The secretion of ions and fluid plays a critical role in a variety of physiological activities that are vital to homeostatic mechanisms in animals. Control of such secretory activity is achieved by a range of neurotransmitters and hormones many of which act intracellularly by generating the second messenger inositol 1,4,5-trisphosphate (InsP3) and increasing cytosolic free calcium ion concentrations ([Ca2+]i). These increases are achieved by a combination of the InsP3-induced release of Ca2+ from specific intracellular stores and the activation of Ca2+ entry from the extracellular environment. The [Ca2+]i signal represents a balance between the adequate activation of components of the secretory mechanism and the avoidance of [Ca2+]i levels that are toxic to the cell. Resting [Ca2+]i is maintained low by the action of Ca2+ pumps on the intracellular stores and plasma membrane, with the result that gradients for Ca2+ movement into the cytosol from either of these two sources are very large and there is considerable potential for achieving rapid increases in [Ca2+]i. Consequently, for successful Ca2+ signalling, it is imperative that these two mechanisms of raising [Ca2+]i (i.e. Ca2+ release and Ca2+ entry) are closely integrated. Current models emphasize the activation of Ca2+ entry as a downstream result of the emptying of the intracellular stores (‘capacitative’ model). Whilst this may be true for situations of maximal stimulation, recent experiments on the oscillatory [Ca2+]i responses typical of more physiological levels of stimulation indicate a previously unsuspected, independent activation of Ca2+ entry involving arachidonic acid.
This arachidonic-acid-activated entry plays a key role, along with InsP3, in inducing the repetitive release of Ca2+ from the stores to produce the [Ca2+]i oscillations. In this way, the two components responsible for the elevation of [Ca2+]i are intimately related and their dual effects closely coordinated, resulting in the finely tuned control of agonist-induced changes in [Ca2+]i.
At first sight, it may seem that the secretion of ions and fluid is a rather specific and narrowly specialized activity that has little impact on overall homeostasis of the body. However, further consideration reveals that this activity underlies the functioning of a wide variety of activities that are vital to key homeostatic processes. These include digestion, respiration, reproduction, body-fluid balance and the sensing of the physical and chemical composition of the external environment. The fact that the simple disruption of the control of ion and fluid secretion underlies such devastating human diseases as cholera and cystic fibrosis serves to illustrate dramatically the critical role this process plays in our lives.
Given the demonstrated importance of ion and fluid secretion in homeostatic processes, and the diverse roles that it plays in the body, it is not surprising that this activity is under the control of a wide range of different neurotransmitters and hormones. Although it is clear that these regulators of secretory activity can operate intracellularly via many different signalling systems, this review will focus on only one of these systems – the system that involves the generation of inositol phosphates and increases in the cytosolic concentrations of free calcium ions ([Ca2+]i). The reasons for this emphasis are threefold. First, this particular system is probably the one most widely applicable to the various cell types involved in ion and fluid secretion, and is certainly the most intensely studied to date. Second, historically speaking, studies on a variety of secretory cells and tissues have played a major role in the development of much of our current understanding of this intracellular signalling system. For example, the very earliest studies showing that activation of receptors for certain agonists (specifically muscarinic cholinergic agonists) leads to an increase in the turnover of inositol-containing membrane phospholipids were carried out on cells from the avian salt-secreting nasal gland (Hokin and Hokin, 1964). The first direct confirmation of the causal association between inositol and the generation of Ca2+ signals, originally proposed by Michell (1975), came from work on the blowfly salivary gland (Berridge and Fain, 1979), as did the first indication of the oscillatory nature of these responses (Rapp and Berridge, 1981). The watershed discovery of InsP3 as the all-important link between the agonist-activated turnover of phosphoinositides, which takes place in the plasma membrane, and the mobilization of Ca2+ from stores in the cytosol was performed on pancreatic acinar cells (Streb et al. 1983), and much of our current thinking about the nature and control of agonist-activated Ca2+ entry in non-excitable cells comes from studies on lacrimal and parotid exocrine cells (Putney, 1990). Finally, from a homeostatic viewpoint, it is clear that this signalling system is also the one that poses the most problems for the cell (see below).
Mechanism of secretion
Despite the diverse roles played by the secretion of ions and fluid in different tissues, the basic mechanism involved shows remarkable consistency. Essentially the process is one of a secondary-active transepithelial transport of chloride (Frizzell et al. 1979). In this process, a basolateral Na+ pump (Na+/K+ P-ATPase) generates a gradient for the entry of Na+ from the extracellular medium across the basolateral membrane. The entry of Na+ down this gradient is coupled to the uphill movement of Cl− into the cell via a Na+/2Cl−/K+ symporter (cotransporter) that is sensitive to inhibition by the so-called loop diuretics, such as bumetanide and furosemide. The operation of this basolateral symporter results in the accumulation of Cl− in the cell to levels significantly in excess of the electrochemical equilibrium. Under appropriate conditions, apical Cl− channels are opened, allowing the exit of Cl− to the apical surface. At the same time, the opening of basolateral K+ channels permits the recycling of this ion across the basolateral membrane. The increase in K+ conductance serves to maintain the cell in a hyperpolarized state vis-a-vis the Cl− equilibrium potential, thereby maintaining the driving force for the apical exit of Cl−. In this way, Cl− is transported from the basal to the apical side of the epithelium, generating an electrical gradient (apical negative). This gradient drives the secretion of Na+via a paracellular route (through the tight junctions) between the cells. The net secretion of both Na+ and Cl− causes water to follow osmotically, resulting in the secretion of fluid. Although the opening of apical Cl− channels and basolateral K+ channels is the primary event in the initiation of secretion, other processes are activated to sustain the response. For example, the loss of both Cl− and K+ from the cell that is associated with the initiation of secretion leads to cell shrinkage as water follows osmotically. It is this reduction in cell volume and/or the fall in cytosolic Cl− concentration that activates the basolateral symporter to maintain the secretory process (Lytle and Forbush, 1996).
Although this basic scheme is applicable to many secretory tissues, some important variations are found. For example, in certain exocrine cells, a basolateral Cl−/HCO3− antiporter (exchanger), usually coupled with a parallel Na+/H+ antiporter, either assists or replaces the Na+/2Cl−/K+ symporter for the uptake of Cl−. In other, more highly specialized tissues, the basic plan is even further modified and other transporters are involved such as the apical Cl−/HCO3− antiporter in the HCO3−-secreting cells of the pancreatic ducts and the apical K+/H+ P-ATPase found in the acid-secreting parietal cells of the stomach. A rather different system is found in certain insect secretory epithelia (e.g. lepidopteran midgut and Malpighian tubules), as well as in specific renal epithelia in vertebrates, where the transport of a variety of ions is secondarily coupled to gradients generated by an apical proton-translocating V-ATPase (Harvey, 1992; Wieczorek, 1992).
The calcium signal
As noted above, many different receptors on the surface of secretory cells exert their effects on the initiation and control of cell function via pathways involving the elevation of [Ca2+]i. It is now known that these increases are achieved by a combination of the InsP3-induced release of Ca2+ from specific intracellular stores and the activation of Ca2+ entry from the extracellular medium. In recent years, considerable advances have been made in detailing the processes involved, particularly the generation of InsP3 and its role in the release of Ca2+ from specific intracellular stores. The generation of InsP3 results from the hydrolysis of an inositol-containing membrane phospholipid, phosphatidylinositol 1,4,5-bisphosphate, in a reaction catalyzed by a receptor-activated phosphoinositide-specific phospholipase C (Berridge and Irvine, 1989; Berridge, 1993). The resulting products are the second messengers inositol 1,4,5-trisphosphate (InsP3) and diacylglycerol. InsP3 is water-soluble and diffuses into the cytosol where it binds to specific receptors on certain intracellular Ca2+ stores, resulting in the mobilization of Ca2+ from these stores. The precise nature of the agonist-sensitive intracellular Ca2+ stores is unclear, other than the fact that they represent some component of the endoplasmic reticulum (Pozzan et al. 1994). In contrast, considerable information now exists regarding the nature of the InsP3 receptors and their functioning (Mikoshiba, 1993; Shuttleworth, 1994a; Taylor and Traynor, 1995). It is known that the InsP3 receptor protein contains both InsP3-binding and Ca2+-release properties, so the receptor can be considered as an example of an intracellular ligand-gated ion channel (Ferris et al. 1989; Mayrleitner et al. 1991). Isolation, cloning and sequencing of the receptor has shown it to consist of a tetramer of approximately 260 kDa subunits. The InsP3-binding domain resides at the extreme N-terminal region (last 400 residues) of the protein, whilst the C-terminal portion contains six putative membrane-spanning domains that make up, or at least contain, the Ca2+ channel (Mignery and Südhof, 1990). Binding of InsP3 induces a large conformational change in the protein (Mignery and Südhof, 1990), which presumably underlies the transduction process responsible for opening of the Ca2+-channel portion. The region between the InsP3-binding domain and the putative Ca2+-channel domain constitutes the so-called ‘coupling’ domain, some 1400–1500 amino acids in length, which lies in the cytoplasm of the cell. It is now clear that this coupling domain is the site of a variety of modulatory actions of potential functional significance, including phosphorylation by protein kinase A (Supattapone et al. 1988). Studies of the conduction properties of the isolated receptor reconstituted in lipid bilayers indicate that the channel is a rather non-specific cation-selective channel (Bezprozvanny and Ehrlich, 1994) with a single-channel conductance of 20–26 pS with 50–53 mmol l−1 Ca2+ as the current carrier (Maeda et al. 1991; Bezprozvanny et al. 1991; Watras et al. 1991). Addition of InsP3 induces complex channel openings with four conductance states that appear as multiples of the basic 20 pS state, with level 3 (60 pS) predominating. Recent studies have revealed that at least three (possibly as many as five) separate genes can encode distinct InsP3 receptor subtypes. The three subtypes whose distribution is most widespread and whose full sequence is known (subtypes 1–3) show significant sequence diversity, particularly in the Ca2+-channel region of the protein and in the so-called coupling domain (Yamamoto-Hino et al. 1994). Such structural diversity raises the possibility of functional implications – an idea that has assumed increased significance with the demonstration of the specific tissue distribution of receptor subtypes and of unique patterns of subtype expression during differentiation (Fujino et al. 1995; Sugiyama et al. 1994). However, the true physiological implications of this subtype-specific expression remain unclear.
Clearly, although many important details remain to be described, recent years have seen considerable advances in our understanding of the InsP3 receptor and its role in Ca2+ release. However, because the intracellular stores only contain a finite amount of Ca2+, the activity of various mechanisms removing Ca2+ from the cytosol (e.g. Ca2+ transporters on the plasma membrane and other organelles) will mean that such release can only result in a transient increase in [Ca2+]i. Sustained [Ca2+]i signals are entirely dependent on an agonist-activated Ca2+ entry and, as will be discussed below, far less is known about this component of the overall [Ca2+]i signal. Despite the uncertainties in the underlying mechanisms, the net result of these two components, Ca2+ release and Ca2+ entry, is the classic biphasic [Ca2+]i signal – an initial rapid rise, largely caused by the InsP3-dependent release of Ca2+ from the stores, followed by a decline to a new sustained elevated [Ca2+]i ‘plateau’, which is entirely dependent on the enhanced entry of extracellular Ca2+.
The role that these changes in [Ca2+]i play in the control of ion and fluid secretion has also been extensively studied and is now fairly clear (Petersen, 1992). Apical Cl− channels and basolateral K+ channels sensitive to increases in [Ca2+]i have been described from a variety of different secretory cell types, and it is the opening of these channels that is responsible for the initiation of secretory activity (see above). For example, in secretory cells from the avian nasal gland, the Ca2+-activated K+ channels are of the so-called ‘maxi-K+’ type – large-conductance, Ca2+- and voltage-dependent channels inhibitable by tetraethylammonium (TEA+) and by charybdotoxin (Martin and Shuttleworth, 1994a; Wu et al. 1996). The Ca2+-activated Cl− channels are kinetically and pharmacologically distinct from the cyclic-AMP-activated Cl− channels which are of the so-called ‘cystic fibrosis transmembrane conductance regulator’ (CFTR) type, although both co-exist within the same cell (Martin et al. 1994).
Spatial and temporal features of Ca2+ signals
In recent years, techniques that allowed changes in intracellular [Ca2+] to be studied within individual cells have revealed that [Ca2+]i signals in non-excitable cells are far more complex and subtle than the simple elevation of [Ca2+]i to a new sustained level as was originally supposed. These complex responses include repetitive oscillations or waves of [Ca2+]i (Berridge and Irvine, 1989; Berridge, 1990; Lechleiter et al. 1991; Rooney and Thomas, 1993). Increases in agonist concentration generally increase the frequency of these oscillations, but with little change in their amplitude, suggesting that the original continuously graded (analog) signal of the hormone or neurotransmitter is converted intracellularly into a frequency-encoded (digital) signal. Yet another manifestation of the spatial and temporal subtlety of [Ca2+]i signals that applies particularly to polarized secretory cells is the presence of highly localized so-called ‘trigger zones’ in cells where agonist-induced changes in [Ca2+]i occur without affecting other regions of the cytosol (Kasai and Augustine, 1990; Toescu et al. 1992). The potential physiological implications of such trigger zones are fairly clear in that they will permit the targetted activation of appropriate effectors within specific areas in the cell. Overall, the current consensus is that these more complex responses, which are characteristically produced by low agonist concentrations, represent the true physiologically relevant signals as they probably reflect the normal situation in the intact tissue. As a consequence, it has become increasingly clear that the specific spatial and temporal patterns of [Ca2+]i signalling in exocrine, as well as other, cells have a major impact on the effective control of cell activity.
Several different models for [Ca2+]i oscillations have been proposed, including those in which the concentration of InsP3 oscillates (Meyer and Stryer, 1991; Cuthbertson and Chay, 1991) and those in which InsP3 concentrations are constant and the oscillations result from the repetitive Ca2+ release from the InsP3-sensitive stores (De Young and Keizer, 1992; Atri et al. 1993) or via a Ca2+-induced Ca2+ release (CICR) from InsP3-insensitive stores (Goldbeter et al. 1990). Although components of each of these models may apply to the responses in different cell types, it is clear that a requirement for oscillating concentrations of InsP3 is not applicable to the oscillatory responses of many cells where, for example, oscillations can be produced by the introduction of non-hydrolyzable InsP3 analogs (Wakui et al. 1989). Similarly, a requirement for InsP3-insensitive pools of releasable Ca2+, such as those sensitive to ryanodine and caffeine, is far from absolute as oscillations are commonly seen in cells where such stores are absent (e.g. avian nasal gland cells; Martin and
Shuttleworth, 1994b). Although the details vary, the ‘minimal’ models for such examples rely on the unique kinetics of the InsP3 receptor and its sensitivity to [Ca2+]i (Fig. 1A). Increases in cytosolic [Ca2+] have a biphasic effect on InsP3-induced Ca2+ release – initially enhancing Ca2+ release and subsequently decreasing it (Iino, 1990; Finch et al. 1991). This behavior is duplicated by the isolated receptor incorporated into lipid bilayers. Here, the curve for the Ca2+-dependency of channel open probability in the presence of InsP3 is ‘bell-shaped’, with a peak at approximately 250–300 nmol l−1 Ca2+ (Bezprozvanny et al. 1991). The consequence is that, initially, Ca2+ released as a result of the action of InsP3 will exert a positive feedback on further Ca2+ release. Then, as cytosolic [Ca2+] rises further (particularly in the region close to the open InsP3 receptor), it will have a negative feedback action, effectively closing the channel and allowing the stores to refill. This cycle can then be repeated to result in a train of oscillations in [Ca2+]i. Computer models of varying complexity based on such biphasic behavior can successfully simulate the different characteristics of the oscillatory [Ca2+]i responses observed in a variety of cells with reasonable accuracy (De Young and Keizer, 1992; Atri et al. 1993).
The precise basis for the biphasic effects of Ca2+ on the InsP3 receptor that are critical to such models is unclear. The possible involvement of phosphorylation–dephosphorylation events has been suggested (Zhang et al. 1993) but this cannot account for all the data, particularly those obtained with the purified receptor. One possibility is that the biphasic Ca2+ effect may be related to the finding that increases in [Ca2+] reversibly switch the receptor to a non-conducting (or poorly conducting) configuration with a high affinity for InsP3 (Pietri et al. 1990). There is considerable evidence for such Ca2+-dependent conformational switches affecting binding affinity (see, for example, Marshall and Taylor, 1994), but it remains to be directly demonstrated that such changes in binding affinity are associated with changes in the Ca2+ conductance of the release channel. The important point is that this property of the InsP3 receptor makes its activity sensitive to the levels of both InsP3 and Ca2+ in the cytosol, and InsP3 and Ca2+ therefore act as co-agonists at the receptor (Finch et al. 1991). Furthermore, because the positive feedback of increases in [Ca2+]i on the channel are induced more rapidly than the subsequent negative feedback effects, the Ca2+-releasing properties of the InsP3 receptor are also influenced by the rate at which cytosolic [Ca2+] is changing (Taylor and Traynor, 1995).
The role of Ca2+ entry in Ca2+ signalling
As pointed out above, the sustained, elevated ‘plateau’ of [Ca2+]i typically seen at high agonist concentrations is entirely dependent on Ca2+ entry. As such, it is readily abolished by removal of extracellular Ca2+ or by maneuvers known to interfere with Ca2+ entry, such as blockage of the entry pathway by a range of trivalent and divalent metal ions (e.g. La3+, Cd2+), by the drug SK&F 96563 or simply by reducing the driving force for Ca2+ entry by depolarizing the cell membrane potential. The role of Ca2+ entry in such responses is therefore clear. In contrast, the role of Ca2+ entry in the more complex and subtle types of response referred to above has received only scant attention. Indeed, examination of the basic models for [Ca2+]i oscillations reveal that, in each case, the underlying mechanism is essentially intrinsic to the cell and there is no absolute requirement for any Ca2+ entry. Such a notion is supported by the ability of at least some cell types to continue oscillating in the absence of extracellular Ca2+. Nevertheless, to a greater or lesser degree, Ca2+ oscillations in most cells are ultimately sensitive to the removal of extracellular Ca2+. The variability seen probably reflects differences in the activities of the transporters removing cytosolic Ca2+ at the plasma membrane and in the stores, and/or simply differences in the surface to volume ratios of different cell types. In this context, it is worth pointing out that some caution should be used in the interpretation of experiments demonstrating the effects, or lack thereof, of extracellular Ca2+ on oscillations where the oscillations are recorded as changes in Ca2+-activated currents using whole-cell patch-clamp techniques. This protocol necessitates that the experimenter impose their own extrinsic level of cytosolic Ca2+ buffering, determined by the composition of the solution in the pipette, and this may or may not match the intrinsic buffering of the intact cell. That this can have dramatic effects on the apparent sensitivity of agonist-induced oscillations to extracellular Ca2+ is shown by data obtained in avian nasal gland cells (Martin and Shuttleworth, 1994b). Here, the sensitivity of oscillations in Ca2+-activated currents to extracellular Ca2+ in cells in a perforated-patch configuration, where the intrinsic cytosolic Ca2+ buffering is maintained, was markedly different from that in the same cells under standard whole-cell patch-clamp conditions. Similarly, measurements of acetylcholine-induced oscillations in pancreatic acinar cells using Fura-2 indicate an acute sensitivity to extracellular Ca2+ (Yule and Gallacher, 1988), whilst measurements from the same cells using whole-cell patch-clamp techniques report that oscillations in Ca2+-dependent currents are, at least initially, unaffected by removal of extracellular Ca2+ (Wakui et al. 1989).
In any event, even in those cells where [Ca2+]i oscillations demonstrate little or no absolute dependency on extracellular Ca2+, Ca2+ entry has been shown to exert a modulating influence on such responses, principally by affecting oscillation frequency (Kawanishi et al. 1989; Girard and Clapham, 1993). Such effects are usually assumed to result from a role of Ca2+ entry in replenishing intracellular Ca2+ lost during the oscillations and in recharging the intracellular stores (Rink and Hallam, 1989; Berridge, 1993). It was hypothesized that the interspike interval represented the time required for the stores to refill to some set point before they could be discharged by the agonist-induced levels of InsP3 (Berridge, 1993, 1994; Thorn, 1995). Reducing Ca2+ entry therefore slowed the rate at which this refilling took place, lengthening the interspike interval and reducing oscillation frequency. In experiments directly testing this hypothesis (Shuttleworth and Thompson, 1996a), we found that, under conditions designed progressively to deplete oscillating stores by modestly reducing Ca2+ entry, the oscillations continued until the stores were completely empty. Clearly, the intrinsic ability of the stores to oscillate was not dependent on their degree of loading. Furthermore, in oscillating cells, the stores apparently refill completely during the downstroke from the peak of each oscillation and are already recharged by the beginning of the interspike interval (Shuttleworth and Thompson, 1996a). The interspike interval cannot, therefore, reflect the period required to refill the oscillating stores. These findings are inconsistent with the hypothesis that the critical role of Ca2+ entry during oscillations is to recharge the oscillating stores during the interspike interval. Indeed, recent reports have indicated that the relative activity of the organellar Ca2+ pumps on the stores is much higher than that of the plasma membrane Ca2+ pumps (Camello et al. 1996). Such activity would favor Ca2+ recycling between the cytosol and the stores during oscillations, with relatively little being lost to the extracellular environment and, hence, little requirement for Ca2+ entry to recharge the stores between each spike.
If the demonstrated role of Ca2+ entry in modulating [Ca2+]i oscillations is not related to the refilling of the stores between each oscillation, then it follows that it must be involved in regulating the emptying of the oscillating stores. Consistent with this hypothesis, in avian nasal gland cells, the acute inhibition of Ca2+ entry in the continued presence of agonist (e.g. by addition of La3+ or SK&F 96365 or by depolarizing the cell in a high-K+ medium) results in the immediate cessation of the oscillations (Martin and Shuttleworth, 1994b). Importantly, this occurs without the measurable depletion of the agonist-sensitive Ca2+ pool. These data reveal two important findings. First, they show that, at low agonist concentrations (i.e. under the very conditions where oscillations are most commonly observed), agonist-generated levels of InsP3 are in themselves inadequate to result in a regenerative [Ca2+]i signal. Second, they show that a component of Ca2+ entry is critically required to drive the generation of such signals, apparently by inducing the repetitive release of Ca2+ from internal stores. This role of Ca2+ entry in the triggering of InsP3-dependent Ca2+ release (Fig. 1B) probably reflects its action as a co-agonist, with InsP3, at the InsP3 receptors (Finch et al. 1991). This effect may be restricted to stores close to the plasma membrane whose discharge then provides additional Ca2+ to induce the InsP3-dependent release at the remaining agonist-sensitive stores.
The nature of Ca2+ entry during Ca2+ signals
On the basis of the above discussions, it would seem that Ca2+ entry plays a critical role in essentially all [Ca2+]i signals whether in the form of a sustained ‘plateau’ response or in the form of oscillations. However, studies on the nature of Ca2+ entry have almost exclusively concerned the type of responses where sustained elevated [Ca2+]i signals are produced and agonist-sensitive intracellular Ca2+ stores undergo extensive and prolonged depletion (e.g. at high agonist concentrations). Despite some continued controversy, current ideas concerning the activation of Ca2+ entry under these conditions have focused on the so-called ‘capacitative’ model (Putney, 1986, 1990) in which the enhanced entry of Ca2+ results solely from the emptying of the agonist-sensitive intracellular Ca2+ stores, without any direct requirement for inositol phosphates. Evidence in support of such a mechanism in a wide variety of different cell types is extensive. More recently, direct electrophysiological measurements of a Ca2+-selective current that is activated by depletion of the intracellular Ca2+ stores (so-called ICRAC) have been made (Hoth and Penner, 1992, 1993; Penner et al. 1993), and its selectivity and activation/inactivation properties characterized (Hoth, 1995; Zweifach and Lewis, 1995a,b). However, the key question as to the nature of the signal responsible for relaying the status of the intracellular stores to the Ca2+ entry pathway in the plasma membrane remains unresolved. Several possible ‘candidate’ messengers have been suggested, including cyclic GMP (Bahnson et al. 1993; Xu et al. 1994), a member of the ‘small G-protein’ family (Fasolato et al. 1993; Bird and Putney, 1993) or a currently poorly characterized ‘Ca2+ influx factor’ (CIF) (Randriamampita and Tsien, 1993).
An intriguing variation on this model has been proposed in which the Ca2+ status of the stores is transmitted to the plasma membrane Ca2+ entry pathways (ICRAC channels) via direct protein–protein interactions (Berridge, 1995). In this ‘conformational coupling’ model, it is envisaged that the large cytoplasmic domain of the InsP3 receptor is responsible for relaying information on the store status directly to the entry channels. Implicit in such a model is the requirement that the InsP3 receptor be able to sense and respond to changes in the Ca2+ content of the store. Unfortunately, studies examining possible influences of the store content on InsP3 receptor function have produced contradictory results, and this is currently an area of considerable controversy (Shuttleworth, 1995). In addition, for a conformational coupling mechanism to operate, the stores involved must lie in close proximity to the Ca2+ entry channels in the plasma membrane. This limitation has led to an additional important requirement for the model. It is known that the Ca2+ entry pathway demonstrates a marked Ca2+-induced inactivation, and a similar inactivation has been recorded directly in the depletion-activated ICRAC channels (Zweifach and Lewis, 1995a,b). Given this inactivation, it would seem that the InsP3 receptors involved in conformational coupling must not be involved in Ca2+ release as the resulting high local concentrations of Ca2+ would inactivate the entry pathway (Berridge, 1995). The suggestion has been made that both the localization at the plasma membrane and the inhibition of the Ca2+ release activity may involve the binding of ankyrin to the receptor (Berridge, 1995). However, the sequence identified as the proposed ankyrin-binding site (amino acids 2546–2556 in the mouse, 2548–2558 in the rat; Bourguignon and Jin, 1995) actually lies between membrane-spanning domains 5 and 6 in most models of the InsP3 receptor sequence (Yamamoto-Hino et al. 1994). This would place this putative binding site at the luminal face of the receptor, a location which is incompatible with the roles proposed. In conclusion, it would seem that, although the conformational coupling model can explain much of the data regarding the activation of Ca2+ entry by depletion of the stores, some significant problems remain and, in the absence of any direct evidence, its validity remains uncertain.
As already discussed, the kinds of responses that typify stimulation with more physiological levels of agonists, such as [Ca2+]i oscillations and [Ca2+]i waves, involve only the transient emptying of the intracellular stores. As Ca2+ entry was generally thought to be of little importance in these responses, it had been largely ignored. The first indication that agonist-activated Ca2+ entry under these conditions may not conform to the capacitative model came from studies where the quenching of intracellular Indo-1 by Mn2+ was used to evaluate the rate of Ca2+ entry during oscillatory responses. In avian nasal gland cells, the quenching rate of intracellular Indo-1 by Mn2+ is enhanced some three-to sixfold during [Ca2+]i oscillations, but the rate of quenching remains constant during the alternate emptying and refilling of the stores (Martin and Shuttleworth, 1994b; Shuttleworth and Thompson, 1996b). Similar findings have been reported in endothelial cells (Jacob, 1990) and HeLa cells (Thorn, 1995). Of course, the precise relationships between Mn2+ quenching and agonist-activated Ca2+ entry are unclear, and such results must therefore be interpreted with caution. Nevertheless, the finding of a non-oscillating Mn2+ quench in cells whose stores are alternately emptying and refilling does suggest that the rate of at least one component of agonist-activated Ca2+ entry is unaffected by the status of the oscillating agonist-sensitive Ca2+ stores. As such, this is inconsistent with capacitative models.
More substantial evidence for a non-capacitative Ca2+ entry during oscillations came from experiments on cells where the agonist-induced oscillations had been stopped by the acute inhibition of Ca2+ entry upon depolarization of the cell (Shuttleworth and Thompson, 1996b). As discussed above, because Ca2+ entry is critically required to trigger the emptying of the stores (Martin and Shuttleworth, 1994b), acute inhibition of Ca2+ entry stops the oscillations but does so without any measurable depletion of the stores, despite the continued presence of agonist. Significantly, returning such cells to conditions where Ca2+ entry can occur (i.e. repolarization) results in the immediate restoration of [Ca2+]i oscillations whose size, shape and frequency are identical to those seen immediately prior to depolarization. As Ca2+ entry is known to be an absolute requirement to drive these oscillations (see above), their immediate restart upon restoration of the driving force for entry implies that the Ca2+ entry pathway must have remained activated under conditions where it is known that the agonist-sensitive stores are not depleted. Such data are clearly difficult to reconcile with current versions of the capacitative model.
In addition, because it is inherent in the capacitative model that the activation of Ca2+ entry must be preceded by the emptying of the stores, it follows that inhibition of Ca2+ entry should be without effect on the initiation of a [Ca2+]i response. Examination of this proposition under conditions where oscillatory responses would be expected (i.e. low agonist concentrations) revealed that reduction of Ca2+ entry, by depolarization of the cell by exposure to a high-K+ medium, significantly increased the time delay before the first [Ca2+]i spike. More dramatically, inhibition of Ca2+ entry by brief pre-exposure of the cells to SK&F 96365 completely blocked the initiation of an oscillatory response on addition of agonist. Oscillations rapidly begin, however, on subsequent washout of the SK&F 96365 (Shuttleworth and Thompson, 1996b). If, as these data suggest, agonist-activated Ca2+ entry at these agonist concentrations is critically involved in the initial release of Ca2+ from intracellular stores, then such entry cannot be gated by any mechanism dependent on depletion of the stores. Hence, such entry cannot be via a capacitative mechanism, at least as currently understood. Finally, we attempted a more direct evaluation of the Ca2+ entry pathway under conditions of [Ca2+]i oscillations. A basic paradigm of the capacitative model is that, following depletion of the stores by exposure to an agonist, Ca2+ entry remains activated even after removal of the agonist and the decay of elevated levels of inositol phosphates, as long as refilling of the stores is prevented. Such a response can readily be demonstrated in nasal gland cells when treated with agonist concentrations sufficient to produce a sustained elevation in [Ca2+]i (Fig. 2A). However, experiments utilizing an exactly parallel protocol in oscillating cells failed to reveal any such response (Fig. 2B). The protocol employed also effectively ruled out the possibility of Ca2+ entry being gated in a capacitative manner but by only a subset of the total agonist-sensitive stores, perhaps those lying closest to the plasma membrane, with only the remaining stores undergoing the cyclical emptying and refilling associated with oscillations. In this case, the sustained depletion of such a subset of the stores should still give rise to a capacitative Ca2+ entry, but no such entry could be detected in these experiments.
To summarize, a variety of different experiments have produced findings that are inconsistent with the proposal that the observed agonist-activated Ca2+ entry during oscillations is occurring via a capacitative mechanism. Possible alternative mechanisms for Ca2+ entry under these circumstances fall into two groups – either entry is being controlled by some component of the PLC/inositol phosphate pathway [such as via an action of InsP3 and/or inositol tetrakisphosphate (InsP4) at the plasma membrane] or the agonist is activating a distinct signalling pathway that is independently responsible for the control of Ca2+ entry. Our efforts to try and resolve this question have recently revealed the presence of a novel signalling mechanism for the activation of Ca2+ entry under conditions of low agonist concentrations that is quite distinct from the ‘classical’ capacitative mechanism. This mechanism involves the receptor-activation of a phospholipase A2 and the subsequent generation of arachidonic acid (Shuttleworth, 1996). Using cells from the avian nasal gland, we found that activation of muscarinic receptors stimulated a marked release of arachidonic acid in cells preloaded with [3H]arachidonic acid. Furthermore, this response was unaffected by preloading the cells with the Ca2+ chelator BAPTA at a concentration sufficient to obliterate completely any agonist-induced changes in [Ca2+]i, showing that the observed agonist-induced activation of arachidonic acid release was not merely a downstream effect of the simultaneous [Ca2+]i signal. It was further shown that the simple addition of low concentrations (2–5 μmol l−1) of exogenous arachidonic acid to otherwise unstimulated cells was sufficient to result in an increase in Ca2+ entry.
Additional evidence came from experiments using the biscoclaurine (bisbenzylisoquinoline) alkaloid isotetrandrine. Although this compound has been shown to produce many different effects in various cell types (see references in Akiba et al. 1992), these diverse actions can generally be ascribed, at least at the low micromolar concentrations employed, either to effects on various voltage-operated Ca2+ channels (specifically by interacting at the benzothiazipine binding site) (Felix et al. 1992) or to effects on the activation of a phospholipase A2 (PLA2) and the consequent generation of arachidonic acid (AA) (Akiba et al. 1992; Hashizume et al. 1991). For example, isotetrandrine has been shown to inhibit amylase release and the [Ca2+]i oscillations induced by agonists acting at the high-affinity cholecystokinin (CCK) receptor in pancreatic acinar cells (Tsunoda and Owyang, 1995) – actions that are believed to involve PLA2 and arachidonic acid. In avian nasal gland cells, we found that isotetrandrine potently and reversibly blocked [Ca2+]i oscillations induced by low agonist concentrations, and did so by inhibiting the associated non-capacitative entry of Ca2+, an effect confirmed by Mn2+ quench studies (Shuttleworth, 1996). This effect clearly did not result from an action on voltage-operated Ca2+ channels as these are absent in avian nasal gland cells, and the classic benzothiazipine receptor antagonist diltiazem was completely without effect on [Ca2+]i oscillations. However, consistent with a proposed action in inhibiting a PLA2, isotetrandrine significantly inhibited the observed receptor-mediated release of arachidonic acid. It did not, however, have any effect on the receptor-mediated generation of InsP3. As noted above, isotetrandrine potently blocked the non-capacitative Ca2+ entry associated with low agonist concentrations. Further experiments showed that this was a specific effect as isotetrandrine was completely without effect on the sustained [Ca2+]i signals seen at high agonist concentrations, signals that are known to reflect a capacitative Ca2+ entry. It also failed to affect the capacitative entry activated following depletion of the stores with the sarco(endo)plasmic reticulum Ca2+-ATPase (SERCA) pump inhibitor thapsigargin. A similar failure to affect agonist-induced or thapsigargin-induced capacitative Ca2+ entry in lacrimal and parotid cells has previously been reported for the related compound tetrandrine (Rossier et al. 1993). As a final confirmation of the role of a PLA2/AA pathway in the activation of the non-capacitative Ca2+ entry, it was shown that the inhibition of [Ca2+]i oscillations by isotetrandrine could be reversed by addition of low concentrations of exogenous arachidonic acid as well as by addition of a small peptide that is part of a recently isolated PLA2-activating protein (PLAP-peptide) (Clark et al. 1991).
The conclusion from these studies is that the agonist-enhanced, non-capacitative Ca2+ entry seen during [Ca2+]i oscillations results from the generation of AA as a result of a receptor-mediated activation of a PLA2 (Shuttleworth, 1996). The observed effects appear to involve an action of arachidonic acid itself as inhibitors of the cyclo-oxygenase and lipoxygenase pathways for arachidonic acid metabolism failed to mimic the effects of isotetrandrine on [Ca2+]i oscillations. In fact, lipoxygenase inhibition changed an oscillatory [Ca2+]i response into a sustained plateau response, an effect that presumably reflects the accumulation of arachidonic acid following inhibition of its metabolism. The precise mechanism by which the PLA2 is activated is, as yet, unclear. For example, it may involve the carbachol-induced activation of a distinct PLA2-coupled G-protein. However, as argued above, our data show that the continued oscillatory signal requires the combined action of increases in [InsP3] and Ca2+ entry. We know that the activation of phospholipase C (PLC) in the nasal gland cells involves a Gq-type G-protein (Hildebrandt and Shuttleworth, 1993), and one way in which such a combined response could be coordinated would be for both actions to be mediated via the same G-protein, with the Gα subunit activating PLC and the Gβγ subunits activating a PLA2. Significantly, activation of a PLA2 by Gβγ subunits has been suggested in other systems (Jelsema and Axelrod, 1987; Kim et al. 1989; Tsunoda and Owyang, 1995).
Implications for cellular Ca2+ homeostasis
What then are the homeostatic implications of these various phenomena for the cell? It is clear that, in successful Ca2+ signalling, a finely tuned balance between the adequate activation of components of the secretory mechanism (e.g. K+ and Cl− channels) and the avoidance of [Ca2+]i levels that are toxic to the cell is achieved. These toxic effects are multifarious and include the inappropriate activation of Ca2+-dependent proteases, phosphatases, kinases, phospholipases and endonucleases, as well as effects on the organization and function of cytoskeletal elements such as microtubule polymerization and microfilament formation (Nicotera et al. 1992). In this context, much has been made of the ‘homeostatic’ advantages of oscillations versus sustained [Ca2+]i signals. Such advantages include an improved signal-to-noise ratio, the avoidance of desensitization or accommodation of the target effector, the possibility of utilizing different components of the multiple coding information inherent in oscillations, and the avoidance of excessive Ca2+ pumping resulting from sustained elevations in [Ca2+]i. On this basis, the claim that Ca2+ signalling at the cellular level may have evolved towards a frequency-modulated mode of action for reasons of economy and safety − both of which are undoubtedly sound homeostatic principles −may have some validity. However, in such discussions, no discrimination is generally made regarding the source or origin of the respective changes in [Ca2+]i and how the mechanisms responsible for producing the changes in [Ca2+]i are carefully controlled to operate in a ‘homeostatically responsible’ manner.
The magnitude of this problem for the cell should not be overlooked. In this review, I have attempted to show that all Ca2+ signals in non-excitable cells ultimately involve components of both intracellular Ca2+ release and extracellular Ca2+ entry. The electrochemical gradients driving Ca2+ into the cytosol via either of these processes are extremely large. Whilst the magnitude of these gradients is part of what makes cytosolic Ca2+ such an excellent rapid signalling mechanism, it does carry the danger of producing excessive increases in [Ca2+]i. Given the requirement for careful control over the changes in [Ca2+]i associated with successful Ca2+ signalling and the avoidance of the lethal effects of high [Ca2+]i discussed above, it is clearly imperative that these two mechanisms of raising [Ca2+]i (i.e. release and entry) be closely integrated. This review has presented evidence for the presence of two distinct mechanisms for the activation of Ca2+ entry by agonists in non-excitable cells – the well-known capacitative mechanism that is dependent on the depletion of agonist-sensitive intracellular Ca2+ stores (Putney, 1990; Berridge, 1995) and a newly described non-capacitative mechanism involving the activation of a PLA2 and the generation of arachidonic acid (Shuttleworth, 1996). Why are there two forms of Ca2+ entry activation, what are their individual properties and how do these properties influence and indicate their respective roles in the cell?
As discussed, current models for agonist-activated Ca2+ entry emphasize the capacitative model where, under normal circumstances, the activation of Ca2+ entry is a result of the InsP3-mediated emptying of the intracellular stores (Putney, 1990). The obvious implication of this sequential mechanism is that depletion of the stores must precede the activation of Ca2+ entry. Although, as already mentioned, the details of the mechanisms involved in this sequence of events are far from clear, there is evidence that it incurs a significant delay in the activation process. For example, in avian nasal gland cells, the activation of Mn2+ quenching is delayed by some 20–30 s following the peak of store discharge at supramaximal agonist concentrations (Shuttleworth, 1994b). Similar delays have been found in several other cell types. More direct measurements of the depletion-dependent ICRAC have indicated that the onset of activation occurs after a delay of approximately 5–6 s and is followed by a rather slow exponential increase to maximal levels with a time constant of some 18–27 s (Hoth and Penner, 1993). Such slow activation may indicate the involvement of a biochemical sequence of events in the transmission of the information on the status of the stores to the entry channels in the plasma membrane (Shuttleworth, 1994b). Alternatively, it could result from the effects of a Ca2+-dependent inactivation mechanism that has been demonstrated in ICRAC channels (Zweifach and Lewis, 1995a,b), although the activation time-course appears to be unaffected by the presence of high concentrations of the Ca2+-chelators BAPTA or EGTA in the cell (Hoth and Penner, 1993). Whatever its origin, such slow activation kinetics are clearly not optimal for the fine control of a [Ca2+]i response, particularly during the transient emptying and refilling of the stores associated with oscillations. Furthermore, at least in some cell types, the steady-state quantitative relationship between store depletion and capacitative Ca2+ entry is markedly non-linear, with the result that the stores must be 30–40 % depleted before entry begins to be significantly enhanced (Alvarez et al. 1994). Overall, given the limitations imposed by these specific properties of the capacitative mechanism, it would seem to be neither very sensitive nor subtle in its response and therefore is not well-adapted for achieving a finely tuned [Ca2+]i signal accurately reflecting agonist action at the plasma membrane receptor. However, the capacitative mechanism is well-suited for maintaining the capacity of the stores – i.e. keeping them recharged with Ca2+. This is obviously necessary for the cell to be able to respond to the next signal. In addition, there is now extensive evidence that the intrastore Ca2+ has a critical influence on a range of other cell functions, including protein folding and trafficking as well as cell division and proliferation (Waldron et al. 1994). Perhaps this maintenance of the overall, long-term status of the intracellular Ca2+ stores is the main function of the capacitative mechanism in cells.
In contrast to the sustained ‘plateau’ [Ca2+]i responses seen with maximal stimulation, it seems clear that the main source of the Ca2+ involved in the oscillatory [Ca2+]i signals typical of physiological levels of stimulation is the intracellular agonist-sensitive stores. This is to be expected given that the Ca2+ pump activity on these stores (SERCA pump) is 2–14 times higher than the plasma membrane Ca2+ pump activity in the physiological range of [Ca2+]i (Camello et al. 1996). Indeed, there would appear to be sound energetic reasons why it is advantageous for Ca2+ signals to be generated principally by release and re-uptake from the stores (Taylor, 1995). If the main source of the Ca2+ involved in oscillations is the stores, then it follows that the point at which fine control of the oscillations is achieved must be the site of Ca2+ release – i.e. the InsP3 receptor. Indeed, it seems that the properties of the InsP3 receptor are uniquely suited for such a role. As already discussed, the operation of the InsP3-mediated Ca2+ channel is controlled by the concentrations both of InsP3 and of cytosolic Ca2+, which act as co-agonists at the receptor, and also by the rate at which [Ca2+]i is changing (Taylor and Traynor, 1995). It would seem, therefore, that the InsP3 receptor is the ideal signal transducer for converting the agonist activation of a receptor in the plasma membrane into controlled changes in [Ca2+]i. These considerations lead to the concept of the InsP3 receptor as ‘the controller of [Ca2+]i’ rather than merely a simple conduit for Ca2+ release. As noted, the activity of the InsP3-gated channel is controlled by the concentration of InsP3, by the concentration of Ca2+ and by the rate of change in [Ca2+]i. For appropriately sensitive signalling and accurate transduction of the primary message, a system is needed whereby these three parameters can be independently modulated but, at the same time, all three should accurately reflect the degree of activation of the plasma membrane receptor by the agonist. It is difficult to envisage how these complex interactions could be achieved with the strictly sequential pathway inherent in the capacitative model. In contrast, it is precisely under these conditions that the separate, but coordinated, activation of InsP3 generation and Ca2+ entry offered by the parallel PLC/InsP3 and PLA2/AA signalling pathways discussed above probably have their relevance (Fig. 3). We have shown that, under physiological conditions of stimulation, the agonist-generated levels of neither InsP3 nor arachidonic acid are adequate on their own to induce a meaningful [Ca2+]i signal. Only when both are activated together are [Ca2+]i oscillations induced (Shuttleworth, 1996). The model proposed (Fig. 3) invokes the known co-agonist properties of InsP3 and cytosolic Ca2+ on the InsP3 receptor. Because of its rapid diffusion through the cytosol, it is likely that cytosolic concentrations of InsP3 are essentially uniform during agonist stimulation (Allbritton et al. 1992). The role of the generated levels of InsP3 is to sensitize the InsP3 receptors on the stores such that a Ca2+ entry activated by arachidonic acid induces the opening of the Ca2+ release channels to initiate a Ca2+ signal. The probability of such a sequence of events occurring is therefore determined by the sustained level of InsP3 which sensitizes the receptors and by the magnitude (and rate) of the local change in [Ca2+]i produced by the AA-induced Ca2+ entry.
Neither of these is effective alone, but both are independently related to the degree of agonist stimulation (Fig. 3). This primary event may be restricted to stores close to the plasma membrane, the InsP3 receptors of which are more likely to be exposed to changes in [Ca2+]i associated with Ca2+ entry. Once initiated, however, the Ca2+ pulse released from these stores will, if large enough, trigger release from other InsP3-sensitized stores deeper in the cytoplasm. In this way, complex [Ca2+]i signals in the form of spatially restricted ‘trigger zones’ or transient oscillations or waves that traverse the cytosol can be generated. In this model, the inherent repetitive oscillatory cycle is dependent on the biphasic influences of [Ca2+]i on the InsP3 receptor much as in the basic oscillation model originally discussed, but now both of the known required components of the [Ca2+]i signal (Ca2+ release and Ca2+ entry), along with the levels of each of the co-agonists responsible for triggering this Ca2+ release (InsP3 and cytosolic [Ca2+]), are coupled separately, but in a co-ordinated manner, to levels of agonist stimulation. The mechanism provides the means whereby low-amplitude analog signals produced by low concentrations of agonist (namely, InsP3 generation and AA-induced Ca2+ entry), when present together, are accurately converted into large-amplitude, but repetitively transient (i.e. frequency-encoded), signals by their combined actions at the InsP3 receptor. In this way, the InsP3 receptor acts as both an amplifier and an analog-to-digital converter, as well as a coincidence detector. If the sensitive and faithful transduction of relevant signals to appropriate effectors is a critical component of any successful homeostatic system, then this intracellular system is elegantly adapted for its role in the control of secretion.
Work from the author’s laboratory included in this review was supported by NIH grant GM40457. The contributions of Dr J.-P. Hildebrandt and Dr S. C. Martin, and the excellent technical assistance of Jill L. Thompson, are gratefully acknowledged.