## Figures

## Abstract

Calcium signals are involved in a large variety of physiological processes. Their versatility relies on the diversity of spatio-temporal behaviors that the calcium concentration can display. Calcium entry through inositol 1,4,5-trisphosphate (IP) receptors (IPR's) is a key component that participates in both local signals such as “puffs” and in global waves. IPR's are usually organized in clusters on the membrane of the endoplasmic reticulum and their spatial distribution has important effects on the resulting signal. Recent high resolution observations [1] of Ca puffs offer a window to study intra-cluster organization. The experiments give the distribution of the number of IPR's that open during each puff without much processing. Here we present a simple model with which we interpret the experimental distribution in terms of two stochastic processes: IP binding and unbinding and Ca-mediated inter-channel coupling. Depending on the parameters of the system, the distribution may be dominated by one or the other process. The transition between both extreme cases is similar to a percolation process. We show how, from an analysis of the experimental distribution, information can be obtained on the relative weight of the two processes. The largest distance over which Ca-mediated coupling acts and the density of IP-bound IPR's of the cluster can also be estimated. The approach allows us to infer properties of the interactions among the channels of the cluster from statistical information on their emergent collective behavior.

**Citation: **Solovey G, Dawson SP (2010) Intra-Cluster Percolation of Calcium Signals. PLoS ONE 5(2):
e8997.
https://doi.org/10.1371/journal.pone.0008997

**Editor: **Giuseppe Chirico, University of Milano-Bicocca, Italy

**Received: **October 28, 2009; **Accepted: **January 11, 2010; **Published: ** February 18, 2010

**Copyright: ** © 2010 Solovey, Dawson. This is an open-access article distributed under
the terms of the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original author and
source are credited.

**Funding: **This work was supported by Proyecto de Investigación
Científica y Tecnológica (PICT) 17-21001 granted by
Agencia Nacional de Promocion Cientifica y Tecnologica (ANPCyT, http://www.agencia.gov.ar/), and by Proyecto de
Investigación Plurianual (PIP) 112-200801-01612 granted by Consejo
Nacional de Investigaciones Científicas y Técnicas
(CONICET) UBA (http://www.uba.ar/secyt/subsidios/index.php), Santa Fe Institute
(http://www.santafe.edu/). The funders had no role in study
design, data collection and analysis, decision to publish, or preparation of the
manuscript.

**Competing interests: ** The authors have declared that no competing interests exist.

## Introduction

The calcium (Ca) ion is a universal second messenger that is involved in a large number of physiological processes [2]. To this end, cells regulate cytosolic Ca concentration ([Ca]) very precisely. At basal conditions free cytosolic [Ca] is very low (). [Ca] is several orders of magnitude higher in the extracellular medium and in internal reservoirs, such as the endoplasmic reticulum. Different signals can induce the opening of specific Ca channels located on the plasma membrane or on the membrane of the internal reservoirs leading to local increments of the cytosolic [Ca] of various durations. This [Ca] change evokes different end responses depending upon the spatio-temporal distribution of [Ca]. Thus, it is of interest to measure the latter and how different factors shape it.

One of the Ca channels involved in intracellular Ca signals is the inositol 1,4,5-trisphosphate (IP) receptor (IPR) which is expressed in many cell types and is located at the surface of intracellular membranes such as the endoplasmic reticulum (ER), the sarcoplasmic reticulum (SR) and the nucleus. The IPR is biphasically regulated by Ca, with a bell-shaped open probability as a function of [Ca]. Kinetic models of the IPR take this dual effect into account by assuming that the receptor has at least one activating and one inhibitory site such that Ca binding to the first one induces channel opening (provided that IP is also bound to the receptor) and Ca binding to the second one induces channel closing [3]–[5]. Given that the affinity for Ca of the activating site is larger than that of the inhibitory site, a local increase of cytosolic Ca in the vicinity of an IPR with IP bound induces channel opening first. This leads to a phenomenon called Ca-induced Ca-release (CICR) because the Ca ions released by one channel can in turn trigger the opening of other nearby channels with IP bound. Ca channels are not uniformly distributed in the cell. IPR's, in particular, are usually organized in clusters on the membrane of the ER that are separated by a few microns [6]. These clusters have been estimated to be in size in oocytes [7], [8]. The simulations of [7] showed that previous observations could be reproduced assuming that between 25 and 35 IPR's opened simultaneously during puffs. A similar estimate was obtained in [8] using a mean-field model that assumed that all channels opened and closed simultaneously. Simulations that include a stochastic description of the individual channel openings and closings, however, show that at most half of the channels with IP bound are simultaneously open during a puff [8]. This implies that even in clusters with 50 IPR's with IP bound, the maximum number of simultaneously open channels is around 20. These results are consistent with observations of Ca signals in the human neuroblastoma SY5Y cell line in which puffs of up to 20 simultaneously open channels were observed [1]. Measurements performed using patches of the outer nuclear envelope of the DT40 cell line give smaller numbers of IPR's in each patch [9]. The non-uniform spatial organization of the IPR's together with the channel coupling induced by CICR gives rise to a large variety of intracellular Ca signals that go from very localized ones to waves that propagate throughout the cell [10].

The hierarchy of intracellular Ca signals that includes Ca “blips” (Ca release through a single IPR), “puffs” (Ca release through several IPR's in a cluster) and waves that propagate globally across
cells by successive cycles of CICR has been observed using fluorescence microscopy
and Ca sensitive dyes [10]–[13]. The
*Xenopus laevis* oocyte has been frequently used for this purpose
because of its relatively large size and because the only Ca channels that are present on the surface of the ER are IPR's. Fluorescent images of these signals obtained with
confocal microscopy do not resolve the inner-cluster structure. Therefore, different
modeling strategies have been presented in order to determine the properties of the
dynamics and spatial organization of IPR's within clusters that are compatible with these
experimental observations [7], [8], [14], [15]. In
particular, in [8], [16] we made the very simple assumption that the
number of IPR's that open during the first puff that occurs at a site
is given by the number of IPR's with IP bound. The underlying assumption was that the Ca released by the first open channel would induce the opening of all
the other IPR's of the cluster with IP bound. Therefore, if all the clusters had approximately the same
number of IPR's and all IPR's were equally sensitive to IP, the distribution of the number of channels that opened during a
puff could be approximated by a binomial or Poisson distribution [8], provided
that the probability that the channels become open were the same immediately before
the occurrence of each puff. This last condition would not be satisfied in a
non-stationary situation, *e.g.* if the concentrations of the
agonists right before the release event differed significantly from puff to puff. It
would not hold, in particular, for data containing sequences of puffs that are
coupled through CICR or to puffs in which the inhibitory effect of the Ca released in a previous event was noticeable, as described in [16]. In
oocytes, the latter is only relevant for very long records containing many puffs at
a site, which is usually not the case in most experiments. Calcium induced calcium
release is also affected by buffers that can trap Ca ions as they diffuse. This not only reduces the [Ca] but also alters the rate of Ca transport [17]. The distances that separate IPR's within a cluster are very small (10–20nm)
[9].
Thus, only large concentrations of very fast buffers could affect Ca-mediated inter-channel coupling in cases with many active channels
[18],
[19].
The assumption that all the channels with IP bound participate of the first puff of their site is the simplest
way of approaching the complex problem of Ca-mediated inter-channel communication. Yet, it is applicable as
long as the distance between IP-bound channels is not too large. In the present paper we drop this
assumption and analyze how Ca-mediated inter-channel coupling affects the distribution of puff
sizes. Our approach provides a simple tool to study some of the effects of buffers
on the intra-cluster dynamics.

The quantal properties of Ca release during puffs have recently been revealed in [1] using
total internal reflection fluorescence (TIRF) microscopy in intact mammalian cells
of the human neuroblastoma SY5Y cell line. The proximity of IPR's to the plasma membrane in this cell type allowed the
use of TIRF microscopy in which fluorescence can be elicited in a very small
(attoliter) volume. This, together with the use of a fast CCD camera, permitted a
much better temporal resolution than the one achieved with confocal microscopy. In
this way, abrupt step-wise transitions between fluorescence levels were observed
during the falling phase of puffs. Furthermore, many puffs could be elicited at each
release site due to the use of a membrane-permeable form of IP
[20]. The
authors then inferred that the step-wise transitions between fluorescence levels
occurred in multiples of a basic unit that they identified with the amplitude
contribution of each channel at the site [1]. Using this relationship
they could readily obtain the distribution of the number of channels that open
during a puff. Given that there is a large variability among cluster sites, they
analyzed the subset of events that occur in clusters with a similar number of IPR's. The authors did not find any sign of an inhibiting
effect of the Ca released in their records. In spite of that and even constraining
the data set as mentioned before, they found that a Poisson distribution failed to
reproduce the observed histogram of event sizes particularly in the region of small
events (*i.e.*, puffs with very few open channels). They could
approximately describe the distribution with a model that assumes a weak
cooperativity among channels. Inter-channel cooperativity is mediated by the Ca released through an open IPR that subsequently diffuses to a neighboring channel. Thus, the
distance between channels is a key factor that regulates the cooperativity strength
[21]. The approach of [1], however, does not take
space into account.

In the present paper we introduce a simple model that takes into account both the stochasticity due to IP binding and the distance-dependent Ca-mediated cooperativity. It can reproduce the event size distribution reported in [1] for events involving any number of open channels. The distribution obtained with our model approaches a binomial or Poisson distribution as the cooperativity strength increases so that the opening of one IPR induces the opening of all other IPR's with IP bound. This transition from Ca-dominated to IP-binding dominated stochasticity is similar to a percolation transition. It also occurs if the number of IPR's with IP bound increases. Therefore, the transition can be reflected on the distribution of the number of IPR's that open at a given release site.

Percolation in connection with Ca signals has been invoked to explain the transition from abortive to propagating waves in cells [22]–[24]. Our paper is the first to identify two limiting regimes of the intra-cluster dynamics that underlies puffs and to characterize the change between them as a percolation transition. Furthermore, we show how information on the transition between both regimes (the IP-binding and the Ca dominated behaviors) can be extracted from the distribution of the number of IPR's that open during a puff. Knowledge on this transition can, in turn, yield information on the largest distance over which Ca-mediated cooperativity acts and on the mean density of IP-bound IPR's of the clusters. In this way, we can estimate biophysical parameters that affect the intra-cluster dynamics from statistical information on the emergent collective behavior of the channels of the cluster.

The aim of the simple model that we introduce in this paper is to characterize the
basic mechanisms that shape the distribution of the number of channels that open
during puffs. In particular, we identify the competition between two stochastic
processes as the main determinant of the form of the distribution. Therefore, an
analysis of this form may give information on the relative weight of the two
competing processes. The model does not include a detailed description of the
dynamics that takes place during or between events. For some time most models of
intracellular Ca dynamics were deterministic (see *e.g.*
[25]).
The observation of local signals such as puffs led to the development of several
models that included a stochastic description of Ca release [14], [15], [26], [27] or of the spatial location of the IPR's [28]. It is currently clear that stochastic effects
are not only relevant for local release events but are a fundamental aspect of the Ca dynamics for the full range of observed signals, including waves
[29]–[32]. More information on
stochastic models of Ca signals can be found in a recent focus issue on the subject [33].
Simulations of these stochastic dynamic models could be used to probe the main
findings of the present paper.

## Results

### The Model

We introduce here a simple model to describe the distribution of puff sizes that
occur at sites with similar numbers of IPR's. The model is simple in the sense that it does not
include a detailed description of the dynamics of the individual channel
openings and closings or of IP or Ca binding and unbinding. However, it does include the
stochasticity associated to IP binding and channel coupling via CICR. Given the estimates of
[8], the model assumes that clusters occupy a fixed
size region (more specifically, a circle of radius, ) and that IPR's are randomly distributed over the cluster region
with uniform probability. Each IPR of the cluster has a probability of having IP bound. is the random variable that represents the number of available IPR's (*i.e.*, of IPR's with IP bound) of the cluster before a puff starts. The distribution
of this variable is a binomial of parameters and , that can be approximated by a Poisson distribution of
parameter for large and small enough (for example, with and the absolute difference between both cumulative distributions
is lower than for each ). The model considers that if an IPR with IP bound becomes open and Ca starts to flow through its pore all other IPR's with IP bound that are within a distance, , of the open IPR will also become open. These newly opened IPR's in turn trigger the opening of new IPR's with IP bound that are within the distance, , from an open one. This scheme triggers a cascade of openings
that stops when there are no more available IPR's within the radius of influence
(*i.e.*, the distance ) of any open IPR. This cascade determines the number, , of channels that open during a puff. We call the probability that there are available IPR's (with IP bound) in a cluster and the conditional probability that channels open during an event given that there are with IP bound in the cluster. Given that we are interested in the
distribution of event sizes, we only consider the situations for which . Therefore, we renormalize the probabilities so that and . In this way, is a binomial or a Poisson distribution divided by one minus
the probability that there are not IPR's with IP bound in the cluster. Using these renormalized versions of and , the probability, , of having a puff with open channels is given by:(1) is approximated by a Poisson distribution of parameter when . can be readily compared with distributions obtained from
experimental observations as the one displayed in Fig. 4D of [1].

### Factors That Shape the Distribution of Event Sizes

The two stochastic components of the model are evident in the expression of . reflects the stochasticity of IP binding and the one due to inter-channel coupling via CICR. The relative weight of both factors on the resulting depends on the relationship between two typical lengthscales of the problem: the radius of influence, (the maximum distance between channels at which CICR still works) and the mean distance between channels with IP bound, , which is a random variable that can be computed in terms of and the number of IPR's with IP bound, , as:(2) can take values between and . Closely related to is the density of available IPR's which is given by:(3) and are related by: .

The relationship between and determines the relative weight of both stochastic components
on . In particular, if is very large, the opening of any channel of the cluster will
eventually lead to the opening of all other available channels. If such a
situation holds for most events, then and will mainly be determined by the stochastic component due to IP binding, *i.e.*, . If, for most events, is very small, then most of will be concentrated near , regardless of how many available IPR's there are in each realization. We will refer to
both extreme behaviors as IP or Ca limited. Depending on the parameters of the model (, , , and ), one or the other situation is favored. However, in many
situations one or the other behavior is favored depending on the value of , *i.e.*, on the realization. In those cases,
the dominant stochastic component of depends on the value of .

We first illustrate how the distribution, , varies with the number of IPR's of the cluster, , while all other parameters are fixed. As increases, the most likely values that can take on also increase. This means that it is more probable to have more available IPR's at any given instance. On the other hand, since the spatial dimensions of the cluster are unchanged ( is fixed) the mean distance between available IPR's, , is more likely to be smaller (see Eq. 2). Given that the typical distance for CICR to occur, , is also fixed, it is more probable that be larger. Therefore, approaches as is increased. This is illustrated in Fig. 1 where we have plotted the distributions (solid circles) and (bars) obtained with 1000 realizations of our model using , , and three values of . In A, , the number of available channels is small for most realizations (its mean value is ) so that is dominated by inter-channel Ca-coupling and concentrated around small values of . In C, , the number of available channels is large for most realizations (its mean value is ) so that is typically smaller than (). In this case, is dominated by the IP-binding stochasticity and almost indistinguishable from the distribution of available channels, . The example of Fig. 1 B corresponds to a situation in between these two extreme cases with . We can observe how, as the number of available channels is more likely to be larger, approaches . We also observe that for and , and differ mainly in the region of small values of . This occurs because it is difficult for one open channel to induce the opening of another one if the mean inter-channel distance is large. Thus, if is small it is very rare that all available channels become open. In this way, the relative frequency of small events becomes larger than the fraction of instances with a small number of available channels.

Solid circles: distribution of puff sizes, , obtained with our model for , , and three values of : (A), (B) and (C). Histograms (in grey): corresponding distributions of available channels, for the same parameter values. All distributions were computed from 1000 realizations for each set of parameters.

A transition from Ca-dominated to IP-binding dominated stochasticity also occurs as is increased, while all other parameters are fixed. In this case, remains unchanged and so does the mean distance between available IPR's, . By changing it is possible to go from a situation in which is small for most events and is Ca-limited to a situation in which is large and is IP-binding limited. This is illustrated in Fig. 2 where we have plotted the distribution of event sizes that we obtain with our model for three different values of . For , the distribution is Ca-coupling limited and is concentrated around . As is increased, the relative frequency of events with small decreases. For , the distribution is IP-binding limited. In this example, is well approximated by a Poisson distribution of parameter (data not shown). The situation in between these extreme cases corresponds to and is able to reproduce reasonably well the experimental distribution of Fig. 4D of [1] (superimposed with bars in Fig. 2).

We show the probabiliy, , of having a puff with open channels obtained with our model for , , and (solid circles), (open circles) and (triangles). Each curve corresponds to 500 realizations of the model. We observe a transition from a Ca-dominated to a IP-binding dominated stochasticity distribution as increases. Superimposed with bars: experimental data taken from Fig. 4D of [1].

In the Ca limited behavior the number of open channels, , is small for most events, regardless of the value of . This implies for almost all events. In the IP-binding limited behavior all available IPR's become open ( in most cases). Therefore, in order to analyze the transition
between the Ca-dominated to IP-binding dominated stochasticity, we study how often events
occur for which all available IPR's become open. This happens trivially for events
with . Here we are interested in situations with . To this end, we compute numerically the probability that all
available IPR's, , become open, , which is a function of and of only one independent parameter, the dimensionless
radius of influence, , (see Methods). We plot
in Fig. 3 A
as a function of , for (circles), (squares) and (triangles). As expected, is an increasing function of for each value of . We also observe that is an increasing (sigmoidal-like) function of that goes from 0 (*i.e.*
in almost all cases, which corresponds to Ca-dominated stochasticity) to 1 (*i.e.*
in almost all cases, which corresponds to IP-binding dominated stochasticity) and that such transition
occurs over a smaller interval of values the larger is.

A: Probability that all available IPR's become open, , as a function of the dimensionless radius of influence, , for (circles), (squares) and (triangles). B: (circles), (squares) and (triangles) as functions of . The values of and for the case with are indicated in A with one and two asterisks, respectively.

We can think of the Ca-limited and the IP-binding limited situations as two phases and the transition between them as a phase transition in the limit of very large . This percolation-like transition occurs at a well defined value of in this limit. For finite values of we introduce two quantities, and , that determine the type of regime that we can expect (Ca-limited if or IP-binding limited if ) for each value of (see Methods). The arrows in Fig. 3 A indicate the values of (*) and (**) for the case. We show in Fig. 3 B plots of , and as functions of (Eq. (3)). It is important to note that these curves are the same, regardless of the specific parameter values of the model. We observe that all of them are decreasing functions of or, equivalently, of . is a stochastic variable that changes from realization to realization. Therefore, even for a given cluster (characterized by fixed values of , and ) and may take on different values depending on the realization. In this way, depending on and the values that may take on, a subset of the events that occur at a cluster may be IP-binding limited (those for which ) while others are not. An analogous situation may hold regarding the Ca-limited behavior. Furthermore, for some clusters, the Ca-limited condition may hold for some events and the IP-binding limited for others. If the parameters , , and are such that most realizations satisfy , then most events will be IP-binding limited. This happens if or are large enough, in which case the distribution of event sizes, , approaches the distribution of available channels, .

### Observing Percolation as a Function of Event Size

The results of Fig. 3 B imply
that there are clusters that can display different types of behaviors depending
on the event. For these clusters, we expect to find, in their distribution, , a trace of the transition to the limiting behavior that they
can display. Here we are interested in the percolation transition,
*i.e.*, the transition to the IP-binding dominated stochasticity. As already discussed, the
larger the more likely it is that all IPR's become open during the puff (see Fig. 3 A). Thus, the
transition to the IP-binding dominated stochasticity should occur as and, consequently, become larger. To study this transition we consider a cluster
with fixed parameters , , and (or in the Poisson limit) and define as the minimum value of such that . The definition of is based on the conditional probability, , which is independent of . In cases with finite , is meaningful provided that it be smaller than . Since decreases with (see Fig. 3
B), taking into account the definitions of and of (see Methods) we conclude
that and for all . Thus, we can approximate:(4)with less than 10% error. Inserting this approximation
in Eq. (1) we obtain:(5)(6)We then conclude that the tail of corresponds to IP-binding dominated events. Therefore, it should be possible to
approximate it by a (renormalized) binomial (provided that ) or Poisson distribution in the region of large . The left border of this IP dominated behavior, , gives information on , *i.e.* on the maximum distance for which
CICR-coupling can work effectively. Therefore, it should be possible to estimate by analyzing , *i.e.*, to infer a biophysical parameter that
characterizes the intra-cluster dynamics from statistical information on the
emergent collective behavior of the channels of the cluster.

### Determining Intra-Cluster Properties from Observations of the Cluster as a Whole

We now discuss how we can estimate from an experimental distribution of event sizes, . For the sake of simplicity, we assume that can be approximated by a renormalized Poisson distribution, , of unknown parameter . The goal of this section is to provide a way to estimate and , the value of at which and depart from one another (see Eq. (6)). Once is inferred, we estimate as using the function displayed in Fig. 3 B. To this end, we focus on the large tails of and by computing the complementary cumulative distribution functions:(7)(8)for . Given that is proportional to a Poisson distribution, there is an analytic expression for . Namely, , where is the incomplete function and is the integer part of . If the cluster is such that exists so that is larger than for and it is smaller otherwise, then, according to the calculation of the previous section, for . Therefore, the complementary cumulative distribution functions of Eqs. (7)–(8) also satisfy for .

We now describe how to estimate and . The aim is to obtain a (renormalized) Poisson distribution, that can approximate in the large region. If we find it, we assume that it is a good approximation of the distribution of available channels, . As illustrated in Fig. 1, the mean value, that is obtained using the experimental distribution, , is smaller than the one that would be obtained if was used instead. On the other hand, if is a good approximation of in the large region, then the mean value obtained with should be smaller than the size of the largest observed event, . This implies that(9)if can be approximated by a renormalized Poisson distribution of parameter . Therefore, we look for the best within a finite set of renormalized Poisson distributions of parameters satisfying (9). In order to estimate from the observations the relevant quantity that we need to obtain is , which is an integer. For this purpose, it is possible to use a rather coarse grid of values within the interval defined in (9). In particular, we have mainly used integer values of obtaining good results. Once the values are chosen, we compute the complementary cumulative distribution functions, given by (8) for each and . We then calculate the error of approximating by over the interval as a function of :(10)We set a threshold for the error, , and choose for each as the smallest value of for which . Finally, we choose the best as the one with the smallest .

The procedure is illustrated in Fig. 4 where the “experimental” distribution comes from a simulation of our model with , , and . In this case, . We show in Fig. 4 A the complementary cumulative distribution functions and in Fig. 4 B the errors for the values of that we have considered: (inverted triangles), (triangles), (squares) and (rhombes). Larger values of give very bad approximations and are not shown. We show in Fig. 4 C the values, , obtained for each using the threshold, (shown with a horizontal line in Fig. 4 B). In this example, the best value is for which . We estimate the density of IP-bound IPR's at which the departure between the experimental and the Poisson distribution occurs as , where we have used . Using the relationship displayed in Fig. 3 B, we estimate from which we get . This provides an estimate of the radius of influence which compares very well with the value that was used to generate the data, . Using the same procedure, we analyzed the data presented in Fig. 4D of [1] and obtained assuming .

A: for data obtained with our model using , , and (solid circles). Complementary cumulative Poisson distributions, , for (inverted triangles), (triangles), (squares), (rhombes). B: Error of approximating by the various for (see text for definition) as a function of . Symbols are the same as in A. From this figure we choose as the one that provides the best fit to the tail of . The error in the case is larger than 0.02 in most cases and falls outside the region displayed in the figure. C: for the four values of that we tested. We see that .

## Discussion

Intracellular Ca signals are built from localized release events in which Ca enters the cytosol through one or several channels. Ca release from the endoplasmic reticulum through IPR's is a key component of the Ca signaling toolkit in many cell types. IPR's are Ca channels that need to bind IP and Ca to become open and are usually organized in clusters on the membrane of the endoplasmic reticulum. The intra-cluster organization and the interactions of the channels within it affect the dynamics and extent of the signals. Therefore, their study is a matter of active research.

Recent experiments [1] that use super-resolution optical techniques are providing detailed data on elementary IPR-mediated Ca release events in mammalian cells. In the experiments, the number of IPR- Ca-channels that open during each event can be inferred from the observed puff amplitudes without much processing. The observations of [1] showed that the variability among clusters affected the shape of the event size distribution, . In order to get rid of this variability, the distribution coming from sites with similar properties was computed in [1]. The distribution, , obtained in this way was not Poisson, as we might have expected if the number of channels that opened during each event was proportional to the number of IPR's with IP bound in the cluster [8]. The authors of [1] could reproduce approximately (for events larger than a certain size) assuming a weak cooperativity among channels. Namely, they assumed that the probability that a channel became open scaled as some power of the number of open channels and obtained that the exponent was 1/3 from a fit to the data. The rationale for the cooperativity assumption relied on the fact that the IPR's of a cluster may be coupled via CICR induced by the Ca ions that travel from an open IPR to a neighboring one. The model of [1], however, did not take space into account and did not provide a mechanistic explanation for the obtained scaling.

In this paper we have presented a simple model that includes a description of the intra-cluster spatial organization with which we can reproduce the observed distribution over all event sizes. In the model the distribution, , is the result of the competition of two stochastic processes: IP binding and distance-dependent CICR. The model assumes a stationarity condition, namely, that the agonists concentration at the release site is the same immediately before the occurrence of each puff. This condition holds as long as puffs are independent of one another. This is consistent with the observations reported in [1] where cluster coupling was prevented using the slow Ca buffer EGTA and where IPR- Ca-inhibition does not play a significant role. In any case, our model is adequate to describe the distribution of first event sizes that occur at each cluster before Ca can exert any inhibiting effect.

There are two limiting cases in which one of the two stochastic processes considered
in the model is the main determinant of the distribution shape. If the mean distance
between IPR's with IP bound in the cluster is much smaller than the typical distance of
inter-channel coupling due to CICR for most events, the distribution is IP-binding limited and it can be approximated by a binomial or
Poisson distribution. In the opposite case, CICR dominates and the distribution is
peaked around . The Ca-limited and the IP-binding limited situations can be thought of as two phases and the
transition between them as a percolation-like transition in the limit in which the
number of IPR's with IP bound, , is very large. This interpretation of the factors that shape the
observed distribution can be tested with simulations of some of the stochastic
models of intracellular Ca signals reported in the literature (see *e.g.*
[33] and
references therein). They can also be tested experimentally. One possibility is to
change the most likely values of by changing the amount of IP that is photo-released in the cell. An alternative option is to
analyze for events coming from clusters that give rise, on average, to
larger events. According to the model, the distribution should approach a binomial
or Poisson distribution as the mean value of becomes larger while other parameters remain the same. Another way
to affect the balance between both stochastic components is to disrupt Ca-mediated inter-channel coupling by means of a fast buffer such as
BAPTA.

Given that is a stochastic variable that varies from event to event, the transition between the Ca-dominated and IP-binding dominated stochasticity described by the model may be reflected in the way that depends on the event size, . In fact, we have used this property to show how a fingerprint of this transition may be encountered in and how information on the inter-channel coupling distance may be extracted from it. This means that a parameter that characterizes the communication between pairs of channels can be estimated from statistical information on the emergent collective behavior of the channels of the cluster. This information could be used to analyze the effect of buffers on the intra-cluster dynamics, a matter that is of active current research [19], [34]. Our model provides a simple tool with which this effect can be analyzed in experiments.

## Methods

Each term of the sum that defines Eq. (1) is the product of two functions. We have an analytic expression for one of them, , but not for . Thus, we compute numerically performing realizations of the model with fixed values of , , and . The location of the channels within the cluster and which of them have IP bound vary among realizations and are chosen randomly (see Results). We only keep realizations with . Once we have the spatial distribution of available IPR's, we start each event by picking at random one of the IPR's with IP bound and assume it is open. If , we assume it gives rise to an event with . By changing the values of , , and we analyze how varies with them. In this way we can determine the values of the parameters that best reproduce the experimental observations. could be measured in units of the cluster spatial extent, , in which case we would get rid of one parameter of the problem, . We keep it to make a connection with the experimental data. However, it is important to note that the number of independent parameters of the model is 3, for finite and 2 in the limit in which can be approximated by a Poisson distribution.

For each value, , of available IPR's, we estimate the fraction of events such that the IPR's become open. This fraction is one for . For , we compute the probability that all available IPR's become open, , numerically, performing 500 stochastic realizations of our model
for each of which we fix the value of
*a priori*. Namely, we fix at the beginning the values of , and and then pick locations at random over the circle where we assume there are
available IPR's. From there on, the model goes on as before,
generating the cascade of openings that determines . The distribution of events with open channels for each value of gives . This function of depends on only one independent parameter, . As expected, it is an increasing function of (see Fig. 3
A).

We define two quantities, and , which are values of for which is either close to 1 or to 0, respectively. We compute them as follows. We first calculate a lower bound for as the minimum value of such that, if is larger than this lower bound, then . We calculate an upper bound for as the minimum value of for which . Then, we compute as the mean between these two bounds. We assume that the distance between the bounds is the error with which can be determined. We proceed analogously in the case of , but in this case the lower bound is the largest value of for which and the upper bound is the maximum value of for which . We compute and in this way using the numerical estimations of for various values of .

## Acknowledgments

We acknowledge useful discussions with Ian Parker and Ian Smith and that they provide us with their experimental data. We also acknowledge Giorgio Rispoli and the other two referees for their careful reading of the manuscript and useful comments and suggestions.

## Author Contributions

Conceived and designed the experiments: GS SPD. Performed the experiments: GS. Analyzed the data: GS SPD. Wrote the paper: SPD.

## References

- 1.
Smith IF, Parker I (2009) Imaging the quantal substructure of single IP
_{3}R channel activity during Ca^{2+}puffs in intact mammalian cells. Proc Natl Acad Sci (USA) 106: 6404–6409.IF SmithI. Parker2009Imaging the quantal substructure of single IP_{3}R channel activity during Ca^{2+}puffs in intact mammalian cells.Proc Natl Acad Sci (USA)10664046409 - 2. Berridge MJ, Bootman MD, Lipp P (1998) Calcium - a life and death signal. Nature 395: 645–648.MJ BerridgeMD BootmanP. Lipp1998Calcium - a life and death signal.Nature395645648
- 3.
DeYoung G, Keizer J (1992) A single-pool inositol 1,4,5-trisphosphate-receptor-based model
for agonist-stimulated oscillations in Ca
^{2+}concentration. Proc Natl Acad Sci (USA) 89: 9895–9899.G. DeYoungJ. Keizer1992A single-pool inositol 1,4,5-trisphosphate-receptor-based model for agonist-stimulated oscillations in Ca^{2+}concentration.Proc Natl Acad Sci (USA)8998959899 - 4.
Fraiman D, Ponce Dawson S (2004) A model of the IP
_{3}receptor with a luminal calcium binding site: stochastic simulations and analysis. Cell Calcium 35: 403–413.D. FraimanS. Ponce Dawson2004A model of the IP_{3}receptor with a luminal calcium binding site: stochastic simulations and analysis.Cell Calcium35403413 - 5.
Shuai J, Pearson JE, Foskett JK, Mak DOD, Parker I (2007) A kinetic model of clustered IP
_{3}receptors in the absence of Ca^{2+}feedback. Biophysical Journal 93: 1151–1162.J. ShuaiJE PearsonJK FoskettDOD MakI. Parker2007A kinetic model of clustered IP_{3}receptors in the absence of Ca^{2+}feedback.Biophysical Journal9311511162 - 6.
Yao Y, Choi J, Parker I (1995) Quantal puffs of intracellular Ca
^{2+}evoked by inositol trisphosphate in xenopus oocytes. J Physiol (Lond) 482: 533–553.Y. YaoJ. ChoiI. Parker1995Quantal puffs of intracellular Ca^{2+}evoked by inositol trisphosphate in xenopus oocytes.J Physiol (Lond)482533553 - 7.
Shuai J, Rose HJ, Parker I (2006) The number and spatial distribution of IP
_{3}receptors underlying calcium puffs in*xenopus*oocytes. Biophys J 91: 4033–4044.J. ShuaiHJ RoseI. Parker2006The number and spatial distribution of IP_{3}receptors underlying calcium puffs in*xenopus*oocytes.Biophys J9140334044 - 8.
Bruno L, Ventura AC, Dargan S, Ponce Dawson S (2009) Quantifying calcium fluxes underlying calcium puffs in
*xenopus laevis*oocytes. Cell Calcium. L. BrunoAC VenturaS. DarganS. Ponce Dawson2009Quantifying calcium fluxes underlying calcium puffs in*xenopus laevis*oocytes.Cell CalciumIn press. In press. - 9.
Ur-Rahman T, Skupin A, Falcke M, Taylor CW (2009) Clustering of IP
_{3}receptors by IP_{3}retunes their regulation by IP_{3}and Ca^{2+}. Nature 458: 655–659.T. Ur-RahmanA. SkupinM. FalckeCW Taylor2009Clustering of IP_{3}receptors by IP_{3}retunes their regulation by IP_{3}and Ca^{2+}.Nature458655659 - 10.
Sun XP, Callamaras N, Marchant JS, Parker I (1998) A continuum of IP
_{3}-mediated elementary Ca^{2+}signalling events in xenopus oocytes. J Physiol (Lond) 509: 67–80.XP SunN. CallamarasJS MarchantI. Parker1998A continuum of IP_{3}-mediated elementary Ca^{2+}signalling events in xenopus oocytes.J Physiol (Lond)5096780 - 11.
Parker I, Yao Y (1996) Ca
^{2+}transients associated with openings of inositol trisphosphate-gated channels in xenopus oocytes. J Physiol 491(Pt 3): 663–8.I. ParkerY. Yao1996Ca^{2+}transients associated with openings of inositol trisphosphate-gated channels in xenopus oocytes.J Physiol491(Pt 3)6638 - 12. Bootman MD, Berridge MJ, Lipp P (1997) Cooking with calcium: the recipes for composing global signals from elementary events. Cell 91: 367–373.MD BootmanMJ BerridgeP. Lipp1997Cooking with calcium: the recipes for composing global signals from elementary events.Cell91367373
- 13.
Callamaras N, Marchant JS, Sun XP, Parker I (1998) Activation and co-ordination of IP
_{3}-mediated elementary Ca^{2+}events during global Ca^{2+}signals in*xenopus oocytes*. J Physiol 509: 81–91.N. CallamarasJS MarchantXP SunI. Parker1998Activation and co-ordination of IP_{3}-mediated elementary Ca^{2+}events during global Ca^{2+}signals in*xenopus oocytes*.J Physiol5098191 - 14. Swillens S, Dupont G, Combettes L, Champell P (1999) From calcium blips to calcium puffs: Theoretical analysis of the requirements for interchannel communication. Proc Natl Acad Sci (USA) 96: 13750–13755.S. SwillensG. DupontL. CombettesP. Champell1999From calcium blips to calcium puffs: Theoretical analysis of the requirements for interchannel communication.Proc Natl Acad Sci (USA)961375013755
- 15.
Falcke M (2003) On the role of stochastic channel behavior in intracellular
Ca
^{2+}dynamics. Biophys J 84: 42–56.M. Falcke2003On the role of stochastic channel behavior in intracellular Ca^{2+}dynamics.Biophys J844256 - 16. Fraiman D, Pando B, Dargan S, Parker I, Ponce Dawson S (2006) Analysis of puff dynamics in oocytes: Interdependence of puff amplitude and interpuff interval. Biophys J 90: 3897–3907.D. FraimanB. PandoS. DarganI. ParkerS. Ponce Dawson2006Analysis of puff dynamics in oocytes: Interdependence of puff amplitude and interpuff interval.Biophys J9038973907
- 17. Allbritton N, Meyer T, Stryer L (1992) Range of messenger action of calcium ion and inositol 1,4,5-trisphosphate. Science 258: 1812–1815.N. AllbrittonT. MeyerL. Stryer1992Range of messenger action of calcium ion and inositol 1,4,5-trisphosphate.Science25818121815
- 18.
Falcke M (2003) Buffers and oscillations in intracellular
Ca
^{2+}dynamics. Biophys J 84: 28–41.M. Falcke2003Buffers and oscillations in intracellular Ca^{2+}dynamics.Biophys J842841 - 19.
Zeller S, Rüdiger S, Engel H, Sneyd J, Warnecke G, et al. (2009) Modeling of the modulation by buffers of Ca
^{2+}release through clusters of IP_{3}receptors. Biophys J 97: 992–1002.S. ZellerS. RüdigerH. EngelJ. SneydG. Warnecke2009Modeling of the modulation by buffers of Ca^{2+}release through clusters of IP_{3}receptors.Biophys J979921002 - 20.
Smith IF, Wiltgen SM, Parker I (2009) Localization of puff sites adjacent to the plasma membrane:
Functional and spatial characterization of Ca
^{2+}signaling in SH-SY5Y cells utilizing membrane-permeant caged IP_{3}. Cell Calcium 45: 65–76.IF SmithSM WiltgenI. Parker2009Localization of puff sites adjacent to the plasma membrane: Functional and spatial characterization of Ca^{2+}signaling in SH-SY5Y cells utilizing membrane-permeant caged IP_{3}.Cell Calcium456576 - 21. Ponce Dawson S, Keizer J, Pearson JE (1999) Fire-diffuse-fire model of dynamics of intracellular calcium waves. Proc Natl Acad Sci (USA) 96: 6060–6063.S. Ponce DawsonJ. KeizerJE Pearson1999Fire-diffuse-fire model of dynamics of intracellular calcium waves.Proc Natl Acad Sci (USA)9660606063
- 22. Bär M, Falcke M, Levine H, Tsimring LS (2000) Discrete stochastic modeling of calcium channel dynamics. Phys Rev Lett 84: 5664–5667.M. BärM. FalckeH. LevineLS Tsimring2000Discrete stochastic modeling of calcium channel dynamics.Phys Rev Lett8456645667
- 23. Pencea CS, Hentschel HGE (2000) Excitable calcium wave propagation in the presence of localized stores. Phys Rev E 62: 8420–8426.CS PenceaHGE Hentschel2000Excitable calcium wave propagation in the presence of localized stores.Phys Rev E6284208426
- 24. Timofeeva Y, Coombes S (2004) Directed percolation in a two-dimensional stochastic fire-diffuse-fire model. Phys Rev E 70: 062901: 1–3.Y. TimofeevaS. Coombes2004Directed percolation in a two-dimensional stochastic fire-diffuse-fire model.Phys Rev E70: 06290113
- 25. Schuster S, Marhl M, Höfer T (2002) Modelling of simple and complex calcium oscillations. from single-cell responses to intercellular signalling. Eur J Biochem 269: 1333–55.S. SchusterM. MarhlT. Höfer2002Modelling of simple and complex calcium oscillations. from single-cell responses to intercellular signalling.Eur J Biochem269133355
- 26. Shuai JW, Jung P (2002) Optimal intracellular calcium signaling. Phys Rev Lett 88: 068102: 1–4.JW ShuaiP. Jung2002Optimal intracellular calcium signaling.Phys Rev Lett88: 06810214
- 27. Shuai JW, Jung P (2003) Optimal ion channel clustering for intracellular calcium signaling. Proc Natl Acad Sci (USA) 100: 506–510.JW ShuaiP. Jung2003Optimal ion channel clustering for intracellular calcium signaling.Proc Natl Acad Sci (USA)100506510
- 28.
Bugrim AE, Zhabotinsky AM, Epstein IR (1997) Calcium waves in a model with a random spatially discrete
distribution of Ca
^{2+}release sites. Biophys J 73: 2897–2906.AE BugrimAM ZhabotinskyIR Epstein1997Calcium waves in a model with a random spatially discrete distribution of Ca^{2+}release sites.Biophys J7328972906 - 29.
Skupin A, Kettenmann H, Winkler U, Wartenberg M, Sauer H, et al. (2008) How does intracellular Ca
^{2+}oscillate: By chance or by the clock? Biophys J 94: 2404–2411.A. SkupinH. KettenmannU. WinklerM. WartenbergH. Sauer2008How does intracellular Ca^{2+}oscillate: By chance or by the clock?Biophys J9424042411 - 30. Perc M, Green AK, Dixon CJ, Marhl M (2008) Establishing the stochastic nature of intracellular calcium oscillations from experimental data. Biophys Chem 132: 33–38.M. PercAK GreenCJ DixonM. Marhl2008Establishing the stochastic nature of intracellular calcium oscillations from experimental data.Biophys Chem1323338
- 31. Perc M, Gosak M, Marhl M (2007) Periodic calcium waves in coupled cells induced by internal noise. Chem Phys Lett 437: 143–147.M. PercM. GosakM. Marhl2007Periodic calcium waves in coupled cells induced by internal noise.Chem Phys Lett437143147
- 32. Perc M, Gosak M, Marhl M (2006) From stochasticity to determinism in the collective dynamics of diffusively coupled cells. Chem Phys Lett 421: 106–110.M. PercM. GosakM. Marhl2006From stochasticity to determinism in the collective dynamics of diffusively coupled cells.Chem Phys Lett421106110
- 33.
Falcke M (2009) Introduction to focus issue: Intracellular
Ca
^{2+}dynamics— a change of modeling paradigm? Chaos: An Interdisciplinary Journal of Nonlinear Science 19: 037101.M. Falcke2009Introduction to focus issue: Intracellular Ca^{2+}dynamics— a change of modeling paradigm?Chaos: An Interdisciplinary Journal of Nonlinear Science19037101 - 34.
Shuai J, Pearson JE, Parker I (2008) Modeling Ca
^{2+}feedback on a single inositol 1,4,5-triphosphate receptor and its modulation by Ca^{2+}buffers. Biophysical Journal 95: 3738–3752.J. ShuaiJE PearsonI. Parker2008Modeling Ca^{2+}feedback on a single inositol 1,4,5-triphosphate receptor and its modulation by Ca^{2+}buffers.Biophysical Journal9537383752