Organelle selection determines agonist-specific Ca2+ signals in pancreatic acinar and beta cells.

How different extracellular stimuli can evoke different spatiotemporal Ca2+ signals is uncertain. We have elucidated a novel paradigm whereby different agonists use different Ca2+-storing organelles ("organelle selection") to evoke unique responses. Some agonists select the endoplasmic reticulum (ER), and others select lysosome-related (acidic) organelles, evoking spatial Ca2+ responses that mirror the organellar distribution. In pancreatic acinar cells, acetylcholine and bombesin exclusively select the ER Ca2+ store, whereas cholecystokinin additionally recruits a lysosome-related organelle. Similarly, in a pancreatic beta cell line MIN6, acetylcholine selects only the ER, whereas glucose mobilizes Ca2+ from a lysosome-related organelle. We also show that the key to organelle selection is the agonist-specific coupling messenger(s) such that the ER is selected by recruitment of inositol 1,4,5-trisphosphate (or cADP-ribose), whereas lysosome-related organelles are selected by NAADP.

It has become clear that a primary factor governing agonist specificity is the release of Ca 2ϩ from intracellular stores, a process that encompasses multiple channel families regulated by the second messengers inositol 1,4,5-trisphosphate (IP 3 ), cyclic ADP-ribose (cADPR), and nicotinic acid adenine dinucleotide phosphate (NAADP). By differential recruitment of second messenger complements, different agonists could, in principle, elicit different Ca 2ϩ signals, and several cellular systems have indeed been shown to exploit messenger diversity, e.g. pancreatic acinar cells (3).
However, the use of different messengers per se is not sufficient to evoke different signals; only if these systems are nonequivalent and differ in their properties, regulation, or spatial distribution will this hold. In spatial terms, IP 3 and cADPR mobilize the endoplasmic reticular (ER) Ca 2ϩ store in nearly all cell types (1,4), although an outstanding issue has been the identity and location of the NAADP-sensitive store. Only recently has this been characterized in sea urchin eggs as a lysosome-related organelle (5), with its mammalian counterpart remaining elusive beyond an isolated report that secretory vesicles support NAADP-induced Ca 2ϩ release in permeabilized cells (6).
Given the uncertainties surrounding the role and properties of different potential Ca 2ϩ stores, we therefore show for the first time that different agonists generate their specific signals by signaling through Ca 2ϩ mobilization from different intracellular stores (organelle selection). This is determined by their second messenger complements, with NAADP-linked agonists coupling to lysosome-related organelles in mammalian cells and those linked to IP 3 /cADPR coupling to ER pools.

EXPERIMENTAL PROCEDURES
Cell Preparation-To obtain pancreatic acinar cells, pancreata were excised from male CD1 mice 8 -10 weeks old, and small clusters of pancreatic acinar cells were prepared by collagenase digestion as described previously (7). MIN6 cells were cultured in Dulbecco's modified Eagle's medium (25 mM glucose) supplemented with 10% fetal calf serum, 2 mM glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, and 50 M ␤-mercaptoethanol equilibrated with 5% CO 2 and 95% air at 37°C. Twenty hours before each experiment, cells were placed in low glucose Dulbecco's modified Eagle's medium (5 mM glucose).
Ca 2ϩ Imaging-Both acinar and MIN6 cells were seeded onto polylysine-coated number 1 glass coverslips and loaded with 1-5 M fura-2 acetoxymethyl ester (fura-2/AM) for 60 min at room temperature. Acinar cells and MIN6 cells were maintained in buffer of the following compositions: for acinar cells (in mM), 140 NaCl, 4.  ), on an inverted Zeiss 35 Axiovert microscope, and imaged with a conventional epifluorescence system, using Metafluor software (Universal Imaging). Cells were excited alternately with 340 and 380 nm light (emission 510 nm), and ratio images of clusters were recorded every 4 -5 s, using a 12-bit CCD camera (MicroMax, Princeton Instruments). All experiments were conducted at room temperature for acinar cells and at 37°C for MIN6 cells.
Imaging Lysosomes-Acidic organelles in both cell types were labeled by incubating cells with 50 nM Lysotracker Red for 20 min at room temperature. Labeling was visualized after 20 -40 min of removing excess dye using a Leica TCS NT laser scanning confocal microscope (excitation 568 nm, emission Ͼ590 nm).
Flash Photolysis-Acinar cells and MIN6 cells were pressure-microinjected (Femtojet, Eppendorf) with Oregon Green BAPTA-1 dextran (OGBD, final concentration of 20 and 5 M, respectively) with caged compounds. In acinar cells the Ca 2ϩ -sensitive dye was imaged (excitation 490 nm, emission 530 nm) as mentioned, and the caged compounds were photolysed with an XF-10 arc lamp (HI-TECH Scientific, the ultraviolet flash efficiency was 0.5-1%). On the other hand, MIN6 cells were imaged by laser-scanning confocal microscopy (Leica TCS NT), and caged compounds were photolysed with an ultraviolet laser (efficiency of uncaging of ϳ50%). Images were processed using Metamorph software (Universal Imaging). Ca 2ϩ concentration is given as the ratio F/F o where F o is the fluorescence before stimulation, and F the fluorescence at a given time. Changes in Ca 2ϩ concentration are given as increases in the mentioned ratio (⌬F/F o ).
Statistical Analysis-Data are presented as means Ϯ S.E. Statistical significance were evaluated by paired Student's t test and, for multiple comparisons, analysis of variance followed by Fisher's Least Significant Difference test (Statview, Abacus Concepts).
Materials-Caged IP 3 was from Calbiochem. Caged cADPR, Fura-2/ AM, OGBD, and Lysotracker Red were from Molecular Probes, and collagenase was from Worthington. Caged NAADP was synthesized essentially as described previously (8). All other reagents were from Sigma.

RESULTS
Of all the mammalian cell models used for investigating agonist-specific Ca 2ϩ signals, one of the most extensively studied non-excitable cells is the pancreatic acinar cell. The gastrointestinal peptides cholecystokinin and bombesin, and the neurotransmitter acetylcholine, evoke subtly different Ca 2ϩ signatures resulting in differential control of fluid secretion, exocytosis, and trophic effects (3). We first addressed whether these agonists differentially recruited lysosome-related or endoplasmic reticular Ca 2ϩ stores by using inhibitors to selectively abrogate Ca 2ϩ storage in each organelle; bafilomycin A1 inhibits the vacuolar H ϩ -ATPase responsible for the proton gradient that drives lysosomal Ca 2ϩ uptake (9), whereas thapsigargin directly blocks the ER Ca 2ϩ -ATPase (SERCA) (10).
Whether or not bafilomycin A1 had an effect upon Ca 2ϩ oscillations depended upon the agonist used. Cholecystokinininduced Ca 2ϩ oscillations were profoundly inhibited, whether bafilomycin A1 was added before (Fig. 1a) or after (Fig. 1b) the agonist. By contrast, bafilomycin A1 had very little effect upon acetylcholine-or bombesin-stimulated oscillations (Fig. 1, c-f). To confirm that bafilomycin A1 was indeed acting at a lysosome-related organelle, we also eliminated such stores with glycyl-phenylalanine 2-naphthylamide (GPN), a substrate of lysosomal cathepsin C whose cleavage results in osmotic lysis (5,11). Essentially identical results were obtained with GPN, notably the selective block of the response to cholecystokinin over acetylcholine or bombesin ( Fig. 1, g-l). Moreover, in cells where cholecystokinin-induced oscillations were blocked by GPN, subsequent addition of acetylcholine could rescue Ca 2ϩ spiking, highlighting the specificity of the response (Fig. 1m). The data show that, of the agonists tested, cholecystokinin is unique in recruiting a lysosome-related organelle that can function as a Ca 2ϩ store.
The ER, on the other hand, is the established Ca 2ϩ reservoir for many G protein-coupled receptors, including those for acetylcholine and bombesin. In pancreatic acinar cells, these stimuli are well documented to release Ca 2ϩ from the ER using both IP 3 and ryanodine receptors (12,13). Furthermore, cholecystokinin-induced Ca 2ϩ signals were also confirmed to derive from the ER as evidenced by the marked inhibition by thapsigargin (Fig. 1n). Together, the evidence supports cholecystokinin recruiting both lysosomes and ER, whereas acetylcholine and bombesin only target the ER in order to generate [Ca 2ϩ ] i signals.
An obvious issue is whether this differential organellar recruitment is specific to acinar cells or a universal blueprint for other mammalian cell types and stimuli. In choosing another model, the pancreatic ␤ cell, we opted for a system with very different properties from the acinar cell, the new one being an excitable cell in which Ca 2ϩ signals are elicited by nutrients as well as by G protein-coupled agonists (14,15). In the ␤ cell line, MIN6 (16), we compared three different stimuli, glucose, acetylcholine, and K ϩ , with regard to their relative sensitivities to bafilomycin A1 or thapsigargin. Mechanistically, glucose metabolism generates intracellular signals that culminate in a complex interplay between Ca 2ϩ influx and Ca 2ϩ release from intracellular stores (17); muscarinic acetylcholine receptors are G protein-coupled to phospholipase C (15), whereas high K ϩ depolarizes the plasma membrane to induce voltage-operated Ca 2ϩ entry (18,19) In MIN6 cells, the sensitivity of Ca 2ϩ responses to bafilomycin A1 was, like acinar cells, highly dependent upon the stimulus. Remarkably glucose responses were profoundly inhibited by a preincubation with bafilomycin A1 (Fig. 2, a and b), whereas neither acetylcholine nor K ϩ stimulation was affected (Fig. 2, d and e and g and h). That the acidic stores of the glucose response were lysosome-related was confirmed by the inhibition by GPN (data not shown). Moreover, these results confirm the specificity of bafilomycin and GPN because neither interacts with the IP 3 -calcium release pathway (acetylcholine) nor calcium influx (K ϩ ). Remarkably, the effects of interfering with ER stores with thapsigargin were almost the mirror of those with bafilomycin A1. Glucose-induced Ca 2ϩ signals were not inhibited by ER depletion (Fig. 2c) but rather appeared to be potentiated because thapsigargin greatly reduced the lag phase and eliminated the initial fall in basal [Ca 2ϩ ] i . Similarly, responses induced by K ϩ were also slightly potentiated (Fig.  2i), attesting to the role of the ER as a Ca 2ϩ sink in ␤ cells (20). On the other hand, the ER appeared to play a major role during acetylcholine-induced Ca 2ϩ mobilization because responses were completely eliminated by thapsigargin (Fig. 2f). Taken together, the data in this cell type support a model of the reciprocal recruitment of different organelles where metabolic activation is heavily reliant upon lysosome-related Ca 2ϩ stores, contrasting with an exclusive ER role in response to neurotransmitter.
Next we provide a mechanism to couple particular extracellular stimuli to specific intracellular organelles with the appropriate fidelity. Interestingly, our data show an absolute correlation between those stimuli that recruit lysosome-related organelles and those known to utilize NAADP as a Ca 2ϩ -mobilizing messenger (21), i.e. cholecystokinin in pancreatic acinar cells (22) and glucose in ␤ cells (16). The agonists that fail to require acidic stores are well known to couple to IP 3 and/or ryanodine receptors that mobilize ER Ca 2ϩ stores (13,23,24). We therefore tested whether NAADP was the unique link to acidic stores, with IP 3 and/or cADPR showing a preference for non-acidic (ER) stores.
First, in pancreatic acinar cells, photorelease of IP 3 or cADPR from their caged precursors evoked robust, monotonic Ca 2ϩ transients in control cells (Fig. 3, a and b) comparable in magnitude to the subsequent response to acetylcholine. Elimination of lysosomal Ca 2ϩ storage by preincubation with GPN had no effect upon the magnitude of the [Ca 2ϩ ] i rise in response to either IP 3 or cADPR (or the following acetylcholine responses) (Fig. 3, d and e). By contrast, GPN profoundly inhibited the Ca 2ϩ oscillations following uncaging of NAADP (Fig. 3, c and f). Note that the NAADP-induced Ca 2ϩ spikes are initially small and became progressively amplified by Ca 2ϩ -induced Ca 2ϩ release mechanism through the recruitment of IP 3 and ryanodine receptors (3,22,(25)(26)(27). In agreement with the effects of GPN, bafilomycin A1 displayed an identical and selective block of NAADP-induced over IP 3 -induced responses (data not shown, n ϭ 4). The very fact that the Ca 2ϩ responses to second messengers alone are inhibited strongly suggests that bafilomycin A1 and GPN are working downstream of NAADP, and not at an upstream element of the signaling cascade initiated by agonist. We conclude that only NAADP couples to the lysosome-related Ca 2ϩ store in acinar cells, whereas cADPR and IP 3 couple to the ER.
Our hypothesis is also supported in experiments with the ␤ cell system; in these cells NAADP-dependent Ca 2ϩ release occurs via specific binding sites (16). Just as glucose and acetylcholine manifest a reciprocal dependence upon lysosomes and ER, so NAADP and IP 3 displayed this mutually exclusive pattern. NAADP photorelease stimulated a Ca 2ϩ increase that was inhibited by bafilomycin A1 but not by thapsigargin (Fig. 4,  a, c and e), whereas the converse occurred when photoreleasing IP 3 (Fig. 4, b, d, and f). Thapsigargin-induced depletion of stores profoundly inhibited IP 3 transients, which were other-wise insensitive to bafilomycin A1. Once again, NAADP selects lysosome-related stores, whereas IP 3 predominantly selects the ER. Hence, the data suggest that lysosome-related stores couple via NAADP to particular extracellular stimuli (16), cholecystokinin and glucose, respectively. Furthermore, by recruiting NAADP, agonists select a novel Ca 2ϩ store with distinct properties, distribution, and ramifications from the ER (primarily the domain of IP 3 and cADPR).
To confirm that the distribution of lysosomal Ca 2ϩ stores indeed has a bearing upon the spatial profile of the Ca 2ϩ response, we compared NAADP-mediated Ca 2ϩ release with the distribution of the organelle in live cells. Lysotracker Red labeling was markedly polarized and confined to the apical region of pancreatic acinar cells reminiscent of secretory vesicle staining (Fig. 5,  a and b) (28), which are highly related if not overlapping organelles (29). The observed pattern with Lysotracker Red faithfully reflected lysosomal staining as confirmed by the elimination of the punctate fluorescence by treatment with either GPN (Fig.  5a) or bafilomycin A1 (Fig. 5b). Interestingly, the organelle distribution coincided with the ensuing NAADP-evoked small Ca 2ϩ oscillations, which do not fully recruit Ca 2ϩ -induced Ca 2ϩ release (22,30) and were confined to the apical pole in acinar cells (Fig.  5c). In contrast, in ␤ cells Lysotracker Red comprised staining of bright punctate bodies, albeit superimposed on a diffuse fluorescent background. This punctate staining was uniformly dispersed throughout the cytoplasm but excluded from the nucleus (Fig. 6, a-c). As in acinar cells, the Lysotracker Red granular staining was eliminated by GPN (Fig. 6a) or bafilomycin A1 (Fig.  6b) but not by thapsigargin (Fig. 6c). Moreover, in MIN6 cells, the response to NAADP in ␤ cells was essentially global (Fig. 6d). Specifically, the close spatial correlation of Ca 2ϩ release and the acidic store distribution strongly imply that there is a substantive rationale for using different Ca 2ϩ -storing organelles. DISCUSSION In this present report we provide evidence for a novel mechanism that explains how agonists evoke their own characteristic Ca 2ϩ response, that of organelle selection (Fig. 7). According to this model, different agonists (even within the same cell) mobilize Ca 2ϩ in different ways in time and space by coupling to (selecting) different Ca 2ϩ -storing organelles with their own unique properties and distribution. Moreover, we reveal that a given organelle couples via a particular second messenger, such that an agonist selects organelles by recruiting the appropriate messenger complement. Although our recent results in sea urchin egg (5) provided a framework (that different messengers mobilize Ca 2ϩ from different organelles), the differential recruitment of these organelles by different agonists has never been shown.
In essence, irrespective of cell type, agonists can be divided into those that recruit lysosome-related organelles (cholecystokinin and glucose) and those that do not (acetylcholine and bombesin). Such a conclusion is drawn from pharmacological studies using two mechanistically and chemically distinct inhibitors of lysosomal function, bafilomycin A1 and GPN.
Whether added before or during Ca 2ϩ oscillations, these agents selectively inhibited responses to the former pair of agonists, while having little or no effect upon the latter. We are confident that these agents act specifically upon acidic Ca 2ϩ stores because of the following: (a) they did not indiscriminately inhibit all agonists, as evidenced by their lack of effect upon the ERcoupled acetylcholine and bombesin; (b) their site of action is likely downstream of second messengers themselves as indicated by photolysis studies (see below) and therefore not an upstream signal; (c) they do not block depolarization-induced Ca 2ϩ entry; and (d) they eliminate Lysotracker Red staining.
We have proceeded to show that cell surface receptors couple to an intracellular store type by virtue of a characteristic selecting messenger, i.e. NAADP was unique in coupling to lysosome-related organelles, whereas IP 3 /cADPR coupled to the ER. Not only does the published messenger profiles of the agonists support our hypothesis (3,15), but the sensitivity of the second messengers themselves to various store inhibitors showed an absolute agreement. Therefore, sea urchin eggs are not anomalous in having acidic stores sensitive to NAADP but rather are vindicated as an excellent model system to study mammalian Ca 2ϩ signaling. Moreover, our data are of interest in the light of a previous study (6) in permeabilized MIN6 cells suggesting that NAADP releases Ca 2ϩ from secretory vesicles, themselves an acidic organelle. It should be noted that our data differ from that by Mitchell et al. (6) because (a) we have used intact cells; (b) we show agonist (glucose) coupling to acidic stores via NAADP; and (c) in our hands NAADP predominantly releases Ca 2ϩ from a bafilomycin A1-sensitive and lysosomalrelated store.
At first sight, it might appear contradictory that selective elimination of acidic Ca 2ϩ stores has such a marked effect upon cholecystokinin when clearly there is an additional ER component ( Fig. 1) (22). More surprisingly, glucose-stimulated Ca 2ϩ signals in ␤ cells also manifest a profound sensitivity to acidic store blockade when there ought to be a substantial residual Ca 2ϩ entry component (17,18) (and perhaps an ER component) (31). Although there is currently no complete mechanistic explanation for this absolute dependence, it has been empirically determined that desensitization of the NAADP receptor by its own ligand ablates both the cholecystokinin-as well as the glucose-induced Ca 2ϩ signals (16,22). Therefore, the effects of bafilomycin A1 and GPN remain entirely consistent with the blockade of the NAADP store. For the acinar cells, it has been suggested that the ER is essential to amplify NAADP-induced Ca 2ϩ release via Ca 2ϩ -induced Ca 2ϩ release at the IP 3 or ryanodine receptors (3,25). It is currently less clear how NAADP might affect Ca 2ϩ entry in ␤ cells, but in sea urchin eggs a link between NAADP signaling and voltage-gated Ca 2ϩ channels has been suggested (32).
The agonist-specific recruitment of different organelles also has ramifications in the spatial domain. It has been clearly shown that the apical pole of pancreatic acinar cells has a high density of zymogen granules (33), with only small fingers of ER penetrating into this region, and that the ER is highly concentrated in the basolateral part of the cell (34). The distribution of acidic vesicles may display a more cell-specific pattern; certainly for pancreatic acinar cells, intense Lysotracker Red staining was confined to the apical pole, whereas MIN6 cells appeared to display a more uniform staining. Supporting our model that these are Ca 2ϩ stores, this pattern mirrored the subsequent Ca 2ϩ responses that were evoked upon uncaging NAADP; in pancreatic acinar cells the region of highest NAADP sensitivity was confirmed as the apical pole (30), whereas the MIN6 response was essentially uniform. It should be noted, however, that a previous study in permeabilized cells FIG. 4. Effect of eliminating Ca 2؉ stores upon the response to photolysis of caged second messengers in pancreatic ␤ cells. In OGBD-injected MIN6 ␤ cells, control responses to flash photolysis of caged NAADP (approximated final intracellular concentration of 100 nM) (a, n ϭ 58) or caged IP 3 (approximated final intracellular concentration of 1 M) (b, n ϭ 38) were similar. 1 M thapsigargin had no effect upon NAADP responses (c and g, n ϭ 18) but eliminated IP 3 responses (d and h, n ϭ 24). Conversely, 2 M bafilomycin A1 inhibited NAADP (e and g, n ϭ 31) but had no effect upon IP 3 (f and h, n ϭ 24).  described the basolateral pole as the region of highest NAADP sensitivity (35). We cannot currently rationalize this apparent discrepancy, but we suggest methodological differences (e.g. permeabilization, uneven distribution of compartmentalized Ca 2ϩ indicator).
In summary, we hypothesize the specific agonist-induced Ca 2ϩ mobilization from lysosomal-like acidic organelles as a universal concept in mammalian cells. The use of distinct, non-contiguous stores would allow the Ca 2ϩ levels therein to be regulated independently, e.g. by altering ATPase (Ca 2ϩ or H ϩ ) expression or activity. Indeed, the use of non-ER stores will have other advantages such as avoiding plummeting Ca 2ϩ levels in the ER lumen which affect nascent protein synthesis (36) and protein phosphorylation (37) as well as increasing cell viability by minimizing ER and mitochondria overload and hence apoptosis (38). Conversely, altering the Ca 2ϩ content of acidic stores may affect processes such as secretory vesicle fusion (39), membrane repair (29), and proteolysis (40). Furthermore, if the NAADP-sensitive Ca 2ϩ store is indeed an acidic secretory vesicle (6) or secretory lysosome (29), Ca 2ϩ is delivered precisely where required to evoke exocytosis of zymogens (41), ATP (28), or insulin (6,42), depending upon cell type, and provides another potential target for treatment of diabetes. FIG. 7. A minimal model for achieving stimulus specificity through organelle selection. Different external stimuli are coupled to different organelles via the stimulus-specific recruitment of nonpromiscuous second messenger complements. In pancreatic acinar cells, cholecystokinin (CCK) selects both lysosome-related organelles and endoplasmic reticulum by recruiting NAADP, cADPR, and IP 3 respectively. By contrast, acetylcholine (ACh) and bombesin only select the ER using IP 3 /cADPR. In ␤ cells, glucose stimulates NAADP synthesis to release Ca 2ϩ from lysosome-related organelles. Although our data do not favor a major role for the endoplasmic reticulum via IP 3 (or cADPR), we cannot formally exclude their involvement (31). Acetylcholine in ␤ cells uses only the ER/IP 3 pathway.