You are on page 1of 15

PHYSICAL REVIEW E 78, 041915 共2008兲

Simplified model of cytosolic Ca2+ dynamics in the presence of one or several clusters
of Ca2+-release channels
G. Solovey,1 D. Fraiman,2 B. Pando,3 and S. Ponce Dawson1
Departamento de Física, FCEN-UBA, Ciudad Universitaria, Pabellón I, (1428) Buenos Aires, Argentina
Departamento de Matemática y Ciencias, Universidad de San Andrés, Buenos Aires, Argentina
Department of Physics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA
共Received 22 April 2008; published 23 October 2008兲
Calcium release from intracellular stores plays a key role in the regulation of a variety of cellular activities.
In various cell types this release occurs through inositol-triphosphate 共IP3兲 receptors which are Ca2+ channels
whose open probability is modulated by the cytosolic Ca2+ concentration itself. Thus, the combination of Ca2+
release and Ca2+ diffusion evokes a variety of Ca2+ signals depending on the number and relative location of
the channels that participate of them. In fact, a hierarchy of Ca2+ signals has been observed in Xenopus laevis
oocytes, ranging from very localized events 共puffs and blips兲 to waves that propagate throughout the cell. In
this cell type channels are organized in clusters. The behavior of individual channels within a cluster cannot be
resolved with current optical techniques. Therefore, a combination of experiments and mathematical modeling
is unavoidable to understand these signals. However, the numerical simulation of a detailed mathematical
model of the problem is very hard given the large range of spatial and temporal scales that must be covered.
In this paper we present an alternative model in which the cluster region is modeled using a relatively fine grid
but where several approximations are made to compute the cytosolic Ca2+ concentration 共关Ca2+兴兲 distribution.
The inner-cluster 关Ca2+兴 distribution is used to determine the openings and closings of the channels of the
cluster. The spatiotemporal 关Ca2+兴 distribution outside the cluster is determined using a coarser grid in which
each 共active兲 cluster is represented by a point source whose current is proportional to the number of open
channels determined before. A full reaction-diffusion system is solved on this coarser grid.

DOI: 10.1103/PhysRevE.78.041915 PACS number共s兲: 87.17.⫺d, 82.33.⫺z, 02.30.Jr, 87.16.Vy

I. INTRODUCTION Fluorescence microscopy techniques allow the visualiza-

tion of Ca2+ signals and, at the same time, provide a means
The calcium ion 共Ca2+兲 is a universal second messenger by which the dynamics of IP3R’s can be studied under more
that is involved in the regulation of a large number of physi- physiological conditions than with other methods. Neverthe-
ological processes 关1兴. To attain this difficult task, the cyto- less, since the spatial resolution of fluorescent images is usu-
solic Ca2+ concentration 共关Ca2+兴兲 is precisely regulated. One ally of hundreds of nanometers, it is impossible to resolve
of the channels that regulates intracellular calcium is the the 关Ca2+兴 spatiotemporal distribution within the cluster re-
inositol 1,4,5-trisphosphate 共IP3兲 receptor 共IP3R兲 which is gion or to study the dynamics of single IP3R’s. The use of
expressed in many cell types and is located at the surface of mathematical models that span the scale that goes from
intracellular membranes such as the endoplasmic reticulum single channel pores 共艋10 nm 关15兴兲 to inter-cluster distances
共ER兲, the sarcoplasmic reticulum 共SR兲, and the nucleus. The 共⬃2 ␮m 关5兴兲 is thus unavoidable in order to obtain a detailed
IP3R is biphasically regulated by Ca2+, with a bell-shaped description of Ca2+ puffs. Furthermore, these models provide
open probability as a function of 关Ca2+兴. This leads to a a framework within which one can test hypothetical models
phenomenon called Ca2+-induced Ca2+ release 共CICR兲 be- of single IP3R’s. In fact, there is a large body of experimental
cause the Ca2+ ions released by one channel can in turn trig- and modeling results on single IP3R’s 关16–20兴 and it is im-
ger the opening of another nearby channel. portant to advance towards a comprehensive description that
The Xenopus laevis oocyte is a useful system to study combines this variety of observations in a unified way. The
intracellular Ca2+ signals because of its large size and be- use of mathematical models is unavoidable for this purpose
cause the only Ca2+ channels that are present at the surface too. On the other hand, a unified description of all the rep-
of the ER are IP3R’s. In oocytes, IP3R’s are organized in ertoire of Ca2+ signals that have been observed requires that
clusters of width ⬃100 nm 关2–4兴 that are separated by a few the mm scale be also included. This diversity of spatial 共and
micrometers 关5兴. As a result of the combination of CICR and temporal兲 scales represents a challenge to theoretical ap-
this spatial organization 关6–8兴, cytosolic Ca2+ signals in oo- proaches. The numerical simulation of the reaction-diffusion
cytes display a hierarchical spatiotemporal organization, systems that are usually used to model the cytosolic Ca2+
which includes Ca2+ “blips” that represent the release of dynamics is a very stiff 共and computationally expensive兲
Ca2+ through a single IP3R 关9–11兴, “puffs” that involve the problem. For this reason, several simplified models and the-
concerted opening of several IP3R’s in a cluster 关10–13兴, and oretical approaches have been proposed to analyze different
Ca2+ waves that propagate globally across cells by succes- aspects of Ca2+ signals.
sive cycles of CICR and Ca2+ diffusion between clusters The simplest signal that can be studied is the one that
关12–14兴. The whole range of signals has been observed using arises due to the opening of a single Ca2+ channel. The
fluorescence microscopy and Ca2+ sensitive dyes. analysis of the 关Ca2+兴 distribution in the presence of a single

1539-3755/2008/78共4兲/041915共15兲 041915-1 ©2008 The American Physical Society

SOLOVEY et al. PHYSICAL REVIEW E 78, 041915 共2008兲

source is actually the first step of our work. Several studies the channel. This method was applied in Ref. 关35兴 to study
have used approximations of the complete reaction-diffusion Ca2+ “blips.” In their work, the authors introduce an hybrid
problem to find analytical expressions for the 关Ca2+兴 close to algorithm to reliably link the stochastic transitions of the
the open channel. These results were put in a rigorous basis state of the channel with the deterministic concentration vari-
and extended in 关21兴. More recently, Bentele and Falcke 关22兴 ables. So far this method has been applied to single channel
have exploited the fast time scales of the local concentrations simulations 关35兴, where it has shown its accuracy compared
build up to find analytical expressions for the current and to standard methods.
关Ca2+兴 close to the open channel. Shuai and Parker 关23兴 also In this work we present an alternative model in which a
simulated microdomains of Ca2+ ions and Ca2+-bound indi- fine grid is used to determine the 关Ca2+兴 concentration distri-
cator dye around an open channel in order to understand the bution within the cluster but its dynamics is described within
processes that determine the temporal resolution and the a quasistationary approximation. In order to determine the
noise level of single-channel Ca2+ fluorescence signals 共SC- contribution of each open channel to the 关Ca2+兴 distribution
CaFTs兲. within the cluster, we analyze first this distribution when
Other signals involve the release of Ca2+ from several there is only one channel in the presence of the same amount
open channels in a cluster. The dynamics of clusters of of buffers as in the full scenario. The total 关Ca2+兴 within the
IP3R’s has been studied with a variety of approaches which cluster when there are several open channels is then approxi-
focus on different properties depending on the spatial and mated by a linear combination of the contributions due to
time scale that they resolve. One type of model, called “fire- each individual open channel. This detailed description of
diffuse-fire,” introduced to study 关Ca2+兴 waves, simplify 关Ca2+兴 within the cluster is used to determine the openings
clusters by considering them as discrete excitable Ca2+ re- and closings of the various channels in each release site. In
lease elements that open when a certain threshold amount of this way we obtain the total Ca2+ current that is released
Ca2+ is achieved 关24兴. This model is useful to understand from any given cluster. The spatiotemporal 关Ca2+兴 distribu-
general properties of wave propagation, such as the transition tion outside the cluster is determined using a coarser grid in
between saltatory and continuous propagation, based on which each 共active兲 cluster is represented by a point source
simple physical ideas, but does not take the internal structure whose current is proportional to the number of open channels
and complexity of the clusters into account. Several math- determined before. A full reaction-diffusion system is solved
ematical models have been proposed to describe Ca2+ release on this coarser grid.
through clustered IP3R’s and some of them rely on the hy- This model allows us to obtain the distribution of Ca2+ in
pothesis that channels are in such close contact that 关Ca2+兴 the presence of clusters of IP3R’s in a much shorter time than
can be considered homogeneous throughout the cluster the one required to solve a complete system of reaction-
关25–30兴. This assumption allows for relatively fast simula- diffusion equations on a very fine grid. The Ca2+-bound dye
tions of cytosolic Ca2+ dynamics since Ca2+ diffusion within distribution, a quantity that is directly related to the fluores-
the cluster can be neglected. This type of models can be cence distribution observed experimentally, is readily deter-
further simplified as done in Ref. 关31,32兴, where it is as- mined with the model. This simplified solution describes cor-
sumed that 关Ca2+兴 at one cluster depends only on the number rectly global Ca2+ signals with the spatial resolution of
of nearest release sites where there are open channels and experiments using the stochastic single channel activity as a
that each neighboring 共active兲 site adds a time-independent building block.
contribution to 关Ca2+兴. In the case of only one cluster the The organization of the paper is as follows. We first in-
system is spherically symmetric, which further reduces the troduce the 共full兲 reaction-diffusion model that describes the
computational effort. In this case the cluster can be consid- dynamics of intracellular Ca2+ for the problem of interest
ered as a localized point source with a Ca2+ current that 共Sec. II兲. In Sec. III we first study the solutions of this model
depends on the number of open channels. However, if one is in the presence of a single localized source. We then analyze
willing to include a more accurate description of the to what extent this single channel solution can be used to
Ca2+-release process kinetics, which depends on the time at determine the 关Ca2+兴 distribution within the cluster region
which each channel of the cluster becomes open, it is neces- when there are several open channels. We approximate the
sary to have a better estimate of the 关Ca2+兴 value at each numerically determined solution in various ways and analyze
channel pore within the cluster. Namely, the Ca2+ concentra- the validity of the approximations in a number of particular
tion nearby an open Ca2+ channel has a sharp spatial decay examples. In Sec. IV we show that the 关Ca2+兴 spatio-
关18,21,23,33,34兴, so even at interchannel distances 共⬃tens of temporal distribution outside the cluster region can be deter-
nm兲, the 关Ca2+兴 gradient is so high that neglecting its inho- mined using the full reaction-diffusion model on a grid that
mogeneities may not be a good approximation. The existence is coarser than the cluster size. We show that, on this coarser
of these deep gradients that usually build up very fast impose grid, each cluster can be represented by a point source whose
a serious difficulty to quantitative modeling approaches. total current is equal to the one determined within the cluster
Keeping the same spatial resolution both inside and outside region. In Sec. V we put everything together to show in a
the cluster is computationally expensive 关2,18,35兴. One alter- step by step manner how to build a complete description of
native is to use a spatial grid that is finer inside than outside Ca2+ dynamics resulting from an isolated puff. In Sec. VI we
the clusters. However, finer spatial grids usually call for generalize the model to the case of multiple clusters and
smaller time steps and, therefore, longer computation times. show that it is able to reproduce efficiently Ca2+ signals that
Another option is to use a finite element method to resolve propagate throughout the cell. Finally, we include a final dis-
the extreme spatial gradients of concentration values close to cussion with conclusions in Sec. VII.


z (µm) (a)
0.1 µm 20
0.1 µm

No. channel
0.8 10
0.4 x
0.2 5
x 1 µm 1 µm y
0 0.005 0.01 0.015 0.02
Time (s)

FIG. 1. 共a兲: Geometry of the cytosolic space. The illustration corresponds to a situation with one cluster of IP3R’s centered around the
origin, r = 0 and when individual IP3R’s are considered as the Ca2+ sources. The plane z = 0 represents the surface of the ER. 共b兲 A detailed
view of a cluster of Ns = 20 channels randomly distributed within a region of 300⫻ 300 nm at the surface of the ER. 共c兲 A graphical
representation of the open and close time of each of the Ns channels. The length of the black lines indicate the time interval during which
the channels remain open. In this case the behavior of the channels was fixed a priori by setting the open and closed time for each of them.



We present here a model of Ca2+ diffusion in a cytosolic ⳵关CaS兴

medium with molecular buffers and localized Ca2+ sources. = RCaS , 共2d兲
The model is a standard reaction diffusion scheme, similar to
others that have been introduced previously 关2,8,23,30,35兴. It where DCa2+, DFluo, and DEGTA are the diffusion coefficients
includes the following species: cytosolic calcium 共Ca2+兲, an of Ca2+, Fluo, and EGTA, respectively. The reaction terms
immobile endogenous buffer 共S兲, a cytosolic Ca2+ indicator RCaX derived from the kinetic scheme, Eq. 共1兲, can be written
共Fluo4兲 and an exogenous mobile buffer 共EGTA, or as
BAPTA兲. The Ca2+ indicator represents the dye usually used
in fluorescent microscopy experiments 关36,37兴. Exogenous
RCaX = kon 关Ca2+兴共关X兴T − 关CaX兴兲 − koff
关CaX兴. 共3兲
buffers are used in experiments to prevent the initiation of The Ca entry through 共open兲 IP3R’s is included in the
Ca2+ waves 关36,37兴. The equations we derive in this section boundary conditions 关21兴. Namely, we consider the cytosolic
are the same whether we use BAPTA or EGTA 共only some space as a parallelepiped of sides Lx, Ly, and Lz. The plane
parameter values change兲. In the following, we use EGTA z = 0 represents the surface of the ER membrane, where there
except explicitly noted. We consider that a single Ca2+ ion are Ns sources of 共small兲 area ␦S centered at positions rs共i兲
can bind to a single buffer or dye molecule according to = 共xsi , y si , 0兲 , i = 1 , . . . , Ns 关see Figs. 1共a兲 and 1共b兲兴. We assume
no flux boundary conditions everywhere for all concentra-
Ca2+ + X CaX, 共1兲 tions except for free Ca2+ at the location of the sources,
where we consider the condition 关21兴

where X represents S, EGTA or Fluo, and kon X X

and koff are the ⳵
− DCa2+ 关Ca2+兴关rs共i兲,t兴 = Jch for i = 1, . . ,Ns , 共4兲
forward and backward binding rate constants of the corre- ⳵z
sponding reaction, respectively. We assume that the total
where Jch is the Ca2+ flux through the sources. Namely, Jch
concentration of Ca2+ indicator, mobile and immobile buffer I
remain constant 共关Fluo兴T, 关EGTA兴T, and 关S兴T, respectively兲 = 共2Fch␦S兲 , where Ich is the single channel current 共which is
and that the diffusion coefficient of their free and equal to zero if the channel is closed兲 and F is Faraday’s
Ca2+-bound forms are equal. Therefore, we can calculate constant 共96, 485 C mol−1兲. At the surface of the ER there are
关Fluo兴, 关S兴, and 关EGTA兴 by subtracting the concentration of also Ca2+ pumps that take Ca2+ ions back into the lumen of
the Ca2+ bound forms to the total concentration. Given these the ER, collaborating to restore and keep 关Ca2+兴 at its basal
assumptions, the set of partial differential equations that give level. Nevertheless, we do not consider Ca2+ pumps because
the spatial and temporal evolution of free and bound Ca2+ is they act in a slower time scale.
the following: In this paper, we solve this full model in two types of
situations. In one of them, the Ns Ca2+ sources correspond to
⳵关Ca2+兴 individual IP3R-Ca2+ channels. In the other one, individual
= DCa2+ⵜ2关Ca2+兴 − 兺 RCaX , 共2a兲 channels are not resolved and the sources correspond to ei-
⳵t X
ther a finite region within the cluster with several channels
inside or to the whole cluster. When individual channels are
⳵关CaFluo兴 resolved, we either model their behavior using the stochastic
= DFluoⵜ2关CaFluo兴 + RCaFluo , 共2b兲
⳵t De Young and Keizer model 关17兴 with a type of simulation

SOLOVEY et al. PHYSICAL REVIEW E 78, 041915 共2008兲

that resembles the ones by Swillens et al. in Ref. 关2兴 or we TABLE I. Numerical parameters used to solve the 共full兲
assume that they remain open for a finite amount of time that reaction-diffusion model that describes the dynamics of intracellular
is fixed a priori. The purpose of the latter is to analyze the Ca2+ 共Sec. II兲 关3,38兴.
quality of certain approximations in different situations. The
kinetic model, on the other hand, assumes that the channels Parameter Value Units
can be in a finite set of states, some of which are open and
Free Calcium:
others are closed, with transition probabilities between states
that can depend on 关Ca2+兴 共see the Appendix兲. When the DCa2+ 200 ␮m2 / s
channels are open we consider that they release a fixed cur- 关Ca2+兴basal 0.1 ␮M
rent Ich. When individual channels are not resolved, the time Calcium indicator 共Fluo4-dextran兲:
course of the released current is computed separately, as ex- DFluo 15 ␮ m 2s
plained later. kFluo
on 150 ␮M−1s−1
All numerical simulations of the model are done using a Fluo
koff 300 1/s
forward Euler method in time and an explicit finite- 关Fluo兴T 40 ␮M
difference formula in space with a second order expression
Exogenous buffer 共EGTA兲:
共first neighbors兲 for the Laplacian. Different sizes of the in-
tegration region and of the spatial grid are used in different DEGTA 200 ␮m2 / s
cases. For the initial condition, we assume that all concen- kEGTA
on 5 ␮M−1s−1
trations are homogeneously distributed, Ca2+ is at basal con- koff 0.75 1/s
centration and all species are in equilibrium among them- 关EGTA兴T 300 ␮M
selves 共RCaX = 0 for all X兲, so that Exogenous buffer 共BAPTA兲:
关Ca2+兴共r,t = 0兲 = 关Ca2+兴basal , 共5兲 DBAPTA 200 ␮m2 / s
kBAPTA 500 ␮M−1s−1

冉 冊
kon 关Ca2+兴basal koff 100 1/s
关CaX兴共r,t = 0兲 = 关X兴T , 共6兲 关BAPTA兴T 300 ␮M
kon关Ca2+兴basal + koff
Endogenous stationary buffer:
where X represents S, EGTA, BAPTA, or Fluo. The physical kSon 400 ␮M−1s−1
parameter values are listed in Table I. They correspond to S
koff 800 1/s
values that have been used in other modeling papers on in-
关S兴T 300 ␮M
tracellular calcium dynamics 关3,38兴.
Calcium current:
Ich 0.1 pA
A. Analyzing the Ca2+ distribution near one open channel we will consider the distance along the x coordinate. The
time scale to achieve such stationary solution depends on x.
In order to build a model capable of describing, in a sim- We show in Fig. 2共b兲 the spatial distribution of
plified manner, the coupled dynamics of cytosolic Ca2+ and 关Ca2+兴共x , 0 , 0 , t⬘兲 for different times, t⬘, and we see that it
of several IP3R’s within a cluster, we first analyze how tends to a stationary solution, Cst共x兲 asymptotically in time.
关Ca2+兴 behaves in the presence of a very localized source that Our results agree with previous simulations of the temporal
opens for a fixed time 共i.e., we do not model the channel and spatial distribution of free Ca2+ in the cytosol around an
kinetics in this case兲. To this end, we performed a series of open channel 关23兴. In addition, theoretical cytosolic 关Ca2+兴
numerical simulations of the model described in Sec. II as- profiles for different times were already reported in Ref.
suming that there is only one Ca2+ channel present 共Ns = 1兲 关22兴, also showing that the steady state near the channel is
located at rs共1兲 = 共0 , 0 , 0兲. We consider a cytosolic space of quickly established. Here we quantify the differences be-
sides Lx = 2.7 ␮m, Ly = 2.7 ␮m, and Lz = 1.5 ␮m, a grid spac- tween the actual profile and the steady profile. We have de-
ing dx = dy = dz = 20 nm and time step dt = 0.3 ␮s. Results ob- termined that if we replace 关Ca2+兴共x , 0 , 0 , t兲 by
tained with this dt were indistinguishable from those ob- 关Ca2+兴共x , 0 , 0 , 10 ms兲 共just before the channel closes兲 the er-
tained with smaller values of dt. In the simulations, the ror inside the cluster region is lower than 10% after a time
channel opens at t = 0 and releases a constant calcium current t ⯝ 3 ms for x 艋 60 nm 关see Fig. 2共c兲兴. Therefore, neglecting
共Ich兲 during 10 ms 共a typical open duration for an IP3R兲 after the ⬃3 ms transient, we can approximate
which the current becomes equal to zero again. A typical
time course of the 关Ca2+兴 at different distances from the 关Ca2+兴共x,0,0,t兲 ⬇ Cst共x兲 ⬅ 关Ca2+兴共x,0,0,10 ms兲,
source and at the surface of the ER 关共x , 0 , 0兲 , x
= 0 , 20, 40, . . . , nm兴, is shown in Fig. 2共a兲. We will analyze 0 ⬍ t 艋 tend, x 艋 60 nm, 共7兲
separately what happens before and after the channel closes.
We can see that after the channel opens, 关Ca2+兴 at the within the cluster region.
surface of the ER tends to a stationary solution that depends In Eq. 共7兲, tend is the time at which the channel closes
only on the distance to the source. Without loss of generality, 共here tend = 10 ms兲 and we approximate the stationary solu-


100 100
(a) x = 0 nm (b) t’ = 0.1 ms
x = 20 nm t’ = 1 ms
x = 40 nm t’ = 5 ms
x = 60 nm t’ = 10 ms
10 x = 80 nm 10
[Ca ] (µM)

[Ca2+] (µM)
x = 120 nm
x = 180 nm
x = 240 nm

1 1

0.1 0.1

-0.005 0 0.005 0.01 0.015 0.02 0.025 -0.6 -0.4 -0.2 0 0.2 0.4 0.6
Time (s) x (µm)
100 100 10
(c) (d) x = 0 nm (e) x = 0 nm
x = 20 nm x = 20 nm
x = 40 nm x = 40 nm
10 x = 60 nm x = 60 nm
Relative error (%)

10 x = 80 nm x = 80 nm
[Ca2+] (µM)

[Ca2+] (µM)
x = 120 nm x = 120 nm
x = 180 nm x = 180 nm
1 x = 240 nm 1 x = 240 nm
x = 0 nm
x = 20 nm
x = 40 nm 1
0.1 x = 60 nm
x = 80 nm
x = 100 nm
x = 120 nm
0.01 0.1 0.1
0 0.002 0.004 0.006 0.008 0.01 0.01 0.0105 0.011 0.0115 0.012 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Time (s) Time (s) Time (s)

FIG. 2. Results of the simulation of the 共full兲 reaction-diffusion model presented in Sec. II in the presence of one localized source at
rs = 共0 , 0 , 0兲. The source releases Ca2+ ions at 0.1 pA between t = 0 and t = 0.01 s. 共a兲 Each trace shows 关Ca2+兴 at a different distance x from
the mouth of the channel. The solid line for t ⬍ 10 ms represents the 关Ca2+兴 value reached after 10 ms, i.e., Cst共x兲. After the channel closes
共tend = 10 ms兲, the 关Ca2+兴 decays slowly to its basal level. 共b兲 Spatial profile of 关Ca2+兴 at four times 0.1, 1, 5, and 10 ms 共just at the time of
the channel closing兲. 共c兲 Relative difference 共in %兲 between Con共x兲 + 关Ca2+兴basal and 关Ca2+兴共x , 0 , 0 , t兲 during the time that the channel is open
at different distances from the location of the source x = 0 , 20, 40, . . . , 120 nm. 共d兲, 共e兲 Results of the simulation of the simplified model 共solid
lines兲 and the full reaction diffusion model 共dotted lines兲, after the source located at rs共1兲 = 共0 , 0 , 0兲 turns off in regions of the size of a cluster
共see Sec. III兲. According to this model, after the channel closes, 关Ca2+兴 collapses linearly to a common value regardless of the value of x 共in
a time tcollapse = 0.1 ms兲 and then it follows a spatially homogeneous decay.

tion in the presence of the open source Cst共x兲 by observed that regardless of these other parameters, Cst共x兲 can
关Ca2+兴共x , 0 , 0 , 10 ms兲. The shape of the 关Ca2+兴 distribution always be approximated by an expression of the form 共8兲 共at
depends on the concentrations of buffers in the cytosol. We least for 关EGTA兴 and 关BAPTA兴 within the range of 0
repeated the simulation described above with other param- − 300 ␮M兲. Equations 共7兲 and 共8兲, on the other hand, imply
eters 共关EGTA兴 , 关BAPTA兴 = 0 , 100, 200, 300 ␮M兲 and we that 兺Ai + 关Ca2+兴basal = 关Ca2+兴共0 , 0 , 0 , 10 ms兲. Other fitting pa-
found that the approximation of Eq. 共7兲 holds even better for rameter values 兵Ai , ␦i其 are obtained for other values of the
other concentrations. Indeed, 关EGTA兴 is such a slow buffer single channel current or for other types or amounts of ex-
共see Table I兲 that for the concentration values we used, the ogenous buffers. The stationary solution near an open chan-
关Ca2+兴 does not change significantly. On the other hand, with nel has been extensively studied in the literature. Using dif-
increasing concentrations of BAPTA the stationary 关Ca2+兴 is ferent approximations, researchers have found analytical
achieved in less than 3 ms, due to the fast kinetics of the expressions for the 关Ca2+兴 near an open channel 关21,22兴.
These expressions may be used instead of Eq. 共8兲 if the cor-
buffer. Changing the diffusion coefficients of the exogenous
responding approximations used to derive the analytical ex-
buffers does not affect the quality of the approximation ei- pressions are valid. Here we choose to use a fitting function
ther. In particular, we have run simulations using values for that can be used, in principle, beyond the range of validity of
these coefficients between 50 and 200 ␮m2 s−1 and did not the various analytic approximations that can be found in the
observe any appreciable changes in 关Ca2+兴. literature.
We will then use the stationary solution Cst共x兲 to approxi- We now analyze what happens after the Ca2+ channel
mate the contribution of each open channel to the 关Ca2+兴 closes. We show in Figs. 2共d兲 and 2共e兲 plots of 关Ca2+兴共x , t兲 as
distribution within the cluster region. In order to expedite the a function of time for various values of x. It can be observed
subsequent computations, we approximate Cst共x兲 as that soon after the channel closes 共⬃0.1 ms, a time we call
4 tcollapse兲, 关Ca2+兴 becomes almost spatially homogeneous, es-
Cst共x兲 ⬇ Con共x兲 + 关Ca 兴basal ⬅ 兺 Aie−x/␦i + 关Ca2+兴basal .
2+ pecially for x 艋 60 nm 关see Fig. 2共d兲兴. For the purpose of
i=1 building the simplified model 共Sec. III兲, we will assume that
a time tcollapse after the channel closes, 关Ca2+兴 becomes ho-
mogeneous. The differences in 关Ca2+兴共x , t = tend + tcollapse兲 are
The fitting function 共8兲 has the purpose of having an analytic smaller than 1 ␮M for x ⬍ 60 nm 共the relative error is
expression that approximates Cst共x兲. It depends on the prop- ⬃18%兲. We approximate the fast decay of 关Ca2+兴 between
erties of the system that is simulated. However, we have tend and tend + tcollapse by a linear function of t. After tcollapse,

SOLOVEY et al. PHYSICAL REVIEW E 78, 041915 共2008兲

关Ca2+兴 follows a slower decay leading to its basal level on ters 关5兴 composed of a few tens of channels within a region
the order of tens of milliseconds 关see Fig. 2共e兲兴. An analytic of only a few hundreds of nm 关2–4兴. In this section we use
expression of this decay can be found by fitting the approximate solutions described in Sec. III A to build a
关Ca2+兴共x , 0 , 0 , t兲 共within the cluster region and for t ⬎ tend simplified model of Ca2+ dynamics within a cluster of IP3
+ tcollapse兲 with the following function that depends only on receptors. We emphasize that given the large gradients of
time 共in practice we use x = x* = 160 nm兲: cytosolic 关Ca2+兴 that build up close to an open channel, we
will not consider that all the channels in the cluster sense the
关Ca2+兴共x,0,0,t兲 ⬇ Cdecay共t兲 + 关Ca2+兴basal same 关Ca2+兴.

冉 冊
3 −␣i We consider that the cluster is composed of Ns channels
t − 共tend + tcollapse兲
⬅ 兺 Bi 1 + distributed around r = 0 at specified locations rs共i兲
i=1 ␶i = 共xsi , y si , 0兲, i = 1 , . . , Ns 关see Fig. 1共b兲兴. The number and the
distribution of channels within the cluster are based on ex-
+ 关Ca2+兴basal, t ⬎ tend + tcollapse . 共9兲
perimental observations and previous theoretical results
We show in Fig. 2共e兲 Cdecay + 关Ca 兴basal and 2+ 关2–4兴. The channels can be open or closed, and while they
关Ca2+兴共x , 0 , 0 , t兲, for different values of x, as functions of are open they release a constant current Ich 共Table I兲. The
time. We can observe the quality of the fit. Again, the pur- model has two main hypothesis. One is to assume that
pose of the function Cdecay is to have an analytic expression 关Ca2+兴共r , t兲 within the cluster region can be approximated by
of 关Ca2+兴 during the decay so as to make all subsequent a sum of the contributions of the individual channels, so that
computations much faster, as it will be clear in Sec. III. The Ns
兺i=1 Bi + 关Ca2+兴basal 关Ca2+兴共r,t兲 ⬇ 兺 C共Ri,t兲 + 关Ca2+兴basal ,
coefficients Bi must satisfy
= 关Ca 兴共x , tend + tcollapse兲. A similar fitting function but with
2+ * i=1
other parameter values can be used in the presence of other
for r within the cluster region 共10兲
types or amounts of buffers and for other values of the single
channel current. The goodness of the fit is almost perfect for where Ri = 兩r − rs共i兲兩 and C共Ri , t兲 is the contribution to 关Ca2+兴,
x* = 160 nm. For other values of x ⬍ 160 nm, the relative er- at r, of the ith channel located at rs共i兲. The second hypoth-
ror is a rapidly decreasing function, being lower than 25% esis is that C共Ri , t兲 can be approximated by a function that
for t − tcollapse ⬎ 5 ms. only depends on the state of the channel located at rs共i兲. We
distinguish four different conditions: the channel has never
opened 共i兲, the channel is open 共ii兲, a time less than tcollapse
B. Determining the [Ca2+] distribution
has passed since the channel has last closed 共iii兲 or a time
within a cluster of IP3R’s
larger than tcollapse has passed since the channel has last
Experimental observations of local Ca2+ events in Xeno- closed 共iv兲. Given the results of Sec. III A, we define C共Ri , t兲
pus laevis oocyte indicate that IP3R’s are organized in clus- as follows:

冦 冧
0 if the channel has never opened 共i兲,

Con共Ri兲 if the channel is open 共ii兲,

C共Ri,t兲 = Ccollapse − Cst共Ri兲 共11兲
Con共Ri兲 + 共t − tend兲 if tend ⬍ t 艋 tend + tcollapse, 共iii兲,

Cdecay共t兲 if t ⬎ tend + tcollapse . 共iv兲.

Also in this case, we either simulate the channel transi- etc. All parameters are as described in Table I, using
tions among states using the De Young and Keizer kinetic 关EGTA兴 = 300 ␮M as the exogenous buffer other than the
model 关17兴 or we fix the opening and closing times a priori. dye. We explain later how we compare both models when the
The purpose of the latter is to analyze the quality of the channel kinetics is included in the simulation.
approximations. For simplicity, we first consider a cluster of three channels
We now compare the distribution of free Ca2+ concentra- 共Ns = 3兲 located at rs共1兲 = 共−120, 0 , 0兲, rs共2兲 = 共0 , 0 , 0兲, rs共3兲
tion that we obtain with the method just described and the = 共120, 0 , 0兲. Channels 1 and 2 open from t = 0 s to t
one we obtain by solving the full reaction-diffusion model, = 0.01 s while channel 3 opens at t = 0.005 s and remains
with exactly the same conditions, i.e., the same number and open for 10 ms. We show in Fig. 3共a兲 the time evolution of
location of the channels within the cluster, buffer parameters, 关Ca2+兴 at r = 0 and in the inset we show the spatial distribu-


50 60 25
(a) (b) (c)
50 90 1.5 ms
40 40 1 ms 20 40 1.5 ms
20 30 20
[Ca ] (µM)

[Ca2+] (µM)

[Ca2+] (µM)
0 90
0 4.5 ms 0
30 40 5.5 ms 60 15 40 4.5 ms
20 30 30 20

0 90
0 8.5 ms 0
20 40 10.5 ms 60 10 40 8.5 ms
20 20 30 20
0 0 0
10 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 5 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
x (µm) x (µm) x (µm)
0 0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
Time (s) Time (s) Time (s)

FIG. 3. 共a兲: Ns = 3 channels are located at the membrane of the ER at points rs共1兲 = 关共−120 nm, 0 , 0兲 , rs共2兲 = 共0 , 0 , 0兲 ,
and rs共3兲 = 共120 nm, 0 , 0兲兴. The channels remain open during 10 ms 共channels 1 and 2 open at t = 0 and channel 3 at t = 0.005 s兲. We show
关Ca2+兴共0 , 0 , 0 , t兲 calculated with the full reaction diffusion model 共solid line兲 and with the SM 共dotted line兲 and in the insets the spatial
distributions of free 关Ca2+兴 along a line 共x , 0 , 0兲 at times, 1, 5.5, and 10.5 ms, respectively. 共b兲 Same as 共a兲 but for a cluster of Ns = 20
channels, located at positions described in Fig. 1共b兲 and with open and closed times as in Fig. 1共c兲. 共c兲 Same as 共b兲 but for a cytosol with
BAPTA at 300 ␮M instead of EGTA. In all cases, solid lines plots correspond to results with the full reaction diffusion model and dotted
lines with the simplified model.

tion of 关Ca2+兴 along a line at the surface of the ER, for three needed to run the full model for one open source and find
different times. In all cases we show the results obtained good fitting functions Cdecay共t兲 and Con共x兲 as explained in
solving the full reaction-diffusion model 共solid line兲 and the Sec. III A 共which is done just one time for each set of con-
simplified model 共dotted line兲. At r = 0 the relative differ- ditions of the system and takes ⬇10 min to simulate 1 ms of
ences were most of the time smaller than 10%. except for model time.
brief periods 共shorter than 1 ms when a channel opens or The error of the simplified model cannot be quantified
closes兲. simply by calculating the difference between the 关Ca2+兴 cal-
We then consider Ns = 20 channels in a cluster of 300 culated with it and with the full reaction-diffusion model.
⫻ 300 nm around r = 0 as in Fig. 1共b兲. The spatial distribu- The behavior of IP3R’s is highly nonlinear. A small differ-
tion and the opening and closing times of each channel is ence in 关Ca2+兴 may produce large differences in the behavior
chosen arbitrarily and are fixed before starting the simula- of the IP3R’s in the cluster, while for other conditions the
tion. The location of the channels and the time they open and channel could be insensitive to changes in 关Ca2+兴. Since we
close are summarized in Figs. 1共b兲 and 1共c兲. With these pa- are interested in the behavior of the IP3R’s in the cluster we
rameters we solve the simplified and the full reaction- have to determine whether or not the difference in 关Ca2+兴
diffusion models. We show in Fig. 3共b兲 the time evolution of affects the stochastic behavior of single IP3R’s.
关Ca2+兴 at 共0,0,0兲 and in the inset the spatial distribution of In order to study to what extent the simplified model is
关Ca2+兴 along a line at the surface of the ER, at three different able to reproduce the statistics of the openings and closings
times. Again, we simulated a cytosol with another buffer in of the individual channels of the cluster we solved the sim-
order to test the approximation. In Fig. 3共c兲 we show the plified and full models in different situations. First, we simu-
results of 关Ca2+兴 simulated in a cytosol with BAPTA at a late a case with two channels, one that was located at r = 0
total concentration of 300 ␮M. We see that the simplified which released a constant Ca2+ current of 0.1 pA for 10 ms
model gives a 关Ca2+兴 very similar to the solution of the full and the other one 共test channel兲, located at a distance x from
reaction diffusion model. The absolute error of the simplified the first one, that was initially closed and which followed the
model is lower than 5 ␮M for a cytosol with EGTA and De Young and Keizer kinetics 关17兴. The initial state of the
lower than 2 ␮M for a cytosol with BAPTA. During the closed channel was chosen randomly according to its station-
period of Ca2+ release, the relative errors are most of the time ary probability with 关Ca2+兴 = 关Ca2+兴basal and 关IP3兴 = 10 ␮M.
lower than 30% in the case of a cytosol with EGTA and Ns By doing so, we are assuming that 关IP3兴 has been constant
= 20 channels, and much lower in the cytosol with the faster for a long time so that the channel can be assume to be in a
buffer 共BAPTA兲. After the first 10 ms the relative errors in- state given by the stationary distribution with the constant
creases because the decay in 关Ca2+兴 is much faster in the full 关IP3兴. Results showing the temporal and spatial distribution
reaction diffusion model than in the simplified model 共espe- of 关Ca2+兴 using the full reaction diffusion and the simplified
cially with EGTA兲. This is not a big problem since it happens models are shown in Fig. 2. We show in Fig. 4共a兲 the mean
after the puff ends, when almost all the channels are closed time for the second channel to open for the first time 共la-
and it is highly unlikely that they will reopen again due to a tency兲 as a function of the distance to the open channel, x.
low increase in 关Ca2+兴. The stochastic transitions were computed using the value of
Regarding the time it takes for the simulation to be com- 关Ca2+兴 at the mouth of the test channel. We can see that the
pleted, for the cluster of Ns = 20 channels the full reaction- difference between the expected latency period using the
diffusion model took ⬇13 min to simulate 1 ms of model 关Ca2+兴 calculated with the full reaction diffusion and simpli-
time, running on an Intel Xeon 2.66 GHz processor. On the fied models is negligible 关even though the differences in
other hand, the time employed by the simplified model was 关Ca2+兴 were not, for example for example for distances from
0.5 s for each ms of model time, in addition to the time the source as large as 0.5 ␮m, as shown in Fig. 2共a兲兴. On the

SOLOVEY et al. PHYSICAL REVIEW E 78, 041915 共2008兲

(b) (d)

[Ca ] (µM)

[Ca2+] (µM)
(a) 60 60
30 40 40

25 20 20
Latency (ms)

0 0
20 0 0.01 0.02 0.03 0.04 0.05 0 0.005 0.01 0.015
Time (s) Time (s)
(c) (e)


10 0.3 0.06
0.2 0.04
0.1 0.02
0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.003 0.006 0.009 0.012 0.015 0 0.003 0.006 0.009
x (µm) Latency (s) Open duration (s)

FIG. 4. Different results of IP3R responses in the presence of different 关Ca2+兴 distributions calculated using the full reaction diffusion and
the simplified models. 共a兲 A closed IP3R 共test channel兲 is located at a distance x from an open channel that releases Ca2+ at a constant rate
during 10 ms. The IP3R follows the DeYoung-Keizer stochastic kinetics, so its dynamics is regulated by the local 关Ca2+兴. We measure the
latency of the receptor 共time to open兲 when 关Ca2+兴 is obtained by a numerical solution of the full reaction diffusion model and when 关Ca2+兴
is calculated with the simplified model. For each distance x we did 1000 stochastic simulations. We plot the average latency of all these
simulations. The error bars are 冑N , where SD is the standard deviation of the latency and N the number of latencies considered 共solid lines
represents latency obtained with 关Ca2+兴 from the full reaction diffusion model and dotted lines with 关Ca2+兴 from the simplified model兲. 共b兲
关Ca2+兴 at 共40, −40, 0兲 nm calculated using the full reaction diffusion and the simplified models for the cluster of Ns = 20 channels. 共c兲 Latency
distribution of the test channel 共initially closed兲 located at 共40, −40, 0兲 nm.共d兲, 共e兲 In this case an open IP3R that follows DeYoung-Keizer
stochastic model is located at the location of an open channel within the cluster of Ns = 20 channels. Horizontal solid line in 共d兲 represents
the time during which the channels remains open in the simulation of the full reaction diffusion and the simplified models 共the stochastic
IP3R cannot have an open channel longer than that兲. Solid lines correspond to simulation with 关Ca2+兴 obtained with the full reaction diffusion
model and dotted line with the simplified model.

other hand, the latency increases with the distance between to the obvious constraint of being lower than the actual open
both channels 共x兲 which follows immediately from the fact duration of channel 2兲. Again we see that the two sets of data
that 关Ca2+兴 tends to 关Ca2+兴basal for large x 关see also Fig. 2共a兲兴. are consistent.
The computation described previously describes a situa-
tion in which there is only one open channel, for instance, IV. THE Ca2+ AND CaFluo SPATIOTEMPORAL
when the puff is starting. We then repeated the same type of DISTRIBUTIONS OUTSIDE THE CLUSTER REGION
computation but in the presence of Ns = 20 IP3R’s in the clus-
ter. We considered a region of 700 nm wide around the clus- We have shown that the simplified model reproduces the
ter and a grid of 20 nm in it. Therefore a total of 352 loca- 关Ca2+兴 distribution inside the cluster region with a maximum
tions are defined, 20 of which are occupied by channels of error of 5 ␮M for a cytosol with EGTA and 2 ␮M for a
the cluster. We placed the test channel 共initially closed兲 at a cytosol with BAPTA. The relative error stays below 20% for
free point of the grid 共where there was not already another a cluster of 20 open channels 共but is much slower for a
channel兲. Then we computed the stochastic transitions of the smaller cluster兲. Furthermore, the errors are small enough so
test IP3R using again the De Young-Keizer model 关17兴, until that the statistics of the openings and closings of the chan-
the channel opens 共latency兲, in the presence of the time de- nels in the cluster are reproduced correctly. In spite of this,
pendent 关Ca2+兴. We computed 200 stochastic latencies using the results are not readily comparable to the experimental
this procedure 共100 with 关Ca2+兴 from the full reaction diffu- observations, even for isolated events such as puffs. Namely,
sion and 100 with the simplified model兲. In Fig. 4共b兲 we optical experiments give information on the Ca2+-bound to
show the time evolution of 关Ca2+兴 at the position of the test dye distribution averaged over a region of size ⬃300 nm
channel, 共40, −40, 0兲 nm. In Fig. 4共c兲 we show the histo- ⫻ 300 nm⫻ 700 nm, which is larger than the one spanned by
grams of latencies of the two sets of data, using simplified or a typical cluster. The simplified model described before is
the full reaction diffusion models for 关Ca2+兴. We see that the not good enough to describe the 关Ca2+兴 distribution within
two sets of latencies are consistent. We repeated this proce- this larger region. Furthermore, if one is willing to model
dure locating the test channel in every free point of the grid global Ca2+ signals that propagate throughout the cell, it is
and we found that the sets of latencies are consistent in more necessary to describe the spatiotemporal distributions of
than 70–80 % of the cases. 关Ca2+兴 and Ca2+ bound to dye over much larger regions 共in-
Finally, we looked at the 关Ca2+兴 distribution at a point volving ⬃ mm distances兲. To this end, we simulate the full
where there was already an open channel and measure the reaction-diffusion model.
open duration of a test channel located at that point in space To obtain the distribution of 关Ca2+兴 and 关CaFluo兴 outside
and being exposed to the Ca2+ obtained with the simplified the cluster region we solve the full reaction-diffusion system
and full reaction diffusion models. We use, again, the De described in Sec. II but using a “coarse” spatial grid 共that can
Young-Keizer model for the channel kinetics. We took the be larger than a typical cluster size兲. In this section we ana-
channel that remains open for the longest time 关see Fig. lyze how coarse the grid may be and still give the 关Ca2+兴 and
1共c兲兴. In Fig. 4共d兲 we show the time evolution of 关Ca2+兴 in 关CaFluo兴 distributions with relatively low errors. We also
both cases and in Fig. 4共e兲 the open time distribution 共limited analyze to what extent the clusters can be replaced by point


18 30
(a) (b)

[CaFluo]CG (µM)
[Ca2+]CG (µM)

12 20
6 10
0 0
0 0.005 0.01 0.015 0.02 0 0.005 0.01 0.015 0.02
Time (s) Time (s)
3.5 12
(c) (d)
3 10

[CaFluo]CG (µM)
[Ca2+]CG (µM)


0.5 2

0 0
0 0.005 0.01 0.015 0.02 0 0.005 0.01 0.015 0.02
Time (s) Time (s)

FIG. 5. Comparison of the time evolution of 关Ca2+兴 and 关CaFluo兴 coarse grained over a region of 共300 nm兲3 around r1 = 0 and around
r2 = 共300, 0 , 0兲 using two discretizations to solve the full reaction diffusion model 关dx = dy = dz = 20 nm 共solid lines兲 and dx = dy = dz
= 100 nm 共dotted lines兲兴. 共a兲, 共b兲 with EGTA at a total concentration of 300 ␮M and 共c兲, 共d兲 with BAPTA also at 300 ␮M of total
concentration. In each plot, the two curves with higher concentration values correspond to r1 and the lower ones to r2

sources for which the total Ca2+ current is the one deter-
mined using the simplified cluster model, regardless of the
number of channels that are open at any given time. To test
关CaFluo兴CG共r = 0,t兲 = 冕
关CaFluo兴共r,t兲dV, 共13兲

the accuracy of the approximation, we compare the solutions

obtained using the full reaction diffusion model with differ- where ⌬V is a volume of 共300 nm兲3 defined as −⌬x / 2 艋 x
ent discretizations: some in which each channel of the cluster 艋 ⌬x / 2, −⌬y / 2 艋 y 艋 ⌬y / 2, and 0 艋 z 艋 ⌬z 共⌬x = ⌬y = ⌬z
is contained in a different grid point and others in which = 300 nm兲.
several channels of the cluster belong to the same grid point. We show in Fig. 5 the time evolution of 关Ca2+兴CG and
In the latter, the Ca2+ current is simply calculated as the sum 关CaFluo兴CG at r1 = 共0 , 0 , 0兲 and at r2 = 共300 nm, 0 , 0兲 obtained
of the currents that flow through each open channel at any for the case with Ns = 20 channels already described in Sec.
III and Figs. 1共b兲 and 1共c兲. Figures 5共a兲 and 5共b兲 are simu-
given time, i.e., in Eq. 共4兲, Jch = 共2F1␦S兲 兺Ich共i兲, where the sum
lations with EGTA and Figs. 5共c兲 and 5共d兲 with BAPTA. The
is over all channels that are located at the same grid point
two solutions displayed correspond to grid sizes: dx = dy
and ␦S = dxdy, with dx and dy the grid spacing. In order to
= dz = 20 nm 共solid lines兲 and dx = dy = dz = 100 nm 共dotted
perform the comparison, we coarse grain all the solutions so
lines兲. In each plot, the couple of curves with higher concen-
as to represent them with a spatial resolution that is similar to
tration values correspond to r1 while the other two corre-
the experimental one 关⬃共300 nm兲3兴. Namely, we compute
spond to r2. The differences are a little larger in r1 than in r2
but in all cases the absolute differences are not significant in
␮M. The relative differences in 关CaFluo兴CG共r = 0 , t兲 is always

below 6% 共at r1 and r2, either with EGTA or BAPTA兲. In the
关Ca2+兴CG共r = 0,t兲 = 关Ca2+兴共r,t兲dV, 共12兲 case of 关Ca2+兴CG共r , t兲 the relative differences are below 10%
⌬V in the case of a cytosol with EGTA and in the case of

SOLOVEY et al. PHYSICAL REVIEW E 78, 041915 共2008兲

BAPTA it is a little larger because the absolute difference is tions showed that the simple model takes ⬃25 ms for each
similar but the total 关Ca2+兴 is lower since it is mostly bound ms of model time, which represents a simulation ⬃60 times
to the fast buffer, BAPTA. faster than the full reaction diffusion model.


Now, the aim of the model is not just to reproduce the
The various approximations described so far can be com- dynamics of isolated signals. We also want to use it to de-
bined to generate the simulation of an isolated Ca2+ puff. The scribe global Ca2+ signals such as waves that propagate
steps involved are as follows throughout the cell as have been observed in various cell
共1兲 Define the components of the cytosol 共concentrations types 关14,39–44兴. These waves can travel in two modes: with
and reaction rates兲 and the single channel Ca2+ current 共here no deformation 共continuous propagation兲 or in a saltatory
given in Table I兲. This last condition can be relaxed and manner 共as a sequence of bursts兲. For instance, in immature
various Ca2+ currents can be considered. In such a case, the Xenopus oocytes the propagation is saltatory while the fer-
second step of this list has to be repeated to analyze the tilization wave in the mature oocyte is continuous. The “fire-
behavior of cytosolic 关Ca2+兴 for the various Ca2+ currents of diffuse-fire” model 关24兴 mentioned in Sec. I was able to ex-
interest. The case in which the evolution of cytosolic Ca2+ plain the transition between these two modes of propagation
needs to be coupled to the evolution of 关Ca2+兴 inside the ER without having a detailed description of the intracluster dy-
could be treated similarly. However, the variety of possible namics. The model introduced in this paper can be used to
Ca2+ currents makes the suggested model not very practical study this problem from a different perspective in which a
in this case. more detailed description of the cluster dynamics is included.
共2兲 Solve Eqs. 共2兲 with one point source at r = 0 that opens Recently, it has been shown that for different cell types, glo-
at t = 0 and remains open for a long time, tend 共tend = 10 ms bal Ca2+ oscillations are stochastic, i.e., the result of random
seemed to be enough in the cases we tried兲 and then becomes occurring Ca2+ spikes 关45兴. Furthermore, the analysis shows
closed again. While the source is on, the cytosolic 关Ca2+兴 in that waves initiate randomly 共not as the result of a determin-
the vicinity of the open channel approximately reaches a sta- istic process兲 关45兴. Our approach of modeling Ca2+ waves
tionary solution. using a stochastic model for the IP3R as the elementary unit
共3兲 Find 共for each current of interest兲 a good fit of the is particularly suited to study this type of processes. In this
stationary spatial distribution of 关Ca2+兴, Cst共x兲 section we show how the model can be generalized to simu-
⬅ 关Ca2+兴共x , 0 , 0 , tend兲 while the channel is open and of its late the dynamics of intracellular Ca2+ in the presence of
decay after the channel closes, for example using a linear several clusters of IP3R’s. We also present some simulations
function for a fast decay and a spatial independent function that reproduce the propagation of saltatory waves with a ve-
关like Cdecay共t兲 + 关Ca2+兴basal兴, for subsequent times. locity that is in agreement with experimental results in Xe-
共4兲 Define the number Ns and location rs of the channels nopus laevis oocytes 关14,42兴.
in the cluster and choose a kinetic model to describe their In the case of considering several clusters we can proceed
states and state transitions. in the following way. We first choose the location and size of
共5兲 Start your simulation with all channels in a closed the clusters and the distribution of IP3R’s inside each of
state and Ca2+ and other components in equilibrium, with them. Let us call dx, dy, dz the grid spacings within the
关Ca2+兴 at its basal level. Run the kinetic model for the chan- cluster region and ⌬x, ⌬y, and ⌬z the grid spacings outside.
nels of the cluster under these conditions. As soon as one Let us consider square clusters of sides of size ⌬x = Ndx and
channel opens, compute 关Ca2+兴 within the cluster using Eqs. ⌬y = Ndy. Only one channel is contained in a dx ⫻ dy region,
共10兲 and 共11兲. Continue the simulation until all channels are while the whole cluster is contained inside the ⌬x ⫻ ⌬y re-
closed again. gion. Again we start from a situation in which all channels
共6兲 The previous step provides information on the time are closed and Ca2+ is uniformly distributed and in equilib-
evolution of the total Ca2+ current released at the cluster of rium with all other species at its basal level. Depending on
interest. Solve Eqs. 共2兲 using this current and a coarse spatial the situation that we want to model, we can either run the
grid so that the cluster is fully contained within one grid kinetic IP3R model, use probability arguments to decide
point. If the time course of the current is known with a finer when the first channel opens or we can put one randomly
temporal resolution than the time step with which the full chosen channel in an open state at t = 0. Once the first chan-
reaction-diffusion model is updated, then, the current needs nel opens we compute separately the 关Ca2+兴 distribution
to be coarse grained in time. If necessary, the Ca2+-bound to within the clusters 共using the dx, dy, dz grid spacings兲 and in
dye distribution obtained with the simulation 关CaFluo兴, can the region outside 共which is resolved up to ⌬x, ⌬y, and ⌬z兲.
be coarse-grained even further to compare with experiments. Let us assume that the channel kinetics within the cluster is
Steps 2 and 3 may be avoided if one has an accurate followed with a time step dt and that the reaction-diffusion
theoretical analytic expression for the stationary 关Ca2+兴 pro- model outside the cluster regions is updated with a larger
file close to an open channel such as the ones reported in time step ⌬t. For simplicity, we take ␯ ⬅ ⌬t / dt to be integer.
Refs. 关21,22兴 and for the temporal decay of 关Ca2+兴 after the Let us call 关Ca2+兴LR the Ca2+ concentration computed with
channel closes. Regarding the time it takes to obtain the fluo- the large grid spacing and large time step 共⌬x, ⌬y, ⌬z and
rescence signal and 关Ca2+兴 during a puff event, our simula- ⌬t, respectively兲 and 关Ca2+兴HR, the high resolution 共dx, dy,


5 (a) (e)

[Ca2+] (µM)
3 (c) 0.5 s (d) 0.5 s 100
y (µm)

1 10
-1 1
-3 0.1
-5 0.001
-20 -15 -10 -5 0 5 10 15 20 -20 -18 -16 -14 -12 -10 -8 -6

10 µm

10 µm
x (µm) x (µm)

(b) (f)

[Ca2+] (µM)
y (µm)

0.15 100
-0.15 1
1 µM 1 µM 0.01

0 µM 0 µM -14 -12 -10 -8 -6 -4 -2 0
x (µm) x (µm)

FIG. 6. Results of the simulation of global Ca2+ signals in the presence of several cluster of channels. 共a兲 A sketch of the cytosolic space
where we simulate Ca2+ dynamics. Note that all clusters are located at the line 共x , 0兲 and separated by 1.2 ␮m. 共b兲 A more detailed look at
the channels and clusters distribution. It is also showed the grid we use for the low spatial resolution step of simulations 共⌬x = 300 ␮m兲. 共c兲,
共d兲 Typical space time 共linescan兲 images of 关Ca2+兴 in the plane 共t , x兲. Brighter regions means elevated 关Ca2+兴. Time advances from left to
right. 共c兲 corresponds to the simplified model and 共d兲 to the full reaction diffusion mode. White vertical line correspond to the time of IP3
release. We can see several traveling waves 共e兲, 共f兲: Four snapshots of the spatial distribution of 关Ca2+兴 showing a saltatory wave, at four
different times: 共e兲 Simplified model: t = 322 ms 共⫹兲, t = 428 ms 共⫻兲, t = 527 ms 共black square兲 and t = 572 ms 共open circle兲. 共f兲 Full reaction
diffusion model: t = 7 ms 共⫹兲, t = 48 ms 共⫻兲, t = 101 ms 共black square兲, t = 145 ms 共open circle兲. Each snapshot corresponds to the time at
which one cluster is firing. Vertical lines shows the location the clusters 共solid if the cluster fired and dashed if the cluster did not fire in any
of those four times兲. Note that at each time there is only one cluster firing.

dz, and dt兲 concentration computed within one particular ing 关Ca2+兴HR inside the cluster at the same time using Eq.
cluster. Let us assume that we have already updated 关Ca2+兴LR 共11兲 and a modified version of Eq. 共10兲 of the form
at time t = n⌬t and that we want to move one 共large兲 time step
with the simulation. Let us describe how we would proceed
with one particular cluster between n⌬t and 共n + 1兲⌬t 共all 关Ca 兴HR共r,t兲 = 兺 C共Ri,t兲 + 关Ca2+兴0共n⌬t兲n⌬t
clusters should be updated “simultaneously,” following the
same approach兲. This implies computing the intracluster dy- + dt 艋 t 艋 n⌬t + ␯dt, 共14兲
namics during ␯ time steps of size dt. To this end, we run the
channel kinetic model for each channel of the cluster updat- with

关Ca2+兴0共n⌬t兲 =
冦 关Ca2+兴LR共n⌬t兲 − 冋冓 Ns

兺 C共Ri,n⌬t兲
冔册 if 关Ca2+兴LR共n⌬t兲 − 冋冓 兺Ns

C共Ri,n⌬t兲 冔册 艌 关Ca2+兴basal ,
冧 共15兲
关Ca 兴basal 2+

where 具¯典 is an average over all grid points inside the clus- located at the same 共coarse兲 grid point and ␦S = ⌬x⌬y. Given
ter. The term 关Ca2+兴0共n⌬t兲 is the average 关Ca2+兴 within the that we do not simulate Ca2+ diffusion inside the cluster but
cluster minus the contribution to this average due to the use some analytic approximation, it is the transition rates
channels of the cluster itself. 关Ca2+兴0 is responsible for the among the channel states that determine the allowed values
possible coupling between clusters. Given that 关Ca2+兴0 rep- of dt. Thus, we do not need to use dt ⬍ ⌬t and in all the
resents the contribution of other clusters to the 关Ca2+兴, we simulations that we present we have used dt = ⌬t, i.e., ␯ = 1.
consider its variation only on the low resolution scale. How- We performed several simulations using the procedure
ever, as explained later, there was no need to consider dt stated above. We simulated global Ca2+ signals generated by
⬍ ⌬t and, all the results that we show have been obtained for a set of Nc clusters arranged in a 1D “chain” within a two-
dt = ⌬t. dimensional cytosolic space. In Fig. 6共a兲 we show the com-
The computation 共14兲 is repeated ␯ dt time steps. At the plete cytosolic space and the clusters. The number of chan-
end, we compute the total current that flew through each nels of each cluster were chosen randomly from a Poisson
cluster between n⌬t and 共n + 1兲⌬t as before, by simply add- distribution with a mean of 25 IP3R’s 关4兴. The channel loca-
ing the current that flew through each open channel at any tions within the clusters were also picked at random using a
given time. We are then in the position of advancing one ⌬t uniform distribution over a circle of radius 150 nm. Configu-
time step the full reaction-diffusion system Eqs. 共2兲 with rations that gave interchannel distances that were smaller
Jch = 共2F1␦S兲 兺Ich共i兲, where the sum is over all channels that are than 40 nm were discarded. We set ⌬x = ⌬y = 300 nm, so as

SOLOVEY et al. PHYSICAL REVIEW E 78, 041915 共2008兲

TABLE II. Parameters used in the simulation of global Ca2+ the full reaction diffusion model 共wave traveling from right
signals such as the waves shown in Fig. 6. The rest of the param- to left兲.
eters are the same as Table I. Again, the improvement in computer time using the sim-
plified model was considerable. The simplified model takes
Parameter value Parameter value ⬃15s per ms of model time while the full reaction diffusion
model takes ⬃6 min to simulate the same model time 共all
dx = dy = dz 15 nm Nc 26
values measures in simulations running on a Intel Xeon
⌬x = ⌬y 300 nm d 1.2 ␮m 2.66 GHz processor兲.
dt = ⌬t 5 ␮s Ns Poisson 共25兲
关Ca兴basal 0.01 ␮M 关EGTA兴T 0

The development of microscopy and fluorescent indica-

tors have made it possible to measure localized Ca2+ signals
to ensure that each cluster corresponded to a single grid point that result from the activity of clustered channels in the
and the intercluster distance is 1.2 ␮m 关see Fig. 6共b兲 for a membrane of the endoplasmic reticulum. On one hand, these
closer look at the clusters and channel distribution where the optical techniques have the advantage over electrophysi-
grid lines correspond to the lattice that we used to solve the ological single channel recordings that they are less invasive
model in low resolution兴. To calculate 关Ca2+兴HR we used dx and provide spatial information. On the other hand, due to
= dy = ⌬20x = 15 nm. The simulations were done with 关EGTA兴 their spatial and temporal resolution, they are still unable to
= 0. The initial condition was such that 关IP3兴 = 0. The chan- resolve in detail the activity of channels within a cluster or to
nels were modeled using DeYoung-Keizer model 关17兴 and allow the study of the kinetics of single IP3R’s as in patch-
their initial state was chosen randomly using the stationary clamp recordings. Therefore, the use of mathematical models
probability distribution with 关Ca2+兴 = 关Ca兴basal and 关IP3兴 = 0 is a useful tool to gain insight and to understand local and
共see the Appendix for more details兲. 关IP3兴 was stepped uni- global Ca2+ signals such as puffs and waves in terms of the
formly in space from 0 ␮M to 关IP3兴stim = 1 ␮M at the time activity of their individual units 共i.e., the IP3R-channels兲.
indicated by the white horizontal line. The parameters used One of the main difficulties of modeling Ca2+ dynamics
in these simulations are summarized in Table II. with discrete source terms resides on the large range of tem-
We first did a simulation with 关Ca2+兴basal = 0.1 ␮M and poral and spatial scales that are involved. Stochastic single
found that very soon after the IP3 was released all the chan- channel transitions occur on time scales of ms while the con-
nels opened 共in a time much shorter than the intercluster certing activity of channels produce Ca2+ signals that may
diffusion time兲. In fact, the latency of the IP3R’s is very short last for seconds. At the same time, Ca2+ release from indi-
after such a step in 关IP3兴 for the De Young and Keizer model. vidual channels produce steep concentration gradients in the
Therefore, we decreased 关Ca2+兴basal in order to enlarge the vicinity of the open channels 共over a ⬃100 nm scale兲 while
latency and allow intercluster “communication” due to Ca2+ the summation of these releases can give rise to Ca2+ waves
diffusion. that travel throughout the entire cell 共1 − 2 ⬃ mm in the case
In Figs. 6共c兲 and 6共d兲, we show linescan images of 关Ca2+兴. of the Xenopus laevis oocytes兲. In this work we introduce a
The vertical axis represents the position along the line y = 0, model of cytosolic Ca2+ dynamics in the presence of one or
−20 ␮m 艋 x 艋 20 ␮m and the horizontal axis represents several clusters of Ca2+-release channels, each of which is
time. We can see that several saltatory waves are elicited. composed of discrete Ca2+ sources that behave stochastically.
The velocity of the waves are between 26 and 42 ␮sm . Figure The model is able to span several orders of magnitude in
6共c兲 corresponds to the simplified model and Fig. 6共d兲 is a space and time, ranging from the single IP3R activity to the
solution using the full reaction diffusion model. We can see generation of Ca2+ waves. We have tested the model by per-
that the results are quite similar. Indeed, in these case the forming extensive numerical simulations that show its accu-
resulting velocities are between 40 and 49 ␮sm . The velocity racy compared to standard methods. Therefore, we think it
of the waves obtained in simulations are in the range of gives a good framework to advance in the direction of un-
saltatory wave velocities found in experimental results in derstanding the local and global dynamics of Ca2+ signals in
Xenopus laevis 关14,42兴. a comprehensive way.
In Figs. 6共e兲 and 6共f兲 we show four snapshots of the spa- A key aspect of our model is its ability to link efficiently
tial distribution of 关Ca2+兴 along a segment of the line 共x , 0兲 at a relatively detailed spatial and temporal description of the
different times, corresponding to the moments at which four stochastic behavior of several IP3R’s in a cluster to the dy-
adjacent clusters fire 共at the peak of its Ca2+ release兲. The namics on a coarser scale in time and space. Previous works
short vertical lines at the horizontal axis indicate the position on modeling Ca2+ dynamics have used other strategies to
of the clusters 共solid lines are clusters with open channels at deal with the problem of the different time and spatial scales
one of the times shown and dotted lines are clusters that are involved. In general, this alternative strategies either simplify
closed at the four times shown兲. Note that only one cluster the diffusion description or the kinetics of single Ca2+ chan-
“fires” 共releases Ca2+兲 at a time 共saltatory propagation兲 and nels. A first simple approach is to focus on the temporal
that after the cluster fired, 关Ca2+兴 at its location decreases dynamics, i.e., to study temporal oscillations discarding dif-
back to its basal value. Figure 6共e兲 correspond to the simpli- fusion as done by De Young and Keizer in Ref. 关17兴. One
fied model 共wave traveling from left to right兲 and Fig. 6共f兲 to step further would be to consider cytosolic Ca2+ diffusion.


Sneyd et al. 关46兴 studied wave propagation in pancreatic and model of Ca2+ dynamics with a realistic ER geometry and
parotid acinal cells which result from Ca2+ release from IP3R where the gating of the IP3R’s was modeled stochastically
and ryanodine receptors 共RyR兲 distributed heterogeneously using a five-state representation. Nevertheless, it was not
through the cell. The authors included Ca2+ diffusion and used so far to study global Ca2+ signals such as waves.
described the state of the channels using rate equations de- Our model is built up upon a detailed analysis of the
rived from mass action kinetics that were coupled to an 关Ca2+兴 profile close to a channel when it is open and after it
equation governing the evolution of the Ca2+ concentration. closes. The profiles due to a single channel has been the
The distribution of the various receptors was described in subject of several works both analytical 关21兴 and numerical
terms of densities that were assumed to be nonhomogeneous 关22,23,35兴. Here we decided to solve numerically the com-
to reflect the nonuniform localization of IP3R’s and RyR’s in plete reaction-diffusion system for a particular cytosolic en-
the cell type of interest. Another approach presented in the vironment with one channel at the center of the cytosol.
literature was to model in more detail the stochastic nature of Finding good fits of both the stationary distribution and the
the channels but simplifying the diffusion of Ca2+ 关31,32兴. In temporal decay of 关Ca2+兴, the solution inside a cluster of
those works, the authors assumed that 关Ca2+兴 at one cluster several channels is approximated by a superposition of these
depended only on the number of nearest release sites where single channel solutions. We simulated a cluster of IP3R’s
there were open channels and that each neighboring 共active兲 with deterministic open and close times and found that the
site added a time-independent contribution to 关Ca2+兴. One model correctly describes the free 关Ca2+兴 within the cluster
useful simplified approach was to discard all detail within the of IP3R’s with an error lower than 30%. Furthermore, we
clusters 关24兴 and describe several aspects of global waves also showed that the errors in 关Ca2+兴 do not affect the statis-
considering the clusters as single excitable units. A much tical properties of the channels behavior. We did it by using a
detailed description was introduced by Falcke 关47,48兴 to specific single IP3R model and exposing it to the 关Ca2+兴 cal-
study intracellular Ca2+ dynamics in the presence of several culated using the model. We compared the statistics of the
clusters of IP3R’s. Each cluster was composed by a certain latencies and open times obtained in this way with those
number of channels that made stochastic state transitions. It obtained using a numerical solution of the complete reaction
was assumed in these works that the size of the cluster was diffusion system.
proportional to the number of open channels and within this The use of steady state solutions for the complete reaction
region, 关Ca2+兴 gradients were neglected, so that 关Ca2+兴 diffusion problem in the presence of a single channel has
within the cluster was considered homogeneous. This as- been used previously 关21,34,47兴. Nevertheless, as pointed
sumption was also used in other works 关25,30兴 and allowed out by Bentele and Falcke 关22兴 this approximation is rather
for faster simulations since then the whole cluster could be poor if used on the whole cell. Therefore, they propose the
considered as a single source. Additionally, in several in- use of a quasi-steady state approximation that avoids this
stances, only one cluster was considered, in which case the problem by using stationary solutions over short length
system is spherically symmetric. This approximation has scales close to the open channel. Our work resembles this
shown to be useful to describe several experimental observa- idea by using a stationary 关Ca2+兴 distribution for each open
tions. Nevertheless, it becomes inadequate if one is inter- channel within the scale of the cluster. Nevertheless, in this
ested in modeling the stochastic behavior of single IP3R’s in work we do not use analytical expressions for the 关Ca2+兴
the presence of Ca2+ feedback. Each channel behavior de- distribution near an open channel 关21,22,34兴 to avoid the
pends on the local Ca2+ concentration which is, as we re- limitations of the hypothesis that underlie analytic solutions.
ported in Sec. III A, subject to high 关Ca2+兴 gradients that We have shown that the model correctly describes the free
develop in the vicinity of an open IP3R, in agreement with Ca2+ concentration within a cluster of IP3R’s, something that
previous works 关18,21,23,33,34兴. If one is willing to include is necessary to connect the reaction-diffusion of Ca2+ with
a more accurate description of the Ca2+-release process ki- the single channel behavior. The model also provides a good
netics, which depends on the time at which each channel of approximation to the 关Ca2+兴 and 关CaFluo兴 distributions in a
the cluster becomes open, it is necessary to have a better large cytosolic region. The former is needed to model Ca2+
estimate of the 关Ca2+兴 value at each channel pore within the signals such as waves that depend on the “communication”
cluster. IP3R’s open and close depending on the local 关Ca2+兴. between clusters and the latter allows us to make the link
Therefore it is important to know 关Ca2+兴 close enough to the with experimental data. In a comparison with a traditional
channel. The model we present in this paper does not rely on 共full兲 reaction diffusion scheme, this model solves the prob-
the hypothesis of 关Ca2+兴 homogeneity within the cluster. It lem between 25–100 times faster 共depending on the param-
simplifies the description of the 关Ca2+兴 otherwise. eters and cytosolic space兲 at a negligible cost. This model
More recently, large scale computer simulations have opens the door for future work on modeling localized Ca2+
been used to model Ca2+ dynamics using finite elements signals such as “puffs” 共Ca2+ release from one cluster兲 in
methods 关49,50兴. This type of approach gives much more order to test hypotheses on IP3R’s. Regarding global Ca2+
detail but is computationally more expensive than ours. In signals, we have shown that the model reproduces saltatory
Ref. 关49兴 the authors studied Ca2+ waves in ventricular and wave propagation with velocities in the experimental range,
atrial myocites focusing on the influence of the RyR distri- using a relatively detailed description of the intracluster dy-
bution on these waves. One of the simplifications of the namics with a stochastic model of the IP3R as the elementary
model was to use a steady state response of the channel, unit. Nevertheless, this is only an example of the capabilities
which allows one to do faster simulations. On the other hand, of the method. In order to advance onto a more detailed
in Ref. 关50兴, the authors introduced a reaction-diffusion comparison with experimental data we would need a better

SOLOVEY et al. PHYSICAL REVIEW E 78, 041915 共2008兲

knowledge of various of the parameters that enter the model The model considers that the channel has 4 identical and
or replace the lack of knowledge with an exploration of the independent subunits. Each subunit has two binding sites for
behaviors it supports for different parameter values. In this Ca2+ 共one activatory and one inhibitory兲 and one site for IP3.
way, our approach could be used, for example, to explore in Therefore each subunit has eight possible states, usually de-
detail the results of Ref. 关45兴, where it has been shown that noted Sijk 共i , j , k = 0 , 1兲 where i = 1 indicates the presence of
Ca2+ oscillations occur stochastically and not as the result of an IP3 molecule bound to the channel, j = 1 a Ca2+ ion bound
a deterministic process. to the activatory site and k = 1 the binding of a Ca2+ ion to the
We think that the model we introduced in this work pro- inhibitory site. The open state of the subunit is S110 and a
vides a good framework that may lead to further work on channel is considered to be open when at least three of the
local and global Ca2+ signals using detailed single IP3R mod- subunits are in the open state. Transitions between states are
els as a building block. This kind of approach may contribute either constant or linearly dependent on 关Ca2+兴 or 关IP3兴 共we
to get a comprehensive understanding of Ca2+ signals over a use the original parameters published in Ref. 关17兴 and, unless
large range of time and spatial scales. In particular, the model explicitly noted, 关IP3兴 = 10 ␮M兲. In most simulations, 关IP3兴 is
could be used to analyze the behaviors that different IP3R constant, but 关Ca2+兴 changes in time due to the coupled
models would predict, contrasting them against experimental reaction-diffusion of Ca2+ released by nearby channels
data on Ca2+ puffs and waves, such as puff latencies or the within the cluster. As described in the paper, the 关Ca2+兴 dis-
dependence of the wave speed on 关IP3兴. tribution within the cluster is determined in two ways: we
either use the simplified model 共Sec. III兲 or the full reaction-
ACKNOWLEDGMENTS diffusion model 共Sec. II兲. In both cases, we simulate the
We acknowledge useful conversations with Luciana stochastic transitions between states of the channel using a
Bruno. D.F. and S.P.D. are members of the Carrera del In- discrete time Markov Chain procedure. Within the frame-
vestigador Científico 共CONICET兲. This research was sup- work of the simplified model, we use dt = 10−5 s. When
关Ca2+兴 within the cluster is simulated using the full reaction
ported by UBA 共Grant No. UBACyT X208兲, Grant No. PICT
diffusion model, the same time step is used to update the
17-21001 of ANPCyT 共Argentina兲, Santa Fe Institute, and
state of the channels.
CONICET 共Grant No. PIP 5131兲.
In Sec. VI we use a step of 关IP3兴 at t = 0 representing the
time at which IP3 is photoreleased. In that case, the states of
the IP3R are initially chosen at random using the stationary
probability distribution, i.e., the probability of being in each
We use the stochastic IP3R model introduced by De state at a given 关Ca2+兴 and 关IP3兴 = 0 which can be obtained
Young and Keizer 关17兴, with its original parameter values. analytically.

关1兴 M. J. Berridge, M. D. Bootman, and P. Lipp, Nature 共London兲 关15兴 P. C. da Fonseca, S. A. Morris, E. P. Nerou, C. W. Taylor, and
395, 645 共1998兲. E. P. Morris, Proc. Natl. Acad. Sci. U.S.A. 100, 3936 共2003兲.
关2兴 S. Swillens, G. Dupont, L. Combettes, and P. Champeil, Proc. 关16兴 I. Bezprozvanny, J. Watras, and B. E. Ehrlich, Nature 共Lon-
Natl. Acad. Sci. U.S.A. 96, 13750 共1999兲. don兲 351, 751 共1991兲.
关3兴 J. Shuai, H. J. Rose, and I. Parker, Biophys. J. 91, 4033 关17兴 G. W. De Young and J. Keizer, Proc. Natl. Acad. Sci. U.S.A.
共2006兲. 89, 9895 共1992兲.
关4兴 L. Bruno, G. Solovey, A. C. Ventura, S. Dargan, I. Parker, and 关18兴 S. Swillens, P. Champeil, L. Combettes, and G. Dupont, Cell
S. Ponce Dawson 共unpublished兲. Calcium 23, 291 共1998兲.
关5兴 Y. Yao, J. Choi, and I. Parker, J. Physiol. 共London兲 482, 533 关19兴 D. Fraiman and S. P. Dawson, Cell Calcium 35, 403 共2004兲.
共1995兲. 关20兴 J. K. Foskett, C. White, K. H. Cheung, and D. O. Mak,
关6兴 N. Callamaras and I. Parker, Cell Calcium 15, 66 共1994兲. Physiol. Rev. 87, 593 共2007兲.
关7兴 D. O. Mak and J. K. Foskett, J. Gen. Physiol. 109, 571 共1997兲. 关21兴 G. D. Smith, L. Dai, R. M. Miura, and A. Sherman, SIAM J.
关8兴 J. W. Shuai and P. Jung, Proc. Natl. Acad. Sci. U.S.A. 100, Appl. Math. 61, 1816 共2001兲.
506 共2003兲. 关22兴 K. Bentele and M. Falcke, Biophys. J. 93, 2597 共2007兲.
关9兴 I. Parker and Y. Yao, J. Physiol. 共London兲 491, 663 共1996兲. 关23兴 J. W. Shuai and I. Parker, Cell Calcium 37, 283 共2005兲.
关10兴 M. Bootman, E. Niggli, M. Berridge, and P. Lipp, J. Physiol. 关24兴 S. P. Dawson, J. Keizer, and J. E. Pearson, Proc. Natl. Acad.
共London兲 499, 307 共1997兲. Sci. U.S.A. 96, 6060 共1999兲.
关11兴 X. P. Sun, N. Callamaras, J. S. Marchant, and I. Parker, J. 关25兴 J. W. Shuai and P. Jung, Biophys. J. 83, 87 共2002兲.
Physiol. 共London兲 509, 67 共1998兲. 关26兴 H. DeRemigio and G. D. Smith, Cell Calcium 38, 73 共2005兲.
关12兴 I. Parker and Y. Yao, Proc. R. Soc. London, Ser. B 246, 269 关27兴 L. Diambra and N. Guisoni, Cell Calcium 37, 321 共2005兲.
共1991兲. 关28兴 G. Ullah and P. Jung, Biophys. J. 90, 3485 共2006兲.
关13兴 I. Parker, J. Choi, and Y. Yao, Cell Calcium 20, 105 共1996兲. 关29兴 J. Shuai and P. Jung, New J. Phys. 5, 132 共2003兲.
关14兴 N. Callamaras, J. S. Marchant, X. P. Sun, and I. Parker, J. 关30兴 R. Thul and M. Falcke, Biophys. J. 86, 2660 共2004兲.
Physiol. 共London兲 509, 81 共1998兲. 关31兴 M. Bar, M. Falcke, H. Levine, and L. S. Tsimring, Phys. Rev.


Lett. 84, 5664 共2000兲. 关42兴 J. Marchant, N. Callamaras, and I. Parker, EMBO J. 18, 5285
关32兴 M. Falcke, L. Tsimring, and H. Levine, Phys. Rev. E 62, 2636 共1999兲.
共2000兲. 关43兴 M. D. Bootman and M. J. Berridge, Curr. Biol. 6, 855 共1996兲.
关33兴 M. Naraghi and E. Neher, J. Neurosci. 17, 6961 共1997兲. 关44兴 H. Cheng, M. R. Lederer, W. J. Lederer, and M. B. Cannell,
关34兴 G. D. Smith, Biophys. J. 71, 3064 共1996兲. Am. J. Physiol. 270, C148 共1996兲.
关35兴 S. Rudiger, J. W. Shuai, W. Huisinga, C. Nagaiah, G. War-
关45兴 A. Skupin, H. Kettenmann, U. Winkler, M. Wartenberg, H.
necke, I. Parker, and M. Falcke, Biophys. J. 93, 1847 共2007兲.
Sauer, S. C. Tovey, C. W. Taylor, and M. Falcke, Biophys. J.
关36兴 H. J. Rose, S. Dargan, J. Shuai, and I. Parker, Biophys. J. 91,
94, 2404 共2008兲.
4024 共2006兲.
关37兴 S. L. Dargan, B. Schwaller, and I. Parker, J. Physiol. 共London兲 关46兴 J. Sneyd, K. Tsaneva-Atanasova, J. I. Bruce, S. V. Straub, D.
556, 447 共2004兲. R. Giovannucci, and D. I. Yule, Biophys. J. 85, 1392 共2003兲.
关38兴 S. L. Dargan and I. Parker, J. Physiol. 共London兲 553, 775 关47兴 M. Falcke, Biophys. J. 84, 28 共2003兲.
共2003兲. 关48兴 M. Falcke, Biophys. J. 84, 42 共2003兲.
关39兴 M. J. Berridge and G. Dupont, Curr. Opin. Cell Biol. 6, 267 关49兴 L. T. Izu, S. A. Means, J. N. Shadid, Y. Chen-Izu, and C. W.
共1994兲. Balke, Biophys. J. 91, 95 共2006兲.
关40兴 M. J. Berridge, J. Exp. Biol. 200, 315 共1997兲. 关50兴 S. Means, A. J. Smith, J. Shepherd, J. Shadid, J. Fowler, R. J.
关41兴 A. P. Thomas, G. S. Bird, G. Hajnoczky, L. D. Robb-Gaspers, Wojcikiewicz, T. Mazel, G. D. Smith, and B. S. Wilson, Bio-
and J. Putney, FASEB J. 10, 1505 共1996兲. phys. J. 91, 537 共2006兲.