Research ArticleCell Biology

Hindered cytoplasmic diffusion of inositol trisphosphate restricts its cellular range of action

See allHide authors and affiliations

Science Signaling  08 Nov 2016:
Vol. 9, Issue 453, pp. ra108
DOI: 10.1126/scisignal.aag1625

A new paradigm for IP3 signaling

Receptors that activate phospholipase C generate the second messenger inositol trisphosphate (IP3). IP3 stimulates calcium release from the endoplasmic and sarcoplasmic reticulum, thereby shaping calcium signals in cells. Dickinson et al. triggered the focal release of IP3 in animal cells and measured calcium “puffs”—intense, localized increases in calcium released from the endoplasmic reticulum. They found that IP3 diffuses much more slowly within cells than had been originally measured in vitro using oocyte cytoplasmic extracts. Thus, rather than functioning as a global cellular signal, IP3 can produce local signals, which increases the complexity of information that can be encoded by cells in response to stimuli that activate receptors that generate this second messenger.


The range of action of intracellular messengers is determined by their rates of diffusion and degradation. Previous measurements in oocyte cytoplasmic extracts indicated that the Ca2+-liberating second messenger inositol trisphosphate (IP3) diffuses with a coefficient (~280 μm2 s−1) similar to that in water, corresponding to a range of action of ~25 μm. Consequently, IP3 is generally considered a “global” cellular messenger. We reexamined this issue by measuring local IP3-evoked Ca2+ puffs to monitor IP3 diffusing from spot photorelease in neuroblastoma cells. Fitting these data by numerical simulations yielded a diffusion coefficient (≤10 μm2 s−1) about 30-fold slower than that previously reported. We propose that diffusion of IP3 in mammalian cells is hindered by binding to immobile, functionally inactive receptors that were diluted in oocyte extracts. The predicted range of action of IP3 (<5 μm) is thus smaller than the size of typical mammalian cells, indicating that IP3 should better be considered as a local rather than a global cellular messenger.


In a classic paper, Allbritton et al. (1) introduced the concept of the “range of action” of an intracellular messenger, specifically considering the distance from a source over which Ca2+ ions and inositol trisphosphate (IP3) can exert their actions. The range of action is determined by how far a messenger can diffuse, on average, before it is removed from the cytosol by degradation (IP3) or sequestration (Ca2+). To experimentally determine the diffusion rates, Allbritton et al. (1) prepared slabs of cytosolic extract from Xenopus oocytes and measured the penetration over time of radiolabeled Ca2+ and IP3 presented to one side of a slab after inhibiting degradation and sequestration mechanisms. The apparent diffusion coefficient that they obtained for IP3 (283 ± 53 μm2 s−1) was similar to that expected for free diffusion in a medium with twice the viscosity of water, whereas the value for Ca2+ was much lower (38 ± 11 μm2 s−1), concordant with hindered diffusion in the presence of immobile Ca2+-binding buffers. Taking into account the respective rates of degradation and sequestration of IP3 and Ca2+, Sims and Allbritton (2) concluded that Ca2+ has a narrow cytosolic range of action and serves as a local signal, whereas IP3 functions as a global signal because its range of action is greater than the dimensions of typical mammalian cells.

Here, we reconsider the issue of diffusion of IP3 in the cytosol in light of observations suggesting that binding to immobile sites may appreciably slow the effective diffusion of IP3 in mammalian cells. In SH-SY5Y neuroblastoma cells and other cell lines, IP3 primarily evokes Ca2+ release from only a few hundred functional IP3 receptors (IP3Rs), clustered at discrete sites (3, 4). Different from this, immunostaining of IP3Rs reveals a dense distribution throughout the cell, and radioligand assays indicate the presence of about 30,000 IP3R monomers (IP3 binding sites) per cell (58). If these functionally “silent” IP3Rs bind IP3, they would hinder its diffusion without contributing to the Ca2+ signal.

We thus set out to determine the effective diffusion coefficient of IP3 in intact mammalian cells. We uniformly loaded SH-SY5Y cells with a caged, poorly metabolized IP3 analog (i-IP3) and selectively photoreleased i-IP3 using flash photolysis by an ultraviolet (UV) laser spot positioned at one end of these elongated cells. This increased the cytosolic concentration of i-IP3 at the location of the laser spot, which subsequently equilibrated throughout the cell as free i-IP3 diffused from the site of origin. To monitor the spread of i-IP3, we used IP3-evoked Ca2+ liberation (Ca2+ puffs) from local clusters of IP3Rs (4, 9) as endogenous and sensitive detectors that are distributed throughout the cell. As previously reported (10), control experiments in which i-IP3 was uniformly photoreleased across a cell showed that the mean latency between photorelease and the occurrence of the first puff at a site shortened in about linear proportion with increasing concentration of i-IP3, whereas the latency did not vary systematically along the length of the cell. Thus, we could use the mean puff latency at sites at different distances from the laser spot as a measure of the temporal profile of the concentration of i-IP3 ([i-IP3]) at those locations. If i-IP3 were to diffuse sufficiently rapidly that its concentration neared equilibrium within the mean first-puff latency, then we would expect similar latencies at all locations. Instead, we observed that the mean first-puff latencies were about 20 times longer at the distal end of the cells as compared to the sites proximate to the UV spot. By comparing these data to model predictions, we derive an estimate for the effective diffusion coefficient of IP3 in the cytosol of 5 to 10 μm2 s−1, about 30-fold lower than the widely accepted value obtained by Allbritton et al. (1). We therefore propose that IP3 is better considered as a local, rather than a global, messenger in mammalian cells.


Puffs evoked by localized and distributed photorelease of i-IP3

We loaded SH-SY5Y cells with the Ca2+ indicator Fluo-4 and a caged precursor of i-IP3, an analog of IP3 that is slowly metabolized (11) and evokes Ca2+ puffs that persist in the cells after flash photorelease with little diminution in mean frequency for more than 2 min (3, 4, 12). During the time course of recordings (30 s), we assumed that degradation of i-IP3 was negligible so that the total amount that was photoreleased could be assumed to remain constant. Moreover, the latencies of puffs evoked by a given, spatially uniform photolysis flash showed only a small variation between different cells (SD, 40% of mean; 24 cells), indicating a relatively consistent loading of caged i-IP3. We also loaded the cells with the slow Ca2+ buffer EGTA to suppress global Ca2+ waves that could be triggered by the released Ca2+ (13). Localized transient Ca2+ puffs were evoked by photoreleasing i-IP3 using a focused laser spot of 405-nm light that was either stationary and positioned so that it illuminated within one end of a cell (spot flash) or was rapidly and uniformly stepped across the length of the cell (distributed flash) (Fig. 1A). In both cases, the intensity and total duration of the laser flash were identical, delivering the same number of photons and, hence, photoreleasing the same average amount of i-IP3. However, because of the difference in spatial profile of the photoreleased i-IP3, we expected that the spot flash stimulation would evoke puffs beginning, on average, after longer latencies at greater distances from the spot because of the time scale and dilution imposed if i-IP3 were to diffuse slowly, whereas the distributed flash stimulation would result in a rapid and near-homogeneous increase in [i-IP3] throughout the cell.

Fig. 1 Latencies of Ca2+ puffs evoked by localized and distributed laser spot photorelease of i-IP3.

(A) The top row of panels shows the location of a stationary laser spot flash (200-ms duration) delivered to an SH-SY5Y cell at time 0 (red spot, left panel), and image frames at different times after the flash showing the first puffs arising at select locations (circled). Fluorescence of Fluo-4 during puffs is depicted on an arbitrary gray scale. The bottom row of panels shows corresponding images in a different cell, where a laser flash of the same intensity and total duration was distributed as 10 individual spots spaced evenly along the length of the cell. Scale bar, 10 μm. (B) Map of the same cell as in the top row of (A), showing the locations of the puffs numbered in (A) as black squares, the locations of all other identified puff sites marked as gray squares, and the location of the stationary laser spot marked by the red circle. Scale bar, 10 μm. (C) Representative traces showing Ca2+-dependent fluorescence ratio changes (ΔF/F0) measured from small regions of interest (1.3 μm × 1.3 μm) centered on the puff sites numbered in (A) and (B). The gray arrow indicates the time of the spot flash, and the open arrows mark the times of the first puff at each site. (D) Scatter plots show measurements of first-puff latencies from all sites in the two cells illustrated in (A) as a function of the distance of that site from the stationary spot flash (top) or from the lower end of the cell in the case of the distributed flash (bottom). Latencies are plotted on a logarithmic scale to better encompass the wide range of values.

We mapped the locations of puffs evoked by photoreleased IP3 throughout a 30-s recording (Fig. 1B) and monitored the fluorescence ratio traces (ΔF/F0) from regions of interest centered on these sites (Fig. 1C). We measured the latency to the first puff at each site as a function of the distance of the site from the spot flash (Fig. 1D, top panel) or to the lower end of cells receiving distributed flash stimulation (Fig. 1D, bottom panel). Although there was considerable scatter in the data, these plots showed that puff latencies were longer at sites more distant from the stationary laser spot, whereas there was no obvious systematic variation of puff latency with the positions along the cell in the case of the distributed stimulation.

Determination of mean first-puff latencies

Puffs are stochastic events, and their latencies after spatially uniform photorelease of a given amount of IP3 are exponentially distributed (9, 10). To obtain estimates of mean latencies by pooling data from multiple puff sites in many cells, we automated the detection and localization of puffs using a freely available custom software algorithm (14). This algorithm reported the latency of each event from the beginning of the photolysis flash together with the distance from the centroid locations of a puff in relation to the spot flash or, arbitrarily, to one end of the cell in trials with distributed flash stimulation. Trials with spot flash and distributed flash stimulation were performed in alternating order, using a constant intensity and differing total flash durations (100, 200, and 500 ms).

Individual first-puff latencies evoked by spot flash of i-IP3 with flash durations of 100, 200, or 500 ms showed considerable variability (Fig. 2A), as expected, but mean latencies calculated by binning the data points over ~5-μm increments revealed a dependence on distance from the site of photorelease for all three flash durations (Fig. 2B). Mean puff latencies at sites near the laser spot were shorter for longer flash durations but lengthened to ~10 s for all flash durations at the end of the cell farthest from the spot flash stimulation site. In contrast, distributed flash stimulation of i-IP3 throughout the cell evoked puffs with latencies that lacked a systematic relationship with the location along the cell but, on average, shortened progressively with increasing flash duration (Fig. 2, C and D). To further validate the difference in spatial distribution of first-puff latencies between the spot flash and distributed photorelease of i-IP3, we performed a linear regression to fit a linear function of the distance along the cell to the logarithm of the individual first-puff latencies, as plotted in Fig. 2 (A and C). Latencies depended strongly on the distance from the spot flash stimulation for all flash durations, as manifested by slopes differing significantly from zero, whereas there was no significant distance dependence for distributed flash stimulation (fig. S1).

Fig. 2 First-puff latencies increase with increasing distance from spot flash photorelease of i-IP3 but show no systematic variation with position along the cell for distributed flash photorelease of i-IP3.

(A) Scatter plot showing first-puff latencies at individual puff sites after spot flash photorelease of i-IP3 as a function of distance from the laser spot after flashes of 100-ms (gray squares, eight cells, 52 puff sites), 200-ms (open circles, seven cells, 75 puff sites), and 500-ms duration (stars, nine cells, 137 puff sites). The y axis is plotted on a logarithmic scale to better display the wide variability in observed puff latencies. (B) Mean measurements of first-puff latency obtained by binning the data in (A) in ~5-μm increments. (C and D) Corresponding plots of individual (C) and mean binned puff latency data (D) after distributed flash photorelease of i-IP3. The spot flash intensity and total exposure times were the same as for the spot flash stimulation in (A) and (B). Data from the 100-, 200-, and 500-ms total exposure durations are from eight cells at 134 puff sites, eight cells at 76 sites, and seven cells at 152 sites, respectively. Error bars indicate ± SEM.

These experiments to determine puff latencies were performed with cells loaded with EGTA to suppress global Ca2+ waves. To verify that the stimulation conditions produced physiologically relevant IP3 concentrations, we photoreleased equivalent amounts of i-IP3 in cells that had not been loaded with EGTA (fig. S2). We observed global Ca2+ signals in 13 of 16 (81%), 16 of 21 (76%), and 19 of 19 (100%) cells stimulated with a distributed flash for 100, 200, and 500 ms, respectively. In parallel experiments in EGTA-loaded cells, the 200-ms flash evoked puffs with a mean latency of 6.3 s, closely matching that in the experiment shown in Fig. 2D. Therefore, the stimulus conditions were in a physiologically relevant range of IP3 concentrations.

The data shown for the SH-SY5Y cells were obtained using the same culture of cells loaded under the same conditions to ensure internal consistency. We observed similar results in two other experiments with SH-SY5Y cells and in one with COS-7 cells in which we measured 17 cells (fig. S3, A and B).

The overall mean latencies of puffs evoked by distributed flash stimulation decreased as the durations of the flash increased and, for each given flash duration, the distributions of puff latencies approximated exponential functions (Fig. 3A), consistent with previous findings (10). In contrast to the lack of spatial variation in puff latencies after distributed flash stimulation, the latency distributions of puffs evoked by a 500-ms spot flash stimulation varied with distance, with a time constant of ~1 s within <15 μm of the spot and ~10 s at sites >15 μm away (Fig. 3B).

Fig. 3 Distributions of first-puff latencies and dependence of mean first-puff latency on photolysis flash duration.

(A) Distributions of first-puff latencies evoked after distributed flash stimulation of i-IP3 with the indicated total flash durations. (B) Distributions of first-puff latencies after a 500-ms spot flash stimulation at sites within (left) and beyond (right) 15 μm of the laser spot. Data in (A) and (B) are fitted by single-exponential functions, with time constants τ as indicated. (C) Mean first-puff latency after distributed flash stimulation of i-IP3 plotted as a function of reciprocal flash duration. Error bars indicate ± SEM and are smaller than the symbol width for the leftmost point. The data are fitted by a regression line, constrained to the origin, with a slope of 0.90.

Given that the latency to the first occurrence of a puff at a site is an inverse function of [IP3] (10), we interpreted the increasing first-puff latency at greater distances from the spot flash stimulation to arise from the time needed for i-IP3 to diffuse along the length of the cell. To derive the diffusion rate of i-IP3 in the cytosol, we needed to know how the latency varies with flash duration (amount of photoreleased i-IP3) under our experimental conditions. To obtain this relationship, we plotted (Fig. 3C) the mean latencies of the first puffs evoked by distributed flash stimulation as a function of reciprocal flash duration, which is a linearly proportional measure of the amount of photoreleased i-IP3, assuming that only a small fraction of the total caged i-IP3 was released by photolysis (15). Concordant with previous findings (10), the data fit to a linear relationship, with the latency shortening in direct proportion to increasing [i-IP3]. Using the slope of the relationship (0.90 ± 0.05 in units relative to flash duration), we constructed a model of the predicted distribution of puff latencies as i-IP3 diffused through the cell.

Simulation of IP3 diffusion and puff triggering

To obtain a quantitative estimate of the effective diffusion coefficient of IP3, we compared our experimental data of first-puff latencies with predictions obtained by model simulations of the diffusion equation using various diffusion coefficients (methods S1). We used a simplified, closed-end, one-dimensional model, representing the average length of cells used in the experiments (40 μm; fig. S4A). In the model, we introduced IP3 at the spot flash stimulation site in relative amounts corresponding to the different flash durations, over a time course matching the flash duration and with a spatial distribution approximating that of the UV laser spot (fig. S4B). After the introduction of IP3, we assumed that the total amount remained constant, mimicking the slow degradation of i-IP3 in the cytosol (3, 4).

Heat maps of simulated spatiotemporal profiles of IP3 concentration resulting from a modeled 500-ms flash illustrate the different results obtained with two effective diffusion coefficients: 10 μm2 s−1 (Fig. 4A) and 280 μm2 s−1, the value obtained by Allbritton et al. (1) (Fig. 4B). From these simulation results, we plotted the change in IP3 concentrations over time at select distances along the modeled cells with the different diffusion constants (Fig. 4, C and D). When the diffusion coefficient was set to 280 μm2 s−1, the IP3 concentration equilibrated across the cell within about 1 s to a final normalized concentration of ~0.11. For a diffusion coefficient of 10 μm2 s−1, the local concentration of IP3 in the vicinity of the modeled spot flash remained considerably greater than the final equilibrium concentration for several seconds, whereas the concentration at the distal end of the cell did not begin to increase appreciably for more than 5 s and was still below the equilibrium level after as long as 20 s.

Fig. 4 Simulation of IP3 diffusion after spot flash stimulation.

Panels illustrate the simulated diffusion of IP3 in a 40-μm-long, one-dimensional model of a cell after spot (Gaussian profile with σ = 1.35 μm) flash stimulation at one end with a flash duration of 500 ms. After the flash, the total amount of IP3 was conserved through the cell (no degradation occurred). The diffusion coefficients used in the simulations are indicated at the top. (A and B) Heat maps representing IP3 concentrations at different times (y axis, with time running from top to bottom) and distances (x axis) after the beginning of the flash at time 0. IP3 concentrations are in arbitrary units but are consistent for both panels. (C and D) Temporal profiles of [IP3] at different distances (indicated in μm) along the modeled cell. Concentrations of IP3 in both panels are expressed normalized to the peak concentration attained at the photolysis spot site for D = 10 μm2 s−1.

We used these predicted spatiotemporal concentration profiles to compute the probability of observing a puff at each position along the length of the one-dimensional cell at any given time after the introduction of IP3 (onset of the spot flash; methods S1). For each spatial element and time point in the model, we calculated the puff-triggering rate that corresponded to the instantaneous concentration of IP3 predicted by the simulations in Fig. 4 using the slope value from Fig. 3C, in which a distributed flash photorelease of IP3 evoked first puffs at a rate (inverse mean latency) of 1.11 s−1 per second of flash duration. The heat maps of the predicted puff-triggering rates for IP3 diffusion coefficients of 10 and 280 μm2 s−1 showed different probability patterns (Fig. 5, A and B). From these simulation results, we derived the corresponding probabilities of observing an initial puff at select distances along the modeled cell as a function of time, taking into account the increasing chance that a puff would have already occurred during previous time points (Fig. 5, C and D). Finally, we computed how the first-puff latencies varied with distance from the spot where IP3 was introduced (the simulated spot flash stimulation) for a 500-ms spot flash duration and respective diffusion coefficients of 10 and 280 μm2 s−1 (Fig. 5, E and F, blue curves).

Fig. 5 Simulation of the probability of observing puffs at different times and distances after spot photorelease of IP3.

Simulations were based on the spatiotemporal concentration profiles in Fig. 4, assuming that the probability of puff triggering is a linear function of [IP3]. The diffusion coefficients used in the simulations are indicated at the top. (A and B) Heat maps representing the probability of observing a puff at different times (y axis, with time running from top to bottom) and distances along the cell (x axis) after the beginning of the flash at time 0. (C and D) Temporal profiles of puff probability at different distances (indicated in μm) along the cell. (E and F) Predicted mean latencies of puffs as a function of distance along the cell from the flash. Horizontal lines indicate the mean puff latency if the same total amount of IP3 were released uniformly along the cell by distributed flash stimulation.

In the case of the 280 μm2 s−1 diffusion coefficient, almost all first puffs were predicted to arise within 5 s of the introduction of IP3 (Fig. 5D), and there was only a small dependence of first-puff latencies on distance from the introduction spot, increasing from about 1.5 s at the introduction site to about 2.5 s at the far end of the modeled cell (Fig. 5F). These predicted values are close to the experimental measurements obtained with distributed flash stimulation across the cell (indicated by the red line in Fig. 5F), concordant with a rapid spatial equilibration of IP3. In marked contrast, simulations with a diffusion coefficient of 10 μm2 s−1 showed large differences in puff latency distributions with distance (Fig. 5C). Mean latencies near the IP3 introduction site (~0.4 s) were shorter than those measured after distributed introduction of the same amount of IP3 (red line), whereas they lengthened to ~11 s at the end of the modeled cell farthest from the site of IP3 introduction (Fig. 5E).

Estimating the IP3 diffusion coefficient by fitting simulated to experimental data

The simulation results indicated that using a diffusion coefficient of 10 μm2 s−1 better matched our experimental results than a diffusion coefficient of 280 μm2 s−1. To refine our estimate of the IP3 diffusion coefficient, we performed additional simulations using a range of diffusion coefficients between 1 and 280 μm2 s−1. Simulations were performed by calculating the probabilities of puff occurrence within 30 s of IP3 introduction (mimicking the experimental recording time for the response to spot flash stimulation of i-IP3). This procedure normalizes the comparison between experimental and simulated data by accounting for missed long-latency puffs that may have occurred after the end of recordings (see methods S1).

Superimposition of the experimental data corresponding to spot flash stimulation for 500 ms (Fig. 2B) with the predicted mean first-puff latencies, as a function of distance from a 500-ms duration introduction of IP3, showed that the experimental data matched best with the simulation curves generated with diffusion constants of 5 or 10 μm2 s−1 (Fig. 6A). Plotting the experimental data together with the simulation curves for all flash durations further indicated that a diffusion coefficient of 280 μm2 s−1 would not match the data well (fig. S5). We quantified the goodness of fit to the simulated relationships by calculating the mean squared differences between experimental and simulated data for different values of IP3 diffusion coefficients (Fig. 6B). For each flash duration and selected diffusion coefficient, the mean squared error is the average of the squared differences between the binned mean observed latencies (Fig. 6A and fig. S5) and the predicted mean latencies at the corresponding distances. This analysis showed that the minimum values (representing the best fit) were obtained with diffusion coefficients between 3 and 10 μm2 s−1, with the best-fit diffusion coefficient tending to higher values with increasing flash duration.

Fig. 6 Comparison of experimental with simulated data indicates an effective intracellular diffusion coefficient for IP3 of 3 to 10 μm2 s−1.

(A) Comparison of experimental and predicted first-puff latencies evoked by a 500-ms spot flash stimulation. Curves plot the predicted mean first-puff latencies as a function of distance along the cell from the stationary photolysis spot for various values of IP3 diffusion coefficient, as indicated in units of μm2 s−1. Data points show mean ± SEM of first-puff latencies from the experiment shown in Fig. 2 (A and B). (B) Mean squared error between observed and predicted puff latencies as a function of the simulated IP3 diffusion coefficient. Curves are shown for experimental and simulated data for flash durations of 100, 200, and 500 ms.

We did not apply this model to estimate a diffusion coefficient for IP3 in COS-7 cells because their more complex morphology cannot be approximated as a one-dimensional situation. Nevertheless, the observed about threefold lengthening of mean puff latency over a 20-μm distance from the spot flash stimulation site indicates that diffusion of IP3 is strongly hindered in COS-7 cells (fig. S3B). If IP3 were to diffuse freely with a diffusion coefficient of 280 μm2 s−1, its concentration would equilibrate over this distance within <1 s, so puff latencies would show little or no apparent dependence on distance.

Range of action of IP3

Allbritton et al. (1) defined the range of action of a messenger as2Dτ(1)where D is the effective diffusion coefficient of the messenger and τ is its mean lifetime before degradation. This expression represents an equilibrium value and does not take into account the time it takes a messenger to diffuse from a localized source. In light of our observations of slow diffusion of IP3, we propose a redefinition of the range of action as (see methods S2)range of action =0xΦT(x)dxΦT(x)dx(2)ΦT(x)=def0T[IP3]D,τ(x,t)dt(3)where x is the distance from the IP3 source, T is any given time (t) after deposition of the messenger, and [IP3]D(x, t) is the IP3 concentration that results from diffusion with effective coefficient D and mean lifetime τ. Intuitively, this can be considered as the distance from a point source over which one-half of the total cumulative actions of the messenger would have occurred (Fig. 7, A and B). Our definition also corresponds to the median distance over which a messenger that is continuously deposited in the one-dimensional model cell at position x = 0 is spread when a time T has elapsed since deposition was initiated (16). In a biological cell, the messenger would be produced (“deposited”) in response to a signal and then its range of action would be dictated by both its degradation and diffusion. Heat maps illustrate the range of action predicted by Eq. 2 as a function of effective diffusion coefficient and degradation rate, calculated for 0.1, 1, 10, and 100 s after an instantaneous appearance of the messenger (Fig. 7, C to F). These heat maps showed that the range of action over short intervals (≤1 s) is narrowly restricted (<5 μm) and dominated by the diffusion coefficient (Fig. 7, C and D), whereas at longer intervals, the range of action becomes more extended and is dependent on both the diffusion coefficient and the rate of degradation of the messenger (Fig. 7, E and F).

Fig. 7 Time dependence of the range of action of an intracellular messenger.

We define the range of action at a given time T as the distance from a transient point source of messenger at which half of its total cumulative actions (given by Eq. 2) will have occurred. (A and B) Illustrations showing the range of action after times T = 1 s (A) and 10 s (B) for a diffusion coefficient D = 10 μm2 s−1 and mean messenger lifetime τ = 10 s. The schematic represents a one-dimensional model cell extending infinitely far on each side from a transient, point source of messenger; the profile and range of action are identical for a one-sided model extending infinitely from a closed end where the messenger is introduced. The y axis is in arbitrary units. (C to F) Heat maps showing the predicted ranges of action of an intracellular messenger at times T = 0.1 s (C), 1 s (D), 10 s (E), and 100 s (F) for different combinations of effective diffusion coefficient and rate of degradation (inverse mean lifetime) of the messenger.


Intracellular diffusion and range of action of IP3

We investigated the diffusion of IP3 in neuroblastoma cells by locally photoreleasing a poorly metabolized IP3 analog at one end of these elongated cells and monitoring its subsequent spread by the timing of discrete IP3-evoked, Ca2+ release events (puffs) at multiple sites along the cell. Using the first-puff latency as an endogenous reporter of local [IP3], we obtained estimates of the effective diffusion coefficient of IP3 in the cytoplasm of intact mammalian cells under physiologically relevant IP3 concentrations, with minimal perturbation of cellular function and structure.

By comparing our experimental puff latency data to predicted puff-triggering probabilities in a one-dimensional model, we obtained the best agreement for an effective diffusion coefficient D of ~3 to 10 μm2 s−1 for IP3. Although our model does not fully replicate the complex three-dimensional cellular architecture, we selected relatively elongated cells for study. Another possible source of error is that basal Ca2+ concentrations were not fully clamped by the EGTA loaded into the cells, so puff triggering may have been accelerated by a slow, progressive increase in basal free cytosolic [Ca2+] (Fig. 1C). This increase in basal [Ca2+] would have the most pronounced effect at sites distant from the photolysis spot where puff latencies were long, with a potential net effect that our values for the IP3 diffusion coefficient are overestimated.

The effective diffusion coefficient of IP3 that we determined is ~30-fold slower than the value (283 ± 53 μm2 s−1) reported by Allbritton et al. (1). The larger value has become widely accepted (1720), leading to the conclusion that the range of action of IP3 is sufficiently large that it functions as a global messenger in mammalian cells of typical size (1, 2). Published values for the half-life (τ) of IP3 in cells range widely from one to several seconds in mammalian cells (2), including 10 to 20 s in neuroblastoma cells (11, 21) to a minute or more in Xenopus oocytes (2, 21). Defining the range of action as in Eq. 1 and assuming that D = 280 μm2 s−1 and τ = 1 s, Allbritton et al. (1) obtained a range of action of about 25 μm. However, taking our value of the effective diffusion coefficient D = 3 to 10 μm2 s−1, the range is reduced to <5 μm, indicating that IP3 should no longer be considered a global messenger and that appreciable spatial gradients of IP3 may arise within even small cells. The range of action of IP3 is comparable to that (~5 μm) for the buffered diffusion of Ca2+ (1, 13), which is considered a local messenger.

Physiological implications of hindered diffusion of IP3

The restricted effective diffusion of IP3 likely has important physiological implications for IP3-mediated signaling. Ca2+ release from the endoplasmic reticulum (ER) through IP3Rs requires binding of both cytosolic IP3 and Ca2+ to receptor sites on the tetrameric IP3R, leading to a regenerative process of Ca2+-induced Ca2+ release (22). This process can support the generation of propagating Ca2+ waves by a fire-diffuse-fire model (23), wherein Ca2+ released through IP3Rs at one site diffuses to activate release through surrounding IP3-bound IP3Rs. Thus, although Ca2+ wave propagation can be considered to overcome the limited effective diffusional range of Ca2+ ions, the extent to which a Ca2+ wave can propagate is set by the diffusional spread of IP3, if IP3 is generated from a localized source. The impact of the diffusional spread will be particularly strong in extended cellular processes, such as neuronal dendrites, and in the propagation of intercellular Ca2+ signals mediated by the diffusion of IP3 through gap junctions (24, 25).

The expression (Eq. 1) for the range of action as formulated by Allbritton et al. (1) does not adequately describe many physiological situations. If the messenger is not degraded, its range of action goes to infinity. Furthermore, the range of action given by Eq. 1 does not take time into account. Our data provide a clear example of these limitations, in that Ca2+ signals at the distal end of the cell arose after latencies ~10-fold longer than those near the site of photoreleased IP3, even though degradation of i-IP3 was negligible (11). Timing is crucial for many physiological processes—it is not merely the final concentration to which a diffusible messenger ultimately increases but how quickly it increases and subsequently decreases is also critical for shaping the cellular response to the messenger. In place of the Allbritton et al. expression, we thus propose a more biologically relevant expression (Eq. 2), equating the range of action as the distance from the source over which one-half of the total cumulative actions of the messenger would have occurred at any given time (Fig. 7, A and B) (methods S2). Using our expression for range of action, at long times (t >> τ) after the appearance of the messenger, the expression converges to Dτ, which yields a range of action shorter by a factor of 2 than the range given by the Allbritton et al. expression. For example, for D = 10 μm2 s−1 and τ = 1 s, the steady-state range of action given by our expression is ~3.3 and that given by the Allbritton et al. expression is ~4.5 μm. Potentially of more biological importance, Eq. 2 indicates that for short times (t < τ), the range of action becomes increasingly independent of messenger degradation and varies approximately as a square root function of time. Considering the same example, the predicted range of action is further reduced to 1.60 μm over an interval of 500 ms after the introduction of messenger.

This dependence on IP3 metabolism may be of particular importance for IP3 signaling in neurons, where the dual gating of IP3 receptors by IP3 and cytosolic Ca2+ has been proposed to function as a coincidence detector. Specifically, the ability of IP3 to evoke intracellular Ca2+ release as it diffuses along a dendrite after local metabotropic receptor activation at one site may be promoted by a subsequent action potential that evokes entry of Ca2+ through plasma membrane channels (26, 27) or by summation with local production of IP3 (28) at other sites further along the dendrite. A lower effective diffusion coefficient of IP3 would narrow the spatiotemporal window over which the coincidence of these signals would occur because the spread of IP3 would be more restricted, and IP3 concentrations at distant sites may rise too slowly to coincide with the increase in Ca2+.

Hindered diffusion of IP3 by binding to silent IP3Rs

What might underlie the markedly hindered diffusion of IP3 in neuroblastoma cells, and why does this markedly differ from the free mobility of IP3 reported by Allbritton et al. in oocyte cytoplasmic extracts? The slow intracellular diffusion of IP3 in the cytosol cannot readily be explained by nonspecific factors, such as tortuosity or viscosity, given that the diffusion coefficient of adenosine 5´-triphosphate (ATP)—a molecule of similar size and identical charge to IP3—is only slightly lower in cells than in free solution [248 and 349 μm2 s−1, respectively (29)]. Instead, we propose that diffusion of IP3 is limited by binding to the large number of IP3Rs estimated to be present in SH-SY5Y and other mammalian cells (5). The motility of IP3Rs in the ER membrane is low with a diffusion coefficient of <0.1 μm2 s−1 (6), so they can be considered as immobile on the time and distance scales of our experiments. Binding to stationary, unoccupied receptor sites is thus expected to slow the effective diffusion of IP3 in a manner analogous to the slowing of the diffusion of Ca2+ ions by binding to immobile cytoplasmic buffers. The ability of IP3Rs to buffer IP3 and slow its effective diffusion will diminish at increasing concentrations of IP3 because an increasing fraction of the binding sites becomes occupied. Our results showed a trend for the effective diffusion coefficient to increase from about 3 to 10 μm2 s−1 over the fivefold range of photoreleased i-IP3 that we explored. We anticipate that the released i-IP3 represents a physiological range of IP3 concentrations because even the weakest photolysis flash evoked robust global Ca2+ waves in cells not loaded with EGTA (fig. S2).

Concordant with a mechanism of physiological IP3 buffering, Finch and Augustine (30) observed that Ca2+ signals evoked by spot photorelease of IP3 (not the poorly metabolized i-IP3 analog) in the dendrites of cerebellar Purkinje cells were restricted to a few micrometers, a finding that they interpreted to result from binding to the high density of IP3Rs in these cells . In contrast, the experiments of Allbritton et al. (1) were made using cytosolic extracts from Xenopus oocytes in which IP3Rs are concentrated in a thin circumferential shell around these giant cells (31, 32). The concentration of IP3Rs in these extracts would have been diluted by the large bulk of cytoplasm from the oocyte interior. Moreover, binding to receptor sites may have been saturated by the concentration of radiolabeled IP3 (100 nM total IP3) applied to monitor diffusion. In this context, we therefore believe that the Xenopus oocyte cytoplasm does not serve well as a model system for small mammalian cells.

The calculated diffusion coefficient for IP3 in a medium with twice the viscosity of water is about 250 μm2 s−1 (1), similar to the measured value for ATP in cells (29). We interpret our finding of an effective diffusion coefficient of ≤10 μm2 s−1 as arising because specific binding of IP3 to immobile sites slows its effective diffusion by a factor of ≥25. Assuming that binding equilibrates rapidly and that only a small fraction of all sites are bound, the effective diffusion coefficient Deff is given byDeff=D1+R(4)where D is the diffusion coefficient in the absence of binding and R is the ratio of concentration of binding sites to the dissociation constant Kd of binding to these sites (33). Our results thus lead to an R value of ≥24. This differs markedly from the value of ~1 calculated from the intracellular IP3R concentration of 100 nM and a Kd of 113 nM reported by Taylor and Konieczny (5) and Ding et al. (34). With an R of 1, the effective diffusion coefficient of IP3 would be reduced by no more than a factor of 2 from that in the absence of binding.

Silent and functional IP3Rs

Ca2+ puffs involve the concerted opening of several IP3Rs, tightly clustered at immobile, sparsely distributed sites throughout the cell (3, 4). A long-standing question is why Ca2+ release is restricted to these specific sites, involving only about 3% of the total number of IP3Rs in the cell, whereas the great majority of IP3Rs appear functionally silent under the same conditions even though the slow diffusion of IP3 suggests that they bind IP3 (5, 35).

Our results, considered together with evidence that the tetrameric IP3R channel requires IP3 to be bound to all four receptor sites before it can be activated (5, 36), cast light on this issue. The hindered diffusion of IP3 and the relative insensitivity of the effective diffusion coefficient to changes in the amount of photoreleased IP3 (from ~3 μm2 s−1 with a 100-ms flash to ~10 μm2 s−1 with a 500-ms flash) indicated that a substantial fraction of binding sites must remain unoccupied. How then can Ca2+ puffs be generated under these conditions, given that they involve the concerted opening of several, tightly clustered IP3Rs (3, 4)? As an example, for a cluster of four channels, the probability that all will be available to open during a puff increases as the 16th power of [IP3] increases, so a concentration that yields a 50% probability that all four IP3Rs are activated (binding four IP3 molecules) corresponds to an overall occupancy of receptor sites of 95.8%. It seems that the IP3Rs active at puff sites must be functionally different from the silent receptors that hinder binding, in addition to their differing spatial localization and motility (5, 6, 35). The mechanism remains unknown, but possibilities include that receptors at puff sites exhibit a much higher affinity of IP3 binding or that their gating properties are modified so that channel opening requires binding of fewer than four IP3 molecules per tetramer.



Unless otherwise indicated, all reagents were obtained from Sigma, and cell culture media were from Invitrogen.

Cell culture

SH-SY5Y human neuroblastoma cells were cultured as previously described (4) in a 1:1 mix of Ham’s F12 medium and Eagle’s minimum essential medium, supplemented with fetal calf serum [10% (v/v)], nonessential amino acids [1% (v/v)], and penicillin-streptomycin [1% (v/v)]. Cells were incubated at 37°C in a humidified incubator with 95% air and 5% CO2 and were passaged every 2 to 3 days to a maximum of 20 passages. Four days before imaging, cells were harvested in Ca2+/Mg2+-free phosphate-buffered saline (pH 7.4) and cultured on glass coverslips in petri dishes (35-mm dish, no. 1.0 coverglass; MatTek) at a density of ~3 × 104 cells/ml. On the day of imaging, cells were washed in Hepes-buffered saline solution [HBS; composition: 135 mM NaCl, 5 mM KCl, 1.2 mM MgCl, 2.5 mM CaCl2, 5 mM Hepes, and 10 mM glucose (pH 7.4)].

Ca2+ imaging and localized photorelease of i-IP3

Cultured SH-SY5Y cells were loaded for imaging by incubation with membrane-permeant esters of Fluo-4 (1 μM; Invitrogen), EGTA (5 μM; Invitrogen), and caged i-IP3 (1 μM; SiChem) in HBS, as described (4). Cells averaged ~40 μm in length along the cell body (fig. S4A). Cytosolic Ca2+ changes were imaged using a total internal reflection fluorescence (TIRF) microscope system (4) constructed around an Olympus IX70 microscope with a 60× TIRF objective (numerical aperture, 1.45). Fluo-4 fluorescence was excited by 488-nm laser light within an evanescent field extending a few hundred nanometers into the cells, and emitted fluorescence (λ > 510 nm) was imaged at a resolution of 256 × 256 pixels (1 pixel = 0.266 μm) at an exposure time of 15 ms (~66 frames s−1) using the center quad of an Evolve 512 electron-multiplying charge-coupled device camera (Roper Scientific). Image data were acquired as stack files using MetaMorph v7.7 (Universal Imaging/Molecular Devices) and were analyzed offline to detect the locations of puff sites and to measure puff latencies. The custom software used for analysis is described in and available from the work of Ellefsen et al. (14). Measurements were exported to Microcal Origin v8.0 (OriginLab) for analysis and graphing. Unless otherwise noted, data are presented as means ± 1 SEM.

Photorelease of i-IP3

Photolysis of caged i-IP3 was evoked by a custom-built system using computer-controlled galvanometer mirrors to direct light from a 405-nm laser diode module. The laser light was focused to a spot in the specimen (fig. S4B), which could be steered to any desired location within the imaging field using dim light from a coaxial 450-nm laser as a “guide star” to excite fluorescence in Fluo-4–loaded cells without causing photorelease of i-IP3. A computer-generated TTL (transistor-transistor logic) signal controlled the duration of the 405-nm photolysis spot flash, and a variable neutral density filter wheel controlled its intensity. Experiments were performed using photolysis flashes of given duration and fixed intensity, with the laser spot either remaining stationary at one end of a cell to evoke localized photorelease of i-IP3 (spot flash) or scanned along the length of a cell during the same flash duration to evoke a spatially distributed photorelease (distributed flash). In the latter case, a line was defined by clicking the computer mouse after positioning the guide laser spot at each end of the cell, and the computer then stepped the 405-nm laser spot in 10 equal increments along the line during the flash duration. For example, a 100-ms flash would be delivered as 10 exposures, each with a 10-ms duration, spaced at 4-μm increments along a cell 40 μm in length. Given that flash durations were short (100 to 500 ms) in comparison to puff latencies, this distributed flash protocol leads to an approximately uniform photorelease of i-IP3 throughout the cell, in an amount equivalent to that photoreleased by a stationary spot flash of the same duration and intensity.

Analytical derivation and numerical simulation of puff latencies

To simulate the diffusion of IP3 and subsequent triggering of initial puffs after spot flash stimulation, we applied a simplified one-dimensional model. Analytical solutions of the model are presented in methods S1. To generate model data with which to fit to experimental observations, we performed numerical simulations of these solutions. The cell was represented as a 40-μm-long array, with closed ends, composed of 100 elements, each containing a single puff site. To simulate uncaging of IP3, we increased the concentration of IP3 to a fixed amount for every time step during the duration of the photolysis flash. The spatial profile of uncaging was simulated as a Gaussian curve with a width (σ = 1.35 μm) corresponding to that of the UV laser spot. Because i-IP3 is metabolically degraded more slowly than endogenous IP3, the rate of metabolism was set to 0, except where otherwise noted; that is to say, the total amount of IP3 remained constant after the uncaging period. To then simulate the probability of a puff being triggered at each spatial element, we assumed a linear relationship between [IP3] and puff triggering (10), with a rate constant (1.11 s−1 per 1 s flash duration) derived from experiments (Fig. 3C), in which corresponding amounts of i-IP3 were photoreleased uniformly throughout cells by the distributed flash protocol. For each spatial element, we then took the triggering rate as given by the concentration of IP3 at a given time step and calculated the probability of observing the first puff at that site over successive time steps. Simulations were terminated after 30 s to match the duration of experimental recordings, and calculations of mean first-puff latencies were made only over this interval. Thus, mean values from both experimental and simulated data were unaffected by “missed” sites that failed to show a puff within 30s.


Methods S1. Theoretical derivation of puff latencies

Methods S2. Calculation of range of action

Fig. S1. First-puff latencies depended strongly on the distance from the spot flash stimulation, whereas there was no significant distance dependence for distributed flash stimulation.

Fig. S2. Ca2+ waves evoked by photoreleased i-IP3 in SH-SY5Y cells without EGTA.

Fig. S3. Latencies of puffs evoked by spot photorelease of i-IP3 in COS-7 cells.

Fig. S4. Lengths of SH-SY5Y cells and spatial extent of the photolysis laser spot.

Fig. S5. Experimental and simulated first-puff latency data for the 100- and 200-ms spot flash durations.

Movie S1. Puffs evoked by localized and distributed flash photorelease of IP3.

Reference (37)


Acknowledgments: We thank N. Hengartner and D. Colquhoun for helpful discussions and B. Settle for help with software development. Funding: This work was supported by the NIH through grants R37 GM048071 (to I.P.), RO1 GM065830 (to J.E.P. and I.P.), and F31 GM119330-01 (to K.E.); and by UBA and FONCyT (Argentina) through grants UBACyT 20020130100480BA and PICT 2013-1301 (to S.D.). Author contributions: G.D.D. performed experiments and analyzed experimental data. K.L.E. wrote analytical and simulation software and together with G.D.D. performed numerical simulations. J.E.P. and S.P.D. formulated the analytical solutions and performed statistical analysis. I.P. conceived and supervised the project. All authors contributed to writing the paper and read and approved the final manuscript. Competing interests: The authors declare that they have no competing interests. Data and software availability: Raw data and custom software are freely available to academic researchers upon request.
View Abstract

Navigate This Article