Local Ca²⁺ releases enable rapid heart rates in developing cardiomyocytes

Abstract

The ability to generate homogeneous intracellular Ca²⁺ oscillations at high frequency is the basis of the rhythmic contractions of mammalian cardiac myocytes. While the specific mechanisms and structures enabling homogeneous high-frequency Ca²⁺ signals in adult cardiomyocytes are well characterized, it is not known how these kind of Ca²⁺ signals are produced in developing cardiomyocytes. Here we investigated the mechanisms reducing spatial and temporal heterogeneity of cytosolic Ca²⁺ signals in mouse embryonic ventricular cardiomyocytes. We show that in developing cardiomyocytes the propagating Ca²⁺ signals are amplified in cytosol by local Ca²⁺ releases. Local releases are based on regular 3-D sarcoplasmic reticulum (SR) structures containing SR Ca²⁺ uptake ATPases (SERCA) and Ca²⁺ release channels (ryanodine receptors, RyRs) at regular intervals throughout the cytosol. By evoking [Ca²⁺]_i-induced Ca²⁺ sparks, the local release sites promote a 3-fold increase in the cytosolic Ca²⁺ propagation speed. We further demonstrate by mathematical modelling that without these local release sites the developing cardiomyocytes lose their ability to generate homogeneous global Ca²⁺ signals at a sufficiently high frequency. The mechanism described here is robust and indispensable for normal mammalian cardiomyocyte function from the first heartbeats during the early embryonic phase till terminal differentiation after birth. These results suggest that local cytosolic Ca²⁺ releases are indispensable for normal cardiomyocyte development and function of developing heart.

Introduction

Mammalian cardiac myocytes are relatively large cells, capable of generating large cytosolic Ca²⁺ signals at high frequency (Bers, 2002). These homogeneous, global and transient Ca²⁺ signals are absolute requirements for the forceful repetitive contractions of cardiac myocytes. In practice, such signals require the existence of efficient mechanisms for Ca²⁺ release to the cytosol and for removal from the cytosol with minimal diffusion distances. For efficient and fast Ca²⁺ release, adult atrial cells have local Ca²⁺ release channels in the cytosol (Hatem et al. 1997) and adult ventricular cardiomyocytes have unique cell membrane invaginations called t-tubules (Brette & Orchard, 2003), both of which increase the spatiotemporal homogeneity of the Ca²⁺ signals. For example, in the t-tubular network of the adult ventricular myocytes the voltage-activated Ca²⁺ channels (VACCs), which produce Ca²⁺ influx during the action potential, are located at the t-tubules and the ryanodine receptors (RyRs), which induce calcium-induced release from the sarcoplasmic reticulum (SR), are located immediately in the near vicinity on the intracellular side of the VACCs. Thereby t-tubules form a 3-dimensional (3-D) structure linking the membranes and enabling Ca²⁺-induced Ca²⁺ release (CICR) to occur simultaneously throughout the whole myocyte (Bers, 2002). Importantly, this fine 3-D membrane architecture minimizes the Ca²⁺ diffusion distances in the cytosol, which fundamentally enables global high-frequency Ca²⁺ signals.

During mammalian heart development, cardiomyocytes start regular beating activity during the first trimester of gestation. Already during embryonic development, the heart has a high basal heart rate: ∼50% of that of adult mouse (Desai et al. 1999; Phoon et al. 2000; Ji et al. 2003; Yu et al. 2008). Surprisingly, developing ventricular cardiomyocytes do not have t-tubules (Seki et al. 2003) until the myocytes reach the mature developmental stage several weeks after birth (Forbes et al. 1984). Unlike adult terminally differentiated cardiomyocytes, developing cardiomyocytes grow, divide, differentiate and even migrate, making it impossible to maintain complex stable structures such as the t-tubular network of adult ventricular myocytes. It is not known how the developing mammalian cardiomyocytes can still generate global Ca²⁺ oscillations at relatively high frequency.

Interestingly, already during early heart development, cardiomyocytes have RyRs and SR Ca²⁺ pump (SERCA) proteins between the perinuclear and sub-sarcolemmal region (Seki et al. 2003; de Diego et al. 2008). We hypothesized that if these proteins form regular functional structures, they could establish a network of Ca²⁺ release sites in the cytosol of developing cardiomyocytes. This structure would make it possible for Ca²⁺ signals to propagate as ‘fire–diffusion–fire’ propagation, i.e. a Ca²⁺ wave (Lipp et al. 1990; Trafford et al. 1995; Cheng et al. 1996; Keizer et al. 1998), providing a more rapid propagation of Ca²⁺ signal and consequently increase the spatiotemporal homogeneity of the cytosolic Ca²⁺ signals.

Here we show how the coupling of sarcolemmal (SL) and SR Ca²⁺ signalling is enforced in developing cardiomyocytes. We report that embryonic mouse cardiac myocytes have extensions of SR, which form regular structures at ∼2 μm intervals with both ryanodine receptors and SR Ca²⁺-ATPases for peripheral Ca²⁺ release and uptake. These Ca²⁺ release units provide local Ca²⁺ releases between SR and SL, and amplify and speed up the propagating Ca²⁺ wave. This mechanism expedites the Ca²⁺ wave propagation, enabling the cells to grow and yet retain their ability to produce whole-cell Ca²⁺ oscillations at a sufficiently high rate. Based on structural and functional characterization of the embryonic myocytes, we propose a model for cytosolic Ca²⁺ propagation of developing cardiomyocytes. We further demonstrate that without this mechanism, developing cardiomyocytes could not achieve high enough beating rates that are required for normal development.

Methods

Cell isolation and culturing

Pregnant CD-1 mice from the Center for Experimental Animals at the University of Oulu were killed by cervical dislocation. Embryos at 12 days old (E12) were rapidly excised and transferred to +2°C isolation buffer, and vertricular cardiomyocytes were isolated as previously described (Rapila et al. 2008). Cells were cultured on laminin-coated, glass-bottom Petri dishes for 12–20 h before using them in the experiments. All procedures were carried out in compliance with the policies and regulations set out by Drummond (2009), and the experimental designs were approved by the Animal Use and Care Committee of the University of Oulu.

Electrophysiology and Ca²⁺ imaging

E12.5 mouse cardiomyocytes were loaded with Fluo-4-acetoxymethyl (AM)-ester (1 μm dissolved in pluronic DMSO; Invitrogen) as previously for Ca²⁺ imaging (Rapila et al. 2008). Loaded cells were placed in a custom-made perfusion system built into a confocal inverted microscope (FluoView 1000, Olympus). Cells were superfused with preheated (T= 34°C) Dulbecco's modified Eagle medium plus glutamax I (Invitrogen) culturing medium (pH 7.4, bubbled with 95% O₂–5% CO₂). Loaded myocytes were excited at 488 nm and the emitted light was collected with a spectral detector from 520 to 620 nm through a 20× or 60× objective lens. To excite the cells, myocytes were stimulated with 1 ms voltage pulses 50% over the excitation threshold through two platinum wires located on both sides of the Petri dish. Cells were line scanned at 300–400 Hz depending on the length of the scanning line, with a pixel time of 4–10 μs.

The whole-cell patch-clamp method (Axopatch-1D with pCLAMP 9.0 software, Axon Instruments) was used to inject 1 μm Ca²⁺ solution into the Fluo-4-loaded cells while measuring the Ca²⁺ signal simultaneously as described above. Glass capillary electrodes (5 MΩ) were filled with solution containing (in mm): 0.84 CaCl₂, 130KCl, 5 Na₂-phosphocreatine, 5 Mg-ATP, 1 EGTA and 10 Hepes (pH adjusted to 7.2 with KOH), which results in 1.042 μm free Ca²⁺ (calculated with K_d(Ca2+)= 0.2 m for EGTA (Smith et al. 1984)). The pipette tip was attached with a giga-seal to the cell membrane. A microinjector (Picopritser II, Parker Instrumentation) connected to the patch-clamp pipette was used to generate a 1–10 ms pressure pulse to the inside of the patch-clamp pipette and consequently rupture the cell membrane and inject a small amount of the pipette solution into the cell. The pressure setting in the microinjector was 68 psi (469 kPa). The cell was held at a −70 mV voltage-clamp.

Immunofluorescence labelling and microscopy

Cells were rinsed with 0.1 mol l⁻¹ Tris-HCl, pH 7.3, fixed with 3% paraformaldehyde for 2 min, and permeabilized for 10 min with 0.5% Triton X-100. After washing with 0.1 mol l⁻¹ Tris-HCl, pH 7.3 twice for 5 min, the primary SERCA2a (goat anti-SERCA2, Santa Cruz sc-8094, Santa Cruz Biotechnology, USA) or RyR (rabbit anti-ryanodine receptor 2, Millipore AB9080, Millipore, USA) antibody was incubated in 0.1 mol l⁻¹ Tris-HCl (pH 6.9) containing 10% FBS, 0.05% Triton X-100 and 100 mmol l⁻¹ NaCl at +4°C overnight. Again the specimens were washed twice and the secondary antibody Alexa Fluor 488 chicken anti-goat (Molecular Probes) or Alexa Fluor 488 chicken anti-rabbit (Molecular Probes) was incubated (pH 7.3) for 1 h at room temperature. Dilution for primary antibodies was 1:100 and secondary antibody 1:750. After labelling, images were taken immediately with an Olympus FV1000 confocal microscope (excitation 488 nm, emission 505–605 nm) using a 60× objective.

Mathematical modelling of Ca²⁺ diffusion

Fick's second law of diffusion with reflecting boundary conditions was used to simulate the 1-D Ca²⁺ diffusion in cartesian coordinates in Figs 1A and 2E. The diffusion coefficient was 0.79 μm² ms⁻¹ (Cussler, 1997) and the volume in which diffusion occurred had the same concentrations of Ca²⁺ buffers as previously used in the cytosol of an embryonic cardiomyocyte model (Korhonen et al. 2008). The initial [Ca²⁺] was 0.2 μm homogeneously in the volume. In Fig. 2E the volume had release sites located at 2 μm intervals, which released Ca²⁺ with a flux of 0.35 μm ms⁻¹ (in cytosol volume) for 10 ms after the threshold of 0.25 μm for local [Ca²⁺] was crossed.

Figure 1. In embryonic cardiomyocytes Ca²⁺ propagation is faster than diffusion.

A, top: line-scan Ca²⁺ image across the cytosol of an E12.5 mouse cardiomyocyte during electrical excitation. Bottom: simulated 1-D diffusion of Ca²⁺ for 8 μm length. During t= 0–1 ms, 7 μm of Ca²⁺, based on our previous estimate of the amount of Ca²⁺ intrusion to the cytosol during electrical excitation of the embryonic cardiomyocytes (Korhonen et al. 2008), was released as a constant flux to the r= 0 μm. The insets show the time-to-target plots from the representative experimental and simulated data. B, a curve with the mean ±s.e.m. (n= 5) slope of experimental time-to-target plots together with the time-to-target plot from the simulated diffusion.

Figure 2. Ryanodine receptors amplify cytosolic Ca²⁺ propagation in mouse embryonic cardiomyocytes.

A, routes for Ca²⁺ signal propagation in an embryonic cardiomyocyte placed in a culture dish during electrical excitation (top) and during injection of Ca²⁺ solution from a patch pipette (bottom). The dashed line shows the placement of the line-scan line used when recording Ca²⁺ signals in the experiment. B, photograph of the experimental setup for recording Ca²⁺ diffusion in the cardiac myocyte. A 1 μm Ca²⁺ solution was injected through a patch-clamp pipette (orange) and the [Ca²⁺] in the myocyte was recorded with a confocal microscope in line-scan mode (dashed line). The scale bar is 10 μm. C, a line-scan recording of a Ca²⁺ signal during injection of 1 μm Ca²⁺ solution into the cell. The recordings in control (top) and after 3 min 50 μm ryanodine treatment (bottom) are shown. The scale bars are 200 ms in the horizontal and 5 μm in the vertical direction. The insets on the right show the traces of Ca²⁺ signals taken from near the point of Ca²⁺ injection and 15 μm away from this point. D, mean propagation times (left) and time-to-target plots (right) of the Ca²⁺ signals after the injection of 1 μm Ca²⁺ solution into the cell (n= 5 cells for control, black; and n= 4 cells for ryanodine, grey). E, simulated 1-D propagation of Ca²⁺ for 10 μm length. At t= 0 the [Ca²⁺] at r= 0 μm was set to 1 μm to simulate the Ca²⁺ injection in the in vitro experimental design in B. The panel on the right shows the Ca²⁺ signals at r= 0 μm and at r= 5 μm, when the local release sites produce 75%, 50%, 25% and 0% release flux compared to the control situation (100%).

Mathematical modelling of E–C coupling of a developing cardiomyocyte

To simulate the excitation–contraction coupling (E–C coupling) of a developing cardiomyocyte, the previously published model of an E9–11 embryonic mouse cardiomyocyte was used in its original form and with three different increased cell sizes (Korhonen et al. 2008). The cell model is spherically shaped, with the nucleus in the centre of the cell and the SR as a thin layer on the surface of the nucleus. Based on confocal microscopy images, the r_nucleus is approximately the same in E12.5 myocytes as that used in the original E9–11 model (∼5% difference, n= 28), but the minimum distance that the Ca²⁺ signal has to propagate between SL and SR in the centre of the cell is ∼54% higher in E12.5 myocytes than that used in the original E9–11 model (5.4 ± 0.4 μm, n= 28, vs. 3.5 μm) (Korhonen et al. 2008). Consequently, the free diffusion distance between SL and SR (r_SL-SR) was increased from the 3.5 μm value to 4, 6 and 8 μm values. The magnitudes of SL ion currents were scaled to maintain the current densities while increasing the cell size. The volume of the SR was increased to the same degree as the volume of the cytosol, based on the assumption that the cell maintains its cytosol volume-related Ca²⁺ release capacity. However, the surface area of the SR facing the cytosol and the distance from the centre of the cell to the surface of the SR were approximated to stay constant. When implementing the maximum increment of r_SL-SR (3.5 μm → 8 μm) studied here to the spherical geometry, the resulting errors of maintaining the SR surface area and cell centre–SL distance were only 1.2% and 2.4%, respectively.

When the local release sites between the centre of the cell and SL were simulated, the SR was divided into independent fractions which all had equal Ca²⁺ storing capacity (i.e. volume and Ca²⁺ buffering) compared to the local cytosol volume into which each SR releases its Ca²⁺. The local release sites were placed 2 μm apart (see Results). The Ca²⁺ flux between the local SR and cytosol at a local release site i, consisting of release and uptake fluxes (J_rel, J_uptake), was modelled as

(1)

where t_i is the time when [Ca²⁺]_i crosses the threshold for the release in the positive direction (d[Ca²⁺]_i/dt > 0) and Θ is the Heaviside step function (Keizer et al. 1998; Shannon et al. 2000; Coombes et al. 2004). The parameters are shown in Table 1. When the local Ca²⁺ concentration exceeds the threshold value, the J_rel generates a release flux, the magnitude of which depends on the local SR Ca²⁺ concentration, for time τ. The J_uptake simulates the function of Ca²⁺ uptake by SERCA to the local SR.

Table 1.

Parameters of the model of local release site

Parameter	Definition	Value	Reference
vrel	Release flux scaling factor	0.05 ms−1
τ	Duration of the release	10 ms	(Coombes et al. 2004)
	[Ca2+]i threshold for the release	0.25 μm	(Keizer & Levine, 1996; Dawson et al. 1999)
vuptake	Maximum uptake flux	1 μm ms−1
H	Hill's coefficient for the forward and reverse SERCA flux	2	(Shannon et al. 2004)
Kmf	Half-maximal activation of the forward flux	0.25 μm	(Shannon & Bers, 1997; Shannon et al. 2000, 2004)
Kmr	Half-maximal activation of the reverse flux	1.75 μm	(Shannon & Bers, 1997; Shannon et al. 2000, 2004)

In pacing simulations the APs were triggered by applying −80 pA pF⁻¹ current for 0.5 ms. The concentration of inositol-3-phosphate [IP₃] in cytosol was 0.06 μm (Fig. 4B) and [IP₃]= 0.075 μm (Fig. 4D and E) in the model unless stated otherwise.

Figure 4. Local release sites are required for homogenous Ca²⁺ signals.

A, schematic diagram of the model structure and Ca²⁺ signal propagation in the model with (left) and without (right) local Ca²⁺ release sites. B, the membrane voltage (top) and intracellular [Ca²⁺] (bottom) from the simulation of spontaneous and externally triggered E–C coupling with (left) and without (right) local Ca²⁺ release sites. The simulated intracellular Ca²⁺ is shown as a line-scan between the cell membrane (distance, 13.5 μm) and the nucleus (distance, 7.5 μm). The insets on the right show [Ca²⁺] from one cycle of spontaneous activity from the model with (top) and without (bottom) local Ca²⁺ release sites in the same scale (50 ms and 1 μm scalebars). C, the spontaneous AP frequency vs.[IP₃] from different sized model cells with and without local Ca²⁺ release sites. D, the range of pacing frequencies to which the Ca²⁺ signalling could be synchronized in the model and in experiments. In the model the frequencies 0.5, 1, 2, 2.5 and 3 Hz and in the experiment the frequencies 0.5, 1, 2 and 3 Hz were tested (n_cells= 12).

Numerical methods

As previously, the partial differential equations in the Ca²⁺ diffusion and E–C coupling models were approximated to a system of ordinary differential equations (ODEs) with spatial step dx= 0.1 μm and simulated in Matlab 6.5 (The MathWorks) with the ‘ode15s’ solver for stiff ODEs (Korhonen et al. 2008, 2009). The ‘Event Location’ feature of Matlab's ODE solver algorithms was used to detect the time when the local [Ca²⁺]_i crossed the threshold for local Ca²⁺ release. In the simulations of the E–C coupling model, the simulation was run for 2000 s to reach steady-state intracellular ion concentrations.

Statistics and data analysis

Experimental data are presented as mean ±s.e.m. unless stated otherwise. Data were analysed with Origin 7.5 (OriginLab Corporation), Matlab 6.5 (The MathWorks), ImageJ 1.36b (http://rsb.info.nih.gov/ij/) and SparkMaster (Picht et al. 2007). Statistical significance of difference was analysed using one-way ANOVA. Linear fits were determined using unweighted linear regression. Moving average filtering was used when necessary in analysing confocal images.

The Fourier transform of line-scanned immunolabel signals was calculated as 128-point fast Fourier transform (FFT). Before applying the FFT, the linear trends, DC components and noise (less than 50% of maximum intensity fluctuations) were removed and signals were band-pass filtered with a 10th order Butterworth filter to contain information from the range 0.2–5 1 μm⁻¹.

The time-to-target plots were used to analyse the Ca²⁺ propagation (Kirk et al. 2003). The half-point between maximal and minimal fluorescence was used as the target value when analysing the experimental data. When analysing the simulated diffusion, the half-point between the steady-state concentration at the end of the simulation and the initial concentration values was used as the target value. Propagation time was obtained as the slope of the linear fit in the time-to-target plot.

Results

Intracellular Ca²⁺ propagation

Early embryonic cardiomyocytes have two modes to initiate global Ca²⁺ signals: spontaneous pacemaking activity which originates from spontaneous SR Ca²⁺ releases and the prototypical E–C coupling mode, where triggered action potentials induce voltage-activated Ca²⁺ influx and subsequent Ca²⁺-induced Ca²⁺ release from the SR (Rapila et al. 2008). In both modes, due to the lack of t-tubules, the Ca²⁺ signal has to propagate through the cytosol between the SR located in the perinuclear part of the cell and the cell membrane. As revealed by confocal line-scanning this results in spatial and temporal Ca²⁺ concentration gradients within the cytosol (Fig. 1A). However, the time delay of the Ca²⁺ signal is linear and surprisingly small (R of linear fit, 0.82 ± 0.05; slope of the fit, 4.4 ± 1.1 ms μm⁻¹; n= 5; Fig. 1B). If the Ca²⁺ signal would propagate by free passive diffusion, it would result in an almost 10-fold time delay and a more exponentially than linearly shaped time-to-target plot (Fig. 1A and B) (Kirk et al. 2003).

However, recording cytosolic Ca²⁺ propagation in the X–Y plane with a confocal microscope may give an overestimation of the speed of the Ca²⁺ wave. As the action potential (AP) spreads almost immediately throughout the cell membrane, the VACCs are activated simultaneously in all parts of the cell membrane. This results in 3-D Ca²⁺ fluxes towards the confocal plane of view from all parts of the cell membrane (Fig. 2A), and the measured line-scan image in the 2-D X–Y plane is a projection of these 3-D fluxes. Especially in flat cells, the Z components of these Ca²⁺ fluxes reach the line-scan faster than the X–Y components, resulting in an overestimation of the Ca²⁺ propagation speed (Fig. 2A). To avoid this methodological pitfall and to measure the propagation of the Ca²⁺ signal more accurately in a more controlled environment, we injected a solution containing 1 μm Ca²⁺ at a single edge of the cell and measured the spread of Ca²⁺ in the cytosol. To do this we made a giga-seal contact between the cell and a patch-clamp pipette filled with Ca²⁺ solution. The line-scan image was recorded while a small pressure injection was applied to the patch-clamp pipette to rupture the cell membrane and at the same time inject a pulse (1 ms duration) of the 1 μm Ca²⁺ solution into the cell (Fig. 2B and C). During and after the injection the membrane potential of the cell was voltage-clamped to −70 mV. With this experimental setting we observed a slower Ca²⁺ signal propagation than with electrical activation of all VACCs of the cell membrane (10.1 ± 2.7 vs. 4.4 ± 1.1 ms μm⁻¹, P < 0.05) (Fig. 2D, left panel). Nevertheless, the propagation of the Ca²⁺ wave in these cells was still several times faster than the maximal speed with pure diffusion, suggesting that Ca²⁺ signal propagation is indeed amplified.

It was shown earlier that local RyR Ca²⁺ releases can amplify propagating Ca²⁺ signals in cardiac myocytes (Huser et al. 1996; Sheehan & Blatter, 2003; Korhonen et al. 2009). Therefore, we next tested if inhibition of ryanodine receptors by pre-application of ryanodine affects the propagation of the Ca²⁺. As expected, the propagation time of the Ca²⁺ signal was ∼3-fold longer when the cell was pre-exposed to ryanodine (50 μm) (Fig. 2C and D, left panel). Furthermore, in control situations the Ca²⁺ signal travelling away from the patch-pipette increases with a high initial slope, whereas the ryanodine pre-treated cells show a slowly rising Ca²⁺ signal with a sigmoidal shape (Fig. 2C). Based on Fick's second law of diffusion, when the Ca²⁺ signal propagates by pure diffusion, the concentration gradient driving the diffusion is diluted along the propagation. This results in a lower initial slope at the points away from the original source of Ca²⁺. Comparison of the time-to-target plots (Fig. 2D, right panel) indicates that pre-application of ryanodine causes the Ca²⁺ signal to propagate by free passive diffusion (exponential curve). On the contrary, in control conditions, the time-to-target plot is linear, which suggests that diffusion is facilitated. The local Ca²⁺ releases prevent this dilution and consequently maintain the concentration gradient so that the initial slope of the Ca²⁺ signal remains high in the front line of the propagation. This phenomenon was reproduced by computer simulations where the local Ca²⁺ releases are gradually inhibited (Fig. 2E). Also, the simulated propagation time with full release capacity (10.9 ms μm⁻¹, measured from a time-to-target plot generated with 0.6 μm target value) is in line with the control situation of the experiments (10.1 ± 2.7 ms μm⁻¹). From the simulated RyR activation levels, the propagation time with 50% release (39.6 ms μm⁻¹) corresponds closest to the propagation time in the experiments (33.6 ± 9.2 ms μm⁻¹).

Structure and localization of SR

The ryanodine-sensitive amplification of Ca²⁺ signals probably requires the presence of both RyRs for Ca²⁺ release and SERCAs for Ca²⁺ reuptake to the local SR. The confocal microscope images from E12.5 cardiomyocytes with antibody-labelled SERCAs or RyRs show clear granular structures in the cytosol area between perinuclear SR and SL (Fig. 3A). The structures seem to be similar in both the radial and perpendicular to radial axis directions. To quantify the possible regularity of these structures, we took the FFT from the line-scan intensity signals from these directions (Fig. 3A and B). In the FFT signal, the intensity peaks are at ∼0.5 μm⁻¹, giving the average distance of ∼2 μm for both RyR and SERCA in the radial and perpendicular to radial directions.

Figure 3. Developing myocytes have a regular arrangement of local Ca²⁺ release sites in the cytosol.

A, left: frame-scan images of SERCA (top) and RyR (bottom) immunolabelled cardiomyocytes. The scale bar is 5 μm. Right: line-scans in the radial and perpendicular to radial axis directions from the images in left. B, mean of FFTs from the SERCA and RyR radial and perpendicular to radial axis direction line-scans, shown as a function 1/distance (1/d) (n_cells= 10, 5, 8 and 9; n_linescans= 10, 10, 8 and 17 from top to down). The colour coding is the same as in A. The amplitudes of the individual FFT signals were normalized before calculating the mean. C, a line-scan image of spontaneous Ca²⁺ sparks in E12.5 cardiomyocytes. D, left: mean of fractions of sparks in different parts of the cytosol (n_cells= 7). Line-scan from the surface of the nucleus to the cell membrane was divided into three equal-width parts (sub-SL, centre and sub-nuclear; see C). The fraction of sparks compared to the total number of sparks in the cell was calculated for each region. Right: the mean of amplitude, full width at half-maximum (FWHM), and full duration at half-maximum (FDHM) of the Ca²⁺ sparks in the sub-SL, centre and sub-nuclear regions (n_cells= 7; n_sparks= 40, 55 and 61 in sub-SL, centre and sub-nuclear groups, respectively; *P < 0.05 compared to sub-nuclear).

These peripheral RyRs appear to form local functional release sites, since spontaneous local releases, i.e. sparks, occur throughout the cytosol in embryonic cardiomyocytes (Fig. 3C) with equal distribution in different regions of the cytosol (Fig. 3D). The duration and width of the sparks are also equal in different regions compared to sparks in the sub-nuclear Ca²⁺ store. The only statistically significant difference was found in the amplitude of sub-SL Ca²⁺ sparks, which was ∼1.5-fold compared to sub-nuclear sparks (Fig. 3D).

Physiological significance of the local releases

To estimate the impact of local Ca²⁺ releases in the E–C coupling in developing mouse cardiomyocytes, we did computer simulations with the mathematical model of a mouse embryonic cardiomyocyte (Korhonen et al. 2008). Different-sized model cells were simulated to study the impact of the developmental growth (Seki et al. 2003) of cardiomyocytes on Ca²⁺ signals. The total SR release capacity was increased in accordance with the increased cell volume, but it was either localized to the centre of the cell or divided also to local release sites between the centre of the cell and SL (Fig. 4A). Thus, the total amount of released Ca²⁺ was equal in both cases. In the latter model, as suggested by our experiments (Figs 1–3), the Ca²⁺ propagates by the fire–diffusion–fire mechanism (Coombes et al. 2004) instead of pure diffusion (Fig. 4B).

The local, peripheral Ca²⁺ releases have a great impact on the cell's capability to generate the spontaneous pacemaking action potentials. With pure diffusion, when the size of model cell is increased, the frequency of spontaneous pacemaking activity is reduced significantly (Fig. 4C). In early embryonic cardiomyocytes, IP₃ modulates the frequency of spontaneous SR Ca²⁺ releases and hence the frequency of spontaneous action potentials (Korhonen et al. 2008; Rapila et al. 2008). An increase in the model cell radius from 11 μm to 13.5 μm results in a decrease to 0.1-fold the frequency range of the [IP₃]-dependent spontaneous activity. However, with local release sites between the nucleus and SL at 2 μm intervals, the model cell maintains its [IP₃]-dependent range of spontaneous pacemaking frequencies (Fig. 4B and C).

Excitation–contraction coupling was also greatly affected by the presence of local Ca²⁺ release sites (Fig. 4B and D). When the size of the cell is increased without local release sites, the SR–SL coupling is too weak to synchronize CICRs with APs (Fig. 4C). However, when local release sites are introduced, the maximal simulated pacing frequencies correspond to those recorded in experiments (Figs 4C–D and 5). Based on simulations, in the cells with local release sites the possible pacing frequencies do not depend on the size of the cell (Fig. 4D).

Figure 5. Local release sites are indispensable for global Ca²⁺ signals at physiological rates.

A, mean cytosolic Ca²⁺ signal from E12.5 cardiomyocyte (top), mathematical model with local release sites (centre) and mathematical model without local release sites (bottom) during 0.5, 1 and 2 Hz electrical pacing. B, average of baseline (top) and amplitude (bottom) of mean Ca²⁺ signals from E12.5 cardiomyocytes (n= 12) and from simulated Ca²⁺ signals from the mathematical model with local release sites.

Discussion

In this study we have shown that in embryonic cardiomyocytes, proteins for Ca²⁺ release (RyR) and uptake (SERCA) form regular and homogeneous structural patterns in the cytosol. These structures establish local Ca²⁺ release sites which amplify Ca²⁺ signal propagation in the cytosol. This amplification is indispensable for the strong coupling of SL and SR Ca²⁺ signals required for both the generation of spontaneous pacemaking and for triggering of CICRs at sufficiently high frequency in developing cardiomyocytes.

The role of local Ca²⁺ releases in the E–C coupling of embryonic cardiomyocytes

In embryonic cardiomyocytes, the Ca²⁺ is either released to the cytosol from SL Ca²⁺ sources (VACC) during electrical excitation or from the SR via RyRs during spontaneous activity (Sasse et al. 2007; Rapila et al. 2008). In both cases, the Ca²⁺ entering the cell propagates in the cytosol by diffusion. The [Ca²⁺] in the front of the propagating Ca²⁺ signal dilutes and consequently the velocity of the diffusion decreases in proportion to the diffusion distance. If the diffused [Ca²⁺] is high enough in the vicinity of the RyRs in the local release site, it induces a local CICR. Consequently, the [Ca²⁺] in the front of the Ca²⁺ propagation is augmented, and the strength and the velocity of the Ca²⁺ signal is thereby amplified. The signal is then propagated by diffusion to the next release site, where the signal is amplified again. By this ‘fire–diffusion–fire’ mechanism (Lipp et al. 1990; Trafford et al. 1995; Cheng et al. 1996; Keizer et al. 1998), the Ca²⁺ signal can be transmitted basically an infinite distance, whereas the propagation based on pure diffusion would have a very limited range. Consequently, based on our simulations, the local release sites are required to maintain the developing cardiomyocytes functional when the cell size increases. Both the generation of spontaneous action potentials and excitation–contraction coupling triggered by electrical pacing were disabled in simulations if the SR remained localized in the centre of the cell. The ‘fire–diffusion–fire’ mechanism seems to be generally present in all kinds of developing cardiomyocytes. In addition to the primary isolated cardiomyocytes from embryos studied here and neonates studied previously (Korhonen et al. 2009), the Ca²⁺ signal propagates as a Ca²⁺ wave in embryonic stem cell-derived cardiomyocytes (Kapur & Banach, 2007; Satin et al. 2008).

Embryonic cardiomyocytes release Ca²⁺ spontaneously from the SR, causing cytosolic Ca²⁺ oscillations. When this Ca²⁺ reaches the SL, Na⁺/Ca²⁺ exchanger (NCX) extrudes Ca²⁺, generating a depolarizing current and consequently triggering AP and voltage-activated Ca²⁺ intrusion via voltage-activated Ca²⁺ channels (Rapila et al. 2008). As the SR has to refill before the next spontaneous release, the SR begins to uptake the Ca²⁺ at the site of the release immediately after the release. However, some of the released Ca²⁺ escapes and diffuses all the way to the SL, where it is extruded out of the cell by NCX. Consequently, the final Ca²⁺ uptake to the SR cannot occur before the compensating amount of Ca²⁺ is first intruded to the cytosol via VACC during the AP. However, with local release and uptake sites, the local Ca²⁺ release amplifies the Ca²⁺ signal propagation, and local uptake sites refill the peripheral SR sites and provide compensation for the released Ca²⁺ (Fig. 4A). Consequently, each release/uptake site amplifies the propagation of Ca²⁺ signal towards the next release site and also provides compensating Ca²⁺ for previous release sites. By this mechanism, the central SR is refilled much faster with local release site(s) than without, enabling faster spontaneous Ca²⁺ oscillations and consequent AP frequencies.

Although there are release sites all over the cytosol, the global spontaneous Ca²⁺ releases seem to originate exclusively from the perinuclear part of the cytosol in early embryonic cardiomyocytes (Rapila et al. 2008). Based on this study, the SR also amplifies the propagation of Ca²⁺ signals within the cytosol with local, peripheral release sites in the cytosol. However, on the basis of our experiments, these peripheral release sites are capable of producing local Ca²⁺ release, but unable to trigger propagating Ca²⁺ waves alone, because sparks occurred always at discrete locations in the cytosol (Fig. 3). In the experiments, we never saw global Ca²⁺ signals without Ca²⁺ through VACC or Ca²⁺ release from the central SR. This suggests that the release capacity of the peripheral release sites is limited and perhaps a single opening of RyRs producing a spark is enough to locally deplete the peripheral SR. This was suggested also by our immunolabelling of SR proteins where peripheral SR appeared as is a thin network structure with presumably limited volume (Fig. 3). Even if the Ca²⁺ release from an individual local peripheral site is small, they have a great impact on the whole-cell Ca²⁺ signals. The cells with local Ca²⁺ release units spread throughout the cytosol can maintain electrical and contractile function with whole-cell Ca²⁺ signals at lower amplitudes and with greater spatiotemporal homogeneity than cells with only central SR release sites. In practice, due to the dilution of signals in signal propagation based on pure diffusion, triggering a large enough Ca²⁺ wave from the centre of the cell to diffuse to the SL and yet be able to trigger AP via NCX current requires a much greater initial Ca²⁺ signal compared to the situation where Ca²⁺ is propagated by the fire–diffusion–fire mechanism (Fig. 4B).

The 3-D structure of t-tubules and VACC–RyR couplings in adult ventricular myocytes is the most finely tuned E–C coupling mechanism for generating homogeneous high-frequency Ca²⁺ signals (Soeller & Cannell, 1999; Bers, 2001). Building and maintaining intimate, 15 nm couplings (Cannell & Soeller, 1997; Soeller & Cannell, 1999) between VACCs in t-tubules and RyRs in SR throughout the whole myocyte is likely to require stable, relatively rigid structures. Transverse tubuli are formed at the very end of cardiomyocyte development (Di Maio et al. 2007), when cells become terminally differentiated. In developing cardiomyocytes the sizes and shapes of individual cells are continuously changed and cells are dividing and migrating (Seki et al. 2003; Buckingham et al. 2005). Therefore, in order to maintain electrical and mechanical activity in developing cells, the E–C coupling mechanism has to be more flexible and robust, like the one suggested in this study. According to our results, establishing such a robust E–C coupling mechanism requires only active growth of peripheral SR extensions establishing local release units without junctional coupling with SL. In this model, the cell membrane can be a simple excitable surface without complex invaginations, being flexible enough to allow the cell to change shape and size.

Although the adult ventricular myocytes develop t-tubules and tight local coupling between VACCs and RyRs, a similar mechanism presented in this study is present in adult atrial myocytes (Kockskamper et al. 2001). Both adult atrial and embryonic ventricular myocytes have a homogeneous structure of local release sites located ∼2 μm apart from each other (Kockskamper et al. 2001; Sheehan et al. 2006). However, in adult atrial myocytes, due to the tight ventricular-like coupling of sub-SL release sites with the VACCs, the sub-SL release sites produce Ca²⁺ sparks more frequently and play a dominant role in the generation of the Ca²⁺ transient compared to release sites at the inner parts of the cell (Sheehan et al. 2006). In addition to the structural homogeneity, our data (Fig. 3D) indicate that in embryonic cardiomyocytes, the functionality of the local release sites is also distributed homogeneously throughout the cell. Again, this might be a beneficial feature, regarding the development of the myocytes, with the cost of less dynamic initial CICR during electrical excitation, since fewer geometrical variables are required to control in a homogeneous cell structure without tight sub-SL coupling region.

Transgenic animal models with altered expression of SR Ca²⁺ handling proteins result in lethality during development (Takeshima et al. 1998; Mesaeli et al. 1999; Nakamura et al. 2001). Previously it has been shown that the SR is indispensable for the embryonic heart beat, since the SR generates the spontaneous Ca²⁺ releases underlying the spontaneous action potentials and produces CICR required for the large synchronized Ca²⁺ transients in forceful contraction (Korhonen et al. 2008; Rapila et al. 2008). In this study we show that the SR is also indispensable since the local CICRs from the local extensions of the SR are required to carry the spontaneous and externally evoked Ca²⁺ signals throughout the cytosol. Without the mechanism presented here, the myocytes in the developing heart are unable to produce coordinated, forceful enough contractions at a sufficiently high rate.

Glossary

Abbreviations

AP

action potential

CICR

Ca²⁺-induced Ca²⁺ release

E

embryonic day

FFT

fast Fourier transform

NCX

Na⁺/Ca²⁺ exchanger

ODE

ordinary differential equation

RyR

ryanodine receptor

SERCA

SR Ca²⁺ ATPase

SL

sarcolemma

SR

sarcoplasmic reticulum

VACC

voltage-activated Ca²⁺ channel

Author contributions

Conception and design of this study was conducted by P.T. and T.K. Experiments were performed by T.K., R.R. and V.-P.R. Mathematical modelling was conducted by T.K. and J.T.K. Data were interpreted and analysed by P.T, T.K and J.T.K. All authors contributed to writing of the manuscript and approved the final version to be published.