plasma membrane. The CCKRs at the plasma membrane were concentrated very near the IP 3 R3 at the apical pole (n ϭ 10). The anti-CCKRs Abs did not recognize a Golgi pool of this receptor.
Quantal Behavior and an All-or-None Ca 2ϩ Release-The independence of the Ca 2ϩ waves initiation site and propagation pattern of IP 3 concentration (Fig. 1) offered us the opportunity to examine whether local Ca 2ϩ release is due to an all-or-none release from a compartmentalized pool (Fig. 5, model A) or a partial release from a continuous pool (Fig. 5, model B). The experiments used to distinguish between the two models and their predicted outcome according to each model are depicted in [Ca 2ϩ ] i Waves and an All-or-None Ca 2ϩ Release trast, model B assumes that at continuous stimulation, the activity of all the IP 3 Rs channels is adapted to the lower Ca 2ϩ content in the entire pool to terminate Ca 2ϩ release so that Ca 2ϩ permeability of the pool is similar to that of resting cells.
The first protocol we used is an agonist concentration jump after state c, long after return of [Ca 2ϩ ] i to near resting level. Model A predicts that the agonist (or IP 3 ) concentration jump will generate a Ca 2ϩ wave with the same initiation site and pattern. Model B predicts generation of Ca 2ϩ waves from random initiation sites and propagation pattern, because of the adaptation of the IP 3 Rs that generated the first wave. Fig. 6 shows the results of such experiments with carbachol and CCK. It is clear that the Ca 2ϩ waves initiated by low agonist and the concentration jump have the same initiation site and propaga- [Ca 2ϩ ] i Waves and an All-or-None Ca 2ϩ Release tion pattern. In five experiments with carbachol and four experiments with CCK, the average distances between the two initiation sites were 0.16 Ϯ 0.03 and 0.10 Ϯ 0.03 m, respectively. Localized generation of IP 3 due to the polarized expression of the muscarinic receptors is likely to contribute to the constancy of the wave. Hence, in addition to testing the first prediction in Fig. 5, the constancy of the wave observed with the protocol of Fig. 6 lend further support for the autonomous behavior of signaling complexes In the second protocol, the second Ca 2ϩ wave was initiated by applying the concentration jump at the end of Ca 2ϩ release evoked by the low agonist and before any reduction in [Ca 2ϩ ] i due to IP 3 Rs channel adaptation. Model A predicts a second Ca 2ϩ wave with different initiation site, due to discharge of all the Ca 2ϩ from the pool mobilized by low agonist. Model B predicts a second Ca 2ϩ wave from the same initiation site because Ca 2ϩ is released prior to adaptation of the IP 3 Rs activated by the first stimulus. The trace in Fig. 7 shows that this particular cell was exposed to the supermaximal concentration of 5 mM carbachol at 1.63 s after simulation with 5 M carbachol. The concentration of 5 mM was used for the second stimulus, which was applied by rapid injection from a needle adjacent to the cell, to minimize the delay between agonist application and cell stimulation. The images in Fig. 7 show the results of four of five independent experiments with similar results. In contrast with the prediction of model B, but as predicted by model A, in all experiments the second Ca 2ϩ wave evoked by the agonist concentration jump was initiated at a   FIG. 6. Successive stimulation by a delayed agonist concentration jump. A, the cell was stimulated with 5 M carbachol, and after return of Ca 2ϩ to basal levels carbachol concentration was rapidly increased to 0.5 mM. B, the cell was stimulated with 0.25 nM CCK, and after return of Ca 2ϩ to resting levels CCK concentration was increased to 10 nM. The white arrowheads show the Ca 2ϩ wave initiation site for each stimulation period. Note the same initiation site by the low and high agonist concentrations in all experiments. different site than the first Ca 2ϩ wave. The calculated distance between the initiation sites averaged 7.3 Ϯ 1.3 m (range 3.3-10.9 m, n ϭ 5), which is statistically different (p Ͻ 0.01) from the 0.19 Ϯ 0.05 m measured after delayed application of carbachol.
A shift in the initiation site of Ca 2ϩ waves when the second wave was initiated at the end of the first Ca 2ϩ release (Fig. 7) can be either because the initiation site still contained Ca 2ϩ but the IP 3 Rs at this site became refractory to a jump in IP 3 concentration, or because all the Ca 2ϩ was released from this site by the first stimulus, and Ca 2ϩ re-uptake is needed to initiate a second Ca 2ϩ wave from the same site. The results in Fig. 8 provide direct evidence that Ca 2ϩ release is an all-ornone process and that Ca 2ϩ uptake must occur to initiate multiple Ca 2ϩ waves from the same site. In these experiments we blocked the SERCA pumps with 10 M thapsigargin (Tg) at different times after initiation of a Ca 2ϩ wave. Model A predicts that inhibition SERCA pumps of cells continuously stimulated with low agonist at any time will trigger a Ca 2ϩ wave with the same initiation site and pattern. Model B predicts that inhibition of SERCA pumps will not trigger a wave but rather will cause slow Ca 2ϩ release because of adaptation of all IP 3 Rs. Fig.   8A shows that in the absence of agonist stimulation, Tg caused a slow, peripheral to central Ca 2ϩ release with no apparent Ca 2ϩ wave or preferential release from any cellular region. As shown by the trace in Fig. 8A, [Ca 2ϩ ] i continued to increase for at least 30 s after exposure to Tg. Hence, in the absence of agonist stimulation Tg never initiated a Ca 2ϩ wave.
Very different results were obtained when the cells were stimulated with submaximal agonist concentration and allowed to reduce [Ca 2ϩ ] i back to near basal level. As shown in return of [Ca 2ϩ ] i to basal levels, much slower than the rate observed with agonist alone (compare red to all other traces in C). Application of Tg at different times after cell stimulation caused a rapid increase in [Ca 2ϩ ] i , to approximately the same level as that caused by the agonist and a slow reduction toward basal levels. Hence, Tg rapidly discharged only the Ca 2ϩ that was incorporated into the initiation site during continuous stimulation and never to a level significantly higher than the initial level caused by the agonists. This behavior indicates that Ca 2ϩ release from the initiation site must be an all-ornone process and that the initiation site reloads with Ca 2ϩ during continuous stimulation with submaximal agonist concentration (model A). Furthermore, this behavior is incompatible with model B. [Ca 2ϩ ] i Waves and an All-or-None Ca 2ϩ Release 44154 DISCUSSION In the present work we addressed two important topics in Ca 2ϩ signaling: how luminal to basal Ca 2ϩ waves can be generated in a receptor-specific manner and whether quantal Ca 2ϩ release reflects partial release from a continuous pool or an all-or-none release from a compartmentalized pool. Polarized exocrine cells are a good model system to address these questions because of the polarized nature of their Ca 2ϩ signals (3), the polarized expression of high levels of Ca 2ϩ transport proteins in the apical pole (7,9,29,33,36), and the generation of receptor-specific Ca 2ϩ signals in these cells (6,37). Our results indicate that signaling specificity in pancreatic acinar cells is aided by polarized expression and autonomous functioning of GPCRs and that quantal Ca 2ϩ release is due to an all-or-none Ca 2ϩ release from a compartmentalized Ca 2ϩ pool.
Signaling specificity is a central topic in cell signaling (38). Several mechanisms have been shown to contribute to generation of specific GPCR-dependent Ca 2ϩ signals, which include selective phosphorylation of IP 3 Rs (39), differential couplings of Ca 2ϩ release to Ca 2ϩ entry (40), and different involvement of cADP-ribose and NADP-mediated Ca 2ϩ release in the response (30). However, ultimately signaling specificity must reside in the communication between the receptor and downstream effectors. Indeed, RGS proteins function in a receptor-specific manner to confer receptor-specific signaling (41). Interaction of receptors with effectors requires their co-localization in cellular microdomains at sites of Ca 2ϩ release. In polarized exocrine cells Ca 2ϩ waves are always initiated at the apical pole. Whereas expression of all known IP 3 R isoforms (7,9), specific isoforms of SERCA pumps (33), and plasma membrane Ca 2ϩ ATPase pumps (36) has been demonstrated, it remains a puzzle as to how IP 3 can be generated at this cellular microdomain because GPCRs are believed to reside largely at the basal pole. The localization of the M3Rs and CCKRs found at the present work clarifies this issue. The use of three anti-M3Rs and two anti-CCKRs, which recognize different epitopes, showed that both GPCR types are expressed at high level at close proximity or at the tight junctions. This localization removes the need to assume long range diffusion of IP 3 from the basal to the apical pole to initiate Ca 2ϩ waves. Moreover, such localization of GPCRs is likely to contribute to the constancy of the Ca 2ϩ waves generated by repetitive stimulation of the same GPCR complex.
In the present work we confirmed and extended the finding of receptor-specific Ca 2ϩ waves by various GPCRs expressed in the same cell. By using the unique property of recruitment and trapping of the 4Box within signaling complexes (27), we demonstrated autonomous functioning of GPCRs. Furthermore, production of IP 3 by multiple GPCRs was nearly additive. The combined findings indicate that different GPCRs communicate with separate portions of the cellular pool of G␣ q and PLC␤ to generate IP 3 at separate cellular microdomains. The finding that the initiation site of Ca 2ϩ waves was independent of IP 3 concentration, remaining constant when initiated by low (weak stimulation) or high IP 3 concentration (intense agonist stimulation), indicates that each GPCR communicates with a separate portion of the Ca 2ϩ pool at the apical pole to generate Ca 2ϩ waves with distinct initiation sites. Communication with a distinct portion of the Ca 2ϩ pool may also reflect localized generation of IP 3 at the locus from which Ca 2ϩ waves are initiated. Measurement of IP 3 in single pancreatic acinar cells is needed to test this point. At present, this is not feasible.
Since the discovery of the quantal behavior of Ca 2ϩ release (15), two models were proposed as possible mechanisms of this phenomenon (see Fig. 5). The first model proposes that the intracellular Ca 2ϩ pool is compartmentalized with respect to Ca 2ϩ release, and the IP 3 R Ca 2ϩ release channels in the different compartments have variable affinity to IP 3 and Ca 2ϩ release from individual compartments in an all-or-none process (15,20). The second model proposed that the intracellular Ca 2ϩ pool is continuous, and the affinity of all IP 3 Rs to IP 3 is the same but is sensitive to Ca 2ϩ content in the stores, and Ca 2ϩ release is incremental due to rapid inactivating adaptation of the IP 3 Rs (21). The evidence in favor of the adaptation model is the finding that stored Ca 2ϩ regulates the affinity of the IP 3 Rs for IP 3 in virtually all cell types examined (42), and a recent report (22) indicates that two rapid successive increases in IP 3 can liberate Ca 2ϩ from the same site. The first evidence (42) is not mutually exclusive with the all-or-none model, and in the second case (22) rapid Ca 2ϩ re-uptake into the stores was not prevented by inhibition of the SERCA pumps which, as shown in the present work, can explain the second Ca 2ϩ release event.
Previous evidence in support of the all-or-none model was developed by measurement of Ca 2ϩ permeability of the stores at increasing stimulus intensity (15). Increased agonist concentration mobilized a larger fraction of the Ca 2ϩ pool. At persistent stimulation with submaximal agonist concentrations, Ca 2ϩ permeability of the mobilized fraction of the pool remained very high, whereas Ca 2ϩ permeability of the immobilized fraction remained very low, identical to that measured in resting cells (15). Another important finding in support of the all-or-none model is the variable sensitivity of IP 3 to Ca 2ϩ release from different regions of pancreatic acinar cells (3,11). The adaptation model predicts uniform IP 3 sensitivity throughout the cell.
In the present work we provide what we believe is strong evidence in support of the all-or-none model by first showing that the initiation site of Ca 2ϩ waves is independent of IP 3 concentration. The implication of this finding is that there must exist a sub-pool that is more sensitive to IP 3 than the remaining cellular Ca 2ϩ pool, and this pool is always discharged first upon cell stimulation. This provides additional evidence that the cellular Ca 2ϩ pool is not uniform with respect to sensitivity to IP 3 . Next, we used an agonist concentration jump to apply two consecutive rapid or delayed applications of IP 3 . This protocol is equivalent to the two consecutive IP 3 application protocols used to conclude multiple Ca 2ϩ release events from the same site (22). Two consecutive applications of IP 3 in pancreatic acini released Ca 2ϩ from two separate sites when the second application was applied immediately at the end of Ca 2ϩ release by the first application (Fig. 7). The difference between our results and those obtained in Xenopus oocytes (22) can be due to species differences or partial reloading of the just released pool with Ca 2ϩ .
The adaptation model requires that at continued exposure to the same IP 3 concentration, the IP 3 R channels adapt, becoming refractory to IP 3 , and Ca 2ϩ permeability of the pool is reduced back to resting level to terminate Ca 2ϩ release. We used Tg to provide compelling evidence that this is not the case, at least in pancreatic acini. In fact, the Ca 2ϩ pool liberated by submaximal receptor stimulation reloads with Ca 2ϩ with continuous exposure to a constant agonist concentration. Furthermore, and most important, the Ca 2ϩ permeability of the released and reloaded pool remains high, resulting in a Tg-evoked Ca 2ϩ wave. The Tg-evoked Ca 2ϩ wave had the same initiation site and propagation pattern as the Ca 2ϩ wave evoked by the agonist (Fig. 8). Hence, the maintained high Ca 2ϩ permeability of the liberated sub-pool, reloading of the sub-pool at continued stimulation, and the shift in the initiation site of Ca 2ϩ wave observed at rapid agonist concentration jump are all compatible only with an all-or-non model of Ca 2ϩ release. Therefore, we conclude that the principal mechanism behind the quantal behavior of Ca 2ϩ release is an all-or-none Ca 2ϩ release from highly compartmentalized intracellular Ca 2ϩ pool.