The versatility of Ca2+ as an intracellular messenger stems largely from the impressive, but complex, spatiotemporal organization of the Ca2+ signals. For example, the latter when initiated by IP3 (inositol 1,4,5-trisphosphate) in many cells manifest hierarchical recruitment of elementary Ca2+ release events (‘blips’ and then ‘puffs’) en route to global regenerative Ca2+ waves as the cellular IP3 concentration rises. The spacing of IP3Rs (IP3 receptors) and their regulation by Ca2+ are key determinants of these spatially organized Ca2+ signals, but neither is adequately understood. IP3Rs have been proposed to be pre-assembled into clusters, but their composition, geometry and whether clustering affects IP3R behaviour are unknown. Using patch-clamp recording from the outer nuclear envelope of DT40 cells expressing rat IP3R1 or IP3R3, we have recently shown that low concentrations of IP3 cause IP3Rs to aggregate rapidly and reversibly into small clusters of approximately four IP3Rs. At resting cytosolic Ca2+ concentrations, clustered IP3Rs open independently, but with lower open probability, shorter open duration and lesser IP3-sensitivity than lone IP3Rs. This inhibitory influence of clustering on IP3R is reversed when the [Ca2+]i (cytosolic free Ca2+ concentration) increases. The gating of clustered IP3Rs exposed to increased [Ca2+]i is coupled: they are more likely to open and close together, and their simultaneous openings are prolonged. Dynamic clustering of IP3Rs by IP3 thus exposes them to local Ca2+ rises and increases their propensity for a CICR (Ca2+-induced Ca2+ rise), thereby facilitating hierarchical recruitment of the elementary events that underlie all IP3-evoked Ca2+ signals.
Every cell needs to cope with its ever changing immediate environment marked by continuous arrival and disappearance of numerous stimuli. Although a rich palette of surface receptors are employed to recognize these exogenous signals of diverse nature, it is highly intriguing that only a handful of intracellular messengers are produced upon transduction of these signals. Perhaps even more impressive is how these messengers commit themselves to trigger (or modulate) appropriate cellular responses without compromising the fidelity. The latter is best epitomized by the universality and versatility of intracellular Ca2+ signals which regulate almost every aspect of cellular life, ranging from fertilization to cell death for every life form .
A substantial amount of cellular energy is continuously spent in keeping the basal [Ca2+]i (intracellular free Ca2+ concentration) at a low level (100–300 nM) and as such also maintaining a large concentration gradient of this ion towards cytosol with respect to internal Ca2+ stores and the extracellular compartment. Transient elevation of [Ca2+]i from the resting low level to a much higher (~1 μM) level is what effectively constitutes a ‘Ca2+ signal’ . Although such a rise in [Ca2+]i occurs on demand, the way it is triggered varies between cells largely depending on the presence or absence of excitable membranes. For the excitable cells such as neurons and those of skeletal muscle, Ca2+ signals are triggered by Ca2+ entry through some voltage-gated highly Ca2+-selective ion channels and, to some extent, ligand-gated Ca2+-permeant channels expressed in their plasma membrane. Non-excitable cells (e.g. lymphocytes, acinar cells of exocrine glands, endothelial cells of blood vessels or fibroblasts) essentially lack this pathway and, following surface stimulation, they instead produce a few soluble second messengers such as IP3 (inositol 1,4,5-trisphosphate) that mobilize Ca2+ from intracellular stores such as the ER (endoplasmic reticulum). IICR (IP3-evoked Ca2+ release) eventually depletes the ER and thus triggers further and more sustained Ca2+ entry via the store-operated pathway in the plasma membrane. I have been particularly interested in IP3 which is the most widely used Ca2+-mobilizing intracellular messenger and, in the rest of the present article, I focus on IP3-evoked Ca2+ signalling.
IP3-evoked Ca2+ signals are spatially organized
IP3 is produced at the plasma membrane as a result of hydrolysis of PIP2 (phosphatidylinositol 4,5-bisphosphate) in a PLC (phospholipase C)-dependent manner. Activation of PLC occurs when cell-surface receptors (receptor tyrosine kinases and those coupled to the Gq/11 family of G-proteins) are stimulated by various hormones, growth factors or neurotransmitters. Upon production, IP3 rapidly diffuses into the cytosol and binds to IP3Rs (IP3 receptors) that are mainly expressed in the membranes of the ER, the major internal Ca2+ store. IP3 binding to IP3Rs causes the pore to open, through which Ca2+ rapidly flows into the cytosol down its concentration gradient. Cells often have numerous and diverse types of surface receptors, many of which, upon stimulation, activate PLC and thus converge on making the same intracellular messenger, IP3, and the Ca2+ signalling that ensues. Yet diverse, but specific, cellular effects are elicited without any chaos. Indeed, Ca2+ signals triggered by IP3 have been specifically implicated in fertilization, proliferation and differentiation, metabolism, contraction, fluid secretion, exocytosis, chemotaxis, platelet aggregation, synaptic plasticity and opening of some Ca2+-activated ion channels . So what imparts such enormous versatility in the action of Ca2+? Growing evidence suggests that this versatility stems largely from the spatiotemporal variations in Ca2+ signals as well as the availability of proteins that can ‘interpret’ different messages encoded in different spatiotemporal patterns of Ca2+ signals.
IP3-evoked Ca2+ signals in most intact cells do manifest complex spatiotemporal patterns such as waves and oscillations that result from the hierarchical recruitment of elementary Ca2+-release events [3–5]. Stimulation of the cell-surface receptors with graded concentrations of IP3-generating stimuli triggers Ca2+ signals of differing amplitudes and spatial dimensions, sequentially growing from elementary to global Ca2+ release events  (Figure 1). When the intracellular [IP3] (IP3 concentration) is low, the fundamental events known as ‘Ca2+ blips’ occur sporadically. These are small localized elevations of [Ca2+]i and are most likely to stem from random openings of single IP3Rs. Such events typically last less than 130 ms, and have amplitudes of less than 40 nM. As [IP3] increases with higher stimulus intensity, more IP3Rs become active, and Ca2+ mobilized through a group of active IP3Rs leads to the appearance of ‘elementary or intermediate events’ known as ‘Ca2+ puffs’. The latter remain localized, spreading no more than 6 μm, have amplitudes of 50–600 nM and last for ~1 s. With much higher [IP3], more and more puffs are ignited, and the spatiotemporal summation of all puffs eventually leads to a global regenerative Ca2+ wave that spreads throughout the cell [7–9]. It is important to appreciate the fact that the elementary Ca2+-release events are not just generic building blocks for global Ca2+ waves. These spatially restricted Ca2+ signals can be delivered locally to some target proteins and thus can regulate distinct cellular processes . It is therefore imperative to have a better understanding of what makes spatial organization of Ca2+ signals possible.
Hierarchical recruitment of IP3-mediated Ca2+ signalling events
For IP3-evoked Ca2+ signals, the sequential progression of signal Ca2+ from an elementary to a global level seems to capitalize on at least two critical processes that are directly related to IP3Rs. First, IP3 tunes the Ca2+-sensitivity of IP3Rs in such a way that Ca2+ released from an active IP3R can activate another IP3R through a process known as CICR (Ca2+-induced Ca2+ release) [10,11]. In order for CICR to operate successfully, it is intuitively obvious that the process should also be dependent on, apart from the cytoplasmic diffusibility of Ca2+ itself governed by the nature of existing buffers, the average distance Ca2+ needs to traverse before it encounters the next IP3R. The latter is dictated by the distribution and spatial arrangement of IP3Rs within the ER membrane. Review of the existing literature on IP3Rs is suggestive of non-random distribution of these channels within the native ER membrane of various cell types . Detection of Ca2+ puffs using single-cell Ca2+ imaging itself implies the existence of IP3R clusters [8,9,13]. Confocal microscopy in several cell types confirmed that both native and tagged IP3Rs form clusters [14–18]. Patch-clamp recordings have also indicated heterogeneous distribution of IP3Rs in the outer nuclear membrane that is continuous with the ER [19–21]. The prevailing view is thus in favour of IP3Rs existing as pre-formed static clusters in the ER membrane and within each cluster, IP3Rs are stimulated by CICR initiated by a Ca2+ blip. Our results have led us to suggest that IP3Rs may assemble ‘on demand’ into clusters in response to IP3 and the process is dynamically regulated [22,23].
What prompts IP3Rs to cluster?
The first clue regarding the possible mechanism of IP3R clustering came from Wojcikiewicz and colleagues , showing that agonists triggered redistribution of IP3Rs (type 2 and 3 predominantly) in some cells. This could be mimicked by increasing [Ca2+]c (cytosolic Ca2+ concentration), using either ionomycin or thapsigargin, and it could be prevented by removal of extracellular Ca2+. Thus an increase in [Ca2+]c was suggested to be the trigger for clustering . However, in a later study, the Mikoshiba group  observed that agonist-evoked clustering of tagged IP3Rs coincided with a peak in IP3 production, with no obvious correlation with the cytosolic Ca2+ transients, and a PLC inhibitor (U-73122) could suppress the ionophore-induced IP3R clustering observed by Wilson et al. . Furthermore, mutant IP3Rs unable to change conformation upon IP3 binding did not cluster . All of these led to the suggestion that the critical trigger for IP3R clustering was IP3 binding to IP3Rs followed by the necessity of the subsequent conformational transition of these channels to their open state [15,18]. We found that IP3Rs recorded from the outer nuclear membranes of DT40 cells expressing recombinant IP3Rs (isoform 1 and 3) are initially randomly distributed, but a low [IP3] (lower than required to maximally activate them, <300 nM) rapidly and reversibly drive these receptors into small clusters [22,23]. The molecular driving force of IP3R clustering is hitherto unknown. It surely requires IP3-evoked structural changes leading towards the open state, but the permeant cation Ca2+ seems not to play any role in triggering the process [15,22,23]. Structural rearrangements of the ER are also unlikely to cause the observed aggregation of these channels [14,15]. FRAP (fluorescence recovery after photobleaching) analysis of fluorescent protein-tagged IP3Rs have shown that these proteins can laterally diffuse within the ER membrane [24–26]. Since clustering occurred regardless of disruption of actin-based cytoskeleton or microtubules, we suggest that the lateral diffusional mobility of IP3R is probably enough to allow such a process to happen .
Possible geometry of IP3R clusters
Although there have been several modelling and stochastic simulation-based studies undertaken to resolve the geometry of IP3R clusters, the first experimental attempt to determine the number of IP3Rs per Ca2+ puff that effectively represents activity from an IP3R cluster was made by Ian Parker and colleagues , who estimated the Ca2+ flux associated with individual events in Xenopus oocytes by integrating fluorescence profiles along the scan line in three dimensions to derive the ‘signal mass’ at each time point. The peak signal mass values during the averaged blip and puff corresponded respectively, to approximately 1.2×10−19 and 2×10−18 mol of Ca2+ (Ca2+ currents of ~0.4 and ~2.5 pA). This estimate of the unitary Ca2+ current was close to a previously estimated value (0.47 pA) from bilayer recordings of cerebellar IP3R1 under physiologically relevant conditions (i.e. voltage-clamped at 0 mV, 2.5 mM luminal Ca2+ and 110 mM symmetrical K+, so that there is no net K+ flux) . This implied that a typical Ca2+ puff might arise from the concerted opening of approximately five IP3Rs. In our recent nuclear patch-clamp studies with DT40 cells expressing recombinant IP3R3, a cluster appears to contain similar average number of active IP3R3 with a maximum of eight observed [22,23]. The maximal number of active IP3R detected in some nuclear patch-clamp studies was similarly small (six to eight) , although as many as 15 channels were detected in nuclear patches from oocytes expressing rat IP3R3 . There is, however, a practical limit beyond which detecting large number of large-conductance channels, such as IP3Rs, electrophysiologically would be challenging due to either possible interference with giga-seal formation or recorded currents becoming too noisy to resolve separate current levels. Nevertheless, our estimates closely match more recent higher-resolution optical analyses of elementary events . The latter confirmed, using a TIRFM (total internal reflection fluorescence microscopy)-based ‘optical patch-clamping’ approach, that approximately six active IP3Rs on an average underlie a puff . It is also interesting to note that Ca2+ sparks which are comparable with Ca2+ puffs may consist of similar low number (<10) of RyRs (ryanodine receptors) .
The precise topology of an IP3R cluster is hitherto unknown. On the basis of the observed effect (discussed below) of clustering on IP3R activity recorded from excised nuclear patches (estimated patch-pipette diameter ~1 μm) of DT40 cells, we reasoned that clustered IP3Rs must be in contact, and this led us to speculate that the average separation of IP3Rs is likely to fall from ~1 μm to ~20 nm after clustering, and that clusters are ~2 μm apart . These estimated spacings somewhat concur with confocal measurements suggesting that a Ca2+ puff originates from a cluster ~50 nm wide and that clusters are ~3 μm apart . Recent TIRFM-based analysis suggests that IP3Rs contributing to puffs are perhaps distributed in clusters with a diameter of less than 500 nm . When expressed at high densities, IP3Rs  and RyRs  form arrays with each tetrameric receptor contacting four others. We speculate that IP3-evoked clusters consisting of approximately four or five IP3Rs exploit similar contacts and so, with single IP3Rs, form the fundamental units of Ca2+ signalling.
Does clustering affect IP3R gating?
Although clustering of different IP3R subtypes has been observed in various cells with some mechanisms proposed, nothing was known of whether clustering affects IP3R behaviour. Multiple IP3Rs have been frequently observed in previous bilayer and nuclear patch-clamp studies  and they were assumed to be kinetically homogeneous so that their collective activity (reported by ‘NPo’ where N is the number of IP3Rs in a patch and Po is the open probability of single IP3R) was considered simply equal to the sum of individual open probabilities. Although this view is perhaps generally true for most ion channels, there are examples for many ion channels for which gating of one channel within a multi-channel patch can influence the gating of its neighbours [33–37]. We, for the first time, carried out detailed kinetic analysis of multiple IP3R activity recorded from the DT40 cell nuclei. We observed that at a [Ca2+]i that mimics resting cytosolic condition (~200 nM), multiple IP3Rs recorded from DT40 cell nuclei were identical and they gated independently. But, to our surprise, each IP3R within a multichannel patch had its Po reduced by ~50% and had attenuated IP3-sensitivity when compared with a lone IP3R. The observed halving of Po seemed to stem from a comparable reduction in mean open time (τo) of each IP3R within a cluster. This implies that IP3-evoked clustering almost doubles the rate of channel closure (1/τo), whereas most other known regulators of IP3Rs under comparable recording conditions affect mainly the mean closed time (τc) and thereby the rate of channel opening . The observed kinetic distinction between lone and clustered IP3Rs we believe was significant as this could affect the time course of the initial Ca2+ release within elementary events and thereby Ca2+-mediated interplay between clustered IP3Rs. This is confirmed by our simulations of intracellular Ca2+ spikes, where the ~50% decrease in τo of clustered IP3Rs results in a 4-fold decrease in Ca2+ spiking frequency . How is it possible for multiple IP3Rs under low [Ca2+]i to gate independently, yet with reduction in their open time? We speculate that lone and clustered IP3Rs are kinetically distinct ion channel entities and clustered IP3Rs perhaps can be envisaged as a multi-barrel gun, with each barrel firing independently of others, but each with reduced force when compared with that of a single-barrelled gun. This sort of multi-barrel gating kinetics for ion channels is not totally unprecedented [38–40]. The fact that IP3-evoked clustering almost doubles the rate of channel closure allowed us to make a novel use of patch-clamp technique to follow the time course of IP3-evoked clustering of IP3R. Using flash photolysis of caged IP3, we found that clustering is complete within 1.5–2 s of the photorelease of IP3, and a pairing of IP3R suffices to have maximal diminution of Po .
Recent analyses by Foskett and colleagues  have challenged our observation on clustering-induced changes in gating behaviour of IP3R. The cause of the apparent discrepancy is not yet clear, but there were major differences in properties of lone IP3R observed by us and them. For example, we (like many others) hardly observed any indication of rapid inactivation or modal gating pattern in IP3R behaviour in contrast with what they invariably found . It is, however, intriguing to notice that in one of their early reports , multiple IP3Rs detected from oocyte nuclei had much slower inactivation kinetics when compared with single IP3Rs.
IP3R clustering retunes Ca2+-sensitivity of IP3Rs
Under physiological conditions, the [Ca2+]i near an open IP3R can almost instantaneously increase to several micromolar . Dispersed IP3Rs are insulated from these increases by effective cytosolic Ca2+ buffering . However, as IP3Rs aggregate and their pores may come to be only ~20 nm apart, each clustered IP3R will be instantly (microseconds) exposed to very high local free [Ca2+] whenever a neighbouring receptor opens. On the basis of this rationale, we repeated our analyses for multiple IP3R activity recorded in presence of higher (~1 μM) [Ca2+]i and observed that the multiple IP3Rs were no longer inhibited and neither were they independent. Instead, their gating was coupled with near synchronous opening and closing transitions. Although high [Ca2+]i was overall stimulatory towards IP3R activity, its effect on clustered IP3Rs was greater than the lone IP3Rs because it reversed the attrition in IP3R activity caused by their clustering. We therefore proposed that IP3R clustering not only ensures, by placing them optimally, that they respond to Ca2+ released by a neighbour instantaneously, but also retunes their regulation to exaggerate the impact of that Ca2+ (Figure 2) .
Clustering retunes Ca2+ regulation of IP3R
The mechanism underlying the co-operative gating or positive coupling of IP3R in the presence of high [Ca2+]i is not clear. An attractive possibility, assuming physical contact between IP3Rs, is that mechanical interactions allow ‘conformational spread’ between channels . Alternatively, as proposed for the nicotinic channel , interacting IP3Rs may not be in contact, but altered surface charge distribution caused by the gating transition in one channel may positively affect the gating of its neighbours. Last, but not least, there may be involvement of accessory protein(s) in this process. For example, coupled gating of RyR1 and RyR2 requires the presence of immunophilins, FKBP (FK506-binding protein) 12/12.6  and exogenous FKBP-induced coupled gating in cerebellar IP3R1 reconstituted in planar lipid bilayers . However, on the basis of studies from other groups, the association of FKBP12 with IP3R remains controversial . The role of other cytosolic or luminal factors in this process remains to be investigated.
On the basis of our experimental findings, we proposed a model where IP3 rapidly drives randomly distributed IP3Rs into small clusters, wherein their IP3- and Ca2+-sensitivities are retuned to exaggerate Ca2+-mediated recruitment of IP3Rs and so to allow hierarchical recruitment of elementary Ca2+-release events. Although our model seems to align comfortably with several studies reporting mobile IP3Rs within the ER membrane [24–26] and also with IP3 triggering assembly of these receptors into clusters (perhaps of higher order than detected electrophysiologically) [15,16,18], it does not seem compatible with TIRFM-based analyses of elementary Ca2+-release events in intact cells [46,47]. Measuring elementary Ca2+-release events in SH-SY5Y neuroblastoma cells using TIRFM, Parker and colleagues have shown that IP3Rs underlying the puff sites are relatively stable entities with hardly any spatial movement observed for single or multiple IP3Rs [46,47]. To reconcile their findings with those studies showing that IP3Rs are mobile, they have suggested that physiologically relevant Ca2+ signals stem from a subset (~200) of thousands of IP3Rs. These chosen IP3Rs are somehow primed for generating puffs being corralled over some anchoring platform. It is clearly imperative now to resolve whether our data with nuclear IP3Rs can really be extrapolated to those in the ER, especially when we cannot completely rule out the possibility of differing microscopic environment between these organelles. It is also equally urgent is to know whether the fixed sites that beget puffs stereotypically within an intact cell contains arrays of immobile IP3Rs or static anchorage to which dynamic IP3Rs can dock and become committed for generating puffs.
Early Career Research Award Delivered at the Biochemical Society Centenary Event held at the Royal Society, London, on 16 December 2011 Taufiq Rahman
cytosolic Ca2+ concentration
intracellular free Ca2+ concentration
Ca2+-induced Ca2+ release
total internal reflection fluorescence microscopy
I thank Professor Colin Taylor (Department of Pharmacology, University of Cambridge) for his great support and supervision. I also thank Dr Alexander Skupin (Luxembourg Centre for Systems Biomedicine, Luxembourg) and Dr Martin Falcke (Mathematical Cell Physiology, Max Delbruck Center for Molecular Medicine, Berlin, Germany) for their collaborative support to my work.
I thank the Yousouf Jameel Foundation for my PhD studentship in Cambridge University. My doctoral and postdoctoral work has been supported by the Jameel Foundation, the Wellcome Trust and the Biotechnology and Biological Sciences Research Council. Finally, I acknowledge a Drapers' Company Research Fellowship at Pembroke College for supporting me at present.