Ca²⁺ microdomains near plasma membrane Ca²⁺ channels: impact on cell function

Abstract

In eukaryotic cells, a rise in cytoplasmic Ca²⁺ can activate a plethora of responses that operate on time scales ranging from milliseconds to days. Inherent to the use of a promiscuous signal like Ca²⁺ is the problem of specificity: how can Ca²⁺ activate some responses but not others? We now know that the spatial profile of the Ca²⁺ signal is important Ca²⁺ does not simply rise uniformly throughout the cytoplasm upon stimulation but can reach very high levels locally, creating spatial gradients. The most fundamental local Ca²⁺ signal is the Ca²⁺ microdomain that develops rapidly near open plasmalemmal Ca²⁺ channels like voltage-gated L-type (Cav1.2) and store-operated CRAC channels. Recent work has revealed that Ca²⁺ microdomains arising from these channels are remarkably versatile in triggering a range of responses that differ enormously in both temporal and spatial profile. Here, I delineate basic features of Ca²⁺ microdomains and then describe how these highly local signals are used by Ca²⁺-permeable channels to drive cellular responses.

Introduction

Eukaryotic cells respond to changes in their environment by generating intracellular signals. Foremost amongst such signals is the Ca²⁺ ion. Although its appreciation by physiologists was rather fortuitous, beginning with Sidney Ringer's serendipitous discovery that it was Ca²⁺ in London's tap water that supported frog heart contractions, we now know that a rise in cytoplasmic Ca²⁺ can activate a plethora of key cellular responses including neurotransmitter release, metabolism, cell growth and proliferation as well as cell death (Berridge et al. 2003; Clapham, 2008). This remarkable versatility of Ca²⁺ raises the old chestnut of specificity: how can such a promiscuous messenger ever elicit a selective response? For many years, this problem could be described by Churchill's aphorism of ‘a riddle wrapped in a mystery inside an enigma’. Over the past few years, however, evidence has revealed that information is contained in both the frequency and the amplitude of Ca²⁺ signals, and these discrete components can be deciphered by intracellular proteins and then translated into distinct responses (Clapham, 2008). There is now a growing awareness that local Ca²⁺ signals or Ca²⁺ microdomains, generated by open Ca²⁺ channels, provide a further means for generating specificity (Neher, 1998b; Rizzuto & Pozzan, 2006). Diffusion of Ca²⁺ through an open Ca²⁺ channel creates a local Ca²⁺ gradient, which falls off steeply with distance due to cytoplasmic Ca²⁺ buffering. Within a few nanometres of the channel pore, buffers are simply not fast enough to capture the incoming Ca²⁺ (Neher, 1998a,b). This, coupled with the very small volume the microdomain occupies, means that local Ca²⁺ can rapidly rise to very high levels, often orders of magnitude greater than the bulk cytoplasmic Ca²⁺. Location of targets close to Ca²⁺ channels therefore provides a means for rapid and selective activation (Berridge et al. 2003). The term Ca²⁺ microdomain is widely used in the literature to refer to a range of Ca²⁺ signals from local plumes of Ca²⁺ near open single Ca²⁺ channels to those restricted to one pole or region of a cell and occupying distances of hundreds of nanometres to micrometres. In this review, I will focus exclusively on how Ca²⁺ microdomains arising from plasma membrane Ca²⁺ channels affect cell function, drawing mainly from two very well-studied systems: voltage-gated L-type Ca²⁺ channels in neurons and store-operated CRAC channels in non-excitable cells. Ca²⁺ microdomains near Ca²⁺ release channels on intracellular stores play important roles in generating inter alia intracellular Ca²⁺ oscillations, Ca²⁺ wave propagation and mitochondrial ATP production, but are not discussed here. The reader is referred instead to two excellent recent reviews on this topic (Petersen, 2005; Rizzuto & Pozzan, 2006).

Measuring and Manipulating Ca²⁺ microdomains

The amplitude and spatial extent of local Ca²⁺ influx ([Ca²⁺]_i through a Ca²⁺ channel is determined by the single channel Ca²⁺ flux and, at distances greater than a few nanometres, local Ca²⁺ buffering (Neher, 1998a,b). Ca²⁺ flux in turn is determined by the electrochemical gradient for Ca²⁺ entry and the single channel conductance.

In the absence of buffers, the spatial profile of the Ca²⁺ signal is determined by simple diffusion and can be described thus (Neher, 1998a):

where, i_Ca is the single channel Ca²⁺ current, D_Ca is the diffusion coefficient of Ca²⁺ in the cytoplasm (300 μm² s⁻¹), r the distance from the channel and [Ca²⁺]_o the bulk Ca²⁺ concentration.

In living cells, mobile and immobile buffers shape the profile of the Ca²⁺ signal. Immobile buffers are fixed in space and hence will only prolong the time needed to reach steady-state, and not affect the final level reached. Once bound with Ca²⁺, immobile buffers can be ignored as they are not replaced by free buffer. Mobile buffers on the other hand bind Ca²⁺ and then, by buffered diffusion, spread the Ca²⁺ deeper into the cytoplasm. Ca²⁺-bound mobile buffers are replaced by Ca²⁺-free ones and thus will bind Ca²⁺ until the total pool is saturated.

Equation (1) can then be modified to take into account the presence of excess unsaturable mobile buffers (Neher, 1998a):

where λ = √D_Ca/k_onB and denotes the mean path length a Ca²⁺ ion travels before being captured by chelator, k_on is the chelator's on-rate and B is the free buffer concentration where buffer is present in excess.

Unfortunately, we still know little about the k_on, concentration and spatial distribution of mobile Ca²⁺ buffers in cells. Moreover, different cell types have different endogenous Ca²⁺ binding capacities (Neher, 1995), and hence it is difficult to compare directly the spatial profile of a Ca²⁺ signal through different Ca²⁺ channels in different systems. In addition, most single channel recordings of voltage-gated Ca²⁺ channels have understandably been conducted with 110 mm Ba²⁺ as the charge carrier, not physiological Ca²⁺. Two channels where the conductance has been studied in physiological Ca²⁺ are L-type Ca²⁺ channels (2.6 pS) and store-operated CRAC channels (unitary chord conductance of 9 fS) (Zweifach & Lewis, 1993; Church & Stanley, 1996). The profile of the Ca²⁺ microdomain as a function of distance from each of these channels is simulated in Fig. 1A and B. Ca²⁺ close to the pore is considerably higher than bulk Ca²⁺ and even a Ca²⁺-selective but low conductance channel like CRAC can nevertheless generate quite a high local Ca²⁺ concentration.

Figure 1. Amplitude and spatial extent of Ca²⁺ microdomains.

A and B, plots of the numerically calculated profile of a Ca²⁺ microdomain near an open L-type Ca²⁺ channel (A) and a CRAC channel (B). Single channel currents were 0.144 pA (Church & Stanley, 1996) and 1.5 fA (Zweifach & Lewis, 1995), respectively. Endogenous Ca²⁺ buffering was ignored. C, the distance (λ) a Ca²⁺ ion diffuses before it is captured by some widely used Ca²⁺ chelators is shown and is stated next to each chelator. All concentrations are in mm. D, various properties of widely used fluorescent dyes (Lattanzio. & Bartschat (1991) and the calculated λ. All concentrations were 100 μm, corresponding to the levels used experimentally. For C and D, resting Ca²⁺ was set at 100 nm and the free chelator concentration then calculated. λ is not a precise calculation because buffer is not present in excess.

The simulations in Fig. 1 neglect Ca²⁺ buffering. How might this sculpt the profile of the Ca²⁺ signal? The mean path length (λ) Ca²⁺ diffuses from a point source for different concentrations of experimentally used chelators is shown in Fig. 1C. Strong spatial restriction of the microdomain is seen with millimolar BAPTA whereas EGTA is much less effective, reflecting its ∼100-fold slower k_on. Figure 1C also shows that the fluorescent dye fura 2, widely used to measure cytoplasmic Ca²⁺, has a marked effect on the lateral spread of a Ca²⁺ microdomain. This reflects its rapid k_on (7.6 × 10⁸m⁻¹ s⁻¹; Lattanzio. & Bartschat, 1991), which is similar to that of BAPTA. Figure 1D presents a similar analysis for commonly used fluorescent Ca²⁺ indicators. All reduce the lateral spread of single microdomains, which needs to be borne in mind when interpreting Ca²⁺ signals measured with such dyes, especially when used in high concentrations. These calculations are somewhat simplistic as they do not correct for buffered diffusion. Nor do they consider local saturation of buffer, although this would be a less acute problem for CRAC channels where local Ca²⁺ is severalfold less than free buffer concentration.

Numerical calculations therefore predict high local Ca²⁺ concentrations near single channels. Can such local Ca²⁺ be measured? Early work comparing the rate and extent of activation of bio-assays such as Ca²⁺-activated K⁺ channels (Roberts et al. 1990; Robitaille et al. 1990; Prakriya et al. 1996) or exocytosis (Heidelberger et al. 1994; Beutner et al. 2004) following a bulk cytoplasmic Ca²⁺ rise with activation of voltage-gated Ca²⁺ channels enabled researchers to conclude that local Ca²⁺ near the Ca²⁺ channels could reach 10–100 μm. Can these microdomains be measured directly? Unfortunately, this has proved very difficult because the resolution of light microscopy is not sufficiently high to detect the Ca²⁺ microdomain near a single open plasma membrane Ca²⁺ channel (see section on Neurotransmitter release). An exciting new development is the use of total internal reflection fluorescence microscopy to measure flux through single channels (Demuro & Parker, 2004), although this method does not permit combined patch clamp and optical recordings of the channel under investigation.

There are several ways to invoke the involvement of local Ca²⁺ entry rather than a bulk Ca²⁺ rise in activating a target. Firstly, one can test whether manipulating the amplitude and spatial extent of single Ca²⁺ channel flux (by changing the driving force) alters the response without changing bulk Ca²⁺. Alternatively, for a similar bulk Ca²⁺, many open Ca²⁺ channels but each with a smaller single channel flux might be less effective in driving a response than fewer open channels but each with a larger single channel flux. Secondly, one can compare the effects of exogenous Ca²⁺ chelators, loaded into the cell, on the response. Very close to the channel pore, chelators have virtually no effect on the Ca²⁺ concentration but do reduce the lateral spread of the microdomain in a manner determined by the chelator's on-rate (Neher, 1998b). Use of chelators with different on-rates can therefore provide information on the effects of lateral spread of the microdomain on a response as well as providing an estimate of how close to the pore an effector is located. However, chelators have effects in addition to their ability to buffer Ca²⁺. For example, BAPTA can interfere with InsP₃ binding to InsP₃ receptors (Richardson & Taylor, 1993) and impedes protein kinase C activation (Chang et al. 2008). Hence effects of BAPTA are not always attributable to Ca²⁺ buffering and these additional effects need to be considered. Third, the kinetics of a response provides clues as to whether it is driven by adjacent Ca²⁺ channels. Fast responses arising within milliseconds, which therefore develop well before bulk Ca²⁺ has reached equilibrium, often point towards a role for local Ca²⁺ signals. Fourth, if the Ca²⁺ requirement of a process is considerably higher than the increase in bulk Ca²⁺ that is reached, then it is difficult to see how it can be controlled by anything other than local Ca²⁺. Finally, in some cases, it has been possible to show the Ca²⁺ channel and its target physically interact (see next section). Such strong colocalization implicates control by local Ca²⁺.

Ca²⁺ microdomains from voltage-gated Ca²⁺ channels and neuronal function

Activation of Ca²⁺-dependent K⁺ channels

Ca²⁺ microdomains near voltage-gated Ca²⁺ channels are tightly coupled to Ca²⁺-activated K⁺ channels in vestibular hair cells (Roberts et al. 1990), adrenal chromaffin cells (Prakriya et al. 1996) and the frog neuromuscular junction (Robitaille et al. 1990). A striking example of this tight coupling has been reported recently by Falker and colleagues (Berkefeld et al. 2006). Affinity purification of large conductance Ca²⁺-activated K⁺ (BK_Ca) channels from plasma membrane of rat brain revealed the additional presence of L-, P/Q- and N-type Ca²⁺ channels, suggesting these form a macromolecular complex. Giant inside-out patches from Xenopus oocytes expressing recombinant BK and P/Q-type channels revealed rapid activation of the K⁺ current by Ca²⁺ influx, consistent with a tight physical interaction (Berkefeld et al. 2006). Comparing BK_Ca kinetics following voltage-dependent Ca²⁺ influx with that seen upon directly varying cytoplasmic Ca²⁺ led the authors to conclude that local Ca²⁺ was ∼10 μm. Finally, high concentrations of cytoplasmic EGTA were ineffective, but BAPTA reduced BK_Ca–P/Q-type channel coupling. The distance between BK_Ca and P/Q-type channels was placed at ∼10–15 nm (Berkefeld et al. 2006). Although the physical basis for coupling is unclear, a major role for the pore forming α subunit was found.

The coupling between BK and P/Q-type channels is a nice example of a tight physical interaction between a Ca²⁺ channel and its target, thus ensuring robust activation of the effector. Other examples include the interaction between N-type channels at active zones with syntaxin 1, a protein that binds to synaptic vesicles and thus locates the vesicle near the channel, and coupling between L-type Ca²⁺ channels and ryanodine type 1 release channels of the sarcoplasmic reticulum at the t-tubules of skeletal muscle. In this latter case, coupling is conformational in nature and not by Ca²⁺ flux through the L-type channel.

Neurotransmitter release

At synapses, neurotransmitter release following an action potential clearly has to be fast in order to maximize information propagation. Hence one might expect a dominant role for local Ca²⁺ entry in driving the fusion of ready releasable vesicles. Imaging of n-aequorin-J, a luminescent protein with low affinity for Ca²⁺, revealed ‘hot-spots’ of subplasmalemmal Ca²⁺ following stimulation of presynaptic terminals of the squid giant synapse and which were ∼200–300 μm in amplitude (Llinas et al. 1992). Action potentials induced localized presynaptic subplasmalemmal Ca²⁺ hot-spots in Xenopus neuromuscular junction preparations and these fluorescent spots decayed within a couple of milliseconds (DiGregorio & Vergara, 1997). The spatial resolution with optical microscopy is limited by the wavelength of light. Even with a high numerical aperture quartz objective lens, it would be more than ∼200 nm, which is severalfold wider than the calculated spatial extent of a Ca²⁺ microdomain (Neher, 1998b). Using total internal reflection fluorescence microscopy, which measures events located within ∼100 nm of the plasma membrane, local ‘hot spots’ of high Ca²⁺ have been seen in several different experimental systems and these are likely to correlate with the fusion event (Becherer et al. 2003; Zenisek et al. 2003). Another approach has been to compare the rate and extent of exocytosis between opening of Ca²⁺ channels and a bulk rise in cytoplasmic Ca²⁺ (Heidelberger et al. 1994; Beutner et al. 2004). The best way to achieve a rapid and uniform rise in bulk Ca²⁺ is to use flash photolysis of caged Ca²⁺, since the Ca²⁺ sensor triggering exocytosis and the fluorescent dye used to measure this Ca²⁺ will be exposed to the same cytoplasmic Ca²⁺ load (Neher, 1998b). Using this approach, it has become clear that the Ca²⁺ concentration at the release sites is in the range 10–300 μm, depending on the type of neuron investigated (Heidelberger et al. 1994; Schneggenburger & Neher, 2000; Beutner et al. 2004; Bollmann & Sakmann, 2005). With these higher values in particular, one might expect secretion to be unaffected by slow chelators like EGTA. Even at tens of millimolar concentrations, EGTA had no effect on exoctyosis at the squid giant synapse whereas BAPTA inhibited release at considerably lower concentrations (Adler et al. 1991). Hence at this synapse, Ca²⁺ influx is very tightly coupled with transmitter release. On the other hand, high local Ca²⁺ does not necessarily mean that ready releasable vesicles are governed by single Ca²⁺ channel microdomains. Work in the calyx of Held and neocortex have shown that EGTA can reduce evoked neurotransmitter release (Borst & Sakmann, 1996; Rozov et al. 2001), demonstrating that Ca²⁺ channels are not as tightly coupled to exocytosis in these synapses as they are in the squid giant synapse. Instead, several Ca²⁺ channels are likely to affect the release of one vesicle at the calyx of Held. This is supported by a morphological and modelling study that reported an average distance of 200 nm between a vesicle and a Ca²⁺ channel and that a cluster of Ca²⁺ channels controlled the fusion of a vesicle (Meinrenken et al. 2002). Local Ca²⁺ is still the trigger, but a stringent colocalization between channel and effector is not required, at least in this synapse.

Neuronal growth

One key aspect of neuronal differentiation involves the targeting of axons to form synapses with other neurons, myocytes or endocrine cells. Such pathfinding occurs through migration of growth cones at the axon's tip (Spitzer, 2006). Growth cones exhibit spontaneous fluctuations in intracellular Ca²⁺ (around 1–10 spikes per hour). These Ca²⁺ signals are spatially restricted to the growth cones but are nevertheless needed for neuronal differentiation. Because these Ca²⁺ signals can occur in the absence of action potentials but depend on Ca²⁺ influx, the Ca²⁺ entry pathway is thought to involve an ion channel active at the resting potential. TRPC channels are Ca²⁺ permeable non-selective cation channels expressed in a range of diverse cell types (Nilius et al. 2007). Recent evidence implicates a major role for TRPC channels in growth cone morphology, motility and neurite elongation. In young hippocampal neurons, TRPC5 forms a complex with stathmin2, a protein that accumulates in growth cones, and this complex is transported to the cones via a vesicular transport mechanism (Greka et al. 2003). Whereas most growth cones had filipodia with perimeters (an index of length) of < 5 μm, overexpression of TRPC5 increased this ∼3-fold.

Growth cones do not simply extend along one axis but can turn in response to diffusible chemoattractants like netrin 1 and brain-derived neurotrophic factor (BDNF). Axon turning is an intracellular Ca²⁺-dependent process and the site of the Ca²⁺ rise is critical (Zheng, 2000). Focal laser-induced photolysis of caged Ca²⁺ has revealed that a local Ca²⁺ rise on one side of a growth cone causes the turning of the growth cone to that side (Zheng, 2000). Physiologically, these local Ca²⁺ rises can be brought about by opening of Ca²⁺ channels. In cultured Xenopus neurons, netrin 1 and BDNF both activated TRPC currents and knockdown of TRPC1 expression prevented both growth cone turning and the Ca²⁺ rise in response to nestrin-1 (Wang & Poo, 2005). In cultured cerebellar granule neurons, growth cone turning to BDNF was suppressed by knockdown of TRPC3 channels (Li et al. 2005). Collectively, these findings all point to a central role for TRPC channels in regulating growth cone motility and turning.

Is a microdomain of Ca²⁺ through a single TRPC channel sufficient to drive growth cone motility and turning? Zheng (2000) measured the Ca²⁺ rise to be ∼78 nm at the site of uncaging Ca²⁺ in the growth cone. Although TRPC channels are non-selective, conducting Ca²⁺ and Na⁺ with a permeability ratio of ∼1.6–9.0, they have a high single channel conductance of the order of 16–75 pS (Nilius et al. 2007). Hence the Ca²⁺ microdomain near an open TRPC channel may be sufficient to trigger the growth turning process.

Gene expression

In neurons, activation of the transcriptional factor CREB (cAMP response element binding protein) is linked to synaptic plasticity, learning and memory (Bailey et al. 1996; Shaywitz & Greenberg, 1999; Ahn et al. 2000). CREB activation is a two step process (Hu et al. 1999; Shaywitz & Greenberg, 1999): first, phosphorylation of serine 133 by either the ERK1/2 pathway or Ca²⁺/calmodulin-dependent kinases occurs and this is followed by binding of the CREB binding protein, a transcriptional coactivator. CREB binding protein in turn is activated by nuclear Ca²⁺ and CaM kinase IV.

A rise in nuceloplasmic Ca²⁺ following depolarization of the plasma membrane is sufficient to activate CREB (Hardingham et al. 1997). Direct injection of dextran-conjugated BAPTA into the nucleus reduced the nucleoplasmic Ca²⁺ rise and prevented activity-induced CREB stimulation without compromising the cytoplasmic Ca²⁺ elevation. Since cytoplasmic Ca²⁺ readily propagates into the nucleus, Ca²⁺ channel opening at the plasma membrane can raise nuclear Ca²⁺ by eliciting a bulk Ca²⁺ rise. In addition to this global Ca²⁺ dependence, a substantial body of evidence has demonstrated that local Ca²⁺ influx can also have an impact upon nuclear events. Ca²⁺ microdomains arising from voltage-gated Ca²⁺ channels (Deisseroth et al. 1996; 1998; Dolmetsch et al. 2001) or Ca²⁺-permeable NMDA receptors (Hardingham et al. 2001) can signal to the nucleus and induce CREB phosphorylation.

Although cortical and hippocampal neurons coexpress L-type, P/Q-type and N-type channels, only the L-type channel consistently couples to CREB activation (Deisseroth et al. 1998; Dolmetsch et al. 2001), providing a poignant example of how local Ca²⁺ signalling can achieve specificity.

The L-type Ca²⁺ channel has a calmodulin-binding isoleucine–glutamine (IQ) motif in the carboxy terminus (Zuehlke et al. 1999). At resting Ca²⁺ (∼100 nm), calmodulin is stably associated with this motif. Hence calmodulin is held in a strategically optimal location adjacent to the Ca²⁺ channel pore, enabling it to detect Ca²⁺ microdomains arising from L-type channel opening. Calmodulin translocates from the cytoplasm to the nucleus in a Ca²⁺-dependent manner, where it stimulates CREB phosphorylation through activation of nuclear Ca²⁺/calmodulin-dependent protein kinases (Deisseroth et al. 1998). In cortical neurons, the IQ motif is required for signalling to the ERK pathway and the subsequent activation of CREB (Dolmetsch et al. 2001). However, voltage-dependent Ca²⁺ channels like the P/Q-type channel also bind calmodulin (Lee et al. 1999) yet are less able to activate CREB (Dolmetsch et al. 2001). Hence additional domains of the L-type channel are likely to be involved in signalling to the nucleus. One such region is the PDZ interaction sequence located within the amino acid sequence valine–serine–asparagine–leucine (VSNL) at the end of the L-type channel (Weick et al. 2003). Overexpression of a GFP-tagged VSNL construct prevented L-type channels in hippocampal CA3–CA1 pyramidal neurons from evoking CREB phosphorlyation and subsequent gene transcription. Hence the L-type channel may be physically integrated into a signalling complex that efficiently links the local Ca²⁺ microdomain to a pathway culminating in CREB activation.

Recently, it has been found that the carboxy terminus of the L-type Ca²⁺ channel can be cleaved to release a peptide that translocates to the nucleus where it regulates expression of numerous genes including the gap junction Cx31.1, netrin 4, TRPV4 and the sodium–calcium exchanger (Gomez-Ospina et al. 2006). How the L-type channel is cleaved and whether this involves a protease activated by local Ca²⁺ remain unclear at present.

Can L-type channels distinguish between different types of excitatory stimulus? Synaptic activity (excitatory postsynaptic potentials) is much more effective than spike activity (back-propagating action potentials) in triggering CREB phosphorylation, despite both stimuli generating similar bulk Ca²⁺ elevations. This is probably because L-type channels behave as kinetic filters, activating slowly upon depolarization (Mermelstein et al. 2000). Moreover, they open at relatively hyperpolarized potentials. Both the voltage- and time-dependent gating of the channels permit greater Ca²⁺ flux during a slow excitatory postsynaptic potential than with a brief action potential, providing a mechanism whereby certain synaptic inputs can trigger more CREB phosphorylation than others.

L-type channels are not the only voltage-gated Ca²⁺ channels that signal to the nucleus though. In cerebellar granule cells, activation of P/Q-type channels resulted in increased transcription and translation of syntaxin-1A. L- and N-type channels were not able to mimic this (Sutton et al. 1999). Ca²⁺ influx through these P/Q-type channels was propagated to the nucleus via Ca²⁺-induced Ca²⁺ release. Presumably, there is a specific association between P/Q-type channels and the internal Ca²⁺ stores, which the other Ca²⁺ channels are excluded from. Whether this tight coupling involves local Ca²⁺ influx through the P/Q-type channels is not known.

Activation of NMDA receptors induces CREB activation and gene expression. As this is still manifest in the presence of voltage-gated Ca²⁺ channel blockers, it is Ca²⁺ influx through the NMDA receptors themselves drives gene expression. In EGTA-loaded hippocampal neurons, NMDA receptor stimulation still activated the ERK pathway leading to CREB phosphorylation. Hence Ca²⁺ microdomains around open NMDA channels are thought to provide the trigger for signal transmission to the nucleus (Hardingham et al. 2001). These local Ca²⁺ signals are apparently propagated to the nucleus as a regenerative Ca²⁺ wave involving intracellular stores, probably through a Ca²⁺-induced Ca²⁺ release mechanism.

Several NMDA receptor subtypes have been described in neurons and are composed of different subunits (Cull-Candy et al. 2001). Subunit composition imparts differences in Mg²⁺ block as well as fractional Ca²⁺ permeability. Moreover, Ca²⁺ flux through NMDA channels can be increased within seconds following phosphorylation by protein kinase A (Skeberdis et al. 2006) or the Src family of tyrosine kinases (Takasu et al. 2002). Since the amplitude and spatial extent of a microdomain depends on Ca²⁺ flux through the channel, it will be interesting to see whether the local Ca²⁺ signals emanating from different NMDA channels, or the same channel in the presence of active protein kinase A for example, differ in their ability to trigger CREB phosphorylation. This could provide a novel mechanism for regulating synaptic plasticity.

Not only is the Ca²⁺ flux through NMDA receptors important in regulating nuclear events, but the location of the channels is also critical. Ca²⁺ microdomains associated with synaptic NMDA receptors triggered CREB phosphorylation and BDNF expression but extra-synaptic NMDA receptors had the opposite effect, switching off the CREB pathway and evoking cell death (Hardingham et al. 2002). These differences were related to the different subunit composition of the channels.

Ca²⁺ microdomains near store-operated Ca²⁺ channels and non-excitable cell function

Store-operated Ca²⁺ channels are activated by the emptying of intracellular Ca²⁺ stores and the ensuing Ca²⁺ entry activates a disparate array of responses (Parekh & Putney, 2005; Parekh, 2007). The most thoroughly studied store-operated channel is the CRAC channel (Hoth & Penner, 1992; Parekh & Putney, 2005). The molecular basis of this ubiquitous pathway has recently been identified (reviewed in Lewis, 2007). STIM1 is the Ca²⁺ sensor that detects the fall in store Ca²⁺ content and then migrates to form discrete puncta just below the plasma membrane. Here it somehow opens Orai1, the CRAC channel pore. The reader is referred to other symposium articles in this issue for more discussion of these proteins.

Local Ca²⁺ entry and ion channels

In a prostate cancer cell line, store-operated Ca²⁺ influx partially inhibited a volume-regulated anion channel that was involved in regulatory volume decrease (Lemonnier et al. 2002). Dialysis with 1 μm free Ca²⁺ failed to mimic this effect, indicating that subplasmalemmal [Ca²⁺] after CRAC channel opening was higher than this. It was suggested that store-operated Ca²⁺ channels were colocalized with the Cl⁻ channels.

Another example was revealed in a detailed study on TRPC7 channels, which are activated by both store-operated and non-store-operated mechanisms, the latter involving diacylglycerol. Diacylglycerol activation of TRPC7 was suppressed by block of SERCA pumps (Lemonnier et al. 2006). It was found that Ca²⁺ entry through TRPC7 fed back to inhibit the channels but this was prevented by Ca²⁺ uptake into the stores. What was particularly striking was that interaction between TRPC7 and SERCA pumps was maintained in the presence of high concentrations of BAPTA, revealing an intimate interaction between these proteins that reside on different membranes (Lemonnier et al. 2006).

Ca²⁺ tunnelling: local refilling of the endoplasmic reticulum

A striking example of how local Ca²⁺ entry can impact upon a response some distance away has been dissected out from studies in pancreatic acinar cells. In these highly polarized epithelia, store-operated Ca²⁺ entry occurs across the basolateral membrane but exocytosis of zymogen-containing granules takes place at the apical pole several micrometres away. Simple diffusion of Ca²⁺ is impeded by the very high Ca²⁺ cytosolic buffering of the acinar cell (Mogami et al. 1999). This problem is circumvented by the ingenious mechanism of Ca²⁺ tunnelling (Mogami et al. 1997). The endoplasmic reticulum is a contiguous organelle in acinar cells and runs very close to the basolateral membrane as well as protruding into the apical pole (Petersen et al. 2001). Local Ca²⁺ entering across the basolateral membrane is taken up into the endoplasmic reticulum by SERCA pumps which are adjacent to the Ca²⁺ influx sites. The endoplasmic reticulum has a very low Ca²⁺ buffering capacity, so the Ca²⁺ can diffuse rapidly through the organelle to reach the apical region, where it is released close to the granules by InsP₃ (Mogami et al. 1997). Local Ca²⁺ influx can therefore have an impact upon events restricted to the opposite pole of the cell.

Local Ca²⁺entry and regulation of adenylyl cyclases

The first real evidence that local Ca²⁺ entry through store-operated Ca²⁺ channels could regulate a cellular response was provided by Cooper and colleagues (reviewed in Willoughby & Cooper, 2007). They studied the dependence of calmodulin-regulated adenylyl cyclases on cytoplasmic Ca²⁺. Two isoforms of adenylyl cyclase are activated by a rise in cytoplasmic Ca²⁺ (types 1 and 8) and two are inhibited by a Ca²⁺ rise (types 5 and 6). These enzymes associate with the plasma membrane. In C6-2B glioma cells, robust Ca²⁺ release from intracellular stores in response to agonist, thapsigargin or the Ca²⁺ ionophore ionomycin were all unable to regulate the Ca²⁺-dependent isoforms (Chiono et al. 1995; Fagan et al. 1996). By contrast, store-operated Ca²⁺ influx significantly altered enzyme activity even though it raised bulk Ca²⁺ to a lesser extent than that achieved by Ca²⁺ release. Hence these isoforms are regulated by a subplasmalemmal Ca²⁺ rise rather than global Ca²⁺. However, these early results did not distinguish between a general subplasmalemmal Ca²⁺ rise or a more restricted Ca²⁺ signal in the vicinity of open store-operated Ca²⁺ channels. Further work by the Cooper laboratory clearly established the concept that local Ca²⁺ tightly associated with store-operated Ca²⁺ channels regulated adenylyl cyclase. Firstly, loading cells with BAPTA prevented store-operated entry from regulating adenylyl cyclase whereas EGTA was ineffective (Fagan et al. 1998). Second, the potency series of divalent cations for regulating adenylyl cyclase correlated with the conductivity through CRAC channels. When Ba²⁺ permeated store-operated channels, no regulation of adenylyl cyclase was detected whereas Sr²⁺ was ∼50% as effective as Ca²⁺ (Gu & Cooper, 2000). This is similar to the conductivity profile of CRAC channels, which over the physiological voltage range (–80 to 0 mV) is Ca²⁺ > Sr²⁺ > Ba²⁺, where Sr²⁺ conductivity is 50–80% that of Ca²⁺ (Zweifach & Lewis, 1993; Fierro & Parekh, 2000). By contrast, the divalent cation equilibrium binding profile to the EF hand of calmodulin is 1 Ca²⁺: 40 Sr²⁺: 1600 Ba²⁺. From equilibrium considerations, Sr²⁺ should therefore be much less effective than Ca²⁺ in regulating calmodulin-dependent adenylyl cyclases. The fact that Sr²⁺ is almost as effective as Ca²⁺ implies that adenylyl cyclase is exposed to very high local Sr²⁺ levels, which could only be the case if the enzyme was located very close to the Ca²⁺ channel. Finally, opening non-store-operated Ca²⁺ channels by application of either OAG (Martin & Cooper, 2006) or arachidonic acid (Shuttleworth & Thompson, 1999) did not affect adenylyl cyclase activity, despite raising bulk Ca²⁺ to a similar extent to that seen following store-operated entry. Hence only a local Ca²⁺ rise in the vicinity of store-operated channels can regulate adenylyl cyclase. What underpins this tight molecular coupling? One possibility is that Ca²⁺-dependent adenylyl cyclases are tethered to the store-operated channels, analogous to anchoring of protein kinase A to channels through AKAP proteins. Alternatively, store-operated channels and Ca²⁺-dependent adenylyl cyclases might associate with specific plasma membrane domains like lipid rafts (Willoughby & Cooper, 2007). Destruction of cholesterol-rich rafts by methyl-β-cyclodextrin prevented store-operated entry from regulating adenylyl cyclase without affecting the extent of Ca²⁺ influx, suggesting recruitment of both proteins to rafts facilitated interaction between them. Although certain TRPC channels have been identified in rafts, it is not clear whether CRAC channels are integrated here too. Now that Orai1 has been identified as the CRAC channel pore, this possibility can be tested directly.

Local Ca²⁺ entry and NO synthase

As with Ca²⁺-regulated adenylyl cyclases, eNOS is preferentially activated by local Ca²⁺ influx through store-operated channels (Lin et al. 2000). In endothelial cells, opening of store-operated channels following stimulation with thapsigargin in the presence of 4 mm external Ca²⁺ raised bulk cytoplasmic [Ca²⁺] to a level comparable to that seen with ionomycin in 30 μm external Ca²⁺. However, nitric oxide production was 10-fold higher with thapsigargin. Lin et al. (2000) then constructed two chimaeras: eNOS linked to aequorin and a myristoylation-deficient eNOS-aequorin. Myristoylation is required for eNOS insertion into the plasma membrane. The eNOS-aequorin construct associated with the plasma membrane whereas the myristoylation-deficient chimaera was cytosolic. Consistent with local Ca²⁺ influx and hence spatial Ca²⁺ gradients, eNOS-aequorin reported a significantly higher Ca²⁺ signal than the cytosolic probe following activation of store-operated Ca²⁺ entry.

Local Ca²⁺ entry and plasma membrane Ca²⁺ATPase pump

Another example of local Ca²⁺ influx tightly controlling a plasma membrane-restricted enzyme has come from work on T lymphocytes (Bautista & Lewis, 2004). A rise in cytoplasmic Ca²⁺ following CRAC channel opening increased plasmalemmal Ca²⁺-ATPase activity by reducing the K_M and increasing V_max of the pump. The increased Ca²⁺-ATPase activity was triggered by local Ca²⁺ microdomains near CRAC channels in addition to bulk cytoplasmic Ca²⁺ (Bautista & Lewis, 2004). Both an increase in the electrical gradient for Ca²⁺ influx by hyperpolarization of the membrane potential and raising the chemical gradient by elevating external Ca²⁺ significantly increased pump activity without changing bulk cytoplasmic Ca²⁺. Furthermore, a nice method was developed to reinforce the importance of microdomains in controlling effector activity. The trivalent cation La³⁺ is a potent blocker of CRAC channels, with an IC₅₀ of ∼20 nm. La³⁺ is likely to inhibit CRAC channels in a way similar to that seen in voltage-gated Ca²⁺ channels, namely by slow permeation resulting in a flickery block (Lansman et al. 1986). La³⁺ does not reduce the driving force for Ca²⁺ entry and hence does not decrease the amplitude of the microdomain, though its duration is likely to be reduced. Bautista & Lewis (2004) found that bulk cytoplasmic Ca²⁺ following CRAC channel activation in 20 mm external Ca²⁺ and 7 nm La³⁺ was similar to that seen in 2 mm external Ca²⁺ without La³⁺. Despite this, pump activity was significantly enhanced in 20 mm Ca²⁺ and La³⁺, providing compelling evidence that Ca²⁺ microdomains near CRAC channels are the critical determinants of enhanced pump activity.

Local Ca²⁺ entry and signalling to cytoplasmic enzymes: insight from the cPLA₂ and 5-lipoxygenase pathway

All the preceding examples of effectors activated by local Ca²⁺ influx are directly regulated by Ca²⁺ and are attached to, or span, the plasma membrane close to CRAC channels, thus facilitating their activation by local Ca²⁺ gradients. An interesting complication arises when the Ca²⁺-dependent target is located in the cytoplasm, some distance away. Recent work in my group has examined how local Ca²⁺ entry through CRAC channels in mast cells stimulates Ca²⁺-dependent phospholipase A₂ (cPLA₂) and 5-lipoxygenase, two cytoplasmic enzymes that generate the intracellular messenger arachidonic acid and the intercellular pro-inflammatory signal cysteinyl leukotriene LTC₄, which is secreted from the cell. Aberrant LTC₄ production is important clinically as it is linked to a variety of allergic diseases including allergic rhinitis and asthma (Peters-Golden et al. 2006).

Stimulation with thapsigargin or agonist in Ca²⁺-free solution elicited significant rises in global Ca²⁺ but these were totally ineffective in activating cPLA₂ or LTC₄ secretion (Chang & Parekh, 2004; Chang et al. 2007). On the other hand, activation of CRAC channels robustly activated cPLA₂ and LTC₄ secretion despite raising global Ca²⁺ to a similar extent (Chang & Parekh, 2004; Chang et al. 2008). We used a nice method developed by Bird & Putney (2005) to raise global Ca²⁺ after Ca²⁺ release from the stores. Expulsion of Ca²⁺ by plasma membrane Ca²⁺-ATPases is a major Ca²⁺ clearance mechanism in most cells types. Bird & Putney showed that block of this Ca²⁺-ATPase with high concentrations of lanthanides rendered cells tight to Ca²⁺ so that Ca²⁺ released by thapsigargin remained within the cytosol (Bird & Putney, 2005). Using this approach, we showed that the extent of the global Ca²⁺ rise following stimulation with thapisigargin and lanthanide in the absence of external Ca²⁺ was indistinguishable from the rise in response to CRAC channel activation but the latter was severalfold more effective in activating cPLA₂ and LTC₄ secretion (Chang et al. 2008). Hence local Ca²⁺ entry through CRAC channels and not global Ca²⁺ drives this signalling pathway. Consistent with this were the findings that (i) manipulating the electrochemical gradient for Ca²⁺ influx strongly impacted upon cPLA₂ activation with little effect on global Ca²⁺; (ii) loading cells with the slow Ca²⁺ chelator EGTA reduced bulk Ca²⁺ following CRAC channel opening but had little effect on cPLA₂ activation, and (iii) although bulk Ca²⁺ was slightly less after CRAC channel activation in 2 mm Ca²⁺ and a low concentration of La³⁺ compared with that seen in 0.5 mm Ca²⁺, the former condition, which supports larger Ca²⁺ microdomains, was much more effective in driving LTC₄ secretion (Chang et al. 2008).

Sensing local Ca²⁺ entry

How does a target detect local Ca²⁺ influx? In general, there are two ways whereby local Ca²⁺ entry can activate a Ca²⁺-dependent target. Because the spatial extent of a Ca²⁺ microdomain can extend tens of nanometres from the channel pore (Neher, 1998b), an effector placed within this range could respond directly to the local Ca²⁺. Examples include certain Ca²⁺-activated ion channels and Ca²⁺-dependent neurotransmitter release (see above), where Ca²⁺ microdomains near N- and P/Q-type Ca²⁺ channels bind to synaptotagmin to promote vesicular fusion. Alternatively, a Ca²⁺ sensor could relay local Ca²⁺ entry to targets not located within the Ca²⁺ plume of the open channel. An example here is the coupling of the IQ motif of L-type Ca²⁺ channels to the ERK pathway (Dolmetsch et al. 2001), which signals to the nucleus. Our work on CRAC channels has identified another way for propagation of local Ca²⁺ signals into the cytoplasm to activate cPLA₂ and leukotriene secretion (Chang et al. 2008). Local Ca²⁺ entry is detected by the non-receptor tyrosine kinase Syk, which then activates the ERK pathway through recruitment of protein kinase C isoforms α and βII. ERK stimulates cPLA₂ by phosphorylating the enzyme on Ser-505 to liberate arachidonic acid (Leslie, 1997). ERK also activates the 5-lipoxygenase enzyme, which metabolizes arachidonic acid to leukotrienes. ERK is therefore an effective and versatile transducer of CRAC channel activity and can bifurcate the local Ca²⁺ signal into the co-ordinated activation of two important downstream catabolic enzymes. Moreover, dual regulation of these two enzymes ensures that as arachidonic acid is generated, it is rapidly metabolized to LTC₄, thereby preventing excessive build-up of this pleiotropic messenger.

Ca²⁺ channel clustering: overlapping microdomains

Although I have focused exclusively on single Ca²⁺ channel microdomains, channels may cluster and produce more complex local Ca²⁺ signals. Now, the local Ca²⁺ buffers are more rapidly saturated and therefore the spatial extent of the Ca²⁺ signal increases substantially. In neurons, both a single Ca²⁺ channel and a cluster of channels drive exocytosis at an active zone. For CRAC channels, the pore-forming subunit Orai1 forms discrete puncta upon store depletion, which would be expected to generate a local Ca²⁺ concentration much larger than that in response to a single channel. Puncta formation therefore might increase the likelihood of activating targets located further away from the extent of a single microdomain. Interesting modes of regulation could involve varying the number of CRAC channels in a punctum as well as how quickly the puncta form and disaggregate.

Concluding remarks

Local Ca²⁺ influx provides a mechanism for generating specificity to the promiscuous second messenger Ca²⁺. Only those effectors located near the channel will be activated by the high ambient Ca²⁺, thus avoiding a need for a more global and therefore indiscriminate Ca²⁺ signal. Moreover, selectivity is further increased by targeting effectors to certain Ca²⁺ channels and not others, as seen with calmodulin tethering to the L-type channel and the tight association between adenylyl cyclases and store-operated channels. Finally, tethering signalling molecules to the Ca²⁺ channel enables long-range signalling to occur, as occurs with diffusion of calmodulin to the nucleus or following recruitment of protein kinase C to CRAC channels by Syk. Figure 2 summarizes a range of cellular responses activated by local Ca²⁺ signals near L-type (Fig. 2A) and CRAC channels (Fig. 2B).

Figure 2. The cartoon summarizing the impact of Ca²⁺ microdomains near L-type (A) and CRAC (B) channels on cell function.

CaM denotes calmodulin, 5LOX 5-lipoxygenase.

Microdomains have other advantages too. Locating the effector near the Ca²⁺ microdomain greatly increases the speed of response, a mechanism which underlies rapid release of primed neurotransmitter-containing vesicles at the active zone by N- and P/Q-type Ca²⁺ channels and the shaping of neuronal firing pattern by Ca²⁺-activated K⁺ channels colocalized with Ca²⁺ channels. Microdomains also increase the fidelity of the response; when the Ca²⁺ channel opens, a juxtaposed effector will generally be exposed to a high local Ca²⁺. By contrast, the same effector located deeper in the cell may not be exposed to incoming Ca²⁺ due to Ca²⁺ buffering and sequestration by intervening mitochondria and ER.

Finally, Ca²⁺ microdomains are not invariant in amplitude and extent but are controlled by several factors. Clearly, changes in single channel flux through alterations in the electrochemical gradient will have an impact directly on the size of the microdomain and hence its ability to couple to a target. However, one could envisage other, hitherto unexplored, ways of regulating ‘microdomain–effector coupling’. L-type Ca²⁺ channels show gating mode shifts (Tsien et al. 1986; Armstrong et al. 1991). Mode 0 represents no channel activity despite depolarization, mode 1 reflects brief channel openings of short duration and mode 2 is characterized by continuous channel openings of long duration. Calmodulin-dependent protein kinase-mediated phosphorylation favours stabilization in mode 2 (Dzhura et al. 2000), whereas phosphatases shift channels to mode 0 (Erxleben et al. 2003). The duration of the microdomain as well as the lateral spread would therefore be increased in mode 2 gating. It would be interesting to see whether agonists that stabilize mode 2 enhance the robustness of ‘microdomain–effector coupling’ or recruit new effectors due to the greater spatial spread of the microdomain. Another intriguing possibility arises from the increase in Ca²⁺ permeability of NMDA receptors upon phosphorylation by protein kinase A, which would increase the size of the Ca²⁺ microdomain. It is certainly intriguing to consider that coupling between Ca²⁺ microdomains near NMDA receptors and local targets might be strengthened or weakened depending on the phosphorylation state of the channel. In this way, agonist-specific responses could be elicited despite using the same local Ca²⁺ signal. Collectively, Ca²⁺ microdomains provide a simple yet effective mechanism for eliciting specific responses to the pleiotropic and versatile Ca²⁺ messenger.